From 62a038d34db26771756cf3689e36de638bedd2c4 Mon Sep 17 00:00:00 2001 From: "K.Prasad" Date: Mon, 1 Jun 2009 23:43:33 +0530 Subject: hw-breakpoints: introducing generic hardware breakpoint handler interfaces This patch introduces the generic Hardware Breakpoint interfaces for both user and kernel space requests. This core Api handles the hardware breakpoints through new helpers. It handles the user-space breakpoints and kernel breakpoints in front of arch implementation. One can choose kernel wide breakpoints using the following helpers and passing them a generic struct hw_breakpoint: - register_kernel_hw_breakpoint() - unregister_kernel_hw_breakpoint() - modify_kernel_hw_breakpoint() On the other side, you can choose per task breakpoints. - register_user_hw_breakpoint() - unregister_user_hw_breakpoint() - modify_user_hw_breakpoint() [ fweisbec@gmail.com: fix conflict against perfcounter ] Original-patch-by: Alan Stern Signed-off-by: K.Prasad Reviewed-by: Alan Stern Signed-off-by: Frederic Weisbecker --- kernel/Makefile | 1 + kernel/hw_breakpoint.c | 378 +++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 379 insertions(+) create mode 100644 kernel/hw_breakpoint.c (limited to 'kernel') diff --git a/kernel/Makefile b/kernel/Makefile index a35eee3..18ad111 100644 --- a/kernel/Makefile +++ b/kernel/Makefile @@ -96,6 +96,7 @@ obj-$(CONFIG_TRACING) += trace/ obj-$(CONFIG_X86_DS) += trace/ obj-$(CONFIG_SMP) += sched_cpupri.o obj-$(CONFIG_SLOW_WORK) += slow-work.o +obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o ifneq ($(CONFIG_SCHED_OMIT_FRAME_POINTER),y) # According to Alan Modra , the -fno-omit-frame-pointer is diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c new file mode 100644 index 0000000..c1f64e6 --- /dev/null +++ b/kernel/hw_breakpoint.c @@ -0,0 +1,378 @@ +/* + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + * + * Copyright (C) 2007 Alan Stern + * Copyright (C) IBM Corporation, 2009 + */ + +/* + * HW_breakpoint: a unified kernel/user-space hardware breakpoint facility, + * using the CPU's debug registers. + * This file contains the arch-independent routines. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include + +#ifdef CONFIG_X86 +#include +#endif +/* + * Spinlock that protects all (un)register operations over kernel/user-space + * breakpoint requests + */ +static DEFINE_SPINLOCK(hw_breakpoint_lock); + +/* Array of kernel-space breakpoint structures */ +struct hw_breakpoint *hbp_kernel[HBP_NUM]; + +/* + * Per-processor copy of hbp_kernel[]. Used only when hbp_kernel is being + * modified but we need the older copy to handle any hbp exceptions. It will + * sync with hbp_kernel[] value after updation is done through IPIs. + */ +DEFINE_PER_CPU(struct hw_breakpoint*, this_hbp_kernel[HBP_NUM]); + +/* + * Kernel breakpoints grow downwards, starting from HBP_NUM + * 'hbp_kernel_pos' denotes lowest numbered breakpoint register occupied for + * kernel-space request. We will initialise it here and not in an __init + * routine because load_debug_registers(), which uses this variable can be + * called very early during CPU initialisation. + */ +unsigned int hbp_kernel_pos = HBP_NUM; + +/* + * An array containing refcount of threads using a given bkpt register + * Accesses are synchronised by acquiring hw_breakpoint_lock + */ +unsigned int hbp_user_refcount[HBP_NUM]; + +/* + * Load the debug registers during startup of a CPU. + */ +void load_debug_registers(void) +{ + unsigned long flags; + struct task_struct *tsk = current; + + spin_lock_bh(&hw_breakpoint_lock); + + /* Prevent IPIs for new kernel breakpoint updates */ + local_irq_save(flags); + arch_update_kernel_hw_breakpoint(NULL); + local_irq_restore(flags); + + if (test_tsk_thread_flag(tsk, TIF_DEBUG)) + arch_install_thread_hw_breakpoint(tsk); + + spin_unlock_bh(&hw_breakpoint_lock); +} + +/* + * Erase all the hardware breakpoint info associated with a thread. + * + * If tsk != current then tsk must not be usable (for example, a + * child being cleaned up from a failed fork). + */ +void flush_thread_hw_breakpoint(struct task_struct *tsk) +{ + int i; + struct thread_struct *thread = &(tsk->thread); + + spin_lock_bh(&hw_breakpoint_lock); + + /* The thread no longer has any breakpoints associated with it */ + clear_tsk_thread_flag(tsk, TIF_DEBUG); + for (i = 0; i < HBP_NUM; i++) { + if (thread->hbp[i]) { + hbp_user_refcount[i]--; + kfree(thread->hbp[i]); + thread->hbp[i] = NULL; + } + } + + arch_flush_thread_hw_breakpoint(tsk); + + /* Actually uninstall the breakpoints if necessary */ + if (tsk == current) + arch_uninstall_thread_hw_breakpoint(); + spin_unlock_bh(&hw_breakpoint_lock); +} + +/* + * Copy the hardware breakpoint info from a thread to its cloned child. + */ +int copy_thread_hw_breakpoint(struct task_struct *tsk, + struct task_struct *child, unsigned long clone_flags) +{ + /* + * We will assume that breakpoint settings are not inherited + * and the child starts out with no debug registers set. + * But what about CLONE_PTRACE? + */ + clear_tsk_thread_flag(child, TIF_DEBUG); + + /* We will call flush routine since the debugregs are not inherited */ + arch_flush_thread_hw_breakpoint(child); + + return 0; +} + +static int __register_user_hw_breakpoint(int pos, struct task_struct *tsk, + struct hw_breakpoint *bp) +{ + struct thread_struct *thread = &(tsk->thread); + int rc; + + /* Do not overcommit. Fail if kernel has used the hbp registers */ + if (pos >= hbp_kernel_pos) + return -ENOSPC; + + rc = arch_validate_hwbkpt_settings(bp, tsk); + if (rc) + return rc; + + thread->hbp[pos] = bp; + hbp_user_refcount[pos]++; + + arch_update_user_hw_breakpoint(pos, tsk); + /* + * Does it need to be installed right now? + * Otherwise it will get installed the next time tsk runs + */ + if (tsk == current) + arch_install_thread_hw_breakpoint(tsk); + + return rc; +} + +/* + * Modify the address of a hbp register already in use by the task + * Do not invoke this in-lieu of a __unregister_user_hw_breakpoint() + */ +static int __modify_user_hw_breakpoint(int pos, struct task_struct *tsk, + struct hw_breakpoint *bp) +{ + struct thread_struct *thread = &(tsk->thread); + + if ((pos >= hbp_kernel_pos) || (arch_validate_hwbkpt_settings(bp, tsk))) + return -EINVAL; + + if (thread->hbp[pos] == NULL) + return -EINVAL; + + thread->hbp[pos] = bp; + /* + * 'pos' must be that of a hbp register already used by 'tsk' + * Otherwise arch_modify_user_hw_breakpoint() will fail + */ + arch_update_user_hw_breakpoint(pos, tsk); + + if (tsk == current) + arch_install_thread_hw_breakpoint(tsk); + + return 0; +} + +static void __unregister_user_hw_breakpoint(int pos, struct task_struct *tsk) +{ + hbp_user_refcount[pos]--; + tsk->thread.hbp[pos] = NULL; + + arch_update_user_hw_breakpoint(pos, tsk); + + if (tsk == current) + arch_install_thread_hw_breakpoint(tsk); +} + +/** + * register_user_hw_breakpoint - register a hardware breakpoint for user space + * @tsk: pointer to 'task_struct' of the process to which the address belongs + * @bp: the breakpoint structure to register + * + * @bp.info->name or @bp.info->address, @bp.info->len, @bp.info->type and + * @bp->triggered must be set properly before invocation + * + */ +int register_user_hw_breakpoint(struct task_struct *tsk, + struct hw_breakpoint *bp) +{ + struct thread_struct *thread = &(tsk->thread); + int i, rc = -ENOSPC; + + spin_lock_bh(&hw_breakpoint_lock); + + for (i = 0; i < hbp_kernel_pos; i++) { + if (!thread->hbp[i]) { + rc = __register_user_hw_breakpoint(i, tsk, bp); + break; + } + } + if (!rc) + set_tsk_thread_flag(tsk, TIF_DEBUG); + + spin_unlock_bh(&hw_breakpoint_lock); + return rc; +} +EXPORT_SYMBOL_GPL(register_user_hw_breakpoint); + +/** + * modify_user_hw_breakpoint - modify a user-space hardware breakpoint + * @tsk: pointer to 'task_struct' of the process to which the address belongs + * @bp: the breakpoint structure to unregister + * + */ +int modify_user_hw_breakpoint(struct task_struct *tsk, struct hw_breakpoint *bp) +{ + struct thread_struct *thread = &(tsk->thread); + int i, ret = -ENOENT; + + spin_lock_bh(&hw_breakpoint_lock); + for (i = 0; i < hbp_kernel_pos; i++) { + if (bp == thread->hbp[i]) { + ret = __modify_user_hw_breakpoint(i, tsk, bp); + break; + } + } + spin_unlock_bh(&hw_breakpoint_lock); + return ret; +} +EXPORT_SYMBOL_GPL(modify_user_hw_breakpoint); + +/** + * unregister_user_hw_breakpoint - unregister a user-space hardware breakpoint + * @tsk: pointer to 'task_struct' of the process to which the address belongs + * @bp: the breakpoint structure to unregister + * + */ +void unregister_user_hw_breakpoint(struct task_struct *tsk, + struct hw_breakpoint *bp) +{ + struct thread_struct *thread = &(tsk->thread); + int i, pos = -1, hbp_counter = 0; + + spin_lock_bh(&hw_breakpoint_lock); + for (i = 0; i < hbp_kernel_pos; i++) { + if (thread->hbp[i]) + hbp_counter++; + if (bp == thread->hbp[i]) + pos = i; + } + if (pos >= 0) { + __unregister_user_hw_breakpoint(pos, tsk); + hbp_counter--; + } + if (!hbp_counter) + clear_tsk_thread_flag(tsk, TIF_DEBUG); + + spin_unlock_bh(&hw_breakpoint_lock); +} +EXPORT_SYMBOL_GPL(unregister_user_hw_breakpoint); + +/** + * register_kernel_hw_breakpoint - register a hardware breakpoint for kernel space + * @bp: the breakpoint structure to register + * + * @bp.info->name or @bp.info->address, @bp.info->len, @bp.info->type and + * @bp->triggered must be set properly before invocation + * + */ +int register_kernel_hw_breakpoint(struct hw_breakpoint *bp) +{ + int rc; + + rc = arch_validate_hwbkpt_settings(bp, NULL); + if (rc) + return rc; + + spin_lock_bh(&hw_breakpoint_lock); + + rc = -ENOSPC; + /* Check if we are over-committing */ + if ((hbp_kernel_pos > 0) && (!hbp_user_refcount[hbp_kernel_pos-1])) { + hbp_kernel_pos--; + hbp_kernel[hbp_kernel_pos] = bp; + on_each_cpu(arch_update_kernel_hw_breakpoint, NULL, 1); + rc = 0; + } + + spin_unlock_bh(&hw_breakpoint_lock); + return rc; +} +EXPORT_SYMBOL_GPL(register_kernel_hw_breakpoint); + +/** + * unregister_kernel_hw_breakpoint - unregister a HW breakpoint for kernel space + * @bp: the breakpoint structure to unregister + * + * Uninstalls and unregisters @bp. + */ +void unregister_kernel_hw_breakpoint(struct hw_breakpoint *bp) +{ + int i, j; + + spin_lock_bh(&hw_breakpoint_lock); + + /* Find the 'bp' in our list of breakpoints for kernel */ + for (i = hbp_kernel_pos; i < HBP_NUM; i++) + if (bp == hbp_kernel[i]) + break; + + /* Check if we did not find a match for 'bp'. If so return early */ + if (i == HBP_NUM) { + spin_unlock_bh(&hw_breakpoint_lock); + return; + } + + /* + * We'll shift the breakpoints one-level above to compact if + * unregistration creates a hole + */ + for (j = i; j > hbp_kernel_pos; j--) + hbp_kernel[j] = hbp_kernel[j-1]; + + hbp_kernel[hbp_kernel_pos] = NULL; + on_each_cpu(arch_update_kernel_hw_breakpoint, NULL, 1); + hbp_kernel_pos++; + + spin_unlock_bh(&hw_breakpoint_lock); +} +EXPORT_SYMBOL_GPL(unregister_kernel_hw_breakpoint); + +static struct notifier_block hw_breakpoint_exceptions_nb = { + .notifier_call = hw_breakpoint_exceptions_notify, + /* we need to be notified first */ + .priority = 0x7fffffff +}; + +static int __init init_hw_breakpoint(void) +{ + return register_die_notifier(&hw_breakpoint_exceptions_nb); +} + +core_initcall(init_hw_breakpoint); -- cgit v1.1 From 0722db015c246204044299eae3b02d18d3ca4faf Mon Sep 17 00:00:00 2001 From: "K.Prasad" Date: Mon, 1 Jun 2009 23:46:40 +0530 Subject: hw-breakpoints: ftrace plugin for kernel symbol tracing using HW Breakpoint interfaces This patch adds an ftrace plugin to detect and profile memory access over kernel variables. It uses HW Breakpoint interfaces to 'watch memory addresses. Signed-off-by: K.Prasad Signed-off-by: Frederic Weisbecker --- kernel/trace/Kconfig | 21 ++ kernel/trace/Makefile | 1 + kernel/trace/trace.h | 23 ++ kernel/trace/trace_ksym.c | 525 ++++++++++++++++++++++++++++++++++++++++++ kernel/trace/trace_selftest.c | 53 +++++ 5 files changed, 623 insertions(+) create mode 100644 kernel/trace/trace_ksym.c (limited to 'kernel') diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig index a508b9d..d7f01e6 100644 --- a/kernel/trace/Kconfig +++ b/kernel/trace/Kconfig @@ -314,6 +314,27 @@ config POWER_TRACER power management decisions, specifically the C-state and P-state behavior. +config KSYM_TRACER + bool "Trace read and write access on kernel memory locations" + depends on HAVE_HW_BREAKPOINT + select TRACING + help + This tracer helps find read and write operations on any given kernel + symbol i.e. /proc/kallsyms. + +config PROFILE_KSYM_TRACER + bool "Profile all kernel memory accesses on 'watched' variables" + depends on KSYM_TRACER + help + This tracer profiles kernel accesses on variables watched through the + ksym tracer ftrace plugin. Depending upon the hardware, all read + and write operations on kernel variables can be monitored for + accesses. + + The results will be displayed in: + /debugfs/tracing/profile_ksym + + Say N if unsure. config STACK_TRACER bool "Trace max stack" diff --git a/kernel/trace/Makefile b/kernel/trace/Makefile index 06b8585..658aace 100644 --- a/kernel/trace/Makefile +++ b/kernel/trace/Makefile @@ -51,5 +51,6 @@ obj-$(CONFIG_EVENT_TRACING) += trace_export.o obj-$(CONFIG_FTRACE_SYSCALLS) += trace_syscalls.o obj-$(CONFIG_EVENT_PROFILE) += trace_event_profile.o obj-$(CONFIG_EVENT_TRACING) += trace_events_filter.o +obj-$(CONFIG_KSYM_TRACER) += trace_ksym.o libftrace-y := ftrace.o diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 6e735d4..7d5cc37 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -15,6 +15,10 @@ #include #include +#ifdef CONFIG_KSYM_TRACER +#include +#endif + enum trace_type { __TRACE_FIRST_TYPE = 0, @@ -40,6 +44,7 @@ enum trace_type { TRACE_KMEM_FREE, TRACE_POWER, TRACE_BLK, + TRACE_KSYM, __TRACE_LAST_TYPE, }; @@ -207,6 +212,21 @@ struct syscall_trace_exit { unsigned long ret; }; +#define KSYM_SELFTEST_ENTRY "ksym_selftest_dummy" +extern int process_new_ksym_entry(char *ksymname, int op, unsigned long addr); + +struct trace_ksym { + struct trace_entry ent; + struct hw_breakpoint *ksym_hbp; + unsigned long ksym_addr; + unsigned long ip; +#ifdef CONFIG_PROFILE_KSYM_TRACER + unsigned long counter; +#endif + struct hlist_node ksym_hlist; + char ksym_name[KSYM_NAME_LEN]; + char p_name[TASK_COMM_LEN]; +}; /* * trace_flag_type is an enumeration that holds different @@ -323,6 +343,7 @@ extern void __ftrace_bad_type(void); TRACE_SYSCALL_ENTER); \ IF_ASSIGN(var, ent, struct syscall_trace_exit, \ TRACE_SYSCALL_EXIT); \ + IF_ASSIGN(var, ent, struct trace_ksym, TRACE_KSYM); \ __ftrace_bad_type(); \ } while (0) @@ -540,6 +561,8 @@ extern int trace_selftest_startup_branch(struct tracer *trace, struct trace_array *tr); extern int trace_selftest_startup_hw_branches(struct tracer *trace, struct trace_array *tr); +extern int trace_selftest_startup_ksym(struct tracer *trace, + struct trace_array *tr); #endif /* CONFIG_FTRACE_STARTUP_TEST */ extern void *head_page(struct trace_array_cpu *data); diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c new file mode 100644 index 0000000..11c74f6 --- /dev/null +++ b/kernel/trace/trace_ksym.c @@ -0,0 +1,525 @@ +/* + * trace_ksym.c - Kernel Symbol Tracer + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + * + * Copyright (C) IBM Corporation, 2009 + */ + +#include +#include +#include +#include +#include +#include + +#include "trace_output.h" +#include "trace_stat.h" +#include "trace.h" + +/* For now, let us restrict the no. of symbols traced simultaneously to number + * of available hardware breakpoint registers. + */ +#define KSYM_TRACER_MAX HBP_NUM + +#define KSYM_TRACER_OP_LEN 3 /* rw- */ +#define KSYM_FILTER_ENTRY_LEN (KSYM_NAME_LEN + KSYM_TRACER_OP_LEN + 1) + +static struct trace_array *ksym_trace_array; + +static unsigned int ksym_filter_entry_count; +static unsigned int ksym_tracing_enabled; + +static HLIST_HEAD(ksym_filter_head); + +#ifdef CONFIG_PROFILE_KSYM_TRACER + +#define MAX_UL_INT 0xffffffff + +static DEFINE_MUTEX(ksym_tracer_mutex); + +void ksym_collect_stats(unsigned long hbp_hit_addr) +{ + struct hlist_node *node; + struct trace_ksym *entry; + + rcu_read_lock(); + hlist_for_each_entry_rcu(entry, node, &ksym_filter_head, ksym_hlist) { + if ((entry->ksym_addr == hbp_hit_addr) && + (entry->counter <= MAX_UL_INT)) { + entry->counter++; + break; + } + } + rcu_read_unlock(); +} +#endif /* CONFIG_PROFILE_KSYM_TRACER */ + +void ksym_hbp_handler(struct hw_breakpoint *hbp, struct pt_regs *regs) +{ + struct ring_buffer_event *event; + struct trace_array *tr; + struct trace_ksym *entry; + int pc; + + if (!ksym_tracing_enabled) + return; + + tr = ksym_trace_array; + pc = preempt_count(); + + event = trace_buffer_lock_reserve(tr, TRACE_KSYM, + sizeof(*entry), 0, pc); + if (!event) + return; + + entry = ring_buffer_event_data(event); + strlcpy(entry->ksym_name, hbp->info.name, KSYM_SYMBOL_LEN); + entry->ksym_hbp = hbp; + entry->ip = instruction_pointer(regs); + strlcpy(entry->p_name, current->comm, TASK_COMM_LEN); +#ifdef CONFIG_PROFILE_KSYM_TRACER + ksym_collect_stats(hbp->info.address); +#endif /* CONFIG_PROFILE_KSYM_TRACER */ + + trace_buffer_unlock_commit(tr, event, 0, pc); +} + +/* Valid access types are represented as + * + * rw- : Set Read/Write Access Breakpoint + * -w- : Set Write Access Breakpoint + * --- : Clear Breakpoints + * --x : Set Execution Break points (Not available yet) + * + */ +static int ksym_trace_get_access_type(char *access_str) +{ + int pos, access = 0; + + for (pos = 0; pos < KSYM_TRACER_OP_LEN; pos++) { + switch (access_str[pos]) { + case 'r': + access += (pos == 0) ? 4 : -1; + break; + case 'w': + access += (pos == 1) ? 2 : -1; + break; + case '-': + break; + default: + return -EINVAL; + } + } + + switch (access) { + case 6: + access = HW_BREAKPOINT_RW; + break; + case 2: + access = HW_BREAKPOINT_WRITE; + break; + case 0: + access = 0; + } + + return access; +} + +/* + * There can be several possible malformed requests and we attempt to capture + * all of them. We enumerate some of the rules + * 1. We will not allow kernel symbols with ':' since it is used as a delimiter. + * i.e. multiple ':' symbols disallowed. Possible uses are of the form + * ::. + * 2. No delimiter symbol ':' in the input string + * 3. Spurious operator symbols or symbols not in their respective positions + * 4. :--- i.e. clear breakpoint request when ksym_name not in file + * 5. Kernel symbol not a part of /proc/kallsyms + * 6. Duplicate requests + */ +static int parse_ksym_trace_str(char *input_string, char **ksymname, + unsigned long *addr) +{ + char *delimiter = ":"; + int ret; + + ret = -EINVAL; + *ksymname = strsep(&input_string, delimiter); + *addr = kallsyms_lookup_name(*ksymname); + + /* Check for malformed request: (2), (1) and (5) */ + if ((!input_string) || + (strlen(input_string) != (KSYM_TRACER_OP_LEN + 1)) || + (*addr == 0)) + goto return_code; + ret = ksym_trace_get_access_type(input_string); + +return_code: + return ret; +} + +int process_new_ksym_entry(char *ksymname, int op, unsigned long addr) +{ + struct trace_ksym *entry; + int ret; + + if (ksym_filter_entry_count >= KSYM_TRACER_MAX) { + printk(KERN_ERR "ksym_tracer: Maximum limit:(%d) reached. No" + " new requests for tracing can be accepted now.\n", + KSYM_TRACER_MAX); + return -ENOSPC; + } + + entry = kzalloc(sizeof(struct trace_ksym), GFP_KERNEL); + if (!entry) + return -ENOMEM; + + entry->ksym_hbp = kzalloc(sizeof(struct hw_breakpoint), GFP_KERNEL); + if (!entry->ksym_hbp) { + kfree(entry); + return -ENOMEM; + } + + entry->ksym_hbp->info.name = ksymname; + entry->ksym_hbp->info.type = op; + entry->ksym_addr = entry->ksym_hbp->info.address = addr; +#ifdef CONFIG_X86 + entry->ksym_hbp->info.len = HW_BREAKPOINT_LEN_4; +#endif + entry->ksym_hbp->triggered = (void *)ksym_hbp_handler; + + ret = register_kernel_hw_breakpoint(entry->ksym_hbp); + if (ret < 0) { + printk(KERN_INFO "ksym_tracer request failed. Try again" + " later!!\n"); + kfree(entry->ksym_hbp); + kfree(entry); + return -EAGAIN; + } + hlist_add_head_rcu(&(entry->ksym_hlist), &ksym_filter_head); + ksym_filter_entry_count++; + + return 0; +} + +static ssize_t ksym_trace_filter_read(struct file *filp, char __user *ubuf, + size_t count, loff_t *ppos) +{ + struct trace_ksym *entry; + struct hlist_node *node; + char buf[KSYM_FILTER_ENTRY_LEN * KSYM_TRACER_MAX]; + ssize_t ret, cnt = 0; + + mutex_lock(&ksym_tracer_mutex); + + hlist_for_each_entry(entry, node, &ksym_filter_head, ksym_hlist) { + cnt += snprintf(&buf[cnt], KSYM_FILTER_ENTRY_LEN - cnt, "%s:", + entry->ksym_hbp->info.name); + if (entry->ksym_hbp->info.type == HW_BREAKPOINT_WRITE) + cnt += snprintf(&buf[cnt], KSYM_FILTER_ENTRY_LEN - cnt, + "-w-\n"); + else if (entry->ksym_hbp->info.type == HW_BREAKPOINT_RW) + cnt += snprintf(&buf[cnt], KSYM_FILTER_ENTRY_LEN - cnt, + "rw-\n"); + } + ret = simple_read_from_buffer(ubuf, count, ppos, buf, strlen(buf)); + mutex_unlock(&ksym_tracer_mutex); + + return ret; +} + +static ssize_t ksym_trace_filter_write(struct file *file, + const char __user *buffer, + size_t count, loff_t *ppos) +{ + struct trace_ksym *entry; + struct hlist_node *node; + char *input_string, *ksymname = NULL; + unsigned long ksym_addr = 0; + int ret, op, changed = 0; + + /* Ignore echo "" > ksym_trace_filter */ + if (count == 0) + return 0; + + input_string = kzalloc(count, GFP_KERNEL); + if (!input_string) + return -ENOMEM; + + if (copy_from_user(input_string, buffer, count)) { + kfree(input_string); + return -EFAULT; + } + + ret = op = parse_ksym_trace_str(input_string, &ksymname, &ksym_addr); + if (ret < 0) { + kfree(input_string); + return ret; + } + + mutex_lock(&ksym_tracer_mutex); + + ret = -EINVAL; + hlist_for_each_entry(entry, node, &ksym_filter_head, ksym_hlist) { + if (entry->ksym_addr == ksym_addr) { + /* Check for malformed request: (6) */ + if (entry->ksym_hbp->info.type != op) + changed = 1; + else + goto err_ret; + break; + } + } + if (changed) { + unregister_kernel_hw_breakpoint(entry->ksym_hbp); + entry->ksym_hbp->info.type = op; + if (op > 0) { + ret = register_kernel_hw_breakpoint(entry->ksym_hbp); + if (ret == 0) { + ret = count; + goto unlock_ret_path; + } + } + ksym_filter_entry_count--; + hlist_del_rcu(&(entry->ksym_hlist)); + synchronize_rcu(); + kfree(entry->ksym_hbp); + kfree(entry); + ret = count; + goto err_ret; + } else { + /* Check for malformed request: (4) */ + if (op == 0) + goto err_ret; + ret = process_new_ksym_entry(ksymname, op, ksym_addr); + if (ret) + goto err_ret; + } + ret = count; + goto unlock_ret_path; + +err_ret: + kfree(input_string); + +unlock_ret_path: + mutex_unlock(&ksym_tracer_mutex); + return ret; +} + +static const struct file_operations ksym_tracing_fops = { + .open = tracing_open_generic, + .read = ksym_trace_filter_read, + .write = ksym_trace_filter_write, +}; + +static void ksym_trace_reset(struct trace_array *tr) +{ + struct trace_ksym *entry; + struct hlist_node *node, *node1; + + ksym_tracing_enabled = 0; + + mutex_lock(&ksym_tracer_mutex); + hlist_for_each_entry_safe(entry, node, node1, &ksym_filter_head, + ksym_hlist) { + unregister_kernel_hw_breakpoint(entry->ksym_hbp); + ksym_filter_entry_count--; + hlist_del_rcu(&(entry->ksym_hlist)); + synchronize_rcu(); + /* Free the 'input_string' only if reset + * after startup self-test + */ +#ifdef CONFIG_FTRACE_SELFTEST + if (strncmp(entry->ksym_hbp->info.name, KSYM_SELFTEST_ENTRY, + strlen(KSYM_SELFTEST_ENTRY)) != 0) +#endif /* CONFIG_FTRACE_SELFTEST*/ + kfree(entry->ksym_hbp->info.name); + kfree(entry->ksym_hbp); + kfree(entry); + } + mutex_unlock(&ksym_tracer_mutex); +} + +static int ksym_trace_init(struct trace_array *tr) +{ + int cpu, ret = 0; + + for_each_online_cpu(cpu) + tracing_reset(tr, cpu); + ksym_tracing_enabled = 1; + ksym_trace_array = tr; + + return ret; +} + +static void ksym_trace_print_header(struct seq_file *m) +{ + + seq_puts(m, + "# TASK-PID CPU# Symbol Type " + "Function \n"); + seq_puts(m, + "# | | | | " + "| \n"); +} + +static enum print_line_t ksym_trace_output(struct trace_iterator *iter) +{ + struct trace_entry *entry = iter->ent; + struct trace_seq *s = &iter->seq; + struct trace_ksym *field; + char str[KSYM_SYMBOL_LEN]; + int ret; + + if (entry->type != TRACE_KSYM) + return TRACE_TYPE_UNHANDLED; + + trace_assign_type(field, entry); + + ret = trace_seq_printf(s, "%-15s %-5d %-3d %-20s ", field->p_name, + entry->pid, iter->cpu, field->ksym_name); + if (!ret) + return TRACE_TYPE_PARTIAL_LINE; + + switch (field->ksym_hbp->info.type) { + case HW_BREAKPOINT_WRITE: + ret = trace_seq_printf(s, " W "); + break; + case HW_BREAKPOINT_RW: + ret = trace_seq_printf(s, " RW "); + break; + default: + return TRACE_TYPE_PARTIAL_LINE; + } + + if (!ret) + return TRACE_TYPE_PARTIAL_LINE; + + sprint_symbol(str, field->ip); + ret = trace_seq_printf(s, "%-20s\n", str); + if (!ret) + return TRACE_TYPE_PARTIAL_LINE; + + return TRACE_TYPE_HANDLED; +} + +struct tracer ksym_tracer __read_mostly = +{ + .name = "ksym_tracer", + .init = ksym_trace_init, + .reset = ksym_trace_reset, +#ifdef CONFIG_FTRACE_SELFTEST + .selftest = trace_selftest_startup_ksym, +#endif + .print_header = ksym_trace_print_header, + .print_line = ksym_trace_output +}; + +__init static int init_ksym_trace(void) +{ + struct dentry *d_tracer; + struct dentry *entry; + + d_tracer = tracing_init_dentry(); + ksym_filter_entry_count = 0; + + entry = debugfs_create_file("ksym_trace_filter", 0644, d_tracer, + NULL, &ksym_tracing_fops); + if (!entry) + pr_warning("Could not create debugfs " + "'ksym_trace_filter' file\n"); + + return register_tracer(&ksym_tracer); +} +device_initcall(init_ksym_trace); + + +#ifdef CONFIG_PROFILE_KSYM_TRACER +static int ksym_tracer_stat_headers(struct seq_file *m) +{ + seq_printf(m, " Access type "); + seq_printf(m, " Symbol Counter \n"); + return 0; +} + +static int ksym_tracer_stat_show(struct seq_file *m, void *v) +{ + struct hlist_node *stat = v; + struct trace_ksym *entry; + int access_type = 0; + char fn_name[KSYM_NAME_LEN]; + + entry = hlist_entry(stat, struct trace_ksym, ksym_hlist); + + if (entry->ksym_hbp) + access_type = entry->ksym_hbp->info.type; + + switch (access_type) { + case HW_BREAKPOINT_WRITE: + seq_printf(m, " W "); + break; + case HW_BREAKPOINT_RW: + seq_printf(m, " RW "); + break; + default: + seq_printf(m, " NA "); + } + + if (lookup_symbol_name(entry->ksym_addr, fn_name) >= 0) + seq_printf(m, " %s ", fn_name); + else + seq_printf(m, " "); + + seq_printf(m, "%15lu\n", entry->counter); + return 0; +} + +static void *ksym_tracer_stat_start(struct tracer_stat *trace) +{ + return &(ksym_filter_head.first); +} + +static void * +ksym_tracer_stat_next(void *v, int idx) +{ + struct hlist_node *stat = v; + + return stat->next; +} + +static struct tracer_stat ksym_tracer_stats = { + .name = "ksym_tracer", + .stat_start = ksym_tracer_stat_start, + .stat_next = ksym_tracer_stat_next, + .stat_headers = ksym_tracer_stat_headers, + .stat_show = ksym_tracer_stat_show +}; + +__init static int ksym_tracer_stat_init(void) +{ + int ret; + + ret = register_stat_tracer(&ksym_tracer_stats); + if (ret) { + printk(KERN_WARNING "Warning: could not register " + "ksym tracer stats\n"); + return 1; + } + + return 0; +} +fs_initcall(ksym_tracer_stat_init); +#endif /* CONFIG_PROFILE_KSYM_TRACER */ diff --git a/kernel/trace/trace_selftest.c b/kernel/trace/trace_selftest.c index 00dd648..71f2edb 100644 --- a/kernel/trace/trace_selftest.c +++ b/kernel/trace/trace_selftest.c @@ -17,6 +17,7 @@ static inline int trace_valid_entry(struct trace_entry *entry) case TRACE_GRAPH_ENT: case TRACE_GRAPH_RET: case TRACE_HW_BRANCHES: + case TRACE_KSYM: return 1; } return 0; @@ -807,3 +808,55 @@ trace_selftest_startup_hw_branches(struct tracer *trace, return ret; } #endif /* CONFIG_HW_BRANCH_TRACER */ + +#ifdef CONFIG_KSYM_TRACER +static int ksym_selftest_dummy; + +int +trace_selftest_startup_ksym(struct tracer *trace, struct trace_array *tr) +{ + unsigned long count; + int ret; + + /* start the tracing */ + ret = tracer_init(trace, tr); + if (ret) { + warn_failed_init_tracer(trace, ret); + return ret; + } + + ksym_selftest_dummy = 0; + /* Register the read-write tracing request */ + ret = process_new_ksym_entry(KSYM_SELFTEST_ENTRY, HW_BREAKPOINT_RW, + (unsigned long)(&ksym_selftest_dummy)); + + if (ret < 0) { + printk(KERN_CONT "ksym_trace read-write startup test failed\n"); + goto ret_path; + } + /* Perform a read and a write operation over the dummy variable to + * trigger the tracer + */ + if (ksym_selftest_dummy == 0) + ksym_selftest_dummy++; + + /* stop the tracing. */ + tracing_stop(); + /* check the trace buffer */ + ret = trace_test_buffer(tr, &count); + trace->reset(tr); + tracing_start(); + + /* read & write operations - one each is performed on the dummy variable + * triggering two entries in the trace buffer + */ + if (!ret && count != 2) { + printk(KERN_CONT "Ksym tracer startup test failed"); + ret = -1; + } + +ret_path: + return ret; +} +#endif /* CONFIG_KSYM_TRACER */ + -- cgit v1.1 From 73874005cd8800440be4299bd095387fff4b90ac Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 3 Jun 2009 01:43:38 +0200 Subject: hw-breakpoints: fix undeclared ksym_tracer_mutex MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit ksym_tracer_mutex is declared inside an #ifdef CONFIG_PROFILE_KSYM_TRACER section. This makes it unavailable for the hardware breakpoint tracer if it is configured without the breakpoint profiler. This patch fixes the following build error: kernel/trace/trace_ksym.c: In function ‘ksym_trace_filter_read’: kernel/trace/trace_ksym.c:226: erreur: ‘ksym_tracer_mutex’ undeclared (first use in this function) kernel/trace/trace_ksym.c:226: erreur: (Each undeclared identifier is reported only once kernel/trace/trace_ksym.c:226: erreur: for each function it appears in.) kernel/trace/trace_ksym.c: In function ‘ksym_trace_filter_write’: kernel/trace/trace_ksym.c:273: erreur: ‘ksym_tracer_mutex’ undeclared (first use in this function) kernel/trace/trace_ksym.c: In function ‘ksym_trace_reset’: kernel/trace/trace_ksym.c:335: erreur: ‘ksym_tracer_mutex’ undeclared (first use in this function) make[1]: *** [kernel/trace/trace_ksym.o] Erreur 1 [ Impact: fix a build error ] Reported-by: Ingo Molnar Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_ksym.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index 11c74f6..eef97e7 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -44,12 +44,12 @@ static unsigned int ksym_tracing_enabled; static HLIST_HEAD(ksym_filter_head); +static DEFINE_MUTEX(ksym_tracer_mutex); + #ifdef CONFIG_PROFILE_KSYM_TRACER #define MAX_UL_INT 0xffffffff -static DEFINE_MUTEX(ksym_tracer_mutex); - void ksym_collect_stats(unsigned long hbp_hit_addr) { struct hlist_node *node; -- cgit v1.1 From db59504d89db1462a5281fb55b1d962cb74a398f Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 7 Jul 2009 13:52:36 +0800 Subject: ksym_tracer: Extract trace entry from struct trace_ksym struct trace_ksym is used as an entry in hbp list, and is also used as trace_entry stored in ring buffer. This is not necessary and is a waste of memory in ring buffer. There is also a bug that dereferencing field->ksym_hbp in ksym_trace_output() can be invalid. Signed-off-by: Li Zefan Acked-by: Frederic Weisbecker Cc: "K.Prasad" Cc: Alan Stern Cc: Steven Rostedt LKML-Reference: <4A52E2A4.4050007@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace.h | 13 ++++--------- kernel/trace/trace_ksym.c | 26 ++++++++++++++++++-------- 2 files changed, 22 insertions(+), 17 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 7d5cc37..ff1ef41 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -215,17 +215,12 @@ struct syscall_trace_exit { #define KSYM_SELFTEST_ENTRY "ksym_selftest_dummy" extern int process_new_ksym_entry(char *ksymname, int op, unsigned long addr); -struct trace_ksym { +struct ksym_trace_entry { struct trace_entry ent; - struct hw_breakpoint *ksym_hbp; - unsigned long ksym_addr; unsigned long ip; -#ifdef CONFIG_PROFILE_KSYM_TRACER - unsigned long counter; -#endif - struct hlist_node ksym_hlist; + unsigned char type; char ksym_name[KSYM_NAME_LEN]; - char p_name[TASK_COMM_LEN]; + char cmd[TASK_COMM_LEN]; }; /* @@ -343,7 +338,7 @@ extern void __ftrace_bad_type(void); TRACE_SYSCALL_ENTER); \ IF_ASSIGN(var, ent, struct syscall_trace_exit, \ TRACE_SYSCALL_EXIT); \ - IF_ASSIGN(var, ent, struct trace_ksym, TRACE_KSYM); \ + IF_ASSIGN(var, ent, struct ksym_trace_entry, TRACE_KSYM);\ __ftrace_bad_type(); \ } while (0) diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index eef97e7..085ff05 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -37,6 +37,15 @@ #define KSYM_TRACER_OP_LEN 3 /* rw- */ #define KSYM_FILTER_ENTRY_LEN (KSYM_NAME_LEN + KSYM_TRACER_OP_LEN + 1) +struct trace_ksym { + struct hw_breakpoint *ksym_hbp; + unsigned long ksym_addr; +#ifdef CONFIG_PROFILE_KSYM_TRACER + unsigned long counter; +#endif + struct hlist_node ksym_hlist; +}; + static struct trace_array *ksym_trace_array; static unsigned int ksym_filter_entry_count; @@ -71,7 +80,7 @@ void ksym_hbp_handler(struct hw_breakpoint *hbp, struct pt_regs *regs) { struct ring_buffer_event *event; struct trace_array *tr; - struct trace_ksym *entry; + struct ksym_trace_entry *entry; int pc; if (!ksym_tracing_enabled) @@ -85,11 +94,12 @@ void ksym_hbp_handler(struct hw_breakpoint *hbp, struct pt_regs *regs) if (!event) return; - entry = ring_buffer_event_data(event); + entry = ring_buffer_event_data(event); + entry->ip = instruction_pointer(regs); + entry->type = hbp->info.type; strlcpy(entry->ksym_name, hbp->info.name, KSYM_SYMBOL_LEN); - entry->ksym_hbp = hbp; - entry->ip = instruction_pointer(regs); - strlcpy(entry->p_name, current->comm, TASK_COMM_LEN); + strlcpy(entry->cmd, current->comm, TASK_COMM_LEN); + #ifdef CONFIG_PROFILE_KSYM_TRACER ksym_collect_stats(hbp->info.address); #endif /* CONFIG_PROFILE_KSYM_TRACER */ @@ -380,7 +390,7 @@ static enum print_line_t ksym_trace_output(struct trace_iterator *iter) { struct trace_entry *entry = iter->ent; struct trace_seq *s = &iter->seq; - struct trace_ksym *field; + struct ksym_trace_entry *field; char str[KSYM_SYMBOL_LEN]; int ret; @@ -389,12 +399,12 @@ static enum print_line_t ksym_trace_output(struct trace_iterator *iter) trace_assign_type(field, entry); - ret = trace_seq_printf(s, "%-15s %-5d %-3d %-20s ", field->p_name, + ret = trace_seq_printf(s, "%-15s %-5d %-3d %-20s ", field->cmd, entry->pid, iter->cpu, field->ksym_name); if (!ret) return TRACE_TYPE_PARTIAL_LINE; - switch (field->ksym_hbp->info.type) { + switch (field->type) { case HW_BREAKPOINT_WRITE: ret = trace_seq_printf(s, " W "); break; -- cgit v1.1 From be9742e6cb107fe1d77db7a081ea4eb25e79e1ad Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 7 Jul 2009 13:52:52 +0800 Subject: ksym_tracer: Rewrite ksym_trace_filter_read() Reading ksym_trace_filter gave me some arbitrary characters, when it should show nothing. It's because buf is not initialized when there's no filter. Also reduce stack usage by about 512 bytes. Signed-off-by: Li Zefan Acked-by: Frederic Weisbecker Cc: "K.Prasad" Cc: Alan Stern Cc: Steven Rostedt LKML-Reference: <4A52E2B4.6030706@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_ksym.c | 29 ++++++++++++++++++----------- 1 file changed, 18 insertions(+), 11 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index 085ff05..b6710d3 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -35,7 +35,6 @@ #define KSYM_TRACER_MAX HBP_NUM #define KSYM_TRACER_OP_LEN 3 /* rw- */ -#define KSYM_FILTER_ENTRY_LEN (KSYM_NAME_LEN + KSYM_TRACER_OP_LEN + 1) struct trace_ksym { struct hw_breakpoint *ksym_hbp; @@ -230,25 +229,33 @@ static ssize_t ksym_trace_filter_read(struct file *filp, char __user *ubuf, { struct trace_ksym *entry; struct hlist_node *node; - char buf[KSYM_FILTER_ENTRY_LEN * KSYM_TRACER_MAX]; - ssize_t ret, cnt = 0; + struct trace_seq *s; + ssize_t cnt = 0; + int ret; + + s = kmalloc(sizeof(*s), GFP_KERNEL); + if (!s) + return -ENOMEM; + trace_seq_init(s); mutex_lock(&ksym_tracer_mutex); hlist_for_each_entry(entry, node, &ksym_filter_head, ksym_hlist) { - cnt += snprintf(&buf[cnt], KSYM_FILTER_ENTRY_LEN - cnt, "%s:", - entry->ksym_hbp->info.name); + ret = trace_seq_printf(s, "%s:", entry->ksym_hbp->info.name); if (entry->ksym_hbp->info.type == HW_BREAKPOINT_WRITE) - cnt += snprintf(&buf[cnt], KSYM_FILTER_ENTRY_LEN - cnt, - "-w-\n"); + ret = trace_seq_puts(s, "-w-\n"); else if (entry->ksym_hbp->info.type == HW_BREAKPOINT_RW) - cnt += snprintf(&buf[cnt], KSYM_FILTER_ENTRY_LEN - cnt, - "rw-\n"); + ret = trace_seq_puts(s, "rw-\n"); + WARN_ON_ONCE(!ret); } - ret = simple_read_from_buffer(ubuf, count, ppos, buf, strlen(buf)); + + cnt = simple_read_from_buffer(ubuf, count, ppos, s->buffer, s->len); + mutex_unlock(&ksym_tracer_mutex); - return ret; + kfree(s); + + return cnt; } static ssize_t ksym_trace_filter_write(struct file *file, -- cgit v1.1 From f088e5471297cc78d7465e1fd997cb1a91a48019 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 7 Jul 2009 13:53:18 +0800 Subject: ksym_tracer: Fix validation of access type # echo 'pid_max:rw-' > ksym_trace_filter # cat ksym_trace_filter pid_max:rw- # echo 'pid_max:ww-' > ksym_trace_filter (should return -EINVAL) # cat ksym_trace_filter (but it ended up removing filter entry) Signed-off-by: Li Zefan Acked-by: Frederic Weisbecker Cc: "K.Prasad" Cc: Alan Stern Cc: Steven Rostedt LKML-Reference: <4A52E2CE.6080409@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_ksym.c | 32 ++++++++++++++------------------ 1 file changed, 14 insertions(+), 18 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index b6710d3..9556009 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -114,24 +114,22 @@ void ksym_hbp_handler(struct hw_breakpoint *hbp, struct pt_regs *regs) * --x : Set Execution Break points (Not available yet) * */ -static int ksym_trace_get_access_type(char *access_str) +static int ksym_trace_get_access_type(char *str) { - int pos, access = 0; + int access = 0; - for (pos = 0; pos < KSYM_TRACER_OP_LEN; pos++) { - switch (access_str[pos]) { - case 'r': - access += (pos == 0) ? 4 : -1; - break; - case 'w': - access += (pos == 1) ? 2 : -1; - break; - case '-': - break; - default: - return -EINVAL; - } - } + if (str[0] == 'r') + access += 4; + else if (str[0] != '-') + return -EINVAL; + + if (str[1] == 'w') + access += 2; + else if (str[1] != '-') + return -EINVAL; + + if (str[2] != '-') + return -EINVAL; switch (access) { case 6: @@ -140,8 +138,6 @@ static int ksym_trace_get_access_type(char *access_str) case 2: access = HW_BREAKPOINT_WRITE; break; - case 0: - access = 0; } return access; -- cgit v1.1 From 92cf9f8f7e89c6bdbb1a724f879b8b18fc0dfe0f Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 7 Jul 2009 13:53:47 +0800 Subject: ksym_tracer: Fix validation of length of access type Don't take newline into account, otherwise: # echo 'pid_max:-w-' > ksym_trace_filter # echo -n 'pid_max:rw-' > ksym_trace_filter bash: echo: write error: Invalid argument Signed-off-by: Li Zefan Acked-by: Frederic Weisbecker Cc: "K.Prasad" Cc: Alan Stern Cc: Steven Rostedt LKML-Reference: <4A52E2EB.9070503@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_ksym.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index 9556009..72fcb46 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -158,21 +158,21 @@ static int ksym_trace_get_access_type(char *str) static int parse_ksym_trace_str(char *input_string, char **ksymname, unsigned long *addr) { - char *delimiter = ":"; int ret; - ret = -EINVAL; - *ksymname = strsep(&input_string, delimiter); + strstrip(input_string); + + *ksymname = strsep(&input_string, ":"); *addr = kallsyms_lookup_name(*ksymname); /* Check for malformed request: (2), (1) and (5) */ if ((!input_string) || - (strlen(input_string) != (KSYM_TRACER_OP_LEN + 1)) || - (*addr == 0)) - goto return_code; + (strlen(input_string) != KSYM_TRACER_OP_LEN) || + (*addr == 0)) + return -EINVAL;; + ret = ksym_trace_get_access_type(input_string); -return_code: return ret; } -- cgit v1.1 From 011ed56853e07e30653d6f1bfddc56b396218664 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 7 Jul 2009 13:54:08 +0800 Subject: ksym_tracer: NIL-terminate user input filter Make sure the user input string is NULL-terminated. Signed-off-by: Li Zefan Acked-by: Frederic Weisbecker Cc: "K.Prasad" Cc: Alan Stern Cc: Steven Rostedt LKML-Reference: <4A52E300.7020601@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_ksym.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index 72fcb46..8cbed5a 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -264,11 +264,7 @@ static ssize_t ksym_trace_filter_write(struct file *file, unsigned long ksym_addr = 0; int ret, op, changed = 0; - /* Ignore echo "" > ksym_trace_filter */ - if (count == 0) - return 0; - - input_string = kzalloc(count, GFP_KERNEL); + input_string = kzalloc(count + 1, GFP_KERNEL); if (!input_string) return -ENOMEM; @@ -276,6 +272,7 @@ static ssize_t ksym_trace_filter_write(struct file *file, kfree(input_string); return -EFAULT; } + input_string[count] = '\0'; ret = op = parse_ksym_trace_str(input_string, &ksymname, &ksym_addr); if (ret < 0) { -- cgit v1.1 From 0d109c8f70eab8b9f693bd5caea23012394e4876 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 7 Jul 2009 13:54:28 +0800 Subject: ksym_tracer: Report error when failed to re-register hbp When access type is changed, the hw break point will be unregistered and then be registered again with new access type. But the registration may fail, in this case, -errno should be returned. Signed-off-by: Li Zefan Acked-by: Frederic Weisbecker Cc: "K.Prasad" Cc: Alan Stern Cc: Steven Rostedt LKML-Reference: <4A52E314.7070004@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_ksym.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index 8cbed5a..891e3b8 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -302,13 +302,13 @@ static ssize_t ksym_trace_filter_write(struct file *file, ret = count; goto unlock_ret_path; } - } + } else + ret = count; ksym_filter_entry_count--; hlist_del_rcu(&(entry->ksym_hlist)); synchronize_rcu(); kfree(entry->ksym_hbp); kfree(entry); - ret = count; goto err_ret; } else { /* Check for malformed request: (4) */ -- cgit v1.1 From 558df6c8f74ac4a0b9026ef85b0028280f364d96 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 7 Jul 2009 13:54:48 +0800 Subject: ksym_tracer: Fix memory leak - When remove a filter, we leak entry->ksym_hbp->info.name. - With CONFIG_FTRAC_SELFTEST enabled, we leak ->info.name: # echo ksym_tracer > current_tracer # echo 'ksym_selftest_dummy:rw-' > ksym_trace_filter # echo nop > current_tracer Signed-off-by: Li Zefan Acked-by: Frederic Weisbecker Cc: "K.Prasad" Cc: Alan Stern Cc: Steven Rostedt LKML-Reference: <4A52E328.8010200@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_ksym.c | 61 +++++++++++++++++++++-------------------------- 1 file changed, 27 insertions(+), 34 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index 891e3b8..7d349d3 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -179,7 +179,7 @@ static int parse_ksym_trace_str(char *input_string, char **ksymname, int process_new_ksym_entry(char *ksymname, int op, unsigned long addr) { struct trace_ksym *entry; - int ret; + int ret = -ENOMEM; if (ksym_filter_entry_count >= KSYM_TRACER_MAX) { printk(KERN_ERR "ksym_tracer: Maximum limit:(%d) reached. No" @@ -193,12 +193,13 @@ int process_new_ksym_entry(char *ksymname, int op, unsigned long addr) return -ENOMEM; entry->ksym_hbp = kzalloc(sizeof(struct hw_breakpoint), GFP_KERNEL); - if (!entry->ksym_hbp) { - kfree(entry); - return -ENOMEM; - } + if (!entry->ksym_hbp) + goto err; + + entry->ksym_hbp->info.name = kstrdup(ksymname, GFP_KERNEL); + if (!entry->ksym_hbp->info.name) + goto err; - entry->ksym_hbp->info.name = ksymname; entry->ksym_hbp->info.type = op; entry->ksym_addr = entry->ksym_hbp->info.address = addr; #ifdef CONFIG_X86 @@ -210,14 +211,18 @@ int process_new_ksym_entry(char *ksymname, int op, unsigned long addr) if (ret < 0) { printk(KERN_INFO "ksym_tracer request failed. Try again" " later!!\n"); - kfree(entry->ksym_hbp); - kfree(entry); - return -EAGAIN; + ret = -EAGAIN; + goto err; } hlist_add_head_rcu(&(entry->ksym_hlist), &ksym_filter_head); ksym_filter_entry_count++; - return 0; +err: + if (entry->ksym_hbp) + kfree(entry->ksym_hbp->info.name); + kfree(entry->ksym_hbp); + kfree(entry); + return ret; } static ssize_t ksym_trace_filter_read(struct file *filp, char __user *ubuf, @@ -289,7 +294,7 @@ static ssize_t ksym_trace_filter_write(struct file *file, if (entry->ksym_hbp->info.type != op) changed = 1; else - goto err_ret; + goto out; break; } } @@ -298,34 +303,29 @@ static ssize_t ksym_trace_filter_write(struct file *file, entry->ksym_hbp->info.type = op; if (op > 0) { ret = register_kernel_hw_breakpoint(entry->ksym_hbp); - if (ret == 0) { - ret = count; - goto unlock_ret_path; - } - } else - ret = count; + if (ret == 0) + goto out; + } ksym_filter_entry_count--; hlist_del_rcu(&(entry->ksym_hlist)); synchronize_rcu(); + kfree(entry->ksym_hbp->info.name); kfree(entry->ksym_hbp); kfree(entry); - goto err_ret; + goto out; } else { /* Check for malformed request: (4) */ if (op == 0) - goto err_ret; + goto out; ret = process_new_ksym_entry(ksymname, op, ksym_addr); - if (ret) - goto err_ret; } - ret = count; - goto unlock_ret_path; +out: + mutex_unlock(&ksym_tracer_mutex); -err_ret: kfree(input_string); -unlock_ret_path: - mutex_unlock(&ksym_tracer_mutex); + if (!ret) + ret = count; return ret; } @@ -349,14 +349,7 @@ static void ksym_trace_reset(struct trace_array *tr) ksym_filter_entry_count--; hlist_del_rcu(&(entry->ksym_hlist)); synchronize_rcu(); - /* Free the 'input_string' only if reset - * after startup self-test - */ -#ifdef CONFIG_FTRACE_SELFTEST - if (strncmp(entry->ksym_hbp->info.name, KSYM_SELFTEST_ENTRY, - strlen(KSYM_SELFTEST_ENTRY)) != 0) -#endif /* CONFIG_FTRACE_SELFTEST*/ - kfree(entry->ksym_hbp->info.name); + kfree(entry->ksym_hbp->info.name); kfree(entry->ksym_hbp); kfree(entry); } -- cgit v1.1 From 9d7e934408b52cd53dd85270eb36941a6a318cc5 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 7 Jul 2009 13:55:18 +0800 Subject: ksym_tracer: Fix the output of stat tracing - make ksym_tracer_stat_start() return head->first instead of &head->first - make the output properly aligned Before: Access type Symbol Counter NA 0 RW pid_max 0 After: Access Type Symbol Counter ----------- ------ ------- RW pid_max 0 Signed-off-by: Li Zefan Acked-by: Frederic Weisbecker Cc: "K.Prasad" Cc: Alan Stern Cc: Steven Rostedt LKML-Reference: <4A52E346.5050608@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_ksym.c | 20 +++++++++++--------- 1 file changed, 11 insertions(+), 9 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index 7d349d3..1256a6e 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -453,8 +453,10 @@ device_initcall(init_ksym_trace); #ifdef CONFIG_PROFILE_KSYM_TRACER static int ksym_tracer_stat_headers(struct seq_file *m) { - seq_printf(m, " Access type "); - seq_printf(m, " Symbol Counter \n"); + seq_puts(m, " Access Type "); + seq_puts(m, " Symbol Counter\n"); + seq_puts(m, " ----------- "); + seq_puts(m, " ------ -------\n"); return 0; } @@ -472,27 +474,27 @@ static int ksym_tracer_stat_show(struct seq_file *m, void *v) switch (access_type) { case HW_BREAKPOINT_WRITE: - seq_printf(m, " W "); + seq_puts(m, " W "); break; case HW_BREAKPOINT_RW: - seq_printf(m, " RW "); + seq_puts(m, " RW "); break; default: - seq_printf(m, " NA "); + seq_puts(m, " NA "); } if (lookup_symbol_name(entry->ksym_addr, fn_name) >= 0) - seq_printf(m, " %s ", fn_name); + seq_printf(m, " %-36s", fn_name); else - seq_printf(m, " "); + seq_printf(m, " %-36s", ""); + seq_printf(m, " %15lu\n", entry->counter); - seq_printf(m, "%15lu\n", entry->counter); return 0; } static void *ksym_tracer_stat_start(struct tracer_stat *trace) { - return &(ksym_filter_head.first); + return ksym_filter_head.first; } static void * -- cgit v1.1 From d857ace143df3884954887e1899a65831ca72ece Mon Sep 17 00:00:00 2001 From: Xiao Guangrong Date: Wed, 22 Jul 2009 11:21:31 +0800 Subject: tracing/ksym_tracer: fix the output of ksym tracer Fix the output format of ksym tracer, make it properly aligned Befor patch: # tracer: ksym_tracer # # TASK-PID CPU# Symbol Type Function # | | | | | bash 1378 1 ksym_tracer_mutex W mutex_lock+0x11/0x27 bash 1378 1 ksym_filter_head W process_new_ksym_entry+0xd2/0x10c bash 1378 1 ksym_tracer_mutex W mutex_unlock+0x12/0x1b cat 1429 0 ksym_tracer_mutex W mutex_lock+0x11/0x27 After patch: # tracer: ksym_tracer # # TASK-PID CPU# Symbol Type Function # | | | | | cat-1423 [000] ksym_tracer_mutex RW mutex_lock+0x11/0x27 cat-1423 [000] ksym_filter_head RW ksym_trace_filter_read+0x6e/0x10d cat-1423 [000] ksym_tracer_mutex RW mutex_unlock+0x12/0x1b cat-1423 [000] ksym_tracer_mutex RW mutex_lock+0x11/0x27 cat-1423 [000] ksym_filter_head RW ksym_trace_filter_read+0x6e/0x10d cat-1423 [000] ksym_tracer_mutex RW mutex_unlock+0x12/0x1b Signed-off-by: Xiao Guangrong LKML-Reference: <4A6685BB.2090809@cn.fujitsu.com> Signed-off-by: Steven Rostedt --- kernel/trace/trace_ksym.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index 1256a6e..fbf3a8e 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -370,13 +370,12 @@ static int ksym_trace_init(struct trace_array *tr) static void ksym_trace_print_header(struct seq_file *m) { - seq_puts(m, - "# TASK-PID CPU# Symbol Type " - "Function \n"); + "# TASK-PID CPU# Symbol " + "Type Function\n"); seq_puts(m, - "# | | | | " - "| \n"); + "# | | | " + " | |\n"); } static enum print_line_t ksym_trace_output(struct trace_iterator *iter) @@ -392,7 +391,7 @@ static enum print_line_t ksym_trace_output(struct trace_iterator *iter) trace_assign_type(field, entry); - ret = trace_seq_printf(s, "%-15s %-5d %-3d %-20s ", field->cmd, + ret = trace_seq_printf(s, "%11s-%-5d [%03d] %-30s ", field->cmd, entry->pid, iter->cpu, field->ksym_name); if (!ret) return TRACE_TYPE_PARTIAL_LINE; @@ -412,7 +411,7 @@ static enum print_line_t ksym_trace_output(struct trace_iterator *iter) return TRACE_TYPE_PARTIAL_LINE; sprint_symbol(str, field->ip); - ret = trace_seq_printf(s, "%-20s\n", str); + ret = trace_seq_printf(s, "%s\n", str); if (!ret) return TRACE_TYPE_PARTIAL_LINE; -- cgit v1.1 From 8e068542a8d9efec55126284d2f5cb32f003d507 Mon Sep 17 00:00:00 2001 From: Xiao Guangrong Date: Wed, 22 Jul 2009 11:23:41 +0800 Subject: tracing/ksym_tracer: fix write operation of ksym_trace_filter This patch fix 2 bugs: - fix the return value of ksym_trace_filter_write() when we want to clear symbol in ksym_trace_filter file for example: # echo global_trace:rw- > /debug/tracing/ksym_trace_filter # echo global_trace:--- > /debug/tracing/ksym_trace_filter -bash: echo: write error: Invalid argument # cat /debug/tracing/ksym_trace_filter # We want to clear 'global_trace' in ksym_trace_filter, it complain with "Invalid argument", but the operation is successful - the "r--" access types is not allowed, but ksym_trace_filter file think it OK for example: # echo ksym_tracer_mutex:r-- > ksym_trace_filter -bash: echo: write error: Resource temporarily unavailable # dmesg ksym_tracer request failed. Try again later!! The error occur at register_kernel_hw_breakpoint(), but It's should at access types parser Signed-off-by: Xiao Guangrong LKML-Reference: <4A66863D.5090802@cn.fujitsu.com> Signed-off-by: Steven Rostedt --- kernel/trace/trace_ksym.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'kernel') diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index fbf3a8e..cd5cb65 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -135,6 +135,9 @@ static int ksym_trace_get_access_type(char *str) case 6: access = HW_BREAKPOINT_RW; break; + case 4: + access = -EINVAL; + break; case 2: access = HW_BREAKPOINT_WRITE; break; @@ -312,6 +315,7 @@ static ssize_t ksym_trace_filter_write(struct file *file, kfree(entry->ksym_hbp->info.name); kfree(entry->ksym_hbp); kfree(entry); + ret = 0; goto out; } else { /* Check for malformed request: (4) */ -- cgit v1.1 From 75e33751ca8bbb72dd6f1a74d2810ddc8cbe4bdf Mon Sep 17 00:00:00 2001 From: Xiao Guangrong Date: Thu, 23 Jul 2009 12:01:22 +0800 Subject: tracing/ksym_tracer: support quick clear for ksym_trace_filter -- v2 It's rather boring to clear symbol one by one in ksym_trace_filter file, so, this patch will let ksym_trace_filter file support quickly clear all break points. We can write "0" to this file and it will clear all symbols for example: # cat ksym_trace_filter ksym_filter_head:rw- global_trace:rw- # echo 0 > ksym_trace_filter # cat ksym_trace_filter # Changelog v1->v2: Add other ways to clear all breakpoints by writing NULL or "*:---" to ksym_trace_filter file base on K.Prasad's suggestion Signed-off-by: Xiao Guangrong LKML-Reference: <4A67E092.3080202@cn.fujitsu.com> Signed-off-by: Steven Rostedt --- kernel/trace/trace_ksym.c | 53 +++++++++++++++++++++++++++++++---------------- 1 file changed, 35 insertions(+), 18 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index cd5cb65..2fde875 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -163,8 +163,6 @@ static int parse_ksym_trace_str(char *input_string, char **ksymname, { int ret; - strstrip(input_string); - *ksymname = strsep(&input_string, ":"); *addr = kallsyms_lookup_name(*ksymname); @@ -262,6 +260,25 @@ static ssize_t ksym_trace_filter_read(struct file *filp, char __user *ubuf, return cnt; } +static void __ksym_trace_reset(void) +{ + struct trace_ksym *entry; + struct hlist_node *node, *node1; + + mutex_lock(&ksym_tracer_mutex); + hlist_for_each_entry_safe(entry, node, node1, &ksym_filter_head, + ksym_hlist) { + unregister_kernel_hw_breakpoint(entry->ksym_hbp); + ksym_filter_entry_count--; + hlist_del_rcu(&(entry->ksym_hlist)); + synchronize_rcu(); + kfree(entry->ksym_hbp->info.name); + kfree(entry->ksym_hbp); + kfree(entry); + } + mutex_unlock(&ksym_tracer_mutex); +} + static ssize_t ksym_trace_filter_write(struct file *file, const char __user *buffer, size_t count, loff_t *ppos) @@ -282,6 +299,21 @@ static ssize_t ksym_trace_filter_write(struct file *file, } input_string[count] = '\0'; + strstrip(input_string); + + /* + * Clear all breakpoints if: + * 1: echo > ksym_trace_filter + * 2: echo 0 > ksym_trace_filter + * 3: echo "*:---" > ksym_trace_filter + */ + if (!input_string[0] || !strcmp(input_string, "0") || + !strcmp(input_string, "*:---")) { + __ksym_trace_reset(); + kfree(input_string); + return count; + } + ret = op = parse_ksym_trace_str(input_string, &ksymname, &ksym_addr); if (ret < 0) { kfree(input_string); @@ -341,23 +373,8 @@ static const struct file_operations ksym_tracing_fops = { static void ksym_trace_reset(struct trace_array *tr) { - struct trace_ksym *entry; - struct hlist_node *node, *node1; - ksym_tracing_enabled = 0; - - mutex_lock(&ksym_tracer_mutex); - hlist_for_each_entry_safe(entry, node, node1, &ksym_filter_head, - ksym_hlist) { - unregister_kernel_hw_breakpoint(entry->ksym_hbp); - ksym_filter_entry_count--; - hlist_del_rcu(&(entry->ksym_hlist)); - synchronize_rcu(); - kfree(entry->ksym_hbp->info.name); - kfree(entry->ksym_hbp); - kfree(entry); - } - mutex_unlock(&ksym_tracer_mutex); + __ksym_trace_reset(); } static int ksym_trace_init(struct trace_array *tr) -- cgit v1.1 From bd1a5c849bdcc5c89e4a6a18216cd2b9a7a8a78f Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 13 Aug 2009 16:34:53 -0400 Subject: tracing: Ftrace dynamic ftrace_event_call support MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add dynamic ftrace_event_call support to ftrace. Trace engines can add new ftrace_event_call to ftrace on the fly. Each operator function of the call takes an ftrace_event_call data structure as an argument, because these functions may be shared among several ftrace_event_calls. Changes from v13: - Define remove_subsystem_dir() always (revirt a2ca5e03), because trace_remove_event_call() uses it. - Modify syscall tracer because of ftrace_event_call change. [fweisbec@gmail.com: Fixed conflict against latest tracing/core] Signed-off-by: Masami Hiramatsu Cc: Ananth N Mavinakayanahalli Cc: Avi Kivity Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: Jim Keniston Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Przemysław Pawełczyk Cc: Roland McGrath Cc: Sam Ravnborg Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi Cc: Vegard Nossum LKML-Reference: <20090813203453.31965.71901.stgit@localhost.localdomain> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_events.c | 121 +++++++++++++++++++++++++++++------------- kernel/trace/trace_export.c | 18 +++---- kernel/trace/trace_syscalls.c | 20 +++---- 3 files changed, 103 insertions(+), 56 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c index d33bcde..8079bb5 100644 --- a/kernel/trace/trace_events.c +++ b/kernel/trace/trace_events.c @@ -27,8 +27,8 @@ DEFINE_MUTEX(event_mutex); LIST_HEAD(ftrace_events); -int trace_define_field(struct ftrace_event_call *call, const char *type, - const char *name, int offset, int size, int is_signed, +int trace_define_field(struct ftrace_event_call *call, char *type, + char *name, int offset, int size, int is_signed, int filter_type) { struct ftrace_event_field *field; @@ -92,9 +92,7 @@ int trace_define_common_fields(struct ftrace_event_call *call) } EXPORT_SYMBOL_GPL(trace_define_common_fields); -#ifdef CONFIG_MODULES - -static void trace_destroy_fields(struct ftrace_event_call *call) +void trace_destroy_fields(struct ftrace_event_call *call) { struct ftrace_event_field *field, *next; @@ -106,8 +104,6 @@ static void trace_destroy_fields(struct ftrace_event_call *call) } } -#endif /* CONFIG_MODULES */ - static void ftrace_event_enable_disable(struct ftrace_event_call *call, int enable) { @@ -116,14 +112,14 @@ static void ftrace_event_enable_disable(struct ftrace_event_call *call, if (call->enabled) { call->enabled = 0; tracing_stop_cmdline_record(); - call->unregfunc(call->data); + call->unregfunc(call); } break; case 1: if (!call->enabled) { call->enabled = 1; tracing_start_cmdline_record(); - call->regfunc(call->data); + call->regfunc(call); } break; } @@ -991,27 +987,43 @@ event_create_dir(struct ftrace_event_call *call, struct dentry *d_events, return 0; } -#define for_each_event(event, start, end) \ - for (event = start; \ - (unsigned long)event < (unsigned long)end; \ - event++) +static int __trace_add_event_call(struct ftrace_event_call *call) +{ + struct dentry *d_events; + int ret; -#ifdef CONFIG_MODULES + if (!call->name) + return -EINVAL; -static LIST_HEAD(ftrace_module_file_list); + if (call->raw_init) { + ret = call->raw_init(call); + if (ret < 0) { + if (ret != -ENOSYS) + pr_warning("Could not initialize trace " + "events/%s\n", call->name); + return ret; + } + } -/* - * Modules must own their file_operations to keep up with - * reference counting. - */ -struct ftrace_module_file_ops { - struct list_head list; - struct module *mod; - struct file_operations id; - struct file_operations enable; - struct file_operations format; - struct file_operations filter; -}; + d_events = event_trace_events_dir(); + if (!d_events) + return -ENOENT; + + list_add(&call->list, &ftrace_events); + return event_create_dir(call, d_events, &ftrace_event_id_fops, + &ftrace_enable_fops, &ftrace_event_filter_fops, + &ftrace_event_format_fops); +} + +/* Add an additional event_call dynamically */ +int trace_add_event_call(struct ftrace_event_call *call) +{ + int ret; + mutex_lock(&event_mutex); + ret = __trace_add_event_call(call); + mutex_unlock(&event_mutex); + return ret; +} static void remove_subsystem_dir(const char *name) { @@ -1039,6 +1051,48 @@ static void remove_subsystem_dir(const char *name) } } +static void __trace_remove_event_call(struct ftrace_event_call *call) +{ + ftrace_event_enable_disable(call, 0); + if (call->event) + __unregister_ftrace_event(call->event); + debugfs_remove_recursive(call->dir); + list_del(&call->list); + trace_destroy_fields(call); + destroy_preds(call); + remove_subsystem_dir(call->system); +} + +/* Remove an event_call */ +void trace_remove_event_call(struct ftrace_event_call *call) +{ + mutex_lock(&event_mutex); + __trace_remove_event_call(call); + mutex_unlock(&event_mutex); +} + +#define for_each_event(event, start, end) \ + for (event = start; \ + (unsigned long)event < (unsigned long)end; \ + event++) + +#ifdef CONFIG_MODULES + +static LIST_HEAD(ftrace_module_file_list); + +/* + * Modules must own their file_operations to keep up with + * reference counting. + */ +struct ftrace_module_file_ops { + struct list_head list; + struct module *mod; + struct file_operations id; + struct file_operations enable; + struct file_operations format; + struct file_operations filter; +}; + static struct ftrace_module_file_ops * trace_create_file_ops(struct module *mod) { @@ -1096,7 +1150,7 @@ static void trace_module_add_events(struct module *mod) if (!call->name) continue; if (call->raw_init) { - ret = call->raw_init(); + ret = call->raw_init(call); if (ret < 0) { if (ret != -ENOSYS) pr_warning("Could not initialize trace " @@ -1131,14 +1185,7 @@ static void trace_module_remove_events(struct module *mod) list_for_each_entry_safe(call, p, &ftrace_events, list) { if (call->mod == mod) { found = true; - ftrace_event_enable_disable(call, 0); - if (call->event) - __unregister_ftrace_event(call->event); - debugfs_remove_recursive(call->dir); - list_del(&call->list); - trace_destroy_fields(call); - destroy_preds(call); - remove_subsystem_dir(call->system); + __trace_remove_event_call(call); } } @@ -1256,7 +1303,7 @@ static __init int event_trace_init(void) if (!call->name) continue; if (call->raw_init) { - ret = call->raw_init(); + ret = call->raw_init(call); if (ret < 0) { if (ret != -ENOSYS) pr_warning("Could not initialize trace " diff --git a/kernel/trace/trace_export.c b/kernel/trace/trace_export.c index 029a91f..9cbe7f1 100644 --- a/kernel/trace/trace_export.c +++ b/kernel/trace/trace_export.c @@ -117,10 +117,16 @@ ftrace_format_##call(struct ftrace_event_call *unused, \ #define TRACE_FIELD_SPECIAL(type_item, item, len, cmd) \ cmd; +static int ftrace_raw_init_event(struct ftrace_event_call *event_call) +{ + INIT_LIST_HEAD(&event_call->fields); + init_preds(event_call); + return 0; +} + #undef TRACE_EVENT_FORMAT #define TRACE_EVENT_FORMAT(call, proto, args, fmt, tstruct, tpfmt) \ int ftrace_define_fields_##call(struct ftrace_event_call *event_call); \ -static int ftrace_raw_init_event_##call(void); \ \ struct ftrace_event_call __used \ __attribute__((__aligned__(4))) \ @@ -128,16 +134,10 @@ __attribute__((section("_ftrace_events"))) event_##call = { \ .name = #call, \ .id = proto, \ .system = __stringify(TRACE_SYSTEM), \ - .raw_init = ftrace_raw_init_event_##call, \ + .raw_init = ftrace_raw_init_event, \ .show_format = ftrace_format_##call, \ .define_fields = ftrace_define_fields_##call, \ -}; \ -static int ftrace_raw_init_event_##call(void) \ -{ \ - INIT_LIST_HEAD(&event_##call.fields); \ - init_preds(&event_##call); \ - return 0; \ -} \ +}; #undef TRACE_EVENT_FORMAT_NOFILTER #define TRACE_EVENT_FORMAT_NOFILTER(call, proto, args, fmt, tstruct, \ diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index 85291c4..5931933 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -193,8 +193,8 @@ int syscall_enter_define_fields(struct ftrace_event_call *call) return ret; for (i = 0; i < meta->nb_args; i++) { - ret = trace_define_field(call, meta->types[i], - meta->args[i], offset, + ret = trace_define_field(call, (char *)meta->types[i], + (char *)meta->args[i], offset, sizeof(unsigned long), 0, FILTER_OTHER); offset += sizeof(unsigned long); @@ -277,13 +277,13 @@ void ftrace_syscall_exit(struct pt_regs *regs, long ret) trace_current_buffer_unlock_commit(event, 0, 0); } -int reg_event_syscall_enter(void *ptr) +int reg_event_syscall_enter(struct ftrace_event_call *call) { int ret = 0; int num; char *name; - name = (char *)ptr; + name = (char *)call->data; num = syscall_name_to_nr(name); if (num < 0 || num >= FTRACE_SYSCALL_MAX) return -ENOSYS; @@ -301,12 +301,12 @@ int reg_event_syscall_enter(void *ptr) return ret; } -void unreg_event_syscall_enter(void *ptr) +void unreg_event_syscall_enter(struct ftrace_event_call *call) { int num; char *name; - name = (char *)ptr; + name = (char *)call->data; num = syscall_name_to_nr(name); if (num < 0 || num >= FTRACE_SYSCALL_MAX) return; @@ -318,13 +318,13 @@ void unreg_event_syscall_enter(void *ptr) mutex_unlock(&syscall_trace_lock); } -int reg_event_syscall_exit(void *ptr) +int reg_event_syscall_exit(struct ftrace_event_call *call) { int ret = 0; int num; char *name; - name = (char *)ptr; + name = (char *)call->data; num = syscall_name_to_nr(name); if (num < 0 || num >= FTRACE_SYSCALL_MAX) return -ENOSYS; @@ -342,12 +342,12 @@ int reg_event_syscall_exit(void *ptr) return ret; } -void unreg_event_syscall_exit(void *ptr) +void unreg_event_syscall_exit(struct ftrace_event_call *call) { int num; char *name; - name = (char *)ptr; + name = (char *)call->data; num = syscall_name_to_nr(name); if (num < 0 || num >= FTRACE_SYSCALL_MAX) return; -- cgit v1.1 From d93f12f3f417e49a175800da85c6fcb2a5096e03 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 13 Aug 2009 16:35:01 -0400 Subject: tracing: Introduce TRACE_FIELD_ZERO() macro MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use TRACE_FIELD_ZERO(type, item) instead of TRACE_FIELD_ZERO_CHAR(item). This also includes a typo fix of TRACE_ZERO_CHAR() macro. Signed-off-by: Masami Hiramatsu Cc: Ananth N Mavinakayanahalli Cc: Avi Kivity Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: Jim Keniston Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Przemysław Pawełczyk Cc: Roland McGrath Cc: Sam Ravnborg Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi Cc: Vegard Nossum LKML-Reference: <20090813203501.31965.30172.stgit@localhost.localdomain> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_event_types.h | 4 ++-- kernel/trace/trace_export.c | 16 ++++++++-------- 2 files changed, 10 insertions(+), 10 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_event_types.h b/kernel/trace/trace_event_types.h index 6db005e..e74f090 100644 --- a/kernel/trace/trace_event_types.h +++ b/kernel/trace/trace_event_types.h @@ -109,7 +109,7 @@ TRACE_EVENT_FORMAT(bprint, TRACE_BPRINT, bprint_entry, ignore, TRACE_STRUCT( TRACE_FIELD(unsigned long, ip, ip) TRACE_FIELD(char *, fmt, fmt) - TRACE_FIELD_ZERO_CHAR(buf) + TRACE_FIELD_ZERO(char, buf) ), TP_RAW_FMT("%08lx (%d) fmt:%p %s") ); @@ -117,7 +117,7 @@ TRACE_EVENT_FORMAT(bprint, TRACE_BPRINT, bprint_entry, ignore, TRACE_EVENT_FORMAT(print, TRACE_PRINT, print_entry, ignore, TRACE_STRUCT( TRACE_FIELD(unsigned long, ip, ip) - TRACE_FIELD_ZERO_CHAR(buf) + TRACE_FIELD_ZERO(char, buf) ), TP_RAW_FMT("%08lx (%d) fmt:%p %s") ); diff --git a/kernel/trace/trace_export.c b/kernel/trace/trace_export.c index 9cbe7f1..f75faec 100644 --- a/kernel/trace/trace_export.c +++ b/kernel/trace/trace_export.c @@ -42,9 +42,9 @@ extern void __bad_type_size(void); if (!ret) \ return 0; -#undef TRACE_FIELD_ZERO_CHAR -#define TRACE_FIELD_ZERO_CHAR(item) \ - ret = trace_seq_printf(s, "\tfield:char " #item ";\t" \ +#undef TRACE_FIELD_ZERO +#define TRACE_FIELD_ZERO(type, item) \ + ret = trace_seq_printf(s, "\tfield:" #type " " #item ";\t" \ "offset:%u;\tsize:0;\n", \ (unsigned int)offsetof(typeof(field), item)); \ if (!ret) \ @@ -92,9 +92,6 @@ ftrace_format_##call(struct ftrace_event_call *unused, \ #include "trace_event_types.h" -#undef TRACE_ZERO_CHAR -#define TRACE_ZERO_CHAR(arg) - #undef TRACE_FIELD #define TRACE_FIELD(type, item, assign)\ entry->item = assign; @@ -107,6 +104,9 @@ ftrace_format_##call(struct ftrace_event_call *unused, \ #define TRACE_FIELD_SIGN(type, item, assign, is_signed) \ TRACE_FIELD(type, item, assign) +#undef TRACE_FIELD_ZERO +#define TRACE_FIELD_ZERO(type, item) + #undef TP_CMD #define TP_CMD(cmd...) cmd @@ -180,8 +180,8 @@ __attribute__((section("_ftrace_events"))) event_##call = { \ if (ret) \ return ret; -#undef TRACE_FIELD_ZERO_CHAR -#define TRACE_FIELD_ZERO_CHAR(item) +#undef TRACE_FIELD_ZERO +#define TRACE_FIELD_ZERO(type, item) #undef TRACE_EVENT_FORMAT #define TRACE_EVENT_FORMAT(call, proto, args, fmt, tstruct, tpfmt) \ -- cgit v1.1 From 413d37d1eb69c1765b9ace0a612dac9b6c990e66 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 13 Aug 2009 16:35:11 -0400 Subject: tracing: Add kprobe-based event tracer MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add kprobes-based event tracer on ftrace. This tracer is similar to the events tracer which is based on Tracepoint infrastructure. Instead of Tracepoint, this tracer is based on kprobes (kprobe and kretprobe). It probes anywhere where kprobes can probe(this means, all functions body except for __kprobes functions). Similar to the events tracer, this tracer doesn't need to be activated via current_tracer, instead of that, just set probe points via /sys/kernel/debug/tracing/kprobe_events. And you can set filters on each probe events via /sys/kernel/debug/tracing/events/kprobes//filter. This tracer supports following probe arguments for each probe. %REG : Fetch register REG sN : Fetch Nth entry of stack (N >= 0) sa : Fetch stack address. @ADDR : Fetch memory at ADDR (ADDR should be in kernel) @SYM[+|-offs] : Fetch memory at SYM +|- offs (SYM should be a data symbol) aN : Fetch function argument. (N >= 0) rv : Fetch return value. ra : Fetch return address. +|-offs(FETCHARG) : fetch memory at FETCHARG +|- offs address. See Documentation/trace/kprobetrace.txt in the next patch for details. Changes from v13: - Support 'sa' for stack address. - Use call->data instead of container_of() macro. [fweisbec@gmail.com: Fixed conflict against latest tracing/core] Signed-off-by: Masami Hiramatsu Acked-by: Ananth N Mavinakayanahalli Cc: Avi Kivity Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: Jim Keniston Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Przemysław Pawełczyk Cc: Roland McGrath Cc: Sam Ravnborg Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi Cc: Vegard Nossum LKML-Reference: <20090813203510.31965.29123.stgit@localhost.localdomain> Signed-off-by: Frederic Weisbecker --- kernel/trace/Kconfig | 12 + kernel/trace/Makefile | 1 + kernel/trace/trace.h | 30 + kernel/trace/trace_event_types.h | 18 + kernel/trace/trace_kprobe.c | 1202 ++++++++++++++++++++++++++++++++++++++ 5 files changed, 1263 insertions(+) create mode 100644 kernel/trace/trace_kprobe.c (limited to 'kernel') diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig index 06be85a..fb5fbf7 100644 --- a/kernel/trace/Kconfig +++ b/kernel/trace/Kconfig @@ -411,6 +411,18 @@ config BLK_DEV_IO_TRACE If unsure, say N. +config KPROBE_TRACER + depends on KPROBES + depends on X86 + bool "Trace kprobes" + select TRACING + select GENERIC_TRACER + help + This tracer probes everywhere where kprobes can probe it, and + records various registers and memories specified by user. + This also allows you to trace kprobe probe points as a dynamic + defined events. It provides per-probe event filtering interface. + config DYNAMIC_FTRACE bool "enable/disable ftrace tracepoints dynamically" depends on FUNCTION_TRACER diff --git a/kernel/trace/Makefile b/kernel/trace/Makefile index 844164d..7c00a1e 100644 --- a/kernel/trace/Makefile +++ b/kernel/trace/Makefile @@ -54,5 +54,6 @@ obj-$(CONFIG_EVENT_TRACING) += trace_export.o obj-$(CONFIG_FTRACE_SYSCALLS) += trace_syscalls.o obj-$(CONFIG_EVENT_PROFILE) += trace_event_profile.o obj-$(CONFIG_EVENT_TRACING) += trace_events_filter.o +obj-$(CONFIG_KPROBE_TRACER) += trace_kprobe.o libftrace-y := ftrace.o diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 654fd65..667f832 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -38,6 +38,8 @@ enum trace_type { TRACE_KMEM_FREE, TRACE_POWER, TRACE_BLK, + TRACE_KPROBE, + TRACE_KRETPROBE, __TRACE_LAST_TYPE, }; @@ -205,6 +207,30 @@ struct syscall_trace_exit { unsigned long ret; }; +struct kprobe_trace_entry { + struct trace_entry ent; + unsigned long ip; + int nargs; + unsigned long args[]; +}; + +#define SIZEOF_KPROBE_TRACE_ENTRY(n) \ + (offsetof(struct kprobe_trace_entry, args) + \ + (sizeof(unsigned long) * (n))) + +struct kretprobe_trace_entry { + struct trace_entry ent; + unsigned long func; + unsigned long ret_ip; + int nargs; + unsigned long args[]; +}; + +#define SIZEOF_KRETPROBE_TRACE_ENTRY(n) \ + (offsetof(struct kretprobe_trace_entry, args) + \ + (sizeof(unsigned long) * (n))) + + /* * trace_flag_type is an enumeration that holds different @@ -317,6 +343,10 @@ extern void __ftrace_bad_type(void); TRACE_KMEM_ALLOC); \ IF_ASSIGN(var, ent, struct kmemtrace_free_entry, \ TRACE_KMEM_FREE); \ + IF_ASSIGN(var, ent, struct kprobe_trace_entry, \ + TRACE_KPROBE); \ + IF_ASSIGN(var, ent, struct kretprobe_trace_entry, \ + TRACE_KRETPROBE); \ __ftrace_bad_type(); \ } while (0) diff --git a/kernel/trace/trace_event_types.h b/kernel/trace/trace_event_types.h index e74f090..186b598 100644 --- a/kernel/trace/trace_event_types.h +++ b/kernel/trace/trace_event_types.h @@ -175,4 +175,22 @@ TRACE_EVENT_FORMAT(kmem_free, TRACE_KMEM_FREE, kmemtrace_free_entry, ignore, TP_RAW_FMT("type:%u call_site:%lx ptr:%p") ); +TRACE_EVENT_FORMAT(kprobe, TRACE_KPROBE, kprobe_trace_entry, ignore, + TRACE_STRUCT( + TRACE_FIELD(unsigned long, ip, ip) + TRACE_FIELD(int, nargs, nargs) + TRACE_FIELD_ZERO(unsigned long, args) + ), + TP_RAW_FMT("%08lx: args:0x%lx ...") +); + +TRACE_EVENT_FORMAT(kretprobe, TRACE_KRETPROBE, kretprobe_trace_entry, ignore, + TRACE_STRUCT( + TRACE_FIELD(unsigned long, func, func) + TRACE_FIELD(unsigned long, ret_ip, ret_ip) + TRACE_FIELD(int, nargs, nargs) + TRACE_FIELD_ZERO(unsigned long, args) + ), + TP_RAW_FMT("%08lx <- %08lx: args:0x%lx ...") +); #undef TRACE_SYSTEM diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c new file mode 100644 index 0000000..0c4f00a --- /dev/null +++ b/kernel/trace/trace_kprobe.c @@ -0,0 +1,1202 @@ +/* + * kprobe based kernel tracer + * + * Created by Masami Hiramatsu + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "trace.h" +#include "trace_output.h" + +#define TRACE_KPROBE_ARGS 6 +#define MAX_ARGSTR_LEN 63 + +/* currently, trace_kprobe only supports X86. */ + +struct fetch_func { + unsigned long (*func)(struct pt_regs *, void *); + void *data; +}; + +static __kprobes unsigned long call_fetch(struct fetch_func *f, + struct pt_regs *regs) +{ + return f->func(regs, f->data); +} + +/* fetch handlers */ +static __kprobes unsigned long fetch_register(struct pt_regs *regs, + void *offset) +{ + return regs_get_register(regs, (unsigned int)((unsigned long)offset)); +} + +static __kprobes unsigned long fetch_stack(struct pt_regs *regs, + void *num) +{ + return regs_get_kernel_stack_nth(regs, + (unsigned int)((unsigned long)num)); +} + +static __kprobes unsigned long fetch_memory(struct pt_regs *regs, void *addr) +{ + unsigned long retval; + + if (probe_kernel_address(addr, retval)) + return 0; + return retval; +} + +static __kprobes unsigned long fetch_argument(struct pt_regs *regs, void *num) +{ + return regs_get_argument_nth(regs, (unsigned int)((unsigned long)num)); +} + +static __kprobes unsigned long fetch_retvalue(struct pt_regs *regs, + void *dummy) +{ + return regs_return_value(regs); +} + +static __kprobes unsigned long fetch_ip(struct pt_regs *regs, void *dummy) +{ + return instruction_pointer(regs); +} + +static __kprobes unsigned long fetch_stack_address(struct pt_regs *regs, + void *dummy) +{ + return kernel_stack_pointer(regs); +} + +/* Memory fetching by symbol */ +struct symbol_cache { + char *symbol; + long offset; + unsigned long addr; +}; + +static unsigned long update_symbol_cache(struct symbol_cache *sc) +{ + sc->addr = (unsigned long)kallsyms_lookup_name(sc->symbol); + if (sc->addr) + sc->addr += sc->offset; + return sc->addr; +} + +static void free_symbol_cache(struct symbol_cache *sc) +{ + kfree(sc->symbol); + kfree(sc); +} + +static struct symbol_cache *alloc_symbol_cache(const char *sym, long offset) +{ + struct symbol_cache *sc; + + if (!sym || strlen(sym) == 0) + return NULL; + sc = kzalloc(sizeof(struct symbol_cache), GFP_KERNEL); + if (!sc) + return NULL; + + sc->symbol = kstrdup(sym, GFP_KERNEL); + if (!sc->symbol) { + kfree(sc); + return NULL; + } + sc->offset = offset; + + update_symbol_cache(sc); + return sc; +} + +static __kprobes unsigned long fetch_symbol(struct pt_regs *regs, void *data) +{ + struct symbol_cache *sc = data; + + if (sc->addr) + return fetch_memory(regs, (void *)sc->addr); + else + return 0; +} + +/* Special indirect memory access interface */ +struct indirect_fetch_data { + struct fetch_func orig; + long offset; +}; + +static __kprobes unsigned long fetch_indirect(struct pt_regs *regs, void *data) +{ + struct indirect_fetch_data *ind = data; + unsigned long addr; + + addr = call_fetch(&ind->orig, regs); + if (addr) { + addr += ind->offset; + return fetch_memory(regs, (void *)addr); + } else + return 0; +} + +static __kprobes void free_indirect_fetch_data(struct indirect_fetch_data *data) +{ + if (data->orig.func == fetch_indirect) + free_indirect_fetch_data(data->orig.data); + else if (data->orig.func == fetch_symbol) + free_symbol_cache(data->orig.data); + kfree(data); +} + +/** + * kprobe_trace_core + */ + +struct trace_probe { + struct list_head list; + union { + struct kprobe kp; + struct kretprobe rp; + }; + const char *symbol; /* symbol name */ + unsigned int nr_args; + struct fetch_func args[TRACE_KPROBE_ARGS]; + struct ftrace_event_call call; +}; + +static int kprobe_trace_func(struct kprobe *kp, struct pt_regs *regs); +static int kretprobe_trace_func(struct kretprobe_instance *ri, + struct pt_regs *regs); + +static __kprobes int probe_is_return(struct trace_probe *tp) +{ + return (tp->rp.handler == kretprobe_trace_func); +} + +static __kprobes const char *probe_symbol(struct trace_probe *tp) +{ + return tp->symbol ? tp->symbol : "unknown"; +} + +static __kprobes long probe_offset(struct trace_probe *tp) +{ + return (probe_is_return(tp)) ? tp->rp.kp.offset : tp->kp.offset; +} + +static __kprobes void *probe_address(struct trace_probe *tp) +{ + return (probe_is_return(tp)) ? tp->rp.kp.addr : tp->kp.addr; +} + +static int trace_arg_string(char *buf, size_t n, struct fetch_func *ff) +{ + int ret = -EINVAL; + + if (ff->func == fetch_argument) + ret = snprintf(buf, n, "a%lu", (unsigned long)ff->data); + else if (ff->func == fetch_register) { + const char *name; + name = regs_query_register_name((unsigned int)((long)ff->data)); + ret = snprintf(buf, n, "%%%s", name); + } else if (ff->func == fetch_stack) + ret = snprintf(buf, n, "s%lu", (unsigned long)ff->data); + else if (ff->func == fetch_memory) + ret = snprintf(buf, n, "@0x%p", ff->data); + else if (ff->func == fetch_symbol) { + struct symbol_cache *sc = ff->data; + ret = snprintf(buf, n, "@%s%+ld", sc->symbol, sc->offset); + } else if (ff->func == fetch_retvalue) + ret = snprintf(buf, n, "rv"); + else if (ff->func == fetch_ip) + ret = snprintf(buf, n, "ra"); + else if (ff->func == fetch_stack_address) + ret = snprintf(buf, n, "sa"); + else if (ff->func == fetch_indirect) { + struct indirect_fetch_data *id = ff->data; + size_t l = 0; + ret = snprintf(buf, n, "%+ld(", id->offset); + if (ret >= n) + goto end; + l += ret; + ret = trace_arg_string(buf + l, n - l, &id->orig); + if (ret < 0) + goto end; + l += ret; + ret = snprintf(buf + l, n - l, ")"); + ret += l; + } +end: + if (ret >= n) + return -ENOSPC; + return ret; +} + +static int register_probe_event(struct trace_probe *tp); +static void unregister_probe_event(struct trace_probe *tp); + +static DEFINE_MUTEX(probe_lock); +static LIST_HEAD(probe_list); + +static struct trace_probe *alloc_trace_probe(const char *symbol, + const char *event) +{ + struct trace_probe *tp; + + tp = kzalloc(sizeof(struct trace_probe), GFP_KERNEL); + if (!tp) + return ERR_PTR(-ENOMEM); + + if (symbol) { + tp->symbol = kstrdup(symbol, GFP_KERNEL); + if (!tp->symbol) + goto error; + } + if (event) { + tp->call.name = kstrdup(event, GFP_KERNEL); + if (!tp->call.name) + goto error; + } + + INIT_LIST_HEAD(&tp->list); + return tp; +error: + kfree(tp->symbol); + kfree(tp); + return ERR_PTR(-ENOMEM); +} + +static void free_trace_probe(struct trace_probe *tp) +{ + int i; + + for (i = 0; i < tp->nr_args; i++) + if (tp->args[i].func == fetch_symbol) + free_symbol_cache(tp->args[i].data); + else if (tp->args[i].func == fetch_indirect) + free_indirect_fetch_data(tp->args[i].data); + + kfree(tp->call.name); + kfree(tp->symbol); + kfree(tp); +} + +static struct trace_probe *find_probe_event(const char *event) +{ + struct trace_probe *tp; + + list_for_each_entry(tp, &probe_list, list) + if (tp->call.name && !strcmp(tp->call.name, event)) + return tp; + return NULL; +} + +static void __unregister_trace_probe(struct trace_probe *tp) +{ + if (probe_is_return(tp)) + unregister_kretprobe(&tp->rp); + else + unregister_kprobe(&tp->kp); +} + +/* Unregister a trace_probe and probe_event: call with locking probe_lock */ +static void unregister_trace_probe(struct trace_probe *tp) +{ + if (tp->call.name) + unregister_probe_event(tp); + __unregister_trace_probe(tp); + list_del(&tp->list); +} + +/* Register a trace_probe and probe_event */ +static int register_trace_probe(struct trace_probe *tp) +{ + struct trace_probe *old_tp; + int ret; + + mutex_lock(&probe_lock); + + if (probe_is_return(tp)) + ret = register_kretprobe(&tp->rp); + else + ret = register_kprobe(&tp->kp); + + if (ret) { + pr_warning("Could not insert probe(%d)\n", ret); + if (ret == -EILSEQ) { + pr_warning("Probing address(0x%p) is not an " + "instruction boundary.\n", + probe_address(tp)); + ret = -EINVAL; + } + goto end; + } + /* register as an event */ + if (tp->call.name) { + old_tp = find_probe_event(tp->call.name); + if (old_tp) { + /* delete old event */ + unregister_trace_probe(old_tp); + free_trace_probe(old_tp); + } + ret = register_probe_event(tp); + if (ret) { + pr_warning("Faild to register probe event(%d)\n", ret); + __unregister_trace_probe(tp); + } + } + list_add_tail(&tp->list, &probe_list); +end: + mutex_unlock(&probe_lock); + return ret; +} + +/* Split symbol and offset. */ +static int split_symbol_offset(char *symbol, long *offset) +{ + char *tmp; + int ret; + + if (!offset) + return -EINVAL; + + tmp = strchr(symbol, '+'); + if (!tmp) + tmp = strchr(symbol, '-'); + + if (tmp) { + /* skip sign because strict_strtol doesn't accept '+' */ + ret = strict_strtol(tmp + 1, 0, offset); + if (ret) + return ret; + if (*tmp == '-') + *offset = -(*offset); + *tmp = '\0'; + } else + *offset = 0; + return 0; +} + +#define PARAM_MAX_ARGS 16 +#define PARAM_MAX_STACK (THREAD_SIZE / sizeof(unsigned long)) + +static int parse_trace_arg(char *arg, struct fetch_func *ff, int is_return) +{ + int ret = 0; + unsigned long param; + long offset; + char *tmp; + + switch (arg[0]) { + case 'a': /* argument */ + ret = strict_strtoul(arg + 1, 10, ¶m); + if (ret || param > PARAM_MAX_ARGS) + ret = -EINVAL; + else { + ff->func = fetch_argument; + ff->data = (void *)param; + } + break; + case 'r': /* retval or retaddr */ + if (is_return && arg[1] == 'v') { + ff->func = fetch_retvalue; + ff->data = NULL; + } else if (is_return && arg[1] == 'a') { + ff->func = fetch_ip; + ff->data = NULL; + } else + ret = -EINVAL; + break; + case '%': /* named register */ + ret = regs_query_register_offset(arg + 1); + if (ret >= 0) { + ff->func = fetch_register; + ff->data = (void *)(unsigned long)ret; + ret = 0; + } + break; + case 's': /* stack */ + if (arg[1] == 'a') { + ff->func = fetch_stack_address; + ff->data = NULL; + } else { + ret = strict_strtoul(arg + 1, 10, ¶m); + if (ret || param > PARAM_MAX_STACK) + ret = -EINVAL; + else { + ff->func = fetch_stack; + ff->data = (void *)param; + } + } + break; + case '@': /* memory or symbol */ + if (isdigit(arg[1])) { + ret = strict_strtoul(arg + 1, 0, ¶m); + if (ret) + break; + ff->func = fetch_memory; + ff->data = (void *)param; + } else { + ret = split_symbol_offset(arg + 1, &offset); + if (ret) + break; + ff->data = alloc_symbol_cache(arg + 1, + offset); + if (ff->data) + ff->func = fetch_symbol; + else + ret = -EINVAL; + } + break; + case '+': /* indirect memory */ + case '-': + tmp = strchr(arg, '('); + if (!tmp) { + ret = -EINVAL; + break; + } + *tmp = '\0'; + ret = strict_strtol(arg + 1, 0, &offset); + if (ret) + break; + if (arg[0] == '-') + offset = -offset; + arg = tmp + 1; + tmp = strrchr(arg, ')'); + if (tmp) { + struct indirect_fetch_data *id; + *tmp = '\0'; + id = kzalloc(sizeof(struct indirect_fetch_data), + GFP_KERNEL); + if (!id) + return -ENOMEM; + id->offset = offset; + ret = parse_trace_arg(arg, &id->orig, is_return); + if (ret) + kfree(id); + else { + ff->func = fetch_indirect; + ff->data = (void *)id; + } + } else + ret = -EINVAL; + break; + default: + /* TODO: support custom handler */ + ret = -EINVAL; + } + return ret; +} + +static int create_trace_probe(int argc, char **argv) +{ + /* + * Argument syntax: + * - Add kprobe: p[:EVENT] SYMBOL[+OFFS|-OFFS]|ADDRESS [FETCHARGS] + * - Add kretprobe: r[:EVENT] SYMBOL[+0] [FETCHARGS] + * Fetch args: + * aN : fetch Nth of function argument. (N:0-) + * rv : fetch return value + * ra : fetch return address + * sa : fetch stack address + * sN : fetch Nth of stack (N:0-) + * @ADDR : fetch memory at ADDR (ADDR should be in kernel) + * @SYM[+|-offs] : fetch memory at SYM +|- offs (SYM is a data symbol) + * %REG : fetch register REG + * Indirect memory fetch: + * +|-offs(ARG) : fetch memory at ARG +|- offs address. + */ + struct trace_probe *tp; + struct kprobe *kp; + int i, ret = 0; + int is_return = 0; + char *symbol = NULL, *event = NULL; + long offset = 0; + void *addr = NULL; + + if (argc < 2) + return -EINVAL; + + if (argv[0][0] == 'p') + is_return = 0; + else if (argv[0][0] == 'r') + is_return = 1; + else + return -EINVAL; + + if (argv[0][1] == ':') { + event = &argv[0][2]; + if (strlen(event) == 0) { + pr_info("Event name is not specifiled\n"); + return -EINVAL; + } + } + + if (isdigit(argv[1][0])) { + if (is_return) + return -EINVAL; + /* an address specified */ + ret = strict_strtoul(&argv[0][2], 0, (unsigned long *)&addr); + if (ret) + return ret; + } else { + /* a symbol specified */ + symbol = argv[1]; + /* TODO: support .init module functions */ + ret = split_symbol_offset(symbol, &offset); + if (ret) + return ret; + if (offset && is_return) + return -EINVAL; + } + + /* setup a probe */ + tp = alloc_trace_probe(symbol, event); + if (IS_ERR(tp)) + return PTR_ERR(tp); + + if (is_return) { + kp = &tp->rp.kp; + tp->rp.handler = kretprobe_trace_func; + } else { + kp = &tp->kp; + tp->kp.pre_handler = kprobe_trace_func; + } + + if (tp->symbol) { + kp->symbol_name = tp->symbol; + kp->offset = offset; + } else + kp->addr = addr; + + /* parse arguments */ + argc -= 2; argv += 2; ret = 0; + for (i = 0; i < argc && i < TRACE_KPROBE_ARGS; i++) { + if (strlen(argv[i]) > MAX_ARGSTR_LEN) { + pr_info("Argument%d(%s) is too long.\n", i, argv[i]); + ret = -ENOSPC; + goto error; + } + ret = parse_trace_arg(argv[i], &tp->args[i], is_return); + if (ret) + goto error; + } + tp->nr_args = i; + + ret = register_trace_probe(tp); + if (ret) + goto error; + return 0; + +error: + free_trace_probe(tp); + return ret; +} + +static void cleanup_all_probes(void) +{ + struct trace_probe *tp; + + mutex_lock(&probe_lock); + /* TODO: Use batch unregistration */ + while (!list_empty(&probe_list)) { + tp = list_entry(probe_list.next, struct trace_probe, list); + unregister_trace_probe(tp); + free_trace_probe(tp); + } + mutex_unlock(&probe_lock); +} + + +/* Probes listing interfaces */ +static void *probes_seq_start(struct seq_file *m, loff_t *pos) +{ + mutex_lock(&probe_lock); + return seq_list_start(&probe_list, *pos); +} + +static void *probes_seq_next(struct seq_file *m, void *v, loff_t *pos) +{ + return seq_list_next(v, &probe_list, pos); +} + +static void probes_seq_stop(struct seq_file *m, void *v) +{ + mutex_unlock(&probe_lock); +} + +static int probes_seq_show(struct seq_file *m, void *v) +{ + struct trace_probe *tp = v; + int i, ret; + char buf[MAX_ARGSTR_LEN + 1]; + + seq_printf(m, "%c", probe_is_return(tp) ? 'r' : 'p'); + if (tp->call.name) + seq_printf(m, ":%s", tp->call.name); + + if (tp->symbol) + seq_printf(m, " %s%+ld", probe_symbol(tp), probe_offset(tp)); + else + seq_printf(m, " 0x%p", probe_address(tp)); + + for (i = 0; i < tp->nr_args; i++) { + ret = trace_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); + if (ret < 0) { + pr_warning("Argument%d decoding error(%d).\n", i, ret); + return ret; + } + seq_printf(m, " %s", buf); + } + seq_printf(m, "\n"); + return 0; +} + +static const struct seq_operations probes_seq_op = { + .start = probes_seq_start, + .next = probes_seq_next, + .stop = probes_seq_stop, + .show = probes_seq_show +}; + +static int probes_open(struct inode *inode, struct file *file) +{ + if ((file->f_mode & FMODE_WRITE) && + (file->f_flags & O_TRUNC)) + cleanup_all_probes(); + + return seq_open(file, &probes_seq_op); +} + +static int command_trace_probe(const char *buf) +{ + char **argv; + int argc = 0, ret = 0; + + argv = argv_split(GFP_KERNEL, buf, &argc); + if (!argv) + return -ENOMEM; + + if (argc) + ret = create_trace_probe(argc, argv); + + argv_free(argv); + return ret; +} + +#define WRITE_BUFSIZE 128 + +static ssize_t probes_write(struct file *file, const char __user *buffer, + size_t count, loff_t *ppos) +{ + char *kbuf, *tmp; + int ret; + size_t done; + size_t size; + + kbuf = kmalloc(WRITE_BUFSIZE, GFP_KERNEL); + if (!kbuf) + return -ENOMEM; + + ret = done = 0; + while (done < count) { + size = count - done; + if (size >= WRITE_BUFSIZE) + size = WRITE_BUFSIZE - 1; + if (copy_from_user(kbuf, buffer + done, size)) { + ret = -EFAULT; + goto out; + } + kbuf[size] = '\0'; + tmp = strchr(kbuf, '\n'); + if (tmp) { + *tmp = '\0'; + size = tmp - kbuf + 1; + } else if (done + size < count) { + pr_warning("Line length is too long: " + "Should be less than %d.", WRITE_BUFSIZE); + ret = -EINVAL; + goto out; + } + done += size; + /* Remove comments */ + tmp = strchr(kbuf, '#'); + if (tmp) + *tmp = '\0'; + + ret = command_trace_probe(kbuf); + if (ret) + goto out; + } + ret = done; +out: + kfree(kbuf); + return ret; +} + +static const struct file_operations kprobe_events_ops = { + .owner = THIS_MODULE, + .open = probes_open, + .read = seq_read, + .llseek = seq_lseek, + .release = seq_release, + .write = probes_write, +}; + +/* Kprobe handler */ +static __kprobes int kprobe_trace_func(struct kprobe *kp, struct pt_regs *regs) +{ + struct trace_probe *tp = container_of(kp, struct trace_probe, kp); + struct kprobe_trace_entry *entry; + struct ring_buffer_event *event; + int size, i, pc; + unsigned long irq_flags; + struct ftrace_event_call *call = &event_kprobe; + + if (&tp->call.name) + call = &tp->call; + + local_save_flags(irq_flags); + pc = preempt_count(); + + size = SIZEOF_KPROBE_TRACE_ENTRY(tp->nr_args); + + event = trace_current_buffer_lock_reserve(TRACE_KPROBE, size, + irq_flags, pc); + if (!event) + return 0; + + entry = ring_buffer_event_data(event); + entry->nargs = tp->nr_args; + entry->ip = (unsigned long)kp->addr; + for (i = 0; i < tp->nr_args; i++) + entry->args[i] = call_fetch(&tp->args[i], regs); + + if (!filter_current_check_discard(call, entry, event)) + trace_nowake_buffer_unlock_commit(event, irq_flags, pc); + return 0; +} + +/* Kretprobe handler */ +static __kprobes int kretprobe_trace_func(struct kretprobe_instance *ri, + struct pt_regs *regs) +{ + struct trace_probe *tp = container_of(ri->rp, struct trace_probe, rp); + struct kretprobe_trace_entry *entry; + struct ring_buffer_event *event; + int size, i, pc; + unsigned long irq_flags; + struct ftrace_event_call *call = &event_kretprobe; + + if (&tp->call.name) + call = &tp->call; + + local_save_flags(irq_flags); + pc = preempt_count(); + + size = SIZEOF_KRETPROBE_TRACE_ENTRY(tp->nr_args); + + event = trace_current_buffer_lock_reserve(TRACE_KRETPROBE, size, + irq_flags, pc); + if (!event) + return 0; + + entry = ring_buffer_event_data(event); + entry->nargs = tp->nr_args; + entry->func = (unsigned long)probe_address(tp); + entry->ret_ip = (unsigned long)ri->ret_addr; + for (i = 0; i < tp->nr_args; i++) + entry->args[i] = call_fetch(&tp->args[i], regs); + + if (!filter_current_check_discard(call, entry, event)) + trace_nowake_buffer_unlock_commit(event, irq_flags, pc); + + return 0; +} + +/* Event entry printers */ +enum print_line_t +print_kprobe_event(struct trace_iterator *iter, int flags) +{ + struct kprobe_trace_entry *field; + struct trace_seq *s = &iter->seq; + int i; + + trace_assign_type(field, iter->ent); + + if (!seq_print_ip_sym(s, field->ip, flags | TRACE_ITER_SYM_OFFSET)) + goto partial; + + if (!trace_seq_puts(s, ":")) + goto partial; + + for (i = 0; i < field->nargs; i++) + if (!trace_seq_printf(s, " 0x%lx", field->args[i])) + goto partial; + + if (!trace_seq_puts(s, "\n")) + goto partial; + + return TRACE_TYPE_HANDLED; +partial: + return TRACE_TYPE_PARTIAL_LINE; +} + +enum print_line_t +print_kretprobe_event(struct trace_iterator *iter, int flags) +{ + struct kretprobe_trace_entry *field; + struct trace_seq *s = &iter->seq; + int i; + + trace_assign_type(field, iter->ent); + + if (!seq_print_ip_sym(s, field->ret_ip, flags | TRACE_ITER_SYM_OFFSET)) + goto partial; + + if (!trace_seq_puts(s, " <- ")) + goto partial; + + if (!seq_print_ip_sym(s, field->func, flags & ~TRACE_ITER_SYM_OFFSET)) + goto partial; + + if (!trace_seq_puts(s, ":")) + goto partial; + + for (i = 0; i < field->nargs; i++) + if (!trace_seq_printf(s, " 0x%lx", field->args[i])) + goto partial; + + if (!trace_seq_puts(s, "\n")) + goto partial; + + return TRACE_TYPE_HANDLED; +partial: + return TRACE_TYPE_PARTIAL_LINE; +} + +static struct trace_event kprobe_trace_event = { + .type = TRACE_KPROBE, + .trace = print_kprobe_event, +}; + +static struct trace_event kretprobe_trace_event = { + .type = TRACE_KRETPROBE, + .trace = print_kretprobe_event, +}; + +static int probe_event_enable(struct ftrace_event_call *call) +{ + struct trace_probe *tp = (struct trace_probe *)call->data; + + if (probe_is_return(tp)) + return enable_kretprobe(&tp->rp); + else + return enable_kprobe(&tp->kp); +} + +static void probe_event_disable(struct ftrace_event_call *call) +{ + struct trace_probe *tp = (struct trace_probe *)call->data; + + if (probe_is_return(tp)) + disable_kretprobe(&tp->rp); + else + disable_kprobe(&tp->kp); +} + +static int probe_event_raw_init(struct ftrace_event_call *event_call) +{ + INIT_LIST_HEAD(&event_call->fields); + init_preds(event_call); + return 0; +} + +#undef DEFINE_FIELD +#define DEFINE_FIELD(type, item, name, is_signed) \ + do { \ + ret = trace_define_field(event_call, #type, name, \ + offsetof(typeof(field), item), \ + sizeof(field.item), is_signed, \ + FILTER_OTHER); \ + if (ret) \ + return ret; \ + } while (0) + +static int kprobe_event_define_fields(struct ftrace_event_call *event_call) +{ + int ret, i; + struct kprobe_trace_entry field; + char buf[MAX_ARGSTR_LEN + 1]; + struct trace_probe *tp = (struct trace_probe *)event_call->data; + + ret = trace_define_common_fields(event_call); + if (!ret) + return ret; + + DEFINE_FIELD(unsigned long, ip, "ip", 0); + DEFINE_FIELD(int, nargs, "nargs", 1); + for (i = 0; i < tp->nr_args; i++) { + /* Set argN as a field */ + sprintf(buf, "arg%d", i); + DEFINE_FIELD(unsigned long, args[i], buf, 0); + /* Set argument string as an alias field */ + ret = trace_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); + if (ret < 0) + return ret; + DEFINE_FIELD(unsigned long, args[i], buf, 0); + } + return 0; +} + +static int kretprobe_event_define_fields(struct ftrace_event_call *event_call) +{ + int ret, i; + struct kretprobe_trace_entry field; + char buf[MAX_ARGSTR_LEN + 1]; + struct trace_probe *tp = (struct trace_probe *)event_call->data; + + ret = trace_define_common_fields(event_call); + if (!ret) + return ret; + + DEFINE_FIELD(unsigned long, func, "func", 0); + DEFINE_FIELD(unsigned long, ret_ip, "ret_ip", 0); + DEFINE_FIELD(int, nargs, "nargs", 1); + for (i = 0; i < tp->nr_args; i++) { + /* Set argN as a field */ + sprintf(buf, "arg%d", i); + DEFINE_FIELD(unsigned long, args[i], buf, 0); + /* Set argument string as an alias field */ + ret = trace_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); + if (ret < 0) + return ret; + DEFINE_FIELD(unsigned long, args[i], buf, 0); + } + return 0; +} + +static int __probe_event_show_format(struct trace_seq *s, + struct trace_probe *tp, const char *fmt, + const char *arg) +{ + int i, ret; + char buf[MAX_ARGSTR_LEN + 1]; + + /* Show aliases */ + for (i = 0; i < tp->nr_args; i++) { + ret = trace_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); + if (ret < 0) + return ret; + if (!trace_seq_printf(s, "\talias: %s;\toriginal: arg%d;\n", + buf, i)) + return 0; + } + /* Show format */ + if (!trace_seq_printf(s, "\nprint fmt: \"%s", fmt)) + return 0; + + for (i = 0; i < tp->nr_args; i++) + if (!trace_seq_puts(s, " 0x%lx")) + return 0; + + if (!trace_seq_printf(s, "\", %s", arg)) + return 0; + + for (i = 0; i < tp->nr_args; i++) + if (!trace_seq_printf(s, ", arg%d", i)) + return 0; + + return trace_seq_puts(s, "\n"); +} + +#undef SHOW_FIELD +#define SHOW_FIELD(type, item, name) \ + do { \ + ret = trace_seq_printf(s, "\tfield: " #type " %s;\t" \ + "offset:%u;tsize:%u;\n", name, \ + (unsigned int)offsetof(typeof(field), item),\ + (unsigned int)sizeof(type)); \ + if (!ret) \ + return 0; \ + } while (0) + +static int kprobe_event_show_format(struct ftrace_event_call *call, + struct trace_seq *s) +{ + struct kprobe_trace_entry field __attribute__((unused)); + int ret, i; + char buf[8]; + struct trace_probe *tp = (struct trace_probe *)call->data; + + SHOW_FIELD(unsigned long, ip, "ip"); + SHOW_FIELD(int, nargs, "nargs"); + + /* Show fields */ + for (i = 0; i < tp->nr_args; i++) { + sprintf(buf, "arg%d", i); + SHOW_FIELD(unsigned long, args[i], buf); + } + trace_seq_puts(s, "\n"); + + return __probe_event_show_format(s, tp, "%lx:", "ip"); +} + +static int kretprobe_event_show_format(struct ftrace_event_call *call, + struct trace_seq *s) +{ + struct kretprobe_trace_entry field __attribute__((unused)); + int ret, i; + char buf[8]; + struct trace_probe *tp = (struct trace_probe *)call->data; + + SHOW_FIELD(unsigned long, func, "func"); + SHOW_FIELD(unsigned long, ret_ip, "ret_ip"); + SHOW_FIELD(int, nargs, "nargs"); + + /* Show fields */ + for (i = 0; i < tp->nr_args; i++) { + sprintf(buf, "arg%d", i); + SHOW_FIELD(unsigned long, args[i], buf); + } + trace_seq_puts(s, "\n"); + + return __probe_event_show_format(s, tp, "%lx <- %lx:", + "func, ret_ip"); +} + +static int register_probe_event(struct trace_probe *tp) +{ + struct ftrace_event_call *call = &tp->call; + int ret; + + /* Initialize ftrace_event_call */ + call->system = "kprobes"; + if (probe_is_return(tp)) { + call->event = &kretprobe_trace_event; + call->id = TRACE_KRETPROBE; + call->raw_init = probe_event_raw_init; + call->show_format = kretprobe_event_show_format; + call->define_fields = kretprobe_event_define_fields; + } else { + call->event = &kprobe_trace_event; + call->id = TRACE_KPROBE; + call->raw_init = probe_event_raw_init; + call->show_format = kprobe_event_show_format; + call->define_fields = kprobe_event_define_fields; + } + call->enabled = 1; + call->regfunc = probe_event_enable; + call->unregfunc = probe_event_disable; + call->data = tp; + ret = trace_add_event_call(call); + if (ret) + pr_info("Failed to register kprobe event: %s\n", call->name); + return ret; +} + +static void unregister_probe_event(struct trace_probe *tp) +{ + /* + * Prevent to unregister event itself because the event is shared + * among other probes. + */ + tp->call.event = NULL; + trace_remove_event_call(&tp->call); +} + +/* Make a debugfs interface for controling probe points */ +static __init int init_kprobe_trace(void) +{ + struct dentry *d_tracer; + struct dentry *entry; + int ret; + + ret = register_ftrace_event(&kprobe_trace_event); + if (!ret) { + pr_warning("Could not register kprobe_trace_event type.\n"); + return 0; + } + ret = register_ftrace_event(&kretprobe_trace_event); + if (!ret) { + pr_warning("Could not register kretprobe_trace_event type.\n"); + return 0; + } + + d_tracer = tracing_init_dentry(); + if (!d_tracer) + return 0; + + entry = debugfs_create_file("kprobe_events", 0644, d_tracer, + NULL, &kprobe_events_ops); + + if (!entry) + pr_warning("Could not create debugfs " + "'kprobe_events' entry\n"); + return 0; +} +fs_initcall(init_kprobe_trace); + + +#ifdef CONFIG_FTRACE_STARTUP_TEST + +static int kprobe_trace_selftest_target(int a1, int a2, int a3, + int a4, int a5, int a6) +{ + return a1 + a2 + a3 + a4 + a5 + a6; +} + +static __init int kprobe_trace_self_tests_init(void) +{ + int ret; + int (*target)(int, int, int, int, int, int); + + target = kprobe_trace_selftest_target; + + pr_info("Testing kprobe tracing: "); + + ret = command_trace_probe("p:testprobe kprobe_trace_selftest_target " + "a1 a2 a3 a4 a5 a6"); + if (WARN_ON_ONCE(ret)) + pr_warning("error enabling function entry\n"); + + ret = command_trace_probe("r:testprobe2 kprobe_trace_selftest_target " + "ra rv"); + if (WARN_ON_ONCE(ret)) + pr_warning("error enabling function return\n"); + + ret = target(1, 2, 3, 4, 5, 6); + + cleanup_all_probes(); + + pr_cont("OK\n"); + return 0; +} + +late_initcall(kprobe_trace_self_tests_init); + +#endif -- cgit v1.1 From a82378d8802717b9776a7d9b54422f65c414d6cc Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 13 Aug 2009 16:35:18 -0400 Subject: tracing: Kprobe-tracer supports more than 6 arguments MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Support up to 128 arguments to fetch for each kprobes event. Signed-off-by: Masami Hiramatsu Cc: Ananth N Mavinakayanahalli Cc: Avi Kivity Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: Jim Keniston Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Przemysław Pawełczyk Cc: Roland McGrath Cc: Sam Ravnborg Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi Cc: Vegard Nossum LKML-Reference: <20090813203518.31965.96979.stgit@localhost.localdomain> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 21 +++++++++++++-------- 1 file changed, 13 insertions(+), 8 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 0c4f00a..6d488ef 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -32,7 +32,7 @@ #include "trace.h" #include "trace_output.h" -#define TRACE_KPROBE_ARGS 6 +#define MAX_TRACE_ARGS 128 #define MAX_ARGSTR_LEN 63 /* currently, trace_kprobe only supports X86. */ @@ -184,11 +184,15 @@ struct trace_probe { struct kretprobe rp; }; const char *symbol; /* symbol name */ - unsigned int nr_args; - struct fetch_func args[TRACE_KPROBE_ARGS]; struct ftrace_event_call call; + unsigned int nr_args; + struct fetch_func args[]; }; +#define SIZEOF_TRACE_PROBE(n) \ + (offsetof(struct trace_probe, args) + \ + (sizeof(struct fetch_func) * (n))) + static int kprobe_trace_func(struct kprobe *kp, struct pt_regs *regs); static int kretprobe_trace_func(struct kretprobe_instance *ri, struct pt_regs *regs); @@ -263,11 +267,11 @@ static DEFINE_MUTEX(probe_lock); static LIST_HEAD(probe_list); static struct trace_probe *alloc_trace_probe(const char *symbol, - const char *event) + const char *event, int nargs) { struct trace_probe *tp; - tp = kzalloc(sizeof(struct trace_probe), GFP_KERNEL); + tp = kzalloc(SIZEOF_TRACE_PROBE(nargs), GFP_KERNEL); if (!tp) return ERR_PTR(-ENOMEM); @@ -573,9 +577,10 @@ static int create_trace_probe(int argc, char **argv) if (offset && is_return) return -EINVAL; } + argc -= 2; argv += 2; /* setup a probe */ - tp = alloc_trace_probe(symbol, event); + tp = alloc_trace_probe(symbol, event, argc); if (IS_ERR(tp)) return PTR_ERR(tp); @@ -594,8 +599,8 @@ static int create_trace_probe(int argc, char **argv) kp->addr = addr; /* parse arguments */ - argc -= 2; argv += 2; ret = 0; - for (i = 0; i < argc && i < TRACE_KPROBE_ARGS; i++) { + ret = 0; + for (i = 0; i < argc && i < MAX_TRACE_ARGS; i++) { if (strlen(argv[i]) > MAX_ARGSTR_LEN) { pr_info("Argument%d(%s) is too long.\n", i, argv[i]); ret = -ENOSPC; -- cgit v1.1 From 4263565d491145b57621a761714f2ca6f1293a45 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 13 Aug 2009 16:35:26 -0400 Subject: tracing: Generate names for each kprobe event automatically MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Generate names for each kprobe event based on the probe point. (SYMBOL+offs or MEMADDR). Also remove generic k*probe event types because there is no user of those types. Signed-off-by: Masami Hiramatsu Cc: Ananth N Mavinakayanahalli Cc: Avi Kivity Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: Jim Keniston Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Przemysław Pawełczyk Cc: Roland McGrath Cc: Sam Ravnborg Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi Cc: Vegard Nossum LKML-Reference: <20090813203526.31965.56672.stgit@localhost.localdomain> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_event_types.h | 18 ----------- kernel/trace/trace_kprobe.c | 64 +++++++++++++++++++++------------------- 2 files changed, 33 insertions(+), 49 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_event_types.h b/kernel/trace/trace_event_types.h index 186b598..e74f090 100644 --- a/kernel/trace/trace_event_types.h +++ b/kernel/trace/trace_event_types.h @@ -175,22 +175,4 @@ TRACE_EVENT_FORMAT(kmem_free, TRACE_KMEM_FREE, kmemtrace_free_entry, ignore, TP_RAW_FMT("type:%u call_site:%lx ptr:%p") ); -TRACE_EVENT_FORMAT(kprobe, TRACE_KPROBE, kprobe_trace_entry, ignore, - TRACE_STRUCT( - TRACE_FIELD(unsigned long, ip, ip) - TRACE_FIELD(int, nargs, nargs) - TRACE_FIELD_ZERO(unsigned long, args) - ), - TP_RAW_FMT("%08lx: args:0x%lx ...") -); - -TRACE_EVENT_FORMAT(kretprobe, TRACE_KRETPROBE, kretprobe_trace_entry, ignore, - TRACE_STRUCT( - TRACE_FIELD(unsigned long, func, func) - TRACE_FIELD(unsigned long, ret_ip, ret_ip) - TRACE_FIELD(int, nargs, nargs) - TRACE_FIELD_ZERO(unsigned long, args) - ), - TP_RAW_FMT("%08lx <- %08lx: args:0x%lx ...") -); #undef TRACE_SYSTEM diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 6d488ef..8aeb24c 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -34,6 +34,7 @@ #define MAX_TRACE_ARGS 128 #define MAX_ARGSTR_LEN 63 +#define MAX_EVENT_NAME_LEN 64 /* currently, trace_kprobe only supports X86. */ @@ -280,11 +281,11 @@ static struct trace_probe *alloc_trace_probe(const char *symbol, if (!tp->symbol) goto error; } - if (event) { - tp->call.name = kstrdup(event, GFP_KERNEL); - if (!tp->call.name) - goto error; - } + if (!event) + goto error; + tp->call.name = kstrdup(event, GFP_KERNEL); + if (!tp->call.name) + goto error; INIT_LIST_HEAD(&tp->list); return tp; @@ -314,7 +315,7 @@ static struct trace_probe *find_probe_event(const char *event) struct trace_probe *tp; list_for_each_entry(tp, &probe_list, list) - if (tp->call.name && !strcmp(tp->call.name, event)) + if (!strcmp(tp->call.name, event)) return tp; return NULL; } @@ -330,8 +331,7 @@ static void __unregister_trace_probe(struct trace_probe *tp) /* Unregister a trace_probe and probe_event: call with locking probe_lock */ static void unregister_trace_probe(struct trace_probe *tp) { - if (tp->call.name) - unregister_probe_event(tp); + unregister_probe_event(tp); __unregister_trace_probe(tp); list_del(&tp->list); } @@ -360,18 +360,16 @@ static int register_trace_probe(struct trace_probe *tp) goto end; } /* register as an event */ - if (tp->call.name) { - old_tp = find_probe_event(tp->call.name); - if (old_tp) { - /* delete old event */ - unregister_trace_probe(old_tp); - free_trace_probe(old_tp); - } - ret = register_probe_event(tp); - if (ret) { - pr_warning("Faild to register probe event(%d)\n", ret); - __unregister_trace_probe(tp); - } + old_tp = find_probe_event(tp->call.name); + if (old_tp) { + /* delete old event */ + unregister_trace_probe(old_tp); + free_trace_probe(old_tp); + } + ret = register_probe_event(tp); + if (ret) { + pr_warning("Faild to register probe event(%d)\n", ret); + __unregister_trace_probe(tp); } list_add_tail(&tp->list, &probe_list); end: @@ -580,7 +578,18 @@ static int create_trace_probe(int argc, char **argv) argc -= 2; argv += 2; /* setup a probe */ - tp = alloc_trace_probe(symbol, event, argc); + if (!event) { + /* Make a new event name */ + char buf[MAX_EVENT_NAME_LEN]; + if (symbol) + snprintf(buf, MAX_EVENT_NAME_LEN, "%c@%s%+ld", + is_return ? 'r' : 'p', symbol, offset); + else + snprintf(buf, MAX_EVENT_NAME_LEN, "%c@0x%p", + is_return ? 'r' : 'p', addr); + tp = alloc_trace_probe(symbol, buf, argc); + } else + tp = alloc_trace_probe(symbol, event, argc); if (IS_ERR(tp)) return PTR_ERR(tp); @@ -661,8 +670,7 @@ static int probes_seq_show(struct seq_file *m, void *v) char buf[MAX_ARGSTR_LEN + 1]; seq_printf(m, "%c", probe_is_return(tp) ? 'r' : 'p'); - if (tp->call.name) - seq_printf(m, ":%s", tp->call.name); + seq_printf(m, ":%s", tp->call.name); if (tp->symbol) seq_printf(m, " %s%+ld", probe_symbol(tp), probe_offset(tp)); @@ -780,10 +788,7 @@ static __kprobes int kprobe_trace_func(struct kprobe *kp, struct pt_regs *regs) struct ring_buffer_event *event; int size, i, pc; unsigned long irq_flags; - struct ftrace_event_call *call = &event_kprobe; - - if (&tp->call.name) - call = &tp->call; + struct ftrace_event_call *call = &tp->call; local_save_flags(irq_flags); pc = preempt_count(); @@ -815,10 +820,7 @@ static __kprobes int kretprobe_trace_func(struct kretprobe_instance *ri, struct ring_buffer_event *event; int size, i, pc; unsigned long irq_flags; - struct ftrace_event_call *call = &event_kretprobe; - - if (&tp->call.name) - call = &tp->call; + struct ftrace_event_call *call = &tp->call; local_save_flags(irq_flags); pc = preempt_count(); -- cgit v1.1 From ff50d99136c3315513ef3b2921e77f35ab04d081 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 13 Aug 2009 16:35:34 -0400 Subject: tracing: Kprobe tracer assigns new event ids for each event MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Assign new event ids for each kprobes event. This doesn't clear ring_buffer when unregistering each kprobe event. Thus, if you mind 'Unknown event' messages, clear the buffer manually after changing kprobe events. Signed-off-by: Masami Hiramatsu Cc: Ananth N Mavinakayanahalli Cc: Avi Kivity Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: Jim Keniston Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Przemysław Pawełczyk Cc: Roland McGrath Cc: Sam Ravnborg Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi Cc: Vegard Nossum LKML-Reference: <20090813203534.31965.49105.stgit@localhost.localdomain> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace.h | 6 ------ kernel/trace/trace_kprobe.c | 51 +++++++++++++-------------------------------- 2 files changed, 15 insertions(+), 42 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 667f832..f5362a0 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -38,8 +38,6 @@ enum trace_type { TRACE_KMEM_FREE, TRACE_POWER, TRACE_BLK, - TRACE_KPROBE, - TRACE_KRETPROBE, __TRACE_LAST_TYPE, }; @@ -343,10 +341,6 @@ extern void __ftrace_bad_type(void); TRACE_KMEM_ALLOC); \ IF_ASSIGN(var, ent, struct kmemtrace_free_entry, \ TRACE_KMEM_FREE); \ - IF_ASSIGN(var, ent, struct kprobe_trace_entry, \ - TRACE_KPROBE); \ - IF_ASSIGN(var, ent, struct kretprobe_trace_entry, \ - TRACE_KRETPROBE); \ __ftrace_bad_type(); \ } while (0) diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 8aeb24c..9c067bf 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -186,6 +186,7 @@ struct trace_probe { }; const char *symbol; /* symbol name */ struct ftrace_event_call call; + struct trace_event event; unsigned int nr_args; struct fetch_func args[]; }; @@ -795,7 +796,7 @@ static __kprobes int kprobe_trace_func(struct kprobe *kp, struct pt_regs *regs) size = SIZEOF_KPROBE_TRACE_ENTRY(tp->nr_args); - event = trace_current_buffer_lock_reserve(TRACE_KPROBE, size, + event = trace_current_buffer_lock_reserve(call->id, size, irq_flags, pc); if (!event) return 0; @@ -827,7 +828,7 @@ static __kprobes int kretprobe_trace_func(struct kretprobe_instance *ri, size = SIZEOF_KRETPROBE_TRACE_ENTRY(tp->nr_args); - event = trace_current_buffer_lock_reserve(TRACE_KRETPROBE, size, + event = trace_current_buffer_lock_reserve(call->id, size, irq_flags, pc); if (!event) return 0; @@ -853,7 +854,7 @@ print_kprobe_event(struct trace_iterator *iter, int flags) struct trace_seq *s = &iter->seq; int i; - trace_assign_type(field, iter->ent); + field = (struct kprobe_trace_entry *)iter->ent; if (!seq_print_ip_sym(s, field->ip, flags | TRACE_ITER_SYM_OFFSET)) goto partial; @@ -880,7 +881,7 @@ print_kretprobe_event(struct trace_iterator *iter, int flags) struct trace_seq *s = &iter->seq; int i; - trace_assign_type(field, iter->ent); + field = (struct kretprobe_trace_entry *)iter->ent; if (!seq_print_ip_sym(s, field->ret_ip, flags | TRACE_ITER_SYM_OFFSET)) goto partial; @@ -906,16 +907,6 @@ partial: return TRACE_TYPE_PARTIAL_LINE; } -static struct trace_event kprobe_trace_event = { - .type = TRACE_KPROBE, - .trace = print_kprobe_event, -}; - -static struct trace_event kretprobe_trace_event = { - .type = TRACE_KRETPROBE, - .trace = print_kretprobe_event, -}; - static int probe_event_enable(struct ftrace_event_call *call) { struct trace_probe *tp = (struct trace_probe *)call->data; @@ -1104,35 +1095,35 @@ static int register_probe_event(struct trace_probe *tp) /* Initialize ftrace_event_call */ call->system = "kprobes"; if (probe_is_return(tp)) { - call->event = &kretprobe_trace_event; - call->id = TRACE_KRETPROBE; + tp->event.trace = print_kretprobe_event; call->raw_init = probe_event_raw_init; call->show_format = kretprobe_event_show_format; call->define_fields = kretprobe_event_define_fields; } else { - call->event = &kprobe_trace_event; - call->id = TRACE_KPROBE; + tp->event.trace = print_kprobe_event; call->raw_init = probe_event_raw_init; call->show_format = kprobe_event_show_format; call->define_fields = kprobe_event_define_fields; } + call->event = &tp->event; + call->id = register_ftrace_event(&tp->event); + if (!call->id) + return -ENODEV; call->enabled = 1; call->regfunc = probe_event_enable; call->unregfunc = probe_event_disable; call->data = tp; ret = trace_add_event_call(call); - if (ret) + if (ret) { pr_info("Failed to register kprobe event: %s\n", call->name); + unregister_ftrace_event(&tp->event); + } return ret; } static void unregister_probe_event(struct trace_probe *tp) { - /* - * Prevent to unregister event itself because the event is shared - * among other probes. - */ - tp->call.event = NULL; + /* tp->event is unregistered in trace_remove_event_call() */ trace_remove_event_call(&tp->call); } @@ -1141,18 +1132,6 @@ static __init int init_kprobe_trace(void) { struct dentry *d_tracer; struct dentry *entry; - int ret; - - ret = register_ftrace_event(&kprobe_trace_event); - if (!ret) { - pr_warning("Could not register kprobe_trace_event type.\n"); - return 0; - } - ret = register_ftrace_event(&kretprobe_trace_event); - if (!ret) { - pr_warning("Could not register kretprobe_trace_event type.\n"); - return 0; - } d_tracer = tracing_init_dentry(); if (!d_tracer) -- cgit v1.1 From cd7e7bd5e44718c7625ce1e1f0fda53d77cd3797 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 13 Aug 2009 16:35:42 -0400 Subject: tracing: Add kprobes event profiling interface MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add profiling interfaces for each kprobes event. This interface provides how many times each probe hit or missed. Signed-off-by: Masami Hiramatsu Cc: Ananth N Mavinakayanahalli Cc: Avi Kivity Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: Jim Keniston Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Przemysław Pawełczyk Cc: Roland McGrath Cc: Sam Ravnborg Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi Cc: Vegard Nossum LKML-Reference: <20090813203541.31965.8452.stgit@localhost.localdomain> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 43 +++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 43 insertions(+) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 9c067bf..ce68197 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -184,6 +184,7 @@ struct trace_probe { struct kprobe kp; struct kretprobe rp; }; + unsigned long nhit; const char *symbol; /* symbol name */ struct ftrace_event_call call; struct trace_event event; @@ -781,6 +782,37 @@ static const struct file_operations kprobe_events_ops = { .write = probes_write, }; +/* Probes profiling interfaces */ +static int probes_profile_seq_show(struct seq_file *m, void *v) +{ + struct trace_probe *tp = v; + + seq_printf(m, " %-44s %15lu %15lu\n", tp->call.name, tp->nhit, + probe_is_return(tp) ? tp->rp.kp.nmissed : tp->kp.nmissed); + + return 0; +} + +static const struct seq_operations profile_seq_op = { + .start = probes_seq_start, + .next = probes_seq_next, + .stop = probes_seq_stop, + .show = probes_profile_seq_show +}; + +static int profile_open(struct inode *inode, struct file *file) +{ + return seq_open(file, &profile_seq_op); +} + +static const struct file_operations kprobe_profile_ops = { + .owner = THIS_MODULE, + .open = profile_open, + .read = seq_read, + .llseek = seq_lseek, + .release = seq_release, +}; + /* Kprobe handler */ static __kprobes int kprobe_trace_func(struct kprobe *kp, struct pt_regs *regs) { @@ -791,6 +823,8 @@ static __kprobes int kprobe_trace_func(struct kprobe *kp, struct pt_regs *regs) unsigned long irq_flags; struct ftrace_event_call *call = &tp->call; + tp->nhit++; + local_save_flags(irq_flags); pc = preempt_count(); @@ -1140,9 +1174,18 @@ static __init int init_kprobe_trace(void) entry = debugfs_create_file("kprobe_events", 0644, d_tracer, NULL, &kprobe_events_ops); + /* Event list interface */ if (!entry) pr_warning("Could not create debugfs " "'kprobe_events' entry\n"); + + /* Profile interface */ + entry = debugfs_create_file("kprobe_profile", 0444, d_tracer, + NULL, &kprobe_profile_ops); + + if (!entry) + pr_warning("Could not create debugfs " + "'kprobe_profile' entry\n"); return 0; } fs_initcall(init_kprobe_trace); -- cgit v1.1 From 38a47497d9e34632abbeb484603cedf10c4b05e4 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Fri, 21 Aug 2009 15:43:43 -0400 Subject: tracing/kprobes: Fix format typo in trace_kprobes MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Fix a format typo in kprobe-tracer. Currently, it shows 'tsize' in format; $ cat /debug/tracing/events/kprobes/event/format ... field: unsigned long ip; offset:16;tsize:8; field: int nargs; offset:24;tsize:4; ... This should be '\tsize'; $ cat /debug/tracing/events/kprobes/event/format ... field: unsigned long ip; offset:16; size:8; field: int nargs; offset:24; size:4; ... Signed-off-by: Masami Hiramatsu Cc: Jim Keniston Cc: H. Peter Anvin Cc: Ananth N Mavinakayanahalli Cc: Avi Kivity Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Przemysław Pawełczyk Cc: Roland McGrath Cc: Sam Ravnborg Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi Cc: Vegard Nossum LKML-Reference: <20090821194343.12478.37618.stgit@localhost.localdomain> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index ce68197..1a9ca79 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -1070,7 +1070,7 @@ static int __probe_event_show_format(struct trace_seq *s, #define SHOW_FIELD(type, item, name) \ do { \ ret = trace_seq_printf(s, "\tfield: " #type " %s;\t" \ - "offset:%u;tsize:%u;\n", name, \ + "offset:%u;\tsize:%u;\n", name, \ (unsigned int)offsetof(typeof(field), item),\ (unsigned int)sizeof(type)); \ if (!ret) \ -- cgit v1.1 From 30a7e073b590ebd1829a906164b0a637e77cc967 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Fri, 21 Aug 2009 15:43:51 -0400 Subject: tracing/kprobes: Change trace_arg to probe_arg MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Change trace_arg_string() and parse_trace_arg() to probe_arg_string() and parse_probe_arg(), since those are kprobe-tracer local functions. Signed-off-by: Masami Hiramatsu Cc: Jim Keniston Cc: H. Peter Anvin Cc: Ananth N Mavinakayanahalli Cc: Avi Kivity Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Przemysław Pawełczyk Cc: Roland McGrath Cc: Sam Ravnborg Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi Cc: Vegard Nossum LKML-Reference: <20090821194351.12478.15247.stgit@localhost.localdomain> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 1a9ca79..f4ec3fc 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -220,7 +220,7 @@ static __kprobes void *probe_address(struct trace_probe *tp) return (probe_is_return(tp)) ? tp->rp.kp.addr : tp->kp.addr; } -static int trace_arg_string(char *buf, size_t n, struct fetch_func *ff) +static int probe_arg_string(char *buf, size_t n, struct fetch_func *ff) { int ret = -EINVAL; @@ -250,7 +250,7 @@ static int trace_arg_string(char *buf, size_t n, struct fetch_func *ff) if (ret >= n) goto end; l += ret; - ret = trace_arg_string(buf + l, n - l, &id->orig); + ret = probe_arg_string(buf + l, n - l, &id->orig); if (ret < 0) goto end; l += ret; @@ -408,7 +408,7 @@ static int split_symbol_offset(char *symbol, long *offset) #define PARAM_MAX_ARGS 16 #define PARAM_MAX_STACK (THREAD_SIZE / sizeof(unsigned long)) -static int parse_trace_arg(char *arg, struct fetch_func *ff, int is_return) +static int parse_probe_arg(char *arg, struct fetch_func *ff, int is_return) { int ret = 0; unsigned long param; @@ -499,7 +499,7 @@ static int parse_trace_arg(char *arg, struct fetch_func *ff, int is_return) if (!id) return -ENOMEM; id->offset = offset; - ret = parse_trace_arg(arg, &id->orig, is_return); + ret = parse_probe_arg(arg, &id->orig, is_return); if (ret) kfree(id); else { @@ -617,7 +617,7 @@ static int create_trace_probe(int argc, char **argv) ret = -ENOSPC; goto error; } - ret = parse_trace_arg(argv[i], &tp->args[i], is_return); + ret = parse_probe_arg(argv[i], &tp->args[i], is_return); if (ret) goto error; } @@ -680,7 +680,7 @@ static int probes_seq_show(struct seq_file *m, void *v) seq_printf(m, " 0x%p", probe_address(tp)); for (i = 0; i < tp->nr_args; i++) { - ret = trace_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); + ret = probe_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); if (ret < 0) { pr_warning("Argument%d decoding error(%d).\n", i, ret); return ret; @@ -997,7 +997,7 @@ static int kprobe_event_define_fields(struct ftrace_event_call *event_call) sprintf(buf, "arg%d", i); DEFINE_FIELD(unsigned long, args[i], buf, 0); /* Set argument string as an alias field */ - ret = trace_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); + ret = probe_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); if (ret < 0) return ret; DEFINE_FIELD(unsigned long, args[i], buf, 0); @@ -1024,7 +1024,7 @@ static int kretprobe_event_define_fields(struct ftrace_event_call *event_call) sprintf(buf, "arg%d", i); DEFINE_FIELD(unsigned long, args[i], buf, 0); /* Set argument string as an alias field */ - ret = trace_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); + ret = probe_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); if (ret < 0) return ret; DEFINE_FIELD(unsigned long, args[i], buf, 0); @@ -1041,7 +1041,7 @@ static int __probe_event_show_format(struct trace_seq *s, /* Show aliases */ for (i = 0; i < tp->nr_args; i++) { - ret = trace_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); + ret = probe_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); if (ret < 0) return ret; if (!trace_seq_printf(s, "\talias: %s;\toriginal: arg%d;\n", -- cgit v1.1 From 24851d2447830e6cba4c4b641cb73e713f312373 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 26 Aug 2009 23:38:30 +0200 Subject: tracing/kprobes: Dump the culprit kprobe in case of kprobe recursion Kprobes can enter into a probing recursion, ie: a kprobe that does an endless loop because one of its core mechanism function used during probing is also probed itself. This patch helps pinpointing the kprobe that raised such recursion by dumping it and raising a BUG instead of a warning (we also disarm the kprobe to try avoiding recursion in BUG itself). Having a BUG instead of a warning stops the stacktrace in the right place and doesn't pollute the logs with hundreds of traces that eventually end up in a stack overflow. Signed-off-by: Frederic Weisbecker Cc: Masami Hiramatsu Cc: Ananth N Mavinakayanahalli --- kernel/kprobes.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'kernel') diff --git a/kernel/kprobes.c b/kernel/kprobes.c index ef177d6..f72e96c 100644 --- a/kernel/kprobes.c +++ b/kernel/kprobes.c @@ -1141,6 +1141,13 @@ static void __kprobes kill_kprobe(struct kprobe *p) arch_remove_kprobe(p); } +void __kprobes dump_kprobe(struct kprobe *kp) +{ + printk(KERN_WARNING "Dumping kprobe:\n"); + printk(KERN_WARNING "Name: %s\nAddress: %p\nOffset: %x\n", + kp->symbol_name, kp->addr, kp->offset); +} + /* Module notifier call back, checking kprobes on the module */ static int __kprobes kprobes_module_callback(struct notifier_block *nb, unsigned long val, void *data) -- cgit v1.1 From aeaeae1187d7520f1c5559623f0a149da6a1c96e Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 27 Aug 2009 05:09:51 +0200 Subject: tracing: Restore the const qualifier for field names and types definition Restore the const qualifier in field's name and type parameters of trace_define_field that was lost while solving a conflict. Fields names and types are defined as builtin constant strings in static TRACE_EVENTs. But kprobes allocates these dynamically. That said, we still want to always pass these strings as const char * in trace_define_fields() to avoid any further accidental writes on the pointed strings. Reported-by: Li Zefan Signed-off-by: Frederic Weisbecker Cc: Steven Rostedt --- kernel/trace/trace_events.c | 4 ++-- kernel/trace/trace_syscalls.c | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c index 8079bb5..197cdaa 100644 --- a/kernel/trace/trace_events.c +++ b/kernel/trace/trace_events.c @@ -27,8 +27,8 @@ DEFINE_MUTEX(event_mutex); LIST_HEAD(ftrace_events); -int trace_define_field(struct ftrace_event_call *call, char *type, - char *name, int offset, int size, int is_signed, +int trace_define_field(struct ftrace_event_call *call, const char *type, + const char *name, int offset, int size, int is_signed, int filter_type) { struct ftrace_event_field *field; diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index 5931933..a928dd0 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -193,8 +193,8 @@ int syscall_enter_define_fields(struct ftrace_event_call *call) return ret; for (i = 0; i < meta->nb_args; i++) { - ret = trace_define_field(call, (char *)meta->types[i], - (char *)meta->args[i], offset, + ret = trace_define_field(call, meta->types[i], + meta->args[i], offset, sizeof(unsigned long), 0, FILTER_OTHER); offset += sizeof(unsigned long); -- cgit v1.1 From f8468f3695209735c1595342f6bd95f7bdab66e1 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 27 Aug 2009 05:23:29 +0200 Subject: tracing: Remove unneeded pointer casts Cleaup uneeded casts from void * to char * in syscalls tracing file. Reported-by: Li Zefan Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_syscalls.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index a928dd0..e7c676e 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -324,7 +324,7 @@ int reg_event_syscall_exit(struct ftrace_event_call *call) int num; char *name; - name = (char *)call->data; + name = call->data; num = syscall_name_to_nr(name); if (num < 0 || num >= FTRACE_SYSCALL_MAX) return -ENOSYS; @@ -347,7 +347,7 @@ void unreg_event_syscall_exit(struct ftrace_event_call *call) int num; char *name; - name = (char *)call->data; + name = call->data; num = syscall_name_to_nr(name); if (num < 0 || num >= FTRACE_SYSCALL_MAX) return; -- cgit v1.1 From 8f270083587a4cb70fa14f0e2fd698eb08a4dd07 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 27 Aug 2009 13:23:18 -0400 Subject: kprobes: Fix to add __kprobes to notify_die Add __kprobes to notify_die() because do_int3() calls notify_die() instead of atomic_notify_call_chain() which is already marked as __kprobes. Signed-off-by: Masami Hiramatsu Acked-by: Ananth N Mavinakayanahalli Cc: Ingo Molnar LKML-Reference: <20090827172318.8246.53702.stgit@localhost.localdomain> Signed-off-by: Frederic Weisbecker --- kernel/notifier.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/notifier.c b/kernel/notifier.c index 61d5aa5..acd24e7 100644 --- a/kernel/notifier.c +++ b/kernel/notifier.c @@ -558,7 +558,7 @@ EXPORT_SYMBOL(unregister_reboot_notifier); static ATOMIC_NOTIFIER_HEAD(die_chain); -int notrace notify_die(enum die_val val, const char *str, +int notrace __kprobes notify_die(enum die_val val, const char *str, struct pt_regs *regs, long err, int trap, int sig) { struct die_args args = { -- cgit v1.1 From 65e234ec2c4a0659ca22531dc1372a185f088517 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 27 Aug 2009 13:23:32 -0400 Subject: kprobes: Prohibit to probe native_get_debugreg Since do_debug() calls get_debugreg(), native_get_debugreg() will be called from singlestepping. This can cause an int3 infinite loop. We can't put it in the .text.kprobes section because it is inlined, then we blacklist its name. Signed-off-by: Masami Hiramatsu Acked-by: Ananth N Mavinakayanahalli Cc: Ingo Molnar LKML-Reference: <20090827172332.8246.34194.stgit@localhost.localdomain> Signed-off-by: Frederic Weisbecker --- kernel/kprobes.c | 1 + 1 file changed, 1 insertion(+) (limited to 'kernel') diff --git a/kernel/kprobes.c b/kernel/kprobes.c index f72e96c..3267d90 100644 --- a/kernel/kprobes.c +++ b/kernel/kprobes.c @@ -90,6 +90,7 @@ static spinlock_t *kretprobe_table_lock_ptr(unsigned long hash) */ static struct kprobe_blackpoint kprobe_blacklist[] = { {"preempt_schedule",}, + {"native_get_debugreg",}, {NULL} /* Terminator */ }; -- cgit v1.1 From d6a65dffb30d8636b1e5d4c201564ef401a246cf Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Mon, 7 Sep 2009 03:23:20 +0200 Subject: tracing: Fix ring-buffer and ksym tracer merge interaction The compiler warns us about: kernel/trace/trace_ksym.c: In function ksym_hbp_handler: kernel/trace/trace_ksym.c:92: attention : passing argument 1 of trace_buffer_lock_reserve from incompatible pointer type kernel/trace/trace_ksym.c:106: attention : passing argument 1 of trace_buffer_unlock_commit from incompatible pointer type Commit "e77405ad" (tracing: pass around ring buffer instead of tracer) has changed the central tracing APIs. And this change has updated every callsites of these APIs except those that aren't in tracing/core, such as the ksym tracer. Cc: Steven Rostedt Signed-off-by: Ingo Molnar --- kernel/trace/trace_ksym.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index 2fde875..6d5609c 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -78,17 +78,18 @@ void ksym_collect_stats(unsigned long hbp_hit_addr) void ksym_hbp_handler(struct hw_breakpoint *hbp, struct pt_regs *regs) { struct ring_buffer_event *event; - struct trace_array *tr; struct ksym_trace_entry *entry; + struct ring_buffer *buffer; int pc; if (!ksym_tracing_enabled) return; - tr = ksym_trace_array; + buffer = ksym_trace_array->buffer; + pc = preempt_count(); - event = trace_buffer_lock_reserve(tr, TRACE_KSYM, + event = trace_buffer_lock_reserve(buffer, TRACE_KSYM, sizeof(*entry), 0, pc); if (!event) return; @@ -103,7 +104,7 @@ void ksym_hbp_handler(struct hw_breakpoint *hbp, struct pt_regs *regs) ksym_collect_stats(hbp->info.address); #endif /* CONFIG_PROFILE_KSYM_TRACER */ - trace_buffer_unlock_commit(tr, event, 0, pc); + trace_buffer_unlock_commit(buffer, event, 0, pc); } /* Valid access types are represented as -- cgit v1.1 From a00e817f42663941ea0aa5f85a9d1c4f8b212839 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Tue, 8 Sep 2009 12:47:55 -0400 Subject: kprobes/x86-32: Move irq-exit functions to kprobes section Move irq-exit functions to .kprobes.text section to protect against kprobes recursion. When I ran kprobe stress test on x86-32, I found below symbols cause unrecoverable recursive probing: ret_from_exception ret_from_intr check_userspace restore_all restore_all_notrace restore_nocheck irq_return And also, I found some interrupt/exception entry points that cause similar problems. This patch moves those symbols (including their container functions) to .kprobes.text section to prevent any kprobes probing. Signed-off-by: Masami Hiramatsu Cc: Frederic Weisbecker Cc: Ananth N Mavinakayanahalli Cc: Jim Keniston Cc: Ingo Molnar LKML-Reference: <20090908164755.24050.81182.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/kprobes.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'kernel') diff --git a/kernel/kprobes.c b/kernel/kprobes.c index 3267d90..00d01b0 100644 --- a/kernel/kprobes.c +++ b/kernel/kprobes.c @@ -91,6 +91,8 @@ static spinlock_t *kretprobe_table_lock_ptr(unsigned long hash) static struct kprobe_blackpoint kprobe_blacklist[] = { {"preempt_schedule",}, {"native_get_debugreg",}, + {"irq_entries_start",}, + {"common_interrupt",}, {NULL} /* Terminator */ }; -- cgit v1.1 From 2fba0c8867af47f6455490e7b59e512dd180c027 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 10 Sep 2009 19:53:14 -0400 Subject: tracing/kprobes: Fix probe offset to be unsigned Prohibit user to specify negative offset from symbols. Since kprobe.offset is unsigned int, the offset must be always positive value. Signed-off-by: Masami Hiramatsu Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi LKML-Reference: <20090910235314.22412.64631.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 19 +++++++------------ 1 file changed, 7 insertions(+), 12 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 19a6de6..c24b7e9 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -210,7 +210,7 @@ static __kprobes const char *probe_symbol(struct trace_probe *tp) return tp->symbol ? tp->symbol : "unknown"; } -static __kprobes long probe_offset(struct trace_probe *tp) +static __kprobes unsigned int probe_offset(struct trace_probe *tp) { return (probe_is_return(tp)) ? tp->rp.kp.offset : tp->kp.offset; } @@ -380,7 +380,7 @@ end: } /* Split symbol and offset. */ -static int split_symbol_offset(char *symbol, long *offset) +static int split_symbol_offset(char *symbol, unsigned long *offset) { char *tmp; int ret; @@ -389,16 +389,11 @@ static int split_symbol_offset(char *symbol, long *offset) return -EINVAL; tmp = strchr(symbol, '+'); - if (!tmp) - tmp = strchr(symbol, '-'); - if (tmp) { /* skip sign because strict_strtol doesn't accept '+' */ - ret = strict_strtol(tmp + 1, 0, offset); + ret = strict_strtoul(tmp + 1, 0, offset); if (ret) return ret; - if (*tmp == '-') - *offset = -(*offset); *tmp = '\0'; } else *offset = 0; @@ -520,7 +515,7 @@ static int create_trace_probe(int argc, char **argv) { /* * Argument syntax: - * - Add kprobe: p[:EVENT] SYMBOL[+OFFS|-OFFS]|ADDRESS [FETCHARGS] + * - Add kprobe: p[:EVENT] SYMBOL[+OFFS]|ADDRESS [FETCHARGS] * - Add kretprobe: r[:EVENT] SYMBOL[+0] [FETCHARGS] * Fetch args: * aN : fetch Nth of function argument. (N:0-) @@ -539,7 +534,7 @@ static int create_trace_probe(int argc, char **argv) int i, ret = 0; int is_return = 0; char *symbol = NULL, *event = NULL; - long offset = 0; + unsigned long offset = 0; void *addr = NULL; if (argc < 2) @@ -605,7 +600,7 @@ static int create_trace_probe(int argc, char **argv) if (tp->symbol) { kp->symbol_name = tp->symbol; - kp->offset = offset; + kp->offset = (unsigned int)offset; } else kp->addr = addr; @@ -675,7 +670,7 @@ static int probes_seq_show(struct seq_file *m, void *v) seq_printf(m, ":%s", tp->call.name); if (tp->symbol) - seq_printf(m, " %s%+ld", probe_symbol(tp), probe_offset(tp)); + seq_printf(m, " %s+%u", probe_symbol(tp), probe_offset(tp)); else seq_printf(m, " 0x%p", probe_address(tp)); -- cgit v1.1 From 4a846b443b4e8633057946a2234e23559a67ce42 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Fri, 11 Sep 2009 05:31:21 +0200 Subject: tracing/kprobes: Cleanup kprobe tracer code. Simplify trace_probe to remove a union, and remove some redundant wrappers. And also, cleanup create_trace_probe() function. Signed-off-by: Masami Hiramatsu Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi LKML-Reference: <20090910235322.22412.52525.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 81 +++++++++++++++++++-------------------------- 1 file changed, 34 insertions(+), 47 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index c24b7e9..4ce728c 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -180,10 +180,7 @@ static __kprobes void free_indirect_fetch_data(struct indirect_fetch_data *data) struct trace_probe { struct list_head list; - union { - struct kprobe kp; - struct kretprobe rp; - }; + struct kretprobe rp; /* Use rp.kp for kprobe use */ unsigned long nhit; const char *symbol; /* symbol name */ struct ftrace_event_call call; @@ -202,7 +199,7 @@ static int kretprobe_trace_func(struct kretprobe_instance *ri, static __kprobes int probe_is_return(struct trace_probe *tp) { - return (tp->rp.handler == kretprobe_trace_func); + return tp->rp.handler != NULL; } static __kprobes const char *probe_symbol(struct trace_probe *tp) @@ -210,16 +207,6 @@ static __kprobes const char *probe_symbol(struct trace_probe *tp) return tp->symbol ? tp->symbol : "unknown"; } -static __kprobes unsigned int probe_offset(struct trace_probe *tp) -{ - return (probe_is_return(tp)) ? tp->rp.kp.offset : tp->kp.offset; -} - -static __kprobes void *probe_address(struct trace_probe *tp) -{ - return (probe_is_return(tp)) ? tp->rp.kp.addr : tp->kp.addr; -} - static int probe_arg_string(char *buf, size_t n, struct fetch_func *ff) { int ret = -EINVAL; @@ -269,8 +256,14 @@ static void unregister_probe_event(struct trace_probe *tp); static DEFINE_MUTEX(probe_lock); static LIST_HEAD(probe_list); -static struct trace_probe *alloc_trace_probe(const char *symbol, - const char *event, int nargs) +/* + * Allocate new trace_probe and initialize it (including kprobes). + */ +static struct trace_probe *alloc_trace_probe(const char *event, + void *addr, + const char *symbol, + unsigned long offs, + int nargs, int is_return) { struct trace_probe *tp; @@ -282,7 +275,16 @@ static struct trace_probe *alloc_trace_probe(const char *symbol, tp->symbol = kstrdup(symbol, GFP_KERNEL); if (!tp->symbol) goto error; - } + tp->rp.kp.symbol_name = tp->symbol; + tp->rp.kp.offset = offs; + } else + tp->rp.kp.addr = addr; + + if (is_return) + tp->rp.handler = kretprobe_trace_func; + else + tp->rp.kp.pre_handler = kprobe_trace_func; + if (!event) goto error; tp->call.name = kstrdup(event, GFP_KERNEL); @@ -327,7 +329,7 @@ static void __unregister_trace_probe(struct trace_probe *tp) if (probe_is_return(tp)) unregister_kretprobe(&tp->rp); else - unregister_kprobe(&tp->kp); + unregister_kprobe(&tp->rp.kp); } /* Unregister a trace_probe and probe_event: call with locking probe_lock */ @@ -349,14 +351,14 @@ static int register_trace_probe(struct trace_probe *tp) if (probe_is_return(tp)) ret = register_kretprobe(&tp->rp); else - ret = register_kprobe(&tp->kp); + ret = register_kprobe(&tp->rp.kp); if (ret) { pr_warning("Could not insert probe(%d)\n", ret); if (ret == -EILSEQ) { pr_warning("Probing address(0x%p) is not an " "instruction boundary.\n", - probe_address(tp)); + tp->rp.kp.addr); ret = -EINVAL; } goto end; @@ -530,12 +532,12 @@ static int create_trace_probe(int argc, char **argv) * +|-offs(ARG) : fetch memory at ARG +|- offs address. */ struct trace_probe *tp; - struct kprobe *kp; int i, ret = 0; int is_return = 0; char *symbol = NULL, *event = NULL; unsigned long offset = 0; void *addr = NULL; + char buf[MAX_EVENT_NAME_LEN]; if (argc < 2) return -EINVAL; @@ -577,33 +579,18 @@ static int create_trace_probe(int argc, char **argv) /* setup a probe */ if (!event) { /* Make a new event name */ - char buf[MAX_EVENT_NAME_LEN]; if (symbol) snprintf(buf, MAX_EVENT_NAME_LEN, "%c@%s%+ld", is_return ? 'r' : 'p', symbol, offset); else snprintf(buf, MAX_EVENT_NAME_LEN, "%c@0x%p", is_return ? 'r' : 'p', addr); - tp = alloc_trace_probe(symbol, buf, argc); - } else - tp = alloc_trace_probe(symbol, event, argc); + event = buf; + } + tp = alloc_trace_probe(event, addr, symbol, offset, argc, is_return); if (IS_ERR(tp)) return PTR_ERR(tp); - if (is_return) { - kp = &tp->rp.kp; - tp->rp.handler = kretprobe_trace_func; - } else { - kp = &tp->kp; - tp->kp.pre_handler = kprobe_trace_func; - } - - if (tp->symbol) { - kp->symbol_name = tp->symbol; - kp->offset = (unsigned int)offset; - } else - kp->addr = addr; - /* parse arguments */ ret = 0; for (i = 0; i < argc && i < MAX_TRACE_ARGS; i++) { @@ -670,9 +657,9 @@ static int probes_seq_show(struct seq_file *m, void *v) seq_printf(m, ":%s", tp->call.name); if (tp->symbol) - seq_printf(m, " %s+%u", probe_symbol(tp), probe_offset(tp)); + seq_printf(m, " %s+%u", probe_symbol(tp), tp->rp.kp.offset); else - seq_printf(m, " 0x%p", probe_address(tp)); + seq_printf(m, " 0x%p", tp->rp.kp.addr); for (i = 0; i < tp->nr_args; i++) { ret = probe_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); @@ -783,7 +770,7 @@ static int probes_profile_seq_show(struct seq_file *m, void *v) struct trace_probe *tp = v; seq_printf(m, " %-44s %15lu %15lu\n", tp->call.name, tp->nhit, - probe_is_return(tp) ? tp->rp.kp.nmissed : tp->kp.nmissed); + tp->rp.kp.nmissed); return 0; } @@ -811,7 +798,7 @@ static const struct file_operations kprobe_profile_ops = { /* Kprobe handler */ static __kprobes int kprobe_trace_func(struct kprobe *kp, struct pt_regs *regs) { - struct trace_probe *tp = container_of(kp, struct trace_probe, kp); + struct trace_probe *tp = container_of(kp, struct trace_probe, rp.kp); struct kprobe_trace_entry *entry; struct ring_buffer_event *event; struct ring_buffer *buffer; @@ -866,7 +853,7 @@ static __kprobes int kretprobe_trace_func(struct kretprobe_instance *ri, entry = ring_buffer_event_data(event); entry->nargs = tp->nr_args; - entry->func = (unsigned long)probe_address(tp); + entry->func = (unsigned long)tp->rp.kp.addr; entry->ret_ip = (unsigned long)ri->ret_addr; for (i = 0; i < tp->nr_args; i++) entry->args[i] = call_fetch(&tp->args[i], regs); @@ -945,7 +932,7 @@ static int probe_event_enable(struct ftrace_event_call *call) if (probe_is_return(tp)) return enable_kretprobe(&tp->rp); else - return enable_kprobe(&tp->kp); + return enable_kprobe(&tp->rp.kp); } static void probe_event_disable(struct ftrace_event_call *call) @@ -955,7 +942,7 @@ static void probe_event_disable(struct ftrace_event_call *call) if (probe_is_return(tp)) disable_kretprobe(&tp->rp); else - disable_kprobe(&tp->kp); + disable_kprobe(&tp->rp.kp); } static int probe_event_raw_init(struct ftrace_event_call *event_call) -- cgit v1.1 From e08d1c657f70bcaca11401cd6ac5c8fe59bd2bb7 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 10 Sep 2009 19:53:30 -0400 Subject: tracing/kprobes: Add event profiling support Add *probe_profile_enable/disable to support kprobes raw events sampling from perf counters, like other ftrace events, when CONFIG_PROFILE_EVENT=y. Signed-off-by: Masami Hiramatsu Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi LKML-Reference: <20090910235329.22412.94731.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 110 +++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 108 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 4ce728c..730e992 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -28,6 +28,7 @@ #include #include #include +#include #include "trace.h" #include "trace_output.h" @@ -280,6 +281,7 @@ static struct trace_probe *alloc_trace_probe(const char *event, } else tp->rp.kp.addr = addr; + /* Set handler here for checking whether this probe is return or not. */ if (is_return) tp->rp.handler = kretprobe_trace_func; else @@ -929,10 +931,13 @@ static int probe_event_enable(struct ftrace_event_call *call) { struct trace_probe *tp = (struct trace_probe *)call->data; - if (probe_is_return(tp)) + if (probe_is_return(tp)) { + tp->rp.handler = kretprobe_trace_func; return enable_kretprobe(&tp->rp); - else + } else { + tp->rp.kp.pre_handler = kprobe_trace_func; return enable_kprobe(&tp->rp.kp); + } } static void probe_event_disable(struct ftrace_event_call *call) @@ -1105,6 +1110,101 @@ static int kretprobe_event_show_format(struct ftrace_event_call *call, "func, ret_ip"); } +#ifdef CONFIG_EVENT_PROFILE + +/* Kprobe profile handler */ +static __kprobes int kprobe_profile_func(struct kprobe *kp, + struct pt_regs *regs) +{ + struct trace_probe *tp = container_of(kp, struct trace_probe, rp.kp); + struct ftrace_event_call *call = &tp->call; + struct kprobe_trace_entry *entry; + int size, i, pc; + unsigned long irq_flags; + + local_save_flags(irq_flags); + pc = preempt_count(); + + size = SIZEOF_KPROBE_TRACE_ENTRY(tp->nr_args); + + do { + char raw_data[size]; + struct trace_entry *ent; + + *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; + entry = (struct kprobe_trace_entry *)raw_data; + ent = &entry->ent; + + tracing_generic_entry_update(ent, irq_flags, pc); + ent->type = call->id; + entry->nargs = tp->nr_args; + entry->ip = (unsigned long)kp->addr; + for (i = 0; i < tp->nr_args; i++) + entry->args[i] = call_fetch(&tp->args[i], regs); + perf_tpcounter_event(call->id, entry->ip, 1, entry, size); + } while (0); + return 0; +} + +/* Kretprobe profile handler */ +static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri, + struct pt_regs *regs) +{ + struct trace_probe *tp = container_of(ri->rp, struct trace_probe, rp); + struct ftrace_event_call *call = &tp->call; + struct kretprobe_trace_entry *entry; + int size, i, pc; + unsigned long irq_flags; + + local_save_flags(irq_flags); + pc = preempt_count(); + + size = SIZEOF_KRETPROBE_TRACE_ENTRY(tp->nr_args); + + do { + char raw_data[size]; + struct trace_entry *ent; + + *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; + entry = (struct kretprobe_trace_entry *)raw_data; + ent = &entry->ent; + + tracing_generic_entry_update(ent, irq_flags, pc); + ent->type = call->id; + entry->nargs = tp->nr_args; + entry->func = (unsigned long)tp->rp.kp.addr; + entry->ret_ip = (unsigned long)ri->ret_addr; + for (i = 0; i < tp->nr_args; i++) + entry->args[i] = call_fetch(&tp->args[i], regs); + perf_tpcounter_event(call->id, entry->ret_ip, 1, entry, size); + } while (0); + return 0; +} + +static int probe_profile_enable(struct ftrace_event_call *call) +{ + struct trace_probe *tp = (struct trace_probe *)call->data; + + if (atomic_inc_return(&call->profile_count)) + return 0; + + if (probe_is_return(tp)) { + tp->rp.handler = kretprobe_profile_func; + return enable_kretprobe(&tp->rp); + } else { + tp->rp.kp.pre_handler = kprobe_profile_func; + return enable_kprobe(&tp->rp.kp); + } +} + +static void probe_profile_disable(struct ftrace_event_call *call) +{ + if (atomic_add_negative(-1, &call->profile_count)) + probe_event_disable(call); +} + +#endif /* CONFIG_EVENT_PROFILE */ + static int register_probe_event(struct trace_probe *tp) { struct ftrace_event_call *call = &tp->call; @@ -1130,6 +1230,12 @@ static int register_probe_event(struct trace_probe *tp) call->enabled = 1; call->regfunc = probe_event_enable; call->unregfunc = probe_event_disable; + +#ifdef CONFIG_EVENT_PROFILE + atomic_set(&call->profile_count, -1); + call->profile_enable = probe_profile_enable; + call->profile_disable = probe_profile_disable; +#endif call->data = tp; ret = trace_add_event_call(call); if (ret) { -- cgit v1.1 From eca0d916f6429785bbc88db3ff66631cde62b432 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 10 Sep 2009 19:53:38 -0400 Subject: tracing/kprobes: Add argument name support Add argument name assignment support and remove "alias" lines from format. This allows user to assign unique name to each argument. For example, $ echo p do_sys_open dfd=a0 filename=a1 flags=a2 mode=a3 > kprobe_events This assigns dfd, filename, flags, and mode to 1st - 4th arguments respectively. Trace buffer shows those names too. <...>-1439 [000] 1200885.933147: do_sys_open+0x0/0xdf: dfd=ffffff9c filename=bfa898ac flags=8000 mode=0 This helps users to know what each value means. Users can filter each events by these names too. Note that you can not filter by argN anymore. Signed-off-by: Masami Hiramatsu Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi LKML-Reference: <20090910235337.22412.77383.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 128 ++++++++++++++++++++++---------------------- 1 file changed, 64 insertions(+), 64 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 730e992..44dad1a 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -176,9 +176,14 @@ static __kprobes void free_indirect_fetch_data(struct indirect_fetch_data *data) } /** - * kprobe_trace_core + * Kprobe tracer core functions */ +struct probe_arg { + struct fetch_func fetch; + const char *name; +}; + struct trace_probe { struct list_head list; struct kretprobe rp; /* Use rp.kp for kprobe use */ @@ -187,12 +192,12 @@ struct trace_probe { struct ftrace_event_call call; struct trace_event event; unsigned int nr_args; - struct fetch_func args[]; + struct probe_arg args[]; }; #define SIZEOF_TRACE_PROBE(n) \ (offsetof(struct trace_probe, args) + \ - (sizeof(struct fetch_func) * (n))) + (sizeof(struct probe_arg) * (n))) static int kprobe_trace_func(struct kprobe *kp, struct pt_regs *regs); static int kretprobe_trace_func(struct kretprobe_instance *ri, @@ -301,15 +306,21 @@ error: return ERR_PTR(-ENOMEM); } +static void free_probe_arg(struct probe_arg *arg) +{ + if (arg->fetch.func == fetch_symbol) + free_symbol_cache(arg->fetch.data); + else if (arg->fetch.func == fetch_indirect) + free_indirect_fetch_data(arg->fetch.data); + kfree(arg->name); +} + static void free_trace_probe(struct trace_probe *tp) { int i; for (i = 0; i < tp->nr_args; i++) - if (tp->args[i].func == fetch_symbol) - free_symbol_cache(tp->args[i].data); - else if (tp->args[i].func == fetch_indirect) - free_indirect_fetch_data(tp->args[i].data); + free_probe_arg(&tp->args[i]); kfree(tp->call.name); kfree(tp->symbol); @@ -532,11 +543,13 @@ static int create_trace_probe(int argc, char **argv) * %REG : fetch register REG * Indirect memory fetch: * +|-offs(ARG) : fetch memory at ARG +|- offs address. + * Alias name of args: + * NAME=FETCHARG : set NAME as alias of FETCHARG. */ struct trace_probe *tp; int i, ret = 0; int is_return = 0; - char *symbol = NULL, *event = NULL; + char *symbol = NULL, *event = NULL, *arg = NULL; unsigned long offset = 0; void *addr = NULL; char buf[MAX_EVENT_NAME_LEN]; @@ -596,12 +609,21 @@ static int create_trace_probe(int argc, char **argv) /* parse arguments */ ret = 0; for (i = 0; i < argc && i < MAX_TRACE_ARGS; i++) { - if (strlen(argv[i]) > MAX_ARGSTR_LEN) { - pr_info("Argument%d(%s) is too long.\n", i, argv[i]); + /* Parse argument name */ + arg = strchr(argv[i], '='); + if (arg) + *arg++ = '\0'; + else + arg = argv[i]; + tp->args[i].name = kstrdup(argv[i], GFP_KERNEL); + + /* Parse fetch argument */ + if (strlen(arg) > MAX_ARGSTR_LEN) { + pr_info("Argument%d(%s) is too long.\n", i, arg); ret = -ENOSPC; goto error; } - ret = parse_probe_arg(argv[i], &tp->args[i], is_return); + ret = parse_probe_arg(arg, &tp->args[i].fetch, is_return); if (ret) goto error; } @@ -664,12 +686,12 @@ static int probes_seq_show(struct seq_file *m, void *v) seq_printf(m, " 0x%p", tp->rp.kp.addr); for (i = 0; i < tp->nr_args; i++) { - ret = probe_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); + ret = probe_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i].fetch); if (ret < 0) { pr_warning("Argument%d decoding error(%d).\n", i, ret); return ret; } - seq_printf(m, " %s", buf); + seq_printf(m, " %s=%s", tp->args[i].name, buf); } seq_printf(m, "\n"); return 0; @@ -824,7 +846,7 @@ static __kprobes int kprobe_trace_func(struct kprobe *kp, struct pt_regs *regs) entry->nargs = tp->nr_args; entry->ip = (unsigned long)kp->addr; for (i = 0; i < tp->nr_args; i++) - entry->args[i] = call_fetch(&tp->args[i], regs); + entry->args[i] = call_fetch(&tp->args[i].fetch, regs); if (!filter_current_check_discard(buffer, call, entry, event)) trace_nowake_buffer_unlock_commit(buffer, event, irq_flags, pc); @@ -858,7 +880,7 @@ static __kprobes int kretprobe_trace_func(struct kretprobe_instance *ri, entry->func = (unsigned long)tp->rp.kp.addr; entry->ret_ip = (unsigned long)ri->ret_addr; for (i = 0; i < tp->nr_args; i++) - entry->args[i] = call_fetch(&tp->args[i], regs); + entry->args[i] = call_fetch(&tp->args[i].fetch, regs); if (!filter_current_check_discard(buffer, call, entry, event)) trace_nowake_buffer_unlock_commit(buffer, event, irq_flags, pc); @@ -872,9 +894,13 @@ print_kprobe_event(struct trace_iterator *iter, int flags) { struct kprobe_trace_entry *field; struct trace_seq *s = &iter->seq; + struct trace_event *event; + struct trace_probe *tp; int i; field = (struct kprobe_trace_entry *)iter->ent; + event = ftrace_find_event(field->ent.type); + tp = container_of(event, struct trace_probe, event); if (!seq_print_ip_sym(s, field->ip, flags | TRACE_ITER_SYM_OFFSET)) goto partial; @@ -883,7 +909,8 @@ print_kprobe_event(struct trace_iterator *iter, int flags) goto partial; for (i = 0; i < field->nargs; i++) - if (!trace_seq_printf(s, " 0x%lx", field->args[i])) + if (!trace_seq_printf(s, " %s=%lx", + tp->args[i].name, field->args[i])) goto partial; if (!trace_seq_puts(s, "\n")) @@ -899,9 +926,13 @@ print_kretprobe_event(struct trace_iterator *iter, int flags) { struct kretprobe_trace_entry *field; struct trace_seq *s = &iter->seq; + struct trace_event *event; + struct trace_probe *tp; int i; field = (struct kretprobe_trace_entry *)iter->ent; + event = ftrace_find_event(field->ent.type); + tp = container_of(event, struct trace_probe, event); if (!seq_print_ip_sym(s, field->ret_ip, flags | TRACE_ITER_SYM_OFFSET)) goto partial; @@ -916,7 +947,8 @@ print_kretprobe_event(struct trace_iterator *iter, int flags) goto partial; for (i = 0; i < field->nargs; i++) - if (!trace_seq_printf(s, " 0x%lx", field->args[i])) + if (!trace_seq_printf(s, " %s=%lx", + tp->args[i].name, field->args[i])) goto partial; if (!trace_seq_puts(s, "\n")) @@ -972,7 +1004,6 @@ static int kprobe_event_define_fields(struct ftrace_event_call *event_call) { int ret, i; struct kprobe_trace_entry field; - char buf[MAX_ARGSTR_LEN + 1]; struct trace_probe *tp = (struct trace_probe *)event_call->data; ret = trace_define_common_fields(event_call); @@ -981,16 +1012,9 @@ static int kprobe_event_define_fields(struct ftrace_event_call *event_call) DEFINE_FIELD(unsigned long, ip, "ip", 0); DEFINE_FIELD(int, nargs, "nargs", 1); - for (i = 0; i < tp->nr_args; i++) { - /* Set argN as a field */ - sprintf(buf, "arg%d", i); - DEFINE_FIELD(unsigned long, args[i], buf, 0); - /* Set argument string as an alias field */ - ret = probe_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); - if (ret < 0) - return ret; - DEFINE_FIELD(unsigned long, args[i], buf, 0); - } + /* Set argument names as fields */ + for (i = 0; i < tp->nr_args; i++) + DEFINE_FIELD(unsigned long, args[i], tp->args[i].name, 0); return 0; } @@ -998,7 +1022,6 @@ static int kretprobe_event_define_fields(struct ftrace_event_call *event_call) { int ret, i; struct kretprobe_trace_entry field; - char buf[MAX_ARGSTR_LEN + 1]; struct trace_probe *tp = (struct trace_probe *)event_call->data; ret = trace_define_common_fields(event_call); @@ -1008,16 +1031,9 @@ static int kretprobe_event_define_fields(struct ftrace_event_call *event_call) DEFINE_FIELD(unsigned long, func, "func", 0); DEFINE_FIELD(unsigned long, ret_ip, "ret_ip", 0); DEFINE_FIELD(int, nargs, "nargs", 1); - for (i = 0; i < tp->nr_args; i++) { - /* Set argN as a field */ - sprintf(buf, "arg%d", i); - DEFINE_FIELD(unsigned long, args[i], buf, 0); - /* Set argument string as an alias field */ - ret = probe_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); - if (ret < 0) - return ret; - DEFINE_FIELD(unsigned long, args[i], buf, 0); - } + /* Set argument names as fields */ + for (i = 0; i < tp->nr_args; i++) + DEFINE_FIELD(unsigned long, args[i], tp->args[i].name, 0); return 0; } @@ -1025,31 +1041,21 @@ static int __probe_event_show_format(struct trace_seq *s, struct trace_probe *tp, const char *fmt, const char *arg) { - int i, ret; - char buf[MAX_ARGSTR_LEN + 1]; + int i; - /* Show aliases */ - for (i = 0; i < tp->nr_args; i++) { - ret = probe_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]); - if (ret < 0) - return ret; - if (!trace_seq_printf(s, "\talias: %s;\toriginal: arg%d;\n", - buf, i)) - return 0; - } /* Show format */ if (!trace_seq_printf(s, "\nprint fmt: \"%s", fmt)) return 0; for (i = 0; i < tp->nr_args; i++) - if (!trace_seq_puts(s, " 0x%lx")) + if (!trace_seq_printf(s, " %s=%%lx", tp->args[i].name)) return 0; if (!trace_seq_printf(s, "\", %s", arg)) return 0; for (i = 0; i < tp->nr_args; i++) - if (!trace_seq_printf(s, ", arg%d", i)) + if (!trace_seq_printf(s, ", REC->%s", tp->args[i].name)) return 0; return trace_seq_puts(s, "\n"); @@ -1071,17 +1077,14 @@ static int kprobe_event_show_format(struct ftrace_event_call *call, { struct kprobe_trace_entry field __attribute__((unused)); int ret, i; - char buf[8]; struct trace_probe *tp = (struct trace_probe *)call->data; SHOW_FIELD(unsigned long, ip, "ip"); SHOW_FIELD(int, nargs, "nargs"); /* Show fields */ - for (i = 0; i < tp->nr_args; i++) { - sprintf(buf, "arg%d", i); - SHOW_FIELD(unsigned long, args[i], buf); - } + for (i = 0; i < tp->nr_args; i++) + SHOW_FIELD(unsigned long, args[i], tp->args[i].name); trace_seq_puts(s, "\n"); return __probe_event_show_format(s, tp, "%lx:", "ip"); @@ -1092,7 +1095,6 @@ static int kretprobe_event_show_format(struct ftrace_event_call *call, { struct kretprobe_trace_entry field __attribute__((unused)); int ret, i; - char buf[8]; struct trace_probe *tp = (struct trace_probe *)call->data; SHOW_FIELD(unsigned long, func, "func"); @@ -1100,10 +1102,8 @@ static int kretprobe_event_show_format(struct ftrace_event_call *call, SHOW_FIELD(int, nargs, "nargs"); /* Show fields */ - for (i = 0; i < tp->nr_args; i++) { - sprintf(buf, "arg%d", i); - SHOW_FIELD(unsigned long, args[i], buf); - } + for (i = 0; i < tp->nr_args; i++) + SHOW_FIELD(unsigned long, args[i], tp->args[i].name); trace_seq_puts(s, "\n"); return __probe_event_show_format(s, tp, "%lx <- %lx:", @@ -1140,7 +1140,7 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp, entry->nargs = tp->nr_args; entry->ip = (unsigned long)kp->addr; for (i = 0; i < tp->nr_args; i++) - entry->args[i] = call_fetch(&tp->args[i], regs); + entry->args[i] = call_fetch(&tp->args[i].fetch, regs); perf_tpcounter_event(call->id, entry->ip, 1, entry, size); } while (0); return 0; @@ -1175,7 +1175,7 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri, entry->func = (unsigned long)tp->rp.kp.addr; entry->ret_ip = (unsigned long)ri->ret_addr; for (i = 0; i < tp->nr_args; i++) - entry->args[i] = call_fetch(&tp->args[i], regs); + entry->args[i] = call_fetch(&tp->args[i].fetch, regs); perf_tpcounter_event(call->id, entry->ret_ip, 1, entry, size); } while (0); return 0; -- cgit v1.1 From 6e9f23d1619f7badaf9090dac09e86a22d6061d8 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 10 Sep 2009 19:53:45 -0400 Subject: tracing/kprobes: Show event name in trace output Show event name in tracing/trace output. This also fixes kprobes events format to comply with other tracepoint events formats. Before patching: <...>-1447 [001] 1038282.286875: do_sys_open+0x0/0xd6: ... <...>-1447 [001] 1038282.286878: sys_openat+0xc/0xe <- do_sys_open: ... After patching: <...>-1447 [001] 1038282.286875: myprobe: (do_sys_open+0x0/0xd6) ... <...>-1447 [001] 1038282.286878: myretprobe: (sys_openat+0xc/0xe <- do_sys_open) ... Signed-off-by: Masami Hiramatsu Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi LKML-Reference: <20090910235345.22412.76527.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 44dad1a..1746afe 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -902,10 +902,13 @@ print_kprobe_event(struct trace_iterator *iter, int flags) event = ftrace_find_event(field->ent.type); tp = container_of(event, struct trace_probe, event); + if (!trace_seq_printf(s, "%s: (", tp->call.name)) + goto partial; + if (!seq_print_ip_sym(s, field->ip, flags | TRACE_ITER_SYM_OFFSET)) goto partial; - if (!trace_seq_puts(s, ":")) + if (!trace_seq_puts(s, ")")) goto partial; for (i = 0; i < field->nargs; i++) @@ -934,6 +937,9 @@ print_kretprobe_event(struct trace_iterator *iter, int flags) event = ftrace_find_event(field->ent.type); tp = container_of(event, struct trace_probe, event); + if (!trace_seq_printf(s, "%s: (", tp->call.name)) + goto partial; + if (!seq_print_ip_sym(s, field->ret_ip, flags | TRACE_ITER_SYM_OFFSET)) goto partial; @@ -943,7 +949,7 @@ print_kretprobe_event(struct trace_iterator *iter, int flags) if (!seq_print_ip_sym(s, field->func, flags & ~TRACE_ITER_SYM_OFFSET)) goto partial; - if (!trace_seq_puts(s, ":")) + if (!trace_seq_puts(s, ")")) goto partial; for (i = 0; i < field->nargs; i++) @@ -1087,7 +1093,7 @@ static int kprobe_event_show_format(struct ftrace_event_call *call, SHOW_FIELD(unsigned long, args[i], tp->args[i].name); trace_seq_puts(s, "\n"); - return __probe_event_show_format(s, tp, "%lx:", "ip"); + return __probe_event_show_format(s, tp, "(%lx)", "REC->ip"); } static int kretprobe_event_show_format(struct ftrace_event_call *call, @@ -1106,8 +1112,8 @@ static int kretprobe_event_show_format(struct ftrace_event_call *call, SHOW_FIELD(unsigned long, args[i], tp->args[i].name); trace_seq_puts(s, "\n"); - return __probe_event_show_format(s, tp, "%lx <- %lx:", - "func, ret_ip"); + return __probe_event_show_format(s, tp, "(%lx <- %lx)", + "REC->func, REC->ret_ip"); } #ifdef CONFIG_EVENT_PROFILE -- cgit v1.1 From f52487e9c0041842eeb77c6c48774414b1cede08 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 10 Sep 2009 19:53:53 -0400 Subject: tracing/kprobes: Support custom subsystem for each kprobe event Support specifying a custom subsystem(group) for each kprobe event. This allows users to create new group to control several probes at once, or add events to existing groups as additional tracepoints. New synopsis: p[:[subsys/]event-name] KADDR|KSYM[+offs] [ARGS] Signed-off-by: Masami Hiramatsu Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi LKML-Reference: <20090910235353.22412.15149.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 33 +++++++++++++++++++++++++++------ 1 file changed, 27 insertions(+), 6 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 1746afe..cbc0870 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -36,6 +36,7 @@ #define MAX_TRACE_ARGS 128 #define MAX_ARGSTR_LEN 63 #define MAX_EVENT_NAME_LEN 64 +#define KPROBE_EVENT_SYSTEM "kprobes" /* currently, trace_kprobe only supports X86. */ @@ -265,7 +266,8 @@ static LIST_HEAD(probe_list); /* * Allocate new trace_probe and initialize it (including kprobes). */ -static struct trace_probe *alloc_trace_probe(const char *event, +static struct trace_probe *alloc_trace_probe(const char *group, + const char *event, void *addr, const char *symbol, unsigned long offs, @@ -298,9 +300,16 @@ static struct trace_probe *alloc_trace_probe(const char *event, if (!tp->call.name) goto error; + if (!group) + goto error; + tp->call.system = kstrdup(group, GFP_KERNEL); + if (!tp->call.system) + goto error; + INIT_LIST_HEAD(&tp->list); return tp; error: + kfree(tp->call.name); kfree(tp->symbol); kfree(tp); return ERR_PTR(-ENOMEM); @@ -322,6 +331,7 @@ static void free_trace_probe(struct trace_probe *tp) for (i = 0; i < tp->nr_args; i++) free_probe_arg(&tp->args[i]); + kfree(tp->call.system); kfree(tp->call.name); kfree(tp->symbol); kfree(tp); @@ -530,8 +540,8 @@ static int create_trace_probe(int argc, char **argv) { /* * Argument syntax: - * - Add kprobe: p[:EVENT] SYMBOL[+OFFS]|ADDRESS [FETCHARGS] - * - Add kretprobe: r[:EVENT] SYMBOL[+0] [FETCHARGS] + * - Add kprobe: p[:[GRP/]EVENT] KSYM[+OFFS]|KADDR [FETCHARGS] + * - Add kretprobe: r[:[GRP/]EVENT] KSYM[+0] [FETCHARGS] * Fetch args: * aN : fetch Nth of function argument. (N:0-) * rv : fetch return value @@ -549,7 +559,7 @@ static int create_trace_probe(int argc, char **argv) struct trace_probe *tp; int i, ret = 0; int is_return = 0; - char *symbol = NULL, *event = NULL, *arg = NULL; + char *symbol = NULL, *event = NULL, *arg = NULL, *group = NULL; unsigned long offset = 0; void *addr = NULL; char buf[MAX_EVENT_NAME_LEN]; @@ -566,6 +576,15 @@ static int create_trace_probe(int argc, char **argv) if (argv[0][1] == ':') { event = &argv[0][2]; + if (strchr(event, '/')) { + group = event; + event = strchr(group, '/') + 1; + event[-1] = '\0'; + if (strlen(group) == 0) { + pr_info("Group name is not specifiled\n"); + return -EINVAL; + } + } if (strlen(event) == 0) { pr_info("Event name is not specifiled\n"); return -EINVAL; @@ -592,6 +611,8 @@ static int create_trace_probe(int argc, char **argv) argc -= 2; argv += 2; /* setup a probe */ + if (!group) + group = KPROBE_EVENT_SYSTEM; if (!event) { /* Make a new event name */ if (symbol) @@ -602,7 +623,8 @@ static int create_trace_probe(int argc, char **argv) is_return ? 'r' : 'p', addr); event = buf; } - tp = alloc_trace_probe(event, addr, symbol, offset, argc, is_return); + tp = alloc_trace_probe(group, event, addr, symbol, offset, argc, + is_return); if (IS_ERR(tp)) return PTR_ERR(tp); @@ -1217,7 +1239,6 @@ static int register_probe_event(struct trace_probe *tp) int ret; /* Initialize ftrace_event_call */ - call->system = "kprobes"; if (probe_is_return(tp)) { tp->event.trace = print_kretprobe_event; call->raw_init = probe_event_raw_init; -- cgit v1.1 From 2d5e067edc4635ff7515bfa9ab3edb38bc344cab Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 14 Sep 2009 16:48:56 -0400 Subject: tracing/kprobes: Fix trace_probe registration order Fix trace_probe registration order. ftrace_event_call and ftrace_event must be registered before kprobe/kretprobe, because tracing/profiling handlers dereference the event-id. Signed-off-by: Masami Hiramatsu Acked-by: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Tom Zanussi LKML-Reference: <20090914204856.18779.52961.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 42 +++++++++++++++++++----------------------- 1 file changed, 19 insertions(+), 23 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index cbc0870..ea0db8e 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -347,20 +347,15 @@ static struct trace_probe *find_probe_event(const char *event) return NULL; } -static void __unregister_trace_probe(struct trace_probe *tp) +/* Unregister a trace_probe and probe_event: call with locking probe_lock */ +static void unregister_trace_probe(struct trace_probe *tp) { if (probe_is_return(tp)) unregister_kretprobe(&tp->rp); else unregister_kprobe(&tp->rp.kp); -} - -/* Unregister a trace_probe and probe_event: call with locking probe_lock */ -static void unregister_trace_probe(struct trace_probe *tp) -{ - unregister_probe_event(tp); - __unregister_trace_probe(tp); list_del(&tp->list); + unregister_probe_event(tp); } /* Register a trace_probe and probe_event */ @@ -371,6 +366,19 @@ static int register_trace_probe(struct trace_probe *tp) mutex_lock(&probe_lock); + /* register as an event */ + old_tp = find_probe_event(tp->call.name); + if (old_tp) { + /* delete old event */ + unregister_trace_probe(old_tp); + free_trace_probe(old_tp); + } + ret = register_probe_event(tp); + if (ret) { + pr_warning("Faild to register probe event(%d)\n", ret); + goto end; + } + if (probe_is_return(tp)) ret = register_kretprobe(&tp->rp); else @@ -384,21 +392,9 @@ static int register_trace_probe(struct trace_probe *tp) tp->rp.kp.addr); ret = -EINVAL; } - goto end; - } - /* register as an event */ - old_tp = find_probe_event(tp->call.name); - if (old_tp) { - /* delete old event */ - unregister_trace_probe(old_tp); - free_trace_probe(old_tp); - } - ret = register_probe_event(tp); - if (ret) { - pr_warning("Faild to register probe event(%d)\n", ret); - __unregister_trace_probe(tp); - } - list_add_tail(&tp->list, &probe_list); + unregister_probe_event(tp); + } else + list_add_tail(&tp->list, &probe_list); end: mutex_unlock(&probe_lock); return ret; -- cgit v1.1 From 588bebb74fe87270f94c2810652bd683d63c4b54 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Wed, 16 Sep 2009 11:42:55 -0400 Subject: ftrace: Fix trace_add_event_call() to initialize list Handle failure path in trace_add_event_call() to fix the below bug which occurred when I tried to add invalid event twice. Could not create debugfs 'kmalloc' directory Failed to register kprobe event: kmalloc Faild to register probe event(-1) ------------[ cut here ]------------ WARNING: at /home/mhiramat/ksrc/random-tracing/lib/list_debug.c:26 __list_add+0x27/0x5c() Hardware name: list_add corruption. next->prev should be prev (c07d78cc), but was 00001000. (next=d854236c). Modules linked in: sunrpc uinput virtio_net virtio_balloon i2c_piix4 pcspkr i2c_core virtio_blk virtio_pci virtio_ring virtio [last unloaded: scsi_wait_scan] Pid: 1394, comm: tee Not tainted 2.6.31-rc9 #51 Call Trace: [] warn_slowpath_common+0x65/0x7c [] ? __list_add+0x27/0x5c [] warn_slowpath_fmt+0x24/0x27 [] __list_add+0x27/0x5c [] list_add+0xa/0xc [] trace_add_event_call+0x60/0x97 [] command_trace_probe+0x42c/0x51b [] ? remove_wait_queue+0x22/0x27 [] ? __wake_up+0x32/0x3b [] probes_write+0xd4/0x10a [] ? probes_write+0x0/0x10a [] vfs_write+0x80/0xdf [] sys_write+0x3b/0x5d [] syscall_call+0x7/0xb ---[ end trace 2b962b5dc1fdc07d ]--- Signed-off-by: Masami Hiramatsu Acked-by: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Tom Zanussi LKML-Reference: <4AB1077F.6020107@redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_events.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c index ba34920..83cc2c0 100644 --- a/kernel/trace/trace_events.c +++ b/kernel/trace/trace_events.c @@ -1010,9 +1010,12 @@ static int __trace_add_event_call(struct ftrace_event_call *call) return -ENOENT; list_add(&call->list, &ftrace_events); - return event_create_dir(call, d_events, &ftrace_event_id_fops, + ret = event_create_dir(call, d_events, &ftrace_event_id_fops, &ftrace_enable_fops, &ftrace_event_filter_fops, &ftrace_event_format_fops); + if (ret < 0) + list_del(&call->list); + return ret; } /* Add an additional event_call dynamically */ -- cgit v1.1 From 4fead8e46fded93cc0d432ced774d9a3a8d21bad Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 14 Sep 2009 16:49:12 -0400 Subject: ftrace: Fix trace_remove_event_call() to lock trace_event_mutex Lock not only event_mutex but also trace_event_mutex in trace_remove_event_call() to protect __unregister_ftrace_event(). Signed-off-by: Masami Hiramatsu Acked-by: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Tom Zanussi LKML-Reference: <20090914204912.18779.68734.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_events.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'kernel') diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c index 83cc2c0..f85b0f1 100644 --- a/kernel/trace/trace_events.c +++ b/kernel/trace/trace_events.c @@ -1054,6 +1054,9 @@ static void remove_subsystem_dir(const char *name) } } +/* + * Must be called under locking both of event_mutex and trace_event_mutex. + */ static void __trace_remove_event_call(struct ftrace_event_call *call) { ftrace_event_enable_disable(call, 0); @@ -1070,7 +1073,9 @@ static void __trace_remove_event_call(struct ftrace_event_call *call) void trace_remove_event_call(struct ftrace_event_call *call) { mutex_lock(&event_mutex); + down_write(&trace_event_mutex); __trace_remove_event_call(call); + up_write(&trace_event_mutex); mutex_unlock(&event_mutex); } -- cgit v1.1 From 50d780560785b068c358675c5f0bf6c83b5c373e Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 14 Sep 2009 16:49:20 -0400 Subject: tracing/kprobes: Add probe handler dispatcher to support perf and ftrace concurrent use Add kprobe_dispatcher and kretprobe_dispatcher to dispatch event in both profile and tracing handlers. This allows simultaneous kprobe uses by ftrace and perf. Signed-off-by: Masami Hiramatsu Acked-by: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Tom Zanussi LKML-Reference: <20090914204920.18779.57555.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 85 +++++++++++++++++++++++++++++++++------------ 1 file changed, 63 insertions(+), 22 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index ea0db8e..70b632c 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -185,10 +185,15 @@ struct probe_arg { const char *name; }; +/* Flags for trace_probe */ +#define TP_FLAG_TRACE 1 +#define TP_FLAG_PROFILE 2 + struct trace_probe { struct list_head list; struct kretprobe rp; /* Use rp.kp for kprobe use */ unsigned long nhit; + unsigned int flags; /* For TP_FLAG_* */ const char *symbol; /* symbol name */ struct ftrace_event_call call; struct trace_event event; @@ -200,10 +205,6 @@ struct trace_probe { (offsetof(struct trace_probe, args) + \ (sizeof(struct probe_arg) * (n))) -static int kprobe_trace_func(struct kprobe *kp, struct pt_regs *regs); -static int kretprobe_trace_func(struct kretprobe_instance *ri, - struct pt_regs *regs); - static __kprobes int probe_is_return(struct trace_probe *tp) { return tp->rp.handler != NULL; @@ -263,6 +264,10 @@ static void unregister_probe_event(struct trace_probe *tp); static DEFINE_MUTEX(probe_lock); static LIST_HEAD(probe_list); +static int kprobe_dispatcher(struct kprobe *kp, struct pt_regs *regs); +static int kretprobe_dispatcher(struct kretprobe_instance *ri, + struct pt_regs *regs); + /* * Allocate new trace_probe and initialize it (including kprobes). */ @@ -288,11 +293,10 @@ static struct trace_probe *alloc_trace_probe(const char *group, } else tp->rp.kp.addr = addr; - /* Set handler here for checking whether this probe is return or not. */ if (is_return) - tp->rp.handler = kretprobe_trace_func; + tp->rp.handler = kretprobe_dispatcher; else - tp->rp.kp.pre_handler = kprobe_trace_func; + tp->rp.kp.pre_handler = kprobe_dispatcher; if (!event) goto error; @@ -379,6 +383,7 @@ static int register_trace_probe(struct trace_probe *tp) goto end; } + tp->flags = TP_FLAG_TRACE; if (probe_is_return(tp)) ret = register_kretprobe(&tp->rp); else @@ -987,23 +992,24 @@ static int probe_event_enable(struct ftrace_event_call *call) { struct trace_probe *tp = (struct trace_probe *)call->data; - if (probe_is_return(tp)) { - tp->rp.handler = kretprobe_trace_func; + tp->flags |= TP_FLAG_TRACE; + if (probe_is_return(tp)) return enable_kretprobe(&tp->rp); - } else { - tp->rp.kp.pre_handler = kprobe_trace_func; + else return enable_kprobe(&tp->rp.kp); - } } static void probe_event_disable(struct ftrace_event_call *call) { struct trace_probe *tp = (struct trace_probe *)call->data; - if (probe_is_return(tp)) - disable_kretprobe(&tp->rp); - else - disable_kprobe(&tp->rp.kp); + tp->flags &= ~TP_FLAG_TRACE; + if (!(tp->flags & (TP_FLAG_TRACE | TP_FLAG_PROFILE))) { + if (probe_is_return(tp)) + disable_kretprobe(&tp->rp); + else + disable_kprobe(&tp->rp.kp); + } } static int probe_event_raw_init(struct ftrace_event_call *event_call) @@ -1212,22 +1218,57 @@ static int probe_profile_enable(struct ftrace_event_call *call) if (atomic_inc_return(&call->profile_count)) return 0; - if (probe_is_return(tp)) { - tp->rp.handler = kretprobe_profile_func; + tp->flags |= TP_FLAG_PROFILE; + if (probe_is_return(tp)) return enable_kretprobe(&tp->rp); - } else { - tp->rp.kp.pre_handler = kprobe_profile_func; + else return enable_kprobe(&tp->rp.kp); - } } static void probe_profile_disable(struct ftrace_event_call *call) { + struct trace_probe *tp = (struct trace_probe *)call->data; + if (atomic_add_negative(-1, &call->profile_count)) - probe_event_disable(call); + tp->flags &= ~TP_FLAG_PROFILE; + + if (!(tp->flags & (TP_FLAG_TRACE | TP_FLAG_PROFILE))) { + if (probe_is_return(tp)) + disable_kretprobe(&tp->rp); + else + disable_kprobe(&tp->rp.kp); + } } +#endif /* CONFIG_EVENT_PROFILE */ + + +static __kprobes +int kprobe_dispatcher(struct kprobe *kp, struct pt_regs *regs) +{ + struct trace_probe *tp = container_of(kp, struct trace_probe, rp.kp); + if (tp->flags & TP_FLAG_TRACE) + kprobe_trace_func(kp, regs); +#ifdef CONFIG_EVENT_PROFILE + if (tp->flags & TP_FLAG_PROFILE) + kprobe_profile_func(kp, regs); #endif /* CONFIG_EVENT_PROFILE */ + return 0; /* We don't tweek kernel, so just return 0 */ +} + +static __kprobes +int kretprobe_dispatcher(struct kretprobe_instance *ri, struct pt_regs *regs) +{ + struct trace_probe *tp = container_of(ri->rp, struct trace_probe, rp); + + if (tp->flags & TP_FLAG_TRACE) + kretprobe_trace_func(ri, regs); +#ifdef CONFIG_EVENT_PROFILE + if (tp->flags & TP_FLAG_PROFILE) + kretprobe_profile_func(ri, regs); +#endif /* CONFIG_EVENT_PROFILE */ + return 0; /* We don't tweek kernel, so just return 0 */ +} static int register_probe_event(struct trace_probe *tp) { -- cgit v1.1 From 74ebb63e7cd25f6fb02a45fc2ea7735bce1217c9 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 14 Sep 2009 16:49:28 -0400 Subject: tracing/kprobes: Fix profiling alignment for perf_counter buffer Fix *probe_profile_func() to align buffer size, since perf_counter requires its buffer entries to be 8 bytes aligned. Signed-off-by: Masami Hiramatsu Acked-by: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Tom Zanussi LKML-Reference: <20090914204928.18779.60029.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 17 ++++++++++++----- 1 file changed, 12 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 70b632c..d8db935 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -1149,18 +1149,23 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp, struct trace_probe *tp = container_of(kp, struct trace_probe, rp.kp); struct ftrace_event_call *call = &tp->call; struct kprobe_trace_entry *entry; - int size, i, pc; + int size, __size, i, pc; unsigned long irq_flags; local_save_flags(irq_flags); pc = preempt_count(); - size = SIZEOF_KPROBE_TRACE_ENTRY(tp->nr_args); + __size = SIZEOF_KPROBE_TRACE_ENTRY(tp->nr_args); + size = ALIGN(__size + sizeof(u32), sizeof(u64)); + size -= sizeof(u32); do { char raw_data[size]; struct trace_entry *ent; - + /* + * Zero dead bytes from alignment to avoid stack leak + * to userspace + */ *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; entry = (struct kprobe_trace_entry *)raw_data; ent = &entry->ent; @@ -1183,13 +1188,15 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri, struct trace_probe *tp = container_of(ri->rp, struct trace_probe, rp); struct ftrace_event_call *call = &tp->call; struct kretprobe_trace_entry *entry; - int size, i, pc; + int size, __size, i, pc; unsigned long irq_flags; local_save_flags(irq_flags); pc = preempt_count(); - size = SIZEOF_KRETPROBE_TRACE_ENTRY(tp->nr_args); + __size = SIZEOF_KRETPROBE_TRACE_ENTRY(tp->nr_args); + size = ALIGN(__size + sizeof(u32), sizeof(u64)); + size -= sizeof(u32); do { char raw_data[size]; -- cgit v1.1 From 5a0d9050db4d1147722b42afef9011251b2651ee Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 14 Sep 2009 16:49:37 -0400 Subject: tracing/kprobes: Disable kprobe events by default after creation Disable newly created kprobe events by default, not to disturb another user using ftrace. "Disturb" means when someone is using ftrace and another user tries to use perf-tools, (in near future) if he defines new kprobe event via perf-tools, then new events will mess up the frace buffer. Fix this to allow proper and transparent kprobes events concurrent usage between ftrace users and perf users. Signed-off-by: Masami Hiramatsu Acked-by: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Tom Zanussi LKML-Reference: <20090914204937.18779.59422.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index d8db935..f6821f1 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -383,7 +383,7 @@ static int register_trace_probe(struct trace_probe *tp) goto end; } - tp->flags = TP_FLAG_TRACE; + tp->rp.kp.flags |= KPROBE_FLAG_DISABLED; if (probe_is_return(tp)) ret = register_kretprobe(&tp->rp); else @@ -1298,7 +1298,7 @@ static int register_probe_event(struct trace_probe *tp) call->id = register_ftrace_event(&tp->event); if (!call->id) return -ENODEV; - call->enabled = 1; + call->enabled = 0; call->regfunc = probe_event_enable; call->unregfunc = probe_event_disable; -- cgit v1.1 From 1f0ab40976460bc4673fa204ce917a725185d8f2 Mon Sep 17 00:00:00 2001 From: Ananth N Mavinakayanahalli Date: Tue, 15 Sep 2009 10:43:07 +0530 Subject: kprobes: Prevent re-registration of the same kprobe Prevent re-registration of the same kprobe. This situation, though unlikely, needs to be flagged since it can lead to a system crash if it's not handled. The core change itself is small, but the helper routine needed to be moved around a bit; hence the diffstat. Signed-off-by: Ananth N Mavinakayanahalli Acked-by: Masami Hiramatsu Cc: Jim Keniston Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: H. Peter Anvin Cc: Ingo Molnar Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Steven Rostedt Cc: Tom Zanussi LKML-Reference: <20090915051307.GB26458@in.ibm.com> Signed-off-by: Frederic Weisbecker --- kernel/kprobes.c | 58 +++++++++++++++++++++++++++++++++++++------------------- 1 file changed, 38 insertions(+), 20 deletions(-) (limited to 'kernel') diff --git a/kernel/kprobes.c b/kernel/kprobes.c index 00d01b0..b946761 100644 --- a/kernel/kprobes.c +++ b/kernel/kprobes.c @@ -676,6 +676,40 @@ static kprobe_opcode_t __kprobes *kprobe_addr(struct kprobe *p) return (kprobe_opcode_t *)(((char *)addr) + p->offset); } +/* Check passed kprobe is valid and return kprobe in kprobe_table. */ +static struct kprobe * __kprobes __get_valid_kprobe(struct kprobe *p) +{ + struct kprobe *old_p, *list_p; + + old_p = get_kprobe(p->addr); + if (unlikely(!old_p)) + return NULL; + + if (p != old_p) { + list_for_each_entry_rcu(list_p, &old_p->list, list) + if (list_p == p) + /* kprobe p is a valid probe */ + goto valid; + return NULL; + } +valid: + return old_p; +} + +/* Return error if the kprobe is being re-registered */ +static inline int check_kprobe_rereg(struct kprobe *p) +{ + int ret = 0; + struct kprobe *old_p; + + mutex_lock(&kprobe_mutex); + old_p = __get_valid_kprobe(p); + if (old_p) + ret = -EINVAL; + mutex_unlock(&kprobe_mutex); + return ret; +} + int __kprobes register_kprobe(struct kprobe *p) { int ret = 0; @@ -688,6 +722,10 @@ int __kprobes register_kprobe(struct kprobe *p) return -EINVAL; p->addr = addr; + ret = check_kprobe_rereg(p); + if (ret) + return ret; + preempt_disable(); if (!kernel_text_address((unsigned long) p->addr) || in_kprobes_functions((unsigned long) p->addr)) { @@ -757,26 +795,6 @@ out: } EXPORT_SYMBOL_GPL(register_kprobe); -/* Check passed kprobe is valid and return kprobe in kprobe_table. */ -static struct kprobe * __kprobes __get_valid_kprobe(struct kprobe *p) -{ - struct kprobe *old_p, *list_p; - - old_p = get_kprobe(p->addr); - if (unlikely(!old_p)) - return NULL; - - if (p != old_p) { - list_for_each_entry_rcu(list_p, &old_p->list, list) - if (list_p == p) - /* kprobe p is a valid probe */ - goto valid; - return NULL; - } -valid: - return old_p; -} - /* * Unregister a kprobe without a scheduler synchronization. */ -- cgit v1.1 From a1a138d05fa060ac4238c19a1e890aacc25ed3ba Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Fri, 25 Sep 2009 11:20:12 -0700 Subject: tracing/kprobes: Use global event perf buffers in kprobe tracer Use new percpu global event buffer instead of stack in kprobe tracer while tracing through perf. Signed-off-by: Masami Hiramatsu Acked-by: Steven Rostedt Acked-by: Ingo Molnar Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: H. Peter Anvin Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Tom Zanussi LKML-Reference: <20090925182011.10157.60140.stgit@omoto> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 115 ++++++++++++++++++++++++++++---------------- 1 file changed, 73 insertions(+), 42 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 09cba270..97309d4 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -1149,35 +1149,49 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp, struct trace_probe *tp = container_of(kp, struct trace_probe, rp.kp); struct ftrace_event_call *call = &tp->call; struct kprobe_trace_entry *entry; - int size, __size, i, pc; + struct trace_entry *ent; + int size, __size, i, pc, __cpu; unsigned long irq_flags; + char *raw_data; - local_save_flags(irq_flags); pc = preempt_count(); - __size = SIZEOF_KPROBE_TRACE_ENTRY(tp->nr_args); size = ALIGN(__size + sizeof(u32), sizeof(u64)); size -= sizeof(u32); + if (WARN_ONCE(size > FTRACE_MAX_PROFILE_SIZE, + "profile buffer not large enough")) + return 0; - do { - char raw_data[size]; - struct trace_entry *ent; - /* - * Zero dead bytes from alignment to avoid stack leak - * to userspace - */ - *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; - entry = (struct kprobe_trace_entry *)raw_data; - ent = &entry->ent; - - tracing_generic_entry_update(ent, irq_flags, pc); - ent->type = call->id; - entry->nargs = tp->nr_args; - entry->ip = (unsigned long)kp->addr; - for (i = 0; i < tp->nr_args; i++) - entry->args[i] = call_fetch(&tp->args[i].fetch, regs); - perf_tp_event(call->id, entry->ip, 1, entry, size); - } while (0); + /* + * Protect the non nmi buffer + * This also protects the rcu read side + */ + local_irq_save(irq_flags); + __cpu = smp_processor_id(); + + if (in_nmi()) + raw_data = rcu_dereference(trace_profile_buf_nmi); + else + raw_data = rcu_dereference(trace_profile_buf); + + if (!raw_data) + goto end; + + raw_data = per_cpu_ptr(raw_data, __cpu); + /* Zero dead bytes from alignment to avoid buffer leak to userspace */ + *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; + entry = (struct kprobe_trace_entry *)raw_data; + ent = &entry->ent; + + tracing_generic_entry_update(ent, irq_flags, pc); + ent->type = call->id; + entry->nargs = tp->nr_args; + entry->ip = (unsigned long)kp->addr; + for (i = 0; i < tp->nr_args; i++) + entry->args[i] = call_fetch(&tp->args[i].fetch, regs); + perf_tp_event(call->id, entry->ip, 1, entry, size); +end: + local_irq_restore(irq_flags); return 0; } @@ -1188,33 +1202,50 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri, struct trace_probe *tp = container_of(ri->rp, struct trace_probe, rp); struct ftrace_event_call *call = &tp->call; struct kretprobe_trace_entry *entry; - int size, __size, i, pc; + struct trace_entry *ent; + int size, __size, i, pc, __cpu; unsigned long irq_flags; + char *raw_data; - local_save_flags(irq_flags); pc = preempt_count(); - __size = SIZEOF_KRETPROBE_TRACE_ENTRY(tp->nr_args); size = ALIGN(__size + sizeof(u32), sizeof(u64)); size -= sizeof(u32); + if (WARN_ONCE(size > FTRACE_MAX_PROFILE_SIZE, + "profile buffer not large enough")) + return 0; + + /* + * Protect the non nmi buffer + * This also protects the rcu read side + */ + local_irq_save(irq_flags); + __cpu = smp_processor_id(); + + if (in_nmi()) + raw_data = rcu_dereference(trace_profile_buf_nmi); + else + raw_data = rcu_dereference(trace_profile_buf); + + if (!raw_data) + goto end; + + raw_data = per_cpu_ptr(raw_data, __cpu); + /* Zero dead bytes from alignment to avoid buffer leak to userspace */ + *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; + entry = (struct kretprobe_trace_entry *)raw_data; + ent = &entry->ent; - do { - char raw_data[size]; - struct trace_entry *ent; - - *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; - entry = (struct kretprobe_trace_entry *)raw_data; - ent = &entry->ent; - - tracing_generic_entry_update(ent, irq_flags, pc); - ent->type = call->id; - entry->nargs = tp->nr_args; - entry->func = (unsigned long)tp->rp.kp.addr; - entry->ret_ip = (unsigned long)ri->ret_addr; - for (i = 0; i < tp->nr_args; i++) - entry->args[i] = call_fetch(&tp->args[i].fetch, regs); - perf_tp_event(call->id, entry->ret_ip, 1, entry, size); - } while (0); + tracing_generic_entry_update(ent, irq_flags, pc); + ent->type = call->id; + entry->nargs = tp->nr_args; + entry->func = (unsigned long)tp->rp.kp.addr; + entry->ret_ip = (unsigned long)ri->ret_addr; + for (i = 0; i < tp->nr_args; i++) + entry->args[i] = call_fetch(&tp->args[i].fetch, regs); + perf_tp_event(call->id, entry->ret_ip, 1, entry, size); +end: + local_irq_restore(irq_flags); return 0; } -- cgit v1.1 From 88f70d7590538e427c8405a2e02ac2624847386c Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Fri, 25 Sep 2009 11:20:54 -0700 Subject: tracing/ftrace: Fix to check create_event_dir() when adding new events Check result of event_create_dir() and add ftrace_event_call to ftrace_events list only if it is succeeded. Thanks to Li for pointing it out. Signed-off-by: Masami Hiramatsu Acked-by: Steven Rostedt Acked-by: Ingo Molnar Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Andi Kleen Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: H. Peter Anvin Cc: Jason Baron Cc: K.Prasad Cc: Lai Jiangshan Cc: Li Zefan Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Tom Zanussi LKML-Reference: <20090925182054.10157.55219.stgit@omoto> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_events.c | 25 ++++++++++++++----------- 1 file changed, 14 insertions(+), 11 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c index a4b7c9a..155b5d5 100644 --- a/kernel/trace/trace_events.c +++ b/kernel/trace/trace_events.c @@ -957,12 +957,12 @@ static int __trace_add_event_call(struct ftrace_event_call *call) if (!d_events) return -ENOENT; - list_add(&call->list, &ftrace_events); ret = event_create_dir(call, d_events, &ftrace_event_id_fops, &ftrace_enable_fops, &ftrace_event_filter_fops, &ftrace_event_format_fops); - if (ret < 0) - list_del(&call->list); + if (!ret) + list_add(&call->list, &ftrace_events); + return ret; } @@ -1124,10 +1124,11 @@ static void trace_module_add_events(struct module *mod) return; } call->mod = mod; - list_add(&call->list, &ftrace_events); - event_create_dir(call, d_events, - &file_ops->id, &file_ops->enable, - &file_ops->filter, &file_ops->format); + ret = event_create_dir(call, d_events, + &file_ops->id, &file_ops->enable, + &file_ops->filter, &file_ops->format); + if (!ret) + list_add(&call->list, &ftrace_events); } } @@ -1267,10 +1268,12 @@ static __init int event_trace_init(void) continue; } } - list_add(&call->list, &ftrace_events); - event_create_dir(call, d_events, &ftrace_event_id_fops, - &ftrace_enable_fops, &ftrace_event_filter_fops, - &ftrace_event_format_fops); + ret = event_create_dir(call, d_events, &ftrace_event_id_fops, + &ftrace_enable_fops, + &ftrace_event_filter_fops, + &ftrace_event_format_fops); + if (!ret) + list_add(&call->list, &ftrace_events); } while (true) { -- cgit v1.1 From 26a50744b21fff65bd754874072857bee8967f4d Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Tue, 6 Oct 2009 01:09:50 -0500 Subject: tracing/events: Add 'signed' field to format files The sign info used for filters in the kernel is also useful to applications that process the trace stream. Add it to the format files and make it available to userspace. Signed-off-by: Tom Zanussi Acked-by: Frederic Weisbecker Cc: rostedt@goodmis.org Cc: lizf@cn.fujitsu.com Cc: hch@infradead.org Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: <1254809398-8078-2-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- kernel/trace/ring_buffer.c | 15 +++++++++------ kernel/trace/trace_events.c | 24 ++++++++++++------------ kernel/trace/trace_export.c | 25 ++++++++++++++----------- kernel/trace/trace_syscalls.c | 20 +++++++++++++------- 4 files changed, 48 insertions(+), 36 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index d4ff019..e43c928 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -397,18 +397,21 @@ int ring_buffer_print_page_header(struct trace_seq *s) int ret; ret = trace_seq_printf(s, "\tfield: u64 timestamp;\t" - "offset:0;\tsize:%u;\n", - (unsigned int)sizeof(field.time_stamp)); + "offset:0;\tsize:%u;\tsigned:%u;\n", + (unsigned int)sizeof(field.time_stamp), + (unsigned int)is_signed_type(u64)); ret = trace_seq_printf(s, "\tfield: local_t commit;\t" - "offset:%u;\tsize:%u;\n", + "offset:%u;\tsize:%u;\tsigned:%u;\n", (unsigned int)offsetof(typeof(field), commit), - (unsigned int)sizeof(field.commit)); + (unsigned int)sizeof(field.commit), + (unsigned int)is_signed_type(long)); ret = trace_seq_printf(s, "\tfield: char data;\t" - "offset:%u;\tsize:%u;\n", + "offset:%u;\tsize:%u;\tsigned:%u;\n", (unsigned int)offsetof(typeof(field), data), - (unsigned int)BUF_PAGE_SIZE); + (unsigned int)BUF_PAGE_SIZE, + (unsigned int)is_signed_type(char)); return ret; } diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c index d128f65..cf3cabf 100644 --- a/kernel/trace/trace_events.c +++ b/kernel/trace/trace_events.c @@ -507,7 +507,7 @@ extern char *__bad_type_size(void); #define FIELD(type, name) \ sizeof(type) != sizeof(field.name) ? __bad_type_size() : \ #type, "common_" #name, offsetof(typeof(field), name), \ - sizeof(field.name) + sizeof(field.name), is_signed_type(type) static int trace_write_header(struct trace_seq *s) { @@ -515,17 +515,17 @@ static int trace_write_header(struct trace_seq *s) /* struct trace_entry */ return trace_seq_printf(s, - "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\n" - "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\n" - "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\n" - "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\n" - "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\n" - "\n", - FIELD(unsigned short, type), - FIELD(unsigned char, flags), - FIELD(unsigned char, preempt_count), - FIELD(int, pid), - FIELD(int, lock_depth)); + "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\tsigned:%u;\n" + "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\tsigned:%u;\n" + "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\tsigned:%u;\n" + "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\tsigned:%u;\n" + "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\tsigned:%u;\n" + "\n", + FIELD(unsigned short, type), + FIELD(unsigned char, flags), + FIELD(unsigned char, preempt_count), + FIELD(int, pid), + FIELD(int, lock_depth)); } static ssize_t diff --git a/kernel/trace/trace_export.c b/kernel/trace/trace_export.c index 9753fcc..31da218 100644 --- a/kernel/trace/trace_export.c +++ b/kernel/trace/trace_export.c @@ -66,44 +66,47 @@ static void __used ____ftrace_check_##name(void) \ #undef __field #define __field(type, item) \ ret = trace_seq_printf(s, "\tfield:" #type " " #item ";\t" \ - "offset:%zu;\tsize:%zu;\n", \ + "offset:%zu;\tsize:%zu;\tsigned:%u;\n", \ offsetof(typeof(field), item), \ - sizeof(field.item)); \ + sizeof(field.item), is_signed_type(type)); \ if (!ret) \ return 0; #undef __field_desc #define __field_desc(type, container, item) \ ret = trace_seq_printf(s, "\tfield:" #type " " #item ";\t" \ - "offset:%zu;\tsize:%zu;\n", \ + "offset:%zu;\tsize:%zu;\tsigned:%u;\n", \ offsetof(typeof(field), container.item), \ - sizeof(field.container.item)); \ + sizeof(field.container.item), \ + is_signed_type(type)); \ if (!ret) \ return 0; #undef __array #define __array(type, item, len) \ ret = trace_seq_printf(s, "\tfield:" #type " " #item "[" #len "];\t" \ - "offset:%zu;\tsize:%zu;\n", \ - offsetof(typeof(field), item), \ - sizeof(field.item)); \ + "offset:%zu;\tsize:%zu;\tsigned:%u;\n", \ + offsetof(typeof(field), item), \ + sizeof(field.item), is_signed_type(type)); \ if (!ret) \ return 0; #undef __array_desc #define __array_desc(type, container, item, len) \ ret = trace_seq_printf(s, "\tfield:" #type " " #item "[" #len "];\t" \ - "offset:%zu;\tsize:%zu;\n", \ + "offset:%zu;\tsize:%zu;\tsigned:%u;\n", \ offsetof(typeof(field), container.item), \ - sizeof(field.container.item)); \ + sizeof(field.container.item), \ + is_signed_type(type)); \ if (!ret) \ return 0; #undef __dynamic_array #define __dynamic_array(type, item) \ ret = trace_seq_printf(s, "\tfield:" #type " " #item ";\t" \ - "offset:%zu;\tsize:0;\n", \ - offsetof(typeof(field), item)); \ + "offset:%zu;\tsize:0;\tsigned:%u;\n", \ + offsetof(typeof(field), item), \ + is_signed_type(type)); \ if (!ret) \ return 0; diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index 527e17e..d99abc4 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -103,7 +103,8 @@ extern char *__bad_type_size(void); #define SYSCALL_FIELD(type, name) \ sizeof(type) != sizeof(trace.name) ? \ __bad_type_size() : \ - #type, #name, offsetof(typeof(trace), name), sizeof(trace.name) + #type, #name, offsetof(typeof(trace), name), \ + sizeof(trace.name), is_signed_type(type) int syscall_enter_format(struct ftrace_event_call *call, struct trace_seq *s) { @@ -120,7 +121,8 @@ int syscall_enter_format(struct ftrace_event_call *call, struct trace_seq *s) if (!entry) return 0; - ret = trace_seq_printf(s, "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\n", + ret = trace_seq_printf(s, "\tfield:%s %s;\toffset:%zu;\tsize:%zu;" + "\tsigned:%u;\n", SYSCALL_FIELD(int, nr)); if (!ret) return 0; @@ -130,8 +132,10 @@ int syscall_enter_format(struct ftrace_event_call *call, struct trace_seq *s) entry->args[i]); if (!ret) return 0; - ret = trace_seq_printf(s, "\toffset:%d;\tsize:%zu;\n", offset, - sizeof(unsigned long)); + ret = trace_seq_printf(s, "\toffset:%d;\tsize:%zu;" + "\tsigned:%u;\n", offset, + sizeof(unsigned long), + is_signed_type(unsigned long)); if (!ret) return 0; offset += sizeof(unsigned long); @@ -163,8 +167,10 @@ int syscall_exit_format(struct ftrace_event_call *call, struct trace_seq *s) struct syscall_trace_exit trace; ret = trace_seq_printf(s, - "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\n" - "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\n", + "\tfield:%s %s;\toffset:%zu;\tsize:%zu;" + "\tsigned:%u;\n" + "\tfield:%s %s;\toffset:%zu;\tsize:%zu;" + "\tsigned:%u;\n", SYSCALL_FIELD(int, nr), SYSCALL_FIELD(long, ret)); if (!ret) @@ -212,7 +218,7 @@ int syscall_exit_define_fields(struct ftrace_event_call *call) if (ret) return ret; - ret = trace_define_field(call, SYSCALL_FIELD(long, ret), 0, + ret = trace_define_field(call, SYSCALL_FIELD(long, ret), FILTER_OTHER); return ret; -- cgit v1.1 From 405b2651e4bedf8d3932b64cad649b4d26b067f5 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Wed, 7 Oct 2009 18:27:40 -0400 Subject: tracing/kprobes: Add $ prefix to special variables Add $ prefix to the special variables(e.g. sa, rv) of kprobe-tracer. This resolves consistency issues between kprobe_events and perf-kprobe. The main goal is to avoid conflicts between local variable names of probed functions, used by perf probe, and special variables used in the kprobe event creation interface (stack values, etc...) and also available from perf probe. ie: we don't want rv (return value) to conflict with a local variable named rv in a probed function. Signed-off-by: Masami Hiramatsu Cc: Ingo Molnar Cc: Thomas Gleixner Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Christoph Hellwig Cc: Ananth N Mavinakayanahalli Cc: Jim Keniston Cc: Frank Ch. Eigler LKML-Reference: <20091007222740.1684.91170.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 60 ++++++++++++++++++++++++++++----------------- 1 file changed, 37 insertions(+), 23 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 97309d4..f63ead0 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -220,24 +220,24 @@ static int probe_arg_string(char *buf, size_t n, struct fetch_func *ff) int ret = -EINVAL; if (ff->func == fetch_argument) - ret = snprintf(buf, n, "a%lu", (unsigned long)ff->data); + ret = snprintf(buf, n, "$a%lu", (unsigned long)ff->data); else if (ff->func == fetch_register) { const char *name; name = regs_query_register_name((unsigned int)((long)ff->data)); ret = snprintf(buf, n, "%%%s", name); } else if (ff->func == fetch_stack) - ret = snprintf(buf, n, "s%lu", (unsigned long)ff->data); + ret = snprintf(buf, n, "$s%lu", (unsigned long)ff->data); else if (ff->func == fetch_memory) ret = snprintf(buf, n, "@0x%p", ff->data); else if (ff->func == fetch_symbol) { struct symbol_cache *sc = ff->data; ret = snprintf(buf, n, "@%s%+ld", sc->symbol, sc->offset); } else if (ff->func == fetch_retvalue) - ret = snprintf(buf, n, "rv"); + ret = snprintf(buf, n, "$rv"); else if (ff->func == fetch_ip) - ret = snprintf(buf, n, "ra"); + ret = snprintf(buf, n, "$ra"); else if (ff->func == fetch_stack_address) - ret = snprintf(buf, n, "sa"); + ret = snprintf(buf, n, "$sa"); else if (ff->func == fetch_indirect) { struct indirect_fetch_data *id = ff->data; size_t l = 0; @@ -429,12 +429,10 @@ static int split_symbol_offset(char *symbol, unsigned long *offset) #define PARAM_MAX_ARGS 16 #define PARAM_MAX_STACK (THREAD_SIZE / sizeof(unsigned long)) -static int parse_probe_arg(char *arg, struct fetch_func *ff, int is_return) +static int parse_probe_vars(char *arg, struct fetch_func *ff, int is_return) { int ret = 0; unsigned long param; - long offset; - char *tmp; switch (arg[0]) { case 'a': /* argument */ @@ -456,14 +454,6 @@ static int parse_probe_arg(char *arg, struct fetch_func *ff, int is_return) } else ret = -EINVAL; break; - case '%': /* named register */ - ret = regs_query_register_offset(arg + 1); - if (ret >= 0) { - ff->func = fetch_register; - ff->data = (void *)(unsigned long)ret; - ret = 0; - } - break; case 's': /* stack */ if (arg[1] == 'a') { ff->func = fetch_stack_address; @@ -478,6 +468,31 @@ static int parse_probe_arg(char *arg, struct fetch_func *ff, int is_return) } } break; + default: + ret = -EINVAL; + } + return ret; +} + +static int parse_probe_arg(char *arg, struct fetch_func *ff, int is_return) +{ + int ret = 0; + unsigned long param; + long offset; + char *tmp; + + switch (arg[0]) { + case '$': + ret = parse_probe_vars(arg + 1, ff, is_return); + break; + case '%': /* named register */ + ret = regs_query_register_offset(arg + 1); + if (ret >= 0) { + ff->func = fetch_register; + ff->data = (void *)(unsigned long)ret; + ret = 0; + } + break; case '@': /* memory or symbol */ if (isdigit(arg[1])) { ret = strict_strtoul(arg + 1, 0, ¶m); @@ -489,8 +504,7 @@ static int parse_probe_arg(char *arg, struct fetch_func *ff, int is_return) ret = split_symbol_offset(arg + 1, &offset); if (ret) break; - ff->data = alloc_symbol_cache(arg + 1, - offset); + ff->data = alloc_symbol_cache(arg + 1, offset); if (ff->data) ff->func = fetch_symbol; else @@ -544,11 +558,11 @@ static int create_trace_probe(int argc, char **argv) * - Add kprobe: p[:[GRP/]EVENT] KSYM[+OFFS]|KADDR [FETCHARGS] * - Add kretprobe: r[:[GRP/]EVENT] KSYM[+0] [FETCHARGS] * Fetch args: - * aN : fetch Nth of function argument. (N:0-) - * rv : fetch return value - * ra : fetch return address - * sa : fetch stack address - * sN : fetch Nth of stack (N:0-) + * $aN : fetch Nth of function argument. (N:0-) + * $rv : fetch return value + * $ra : fetch return address + * $sa : fetch stack address + * $sN : fetch Nth of stack (N:0-) * @ADDR : fetch memory at ADDR (ADDR should be in kernel) * @SYM[+|-offs] : fetch memory at SYM +|- offs (SYM is a data symbol) * %REG : fetch register REG -- cgit v1.1 From 99329c44f28a1b7ac83beebfb4319e612042e319 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Wed, 7 Oct 2009 18:27:48 -0400 Subject: tracing/kprobes: Remove '$ra' special variable Remove '$ra' (return address) because it is already shown at the head of each entry. Signed-off-by: Masami Hiramatsu Cc: Ingo Molnar Cc: Thomas Gleixner Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Christoph Hellwig Cc: Ananth N Mavinakayanahalli Cc: Jim Keniston Cc: Frank Ch. Eigler LKML-Reference: <20091007222748.1684.12711.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 11 ----------- 1 file changed, 11 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index f63ead0..ba6d3bd 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -85,11 +85,6 @@ static __kprobes unsigned long fetch_retvalue(struct pt_regs *regs, return regs_return_value(regs); } -static __kprobes unsigned long fetch_ip(struct pt_regs *regs, void *dummy) -{ - return instruction_pointer(regs); -} - static __kprobes unsigned long fetch_stack_address(struct pt_regs *regs, void *dummy) { @@ -234,8 +229,6 @@ static int probe_arg_string(char *buf, size_t n, struct fetch_func *ff) ret = snprintf(buf, n, "@%s%+ld", sc->symbol, sc->offset); } else if (ff->func == fetch_retvalue) ret = snprintf(buf, n, "$rv"); - else if (ff->func == fetch_ip) - ret = snprintf(buf, n, "$ra"); else if (ff->func == fetch_stack_address) ret = snprintf(buf, n, "$sa"); else if (ff->func == fetch_indirect) { @@ -448,9 +441,6 @@ static int parse_probe_vars(char *arg, struct fetch_func *ff, int is_return) if (is_return && arg[1] == 'v') { ff->func = fetch_retvalue; ff->data = NULL; - } else if (is_return && arg[1] == 'a') { - ff->func = fetch_ip; - ff->data = NULL; } else ret = -EINVAL; break; @@ -560,7 +550,6 @@ static int create_trace_probe(int argc, char **argv) * Fetch args: * $aN : fetch Nth of function argument. (N:0-) * $rv : fetch return value - * $ra : fetch return address * $sa : fetch stack address * $sN : fetch Nth of stack (N:0-) * @ADDR : fetch memory at ADDR (ADDR should be in kernel) -- cgit v1.1 From 2e06ff6389aedafc4a3a374344ac70672252f9b5 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Wed, 7 Oct 2009 18:27:59 -0400 Subject: tracing/kprobes: Make special variable names more self-explainable Rename special variables to more self-explainable names as below: - $rv to $retval - $sa to $stack - $aN to $argN - $sN to $stackN Signed-off-by: Masami Hiramatsu Cc: Ingo Molnar Cc: Thomas Gleixner Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Christoph Hellwig Cc: Ananth N Mavinakayanahalli Cc: Jim Keniston Cc: Frank Ch. Eigler LKML-Reference: <20091007222759.1684.3319.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 52 +++++++++++++++++++++------------------------ 1 file changed, 24 insertions(+), 28 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index ba6d3bd..3313fa7 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -215,22 +215,22 @@ static int probe_arg_string(char *buf, size_t n, struct fetch_func *ff) int ret = -EINVAL; if (ff->func == fetch_argument) - ret = snprintf(buf, n, "$a%lu", (unsigned long)ff->data); + ret = snprintf(buf, n, "$arg%lu", (unsigned long)ff->data); else if (ff->func == fetch_register) { const char *name; name = regs_query_register_name((unsigned int)((long)ff->data)); ret = snprintf(buf, n, "%%%s", name); } else if (ff->func == fetch_stack) - ret = snprintf(buf, n, "$s%lu", (unsigned long)ff->data); + ret = snprintf(buf, n, "$stack%lu", (unsigned long)ff->data); else if (ff->func == fetch_memory) ret = snprintf(buf, n, "@0x%p", ff->data); else if (ff->func == fetch_symbol) { struct symbol_cache *sc = ff->data; ret = snprintf(buf, n, "@%s%+ld", sc->symbol, sc->offset); } else if (ff->func == fetch_retvalue) - ret = snprintf(buf, n, "$rv"); + ret = snprintf(buf, n, "$retval"); else if (ff->func == fetch_stack_address) - ret = snprintf(buf, n, "$sa"); + ret = snprintf(buf, n, "$stack"); else if (ff->func == fetch_indirect) { struct indirect_fetch_data *id = ff->data; size_t l = 0; @@ -427,40 +427,36 @@ static int parse_probe_vars(char *arg, struct fetch_func *ff, int is_return) int ret = 0; unsigned long param; - switch (arg[0]) { - case 'a': /* argument */ - ret = strict_strtoul(arg + 1, 10, ¶m); - if (ret || param > PARAM_MAX_ARGS) - ret = -EINVAL; - else { - ff->func = fetch_argument; - ff->data = (void *)param; - } - break; - case 'r': /* retval or retaddr */ - if (is_return && arg[1] == 'v') { + if (strcmp(arg, "retval") == 0) { + if (is_return) { ff->func = fetch_retvalue; ff->data = NULL; } else ret = -EINVAL; - break; - case 's': /* stack */ - if (arg[1] == 'a') { + } else if (strncmp(arg, "stack", 5) == 0) { + if (arg[5] == '\0') { ff->func = fetch_stack_address; ff->data = NULL; - } else { - ret = strict_strtoul(arg + 1, 10, ¶m); + } else if (isdigit(arg[5])) { + ret = strict_strtoul(arg + 5, 10, ¶m); if (ret || param > PARAM_MAX_STACK) ret = -EINVAL; else { ff->func = fetch_stack; ff->data = (void *)param; } + } else + ret = -EINVAL; + } else if (strncmp(arg, "arg", 3) == 0 && isdigit(arg[3])) { + ret = strict_strtoul(arg + 3, 10, ¶m); + if (ret || param > PARAM_MAX_ARGS) + ret = -EINVAL; + else { + ff->func = fetch_argument; + ff->data = (void *)param; } - break; - default: + } else ret = -EINVAL; - } return ret; } @@ -548,10 +544,10 @@ static int create_trace_probe(int argc, char **argv) * - Add kprobe: p[:[GRP/]EVENT] KSYM[+OFFS]|KADDR [FETCHARGS] * - Add kretprobe: r[:[GRP/]EVENT] KSYM[+0] [FETCHARGS] * Fetch args: - * $aN : fetch Nth of function argument. (N:0-) - * $rv : fetch return value - * $sa : fetch stack address - * $sN : fetch Nth of stack (N:0-) + * $argN : fetch Nth of function argument. (N:0-) + * $retval : fetch return value + * $stack : fetch stack address + * $stackN : fetch Nth of stack (N:0-) * @ADDR : fetch memory at ADDR (ADDR should be in kernel) * @SYM[+|-offs] : fetch memory at SYM +|- offs (SYM is a data symbol) * %REG : fetch register REG -- cgit v1.1 From a703d946e883d8e447d0597de556e2effd110372 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Wed, 7 Oct 2009 18:28:07 -0400 Subject: tracing/kprobes: Avoid field name confliction Check whether the argument name is in conflict with other field names while creating a kprobe through the debugfs interface. Changes in v3: - Check strcmp() == 0 instead of !strcmp(). Changes in v2: - Add common_lock_depth to reserved name list. Signed-off-by: Masami Hiramatsu Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Thomas Gleixner Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Christoph Hellwig Cc: Ananth N Mavinakayanahalli Cc: Jim Keniston Cc: Frank Ch. Eigler LKML-Reference: <20091007222807.1684.26880.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 65 ++++++++++++++++++++++++++++++++++++--------- 1 file changed, 53 insertions(+), 12 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 3313fa7..bb6cb2b 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -38,6 +38,25 @@ #define MAX_EVENT_NAME_LEN 64 #define KPROBE_EVENT_SYSTEM "kprobes" +/* Reserved field names */ +#define FIELD_STRING_IP "ip" +#define FIELD_STRING_NARGS "nargs" +#define FIELD_STRING_RETIP "ret_ip" +#define FIELD_STRING_FUNC "func" + +const char *reserved_field_names[] = { + "common_type", + "common_flags", + "common_preempt_count", + "common_pid", + "common_tgid", + "common_lock_depth", + FIELD_STRING_IP, + FIELD_STRING_NARGS, + FIELD_STRING_RETIP, + FIELD_STRING_FUNC, +}; + /* currently, trace_kprobe only supports X86. */ struct fetch_func { @@ -537,6 +556,20 @@ static int parse_probe_arg(char *arg, struct fetch_func *ff, int is_return) return ret; } +/* Return 1 if name is reserved or already used by another argument */ +static int conflict_field_name(const char *name, + struct probe_arg *args, int narg) +{ + int i; + for (i = 0; i < ARRAY_SIZE(reserved_field_names); i++) + if (strcmp(reserved_field_names[i], name) == 0) + return 1; + for (i = 0; i < narg; i++) + if (strcmp(args[i].name, name) == 0) + return 1; + return 0; +} + static int create_trace_probe(int argc, char **argv) { /* @@ -637,6 +670,12 @@ static int create_trace_probe(int argc, char **argv) *arg++ = '\0'; else arg = argv[i]; + + if (conflict_field_name(argv[i], tp->args, i)) { + ret = -EINVAL; + goto error; + } + tp->args[i].name = kstrdup(argv[i], GFP_KERNEL); /* Parse fetch argument */ @@ -1039,8 +1078,8 @@ static int kprobe_event_define_fields(struct ftrace_event_call *event_call) if (!ret) return ret; - DEFINE_FIELD(unsigned long, ip, "ip", 0); - DEFINE_FIELD(int, nargs, "nargs", 1); + DEFINE_FIELD(unsigned long, ip, FIELD_STRING_IP, 0); + DEFINE_FIELD(int, nargs, FIELD_STRING_NARGS, 1); /* Set argument names as fields */ for (i = 0; i < tp->nr_args; i++) DEFINE_FIELD(unsigned long, args[i], tp->args[i].name, 0); @@ -1057,9 +1096,9 @@ static int kretprobe_event_define_fields(struct ftrace_event_call *event_call) if (!ret) return ret; - DEFINE_FIELD(unsigned long, func, "func", 0); - DEFINE_FIELD(unsigned long, ret_ip, "ret_ip", 0); - DEFINE_FIELD(int, nargs, "nargs", 1); + DEFINE_FIELD(unsigned long, func, FIELD_STRING_FUNC, 0); + DEFINE_FIELD(unsigned long, ret_ip, FIELD_STRING_RETIP, 0); + DEFINE_FIELD(int, nargs, FIELD_STRING_NARGS, 1); /* Set argument names as fields */ for (i = 0; i < tp->nr_args; i++) DEFINE_FIELD(unsigned long, args[i], tp->args[i].name, 0); @@ -1108,15 +1147,16 @@ static int kprobe_event_show_format(struct ftrace_event_call *call, int ret, i; struct trace_probe *tp = (struct trace_probe *)call->data; - SHOW_FIELD(unsigned long, ip, "ip"); - SHOW_FIELD(int, nargs, "nargs"); + SHOW_FIELD(unsigned long, ip, FIELD_STRING_IP); + SHOW_FIELD(int, nargs, FIELD_STRING_NARGS); /* Show fields */ for (i = 0; i < tp->nr_args; i++) SHOW_FIELD(unsigned long, args[i], tp->args[i].name); trace_seq_puts(s, "\n"); - return __probe_event_show_format(s, tp, "(%lx)", "REC->ip"); + return __probe_event_show_format(s, tp, "(%lx)", + "REC->" FIELD_STRING_IP); } static int kretprobe_event_show_format(struct ftrace_event_call *call, @@ -1126,9 +1166,9 @@ static int kretprobe_event_show_format(struct ftrace_event_call *call, int ret, i; struct trace_probe *tp = (struct trace_probe *)call->data; - SHOW_FIELD(unsigned long, func, "func"); - SHOW_FIELD(unsigned long, ret_ip, "ret_ip"); - SHOW_FIELD(int, nargs, "nargs"); + SHOW_FIELD(unsigned long, func, FIELD_STRING_FUNC); + SHOW_FIELD(unsigned long, ret_ip, FIELD_STRING_RETIP); + SHOW_FIELD(int, nargs, FIELD_STRING_NARGS); /* Show fields */ for (i = 0; i < tp->nr_args; i++) @@ -1136,7 +1176,8 @@ static int kretprobe_event_show_format(struct ftrace_event_call *call, trace_seq_puts(s, "\n"); return __probe_event_show_format(s, tp, "(%lx <- %lx)", - "REC->func, REC->ret_ip"); + "REC->" FIELD_STRING_FUNC + ", REC->" FIELD_STRING_RETIP); } #ifdef CONFIG_EVENT_PROFILE -- cgit v1.1 From e93f4d8539d5e9dd59f4af9d8ef4e9b62cfa1f81 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Wed, 7 Oct 2009 18:28:14 -0400 Subject: tracing/kprobes: Robustify fixed field names against variable field names conflicts Rename probe-common fixed field names to harder conflictable names, because current 'ip', 'func', and other probe field names are easily in conflict with user-specified variable names. Signed-off-by: Masami Hiramatsu Cc: Ingo Molnar Cc: Thomas Gleixner Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Christoph Hellwig Cc: Ananth N Mavinakayanahalli Cc: Jim Keniston Cc: Frank Ch. Eigler LKML-Reference: <20091007222814.1684.407.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_kprobe.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index bb6cb2b..739f70e 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -39,10 +39,10 @@ #define KPROBE_EVENT_SYSTEM "kprobes" /* Reserved field names */ -#define FIELD_STRING_IP "ip" -#define FIELD_STRING_NARGS "nargs" -#define FIELD_STRING_RETIP "ret_ip" -#define FIELD_STRING_FUNC "func" +#define FIELD_STRING_IP "__probe_ip" +#define FIELD_STRING_NARGS "__probe_nargs" +#define FIELD_STRING_RETIP "__probe_ret_ip" +#define FIELD_STRING_FUNC "__probe_func" const char *reserved_field_names[] = { "common_type", -- cgit v1.1 From fce29d15b59245597f7f320db4a9f2be0f5fb512 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Thu, 15 Oct 2009 11:20:34 +0800 Subject: tracing/filters: Refactor subsystem filter code Change: for_each_pred for_each_subsystem To: for_each_subsystem for_each_pred This change also prepares for later patches. Signed-off-by: Li Zefan Acked-by: Peter Zijlstra Acked-by: Frederic Weisbecker Cc: Steven Rostedt Cc: Tom Zanussi LKML-Reference: <4AD69502.8060903@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace.h | 1 - kernel/trace/trace_events_filter.c | 124 ++++++++++++++----------------------- 2 files changed, 45 insertions(+), 80 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index acef8b4..6286145 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -683,7 +683,6 @@ struct event_filter { int n_preds; struct filter_pred **preds; char *filter_string; - bool no_reset; }; struct event_subsystem { diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c index 9267201..9c4f9a0 100644 --- a/kernel/trace/trace_events_filter.c +++ b/kernel/trace/trace_events_filter.c @@ -615,14 +615,7 @@ static int init_subsystem_preds(struct event_subsystem *system) return 0; } -enum { - FILTER_DISABLE_ALL, - FILTER_INIT_NO_RESET, - FILTER_SKIP_NO_RESET, -}; - -static void filter_free_subsystem_preds(struct event_subsystem *system, - int flag) +static void filter_free_subsystem_preds(struct event_subsystem *system) { struct ftrace_event_call *call; @@ -633,14 +626,6 @@ static void filter_free_subsystem_preds(struct event_subsystem *system, if (strcmp(call->system, system->name) != 0) continue; - if (flag == FILTER_INIT_NO_RESET) { - call->filter->no_reset = false; - continue; - } - - if (flag == FILTER_SKIP_NO_RESET && call->filter->no_reset) - continue; - filter_disable_preds(call); remove_filter_string(call->filter); } @@ -817,44 +802,6 @@ add_pred_fn: return 0; } -static int filter_add_subsystem_pred(struct filter_parse_state *ps, - struct event_subsystem *system, - struct filter_pred *pred, - char *filter_string, - bool dry_run) -{ - struct ftrace_event_call *call; - int err = 0; - bool fail = true; - - list_for_each_entry(call, &ftrace_events, list) { - - if (!call->define_fields) - continue; - - if (strcmp(call->system, system->name)) - continue; - - if (call->filter->no_reset) - continue; - - err = filter_add_pred(ps, call, pred, dry_run); - if (err) - call->filter->no_reset = true; - else - fail = false; - - if (!dry_run) - replace_filter_string(call->filter, filter_string); - } - - if (fail) { - parse_error(ps, FILT_ERR_BAD_SUBSYS_FILTER, 0); - return err; - } - return 0; -} - static void parse_init(struct filter_parse_state *ps, struct filter_op *ops, char *infix_string) @@ -1209,8 +1156,7 @@ static int check_preds(struct filter_parse_state *ps) return 0; } -static int replace_preds(struct event_subsystem *system, - struct ftrace_event_call *call, +static int replace_preds(struct ftrace_event_call *call, struct filter_parse_state *ps, char *filter_string, bool dry_run) @@ -1257,11 +1203,7 @@ static int replace_preds(struct event_subsystem *system, add_pred: if (!pred) return -ENOMEM; - if (call) - err = filter_add_pred(ps, call, pred, false); - else - err = filter_add_subsystem_pred(ps, system, pred, - filter_string, dry_run); + err = filter_add_pred(ps, call, pred, dry_run); filter_free_pred(pred); if (err) return err; @@ -1272,6 +1214,44 @@ add_pred: return 0; } +static int replace_system_preds(struct event_subsystem *system, + struct filter_parse_state *ps, + char *filter_string) +{ + struct ftrace_event_call *call; + int err; + bool fail = true; + + list_for_each_entry(call, &ftrace_events, list) { + + if (!call->define_fields) + continue; + + if (strcmp(call->system, system->name) != 0) + continue; + + /* try to see if the filter can be applied */ + err = replace_preds(call, ps, filter_string, true); + if (err) + continue; + + /* really apply the filter */ + filter_disable_preds(call); + err = replace_preds(call, ps, filter_string, false); + if (err) + filter_disable_preds(call); + else + replace_filter_string(call->filter, filter_string); + fail = false; + } + + if (fail) { + parse_error(ps, FILT_ERR_BAD_SUBSYS_FILTER, 0); + return err; + } + return 0; +} + int apply_event_filter(struct ftrace_event_call *call, char *filter_string) { int err; @@ -1306,7 +1286,7 @@ int apply_event_filter(struct ftrace_event_call *call, char *filter_string) goto out; } - err = replace_preds(NULL, call, ps, filter_string, false); + err = replace_preds(call, ps, filter_string, false); if (err) append_filter_err(ps, call->filter); @@ -1334,7 +1314,7 @@ int apply_subsystem_event_filter(struct event_subsystem *system, goto out_unlock; if (!strcmp(strstrip(filter_string), "0")) { - filter_free_subsystem_preds(system, FILTER_DISABLE_ALL); + filter_free_subsystem_preds(system); remove_filter_string(system->filter); mutex_unlock(&event_mutex); return 0; @@ -1354,23 +1334,9 @@ int apply_subsystem_event_filter(struct event_subsystem *system, goto out; } - filter_free_subsystem_preds(system, FILTER_INIT_NO_RESET); - - /* try to see the filter can be applied to which events */ - err = replace_preds(system, NULL, ps, filter_string, true); - if (err) { - append_filter_err(ps, system->filter); - goto out; - } - - filter_free_subsystem_preds(system, FILTER_SKIP_NO_RESET); - - /* really apply the filter to the events */ - err = replace_preds(system, NULL, ps, filter_string, false); - if (err) { + err = replace_system_preds(system, ps, filter_string); + if (err) append_filter_err(ps, system->filter); - filter_free_subsystem_preds(system, 2); - } out: filter_opstack_clear(ps); -- cgit v1.1 From b0f1a59a98d7ac2102e7e4f22904c26d564a5628 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Thu, 15 Oct 2009 11:21:12 +0800 Subject: tracing/filters: Use a different op for glob match "==" will always do a full match, and "~" will do a glob match. In the future, we may add "=~" for regex match. Signed-off-by: Li Zefan Acked-by: Peter Zijlstra Acked-by: Frederic Weisbecker Cc: Steven Rostedt Cc: Tom Zanussi LKML-Reference: <4AD69528.3050309@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace.h | 2 +- kernel/trace/trace_events_filter.c | 59 ++++++++++++++++++-------------------- 2 files changed, 29 insertions(+), 32 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 6286145..ffe53dd 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -702,7 +702,7 @@ typedef int (*filter_pred_fn_t) (struct filter_pred *pred, void *event, typedef int (*regex_match_func)(char *str, struct regex *r, int len); enum regex_type { - MATCH_FULL, + MATCH_FULL = 0, MATCH_FRONT_ONLY, MATCH_MIDDLE_ONLY, MATCH_END_ONLY, diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c index 9c4f9a0..273845f 100644 --- a/kernel/trace/trace_events_filter.c +++ b/kernel/trace/trace_events_filter.c @@ -29,6 +29,7 @@ enum filter_op_ids { OP_OR, OP_AND, + OP_GLOB, OP_NE, OP_EQ, OP_LT, @@ -46,16 +47,17 @@ struct filter_op { }; static struct filter_op filter_ops[] = { - { OP_OR, "||", 1 }, - { OP_AND, "&&", 2 }, - { OP_NE, "!=", 4 }, - { OP_EQ, "==", 4 }, - { OP_LT, "<", 5 }, - { OP_LE, "<=", 5 }, - { OP_GT, ">", 5 }, - { OP_GE, ">=", 5 }, - { OP_NONE, "OP_NONE", 0 }, - { OP_OPEN_PAREN, "(", 0 }, + { OP_OR, "||", 1 }, + { OP_AND, "&&", 2 }, + { OP_GLOB, "~", 4 }, + { OP_NE, "!=", 4 }, + { OP_EQ, "==", 4 }, + { OP_LT, "<", 5 }, + { OP_LE, "<=", 5 }, + { OP_GT, ">", 5 }, + { OP_GE, ">=", 5 }, + { OP_NONE, "OP_NONE", 0 }, + { OP_OPEN_PAREN, "(", 0 }, }; enum { @@ -329,22 +331,18 @@ enum regex_type filter_parse_regex(char *buff, int len, char **search, int *not) return type; } -static int filter_build_regex(struct filter_pred *pred) +static void filter_build_regex(struct filter_pred *pred) { struct regex *r = &pred->regex; - char *search, *dup; - enum regex_type type; - int not; - - type = filter_parse_regex(r->pattern, r->len, &search, ¬); - dup = kstrdup(search, GFP_KERNEL); - if (!dup) - return -ENOMEM; - - strcpy(r->pattern, dup); - kfree(dup); - - r->len = strlen(r->pattern); + char *search; + enum regex_type type = MATCH_FULL; + int not = 0; + + if (pred->op == OP_GLOB) { + type = filter_parse_regex(r->pattern, r->len, &search, ¬); + r->len = strlen(search); + memmove(r->pattern, search, r->len+1); + } switch (type) { case MATCH_FULL: @@ -362,8 +360,6 @@ static int filter_build_regex(struct filter_pred *pred) } pred->not ^= not; - - return 0; } /* return 1 if event matches, 0 otherwise (discard) */ @@ -676,7 +672,10 @@ static bool is_string_field(struct ftrace_event_field *field) static int is_legal_op(struct ftrace_event_field *field, int op) { - if (is_string_field(field) && (op != OP_EQ && op != OP_NE)) + if (is_string_field(field) && + (op != OP_EQ && op != OP_NE && op != OP_GLOB)) + return 0; + if (!is_string_field(field) && op == OP_GLOB) return 0; return 1; @@ -761,15 +760,13 @@ static int filter_add_pred(struct filter_parse_state *ps, } if (is_string_field(field)) { - ret = filter_build_regex(pred); - if (ret) - return ret; + filter_build_regex(pred); if (field->filter_type == FILTER_STATIC_STRING) { fn = filter_pred_string; pred->regex.field_len = field->size; } else if (field->filter_type == FILTER_DYN_STRING) - fn = filter_pred_strloc; + fn = filter_pred_strloc; else { fn = filter_pred_pchar; pred->regex.field_len = strlen(pred->regex.pattern); -- cgit v1.1 From 6fb2915df7f0747d9044da9dbff5b46dc2e20830 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Thu, 15 Oct 2009 11:21:42 +0800 Subject: tracing/profile: Add filter support - Add an ioctl to allocate a filter for a perf event. - Free the filter when the associated perf event is to be freed. - Do the filtering in perf_swevent_match(). Signed-off-by: Li Zefan Acked-by: Peter Zijlstra Acked-by: Frederic Weisbecker Cc: Steven Rostedt Cc: Tom Zanussi LKML-Reference: <4AD69546.8050401@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 80 ++++++++++++++++++++-- kernel/trace/trace.h | 3 +- kernel/trace/trace_events_filter.c | 133 +++++++++++++++++++++++++++++-------- 3 files changed, 182 insertions(+), 34 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 9d0b5c6..12b5ec3 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -28,6 +28,7 @@ #include #include #include +#include #include @@ -1658,6 +1659,8 @@ static struct perf_event_context *find_get_context(pid_t pid, int cpu) return ERR_PTR(err); } +static void perf_event_free_filter(struct perf_event *event); + static void free_event_rcu(struct rcu_head *head) { struct perf_event *event; @@ -1665,6 +1668,7 @@ static void free_event_rcu(struct rcu_head *head) event = container_of(head, struct perf_event, rcu_head); if (event->ns) put_pid_ns(event->ns); + perf_event_free_filter(event); kfree(event); } @@ -1974,7 +1978,8 @@ unlock: return ret; } -int perf_event_set_output(struct perf_event *event, int output_fd); +static int perf_event_set_output(struct perf_event *event, int output_fd); +static int perf_event_set_filter(struct perf_event *event, void __user *arg); static long perf_ioctl(struct file *file, unsigned int cmd, unsigned long arg) { @@ -2002,6 +2007,9 @@ static long perf_ioctl(struct file *file, unsigned int cmd, unsigned long arg) case PERF_EVENT_IOC_SET_OUTPUT: return perf_event_set_output(event, arg); + case PERF_EVENT_IOC_SET_FILTER: + return perf_event_set_filter(event, (void __user *)arg); + default: return -ENOTTY; } @@ -3806,9 +3814,14 @@ static int perf_swevent_is_counting(struct perf_event *event) return 1; } +static int perf_tp_event_match(struct perf_event *event, + struct perf_sample_data *data); + static int perf_swevent_match(struct perf_event *event, enum perf_type_id type, - u32 event_id, struct pt_regs *regs) + u32 event_id, + struct perf_sample_data *data, + struct pt_regs *regs) { if (!perf_swevent_is_counting(event)) return 0; @@ -3826,6 +3839,10 @@ static int perf_swevent_match(struct perf_event *event, return 0; } + if (event->attr.type == PERF_TYPE_TRACEPOINT && + !perf_tp_event_match(event, data)) + return 0; + return 1; } @@ -3842,7 +3859,7 @@ static void perf_swevent_ctx_event(struct perf_event_context *ctx, rcu_read_lock(); list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { - if (perf_swevent_match(event, type, event_id, regs)) + if (perf_swevent_match(event, type, event_id, data, regs)) perf_swevent_add(event, nr, nmi, data, regs); } rcu_read_unlock(); @@ -4086,6 +4103,7 @@ static const struct pmu perf_ops_task_clock = { }; #ifdef CONFIG_EVENT_PROFILE + void perf_tp_event(int event_id, u64 addr, u64 count, void *record, int entry_size) { @@ -4109,8 +4127,15 @@ void perf_tp_event(int event_id, u64 addr, u64 count, void *record, } EXPORT_SYMBOL_GPL(perf_tp_event); -extern int ftrace_profile_enable(int); -extern void ftrace_profile_disable(int); +static int perf_tp_event_match(struct perf_event *event, + struct perf_sample_data *data) +{ + void *record = data->raw->data; + + if (likely(!event->filter) || filter_match_preds(event->filter, record)) + return 1; + return 0; +} static void tp_perf_event_destroy(struct perf_event *event) { @@ -4135,12 +4160,53 @@ static const struct pmu *tp_perf_event_init(struct perf_event *event) return &perf_ops_generic; } + +static int perf_event_set_filter(struct perf_event *event, void __user *arg) +{ + char *filter_str; + int ret; + + if (event->attr.type != PERF_TYPE_TRACEPOINT) + return -EINVAL; + + filter_str = strndup_user(arg, PAGE_SIZE); + if (IS_ERR(filter_str)) + return PTR_ERR(filter_str); + + ret = ftrace_profile_set_filter(event, event->attr.config, filter_str); + + kfree(filter_str); + return ret; +} + +static void perf_event_free_filter(struct perf_event *event) +{ + ftrace_profile_free_filter(event); +} + #else + +static int perf_tp_event_match(struct perf_event *event, + struct perf_sample_data *data) +{ + return 1; +} + static const struct pmu *tp_perf_event_init(struct perf_event *event) { return NULL; } -#endif + +static int perf_event_set_filter(struct perf_event *event, void __user *arg) +{ + return -ENOENT; +} + +static void perf_event_free_filter(struct perf_event *event) +{ +} + +#endif /* CONFIG_EVENT_PROFILE */ atomic_t perf_swevent_enabled[PERF_COUNT_SW_MAX]; @@ -4394,7 +4460,7 @@ err_size: goto out; } -int perf_event_set_output(struct perf_event *event, int output_fd) +static int perf_event_set_output(struct perf_event *event, int output_fd) { struct perf_event *output_event = NULL; struct file *output_file = NULL; diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index ffe53dd..4959ada 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -743,7 +743,8 @@ filter_check_discard(struct ftrace_event_call *call, void *rec, struct ring_buffer *buffer, struct ring_buffer_event *event) { - if (unlikely(call->filter_active) && !filter_match_preds(call, rec)) { + if (unlikely(call->filter_active) && + !filter_match_preds(call->filter, rec)) { ring_buffer_discard_commit(buffer, event); return 1; } diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c index 273845f..e27bb6a 100644 --- a/kernel/trace/trace_events_filter.c +++ b/kernel/trace/trace_events_filter.c @@ -21,6 +21,7 @@ #include #include #include +#include #include "trace.h" #include "trace_output.h" @@ -363,9 +364,8 @@ static void filter_build_regex(struct filter_pred *pred) } /* return 1 if event matches, 0 otherwise (discard) */ -int filter_match_preds(struct ftrace_event_call *call, void *rec) +int filter_match_preds(struct event_filter *filter, void *rec) { - struct event_filter *filter = call->filter; int match, top = 0, val1 = 0, val2 = 0; int stack[MAX_FILTER_PRED]; struct filter_pred *pred; @@ -538,9 +538,8 @@ static void filter_disable_preds(struct ftrace_event_call *call) filter->preds[i]->fn = filter_pred_none; } -void destroy_preds(struct ftrace_event_call *call) +static void __free_preds(struct event_filter *filter) { - struct event_filter *filter = call->filter; int i; if (!filter) @@ -553,21 +552,24 @@ void destroy_preds(struct ftrace_event_call *call) kfree(filter->preds); kfree(filter->filter_string); kfree(filter); +} + +void destroy_preds(struct ftrace_event_call *call) +{ + __free_preds(call->filter); call->filter = NULL; + call->filter_active = 0; } -static int init_preds(struct ftrace_event_call *call) +static struct event_filter *__alloc_preds(void) { struct event_filter *filter; struct filter_pred *pred; int i; - if (call->filter) - return 0; - - filter = call->filter = kzalloc(sizeof(*filter), GFP_KERNEL); - if (!call->filter) - return -ENOMEM; + filter = kzalloc(sizeof(*filter), GFP_KERNEL); + if (!filter) + return ERR_PTR(-ENOMEM); filter->n_preds = 0; @@ -583,12 +585,24 @@ static int init_preds(struct ftrace_event_call *call) filter->preds[i] = pred; } - return 0; + return filter; oom: - destroy_preds(call); + __free_preds(filter); + return ERR_PTR(-ENOMEM); +} + +static int init_preds(struct ftrace_event_call *call) +{ + if (call->filter) + return 0; + + call->filter_active = 0; + call->filter = __alloc_preds(); + if (IS_ERR(call->filter)) + return PTR_ERR(call->filter); - return -ENOMEM; + return 0; } static int init_subsystem_preds(struct event_subsystem *system) @@ -629,10 +643,10 @@ static void filter_free_subsystem_preds(struct event_subsystem *system) static int filter_add_pred_fn(struct filter_parse_state *ps, struct ftrace_event_call *call, + struct event_filter *filter, struct filter_pred *pred, filter_pred_fn_t fn) { - struct event_filter *filter = call->filter; int idx, err; if (filter->n_preds == MAX_FILTER_PRED) { @@ -647,7 +661,6 @@ static int filter_add_pred_fn(struct filter_parse_state *ps, return err; filter->n_preds++; - call->filter_active = 1; return 0; } @@ -726,6 +739,7 @@ static filter_pred_fn_t select_comparison_fn(int op, int field_size, static int filter_add_pred(struct filter_parse_state *ps, struct ftrace_event_call *call, + struct event_filter *filter, struct filter_pred *pred, bool dry_run) { @@ -795,7 +809,7 @@ static int filter_add_pred(struct filter_parse_state *ps, add_pred_fn: if (!dry_run) - return filter_add_pred_fn(ps, call, pred, fn); + return filter_add_pred_fn(ps, call, filter, pred, fn); return 0; } @@ -1154,6 +1168,7 @@ static int check_preds(struct filter_parse_state *ps) } static int replace_preds(struct ftrace_event_call *call, + struct event_filter *filter, struct filter_parse_state *ps, char *filter_string, bool dry_run) @@ -1200,7 +1215,7 @@ static int replace_preds(struct ftrace_event_call *call, add_pred: if (!pred) return -ENOMEM; - err = filter_add_pred(ps, call, pred, dry_run); + err = filter_add_pred(ps, call, filter, pred, dry_run); filter_free_pred(pred); if (err) return err; @@ -1216,6 +1231,7 @@ static int replace_system_preds(struct event_subsystem *system, char *filter_string) { struct ftrace_event_call *call; + struct event_filter *filter; int err; bool fail = true; @@ -1228,17 +1244,19 @@ static int replace_system_preds(struct event_subsystem *system, continue; /* try to see if the filter can be applied */ - err = replace_preds(call, ps, filter_string, true); + err = replace_preds(call, filter, ps, filter_string, true); if (err) continue; /* really apply the filter */ filter_disable_preds(call); - err = replace_preds(call, ps, filter_string, false); + err = replace_preds(call, filter, ps, filter_string, false); if (err) filter_disable_preds(call); - else - replace_filter_string(call->filter, filter_string); + else { + call->filter_active = 1; + replace_filter_string(filter, filter_string); + } fail = false; } @@ -1252,7 +1270,6 @@ static int replace_system_preds(struct event_subsystem *system, int apply_event_filter(struct ftrace_event_call *call, char *filter_string) { int err; - struct filter_parse_state *ps; mutex_lock(&event_mutex); @@ -1283,10 +1300,11 @@ int apply_event_filter(struct ftrace_event_call *call, char *filter_string) goto out; } - err = replace_preds(call, ps, filter_string, false); + err = replace_preds(call, call->filter, ps, filter_string, false); if (err) append_filter_err(ps, call->filter); - + else + call->filter_active = 1; out: filter_opstack_clear(ps); postfix_clear(ps); @@ -1301,7 +1319,6 @@ int apply_subsystem_event_filter(struct event_subsystem *system, char *filter_string) { int err; - struct filter_parse_state *ps; mutex_lock(&event_mutex); @@ -1345,3 +1362,67 @@ out_unlock: return err; } +#ifdef CONFIG_EVENT_PROFILE + +void ftrace_profile_free_filter(struct perf_event *event) +{ + struct event_filter *filter = event->filter; + + event->filter = NULL; + __free_preds(filter); +} + +int ftrace_profile_set_filter(struct perf_event *event, int event_id, + char *filter_str) +{ + int err; + struct event_filter *filter; + struct filter_parse_state *ps; + struct ftrace_event_call *call = NULL; + + mutex_lock(&event_mutex); + + list_for_each_entry(call, &ftrace_events, list) { + if (call->id == event_id) + break; + } + if (!call) + return -EINVAL; + + if (event->filter) + return -EEXIST; + + filter = __alloc_preds(); + if (IS_ERR(filter)) + return PTR_ERR(filter); + + err = -ENOMEM; + ps = kzalloc(sizeof(*ps), GFP_KERNEL); + if (!ps) + goto free_preds; + + parse_init(ps, filter_ops, filter_str); + err = filter_parse(ps); + if (err) + goto free_ps; + + err = replace_preds(call, filter, ps, filter_str, false); + if (!err) + event->filter = filter; + +free_ps: + filter_opstack_clear(ps); + postfix_clear(ps); + kfree(ps); + +free_preds: + if (err) + __free_preds(filter); + + mutex_unlock(&event_mutex); + + return err; +} + +#endif /* CONFIG_EVENT_PROFILE */ + -- cgit v1.1 From a66abe7fbf7805a1a02f241bd5283265ff6706ec Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 15 Oct 2009 12:24:04 +0200 Subject: tracing/events: Fix locking imbalance in the filter code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Américo Wang noticed that we have a locking imbalance in the error paths of ftrace_profile_set_filter(), causing potential leakage of event_mutex. Also clean up other error codepaths related to event_mutex while at it. Plus fix an initialized variable in the subsystem filter code. Reported-by: Américo Wang Cc: Li Zefan Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Steven Rostedt Cc: Tom Zanussi LKML-Reference: <2375c9f90910150247u5ccb8e2at58c764e385ffa490@mail.gmail.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_events_filter.c | 26 +++++++++++++++----------- 1 file changed, 15 insertions(+), 11 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c index e27bb6a..21d3475 100644 --- a/kernel/trace/trace_events_filter.c +++ b/kernel/trace/trace_events_filter.c @@ -1230,10 +1230,10 @@ static int replace_system_preds(struct event_subsystem *system, struct filter_parse_state *ps, char *filter_string) { + struct event_filter *filter = system->filter; struct ftrace_event_call *call; - struct event_filter *filter; - int err; bool fail = true; + int err; list_for_each_entry(call, &ftrace_events, list) { @@ -1262,7 +1262,7 @@ static int replace_system_preds(struct event_subsystem *system, if (fail) { parse_error(ps, FILT_ERR_BAD_SUBSYS_FILTER, 0); - return err; + return -EINVAL; } return 0; } @@ -1281,8 +1281,7 @@ int apply_event_filter(struct ftrace_event_call *call, char *filter_string) if (!strcmp(strstrip(filter_string), "0")) { filter_disable_preds(call); remove_filter_string(call->filter); - mutex_unlock(&event_mutex); - return 0; + goto out_unlock; } err = -ENOMEM; @@ -1330,8 +1329,7 @@ int apply_subsystem_event_filter(struct event_subsystem *system, if (!strcmp(strstrip(filter_string), "0")) { filter_free_subsystem_preds(system); remove_filter_string(system->filter); - mutex_unlock(&event_mutex); - return 0; + goto out_unlock; } err = -ENOMEM; @@ -1386,15 +1384,20 @@ int ftrace_profile_set_filter(struct perf_event *event, int event_id, if (call->id == event_id) break; } + + err = -EINVAL; if (!call) - return -EINVAL; + goto out_unlock; + err = -EEXIST; if (event->filter) - return -EEXIST; + goto out_unlock; filter = __alloc_preds(); - if (IS_ERR(filter)) - return PTR_ERR(filter); + if (IS_ERR(filter)) { + err = PTR_ERR(filter); + goto out_unlock; + } err = -ENOMEM; ps = kzalloc(sizeof(*ps), GFP_KERNEL); @@ -1419,6 +1422,7 @@ free_preds: if (err) __free_preds(filter); +out_unlock: mutex_unlock(&event_mutex); return err; -- cgit v1.1 From f397af06e4c9bf5a0bc92facb8cb29905e338ab0 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Fri, 16 Oct 2009 20:07:20 -0400 Subject: tracing/kprobes: Update kprobe-tracer selftest against new syntax Update kprobe-tracer selftest since command syntax has been changed. Signed-off-by: Masami Hiramatsu Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <20091017000720.16556.26343.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_kprobe.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 739f70e..cdacdab 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -1438,12 +1438,12 @@ static __init int kprobe_trace_self_tests_init(void) pr_info("Testing kprobe tracing: "); ret = command_trace_probe("p:testprobe kprobe_trace_selftest_target " - "a1 a2 a3 a4 a5 a6"); + "$arg1 $arg2 $arg3 $arg4 $stack $stack0"); if (WARN_ON_ONCE(ret)) pr_warning("error enabling function entry\n"); ret = command_trace_probe("r:testprobe2 kprobe_trace_selftest_target " - "ra rv"); + "$retval"); if (WARN_ON_ONCE(ret)) pr_warning("error enabling function return\n"); -- cgit v1.1 From e63cc2397ecc0f2b604f22fb9cdbb05911c1e5d4 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Fri, 16 Oct 2009 20:07:28 -0400 Subject: tracing/kprobes: Add failure messages for debugging Add verbose failure messages to kprobe-tracer for debugging. Signed-off-by: Masami Hiramatsu Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <20091017000728.16556.16713.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_kprobe.c | 35 +++++++++++++++++++++++++++-------- 1 file changed, 27 insertions(+), 8 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index cdacdab..b8ef707 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -597,15 +597,19 @@ static int create_trace_probe(int argc, char **argv) void *addr = NULL; char buf[MAX_EVENT_NAME_LEN]; - if (argc < 2) + if (argc < 2) { + pr_info("Probe point is not specified.\n"); return -EINVAL; + } if (argv[0][0] == 'p') is_return = 0; else if (argv[0][0] == 'r') is_return = 1; - else + else { + pr_info("Probe definition must be started with 'p' or 'r'.\n"); return -EINVAL; + } if (argv[0][1] == ':') { event = &argv[0][2]; @@ -625,21 +629,29 @@ static int create_trace_probe(int argc, char **argv) } if (isdigit(argv[1][0])) { - if (is_return) + if (is_return) { + pr_info("Return probe point must be a symbol.\n"); return -EINVAL; + } /* an address specified */ ret = strict_strtoul(&argv[0][2], 0, (unsigned long *)&addr); - if (ret) + if (ret) { + pr_info("Failed to parse address.\n"); return ret; + } } else { /* a symbol specified */ symbol = argv[1]; /* TODO: support .init module functions */ ret = split_symbol_offset(symbol, &offset); - if (ret) + if (ret) { + pr_info("Failed to parse symbol.\n"); return ret; - if (offset && is_return) + } + if (offset && is_return) { + pr_info("Return probe must be used without offset.\n"); return -EINVAL; + } } argc -= 2; argv += 2; @@ -658,8 +670,11 @@ static int create_trace_probe(int argc, char **argv) } tp = alloc_trace_probe(group, event, addr, symbol, offset, argc, is_return); - if (IS_ERR(tp)) + if (IS_ERR(tp)) { + pr_info("Failed to allocate trace_probe.(%d)\n", + (int)PTR_ERR(tp)); return PTR_ERR(tp); + } /* parse arguments */ ret = 0; @@ -672,6 +687,8 @@ static int create_trace_probe(int argc, char **argv) arg = argv[i]; if (conflict_field_name(argv[i], tp->args, i)) { + pr_info("Argument%d name '%s' conflicts with " + "another field.\n", i, argv[i]); ret = -EINVAL; goto error; } @@ -685,8 +702,10 @@ static int create_trace_probe(int argc, char **argv) goto error; } ret = parse_probe_arg(arg, &tp->args[i].fetch, is_return); - if (ret) + if (ret) { + pr_info("Parse error at argument%d. (%d)\n", i, ret); goto error; + } } tp->nr_args = i; -- cgit v1.1 From f7d7986060b2890fc26db6ab5203efbd33aa2497 Mon Sep 17 00:00:00 2001 From: Anton Blanchard Date: Sun, 18 Oct 2009 01:09:29 +0000 Subject: perf_event: Add alignment-faults and emulation-faults software events Add two more software events that are common to many cpus. Alignment faults: When a load or store is not aligned properly. Emulation faults: When an instruction is emulated in software. Both cause a very significant slowdown (100x or worse), so identifying and fixing them is very important. Signed-off-by: Anton Blanchard Signed-off-by: Paul Mackerras --- kernel/perf_event.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 9d0b5c6..0683b33 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -4186,6 +4186,8 @@ static const struct pmu *sw_perf_event_init(struct perf_event *event) case PERF_COUNT_SW_PAGE_FAULTS_MAJ: case PERF_COUNT_SW_CONTEXT_SWITCHES: case PERF_COUNT_SW_CPU_MIGRATIONS: + case PERF_COUNT_SW_ALIGNMENT_FAULTS: + case PERF_COUNT_SW_EMULATION_FAULTS: if (!event->parent) { atomic_inc(&perf_swevent_enabled[event_id]); event->destroy = sw_perf_event_destroy; -- cgit v1.1 From dd004c475cd15a5749b04b0283d41ffdfa57d658 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Tue, 27 Oct 2009 16:42:44 -0400 Subject: kprobe-tracer: Compare both of event-name and event-group to find probe Fix find_probe_event() to compare both of event-name and event-group. Without this fix, kprobe-tracer overwrites existing same event-name probe even if its group-name is different. Signed-off-by: Masami Hiramatsu Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju LKML-Reference: <20091027204244.30545.27516.stgit@harusame> Signed-off-by: Ingo Molnar --- kernel/trace/trace_kprobe.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index b8ef707..a86c3ac 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -353,12 +353,14 @@ static void free_trace_probe(struct trace_probe *tp) kfree(tp); } -static struct trace_probe *find_probe_event(const char *event) +static struct trace_probe *find_probe_event(const char *event, + const char *group) { struct trace_probe *tp; list_for_each_entry(tp, &probe_list, list) - if (!strcmp(tp->call.name, event)) + if (strcmp(tp->call.name, event) == 0 && + strcmp(tp->call.system, group) == 0) return tp; return NULL; } @@ -383,7 +385,7 @@ static int register_trace_probe(struct trace_probe *tp) mutex_lock(&probe_lock); /* register as an event */ - old_tp = find_probe_event(tp->call.name); + old_tp = find_probe_event(tp->call.name, tp->call.system); if (old_tp) { /* delete old event */ unregister_trace_probe(old_tp); -- cgit v1.1 From 3ed67776fc23061180896086a206a02be649dd26 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Wed, 28 Oct 2009 17:37:01 +0800 Subject: tracing/filters: Fix to make system filter work commit fce29d15b59245597f7f320db4a9f2be0f5fb512 ("tracing/filters: Refactor subsystem filter code") broke system filter accidentally. Signed-off-by: Li Zefan Cc: Steven Rostedt Cc: Frederic Weisbecker Cc: Tom Zanussi LKML-Reference: <4AE810BD.3070009@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_events_filter.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c index 21d3475..50504cb 100644 --- a/kernel/trace/trace_events_filter.c +++ b/kernel/trace/trace_events_filter.c @@ -1230,12 +1230,12 @@ static int replace_system_preds(struct event_subsystem *system, struct filter_parse_state *ps, char *filter_string) { - struct event_filter *filter = system->filter; struct ftrace_event_call *call; bool fail = true; int err; list_for_each_entry(call, &ftrace_events, list) { + struct event_filter *filter = call->filter; if (!call->define_fields) continue; -- cgit v1.1 From fb0459d75c1d0a4ba3cafdd2c754e7486968a676 Mon Sep 17 00:00:00 2001 From: Arjan van de Ven Date: Fri, 25 Sep 2009 12:25:56 +0200 Subject: perf/core: Provide a kernel-internal interface to get to performance counters There are reasons for kernel code to ask for, and use, performance counters. For example, in CPU freq governors this tends to be a good idea, but there are other examples possible as well of course. This patch adds the needed bits to do enable this functionality; they have been tested in an experimental cpufreq driver that I'm working on, and the changes are all that I needed to access counters properly. [fweisbec@gmail.com: added pid to perf_event_create_kernel_counter so that we can profile a particular task too TODO: Have a better error reporting, don't just return NULL in fail case.] v2: Remove the wrong comment about the fact perf_event_create_kernel_counter must be called from a kernel thread. Signed-off-by: Arjan van de Ven Acked-by: Peter Zijlstra Cc: "K.Prasad" Cc: Alan Stern Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Ingo Molnar Cc: Jan Kiszka Cc: Jiri Slaby Cc: Li Zefan Cc: Avi Kivity Cc: Paul Mackerras Cc: Mike Galbraith Cc: Masami Hiramatsu Cc: Paul Mundt Cc: Jan Kiszka Cc: Avi Kivity LKML-Reference: <20090925122556.2f8bd939@infradead.org> Signed-off-by: Frederic Weisbecker --- kernel/perf_event.c | 75 ++++++++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 74 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 12b5ec3..02d4ff0 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -1725,6 +1725,26 @@ static int perf_release(struct inode *inode, struct file *file) return 0; } +int perf_event_release_kernel(struct perf_event *event) +{ + struct perf_event_context *ctx = event->ctx; + + WARN_ON_ONCE(ctx->parent_ctx); + mutex_lock(&ctx->mutex); + perf_event_remove_from_context(event); + mutex_unlock(&ctx->mutex); + + mutex_lock(&event->owner->perf_event_mutex); + list_del_init(&event->owner_entry); + mutex_unlock(&event->owner->perf_event_mutex); + put_task_struct(event->owner); + + free_event(event); + + return 0; +} +EXPORT_SYMBOL_GPL(perf_event_release_kernel); + static int perf_event_read_size(struct perf_event *event) { int entry = sizeof(u64); /* value */ @@ -1750,7 +1770,7 @@ static int perf_event_read_size(struct perf_event *event) return size; } -static u64 perf_event_read_value(struct perf_event *event) +u64 perf_event_read_value(struct perf_event *event) { struct perf_event *child; u64 total = 0; @@ -1761,6 +1781,7 @@ static u64 perf_event_read_value(struct perf_event *event) return total; } +EXPORT_SYMBOL_GPL(perf_event_read_value); static int perf_event_read_entry(struct perf_event *event, u64 read_format, char __user *buf) @@ -4638,6 +4659,58 @@ err_put_context: return err; } +/** + * perf_event_create_kernel_counter + * + * @attr: attributes of the counter to create + * @cpu: cpu in which the counter is bound + * @pid: task to profile + */ +struct perf_event * +perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, + pid_t pid) +{ + struct perf_event *event; + struct perf_event_context *ctx; + int err; + + /* + * Get the target context (task or percpu): + */ + + ctx = find_get_context(pid, cpu); + if (IS_ERR(ctx)) + return NULL ; + + event = perf_event_alloc(attr, cpu, ctx, NULL, + NULL, GFP_KERNEL); + err = PTR_ERR(event); + if (IS_ERR(event)) + goto err_put_context; + + event->filp = NULL; + WARN_ON_ONCE(ctx->parent_ctx); + mutex_lock(&ctx->mutex); + perf_install_in_context(ctx, event, cpu); + ++ctx->generation; + mutex_unlock(&ctx->mutex); + + event->owner = current; + get_task_struct(current); + mutex_lock(¤t->perf_event_mutex); + list_add_tail(&event->owner_entry, ¤t->perf_event_list); + mutex_unlock(¤t->perf_event_mutex); + + return event; + +err_put_context: + if (err < 0) + put_ctx(ctx); + + return NULL; +} +EXPORT_SYMBOL_GPL(perf_event_create_kernel_counter); + /* * inherit a event from parent task to child task: */ -- cgit v1.1 From 97eaf5300b9d0cd99c310bf8c4a0f2f3296d88a3 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Sun, 18 Oct 2009 15:33:50 +0200 Subject: perf/core: Add a callback to perf events A simple callback in a perf event can be used for multiple purposes. For example it is useful for triggered based events like hardware breakpoints that need a callback to dispatch a triggered breakpoint event. v2: Simplify a bit the callback attribution as suggested by Paul Mackerras Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: "K.Prasad" Cc: Alan Stern Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Ingo Molnar Cc: Paul Mackerras Cc: Mike Galbraith Cc: Paul Mundt --- kernel/perf_event.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 02d4ff0..5087125 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -4293,6 +4293,7 @@ perf_event_alloc(struct perf_event_attr *attr, struct perf_event_context *ctx, struct perf_event *group_leader, struct perf_event *parent_event, + perf_callback_t callback, gfp_t gfpflags) { const struct pmu *pmu; @@ -4335,6 +4336,11 @@ perf_event_alloc(struct perf_event_attr *attr, event->state = PERF_EVENT_STATE_INACTIVE; + if (!callback && parent_event) + callback = parent_event->callback; + + event->callback = callback; + if (attr->disabled) event->state = PERF_EVENT_STATE_OFF; @@ -4611,7 +4617,7 @@ SYSCALL_DEFINE5(perf_event_open, } event = perf_event_alloc(&attr, cpu, ctx, group_leader, - NULL, GFP_KERNEL); + NULL, NULL, GFP_KERNEL); err = PTR_ERR(event); if (IS_ERR(event)) goto err_put_context; @@ -4668,7 +4674,7 @@ err_put_context: */ struct perf_event * perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, - pid_t pid) + pid_t pid, perf_callback_t callback) { struct perf_event *event; struct perf_event_context *ctx; @@ -4683,7 +4689,7 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, return NULL ; event = perf_event_alloc(attr, cpu, ctx, NULL, - NULL, GFP_KERNEL); + NULL, callback, GFP_KERNEL); err = PTR_ERR(event); if (IS_ERR(event)) goto err_put_context; @@ -4736,7 +4742,7 @@ inherit_event(struct perf_event *parent_event, child_event = perf_event_alloc(&parent_event->attr, parent_event->cpu, child_ctx, group_leader, parent_event, - GFP_KERNEL); + NULL, GFP_KERNEL); if (IS_ERR(child_event)) return child_event; get_ctx(child_ctx); -- cgit v1.1 From 77b44d1b7c28360910cdbd427fb62d485c08674c Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Tue, 3 Nov 2009 19:12:47 -0500 Subject: tracing/kprobes: Rename Kprobe-tracer to kprobe-event Rename Kprobes-based event tracer to kprobes-based tracing event (kprobe-event), since it is not a tracer but an extensible tracing event interface. This also changes CONFIG_KPROBE_TRACER to CONFIG_KPROBE_EVENT and sets it y by default. Signed-off-by: Masami Hiramatsu Acked-by: Frederic Weisbecker Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju LKML-Reference: <20091104001247.3454.14131.stgit@harusame> Signed-off-by: Ingo Molnar --- kernel/trace/Kconfig | 19 ++++++++++++------- kernel/trace/Makefile | 2 +- kernel/trace/trace_kprobe.c | 6 ++---- 3 files changed, 15 insertions(+), 12 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig index 15372a9..f056716 100644 --- a/kernel/trace/Kconfig +++ b/kernel/trace/Kconfig @@ -428,17 +428,22 @@ config BLK_DEV_IO_TRACE If unsure, say N. -config KPROBE_TRACER +config KPROBE_EVENT depends on KPROBES depends on X86 - bool "Trace kprobes" + bool "Enable kprobes-based dynamic events" select TRACING - select GENERIC_TRACER + default y help - This tracer probes everywhere where kprobes can probe it, and - records various registers and memories specified by user. - This also allows you to trace kprobe probe points as a dynamic - defined events. It provides per-probe event filtering interface. + This allows the user to add tracing events (similar to tracepoints) on the fly + via the ftrace interface. See Documentation/trace/kprobetrace.txt + for more details. + + Those events can be inserted wherever kprobes can probe, and record + various register and memory values. + + This option is also required by perf-probe subcommand of perf tools. If + you want to use perf tools, this option is strongly recommended. config DYNAMIC_FTRACE bool "enable/disable ftrace tracepoints dynamically" diff --git a/kernel/trace/Makefile b/kernel/trace/Makefile index c8cb75d..edc3a3c 100644 --- a/kernel/trace/Makefile +++ b/kernel/trace/Makefile @@ -53,7 +53,7 @@ obj-$(CONFIG_EVENT_TRACING) += trace_export.o obj-$(CONFIG_FTRACE_SYSCALLS) += trace_syscalls.o obj-$(CONFIG_EVENT_PROFILE) += trace_event_profile.o obj-$(CONFIG_EVENT_TRACING) += trace_events_filter.o -obj-$(CONFIG_KPROBE_TRACER) += trace_kprobe.o +obj-$(CONFIG_KPROBE_EVENT) += trace_kprobe.o obj-$(CONFIG_EVENT_TRACING) += power-traces.o libftrace-y := ftrace.o diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index a86c3ac..cf17a66 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -1,5 +1,5 @@ /* - * kprobe based kernel tracer + * Kprobes-based tracing events * * Created by Masami Hiramatsu * @@ -57,8 +57,6 @@ const char *reserved_field_names[] = { FIELD_STRING_FUNC, }; -/* currently, trace_kprobe only supports X86. */ - struct fetch_func { unsigned long (*func)(struct pt_regs *, void *); void *data; @@ -191,7 +189,7 @@ static __kprobes void free_indirect_fetch_data(struct indirect_fetch_data *data) } /** - * Kprobe tracer core functions + * Kprobe event core functions */ struct probe_arg { -- cgit v1.1 From 444a2a3bcd6d5bed5c823136f68fcc93c0fe283f Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Fri, 6 Nov 2009 04:13:05 +0100 Subject: tracing, perf_events: Protect the buffer from recursion in perf While tracing using events with perf, if one enables the lockdep:lock_acquire event, it will infect every other perf trace events. Basically, you can enable whatever set of trace events through perf but if this event is part of the set, the only result we can get is a long list of lock_acquire events of rcu read lock, and only that. This is because of a recursion inside perf. 1) When a trace event is triggered, it will fill a per cpu buffer and submit it to perf. 2) Perf will commit this event but will also protect some data using rcu_read_lock 3) A recursion appears: rcu_read_lock triggers a lock_acquire event that will fill the per cpu event and then submit the buffer to perf. 4) Perf detects a recursion and ignores it 5) Perf continues its work on the previous event, but its buffer has been overwritten by the lock_acquire event, it has then been turned into a lock_acquire event of rcu read lock Such scenario also happens with lock_release with rcu_read_unlock(). We could turn the rcu_read_lock() into __rcu_read_lock() to drop the lock debugging from perf fast path, but that would make us lose the rcu debugging and that doesn't prevent from other possible kind of recursion from perf in the future. This patch adds a recursion protection based on a counter on the perf trace per cpu buffers to solve the problem. -v2: Fixed lost whitespace, added reviewed-by tag Signed-off-by: Frederic Weisbecker Reviewed-by: Masami Hiramatsu Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Steven Rostedt Cc: Li Zefan Cc: Jason Baron LKML-Reference: <1257477185-7838-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_event_profile.c | 41 ++++++++++++++----------------- kernel/trace/trace_kprobe.c | 50 ++++++++++++++++++++++++++++++++------ kernel/trace/trace_syscalls.c | 44 +++++++++++++++++++++++++++------ 3 files changed, 96 insertions(+), 39 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_event_profile.c b/kernel/trace/trace_event_profile.c index c9f687a..e0d351b 100644 --- a/kernel/trace/trace_event_profile.c +++ b/kernel/trace/trace_event_profile.c @@ -8,41 +8,36 @@ #include #include "trace.h" -/* - * We can't use a size but a type in alloc_percpu() - * So let's create a dummy type that matches the desired size - */ -typedef struct {char buf[FTRACE_MAX_PROFILE_SIZE];} profile_buf_t; -char *trace_profile_buf; -EXPORT_SYMBOL_GPL(trace_profile_buf); +struct perf_trace_buf *perf_trace_buf; +EXPORT_SYMBOL_GPL(perf_trace_buf); -char *trace_profile_buf_nmi; -EXPORT_SYMBOL_GPL(trace_profile_buf_nmi); +struct perf_trace_buf *perf_trace_buf_nmi; +EXPORT_SYMBOL_GPL(perf_trace_buf_nmi); /* Count the events in use (per event id, not per instance) */ static int total_profile_count; static int ftrace_profile_enable_event(struct ftrace_event_call *event) { - char *buf; + struct perf_trace_buf *buf; int ret = -ENOMEM; if (atomic_inc_return(&event->profile_count)) return 0; if (!total_profile_count) { - buf = (char *)alloc_percpu(profile_buf_t); + buf = alloc_percpu(struct perf_trace_buf); if (!buf) goto fail_buf; - rcu_assign_pointer(trace_profile_buf, buf); + rcu_assign_pointer(perf_trace_buf, buf); - buf = (char *)alloc_percpu(profile_buf_t); + buf = alloc_percpu(struct perf_trace_buf); if (!buf) goto fail_buf_nmi; - rcu_assign_pointer(trace_profile_buf_nmi, buf); + rcu_assign_pointer(perf_trace_buf_nmi, buf); } ret = event->profile_enable(event); @@ -53,10 +48,10 @@ static int ftrace_profile_enable_event(struct ftrace_event_call *event) fail_buf_nmi: if (!total_profile_count) { - free_percpu(trace_profile_buf_nmi); - free_percpu(trace_profile_buf); - trace_profile_buf_nmi = NULL; - trace_profile_buf = NULL; + free_percpu(perf_trace_buf_nmi); + free_percpu(perf_trace_buf); + perf_trace_buf_nmi = NULL; + perf_trace_buf = NULL; } fail_buf: atomic_dec(&event->profile_count); @@ -84,7 +79,7 @@ int ftrace_profile_enable(int event_id) static void ftrace_profile_disable_event(struct ftrace_event_call *event) { - char *buf, *nmi_buf; + struct perf_trace_buf *buf, *nmi_buf; if (!atomic_add_negative(-1, &event->profile_count)) return; @@ -92,11 +87,11 @@ static void ftrace_profile_disable_event(struct ftrace_event_call *event) event->profile_disable(event); if (!--total_profile_count) { - buf = trace_profile_buf; - rcu_assign_pointer(trace_profile_buf, NULL); + buf = perf_trace_buf; + rcu_assign_pointer(perf_trace_buf, NULL); - nmi_buf = trace_profile_buf_nmi; - rcu_assign_pointer(trace_profile_buf_nmi, NULL); + nmi_buf = perf_trace_buf_nmi; + rcu_assign_pointer(perf_trace_buf_nmi, NULL); /* * Ensure every events in profiling have finished before diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index cf17a66..3696476 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -1208,6 +1208,7 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp, struct trace_probe *tp = container_of(kp, struct trace_probe, rp.kp); struct ftrace_event_call *call = &tp->call; struct kprobe_trace_entry *entry; + struct perf_trace_buf *trace_buf; struct trace_entry *ent; int size, __size, i, pc, __cpu; unsigned long irq_flags; @@ -1229,14 +1230,26 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp, __cpu = smp_processor_id(); if (in_nmi()) - raw_data = rcu_dereference(trace_profile_buf_nmi); + trace_buf = rcu_dereference(perf_trace_buf_nmi); else - raw_data = rcu_dereference(trace_profile_buf); + trace_buf = rcu_dereference(perf_trace_buf); - if (!raw_data) + if (!trace_buf) goto end; - raw_data = per_cpu_ptr(raw_data, __cpu); + trace_buf = per_cpu_ptr(trace_buf, __cpu); + + if (trace_buf->recursion++) + goto end_recursion; + + /* + * Make recursion update visible before entering perf_tp_event + * so that we protect from perf recursions. + */ + barrier(); + + raw_data = trace_buf->buf; + /* Zero dead bytes from alignment to avoid buffer leak to userspace */ *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; entry = (struct kprobe_trace_entry *)raw_data; @@ -1249,8 +1262,12 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp, for (i = 0; i < tp->nr_args; i++) entry->args[i] = call_fetch(&tp->args[i].fetch, regs); perf_tp_event(call->id, entry->ip, 1, entry, size); + +end_recursion: + trace_buf->recursion--; end: local_irq_restore(irq_flags); + return 0; } @@ -1261,6 +1278,7 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri, struct trace_probe *tp = container_of(ri->rp, struct trace_probe, rp); struct ftrace_event_call *call = &tp->call; struct kretprobe_trace_entry *entry; + struct perf_trace_buf *trace_buf; struct trace_entry *ent; int size, __size, i, pc, __cpu; unsigned long irq_flags; @@ -1282,14 +1300,26 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri, __cpu = smp_processor_id(); if (in_nmi()) - raw_data = rcu_dereference(trace_profile_buf_nmi); + trace_buf = rcu_dereference(perf_trace_buf_nmi); else - raw_data = rcu_dereference(trace_profile_buf); + trace_buf = rcu_dereference(perf_trace_buf); - if (!raw_data) + if (!trace_buf) goto end; - raw_data = per_cpu_ptr(raw_data, __cpu); + trace_buf = per_cpu_ptr(trace_buf, __cpu); + + if (trace_buf->recursion++) + goto end_recursion; + + /* + * Make recursion update visible before entering perf_tp_event + * so that we protect from perf recursions. + */ + barrier(); + + raw_data = trace_buf->buf; + /* Zero dead bytes from alignment to avoid buffer leak to userspace */ *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; entry = (struct kretprobe_trace_entry *)raw_data; @@ -1303,8 +1333,12 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri, for (i = 0; i < tp->nr_args; i++) entry->args[i] = call_fetch(&tp->args[i].fetch, regs); perf_tp_event(call->id, entry->ret_ip, 1, entry, size); + +end_recursion: + trace_buf->recursion--; end: local_irq_restore(irq_flags); + return 0; } diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index 58b8e53..51213b0 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -477,6 +477,7 @@ static int sys_prof_refcount_exit; static void prof_syscall_enter(struct pt_regs *regs, long id) { struct syscall_metadata *sys_data; + struct perf_trace_buf *trace_buf; struct syscall_trace_enter *rec; unsigned long flags; char *raw_data; @@ -507,14 +508,25 @@ static void prof_syscall_enter(struct pt_regs *regs, long id) cpu = smp_processor_id(); if (in_nmi()) - raw_data = rcu_dereference(trace_profile_buf_nmi); + trace_buf = rcu_dereference(perf_trace_buf_nmi); else - raw_data = rcu_dereference(trace_profile_buf); + trace_buf = rcu_dereference(perf_trace_buf); - if (!raw_data) + if (!trace_buf) goto end; - raw_data = per_cpu_ptr(raw_data, cpu); + trace_buf = per_cpu_ptr(trace_buf, cpu); + + if (trace_buf->recursion++) + goto end_recursion; + + /* + * Make recursion update visible before entering perf_tp_event + * so that we protect from perf recursions. + */ + barrier(); + + raw_data = trace_buf->buf; /* zero the dead bytes from align to not leak stack to user */ *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; @@ -527,6 +539,8 @@ static void prof_syscall_enter(struct pt_regs *regs, long id) (unsigned long *)&rec->args); perf_tp_event(sys_data->enter_id, 0, 1, rec, size); +end_recursion: + trace_buf->recursion--; end: local_irq_restore(flags); } @@ -574,6 +588,7 @@ static void prof_syscall_exit(struct pt_regs *regs, long ret) { struct syscall_metadata *sys_data; struct syscall_trace_exit *rec; + struct perf_trace_buf *trace_buf; unsigned long flags; int syscall_nr; char *raw_data; @@ -605,14 +620,25 @@ static void prof_syscall_exit(struct pt_regs *regs, long ret) cpu = smp_processor_id(); if (in_nmi()) - raw_data = rcu_dereference(trace_profile_buf_nmi); + trace_buf = rcu_dereference(perf_trace_buf_nmi); else - raw_data = rcu_dereference(trace_profile_buf); + trace_buf = rcu_dereference(perf_trace_buf); - if (!raw_data) + if (!trace_buf) goto end; - raw_data = per_cpu_ptr(raw_data, cpu); + trace_buf = per_cpu_ptr(trace_buf, cpu); + + if (trace_buf->recursion++) + goto end_recursion; + + /* + * Make recursion update visible before entering perf_tp_event + * so that we protect from perf recursions. + */ + barrier(); + + raw_data = trace_buf->buf; /* zero the dead bytes from align to not leak stack to user */ *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; @@ -626,6 +652,8 @@ static void prof_syscall_exit(struct pt_regs *regs, long ret) perf_tp_event(sys_data->exit_id, 0, 1, rec, size); +end_recursion: + trace_buf->recursion--; end: local_irq_restore(flags); } -- cgit v1.1 From 24f1e32c60c45c89a997c73395b69c8af6f0a84e Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 9 Sep 2009 19:22:48 +0200 Subject: hw-breakpoints: Rewrite the hw-breakpoints layer on top of perf events This patch rebase the implementation of the breakpoints API on top of perf events instances. Each breakpoints are now perf events that handle the register scheduling, thread/cpu attachment, etc.. The new layering is now made as follows: ptrace kgdb ftrace perf syscall \ | / / \ | / / / Core breakpoint API / / | / | / Breakpoints perf events | | Breakpoints PMU ---- Debug Register constraints handling (Part of core breakpoint API) | | Hardware debug registers Reasons of this rewrite: - Use the centralized/optimized pmu registers scheduling, implying an easier arch integration - More powerful register handling: perf attributes (pinned/flexible events, exclusive/non-exclusive, tunable period, etc...) Impact: - New perf ABI: the hardware breakpoints counters - Ptrace breakpoints setting remains tricky and still needs some per thread breakpoints references. Todo (in the order): - Support breakpoints perf counter events for perf tools (ie: implement perf_bpcounter_event()) - Support from perf tools Changes in v2: - Follow the perf "event " rename - The ptrace regression have been fixed (ptrace breakpoint perf events weren't released when a task ended) - Drop the struct hw_breakpoint and store generic fields in perf_event_attr. - Separate core and arch specific headers, drop asm-generic/hw_breakpoint.h and create linux/hw_breakpoint.h - Use new generic len/type for breakpoint - Handle off case: when breakpoints api is not supported by an arch Changes in v3: - Fix broken CONFIG_KVM, we need to propagate the breakpoint api changes to kvm when we exit the guest and restore the bp registers to the host. Changes in v4: - Drop the hw_breakpoint_restore() stub as it is only used by KVM - EXPORT_SYMBOL_GPL hw_breakpoint_restore() as KVM can be built as a module - Restore the breakpoints unconditionally on kvm guest exit: TIF_DEBUG_THREAD doesn't anymore cover every cases of running breakpoints and vcpu->arch.switch_db_regs might not always be set when the guest used debug registers. (Waiting for a reliable optimization) Changes in v5: - Split-up the asm-generic/hw-breakpoint.h moving to linux/hw_breakpoint.h into a separate patch - Optimize the breakpoints restoring while switching from kvm guest to host. We only want to restore the state if we have active breakpoints to the host, otherwise we don't care about messed-up address registers. - Add asm/hw_breakpoint.h to Kbuild - Fix bad breakpoint type in trace_selftest.c Changes in v6: - Fix wrong header inclusion in trace.h (triggered a build error with CONFIG_FTRACE_SELFTEST Signed-off-by: Frederic Weisbecker Cc: Prasad Cc: Alan Stern Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Ingo Molnar Cc: Jan Kiszka Cc: Jiri Slaby Cc: Li Zefan Cc: Avi Kivity Cc: Paul Mackerras Cc: Mike Galbraith Cc: Masami Hiramatsu Cc: Paul Mundt --- kernel/exit.c | 5 + kernel/hw_breakpoint.c | 424 +++++++++++++++++------------------------- kernel/perf_event.c | 53 +++++- kernel/trace/trace.h | 5 +- kernel/trace/trace_entries.h | 6 +- kernel/trace/trace_ksym.c | 126 ++++++------- kernel/trace/trace_selftest.c | 3 +- 7 files changed, 293 insertions(+), 329 deletions(-) (limited to 'kernel') diff --git a/kernel/exit.c b/kernel/exit.c index e61891f..266f892 100644 --- a/kernel/exit.c +++ b/kernel/exit.c @@ -49,6 +49,7 @@ #include #include #include +#include #include #include @@ -980,6 +981,10 @@ NORET_TYPE void do_exit(long code) proc_exit_connector(tsk); /* + * FIXME: do that only when needed, using sched_exit tracepoint + */ + flush_ptrace_hw_breakpoint(tsk); + /* * Flush inherited counters to the parent - before the parent * gets woken up by child-exit notifications. */ diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index c1f64e6..08f6d01 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -15,6 +15,7 @@ * * Copyright (C) 2007 Alan Stern * Copyright (C) IBM Corporation, 2009 + * Copyright (C) 2009, Frederic Weisbecker */ /* @@ -35,334 +36,242 @@ #include #include -#include +#include + #include #ifdef CONFIG_X86 #include #endif -/* - * Spinlock that protects all (un)register operations over kernel/user-space - * breakpoint requests - */ -static DEFINE_SPINLOCK(hw_breakpoint_lock); - -/* Array of kernel-space breakpoint structures */ -struct hw_breakpoint *hbp_kernel[HBP_NUM]; - -/* - * Per-processor copy of hbp_kernel[]. Used only when hbp_kernel is being - * modified but we need the older copy to handle any hbp exceptions. It will - * sync with hbp_kernel[] value after updation is done through IPIs. - */ -DEFINE_PER_CPU(struct hw_breakpoint*, this_hbp_kernel[HBP_NUM]); - -/* - * Kernel breakpoints grow downwards, starting from HBP_NUM - * 'hbp_kernel_pos' denotes lowest numbered breakpoint register occupied for - * kernel-space request. We will initialise it here and not in an __init - * routine because load_debug_registers(), which uses this variable can be - * called very early during CPU initialisation. - */ -unsigned int hbp_kernel_pos = HBP_NUM; -/* - * An array containing refcount of threads using a given bkpt register - * Accesses are synchronised by acquiring hw_breakpoint_lock - */ -unsigned int hbp_user_refcount[HBP_NUM]; +static atomic_t bp_slot; -/* - * Load the debug registers during startup of a CPU. - */ -void load_debug_registers(void) +int reserve_bp_slot(struct perf_event *bp) { - unsigned long flags; - struct task_struct *tsk = current; - - spin_lock_bh(&hw_breakpoint_lock); - - /* Prevent IPIs for new kernel breakpoint updates */ - local_irq_save(flags); - arch_update_kernel_hw_breakpoint(NULL); - local_irq_restore(flags); - - if (test_tsk_thread_flag(tsk, TIF_DEBUG)) - arch_install_thread_hw_breakpoint(tsk); - - spin_unlock_bh(&hw_breakpoint_lock); -} + if (atomic_inc_return(&bp_slot) == HBP_NUM) { + atomic_dec(&bp_slot); -/* - * Erase all the hardware breakpoint info associated with a thread. - * - * If tsk != current then tsk must not be usable (for example, a - * child being cleaned up from a failed fork). - */ -void flush_thread_hw_breakpoint(struct task_struct *tsk) -{ - int i; - struct thread_struct *thread = &(tsk->thread); - - spin_lock_bh(&hw_breakpoint_lock); - - /* The thread no longer has any breakpoints associated with it */ - clear_tsk_thread_flag(tsk, TIF_DEBUG); - for (i = 0; i < HBP_NUM; i++) { - if (thread->hbp[i]) { - hbp_user_refcount[i]--; - kfree(thread->hbp[i]); - thread->hbp[i] = NULL; - } + return -ENOSPC; } - arch_flush_thread_hw_breakpoint(tsk); - - /* Actually uninstall the breakpoints if necessary */ - if (tsk == current) - arch_uninstall_thread_hw_breakpoint(); - spin_unlock_bh(&hw_breakpoint_lock); + return 0; } -/* - * Copy the hardware breakpoint info from a thread to its cloned child. - */ -int copy_thread_hw_breakpoint(struct task_struct *tsk, - struct task_struct *child, unsigned long clone_flags) +void release_bp_slot(struct perf_event *bp) { - /* - * We will assume that breakpoint settings are not inherited - * and the child starts out with no debug registers set. - * But what about CLONE_PTRACE? - */ - clear_tsk_thread_flag(child, TIF_DEBUG); - - /* We will call flush routine since the debugregs are not inherited */ - arch_flush_thread_hw_breakpoint(child); - - return 0; + atomic_dec(&bp_slot); } -static int __register_user_hw_breakpoint(int pos, struct task_struct *tsk, - struct hw_breakpoint *bp) +int __register_perf_hw_breakpoint(struct perf_event *bp) { - struct thread_struct *thread = &(tsk->thread); - int rc; + int ret; - /* Do not overcommit. Fail if kernel has used the hbp registers */ - if (pos >= hbp_kernel_pos) - return -ENOSPC; + ret = reserve_bp_slot(bp); + if (ret) + return ret; - rc = arch_validate_hwbkpt_settings(bp, tsk); - if (rc) - return rc; + if (!bp->attr.disabled) + ret = arch_validate_hwbkpt_settings(bp, bp->ctx->task); - thread->hbp[pos] = bp; - hbp_user_refcount[pos]++; + return ret; +} - arch_update_user_hw_breakpoint(pos, tsk); - /* - * Does it need to be installed right now? - * Otherwise it will get installed the next time tsk runs - */ - if (tsk == current) - arch_install_thread_hw_breakpoint(tsk); +int register_perf_hw_breakpoint(struct perf_event *bp) +{ + bp->callback = perf_bp_event; - return rc; + return __register_perf_hw_breakpoint(bp); } /* - * Modify the address of a hbp register already in use by the task - * Do not invoke this in-lieu of a __unregister_user_hw_breakpoint() + * Register a breakpoint bound to a task and a given cpu. + * If cpu is -1, the breakpoint is active for the task in every cpu + * If the task is -1, the breakpoint is active for every tasks in the given + * cpu. */ -static int __modify_user_hw_breakpoint(int pos, struct task_struct *tsk, - struct hw_breakpoint *bp) +static struct perf_event * +register_user_hw_breakpoint_cpu(unsigned long addr, + int len, + int type, + perf_callback_t triggered, + pid_t pid, + int cpu, + bool active) { - struct thread_struct *thread = &(tsk->thread); - - if ((pos >= hbp_kernel_pos) || (arch_validate_hwbkpt_settings(bp, tsk))) - return -EINVAL; - - if (thread->hbp[pos] == NULL) - return -EINVAL; - - thread->hbp[pos] = bp; + struct perf_event_attr *attr; + struct perf_event *bp; + + attr = kzalloc(sizeof(*attr), GFP_KERNEL); + if (!attr) + return ERR_PTR(-ENOMEM); + + attr->type = PERF_TYPE_BREAKPOINT; + attr->size = sizeof(*attr); + attr->bp_addr = addr; + attr->bp_len = len; + attr->bp_type = type; /* - * 'pos' must be that of a hbp register already used by 'tsk' - * Otherwise arch_modify_user_hw_breakpoint() will fail + * Such breakpoints are used by debuggers to trigger signals when + * we hit the excepted memory op. We can't miss such events, they + * must be pinned. */ - arch_update_user_hw_breakpoint(pos, tsk); + attr->pinned = 1; - if (tsk == current) - arch_install_thread_hw_breakpoint(tsk); + if (!active) + attr->disabled = 1; - return 0; -} - -static void __unregister_user_hw_breakpoint(int pos, struct task_struct *tsk) -{ - hbp_user_refcount[pos]--; - tsk->thread.hbp[pos] = NULL; + bp = perf_event_create_kernel_counter(attr, cpu, pid, triggered); + kfree(attr); - arch_update_user_hw_breakpoint(pos, tsk); - - if (tsk == current) - arch_install_thread_hw_breakpoint(tsk); + return bp; } /** * register_user_hw_breakpoint - register a hardware breakpoint for user space + * @addr: is the memory address that triggers the breakpoint + * @len: the length of the access to the memory (1 byte, 2 bytes etc...) + * @type: the type of the access to the memory (read/write/exec) + * @triggered: callback to trigger when we hit the breakpoint * @tsk: pointer to 'task_struct' of the process to which the address belongs - * @bp: the breakpoint structure to register - * - * @bp.info->name or @bp.info->address, @bp.info->len, @bp.info->type and - * @bp->triggered must be set properly before invocation + * @active: should we activate it while registering it * */ -int register_user_hw_breakpoint(struct task_struct *tsk, - struct hw_breakpoint *bp) +struct perf_event * +register_user_hw_breakpoint(unsigned long addr, + int len, + int type, + perf_callback_t triggered, + struct task_struct *tsk, + bool active) { - struct thread_struct *thread = &(tsk->thread); - int i, rc = -ENOSPC; - - spin_lock_bh(&hw_breakpoint_lock); - - for (i = 0; i < hbp_kernel_pos; i++) { - if (!thread->hbp[i]) { - rc = __register_user_hw_breakpoint(i, tsk, bp); - break; - } - } - if (!rc) - set_tsk_thread_flag(tsk, TIF_DEBUG); - - spin_unlock_bh(&hw_breakpoint_lock); - return rc; + return register_user_hw_breakpoint_cpu(addr, len, type, triggered, + tsk->pid, -1, active); } EXPORT_SYMBOL_GPL(register_user_hw_breakpoint); /** * modify_user_hw_breakpoint - modify a user-space hardware breakpoint + * @bp: the breakpoint structure to modify + * @addr: is the memory address that triggers the breakpoint + * @len: the length of the access to the memory (1 byte, 2 bytes etc...) + * @type: the type of the access to the memory (read/write/exec) + * @triggered: callback to trigger when we hit the breakpoint * @tsk: pointer to 'task_struct' of the process to which the address belongs - * @bp: the breakpoint structure to unregister - * + * @active: should we activate it while registering it */ -int modify_user_hw_breakpoint(struct task_struct *tsk, struct hw_breakpoint *bp) +struct perf_event * +modify_user_hw_breakpoint(struct perf_event *bp, + unsigned long addr, + int len, + int type, + perf_callback_t triggered, + struct task_struct *tsk, + bool active) { - struct thread_struct *thread = &(tsk->thread); - int i, ret = -ENOENT; + /* + * FIXME: do it without unregistering + * - We don't want to lose our slot + * - If the new bp is incorrect, don't lose the older one + */ + unregister_hw_breakpoint(bp); - spin_lock_bh(&hw_breakpoint_lock); - for (i = 0; i < hbp_kernel_pos; i++) { - if (bp == thread->hbp[i]) { - ret = __modify_user_hw_breakpoint(i, tsk, bp); - break; - } - } - spin_unlock_bh(&hw_breakpoint_lock); - return ret; + return register_user_hw_breakpoint(addr, len, type, triggered, + tsk, active); } EXPORT_SYMBOL_GPL(modify_user_hw_breakpoint); /** - * unregister_user_hw_breakpoint - unregister a user-space hardware breakpoint - * @tsk: pointer to 'task_struct' of the process to which the address belongs + * unregister_hw_breakpoint - unregister a user-space hardware breakpoint * @bp: the breakpoint structure to unregister - * */ -void unregister_user_hw_breakpoint(struct task_struct *tsk, - struct hw_breakpoint *bp) +void unregister_hw_breakpoint(struct perf_event *bp) { - struct thread_struct *thread = &(tsk->thread); - int i, pos = -1, hbp_counter = 0; - - spin_lock_bh(&hw_breakpoint_lock); - for (i = 0; i < hbp_kernel_pos; i++) { - if (thread->hbp[i]) - hbp_counter++; - if (bp == thread->hbp[i]) - pos = i; - } - if (pos >= 0) { - __unregister_user_hw_breakpoint(pos, tsk); - hbp_counter--; - } - if (!hbp_counter) - clear_tsk_thread_flag(tsk, TIF_DEBUG); - - spin_unlock_bh(&hw_breakpoint_lock); + if (!bp) + return; + perf_event_release_kernel(bp); +} +EXPORT_SYMBOL_GPL(unregister_hw_breakpoint); + +static struct perf_event * +register_kernel_hw_breakpoint_cpu(unsigned long addr, + int len, + int type, + perf_callback_t triggered, + int cpu, + bool active) +{ + return register_user_hw_breakpoint_cpu(addr, len, type, triggered, + -1, cpu, active); } -EXPORT_SYMBOL_GPL(unregister_user_hw_breakpoint); /** - * register_kernel_hw_breakpoint - register a hardware breakpoint for kernel space - * @bp: the breakpoint structure to register - * - * @bp.info->name or @bp.info->address, @bp.info->len, @bp.info->type and - * @bp->triggered must be set properly before invocation + * register_wide_hw_breakpoint - register a wide breakpoint in the kernel + * @addr: is the memory address that triggers the breakpoint + * @len: the length of the access to the memory (1 byte, 2 bytes etc...) + * @type: the type of the access to the memory (read/write/exec) + * @triggered: callback to trigger when we hit the breakpoint + * @active: should we activate it while registering it * + * @return a set of per_cpu pointers to perf events */ -int register_kernel_hw_breakpoint(struct hw_breakpoint *bp) +struct perf_event ** +register_wide_hw_breakpoint(unsigned long addr, + int len, + int type, + perf_callback_t triggered, + bool active) { - int rc; + struct perf_event **cpu_events, **pevent, *bp; + long err; + int cpu; + + cpu_events = alloc_percpu(typeof(*cpu_events)); + if (!cpu_events) + return ERR_PTR(-ENOMEM); - rc = arch_validate_hwbkpt_settings(bp, NULL); - if (rc) - return rc; + for_each_possible_cpu(cpu) { + pevent = per_cpu_ptr(cpu_events, cpu); + bp = register_kernel_hw_breakpoint_cpu(addr, len, type, + triggered, cpu, active); - spin_lock_bh(&hw_breakpoint_lock); + *pevent = bp; - rc = -ENOSPC; - /* Check if we are over-committing */ - if ((hbp_kernel_pos > 0) && (!hbp_user_refcount[hbp_kernel_pos-1])) { - hbp_kernel_pos--; - hbp_kernel[hbp_kernel_pos] = bp; - on_each_cpu(arch_update_kernel_hw_breakpoint, NULL, 1); - rc = 0; + if (IS_ERR(bp) || !bp) { + err = PTR_ERR(bp); + goto fail; + } } - spin_unlock_bh(&hw_breakpoint_lock); - return rc; + return cpu_events; + +fail: + for_each_possible_cpu(cpu) { + pevent = per_cpu_ptr(cpu_events, cpu); + if (IS_ERR(*pevent) || !*pevent) + break; + unregister_hw_breakpoint(*pevent); + } + free_percpu(cpu_events); + /* return the error if any */ + return ERR_PTR(err); } -EXPORT_SYMBOL_GPL(register_kernel_hw_breakpoint); /** - * unregister_kernel_hw_breakpoint - unregister a HW breakpoint for kernel space - * @bp: the breakpoint structure to unregister - * - * Uninstalls and unregisters @bp. + * unregister_wide_hw_breakpoint - unregister a wide breakpoint in the kernel + * @cpu_events: the per cpu set of events to unregister */ -void unregister_kernel_hw_breakpoint(struct hw_breakpoint *bp) +void unregister_wide_hw_breakpoint(struct perf_event **cpu_events) { - int i, j; - - spin_lock_bh(&hw_breakpoint_lock); - - /* Find the 'bp' in our list of breakpoints for kernel */ - for (i = hbp_kernel_pos; i < HBP_NUM; i++) - if (bp == hbp_kernel[i]) - break; + int cpu; + struct perf_event **pevent; - /* Check if we did not find a match for 'bp'. If so return early */ - if (i == HBP_NUM) { - spin_unlock_bh(&hw_breakpoint_lock); - return; + for_each_possible_cpu(cpu) { + pevent = per_cpu_ptr(cpu_events, cpu); + unregister_hw_breakpoint(*pevent); } - - /* - * We'll shift the breakpoints one-level above to compact if - * unregistration creates a hole - */ - for (j = i; j > hbp_kernel_pos; j--) - hbp_kernel[j] = hbp_kernel[j-1]; - - hbp_kernel[hbp_kernel_pos] = NULL; - on_each_cpu(arch_update_kernel_hw_breakpoint, NULL, 1); - hbp_kernel_pos++; - - spin_unlock_bh(&hw_breakpoint_lock); + free_percpu(cpu_events); } -EXPORT_SYMBOL_GPL(unregister_kernel_hw_breakpoint); + static struct notifier_block hw_breakpoint_exceptions_nb = { .notifier_call = hw_breakpoint_exceptions_notify, @@ -374,5 +283,12 @@ static int __init init_hw_breakpoint(void) { return register_die_notifier(&hw_breakpoint_exceptions_nb); } - core_initcall(init_hw_breakpoint); + + +struct pmu perf_ops_bp = { + .enable = arch_install_hw_breakpoint, + .disable = arch_uninstall_hw_breakpoint, + .read = hw_breakpoint_pmu_read, + .unthrottle = hw_breakpoint_pmu_unthrottle +}; diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 5087125..98dc56b 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -29,6 +29,7 @@ #include #include #include +#include #include @@ -4229,6 +4230,51 @@ static void perf_event_free_filter(struct perf_event *event) #endif /* CONFIG_EVENT_PROFILE */ +#ifdef CONFIG_HAVE_HW_BREAKPOINT +static void bp_perf_event_destroy(struct perf_event *event) +{ + release_bp_slot(event); +} + +static const struct pmu *bp_perf_event_init(struct perf_event *bp) +{ + int err; + /* + * The breakpoint is already filled if we haven't created the counter + * through perf syscall + * FIXME: manage to get trigerred to NULL if it comes from syscalls + */ + if (!bp->callback) + err = register_perf_hw_breakpoint(bp); + else + err = __register_perf_hw_breakpoint(bp); + if (err) + return ERR_PTR(err); + + bp->destroy = bp_perf_event_destroy; + + return &perf_ops_bp; +} + +void perf_bp_event(struct perf_event *bp, void *regs) +{ + /* TODO */ +} +#else +static void bp_perf_event_destroy(struct perf_event *event) +{ +} + +static const struct pmu *bp_perf_event_init(struct perf_event *bp) +{ + return NULL; +} + +void perf_bp_event(struct perf_event *bp, void *regs) +{ +} +#endif + atomic_t perf_swevent_enabled[PERF_COUNT_SW_MAX]; static void sw_perf_event_destroy(struct perf_event *event) @@ -4375,6 +4421,11 @@ perf_event_alloc(struct perf_event_attr *attr, pmu = tp_perf_event_init(event); break; + case PERF_TYPE_BREAKPOINT: + pmu = bp_perf_event_init(event); + break; + + default: break; } @@ -4686,7 +4737,7 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, ctx = find_get_context(pid, cpu); if (IS_ERR(ctx)) - return NULL ; + return NULL; event = perf_event_alloc(attr, cpu, ctx, NULL, NULL, callback, GFP_KERNEL); diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 91c3d0e..d72f06f 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -11,14 +11,11 @@ #include #include #include +#include #include #include -#ifdef CONFIG_KSYM_TRACER -#include -#endif - enum trace_type { __TRACE_FIRST_TYPE = 0, diff --git a/kernel/trace/trace_entries.h b/kernel/trace/trace_entries.h index e19747d..c16a08f 100644 --- a/kernel/trace/trace_entries.h +++ b/kernel/trace/trace_entries.h @@ -372,11 +372,11 @@ FTRACE_ENTRY(ksym_trace, ksym_trace_entry, F_STRUCT( __field( unsigned long, ip ) __field( unsigned char, type ) - __array( char , ksym_name, KSYM_NAME_LEN ) __array( char , cmd, TASK_COMM_LEN ) + __field( unsigned long, addr ) ), - F_printk("ip: %pF type: %d ksym_name: %s cmd: %s", + F_printk("ip: %pF type: %d ksym_name: %pS cmd: %s", (void *)__entry->ip, (unsigned int)__entry->type, - __entry->ksym_name, __entry->cmd) + (void *)__entry->addr, __entry->cmd) ); diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index 6d5609c..fea83ee 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -29,7 +29,11 @@ #include "trace_stat.h" #include "trace.h" -/* For now, let us restrict the no. of symbols traced simultaneously to number +#include +#include + +/* + * For now, let us restrict the no. of symbols traced simultaneously to number * of available hardware breakpoint registers. */ #define KSYM_TRACER_MAX HBP_NUM @@ -37,8 +41,10 @@ #define KSYM_TRACER_OP_LEN 3 /* rw- */ struct trace_ksym { - struct hw_breakpoint *ksym_hbp; + struct perf_event **ksym_hbp; unsigned long ksym_addr; + int type; + int len; #ifdef CONFIG_PROFILE_KSYM_TRACER unsigned long counter; #endif @@ -75,10 +81,11 @@ void ksym_collect_stats(unsigned long hbp_hit_addr) } #endif /* CONFIG_PROFILE_KSYM_TRACER */ -void ksym_hbp_handler(struct hw_breakpoint *hbp, struct pt_regs *regs) +void ksym_hbp_handler(struct perf_event *hbp, void *data) { struct ring_buffer_event *event; struct ksym_trace_entry *entry; + struct pt_regs *regs = data; struct ring_buffer *buffer; int pc; @@ -96,12 +103,12 @@ void ksym_hbp_handler(struct hw_breakpoint *hbp, struct pt_regs *regs) entry = ring_buffer_event_data(event); entry->ip = instruction_pointer(regs); - entry->type = hbp->info.type; - strlcpy(entry->ksym_name, hbp->info.name, KSYM_SYMBOL_LEN); + entry->type = hw_breakpoint_type(hbp); + entry->addr = hw_breakpoint_addr(hbp); strlcpy(entry->cmd, current->comm, TASK_COMM_LEN); #ifdef CONFIG_PROFILE_KSYM_TRACER - ksym_collect_stats(hbp->info.address); + ksym_collect_stats(hw_breakpoint_addr(hbp)); #endif /* CONFIG_PROFILE_KSYM_TRACER */ trace_buffer_unlock_commit(buffer, event, 0, pc); @@ -120,31 +127,21 @@ static int ksym_trace_get_access_type(char *str) int access = 0; if (str[0] == 'r') - access += 4; - else if (str[0] != '-') - return -EINVAL; + access |= HW_BREAKPOINT_R; if (str[1] == 'w') - access += 2; - else if (str[1] != '-') - return -EINVAL; + access |= HW_BREAKPOINT_W; - if (str[2] != '-') - return -EINVAL; + if (str[2] == 'x') + access |= HW_BREAKPOINT_X; switch (access) { - case 6: - access = HW_BREAKPOINT_RW; - break; - case 4: - access = -EINVAL; - break; - case 2: - access = HW_BREAKPOINT_WRITE; - break; + case HW_BREAKPOINT_W: + case HW_BREAKPOINT_W | HW_BREAKPOINT_R: + return access; + default: + return -EINVAL; } - - return access; } /* @@ -194,36 +191,33 @@ int process_new_ksym_entry(char *ksymname, int op, unsigned long addr) if (!entry) return -ENOMEM; - entry->ksym_hbp = kzalloc(sizeof(struct hw_breakpoint), GFP_KERNEL); - if (!entry->ksym_hbp) - goto err; - - entry->ksym_hbp->info.name = kstrdup(ksymname, GFP_KERNEL); - if (!entry->ksym_hbp->info.name) - goto err; - - entry->ksym_hbp->info.type = op; - entry->ksym_addr = entry->ksym_hbp->info.address = addr; -#ifdef CONFIG_X86 - entry->ksym_hbp->info.len = HW_BREAKPOINT_LEN_4; -#endif - entry->ksym_hbp->triggered = (void *)ksym_hbp_handler; + entry->type = op; + entry->ksym_addr = addr; + entry->len = HW_BREAKPOINT_LEN_4; + + ret = -EAGAIN; + entry->ksym_hbp = register_wide_hw_breakpoint(entry->ksym_addr, + entry->len, entry->type, + ksym_hbp_handler, true); + if (IS_ERR(entry->ksym_hbp)) { + entry->ksym_hbp = NULL; + ret = PTR_ERR(entry->ksym_hbp); + } - ret = register_kernel_hw_breakpoint(entry->ksym_hbp); - if (ret < 0) { + if (!entry->ksym_hbp) { printk(KERN_INFO "ksym_tracer request failed. Try again" " later!!\n"); - ret = -EAGAIN; goto err; } + hlist_add_head_rcu(&(entry->ksym_hlist), &ksym_filter_head); ksym_filter_entry_count++; + return 0; + err: - if (entry->ksym_hbp) - kfree(entry->ksym_hbp->info.name); - kfree(entry->ksym_hbp); kfree(entry); + return ret; } @@ -244,10 +238,10 @@ static ssize_t ksym_trace_filter_read(struct file *filp, char __user *ubuf, mutex_lock(&ksym_tracer_mutex); hlist_for_each_entry(entry, node, &ksym_filter_head, ksym_hlist) { - ret = trace_seq_printf(s, "%s:", entry->ksym_hbp->info.name); - if (entry->ksym_hbp->info.type == HW_BREAKPOINT_WRITE) + ret = trace_seq_printf(s, "%pS:", (void *)entry->ksym_addr); + if (entry->type == HW_BREAKPOINT_W) ret = trace_seq_puts(s, "-w-\n"); - else if (entry->ksym_hbp->info.type == HW_BREAKPOINT_RW) + else if (entry->type == (HW_BREAKPOINT_W | HW_BREAKPOINT_R)) ret = trace_seq_puts(s, "rw-\n"); WARN_ON_ONCE(!ret); } @@ -269,12 +263,10 @@ static void __ksym_trace_reset(void) mutex_lock(&ksym_tracer_mutex); hlist_for_each_entry_safe(entry, node, node1, &ksym_filter_head, ksym_hlist) { - unregister_kernel_hw_breakpoint(entry->ksym_hbp); + unregister_wide_hw_breakpoint(entry->ksym_hbp); ksym_filter_entry_count--; hlist_del_rcu(&(entry->ksym_hlist)); synchronize_rcu(); - kfree(entry->ksym_hbp->info.name); - kfree(entry->ksym_hbp); kfree(entry); } mutex_unlock(&ksym_tracer_mutex); @@ -327,7 +319,7 @@ static ssize_t ksym_trace_filter_write(struct file *file, hlist_for_each_entry(entry, node, &ksym_filter_head, ksym_hlist) { if (entry->ksym_addr == ksym_addr) { /* Check for malformed request: (6) */ - if (entry->ksym_hbp->info.type != op) + if (entry->type != op) changed = 1; else goto out; @@ -335,18 +327,21 @@ static ssize_t ksym_trace_filter_write(struct file *file, } } if (changed) { - unregister_kernel_hw_breakpoint(entry->ksym_hbp); - entry->ksym_hbp->info.type = op; + unregister_wide_hw_breakpoint(entry->ksym_hbp); + entry->type = op; if (op > 0) { - ret = register_kernel_hw_breakpoint(entry->ksym_hbp); - if (ret == 0) + entry->ksym_hbp = + register_wide_hw_breakpoint(entry->ksym_addr, + entry->len, entry->type, + ksym_hbp_handler, true); + if (IS_ERR(entry->ksym_hbp)) + entry->ksym_hbp = NULL; + if (!entry->ksym_hbp) goto out; } ksym_filter_entry_count--; hlist_del_rcu(&(entry->ksym_hlist)); synchronize_rcu(); - kfree(entry->ksym_hbp->info.name); - kfree(entry->ksym_hbp); kfree(entry); ret = 0; goto out; @@ -413,16 +408,16 @@ static enum print_line_t ksym_trace_output(struct trace_iterator *iter) trace_assign_type(field, entry); - ret = trace_seq_printf(s, "%11s-%-5d [%03d] %-30s ", field->cmd, - entry->pid, iter->cpu, field->ksym_name); + ret = trace_seq_printf(s, "%11s-%-5d [%03d] %pS", field->cmd, + entry->pid, iter->cpu, (char *)field->addr); if (!ret) return TRACE_TYPE_PARTIAL_LINE; switch (field->type) { - case HW_BREAKPOINT_WRITE: + case HW_BREAKPOINT_W: ret = trace_seq_printf(s, " W "); break; - case HW_BREAKPOINT_RW: + case HW_BREAKPOINT_R | HW_BREAKPOINT_W: ret = trace_seq_printf(s, " RW "); break; default: @@ -490,14 +485,13 @@ static int ksym_tracer_stat_show(struct seq_file *m, void *v) entry = hlist_entry(stat, struct trace_ksym, ksym_hlist); - if (entry->ksym_hbp) - access_type = entry->ksym_hbp->info.type; + access_type = entry->type; switch (access_type) { - case HW_BREAKPOINT_WRITE: + case HW_BREAKPOINT_W: seq_puts(m, " W "); break; - case HW_BREAKPOINT_RW: + case HW_BREAKPOINT_R | HW_BREAKPOINT_W: seq_puts(m, " RW "); break; default: diff --git a/kernel/trace/trace_selftest.c b/kernel/trace/trace_selftest.c index 7179c12..27c5072 100644 --- a/kernel/trace/trace_selftest.c +++ b/kernel/trace/trace_selftest.c @@ -828,7 +828,8 @@ trace_selftest_startup_ksym(struct tracer *trace, struct trace_array *tr) ksym_selftest_dummy = 0; /* Register the read-write tracing request */ - ret = process_new_ksym_entry(KSYM_SELFTEST_ENTRY, HW_BREAKPOINT_RW, + ret = process_new_ksym_entry(KSYM_SELFTEST_ENTRY, + HW_BREAKPOINT_R | HW_BREAKPOINT_W, (unsigned long)(&ksym_selftest_dummy)); if (ret < 0) { -- cgit v1.1 From ba1c813a6b9a0ef14d7112daf51270eff326f037 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 10 Sep 2009 09:26:21 +0200 Subject: hw-breakpoints: Arbitrate access to pmu following registers constraints Allow or refuse to build a counter using the breakpoints pmu following given constraints. We keep track of the pmu users by using three per cpu variables: - nr_cpu_bp_pinned stores the number of pinned cpu breakpoints counters in the given cpu - nr_bp_flexible stores the number of non-pinned breakpoints counters in the given cpu. - task_bp_pinned stores the number of pinned task breakpoints in a cpu The latter is not a simple counter but gathers the number of tasks that have n pinned breakpoints. Considering HBP_NUM the number of available breakpoint address registers: task_bp_pinned[0] is the number of tasks having 1 breakpoint task_bp_pinned[1] is the number of tasks having 2 breakpoints [...] task_bp_pinned[HBP_NUM - 1] is the number of tasks having the maximum number of registers (HBP_NUM). When a breakpoint counter is created and wants an access to the pmu, we evaluate the following constraints: == Non-pinned counter == - If attached to a single cpu, check: (per_cpu(nr_bp_flexible, cpu) || (per_cpu(nr_cpu_bp_pinned, cpu) + max(per_cpu(task_bp_pinned, cpu)))) < HBP_NUM -> If there are already non-pinned counters in this cpu, it means there is already a free slot for them. Otherwise, we check that the maximum number of per task breakpoints (for this cpu) plus the number of per cpu breakpoint (for this cpu) doesn't cover every registers. - If attached to every cpus, check: (per_cpu(nr_bp_flexible, *) || (max(per_cpu(nr_cpu_bp_pinned, *)) + max(per_cpu(task_bp_pinned, *)))) < HBP_NUM -> This is roughly the same, except we check the number of per cpu bp for every cpu and we keep the max one. Same for the per tasks breakpoints. == Pinned counter == - If attached to a single cpu, check: ((per_cpu(nr_bp_flexible, cpu) > 1) + per_cpu(nr_cpu_bp_pinned, cpu) + max(per_cpu(task_bp_pinned, cpu))) < HBP_NUM -> Same checks as before. But now the nr_bp_flexible, if any, must keep one register at least (or flexible breakpoints will never be be fed). - If attached to every cpus, check: ((per_cpu(nr_bp_flexible, *) > 1) + max(per_cpu(nr_cpu_bp_pinned, *)) + max(per_cpu(task_bp_pinned, *))) < HBP_NUM Changes in v2: - Counter -> event rename Changes in v5: - Fix unreleased non-pinned task-bound-only counters. We only released it in the first cpu. (Thanks to Paul Mackerras for reporting that) Changes in v6: - Currently, events scheduling are done in this order: cpu context pinned + cpu context non-pinned + task context pinned + task context non-pinned events. Then our current constraints are right theoretically but not in practice, because non-pinned counters may be scheduled before we can apply every possible pinned counters. So consider non-pinned counters as pinned for now. Signed-off-by: Frederic Weisbecker Cc: Prasad Cc: Alan Stern Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Ingo Molnar Cc: Jan Kiszka Cc: Jiri Slaby Cc: Li Zefan Cc: Avi Kivity Cc: Paul Mackerras Cc: Mike Galbraith Cc: Masami Hiramatsu Cc: Paul Mundt --- kernel/hw_breakpoint.c | 211 +++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 205 insertions(+), 6 deletions(-) (limited to 'kernel') diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index 08f6d01..e662dc9 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -16,6 +16,8 @@ * Copyright (C) 2007 Alan Stern * Copyright (C) IBM Corporation, 2009 * Copyright (C) 2009, Frederic Weisbecker + * + * Thanks to Ingo Molnar for his many suggestions. */ /* @@ -44,24 +46,221 @@ #include #endif -static atomic_t bp_slot; +/* + * Constraints data + */ + +/* Number of pinned cpu breakpoints in a cpu */ +static DEFINE_PER_CPU(unsigned int, nr_cpu_bp_pinned); -int reserve_bp_slot(struct perf_event *bp) +/* Number of pinned task breakpoints in a cpu */ +static DEFINE_PER_CPU(unsigned int, task_bp_pinned[HBP_NUM]); + +/* Number of non-pinned cpu/task breakpoints in a cpu */ +static DEFINE_PER_CPU(unsigned int, nr_bp_flexible); + +/* Gather the number of total pinned and un-pinned bp in a cpuset */ +struct bp_busy_slots { + unsigned int pinned; + unsigned int flexible; +}; + +/* Serialize accesses to the above constraints */ +static DEFINE_MUTEX(nr_bp_mutex); + +/* + * Report the maximum number of pinned breakpoints a task + * have in this cpu + */ +static unsigned int max_task_bp_pinned(int cpu) { - if (atomic_inc_return(&bp_slot) == HBP_NUM) { - atomic_dec(&bp_slot); + int i; + unsigned int *tsk_pinned = per_cpu(task_bp_pinned, cpu); - return -ENOSPC; + for (i = HBP_NUM -1; i >= 0; i--) { + if (tsk_pinned[i] > 0) + return i + 1; } return 0; } +/* + * Report the number of pinned/un-pinned breakpoints we have in + * a given cpu (cpu > -1) or in all of them (cpu = -1). + */ +static void fetch_bp_busy_slots(struct bp_busy_slots *slots, int cpu) +{ + if (cpu >= 0) { + slots->pinned = per_cpu(nr_cpu_bp_pinned, cpu); + slots->pinned += max_task_bp_pinned(cpu); + slots->flexible = per_cpu(nr_bp_flexible, cpu); + + return; + } + + for_each_online_cpu(cpu) { + unsigned int nr; + + nr = per_cpu(nr_cpu_bp_pinned, cpu); + nr += max_task_bp_pinned(cpu); + + if (nr > slots->pinned) + slots->pinned = nr; + + nr = per_cpu(nr_bp_flexible, cpu); + + if (nr > slots->flexible) + slots->flexible = nr; + } +} + +/* + * Add a pinned breakpoint for the given task in our constraint table + */ +static void toggle_bp_task_slot(struct task_struct *tsk, int cpu, bool enable) +{ + int count = 0; + struct perf_event *bp; + struct perf_event_context *ctx = tsk->perf_event_ctxp; + unsigned int *task_bp_pinned; + struct list_head *list; + unsigned long flags; + + if (WARN_ONCE(!ctx, "No perf context for this task")) + return; + + list = &ctx->event_list; + + spin_lock_irqsave(&ctx->lock, flags); + + /* + * The current breakpoint counter is not included in the list + * at the open() callback time + */ + list_for_each_entry(bp, list, event_entry) { + if (bp->attr.type == PERF_TYPE_BREAKPOINT) + count++; + } + + spin_unlock_irqrestore(&ctx->lock, flags); + + if (WARN_ONCE(count < 0, "No breakpoint counter found in the counter list")) + return; + + task_bp_pinned = per_cpu(task_bp_pinned, cpu); + if (enable) { + task_bp_pinned[count]++; + if (count > 0) + task_bp_pinned[count-1]--; + } else { + task_bp_pinned[count]--; + if (count > 0) + task_bp_pinned[count-1]++; + } +} + +/* + * Add/remove the given breakpoint in our constraint table + */ +static void toggle_bp_slot(struct perf_event *bp, bool enable) +{ + int cpu = bp->cpu; + struct task_struct *tsk = bp->ctx->task; + + /* Pinned counter task profiling */ + if (tsk) { + if (cpu >= 0) { + toggle_bp_task_slot(tsk, cpu, enable); + return; + } + + for_each_online_cpu(cpu) + toggle_bp_task_slot(tsk, cpu, enable); + return; + } + + /* Pinned counter cpu profiling */ + if (enable) + per_cpu(nr_cpu_bp_pinned, bp->cpu)++; + else + per_cpu(nr_cpu_bp_pinned, bp->cpu)--; +} + +/* + * Contraints to check before allowing this new breakpoint counter: + * + * == Non-pinned counter == (Considered as pinned for now) + * + * - If attached to a single cpu, check: + * + * (per_cpu(nr_bp_flexible, cpu) || (per_cpu(nr_cpu_bp_pinned, cpu) + * + max(per_cpu(task_bp_pinned, cpu)))) < HBP_NUM + * + * -> If there are already non-pinned counters in this cpu, it means + * there is already a free slot for them. + * Otherwise, we check that the maximum number of per task + * breakpoints (for this cpu) plus the number of per cpu breakpoint + * (for this cpu) doesn't cover every registers. + * + * - If attached to every cpus, check: + * + * (per_cpu(nr_bp_flexible, *) || (max(per_cpu(nr_cpu_bp_pinned, *)) + * + max(per_cpu(task_bp_pinned, *)))) < HBP_NUM + * + * -> This is roughly the same, except we check the number of per cpu + * bp for every cpu and we keep the max one. Same for the per tasks + * breakpoints. + * + * + * == Pinned counter == + * + * - If attached to a single cpu, check: + * + * ((per_cpu(nr_bp_flexible, cpu) > 1) + per_cpu(nr_cpu_bp_pinned, cpu) + * + max(per_cpu(task_bp_pinned, cpu))) < HBP_NUM + * + * -> Same checks as before. But now the nr_bp_flexible, if any, must keep + * one register at least (or they will never be fed). + * + * - If attached to every cpus, check: + * + * ((per_cpu(nr_bp_flexible, *) > 1) + max(per_cpu(nr_cpu_bp_pinned, *)) + * + max(per_cpu(task_bp_pinned, *))) < HBP_NUM + */ +int reserve_bp_slot(struct perf_event *bp) +{ + struct bp_busy_slots slots = {0}; + int ret = 0; + + mutex_lock(&nr_bp_mutex); + + fetch_bp_busy_slots(&slots, bp->cpu); + + /* Flexible counters need to keep at least one slot */ + if (slots.pinned + (!!slots.flexible) == HBP_NUM) { + ret = -ENOSPC; + goto end; + } + + toggle_bp_slot(bp, true); + +end: + mutex_unlock(&nr_bp_mutex); + + return ret; +} + void release_bp_slot(struct perf_event *bp) { - atomic_dec(&bp_slot); + mutex_lock(&nr_bp_mutex); + + toggle_bp_slot(bp, false); + + mutex_unlock(&nr_bp_mutex); } + int __register_perf_hw_breakpoint(struct perf_event *bp) { int ret; -- cgit v1.1 From 30ff21e31fe5c8b7b1b7d30cc41e32bc4ee9f175 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Thu, 10 Sep 2009 09:35:20 +0800 Subject: ksym_tracer: Remove KSYM_SELFTEST_ENTRY The macro used to be used in both trace_selftest.c and trace_ksym.c, but no longer, so remove it from header file. Signed-off-by: Li Zefan Cc: Prasad Signed-off-by: Frederic Weisbecker --- kernel/trace/trace.h | 1 - kernel/trace/trace_selftest.c | 3 ++- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index d72f06f..ee00475 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -367,7 +367,6 @@ int register_tracer(struct tracer *type); void unregister_tracer(struct tracer *type); int is_tracing_stopped(void); -#define KSYM_SELFTEST_ENTRY "ksym_selftest_dummy" extern int process_new_ksym_entry(char *ksymname, int op, unsigned long addr); extern unsigned long nsecs_to_usecs(unsigned long nsecs); diff --git a/kernel/trace/trace_selftest.c b/kernel/trace/trace_selftest.c index 27c5072..dc98309 100644 --- a/kernel/trace/trace_selftest.c +++ b/kernel/trace/trace_selftest.c @@ -828,7 +828,8 @@ trace_selftest_startup_ksym(struct tracer *trace, struct trace_array *tr) ksym_selftest_dummy = 0; /* Register the read-write tracing request */ - ret = process_new_ksym_entry(KSYM_SELFTEST_ENTRY, + + ret = process_new_ksym_entry("ksym_selftest_dummy", HW_BREAKPOINT_R | HW_BREAKPOINT_W, (unsigned long)(&ksym_selftest_dummy)); -- cgit v1.1 From 676c0dbe6e514fdd8e434a9e623c781aa9b40b15 Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Mon, 9 Nov 2009 17:37:34 +0900 Subject: ksym_tracer: Support read accesses independent of read/write. All of the infrastructure already exists to support read accesses for platforms that support a read access independently of read/write (such as in the case of the SuperH UBC). This just trivially hooks up the read case by itself. Signed-off-by: Paul Mundt Cc: Ingo Molnar Cc: Li Zefan Cc: Prasad Cc: Alan Stern Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Jan Kiszka Cc: Jiri Slaby Cc: Avi Kivity Cc: Paul Mackerras Cc: Mike Galbraith Cc: Masami Hiramatsu Cc: Arjan van de Ven LKML-Reference: <20091109083733.GA25848@linux-sh.org> Signed-off-by: Frederic Weisbecker --- kernel/trace/trace_ksym.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index fea83ee..11935b5 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -136,6 +136,7 @@ static int ksym_trace_get_access_type(char *str) access |= HW_BREAKPOINT_X; switch (access) { + case HW_BREAKPOINT_R: case HW_BREAKPOINT_W: case HW_BREAKPOINT_W | HW_BREAKPOINT_R: return access; @@ -239,7 +240,9 @@ static ssize_t ksym_trace_filter_read(struct file *filp, char __user *ubuf, hlist_for_each_entry(entry, node, &ksym_filter_head, ksym_hlist) { ret = trace_seq_printf(s, "%pS:", (void *)entry->ksym_addr); - if (entry->type == HW_BREAKPOINT_W) + if (entry->type == HW_BREAKPOINT_R) + ret = trace_seq_puts(s, "r--\n"); + else if (entry->type == HW_BREAKPOINT_W) ret = trace_seq_puts(s, "-w-\n"); else if (entry->type == (HW_BREAKPOINT_W | HW_BREAKPOINT_R)) ret = trace_seq_puts(s, "rw-\n"); @@ -414,6 +417,9 @@ static enum print_line_t ksym_trace_output(struct trace_iterator *iter) return TRACE_TYPE_PARTIAL_LINE; switch (field->type) { + case HW_BREAKPOINT_R: + ret = trace_seq_printf(s, " R "); + break; case HW_BREAKPOINT_W: ret = trace_seq_printf(s, " W "); break; @@ -488,6 +494,9 @@ static int ksym_tracer_stat_show(struct seq_file *m, void *v) access_type = entry->type; switch (access_type) { + case HW_BREAKPOINT_R: + seq_puts(m, " R "); + break; case HW_BREAKPOINT_W: seq_puts(m, " W "); break; -- cgit v1.1 From f60d24d2ad04977b0bd9e3eb35dba2d2fa569af9 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Tue, 10 Nov 2009 10:17:07 +0100 Subject: hw-breakpoints: Fix broken hw-breakpoint sample module The hw-breakpoint sample module has been broken during the hw-breakpoint internals refactoring. Propagate the changes to it. Reported-by: "K. Prasad" Signed-off-by: Frederic Weisbecker --- kernel/hw_breakpoint.c | 3 ++- kernel/kallsyms.c | 1 + 2 files changed, 3 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index e662dc9..9ea9414 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -454,6 +454,7 @@ fail: /* return the error if any */ return ERR_PTR(err); } +EXPORT_SYMBOL_GPL(register_wide_hw_breakpoint); /** * unregister_wide_hw_breakpoint - unregister a wide breakpoint in the kernel @@ -470,7 +471,7 @@ void unregister_wide_hw_breakpoint(struct perf_event **cpu_events) } free_percpu(cpu_events); } - +EXPORT_SYMBOL_GPL(unregister_wide_hw_breakpoint); static struct notifier_block hw_breakpoint_exceptions_nb = { .notifier_call = hw_breakpoint_exceptions_notify, diff --git a/kernel/kallsyms.c b/kernel/kallsyms.c index 8b6b8b6..8e5288a 100644 --- a/kernel/kallsyms.c +++ b/kernel/kallsyms.c @@ -181,6 +181,7 @@ unsigned long kallsyms_lookup_name(const char *name) } return module_kallsyms_lookup_name(name); } +EXPORT_SYMBOL_GPL(kallsyms_lookup_name); int kallsyms_on_each_symbol(int (*fn)(void *, const char *, struct module *, unsigned long), -- cgit v1.1 From 67178767b936fb47a3a5e88097cff41ccbda7acb Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Fri, 13 Nov 2009 10:06:34 +0100 Subject: tracing: Rename 'lockdep' event subsystem into 'lock' Lockdep events subsystem gathers various locking related events such as a request, release, contention or acquisition of a lock. The name of this event subsystem is a bit of a misnomer since these events are not quite related to lockdep but more generally to locking, ie: these events are not reporting lock dependencies or possible deadlock scenario but pure locking events. Hence this rename. Signed-off-by: Frederic Weisbecker Acked-by: Peter Zijlstra Acked-by: Hitoshi Mitake Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Steven Rostedt Cc: Li Zefan LKML-Reference: <1258103194-843-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/lockdep.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/lockdep.c b/kernel/lockdep.c index 9af5672..f5dcd36 100644 --- a/kernel/lockdep.c +++ b/kernel/lockdep.c @@ -49,7 +49,7 @@ #include "lockdep_internals.h" #define CREATE_TRACE_POINTS -#include +#include #ifdef CONFIG_PROVE_LOCKING int prove_locking = 1; -- cgit v1.1 From 559fdc3c1b624edb1933a875022fe7e27934d11c Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 16 Nov 2009 12:45:14 +0100 Subject: perf_event: Optimize perf_output_lock() The purpose of perf_output_{un,}lock() is to: 1) avoid publishing incomplete data [ possible when publishing a head that is ahead of an entry that is still being written ] 2) guarantee fwd progress [ a simple refcount on pending writers doesn't need to drop to 0, making it so would end up implementing something like forced quiecent states of RCU ] To satisfy the above without undue complexity it serializes between CPUs, this means that a pending writer can only be the same cpu in a nested context, and since (under normal operation) a cpu always makes progress we're good -- if the head is only published when the bottom most writer completes. Now we don't need to disable IRQs in order to serialize between CPUs, disabling preemption ought to be sufficient, esp since we already deal with nesting due to NMIs. This avoids potentially expensive (and needless) local IRQ disable/enable ops. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras Cc: Steven Rostedt Cc: Frederic Weisbecker LKML-Reference: <1258373161.26714.254.camel@laptop> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 21 +++++++++++---------- 1 file changed, 11 insertions(+), 10 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 6f4ed3b..3256e36 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -2674,20 +2674,21 @@ static void perf_output_wakeup(struct perf_output_handle *handle) static void perf_output_lock(struct perf_output_handle *handle) { struct perf_mmap_data *data = handle->data; - int cpu; + int cur, cpu = get_cpu(); handle->locked = 0; - local_irq_save(handle->flags); - cpu = smp_processor_id(); - - if (in_nmi() && atomic_read(&data->lock) == cpu) - return; + for (;;) { + cur = atomic_cmpxchg(&data->lock, -1, cpu); + if (cur == -1) { + handle->locked = 1; + break; + } + if (cur == cpu) + break; - while (atomic_cmpxchg(&data->lock, -1, cpu) != -1) cpu_relax(); - - handle->locked = 1; + } } static void perf_output_unlock(struct perf_output_handle *handle) @@ -2733,7 +2734,7 @@ again: if (atomic_xchg(&data->wakeup, 0)) perf_output_wakeup(handle); out: - local_irq_restore(handle->flags); + put_cpu(); } void perf_output_copy(struct perf_output_handle *handle, -- cgit v1.1 From 453f19eea7dbad837425e9b07d84568d14898794 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 20 Nov 2009 22:19:43 +0100 Subject: perf: Allow for custom overflow handlers in-kernel perf users might wish to have custom actions on the sample interrupt. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212508.222339539@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 3852e26..1dfb6cc 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -3710,7 +3710,11 @@ static int __perf_event_overflow(struct perf_event *event, int nmi, perf_event_disable(event); } - perf_event_output(event, nmi, data, regs); + if (event->overflow_handler) + event->overflow_handler(event, nmi, data, regs); + else + perf_event_output(event, nmi, data, regs); + return ret; } @@ -4836,6 +4840,8 @@ inherit_event(struct perf_event *parent_event, if (parent_event->attr.freq) child_event->hw.sample_period = parent_event->hw.sample_period; + child_event->overflow_handler = parent_event->overflow_handler; + /* * Link it up in the child's context: */ -- cgit v1.1 From 0cff784ae41cc125368ae77f1c01328ae2fdc6b3 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 20 Nov 2009 22:19:44 +0100 Subject: perf: Optimize some swcounter attr.sample_period==1 paths Avoid the rather expensive perf_swevent_set_period() if we know we have to sample every single event anyway. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212508.299508332@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 17 +++++++++++------ 1 file changed, 11 insertions(+), 6 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 1dfb6cc..8e55b44 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -3759,16 +3759,16 @@ again: return nr; } -static void perf_swevent_overflow(struct perf_event *event, +static void perf_swevent_overflow(struct perf_event *event, u64 overflow, int nmi, struct perf_sample_data *data, struct pt_regs *regs) { struct hw_perf_event *hwc = &event->hw; int throttle = 0; - u64 overflow; data->period = event->hw.last_period; - overflow = perf_swevent_set_period(event); + if (!overflow) + overflow = perf_swevent_set_period(event); if (hwc->interrupts == MAX_INTERRUPTS) return; @@ -3801,14 +3801,19 @@ static void perf_swevent_add(struct perf_event *event, u64 nr, atomic64_add(nr, &event->count); + if (!regs) + return; + if (!hwc->sample_period) return; - if (!regs) + if (nr == 1 && hwc->sample_period == 1 && !event->attr.freq) + return perf_swevent_overflow(event, 1, nmi, data, regs); + + if (atomic64_add_negative(nr, &hwc->period_left)) return; - if (!atomic64_add_negative(nr, &hwc->period_left)) - perf_swevent_overflow(event, nmi, data, regs); + perf_swevent_overflow(event, 0, nmi, data, regs); } static int perf_swevent_is_counting(struct perf_event *event) -- cgit v1.1 From 81520183878a8813c71c9372de28bb70913ba549 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 20 Nov 2009 22:19:45 +0100 Subject: perf: Optimize perf_swevent_ctx_event() Remove a rcu_read_{,un}lock() pair and a few conditionals. We can remove the rcu_read_lock() by increasing the scope of one in the calling function. We can do away with the system_state check if the machine still boots after this patch (seems to be the case). We can do away with the list_empty() check because the bare list_for_each_entry_rcu() reduces to that now that we've removed everything else. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212508.378188589@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 8e55b44..cda17ac 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -3886,15 +3886,10 @@ static void perf_swevent_ctx_event(struct perf_event_context *ctx, { struct perf_event *event; - if (system_state != SYSTEM_RUNNING || list_empty(&ctx->event_list)) - return; - - rcu_read_lock(); list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { if (perf_swevent_match(event, type, event_id, data, regs)) perf_swevent_add(event, nr, nmi, data, regs); } - rcu_read_unlock(); } static int *perf_swevent_recursion_context(struct perf_cpu_context *cpuctx) @@ -3926,9 +3921,9 @@ static void do_perf_sw_event(enum perf_type_id type, u32 event_id, (*recursion)++; barrier(); + rcu_read_lock(); perf_swevent_ctx_event(&cpuctx->ctx, type, event_id, nr, nmi, data, regs); - rcu_read_lock(); /* * doesn't really matter which of the child contexts the * events ends up in. -- cgit v1.1 From d6ff86cfb50a72df820e7e839836d55d245306fb Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 20 Nov 2009 22:19:46 +0100 Subject: perf: Optimize perf_event_task_ctx() Remove a rcu_read_{,un}lock() pair and a few conditionals. We can remove the rcu_read_lock() by increasing the scope of one in the calling function. We can do away with the system_state check if the machine still boots after this patch (seems to be the case). We can do away with the list_empty() check because the bare list_for_each_entry_rcu() reduces to that now that we've removed everything else. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212508.452227115@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index cda17ac..2afb305 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -3267,15 +3267,10 @@ static void perf_event_task_ctx(struct perf_event_context *ctx, { struct perf_event *event; - if (system_state != SYSTEM_RUNNING || list_empty(&ctx->event_list)) - return; - - rcu_read_lock(); list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { if (perf_event_task_match(event)) perf_event_task_output(event, task_event); } - rcu_read_unlock(); } static void perf_event_task_event(struct perf_task_event *task_event) @@ -3283,11 +3278,11 @@ static void perf_event_task_event(struct perf_task_event *task_event) struct perf_cpu_context *cpuctx; struct perf_event_context *ctx = task_event->task_ctx; + rcu_read_lock(); cpuctx = &get_cpu_var(perf_cpu_context); perf_event_task_ctx(&cpuctx->ctx, task_event); put_cpu_var(perf_cpu_context); - rcu_read_lock(); if (!ctx) ctx = rcu_dereference(task_event->task->perf_event_ctxp); if (ctx) -- cgit v1.1 From f6595f3a9680c86b6332f881a7ae2cbbcfdc8619 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 20 Nov 2009 22:19:47 +0100 Subject: perf: Optimize perf_event_comm_ctx() Remove a rcu_read_{,un}lock() pair and a few conditionals. We can remove the rcu_read_lock() by increasing the scope of one in the calling function. We can do away with the system_state check if the machine still boots after this patch (seems to be the case). We can do away with the list_empty() check because the bare list_for_each_entry_rcu() reduces to that now that we've removed everything else. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212508.527608793@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 2afb305..4deefaa 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -3374,15 +3374,10 @@ static void perf_event_comm_ctx(struct perf_event_context *ctx, { struct perf_event *event; - if (system_state != SYSTEM_RUNNING || list_empty(&ctx->event_list)) - return; - - rcu_read_lock(); list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { if (perf_event_comm_match(event)) perf_event_comm_output(event, comm_event); } - rcu_read_unlock(); } static void perf_event_comm_event(struct perf_comm_event *comm_event) @@ -3401,11 +3396,11 @@ static void perf_event_comm_event(struct perf_comm_event *comm_event) comm_event->event_id.header.size = sizeof(comm_event->event_id) + size; + rcu_read_lock(); cpuctx = &get_cpu_var(perf_cpu_context); perf_event_comm_ctx(&cpuctx->ctx, comm_event); put_cpu_var(perf_cpu_context); - rcu_read_lock(); /* * doesn't really matter which of the child contexts the * events ends up in. -- cgit v1.1 From f6d9dd237da400effb265f3554c64413f8a3e7b4 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 20 Nov 2009 22:19:48 +0100 Subject: perf: Optimize perf_event_mmap_ctx() Remove a rcu_read_{,un}lock() pair and a few conditionals. We can remove the rcu_read_lock() by increasing the scope of one in the calling function. We can do away with the system_state check if the machine still boots after this patch (seems to be the case). We can do away with the list_empty() check because the bare list_for_each_entry_rcu() reduces to that now that we've removed everything else. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212508.606459548@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 4deefaa..68fbf4f 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -3493,15 +3493,10 @@ static void perf_event_mmap_ctx(struct perf_event_context *ctx, { struct perf_event *event; - if (system_state != SYSTEM_RUNNING || list_empty(&ctx->event_list)) - return; - - rcu_read_lock(); list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { if (perf_event_mmap_match(event, mmap_event)) perf_event_mmap_output(event, mmap_event); } - rcu_read_unlock(); } static void perf_event_mmap_event(struct perf_mmap_event *mmap_event) @@ -3557,11 +3552,11 @@ got_name: mmap_event->event_id.header.size = sizeof(mmap_event->event_id) + size; + rcu_read_lock(); cpuctx = &get_cpu_var(perf_cpu_context); perf_event_mmap_ctx(&cpuctx->ctx, mmap_event); put_cpu_var(perf_cpu_context); - rcu_read_lock(); /* * doesn't really matter which of the child contexts the * events ends up in. -- cgit v1.1 From abf4868b8548cae18d4fe8bbfb4e207443be01be Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 20 Nov 2009 22:19:49 +0100 Subject: perf: Fix PERF_FORMAT_GROUP scale info As Corey reported, the total_enabled and total_running times could occasionally be 0, even though there were events counted. It turns out this is because we record the times before reading the counter while the latter updates the times. This patch corrects that. While looking at this code I found that there is a lot of locking iffyness around, the following patches correct most of that. Reported-by: Corey Ashford Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212508.685559857@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 52 +++++++++++++++++++++------------------------------- 1 file changed, 21 insertions(+), 31 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 68fbf4f..9a18ff2 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -1784,30 +1784,15 @@ u64 perf_event_read_value(struct perf_event *event) } EXPORT_SYMBOL_GPL(perf_event_read_value); -static int perf_event_read_entry(struct perf_event *event, - u64 read_format, char __user *buf) -{ - int n = 0, count = 0; - u64 values[2]; - - values[n++] = perf_event_read_value(event); - if (read_format & PERF_FORMAT_ID) - values[n++] = primary_event_id(event); - - count = n * sizeof(u64); - - if (copy_to_user(buf, values, count)) - return -EFAULT; - - return count; -} - static int perf_event_read_group(struct perf_event *event, u64 read_format, char __user *buf) { struct perf_event *leader = event->group_leader, *sub; - int n = 0, size = 0, err = -EFAULT; - u64 values[3]; + int n = 0, size = 0, ret = 0; + u64 values[5]; + u64 count; + + count = perf_event_read_value(leader); values[n++] = 1 + leader->nr_siblings; if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) { @@ -1818,28 +1803,33 @@ static int perf_event_read_group(struct perf_event *event, values[n++] = leader->total_time_running + atomic64_read(&leader->child_total_time_running); } + values[n++] = count; + if (read_format & PERF_FORMAT_ID) + values[n++] = primary_event_id(leader); size = n * sizeof(u64); if (copy_to_user(buf, values, size)) return -EFAULT; - err = perf_event_read_entry(leader, read_format, buf + size); - if (err < 0) - return err; - - size += err; + ret += size; list_for_each_entry(sub, &leader->sibling_list, group_entry) { - err = perf_event_read_entry(sub, read_format, - buf + size); - if (err < 0) - return err; + n = 0; - size += err; + values[n++] = perf_event_read_value(sub); + if (read_format & PERF_FORMAT_ID) + values[n++] = primary_event_id(sub); + + size = n * sizeof(u64); + + if (copy_to_user(buf + size, values, size)) + return -EFAULT; + + ret += size; } - return size; + return ret; } static int perf_event_read_one(struct perf_event *event, -- cgit v1.1 From 02ffdbc866c8b1c8644601e9aa6155700eed4c91 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 20 Nov 2009 22:19:50 +0100 Subject: perf: Optimize perf_event_task_sched_out Remove an update_context_time() call from the perf_event_task_sched_out() path and into the branch its needed. The call was both superfluous, because __perf_event_sched_out() already does it, and wrong, because it was done without holding ctx->lock. Place it in perf_event_sync_stat(), which is the only place it is needed and which does already hold ctx->lock. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212508.779516394@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 9a18ff2..65f4dab 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -1120,6 +1120,8 @@ static void perf_event_sync_stat(struct perf_event_context *ctx, if (!ctx->nr_stat) return; + update_context_time(ctx); + event = list_first_entry(&ctx->event_list, struct perf_event, event_entry); @@ -1163,8 +1165,6 @@ void perf_event_task_sched_out(struct task_struct *task, if (likely(!ctx || !cpuctx->task_ctx)) return; - update_context_time(ctx); - rcu_read_lock(); parent = rcu_dereference(ctx->parent_ctx); next_ctx = next->perf_event_ctxp; -- cgit v1.1 From f6f83785222b0ee037f7be90731f62a649292b5e Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 20 Nov 2009 22:19:51 +0100 Subject: perf: Optimize __perf_event_read() Both callers actually have IRQs disabled, no need doing so again. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212508.863685796@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 65f4dab..e66f6c4 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -1517,7 +1517,6 @@ static void __perf_event_read(void *info) struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context); struct perf_event *event = info; struct perf_event_context *ctx = event->ctx; - unsigned long flags; /* * If this is a task context, we need to check whether it is @@ -1529,12 +1528,10 @@ static void __perf_event_read(void *info) if (ctx->task && cpuctx->task_ctx != ctx) return; - local_irq_save(flags); if (ctx->is_active) update_context_time(ctx); event->pmu->read(event); update_event_times(event); - local_irq_restore(flags); } static u64 perf_event_read(struct perf_event *event) -- cgit v1.1 From 3dbebf15c5d3e265f751eec72c1538a00da4be27 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 20 Nov 2009 22:19:52 +0100 Subject: perf: Simplify __perf_event_sync_stat Removes constraints from __perf_event_read() by leaving it with a single callsite; this callsite had ctx->lock held, the other one does not. Removes some superfluous code from __perf_event_sync_stat(). Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212508.918544317@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index e66f6c4..af150bb 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -1061,8 +1061,6 @@ static int context_equiv(struct perf_event_context *ctx1, && !ctx1->pin_count && !ctx2->pin_count; } -static void __perf_event_read(void *event); - static void __perf_event_sync_stat(struct perf_event *event, struct perf_event *next_event) { @@ -1080,8 +1078,8 @@ static void __perf_event_sync_stat(struct perf_event *event, */ switch (event->state) { case PERF_EVENT_STATE_ACTIVE: - __perf_event_read(event); - break; + event->pmu->read(event); + /* fall-through */ case PERF_EVENT_STATE_INACTIVE: update_event_times(event); -- cgit v1.1 From 58e5ad1de3d6ad931c84f0cc8ef0655c922f30ad Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 20 Nov 2009 22:19:53 +0100 Subject: perf: Simplify __perf_event_read cpuctx is always active, task context is always active for current the previous condition verifies that if its a task context its for current, hence we can assume ctx->is_active. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212509.000272254@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index af150bb..028619d 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -1526,10 +1526,9 @@ static void __perf_event_read(void *info) if (ctx->task && cpuctx->task_ctx != ctx) return; - if (ctx->is_active) - update_context_time(ctx); - event->pmu->read(event); + update_context_time(ctx); update_event_times(event); + event->pmu->read(event); } static u64 perf_event_read(struct perf_event *event) -- cgit v1.1 From 2b8988c9f7defe319cffe0cd362a7cd356c86f62 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 20 Nov 2009 22:19:54 +0100 Subject: perf: Fix time locking Most sites updating ctx->time and event times do so under ctx->lock, make sure they all do. This was made possible by removing the __perf_event_read() call from __perf_event_sync_stat(), which already had this lock taken. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212509.102316434@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 028619d..fdfae88 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -1526,8 +1526,11 @@ static void __perf_event_read(void *info) if (ctx->task && cpuctx->task_ctx != ctx) return; + spin_lock(&ctx->lock); update_context_time(ctx); update_event_times(event); + spin_unlock(&ctx->lock); + event->pmu->read(event); } @@ -1541,7 +1544,13 @@ static u64 perf_event_read(struct perf_event *event) smp_call_function_single(event->oncpu, __perf_event_read, event, 1); } else if (event->state == PERF_EVENT_STATE_INACTIVE) { + struct perf_event_context *ctx = event->ctx; + unsigned long flags; + + spin_lock_irqsave(&ctx->lock, flags); + update_context_time(ctx); update_event_times(event); + spin_unlock_irqrestore(&ctx->lock, flags); } return atomic64_read(&event->count); -- cgit v1.1 From 59ed446f792cc07d37b1536b9c4664d14e25e425 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 20 Nov 2009 22:19:55 +0100 Subject: perf: Fix event scaling for inherited counters Properly account the full hierarchy of counters for both the count (we already did so) and the scale times (new). Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212509.153379276@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 48 ++++++++++++++++++++++++++---------------------- 1 file changed, 26 insertions(+), 22 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index fdfae88..80f40da 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -1774,14 +1774,25 @@ static int perf_event_read_size(struct perf_event *event) return size; } -u64 perf_event_read_value(struct perf_event *event) +u64 perf_event_read_value(struct perf_event *event, u64 *enabled, u64 *running) { struct perf_event *child; u64 total = 0; + *enabled = 0; + *running = 0; + total += perf_event_read(event); - list_for_each_entry(child, &event->child_list, child_list) + *enabled += event->total_time_enabled + + atomic64_read(&event->child_total_time_enabled); + *running += event->total_time_running + + atomic64_read(&event->child_total_time_running); + + list_for_each_entry(child, &event->child_list, child_list) { total += perf_event_read(child); + *enabled += child->total_time_enabled; + *running += child->total_time_running; + } return total; } @@ -1793,19 +1804,15 @@ static int perf_event_read_group(struct perf_event *event, struct perf_event *leader = event->group_leader, *sub; int n = 0, size = 0, ret = 0; u64 values[5]; - u64 count; + u64 count, enabled, running; - count = perf_event_read_value(leader); + count = perf_event_read_value(leader, &enabled, &running); values[n++] = 1 + leader->nr_siblings; - if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) { - values[n++] = leader->total_time_enabled + - atomic64_read(&leader->child_total_time_enabled); - } - if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) { - values[n++] = leader->total_time_running + - atomic64_read(&leader->child_total_time_running); - } + if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) + values[n++] = enabled; + if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) + values[n++] = running; values[n++] = count; if (read_format & PERF_FORMAT_ID) values[n++] = primary_event_id(leader); @@ -1820,7 +1827,7 @@ static int perf_event_read_group(struct perf_event *event, list_for_each_entry(sub, &leader->sibling_list, group_entry) { n = 0; - values[n++] = perf_event_read_value(sub); + values[n++] = perf_event_read_value(sub, &enabled, &running); if (read_format & PERF_FORMAT_ID) values[n++] = primary_event_id(sub); @@ -1838,18 +1845,15 @@ static int perf_event_read_group(struct perf_event *event, static int perf_event_read_one(struct perf_event *event, u64 read_format, char __user *buf) { + u64 enabled, running; u64 values[4]; int n = 0; - values[n++] = perf_event_read_value(event); - if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) { - values[n++] = event->total_time_enabled + - atomic64_read(&event->child_total_time_enabled); - } - if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) { - values[n++] = event->total_time_running + - atomic64_read(&event->child_total_time_running); - } + values[n++] = perf_event_read_value(event, &enabled, &running); + if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) + values[n++] = enabled; + if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) + values[n++] = running; if (read_format & PERF_FORMAT_ID) values[n++] = primary_event_id(event); -- cgit v1.1 From 6f10581aeaa5543a3b7a8c7a87a064375ec357f8 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 20 Nov 2009 22:19:56 +0100 Subject: perf: Fix locking for PERF_FORMAT_GROUP We should hold event->child_mutex when iterating the inherited counters, we should hold ctx->mutex when iterating siblings. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212509.251030114@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 20 +++++++++++++------- 1 file changed, 13 insertions(+), 7 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 80f40da..3ede098 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -1782,6 +1782,7 @@ u64 perf_event_read_value(struct perf_event *event, u64 *enabled, u64 *running) *enabled = 0; *running = 0; + mutex_lock(&event->child_mutex); total += perf_event_read(event); *enabled += event->total_time_enabled + atomic64_read(&event->child_total_time_enabled); @@ -1793,6 +1794,7 @@ u64 perf_event_read_value(struct perf_event *event, u64 *enabled, u64 *running) *enabled += child->total_time_enabled; *running += child->total_time_running; } + mutex_unlock(&event->child_mutex); return total; } @@ -1802,10 +1804,12 @@ static int perf_event_read_group(struct perf_event *event, u64 read_format, char __user *buf) { struct perf_event *leader = event->group_leader, *sub; - int n = 0, size = 0, ret = 0; + int n = 0, size = 0, ret = -EFAULT; + struct perf_event_context *ctx = leader->ctx; u64 values[5]; u64 count, enabled, running; + mutex_lock(&ctx->mutex); count = perf_event_read_value(leader, &enabled, &running); values[n++] = 1 + leader->nr_siblings; @@ -1820,9 +1824,9 @@ static int perf_event_read_group(struct perf_event *event, size = n * sizeof(u64); if (copy_to_user(buf, values, size)) - return -EFAULT; + goto unlock; - ret += size; + ret = size; list_for_each_entry(sub, &leader->sibling_list, group_entry) { n = 0; @@ -1833,11 +1837,15 @@ static int perf_event_read_group(struct perf_event *event, size = n * sizeof(u64); - if (copy_to_user(buf + size, values, size)) - return -EFAULT; + if (copy_to_user(buf + size, values, size)) { + ret = -EFAULT; + goto unlock; + } ret += size; } +unlock: + mutex_unlock(&ctx->mutex); return ret; } @@ -1884,12 +1892,10 @@ perf_read_hw(struct perf_event *event, char __user *buf, size_t count) return -ENOSPC; WARN_ON_ONCE(event->ctx->parent_ctx); - mutex_lock(&event->child_mutex); if (read_format & PERF_FORMAT_GROUP) ret = perf_event_read_group(event, read_format, buf); else ret = perf_event_read_one(event, read_format, buf); - mutex_unlock(&event->child_mutex); return ret; } -- cgit v1.1 From 8904b18046c2f050107f6449e887e7c1142b9ab9 Mon Sep 17 00:00:00 2001 From: Stephane Eranian Date: Fri, 20 Nov 2009 22:19:57 +0100 Subject: perf_events: Fix default watermark calculation This patch fixes the default watermark value for the sampling buffer. With the existing calculation (watermark = max(PAGE_SIZE, max_size / 2)), no notification was ever received when the buffer was exactly 1 page. This was because you would never cross the threshold (there is no partial samples). In certain configuration, there was no possibilty detecting the problem because there was not enough space left to store the LOST record.In fact, there may be a more generic problem here. The kernel should ensure that there is alaways enough space to store one LOST record. This patch sets the default watermark to half the buffer size. With such limit, we are guaranteed to get a notification even with a single page buffer assuming no sample is bigger than a page. Signed-off-by: Stephane Eranian Signed-off-by: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091120212509.344964101@chello.nl> Signed-off-by: Ingo Molnar LKML-Reference: <1256302576-6169-1-git-send-email-eranian@gmail.com> --- kernel/perf_event.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 3ede098..718fa93 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -2340,7 +2340,7 @@ perf_mmap_data_init(struct perf_event *event, struct perf_mmap_data *data) } if (!data->watermark) - data->watermark = max_t(long, PAGE_SIZE, max_size / 2); + data->watermark = max_size / 2; rcu_assign_pointer(event->data, data); -- cgit v1.1 From ce71b9df8893ec954e56c5979df6da274f20f65e Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Sun, 22 Nov 2009 05:26:55 +0100 Subject: tracing: Use the perf recursion protection from trace event When we commit a trace to perf, we first check if we are recursing in the same buffer so that we don't mess-up the buffer with a recursing trace. But later on, we do the same check from perf to avoid commit recursion. The recursion check is desired early before we touch the buffer but we want to do this check only once. Then export the recursion protection from perf and use it from the trace events before submitting a trace. v2: Put appropriate Reported-by tag Reported-by: Peter Zijlstra Signed-off-by: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo Cc: Paul Mackerras Cc: Steven Rostedt Cc: Masami Hiramatsu Cc: Jason Baron LKML-Reference: <1258864015-10579-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 68 +++++++++++++++++++++++++------------- kernel/trace/trace_event_profile.c | 14 ++++---- kernel/trace/trace_kprobe.c | 48 ++++++++++----------------- kernel/trace/trace_syscalls.c | 47 ++++++++++---------------- 4 files changed, 88 insertions(+), 89 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 718fa93..aba8227 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -3880,34 +3880,42 @@ static void perf_swevent_ctx_event(struct perf_event_context *ctx, } } -static int *perf_swevent_recursion_context(struct perf_cpu_context *cpuctx) +/* + * Must be called with preemption disabled + */ +int perf_swevent_get_recursion_context(int **recursion) { + struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context); + if (in_nmi()) - return &cpuctx->recursion[3]; + *recursion = &cpuctx->recursion[3]; + else if (in_irq()) + *recursion = &cpuctx->recursion[2]; + else if (in_softirq()) + *recursion = &cpuctx->recursion[1]; + else + *recursion = &cpuctx->recursion[0]; - if (in_irq()) - return &cpuctx->recursion[2]; + if (**recursion) + return -1; - if (in_softirq()) - return &cpuctx->recursion[1]; + (**recursion)++; - return &cpuctx->recursion[0]; + return 0; } -static void do_perf_sw_event(enum perf_type_id type, u32 event_id, - u64 nr, int nmi, - struct perf_sample_data *data, - struct pt_regs *regs) +void perf_swevent_put_recursion_context(int *recursion) { - struct perf_cpu_context *cpuctx = &get_cpu_var(perf_cpu_context); - int *recursion = perf_swevent_recursion_context(cpuctx); - struct perf_event_context *ctx; - - if (*recursion) - goto out; + (*recursion)--; +} - (*recursion)++; - barrier(); +static void __do_perf_sw_event(enum perf_type_id type, u32 event_id, + u64 nr, int nmi, + struct perf_sample_data *data, + struct pt_regs *regs) +{ + struct perf_event_context *ctx; + struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context); rcu_read_lock(); perf_swevent_ctx_event(&cpuctx->ctx, type, event_id, @@ -3920,12 +3928,25 @@ static void do_perf_sw_event(enum perf_type_id type, u32 event_id, if (ctx) perf_swevent_ctx_event(ctx, type, event_id, nr, nmi, data, regs); rcu_read_unlock(); +} - barrier(); - (*recursion)--; +static void do_perf_sw_event(enum perf_type_id type, u32 event_id, + u64 nr, int nmi, + struct perf_sample_data *data, + struct pt_regs *regs) +{ + int *recursion; + + preempt_disable(); + + if (perf_swevent_get_recursion_context(&recursion)) + goto out; + + __do_perf_sw_event(type, event_id, nr, nmi, data, regs); + perf_swevent_put_recursion_context(recursion); out: - put_cpu_var(perf_cpu_context); + preempt_enable(); } void __perf_sw_event(u32 event_id, u64 nr, int nmi, @@ -4159,7 +4180,8 @@ void perf_tp_event(int event_id, u64 addr, u64 count, void *record, if (!regs) regs = task_pt_regs(current); - do_perf_sw_event(PERF_TYPE_TRACEPOINT, event_id, count, 1, + /* Trace events already protected against recursion */ + __do_perf_sw_event(PERF_TYPE_TRACEPOINT, event_id, count, 1, &data, regs); } EXPORT_SYMBOL_GPL(perf_tp_event); diff --git a/kernel/trace/trace_event_profile.c b/kernel/trace/trace_event_profile.c index e0d351b..d9c60f8 100644 --- a/kernel/trace/trace_event_profile.c +++ b/kernel/trace/trace_event_profile.c @@ -9,31 +9,33 @@ #include "trace.h" -struct perf_trace_buf *perf_trace_buf; +char *perf_trace_buf; EXPORT_SYMBOL_GPL(perf_trace_buf); -struct perf_trace_buf *perf_trace_buf_nmi; +char *perf_trace_buf_nmi; EXPORT_SYMBOL_GPL(perf_trace_buf_nmi); +typedef typeof(char [FTRACE_MAX_PROFILE_SIZE]) perf_trace_t ; + /* Count the events in use (per event id, not per instance) */ static int total_profile_count; static int ftrace_profile_enable_event(struct ftrace_event_call *event) { - struct perf_trace_buf *buf; + char *buf; int ret = -ENOMEM; if (atomic_inc_return(&event->profile_count)) return 0; if (!total_profile_count) { - buf = alloc_percpu(struct perf_trace_buf); + buf = (char *)alloc_percpu(perf_trace_t); if (!buf) goto fail_buf; rcu_assign_pointer(perf_trace_buf, buf); - buf = alloc_percpu(struct perf_trace_buf); + buf = (char *)alloc_percpu(perf_trace_t); if (!buf) goto fail_buf_nmi; @@ -79,7 +81,7 @@ int ftrace_profile_enable(int event_id) static void ftrace_profile_disable_event(struct ftrace_event_call *event) { - struct perf_trace_buf *buf, *nmi_buf; + char *buf, *nmi_buf; if (!atomic_add_negative(-1, &event->profile_count)) return; diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 3696476..22e6f68 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -1208,11 +1208,12 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp, struct trace_probe *tp = container_of(kp, struct trace_probe, rp.kp); struct ftrace_event_call *call = &tp->call; struct kprobe_trace_entry *entry; - struct perf_trace_buf *trace_buf; struct trace_entry *ent; int size, __size, i, pc, __cpu; unsigned long irq_flags; + char *trace_buf; char *raw_data; + int *recursion; pc = preempt_count(); __size = SIZEOF_KPROBE_TRACE_ENTRY(tp->nr_args); @@ -1227,6 +1228,10 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp, * This also protects the rcu read side */ local_irq_save(irq_flags); + + if (perf_swevent_get_recursion_context(&recursion)) + goto end_recursion; + __cpu = smp_processor_id(); if (in_nmi()) @@ -1237,18 +1242,7 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp, if (!trace_buf) goto end; - trace_buf = per_cpu_ptr(trace_buf, __cpu); - - if (trace_buf->recursion++) - goto end_recursion; - - /* - * Make recursion update visible before entering perf_tp_event - * so that we protect from perf recursions. - */ - barrier(); - - raw_data = trace_buf->buf; + raw_data = per_cpu_ptr(trace_buf, __cpu); /* Zero dead bytes from alignment to avoid buffer leak to userspace */ *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; @@ -1263,9 +1257,9 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp, entry->args[i] = call_fetch(&tp->args[i].fetch, regs); perf_tp_event(call->id, entry->ip, 1, entry, size); -end_recursion: - trace_buf->recursion--; end: + perf_swevent_put_recursion_context(recursion); +end_recursion: local_irq_restore(irq_flags); return 0; @@ -1278,10 +1272,11 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri, struct trace_probe *tp = container_of(ri->rp, struct trace_probe, rp); struct ftrace_event_call *call = &tp->call; struct kretprobe_trace_entry *entry; - struct perf_trace_buf *trace_buf; struct trace_entry *ent; int size, __size, i, pc, __cpu; unsigned long irq_flags; + char *trace_buf; + int *recursion; char *raw_data; pc = preempt_count(); @@ -1297,6 +1292,10 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri, * This also protects the rcu read side */ local_irq_save(irq_flags); + + if (perf_swevent_get_recursion_context(&recursion)) + goto end_recursion; + __cpu = smp_processor_id(); if (in_nmi()) @@ -1307,18 +1306,7 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri, if (!trace_buf) goto end; - trace_buf = per_cpu_ptr(trace_buf, __cpu); - - if (trace_buf->recursion++) - goto end_recursion; - - /* - * Make recursion update visible before entering perf_tp_event - * so that we protect from perf recursions. - */ - barrier(); - - raw_data = trace_buf->buf; + raw_data = per_cpu_ptr(trace_buf, __cpu); /* Zero dead bytes from alignment to avoid buffer leak to userspace */ *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; @@ -1334,9 +1322,9 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri, entry->args[i] = call_fetch(&tp->args[i].fetch, regs); perf_tp_event(call->id, entry->ret_ip, 1, entry, size); -end_recursion: - trace_buf->recursion--; end: + perf_swevent_put_recursion_context(recursion); +end_recursion: local_irq_restore(irq_flags); return 0; diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index 51213b0..0bb9348 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -477,10 +477,11 @@ static int sys_prof_refcount_exit; static void prof_syscall_enter(struct pt_regs *regs, long id) { struct syscall_metadata *sys_data; - struct perf_trace_buf *trace_buf; struct syscall_trace_enter *rec; unsigned long flags; + char *trace_buf; char *raw_data; + int *recursion; int syscall_nr; int size; int cpu; @@ -505,6 +506,9 @@ static void prof_syscall_enter(struct pt_regs *regs, long id) /* Protect the per cpu buffer, begin the rcu read side */ local_irq_save(flags); + if (perf_swevent_get_recursion_context(&recursion)) + goto end_recursion; + cpu = smp_processor_id(); if (in_nmi()) @@ -515,18 +519,7 @@ static void prof_syscall_enter(struct pt_regs *regs, long id) if (!trace_buf) goto end; - trace_buf = per_cpu_ptr(trace_buf, cpu); - - if (trace_buf->recursion++) - goto end_recursion; - - /* - * Make recursion update visible before entering perf_tp_event - * so that we protect from perf recursions. - */ - barrier(); - - raw_data = trace_buf->buf; + raw_data = per_cpu_ptr(trace_buf, cpu); /* zero the dead bytes from align to not leak stack to user */ *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; @@ -539,9 +532,9 @@ static void prof_syscall_enter(struct pt_regs *regs, long id) (unsigned long *)&rec->args); perf_tp_event(sys_data->enter_id, 0, 1, rec, size); -end_recursion: - trace_buf->recursion--; end: + perf_swevent_put_recursion_context(recursion); +end_recursion: local_irq_restore(flags); } @@ -588,10 +581,11 @@ static void prof_syscall_exit(struct pt_regs *regs, long ret) { struct syscall_metadata *sys_data; struct syscall_trace_exit *rec; - struct perf_trace_buf *trace_buf; unsigned long flags; int syscall_nr; + char *trace_buf; char *raw_data; + int *recursion; int size; int cpu; @@ -617,6 +611,10 @@ static void prof_syscall_exit(struct pt_regs *regs, long ret) /* Protect the per cpu buffer, begin the rcu read side */ local_irq_save(flags); + + if (perf_swevent_get_recursion_context(&recursion)) + goto end_recursion; + cpu = smp_processor_id(); if (in_nmi()) @@ -627,18 +625,7 @@ static void prof_syscall_exit(struct pt_regs *regs, long ret) if (!trace_buf) goto end; - trace_buf = per_cpu_ptr(trace_buf, cpu); - - if (trace_buf->recursion++) - goto end_recursion; - - /* - * Make recursion update visible before entering perf_tp_event - * so that we protect from perf recursions. - */ - barrier(); - - raw_data = trace_buf->buf; + raw_data = per_cpu_ptr(trace_buf, cpu); /* zero the dead bytes from align to not leak stack to user */ *(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL; @@ -652,9 +639,9 @@ static void prof_syscall_exit(struct pt_regs *regs, long ret) perf_tp_event(sys_data->exit_id, 0, 1, rec, size); -end_recursion: - trace_buf->recursion--; end: + perf_swevent_put_recursion_context(recursion); +end_recursion: local_irq_restore(flags); } -- cgit v1.1 From 28889bf9e2db29747d58cd47a92d727f927c3aee Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Sun, 22 Nov 2009 05:21:33 +0100 Subject: tracing: Forget about the NMI buffer for syscall events We are never in an NMI context when we commit a syscall trace to perf. So just forget about the nmi buffer there. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Paul Mackerras Cc: Steven Rostedt Cc: Jason Baron LKML-Reference: <1258863695-10464-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_syscalls.c | 10 ++-------- 1 file changed, 2 insertions(+), 8 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index 0bb9348..41b6dd9 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -511,10 +511,7 @@ static void prof_syscall_enter(struct pt_regs *regs, long id) cpu = smp_processor_id(); - if (in_nmi()) - trace_buf = rcu_dereference(perf_trace_buf_nmi); - else - trace_buf = rcu_dereference(perf_trace_buf); + trace_buf = rcu_dereference(perf_trace_buf); if (!trace_buf) goto end; @@ -617,10 +614,7 @@ static void prof_syscall_exit(struct pt_regs *regs, long ret) cpu = smp_processor_id(); - if (in_nmi()) - trace_buf = rcu_dereference(perf_trace_buf_nmi); - else - trace_buf = rcu_dereference(perf_trace_buf); + trace_buf = rcu_dereference(perf_trace_buf); if (!trace_buf) goto end; -- cgit v1.1 From b3a75542d329ce4e1c66b293cefeb4429a2af043 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Sun, 22 Nov 2009 05:21:34 +0100 Subject: hw-breakpoints: Remove x86 specific headers from core file Remove asm/processor.h and asm/debugreg.h as these headers are not used anymore in the hw-breakpoints core file. Signed-off-by: Frederic Weisbecker Cc: Prasad LKML-Reference: <1258863695-10464-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/hw_breakpoint.c | 6 ------ 1 file changed, 6 deletions(-) (limited to 'kernel') diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index 9ea9414..b6d6fa2 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -40,12 +40,6 @@ #include -#include - -#ifdef CONFIG_X86 -#include -#endif - /* * Constraints data */ -- cgit v1.1 From 96b02d78a7e47cd189f6b307c5513fec6b2155dc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?M=C3=A1rton=20N=C3=A9meth?= Date: Sat, 21 Nov 2009 23:10:15 +0100 Subject: perf_event: Remove redundant zero fill MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The buffer is first zeroed out by memset(). Then strncpy() is used to fill the content. The strncpy() function also pads the string till the end of the specified length, which is redundant. The strncpy() does not ensures that the string will be properly closed with 0. Use strlcpy() instead. The semantic match that finds this kind of pattern is as follows: (http://coccinelle.lip6.fr/) // @@ expression buffer; expression size; expression str; @@ memset(buffer, 0, size); ... - strncpy( + strlcpy( buffer, str, sizeof(buffer) ); @@ expression buffer; expression size; expression str; @@ memset(&buffer, 0, size); ... - strncpy( + strlcpy( &buffer, str, sizeof(buffer)); @@ expression buffer; identifier field; expression size; expression str; @@ memset(buffer, 0, size); ... - strncpy( + strlcpy( buffer->field, str, sizeof(buffer->field) ); @@ expression buffer; identifier field; expression size; expression str; @@ memset(&buffer, 0, size); ... - strncpy( + strlcpy( buffer.field, str, sizeof(buffer.field)); // On strncpy() vs strlcpy() see http://www.gratisoft.us/todd/papers/strlcpy.html . Signed-off-by: Márton Németh Cc: Julia Lawall Cc: cocci@diku.dk Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <4B086547.5040100@freemail.hu> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index aba8227..b26cb03 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -3391,7 +3391,7 @@ static void perf_event_comm_event(struct perf_comm_event *comm_event) char comm[TASK_COMM_LEN]; memset(comm, 0, sizeof(comm)); - strncpy(comm, comm_event->task->comm, sizeof(comm)); + strlcpy(comm, comm_event->task->comm, sizeof(comm)); size = ALIGN(strlen(comm)+1, sizeof(u64)); comm_event->comm = comm; -- cgit v1.1 From 645e8cc0c9f01f07f384fd522b782e5e6ae9de18 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Sun, 22 Nov 2009 12:20:19 +0100 Subject: perf_events: Fix modular build Fix: ERROR: "perf_swevent_put_recursion_context" [fs/ext4/ext4.ko] undefined! ERROR: "perf_swevent_get_recursion_context" [fs/ext4/ext4.ko] undefined! Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Paul Mackerras Cc: Steven Rostedt Cc: Masami Hiramatsu Cc: Jason Baron LKML-Reference: <1258864015-10579-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index b26cb03..abe1ef4 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -3903,11 +3903,13 @@ int perf_swevent_get_recursion_context(int **recursion) return 0; } +EXPORT_SYMBOL_GPL(perf_swevent_get_recursion_context); void perf_swevent_put_recursion_context(int *recursion) { (*recursion)--; } +EXPORT_SYMBOL_GPL(perf_swevent_put_recursion_context); static void __do_perf_sw_event(enum perf_type_id type, u32 event_id, u64 nr, int nmi, -- cgit v1.1 From 6e3d8330ae2c4b2c11a9577a0130d2ecda1c610d Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Mon, 23 Nov 2009 10:19:20 +0100 Subject: perf events: Do not generate function trace entries in perf code Decreases perf overhead when function tracing is enabled, by about 50%. Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: Signed-off-by: Ingo Molnar --- kernel/Makefile | 1 + 1 file changed, 1 insertion(+) (limited to 'kernel') diff --git a/kernel/Makefile b/kernel/Makefile index 17b575e..6b7ce81 100644 --- a/kernel/Makefile +++ b/kernel/Makefile @@ -21,6 +21,7 @@ CFLAGS_REMOVE_mutex-debug.o = -pg CFLAGS_REMOVE_rtmutex-debug.o = -pg CFLAGS_REMOVE_cgroup-debug.o = -pg CFLAGS_REMOVE_sched_clock.o = -pg +CFLAGS_REMOVE_perf_event.o = -pg endif obj-$(CONFIG_FREEZER) += freezer.o -- cgit v1.1 From a4234bfcf4d72a10a99176cdef007345e9c3b4aa Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Mon, 23 Nov 2009 10:57:59 +0100 Subject: perf_events: Optimize the swcounter hotpath The structure init creates a bit memcpy, which shows up big time in perf annotate output: : ffffffff810a859d <__perf_sw_event>: 1.68 : ffffffff810a859d: 55 push %rbp 1.69 : ffffffff810a859e: 41 89 fa mov %edi,%r10d 0.01 : ffffffff810a85a1: 49 89 c9 mov %rcx,%r9 0.00 : ffffffff810a85a4: 31 c0 xor %eax,%eax 1.71 : ffffffff810a85a6: b9 16 00 00 00 mov $0x16,%ecx 0.00 : ffffffff810a85ab: 48 89 e5 mov %rsp,%rbp 0.00 : ffffffff810a85ae: 48 83 ec 60 sub $0x60,%rsp 1.52 : ffffffff810a85b2: 48 8d 7d a0 lea -0x60(%rbp),%rdi 85.20 : ffffffff810a85b6: f3 ab rep stos %eax,%es:(%rdi) None of the callees depends on the structure being pre-initialized, so only initialize ->addr. This gets rid of the memcpy overhead. Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index abe1ef4..20df8ab 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -3954,12 +3954,12 @@ out: void __perf_sw_event(u32 event_id, u64 nr, int nmi, struct pt_regs *regs, u64 addr) { - struct perf_sample_data data = { - .addr = addr, - }; + struct perf_sample_data data; - do_perf_sw_event(PERF_TYPE_SOFTWARE, event_id, nr, nmi, - &data, regs); + data.addr = addr; + data.raw = NULL; + + do_perf_sw_event(PERF_TYPE_SOFTWARE, event_id, nr, nmi, &data, regs); } static void perf_swevent_read(struct perf_event *event) -- cgit v1.1 From a66a3052e2d4c5815d7ad26887b1d4193206e691 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 23 Nov 2009 11:37:23 +0100 Subject: perf_events: Undo copy/paste damage We had two almost identical functions, avoid the duplication. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091123103819.537537928@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 31 +++++++++---------------------- 1 file changed, 9 insertions(+), 22 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 20df8ab..e2daa10 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -1704,16 +1704,10 @@ static void free_event(struct perf_event *event) call_rcu(&event->rcu_head, free_event_rcu); } -/* - * Called when the last reference to the file is gone. - */ -static int perf_release(struct inode *inode, struct file *file) +int perf_event_release_kernel(struct perf_event *event) { - struct perf_event *event = file->private_data; struct perf_event_context *ctx = event->ctx; - file->private_data = NULL; - WARN_ON_ONCE(ctx->parent_ctx); mutex_lock(&ctx->mutex); perf_event_remove_from_context(event); @@ -1728,26 +1722,19 @@ static int perf_release(struct inode *inode, struct file *file) return 0; } +EXPORT_SYMBOL_GPL(perf_event_release_kernel); -int perf_event_release_kernel(struct perf_event *event) +/* + * Called when the last reference to the file is gone. + */ +static int perf_release(struct inode *inode, struct file *file) { - struct perf_event_context *ctx = event->ctx; - - WARN_ON_ONCE(ctx->parent_ctx); - mutex_lock(&ctx->mutex); - perf_event_remove_from_context(event); - mutex_unlock(&ctx->mutex); - - mutex_lock(&event->owner->perf_event_mutex); - list_del_init(&event->owner_entry); - mutex_unlock(&event->owner->perf_event_mutex); - put_task_struct(event->owner); + struct perf_event *event = file->private_data; - free_event(event); + file->private_data = NULL; - return 0; + return perf_event_release_kernel(event); } -EXPORT_SYMBOL_GPL(perf_event_release_kernel); static int perf_event_read_size(struct perf_event *event) { -- cgit v1.1 From 6c2bfcbe58e0dd39554be88940149f5aa11e17d1 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 23 Nov 2009 11:37:24 +0100 Subject: perf_events: Fix style nits Signed-off-by: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker LKML-Reference: <20091123103819.613427378@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index e2daa10..1f14481 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -447,9 +447,8 @@ retry: * can remove the event safely, if the call above did not * succeed. */ - if (!list_empty(&event->group_entry)) { + if (!list_empty(&event->group_entry)) list_del_event(event, ctx); - } spin_unlock_irq(&ctx->lock); } @@ -1033,10 +1032,10 @@ void __perf_event_sched_out(struct perf_event_context *ctx, update_context_time(ctx); perf_disable(); - if (ctx->nr_active) + if (ctx->nr_active) { list_for_each_entry(event, &ctx->group_list, group_entry) group_sched_out(event, cpuctx, ctx); - + } perf_enable(); out: spin_unlock(&ctx->lock); -- cgit v1.1 From 2e2af50b1fab3c40636839a7f439c167ae559533 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 23 Nov 2009 11:37:25 +0100 Subject: perf_events: Disable events when we detach them If we leave the event in STATE_INACTIVE, any read of the event after the detach will increase the running count but not the enabled count and cause funny scaling artefacts. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker LKML-Reference: <20091123103819.689055515@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 1f14481..fb851ec 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -294,6 +294,8 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx) if (event->group_leader != event) event->group_leader->nr_siblings--; + event->state = PERF_EVENT_STATE_OFF; + /* * If this was a group event with sibling events then * upgrade the siblings to singleton events by adding them -- cgit v1.1 From 5e942bb33371254a474653123cd9e13a4c89ee44 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 23 Nov 2009 11:37:26 +0100 Subject: perf_events: Update the context time on exit It appeared we did call update_event_times() on exit, but we failed to update the context time, which renders the former moot. Locking is a bit iffy, we call update_event_times under ctx->mutex instead of ctx->lock - the next patch fixes this. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker LKML-Reference: <20091123103819.764207355@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 1 + 1 file changed, 1 insertion(+) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index fb851ec..8be2574 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -4983,6 +4983,7 @@ void perf_event_exit_task(struct task_struct *child) * the events from it. */ unclone_ctx(child_ctx); + update_context_time(child_ctx); spin_unlock_irqrestore(&child_ctx->lock, flags); /* -- cgit v1.1 From f67218c3e93abaf0f480bb94b53d234853ffe4de Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 23 Nov 2009 11:37:27 +0100 Subject: perf_events: Fix __perf_event_exit_task() vs. update_event_times() locking Move the update_event_times() call in __perf_event_exit_task() into list_del_event() because that holds the proper lock (ctx->lock) and seems a more natural place to do the last time update. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker LKML-Reference: <20091123103819.842455480@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 78 ++++++++++++++++++++++++++--------------------------- 1 file changed, 39 insertions(+), 39 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 8be2574..50f11b5 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -246,6 +246,44 @@ static void perf_unpin_context(struct perf_event_context *ctx) put_ctx(ctx); } +static inline u64 perf_clock(void) +{ + return cpu_clock(smp_processor_id()); +} + +/* + * Update the record of the current time in a context. + */ +static void update_context_time(struct perf_event_context *ctx) +{ + u64 now = perf_clock(); + + ctx->time += now - ctx->timestamp; + ctx->timestamp = now; +} + +/* + * Update the total_time_enabled and total_time_running fields for a event. + */ +static void update_event_times(struct perf_event *event) +{ + struct perf_event_context *ctx = event->ctx; + u64 run_end; + + if (event->state < PERF_EVENT_STATE_INACTIVE || + event->group_leader->state < PERF_EVENT_STATE_INACTIVE) + return; + + event->total_time_enabled = ctx->time - event->tstamp_enabled; + + if (event->state == PERF_EVENT_STATE_INACTIVE) + run_end = event->tstamp_stopped; + else + run_end = ctx->time; + + event->total_time_running = run_end - event->tstamp_running; +} + /* * Add a event from the lists for its context. * Must be called with ctx->mutex and ctx->lock held. @@ -294,6 +332,7 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx) if (event->group_leader != event) event->group_leader->nr_siblings--; + update_event_times(event); event->state = PERF_EVENT_STATE_OFF; /* @@ -454,44 +493,6 @@ retry: spin_unlock_irq(&ctx->lock); } -static inline u64 perf_clock(void) -{ - return cpu_clock(smp_processor_id()); -} - -/* - * Update the record of the current time in a context. - */ -static void update_context_time(struct perf_event_context *ctx) -{ - u64 now = perf_clock(); - - ctx->time += now - ctx->timestamp; - ctx->timestamp = now; -} - -/* - * Update the total_time_enabled and total_time_running fields for a event. - */ -static void update_event_times(struct perf_event *event) -{ - struct perf_event_context *ctx = event->ctx; - u64 run_end; - - if (event->state < PERF_EVENT_STATE_INACTIVE || - event->group_leader->state < PERF_EVENT_STATE_INACTIVE) - return; - - event->total_time_enabled = ctx->time - event->tstamp_enabled; - - if (event->state == PERF_EVENT_STATE_INACTIVE) - run_end = event->tstamp_stopped; - else - run_end = ctx->time; - - event->total_time_running = run_end - event->tstamp_running; -} - /* * Update total_time_enabled and total_time_running for all events in a group. */ @@ -4931,7 +4932,6 @@ __perf_event_exit_task(struct perf_event *child_event, { struct perf_event *parent_event; - update_event_times(child_event); perf_event_remove_from_context(child_event); parent_event = child_event->parent; -- cgit v1.1 From 4ed7c92d68a5387ba5f7030dc76eab03558e27f5 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 23 Nov 2009 11:37:29 +0100 Subject: perf_events: Undo some recursion damage Make perf_swevent_get_recursion_context return a context number and disable preemption. This could be used to remove the IRQ disable from the trace bit and index the per-cpu buffer with. Signed-off-by: Peter Zijlstra Cc: Frederic Weisbecker Cc: Paul Mackerras LKML-Reference: <20091123103819.993226816@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 71 +++++++++++++++++++------------------------ kernel/trace/trace_kprobe.c | 14 +++++---- kernel/trace/trace_syscalls.c | 14 +++++---- 3 files changed, 48 insertions(+), 51 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 50f11b5..0b0d5f7 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -3869,45 +3869,50 @@ static void perf_swevent_ctx_event(struct perf_event_context *ctx, } } -/* - * Must be called with preemption disabled - */ -int perf_swevent_get_recursion_context(int **recursion) +int perf_swevent_get_recursion_context(void) { - struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context); + struct perf_cpu_context *cpuctx = &get_cpu_var(perf_cpu_context); + int rctx; if (in_nmi()) - *recursion = &cpuctx->recursion[3]; + rctx = 3; else if (in_irq()) - *recursion = &cpuctx->recursion[2]; + rctx = 2; else if (in_softirq()) - *recursion = &cpuctx->recursion[1]; + rctx = 1; else - *recursion = &cpuctx->recursion[0]; + rctx = 0; - if (**recursion) + if (cpuctx->recursion[rctx]) { + put_cpu_var(perf_cpu_context); return -1; + } - (**recursion)++; + cpuctx->recursion[rctx]++; + barrier(); - return 0; + return rctx; } EXPORT_SYMBOL_GPL(perf_swevent_get_recursion_context); -void perf_swevent_put_recursion_context(int *recursion) +void perf_swevent_put_recursion_context(int rctx) { - (*recursion)--; + struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context); + barrier(); + cpuctx->recursion[rctx]++; + put_cpu_var(perf_cpu_context); } EXPORT_SYMBOL_GPL(perf_swevent_put_recursion_context); -static void __do_perf_sw_event(enum perf_type_id type, u32 event_id, - u64 nr, int nmi, - struct perf_sample_data *data, - struct pt_regs *regs) +static void do_perf_sw_event(enum perf_type_id type, u32 event_id, + u64 nr, int nmi, + struct perf_sample_data *data, + struct pt_regs *regs) { + struct perf_cpu_context *cpuctx; struct perf_event_context *ctx; - struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context); + cpuctx = &__get_cpu_var(perf_cpu_context); rcu_read_lock(); perf_swevent_ctx_event(&cpuctx->ctx, type, event_id, nr, nmi, data, regs); @@ -3921,34 +3926,22 @@ static void __do_perf_sw_event(enum perf_type_id type, u32 event_id, rcu_read_unlock(); } -static void do_perf_sw_event(enum perf_type_id type, u32 event_id, - u64 nr, int nmi, - struct perf_sample_data *data, - struct pt_regs *regs) -{ - int *recursion; - - preempt_disable(); - - if (perf_swevent_get_recursion_context(&recursion)) - goto out; - - __do_perf_sw_event(type, event_id, nr, nmi, data, regs); - - perf_swevent_put_recursion_context(recursion); -out: - preempt_enable(); -} - void __perf_sw_event(u32 event_id, u64 nr, int nmi, struct pt_regs *regs, u64 addr) { struct perf_sample_data data; + int rctx; + + rctx = perf_swevent_get_recursion_context(); + if (rctx < 0) + return; data.addr = addr; data.raw = NULL; do_perf_sw_event(PERF_TYPE_SOFTWARE, event_id, nr, nmi, &data, regs); + + perf_swevent_put_recursion_context(rctx); } static void perf_swevent_read(struct perf_event *event) @@ -4172,7 +4165,7 @@ void perf_tp_event(int event_id, u64 addr, u64 count, void *record, regs = task_pt_regs(current); /* Trace events already protected against recursion */ - __do_perf_sw_event(PERF_TYPE_TRACEPOINT, event_id, count, 1, + do_perf_sw_event(PERF_TYPE_TRACEPOINT, event_id, count, 1, &data, regs); } EXPORT_SYMBOL_GPL(perf_tp_event); diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 22e6f68..79ce6a2 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -1213,7 +1213,7 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp, unsigned long irq_flags; char *trace_buf; char *raw_data; - int *recursion; + int rctx; pc = preempt_count(); __size = SIZEOF_KPROBE_TRACE_ENTRY(tp->nr_args); @@ -1229,7 +1229,8 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp, */ local_irq_save(irq_flags); - if (perf_swevent_get_recursion_context(&recursion)) + rctx = perf_swevent_get_recursion_context(); + if (rctx < 0) goto end_recursion; __cpu = smp_processor_id(); @@ -1258,7 +1259,7 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp, perf_tp_event(call->id, entry->ip, 1, entry, size); end: - perf_swevent_put_recursion_context(recursion); + perf_swevent_put_recursion_context(rctx); end_recursion: local_irq_restore(irq_flags); @@ -1276,8 +1277,8 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri, int size, __size, i, pc, __cpu; unsigned long irq_flags; char *trace_buf; - int *recursion; char *raw_data; + int rctx; pc = preempt_count(); __size = SIZEOF_KRETPROBE_TRACE_ENTRY(tp->nr_args); @@ -1293,7 +1294,8 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri, */ local_irq_save(irq_flags); - if (perf_swevent_get_recursion_context(&recursion)) + rctx = perf_swevent_get_recursion_context(); + if (rctx < 0) goto end_recursion; __cpu = smp_processor_id(); @@ -1323,7 +1325,7 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri, perf_tp_event(call->id, entry->ret_ip, 1, entry, size); end: - perf_swevent_put_recursion_context(recursion); + perf_swevent_put_recursion_context(rctx); end_recursion: local_irq_restore(irq_flags); diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index 41b6dd9..9189cbe 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -481,8 +481,8 @@ static void prof_syscall_enter(struct pt_regs *regs, long id) unsigned long flags; char *trace_buf; char *raw_data; - int *recursion; int syscall_nr; + int rctx; int size; int cpu; @@ -506,7 +506,8 @@ static void prof_syscall_enter(struct pt_regs *regs, long id) /* Protect the per cpu buffer, begin the rcu read side */ local_irq_save(flags); - if (perf_swevent_get_recursion_context(&recursion)) + rctx = perf_swevent_get_recursion_context(); + if (rctx < 0) goto end_recursion; cpu = smp_processor_id(); @@ -530,7 +531,7 @@ static void prof_syscall_enter(struct pt_regs *regs, long id) perf_tp_event(sys_data->enter_id, 0, 1, rec, size); end: - perf_swevent_put_recursion_context(recursion); + perf_swevent_put_recursion_context(rctx); end_recursion: local_irq_restore(flags); } @@ -582,7 +583,7 @@ static void prof_syscall_exit(struct pt_regs *regs, long ret) int syscall_nr; char *trace_buf; char *raw_data; - int *recursion; + int rctx; int size; int cpu; @@ -609,7 +610,8 @@ static void prof_syscall_exit(struct pt_regs *regs, long ret) /* Protect the per cpu buffer, begin the rcu read side */ local_irq_save(flags); - if (perf_swevent_get_recursion_context(&recursion)) + rctx = perf_swevent_get_recursion_context(); + if (rctx < 0) goto end_recursion; cpu = smp_processor_id(); @@ -634,7 +636,7 @@ static void prof_syscall_exit(struct pt_regs *regs, long ret) perf_tp_event(sys_data->exit_id, 0, 1, rec, size); end: - perf_swevent_put_recursion_context(recursion); + perf_swevent_put_recursion_context(rctx); end_recursion: local_irq_restore(flags); } -- cgit v1.1 From acd1d7c1f8f3d848a3c5327dc09f8c1efb971678 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 23 Nov 2009 15:00:36 +0100 Subject: perf_events: Restore sanity to scaling land It is quite possible to call update_event_times() on a context that isn't actually running and thereby confuse the thing. perf stat was reporting !100% scale values for software counters (2e2af50b perf_events: Disable events when we detach them, solved the worst of that, but there was still some left). The thing that happens is that because we are not self-reaping (we have a caring parent) there is a time between the last schedule (out) and having do_exit() called which will detach the events. This period would be accounted as enabled,!running because the event->state==INACTIVE, even though !event->ctx->is_active. Similar issues could have been observed by calling read() on a event while the attached task was not scheduled in. Solve this by teaching update_event_times() about ctx->is_active. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <1258984836.4531.480.camel@laptop> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 0b0d5f7..0aafe85 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -274,7 +274,12 @@ static void update_event_times(struct perf_event *event) event->group_leader->state < PERF_EVENT_STATE_INACTIVE) return; - event->total_time_enabled = ctx->time - event->tstamp_enabled; + if (ctx->is_active) + run_end = ctx->time; + else + run_end = event->tstamp_stopped; + + event->total_time_enabled = run_end - event->tstamp_enabled; if (event->state == PERF_EVENT_STATE_INACTIVE) run_end = event->tstamp_stopped; -- cgit v1.1 From ba6909b719a5ccc0c8100d2895bb7ff557b2eeae Mon Sep 17 00:00:00 2001 From: "K.Prasad" Date: Mon, 23 Nov 2009 21:17:13 +0530 Subject: hw-breakpoint: Attribute authorship of hw-breakpoint related files Attribute authorship to developers of hw-breakpoint related files. Signed-off-by: K.Prasad Cc: Alan Stern Cc: Frederic Weisbecker LKML-Reference: <20091123154713.GA5593@in.ibm.com> [ v2: moved it to latest -tip ] Signed-off-by: Ingo Molnar --- kernel/hw_breakpoint.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'kernel') diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index b6d6fa2..c166622 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -18,6 +18,10 @@ * Copyright (C) 2009, Frederic Weisbecker * * Thanks to Ingo Molnar for his many suggestions. + * + * Authors: Alan Stern + * K.Prasad + * Frederic Weisbecker */ /* -- cgit v1.1 From fdf6bc95229821e3d9405eba28925b76e92b74d0 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Mon, 23 Nov 2009 15:42:33 +0100 Subject: hw-breakpoints: Check the breakpoint params from perf tools Perf tools create perf events as disabled in the beginning. Breakpoints are then considered like ptrace temporary breakpoints, only meant to reserve a breakpoint slot until we get all the necessary informations from the user. In this case, we don't check the address that is breakpointed as it is NULL in the ptrace case. But perf tools don't have the same purpose, events are created disabled to wait for all events to be created before enabling all of them. We want to check the breakpoint parameters in this case. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Paul Mackerras Cc: Prasad LKML-Reference: <1258987355-8751-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/hw_breakpoint.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index c166622..06d372f 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -267,7 +267,16 @@ int __register_perf_hw_breakpoint(struct perf_event *bp) if (ret) return ret; - if (!bp->attr.disabled) + /* + * Ptrace breakpoints can be temporary perf events only + * meant to reserve a slot. In this case, it is created disabled and + * we don't want to check the params right now (as we put a null addr) + * But perf tools create events as disabled and we want to check + * the params for them. + * This is a quick hack that will be removed soon, once we remove + * the tmp breakpoints from ptrace + */ + if (!bp->attr.disabled || bp->callback == perf_bp_event) ret = arch_validate_hwbkpt_settings(bp, bp->ctx->task); return ret; -- cgit v1.1 From f5ffe02e5046003ae7e2ce70d3d1c2a73331268b Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Mon, 23 Nov 2009 15:42:34 +0100 Subject: perf: Add kernel side syscall events support for breakpoints Add the remaining necessary bits to support breakpoints created through perf syscall. We don't use the software counter interface as: - We don't need to check against recursion, this is already done in hardware breakpoints arch level. - We already know the perf event we are dealing with when the event is to be committed. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Paul Mackerras Cc: Prasad LKML-Reference: <1258987355-8751-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 34 +++++++++++++++++++++++++--------- 1 file changed, 25 insertions(+), 9 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 0aafe85..9425c96 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -3831,6 +3831,20 @@ static int perf_swevent_is_counting(struct perf_event *event) static int perf_tp_event_match(struct perf_event *event, struct perf_sample_data *data); +static int perf_exclude_event(struct perf_event *event, + struct pt_regs *regs) +{ + if (regs) { + if (event->attr.exclude_user && user_mode(regs)) + return 1; + + if (event->attr.exclude_kernel && !user_mode(regs)) + return 1; + } + + return 0; +} + static int perf_swevent_match(struct perf_event *event, enum perf_type_id type, u32 event_id, @@ -3842,16 +3856,12 @@ static int perf_swevent_match(struct perf_event *event, if (event->attr.type != type) return 0; + if (event->attr.config != event_id) return 0; - if (regs) { - if (event->attr.exclude_user && user_mode(regs)) - return 0; - - if (event->attr.exclude_kernel && !user_mode(regs)) - return 0; - } + if (perf_exclude_event(event, regs)) + return 0; if (event->attr.type == PERF_TYPE_TRACEPOINT && !perf_tp_event_match(event, data)) @@ -4282,9 +4292,15 @@ static const struct pmu *bp_perf_event_init(struct perf_event *bp) return &perf_ops_bp; } -void perf_bp_event(struct perf_event *bp, void *regs) +void perf_bp_event(struct perf_event *bp, void *data) { - /* TODO */ + struct perf_sample_data sample; + struct pt_regs *regs = data; + + sample.addr = bp->attr.bp_addr; + + if (!perf_exclude_event(bp, regs)) + perf_swevent_add(bp, 1, 1, &sample, regs); } #else static void bp_perf_event_destroy(struct perf_event *event) -- cgit v1.1 From 184d3da8ef0ca552dffa0fdd35c046e058a2cf9a Mon Sep 17 00:00:00 2001 From: Stephane Eranian Date: Mon, 23 Nov 2009 21:40:49 -0800 Subject: perf_events: Fix bogus copy_to_user() in perf_event_read_group() When using an event group, the value and id for non leaders events were wrong due to invalid offset into the outgoing buffer. Signed-off-by: Stephane Eranian Acked-by: Peter Zijlstra Cc: paulus@samba.org Cc: perfmon2-devel@lists.sourceforge.net LKML-Reference: <4b0b71e1.0508d00a.075e.ffff84a3@mx.google.com> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 9425c96..accfd7bf 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -1831,7 +1831,7 @@ static int perf_event_read_group(struct perf_event *event, size = n * sizeof(u64); - if (copy_to_user(buf + size, values, size)) { + if (copy_to_user(buf + ret, values, size)) { ret = -EFAULT; goto unlock; } -- cgit v1.1 From fe6126722718e51fba4879517c11ac12d9775bcc Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Tue, 24 Nov 2009 20:38:22 +0100 Subject: perf_events: Fix bad software/trace event recursion counting Commit 4ed7c92d68a5387ba5f7030dc76eab03558e27f5 (perf_events: Undo some recursion damage) has introduced a bad reference counting of the recursion context. putting the context behaves like getting it, dropping every software/trace events after the first one in a context. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Paul Mackerras Cc: Arjan van de Ven Cc: Li Zefan Cc: Steven Rostedt LKML-Reference: <1259091502-5171-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index accfd7bf..35df94e 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -3914,7 +3914,7 @@ void perf_swevent_put_recursion_context(int rctx) { struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context); barrier(); - cpuctx->recursion[rctx]++; + cpuctx->recursion[rctx]--; put_cpu_var(perf_cpu_context); } EXPORT_SYMBOL_GPL(perf_swevent_put_recursion_context); -- cgit v1.1 From 99df5a6a215f026e62287083de2b44b22edd3623 Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Wed, 25 Nov 2009 01:14:59 -0600 Subject: trace/syscalls: Change ret param in struct syscall_trace_exit to long Commit ee949a86b3aef15845ea677aa60231008de62672 ("tracing/syscalls: Use long for syscall ret format and field definitions") changed the syscall exit return type to long, but forgot to change it in the struct. Signed-off-by: Tom Zanussi Cc: Steven Rostedt Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <1259133299-23594-3-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 4da6ede..1d7f483 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -100,7 +100,7 @@ struct syscall_trace_enter { struct syscall_trace_exit { struct trace_entry ent; int nr; - unsigned long ret; + long ret; }; struct kprobe_trace_entry { -- cgit v1.1 From d99be40aff88722ab03ee295e4f6c13a4cca9a3d Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 26 Nov 2009 05:35:40 +0100 Subject: ksym_tracer: Fix breakpoint removal after modification The error path of a breakpoint modification is broken in the ksym tracer. A modified breakpoint hlist node is immediately released after its removal. Also we leak a breakpoint in this case. Fix the path. Signed-off-by: Frederic Weisbecker Cc: Steven Rostedt Cc: Prasad LKML-Reference: <1259210142-5714-1-git-send-regression-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_ksym.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index 11935b5..9f040e4 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -339,14 +339,20 @@ static ssize_t ksym_trace_filter_write(struct file *file, ksym_hbp_handler, true); if (IS_ERR(entry->ksym_hbp)) entry->ksym_hbp = NULL; - if (!entry->ksym_hbp) + + /* modified without problem */ + if (entry->ksym_hbp) { + ret = 0; goto out; + } + } else { + ret = 0; } + /* Error or "symbol:---" case: drop it */ ksym_filter_entry_count--; hlist_del_rcu(&(entry->ksym_hlist)); synchronize_rcu(); kfree(entry); - ret = 0; goto out; } else { /* Check for malformed request: (4) */ -- cgit v1.1 From c6567f642e20bcc79abed030f44be5b0d6da2ded Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 26 Nov 2009 05:35:41 +0100 Subject: hw-breakpoints: Improve in-kernel event creation error granularity In fail case, perf_event_create_kernel_counter() returns NULL instead of an error, which doesn't help us to inform the user about the origin of the problem from the outer most callers. Often we can just return -EINVAL, which doesn't help anyone when it's eventually about a memory allocation failure. Then, this patch makes perf_event_create_kernel_counter() always return a detailed error code. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Paul Mackerras Cc: Prasad LKML-Reference: <1259210142-5714-2-git-send-regression-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 20 +++++++++++--------- 1 file changed, 11 insertions(+), 9 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 35df94e..34a1b9d 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -4780,14 +4780,17 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, */ ctx = find_get_context(pid, cpu); - if (IS_ERR(ctx)) - return NULL; + if (IS_ERR(ctx)) { + err = PTR_ERR(ctx); + goto err_exit; + } event = perf_event_alloc(attr, cpu, ctx, NULL, NULL, callback, GFP_KERNEL); - err = PTR_ERR(event); - if (IS_ERR(event)) + if (IS_ERR(event)) { + err = PTR_ERR(event); goto err_put_context; + } event->filp = NULL; WARN_ON_ONCE(ctx->parent_ctx); @@ -4804,11 +4807,10 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, return event; -err_put_context: - if (err < 0) - put_ctx(ctx); - - return NULL; + err_put_context: + put_ctx(ctx); + err_exit: + return ERR_PTR(err); } EXPORT_SYMBOL_GPL(perf_event_create_kernel_counter); -- cgit v1.1 From 605bfaee9078cd0b01d83402315389839ee4bb5c Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 26 Nov 2009 05:35:42 +0100 Subject: hw-breakpoints: Simplify error handling in breakpoint creation requests This simplifies the error handling when we create a breakpoint. We don't need to check the NULL return value corner case anymore since we have improved perf_event_create_kernel_counter() to always return an error code in the failure case. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Paul Mackerras Cc: Steven Rostedt Cc: Prasad LKML-Reference: <1259210142-5714-3-git-send-regression-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/hw_breakpoint.c | 4 ++-- kernel/trace/trace_ksym.c | 16 ++++------------ 2 files changed, 6 insertions(+), 14 deletions(-) (limited to 'kernel') diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index 06d372f..dd3fb4a 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -442,7 +442,7 @@ register_wide_hw_breakpoint(unsigned long addr, *pevent = bp; - if (IS_ERR(bp) || !bp) { + if (IS_ERR(bp)) { err = PTR_ERR(bp); goto fail; } @@ -453,7 +453,7 @@ register_wide_hw_breakpoint(unsigned long addr, fail: for_each_possible_cpu(cpu) { pevent = per_cpu_ptr(cpu_events, cpu); - if (IS_ERR(*pevent) || !*pevent) + if (IS_ERR(*pevent)) break; unregister_hw_breakpoint(*pevent); } diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index 9f040e4..c538b15 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -200,12 +200,9 @@ int process_new_ksym_entry(char *ksymname, int op, unsigned long addr) entry->ksym_hbp = register_wide_hw_breakpoint(entry->ksym_addr, entry->len, entry->type, ksym_hbp_handler, true); + if (IS_ERR(entry->ksym_hbp)) { - entry->ksym_hbp = NULL; ret = PTR_ERR(entry->ksym_hbp); - } - - if (!entry->ksym_hbp) { printk(KERN_INFO "ksym_tracer request failed. Try again" " later!!\n"); goto err; @@ -332,21 +329,16 @@ static ssize_t ksym_trace_filter_write(struct file *file, if (changed) { unregister_wide_hw_breakpoint(entry->ksym_hbp); entry->type = op; + ret = 0; if (op > 0) { entry->ksym_hbp = register_wide_hw_breakpoint(entry->ksym_addr, entry->len, entry->type, ksym_hbp_handler, true); if (IS_ERR(entry->ksym_hbp)) - entry->ksym_hbp = NULL; - - /* modified without problem */ - if (entry->ksym_hbp) { - ret = 0; + ret = PTR_ERR(entry->ksym_hbp); + else goto out; - } - } else { - ret = 0; } /* Error or "symbol:---" case: drop it */ ksym_filter_entry_count--; -- cgit v1.1 From 11e6635763bdc0e24b39a38876574660755acffc Mon Sep 17 00:00:00 2001 From: Andrew Morton Date: Wed, 25 Nov 2009 23:01:50 -0800 Subject: kernel/hw_breakpoint.c: Fix local/global shadowing If the new percpu tree is combined with the perf events tree the following new warning triggers: kernel/hw_breakpoint.c: In function 'toggle_bp_task_slot': kernel/hw_breakpoint.c:151: warning: 'task_bp_pinned' is used uninitialized in this function Because it's not valid anymore to define a local variable and a percpu variable (even if it's file scope local) with the same name. Rename the local variable to resolve this. Signed-off-by: Andrew Morton Cc: Frederic Weisbecker Cc: K.Prasad Cc: Tejun Heo Cc: Linus Torvalds LKML-Reference: <200911260701.nAQ71owx016356@imap1.linux-foundation.org> [ v2: added changelog ] Signed-off-by: Ingo Molnar --- kernel/hw_breakpoint.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'kernel') diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index dd3fb4a..32e1018 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -121,7 +121,7 @@ static void toggle_bp_task_slot(struct task_struct *tsk, int cpu, bool enable) int count = 0; struct perf_event *bp; struct perf_event_context *ctx = tsk->perf_event_ctxp; - unsigned int *task_bp_pinned; + unsigned int *tsk_pinned; struct list_head *list; unsigned long flags; @@ -146,15 +146,15 @@ static void toggle_bp_task_slot(struct task_struct *tsk, int cpu, bool enable) if (WARN_ONCE(count < 0, "No breakpoint counter found in the counter list")) return; - task_bp_pinned = per_cpu(task_bp_pinned, cpu); + tsk_pinned = per_cpu(task_bp_pinned, cpu); if (enable) { - task_bp_pinned[count]++; + tsk_pinned[count]++; if (count > 0) - task_bp_pinned[count-1]--; + tsk_pinned[count-1]--; } else { - task_bp_pinned[count]--; + tsk_pinned[count]--; if (count > 0) - task_bp_pinned[count-1]++; + tsk_pinned[count-1]++; } } -- cgit v1.1 From 80bbf6b641c8843b9d751a1f299aa7ee073ab9d4 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 25 Nov 2009 21:20:53 +0100 Subject: hw-breakpoints: Fix unused function in off-case bp_perf_event_destroy() is unused in its off-case version, let's remove it to fix the following warning reported by Stephen Rothwell in linux-next: kernel/perf_event.c:4306: warning: 'bp_perf_event_destroy' defined but not used Reported-by: Stephen Rothwell Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra LKML-Reference: <1259180453-5813-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 34a1b9d..f8c7939 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -4303,10 +4303,6 @@ void perf_bp_event(struct perf_event *bp, void *data) perf_swevent_add(bp, 1, 1, &sample, regs); } #else -static void bp_perf_event_destroy(struct perf_event *event) -{ -} - static const struct pmu *bp_perf_event_init(struct perf_event *bp) { return NULL; -- cgit v1.1 From d1eb650ff4130972fa21462fa49cd35a2865403b Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Tue, 24 Nov 2009 16:56:45 -0500 Subject: tracepoint: Move signal sending tracepoint to events/signal.h Move signal sending event to events/signal.h. This patch also renames sched_signal_send event to signal_generate. Changes in v4: - Fix a typo of task_struct pointer. Changes in v3: - Add docbook style comments Changes in v2: - Add siginfo argument - Add siginfo storing macro Signed-off-by: Masami Hiramatsu Reviewed-by: Jason Baron Acked-by: Roland McGrath Cc: systemtap Cc: DLE Cc: Oleg Nesterov LKML-Reference: <20091124215645.30449.60208.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Ingo Molnar --- kernel/signal.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/signal.c b/kernel/signal.c index 6705320..a1e0cc6 100644 --- a/kernel/signal.c +++ b/kernel/signal.c @@ -27,7 +27,8 @@ #include #include #include -#include +#define CREATE_TRACE_POINTS +#include #include #include @@ -834,7 +835,7 @@ static int __send_signal(int sig, struct siginfo *info, struct task_struct *t, struct sigqueue *q; int override_rlimit; - trace_sched_signal_send(sig, t); + trace_signal_generate(sig, info, t); assert_spin_locked(&t->sighand->siglock); -- cgit v1.1 From f9d4257e01d266e67420cc99d456b6d4c8464f54 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Tue, 24 Nov 2009 16:56:51 -0500 Subject: tracepoint: Add signal deliver event Add a tracepoint where a process gets a signal. This tracepoint shows signal-number, sa-handler and sa-flag. Changes in v3: - Add docbook style comments Changes in v2: - Add siginfo argument - Fix comment Signed-off-by: Masami Hiramatsu Reviewed-by: Jason Baron Acked-by: Roland McGrath Cc: systemtap Cc: DLE Cc: Oleg Nesterov LKML-Reference: <20091124215651.30449.20926.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Ingo Molnar --- kernel/signal.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'kernel') diff --git a/kernel/signal.c b/kernel/signal.c index a1e0cc6..349d449 100644 --- a/kernel/signal.c +++ b/kernel/signal.c @@ -1840,6 +1840,9 @@ relock: ka = &sighand->action[signr-1]; } + /* Trace actually delivered signals. */ + trace_signal_deliver(signr, info, ka); + if (ka->sa.sa_handler == SIG_IGN) /* Do nothing. */ continue; if (ka->sa.sa_handler != SIG_DFL) { -- cgit v1.1 From ba005e1f417295d28cd1563ab82bc33af07fb16a Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Tue, 24 Nov 2009 16:56:58 -0500 Subject: tracepoint: Add signal loss events Add signal_overflow_fail and signal_lose_info tracepoints for signal-lost events. Changes in v3: - Add docbook style comments Changes in v2: - Use siginfo string macro Suggested-by: Roland McGrath Reviewed-by: Jason Baron Signed-off-by: Masami Hiramatsu Acked-by: Roland McGrath Cc: systemtap Cc: DLE Cc: Oleg Nesterov LKML-Reference: <20091124215658.30449.9934.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Ingo Molnar --- kernel/signal.c | 19 ++++++++++++++----- 1 file changed, 14 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/signal.c b/kernel/signal.c index 349d449..93e72e5 100644 --- a/kernel/signal.c +++ b/kernel/signal.c @@ -897,12 +897,21 @@ static int __send_signal(int sig, struct siginfo *info, struct task_struct *t, break; } } else if (!is_si_special(info)) { - if (sig >= SIGRTMIN && info->si_code != SI_USER) - /* - * Queue overflow, abort. We may abort if the signal was rt - * and sent by user using something other than kill(). - */ + if (sig >= SIGRTMIN && info->si_code != SI_USER) { + /* + * Queue overflow, abort. We may abort if the + * signal was rt and sent by user using something + * other than kill(). + */ + trace_signal_overflow_fail(sig, group, info); return -EAGAIN; + } else { + /* + * This is a silent loss of information. We still + * send the signal, but the *info bits are lost. + */ + trace_signal_lose_info(sig, group, info); + } } out_set: -- cgit v1.1 From b2e74a265ded1a185f762ebaab967e9e0d008dd8 Mon Sep 17 00:00:00 2001 From: Stephane Eranian Date: Thu, 26 Nov 2009 09:24:30 -0800 Subject: perf_events: Fix read() bogus counts when in error state When a pinned group cannot be scheduled it goes into error state. Normally a group cannot go out of error state without being explicitly re-enabled or disabled. There was a bug in per-thread mode, whereby upon termination of the thread, the group would transition from error to off leading to bogus counts and timing information returned by read(). Fix it by clearing the error state. Signed-off-by: Stephane Eranian Acked-by: Peter Zijlstra Cc: Paul Mackerras Cc: perfmon2-devel@lists.sourceforge.net LKML-Reference: <4b0eb9ce.0508d00a.573b.ffffeab6@mx.google.com> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index f8c7939..0b9ca2d 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -338,7 +338,16 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx) event->group_leader->nr_siblings--; update_event_times(event); - event->state = PERF_EVENT_STATE_OFF; + + /* + * If event was in error state, then keep it + * that way, otherwise bogus counts will be + * returned on read(). The only way to get out + * of error state is by explicit re-enabling + * of the event + */ + if (event->state > PERF_EVENT_STATE_OFF) + event->state = PERF_EVENT_STATE_OFF; /* * If this was a group event with sibling events then -- cgit v1.1 From 5fa10b28e57f94a90535cfeafe89dcee9f47d540 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Fri, 27 Nov 2009 04:55:53 +0100 Subject: hw-breakpoints: Use struct perf_event_attr to define user breakpoints In-kernel user breakpoints are created using functions in which we pass breakpoint parameters as individual variables: address, length and type. Although it fits well for x86, this just does not scale across archictectures that may support this api later as these may have more or different needs. Pass in a perf_event_attr structure instead because it is meant to evolve as much as possible into a generic hardware breakpoint parameter structure. Reported-by: K.Prasad Signed-off-by: Frederic Weisbecker LKML-Reference: <1259294154-5197-1-git-send-regression-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/hw_breakpoint.c | 87 +++++++++++--------------------------------------- 1 file changed, 18 insertions(+), 69 deletions(-) (limited to 'kernel') diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index 32e1018..2a47514 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -289,90 +289,32 @@ int register_perf_hw_breakpoint(struct perf_event *bp) return __register_perf_hw_breakpoint(bp); } -/* - * Register a breakpoint bound to a task and a given cpu. - * If cpu is -1, the breakpoint is active for the task in every cpu - * If the task is -1, the breakpoint is active for every tasks in the given - * cpu. - */ -static struct perf_event * -register_user_hw_breakpoint_cpu(unsigned long addr, - int len, - int type, - perf_callback_t triggered, - pid_t pid, - int cpu, - bool active) -{ - struct perf_event_attr *attr; - struct perf_event *bp; - - attr = kzalloc(sizeof(*attr), GFP_KERNEL); - if (!attr) - return ERR_PTR(-ENOMEM); - - attr->type = PERF_TYPE_BREAKPOINT; - attr->size = sizeof(*attr); - attr->bp_addr = addr; - attr->bp_len = len; - attr->bp_type = type; - /* - * Such breakpoints are used by debuggers to trigger signals when - * we hit the excepted memory op. We can't miss such events, they - * must be pinned. - */ - attr->pinned = 1; - - if (!active) - attr->disabled = 1; - - bp = perf_event_create_kernel_counter(attr, cpu, pid, triggered); - kfree(attr); - - return bp; -} - /** * register_user_hw_breakpoint - register a hardware breakpoint for user space - * @addr: is the memory address that triggers the breakpoint - * @len: the length of the access to the memory (1 byte, 2 bytes etc...) - * @type: the type of the access to the memory (read/write/exec) + * @attr: breakpoint attributes * @triggered: callback to trigger when we hit the breakpoint * @tsk: pointer to 'task_struct' of the process to which the address belongs - * @active: should we activate it while registering it - * */ struct perf_event * -register_user_hw_breakpoint(unsigned long addr, - int len, - int type, +register_user_hw_breakpoint(struct perf_event_attr *attr, perf_callback_t triggered, - struct task_struct *tsk, - bool active) + struct task_struct *tsk) { - return register_user_hw_breakpoint_cpu(addr, len, type, triggered, - tsk->pid, -1, active); + return perf_event_create_kernel_counter(attr, -1, tsk->pid, triggered); } EXPORT_SYMBOL_GPL(register_user_hw_breakpoint); /** * modify_user_hw_breakpoint - modify a user-space hardware breakpoint * @bp: the breakpoint structure to modify - * @addr: is the memory address that triggers the breakpoint - * @len: the length of the access to the memory (1 byte, 2 bytes etc...) - * @type: the type of the access to the memory (read/write/exec) + * @attr: new breakpoint attributes * @triggered: callback to trigger when we hit the breakpoint * @tsk: pointer to 'task_struct' of the process to which the address belongs - * @active: should we activate it while registering it */ struct perf_event * -modify_user_hw_breakpoint(struct perf_event *bp, - unsigned long addr, - int len, - int type, +modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr, perf_callback_t triggered, - struct task_struct *tsk, - bool active) + struct task_struct *tsk) { /* * FIXME: do it without unregistering @@ -381,8 +323,7 @@ modify_user_hw_breakpoint(struct perf_event *bp, */ unregister_hw_breakpoint(bp); - return register_user_hw_breakpoint(addr, len, type, triggered, - tsk, active); + return perf_event_create_kernel_counter(attr, -1, tsk->pid, triggered); } EXPORT_SYMBOL_GPL(modify_user_hw_breakpoint); @@ -406,8 +347,16 @@ register_kernel_hw_breakpoint_cpu(unsigned long addr, int cpu, bool active) { - return register_user_hw_breakpoint_cpu(addr, len, type, triggered, - -1, cpu, active); + DEFINE_BREAKPOINT_ATTR(attr); + + attr.bp_addr = addr; + attr.bp_len = len; + attr.bp_type = type; + + if (!active) + attr.disabled = 1; + + return perf_event_create_kernel_counter(&attr, cpu, -1, triggered); } /** -- cgit v1.1 From dd1853c3f493f6d22d9e5390b192a07b73d2ac0a Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Fri, 27 Nov 2009 04:55:54 +0100 Subject: hw-breakpoints: Use struct perf_event_attr to define kernel breakpoints Kernel breakpoints are created using functions in which we pass breakpoint parameters as individual variables: address, length and type. Although it fits well for x86, this just does not scale across architectures that may support this api later as these may have more or different needs. Pass in a perf_event_attr structure instead because it is meant to evolve as much as possible into a generic hardware breakpoint parameter structure. Reported-by: K.Prasad Signed-off-by: Frederic Weisbecker LKML-Reference: <1259294154-5197-2-git-send-regression-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- kernel/hw_breakpoint.c | 35 ++++------------------------------- kernel/trace/trace_ksym.c | 42 ++++++++++++++++++++---------------------- 2 files changed, 24 insertions(+), 53 deletions(-) (limited to 'kernel') diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index 2a47514..cf5ee16 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -339,42 +339,16 @@ void unregister_hw_breakpoint(struct perf_event *bp) } EXPORT_SYMBOL_GPL(unregister_hw_breakpoint); -static struct perf_event * -register_kernel_hw_breakpoint_cpu(unsigned long addr, - int len, - int type, - perf_callback_t triggered, - int cpu, - bool active) -{ - DEFINE_BREAKPOINT_ATTR(attr); - - attr.bp_addr = addr; - attr.bp_len = len; - attr.bp_type = type; - - if (!active) - attr.disabled = 1; - - return perf_event_create_kernel_counter(&attr, cpu, -1, triggered); -} - /** * register_wide_hw_breakpoint - register a wide breakpoint in the kernel - * @addr: is the memory address that triggers the breakpoint - * @len: the length of the access to the memory (1 byte, 2 bytes etc...) - * @type: the type of the access to the memory (read/write/exec) + * @attr: breakpoint attributes * @triggered: callback to trigger when we hit the breakpoint - * @active: should we activate it while registering it * * @return a set of per_cpu pointers to perf events */ struct perf_event ** -register_wide_hw_breakpoint(unsigned long addr, - int len, - int type, - perf_callback_t triggered, - bool active) +register_wide_hw_breakpoint(struct perf_event_attr *attr, + perf_callback_t triggered) { struct perf_event **cpu_events, **pevent, *bp; long err; @@ -386,8 +360,7 @@ register_wide_hw_breakpoint(unsigned long addr, for_each_possible_cpu(cpu) { pevent = per_cpu_ptr(cpu_events, cpu); - bp = register_kernel_hw_breakpoint_cpu(addr, len, type, - triggered, cpu, active); + bp = perf_event_create_kernel_counter(attr, cpu, -1, triggered); *pevent = bp; diff --git a/kernel/trace/trace_ksym.c b/kernel/trace/trace_ksym.c index c538b15..ddfa0fd 100644 --- a/kernel/trace/trace_ksym.c +++ b/kernel/trace/trace_ksym.c @@ -42,9 +42,7 @@ struct trace_ksym { struct perf_event **ksym_hbp; - unsigned long ksym_addr; - int type; - int len; + struct perf_event_attr attr; #ifdef CONFIG_PROFILE_KSYM_TRACER unsigned long counter; #endif @@ -71,7 +69,7 @@ void ksym_collect_stats(unsigned long hbp_hit_addr) rcu_read_lock(); hlist_for_each_entry_rcu(entry, node, &ksym_filter_head, ksym_hlist) { - if ((entry->ksym_addr == hbp_hit_addr) && + if ((entry->attr.bp_addr == hbp_hit_addr) && (entry->counter <= MAX_UL_INT)) { entry->counter++; break; @@ -192,14 +190,15 @@ int process_new_ksym_entry(char *ksymname, int op, unsigned long addr) if (!entry) return -ENOMEM; - entry->type = op; - entry->ksym_addr = addr; - entry->len = HW_BREAKPOINT_LEN_4; + hw_breakpoint_init(&entry->attr); + + entry->attr.bp_type = op; + entry->attr.bp_addr = addr; + entry->attr.bp_len = HW_BREAKPOINT_LEN_4; ret = -EAGAIN; - entry->ksym_hbp = register_wide_hw_breakpoint(entry->ksym_addr, - entry->len, entry->type, - ksym_hbp_handler, true); + entry->ksym_hbp = register_wide_hw_breakpoint(&entry->attr, + ksym_hbp_handler); if (IS_ERR(entry->ksym_hbp)) { ret = PTR_ERR(entry->ksym_hbp); @@ -236,12 +235,12 @@ static ssize_t ksym_trace_filter_read(struct file *filp, char __user *ubuf, mutex_lock(&ksym_tracer_mutex); hlist_for_each_entry(entry, node, &ksym_filter_head, ksym_hlist) { - ret = trace_seq_printf(s, "%pS:", (void *)entry->ksym_addr); - if (entry->type == HW_BREAKPOINT_R) + ret = trace_seq_printf(s, "%pS:", (void *)entry->attr.bp_addr); + if (entry->attr.bp_type == HW_BREAKPOINT_R) ret = trace_seq_puts(s, "r--\n"); - else if (entry->type == HW_BREAKPOINT_W) + else if (entry->attr.bp_type == HW_BREAKPOINT_W) ret = trace_seq_puts(s, "-w-\n"); - else if (entry->type == (HW_BREAKPOINT_W | HW_BREAKPOINT_R)) + else if (entry->attr.bp_type == (HW_BREAKPOINT_W | HW_BREAKPOINT_R)) ret = trace_seq_puts(s, "rw-\n"); WARN_ON_ONCE(!ret); } @@ -317,9 +316,9 @@ static ssize_t ksym_trace_filter_write(struct file *file, ret = -EINVAL; hlist_for_each_entry(entry, node, &ksym_filter_head, ksym_hlist) { - if (entry->ksym_addr == ksym_addr) { + if (entry->attr.bp_addr == ksym_addr) { /* Check for malformed request: (6) */ - if (entry->type != op) + if (entry->attr.bp_type != op) changed = 1; else goto out; @@ -328,13 +327,12 @@ static ssize_t ksym_trace_filter_write(struct file *file, } if (changed) { unregister_wide_hw_breakpoint(entry->ksym_hbp); - entry->type = op; + entry->attr.bp_type = op; ret = 0; if (op > 0) { entry->ksym_hbp = - register_wide_hw_breakpoint(entry->ksym_addr, - entry->len, entry->type, - ksym_hbp_handler, true); + register_wide_hw_breakpoint(&entry->attr, + ksym_hbp_handler); if (IS_ERR(entry->ksym_hbp)) ret = PTR_ERR(entry->ksym_hbp); else @@ -489,7 +487,7 @@ static int ksym_tracer_stat_show(struct seq_file *m, void *v) entry = hlist_entry(stat, struct trace_ksym, ksym_hlist); - access_type = entry->type; + access_type = entry->attr.bp_type; switch (access_type) { case HW_BREAKPOINT_R: @@ -505,7 +503,7 @@ static int ksym_tracer_stat_show(struct seq_file *m, void *v) seq_puts(m, " NA "); } - if (lookup_symbol_name(entry->ksym_addr, fn_name) >= 0) + if (lookup_symbol_name(entry->attr.bp_addr, fn_name) >= 0) seq_printf(m, " %-36s", fn_name); else seq_printf(m, " %-36s", ""); -- cgit v1.1 From 0f1ef51d244809f417bdf45cdb00109fb6005672 Mon Sep 17 00:00:00 2001 From: Lai Jiangshan Date: Thu, 26 Nov 2009 15:49:33 +0800 Subject: trace_syscalls: Add syscall nr field Field syscall number is missed in syscall_enter_define_fields()/ syscall_exit_define_fields(). Syscall number is also needed for event filter or other users. Signed-off-by: Lai Jiangshan Acked-by: Frederic Weisbecker Cc: Jason Baron Cc: Steven Rostedt LKML-Reference: <4B0E330D.1070206@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_syscalls.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'kernel') diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index 9189cbe..63aa807 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -261,6 +261,10 @@ int syscall_enter_define_fields(struct ftrace_event_call *call) if (ret) return ret; + ret = trace_define_field(call, SYSCALL_FIELD(int, nr), FILTER_OTHER); + if (ret) + return ret; + for (i = 0; i < meta->nb_args; i++) { ret = trace_define_field(call, meta->types[i], meta->args[i], offset, @@ -281,6 +285,10 @@ int syscall_exit_define_fields(struct ftrace_event_call *call) if (ret) return ret; + ret = trace_define_field(call, SYSCALL_FIELD(int, nr), FILTER_OTHER); + if (ret) + return ret; + ret = trace_define_field(call, SYSCALL_FIELD(long, ret), FILTER_OTHER); -- cgit v1.1 From abab9d37d2a826fcf588c5f30152dbe05c40111c Mon Sep 17 00:00:00 2001 From: Lai Jiangshan Date: Wed, 25 Nov 2009 16:32:21 +0800 Subject: trace_kprobes: Fix memory leak tp->nr_args is not set before we "goto error", it causes memory leak for free_trace_probe() use tp->nr_args to free memory of args. Signed-off-by: Lai Jiangshan Acked-by: Masami Hiramatsu Cc: Steven Rostedt Cc: Frederic Weisbecker LKML-Reference: <4B0CEB95.2060107@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_kprobe.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 79ce6a2..82e8583 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -704,10 +704,12 @@ static int create_trace_probe(int argc, char **argv) ret = parse_probe_arg(arg, &tp->args[i].fetch, is_return); if (ret) { pr_info("Parse error at argument%d. (%d)\n", i, ret); + kfree(tp->args[i].name); goto error; } + + tp->nr_args++; } - tp->nr_args = i; ret = register_trace_probe(tp); if (ret) -- cgit v1.1 From 3d9b2e1ddf42dd3df38af7794fa5e39cce760f3b Mon Sep 17 00:00:00 2001 From: Lai Jiangshan Date: Wed, 25 Nov 2009 16:32:47 +0800 Subject: trace_kprobes: Always show group name Sometimes the group name is not "kprobes", It'll be better if we can read it from tracing/kprobe_events. # echo 'r:laijs/vfs_read vfs_read %ax' > kprobe_events # cat kprobe_events r:laijs/vfs_read vfs_read %ax=%ax Signed-off-by: Lai Jiangshan Acked-by: Masami Hiramatsu Cc: Steven Rostedt Cc: Frederic Weisbecker LKML-Reference: <4B0CEBAF.6000104@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_kprobe.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 82e8583..96e1944 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -760,7 +760,7 @@ static int probes_seq_show(struct seq_file *m, void *v) char buf[MAX_ARGSTR_LEN + 1]; seq_printf(m, "%c", probe_is_return(tp) ? 'r' : 'p'); - seq_printf(m, ":%s", tp->call.name); + seq_printf(m, ":%s/%s", tp->call.system, tp->call.name); if (tp->symbol) seq_printf(m, " %s+%u", probe_symbol(tp), tp->rp.kp.offset); -- cgit v1.1 From 52a11f354970e7301e1d1a029b87535be45abec9 Mon Sep 17 00:00:00 2001 From: Lai Jiangshan Date: Wed, 25 Nov 2009 16:33:15 +0800 Subject: trace_kprobes: Don't output zero offset "symbol_name+0" is not so friendly. It makes the output longer. Signed-off-by: Lai Jiangshan Acked-by: Masami Hiramatsu Cc: Steven Rostedt Cc: Frederic Weisbecker LKML-Reference: <4B0CEBCB.7080309@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_kprobe.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 96e1944..72d0c65 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -243,7 +243,11 @@ static int probe_arg_string(char *buf, size_t n, struct fetch_func *ff) ret = snprintf(buf, n, "@0x%p", ff->data); else if (ff->func == fetch_symbol) { struct symbol_cache *sc = ff->data; - ret = snprintf(buf, n, "@%s%+ld", sc->symbol, sc->offset); + if (sc->offset) + ret = snprintf(buf, n, "@%s%+ld", sc->symbol, + sc->offset); + else + ret = snprintf(buf, n, "@%s", sc->symbol); } else if (ff->func == fetch_retvalue) ret = snprintf(buf, n, "$retval"); else if (ff->func == fetch_stack_address) @@ -762,10 +766,12 @@ static int probes_seq_show(struct seq_file *m, void *v) seq_printf(m, "%c", probe_is_return(tp) ? 'r' : 'p'); seq_printf(m, ":%s/%s", tp->call.system, tp->call.name); - if (tp->symbol) + if (!tp->symbol) + seq_printf(m, " 0x%p", tp->rp.kp.addr); + else if (tp->rp.kp.offset) seq_printf(m, " %s+%u", probe_symbol(tp), tp->rp.kp.offset); else - seq_printf(m, " 0x%p", tp->rp.kp.addr); + seq_printf(m, " %s", probe_symbol(tp)); for (i = 0; i < tp->nr_args; i++) { ret = probe_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i].fetch); -- cgit v1.1 From ba8665d7dd95eb6093ee06f8f624b6acb1e73206 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 30 Nov 2009 19:19:20 -0500 Subject: trace_kprobes: Fix a memory leak bug and check kstrdup() return value Fix a memory leak case in create_trace_probe(). When an argument is too long (> MAX_ARGSTR_LEN), it just jumps to error path. In that case tp->args[i].name is not released. This also fixes a bug to check kstrdup()'s return value. Signed-off-by: Masami Hiramatsu Cc: systemtap Cc: DLE Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <20091201001919.10235.56455.stgit@harusame> Signed-off-by: Ingo Molnar --- kernel/trace/trace_kprobe.c | 26 +++++++++++++++++++------- 1 file changed, 19 insertions(+), 7 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 72d0c65..aff5f80 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -483,7 +483,8 @@ static int parse_probe_vars(char *arg, struct fetch_func *ff, int is_return) return ret; } -static int parse_probe_arg(char *arg, struct fetch_func *ff, int is_return) +/* Recursive argument parser */ +static int __parse_probe_arg(char *arg, struct fetch_func *ff, int is_return) { int ret = 0; unsigned long param; @@ -543,7 +544,7 @@ static int parse_probe_arg(char *arg, struct fetch_func *ff, int is_return) if (!id) return -ENOMEM; id->offset = offset; - ret = parse_probe_arg(arg, &id->orig, is_return); + ret = __parse_probe_arg(arg, &id->orig, is_return); if (ret) kfree(id); else { @@ -560,6 +561,16 @@ static int parse_probe_arg(char *arg, struct fetch_func *ff, int is_return) return ret; } +/* String length checking wrapper */ +static int parse_probe_arg(char *arg, struct fetch_func *ff, int is_return) +{ + if (strlen(arg) > MAX_ARGSTR_LEN) { + pr_info("Argument is too long.: %s\n", arg); + return -ENOSPC; + } + return __parse_probe_arg(arg, ff, is_return); +} + /* Return 1 if name is reserved or already used by another argument */ static int conflict_field_name(const char *name, struct probe_arg *args, int narg) @@ -698,13 +709,14 @@ static int create_trace_probe(int argc, char **argv) } tp->args[i].name = kstrdup(argv[i], GFP_KERNEL); - - /* Parse fetch argument */ - if (strlen(arg) > MAX_ARGSTR_LEN) { - pr_info("Argument%d(%s) is too long.\n", i, arg); - ret = -ENOSPC; + if (!tp->args[i].name) { + pr_info("Failed to allocate argument%d name '%s'.\n", + i, argv[i]); + ret = -ENOMEM; goto error; } + + /* Parse fetch argument */ ret = parse_probe_arg(arg, &tp->args[i].fetch, is_return); if (ret) { pr_info("Parse error at argument%d. (%d)\n", i, ret); -- cgit v1.1 From 59d069eb5ae9b033ed1c124c92e1532c4a958991 Mon Sep 17 00:00:00 2001 From: Xiao Guangrong Date: Tue, 1 Dec 2009 17:30:08 +0800 Subject: perf_event: Initialize data.period in perf_swevent_hrtimer() In current code in perf_swevent_hrtimer(), data.period is not initialized, The result is obvious wrong: # ./perf record -f -e cpu-clock make # ./perf report # Samples: 1740 # # Overhead Command ...... # ........ ........ .......................................... # 1025422183050275328.00% sh libc-2.9.90.so ... 1025422183050275328.00% perl libperl.so ... 1025422168240043264.00% perl [kernel] ... 1025422030011210752.00% perl [kernel] ... Signed-off-by: Xiao Guangrong Acked-by: Peter Zijlstra Cc: Frederic Weisbecker Cc: LKML-Reference: <4B14E220.2050107@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 1 + 1 file changed, 1 insertion(+) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 0b9ca2d..040ee51 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -4010,6 +4010,7 @@ static enum hrtimer_restart perf_swevent_hrtimer(struct hrtimer *hrtimer) event->pmu->read(event); data.addr = 0; + data.period = event->hw.last_period; regs = get_irq_regs(); /* * In case we exclude kernel IPs or are somehow not in interrupt -- cgit v1.1 From bf56a4ea9f1683c5b223fd3a5dbea23f1fa91c34 Mon Sep 17 00:00:00 2001 From: Lai Jiangshan Date: Tue, 1 Dec 2009 16:23:20 +0800 Subject: trace_syscalls: Remove unused event_syscall_enter and event_syscall_exit fix event_enter_##sname->event fix event_exit_##sname->event remove unused event_syscall_enter and event_syscall_exit Signed-off-by: Lai Jiangshan Acked-by: Jason Baron Cc: Steven Rostedt Cc: Frederic Weisbecker LKML-Reference: <4B14D278.4090209@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_syscalls.c | 8 -------- 1 file changed, 8 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index 63aa807..00d6e17 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -444,14 +444,6 @@ void unreg_event_syscall_exit(struct ftrace_event_call *call) mutex_unlock(&syscall_trace_lock); } -struct trace_event event_syscall_enter = { - .trace = print_syscall_enter, -}; - -struct trace_event event_syscall_exit = { - .trace = print_syscall_exit, -}; - int __init init_ftrace_syscalls(void) { struct syscall_metadata *meta; -- cgit v1.1 From 31c16b13349970b2684248c7d8608d2a96ae135d Mon Sep 17 00:00:00 2001 From: Lai Jiangshan Date: Tue, 1 Dec 2009 16:23:30 +0800 Subject: trace_syscalls: Set event_enter_##sname->data to its metadata Set event_enter_##sname->data to its metadata, it makes codes simpler. Signed-off-by: Lai Jiangshan Acked-by: Jason Baron Cc: Steven Rostedt Cc: Frederic Weisbecker LKML-Reference: <4B14D282.7050709@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_syscalls.c | 36 +++++++++++------------------------- 1 file changed, 11 insertions(+), 25 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index 00d6e17..39649b1 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -51,7 +51,7 @@ static struct syscall_metadata *syscall_nr_to_meta(int nr) return syscalls_metadata[nr]; } -int syscall_name_to_nr(char *name) +int syscall_name_to_nr(const char *name) { int i; @@ -172,18 +172,11 @@ extern char *__bad_type_size(void); int syscall_enter_format(struct ftrace_event_call *call, struct trace_seq *s) { int i; - int nr; int ret; - struct syscall_metadata *entry; + struct syscall_metadata *entry = call->data; struct syscall_trace_enter trace; int offset = offsetof(struct syscall_trace_enter, args); - nr = syscall_name_to_nr(call->data); - entry = syscall_nr_to_meta(nr); - - if (!entry) - return 0; - ret = trace_seq_printf(s, "\tfield:%s %s;\toffset:%zu;\tsize:%zu;" "\tsigned:%u;\n", SYSCALL_FIELD(int, nr)); @@ -245,18 +238,11 @@ int syscall_exit_format(struct ftrace_event_call *call, struct trace_seq *s) int syscall_enter_define_fields(struct ftrace_event_call *call) { struct syscall_trace_enter trace; - struct syscall_metadata *meta; + struct syscall_metadata *meta = call->data; int ret; - int nr; int i; int offset = offsetof(typeof(trace), args); - nr = syscall_name_to_nr(call->data); - meta = syscall_nr_to_meta(nr); - - if (!meta) - return 0; - ret = trace_define_common_fields(call); if (ret) return ret; @@ -366,9 +352,9 @@ int reg_event_syscall_enter(struct ftrace_event_call *call) { int ret = 0; int num; - char *name; + const char *name; - name = (char *)call->data; + name = ((struct syscall_metadata *)call->data)->name; num = syscall_name_to_nr(name); if (num < 0 || num >= NR_syscalls) return -ENOSYS; @@ -389,9 +375,9 @@ int reg_event_syscall_enter(struct ftrace_event_call *call) void unreg_event_syscall_enter(struct ftrace_event_call *call) { int num; - char *name; + const char *name; - name = (char *)call->data; + name = ((struct syscall_metadata *)call->data)->name; num = syscall_name_to_nr(name); if (num < 0 || num >= NR_syscalls) return; @@ -407,9 +393,9 @@ int reg_event_syscall_exit(struct ftrace_event_call *call) { int ret = 0; int num; - char *name; + const char *name; - name = call->data; + name = ((struct syscall_metadata *)call->data)->name; num = syscall_name_to_nr(name); if (num < 0 || num >= NR_syscalls) return -ENOSYS; @@ -430,9 +416,9 @@ int reg_event_syscall_exit(struct ftrace_event_call *call) void unreg_event_syscall_exit(struct ftrace_event_call *call) { int num; - char *name; + const char *name; - name = call->data; + name = ((struct syscall_metadata *)call->data)->name; num = syscall_name_to_nr(name); if (num < 0 || num >= NR_syscalls) return; -- cgit v1.1 From fcc19438dda38dacc8c144e2db3ebc6b9fd4f8b8 Mon Sep 17 00:00:00 2001 From: Lai Jiangshan Date: Tue, 1 Dec 2009 16:23:36 +0800 Subject: trace_syscalls: Remove enter_id exit_id use ->enter_event->id instead of ->enter_id use ->exit_event->id instead of ->exit_id Signed-off-by: Lai Jiangshan Acked-by: Jason Baron Cc: Steven Rostedt Cc: Frederic Weisbecker LKML-Reference: <4B14D288.7030001@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_syscalls.c | 30 ++++++++++-------------------- 1 file changed, 10 insertions(+), 20 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index 39649b1..27eb18d 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -67,16 +67,6 @@ int syscall_name_to_nr(const char *name) return -1; } -void set_syscall_enter_id(int num, int id) -{ - syscalls_metadata[num]->enter_id = id; -} - -void set_syscall_exit_id(int num, int id) -{ - syscalls_metadata[num]->exit_id = id; -} - enum print_line_t print_syscall_enter(struct trace_iterator *iter, int flags) { @@ -93,7 +83,7 @@ print_syscall_enter(struct trace_iterator *iter, int flags) if (!entry) goto end; - if (entry->enter_id != ent->type) { + if (entry->enter_event->id != ent->type) { WARN_ON_ONCE(1); goto end; } @@ -148,7 +138,7 @@ print_syscall_exit(struct trace_iterator *iter, int flags) return TRACE_TYPE_HANDLED; } - if (entry->exit_id != ent->type) { + if (entry->exit_event->id != ent->type) { WARN_ON_ONCE(1); return TRACE_TYPE_UNHANDLED; } @@ -302,8 +292,8 @@ void ftrace_syscall_enter(struct pt_regs *regs, long id) size = sizeof(*entry) + sizeof(unsigned long) * sys_data->nb_args; - event = trace_current_buffer_lock_reserve(&buffer, sys_data->enter_id, - size, 0, 0); + event = trace_current_buffer_lock_reserve(&buffer, + sys_data->enter_event->id, size, 0, 0); if (!event) return; @@ -334,8 +324,8 @@ void ftrace_syscall_exit(struct pt_regs *regs, long ret) if (!sys_data) return; - event = trace_current_buffer_lock_reserve(&buffer, sys_data->exit_id, - sizeof(*entry), 0, 0); + event = trace_current_buffer_lock_reserve(&buffer, + sys_data->exit_event->id, sizeof(*entry), 0, 0); if (!event) return; @@ -510,11 +500,11 @@ static void prof_syscall_enter(struct pt_regs *regs, long id) rec = (struct syscall_trace_enter *) raw_data; tracing_generic_entry_update(&rec->ent, 0, 0); - rec->ent.type = sys_data->enter_id; + rec->ent.type = sys_data->enter_event->id; rec->nr = syscall_nr; syscall_get_arguments(current, regs, 0, sys_data->nb_args, (unsigned long *)&rec->args); - perf_tp_event(sys_data->enter_id, 0, 1, rec, size); + perf_tp_event(sys_data->enter_event->id, 0, 1, rec, size); end: perf_swevent_put_recursion_context(rctx); @@ -615,11 +605,11 @@ static void prof_syscall_exit(struct pt_regs *regs, long ret) rec = (struct syscall_trace_exit *)raw_data; tracing_generic_entry_update(&rec->ent, 0, 0); - rec->ent.type = sys_data->exit_id; + rec->ent.type = sys_data->exit_event->id; rec->nr = syscall_nr; rec->ret = syscall_get_return_value(current, regs); - perf_tp_event(sys_data->exit_id, 0, 1, rec, size); + perf_tp_event(sys_data->exit_event->id, 0, 1, rec, size); end: perf_swevent_put_recursion_context(rctx); -- cgit v1.1 From c252f65793874b56d50395ab604db465ce688665 Mon Sep 17 00:00:00 2001 From: Lai Jiangshan Date: Tue, 1 Dec 2009 16:23:47 +0800 Subject: trace_syscalls: Add syscall_nr field to struct syscall_metadata Add syscall_nr field to struct syscall_metadata, it helps us to get syscall number easier. Signed-off-by: Lai Jiangshan Acked-by: Jason Baron Cc: Steven Rostedt Cc: Frederic Weisbecker LKML-Reference: <4B14D293.6090800@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_syscalls.c | 22 +++++++++------------- 1 file changed, 9 insertions(+), 13 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index 27eb18d..144cc14 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -51,7 +51,7 @@ static struct syscall_metadata *syscall_nr_to_meta(int nr) return syscalls_metadata[nr]; } -int syscall_name_to_nr(const char *name) +static int syscall_name_to_nr(const char *name) { int i; @@ -342,10 +342,8 @@ int reg_event_syscall_enter(struct ftrace_event_call *call) { int ret = 0; int num; - const char *name; - name = ((struct syscall_metadata *)call->data)->name; - num = syscall_name_to_nr(name); + num = ((struct syscall_metadata *)call->data)->syscall_nr; if (num < 0 || num >= NR_syscalls) return -ENOSYS; mutex_lock(&syscall_trace_lock); @@ -365,10 +363,8 @@ int reg_event_syscall_enter(struct ftrace_event_call *call) void unreg_event_syscall_enter(struct ftrace_event_call *call) { int num; - const char *name; - name = ((struct syscall_metadata *)call->data)->name; - num = syscall_name_to_nr(name); + num = ((struct syscall_metadata *)call->data)->syscall_nr; if (num < 0 || num >= NR_syscalls) return; mutex_lock(&syscall_trace_lock); @@ -383,10 +379,8 @@ int reg_event_syscall_exit(struct ftrace_event_call *call) { int ret = 0; int num; - const char *name; - name = ((struct syscall_metadata *)call->data)->name; - num = syscall_name_to_nr(name); + num = ((struct syscall_metadata *)call->data)->syscall_nr; if (num < 0 || num >= NR_syscalls) return -ENOSYS; mutex_lock(&syscall_trace_lock); @@ -406,10 +400,8 @@ int reg_event_syscall_exit(struct ftrace_event_call *call) void unreg_event_syscall_exit(struct ftrace_event_call *call) { int num; - const char *name; - name = ((struct syscall_metadata *)call->data)->name; - num = syscall_name_to_nr(name); + num = ((struct syscall_metadata *)call->data)->syscall_nr; if (num < 0 || num >= NR_syscalls) return; mutex_lock(&syscall_trace_lock); @@ -436,6 +428,10 @@ int __init init_ftrace_syscalls(void) for (i = 0; i < NR_syscalls; i++) { addr = arch_syscall_addr(i); meta = find_syscall_meta(addr); + if (!meta) + continue; + + meta->syscall_nr = i; syscalls_metadata[i] = meta; } -- cgit v1.1 From a1301da0997bf73c44dbe584e9070a13adc89672 Mon Sep 17 00:00:00 2001 From: Lai Jiangshan Date: Tue, 1 Dec 2009 16:23:55 +0800 Subject: trace_syscalls: Remove duplicate init_enter_##sname() use only one init_syscall_trace instead of many init_enter_##sname()/init_exit_##sname() Signed-off-by: Lai Jiangshan Acked-by: Jason Baron Cc: Steven Rostedt Cc: Frederic Weisbecker LKML-Reference: <4B14D29B.6090708@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_syscalls.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'kernel') diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index 144cc14..c651409 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -412,6 +412,18 @@ void unreg_event_syscall_exit(struct ftrace_event_call *call) mutex_unlock(&syscall_trace_lock); } +int init_syscall_trace(struct ftrace_event_call *call) +{ + int id; + + id = register_ftrace_event(call->event); + if (!id) + return -ENODEV; + call->id = id; + INIT_LIST_HEAD(&call->fields); + return 0; +} + int __init init_ftrace_syscalls(void) { struct syscall_metadata *meta; -- cgit v1.1 From 3bbe84e9d385205d638035ee9dcc4db1b486ea08 Mon Sep 17 00:00:00 2001 From: Lai Jiangshan Date: Tue, 1 Dec 2009 16:24:01 +0800 Subject: trace_syscalls: Simplify syscall profile use only one prof_sysenter_enable() instead of prof_sysenter_enable_##sname() use only one prof_sysenter_disable() instead of prof_sysenter_disable_##sname() use only one prof_sysexit_enable() instead of prof_sysexit_enable_##sname() use only one prof_sysexit_disable() instead of prof_sysexit_disable_##sname() Signed-off-by: Lai Jiangshan Acked-by: Jason Baron Cc: Steven Rostedt Cc: Frederic Weisbecker LKML-Reference: <4B14D2A1.8060304@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_syscalls.c | 24 ++++++++---------------- 1 file changed, 8 insertions(+), 16 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index c651409..1e85b6c 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -520,14 +520,12 @@ end_recursion: local_irq_restore(flags); } -int reg_prof_syscall_enter(char *name) +int prof_sysenter_enable(struct ftrace_event_call *call) { int ret = 0; int num; - num = syscall_name_to_nr(name); - if (num < 0 || num >= NR_syscalls) - return -ENOSYS; + num = ((struct syscall_metadata *)call->data)->syscall_nr; mutex_lock(&syscall_trace_lock); if (!sys_prof_refcount_enter) @@ -543,13 +541,11 @@ int reg_prof_syscall_enter(char *name) return ret; } -void unreg_prof_syscall_enter(char *name) +void prof_sysenter_disable(struct ftrace_event_call *call) { int num; - num = syscall_name_to_nr(name); - if (num < 0 || num >= NR_syscalls) - return; + num = ((struct syscall_metadata *)call->data)->syscall_nr; mutex_lock(&syscall_trace_lock); sys_prof_refcount_enter--; @@ -625,14 +621,12 @@ end_recursion: local_irq_restore(flags); } -int reg_prof_syscall_exit(char *name) +int prof_sysexit_enable(struct ftrace_event_call *call) { int ret = 0; int num; - num = syscall_name_to_nr(name); - if (num < 0 || num >= NR_syscalls) - return -ENOSYS; + num = ((struct syscall_metadata *)call->data)->syscall_nr; mutex_lock(&syscall_trace_lock); if (!sys_prof_refcount_exit) @@ -648,13 +642,11 @@ int reg_prof_syscall_exit(char *name) return ret; } -void unreg_prof_syscall_exit(char *name) +void prof_sysexit_disable(struct ftrace_event_call *call) { int num; - num = syscall_name_to_nr(name); - if (num < 0 || num >= NR_syscalls) - return; + num = ((struct syscall_metadata *)call->data)->syscall_nr; mutex_lock(&syscall_trace_lock); sys_prof_refcount_exit--; -- cgit v1.1 From 7be077f56370cd52c48c08272b0867132f87bc48 Mon Sep 17 00:00:00 2001 From: Lai Jiangshan Date: Tue, 1 Dec 2009 16:24:06 +0800 Subject: trace_syscalls: Remove unused syscall_name_to_nr() After duplications are removed, syscall_name_to_nr() is unused. Signed-off-by: Lai Jiangshan Acked-by: Jason Baron Cc: Steven Rostedt Cc: Frederic Weisbecker LKML-Reference: <4B14D2A6.6060803@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/trace/trace_syscalls.c | 16 ---------------- 1 file changed, 16 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index 1e85b6c..57501d9 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c @@ -51,22 +51,6 @@ static struct syscall_metadata *syscall_nr_to_meta(int nr) return syscalls_metadata[nr]; } -static int syscall_name_to_nr(const char *name) -{ - int i; - - if (!syscalls_metadata) - return -1; - - for (i = 0; i < NR_syscalls; i++) { - if (syscalls_metadata[i]) { - if (!strcmp(syscalls_metadata[i]->name, name)) - return i; - } - } - return -1; -} - enum print_line_t print_syscall_enter(struct trace_iterator *iter, int flags) { -- cgit v1.1 From ec70ccd806111ba3caf596def91a8580138b12db Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Kristian=20H=C3=B8gsberg?= Date: Tue, 1 Dec 2009 15:05:01 -0500 Subject: perf: Don't free perf_mmap_data until work has been done MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In the CONFIG_PERF_USE_VMALLOC case, perf_mmap_data_free() only schedules the cleanup of the perf_mmap_data struct. In that case we have to wait until the work has been done before we free data. Signed-off-by: Kristian Høgsberg Cc: David S. Miller Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker Cc: LKML-Reference: <1259697901-1747-1-git-send-email-krh@bitplanet.net> Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 040ee51..6b7ddba 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -2210,6 +2210,7 @@ static void perf_mmap_data_free(struct perf_mmap_data *data) perf_mmap_free_page((unsigned long)data->user_page); for (i = 0; i < data->nr_pages; i++) perf_mmap_free_page((unsigned long)data->data_pages[i]); + kfree(data); } #else @@ -2250,6 +2251,7 @@ static void perf_mmap_data_free_work(struct work_struct *work) perf_mmap_unmark_page(base + (i * PAGE_SIZE)); vfree(base); + kfree(data); } static void perf_mmap_data_free(struct perf_mmap_data *data) @@ -2355,7 +2357,6 @@ static void perf_mmap_data_free_rcu(struct rcu_head *rcu_head) data = container_of(rcu_head, struct perf_mmap_data, rcu_head); perf_mmap_data_free(data); - kfree(data); } static void perf_mmap_data_release(struct perf_event *event) -- cgit v1.1