diff options
-rw-r--r-- | arch/i386/kernel/cpu/amd.c | 2 | ||||
-rw-r--r-- | arch/x86_64/kernel/process.c | 4 | ||||
-rw-r--r-- | arch/x86_64/kernel/setup.c | 4 | ||||
-rw-r--r-- | include/asm-i386/cpufeature.h | 1 | ||||
-rw-r--r-- | include/asm-i386/i387.h | 30 | ||||
-rw-r--r-- | include/asm-x86_64/cpufeature.h | 1 | ||||
-rw-r--r-- | include/asm-x86_64/i387.h | 20 |
7 files changed, 56 insertions, 6 deletions
diff --git a/arch/i386/kernel/cpu/amd.c b/arch/i386/kernel/cpu/amd.c index ff2b215..786d1a5 100644 --- a/arch/i386/kernel/cpu/amd.c +++ b/arch/i386/kernel/cpu/amd.c @@ -207,6 +207,8 @@ static void __init init_amd(struct cpuinfo_x86 *c) set_bit(X86_FEATURE_K7, c->x86_capability); break; } + if (c->x86 >= 6) + set_bit(X86_FEATURE_FXSAVE_LEAK, c->x86_capability); display_cacheinfo(c); diff --git a/arch/x86_64/kernel/process.c b/arch/x86_64/kernel/process.c index 1c44b53..fb903e6 100644 --- a/arch/x86_64/kernel/process.c +++ b/arch/x86_64/kernel/process.c @@ -575,8 +575,10 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) prev->userrsp = read_pda(oldrsp); write_pda(oldrsp, next->userrsp); write_pda(pcurrent, next_p); + /* This must be here to ensure both math_state_restore() and - kernel_fpu_begin() work consistently. */ + kernel_fpu_begin() work consistently. + And the AMD workaround requires it to be after DS reload. */ unlazy_fpu(prev_p); write_pda(kernelstack, task_stack_page(next_p) + THREAD_SIZE - PDA_STACKOFFSET); diff --git a/arch/x86_64/kernel/setup.c b/arch/x86_64/kernel/setup.c index c50b067..759070c 100644 --- a/arch/x86_64/kernel/setup.c +++ b/arch/x86_64/kernel/setup.c @@ -930,6 +930,10 @@ static int __init init_amd(struct cpuinfo_x86 *c) if (c->x86 == 15 && ((level >= 0x0f48 && level < 0x0f50) || level >= 0x0f58)) set_bit(X86_FEATURE_REP_GOOD, &c->x86_capability); + /* Enable workaround for FXSAVE leak */ + if (c->x86 >= 6) + set_bit(X86_FEATURE_FXSAVE_LEAK, &c->x86_capability); + r = get_model_name(c); if (!r) { switch (c->x86) { diff --git a/include/asm-i386/cpufeature.h b/include/asm-i386/cpufeature.h index 5c0b587..b44bfc6 100644 --- a/include/asm-i386/cpufeature.h +++ b/include/asm-i386/cpufeature.h @@ -71,6 +71,7 @@ #define X86_FEATURE_P4 (3*32+ 7) /* P4 */ #define X86_FEATURE_CONSTANT_TSC (3*32+ 8) /* TSC ticks at a constant rate */ #define X86_FEATURE_UP (3*32+ 9) /* smp kernel running on up */ +#define X86_FEATURE_FXSAVE_LEAK (3*32+10) /* FXSAVE leaks FOP/FIP/FOP */ /* Intel-defined CPU features, CPUID level 0x00000001 (ecx), word 4 */ #define X86_FEATURE_XMM3 (4*32+ 0) /* Streaming SIMD Extensions-3 */ diff --git a/include/asm-i386/i387.h b/include/asm-i386/i387.h index 152d0ba..7b1f011 100644 --- a/include/asm-i386/i387.h +++ b/include/asm-i386/i387.h @@ -13,6 +13,7 @@ #include <linux/sched.h> #include <linux/init.h> +#include <linux/kernel_stat.h> #include <asm/processor.h> #include <asm/sigcontext.h> #include <asm/user.h> @@ -38,17 +39,38 @@ extern void init_fpu(struct task_struct *); extern void kernel_fpu_begin(void); #define kernel_fpu_end() do { stts(); preempt_enable(); } while(0) +/* We need a safe address that is cheap to find and that is already + in L1 during context switch. The best choices are unfortunately + different for UP and SMP */ +#ifdef CONFIG_SMP +#define safe_address (__per_cpu_offset[0]) +#else +#define safe_address (kstat_cpu(0).cpustat.user) +#endif + /* * These must be called with preempt disabled */ static inline void __save_init_fpu( struct task_struct *tsk ) { + /* Use more nops than strictly needed in case the compiler + varies code */ alternative_input( - "fnsave %1 ; fwait ;" GENERIC_NOP2, - "fxsave %1 ; fnclex", + "fnsave %[fx] ;fwait;" GENERIC_NOP8 GENERIC_NOP4, + "fxsave %[fx]\n" + "bt $7,%[fsw] ; jc 1f ; fnclex\n1:", X86_FEATURE_FXSR, - "m" (tsk->thread.i387.fxsave) - :"memory"); + [fx] "m" (tsk->thread.i387.fxsave), + [fsw] "m" (tsk->thread.i387.fxsave.swd) : "memory"); + /* AMD K7/K8 CPUs don't save/restore FDP/FIP/FOP unless an exception + is pending. Clear the x87 state here by setting it to fixed + values. __per_cpu_offset[0] is a random variable that should be in L1 */ + alternative_input( + GENERIC_NOP8 GENERIC_NOP2, + "emms\n\t" /* clear stack tags */ + "fildl %[addr]", /* set F?P to defined value */ + X86_FEATURE_FXSAVE_LEAK, + [addr] "m" (safe_address)); task_thread_info(tsk)->status &= ~TS_USEDFPU; } diff --git a/include/asm-x86_64/cpufeature.h b/include/asm-x86_64/cpufeature.h index 76bb619..662964b 100644 --- a/include/asm-x86_64/cpufeature.h +++ b/include/asm-x86_64/cpufeature.h @@ -64,6 +64,7 @@ #define X86_FEATURE_REP_GOOD (3*32+ 4) /* rep microcode works well on this CPU */ #define X86_FEATURE_CONSTANT_TSC (3*32+5) /* TSC runs at constant rate */ #define X86_FEATURE_SYNC_RDTSC (3*32+6) /* RDTSC syncs CPU core */ +#define X86_FEATURE_FXSAVE_LEAK (3*32+7) /* FIP/FOP/FDP leaks through FXSAVE */ /* Intel-defined CPU features, CPUID level 0x00000001 (ecx), word 4 */ #define X86_FEATURE_XMM3 (4*32+ 0) /* Streaming SIMD Extensions-3 */ diff --git a/include/asm-x86_64/i387.h b/include/asm-x86_64/i387.h index 876eb9a..cba8a3b 100644 --- a/include/asm-x86_64/i387.h +++ b/include/asm-x86_64/i387.h @@ -72,6 +72,23 @@ extern int set_fpregs(struct task_struct *tsk, #define set_fpu_swd(t,val) ((t)->thread.i387.fxsave.swd = (val)) #define set_fpu_fxsr_twd(t,val) ((t)->thread.i387.fxsave.twd = (val)) +#define X87_FSW_ES (1 << 7) /* Exception Summary */ + +/* AMD CPUs don't save/restore FDP/FIP/FOP unless an exception + is pending. Clear the x87 state here by setting it to fixed + values. The kernel data segment can be sometimes 0 and sometimes + new user value. Both should be ok. + Use the PDA as safe address because it should be already in L1. */ +static inline void clear_fpu_state(struct i387_fxsave_struct *fx) +{ + if (unlikely(fx->swd & X87_FSW_ES)) + asm volatile("fnclex"); + alternative_input(ASM_NOP8 ASM_NOP2, + " emms\n" /* clear stack tags */ + " fildl %%gs:0", /* load to clear state */ + X86_FEATURE_FXSAVE_LEAK); +} + static inline int restore_fpu_checking(struct i387_fxsave_struct *fx) { int err; @@ -119,6 +136,7 @@ static inline int save_i387_checking(struct i387_fxsave_struct __user *fx) #endif if (unlikely(err)) __clear_user(fx, sizeof(struct i387_fxsave_struct)); + /* No need to clear here because the caller clears USED_MATH */ return err; } @@ -149,7 +167,7 @@ static inline void __fxsave_clear(struct task_struct *tsk) "i" (offsetof(__typeof__(*tsk), thread.i387.fxsave))); #endif - __asm__ __volatile__("fnclex"); + clear_fpu_state(&tsk->thread.i387.fxsave); } static inline void kernel_fpu_begin(void) |