Blackfin: SMP: work around anomaly 05000491

In order to safely work around anomaly 05000491, we have to execute IFLUSH
from L1 instruction sram.  The trouble with multi-core systems is that all
L1 sram is visible only to the active core.  So we can't just place the
functions into L1 and call it directly.  We need to setup a jump table and
place the entry point in external memory.  This will call the right func
based on the active core.

In the process, convert from the manual relocation of a small bit of code
into Core B's L1 to the more general framework we already have in place
for loading arbitrary pieces of code into L1.

Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

8 files changed, 74 insertions, 12 deletions

diff --git a/arch/blackfin/Kconfig b/arch/blackfin/Kconfig
index 0f34ec5..c4cda6f 100644
--- a/arch/blackfin/Kconfig
+++ b/arch/blackfin/Kconfig
@@ -850,7 +850,6 @@ config CPLB_SWITCH_TAB_L1
 config ICACHE_FLUSH_L1
 	bool "Locate icache flush funcs in L1 Inst Memory"
 	default y
-	depends on !SMP
 	help
 	  If enabled, the Blackfin icache flushing functions are linked
 	  into L1 instruction memory.
diff --git a/arch/blackfin/include/asm/smp.h b/arch/blackfin/include/asm/smp.h
index 9dd4873..af6c0aa 100644
--- a/arch/blackfin/include/asm/smp.h
+++ b/arch/blackfin/include/asm/smp.h
@@ -17,7 +17,12 @@
 
 #define raw_smp_processor_id()  blackfin_core_id()
 
-extern char coreb_trampoline_start, coreb_trampoline_end;
+extern void bfin_relocate_coreb_l1_mem(void);
+
+#if defined(CONFIG_SMP) && defined(CONFIG_ICACHE_FLUSH_L1)
+asmlinkage void blackfin_icache_flush_range_l1(unsigned long *ptr);
+extern unsigned long blackfin_iflush_l1_entry[NR_CPUS];
+#endif
 
 struct corelock_slot {
 	int lock;
diff --git a/arch/blackfin/kernel/setup.c b/arch/blackfin/kernel/setup.c
index ac71dc1..805c613 100644
--- a/arch/blackfin/kernel/setup.c
+++ b/arch/blackfin/kernel/setup.c
@@ -215,11 +215,48 @@ void __init bfin_relocate_l1_mem(void)
 
 	early_dma_memcpy_done();
 
+#if defined(CONFIG_SMP) && defined(CONFIG_ICACHE_FLUSH_L1)
+	blackfin_iflush_l1_entry[0] = (unsigned long)blackfin_icache_flush_range_l1;
+#endif
+
 	/* if necessary, copy L2 text/data to L2 SRAM */
 	if (L2_LENGTH && l2_len)
 		memcpy(_stext_l2, _l2_lma, l2_len);
 }
 
+#ifdef CONFIG_SMP
+void __init bfin_relocate_coreb_l1_mem(void)
+{
+	unsigned long text_l1_len = (unsigned long)_text_l1_len;
+	unsigned long data_l1_len = (unsigned long)_data_l1_len;
+	unsigned long data_b_l1_len = (unsigned long)_data_b_l1_len;
+
+	blackfin_dma_early_init();
+
+	/* if necessary, copy L1 text to L1 instruction SRAM */
+	if (L1_CODE_LENGTH && text_l1_len)
+		early_dma_memcpy((void *)COREB_L1_CODE_START, _text_l1_lma,
+				text_l1_len);
+
+	/* if necessary, copy L1 data to L1 data bank A SRAM */
+	if (L1_DATA_A_LENGTH && data_l1_len)
+		early_dma_memcpy((void *)COREB_L1_DATA_A_START, _data_l1_lma,
+				data_l1_len);
+
+	/* if necessary, copy L1 data B to L1 data bank B SRAM */
+	if (L1_DATA_B_LENGTH && data_b_l1_len)
+		early_dma_memcpy((void *)COREB_L1_DATA_B_START, _data_b_l1_lma,
+				data_b_l1_len);
+
+	early_dma_memcpy_done();
+
+#ifdef CONFIG_ICACHE_FLUSH_L1
+	blackfin_iflush_l1_entry[1] = (unsigned long)blackfin_icache_flush_range_l1 -
+			(unsigned long)_stext_l1 + COREB_L1_CODE_START;
+#endif
+}
+#endif
+
 #ifdef CONFIG_ROMKERNEL
 void __init bfin_relocate_xip_data(void)
 {
diff --git a/arch/blackfin/kernel/vmlinux.lds.S b/arch/blackfin/kernel/vmlinux.lds.S
index f5fd234..854fa49 100644
--- a/arch/blackfin/kernel/vmlinux.lds.S
+++ b/arch/blackfin/kernel/vmlinux.lds.S
@@ -176,6 +176,7 @@ SECTIONS
 	{
 		. = ALIGN(4);
 		__stext_l1 = .;
+		*(.l1.text.head)
 		*(.l1.text)
 #ifdef CONFIG_SCHEDULE_L1
 		SCHED_TEXT
diff --git a/arch/blackfin/mach-bf561/secondary.S b/arch/blackfin/mach-bf561/secondary.S
index 148e507..4c46283 100644
--- a/arch/blackfin/mach-bf561/secondary.S
+++ b/arch/blackfin/mach-bf561/secondary.S
@@ -13,7 +13,11 @@
 #include <asm/asm-offsets.h>
 #include <asm/trace.h>
 
-__INIT
+/*
+ * This code must come first as CoreB is hardcoded (in hardware)
+ * to start at the beginning of its L1 instruction memory.
+ */
+.section .l1.text.head
 
 /* Lay the initial stack into the L1 scratch area of Core B */
 #define INITIAL_STACK	(COREB_L1_SCRATCH_START + L1_SCRATCH_LENGTH - 12)
@@ -160,7 +164,6 @@ ENTRY(_coreb_trampoline_start)
 .LWAIT_HERE:
 	jump .LWAIT_HERE;
 ENDPROC(_coreb_trampoline_start)
-ENTRY(_coreb_trampoline_end)
 
 #ifdef CONFIG_HOTPLUG_CPU
 .section ".text"
diff --git a/arch/blackfin/mach-bf561/smp.c b/arch/blackfin/mach-bf561/smp.c
index 1074a7e..82b94e1 100644
--- a/arch/blackfin/mach-bf561/smp.c
+++ b/arch/blackfin/mach-bf561/smp.c
@@ -30,18 +30,11 @@ void __init platform_init_cpus(void)
 
 void __init platform_prepare_cpus(unsigned int max_cpus)
 {
-	int len;
-
-	len = &coreb_trampoline_end - &coreb_trampoline_start + 1;
-	BUG_ON(len > L1_CODE_LENGTH);
-
-	dma_memcpy((void *)COREB_L1_CODE_START, &coreb_trampoline_start, len);
+	bfin_relocate_coreb_l1_mem();
 
 	/* Both cores ought to be present on a bf561! */
 	cpu_set(0, cpu_present_map); /* CoreA */
 	cpu_set(1, cpu_present_map); /* CoreB */
-
-	printk(KERN_INFO "CoreB bootstrap code to SRAM %p via DMA.\n", (void *)COREB_L1_CODE_START);
 }
 
 int __init setup_profiling_timer(unsigned int multiplier) /* not supported */
diff --git a/arch/blackfin/mach-common/cache.S b/arch/blackfin/mach-common/cache.S
index 85aadeb..9f4dd35 100644
--- a/arch/blackfin/mach-common/cache.S
+++ b/arch/blackfin/mach-common/cache.S
@@ -69,10 +69,30 @@
 #endif
 
 /* Invalidate all instruction cache lines assocoiated with this memory area */
+#ifdef CONFIG_SMP
+# define _blackfin_icache_flush_range _blackfin_icache_flush_range_l1
+#endif
 ENTRY(_blackfin_icache_flush_range)
 	do_flush IFLUSH
 ENDPROC(_blackfin_icache_flush_range)
 
+#ifdef CONFIG_SMP
+.text
+# undef _blackfin_icache_flush_range
+ENTRY(_blackfin_icache_flush_range)
+	p0.L = LO(DSPID);
+	p0.H = HI(DSPID);
+	r3 = [p0];
+	r3 = r3.b (z);
+	p2 = r3;
+	p0.L = _blackfin_iflush_l1_entry;
+	p0.H = _blackfin_iflush_l1_entry;
+	p0 = p0 + (p2 << 2);
+	p1 = [p0];
+	jump (p1);
+ENDPROC(_blackfin_icache_flush_range)
+#endif
+
 #ifdef CONFIG_DCACHE_FLUSH_L1
 .section .l1.text
 #else
diff --git a/arch/blackfin/mach-common/smp.c b/arch/blackfin/mach-common/smp.c
index 5f7617d..6e17a26 100644
--- a/arch/blackfin/mach-common/smp.c
+++ b/arch/blackfin/mach-common/smp.c
@@ -40,6 +40,10 @@
  */
 struct corelock_slot corelock __attribute__ ((__section__(".l2.bss")));
 
+#ifdef CONFIG_ICACHE_FLUSH_L1
+unsigned long blackfin_iflush_l1_entry[NR_CPUS];
+#endif
+
 void __cpuinitdata *init_retx_coreb, *init_saved_retx_coreb,
 	*init_saved_seqstat_coreb, *init_saved_icplb_fault_addr_coreb,
 	*init_saved_dcplb_fault_addr_coreb;