Check whether the TLB operations need broadcasting on SMP systems

ARMv7 SMP hardware can handle the TLB maintenance operations
broadcasting in hardware so that the software can avoid the costly IPIs.
This patch adds the necessary checks (the MMFR3 CPUID register) to avoid
the broadcasting if already supported by the hardware.

(this patch is based on the work done by Tony Thompson @ ARM)
Signed-off-by: NCatalin Marinas <catalin.marinas@arm.com>

arch/arm/include/asm/cputype.h

@@ -8,6 +8,21 @@
 #define CPUID_TCM	2
 #define CPUID_TLBTYPE	3
 
+#define CPUID_EXT_PFR0	"c1, 0"
+#define CPUID_EXT_PFR1	"c1, 1"
+#define CPUID_EXT_DFR0	"c1, 2"
+#define CPUID_EXT_AFR0	"c1, 3"
+#define CPUID_EXT_MMFR0	"c1, 4"
+#define CPUID_EXT_MMFR1	"c1, 5"
+#define CPUID_EXT_MMFR2	"c1, 6"
+#define CPUID_EXT_MMFR3	"c1, 7"
+#define CPUID_EXT_ISAR0	"c2, 0"
+#define CPUID_EXT_ISAR1	"c2, 1"
+#define CPUID_EXT_ISAR2	"c2, 2"
+#define CPUID_EXT_ISAR3	"c2, 3"
+#define CPUID_EXT_ISAR4	"c2, 4"
+#define CPUID_EXT_ISAR5	"c2, 5"
+
 #ifdef CONFIG_CPU_CP15
 #define read_cpuid(reg)							\
 	({								\
@@ -18,9 +33,19 @@
 		    : "cc");						\
 		__val;							\
 	})
+#define read_cpuid_ext(ext_reg)						\
+	({								\
+		unsigned int __val;					\
+		asm("mrc	p15, 0, %0, c0, " ext_reg		\
+		    : "=r" (__val)					\
+		    :							\
+		    : "cc");						\
+		__val;							\
+	})
 #else
 extern unsigned int processor_id;
 #define read_cpuid(reg) (processor_id)
+#define read_cpuid_ext(reg) 0
 #endif
 
 /*

arch/arm/include/asm/tlbflush.h

@@ -40,6 +40,12 @@
 #define TLB_V6_I_ASID	(1 << 18)
 
 #define TLB_BTB		(1 << 28)
+
+/* Unified Inner Shareable TLB operations (ARMv7 MP extensions) */
+#define TLB_V7_UIS_PAGE	(1 << 19)
+#define TLB_V7_UIS_FULL (1 << 20)
+#define TLB_V7_UIS_ASID (1 << 21)
+
 #define TLB_L2CLEAN_FR	(1 << 29)		/* Feroceon */
 #define TLB_DCLEAN	(1 << 30)
 #define TLB_WB		(1 << 31)
@@ -176,9 +182,17 @@
 # define v6wbi_always_flags	(-1UL)
 #endif
 
+#ifdef CONFIG_SMP
+#define v7wbi_tlb_flags (TLB_WB | TLB_DCLEAN | TLB_BTB | \
+			 TLB_V7_UIS_FULL | TLB_V7_UIS_PAGE | TLB_V7_UIS_ASID)
+#else
+#define v7wbi_tlb_flags (TLB_WB | TLB_DCLEAN | TLB_BTB | \
+			 TLB_V6_U_FULL | TLB_V6_U_PAGE | TLB_V6_U_ASID)
+#endif
+
 #ifdef CONFIG_CPU_TLB_V7
-# define v7wbi_possible_flags	v6wbi_tlb_flags
-# define v7wbi_always_flags	v6wbi_tlb_flags
+# define v7wbi_possible_flags	v7wbi_tlb_flags
+# define v7wbi_always_flags	v7wbi_tlb_flags
 # ifdef _TLB
 #  define MULTI_TLB 1
 # else
@@ -316,6 +330,8 @@ static inline void local_flush_tlb_all(void)
 		asm("mcr p15, 0, %0, c8, c6, 0" : : "r" (zero) : "cc");
 	if (tlb_flag(TLB_V4_I_FULL | TLB_V6_I_FULL))
 		asm("mcr p15, 0, %0, c8, c5, 0" : : "r" (zero) : "cc");
+	if (tlb_flag(TLB_V7_UIS_FULL))
+		asm("mcr p15, 0, %0, c8, c3, 0" : : "r" (zero) : "cc");
 
 	if (tlb_flag(TLB_BTB)) {
 		/* flush the branch target cache */
@@ -351,6 +367,8 @@ static inline void local_flush_tlb_mm(struct mm_struct *mm)
 		asm("mcr p15, 0, %0, c8, c6, 2" : : "r" (asid) : "cc");
 	if (tlb_flag(TLB_V6_I_ASID))
 		asm("mcr p15, 0, %0, c8, c5, 2" : : "r" (asid) : "cc");
+	if (tlb_flag(TLB_V7_UIS_ASID))
+		asm("mcr p15, 0, %0, c8, c3, 2" : : "r" (asid) : "cc");
 
 	if (tlb_flag(TLB_BTB)) {
 		/* flush the branch target cache */
@@ -389,6 +407,8 @@ local_flush_tlb_page(struct vm_area_struct *vma, unsigned long uaddr)
 		asm("mcr p15, 0, %0, c8, c6, 1" : : "r" (uaddr) : "cc");
 	if (tlb_flag(TLB_V6_I_PAGE))
 		asm("mcr p15, 0, %0, c8, c5, 1" : : "r" (uaddr) : "cc");
+	if (tlb_flag(TLB_V7_UIS_PAGE))
+		asm("mcr p15, 0, %0, c8, c3, 1" : : "r" (uaddr) : "cc");
 
 	if (tlb_flag(TLB_BTB)) {
 		/* flush the branch target cache */
@@ -424,6 +444,8 @@ static inline void local_flush_tlb_kernel_page(unsigned long kaddr)
 		asm("mcr p15, 0, %0, c8, c6, 1" : : "r" (kaddr) : "cc");
 	if (tlb_flag(TLB_V6_I_PAGE))
 		asm("mcr p15, 0, %0, c8, c5, 1" : : "r" (kaddr) : "cc");
+	if (tlb_flag(TLB_V7_UIS_PAGE))
+		asm("mcr p15, 0, %0, c8, c3, 1" : : "r" (kaddr) : "cc");
 
 	if (tlb_flag(TLB_BTB)) {
 		/* flush the branch target cache */

arch/arm/kernel/smp.c

@@ -32,6 +32,7 @@
 #include <asm/processor.h>
 #include <asm/tlbflush.h>
 #include <asm/ptrace.h>
+#include <asm/cputype.h>
 
 /*
  * as from 2.5, kernels no longer have an init_tasks structure
@@ -545,6 +546,12 @@ struct tlb_args {
 	unsigned long ta_end;
 };
 
+/* all SMP configurations have the extended CPUID registers */
+static inline int tlb_ops_need_broadcast(void)
+{
+	return ((read_cpuid_ext(CPUID_EXT_MMFR3) >> 12) & 0xf) < 2;
+}
+
 static inline void ipi_flush_tlb_all(void *ignored)
 {
 	local_flush_tlb_all();
@@ -587,51 +594,61 @@ static inline void ipi_flush_tlb_kernel_range(void *arg)
 
 void flush_tlb_all(void)
 {
-	on_each_cpu(ipi_flush_tlb_all, NULL, 1);
+	if (tlb_ops_need_broadcast())
+		on_each_cpu(ipi_flush_tlb_all, NULL, 1);
+	else
+		local_flush_tlb_all();
 }
 
 void flush_tlb_mm(struct mm_struct *mm)
 {
-	on_each_cpu_mask(ipi_flush_tlb_mm, mm, 1, &mm->cpu_vm_mask);
+	if (tlb_ops_need_broadcast())
+		on_each_cpu_mask(ipi_flush_tlb_mm, mm, 1, &mm->cpu_vm_mask);
+	else
+		local_flush_tlb_mm(mm);
 }
 
 void flush_tlb_page(struct vm_area_struct *vma, unsigned long uaddr)
 {
-	struct tlb_args ta;
-
-	ta.ta_vma = vma;
-	ta.ta_start = uaddr;
-
-	on_each_cpu_mask(ipi_flush_tlb_page, &ta, 1, &vma->vm_mm->cpu_vm_mask);
+	if (tlb_ops_need_broadcast()) {
+		struct tlb_args ta;
+		ta.ta_vma = vma;
+		ta.ta_start = uaddr;
+		on_each_cpu_mask(ipi_flush_tlb_page, &ta, 1, &vma->vm_mm->cpu_vm_mask);
+	} else
+		local_flush_tlb_page(vma, uaddr);
 }
 
 void flush_tlb_kernel_page(unsigned long kaddr)
 {
-	struct tlb_args ta;
-
-	ta.ta_start = kaddr;
-
-	on_each_cpu(ipi_flush_tlb_kernel_page, &ta, 1);
+	if (tlb_ops_need_broadcast()) {
+		struct tlb_args ta;
+		ta.ta_start = kaddr;
+		on_each_cpu(ipi_flush_tlb_kernel_page, &ta, 1);
+	} else
+		local_flush_tlb_kernel_page(kaddr);
 }
 
 void flush_tlb_range(struct vm_area_struct *vma,
                      unsigned long start, unsigned long end)
 {
-	struct tlb_args ta;
-
-	ta.ta_vma = vma;
-	ta.ta_start = start;
-	ta.ta_end = end;
-
-	on_each_cpu_mask(ipi_flush_tlb_range, &ta, 1, &vma->vm_mm->cpu_vm_mask);
+	if (tlb_ops_need_broadcast()) {
+		struct tlb_args ta;
+		ta.ta_vma = vma;
+		ta.ta_start = start;
+		ta.ta_end = end;
+		on_each_cpu_mask(ipi_flush_tlb_range, &ta, 1, &vma->vm_mm->cpu_vm_mask);
+	} else
+		local_flush_tlb_range(vma, start, end);
 }
 
 void flush_tlb_kernel_range(unsigned long start, unsigned long end)
 {
-	struct tlb_args ta;
-
-	ta.ta_start = start;
-	ta.ta_end = end;
-
-	on_each_cpu(ipi_flush_tlb_kernel_range, &ta, 1);
+	if (tlb_ops_need_broadcast()) {
+		struct tlb_args ta;
+		ta.ta_start = start;
+		ta.ta_end = end;
+		on_each_cpu(ipi_flush_tlb_kernel_range, &ta, 1);
+	} else
+		local_flush_tlb_kernel_range(start, end);
 }

arch/arm/mm/proc-v7.S

@@ -176,8 +176,8 @@ cpu_v7_name:
  */
 __v7_setup:
 #ifdef CONFIG_SMP
-	mrc	p15, 0, r0, c1, c0, 1		@ Enable SMP/nAMP mode
-	orr	r0, r0, #(0x1 << 6)
+	mrc	p15, 0, r0, c1, c0, 1		@ Enable SMP/nAMP mode and
+	orr	r0, r0, #(1 << 6) | (1 << 0)	@ TLB ops broadcasting
 	mcr	p15, 0, r0, c1, c0, 1
 #endif
 	adr	r12, __v7_setup_stack		@ the local stack

arch/arm/mm/tlb-v7.S

@@ -42,9 +42,11 @@ ENTRY(v7wbi_flush_user_tlb_range)
 	mov	r1, r1, lsl #PAGE_SHIFT
 	vma_vm_flags r2, r2			@ get vma->vm_flags
 1:
-	mcr	p15, 0, r0, c8, c6, 1		@ TLB invalidate D MVA (was 1)
-	tst	r2, #VM_EXEC			@ Executable area ?
-	mcrne	p15, 0, r0, c8, c5, 1		@ TLB invalidate I MVA (was 1)
+#ifdef CONFIG_SMP
+	mcr	p15, 0, r0, c8, c3, 1		@ TLB invalidate U MVA (shareable) 
+#else
+	mcr	p15, 0, r0, c8, c7, 1		@ TLB invalidate U MVA
+#endif
 	add	r0, r0, #PAGE_SZ
 	cmp	r0, r1
 	blo	1b
@@ -69,8 +71,11 @@ ENTRY(v7wbi_flush_kern_tlb_range)
 	mov	r0, r0, lsl #PAGE_SHIFT
 	mov	r1, r1, lsl #PAGE_SHIFT
 1:
-	mcr	p15, 0, r0, c8, c6, 1		@ TLB invalidate D MVA
-	mcr	p15, 0, r0, c8, c5, 1		@ TLB invalidate I MVA
+#ifdef CONFIG_SMP
+	mcr	p15, 0, r0, c8, c3, 1		@ TLB invalidate U MVA (shareable)
+#else
+	mcr	p15, 0, r0, c8, c7, 1		@ TLB invalidate U MVA
+#endif
 	add	r0, r0, #PAGE_SZ
 	cmp	r0, r1
 	blo	1b
@@ -87,5 +92,5 @@ ENDPROC(v7wbi_flush_kern_tlb_range)
 ENTRY(v7wbi_tlb_fns)
 	.long	v7wbi_flush_user_tlb_range
 	.long	v7wbi_flush_kern_tlb_range
-	.long	v6wbi_tlb_flags
+	.long	v7wbi_tlb_flags
 	.size	v7wbi_tlb_fns, . - v7wbi_tlb_fns