Skip to contentSkip to conversation

powerpc/pseries: Use jump labels for hcall tracepoints

PATCH New Anton Blanchard 1 participant 2 messages Jul 3, 2014

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 powerpc/pseries: Use jump labels for hcall tracepoints

hcall tracepoints add quite a few instructions to our hcall path:

plpar_hcall:

	mr      r2,r2
	mfcr    r0
	stw     r0,8(r1)
	b       164		<---- start
	ld      r12,0(r2)
	std     r12,32(r1)
	cmpdi   r12,0
	beq     164		<---- end

...

We have an unconditional branch that gets noped out during boot and a load/compare/branch. We also store the tracepoint value to the stack for the hcall_exit path to use.

By using jump labels we can simplify this to just a single nop that gets replaced with a branch when the tracepoint is enabled:

plpar_hcall:

	mr      r2,r2
	mfcr    r0
	stw     r0,8(r1)
	nop			<----

...

If jump labels are not enabled, we fall back to the old method.

---

arch/powerpc/include/asm/jump_label.h +9 -0
Index: b/arch/powerpc/include/asm/jump_label.h
===================================================================
@@ -10,6 +10,7 @@
  * 2 of the License, or (at your option) any later version.
  */
 
+#ifndef __ASSEMBLY__
 #include <linux/types.h>
 
 #include <asm/feature-fixups.h>
@@ -42,4 +43,12 @@ struct jump_entry {
 	jump_label_t key;
 };
 
+#else
+#define ARCH_STATIC_BRANCH(LABEL, KEY)		\
+1098:	nop;					\
+	.pushsection __jump_table, "aw";	\
+	FTR_ENTRY_LONG 1098b, LABEL, KEY;	\
+	.popsection
+#endif
+
 #endif /* _ASM_POWERPC_JUMP_LABEL_H */
arch/powerpc/platforms/pseries/hvCall.S +110 -34

238 lines not on this page · raw message

arch/powerpc/platforms/pseries/lpar.c +23 -7
Index: b/arch/powerpc/platforms/pseries/lpar.c
===================================================================
@@ -26,6 +26,7 @@
 #include <linux/dma-mapping.h>
 #include <linux/console.h>
 #include <linux/export.h>
+#include <linux/static_key.h>
 #include <asm/processor.h>
 #include <asm/mmu.h>
 #include <asm/page.h>
@@ -649,6 +650,19 @@ EXPORT_SYMBOL(arch_free_page);
 #endif
 
 #ifdef CONFIG_TRACEPOINTS
+#ifdef CONFIG_JUMP_LABEL
+struct static_key hcall_tracepoint_key = STATIC_KEY_INIT;
+
+void hcall_tracepoint_regfunc(void)
+{
+	static_key_slow_inc(&hcall_tracepoint_key);
+}
+
+void hcall_tracepoint_unregfunc(void)
+{
+	static_key_slow_dec(&hcall_tracepoint_key);
+}
+#else
 /*
  * We optimise our hcall path by placing hcall_tracepoint_refcount
  * directly in the TOC so we can check if the hcall tracepoints are
@@ -658,13 +672,6 @@ EXPORT_SYMBOL(arch_free_page);
 /* NB: reg/unreg are called while guarded with the tracepoints_mutex */
 extern long hcall_tracepoint_refcount;
 
-/* 
- * Since the tracing code might execute hcalls we need to guard against
- * recursion. One example of this are spinlocks calling H_YIELD on
- * shared processor partitions.
- */
-static DEFINE_PER_CPU(unsigned int, hcall_trace_depth);
-
 void hcall_tracepoint_regfunc(void)
 {
 	hcall_tracepoint_refcount++;
@@ -674,6 +681,15 @@ void hcall_tracepoint_unregfunc(void)
 {
 	hcall_tracepoint_refcount--;
 }
+#endif
+
+/*
+ * Since the tracing code might execute hcalls we need to guard against
+ * recursion. One example of this are spinlocks calling H_YIELD on
+ * shared processor partitions.
+ */
+static DEFINE_PER_CPU(unsigned int, hcall_trace_depth);
+
 
 void __trace_hcall_entry(unsigned long opcode, unsigned long *args)
 {
#1 lore raw

PATCH 2/2 powerpc/pseries: optimise hcall tracepoints

Now that we execute the hcall tracepoint entry and exit code out of line, we can use the same stack across both functions.

---

arch/powerpc/platforms/pseries/hvCall.S +16 -22
Index: b/arch/powerpc/platforms/pseries/hvCall.S
===================================================================
@@ -46,33 +46,27 @@ hcall_tracepoint_refcount:
 	addi	r4,r1,STK_PARAM(FIRST_REG);			\
 	stdu	r1,-STACK_FRAME_OVERHEAD(r1);			\
 	bl	__trace_hcall_entry;				\
-	addi	r1,r1,STACK_FRAME_OVERHEAD;			\
-	ld	r0,16(r1);					\
-	ld	r3,STK_PARAM(R3)(r1);				\
-	ld	r4,STK_PARAM(R4)(r1);				\
-	ld	r5,STK_PARAM(R5)(r1);				\
-	ld	r6,STK_PARAM(R6)(r1);				\
-	ld	r7,STK_PARAM(R7)(r1);				\
-	ld	r8,STK_PARAM(R8)(r1);				\
-	ld	r9,STK_PARAM(R9)(r1);				\
-	ld	r10,STK_PARAM(R10)(r1);				\
-	mtlr	r0
+	ld	r3,STACK_FRAME_OVERHEAD+STK_PARAM(R3)(r1);	\
+	ld	r4,STACK_FRAME_OVERHEAD+STK_PARAM(R4)(r1);	\
+	ld	r5,STACK_FRAME_OVERHEAD+STK_PARAM(R5)(r1);	\
+	ld	r6,STACK_FRAME_OVERHEAD+STK_PARAM(R6)(r1);	\
+	ld	r7,STACK_FRAME_OVERHEAD+STK_PARAM(R7)(r1);	\
+	ld	r8,STACK_FRAME_OVERHEAD+STK_PARAM(R8)(r1);	\
+	ld	r9,STACK_FRAME_OVERHEAD+STK_PARAM(R9)(r1);	\
+	ld	r10,STACK_FRAME_OVERHEAD+STK_PARAM(R10)(r1)
 
 /*
  * postcall is performed immediately before function return which
  * allows liberal use of volatile registers.
  */
 #define __HCALL_INST_POSTCALL					\
-	mflr	r0;						\
-	ld	r6,STK_PARAM(R3)(r1);				\
-	std	r3,STK_PARAM(R3)(r1);				\
+	ld	r0,STACK_FRAME_OVERHEAD+STK_PARAM(R3)(r1);	\
+	std	r3,STACK_FRAME_OVERHEAD+STK_PARAM(R3)(r1);	\
 	mr	r4,r3;						\
-	mr	r3,r6;						\
-	std	r0,16(r1);					\
-	stdu	r1,-STACK_FRAME_OVERHEAD(r1);			\
+	mr	r3,r0;						\
 	bl	__trace_hcall_exit;				\
+	ld	r0,STACK_FRAME_OVERHEAD+16(r1);			\
 	addi	r1,r1,STACK_FRAME_OVERHEAD;			\
-	ld	r0,16(r1);					\
 	ld	r3,STK_PARAM(R3)(r1);				\
 	mtlr	r0
 
@@ -279,14 +273,14 @@ plpar_hcall9_trace:
 	mr	r7,r8
 	mr	r8,r9
 	mr	r9,r10
-	ld	r10,STK_PARAM(R11)(r1)
-	ld	r11,STK_PARAM(R12)(r1)
-	ld	r12,STK_PARAM(R13)(r1)
+	ld	r10,STACK_FRAME_OVERHEAD+STK_PARAM(R11)(r1)
+	ld	r11,STACK_FRAME_OVERHEAD+STK_PARAM(R12)(r1)
+	ld	r12,STACK_FRAME_OVERHEAD+STK_PARAM(R13)(r1)
 
 	HVSC
 
 	mr	r0,r12
-	ld	r12,STK_PARAM(R4)(r1)
+	ld	r12,STACK_FRAME_OVERHEAD+STK_PARAM(R4)(r1)
 	std	r4,0(r12)
 	std	r5,8(r12)
 	std	r6,16(r12)