Skip to contentSkip to conversation

powerpc/pseries: Fix hcall tracepoints with JUMP_LABEL=n

PATCH Applied Nicholas Piggin 2 participants 3 messages May 9, 2023 – Aug 31, 2023 PowerPC

lore
Share & tools
lore

Nicholas Piggin

author reviewer
#0 lore raw

PATCH 1/2 powerpc/pseries: Fix hcall tracepoints with JUMP_LABEL=n

With JUMP_LABEL=n, hcall_tracepoint_refcount's address is being tested instead of its value. This results in the tracing slowpath always being taken unnecessarily.

Fixes 9a10ccb29c0a2 ("powerpc/pseries: move hcall_tracepoint_refcount out of .toc")
---
 arch/powerpc/platforms/pseries/hvCall.S | 1 +
 1 file changed, 1 insertion(+)
arch/powerpc/platforms/pseries/hvCall.S +1 -0
index 35254ac7af5e..ca0674b0b683 100644
@@ -91,6 +91,7 @@ BEGIN_FTR_SECTION;						\
 	b	1f;						\
 END_FTR_SECTION(0, 1);						\
 	LOAD_REG_ADDR(r12, hcall_tracepoint_refcount) ;		\
+	ld	r12,0(r12);					\
 	std	r12,32(r1);					\
 	cmpdi	r12,0;						\
 	bne-	LABEL;						\
Signature · 1 line
2.40.1
author reviewer
#1 lore raw

PATCH 2/2 powerpc/pseries: Remove unused hcall tracing instruction

When JUMP_LABEL=n, the tracepoint refcount test in the pre-call stores the refcount value to the stack, so the same value can be used for the post-call (presumably to avoid racing with the value concurrently changing).

On little-endian (ELFv2) that might have just worked by luck, because 32(r1) is STK_PARAM(R3) there and so the value save gets clobbered by the tracing code when it's non-zero, but fortunately r3 is the hcall number and 0 is an invalid hcall number so it should get clobbered by another non-zero value. In any case, commit cc1adb5f32557 ("powerpc/pseries: Use jump labels for hcall tracepoints") removed the code that actually used the value stored, so now it's just dead code.

It's fragile to be storing to the stack like this, and confusing. Better remove it.

---
 arch/powerpc/platforms/pseries/hvCall.S | 1 -
 1 file changed, 1 deletion(-)
arch/powerpc/platforms/pseries/hvCall.S +0 -1
index ca0674b0b683..bae45b358a09 100644
@@ -92,7 +92,6 @@ BEGIN_FTR_SECTION;						\
 END_FTR_SECTION(0, 1);						\
 	LOAD_REG_ADDR(r12, hcall_tracepoint_refcount) ;		\
 	ld	r12,0(r12);					\
-	std	r12,32(r1);					\
 	cmpdi	r12,0;						\
 	bne-	LABEL;						\
 1:
Signature · 1 line
2.40.1
#2 lore raw
In reply to Nicholas Piggin
Quoted 5 lines · Nicholas Piggin

With JUMP_LABEL=n, hcall_tracepoint_refcount's address is being tested instead of its value. This results in the tracing slowpath always being taken unnecessarily.

Applied to powerpc/next.

[1/2] powerpc/pseries: Fix hcall tracepoints with JUMP_LABEL=n

      https://git.kernel.org/powerpc/c/750bd41aeaeb1f0e0128aa4f8fcd6dd759713641
[2/2] powerpc/pseries: Remove unused hcall tracing instruction
      https://git.kernel.org/powerpc/c/61d7ebe0376e2640ba77be16e186b1a6c77eb3f7

cheers