perf/x86/p4: Fix "Wunused-but-set-variable" warning
Subsystems (2)
PATCH 1/1 perf/x86/p4: Fix "Wunused-but-set-variable" warning
This fixes a compiler warning when compiling an allyesconfig with W=1:
warning: variable ´high´ set but not used [-Wunused-but-set-variable]
--- arch/x86/events/intel/p4.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/events/intel/p4.c +1 -1
@@ -1366,7 +1366,7 @@ static __initconst const struct x86_pmu p4_pmu = {__init int p4_pmu_init (void ) { -unsigned int low, high; +unsigned int low,__maybe_unused high;int i, reg;/* If we get stripped -- indexing fails */
Signature · 1 line
2.42.0
Peter Zijlstra
maintainerQuoted 2 lines · Lucy Mielke
This fixes a compiler warning when compiling an allyesconfig with W=1:
warning: variable ´high´ set but not used [-Wunused-but-set-variable]
What compiler and what .config?
Quoted 10 lines
Signed-off-by Lucy Mielke <[email protected]>--- arch/x86/events/intel/p4.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)arch/x86/events/intel/p4.c +0
-0
arch/x86/events/intel/p4.c static __initconst const struct x86_pmu p4_pmu = { · line 1372
Quoted 8 lines
arch/x86/events/intel/p4.c +1
-1@@ -1366,7 +1366,7 @@ static __initconst const struct x86_pmu p4_pmu = {__init int p4_pmu_init (void ) { -unsigned int low, high; +unsigned int low,__maybe_unused high;int i, reg;/* If we get stripped -- indexing fails */
Right after this we have:
rdmsr(MSR_IA32_MISC_ENABLE, low, high);
which should get high unconditionally used. If there's a problem then it's probably inside that rdmsr macro.
Quoted 23 lines · Peter Zijlstra
Quoted 2 lines · Lucy Mielke
This fixes a compiler warning when compiling an allyesconfig with W=1:
warning: variable ´high´ set but not used [-Wunused-but-set-variable]What compiler and what .config?
Quoted 17 lines
Signed-off-by Lucy Mielke <[email protected]>--- arch/x86/events/intel/p4.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)arch/x86/events/intel/p4.c +1
-1@@ -1366,7 +1366,7 @@ static __initconst const struct x86_pmu p4_pmu = {__init int p4_pmu_init (void ) { -unsigned int low, high; +unsigned int low,__maybe_unused high;int i, reg;
Quoted 8 lines
arch/x86/events/intel/p4.c static __initconst const struct x86_pmu p4_pmu = { · line 1372
Quoted 1 line
arch/x86/events/intel/p4.c +0
-0/* If we get stripped -- indexing fails */ Right after this we have:
rdmsr(MSR_IA32_MISC_ENABLE, low, high);which should get high unconditionally used. If there's a problem then it's probably inside that rdmsr macro.
rdmsr() writes to "high", but nothing ever reads from high. FWIW, I would _love_ for rdmsrl() to have return semantics, e.g. to be able to do:
low = (u32)rdmsrl(MSR_IA32_MISC_ENABLE);
or even
if (!(rdmsrl(MSR_IA32_MISC_ENABLE) & BIT(7)))
Am Mon, Oct 09, 2023 at 09:29:50AM -0700 schrieb Sean Christopherson:
Quoted 9 lines
rdmsr()writes to "high", but nothing ever reads from high. FWIW, I would _love_ forrdmsrl()to have return semantics, e.g. to be able to do:low = (u32)rdmsrl(MSR_IA32_MISC_ENABLE);or even
if (!(rdmsrl(MSR_IA32_MISC_ENABLE) & BIT(7)))
I have taken a look and it seems to me like this macro is called quite a lot for different things thoughout the kernel tree, including drivers. If one were to change it to have return semantics instead of the way it currently works, you'd have to change around 300 occurences, right? (Let me know if I misunderstood something.)
Return semantics or not, since the only way "high" is used in p4_pmu_init() is by being written to by rdmsr(), the variable can be completely removed by just using rdmsrl(). Would this be a patch you'd be interested in?
Mit freundlichen Grüßen / Best regards,
Lucy
Quoted 7 lines · Lucy Mielke
Am Mon, Oct 09, 2023 at 09:29:50AM -0700 schrieb Sean Christopherson:
Quoted 6 lines
rdmsr()writes to "high", but nothing ever reads from high. FWIW, I would _love_ forrdmsrl()to have return semantics, e.g. to be able to do:low = (u32)rdmsrl(MSR_IA32_MISC_ENABLE);
Quoted 8 lines
Quoted 3 lines
or even
if (!(rdmsrl(MSR_IA32_MISC_ENABLE) & BIT(7)))I have taken a look and it seems to me like this macro is called quite a lot for different things thoughout the kernel tree, including drivers. If one were to change it to have return semantics instead of the way it currently works, you'd have to change around 300 occurences, right?
Yep, which is the only reason I haven't force the issue.