Skip to contentSkip to conversation

Remove indirect variable usage at arch/x86/kernel/tsc.c

PATCH Superseded Subrata Modak 3 participants 4 messages May 19, 2009 – May 24, 2009 x86

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Remove indirect variable usage at arch/x86/kernel/tsc.c

Hi,

The following warning is generated on compilation:

CC arch/x86/kernel/tsc.o
arch/x86/kernel/tsc.c: In function 'time_cpufreq_notifier':
arch/x86/kernel/tsc.c:634: warning: 'dummy' may be used uninitialized in this function

However, there seems to be no practical usage of variable 'dummy' in the following piece of code:

630 static int time_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
631 void *data)
632 {
633 struct cpufreq_freqs *freq = data;
634 unsigned long *lpj, dummy;
635
636 if (cpu_has(&cpu_data(freq->cpu), X86_FEATURE_CONSTANT_TSC))
637 return 0;
638
639 lpj = &dummy;
640 if (!(freq->flags & CPUFREQ_CONST_LOOPS))
641 #ifdef CONFIG_SMP
642 lpj = &cpu_data(freq->cpu).loops_per_jiffy;
643 #else
644 lpj = &boot_cpu_data.loops_per_jiffy;
645 #endif
646

'lpj' probably will get to point to some address after this if() statement.

647 if (!ref_freq) {
648 ref_freq = freq->old;
649 loops_per_jiffy_ref = *lpj;

And, if it does, then "loops_per_jiffy_ref" will have a proper value, else, even with "lpj = &dummy" will not gurantee "loops_per_jiffy_ref = *lpj" to have the expected value.

650 tsc_khz_ref = tsc_khz;
651 }
652 if ((val == CPUFREQ_PRECHANGE && freq->old < freq->new) ||

653                         (val == CPUFREQ_POSTCHANGE && freq->old > freq->new) ||
654                         (val == CPUFREQ_RESUMECHANGE)) {
655                 *lpj =  cpufreq_scale(loops_per_jiffy_ref, ref_freq, freq->new);

656
657 tsc_khz = cpufreq_scale(tsc_khz_ref, ref_freq, freq->new);
658 if (!(freq->flags & CPUFREQ_CONST_LOOPS))
659 mark_tsc_unstable("cpufreq changes");
660 }
661
662 set_cyc2ns_scale(tsc_khz, freq->cpu);
663
664 return 0;
665 }

Is there any specific reason for 'dummy' to exist in this function ? If not, i would like to propose the following fix.

To: <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Balbir Singh <[email protected]>
Cc: Sachin P Sant <[email protected]>
Cc: Linux Kernel <[email protected]>
Cc: Andi Kleen <[email protected]>
Subject: [PATCH] Remove indirect variable usage at arch/x86/kernel/tsc.c
---

arch/x86/kernel/tsc.c +1 -2
@@ -631,12 +631,11 @@ static int time_cpufreq_notifier(struct 
 				void *data)
 {
 	struct cpufreq_freqs *freq = data;
-	unsigned long *lpj, dummy;
+	unsigned long *lpj = NULL;
 
 	if (cpu_has(&cpu_data(freq->cpu), X86_FEATURE_CONSTANT_TSC))
 		return 0;
 
-	lpj = &dummy;
 	if (!(freq->flags & CPUFREQ_CONST_LOOPS))
 #ifdef CONFIG_SMP
 		lpj = &cpu_data(freq->cpu).loops_per_jiffy;

---
Regards--
Subrata

#1 lore raw
Quoted 6 lines · Subrata Modak

640 if (!(freq->flags & CPUFREQ_CONST_LOOPS))
641 #ifdef CONFIG_SMP
642 lpj = &cpu_data(freq->cpu).loops_per_jiffy;
643 #else
644 lpj = &boot_cpu_data.loops_per_jiffy;
645 #endif

No comments on your patch, but it might be good to also fix the incorrect indentation here: line 644 needs an extra tab.

Cheers,
FJP

#2 lore raw
Quoted 10 lines · Frans Pop
Quoted 6 lines · Subrata Modak

640 if (!(freq->flags & CPUFREQ_CONST_LOOPS))
641 #ifdef CONFIG_SMP
642 lpj = &cpu_data(freq->cpu).loops_per_jiffy;
643 #else
644 lpj = &boot_cpu_data.loops_per_jiffy;
645 #endif

No comments on your patch, but it might be good to also fix the incorrect indentation here: line 644 needs an extra tab.

Sure, i will do that in the updated patch.

Regards--
Subrata

Quoted 3 lines

Cheers,
FJP

#3 lore raw
In reply to Subrata Modak

Hi!

Quoted 27 lines

CC arch/x86/kernel/tsc.o
arch/x86/kernel/tsc.c: In function 'time_cpufreq_notifier':
arch/x86/kernel/tsc.c:634: warning: 'dummy' may be used uninitialized in this function

However, there seems to be no practical usage of variable 'dummy' in the following piece of code:

630 static int time_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
631 void *data)
632 {
633 struct cpufreq_freqs *freq = data;
634 unsigned long *lpj, dummy;
635
636 if (cpu_has(&cpu_data(freq->cpu), X86_FEATURE_CONSTANT_TSC))
637 return 0;
638
639 lpj = &dummy;
640 if (!(freq->flags & CPUFREQ_CONST_LOOPS))
641 #ifdef CONFIG_SMP
642 lpj = &cpu_data(freq->cpu).loops_per_jiffy;
643 #else
644 lpj = &boot_cpu_data.loops_per_jiffy;
645 #endif
646

'lpj' probably will get to point to some address after this if() statement.

Quoted 7 lines

647 if (!ref_freq) {
648 ref_freq = freq->old;
649 loops_per_jiffy_ref = *lpj;

And, if it does, then "loops_per_jiffy_ref" will have a proper value, else, even with "lpj = &dummy" will not gurantee "loops_per_jiffy_ref = *lpj" to have the expected value.

But that's a bug to be fixed, I'd say? ... actually I believe you are introducing a bug here. Yes, old code would put random numbers in loops_per_jiffy_ref for !CPUFREQ_CONST_LOOPS, but you are introducing oops there.

Have you tested the code?

Quoted 15 lines
arch/x86/kernel/tsc.c +1 -2
@@ -631,12 +631,11 @@ static int time_cpufreq_notifier(struct 
 				void *data)
 {
 	struct cpufreq_freqs *freq = data;
-	unsigned long *lpj, dummy;
+	unsigned long *lpj = NULL;
 
 	if (cpu_has(&cpu_data(freq->cpu), X86_FEATURE_CONSTANT_TSC))
 		return 0;
 
-	lpj = &dummy;
 	if (!(freq->flags & CPUFREQ_CONST_LOOPS))
 #ifdef CONFIG_SMP
 		lpj = &cpu_data(freq->cpu).loops_per_jiffy;
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html