perf bench: Fix undefined behavior in cmpworker()
Subsystems (2)
PATCH v2 1/1 perf bench: Fix undefined behavior in cmpworker()
The comparison function cmpworker() violates the C standard's requirements for qsort() comparison functions, which mandate symmetry and transitivity:
Symmetry: If x < y, then y > x.
Transitivity: If x < y and y < z, then x < z.
In its current implementation, cmpworker() incorrectly returns 0 when w1->tid < w2->tid, which breaks both symmetry and transitivity. This violation causes undefined behavior, potentially leading to issues such as memory corruption in glibc [1].
Fix the issue by returning -1 when w1->tid < w2->tid, ensuring compliance with the C standard and preventing undefined behavior.
---
Changes in v2:
- Rewrite commit message
tools/perf/bench/epoll-wait.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/bench/epoll-wait.c +1 -1
@@ -420,7 +420,7 @@ static int cmpworker(const void *p1, const void *p2)struct worker *w1 = (struct worker *) p1;struct worker *w2 = (struct worker *) p2; -return w1->tid > w2->tid; +return w1->tid > w2->tid ?1 :-1 ; }int bench_epoll_wait (int argc,const char **argv)
Signature · 1 line
2.34.1
James Clark
reviewerQuoted 31 lines · Kuan-Wei Chiu
The comparison function
cmpworker()violates the C standard's requirements forqsort()comparison functions, which mandate symmetry and transitivity:Symmetry: If x < y, then y > x.
Transitivity: If x < y and y < z, then x < z.In its current implementation,
cmpworker()incorrectly returns 0 whenw1->tid<w2->tid, which breaks both symmetry and transitivity. This violation causes undefined behavior, potentially leading to issues such as memory corruption in glibc [1].Fix the issue by returning -1 when
w1->tid<w2->tid, ensuring compliance with the C standard and preventing undefined behavior.Fixes 121dd9ea0116 ("perf bench: Add epoll parallel epoll_wait benchmark")Signed-off-by Kuan-Wei Chiu <[email protected]>---
Changes in v2:
- Rewrite commit messagetools/perf/bench/epoll-wait.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)tools/perf/bench/epoll-wait.c +0
-0
tools/perf/bench/epoll-wait.c static int cmpworker(const void *p1, const void *p2) · line 423
Quoted 5 lines
tools/perf/bench/epoll-wait.c +1
-1@@ -420,7 +420,7 @@ static int cmpworker(const void *p1, const void *p2)struct worker *w1 = (struct worker *) p1;struct worker *w2 = (struct worker *) p2; -return w1->tid > w2->tid; +return w1->tid > w2->tid ?1 :-1 ;
I suppose you can skip the 0 for equality because you know that no two tids are the same?
Anyone looking at this in the future might still think it's still wrong unless it does the full comparison. Even if it's not technically required I would write it like a "normal" one now that we're here:
if (w1->tid > w2->tid) return 1; if (w1->tid < w2->tid) return -1; return 0;
Quoted 29 lines · James Clark
Quoted 26 lines · Kuan-Wei Chiu
The comparison function
cmpworker()violates the C standard's requirements forqsort()comparison functions, which mandate symmetry and transitivity:Symmetry: If x < y, then y > x.
Transitivity: If x < y and y < z, then x < z.In its current implementation,
cmpworker()incorrectly returns 0 whenw1->tid<w2->tid, which breaks both symmetry and transitivity. This violation causes undefined behavior, potentially leading to issues such as memory corruption in glibc [1].Fix the issue by returning -1 when
w1->tid<w2->tid, ensuring compliance with the C standard and preventing undefined behavior.Fixes 121dd9ea0116 ("perf bench: Add epoll parallel epoll_wait benchmark")Signed-off-by Kuan-Wei Chiu <[email protected]>---
Changes in v2:
- Rewrite commit messagetools/perf/bench/epoll-wait.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 13 lines
tools/perf/bench/epoll-wait.c static int cmpworker(const void *p1, const void *p2) · line 422
Quoted 9 lines
tools/perf/bench/epoll-wait.c +1
-1@@ -420,7 +420,7 @@ static int cmpworker(const void *p1, const void *p2)struct worker *w1 = (struct worker *) p1;struct worker *w2 = (struct worker *) p2; -return w1->tid > w2->tid; +return w1->tid > w2->tid ?1 :-1 ;I suppose you can skip the 0 for equality because you know that no two tids are the same?
Yes, exactly.
Quoted 8 lines
Anyone looking at this in the future might still think it's still wrong unless it does the full comparison. Even if it's not technically required I would write it like a "normal" one now that we're here:
if (w1->tid > w2->tid) return 1; if (w1->tid < w2->tid) return -1; return 0;
Sure. I'll make that change in v3.
Regards,
Kuan-Wei