Skip to contentSkip to conversation

sched: fix user time incorrectly accounted as system time on 32 bit

PATCH In review Stanislaw Gruszka 2 participants 3 messages Sep 14, 2010 – Sep 15, 2010

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 sched: fix user time incorrectly accounted as system time on 32 bit

We have 32 bit variable overflow possibility when multiply in task_times() and thread_group_times() functions. If overflow happens calculated scaled utime value become wrongly small and scaled stime wrongly big.

Reported here:
https://bugzilla.redhat.com/show_bug.cgi?id=633037
https://bugzilla.kernel.org/show_bug.cgi?id=16559

Cc Hidetoshi Seto <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc <[email protected]> # 2.6.32.19+ (partially) and 2.6.33+

---
v1 -> v2
Fix mail recipients.

 kernel/sched.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)
kernel/sched.c +4 -4
index 09b574e..eee9470 100644
@@ -3507,9 +3507,9 @@ void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
 	rtime = nsecs_to_cputime(p->se.sum_exec_runtime);
 
 	if (total) {
-		u64 temp;
+		u64 temp = rtime;
 
-		temp = (u64)(rtime * utime);
+		temp *= utime;
 		do_div(temp, total);
 		utime = (cputime_t)temp;
 	} else
@@ -3540,9 +3540,9 @@ void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
 	rtime = nsecs_to_cputime(cputime.sum_exec_runtime);
 
 	if (total) {
-		u64 temp;
+		u64 temp = rtime;
 
-		temp = (u64)(rtime * cputime.utime);
+		temp *= cputime.utime;
 		do_div(temp, total);
 		utime = (cputime_t)temp;
 	} else
Signature · 1 line
1.7.1
#1 lore raw
Quoted 4 lines · Stanislaw Gruszka

We have 32 bit variable overflow possibility when multiply in task_times() and thread_group_times() functions. If overflow happens calculated scaled utime value become wrongly small and scaled stime wrongly big.

Ah, right, (u64)foo * bar would have worked too I guess, but yeah, the used (u64)(foo * bar) is wrong.

Quoted 10 lines

Reported here:
https://bugzilla.redhat.com/show_bug.cgi?id=633037
https://bugzilla.kernel.org/show_bug.cgi?id=16559

Cc Hidetoshi Seto <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc <[email protected]> # 2.6.32.19+ (partially) and 2.6.33+

Thanks, got it.

#2 lore raw

[tip:sched/urgent] sched: Fix user time incorrectly accounted as system time on 32-bit

In reply to Stanislaw Gruszka
Commit-ID:  e75e863dd5c7d96b91ebbd241da5328fc38a78cc
Gitweb:     http://git.kernel.org/tip/e75e863dd5c7d96b91ebbd241da5328fc38a78cc
Author:     Stanislaw Gruszka <[email protected]>
AuthorDate: Tue, 14 Sep 2010 16:35:14 +0200
Committer:  Ingo Molnar <[email protected]>
CommitDate: Wed, 15 Sep 2010 10:41:36 +0200

sched: Fix user time incorrectly accounted as system time on 32-bit

We have 32-bit variable overflow possibility when multiply in task_times() and thread_group_times() functions. When the overflow happens then the scaled utime value becomes erroneously small and the scaled stime becomes i erroneously big.

Reported here:

https://bugzilla.redhat.com/show_bug.cgi?id=633037
https://bugzilla.kernel.org/show_bug.cgi?id=16559

LKML-Reference: <[email protected]>

---
 kernel/sched.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)
kernel/sched.c +4 -4
index ed09d4f..dc85ceb 100644
@@ -3513,9 +3513,9 @@ void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
 	rtime = nsecs_to_cputime(p->se.sum_exec_runtime);
 
 	if (total) {
-		u64 temp;
+		u64 temp = rtime;
 
-		temp = (u64)(rtime * utime);
+		temp *= utime;
 		do_div(temp, total);
 		utime = (cputime_t)temp;
 	} else
@@ -3546,9 +3546,9 @@ void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
 	rtime = nsecs_to_cputime(cputime.sum_exec_runtime);
 
 	if (total) {
-		u64 temp;
+		u64 temp = rtime;
 
-		temp = (u64)(rtime * cputime.utime);
+		temp *= cputime.utime;
 		do_div(temp, total);
 		utime = (cputime_t)temp;
 	} else