Bug 2843 - Race between pthread_cancel and pthread_join (tst-join5)
Summary: Race between pthread_cancel and pthread_join (tst-join5)
Status: RESOLVED FIXED
Alias: None
Product: glibc
Classification: Unclassified
Component: nptl (show other bugs)
Version: unspecified
: P2 normal
Target Milestone: ---
Assignee: Ulrich Drepper
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2006-06-24 20:36 UTC by Paul Brook
Modified: 2016-05-17 18:21 UTC (History)
1 user (show)

See Also:
Host: i686-pc-linux-gnu
Target: i686-pc-linux-gnu
Build: i686-pc-linux-gnu
Last reconfirmed:
fweimer: security-

Attachments
Testcase (893 bytes, text/x-csrc)
2006-06-24 20:37 UTC, Paul Brook 		Details
Patch (591 bytes, patch)
2006-06-24 20:37 UTC, Paul Brook 		Details | Diff
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Paul Brook 2006-06-24 20:36:14 UTC 
I'm seeing intermittent failures in nptl/ tst-join5

Consider two threads, A and B:

Thread A calls pthread_cancel(B); pthread_join(B).
Some time later thread B calls pthread_join(A).
pthread_join is a cancellation point, so I would expect thread B to be cancelled
at that point. However what thread B sees thread A waiting and mistakenly
returns EDEADLK.

The attached testcase is based on tst-join5, with an extra busy look added to
make the failure reproducible.

The patch fixes this by checking if the current thread has been cancelled before
returning EDEADLK.
Comment 1 Paul Brook 2006-06-24 20:37:15 UTC 
Created attachment 1122 [details]
Testcase

Based on nptl/tst-join5.c
Comment 2 Paul Brook 2006-06-24 20:37:57 UTC 
Created attachment 1123 [details]
Patch
Comment 3 Ulrich Drepper 2006-08-13 01:56:17 UTC 
There is a problem but the patch is wrong.  I checked in a correct patch.