Bug 516469

Summary: PTHREAD_PRIO_INHERIT does not mix well with POSIX condition variables
Product: [Fedora] Fedora Reporter: Lennart Poettering <lpoetter>
Component: glibcAssignee: Andreas Schwab <schwab>
Status: CLOSED RAWHIDE QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: medium Docs Contact:
Priority: low    
Version: rawhideCC: dino, dvhltc, itamar, jakub, kernel-maint, pmuller, schwab
Target Milestone: ---   
Target Release: ---   
Hardware: All   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2009-08-17 16:10:42 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Bug Depends On:    
Bug Blocks: 473303, 516995    
Attachments:
Description Flags
test case none

Description Lennart Poettering 2009-08-10 02:05:25 UTC 
The attached test case blocks forever on kernel-2.6.31-0.139.rc5.git3.fc12.x86_64/glibc-2.10.90-12.x86_64. It works fine on Program works as intended on kernel-2.6.31-0.38.rc1.git7.fc12.i586/glibc-2.10.1-2.i686.

Disabling PTHREAD_PRIO_INHERIT for the mutex in question makes the test case work properly.

Not sure if this is a kernel or glibc problem.
Comment 1 Lennart Poettering 2009-08-10 02:06:22 UTC 
Created attachment 356836 [details]
test case
Comment 2 Lennart Poettering 2009-08-10 14:25:09 UTC 
peterz thinks it's a kernel problem. Reassigning.
Comment 3 Andreas Schwab 2009-08-10 14:55:26 UTC 
No, it isn't.

<http://sourceware.org/ml/libc-alpha/2009-08/msg00024.html>
Comment 4 Darren Hart 2009-08-10 16:21:02 UTC 
Took a look at the trace, the kernel is doing what it should.  If you put the thread to sleep with FUTEX_WAIT_REQUEUE_PI_PRIVATE (0x8b) you have to wake it up with FUTEX_CMP_REQUEUE_PI_PRIVATE (0x8c), but the trace shows a call to FUTEX_WAKE_PRIVATE instead.  This causes the kernel routine the thread is blocked in to wake, detect invalid usage, and return EINVAL.  This is a glibc bug.
Comment 5 Andreas Schwab 2009-08-17 16:10:42 UTC 
Fixed in 2.10.90-14.