Bug 9371 - Reproducable hang with 2.6.24-rc2 using oprofile
Reproducable hang with 2.6.24-rc2 using oprofile
Status: CLOSED CODE_FIX
Product: Platform Specific/Hardware
Classification: Unclassified
Component: i386
All Linux
: P1 normal
Assigned To: Ingo Molnar
:
Depends on:
Blocks: 9243
  Show dependency treegraph
 
Reported: 2007-11-13 13:18 UTC by Rafael J. Wysocki
Modified: 2007-11-20 02:39 UTC (History)
3 users (show)

See Also:
Kernel Version: 2.6.24-rc2
Tree: Mainline
Regression: Yes


Attachments

Description Rafael J. Wysocki 2007-11-13 13:18:47 UTC
Subject         : Reproducable hang with 2.6.24-rc2 using oprofile
Submitter       : Jan Kara <jack@suse.cz>
References      : http://lkml.org/lkml/2007/11/13/225
Comment 1 Jan Kara 2007-11-19 08:35:13 UTC
I've bisected a bit and the following commit is the one causing it:

commit 574a60421c8ea5383a54ebee1f37fa871d00e1b9
Author: Jan Blunck <jblunck@suse.de>
Date:   Fri Oct 19 20:35:03 2007 +0200

    i386: make callgraph use dump_trace() on i386/x86_64

Comment 2 Ingo Molnar 2007-11-20 01:07:50 UTC
does 2.6.24-rc3 fix it? It has an oprofile fix:

| commit df9d177aa28d50e64bae6fbd6b263833079e3571
| Author: Philippe Elie <phil.el@wanadoo.fr>
| Date:   Wed Nov 14 16:58:48 2007 -0800
|
|    oProfile: oops when profile_pc() returns ~0LU

if that does not help then we'll revert the commit you bisected so excellently.
Comment 3 Jan Kara 2007-11-20 02:07:31 UTC
It helps! The kernel does not hang any more and I'm even able to get some numbers from oprofile :). Thanks. I just wonder how come I didn't see any message on the console - I used standard VGA console to maximize the possibility I'd see something before hang...
Comment 4 Ingo Molnar 2007-11-20 02:39:50 UTC
profiling might crash in places where the crash handler itself profiles too, etc. - resulting in a hard hang.

nmi_watchdog=2 (or =1) could perhaps have gotten a backtrace.

Note You need to log in before you can comment on or make changes to this bug.