wait: No child processes

Denys Vlasenko vda.linux at googlemail.com
Tue Dec 27 16:53:15 UTC 2011


2011/12/27 Dmitry V. Levin <ldv at altlinux.org>:
> On Thu, Nov 17, 2011 at 10:25:26AM +0100, Łukasz Michalik wrote:
>> I'm currently debugging an issue in sydbox wrt recent linux kernels.
>> I initially thought strace wasn't affected by it, but then debian bug
>> #622863 [1] made me investigate further, althought it's impossible to
>> say what version was that user using.  It seems that unreleased code
>> clean up in HEAD exposes the bug, namely the 19cdada5b commit.
>>
>> The bug only happens on recent linux.  I've bisected the kernel and
>> found that it was introduced by 9b84cca2564b9 [2] in linux.git, so
>> pretty much everything since 3.0-rc2 is affected.
>
> Łukasz, thanks for letting us aware of the issue.  I had a hope that
> Denys as the author of that cleanup series would comment it, but
> I haven't heard from him yet.

I can confirm that bug triggers with current git and kernel 3.1.6;
sometime later I will try reverting kernel commit 9b84cca2564b9.

>> I'm attaching a testcase that makes strace -f fail within seconds
>> (compile with g++ -std=c++0x -pthread); it just fires whole bunch of
>> threads that fork and wait for exiting children.

Please find modified testcase attached. Modifications make it
compilable with only plain C, not C++, compiler; also,
I changed exit codes so it's a bit more easy to see under strace
who's exiting.

Attached LOG.bz2 is an example of a full strace -f session
which caught a bug.

-- 
vda
-------------- next part --------------
A non-text attachment was scrubbed...
Name: testcase1.c
Type: text/x-csrc
Size: 910 bytes
Desc: not available
URL: <http://lists.strace.io/pipermail/strace-devel/attachments/20111227/5e782524/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: LOG.bz2
Type: application/x-bzip2
Size: 5316 bytes
Desc: not available
URL: <http://lists.strace.io/pipermail/strace-devel/attachments/20111227/5e782524/attachment-0001.bin>


More information about the Strace-devel mailing list