Re: [PATCH] [RFC] remove ext3 inode from orphan list when link and unlink race

[Eric Sandeen <sandeen@redhat.com>]

Dmitriy Monakhov wrote:
> Eric Sandeen <sandeen@redhat.com> writes:
> 
>> I've been looking at a case where many threads are opening, unlinking, and
>> hardlinking files on ext3 .
> How many concurent threads do you use and how long does it takes to trigger 
> this race? I've tried to reproduce this with two threads, but not succeed.
> <thread 1>  
>         fd = create("src")
>         close(fd)
>         unlink("src")
> <thread 2>
>         link("src", "dst")
>         unlink("dst")
> 
> Original testcase will be the best answer :).

Sure :)  Though I didn't write it... see this collection of bash scripts:

http://people.redhat.com/esandeen/testcases/orphan-repro.tar.bz2

I didn't write it, but it exposed the bug for me.  The VAR file contains 
variables to specify mountpoint and a device, which the script starts by 
mkfs'ing, so be warned.

It spawns -many- threads, and on my 4 CPU opteron I can hit it in a 
reasonable amount of time.  It would probably be nice to have a more 
targeted testcase but it did the trick for me.

Thanks,
-Eric

> Thanks.