Re: Kernel 2.6.30: Memory/XFS leak, OOM killer kills many processes

[Felix Blyakher]

On Jun 11, 2009, at 12:31 PM, Eric Sandeen wrote:

> Justin Piszcz wrote:
> > On Thu, 11 Jun 2009, Justin Piszcz wrote:
> >
> > > Hello,
> > >
> > > I have a daily cron that backs up my root filesystem using  
> > > xfsdump, it has
> > > remain unchanged for at least 7-10 kernel versions.  When I  
> > > migrated to
> > > 2.6.30, when the xfsdump ran at its scheduled time, nearly all of my
> > > processes were killed due to an OOM situation, I can reproduce the  
> > > situation.
> > >
> > > Kernel: 2.6.30
> > > Dist: Debian Testing
> > > xfsdump: 2.2.48-1
> >
> > Kernel 2.6.29.4 does not exhibit this problem:
> >
> > xfsdump: estimated dump size: 8694781376 bytes
> > xfsdump: creating dump session media file 0 (media 0, file 0)
> > xfsdump: dumping ino map
> > xfsdump: dumping directories
> > xfsdump: dumping non-directory files
> > xfsdump: ending media file
> > xfsdump: media file size 8294709848 bytes
> > xfsdump: dump size (non-dir files) : 8208863560 bytes
> > xfsdump: dump complete: 102 seconds elapsed
> > xfsdump: Dump Status: SUCCESS
> >
> > XFS(?) bug in 2.6.30.
>
> Any chance for a bisect run? :)

Well, Hedi (@sgi) pointed out to the problem without
bisect :)

commit 28e211700a81b0a934b6c7a4b8e7dda843634d2f
Author: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date:   Tue Feb 24 08:39:02 2009 -0500

   xfs: fix getbmap vs mmap deadlock


we do allocate memory for out

   out = kmem_zalloc(bmv->bmv_count * sizeof(struct getbmapx),  
KM_MAYFAIL);

but I am not seeing where it's being released.

If I am reading the code correctly we need to handle the freeing in
in out_unlock_iolock.

The following should fix it:

diff --git a/fs/xfs/xfs_bmap.c b/fs/xfs/xfs_bmap.c
index 4b0f6ef..7928b99 100644
--- a/fs/xfs/xfs_bmap.c
+++ b/fs/xfs/xfs_bmap.c
@@ -6086,6 +6086,7 @@ xfs_getbmap(
                        break;
        }

+       kmem_free(out);
        return error;
 }


Felix

> Or, just as a thought, watch slabtop while you run the dump?
>
> -Eric
> --
> To unsubscribe from this list: send the line "unsubscribe linux- 
> kernel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/