Re: kernel BUG at fs/btrfs/delayed-inode.c:1466!

[Jacek Luczak]

2012/3/9 David Sterba <dave@xxxxxxxx>:
> On Fri, Mar 09, 2012 at 12:08:12PM +0100, Jacek Luczak wrote:
>> For this one I've created also a report [1].
>> >
>> > so there is probably other problem in reservations and it just blew up during
>> > the unlink call.
>>
>> Could be as this came up after a longer time of throwing above WARN_ON.
>>
>> I'm now cloning the Linus tree. Lets see if both will pop up on there.
>
> The 3.3-rc6 should help in one case, with
>
> http://thread.gmane.org/gmane.comp.file-systems.btrfs/15268
>
> but I was able to reproduce the WARN_ON even with this patch, didn't get
> to debugging it again yet.
>


The story so far looks like this:
1) kernel 3.2.7:
- on the BUG_ON triggers after a longer while of CI env (doing builds)
running. This has been already reproduced twice.
- WARN_ON spams heavily, even after BUG_ON pop up.
- possible relation between WARN_ON and BUG_ON.
2) A *regression* in 3.3.0-rc6-00197-g9f8050c
- completely unusable as reports ENOSPC
- to reproduce, mount volume and issue:
# CNT=1 ; while [ $CNT -lt 10000 ] ; do  rm -f /btrfs/dd ; ! touch
/btrfs/dd && echo "$CNT" && break  ; CNT=$(( $CNT + 1 )) ; done
On my host this shows:
# CNT=1 ; while [ $CNT -lt 10000 ] ; do  rm -f /btrfs/dd ; ! touch
/btrfs/dd && echo "$CNT" && break  ; CNT=$(( $CNT + 1 )) ; done
touch: cannot touch `/btrfs/dd': No space left on device
423
- remount to reset:
# CNT=1 ; while [ $CNT -lt 10000 ] ; do  rm -f /btrfs/dd ; ! touch
/btrfs/dd && echo "$CNT" && break  ; CNT=$(( $CNT + 1 )) ; done
touch: cannot touch `/btrfs/dd': No space left on device
1
# umount /btrfs/
# mount -t btrfs /dev/vg00/btrfs /btrfs/ -o noatime,nodatacow,defaults
# CNT=1 ; while [ $CNT -lt 10000 ] ; do  rm -f /btrfs/dd ; ! touch
/vdd && echo "$CNT" && break  ; CNT=$(( $CNT + 1 )) ; done
touch: cannot touch `/btrfs/dd': No space left on device
423
- bisected down to 5500cdb (Btrfs: increase the global block reserve
estimates). After reverting this one Linus master works for me again.

-Jacek
--
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/