Re: [Bugme-new] [Bug 29302] New: Null pointer dereference with large max_sectors_kb

[Chris Mason <chris.mason@oracle.com>]

Excerpts from Tomas Zvala's message of 2011-02-20 06:06:42 -0500:
> On 18.2.2011 21:18, Chris Mason wrote:
> > Ok, so it isn't part of the open devices code that prints errors, my
> > guess is we're failing to read a good super.
> >
> > Could you please mkfs.btrfs /dev/xxx, sync, then btrfsck /dev/xxx, I want
> > to make sure things are really getting written.
> >
> > Here's a patch that makes sure we have a good bdev after scanning,
> > hopefully it will let us debug things without your box going boom.
> >
> > -chris
> >
> >
> Hello Chris,
> I patched the kernel, recompiled and installed, but I don't see the 
> message in the kernel log.  I forgot to mention before, that 
> max_hw_sectors_kb and thus max_sectors_kb is 4096.

ok, more below.

> 
> Linux stg-top 2.6.37stg #7 SMP Sun Feb 20 10:48:57 CET 2011 x86_64 
> Intel(R) Xeon(R) CPU E5620 @ 2.40GHz GenuineIntel GNU/Linux
> 
> 
> With default max_sectors_kb = 512
> mkfs, sync, btrfsck going through fine
> first mount is fine
> 
> [112005]stg-top ~ # cat /sys/block/sdb/queue/max_sectors_kb
> 512
> [112012]stg-top ~ # mkfs.btrfs /dev/sdb
> 
> WARNING! - Btrfs Btrfs v0.19 IS EXPERIMENTAL
> WARNING! - see http://btrfs.wiki.kernel.org before using
> 
> fs created label (null) on /dev/sdb
>          nodesize 4096 leafsize 4096 sectorsize 4096 size 2.73TB
> Btrfs Btrfs v0.19
> [112045]stg-top ~ # sync
> [112046]stg-top ~ # btrfsck /dev/sdb
> found 28672 bytes used err is 0
> total csum bytes: 0
> total tree bytes: 28672
> total fs tree bytes: 8192
> btree space waste bytes: 23875
> file data blocks allocated: 0
>   referenced 0
> Btrfs Btrfs v0.19
> [112050]stg-top ~ # mount /dev/sdb /mnt/
> [112054]stg-top ~ # df /mnt/
> Filesystem           1K-blocks      Used Available Use% Mounted on
> /dev/sdb             2929685760        56 2927559936   1% /mnt
> [112057]stg-top ~ # umount /mnt
> 
> 
> With max_sectors_kb = 4096
> mkfs, sync, btrfsck is fine
> second mount  BUGs out
> 
> [112100]stg-top ~ # echo 4096 > /sys/block/sdb/queue/max_sectors_kb
> [112107]stg-top ~ # mkfs.btrfs /dev/sdb
> 
> WARNING! - Btrfs Btrfs v0.19 IS EXPERIMENTAL
> WARNING! - see http://btrfs.wiki.kernel.org before using
> 
> fs created label (null) on /dev/sdb
>          nodesize 4096 leafsize 4096 sectorsize 4096 size 2.73TB
> Btrfs Btrfs v0.19
> [112113]stg-top ~ # sync
> [112117]stg-top ~ # btrfsck /dev/sdb
> found 28672 bytes used err is 0
> total csum bytes: 0
> total tree bytes: 28672
> total fs tree bytes: 8192
> btree space waste bytes: 23875
> file data blocks allocated: 0
>   referenced 0
> Btrfs Btrfs v0.19
> [112123]stg-top ~ # mount /dev/sdb /mnt/
> Killed

At this point, your box is broken.  You have to reboot before anything
new is valid (sorry).  Could you please:

Set max_sectors_kb to 4096
mkfs.btrfs /dev/xxx
sync
echo 3 > /proc/sys/vm/drop_caches
btrfsck /dev/xxx

Don't try and mount, clearly my patch isn't saving us.

-chris