Re: hard freezes with 3.9.0 during io-intensive loads

[Josef Bacik <jbacik@fusionio.com>]

On Sun, May 05, 2013 at 04:25:45AM -0600, Kai Krakow wrote:
> Hello list,
> 
> I've upgraded to 3.9.0 mainly for the snapshot-aware defragging patches. I'm 
> running bedup[1] on a regular basis and it is now the third time that I got 
> back to my PC just to find it hard-frozen and I needed to use the reset 
> button.
> 
> It looks like this happens only while running bedup on my two btrfs 
> filesystems but I'm not sure if it happens for any of the filesystems or 
> only one. This is my setup:
> 
> # cat /etc/fstab (shortened)
> UUID=d2bb232a-2e8f-4951-8bcc-97e237f1b536 / btrfs compress=lzo,subvol=root64 
> 0 1 # /dev/sd{a,b,c}3
> LABEL=usb-backup /mnt/private/usb-backup btrfs noauto,compress-
> force=zlib,subvolid=0,autodefrag,comment=systemd.automount 0 0 # external 
> usb3 disk
> 
> # btrfs filesystem show
> Label: 'usb-backup'  uuid: 7038c8fa-4293-49e9-b493-a9c46e5663ca
>         Total devices 1 FS bytes used 1.13TB
>         devid    1 size 1.82TB used 1.75TB path /dev/sdd1
> 
> Label: 'system'  uuid: d2bb232a-2e8f-4951-8bcc-97e237f1b536
>         Total devices 3 FS bytes used 914.43GB
>         devid    3 size 927.26GB used 426.03GB path /dev/sdc3
>         devid    2 size 927.26GB used 426.03GB path /dev/sdb3
>         devid    1 size 927.26GB used 427.07GB path /dev/sda3
> 
> Btrfs v0.20-rc1
> 
> Since the system hard-freezes I have no messages from dmesg. But I suspect 
> it to be related to the defragmentation option in bedup (I've switched to 
> bedub with --defrag since 3.9.0, and autodefrag for the backup drive). Just 
> in case, I'm going to try without this option now and see if it won't 
> freeze.
> 
> I was able to take a "physical" screenshot with a real camera of a kernel 
> backtrace one time when the freeze happened. I wonder if it is useful to you 
> and where to send it. I just don't want to upload jpegs right here to the 
> list without asking first.
> 
> The big plus is: Altough I had to hard-reset the frozen system several times 
> now, btrfs survived the procedure without any impact (just boot times 
> increases noticeably, probably due to log-replays or something). So thumbs 
> up for the developers on that point.
> 
> [1]: https://github.com/g2p/bedup
> 

Can you please file a bug for this issue on bugzilla.kernel.org so I can make
sure we don't lose track of it?  Make sure the component is set to Btrfs.
Thanks,

Josef