On Mon, 2012-07-02 at 11:14 -0400, Theodore Ts'o wrote:
> On Mon, Jul 02, 2012 at 05:54:22PM +0300, Artem Bityutskiy wrote:
> > On Mon, 2012-07-02 at 10:44 -0400, Theodore Ts'o wrote:
> > > Can you tell me *which* xfstest this was blocking on?
> > 
> > Test 068 which runs fsstress.
> 
> Hmmm, I had just started an xfstests -g auto run using standard ext4
> defaults with 3.5-rc2, and I'm not noting any problems.  And I've
> gotten past test 68 w/o any problems.
> 
> Can you tell me more about your KVM config?
> 
> I'm using a 512 megs of memory and 2 CPU's under KVM, with a 5 gig
> test partition running on a 5400 RPM laptop drive.

I have to run home now, but quickly few words about my setup. Yes, it is
100% reproducible, I can probably even try to bisect it. It worked, not
sure in which kernel, but I had xfstests finish successfully. I
reproduce this on stock v3.5-r5 kernel.

This is how I run kvm:

IMAGE="np.img"
DEVEL_DISK1="devel-disk-1.img"
DEVEL_DISK2="devel-disk-2.img"

mv np.output np.output.prev

kvm -m 12384 -spice port=7837,disable-ticketing \
        -cpu core2duo -smp 12,sockets=1,cores=6,threads=2 \
        -enable-kvm -name np \
        -drive file=$IMAGE,if=virtio,cache=none,boot=on \
        -drive file=$DEVEL_DISK1,if=virtio,cache=none \
        -drive file=$DEVEL_DISK2,if=virtio,cache=none \
        -rtc base=utc \
        -net nic,model=virtio -net user \
        -device virtio-balloon-pci \
        -nographic \
        -redir tcp:7838::22 \
| tee np.output

The cpuinfo in kvm is attached.

My host system is a SandyBridge with 6 cores, hyperthreding gives 12.

The disk where the images sit is an Intel SSD.

My hacky scripts to run the tests are also attached. I just run
"./test-ext4". My .config is also attached. Yes, I just create the
file-system with mkfs.ext4 without any additional options.

-- 
Best Regards,
Artem Bityutskiy