[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#637234: linux-image-3.0.0-1-686-pae: I/O errors using ext4 under xen (also affects ext3 as of linux-image-3.1.0-1-amd64 et al)



notfound 637234 3.2.6-1
notfound 637234 2.6.32-41
thanks

Timo & all,

This has been a bug in the dom0 kernel. If I've been following this bug correctly, then the only thing about the 3.0+ kernels is that they have barriers enabled by default, which we then disable as a workaround. But this is supposed to work, and it was mishandled by the kernel in dom0. I therefore see no point in marking this bug as found in 3.0+ kernels used in domU's.

This bug was apparently fixed in 2.6.32-40 or 2.6.32-41, I waited for 2.6.32-41 and now I'm running that kernel in my dom0. You have to upgrade your dom0 to the latest kernel. I now have various wheezy domU's with barriers enabled again, running with no issues.

Regards,

Gedalya


On 02/07/2012 05:17 AM, Timo Juhani Lindfors wrote:
package linux-2.6
notfound 637234 3.0.0-3
found 637234 3.1.0-1~experimental.1
found 637234 3.1.1-1
found 637234 3.1.8-2
found 637234 3.2.4-1
thanks

dom0
====

amd64 squeeze with
Linux version 2.6.32-5-xen-amd64 (Debian 2.6.32-39) (dannf@debian.org) (gcc version 4.3.5 (Debian 4.3.5-4) ) #1 SMP Thu Nov 3 05:42:31 UTC 2011

config
------

name = 'lindi2'
vcpus = '4'
memory = '1024'
disk = [ 'file:/local/xen/lindi2/disk.img,xvda,w' ]
vif = [ 'mac=00:01:01:99:80:02' ]
_on_crash_ = 'restart'

domU
====

amd64 wheezy with
Linux version 3.1.0-1-amd64 (Debian 3.1.0-1~experimental.1) (waldi@debian.org) (gcc version 4.6.1 (Debian 4.6.1-16) ) #1 SMP Thu Nov 3 19:35:59 UTC 2011

(also occurs with 3.1.1-1, 3.1.8-2 and 3.2.4-1 but listing the first
version where the bug occurs)

Using ext3 for rootfs. dmesg has

...
blkfront: barrier: empty write xvda op failed
blkfront: xvda: barrier or flush: disabled
end_request: I/O error, dev xvda, sector 16519664
end_request: I/O error, dev xvda, sector 16519664
...

Full dmesg attached.



workaround
==========

Setting barrier=0 for / in fstab seems to help. Thanks Konrad for the
tip, now I can continue working :-)



Reply to: