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

Re: [Xen-devel] bug disabling guest interface

On Thu, Mar 7, 2013 at 2:34 AM, Jacek Milewicz
<jacekowski@xxxxxxxxxxxxxx> wrote:
>> I'm interested in knowing more information. Have you always been using
>> MAX_SKB_FRAGS > default_value (16 or 17) for you DomUs?
> 18 used to be default value on older kernels (anything older than
> https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/commit/include/linux/skbuff.h?id=9d4dde5215779f4099730194ad30624fdba3d8b2
>  -
> 2011-12-23 21:51:18)
>> If it is this case, you're likely to be hitting this error path all the
>> time. It is just
>> that XSA-39 chooses to shutdown vif if there is something wrong. So my
>> further questions is, do you notice any performance problem in the case
>> that
>> netback silently drops your packets?
> Basically what happened in our case is VMs were migrated to new host with
> new kernel, before that dom0 used to run kernel with MAX_SKB_FRAGS==18 so it
> was same for both dom0 and domU.
> New host has kernel with XSA-39 fixes and smaller MAX_SKB_FRAGS (17), and
> that's when we encountered that bug, we've never tried kernel with smaller
> MAX_SKB_FRAGS but without XSA_39 fix.

Sorry for my intrusion of this thread.  I would like to highlight that
this also affect guests/domUs running Windows Server 2003 in HVM and
the logs are similar with the OP:

kernel: vif vif-2-0 vifvm104.0: Too many frags
kernel: vif vif-2-0 vifvm104.0: fatal error; disabling device
kernel: eth0: port 4(vifvm104.0) entered disabled state


Kindest regards,
Giam Teck Choon

Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.