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

Re: [Xen-devel] [BUG] Kernel BUG in page_alloc.c (mismatched start and end zone) using xl generated e820 map



On Tue, Jun 02, 2015 at 02:58:17PM +0000, Simon Waterman wrote:
> Hi,
> 
> We're hitting the kernel BUG below in one of our VMs running on Xen 4.4 and
> Linux kernel 3.13.0.  We use the xl toolstack and are using PCI pass-through
> to pass network cards and a disk controller.  It happens on a variety of our
> hardware but not all servers and it seems to be related to the e820 map
> passed by xl.
> 
> The problem occurs when we put the server under heavy load - the 'dd' command
> at the top of the stack trace seems to be sufficient to cause the problem if
> run a few times.
> 
> We didn't get a problem with previous versions of Xen (we were using 4.2.2)
> but at that time we were using xend and as I understand it the RAM map
> provided to the guest is fabricated rather than based upon the real hardware 
> map.

Right - older kenrnels did not have it. You can also do 'e820_hole=0' and 
that should make the E820 be fabricated.

Have you tried with an later kernel? Say 4.0? There were some boundary
problems with the older kernels, but sadly I do not remember the exact
commits.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel


 


Rackspace

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