Re: [Xen-devel] Xen pciback not working: "address space collision: [mem ...] conflicts with System RAM [mem ...]"

[Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>]

On Fri, Aug 02, 2013 at 02:30:24PM +0200, Andrea Brugiolo wrote:
> On Fri, Aug 02, 2013 at 08:04:09AM -0400, Konrad Rzeszutek Wilk wrote:
> > On Fri, Aug 02, 2013 at 11:07:35AM +0200, Andrea Brugiolo wrote:
> > > On Mon, Jul 29, 2013 at 01:55:16PM -0400, Konrad Rzeszutek Wilk wrote:
> > > > On Mon, Jul 29, 2013 at 10:02:03AM +0100, Ian Campbell wrote:
> > > > > On Fri, 2013-07-26 at 12:32 +0200, Andrea Brugiolo wrote:
> > > > > > Good Morning
> > > > > > 
> > > > > > I cannot do pciback anymore for both my second scsi controller and 
> > > > > > my
> > > > > > second network card: when I try to pass the device to the domU I get
> > > > > > this error in system logs:
> > > > > > 
> > > > > >   ... address space collision: [mem ...] conflicts with System RAM 
> > > > > > [mem ...]
> > > > > 
> > > > > By eliding the actually addresses you've omitted something which I 
> > > > > think
> > > > > might be interesting:
> > > > >         [mem 0xf9e00000-0xf9e1ffff 64bit] conflicts with System RAM 
> > > > > [mem 0x00100000-0x4007fffff]
> > > > > 
> > > > > Note that there is not any actual overlap in those two sets of 
> > > > > addresses...
> > > > 
> > > > I think it is:
> > > > mem 0xf9e00000-0xf9e1ffff
> > > > mem 0x00100000-0x4007fffff
> > > > 
> > > > The RAM region is pretty much all of the memory. This looks like the 
> > > > 'e820_hole'
> > > > parameter is not being used? (It only works for xl btw).
> > > > 
> > > > > 
> > > > > Might be that the check is truncating something, or maybe it is
> > > > > confusing MFN and PFN and so getting a false +ve. Both wild guesses
> > > > > having not even looked at the code...
> > > > > 
> > > > > > The problem is described here:
> > > > > > http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=717353 and is
> > > > > > blocking my backup system which is based on a scsi tape changer
> > > > > > attached to the domU.
> > > > > 
> > > > > What do the guest and host e820 map look like? Actually the full dmesg
> > > > > for the hypervisor, dom0 and domU kernels would be useful to provide, 
> > > > > I
> > > > > expect.
> > > > 
> > > > And the guest config pls.
> > > 
> > > At last I have managed to reboot the system. Please find the attachments:
> > > 
> > >  - dom0 dmesg
> > >  - domU dmesg
> > >  - domU configuration
> > > 
> > > Recall:
> > > 
> > >  - dom0 is Debian GNU/Linux 7.1 with Debian kernel 3.2.0-4-amd64 
> > > (3.2.46-1)
> > >  - domU is Debian GNU/Linux 6.0.7 with the same kernel
> > > 
> > > The "address space collision" shows up for both the devices I am
> > > trying to pass as I have been doing for years.
> > 
> > So you are using 'xm', but 'xm' does not support 'e820_hole=1'. You need
> > to use 'xl'. The domU E820 is as I suspected without the host E820 which
> > is why you are hitting the issue.
> > 
> > Note, I did at some point post an implementation of 'e820_hole=1'
> > argument for Xend, but since Xen is being deprecated ... it didn't make
> > much sense adding it in.
> > 
> > Awaiting your response with the usage of 'xl'.
> 
> Do you really mean e820_hole=1 or do you mean e820_host=1?

Oh boy. Thank you for pointing this out to me. I meant 'e820_host'
<sigh> No wonder my suggestions did not work :-(

> 
> Thank you
> 
> Andrea

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