RE: [Xen-users] HVM memory upper limit

["Petersson, Mats" <Mats.Petersson@xxxxxxx>]

> -----Original Message-----
> From: xen-users-bounces@xxxxxxxxxxxxxxxxxxx 
> [mailto:xen-users-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of 
> Marcos Paulo Serafim
> Sent: 07 December 2006 10:33
> To: xen-users@xxxxxxxxxxxxxxxxxxx
> Subject: Re: [Xen-users] HVM memory upper limit
> 
> Petersson, Mats escreveu:
> > Is this 32-bit PAE or 64-bit guest? I believe a 64-bit 
> guest would work, but 32-bit guest would not, as the entire 
> guest-memory is also mapped into QEMU, which fails to find a 
> contiguous memory block of 3GB (as the overall limit of 
> user-space memory is 3GB, and that space has already been 
> fragmented by loading the qemu-dm executable as well as any 
> other memory allocation happening before the mapping of the 
> guest memory. 2GB may well work, but 3GB is definitely more 
> than a 32-bit guest can handle). There was a patch to 
> dynamically map memory, which was sent in the last couple of 
> days - that will solve this problem, and you could, should 
> you wish, give a 32-bit PAE guest 8GB of memory - at least if 
> the patch "does what it says on the label". I haven't seen it 
> go into the tree itself, so I guess it's not going to make 
> 3.0.4, as it's been feature frozen. 
> >
> > --
> > Mats 
> 
> Mats, thanks for your answer and I'm sorry about the lack of
> information. I'm using Xen 3.0.3 with a 32-bit PAE system.
> 
[snip other info on the domain]
> 
> The Xen qemu-dm log points a error:
> 
> domid: 19
> qemu: the number of cpus is 1
> malloc returned error 12

Error 12 = ENOMEM - out of memory, which is what I'd expect "map memory" to say 
if you haven't got a contiguous space big enough to map the memory... 

Error handling with QEMU could be slightly better - particularly, I think xend 
should monitor the qemu-dm process and detect if it's exiting - perhaps kill 
the domain at that point?

Does the processor in your system support 64-bit? If so, you could try a 64-bit 
Xen + Dom0 - that would give qemu all the memory in the world (well, not quite, 
but about 1024TB less some hundreds of gigabyte for the Xen + Linux kernel, 
should be enough for most things in the near future... ;-)

> 
> 
> Well, I'll look for this patch. :D

There was a note from Keir Fraser on xen-devel list:
-- Start Keir's mail --
[Original post by Cui, Dexuan]
<my snipping some other stuff>
> Btw, it seems the patch has not been checked in. Could you give some commetns?

I'm reworking it a bit. It'll go in this morning.

 -- Keir
-- End Keirs mail --

So keep an eye on the unstable tree for the next several hours... Just goes to 
show that I was wrong about it making 3.0.4 (I think).

--
Mats

> 
> Thanks again!
> 
> -- 
> Marcos Paulo Serafim
> Analista de Suporte Técnico / Support Analyst
> mpserafim (at) gmail (dot) com
> Orlandia/SP - Brazil
> MSN: mpserafim (at) gmail (dot) com; Linux User: #171191;
> FP: EB38 D704 A4FD 2677 A2B2 A052 F44E CAA5 45FB 1F42
> 
> 
> _______________________________________________
> Xen-users mailing list
> Xen-users@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-users
> 
> 
> 



_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users