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

Re: [Xen-devel] [PATCH] mm: sync vmalloc address space page tables in alloc_vm_area()



On Mon, 2011-11-28 at 09:36 +0000, Jan Beulich wrote:
> >>> On 07.11.11 at 21:36, Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx> 
> >>> wrote:
> > Patch included as attachment for easier review..
> 
> I just noticed that this patch made it upstream meanwhile, and that I
> should have paid more attention earlier: Once again this adds x86
> specific bits to (supposedly) architecture independent Xen code
> (lookup_address(), use of GNTMAP_contains_pte). Unless everyone
> agrees that x86 is going to remain the only architecture Xen will support
> going forward (no ia64, no ARM, nothing else), patches doing so (at
> least outside proper #ifdef-s or alike) should really be rejected.

I think it is up to those interested in such architectures to ensure
that a working baseline exists in the first place.

The ARM stuff hasn't even been submitted yet. When the arm stuff is
submitted it will naturally include fixes for these sorts issues as
necessary and at that point we can talk about regressions and reviewing
patches in order to avoid them (until then we can't really know what a
"regression" is). There is nothing unusual about that and nothing about
patches we take right now commit us to never supporting another arch in
the future so lets not get carried away here.

The IA64 support in mainline Linux does not appear to have anyone
interested in working on it. AFAICT it hasn't really been touched (other
than odd fixes and tree-wide cleanups) since it was first committed
(circa 2.6.25 IIRC). It doesn't even build right now and looks like it
hasn't built since at least 2.6.36, based on the one failure I happened
to look at.

I know you've been working on fixing up the hypervisor side of ia64
support things recently but it is not clear that there is an existing or
viable user or developer base for that port right now.

> Besides that, the patch also appears to close the road to running
> backends in HVM - use of GNTMAP_contains_pte isn't permitted for
> paging_mode_external() guests, so it's even a step backwards for
> x86.

That is a bigger concern.

Ian.



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


 


Rackspace

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