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

[Keir Fraser <keir.xen@xxxxxxxxx>]

On 02/09/2011 08:22, "Ian Campbell" <Ian.Campbell@xxxxxxxxxx> wrote:

> On Thu, 2011-09-01 at 21:37 +0100, Jeremy Fitzhardinge wrote:
>> On 09/01/2011 09:11 AM, Konrad Rzeszutek Wilk wrote:
>>> On Thu, Sep 01, 2011 at 12:51:03PM +0100, David Vrabel wrote:
>>>> From: David Vrabel <david.vrabel@xxxxxxxxxx>
>>> Andrew,
>>> 
>>> I was wondering if you would be Ok with this patch for 3.1.
>>> 
>>> It is a revert (I can prepare a proper revert if you would like
>>> that instead of this patch).
>>> 
>>> The users of this particular function (alloc_vm_area) are just
>>> Xen. There are no others.
>> 
>> I'd prefer to put explicit vmalloc_sync_all()s in the callsites where
>> necessary, and ultimately try to work out ways of avoiding it altogether
>> (like have some hypercall wrapper which touches the arg memory to make
>> sure its mapped?).
> 
> That only syncs the current pagetable though. If that is sufficient (and
> it could well be) then perhaps just doing a vmalloc_sync_one on the
> current page tables directly would be better than faulting to do it?

It's probably sufficient unless the kernel has non-voluntary preemption, and
we are not preempt_disable()d.

 -- Keir

> It's the sort of thing you could hide inside the gnttab_set_map_op type
> helpers I guess?
> 
> Ian.
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-devel



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