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

Re: [Xen-devel] [PATCH v3 2/2] x86/hvm: finish IOREQs correctly on completion path



>>> On 15.03.19 at 14:05, <igor.druzhinin@xxxxxxxxxx> wrote:
> Thanks for clarification. I discussed with Paul - there is definitely
> still a hole in general case where 1st half of the instruction is memory
> and 2nd half is MMIO and the 1st half is changed *to* MMIO. But it's
> hard to deal with these types of accesses without a complete re-write of
> MMIO cache into general insn access cache - so to lift it up to
> linear_{read,write} layer. I hope my understanding is now correct and
> I'll put into the description.

Well, mostly. With patch 1 there's no dependency anymore on an
access to be page straddling afaict. I.e. the scenarios I gave
should apply also to aligned accesses.

> Until then the fix should do fine with scenarios we're seeing.

Indeed.

Jan



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel

 


Rackspace

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