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

Re: [Xen-devel] [PATCH v5] x86: properly handle MSI-X unmask operation from guests




> -----Original Message-----
> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> Sent: Tuesday, November 26, 2013 5:35 PM
> To: Wu, Feng
> Cc: Andrew Cooper (andrew.cooper3@xxxxxxxxxx); Nakajima, Jun; Auld, Will;
> Zhang, Xiantao; xen-devel
> Subject: RE: [PATCH v5] x86: properly handle MSI-X unmask operation from
> guests
> 
> >>> On 26.11.13 at 09:58, "Wu, Feng" <feng.wu@xxxxxxxxx> wrote:
> >> I'm fine with this now, except for some cosmetic things which I
> >> would take the liberty of changing on the fly while committing,
> >> assuming there aren't going to be other comments requiring
> >> another revision.
> >
> > Jan, thank you very much for your review of this patch!
> 
> Actually, as you may have seen already, I just posted v6,
> which - apart from having all the intended cleanup - fixes a
> bug I just spotted.

Yes, I see your changes, the check is needed. 
Normally OS follows the sequences below to operate interrupt:
1) mask
2) change msix data / address
3) unmask
In this case, patch v5 works fine. 

However, If guest operate it like this way:
1) mask
2) change msix data / address
3) mask
4) change msix data /address again
3) unmask
Patch v5 will latch the address the step #3, which will cause problem.

Thanks a lot for this finding.

> 
> Jan


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


 


Rackspace

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