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

Re: [Xen-devel] [PATCH 7/7] xl: Fix 'free_memory' to include outstanding_claims value.



On Tue, Apr 16, 2013 at 05:45:20PM +0100, Tim Deegan wrote:
> At 16:33 +0100 on 16 Apr (1366129996), Ian Jackson wrote:
> > konrad wilk writes ("Re: [PATCH 7/7] xl: Fix 'free_memory' to include 
> > outstanding_claims value."):
> > > On 4/15/2013 12:50 PM, Ian Jackson wrote:
> > > > I have checked this and the race is in the hypercall API.  The
> > > > hypercall API has already been checked in.  So, under the
> > > > circumstances, for this patch:
> > ...
> > > Right. Let me enumerate some ideas for fixing this on a different thread 
> > > (if we have the same race in mind that you spotted).
> > 
> > The race I'm thinking of is this:
> > 
> > When we display the effective amount of free memory (in "xl info"
> > etc.), we calculate it as
> >   physinfo->free_pages - xc_domain_get_outstanding_pages()
> > 
> > But these two values have been retrieved at different times.  So while
> > a domain is being built and memory moves from "free but claimed" to
> > "in use", the free memory visible via the libxl API will sometimes be
> > "wrong".
> > 
> > This may seem like a minor point, but it will show up as occasional
> > glitches in automatic monitoring and graphing systems; it might even
> > trigger spurious nagios alerts in higher layers etc.  If that was all
> > there was to it I wouldn't regard it as a release-critical bug - a
> > race like that would be annoying but could be fixed later.
> > 
> > However, the race is baked into the hypercall API/ABI which we want to
> > keep relatively stable at least in stable releases.
> > 
> > I think the right answer is probably simply to move the total claim
> > value into the physinfo struct, and do away with the separate
> > XENMEM_get_outstanding_pages memory op hypercall.  Do you agree ?
> 
> FWIW, I think this is a good idea.  You might have to be a little
> careful on the hypervisor side to make sure the two values are actually
> a matched pair (say by taking the page allocator lock).

<nods> Going to prep a patch for that. I might not have it ready this week
as Linus merge window is immienient and I need to queue up patches (And test).

And then right after that I am out for a week.

But after that - I will have the patch ready.
> 
> Tim.

_______________________________________________
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®.