Re: [Xen-devel] [PATCH v7 01/21] xen: make two memory hypercalls vNUMA-aware

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 09.03.15 at 14:29, <wei.liu2@xxxxxxxxxx> wrote:
> On Mon, Mar 09, 2015 at 01:22:55PM +0000, Jan Beulich wrote:
>> >>> On 09.03.15 at 13:51, <wei.liu2@xxxxxxxxxx> wrote:
>> > @@ -759,6 +784,14 @@ long do_memory_op(unsigned long cmd, 
> XEN_GUEST_HANDLE_PARAM(void) arg)
>> >              return start_extent;
>> >          args.domain = d;
>> >  
>> > +        if ( construct_memop_from_reservation(d, &reservation, &args) )
>> > +        {
>> > +            rcu_unlock_domain(d);
>> > +            return start_extent;
>> > +        }
>> > +        args.nr_done      = start_extent;
>> > +        args.preempted    = 0;
>> 
>> Remembering Andrew's comment on your patch introducing
>> construct_memop_from_reservation(), the setting of
>> args.domain visible in the context above should either be
>> moved past that call, or be used by the function instead of
>> passing d explicitly (in which case it is pretty clear that the
>> function isn't free to clobber that structure field).
>> 
> 
> My understanding was that he wanted me to move setting args.nr_done and
> args.preempted after calling construct_memop_from_reservation. But then
> here I missed args.domain. :-/
> 
> I will just move "args.domain = d" after the function call, if that's OK
> with you.

Yes, as said, either adjustment is fine with me.

But looking at this another time points out another issue with the
patch - you now clobber MEMF_populate_on_demand possibly
being set in args.memflags, i.e. you need to move down that part
too.

Jan


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