Re: [PATCH RFC v1 57/74] x86/pv-shim: shadow PV console's page for L2 DomU

[Sergey Dyasli <sergey.dyasli@citrix.com>]

On Tue, 2018-01-09 at 02:13 -0700, Jan Beulich wrote:
> > > > On 04.01.18 at 14:06, <wei.liu2@citrix.com> wrote:
> > +size_t consoled_guest_rx(void)
> > +{
> > +    size_t recv = 0, idx = 0;
> > +    XENCONS_RING_IDX cons, prod;
> > +
> > +    if ( !cons_ring )
> > +        return 0;
> > +
> > +    spin_lock(&rx_lock);
> > +
> > +    cons = cons_ring->out_cons;
> > +    prod = ACCESS_ONCE(cons_ring->out_prod);
> > +    ASSERT((prod - cons) <= sizeof(cons_ring->out));
> > +
> > +    /* Is the ring empty? */
> > +    if ( cons == prod )
> > +        goto out;
> > +
> > +    /* Update pointers before accessing the ring */
> > +    smp_rmb();
> 
> I think this need to move up ahead of the if(). In the comment
> perhaps s/Update/Latch/?

The read/write memory barriers here are between read/write accesses to
ring->out_prod and ring->out array. So there is no need to move them.
(the same goes for the input ring)

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