Re: [Qemu-devel] [PATCH V16 09/11] NUMA: set guest numa nodes memory policy

[Andrew Jones <drjones@redhat.com>]

On Wed, Nov 27, 2013 at 03:35:53PM +0100, Paolo Bonzini wrote:
> > +
> > +    len = numa_info[nodeid].node_mem;
> > +    bind_mode = node_parse_bind_mode(nodeid);
> > +    unsigned long *nodes = numa_info[nodeid].host_mem;
> > +
> > +    /* This is a workaround for a long standing bug in Linux'
> > +     * mbind implementation, which cuts off the last specified
> > +     * node. To stay compatible should this bug be fixed, we
> > +     * specify one more node and zero this one out.
> > +     */
> > +    unsigned long maxnode = find_last_bit(nodes, MAX_NODES);
> > +    if (syscall(SYS_mbind, ram_ptr + ram_offset, len, bind_mode,
> > +                nodes, maxnode + 2, 0)) {
> > +            perror("mbind");
> > +            return -1;
> > +    }
> 
> Also, it's still not clear to me why we're not using libnuma.
>

There only seem to be two reasons to use it right now; 1)
it provides a wrapper around mbind 2) it exists. IMHO libnuma
needs some work before it would be suitable for qemu to marry
it. Its interfaces don't buy a lot of elegance, it only reads
system information on link time (i.e. no good if you want to
consider hotplug of cpus and nodes), and it's currently linux
specific anyway.  It's a good idea, but is it worth yet another
qemu build dependency?

drew