On 01.07.2014 [12:24:48 -0300], Eduardo Habkost wrote:
> On Mon, Jun 30, 2014 at 02:34:56PM -0700, Nishanth Aravamudan wrote:
> [...]
> > }
> > + assert(numa_info[i].present);
> > numa_info[i].node_mem = ram_size - usedmem;
> > }
>
> Tab here.
>
> >
> [.
On Mon, Jun 30, 2014 at 02:34:56PM -0700, Nishanth Aravamudan wrote:
[...]
> }
> + assert(numa_info[i].present);
> numa_info[i].node_mem = ram_size - usedmem;
> }
Tab here.
>
[...]
> +for (i = 0, j = 0; i < max_cpus; i++) {
> +
Sparse node numbering occurs on powerpc in practice under PowerVM. In
order to emulate the same NUMA topology under qemu, the assumption that
NUMA nodes are linearly ordered has to be removed. qemu was recently
modified to reject requests for sparse NUMA node numbering.
Leverage the present field
