On Fri, 26 May 2017 12:46:25 -0300
Eduardo Habkost <ehabk...@redhat.com> wrote:
> On Wed, May 10, 2017 at 01:29:50PM +0200, Igor Mammedov wrote:
> [...]
> > diff --git a/hw/core/machine.c b/hw/core/machine.c
> > index 2482c63..420c8c4 100644
> > --- a/hw/core/machine.c
> > +++ b/hw/core/machine.c
> > @@ -389,6 +389,102 @@ HotpluggableCPUList
> > *machine_query_hotpluggable_cpus(MachineState *machine)
> [...]
> > +void machine_set_cpu_numa_node(MachineState *machine,
> > + const CpuInstanceProperties *props, Error
> > **errp)
> > +{
> [...]
> > + /* force board to initialize possible_cpus if it hasn't been done yet
> > */
> > + mc->possible_cpu_arch_ids(machine);
> [...]
> > diff --git a/numa.c b/numa.c
> > index 7182481..7db5dde 100644
> > --- a/numa.c
> > +++ b/numa.c
> > @@ -170,6 +170,7 @@ static void parse_numa_node(MachineState *ms,
> > NumaNodeOptions *node,
> > exit(1);
> > }
> > for (cpus = node->cpus; cpus; cpus = cpus->next) {
> > + CpuInstanceProperties props;
> > if (cpus->value >= max_cpus) {
> > error_setg(errp,
> > "CPU index (%" PRIu16 ")"
> > @@ -178,6 +179,10 @@ static void parse_numa_node(MachineState *ms,
> > NumaNodeOptions *node,
> > return;
> > }
> > bitmap_set(numa_info[nodenr].node_cpu, cpus->value, 1);
> > + props = mc->cpu_index_to_instance_props(ms, cpus->value);
> > + props.node_id = nodenr;
> > + props.has_node_id = true;
> > + machine_set_cpu_numa_node(ms, &props, &error_fatal);
>
> This triggers a call to possible_cpu_arch_ids() before
> nb_numa_nodes is set to the actual number of NUMA nodes in the
> machine, breaking the "node_id = ... % nb_numa_nodes"
> initialization logic in pc, virt, and spapr.
Looking at it more, in current code this problem theoretically
affects only partial mapping case. And even there it's masked by
fixup to 0 node for non explicitly mapped CPUs and the bug which
*) "[PATCH v2 3/5] numa: make sure that all cpus in has has_node_id set if
numa is enabled"
fixes. So invalid default values for non mapped cpus are stored
in possible_cpus but not used nor visible elsewhere.
[*] as side effect fixes invalid defaults in possible_cpus as well
by fixup to 0 node that's moved from pre_plug handler.
In future we are going to remove partial mapping with fixup to 0
and fetch default mappings from possible_cpus where these wrong
defaults won't exists (or matter) as CPUs are going to either
1) be all explicitly mapped (so all node_id values are overwritten)
or
2) all default mapped (where defaults will be correct since no
-numa cpu=/-numa node,cpus= were called), but that's fragile
so I'd add calculate defaults at machine_run_board_init()
time anyway.
