Re: [Qemu-devel] [PATCH 0/3] cpu: numa: Fix the mapping initialization of VCPUs and NUMA nodes

[Eduardo Habkost]

On Thu, Jan 19, 2017 at 08:17:02PM +0800, Dou Liyang wrote:
> Hi, Eduardo
> 
> At 01/19/2017 01:06 AM, Eduardo Habkost wrote:
> > On Wed, Jan 18, 2017 at 09:26:36PM +0800, Dou Liyang wrote:
> > > Hi, All
> > > 
> > > > **
> > > > ERROR:/tmp/qemu-test/src/tests/vhost-user-test.c:668:test_migrate: 
> > > > assertion failed: (qdict_haskey(rsp, "return"))
> > > > GTester: last random seed: R02Sf52546c4daff8087416f43fa7c146db8
> > > > ftruncate: Permission denied
> > > > ftruncate: Permission denied
> > > > qemu-system-aarch64: /tmp/qemu-test/src/qom/cpu.c:346: 
> > > > cpu_common_map_numa_node: Assertion `cpu->cpu_index < max_cpus' failed.
> > > > Broken pipe
> > > 
> > > I don't know What's the meaning of this log ?
> > > 
> > > Is the qemu-system-aarch64 can't recognize the
> > > qom/cpu.c:346: assert(cpu->cpu_index < max_cpus);
> > 
> > This means the assert() line is being triggered for some reason,
> > and cpu_index is >= max_cpus when we cpu_common_map_numa_node()
> > gets called. We need to investigate why.
> > 
> 
> I have investigated the reason why it is failed.
> 
> Because not all targets(aarch64-linux-user, x86_64-linux-user ...)
> need to compile the vl.c(include the max_cpus) and numa.c, but the all
> may compile the ./qom/cpu.c.
> So, when we Link those targets, we may can't find the vl.o or numa.o
> that we want.
> 
> Add "#ifdef CONFIG_NUMA" to fix it.
> 
> +static void cpu_common_map_numa_node(CPUState *cpu)
> +{
> +    #ifdef CONFIG_NUMA
> +    int i;
> +
> +    assert(cpu->cpu_index < max_cpus);
> +    for (i = 0; i < nb_numa_nodes; i++) {
> +        if (test_bit(cpu->cpu_index, numa_info[i].node_cpu)) {
> +            cpu->numa_node = i;
> +            return;
> +        }
> +    }
> +    #endif
> +}
> +
> 
> And I am not sure if it is necessary to resend this patch for
> fixing the bug before Igor's patches is OK completely?

No need to resend this patch: I have already applied your
previous patch that moved the numa_post_machine_init() call, and
Igor is taking care of removing numa_post_machine_init() in a
different way.

Thanks!

-- 
Eduardo