Re: [Qemu-devel] [PATCH v3 3/5] numa: reduce code duplication by adding helper numa_get_node_for_cpu()

[Igor Mammedov]

On Fri, 1 Jul 2016 14:30:12 +0200
Andrew Jones <address@hidden> wrote:

> On Fri, Jul 01, 2016 at 01:50:24PM +0200, Igor Mammedov wrote:
> > Replace repeated pattern
> > 
> >     for (i = 0; i < nb_numa_nodes; i++) {
> >         if (test_bit(idx, numa_info[i].node_cpu)) {
> >            ...
> >            break;
> > 
> > with a helper function to lookup numa node index for cpu.
> > 
> > Suggested-by: Michael S. Tsirkin <address@hidden>
> > Signed-off-by: Igor Mammedov <address@hidden>
> > ---
> >  hw/arm/virt-acpi-build.c |  6 ++----
> >  hw/arm/virt.c            |  7 +++----
> >  hw/i386/acpi-build.c     |  7 ++-----
> >  hw/i386/pc.c             |  8 +++-----
> >  hw/ppc/spapr_cpu_core.c  |  6 ++----
> >  include/sysemu/numa.h    |  3 +++
> >  numa.c                   | 12 ++++++++++++
> >  7 files changed, 27 insertions(+), 22 deletions(-)
> > 
> > diff --git a/hw/arm/virt-acpi-build.c b/hw/arm/virt-acpi-build.c
> > index 28fc59c..5923b3d 100644
> > --- a/hw/arm/virt-acpi-build.c
> > +++ b/hw/arm/virt-acpi-build.c
> > @@ -426,11 +426,9 @@ build_srat(GArray *table_data, BIOSLinker
> > *linker, VirtGuestInfo *guest_info) uint32_t *cpu_node =
> > g_malloc0(guest_info->smp_cpus * sizeof(uint32_t)); 
> >      for (i = 0; i < guest_info->smp_cpus; i++) {
> > -        for (j = 0; j < nb_numa_nodes; j++) {
> > -            if (test_bit(i, numa_info[j].node_cpu)) {
> > +        j = numa_get_node_for_cpu(i);
> > +        if (j < nb_numa_nodes) {
> >                  cpu_node[i] = j;
> 
> I think this, and all other occurrences, would read nicer like
> 
>  if (numa_enabled()) {
>      cpu_node[i] = numa_get_node_for_cpu(i);
>  }
it would be nicer but it could be a guest visible change as in case of
if cpu is not in numa_info[].node_cpu then old code won't do
assignment, if done as suggested it will assign whatever value
numa_get_node_for_cpu(i) returns.
  
> 
> We just need to also add
> 
>  bool numa_enabled()
>  {
>     return nb_numa_nodes != 0;
>  }
> 
> > -                break;
> > -            }
> >          }
> >      }
> >  
> > diff --git a/hw/arm/virt.c b/hw/arm/virt.c
> > index c5c125e..b066f15 100644
> > --- a/hw/arm/virt.c
> > +++ b/hw/arm/virt.c
> > @@ -411,10 +411,9 @@ static void fdt_add_cpu_nodes(const
> > VirtBoardInfo *vbi) armcpu->mp_affinity);
> >          }
> >  
> > -        for (i = 0; i < nb_numa_nodes; i++) {
> > -            if (test_bit(cpu, numa_info[i].node_cpu)) {
> > -                qemu_fdt_setprop_cell(vbi->fdt, nodename,
> > "numa-node-id", i);
> 
> We were missing the break here, which I had queued to send some day,
> now I don't need to :-)
> 
> > -            }
> > +        i = numa_get_node_for_cpu(cpu);
> > +        if (i < nb_numa_nodes) {
> > +            qemu_fdt_setprop_cell(vbi->fdt, nodename,
> > "numa-node-id", i); }
> >  
> >          g_free(nodename);
> > diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
> > index 5a594be..60be550 100644
> > --- a/hw/i386/acpi-build.c
> > +++ b/hw/i386/acpi-build.c
> > @@ -2344,18 +2344,15 @@ build_srat(GArray *table_data, BIOSLinker
> > *linker, MachineState *machine) srat->reserved1 = cpu_to_le32(1);
> >  
> >      for (i = 0; i < apic_ids->len; i++) {
> > -        int j;
> > +        int j = numa_get_node_for_cpu(i);
> >          int apic_id = apic_ids->cpus[i].arch_id;
> >  
> >          core = acpi_data_push(table_data, sizeof *core);
> >          core->type = ACPI_SRAT_PROCESSOR_APIC;
> >          core->length = sizeof(*core);
> >          core->local_apic_id = apic_id;
> > -        for (j = 0; j < nb_numa_nodes; j++) {
> > -            if (test_bit(i, numa_info[j].node_cpu)) {
> > +        if (j < nb_numa_nodes) {
> >                  core->proximity_lo = j;
> > -                break;
> > -            }
> >          }
> >          memset(core->proximity_hi, 0, 3);
> >          core->local_sapic_eid = 0;
> > diff --git a/hw/i386/pc.c b/hw/i386/pc.c
> > index 44a8f3b..fef34e7 100644
> > --- a/hw/i386/pc.c
> > +++ b/hw/i386/pc.c
> > @@ -780,11 +780,9 @@ static FWCfgState
> > *bochs_bios_init(AddressSpace *as, PCMachineState *pcms) for (i =
> > 0; i < max_cpus; i++) { unsigned int apic_id =
> > x86_cpu_apic_id_from_index(i); assert(apic_id <
> > pcms->apic_id_limit);
> > -        for (j = 0; j < nb_numa_nodes; j++) {
> > -            if (test_bit(i, numa_info[j].node_cpu)) {
> > -                numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
> > -                break;
> > -            }
> > +        j = numa_get_node_for_cpu(i);
> > +        if (j < nb_numa_nodes) {
> > +            numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
> >          }
> >      }
> >      for (i = 0; i < nb_numa_nodes; i++) {
> > diff --git a/hw/ppc/spapr_cpu_core.c b/hw/ppc/spapr_cpu_core.c
> > index 3a5da09..030016c 100644
> > --- a/hw/ppc/spapr_cpu_core.c
> > +++ b/hw/ppc/spapr_cpu_core.c
> > @@ -69,11 +69,9 @@ void spapr_cpu_init(sPAPRMachineState *spapr,
> > PowerPCCPU *cpu, Error **errp) }
> >  
> >      /* Set NUMA node for the added CPUs  */
> > -    for (i = 0; i < nb_numa_nodes; i++) {
> > -        if (test_bit(cs->cpu_index, numa_info[i].node_cpu)) {
> > +    i = numa_get_node_for_cpu(cs->cpu_index);
> > +    if (i < nb_numa_nodes) {
> >              cs->numa_node = i;
> > -            break;
> > -        }
> >      }
> >  
> >      xics_cpu_setup(spapr->icp, cpu);
> > diff --git a/include/sysemu/numa.h b/include/sysemu/numa.h
> > index bb184c9..4da808a 100644
> > --- a/include/sysemu/numa.h
> > +++ b/include/sysemu/numa.h
> > @@ -32,4 +32,7 @@ void numa_set_mem_node_id(ram_addr_t addr,
> > uint64_t size, uint32_t node); void
> > numa_unset_mem_node_id(ram_addr_t addr, uint64_t size, uint32_t
> > node); uint32_t numa_get_node(ram_addr_t addr, Error **errp); 
> > +/* on success returns node index in numa_info,
> > + * on failure returns nb_numa_nodes */
> 
> Or zero when nb_numa_nodes == 0, which requires another
> condition to disambiguate it from node=0. You could
> return -1 for both failure to find the node and for
> when nb_numa_nodes == 0, but no users are checking
> for error right now, only for nb_numa_nodes == 0, which,
> as I said above, would look better with numa_enabled()
> 
> > +int numa_get_node_for_cpu(int idx);
> >  #endif
> > diff --git a/numa.c b/numa.c
> > index 572712c..5790858 100644
> > --- a/numa.c
> > +++ b/numa.c
> > @@ -548,3 +548,15 @@ MemdevList *qmp_query_memdev(Error **errp)
> >      object_child_foreach(obj, query_memdev, &list);
> >      return list;
> >  }
> > +
> > +int numa_get_node_for_cpu(int idx)
> > +{
> > +    int i;
> > +
> > +    for (i = 0; i < nb_numa_nodes; i++) {
> > +        if (test_bit(idx, numa_info[i].node_cpu)) {
> > +            break;
> > +        }
> > +    }
> > +    return i;
> > +}
> > -- 
> > 1.8.3.1
> > 
> >
> 
> numa.c has one of these patterns too in numa_post_machine_init that
> could be replaced.
I didn't touch it intentionally as numa_post_machine_init is
broken code when combined with sparse cpu_index and I plan to remove it
in favor of explicit assignment (spapr does it, pc needs to be taught)
or may be event redo sole user of CPUState:numa_node (HMP command info
numa) and drop CPUState:numa_node field along with numa_post_machine_init()

but all that said it's topic for another not related patch. 

> 
> Thanks,
> drew