Re: [Qemu-devel] [PATCH v5] numa: make 'info numa' take into account hotplugged memory

[zhanghailiang <zhang.zhanghailiang@huawei.com>]

On 2014/10/16 20:50, Gonglei wrote:
> On 2014/10/16 20:02, zhanghailiang wrote:
>
>> When do memory hotplug, if there is numa node, we should add
>> the memory size to the corresponding node memory size.
>>
>> For now, it mainly affects the result of hmp command "info numa".
>>
>> Reviewed-by: Igor Mammedov <imammedo@redhat.com>
>> Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
>> ---
>>   v5:
>> - reword the subject (Igor Mammedov)
>> - turn query_numa_node_mem to void (Igor Mammedov)
>>   v4:
>> - s/pc_dimm_stat_node_mem/numa_stat_memory_devices/ (Igor Mammedov)
>> - rewrite numa_stat_memory_devices and this will also fix compile error for
>>    targets that don't support memory hotplug
>>   v3:
>> - cold-plugged memory should not be excluded (Igor Mammedov)
>>   v2:
>> - Don't modify the numa_info.node_mem directly when treating hotplug memory,
>>    fix the "info numa" instead (Igor Mammedov)
>>
>> Thanks for review!;)
>> ---
>>   include/sysemu/sysemu.h |  1 +
>>   monitor.c               |  6 +++++-
>>   numa.c                  | 41 +++++++++++++++++++++++++++++++++++++++++
>>   3 files changed, 47 insertions(+), 1 deletion(-)
>>
>> diff --git a/include/sysemu/sysemu.h b/include/sysemu/sysemu.h
>> index 0037a69..ef5eaf4 100644
>> --- a/include/sysemu/sysemu.h
>> +++ b/include/sysemu/sysemu.h
>> @@ -161,6 +161,7 @@ typedef struct node_info {
>>   extern NodeInfo numa_info[MAX_NODES];
>>   void set_numa_nodes(void);
>>   void set_numa_modes(void);
>> +void query_numa_node_mem(uint64_t *node_mem);
>>   extern QemuOptsList qemu_numa_opts;
>>   int numa_init_func(QemuOpts *opts, void *opaque);
>>
>> diff --git a/monitor.c b/monitor.c
>> index 2d14f39..d45b0a3 100644
>> --- a/monitor.c
>> +++ b/monitor.c
>> @@ -1949,7 +1949,10 @@ static void do_info_numa(Monitor *mon, const QDict *qdict)
>>   {
>>       int i;
>>       CPUState *cpu;
>> +    uint64_t *node_mem;
>>
>> +    node_mem = g_new0(uint64_t, nb_numa_nodes);
>> +    query_numa_node_mem(node_mem);
>>       monitor_printf(mon, "%d nodes\n", nb_numa_nodes);
>>       for (i = 0; i < nb_numa_nodes; i++) {
>>           monitor_printf(mon, "node %d cpus:", i);
>> @@ -1960,8 +1963,9 @@ static void do_info_numa(Monitor *mon, const QDict *qdict)
>>           }
>>           monitor_printf(mon, "\n");
>>           monitor_printf(mon, "node %d size: %" PRId64 " MB\n", i,
>> -            numa_info[i].node_mem >> 20);
>> +                       node_mem[i] >> 20);
>>       }
>> +    g_free(node_mem);
>>   }
>>
>>   #ifdef CONFIG_PROFILER
>> diff --git a/numa.c b/numa.c
>> index 3b98135..5b84ffb 100644
>> --- a/numa.c
>> +++ b/numa.c
>> @@ -35,6 +35,7 @@
>>   #include "hw/boards.h"
>>   #include "sysemu/hostmem.h"
>>   #include "qmp-commands.h"
>> +#include "hw/mem/pc-dimm.h"
>>
>>   QemuOptsList qemu_numa_opts = {
>>       .name = "numa",
>> @@ -315,6 +316,46 @@ void memory_region_allocate_system_memory(MemoryRegion *mr, Object *owner,
>>       }
>>   }
>>
>> +static void numa_stat_memory_devices(uint64_t *node_mem)
>> +{
>> +    MemoryDeviceInfoList *info_list = NULL;
>> +    MemoryDeviceInfoList **prev = &info_list;
>> +    MemoryDeviceInfoList *info;
>> +
>> +    qmp_pc_dimm_device_list(qdev_get_machine(), &prev);
>> +    for (info = info_list; info; info = info->next) {
>> +        MemoryDeviceInfo *value = info->value;
>> +
>> +        if (value) {
>> +            switch (value->kind) {
>> +            case MEMORY_DEVICE_INFO_KIND_DIMM: {
>
> I think this '{}' symbol is superfluous.
>

No, without it, compiler will complain 'error: a label can only be part of
statement and a declaration is not a statement'.

So it is necessary for defining a variable inside switch.
Actually, you will see lots of such statements in kernel source;)

> Best regards,
> -Gonglei
>
>
>
> .
>