Re: [Qemu-devel] [PATCH V9 07/12] NUMA: parse guest numa nodes memory policy

[Wanlong Gao]

On 08/23/2013 10:11 PM, Andrew Jones wrote:
> 
> 
> ----- Original Message -----
>> The memory policy setting format is like:
>>     
>> policy={default|membind|interleave|preferred}[,relative=true],host-nodes=N-N
>> And we are adding this setting as a suboption of "-numa mem,",
>> the memory policy then can be set like following:
>>     -numa node,nodeid=0,cpus=0 \
>>     -numa node,nodeid=1,cpus=1 \
>>     -numa mem,nodeid=0,size=1G,policy=membind,host-nodes=0-1 \
>>     -numa mem,nodeid=1,size=1G,policy=interleave,relative=true,host-nodes=1
>>
>> Signed-off-by: Wanlong Gao <address@hidden>
>> ---
>>  include/sysemu/sysemu.h |  3 +++
>>  numa.c                  | 13 +++++++++++++
>>  qapi-schema.json        | 31 +++++++++++++++++++++++++++++--
>>  vl.c                    |  3 +++
>>  4 files changed, 48 insertions(+), 2 deletions(-)
>>
>> diff --git a/include/sysemu/sysemu.h b/include/sysemu/sysemu.h
>> index b683d08..81d16a5 100644
>> --- a/include/sysemu/sysemu.h
>> +++ b/include/sysemu/sysemu.h
>> @@ -134,6 +134,9 @@ extern int nb_numa_mem_nodes;
>>  typedef struct node_info {
>>      uint64_t node_mem;
>>      DECLARE_BITMAP(node_cpu, MAX_CPUMASK_BITS);
>> +    DECLARE_BITMAP(host_mem, MAX_CPUMASK_BITS);
>> +    NumaNodePolicy policy;
>> +    bool relative;
>>  } NodeInfo;
>>  extern NodeInfo numa_info[MAX_NODES];
>>  extern QemuOptsList qemu_numa_opts;
>> diff --git a/numa.c b/numa.c
>> index 3e2dfc1..4ccc6cb 100644
>> --- a/numa.c
>> +++ b/numa.c
>> @@ -74,6 +74,7 @@ static int numa_mem_parse(NumaMemOptions *opts)
>>  {
>>      uint16_t nodenr;
>>      uint64_t mem_size;
>> +    uint16List *nodes;
>>  
>>      if (opts->has_nodeid) {
>>          nodenr = opts->nodeid;
>> @@ -91,6 +92,18 @@ static int numa_mem_parse(NumaMemOptions *opts)
>>          numa_info[nodenr].node_mem = mem_size;
>>      }
>>  
>> +    if (opts->has_policy) {
>> +        numa_info[nodenr].policy = opts->policy;
>> +    }
>> +
>> +    if (opts->has_relative) {
>> +        numa_info[nodenr].relative = opts->relative;
>> +    }
>> +
>> +    for (nodes = opts->host_nodes; nodes; nodes = nodes->next) {
>> +        bitmap_set(numa_info[nodenr].host_mem, nodes->value, 1);
>> +    }
>> +
>>      return 0;
>>  }
>>  
>> diff --git a/qapi-schema.json b/qapi-schema.json
>> index 11851a1..650741f 100644
>> --- a/qapi-schema.json
>> +++ b/qapi-schema.json
>> @@ -3806,6 +3806,24 @@
>>     '*mem':    'str' }}
>>  
>>  ##
>> +# @NumaNodePolicy
>> +#
>> +# NUMA node policy types
>> +#
>> +# @default: restore default policy, remove any nondefault policy
>> +#
>> +# @membind: a strict policy that restricts memory allocation to the
>> +#           nodes specified
>> +#
>> +# @interleave: the page allocations is interleaved across the set
>> +#              of nodes specified
>> +#
>> +# @preferred: set the preferred node for allocation
>> +##
>> +{ 'enum': 'NumaNodePolicy',
>> +  'data': [ 'default', 'membind', 'interleave', 'preferred' ] }
>> +
>> +##
>>  # @NumaMemOptions
>>  #
>>  # Set memory information of guest NUMA node. (for OptsVisitor)
>> @@ -3814,9 +3832,18 @@
>>  #
>>  # @size: #optional memory size of this node
>>  #
>> +# @policy: #optional memory policy of this node
>> +#
>> +# @relative: #optional if the nodes specified are relative
>> +#
>> +# @host-nodes: #optional host nodes for its memory policy
>> +#
>>  # Since 1.7
>>  ##
>>  { 'type': 'NumaMemOptions',
>>    'data': {
>> -   '*nodeid': 'uint16',
>> -   '*size':   'size' }}
>> +   '*nodeid':     'uint16',
>> +   '*size':       'size',
>> +   '*policy':     'NumaNodePolicy',
>> +   '*relative':   'bool',
>> +   '*host-nodes': ['uint16'] }}
>> diff --git a/vl.c b/vl.c
>> index 2377b67..91b0d76 100644
>> --- a/vl.c
>> +++ b/vl.c
>> @@ -2888,6 +2888,9 @@ int main(int argc, char **argv, char **envp)
>>      for (i = 0; i < MAX_NODES; i++) {
>>          numa_info[i].node_mem = 0;
>>          bitmap_zero(numa_info[i].node_cpu, MAX_CPUMASK_BITS);
>> +        bitmap_zero(numa_info[i].host_mem, MAX_CPUMASK_BITS);
> 
> Shouldn't the bitmap size of host_mem be MAX_NODES? If so, and you

MAX_NODES is for guest numa nodes number, but this bitmap is for host
numa nodes. AFAIK, this MAX_NODES is not big enough for host nodes number,
the default host kernel NODES_SHIFT is 9.

Thanks,
Wanlong Gao

> change it, then make sure the find_last_bit() call also gets
> updated in patch 8/12, and anywhere else needed.
> 
> drew
> 
>> +        numa_info[i].policy = NUMA_NODE_POLICY_DEFAULT;
>> +        numa_info[i].relative = false;
>>      }
>>  
>>      nb_numa_nodes = 0;
>> --
>> 1.8.4.rc4
>>
>>
>>
>