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

[Andrew Jones]

----- 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
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
> 
> 
>