Re: [Qemu-devel] [PATCH v3] Allow setting NUMA distance for different NUMA nodes

[Igor Mammedov]

On Wed, 22 Mar 2017 17:32:46 +0800
He Chen <address@hidden> wrote:

> Current, QEMU does not provide a clear command to set vNUMA distance for
> guest although we already have `-numa` command to set vNUMA nodes.
> 
> vNUMA distance makes sense in certain scenario.
> But now, if we create a guest that has 4 vNUMA nodes, when we check NUMA
> info via `numactl -H`, we will see:
> 
> node distance:
> node    0    1    2    3
>   0:   10   20   20   20
>   1:   20   10   20   20
>   2:   20   20   10   20
>   3:   20   20   20   10
> 
> Guest kernel regards all local node as distance 10, and all remote node
> as distance 20 when there is no SLIT table since QEMU doesn't build it.
> It looks like a little strange when you have seen the distance in an
> actual physical machine that contains 4 NUMA nodes. My machine shows:
> 
> node distance:
> node    0    1    2    3
>   0:   10   21   31   41
>   1:   21   10   21   31
>   2:   31   21   10   21
>   3:   41   31   21   10
> 
> To set vNUMA distance, guest should see a complete SLIT table.
> I found QEMU has provide `-acpitable` command that allows users to add
> a ACPI table into guest, but it requires users building ACPI table by
> themselves first. Using `-acpitable` to add a SLIT table may be not so
> straightforward or flexible, imagine that when the vNUMA configuration
> is changes and we need to generate another SLIT table manually. It may
> not be friendly to users or upper software like libvirt.
> 
> This patch is going to add SLIT table support in QEMU, and provides
> additional option `dist` for command `-numa` to allow user set vNUMA
> distance by QEMU command.
> 
> With this patch, when a user wants to create a guest that contains
> several vNUMA nodes and also wants to set distance among those nodes,
> the QEMU command would like:
> 
> ```
> -object 
> memory-backend-ram,size=1G,prealloc=yes,host-nodes=0,policy=bind,id=node0 \
> -numa node,nodeid=0,cpus=0,memdev=node0 \
> -object 
> memory-backend-ram,size=1G,prealloc=yes,host-nodes=1,policy=bind,id=node1 \
> -numa node,nodeid=1,cpus=1,memdev=node1 \
> -object 
> memory-backend-ram,size=1G,prealloc=yes,host-nodes=2,policy=bind,id=node2 \
> -numa node,nodeid=2,cpus=2,memdev=node2 \
> -object 
> memory-backend-ram,size=1G,prealloc=yes,host-nodes=3,policy=bind,id=node3 \
> -numa node,nodeid=3,cpus=3,memdev=node3 \
> -numa dist,src=0,dst=1,val=21 \
> -numa dist,src=0,dst=2,val=31 \
> -numa dist,src=0,dst=3,val=41 \
> -numa dist,src=1,dst=0,val=21 \
> ...
> ```
> 
> Signed-off-by: He Chen <address@hidden>
> 
> fix
> ---
>  hw/acpi/aml-build.c         | 26 +++++++++++++++++++++++++
>  hw/arm/virt-acpi-build.c    |  2 ++
>  hw/i386/acpi-build.c        |  2 ++
>  include/hw/acpi/aml-build.h |  1 +
>  include/sysemu/numa.h       |  1 +
>  include/sysemu/sysemu.h     |  4 ++++
>  numa.c                      | 47 
> +++++++++++++++++++++++++++++++++++++++++++++
>  qapi-schema.json            | 30 ++++++++++++++++++++++++++---
>  qemu-options.hx             | 12 +++++++++++-
>  9 files changed, 121 insertions(+), 4 deletions(-)
> 
> diff --git a/hw/acpi/aml-build.c b/hw/acpi/aml-build.c
> index c6f2032..410b30e 100644
> --- a/hw/acpi/aml-build.c
> +++ b/hw/acpi/aml-build.c
> @@ -24,6 +24,7 @@
>  #include "hw/acpi/aml-build.h"
>  #include "qemu/bswap.h"
>  #include "qemu/bitops.h"
> +#include "sysemu/numa.h"
>  
>  static GArray *build_alloc_array(void)
>  {
> @@ -1609,3 +1610,28 @@ void build_srat_memory(AcpiSratMemoryAffinity 
> *numamem, uint64_t base,
>      numamem->base_addr = cpu_to_le64(base);
>      numamem->range_length = cpu_to_le64(len);
>  }
> +
> +/*
> + * ACPI spec 5.2.17 System Locality Distance Information Table
> + * (Revision 2.0 or later)
> + */
> +void build_slit(GArray *table_data, BIOSLinker *linker)
> +{
> +    int slit_start, i, j;
> +    slit_start = table_data->len;
> +
> +    acpi_data_push(table_data, sizeof(AcpiTableHeader));
> +
> +    build_append_int_noprefix(table_data, nb_numa_nodes, 8);
> +    for (i = 0; i < nb_numa_nodes; i++) {
> +        for (j = 0; j < nb_numa_nodes; j++) {
> +            build_append_int_noprefix(table_data, numa_info[i].distance[j], 
> 1);
> +        }
> +    }
> +
> +    build_header(linker, table_data,
> +                 (void *)(table_data->data + slit_start),
> +                 "SLIT",
> +                 table_data->len - slit_start, 1, NULL, NULL);
> +}
> +
> diff --git a/hw/arm/virt-acpi-build.c b/hw/arm/virt-acpi-build.c
> index 0835e59..d9e6828 100644
> --- a/hw/arm/virt-acpi-build.c
> +++ b/hw/arm/virt-acpi-build.c
> @@ -781,6 +781,8 @@ void virt_acpi_build(VirtMachineState *vms, 
> AcpiBuildTables *tables)
>      if (nb_numa_nodes > 0) {
>          acpi_add_table(table_offsets, tables_blob);
>          build_srat(tables_blob, tables->linker, vms);
> +        acpi_add_table(table_offsets, tables_blob);
> +        build_slit(tables_blob, tables->linker);
>      }
>  
>      if (its_class_name() && !vmc->no_its) {
> diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
> index 2073108..12730ea 100644
> --- a/hw/i386/acpi-build.c
> +++ b/hw/i386/acpi-build.c
> @@ -2678,6 +2678,8 @@ void acpi_build(AcpiBuildTables *tables, MachineState 
> *machine)
>      if (pcms->numa_nodes) {
>          acpi_add_table(table_offsets, tables_blob);
>          build_srat(tables_blob, tables->linker, machine);
> +        acpi_add_table(table_offsets, tables_blob);
> +        build_slit(tables_blob, tables->linker);
>      }
>      if (acpi_get_mcfg(&mcfg)) {
>          acpi_add_table(table_offsets, tables_blob);
> diff --git a/include/hw/acpi/aml-build.h b/include/hw/acpi/aml-build.h
> index 00c21f1..329a0d0 100644
> --- a/include/hw/acpi/aml-build.h
> +++ b/include/hw/acpi/aml-build.h
> @@ -389,4 +389,5 @@ GCC_FMT_ATTR(2, 3);
>  void build_srat_memory(AcpiSratMemoryAffinity *numamem, uint64_t base,
>                         uint64_t len, int node, MemoryAffinityFlags flags);
>  
> +void build_slit(GArray *table_data, BIOSLinker *linker);
>  #endif
> diff --git a/include/sysemu/numa.h b/include/sysemu/numa.h
> index 8f09dcf..2f7a941 100644
> --- a/include/sysemu/numa.h
> +++ b/include/sysemu/numa.h
> @@ -21,6 +21,7 @@ typedef struct node_info {
>      struct HostMemoryBackend *node_memdev;
>      bool present;
>      QLIST_HEAD(, numa_addr_range) addr; /* List to store address ranges */
> +    uint8_t distance[MAX_NODES];
>  } NodeInfo;
>  
>  extern NodeInfo numa_info[MAX_NODES];
> diff --git a/include/sysemu/sysemu.h b/include/sysemu/sysemu.h
> index 576c7ce..a4e328d 100644
> --- a/include/sysemu/sysemu.h
> +++ b/include/sysemu/sysemu.h
> @@ -169,6 +169,10 @@ extern int mem_prealloc;
>  
>  #define MAX_NODES 128
>  #define NUMA_NODE_UNASSIGNED MAX_NODES
> +#define MIN_NUMA_DISTANCE 10
> +#define DEF_NUMA_DISTANCE 20
> +#define MAX_NUMA_DISTANCE 254
> +#define NUMA_DISTANCE_UNREACHABLE 255
>  
>  #define MAX_OPTION_ROMS 16
>  typedef struct QEMUOptionRom {
> diff --git a/numa.c b/numa.c
> index e01cb54..425a320 100644
> --- a/numa.c
> +++ b/numa.c
> @@ -212,6 +212,28 @@ static void numa_node_parse(NumaNodeOptions *node, 
> QemuOpts *opts, Error **errp)
>      max_numa_nodeid = MAX(max_numa_nodeid, nodenr + 1);
>  }
>  
> +static void numa_distance_parse(NumaDistOptions *dist, QemuOpts *opts, Error 
> **errp)
> +{
> +    uint64_t src = dist->src;
> +    uint64_t dst = dist->dst;
> +    uint8_t val = dist->val;
> +
> +    if (src >= MAX_NODES || dst >= MAX_NODES) {
this allows to specify any src/dst instead of present nodes only,
you should probably check that src/dst are present here,
see: numa_info[x].present

if node is not present complain that it should be declared with
'-numa node' option first and exit.

> +        error_setg(errp, "Max number of NUMA nodes reached: %"
> +                   PRIu64 "", src > dst ? src : dst);
> +        return;
> +    }
> +
> +    if (val < MIN_NUMA_DISTANCE || val > MAX_NUMA_DISTANCE) {
> +        error_setg(errp,
> +                "NUMA distance (%" PRIu8 ") out of range (%d) ~ (%d)",
> +                dist->val, MIN_NUMA_DISTANCE, MAX_NUMA_DISTANCE);
> +        return;
> +    }
> +
> +    numa_info[src].distance[dst] = val;
> +}
> +
>  static int parse_numa(void *opaque, QemuOpts *opts, Error **errp)
>  {
>      NumaOptions *object = NULL;
> @@ -235,6 +257,12 @@ static int parse_numa(void *opaque, QemuOpts *opts, 
> Error **errp)
>          }
>          nb_numa_nodes++;
>          break;
> +    case NUMA_OPTIONS_TYPE_DIST:
> +        numa_distance_parse(&object->u.dist, opts, &err);
> +        if (err) {
> +            goto end;
> +        }
> +        break;
>      default:
>          abort();
>      }
> @@ -294,6 +322,24 @@ static void validate_numa_cpus(void)
>      g_free(seen_cpus);
>  }
>  
> +static void default_numa_distance(void)
> +{
> +    int src, dst;
> +
fill in defaults only if there weren't any '-numa dist' on CLI
and refuse to start if partial filled table were explicitly provided on CLI

> +    for (src = 0; src < nb_numa_nodes; src++) {
> +        for (dst = 0; dst < nb_numa_nodes; dst++) {
> +            if (src == dst && numa_info[src].distance[dst] != 
> MIN_NUMA_DISTANCE) {
> +                numa_info[src].distance[dst] = MIN_NUMA_DISTANCE;
> +            } else if (numa_info[src].distance[dst] <= MIN_NUMA_DISTANCE) {
> +                if (numa_info[dst].distance[src] > MIN_NUMA_DISTANCE)
> +                    numa_info[src].distance[dst] = 
> numa_info[dst].distance[src];
> +                else
> +                    numa_info[src].distance[dst] = DEF_NUMA_DISTANCE;
> +            }
> +        }
> +    }
> +}
> +
>  void parse_numa_opts(MachineClass *mc)
>  {
>      int i;
> @@ -390,6 +436,7 @@ void parse_numa_opts(MachineClass *mc)
>          }
>  
>          validate_numa_cpus();
> +        default_numa_distance();
>      } else {
>          numa_set_mem_node_id(0, ram_size, 0);
>      }
> diff --git a/qapi-schema.json b/qapi-schema.json
> index 32b4a4b..21ad94a 100644
> --- a/qapi-schema.json
> +++ b/qapi-schema.json
> @@ -5644,15 +5644,19 @@
>  ##
>  # @NumaOptionsType:
>  #
> +# @node: NUMA nodes configuration
> +#
> +# @dist: NUMA distance configuration
> +#
>  # Since: 2.1
>  ##
>  { 'enum': 'NumaOptionsType',
> -  'data': [ 'node' ] }
> +  'data': [ 'node', 'dist' ] }
>  
>  ##
>  # @NumaOptions:
>  #
> -# A discriminated record of NUMA options. (for OptsVisitor)
> +# A discriminated record of NUMA options.
unrelated change?

>  #
>  # Since: 2.1
>  ##
> @@ -5660,7 +5664,8 @@
>    'base': { 'type': 'NumaOptionsType' },
>    'discriminator': 'type',
>    'data': {
> -    'node': 'NumaNodeOptions' }}
> +    'node': 'NumaNodeOptions',
> +    'dist': 'NumaDistOptions' }}
>  
>  ##
>  # @NumaNodeOptions:
> @@ -5689,6 +5694,25 @@
>     '*memdev': 'str' }}
>  
>  ##
> +# @NumaDistOptions:
> +#
> +# Set distance between 2 NUMA nodes. (for OptsVisitor)
> +#
> +# @src: source NUMA node.
> +#
> +# @dst: destination NUMA node.
> +#
> +# @val: NUMA distance from source node to destination node.
> +#
> +# Since: 2.10
> +##
> +{ 'struct': 'NumaDistOptions',
> +  'data': {
> +   'src': 'uint64',
> +   'dst': 'uint64',
nodeid is uint16, so probably these should be also uint16


> +   'val': 'uint8' }}
> +
> +##
>  # @HostMemPolicy:
>  #
>  # Host memory policy types
> diff --git a/qemu-options.hx b/qemu-options.hx
> index 8dd8ee3..43c3950 100644
> --- a/qemu-options.hx
> +++ b/qemu-options.hx
> @@ -139,12 +139,15 @@ ETEXI
>  
>  DEF("numa", HAS_ARG, QEMU_OPTION_numa,
>      "-numa node[,mem=size][,cpus=firstcpu[-lastcpu]][,nodeid=node]\n"
> -    "-numa node[,memdev=id][,cpus=firstcpu[-lastcpu]][,nodeid=node]\n", 
> QEMU_ARCH_ALL)
> +    "-numa node[,memdev=id][,cpus=firstcpu[-lastcpu]][,nodeid=node]\n"
> +    "-numa dist,src=source,dst=destination,val=distance\n", QEMU_ARCH_ALL)
>  STEXI
>  @item -numa 
> node[,address@hidden,address@hidden@var{lastcpu}]][,address@hidden
>  @itemx -numa 
> node[,address@hidden,address@hidden@var{lastcpu}]][,address@hidden
> address@hidden -numa dist,address@hidden,address@hidden,address@hidden
>  @findex -numa
>  Define a NUMA node and assign RAM and VCPUs to it.
> +Set NUMA distance from source node to destination node.
>  
>  @var{firstcpu} and @var{lastcpu} are CPU indexes. Each
>  @samp{cpus} option represent a contiguous range of CPU indexes
> @@ -167,6 +170,13 @@ split equally between them.
>  @samp{mem} and @samp{memdev} are mutually exclusive. Furthermore,
>  if one node uses @samp{memdev}, all of them have to use it.
>  
> address@hidden and @var{destination} are NUMA node IDs.
> address@hidden is the NUMA distance from @var{source} to @var{destination}.
> +The distance from node A to node B may be different from the distance from
> +node B to node A since the distance can to be asymmetry.
> +If the distance is not set, the default distance for a local NUMA node is 10,
> +and 20 for a remote node.
> +
>  Note that the address@hidden option doesn't allocate any of the
>  specified resources, it just assigns existing resources to NUMA
>  nodes. This means that one still has to use the @option{-m},