[Qemu-devel] [PATCH 7/8] v2: add numa monitor command

[Qemu-devel] [PATCH 7/8] v2: add numa monitor command

[Andre Przywara]

[-- Attachment #1: Type: text/plain, Size: 491 bytes --]

Signed-off-by: Andre Przywara <andre.przywara@amd.com>

-- 
Andre Przywara
AMD-Operating System Research Center (OSRC), Dresden, Germany
Tel: +49 351 277-84917
----to satisfy European Law for business letters:
AMD Saxony Limited Liability Company & Co. KG,
Wilschdorfer Landstr. 101, 01109 Dresden, Germany
Register Court Dresden: HRA 4896, General Partner authorized
to represent: AMD Saxony LLC (Wilmington, Delaware, US)
General Manager of AMD Saxony LLC: Dr. Hans-R. Deppe, Thomas McCoy

[-- Attachment #2: qemunuma_v2_monitor_numa.patch --]
[-- Type: text/x-patch, Size: 2709 bytes --]

# HG changeset patch
# User Andre Przywara <andre.przywara@amd.com>
# Date 1229435568 -3600
# Node ID 4f0a8ac2d88ffffc1dcd82785c1620553baa86da
# Parent  5a74bd76931b79713803795b3943aa8383946521
add -numa monitor command to repin guest nodes

diff -r 5a74bd76931b -r 4f0a8ac2d88f monitor.c
--- a/monitor.c	Tue Dec 16 14:51:24 2008 +0100
+++ b/monitor.c	Tue Dec 16 14:52:48 2008 +0100
@@ -39,6 +39,9 @@
 #include "qemu-timer.h"
 #include "migration.h"
 #include "kvm.h"
+#ifdef CONFIG_NUMA
+#include <numa.h>
+#endif
 
 //#define DEBUG
 //#define DEBUG_COMPLETION
@@ -1285,6 +1288,9 @@ static void do_info_numa(void)
 {
     int i, j;
 
+#ifndef CONFIG_NUMA
+    term_printf("NUMA placement support: not compiled\n");
+#endif
     term_printf("%d nodes\n", numnumanodes);
     for (i = 0; i < numnumanodes; i++) {
         term_printf("node %d cpus:", i);
@@ -1292,7 +1298,40 @@ static void do_info_numa(void)
             if (node_to_cpus[i] & (1ULL << j)) term_printf(" %d", j);
         term_printf("\n");
         term_printf("node %d size: %" PRId64 " MB\n", i, node_mem[i] >> 20);
-    }
+        term_printf("node %d host: ", i);
+        if (hostnodes[i] == (uint64_t)-1) term_printf("*\n"); else
+            term_printf("%" PRId64 "\n", hostnodes[i]);
+    }
+}
+
+static void do_numa(const char *affinity)
+{
+#ifdef CONFIG_NUMA
+    uint64_t newaff[MAX_NODES];
+    int i;
+    unsigned long offset = 0;
+
+    if (numnumanodes <= 0) term_printf("No NUMA nodes defined.\n"); else
+    if (numa_available() == -1) term_printf("Not a NUMA host.\n"); else
+    {
+        for (i = 0; i < numnumanodes; i++) newaff[i] = hostnodes[i];
+        parse_numa_args(affinity, newaff, NULL, NULL, MAX_NODES, 0);
+        for (i = 0; i < numnumanodes; i++) {
+            if (newaff[i] != hostnodes[i]) {
+                 hostnodes[i] = newaff[i];
+                 if (hostnodes[i] == (uint64_t)-1)
+                     numa_tonodemask_memory (phys_ram_base + offset,
+                         node_mem[i], &numa_all_nodes);
+                 else
+                     numa_tonode_memory(phys_ram_base + offset,
+                         node_mem[i], hostnodes[i] % (numa_max_node() + 1));
+            }
+            offset += node_mem[i];
+        }
+    }
+#else
+    term_printf ("NUMA host affinity support not compiled in\n");
+#endif
 }
 
 #ifdef CONFIG_PROFILER
@@ -1514,6 +1553,8 @@ static const term_cmd_t term_cmds[] = {
       "value", "set maximum speed (in bytes) for migrations" },
     { "balloon", "i", do_balloon,
       "target", "request VM to change it's memory allocation (in MB)" },
+    { "numa", "s", do_numa,
+      "affinity", "change NUMA affinity" },
     { NULL, NULL, },
 };
 

[Qemu-devel] Re: [PATCH 7/8] v2: add numa monitor command

[Anthony Liguori]

Andre Przywara wrote:
> Signed-off-by: Andre Przywara <andre.przywara@amd.com>
>
> # HG changeset patch
> # User Andre Przywara <andre.przywara@amd.com>
> # Date 1229435568 -3600
> # Node ID 4f0a8ac2d88ffffc1dcd82785c1620553baa86da
> # Parent  5a74bd76931b79713803795b3943aa8383946521
> add -numa monitor command to repin guest nodes
>
> diff -r 5a74bd76931b -r 4f0a8ac2d88f monitor.c
> --- a/monitor.c	Tue Dec 16 14:51:24 2008 +0100
> +++ b/monitor.c	Tue Dec 16 14:52:48 2008 +0100
> @@ -39,6 +39,9 @@
>  #include "qemu-timer.h"
>  #include "migration.h"
>  #include "kvm.h"
> +#ifdef CONFIG_NUMA
> +#include <numa.h>
> +#endif
>  
>  //#define DEBUG
>  //#define DEBUG_COMPLETION
> @@ -1285,6 +1288,9 @@ static void do_info_numa(void)
>  {
>      int i, j;
>  
> +#ifndef CONFIG_NUMA
> +    term_printf("NUMA placement support: not compiled\n");
> +#endif
>      term_printf("%d nodes\n", numnumanodes);
>      for (i = 0; i < numnumanodes; i++) {
>          term_printf("node %d cpus:", i);
> @@ -1292,7 +1298,40 @@ static void do_info_numa(void)
>              if (node_to_cpus[i] & (1ULL << j)) term_printf(" %d", j);
>          term_printf("\n");
>          term_printf("node %d size: %" PRId64 " MB\n", i, node_mem[i] >> 20);
> -    }
> +        term_printf("node %d host: ", i);
> +        if (hostnodes[i] == (uint64_t)-1) term_printf("*\n"); else
> +            term_printf("%" PRId64 "\n", hostnodes[i]);
> +    }
> +}
> +
> +static void do_numa(const char *affinity)
> +{
> +#ifdef CONFIG_NUMA
> +    uint64_t newaff[MAX_NODES];
> +    int i;
> +    unsigned long offset = 0;
> +
> +    if (numnumanodes <= 0) term_printf("No NUMA nodes defined.\n"); else
> +    if (numa_available() == -1) term_printf("Not a NUMA host.\n"); else
> +    {
>   

This should be reformatted.

> +        for (i = 0; i < numnumanodes; i++) newaff[i] = hostnodes[i];
>   

This should be a separate line.

> +        parse_numa_args(affinity, newaff, NULL, NULL, MAX_NODES, 0);
> +        for (i = 0; i < numnumanodes; i++) {
> +            if (newaff[i] != hostnodes[i]) {
> +                 hostnodes[i] = newaff[i];
> +                 if (hostnodes[i] == (uint64_t)-1)
>   

I don't think using u64's for cpu=>node mappings is a reasonable thing 
to do at this stage.  64 processors systems are relatively common these 
days.

> +                     numa_tonodemask_memory (phys_ram_base + offset,
> +                         node_mem[i], &numa_all_nodes);
> +                 else
> +                     numa_tonode_memory(phys_ram_base + offset,
> +                         node_mem[i], hostnodes[i] % (numa_max_node() + 1));
> +            }
> +            offset += node_mem[i];
> +        }
> +    }
> +#else
> +    term_printf ("NUMA host affinity support not compiled in\n");
> +#endif
>  }

Regards,

Anthony Liguori