nodeinfo: Get the correct CPU number on AMD Magny Cours platform

"Instead of developing one CPU with 12 cores, the Magny Cours is
actually two 6 core “Bulldozer” CPUs combined in to one package"

I.e, each package has two NUMA nodes, and the two numa nodes share
the same core ID set (0-6), which means parsing the cores number
from sysfs doesn't work in this case.

And the wrong CPU number could cause three problems for libvirt:

1) performance lost

  A domain without "cpuset" or "placement='auto'" (to drive numad)
specified will be only pinned to part of the CPUs.

2) domain can be started

  If a domain uses numad, and the advisory nodeset returned from
numad contains node which exceeds the range of wrong total CPU
number. The domain will fail to start, as the bitmask passed to
sched_setaffinity could be fully filled with zero.

3) wrong CPU number affects lots of stuffs.

  E.g. for command "virsh vcpuinfo", "virsh vcpupin", it will always
output with the truncated CPU list.

For more details:

https://www.redhat.com/archives/libvir-list/2012-May/msg00607.html

This patch is to fix the problem by parsing /proc/cpuinfo to get
the value of field "cpu cores", and use it as nodeinfo->cores if
it's greater than the cores number from sysfs.

diff --git a/src/nodeinfo.c b/src/nodeinfo.c
index cb047c40f451597b20a83b80a61c2b23044a9881..1c4272b0c25980e448890fa1c2760a0481205907 100644 (file)
--- a/src/nodeinfo.c
+++ b/src/nodeinfo.c
@@ -237,6 +237,7 @@ int linuxNodeInfoCPUPopulate(FILE *cpuinfo,
     int online;
     int ret = -1;
     char *sysfs_cpudir = NULL;
+    unsigned int cpu_cores = 0;
 
     nodeinfo->cpus = 0;
     nodeinfo->mhz = 0;
@@ -252,33 +253,58 @@ int linuxNodeInfoCPUPopulate(FILE *cpuinfo,
         if (STRPREFIX(buf, "cpu MHz")) {
             char *p;
             unsigned int ui;
+
             buf += 9;
             while (*buf && c_isspace(*buf))
                 buf++;
+
             if (*buf != ':' || !buf[1]) {
                 nodeReportError(VIR_ERR_INTERNAL_ERROR,
-                                "%s", _("parsing cpuinfo cpu MHz"));
+                                "%s", _("parsing cpu MHz from cpuinfo"));
                 goto cleanup;
             }
+
             if (virStrToLong_ui(buf+1, &p, 10, &ui) == 0
                 /* Accept trailing fractional part.  */
                 && (*p == '\0' || *p == '.' || c_isspace(*p)))
                 nodeinfo->mhz = ui;
         }
+
+        if (STRPREFIX(buf, "cpu cores")) {
+            char *p;
+            unsigned int ui;
+
+            buf += 9;
+            while (*buf && c_isspace(*buf))
+                buf++;
+
+            if (*buf != ':' || !buf[1]) {
+                nodeReportError(VIR_ERR_INTERNAL_ERROR,
+                                "%s", _("parsing cpu cores from cpuinfo"));
+                return -1;
+            }
+
+            if (virStrToLong_ui(buf+1, &p, 10, &ui) == 0
+                && (*p == '\0' || c_isspace(*p)))
+                cpu_cores = ui;
+         }
 # elif defined(__powerpc__) || \
       defined(__powerpc64__)
         char *buf = line;
         if (STRPREFIX(buf, "clock")) {
             char *p;
             unsigned int ui;
+
             buf += 5;
             while (*buf && c_isspace(*buf))
                 buf++;
+
             if (*buf != ':' || !buf[1]) {
                 nodeReportError(VIR_ERR_INTERNAL_ERROR,
-                                "%s", _("parsing cpuinfo cpu MHz"));
+                                "%s", _("parsing cpu MHz from cpuinfo"));
                 goto cleanup;
             }
+
             if (virStrToLong_ui(buf+1, &p, 10, &ui) == 0
                 /* Accept trailing fractional part.  */
                 && (*p == '\0' || *p == '.' || c_isspace(*p)))
@@ -377,6 +403,16 @@ int linuxNodeInfoCPUPopulate(FILE *cpuinfo,
         goto cleanup;
     }
 
+    /* Platform like AMD Magny Cours has two NUMA nodes each package, and
+     * the two nodes share the same core ID set, it results in the cores
+     * number calculated from sysfs is not the actual cores number. Use
+     * "cpu cores" in /proc/cpuinfo as the cores number instead in this case.
+     * More details about the problem:
+     * https://www.redhat.com/archives/libvir-list/2012-May/msg00607.html
+     */
+    if (cpu_cores && (cpu_cores > nodeinfo->cores))
+        nodeinfo->cores = cpu_cores;
+
     /* nodeinfo->sockets is supposed to be a number of sockets per NUMA node,
      * however if NUMA nodes are not composed of whole sockets, we just lie
      * about the number of NUMA nodes and force apps to check capabilities XML

diff --git a/tests/nodeinfodata/linux-nodeinfo-sysfs-test-3-cpu-x86-output.txt b/tests/nodeinfodata/linux-nodeinfo-sysfs-test-3-cpu-x86-output.txt
index e3c046b1e52f263ed70d7fb06e0449f612dd08f4..7db386451d8ce2d38ce30cdf6ae560a7612f9641 100644 (file)
--- a/tests/nodeinfodata/linux-nodeinfo-sysfs-test-3-cpu-x86-output.txt
+++ b/tests/nodeinfodata/linux-nodeinfo-sysfs-test-3-cpu-x86-output.txt
@@ -1 +1 @@
-CPUs: 48, MHz: 2100, Nodes: 1, Cores: 6
+CPUs: 48, MHz: 2100, Nodes: 1, Cores: 12