MEDIUM: cfgparse: assign NUMA affinity to cpu-maps

Do not force affinity on the process, instead let's just apply it to
cpu-map, it will automatically be used later in the init process. We
can do this because we know that cpu-map was not set when we're using
this detection code.

This is much saner, as we don't need to manipulate the process' affinity
at this point in time, and just update the info that the user omitted to
set by themselves, which guarantees a better long-term consistency with
the documented feature.

diff --git a/src/cfgparse.c b/src/cfgparse.c
index dfbecfe17e43d59199b00d89618a7428294ff7c7..bb7fb71ed7d9c7e34163e8d095c3ea4f05231ac1 100644 (file)
--- a/src/cfgparse.c
+++ b/src/cfgparse.c
@@ -2637,6 +2637,7 @@ static int numa_detect_topology()
        struct hap_cpuset active_cpus, node_cpu_set;
        const char *parse_cpu_set_args[2];
        char *err = NULL;
+       int grp, thr;
 
        /* node_cpu_set count is used as return value */
        ha_cpuset_zero(&node_cpu_set);
@@ -2682,12 +2683,9 @@ static int numa_detect_topology()
                }
 
                ha_diag_warning("Multi-socket cpu detected, automatically binding on active CPUs of '%s' (%u active cpu(s))\n", node, ha_cpuset_count(&node_cpu_set));
-               if (sched_setaffinity(getpid(), sizeof(node_cpu_set.cpuset), &node_cpu_set.cpuset) == -1) {
-                       ha_warning("Cannot set the cpu affinity for this multi-cpu machine\n");
-
-                       /* clear the cpuset used as return value */
-                       ha_cpuset_zero(&node_cpu_set);
-               }
+               for (grp = 0; grp < MAX_TGROUPS; grp++)
+                       for (thr = 0; thr < MAX_THREADS_PER_GROUP; thr++)
+                               ha_cpuset_assign(&cpu_map[grp].thread[thr], &node_cpu_set);
 
                free(node_dirlist[node_dirlist_size]);
                break;
@@ -2707,6 +2705,7 @@ static int numa_detect_topology()
        struct hap_cpuset node_cpu_set;
        int ndomains = 0, i;
        size_t len = sizeof(ndomains);
+       int grp, thr;
 
        if (sysctlbyname("vm.ndomains", &ndomains, &len, NULL, 0) == -1) {
                ha_notice("Cannot assess the number of CPUs domains\n");
@@ -2736,12 +2735,9 @@ static int numa_detect_topology()
                ha_cpuset_assign(&node_cpu_set, &dom);
 
                ha_diag_warning("Multi-socket cpu detected, automatically binding on active CPUs of '%d' (%u active cpu(s))\n", i, ha_cpuset_count(&node_cpu_set));
-               if (cpuset_setaffinity(CPU_LEVEL_WHICH, CPU_WHICH_PID, -1, sizeof(node_cpu_set.cpuset), &node_cpu_set.cpuset) == -1) {
-                       ha_warning("Cannot set the cpu affinity for this multi-cpu machine\n");
-
-                       /* clear the cpuset used as return value */
-                       ha_cpuset_zero(&node_cpu_set);
-               }
+               for (grp = 0; grp < MAX_TGROUPS; grp++)
+                       for (thr = 0; thr < MAX_THREADS_PER_GROUP; thr++)
+                               ha_cpuset_assign(&cpu_map[grp].thread[thr], &node_cpu_set);
                break;
        }
  leave: