x86: fix a crash when adding both legacy AMD and CPUID 0x04 caches

The old code increased numcaches in the second case
but it added additional caches at the beginning of the array.
Uninitialized caches in the array caused a divide by zero
(cache->nbthreads_sharing) when used later.

Only occurs if the CPUID vendor isn't recognized
(neither Intel, nor AMD, nor Zhaoxin) or in case
of clang 4.0 bug on FreeBSD11.1 (#282).

That's also why the code was crashing on Zhaoxin
instead of just reporting wrong topology (#279).

Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr>

hwloc/topology-x86.c

@@ -360,6 +360,9 @@ static void look_proc(struct hwloc_backend *backend, struct procinfo *infos, uns
   if (cpuid_type != intel && cpuid_type != zhaoxin && has_topoext(features)) {
     unsigned apic_id, node_id, nodes_per_proc;
 
+    /* the code below doesn't want any other cache yet */
+    assert(!infos->numcaches);
+
     eax = 0x8000001e;
     cpuid_or_from_dump(&eax, &ebx, &ecx, &edx, src_cpuiddump);
     infos->apicid = apic_id = eax;
@@ -476,6 +479,8 @@ static void look_proc(struct hwloc_backend *backend, struct procinfo *infos, uns
     unsigned max_nbthreads;
     unsigned level;
 
+    unsigned oldnumcaches = infos->numcaches; /* in case we got caches above */
+
     for (cachenum = 0; ; cachenum++) {
       eax = 0x04;
       ecx = cachenum;
@@ -501,7 +506,8 @@ static void look_proc(struct hwloc_backend *backend, struct procinfo *infos, uns
       }
     }
 
-    cache = infos->cache = malloc(infos->numcaches * sizeof(*infos->cache));
+    infos->cache = realloc(infos->cache, infos->numcaches * sizeof(*infos->cache));
+    cache = &infos->cache[oldnumcaches];
 
     for (cachenum = 0; ; cachenum++) {
       unsigned long linesize, linepart, ways, sets;