cpu/cacheinfo: Fix teardown path
authorBorislav Petkov <bp@suse.de>
Sat, 8 Aug 2015 08:46:02 +0000 (10:46 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 17 Sep 2015 13:06:54 +0000 (06:06 -0700)
Philip Müller reported a hang when booting 32-bit 4.1 kernel on an AMD
box. A fragment of the splat was enough to pinpoint the issue:

  task: f58e0000 ti: f58e8000 task.ti: f58e800
  EIP: 0060:[<c135a903>] EFLAGS: 00010206 CPU: 0
  EIP is at free_cache_attributes+0x83/0xd0
  EAX: 00000001 EBX: f589d46c ECX: 00000090 EDX: 360c2000
  ESI: 00000000 EDI: c1724a80 EBP: f58e9ec0 ESP: f58e9ea0
   DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
  CR0: 8005003b CR2: 000000ac CR3: 01731000 CR4: 000006d0

cache_shared_cpu_map_setup() did check sibling CPUs cacheinfo descriptor
while the respective teardown path cache_shared_cpu_map_remove() didn't.
Fix that.

>From tglx's version: to be on the safe side, move the cacheinfo
descriptor check to free_cache_attributes(), thus cleaning up the
hotplug path a little and making this even more robust.

Reported-and-tested-by: Philip Müller <philm@manjaro.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Sudeep Holla <sudeep.holla@arm.com>
Cc: <stable@vger.kernel.org> # 4.1
Cc: Andre Przywara <andre.przywara@arm.com>
Cc: Guenter Roeck <linux@roeck-us.net>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: linux-kernel@vger.kernel.org
Cc: manjaro-dev@manjaro.org
Cc: Philip Müller <philm@manjaro.org>
Link: https://lkml.kernel.org/r/55B47BB8.6080202@manjaro.org
Signed-off-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/base/cacheinfo.c

index 764280a917761e9bf3360111bb92709d676c69e4..e9fd32e91668992e478f31423c1e56be0724e1d2 100644 (file)
@@ -148,7 +148,11 @@ static void cache_shared_cpu_map_remove(unsigned int cpu)
 
                        if (sibling == cpu) /* skip itself */
                                continue;
+
                        sib_cpu_ci = get_cpu_cacheinfo(sibling);
+                       if (!sib_cpu_ci->info_list)
+                               continue;
+
                        sib_leaf = sib_cpu_ci->info_list + index;
                        cpumask_clear_cpu(cpu, &sib_leaf->shared_cpu_map);
                        cpumask_clear_cpu(sibling, &this_leaf->shared_cpu_map);
@@ -159,6 +163,9 @@ static void cache_shared_cpu_map_remove(unsigned int cpu)
 
 static void free_cache_attributes(unsigned int cpu)
 {
+       if (!per_cpu_cacheinfo(cpu))
+               return;
+
        cache_shared_cpu_map_remove(cpu);
 
        kfree(per_cpu_cacheinfo(cpu));
@@ -514,8 +521,7 @@ static int cacheinfo_cpu_callback(struct notifier_block *nfb,
                break;
        case CPU_DEAD:
                cache_remove_dev(cpu);
-               if (per_cpu_cacheinfo(cpu))
-                       free_cache_attributes(cpu);
+               free_cache_attributes(cpu);
                break;
        }
        return notifier_from_errno(rc);