zsmalloc: Fix CPU hotplug callback registration
authorSrivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Mon, 10 Mar 2014 20:39:59 +0000 (02:09 +0530)
committerAlex Shi <alex.shi@linaro.org>
Mon, 11 May 2015 12:40:31 +0000 (20:40 +0800)
Subsystems that want to register CPU hotplug callbacks, as well as perform
initialization for the CPUs that are already online, often do it as shown
below:

get_online_cpus();

for_each_online_cpu(cpu)
init_cpu(cpu);

register_cpu_notifier(&foobar_cpu_notifier);

put_online_cpus();

This is wrong, since it is prone to ABBA deadlocks involving the
cpu_add_remove_lock and the cpu_hotplug.lock (when running concurrently
with CPU hotplug operations).

Instead, the correct and race-free way of performing the callback
registration is:

cpu_notifier_register_begin();

for_each_online_cpu(cpu)
init_cpu(cpu);

/* Note the use of the double underscored version of the API */
__register_cpu_notifier(&foobar_cpu_notifier);

cpu_notifier_register_done();

Fix the zsmalloc code by using this latter form of callback registration.

Cc: Nitin Gupta <ngupta@vflare.org>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
(cherry picked from commit f0e71fcd0fa6f3f5495cd9ad3f1e4acd94446a55)
Signed-off-by: Alex Shi <alex.shi@linaro.org>
mm/zsmalloc.c

index 7a8d161b4cd76350b3e8c975a269d1f2596c1008..b728f10e353ba455f585ee47319bbdeb5a9a7f21 100644 (file)
@@ -780,21 +780,32 @@ static void zs_exit(void)
 {
        int cpu;
 
+       cpu_notifier_register_begin();
+
        for_each_online_cpu(cpu)
                zs_cpu_notifier(NULL, CPU_DEAD, (void *)(long)cpu);
-       unregister_cpu_notifier(&zs_cpu_nb);
+       __unregister_cpu_notifier(&zs_cpu_nb);
+
+       cpu_notifier_register_done();
 }
 
 static int zs_init(void)
 {
        int cpu, ret;
 
-       register_cpu_notifier(&zs_cpu_nb);
+       cpu_notifier_register_begin();
+
+       __register_cpu_notifier(&zs_cpu_nb);
        for_each_online_cpu(cpu) {
                ret = zs_cpu_notifier(NULL, CPU_UP_PREPARE, (void *)(long)cpu);
-               if (notifier_to_errno(ret))
+               if (notifier_to_errno(ret)) {
+                       cpu_notifier_register_done();
                        goto fail;
+               }
        }
+
+       cpu_notifier_register_done();
+
        return 0;
 fail:
        zs_exit();