Hotplug CPU: don't check cpu_online after take_cpu_down
Rusty Russell [Mon, 28 Jul 2008 17:16:29 +0000 (12:16 -0500)]
Akinobu points out that if take_cpu_down() succeeds, the cpu must be offline.
Remove the cpu_online() check, and put a BUG_ON().

Quoting Akinobu Mita:
   Actually the cpu_online() check was necessary before appling this
   stop_machine: simplify patch.

   With old __stop_machine_run(), __stop_machine_run() could succeed
   (return !IS_ERR(p) value) even if take_cpu_down() returned non-zero value.
   The return value of take_cpu_down() was obtained through kthread_stop()..

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: "Akinobu Mita" <akinobu.mita@gmail.com>

kernel/cpu.c

index cf79bb9..53cf508 100644 (file)
@@ -250,8 +250,7 @@ static int __ref _cpu_down(unsigned int cpu, int tasks_frozen)
        set_cpus_allowed_ptr(current, &tmp);
 
        err = __stop_machine_run(take_cpu_down, &tcd_param, cpu);
-
-       if (err || cpu_online(cpu)) {
+       if (err) {
                /* CPU didn't die: tell everyone.  Can't complain. */
                if (raw_notifier_call_chain(&cpu_chain, CPU_DOWN_FAILED | mod,
                                            hcpu) == NOTIFY_BAD)
@@ -259,6 +258,7 @@ static int __ref _cpu_down(unsigned int cpu, int tasks_frozen)
 
                goto out_allowed;
        }
+       BUG_ON(cpu_online(cpu));
 
        /* Wait for it to sleep (leaving idle task). */
        while (!idle_cpu(cpu))