cpu hotplug: fix ksoftirqd termination on cpu hotplug with naughty realtime process
authorSatoru Takeuchi <takeuchi_satoru@jp.fujitsu.com>
Mon, 16 Jul 2007 06:39:48 +0000 (23:39 -0700)
committerLinus Torvalds <torvalds@woody.linux-foundation.org>
Mon, 16 Jul 2007 16:05:41 +0000 (09:05 -0700)
Fix ksoftirqd termination on cpu hotplug with naughty real time process.

Assuming the following case:

 - Try to hot remove CPU2 from CPU1.
 - There is a real time process on CPU2, and that process doesn't sleep at all.
 - That rt process and ksoftirqd/2 is migrated to the CPU0

Then ksoftirqd/2 can't stop becasue that rt process runs everlastingly on
CPU0, and CPU1 waiting the ksoftirqd/2's termination hangs up.  To fix this
problem, set the priority of ksoftirqd/2 to max one before kthread_stop().

[akpm@linux-foundation.org: fix warning]
Signed-off-by: Satoru Takeuchi <takeuchi_satoru@jp.fujitsu.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Ashok Raj <ashok.raj@intel.com>
Cc: Gautham R Shenoy <ego@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
kernel/softirq.c

index 73217a9e2875b27909cb4c9db09a6ec2d4284a7c..8de267790166fe1c062dabaae066c578829e6dba 100644 (file)
@@ -614,12 +614,16 @@ static int __cpuinit cpu_callback(struct notifier_block *nfb,
                kthread_bind(per_cpu(ksoftirqd, hotcpu),
                             any_online_cpu(cpu_online_map));
        case CPU_DEAD:
-       case CPU_DEAD_FROZEN:
+       case CPU_DEAD_FROZEN: {
+               struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 };
+
                p = per_cpu(ksoftirqd, hotcpu);
                per_cpu(ksoftirqd, hotcpu) = NULL;
+               sched_setscheduler(p, SCHED_FIFO, &param);
                kthread_stop(p);
                takeover_tasklets(hotcpu);
                break;
+       }
 #endif /* CONFIG_HOTPLUG_CPU */
        }
        return NOTIFY_OK;