[PATCH] workqueues: do CPU_UP_CANCELED if CPU_UP_PREPARE fails

From: Oleg Nesterov
Date: Tue Jul 22 2008 - 06:23:47 EST

(replaces workqueue-proper-error-unwinding-in-cpu-hotplug-error-path.patch)

The bug was pointed out by Akinobu Mita <akinobu.mita@xxxxxxxxx>, and this
patch is based on his original patch.

workqueue_cpu_callback(CPU_UP_PREPARE) expects that if it returns NOTIFY_BAD,
_cpu_up() will send CPU_UP_CANCELED then.

However, this is not true since

"cpu hotplug: cpu: deliver CPU_UP_CANCELED only to NOTIFY_OKed callbacks with CPU_UP_PREPARE"
commit: a0d8cdb652d35af9319a9e0fb7134de2a276c636

The callback which has returned NOTIFY_BAD will not receive CPU_UP_CANCELED.
Change the code to fulfil the CPU_UP_CANCELED logic if CPU_UP_PREPARE fails.

Signed-off-by: Oleg Nesterov <oleg@xxxxxxxxxx>

--- 26-rc2/kernel/workqueue.c~WQ_CPU_UP_PREPARE 2008-07-12 19:40:57.000000000 +0400
+++ 26-rc2/kernel/workqueue.c 2008-07-22 13:43:47.000000000 +0400
@@ -911,6 +911,7 @@ static int __devinit workqueue_cpu_callb
unsigned int cpu = (unsigned long)hcpu;
struct cpu_workqueue_struct *cwq;
struct workqueue_struct *wq;
+ int ret = NOTIFY_OK;

action &= ~CPU_TASKS_FROZEN;

@@ -918,7 +919,7 @@ static int __devinit workqueue_cpu_callb
cpu_set(cpu, cpu_populated_map);
list_for_each_entry(wq, &workqueues, list) {
cwq = per_cpu_ptr(wq->cpu_wq, cpu);

@@ -928,7 +929,9 @@ static int __devinit workqueue_cpu_callb
printk(KERN_ERR "workqueue [%s] for %i failed\n",
wq->name, cpu);
- return NOTIFY_BAD;
+ action = CPU_UP_CANCELED;
+ ret = NOTIFY_BAD;
+ goto undo;

start_workqueue_thread(cwq, cpu);
@@ -948,7 +951,7 @@ static int __devinit workqueue_cpu_callb
cpu_clear(cpu, cpu_populated_map);

- return NOTIFY_OK;
+ return ret;

void __init init_workqueues(void)

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/