aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorOleg Nesterov <oleg@redhat.com>2014-02-05 17:12:45 -0500
committerNeilBrown <neilb@suse.de>2014-02-12 21:46:45 -0500
commit789b5e0315284463617e106baad360cb9e8db3ac (patch)
tree3014756d06cac24b2e0692f0ab05a351b7323677
parent1877db75589a895bbdc4c4c3f23558e57b521141 (diff)
md/raid5: Fix CPU hotplug callback registration
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). Interestingly, the raid5 code can actually prevent double initialization and hence can use the following simplified form of callback registration: register_cpu_notifier(&foobar_cpu_notifier); get_online_cpus(); for_each_online_cpu(cpu) init_cpu(cpu); put_online_cpus(); A hotplug operation that occurs between registering the notifier and calling get_online_cpus(), won't disrupt anything, because the code takes care to perform the memory allocations only once. So reorganize the code in raid5 this way to fix the deadlock with callback registration. Cc: linux-raid@vger.kernel.org Cc: stable@vger.kernel.org (v2.6.32+) Fixes: 36d1c6476be51101778882897b315bd928c8c7b5 Signed-off-by: Oleg Nesterov <oleg@redhat.com> [Srivatsa: Fixed the unregister_cpu_notifier() deadlock, added the free_scratch_buffer() helper to condense code further and wrote the changelog.] Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: NeilBrown <neilb@suse.de>
-rw-r--r--drivers/md/raid5.c90
1 files changed, 44 insertions, 46 deletions
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index f1feadeb7bb2..16f5c21963db 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -5514,23 +5514,43 @@ raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks)
5514 return sectors * (raid_disks - conf->max_degraded); 5514 return sectors * (raid_disks - conf->max_degraded);
5515} 5515}
5516 5516
5517static void free_scratch_buffer(struct r5conf *conf, struct raid5_percpu *percpu)
5518{
5519 safe_put_page(percpu->spare_page);
5520 kfree(percpu->scribble);
5521 percpu->spare_page = NULL;
5522 percpu->scribble = NULL;
5523}
5524
5525static int alloc_scratch_buffer(struct r5conf *conf, struct raid5_percpu *percpu)
5526{
5527 if (conf->level == 6 && !percpu->spare_page)
5528 percpu->spare_page = alloc_page(GFP_KERNEL);
5529 if (!percpu->scribble)
5530 percpu->scribble = kmalloc(conf->scribble_len, GFP_KERNEL);
5531
5532 if (!percpu->scribble || (conf->level == 6 && !percpu->spare_page)) {
5533 free_scratch_buffer(conf, percpu);
5534 return -ENOMEM;
5535 }
5536
5537 return 0;
5538}
5539
5517static void raid5_free_percpu(struct r5conf *conf) 5540static void raid5_free_percpu(struct r5conf *conf)
5518{ 5541{
5519 struct raid5_percpu *percpu;
5520 unsigned long cpu; 5542 unsigned long cpu;
5521 5543
5522 if (!conf->percpu) 5544 if (!conf->percpu)
5523 return; 5545 return;
5524 5546
5525 get_online_cpus();
5526 for_each_possible_cpu(cpu) {
5527 percpu = per_cpu_ptr(conf->percpu, cpu);
5528 safe_put_page(percpu->spare_page);
5529 kfree(percpu->scribble);
5530 }
5531#ifdef CONFIG_HOTPLUG_CPU 5547#ifdef CONFIG_HOTPLUG_CPU
5532 unregister_cpu_notifier(&conf->cpu_notify); 5548 unregister_cpu_notifier(&conf->cpu_notify);
5533#endif 5549#endif
5550
5551 get_online_cpus();
5552 for_each_possible_cpu(cpu)
5553 free_scratch_buffer(conf, per_cpu_ptr(conf->percpu, cpu));
5534 put_online_cpus(); 5554 put_online_cpus();
5535 5555
5536 free_percpu(conf->percpu); 5556 free_percpu(conf->percpu);
@@ -5557,15 +5577,7 @@ static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
5557 switch (action) { 5577 switch (action) {
5558 case CPU_UP_PREPARE: 5578 case CPU_UP_PREPARE:
5559 case CPU_UP_PREPARE_FROZEN: 5579 case CPU_UP_PREPARE_FROZEN:
5560 if (conf->level == 6 && !percpu->spare_page) 5580 if (alloc_scratch_buffer(conf, percpu)) {
5561 percpu->spare_page = alloc_page(GFP_KERNEL);
5562 if (!percpu->scribble)
5563 percpu->scribble = kmalloc(conf->scribble_len, GFP_KERNEL);
5564
5565 if (!percpu->scribble ||
5566 (conf->level == 6 && !percpu->spare_page)) {
5567 safe_put_page(percpu->spare_page);
5568 kfree(percpu->scribble);
5569 pr_err("%s: failed memory allocation for cpu%ld\n", 5581 pr_err("%s: failed memory allocation for cpu%ld\n",
5570 __func__, cpu); 5582 __func__, cpu);
5571 return notifier_from_errno(-ENOMEM); 5583 return notifier_from_errno(-ENOMEM);
@@ -5573,10 +5585,7 @@ static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
5573 break; 5585 break;
5574 case CPU_DEAD: 5586 case CPU_DEAD:
5575 case CPU_DEAD_FROZEN: 5587 case CPU_DEAD_FROZEN:
5576 safe_put_page(percpu->spare_page); 5588 free_scratch_buffer(conf, per_cpu_ptr(conf->percpu, cpu));
5577 kfree(percpu->scribble);
5578 percpu->spare_page = NULL;
5579 percpu->scribble = NULL;
5580 break; 5589 break;
5581 default: 5590 default:
5582 break; 5591 break;
@@ -5588,40 +5597,29 @@ static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
5588static int raid5_alloc_percpu(struct r5conf *conf) 5597static int raid5_alloc_percpu(struct r5conf *conf)
5589{ 5598{
5590 unsigned long cpu; 5599 unsigned long cpu;
5591 struct page *spare_page; 5600 int err = 0;
5592 struct raid5_percpu __percpu *allcpus;
5593 void *scribble;
5594 int err;
5595 5601
5596 allcpus = alloc_percpu(struct raid5_percpu); 5602 conf->percpu = alloc_percpu(struct raid5_percpu);
5597 if (!allcpus) 5603 if (!conf->percpu)
5598 return -ENOMEM; 5604 return -ENOMEM;
5599 conf->percpu = allcpus; 5605
5606#ifdef CONFIG_HOTPLUG_CPU
5607 conf->cpu_notify.notifier_call = raid456_cpu_notify;
5608 conf->cpu_notify.priority = 0;
5609 err = register_cpu_notifier(&conf->cpu_notify);
5610 if (err)
5611 return err;
5612#endif
5600 5613
5601 get_online_cpus(); 5614 get_online_cpus();
5602 err = 0;
5603 for_each_present_cpu(cpu) { 5615 for_each_present_cpu(cpu) {
5604 if (conf->level == 6) { 5616 err = alloc_scratch_buffer(conf, per_cpu_ptr(conf->percpu, cpu));
5605 spare_page = alloc_page(GFP_KERNEL); 5617 if (err) {
5606 if (!spare_page) { 5618 pr_err("%s: failed memory allocation for cpu%ld\n",
5607 err = -ENOMEM; 5619 __func__, cpu);
5608 break;
5609 }
5610 per_cpu_ptr(conf->percpu, cpu)->spare_page = spare_page;
5611 }
5612 scribble = kmalloc(conf->scribble_len, GFP_KERNEL);
5613 if (!scribble) {
5614 err = -ENOMEM;
5615 break; 5620 break;
5616 } 5621 }
5617 per_cpu_ptr(conf->percpu, cpu)->scribble = scribble;
5618 } 5622 }
5619#ifdef CONFIG_HOTPLUG_CPU
5620 conf->cpu_notify.notifier_call = raid456_cpu_notify;
5621 conf->cpu_notify.priority = 0;
5622 if (err == 0)
5623 err = register_cpu_notifier(&conf->cpu_notify);
5624#endif
5625 put_online_cpus(); 5623 put_online_cpus();
5626 5624
5627 return err; 5625 return err;