Skip to contentSkip to conversation

Patch "RAID5: revert e9e4c377e2f563 to fix a livelock" has been added to the 4.5-stable tree

PATCH [email protected] 1 participant 1 message Apr 10, 2016

Subsystems (2)
Stable branch MD RAID
lore
Share & tools
lore

[email protected]

author maintainer Applied
#0 lore raw

This is a note to let you know that I've just added the patch titled

RAID5: revert e9e4c377e2f563 to fix a livelock

to the 4.5-stable tree which can be found at:
http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary

The filename of the patch is:

     raid5-revert-e9e4c377e2f563-to-fix-a-livelock.patch

and it can be found in the queue-4.5 subdirectory.

If you, or anyone else, feels it should not be added to the stable tree, please let <[email protected]> know about it.

Quoted 1 line

From 6ab2a4b806ae21b6c3e47c5ff1285ec06d505325 Mon Sep 17 00:00:00 2001

From: Shaohua Li <[email protected]>
Date: Thu, 25 Feb 2016 16:24:42 -0800
Subject: RAID5: revert e9e4c377e2f563 to fix a livelock

From: Shaohua Li <[email protected]>

commit 6ab2a4b806ae21b6c3e47c5ff1285ec06d505325 upstream.

Revert commit
e9e4c377e2f563(md/raid5: per hash value and exclusive wait_for_stripe)

The problem is raid5_get_active_stripe waits on conf->wait_for_stripe[hash]. Assume hash is 0. My test release stripes in this order:
- release all stripes with hash 0
- raid5_get_active_stripe still sleeps since active_stripes > max_nr_stripes * 3 / 4
- release all stripes with hash other than 0. active_stripes becomes 0
- raid5_get_active_stripe still sleeps, since nobody wakes up wait_for_stripe[0]
The system live locks. The problem is active_stripes isn't a per-hash count. Revert the patch makes the live lock go away.

Cc Yuanhan Liu <[email protected]>
Cc NeilBrown <[email protected]>
Signed-off-by Shaohua Li <[email protected]>
Signed-off-by Greg Kroah-Hartman <[email protected]>
---
 drivers/md/raid5.c |   27 ++++++++-------------------
 drivers/md/raid5.h |    2 +-
 2 files changed, 9 insertions(+), 20 deletions(-)
drivers/md/raid5.c +8 -19
@@ -340,8 +340,7 @@ static void release_inactive_stripe_list
 					 int hash)
 {
 	int size;
-	unsigned long do_wakeup = 0;
-	int i = 0;
+	bool do_wakeup = false;
 	unsigned long flags;
 
 	if (hash == NR_STRIPE_HASH_LOCKS) {

@@ -362,19 +361,15 @@ static void release_inactive_stripe_list
 			    !list_empty(list))
 				atomic_dec(&conf->empty_inactive_list_nr);
 			list_splice_tail_init(list, conf->inactive_list + hash);
-			do_wakeup |= 1 << hash;
+			do_wakeup = true;
 			spin_unlock_irqrestore(conf->hash_locks + hash, flags);
 		}
 		size--;
 		hash--;
 	}
 
-	for (i = 0; i < NR_STRIPE_HASH_LOCKS; i++) {
-		if (do_wakeup & (1 << i))
-			wake_up(&conf->wait_for_stripe[i]);
-	}
-
 	if (do_wakeup) {
+		wake_up(&conf->wait_for_stripe);
 		if (atomic_read(&conf->active_stripes) == 0)
 			wake_up(&conf->wait_for_quiescent);
 		if (conf->retry_read_aligned)

@@ -687,15 +682,14 @@ raid5_get_active_stripe(struct r5conf *c
 			if (!sh) {
 				set_bit(R5_INACTIVE_BLOCKED,
 					&conf->cache_state);
-				wait_event_exclusive_cmd(
-					conf->wait_for_stripe[hash],
+				wait_event_lock_irq(
+					conf->wait_for_stripe,
 					!list_empty(conf->inactive_list + hash) &&
 					(atomic_read(&conf->active_stripes)
 					 < (conf->max_nr_stripes * 3 / 4)
 					 || !test_bit(R5_INACTIVE_BLOCKED,
 						      &conf->cache_state)),
-					spin_unlock_irq(conf->hash_locks + hash),
-					spin_lock_irq(conf->hash_locks + hash));
+					*(conf->hash_locks + hash));
 				clear_bit(R5_INACTIVE_BLOCKED,
 					  &conf->cache_state);
 			} else {

@@ -720,9 +714,6 @@ raid5_get_active_stripe(struct r5conf *c
 		}
 	} while (sh == NULL);
 
-	if (!list_empty(conf->inactive_list + hash))
-		wake_up(&conf->wait_for_stripe[hash]);
-
 	spin_unlock_irq(conf->hash_locks + hash);
 	return sh;
 }

@@ -2202,7 +2193,7 @@ static int resize_stripes(struct r5conf
 	cnt = 0;
 	list_for_each_entry(nsh, &newstripes, lru) {
 		lock_device_hash_lock(conf, hash);
-		wait_event_exclusive_cmd(conf->wait_for_stripe[hash],
+		wait_event_cmd(conf->wait_for_stripe,
 				    !list_empty(conf->inactive_list + hash),
 				    unlock_device_hash_lock(conf, hash),
 				    lock_device_hash_lock(conf, hash));

@@ -6521,9 +6512,7 @@ static struct r5conf *setup_conf(struct
 	seqcount_init(&conf->gen_lock);
 	mutex_init(&conf->cache_size_mutex);
 	init_waitqueue_head(&conf->wait_for_quiescent);
-	for (i = 0; i < NR_STRIPE_HASH_LOCKS; i++) {
-		init_waitqueue_head(&conf->wait_for_stripe[i]);
-	}
+	init_waitqueue_head(&conf->wait_for_stripe);
 	init_waitqueue_head(&conf->wait_for_overlap);
 	INIT_LIST_HEAD(&conf->handle_list);
 	INIT_LIST_HEAD(&conf->hold_list);
drivers/md/raid5.h +1 -1
@@ -524,7 +524,7 @@ struct r5conf {
 	atomic_t		empty_inactive_list_nr;
 	struct llist_head	released_stripes;
 	wait_queue_head_t	wait_for_quiescent;
-	wait_queue_head_t	wait_for_stripe[NR_STRIPE_HASH_LOCKS];
+	wait_queue_head_t	wait_for_stripe;
 	wait_queue_head_t	wait_for_overlap;
 	unsigned long		cache_state;
 #define R5_INACTIVE_BLOCKED	1	/* release of inactive stripes blocked,

Patches currently in stable-queue which might be from [email protected] are

queue-4.5/raid10-include-bio_end_io_list-in-nr_queued-to-prevent-freeze_array-hang.patch queue-4.5/md-multipath-don-t-hardcopy-bio-in-.make_request-path.patch queue-4.5/raid5-revert-e9e4c377e2f563-to-fix-a-livelock.patch queue-4.5/raid1-include-bio_end_io_list-in-nr_queued-to-prevent-freeze_array-hang.patch queue-4.5/raid5-check_reshape-shouldn-t-call-mddev_suspend.patch queue-4.5/md-raid5-compare-apples-to-apples-or-sectors-to-sectors.patch queue-4.5/md-raid5-preserve-stripe_preread_active-in-break_stripe_batch_list.patch