14 files changed, 105 insertions, 85 deletions

diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c
index 77619a56e2bf..0dd6c2b5391b 100644
--- a/drivers/md/dm-crypt.c
+++ b/drivers/md/dm-crypt.c
@@ -331,25 +331,19 @@ crypt_alloc_buffer(struct crypt_config *cc, unsigned int size,
 	struct bio *bio;
 	unsigned int nr_iovecs = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
 	int gfp_mask = GFP_NOIO | __GFP_HIGHMEM;
-	unsigned long flags = current->flags;
 	unsigned int i;
 
 	/*
-	 * Tell VM to act less aggressively and fail earlier.
-	 * This is not necessary but increases throughput.
+	 * Use __GFP_NOMEMALLOC to tell the VM to act less aggressively and
+	 * to fail earlier.  This is not necessary but increases throughput.
 	 * FIXME: Is this really intelligent?
 	 */
-	current->flags &= ~PF_MEMALLOC;
-
 	if (base_bio)
-		bio = bio_clone(base_bio, GFP_NOIO);
+		bio = bio_clone(base_bio, GFP_NOIO|__GFP_NOMEMALLOC);
 	else
-		bio = bio_alloc(GFP_NOIO, nr_iovecs);
-	if (!bio) {
-		if (flags & PF_MEMALLOC)
-			current->flags |= PF_MEMALLOC;
+		bio = bio_alloc(GFP_NOIO|__GFP_NOMEMALLOC, nr_iovecs);
+	if (!bio)
 		return NULL;
-	}
 
 	/* if the last bio was not complete, continue where that one ended */
 	bio->bi_idx = *bio_vec_idx;
@@ -386,9 +380,6 @@ crypt_alloc_buffer(struct crypt_config *cc, unsigned int size,
 		size -= bv->bv_len;
 	}
 
-	if (flags & PF_MEMALLOC)
-		current->flags |= PF_MEMALLOC;
-
 	if (!bio->bi_size) {
 		bio_put(bio);
 		return NULL;
diff --git a/drivers/md/dm-emc.c b/drivers/md/dm-emc.c
index 700658664594..c7067674dcb7 100644
--- a/drivers/md/dm-emc.c
+++ b/drivers/md/dm-emc.c
@@ -223,8 +223,10 @@ static struct emc_handler *alloc_emc_handler(void)
 {
 	struct emc_handler *h = kmalloc(sizeof(*h), GFP_KERNEL);
 
-	if (h)
+	if (h) {
+		memset(h, 0, sizeof(*h));
 		spin_lock_init(&h->lock);
+	}
 
 	return h;
 }
@@ -259,8 +261,6 @@ static int emc_create(struct hw_handler *hwh, unsigned argc, char **argv)
 	if (!h)
 		return -ENOMEM;
 
-	memset(h, 0, sizeof(*h));
-
 	hwh->context = h;
 
 	if ((h->short_trespass = short_trespass))
diff --git a/drivers/md/dm-hw-handler.c b/drivers/md/dm-hw-handler.c
index ae63772e44c9..4cc0010e0156 100644
--- a/drivers/md/dm-hw-handler.c
+++ b/drivers/md/dm-hw-handler.c
@@ -23,7 +23,7 @@ struct hwh_internal {
 static LIST_HEAD(_hw_handlers);
 static DECLARE_RWSEM(_hwh_lock);
 
-struct hwh_internal *__find_hw_handler_type(const char *name)
+static struct hwh_internal *__find_hw_handler_type(const char *name)
 {
 	struct hwh_internal *hwhi;
 
diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c
index 43763a0bd096..1e97b3c12bd5 100644
--- a/drivers/md/dm-mpath.c
+++ b/drivers/md/dm-mpath.c
@@ -101,6 +101,7 @@ typedef int (*action_fn) (struct pgpath *pgpath);
 
 static kmem_cache_t *_mpio_cache;
 
+struct workqueue_struct *kmultipathd;
 static void process_queued_ios(void *data);
 static void trigger_event(void *data);
 
@@ -308,7 +309,7 @@ static int map_io(struct multipath *m, struct bio *bio, struct mpath_io *mpio,
 		bio_list_add(&m->queued_ios, bio);
 		m->queue_size++;
 		if (m->pg_init_required || !m->queue_io)
-			schedule_work(&m->process_queued_ios);
+			queue_work(kmultipathd, &m->process_queued_ios);
 		pgpath = NULL;
 		r = 0;
 	} else if (!pgpath)
@@ -334,7 +335,7 @@ static int queue_if_no_path(struct multipath *m, unsigned queue_if_no_path)
 
 	m->queue_if_no_path = queue_if_no_path;
 	if (!m->queue_if_no_path)
-		schedule_work(&m->process_queued_ios);
+		queue_work(kmultipathd, &m->process_queued_ios);
 
 	spin_unlock_irqrestore(&m->lock, flags);
 
@@ -800,7 +801,7 @@ static int fail_path(struct pgpath *pgpath)
 	if (pgpath == m->current_pgpath)
 		m->current_pgpath = NULL;
 
-	schedule_work(&m->trigger_event);
+	queue_work(kmultipathd, &m->trigger_event);
 
 out:
 	spin_unlock_irqrestore(&m->lock, flags);
@@ -837,9 +838,9 @@ static int reinstate_path(struct pgpath *pgpath)
 
 	m->current_pgpath = NULL;
 	if (!m->nr_valid_paths++)
-		schedule_work(&m->process_queued_ios);
+		queue_work(kmultipathd, &m->process_queued_ios);
 
-	schedule_work(&m->trigger_event);
+	queue_work(kmultipathd, &m->trigger_event);
 
 out:
 	spin_unlock_irqrestore(&m->lock, flags);
@@ -883,7 +884,7 @@ static void bypass_pg(struct multipath *m, struct priority_group *pg,
 
 	spin_unlock_irqrestore(&m->lock, flags);
 
-	schedule_work(&m->trigger_event);
+	queue_work(kmultipathd, &m->trigger_event);
 }
 
 /*
@@ -913,7 +914,7 @@ static int switch_pg_num(struct multipath *m, const char *pgstr)
 	}
 	spin_unlock_irqrestore(&m->lock, flags);
 
-	schedule_work(&m->trigger_event);
+	queue_work(kmultipathd, &m->trigger_event);
 	return 0;
 }
 
@@ -968,7 +969,7 @@ void dm_pg_init_complete(struct path *path, unsigned err_flags)
 		m->current_pgpath = NULL;
 		m->current_pg = NULL;
 	}
-	schedule_work(&m->process_queued_ios);
+	queue_work(kmultipathd, &m->process_queued_ios);
 	spin_unlock_irqrestore(&m->lock, flags);
 }
 
@@ -1018,7 +1019,7 @@ static int do_end_io(struct multipath *m, struct bio *bio,
 	bio_list_add(&m->queued_ios, bio);
 	m->queue_size++;
 	if (!m->queue_io)
-		schedule_work(&m->process_queued_ios);
+		queue_work(kmultipathd, &m->process_queued_ios);
 	spin_unlock(&m->lock);
 
 	return 1;	/* io not complete */
@@ -1057,7 +1058,7 @@ static void multipath_presuspend(struct dm_target *ti)
 	spin_lock_irqsave(&m->lock, flags);
 	m->suspended = 1;
 	if (m->queue_if_no_path)
-		schedule_work(&m->process_queued_ios);
+		queue_work(kmultipathd, &m->process_queued_ios);
 	spin_unlock_irqrestore(&m->lock, flags);
 }
 
@@ -1274,6 +1275,15 @@ static int __init dm_multipath_init(void)
 		return -EINVAL;
 	}
 
+	kmultipathd = create_workqueue("kmpathd");
+	if (!kmultipathd) {
+		DMERR("%s: failed to create workqueue kmpathd",
+				multipath_target.name);
+		dm_unregister_target(&multipath_target);
+		kmem_cache_destroy(_mpio_cache);
+		return -ENOMEM;
+	}
+
 	DMINFO("dm-multipath version %u.%u.%u loaded",
 	       multipath_target.version[0], multipath_target.version[1],
 	       multipath_target.version[2]);
@@ -1285,6 +1295,8 @@ static void __exit dm_multipath_exit(void)
 {
 	int r;
 
+	destroy_workqueue(kmultipathd);
+
 	r = dm_unregister_target(&multipath_target);
 	if (r < 0)
 		DMERR("%s: target unregister failed %d",
diff --git a/drivers/md/dm-path-selector.c b/drivers/md/dm-path-selector.c
index ac5c4bbec6c1..a28c1c2b4ef5 100644
--- a/drivers/md/dm-path-selector.c
+++ b/drivers/md/dm-path-selector.c
@@ -26,7 +26,7 @@ struct ps_internal {
 static LIST_HEAD(_path_selectors);
 static DECLARE_RWSEM(_ps_lock);
 
-struct ps_internal *__find_path_selector_type(const char *name)
+static struct ps_internal *__find_path_selector_type(const char *name)
 {
 	struct ps_internal *psi;
 
diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c
index ee175d4906c4..18e9b9953fcd 100644
--- a/drivers/md/dm-table.c
+++ b/drivers/md/dm-table.c
@@ -242,7 +242,7 @@ static void free_devices(struct list_head *devices)
 	}
 }
 
-void table_destroy(struct dm_table *t)
+static void table_destroy(struct dm_table *t)
 {
 	unsigned int i;
 
diff --git a/drivers/md/dm-zero.c b/drivers/md/dm-zero.c
index 7febc2cac73d..51c0639b2487 100644
--- a/drivers/md/dm-zero.c
+++ b/drivers/md/dm-zero.c
@@ -55,7 +55,7 @@ static struct target_type zero_target = {
 	.map    = zero_map,
 };
 
-int __init dm_zero_init(void)
+static int __init dm_zero_init(void)
 {
 	int r = dm_register_target(&zero_target);
 
@@ -65,7 +65,7 @@ int __init dm_zero_init(void)
 	return r;
 }
 
-void __exit dm_zero_exit(void)
+static void __exit dm_zero_exit(void)
 {
 	int r = dm_unregister_target(&zero_target);
 
diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index 243ff6884e83..f6b03957efc7 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -97,6 +97,7 @@ struct mapped_device {
 	 * freeze/thaw support require holding onto a super block
 	 */
 	struct super_block *frozen_sb;
+	struct block_device *frozen_bdev;
 };
 
 #define MIN_IOS 256
@@ -990,44 +991,50 @@ int dm_swap_table(struct mapped_device *md, struct dm_table *table)
  */
 static int __lock_fs(struct mapped_device *md)
 {
-	struct block_device *bdev;
+	int error = -ENOMEM;
 
 	if (test_and_set_bit(DMF_FS_LOCKED, &md->flags))
 		return 0;
 
-	bdev = bdget_disk(md->disk, 0);
-	if (!bdev) {
+	md->frozen_bdev = bdget_disk(md->disk, 0);
+	if (!md->frozen_bdev) {
 		DMWARN("bdget failed in __lock_fs");
-		return -ENOMEM;
+		goto out;
 	}
 
 	WARN_ON(md->frozen_sb);
-	md->frozen_sb = freeze_bdev(bdev);
+
+	md->frozen_sb = freeze_bdev(md->frozen_bdev);
+	if (IS_ERR(md->frozen_sb)) {
+		error = PTR_ERR(md->frozen_sb);
+		goto out_bdput;
+	}
+
 	/* don't bdput right now, we don't want the bdev
 	 * to go away while it is locked.  We'll bdput
 	 * in __unlock_fs
 	 */
 	return 0;
+
+out_bdput:
+	bdput(md->frozen_bdev);
+	md->frozen_sb = NULL;
+	md->frozen_bdev = NULL;
+out:
+	clear_bit(DMF_FS_LOCKED, &md->flags);
+	return error;
 }
 
-static int __unlock_fs(struct mapped_device *md)
+static void __unlock_fs(struct mapped_device *md)
 {
-	struct block_device *bdev;
-
 	if (!test_and_clear_bit(DMF_FS_LOCKED, &md->flags))
-		return 0;
+		return;
 
-	bdev = bdget_disk(md->disk, 0);
-	if (!bdev) {
-		DMWARN("bdget failed in __unlock_fs");
-		return -ENOMEM;
-	}
+	thaw_bdev(md->frozen_bdev, md->frozen_sb);
+	bdput(md->frozen_bdev);
 
-	thaw_bdev(bdev, md->frozen_sb);
 	md->frozen_sb = NULL;
-	bdput(bdev);
-	bdput(bdev);
-	return 0;
+	md->frozen_bdev = NULL;
 }
 
 /*
@@ -1041,37 +1048,37 @@ int dm_suspend(struct mapped_device *md)
 {
 	struct dm_table *map;
 	DECLARE_WAITQUEUE(wait, current);
+	int error = -EINVAL;
 
 	/* Flush I/O to the device. */
 	down_read(&md->lock);
-	if (test_bit(DMF_BLOCK_IO, &md->flags)) {
-		up_read(&md->lock);
-		return -EINVAL;
-	}
+	if (test_bit(DMF_BLOCK_IO, &md->flags))
+		goto out_read_unlock;
+
+	error = __lock_fs(md);
+	if (error)
+		goto out_read_unlock;
 
 	map = dm_get_table(md);
 	if (map)
 		dm_table_presuspend_targets(map);
-	__lock_fs(md);
 
 	up_read(&md->lock);
 
 	/*
-	 * First we set the BLOCK_IO flag so no more ios will be
-	 * mapped.
+	 * First we set the BLOCK_IO flag so no more ios will be mapped.
+	 *
+	 * If the flag is already set we know another thread is trying to
+	 * suspend as well, so we leave the fs locked for this thread.
 	 */
+	error = -EINVAL;
 	down_write(&md->lock);
-	if (test_bit(DMF_BLOCK_IO, &md->flags)) {
-		/*
-		 * If we get here we know another thread is
-		 * trying to suspend as well, so we leave the fs
-		 * locked for this thread.
-		 */
-		up_write(&md->lock);
-		return -EINVAL;
+	if (test_and_set_bit(DMF_BLOCK_IO, &md->flags)) {
+		if (map)
+			dm_table_put(map);
+		goto out_write_unlock;
 	}
 
-	set_bit(DMF_BLOCK_IO, &md->flags);
 	add_wait_queue(&md->wait, &wait);
 	up_write(&md->lock);
 
@@ -1099,12 +1106,9 @@ int dm_suspend(struct mapped_device *md)
 	remove_wait_queue(&md->wait, &wait);
 
 	/* were we interrupted ? */
-	if (atomic_read(&md->pending)) {
-		__unlock_fs(md);
-		clear_bit(DMF_BLOCK_IO, &md->flags);
-		up_write(&md->lock);
-		return -EINTR;
-	}
+	error = -EINTR;
+	if (atomic_read(&md->pending))
+		goto out_unfreeze;
 
 	set_bit(DMF_SUSPENDED, &md->flags);
 
@@ -1115,6 +1119,18 @@ int dm_suspend(struct mapped_device *md)
 	up_write(&md->lock);
 
 	return 0;
+
+out_unfreeze:
+	/* FIXME Undo dm_table_presuspend_targets */
+	__unlock_fs(md);
+	clear_bit(DMF_BLOCK_IO, &md->flags);
+out_write_unlock:
+	up_write(&md->lock);
+	return error;
+
+out_read_unlock:
+	up_read(&md->lock);
+	return error;
 }
 
 int dm_resume(struct mapped_device *md)
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 97af857d8a88..d899204d3743 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -957,7 +957,7 @@ static void super_1_sync(mddev_t *mddev, mdk_rdev_t *rdev)
 }
 
 
-struct super_type super_types[] = {
+static struct super_type super_types[] = {
 	[0] = {
 		.name	= "0.90.0",
 		.owner	= THIS_MODULE,
@@ -2740,7 +2740,7 @@ static struct block_device_operations md_fops =
 	.revalidate_disk= md_revalidate,
 };
 
-int md_thread(void * arg)
+static int md_thread(void * arg)
 {
 	mdk_thread_t *thread = arg;
 
@@ -3232,7 +3232,7 @@ void md_handle_safemode(mddev_t *mddev)
 }
 
 
-DECLARE_WAIT_QUEUE_HEAD(resync_wait);
+static DECLARE_WAIT_QUEUE_HEAD(resync_wait);
 
 #define SYNC_MARKS	10
 #define	SYNC_MARK_STEP	(3*HZ)
@@ -3575,8 +3575,8 @@ void md_check_recovery(mddev_t *mddev)
 	}
 }
 
-int md_notify_reboot(struct notifier_block *this,
-					unsigned long code, void *x)
+static int md_notify_reboot(struct notifier_block *this,
+			    unsigned long code, void *x)
 {
 	struct list_head *tmp;
 	mddev_t *mddev;
@@ -3599,7 +3599,7 @@ int md_notify_reboot(struct notifier_block *this,
 	return NOTIFY_DONE;
 }
 
-struct notifier_block md_notifier = {
+static struct notifier_block md_notifier = {
 	.notifier_call	= md_notify_reboot,
 	.next		= NULL,
 	.priority	= INT_MAX, /* before any real devices */
@@ -3616,7 +3616,7 @@ static void md_geninit(void)
 		p->proc_fops = &md_seq_fops;
 }
 
-int __init md_init(void)
+static int __init md_init(void)
 {
 	int minor;
 
diff --git a/drivers/md/multipath.c b/drivers/md/multipath.c
index c9b134cd1532..4e4bfde3db5d 100644
--- a/drivers/md/multipath.c
+++ b/drivers/md/multipath.c
@@ -103,7 +103,8 @@ static void multipath_end_bh_io (struct multipath_bh *mp_bh, int err)
 	mempool_free(mp_bh, conf->pool);
 }
 
-int multipath_end_request(struct bio *bio, unsigned int bytes_done, int error)
+static int multipath_end_request(struct bio *bio, unsigned int bytes_done,
+				 int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
 	struct multipath_bh * mp_bh = (struct multipath_bh *)(bio->bi_private);
@@ -355,7 +356,7 @@ static int multipath_remove_disk(mddev_t *mddev, int number)
 			goto abort;
 		}
 		p->rdev = NULL;
-		synchronize_kernel();
+		synchronize_rcu();
 		if (atomic_read(&rdev->nr_pending)) {
 			/* lost the race, try later */
 			err = -EBUSY;
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index a389394b52f6..83380b5d6593 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -797,7 +797,7 @@ static int raid1_remove_disk(mddev_t *mddev, int number)
 			goto abort;
 		}
 		p->rdev = NULL;
-		synchronize_kernel();
+		synchronize_rcu();
 		if (atomic_read(&rdev->nr_pending)) {
 			/* lost the race, try later */
 			err = -EBUSY;
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index b100bfe4fdca..e9dc2876a626 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -977,7 +977,7 @@ static int raid10_remove_disk(mddev_t *mddev, int number)
 			goto abort;
 		}
 		p->rdev = NULL;
-		synchronize_kernel();
+		synchronize_rcu();
 		if (atomic_read(&rdev->nr_pending)) {
 			/* lost the race, try later */
 			err = -EBUSY;
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 52c3a81c4aa7..e96e2a10a9c9 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -1873,7 +1873,7 @@ static int raid5_remove_disk(mddev_t *mddev, int number)
 			goto abort;
 		}
 		p->rdev = NULL;
-		synchronize_kernel();
+		synchronize_rcu();
 		if (atomic_read(&rdev->nr_pending)) {
 			/* lost the race, try later */
 			err = -EBUSY;
diff --git a/drivers/md/raid6main.c b/drivers/md/raid6main.c
index 7e30ab29691a..8a33f351e092 100644
--- a/drivers/md/raid6main.c
+++ b/drivers/md/raid6main.c
@@ -2038,7 +2038,7 @@ static int raid6_remove_disk(mddev_t *mddev, int number)
 			goto abort;
 		}
 		p->rdev = NULL;
-		synchronize_kernel();
+		synchronize_rcu();
 		if (atomic_read(&rdev->nr_pending)) {
 			/* lost the race, try later */
 			err = -EBUSY;