Merge commit 'v3.2-rc5' into sched/core

Merge reason: Pick up the latest fixes.

Signed-off-by: Ingo Molnar <mingo@elte.hu>

91 files changed, 2076 insertions, 1205 deletions

diff --git a/fs/bio.c b/fs/bio.c
index 41c93c72224..b1fe82cf88c 100644
--- a/fs/bio.c
+++ b/fs/bio.c
@@ -337,7 +337,7 @@ static void bio_fs_destructor(struct bio *bio)
  *	RETURNS:
  *	Pointer to new bio on success, NULL on failure.
  */
-struct bio *bio_alloc(gfp_t gfp_mask, int nr_iovecs)
+struct bio *bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs)
 {
 	struct bio *bio = bio_alloc_bioset(gfp_mask, nr_iovecs, fs_bio_set);
 
@@ -365,7 +365,7 @@ static void bio_kmalloc_destructor(struct bio *bio)
  *   %__GFP_WAIT, the allocation is guaranteed to succeed.
  *
  **/
-struct bio *bio_kmalloc(gfp_t gfp_mask, int nr_iovecs)
+struct bio *bio_kmalloc(gfp_t gfp_mask, unsigned int nr_iovecs)
 {
 	struct bio *bio;
 
@@ -696,7 +696,8 @@ static void bio_free_map_data(struct bio_map_data *bmd)
 	kfree(bmd);
 }
 
-static struct bio_map_data *bio_alloc_map_data(int nr_segs, int iov_count,
+static struct bio_map_data *bio_alloc_map_data(int nr_segs,
+					       unsigned int iov_count,
 					       gfp_t gfp_mask)
 {
 	struct bio_map_data *bmd;
diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index 8855aad3929..22c64fff1bd 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -683,7 +683,7 @@ static int inode_to_path(u64 inum, struct btrfs_inode_ref *iref,
 		return PTR_ERR(fspath);
 
 	if (fspath > fspath_min) {
-		ipath->fspath->val[i] = (u64)fspath;
+		ipath->fspath->val[i] = (u64)(unsigned long)fspath;
 		++ipath->fspath->elem_cnt;
 		ipath->fspath->bytes_left = fspath - fspath_min;
 	} else {
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 0fe615e4ea3..dede441bdee 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -514,10 +514,25 @@ static inline int should_cow_block(struct btrfs_trans_handle *trans,
 				   struct btrfs_root *root,
 				   struct extent_buffer *buf)
 {
+	/* ensure we can see the force_cow */
+	smp_rmb();
+
+	/*
+	 * We do not need to cow a block if
+	 * 1) this block is not created or changed in this transaction;
+	 * 2) this block does not belong to TREE_RELOC tree;
+	 * 3) the root is not forced COW.
+	 *
+	 * What is forced COW:
+	 *    when we create snapshot during commiting the transaction,
+	 *    after we've finished coping src root, we must COW the shared
+	 *    block to ensure the metadata consistency.
+	 */
 	if (btrfs_header_generation(buf) == trans->transid &&
 	    !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN) &&
 	    !(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID &&
-	      btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)))
+	      btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)) &&
+	    !root->force_cow)
 		return 0;
 	return 1;
 }
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index b9ba59ff929..50634abef9b 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -848,7 +848,8 @@ struct btrfs_free_cluster {
 enum btrfs_caching_type {
 	BTRFS_CACHE_NO		= 0,
 	BTRFS_CACHE_STARTED	= 1,
-	BTRFS_CACHE_FINISHED	= 2,
+	BTRFS_CACHE_FAST	= 2,
+	BTRFS_CACHE_FINISHED	= 3,
 };
 
 enum btrfs_disk_cache_state {
@@ -1271,6 +1272,8 @@ struct btrfs_root {
 	 * for stat.  It may be used for more later
 	 */
 	dev_t anon_dev;
+
+	int force_cow;
 };
 
 struct btrfs_ioctl_defrag_range_args {
@@ -2366,6 +2369,9 @@ int btrfs_block_rsv_check(struct btrfs_root *root,
 int btrfs_block_rsv_refill(struct btrfs_root *root,
 			  struct btrfs_block_rsv *block_rsv,
 			  u64 min_reserved);
+int btrfs_block_rsv_refill_noflush(struct btrfs_root *root,
+				   struct btrfs_block_rsv *block_rsv,
+				   u64 min_reserved);
 int btrfs_block_rsv_migrate(struct btrfs_block_rsv *src_rsv,
 			    struct btrfs_block_rsv *dst_rsv,
 			    u64 num_bytes);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 62afe5c5694..632f8f3cc9d 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -620,7 +620,7 @@ out:
 
 static int btree_io_failed_hook(struct bio *failed_bio,
 			 struct page *page, u64 start, u64 end,
-			 u64 mirror_num, struct extent_state *state)
+			 int mirror_num, struct extent_state *state)
 {
 	struct extent_io_tree *tree;
 	unsigned long len;
@@ -2573,22 +2573,10 @@ static int write_dev_supers(struct btrfs_device *device,
 	int errors = 0;
 	u32 crc;
 	u64 bytenr;
-	int last_barrier = 0;
 
 	if (max_mirrors == 0)
 		max_mirrors = BTRFS_SUPER_MIRROR_MAX;
 
-	/* make sure only the last submit_bh does a barrier */
-	if (do_barriers) {
-		for (i = 0; i < max_mirrors; i++) {
-			bytenr = btrfs_sb_offset(i);
-			if (bytenr + BTRFS_SUPER_INFO_SIZE >=
-			    device->total_bytes)
-				break;
-			last_barrier = i;
-		}
-	}
-
 	for (i = 0; i < max_mirrors; i++) {
 		bytenr = btrfs_sb_offset(i);
 		if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
@@ -2634,17 +2622,136 @@ static int write_dev_supers(struct btrfs_device *device,
 			bh->b_end_io = btrfs_end_buffer_write_sync;
 		}
 
-		if (i == last_barrier && do_barriers)
-			ret = submit_bh(WRITE_FLUSH_FUA, bh);
-		else
-			ret = submit_bh(WRITE_SYNC, bh);
-
+		/*
+		 * we fua the first super.  The others we allow
+		 * to go down lazy.
+		 */
+		ret = submit_bh(WRITE_FUA, bh);
 		if (ret)
 			errors++;
 	}
 	return errors < i ? 0 : -1;
 }
 
+/*
+ * endio for the write_dev_flush, this will wake anyone waiting
+ * for the barrier when it is done
+ */
+static void btrfs_end_empty_barrier(struct bio *bio, int err)
+{
+	if (err) {
+		if (err == -EOPNOTSUPP)
+			set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
+		clear_bit(BIO_UPTODATE, &bio->bi_flags);
+	}
+	if (bio->bi_private)
+		complete(bio->bi_private);
+	bio_put(bio);
+}
+
+/*
+ * trigger flushes for one the devices.  If you pass wait == 0, the flushes are
+ * sent down.  With wait == 1, it waits for the previous flush.
+ *
+ * any device where the flush fails with eopnotsupp are flagged as not-barrier
+ * capable
+ */
+static int write_dev_flush(struct btrfs_device *device, int wait)
+{
+	struct bio *bio;
+	int ret = 0;
+
+	if (device->nobarriers)
+		return 0;
+
+	if (wait) {
+		bio = device->flush_bio;
+		if (!bio)
+			return 0;
+
+		wait_for_completion(&device->flush_wait);
+
+		if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
+			printk("btrfs: disabling barriers on dev %s\n",
+			       device->name);
+			device->nobarriers = 1;
+		}
+		if (!bio_flagged(bio, BIO_UPTODATE)) {
+			ret = -EIO;
+		}
+
+		/* drop the reference from the wait == 0 run */
+		bio_put(bio);
+		device->flush_bio = NULL;
+
+		return ret;
+	}
+
+	/*
+	 * one reference for us, and we leave it for the
+	 * caller
+	 */
+	device->flush_bio = NULL;;
+	bio = bio_alloc(GFP_NOFS, 0);
+	if (!bio)
+		return -ENOMEM;
+
+	bio->bi_end_io = btrfs_end_empty_barrier;
+	bio->bi_bdev = device->bdev;
+	init_completion(&device->flush_wait);
+	bio->bi_private = &device->flush_wait;
+	device->flush_bio = bio;
+
+	bio_get(bio);
+	submit_bio(WRITE_FLUSH, bio);
+
+	return 0;
+}
+
+/*
+ * send an empty flush down to each device in parallel,
+ * then wait for them
+ */
+static int barrier_all_devices(struct btrfs_fs_info *info)
+{
+	struct list_head *head;
+	struct btrfs_device *dev;
+	int errors = 0;
+	int ret;
+
+	/* send down all the barriers */
+	head = &info->fs_devices->devices;
+	list_for_each_entry_rcu(dev, head, dev_list) {
+		if (!dev->bdev) {
+			errors++;
+			continue;
+		}
+		if (!dev->in_fs_metadata || !dev->writeable)
+			continue;
+
+		ret = write_dev_flush(dev, 0);
+		if (ret)
+			errors++;
+	}
+
+	/* wait for all the barriers */
+	list_for_each_entry_rcu(dev, head, dev_list) {
+		if (!dev->bdev) {
+			errors++;
+			continue;
+		}
+		if (!dev->in_fs_metadata || !dev->writeable)
+			continue;
+
+		ret = write_dev_flush(dev, 1);
+		if (ret)
+			errors++;
+	}
+	if (errors)
+		return -EIO;
+	return 0;
+}
+
 int write_all_supers(struct btrfs_root *root, int max_mirrors)
 {
 	struct list_head *head;
@@ -2666,6 +2773,10 @@ int write_all_supers(struct btrfs_root *root, int max_mirrors)
 
 	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
 	head = &root->fs_info->fs_devices->devices;
+
+	if (do_barriers)
+		barrier_all_devices(root->fs_info);
+
 	list_for_each_entry_rcu(dev, head, dev_list) {
 		if (!dev->bdev) {
 			total_errors++;
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index b232150b5b6..2ad813674d7 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -467,13 +467,59 @@ static int cache_block_group(struct btrfs_block_group_cache *cache,
 			     struct btrfs_root *root,
 			     int load_cache_only)
 {
+	DEFINE_WAIT(wait);
 	struct btrfs_fs_info *fs_info = cache->fs_info;
 	struct btrfs_caching_control *caching_ctl;
 	int ret = 0;
 
-	smp_mb();
-	if (cache->cached != BTRFS_CACHE_NO)
+	caching_ctl = kzalloc(sizeof(*caching_ctl), GFP_NOFS);
+	BUG_ON(!caching_ctl);
+
+	INIT_LIST_HEAD(&caching_ctl->list);
+	mutex_init(&caching_ctl->mutex);
+	init_waitqueue_head(&caching_ctl->wait);
+	caching_ctl->block_group = cache;
+	caching_ctl->progress = cache->key.objectid;
+	atomic_set(&caching_ctl->count, 1);
+	caching_ctl->work.func = caching_thread;
+
+	spin_lock(&cache->lock);
+	/*
+	 * This should be a rare occasion, but this could happen I think in the
+	 * case where one thread starts to load the space cache info, and then
+	 * some other thread starts a transaction commit which tries to do an
+	 * allocation while the other thread is still loading the space cache
+	 * info.  The previous loop should have kept us from choosing this block
+	 * group, but if we've moved to the state where we will wait on caching
+	 * block groups we need to first check if we're doing a fast load here,
+	 * so we can wait for it to finish, otherwise we could end up allocating
+	 * from a block group who's cache gets evicted for one reason or
+	 * another.
+	 */
+	while (cache->cached == BTRFS_CACHE_FAST) {
+		struct btrfs_caching_control *ctl;
+
+		ctl = cache->caching_ctl;
+		atomic_inc(&ctl->count);
+		prepare_to_wait(&ctl->wait, &wait, TASK_UNINTERRUPTIBLE);
+		spin_unlock(&cache->lock);
+
+		schedule();
+
+		finish_wait(&ctl->wait, &wait);
+		put_caching_control(ctl);
+		spin_lock(&cache->lock);
+	}
+
+	if (cache->cached != BTRFS_CACHE_NO) {
+		spin_unlock(&cache->lock);
+		kfree(caching_ctl);
 		return 0;
+	}
+	WARN_ON(cache->caching_ctl);
+	cache->caching_ctl = caching_ctl;
+	cache->cached = BTRFS_CACHE_FAST;
+	spin_unlock(&cache->lock);
 
 	/*
 	 * We can't do the read from on-disk cache during a commit since we need
@@ -484,56 +530,51 @@ static int cache_block_group(struct btrfs_block_group_cache *cache,
 	if (trans && (!trans->transaction->in_commit) &&
 	    (root && root != root->fs_info->tree_root) &&
 	    btrfs_test_opt(root, SPACE_CACHE)) {
-		spin_lock(&cache->lock);
-		if (cache->cached != BTRFS_CACHE_NO) {
-			spin_unlock(&cache->lock);
-			return 0;
-		}
-		cache->cached = BTRFS_CACHE_STARTED;
-		spin_unlock(&cache->lock);
-
 		ret = load_free_space_cache(fs_info, cache);
 
 		spin_lock(&cache->lock);
 		if (ret == 1) {
+			cache->caching_ctl = NULL;
 			cache->cached = BTRFS_CACHE_FINISHED;
 			cache->last_byte_to_unpin = (u64)-1;
 		} else {
-			cache->cached = BTRFS_CACHE_NO;
+			if (load_cache_only) {
+				cache->caching_ctl = NULL;
+				cache->cached = BTRFS_CACHE_NO;
+			} else {
+				cache->cached = BTRFS_CACHE_STARTED;
+			}
 		}
 		spin_unlock(&cache->lock);
+		wake_up(&caching_ctl->wait);
 		if (ret == 1) {
+			put_caching_control(caching_ctl);
 			free_excluded_extents(fs_info->extent_root, cache);
 			return 0;
 		}
+	} else {
+		/*
+		 * We are not going to do the fast caching, set cached to the
+		 * appropriate value and wakeup any waiters.
+		 */
+		spin_lock(&cache->lock);
+		if (load_cache_only) {
+			cache->caching_ctl = NULL;
+			cache->cached = BTRFS_CACHE_NO;
+		} else {
+			cache-