for-5.12-rc3-tag

-----BEGIN PGP SIGNATURE-----
 
 iQIzBAABCgAdFiEE8rQSAMVO+zA4DBdWxWXV+ddtWDsFAmBTeBsACgkQxWXV+ddt
 WDtwcBAAoto5Pbc3Lvt0aha3qn9q/Ms9lNU3YIwTjqXV3lIRKksWCS7kQmWlFmLz
 dILhdRBg1iWVh8qbeqpL5su7yNJduypsY/ImJroukb/BzwQViFRDGy5qIc56qLH2
 OVTx4LQ0zdqVdD86Qj0mt9ilSjgXYN+J53IUjsSSyJIpgt3vVcfjCYSkFO8zBiMH
 eliRtYShzJHkjEwVWLZRzk76oTnFQEC28IdYJ4y95mYl2wCABfTU2ylSeVDTtc6O
 x+fNMHHRmde2nbsHc+0eMm7rYLXuzvyx/tY17u6A6iwEQLGjE4rXOVZ7kA93WgAd
 YTXhM/B+YFfirNh029Av/MJP+2t9YBEODAHl1tnOdM0mfvXkpimaW0jvUEhi5f6I
 ZGu5FytscsgjyUK827WL7bZKO8WMzTLQvB3ryZ9UcrHm3QbZ7xGdoBE2L86p4Euw
 LiXUALdOWeYjFKSW9WWKrtQBtdjlLQYqJt+hL0ifaGlnfoi2G+DQeKtL9ZAKH5Cu
 gcjDUewnJtYPLyDOCRjQPFcts/MD5o81qMLeEwshmZT/bNMD9JOGEppCxBWGWSCx
 dYGq04Wib/dN710i5jB1XbJboBmT2SZDyBeiKTpCXs5mECBU00uWkkO98oId1YS3
 wHu9qyGUOi2g88V27jH593/JstUYn6zyxJYIZX84mzcxOqZlKuo=
 =auMP
 -----END PGP SIGNATURE-----

Merge tag 'for-5.12-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:
 "There are still regressions being found and fixed in the zoned mode
  and subpage code, the rest are fixes for bugs reported by users.

  Regressions:

   - subpage block support:
      - readahead works on the proper block size
      - fix last page zeroing

   - zoned mode:
      - linked list corruption for tree log

  Fixes:

   - qgroup leak after falloc failure

   - tree mod log and backref resolving:
      - extent buffer cloning race when resolving backrefs
      - pin deleted leaves with active tree mod log users

   - drop debugging flag from slab cache"

* tag 'for-5.12-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: always pin deleted leaves when there are active tree mod log users
  btrfs: fix race when cloning extent buffer during rewind of an old root
  btrfs: fix slab cache flags for free space tree bitmap
  btrfs: subpage: make readahead work properly
  btrfs: subpage: fix wild pointer access during metadata read failure
  btrfs: zoned: fix linked list corruption after log root tree allocation failure
  btrfs: fix qgroup data rsv leak caused by falloc failure
  btrfs: track qgroup released data in own variable in insert_prealloc_file_extent
  btrfs: fix wrong offset to zero out range beyond i_size

fs/btrfs/ctree.c

@@ -1365,7 +1365,9 @@ get_old_root(struct btrfs_root *root, u64 time_seq)
 				   "failed to read tree block %llu from get_old_root",
 				   logical);
 		} else {
+			btrfs_tree_read_lock(old);
 			eb = btrfs_clone_extent_buffer(old);
+			btrfs_tree_read_unlock(old);
 			free_extent_buffer(old);
 		}
 	} else if (old_root) {

fs/btrfs/extent-tree.c

@@ -3323,6 +3323,7 @@ void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
 
 	if (last_ref && btrfs_header_generation(buf) == trans->transid) {
 		struct btrfs_block_group *cache;
+		bool must_pin = false;
 
 		if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
 			ret = check_ref_cleanup(trans, buf->start);
@@ -3340,7 +3341,27 @@ void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
 			goto out;
 		}
 
-		if (btrfs_is_zoned(fs_info)) {
+		/*
+		 * If this is a leaf and there are tree mod log users, we may
+		 * have recorded mod log operations that point to this leaf.
+		 * So we must make sure no one reuses this leaf's extent before
+		 * mod log operations are applied to a node, otherwise after
+		 * rewinding a node using the mod log operations we get an
+		 * inconsistent btree, as the leaf's extent may now be used as
+		 * a node or leaf for another different btree.
+		 * We are safe from races here because at this point no other
+		 * node or root points to this extent buffer, so if after this
+		 * check a new tree mod log user joins, it will not be able to
+		 * find a node pointing to this leaf and record operations that
+		 * point to this leaf.
+		 */
+		if (btrfs_header_level(buf) == 0) {
+			read_lock(&fs_info->tree_mod_log_lock);
+			must_pin = !list_empty(&fs_info->tree_mod_seq_list);
+			read_unlock(&fs_info->tree_mod_log_lock);
+		}
+
+		if (must_pin || btrfs_is_zoned(fs_info)) {
 			btrfs_redirty_list_add(trans->transaction, buf);
 			pin_down_extent(trans, cache, buf->start, buf->len, 1);
 			btrfs_put_block_group(cache);

fs/btrfs/extent_io.c

@@ -2885,6 +2885,35 @@ static void end_page_read(struct page *page, bool uptodate, u64 start, u32 len)
 		btrfs_subpage_end_reader(fs_info, page, start, len);
 }
 
+/*
+ * Find extent buffer for a givne bytenr.
+ *
+ * This is for end_bio_extent_readpage(), thus we can't do any unsafe locking
+ * in endio context.
+ */
+static struct extent_buffer *find_extent_buffer_readpage(
+		struct btrfs_fs_info *fs_info, struct page *page, u64 bytenr)
+{
+	struct extent_buffer *eb;
+
+	/*
+	 * For regular sectorsize, we can use page->private to grab extent
+	 * buffer
+	 */
+	if (fs_info->sectorsize == PAGE_SIZE) {
+		ASSERT(PagePrivate(page) && page->private);
+		return (struct extent_buffer *)page->private;
+	}
+
+	/* For subpage case, we need to lookup buffer radix tree */
+	rcu_read_lock();
+	eb = radix_tree_lookup(&fs_info->buffer_radix,
+			       bytenr >> fs_info->sectorsize_bits);
+	rcu_read_unlock();
+	ASSERT(eb);
+	return eb;
+}
+
 /*
  * after a readpage IO is done, we need to:
  * clear the uptodate bits on error
@@ -2996,7 +3025,7 @@ static void end_bio_extent_readpage(struct bio *bio)
 		} else {
 			struct extent_buffer *eb;
 
-			eb = (struct extent_buffer *)page->private;
+			eb = find_extent_buffer_readpage(fs_info, page, start);
 			set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
 			eb->read_mirror = mirror;
 			atomic_dec(&eb->io_pages);
@@ -3020,7 +3049,7 @@ readpage_ok:
 			 */
 			if (page->index == end_index && i_size <= end) {
 				u32 zero_start = max(offset_in_page(i_size),
-						     offset_in_page(end));
+						     offset_in_page(start));
 
 				zero_user_segment(page, zero_start,
 						  offset_in_page(end) + 1);

fs/btrfs/inode.c

@@ -9008,7 +9008,7 @@ int __init btrfs_init_cachep(void)
 
 	btrfs_free_space_bitmap_cachep = kmem_cache_create("btrfs_free_space_bitmap",
 							PAGE_SIZE, PAGE_SIZE,
-							SLAB_RED_ZONE, NULL);
+							SLAB_MEM_SPREAD, NULL);
 	if (!btrfs_free_space_bitmap_cachep)
 		goto fail;
 
@@ -9877,6 +9877,7 @@ static struct btrfs_trans_handle *insert_prealloc_file_extent(
 	struct btrfs_path *path;
 	u64 start = ins->objectid;
 	u64 len = ins->offset;
+	int qgroup_released;
 	int ret;
 
 	memset(&stack_fi, 0, sizeof(stack_fi));
@@ -9889,16 +9890,16 @@ static struct btrfs_trans_handle *insert_prealloc_file_extent(
 	btrfs_set_stack_file_extent_compression(&stack_fi, BTRFS_COMPRESS_NONE);
 	/* Encryption and other encoding is reserved and all 0 */
 
-	ret = btrfs_qgroup_release_data(inode, file_offset, len);
+	qgroup_released = btrfs_qgroup_release_data(inode, file_offset, len);
-	if (ret < 0)
+	if (qgroup_released < 0)
-		return ERR_PTR(ret);
+		return ERR_PTR(qgroup_released);
 
 	if (trans) {
 		ret = insert_reserved_file_extent(trans, inode,
 						  file_offset, &stack_fi,
-						  true, ret);
+						  true, qgroup_released);
 		if (ret)
-			return ERR_PTR(ret);
+			goto free_qgroup;
 		return trans;
 	}
 
@@ -9909,21 +9910,35 @@ static struct btrfs_trans_handle *insert_prealloc_file_extent(
 	extent_info.file_offset = file_offset;
 	extent_info.extent_buf = (char *)&stack_fi;
 	extent_info.is_new_extent = true;
-	extent_info.qgroup_reserved = ret;
+	extent_info.qgroup_reserved = qgroup_released;
 	extent_info.insertions = 0;
 
 	path = btrfs_alloc_path();
-	if (!path)
+	if (!path) {
-		return ERR_PTR(-ENOMEM);
+		ret = -ENOMEM;
+		goto free_qgroup;
+	}
 
 	ret = btrfs_replace_file_extents(&inode->vfs_inode, path, file_offset,
 				     file_offset + len - 1, &extent_info,
 				     &trans);
 	btrfs_free_path(path);
 	if (ret)
-		return ERR_PTR(ret);
+		goto free_qgroup;
-
 	return trans;
+
+free_qgroup:
+	/*
+	 * We have released qgroup data range at the beginning of the function,
+	 * and normally qgroup_released bytes will be freed when committing
+	 * transaction.
+	 * But if we error out early, we have to free what we have released
+	 * or we leak qgroup data reservation.
+	 */
+	btrfs_qgroup_free_refroot(inode->root->fs_info,
+			inode->root->root_key.objectid, qgroup_released,
+			BTRFS_QGROUP_RSV_DATA);
+	return ERR_PTR(ret);
 }
 
 static int __btrfs_prealloc_file_range(struct inode *inode, int mode,

fs/btrfs/reada.c

@@ -209,7 +209,7 @@ int btree_readahead_hook(struct extent_buffer *eb, int err)
 	/* find extent */
 	spin_lock(&fs_info->reada_lock);
 	re = radix_tree_lookup(&fs_info->reada_tree,
-			       eb->start >> PAGE_SHIFT);
+			       eb->start >> fs_info->sectorsize_bits);
 	if (re)
 		re->refcnt++;
 	spin_unlock(&fs_info->reada_lock);
@@ -240,7 +240,7 @@ static struct reada_zone *reada_find_zone(struct btrfs_device *dev, u64 logical,
 	zone = NULL;
 	spin_lock(&fs_info->reada_lock);
 	ret = radix_tree_gang_lookup(&dev->reada_zones, (void **)&zone,
-				     logical >> PAGE_SHIFT, 1);
+				     logical >> fs_info->sectorsize_bits, 1);
 	if (ret == 1 && logical >= zone->start && logical <= zone->end) {
 		kref_get(&zone->refcnt);
 		spin_unlock(&fs_info->reada_lock);
@@ -283,13 +283,13 @@ static struct reada_zone *reada_find_zone(struct btrfs_device *dev, u64 logical,
 
 	spin_lock(&fs_info->reada_lock);
 	ret = radix_tree_insert(&dev->reada_zones,
-				(unsigned long)(zone->end >> PAGE_SHIFT),
+			(unsigned long)(zone->end >> fs_info->sectorsize_bits),
-				zone);
+			zone);
 
 	if (ret == -EEXIST) {
 		kfree(zone);
 		ret = radix_tree_gang_lookup(&dev->reada_zones, (void **)&zone,
-					     logical >> PAGE_SHIFT, 1);
+					logical >> fs_info->sectorsize_bits, 1);
 		if (ret == 1 && logical >= zone->start && logical <= zone->end)
 			kref_get(&zone->refcnt);
 		else
@@ -315,7 +315,7 @@ static struct reada_extent *reada_find_extent(struct btrfs_fs_info *fs_info,
 	u64 length;
 	int real_stripes;
 	int nzones = 0;
-	unsigned long index = logical >> PAGE_SHIFT;
+	unsigned long index = logical >> fs_info->sectorsize_bits;
 	int dev_replace_is_ongoing;
 	int have_zone = 0;
 
@@ -497,7 +497,7 @@ static void reada_extent_put(struct btrfs_fs_info *fs_info,
 			     struct reada_extent *re)
 {
 	int i;
-	unsigned long index = re->logical >> PAGE_SHIFT;
+	unsigned long index = re->logical >> fs_info->sectorsize_bits;
 
 	spin_lock(&fs_info->reada_lock);
 	if (--re->refcnt) {
@@ -538,11 +538,12 @@ static void reada_extent_put(struct btrfs_fs_info *fs_info,
 static void reada_zone_release(struct kref *kref)
 {
 	struct reada_zone *zone = container_of(kref, struct reada_zone, refcnt);
+	struct btrfs_fs_info *fs_info = zone->device->fs_info;
 
-	lockdep_assert_held(&zone->device->fs_info->reada_lock);
+	lockdep_assert_held(&fs_info->reada_lock);
 
 	radix_tree_delete(&zone->device->reada_zones,
-			  zone->end >> PAGE_SHIFT);
+			  zone->end >> fs_info->sectorsize_bits);
 
 	kfree(zone);
 }
@@ -593,7 +594,7 @@ static int reada_add_block(struct reada_control *rc, u64 logical,
 static void reada_peer_zones_set_lock(struct reada_zone *zone, int lock)
 {
 	int i;
-	unsigned long index = zone->end >> PAGE_SHIFT;
+	unsigned long index = zone->end >> zone->device->fs_info->sectorsize_bits;
 
 	for (i = 0; i < zone->ndevs; ++i) {
 		struct reada_zone *peer;
@@ -628,7 +629,7 @@ static int reada_pick_zone(struct btrfs_device *dev)
 					     (void **)&zone, index, 1);
 		if (ret == 0)
 			break;
-		index = (zone->end >> PAGE_SHIFT) + 1;
+		index = (zone->end >> dev->fs_info->sectorsize_bits) + 1;
 		if (zone->locked) {
 			if (zone->elems > top_locked_elems) {
 				top_locked_elems = zone->elems;
@@ -709,7 +710,7 @@ static int reada_start_machine_dev(struct btrfs_device *dev)
 	 * plugging to speed things up
 	 */
 	ret = radix_tree_gang_lookup(&dev->reada_extents, (void **)&re,
-				     dev->reada_next >> PAGE_SHIFT, 1);
+				dev->reada_next >> fs_info->sectorsize_bits, 1);
 	if (ret == 0 || re->logical > dev->reada_curr_zone->end) {
 		ret = reada_pick_zone(dev);
 		if (!ret) {
@@ -718,7 +719,7 @@ static int reada_start_machine_dev(struct btrfs_device *dev)
 		}
 		re = NULL;
 		ret = radix_tree_gang_lookup(&dev->reada_extents, (void **)&re,
-					dev->reada_next >> PAGE_SHIFT, 1);
+				dev->reada_next >> fs_info->sectorsize_bits, 1);
 	}
 	if (ret == 0) {
 		spin_unlock(&fs_info->reada_lock);
@@ -885,7 +886,7 @@ static void dump_devs(struct btrfs_fs_info *fs_info, int all)
 				pr_cont(" curr off %llu",
 					device->reada_next - zone->start);
 			pr_cont("\n");
-			index = (zone->end >> PAGE_SHIFT) + 1;
+			index = (zone->end >> fs_info->sectorsize_bits) + 1;
 		}
 		cnt = 0;
 		index = 0;
@@ -910,7 +911,7 @@ static void dump_devs(struct btrfs_fs_info *fs_info, int all)
 				}
 			}
 			pr_cont("\n");
-			index = (re->logical >> PAGE_SHIFT) + 1;
+			index = (re->logical >> fs_info->sectorsize_bits) + 1;
 			if (++cnt > 15)
 				break;
 		}
@@ -926,7 +927,7 @@ static void dump_devs(struct btrfs_fs_info *fs_info, int all)
 		if (ret == 0)
 			break;
 		if (!re->scheduled) {
-			index = (re->logical >> PAGE_SHIFT) + 1;
+			index = (re->logical >> fs_info->sectorsize_bits) + 1;
 			continue;
 		}
 		pr_debug("re: logical %llu size %u list empty %d scheduled %d",
@@ -942,7 +943,7 @@ static void dump_devs(struct btrfs_fs_info *fs_info, int all)
 			}
 		}
 		pr_cont("\n");
-		index = (re->logical >> PAGE_SHIFT) + 1;
+		index = (re->logical >> fs_info->sectorsize_bits) + 1;
 	}
 	spin_unlock(&fs_info->reada_lock);
 }

fs/btrfs/tree-log.c

@@ -3169,10 +3169,6 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 
 	mutex_lock(&log_root_tree->log_mutex);
 
-	index2 = log_root_tree->log_transid % 2;
-	list_add_tail(&root_log_ctx.list, &log_root_tree->log_ctxs[index2]);
-	root_log_ctx.log_transid = log_root_tree->log_transid;
-
 	if (btrfs_is_zoned(fs_info)) {
 		if (!log_root_tree->node) {
 			ret = btrfs_alloc_log_tree_node(trans, log_root_tree);
@@ -3183,6 +3179,10 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 		}
 	}
 
+	index2 = log_root_tree->log_transid % 2;
+	list_add_tail(&root_log_ctx.list, &log_root_tree->log_ctxs[index2]);
+	root_log_ctx.log_transid = log_root_tree->log_transid;
+
 	/*
 	 * Now we are safe to update the log_root_tree because we're under the
 	 * log_mutex, and we're a current writer so we're holding the commit