1 files changed, 149 insertions, 37 deletions

diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index cfd2967f04a2..8b4bef05e222 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -2480,6 +2480,7 @@ static int emit_fiemap_extent(struct fiemap_extent_info *fieinfo,
 				struct fiemap_cache *cache,
 				u64 offset, u64 phys, u64 len, u32 flags)
 {
+	u64 cache_end;
 	int ret = 0;
 
 	/* Set at the end of extent_fiemap(). */
@@ -2489,15 +2490,102 @@ static int emit_fiemap_extent(struct fiemap_extent_info *fieinfo,
 		goto assign;
 
 	/*
-	 * Sanity check, extent_fiemap() should have ensured that new
-	 * fiemap extent won't overlap with cached one.
-	 * Not recoverable.
+	 * When iterating the extents of the inode, at extent_fiemap(), we may
+	 * find an extent that starts at an offset behind the end offset of the
+	 * previous extent we processed. This happens if fiemap is called
+	 * without FIEMAP_FLAG_SYNC and there are ordered extents completing
+	 * while we call btrfs_next_leaf() (through fiemap_next_leaf_item()).
 	 *
-	 * NOTE: Physical address can overlap, due to compression
+	 * For example we are in leaf X processing its last item, which is the
+	 * file extent item for file range [512K, 1M[, and after
+	 * btrfs_next_leaf() releases the path, there's an ordered extent that
+	 * completes for the file range [768K, 2M[, and that results in trimming
+	 * the file extent item so that it now corresponds to the file range
+	 * [512K, 768K[ and a new file extent item is inserted for the file
+	 * range [768K, 2M[, which may end up as the last item of leaf X or as
+	 * the first item of the next leaf - in either case btrfs_next_leaf()
+	 * will leave us with a path pointing to the new extent item, for the
+	 * file range [768K, 2M[, since that's the first key that follows the
+	 * last one we processed. So in order not to report overlapping extents
+	 * to user space, we trim the length of the previously cached extent and
+	 * emit it.
+	 *
+	 * Upon calling btrfs_next_leaf() we may also find an extent with an
+	 * offset smaller than or equals to cache->offset, and this happens
+	 * when we had a hole or prealloc extent with several delalloc ranges in
+	 * it, but after btrfs_next_leaf() released the path, delalloc was
+	 * flushed and the resulting ordered extents were completed, so we can
+	 * now have found a file extent item for an offset that is smaller than
+	 * or equals to what we have in cache->offset. We deal with this as
+	 * described below.
 	 */
-	if (cache->offset + cache->len > offset) {
-		WARN_ON(1);
-		return -EINVAL;
+	cache_end = cache->offset + cache->len;
+	if (cache_end > offset) {
+		if (offset == cache->offset) {
+			/*
+			 * We cached a dealloc range (found in the io tree) for
+			 * a hole or prealloc extent and we have now found a
+			 * file extent item for the same offset. What we have
+			 * now is more recent and up to date, so discard what
+			 * we had in the cache and use what we have just found.
+			 */
+			goto assign;
+		} else if (offset > cache->offset) {
+			/*
+			 * The extent range we previously found ends after the
+			 * offset of the file extent item we found and that
+			 * offset falls somewhere in the middle of that previous
+			 * extent range. So adjust the range we previously found
+			 * to end at the offset of the file extent item we have
+			 * just found, since this extent is more up to date.
+			 * Emit that adjusted range and cache the file extent
+			 * item we have just found. This corresponds to the case
+			 * where a previously found file extent item was split
+			 * due to an ordered extent completing.
+			 */
+			cache->len = offset - cache->offset;
+			goto emit;
+		} else {
+			const u64 range_end = offset + len;
+
+			/*
+			 * The offset of the file extent item we have just found
+			 * is behind the cached offset. This means we were
+			 * processing a hole or prealloc extent for which we
+			 * have found delalloc ranges (in the io tree), so what
+			 * we have in the cache is the last delalloc range we
+			 * found while the file extent item we found can be
+			 * either for a whole delalloc range we previously
+			 * emmitted or only a part of that range.
+			 *
+			 * We have two cases here:
+			 *
+			 * 1) The file extent item's range ends at or behind the
+			 *    cached extent's end. In this case just ignore the
+			 *    current file extent item because we don't want to
+			 *    overlap with previous ranges that may have been
+			 *    emmitted already;
+			 *
+			 * 2) The file extent item starts behind the currently
+			 *    cached extent but its end offset goes beyond the
+			 *    end offset of the cached extent. We don't want to
+			 *    overlap with a previous range that may have been
+			 *    emmitted already, so we emit the currently cached
+			 *    extent and then partially store the current file
+			 *    extent item's range in the cache, for the subrange
+			 *    going the cached extent's end to the end of the
+			 *    file extent item.
+			 */
+			if (range_end <= cache_end)
+				return 0;
+
+			if (!(flags & (FIEMAP_EXTENT_ENCODED | FIEMAP_EXTENT_DELALLOC)))
+				phys += cache_end - offset;
+
+			offset = cache_end;
+			len = range_end - cache_end;
+			goto emit;
+		}
 	}
 
 	/*
@@ -2517,6 +2605,7 @@ static int emit_fiemap_extent(struct fiemap_extent_info *fieinfo,
 		return 0;
 	}
 
+emit:
 	/* Not mergeable, need to submit cached one */
 	ret = fiemap_fill_next_extent(fieinfo, cache->offset, cache->phys,
 				      cache->len, cache->flags);
@@ -2689,16 +2778,34 @@ static int fiemap_process_hole(struct btrfs_inode *inode,
 	 * it beyond i_size.
 	 */
 	while (cur_offset < end && cur_offset < i_size) {
+		struct extent_state *cached_state = NULL;
 		u64 delalloc_start;
 		u64 delalloc_end;
 		u64 prealloc_start;
+		u64 lockstart;
+		u64 lockend;
 		u64 prealloc_len = 0;
 		bool delalloc;
 
+		lockstart = round_down(cur_offset, inode->root->fs_info->sectorsize);
+		lockend = round_up(end, inode->root->fs_info->sectorsize);
+
+		/*
+		 * We are only locking for the delalloc range because that's the
+		 * only thing that can change here.  With fiemap we have a lock
+		 * on the inode, so no buffered or direct writes can happen.
+		 *
+		 * However mmaps and normal page writeback will cause this to
+		 * change arbitrarily.  We have to lock the extent lock here to
+		 * make sure that nobody messes with the tree while we're doing
+		 * btrfs_find_delalloc_in_range.
+		 */
+		lock_extent(&inode->io_tree, lockstart, lockend, &cached_state);
 		delalloc = btrfs_find_delalloc_in_range(inode, cur_offset, end,
 							delalloc_cached_state,
 							&delalloc_start,
 							&delalloc_end);
+		unlock_extent(&inode->io_tree, lockstart, lockend, &cached_state);
 		if (!delalloc)
 			break;
 
@@ -2866,15 +2973,15 @@ int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo,
 		  u64 start, u64 len)
 {
 	const u64 ino = btrfs_ino(inode);
-	struct extent_state *cached_state = NULL;
 	struct extent_state *delalloc_cached_state = NULL;
 	struct btrfs_path *path;
 	struct fiemap_cache cache = { 0 };
 	struct btrfs_backref_share_check_ctx *backref_ctx;
 	u64 last_extent_end;
 	u64 prev_extent_end;
-	u64 lockstart;
-	u64 lockend;
+	u64 range_start;
+	u64 range_end;
+	const u64 sectorsize = inode->root->fs_info->sectorsize;
 	bool stopped = false;
 	int ret;
 
@@ -2885,22 +2992,19 @@ int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo,
 		goto out;
 	}
 
-	lockstart = round_down(start, inode->root->fs_info->sectorsize);
-	lockend = round_up(start + len, inode->root->fs_info->sectorsize);
-	prev_extent_end = lockstart;
-
-	btrfs_inode_lock(inode, BTRFS_ILOCK_SHARED);
-	lock_extent(&inode->io_tree, lockstart, lockend, &cached_state);
+	range_start = round_down(start, sectorsize);
+	range_end = round_up(start + len, sectorsize);
+	prev_extent_end = range_start;
 
 	ret = fiemap_find_last_extent_offset(inode, path, &last_extent_end);
 	if (ret < 0)
-		goto out_unlock;
+		goto out;
 	btrfs_release_path(path);
 
 	path->reada = READA_FORWARD;
-	ret = fiemap_search_slot(inode, path, lockstart);
+	ret = fiemap_search_slot(inode, path, range_start);
 	if (ret < 0) {
-		goto out_unlock;
+		goto out;
 	} else if (ret > 0) {
 		/*
 		 * No file extent item found, but we may have delalloc between
@@ -2910,7 +3014,7 @@ int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo,
 		goto check_eof_delalloc;
 	}
 
-	while (prev_extent_end < lockend) {
+	while (prev_extent_end < range_end) {
 		struct extent_buffer *leaf = path->nodes[0];
 		struct btrfs_file_extent_item *ei;
 		struct btrfs_key key;
@@ -2933,21 +3037,21 @@ int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo,
 		 * The first iteration can leave us at an extent item that ends
 		 * before our range's start. Move to the next item.
 		 */
-		if (extent_end <= lockstart)
+		if (extent_end <= range_start)
 			goto next_item;
 
 		backref_ctx->curr_leaf_bytenr = leaf->start;
 
 		/* We have in implicit hole (NO_HOLES feature enabled). */
 		if (prev_extent_end < key.offset) {
-			const u64 range_end = min(key.offset, lockend) - 1;
+			const u64 hole_end = min(key.offset, range_end) - 1;
 
 			ret = fiemap_process_hole(inode, fieinfo, &cache,
 						  &delalloc_cached_state,
 						  backref_ctx, 0, 0, 0,
-						  prev_extent_end, range_end);
+						  prev_extent_end, hole_end);
 			if (ret < 0) {
-				goto out_unlock;
+				goto out;
 			} else if (ret > 0) {
 				/* fiemap_fill_next_extent() told us to stop. */
 				stopped = true;
@@ -2955,7 +3059,7 @@ int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo,
 			}
 
 			/* We've reached the end of the fiemap range, stop. */
-			if (key.offset >= lockend) {
+			if (key.offset >= range_end) {
 				stopped = true;
 				break;
 			}
@@ -3003,7 +3107,7 @@ int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo,
 								  extent_gen,
 								  backref_ctx);
 				if (ret < 0)
-					goto out_unlock;
+					goto out;
 				else if (ret > 0)
 					flags |= FIEMAP_EXTENT_SHARED;
 			}
@@ -3014,7 +3118,7 @@ int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo,
 		}
 
 		if (ret < 0) {
-			goto out_unlock;
+			goto out;
 		} else if (ret > 0) {
 			/* fiemap_fill_next_extent() told us to stop. */
 			stopped = true;
@@ -3025,12 +3129,12 @@ int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo,
 next_item:
 		if (fatal_signal_pending(current)) {
 			ret = -EINTR;
-			goto out_unlock;
+			goto out;
 		}
 
 		ret = fiemap_next_leaf_item(inode, path);
 		if (ret < 0) {
-			goto out_unlock;
+			goto out;
 		} else if (ret > 0) {
 			/* No more file extent items for this inode. */
 			break;
@@ -3049,29 +3153,41 @@ check_eof_delalloc:
 	btrfs_free_path(path);
 	path = NULL;
 
-	if (!stopped && prev_extent_end < lockend) {
+	if (!stopped && prev_extent_end < range_end) {
 		ret = fiemap_process_hole(inode, fieinfo, &cache,
 					  &delalloc_cached_state, backref_ctx,
-					  0, 0, 0, prev_extent_end, lockend - 1);
+					  0, 0, 0, prev_extent_end, range_end - 1);
 		if (ret < 0)
-			goto out_unlock;
-		prev_extent_end = lockend;
+			goto out;
+		prev_extent_end = range_end;
 	}
 
 	if (cache.cached && cache.offset + cache.len >= last_extent_end) {
 		const u64 i_size = i_size_read(&inode->vfs_inode);
 
 		if (prev_extent_end < i_size) {
+			struct extent_state *cached_state = NULL;
 			u64 delalloc_start;
 			u64 delalloc_end;
+			u64 lockstart;
+			u64 lockend;
 			bool delalloc;
 
+			lockstart = round_down(prev_extent_end, sectorsize);
+			lockend = round_up(i_size, sectorsize);
+
+			/*
+			 * See the comment in fiemap_process_hole as to why
+			 * we're doing the locking here.
+			 */
+			lock_extent(&inode->io_tree, lockstart, lockend, &cached_state);
 			delalloc = btrfs_find_delalloc_in_range(inode,
 								prev_extent_end,
 								i_size - 1,
 								&delalloc_cached_state,
 								&delalloc_start,
 								&delalloc_end);
+			unlock_extent(&inode->io_tree, lockstart, lockend, &cached_state);
 			if (!delalloc)
 				cache.flags |= FIEMAP_EXTENT_LAST;
 		} else {
@@ -3080,10 +3196,6 @@ check_eof_delalloc:
 	}
 
 	ret = emit_last_fiemap_cache(fieinfo, &cache);
-
-out_unlock:
-	unlock_extent(&inode->io_tree, lockstart, lockend, &cached_state);
-	btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED);
 out:
 	free_extent_state(delalloc_cached_state);
 	btrfs_free_backref_share_ctx(backref_ctx);