From: Josef Bacik <josef@toxicpanda.com>
Date: Fri, 17 Jan 2020 09:02:22 -0500
Git-commit: 9ddc959e802bf7555a0be543205ddcba2bae98bf
Patch-mainline: 5.7-rc1
References: bsc#1175377
Subject: [PATCH] btrfs: use the file extent tree infrastructure
We want to use this everywhere we modify the file extent items
permanently. These include:
1) Inserting new file extents for writes and prealloc extents.
2) Truncating inode items.
3) btrfs_cont_expand().
4) Insert inline extents.
5) Insert new extents from log replay.
6) Insert a new extent for clone, as it could be past i_size.
7) Hole punching
For hole punching in particular it might seem it's not necessary because
anybody extending would use btrfs_cont_expand, however there is a corner
that still can give us trouble. Start with an empty file and
fallocate KEEP_SIZE 1M-2M
We now have a 0 length file, and a hole file extent from 0-1M, and a
prealloc extent from 1M-2M. Now
punch 1M-1.5M
Because this is past i_size we have
[HOLE EXTENT][ NOTHING ][PREALLOC]
[0 1M][1M 1.5M][1.5M 2M]
with an i_size of 0. Now if we pwrite 0-1.5M we'll increas our i_size
to 1.5M, but our disk_i_size is still 0 until the ordered extent
completes.
However if we now immediately truncate 2M on the file we'll just call
btrfs_cont_expand(inode, 1.5M, 2M), since our old i_size is 1.5M. If we
commit the transaction here and crash we'll expose the gap.
To fix this we need to clear the file extent mapping for the range that
we punched but didn't insert a corresponding file extent for. This will
mean the truncate will only get an disk_i_size set to 1M if we crash
before the finish ordered io happens.
I've written an xfstest to reproduce the problem and validate this fix.
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
---
fs/btrfs/delayed-inode.c | 3 +++
fs/btrfs/file.c | 32 +++++++++++++++++++++++
fs/btrfs/inode.c | 55 +++++++++++++++++++++++++++++++++++++++-
fs/btrfs/tree-log.c | 5 ++++
4 files changed, 94 insertions(+), 1 deletion(-)
diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c
index d3e15e1d4a91..9b23883a1086 100644
--- a/fs/btrfs/delayed-inode.c
+++ b/fs/btrfs/delayed-inode.c
@@ -1760,6 +1760,7 @@ static void fill_stack_inode_item(struct btrfs_trans_handle *trans,
int btrfs_fill_inode(struct inode *inode, u32 *rdev)
{
+ struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
struct btrfs_delayed_node *delayed_node;
struct btrfs_inode_item *inode_item;
@@ -1779,6 +1780,8 @@ int btrfs_fill_inode(struct inode *inode, u32 *rdev)
i_uid_write(inode, btrfs_stack_inode_uid(inode_item));
i_gid_write(inode, btrfs_stack_inode_gid(inode_item));
btrfs_i_size_write(BTRFS_I(inode), btrfs_stack_inode_size(inode_item));
+ btrfs_inode_set_file_extent_range(BTRFS_I(inode), 0,
+ round_up(i_size_read(inode), fs_info->sectorsize));
inode->i_mode = btrfs_stack_inode_mode(inode_item);
set_nlink(inode, btrfs_stack_inode_nlink(inode_item));
inode_set_bytes(inode, btrfs_stack_inode_nbytes(inode_item));
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index a16da274c9aa..200eb14b6721 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -2486,6 +2486,11 @@ static int btrfs_insert_clone_extent(struct btrfs_trans_handle *trans,
btrfs_mark_buffer_dirty(leaf);
btrfs_release_path(path);
+ ret = btrfs_inode_set_file_extent_range(BTRFS_I(inode),
+ clone_info->file_offset, clone_len);
+ if (ret)
+ return ret;
+
/* If it's a hole, nothing more needs to be done. */
if (clone_info->disk_offset == 0)
return 0;
@@ -2596,6 +2601,24 @@ int btrfs_punch_hole_range(struct inode *inode, struct btrfs_path *path,
btrfs_abort_transaction(trans, ret);
break;
}
+ } else if (!clone_info && cur_offset < drop_end) {
+ /*
+ * We are past the i_size here, but since we didn't
+ * insert holes we need to clear the mapped area so we
+ * know to not set disk_i_size in this area until a new
+ * file extent is inserted here.
+ */
+ ret = btrfs_inode_clear_file_extent_range(BTRFS_I(inode),
+ cur_offset, drop_end - cur_offset);
+ if (ret) {
+ /*
+ * We couldn't clear our area, so we could
+ * presumably adjust up and corrupt the fs, so
+ * we need to abort.
+ */
+ btrfs_abort_transaction(trans, ret);
+ break;
+ }
}
if (clone_info && drop_end > clone_info->file_offset) {
@@ -2686,6 +2709,15 @@ int btrfs_punch_hole_range(struct inode *inode, struct btrfs_path *path,
btrfs_abort_transaction(trans, ret);
goto out_trans;
}
+ } else if (!clone_info && cur_offset < drop_end) {
+ /* See the comment in the loop above for the reasoning here. */
+ ret = btrfs_inode_clear_file_extent_range(BTRFS_I(inode),
+ cur_offset, drop_end - cur_offset);
+ if (ret) {
+ btrfs_abort_transaction(trans, ret);
+ goto out_trans;
+ }
+
}
if (clone_info) {
ret = btrfs_insert_clone_extent(trans, inode, path, clone_info,
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index c481450dc76e..2865de52dc22 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -241,6 +241,15 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans,
btrfs_mark_buffer_dirty(leaf);
btrfs_release_path(path);
+ /*
+ * We align size to sectorsize for inline extents just for simplicity
+ * sake.
+ */
+ size = ALIGN(size, root->fs_info->sectorsize);
+ ret = btrfs_inode_set_file_extent_range(BTRFS_I(inode), start, size);
+ if (ret)
+ goto fail;
+
/*
* we're an inline extent, so nobody can
* extend the file past i_size without locking
@@ -2446,6 +2455,11 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
ins.offset = disk_num_bytes;
ins.type = BTRFS_EXTENT_ITEM_KEY;
+ ret = btrfs_inode_set_file_extent_range(BTRFS_I(inode), file_pos,
+ ram_bytes);
+ if (ret)
+ goto out;
+
/*
* Release the reserved range from inode dirty range map, as it is
* already moved into delayed_ref_head
@@ -4160,6 +4174,8 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
}
while (1) {
+ u64 clear_start = 0, clear_len = 0;
+
fi = NULL;
leaf = path->nodes[0];
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
@@ -4210,6 +4226,8 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
u64 num_dec;
+
+ clear_start = found_key.offset;
extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
if (!del_item) {
u64 orig_num_bytes =
@@ -4217,6 +4235,7 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
extent_num_bytes = ALIGN(new_size -
found_key.offset,
fs_info->sectorsize);
+ clear_start = ALIGN(new_size, fs_info->sectorsize);
btrfs_set_file_extent_num_bytes(leaf, fi,
extent_num_bytes);
num_dec = (orig_num_bytes -
@@ -4242,6 +4261,7 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
inode_sub_bytes(inode, num_dec);
}
}
+ clear_len = num_dec;
} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
/*
* we can't truncate inline items that have had
@@ -4263,12 +4283,33 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
*/
ret = NEED_TRUNCATE_BLOCK;
break;
+ } else {
+ /*
+ * Inline extents are special, we just treat
+ * them as a full sector worth in the file
+ * extent tree just for simplicity sake.
+ */
+ clear_len = fs_info->sectorsize;
}
if (test_bit(BTRFS_ROOT_REF_COWS, &root->state))
inode_sub_bytes(inode, item_end + 1 - new_size);
}
delete:
+ /*
+ * We use btrfs_truncate_inode_items() to clean up log trees for
+ * multiple fsyncs, and in this case we don't want to clear the
+ * file extent range because it's just the log.
+ */
+ if (root == BTRFS_I(inode)->root) {
+ ret = btrfs_inode_clear_file_extent_range(BTRFS_I(inode),
+ clear_start, clear_len);
+ if (ret) {
+ btrfs_abort_transaction(trans, ret);
+ break;
+ }
+ }
+
if (del_item)
last_size = found_key.offset;
else
@@ -4591,14 +4632,21 @@ int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size)
}
last_byte = min(extent_map_end(em), block_end);
last_byte = ALIGN(last_byte, fs_info->sectorsize);
+ hole_size = last_byte - cur_offset;
+
if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) {
struct extent_map *hole_em;
- hole_size = last_byte - cur_offset;
err = maybe_insert_hole(root, inode, cur_offset,
hole_size);
if (err)
break;
+
+ err = btrfs_inode_set_file_extent_range(BTRFS_I(inode),
+ cur_offset, hole_size);
+ if (err)
+ break;
+
btrfs_drop_extent_cache(BTRFS_I(inode), cur_offset,
cur_offset + hole_size - 1, 0);
hole_em = alloc_extent_map();
@@ -4630,6 +4678,11 @@ int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size)
hole_size - 1, 0);
}
free_extent_map(hole_em);
+ } else {
+ err = btrfs_inode_set_file_extent_range(BTRFS_I(inode),
+ cur_offset, hole_size);
+ if (err)
+ break;
}
next:
free_extent_map(em);
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 7dd7552f53a4..d3b101f2816a 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -830,6 +830,11 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
goto out;
}
+ ret = btrfs_inode_set_file_extent_range(BTRFS_I(inode), start,
+ extent_end - start);
+ if (ret)
+ goto out;
+
inode_add_bytes(inode, nbytes);
update_inode:
ret = btrfs_update_inode(trans, root, inode);
--
2.26.2