summaryrefslogtreecommitdiff
path: root/fs/btrfs/file.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/btrfs/file.c')
-rw-r--r--fs/btrfs/file.c435
1 files changed, 288 insertions, 147 deletions
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 58a18ed11546..8fe4eb7e5045 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -537,8 +537,8 @@ int btrfs_dirty_pages(struct inode *inode, struct page **pages,
* we can set things up properly
*/
clear_extent_bit(&BTRFS_I(inode)->io_tree, start_pos, end_of_last_block,
- EXTENT_DIRTY | EXTENT_DELALLOC |
- EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, cached);
+ EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
+ 0, 0, cached);
if (!btrfs_is_free_space_inode(BTRFS_I(inode))) {
if (start_pos >= isize &&
@@ -559,7 +559,7 @@ int btrfs_dirty_pages(struct inode *inode, struct page **pages,
}
err = btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block,
- extra_bits, cached, 0);
+ extra_bits, cached);
if (err)
return err;
@@ -1882,10 +1882,10 @@ static ssize_t btrfs_file_write_iter(struct kiocb *iocb,
u64 start_pos;
u64 end_pos;
ssize_t num_written = 0;
- bool sync = (file->f_flags & O_DSYNC) || IS_SYNC(file->f_mapping->host);
+ const bool sync = iocb->ki_flags & IOCB_DSYNC;
ssize_t err;
loff_t pos;
- size_t count = iov_iter_count(from);
+ size_t count;
loff_t oldsize;
int clean_page = 0;
@@ -1906,6 +1906,7 @@ static ssize_t btrfs_file_write_iter(struct kiocb *iocb,
}
pos = iocb->ki_pos;
+ count = iov_iter_count(from);
if (iocb->ki_flags & IOCB_NOWAIT) {
/*
* We will allocate space in case nodatacow is not set,
@@ -2439,27 +2440,286 @@ static int btrfs_punch_hole_lock_range(struct inode *inode,
return 0;
}
+static int btrfs_insert_clone_extent(struct btrfs_trans_handle *trans,
+ struct inode *inode,
+ struct btrfs_path *path,
+ struct btrfs_clone_extent_info *clone_info,
+ const u64 clone_len)
+{
+ struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct btrfs_file_extent_item *extent;
+ struct extent_buffer *leaf;
+ struct btrfs_key key;
+ int slot;
+ struct btrfs_ref ref = { 0 };
+ u64 ref_offset;
+ int ret;
+
+ if (clone_len == 0)
+ return 0;
+
+ if (clone_info->disk_offset == 0 &&
+ btrfs_fs_incompat(fs_info, NO_HOLES))
+ return 0;
+
+ key.objectid = btrfs_ino(BTRFS_I(inode));
+ key.type = BTRFS_EXTENT_DATA_KEY;
+ key.offset = clone_info->file_offset;
+ ret = btrfs_insert_empty_item(trans, root, path, &key,
+ clone_info->item_size);
+ if (ret)
+ return ret;
+ leaf = path->nodes[0];
+ slot = path->slots[0];
+ write_extent_buffer(leaf, clone_info->extent_buf,
+ btrfs_item_ptr_offset(leaf, slot),
+ clone_info->item_size);
+ extent = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
+ btrfs_set_file_extent_offset(leaf, extent, clone_info->data_offset);
+ btrfs_set_file_extent_num_bytes(leaf, extent, clone_len);
+ btrfs_mark_buffer_dirty(leaf);
+ btrfs_release_path(path);
+
+ /* If it's a hole, nothing more needs to be done. */
+ if (clone_info->disk_offset == 0)
+ return 0;
+
+ inode_add_bytes(inode, clone_len);
+ btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF,
+ clone_info->disk_offset,
+ clone_info->disk_len, 0);
+ ref_offset = clone_info->file_offset - clone_info->data_offset;
+ btrfs_init_data_ref(&ref, root->root_key.objectid,
+ btrfs_ino(BTRFS_I(inode)), ref_offset);
+ ret = btrfs_inc_extent_ref(trans, &ref);
+
+ return ret;
+}
+
+/*
+ * The respective range must have been previously locked, as well as the inode.
+ * The end offset is inclusive (last byte of the range).
+ * @clone_info is NULL for fallocate's hole punching and non-NULL for extent
+ * cloning.
+ * When cloning, we don't want to end up in a state where we dropped extents
+ * without inserting a new one, so we must abort the transaction to avoid a
+ * corruption.
+ */
+int btrfs_punch_hole_range(struct inode *inode, struct btrfs_path *path,
+ const u64 start, const u64 end,
+ struct btrfs_clone_extent_info *clone_info,
+ struct btrfs_trans_handle **trans_out)
+{
+ struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
+ u64 min_size = btrfs_calc_insert_metadata_size(fs_info, 1);
+ u64 ino_size = round_up(inode->i_size, fs_info->sectorsize);
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct btrfs_trans_handle *trans = NULL;
+ struct btrfs_block_rsv *rsv;
+ unsigned int rsv_count;
+ u64 cur_offset;
+ u64 drop_end;
+ u64 len = end - start;
+ int ret = 0;
+
+ if (end <= start)
+ return -EINVAL;
+
+ rsv = btrfs_alloc_block_rsv(fs_info, BTRFS_BLOCK_RSV_TEMP);
+ if (!rsv) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ rsv->size = btrfs_calc_insert_metadata_size(fs_info, 1);
+ rsv->failfast = 1;
+
+ /*
+ * 1 - update the inode
+ * 1 - removing the extents in the range
+ * 1 - adding the hole extent if no_holes isn't set or if we are cloning
+ * an extent
+ */
+ if (!btrfs_fs_incompat(fs_info, NO_HOLES) || clone_info)
+ rsv_count = 3;
+ else
+ rsv_count = 2;
+
+ trans = btrfs_start_transaction(root, rsv_count);
+ if (IS_ERR(trans)) {
+ ret = PTR_ERR(trans);
+ trans = NULL;
+ goto out_free;
+ }
+
+ ret = btrfs_block_rsv_migrate(&fs_info->trans_block_rsv, rsv,
+ min_size, false);
+ BUG_ON(ret);
+ trans->block_rsv = rsv;
+
+ cur_offset = start;
+ while (cur_offset < end) {
+ ret = __btrfs_drop_extents(trans, root, inode, path,
+ cur_offset, end + 1, &drop_end,
+ 1, 0, 0, NULL);
+ if (ret != -ENOSPC) {
+ /*
+ * When cloning we want to avoid transaction aborts when
+ * nothing was done and we are attempting to clone parts
+ * of inline extents, in such cases -EOPNOTSUPP is
+ * returned by __btrfs_drop_extents() without having
+ * changed anything in the file.
+ */
+ if (clone_info && ret && ret != -EOPNOTSUPP)
+ btrfs_abort_transaction(trans, ret);
+ break;
+ }
+
+ trans->block_rsv = &fs_info->trans_block_rsv;
+
+ if (!clone_info && cur_offset < drop_end &&
+ cur_offset < ino_size) {
+ ret = fill_holes(trans, BTRFS_I(inode), path,
+ cur_offset, drop_end);
+ if (ret) {
+ /*
+ * If we failed then we didn't insert our hole
+ * entries for the area we dropped, so now the
+ * fs is corrupted, so we must abort the
+ * transaction.
+ */
+ btrfs_abort_transaction(trans, ret);
+ break;
+ }
+ }
+
+ if (clone_info) {
+ u64 clone_len = drop_end - cur_offset;
+
+ ret = btrfs_insert_clone_extent(trans, inode, path,
+ clone_info, clone_len);
+ if (ret) {
+ btrfs_abort_transaction(trans, ret);
+ break;
+ }
+ clone_info->data_len -= clone_len;
+ clone_info->data_offset += clone_len;
+ clone_info->file_offset += clone_len;
+ }
+
+ cur_offset = drop_end;
+
+ ret = btrfs_update_inode(trans, root, inode);
+ if (ret)
+ break;
+
+ btrfs_end_transaction(trans);
+ btrfs_btree_balance_dirty(fs_info);
+
+ trans = btrfs_start_transaction(root, rsv_count);
+ if (IS_ERR(trans)) {
+ ret = PTR_ERR(trans);
+ trans = NULL;
+ break;
+ }
+
+ ret = btrfs_block_rsv_migrate(&fs_info->trans_block_rsv,
+ rsv, min_size, false);
+ BUG_ON(ret); /* shouldn't happen */
+ trans->block_rsv = rsv;
+
+ if (!clone_info) {
+ ret = find_first_non_hole(inode, &cur_offset, &len);
+ if (unlikely(ret < 0))
+ break;
+ if (ret && !len) {
+ ret = 0;
+ break;
+ }
+ }
+ }
+
+ /*
+ * If we were cloning, force the next fsync to be a full one since we
+ * we replaced (or just dropped in the case of cloning holes when
+ * NO_HOLES is enabled) extents and extent maps.
+ * This is for the sake of simplicity, and cloning into files larger
+ * than 16Mb would force the full fsync any way (when
+ * try_release_extent_mapping() is invoked during page cache truncation.
+ */
+ if (clone_info)
+ set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+ &BTRFS_I(inode)->runtime_flags);
+
+ if (ret)
+ goto out_trans;
+
+ trans->block_rsv = &fs_info->trans_block_rsv;
+ /*
+ * If we are using the NO_HOLES feature we might have had already an
+ * hole that overlaps a part of the region [lockstart, lockend] and
+ * ends at (or beyond) lockend. Since we have no file extent items to
+ * represent holes, drop_end can be less than lockend and so we must
+ * make sure we have an extent map representing the existing hole (the
+ * call to __btrfs_drop_extents() might have dropped the existing extent
+ * map representing the existing hole), otherwise the fast fsync path
+ * will not record the existence of the hole region
+ * [existing_hole_start, lockend].
+ */
+ if (drop_end <= end)
+ drop_end = end + 1;
+ /*
+ * Don't insert file hole extent item if it's for a range beyond eof
+ * (because it's useless) or if it represents a 0 bytes range (when
+ * cur_offset == drop_end).
+ */
+ if (!clone_info && cur_offset < ino_size && cur_offset < drop_end) {
+ ret = fill_holes(trans, BTRFS_I(inode), path,
+ cur_offset, drop_end);
+ if (ret) {
+ /* Same comment as above. */
+ btrfs_abort_transaction(trans, ret);
+ goto out_trans;
+ }
+ }
+ if (clone_info) {
+ ret = btrfs_insert_clone_extent(trans, inode, path, clone_info,
+ clone_info->data_len);
+ if (ret) {
+ btrfs_abort_transaction(trans, ret);
+ goto out_trans;
+ }
+ }
+
+out_trans:
+ if (!trans)
+ goto out_free;
+
+ trans->block_rsv = &fs_info->trans_block_rsv;
+ if (ret)
+ btrfs_end_transaction(trans);
+ else
+ *trans_out = trans;
+out_free:
+ btrfs_free_block_rsv(fs_info, rsv);
+out:
+ return ret;
+}
+
static int btrfs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
{
struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
struct btrfs_root *root = BTRFS_I(inode)->root;
struct extent_state *cached_state = NULL;
struct btrfs_path *path;
- struct btrfs_block_rsv *rsv;
- struct btrfs_trans_handle *trans;
+ struct btrfs_trans_handle *trans = NULL;
u64 lockstart;
u64 lockend;
u64 tail_start;
u64 tail_len;
u64 orig_start = offset;
- u64 cur_offset;
- u64 min_size = btrfs_calc_trans_metadata_size(fs_info, 1);
- u64 drop_end;
int ret = 0;
- int err = 0;
- unsigned int rsv_count;
bool same_block;
- bool no_holes = btrfs_fs_incompat(fs_info, NO_HOLES);
u64 ino_size;
bool truncated_block = false;
bool updated_inode = false;
@@ -2566,145 +2826,24 @@ static int btrfs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
goto out;
}
- rsv = btrfs_alloc_block_rsv(fs_info, BTRFS_BLOCK_RSV_TEMP);
- if (!rsv) {
- ret = -ENOMEM;
- goto out_free;
- }
- rsv->size = btrfs_calc_trans_metadata_size(fs_info, 1);
- rsv->failfast = 1;
-
- /*
- * 1 - update the inode
- * 1 - removing the extents in the range
- * 1 - adding the hole extent if no_holes isn't set
- */
- rsv_count = no_holes ? 2 : 3;
- trans = btrfs_start_transaction(root, rsv_count);
- if (IS_ERR(trans)) {
- err = PTR_ERR(trans);
- goto out_free;
- }
-
- ret = btrfs_block_rsv_migrate(&fs_info->trans_block_rsv, rsv,
- min_size, false);
- BUG_ON(ret);
- trans->block_rsv = rsv;
-
- cur_offset = lockstart;
- len = lockend - cur_offset;
- while (cur_offset < lockend) {
- ret = __btrfs_drop_extents(trans, root, inode, path,
- cur_offset, lockend + 1,
- &drop_end, 1, 0, 0, NULL);
- if (ret != -ENOSPC)
- break;
-
- trans->block_rsv = &fs_info->trans_block_rsv;
-
- if (cur_offset < drop_end && cur_offset < ino_size) {
- ret = fill_holes(trans, BTRFS_I(inode), path,
- cur_offset, drop_end);
- if (ret) {
- /*
- * If we failed then we didn't insert our hole
- * entries for the area we dropped, so now the
- * fs is corrupted, so we must abort the
- * transaction.
- */
- btrfs_abort_transaction(trans, ret);
- err = ret;
- break;
- }
- }
-
- cur_offset = drop_end;
-
- ret = btrfs_update_inode(trans, root, inode);
- if (ret) {
- err = ret;
- break;
- }
-
- btrfs_end_transaction(trans);
- btrfs_btree_balance_dirty(fs_info);
-
- trans = btrfs_start_transaction(root, rsv_count);
- if (IS_ERR(trans)) {
- ret = PTR_ERR(trans);
- trans = NULL;
- break;
- }
-
- ret = btrfs_block_rsv_migrate(&fs_info->trans_block_rsv,
- rsv, min_size, false);
- BUG_ON(ret); /* shouldn't happen */
- trans->block_rsv = rsv;
-
- ret = find_first_non_hole(inode, &cur_offset, &len);
- if (unlikely(ret < 0))
- break;
- if (ret && !len) {
- ret = 0;
- break;
- }
- }
-
- if (ret) {
- err = ret;
- goto out_trans;
- }
-
- trans->block_rsv = &fs_info->trans_block_rsv;
- /*
- * If we are using the NO_HOLES feature we might have had already an
- * hole that overlaps a part of the region [lockstart, lockend] and
- * ends at (or beyond) lockend. Since we have no file extent items to
- * represent holes, drop_end can be less than lockend and so we must
- * make sure we have an extent map representing the existing hole (the
- * call to __btrfs_drop_extents() might have dropped the existing extent
- * map representing the existing hole), otherwise the fast fsync path
- * will not record the existence of the hole region
- * [existing_hole_start, lockend].
- */
- if (drop_end <= lockend)
- drop_end = lockend + 1;
- /*
- * Don't insert file hole extent item if it's for a range beyond eof
- * (because it's useless) or if it represents a 0 bytes range (when
- * cur_offset == drop_end).
- */
- if (cur_offset < ino_size && cur_offset < drop_end) {
- ret = fill_holes(trans, BTRFS_I(inode), path,
- cur_offset, drop_end);
- if (ret) {
- /* Same comment as above. */
- btrfs_abort_transaction(trans, ret);
- err = ret;
- goto out_trans;
- }
- }
-
-out_trans:
- if (!trans)
- goto out_free;
+ ret = btrfs_punch_hole_range(inode, path, lockstart, lockend, NULL,
+ &trans);
+ btrfs_free_path(path);
+ if (ret)
+ goto out;
+ ASSERT(trans != NULL);
inode_inc_iversion(inode);
inode->i_mtime = inode->i_ctime = current_time(inode);
-
- trans->block_rsv = &fs_info->trans_block_rsv;
ret = btrfs_update_inode(trans, root, inode);
updated_inode = true;
btrfs_end_transaction(trans);
btrfs_btree_balance_dirty(fs_info);
-out_free:
- btrfs_free_path(path);
- btrfs_free_block_rsv(fs_info, rsv);
out:
unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
&cached_state);
out_only_mutex:
- if (!updated_inode && truncated_block && !ret && !err) {
+ if (!updated_inode && truncated_block && !ret) {
/*
* If we only end up zeroing part of a page, we still need to
* update the inode item, so that all the time fields are
@@ -2719,16 +2858,18 @@ out_only_mutex:
inode->i_ctime = now;
trans = btrfs_start_transaction(root, 1);
if (IS_ERR(trans)) {
- err = PTR_ERR(trans);
+ ret = PTR_ERR(trans);
} else {
- err = btrfs_update_inode(trans, root, inode);
- ret = btrfs_end_transaction(trans);
+ int ret2;
+
+ ret = btrfs_update_inode(trans, root, inode);
+ ret2 = btrfs_end_transaction(trans);
+ if (!ret)
+ ret = ret2;
}
}
inode_unlock(inode);
- if (ret && !err)
- err = ret;
- return err;
+ return ret;
}
/* Helper structure to record which range is already reserved */