summaryrefslogtreecommitdiff
path: root/fs/bcachefs/io_write.c
diff options
context:
space:
mode:
authorKent Overstreet <kent.overstreet@linux.dev>2023-11-04 20:49:31 +0300
committerKent Overstreet <kent.overstreet@linux.dev>2023-11-05 05:19:13 +0300
commit9fcdd23b6eea3f04475cd9cfb4497f6e26906061 (patch)
treee17bdb92371b1c893b47d669faf4f41a070b0ebb /fs/bcachefs/io_write.c
parentd3c7727bb9269c7f7a2f17ef76b9e5c9b8cc8863 (diff)
downloadlinux-9fcdd23b6eea3f04475cd9cfb4497f6e26906061.tar.xz
bcachefs: Add a comment for BTREE_INSERT_NOJOURNAL usage
BTREE_INSERT_NOJOURNAL is primarily used for a performance optimization related to inode updates and fsync - document it. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Diffstat (limited to 'fs/bcachefs/io_write.c')
-rw-r--r--fs/bcachefs/io_write.c11
1 files changed, 11 insertions, 0 deletions
diff --git a/fs/bcachefs/io_write.c b/fs/bcachefs/io_write.c
index 613f38436640..4eb0eda723ab 100644
--- a/fs/bcachefs/io_write.c
+++ b/fs/bcachefs/io_write.c
@@ -202,6 +202,17 @@ static inline int bch2_extent_update_i_size_sectors(struct btree_trans *trans,
struct btree_iter iter;
struct bkey_i *k;
struct bkey_i_inode_v3 *inode;
+ /*
+ * Crazy performance optimization:
+ * Every extent update needs to also update the inode: the inode trigger
+ * will set bi->journal_seq to the journal sequence number of this
+ * transaction - for fsync.
+ *
+ * But if that's the only reason we're updating the inode (we're not
+ * updating bi_size or bi_sectors), then we don't need the inode update
+ * to be journalled - if we crash, the bi_journal_seq update will be
+ * lost, but that's fine.
+ */
unsigned inode_update_flags = BTREE_UPDATE_NOJOURNAL;
int ret;