Re: [PATCH 2/5] vfs: use writeback lists to provide lazytime one day timeout
From: Jan Kara <jack@suse.cz>
Date: 2014-11-28 17:20:46
Also in:
linux-fsdevel
On Fri 28-11-14 01:00:08, Ted Tso wrote:
Queue inodes with dirty timestamps for writeback 24 hours after they were initially dirtied.
Oh I see, this patch should probably replace the other 2/5 patch.
quoted hunk ↗ jump to hunk
Signed-off-by: Theodore Ts'o <tytso@mit.edu> --- fs/fs-writeback.c | 29 +++++++++++++++++++++++------ 1 file changed, 23 insertions(+), 6 deletions(-)diff --git a/fs/fs-writeback.c b/fs/fs-writeback.c index 3d562e2..bd76590 100644 --- a/fs/fs-writeback.c +++ b/fs/fs-writeback.c@@ -253,7 +253,7 @@ static bool inode_dirtied_after(struct inode *inode, unsigned long t) */ static int move_expired_inodes(struct list_head *delaying_queue, struct list_head *dispatch_queue, - struct wb_writeback_work *work) + unsigned long *older_than_this) { LIST_HEAD(tmp); struct list_head *pos, *node;@@ -264,8 +264,8 @@ static int move_expired_inodes(struct list_head *delaying_queue, while (!list_empty(delaying_queue)) { inode = wb_inode(delaying_queue->prev); - if (work->older_than_this && - inode_dirtied_after(inode, *work->older_than_this)) + if (older_than_this && + inode_dirtied_after(inode, *older_than_this)) break; list_move(&inode->i_wb_list, &tmp); moved++;@@ -309,9 +309,14 @@ out: static void queue_io(struct bdi_writeback *wb, struct wb_writeback_work *work) { int moved; + unsigned long one_day_later = jiffies + (HZ * 86400); + assert_spin_locked(&wb->list_lock); list_splice_init(&wb->b_more_io, &wb->b_io); - moved = move_expired_inodes(&wb->b_dirty, &wb->b_io, work); + moved = move_expired_inodes(&wb->b_dirty, &wb->b_io, + work->older_than_this); + moved += move_expired_inodes(&wb->b_dirty_time, &wb->b_io, + &one_day_later);
I'd change this to:
if (work->sync_mode == WB_SYNC_ALL) {
moved += move_expired_inodes(&wb->b_dirty_time, &wb->b_io,
work->older_than_this);
} else {
moved += move_expired_inodes(&wb->b_dirty_time, &wb->b_io,
&one_day_later);
}
That way you can get rid of special sync handling. Also just fold this
patch into 1/5...
quoted hunk ↗ jump to hunk
trace_writeback_queue_io(wb, work, moved); }@@ -679,6 +684,17 @@ static long writeback_sb_inodes(struct super_block *sb, inode->i_state |= I_SYNC; spin_unlock(&inode->i_lock); + /* + * If the inode is marked dirty time but is not dirty, + * then at last for ext3 and ext4 we need to call + * mark_inode_dirty_sync in order to get the inode + * timestamp transferred to the on disk inode, since + * write_inode is a no-op for those file systems. + */ + if ((inode->i_state & I_DIRTY_TIME) && + ((inode->i_state & (I_DIRTY_SYNC | I_DIRTY_DATASYNC)) == 0)) + mark_inode_dirty_sync(inode); +
This isn't necessary - you already handle this in __writeback_single_inode(). Or am I missing something? Honza -- Jan Kara [off-list ref] SUSE Labs, CR