On 08/16/2011 10:04 PM, Eric Sandeen wrote:

On 8/16/11 2:06 AM, Tao Ma wrote:

quoted

From: Tao Ma <redacted>

EXT4_IO_END_UNWRITTEN flag set and the increase of i_aiodio_unwritten should
be done simultaneously since ext4_end_io_nolock always clear the flag and
decrease the counter in the same time.

We have found some bugs that the flag is set while leaving i_aiodio_unwritten
unchanged. So this patch just tries to creat a helper function to wrap them
to avoid any future bug. The idea is inspired by Eric.

Although I like it a little less now that I see it in practice, sorry.   ;)

This:
                /*
                 * Flag the inode(non aio case) or end_io struct (aio case)
                 * that this IO needs to conversion to written when IO is
                 * completed
                 */
		if (!ext4_set_io_unwritten_flag(inode, io))
			ext4_set_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);

just reads weirdly to me.

yeah, this also looks a little bit strange. Another possible way is that
we just wrap the set of the flag and the inc of aiodio_unwritten while
leave the check there. It should look more natural and good for the reader.

It encapsulates all the steps, but it really isn't intuitive to read, I think.

"If we (can't? don't?) set the io unwritten flag, set the dio unwritten flag
on the inode."

I wonder if changing "io" to "endio" would make it a little more obvious.

I'm not sure, maybe it's ok.  What do others think?

I also wonder about this, but I get lost in this code.  It was originally:

                if (io && !(io->flag & EXT4_IO_END_UNWRITTEN)) {
                        io->flag = EXT4_IO_END_UNWRITTEN;
                        atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten);
                } else
                        ext4_set_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);

That means that if it is AIO ("if (io)"), but the flag is already set
(io->flag & EXT4_IO_END_UNWRITTEN), we don't re-set the flag or increase
the count, which is fine - but is it then correct to set the state on the
inode?

I will check it later.

Thanks
Tao

thanks,
-Eric

quoted

Cc: Eric Sandeen <redacted>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Tao Ma <redacted>
---
Ted,
	This patch is based on the patch with subject "ext4: Resolve the hang
of direct i/o read in handling EXT4_IO_END_UNWRITTEN.". I meant it for the next
merge window, but if you think it is also ok for 3.1, go ahead.

 fs/ext4/ext4.h    |   11 +++++++++++
 fs/ext4/extents.c |   10 ++--------
 fs/ext4/inode.c   |    5 +----
 fs/ext4/page-io.c |    6 ++----
 4 files changed, 16 insertions(+), 16 deletions(-)

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index e717dfd..514f670 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h

@@ -1247,6 +1247,17 @@ static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
 		 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
 }
 
+static inline int ext4_set_io_unwritten_flag(struct inode *inode,
+					     struct ext4_io_end *io_end)
+{
+	if (io_end && !(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
+		io_end->flag |= EXT4_IO_END_UNWRITTEN;
+		atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten);
+		return 1;
+	}
+
+	return 0;
+}
 /*
  * Inode dynamic state flags
  */

diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 57cf568..c0f8655 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c

@@ -3190,10 +3190,7 @@ ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode,
 		 * that this IO needs to conversion to written when IO is
 		 * completed
 		 */
-		if (io && !(io->flag & EXT4_IO_END_UNWRITTEN)) {
-			io->flag = EXT4_IO_END_UNWRITTEN;
-			atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten);
-		} else
+		if (!ext4_set_io_unwritten_flag(inode, io))
 			ext4_set_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);
 		if (ext4_should_dioread_nolock(inode))
 			map->m_flags |= EXT4_MAP_UNINIT;

@@ -3572,10 +3569,7 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
 		 * that we need to perform conversion when IO is done.
 		 */
 		if ((flags & EXT4_GET_BLOCKS_PRE_IO)) {
-			if (io && !(io->flag & EXT4_IO_END_UNWRITTEN)) {
-				io->flag = EXT4_IO_END_UNWRITTEN;
-				atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten);
-			} else
+			if (!ext4_set_io_unwritten_flag(inode, io))
 				ext4_set_inode_state(inode,
 						     EXT4_STATE_DIO_UNWRITTEN);
 		}

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 40c0b10..128493b 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c

@@ -2673,10 +2673,7 @@ static void ext4_end_io_buffer_write(struct buffer_head *bh, int uptodate)
 	 * but being more careful is always safe for the future change.
 	 */
 	inode = io_end->inode;
-	if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
-		io_end->flag |= EXT4_IO_END_UNWRITTEN;
-		atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten);
-	}
+	ext4_set_io_unwritten_flag(inode, io_end);
 
 	/* Add the io_end to per-inode completed io list*/
 	spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);

diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c
index 78839af..edd2f7b 100644
--- a/fs/ext4/page-io.c
+++ b/fs/ext4/page-io.c

@@ -334,10 +334,8 @@ submit_and_retry:
 	if ((io_end->num_io_pages >= MAX_IO_PAGES) &&
 	    (io_end->pages[io_end->num_io_pages-1] != io_page))
 		goto submit_and_retry;
-	if (buffer_uninit(bh) && !(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
-		io_end->flag |= EXT4_IO_END_UNWRITTEN;
-		atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten);
-	}
+	if (buffer_uninit(bh))
+		ext4_set_io_unwritten_flag(inode, io_end);
 	io->io_end->size += bh->b_size;
 	io->io_next_block++;
 	ret = bio_add_page(io->io_bio, bh->b_page, bh->b_size, bh_offset(bh));