[PATCH 2/2] Fix IO error reporting on fsync() (2nd try)

[Jan Kara <jack@suse.cz>]

When IO error happens on metadata buffer, buffer is freed from memory and later
fsync() is called, filesystems like ext2 fail to report EIO.  We solve the
problem by introdusing a pointer to associated address space to buffer_head.
When buffer is removed from a list of metadata buffers associated with an
address space, IO error is transferred from the buffer to the address space,
so that fsync can later report it.

Signed-off-by: Jan Kara <jack@suse.cz>

diff -rupX /home/jack/.kerndiffexclude linux-2.6.18-1-mark_error_buffer/fs/buffer.c linux-2.6.18-2-fsync_bh_fix/fs/buffer.c
--- linux-2.6.18-1-mark_error_buffer/fs/buffer.c	2006-10-06 13:05:29.000000000 +0200
+++ linux-2.6.18-2-fsync_bh_fix/fs/buffer.c	2006-10-10 17:00:04.000000000 +0200
@@ -709,6 +709,10 @@ EXPORT_SYMBOL(mark_buffer_async_write);
 static inline void __remove_assoc_queue(struct buffer_head *bh)
 {
 	list_del_init(&bh->b_assoc_buffers);
+	WARN_ON(!bh->b_assoc_map);
+	if (buffer_write_io_error(bh))
+		set_bit(AS_EIO, &bh->b_assoc_map->flags);
+	bh->b_assoc_map = NULL;
 }
 
 int inode_has_buffers(struct inode *inode)
@@ -807,6 +811,7 @@ void mark_buffer_dirty_inode(struct buff
 		spin_lock(&buffer_mapping->private_lock);
 		list_move_tail(&bh->b_assoc_buffers,
 				&mapping->private_list);
+		bh->b_assoc_map = mapping;
 		spin_unlock(&buffer_mapping->private_lock);
 	}
 }
@@ -900,7 +905,7 @@ static int fsync_buffers_list(spinlock_t
 	spin_lock(lock);
 	while (!list_empty(list)) {
 		bh = BH_ENTRY(list->next);
-		list_del_init(&bh->b_assoc_buffers);
+		__remove_assoc_queue(bh);
 		if (buffer_dirty(bh) || buffer_locked(bh)) {
 			list_add(&bh->b_assoc_buffers, &tmp);
 			if (buffer_dirty(bh)) {
@@ -921,7 +926,7 @@ static int fsync_buffers_list(spinlock_t
 
 	while (!list_empty(&tmp)) {
 		bh = BH_ENTRY(tmp.prev);
-		__remove_assoc_queue(bh);
+		list_del_init(&bh->b_assoc_buffers);
 		get_bh(bh);
 		spin_unlock(lock);
 		wait_on_buffer(bh);
@@ -1285,6 +1290,7 @@ void __bforget(struct buffer_head *bh)
 
 		spin_lock(&buffer_mapping->private_lock);
 		list_del_init(&bh->b_assoc_buffers);
+		bh->b_assoc_map = NULL;
 		spin_unlock(&buffer_mapping->private_lock);
 	}
 	__brelse(bh);
diff -rupX /home/jack/.kerndiffexclude linux-2.6.18-1-mark_error_buffer/include/linux/buffer_head.h linux-2.6.18-2-fsync_bh_fix/include/linux/buffer_head.h
--- linux-2.6.18-1-mark_error_buffer/include/linux/buffer_head.h	2006-09-27 13:09:04.000000000 +0200
+++ linux-2.6.18-2-fsync_bh_fix/include/linux/buffer_head.h	2006-10-10 15:37:46.000000000 +0200
@@ -67,6 +67,8 @@ struct buffer_head {
 	bh_end_io_t *b_end_io;		/* I/O completion */
  	void *b_private;		/* reserved for b_end_io */
 	struct list_head b_assoc_buffers; /* associated with another mapping */
+	struct address_space *b_assoc_map;	/* mapping this buffer is
+						   associated with */
 	atomic_t b_count;		/* users using this buffer_head */
 };

-- 
Jan Kara <jack@suse.cz>
SuSE CR Labs
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/