[PATCH] Fix IO error reporting on fsync()

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 introducing a pointer to associated address space into
the buffer_head.  When a 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>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

diff --git a/fs/buffer.c b/fs/buffer.c
index f65ef8821c7364a718b64948166340a2fc418bdd..35527dca1dbcc415d95d76780537c577808df9ff 100644 (file)
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -452,6 +452,7 @@ static void end_buffer_async_write(struct buffer_head *bh, int uptodate)
                               bdevname(bh->b_bdev, b));
                }
                set_bit(AS_EIO, &page->mapping->flags);
+               set_buffer_write_io_error(bh);
                clear_buffer_uptodate(bh);
                SetPageError(page);
        }
@@ -571,6 +572,10 @@ 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)
@@ -669,6 +674,7 @@ void mark_buffer_dirty_inode(struct buffer_head *bh, struct inode *inode)
                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);
        }
 }
@@ -765,7 +771,7 @@ static int fsync_buffers_list(spinlock_t *lock, struct list_head *list)
        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)) {
@@ -786,7 +792,7 @@ static int fsync_buffers_list(spinlock_t *lock, struct list_head *list)
 
        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);
@@ -1167,6 +1173,7 @@ 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 --git a/include/linux/buffer_head.h b/include/linux/buffer_head.h
index 131ffd37e716fb7ac4a386eaea9ff58c26bf639d..5d9fb0e94156235eca46e199352933fc638b9596 100644 (file)
--- a/include/linux/buffer_head.h
+++ b/include/linux/buffer_head.h
@@ -69,6 +69,8 @@ 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 */
 };