ocfs2: fix ip_unaligned_aio deadlock with dio work queue

In the current implementation of unaligned aio+dio, lock order behave as
follow:

in user process context:
  -> call io_submit()
    -> get i_mutex
		<== window1
      -> get ip_unaligned_aio
        -> submit direct io to block device
    -> release i_mutex
  -> io_submit() return

in dio work queue context(the work queue is created in __blockdev_direct_IO):
  -> release ip_unaligned_aio
		<== window2
    -> get i_mutex
      -> clear unwritten flag & change i_size
    -> release i_mutex

There is a limitation to the thread number of dio work queue.  256 at
default.  If all 256 thread are in the above 'window2' stage, and there
is a user process in the 'window1' stage, the system will became
deadlock.  Since the user process hold i_mutex to wait ip_unaligned_aio
lock, while there is a direct bio hold ip_unaligned_aio mutex who is
waiting for a dio work queue thread to be schedule.  But all the dio
work queue thread is waiting for i_mutex lock in 'window2'.

This case only happened in a test which send a large number(more than
256) of aio at one io_submit() call.

My design is to remove ip_unaligned_aio lock.  Change it to a sync io
instead.  Just like ip_unaligned_aio lock, serialize the unaligned aio
dio.

[akpm@linux-foundation.org: remove OCFS2_IOCB_UNALIGNED_IO, per Junxiao Bi]
Signed-off-by: Ryan Ding <ryan.ding@oracle.com>
Reviewed-by: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <joseph.qi@huawei.com>
Cc: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c
index 578ce225cf83435f34f202bad346a67d6c783419..843475793504299aea8348bf2312eede339e4846 100644 (file)
--- a/fs/ocfs2/aops.c
+++ b/fs/ocfs2/aops.c
@@ -2391,12 +2391,6 @@ static int ocfs2_dio_end_io(struct kiocb *iocb,
        /* this io's submitter should not have unlocked this before we could */
        BUG_ON(!ocfs2_iocb_is_rw_locked(iocb));
 
-       if (ocfs2_iocb_is_unaligned_aio(iocb)) {
-               ocfs2_iocb_clear_unaligned_aio(iocb);
-
-               mutex_unlock(&OCFS2_I(inode)->ip_unaligned_aio);
-       }
-
        if (private)
                ocfs2_dio_end_io_write(inode, private, offset, bytes);
 

diff --git a/fs/ocfs2/aops.h b/fs/ocfs2/aops.h
index d06b80f58f83d162ffdeb757a80e5b047fe13911..b1c9f28a57b16efb3ea7f7cbe34d43cf9e1b43b5 100644 (file)
--- a/fs/ocfs2/aops.h
+++ b/fs/ocfs2/aops.h
@@ -84,7 +84,6 @@ static inline void ocfs2_iocb_set_rw_locked(struct kiocb *iocb, int level)
 enum ocfs2_iocb_lock_bits {
        OCFS2_IOCB_RW_LOCK = 0,
        OCFS2_IOCB_RW_LOCK_LEVEL,
-       OCFS2_IOCB_UNALIGNED_IO,
        OCFS2_IOCB_NUM_LOCKS
 };
 
@@ -93,11 +92,4 @@ enum ocfs2_iocb_lock_bits {
 #define ocfs2_iocb_rw_locked_level(iocb) \
        test_bit(OCFS2_IOCB_RW_LOCK_LEVEL, (unsigned long *)&iocb->private)
 
-#define ocfs2_iocb_set_unaligned_aio(iocb) \
-       set_bit(OCFS2_IOCB_UNALIGNED_IO, (unsigned long *)&iocb->private)
-#define ocfs2_iocb_clear_unaligned_aio(iocb) \
-       clear_bit(OCFS2_IOCB_UNALIGNED_IO, (unsigned long *)&iocb->private)
-#define ocfs2_iocb_is_unaligned_aio(iocb) \
-       test_bit(OCFS2_IOCB_UNALIGNED_IO, (unsigned long *)&iocb->private)
-
 #endif /* OCFS2_FILE_H */

diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
index 1ab182321b18c79b287158b0a6a6208cb64b818e..c18ab45f8d210ca2032864069b3701b948aa0cbd 100644 (file)
--- a/fs/ocfs2/file.c
+++ b/fs/ocfs2/file.c
@@ -2178,7 +2178,7 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb,
        struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
        int full_coherency = !(osb->s_mount_opt &
                               OCFS2_MOUNT_COHERENCY_BUFFERED);
-       int unaligned_dio = 0;
+       void *saved_ki_complete = NULL;
        int append_write = ((iocb->ki_pos + count) >=
                        i_size_read(inode) ? 1 : 0);
 
@@ -2241,17 +2241,12 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb,
                goto out;
        }
 
-       if (direct_io && !is_sync_kiocb(iocb))
-               unaligned_dio = ocfs2_is_io_unaligned(inode, count, iocb->ki_pos);
-
-       if (unaligned_dio) {
+       if (direct_io && !is_sync_kiocb(iocb) &&
+           ocfs2_is_io_unaligned(inode, count, iocb->ki_pos)) {
                /*
-                * Wait on previous unaligned aio to complete before
-                * proceeding.
+                * Make it a sync io if it's an unaligned aio.
                 */
-               mutex_lock(&OCFS2_I(inode)->ip_unaligned_aio);
-               /* Mark the iocb as needing an unlock in ocfs2_dio_end_io */
-               ocfs2_iocb_set_unaligned_aio(iocb);
+               saved_ki_complete = xchg(&iocb->ki_complete, NULL);
        }
 
        /* communicate with ocfs2_dio_end_io */
@@ -2272,11 +2267,10 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb,
         */
        if ((written == -EIOCBQUEUED) || (!ocfs2_iocb_is_rw_locked(iocb))) {
                rw_level = -1;
-               unaligned_dio = 0;
        }
 
        if (unlikely(written <= 0))
-               goto no_sync;
+               goto out;
 
        if (((file->f_flags & O_DSYNC) && !direct_io) ||
            IS_SYNC(inode)) {
@@ -2298,13 +2292,10 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb,
                                                      iocb->ki_pos - 1);
        }
 
-no_sync:
-       if (unaligned_dio && ocfs2_iocb_is_unaligned_aio(iocb)) {
-               ocfs2_iocb_clear_unaligned_aio(iocb);
-               mutex_unlock(&OCFS2_I(inode)->ip_unaligned_aio);
-       }
-
 out:
+       if (saved_ki_complete)
+               xchg(&iocb->ki_complete, saved_ki_complete);
+
        if (rw_level != -1)
                ocfs2_rw_unlock(inode, rw_level);
 

diff --git a/fs/ocfs2/inode.h b/fs/ocfs2/inode.h
index 68e8cf9bda973d0446d3dbc2c70f68603b95c8e8..d8f3fc8d255157bc1fc41eaac72717cf01bbd177 100644 (file)
--- a/fs/ocfs2/inode.h
+++ b/fs/ocfs2/inode.h
@@ -43,9 +43,6 @@ struct ocfs2_inode_info
        /* protects extended attribute changes on this inode */
        struct rw_semaphore             ip_xattr_sem;
 
-       /* Number of outstanding AIO's which are not page aligned */
-       struct mutex                    ip_unaligned_aio;
-
        /* These fields are protected by ip_lock */
        spinlock_t                      ip_lock;
        u32                             ip_open_count;

diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c
index d142fb75740ee89d3d51c62e626e660640fa32cd..e2c4e3839d7bd12bbd17a5806448a445b7e10668 100644 (file)
--- a/fs/ocfs2/super.c
+++ b/fs/ocfs2/super.c
@@ -1747,7 +1747,6 @@ static void ocfs2_inode_init_once(void *data)
        INIT_LIST_HEAD(&oi->ip_io_markers);
        INIT_LIST_HEAD(&oi->ip_unwritten_list);
        oi->ip_dir_start_lookup = 0;
-       mutex_init(&oi->ip_unaligned_aio);
        init_rwsem(&oi->ip_alloc_sem);
        init_rwsem(&oi->ip_xattr_sem);
        mutex_init(&oi->ip_io_mutex);