[MTD] NAND: Reorganize chip locking

The code was wrong in several aspects. The locking order was
inconsistent, the device aquire code did not reset a variable
after a wakeup and the wakeup handling was not working for
applications where multiple chips are sharing a single
hardware controller.
When a hardware controller is available the locking is now
reduced to the hardware controller lock and the waitqueue is
moved to the hardware controller structure in order to avoid
a wake_up_all().

The problem was pointed out by Ben Dooks, who also found the
missing variable reset as main cause for his deadlock problem.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

1 files changed, 4 insertions, 1 deletions

diff --git a/include/linux/mtd/nand.h b/include/linux/mtd/nand.h
index bee78969cb21..9b5b76217584 100644
--- a/include/linux/mtd/nand.h
+++ b/include/linux/mtd/nand.h
@@ -5,7 +5,7 @@
  *                     Steven J. Hill <sjhill@realitydiluted.com>
  *                     Thomas Gleixner <tglx@linutronix.de>
  *
- * $Id: nand.h,v 1.71 2005/02/09 12:12:59 gleixner Exp $
+ * $Id: nand.h,v 1.73 2005/05/31 19:39:17 gleixner Exp $
  *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License version 2 as
@@ -253,10 +253,13 @@ struct nand_chip;
  * struct nand_hw_control - Control structure for hardware controller (e.g ECC generator) shared among independend devices
  * @lock:               protection lock  
  * @active:             the mtd device which holds the controller currently
+ * @wq:                 wait queue to sleep on if a NAND operation is in progress
+ *                      used instead of the per chip wait queue when a hw controller is available
  */
 struct nand_hw_control {
         spinlock_t       lock;
         struct nand_chip *active;
+        wait_queue_head_t wq;
 };
 
 /**