Merge tag 'md/3.13' of git://neil.brown.name/md

Pull md update from Neil Brown:
 "Mostly optimisations and obscure bug fixes.
   - raid5 gets less lock contention
   - raid1 gets less contention between normal-io and resync-io during
     resync"

* tag 'md/3.13' of git://neil.brown.name/md:
  md/raid5: Use conf->device_lock protect changing of multi-thread resources.
  md/raid5: Before freeing old multi-thread worker, it should flush them.
  md/raid5: For stripe with R5_ReadNoMerge, we replace REQ_FLUSH with REQ_NOMERGE.
  UAPI: include <asm/byteorder.h> in linux/raid/md_p.h
  raid1: Rewrite the implementation of iobarrier.
  raid1: Add some macros to make code clearly.
  raid1: Replace raise_barrier/lower_barrier with freeze_array/unfreeze_array when reconfiguring the array.
  raid1: Add a field array_frozen to indicate whether raid in freeze state.
  md: Convert use of typedef ctl_table to struct ctl_table
  md/raid5: avoid deadlock when raid5 array has unack badblocks during md_stop_writes.
  md: use MD_RECOVERY_INTR instead of kthread_should_stop in resync thread.
  md: fix some places where mddev_lock return value is not checked.
  raid5: Retry R5_ReadNoMerge flag when hit a read error.
  raid5: relieve lock contention in get_active_stripe()
  raid5: relieve lock contention in get_active_stripe()
  wait: add wait_event_cmd()
  md/raid5.c: add proper locking to error path of raid5_start_reshape.
  md: fix calculation of stacking limits on level change.
  raid5: Use slow_path to release stripe when mddev->thread is null

drivers/md/md.c

@@ -112,7 +112,7 @@ static inline int speed_max(struct mddev *mddev)
 
 static struct ctl_table_header *raid_table_header;
 
-static ctl_table raid_table[] = {
+static struct ctl_table raid_table[] = {
 	{
 		.procname	= "speed_limit_min",
 		.data		= &sysctl_speed_limit_min,
@@ -130,7 +130,7 @@ static ctl_table raid_table[] = {
 	{ }
 };
 
-static ctl_table raid_dir_table[] = {
+static struct ctl_table raid_dir_table[] = {
 	{
 		.procname	= "raid",
 		.maxlen		= 0,
@@ -140,7 +140,7 @@ static ctl_table raid_dir_table[] = {
 	{ }
 };
 
-static ctl_table raid_root_table[] = {
+static struct ctl_table raid_root_table[] = {
 	{
 		.procname	= "dev",
 		.maxlen		= 0,
@@ -562,11 +562,19 @@ static struct mddev * mddev_find(dev_t unit)
 	goto retry;
 }
 
-static inline int mddev_lock(struct mddev * mddev)
+static inline int __must_check mddev_lock(struct mddev * mddev)
 {
 	return mutex_lock_interruptible(&mddev->reconfig_mutex);
 }
 
+/* Sometimes we need to take the lock in a situation where
+ * failure due to interrupts is not acceptable.
+ */
+static inline void mddev_lock_nointr(struct mddev * mddev)
+{
+	mutex_lock(&mddev->reconfig_mutex);
+}
+
 static inline int mddev_is_locked(struct mddev *mddev)
 {
 	return mutex_is_locked(&mddev->reconfig_mutex);
@@ -2978,7 +2986,7 @@ rdev_size_store(struct md_rdev *rdev, const char *buf, size_t len)
 		for_each_mddev(mddev, tmp) {
 			struct md_rdev *rdev2;
 
-			mddev_lock(mddev);
+			mddev_lock_nointr(mddev);
 			rdev_for_each(rdev2, mddev)
 				if (rdev->bdev == rdev2->bdev &&
 				    rdev != rdev2 &&
@@ -2994,7 +3002,7 @@ rdev_size_store(struct md_rdev *rdev, const char *buf, size_t len)
 				break;
 			}
 		}
-		mddev_lock(my_mddev);
+		mddev_lock_nointr(my_mddev);
 		if (overlap) {
 			/* Someone else could have slipped in a size
 			 * change here, but doing so is just silly.
@@ -3580,6 +3588,7 @@ level_store(struct mddev *mddev, const char *buf, size_t len)
 		mddev->in_sync = 1;
 		del_timer_sync(&mddev->safemode_timer);
 	}
+	blk_set_stacking_limits(&mddev->queue->limits);
 	pers->run(mddev);
 	set_bit(MD_CHANGE_DEVS, &mddev->flags);
 	mddev_resume(mddev);
@@ -5258,7 +5267,7 @@ static void __md_stop_writes(struct mddev *mddev)
 
 void md_stop_writes(struct mddev *mddev)
 {
-	mddev_lock(mddev);
+	mddev_lock_nointr(mddev);
 	__md_stop_writes(mddev);
 	mddev_unlock(mddev);
 }
@@ -5291,20 +5300,35 @@ EXPORT_SYMBOL_GPL(md_stop);
 static int md_set_readonly(struct mddev *mddev, struct block_device *bdev)
 {
 	int err = 0;
+	int did_freeze = 0;
+
+	if (!test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) {
+		did_freeze = 1;
+		set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+		md_wakeup_thread(mddev->thread);
+	}
+	if (mddev->sync_thread) {
+		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+		/* Thread might be blocked waiting for metadata update
+		 * which will now never happen */
+		wake_up_process(mddev->sync_thread->tsk);
+	}
+	mddev_unlock(mddev);
+	wait_event(resync_wait, mddev->sync_thread == NULL);
+	mddev_lock_nointr(mddev);
+
 	mutex_lock(&mddev->open_mutex);
-	if (atomic_read(&mddev->openers) > !!bdev) {
+	if (atomic_read(&mddev->openers) > !!bdev ||
+	    mddev->sync_thread ||
+	    (bdev && !test_bit(MD_STILL_CLOSED, &mddev->flags))) {
 		printk("md: %s still in use.\n",mdname(mddev));
+		if (did_freeze) {
+			clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+			md_wakeup_thread(mddev->thread);
+		}
 		err = -EBUSY;
 		goto out;
 	}
-	if (bdev && !test_bit(MD_STILL_CLOSED, &mddev->flags)) {
-		/* Someone opened the device since we flushed it
-		 * so page cache could be dirty and it is too late
-		 * to flush.  So abort
-		 */
-		mutex_unlock(&mddev->open_mutex);
-		return -EBUSY;
-	}
 	if (mddev->pers) {
 		__md_stop_writes(mddev);
 
@@ -5331,20 +5355,34 @@ static int do_md_stop(struct mddev * mddev, int mode,
 {
 	struct gendisk *disk = mddev->gendisk;
 	struct md_rdev *rdev;
+	int did_freeze = 0;
+
+	if (!test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) {
+		did_freeze = 1;
+		set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+		md_wakeup_thread(mddev->thread);
+	}
+	if (mddev->sync_thread) {
+		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+		/* Thread might be blocked waiting for metadata update
+		 * which will now never happen */
+		wake_up_process(mddev->sync_thread->tsk);
+	}
+	mddev_unlock(mddev);
+	wait_event(resync_wait, mddev->sync_thread == NULL);
+	mddev_lock_nointr(mddev);
 
 	mutex_lock(&mddev->open_mutex);
 	if (atomic_read(&mddev->openers) > !!bdev ||
-	    mddev->sysfs_active) {
+	    mddev->sysfs_active ||
+	    mddev->sync_thread ||
+	    (bdev && !test_bit(MD_STILL_CLOSED, &mddev->flags))) {
 		printk("md: %s still in use.\n",mdname(mddev));
 		mutex_unlock(&mddev->open_mutex);
-		return -EBUSY;
+		if (did_freeze) {
+			clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+			md_wakeup_thread(mddev->thread);
 		}
-	if (bdev && !test_bit(MD_STILL_CLOSED, &mddev->flags)) {
-		/* Someone opened the device since we flushed it
-		 * so page cache could be dirty and it is too late
-		 * to flush.  So abort
-		 */
-		mutex_unlock(&mddev->open_mutex);
 		return -EBUSY;
 	}
 	if (mddev->pers) {
@@ -6551,7 +6589,7 @@ static int md_ioctl(struct block_device *bdev, fmode_t mode,
 				wait_event(mddev->sb_wait,
 					   !test_bit(MD_CHANGE_DEVS, &mddev->flags) &&
 					   !test_bit(MD_CHANGE_PENDING, &mddev->flags));
-				mddev_lock(mddev);
+				mddev_lock_nointr(mddev);
 			}
 		} else {
 			err = -EROFS;
@@ -7361,9 +7399,6 @@ void md_do_sync(struct md_thread *thread)
 		mddev->curr_resync = 2;
 
 	try_again:
-		if (kthread_should_stop())
-			set_bit(MD_RECOVERY_INTR, &mddev->recovery);
-
 		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
 			goto skip;
 		for_each_mddev(mddev2, tmp) {
@@ -7388,7 +7423,7 @@ void md_do_sync(struct md_thread *thread)
 				 * be caught by 'softlockup'
 				 */
 				prepare_to_wait(&resync_wait, &wq, TASK_INTERRUPTIBLE);
-				if (!kthread_should_stop() &&
+				if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
 				    mddev2->curr_resync >= mddev->curr_resync) {
 					printk(KERN_INFO "md: delaying %s of %s"
 					       " until %s has finished (they"
@@ -7501,7 +7536,8 @@ void md_do_sync(struct md_thread *thread)
 			sysfs_notify(&mddev->kobj, NULL, "sync_completed");
 		}
 
-		while (j >= mddev->resync_max && !kthread_should_stop()) {
+		while (j >= mddev->resync_max &&
+		       !test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
 			/* As this condition is controlled by user-space,
 			 * we can block indefinitely, so use '_interruptible'
 			 * to avoid triggering warnings.
@@ -7509,17 +7545,18 @@ void md_do_sync(struct md_thread *thread)
 			flush_signals(current); /* just in case */
 			wait_event_interruptible(mddev->recovery_wait,
 						 mddev->resync_max > j
-						 || kthread_should_stop());
+						 || test_bit(MD_RECOVERY_INTR,
+							     &mddev->recovery));
 		}
 
-		if (kthread_should_stop())
-			goto interrupted;
+		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
+			break;
 
 		sectors = mddev->pers->sync_request(mddev, j, &skipped,
 						  currspeed < speed_min(mddev));
 		if (sectors == 0) {
 			set_bit(MD_RECOVERY_INTR, &mddev->recovery);
-			goto out;
+			break;
 		}
 
 		if (!skipped) { /* actual IO requested */
@@ -7556,10 +7593,8 @@ void md_do_sync(struct md_thread *thread)
 			last_mark = next;
 		}
 
-
-		if (kthread_should_stop())
-			goto interrupted;
-
+		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
+			break;
 
 		/*
 		 * this loop exits only if either when we are slower than
@@ -7582,11 +7617,12 @@ void md_do_sync(struct md_thread *thread)
 			}
 		}
 	}
-	printk(KERN_INFO "md: %s: %s done.\n",mdname(mddev), desc);
+	printk(KERN_INFO "md: %s: %s %s.\n",mdname(mddev), desc,
+	       test_bit(MD_RECOVERY_INTR, &mddev->recovery)
+	       ? "interrupted" : "done");
 	/*
 	 * this also signals 'finished resyncing' to md_stop
 	 */
- out:
 	blk_finish_plug(&plug);
 	wait_event(mddev->recovery_wait, !atomic_read(&mddev->recovery_active));
 
@@ -7640,16 +7676,6 @@ void md_do_sync(struct md_thread *thread)
 	set_bit(MD_RECOVERY_DONE, &mddev->recovery);
 	md_wakeup_thread(mddev->thread);
 	return;
-
- interrupted:
-	/*
-	 * got a signal, exit.
-	 */
-	printk(KERN_INFO
-	       "md: md_do_sync() got signal ... exiting\n");
-	set_bit(MD_RECOVERY_INTR, &mddev->recovery);
-	goto out;
-
 }
 EXPORT_SYMBOL_GPL(md_do_sync);
 
@@ -7894,6 +7920,7 @@ void md_reap_sync_thread(struct mddev *mddev)
 
 	/* resync has finished, collect result */
 	md_unregister_thread(&mddev->sync_thread);
+	wake_up(&resync_wait);
 	if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
 	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
 		/* success...*/

drivers/md/raid1.h

@@ -41,6 +41,19 @@ struct r1conf {
 	 */
 	sector_t		next_resync;
 
+	/* When raid1 starts resync, we divide array into four partitions
+	 * |---------|--------------|---------------------|-------------|
+	 *        next_resync   start_next_window       end_window
+	 * start_next_window = next_resync + NEXT_NORMALIO_DISTANCE
+	 * end_window = start_next_window + NEXT_NORMALIO_DISTANCE
+	 * current_window_requests means the count of normalIO between
+	 *   start_next_window and end_window.
+	 * next_window_requests means the count of normalIO after end_window.
+	 * */
+	sector_t		start_next_window;
+	int			current_window_requests;
+	int			next_window_requests;
+
 	spinlock_t		device_lock;
 
 	/* list of 'struct r1bio' that need to be processed by raid1d,
@@ -65,6 +78,7 @@ struct r1conf {
 	int			nr_waiting;
 	int			nr_queued;
 	int			barrier;
+	int			array_frozen;
 
 	/* Set to 1 if a full sync is needed, (fresh device added).
 	 * Cleared when a sync completes.
@@ -111,6 +125,7 @@ struct r1bio {
 						 * in this BehindIO request
 						 */
 	sector_t		sector;
+	sector_t		start_next_window;
 	int			sectors;
 	unsigned long		state;
 	struct mddev		*mddev;

drivers/md/raid10.c

@@ -4384,7 +4384,11 @@ static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr,
 		set_bit(MD_CHANGE_DEVS, &mddev->flags);
 		md_wakeup_thread(mddev->thread);
 		wait_event(mddev->sb_wait, mddev->flags == 0 ||
-			   kthread_should_stop());
+			   test_bit(MD_RECOVERY_INTR, &mddev->recovery));
+		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
+			allow_barrier(conf);
+			return sectors_done;
+		}
 		conf->reshape_safe = mddev->reshape_position;
 		allow_barrier(conf);
 	}

drivers/md/raid5.h

@@ -205,6 +205,7 @@ struct stripe_head {
 	short			pd_idx;		/* parity disk index */
 	short			qd_idx;		/* 'Q' disk index for raid6 */
 	short			ddf_layout;/* use DDF ordering to calculate Q */
+	short			hash_lock_index;
 	unsigned long		state;		/* state flags */
 	atomic_t		count;	      /* nr of active thread/requests */
 	int			bm_seq;	/* sequence number for bitmap flushes */
@@ -367,9 +368,18 @@ struct disk_info {
 	struct md_rdev	*rdev, *replacement;
 };
 
+/* NOTE NR_STRIPE_HASH_LOCKS must remain below 64.
+ * This is because we sometimes take all the spinlocks
+ * and creating that much locking depth can cause
+ * problems.
+ */
+#define NR_STRIPE_HASH_LOCKS 8
+#define STRIPE_HASH_LOCKS_MASK (NR_STRIPE_HASH_LOCKS - 1)
+
 struct r5worker {
 	struct work_struct work;
 	struct r5worker_group *group;
+	struct list_head temp_inactive_list[NR_STRIPE_HASH_LOCKS];
 	bool working;
 };
 
@@ -382,6 +392,8 @@ struct r5worker_group {
 
 struct r5conf {
 	struct hlist_head	*stripe_hashtbl;
+	/* only protect corresponding hash list and inactive_list */
+	spinlock_t		hash_locks[NR_STRIPE_HASH_LOCKS];
 	struct mddev		*mddev;
 	int			chunk_sectors;
 	int			level, algorithm;
@@ -462,7 +474,8 @@ struct r5conf {
 	 * Free stripes pool
 	 */
 	atomic_t		active_stripes;
-	struct list_head	inactive_list;
+	struct list_head	inactive_list[NR_STRIPE_HASH_LOCKS];
+	atomic_t		empty_inactive_list_nr;
 	struct llist_head	released_stripes;
 	wait_queue_head_t	wait_for_stripe;
 	wait_queue_head_t	wait_for_overlap;
@@ -477,6 +490,7 @@ struct r5conf {
 	 * the new thread here until we fully activate the array.
 	 */
 	struct md_thread	*thread;
+	struct list_head	temp_inactive_list[NR_STRIPE_HASH_LOCKS];
 	struct r5worker_group	*worker_groups;
 	int			group_cnt;
 	int			worker_cnt_per_group;

include/linux/wait.h

@@ -278,6 +278,31 @@ do {									\
 	__ret;								\
 })
 
+#define __wait_event_cmd(wq, condition, cmd1, cmd2)			\
+	(void)___wait_event(wq, condition, TASK_UNINTERRUPTIBLE, 0, 0,	\
+			    cmd1; schedule(); cmd2)
+
+/**
+ * wait_event_cmd - sleep until a condition gets true
+ * @wq: the waitqueue to wait on
+ * @condition: a C expression for the event to wait for
+ * cmd1: the command will be executed before sleep
+ * cmd2: the command will be executed after sleep
+ *
+ * The process is put to sleep (TASK_UNINTERRUPTIBLE) until the
+ * @condition evaluates to true. The @condition is checked each time
+ * the waitqueue @wq is woken up.
+ *
+ * wake_up() has to be called after changing any variable that could
+ * change the result of the wait condition.
+ */
+#define wait_event_cmd(wq, condition, cmd1, cmd2)			\
+do {									\
+	if (condition)							\
+		break;							\
+	__wait_event_cmd(wq, condition, cmd1, cmd2);			\
+} while (0)
+
 #define __wait_event_interruptible(wq, condition)			\
 	___wait_event(wq, condition, TASK_INTERRUPTIBLE, 0, 0,		\
 		      schedule())

include/uapi/linux/raid/md_p.h

@@ -16,6 +16,7 @@
 #define _MD_P_H
 
 #include <linux/types.h>
+#include <asm/byteorder.h>
 
 /*
  * RAID superblock.