Merge branch 'for-linus' of git://neil.brown.name/md
[sfrench/cifs-2.6.git] / drivers / md / raid5.c
index 96c690279fc6b58c7147c5860bf16417baa2149e..866d4b5a144c465daf21e439b9b0e6ef36571d6a 100644 (file)
@@ -201,11 +201,11 @@ static void __release_stripe(raid5_conf_t *conf, struct stripe_head *sh)
                if (test_bit(STRIPE_HANDLE, &sh->state)) {
                        if (test_bit(STRIPE_DELAYED, &sh->state)) {
                                list_add_tail(&sh->lru, &conf->delayed_list);
-                               blk_plug_device(conf->mddev->queue);
+                               plugger_set_plug(&conf->plug);
                        } else if (test_bit(STRIPE_BIT_DELAY, &sh->state) &&
                                   sh->bm_seq - conf->seq_write > 0) {
                                list_add_tail(&sh->lru, &conf->bitmap_list);
-                               blk_plug_device(conf->mddev->queue);
+                               plugger_set_plug(&conf->plug);
                        } else {
                                clear_bit(STRIPE_BIT_DELAY, &sh->state);
                                list_add_tail(&sh->lru, &conf->handle_list);
@@ -434,7 +434,6 @@ static int has_failed(raid5_conf_t *conf)
 }
 
 static void unplug_slaves(mddev_t *mddev);
-static void raid5_unplug_device(struct request_queue *q);
 
 static struct stripe_head *
 get_active_stripe(raid5_conf_t *conf, sector_t sector,
@@ -464,7 +463,7 @@ get_active_stripe(raid5_conf_t *conf, sector_t sector,
                                                     < (conf->max_nr_stripes *3/4)
                                                     || !conf->inactive_blocked),
                                                    conf->device_lock,
-                                                   raid5_unplug_device(conf->mddev->queue)
+                                                   md_raid5_unplug_device(conf)
                                        );
                                conf->inactive_blocked = 0;
                        } else
@@ -1337,10 +1336,14 @@ static int grow_stripes(raid5_conf_t *conf, int num)
        struct kmem_cache *sc;
        int devs = max(conf->raid_disks, conf->previous_raid_disks);
 
-       sprintf(conf->cache_name[0],
-               "raid%d-%s", conf->level, mdname(conf->mddev));
-       sprintf(conf->cache_name[1],
-               "raid%d-%s-alt", conf->level, mdname(conf->mddev));
+       if (conf->mddev->gendisk)
+               sprintf(conf->cache_name[0],
+                       "raid%d-%s", conf->level, mdname(conf->mddev));
+       else
+               sprintf(conf->cache_name[0],
+                       "raid%d-%p", conf->level, conf->mddev);
+       sprintf(conf->cache_name[1], "%s-alt", conf->cache_name[0]);
+
        conf->active_name = 0;
        sc = kmem_cache_create(conf->cache_name[conf->active_name],
                               sizeof(struct stripe_head)+(devs-1)*sizeof(struct r5dev),
@@ -3614,7 +3617,7 @@ static void raid5_activate_delayed(raid5_conf_t *conf)
                        list_add_tail(&sh->lru, &conf->hold_list);
                }
        } else
-               blk_plug_device(conf->mddev->queue);
+               plugger_set_plug(&conf->plug);
 }
 
 static void activate_bit_delay(raid5_conf_t *conf)
@@ -3655,36 +3658,44 @@ static void unplug_slaves(mddev_t *mddev)
        rcu_read_unlock();
 }
 
-static void raid5_unplug_device(struct request_queue *q)
+void md_raid5_unplug_device(raid5_conf_t *conf)
 {
-       mddev_t *mddev = q->queuedata;
-       raid5_conf_t *conf = mddev->private;
        unsigned long flags;
 
        spin_lock_irqsave(&conf->device_lock, flags);
 
-       if (blk_remove_plug(q)) {
+       if (plugger_remove_plug(&conf->plug)) {
                conf->seq_flush++;
                raid5_activate_delayed(conf);
        }
-       md_wakeup_thread(mddev->thread);
+       md_wakeup_thread(conf->mddev->thread);
 
        spin_unlock_irqrestore(&conf->device_lock, flags);
 
-       unplug_slaves(mddev);
+       unplug_slaves(conf->mddev);
 }
+EXPORT_SYMBOL_GPL(md_raid5_unplug_device);
 
-static int raid5_congested(void *data, int bits)
+static void raid5_unplug(struct plug_handle *plug)
+{
+       raid5_conf_t *conf = container_of(plug, raid5_conf_t, plug);
+       md_raid5_unplug_device(conf);
+}
+
+static void raid5_unplug_queue(struct request_queue *q)
+{
+       mddev_t *mddev = q->queuedata;
+       md_raid5_unplug_device(mddev->private);
+}
+
+int md_raid5_congested(mddev_t *mddev, int bits)
 {
-       mddev_t *mddev = data;
        raid5_conf_t *conf = mddev->private;
 
        /* No difference between reads and writes.  Just check
         * how busy the stripe_cache is
         */
 
-       if (mddev_congested(mddev, bits))
-               return 1;
        if (conf->inactive_blocked)
                return 1;
        if (conf->quiesce)
@@ -3694,6 +3705,15 @@ static int raid5_congested(void *data, int bits)
 
        return 0;
 }
+EXPORT_SYMBOL_GPL(md_raid5_congested);
+
+static int raid5_congested(void *data, int bits)
+{
+       mddev_t *mddev = data;
+
+       return mddev_congested(mddev, bits) ||
+               md_raid5_congested(mddev, bits);
+}
 
 /* We want read requests to align with chunks where possible,
  * but write requests don't need to.
@@ -3958,7 +3978,7 @@ static int make_request(mddev_t *mddev, struct bio * bi)
        const int rw = bio_data_dir(bi);
        int remaining;
 
-       if (unlikely(bio_rw_flagged(bi, BIO_RW_BARRIER))) {
+       if (unlikely(bi->bi_rw & REQ_HARDBARRIER)) {
                /* Drain all pending writes.  We only really need
                 * to ensure they have been submitted, but this is
                 * easier.
@@ -4075,7 +4095,7 @@ static int make_request(mddev_t *mddev, struct bio * bi)
                                 * add failed due to overlap.  Flush everything
                                 * and wait a while
                                 */
-                               raid5_unplug_device(mddev->queue);
+                               md_raid5_unplug_device(conf);
                                release_stripe(sh);
                                schedule();
                                goto retry;
@@ -4566,23 +4586,15 @@ raid5_show_stripe_cache_size(mddev_t *mddev, char *page)
                return 0;
 }
 
-static ssize_t
-raid5_store_stripe_cache_size(mddev_t *mddev, const char *page, size_t len)
+int
+raid5_set_cache_size(mddev_t *mddev, int size)
 {
        raid5_conf_t *conf = mddev->private;
-       unsigned long new;
        int err;
 
-       if (len >= PAGE_SIZE)
+       if (size <= 16 || size > 32768)
                return -EINVAL;
-       if (!conf)
-               return -ENODEV;
-
-       if (strict_strtoul(page, 10, &new))
-               return -EINVAL;
-       if (new <= 16 || new > 32768)
-               return -EINVAL;
-       while (new < conf->max_nr_stripes) {
+       while (size < conf->max_nr_stripes) {
                if (drop_one_stripe(conf))
                        conf->max_nr_stripes--;
                else
@@ -4591,11 +4603,32 @@ raid5_store_stripe_cache_size(mddev_t *mddev, const char *page, size_t len)
        err = md_allow_write(mddev);
        if (err)
                return err;
-       while (new > conf->max_nr_stripes) {
+       while (size > conf->max_nr_stripes) {
                if (grow_one_stripe(conf))
                        conf->max_nr_stripes++;
                else break;
        }
+       return 0;
+}
+EXPORT_SYMBOL(raid5_set_cache_size);
+
+static ssize_t
+raid5_store_stripe_cache_size(mddev_t *mddev, const char *page, size_t len)
+{
+       raid5_conf_t *conf = mddev->private;
+       unsigned long new;
+       int err;
+
+       if (len >= PAGE_SIZE)
+               return -EINVAL;
+       if (!conf)
+               return -ENODEV;
+
+       if (strict_strtoul(page, 10, &new))
+               return -EINVAL;
+       err = raid5_set_cache_size(mddev, new);
+       if (err)
+               return err;
        return len;
 }
 
@@ -4958,7 +4991,7 @@ static int only_parity(int raid_disk, int algo, int raid_disks, int max_degraded
 static int run(mddev_t *mddev)
 {
        raid5_conf_t *conf;
-       int working_disks = 0, chunk_size;
+       int working_disks = 0;
        int dirty_parity_disks = 0;
        mdk_rdev_t *rdev;
        sector_t reshape_offset = 0;
@@ -5144,42 +5177,47 @@ static int run(mddev_t *mddev)
                                                        "reshape");
        }
 
-       /* read-ahead size must cover two whole stripes, which is
-        * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
-        */
-       {
-               int data_disks = conf->previous_raid_disks - conf->max_degraded;
-               int stripe = data_disks *
-                       ((mddev->chunk_sectors << 9) / PAGE_SIZE);
-               if (mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
-                       mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
-       }
 
        /* Ok, everything is just fine now */
        if (mddev->to_remove == &raid5_attrs_group)
                mddev->to_remove = NULL;
-       else if (sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
+       else if (mddev->kobj.sd &&
+           sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
                printk(KERN_WARNING
-                      "md/raid:%s: failed to create sysfs attributes.\n",
+                      "raid5: failed to create sysfs attributes for %s\n",
                       mdname(mddev));
+       md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
 
-       mddev->queue->queue_lock = &conf->device_lock;
+       plugger_init(&conf->plug, raid5_unplug);
+       mddev->plug = &conf->plug;
+       if (mddev->queue) {
+               int chunk_size;
+               /* read-ahead size must cover two whole stripes, which
+                * is 2 * (datadisks) * chunksize where 'n' is the
+                * number of raid devices
+                */
+               int data_disks = conf->previous_raid_disks - conf->max_degraded;
+               int stripe = data_disks *
+                       ((mddev->chunk_sectors << 9) / PAGE_SIZE);
+               if (mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
+                       mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
 
-       mddev->queue->unplug_fn = raid5_unplug_device;
-       mddev->queue->backing_dev_info.congested_data = mddev;
-       mddev->queue->backing_dev_info.congested_fn = raid5_congested;
+               blk_queue_merge_bvec(mddev->queue, raid5_mergeable_bvec);
 
-       md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
+               mddev->queue->backing_dev_info.congested_data = mddev;
+               mddev->queue->backing_dev_info.congested_fn = raid5_congested;
+               mddev->queue->queue_lock = &conf->device_lock;
+               mddev->queue->unplug_fn = raid5_unplug_queue;
 
-       blk_queue_merge_bvec(mddev->queue, raid5_mergeable_bvec);
-       chunk_size = mddev->chunk_sectors << 9;
-       blk_queue_io_min(mddev->queue, chunk_size);
-       blk_queue_io_opt(mddev->queue, chunk_size *
-                        (conf->raid_disks - conf->max_degraded));
+               chunk_size = mddev->chunk_sectors << 9;
+               blk_queue_io_min(mddev->queue, chunk_size);
+               blk_queue_io_opt(mddev->queue, chunk_size *
+                                (conf->raid_disks - conf->max_degraded));
 
-       list_for_each_entry(rdev, &mddev->disks, same_set)
-               disk_stack_limits(mddev->gendisk, rdev->bdev,
-                                 rdev->data_offset << 9);
+               list_for_each_entry(rdev, &mddev->disks, same_set)
+                       disk_stack_limits(mddev->gendisk, rdev->bdev,
+                                         rdev->data_offset << 9);
+       }
 
        return 0;
 abort:
@@ -5200,8 +5238,9 @@ static int stop(mddev_t *mddev)
 
        md_unregister_thread(mddev->thread);
        mddev->thread = NULL;
-       mddev->queue->backing_dev_info.congested_fn = NULL;
-       blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
+       if (mddev->queue)
+               mddev->queue->backing_dev_info.congested_fn = NULL;
+       plugger_flush(&conf->plug); /* the unplug fn references 'conf'*/
        free_conf(conf);
        mddev->private = NULL;
        mddev->to_remove = &raid5_attrs_group;
@@ -5545,10 +5584,7 @@ static int raid5_start_reshape(mddev_t *mddev)
                                sprintf(nm, "rd%d", rdev->raid_disk);
                                if (sysfs_create_link(&mddev->kobj,
                                                      &rdev->kobj, nm))
-                                       printk(KERN_WARNING
-                                              "md/raid:%s: failed to create "
-                                              " link %s\n",
-                                              mdname(mddev), nm);
+                                       /* Failure here is OK */;
                        } else
                                break;
                }
@@ -5603,7 +5639,7 @@ static void end_reshape(raid5_conf_t *conf)
                /* read-ahead size must cover two whole stripes, which is
                 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
                 */
-               {
+               if (conf->mddev->queue) {
                        int data_disks = conf->raid_disks - conf->max_degraded;
                        int stripe = data_disks * ((conf->chunk_sectors << 9)
                                                   / PAGE_SIZE);