Merge branch 'for-3.6/drivers' of git://git.kernel.dk/linux-block
[linux-drm-fsl-dcu.git] / drivers / md / raid10.c
index 5d33603a497d6b7a0ee53c54d51501b99c4ee2ac..de5ed6fd8806c3c3db39bbd894fabb04734b4523 100644 (file)
  */
 #define        NR_RAID10_BIOS 256
 
-/* When there are this many requests queue to be written by
+/* when we get a read error on a read-only array, we redirect to another
+ * device without failing the first device, or trying to over-write to
+ * correct the read error.  To keep track of bad blocks on a per-bio
+ * level, we store IO_BLOCKED in the appropriate 'bios' pointer
+ */
+#define IO_BLOCKED ((struct bio *)1)
+/* When we successfully write to a known bad-block, we need to remove the
+ * bad-block marking which must be done from process context.  So we record
+ * the success by setting devs[n].bio to IO_MADE_GOOD
+ */
+#define IO_MADE_GOOD ((struct bio *)2)
+
+#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
+
+/* When there are this many requests queued to be written by
  * the raid10 thread, we become 'congested' to provide back-pressure
  * for writeback.
  */
@@ -717,7 +731,7 @@ static struct md_rdev *read_balance(struct r10conf *conf,
        int sectors = r10_bio->sectors;
        int best_good_sectors;
        sector_t new_distance, best_dist;
-       struct md_rdev *rdev, *best_rdev;
+       struct md_rdev *best_rdev, *rdev = NULL;
        int do_balance;
        int best_slot;
        struct geom *geo = &conf->geo;
@@ -839,9 +853,8 @@ retry:
        return rdev;
 }
 
-static int raid10_congested(void *data, int bits)
+int md_raid10_congested(struct mddev *mddev, int bits)
 {
-       struct mddev *mddev = data;
        struct r10conf *conf = mddev->private;
        int i, ret = 0;
 
@@ -849,8 +862,6 @@ static int raid10_congested(void *data, int bits)
            conf->pending_count >= max_queued_requests)
                return 1;
 
-       if (mddev_congested(mddev, bits))
-               return 1;
        rcu_read_lock();
        for (i = 0;
             (i < conf->geo.raid_disks || i < conf->prev.raid_disks)
@@ -866,6 +877,15 @@ static int raid10_congested(void *data, int bits)
        rcu_read_unlock();
        return ret;
 }
+EXPORT_SYMBOL_GPL(md_raid10_congested);
+
+static int raid10_congested(void *data, int bits)
+{
+       struct mddev *mddev = data;
+
+       return mddev_congested(mddev, bits) ||
+               md_raid10_congested(mddev, bits);
+}
 
 static void flush_pending_writes(struct r10conf *conf)
 {
@@ -1546,7 +1566,7 @@ static void error(struct mddev *mddev, struct md_rdev *rdev)
 static void print_conf(struct r10conf *conf)
 {
        int i;
-       struct mirror_info *tmp;
+       struct raid10_info *tmp;
 
        printk(KERN_DEBUG "RAID10 conf printout:\n");
        if (!conf) {
@@ -1580,7 +1600,7 @@ static int raid10_spare_active(struct mddev *mddev)
 {
        int i;
        struct r10conf *conf = mddev->private;
-       struct mirror_info *tmp;
+       struct raid10_info *tmp;
        int count = 0;
        unsigned long flags;
 
@@ -1655,7 +1675,7 @@ static int raid10_add_disk(struct mddev *mddev, struct md_rdev *rdev)
        else
                mirror = first;
        for ( ; mirror <= last ; mirror++) {
-               struct mirror_info *p = &conf->mirrors[mirror];
+               struct raid10_info *p = &conf->mirrors[mirror];
                if (p->recovery_disabled == mddev->recovery_disabled)
                        continue;
                if (p->rdev) {
@@ -1709,7 +1729,7 @@ static int raid10_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
        int err = 0;
        int number = rdev->raid_disk;
        struct md_rdev **rdevp;
-       struct mirror_info *p = conf->mirrors + number;
+       struct raid10_info *p = conf->mirrors + number;
 
        print_conf(conf);
        if (rdev == p->rdev)
@@ -2875,7 +2895,7 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
                        sector_t sect;
                        int must_sync;
                        int any_working;
-                       struct mirror_info *mirror = &conf->mirrors[i];
+                       struct raid10_info *mirror = &conf->mirrors[i];
 
                        if ((mirror->rdev == NULL ||
                             test_bit(In_sync, &mirror->rdev->flags))
@@ -3387,7 +3407,7 @@ static struct r10conf *setup_conf(struct mddev *mddev)
                goto out;
 
        /* FIXME calc properly */
-       conf->mirrors = kzalloc(sizeof(struct mirror_info)*(mddev->raid_disks +
+       conf->mirrors = kzalloc(sizeof(struct raid10_info)*(mddev->raid_disks +
                                                            max(0,mddev->delta_disks)),
                                GFP_KERNEL);
        if (!conf->mirrors)
@@ -3451,7 +3471,7 @@ static int run(struct mddev *mddev)
 {
        struct r10conf *conf;
        int i, disk_idx, chunk_size;
-       struct mirror_info *disk;
+       struct raid10_info *disk;
        struct md_rdev *rdev;
        sector_t size;
        sector_t min_offset_diff = 0;
@@ -3471,12 +3491,14 @@ static int run(struct mddev *mddev)
        conf->thread = NULL;
 
        chunk_size = mddev->chunk_sectors << 9;
-       blk_queue_io_min(mddev->queue, chunk_size);
-       if (conf->geo.raid_disks % conf->geo.near_copies)
-               blk_queue_io_opt(mddev->queue, chunk_size * conf->geo.raid_disks);
-       else
-               blk_queue_io_opt(mddev->queue, chunk_size *
-                                (conf->geo.raid_disks / conf->geo.near_copies));
+       if (mddev->queue) {
+               blk_queue_io_min(mddev->queue, chunk_size);
+               if (conf->geo.raid_disks % conf->geo.near_copies)
+                       blk_queue_io_opt(mddev->queue, chunk_size * conf->geo.raid_disks);
+               else
+                       blk_queue_io_opt(mddev->queue, chunk_size *
+                                        (conf->geo.raid_disks / conf->geo.near_copies));
+       }
 
        rdev_for_each(rdev, mddev) {
                long long diff;
@@ -3510,8 +3532,9 @@ static int run(struct mddev *mddev)
                if (first || diff < min_offset_diff)
                        min_offset_diff = diff;
 
-               disk_stack_limits(mddev->gendisk, rdev->bdev,
-                                 rdev->data_offset << 9);
+               if (mddev->gendisk)
+                       disk_stack_limits(mddev->gendisk, rdev->bdev,
+                                         rdev->data_offset << 9);
 
                disk->head_position = 0;
        }
@@ -3574,22 +3597,22 @@ static int run(struct mddev *mddev)
        md_set_array_sectors(mddev, size);
        mddev->resync_max_sectors = size;
 
-       mddev->queue->backing_dev_info.congested_fn = raid10_congested;
-       mddev->queue->backing_dev_info.congested_data = mddev;
-
-       /* Calculate max read-ahead size.
-        * We need to readahead at least twice a whole stripe....
-        * maybe...
-        */
-       {
+       if (mddev->queue) {
                int stripe = conf->geo.raid_disks *
                        ((mddev->chunk_sectors << 9) / PAGE_SIZE);
+               mddev->queue->backing_dev_info.congested_fn = raid10_congested;
+               mddev->queue->backing_dev_info.congested_data = mddev;
+
+               /* Calculate max read-ahead size.
+                * We need to readahead at least twice a whole stripe....
+                * maybe...
+                */
                stripe /= conf->geo.near_copies;
                if (mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
                        mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
+               blk_queue_merge_bvec(mddev->queue, raid10_mergeable_bvec);
        }
 
-       blk_queue_merge_bvec(mddev->queue, raid10_mergeable_bvec);
 
        if (md_integrity_register(mddev))
                goto out_free_conf;
@@ -3640,7 +3663,10 @@ static int stop(struct mddev *mddev)
        lower_barrier(conf);
 
        md_unregister_thread(&mddev->thread);
-       blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
+       if (mddev->queue)
+               /* the unplug fn references 'conf'*/
+               blk_sync_queue(mddev->queue);
+
        if (conf->r10bio_pool)
                mempool_destroy(conf->r10bio_pool);
        kfree(conf->mirrors);
@@ -3804,7 +3830,7 @@ static int raid10_check_reshape(struct mddev *mddev)
        if (mddev->delta_disks > 0) {
                /* allocate new 'mirrors' list */
                conf->mirrors_new = kzalloc(
-                       sizeof(struct mirror_info)
+                       sizeof(struct raid10_info)
                        *(mddev->raid_disks +
                          mddev->delta_disks),
                        GFP_KERNEL);
@@ -3929,7 +3955,7 @@ static int raid10_start_reshape(struct mddev *mddev)
        spin_lock_irq(&conf->device_lock);
        if (conf->mirrors_new) {
                memcpy(conf->mirrors_new, conf->mirrors,
-                      sizeof(struct mirror_info)*conf->prev.raid_disks);
+                      sizeof(struct raid10_info)*conf->prev.raid_disks);
                smp_mb();
                kfree(conf->mirrors_old); /* FIXME and elsewhere */
                conf->mirrors_old = conf->mirrors;