md/raid10: Introduce 'prev' geometry to support reshape.

When RAID10 supports reshape it will need a 'previous' and a 'current'
geometry, so introduce that here.
Use the 'prev' geometry when before the reshape_position, and the
current 'geo' when beyond it.  At other times, use both as
appropriate.

For now, both are identical (And reshape_position is never set).

When we use the 'prev' geometry, we must use the old data_offset.
When we use the current (And a reshape is happening) we must use
the new_data_offset.

Signed-off-by: NeilBrown <neilb@suse.de>

2 files changed, 92 insertions, 23 deletions

diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index 36f445f9e11d..1c90005ab343 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -504,15 +504,13 @@ static void raid10_end_write_request(struct bio *bio, int error)
  * sector offset to a virtual address
  */
 
-static void raid10_find_phys(struct r10conf *conf, struct r10bio *r10bio)
+static void __raid10_find_phys(struct geom *geo, struct r10bio *r10bio)
 {
         int n,f;
         sector_t sector;
         sector_t chunk;
         sector_t stripe;
         int dev;
-        struct geom *geo = &conf->geo;
-
         int slot = 0;
 
         /* now calculate first sector/dev */
@@ -550,12 +548,29 @@ static void raid10_find_phys(struct r10conf *conf, struct r10bio *r10bio)
                         sector += (geo->chunk_mask + 1);
                 }
         }
-        BUG_ON(slot != conf->copies);
+}
+
+static void raid10_find_phys(struct r10conf *conf, struct r10bio *r10bio)
+{
+        struct geom *geo = &conf->geo;
+
+        if (conf->reshape_progress != MaxSector &&
+            ((r10bio->sector >= conf->reshape_progress) !=
+             conf->mddev->reshape_backwards)) {
+                set_bit(R10BIO_Previous, &r10bio->state);
+                geo = &conf->prev;
+        } else
+                clear_bit(R10BIO_Previous, &r10bio->state);
+
+        __raid10_find_phys(geo, r10bio);
 }
 
 static sector_t raid10_find_virt(struct r10conf *conf, sector_t sector, int dev)
 {
         sector_t offset, chunk, vchunk;
+        /* Never use conf->prev as this is only called during resync
+         * or recovery, so reshape isn't happening
+         */
         struct geom *geo = &conf->geo;
 
         offset = sector & geo->chunk_mask;
@@ -603,6 +618,11 @@ static int raid10_mergeable_bvec(struct request_queue *q,
         unsigned int bio_sectors = bvm->bi_size >> 9;
         struct geom *geo = &conf->geo;
 
+        if (conf->reshape_progress != MaxSector &&
+            ((sector >= conf->reshape_progress) !=
+             conf->mddev->reshape_backwards))
+                geo = &conf->prev;
+
         if (geo->near_copies < geo->raid_disks) {
                 max = (chunk_sectors - ((sector & (chunk_sectors - 1))
                                         + bio_sectors)) << 9;
@@ -617,6 +637,12 @@ static int raid10_mergeable_bvec(struct request_queue *q,
         if (mddev->merge_check_needed) {
                 struct r10bio r10_bio;
                 int s;
+                if (conf->reshape_progress != MaxSector) {
+                        /* Cannot give any guidance during reshape */
+                        if (max <= biovec->bv_len && bio_sectors == 0)
+                                return biovec->bv_len;
+                        return 0;
+                }
                 r10_bio.sector = sector;
                 raid10_find_phys(conf, &r10_bio);
                 rcu_read_lock();
@@ -816,7 +842,10 @@ static int raid10_congested(void *data, int bits)
         if (mddev_congested(mddev, bits))
                 return 1;
         rcu_read_lock();
-        for (i = 0; i < conf->geo.raid_disks && ret == 0; i++) {
+        for (i = 0;
+             (i < conf->geo.raid_disks || i < conf->prev.raid_disks)
+                     && ret == 0;
+             i++) {
                 struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
                 if (rdev && !test_bit(Faulty, &rdev->flags)) {
                         struct request_queue *q = bdev_get_queue(rdev->bdev);
@@ -977,13 +1006,23 @@ static void unfreeze_array(struct r10conf *conf)
         spin_unlock_irq(&conf->resync_lock);
 }
 
+static sector_t choose_data_offset(struct r10bio *r10_bio,
+                                   struct md_rdev *rdev)
+{
+        if (!test_bit(MD_RECOVERY_RESHAPE, &rdev->mddev->recovery) ||
+            test_bit(R10BIO_Previous, &r10_bio->state))
+                return rdev->data_offset;
+        else
+                return rdev->new_data_offset;
+}
+
 static void make_request(struct mddev *mddev, struct bio * bio)
 {
         struct r10conf *conf = mddev->private;
         struct r10bio *r10_bio;
         struct bio *read_bio;
         int i;
-        sector_t chunk_mask = conf->geo.chunk_mask;
+        sector_t chunk_mask = (conf->geo.chunk_mask & conf->prev.chunk_mask);
         int chunk_sects = chunk_mask + 1;
         const int rw = bio_data_dir(bio);
         const unsigned long do_sync = (bio->bi_rw & REQ_SYNC);
@@ -1004,7 +1043,8 @@ static void make_request(struct mddev *mddev, struct bio * bio)
          */
         if (unlikely((bio->bi_sector & chunk_mask) + (bio->bi_size >> 9)
                      > chunk_sects
-                     && conf->geo.near_copies < conf->geo.raid_disks)) {
+                     && (conf->geo.near_copies < conf->geo.raid_disks
+                         || conf->prev.near_copies < conf->prev.raid_disks))) {
                 struct bio_pair *bp;
                 /* Sanity check -- queue functions should prevent this happening */
                 if (bio->bi_vcnt != 1 ||
@@ -1098,7 +1138,7 @@ read_again:
                 r10_bio->devs[slot].rdev = rdev;
 
                 read_bio->bi_sector = r10_bio->devs[slot].addr +
-                        rdev->data_offset;
+                        choose_data_offset(r10_bio, rdev);
                 read_bio->bi_bdev = rdev->bdev;
                 read_bio->bi_end_io = raid10_end_read_request;
                 read_bio->bi_rw = READ | do_sync;
@@ -1302,7 +1342,8 @@ retry_write:
                 r10_bio->devs[i].bio = mbio;
 
                 mbio->bi_sector = (r10_bio->devs[i].addr+
-                                   conf->mirrors[d].rdev->data_offset);
+                                   choose_data_offset(r10_bio,
+                                                      conf->mirrors[d].rdev));
                 mbio->bi_bdev = conf->mirrors[d].rdev->bdev;
                 mbio->bi_end_io = raid10_end_write_request;
                 mbio->bi_rw = WRITE | do_sync | do_fua;
@@ -1326,8 +1367,10 @@ retry_write:
                  * so it cannot disappear, so the replacement cannot
                  * become NULL here
                  */
-                mbio->bi_sector = (r10_bio->devs[i].addr+
-                                   conf->mirrors[d].replacement->data_offset);
+                mbio->bi_sector = (r10_bio->devs[i].addr +
+                                   choose_data_offset(
+                                           r10_bio,
+                                           conf->mirrors[d].replacement));
                 mbio->bi_bdev = conf->mirrors[d].replacement->bdev;
                 mbio->bi_end_io = raid10_end_write_request;
                 mbio->bi_rw = WRITE | do_sync | do_fua;
@@ -1397,7 +1440,7 @@ static void status(struct seq_file *seq, struct mddev *mddev)
  * Don't consider the device numbered 'ignore'
  * as we might be about to remove it.
  */
-static int enough(struct r10conf *conf, int ignore)
+static int _enough(struct r10conf *conf, struct geom *geo, int ignore)
 {
         int first = 0;
 
@@ -1408,7 +1451,7 @@ static int enough(struct r10conf *conf, int ignore)
                         if (conf->mirrors[first].rdev &&
                             first != ignore)
                                 cnt++;
-                        first = (first+1) % conf->geo.raid_disks;
+                        first = (first+1) % geo->raid_disks;
                 }
                 if (cnt == 0)
                         return 0;
@@ -1416,6 +1459,12 @@ static int enough(struct r10conf *conf, int ignore)
         return 1;
 }
 
+static int enough(struct r10conf *conf, int ignore)
+{
+        return _enough(conf, &conf->geo, ignore) &&
+                _enough(conf, &conf->prev, ignore);
+}
+
 static void error(struct mddev *mddev, struct md_rdev *rdev)
 {
         char b[BDEVNAME_SIZE];
@@ -1548,7 +1597,7 @@ static int raid10_add_disk(struct mddev *mddev, struct md_rdev *rdev)
                  * very different from resync
                  */
                 return -EBUSY;
-        if (rdev->saved_raid_disk < 0 && !enough(conf, -1))
+        if (rdev->saved_raid_disk < 0 && !_enough(conf, &conf->prev, -1))
                 return -EINVAL;
 
         if (rdev->raid_disk >= 0)
@@ -2223,7 +2272,9 @@ static void fix_read_error(struct r10conf *conf, struct mddev *mddev, struct r10
                                        " (%d sectors at %llu on %s)\n",
                                        mdname(mddev), s,
                                        (unsigned long long)(
-                                               sect + rdev->data_offset),
+                                               sect +
+                                               choose_data_offset(r10_bio,
+                                                                  rdev)),
                                        bdevname(rdev->bdev, b));
                                 printk(KERN_NOTICE "md/raid10:%s: %s: failing "
                                        "drive\n",
@@ -2261,7 +2312,8 @@ static void fix_read_error(struct r10conf *conf, struct mddev *mddev, struct r10
                                        " (%d sectors at %llu on %s)\n",
                                        mdname(mddev), s,
                                        (unsigned long long)(
-                                               sect + rdev->data_offset),
+                                               sect +
+                                               choose_data_offset(r10_bio, rdev)),
                                        bdevname(rdev->bdev, b));
                                 printk(KERN_NOTICE "md/raid10:%s: %s: failing "
                                        "drive\n",
@@ -2274,7 +2326,8 @@ static void fix_read_error(struct r10conf *conf, struct mddev *mddev, struct r10
                                        " (%d sectors at %llu on %s)\n",
                                        mdname(mddev), s,
                                        (unsigned long long)(
-                                               sect + rdev->data_offset),
+                                               sect +
+                                               choose_data_offset(r10_bio, rdev)),
                                        bdevname(rdev->bdev, b));
                                 atomic_add(s, &rdev->corrected_errors);
                         }
@@ -2348,7 +2401,7 @@ static int narrow_write_error(struct r10bio *r10_bio, int i)
                 wbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
                 md_trim_bio(wbio, sector - bio->bi_sector, sectors);
                 wbio->bi_sector = (r10_bio->devs[i].addr+
-                                   rdev->data_offset+
+                                   choose_data_offset(r10_bio, rdev) +
                                    (sector - r10_bio->sector));
                 wbio->bi_bdev = rdev->bdev;
                 if (submit_bio_wait(WRITE, wbio) == 0)
@@ -2425,7 +2478,7 @@ read_more:
         r10_bio->devs[slot].bio = bio;
         r10_bio->devs[slot].rdev = rdev;
         bio->bi_sector = r10_bio->devs[slot].addr
-                + rdev->data_offset;
+                + choose_data_offset(r10_bio, rdev);
         bio->bi_bdev = rdev->bdev;
         bio->bi_rw = READ | do_sync;
         bio->bi_private = r10_bio;
@@ -3254,6 +3307,8 @@ static struct r10conf *setup_conf(struct mddev *mddev)
                 goto out;
 
         calc_sectors(conf, mddev->dev_sectors);
+        conf->prev = conf->geo;
+        conf->reshape_progress = MaxSector;
 
         spin_lock_init(&conf->device_lock);
         INIT_LIST_HEAD(&conf->retry_list);
@@ -3319,8 +3374,10 @@ static int run(struct mddev *mddev)
         rdev_for_each(rdev, mddev) {
 
                 disk_idx = rdev->raid_disk;
-                if (disk_idx >= conf->geo.raid_disks
-                    || disk_idx < 0)
+                if (disk_idx < 0)
+                        continue;
+                if (disk_idx >= conf->geo.raid_disks &&
+                    disk_idx >= conf->prev.raid_disks)
                         continue;
                 disk = conf->mirrors + disk_idx;
 
@@ -3347,7 +3404,10 @@ static int run(struct mddev *mddev)
         }
 
         mddev->degraded = 0;
-        for (i = 0; i < conf->geo.raid_disks; i++) {
+        for (i = 0;
+             i < conf->geo.raid_disks
+                     || i < conf->prev.raid_disks;
+             i++) {
 
                 disk = conf->mirrors + i;
 
@@ -3466,6 +3526,9 @@ static int raid10_resize(struct mddev *mddev, sector_t sectors)
         struct r10conf *conf = mddev->private;
         sector_t oldsize, size;
 
+        if (mddev->reshape_position != MaxSector)
+                return -EBUSY;
+
         if (conf->geo.far_copies > 1 && !conf->geo.far_offset)
                 return -EINVAL;
 
diff --git a/drivers/md/raid10.h b/drivers/md/raid10.h
index 4c4942ac46fc..37509d7134aa 100644
--- a/drivers/md/raid10.h
+++ b/drivers/md/raid10.h
@@ -34,13 +34,14 @@ struct r10conf {
                                                */
                 int             chunk_shift; /* shift from chunks to sectors */
                 sector_t        chunk_mask;
-        } geo;
+        } prev, geo;
         int                     copies;       /* near_copies * far_copies.
                                                * must be <= raid_disks
                                                */
 
         sector_t                dev_sectors;  /* temp copy of
                                                * mddev->dev_sectors */
+        sector_t                reshape_progress;
 
         struct list_head        retry_list;
         /* queue pending writes and submit them on unplug */
@@ -147,5 +148,10 @@ enum r10bio_state {
  */
         R10BIO_MadeGood,
         R10BIO_WriteError,
+/* During a reshape we might be performing IO on the
+ * 'previous' part of the array, in which case this
+ * flag is set
+ */
+        R10BIO_Previous,
 };
 #endif