]> git.sesse.net Git - bcachefs-tools-debian/blobdiff - libbcachefs/move.c
New upstream release
[bcachefs-tools-debian] / libbcachefs / move.c
index 05272673901d2ba93cccd69c0b1e9febf4042803..4f7d1758d8a97588a2e73a7397ea9880a356b414 100644 (file)
 #include "errcode.h"
 #include "error.h"
 #include "inode.h"
-#include "io.h"
+#include "io_read.h"
+#include "io_write.h"
 #include "journal_reclaim.h"
 #include "keylist.h"
 #include "move.h"
 #include "replicas.h"
+#include "snapshot.h"
 #include "super-io.h"
 #include "trace.h"
 
@@ -58,20 +60,6 @@ static void trace_move_extent_alloc_mem_fail2(struct bch_fs *c, struct bkey_s_c
        }
 }
 
-static void progress_list_add(struct bch_fs *c, struct bch_move_stats *stats)
-{
-       mutex_lock(&c->data_progress_lock);
-       list_add(&stats->list, &c->data_progress_list);
-       mutex_unlock(&c->data_progress_lock);
-}
-
-static void progress_list_del(struct bch_fs *c, struct bch_move_stats *stats)
-{
-       mutex_lock(&c->data_progress_lock);
-       list_del(&stats->list);
-       mutex_unlock(&c->data_progress_lock);
-}
-
 struct moving_io {
        struct list_head                read_list;
        struct list_head                io_list;
@@ -155,35 +143,31 @@ static void move_read_endio(struct bio *bio)
        closure_put(&ctxt->cl);
 }
 
-void bch2_moving_ctxt_do_pending_writes(struct moving_context *ctxt,
-                                       struct btree_trans *trans)
+void bch2_moving_ctxt_do_pending_writes(struct moving_context *ctxt)
 {
        struct moving_io *io;
 
-       if (trans)
-               bch2_trans_unlock(trans);
-
        while ((io = bch2_moving_ctxt_next_pending_write(ctxt))) {
+               bch2_trans_unlock_long(ctxt->trans);
                list_del(&io->read_list);
                move_write(io);
        }
 }
 
-static void bch2_move_ctxt_wait_for_io(struct moving_context *ctxt,
-                                      struct btree_trans *trans)
+void bch2_move_ctxt_wait_for_io(struct moving_context *ctxt)
 {
        unsigned sectors_pending = atomic_read(&ctxt->write_sectors);
 
-       move_ctxt_wait_event(ctxt, trans,
+       move_ctxt_wait_event(ctxt,
                !atomic_read(&ctxt->write_sectors) ||
                atomic_read(&ctxt->write_sectors) != sectors_pending);
 }
 
 void bch2_moving_ctxt_exit(struct moving_context *ctxt)
 {
-       struct bch_fs *c = ctxt->c;
+       struct bch_fs *c = ctxt->trans->c;
 
-       move_ctxt_wait_event(ctxt, NULL, list_empty(&ctxt->reads));
+       move_ctxt_wait_event(ctxt, list_empty(&ctxt->reads));
        closure_sync(&ctxt->cl);
 
        EBUG_ON(atomic_read(&ctxt->write_sectors));
@@ -191,16 +175,12 @@ void bch2_moving_ctxt_exit(struct moving_context *ctxt)
        EBUG_ON(atomic_read(&ctxt->read_sectors));
        EBUG_ON(atomic_read(&ctxt->read_ios));
 
-       if (ctxt->stats) {
-               progress_list_del(c, ctxt->stats);
-               trace_move_data(c,
-                               atomic64_read(&ctxt->stats->sectors_moved),
-                               atomic64_read(&ctxt->stats->keys_moved));
-       }
-
        mutex_lock(&c->moving_context_lock);
        list_del(&ctxt->list);
        mutex_unlock(&c->moving_context_lock);
+
+       bch2_trans_put(ctxt->trans);
+       memset(ctxt, 0, sizeof(*ctxt));
 }
 
 void bch2_moving_ctxt_init(struct moving_context *ctxt,
@@ -212,7 +192,7 @@ void bch2_moving_ctxt_init(struct moving_context *ctxt,
 {
        memset(ctxt, 0, sizeof(*ctxt));
 
-       ctxt->c         = c;
+       ctxt->trans     = bch2_trans_get(c);
        ctxt->fn        = (void *) _RET_IP_;
        ctxt->rate      = rate;
        ctxt->stats     = stats;
@@ -229,16 +209,17 @@ void bch2_moving_ctxt_init(struct moving_context *ctxt,
        mutex_lock(&c->moving_context_lock);
        list_add(&ctxt->list, &c->moving_context_list);
        mutex_unlock(&c->moving_context_lock);
+}
 
-       if (stats) {
-               progress_list_add(c, stats);
-               stats->data_type = BCH_DATA_user;
-       }
+void bch2_move_stats_exit(struct bch_move_stats *stats, struct bch_fs *c)
+{
+       trace_move_data(c, stats);
 }
 
 void bch2_move_stats_init(struct bch_move_stats *stats, char *name)
 {
        memset(stats, 0, sizeof(*stats));
+       stats->data_type = BCH_DATA_user;
        scnprintf(stats->name, sizeof(stats->name), "%s", name);
 }
 
@@ -282,18 +263,17 @@ static int bch2_extent_drop_ptrs(struct btree_trans *trans,
 
        return bch2_trans_relock(trans) ?:
                bch2_trans_update(trans, iter, n, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
-               bch2_trans_commit(trans, NULL, NULL, BTREE_INSERT_NOFAIL);
+               bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc);
 }
 
-static int bch2_move_extent(struct btree_trans *trans,
-                           struct btree_iter *iter,
-                           struct moving_context *ctxt,
-                           struct move_bucket_in_flight *bucket_in_flight,
-                           struct bch_io_opts io_opts,
-                           enum btree_id btree_id,
-                           struct bkey_s_c k,
-                           struct data_update_opts data_opts)
+int bch2_move_extent(struct moving_context *ctxt,
+                    struct move_bucket_in_flight *bucket_in_flight,
+                    struct btree_iter *iter,
+                    struct bkey_s_c k,
+                    struct bch_io_opts io_opts,
+                    struct data_update_opts data_opts)
 {
+       struct btree_trans *trans = ctxt->trans;
        struct bch_fs *c = trans->c;
        struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
        struct moving_io *io;
@@ -302,6 +282,8 @@ static int bch2_move_extent(struct btree_trans *trans,
        unsigned sectors = k.k->size, pages;
        int ret = -ENOMEM;
 
+       if (ctxt->stats)
+               ctxt->stats->pos = BBPOS(iter->btree_id, iter->pos);
        trace_move_extent2(c, k);
 
        bch2_data_update_opts_normalize(k, &data_opts);
@@ -354,7 +336,7 @@ static int bch2_move_extent(struct btree_trans *trans,
        io->rbio.bio.bi_end_io          = move_read_endio;
 
        ret = bch2_data_update_init(trans, ctxt, &io->write, ctxt->wp,
-                                   io_opts, data_opts, btree_id, k);
+                                   io_opts, data_opts, iter->btree_id, k);
        if (ret && ret != -BCH_ERR_unwritten_extent_update)
                goto err_free_pages;
 
@@ -366,9 +348,11 @@ static int bch2_move_extent(struct btree_trans *trans,
 
        BUG_ON(ret);
 
-       io->write.ctxt = ctxt;
        io->write.op.end_io = move_write_done;
 
+       if (ctxt->rate)
+               bch2_ratelimit_increment(ctxt->rate, k.k->size);
+
        if (ctxt->stats) {
                atomic64_inc(&ctxt->stats->keys_moved);
                atomic64_add(k.k->size, &ctxt->stats->sectors_moved);
@@ -398,7 +382,7 @@ static int bch2_move_extent(struct btree_trans *trans,
        closure_get(&ctxt->cl);
        bch2_read_extent(trans, &io->rbio,
                         bkey_start_pos(k.k),
-                        btree_id, k, 0,
+                        iter->btree_id, k, 0,
                         BCH_READ_NODECODE|
                         BCH_READ_LAST_FRAGMENT);
        return 0;
@@ -412,45 +396,96 @@ err:
        return ret;
 }
 
-static int lookup_inode(struct btree_trans *trans, struct bpos pos,
-                       struct bch_inode_unpacked *inode)
+struct bch_io_opts *bch2_move_get_io_opts(struct btree_trans *trans,
+                         struct per_snapshot_io_opts *io_opts,
+                         struct bkey_s_c extent_k)
+{
+       struct bch_fs *c = trans->c;
+       u32 restart_count = trans->restart_count;
+       int ret = 0;
+
+       if (io_opts->cur_inum != extent_k.k->p.inode) {
+               struct btree_iter iter;
+               struct bkey_s_c k;
+
+               io_opts->d.nr = 0;
+
+               for_each_btree_key(trans, iter, BTREE_ID_inodes, POS(0, extent_k.k->p.inode),
+                                  BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
+                       if (k.k->p.offset != extent_k.k->p.inode)
+                               break;
+
+                       if (!bkey_is_inode(k.k))
+                               continue;
+
+                       struct bch_inode_unpacked inode;
+                       BUG_ON(bch2_inode_unpack(k, &inode));
+
+                       struct snapshot_io_opts_entry e = { .snapshot = k.k->p.snapshot };
+                       bch2_inode_opts_get(&e.io_opts, trans->c, &inode);
+
+                       ret = darray_push(&io_opts->d, e);
+                       if (ret)
+                               break;
+               }
+               bch2_trans_iter_exit(trans, &iter);
+               io_opts->cur_inum = extent_k.k->p.inode;
+       }
+
+       ret = ret ?: trans_was_restarted(trans, restart_count);
+       if (ret)
+               return ERR_PTR(ret);
+
+       if (extent_k.k->p.snapshot) {
+               struct snapshot_io_opts_entry *i;
+               darray_for_each(io_opts->d, i)
+                       if (bch2_snapshot_is_ancestor(c, extent_k.k->p.snapshot, i->snapshot))
+                               return &i->io_opts;
+       }
+
+       return &io_opts->fs_io_opts;
+}
+
+int bch2_move_get_io_opts_one(struct btree_trans *trans,
+                             struct bch_io_opts *io_opts,
+                             struct bkey_s_c extent_k)
 {
        struct btree_iter iter;
        struct bkey_s_c k;
        int ret;
 
-       bch2_trans_iter_init(trans, &iter, BTREE_ID_inodes, pos,
-                            BTREE_ITER_ALL_SNAPSHOTS);
-       k = bch2_btree_iter_peek(&iter);
+       /* reflink btree? */
+       if (!extent_k.k->p.inode) {
+               *io_opts = bch2_opts_to_inode_opts(trans->c->opts);
+               return 0;
+       }
+
+       k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes,
+                              SPOS(0, extent_k.k->p.inode, extent_k.k->p.snapshot),
+                              BTREE_ITER_CACHED);
        ret = bkey_err(k);
-       if (ret)
-               goto err;
+       if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
+               return ret;
 
-       if (!k.k || !bkey_eq(k.k->p, pos)) {
-               ret = -BCH_ERR_ENOENT_inode;
-               goto err;
+       if (!ret && bkey_is_inode(k.k)) {
+               struct bch_inode_unpacked inode;
+               bch2_inode_unpack(k, &inode);
+               bch2_inode_opts_get(io_opts, trans->c, &inode);
+       } else {
+               *io_opts = bch2_opts_to_inode_opts(trans->c->opts);
        }
 
-       ret = bkey_is_inode(k.k) ? 0 : -EIO;
-       if (ret)
-               goto err;
-
-       ret = bch2_inode_unpack(k, inode);
-       if (ret)
-               goto err;
-err:
        bch2_trans_iter_exit(trans, &iter);
-       return ret;
+       return 0;
 }
 
-static int move_ratelimit(struct btree_trans *trans,
-                         struct moving_context *ctxt)
+int bch2_move_ratelimit(struct moving_context *ctxt)
 {
-       struct bch_fs *c = trans->c;
+       struct bch_fs *c = ctxt->trans->c;
        u64 delay;
 
-       if (ctxt->wait_on_copygc) {
-               bch2_trans_unlock(trans);
+       if (ctxt->wait_on_copygc && !c->copygc_running) {
+               bch2_trans_unlock_long(ctxt->trans);
                wait_event_killable(c->copygc_running_wq,
                                    !c->copygc_running ||
                                    kthread_should_stop());
@@ -459,8 +494,12 @@ static int move_ratelimit(struct btree_trans *trans,
        do {
                delay = ctxt->rate ? bch2_ratelimit_delay(ctxt->rate) : 0;
 
+
                if (delay) {
-                       bch2_trans_unlock(trans);
+                       if (delay > HZ / 10)
+                               bch2_trans_unlock_long(ctxt->trans);
+                       else
+                               bch2_trans_unlock(ctxt->trans);
                        set_current_state(TASK_INTERRUPTIBLE);
                }
 
@@ -473,7 +512,7 @@ static int move_ratelimit(struct btree_trans *trans,
                        schedule_timeout(delay);
 
                if (unlikely(freezing(current))) {
-                       move_ctxt_wait_event(ctxt, trans, list_empty(&ctxt->reads));
+                       move_ctxt_wait_event(ctxt, list_empty(&ctxt->reads));
                        try_to_freeze();
                }
        } while (delay);
@@ -482,7 +521,7 @@ static int move_ratelimit(struct btree_trans *trans,
         * XXX: these limits really ought to be per device, SSDs and hard drives
         * will want different limits
         */
-       move_ctxt_wait_event(ctxt, trans,
+       move_ctxt_wait_event(ctxt,
                atomic_read(&ctxt->write_sectors) < c->opts.move_bytes_in_flight >> 9 &&
                atomic_read(&ctxt->read_sectors) < c->opts.move_bytes_in_flight >> 9 &&
                atomic_read(&ctxt->write_ios) < c->opts.move_ios_in_flight &&
@@ -491,64 +530,39 @@ static int move_ratelimit(struct btree_trans *trans,
        return 0;
 }
 
-static int move_get_io_opts(struct btree_trans *trans,
-                           struct bch_io_opts *io_opts,
-                           struct bkey_s_c k, u64 *cur_inum)
-{
-       struct bch_inode_unpacked inode;
-       int ret;
-
-       if (*cur_inum == k.k->p.inode)
-               return 0;
-
-       ret = lookup_inode(trans,
-                          SPOS(0, k.k->p.inode, k.k->p.snapshot),
-                          &inode);
-       if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
-               return ret;
-
-       if (!ret)
-               bch2_inode_opts_get(io_opts, trans->c, &inode);
-       else
-               *io_opts = bch2_opts_to_inode_opts(trans->c->opts);
-       *cur_inum = k.k->p.inode;
-       return 0;
-}
-
-static int __bch2_move_data(struct moving_context *ctxt,
-                           struct bpos start,
-                           struct bpos end,
-                           move_pred_fn pred, void *arg,
-                           enum btree_id btree_id)
+static int bch2_move_data_btree(struct moving_context *ctxt,
+                               struct bpos start,
+                               struct bpos end,
+                               move_pred_fn pred, void *arg,
+                               enum btree_id btree_id)
 {
-       struct bch_fs *c = ctxt->c;
-       struct bch_io_opts io_opts = bch2_opts_to_inode_opts(c->opts);
+       struct btree_trans *trans = ctxt->trans;
+       struct bch_fs *c = trans->c;
+       struct per_snapshot_io_opts snapshot_io_opts;
+       struct bch_io_opts *io_opts;
        struct bkey_buf sk;
-       struct btree_trans trans;
        struct btree_iter iter;
        struct bkey_s_c k;
        struct data_update_opts data_opts;
-       u64 cur_inum = U64_MAX;
        int ret = 0, ret2;
 
+       per_snapshot_io_opts_init(&snapshot_io_opts, c);
        bch2_bkey_buf_init(&sk);
-       bch2_trans_init(&trans, c, 0, 0);
 
        if (ctxt->stats) {
                ctxt->stats->data_type  = BCH_DATA_user;
-               ctxt->stats->btree_id   = btree_id;
-               ctxt->stats->pos        = start;
+               ctxt->stats->pos        = BBPOS(btree_id, start);
        }
 
-       bch2_trans_iter_init(&trans, &iter, btree_id, start,
+       bch2_trans_iter_init(trans, &iter, btree_id, start,
                             BTREE_ITER_PREFETCH|
                             BTREE_ITER_ALL_SNAPSHOTS);
 
        if (ctxt->rate)
                bch2_ratelimit_reset(ctxt->rate);
 
-       while (!move_ratelimit(&trans, ctxt)) {
-               bch2_trans_begin(&trans);
+       while (!bch2_move_ratelimit(ctxt)) {
+               bch2_trans_begin(trans);
 
                k = bch2_btree_iter_peek(&iter);
                if (!k.k)
@@ -564,17 +578,18 @@ static int __bch2_move_data(struct moving_context *ctxt,
                        break;
 
                if (ctxt->stats)
-                       ctxt->stats->pos = iter.pos;
+                       ctxt->stats->pos = BBPOS(iter.btree_id, iter.pos);
 
                if (!bkey_extent_is_direct_data(k.k))
                        goto next_nondata;
 
-               ret = move_get_io_opts(&trans, &io_opts, k, &cur_inum);
+               io_opts = bch2_move_get_io_opts(trans, &snapshot_io_opts, k);
+               ret = PTR_ERR_OR_ZERO(io_opts);
                if (ret)
                        continue;
 
                memset(&data_opts, 0, sizeof(data_opts));
-               if (!pred(c, arg, k, &io_opts, &data_opts))
+               if (!pred(c, arg, k, io_opts, &data_opts))
                        goto next;
 
                /*
@@ -584,24 +599,20 @@ static int __bch2_move_data(struct moving_context *ctxt,
                bch2_bkey_buf_reassemble(&sk, c, k);
                k = bkey_i_to_s_c(sk.k);
 
-               ret2 = bch2_move_extent(&trans, &iter, ctxt, NULL,
-                                       io_opts, btree_id, k, data_opts);
+               ret2 = bch2_move_extent(ctxt, NULL, &iter, k, *io_opts, data_opts);
                if (ret2) {
                        if (bch2_err_matches(ret2, BCH_ERR_transaction_restart))
                                continue;
 
                        if (ret2 == -ENOMEM) {
                                /* memory allocation failure, wait for some IO to finish */
-                               bch2_move_ctxt_wait_for_io(ctxt, &trans);
+                               bch2_move_ctxt_wait_for_io(ctxt);
                                continue;
                        }
 
                        /* XXX signal failure */
                        goto next;
                }
-
-               if (ctxt->rate)
-                       bch2_ratelimit_increment(ctxt->rate, k.k->size);
 next:
                if (ctxt->stats)
                        atomic64_add(k.k->size, &ctxt->stats->sectors_seen);
@@ -609,60 +620,69 @@ next_nondata:
                bch2_btree_iter_advance(&iter);
        }
 
-       bch2_trans_iter_exit(&trans, &iter);
-       bch2_trans_exit(&trans);
+       bch2_trans_iter_exit(trans, &iter);
        bch2_bkey_buf_exit(&sk, c);
+       per_snapshot_io_opts_exit(&snapshot_io_opts);
 
        return ret;
 }
 
-int bch2_move_data(struct bch_fs *c,
-                  enum btree_id start_btree_id, struct bpos start_pos,
-                  enum btree_id end_btree_id,   struct bpos end_pos,
-                  struct bch_ratelimit *rate,
-                  struct bch_move_stats *stats,
-                  struct write_point_specifier wp,
-                  bool wait_on_copygc,
-                  move_pred_fn pred, void *arg)
+int __bch2_move_data(struct moving_context *ctxt,
+                    struct bbpos start,
+                    struct bbpos end,
+                    move_pred_fn pred, void *arg)
 {
-       struct moving_context ctxt;
+       struct bch_fs *c = ctxt->trans->c;
        enum btree_id id;
-       int ret;
-
-       bch2_moving_ctxt_init(&ctxt, c, rate, stats, wp, wait_on_copygc);
+       int ret = 0;
 
-       for (id = start_btree_id;
-            id <= min_t(unsigned, end_btree_id, btree_id_nr_alive(c) - 1);
+       for (id = start.btree;
+            id <= min_t(unsigned, end.btree, btree_id_nr_alive(c) - 1);
             id++) {
-               stats->btree_id = id;
-
-               if (id != BTREE_ID_extents &&
-                   id != BTREE_ID_reflink)
-                       continue;
+               ctxt->stats->pos = BBPOS(id, POS_MIN);
 
-               if (!bch2_btree_id_root(c, id)->b)
+               if (!btree_type_has_ptrs(id) ||
+                   !bch2_btree_id_root(c, id)->b)
                        continue;
 
-               ret = __bch2_move_data(&ctxt,
-                                      id == start_btree_id ? start_pos : POS_MIN,
-                                      id == end_btree_id   ? end_pos   : POS_MAX,
+               ret = bch2_move_data_btree(ctxt,
+                                      id == start.btree ? start.pos : POS_MIN,
+                                      id == end.btree   ? end.pos   : POS_MAX,
                                       pred, arg, id);
                if (ret)
                        break;
        }
 
+       return ret;
+}
+
+int bch2_move_data(struct bch_fs *c,
+                  struct bbpos start,
+                  struct bbpos end,
+                  struct bch_ratelimit *rate,
+                  struct bch_move_stats *stats,
+                  struct write_point_specifier wp,
+                  bool wait_on_copygc,
+                  move_pred_fn pred, void *arg)
+{
+
+       struct moving_context ctxt;
+       int ret;
+
+       bch2_moving_ctxt_init(&ctxt, c, rate, stats, wp, wait_on_copygc);
+       ret = __bch2_move_data(&ctxt, start, end, pred, arg);
        bch2_moving_ctxt_exit(&ctxt);
 
        return ret;
 }
 
-int __bch2_evacuate_bucket(struct btree_trans *trans,
-                          struct moving_context *ctxt,
+int __bch2_evacuate_bucket(struct moving_context *ctxt,
                           struct move_bucket_in_flight *bucket_in_flight,
                           struct bpos bucket, int gen,
                           struct data_update_opts _data_opts)
 {
-       struct bch_fs *c = ctxt->c;
+       struct btree_trans *trans = ctxt->trans;
+       struct bch_fs *c = trans->c;
        struct bch_io_opts io_opts = bch2_opts_to_inode_opts(c->opts);
        struct btree_iter iter;
        struct bkey_buf sk;
@@ -673,7 +693,6 @@ int __bch2_evacuate_bucket(struct btree_trans *trans,
        struct data_update_opts data_opts;
        unsigned dirty_sectors, bucket_size;
        u64 fragmentation;
-       u64 cur_inum = U64_MAX;
        struct bpos bp_pos = POS_MIN;
        int ret = 0;
 
@@ -708,7 +727,7 @@ int __bch2_evacuate_bucket(struct btree_trans *trans,
                goto err;
        }
 
-       while (!(ret = move_ratelimit(trans, ctxt))) {
+       while (!(ret = bch2_move_ratelimit(ctxt))) {
                bch2_trans_begin(trans);
 
                ret = bch2_get_next_backpointer(trans, bucket, gen,
@@ -723,7 +742,6 @@ int __bch2_evacuate_bucket(struct btree_trans *trans,
 
                if (!bp.level) {
                        const struct bch_extent_ptr *ptr;
-                       struct bkey_s_c k;
                        unsigned i = 0;
 
                        k = bch2_backpointer_get_key(trans, &iter, bp_pos, bp, 0);
@@ -738,7 +756,7 @@ int __bch2_evacuate_bucket(struct btree_trans *trans,
                        bch2_bkey_buf_reassemble(&sk, c, k);
                        k = bkey_i_to_s_c(sk.k);
 
-                       ret = move_get_io_opts(trans, &io_opts, k, &cur_inum);
+                       ret = bch2_move_get_io_opts_one(trans, &io_opts, k);
                        if (ret) {
                                bch2_trans_iter_exit(trans, &iter);
                                continue;
@@ -759,23 +777,20 @@ int __bch2_evacuate_bucket(struct btree_trans *trans,
                                i++;
                        }
 
-                       ret = bch2_move_extent(trans, &iter, ctxt,
-                                       bucket_in_flight,
-                                       io_opts, bp.btree_id, k, data_opts);
+                       ret = bch2_move_extent(ctxt, bucket_in_flight,
+                                              &iter, k, io_opts, data_opts);
                        bch2_trans_iter_exit(trans, &iter);
 
                        if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
                                continue;
                        if (ret == -ENOMEM) {
                                /* memory allocation failure, wait for some IO to finish */
-                               bch2_move_ctxt_wait_for_io(ctxt, trans);
+                               bch2_move_ctxt_wait_for_io(ctxt);
                                continue;
                        }
                        if (ret)
                                goto err;
 
-                       if (ctxt->rate)
-                               bch2_ratelimit_increment(ctxt->rate, k.k->size);
                        if (ctxt->stats)
                                atomic64_add(k.k->size, &ctxt->stats->sectors_seen);
                } else {
@@ -826,15 +841,12 @@ int bch2_evacuate_bucket(struct bch_fs *c,
                         struct write_point_specifier wp,
                         bool wait_on_copygc)
 {
-       struct btree_trans trans;
        struct moving_context ctxt;
        int ret;
 
-       bch2_trans_init(&trans, c, 0, 0);
        bch2_moving_ctxt_init(&ctxt, c, rate, stats, wp, wait_on_copygc);
-       ret = __bch2_evacuate_bucket(&trans, &ctxt, NULL, bucket, gen, data_opts);
+       ret = __bch2_evacuate_bucket(&ctxt, NULL, bucket, gen, data_opts);
        bch2_moving_ctxt_exit(&ctxt);
-       bch2_trans_exit(&trans);
 
        return ret;
 }
@@ -851,31 +863,34 @@ static int bch2_move_btree(struct bch_fs *c,
 {
        bool kthread = (current->flags & PF_KTHREAD) != 0;
        struct bch_io_opts io_opts = bch2_opts_to_inode_opts(c->opts);
-       struct btree_trans trans;
+       struct moving_context ctxt;
+       struct btree_trans *trans;
        struct btree_iter iter;
        struct btree *b;
        enum btree_id id;
        struct data_update_opts data_opts;
        int ret = 0;
 
-       bch2_trans_init(&trans, c, 0, 0);
-       progress_list_add(c, stats);
+       bch2_moving_ctxt_init(&ctxt, c, NULL, stats,
+                             writepoint_ptr(&c->btree_write_point),
+                             true);
+       trans = ctxt.trans;
 
        stats->data_type = BCH_DATA_btree;
 
        for (id = start_btree_id;
             id <= min_t(unsigned, end_btree_id, btree_id_nr_alive(c) - 1);
             id++) {
-               stats->btree_id = id;
+               stats->pos = BBPOS(id, POS_MIN);
 
                if (!bch2_btree_id_root(c, id)->b)
                        continue;
 
-               bch2_trans_node_iter_init(&trans, &iter, id, POS_MIN, 0, 0,
+               bch2_trans_node_iter_init(trans, &iter, id, POS_MIN, 0, 0,
                                          BTREE_ITER_PREFETCH);
 retry:
                ret = 0;
-               while (bch2_trans_begin(&trans),
+               while (bch2_trans_begin(trans),
                       (b = bch2_btree_iter_peek_node(&iter)) &&
                       !(ret = PTR_ERR_OR_ZERO(b))) {
                        if (kthread && kthread_should_stop())
@@ -885,12 +900,12 @@ retry:
                             bpos_cmp(b->key.k.p, end_pos)) > 0)
                                break;
 
-                       stats->pos = iter.pos;
+                       stats->pos = BBPOS(iter.btree_id, iter.pos);
 
                        if (!pred(c, arg, b, &io_opts, &data_opts))
                                goto next;
 
-                       ret = bch2_btree_node_rewrite(&trans, &iter, b, 0) ?: ret;
+                       ret = bch2_btree_node_rewrite(trans, &iter, b, 0) ?: ret;
                        if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
                                continue;
                        if (ret)
@@ -901,20 +916,16 @@ next:
                if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
                        goto retry;
 
-               bch2_trans_iter_exit(&trans, &iter);
+               bch2_trans_iter_exit(trans, &iter);
 
                if (kthread && kthread_should_stop())
                        break;
        }
 
-       bch2_trans_exit(&trans);
-
-       if (ret)
-               bch_err_fn(c, ret);
-
+       bch_err_fn(c, ret);
+       bch2_moving_ctxt_exit(&ctxt);
        bch2_btree_interior_updates_flush(c);
 
-       progress_list_del(c, stats);
        return ret;
 }
 
@@ -1035,8 +1046,7 @@ int bch2_scan_old_btree_nodes(struct bch_fs *c, struct bch_move_stats *stats)
                mutex_unlock(&c->sb_lock);
        }
 
-       if (ret)
-               bch_err_fn(c, ret);
+       bch_err_fn(c, ret);
        return ret;
 }
 
@@ -1059,14 +1069,16 @@ int bch2_data_job(struct bch_fs *c,
                ret = bch2_replicas_gc2(c) ?: ret;
 
                ret = bch2_move_data(c,
-                                    op.start_btree,    op.start_pos,
-                                    op.end_btree,      op.end_pos,
+                                    (struct bbpos) { op.start_btree,   op.start_pos },
+                                    (struct bbpos) { op.end_btree,     op.end_pos },
                                     NULL,
                                     stats,
                                     writepoint_hashed((unsigned long) current),
                                     true,
                                     rereplicate_pred, c) ?: ret;
                ret = bch2_replicas_gc2(c) ?: ret;
+
+               bch2_move_stats_exit(stats, c);
                break;
        case BCH_DATA_OP_MIGRATE:
                if (op.migrate.dev >= c->sb.nr_devices)
@@ -1083,18 +1095,21 @@ int bch2_data_job(struct bch_fs *c,
                ret = bch2_replicas_gc2(c) ?: ret;
 
                ret = bch2_move_data(c,
-                                    op.start_btree,    op.start_pos,
-                                    op.end_btree,      op.end_pos,
+                                    (struct bbpos) { op.start_btree,   op.start_pos },
+                                    (struct bbpos) { op.end_btree,     op.end_pos },
                                     NULL,
                                     stats,
                                     writepoint_hashed((unsigned long) current),
                                     true,
                                     migrate_pred, &op) ?: ret;
                ret = bch2_replicas_gc2(c) ?: ret;
+
+               bch2_move_stats_exit(stats, c);
                break;
        case BCH_DATA_OP_REWRITE_OLD_NODES:
                bch2_move_stats_init(stats, "rewrite_old_nodes");
                ret = bch2_scan_old_btree_nodes(c, stats);
+               bch2_move_stats_exit(stats, c);
                break;
        default:
                ret = -EINVAL;
@@ -1103,46 +1118,64 @@ int bch2_data_job(struct bch_fs *c,
        return ret;
 }
 
-void bch2_data_jobs_to_text(struct printbuf *out, struct bch_fs *c)
+void bch2_move_stats_to_text(struct printbuf *out, struct bch_move_stats *stats)
 {
-       struct bch_move_stats *stats;
-
-       mutex_lock(&c->data_progress_lock);
-       list_for_each_entry(stats, &c->data_progress_list, list) {
-               prt_printf(out, "%s: data type %s btree_id %s position: ",
-                      stats->name,
-                      bch2_data_types[stats->data_type],
-                      bch2_btree_ids[stats->btree_id]);
-               bch2_bpos_to_text(out, stats->pos);
-               prt_printf(out, "%s", "\n");
-       }
-       mutex_unlock(&c->data_progress_lock);
+       prt_printf(out, "%s: data type=%s pos=",
+                  stats->name,
+                  bch2_data_types[stats->data_type]);
+       bch2_bbpos_to_text(out, stats->pos);
+       prt_newline(out);
+       printbuf_indent_add(out, 2);
+
+       prt_str(out, "keys moved:  ");
+       prt_u64(out, atomic64_read(&stats->keys_moved));
+       prt_newline(out);
+
+       prt_str(out, "keys raced:  ");
+       prt_u64(out, atomic64_read(&stats->keys_raced));
+       prt_newline(out);
+
+       prt_str(out, "bytes seen:  ");
+       prt_human_readable_u64(out, atomic64_read(&stats->sectors_seen) << 9);
+       prt_newline(out);
+
+       prt_str(out, "bytes moved: ");
+       prt_human_readable_u64(out, atomic64_read(&stats->sectors_moved) << 9);
+       prt_newline(out);
+
+       prt_str(out, "bytes raced: ");
+       prt_human_readable_u64(out, atomic64_read(&stats->sectors_raced) << 9);
+       prt_newline(out);
+
+       printbuf_indent_sub(out, 2);
 }
 
-static void bch2_moving_ctxt_to_text(struct printbuf *out, struct moving_context *ctxt)
+static void bch2_moving_ctxt_to_text(struct printbuf *out, struct bch_fs *c, struct moving_context *ctxt)
 {
        struct moving_io *io;
 
-       prt_printf(out, "%ps:", ctxt->fn);
-       prt_newline(out);
+       bch2_move_stats_to_text(out, ctxt->stats);
        printbuf_indent_add(out, 2);
 
-       prt_printf(out, "reads: %u sectors %u",
+       prt_printf(out, "reads: ios %u/%u sectors %u/%u",
                   atomic_read(&ctxt->read_ios),
-                  atomic_read(&ctxt->read_sectors));
+                  c->opts.move_ios_in_flight,
+                  atomic_read(&ctxt->read_sectors),
+                  c->opts.move_bytes_in_flight >> 9);
        prt_newline(out);
 
-       prt_printf(out, "writes: %u sectors %u",
+       prt_printf(out, "writes: ios %u/%u sectors %u/%u",
                   atomic_read(&ctxt->write_ios),
-                  atomic_read(&ctxt->write_sectors));
+                  c->opts.move_ios_in_flight,
+                  atomic_read(&ctxt->write_sectors),
+                  c->opts.move_bytes_in_flight >> 9);
        prt_newline(out);
 
        printbuf_indent_add(out, 2);
 
        mutex_lock(&ctxt->lock);
-       list_for_each_entry(io, &ctxt->ios, io_list) {
+       list_for_each_entry(io, &ctxt->ios, io_list)
                bch2_write_op_to_text(out, &io->write.op);
-       }
        mutex_unlock(&ctxt->lock);
 
        printbuf_indent_sub(out, 4);
@@ -1154,7 +1187,7 @@ void bch2_fs_moving_ctxts_to_text(struct printbuf *out, struct bch_fs *c)
 
        mutex_lock(&c->moving_context_lock);
        list_for_each_entry(ctxt, &c->moving_context_list, list)
-               bch2_moving_ctxt_to_text(out, ctxt);
+               bch2_moving_ctxt_to_text(out, c, ctxt);
        mutex_unlock(&c->moving_context_lock);
 }
 
@@ -1162,7 +1195,4 @@ void bch2_fs_move_init(struct bch_fs *c)
 {
        INIT_LIST_HEAD(&c->moving_context_list);
        mutex_init(&c->moving_context_lock);
-
-       INIT_LIST_HEAD(&c->data_progress_list);
-       mutex_init(&c->data_progress_lock);
 }