-struct moving_context {
- /* Closure for waiting on all reads and writes to complete */
- struct closure cl;
-
- struct bch_move_stats *stats;
-
- struct list_head reads;
-
- /* in flight sectors: */
- atomic_t read_sectors;
- atomic_t write_sectors;
-
- wait_queue_head_t wait;
-};
-
-static int insert_snapshot_whiteouts(struct btree_trans *trans,
- enum btree_id id,
- struct bpos old_pos,
- struct bpos new_pos)
-{
- struct bch_fs *c = trans->c;
- struct btree_iter iter, update_iter;
- struct bkey_s_c k;
- struct snapshots_seen s;
- int ret;
-
- if (!btree_type_has_snapshots(id))
- return 0;
-
- snapshots_seen_init(&s);
-
- if (!bkey_cmp(old_pos, new_pos))
- return 0;
-
- if (!snapshot_t(c, old_pos.snapshot)->children[0])
- return 0;
-
- bch2_trans_iter_init(trans, &iter, id, old_pos,
- BTREE_ITER_NOT_EXTENTS|
- BTREE_ITER_ALL_SNAPSHOTS);
- while (1) {
-next:
- k = bch2_btree_iter_prev(&iter);
- ret = bkey_err(k);
- if (ret)
- break;
-
- if (bkey_cmp(old_pos, k.k->p))
- break;
-
- if (bch2_snapshot_is_ancestor(c, k.k->p.snapshot, old_pos.snapshot)) {
- struct bkey_i *update;
- u32 *i;
-
- darray_for_each(s.ids, i)
- if (bch2_snapshot_is_ancestor(c, k.k->p.snapshot, *i))
- goto next;
-
- update = bch2_trans_kmalloc(trans, sizeof(struct bkey_i));
-
- ret = PTR_ERR_OR_ZERO(update);
- if (ret)
- break;
-
- bkey_init(&update->k);
- update->k.p = new_pos;
- update->k.p.snapshot = k.k->p.snapshot;
-
- bch2_trans_iter_init(trans, &update_iter, id, update->k.p,
- BTREE_ITER_NOT_EXTENTS|
- BTREE_ITER_ALL_SNAPSHOTS|
- BTREE_ITER_INTENT);
- ret = bch2_btree_iter_traverse(&update_iter) ?:
- bch2_trans_update(trans, &update_iter, update,
- BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
- bch2_trans_iter_exit(trans, &update_iter);
- if (ret)
- break;
-
- ret = snapshots_seen_add(c, &s, k.k->p.snapshot);
- if (ret)
- break;
- }
- }
- bch2_trans_iter_exit(trans, &iter);
- darray_exit(&s.ids);
-
- return ret;
-}
-
-static int bch2_migrate_index_update(struct bch_write_op *op)
-{
- struct bch_fs *c = op->c;
- struct btree_trans trans;
- struct btree_iter iter;
- struct migrate_write *m =
- container_of(op, struct migrate_write, op);
- struct open_bucket *ec_ob = ec_open_bucket(c, &op->open_buckets);
- struct keylist *keys = &op->insert_keys;
- struct bkey_buf _new, _insert;
- int ret = 0;
-
- bch2_bkey_buf_init(&_new);
- bch2_bkey_buf_init(&_insert);
- bch2_bkey_buf_realloc(&_insert, c, U8_MAX);
-
- bch2_trans_init(&trans, c, BTREE_ITER_MAX, 1024);
-
- bch2_trans_iter_init(&trans, &iter, m->btree_id,
- bkey_start_pos(&bch2_keylist_front(keys)->k),
- BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
-
- while (1) {
- struct bkey_s_c k;
- struct bkey_i *insert;
- struct bkey_i_extent *new;
- const union bch_extent_entry *entry;
- struct extent_ptr_decoded p;
- struct bpos next_pos;
- bool did_work = false;
- bool should_check_enospc;
- s64 i_sectors_delta = 0, disk_sectors_delta = 0;
-
- bch2_trans_begin(&trans);
-
- k = bch2_btree_iter_peek_slot(&iter);
- ret = bkey_err(k);
- if (ret)
- goto err;
-
- new = bkey_i_to_extent(bch2_keylist_front(keys));
-
- if (bversion_cmp(k.k->version, new->k.version) ||
- !bch2_bkey_matches_ptr(c, k, m->ptr, m->offset))
- goto nomatch;
-
- bkey_reassemble(_insert.k, k);
- insert = _insert.k;
-
- bch2_bkey_buf_copy(&_new, c, bch2_keylist_front(keys));
- new = bkey_i_to_extent(_new.k);
- bch2_cut_front(iter.pos, &new->k_i);
-
- bch2_cut_front(iter.pos, insert);
- bch2_cut_back(new->k.p, insert);
- bch2_cut_back(insert->k.p, &new->k_i);
-
- if (m->data_cmd == DATA_REWRITE) {
- struct bch_extent_ptr *new_ptr, *old_ptr = (void *)
- bch2_bkey_has_device(bkey_i_to_s_c(insert),
- m->data_opts.rewrite_dev);
- if (!old_ptr)
- goto nomatch;
-
- if (old_ptr->cached)
- extent_for_each_ptr(extent_i_to_s(new), new_ptr)
- new_ptr->cached = true;
-
- __bch2_bkey_drop_ptr(bkey_i_to_s(insert), old_ptr);
- }
-
- extent_for_each_ptr_decode(extent_i_to_s(new), p, entry) {
- if (bch2_bkey_has_device(bkey_i_to_s_c(insert), p.ptr.dev)) {
- /*
- * raced with another move op? extent already
- * has a pointer to the device we just wrote
- * data to
- */
- continue;
- }
-
- bch2_extent_ptr_decoded_append(insert, &p);
- did_work = true;
- }
-
- if (!did_work)
- goto nomatch;
-
- bch2_bkey_narrow_crcs(insert,
- (struct bch_extent_crc_unpacked) { 0 });
- bch2_extent_normalize(c, bkey_i_to_s(insert));
- bch2_bkey_mark_replicas_cached(c, bkey_i_to_s(insert),
- op->opts.background_target,
- op->opts.data_replicas);
-
- ret = bch2_sum_sector_overwrites(&trans, &iter, insert,
- &should_check_enospc,
- &i_sectors_delta,
- &disk_sectors_delta);
- if (ret)
- goto err;
-
- if (disk_sectors_delta > (s64) op->res.sectors) {
- ret = bch2_disk_reservation_add(c, &op->res,
- disk_sectors_delta - op->res.sectors,
- !should_check_enospc
- ? BCH_DISK_RESERVATION_NOFAIL : 0);
- if (ret)
- goto out;
- }
-
- next_pos = insert->k.p;
-
- ret = insert_snapshot_whiteouts(&trans, m->btree_id,
- k.k->p, insert->k.p) ?:
- bch2_trans_update(&trans, &iter, insert,
- BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
- bch2_trans_commit(&trans, &op->res,
- op_journal_seq(op),
- BTREE_INSERT_NOFAIL|
- m->data_opts.btree_insert_flags);
- if (!ret) {
- bch2_btree_iter_set_pos(&iter, next_pos);
- atomic_long_inc(&c->extent_migrate_done);
- if (ec_ob)
- bch2_ob_add_backpointer(c, ec_ob, &insert->k);
- }
-err:
- if (ret == -EINTR)
- ret = 0;
- if (ret)
- break;
-next:
- while (bkey_cmp(iter.pos, bch2_keylist_front(keys)->k.p) >= 0) {
- bch2_keylist_pop_front(keys);
- if (bch2_keylist_empty(keys))
- goto out;
- }
- continue;
-nomatch:
- if (m->ctxt) {
- BUG_ON(k.k->p.offset <= iter.pos.offset);
- atomic64_inc(&m->ctxt->stats->keys_raced);
- atomic64_add(k.k->p.offset - iter.pos.offset,
- &m->ctxt->stats->sectors_raced);
- }
- atomic_long_inc(&c->extent_migrate_raced);
- trace_move_race(&new->k);
- bch2_btree_iter_advance(&iter);
- goto next;
- }
-out:
- bch2_trans_iter_exit(&trans, &iter);
- bch2_trans_exit(&trans);
- bch2_bkey_buf_exit(&_insert, c);
- bch2_bkey_buf_exit(&_new, c);
- BUG_ON(ret == -EINTR);
- return ret;
-}
-
-void bch2_migrate_read_done(struct migrate_write *m, struct bch_read_bio *rbio)