Update bcachefs sources to 95ff72a6c1 fixup! mm: Centralize & improve oom reporting in show_mem.c

1 files changed, 225 insertions, 587 deletions

diff --git a/libbcachefs/move.c b/libbcachefs/move.c
index 36d20dc8..093efb09 100644
--- a/libbcachefs/move.c
+++ b/libbcachefs/move.c
@@ -1,14 +1,12 @@
 // SPDX-License-Identifier: GPL-2.0
 
 #include "bcachefs.h"
-#include "alloc_background.h"
 #include "alloc_foreground.h"
 #include "backpointers.h"
 #include "bkey_buf.h"
 #include "btree_gc.h"
 #include "btree_update.h"
 #include "btree_update_interior.h"
-#include "buckets.h"
 #include "disk_groups.h"
 #include "ec.h"
 #include "error.h"
@@ -17,7 +15,6 @@
 #include "journal_reclaim.h"
 #include "move.h"
 #include "replicas.h"
-#include "subvolume.h"
 #include "super-io.h"
 #include "keylist.h"
 
@@ -26,7 +23,19 @@
 
 #include <trace/events/bcachefs.h>
 
-#define SECTORS_IN_FLIGHT_PER_DEVICE	2048
+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	list;
@@ -38,414 +47,30 @@ struct moving_io {
 
 	struct bch_read_bio	rbio;
 
-	struct migrate_write	write;
+	struct data_update	write;
 	/* Must be last since it is variable size */
 	struct bio_vec		bi_inline_vecs[0];
 };
 
-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)
-{
-	/* write bio must own pages: */
-	BUG_ON(!m->op.wbio.bio.bi_vcnt);
-
-	m->ptr		= rbio->pick.ptr;
-	m->offset	= rbio->data_pos.offset - rbio->pick.crc.offset;
-	m->op.devs_have	= rbio->devs_have;
-	m->op.pos	= rbio->data_pos;
-	m->op.version	= rbio->version;
-	m->op.crc	= rbio->pick.crc;
-	m->op.wbio.bio.bi_iter.bi_size = m->op.crc.compressed_size << 9;
-
-	if (m->data_cmd == DATA_REWRITE)
-		bch2_dev_list_drop_dev(&m->op.devs_have, m->data_opts.rewrite_dev);
-}
-
-int bch2_migrate_write_init(struct bch_fs *c, struct migrate_write *m,
-			    struct write_point_specifier wp,
-			    struct bch_io_opts io_opts,
-			    enum data_cmd data_cmd,
-			    struct data_opts data_opts,
-			    enum btree_id btree_id,
-			    struct bkey_s_c k)
-{
-	struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
-	const union bch_extent_entry *entry;
-	struct bch_extent_crc_unpacked crc;
-	struct extent_ptr_decoded p;
-	int ret;
-
-	m->btree_id	= btree_id;
-	m->data_cmd	= data_cmd;
-	m->data_opts	= data_opts;
-	m->nr_ptrs_reserved = 0;
-
-	bch2_write_op_init(&m->op, c, io_opts);
-
-	if (!bch2_bkey_is_incompressible(k))
-		m->op.compression_type =
-			bch2_compression_opt_to_type[io_opts.background_compression ?:
-						     io_opts.compression];
-	else
-		m->op.incompressible = true;
-
-	m->op.target	= data_opts.target,
-	m->op.write_point = wp;
-
-	/*
-	 * op->csum_type is normally initialized from the fs/file's current
-	 * options - but if an extent is encrypted, we require that it stays
-	 * encrypted:
-	 */
-	bkey_for_each_crc(k.k, ptrs, crc, entry)
-		if (bch2_csum_type_is_encryption(crc.csum_type)) {
-			m->op.nonce	= crc.nonce + crc.offset;
-			m->op.csum_type = crc.csum_type;
-			break;
-		}
-
-	if (m->data_opts.btree_insert_flags & BTREE_INSERT_USE_RESERVE) {
-		m->op.alloc_reserve = RESERVE_movinggc;
-	} else {
-		/* XXX: this should probably be passed in */
-		m->op.flags |= BCH_WRITE_ONLY_SPECIFIED_DEVS;
-	}
-
-	m->op.flags |= BCH_WRITE_PAGES_STABLE|
-		BCH_WRITE_PAGES_OWNED|
-		BCH_WRITE_DATA_ENCODED|
-		BCH_WRITE_FROM_INTERNAL;
-
-	m->op.nr_replicas	= data_opts.nr_replicas;
-	m->op.nr_replicas_required = data_opts.nr_replicas;
-	m->op.index_update_fn	= bch2_migrate_index_update;
-
-	switch (data_cmd) {
-	case DATA_ADD_REPLICAS: {
-		/*
-		 * DATA_ADD_REPLICAS is used for moving data to a different
-		 * device in the background, and due to compression the new copy
-		 * might take up more space than the old copy:
-		 */
-#if 0
-		int nr = (int) io_opts.data_replicas -
-			bch2_bkey_nr_ptrs_allocated(k);
-#endif
-		int nr = (int) io_opts.data_replicas;
-
-		if (nr > 0) {
-			m->op.nr_replicas = m->nr_ptrs_reserved = nr;
-
-			ret = bch2_disk_reservation_get(c, &m->op.res,
-					k.k->size, m->op.nr_replicas, 0);
-			if (ret)
-				return ret;
-		}
-		break;
-	}
-	case DATA_REWRITE: {
-		unsigned compressed_sectors = 0;
-
-		bkey_for_each_ptr_decode(k.k, ptrs, p, entry)
-			if (p.ptr.dev == data_opts.rewrite_dev) {
-				if (p.ptr.cached)
-					m->op.flags |= BCH_WRITE_CACHED;
-
-				if (!p.ptr.cached &&
-				    crc_is_compressed(p.crc))
-					compressed_sectors += p.crc.compressed_size;
-			}
-
-		if (compressed_sectors) {
-			ret = bch2_disk_reservation_add(c, &m->op.res,
-					k.k->size * m->op.nr_replicas,
-					BCH_DISK_RESERVATION_NOFAIL);
-			if (ret)
-				return ret;
-		}
-		break;
-	}
-	case DATA_PROMOTE:
-		m->op.flags	|= BCH_WRITE_ALLOC_NOWAIT;
-		m->op.flags	|= BCH_WRITE_CACHED;
-		break;
-	default:
-		BUG();
-	}
-
-	return 0;
-}
-
 static void move_free(struct closure *cl)
 {
 	struct moving_io *io = container_of(cl, struct moving_io, cl);
 	struct moving_context *ctxt = io->write.ctxt;
-	struct bvec_iter_all iter;
-	struct bio_vec *bv;
-
-	bch2_disk_reservation_put(io->write.op.c, &io->write.op.res);
-
-	bio_for_each_segment_all(bv, &io->write.op.wbio.bio, iter)
-		if (bv->bv_page)
-			__free_page(bv->bv_page);
+	struct bch_fs *c = ctxt->c;
 
+	bch2_data_update_exit(&io->write);
 	wake_up(&ctxt->wait);
-
+	percpu_ref_put(&c->writes);
 	kfree(io);
 }
 
 static void move_write_done(struct closure *cl)
 {
 	struct moving_io *io = container_of(cl, struct moving_io, cl);
+	struct moving_context *ctxt = io->write.ctxt;
+
+	if (io->write.op.error)
+		ctxt->write_error = true;
 
 	atomic_sub(io->write_sectors, &io->write.ctxt->write_sectors);
 	closure_return_with_destructor(cl, move_free);
@@ -460,10 +85,9 @@ static void move_write(struct closure *cl)
 		return;
 	}
 
-	bch2_migrate_read_done(&io->write, &io->rbio);
-
 	atomic_add(io->write_sectors, &io->write.ctxt->write_sectors);
-	closure_call(&io->write.op.cl, bch2_write, NULL, cl);
+
+	bch2_data_update_read_done(&io->write, io->rbio.pick.crc, cl);
 	continue_at(cl, move_write_done, NULL);
 }
 
@@ -520,14 +144,55 @@ static void bch2_move_ctxt_wait_for_io(struct moving_context *ctxt,
 		atomic_read(&ctxt->write_sectors) != sectors_pending);
 }
 
+void bch2_moving_ctxt_exit(struct moving_context *ctxt)
+{
+	move_ctxt_wait_event(ctxt, NULL, list_empty(&ctxt->reads));
+	closure_sync(&ctxt->cl);
+	progress_list_del(ctxt->c, ctxt->stats);
+
+	EBUG_ON(atomic_read(&ctxt->write_sectors));
+
+	trace_move_data(ctxt->c,
+			atomic64_read(&ctxt->stats->sectors_moved),
+			atomic64_read(&ctxt->stats->keys_moved));
+}
+
+void bch2_moving_ctxt_init(struct moving_context *ctxt,
+			   struct bch_fs *c,
+			   struct bch_ratelimit *rate,
+			   struct bch_move_stats *stats,
+			   struct write_point_specifier wp,
+			   bool wait_on_copygc)
+{
+	memset(ctxt, 0, sizeof(*ctxt));
+
+	ctxt->c		= c;
+	ctxt->rate	= rate;
+	ctxt->stats	= stats;
+	ctxt->wp	= wp;
+	ctxt->wait_on_copygc = wait_on_copygc;
+
+	progress_list_add(c, stats);
+	closure_init_stack(&ctxt->cl);
+	INIT_LIST_HEAD(&ctxt->reads);
+	init_waitqueue_head(&ctxt->wait);
+
+	if (stats)
+		stats->data_type = BCH_DATA_user;
+}
+
+void bch_move_stats_init(struct bch_move_stats *stats, char *name)
+{
+	memset(stats, 0, sizeof(*stats));
+	scnprintf(stats->name, sizeof(stats->name), "%s", name);
+}
+
 static int bch2_move_extent(struct btree_trans *trans,
 			    struct moving_context *ctxt,
-			    struct write_point_specifier wp,
 			    struct bch_io_opts io_opts,
 			    enum btree_id btree_id,
 			    struct bkey_s_c k,
-			    enum data_cmd data_cmd,
-			    struct data_opts data_opts)
+			    struct data_update_opts data_opts)
 {
 	struct bch_fs *c = trans->c;
 	struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
@@ -537,6 +202,9 @@ static int bch2_move_extent(struct btree_trans *trans,
 	unsigned sectors = k.k->size, pages;
 	int ret = -ENOMEM;
 
+	if (!percpu_ref_tryget_live(&c->writes))
+		return -EROFS;
+
 	/* write path might have to decompress data: */
 	bkey_for_each_ptr_decode(k.k, ptrs, p, entry)
 		sectors = max_t(unsigned, sectors, p.crc.uncompressed_size);
@@ -570,11 +238,13 @@ static int bch2_move_extent(struct btree_trans *trans,
 	io->rbio.bio.bi_iter.bi_sector	= bkey_start_offset(k.k);
 	io->rbio.bio.bi_end_io		= move_read_endio;
 
-	ret = bch2_migrate_write_init(c, &io->write, wp, io_opts,
-				      data_cmd, data_opts, btree_id, k);
+	ret = bch2_data_update_init(c, &io->write, ctxt->wp, io_opts,
+				    data_opts, btree_id, k);
 	if (ret)
 		goto err_free_pages;
 
+	io->write.ctxt = ctxt;
+
 	atomic64_inc(&ctxt->stats->keys_moved);
 	atomic64_add(k.k->size, &ctxt->stats->sectors_moved);
 	this_cpu_add(c->counters[BCH_COUNTER_io_move], k.k->size);
@@ -600,6 +270,7 @@ err_free_pages:
 err_free:
 	kfree(io);
 err:
+	percpu_ref_put(&c->writes);
 	trace_move_alloc_mem_fail(k.k);
 	return ret;
 }
@@ -636,13 +307,20 @@ err:
 }
 
 static int move_ratelimit(struct btree_trans *trans,
-			  struct moving_context *ctxt,
-			  struct bch_ratelimit *rate)
+			  struct moving_context *ctxt)
 {
+	struct bch_fs *c = trans->c;
 	u64 delay;
 
+	if (ctxt->wait_on_copygc) {
+		bch2_trans_unlock(trans);
+		wait_event_killable(c->copygc_running_wq,
+				    !c->copygc_running ||
+				    kthread_should_stop());
+	}
+
 	do {
-		delay = rate ? bch2_ratelimit_delay(rate) : 0;
+		delay = ctxt->rate ? bch2_ratelimit_delay(ctxt->rate) : 0;
 
 		if (delay) {
 			bch2_trans_unlock(trans);
@@ -665,11 +343,11 @@ static int move_ratelimit(struct btree_trans *trans,
 
 	move_ctxt_wait_event(ctxt, trans,
 		atomic_read(&ctxt->write_sectors) <
-		SECTORS_IN_FLIGHT_PER_DEVICE);
+		c->opts.move_bytes_in_flight >> 9);
 
 	move_ctxt_wait_event(ctxt, trans,
 		atomic_read(&ctxt->read_sectors) <
-		SECTORS_IN_FLIGHT_PER_DEVICE);
+		c->opts.move_bytes_in_flight >> 9);
 
 	return 0;
 }
@@ -699,41 +377,37 @@ static int move_get_io_opts(struct btree_trans *trans,
 	return 0;
 }
 
-static int __bch2_move_data(struct bch_fs *c,
-		struct moving_context *ctxt,
-		struct bch_ratelimit *rate,
-		struct write_point_specifier wp,
-		struct bpos start,
-		struct bpos end,
-		move_pred_fn pred, void *arg,
-		struct bch_move_stats *stats,
-		enum btree_id btree_id)
+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)
 {
+	struct bch_fs *c = ctxt->c;
 	struct bch_io_opts io_opts = bch2_opts_to_inode_opts(c->opts);
 	struct bkey_buf sk;
 	struct btree_trans trans;
 	struct btree_iter iter;
 	struct bkey_s_c k;
-	struct data_opts data_opts;
-	enum data_cmd data_cmd;
+	struct data_update_opts data_opts;
 	u64 cur_inum = U64_MAX;
 	int ret = 0, ret2;
 
 	bch2_bkey_buf_init(&sk);
 	bch2_trans_init(&trans, c, 0, 0);
 
-	stats->data_type = BCH_DATA_user;
-	stats->btree_id	= btree_id;
-	stats->pos	= start;
+	ctxt->stats->data_type	= BCH_DATA_user;
+	ctxt->stats->btree_id	= btree_id;
+	ctxt->stats->pos	= start;
 
 	bch2_trans_iter_init(&trans, &iter, btree_id, start,
 			     BTREE_ITER_PREFETCH|
 			     BTREE_ITER_ALL_SNAPSHOTS);
 
-	if (rate)
-		bch2_ratelimit_reset(rate);
+	if (ctxt->rate)
+		bch2_ratelimit_reset(ctxt->rate);
 
-	while (!move_ratelimit(&trans, ctxt, rate)) {
+	while (!move_ratelimit(&trans, ctxt)) {
 		bch2_trans_begin(&trans);
 
 		k = bch2_btree_iter_peek(&iter);
@@ -749,7 +423,7 @@ static int __bch2_move_data(struct bch_fs *c,
 		if (bkey_cmp(bkey_start_pos(k.k), end) >= 0)
 			break;
 
-		stats->pos = iter.pos;
+		ctxt->stats->pos = iter.pos;
 
 		if (!bkey_extent_is_direct_data(k.k))
 			goto next_nondata;
@@ -758,18 +432,9 @@ static int __bch2_move_data(struct bch_fs *c,
 		if (ret)
 			continue;
 
-		switch ((data_cmd = pred(c, arg, k, &io_opts, &data_opts))) {
-		case DATA_SKIP:
+		memset(&data_opts, 0, sizeof(data_opts));
+		if (!pred(c, arg, k, &io_opts, &data_opts))
 			goto next;
-		case DATA_SCRUB:
-			BUG();
-		case DATA_ADD_REPLICAS:
-		case DATA_REWRITE:
-		case DATA_PROMOTE:
-			break;
-		default:
-			BUG();
-		}
 
 		/*
 		 * The iterator gets unlocked by __bch2_read_extent - need to
@@ -778,8 +443,8 @@ static int __bch2_move_data(struct bch_fs *c,
 		bch2_bkey_buf_reassemble(&sk, c, k);
 		k = bkey_i_to_s_c(sk.k);
 
-		ret2 = bch2_move_extent(&trans, ctxt, wp, io_opts, btree_id, k,
-					data_cmd, data_opts);
+		ret2 = bch2_move_extent(&trans, ctxt, io_opts,
+					btree_id, k, data_opts);
 		if (ret2) {
 			if (ret2 == -EINTR)
 				continue;
@@ -794,10 +459,10 @@ static int __bch2_move_data(struct bch_fs *c,
 			goto next;
 		}
 
-		if (rate)
-			bch2_ratelimit_increment(rate, k.k->size);
+		if (ctxt->rate)
+			bch2_ratelimit_increment(ctxt->rate, k.k->size);
 next:
-		atomic64_add(k.k->size, &stats->sectors_seen);
+		atomic64_add(k.k->size, &ctxt->stats->sectors_seen);
 next_nondata:
 		bch2_btree_iter_advance(&iter);
 	}
@@ -809,48 +474,20 @@ next_nondata:
 	return ret;
 }
 
-inline void bch_move_stats_init(struct bch_move_stats *stats, char *name)
-{
-	memset(stats, 0, sizeof(*stats));
-
-	scnprintf(stats->name, sizeof(stats->name),
-			"%s", name);
-}
-
-static inline 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 inline 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);
-}
-
 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,
-		   move_pred_fn pred, void *arg,
-		   struct bch_move_stats *stats)
+		   bool wait_on_copygc,
+		   move_pred_fn pred, void *arg)
 {
-	struct moving_context ctxt = { .stats = stats };
+	struct moving_context ctxt;
 	enum btree_id id;
 	int ret;
 
-	progress_list_add(c, stats);
-	closure_init_stack(&ctxt.cl);
-	INIT_LIST_HEAD(&ctxt.reads);
-	init_waitqueue_head(&ctxt.wait);
-
-	stats->data_type = BCH_DATA_user;
+	bch2_moving_ctxt_init(&ctxt, c, rate, stats, wp, wait_on_copygc);
 
 	for (id = start_btree_id;
 	     id <= min_t(unsigned, end_btree_id, BTREE_ID_NR - 1);
@@ -861,24 +498,16 @@ int bch2_move_data(struct bch_fs *c,
 		    id != BTREE_ID_reflink)
 			continue;
 
-		ret = __bch2_move_data(c, &ctxt, rate, wp,
+		ret = __bch2_move_data(&ctxt,
 				       id == start_btree_id ? start_pos : POS_MIN,
 				       id == end_btree_id   ? end_pos   : POS_MAX,
-				       pred, arg, stats, id);
+				       pred, arg, id);
 		if (ret)
 			break;
 	}
 
-	move_ctxt_wait_event(&ctxt, NULL, list_empty(&ctxt.reads));
-	closure_sync(&ctxt.cl);
-
-	EBUG_ON(atomic_read(&ctxt.write_sectors));
+	bch2_moving_ctxt_exit(&ctxt);
 
-	trace_move_data(c,
-			atomic64_read(&stats->sectors_moved),
-			atomic64_read(&stats->keys_moved));
-
-	progress_list_del(c, stats);
 	return ret;
 }
 
@@ -891,6 +520,7 @@ static int verify_bucket_evacuated(struct btree_trans *trans, struct bpos bucket
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc,
 			     bucket, BTREE_ITER_CACHED);
+again:
 	k = bch2_btree_iter_peek_slot(&iter);
 	ret = bkey_err(k);
 
@@ -901,10 +531,16 @@ static int verify_bucket_evacuated(struct btree_trans *trans, struct bpos bucket
 		    a.v->dirty_sectors) {
 			struct printbuf buf = PRINTBUF;
 
+			if (a.v->data_type == BCH_DATA_btree) {
+				bch2_trans_unlock(trans);
+				if (bch2_btree_interior_updates_flush(c))
+					goto again;
+			}
+
 			prt_str(&buf, "failed to evacuate bucket ");
 			bch2_bkey_val_to_text(&buf, c, k);
 
-			bch_err_ratelimited(c, "%s", buf.buf);
+			bch2_trans_inconsistent(trans, "%s", buf.buf);
 			printbuf_exit(&buf);
 		}
 	}
@@ -913,33 +549,24 @@ static int verify_bucket_evacuated(struct btree_trans *trans, struct bpos bucket
 	return ret;
 }
 
-int bch2_evacuate_bucket(struct bch_fs *c,
-			 struct bpos bucket, int gen,
-			 struct bch_ratelimit *rate,
-			 struct write_point_specifier wp,
-			 enum data_cmd data_cmd,
-			 struct data_opts *data_opts,
-			 struct bch_move_stats *stats)
+int __bch2_evacuate_bucket(struct moving_context *ctxt,
+			   struct bpos bucket, int gen,
+			   struct data_update_opts _data_opts)
 {
+	struct bch_fs *c = ctxt->c;
 	struct bch_io_opts io_opts = bch2_opts_to_inode_opts(c->opts);
-	struct moving_context ctxt = { .stats = stats };
 	struct btree_trans trans;
 	struct btree_iter iter;
 	struct bkey_buf sk;
 	struct bch_backpointer bp;
+	struct data_update_opts data_opts;
 	u64 bp_offset = 0, cur_inum = U64_MAX;
 	int ret = 0;
 
 	bch2_bkey_buf_init(&sk);
 	bch2_trans_init(&trans, c, 0, 0);
-	progress_list_add(c, stats);
-	closure_init_stack(&ctxt.cl);
-	INIT_LIST_HEAD(&ctxt.reads);
-	init_waitqueue_head(&ctxt.wait);
 
-	stats->data_type = BCH_DATA_user;
-
-	while (!(ret = move_ratelimit(&trans, &ctxt, rate))) {
+	while (!(ret = move_ratelimit(&trans, ctxt))) {
 		bch2_trans_begin(&trans);
 
 		ret = bch2_get_next_backpointer(&trans, bucket, gen,
@@ -952,7 +579,9 @@ int bch2_evacuate_bucket(struct bch_fs *c,
 			break;
 
 		if (!bp.level) {
+			const struct bch_extent_ptr *ptr;
 			struct bkey_s_c k;
+			unsigned i = 0;
 
 			k = bch2_backpointer_get_key(&trans, &iter,
 						bucket, bp_offset, bp);
@@ -972,24 +601,31 @@ int bch2_evacuate_bucket(struct bch_fs *c,
 			if (ret)
 				continue;
 
-			data_opts->target	= io_opts.background_target;
-			data_opts->rewrite_dev	= bucket.inode;
+			data_opts = _data_opts;
+			data_opts.target	= io_opts.background_target;
+			data_opts.rewrite_ptrs = 0;
+
+			bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr) {
+				if (ptr->dev == bucket.inode)
+					data_opts.rewrite_ptrs |= 1U << i;
+				i++;
+			}
 
-			ret = bch2_move_extent(&trans, &ctxt, wp, io_opts, bp.btree_id, k,
-						data_cmd, *data_opts);
+			ret = bch2_move_extent(&trans, ctxt, io_opts,
+					       bp.btree_id, k, data_opts);
 			if (ret == -EINTR)
 				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, &trans);
 				continue;
 			}
 			if (ret)
 				goto err;
 
-			if (rate)
-				bch2_ratelimit_increment(rate, k.k->size);
-			atomic64_add(k.k->size, &stats->sectors_seen);
+			if (ctxt->rate)
+				bch2_ratelimit_increment(ctxt->rate, k.k->size);
+			atomic64_add(k.k->size, &ctxt->stats->sectors_seen);
 		} else {
 			struct btree *b;
 
@@ -1011,10 +647,11 @@ int bch2_evacuate_bucket(struct bch_fs *c,
 			if (ret)
 				goto err;
 
-			if (rate)
-				bch2_ratelimit_increment(rate, c->opts.btree_node_size >> 9);
-			atomic64_add(c->opts.btree_node_size >> 9, &stats->sectors_seen);
-			atomic64_add(c->opts.btree_node_size >> 9, &stats->sectors_moved);
+			if (ctxt->rate)
+				bch2_ratelimit_increment(ctxt->rate,
+							 c->opts.btree_node_size >> 9);
+			atomic64_add(c->opts.btree_node_size >> 9, &ctxt->stats->sectors_seen);
+			atomic64_add(c->opts.btree_node_size >> 9, &ctxt->stats->sectors_moved);
 		}
 
 		bp_offset++;
@@ -1022,30 +659,38 @@ int bch2_evacuate_bucket(struct bch_fs *c,
 
 	if (IS_ENABLED(CONFIG_BCACHEFS_DEBUG) && gen >= 0) {
 		bch2_trans_unlock(&trans);
-		move_ctxt_wait_event(&ctxt, NULL, list_empty(&ctxt.reads));
-		closure_sync(&ctxt.cl);
-		lockrestart_do(&trans, verify_bucket_evacuated(&trans, bucket, gen));
+		move_ctxt_wait_event(ctxt, NULL, list_empty(&ctxt->reads));
+		closure_sync(&ctxt->cl);
+		if (!ctxt->write_error)
+			lockrestart_do(&trans, verify_bucket_evacuated(&trans, bucket, gen));
 	}
 err:
 	bch2_trans_exit(&trans);
 	bch2_bkey_buf_exit(&sk, c);
+	return ret;
+}
 
-	move_ctxt_wait_event(&ctxt, NULL, list_empty(&ctxt.reads));
-	closure_sync(&ctxt.cl);
-	progress_list_del(c, stats);
-
-	EBUG_ON(atomic_read(&ctxt.write_sectors));
+int bch2_evacuate_bucket(struct bch_fs *c,
+			 struct bpos bucket, int gen,
+			 struct data_update_opts data_opts,
+			 struct bch_ratelimit *rate,
+			 struct bch_move_stats *stats,
+			 struct write_point_specifier wp,
+			 bool wait_on_copygc)
+{
+	struct moving_context ctxt;
+	int ret;
 
-	trace_move_data(c,
-			atomic64_read(&stats->sectors_moved),
-			atomic64_read(&stats->keys_moved));
+	bch2_moving_ctxt_init(&ctxt, c, rate, stats, wp, wait_on_copygc);
+	ret = __bch2_evacuate_bucket(&ctxt, bucket, gen, data_opts);
+	bch2_moving_ctxt_exit(&ctxt);
 
 	return ret;
 }
 
-typedef enum data_cmd (*move_btree_pred)(struct bch_fs *, void *,
-					 struct btree *, struct bch_io_opts *,
-					 struct data_opts *);
+typedef bool (*move_btree_pred)(struct bch_fs *, void *,
+				struct btree *, struct bch_io_opts *,
+				struct data_update_opts *);
 
 static int bch2_move_btree(struct bch_fs *c,
 			   enum btree_id start_btree_id, struct bpos start_pos,
@@ -1059,8 +704,7 @@ static int bch2_move_btree(struct bch_fs *c,
 	struct btree_iter iter;
 	struct btree *b;
 	enum btree_id id;
-	struct data_opts data_opts;
-	enum data_cmd cmd;
+	struct data_update_opts data_opts;
 	int ret = 0;
 
 	bch2_trans_init(&trans, c, 0, 0);
@@ -1089,17 +733,8 @@ retry:
 
 			stats->pos = iter.pos;
 
-			switch ((cmd = pred(c, arg, b, &io_opts, &data_opts))) {
-			case DATA_SKIP:
+			if (!pred(c, arg, b, &io_opts, &data_opts))
 				goto next;
-			case DATA_SCRUB:
-				BUG();
-			case DATA_ADD_REPLICAS:
-			case DATA_REWRITE:
-				break;
-			default:
-				BUG();
-			}
 
 			ret = bch2_btree_node_rewrite(&trans, &iter, b, 0) ?: ret;
 			if (ret == -EINTR)
@@ -1129,20 +764,10 @@ next:
 	return ret;
 }
 
-#if 0
-static enum data_cmd scrub_pred(struct bch_fs *c, void *arg,
-				struct bkey_s_c k,
-				struct bch_io_opts *io_opts,
-				struct data_opts *data_opts)
-{
-	return DATA_SCRUB;
-}
-#endif
-
-static enum data_cmd rereplicate_pred(struct bch_fs *c, void *arg,
-				      struct bkey_s_c k,
-				      struct bch_io_opts *io_opts,
-				      struct data_opts *data_opts)
+static bool rereplicate_pred(struct bch_fs *c, void *arg,
+			     struct bkey_s_c k,
+			     struct bch_io_opts *io_opts,
+			     struct data_update_opts *data_opts)
 {
 	unsigned nr_good = bch2_bkey_durability(c, k);
 	unsigned replicas = bkey_is_btree_ptr(k.k)
@@ -1150,43 +775,50 @@ static enum data_cmd rereplicate_pred(struct bch_fs *c, void *arg,
 		: io_opts->data_replicas;
 
 	if (!nr_good || nr_good >= replicas)
-		return DATA_SKIP;
+		return false;
 
 	data_opts->target		= 0;
-	data_opts->nr_replicas		= 1;
+	data_opts->extra_replicas	= replicas - nr_good;
 	data_opts->btree_insert_flags	= 0;
-	return DATA_ADD_REPLICAS;
+	return true;
 }
 
-static enum data_cmd migrate_pred(struct bch_fs *c, void *arg,
-				  struct bkey_s_c k,
-				  struct bch_io_opts *io_opts,
-				  struct data_opts *data_opts)
+static bool migrate_pred(struct bch_fs *c, void *arg,
+			 struct bkey_s_c k,
+			 struct bch_io_opts *io_opts,
+			 struct data_update_opts *data_opts)
 {
+	struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
+	const struct bch_extent_ptr *ptr;
 	struct bch_ioctl_data *op = arg;
+	unsigned i = 0;
 
-	if (!bch2_bkey_has_device(k, op->migrate.dev))
-		return DATA_SKIP;
-
+	data_opts->rewrite_ptrs		= 0;
 	data_opts->target		= 0;
-	data_opts->nr_replicas		= 1;
+	data_opts->extra_replicas	= 0;
 	data_opts->btree_insert_flags	= 0;
-	data_opts->rewrite_dev		= op->migrate.dev;
-	return DATA_REWRITE;
+
+	bkey_for_each_ptr(ptrs, ptr) {
+		if (ptr->dev == op->migrate.dev)
+			data_opts->rewrite_ptrs |= 1U << i;
+		i++;
+	}
+
+	return data_opts->rewrite_ptrs != 0;;
 }
 
-static enum data_cmd rereplicate_btree_pred(struct bch_fs *c, void *arg,
-					    struct btree *b,
-					    struct bch_io_opts *io_opts,
-					    struct data_opts *data_opts)
+static bool rereplicate_btree_pred(struct bch_fs *c, void *arg,
+				   struct btree *b,
+				   struct bch_io_opts *io_opts,
+				   struct data_update_opts *data_opts)
 {
 	return rereplicate_pred(c, arg, bkey_i_to_s_c(&b->key), io_opts, data_opts);
 }
 
-static enum data_cmd migrate_btree_pred(struct bch_fs *c, void *arg,
-					struct btree *b,
-					struct bch_io_opts *io_opts,
-					struct data_opts *data_opts)
+static bool migrate_btree_pred(struct bch_fs *c, void *arg,
+			       struct btree *b,
+			       struct bch_io_opts *io_opts,
+			       struct data_update_opts *data_opts)
 {
 	return migrate_pred(c, arg, bkey_i_to_s_c(&b->key), io_opts, data_opts);
 }
@@ -1215,21 +847,21 @@ static bool bformat_needs_redo(struct bkey_format *f)
 	return false;
 }
 
-static enum data_cmd rewrite_old_nodes_pred(struct bch_fs *c, void *arg,
-					    struct btree *b,
-					    struct bch_io_opts *io_opts,
-					    struct data_opts *data_opts)
+static bool rewrite_old_nodes_pred(struct bch_fs *c, void *arg,
+				   struct btree *b,
+				   struct bch_io_opts *io_opts,
+				   struct data_update_opts *data_opts)
 {
 	if (b->version_ondisk != c->sb.version ||
 	    btree_node_need_rewrite(b) ||
 	    bformat_needs_redo(&b->format)) {
 		data_opts->target		= 0;
-		data_opts->nr_replicas		= 1;
+		data_opts->extra_replicas	= 0;
 		data_opts->btree_insert_flags	= 0;
-		return DATA_REWRITE;
+		return true;
 	}
 
-	return DATA_SKIP;
+	return false;
 }
 
 int bch2_scan_old_btree_nodes(struct bch_fs *c, struct bch_move_stats *stats)
@@ -1273,8 +905,11 @@ int bch2_data_job(struct bch_fs *c,
 		ret = bch2_move_data(c,
 				     op.start_btree,	op.start_pos,
 				     op.end_btree,	op.end_pos,
-				     NULL, writepoint_hashed((unsigned long) current),
-				     rereplicate_pred, c, stats) ?: ret;
+				     NULL,
+				     stats,
+				     writepoint_hashed((unsigned long) current),
+				     true,
+				     rereplicate_pred, c) ?: ret;
 		ret = bch2_replicas_gc2(c) ?: ret;
 		break;
 	case BCH_DATA_OP_MIGRATE:
@@ -1294,8 +929,11 @@ int bch2_data_job(struct bch_fs *c,
 		ret = bch2_move_data(c,
 				     op.start_btree,	op.start_pos,
 				     op.end_btree,	op.end_pos,
-				     NULL, writepoint_hashed((unsigned long) current),
-				     migrate_pred, &op, stats) ?: ret;
+				     NULL,
+				     stats,
+				     writepoint_hashed((unsigned long) current),
+				     true,
+				     migrate_pred, &op) ?: ret;
 		ret = bch2_replicas_gc2(c) ?: ret;
 		break;
 	case BCH_DATA_OP_REWRITE_OLD_NODES: