bcachefs: Handle dropping pointers in data_update path

Cached pointers are generally dropped, not moved: this led to an
assertion firing in the data update path when there were no new replicas
being written.

This path adds a data_options field for pointers to be dropped, and
tweaks move_extent() to check if we're only dropping pointers, not
writing new ones, before kicking off a data update operation.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

3 files changed, 84 insertions, 6 deletions

diff --git a/fs/bcachefs/data_update.c b/fs/bcachefs/data_update.c
index 6cda97bfeb17..08b33c845af3 100644
--- a/fs/bcachefs/data_update.c
+++ b/fs/bcachefs/data_update.c
@@ -328,8 +328,9 @@ int bch2_data_update_init(struct bch_fs *c, struct data_update *m,
 
 	i = 0;
 	bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
-		if (p.ptr.cached)
-			m->data_opts.rewrite_ptrs &= ~(1U << i);
+		if (((1U << i) & m->data_opts.rewrite_ptrs) &&
+		    p.ptr.cached)
+			BUG();
 
 		if (!((1U << i) & m->data_opts.rewrite_ptrs))
 			bch2_dev_list_add_dev(&m->op.devs_have, p.ptr.dev);
@@ -365,5 +366,23 @@ int bch2_data_update_init(struct bch_fs *c, struct data_update *m,
 
 	m->op.nr_replicas = m->op.nr_replicas_required =
 		hweight32(m->data_opts.rewrite_ptrs) + m->data_opts.extra_replicas;
+
+	BUG_ON(!m->op.nr_replicas);
 	return 0;
 }
+
+void bch2_data_update_opts_normalize(struct bkey_s_c k, struct data_update_opts *opts)
+{
+	struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
+	const struct bch_extent_ptr *ptr;
+	unsigned i = 0;
+
+	bkey_for_each_ptr(ptrs, ptr) {
+		if ((opts->rewrite_ptrs & (1U << i)) && ptr->cached) {
+			opts->kill_ptrs |= 1U << i;
+			opts->rewrite_ptrs ^= 1U << i;
+		}
+
+		i++;
+	}
+}
diff --git a/fs/bcachefs/data_update.h b/fs/bcachefs/data_update.h
index ee38bd655af1..5d8690795959 100644
--- a/fs/bcachefs/data_update.h
+++ b/fs/bcachefs/data_update.h
@@ -10,6 +10,7 @@ struct moving_context;
 
 struct data_update_opts {
 	unsigned	rewrite_ptrs;
+	unsigned	kill_ptrs;
 	u16		target;
 	u8		extra_replicas;
 	unsigned	btree_insert_flags;
@@ -35,5 +36,6 @@ int bch2_data_update_init(struct bch_fs *, struct data_update *,
 			  struct write_point_specifier,
 			  struct bch_io_opts, struct data_update_opts,
 			  enum btree_id, struct bkey_s_c);
+void bch2_data_update_opts_normalize(struct bkey_s_c, struct data_update_opts *);
 
 #endif /* _BCACHEFS_DATA_UPDATE_H */
diff --git a/fs/bcachefs/move.c b/fs/bcachefs/move.c
index 3b01579bd138..8f9c817eed7b 100644
--- a/fs/bcachefs/move.c
+++ b/fs/bcachefs/move.c
@@ -189,7 +189,52 @@ void bch_move_stats_init(struct bch_move_stats *stats, char *name)
 	scnprintf(stats->name, sizeof(stats->name), "%s", name);
 }
 
+static int bch2_extent_drop_ptrs(struct btree_trans *trans,
+				 struct btree_iter *iter,
+				 struct bkey_s_c k,
+				 struct data_update_opts data_opts)
+{
+	struct bch_fs *c = trans->c;
+	struct bkey_i *n;
+	int ret;
+
+	n = bch2_trans_kmalloc(trans, bkey_bytes(k.k));
+	ret = PTR_ERR_OR_ZERO(n);
+	if (ret)
+		return ret;
+
+	bkey_reassemble(n, k);
+
+	while (data_opts.kill_ptrs) {
+		unsigned i = 0, drop = __fls(data_opts.kill_ptrs);
+		struct bch_extent_ptr *ptr;
+
+		bch2_bkey_drop_ptrs(bkey_i_to_s(n), ptr, i++ == drop);
+		data_opts.kill_ptrs ^= 1U << drop;
+	}
+
+	/*
+	 * If the new extent no longer has any pointers, bch2_extent_normalize()
+	 * will do the appropriate thing with it (turning it into a
+	 * KEY_TYPE_error key, or just a discard if it was a cached extent)
+	 */
+	bch2_extent_normalize(c, bkey_i_to_s(n));
+
+	/*
+	 * Since we're not inserting through an extent iterator
+	 * (BTREE_ITER_ALL_SNAPSHOTS iterators aren't extent iterators),
+	 * we aren't using the extent overwrite path to delete, we're
+	 * just using the normal key deletion path:
+	 */
+	if (bkey_deleted(&n->k))
+		n->k.size = 0;
+
+	return bch2_trans_update(trans, iter, n, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
+		bch2_trans_commit(trans, NULL, NULL, BTREE_INSERT_NOFAIL);
+}
+
 static int bch2_move_extent(struct btree_trans *trans,
+			    struct btree_iter *iter,
 			    struct moving_context *ctxt,
 			    struct bch_io_opts io_opts,
 			    enum btree_id btree_id,
@@ -204,6 +249,15 @@ static int bch2_move_extent(struct btree_trans *trans,
 	unsigned sectors = k.k->size, pages;
 	int ret = -ENOMEM;
 
+	bch2_data_update_opts_normalize(k, &data_opts);
+
+	if (!data_opts.rewrite_ptrs &&
+	    !data_opts.extra_replicas) {
+		if (data_opts.kill_ptrs)
+			return bch2_extent_drop_ptrs(trans, iter, k, data_opts);
+		return 0;
+	}
+
 	if (!percpu_ref_tryget_live(&c->writes))
 		return -EROFS;
 
@@ -446,7 +500,7 @@ 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, ctxt, io_opts,
+		ret2 = bch2_move_extent(&trans, &iter, ctxt, io_opts,
 					btree_id, k, data_opts);
 		if (ret2) {
 			if (bch2_err_matches(ret2, BCH_ERR_transaction_restart))
@@ -598,11 +652,12 @@ int __bch2_evacuate_bucket(struct moving_context *ctxt,
 
 			bch2_bkey_buf_reassemble(&sk, c, k);
 			k = bkey_i_to_s_c(sk.k);
-			bch2_trans_iter_exit(&trans, &iter);
 
 			ret = move_get_io_opts(&trans, &io_opts, k, &cur_inum);
-			if (ret)
+			if (ret) {
+				bch2_trans_iter_exit(&trans, &iter);
 				continue;
+			}
 
 			data_opts = _data_opts;
 			data_opts.target	= io_opts.background_target;
@@ -614,8 +669,10 @@ int __bch2_evacuate_bucket(struct moving_context *ctxt,
 				i++;
 			}
 
-			ret = bch2_move_extent(&trans, ctxt, io_opts,
+			ret = bch2_move_extent(&trans, &iter, ctxt, io_opts,
 					       bp.btree_id, k, data_opts);
+			bch2_trans_iter_exit(&trans, &iter);
+
 			if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
 				continue;
 			if (ret == -ENOMEM) {