bcachefs: bch2_dev_remove_stripes()

We can now correctly force-remove a device that has stripes on it; this
uses the new BCH_SB_MEMBER_INVALID sentinal value.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

fs/bcachefs/alloc_background.c

@@ -2310,7 +2310,8 @@ int bch2_dev_remove_alloc(struct bch_fs *c, struct bch_dev *ca)
 	 * We clear the LRU and need_discard btrees first so that we don't race
 	 * with bch2_do_invalidates() and bch2_do_discards()
 	 */
-	ret =   bch2_btree_delete_range(c, BTREE_ID_lru, start, end,
+	ret =   bch2_dev_remove_stripes(c, ca->dev_idx) ?:
+		bch2_btree_delete_range(c, BTREE_ID_lru, start, end,
 					BTREE_TRIGGER_norun, NULL) ?:
 		bch2_btree_delete_range(c, BTREE_ID_need_discard, start, end,
 					BTREE_TRIGGER_norun, NULL) ?:
@@ -2318,10 +2319,10 @@ int bch2_dev_remove_alloc(struct bch_fs *c, struct bch_dev *ca)
 					BTREE_TRIGGER_norun, NULL) ?:
 		bch2_btree_delete_range(c, BTREE_ID_backpointers, start, end,
 					BTREE_TRIGGER_norun, NULL) ?:
-		bch2_btree_delete_range(c, BTREE_ID_alloc, start, end,
-					BTREE_TRIGGER_norun, NULL) ?:
 		bch2_btree_delete_range(c, BTREE_ID_bucket_gens, start, end,
 					BTREE_TRIGGER_norun, NULL) ?:
+		bch2_btree_delete_range(c, BTREE_ID_alloc, start, end,
+					BTREE_TRIGGER_norun, NULL) ?:
 		bch2_dev_usage_remove(c, ca->dev_idx);
 	bch_err_msg(ca, ret, "removing dev alloc info");
 	return ret;

fs/bcachefs/ec.c

@@ -2169,6 +2169,73 @@ struct ec_stripe_head *bch2_ec_stripe_head_get(struct btree_trans *trans,
 	return ERR_PTR(ret);
 }
 
+/* device removal */
+
+static int bch2_invalidate_stripe_to_dev(struct btree_trans *trans, struct bkey_s_c k_a)
+{
+	struct bch_alloc_v4 a_convert;
+	const struct bch_alloc_v4 *a = bch2_alloc_to_v4(k_a, &a_convert);
+
+	if (!a->stripe)
+		return 0;
+
+	if (a->stripe_sectors) {
+		bch_err(trans->c, "trying to invalidate device in stripe when bucket has stripe data");
+		return -BCH_ERR_invalidate_stripe_to_dev;
+	}
+
+	struct btree_iter iter;
+	struct bkey_i_stripe *s =
+		bch2_bkey_get_mut_typed(trans, &iter, BTREE_ID_stripes, POS(0, a->stripe),
+					BTREE_ITER_slots, stripe);
+	int ret = PTR_ERR_OR_ZERO(s);
+	if (ret)
+		return ret;
+
+	struct disk_accounting_pos acc = {
+		.type = BCH_DISK_ACCOUNTING_replicas,
+	};
+
+	s64 sectors = 0;
+	for (unsigned i = 0; i < s->v.nr_blocks; i++)
+		sectors -= stripe_blockcount_get(&s->v, i);
+
+	bch2_bkey_to_replicas(&acc.replicas, bkey_i_to_s_c(&s->k_i));
+	acc.replicas.data_type = BCH_DATA_user;
+	ret = bch2_disk_accounting_mod(trans, &acc, &sectors, 1, false);
+	if (ret)
+		goto err;
+
+	struct bkey_ptrs ptrs = bch2_bkey_ptrs(bkey_i_to_s(&s->k_i));
+	bkey_for_each_ptr(ptrs, ptr)
+		if (ptr->dev == k_a.k->p.inode)
+			ptr->dev = BCH_SB_MEMBER_INVALID;
+
+	sectors = -sectors;
+
+	bch2_bkey_to_replicas(&acc.replicas, bkey_i_to_s_c(&s->k_i));
+	acc.replicas.data_type = BCH_DATA_user;
+	ret = bch2_disk_accounting_mod(trans, &acc, &sectors, 1, false);
+	if (ret)
+		goto err;
+err:
+	bch2_trans_iter_exit(trans, &iter);
+	return ret;
+}
+
+int bch2_dev_remove_stripes(struct bch_fs *c, unsigned dev_idx)
+{
+	return bch2_trans_run(c,
+		for_each_btree_key_upto_commit(trans, iter,
+				  BTREE_ID_alloc, POS(dev_idx, 0), POS(dev_idx, U64_MAX),
+				  BTREE_ITER_intent, k,
+				  NULL, NULL, 0, ({
+			bch2_invalidate_stripe_to_dev(trans, k);
+	})));
+}
+
+/* startup/shutdown */
+
 static void __bch2_ec_stop(struct bch_fs *c, struct bch_dev *ca)
 {
 	struct ec_stripe_head *h;

fs/bcachefs/ec.h

@@ -251,6 +251,8 @@ static inline void ec_stripe_new_put(struct bch_fs *c, struct ec_stripe_new *s,
 		}
 }
 
+int bch2_dev_remove_stripes(struct bch_fs *, unsigned);
+
 void bch2_ec_stop_dev(struct bch_fs *, struct bch_dev *);
 void bch2_fs_ec_stop(struct bch_fs *);
 void bch2_fs_ec_flush(struct bch_fs *);

fs/bcachefs/errcode.h

@@ -253,6 +253,7 @@
 	x(EIO,				key_type_error)				\
 	x(EIO,				no_device_to_read_from)			\
 	x(EIO,				missing_indirect_extent)		\
+	x(EIO,				invalidate_stripe_to_dev)		\
 	x(BCH_ERR_btree_node_read_err,	btree_node_read_err_fixable)		\
 	x(BCH_ERR_btree_node_read_err,	btree_node_read_err_want_retry)		\
 	x(BCH_ERR_btree_node_read_err,	btree_node_read_err_must_retry)		\