Update bcachefs sources to 4a7a003763f5 bcachefs: Opt_durability can …

…now be set via bch2_opt_set_sb()

.bcachefs_revision

@@ -1 +1 @@
-4d1d53862afb768e4edcf7256f8fe0634c96b40e
+4a7a003763f594fcb798c13b4644521083f885b3

c_src/libbcachefs.c

@@ -227,7 +227,7 @@ struct bch_sb *bch2_format(struct bch_opt_strs	fs_opt_strs,
 			? bch2_opt_get_by_id(&fs_opts, opt_id)
 			: bch2_opt_get_by_id(&bch2_opts_default, opt_id);
 
-		__bch2_opt_set_sb(sb.sb, &bch2_opt_table[opt_id], v);
+		__bch2_opt_set_sb(sb.sb, -1, &bch2_opt_table[opt_id], v);
 	}
 
 	struct timespec now;

c_src/posix_to_bcachefs.c

@@ -166,7 +166,7 @@ static void write_data(struct bch_fs *c,
 
 	closure_call(&op.cl, bch2_write, NULL, NULL);
 
-	BUG_ON(!(op.flags & BCH_WRITE_DONE));
+	BUG_ON(!(op.flags & BCH_WRITE_SUBMITTED));
 	dst_inode->bi_sectors += len >> 9;
 
 	if (op.error)

libbcachefs/alloc_foreground.c

@@ -662,8 +662,14 @@ static struct open_bucket *bch2_bucket_alloc_trans(struct btree_trans *trans,
 		goto alloc;
 	}
 err:
-	if (!ob)
+	if (!ob) {
+		rcu_read_lock();
+		struct task_struct *t = rcu_dereference(c->copygc_thread);
+		if (t)
+			wake_up_process(t);
+		rcu_read_unlock();
 		ob = ERR_PTR(-BCH_ERR_no_buckets_found);
+	}
 
 	if (!IS_ERR(ob))
 		ob->data_type = data_type;

libbcachefs/bcachefs.h

@@ -981,7 +981,7 @@ struct bch_fs {
 	struct bch_fs_rebalance	rebalance;
 
 	/* COPYGC */
-	struct task_struct	*copygc_thread;
+	struct task_struct __rcu *copygc_thread;
 	struct write_point	copygc_write_point;
 	s64			copygc_wait_at;
 	s64			copygc_wait;

libbcachefs/btree_io.c

@@ -46,8 +46,6 @@ void bch2_btree_node_io_unlock(struct btree *b)
 
 void bch2_btree_node_io_lock(struct btree *b)
 {
-	bch2_assert_btree_nodes_not_locked();
-
 	wait_on_bit_lock_io(&b->flags, BTREE_NODE_write_in_flight,
 			    TASK_UNINTERRUPTIBLE);
 }
@@ -66,16 +64,12 @@ void __bch2_btree_node_wait_on_write(struct btree *b)
 
 void bch2_btree_node_wait_on_read(struct btree *b)
 {
-	bch2_assert_btree_nodes_not_locked();
-
 	wait_on_bit_io(&b->flags, BTREE_NODE_read_in_flight,
 		       TASK_UNINTERRUPTIBLE);
 }
 
 void bch2_btree_node_wait_on_write(struct btree *b)
 {
-	bch2_assert_btree_nodes_not_locked();
-
 	wait_on_bit_io(&b->flags, BTREE_NODE_write_in_flight,
 		       TASK_UNINTERRUPTIBLE);
 }

libbcachefs/btree_iter.c

@@ -3282,7 +3282,8 @@ bool bch2_current_has_btree_trans(struct bch_fs *c)
 	struct btree_trans *trans;
 	bool ret = false;
 	list_for_each_entry(trans, &c->btree_trans_list, list)
-		if (trans->locking_wait.task == current) {
+		if (trans->locking_wait.task == current &&
+		    trans->locked) {
 			ret = true;
 			break;
 		}

libbcachefs/btree_locking.c

@@ -13,16 +13,6 @@ void bch2_btree_lock_init(struct btree_bkey_cached_common *b,
 	lockdep_set_notrack_class(&b->lock);
 }
 
-#ifdef CONFIG_LOCKDEP
-void bch2_assert_btree_nodes_not_locked(void)
-{
-#if 0
-	//Re-enable when lock_class_is_held() is merged:
-	BUG_ON(lock_class_is_held(&bch2_btree_node_lock_key));
-#endif
-}
-#endif
-
 /* Btree node locking: */
 
 struct six_lock_count bch2_btree_node_lock_counts(struct btree_trans *trans,

libbcachefs/btree_locking.h

@@ -15,12 +15,6 @@
 
 void bch2_btree_lock_init(struct btree_bkey_cached_common *, enum six_lock_init_flags);
 
-#ifdef CONFIG_LOCKDEP
-void bch2_assert_btree_nodes_not_locked(void);
-#else
-static inline void bch2_assert_btree_nodes_not_locked(void) {}
-#endif
-
 void bch2_trans_unlock_noassert(struct btree_trans *);
 
 static inline bool is_btree_node(struct btree_path *path, unsigned l)

libbcachefs/darray.c

@@ -13,7 +13,8 @@ int __bch2_darray_resize(darray_char *d, size_t element_size, size_t new_size, g
 		if (!data)
 			return -ENOMEM;
 
-		memcpy(data, d->data, d->size * element_size);
+		if (d->size)
+			memcpy(data, d->data, d->size * element_size);
 		if (d->data != d->preallocated)
 			kvfree(d->data);
 		d->data	= data;

libbcachefs/io_read.c

@@ -1147,42 +1147,35 @@ void __bch2_read(struct bch_fs *c, struct bch_read_bio *rbio,
 	struct btree_iter iter;
 	struct bkey_buf sk;
 	struct bkey_s_c k;
-	u32 snapshot;
 	int ret;
 
 	BUG_ON(flags & BCH_READ_NODECODE);
 
 	bch2_bkey_buf_init(&sk);
-retry:
-	bch2_trans_begin(trans);
-	iter = (struct btree_iter) { NULL };
-
-	ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot);
-	if (ret)
-		goto err;
-
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
-			     SPOS(inum.inum, bvec_iter.bi_sector, snapshot),
+			     POS(inum.inum, bvec_iter.bi_sector),
 			     BTREE_ITER_slots);
+
 	while (1) {
 		unsigned bytes, sectors, offset_into_extent;
 		enum btree_id data_btree = BTREE_ID_extents;
 
-		/*
-		 * read_extent -> io_time_reset may cause a transaction restart
-		 * without returning an error, we need to check for that here:
-		 */
-		ret = bch2_trans_relock(trans);
+		bch2_trans_begin(trans);
+
+		u32 snapshot;
+		ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot);
 		if (ret)
-			break;
+			goto err;
+
+		bch2_btree_iter_set_snapshot(&iter, snapshot);
 
 		bch2_btree_iter_set_pos(&iter,
 				POS(inum.inum, bvec_iter.bi_sector));
 
 		k = bch2_btree_iter_peek_slot(&iter);
 		ret = bkey_err(k);
 		if (ret)
-			break;
+			goto err;
 
 		offset_into_extent = iter.pos.offset -
 			bkey_start_offset(k.k);
@@ -1193,7 +1186,7 @@ void __bch2_read(struct bch_fs *c, struct bch_read_bio *rbio,
 		ret = bch2_read_indirect_extent(trans, &data_btree,
 					&offset_into_extent, &sk);
 		if (ret)
-			break;
+			goto err;
 
 		k = bkey_i_to_s_c(sk.k);
 
@@ -1213,7 +1206,7 @@ void __bch2_read(struct bch_fs *c, struct bch_read_bio *rbio,
 					 data_btree, k,
 					 offset_into_extent, failed, flags);
 		if (ret)
-			break;
+			goto err;
 
 		if (flags & BCH_READ_LAST_FRAGMENT)
 			break;
@@ -1223,16 +1216,16 @@ void __bch2_read(struct bch_fs *c, struct bch_read_bio *rbio,
 
 		ret = btree_trans_too_many_iters(trans);
 		if (ret)
+			goto err;
+err:
+		if (ret &&
+		    !bch2_err_matches(ret, BCH_ERR_transaction_restart) &&
+		    ret != READ_RETRY &&
+		    ret != READ_RETRY_AVOID)
 			break;
 	}
-err:
-	bch2_trans_iter_exit(trans, &iter);
-
-	if (bch2_err_matches(ret, BCH_ERR_transaction_restart) ||
-	    ret == READ_RETRY ||
-	    ret == READ_RETRY_AVOID)
-		goto retry;
 
+	bch2_trans_iter_exit(trans, &iter);
 	bch2_trans_put(trans);
 	bch2_bkey_buf_exit(&sk, c);
 

libbcachefs/io_write.c

@@ -554,7 +554,7 @@ static void __bch2_write_index(struct bch_write_op *op)
 err:
 	keys->top = keys->keys;
 	op->error = ret;
-	op->flags |= BCH_WRITE_DONE;
+	op->flags |= BCH_WRITE_SUBMITTED;
 	goto out;
 }
 
@@ -589,7 +589,7 @@ static CLOSURE_CALLBACK(bch2_write_index)
 	struct workqueue_struct *wq = index_update_wq(op);
 	unsigned long flags;
 
-	if ((op->flags & BCH_WRITE_DONE) &&
+	if ((op->flags & BCH_WRITE_SUBMITTED) &&
 	    (op->flags & BCH_WRITE_MOVE))
 		bch2_bio_free_pages_pool(op->c, &op->wbio.bio);
 
@@ -634,7 +634,7 @@ void bch2_write_point_do_index_updates(struct work_struct *work)
 
 		__bch2_write_index(op);
 
-		if (!(op->flags & BCH_WRITE_DONE))
+		if (!(op->flags & BCH_WRITE_SUBMITTED))
 			__bch2_write(op);
 		else
 			bch2_write_done(&op->cl);
@@ -1318,7 +1318,7 @@ static void bch2_nocow_write(struct bch_write_op *op)
 			wbio_init(bio)->put_bio = true;
 			bio->bi_opf = op->wbio.bio.bi_opf;
 		} else {
-			op->flags |= BCH_WRITE_DONE;
+			op->flags |= BCH_WRITE_SUBMITTED;
 		}
 
 		op->pos.offset += bio_sectors(bio);
@@ -1332,7 +1332,7 @@ static void bch2_nocow_write(struct bch_write_op *op)
 					  op->insert_keys.top, true);
 
 		bch2_keylist_push(&op->insert_keys);
-		if (op->flags & BCH_WRITE_DONE)
+		if (op->flags & BCH_WRITE_SUBMITTED)
 			break;
 		bch2_btree_iter_advance(&iter);
 	}
@@ -1347,14 +1347,14 @@ static void bch2_nocow_write(struct bch_write_op *op)
 			op->pos.inode, op->pos.offset << 9,
 			"%s: btree lookup error %s", __func__, bch2_err_str(ret));
 		op->error = ret;
-		op->flags |= BCH_WRITE_DONE;
+		op->flags |= BCH_WRITE_SUBMITTED;
 	}
 
 	bch2_trans_put(trans);
 	darray_exit(&buckets);
 
 	/* fallback to cow write path? */
-	if (!(op->flags & BCH_WRITE_DONE)) {
+	if (!(op->flags & BCH_WRITE_SUBMITTED)) {
 		closure_sync(&op->cl);
 		__bch2_nocow_write_done(op);
 		op->insert_keys.top = op->insert_keys.keys;
@@ -1410,7 +1410,7 @@ static void __bch2_write(struct bch_write_op *op)
 
 	if (unlikely(op->opts.nocow && c->opts.nocow_enabled)) {
 		bch2_nocow_write(op);
-		if (op->flags & BCH_WRITE_DONE)
+		if (op->flags & BCH_WRITE_SUBMITTED)
 			goto out_nofs_restore;
 	}
 again:
@@ -1465,7 +1465,7 @@ static void __bch2_write(struct bch_write_op *op)
 		bch2_alloc_sectors_done_inlined(c, wp);
 err:
 		if (ret <= 0) {
-			op->flags |= BCH_WRITE_DONE;
+			op->flags |= BCH_WRITE_SUBMITTED;
 
 			if (ret < 0) {
 				if (!(op->flags & BCH_WRITE_ALLOC_NOWAIT))
@@ -1501,7 +1501,7 @@ static void __bch2_write(struct bch_write_op *op)
 	 * once, as that signals backpressure to the caller.
 	 */
 	if ((op->flags & BCH_WRITE_SYNC) ||
-	    (!(op->flags & BCH_WRITE_DONE) &&
+	    (!(op->flags & BCH_WRITE_SUBMITTED) &&
 	     !(op->flags & BCH_WRITE_IN_WORKER))) {
 		if (closure_sync_timeout(&op->cl, HZ * 10)) {
 			bch2_print_allocator_stuck(c);
@@ -1510,7 +1510,7 @@ static void __bch2_write(struct bch_write_op *op)
 
 		__bch2_write_index(op);
 
-		if (!(op->flags & BCH_WRITE_DONE))
+		if (!(op->flags & BCH_WRITE_SUBMITTED))
 			goto again;
 		bch2_write_done(&op->cl);
 	} else {
@@ -1532,7 +1532,7 @@ static void bch2_write_data_inline(struct bch_write_op *op, unsigned data_len)
 	memset(&op->failed, 0, sizeof(op->failed));
 
 	op->flags |= BCH_WRITE_WROTE_DATA_INLINE;
-	op->flags |= BCH_WRITE_DONE;
+	op->flags |= BCH_WRITE_SUBMITTED;
 
 	bch2_check_set_feature(op->c, BCH_FEATURE_inline_data);
 

libbcachefs/io_write.h

@@ -33,7 +33,7 @@ void bch2_submit_wbio_replicas(struct bch_write_bio *, struct bch_fs *,
 	x(SYNC)				\
 	x(MOVE)				\
 	x(IN_WORKER)			\
-	x(DONE)				\
+	x(SUBMITTED)			\
 	x(IO_ERROR)			\
 	x(CONVERT_UNWRITTEN)
 

libbcachefs/journal_reclaim.c

@@ -205,6 +205,17 @@ void bch2_journal_space_available(struct journal *j)
 	j->can_discard = can_discard;
 
 	if (nr_online < metadata_replicas_required(c)) {
+		struct printbuf buf = PRINTBUF;
+		prt_printf(&buf, "insufficient writeable journal devices available: have %u, need %u\n"
+			   "rw journal devs:", nr_online, metadata_replicas_required(c));
+
+		rcu_read_lock();
+		for_each_member_device_rcu(c, ca, &c->rw_devs[BCH_DATA_journal])
+			prt_printf(&buf, " %s", ca->name);
+		rcu_read_unlock();
+
+		bch_err(c, "%s", buf.buf);
+		printbuf_exit(&buf);
 		ret = JOURNAL_ERR_insufficient_devices;
 		goto out;
 	}

libbcachefs/mean_and_variance.h

@@ -111,11 +111,11 @@ static inline u128_u u128_shl(u128_u i, s8 shift)
 {
 	u128_u r;
 
-	r.lo = i.lo << shift;
+	r.lo = i.lo << (shift & 63);
 	if (shift < 64)
-		r.hi = (i.hi << shift) | (i.lo >> (64 - shift));
+		r.hi = (i.hi << (shift & 63)) | (i.lo >> (-shift & 63));
 	else {
-		r.hi = i.lo << (shift - 64);
+		r.hi = i.lo << (-shift & 63);
 		r.lo = 0;
 	}
 	return r;