bcachefs: Data move path now uses bch2_trans_unlock_long()

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

5 files changed, 23 insertions, 9 deletions

diff --git a/fs/bcachefs/btree_iter.c b/fs/bcachefs/btree_iter.c
index b22fd395a1fd..af98545e0f35 100644
--- a/fs/bcachefs/btree_iter.c
+++ b/fs/bcachefs/btree_iter.c
@@ -2833,6 +2833,13 @@ void *__bch2_trans_kmalloc(struct btree_trans *trans, size_t size)
 	return p;
 }
 
+static inline void check_srcu_held_too_long(struct btree_trans *trans)
+{
+	WARN(trans->srcu_held && time_after(jiffies, trans->srcu_lock_time + HZ * 10),
+	     "btree trans held srcu lock (delaying memory reclaim) for %lu seconds",
+	     (jiffies - trans->srcu_lock_time) / HZ);
+}
+
 void bch2_trans_srcu_unlock(struct btree_trans *trans)
 {
 	if (trans->srcu_held) {
@@ -2843,6 +2850,7 @@ void bch2_trans_srcu_unlock(struct btree_trans *trans)
 			if (path->cached && !btree_node_locked(path, 0))
 				path->l[0].b = ERR_PTR(-BCH_ERR_no_btree_node_srcu_reset);
 
+		check_srcu_held_too_long(trans);
 		srcu_read_unlock(&c->btree_trans_barrier, trans->srcu_idx);
 		trans->srcu_held = false;
 	}
@@ -3074,8 +3082,10 @@ void bch2_trans_put(struct btree_trans *trans)
 
 	check_btree_paths_leaked(trans);
 
-	if (trans->srcu_held)
+	if (trans->srcu_held) {
+		check_srcu_held_too_long(trans);
 		srcu_read_unlock(&c->btree_trans_barrier, trans->srcu_idx);
+	}
 
 	bch2_journal_preres_put(&c->journal, &trans->journal_preres);
 
diff --git a/fs/bcachefs/move.c b/fs/bcachefs/move.c
index 1b15b010461a..ab749bf2fcbc 100644
--- a/fs/bcachefs/move.c
+++ b/fs/bcachefs/move.c
@@ -147,9 +147,8 @@ void bch2_moving_ctxt_do_pending_writes(struct moving_context *ctxt)
 {
 	struct moving_io *io;
 
-	bch2_trans_unlock(ctxt->trans);
-
 	while ((io = bch2_moving_ctxt_next_pending_write(ctxt))) {
+		bch2_trans_unlock_long(ctxt->trans);
 		list_del(&io->read_list);
 		move_write(io);
 	}
@@ -485,8 +484,8 @@ int bch2_move_ratelimit(struct moving_context *ctxt)
 	struct bch_fs *c = ctxt->trans->c;
 	u64 delay;
 
-	if (ctxt->wait_on_copygc) {
-		bch2_trans_unlock(ctxt->trans);
+	if (ctxt->wait_on_copygc && !c->copygc_running) {
+		bch2_trans_unlock_long(ctxt->trans);
 		wait_event_killable(c->copygc_running_wq,
 				    !c->copygc_running ||
 				    kthread_should_stop());
@@ -495,8 +494,12 @@ int bch2_move_ratelimit(struct moving_context *ctxt)
 	do {
 		delay = ctxt->rate ? bch2_ratelimit_delay(ctxt->rate) : 0;
 
+
 		if (delay) {
-			bch2_trans_unlock(ctxt->trans);
+			if (delay > HZ / 10)
+				bch2_trans_unlock_long(ctxt->trans);
+			else
+				bch2_trans_unlock(ctxt->trans);
 			set_current_state(TASK_INTERRUPTIBLE);
 		}
 
diff --git a/fs/bcachefs/move.h b/fs/bcachefs/move.h
index 1b1e8678bfae..07cf9d42643b 100644
--- a/fs/bcachefs/move.h
+++ b/fs/bcachefs/move.h
@@ -45,6 +45,7 @@ do {									\
 									\
 	if (_cond)							\
 		break;							\
+	bch2_trans_unlock_long((_ctxt)->trans);				\
 	__wait_event((_ctxt)->wait,					\
 		     bch2_moving_ctxt_next_pending_write(_ctxt) ||	\
 		     (cond_finished = (_cond)));			\
diff --git a/fs/bcachefs/movinggc.c b/fs/bcachefs/movinggc.c
index e0efa5282a77..4d955f3cc5b2 100644
--- a/fs/bcachefs/movinggc.c
+++ b/fs/bcachefs/movinggc.c
@@ -128,7 +128,7 @@ static void move_buckets_wait(struct moving_context *ctxt,
 		kfree(i);
 	}
 
-	bch2_trans_unlock(ctxt->trans);
+	bch2_trans_unlock_long(ctxt->trans);
 }
 
 static bool bucket_in_flight(struct buckets_in_flight *list,
@@ -327,7 +327,7 @@ static int bch2_copygc_thread(void *arg)
 	while (!ret && !kthread_should_stop()) {
 		bool did_work = false;
 
-		bch2_trans_unlock(ctxt.trans);
+		bch2_trans_unlock_long(ctxt.trans);
 		cond_resched();
 
 		if (!c->copy_gc_enabled) {
diff --git a/fs/bcachefs/rebalance.c b/fs/bcachefs/rebalance.c
index 82014cc6e271..3319190b8d9c 100644
--- a/fs/bcachefs/rebalance.c
+++ b/fs/bcachefs/rebalance.c
@@ -348,7 +348,7 @@ static int do_rebalance(struct moving_context *ctxt)
 	    !kthread_should_stop() &&
 	    !atomic64_read(&r->work_stats.sectors_seen) &&
 	    !atomic64_read(&r->scan_stats.sectors_seen)) {
-		bch2_trans_unlock(trans);
+		bch2_trans_unlock_long(trans);
 		rebalance_wait(c);
 	}