bcachefs: Improve btree iterator tracepoints

This patch adds some new tracepoints to the btree iterator code, and
adds new fields to the existing tracepoints - primarily for the iterator
position.

Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

7 files changed, 429 insertions, 198 deletions

diff --git a/fs/bcachefs/btree_cache.c b/fs/bcachefs/btree_cache.c
index f6adbe8955d7..013cf0b53857 100644
--- a/fs/bcachefs/btree_cache.c
+++ b/fs/bcachefs/btree_cache.c
@@ -807,7 +807,10 @@ lock_node:
 			if (bch2_btree_node_relock(iter, level + 1))
 				goto retry;
 
-			trace_trans_restart_btree_node_reused(iter->trans->ip);
+			trace_trans_restart_btree_node_reused(iter->trans->ip,
+							      trace_ip,
+							      iter->btree_id,
+							      &iter->real_pos);
 			return ERR_PTR(-EINTR);
 		}
 	}
diff --git a/fs/bcachefs/btree_iter.c b/fs/bcachefs/btree_iter.c
index bdb068e9d263..8840531266d7 100644
--- a/fs/bcachefs/btree_iter.c
+++ b/fs/bcachefs/btree_iter.c
@@ -170,8 +170,8 @@ success:
 	return true;
 }
 
-static inline bool btree_iter_get_locks(struct btree_iter *iter,
-					bool upgrade, bool trace)
+static inline bool btree_iter_get_locks(struct btree_iter *iter, bool upgrade,
+					unsigned long trace_ip)
 {
 	unsigned l = iter->level;
 	int fail_idx = -1;
@@ -183,16 +183,17 @@ static inline bool btree_iter_get_locks(struct btree_iter *iter,
 		if (!(upgrade
 		      ? bch2_btree_node_upgrade(iter, l)
 		      : bch2_btree_node_relock(iter, l))) {
-			if (trace)
-				(upgrade
-				 ? trace_node_upgrade_fail
-				 : trace_node_relock_fail)(l, iter->l[l].lock_seq,
-						is_btree_node(iter, l)
-						? 0
-						: (unsigned long) iter->l[l].b,
-						is_btree_node(iter, l)
-						? iter->l[l].b->c.lock.state.seq
-						: 0);
+			(upgrade
+			 ? trace_node_upgrade_fail
+			 : trace_node_relock_fail)(iter->trans->ip, trace_ip,
+					iter->btree_id, &iter->real_pos,
+					l, iter->l[l].lock_seq,
+					is_btree_node(iter, l)
+					? 0
+					: (unsigned long) iter->l[l].b,
+					is_btree_node(iter, l)
+					? iter->l[l].b->c.lock.state.seq
+					: 0);
 
 			fail_idx = l;
 			btree_iter_set_dirty(iter, BTREE_ITER_NEED_TRAVERSE);
@@ -372,9 +373,9 @@ static inline void bch2_btree_iter_verify_locks(struct btree_iter *iter) {}
 #endif
 
 __flatten
-bool bch2_btree_iter_relock(struct btree_iter *iter, bool trace)
+static bool bch2_btree_iter_relock(struct btree_iter *iter, unsigned long trace_ip)
 {
-	return btree_iter_get_locks(iter, false, trace);
+	return btree_iter_get_locks(iter, false, trace_ip);
 }
 
 bool __bch2_btree_iter_upgrade(struct btree_iter *iter,
@@ -386,7 +387,7 @@ bool __bch2_btree_iter_upgrade(struct btree_iter *iter,
 
 	iter->locks_want = new_locks_want;
 
-	if (btree_iter_get_locks(iter, true, true))
+	if (btree_iter_get_locks(iter, true, _THIS_IP_))
 		return true;
 
 	/*
@@ -414,7 +415,7 @@ bool __bch2_btree_iter_upgrade(struct btree_iter *iter,
 		    linked->btree_id == iter->btree_id &&
 		    linked->locks_want < new_locks_want) {
 			linked->locks_want = new_locks_want;
-			btree_iter_get_locks(linked, true, false);
+			btree_iter_get_locks(linked, true, _THIS_IP_);
 		}
 
 	return false;
@@ -460,8 +461,9 @@ bool bch2_trans_relock(struct btree_trans *trans)
 	struct btree_iter *iter;
 
 	trans_for_each_iter(trans, iter)
-		if (!bch2_btree_iter_relock(iter, true)) {
-			trace_trans_restart_relock(trans->ip);
+		if (!bch2_btree_iter_relock(iter, _RET_IP_)) {
+			trace_trans_restart_relock(trans->ip, _RET_IP_,
+					iter->btree_id, &iter->real_pos);
 			return false;
 		}
 	return true;
@@ -1175,7 +1177,8 @@ err:
 
 static int btree_iter_traverse_one(struct btree_iter *, unsigned long);
 
-static int __btree_iter_traverse_all(struct btree_trans *trans, int ret)
+static int __btree_iter_traverse_all(struct btree_trans *trans, int ret,
+				     unsigned long trace_ip)
 {
 	struct bch_fs *c = trans->c;
 	struct btree_iter *iter;
@@ -1192,7 +1195,7 @@ retry_all:
 	relock_fail = false;
 
 	trans_for_each_iter(trans, iter) {
-		if (!bch2_btree_iter_relock(iter, true))
+		if (!bch2_btree_iter_relock(iter, _THIS_IP_))
 			relock_fail = true;
 		sorted[nr_sorted++] = iter->idx;
 	}
@@ -1269,13 +1272,13 @@ out:
 
 	trans->in_traverse_all = false;
 
-	trace_trans_traverse_all(trans->ip);
+	trace_trans_traverse_all(trans->ip, trace_ip);
 	return ret;
 }
 
 int bch2_btree_iter_traverse_all(struct btree_trans *trans)
 {
-	return __btree_iter_traverse_all(trans, 0);
+	return __btree_iter_traverse_all(trans, 0, _RET_IP_);
 }
 
 static inline bool btree_iter_good_node(struct btree_iter *iter,
@@ -1320,6 +1323,7 @@ static int btree_iter_traverse_one(struct btree_iter *iter,
 				   unsigned long trace_ip)
 {
 	unsigned depth_want = iter->level;
+	int ret = 0;
 
 	/*
 	 * if we need interior nodes locked, call btree_iter_relock() to make
@@ -1327,16 +1331,18 @@ static int btree_iter_traverse_one(struct btree_iter *iter,
 	 */
 	if (iter->uptodate == BTREE_ITER_NEED_RELOCK ||
 	    iter->locks_want > 1)
-		bch2_btree_iter_relock(iter, false);
+		bch2_btree_iter_relock(iter, _THIS_IP_);
 
-	if (btree_iter_type(iter) == BTREE_ITER_CACHED)
-		return bch2_btree_iter_traverse_cached(iter);
+	if (btree_iter_type(iter) == BTREE_ITER_CACHED) {
+		ret = bch2_btree_iter_traverse_cached(iter);
+		goto out;
+	}
 
 	if (iter->uptodate < BTREE_ITER_NEED_RELOCK)
-		return 0;
+		goto out;
 
 	if (unlikely(iter->level >= BTREE_MAX_DEPTH))
-		return 0;
+		goto out;
 
 	iter->level = btree_iter_up_until_good_node(iter, 0);
 
@@ -1347,12 +1353,18 @@ static int btree_iter_traverse_one(struct btree_iter *iter,
 	 * btree_iter_lock_root() comes next and that it can't fail
 	 */
 	while (iter->level > depth_want) {
-		int ret = btree_iter_node(iter, iter->level)
+		ret = btree_iter_node(iter, iter->level)
 			? btree_iter_down(iter, trace_ip)
 			: btree_iter_lock_root(iter, depth_want, trace_ip);
 		if (unlikely(ret)) {
-			if (ret == 1)
-				return 0;
+			if (ret == 1) {
+				/*
+				 * Got to the end of the btree (in
+				 * BTREE_ITER_NODES mode)
+				 */
+				ret = 0;
+				goto out;
+			}
 
 			iter->level = depth_want;
 
@@ -1364,14 +1376,16 @@ static int btree_iter_traverse_one(struct btree_iter *iter,
 				iter->l[iter->level].b =
 					BTREE_ITER_NO_NODE_DOWN;
 			}
-			return ret;
+			goto out;
 		}
 	}
 
 	iter->uptodate = BTREE_ITER_NEED_PEEK;
-
+out:
+	trace_iter_traverse(iter->trans->ip, trace_ip,
+			    iter->btree_id, &iter->real_pos, ret);
 	bch2_btree_iter_verify(iter);
-	return 0;
+	return ret;
 }
 
 static int __must_check __bch2_btree_iter_traverse(struct btree_iter *iter)
@@ -1382,7 +1396,7 @@ static int __must_check __bch2_btree_iter_traverse(struct btree_iter *iter)
 	ret =   bch2_trans_cond_resched(trans) ?:
 		btree_iter_traverse_one(iter, _RET_IP_);
 	if (unlikely(ret))
-		ret = __btree_iter_traverse_all(trans, ret);
+		ret = __btree_iter_traverse_all(trans, ret, _RET_IP_);
 
 	return ret;
 }
@@ -1498,6 +1512,7 @@ struct btree *bch2_btree_iter_next_node(struct btree_iter *iter)
 
 static void btree_iter_set_search_pos(struct btree_iter *iter, struct bpos new_pos)
 {
+	struct bpos old_pos = iter->real_pos;
 	int cmp = bpos_cmp(new_pos, iter->real_pos);
 	unsigned l = iter->level;
 
@@ -1508,7 +1523,7 @@ static void btree_iter_set_search_pos(struct btree_iter *iter, struct bpos new_p
 
 	if (unlikely(btree_iter_type(iter) == BTREE_ITER_CACHED)) {
 		btree_node_unlock(iter, 0);
-		iter->l[0].b = BTREE_ITER_NO_NODE_UP;
+		iter->l[0].b = BTREE_ITER_NO_NODE_CACHED;
 		btree_iter_set_dirty(iter, BTREE_ITER_NEED_TRAVERSE);
 		return;
 	}
@@ -1537,6 +1552,11 @@ out:
 		btree_iter_set_dirty(iter, BTREE_ITER_NEED_PEEK);
 
 	bch2_btree_iter_verify(iter);
+#ifdef CONFIG_BCACHEFS_DEBUG
+	trace_iter_set_search_pos(iter->trans->ip, _RET_IP_,
+				  iter->btree_id,
+				  &old_pos, &new_pos, l);
+#endif
 }
 
 inline bool bch2_btree_iter_advance(struct btree_iter *iter)
@@ -2055,13 +2075,6 @@ struct btree_iter *__bch2_trans_get_iter(struct btree_trans *trans,
 		best = iter;
 	}
 
-	trace_trans_get_iter(_RET_IP_, trans->ip,
-			     btree_id,
-			     &real_pos, locks_want,
-			     best ? &best->real_pos : &pos_min,
-			     best ? best->locks_want : 0,
-			     best ? best->uptodate : BTREE_ITER_NEED_TRAVERSE);
-
 	if (!best) {
 		iter = btree_trans_iter_alloc(trans);
 		bch2_btree_iter_init(trans, iter, btree_id);
@@ -2090,7 +2103,7 @@ struct btree_iter *__bch2_trans_get_iter(struct btree_trans *trans,
 	locks_want = min(locks_want, BTREE_MAX_DEPTH);
 	if (locks_want > iter->locks_want) {
 		iter->locks_want = locks_want;
-		btree_iter_get_locks(iter, true, false);
+		btree_iter_get_locks(iter, true, _THIS_IP_);
 	}
 
 	while (iter->level != depth) {
@@ -2108,6 +2121,13 @@ struct btree_iter *__bch2_trans_get_iter(struct btree_trans *trans,
 	bch2_btree_iter_set_pos(iter, pos);
 	btree_iter_set_search_pos(iter, real_pos);
 
+	trace_trans_get_iter(_RET_IP_, trans->ip,
+			     btree_id,
+			     &real_pos, locks_want, iter->uptodate,
+			     best ? &best->real_pos	: &pos_min,
+			     best ? best->locks_want	: U8_MAX,
+			     best ? best->uptodate	: U8_MAX);
+
 	return iter;
 }
 
diff --git a/fs/bcachefs/btree_iter.h b/fs/bcachefs/btree_iter.h
index 2f63adb9e420..01b834bf79f7 100644
--- a/fs/bcachefs/btree_iter.h
+++ b/fs/bcachefs/btree_iter.h
@@ -111,7 +111,6 @@ void bch2_btree_node_iter_fix(struct btree_iter *, struct btree *,
 			      struct btree_node_iter *, struct bkey_packed *,
 			      unsigned, unsigned);
 
-bool bch2_btree_iter_relock(struct btree_iter *, bool);
 bool bch2_trans_relock(struct btree_trans *);
 void bch2_trans_unlock(struct btree_trans *);
 
diff --git a/fs/bcachefs/btree_types.h b/fs/bcachefs/btree_types.h
index 06a2c412db7a..172a90f4db4f 100644
--- a/fs/bcachefs/btree_types.h
+++ b/fs/bcachefs/btree_types.h
@@ -230,6 +230,7 @@ enum btree_iter_uptodate {
 #define BTREE_ITER_NO_NODE_DOWN		((struct btree *) 5)
 #define BTREE_ITER_NO_NODE_INIT		((struct btree *) 6)
 #define BTREE_ITER_NO_NODE_ERROR	((struct btree *) 7)
+#define BTREE_ITER_NO_NODE_CACHED	((struct btree *) 8)
 
 /*
  * @pos			- iterator's current position
diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c
index 95e6d21dac2a..2d8093d1bf00 100644
--- a/fs/bcachefs/btree_update_interior.c
+++ b/fs/bcachefs/btree_update_interior.c
@@ -955,7 +955,9 @@ retry:
 	 * instead of locking/reserving all the way to the root:
 	 */
 	if (!bch2_btree_iter_upgrade(iter, U8_MAX)) {
-		trace_trans_restart_iter_upgrade(trans->ip);
+		trace_trans_restart_iter_upgrade(trans->ip, _RET_IP_,
+						 iter->btree_id,
+						 &iter->real_pos);
 		return ERR_PTR(-EINTR);
 	}
 
@@ -996,7 +998,7 @@ retry:
 		 * closure argument
 		 */
 		if (flags & BTREE_INSERT_NOUNLOCK) {
-			trace_trans_restart_journal_preres_get(trans->ip);
+			trace_trans_restart_journal_preres_get(trans->ip, _RET_IP_);
 			ret = -EINTR;
 			goto err;
 		}
@@ -1012,7 +1014,7 @@ retry:
 				BTREE_UPDATE_JOURNAL_RES,
 				journal_flags);
 		if (ret) {
-			trace_trans_restart_journal_preres_get(trans->ip);
+			trace_trans_restart_journal_preres_get(trans->ip, _RET_IP_);
 			goto err;
 		}
 
diff --git a/fs/bcachefs/btree_update_leaf.c b/fs/bcachefs/btree_update_leaf.c
index cc7acce171a2..0d566be7455e 100644
--- a/fs/bcachefs/btree_update_leaf.c
+++ b/fs/bcachefs/btree_update_leaf.c
@@ -228,7 +228,8 @@ static inline void btree_insert_entry_checks(struct btree_trans *trans,
 }
 
 static noinline int
-bch2_trans_journal_preres_get_cold(struct btree_trans *trans, unsigned u64s)
+bch2_trans_journal_preres_get_cold(struct btree_trans *trans, unsigned u64s,
+				   unsigned long trace_ip)
 {
 	struct bch_fs *c = trans->c;
 	int ret;
@@ -241,7 +242,7 @@ bch2_trans_journal_preres_get_cold(struct btree_trans *trans, unsigned u64s)
 		return ret;
 
 	if (!bch2_trans_relock(trans)) {
-		trace_trans_restart_journal_preres_get(trans->ip);
+		trace_trans_restart_journal_preres_get(trans->ip, trace_ip);
 		return -EINTR;
 	}
 
@@ -368,7 +369,8 @@ static noinline void bch2_trans_mark_gc(struct btree_trans *trans)
 
 static inline int
 bch2_trans_commit_write_locked(struct btree_trans *trans,
-			       struct btree_insert_entry **stopped_at)
+			       struct btree_insert_entry **stopped_at,
+			       unsigned long trace_ip)
 {
 	struct bch_fs *c = trans->c;
 	struct btree_insert_entry *i;
@@ -378,7 +380,7 @@ bch2_trans_commit_write_locked(struct btree_trans *trans,
 	int ret;
 
 	if (race_fault()) {
-		trace_trans_restart_fault_inject(trans->ip);
+		trace_trans_restart_fault_inject(trans->ip, trace_ip);
 		return -EINTR;
 	}
 
@@ -525,7 +527,8 @@ static noinline int maybe_do_btree_merge(struct btree_trans *trans, struct btree
  * Get journal reservation, take write locks, and attempt to do btree update(s):
  */
 static inline int do_bch2_trans_commit(struct btree_trans *trans,
-				       struct btree_insert_entry **stopped_at)
+				       struct btree_insert_entry **stopped_at,
+				       unsigned long trace_ip)
 {
 	struct bch_fs *c = trans->c;
 	struct btree_insert_entry *i;
@@ -559,7 +562,7 @@ static inline int do_bch2_trans_commit(struct btree_trans *trans,
 			 ? JOURNAL_RES_GET_RESERVED : 0));
 	if (unlikely(ret == -EAGAIN))
 		ret = bch2_trans_journal_preres_get_cold(trans,
-						trans->journal_preres_u64s);
+						trans->journal_preres_u64s, trace_ip);
 	if (unlikely(ret))
 		return ret;
 
@@ -578,7 +581,9 @@ static inline int do_bch2_trans_commit(struct btree_trans *trans,
 		if (iter->nodes_locked != iter->nodes_intent_locked) {
 			if (btree_iter_keep(trans, iter)) {
 				if (!bch2_btree_iter_upgrade(iter, 1)) {
-					trace_trans_restart_upgrade(trans->ip);
+					trace_trans_restart_upgrade(trans->ip, trace_ip,
+								    iter->btree_id,
+								    &iter->real_pos);
 					return -EINTR;
 				}
 			} else {
@@ -606,7 +611,7 @@ static inline int do_bch2_trans_commit(struct btree_trans *trans,
 			bch2_btree_node_lock_for_insert(c,
 					iter_l(i->iter)->b, i->iter);
 
-	ret = bch2_trans_commit_write_locked(trans, stopped_at);
+	ret = bch2_trans_commit_write_locked(trans, stopped_at, trace_ip);
 
 	trans_for_each_update2(trans, i)
 		if (!same_leaf_as_prev(trans, i))
@@ -644,7 +649,7 @@ static int journal_reclaim_wait_done(struct bch_fs *c)
 static noinline
 int bch2_trans_commit_error(struct btree_trans *trans,
 			    struct btree_insert_entry *i,
-			    int ret)
+			    int ret, unsigned long trace_ip)
 {
 	struct bch_fs *c = trans->c;
 	unsigned flags = trans->flags;
@@ -685,7 +690,9 @@ int bch2_trans_commit_error(struct btree_trans *trans,
 		if (!ret ||
 		    ret == -EINTR ||
 		    (flags & BTREE_INSERT_NOUNLOCK)) {
-			trace_trans_restart_btree_node_split(trans->ip);
+			trace_trans_restart_btree_node_split(trans->ip, trace_ip,
+							     i->iter->btree_id,
+							     &i->iter->real_pos);
 			ret = -EINTR;
 		}
 		break;
@@ -703,7 +710,7 @@ int bch2_trans_commit_error(struct btree_trans *trans,
 		if (bch2_trans_relock(trans))
 			return 0;
 
-		trace_trans_restart_mark_replicas(trans->ip);
+		trace_trans_restart_mark_replicas(trans->ip, trace_ip);
 		ret = -EINTR;
 		break;
 	case BTREE_INSERT_NEED_JOURNAL_RES:
@@ -720,13 +727,13 @@ int bch2_trans_commit_error(struct btree_trans *trans,
 		if (bch2_trans_relock(trans))
 			return 0;
 
-		trace_trans_restart_journal_res_get(trans->ip);
+		trace_trans_restart_journal_res_get(trans->ip, trace_ip);
 		ret = -EINTR;
 		break;
 	case BTREE_INSERT_NEED_JOURNAL_RECLAIM:
 		bch2_trans_unlock(trans);
 
-		trace_trans_blocked_journal_reclaim(trans->ip);
+		trace_trans_blocked_journal_reclaim(trans->ip, trace_ip);
 
 		wait_event_freezable(c->journal.reclaim_wait,
 				     (ret = journal_reclaim_wait_done(c)));
@@ -736,7 +743,7 @@ int bch2_trans_commit_error(struct btree_trans *trans,
 		if (bch2_trans_relock(trans))
 			return 0;
 
-		trace_trans_restart_journal_reclaim(trans->ip);
+		trace_trans_restart_journal_reclaim(trans->ip, trace_ip);
 		ret = -EINTR;
 		break;
 	default:
@@ -950,7 +957,9 @@ int __bch2_trans_commit(struct btree_trans *trans)
 							     i->trigger_flags);
 				if (unlikely(ret)) {
 					if (ret == -EINTR)
-						trace_trans_restart_mark(trans->ip);
+						trace_trans_restart_mark(trans->ip, _RET_IP_,
+									 i->iter->btree_id,
+									 &i->iter->pos);
 					goto out;
 				}
 			}
@@ -976,12 +985,16 @@ int __bch2_trans_commit(struct btree_trans *trans)
 	trans_for_each_update2(trans, i) {
 		ret = bch2_btree_iter_traverse(i->iter);
 		if (unlikely(ret)) {
-			trace_trans_restart_traverse(trans->ip);
+			trace_trans_restart_traverse(trans->ip, _RET_IP_,
+						     i->iter->btree_id,
+						     &i->iter->pos);
 			goto out;
 		}
 
 		if (unlikely(!bch2_btree_iter_upgrade(i->iter, i->level + 1))) {
-			trace_trans_restart_upgrade(trans->ip);
+			trace_trans_restart_upgrade(trans->ip, _RET_IP_,
+						    i->iter->btree_id,
+						    &i->iter->pos);
 			ret = -EINTR;
 			goto out;
 		}
@@ -997,7 +1010,7 @@ int __bch2_trans_commit(struct btree_trans *trans)
 retry:
 	memset(&trans->journal_res, 0, sizeof(trans->journal_res));
 
-	ret = do_bch2_trans_commit(trans, &i);
+	ret = do_bch2_trans_commit(trans, &i, _RET_IP_);
 
 	/* make sure we didn't drop or screw up locks: */
 	bch2_btree_trans_verify_locks(trans);
@@ -1023,7 +1036,7 @@ out_reset:
 
 	return ret;
 err:
-	ret = bch2_trans_commit_error(trans, i, ret);
+	ret = bch2_trans_commit_error(trans, i, ret, _RET_IP_);
 	if (ret)
 		goto out;
 
diff --git a/include/trace/events/bcachefs.h b/include/trace/events/bcachefs.h
index 7c90ba01510a..b5fcda9e65d8 100644
--- a/include/trace/events/bcachefs.h
+++ b/include/trace/events/bcachefs.h
@@ -541,59 +541,66 @@ TRACE_EVENT(copygc_wait,
 );
 
 TRACE_EVENT(trans_get_iter,
-	TP_PROTO(unsigned long caller, unsigned long ip,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip,
 		 enum btree_id btree_id,
-		 struct bpos *pos_want,
-		 unsigned locks_want,
-		 struct bpos *pos_found,
-		 unsigned locks_found,
-		 unsigned uptodate),
-	TP_ARGS(caller, ip, btree_id,
-		pos_want, locks_want,
-		pos_found, locks_found,
-		uptodate),
+		 struct bpos *got_pos,
+		 unsigned got_locks,
+		 unsigned got_uptodate,
+		 struct bpos *src_pos,
+		 unsigned src_locks,
+		 unsigned src_uptodate),
+	TP_ARGS(trans_ip, caller_ip, btree_id,
+		got_pos, got_locks, got_uptodate,
+		src_pos, src_locks, src_uptodate),
 
 	TP_STRUCT__entry(
-		__field(unsigned long,	caller			)
-		__field(unsigned long,	ip			)
-		__field(u8,		btree_id		)
-		__field(u8,		uptodate		)
-		__field(u8,		locks_want		)
-		__field(u8,		locks_found		)
-		__field(u64,		pos_want_inode		)
-		__field(u64,		pos_want_offset		)
-		__field(u32,		pos_want_snapshot	)
-		__field(u64,		pos_found_inode		)
-		__field(u64,		pos_found_offset	)
-		__field(u32,		pos_found_snapshot	)
+		__field(unsigned long,		trans_ip		)
+		__field(unsigned long,		caller_ip		)
+		__field(u8,			btree_id		)
+		__field(u64,			got_pos_inode		)
+		__field(u64,			got_pos_offset		)
+		__field(u32,			got_pos_snapshot	)
+		__field(u8,			got_locks		)
+		__field(u8,			got_uptodate		)
+		__field(u64,			src_pos_inode		)
+		__field(u64,			src_pos_offset		)
+		__field(u32,			src_pos_snapshot	)
+		__field(u8,			src_locks		)
+		__field(u8,			src_uptodate		)
 	),
 
 	TP_fast_assign(
-		__entry->caller			= caller;
-		__entry->ip			= ip;
+		__entry->trans_ip		= trans_ip;
+		__entry->caller_ip		= caller_ip;
 		__entry->btree_id		= btree_id;
-		__entry->uptodate		= uptodate;
-		__entry->pos_want_inode		= pos_want->inode;
-		__entry->pos_want_offset	= pos_want->offset;
-		__entry->pos_want_snapshot	= pos_want->snapshot;
-		__entry->pos_found_inode	= pos_found->inode;
-		__entry->pos_found_offset	= pos_found->offset;
-		__entry->pos_found_snapshot	= pos_found->snapshot;
-	),
-
-	TP_printk("%ps %pS btree %u uptodate %u want %llu:%llu:%u locks %u found %llu:%llu:%u locks %u",
-		  (void *) __entry->caller,
-		  (void *) __entry->ip,
+		__entry->got_pos_inode		= got_pos->inode;
+		__entry->got_pos_offset		= got_pos->offset;
+		__entry->got_pos_snapshot	= got_pos->snapshot;
+		__entry->got_locks		= got_locks;
+		__entry->got_uptodate		= got_uptodate;
+		__entry->src_pos_inode		= src_pos->inode;
+		__entry->src_pos_offset		= src_pos->offset;
+		__entry->src_pos_snapshot	= src_pos->snapshot;
+		__entry->src_locks		= src_locks;
+		__entry->src_uptodate		= src_uptodate;
+	),
+
+	TP_printk("%ps %pS btree %u got %llu:%llu:%u l %u u %u "
+		  "src %llu:%llu:%u l %u u %u",
+		  (void *) __entry->trans_ip,
+		  (void *) __entry->caller_ip,
 		  __entry->btree_id,
-		  __entry->uptodate,
-		  __entry->pos_want_inode,
-		  __entry->pos_want_offset,
-		  __entry->pos_want_snapshot,
-		  __entry->locks_want,
-		  __entry->pos_found_inode,
-		  __entry->pos_found_offset,
-		  __entry->pos_found_snapshot,
-		  __entry->locks_found)
+		  __entry->got_pos_inode,
+		  __entry->got_pos_offset,
+		  __entry->got_pos_snapshot,
+		  __entry->got_locks,
+		  __entry->got_uptodate,
+		  __entry->src_pos_inode,
+		  __entry->src_pos_offset,
+		  __entry->src_pos_snapshot,
+		  __entry->src_locks,
+		  __entry->src_uptodate)
 );
 
 TRACE_EVENT(transaction_restart_ip,
@@ -614,28 +621,241 @@ TRACE_EVENT(transaction_restart_ip,
 );
 
 DECLARE_EVENT_CLASS(transaction_restart,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip),
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip),
+	TP_ARGS(trans_ip, caller_ip),
 
 	TP_STRUCT__entry(
-		__field(unsigned long,		ip	)
+		__field(unsigned long,		trans_ip	)
+		__field(unsigned long,		caller_ip	)
 	),
 
 	TP_fast_assign(
-		__entry->ip = ip;
+		__entry->trans_ip		= trans_ip;
+		__entry->caller_ip		= caller_ip;
 	),
 
-	TP_printk("%ps", (void *) __entry->ip)
+	TP_printk("%ps %pS",
+		  (void *) __entry->trans_ip,
+		  (void *) __entry->caller_ip)
 );
 
-DEFINE_EVENT(transaction_restart,	trans_restart_btree_node_reused,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip)
+DEFINE_EVENT(transaction_restart,	trans_blocked_journal_reclaim,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip),
+	TP_ARGS(trans_ip, caller_ip)
 );
 
-DEFINE_EVENT(transaction_restart,	trans_blocked_journal_reclaim,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip)
+DEFINE_EVENT(transaction_restart,	trans_restart_journal_res_get,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip),
+	TP_ARGS(trans_ip, caller_ip)
+);
+
+DEFINE_EVENT(transaction_restart,	trans_restart_journal_preres_get,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip),
+	TP_ARGS(trans_ip, caller_ip)
+);
+
+DEFINE_EVENT(transaction_restart,	trans_restart_journal_reclaim,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip),
+	TP_ARGS(trans_ip, caller_ip)
+);
+
+DEFINE_EVENT(transaction_restart,	trans_restart_fault_inject,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip),
+	TP_ARGS(trans_ip, caller_ip)
+);
+
+DEFINE_EVENT(transaction_restart,	trans_traverse_all,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip),
+	TP_ARGS(trans_ip, caller_ip)
+);
+
+DEFINE_EVENT(transaction_restart,	trans_restart_mark_replicas,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip),
+	TP_ARGS(trans_ip, caller_ip)
+);
+
+DECLARE_EVENT_CLASS(transaction_restart_iter,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip,
+		 enum btree_id btree_id,
+		 struct bpos *pos),
+	TP_ARGS(trans_ip, caller_ip, btree_id, pos),
+
+	TP_STRUCT__entry(
+		__field(unsigned long,		trans_ip	)
+		__field(unsigned long,		caller_ip	)
+		__field(u8,			btree_id	)
+		__field(u64,			pos_inode	)
+		__field(u64,			pos_offset	)
+		__field(u32,			pos_snapshot	)
+	),
+
+	TP_fast_assign(
+		__entry->trans_ip		= trans_ip;
+		__entry->caller_ip		= caller_ip;
+		__entry->btree_id		= btree_id;
+		__entry->pos_inode		= pos->inode;
+		__entry->pos_offset		= pos->offset;
+		__entry->pos_snapshot		= pos->snapshot;
+	),
+
+	TP_printk("%ps %pS btree %u pos %llu:%llu:%u",
+		  (void *) __entry->trans_ip,
+		  (void *) __entry->caller_ip,
+		  __entry->btree_id,
+		  __entry->pos_inode,
+		  __entry->pos_offset,
+		  __entry->pos_snapshot)
+);
+
+DEFINE_EVENT(transaction_restart_iter,	trans_restart_btree_node_reused,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip,
+		 enum btree_id btree_id,
+		 struct bpos *pos),
+	TP_ARGS(trans_ip, caller_ip, btree_id, pos)
+);
+
+DEFINE_EVENT(transaction_restart_iter,	trans_restart_btree_node_split,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip,
+		 enum btree_id btree_id,
+		 struct bpos *pos),
+	TP_ARGS(trans_ip, caller_ip, btree_id, pos)
+);
+
+DEFINE_EVENT(transaction_restart_iter,	trans_restart_mark,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip,
+		 enum btree_id btree_id,
+		 struct bpos *pos),
+	TP_ARGS(trans_ip, caller_ip, btree_id, pos)
+);
+
+DEFINE_EVENT(transaction_restart_iter,	trans_restart_upgrade,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip,
+		 enum btree_id btree_id,
+		 struct bpos *pos),
+	TP_ARGS(trans_ip, caller_ip, btree_id, pos)
+);
+
+DEFINE_EVENT(transaction_restart_iter,	trans_restart_iter_upgrade,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip,
+		 enum btree_id btree_id,
+		 struct bpos *pos),
+	TP_ARGS(trans_ip, caller_ip, btree_id, pos)
+);
+
+DEFINE_EVENT(transaction_restart_iter,	trans_restart_relock,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip,
+		 enum btree_id btree_id,
+		 struct bpos *pos),
+	TP_ARGS(trans_ip, caller_ip, btree_id, pos)
+);
+
+DEFINE_EVENT(transaction_restart_iter,	trans_restart_traverse,
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip,
+		 enum btree_id btree_id,
+		 struct bpos *pos),
+	TP_ARGS(trans_ip, caller_ip, btree_id, pos)
+);
+
+TRACE_EVENT(iter_traverse,
+	TP_PROTO(unsigned long	trans_ip,
+		 unsigned long	caller_ip,
+		 enum btree_id	btree_id,
+		 struct bpos	*pos,
+		 int ret),
+	TP_ARGS(trans_ip, caller_ip, btree_id, pos, ret),
+
+	TP_STRUCT__entry(
+		__field(unsigned long,		trans_ip	)
+		__field(unsigned long,		caller_ip	)
+		__field(u8,			btree_id	)
+		__field(u64,			pos_inode	)
+		__field(u64,			pos_offset	)
+		__field(u32,			pos_snapshot	)
+		__field(s32,			ret		)
+	),
+
+	TP_fast_assign(
+		__entry->trans_ip		= trans_ip;
+		__entry->caller_ip		= caller_ip;
+		__entry->btree_id		= btree_id;
+		__entry->pos_inode		= pos->inode;
+		__entry->pos_offset		= pos->offset;
+		__entry->pos_snapshot		= pos->snapshot;
+		__entry->ret			= ret;
+	),
+
+	TP_printk("%ps %pS pos %u %llu:%llu:%u ret %i",
+		  (void *) __entry->trans_ip,
+		  (void *) __entry->caller_ip,
+		  __entry->btree_id,
+		  __entry->pos_inode,
+		  __entry->pos_offset,
+		  __entry->pos_snapshot,
+		  __entry->ret)
+);
+
+TRACE_EVENT(iter_set_search_pos,
+	TP_PROTO(unsigned long	trans_ip,
+		 unsigned long	caller_ip,
+		 enum btree_id	btree_id,
+		 struct bpos	*old_pos,
+		 struct bpos	*new_pos,
+		 unsigned	good_level),
+	TP_ARGS(trans_ip, caller_ip, btree_id, old_pos, new_pos, good_level),
+
+	TP_STRUCT__entry(
+		__field(unsigned long,		trans_ip		)
+		__field(unsigned long,		caller_ip		)
+		__field(u8,			btree_id		)
+		__field(u64,			old_pos_inode		)
+		__field(u64,			old_pos_offset		)
+		__field(u32,			old_pos_snapshot	)
+		__field(u64,			new_pos_inode		)
+		__field(u64,			new_pos_offset		)
+		__field(u32,			new_pos_snapshot	)
+		__field(u8,			good_level		)
+	),
+
+	TP_fast_assign(
+		__entry->trans_ip		= trans_ip;
+		__entry->caller_ip		= caller_ip;
+		__entry->btree_id		= btree_id;
+		__entry->old_pos_inode		= old_pos->inode;
+		__entry->old_pos_offset		= old_pos->offset;
+		__entry->old_pos_snapshot	= old_pos->snapshot;
+		__entry->new_pos_inode		= new_pos->inode;
+		__entry->new_pos_offset		= new_pos->offset;
+		__entry->new_pos_snapshot	= new_pos->snapshot;
+		__entry->good_level		= good_level;
+	),
+
+	TP_printk("%ps %pS btree %u old pos %llu:%llu:%u new pos %llu:%llu:%u l %u",
+		  (void *) __entry->trans_ip,
+		  (void *) __entry->caller_ip,
+		  __entry->btree_id,
+		  __entry->old_pos_inode,
+		  __entry->old_pos_offset,
+		  __entry->old_pos_snapshot,
+		  __entry->new_pos_inode,
+		  __entry->new_pos_offset,
+		  __entry->new_pos_snapshot,
+		  __entry->good_level)
 );
 
 TRACE_EVENT(trans_restart_would_deadlock,
@@ -730,97 +950,70 @@ TRACE_EVENT(trans_restart_mem_realloced,
 		  __entry->bytes)
 );
 
-DEFINE_EVENT(transaction_restart,	trans_restart_journal_res_get,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip)
-);
-
-DEFINE_EVENT(transaction_restart,	trans_restart_journal_preres_get,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip)
-);
-
-DEFINE_EVENT(transaction_restart,	trans_restart_journal_reclaim,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip)
-);
-
-DEFINE_EVENT(transaction_restart,	trans_restart_mark_replicas,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip)
-);
-
-DEFINE_EVENT(transaction_restart,	trans_restart_fault_inject,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip)
-);
-
-DEFINE_EVENT(transaction_restart,	trans_restart_btree_node_split,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip)
-);
-
-DEFINE_EVENT(transaction_restart,	trans_restart_mark,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip)
-);
-
-DEFINE_EVENT(transaction_restart,	trans_restart_upgrade,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip)
-);
-
-DEFINE_EVENT(transaction_restart,	trans_restart_iter_upgrade,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip)
-);
-
-DEFINE_EVENT(transaction_restart,	trans_restart_relock,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip)
-);
-
-DEFINE_EVENT(transaction_restart,	trans_restart_traverse,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip)
-);
-
-DEFINE_EVENT(transaction_restart,	trans_traverse_all,
-	TP_PROTO(unsigned long ip),
-	TP_ARGS(ip)
-);
-
 DECLARE_EVENT_CLASS(node_lock_fail,
-	TP_PROTO(unsigned level, u32 iter_seq, unsigned node, u32 node_seq),
-	TP_ARGS(level, iter_seq, node, node_seq),
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip,
+		 enum btree_id btree_id,
+		 struct bpos *pos,
+		 unsigned level, u32 iter_seq, unsigned node, u32 node_seq),
+	TP_ARGS(trans_ip, caller_ip, btree_id, pos,
+		level, iter_seq, node, node_seq),
 
 	TP_STRUCT__entry(
-		__field(u32,		level)
-		__field(u32,		iter_seq)
-		__field(u32,		node)
-		__field(u32,		node_seq)
+		__field(unsigned long,		trans_ip	)
+		__field(unsigned long,		caller_ip	)
+		__field(u8,			btree_id	)
+		__field(u64,			pos_inode	)
+		__field(u64,			pos_offset	)
+		__field(u32,			pos_snapshot	)
+		__field(u32,			level		)
+		__field(u32,			iter_seq	)
+		__field(u32,			node		)
+		__field(u32,			node_seq	)
 	),
 
 	TP_fast_assign(
-		__entry->level		= level;
-		__entry->iter_seq	= iter_seq;
-		__entry->node		= node;
-		__entry->node_seq	= node_seq;
+		__entry->trans_ip		= trans_ip;
+		__entry->caller_ip		= caller_ip;
+		__entry->btree_id		= btree_id;
+		__entry->pos_inode		= pos->inode;
+		__entry->pos_offset		= pos->offset;
+		__entry->pos_snapshot		= pos->snapshot;
+		__entry->level			= level;
+		__entry->iter_seq		= iter_seq;
+		__entry->node			= node;
+		__entry->node_seq		= node_seq;
 	),
 
-	TP_printk("level %u iter seq %u node %u node seq %u",
+	TP_printk("%ps %pS btree %u pos %llu:%llu:%u level %u iter seq %u node %u node seq %u",
+		  (void *) __entry->trans_ip,
+		  (void *) __entry->caller_ip,
+		  __entry->btree_id,
+		  __entry->pos_inode,
+		  __entry->pos_offset,
+		  __entry->pos_snapshot,
 		  __entry->level, __entry->iter_seq,
 		  __entry->node, __entry->node_seq)
 );
 
 DEFINE_EVENT(node_lock_fail, node_upgrade_fail,
-	TP_PROTO(unsigned level, u32 iter_seq, unsigned node, u32 node_seq),
-	TP_ARGS(level, iter_seq, node, node_seq)
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip,
+		 enum btree_id btree_id,
+		 struct bpos *pos,
+		 unsigned level, u32 iter_seq, unsigned node, u32 node_seq),
+	TP_ARGS(trans_ip, caller_ip, btree_id, pos,
+		level, iter_seq, node, node_seq)
 );
 
 DEFINE_EVENT(node_lock_fail, node_relock_fail,
-	TP_PROTO(unsigned level, u32 iter_seq, unsigned node, u32 node_seq),
-	TP_ARGS(level, iter_seq, node, node_seq)
+	TP_PROTO(unsigned long trans_ip,
+		 unsigned long caller_ip,
+		 enum btree_id btree_id,
+		 struct bpos *pos,
+		 unsigned level, u32 iter_seq, unsigned node, u32 node_seq),
+	TP_ARGS(trans_ip, caller_ip, btree_id, pos,
+		level, iter_seq, node, node_seq)
 );
 
 #endif /* _TRACE_BCACHE_H */