X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Fbtree_iter.h;h=4ec873aa1f58923bd4ff0edc18799e0e6e2e3443;hb=99caca2c70f312c4a2504a7e7a9c92a91426d885;hp=c05b2dac1726d157cfdcb74eb8779ca02c8c36c4;hpb=133dfeb64843af65a018be770eb005363ca59bd9;p=bcachefs-tools-debian

diff --git a/libbcachefs/btree_iter.h b/libbcachefs/btree_iter.h
index c05b2da..4ec873a 100644
--- a/libbcachefs/btree_iter.h
+++ b/libbcachefs/btree_iter.h
@@ -1,327 +1,563 @@
+/* SPDX-License-Identifier: GPL-2.0 */
 #ifndef _BCACHEFS_BTREE_ITER_H
 #define _BCACHEFS_BTREE_ITER_H
 
-#include <linux/dynamic_fault.h>
-
 #include "bset.h"
 #include "btree_types.h"
 
-static inline void btree_iter_set_dirty(struct btree_iter *iter,
-					enum btree_iter_uptodate u)
+#include <trace/events/bcachefs.h>
+
+static inline void __btree_path_get(struct btree_path *path, bool intent)
+{
+	path->ref++;
+	path->intent_ref += intent;
+}
+
+static inline bool __btree_path_put(struct btree_path *path, bool intent)
+{
+	EBUG_ON(!path->ref);
+	EBUG_ON(!path->intent_ref && intent);
+	path->intent_ref -= intent;
+	return --path->ref == 0;
+}
+
+static inline void btree_path_set_dirty(struct btree_path *path,
+					enum btree_path_uptodate u)
 {
-	iter->uptodate = max_t(unsigned, iter->uptodate, u);
+	path->uptodate = max_t(unsigned, path->uptodate, u);
 }
 
-static inline struct btree *btree_iter_node(struct btree_iter *iter,
+static inline struct btree *btree_path_node(struct btree_path *path,
 					    unsigned level)
 {
-	return level < BTREE_MAX_DEPTH ? iter->l[level].b : NULL;
+	return level < BTREE_MAX_DEPTH ? path->l[level].b : NULL;
 }
 
-static inline struct btree *btree_node_parent(struct btree_iter *iter,
-					      struct btree *b)
+static inline bool btree_node_lock_seq_matches(const struct btree_path *path,
+					const struct btree *b, unsigned level)
 {
-	return btree_iter_node(iter, b->level + 1);
+	/*
+	 * We don't compare the low bits of the lock sequence numbers because
+	 * @path might have taken a write lock on @b, and we don't want to skip
+	 * the linked path if the sequence numbers were equal before taking that
+	 * write lock. The lock sequence number is incremented by taking and
+	 * releasing write locks and is even when unlocked:
+	 */
+	return path->l[level].lock_seq >> 1 == b->c.lock.state.seq >> 1;
 }
 
-static inline bool btree_trans_has_multiple_iters(const struct btree_trans *trans)
+static inline struct btree *btree_node_parent(struct btree_path *path,
+					      struct btree *b)
 {
-	return hweight64(trans->iters_linked) > 1;
+	return btree_path_node(path, b->c.level + 1);
 }
 
-static inline int btree_iter_err(const struct btree_iter *iter)
+/* Iterate over paths within a transaction: */
+
+static inline struct btree_path *
+__trans_next_path(struct btree_trans *trans, unsigned idx)
 {
-	return iter->flags & BTREE_ITER_ERROR ? -EIO : 0;
+	u64 l;
+
+	if (idx == BTREE_ITER_MAX)
+		return NULL;
+
+	l = trans->paths_allocated >> idx;
+	if (!l)
+		return NULL;
+
+	idx += __ffs64(l);
+	EBUG_ON(idx >= BTREE_ITER_MAX);
+	EBUG_ON(trans->paths[idx].idx != idx);
+	return &trans->paths[idx];
 }
 
-/* Iterate over iters within a transaction: */
+void bch2_btree_path_check_sort(struct btree_trans *, struct btree_path *, int);
+
+#define trans_for_each_path_from(_trans, _path, _start)			\
+	for (_path = __trans_next_path((_trans), _start);		\
+	     (_path);							\
+	     _path = __trans_next_path((_trans), (_path)->idx + 1))
 
-static inline struct btree_iter *
-__trans_next_iter(struct btree_trans *trans, unsigned idx)
+#define trans_for_each_path(_trans, _path)				\
+	trans_for_each_path_from(_trans, _path, 0)
+
+static inline struct btree_path *next_btree_path(struct btree_trans *trans, struct btree_path *path)
 {
-	EBUG_ON(idx < trans->nr_iters && trans->iters[idx].idx != idx);
+	unsigned idx = path ? path->sorted_idx + 1 : 0;
+
+	EBUG_ON(idx > trans->nr_sorted);
 
-	for (; idx < trans->nr_iters; idx++)
-		if (trans->iters_linked & (1ULL << idx))
-			return &trans->iters[idx];
+	return idx < trans->nr_sorted
+		? trans->paths + trans->sorted[idx]
+		: NULL;
+}
 
-	return NULL;
+static inline struct btree_path *prev_btree_path(struct btree_trans *trans, struct btree_path *path)
+{
+	EBUG_ON(path->sorted_idx >= trans->nr_sorted);
+	return path->sorted_idx
+		? trans->paths + trans->sorted[path->sorted_idx - 1]
+		: NULL;
 }
 
-#define trans_for_each_iter(_trans, _iter)				\
-	for (_iter = __trans_next_iter((_trans), 0);			\
-	     (_iter);							\
-	     _iter = __trans_next_iter((_trans), (_iter)->idx + 1))
+#define trans_for_each_path_inorder(_trans, _path, _i)			\
+	for (_i = 0;							\
+	     ((_path) = (_trans)->paths + trans->sorted[_i]), (_i) < (_trans)->nr_sorted;\
+	     _i++)
 
-static inline bool __iter_has_node(const struct btree_iter *iter,
+static inline bool __path_has_node(const struct btree_path *path,
 				   const struct btree *b)
 {
-	/*
-	 * We don't compare the low bits of the lock sequence numbers because
-	 * @iter might have taken a write lock on @b, and we don't want to skip
-	 * the linked iterator if the sequence numbers were equal before taking
-	 * that write lock. The lock sequence number is incremented by taking
-	 * and releasing write locks and is even when unlocked:
-	 */
-
-	return iter->l[b->level].b == b &&
-		iter->l[b->level].lock_seq >> 1 == b->lock.state.seq >> 1;
+	return path->l[b->c.level].b == b &&
+		btree_node_lock_seq_matches(path, b, b->c.level);
 }
 
-static inline struct btree_iter *
-__trans_next_iter_with_node(struct btree_trans *trans, struct btree *b,
+static inline struct btree_path *
+__trans_next_path_with_node(struct btree_trans *trans, struct btree *b,
 			    unsigned idx)
 {
-	EBUG_ON(idx < trans->nr_iters && trans->iters[idx].idx != idx);
+	struct btree_path *path = __trans_next_path(trans, idx);
 
-	for (; idx < trans->nr_iters; idx++)
-		if ((trans->iters_linked & (1ULL << idx)) &&
-		    __iter_has_node(&trans->iters[idx], b))
-			return &trans->iters[idx];
+	while (path && !__path_has_node(path, b))
+		path = __trans_next_path(trans, path->idx + 1);
 
-	return NULL;
+	return path;
 }
 
-#define trans_for_each_iter_with_node(_trans, _b, _iter)		\
-	for (_iter = __trans_next_iter_with_node((_trans), (_b), 0);	\
-	     (_iter);							\
-	     _iter = __trans_next_iter_with_node((_trans), (_b),	\
-						 (_iter)->idx + 1))
+#define trans_for_each_path_with_node(_trans, _b, _path)		\
+	for (_path = __trans_next_path_with_node((_trans), (_b), 0);	\
+	     (_path);							\
+	     _path = __trans_next_path_with_node((_trans), (_b),	\
+						 (_path)->idx + 1))
+
+struct btree_path * __must_check
+bch2_btree_path_make_mut(struct btree_trans *, struct btree_path *,
+			 bool, unsigned long);
+struct btree_path * __must_check
+bch2_btree_path_set_pos(struct btree_trans *, struct btree_path *,
+			struct bpos, bool, unsigned long);
+int __must_check bch2_btree_path_traverse(struct btree_trans *,
+					  struct btree_path *, unsigned);
+struct btree_path *bch2_path_get(struct btree_trans *, enum btree_id, struct bpos,
+				 unsigned, unsigned, unsigned, unsigned long);
+inline struct bkey_s_c bch2_btree_path_peek_slot(struct btree_path *, struct bkey *);
+
+struct bkey_i *bch2_btree_journal_peek_slot(struct btree_trans *,
+					struct btree_iter *, struct bpos);
+
+inline void bch2_btree_path_level_init(struct btree_trans *,
+				       struct btree_path *, struct btree *);
 
 #ifdef CONFIG_BCACHEFS_DEBUG
-void bch2_btree_iter_verify(struct btree_iter *, struct btree *);
-void bch2_btree_trans_verify_locks(struct btree_trans *);
+void bch2_trans_verify_paths(struct btree_trans *);
+void bch2_assert_pos_locked(struct btree_trans *, enum btree_id,
+			    struct bpos, bool);
 #else
-static inline void bch2_btree_iter_verify(struct btree_iter *iter,
-					  struct btree *b) {}
-static inline void bch2_btree_trans_verify_locks(struct btree_trans *iter) {}
+static inline void bch2_trans_verify_paths(struct btree_trans *trans) {}
+static inline void bch2_assert_pos_locked(struct btree_trans *trans, enum btree_id id,
+					  struct bpos pos, bool key_cache) {}
 #endif
 
-void bch2_btree_node_iter_fix(struct btree_iter *, struct btree *,
-			      struct btree_node_iter *, struct bkey_packed *,
-			      unsigned, unsigned);
+void bch2_btree_path_fix_key_modified(struct btree_trans *trans,
+				      struct btree *, struct bkey_packed *);
+void bch2_btree_node_iter_fix(struct btree_trans *trans, struct btree_path *,
+			      struct btree *, struct btree_node_iter *,
+			      struct bkey_packed *, unsigned, unsigned);
+
+int bch2_btree_path_relock_intent(struct btree_trans *, struct btree_path *);
 
-int bch2_btree_iter_unlock(struct btree_iter *);
+void bch2_path_put(struct btree_trans *, struct btree_path *, bool);
 
-bool bch2_btree_trans_relock(struct btree_trans *);
-void bch2_btree_trans_unlock(struct btree_trans *);
+int bch2_trans_relock(struct btree_trans *);
+void bch2_trans_unlock(struct btree_trans *);
+
+static inline bool trans_was_restarted(struct btree_trans *trans, u32 restart_count)
+{
+	return restart_count != trans->restart_count;
+}
 
-bool __bch2_btree_iter_upgrade(struct btree_iter *, unsigned);
-bool __bch2_btree_iter_upgrade_nounlock(struct btree_iter *, unsigned);
+void bch2_trans_verify_not_restarted(struct btree_trans *, u32);
 
-static inline bool bch2_btree_iter_upgrade(struct btree_iter *iter,
-					   unsigned new_locks_want,
-					   bool may_drop_locks)
+__always_inline
+static inline int btree_trans_restart_nounlock(struct btree_trans *trans, int err)
 {
-	new_locks_want = min(new_locks_want, BTREE_MAX_DEPTH);
+	BUG_ON(err <= 0);
+	BUG_ON(!bch2_err_matches(err, BCH_ERR_transaction_restart));
 
-	return iter->locks_want < new_locks_want
-		? (may_drop_locks
-		   ? __bch2_btree_iter_upgrade(iter, new_locks_want)
-		   : __bch2_btree_iter_upgrade_nounlock(iter, new_locks_want))
-		: iter->uptodate <= BTREE_ITER_NEED_PEEK;
+	trans->restarted = err;
+	return -err;
 }
 
-void __bch2_btree_iter_downgrade(struct btree_iter *, unsigned);
+__always_inline
+static inline int btree_trans_restart(struct btree_trans *trans, int err)
+{
+	btree_trans_restart_nounlock(trans, err);
+	return -err;
+}
+
+bool bch2_btree_node_upgrade(struct btree_trans *,
+			     struct btree_path *, unsigned);
 
-static inline void bch2_btree_iter_downgrade(struct btree_iter *iter)
+void __bch2_btree_path_downgrade(struct btree_trans *, struct btree_path *, unsigned);
+
+static inline void bch2_btree_path_downgrade(struct btree_trans *trans,
+					     struct btree_path *path)
 {
-	if (iter->locks_want > (iter->flags & BTREE_ITER_INTENT) ? 1 : 0)
-		__bch2_btree_iter_downgrade(iter, 0);
+	unsigned new_locks_want = path->level + !!path->intent_ref;
+
+	if (path->locks_want > new_locks_want)
+		__bch2_btree_path_downgrade(trans, path, new_locks_want);
 }
 
-void bch2_btree_iter_node_replace(struct btree_iter *, struct btree *);
-void bch2_btree_iter_node_drop(struct btree_iter *, struct btree *);
+void bch2_trans_downgrade(struct btree_trans *);
 
-void bch2_btree_iter_reinit_node(struct btree_iter *, struct btree *);
+void bch2_trans_node_add(struct btree_trans *trans, struct btree *);
+void bch2_trans_node_reinit_iter(struct btree_trans *, struct btree *);
 
+int __must_check __bch2_btree_iter_traverse(struct btree_iter *iter);
 int __must_check bch2_btree_iter_traverse(struct btree_iter *);
-int bch2_btree_iter_traverse_all(struct btree_trans *);
 
 struct btree *bch2_btree_iter_peek_node(struct btree_iter *);
-struct btree *bch2_btree_iter_next_node(struct btree_iter *, unsigned);
+struct btree *bch2_btree_iter_next_node(struct btree_iter *);
 
-struct bkey_s_c bch2_btree_iter_peek(struct btree_iter *);
+struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *, struct bpos);
 struct bkey_s_c bch2_btree_iter_next(struct btree_iter *);
+
+struct bkey_s_c bch2_btree_iter_peek_all_levels(struct btree_iter *);
+
+static inline struct bkey_s_c bch2_btree_iter_peek(struct btree_iter *iter)
+{
+	return bch2_btree_iter_peek_upto(iter, SPOS_MAX);
+}
+
+struct bkey_s_c bch2_btree_iter_peek_prev(struct btree_iter *);
 struct bkey_s_c bch2_btree_iter_prev(struct btree_iter *);
 
 struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_iter *);
 struct bkey_s_c bch2_btree_iter_next_slot(struct btree_iter *);
+struct bkey_s_c bch2_btree_iter_prev_slot(struct btree_iter *);
 
-void bch2_btree_iter_set_pos_same_leaf(struct btree_iter *, struct bpos);
-void bch2_btree_iter_set_pos(struct btree_iter *, struct bpos);
+bool bch2_btree_iter_advance(struct btree_iter *);
+bool bch2_btree_iter_rewind(struct btree_iter *);
 
-static inline struct bpos btree_type_successor(enum btree_id id,
-					       struct bpos pos)
+static inline void __bch2_btree_iter_set_pos(struct btree_iter *iter, struct bpos new_pos)
 {
-	if (id == BTREE_ID_INODES) {
-		pos.inode++;
-		pos.offset = 0;
-	} else if (id != BTREE_ID_EXTENTS) {
-		pos = bkey_successor(pos);
-	}
-
-	return pos;
+	iter->k.type = KEY_TYPE_deleted;
+	iter->k.p.inode		= iter->pos.inode	= new_pos.inode;
+	iter->k.p.offset	= iter->pos.offset	= new_pos.offset;
+	iter->k.p.snapshot	= iter->pos.snapshot	= new_pos.snapshot;
+	iter->k.size = 0;
 }
 
-static inline struct bpos btree_type_predecessor(enum btree_id id,
-					       struct bpos pos)
+static inline void bch2_btree_iter_set_pos(struct btree_iter *iter, struct bpos new_pos)
 {
-	if (id == BTREE_ID_INODES) {
-		--pos.inode;
-		pos.offset = 0;
-	} else /* if (id != BTREE_ID_EXTENTS) */ {
-		pos = bkey_predecessor(pos);
-	}
+	if (unlikely(iter->update_path))
+		bch2_path_put(iter->trans, iter->update_path,
+			      iter->flags & BTREE_ITER_INTENT);
+	iter->update_path = NULL;
+
+	if (!(iter->flags & BTREE_ITER_ALL_SNAPSHOTS))
+		new_pos.snapshot = iter->snapshot;
 
-	return pos;
+	__bch2_btree_iter_set_pos(iter, new_pos);
 }
 
-static inline int __btree_iter_cmp(enum btree_id id,
-				   struct bpos pos,
-				   const struct btree_iter *r)
+static inline void bch2_btree_iter_set_pos_to_extent_start(struct btree_iter *iter)
 {
-	if (id != r->btree_id)
-		return id < r->btree_id ? -1 : 1;
-	return bkey_cmp(pos, r->pos);
+	BUG_ON(!(iter->flags & BTREE_ITER_IS_EXTENTS));
+	iter->pos = bkey_start_pos(&iter->k);
 }
 
-static inline int btree_iter_cmp(const struct btree_iter *l,
-				 const struct btree_iter *r)
+static inline void bch2_btree_iter_set_snapshot(struct btree_iter *iter, u32 snapshot)
 {
-	return __btree_iter_cmp(l->btree_id, l->pos, r);
+	struct bpos pos = iter->pos;
+
+	iter->snapshot = snapshot;
+	pos.snapshot = snapshot;
+	bch2_btree_iter_set_pos(iter, pos);
 }
 
-int bch2_trans_unlock(struct btree_trans *);
+void bch2_trans_iter_exit(struct btree_trans *, struct btree_iter *);
+void bch2_trans_iter_init(struct btree_trans *, struct btree_iter *,
+			  unsigned, struct bpos, unsigned);
+void bch2_trans_node_iter_init(struct btree_trans *, struct btree_iter *,
+			       enum btree_id, struct bpos,
+			       unsigned, unsigned, unsigned);
+void bch2_trans_copy_iter(struct btree_iter *, struct btree_iter *);
 
-/*
- * Unlocks before scheduling
- * Note: does not revalidate iterator
- */
-static inline void bch2_trans_cond_resched(struct btree_trans *trans)
+static inline void set_btree_iter_dontneed(struct btree_iter *iter)
 {
-	if (need_resched()) {
-		bch2_trans_unlock(trans);
-		schedule();
-	} else if (race_fault()) {
-		bch2_trans_unlock(trans);
-	}
+	if (!iter->trans->restarted)
+		iter->path->preserve = false;
 }
 
-#define __for_each_btree_node(_trans, _iter, _btree_id, _start,	\
-			      _locks_want, _depth, _flags, _b)		\
-	for (iter = bch2_trans_get_node_iter((_trans), (_btree_id),	\
-				_start, _locks_want, _depth, _flags),	\
-	     _b = bch2_btree_iter_peek_node(_iter);			\
-	     (_b);							\
-	     (_b) = bch2_btree_iter_next_node(_iter, _depth))
+void *__bch2_trans_kmalloc(struct btree_trans *, size_t);
 
-#define for_each_btree_node(_trans, _iter, _btree_id, _start,		\
-			    _flags, _b)					\
-	__for_each_btree_node(_trans, _iter, _btree_id, _start,		\
-			      0, 0, _flags, _b)
-
-static inline struct bkey_s_c __bch2_btree_iter_peek(struct btree_iter *iter,
-						     unsigned flags)
+static inline void *bch2_trans_kmalloc(struct btree_trans *trans, size_t size)
 {
-	return flags & BTREE_ITER_SLOTS
-		? bch2_btree_iter_peek_slot(iter)
-		: bch2_btree_iter_peek(iter);
+	unsigned new_top = trans->mem_top + size;
+	void *p = trans->mem + trans->mem_top;
+
+	if (likely(new_top <= trans->mem_bytes)) {
+		trans->mem_top += size;
+		memset(p, 0, size);
+		return p;
+	} else {
+		return __bch2_trans_kmalloc(trans, size);
+
+	}
 }
 
-static inline struct bkey_s_c __bch2_btree_iter_next(struct btree_iter *iter,
-						     unsigned flags)
+u32 bch2_trans_begin(struct btree_trans *);
+
+static inline struct btree *
+__btree_iter_peek_node_and_restart(struct btree_trans *trans, struct btree_iter *iter)
 {
-	bch2_trans_cond_resched(iter->trans);
+	struct btree *b;
+
+	while (b = bch2_btree_iter_peek_node(iter),
+	       bch2_err_matches(PTR_ERR_OR_ZERO(b), BCH_ERR_transaction_restart))
+		bch2_trans_begin(trans);
 
-	return flags & BTREE_ITER_SLOTS
-		? bch2_btree_iter_next_slot(iter)
-		: bch2_btree_iter_next(iter);
+	return b;
 }
 
-#define for_each_btree_key(_trans, _iter, _btree_id,  _start, _flags, _k)\
-	for (iter = bch2_trans_get_iter((_trans), (_btree_id),		\
-					(_start), (_flags)),		\
-	     (_k) = __bch2_btree_iter_peek(_iter, _flags);		\
-	     !IS_ERR_OR_NULL((_k).k);					\
-	     (_k) = __bch2_btree_iter_next(_iter, _flags))
+#define __for_each_btree_node(_trans, _iter, _btree_id, _start,		\
+			      _locks_want, _depth, _flags, _b, _ret)	\
+	for (bch2_trans_node_iter_init((_trans), &(_iter), (_btree_id),	\
+				_start, _locks_want, _depth, _flags);	\
+	     (_b) = __btree_iter_peek_node_and_restart((_trans), &(_iter)),\
+	     !((_ret) = PTR_ERR_OR_ZERO(_b)) && (_b);			\
+	     (_b) = bch2_btree_iter_next_node(&(_iter)))
 
-#define for_each_btree_key_continue(_iter, _flags, _k)			\
-	for ((_k) = __bch2_btree_iter_peek(_iter, _flags);		\
-	     !IS_ERR_OR_NULL((_k).k);					\
-	     (_k) = __bch2_btree_iter_next(_iter, _flags))
+#define for_each_btree_node(_trans, _iter, _btree_id, _start,		\
+			    _flags, _b, _ret)				\
+	__for_each_btree_node(_trans, _iter, _btree_id, _start,		\
+			      0, 0, _flags, _b, _ret)
 
 static inline int bkey_err(struct bkey_s_c k)
 {
 	return PTR_ERR_OR_ZERO(k.k);
 }
 
-/* new multiple iterator interface: */
-
-void bch2_trans_preload_iters(struct btree_trans *);
-int bch2_trans_iter_put(struct btree_trans *, struct btree_iter *);
-int bch2_trans_iter_free(struct btree_trans *, struct btree_iter *);
-int bch2_trans_iter_free_on_commit(struct btree_trans *, struct btree_iter *);
+static inline struct bkey_s_c bch2_btree_iter_peek_prev_type(struct btree_iter *iter,
+							     unsigned flags)
+{
+	BUG_ON(flags & BTREE_ITER_ALL_LEVELS);
 
-void bch2_trans_unlink_iters(struct btree_trans *, u64);
+	return  flags & BTREE_ITER_SLOTS      ? bch2_btree_iter_peek_slot(iter) :
+						bch2_btree_iter_peek_prev(iter);
+}
 
-struct btree_iter *__bch2_trans_get_iter(struct btree_trans *, enum btree_id,
-					 struct bpos, unsigned, u64);
-struct btree_iter *bch2_trans_copy_iter(struct btree_trans *,
-					struct btree_iter *);
+static inline struct bkey_s_c bch2_btree_iter_peek_type(struct btree_iter *iter,
+							unsigned flags)
+{
+	return  flags & BTREE_ITER_ALL_LEVELS ? bch2_btree_iter_peek_all_levels(iter) :
+		flags & BTREE_ITER_SLOTS      ? bch2_btree_iter_peek_slot(iter) :
+						bch2_btree_iter_peek(iter);
+}
 
-static __always_inline u64 __btree_iter_id(void)
+static inline struct bkey_s_c bch2_btree_iter_peek_upto_type(struct btree_iter *iter,
+							     struct bpos end,
+							     unsigned flags)
 {
-	u64 ret = 0;
+	if (!(flags & BTREE_ITER_SLOTS))
+		return bch2_btree_iter_peek_upto(iter, end);
+
+	if (bkey_cmp(iter->pos, end) > 0)
+		return bkey_s_c_null;
 
-	ret <<= 32;
-	ret |= _RET_IP_ & U32_MAX;
-	ret <<= 32;
-	ret |= _THIS_IP_ & U32_MAX;
-	return ret;
+	return bch2_btree_iter_peek_slot(iter);
 }
 
-static __always_inline struct btree_iter *
-bch2_trans_get_iter(struct btree_trans *trans, enum btree_id btree_id,
-		    struct bpos pos, unsigned flags)
+static inline int btree_trans_too_many_iters(struct btree_trans *trans)
 {
-	return __bch2_trans_get_iter(trans, btree_id, pos, flags,
-				     __btree_iter_id());
+	if (hweight64(trans->paths_allocated) > BTREE_ITER_MAX - 8) {
+		trace_and_count(trans->c, trans_restart_too_many_iters, trans, _THIS_IP_);
+		return btree_trans_restart(trans, BCH_ERR_transaction_restart_too_many_iters);
+	}
+
+	return 0;
 }
 
-struct btree_iter *bch2_trans_get_node_iter(struct btree_trans *,
-				enum btree_id, struct bpos,
-				unsigned, unsigned, unsigned);
+static inline struct bkey_s_c
+__bch2_btree_iter_peek_and_restart(struct btree_trans *trans,
+				   struct btree_iter *iter, unsigned flags)
+{
+	struct bkey_s_c k;
 
-void __bch2_trans_begin(struct btree_trans *);
+	while (btree_trans_too_many_iters(trans) ||
+	       (k = bch2_btree_iter_peek_type(iter, flags),
+		bch2_err_matches(bkey_err(k), BCH_ERR_transaction_restart)))
+		bch2_trans_begin(trans);
 
-static inline void bch2_trans_begin_updates(struct btree_trans *trans)
-{
-	trans->nr_updates = 0;
+	return k;
 }
 
-void *bch2_trans_kmalloc(struct btree_trans *, size_t);
-void bch2_trans_init(struct btree_trans *, struct bch_fs *);
-int bch2_trans_exit(struct btree_trans *);
+#define lockrestart_do(_trans, _do)					\
+({									\
+	u32 _restart_count;						\
+	int _ret;							\
+									\
+	do {								\
+		_restart_count = bch2_trans_begin(_trans);		\
+		_ret = (_do);						\
+	} while (bch2_err_matches(_ret, BCH_ERR_transaction_restart));	\
+									\
+	if (!_ret)							\
+		bch2_trans_verify_not_restarted(_trans, _restart_count);\
+									\
+	_ret;								\
+})
+
+/*
+ * nested_lockrestart_do(), nested_commit_do():
+ *
+ * These are like lockrestart_do() and commit_do(), with two differences:
+ *
+ *  - We don't call bch2_trans_begin() unless we had a transaction restart
+ *  - We return -BCH_ERR_transaction_restart_nested if we succeeded after a
+ *  transaction restart
+ */
+#define nested_lockrestart_do(_trans, _do)				\
+({									\
+	u32 _restart_count, _orig_restart_count;			\
+	int _ret;							\
+									\
+	_restart_count = _orig_restart_count = (_trans)->restart_count;	\
+									\
+	while (bch2_err_matches(_ret = (_do), BCH_ERR_transaction_restart))\
+		_restart_count = bch2_trans_begin(_trans);		\
+									\
+	if (!_ret)							\
+		bch2_trans_verify_not_restarted(_trans, _restart_count);\
+									\
+	if (!_ret && trans_was_restarted(_trans, _orig_restart_count))	\
+		_ret = -BCH_ERR_transaction_restart_nested;		\
+									\
+	_ret;								\
+})
 
-#ifdef TRACE_TRANSACTION_RESTARTS
-#define bch2_trans_begin(_trans)					\
-do {									\
-	if (is_power_of_2((_trans)->nr_restarts) &&			\
-	    (_trans)->nr_restarts >= 8)					\
-		pr_info("nr restarts: %zu", (_trans)->nr_restarts);	\
+#define for_each_btree_key2(_trans, _iter, _btree_id,			\
+			    _start, _flags, _k, _do)			\
+({									\
+	int _ret = 0;							\
 									\
-	(_trans)->nr_restarts++;					\
-	__bch2_trans_begin(_trans);					\
-} while (0)
-#else
-#define bch2_trans_begin(_trans)	__bch2_trans_begin(_trans)
-#endif
+	bch2_trans_iter_init((_trans), &(_iter), (_btree_id),		\
+			     (_start), (_flags));			\
+									\
+	while (1) {							\
+		u32 _restart_count = bch2_trans_begin(_trans);		\
+		(_k) = bch2_btree_iter_peek_type(&(_iter), (_flags));	\
+		if (!(_k).k) {						\
+			_ret = 0;					\
+			break;						\
+		}							\
+									\
+		_ret = bkey_err(_k) ?: (_do);				\
+		if (bch2_err_matches(_ret, BCH_ERR_transaction_restart))\
+			continue;					\
+		if (_ret)						\
+			break;						\
+		bch2_trans_verify_not_restarted(_trans, _restart_count);\
+		if (!bch2_btree_iter_advance(&(_iter)))			\
+			break;						\
+	}								\
+									\
+	bch2_trans_iter_exit((_trans), &(_iter));			\
+	_ret;								\
+})
+
+#define for_each_btree_key_reverse(_trans, _iter, _btree_id,		\
+				   _start, _flags, _k, _do)		\
+({									\
+	int _ret = 0;							\
+									\
+	bch2_trans_iter_init((_trans), &(_iter), (_btree_id),		\
+			     (_start), (_flags));			\
+									\
+	while (1) {							\
+		u32 _restart_count = bch2_trans_begin(_trans);		\
+		(_k) = bch2_btree_iter_peek_prev_type(&(_iter), (_flags));\
+		if (!(_k).k) {						\
+			_ret = 0;					\
+			break;						\
+		}							\
+									\
+		_ret = bkey_err(_k) ?: (_do);				\
+		if (bch2_err_matches(_ret, BCH_ERR_transaction_restart))\
+			continue;					\
+		if (_ret)						\
+			break;						\
+		bch2_trans_verify_not_restarted(_trans, _restart_count);\
+		if (!bch2_btree_iter_rewind(&(_iter)))			\
+			break;						\
+	}								\
+									\
+	bch2_trans_iter_exit((_trans), &(_iter));			\
+	_ret;								\
+})
+
+#define for_each_btree_key_commit(_trans, _iter, _btree_id,		\
+				  _start, _iter_flags, _k,		\
+				  _disk_res, _journal_seq, _commit_flags,\
+				  _do)					\
+	for_each_btree_key2(_trans, _iter, _btree_id, _start, _iter_flags, _k,\
+			    (_do) ?: bch2_trans_commit(_trans, (_disk_res),\
+					(_journal_seq), (_commit_flags)))
+
+#define for_each_btree_key(_trans, _iter, _btree_id,			\
+			   _start, _flags, _k, _ret)			\
+	for (bch2_trans_iter_init((_trans), &(_iter), (_btree_id),	\
+				  (_start), (_flags));			\
+	     (_k) = __bch2_btree_iter_peek_and_restart((_trans), &(_iter), _flags),\
+	     !((_ret) = bkey_err(_k)) && (_k).k;			\
+	     bch2_btree_iter_advance(&(_iter)))
+
+#define for_each_btree_key_norestart(_trans, _iter, _btree_id,		\
+			   _start, _flags, _k, _ret)			\
+	for (bch2_trans_iter_init((_trans), &(_iter), (_btree_id),	\
+				  (_start), (_flags));			\
+	     (_k) = bch2_btree_iter_peek_type(&(_iter), _flags),	\
+	     !((_ret) = bkey_err(_k)) && (_k).k;			\
+	     bch2_btree_iter_advance(&(_iter)))
+
+#define for_each_btree_key_upto_norestart(_trans, _iter, _btree_id,	\
+			   _start, _end, _flags, _k, _ret)		\
+	for (bch2_trans_iter_init((_trans), &(_iter), (_btree_id),	\
+				  (_start), (_flags));			\
+	     (_k) = bch2_btree_iter_peek_upto_type(&(_iter), _end, _flags),\
+	     !((_ret) = bkey_err(_k)) && (_k).k;			\
+	     bch2_btree_iter_advance(&(_iter)))
+
+#define for_each_btree_key_continue(_trans, _iter, _flags, _k, _ret)	\
+	for (;								\
+	     (_k) = __bch2_btree_iter_peek_and_restart((_trans), &(_iter), _flags),\
+	     !((_ret) = bkey_err(_k)) && (_k).k;			\
+	     bch2_btree_iter_advance(&(_iter)))
+
+#define for_each_btree_key_continue_norestart(_iter, _flags, _k, _ret)	\
+	for (;								\
+	     (_k) = bch2_btree_iter_peek_type(&(_iter), _flags),	\
+	     !((_ret) = bkey_err(_k)) && (_k).k;			\
+	     bch2_btree_iter_advance(&(_iter)))
 
-#ifdef TRACE_TRANSACTION_RESTARTS_ALL
-#define trans_restart(...) pr_info("transaction restart" __VA_ARGS__)
-#else
-#define trans_restart(...) no_printk("transaction restart" __VA_ARGS__)
-#endif
+/* new multiple iterator interface: */
+
+void bch2_trans_updates_to_text(struct printbuf *, struct btree_trans *);
+void bch2_btree_path_to_text(struct printbuf *, struct btree_path *);
+void bch2_trans_paths_to_text(struct printbuf *, struct btree_trans *);
+void bch2_dump_trans_updates(struct btree_trans *);
+void bch2_dump_trans_paths_updates(struct btree_trans *);
+void __bch2_trans_init(struct btree_trans *, struct bch_fs *, const char *);
+void bch2_trans_exit(struct btree_trans *);
+
+#define bch2_trans_init(_trans, _c, _nr_iters, _mem) __bch2_trans_init(_trans, _c, __func__)
+
+void bch2_btree_trans_to_text(struct printbuf *, struct btree_trans *);
+
+void bch2_fs_btree_iter_exit(struct bch_fs *);
+int bch2_fs_btree_iter_init(struct bch_fs *);
 
 #endif /* _BCACHEFS_BTREE_ITER_H */