From 2b3e79fea66e166622a454715ce981432ac8c6e3 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 11 Apr 2024 01:01:11 -0400
Subject: [PATCH 01/24] bcachefs: Don't use bch2_btree_node_lock_write_nofail()
 in btree split path

It turns out - btree splits happen with the rest of the transaction
still locked, to avoid unnecessary restarts, which means using nofail
doesn't work here - we can deadlock.

Fortunately, we now have the ability to return errors here.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_update_interior.c | 41 ++++++++++++++++++-----------
 1 file changed, 26 insertions(+), 15 deletions(-)

diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c
index c4a5e83a56a4..29a5bc7d8789 100644
--- a/fs/bcachefs/btree_update_interior.c
+++ b/fs/bcachefs/btree_update_interior.c
@@ -1280,23 +1280,29 @@ static void bch2_btree_set_root_inmem(struct bch_fs *c, struct btree *b)
 	bch2_recalc_btree_reserve(c);
 }
 
-static void bch2_btree_set_root(struct btree_update *as,
-				struct btree_trans *trans,
-				struct btree_path *path,
-				struct btree *b)
+static int bch2_btree_set_root(struct btree_update *as,
+			       struct btree_trans *trans,
+			       struct btree_path *path,
+			       struct btree *b,
+			       bool nofail)
 {
 	struct bch_fs *c = as->c;
-	struct btree *old;
 
 	trace_and_count(c, btree_node_set_root, trans, b);
 
-	old = btree_node_root(c, b);
+	struct btree *old = btree_node_root(c, b);
 
 	/*
 	 * Ensure no one is using the old root while we switch to the
 	 * new root:
 	 */
-	bch2_btree_node_lock_write_nofail(trans, path, &old->c);
+	if (nofail) {
+		bch2_btree_node_lock_write_nofail(trans, path, &old->c);
+	} else {
+		int ret = bch2_btree_node_lock_write(trans, path, &old->c);
+		if (ret)
+			return ret;
+	}
 
 	bch2_btree_set_root_inmem(c, b);
 
@@ -1310,6 +1316,7 @@ static void bch2_btree_set_root(struct btree_update *as,
 	 * depend on the new root would have to update the new root.
 	 */
 	bch2_btree_node_unlock_write(trans, path, old);
+	return 0;
 }
 
 /* Interior node updates: */
@@ -1652,15 +1659,16 @@ static int btree_split(struct btree_update *as, struct btree_trans *trans,
 	if (parent) {
 		/* Split a non root node */
 		ret = bch2_btree_insert_node(as, trans, path, parent, &as->parent_keys);
-		if (ret)
-			goto err;
 	} else if (n3) {
-		bch2_btree_set_root(as, trans, trans->paths + path, n3);
+		ret = bch2_btree_set_root(as, trans, trans->paths + path, n3, false);
 	} else {
 		/* Root filled up but didn't need to be split */
-		bch2_btree_set_root(as, trans, trans->paths + path, n1);
+		ret = bch2_btree_set_root(as, trans, trans->paths + path, n1, false);
 	}
 
+	if (ret)
+		goto err;
+
 	if (n3) {
 		bch2_btree_update_get_open_buckets(as, n3);
 		bch2_btree_node_write(c, n3, SIX_LOCK_intent, 0);
@@ -1863,7 +1871,9 @@ static void __btree_increase_depth(struct btree_update *as, struct btree_trans *
 	bch2_keylist_add(&as->parent_keys, &b->key);
 	btree_split_insert_keys(as, trans, path_idx, n, &as->parent_keys);
 
-	bch2_btree_set_root(as, trans, path, n);
+	int ret = bch2_btree_set_root(as, trans, path, n, true);
+	BUG_ON(ret);
+
 	bch2_btree_update_get_open_buckets(as, n);
 	bch2_btree_node_write(c, n, SIX_LOCK_intent, 0);
 	bch2_trans_node_add(trans, path, n);
@@ -2106,12 +2116,13 @@ int bch2_btree_node_rewrite(struct btree_trans *trans,
 	if (parent) {
 		bch2_keylist_add(&as->parent_keys, &n->key);
 		ret = bch2_btree_insert_node(as, trans, iter->path, parent, &as->parent_keys);
-		if (ret)
-			goto err;
 	} else {
-		bch2_btree_set_root(as, trans, btree_iter_path(trans, iter), n);
+		ret = bch2_btree_set_root(as, trans, btree_iter_path(trans, iter), n, false);
 	}
 
+	if (ret)
+		goto err;
+
 	bch2_btree_update_get_open_buckets(as, n);
 	bch2_btree_node_write(c, n, SIX_LOCK_intent, 0);
 

From 58caa786f1c02fd84919fb6db9eaecb22e8f7983 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 11 Apr 2024 17:47:42 -0400
Subject: [PATCH 02/24] bcachefs: Fix UAFs of btree_insert_entry array

The btree paths array is now dynamically resizable - and as well the
btree_insert_entries array, as it needs to be the same size.

The merge path (and interior update path) allocates new btree paths,
thus can trigger a resize; thus we need to not retain direct pointers
after invoking merge; similarly when running btree node triggers.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_trans_commit.c | 27 ++++++++++++++-------------
 1 file changed, 14 insertions(+), 13 deletions(-)

diff --git a/fs/bcachefs/btree_trans_commit.c b/fs/bcachefs/btree_trans_commit.c
index aa9da4970740..9609ad370d71 100644
--- a/fs/bcachefs/btree_trans_commit.c
+++ b/fs/bcachefs/btree_trans_commit.c
@@ -499,9 +499,8 @@ static int run_one_trans_trigger(struct btree_trans *trans, struct btree_insert_
 }
 
 static int run_btree_triggers(struct btree_trans *trans, enum btree_id btree_id,
-			      struct btree_insert_entry *btree_id_start)
+			      unsigned btree_id_start)
 {
-	struct btree_insert_entry *i;
 	bool trans_trigger_run;
 	int ret, overwrite;
 
@@ -514,13 +513,13 @@ static int run_btree_triggers(struct btree_trans *trans, enum btree_id btree_id,
 		do {
 			trans_trigger_run = false;
 
-			for (i = btree_id_start;
-			     i < trans->updates + trans->nr_updates && i->btree_id <= btree_id;
+			for (unsigned i = btree_id_start;
+			     i < trans->nr_updates && trans->updates[i].btree_id <= btree_id;
 			     i++) {
-				if (i->btree_id != btree_id)
+				if (trans->updates[i].btree_id != btree_id)
 					continue;
 
-				ret = run_one_trans_trigger(trans, i, overwrite);
+				ret = run_one_trans_trigger(trans, trans->updates + i, overwrite);
 				if (ret < 0)
 					return ret;
 				if (ret)
@@ -534,8 +533,7 @@ static int run_btree_triggers(struct btree_trans *trans, enum btree_id btree_id,
 
 static int bch2_trans_commit_run_triggers(struct btree_trans *trans)
 {
-	struct btree_insert_entry *btree_id_start = trans->updates;
-	unsigned btree_id = 0;
+	unsigned btree_id = 0, btree_id_start = 0;
 	int ret = 0;
 
 	/*
@@ -549,8 +547,8 @@ static int bch2_trans_commit_run_triggers(struct btree_trans *trans)
 		if (btree_id == BTREE_ID_alloc)
 			continue;
 
-		while (btree_id_start < trans->updates + trans->nr_updates &&
-		       btree_id_start->btree_id < btree_id)
+		while (btree_id_start < trans->nr_updates &&
+		       trans->updates[btree_id_start].btree_id < btree_id)
 			btree_id_start++;
 
 		ret = run_btree_triggers(trans, btree_id, btree_id_start);
@@ -558,11 +556,13 @@ static int bch2_trans_commit_run_triggers(struct btree_trans *trans)
 			return ret;
 	}
 
-	trans_for_each_update(trans, i) {
+	for (unsigned idx = 0; idx < trans->nr_updates; idx++) {
+		struct btree_insert_entry *i = trans->updates + idx;
+
 		if (i->btree_id > BTREE_ID_alloc)
 			break;
 		if (i->btree_id == BTREE_ID_alloc) {
-			ret = run_btree_triggers(trans, BTREE_ID_alloc, i);
+			ret = run_btree_triggers(trans, BTREE_ID_alloc, idx);
 			if (ret)
 				return ret;
 			break;
@@ -826,7 +826,8 @@ static inline int do_bch2_trans_commit(struct btree_trans *trans, unsigned flags
 	struct bch_fs *c = trans->c;
 	int ret = 0, u64s_delta = 0;
 
-	trans_for_each_update(trans, i) {
+	for (unsigned idx = 0; idx < trans->nr_updates; idx++) {
+		struct btree_insert_entry *i = trans->updates + idx;
 		if (i->cached)
 			continue;
 

From 031ad9e7dbd18c63e671fc4d98be3082189b8a63 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 11 Apr 2024 21:30:43 -0400
Subject: [PATCH 03/24] bcachefs: Check for packed bkeys that are too big

add missing validation; fixes assertion pop in bkey unpack

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/bkey.h     |  6 ++++++
 fs/bcachefs/btree_io.c | 15 ++++++++-------
 2 files changed, 14 insertions(+), 7 deletions(-)

diff --git a/fs/bcachefs/bkey.h b/fs/bcachefs/bkey.h
index cf23ff47bed8..3a45d128f608 100644
--- a/fs/bcachefs/bkey.h
+++ b/fs/bcachefs/bkey.h
@@ -314,6 +314,12 @@ static inline unsigned bkeyp_key_u64s(const struct bkey_format *format,
 	return bkey_packed(k) ? format->key_u64s : BKEY_U64s;
 }
 
+static inline bool bkeyp_u64s_valid(const struct bkey_format *f,
+				    const struct bkey_packed *k)
+{
+	return ((unsigned) k->u64s - bkeyp_key_u64s(f, k) <= U8_MAX - BKEY_U64s);
+}
+
 static inline unsigned bkeyp_key_bytes(const struct bkey_format *format,
 				       const struct bkey_packed *k)
 {
diff --git a/fs/bcachefs/btree_io.c b/fs/bcachefs/btree_io.c
index d7de82ac3893..b1aaf0550644 100644
--- a/fs/bcachefs/btree_io.c
+++ b/fs/bcachefs/btree_io.c
@@ -831,7 +831,7 @@ static int bset_key_invalid(struct bch_fs *c, struct btree *b,
 		(rw == WRITE ? bch2_bkey_val_invalid(c, k, READ, err) : 0);
 }
 
-static bool __bkey_valid(struct bch_fs *c, struct btree *b,
+static bool bkey_packed_valid(struct bch_fs *c, struct btree *b,
 			 struct bset *i, struct bkey_packed *k)
 {
 	if (bkey_p_next(k) > vstruct_last(i))
@@ -840,7 +840,7 @@ static bool __bkey_valid(struct bch_fs *c, struct btree *b,
 	if (k->format > KEY_FORMAT_CURRENT)
 		return false;
 
-	if (k->u64s < bkeyp_key_u64s(&b->format, k))
+	if (!bkeyp_u64s_valid(&b->format, k))
 		return false;
 
 	struct printbuf buf = PRINTBUF;
@@ -884,11 +884,13 @@ static int validate_bset_keys(struct bch_fs *c, struct btree *b,
 				 "invalid bkey format %u", k->format))
 			goto drop_this_key;
 
-		if (btree_err_on(k->u64s < bkeyp_key_u64s(&b->format, k),
+		if (btree_err_on(!bkeyp_u64s_valid(&b->format, k),
 				 -BCH_ERR_btree_node_read_err_fixable,
 				 c, NULL, b, i,
 				 btree_node_bkey_bad_u64s,
-				 "k->u64s too small (%u < %u)", k->u64s, bkeyp_key_u64s(&b->format, k)))
+				 "bad k->u64s %u (min %u max %lu)", k->u64s,
+				 bkeyp_key_u64s(&b->format, k),
+				 U8_MAX - BKEY_U64s + bkeyp_key_u64s(&b->format, k)))
 			goto drop_this_key;
 
 		if (!write)
@@ -947,13 +949,12 @@ drop_this_key:
 			 * do
 			 */
 
-			if (!__bkey_valid(c, b, i, (void *) ((u64 *) k + next_good_key))) {
+			if (!bkey_packed_valid(c, b, i, (void *) ((u64 *) k + next_good_key))) {
 				for (next_good_key = 1;
 				     next_good_key < (u64 *) vstruct_last(i) - (u64 *) k;
 				     next_good_key++)
-					if (__bkey_valid(c, b, i, (void *) ((u64 *) k + next_good_key)))
+					if (bkey_packed_valid(c, b, i, (void *) ((u64 *) k + next_good_key)))
 						goto got_good_key;
-
 			}
 
 			/*

From 87cb0239c87f608fd48fb50f9f53f129dcfd73f4 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 11 Apr 2024 23:38:07 -0400
Subject: [PATCH 04/24] bcachefs: btree node scan: handle encrypted nodes

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_node_scan.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/fs/bcachefs/btree_node_scan.c b/fs/bcachefs/btree_node_scan.c
index 556f76f5c84e..20f2b37c4474 100644
--- a/fs/bcachefs/btree_node_scan.c
+++ b/fs/bcachefs/btree_node_scan.c
@@ -133,9 +133,19 @@ static void try_read_btree_node(struct find_btree_nodes *f, struct bch_dev *ca,
 	if (le64_to_cpu(bn->magic) != bset_magic(c))
 		return;
 
+	if (bch2_csum_type_is_encryption(BSET_CSUM_TYPE(&bn->keys))) {
+		struct nonce nonce = btree_nonce(&bn->keys, 0);
+		unsigned bytes = (void *) &bn->keys - (void *) &bn->flags;
+
+		bch2_encrypt(c, BSET_CSUM_TYPE(&bn->keys), nonce, &bn->flags, bytes);
+	}
+
 	if (btree_id_is_alloc(BTREE_NODE_ID(bn)))
 		return;
 
+	if (BTREE_NODE_LEVEL(bn) >= BTREE_MAX_DEPTH)
+		return;
+
 	rcu_read_lock();
 	struct found_btree_node n = {
 		.btree_id	= BTREE_NODE_ID(bn),

From dc32c118ec6b1032693c489a0aa9e011f0acdb1a Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 11 Apr 2024 21:20:27 -0400
Subject: [PATCH 05/24] bcachefs: fix unsafety in bch2_extent_ptr_to_text()

Need to check if we have a valid bucket before checking if ptr is stale

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/extents.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/extents.c b/fs/bcachefs/extents.c
index 0e3ca99fbd2d..36d12d2adb81 100644
--- a/fs/bcachefs/extents.c
+++ b/fs/bcachefs/extents.c
@@ -998,7 +998,9 @@ void bch2_extent_ptr_to_text(struct printbuf *out, struct bch_fs *c, const struc
 			prt_str(out, " cached");
 		if (ptr->unwritten)
 			prt_str(out, " unwritten");
-		if (ca && ptr_stale(ca, ptr))
+		if (b >= ca->mi.first_bucket &&
+		    b <  ca->mi.nbuckets &&
+		    ptr_stale(ca, ptr))
 			prt_printf(out, " stale");
 	}
 }

From 2aeed876d7c2c2fb4c6b92f59bdf7e73cfe5e098 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 11 Apr 2024 23:37:24 -0400
Subject: [PATCH 06/24] bcachefs: fix unsafety in bch2_stripe_to_text()

.to_text() functions need to work on key values that didn't pass .valid

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/ec.c | 42 +++++++++++++++++++++++-------------------
 fs/bcachefs/ec.h |  2 ++
 2 files changed, 25 insertions(+), 19 deletions(-)

diff --git a/fs/bcachefs/ec.c b/fs/bcachefs/ec.c
index 082075244e16..c0b2b2f180e0 100644
--- a/fs/bcachefs/ec.c
+++ b/fs/bcachefs/ec.c
@@ -131,29 +131,33 @@ fsck_err:
 void bch2_stripe_to_text(struct printbuf *out, struct bch_fs *c,
 			 struct bkey_s_c k)
 {
-	const struct bch_stripe *s = bkey_s_c_to_stripe(k).v;
-	unsigned i, nr_data = s->nr_blocks - s->nr_redundant;
+	const struct bch_stripe *sp = bkey_s_c_to_stripe(k).v;
+	struct bch_stripe s = {};
+
+	memcpy(&s, sp, min(sizeof(s), bkey_val_bytes(k.k)));
+
+	unsigned nr_data = s.nr_blocks - s.nr_redundant;
 
 	prt_printf(out, "algo %u sectors %u blocks %u:%u csum %u gran %u",
-	       s->algorithm,
-	       le16_to_cpu(s->sectors),
-	       nr_data,
-	       s->nr_redundant,
-	       s->csum_type,
-	       1U << s->csum_granularity_bits);
+		   s.algorithm,
+		   le16_to_cpu(s.sectors),
+		   nr_data,
+		   s.nr_redundant,
+		   s.csum_type,
+		   1U << s.csum_granularity_bits);
 
-	for (i = 0; i < s->nr_blocks; i++) {
-		const struct bch_extent_ptr *ptr = s->ptrs + i;
-		struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
-		u32 offset;
-		u64 b = sector_to_bucket_and_offset(ca, ptr->offset, &offset);
+	for (unsigned i = 0; i < s.nr_blocks; i++) {
+		const struct bch_extent_ptr *ptr = sp->ptrs + i;
 
-		prt_printf(out, " %u:%llu:%u", ptr->dev, b, offset);
-		if (i < nr_data)
-			prt_printf(out, "#%u", stripe_blockcount_get(s, i));
-		prt_printf(out, " gen %u", ptr->gen);
-		if (ptr_stale(ca, ptr))
-			prt_printf(out, " stale");
+		if ((void *) ptr >= bkey_val_end(k))
+			break;
+
+		bch2_extent_ptr_to_text(out, c, ptr);
+
+		if (s.csum_type < BCH_CSUM_NR &&
+		    i < nr_data &&
+		    stripe_blockcount_offset(&s, i) < bkey_val_bytes(k.k))
+			prt_printf(out,  "#%u", stripe_blockcount_get(sp, i));
 	}
 }
 
diff --git a/fs/bcachefs/ec.h b/fs/bcachefs/ec.h
index f4369b02e805..f042616888b0 100644
--- a/fs/bcachefs/ec.h
+++ b/fs/bcachefs/ec.h
@@ -32,6 +32,8 @@ static inline unsigned stripe_csums_per_device(const struct bch_stripe *s)
 static inline unsigned stripe_csum_offset(const struct bch_stripe *s,
 					  unsigned dev, unsigned csum_idx)
 {
+	EBUG_ON(s->csum_type >= BCH_CSUM_NR);
+
 	unsigned csum_bytes = bch_crc_bytes[s->csum_type];
 
 	return sizeof(struct bch_stripe) +

From 7b4c4ccf848b359762c1525bab5a12f008f14a65 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 11 Apr 2024 23:58:36 -0400
Subject: [PATCH 07/24] bcachefs: fix race in bch2_btree_node_evict()

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_cache.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/btree_cache.c b/fs/bcachefs/btree_cache.c
index 84474324dba9..c7f156320a35 100644
--- a/fs/bcachefs/btree_cache.c
+++ b/fs/bcachefs/btree_cache.c
@@ -1148,6 +1148,8 @@ wait_on_io:
 
 	btree_node_lock_nopath_nofail(trans, &b->c, SIX_LOCK_intent);
 	btree_node_lock_nopath_nofail(trans, &b->c, SIX_LOCK_write);
+	if (unlikely(b->hash_val != btree_ptr_hash_val(k)))
+		goto out;
 
 	if (btree_node_dirty(b)) {
 		__bch2_btree_node_write(c, b, BTREE_WRITE_cache_reclaim);
@@ -1162,7 +1164,7 @@ wait_on_io:
 	btree_node_data_free(c, b);
 	bch2_btree_node_hash_remove(bc, b);
 	mutex_unlock(&bc->lock);
-
+out:
 	six_unlock_write(&b->c.lock);
 	six_unlock_intent(&b->c.lock);
 }

From ba8ed36e72033dd6b89d44145f323e6c4508bfc9 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Fri, 12 Apr 2024 00:09:08 -0400
Subject: [PATCH 08/24] bcachefs: don't queue btree nodes for rewrites during
 scan

many nodes found during scan will be old nodes, overwritten by newer
nodes

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_io.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/btree_io.c b/fs/bcachefs/btree_io.c
index b1aaf0550644..9678b2375bed 100644
--- a/fs/bcachefs/btree_io.c
+++ b/fs/bcachefs/btree_io.c
@@ -1340,7 +1340,9 @@ start:
 			       rb->start_time);
 	bio_put(&rb->bio);
 
-	if (saw_error && !btree_node_read_error(b)) {
+	if (saw_error &&
+	    !btree_node_read_error(b) &&
+	    c->curr_recovery_pass != BCH_RECOVERY_PASS_scan_for_btree_nodes) {
 		printbuf_reset(&buf);
 		bch2_bpos_to_text(&buf, b->key.k.p);
 		bch_err_ratelimited(c, "%s: rewriting btree node at btree=%s level=%u %s due to error",

From 9abb6dd7ce5a261f7aebf6f396b50a63db71133f Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Fri, 12 Apr 2024 15:17:00 -0400
Subject: [PATCH 09/24] bcachefs: Standardize helpers for printing enum strs
 with bounds checks

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/bcachefs_format.h |  4 ++--
 fs/bcachefs/buckets.h         |  8 --------
 fs/bcachefs/checksum.c        | 23 ++++++++++++++---------
 fs/bcachefs/checksum.h        |  5 +++--
 fs/bcachefs/compress.h        |  8 --------
 fs/bcachefs/ec.c              | 14 ++++++--------
 fs/bcachefs/extents.c         |  7 ++++---
 fs/bcachefs/journal_io.c      | 17 +++++++++--------
 fs/bcachefs/opts.c            | 29 +++++++++++++++++++++++++----
 fs/bcachefs/opts.h            | 10 ++++++----
 10 files changed, 69 insertions(+), 56 deletions(-)

diff --git a/fs/bcachefs/bcachefs_format.h b/fs/bcachefs/bcachefs_format.h
index 364ae42022af..9480f1b44f07 100644
--- a/fs/bcachefs/bcachefs_format.h
+++ b/fs/bcachefs/bcachefs_format.h
@@ -1314,7 +1314,7 @@ static inline __u64 __bset_magic(struct bch_sb *sb)
 	x(write_buffer_keys,	11)		\
 	x(datetime,		12)
 
-enum {
+enum bch_jset_entry_type {
 #define x(f, nr)	BCH_JSET_ENTRY_##f	= nr,
 	BCH_JSET_ENTRY_TYPES()
 #undef x
@@ -1360,7 +1360,7 @@ struct jset_entry_blacklist_v2 {
 	x(inodes,		1)		\
 	x(key_version,		2)
 
-enum {
+enum bch_fs_usage_type {
 #define x(f, nr)	BCH_FS_USAGE_##f	= nr,
 	BCH_FS_USAGE_TYPES()
 #undef x
diff --git a/fs/bcachefs/buckets.h b/fs/bcachefs/buckets.h
index 00aaf4bb5139..f9af5adabe83 100644
--- a/fs/bcachefs/buckets.h
+++ b/fs/bcachefs/buckets.h
@@ -395,14 +395,6 @@ static inline const char *bch2_data_type_str(enum bch_data_type type)
 		: "(invalid data type)";
 }
 
-static inline void bch2_prt_data_type(struct printbuf *out, enum bch_data_type type)
-{
-	if (type < BCH_DATA_NR)
-		prt_str(out, __bch2_data_types[type]);
-	else
-		prt_printf(out, "(invalid data type %u)", type);
-}
-
 /* disk reservations: */
 
 static inline void bch2_disk_reservation_put(struct bch_fs *c,
diff --git a/fs/bcachefs/checksum.c b/fs/bcachefs/checksum.c
index 4701457f6381..7ed779b411f6 100644
--- a/fs/bcachefs/checksum.c
+++ b/fs/bcachefs/checksum.c
@@ -429,15 +429,20 @@ int bch2_rechecksum_bio(struct bch_fs *c, struct bio *bio,
 				extent_nonce(version, crc_old), bio);
 
 	if (bch2_crc_cmp(merged, crc_old.csum) && !c->opts.no_data_io) {
-		bch_err(c, "checksum error in %s() (memory corruption or bug?)\n"
-			"expected %0llx:%0llx got %0llx:%0llx (old type %s new type %s)",
-			__func__,
-			crc_old.csum.hi,
-			crc_old.csum.lo,
-			merged.hi,
-			merged.lo,
-			bch2_csum_types[crc_old.csum_type],
-			bch2_csum_types[new_csum_type]);
+		struct printbuf buf = PRINTBUF;
+		prt_printf(&buf, "checksum error in %s() (memory corruption or bug?)\n"
+			   "expected %0llx:%0llx got %0llx:%0llx (old type ",
+			   __func__,
+			   crc_old.csum.hi,
+			   crc_old.csum.lo,
+			   merged.hi,
+			   merged.lo);
+		bch2_prt_csum_type(&buf, crc_old.csum_type);
+		prt_str(&buf, " new type ");
+		bch2_prt_csum_type(&buf, new_csum_type);
+		prt_str(&buf, ")");
+		bch_err(c, "%s", buf.buf);
+		printbuf_exit(&buf);
 		return -EIO;
 	}
 
diff --git a/fs/bcachefs/checksum.h b/fs/bcachefs/checksum.h
index 1b8c2c1016dc..e40499fde9a4 100644
--- a/fs/bcachefs/checksum.h
+++ b/fs/bcachefs/checksum.h
@@ -61,11 +61,12 @@ static inline void bch2_csum_err_msg(struct printbuf *out,
 				     struct bch_csum expected,
 				     struct bch_csum got)
 {
-	prt_printf(out, "checksum error: got ");
+	prt_str(out, "checksum error, type ");
+	bch2_prt_csum_type(out, type);
+	prt_str(out, ": got ");
 	bch2_csum_to_text(out, type, got);
 	prt_str(out, " should be ");
 	bch2_csum_to_text(out, type, expected);
-	prt_printf(out, " type %s", bch2_csum_types[type]);
 }
 
 int bch2_chacha_encrypt_key(struct bch_key *, struct nonce, void *, size_t);
diff --git a/fs/bcachefs/compress.h b/fs/bcachefs/compress.h
index 58c2eb45570f..607fd5e232c9 100644
--- a/fs/bcachefs/compress.h
+++ b/fs/bcachefs/compress.h
@@ -47,14 +47,6 @@ static inline enum bch_compression_type bch2_compression_opt_to_type(unsigned v)
 	return __bch2_compression_opt_to_type[bch2_compression_decode(v).type];
 }
 
-static inline void bch2_prt_compression_type(struct printbuf *out, enum bch_compression_type type)
-{
-	if (type < BCH_COMPRESSION_TYPE_NR)
-		prt_str(out, __bch2_compression_types[type]);
-	else
-		prt_printf(out, "(invalid compression type %u)", type);
-}
-
 int bch2_bio_uncompress_inplace(struct bch_fs *, struct bio *,
 				struct bch_extent_crc_unpacked *);
 int bch2_bio_uncompress(struct bch_fs *, struct bio *, struct bio *,
diff --git a/fs/bcachefs/ec.c b/fs/bcachefs/ec.c
index c0b2b2f180e0..556a217108d3 100644
--- a/fs/bcachefs/ec.c
+++ b/fs/bcachefs/ec.c
@@ -138,13 +138,13 @@ void bch2_stripe_to_text(struct printbuf *out, struct bch_fs *c,
 
 	unsigned nr_data = s.nr_blocks - s.nr_redundant;
 
-	prt_printf(out, "algo %u sectors %u blocks %u:%u csum %u gran %u",
+	prt_printf(out, "algo %u sectors %u blocks %u:%u csum ",
 		   s.algorithm,
 		   le16_to_cpu(s.sectors),
 		   nr_data,
-		   s.nr_redundant,
-		   s.csum_type,
-		   1U << s.csum_granularity_bits);
+		   s.nr_redundant);
+	bch2_prt_csum_type(out, s.csum_type);
+	prt_printf(out, " gran %u", 1U << s.csum_granularity_bits);
 
 	for (unsigned i = 0; i < s.nr_blocks; i++) {
 		const struct bch_extent_ptr *ptr = sp->ptrs + i;
@@ -611,10 +611,8 @@ static void ec_validate_checksums(struct bch_fs *c, struct ec_stripe_buf *buf)
 				struct printbuf err = PRINTBUF;
 				struct bch_dev *ca = bch_dev_bkey_exists(c, v->ptrs[i].dev);
 
-				prt_printf(&err, "stripe checksum error: expected %0llx:%0llx got %0llx:%0llx (type %s)\n",
-					   want.hi, want.lo,
-					   got.hi, got.lo,
-					   bch2_csum_types[v->csum_type]);
+				prt_str(&err, "stripe ");
+				bch2_csum_err_msg(&err, v->csum_type, want, got);
 				prt_printf(&err, "  for %ps at %u of\n  ", (void *) _RET_IP_, i);
 				bch2_bkey_val_to_text(&err, c, bkey_i_to_s_c(&buf->key));
 				bch_err_ratelimited(ca, "%s", err.buf);
diff --git a/fs/bcachefs/extents.c b/fs/bcachefs/extents.c
index 36d12d2adb81..1a331e539204 100644
--- a/fs/bcachefs/extents.c
+++ b/fs/bcachefs/extents.c
@@ -1030,11 +1030,12 @@ void bch2_bkey_ptrs_to_text(struct printbuf *out, struct bch_fs *c,
 			struct bch_extent_crc_unpacked crc =
 				bch2_extent_crc_unpack(k.k, entry_to_crc(entry));
 
-			prt_printf(out, "crc: c_size %u size %u offset %u nonce %u csum %s compress ",
+			prt_printf(out, "crc: c_size %u size %u offset %u nonce %u csum ",
 			       crc.compressed_size,
 			       crc.uncompressed_size,
-			       crc.offset, crc.nonce,
-			       bch2_csum_types[crc.csum_type]);
+			       crc.offset, crc.nonce);
+			bch2_prt_csum_type(out, crc.csum_type);
+			prt_str(out, " compress ");
 			bch2_prt_compression_type(out, crc.compression_type);
 			break;
 		}
diff --git a/fs/bcachefs/journal_io.c b/fs/bcachefs/journal_io.c
index 725fcf46f631..9aa28b52ab92 100644
--- a/fs/bcachefs/journal_io.c
+++ b/fs/bcachefs/journal_io.c
@@ -247,7 +247,7 @@ static void journal_entry_err_msg(struct printbuf *out,
 
 	if (entry) {
 		prt_str(out, " type=");
-		prt_str(out, bch2_jset_entry_types[entry->type]);
+		bch2_prt_jset_entry_type(out, entry->type);
 	}
 
 	if (!jset) {
@@ -403,7 +403,8 @@ static void journal_entry_btree_keys_to_text(struct printbuf *out, struct bch_fs
 	jset_entry_for_each_key(entry, k) {
 		if (!first) {
 			prt_newline(out);
-			prt_printf(out, "%s: ", bch2_jset_entry_types[entry->type]);
+			bch2_prt_jset_entry_type(out, entry->type);
+			prt_str(out, ": ");
 		}
 		prt_printf(out, "btree=%s l=%u ", bch2_btree_id_str(entry->btree_id), entry->level);
 		bch2_bkey_val_to_text(out, c, bkey_i_to_s_c(k));
@@ -563,9 +564,9 @@ static void journal_entry_usage_to_text(struct printbuf *out, struct bch_fs *c,
 	struct jset_entry_usage *u =
 		container_of(entry, struct jset_entry_usage, entry);
 
-	prt_printf(out, "type=%s v=%llu",
-	       bch2_fs_usage_types[u->entry.btree_id],
-	       le64_to_cpu(u->v));
+	prt_str(out, "type=");
+	bch2_prt_fs_usage_type(out, u->entry.btree_id);
+	prt_printf(out, " v=%llu", le64_to_cpu(u->v));
 }
 
 static int journal_entry_data_usage_validate(struct bch_fs *c,
@@ -827,11 +828,11 @@ int bch2_journal_entry_validate(struct bch_fs *c,
 void bch2_journal_entry_to_text(struct printbuf *out, struct bch_fs *c,
 				struct jset_entry *entry)
 {
+	bch2_prt_jset_entry_type(out, entry->type);
+
 	if (entry->type < BCH_JSET_ENTRY_NR) {
-		prt_printf(out, "%s: ", bch2_jset_entry_types[entry->type]);
+		prt_str(out, ": ");
 		bch2_jset_entry_ops[entry->type].to_text(out, c, entry);
-	} else {
-		prt_printf(out, "(unknown type %u)", entry->type);
 	}
 }
 
diff --git a/fs/bcachefs/opts.c b/fs/bcachefs/opts.c
index e1800c4119b5..bb068fd72465 100644
--- a/fs/bcachefs/opts.c
+++ b/fs/bcachefs/opts.c
@@ -43,7 +43,7 @@ const char * const __bch2_btree_ids[] = {
 	NULL
 };
 
-const char * const bch2_csum_types[] = {
+static const char * const __bch2_csum_types[] = {
 	BCH_CSUM_TYPES()
 	NULL
 };
@@ -53,7 +53,7 @@ const char * const bch2_csum_opts[] = {
 	NULL
 };
 
-const char * const __bch2_compression_types[] = {
+static const char * const __bch2_compression_types[] = {
 	BCH_COMPRESSION_TYPES()
 	NULL
 };
@@ -83,18 +83,39 @@ const char * const bch2_member_states[] = {
 	NULL
 };
 
-const char * const bch2_jset_entry_types[] = {
+static const char * const __bch2_jset_entry_types[] = {
 	BCH_JSET_ENTRY_TYPES()
 	NULL
 };
 
-const char * const bch2_fs_usage_types[] = {
+static const char * const __bch2_fs_usage_types[] = {
 	BCH_FS_USAGE_TYPES()
 	NULL
 };
 
 #undef x
 
+static void prt_str_opt_boundscheck(struct printbuf *out, const char * const opts[],
+				    unsigned nr, const char *type, unsigned idx)
+{
+	if (idx < nr)
+		prt_str(out, opts[idx]);
+	else
+		prt_printf(out, "(unknown %s %u)", type, idx);
+}
+
+#define PRT_STR_OPT_BOUNDSCHECKED(name, type)					\
+void bch2_prt_##name(struct printbuf *out, type t)				\
+{										\
+	prt_str_opt_boundscheck(out, __bch2_##name##s, ARRAY_SIZE(__bch2_##name##s) - 1, #name, t);\
+}
+
+PRT_STR_OPT_BOUNDSCHECKED(jset_entry_type,	enum bch_jset_entry_type);
+PRT_STR_OPT_BOUNDSCHECKED(fs_usage_type,	enum bch_fs_usage_type);
+PRT_STR_OPT_BOUNDSCHECKED(data_type,		enum bch_data_type);
+PRT_STR_OPT_BOUNDSCHECKED(csum_type,		enum bch_csum_type);
+PRT_STR_OPT_BOUNDSCHECKED(compression_type,	enum bch_compression_type);
+
 static int bch2_opt_fix_errors_parse(struct bch_fs *c, const char *val, u64 *res,
 				     struct printbuf *err)
 {
diff --git a/fs/bcachefs/opts.h b/fs/bcachefs/opts.h
index 1ac4135cca1c..84e452835a17 100644
--- a/fs/bcachefs/opts.h
+++ b/fs/bcachefs/opts.h
@@ -16,18 +16,20 @@ extern const char * const bch2_version_upgrade_opts[];
 extern const char * const bch2_sb_features[];
 extern const char * const bch2_sb_compat[];
 extern const char * const __bch2_btree_ids[];
-extern const char * const bch2_csum_types[];
 extern const char * const bch2_csum_opts[];
-extern const char * const __bch2_compression_types[];
 extern const char * const bch2_compression_opts[];
 extern const char * const bch2_str_hash_types[];
 extern const char * const bch2_str_hash_opts[];
 extern const char * const __bch2_data_types[];
 extern const char * const bch2_member_states[];
-extern const char * const bch2_jset_entry_types[];
-extern const char * const bch2_fs_usage_types[];
 extern const char * const bch2_d_types[];
 
+void bch2_prt_jset_entry_type(struct printbuf *,	enum bch_jset_entry_type);
+void bch2_prt_fs_usage_type(struct printbuf *,		enum bch_fs_usage_type);
+void bch2_prt_data_type(struct printbuf *,		enum bch_data_type);
+void bch2_prt_csum_type(struct printbuf *,		enum bch_csum_type);
+void bch2_prt_compression_type(struct printbuf *,	enum bch_compression_type);
+
 static inline const char *bch2_d_type_str(unsigned d_type)
 {
 	return (d_type < BCH_DT_MAX ? bch2_d_types[d_type] : NULL) ?: "(bad d_type)";

From 4518e80adfdbfdec1cc79c98bc73677ff44d96bd Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Fri, 12 Apr 2024 14:05:36 -0400
Subject: [PATCH 10/24] bcachefs: Go rw if running any explicit recovery passes

This fixes a bug where we fail to start when upgrading/downgrading
because we forgot we needed to go rw.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/recovery_passes.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/bcachefs/recovery_passes.c b/fs/bcachefs/recovery_passes.c
index cb501460d615..0cec0f7d9703 100644
--- a/fs/bcachefs/recovery_passes.c
+++ b/fs/bcachefs/recovery_passes.c
@@ -44,7 +44,7 @@ static int bch2_set_may_go_rw(struct bch_fs *c)
 
 	set_bit(BCH_FS_may_go_rw, &c->flags);
 
-	if (keys->nr || c->opts.fsck || !c->sb.clean)
+	if (keys->nr || c->opts.fsck || !c->sb.clean || c->recovery_passes_explicit)
 		return bch2_fs_read_write_early(c);
 	return 0;
 }

From 82cf18f23e1ae17053983e325e550194f947e1f1 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Fri, 12 Apr 2024 21:07:05 -0400
Subject: [PATCH 11/24] bcachefs: Fix deadlock in journal replay

btree_key_can_insert_cached() should be checking the watermark -
BCH_TRANS_COMMIT_journal_replay really means nonblocking mode when
watermark < reclaim, it was being used incorrectly.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_trans_commit.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/fs/bcachefs/btree_trans_commit.c b/fs/bcachefs/btree_trans_commit.c
index 9609ad370d71..bbec91e8e650 100644
--- a/fs/bcachefs/btree_trans_commit.c
+++ b/fs/bcachefs/btree_trans_commit.c
@@ -397,12 +397,13 @@ static int btree_key_can_insert_cached(struct btree_trans *trans, unsigned flags
 	struct bkey_cached *ck = (void *) path->l[0].b;
 	unsigned new_u64s;
 	struct bkey_i *new_k;
+	unsigned watermark = flags & BCH_WATERMARK_MASK;
 
 	EBUG_ON(path->level);
 
-	if (!test_bit(BKEY_CACHED_DIRTY, &ck->flags) &&
-	    bch2_btree_key_cache_must_wait(c) &&
-	    !(flags & BCH_TRANS_COMMIT_journal_reclaim))
+	if (watermark < BCH_WATERMARK_reclaim &&
+	    !test_bit(BKEY_CACHED_DIRTY, &ck->flags) &&
+	    bch2_btree_key_cache_must_wait(c))
 		return -BCH_ERR_btree_insert_need_journal_reclaim;
 
 	/*

From 9e203c43dc1cbaefb3888ee0ba885b2d20d47526 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sat, 13 Apr 2024 00:26:01 -0400
Subject: [PATCH 12/24] bcachefs: Fix missing write refs in fs fio paths

bch2_journal_flush_seq requires us to have a write ref

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/bcachefs.h     |  2 ++
 fs/bcachefs/fs-io-direct.c | 19 +++++++++++++------
 fs/bcachefs/fs-io.c        | 16 ++++++++--------
 3 files changed, 23 insertions(+), 14 deletions(-)

diff --git a/fs/bcachefs/bcachefs.h b/fs/bcachefs/bcachefs.h
index a31a5f706929..91c3c1fef233 100644
--- a/fs/bcachefs/bcachefs.h
+++ b/fs/bcachefs/bcachefs.h
@@ -709,6 +709,8 @@ struct btree_trans_buf {
 	x(stripe_delete)						\
 	x(reflink)							\
 	x(fallocate)							\
+	x(fsync)							\
+	x(dio_write)							\
 	x(discard)							\
 	x(discard_fast)							\
 	x(invalidate)							\
diff --git a/fs/bcachefs/fs-io-direct.c b/fs/bcachefs/fs-io-direct.c
index f49e6c0f0f68..b889370a5088 100644
--- a/fs/bcachefs/fs-io-direct.c
+++ b/fs/bcachefs/fs-io-direct.c
@@ -387,6 +387,8 @@ static __always_inline long bch2_dio_write_done(struct dio_write *dio)
 	ret = dio->op.error ?: ((long) dio->written << 9);
 	bio_put(&dio->op.wbio.bio);
 
+	bch2_write_ref_put(dio->op.c, BCH_WRITE_REF_dio_write);
+
 	/* inode->i_dio_count is our ref on inode and thus bch_fs */
 	inode_dio_end(&inode->v);
 
@@ -590,22 +592,25 @@ ssize_t bch2_direct_write(struct kiocb *req, struct iov_iter *iter)
 	prefetch(&inode->ei_inode);
 	prefetch((void *) &inode->ei_inode + 64);
 
+	if (!bch2_write_ref_tryget(c, BCH_WRITE_REF_dio_write))
+		return -EROFS;
+
 	inode_lock(&inode->v);
 
 	ret = generic_write_checks(req, iter);
 	if (unlikely(ret <= 0))
-		goto err;
+		goto err_put_write_ref;
 
 	ret = file_remove_privs(file);
 	if (unlikely(ret))
-		goto err;
+		goto err_put_write_ref;
 
 	ret = file_update_time(file);
 	if (unlikely(ret))
-		goto err;
+		goto err_put_write_ref;
 
 	if (unlikely((req->ki_pos|iter->count) & (block_bytes(c) - 1)))
-		goto err;
+		goto err_put_write_ref;
 
 	inode_dio_begin(&inode->v);
 	bch2_pagecache_block_get(inode);
@@ -645,7 +650,7 @@ ssize_t bch2_direct_write(struct kiocb *req, struct iov_iter *iter)
 	}
 
 	ret = bch2_dio_write_loop(dio);
-err:
+out:
 	if (locked)
 		inode_unlock(&inode->v);
 	return ret;
@@ -653,7 +658,9 @@ err_put_bio:
 	bch2_pagecache_block_put(inode);
 	bio_put(bio);
 	inode_dio_end(&inode->v);
-	goto err;
+err_put_write_ref:
+	bch2_write_ref_put(c, BCH_WRITE_REF_dio_write);
+	goto out;
 }
 
 void bch2_fs_fs_io_direct_exit(struct bch_fs *c)
diff --git a/fs/bcachefs/fs-io.c b/fs/bcachefs/fs-io.c
index 8c70123b6a0c..20b40477425f 100644
--- a/fs/bcachefs/fs-io.c
+++ b/fs/bcachefs/fs-io.c
@@ -174,18 +174,18 @@ void __bch2_i_sectors_acct(struct bch_fs *c, struct bch_inode_info *inode,
 static int bch2_flush_inode(struct bch_fs *c,
 			    struct bch_inode_info *inode)
 {
-	struct bch_inode_unpacked u;
-	int ret;
-
 	if (c->opts.journal_flush_disabled)
 		return 0;
 
-	ret = bch2_inode_find_by_inum(c, inode_inum(inode), &u);
-	if (ret)
-		return ret;
+	if (!bch2_write_ref_tryget(c, BCH_WRITE_REF_fsync))
+		return -EROFS;
 
-	return bch2_journal_flush_seq(&c->journal, u.bi_journal_seq) ?:
-		bch2_inode_flush_nocow_writes(c, inode);
+	struct bch_inode_unpacked u;
+	int ret = bch2_inode_find_by_inum(c, inode_inum(inode), &u) ?:
+		  bch2_journal_flush_seq(&c->journal, u.bi_journal_seq) ?:
+		  bch2_inode_flush_nocow_writes(c, inode);
+	bch2_write_ref_put(c, BCH_WRITE_REF_fsync);
+	return ret;
 }
 
 int bch2_fsync(struct file *file, loff_t start, loff_t end, int datasync)

From 9054ef2ea944528b7935d0ce3f540d4dc0bc37ba Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sat, 13 Apr 2024 16:13:13 -0400
Subject: [PATCH 13/24] bcachefs: Run merges at BCH_WATERMARK_btree

This fixes a deadlock where the interior update path during journal
replay ends up doing a ton of merges on the backpointers btree, and
deadlocking.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_update_interior.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c
index 29a5bc7d8789..1a7537ee4329 100644
--- a/fs/bcachefs/btree_update_interior.c
+++ b/fs/bcachefs/btree_update_interior.c
@@ -1926,6 +1926,8 @@ int __bch2_foreground_maybe_merge(struct btree_trans *trans,
 	BUG_ON(!trans->paths[path].should_be_locked);
 	BUG_ON(!btree_node_locked(&trans->paths[path], level));
 
+	flags &= ~BCH_WATERMARK_MASK;
+
 	b = trans->paths[path].l[level].b;
 
 	if ((sib == btree_prev_sib && bpos_eq(b->data->min_key, POS_MIN)) ||
@@ -2071,6 +2073,10 @@ err:
 		bch2_path_put(trans, new_path, true);
 	bch2_path_put(trans, sib_path, true);
 	bch2_trans_verify_locks(trans);
+	if (ret == -BCH_ERR_journal_reclaim_would_deadlock)
+		ret = 0;
+	if (!ret)
+		ret = bch2_trans_relock(trans);
 	return ret;
 err_free_update:
 	bch2_btree_node_free_never_used(as, trans, n);

From 3f10048973c8366e27147d72bf3fd8d0e1d929f2 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sat, 13 Apr 2024 18:39:03 -0400
Subject: [PATCH 14/24] bcachefs: Disable merges from interior update path

There's been a bug in the btree write buffer where it wasn't triggering
btree node merges - and leaving behind a bunch of nearly empty btree
nodes.

Then during journal replay, when updates to the backpointers btree
aren't using the btree write buffer (because we require synchronization
with journal replay), we end up doing those merges all at once.

Then if it's the interior update path running them, we deadlock because
those run with the highest watermark.

There's no real need for the interior update path to be doing btree node
merges; other code paths can handle that at lower watermarks.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_update_interior.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c
index 1a7537ee4329..82ac00aa0f7b 100644
--- a/fs/bcachefs/btree_update_interior.c
+++ b/fs/bcachefs/btree_update_interior.c
@@ -1926,6 +1926,16 @@ int __bch2_foreground_maybe_merge(struct btree_trans *trans,
 	BUG_ON(!trans->paths[path].should_be_locked);
 	BUG_ON(!btree_node_locked(&trans->paths[path], level));
 
+	/*
+	 * Work around a deadlock caused by the btree write buffer not doing
+	 * merges and leaving tons of merges for us to do - we really don't need
+	 * to be doing merges at all from the interior update path, and if the
+	 * interior update path is generating too many new interior updates we
+	 * deadlock:
+	 */
+	if ((flags & BCH_WATERMARK_MASK) == BCH_WATERMARK_interior_updates)
+		return 0;
+
 	flags &= ~BCH_WATERMARK_MASK;
 
 	b = trans->paths[path].l[level].b;

From 86dbf8c566417afd62087ecd3bedcf84f91ee4e4 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Tue, 26 Dec 2023 22:42:34 -0500
Subject: [PATCH 15/24] bcachefs: Fix btree node merging on write buffer btrees

The btree write buffer flush fastpath that avoids the main transaction
commit path had the unfortunate side effect of not doing btree node
merging.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_write_buffer.c | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/fs/bcachefs/btree_write_buffer.c b/fs/bcachefs/btree_write_buffer.c
index baf63e2fddb6..36a6f42aba5e 100644
--- a/fs/bcachefs/btree_write_buffer.c
+++ b/fs/bcachefs/btree_write_buffer.c
@@ -316,6 +316,16 @@ static int bch2_btree_write_buffer_flush_locked(struct btree_trans *trans)
 			    bpos_gt(k->k.k.p, path->l[0].b->key.k.p)) {
 				bch2_btree_node_unlock_write(trans, path, path->l[0].b);
 				write_locked = false;
+
+				ret = lockrestart_do(trans,
+					bch2_btree_iter_traverse(&iter) ?:
+					bch2_foreground_maybe_merge(trans, iter.path, 0,
+							BCH_WATERMARK_reclaim|
+							BCH_TRANS_COMMIT_journal_reclaim|
+							BCH_TRANS_COMMIT_no_check_rw|
+							BCH_TRANS_COMMIT_no_enospc));
+				if (ret)
+					goto err;
 			}
 		}
 
@@ -382,10 +392,10 @@ static int bch2_btree_write_buffer_flush_locked(struct btree_trans *trans)
 
 			ret = commit_do(trans, NULL, NULL,
 					BCH_WATERMARK_reclaim|
+					BCH_TRANS_COMMIT_journal_reclaim|
 					BCH_TRANS_COMMIT_no_check_rw|
 					BCH_TRANS_COMMIT_no_enospc|
-					BCH_TRANS_COMMIT_no_journal_res|
-					BCH_TRANS_COMMIT_journal_reclaim,
+					BCH_TRANS_COMMIT_no_journal_res ,
 					btree_write_buffered_insert(trans, i));
 			if (ret)
 				goto err;

From bceb86be9e970bc6d77bfd1f8e5272d9a2007c16 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sat, 13 Apr 2024 23:59:06 -0400
Subject: [PATCH 16/24] bcachefs: add missing bounds check in
 __bch2_bkey_val_invalid()

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/bkey_methods.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/bkey_methods.c b/fs/bcachefs/bkey_methods.c
index 5e52684764eb..e4481e4d5f38 100644
--- a/fs/bcachefs/bkey_methods.c
+++ b/fs/bcachefs/bkey_methods.c
@@ -175,7 +175,10 @@ int __bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
 			 !(bch2_key_types_allowed[type] & BIT_ULL(k.k->type)), c, err,
 			 bkey_invalid_type_for_btree,
 			 "invalid key type for btree %s (%s)",
-			 bch2_btree_node_type_str(type), bch2_bkey_types[k.k->type]);
+			 bch2_btree_node_type_str(type),
+			 k.k->type < KEY_TYPE_MAX
+			 ? bch2_bkey_types[k.k->type]
+			 : "(unknown)");
 
 	if (btree_node_type_is_extents(type) && !bkey_whiteout(k.k)) {
 		bkey_fsck_err_on(k.k->size == 0, c, err,

From d789e9a7d5e2799f4d5425b0b620210d2fcad529 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sat, 13 Apr 2024 23:59:28 -0400
Subject: [PATCH 17/24] bcachefs: Interior known are required to have known key
 types

For forwards compatibilyt, we allow bkeys of unknown type in leaf nodes;
we can simply ignore metadata we don't understand. Pointers to btree
nodes must always be of known types, howwever.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/bkey_methods.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/bkey_methods.c b/fs/bcachefs/bkey_methods.c
index e4481e4d5f38..db336a43fc08 100644
--- a/fs/bcachefs/bkey_methods.c
+++ b/fs/bcachefs/bkey_methods.c
@@ -171,7 +171,8 @@ int __bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
 	if (type >= BKEY_TYPE_NR)
 		return 0;
 
-	bkey_fsck_err_on((flags & BKEY_INVALID_COMMIT) &&
+	bkey_fsck_err_on((type == BKEY_TYPE_btree ||
+			  (flags & BKEY_INVALID_COMMIT)) &&
 			 !(bch2_key_types_allowed[type] & BIT_ULL(k.k->type)), c, err,
 			 bkey_invalid_type_for_btree,
 			 "invalid key type for btree %s (%s)",

From 8cf2036e7b557282667a437d409e6307d55366ab Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Fri, 12 Apr 2024 15:34:14 -0400
Subject: [PATCH 18/24] bcachefs: add safety checks in bch2_btree_node_fill()

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_cache.c | 25 ++++++++++++++++++++++++-
 1 file changed, 24 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/btree_cache.c b/fs/bcachefs/btree_cache.c
index c7f156320a35..c347d08d80bc 100644
--- a/fs/bcachefs/btree_cache.c
+++ b/fs/bcachefs/btree_cache.c
@@ -711,7 +711,30 @@ static noinline struct btree *bch2_btree_node_fill(struct btree_trans *trans,
 	struct btree *b;
 	u32 seq;
 
-	BUG_ON(level + 1 >= BTREE_MAX_DEPTH);
+	if (unlikely(level >= BTREE_MAX_DEPTH)) {
+		int ret = bch2_fs_topology_error(c, "attempting to get btree node at level %u, >= max depth %u",
+						 level, BTREE_MAX_DEPTH);
+		return ERR_PTR(ret);
+	}
+
+	if (unlikely(!bkey_is_btree_ptr(&k->k))) {
+		struct printbuf buf = PRINTBUF;
+		bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(k));
+
+		int ret = bch2_fs_topology_error(c, "attempting to get btree node with non-btree key %s", buf.buf);
+		printbuf_exit(&buf);
+		return ERR_PTR(ret);
+	}
+
+	if (unlikely(k->k.u64s > BKEY_BTREE_PTR_U64s_MAX)) {
+		struct printbuf buf = PRINTBUF;
+		bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(k));
+
+		int ret = bch2_fs_topology_error(c, "attempting to get btree node with too big key %s", buf.buf);
+		printbuf_exit(&buf);
+		return ERR_PTR(ret);
+	}
+
 	/*
 	 * Parent node must be locked, else we could read in a btree node that's
 	 * been freed:

From e879389f5777de5d94f27f8d88d7e92341afa3ef Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Fri, 12 Apr 2024 15:54:33 -0400
Subject: [PATCH 19/24] bcachefs: Fix bch2_btree_node_fill() for !path

We shouldn't be doing the unlock/relock dance when we're not using a
path - this fixes an assertion pop when called from btree node scan.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_cache.c | 46 ++++++++++++++++-----------------------
 1 file changed, 19 insertions(+), 27 deletions(-)

diff --git a/fs/bcachefs/btree_cache.c b/fs/bcachefs/btree_cache.c
index c347d08d80bc..02c70e813fac 100644
--- a/fs/bcachefs/btree_cache.c
+++ b/fs/bcachefs/btree_cache.c
@@ -709,7 +709,6 @@ static noinline struct btree *bch2_btree_node_fill(struct btree_trans *trans,
 	struct bch_fs *c = trans->c;
 	struct btree_cache *bc = &c->btree_cache;
 	struct btree *b;
-	u32 seq;
 
 	if (unlikely(level >= BTREE_MAX_DEPTH)) {
 		int ret = bch2_fs_topology_error(c, "attempting to get btree node at level %u, >= max depth %u",
@@ -775,34 +774,26 @@ static noinline struct btree *bch2_btree_node_fill(struct btree_trans *trans,
 	}
 
 	set_btree_node_read_in_flight(b);
-
 	six_unlock_write(&b->c.lock);
-	seq = six_lock_seq(&b->c.lock);
-	six_unlock_intent(&b->c.lock);
-
-	/* Unlock before doing IO: */
-	if (path && sync)
-		bch2_trans_unlock_noassert(trans);
-
-	bch2_btree_node_read(trans, b, sync);
-
-	if (!sync)
-		return NULL;
 
 	if (path) {
-		int ret = bch2_trans_relock(trans) ?:
-			bch2_btree_path_relock_intent(trans, path);
-		if (ret) {
-			BUG_ON(!trans->restarted);
-			return ERR_PTR(ret);
-		}
-	}
+		u32 seq = six_lock_seq(&b->c.lock);
 
-	if (!six_relock_type(&b->c.lock, lock_type, seq)) {
-		BUG_ON(!path);
+		/* Unlock before doing IO: */
+		six_unlock_intent(&b->c.lock);
+		bch2_trans_unlock_noassert(trans);
 
-		trace_and_count(c, trans_restart_relock_after_fill, trans, _THIS_IP_, path);
-		return ERR_PTR(btree_trans_restart(trans, BCH_ERR_transaction_restart_relock_after_fill));
+		bch2_btree_node_read(trans, b, sync);
+
+		if (!sync)
+			return NULL;
+
+		if (!six_relock_type(&b->c.lock, lock_type, seq))
+			b = NULL;
+	} else {
+		bch2_btree_node_read(trans, b, sync);
+		if (lock_type == SIX_LOCK_read)
+			six_lock_downgrade(&b->c.lock);
 	}
 
 	return b;
@@ -1135,18 +1126,19 @@ int bch2_btree_node_prefetch(struct btree_trans *trans,
 {
 	struct bch_fs *c = trans->c;
 	struct btree_cache *bc = &c->btree_cache;
-	struct btree *b;
 
 	BUG_ON(path && !btree_node_locked(path, level + 1));
 	BUG_ON(level >= BTREE_MAX_DEPTH);
 
-	b = btree_cache_find(bc, k);
+	struct btree *b = btree_cache_find(bc, k);
 	if (b)
 		return 0;
 
 	b = bch2_btree_node_fill(trans, path, k, btree_id,
 				 level, SIX_LOCK_read, false);
-	return PTR_ERR_OR_ZERO(b);
+	if (!IS_ERR_OR_NULL(b))
+		six_unlock_read(&b->c.lock);
+	return bch2_trans_relock(trans) ?: PTR_ERR_OR_ZERO(b);
 }
 
 void bch2_btree_node_evict(struct btree_trans *trans, const struct bkey_i *k)

From bdae2a7e6020e1cf864a30dab2af323575569dc7 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sat, 13 Apr 2024 22:43:11 -0400
Subject: [PATCH 20/24] bcachefs: sysfs internal/trigger_journal_flush

Add a sysfs knob for immediately flushing the entire journal.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/sysfs.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/sysfs.c b/fs/bcachefs/sysfs.c
index b18b0cc81b59..5be92fe3f4ea 100644
--- a/fs/bcachefs/sysfs.c
+++ b/fs/bcachefs/sysfs.c
@@ -25,6 +25,7 @@
 #include "ec.h"
 #include "inode.h"
 #include "journal.h"
+#include "journal_reclaim.h"
 #include "keylist.h"
 #include "move.h"
 #include "movinggc.h"
@@ -138,6 +139,7 @@ do {									\
 write_attribute(trigger_gc);
 write_attribute(trigger_discards);
 write_attribute(trigger_invalidates);
+write_attribute(trigger_journal_flush);
 write_attribute(prune_cache);
 write_attribute(btree_wakeup);
 rw_attribute(btree_gc_periodic);
@@ -500,7 +502,7 @@ STORE(bch2_fs)
 
 	/* Debugging: */
 
-	if (!test_bit(BCH_FS_rw, &c->flags))
+	if (!bch2_write_ref_tryget(c, BCH_WRITE_REF_sysfs))
 		return -EROFS;
 
 	if (attr == &sysfs_prune_cache) {
@@ -533,6 +535,11 @@ STORE(bch2_fs)
 	if (attr == &sysfs_trigger_invalidates)
 		bch2_do_invalidates(c);
 
+	if (attr == &sysfs_trigger_journal_flush) {
+		bch2_journal_flush_all_pins(&c->journal);
+		bch2_journal_meta(&c->journal);
+	}
+
 #ifdef CONFIG_BCACHEFS_TESTS
 	if (attr == &sysfs_perf_test) {
 		char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
@@ -553,6 +560,7 @@ STORE(bch2_fs)
 			size = ret;
 	}
 #endif
+	bch2_write_ref_put(c, BCH_WRITE_REF_sysfs);
 	return size;
 }
 SYSFS_OPS(bch2_fs);
@@ -651,6 +659,7 @@ struct attribute *bch2_fs_internal_files[] = {
 	&sysfs_trigger_gc,
 	&sysfs_trigger_discards,
 	&sysfs_trigger_invalidates,
+	&sysfs_trigger_journal_flush,
 	&sysfs_prune_cache,
 	&sysfs_btree_wakeup,
 

From 27c15ed297cb71c2e7a839439b5a097081a32605 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Fri, 12 Apr 2024 18:45:47 -0400
Subject: [PATCH 21/24] bcachefs: bch_member.btree_allocated_bitmap

This adds a small (64 bit) per-device bitmap that tracks ranges that
have btree nodes, for accelerating btree node scan if it is ever needed.

- New helpers, bch2_dev_btree_bitmap_marked() and
  bch2_dev_bitmap_mark(), for checking and updating the bitmap

- Interior btree update path updates the bitmaps when required

- The check_allocations pass has a new fsck_err check,
  btree_bitmap_not_marked

- New on disk format version, mi_btree_mitmap, which indicates the new
  bitmap is present

- Upgrade table lists the required recovery pass and expected fsck error

- Btree node scan uses the bitmap to skip ranges if we're on the new
  version

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/bcachefs_format.h       |  7 ++--
 fs/bcachefs/btree_gc.c              | 13 +++++++
 fs/bcachefs/btree_node_scan.c       |  9 +++--
 fs/bcachefs/btree_update_interior.c | 24 +++++++++++++
 fs/bcachefs/sb-downgrade.c          |  5 ++-
 fs/bcachefs/sb-errors_types.h       |  3 +-
 fs/bcachefs/sb-members.c            | 53 +++++++++++++++++++++++++++++
 fs/bcachefs/sb-members.h            | 21 ++++++++++++
 fs/bcachefs/super_types.h           |  2 ++
 9 files changed, 131 insertions(+), 6 deletions(-)

diff --git a/fs/bcachefs/bcachefs_format.h b/fs/bcachefs/bcachefs_format.h
index 9480f1b44f07..085987435a5e 100644
--- a/fs/bcachefs/bcachefs_format.h
+++ b/fs/bcachefs/bcachefs_format.h
@@ -578,7 +578,8 @@ struct bch_member {
 	__le64			nbuckets;	/* device size */
 	__le16			first_bucket;   /* index of first bucket used */
 	__le16			bucket_size;	/* sectors */
-	__le32			pad;
+	__u8			btree_bitmap_shift;
+	__u8			pad[3];
 	__le64			last_mount;	/* time_t */
 
 	__le64			flags;
@@ -587,6 +588,7 @@ struct bch_member {
 	__le64			errors_at_reset[BCH_MEMBER_ERROR_NR];
 	__le64			errors_reset_time;
 	__le64			seq;
+	__le64			btree_allocated_bitmap;
 };
 
 #define BCH_MEMBER_V1_BYTES	56
@@ -876,7 +878,8 @@ struct bch_sb_field_downgrade {
 	x(rebalance_work,		BCH_VERSION(1,  3))		\
 	x(member_seq,			BCH_VERSION(1,  4))		\
 	x(subvolume_fs_parent,		BCH_VERSION(1,  5))		\
-	x(btree_subvolume_children,	BCH_VERSION(1,  6))
+	x(btree_subvolume_children,	BCH_VERSION(1,  6))		\
+	x(mi_btree_bitmap,		BCH_VERSION(1,  7))
 
 enum bcachefs_metadata_version {
 	bcachefs_metadata_version_min = 9,
diff --git a/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c
index d2555da55c6d..ecbd9598f69f 100644
--- a/fs/bcachefs/btree_gc.c
+++ b/fs/bcachefs/btree_gc.c
@@ -828,6 +828,7 @@ static int bch2_gc_mark_key(struct btree_trans *trans, enum btree_id btree_id,
 	struct bch_fs *c = trans->c;
 	struct bkey deleted = KEY(0, 0, 0);
 	struct bkey_s_c old = (struct bkey_s_c) { &deleted, NULL };
+	struct printbuf buf = PRINTBUF;
 	int ret = 0;
 
 	deleted.p = k->k->p;
@@ -848,11 +849,23 @@ static int bch2_gc_mark_key(struct btree_trans *trans, enum btree_id btree_id,
 	if (ret)
 		goto err;
 
+	if (mustfix_fsck_err_on(level && !bch2_dev_btree_bitmap_marked(c, *k),
+				c, btree_bitmap_not_marked,
+				"btree ptr not marked in member info btree allocated bitmap\n  %s",
+				(bch2_bkey_val_to_text(&buf, c, *k),
+				 buf.buf))) {
+		mutex_lock(&c->sb_lock);
+		bch2_dev_btree_bitmap_mark(c, *k);
+		bch2_write_super(c);
+		mutex_unlock(&c->sb_lock);
+	}
+
 	ret = commit_do(trans, NULL, NULL, 0,
 			bch2_key_trigger(trans, btree_id, level, old,
 					 unsafe_bkey_s_c_to_s(*k), BTREE_TRIGGER_GC));
 fsck_err:
 err:
+	printbuf_exit(&buf);
 	bch_err_fn(c, ret);
 	return ret;
 }
diff --git a/fs/bcachefs/btree_node_scan.c b/fs/bcachefs/btree_node_scan.c
index 20f2b37c4474..866bd278439f 100644
--- a/fs/bcachefs/btree_node_scan.c
+++ b/fs/bcachefs/btree_node_scan.c
@@ -205,8 +205,13 @@ static int read_btree_nodes_worker(void *p)
 				last_print = jiffies;
 			}
 
-			try_read_btree_node(w->f, ca, bio, buf,
-					    bucket * ca->mi.bucket_size + bucket_offset);
+			u64 sector = bucket * ca->mi.bucket_size + bucket_offset;
+
+			if (c->sb.version_upgrade_complete >= bcachefs_metadata_version_mi_btree_bitmap &&
+			    !bch2_dev_btree_bitmap_marked_sectors(ca, sector, btree_sectors(c)))
+				continue;
+
+			try_read_btree_node(w->f, ca, bio, buf, sector);
 		}
 err:
 	bio_put(bio);
diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c
index 82ac00aa0f7b..6030c396754f 100644
--- a/fs/bcachefs/btree_update_interior.c
+++ b/fs/bcachefs/btree_update_interior.c
@@ -21,6 +21,7 @@
 #include "keylist.h"
 #include "recovery_passes.h"
 #include "replicas.h"
+#include "sb-members.h"
 #include "super-io.h"
 #include "trace.h"
 
@@ -605,6 +606,26 @@ static void btree_update_add_key(struct btree_update *as,
 	bch2_keylist_push(keys);
 }
 
+static bool btree_update_new_nodes_marked_sb(struct btree_update *as)
+{
+	for_each_keylist_key(&as->new_keys, k)
+		if (!bch2_dev_btree_bitmap_marked(as->c, bkey_i_to_s_c(k)))
+			return false;
+	return true;
+}
+
+static void btree_update_new_nodes_mark_sb(struct btree_update *as)
+{
+	struct bch_fs *c = as->c;
+
+	mutex_lock(&c->sb_lock);
+	for_each_keylist_key(&as->new_keys, k)
+		bch2_dev_btree_bitmap_mark(c, bkey_i_to_s_c(k));
+
+	bch2_write_super(c);
+	mutex_unlock(&c->sb_lock);
+}
+
 /*
  * The transactional part of an interior btree node update, where we journal the
  * update we did to the interior node and update alloc info:
@@ -662,6 +683,9 @@ static void btree_update_nodes_written(struct btree_update *as)
 	if (ret)
 		goto err;
 
+	if (!btree_update_new_nodes_marked_sb(as))
+		btree_update_new_nodes_mark_sb(as);
+
 	/*
 	 * Wait for any in flight writes to finish before we free the old nodes
 	 * on disk:
diff --git a/fs/bcachefs/sb-downgrade.c b/fs/bcachefs/sb-downgrade.c
index d6f81179c3a2..a98ef940b7a3 100644
--- a/fs/bcachefs/sb-downgrade.c
+++ b/fs/bcachefs/sb-downgrade.c
@@ -51,7 +51,10 @@
 	  BCH_FSCK_ERR_subvol_fs_path_parent_wrong)		\
 	x(btree_subvolume_children,				\
 	  BIT_ULL(BCH_RECOVERY_PASS_check_subvols),		\
-	  BCH_FSCK_ERR_subvol_children_not_set)
+	  BCH_FSCK_ERR_subvol_children_not_set)			\
+	x(mi_btree_bitmap,					\
+	  BIT_ULL(BCH_RECOVERY_PASS_check_allocations),		\
+	  BCH_FSCK_ERR_btree_bitmap_not_marked)
 
 #define DOWNGRADE_TABLE()
 
diff --git a/fs/bcachefs/sb-errors_types.h b/fs/bcachefs/sb-errors_types.h
index d7d609131030..5b600c6d7aca 100644
--- a/fs/bcachefs/sb-errors_types.h
+++ b/fs/bcachefs/sb-errors_types.h
@@ -270,7 +270,8 @@
 	x(btree_ptr_v2_min_key_bad,				262)	\
 	x(btree_root_unreadable_and_scan_found_nothing,		263)	\
 	x(snapshot_node_missing,				264)	\
-	x(dup_backpointer_to_bad_csum_extent,			265)
+	x(dup_backpointer_to_bad_csum_extent,			265)	\
+	x(btree_bitmap_not_marked,				266)
 
 enum bch_sb_error_id {
 #define x(t, n) BCH_FSCK_ERR_##t = n,
diff --git a/fs/bcachefs/sb-members.c b/fs/bcachefs/sb-members.c
index eff5ce18c69c..522a969345e5 100644
--- a/fs/bcachefs/sb-members.c
+++ b/fs/bcachefs/sb-members.c
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0
 
 #include "bcachefs.h"
+#include "btree_cache.h"
 #include "disk_groups.h"
 #include "opts.h"
 #include "replicas.h"
@@ -426,3 +427,55 @@ void bch2_dev_errors_reset(struct bch_dev *ca)
 	bch2_write_super(c);
 	mutex_unlock(&c->sb_lock);
 }
+
+/*
+ * Per member "range has btree nodes" bitmap:
+ *
+ * This is so that if we ever have to run the btree node scan to repair we don't
+ * have to scan full devices:
+ */
+
+bool bch2_dev_btree_bitmap_marked(struct bch_fs *c, struct bkey_s_c k)
+{
+	bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr)
+		if (!bch2_dev_btree_bitmap_marked_sectors(bch_dev_bkey_exists(c, ptr->dev),
+							  ptr->offset, btree_sectors(c)))
+			return false;
+	return true;
+}
+
+static void __bch2_dev_btree_bitmap_mark(struct bch_sb_field_members_v2 *mi, unsigned dev,
+				u64 start, unsigned sectors)
+{
+	struct bch_member *m = __bch2_members_v2_get_mut(mi, dev);
+	u64 bitmap = le64_to_cpu(m->btree_allocated_bitmap);
+
+	u64 end = start + sectors;
+
+	int resize = ilog2(roundup_pow_of_two(end)) - (m->btree_bitmap_shift + 6);
+	if (resize > 0) {
+		u64 new_bitmap = 0;
+
+		for (unsigned i = 0; i < 64; i++)
+			if (bitmap & BIT_ULL(i))
+				new_bitmap |= BIT_ULL(i >> resize);
+		bitmap = new_bitmap;
+		m->btree_bitmap_shift += resize;
+	}
+
+	for (unsigned bit = sectors >> m->btree_bitmap_shift;
+	     bit << m->btree_bitmap_shift < end;
+	     bit++)
+		bitmap |= BIT_ULL(bit);
+
+	m->btree_allocated_bitmap = cpu_to_le64(bitmap);
+}
+
+void bch2_dev_btree_bitmap_mark(struct bch_fs *c, struct bkey_s_c k)
+{
+	lockdep_assert_held(&c->sb_lock);
+
+	struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
+	bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr)
+		__bch2_dev_btree_bitmap_mark(mi, ptr->dev, ptr->offset, btree_sectors(c));
+}
diff --git a/fs/bcachefs/sb-members.h b/fs/bcachefs/sb-members.h
index be0a94183271..b27c3e4467cf 100644
--- a/fs/bcachefs/sb-members.h
+++ b/fs/bcachefs/sb-members.h
@@ -3,6 +3,7 @@
 #define _BCACHEFS_SB_MEMBERS_H
 
 #include "darray.h"
+#include "bkey_types.h"
 
 extern char * const bch2_member_error_strs[];
 
@@ -220,6 +221,8 @@ static inline struct bch_member_cpu bch2_mi_to_cpu(struct bch_member *mi)
 			: 1,
 		.freespace_initialized = BCH_MEMBER_FREESPACE_INITIALIZED(mi),
 		.valid		= bch2_member_exists(mi),
+		.btree_bitmap_shift	= mi->btree_bitmap_shift,
+		.btree_allocated_bitmap = le64_to_cpu(mi->btree_allocated_bitmap),
 	};
 }
 
@@ -228,4 +231,22 @@ void bch2_sb_members_from_cpu(struct bch_fs *);
 void bch2_dev_io_errors_to_text(struct printbuf *, struct bch_dev *);
 void bch2_dev_errors_reset(struct bch_dev *);
 
+static inline bool bch2_dev_btree_bitmap_marked_sectors(struct bch_dev *ca, u64 start, unsigned sectors)
+{
+	u64 end = start + sectors;
+
+	if (end > 64 << ca->mi.btree_bitmap_shift)
+		return false;
+
+	for (unsigned bit = sectors >> ca->mi.btree_bitmap_shift;
+	     bit << ca->mi.btree_bitmap_shift < end;
+	     bit++)
+		if (!(ca->mi.btree_allocated_bitmap & BIT_ULL(bit)))
+			return false;
+	return true;
+}
+
+bool bch2_dev_btree_bitmap_marked(struct bch_fs *, struct bkey_s_c);
+void bch2_dev_btree_bitmap_mark(struct bch_fs *, struct bkey_s_c);
+
 #endif /* _BCACHEFS_SB_MEMBERS_H */
diff --git a/fs/bcachefs/super_types.h b/fs/bcachefs/super_types.h
index ec784d975f66..11bcef170c2c 100644
--- a/fs/bcachefs/super_types.h
+++ b/fs/bcachefs/super_types.h
@@ -37,6 +37,8 @@ struct bch_member_cpu {
 	u8			durability;
 	u8			freespace_initialized;
 	u8			valid;
+	u8			btree_bitmap_shift;
+	u64			btree_allocated_bitmap;
 };
 
 #endif /* _BCACHEFS_SUPER_TYPES_H */

From f0a73d4fde5b285d94a702026216d9fd1fd2733d Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sun, 14 Apr 2024 00:51:48 -0400
Subject: [PATCH 22/24] bcachefs: Check for backpointer bucket_offset >= bucket
 size

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/backpointers.c    | 8 +++++---
 fs/bcachefs/backpointers.h    | 9 +++------
 fs/bcachefs/sb-errors_types.h | 2 +-
 3 files changed, 9 insertions(+), 10 deletions(-)

diff --git a/fs/bcachefs/backpointers.c b/fs/bcachefs/backpointers.c
index 114328acde72..fadb1078903d 100644
--- a/fs/bcachefs/backpointers.c
+++ b/fs/bcachefs/backpointers.c
@@ -49,13 +49,15 @@ int bch2_backpointer_invalid(struct bch_fs *c, struct bkey_s_c k,
 	if (!bch2_dev_exists2(c, bp.k->p.inode))
 		return 0;
 
+	struct bch_dev *ca = bch_dev_bkey_exists(c, bp.k->p.inode);
 	struct bpos bucket = bp_pos_to_bucket(c, bp.k->p);
 	int ret = 0;
 
-	bkey_fsck_err_on(!bpos_eq(bp.k->p, bucket_pos_to_bp(c, bucket, bp.v->bucket_offset)),
+	bkey_fsck_err_on((bp.v->bucket_offset >> MAX_EXTENT_COMPRESS_RATIO_SHIFT) >= ca->mi.bucket_size ||
+			 !bpos_eq(bp.k->p, bucket_pos_to_bp(c, bucket, bp.v->bucket_offset)),
 			 c, err,
-			 backpointer_pos_wrong,
-			 "backpointer at wrong pos");
+			 backpointer_bucket_offset_wrong,
+			 "backpointer bucket_offset wrong");
 fsck_err:
 	return ret;
 }
diff --git a/fs/bcachefs/backpointers.h b/fs/bcachefs/backpointers.h
index da012ca7daee..85949b9fd880 100644
--- a/fs/bcachefs/backpointers.h
+++ b/fs/bcachefs/backpointers.h
@@ -53,14 +53,11 @@ static inline struct bpos bucket_pos_to_bp(const struct bch_fs *c,
 					   u64 bucket_offset)
 {
 	struct bch_dev *ca = bch_dev_bkey_exists(c, bucket.inode);
-	struct bpos ret;
-
-	ret = POS(bucket.inode,
-		  (bucket_to_sector(ca, bucket.offset) <<
-		   MAX_EXTENT_COMPRESS_RATIO_SHIFT) + bucket_offset);
+	struct bpos ret = POS(bucket.inode,
+			      (bucket_to_sector(ca, bucket.offset) <<
+			       MAX_EXTENT_COMPRESS_RATIO_SHIFT) + bucket_offset);
 
 	EBUG_ON(!bkey_eq(bucket, bp_pos_to_bucket(c, ret)));
-
 	return ret;
 }
 
diff --git a/fs/bcachefs/sb-errors_types.h b/fs/bcachefs/sb-errors_types.h
index 5b600c6d7aca..4ca6e7b0d8aa 100644
--- a/fs/bcachefs/sb-errors_types.h
+++ b/fs/bcachefs/sb-errors_types.h
@@ -130,7 +130,7 @@
 	x(bucket_gens_nonzero_for_invalid_buckets,		122)	\
 	x(need_discard_freespace_key_to_invalid_dev_bucket,	123)	\
 	x(need_discard_freespace_key_bad,			124)	\
-	x(backpointer_pos_wrong,				125)	\
+	x(backpointer_bucket_offset_wrong,			125)	\
 	x(backpointer_to_missing_device,			126)	\
 	x(backpointer_to_missing_alloc,				127)	\
 	x(backpointer_to_missing_ptr,				128)	\

From 3078e059a5e984663c5c2b04485375c84c2700f9 Mon Sep 17 00:00:00 2001
From: Chao Yu <chao@kernel.org>
Date: Fri, 12 Apr 2024 14:36:38 +0800
Subject: [PATCH 23/24] bcachefs: fix error path of __bch2_read_super()

In __bch2_read_super(), if kstrdup() fails, it needs to release memory
in sb->holder, fix to call bch2_free_super() in the error path.

Signed-off-by: Chao Yu <chao@kernel.org>
Reviewed-by: Hongbo Li <lihongbo22@huawei.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/super-io.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/fs/bcachefs/super-io.c b/fs/bcachefs/super-io.c
index e0aa3655b63b..648986a7dbde 100644
--- a/fs/bcachefs/super-io.c
+++ b/fs/bcachefs/super-io.c
@@ -700,8 +700,11 @@ retry:
 		return -ENOMEM;
 
 	sb->sb_name = kstrdup(path, GFP_KERNEL);
-	if (!sb->sb_name)
-		return -ENOMEM;
+	if (!sb->sb_name) {
+		ret = -ENOMEM;
+		prt_printf(&err, "error allocating memory for sb_name");
+		goto err;
+	}
 
 #ifndef __KERNEL__
 	if (opt_get(*opts, direct_io) == false)

From ad29cf999a9161e7849aa229d2028854f90728c2 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sat, 13 Apr 2024 18:02:15 -0400
Subject: [PATCH 24/24] bcachefs: set_btree_iter_dontneed also clears
 should_be_locked

This is part of a larger series cleaning up the semantics of
should_be_locked and adding assertions around it; if we don't need an
iterator/path anymore, it clearly doesn't need to be locked.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_iter.h | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/fs/bcachefs/btree_iter.h b/fs/bcachefs/btree_iter.h
index 1d58d447b386..1c70836dd7cc 100644
--- a/fs/bcachefs/btree_iter.h
+++ b/fs/bcachefs/btree_iter.h
@@ -498,8 +498,13 @@ static inline void set_btree_iter_dontneed(struct btree_iter *iter)
 {
 	struct btree_trans *trans = iter->trans;
 
-	if (!trans->restarted)
-		btree_iter_path(trans, iter)->preserve = false;
+	if (!iter->path || trans->restarted)
+		return;
+
+	struct btree_path *path = btree_iter_path(trans, iter);
+	path->preserve		= false;
+	if (path->ref == 1)
+		path->should_be_locked	= false;
 }
 
 void *__bch2_trans_kmalloc(struct btree_trans *, size_t);