From 74ec2f302402c4cfd172f7254ae0e5b851e31b0a Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Mon, 14 Oct 2024 22:18:12 -0400
Subject: [PATCH 01/26] bcachefs: fix restart handling in bch2_rename2()

This should be impossible to hit in practice; the first lookup within a
transaction won't return a restart due to lock ordering, but we're
adding fault injection for transaction restarts and shaking out bugs.

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

diff --git a/fs/bcachefs/fs.c b/fs/bcachefs/fs.c
index 857175f4bbbe..d5d16d12e03a 100644
--- a/fs/bcachefs/fs.c
+++ b/fs/bcachefs/fs.c
@@ -871,7 +871,7 @@ static int bch2_rename2(struct mnt_idmap *idmap,
 	ret   = bch2_subvol_is_ro_trans(trans, src_dir->ei_inum.subvol) ?:
 		bch2_subvol_is_ro_trans(trans, dst_dir->ei_inum.subvol);
 	if (ret)
-		goto err;
+		goto err_tx_restart;
 
 	if (inode_attr_changing(dst_dir, src_inode, Inode_opt_project)) {
 		ret = bch2_fs_quota_transfer(c, src_inode,

From 94bdeec8f5303a8e445c9a7b9b2b2d69a29d7f38 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Mon, 14 Oct 2024 22:40:20 -0400
Subject: [PATCH 02/26] bcachefs: fix bch2_hash_delete() error path

we were exiting an iterator that hadn't been initialized

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/str_hash.h | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/fs/bcachefs/str_hash.h b/fs/bcachefs/str_hash.h
index 215eed4cce6d..cc2084e9f1cc 100644
--- a/fs/bcachefs/str_hash.h
+++ b/fs/bcachefs/str_hash.h
@@ -363,8 +363,11 @@ int bch2_hash_delete(struct btree_trans *trans,
 	struct btree_iter iter;
 	struct bkey_s_c k = bch2_hash_lookup(trans, &iter, desc, info, inum, key,
 					     BTREE_ITER_intent);
-	int ret = bkey_err(k) ?:
-		  bch2_hash_delete_at(trans, desc, info, &iter, 0);
+	int ret = bkey_err(k);
+	if (ret)
+		return ret;
+
+	ret = bch2_hash_delete_at(trans, desc, info, &iter, 0);
 	bch2_trans_iter_exit(trans, &iter);
 	return ret;
 }

From e1c4d2f0826d6815a268f42395b61962a05d3c51 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Mon, 14 Oct 2024 23:32:23 -0400
Subject: [PATCH 03/26] bcachefs: fix restart handling in bch2_fiemap()

We were leaking transaction restart errors to userspace.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/fs.c | 12 ++++--------
 1 file changed, 4 insertions(+), 8 deletions(-)

diff --git a/fs/bcachefs/fs.c b/fs/bcachefs/fs.c
index d5d16d12e03a..165d95994bd2 100644
--- a/fs/bcachefs/fs.c
+++ b/fs/bcachefs/fs.c
@@ -1268,7 +1268,7 @@ static int bch2_fiemap(struct inode *vinode, struct fiemap_extent_info *info,
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
 			     POS(ei->v.i_ino, start), 0);
 
-	while (true) {
+	while (!ret || bch2_err_matches(ret, BCH_ERR_transaction_restart)) {
 		enum btree_id data_btree = BTREE_ID_extents;
 
 		bch2_trans_begin(trans);
@@ -1276,14 +1276,14 @@ static int bch2_fiemap(struct inode *vinode, struct fiemap_extent_info *info,
 		u32 snapshot;
 		ret = bch2_subvolume_get_snapshot(trans, ei->ei_inum.subvol, &snapshot);
 		if (ret)
-			goto err;
+			continue;
 
 		bch2_btree_iter_set_snapshot(&iter, snapshot);
 
 		k = bch2_btree_iter_peek_upto(&iter, end);
 		ret = bkey_err(k);
 		if (ret)
-			goto err;
+			continue;
 
 		if (!k.k)
 			break;
@@ -1303,7 +1303,7 @@ static int bch2_fiemap(struct inode *vinode, struct fiemap_extent_info *info,
 		ret = bch2_read_indirect_extent(trans, &data_btree,
 					&offset_into_extent, &cur);
 		if (ret)
-			break;
+			continue;
 
 		k = bkey_i_to_s_c(cur.k);
 		bch2_bkey_buf_realloc(&prev, c, k.k->u64s);
@@ -1331,10 +1331,6 @@ static int bch2_fiemap(struct inode *vinode, struct fiemap_extent_info *info,
 
 		bch2_btree_iter_set_pos(&iter,
 			POS(iter.pos.inode, iter.pos.offset + sectors));
-err:
-		if (ret &&
-		    !bch2_err_matches(ret, BCH_ERR_transaction_restart))
-			break;
 	}
 	bch2_trans_iter_exit(trans, &iter);
 

From 028f3c1d9b3cddfbe327a3e6f09c3229aefc366c Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Mon, 14 Oct 2024 23:52:38 -0400
Subject: [PATCH 04/26] bcachefs: fix missing restart handling in
 bch2_read_retry_nodecode()

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/io_read.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/fs/bcachefs/io_read.c b/fs/bcachefs/io_read.c
index e4fc17c548fd..f00beff9ca0e 100644
--- a/fs/bcachefs/io_read.c
+++ b/fs/bcachefs/io_read.c
@@ -409,8 +409,8 @@ retry:
 	bch2_trans_begin(trans);
 	rbio->bio.bi_status = 0;
 
-	k = bch2_btree_iter_peek_slot(&iter);
-	if (bkey_err(k))
+	ret = lockrestart_do(trans, bkey_err(k = bch2_btree_iter_peek_slot(&iter)));
+	if (ret)
 		goto err;
 
 	bch2_bkey_buf_reassemble(&sk, c, k);

From 7ee4be9c621e4cd6a87d134a46d7c56debdf6664 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Tue, 15 Oct 2024 02:13:22 -0400
Subject: [PATCH 05/26] bcachefs: fix restart handling in
 bch2_do_invalidates_work()

this one is fairly harmless since the invalidate worker will just run
again later if it needs to, but still worth fixing

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

diff --git a/fs/bcachefs/alloc_background.c b/fs/bcachefs/alloc_background.c
index 6e161f8ffe8d..50f1919f4cbf 100644
--- a/fs/bcachefs/alloc_background.c
+++ b/fs/bcachefs/alloc_background.c
@@ -2137,14 +2137,15 @@ static void bch2_do_invalidates_work(struct work_struct *work)
 
 		struct bkey_s_c k = next_lru_key(trans, &iter, ca, &wrapped);
 		ret = bkey_err(k);
-		if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
-			continue;
 		if (ret)
-			break;
+			goto restart_err;
 		if (!k.k)
 			break;
 
 		ret = invalidate_one_bucket(trans, &iter, k, &nr_to_invalidate);
+restart_err:
+		if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
+			continue;
 		if (ret)
 			break;
 

From d8b50597748d6d960c936c89b972516c5e39ee96 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Wed, 16 Oct 2024 03:36:40 -0400
Subject: [PATCH 06/26] bcachefs: fix restart handling in
 bch2_alloc_write_key()

This is ugly:

We may discover in alloc_write_key that the data type we calculated is
wrong, because BCH_DATA_need_discard is checked/set elsewhere, and the
disk accounting counters we calculated need to be updated.

But bch2_alloc_key_to_dev_counters(..., BTREE_TRIGGER_gc) is not safe
w.r.t. transaction restarts, so we need to propagate the fixup back to
our gc state in case we take a transaction restart.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_gc.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c
index 94bbd8505582..0ca3feeb42c8 100644
--- a/fs/bcachefs/btree_gc.c
+++ b/fs/bcachefs/btree_gc.c
@@ -820,12 +820,22 @@ static int bch2_alloc_write_key(struct btree_trans *trans,
 	 * fix that here:
 	 */
 	alloc_data_type_set(&gc, gc.data_type);
-
 	if (gc.data_type != old_gc.data_type ||
 	    gc.dirty_sectors != old_gc.dirty_sectors) {
 		ret = bch2_alloc_key_to_dev_counters(trans, ca, &old_gc, &gc, BTREE_TRIGGER_gc);
 		if (ret)
 			return ret;
+
+		/*
+		 * Ugly: alloc_key_to_dev_counters(..., BTREE_TRIGGER_gc) is not
+		 * safe w.r.t. transaction restarts, so fixup the gc_bucket so
+		 * we don't run it twice:
+		 */
+		percpu_down_read(&c->mark_lock);
+		struct bucket *gc_m = gc_bucket(ca, iter->pos.offset);
+		gc_m->data_type = gc.data_type;
+		gc_m->dirty_sectors = gc.dirty_sectors;
+		percpu_up_read(&c->mark_lock);
 	}
 
 	if (fsck_err_on(new.data_type != gc.data_type,

From 29fd10a36a6326aba134cc0ea9def8d4601279d0 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Wed, 16 Oct 2024 04:11:15 -0400
Subject: [PATCH 07/26] bcachefs: fix restart handling in
 __bch2_resume_logged_op_finsert()

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

diff --git a/fs/bcachefs/io_misc.c b/fs/bcachefs/io_misc.c
index 307ed0a45184..f283051758d6 100644
--- a/fs/bcachefs/io_misc.c
+++ b/fs/bcachefs/io_misc.c
@@ -377,7 +377,7 @@ static int __bch2_resume_logged_op_finsert(struct btree_trans *trans,
 	 * check for missing subvolume before fpunch, as in resume we don't want
 	 * it to be a fatal error
 	 */
-	ret = __bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot, warn_errors);
+	ret = lockrestart_do(trans, __bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot, warn_errors));
 	if (ret)
 		return ret;
 

From 6bee2a04c5473d6d463c64e9a1f77c875b83401d Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Mon, 14 Oct 2024 23:58:45 -0400
Subject: [PATCH 08/26] bcachefs: handle restarts in
 bch2_bucket_io_time_reset()

bch2_bucket_io_time_reset() doesn't need to succeed, which is why it
didn't previously retry on transaction restart - but we're now treating
these as errors.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/alloc_background.c | 28 ++++++++++++++++------------
 1 file changed, 16 insertions(+), 12 deletions(-)

diff --git a/fs/bcachefs/alloc_background.c b/fs/bcachefs/alloc_background.c
index 50f1919f4cbf..69c7fc97e467 100644
--- a/fs/bcachefs/alloc_background.c
+++ b/fs/bcachefs/alloc_background.c
@@ -2351,24 +2351,19 @@ int bch2_dev_remove_alloc(struct bch_fs *c, struct bch_dev *ca)
 
 /* Bucket IO clocks: */
 
-int bch2_bucket_io_time_reset(struct btree_trans *trans, unsigned dev,
-			      size_t bucket_nr, int rw)
+static int __bch2_bucket_io_time_reset(struct btree_trans *trans, unsigned dev,
+				size_t bucket_nr, int rw)
 {
 	struct bch_fs *c = trans->c;
+
 	struct btree_iter iter;
-	struct bkey_i_alloc_v4 *a;
-	u64 now;
-	int ret = 0;
-
-	if (bch2_trans_relock(trans))
-		bch2_trans_begin(trans);
-
-	a = bch2_trans_start_alloc_update_noupdate(trans, &iter, POS(dev, bucket_nr));
-	ret = PTR_ERR_OR_ZERO(a);
+	struct bkey_i_alloc_v4 *a =
+		bch2_trans_start_alloc_update_noupdate(trans, &iter, POS(dev, bucket_nr));
+	int ret = PTR_ERR_OR_ZERO(a);
 	if (ret)
 		return ret;
 
-	now = bch2_current_io_time(c, rw);
+	u64 now = bch2_current_io_time(c, rw);
 	if (a->v.io_time[rw] == now)
 		goto out;
 
@@ -2381,6 +2376,15 @@ out:
 	return ret;
 }
 
+int bch2_bucket_io_time_reset(struct btree_trans *trans, unsigned dev,
+			      size_t bucket_nr, int rw)
+{
+	if (bch2_trans_relock(trans))
+		bch2_trans_begin(trans);
+
+	return nested_lockrestart_do(trans, __bch2_bucket_io_time_reset(trans, dev, bucket_nr, rw));
+}
+
 /* Startup/shutdown (ro/rw): */
 
 void bch2_recalc_capacity(struct bch_fs *c)

From a0d11feefb1998204f095fa0400024403d233108 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sun, 13 Oct 2024 21:53:26 -0400
Subject: [PATCH 09/26] bcachefs: Don't use commit_do() unnecessarily

Using commit_do() to call alloc_sectors_start_trans() breaks when we're
randomly injecting transaction restarts - the restart in the commit
causes us to leak the lock that alloc_sectorS_start_trans() takes.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/alloc_background.c      |  2 +-
 fs/bcachefs/alloc_foreground.c      |  2 +-
 fs/bcachefs/btree_io.c              |  2 +-
 fs/bcachefs/btree_iter.h            |  2 ++
 fs/bcachefs/btree_update.c          |  4 ++--
 fs/bcachefs/btree_update.h          |  2 +-
 fs/bcachefs/btree_update_interior.c |  4 +---
 fs/bcachefs/disk_accounting.c       |  6 ++++--
 fs/bcachefs/ec.c                    | 18 +++++++++---------
 fs/bcachefs/fs.c                    |  2 +-
 fs/bcachefs/fsck.c                  |  2 +-
 fs/bcachefs/inode.c                 |  3 +--
 fs/bcachefs/io_read.c               |  4 ++--
 fs/bcachefs/io_write.c              |  4 ++--
 fs/bcachefs/quota.c                 |  2 +-
 fs/bcachefs/rebalance.c             |  4 +++-
 fs/bcachefs/recovery.c              |  2 +-
 fs/bcachefs/subvolume.c             |  7 +++----
 fs/bcachefs/super.c                 |  2 +-
 fs/bcachefs/tests.c                 |  4 ++--
 fs/bcachefs/xattr.c                 |  2 +-
 21 files changed, 41 insertions(+), 39 deletions(-)

diff --git a/fs/bcachefs/alloc_background.c b/fs/bcachefs/alloc_background.c
index 69c7fc97e467..c84a91572a1d 100644
--- a/fs/bcachefs/alloc_background.c
+++ b/fs/bcachefs/alloc_background.c
@@ -1977,7 +1977,7 @@ static void bch2_do_discards_fast_work(struct work_struct *work)
 					     ca->mi.bucket_size,
 					     GFP_KERNEL);
 
-		int ret = bch2_trans_do(c, NULL, NULL,
+		int ret = bch2_trans_commit_do(c, NULL, NULL,
 			BCH_WATERMARK_btree|
 			BCH_TRANS_COMMIT_no_enospc,
 			bch2_clear_bucket_needs_discard(trans, POS(ca->dev_idx, bucket)));
diff --git a/fs/bcachefs/alloc_foreground.c b/fs/bcachefs/alloc_foreground.c
index d0e0b56892e3..5836870ab882 100644
--- a/fs/bcachefs/alloc_foreground.c
+++ b/fs/bcachefs/alloc_foreground.c
@@ -684,7 +684,7 @@ struct open_bucket *bch2_bucket_alloc(struct bch_fs *c, struct bch_dev *ca,
 	struct bch_dev_usage usage;
 	struct open_bucket *ob;
 
-	bch2_trans_do(c, NULL, NULL, 0,
+	bch2_trans_do(c,
 		      PTR_ERR_OR_ZERO(ob = bch2_bucket_alloc_trans(trans, ca, watermark,
 							data_type, cl, false, &usage)));
 	return ob;
diff --git a/fs/bcachefs/btree_io.c b/fs/bcachefs/btree_io.c
index cf933409d385..6296a11ccb09 100644
--- a/fs/bcachefs/btree_io.c
+++ b/fs/bcachefs/btree_io.c
@@ -1871,7 +1871,7 @@ static void btree_node_write_work(struct work_struct *work)
 
 		}
 	} else {
-		ret = bch2_trans_do(c, NULL, NULL, 0,
+		ret = bch2_trans_do(c,
 			bch2_btree_node_update_key_get_iter(trans, b, &wbio->key,
 					BCH_WATERMARK_interior_updates|
 					BCH_TRANS_COMMIT_journal_reclaim|
diff --git a/fs/bcachefs/btree_iter.h b/fs/bcachefs/btree_iter.h
index 31a58bf46fdb..0bda054f80d7 100644
--- a/fs/bcachefs/btree_iter.h
+++ b/fs/bcachefs/btree_iter.h
@@ -912,6 +912,8 @@ struct bkey_s_c bch2_btree_iter_peek_and_restart_outlined(struct btree_iter *);
 	_ret;								\
 })
 
+#define bch2_trans_do(_c, _do)	bch2_trans_run(_c, lockrestart_do(trans, _do))
+
 struct btree_trans *__bch2_trans_get(struct bch_fs *, unsigned);
 void bch2_trans_put(struct btree_trans *);
 
diff --git a/fs/bcachefs/btree_update.c b/fs/bcachefs/btree_update.c
index 514df618548e..5d809e8bd170 100644
--- a/fs/bcachefs/btree_update.c
+++ b/fs/bcachefs/btree_update.c
@@ -668,7 +668,7 @@ int bch2_btree_insert(struct bch_fs *c, enum btree_id id, struct bkey_i *k,
 		      struct disk_reservation *disk_res, int flags,
 		      enum btree_iter_update_trigger_flags iter_flags)
 {
-	return bch2_trans_do(c, disk_res, NULL, flags,
+	return bch2_trans_commit_do(c, disk_res, NULL, flags,
 			     bch2_btree_insert_trans(trans, id, k, iter_flags));
 }
 
@@ -865,7 +865,7 @@ __bch2_fs_log_msg(struct bch_fs *c, unsigned commit_flags, const char *fmt,
 		memcpy(l->d, buf.buf, buf.pos);
 		c->journal.early_journal_entries.nr += jset_u64s(u64s);
 	} else {
-		ret = bch2_trans_do(c, NULL, NULL,
+		ret = bch2_trans_commit_do(c, NULL, NULL,
 			BCH_TRANS_COMMIT_lazy_rw|commit_flags,
 			__bch2_trans_log_msg(trans, &buf, u64s));
 	}
diff --git a/fs/bcachefs/btree_update.h b/fs/bcachefs/btree_update.h
index 6a454f2fa005..70b3c989fac2 100644
--- a/fs/bcachefs/btree_update.h
+++ b/fs/bcachefs/btree_update.h
@@ -192,7 +192,7 @@ static inline int bch2_trans_commit(struct btree_trans *trans,
 	nested_lockrestart_do(_trans, _do ?: bch2_trans_commit(_trans, (_disk_res),\
 					(_journal_seq), (_flags)))
 
-#define bch2_trans_do(_c, _disk_res, _journal_seq, _flags, _do)		\
+#define bch2_trans_commit_do(_c, _disk_res, _journal_seq, _flags, _do)		\
 	bch2_trans_run(_c, commit_do(trans, _disk_res, _journal_seq, _flags, _do))
 
 #define trans_for_each_update(_trans, _i)				\
diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c
index 190bc1e81756..64f0928e1137 100644
--- a/fs/bcachefs/btree_update_interior.c
+++ b/fs/bcachefs/btree_update_interior.c
@@ -2239,10 +2239,8 @@ static void async_btree_node_rewrite_work(struct work_struct *work)
 	struct async_btree_rewrite *a =
 		container_of(work, struct async_btree_rewrite, work);
 	struct bch_fs *c = a->c;
-	int ret;
 
-	ret = bch2_trans_do(c, NULL, NULL, 0,
-		      async_btree_node_rewrite_trans(trans, a));
+	int ret = bch2_trans_do(c, async_btree_node_rewrite_trans(trans, a));
 	bch_err_fn_ratelimited(c, ret);
 	bch2_write_ref_put(c, BCH_WRITE_REF_node_rewrite);
 	kfree(a);
diff --git a/fs/bcachefs/disk_accounting.c b/fs/bcachefs/disk_accounting.c
index e309fb78529b..07eb8fa1b026 100644
--- a/fs/bcachefs/disk_accounting.c
+++ b/fs/bcachefs/disk_accounting.c
@@ -856,8 +856,10 @@ int bch2_dev_usage_init(struct bch_dev *ca, bool gc)
 	};
 	u64 v[3] = { ca->mi.nbuckets - ca->mi.first_bucket, 0, 0 };
 
-	int ret = bch2_trans_do(c, NULL, NULL, 0,
-			bch2_disk_accounting_mod(trans, &acc, v, ARRAY_SIZE(v), gc));
+	int ret = bch2_trans_do(c, ({
+		bch2_disk_accounting_mod(trans, &acc, v, ARRAY_SIZE(v), gc) ?:
+		(!gc ? bch2_trans_commit(trans, NULL, NULL, 0) : 0);
+	}));
 	bch_err_fn(c, ret);
 	return ret;
 }
diff --git a/fs/bcachefs/ec.c b/fs/bcachefs/ec.c
index e410cfe37b1a..a2303425959d 100644
--- a/fs/bcachefs/ec.c
+++ b/fs/bcachefs/ec.c
@@ -1186,7 +1186,7 @@ static void ec_stripe_delete_work(struct work_struct *work)
 		if (!idx)
 			break;
 
-		int ret = bch2_trans_do(c, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
+		int ret = bch2_trans_commit_do(c, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 					ec_stripe_delete(trans, idx));
 		bch_err_fn(c, ret);
 		if (ret)
@@ -1519,14 +1519,14 @@ static void ec_stripe_create(struct ec_stripe_new *s)
 		goto err;
 	}
 
-	ret = bch2_trans_do(c, &s->res, NULL,
-			    BCH_TRANS_COMMIT_no_check_rw|
-			    BCH_TRANS_COMMIT_no_enospc,
-			    ec_stripe_key_update(trans,
-					s->have_existing_stripe
-					? bkey_i_to_stripe(&s->existing_stripe.key)
-					: NULL,
-					bkey_i_to_stripe(&s->new_stripe.key)));
+	ret = bch2_trans_commit_do(c, &s->res, NULL,
+		BCH_TRANS_COMMIT_no_check_rw|
+		BCH_TRANS_COMMIT_no_enospc,
+		ec_stripe_key_update(trans,
+				     s->have_existing_stripe
+				     ? bkey_i_to_stripe(&s->existing_stripe.key)
+				     : NULL,
+				     bkey_i_to_stripe(&s->new_stripe.key)));
 	bch_err_msg(c, ret, "creating stripe key");
 	if (ret) {
 		goto err;
diff --git a/fs/bcachefs/fs.c b/fs/bcachefs/fs.c
index 165d95994bd2..506b9a2b5bd7 100644
--- a/fs/bcachefs/fs.c
+++ b/fs/bcachefs/fs.c
@@ -658,7 +658,7 @@ static struct dentry *bch2_lookup(struct inode *vdir, struct dentry *dentry,
 	struct bch_hash_info hash = bch2_hash_info_init(c, &dir->ei_inode);
 
 	struct bch_inode_info *inode;
-	bch2_trans_do(c, NULL, NULL, 0,
+	bch2_trans_do(c,
 		PTR_ERR_OR_ZERO(inode = bch2_lookup_trans(trans, inode_inum(dir),
 							  &hash, &dentry->d_name)));
 	if (IS_ERR(inode))
diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c
index a1087fd292e4..807098e76e7a 100644
--- a/fs/bcachefs/fsck.c
+++ b/fs/bcachefs/fsck.c
@@ -2509,7 +2509,7 @@ fsck_err:
 /* Get root directory, create if it doesn't exist: */
 int bch2_check_root(struct bch_fs *c)
 {
-	int ret = bch2_trans_do(c, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
+	int ret = bch2_trans_commit_do(c, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 		check_root_trans(trans));
 	bch_err_fn(c, ret);
 	return ret;
diff --git a/fs/bcachefs/inode.c b/fs/bcachefs/inode.c
index 2c037e84fbae..47544dc91d9e 100644
--- a/fs/bcachefs/inode.c
+++ b/fs/bcachefs/inode.c
@@ -1087,8 +1087,7 @@ int bch2_inode_find_by_inum_trans(struct btree_trans *trans,
 int bch2_inode_find_by_inum(struct bch_fs *c, subvol_inum inum,
 			    struct bch_inode_unpacked *inode)
 {
-	return bch2_trans_do(c, NULL, NULL, 0,
-		bch2_inode_find_by_inum_trans(trans, inum, inode));
+	return bch2_trans_do(c, bch2_inode_find_by_inum_trans(trans, inum, inode));
 }
 
 int bch2_inode_nlink_inc(struct bch_inode_unpacked *bi)
diff --git a/fs/bcachefs/io_read.c b/fs/bcachefs/io_read.c
index f00beff9ca0e..fc246f342820 100644
--- a/fs/bcachefs/io_read.c
+++ b/fs/bcachefs/io_read.c
@@ -557,8 +557,8 @@ out:
 
 static noinline void bch2_rbio_narrow_crcs(struct bch_read_bio *rbio)
 {
-	bch2_trans_do(rbio->c, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
-		      __bch2_rbio_narrow_crcs(trans, rbio));
+	bch2_trans_commit_do(rbio->c, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
+			     __bch2_rbio_narrow_crcs(trans, rbio));
 }
 
 /* Inner part that may run in process context */
diff --git a/fs/bcachefs/io_write.c b/fs/bcachefs/io_write.c
index b5fe9e0dc155..8609e25e450f 100644
--- a/fs/bcachefs/io_write.c
+++ b/fs/bcachefs/io_write.c
@@ -1437,7 +1437,7 @@ again:
 		 * freeing up space on specific disks, which means that
 		 * allocations for specific disks may hang arbitrarily long:
 		 */
-		ret = bch2_trans_do(c, NULL, NULL, 0,
+		ret = bch2_trans_run(c, lockrestart_do(trans,
 			bch2_alloc_sectors_start_trans(trans,
 				op->target,
 				op->opts.erasure_code && !(op->flags & BCH_WRITE_CACHED),
@@ -1447,7 +1447,7 @@ again:
 				op->nr_replicas_required,
 				op->watermark,
 				op->flags,
-				&op->cl, &wp));
+				&op->cl, &wp)));
 		if (unlikely(ret)) {
 			if (bch2_err_matches(ret, BCH_ERR_operation_blocked))
 				break;
diff --git a/fs/bcachefs/quota.c b/fs/bcachefs/quota.c
index c32a05e252e2..74f45a8162ad 100644
--- a/fs/bcachefs/quota.c
+++ b/fs/bcachefs/quota.c
@@ -869,7 +869,7 @@ static int bch2_set_quota(struct super_block *sb, struct kqid qid,
 	bkey_quota_init(&new_quota.k_i);
 	new_quota.k.p = POS(qid.type, from_kqid(&init_user_ns, qid));
 
-	ret = bch2_trans_do(c, NULL, NULL, 0,
+	ret = bch2_trans_commit_do(c, NULL, NULL, 0,
 			    bch2_set_quota_trans(trans, &new_quota, qdq)) ?:
 		__bch2_quota_set(c, bkey_i_to_s_c(&new_quota.k_i), qdq);
 
diff --git a/fs/bcachefs/rebalance.c b/fs/bcachefs/rebalance.c
index 2d299a37cf07..cd6647374353 100644
--- a/fs/bcachefs/rebalance.c
+++ b/fs/bcachefs/rebalance.c
@@ -70,7 +70,9 @@ err:
 
 int bch2_set_rebalance_needs_scan(struct bch_fs *c, u64 inum)
 {
-	int ret = bch2_trans_do(c, NULL, NULL, BCH_TRANS_COMMIT_no_enospc|BCH_TRANS_COMMIT_lazy_rw,
+	int ret = bch2_trans_commit_do(c, NULL, NULL,
+				       BCH_TRANS_COMMIT_no_enospc|
+				       BCH_TRANS_COMMIT_lazy_rw,
 			    __bch2_set_rebalance_needs_scan(trans, inum));
 	rebalance_wakeup(c);
 	return ret;
diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c
index 55e1504a8130..454b5a32dd7f 100644
--- a/fs/bcachefs/recovery.c
+++ b/fs/bcachefs/recovery.c
@@ -1091,7 +1091,7 @@ int bch2_fs_initialize(struct bch_fs *c)
 
 	bch2_inode_init_early(c, &lostfound_inode);
 
-	ret = bch2_trans_do(c, NULL, NULL, 0,
+	ret = bch2_trans_commit_do(c, NULL, NULL, 0,
 		bch2_create_trans(trans,
 				  BCACHEFS_ROOT_SUBVOL_INUM,
 				  &root_inode, &lostfound_inode,
diff --git a/fs/bcachefs/subvolume.c b/fs/bcachefs/subvolume.c
index 91d8187ee168..80e5efaff524 100644
--- a/fs/bcachefs/subvolume.c
+++ b/fs/bcachefs/subvolume.c
@@ -319,8 +319,7 @@ int bch2_subvol_is_ro_trans(struct btree_trans *trans, u32 subvol)
 
 int bch2_subvol_is_ro(struct bch_fs *c, u32 subvol)
 {
-	return bch2_trans_do(c, NULL, NULL, 0,
-		bch2_subvol_is_ro_trans(trans, subvol));
+	return bch2_trans_do(c, bch2_subvol_is_ro_trans(trans, subvol));
 }
 
 int bch2_snapshot_get_subvol(struct btree_trans *trans, u32 snapshot,
@@ -676,8 +675,8 @@ err:
 /* set bi_subvol on root inode */
 int bch2_fs_upgrade_for_subvolumes(struct bch_fs *c)
 {
-	int ret = bch2_trans_do(c, NULL, NULL, BCH_TRANS_COMMIT_lazy_rw,
-				__bch2_fs_upgrade_for_subvolumes(trans));
+	int ret = bch2_trans_commit_do(c, NULL, NULL, BCH_TRANS_COMMIT_lazy_rw,
+				       __bch2_fs_upgrade_for_subvolumes(trans));
 	bch_err_fn(c, ret);
 	return ret;
 }
diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c
index 77d811a539af..657fd3759e7b 100644
--- a/fs/bcachefs/super.c
+++ b/fs/bcachefs/super.c
@@ -1972,7 +1972,7 @@ int bch2_dev_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
 		};
 		u64 v[3] = { nbuckets - old_nbuckets, 0, 0 };
 
-		ret   = bch2_trans_do(ca->fs, NULL, NULL, 0,
+		ret   = bch2_trans_commit_do(ca->fs, NULL, NULL, 0,
 				bch2_disk_accounting_mod(trans, &acc, v, ARRAY_SIZE(v), false)) ?:
 			bch2_dev_freespace_init(c, ca, old_nbuckets, nbuckets);
 		if (ret)
diff --git a/fs/bcachefs/tests.c b/fs/bcachefs/tests.c
index b2f209743afe..315038a0a92d 100644
--- a/fs/bcachefs/tests.c
+++ b/fs/bcachefs/tests.c
@@ -450,7 +450,7 @@ static int insert_test_overlapping_extent(struct bch_fs *c, u64 inum, u64 start,
 	k.k_i.k.p.snapshot = snapid;
 	k.k_i.k.size = len;
 
-	ret = bch2_trans_do(c, NULL, NULL, 0,
+	ret = bch2_trans_commit_do(c, NULL, NULL, 0,
 		bch2_btree_insert_nonextent(trans, BTREE_ID_extents, &k.k_i,
 					    BTREE_UPDATE_internal_snapshot_node));
 	bch_err_fn(c, ret);
@@ -510,7 +510,7 @@ static int test_snapshots(struct bch_fs *c, u64 nr)
 	if (ret)
 		return ret;
 
-	ret = bch2_trans_do(c, NULL, NULL, 0,
+	ret = bch2_trans_commit_do(c, NULL, NULL, 0,
 		      bch2_snapshot_node_create(trans, U32_MAX,
 						snapids,
 						snapid_subvols,
diff --git a/fs/bcachefs/xattr.c b/fs/bcachefs/xattr.c
index 56c8d3fe55a4..952aca400faf 100644
--- a/fs/bcachefs/xattr.c
+++ b/fs/bcachefs/xattr.c
@@ -330,7 +330,7 @@ static int bch2_xattr_get_handler(const struct xattr_handler *handler,
 {
 	struct bch_inode_info *inode = to_bch_ei(vinode);
 	struct bch_fs *c = inode->v.i_sb->s_fs_info;
-	int ret = bch2_trans_do(c, NULL, NULL, 0,
+	int ret = bch2_trans_do(c,
 		bch2_xattr_get_trans(trans, inode, name, buffer, size, handler->flags));
 
 	if (ret < 0 && bch2_err_matches(ret, ENOENT))

From 4007bbb203a0c36e66bb2e785e1b2cb7266179d5 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Wed, 16 Oct 2024 06:32:12 -0400
Subject: [PATCH 10/26] bcachefS: ec: fix data type on stripe deletion

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/ec.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/fs/bcachefs/ec.c b/fs/bcachefs/ec.c
index a2303425959d..a0aa5bb467d9 100644
--- a/fs/bcachefs/ec.c
+++ b/fs/bcachefs/ec.c
@@ -266,12 +266,12 @@ static int __mark_stripe_bucket(struct btree_trans *trans,
 	if (!deleting) {
 		a->stripe		= s.k->p.offset;
 		a->stripe_redundancy	= s.v->nr_redundant;
+		alloc_data_type_set(a, data_type);
 	} else {
 		a->stripe		= 0;
 		a->stripe_redundancy	= 0;
+		alloc_data_type_set(a, BCH_DATA_user);
 	}
-
-	alloc_data_type_set(a, data_type);
 err:
 	printbuf_exit(&buf);
 	return ret;

From 81e0b6c7c1f75c914dede330121391f57f216a6e Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Mon, 14 Oct 2024 17:55:48 -0400
Subject: [PATCH 11/26] bcachefs: fix disk reservation accounting in
 bch2_folio_reservation_get()

bch2_disk_reservation_put() zeroes out the reservation - oops.

This fixes a disk reservation leak when getting a quota reservation
returned an error.

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

diff --git a/fs/bcachefs/fs-io-pagecache.c b/fs/bcachefs/fs-io-pagecache.c
index af3a24546aa3..cde3a0445ee9 100644
--- a/fs/bcachefs/fs-io-pagecache.c
+++ b/fs/bcachefs/fs-io-pagecache.c
@@ -431,9 +431,9 @@ int bch2_folio_reservation_get(struct bch_fs *c,
 		ret = bch2_quota_reservation_add(c, inode, &res->quota, quota_sectors, true);
 		if (unlikely(ret)) {
 			struct disk_reservation tmp = { .sectors = disk_sectors };
+			res->disk.sectors -= disk_sectors;
 
 			bch2_disk_reservation_put(c, &tmp);
-			res->disk.sectors -= disk_sectors;
 			return ret;
 		}
 	}

From 335d318ef5329e0b500e0f3394bbe0a14cd7f99b Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 17 Oct 2024 01:05:17 -0400
Subject: [PATCH 12/26] bcachefs: bch2_folio_reservation_get_partial() is now
 better behaved

bch2_folio_reservation_get_partial(), on partial success, will now
return a reservation that's aligned to the filesystem blocksize.

This is a partial fix for fstests generic/299 - fio verify is badly
behaved in the presence of short writes that aren't aligned to its
blocksize.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/buckets.c         |  7 +++-
 fs/bcachefs/buckets.h         | 12 +++---
 fs/bcachefs/fs-io-pagecache.c | 70 ++++++++++++++++++++++-------------
 3 files changed, 57 insertions(+), 32 deletions(-)

diff --git a/fs/bcachefs/buckets.c b/fs/bcachefs/buckets.c
index 546cd01a72e3..ec7d9a59bea9 100644
--- a/fs/bcachefs/buckets.c
+++ b/fs/bcachefs/buckets.c
@@ -1160,11 +1160,11 @@ int bch2_trans_mark_dev_sbs(struct bch_fs *c)
 #define SECTORS_CACHE	1024
 
 int __bch2_disk_reservation_add(struct bch_fs *c, struct disk_reservation *res,
-			      u64 sectors, int flags)
+				u64 sectors, enum bch_reservation_flags flags)
 {
 	struct bch_fs_pcpu *pcpu;
 	u64 old, get;
-	s64 sectors_available;
+	u64 sectors_available;
 	int ret;
 
 	percpu_down_read(&c->mark_lock);
@@ -1202,6 +1202,9 @@ recalculate:
 	percpu_u64_set(&c->pcpu->sectors_available, 0);
 	sectors_available = avail_factor(__bch2_fs_usage_read_short(c).free);
 
+	if (sectors_available && (flags & BCH_DISK_RESERVATION_PARTIAL))
+		sectors = min(sectors, sectors_available);
+
 	if (sectors <= sectors_available ||
 	    (flags & BCH_DISK_RESERVATION_NOFAIL)) {
 		atomic64_set(&c->sectors_available,
diff --git a/fs/bcachefs/buckets.h b/fs/bcachefs/buckets.h
index e2cb7b24b220..fd5e6ccad45e 100644
--- a/fs/bcachefs/buckets.h
+++ b/fs/bcachefs/buckets.h
@@ -344,14 +344,16 @@ static inline void bch2_disk_reservation_put(struct bch_fs *c,
 	}
 }
 
-#define BCH_DISK_RESERVATION_NOFAIL		(1 << 0)
+enum bch_reservation_flags {
+	BCH_DISK_RESERVATION_NOFAIL	= 1 << 0,
+	BCH_DISK_RESERVATION_PARTIAL	= 1 << 1,
+};
 
-int __bch2_disk_reservation_add(struct bch_fs *,
-				struct disk_reservation *,
-				u64, int);
+int __bch2_disk_reservation_add(struct bch_fs *, struct disk_reservation *,
+				u64, enum bch_reservation_flags);
 
 static inline int bch2_disk_reservation_add(struct bch_fs *c, struct disk_reservation *res,
-					    u64 sectors, int flags)
+					    u64 sectors, enum bch_reservation_flags flags)
 {
 #ifdef __KERNEL__
 	u64 old, new;
diff --git a/fs/bcachefs/fs-io-pagecache.c b/fs/bcachefs/fs-io-pagecache.c
index cde3a0445ee9..1d4910ea0f1d 100644
--- a/fs/bcachefs/fs-io-pagecache.c
+++ b/fs/bcachefs/fs-io-pagecache.c
@@ -399,14 +399,17 @@ void bch2_folio_reservation_put(struct bch_fs *c,
 	bch2_quota_reservation_put(c, inode, &res->quota);
 }
 
-int bch2_folio_reservation_get(struct bch_fs *c,
+static int __bch2_folio_reservation_get(struct bch_fs *c,
 			struct bch_inode_info *inode,
 			struct folio *folio,
 			struct bch2_folio_reservation *res,
-			size_t offset, size_t len)
+			size_t offset, size_t len,
+			bool partial)
 {
 	struct bch_folio *s = bch2_folio_create(folio, 0);
 	unsigned i, disk_sectors = 0, quota_sectors = 0;
+	struct disk_reservation disk_res = {};
+	size_t reserved = len;
 	int ret;
 
 	if (!s)
@@ -422,23 +425,56 @@ int bch2_folio_reservation_get(struct bch_fs *c,
 	}
 
 	if (disk_sectors) {
-		ret = bch2_disk_reservation_add(c, &res->disk, disk_sectors, 0);
+		ret = bch2_disk_reservation_add(c, &disk_res, disk_sectors,
+				partial ? BCH_DISK_RESERVATION_PARTIAL : 0);
 		if (unlikely(ret))
 			return ret;
+
+		if (unlikely(disk_res.sectors != disk_sectors)) {
+			disk_sectors = quota_sectors = 0;
+
+			for (i = round_down(offset, block_bytes(c)) >> 9;
+			     i < round_up(offset + len, block_bytes(c)) >> 9;
+			     i++) {
+				disk_sectors += sectors_to_reserve(&s->s[i], res->disk.nr_replicas);
+				if (disk_sectors > disk_res.sectors) {
+					/*
+					 * Make sure to get a reservation that's
+					 * aligned to the filesystem blocksize:
+					 */
+					unsigned reserved_offset = round_down(i << 9, block_bytes(c));
+					reserved = clamp(reserved_offset, offset, offset + len) - offset;
+
+					if (!reserved) {
+						bch2_disk_reservation_put(c, &disk_res);
+						return -BCH_ERR_ENOSPC_disk_reservation;
+					}
+					break;
+				}
+				quota_sectors += s->s[i].state == SECTOR_unallocated;
+			}
+		}
 	}
 
 	if (quota_sectors) {
 		ret = bch2_quota_reservation_add(c, inode, &res->quota, quota_sectors, true);
 		if (unlikely(ret)) {
-			struct disk_reservation tmp = { .sectors = disk_sectors };
-			res->disk.sectors -= disk_sectors;
-
-			bch2_disk_reservation_put(c, &tmp);
+			bch2_disk_reservation_put(c, &disk_res);
 			return ret;
 		}
 	}
 
-	return 0;
+	res->disk.sectors += disk_res.sectors;
+	return partial ? reserved : 0;
+}
+
+int bch2_folio_reservation_get(struct bch_fs *c,
+			struct bch_inode_info *inode,
+			struct folio *folio,
+			struct bch2_folio_reservation *res,
+			size_t offset, size_t len)
+{
+	return __bch2_folio_reservation_get(c, inode, folio, res, offset, len, false);
 }
 
 ssize_t bch2_folio_reservation_get_partial(struct bch_fs *c,
@@ -447,23 +483,7 @@ ssize_t bch2_folio_reservation_get_partial(struct bch_fs *c,
 			struct bch2_folio_reservation *res,
 			size_t offset, size_t len)
 {
-	size_t l, reserved = 0;
-	int ret;
-
-	while ((l = len - reserved)) {
-		while ((ret = bch2_folio_reservation_get(c, inode, folio, res, offset, l))) {
-			if ((offset & (block_bytes(c) - 1)) + l <= block_bytes(c))
-				return reserved ?: ret;
-
-			len = reserved + l;
-			l /= 2;
-		}
-
-		offset += l;
-		reserved += l;
-	}
-
-	return reserved;
+	return __bch2_folio_reservation_get(c, inode, folio, res, offset, len, true);
 }
 
 static void bch2_clear_folio_bits(struct folio *folio)

From 97535cd84f189248ea0fe14544628d480908e99b Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 17 Oct 2024 01:10:49 -0400
Subject: [PATCH 13/26] bcachefs: Fix data corruption on -ENOSPC in buffered
 write path

Found by generic/299: When we have to truncate a write due to -ENOSPC,
we may have to read in the folio we're writing to if we're now no longer
doing a complete write to a !uptodate folio.

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

diff --git a/fs/bcachefs/fs-io-buffered.c b/fs/bcachefs/fs-io-buffered.c
index 48a1ab9a649b..95972809e76d 100644
--- a/fs/bcachefs/fs-io-buffered.c
+++ b/fs/bcachefs/fs-io-buffered.c
@@ -856,6 +856,12 @@ static int __bch2_buffered_write(struct bch_inode_info *inode,
 				folios_trunc(&fs, fi);
 				end = min(end, folio_end_pos(darray_last(fs)));
 			} else {
+				if (!folio_test_uptodate(f)) {
+					ret = bch2_read_single_folio(f, mapping);
+					if (ret)
+						goto out;
+				}
+
 				folios_trunc(&fs, fi + 1);
 				end = f_pos + f_reserved;
 			}

From 07cf8bac2d3efa8d3fb62cd4d98c00087efd7fe8 Mon Sep 17 00:00:00 2001
From: Hongbo Li <lihongbo22@huawei.com>
Date: Thu, 26 Sep 2024 10:00:01 +0800
Subject: [PATCH 14/26] bcachefs: fix incorrect show_options results

When call show_options in bcachefs, the options buffer is appeneded
to the seq variable. In fact, it requires an additional comma to be
appended first. This will affect the remount process when reading
existing mount options.

Fixes: 9305cf91d05e ("bcachefs: bch2_opts_to_text()")
Signed-off-by: Hongbo Li <lihongbo22@huawei.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/fs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/bcachefs/fs.c b/fs/bcachefs/fs.c
index 506b9a2b5bd7..0ea443eeb8cd 100644
--- a/fs/bcachefs/fs.c
+++ b/fs/bcachefs/fs.c
@@ -2038,7 +2038,7 @@ static int bch2_show_options(struct seq_file *seq, struct dentry *root)
 	bch2_opts_to_text(&buf, c->opts, c, c->disk_sb.sb,
 			  OPT_MOUNT, OPT_HIDDEN, OPT_SHOW_MOUNT_STYLE);
 	printbuf_nul_terminate(&buf);
-	seq_puts(seq, buf.buf);
+	seq_printf(seq, ",%s", buf.buf);
 
 	int ret = buf.allocation_failure ? -ENOMEM : 0;
 	printbuf_exit(&buf);

From 489ecc4cfddada303bd9c2b287a3c8744c324ed3 Mon Sep 17 00:00:00 2001
From: Hongbo Li <lihongbo22@huawei.com>
Date: Thu, 10 Oct 2024 12:01:48 +0800
Subject: [PATCH 15/26] bcachefs: skip mount option handle for empty string.

The options parse in get_tree will split the options buffer, it will
get the empty string for last one by strsep(). After commit
ea0eeb89b1d5 ("bcachefs: reject unknown mount options") is merged,
unknown mount options is not allowed (here is empty string), and this
causes this errors. This can be reproduced just by the following steps:

    bcachefs format /dev/loop
    mount -t bcachefs -o metadata_target=loop1 /dev/loop1 /mnt/bcachefs/

Fixes: ea0eeb89b1d5 ("bcachefs: reject unknown mount options")
Signed-off-by: Hongbo Li <lihongbo22@huawei.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/opts.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/fs/bcachefs/opts.c b/fs/bcachefs/opts.c
index 84097235eea9..68d240126522 100644
--- a/fs/bcachefs/opts.c
+++ b/fs/bcachefs/opts.c
@@ -596,6 +596,9 @@ int bch2_parse_mount_opts(struct bch_fs *c, struct bch_opts *opts,
 	copied_opts_start = copied_opts;
 
 	while ((opt = strsep(&copied_opts, ",")) != NULL) {
+		if (!*opt)
+			continue;
+
 		name	= strsep(&opt, "=");
 		val	= opt;
 

From b96f8cd3870a140524fb8cec58790a4be04f41d9 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 17 Oct 2024 22:29:23 -0400
Subject: [PATCH 16/26] bcachefs: Run in-kernel offline fsck without ratelimit
 errors

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

diff --git a/fs/bcachefs/chardev.c b/fs/bcachefs/chardev.c
index cbfd88f98472..2182b555c112 100644
--- a/fs/bcachefs/chardev.c
+++ b/fs/bcachefs/chardev.c
@@ -225,6 +225,7 @@ static long bch2_ioctl_fsck_offline(struct bch_ioctl_fsck_offline __user *user_a
 
 	opt_set(thr->opts, stdio, (u64)(unsigned long)&thr->thr.stdio);
 	opt_set(thr->opts, read_only, 1);
+	opt_set(thr->opts, ratelimit_errors, 0);
 
 	/* We need request_key() to be called before we punt to kthread: */
 	opt_set(thr->opts, nostart, true);

From 78cf0ae636a55e0bef99308d305d4e1f8a6c4147 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 17 Oct 2024 22:55:59 -0400
Subject: [PATCH 17/26] bcachefs: INODE_STR_HASH() for bch_inode_unpacked

Trivial cleanup - add a normal BITMASK() helper for bch_inode_unpacked.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/inode.c        | 18 ++++++++----------
 fs/bcachefs/inode.h        |  1 +
 fs/bcachefs/inode_format.h |  6 +++---
 fs/bcachefs/str_hash.h     |  3 +--
 4 files changed, 13 insertions(+), 15 deletions(-)

diff --git a/fs/bcachefs/inode.c b/fs/bcachefs/inode.c
index 47544dc91d9e..3662f648a6c3 100644
--- a/fs/bcachefs/inode.c
+++ b/fs/bcachefs/inode.c
@@ -163,8 +163,8 @@ static noinline int bch2_inode_unpack_v1(struct bkey_s_c_inode inode,
 	unsigned fieldnr = 0, field_bits;
 	int ret;
 
-#define x(_name, _bits)					\
-	if (fieldnr++ == INODE_NR_FIELDS(inode.v)) {			\
+#define x(_name, _bits)							\
+	if (fieldnr++ == INODEv1_NR_FIELDS(inode.v)) {			\
 		unsigned offset = offsetof(struct bch_inode_unpacked, _name);\
 		memset((void *) unpacked + offset, 0,			\
 		       sizeof(*unpacked) - offset);			\
@@ -293,10 +293,10 @@ static noinline int bch2_inode_unpack_slowpath(struct bkey_s_c k,
 		unpacked->bi_flags	= le32_to_cpu(inode.v->bi_flags);
 		unpacked->bi_mode	= le16_to_cpu(inode.v->bi_mode);
 
-		if (INODE_NEW_VARINT(inode.v)) {
+		if (INODEv1_NEW_VARINT(inode.v)) {
 			return bch2_inode_unpack_v2(unpacked, inode.v->fields,
 						    bkey_val_end(inode),
-						    INODE_NR_FIELDS(inode.v));
+						    INODEv1_NR_FIELDS(inode.v));
 		} else {
 			return bch2_inode_unpack_v1(inode, unpacked);
 		}
@@ -471,10 +471,10 @@ int bch2_inode_validate(struct bch_fs *c, struct bkey_s_c k,
 	struct bkey_s_c_inode inode = bkey_s_c_to_inode(k);
 	int ret = 0;
 
-	bkey_fsck_err_on(INODE_STR_HASH(inode.v) >= BCH_STR_HASH_NR,
+	bkey_fsck_err_on(INODEv1_STR_HASH(inode.v) >= BCH_STR_HASH_NR,
 			 c, inode_str_hash_invalid,
 			 "invalid str hash type (%llu >= %u)",
-			 INODE_STR_HASH(inode.v), BCH_STR_HASH_NR);
+			 INODEv1_STR_HASH(inode.v), BCH_STR_HASH_NR);
 
 	ret = __bch2_inode_validate(c, k, flags);
 fsck_err:
@@ -800,10 +800,8 @@ void bch2_inode_init_early(struct bch_fs *c,
 
 	memset(inode_u, 0, sizeof(*inode_u));
 
-	/* ick */
-	inode_u->bi_flags |= str_hash << INODE_STR_HASH_OFFSET;
-	get_random_bytes(&inode_u->bi_hash_seed,
-			 sizeof(inode_u->bi_hash_seed));
+	SET_INODE_STR_HASH(inode_u, str_hash);
+	get_random_bytes(&inode_u->bi_hash_seed, sizeof(inode_u->bi_hash_seed));
 }
 
 void bch2_inode_init_late(struct bch_inode_unpacked *inode_u, u64 now,
diff --git a/fs/bcachefs/inode.h b/fs/bcachefs/inode.h
index c8e98443e2d4..eab82b5eb897 100644
--- a/fs/bcachefs/inode.h
+++ b/fs/bcachefs/inode.h
@@ -92,6 +92,7 @@ struct bch_inode_unpacked {
 	BCH_INODE_FIELDS_v3()
 #undef  x
 };
+BITMASK(INODE_STR_HASH,	struct bch_inode_unpacked, bi_flags, 20, 24);
 
 struct bkey_inode_buf {
 	struct bkey_i_inode_v3	inode;
diff --git a/fs/bcachefs/inode_format.h b/fs/bcachefs/inode_format.h
index a204e46b6b47..7928d0c6954f 100644
--- a/fs/bcachefs/inode_format.h
+++ b/fs/bcachefs/inode_format.h
@@ -150,9 +150,9 @@ enum __bch_inode_flags {
 #undef x
 };
 
-LE32_BITMASK(INODE_STR_HASH,	struct bch_inode, bi_flags, 20, 24);
-LE32_BITMASK(INODE_NR_FIELDS,	struct bch_inode, bi_flags, 24, 31);
-LE32_BITMASK(INODE_NEW_VARINT,	struct bch_inode, bi_flags, 31, 32);
+LE32_BITMASK(INODEv1_STR_HASH,	struct bch_inode, bi_flags, 20, 24);
+LE32_BITMASK(INODEv1_NR_FIELDS,	struct bch_inode, bi_flags, 24, 31);
+LE32_BITMASK(INODEv1_NEW_VARINT,struct bch_inode, bi_flags, 31, 32);
 
 LE64_BITMASK(INODEv2_STR_HASH,	struct bch_inode_v2, bi_flags, 20, 24);
 LE64_BITMASK(INODEv2_NR_FIELDS,	struct bch_inode_v2, bi_flags, 24, 31);
diff --git a/fs/bcachefs/str_hash.h b/fs/bcachefs/str_hash.h
index cc2084e9f1cc..d4b5718860bd 100644
--- a/fs/bcachefs/str_hash.h
+++ b/fs/bcachefs/str_hash.h
@@ -46,8 +46,7 @@ bch2_hash_info_init(struct bch_fs *c, const struct bch_inode_unpacked *bi)
 {
 	/* XXX ick */
 	struct bch_hash_info info = {
-		.type = (bi->bi_flags >> INODE_STR_HASH_OFFSET) &
-			~(~0U << INODE_STR_HASH_BITS),
+		.type = INODE_STR_HASH(bi),
 		.siphash_key = { .k0 = bi->bi_hash_seed }
 	};
 

From d8e879377ffb37ba0d3afa0c92bd3b88b849a0a9 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sat, 28 Sep 2024 14:44:06 -0400
Subject: [PATCH 18/26] bcachefs: Add hash seed, type to inode_to_text()

This helped with discovering some filesystem corruption fsck has having
trouble with: the str_hash type had gotten flipped on one snapshot's
version of an inode.

All versions of a given inode number have the same hash seed and hash
type, since lookups will be done with a single hash/seed and type and
see dirents/xattrs from multiple snapshots.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/inode.c | 4 ++++
 fs/bcachefs/opts.c  | 3 ++-
 fs/bcachefs/opts.h  | 3 ++-
 3 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/fs/bcachefs/inode.c b/fs/bcachefs/inode.c
index 3662f648a6c3..b9fde1566df1 100644
--- a/fs/bcachefs/inode.c
+++ b/fs/bcachefs/inode.c
@@ -533,6 +533,10 @@ static void __bch2_inode_unpacked_to_text(struct printbuf *out,
 	prt_printf(out, "(%x)\n", inode->bi_flags);
 
 	prt_printf(out, "journal_seq=%llu\n",	inode->bi_journal_seq);
+	prt_printf(out, "hash_seed=%llx\n",	inode->bi_hash_seed);
+	prt_printf(out, "hash_type=");
+	bch2_prt_str_hash_type(out, INODE_STR_HASH(inode));
+	prt_newline(out);
 	prt_printf(out, "bi_size=%llu\n",	inode->bi_size);
 	prt_printf(out, "bi_sectors=%llu\n",	inode->bi_sectors);
 	prt_printf(out, "bi_version=%llu\n",	inode->bi_version);
diff --git a/fs/bcachefs/opts.c b/fs/bcachefs/opts.c
index 68d240126522..6673cbd8bdb9 100644
--- a/fs/bcachefs/opts.c
+++ b/fs/bcachefs/opts.c
@@ -63,7 +63,7 @@ const char * const bch2_compression_opts[] = {
 	NULL
 };
 
-const char * const bch2_str_hash_types[] = {
+const char * const __bch2_str_hash_types[] = {
 	BCH_STR_HASH_TYPES()
 	NULL
 };
@@ -115,6 +115,7 @@ 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);
+PRT_STR_OPT_BOUNDSCHECKED(str_hash_type,	enum bch_str_hash_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 cb2e244a2429..23dda014e331 100644
--- a/fs/bcachefs/opts.h
+++ b/fs/bcachefs/opts.h
@@ -18,7 +18,7 @@ extern const char * const bch2_sb_compat[];
 extern const char * const __bch2_btree_ids[];
 extern const char * const bch2_csum_opts[];
 extern const char * const bch2_compression_opts[];
-extern const char * const bch2_str_hash_types[];
+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[];
@@ -29,6 +29,7 @@ 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);
+void bch2_prt_str_hash_type(struct printbuf *,		enum bch_str_hash_type);
 
 static inline const char *bch2_d_type_str(unsigned d_type)
 {

From 15a3836c8ed7bf102159e70ed380b8158651df8e Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 17 Oct 2024 23:00:14 -0400
Subject: [PATCH 19/26] bcachefs: Repair mismatches in inode hash seed, type

Different versions of the same inode (same inode number, different
snapshot ID) must have the same hash seed and type - lookups require
this, since they see keys from different snapshots simultaneously.

To repair we only need to make the inodes consistent, hash_check_key()
will do the rest.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/fsck.c | 49 ++++++++++++++++++++++++++++++++++++----------
 1 file changed, 39 insertions(+), 10 deletions(-)

diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c
index 807098e76e7a..16a408210bb7 100644
--- a/fs/bcachefs/fsck.c
+++ b/fs/bcachefs/fsck.c
@@ -1096,10 +1096,36 @@ fsck_err:
 	return ret;
 }
 
+static int get_snapshot_root_inode(struct btree_trans *trans,
+				   struct bch_inode_unpacked *root,
+				   u64 inum)
+{
+	struct btree_iter iter;
+	struct bkey_s_c k;
+	int ret = 0;
+
+	for_each_btree_key_reverse_norestart(trans, iter, BTREE_ID_inodes,
+					     SPOS(0, inum, U32_MAX),
+					     BTREE_ITER_all_snapshots, k, ret) {
+		if (k.k->p.offset != inum)
+			break;
+		if (bkey_is_inode(k.k))
+			goto found_root;
+	}
+	if (ret)
+		goto err;
+	BUG();
+found_root:
+	BUG_ON(bch2_inode_unpack(k, root));
+err:
+	bch2_trans_iter_exit(trans, &iter);
+	return ret;
+}
+
 static int check_inode(struct btree_trans *trans,
 		       struct btree_iter *iter,
 		       struct bkey_s_c k,
-		       struct bch_inode_unpacked *prev,
+		       struct bch_inode_unpacked *snapshot_root,
 		       struct snapshots_seen *s)
 {
 	struct bch_fs *c = trans->c;
@@ -1123,16 +1149,19 @@ static int check_inode(struct btree_trans *trans,
 
 	BUG_ON(bch2_inode_unpack(k, &u));
 
-	if (prev->bi_inum != u.bi_inum)
-		*prev = u;
+	if (snapshot_root->bi_inum != u.bi_inum) {
+		ret = get_snapshot_root_inode(trans, snapshot_root, u.bi_inum);
+		if (ret)
+			goto err;
+	}
 
-	if (fsck_err_on(prev->bi_hash_seed	!= u.bi_hash_seed ||
-			inode_d_type(prev)	!= inode_d_type(&u),
+	if (fsck_err_on(u.bi_hash_seed		!= snapshot_root->bi_hash_seed ||
+			INODE_STR_HASH(&u)	!= INODE_STR_HASH(snapshot_root),
 			trans, inode_snapshot_mismatch,
 			"inodes in different snapshots don't match")) {
-		bch_err(c, "repair not implemented yet");
-		ret = -BCH_ERR_fsck_repair_unimplemented;
-		goto err_noprint;
+		u.bi_hash_seed = snapshot_root->bi_hash_seed;
+		SET_INODE_STR_HASH(&u, INODE_STR_HASH(snapshot_root));
+		do_update = true;
 	}
 
 	if (u.bi_dir || u.bi_dir_offset) {
@@ -1285,7 +1314,7 @@ err_noprint:
 
 int bch2_check_inodes(struct bch_fs *c)
 {
-	struct bch_inode_unpacked prev = { 0 };
+	struct bch_inode_unpacked snapshot_root = {};
 	struct snapshots_seen s;
 
 	snapshots_seen_init(&s);
@@ -1295,7 +1324,7 @@ int bch2_check_inodes(struct bch_fs *c)
 				POS_MIN,
 				BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
-			check_inode(trans, &iter, k, &prev, &s)));
+			check_inode(trans, &iter, k, &snapshot_root, &s)));
 
 	snapshots_seen_exit(&s);
 	bch_err_fn(c, ret);

From dc96656b20eb6f7dea0ccd220541b4af49cf5789 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Fri, 18 Oct 2024 00:19:12 -0400
Subject: [PATCH 20/26] bcachefs: bch2_hash_set_or_get_in_snapshot()

Add a variant of bch2_hash_set_in_snapshot() that returns the existing
key on -EEXIST.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/str_hash.h | 50 ++++++++++++++++++++++++++++++------------
 1 file changed, 36 insertions(+), 14 deletions(-)

diff --git a/fs/bcachefs/str_hash.h b/fs/bcachefs/str_hash.h
index d4b5718860bd..ec2b1feea520 100644
--- a/fs/bcachefs/str_hash.h
+++ b/fs/bcachefs/str_hash.h
@@ -252,19 +252,20 @@ int bch2_hash_needs_whiteout(struct btree_trans *trans,
 }
 
 static __always_inline
-int bch2_hash_set_in_snapshot(struct btree_trans *trans,
+struct bkey_s_c bch2_hash_set_or_get_in_snapshot(struct btree_trans *trans,
+			   struct btree_iter *iter,
 			   const struct bch_hash_desc desc,
 			   const struct bch_hash_info *info,
 			   subvol_inum inum, u32 snapshot,
 			   struct bkey_i *insert,
 			   enum btree_iter_update_trigger_flags flags)
 {
-	struct btree_iter iter, slot = { NULL };
+	struct btree_iter slot = {};
 	struct bkey_s_c k;
 	bool found = false;
 	int ret;
 
-	for_each_btree_key_upto_norestart(trans, iter, desc.btree_id,
+	for_each_btree_key_upto_norestart(trans, *iter, desc.btree_id,
 			   SPOS(insert->k.p.inode,
 				desc.hash_bkey(info, bkey_i_to_s_c(insert)),
 				snapshot),
@@ -279,7 +280,7 @@ int bch2_hash_set_in_snapshot(struct btree_trans *trans,
 		}
 
 		if (!slot.path && !(flags & STR_HASH_must_replace))
-			bch2_trans_copy_iter(&slot, &iter);
+			bch2_trans_copy_iter(&slot, iter);
 
 		if (k.k->type != KEY_TYPE_hash_whiteout)
 			goto not_found;
@@ -289,28 +290,49 @@ int bch2_hash_set_in_snapshot(struct btree_trans *trans,
 		ret = -BCH_ERR_ENOSPC_str_hash_create;
 out:
 	bch2_trans_iter_exit(trans, &slot);
-	bch2_trans_iter_exit(trans, &iter);
-
-	return ret;
+	bch2_trans_iter_exit(trans, iter);
+	return ret ? bkey_s_c_err(ret) : bkey_s_c_null;
 found:
 	found = true;
 not_found:
-
-	if (!found && (flags & STR_HASH_must_replace)) {
+	if (found && (flags & STR_HASH_must_create)) {
+		bch2_trans_iter_exit(trans, &slot);
+		return k;
+	} else if (!found && (flags & STR_HASH_must_replace)) {
 		ret = -BCH_ERR_ENOENT_str_hash_set_must_replace;
-	} else if (found && (flags & STR_HASH_must_create)) {
-		ret = -BCH_ERR_EEXIST_str_hash_set;
 	} else {
 		if (!found && slot.path)
-			swap(iter, slot);
+			swap(*iter, slot);
 
-		insert->k.p = iter.pos;
-		ret = bch2_trans_update(trans, &iter, insert, flags);
+		insert->k.p = iter->pos;
+		ret = bch2_trans_update(trans, iter, insert, flags);
 	}
 
 	goto out;
 }
 
+static __always_inline
+int bch2_hash_set_in_snapshot(struct btree_trans *trans,
+			   const struct bch_hash_desc desc,
+			   const struct bch_hash_info *info,
+			   subvol_inum inum, u32 snapshot,
+			   struct bkey_i *insert,
+			   enum btree_iter_update_trigger_flags flags)
+{
+	struct btree_iter iter;
+	struct bkey_s_c k = bch2_hash_set_or_get_in_snapshot(trans, &iter, desc, info, inum,
+							     snapshot, insert, flags);
+	int ret = bkey_err(k);
+	if (ret)
+		return ret;
+	if (k.k) {
+		bch2_trans_iter_exit(trans, &iter);
+		return -BCH_ERR_EEXIST_str_hash_set;
+	}
+
+	return 0;
+}
+
 static __always_inline
 int bch2_hash_set(struct btree_trans *trans,
 		  const struct bch_hash_desc desc,

From bc6d2d10418e1bfdb95b16f5dd4cca42d5dec766 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Fri, 18 Oct 2024 00:22:09 -0400
Subject: [PATCH 21/26] bcachefs: fsck: Improve hash_check_key()

hash_check_key() checks and repairs the hash table btrees: dirents and
xattrs are open addressing hash tables.

We recently had a corruption reported where the hash type on an inode
somehow got flipped, which made the existing dirents invisible and
allowed new ones to be created with the same name.

Now, hash_check_key() can repair duplicates: it will delete one of them,
if it has an xattr or dangling dirent, but if it has two valid dirents
one of them gets renamed.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/dirent.c |   7 --
 fs/bcachefs/dirent.h |   7 ++
 fs/bcachefs/fsck.c   | 222 +++++++++++++++++++++++++++++++++++--------
 3 files changed, 190 insertions(+), 46 deletions(-)

diff --git a/fs/bcachefs/dirent.c b/fs/bcachefs/dirent.c
index 84dd4a879d98..faffc98d5605 100644
--- a/fs/bcachefs/dirent.c
+++ b/fs/bcachefs/dirent.c
@@ -250,13 +250,6 @@ int bch2_dirent_create(struct btree_trans *trans, subvol_inum dir,
 	return ret;
 }
 
-static void dirent_copy_target(struct bkey_i_dirent *dst,
-			       struct bkey_s_c_dirent src)
-{
-	dst->v.d_inum = src.v->d_inum;
-	dst->v.d_type = src.v->d_type;
-}
-
 int bch2_dirent_read_target(struct btree_trans *trans, subvol_inum dir,
 			    struct bkey_s_c_dirent d, subvol_inum *target)
 {
diff --git a/fs/bcachefs/dirent.h b/fs/bcachefs/dirent.h
index 8945145865c5..53ad99666022 100644
--- a/fs/bcachefs/dirent.h
+++ b/fs/bcachefs/dirent.h
@@ -34,6 +34,13 @@ static inline unsigned dirent_val_u64s(unsigned len)
 int bch2_dirent_read_target(struct btree_trans *, subvol_inum,
 			    struct bkey_s_c_dirent, subvol_inum *);
 
+static inline void dirent_copy_target(struct bkey_i_dirent *dst,
+				      struct bkey_s_c_dirent src)
+{
+	dst->v.d_inum = src.v->d_inum;
+	dst->v.d_type = src.v->d_type;
+}
+
 int bch2_dirent_create_snapshot(struct btree_trans *, u32, u64, u32,
 			const struct bch_hash_info *, u8,
 			const struct qstr *, u64, u64 *,
diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c
index 16a408210bb7..75c8a97a6954 100644
--- a/fs/bcachefs/fsck.c
+++ b/fs/bcachefs/fsck.c
@@ -929,35 +929,138 @@ static int get_visible_inodes(struct btree_trans *trans,
 	return ret;
 }
 
-static int hash_redo_key(struct btree_trans *trans,
-			 const struct bch_hash_desc desc,
-			 struct bch_hash_info *hash_info,
-			 struct btree_iter *k_iter, struct bkey_s_c k)
+static int dirent_has_target(struct btree_trans *trans, struct bkey_s_c_dirent d)
 {
-	struct bkey_i *delete;
-	struct bkey_i *tmp;
+	if (d.v->d_type == DT_SUBVOL) {
+		u32 snap;
+		u64 inum;
+		int ret = subvol_lookup(trans, le32_to_cpu(d.v->d_child_subvol), &snap, &inum);
+		if (ret && !bch2_err_matches(ret, ENOENT))
+			return ret;
+		return !ret;
+	} else {
+		struct btree_iter iter;
+		struct bkey_s_c k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes,
+				SPOS(0, le64_to_cpu(d.v->d_inum), d.k->p.snapshot), 0);
+		int ret = bkey_err(k);
+		if (ret)
+			return ret;
 
-	delete = bch2_trans_kmalloc(trans, sizeof(*delete));
-	if (IS_ERR(delete))
-		return PTR_ERR(delete);
+		ret = bkey_is_inode(k.k);
+		bch2_trans_iter_exit(trans, &iter);
+		return ret;
+	}
+}
 
-	tmp = bch2_bkey_make_mut_noupdate(trans, k);
-	if (IS_ERR(tmp))
-		return PTR_ERR(tmp);
+/*
+ * Prefer to delete the first one, since that will be the one at the wrong
+ * offset:
+ * return value: 0 -> delete k1, 1 -> delete k2
+ */
+static int hash_pick_winner(struct btree_trans *trans,
+			    const struct bch_hash_desc desc,
+			    struct bch_hash_info *hash_info,
+			    struct bkey_s_c k1,
+			    struct bkey_s_c k2)
+{
+	if (bkey_val_bytes(k1.k) == bkey_val_bytes(k2.k) &&
+	    !memcmp(k1.v, k2.v, bkey_val_bytes(k1.k)))
+		return 0;
 
-	bkey_init(&delete->k);
-	delete->k.p = k_iter->pos;
-	return  bch2_btree_iter_traverse(k_iter) ?:
-		bch2_trans_update(trans, k_iter, delete, 0) ?:
-		bch2_hash_set_in_snapshot(trans, desc, hash_info,
-				       (subvol_inum) { 0, k.k->p.inode },
-				       k.k->p.snapshot, tmp,
-				       STR_HASH_must_create|
-				       BTREE_UPDATE_internal_snapshot_node) ?:
-		bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc);
+	switch (desc.btree_id) {
+	case BTREE_ID_dirents: {
+		int ret = dirent_has_target(trans, bkey_s_c_to_dirent(k1));
+		if (ret < 0)
+			return ret;
+		if (!ret)
+			return 0;
+
+		ret = dirent_has_target(trans, bkey_s_c_to_dirent(k2));
+		if (ret < 0)
+			return ret;
+		if (!ret)
+			return 1;
+		return 2;
+	}
+	default:
+		return 0;
+	}
+}
+
+static int fsck_update_backpointers(struct btree_trans *trans,
+				    struct snapshots_seen *s,
+				    const struct bch_hash_desc desc,
+				    struct bch_hash_info *hash_info,
+				    struct bkey_i *new)
+{
+	if (new->k.type != KEY_TYPE_dirent)
+		return 0;
+
+	struct bkey_i_dirent *d = bkey_i_to_dirent(new);
+	struct inode_walker target = inode_walker_init();
+	int ret = 0;
+
+	if (d->v.d_type == DT_SUBVOL) {
+		BUG();
+	} else {
+		ret = get_visible_inodes(trans, &target, s, le64_to_cpu(d->v.d_inum));
+		if (ret)
+			goto err;
+
+		darray_for_each(target.inodes, i) {
+			i->inode.bi_dir_offset = d->k.p.offset;
+			ret = __bch2_fsck_write_inode(trans, &i->inode);
+			if (ret)
+				goto err;
+		}
+	}
+err:
+	inode_walker_exit(&target);
+	return ret;
+}
+
+static int fsck_rename_dirent(struct btree_trans *trans,
+			      struct snapshots_seen *s,
+			      const struct bch_hash_desc desc,
+			      struct bch_hash_info *hash_info,
+			      struct bkey_s_c_dirent old)
+{
+	struct qstr old_name = bch2_dirent_get_name(old);
+	struct bkey_i_dirent *new = bch2_trans_kmalloc(trans, bkey_bytes(old.k) + 32);
+	int ret = PTR_ERR_OR_ZERO(new);
+	if (ret)
+		return ret;
+
+	bkey_dirent_init(&new->k_i);
+	dirent_copy_target(new, old);
+	new->k.p = old.k->p;
+
+	for (unsigned i = 0; i < 1000; i++) {
+		unsigned len = sprintf(new->v.d_name, "%.*s.fsck_renamed-%u",
+				       old_name.len, old_name.name, i);
+		unsigned u64s = BKEY_U64s + dirent_val_u64s(len);
+
+		if (u64s > U8_MAX)
+			return -EINVAL;
+
+		new->k.u64s = u64s;
+
+		ret = bch2_hash_set_in_snapshot(trans, bch2_dirent_hash_desc, hash_info,
+						(subvol_inum) { 0, old.k->p.inode },
+						old.k->p.snapshot, &new->k_i,
+						BTREE_UPDATE_internal_snapshot_node);
+		if (!bch2_err_matches(ret, EEXIST))
+			break;
+	}
+
+	if (ret)
+		return ret;
+
+	return fsck_update_backpointers(trans, s, desc, hash_info, &new->k_i);
 }
 
 static int hash_check_key(struct btree_trans *trans,
+			  struct snapshots_seen *s,
 			  const struct bch_hash_desc desc,
 			  struct bch_hash_info *hash_info,
 			  struct btree_iter *k_iter, struct bkey_s_c hash_k)
@@ -986,16 +1089,9 @@ static int hash_check_key(struct btree_trans *trans,
 		if (bkey_eq(k.k->p, hash_k.k->p))
 			break;
 
-		if (fsck_err_on(k.k->type == desc.key_type &&
-				!desc.cmp_bkey(k, hash_k),
-				trans, hash_table_key_duplicate,
-				"duplicate hash table keys:\n%s",
-				(printbuf_reset(&buf),
-				 bch2_bkey_val_to_text(&buf, c, hash_k),
-				 buf.buf))) {
-			ret = bch2_hash_delete_at(trans, desc, hash_info, k_iter, 0) ?: 1;
-			break;
-		}
+		if (k.k->type == desc.key_type &&
+		    !desc.cmp_bkey(k, hash_k))
+			goto duplicate_entries;
 
 		if (bkey_deleted(k.k)) {
 			bch2_trans_iter_exit(trans, &iter);
@@ -1008,18 +1104,66 @@ out:
 	return ret;
 bad_hash:
 	if (fsck_err(trans, hash_table_key_wrong_offset,
-		     "hash table key at wrong offset: btree %s inode %llu offset %llu, hashed to %llu\n%s",
+		     "hash table key at wrong offset: btree %s inode %llu offset %llu, hashed to %llu\n  %s",
 		     bch2_btree_id_str(desc.btree_id), hash_k.k->p.inode, hash_k.k->p.offset, hash,
 		     (printbuf_reset(&buf),
 		      bch2_bkey_val_to_text(&buf, c, hash_k), buf.buf))) {
-		ret = hash_redo_key(trans, desc, hash_info, k_iter, hash_k);
-		bch_err_fn(c, ret);
+		struct bkey_i *new = bch2_bkey_make_mut_noupdate(trans, hash_k);
+		if (IS_ERR(new))
+			return PTR_ERR(new);
+
+		k = bch2_hash_set_or_get_in_snapshot(trans, &iter, desc, hash_info,
+				       (subvol_inum) { 0, hash_k.k->p.inode },
+				       hash_k.k->p.snapshot, new,
+				       STR_HASH_must_create|
+				       BTREE_ITER_with_updates|
+				       BTREE_UPDATE_internal_snapshot_node);
+		ret = bkey_err(k);
 		if (ret)
-			return ret;
-		ret = -BCH_ERR_transaction_restart_nested;
+			goto out;
+		if (k.k)
+			goto duplicate_entries;
+
+		ret =   bch2_hash_delete_at(trans, desc, hash_info, k_iter,
+					    BTREE_UPDATE_internal_snapshot_node) ?:
+			fsck_update_backpointers(trans, s, desc, hash_info, new) ?:
+			bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc) ?:
+			-BCH_ERR_transaction_restart_nested;
+		goto out;
 	}
 fsck_err:
 	goto out;
+duplicate_entries:
+	ret = hash_pick_winner(trans, desc, hash_info, hash_k, k);
+	if (ret < 0)
+		goto out;
+
+	if (!fsck_err(trans, hash_table_key_duplicate,
+		      "duplicate hash table keys%s:\n%s",
+		      ret != 2 ? "" : ", both point to valid inodes",
+		      (printbuf_reset(&buf),
+		       bch2_bkey_val_to_text(&buf, c, hash_k),
+		       prt_newline(&buf),
+		       bch2_bkey_val_to_text(&buf, c, k),
+		       buf.buf)))
+		goto out;
+
+	switch (ret) {
+	case 0:
+		ret = bch2_hash_delete_at(trans, desc, hash_info, k_iter, 0);
+		break;
+	case 1:
+		ret = bch2_hash_delete_at(trans, desc, hash_info, &iter, 0);
+		break;
+	case 2:
+		ret = fsck_rename_dirent(trans, s, desc, hash_info, bkey_s_c_to_dirent(hash_k)) ?:
+			bch2_hash_delete_at(trans, desc, hash_info, k_iter, 0);
+		goto out;
+	}
+
+	ret = bch2_trans_commit(trans, NULL, NULL, 0) ?:
+		-BCH_ERR_transaction_restart_nested;
+	goto out;
 }
 
 static struct bkey_s_c_dirent dirent_get_by_pos(struct btree_trans *trans,
@@ -2336,7 +2480,7 @@ static int check_dirent(struct btree_trans *trans, struct btree_iter *iter,
 		*hash_info = bch2_hash_info_init(c, &i->inode);
 	dir->first_this_inode = false;
 
-	ret = hash_check_key(trans, bch2_dirent_hash_desc, hash_info, iter, k);
+	ret = hash_check_key(trans, s, bch2_dirent_hash_desc, hash_info, iter, k);
 	if (ret < 0)
 		goto err;
 	if (ret) {
@@ -2450,7 +2594,7 @@ static int check_xattr(struct btree_trans *trans, struct btree_iter *iter,
 		*hash_info = bch2_hash_info_init(c, &i->inode);
 	inode->first_this_inode = false;
 
-	ret = hash_check_key(trans, bch2_xattr_hash_desc, hash_info, iter, k);
+	ret = hash_check_key(trans, NULL, bch2_xattr_hash_desc, hash_info, iter, k);
 	bch_err_fn(c, ret);
 	return ret;
 }

From eb5db64c4570948e6ee0b0f53d658e136e06cd04 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sat, 19 Oct 2024 17:23:10 -0400
Subject: [PATCH 22/26] bcachefs: Fix __bch2_fsck_err() warning

We only warn about having a btree_trans that wasn't passed in if we'll
be prompting.

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

diff --git a/fs/bcachefs/error.c b/fs/bcachefs/error.c
index 7a79f695ba2e..b679def8fb98 100644
--- a/fs/bcachefs/error.c
+++ b/fs/bcachefs/error.c
@@ -251,7 +251,10 @@ int __bch2_fsck_err(struct bch_fs *c,
 	 *   delete the key)
 	 * - and we don't need to warn if we're not prompting
 	 */
-	WARN_ON(!(flags & FSCK_AUTOFIX) && !trans && bch2_current_has_btree_trans(c));
+	WARN_ON((flags & FSCK_CAN_FIX) &&
+		!(flags & FSCK_AUTOFIX) &&
+		!trans &&
+		bch2_current_has_btree_trans(c));
 
 	if ((flags & FSCK_CAN_FIX) &&
 	    test_bit(err, c->sb.errors_silent))

From 3956ff8bc2f39a7e77d7a6da8d95c7ffc4928d64 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sat, 19 Oct 2024 17:50:41 -0400
Subject: [PATCH 23/26] bcachefs: Don't use wait_event_interruptible() in
 recovery

Fix a bug where mount was failing with -ERESTARTSYS:
https://github.com/koverstreet/bcachefs/issues/741

We only want the interruptible wait when called from fsync.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/fs-io.c   |  2 +-
 fs/bcachefs/journal.c | 10 ++++++----
 fs/bcachefs/journal.h |  2 +-
 3 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/fs/bcachefs/fs-io.c b/fs/bcachefs/fs-io.c
index 71d0fa387509..15d3f073b824 100644
--- a/fs/bcachefs/fs-io.c
+++ b/fs/bcachefs/fs-io.c
@@ -182,7 +182,7 @@ static int bch2_flush_inode(struct bch_fs *c,
 
 	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_journal_flush_seq(&c->journal, u.bi_journal_seq, TASK_INTERRUPTIBLE) ?:
 		  bch2_inode_flush_nocow_writes(c, inode);
 	bch2_write_ref_put(c, BCH_WRITE_REF_fsync);
 	return ret;
diff --git a/fs/bcachefs/journal.c b/fs/bcachefs/journal.c
index dc099f06341f..2dc0d60c1745 100644
--- a/fs/bcachefs/journal.c
+++ b/fs/bcachefs/journal.c
@@ -758,7 +758,7 @@ out:
 	return ret;
 }
 
-int bch2_journal_flush_seq(struct journal *j, u64 seq)
+int bch2_journal_flush_seq(struct journal *j, u64 seq, unsigned task_state)
 {
 	u64 start_time = local_clock();
 	int ret, ret2;
@@ -769,7 +769,9 @@ int bch2_journal_flush_seq(struct journal *j, u64 seq)
 	if (seq <= j->flushed_seq_ondisk)
 		return 0;
 
-	ret = wait_event_interruptible(j->wait, (ret2 = bch2_journal_flush_seq_async(j, seq, NULL)));
+	ret = wait_event_state(j->wait,
+			       (ret2 = bch2_journal_flush_seq_async(j, seq, NULL)),
+			       task_state);
 
 	if (!ret)
 		bch2_time_stats_update(j->flush_seq_time, start_time);
@@ -788,7 +790,7 @@ void bch2_journal_flush_async(struct journal *j, struct closure *parent)
 
 int bch2_journal_flush(struct journal *j)
 {
-	return bch2_journal_flush_seq(j, atomic64_read(&j->seq));
+	return bch2_journal_flush_seq(j, atomic64_read(&j->seq), TASK_UNINTERRUPTIBLE);
 }
 
 /*
@@ -851,7 +853,7 @@ int bch2_journal_meta(struct journal *j)
 
 	bch2_journal_res_put(j, &res);
 
-	return bch2_journal_flush_seq(j, res.seq);
+	return bch2_journal_flush_seq(j, res.seq, TASK_UNINTERRUPTIBLE);
 }
 
 /* block/unlock the journal: */
diff --git a/fs/bcachefs/journal.h b/fs/bcachefs/journal.h
index 377a3750406e..2762be6f9814 100644
--- a/fs/bcachefs/journal.h
+++ b/fs/bcachefs/journal.h
@@ -401,7 +401,7 @@ void bch2_journal_entry_res_resize(struct journal *,
 int bch2_journal_flush_seq_async(struct journal *, u64, struct closure *);
 void bch2_journal_flush_async(struct journal *, struct closure *);
 
-int bch2_journal_flush_seq(struct journal *, u64);
+int bch2_journal_flush_seq(struct journal *, u64, unsigned);
 int bch2_journal_flush(struct journal *);
 bool bch2_journal_noflush_seq(struct journal *, u64);
 int bch2_journal_meta(struct journal *);

From f0d3302073e60b55318c941cd10b225a3cde9d32 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sat, 19 Oct 2024 18:27:09 -0400
Subject: [PATCH 24/26] bcachefs: Workaround for kvmalloc() not supporting >
 INT_MAX allocations

kvmalloc() doesn't support allocations > INT_MAX, but vmalloc() does -
the limit should be lifted, but we can work around this for now.

A user with a 75 TB filesystem reported the following journal replay
error:
https://github.com/koverstreet/bcachefs/issues/769

In journal replay we have to sort and dedup all the keys from the
journal, which means we need a large contiguous allocation. Given that
the user has 128GB of ram, the 2GB limit on allocation size has become
far too small.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/darray.c | 15 ++++++++++++++-
 1 file changed, 14 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/darray.c b/fs/bcachefs/darray.c
index 4f06cd8bbbe1..e86d36d23e9e 100644
--- a/fs/bcachefs/darray.c
+++ b/fs/bcachefs/darray.c
@@ -2,6 +2,7 @@
 
 #include <linux/log2.h>
 #include <linux/slab.h>
+#include <linux/vmalloc.h>
 #include "darray.h"
 
 int __bch2_darray_resize_noprof(darray_char *d, size_t element_size, size_t new_size, gfp_t gfp)
@@ -9,7 +10,19 @@ int __bch2_darray_resize_noprof(darray_char *d, size_t element_size, size_t new_
 	if (new_size > d->size) {
 		new_size = roundup_pow_of_two(new_size);
 
-		void *data = kvmalloc_array_noprof(new_size, element_size, gfp);
+		/*
+		 * This is a workaround: kvmalloc() doesn't support > INT_MAX
+		 * allocations, but vmalloc() does.
+		 * The limit needs to be lifted from kvmalloc, and when it does
+		 * we'll go back to just using that.
+		 */
+		size_t bytes;
+		if (unlikely(check_mul_overflow(new_size, element_size, &bytes)))
+			return -ENOMEM;
+
+		void *data = likely(bytes < INT_MAX)
+			? kvmalloc_noprof(bytes, gfp)
+			: vmalloc_noprof(bytes);
 		if (!data)
 			return -ENOMEM;
 

From e04ee8608914d00812e4a10288cf8d8905ced849 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sun, 20 Oct 2024 16:48:31 -0400
Subject: [PATCH 25/26] bcachefs: Mark more errors as AUTOFIX
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Reported-by: Marcin Mirosław <marcin@mejor.pl>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/sb-errors_format.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/fs/bcachefs/sb-errors_format.h b/fs/bcachefs/sb-errors_format.h
index aab328ac6dfa..937275d061fe 100644
--- a/fs/bcachefs/sb-errors_format.h
+++ b/fs/bcachefs/sb-errors_format.h
@@ -267,8 +267,8 @@ enum bch_fsck_flags {
 	x(journal_entry_dup_same_device,			246,	0)		\
 	x(inode_bi_subvol_missing,				247,	0)		\
 	x(inode_bi_subvol_wrong,				248,	0)		\
-	x(inode_points_to_missing_dirent,			249,	0)		\
-	x(inode_points_to_wrong_dirent,				250,	0)		\
+	x(inode_points_to_missing_dirent,			249,	FSCK_AUTOFIX)	\
+	x(inode_points_to_wrong_dirent,				250,	FSCK_AUTOFIX)	\
 	x(inode_bi_parent_nonzero,				251,	0)		\
 	x(dirent_to_missing_parent_subvol,			252,	0)		\
 	x(dirent_not_visible_in_parent_subvol,			253,	0)		\

From a069f014797fdef8757f3adebc1c16416271a599 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sun, 20 Oct 2024 18:00:13 -0400
Subject: [PATCH 26/26] bcachefs: Set bch_inode_unpacked.bi_snapshot in old
 inode path
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

This fixes a fsck bug on a very old filesystem (pre mainline merge).

Fixes: 72350ee0ea22 ("bcachefs: Kill snapshot arg to fsck_write_inode()")
Reported-by: Marcin Mirosław <marcin@mejor.pl>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/inode.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/fs/bcachefs/inode.c b/fs/bcachefs/inode.c
index b9fde1566df1..6b7143e6abca 100644
--- a/fs/bcachefs/inode.c
+++ b/fs/bcachefs/inode.c
@@ -283,6 +283,8 @@ static noinline int bch2_inode_unpack_slowpath(struct bkey_s_c k,
 {
 	memset(unpacked, 0, sizeof(*unpacked));
 
+	unpacked->bi_snapshot = k.k->p.snapshot;
+
 	switch (k.k->type) {
 	case KEY_TYPE_inode: {
 		struct bkey_s_c_inode inode = bkey_s_c_to_inode(k);