Merge tag 'bcachefs-2024-01-26' of https://evilpiepirate.org/git/bcachefs

Pull bcachefs fixes from Kent Overstreet:

 - fix for REQ_OP_FLUSH usage; this fixes filesystems going read only
   with -EOPNOTSUPP from the block layer.

   (this really should have gone in with the block layer patch causing
   the -EOPNOTSUPP, or should have gone in before).

 - fix an allocation in non-sleepable context

 - fix one source of srcu lock latency, on devices with terrible discard
   latency

 - fix a reattach_inode() issue in fsck

* tag 'bcachefs-2024-01-26' of https://evilpiepirate.org/git/bcachefs:
  bcachefs: __lookup_dirent() works in snapshot, not subvol
  bcachefs: discard path uses unlock_long()
  bcachefs: fix incorrect usage of REQ_OP_FLUSH
  bcachefs: Add gfp flags param to bch2_prt_task_backtrace()

10 files changed, 42 insertions, 32 deletions

diff --git a/fs/bcachefs/alloc_background.c b/fs/bcachefs/alloc_background.c
index 10704f2d3af5..fd3e175d8342 100644
--- a/fs/bcachefs/alloc_background.c
+++ b/fs/bcachefs/alloc_background.c
@@ -1715,7 +1715,7 @@ static int bch2_discard_one_bucket(struct btree_trans *trans,
 		 * This works without any other locks because this is the only
 		 * thread that removes items from the need_discard tree
 		 */
-		bch2_trans_unlock(trans);
+		bch2_trans_unlock_long(trans);
 		blkdev_issue_discard(ca->disk_sb.bdev,
 				     k.k->p.offset * ca->mi.bucket_size,
 				     ca->mi.bucket_size,
diff --git a/fs/bcachefs/btree_locking.c b/fs/bcachefs/btree_locking.c
index bed75c93c069..684397442338 100644
--- a/fs/bcachefs/btree_locking.c
+++ b/fs/bcachefs/btree_locking.c
@@ -92,7 +92,7 @@ static noinline void print_cycle(struct printbuf *out, struct lock_graph *g)
 			continue;
 
 		bch2_btree_trans_to_text(out, i->trans);
-		bch2_prt_task_backtrace(out, task, i == g->g ? 5 : 1);
+		bch2_prt_task_backtrace(out, task, i == g->g ? 5 : 1, GFP_NOWAIT);
 	}
 }
 
@@ -227,7 +227,7 @@ static noinline int break_cycle(struct lock_graph *g, struct printbuf *cycle)
 			prt_printf(&buf, "backtrace:");
 			prt_newline(&buf);
 			printbuf_indent_add(&buf, 2);
-			bch2_prt_task_backtrace(&buf, trans->locking_wait.task, 2);
+			bch2_prt_task_backtrace(&buf, trans->locking_wait.task, 2, GFP_NOWAIT);
 			printbuf_indent_sub(&buf, 2);
 			prt_newline(&buf);
 		}
diff --git a/fs/bcachefs/debug.c b/fs/bcachefs/debug.c
index cadda9bbe4a4..7bdba8507fc9 100644
--- a/fs/bcachefs/debug.c
+++ b/fs/bcachefs/debug.c
@@ -627,7 +627,7 @@ restart:
 		prt_printf(&i->buf, "backtrace:");
 		prt_newline(&i->buf);
 		printbuf_indent_add(&i->buf, 2);
-		bch2_prt_task_backtrace(&i->buf, task, 0);
+		bch2_prt_task_backtrace(&i->buf, task, 0, GFP_KERNEL);
 		printbuf_indent_sub(&i->buf, 2);
 		prt_newline(&i->buf);
 
diff --git a/fs/bcachefs/fs-io.c b/fs/bcachefs/fs-io.c
index dc52918d06ef..8c70123b6a0c 100644
--- a/fs/bcachefs/fs-io.c
+++ b/fs/bcachefs/fs-io.c
@@ -79,7 +79,7 @@ void bch2_inode_flush_nocow_writes_async(struct bch_fs *c,
 			continue;
 
 		bio = container_of(bio_alloc_bioset(ca->disk_sb.bdev, 0,
-						    REQ_OP_FLUSH,
+						    REQ_OP_WRITE|REQ_PREFLUSH,
 						    GFP_KERNEL,
 						    &c->nocow_flush_bioset),
 				   struct nocow_flush, bio);
diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c
index 4f0ecd605675..6a760777bafb 100644
--- a/fs/bcachefs/fsck.c
+++ b/fs/bcachefs/fsck.c
@@ -119,22 +119,19 @@ static int lookup_inode(struct btree_trans *trans, u64 inode_nr,
 	if (!ret)
 		*snapshot = iter.pos.snapshot;
 err:
-	bch_err_msg(trans->c, ret, "fetching inode %llu:%u", inode_nr, *snapshot);
 	bch2_trans_iter_exit(trans, &iter);
 	return ret;
 }
 
-static int __lookup_dirent(struct btree_trans *trans,
+static int lookup_dirent_in_snapshot(struct btree_trans *trans,
 			   struct bch_hash_info hash_info,
 			   subvol_inum dir, struct qstr *name,
-			   u64 *target, unsigned *type)
+			   u64 *target, unsigned *type, u32 snapshot)
 {
 	struct btree_iter iter;
 	struct bkey_s_c_dirent d;
-	int ret;
-
-	ret = bch2_hash_lookup(trans, &iter, bch2_dirent_hash_desc,
-			       &hash_info, dir, name, 0);
+	int ret = bch2_hash_lookup_in_snapshot(trans, &iter, bch2_dirent_hash_desc,
+			       &hash_info, dir, name, 0, snapshot);
 	if (ret)
 		return ret;
 
@@ -225,15 +222,16 @@ static int lookup_lostfound(struct btree_trans *trans, u32 snapshot,
 
 	struct bch_inode_unpacked root_inode;
 	struct bch_hash_info root_hash_info;
-	ret = lookup_inode(trans, root_inum.inum, &root_inode, &snapshot);
+	u32 root_inode_snapshot = snapshot;
+	ret = lookup_inode(trans, root_inum.inum, &root_inode, &root_inode_snapshot);
 	bch_err_msg(c, ret, "looking up root inode");
 	if (ret)
 		return ret;
 
 	root_hash_info = bch2_hash_info_init(c, &root_inode);
 
-	ret = __lookup_dirent(trans, root_hash_info, root_inum,
-			      &lostfound_str, &inum, &d_type);
+	ret = lookup_dirent_in_snapshot(trans, root_hash_info, root_inum,
+			      &lostfound_str, &inum, &d_type, snapshot);
 	if (bch2_err_matches(ret, ENOENT))
 		goto create_lostfound;
 
@@ -250,7 +248,10 @@ static int lookup_lostfound(struct btree_trans *trans, u32 snapshot,
 	 * The bch2_check_dirents pass has already run, dangling dirents
 	 * shouldn't exist here:
 	 */
-	return lookup_inode(trans, inum, lostfound, &snapshot);
+	ret = lookup_inode(trans, inum, lostfound, &snapshot);
+	bch_err_msg(c, ret, "looking up lost+found %llu:%u in (root inode %llu, snapshot root %u)",
+		    inum, snapshot, root_inum.inum, bch2_snapshot_root(c, snapshot));
+	return ret;
 
 create_lostfound:
 	/*
diff --git a/fs/bcachefs/journal.c b/fs/bcachefs/journal.c
index d71d26e39521..bc890776eb57 100644
--- a/fs/bcachefs/journal.c
+++ b/fs/bcachefs/journal.c
@@ -233,7 +233,7 @@ static void __journal_entry_close(struct journal *j, unsigned closed_val, bool t
 		prt_str(&pbuf, "entry size: ");
 		prt_human_readable_u64(&pbuf, vstruct_bytes(buf->data));
 		prt_newline(&pbuf);
-		bch2_prt_task_backtrace(&pbuf, current, 1);
+		bch2_prt_task_backtrace(&pbuf, current, 1, GFP_NOWAIT);
 		trace_journal_entry_close(c, pbuf.buf);
 		printbuf_exit(&pbuf);
 	}
diff --git a/fs/bcachefs/journal_io.c b/fs/bcachefs/journal_io.c
index 04a1e79a5ed3..bfd6585e746d 100644
--- a/fs/bcachefs/journal_io.c
+++ b/fs/bcachefs/journal_io.c
@@ -1988,7 +1988,8 @@ CLOSURE_CALLBACK(bch2_journal_write)
 			percpu_ref_get(&ca->io_ref);
 
 			bio = ca->journal.bio;
-			bio_reset(bio, ca->disk_sb.bdev, REQ_OP_FLUSH);
+			bio_reset(bio, ca->disk_sb.bdev,
+				  REQ_OP_WRITE|REQ_PREFLUSH);
 			bio->bi_end_io		= journal_write_endio;
 			bio->bi_private		= ca;
 			closure_bio_submit(bio, cl);
diff --git a/fs/bcachefs/str_hash.h b/fs/bcachefs/str_hash.h
index 89fdb7c21134..fcaa5a888744 100644
--- a/fs/bcachefs/str_hash.h
+++ b/fs/bcachefs/str_hash.h
@@ -160,21 +160,16 @@ static inline bool is_visible_key(struct bch_hash_desc desc, subvol_inum inum, s
 }
 
 static __always_inline int
-bch2_hash_lookup(struct btree_trans *trans,
+bch2_hash_lookup_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, const void *key,
-		 unsigned flags)
+		 unsigned flags, u32 snapshot)
 {
 	struct bkey_s_c k;
-	u32 snapshot;
 	int ret;
 
-	ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot);
-	if (ret)
-		return ret;
-
 	for_each_btree_key_upto_norestart(trans, *iter, desc.btree_id,
 			   SPOS(inum.inum, desc.hash_key(info, key), snapshot),
 			   POS(inum.inum, U64_MAX),
@@ -195,6 +190,19 @@ bch2_hash_lookup(struct btree_trans *trans,
 }
 
 static __always_inline int
+bch2_hash_lookup(struct btree_trans *trans,
+		 struct btree_iter *iter,
+		 const struct bch_hash_desc desc,
+		 const struct bch_hash_info *info,
+		 subvol_inum inum, const void *key,
+		 unsigned flags)
+{
+	u32 snapshot;
+	return  bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot) ?:
+		bch2_hash_lookup_in_snapshot(trans, iter, desc, info, inum, key, flags, snapshot);
+}
+
+static __always_inline int
 bch2_hash_hole(struct btree_trans *trans,
 	       struct btree_iter *iter,
 	       const struct bch_hash_desc desc,
diff --git a/fs/bcachefs/util.c b/fs/bcachefs/util.c
index a135136adeee..56b815fd9fc6 100644
--- a/fs/bcachefs/util.c
+++ b/fs/bcachefs/util.c
@@ -272,14 +272,14 @@ void bch2_print_string_as_lines(const char *prefix, const char *lines)
 	console_unlock();
 }
 
-int bch2_save_backtrace(bch_stacktrace *stack, struct task_struct *task, unsigned skipnr)
+int bch2_save_backtrace(bch_stacktrace *stack, struct task_struct *task, unsigned skipnr,
+			gfp_t gfp)
 {
 #ifdef CONFIG_STACKTRACE
 	unsigned nr_entries = 0;
-	int ret = 0;
 
 	stack->nr = 0;
-	ret = darray_make_room(stack, 32);
+	int ret = darray_make_room_gfp(stack, 32, gfp);
 	if (ret)
 		return ret;
 
@@ -308,10 +308,10 @@ void bch2_prt_backtrace(struct printbuf *out, bch_stacktrace *stack)
 	}
 }
 
-int bch2_prt_task_backtrace(struct printbuf *out, struct task_struct *task, unsigned skipnr)
+int bch2_prt_task_backtrace(struct printbuf *out, struct task_struct *task, unsigned skipnr, gfp_t gfp)
 {
 	bch_stacktrace stack = { 0 };
-	int ret = bch2_save_backtrace(&stack, task, skipnr + 1);
+	int ret = bch2_save_backtrace(&stack, task, skipnr + 1, gfp);
 
 	bch2_prt_backtrace(out, &stack);
 	darray_exit(&stack);
diff --git a/fs/bcachefs/util.h b/fs/bcachefs/util.h
index df67bf55fe2b..b414736d59a5 100644
--- a/fs/bcachefs/util.h
+++ b/fs/bcachefs/util.h
@@ -348,9 +348,9 @@ void bch2_prt_u64_base2(struct printbuf *, u64);
 void bch2_print_string_as_lines(const char *prefix, const char *lines);
 
 typedef DARRAY(unsigned long) bch_stacktrace;
-int bch2_save_backtrace(bch_stacktrace *stack, struct task_struct *, unsigned);
+int bch2_save_backtrace(bch_stacktrace *stack, struct task_struct *, unsigned, gfp_t);
 void bch2_prt_backtrace(struct printbuf *, bch_stacktrace *);
-int bch2_prt_task_backtrace(struct printbuf *, struct task_struct *, unsigned);
+int bch2_prt_task_backtrace(struct printbuf *, struct task_struct *, unsigned, gfp_t);
 
 static inline void prt_bdevname(struct printbuf *out, struct block_device *bdev)
 {