Merge tag 'bcachefs-2024-07-10' of https://evilpiepirate.org/git/bcachefs

Pull bcachefs fixes from Kent Overstreet:

 - Switch some asserts to WARN()

 - Fix a few "transaction not locked" asserts in the data read retry
   paths and backpointers gc

 - Fix a race that would cause the journal to get stuck on a flush
   commit

 - Add missing fsck checks for the fragmentation LRU

 - The usual assorted ssorted syzbot fixes

* tag 'bcachefs-2024-07-10' of https://evilpiepirate.org/git/bcachefs: (22 commits)
  bcachefs: Add missing bch2_trans_begin()
  bcachefs: Fix missing error check in journal_entry_btree_keys_validate()
  bcachefs: Warn on attempting a move with no replicas
  bcachefs: bch2_data_update_to_text()
  bcachefs: Log mount failure error code
  bcachefs: Fix undefined behaviour in eytzinger1_first()
  bcachefs: Mark bch_inode_info as SLAB_ACCOUNT
  bcachefs: Fix bch2_inode_insert() race path for tmpfiles
  closures: fix closure_sync + closure debugging
  bcachefs: Fix journal getting stuck on a flush commit
  bcachefs: io clock: run timer fns under clock lock
  bcachefs: Repair fragmentation_lru in alloc_write_key()
  bcachefs: add check for missing fragmentation in check_alloc_to_lru_ref()
  bcachefs: bch2_btree_write_buffer_maybe_flush()
  bcachefs: Add missing printbuf_tabstops_reset() calls
  bcachefs: Fix loop restart in bch2_btree_transactions_read()
  bcachefs: Fix bch2_read_retry_nodecode()
  bcachefs: Don't use the new_fs() bucket alloc path on an initialized fs
  bcachefs: Fix shift greater than integer size
  bcachefs: Change bch2_fs_journal_stop() BUG_ON() to warning
  ...

Author: torvalds

fs/bcachefs/alloc_background.c

@@ -3,6 +3,7 @@
 #include "alloc_background.h"
 #include "alloc_foreground.h"
 #include "backpointers.h"
+#include "bkey_buf.h"
 #include "btree_cache.h"
 #include "btree_io.h"
 #include "btree_key_cache.h"
@@ -1553,13 +1554,13 @@ int bch2_check_alloc_info(struct bch_fs *c)
 }
 
 static int bch2_check_alloc_to_lru_ref(struct btree_trans *trans,
-				       struct btree_iter *alloc_iter)
+				       struct btree_iter *alloc_iter,
+				       struct bkey_buf *last_flushed)
 {
 	struct bch_fs *c = trans->c;
-	struct btree_iter lru_iter;
 	struct bch_alloc_v4 a_convert;
 	const struct bch_alloc_v4 *a;
-	struct bkey_s_c alloc_k, lru_k;
+	struct bkey_s_c alloc_k;
 	struct printbuf buf = PRINTBUF;
 	int ret;
 
@@ -1573,6 +1574,14 @@ static int bch2_check_alloc_to_lru_ref(struct btree_trans *trans,
 
 	a = bch2_alloc_to_v4(alloc_k, &a_convert);
 
+	if (a->fragmentation_lru) {
+		ret = bch2_lru_check_set(trans, BCH_LRU_FRAGMENTATION_START,
+					 a->fragmentation_lru,
+					 alloc_k, last_flushed);
+		if (ret)
+			return ret;
+	}
+
 	if (a->data_type != BCH_DATA_cached)
 		return 0;
 
@@ -1597,41 +1606,30 @@ static int bch2_check_alloc_to_lru_ref(struct btree_trans *trans,
 		a = &a_mut->v;
 	}
 
-	lru_k = bch2_bkey_get_iter(trans, &lru_iter, BTREE_ID_lru,
-			     lru_pos(alloc_k.k->p.inode,
-				     bucket_to_u64(alloc_k.k->p),
-				     a->io_time[READ]), 0);
-	ret = bkey_err(lru_k);
+	ret = bch2_lru_check_set(trans, alloc_k.k->p.inode, a->io_time[READ],
+				 alloc_k, last_flushed);
 	if (ret)
-		return ret;
-
-	if (fsck_err_on(lru_k.k->type != KEY_TYPE_set, c,
-			alloc_key_to_missing_lru_entry,
-			"missing lru entry\n"
-			"  %s",
-			(printbuf_reset(&buf),
-			 bch2_bkey_val_to_text(&buf, c, alloc_k), buf.buf))) {
-		ret = bch2_lru_set(trans,
-				   alloc_k.k->p.inode,
-				   bucket_to_u64(alloc_k.k->p),
-				   a->io_time[READ]);
-		if (ret)
-			goto err;
-	}
+		goto err;
 err:
 fsck_err:
-	bch2_trans_iter_exit(trans, &lru_iter);
 	printbuf_exit(&buf);
 	return ret;
 }
 
 int bch2_check_alloc_to_lru_refs(struct bch_fs *c)
 {
+	struct bkey_buf last_flushed;
+
+	bch2_bkey_buf_init(&last_flushed);
+	bkey_init(&last_flushed.k->k);
+
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter, BTREE_ID_alloc,
 				POS_MIN, BTREE_ITER_prefetch, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
-			bch2_check_alloc_to_lru_ref(trans, &iter)));
+			bch2_check_alloc_to_lru_ref(trans, &iter, &last_flushed)));
+
+	bch2_bkey_buf_exit(&last_flushed, c);
 	bch_err_fn(c, ret);
 	return ret;
 }

fs/bcachefs/alloc_foreground.c

@@ -1703,6 +1703,7 @@ void bch2_fs_alloc_debug_to_text(struct printbuf *out, struct bch_fs *c)
 	for (unsigned i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
 		nr[c->open_buckets[i].data_type]++;
 
+	printbuf_tabstops_reset(out);
 	printbuf_tabstop_push(out, 24);
 
 	percpu_down_read(&c->mark_lock);
@@ -1736,6 +1737,7 @@ void bch2_dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
 	for (unsigned i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
 		nr[c->open_buckets[i].data_type]++;
 
+	printbuf_tabstops_reset(out);
 	printbuf_tabstop_push(out, 12);
 	printbuf_tabstop_push(out, 16);
 	printbuf_tabstop_push(out, 16);

fs/bcachefs/backpointers.c

@@ -434,13 +434,6 @@ int bch2_check_btree_backpointers(struct bch_fs *c)
 	return ret;
 }
 
-static inline bool bkey_and_val_eq(struct bkey_s_c l, struct bkey_s_c r)
-{
-	return bpos_eq(l.k->p, r.k->p) &&
-		bkey_bytes(l.k) == bkey_bytes(r.k) &&
-		!memcmp(l.v, r.v, bkey_val_bytes(l.k));
-}
-
 struct extents_to_bp_state {
 	struct bpos	bucket_start;
 	struct bpos	bucket_end;
@@ -536,11 +529,8 @@ static int check_bp_exists(struct btree_trans *trans,
 	struct btree_iter other_extent_iter = {};
 	struct printbuf buf = PRINTBUF;
 	struct bkey_s_c bp_k;
-	struct bkey_buf tmp;
 	int ret = 0;
 
-	bch2_bkey_buf_init(&tmp);
-
 	struct bch_dev *ca = bch2_dev_bucket_tryget(c, bucket);
 	if (!ca) {
 		prt_str(&buf, "extent for nonexistent device:bucket ");
@@ -565,22 +555,9 @@ static int check_bp_exists(struct btree_trans *trans,
 
 	if (bp_k.k->type != KEY_TYPE_backpointer ||
 	    memcmp(bkey_s_c_to_backpointer(bp_k).v, &bp, sizeof(bp))) {
-		bch2_bkey_buf_reassemble(&tmp, c, orig_k);
-
-		if (!bkey_and_val_eq(orig_k, bkey_i_to_s_c(s->last_flushed.k))) {
-			if (bp.level) {
-				bch2_trans_unlock(trans);
-				bch2_btree_interior_updates_flush(c);
-			}
-
-			ret = bch2_btree_write_buffer_flush_sync(trans);
-			if (ret)
-				goto err;
-
-			bch2_bkey_buf_copy(&s->last_flushed, c, tmp.k);
-			ret = -BCH_ERR_transaction_restart_write_buffer_flush;
-			goto out;
-		}
+		ret = bch2_btree_write_buffer_maybe_flush(trans, orig_k, &s->last_flushed);
+		if (ret)
+			goto err;
 
 		goto check_existing_bp;
 	}
@@ -589,7 +566,6 @@ static int check_bp_exists(struct btree_trans *trans,
 fsck_err:
 	bch2_trans_iter_exit(trans, &other_extent_iter);
 	bch2_trans_iter_exit(trans, &bp_iter);
-	bch2_bkey_buf_exit(&tmp, c);
 	bch2_dev_put(ca);
 	printbuf_exit(&buf);
 	return ret;
@@ -794,6 +770,8 @@ static int bch2_get_btree_in_memory_pos(struct btree_trans *trans,
 		    !((1U << btree) & btree_interior_mask))
 			continue;
 
+		bch2_trans_begin(trans);
+
 		__for_each_btree_node(trans, iter, btree,
 				      btree == start.btree ? start.pos : POS_MIN,
 				      0, depth, BTREE_ITER_prefetch, b, ret) {
@@ -905,7 +883,7 @@ static int check_one_backpointer(struct btree_trans *trans,
 				 struct bbpos start,
 				 struct bbpos end,
 				 struct bkey_s_c_backpointer bp,
-				 struct bpos *last_flushed_pos)
+				 struct bkey_buf *last_flushed)
 {
 	struct bch_fs *c = trans->c;
 	struct btree_iter iter;
@@ -925,20 +903,18 @@ static int check_one_backpointer(struct btree_trans *trans,
 	if (ret)
 		return ret;
 
-	if (!k.k && !bpos_eq(*last_flushed_pos, bp.k->p)) {
-		*last_flushed_pos = bp.k->p;
-		ret = bch2_btree_write_buffer_flush_sync(trans) ?:
-			-BCH_ERR_transaction_restart_write_buffer_flush;
-		goto out;
-	}
+	if (!k.k) {
+		ret = bch2_btree_write_buffer_maybe_flush(trans, bp.s_c, last_flushed);
+		if (ret)
+			goto out;
 
-	if (fsck_err_on(!k.k, c,
-			backpointer_to_missing_ptr,
-			"backpointer for missing %s\n  %s",
-			bp.v->level ? "btree node" : "extent",
-			(bch2_bkey_val_to_text(&buf, c, bp.s_c), buf.buf))) {
-		ret = bch2_btree_delete_at_buffered(trans, BTREE_ID_backpointers, bp.k->p);
-		goto out;
+		if (fsck_err(c, backpointer_to_missing_ptr,
+			     "backpointer for missing %s\n  %s",
+			     bp.v->level ? "btree node" : "extent",
+			     (bch2_bkey_val_to_text(&buf, c, bp.s_c), buf.buf))) {
+			ret = bch2_btree_delete_at_buffered(trans, BTREE_ID_backpointers, bp.k->p);
+			goto out;
+		}
 	}
 out:
 fsck_err:
@@ -951,14 +927,20 @@ static int bch2_check_backpointers_to_extents_pass(struct btree_trans *trans,
 						   struct bbpos start,
 						   struct bbpos end)
 {
-	struct bpos last_flushed_pos = SPOS_MAX;
+	struct bkey_buf last_flushed;
 
-	return for_each_btree_key_commit(trans, iter, BTREE_ID_backpointers,
+	bch2_bkey_buf_init(&last_flushed);
+	bkey_init(&last_flushed.k->k);
+
+	int ret = for_each_btree_key_commit(trans, iter, BTREE_ID_backpointers,
 				  POS_MIN, BTREE_ITER_prefetch, k,
 				  NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 		check_one_backpointer(trans, start, end,
 				      bkey_s_c_to_backpointer(k),
-				      &last_flushed_pos));
+				      &last_flushed));
+
+	bch2_bkey_buf_exit(&last_flushed, trans->c);
+	return ret;
 }
 
 int bch2_check_backpointers_to_extents(struct bch_fs *c)

fs/bcachefs/bkey.c

@@ -660,8 +660,9 @@ int bch2_bkey_format_invalid(struct bch_fs *c,
 		    bch2_bkey_format_field_overflows(f, i)) {
 			unsigned unpacked_bits = bch2_bkey_format_current.bits_per_field[i];
 			u64 unpacked_max = ~((~0ULL << 1) << (unpacked_bits - 1));
-			u64 packed_max = f->bits_per_field[i]
-				? ~((~0ULL << 1) << (f->bits_per_field[i] - 1))
+			unsigned packed_bits = min(64, f->bits_per_field[i]);
+			u64 packed_max = packed_bits
+				? ~((~0ULL << 1) << (packed_bits - 1))
 				: 0;
 
 			prt_printf(err, "field %u too large: %llu + %llu > %llu",

fs/bcachefs/bkey.h

@@ -194,6 +194,13 @@ static inline struct bpos bkey_max(struct bpos l, struct bpos r)
 	return bkey_gt(l, r) ? l : r;
 }
 
+static inline bool bkey_and_val_eq(struct bkey_s_c l, struct bkey_s_c r)
+{
+	return bpos_eq(l.k->p, r.k->p) &&
+		bkey_bytes(l.k) == bkey_bytes(r.k) &&
+		!memcmp(l.v, r.v, bkey_val_bytes(l.k));
+}
+
 void bch2_bpos_swab(struct bpos *);
 void bch2_bkey_swab_key(const struct bkey_format *, struct bkey_packed *);
 

fs/bcachefs/btree_gc.c

@@ -903,6 +903,8 @@ static int bch2_alloc_write_key(struct btree_trans *trans,
 		bch2_dev_usage_update(c, ca, &old_gc, &gc, 0, true);
 	percpu_up_read(&c->mark_lock);
 
+	gc.fragmentation_lru = alloc_lru_idx_fragmentation(gc, ca);
+
 	if (fsck_err_on(new.data_type != gc.data_type, c,
 			alloc_key_data_type_wrong,
 			"bucket %llu:%llu gen %u has wrong data_type"
@@ -916,23 +918,19 @@ static int bch2_alloc_write_key(struct btree_trans *trans,
 #define copy_bucket_field(_errtype, _f)					\
 	if (fsck_err_on(new._f != gc._f, c, _errtype,			\
 			"bucket %llu:%llu gen %u data type %s has wrong " #_f	\
-			": got %u, should be %u",			\
+			": got %llu, should be %llu",			\
 			iter->pos.inode, iter->pos.offset,		\
 			gc.gen,						\
 			bch2_data_type_str(gc.data_type),		\
-			new._f, gc._f))					\
+			(u64) new._f, (u64) gc._f))				\
 		new._f = gc._f;						\
 
-	copy_bucket_field(alloc_key_gen_wrong,
-			  gen);
-	copy_bucket_field(alloc_key_dirty_sectors_wrong,
-			  dirty_sectors);
-	copy_bucket_field(alloc_key_cached_sectors_wrong,
-			  cached_sectors);
-	copy_bucket_field(alloc_key_stripe_wrong,
-			  stripe);
-	copy_bucket_field(alloc_key_stripe_redundancy_wrong,
-			  stripe_redundancy);
+	copy_bucket_field(alloc_key_gen_wrong,			gen);
+	copy_bucket_field(alloc_key_dirty_sectors_wrong,	dirty_sectors);
+	copy_bucket_field(alloc_key_cached_sectors_wrong,	cached_sectors);
+	copy_bucket_field(alloc_key_stripe_wrong,		stripe);
+	copy_bucket_field(alloc_key_stripe_redundancy_wrong,	stripe_redundancy);
+	copy_bucket_field(alloc_key_fragmentation_lru_wrong,	fragmentation_lru);
 #undef copy_bucket_field
 
 	if (!bch2_alloc_v4_cmp(*old, new))
@@ -946,7 +944,7 @@ static int bch2_alloc_write_key(struct btree_trans *trans,
 	a->v = new;
 
 	/*
-	 * The trigger normally makes sure this is set, but we're not running
+	 * The trigger normally makes sure these are set, but we're not running
 	 * triggers:
 	 */
 	if (a->v.data_type == BCH_DATA_cached && !a->v.io_time[READ])

fs/bcachefs/btree_write_buffer.c

@@ -1,11 +1,13 @@
 // SPDX-License-Identifier: GPL-2.0
 
 #include "bcachefs.h"
+#include "bkey_buf.h"
 #include "btree_locking.h"
 #include "btree_update.h"
 #include "btree_update_interior.h"
 #include "btree_write_buffer.h"
 #include "error.h"
+#include "extents.h"
 #include "journal.h"
 #include "journal_io.h"
 #include "journal_reclaim.h"
@@ -492,6 +494,41 @@ int bch2_btree_write_buffer_tryflush(struct btree_trans *trans)
 	return ret;
 }
 
+/**
+ * In check and repair code, when checking references to write buffer btrees we
+ * need to issue a flush before we have a definitive error: this issues a flush
+ * if this is a key we haven't yet checked.
+ */
+int bch2_btree_write_buffer_maybe_flush(struct btree_trans *trans,
+					struct bkey_s_c referring_k,
+					struct bkey_buf *last_flushed)
+{
+	struct bch_fs *c = trans->c;
+	struct bkey_buf tmp;
+	int ret = 0;
+
+	bch2_bkey_buf_init(&tmp);
+
+	if (!bkey_and_val_eq(referring_k, bkey_i_to_s_c(last_flushed->k))) {
+		bch2_bkey_buf_reassemble(&tmp, c, referring_k);
+
+		if (bkey_is_btree_ptr(referring_k.k)) {
+			bch2_trans_unlock(trans);
+			bch2_btree_interior_updates_flush(c);
+		}
+
+		ret = bch2_btree_write_buffer_flush_sync(trans);
+		if (ret)
+			goto err;
+
+		bch2_bkey_buf_copy(last_flushed, c, tmp.k);
+		ret = -BCH_ERR_transaction_restart_write_buffer_flush;
+	}
+err:
+	bch2_bkey_buf_exit(&tmp, c);
+	return ret;
+}
+
 static void bch2_btree_write_buffer_flush_work(struct work_struct *work)
 {
 	struct bch_fs *c = container_of(work, struct bch_fs, btree_write_buffer.flush_work);

fs/bcachefs/btree_write_buffer.h

@@ -23,6 +23,9 @@ int bch2_btree_write_buffer_flush_sync(struct btree_trans *);
 int bch2_btree_write_buffer_flush_nocheck_rw(struct btree_trans *);
 int bch2_btree_write_buffer_tryflush(struct btree_trans *);
 
+struct bkey_buf;
+int bch2_btree_write_buffer_maybe_flush(struct btree_trans *, struct bkey_s_c, struct bkey_buf *);
+
 struct journal_keys_to_wb {
 	struct btree_write_buffer_keys	*wb;
 	size_t				room;