Update bcachefs sources to a5c0e1bb30 bcachefs: Clean up bch2_btree_and_journal_walk()

.bcachefs_revision

@@ -1 +1 @@
-8d3093bd9b9254957badce4a4ff178baeb3632ed
+a5c0e1bb306e79b40b2432a22f164697c8b22110

include/linux/blk_types.h

@@ -8,12 +8,43 @@
 #include <linux/atomic.h>
 #include <linux/types.h>
 #include <linux/bvec.h>
+#include <linux/kobject.h>
 
 struct bio_set;
 struct bio;
-struct block_device;
 typedef void (bio_end_io_t) (struct bio *);
 
+#define BDEVNAME_SIZE	32
+
+struct request_queue {
+	struct backing_dev_info *backing_dev_info;
+};
+
+struct gendisk {
+};
+
+struct hd_struct {
+	struct kobject		kobj;
+};
+
+struct block_device {
+	struct kobject		kobj;
+	dev_t			bd_dev;
+	char			name[BDEVNAME_SIZE];
+	struct inode		*bd_inode;
+	struct request_queue	queue;
+	void			*bd_holder;
+	struct gendisk *	bd_disk;
+	struct gendisk		__bd_disk;
+	int			bd_fd;
+	int			bd_sync_fd;
+
+	struct backing_dev_info	*bd_bdi;
+	struct backing_dev_info	__bd_bdi;
+};
+
+#define bdev_kobj(_bdev) (&((_bdev)->kobj))
+
 /*
  * Block error status values.  See block/blk-core:blk_errors for the details.
  */

include/linux/blkdev.h

@@ -59,36 +59,8 @@ static inline struct inode *file_inode(const struct file *f)
 	return f->f_inode;
 }
 
-#define BDEVNAME_SIZE	32
-
-struct request_queue {
-	struct backing_dev_info *backing_dev_info;
-};
-
-struct gendisk {
-};
-
-struct hd_struct {
-	struct kobject		kobj;
-};
-
 #define part_to_dev(part)	(part)
 
-struct block_device {
-	char			name[BDEVNAME_SIZE];
-	struct inode		*bd_inode;
-	struct request_queue	queue;
-	void			*bd_holder;
-	struct hd_struct	*bd_part;
-	struct gendisk		*bd_disk;
-	struct gendisk		__bd_disk;
-	int			bd_fd;
-	int			bd_sync_fd;
-
-	struct backing_dev_info	*bd_bdi;
-	struct backing_dev_info	__bd_bdi;
-};
-
 void generic_make_request(struct bio *);
 int submit_bio_wait(struct bio *);
 
@@ -111,7 +83,7 @@ sector_t get_capacity(struct gendisk *disk);
 void blkdev_put(struct block_device *bdev, fmode_t mode);
 void bdput(struct block_device *bdev);
 struct block_device *blkdev_get_by_path(const char *path, fmode_t mode, void *holder);
-struct block_device *lookup_bdev(const char *path);
+int lookup_bdev(const char *path, dev_t *);
 
 struct super_block {
 	void			*s_fs_info;

include/linux/freezer.h

@@ -4,5 +4,6 @@
 #define try_to_freeze()
 #define set_freezable()
 #define freezing(task)		false
+#define freezable_schedule_timeout(_t) schedule_timeout(_t);
 
 #endif /* __TOOLS_LINUX_FREEZER_H */

include/linux/poison.h

@@ -27,11 +27,7 @@
 #define TIMER_ENTRY_STATIC	((void *) 0x300 + POISON_POINTER_DELTA)
 
 /********** mm/page_poison.c **********/
-#ifdef CONFIG_PAGE_POISONING_ZERO
-#define PAGE_POISON 0x00
-#else
 #define PAGE_POISON 0xaa
-#endif
 
 /********** mm/page_alloc.c ************/
 

include/trace/events/bcachefs.h

@@ -528,6 +528,62 @@ TRACE_EVENT(copygc,
 		__entry->buckets_moved, __entry->buckets_not_moved)
 );
 
+TRACE_EVENT(trans_get_iter,
+	TP_PROTO(unsigned long caller, unsigned long ip,
+		 enum btree_id btree_id,
+		 struct bpos *pos_want,
+		 unsigned locks_want,
+		 struct bpos *pos_found,
+		 unsigned locks_found,
+		 unsigned uptodate),
+	TP_ARGS(caller, ip, btree_id,
+		pos_want, locks_want,
+		pos_found, locks_found,
+		uptodate),
+
+	TP_STRUCT__entry(
+		__field(unsigned long,	caller			)
+		__field(unsigned long,	ip			)
+		__field(u8,		btree_id		)
+		__field(u8,		uptodate		)
+		__field(u8,		locks_want		)
+		__field(u8,		locks_found		)
+		__field(u64,		pos_want_inode		)
+		__field(u64,		pos_want_offset		)
+		__field(u32,		pos_want_snapshot	)
+		__field(u64,		pos_found_inode		)
+		__field(u64,		pos_found_offset	)
+		__field(u32,		pos_found_snapshot	)
+	),
+
+	TP_fast_assign(
+		__entry->caller			= caller;
+		__entry->ip			= ip;
+		__entry->btree_id		= btree_id;
+		__entry->uptodate		= uptodate;
+		__entry->pos_want_inode		= pos_want->inode;
+		__entry->pos_want_offset	= pos_want->offset;
+		__entry->pos_want_snapshot	= pos_want->snapshot;
+		__entry->pos_found_inode	= pos_found->inode;
+		__entry->pos_found_offset	= pos_found->offset;
+		__entry->pos_found_snapshot	= pos_found->snapshot;
+	),
+
+	TP_printk("%ps %pS btree %u uptodate %u want %llu:%llu:%u locks %u found %llu:%llu:%u locks %u",
+		  (void *) __entry->caller,
+		  (void *) __entry->ip,
+		  __entry->btree_id,
+		  __entry->uptodate,
+		  __entry->pos_want_inode,
+		  __entry->pos_want_offset,
+		  __entry->pos_want_snapshot,
+		  __entry->locks_want,
+		  __entry->pos_found_inode,
+		  __entry->pos_found_offset,
+		  __entry->pos_found_snapshot,
+		  __entry->locks_found)
+);
+
 TRACE_EVENT(transaction_restart_ip,
 	TP_PROTO(unsigned long caller, unsigned long ip),
 	TP_ARGS(caller, ip),
@@ -565,6 +621,11 @@ DEFINE_EVENT(transaction_restart,	trans_restart_btree_node_reused,
 	TP_ARGS(ip)
 );
 
+DEFINE_EVENT(transaction_restart,	trans_blocked_journal_reclaim,
+	TP_PROTO(unsigned long ip),
+	TP_ARGS(ip)
+);
+
 TRACE_EVENT(trans_restart_would_deadlock,
 	TP_PROTO(unsigned long	trans_ip,
 		 unsigned long	caller_ip,

libbcachefs/alloc_background.c

@@ -261,16 +261,14 @@ void bch2_alloc_to_text(struct printbuf *out, struct bch_fs *c,
 #undef  x
 }
 
-static int bch2_alloc_read_fn(struct bch_fs *c, enum btree_id id,
+static int bch2_alloc_read_fn(struct bch_fs *c, struct bkey_s_c k)
-			      unsigned level, struct bkey_s_c k)
 {
 	struct bch_dev *ca;
 	struct bucket *g;
 	struct bkey_alloc_unpacked u;
 
-	if (level ||
+	if (k.k->type != KEY_TYPE_alloc &&
-	    (k.k->type != KEY_TYPE_alloc &&
+	    k.k->type != KEY_TYPE_alloc_v2)
-	     k.k->type != KEY_TYPE_alloc_v2))
 		return 0;
 
 	ca = bch_dev_bkey_exists(c, k.k->p.inode);
@@ -289,13 +287,12 @@ static int bch2_alloc_read_fn(struct bch_fs *c, enum btree_id id,
 	return 0;
 }
 
-int bch2_alloc_read(struct bch_fs *c, struct journal_keys *journal_keys)
+int bch2_alloc_read(struct bch_fs *c)
 {
 	int ret;
 
 	down_read(&c->gc_lock);
-	ret = bch2_btree_and_journal_walk(c, journal_keys, BTREE_ID_alloc,
+	ret = bch2_btree_and_journal_walk(c, BTREE_ID_alloc, bch2_alloc_read_fn);
-					  NULL, bch2_alloc_read_fn);
 	up_read(&c->gc_lock);
 	if (ret) {
 		bch_err(c, "error reading alloc info: %i", ret);

libbcachefs/alloc_background.h

@@ -91,8 +91,7 @@ void bch2_alloc_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 	.val_to_text	= bch2_alloc_to_text,		\
 }
 
-struct journal_keys;
+int bch2_alloc_read(struct bch_fs *);
-int bch2_alloc_read(struct bch_fs *, struct journal_keys *);
 
 static inline void bch2_wake_allocator(struct bch_dev *ca)
 {

libbcachefs/bcachefs.h

@@ -605,11 +605,13 @@ struct bch_fs {
 
 		u64		time_base_lo;
 		u32		time_base_hi;
-		u32		time_precision;
+		unsigned	time_units_per_sec;
+		unsigned	nsec_per_time_unit;
 		u64		features;
 		u64		compat;
 	}			sb;
 
+
 	struct bch_sb_handle	disk_sb;
 
 	unsigned short		block_bits;	/* ilog2(block_size) */
@@ -873,19 +875,22 @@ static inline unsigned block_bytes(const struct bch_fs *c)
 	return c->opts.block_size << 9;
 }
 
-static inline struct timespec64 bch2_time_to_timespec(struct bch_fs *c, u64 time)
+static inline struct timespec64 bch2_time_to_timespec(struct bch_fs *c, s64 time)
 {
-	return ns_to_timespec64(time * c->sb.time_precision + c->sb.time_base_lo);
+	struct timespec64 t;
+	s32 rem;
+
+	time += c->sb.time_base_lo;
+
+	t.tv_sec = div_s64_rem(time, c->sb.time_units_per_sec, &rem);
+	t.tv_nsec = rem * c->sb.nsec_per_time_unit;
+	return t;
 }
 
 static inline s64 timespec_to_bch2_time(struct bch_fs *c, struct timespec64 ts)
 {
-	s64 ns = timespec64_to_ns(&ts) - c->sb.time_base_lo;
+	return (ts.tv_sec * c->sb.time_units_per_sec +
-
+		(int) ts.tv_nsec / c->sb.nsec_per_time_unit) - c->sb.time_base_lo;
-	if (c->sb.time_precision == 1)
-		return ns;
-
-	return div_s64(ns, c->sb.time_precision);
 }
 
 static inline s64 bch2_current_time(struct bch_fs *c)

libbcachefs/btree_cache.c

@@ -581,6 +581,7 @@ out:
 	b->sib_u64s[1]		= 0;
 	b->whiteout_u64s	= 0;
 	bch2_btree_keys_init(b);
+	set_btree_node_accessed(b);
 
 	bch2_time_stats_update(&c->times[BCH_TIME_btree_node_mem_alloc],
 			       start_time);
@@ -653,9 +654,13 @@ static noinline struct btree *bch2_btree_node_fill(struct bch_fs *c,
 		return NULL;
 	}
 
-	/* Unlock before doing IO: */
+	/*
-	if (iter && sync)
+	 * Unlock before doing IO:
-		bch2_trans_unlock(iter->trans);
+	 *
+	 * XXX: ideally should be dropping all btree node locks here
+	 */
+	if (iter && btree_node_read_locked(iter, level + 1))
+		btree_node_unlock(iter, level + 1);
 
 	bch2_btree_node_read(c, b, sync);
 
@@ -666,16 +671,6 @@ static noinline struct btree *bch2_btree_node_fill(struct bch_fs *c,
 		return NULL;
 	}
 
-	/*
-	 * XXX: this will probably always fail because btree_iter_relock()
-	 * currently fails for iterators that aren't pointed at a valid btree
-	 * node
-	 */
-	if (iter && !bch2_trans_relock(iter->trans)) {
-		six_unlock_intent(&b->c.lock);
-		return ERR_PTR(-EINTR);
-	}
-
 	if (lock_type == SIX_LOCK_read)
 		six_lock_downgrade(&b->c.lock);
 
@@ -817,22 +812,9 @@ lock_node:
 		}
 	}
 
-	if (unlikely(btree_node_read_in_flight(b))) {
+	/* XXX: waiting on IO with btree locks held: */
-		six_unlock_type(&b->c.lock, lock_type);
+	wait_on_bit_io(&b->flags, BTREE_NODE_read_in_flight,
-		bch2_trans_unlock(iter->trans);
+		       TASK_UNINTERRUPTIBLE);
-
-		wait_on_bit_io(&b->flags, BTREE_NODE_read_in_flight,
-			       TASK_UNINTERRUPTIBLE);
-
-		/*
-		 * XXX: check if this always fails - btree_iter_relock()
-		 * currently fails for iterators that aren't pointed at a valid
-		 * btree node
-		 */
-		if (iter && !bch2_trans_relock(iter->trans))
-			return ERR_PTR(-EINTR);
-		goto retry;
-	}
 
 	prefetch(b->aux_data);
 

libbcachefs/btree_iter.c

@@ -2013,6 +2013,13 @@ struct btree_iter *__bch2_trans_get_iter(struct btree_trans *trans,
 					 unsigned flags)
 {
 	struct btree_iter *iter, *best = NULL;
+	struct bpos real_pos, pos_min = POS_MIN;
+
+	if ((flags & BTREE_ITER_TYPE) != BTREE_ITER_NODES &&
+	    btree_node_type_is_extents(btree_id) &&
+	    !(flags & BTREE_ITER_NOT_EXTENTS) &&
+	    !(flags & BTREE_ITER_ALL_SNAPSHOTS))
+		flags |= BTREE_ITER_IS_EXTENTS;
 
 	if ((flags & BTREE_ITER_TYPE) != BTREE_ITER_NODES &&
 	    !btree_type_has_snapshots(btree_id))
@@ -2022,6 +2029,12 @@ struct btree_iter *__bch2_trans_get_iter(struct btree_trans *trans,
 		pos.snapshot = btree_type_has_snapshots(btree_id)
 			? U32_MAX : 0;
 
+	real_pos = pos;
+
+	if ((flags & BTREE_ITER_IS_EXTENTS) &&
+	    bkey_cmp(pos, POS_MAX))
+		real_pos = bpos_nosnap_successor(pos);
+
 	trans_for_each_iter(trans, iter) {
 		if (btree_iter_type(iter) != (flags & BTREE_ITER_TYPE))
 			continue;
@@ -2030,8 +2043,8 @@ struct btree_iter *__bch2_trans_get_iter(struct btree_trans *trans,
 			continue;
 
 		if (best) {
-			int cmp = bkey_cmp(bpos_diff(best->real_pos, pos),
+			int cmp = bkey_cmp(bpos_diff(best->real_pos, real_pos),
-					   bpos_diff(iter->real_pos, pos));
+					   bpos_diff(iter->real_pos, real_pos));
 
 			if (cmp < 0 ||
 			    ((cmp == 0 && btree_iter_keep(trans, iter))))
@@ -2041,6 +2054,13 @@ struct btree_iter *__bch2_trans_get_iter(struct btree_trans *trans,
 		best = iter;
 	}
 
+	trace_trans_get_iter(_RET_IP_, trans->ip,
+			     btree_id,
+			     &real_pos, locks_want,
+			     best ? &best->real_pos : &pos_min,
+			     best ? best->locks_want : 0,
+			     best ? best->uptodate : BTREE_ITER_NEED_TRAVERSE);
+
 	if (!best) {
 		iter = btree_trans_iter_alloc(trans);
 		bch2_btree_iter_init(trans, iter, btree_id);
@@ -2054,12 +2074,6 @@ struct btree_iter *__bch2_trans_get_iter(struct btree_trans *trans,
 	trans->iters_live	|= 1ULL << iter->idx;
 	trans->iters_touched	|= 1ULL << iter->idx;
 
-	if ((flags & BTREE_ITER_TYPE) != BTREE_ITER_NODES &&
-	    btree_node_type_is_extents(btree_id) &&
-	    !(flags & BTREE_ITER_NOT_EXTENTS) &&
-	    !(flags & BTREE_ITER_ALL_SNAPSHOTS))
-		flags |= BTREE_ITER_IS_EXTENTS;
-
 	iter->flags = flags;
 
 	iter->snapshot = pos.snapshot;
@@ -2078,19 +2092,20 @@ struct btree_iter *__bch2_trans_get_iter(struct btree_trans *trans,
 		btree_iter_get_locks(iter, true, false);
 	}
 
-	while (iter->level < depth) {
+	while (iter->level != depth) {
 		btree_node_unlock(iter, iter->level);
 		iter->l[iter->level].b = BTREE_ITER_NO_NODE_INIT;
-		iter->level++;
+		iter->uptodate = BTREE_ITER_NEED_TRAVERSE;
+		if (iter->level < depth)
+			iter->level++;
+		else
+			iter->level--;
 	}
 
-	while (iter->level > depth)
-		iter->l[--iter->level].b = BTREE_ITER_NO_NODE_INIT;
-
 	iter->min_depth	= depth;
 
 	bch2_btree_iter_set_pos(iter, pos);
-	btree_iter_set_search_pos(iter, btree_iter_search_key(iter));
+	btree_iter_set_search_pos(iter, real_pos);
 
 	return iter;
 }

libbcachefs/btree_key_cache.c

@@ -645,8 +645,10 @@ static unsigned long bch2_btree_key_cache_count(struct shrinker *shrink,
 	struct bch_fs *c = container_of(shrink, struct bch_fs,
 					btree_key_cache.shrink);
 	struct btree_key_cache *bc = &c->btree_key_cache;
+	long nr = atomic_long_read(&bc->nr_keys) -
+		atomic_long_read(&bc->nr_dirty);
 
-	return atomic_long_read(&bc->nr_keys);
+	return max(0L, nr);
 }
 
 void bch2_fs_btree_key_cache_exit(struct btree_key_cache *bc)

libbcachefs/btree_update_leaf.c

@@ -725,6 +725,8 @@ int bch2_trans_commit_error(struct btree_trans *trans,
 	case BTREE_INSERT_NEED_JOURNAL_RECLAIM:
 		bch2_trans_unlock(trans);
 
+		trace_trans_blocked_journal_reclaim(trans->ip);
+
 		wait_event_freezable(c->journal.reclaim_wait,
 				     (ret = journal_reclaim_wait_done(c)));
 		if (ret < 0)

libbcachefs/buckets.c

@@ -898,6 +898,7 @@ static int bch2_mark_stripe_ptr(struct bch_fs *c,
 		spin_unlock(&c->ec_stripes_heap_lock);
 		bch_err_ratelimited(c, "pointer to nonexistent stripe %llu",
 				    (u64) p.idx);
+		bch2_inconsistent_error(c);
 		return -EIO;
 	}
 
@@ -1015,6 +1016,7 @@ static int bch2_mark_stripe(struct bch_fs *c,
 	if (!m || (old_s && !m->alive)) {
 		bch_err_ratelimited(c, "error marking nonexistent stripe %zu",
 				    idx);
+		bch2_inconsistent_error(c);
 		return -1;
 	}
 
@@ -1499,6 +1501,7 @@ static int bch2_trans_mark_stripe_ptr(struct btree_trans *trans,
 		bch2_fs_inconsistent(c,
 			"pointer to nonexistent stripe %llu",
 			(u64) p.ec.idx);
+		bch2_inconsistent_error(c);
 		ret = -EIO;
 		goto out;
 	}
@@ -1739,6 +1742,7 @@ static int __bch2_trans_mark_reflink_p(struct btree_trans *trans,
 		bch2_fs_inconsistent(c,
 			"%llu:%llu len %u points to nonexistent indirect extent %llu",
 			p.k->p.inode, p.k->p.offset, p.k->size, idx);
+		bch2_inconsistent_error(c);
 		ret = -EIO;
 		goto err;
 	}

libbcachefs/ec.c

@@ -1630,26 +1630,22 @@ int bch2_stripes_write(struct bch_fs *c, unsigned flags)
 	return ret;
 }
 
-static int bch2_stripes_read_fn(struct bch_fs *c, enum btree_id id,
+static int bch2_stripes_read_fn(struct bch_fs *c, struct bkey_s_c k)
-			      unsigned level, struct bkey_s_c k)
 {
 	int ret = 0;
 
-	if (k.k->type == KEY_TYPE_stripe) {
+	if (k.k->type == KEY_TYPE_stripe)
 		ret = __ec_stripe_mem_alloc(c, k.k->p.offset, GFP_KERNEL) ?:
 			bch2_mark_key(c, k, 0, 0, NULL, 0,
 				      BTREE_TRIGGER_NOATOMIC);
-		if (ret)
-			return ret;
-	}
 
 	return ret;
 }
 
-int bch2_stripes_read(struct bch_fs *c, struct journal_keys *journal_keys)
+int bch2_stripes_read(struct bch_fs *c)
 {
-	int ret = bch2_btree_and_journal_walk(c, journal_keys, BTREE_ID_stripes,
+	int ret = bch2_btree_and_journal_walk(c, BTREE_ID_stripes,
-					  NULL, bch2_stripes_read_fn);
+					      bch2_stripes_read_fn);
 	if (ret)
 		bch_err(c, "error reading stripes: %i", ret);
 

libbcachefs/ec.h

@@ -215,8 +215,7 @@ void bch2_ec_flush_new_stripes(struct bch_fs *);
 
 void bch2_stripes_heap_start(struct bch_fs *);
 
-struct journal_keys;
+int bch2_stripes_read(struct bch_fs *);
-int bch2_stripes_read(struct bch_fs *, struct journal_keys *);
 int bch2_stripes_write(struct bch_fs *, unsigned);
 
 int bch2_ec_mem_alloc(struct bch_fs *, bool);

libbcachefs/fs-io.c

@@ -1931,8 +1931,9 @@ loop:
 			i_size_write(&inode->v, req->ki_pos);
 		spin_unlock(&inode->v.i_lock);
 
-		bio_for_each_segment_all(bv, bio, iter)
+		if (likely(!bio_flagged(bio, BIO_NO_PAGE_REF)))
-			put_page(bv->bv_page);
+			bio_for_each_segment_all(bv, bio, iter)
+				put_page(bv->bv_page);
 
 		if (dio->op.error) {
 			set_bit(EI_INODE_ERROR, &inode->ei_flags);
@@ -2393,6 +2394,15 @@ err:
 
 /* fallocate: */
 
+static int inode_update_times_fn(struct bch_inode_info *inode,
+				 struct bch_inode_unpacked *bi, void *p)
+{
+	struct bch_fs *c = inode->v.i_sb->s_fs_info;
+
+	bi->bi_mtime = bi->bi_ctime = bch2_current_time(c);
+	return 0;
+}
+
 static long bchfs_fpunch(struct bch_inode_info *inode, loff_t offset, loff_t len)
 {
 	struct bch_fs *c = inode->v.i_sb->s_fs_info;
@@ -2430,6 +2440,11 @@ static long bchfs_fpunch(struct bch_inode_info *inode, loff_t offset, loff_t len
 				  &i_sectors_delta);
 		i_sectors_acct(c, inode, NULL, i_sectors_delta);
 	}
+
+	mutex_lock(&inode->ei_update_lock);
+	ret = bch2_write_inode(c, inode, inode_update_times_fn, NULL,
+			       ATTR_MTIME|ATTR_CTIME) ?: ret;
+	mutex_unlock(&inode->ei_update_lock);
 err:
 	bch2_pagecache_block_put(&inode->ei_pagecache_lock);
 	inode_unlock(&inode->v);

libbcachefs/fs.c

@@ -143,7 +143,7 @@ int __must_check bch2_write_inode(struct bch_fs *c,
 	struct bch_inode_unpacked inode_u;
 	int ret;
 
-	bch2_trans_init(&trans, c, 0, 0);
+	bch2_trans_init(&trans, c, 0, 256);
 retry:
 	bch2_trans_begin(&trans);
 
@@ -998,10 +998,7 @@ static const struct file_operations bch_file_operations = {
 	.open		= generic_file_open,
 	.fsync		= bch2_fsync,
 	.splice_read	= generic_file_splice_read,
-#if 0
-	/* Busted: */
 	.splice_write	= iter_file_splice_write,
-#endif
 	.fallocate	= bch2_fallocate_dispatch,
 	.unlocked_ioctl = bch2_fs_file_ioctl,
 #ifdef CONFIG_COMPAT
@@ -1293,16 +1290,17 @@ static int bch2_sync_fs(struct super_block *sb, int wait)
 	return bch2_journal_flush(&c->journal);
 }
 
-static struct bch_fs *bch2_path_to_fs(const char *dev)
+static struct bch_fs *bch2_path_to_fs(const char *path)
 {
 	struct bch_fs *c;
-	struct block_device *bdev = lookup_bdev(dev);
+	dev_t dev;
+	int ret;
 
-	if (IS_ERR(bdev))
+	ret = lookup_bdev(path, &dev);
-		return ERR_CAST(bdev);
+	if (ret)
+		return ERR_PTR(ret);
 
-	c = bch2_bdev_to_fs(bdev);
+	c = bch2_dev_to_fs(dev);
-	bdput(bdev);
 	if (c)
 		closure_put(&c->cl);
 	return c ?: ERR_PTR(-ENOENT);
@@ -1554,7 +1552,9 @@ got_sb:
 #endif
 	sb->s_xattr		= bch2_xattr_handlers;
 	sb->s_magic		= BCACHEFS_STATFS_MAGIC;
-	sb->s_time_gran		= c->sb.time_precision;
+	sb->s_time_gran		= c->sb.nsec_per_time_unit;
+	sb->s_time_min		= div_s64(S64_MIN, c->sb.time_units_per_sec) + 1;
+	sb->s_time_max		= div_s64(S64_MAX, c->sb.time_units_per_sec);
 	c->vfs_sb		= sb;
 	strlcpy(sb->s_id, c->name, sizeof(sb->s_id));
 

libbcachefs/io.c

@@ -1968,6 +1968,7 @@ int __bch2_read_indirect_extent(struct btree_trans *trans,
 	    k.k->type != KEY_TYPE_indirect_inline_data) {
 		bch_err_inum_ratelimited(trans->c, orig_k->k->k.p.inode,
 				"pointer to nonexistent indirect extent");
+		bch2_inconsistent_error(trans->c);
 		ret = -EIO;
 		goto err;
 	}

libbcachefs/journal.c

@@ -1187,6 +1187,8 @@ void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
 	       "nr noflush writes:\t%llu\n"
 	       "nr direct reclaim:\t%llu\n"
 	       "nr background reclaim:\t%llu\n"
+	       "reclaim kicked:\t\t%u\n"
+	       "reclaim runs in:\t%u ms\n"
 	       "current entry sectors:\t%u\n"
 	       "current entry error:\t%u\n"
 	       "current entry:\t\t",
@@ -1202,6 +1204,8 @@ void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
 	       j->nr_noflush_writes,
 	       j->nr_direct_reclaim,
 	       j->nr_background_reclaim,
+	       j->reclaim_kicked,
+	       jiffies_to_msecs(j->next_reclaim - jiffies),
 	       j->cur_entry_sectors,
 	       j->cur_entry_error);
 

libbcachefs/journal_reclaim.c

@@ -677,13 +677,15 @@ int bch2_journal_reclaim(struct journal *j)
 static int bch2_journal_reclaim_thread(void *arg)
 {
 	struct journal *j = arg;
-	unsigned long next;
+	unsigned long delay, now;
 	int ret = 0;
 
 	set_freezable();
 
 	kthread_wait_freezable(test_bit(JOURNAL_RECLAIM_STARTED, &j->flags));
 
+	j->last_flushed = jiffies;
+
 	while (!ret && !kthread_should_stop()) {
 		j->reclaim_kicked = false;
 
@@ -691,7 +693,12 @@ static int bch2_journal_reclaim_thread(void *arg)
 		ret = __bch2_journal_reclaim(j, false);
 		mutex_unlock(&j->reclaim_lock);
 
-		next = j->last_flushed + msecs_to_jiffies(j->reclaim_delay_ms);
+		now = jiffies;
+		delay = msecs_to_jiffies(j->reclaim_delay_ms);
+		j->next_reclaim = j->last_flushed + delay;
+
+		if (!time_in_range(j->next_reclaim, now, now + delay))
+			j->next_reclaim = now + delay;
 
 		while (1) {
 			set_current_state(TASK_INTERRUPTIBLE);
@@ -699,10 +706,9 @@ static int bch2_journal_reclaim_thread(void *arg)
 				break;
 			if (j->reclaim_kicked)
 				break;
-			if (time_after_eq(jiffies, next))
+			if (time_after_eq(jiffies, j->next_reclaim))
 				break;
-			schedule_timeout(next - jiffies);
+			freezable_schedule_timeout(j->next_reclaim - jiffies);
-			try_to_freeze();
 
 		}
 		__set_current_state(TASK_RUNNING);

libbcachefs/journal_reclaim.h

@@ -8,11 +8,9 @@ static inline void journal_reclaim_kick(struct journal *j)
 {
 	struct task_struct *p = READ_ONCE(j->reclaim_thread);
 
-	if (p && !j->reclaim_kicked) {
+	j->reclaim_kicked = true;
-		j->reclaim_kicked = true;
+	if (p)
-		if (p)
+		wake_up_process(p);
-			wake_up_process(p);
-	}
 }
 
 unsigned bch2_journal_dev_buckets_available(struct journal *,

libbcachefs/journal_types.h

@@ -248,6 +248,7 @@ struct journal {
 	wait_queue_head_t	reclaim_wait;
 	struct task_struct	*reclaim_thread;
 	bool			reclaim_kicked;
+	unsigned long		next_reclaim;
 	u64			nr_direct_reclaim;
 	u64			nr_background_reclaim;
 

libbcachefs/movinggc.c

@@ -293,17 +293,19 @@ unsigned long bch2_copygc_wait_amount(struct bch_fs *c)
 {
 	struct bch_dev *ca;
 	unsigned dev_idx;
-	u64 fragmented_allowed = 0, fragmented = 0;
+	s64 wait = S64_MAX, fragmented_allowed, fragmented;
 
 	for_each_rw_member(ca, c, dev_idx) {
 		struct bch_dev_usage usage = bch2_dev_usage_read(ca);
 
-		fragmented_allowed += ((__dev_buckets_reclaimable(ca, usage) *
+		fragmented_allowed = ((__dev_buckets_reclaimable(ca, usage) *
 					ca->mi.bucket_size) >> 1);
-		fragmented += usage.d[BCH_DATA_user].fragmented;
+		fragmented = usage.d[BCH_DATA_user].fragmented;
+
+		wait = min(wait, max(0LL, fragmented_allowed - fragmented));
 	}
 
-	return max_t(s64, 0, fragmented_allowed - fragmented);
+	return wait;
 }
 
 static int bch2_copygc_thread(void *arg)

libbcachefs/quota.c

@@ -372,6 +372,7 @@ static int bch2_quota_init_type(struct bch_fs *c, enum quota_types type)
 		if (ret)
 			break;
 	}
+	bch2_trans_iter_put(&trans, iter);
 
 	return bch2_trans_exit(&trans) ?: ret;
 }
@@ -449,6 +450,8 @@ int bch2_fs_quota_read(struct bch_fs *c)
 					KEY_TYPE_QUOTA_NOCHECK);
 		}
 	}
+	bch2_trans_iter_put(&trans, iter);
+
 	return bch2_trans_exit(&trans) ?: ret;
 }
 

libbcachefs/recovery.c

@@ -323,9 +323,7 @@ static void btree_and_journal_iter_prefetch(struct bch_fs *c, struct btree *b,
 }
 
 static int bch2_btree_and_journal_walk_recurse(struct bch_fs *c, struct btree *b,
-				struct journal_keys *journal_keys,
 				enum btree_id btree_id,
-				btree_walk_node_fn node_fn,
 				btree_walk_key_fn key_fn)
 {
 	struct btree_and_journal_iter iter;
@@ -338,15 +336,9 @@ static int bch2_btree_and_journal_walk_recurse(struct bch_fs *c, struct btree *b
 	bch2_btree_and_journal_iter_init_node_iter(&iter, c, b);
 
 	while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) {
-		ret = key_fn(c, btree_id, b->c.level, k);
-		if (ret)
-			break;
-
 		if (b->c.level) {
 			bch2_bkey_buf_reassemble(&tmp, c, k);
 
-			bch2_btree_and_journal_iter_advance(&iter);
-
 			child = bch2_btree_node_get_noiter(c, tmp.k,
 						b->c.btree_id, b->c.level - 1,
 						false);
@@ -357,16 +349,17 @@ static int bch2_btree_and_journal_walk_recurse(struct bch_fs *c, struct btree *b
 
 			btree_and_journal_iter_prefetch(c, b, iter);
 
-			ret   = (node_fn ? node_fn(c, b) : 0) ?:
+			ret = bch2_btree_and_journal_walk_recurse(c, child,
-				bch2_btree_and_journal_walk_recurse(c, child,
+					btree_id, key_fn);
-					journal_keys, btree_id, node_fn, key_fn);
 			six_unlock_read(&child->c.lock);
-
-			if (ret)
-				break;
 		} else {
-			bch2_btree_and_journal_iter_advance(&iter);
+			ret = key_fn(c, k);
 		}
+
+		if (ret)
+			break;
+
+		bch2_btree_and_journal_iter_advance(&iter);
 	}
 
 	bch2_btree_and_journal_iter_exit(&iter);
@@ -374,9 +367,7 @@ static int bch2_btree_and_journal_walk_recurse(struct bch_fs *c, struct btree *b
 	return ret;
 }
 
-int bch2_btree_and_journal_walk(struct bch_fs *c, struct journal_keys *journal_keys,
+int bch2_btree_and_journal_walk(struct bch_fs *c, enum btree_id btree_id,
-				enum btree_id btree_id,
-				btree_walk_node_fn node_fn,
 				btree_walk_key_fn key_fn)
 {
 	struct btree *b = c->btree_roots[btree_id].b;
@@ -386,10 +377,7 @@ int bch2_btree_and_journal_walk(struct bch_fs *c, struct journal_keys *journal_k
 		return 0;
 
 	six_lock_read(&b->c.lock, NULL, NULL);
-	ret   = (node_fn ? node_fn(c, b) : 0) ?:
+	ret = bch2_btree_and_journal_walk_recurse(c, b, btree_id, key_fn);
-		bch2_btree_and_journal_walk_recurse(c, b, journal_keys, btree_id,
-						    node_fn, key_fn) ?:
-		key_fn(c, btree_id, b->c.level + 1, bkey_i_to_s_c(&b->key));
 	six_unlock_read(&b->c.lock);
 
 	return ret;
@@ -1120,14 +1108,14 @@ use_clean:
 
 	bch_verbose(c, "starting alloc read");
 	err = "error reading allocation information";
-	ret = bch2_alloc_read(c, &c->journal_keys);
+	ret = bch2_alloc_read(c);
 	if (ret)
 		goto err;
 	bch_verbose(c, "alloc read done");
 
 	bch_verbose(c, "starting stripes_read");
 	err = "error reading stripes";
-	ret = bch2_stripes_read(c, &c->journal_keys);
+	ret = bch2_stripes_read(c);
 	if (ret)
 		goto err;
 	bch_verbose(c, "stripes_read done");

libbcachefs/recovery.h

@@ -45,12 +45,9 @@ void bch2_btree_and_journal_iter_init_node_iter(struct btree_and_journal_iter *,
 						struct bch_fs *,
 						struct btree *);
 
-typedef int (*btree_walk_node_fn)(struct bch_fs *c, struct btree *b);
+typedef int (*btree_walk_key_fn)(struct bch_fs *c, struct bkey_s_c k);
-typedef int (*btree_walk_key_fn)(struct bch_fs *c, enum btree_id id,
-				 unsigned level, struct bkey_s_c k);
 
-int bch2_btree_and_journal_walk(struct bch_fs *, struct journal_keys *, enum btree_id,
+int bch2_btree_and_journal_walk(struct bch_fs *, enum btree_id, btree_walk_key_fn);
-				btree_walk_node_fn, btree_walk_key_fn);
 
 void bch2_journal_keys_free(struct journal_keys *);
 void bch2_journal_entries_free(struct list_head *);

libbcachefs/s128.h

@@ -0,0 +1,29 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _BCACHEFS_S128_H
+#define _BCACHEFS_S128_H
+
+#include <linux/math64.h>
+
+typedef struct {
+	s64	lo;
+	s64	hi;
+} s128;
+
+typedef struct {
+	s64	lo;
+	s32	hi;
+} s96;
+
+static inline s128 s128_mul(s128 a, s128 b)
+{
+	return a.lo
+
+}
+
+static inline s96 s96_mul(s96 a, s96 b)
+{
+	return a.lo
+
+}
+
+#endif /* _BCACHEFS_S128_H */

libbcachefs/str_hash.h

@@ -12,7 +12,7 @@
 
 #include <linux/crc32c.h>
 #include <crypto/hash.h>
-#include <crypto/sha.h>
+#include <crypto/sha2.h>
 
 static inline enum bch_str_hash_type
 bch2_str_hash_opt_to_type(struct bch_fs *c, enum bch_str_hash_opts opt)

libbcachefs/super-io.c

@@ -367,9 +367,15 @@ static void bch2_sb_update(struct bch_fs *c)
 	c->sb.clean		= BCH_SB_CLEAN(src);
 	c->sb.encryption_type	= BCH_SB_ENCRYPTION_TYPE(src);
 	c->sb.encoded_extent_max= 1 << BCH_SB_ENCODED_EXTENT_MAX_BITS(src);
-	c->sb.time_base_lo	= le64_to_cpu(src->time_base_lo);
+
+	c->sb.nsec_per_time_unit = le32_to_cpu(src->time_precision);
+	c->sb.time_units_per_sec = NSEC_PER_SEC / c->sb.nsec_per_time_unit;
+
+	/* XXX this is wrong, we need a 96 or 128 bit integer type */
+	c->sb.time_base_lo	= div_u64(le64_to_cpu(src->time_base_lo),
+					  c->sb.nsec_per_time_unit);
 	c->sb.time_base_hi	= le32_to_cpu(src->time_base_hi);
-	c->sb.time_precision	= le32_to_cpu(src->time_precision);
+
 	c->sb.features		= le64_to_cpu(src->features[0]);
 	c->sb.compat		= le64_to_cpu(src->compat[0]);
 

libbcachefs/super.c

@@ -99,7 +99,7 @@ static int bch2_dev_alloc(struct bch_fs *, unsigned);
 static int bch2_dev_sysfs_online(struct bch_fs *, struct bch_dev *);
 static void __bch2_dev_read_only(struct bch_fs *, struct bch_dev *);
 
-struct bch_fs *bch2_bdev_to_fs(struct block_device *bdev)
+struct bch_fs *bch2_dev_to_fs(dev_t dev)
 {
 	struct bch_fs *c;
 	struct bch_dev *ca;
@@ -110,7 +110,7 @@ struct bch_fs *bch2_bdev_to_fs(struct block_device *bdev)
 
 	list_for_each_entry(c, &bch_fs_list, list)
 		for_each_member_device_rcu(ca, c, i, NULL)
-			if (ca->disk_sb.bdev == bdev) {
+			if (ca->disk_sb.bdev->bd_dev == dev) {
 				closure_get(&c->cl);
 				goto found;
 			}
@@ -544,8 +544,7 @@ void __bch2_fs_stop(struct bch_fs *c)
 	for_each_member_device(ca, c, i)
 		if (ca->kobj.state_in_sysfs &&
 		    ca->disk_sb.bdev)
-			sysfs_remove_link(&part_to_dev(ca->disk_sb.bdev->bd_part)->kobj,
+			sysfs_remove_link(bdev_kobj(ca->disk_sb.bdev), "bcachefs");
-					  "bcachefs");
 
 	if (c->kobj.state_in_sysfs)
 		kobject_del(&c->kobj);
@@ -1017,8 +1016,7 @@ static void bch2_dev_free(struct bch_dev *ca)
 
 	if (ca->kobj.state_in_sysfs &&
 	    ca->disk_sb.bdev)
-		sysfs_remove_link(&part_to_dev(ca->disk_sb.bdev->bd_part)->kobj,
+		sysfs_remove_link(bdev_kobj(ca->disk_sb.bdev), "bcachefs");
-				  "bcachefs");
 
 	if (ca->kobj.state_in_sysfs)
 		kobject_del(&ca->kobj);
@@ -1054,10 +1052,7 @@ static void __bch2_dev_offline(struct bch_fs *c, struct bch_dev *ca)
 	wait_for_completion(&ca->io_ref_completion);
 
 	if (ca->kobj.state_in_sysfs) {
-		struct kobject *block =
+		sysfs_remove_link(bdev_kobj(ca->disk_sb.bdev), "bcachefs");
-			&part_to_dev(ca->disk_sb.bdev->bd_part)->kobj;
-
-		sysfs_remove_link(block, "bcachefs");
 		sysfs_remove_link(&ca->kobj, "block");
 	}
 
@@ -1094,12 +1089,12 @@ static int bch2_dev_sysfs_online(struct bch_fs *c, struct bch_dev *ca)
 	}
 
 	if (ca->disk_sb.bdev) {
-		struct kobject *block =
+		struct kobject *block = bdev_kobj(ca->disk_sb.bdev);
-			&part_to_dev(ca->disk_sb.bdev->bd_part)->kobj;
 
 		ret = sysfs_create_link(block, &ca->kobj, "bcachefs");
 		if (ret)
 			return ret;
+
 		ret = sysfs_create_link(&ca->kobj, block, "block");
 		if (ret)
 			return ret;
@@ -1837,20 +1832,21 @@ err:
 /* return with ref on ca->ref: */
 struct bch_dev *bch2_dev_lookup(struct bch_fs *c, const char *path)
 {
-	struct block_device *bdev = lookup_bdev(path);
 	struct bch_dev *ca;
+	dev_t dev;
 	unsigned i;
+	int ret;
 
-	if (IS_ERR(bdev))
+	ret = lookup_bdev(path, &dev);
-		return ERR_CAST(bdev);
+	if (ret)
+		return ERR_PTR(ret);
 
 	for_each_member_device(ca, c, i)
-		if (ca->disk_sb.bdev == bdev)
+		if (ca->disk_sb.bdev->bd_dev == dev)
 			goto found;
 
 	ca = ERR_PTR(-ENOENT);
 found:
-	bdput(bdev);
 	return ca;
 }
 

libbcachefs/super.h

@@ -197,7 +197,7 @@ static inline struct bch_devs_mask bch2_online_devs(struct bch_fs *c)
 	return devs;
 }
 
-struct bch_fs *bch2_bdev_to_fs(struct block_device *);
+struct bch_fs *bch2_dev_to_fs(dev_t);
 struct bch_fs *bch2_uuid_to_fs(uuid_le);
 
 bool bch2_dev_state_allowed(struct bch_fs *, struct bch_dev *,

linux/blkdev.c

@@ -215,6 +215,7 @@ struct block_device *blkdev_get_by_path(const char *path, fmode_t mode,
 	strncpy(bdev->name, path, sizeof(bdev->name));
 	bdev->name[sizeof(bdev->name) - 1] = '\0';
 
+	bdev->bd_dev		= xfstat(fd).st_rdev;
 	bdev->bd_fd		= fd;
 	bdev->bd_sync_fd	= sync_fd;
 	bdev->bd_holder		= holder;
@@ -230,9 +231,9 @@ void bdput(struct block_device *bdev)
 	BUG();
 }
 
-struct block_device *lookup_bdev(const char *path)
+int lookup_bdev(const char *path, dev_t *dev)
 {
-	return ERR_PTR(-EINVAL);
+	return -EINVAL;
 }
 
 static int aio_completion_thread(void *arg)