X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Fjournal.c;h=433c97844f36f7058f5501600e6651f256706add;hb=5ef62f56ab50c5799f713e3a42f5c7ad7e8283d3;hp=505e8367b5f2c867a7e2eb94ff74a39ca8457f4e;hpb=733a17f637dccb7b47bcb3016dd867974db7088d;p=bcachefs-tools-debian

diff --git a/libbcachefs/journal.c b/libbcachefs/journal.c
index 505e836..433c978 100644
--- a/libbcachefs/journal.c
+++ b/libbcachefs/journal.c
@@ -17,8 +17,7 @@
 #include "journal_reclaim.h"
 #include "journal_sb.h"
 #include "journal_seq_blacklist.h"
-
-#include <trace/events/bcachefs.h>
+#include "trace.h"
 
 #define x(n)	#n,
 static const char * const bch2_journal_watermarks[] = {
@@ -68,13 +67,75 @@ journal_seq_to_buf(struct journal *j, u64 seq)
 
 static void journal_pin_list_init(struct journal_entry_pin_list *p, int count)
 {
-	INIT_LIST_HEAD(&p->list);
-	INIT_LIST_HEAD(&p->key_cache_list);
+	unsigned i;
+	for (i = 0; i < ARRAY_SIZE(p->list); i++)
+		INIT_LIST_HEAD(&p->list[i]);
 	INIT_LIST_HEAD(&p->flushed);
 	atomic_set(&p->count, count);
 	p->devs.nr = 0;
 }
 
+/*
+ * Detect stuck journal conditions and trigger shutdown. Technically the journal
+ * can end up stuck for a variety of reasons, such as a blocked I/O, journal
+ * reservation lockup, etc. Since this is a fatal error with potentially
+ * unpredictable characteristics, we want to be fairly conservative before we
+ * decide to shut things down.
+ *
+ * Consider the journal stuck when it appears full with no ability to commit
+ * btree transactions, to discard journal buckets, nor acquire priority
+ * (reserved watermark) reservation.
+ */
+static inline bool
+journal_error_check_stuck(struct journal *j, int error, unsigned flags)
+{
+	struct bch_fs *c = container_of(j, struct bch_fs, journal);
+	bool stuck = false;
+	struct printbuf buf = PRINTBUF;
+
+	if (!(error == JOURNAL_ERR_journal_full ||
+	      error == JOURNAL_ERR_journal_pin_full) ||
+	    nr_unwritten_journal_entries(j) ||
+	    (flags & JOURNAL_WATERMARK_MASK) != JOURNAL_WATERMARK_reserved)
+		return stuck;
+
+	spin_lock(&j->lock);
+
+	if (j->can_discard) {
+		spin_unlock(&j->lock);
+		return stuck;
+	}
+
+	stuck = true;
+
+	/*
+	 * The journal shutdown path will set ->err_seq, but do it here first to
+	 * serialize against concurrent failures and avoid duplicate error
+	 * reports.
+	 */
+	if (j->err_seq) {
+		spin_unlock(&j->lock);
+		return stuck;
+	}
+	j->err_seq = journal_cur_seq(j);
+	spin_unlock(&j->lock);
+
+	bch_err(c, "Journal stuck! Hava a pre-reservation but journal full (error %s)",
+		bch2_journal_errors[error]);
+	bch2_journal_debug_to_text(&buf, j);
+	bch_err(c, "%s", buf.buf);
+
+	printbuf_reset(&buf);
+	bch2_journal_pins_to_text(&buf, j);
+	bch_err(c, "Journal pins:\n%s", buf.buf);
+	printbuf_exit(&buf);
+
+	bch2_fatal_error(c);
+	dump_stack();
+
+	return stuck;
+}
+
 /* journal entry close/open: */
 
 void __bch2_journal_buf_put(struct journal *j)
@@ -162,6 +223,7 @@ void bch2_journal_halt(struct journal *j)
 	__journal_entry_close(j, JOURNAL_ENTRY_ERROR_VAL);
 	if (!j->err_seq)
 		j->err_seq = journal_cur_seq(j);
+	journal_wake(j);
 	spin_unlock(&j->lock);
 }
 
@@ -199,12 +261,6 @@ static bool journal_entry_close(struct journal *j)
 /*
  * should _only_ called from journal_res_get() - when we actually want a
  * journal reservation - journal entry is open means journal is dirty:
- *
- * returns:
- * 0:		success
- * -ENOSPC:	journal currently full, must invoke reclaim
- * -EAGAIN:	journal blocked, must wait
- * -EROFS:	insufficient rw devices or journal error
  */
 static int journal_entry_open(struct journal *j)
 {
@@ -231,7 +287,7 @@ static int journal_entry_open(struct journal *j)
 	if (!fifo_free(&j->pin))
 		return JOURNAL_ERR_journal_pin_full;
 
-	if (nr_unwritten_journal_entries(j) == ARRAY_SIZE(j->buf) - 1)
+	if (nr_unwritten_journal_entries(j) == ARRAY_SIZE(j->buf))
 		return JOURNAL_ERR_max_in_flight;
 
 	BUG_ON(!j->cur_entry_sectors);
@@ -250,7 +306,7 @@ static int journal_entry_open(struct journal *j)
 		journal_entry_overhead(j);
 	u64s = clamp_t(int, u64s, 0, JOURNAL_ENTRY_CLOSED_VAL - 1);
 
-	if (u64s <= 0)
+	if (u64s <= (ssize_t) j->early_journal_entries.nr)
 		return JOURNAL_ERR_journal_full;
 
 	if (fifo_empty(&j->pin) && j->reclaim_thread)
@@ -275,6 +331,12 @@ static int journal_entry_open(struct journal *j)
 	buf->data->seq	= cpu_to_le64(journal_cur_seq(j));
 	buf->data->u64s	= 0;
 
+	if (j->early_journal_entries.nr) {
+		memcpy(buf->data->_data, j->early_journal_entries.data,
+		       j->early_journal_entries.nr * sizeof(u64));
+		le32_add_cpu(&buf->data->u64s, j->early_journal_entries.nr);
+	}
+
 	/*
 	 * Must be set before marking the journal entry as open:
 	 */
@@ -291,7 +353,9 @@ static int journal_entry_open(struct journal *j)
 		BUG_ON(new.idx != (journal_cur_seq(j) & JOURNAL_BUF_MASK));
 
 		journal_state_inc(&new);
-		new.cur_entry_offset = 0;
+
+		/* Handle any already added entries */
+		new.cur_entry_offset = le32_to_cpu(buf->data->u64s);
 	} while ((v = atomic64_cmpxchg(&j->reservations.counter,
 				       old.v, new.v)) != old.v);
 
@@ -304,6 +368,9 @@ static int journal_entry_open(struct journal *j)
 			 &j->write_work,
 			 msecs_to_jiffies(c->opts.journal_flush_delay));
 	journal_wake(j);
+
+	if (j->early_journal_entries.nr)
+		darray_exit(&j->early_journal_entries);
 	return 0;
 }
 
@@ -353,10 +420,16 @@ retry:
 		return 0;
 
 	if (bch2_journal_error(j))
-		return -EROFS;
+		return -BCH_ERR_erofs_journal_err;
 
 	spin_lock(&j->lock);
 
+	/* check once more in case somebody else shut things down... */
+	if (bch2_journal_error(j)) {
+		spin_unlock(&j->lock);
+		return -BCH_ERR_erofs_journal_err;
+	}
+
 	/*
 	 * Recheck after taking the lock, so we don't race with another thread
 	 * that just did journal_entry_open() and call journal_entry_close()
@@ -391,12 +464,12 @@ retry:
 	ret = journal_entry_open(j);
 
 	if (ret == JOURNAL_ERR_max_in_flight)
-		trace_journal_entry_full(c);
+		trace_and_count(c, journal_entry_full, c);
 unlock:
 	if ((ret && ret != JOURNAL_ERR_insufficient_devices) &&
 	    !j->res_get_blocked_start) {
 		j->res_get_blocked_start = local_clock() ?: 1;
-		trace_journal_full(c);
+		trace_and_count(c, journal_full, c);
 	}
 
 	can_discard = j->can_discard;
@@ -404,28 +477,8 @@ unlock:
 
 	if (!ret)
 		goto retry;
-
-	if ((ret == JOURNAL_ERR_journal_full ||
-	     ret == JOURNAL_ERR_journal_pin_full) &&
-	    !can_discard &&
-	    !nr_unwritten_journal_entries(j) &&
-	    (flags & JOURNAL_WATERMARK_MASK) == JOURNAL_WATERMARK_reserved) {
-		struct printbuf buf = PRINTBUF;
-
-		bch_err(c, "Journal stuck! Hava a pre-reservation but journal full (ret %s)",
-			bch2_journal_errors[ret]);
-
-		bch2_journal_debug_to_text(&buf, j);
-		bch_err(c, "%s", buf.buf);
-
-		printbuf_reset(&buf);
-		bch2_journal_pins_to_text(&buf, j);
-		bch_err(c, "Journal pins:\n%s", buf.buf);
-
-		printbuf_exit(&buf);
-		bch2_fatal_error(c);
-		dump_stack();
-	}
+	if (journal_error_check_stuck(j, ret, flags))
+		ret = -BCH_ERR_journal_res_get_blocked;
 
 	/*
 	 * Journal is full - can't rely on reclaim from work item due to
@@ -445,7 +498,9 @@ unlock:
 		}
 	}
 
-	return ret == JOURNAL_ERR_insufficient_devices ? -EROFS : -EAGAIN;
+	return ret == JOURNAL_ERR_insufficient_devices
+		? -EROFS
+		: -BCH_ERR_journal_res_get_blocked;
 }
 
 /*
@@ -464,7 +519,8 @@ int bch2_journal_res_get_slowpath(struct journal *j, struct journal_res *res,
 	int ret;
 
 	closure_wait_event(&j->async_wait,
-		   (ret = __journal_res_get(j, res, flags)) != -EAGAIN ||
+		   (ret = __journal_res_get(j, res, flags)) !=
+		   -BCH_ERR_journal_res_get_blocked||
 		   (flags & JOURNAL_RES_GET_NONBLOCK));
 	return ret;
 }
@@ -720,39 +776,6 @@ int bch2_journal_meta(struct journal *j)
 	return bch2_journal_flush_seq(j, res.seq);
 }
 
-int bch2_journal_log_msg(struct journal *j, const char *fmt, ...)
-{
-	struct jset_entry_log *entry;
-	struct journal_res res = { 0 };
-	unsigned msglen, u64s;
-	va_list args;
-	int ret;
-
-	va_start(args, fmt);
-	msglen = vsnprintf(NULL, 0, fmt, args) + 1;
-	va_end(args);
-
-	u64s = jset_u64s(DIV_ROUND_UP(msglen, sizeof(u64)));
-
-	ret = bch2_journal_res_get(j, &res, u64s, 0);
-	if (ret)
-		return ret;
-
-	entry = container_of(journal_res_entry(j, &res),
-			     struct jset_entry_log, entry);;
-	memset(entry, 0, u64s * sizeof(u64));
-	entry->entry.type = BCH_JSET_ENTRY_log;
-	entry->entry.u64s = u64s - 1;
-
-	va_start(args, fmt);
-	vsnprintf(entry->d, INT_MAX, fmt, args);
-	va_end(args);
-
-	bch2_journal_res_put(j, &res);
-
-	return bch2_journal_flush_seq(j, res.seq);
-}
-
 /* block/unlock the journal: */
 
 void bch2_journal_unblock(struct journal *j)
@@ -783,40 +806,40 @@ static int __bch2_set_nr_journal_buckets(struct bch_dev *ca, unsigned nr,
 	u64 *new_bucket_seq = NULL, *new_buckets = NULL;
 	struct open_bucket **ob = NULL;
 	long *bu = NULL;
-	unsigned i, nr_got = 0, nr_want = nr - ja->nr;
-	unsigned old_nr			= ja->nr;
-	unsigned old_discard_idx	= ja->discard_idx;
-	unsigned old_dirty_idx_ondisk	= ja->dirty_idx_ondisk;
-	unsigned old_dirty_idx		= ja->dirty_idx;
-	unsigned old_cur_idx		= ja->cur_idx;
+	unsigned i, pos, nr_got = 0, nr_want = nr - ja->nr;
 	int ret = 0;
 
-	if (c) {
-		bch2_journal_block(&c->journal);
-		bch2_journal_flush_all_pins(&c->journal);
-	}
+	BUG_ON(nr <= ja->nr);
 
-	bu		= kzalloc(nr_want * sizeof(*bu), GFP_KERNEL);
-	ob		= kzalloc(nr_want * sizeof(*ob), GFP_KERNEL);
-	new_buckets	= kzalloc(nr * sizeof(u64), GFP_KERNEL);
-	new_bucket_seq	= kzalloc(nr * sizeof(u64), GFP_KERNEL);
+	bu		= kcalloc(nr_want, sizeof(*bu), GFP_KERNEL);
+	ob		= kcalloc(nr_want, sizeof(*ob), GFP_KERNEL);
+	new_buckets	= kcalloc(nr, sizeof(u64), GFP_KERNEL);
+	new_bucket_seq	= kcalloc(nr, sizeof(u64), GFP_KERNEL);
 	if (!bu || !ob || !new_buckets || !new_bucket_seq) {
-		ret = -ENOMEM;
-		goto err_unblock;
+		ret = -BCH_ERR_ENOMEM_set_nr_journal_buckets;
+		goto err_free;
 	}
 
 	for (nr_got = 0; nr_got < nr_want; nr_got++) {
 		if (new_fs) {
 			bu[nr_got] = bch2_bucket_alloc_new_fs(ca);
 			if (bu[nr_got] < 0) {
-				ret = -ENOSPC;
+				ret = -BCH_ERR_ENOSPC_bucket_alloc;
 				break;
 			}
 		} else {
-			ob[nr_got] = bch2_bucket_alloc(c, ca, RESERVE_none,
-					       false, cl);
-			if (IS_ERR(ob[nr_got])) {
-				ret = cl ? -EAGAIN : -ENOSPC;
+			ob[nr_got] = bch2_bucket_alloc(c, ca, RESERVE_none, cl);
+			ret = PTR_ERR_OR_ZERO(ob[nr_got]);
+			if (ret)
+				break;
+
+			ret = bch2_trans_run(c,
+				bch2_trans_mark_metadata_bucket(&trans, ca,
+						ob[nr_got]->bucket, BCH_DATA_journal,
+						ca->mi.bucket_size));
+			if (ret) {
+				bch2_open_bucket_put(c, ob[nr_got]);
+				bch_err(c, "error marking new journal buckets: %s", bch2_err_str(ret));
 				break;
 			}
 
@@ -825,76 +848,77 @@ static int __bch2_set_nr_journal_buckets(struct bch_dev *ca, unsigned nr,
 	}
 
 	if (!nr_got)
-		goto err_unblock;
+		goto err_free;
 
-	/*
-	 * We may be called from the device add path, before the new device has
-	 * actually been added to the running filesystem:
-	 */
-	if (!new_fs)
-		spin_lock(&c->journal.lock);
+	/* Don't return an error if we successfully allocated some buckets: */
+	ret = 0;
+
+	if (c) {
+		bch2_journal_flush_all_pins(&c->journal);
+		bch2_journal_block(&c->journal);
+		mutex_lock(&c->sb_lock);
+	}
 
 	memcpy(new_buckets,	ja->buckets,	ja->nr * sizeof(u64));
 	memcpy(new_bucket_seq,	ja->bucket_seq,	ja->nr * sizeof(u64));
-	swap(new_buckets,	ja->buckets);
-	swap(new_bucket_seq,	ja->bucket_seq);
+
+	BUG_ON(ja->discard_idx > ja->nr);
+
+	pos = ja->discard_idx ?: ja->nr;
+
+	memmove(new_buckets + pos + nr_got,
+		new_buckets + pos,
+		sizeof(new_buckets[0]) * (ja->nr - pos));
+	memmove(new_bucket_seq + pos + nr_got,
+		new_bucket_seq + pos,
+		sizeof(new_bucket_seq[0]) * (ja->nr - pos));
 
 	for (i = 0; i < nr_got; i++) {
-		unsigned pos = ja->discard_idx ?: ja->nr;
-		long b = bu[i];
-
-		__array_insert_item(ja->buckets,		ja->nr, pos);
-		__array_insert_item(ja->bucket_seq,		ja->nr, pos);
-		ja->nr++;
-
-		ja->buckets[pos] = b;
-		ja->bucket_seq[pos] = 0;
-
-		if (pos <= ja->discard_idx)
-			ja->discard_idx = (ja->discard_idx + 1) % ja->nr;
-		if (pos <= ja->dirty_idx_ondisk)
-			ja->dirty_idx_ondisk = (ja->dirty_idx_ondisk + 1) % ja->nr;
-		if (pos <= ja->dirty_idx)
-			ja->dirty_idx = (ja->dirty_idx + 1) % ja->nr;
-		if (pos <= ja->cur_idx)
-			ja->cur_idx = (ja->cur_idx + 1) % ja->nr;
+		new_buckets[pos + i] = bu[i];
+		new_bucket_seq[pos + i] = 0;
 	}
 
-	ret = bch2_journal_buckets_to_sb(c, ca);
-	if (ret) {
-		/* Revert: */
-		swap(new_buckets,	ja->buckets);
-		swap(new_bucket_seq,	ja->bucket_seq);
-		ja->nr			= old_nr;
-		ja->discard_idx		= old_discard_idx;
-		ja->dirty_idx_ondisk	= old_dirty_idx_ondisk;
-		ja->dirty_idx		= old_dirty_idx;
-		ja->cur_idx		= old_cur_idx;
-	}
+	nr = ja->nr + nr_got;
+
+	ret = bch2_journal_buckets_to_sb(c, ca, new_buckets, nr);
+	if (ret)
+		goto err_unblock;
 
 	if (!new_fs)
-		spin_unlock(&c->journal.lock);
+		bch2_write_super(c);
 
+	/* Commit: */
 	if (c)
-		bch2_journal_unblock(&c->journal);
+		spin_lock(&c->journal.lock);
 
-	if (ret)
-		goto err;
+	swap(new_buckets,	ja->buckets);
+	swap(new_bucket_seq,	ja->bucket_seq);
+	ja->nr = nr;
 
-	if (!new_fs) {
-		for (i = 0; i < nr_got; i++) {
-			ret = bch2_trans_do(c, NULL, NULL, BTREE_INSERT_NOFAIL,
-				bch2_trans_mark_metadata_bucket(&trans, ca,
-						bu[i], BCH_DATA_journal,
-						ca->mi.bucket_size));
-			if (ret) {
-				bch2_fs_inconsistent(c, "error marking new journal buckets: %i", ret);
-				goto err;
-			}
-		}
+	if (pos <= ja->discard_idx)
+		ja->discard_idx = (ja->discard_idx + nr_got) % ja->nr;
+	if (pos <= ja->dirty_idx_ondisk)
+		ja->dirty_idx_ondisk = (ja->dirty_idx_ondisk + nr_got) % ja->nr;
+	if (pos <= ja->dirty_idx)
+		ja->dirty_idx = (ja->dirty_idx + nr_got) % ja->nr;
+	if (pos <= ja->cur_idx)
+		ja->cur_idx = (ja->cur_idx + nr_got) % ja->nr;
+
+	if (c)
+		spin_unlock(&c->journal.lock);
+err_unblock:
+	if (c) {
+		bch2_journal_unblock(&c->journal);
+		mutex_unlock(&c->sb_lock);
 	}
-err:
-	if (ob && !new_fs)
+
+	if (ret && !new_fs)
+		for (i = 0; i < nr_got; i++)
+			bch2_trans_run(c,
+				bch2_trans_mark_metadata_bucket(&trans, ca,
+						bu[i], BCH_DATA_free, 0));
+err_free:
+	if (!new_fs)
 		for (i = 0; i < nr_got; i++)
 			bch2_open_bucket_put(c, ob[i]);
 
@@ -902,12 +926,7 @@ err:
 	kfree(new_buckets);
 	kfree(ob);
 	kfree(bu);
-
 	return ret;
-err_unblock:
-	if (c)
-		bch2_journal_unblock(&c->journal);
-	goto err;
 }
 
 /*
@@ -919,45 +938,49 @@ int bch2_set_nr_journal_buckets(struct bch_fs *c, struct bch_dev *ca,
 {
 	struct journal_device *ja = &ca->journal;
 	struct closure cl;
-	unsigned current_nr;
 	int ret = 0;
 
-	/* don't handle reducing nr of buckets yet: */
-	if (nr < ja->nr)
-		return 0;
-
 	closure_init_stack(&cl);
 
-	while (ja->nr != nr && (ret == 0 || ret == -EAGAIN)) {
-		struct disk_reservation disk_res = { 0, 0 };
+	down_write(&c->state_lock);
 
-		closure_sync(&cl);
+	/* don't handle reducing nr of buckets yet: */
+	if (nr < ja->nr)
+		goto unlock;
 
-		mutex_lock(&c->sb_lock);
-		current_nr = ja->nr;
+	while (ja->nr < nr) {
+		struct disk_reservation disk_res = { 0, 0 };
 
 		/*
 		 * note: journal buckets aren't really counted as _sectors_ used yet, so
 		 * we don't need the disk reservation to avoid the BUG_ON() in buckets.c
 		 * when space used goes up without a reservation - but we do need the
 		 * reservation to ensure we'll actually be able to allocate:
+		 *
+		 * XXX: that's not right, disk reservations only ensure a
+		 * filesystem-wide allocation will succeed, this is a device
+		 * specific allocation - we can hang here:
 		 */
 
-		if (bch2_disk_reservation_get(c, &disk_res,
-					      bucket_to_sector(ca, nr - ja->nr), 1, 0)) {
-			mutex_unlock(&c->sb_lock);
-			return -ENOSPC;
-		}
+		ret = bch2_disk_reservation_get(c, &disk_res,
+						bucket_to_sector(ca, nr - ja->nr), 1, 0);
+		if (ret)
+			break;
 
 		ret = __bch2_set_nr_journal_buckets(ca, nr, false, &cl);
 
 		bch2_disk_reservation_put(c, &disk_res);
 
-		if (ja->nr != current_nr)
-			bch2_write_super(c);
-		mutex_unlock(&c->sb_lock);
+		closure_sync(&cl);
+
+		if (ret && ret != -BCH_ERR_bucket_alloc_blocked)
+			break;
 	}
 
+	if (ret)
+		bch_err(c, "%s: err %s", __func__, bch2_err_str(ret));
+unlock:
+	up_write(&c->state_lock);
 	return ret;
 }
 
@@ -966,7 +989,7 @@ int bch2_dev_journal_alloc(struct bch_dev *ca)
 	unsigned nr;
 
 	if (dynamic_fault("bcachefs:add:journal_alloc"))
-		return -ENOMEM;
+		return -BCH_ERR_ENOMEM_set_nr_journal_buckets;
 
 	/* 1/128th of the device by default: */
 	nr = ca->mi.nbuckets >> 7;
@@ -996,7 +1019,7 @@ static bool bch2_journal_writing_to_device(struct journal *j, unsigned dev_idx)
 	     seq++) {
 		struct journal_buf *buf = journal_seq_to_buf(j, seq);
 
-		if (bch2_bkey_has_device(bkey_i_to_s_c(&buf->key), dev_idx))
+		if (bch2_bkey_has_device_c(bkey_i_to_s_c(&buf->key), dev_idx))
 			ret = true;
 	}
 	spin_unlock(&j->lock);
@@ -1031,17 +1054,25 @@ void bch2_fs_journal_stop(struct journal *j)
 	cancel_delayed_work_sync(&j->write_work);
 }
 
-int bch2_fs_journal_start(struct journal *j, u64 cur_seq,
-			  struct list_head *journal_entries)
+int bch2_fs_journal_start(struct journal *j, u64 cur_seq)
 {
 	struct bch_fs *c = container_of(j, struct bch_fs, journal);
 	struct journal_entry_pin_list *p;
-	struct journal_replay *i;
+	struct journal_replay *i, **_i;
+	struct genradix_iter iter;
+	bool had_entries = false;
+	unsigned ptr;
 	u64 last_seq = cur_seq, nr, seq;
 
-	if (!list_empty(journal_entries))
-		last_seq = le64_to_cpu(list_last_entry(journal_entries,
-				struct journal_replay, list)->j.last_seq);
+	genradix_for_each_reverse(&c->journal_entries, iter, _i) {
+		i = *_i;
+
+		if (!i || i->ignore)
+			continue;
+
+		last_seq = le64_to_cpu(i->j.last_seq);
+		break;
+	}
 
 	nr = cur_seq - last_seq;
 
@@ -1050,7 +1081,7 @@ int bch2_fs_journal_start(struct journal *j, u64 cur_seq,
 		init_fifo(&j->pin, roundup_pow_of_two(nr + 1), GFP_KERNEL);
 		if (!j->pin.data) {
 			bch_err(c, "error reallocating journal fifo (%llu open entries)", nr);
-			return -ENOMEM;
+			return -BCH_ERR_ENOMEM_journal_pin_fifo;
 		}
 	}
 
@@ -1063,14 +1094,14 @@ int bch2_fs_journal_start(struct journal *j, u64 cur_seq,
 	j->pin.back		= cur_seq;
 	atomic64_set(&j->seq, cur_seq - 1);
 
-	if (list_empty(journal_entries))
-		j->last_empty_seq = cur_seq - 1;
-
 	fifo_for_each_entry_ptr(p, &j->pin, seq)
 		journal_pin_list_init(p, 1);
 
-	list_for_each_entry(i, journal_entries, list) {
-		unsigned ptr;
+	genradix_for_each(&c->journal_entries, iter, _i) {
+		i = *_i;
+
+		if (!i || i->ignore)
+			continue;
 
 		seq = le64_to_cpu(i->j.seq);
 		BUG_ON(seq >= cur_seq);
@@ -1086,9 +1117,11 @@ int bch2_fs_journal_start(struct journal *j, u64 cur_seq,
 		p->devs.nr = 0;
 		for (ptr = 0; ptr < i->nr_ptrs; ptr++)
 			bch2_dev_list_add_dev(&p->devs, i->ptrs[ptr].dev);
+
+		had_entries = true;
 	}
 
-	if (list_empty(journal_entries))
+	if (!had_entries)
 		j->last_empty_seq = cur_seq;
 
 	spin_lock(&j->lock);
@@ -1127,7 +1160,7 @@ int bch2_dev_journal_init(struct bch_dev *ca, struct bch_sb *sb)
 		bch2_sb_get_journal(sb);
 	struct bch_sb_field_journal_v2 *journal_buckets_v2 =
 		bch2_sb_get_journal_v2(sb);
-	unsigned i;
+	unsigned i, nr_bvecs;
 
 	ja->nr = 0;
 
@@ -1142,16 +1175,19 @@ int bch2_dev_journal_init(struct bch_dev *ca, struct bch_sb *sb)
 
 	ja->bucket_seq = kcalloc(ja->nr, sizeof(u64), GFP_KERNEL);
 	if (!ja->bucket_seq)
-		return -ENOMEM;
+		return -BCH_ERR_ENOMEM_dev_journal_init;
+
+	nr_bvecs = DIV_ROUND_UP(JOURNAL_ENTRY_SIZE_MAX, PAGE_SIZE);
 
-	ca->journal.bio = bio_kmalloc(GFP_KERNEL,
-			DIV_ROUND_UP(JOURNAL_ENTRY_SIZE_MAX, PAGE_SIZE));
+	ca->journal.bio = bio_kmalloc(nr_bvecs, GFP_KERNEL);
 	if (!ca->journal.bio)
-		return -ENOMEM;
+		return -BCH_ERR_ENOMEM_dev_journal_init;
+
+	bio_init(ca->journal.bio, NULL, ca->journal.bio->bi_inline_vecs, nr_bvecs, 0);
 
 	ja->buckets = kcalloc(ja->nr, sizeof(u64), GFP_KERNEL);
 	if (!ja->buckets)
-		return -ENOMEM;
+		return -BCH_ERR_ENOMEM_dev_journal_init;
 
 	if (journal_buckets_v2) {
 		unsigned nr = bch2_sb_field_journal_v2_nr_entries(journal_buckets_v2);
@@ -1173,6 +1209,8 @@ void bch2_fs_journal_exit(struct journal *j)
 {
 	unsigned i;
 
+	darray_exit(&j->early_journal_entries);
+
 	for (i = 0; i < ARRAY_SIZE(j->buf); i++)
 		kvpfree(j->buf[i].data, j->buf[i].buf_size);
 	free_fifo(&j->pin);
@@ -1203,7 +1241,7 @@ int bch2_fs_journal_init(struct journal *j)
 		 { .cur_entry_offset = JOURNAL_ENTRY_CLOSED_VAL }).v);
 
 	if (!(init_fifo(&j->pin, JOURNAL_PIN, GFP_KERNEL))) {
-		ret = -ENOMEM;
+		ret = -BCH_ERR_ENOMEM_journal_pin_fifo;
 		goto out;
 	}
 
@@ -1211,7 +1249,7 @@ int bch2_fs_journal_init(struct journal *j)
 		j->buf[i].buf_size = JOURNAL_ENTRY_SIZE_MIN;
 		j->buf[i].data = kvpmalloc(j->buf[i].buf_size, GFP_KERNEL);
 		if (!j->buf[i].data) {
-			ret = -ENOMEM;
+			ret = -BCH_ERR_ENOMEM_journal_buf;
 			goto out;
 		}
 	}
@@ -1233,90 +1271,91 @@ void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
 	u64 seq;
 	unsigned i;
 
+	if (!out->nr_tabstops)
+		printbuf_tabstop_push(out, 24);
 	out->atomic++;
-	out->tabstops[0] = 24;
 
 	rcu_read_lock();
 	s = READ_ONCE(j->reservations);
 
-	pr_buf(out, "dirty journal entries:\t%llu/%llu\n",fifo_used(&j->pin), j->pin.size);
-	pr_buf(out, "seq:\t\t\t%llu\n",			journal_cur_seq(j));
-	pr_buf(out, "seq_ondisk:\t\t%llu\n",		j->seq_ondisk);
-	pr_buf(out, "last_seq:\t\t%llu\n",		journal_last_seq(j));
-	pr_buf(out, "last_seq_ondisk:\t%llu\n",		j->last_seq_ondisk);
-	pr_buf(out, "flushed_seq_ondisk:\t%llu\n",	j->flushed_seq_ondisk);
-	pr_buf(out, "prereserved:\t\t%u/%u\n",		j->prereserved.reserved, j->prereserved.remaining);
-	pr_buf(out, "watermark:\t\t%s\n",		bch2_journal_watermarks[j->watermark]);
-	pr_buf(out, "each entry reserved:\t%u\n",	j->entry_u64s_reserved);
-	pr_buf(out, "nr flush writes:\t%llu\n",		j->nr_flush_writes);
-	pr_buf(out, "nr noflush writes:\t%llu\n",	j->nr_noflush_writes);
-	pr_buf(out, "nr direct reclaim:\t%llu\n",	j->nr_direct_reclaim);
-	pr_buf(out, "nr background reclaim:\t%llu\n",	j->nr_background_reclaim);
-	pr_buf(out, "reclaim kicked:\t\t%u\n",		j->reclaim_kicked);
-	pr_buf(out, "reclaim runs in:\t%u ms\n",	time_after(j->next_reclaim, now)
+	prt_printf(out, "dirty journal entries:\t%llu/%llu\n",	fifo_used(&j->pin), j->pin.size);
+	prt_printf(out, "seq:\t\t\t%llu\n",			journal_cur_seq(j));
+	prt_printf(out, "seq_ondisk:\t\t%llu\n",		j->seq_ondisk);
+	prt_printf(out, "last_seq:\t\t%llu\n",		journal_last_seq(j));
+	prt_printf(out, "last_seq_ondisk:\t%llu\n",		j->last_seq_ondisk);
+	prt_printf(out, "flushed_seq_ondisk:\t%llu\n",	j->flushed_seq_ondisk);
+	prt_printf(out, "prereserved:\t\t%u/%u\n",		j->prereserved.reserved, j->prereserved.remaining);
+	prt_printf(out, "watermark:\t\t%s\n",		bch2_journal_watermarks[j->watermark]);
+	prt_printf(out, "each entry reserved:\t%u\n",	j->entry_u64s_reserved);
+	prt_printf(out, "nr flush writes:\t%llu\n",		j->nr_flush_writes);
+	prt_printf(out, "nr noflush writes:\t%llu\n",	j->nr_noflush_writes);
+	prt_printf(out, "nr direct reclaim:\t%llu\n",	j->nr_direct_reclaim);
+	prt_printf(out, "nr background reclaim:\t%llu\n",	j->nr_background_reclaim);
+	prt_printf(out, "reclaim kicked:\t\t%u\n",		j->reclaim_kicked);
+	prt_printf(out, "reclaim runs in:\t%u ms\n",	time_after(j->next_reclaim, now)
 	       ? jiffies_to_msecs(j->next_reclaim - jiffies) : 0);
-	pr_buf(out, "current entry sectors:\t%u\n",	j->cur_entry_sectors);
-	pr_buf(out, "current entry error:\t%s\n",	bch2_journal_errors[j->cur_entry_error]);
-	pr_buf(out, "current entry:\t\t");
+	prt_printf(out, "current entry sectors:\t%u\n",	j->cur_entry_sectors);
+	prt_printf(out, "current entry error:\t%s\n",	bch2_journal_errors[j->cur_entry_error]);
+	prt_printf(out, "current entry:\t\t");
 
 	switch (s.cur_entry_offset) {
 	case JOURNAL_ENTRY_ERROR_VAL:
-		pr_buf(out, "error");
+		prt_printf(out, "error");
 		break;
 	case JOURNAL_ENTRY_CLOSED_VAL:
-		pr_buf(out, "closed");
+		prt_printf(out, "closed");
 		break;
 	default:
-		pr_buf(out, "%u/%u", s.cur_entry_offset, j->cur_entry_u64s);
+		prt_printf(out, "%u/%u", s.cur_entry_offset, j->cur_entry_u64s);
 		break;
 	}
 
-	pr_newline(out);
+	prt_newline(out);
 
 	for (seq = journal_cur_seq(j);
 	     seq >= journal_last_unwritten_seq(j);
 	     --seq) {
 		i = seq & JOURNAL_BUF_MASK;
 
-		pr_buf(out, "unwritten entry:");
-		pr_tab(out);
-		pr_buf(out, "%llu", seq);
-		pr_newline(out);
-		pr_indent_push(out, 2);
+		prt_printf(out, "unwritten entry:");
+		prt_tab(out);
+		prt_printf(out, "%llu", seq);
+		prt_newline(out);
+		printbuf_indent_add(out, 2);
 
-		pr_buf(out, "refcount:");
-		pr_tab(out);
-		pr_buf(out, "%u", journal_state_count(s, i));
-		pr_newline(out);
+		prt_printf(out, "refcount:");
+		prt_tab(out);
+		prt_printf(out, "%u", journal_state_count(s, i));
+		prt_newline(out);
 
-		pr_buf(out, "sectors:");
-		pr_tab(out);
-		pr_buf(out, "%u", j->buf[i].sectors);
-		pr_newline(out);
+		prt_printf(out, "sectors:");
+		prt_tab(out);
+		prt_printf(out, "%u", j->buf[i].sectors);
+		prt_newline(out);
 
-		pr_buf(out, "expires");
-		pr_tab(out);
-		pr_buf(out, "%li jiffies", j->buf[i].expires - jiffies);
-		pr_newline(out);
+		prt_printf(out, "expires");
+		prt_tab(out);
+		prt_printf(out, "%li jiffies", j->buf[i].expires - jiffies);
+		prt_newline(out);
 
-		pr_indent_pop(out, 2);
+		printbuf_indent_sub(out, 2);
 	}
 
-	pr_buf(out,
+	prt_printf(out,
 	       "replay done:\t\t%i\n",
 	       test_bit(JOURNAL_REPLAY_DONE,	&j->flags));
 
-	pr_buf(out, "space:\n");
-	pr_buf(out, "\tdiscarded\t%u:%u\n",
+	prt_printf(out, "space:\n");
+	prt_printf(out, "\tdiscarded\t%u:%u\n",
 	       j->space[journal_space_discarded].next_entry,
 	       j->space[journal_space_discarded].total);
-	pr_buf(out, "\tclean ondisk\t%u:%u\n",
+	prt_printf(out, "\tclean ondisk\t%u:%u\n",
 	       j->space[journal_space_clean_ondisk].next_entry,
 	       j->space[journal_space_clean_ondisk].total);
-	pr_buf(out, "\tclean\t\t%u:%u\n",
+	prt_printf(out, "\tclean\t\t%u:%u\n",
 	       j->space[journal_space_clean].next_entry,
 	       j->space[journal_space_clean].total);
-	pr_buf(out, "\ttotal\t\t%u:%u\n",
+	prt_printf(out, "\ttotal\t\t%u:%u\n",
 	       j->space[journal_space_total].next_entry,
 	       j->space[journal_space_total].total);
 
@@ -1330,14 +1369,14 @@ void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
 		if (!ja->nr)
 			continue;
 
-		pr_buf(out, "dev %u:\n",		i);
-		pr_buf(out, "\tnr\t\t%u\n",		ja->nr);
-		pr_buf(out, "\tbucket size\t%u\n",	ca->mi.bucket_size);
-		pr_buf(out, "\tavailable\t%u:%u\n",	bch2_journal_dev_buckets_available(j, ja, journal_space_discarded), ja->sectors_free);
-		pr_buf(out, "\tdiscard_idx\t%u\n",	ja->discard_idx);
-		pr_buf(out, "\tdirty_ondisk\t%u (seq %llu)\n", ja->dirty_idx_ondisk,	ja->bucket_seq[ja->dirty_idx_ondisk]);
-		pr_buf(out, "\tdirty_idx\t%u (seq %llu)\n", ja->dirty_idx,		ja->bucket_seq[ja->dirty_idx]);
-		pr_buf(out, "\tcur_idx\t\t%u (seq %llu)\n", ja->cur_idx,		ja->bucket_seq[ja->cur_idx]);
+		prt_printf(out, "dev %u:\n",		i);
+		prt_printf(out, "\tnr\t\t%u\n",		ja->nr);
+		prt_printf(out, "\tbucket size\t%u\n",	ca->mi.bucket_size);
+		prt_printf(out, "\tavailable\t%u:%u\n",	bch2_journal_dev_buckets_available(j, ja, journal_space_discarded), ja->sectors_free);
+		prt_printf(out, "\tdiscard_idx\t%u\n",	ja->discard_idx);
+		prt_printf(out, "\tdirty_ondisk\t%u (seq %llu)\n", ja->dirty_idx_ondisk,	ja->bucket_seq[ja->dirty_idx_ondisk]);
+		prt_printf(out, "\tdirty_idx\t%u (seq %llu)\n", ja->dirty_idx,		ja->bucket_seq[ja->dirty_idx]);
+		prt_printf(out, "\tcur_idx\t\t%u (seq %llu)\n", ja->cur_idx,		ja->bucket_seq[ja->cur_idx]);
 	}
 
 	rcu_read_unlock();
@@ -1356,6 +1395,7 @@ bool bch2_journal_seq_pins_to_text(struct printbuf *out, struct journal *j, u64
 {
 	struct journal_entry_pin_list *pin_list;
 	struct journal_entry_pin *pin;
+	unsigned i;
 
 	spin_lock(&j->lock);
 	*seq = max(*seq, j->pin.front);
@@ -1369,31 +1409,27 @@ bool bch2_journal_seq_pins_to_text(struct printbuf *out, struct journal *j, u64
 
 	pin_list = journal_seq_pin(j, *seq);
 
-	pr_buf(out, "%llu: count %u", *seq, atomic_read(&pin_list->count));
-	pr_newline(out);
-	pr_indent_push(out, 2);
+	prt_printf(out, "%llu: count %u", *seq, atomic_read(&pin_list->count));
+	prt_newline(out);
+	printbuf_indent_add(out, 2);
 
-	list_for_each_entry(pin, &pin_list->list, list) {
-		pr_buf(out, "\t%px %ps", pin, pin->flush);
-		pr_newline(out);
-	}
-
-	list_for_each_entry(pin, &pin_list->key_cache_list, list) {
-		pr_buf(out, "\t%px %ps", pin, pin->flush);
-		pr_newline(out);
-	}
+	for (i = 0; i < ARRAY_SIZE(pin_list->list); i++)
+		list_for_each_entry(pin, &pin_list->list[i], list) {
+			prt_printf(out, "\t%px %ps", pin, pin->flush);
+			prt_newline(out);
+		}
 
 	if (!list_empty(&pin_list->flushed)) {
-		pr_buf(out, "flushed:");
-		pr_newline(out);
+		prt_printf(out, "flushed:");
+		prt_newline(out);
 	}
 
 	list_for_each_entry(pin, &pin_list->flushed, list) {
-		pr_buf(out, "\t%px %ps", pin, pin->flush);
-		pr_newline(out);
+		prt_printf(out, "\t%px %ps", pin, pin->flush);
+		prt_newline(out);
 	}
 
-	pr_indent_pop(out, 2);
+	printbuf_indent_sub(out, 2);
 
 	--out->atomic;
 	spin_unlock(&j->lock);