1 files changed, 137 insertions, 54 deletions

diff --git a/fs/bcachefs/journal.c b/fs/bcachefs/journal.c
index bb45d3634194..dda802a656cf 100644
--- a/fs/bcachefs/journal.c
+++ b/fs/bcachefs/journal.c
@@ -12,6 +12,7 @@
 #include "btree_update.h"
 #include "btree_write_buffer.h"
 #include "buckets.h"
+#include "enumerated_ref.h"
 #include "error.h"
 #include "journal.h"
 #include "journal_io.h"
@@ -173,7 +174,7 @@ journal_error_check_stuck(struct journal *j, int error, unsigned flags)
 	spin_unlock(&j->lock);
 	prt_printf(&buf, bch2_fmt(c, "Journal stuck! Hava a pre-reservation but journal full (error %s)"),
 				  bch2_err_str(error));
-	bch2_print_string_as_lines(KERN_ERR, buf.buf);
+	bch2_print_str(c, KERN_ERR, buf.buf);
 
 	printbuf_reset(&buf);
 	bch2_journal_pins_to_text(&buf, j);
@@ -331,16 +332,6 @@ static void __journal_entry_close(struct journal *j, unsigned closed_val, bool t
 	__bch2_journal_buf_put(j, le64_to_cpu(buf->data->seq));
 }
 
-void bch2_journal_halt(struct journal *j)
-{
-	spin_lock(&j->lock);
-	__journal_entry_close(j, JOURNAL_ENTRY_ERROR_VAL, true);
-	if (!j->err_seq)
-		j->err_seq = journal_cur_seq(j);
-	journal_wake(j);
-	spin_unlock(&j->lock);
-}
-
 void bch2_journal_halt_locked(struct journal *j)
 {
 	lockdep_assert_held(&j->lock);
@@ -351,6 +342,13 @@ void bch2_journal_halt_locked(struct journal *j)
 	journal_wake(j);
 }
 
+void bch2_journal_halt(struct journal *j)
+{
+	spin_lock(&j->lock);
+	bch2_journal_halt_locked(j);
+	spin_unlock(&j->lock);
+}
+
 static bool journal_entry_want_write(struct journal *j)
 {
 	bool ret = !journal_entry_is_open(j) ||
@@ -399,7 +397,7 @@ static int journal_entry_open(struct journal *j)
 	BUG_ON(BCH_SB_CLEAN(c->disk_sb.sb));
 
 	if (j->blocked)
-		return -BCH_ERR_journal_blocked;
+		return bch_err_throw(c, journal_blocked);
 
 	if (j->cur_entry_error)
 		return j->cur_entry_error;
@@ -409,23 +407,23 @@ static int journal_entry_open(struct journal *j)
 		return ret;
 
 	if (!fifo_free(&j->pin))
-		return -BCH_ERR_journal_pin_full;
+		return bch_err_throw(c, journal_pin_full);
 
 	if (nr_unwritten_journal_entries(j) == ARRAY_SIZE(j->buf))
-		return -BCH_ERR_journal_max_in_flight;
+		return bch_err_throw(c, journal_max_in_flight);
 
 	if (atomic64_read(&j->seq) - j->seq_write_started == JOURNAL_STATE_BUF_NR)
-		return -BCH_ERR_journal_max_open;
+		return bch_err_throw(c, journal_max_open);
 
-	if (journal_cur_seq(j) >= JOURNAL_SEQ_MAX) {
+	if (unlikely(journal_cur_seq(j) >= JOURNAL_SEQ_MAX)) {
 		bch_err(c, "cannot start: journal seq overflow");
 		if (bch2_fs_emergency_read_only_locked(c))
 			bch_err(c, "fatal error - emergency read only");
-		return -BCH_ERR_journal_shutdown;
+		return bch_err_throw(c, journal_shutdown);
 	}
 
 	if (!j->free_buf && !buf->data)
-		return -BCH_ERR_journal_buf_enomem; /* will retry after write completion frees up a buf */
+		return bch_err_throw(c, journal_buf_enomem); /* will retry after write completion frees up a buf */
 
 	BUG_ON(!j->cur_entry_sectors);
 
@@ -449,7 +447,7 @@ static int journal_entry_open(struct journal *j)
 	u64s = clamp_t(int, u64s, 0, JOURNAL_ENTRY_CLOSED_VAL - 1);
 
 	if (u64s <= (ssize_t) j->early_journal_entries.nr)
-		return -BCH_ERR_journal_full;
+		return bch_err_throw(c, journal_full);
 
 	if (fifo_empty(&j->pin) && j->reclaim_thread)
 		wake_up_process(j->reclaim_thread);
@@ -461,6 +459,14 @@ static int journal_entry_open(struct journal *j)
 	atomic64_inc(&j->seq);
 	journal_pin_list_init(fifo_push_ref(&j->pin), 1);
 
+	if (unlikely(bch2_journal_seq_is_blacklisted(c, journal_cur_seq(j), false))) {
+		bch_err(c, "attempting to open blacklisted journal seq %llu",
+			journal_cur_seq(j));
+		if (bch2_fs_emergency_read_only_locked(c))
+			bch_err(c, "fatal error - emergency read only");
+		return bch_err_throw(c, journal_shutdown);
+	}
+
 	BUG_ON(j->pin.back - 1 != atomic64_read(&j->seq));
 
 	BUG_ON(j->buf + (journal_cur_seq(j) & JOURNAL_BUF_MASK) != buf);
@@ -591,16 +597,16 @@ retry:
 		return ret;
 
 	if (j->blocked)
-		return -BCH_ERR_journal_blocked;
+		return bch_err_throw(c, journal_blocked);
 
 	if ((flags & BCH_WATERMARK_MASK) < j->watermark) {
-		ret = -BCH_ERR_journal_full;
+		ret = bch_err_throw(c, journal_full);
 		can_discard = j->can_discard;
 		goto out;
 	}
 
 	if (nr_unwritten_journal_entries(j) == ARRAY_SIZE(j->buf) && !journal_entry_is_open(j)) {
-		ret = -BCH_ERR_journal_max_in_flight;
+		ret = bch_err_throw(c, journal_max_in_flight);
 		goto out;
 	}
 
@@ -641,7 +647,7 @@ out:
 		goto retry;
 
 	if (journal_error_check_stuck(j, ret, flags))
-		ret = -BCH_ERR_journal_stuck;
+		ret = bch_err_throw(c, journal_stuck);
 
 	if (ret == -BCH_ERR_journal_max_in_flight &&
 	    track_event_change(&c->times[BCH_TIME_blocked_journal_max_in_flight], true) &&
@@ -702,7 +708,8 @@ static unsigned max_dev_latency(struct bch_fs *c)
 {
 	u64 nsecs = 0;
 
-	for_each_rw_member(c, ca)
+	guard(rcu)();
+	for_each_rw_member_rcu(c, ca)
 		nsecs = max(nsecs, ca->io_latency[WRITE].stats.max_duration);
 
 	return nsecs_to_jiffies(nsecs);
@@ -746,7 +753,7 @@ int bch2_journal_res_get_slowpath(struct journal *j, struct journal_res *res,
 
 	struct printbuf buf = PRINTBUF;
 	bch2_journal_debug_to_text(&buf, j);
-	bch2_print_string_as_lines(KERN_ERR, buf.buf);
+	bch2_print_str(c, KERN_ERR, buf.buf);
 	prt_printf(&buf, bch2_fmt(c, "Journal stuck? Waited for 10 seconds, err %s"), bch2_err_str(ret));
 	printbuf_exit(&buf);
 
@@ -805,6 +812,7 @@ out:
 int bch2_journal_flush_seq_async(struct journal *j, u64 seq,
 				 struct closure *parent)
 {
+	struct bch_fs *c = container_of(j, struct bch_fs, journal);
 	struct journal_buf *buf;
 	int ret = 0;
 
@@ -820,7 +828,7 @@ int bch2_journal_flush_seq_async(struct journal *j, u64 seq,
 
 	/* Recheck under lock: */
 	if (j->err_seq && seq >= j->err_seq) {
-		ret = -BCH_ERR_journal_flush_err;
+		ret = bch_err_throw(c, journal_flush_err);
 		goto out;
 	}
 
@@ -990,11 +998,11 @@ int bch2_journal_meta(struct journal *j)
 {
 	struct bch_fs *c = container_of(j, struct bch_fs, journal);
 
-	if (!bch2_write_ref_tryget(c, BCH_WRITE_REF_journal))
-		return -BCH_ERR_erofs_no_writes;
+	if (!enumerated_ref_tryget(&c->writes, BCH_WRITE_REF_journal))
+		return bch_err_throw(c, erofs_no_writes);
 
 	int ret = __bch2_journal_meta(j);
-	bch2_write_ref_put(c, BCH_WRITE_REF_journal);
+	enumerated_ref_put(&c->writes, BCH_WRITE_REF_journal);
 	return ret;
 }
 
@@ -1124,7 +1132,7 @@ static int bch2_set_nr_journal_buckets_iter(struct bch_dev *ca, unsigned nr,
 	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 = -BCH_ERR_ENOMEM_set_nr_journal_buckets;
+		ret = bch_err_throw(c, ENOMEM_set_nr_journal_buckets);
 		goto err_free;
 	}
 
@@ -1296,13 +1304,83 @@ int bch2_set_nr_journal_buckets(struct bch_fs *c, struct bch_dev *ca,
 	return ret;
 }
 
+int bch2_dev_journal_bucket_delete(struct bch_dev *ca, u64 b)
+{
+	struct bch_fs *c = ca->fs;
+	struct journal *j = &c->journal;
+	struct journal_device *ja = &ca->journal;
+
+	guard(mutex)(&c->sb_lock);
+	unsigned pos;
+	for (pos = 0; pos < ja->nr; pos++)
+		if (ja->buckets[pos] == b)
+			break;
+
+	if (pos == ja->nr) {
+		bch_err(ca, "journal bucket %llu not found when deleting", b);
+		return -EINVAL;
+	}
+
+	u64 *new_buckets = kcalloc(ja->nr, sizeof(u64), GFP_KERNEL);;
+	if (!new_buckets)
+		return bch_err_throw(c, ENOMEM_set_nr_journal_buckets);
+
+	memcpy(new_buckets, ja->buckets, ja->nr * sizeof(u64));
+	memmove(&new_buckets[pos],
+		&new_buckets[pos + 1],
+		(ja->nr - 1 - pos) * sizeof(new_buckets[0]));
+
+	int ret = bch2_journal_buckets_to_sb(c, ca, ja->buckets, ja->nr - 1) ?:
+		bch2_write_super(c);
+	if (ret) {
+		kfree(new_buckets);
+		return ret;
+	}
+
+	scoped_guard(spinlock, &j->lock) {
+		if (pos < ja->discard_idx)
+			--ja->discard_idx;
+		if (pos < ja->dirty_idx_ondisk)
+			--ja->dirty_idx_ondisk;
+		if (pos < ja->dirty_idx)
+			--ja->dirty_idx;
+		if (pos < ja->cur_idx)
+			--ja->cur_idx;
+
+		ja->nr--;
+
+		memmove(&ja->buckets[pos],
+			&ja->buckets[pos + 1],
+			(ja->nr - pos) * sizeof(ja->buckets[0]));
+
+		memmove(&ja->bucket_seq[pos],
+			&ja->bucket_seq[pos + 1],
+			(ja->nr - pos) * sizeof(ja->bucket_seq[0]));
+
+		bch2_journal_space_available(j);
+	}
+
+	kfree(new_buckets);
+	return 0;
+}
+
 int bch2_dev_journal_alloc(struct bch_dev *ca, bool new_fs)
 {
+	struct bch_fs *c = ca->fs;
+
+	if (!(ca->mi.data_allowed & BIT(BCH_DATA_journal)))
+		return 0;
+
+	if (c->sb.features & BIT_ULL(BCH_FEATURE_small_image)) {
+		bch_err(c, "cannot allocate journal, filesystem is an unresized image file");
+		return bch_err_throw(c, erofs_filesystem_full);
+	}
+
 	unsigned nr;
 	int ret;
 
 	if (dynamic_fault("bcachefs:add:journal_alloc")) {
-		ret = -BCH_ERR_ENOMEM_set_nr_journal_buckets;
+		ret = bch_err_throw(c, ENOMEM_set_nr_journal_buckets);
 		goto err;
 	}
 
@@ -1318,7 +1396,7 @@ int bch2_dev_journal_alloc(struct bch_dev *ca, bool new_fs)
 		     min(1 << 13,
 			 (1 << 24) / ca->mi.bucket_size));
 
-	ret = bch2_set_nr_journal_buckets_loop(ca->fs, ca, nr, new_fs);
+	ret = bch2_set_nr_journal_buckets_loop(c, ca, nr, new_fs);
 err:
 	bch_err_fn(ca, ret);
 	return ret;
@@ -1326,13 +1404,14 @@ err:
 
 int bch2_fs_journal_alloc(struct bch_fs *c)
 {
-	for_each_online_member(c, ca) {
+	for_each_online_member(c, ca, BCH_DEV_READ_REF_fs_journal_alloc) {
 		if (ca->journal.nr)
 			continue;
 
 		int ret = bch2_dev_journal_alloc(ca, true);
 		if (ret) {
-			percpu_ref_put(&ca->io_ref[READ]);
+			enumerated_ref_put(&ca->io_ref[READ],
+					   BCH_DEV_READ_REF_fs_journal_alloc);
 			return ret;
 		}
 	}
@@ -1404,6 +1483,13 @@ int bch2_fs_journal_start(struct journal *j, u64 cur_seq)
 	bool had_entries = false;
 	u64 last_seq = cur_seq, nr, seq;
 
+	/*
+	 *
+	 * XXX pick most recent non blacklisted sequence number
+	 */
+
+	cur_seq = max(cur_seq, bch2_journal_last_blacklisted_seq(c));
+
 	if (cur_seq >= JOURNAL_SEQ_MAX) {
 		bch_err(c, "cannot start: journal seq overflow");
 		return -EINVAL;
@@ -1429,13 +1515,11 @@ int bch2_fs_journal_start(struct journal *j, u64 cur_seq)
 	 */
 	nr += nr / 4;
 
-	if (nr + 1 > j->pin.size) {
-		free_fifo(&j->pin);
-		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 -BCH_ERR_ENOMEM_journal_pin_fifo;
-		}
+	nr = max(nr, JOURNAL_PIN);
+	init_fifo(&j->pin, roundup_pow_of_two(nr), GFP_KERNEL);
+	if (!j->pin.data) {
+		bch_err(c, "error reallocating journal fifo (%llu open entries)", nr);
+		return bch_err_throw(c, ENOMEM_journal_pin_fifo);
 	}
 
 	j->replay_journal_seq	= last_seq;
@@ -1523,6 +1607,7 @@ void bch2_dev_journal_exit(struct bch_dev *ca)
 
 int bch2_dev_journal_init(struct bch_dev *ca, struct bch_sb *sb)
 {
+	struct bch_fs *c = ca->fs;
 	struct journal_device *ja = &ca->journal;
 	struct bch_sb_field_journal *journal_buckets =
 		bch2_sb_field_get(sb, journal);
@@ -1542,7 +1627,7 @@ 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 -BCH_ERR_ENOMEM_dev_journal_init;
+		return bch_err_throw(c, ENOMEM_dev_journal_init);
 
 	unsigned nr_bvecs = DIV_ROUND_UP(JOURNAL_ENTRY_SIZE_MAX, PAGE_SIZE);
 
@@ -1550,7 +1635,7 @@ int bch2_dev_journal_init(struct bch_dev *ca, struct bch_sb *sb)
 		ja->bio[i] = kzalloc(struct_size(ja->bio[i], bio.bi_inline_vecs,
 				     nr_bvecs), GFP_KERNEL);
 		if (!ja->bio[i])
-			return -BCH_ERR_ENOMEM_dev_journal_init;
+			return bch_err_throw(c, ENOMEM_dev_journal_init);
 
 		ja->bio[i]->ca = ca;
 		ja->bio[i]->buf_idx = i;
@@ -1559,7 +1644,7 @@ int bch2_dev_journal_init(struct bch_dev *ca, struct bch_sb *sb)
 
 	ja->buckets = kcalloc(ja->nr, sizeof(u64), GFP_KERNEL);
 	if (!ja->buckets)
-		return -BCH_ERR_ENOMEM_dev_journal_init;
+		return bch_err_throw(c, ENOMEM_dev_journal_init);
 
 	if (journal_buckets_v2) {
 		unsigned nr = bch2_sb_field_journal_v2_nr_entries(journal_buckets_v2);
@@ -1590,7 +1675,7 @@ void bch2_fs_journal_exit(struct journal *j)
 	free_fifo(&j->pin);
 }
 
-int bch2_fs_journal_init(struct journal *j)
+void bch2_fs_journal_init_early(struct journal *j)
 {
 	static struct lock_class_key res_key;
 
@@ -1609,24 +1694,24 @@ int bch2_fs_journal_init(struct journal *j)
 	atomic64_set(&j->reservations.counter,
 		((union journal_res_state)
 		 { .cur_entry_offset = JOURNAL_ENTRY_CLOSED_VAL }).v);
+}
 
-	if (!(init_fifo(&j->pin, JOURNAL_PIN, GFP_KERNEL)))
-		return -BCH_ERR_ENOMEM_journal_pin_fifo;
+int bch2_fs_journal_init(struct journal *j)
+{
+	struct bch_fs *c = container_of(j, struct bch_fs, journal);
 
 	j->free_buf_size = j->buf_size_want = JOURNAL_ENTRY_SIZE_MIN;
 	j->free_buf = kvmalloc(j->free_buf_size, GFP_KERNEL);
 	if (!j->free_buf)
-		return -BCH_ERR_ENOMEM_journal_buf;
+		return bch_err_throw(c, ENOMEM_journal_buf);
 
 	for (unsigned i = 0; i < ARRAY_SIZE(j->buf); i++)
 		j->buf[i].idx = i;
 
-	j->pin.front = j->pin.back = 1;
-
 	j->wq = alloc_workqueue("bcachefs_journal",
 				WQ_HIGHPRI|WQ_FREEZABLE|WQ_UNBOUND|WQ_MEM_RECLAIM, 512);
 	if (!j->wq)
-		return -BCH_ERR_ENOMEM_fs_other_alloc;
+		return bch_err_throw(c, ENOMEM_fs_other_alloc);
 	return 0;
 }
 
@@ -1650,7 +1735,7 @@ void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
 	printbuf_tabstop_push(out, 28);
 	out->atomic++;
 
-	rcu_read_lock();
+	guard(rcu)();
 	s = READ_ONCE(j->reservations);
 
 	prt_printf(out, "flags:\t");
@@ -1741,8 +1826,6 @@ void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
 
 	prt_printf(out, "replicas want %u need %u\n", c->opts.metadata_replicas, c->opts.metadata_replicas_required);
 
-	rcu_read_unlock();
-
 	--out->atomic;
 }