bcachefs: opts.journal_rewind

Add a mount option for rewinding the journal, bringing the entire
filesystem to where it was at a previous point in time.

This is for extreme disaster recovery scenarios - it's not intended as
an undelete operation.

The option takes a journal sequence number; the desired sequence number
can be determined with 'bcachefs list_journal'

Caveats:

- The 'journal_transaction_names' option must have been enabled (it's on
  by default). The option controls emitting of extra debug info in the
  journal, so we can see what individual transactions were doing;
  It also enables journalling of keys being overwritten, which is what
  we rely on here.

- A full fsck run will be automatically triggered since alloc info will
  be inconsistent. Only leaf node updates to non-alloc btrees are
  rewound, since rewinding interior btree updates isn't possible or
  desirable.

- We can't do anything about data that was deleted and overwritten.

  Lots of metadata updates after the point in time we're rewinding to
  shouldn't cause a problem, since we segragate data and metadata
  allocations (this is in order to make repair by btree node scan
  practical on larger filesystems; there's a small 64-bit per device
  bitmap in the superblock of device ranges with btree nodes, and we try
  to keep this small).

  However, having discards enabled will cause problems, since buckets
  are discarded as soon as they become empty (this is why we don't
  implement fstrim: we don't need it).

  Hopefully, this feature will be a one-off thing that's never used
  again: this was implemented for recovering from the "vfs i_nlink 0 ->
  subvol deletion" bug, and that bug was unusually disastrous and
  additional safeguards have since been implemented.

  But if it does turn out that we need this more in the future, I'll
  have to implement an option so that empty buckets aren't discarded
  immediately - lagging by perhaps 1% of device capacity.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

fs/bcachefs/btree_journal_iter.c

@@ -641,10 +641,11 @@ static int journal_sort_key_cmp(const void *_l, const void *_r)
 {
 	const struct journal_key *l = _l;
 	const struct journal_key *r = _r;
+	int rewind = l->rewind && r->rewind ? -1 : 1;
 
 	return  journal_key_cmp(l, r) ?:
-		cmp_int(l->journal_seq, r->journal_seq) ?:
-		cmp_int(l->journal_offset, r->journal_offset);
+		((cmp_int(l->journal_seq, r->journal_seq) ?:
+		  cmp_int(l->journal_offset, r->journal_offset)) * rewind);
 }
 
 void bch2_journal_keys_put(struct bch_fs *c)
@@ -713,6 +714,8 @@ int bch2_journal_keys_sort(struct bch_fs *c)
 	struct journal_keys *keys = &c->journal_keys;
 	size_t nr_read = 0;
 
+	u64 rewind_seq = c->opts.journal_rewind ?: U64_MAX;
+
 	genradix_for_each(&c->journal_entries, iter, _i) {
 		i = *_i;
 
@@ -721,28 +724,43 @@ int bch2_journal_keys_sort(struct bch_fs *c)
 
 		cond_resched();
 
-		for_each_jset_key(k, entry, &i->j) {
-			struct journal_key n = (struct journal_key) {
-				.btree_id	= entry->btree_id,
-				.level		= entry->level,
-				.k		= k,
-				.journal_seq	= le64_to_cpu(i->j.seq),
-				.journal_offset	= k->_data - i->j._data,
-			};
+		vstruct_for_each(&i->j, entry) {
+			bool rewind = !entry->level &&
+				!btree_id_is_alloc(entry->btree_id) &&
+				le64_to_cpu(i->j.seq) >= rewind_seq;
 
-			if (darray_push(keys, n)) {
-				__journal_keys_sort(keys);
+			if (entry->type != (rewind
+					    ? BCH_JSET_ENTRY_overwrite
+					    : BCH_JSET_ENTRY_btree_keys))
+				continue;
 
-				if (keys->nr * 8 > keys->size * 7) {
-					bch_err(c, "Too many journal keys for slowpath; have %zu compacted, buf size %zu, processed %zu keys at seq %llu",
-						keys->nr, keys->size, nr_read, le64_to_cpu(i->j.seq));
-					return bch_err_throw(c, ENOMEM_journal_keys_sort);
+			if (!rewind && le64_to_cpu(i->j.seq) < c->journal_replay_seq_start)
+				continue;
+
+			jset_entry_for_each_key(entry, k) {
+				struct journal_key n = (struct journal_key) {
+					.btree_id	= entry->btree_id,
+					.level		= entry->level,
+					.rewind		= rewind,
+					.k		= k,
+					.journal_seq	= le64_to_cpu(i->j.seq),
+					.journal_offset	= k->_data - i->j._data,
+				};
+
+				if (darray_push(keys, n)) {
+					__journal_keys_sort(keys);
+
+					if (keys->nr * 8 > keys->size * 7) {
+						bch_err(c, "Too many journal keys for slowpath; have %zu compacted, buf size %zu, processed %zu keys at seq %llu",
+							keys->nr, keys->size, nr_read, le64_to_cpu(i->j.seq));
+						return bch_err_throw(c, ENOMEM_journal_keys_sort);
+					}
+
+					BUG_ON(darray_push(keys, n));
 				}
 
-				BUG_ON(darray_push(keys, n));
+				nr_read++;
 			}
-
-			nr_read++;
 		}
 	}
 

fs/bcachefs/btree_journal_iter_types.h

@@ -11,8 +11,9 @@ struct journal_key {
 	u32			journal_offset;
 	enum btree_id		btree_id:8;
 	unsigned		level:8;
-	bool			allocated;
-	bool			overwritten;
+	bool			allocated:1;
+	bool			overwritten:1;
+	bool			rewind:1;
 	struct journal_key_range_overwritten __rcu *
 				overwritten_range;
 	struct bkey_i		*k;

fs/bcachefs/journal_io.c

@@ -160,6 +160,9 @@ static int journal_entry_add(struct bch_fs *c, struct bch_dev *ca,
 	struct printbuf buf = PRINTBUF;
 	int ret = JOURNAL_ENTRY_ADD_OK;
 
+	if (last_seq && c->opts.journal_rewind)
+		last_seq = min(last_seq, c->opts.journal_rewind);
+
 	if (!c->journal.oldest_seq_found_ondisk ||
 	    le64_to_cpu(j->seq) < c->journal.oldest_seq_found_ondisk)
 		c->journal.oldest_seq_found_ondisk = le64_to_cpu(j->seq);
@@ -1430,11 +1433,21 @@ int bch2_journal_read(struct bch_fs *c,
 	printbuf_reset(&buf);
 	prt_printf(&buf, "journal read done, replaying entries %llu-%llu",
 		   *last_seq, *blacklist_seq - 1);
+
+	/*
+	 * Drop blacklisted entries and entries older than last_seq (or start of
+	 * journal rewind:
+	 */
+	u64 drop_before = *last_seq;
+	if (c->opts.journal_rewind) {
+		drop_before = min(drop_before, c->opts.journal_rewind);
+		prt_printf(&buf, " (rewinding from %llu)", c->opts.journal_rewind);
+	}
+
+	*last_seq = drop_before;
 	if (*start_seq != *blacklist_seq)
 		prt_printf(&buf, " (unflushed %llu-%llu)", *blacklist_seq, *start_seq - 1);
 	bch_info(c, "%s", buf.buf);
-
-	/* Drop blacklisted entries and entries older than last_seq: */
 	genradix_for_each(&c->journal_entries, radix_iter, _i) {
 		i = *_i;
 
@@ -1442,7 +1455,7 @@ int bch2_journal_read(struct bch_fs *c,
 			continue;
 
 		seq = le64_to_cpu(i->j.seq);
-		if (seq < *last_seq) {
+		if (seq < drop_before) {
 			journal_replay_free(c, i, false);
 			continue;
 		}
@@ -1455,7 +1468,7 @@ int bch2_journal_read(struct bch_fs *c,
 		}
 	}
 
-	ret = bch2_journal_check_for_missing(c, *last_seq, *blacklist_seq - 1);
+	ret = bch2_journal_check_for_missing(c, drop_before, *blacklist_seq - 1);
 	if (ret)
 		goto err;
 

fs/bcachefs/opts.h

@@ -379,6 +379,11 @@ enum fsck_err_opts {
 	  OPT_BOOL(),							\
 	  BCH2_NO_SB_OPT,		false,				\
 	  NULL,		"Exit recovery immediately prior to journal replay")\
+	x(journal_rewind,		u64,				\
+	  OPT_FS|OPT_MOUNT,						\
+	  OPT_UINT(0, U64_MAX),						\
+	  BCH2_NO_SB_OPT,		0,				\
+	  NULL,		"Rewind journal")				\
 	x(recovery_passes,		u64,				\
 	  OPT_FS|OPT_MOUNT,						\
 	  OPT_BITFIELD(bch2_recovery_passes),				\

fs/bcachefs/recovery.c

@@ -757,6 +757,11 @@ int bch2_fs_recovery(struct bch_fs *c)
 	if (c->opts.nochanges)
 		c->opts.read_only = true;
 
+	if (c->opts.journal_rewind) {
+		bch_info(c, "rewinding journal, fsck required");
+		c->opts.fsck = true;
+	}
+
 	mutex_lock(&c->sb_lock);
 	struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext);
 	bool write_sb = false;