bcachefs: Add an internal option for reading entire journal

To be used the debug tool that dumps the contents of the journal.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

fs/bcachefs/journal.c

@@ -986,9 +986,8 @@ int bch2_fs_journal_start(struct journal *j, u64 cur_seq,
 	u64 last_seq = cur_seq, nr, seq;
 
 	if (!list_empty(journal_entries))
-		last_seq = le64_to_cpu(list_first_entry(journal_entries,
-							struct journal_replay,
-							list)->j.seq);
+		last_seq = le64_to_cpu(list_last_entry(journal_entries,
+				struct journal_replay, list)->j.last_seq);
 
 	nr = cur_seq - last_seq;
 
@@ -1017,8 +1016,10 @@ int bch2_fs_journal_start(struct journal *j, u64 cur_seq,
 
 	list_for_each_entry(i, journal_entries, list) {
 		seq = le64_to_cpu(i->j.seq);
+		BUG_ON(seq >= cur_seq);
 
-		BUG_ON(seq < last_seq || seq >= cur_seq);
+		if (seq < last_seq)
+			continue;
 
 		journal_seq_pin(j, seq)->devs = i->devs;
 	}

fs/bcachefs/journal_io.c

@@ -40,19 +40,21 @@ static int journal_entry_add(struct bch_fs *c, struct bch_dev *ca,
 				  list)->j.last_seq
 		: 0;
 
-	/* Is this entry older than the range we need? */
-	if (le64_to_cpu(j->seq) < le64_to_cpu(last_seq)) {
-		ret = JOURNAL_ENTRY_ADD_OUT_OF_RANGE;
-		goto out;
-	}
+	if (!c->opts.read_entire_journal) {
+		/* Is this entry older than the range we need? */
+		if (le64_to_cpu(j->seq) < le64_to_cpu(last_seq)) {
+			ret = JOURNAL_ENTRY_ADD_OUT_OF_RANGE;
+			goto out;
+		}
 
-	/* Drop entries we don't need anymore */
-	list_for_each_entry_safe(i, pos, jlist->head, list) {
-		if (le64_to_cpu(i->j.seq) >= le64_to_cpu(j->last_seq))
-			break;
-		list_del(&i->list);
-		kvpfree(i, offsetof(struct journal_replay, j) +
-			vstruct_bytes(&i->j));
+		/* Drop entries we don't need anymore */
+		list_for_each_entry_safe(i, pos, jlist->head, list) {
+			if (le64_to_cpu(i->j.seq) >= le64_to_cpu(j->last_seq))
+				break;
+			list_del(&i->list);
+			kvpfree(i, offsetof(struct journal_replay, j) +
+				vstruct_bytes(&i->j));
+		}
 	}
 
 	list_for_each_entry_reverse(i, jlist->head, list) {

fs/bcachefs/opts.h

@@ -265,6 +265,11 @@ enum opt_type {
 	  OPT_BOOL(),							\
 	  NO_SB_OPT,			false,				\
 	  NULL,		"Don't free journal entries/keys after startup")\
+	x(read_entire_journal,		u8,				\
+	  0,								\
+	  OPT_BOOL(),							\
+	  NO_SB_OPT,			false,				\
+	  NULL,		"Read all journal entries, not just dirty ones")\
 	x(noexcl,			u8,				\
 	  OPT_MOUNT,							\
 	  OPT_BOOL(),							\

fs/bcachefs/recovery.c

@@ -319,20 +319,30 @@ static struct journal_keys journal_keys_sort(struct list_head *journal_entries)
 	struct journal_key *src, *dst;
 	size_t nr_keys = 0;
 
-	list_for_each_entry(p, journal_entries, list)
+	if (list_empty(journal_entries))
+		return keys;
+
+	keys.journal_seq_base =
+		le64_to_cpu(list_last_entry(journal_entries,
+				struct journal_replay, list)->j.last_seq);
+
+	list_for_each_entry(p, journal_entries, list) {
+		if (le64_to_cpu(p->j.seq) < keys.journal_seq_base)
+			continue;
+
 		for_each_jset_key(k, _n, entry, &p->j)
 			nr_keys++;
+	}
 
-	keys.journal_seq_base =
-		le64_to_cpu(list_first_entry(journal_entries,
-					     struct journal_replay,
-					     list)->j.seq);
 
 	keys.d = kvmalloc(sizeof(keys.d[0]) * nr_keys, GFP_KERNEL);
 	if (!keys.d)
 		goto err;
 
-	list_for_each_entry(p, journal_entries, list)
+	list_for_each_entry(p, journal_entries, list) {
+		if (le64_to_cpu(p->j.seq) < keys.journal_seq_base)
+			continue;
+
 		for_each_jset_key(k, _n, entry, &p->j)
 			keys.d[keys.nr++] = (struct journal_key) {
 				.btree_id	= entry->btree_id,
@@ -342,6 +352,7 @@ static struct journal_keys journal_keys_sort(struct list_head *journal_entries)
 					keys.journal_seq_base,
 				.journal_offset	= k->_data - p->j._data,
 			};
+	}
 
 	sort(keys.d, keys.nr, sizeof(keys.d[0]), journal_sort_key_cmp, NULL);
 
@@ -568,6 +579,9 @@ verify_journal_entries_not_blacklisted_or_missing(struct bch_fs *c,
 	int ret = 0;
 
 	list_for_each_entry(i, journal, list) {
+		if (le64_to_cpu(i->j.seq) < start_seq)
+			continue;
+
 		fsck_err_on(seq != le64_to_cpu(i->j.seq), c,
 			"journal entries %llu-%llu missing! (replaying %llu-%llu)",
 			seq, le64_to_cpu(i->j.seq) - 1,