__bch2_fsck_err() warns if the current task has a btree_trans object and
it wasn't passed in, because if it has to prompt for user input it has
to be able to unlock it.
But plumbing the btree_trans through bkey_validate(), as well as
transaction restarts, is problematic - so instead make bkey fsck errors
FSCK_AUTOFIX, which doesn't need to warn.
Signed-off-by: Kent Overstreet <[email protected]>
if (!c)
c = trans->c;
- WARN_ON(!trans && bch2_current_has_btree_trans(c));
+ /*
+ * Ugly: if there's a transaction in the current task it has to be
+ * passed in to unlock if we prompt for user input.
+ *
+ * But, plumbing a transaction and transaction restarts into
+ * bkey_validate() is problematic.
+ *
+ * So:
+ * - make all bkey errors AUTOFIX, they're simple anyways (we just
+ * delete the key)
+ * - and we don't need to warn if we're not prompting
+ */
+ WARN_ON(!(flags & FSCK_AUTOFIX) && !trans && bch2_current_has_btree_trans(c));
if ((flags & FSCK_CAN_FIX) &&
test_bit(err, c->sb.errors_silent))
ret = -BCH_ERR_fsck_delete_bkey; \
goto fsck_err; \
} \
- int _ret = __bch2_bkey_fsck_err(c, k, FSCK_CAN_FIX, \
+ int _ret = __bch2_bkey_fsck_err(c, k, FSCK_CAN_FIX|FSCK_AUTOFIX,\
BCH_FSCK_ERR_##_err_type, \
_err_msg, ##__VA_ARGS__); \
if (_ret != -BCH_ERR_fsck_fix && \