Commit ef4f6c32 authored Oct 31, 2024 by Kent Overstreet

bcachefs: Ensure BCH_FS_may_go_rw is set before exiting recovery



If BCH_FS_may_go_rw is not yet set, it indicates to the transaction
commit path that updates should be done via the list of journal replay
keys.

This must be set before multithreaded use commences.

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

parent cec136d3

fs/bcachefs/recovery.c

+7 −0

Original line number	Diff line number	Diff line
		@@ -862,6 +862,13 @@ int bch2_fs_recovery(struct bch_fs *c)
		if (ret)
		goto err;

		/*
		* Normally set by the appropriate recovery pass: when cleared, this
		* indicates we're in early recovery and btree updates should be done by
		* being applied to the journal replay keys. _Must_ be cleared before
		* multithreaded use:
		*/
		set_bit(BCH_FS_may_go_rw, &c->flags);
		clear_bit(BCH_FS_fsck_running, &c->flags);

		/* in case we don't run journal replay, i.e. norecovery mode */

fs/bcachefs/recovery_passes.c

+6 −0

Original line number	Diff line number	Diff line
		@@ -221,6 +221,12 @@ int bch2_run_recovery_passes(struct bch_fs *c)
		{
		int ret = 0;

		/*
		* We can't allow set_may_go_rw to be excluded; that would cause us to
		* use the journal replay keys for updates where it's not expected.
		*/
		c->opts.recovery_passes_exclude &= ~BCH_RECOVERY_PASS_set_may_go_rw;

		while (c->curr_recovery_pass < ARRAY_SIZE(recovery_pass_fns)) {
		if (c->opts.recovery_pass_last &&
		c->curr_recovery_pass > c->opts.recovery_pass_last)