btrfs: use monotonic time for transaction handling

The transaction times were changed to ktime_get_real_seconds to avoid
the y2038 overflow, but they still have a minor problem when they go
backwards or jump due to settimeofday() or leap seconds.

This changes the transaction handling to instead use ktime_get_seconds(),
which returns a CLOCK_MONOTONIC timestamp that has neither of those
problems.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This commit is contained in:
Arnd Bergmann 2018-06-21 18:04:05 +02:00 committed by David Sterba
parent e41ca58974
commit afd48513f0
2 changed files with 2 additions and 2 deletions

View file

@ -1805,7 +1805,7 @@ static int transaction_kthread(void *arg)
goto sleep;
}
now = ktime_get_real_seconds();
now = ktime_get_seconds();
if (cur->state < TRANS_STATE_BLOCKED &&
!test_bit(BTRFS_FS_NEED_ASYNC_COMMIT, &fs_info->flags) &&
(now < cur->start_time ||

View file

@ -241,7 +241,7 @@ static noinline int join_transaction(struct btrfs_fs_info *fs_info,
refcount_set(&cur_trans->use_count, 2);
atomic_set(&cur_trans->pending_ordered, 0);
cur_trans->flags = 0;
cur_trans->start_time = ktime_get_real_seconds();
cur_trans->start_time = ktime_get_seconds();
memset(&cur_trans->delayed_refs, 0, sizeof(cur_trans->delayed_refs));