On Thu, 2018-04-12 at 09:12 -0700, tj@xxxxxxxxxx wrote: > > Did you perhaps mean blkg_lookup_create()? That function has one caller, > > namely blkcg_bio_issue_check(). The only caller of that function is > > generic_make_request_checks(). A patch was posted on the linux-block mailing > > list recently that surrounds that call with blk_queue_enter() / blk_queue_exit(). > > I think that prevents that blkcg_exit_queue() is called concurrently with > > blkg_lookup_create(). > > Yeah, that'd solve the problem for that particular path, but what > that's doing is adding another layer of refcnting which is used to > implement the revoke (or sever) semantics. This is a fragile approach > tho because it isn't clear who's protecting what and there's nothing > in blkg's usage which suggests it'd be protected that way and we're > basically working around a different underlying problem (lock > switching) by expanding the coverage of a different lock. Hello Tejun, Any code that submits a bio or request needs blk_queue_enter() / blk_queue_exit() anyway. Please have a look at the following commit - you will see that that commit reduces the number of blk_queue_enter() / blk_queue_exit() calls in the hot path: https://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux-block.git/commit/?h=for-linus&id=37f9579f4c31a6d698dbf3016d7bf132f9288d30 Thanks, Bart.
