On Thu, Mar 02, 2017 at 09:13:30AM -0700, Jens Axboe wrote: > I was worried about that. How about the below? We need to grab the lock > at some point for legacy scheduling, but the ordering should be correct. Makes sense, now the locking is consistent with the other place we call ioc_exit_icq(). One nit below, and you can add Reviewed-by: Omar Sandoval <osandov@xxxxxx> > diff --git a/block/blk-ioc.c b/block/blk-ioc.c > index b12f9c87b4c3..6fd633b5d567 100644 > --- a/block/blk-ioc.c > +++ b/block/blk-ioc.c > @@ -54,7 +54,7 @@ static void ioc_exit_icq(struct io_cq *icq) > icq->flags |= ICQ_EXITED; > } > > -/* Release an icq. Called with both ioc and q locked. */ > +/* Release an icq. Called with ioc locked. */ For ioc_exit_icq(), we have the more explicit comment /* * Exit an icq. Called with both ioc and q locked for sq, only ioc locked for * mq. */ Could you document that here, too? > static void ioc_destroy_icq(struct io_cq *icq) > { > struct io_context *ioc = icq->ioc; > @@ -62,7 +62,6 @@ static void ioc_destroy_icq(struct io_cq *icq) > struct elevator_type *et = q->elevator->type; > > lockdep_assert_held(&ioc->lock); > - lockdep_assert_held(q->queue_lock); > > radix_tree_delete(&ioc->icq_tree, icq->q->id); > hlist_del_init(&icq->ioc_node); > @@ -222,24 +221,40 @@ void exit_io_context(struct task_struct *task) > put_io_context_active(ioc); > }