| Message ID | 20230221092436.3570192-1-yukuai1@huaweicloud.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | blk-mq: quiesce queue while reallocating hctxs | expand |
Hi, friendly ping ... Thanks, Kuai 在 2023/02/21 17:24, Yu Kuai 写道: > From: Yu Kuai <yukuai3@huawei.com> > > commit 8237c01f1696 ("blk-mq: use quiesced elevator switch when > reinitializing queues") add quiesce queue while switching elevator, > however, if old elevator is none, queue is still not quiesced. Hence > reallocating hctxs can concurrent with run queue. Fix it by also > quiesce queue in the beginning of __blk_mq_update_nr_hw_queues(). > > Signed-off-by: Yu Kuai <yukuai3@huawei.com> > --- > block/blk-mq.c | 8 ++++++-- > 1 file changed, 6 insertions(+), 2 deletions(-) > > diff --git a/block/blk-mq.c b/block/blk-mq.c > index d3494a796ba8..fb44ef0dff8a 100644 > --- a/block/blk-mq.c > +++ b/block/blk-mq.c > @@ -4691,8 +4691,10 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, > if (set->nr_maps == 1 && nr_hw_queues == set->nr_hw_queues) > return; > > - list_for_each_entry(q, &set->tag_list, tag_set_list) > + list_for_each_entry(q, &set->tag_list, tag_set_list) { > blk_mq_freeze_queue(q); > + blk_mq_quiesce_queue(q); > + } > /* > * Switch IO scheduler to 'none', cleaning up the data associated > * with the previous scheduler. We will switch back once we are done > @@ -4741,8 +4743,10 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, > list_for_each_entry(q, &set->tag_list, tag_set_list) > blk_mq_elv_switch_back(&head, q); > > - list_for_each_entry(q, &set->tag_list, tag_set_list) > + list_for_each_entry(q, &set->tag_list, tag_set_list) { > + blk_mq_unquiesce_queue(q); > blk_mq_unfreeze_queue(q); > + } > } > > void blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, int nr_hw_queues) >
On Tue, Feb 21, 2023 at 05:24:36PM +0800, Yu Kuai wrote: > From: Yu Kuai <yukuai3@huawei.com> > > commit 8237c01f1696 ("blk-mq: use quiesced elevator switch when > reinitializing queues") add quiesce queue while switching elevator, > however, if old elevator is none, queue is still not quiesced. Hence > reallocating hctxs can concurrent with run queue. Fix it by also > quiesce queue in the beginning of __blk_mq_update_nr_hw_queues(). Is this actually fixing anything? The quiesced elevator switch was to prevent use-after-free from an elevator being torn down, but if you are not switching elevators, then what resource does quiescing protect?
Hi, 在 2023/03/03 9:24, Keith Busch 写道: > On Tue, Feb 21, 2023 at 05:24:36PM +0800, Yu Kuai wrote: >> From: Yu Kuai <yukuai3@huawei.com> >> >> commit 8237c01f1696 ("blk-mq: use quiesced elevator switch when >> reinitializing queues") add quiesce queue while switching elevator, >> however, if old elevator is none, queue is still not quiesced. Hence >> reallocating hctxs can concurrent with run queue. Fix it by also >> quiesce queue in the beginning of __blk_mq_update_nr_hw_queues(). > > Is this actually fixing anything? The quiesced elevator switch was to prevent > use-after-free from an elevator being torn down, but if you are not switching > elevators, then what resource does quiescing protect? > . > What I can think of is hctx itself. run queue can fetch a hctx that will be exited, I did't found any real problems yet, but I don't think it is not good to run work for a exited or reused hctx. Thanks, Kuai
diff --git a/block/blk-mq.c b/block/blk-mq.c index d3494a796ba8..fb44ef0dff8a 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -4691,8 +4691,10 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, if (set->nr_maps == 1 && nr_hw_queues == set->nr_hw_queues) return; - list_for_each_entry(q, &set->tag_list, tag_set_list) + list_for_each_entry(q, &set->tag_list, tag_set_list) { blk_mq_freeze_queue(q); + blk_mq_quiesce_queue(q); + } /* * Switch IO scheduler to 'none', cleaning up the data associated * with the previous scheduler. We will switch back once we are done @@ -4741,8 +4743,10 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, list_for_each_entry(q, &set->tag_list, tag_set_list) blk_mq_elv_switch_back(&head, q); - list_for_each_entry(q, &set->tag_list, tag_set_list) + list_for_each_entry(q, &set->tag_list, tag_set_list) { + blk_mq_unquiesce_queue(q); blk_mq_unfreeze_queue(q); + } } void blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, int nr_hw_queues)