| Message ID | 20220305091205.4188398-11-yukuai3@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | support concurrent sync io for bfq on a specail occasion | expand |
On Sat 05-03-22 17:12:04, Yu Kuai wrote: > Currently 'num_groups_with_pending_reqs' won't be decreased when > the group doesn't have any pending requests, while some child group > still have pending requests. The decrement is delayed to when all the > child groups doesn't have any pending requests. > > For example: > 1) t1 issue sync io on root group, t2 and t3 issue sync io on the same > child group. num_groups_with_pending_reqs is 2 now. > 2) t1 stopped, num_groups_with_pending_reqs is still 2. io from t2 and > t3 still can't be handled concurrently. > > Fix the problem by decreasing 'num_groups_with_pending_reqs' > immediately upon the weights_tree removal of last bfqq of the group. > > Signed-off-by: Yu Kuai <yukuai3@huawei.com> So I'd find the logic easier to follow if you completely removed entity->in_groups_with_pending_reqs and did updates of bfqd->num_groups_with_pending_reqs like: if (!bfqg->num_entities_with_pending_reqs++) bfqd->num_groups_with_pending_reqs++; and similarly on the remove side. And there would we literally two places (addition & removal from weight tree) that would need to touch these counters. Pretty obvious and all can be done in patch 9. Honza > --- > block/bfq-iosched.c | 56 +++++++++++++++------------------------------ > block/bfq-iosched.h | 16 ++++++------- > 2 files changed, 27 insertions(+), 45 deletions(-) > > diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c > index f221e9cab4d0..119b64c9c1d9 100644 > --- a/block/bfq-iosched.c > +++ b/block/bfq-iosched.c > @@ -970,6 +970,24 @@ void __bfq_weights_tree_remove(struct bfq_data *bfqd, > bfq_put_queue(bfqq); > } > > +static void decrease_groups_with_pending_reqs(struct bfq_data *bfqd, > + struct bfq_queue *bfqq) > +{ > +#ifdef CONFIG_BFQ_GROUP_IOSCHED > + struct bfq_entity *entity = bfqq->entity.parent; > + > + /* > + * The decrement of num_groups_with_pending_reqs is performed > + * immediately when the last bfqq completes all the requests. > + */ > + if (!bfqq_group(bfqq)->num_entities_with_pending_reqs && > + entity->in_groups_with_pending_reqs) { > + entity->in_groups_with_pending_reqs = false; > + bfqd->num_groups_with_pending_reqs--; > + } > +#endif > +} > + > /* > * Invoke __bfq_weights_tree_remove on bfqq and decrement the number > * of active groups for each queue's inactive parent entity. > @@ -977,8 +995,6 @@ void __bfq_weights_tree_remove(struct bfq_data *bfqd, > void bfq_weights_tree_remove(struct bfq_data *bfqd, > struct bfq_queue *bfqq) > { > - struct bfq_entity *entity = bfqq->entity.parent; > - > /* > * grab a ref to prevent bfqq to be freed in > * __bfq_weights_tree_remove > @@ -991,41 +1007,7 @@ void bfq_weights_tree_remove(struct bfq_data *bfqd, > */ > __bfq_weights_tree_remove(bfqd, bfqq, > &bfqd->queue_weights_tree); > - > - for_each_entity(entity) { > - struct bfq_sched_data *sd = entity->my_sched_data; > - > - if (sd->next_in_service || sd->in_service_entity) { > - /* > - * entity is still active, because either > - * next_in_service or in_service_entity is not > - * NULL (see the comments on the definition of > - * next_in_service for details on why > - * in_service_entity must be checked too). > - * > - * As a consequence, its parent entities are > - * active as well, and thus this loop must > - * stop here. > - */ > - break; > - } > - > - /* > - * The decrement of num_groups_with_pending_reqs is > - * not performed immediately upon the deactivation of > - * entity, but it is delayed to when it also happens > - * that the first leaf descendant bfqq of entity gets > - * all its pending requests completed. The following > - * instructions perform this delayed decrement, if > - * needed. See the comments on > - * num_groups_with_pending_reqs for details. > - */ > - if (entity->in_groups_with_pending_reqs) { > - entity->in_groups_with_pending_reqs = false; > - bfqd->num_groups_with_pending_reqs--; > - } > - } > - > + decrease_groups_with_pending_reqs(bfqd, bfqq); > bfq_put_queue(bfqq); > } > > diff --git a/block/bfq-iosched.h b/block/bfq-iosched.h > index 5d904851519c..9ec72bd24fc2 100644 > --- a/block/bfq-iosched.h > +++ b/block/bfq-iosched.h > @@ -495,7 +495,7 @@ struct bfq_data { > struct rb_root_cached queue_weights_tree; > > /* > - * Number of groups with at least one descendant process that > + * Number of groups with at least one process that > * has at least one request waiting for completion. Note that > * this accounts for also requests already dispatched, but not > * yet completed. Therefore this number of groups may differ > @@ -508,14 +508,14 @@ struct bfq_data { > * bfq_better_to_idle(). > * > * However, it is hard to compute this number exactly, for > - * groups with multiple descendant processes. Consider a group > - * that is inactive, i.e., that has no descendant process with > + * groups with multiple processes. Consider a group > + * that is inactive, i.e., that has no process with > * pending I/O inside BFQ queues. Then suppose that > * num_groups_with_pending_reqs is still accounting for this > - * group, because the group has descendant processes with some > + * group, because the group has processes with some > * I/O request still in flight. num_groups_with_pending_reqs > * should be decremented when the in-flight request of the > - * last descendant process is finally completed (assuming that > + * last process is finally completed (assuming that > * nothing else has changed for the group in the meantime, in > * terms of composition of the group and active/inactive state of child > * groups and processes). To accomplish this, an additional > @@ -524,7 +524,7 @@ struct bfq_data { > * we resort to the following tradeoff between simplicity and > * accuracy: for an inactive group that is still counted in > * num_groups_with_pending_reqs, we decrement > - * num_groups_with_pending_reqs when the first descendant > + * num_groups_with_pending_reqs when the last > * process of the group remains with no request waiting for > * completion. > * > @@ -532,12 +532,12 @@ struct bfq_data { > * carefulness: to avoid multiple decrements, we flag a group, > * more precisely an entity representing a group, as still > * counted in num_groups_with_pending_reqs when it becomes > - * inactive. Then, when the first descendant queue of the > + * inactive. Then, when the last queue of the > * entity remains with no request waiting for completion, > * num_groups_with_pending_reqs is decremented, and this flag > * is reset. After this flag is reset for the entity, > * num_groups_with_pending_reqs won't be decremented any > - * longer in case a new descendant queue of the entity remains > + * longer in case a new queue of the entity remains > * with no request waiting for completion. > */ > unsigned int num_groups_with_pending_reqs; > -- > 2.31.1 >
在 2022/04/13 19:28, Jan Kara 写道: > On Sat 05-03-22 17:12:04, Yu Kuai wrote: >> Currently 'num_groups_with_pending_reqs' won't be decreased when >> the group doesn't have any pending requests, while some child group >> still have pending requests. The decrement is delayed to when all the >> child groups doesn't have any pending requests. >> >> For example: >> 1) t1 issue sync io on root group, t2 and t3 issue sync io on the same >> child group. num_groups_with_pending_reqs is 2 now. >> 2) t1 stopped, num_groups_with_pending_reqs is still 2. io from t2 and >> t3 still can't be handled concurrently. >> >> Fix the problem by decreasing 'num_groups_with_pending_reqs' >> immediately upon the weights_tree removal of last bfqq of the group. >> >> Signed-off-by: Yu Kuai <yukuai3@huawei.com> > > So I'd find the logic easier to follow if you completely removed > entity->in_groups_with_pending_reqs and did updates of > bfqd->num_groups_with_pending_reqs like: > > if (!bfqg->num_entities_with_pending_reqs++) > bfqd->num_groups_with_pending_reqs++; > Hi, Indeed, this is an excellent idle, and much better than the way I did. Thanks, Kuai > and similarly on the remove side. And there would we literally two places > (addition & removal from weight tree) that would need to touch these > counters. Pretty obvious and all can be done in patch 9. > > Honza >
在 2022/04/13 19:40, yukuai (C) 写道: > 在 2022/04/13 19:28, Jan Kara 写道: >> On Sat 05-03-22 17:12:04, Yu Kuai wrote: >>> Currently 'num_groups_with_pending_reqs' won't be decreased when >>> the group doesn't have any pending requests, while some child group >>> still have pending requests. The decrement is delayed to when all the >>> child groups doesn't have any pending requests. >>> >>> For example: >>> 1) t1 issue sync io on root group, t2 and t3 issue sync io on the same >>> child group. num_groups_with_pending_reqs is 2 now. >>> 2) t1 stopped, num_groups_with_pending_reqs is still 2. io from t2 and >>> t3 still can't be handled concurrently. >>> >>> Fix the problem by decreasing 'num_groups_with_pending_reqs' >>> immediately upon the weights_tree removal of last bfqq of the group. >>> >>> Signed-off-by: Yu Kuai <yukuai3@huawei.com> >> >> So I'd find the logic easier to follow if you completely removed >> entity->in_groups_with_pending_reqs and did updates of >> bfqd->num_groups_with_pending_reqs like: >> >> if (!bfqg->num_entities_with_pending_reqs++) >> bfqd->num_groups_with_pending_reqs++; >> > Hi, > > Indeed, this is an excellent idle, and much better than the way I did. > > Thanks, > Kuai > >> and similarly on the remove side. And there would we literally two places >> (addition & removal from weight tree) that would need to touch these >> counters. Pretty obvious and all can be done in patch 9. >> >> Honza Hi, Jan I think with this change, we can count root_group while activating bfqqs that are under root_group, thus there is no need to modify for_each_entity(or fake bfq_sched_data) any more. The special case is that weight racing bfqqs are not inserted into weights tree, and I think this can be handled by adding a fake bfq_weight_counter for such bfqqs. What do you think ? Kuai
On Fri 15-04-22 09:10:06, yukuai (C) wrote: > 在 2022/04/13 19:40, yukuai (C) 写道: > > 在 2022/04/13 19:28, Jan Kara 写道: > > > On Sat 05-03-22 17:12:04, Yu Kuai wrote: > > > > Currently 'num_groups_with_pending_reqs' won't be decreased when > > > > the group doesn't have any pending requests, while some child group > > > > still have pending requests. The decrement is delayed to when all the > > > > child groups doesn't have any pending requests. > > > > > > > > For example: > > > > 1) t1 issue sync io on root group, t2 and t3 issue sync io on the same > > > > child group. num_groups_with_pending_reqs is 2 now. > > > > 2) t1 stopped, num_groups_with_pending_reqs is still 2. io from t2 and > > > > t3 still can't be handled concurrently. > > > > > > > > Fix the problem by decreasing 'num_groups_with_pending_reqs' > > > > immediately upon the weights_tree removal of last bfqq of the group. > > > > > > > > Signed-off-by: Yu Kuai <yukuai3@huawei.com> > > > > > > So I'd find the logic easier to follow if you completely removed > > > entity->in_groups_with_pending_reqs and did updates of > > > bfqd->num_groups_with_pending_reqs like: > > > > > > if (!bfqg->num_entities_with_pending_reqs++) > > > bfqd->num_groups_with_pending_reqs++; > > > > > Hi, > > > > Indeed, this is an excellent idle, and much better than the way I did. > > > > Thanks, > > Kuai > > > > > and similarly on the remove side. And there would we literally two places > > > (addition & removal from weight tree) that would need to touch these > > > counters. Pretty obvious and all can be done in patch 9. > > I think with this change, we can count root_group while activating bfqqs > that are under root_group, thus there is no need to modify > for_each_entity(or fake bfq_sched_data) any more. Sure, if you can make this work, it would be easier :) > The special case is that weight racing bfqqs are not inserted into > weights tree, and I think this can be handled by adding a fake > bfq_weight_counter for such bfqqs. Do you mean "weight raised bfqqs"? Yes, you are right they would need special treatment - maybe bfq_weights_tree_add() is not the best function to use for this and we should rather use insertion / removal from the service tree for maintaining num_entities_with_pending_reqs counter? I can even see we already have bfqg->active_entities so maybe we could just somehow tweak that accounting and use it for our purposes? Honza
在 2022/04/19 17:49, Jan Kara 写道: > On Fri 15-04-22 09:10:06, yukuai (C) wrote: >> 在 2022/04/13 19:40, yukuai (C) 写道: >>> 在 2022/04/13 19:28, Jan Kara 写道: >>>> On Sat 05-03-22 17:12:04, Yu Kuai wrote: >>>>> Currently 'num_groups_with_pending_reqs' won't be decreased when >>>>> the group doesn't have any pending requests, while some child group >>>>> still have pending requests. The decrement is delayed to when all the >>>>> child groups doesn't have any pending requests. >>>>> >>>>> For example: >>>>> 1) t1 issue sync io on root group, t2 and t3 issue sync io on the same >>>>> child group. num_groups_with_pending_reqs is 2 now. >>>>> 2) t1 stopped, num_groups_with_pending_reqs is still 2. io from t2 and >>>>> t3 still can't be handled concurrently. >>>>> >>>>> Fix the problem by decreasing 'num_groups_with_pending_reqs' >>>>> immediately upon the weights_tree removal of last bfqq of the group. >>>>> >>>>> Signed-off-by: Yu Kuai <yukuai3@huawei.com> >>>> >>>> So I'd find the logic easier to follow if you completely removed >>>> entity->in_groups_with_pending_reqs and did updates of >>>> bfqd->num_groups_with_pending_reqs like: >>>> >>>> if (!bfqg->num_entities_with_pending_reqs++) >>>> bfqd->num_groups_with_pending_reqs++; >>>> >>> Hi, >>> >>> Indeed, this is an excellent idle, and much better than the way I did. >>> >>> Thanks, >>> Kuai >>> >>>> and similarly on the remove side. And there would we literally two places >>>> (addition & removal from weight tree) that would need to touch these >>>> counters. Pretty obvious and all can be done in patch 9. >> >> I think with this change, we can count root_group while activating bfqqs >> that are under root_group, thus there is no need to modify >> for_each_entity(or fake bfq_sched_data) any more. > > Sure, if you can make this work, it would be easier :) > >> The special case is that weight racing bfqqs are not inserted into >> weights tree, and I think this can be handled by adding a fake >> bfq_weight_counter for such bfqqs. > > Do you mean "weight raised bfqqs"? Yes, you are right they would need > special treatment - maybe bfq_weights_tree_add() is not the best function > to use for this and we should rather use insertion / removal from the > service tree for maintaining num_entities_with_pending_reqs counter? > I can even see we already have bfqg->active_entities so maybe we could just > somehow tweak that accounting and use it for our purposes? The problem to use 'active_entities' is that bfqq can be deactivated while it still has pending requests. Anyway, I posted a new version aready, which still use weights_tree insertion / removal to count pending bfqqs. I'll be great if you can take a look: https://patchwork.kernel.org/project/linux-block/cover/20220416093753.3054696-1-yukuai3@huawei.com/ BTW, I was worried that you can't receive the emails because I got warnings that mails can't deliver to you: Your message could not be delivered for more than 6 hour(s). It will be retried until it is 1 day(s) old. For further assistance, please send mail to postmaster. If you do so, please include this problem report. You can delete your own text from the attached returned message. The mail system <jack@imap.suse.de> (expanded from <jack@suse.cz>): host mail2.suse.de[149.44.160.157] said: 452 4.3.1 Insufficient system storage Thanks, Kuai > > Honza >
On Tue 19-04-22 19:37:11, yukuai (C) wrote: > 在 2022/04/19 17:49, Jan Kara 写道: > > On Fri 15-04-22 09:10:06, yukuai (C) wrote: > > > 在 2022/04/13 19:40, yukuai (C) 写道: > > > > 在 2022/04/13 19:28, Jan Kara 写道: > > > > > On Sat 05-03-22 17:12:04, Yu Kuai wrote: > > > > > > Currently 'num_groups_with_pending_reqs' won't be decreased when > > > > > > the group doesn't have any pending requests, while some child group > > > > > > still have pending requests. The decrement is delayed to when all the > > > > > > child groups doesn't have any pending requests. > > > > > > > > > > > > For example: > > > > > > 1) t1 issue sync io on root group, t2 and t3 issue sync io on the same > > > > > > child group. num_groups_with_pending_reqs is 2 now. > > > > > > 2) t1 stopped, num_groups_with_pending_reqs is still 2. io from t2 and > > > > > > t3 still can't be handled concurrently. > > > > > > > > > > > > Fix the problem by decreasing 'num_groups_with_pending_reqs' > > > > > > immediately upon the weights_tree removal of last bfqq of the group. > > > > > > > > > > > > Signed-off-by: Yu Kuai <yukuai3@huawei.com> > > > > > > > > > > So I'd find the logic easier to follow if you completely removed > > > > > entity->in_groups_with_pending_reqs and did updates of > > > > > bfqd->num_groups_with_pending_reqs like: > > > > > > > > > > if (!bfqg->num_entities_with_pending_reqs++) > > > > > bfqd->num_groups_with_pending_reqs++; > > > > > > > > > Hi, > > > > > > > > Indeed, this is an excellent idle, and much better than the way I did. > > > > > > > > Thanks, > > > > Kuai > > > > > > > > > and similarly on the remove side. And there would we literally two places > > > > > (addition & removal from weight tree) that would need to touch these > > > > > counters. Pretty obvious and all can be done in patch 9. > > > > > > I think with this change, we can count root_group while activating bfqqs > > > that are under root_group, thus there is no need to modify > > > for_each_entity(or fake bfq_sched_data) any more. > > > > Sure, if you can make this work, it would be easier :) > > > > > The special case is that weight racing bfqqs are not inserted into > > > weights tree, and I think this can be handled by adding a fake > > > bfq_weight_counter for such bfqqs. > > > > Do you mean "weight raised bfqqs"? Yes, you are right they would need > > special treatment - maybe bfq_weights_tree_add() is not the best function > > to use for this and we should rather use insertion / removal from the > > service tree for maintaining num_entities_with_pending_reqs counter? > > I can even see we already have bfqg->active_entities so maybe we could just > > somehow tweak that accounting and use it for our purposes? > > The problem to use 'active_entities' is that bfqq can be deactivated > while it still has pending requests. > > Anyway, I posted a new version aready, which still use weights_tree > insertion / removal to count pending bfqqs. I'll be great if you can > take a look: > > https://patchwork.kernel.org/project/linux-block/cover/20220416093753.3054696-1-yukuai3@huawei.com/ Thanks, I'll have a look. > BTW, I was worried that you can't receive the emails because I got > warnings that mails can't deliver to you: > > Your message could not be delivered for more than 6 hour(s). > It will be retried until it is 1 day(s) old. Yes, I didn't get those emails because our mail system ran out of disk space and it took a few days to resolve so emails got bounced... Honza
diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c index f221e9cab4d0..119b64c9c1d9 100644 --- a/block/bfq-iosched.c +++ b/block/bfq-iosched.c @@ -970,6 +970,24 @@ void __bfq_weights_tree_remove(struct bfq_data *bfqd, bfq_put_queue(bfqq); } +static void decrease_groups_with_pending_reqs(struct bfq_data *bfqd, + struct bfq_queue *bfqq) +{ +#ifdef CONFIG_BFQ_GROUP_IOSCHED + struct bfq_entity *entity = bfqq->entity.parent; + + /* + * The decrement of num_groups_with_pending_reqs is performed + * immediately when the last bfqq completes all the requests. + */ + if (!bfqq_group(bfqq)->num_entities_with_pending_reqs && + entity->in_groups_with_pending_reqs) { + entity->in_groups_with_pending_reqs = false; + bfqd->num_groups_with_pending_reqs--; + } +#endif +} + /* * Invoke __bfq_weights_tree_remove on bfqq and decrement the number * of active groups for each queue's inactive parent entity. @@ -977,8 +995,6 @@ void __bfq_weights_tree_remove(struct bfq_data *bfqd, void bfq_weights_tree_remove(struct bfq_data *bfqd, struct bfq_queue *bfqq) { - struct bfq_entity *entity = bfqq->entity.parent; - /* * grab a ref to prevent bfqq to be freed in * __bfq_weights_tree_remove @@ -991,41 +1007,7 @@ void bfq_weights_tree_remove(struct bfq_data *bfqd, */ __bfq_weights_tree_remove(bfqd, bfqq, &bfqd->queue_weights_tree); - - for_each_entity(entity) { - struct bfq_sched_data *sd = entity->my_sched_data; - - if (sd->next_in_service || sd->in_service_entity) { - /* - * entity is still active, because either - * next_in_service or in_service_entity is not - * NULL (see the comments on the definition of - * next_in_service for details on why - * in_service_entity must be checked too). - * - * As a consequence, its parent entities are - * active as well, and thus this loop must - * stop here. - */ - break; - } - - /* - * The decrement of num_groups_with_pending_reqs is - * not performed immediately upon the deactivation of - * entity, but it is delayed to when it also happens - * that the first leaf descendant bfqq of entity gets - * all its pending requests completed. The following - * instructions perform this delayed decrement, if - * needed. See the comments on - * num_groups_with_pending_reqs for details. - */ - if (entity->in_groups_with_pending_reqs) { - entity->in_groups_with_pending_reqs = false; - bfqd->num_groups_with_pending_reqs--; - } - } - + decrease_groups_with_pending_reqs(bfqd, bfqq); bfq_put_queue(bfqq); } diff --git a/block/bfq-iosched.h b/block/bfq-iosched.h index 5d904851519c..9ec72bd24fc2 100644 --- a/block/bfq-iosched.h +++ b/block/bfq-iosched.h @@ -495,7 +495,7 @@ struct bfq_data { struct rb_root_cached queue_weights_tree; /* - * Number of groups with at least one descendant process that + * Number of groups with at least one process that * has at least one request waiting for completion. Note that * this accounts for also requests already dispatched, but not * yet completed. Therefore this number of groups may differ @@ -508,14 +508,14 @@ struct bfq_data { * bfq_better_to_idle(). * * However, it is hard to compute this number exactly, for - * groups with multiple descendant processes. Consider a group - * that is inactive, i.e., that has no descendant process with + * groups with multiple processes. Consider a group + * that is inactive, i.e., that has no process with * pending I/O inside BFQ queues. Then suppose that * num_groups_with_pending_reqs is still accounting for this - * group, because the group has descendant processes with some + * group, because the group has processes with some * I/O request still in flight. num_groups_with_pending_reqs * should be decremented when the in-flight request of the - * last descendant process is finally completed (assuming that + * last process is finally completed (assuming that * nothing else has changed for the group in the meantime, in * terms of composition of the group and active/inactive state of child * groups and processes). To accomplish this, an additional @@ -524,7 +524,7 @@ struct bfq_data { * we resort to the following tradeoff between simplicity and * accuracy: for an inactive group that is still counted in * num_groups_with_pending_reqs, we decrement - * num_groups_with_pending_reqs when the first descendant + * num_groups_with_pending_reqs when the last * process of the group remains with no request waiting for * completion. * @@ -532,12 +532,12 @@ struct bfq_data { * carefulness: to avoid multiple decrements, we flag a group, * more precisely an entity representing a group, as still * counted in num_groups_with_pending_reqs when it becomes - * inactive. Then, when the first descendant queue of the + * inactive. Then, when the last queue of the * entity remains with no request waiting for completion, * num_groups_with_pending_reqs is decremented, and this flag * is reset. After this flag is reset for the entity, * num_groups_with_pending_reqs won't be decremented any - * longer in case a new descendant queue of the entity remains + * longer in case a new queue of the entity remains * with no request waiting for completion. */ unsigned int num_groups_with_pending_reqs;
Currently 'num_groups_with_pending_reqs' won't be decreased when the group doesn't have any pending requests, while some child group still have pending requests. The decrement is delayed to when all the child groups doesn't have any pending requests. For example: 1) t1 issue sync io on root group, t2 and t3 issue sync io on the same child group. num_groups_with_pending_reqs is 2 now. 2) t1 stopped, num_groups_with_pending_reqs is still 2. io from t2 and t3 still can't be handled concurrently. Fix the problem by decreasing 'num_groups_with_pending_reqs' immediately upon the weights_tree removal of last bfqq of the group. Signed-off-by: Yu Kuai <yukuai3@huawei.com> --- block/bfq-iosched.c | 56 +++++++++++++++------------------------------ block/bfq-iosched.h | 16 ++++++------- 2 files changed, 27 insertions(+), 45 deletions(-)