| Message ID | 4A44833F.8040308@cn.fujitsu.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Fri, Jun 26, 2009 at 04:13:51PM +0800, Gui Jianfeng wrote: > let rt queue preempt non-rt queue if needed. > Make sure comparision happens at the same level. > > Signed-off-by: Gui Jianfeng <guijianfeng@cn.fujitsu.com> > --- > block/elevator-fq.c | 28 +++++++++++++++++++++++++++- > 1 files changed, 27 insertions(+), 1 deletions(-) > > diff --git a/block/elevator-fq.c b/block/elevator-fq.c > index 1d4ec1f..21d38f5 100644 > --- a/block/elevator-fq.c > +++ b/block/elevator-fq.c > @@ -3742,6 +3742,31 @@ int elv_iosched_expire_ioq(struct request_queue *q, int slice_expired, > return ret; > } > > +static int check_rt_preemption(struct io_queue *ioq) > +{ > + struct hlist_node *node; > + struct hlist_head *hhead = &ioq->efqd->rt_ioq_list; > + struct io_queue *rt_ioq; > + struct io_entity *entity = &ioq->entity; > + struct io_entity *new_entity; > + > + rcu_read_lock(); > + hlist_for_each_entry_rcu(rt_ioq, node, hhead, rt_node) { > + new_entity = &rt_ioq->entity; > + > + bfq_find_matching_entity(&entity, &new_entity); > + > + if (new_entity->ioprio_class == IOPRIO_CLASS_RT && > + entity->ioprio_class != IOPRIO_CLASS_RT) { > + rcu_read_unlock(); > + return 1; > + } > + } > + rcu_read_unlock(); > + > + return 0; > +} > + > /* Common layer function to select the next queue to dispatch from */ > void *elv_fq_select_ioq(struct request_queue *q, int force) > { > @@ -3823,7 +3848,8 @@ void *elv_fq_select_ioq(struct request_queue *q, int force) > */ > iog = ioq_to_io_group(ioq); > > - if (!elv_ioq_class_rt(ioq) && iog->busy_rt_queues) { > + if (!elv_ioq_class_rt(ioq) && > + (iog->busy_rt_queues || check_rt_preemption(ioq))) { > /* Hi Gui, I am not able to understand why do we need above changes? BFQ scheduler already takes care of selecting an RT queue for dispatch (if the queue is entitled to). In case a new RT queue backlogged while a BE queue is being served, we do preemtion check to make sure RT queue gets to run as soon as possible. In fact I think that busy_rt_queues infrastructure is also redundant and I plan to get rid of it. Can you please help me understand what use case are you addressing with above patch? Thanks Vivek -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel
diff --git a/block/elevator-fq.c b/block/elevator-fq.c index 1d4ec1f..21d38f5 100644 --- a/block/elevator-fq.c +++ b/block/elevator-fq.c @@ -3742,6 +3742,31 @@ int elv_iosched_expire_ioq(struct request_queue *q, int slice_expired, return ret; } +static int check_rt_preemption(struct io_queue *ioq) +{ + struct hlist_node *node; + struct hlist_head *hhead = &ioq->efqd->rt_ioq_list; + struct io_queue *rt_ioq; + struct io_entity *entity = &ioq->entity; + struct io_entity *new_entity; + + rcu_read_lock(); + hlist_for_each_entry_rcu(rt_ioq, node, hhead, rt_node) { + new_entity = &rt_ioq->entity; + + bfq_find_matching_entity(&entity, &new_entity); + + if (new_entity->ioprio_class == IOPRIO_CLASS_RT && + entity->ioprio_class != IOPRIO_CLASS_RT) { + rcu_read_unlock(); + return 1; + } + } + rcu_read_unlock(); + + return 0; +} + /* Common layer function to select the next queue to dispatch from */ void *elv_fq_select_ioq(struct request_queue *q, int force) { @@ -3823,7 +3848,8 @@ void *elv_fq_select_ioq(struct request_queue *q, int force) */ iog = ioq_to_io_group(ioq); - if (!elv_ioq_class_rt(ioq) && iog->busy_rt_queues) { + if (!elv_ioq_class_rt(ioq) && + (iog->busy_rt_queues || check_rt_preemption(ioq))) { /* * We simulate this as cfqq timed out so that it gets to bank * the remaining of its time slice.
let rt queue preempt non-rt queue if needed. Make sure comparision happens at the same level. Signed-off-by: Gui Jianfeng <guijianfeng@cn.fujitsu.com> --- block/elevator-fq.c | 28 +++++++++++++++++++++++++++- 1 files changed, 27 insertions(+), 1 deletions(-)