| Message ID | 87twnjb7lq.fsf@notabene.neil.brown.name (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Tue, Dec 15, 2015 at 6:44 PM, NeilBrown <neilb@suse.com> wrote: > > Commit: c05eecf63610 ("SUNRPC: Don't allow low priority tasks to pre-empt higher priority ones") > > removed the 'fair scheduling' feature from SUNRPC priority queues. > This feature caused problems for some queues (send queue and session slot queue) > but is still needed for others, particularly the tcp slot queue. > > Without fairness, reads (priority 1) can starve background writes > (priority 0) so a streaming read can cause writeback to block > indefinitely. This is not easy to measure with default settings as > the current slot table size is much larger than the read-ahead size. > However if the slot-table size is reduced (seen when backporting to > older kernels with a limited size) the problem is easily demonstrated. > > This patch conditionally restores fair scheduling. It is now the > default unless rpc_sleep_on_priority() is called directly. Then the > queue switches to strict priority observance. > > As that function is called for both the send queue and the session > slot queue and not for any others, this has exactly the desired > effect. > > The "count" field that was removed by the previous patch is restored. > A value for '255' means "strict priority queuing, no fair queuing". > Any other value is a could of owners to be processed before switching > to a different priority level, just like before. > > Signed-off-by: NeilBrown <neilb@suse.com> > --- > > It is quite possible that you won't like the overloading of > rpc_sleep_on_priority() to disable fair-scheduling and would prefer an > extra arg to rpc_init_priority_wait_queue(). I can do it that way if > you like. > NeilBrown > > > include/linux/sunrpc/sched.h | 1 + > net/sunrpc/sched.c | 12 +++++++++--- > 2 files changed, 10 insertions(+), 3 deletions(-) > > diff --git a/include/linux/sunrpc/sched.h b/include/linux/sunrpc/sched.h > index d703f0ef37d8..985efe8d7e26 100644 > --- a/include/linux/sunrpc/sched.h > +++ b/include/linux/sunrpc/sched.h > @@ -184,6 +184,7 @@ struct rpc_wait_queue { > pid_t owner; /* process id of last task serviced */ > unsigned char maxpriority; /* maximum priority (0 if queue is not a priority queue) */ > unsigned char priority; /* current priority */ > + unsigned char count; /* # task groups remaining to be serviced */ > unsigned char nr; /* # tasks remaining for cookie */ > unsigned short qlen; /* total # tasks waiting in queue */ > struct rpc_timer timer_list; > diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c > index 73ad57a59989..e8fcd4f098bb 100644 > --- a/net/sunrpc/sched.c > +++ b/net/sunrpc/sched.c > @@ -117,6 +117,8 @@ static void rpc_set_waitqueue_priority(struct rpc_wait_queue *queue, int priorit > rpc_rotate_queue_owner(queue); > queue->priority = priority; > } > + if (queue->count != 255) > + queue->count = 1 << (priority * 2); > } > > static void rpc_set_waitqueue_owner(struct rpc_wait_queue *queue, pid_t pid) > @@ -144,8 +146,10 @@ static void __rpc_add_wait_queue_priority(struct rpc_wait_queue *queue, > INIT_LIST_HEAD(&task->u.tk_wait.links); > if (unlikely(queue_priority > queue->maxpriority)) > queue_priority = queue->maxpriority; > - if (queue_priority > queue->priority) > - rpc_set_waitqueue_priority(queue, queue_priority); > + if (queue->count == 255) { > + if (queue_priority > queue->priority) > + rpc_set_waitqueue_priority(queue, queue_priority); > + } > q = &queue->tasks[queue_priority]; > list_for_each_entry(t, q, u.tk_wait.list) { > if (t->tk_owner == task->tk_owner) { > @@ -401,6 +405,7 @@ void rpc_sleep_on_priority(struct rpc_wait_queue *q, struct rpc_task *task, > * Protect the queue operations. > */ > spin_lock_bh(&q->lock); > + q->count = 255; > __rpc_sleep_on_priority(q, task, action, priority - RPC_PRIORITY_LOW); > spin_unlock_bh(&q->lock); > } > @@ -478,7 +483,8 @@ static struct rpc_task *__rpc_find_next_queued_priority(struct rpc_wait_queue *q > /* > * Check if we need to switch queues. > */ > - goto new_owner; > + if (queue->count == 255 || --queue->count) > + goto new_owner; > } > > /* > Are we sure there is value in keeping FLUSH_LOWPRI for background writes? Cheers Trond -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, Dec 16 2015, Trond Myklebust wrote: > On Tue, Dec 15, 2015 at 6:44 PM, NeilBrown <neilb@suse.com> wrote: >> >> Commit: c05eecf63610 ("SUNRPC: Don't allow low priority tasks to pre-empt higher priority ones") >> >> removed the 'fair scheduling' feature from SUNRPC priority queues. >> This feature caused problems for some queues (send queue and session slot queue) >> but is still needed for others, particularly the tcp slot queue. >> >> Without fairness, reads (priority 1) can starve background writes >> (priority 0) so a streaming read can cause writeback to block >> indefinitely. This is not easy to measure with default settings as >> the current slot table size is much larger than the read-ahead size. >> However if the slot-table size is reduced (seen when backporting to >> older kernels with a limited size) the problem is easily demonstrated. >> >> This patch conditionally restores fair scheduling. It is now the >> default unless rpc_sleep_on_priority() is called directly. Then the >> queue switches to strict priority observance. >> >> As that function is called for both the send queue and the session >> slot queue and not for any others, this has exactly the desired >> effect. >> >> The "count" field that was removed by the previous patch is restored. >> A value for '255' means "strict priority queuing, no fair queuing". >> Any other value is a could of owners to be processed before switching >> to a different priority level, just like before. >> >> Signed-off-by: NeilBrown <neilb@suse.com> >> --- >> >> It is quite possible that you won't like the overloading of >> rpc_sleep_on_priority() to disable fair-scheduling and would prefer an >> extra arg to rpc_init_priority_wait_queue(). I can do it that way if >> you like. >> NeilBrown >> >> >> include/linux/sunrpc/sched.h | 1 + >> net/sunrpc/sched.c | 12 +++++++++--- >> 2 files changed, 10 insertions(+), 3 deletions(-) >> >> diff --git a/include/linux/sunrpc/sched.h b/include/linux/sunrpc/sched.h >> index d703f0ef37d8..985efe8d7e26 100644 >> --- a/include/linux/sunrpc/sched.h >> +++ b/include/linux/sunrpc/sched.h >> @@ -184,6 +184,7 @@ struct rpc_wait_queue { >> pid_t owner; /* process id of last task serviced */ >> unsigned char maxpriority; /* maximum priority (0 if queue is not a priority queue) */ >> unsigned char priority; /* current priority */ >> + unsigned char count; /* # task groups remaining to be serviced */ >> unsigned char nr; /* # tasks remaining for cookie */ >> unsigned short qlen; /* total # tasks waiting in queue */ >> struct rpc_timer timer_list; >> diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c >> index 73ad57a59989..e8fcd4f098bb 100644 >> --- a/net/sunrpc/sched.c >> +++ b/net/sunrpc/sched.c >> @@ -117,6 +117,8 @@ static void rpc_set_waitqueue_priority(struct rpc_wait_queue *queue, int priorit >> rpc_rotate_queue_owner(queue); >> queue->priority = priority; >> } >> + if (queue->count != 255) >> + queue->count = 1 << (priority * 2); >> } >> >> static void rpc_set_waitqueue_owner(struct rpc_wait_queue *queue, pid_t pid) >> @@ -144,8 +146,10 @@ static void __rpc_add_wait_queue_priority(struct rpc_wait_queue *queue, >> INIT_LIST_HEAD(&task->u.tk_wait.links); >> if (unlikely(queue_priority > queue->maxpriority)) >> queue_priority = queue->maxpriority; >> - if (queue_priority > queue->priority) >> - rpc_set_waitqueue_priority(queue, queue_priority); >> + if (queue->count == 255) { >> + if (queue_priority > queue->priority) >> + rpc_set_waitqueue_priority(queue, queue_priority); >> + } >> q = &queue->tasks[queue_priority]; >> list_for_each_entry(t, q, u.tk_wait.list) { >> if (t->tk_owner == task->tk_owner) { >> @@ -401,6 +405,7 @@ void rpc_sleep_on_priority(struct rpc_wait_queue *q, struct rpc_task *task, >> * Protect the queue operations. >> */ >> spin_lock_bh(&q->lock); >> + q->count = 255; >> __rpc_sleep_on_priority(q, task, action, priority - RPC_PRIORITY_LOW); >> spin_unlock_bh(&q->lock); >> } >> @@ -478,7 +483,8 @@ static struct rpc_task *__rpc_find_next_queued_priority(struct rpc_wait_queue *q >> /* >> * Check if we need to switch queues. >> */ >> - goto new_owner; >> + if (queue->count == 255 || --queue->count) >> + goto new_owner; >> } >> >> /* >> > > Are we sure there is value in keeping FLUSH_LOWPRI for background writes? There is currently also FLUSH_HIGHPRI for "for_reclaim" writes. Should they be allowed to starve reads? If you treated all reads and writed the same, then I can't see value in restoring fair scheduling. If there is any difference, then I suspect we do need the fairness. Thanks, NeilBrown
On Tue, Dec 15, 2015 at 10:10 PM, NeilBrown <neilb@suse.com> wrote: > On Wed, Dec 16 2015, Trond Myklebust wrote: > >> On Tue, Dec 15, 2015 at 6:44 PM, NeilBrown <neilb@suse.com> wrote: >>> >>> Commit: c05eecf63610 ("SUNRPC: Don't allow low priority tasks to pre-empt higher priority ones") >>> >>> removed the 'fair scheduling' feature from SUNRPC priority queues. >>> This feature caused problems for some queues (send queue and session slot queue) >>> but is still needed for others, particularly the tcp slot queue. >>> >>> Without fairness, reads (priority 1) can starve background writes >>> (priority 0) so a streaming read can cause writeback to block >>> indefinitely. This is not easy to measure with default settings as >>> the current slot table size is much larger than the read-ahead size. >>> However if the slot-table size is reduced (seen when backporting to >>> older kernels with a limited size) the problem is easily demonstrated. >>> >>> This patch conditionally restores fair scheduling. It is now the >>> default unless rpc_sleep_on_priority() is called directly. Then the >>> queue switches to strict priority observance. >>> >>> As that function is called for both the send queue and the session >>> slot queue and not for any others, this has exactly the desired >>> effect. >>> >>> The "count" field that was removed by the previous patch is restored. >>> A value for '255' means "strict priority queuing, no fair queuing". >>> Any other value is a could of owners to be processed before switching >>> to a different priority level, just like before. <snip> >> Are we sure there is value in keeping FLUSH_LOWPRI for background writes? > > There is currently also FLUSH_HIGHPRI for "for_reclaim" writes. > Should they be allowed to starve reads? > > If you treated all reads and writed the same, then I can't see value in > restoring fair scheduling. If there is any difference, then I suspect > we do need the fairness. I disagree. Reclaiming memory should always be able to pre-empt "interactive" features such as read. Everything goes down the toilet when we force the kernel into situations where it needs to swap. Cheers Trond -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Sun, Dec 27 2015, Trond Myklebust wrote: > On Tue, Dec 15, 2015 at 10:10 PM, NeilBrown <neilb@suse.com> wrote: >> On Wed, Dec 16 2015, Trond Myklebust wrote: >> >>> On Tue, Dec 15, 2015 at 6:44 PM, NeilBrown <neilb@suse.com> wrote: >>>> >>>> Commit: c05eecf63610 ("SUNRPC: Don't allow low priority tasks to pre-empt higher priority ones") >>>> >>>> removed the 'fair scheduling' feature from SUNRPC priority queues. >>>> This feature caused problems for some queues (send queue and session slot queue) >>>> but is still needed for others, particularly the tcp slot queue. >>>> >>>> Without fairness, reads (priority 1) can starve background writes >>>> (priority 0) so a streaming read can cause writeback to block >>>> indefinitely. This is not easy to measure with default settings as >>>> the current slot table size is much larger than the read-ahead size. >>>> However if the slot-table size is reduced (seen when backporting to >>>> older kernels with a limited size) the problem is easily demonstrated. >>>> >>>> This patch conditionally restores fair scheduling. It is now the >>>> default unless rpc_sleep_on_priority() is called directly. Then the >>>> queue switches to strict priority observance. >>>> >>>> As that function is called for both the send queue and the session >>>> slot queue and not for any others, this has exactly the desired >>>> effect. >>>> >>>> The "count" field that was removed by the previous patch is restored. >>>> A value for '255' means "strict priority queuing, no fair queuing". >>>> Any other value is a could of owners to be processed before switching >>>> to a different priority level, just like before. > <snip> >>> Are we sure there is value in keeping FLUSH_LOWPRI for background writes? >> >> There is currently also FLUSH_HIGHPRI for "for_reclaim" writes. >> Should they be allowed to starve reads? >> >> If you treated all reads and writed the same, then I can't see value in >> restoring fair scheduling. If there is any difference, then I suspect >> we do need the fairness. > > I disagree. Reclaiming memory should always be able to pre-empt > "interactive" features such as read. Everything goes down the toilet > when we force the kernel into situations where it needs to swap. That's your call I guess. I certainly agree that memory-reclaim writes should get some priority (e.g. two writes serviced for every read). Whether they should be allowed to completely block reads I'm less sure of. But it is probably purely academic as if the system is busy reclaiming you are unlikely to have any reads to want to send. My problem would be solved (I think) by treating reads and non-reclaim writes as equals. I'll make a patch, see if I can test it, and let you know. Thanks, NeilBrown
On Wed, Feb 10 2016, NeilBrown wrote: > On Sun, Dec 27 2015, Trond Myklebust wrote: > >> On Tue, Dec 15, 2015 at 10:10 PM, NeilBrown <neilb@suse.com> wrote: >>> If you treated all reads and writed the same, then I can't see value in >>> restoring fair scheduling. If there is any difference, then I suspect >>> we do need the fairness. >> >> I disagree. Reclaiming memory should always be able to pre-empt >> "interactive" features such as read. Everything goes down the toilet >> when we force the kernel into situations where it needs to swap. > > That's your call I guess. I certainly agree that memory-reclaim writes > should get some priority (e.g. two writes serviced for every read). > Whether they should be allowed to completely block reads I'm less sure > of. But it is probably purely academic as if the system is busy > reclaiming you are unlikely to have any reads to want to send. > > My problem would be solved (I think) by treating reads and non-reclaim > writes as equals. I'll make a patch, see if I can test it, and let you > know. ahh.... I just discovered Commit: b0ac1bd2bbfd ("NFS: Background flush should not be low priority") I didn't notice that before. I suspect that will fix the problem - thanks. I'll try testing and let you know if there is a problem. Thanks, NeilBrown
diff --git a/include/linux/sunrpc/sched.h b/include/linux/sunrpc/sched.h index d703f0ef37d8..985efe8d7e26 100644 --- a/include/linux/sunrpc/sched.h +++ b/include/linux/sunrpc/sched.h @@ -184,6 +184,7 @@ struct rpc_wait_queue { pid_t owner; /* process id of last task serviced */ unsigned char maxpriority; /* maximum priority (0 if queue is not a priority queue) */ unsigned char priority; /* current priority */ + unsigned char count; /* # task groups remaining to be serviced */ unsigned char nr; /* # tasks remaining for cookie */ unsigned short qlen; /* total # tasks waiting in queue */ struct rpc_timer timer_list; diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c index 73ad57a59989..e8fcd4f098bb 100644 --- a/net/sunrpc/sched.c +++ b/net/sunrpc/sched.c @@ -117,6 +117,8 @@ static void rpc_set_waitqueue_priority(struct rpc_wait_queue *queue, int priorit rpc_rotate_queue_owner(queue); queue->priority = priority; } + if (queue->count != 255) + queue->count = 1 << (priority * 2); } static void rpc_set_waitqueue_owner(struct rpc_wait_queue *queue, pid_t pid) @@ -144,8 +146,10 @@ static void __rpc_add_wait_queue_priority(struct rpc_wait_queue *queue, INIT_LIST_HEAD(&task->u.tk_wait.links); if (unlikely(queue_priority > queue->maxpriority)) queue_priority = queue->maxpriority; - if (queue_priority > queue->priority) - rpc_set_waitqueue_priority(queue, queue_priority); + if (queue->count == 255) { + if (queue_priority > queue->priority) + rpc_set_waitqueue_priority(queue, queue_priority); + } q = &queue->tasks[queue_priority]; list_for_each_entry(t, q, u.tk_wait.list) { if (t->tk_owner == task->tk_owner) { @@ -401,6 +405,7 @@ void rpc_sleep_on_priority(struct rpc_wait_queue *q, struct rpc_task *task, * Protect the queue operations. */ spin_lock_bh(&q->lock); + q->count = 255; __rpc_sleep_on_priority(q, task, action, priority - RPC_PRIORITY_LOW); spin_unlock_bh(&q->lock); } @@ -478,7 +483,8 @@ static struct rpc_task *__rpc_find_next_queued_priority(struct rpc_wait_queue *q /* * Check if we need to switch queues. */ - goto new_owner; + if (queue->count == 255 || --queue->count) + goto new_owner; } /*
Commit: c05eecf63610 ("SUNRPC: Don't allow low priority tasks to pre-empt higher priority ones") removed the 'fair scheduling' feature from SUNRPC priority queues. This feature caused problems for some queues (send queue and session slot queue) but is still needed for others, particularly the tcp slot queue. Without fairness, reads (priority 1) can starve background writes (priority 0) so a streaming read can cause writeback to block indefinitely. This is not easy to measure with default settings as the current slot table size is much larger than the read-ahead size. However if the slot-table size is reduced (seen when backporting to older kernels with a limited size) the problem is easily demonstrated. This patch conditionally restores fair scheduling. It is now the default unless rpc_sleep_on_priority() is called directly. Then the queue switches to strict priority observance. As that function is called for both the send queue and the session slot queue and not for any others, this has exactly the desired effect. The "count" field that was removed by the previous patch is restored. A value for '255' means "strict priority queuing, no fair queuing". Any other value is a could of owners to be processed before switching to a different priority level, just like before. Signed-off-by: NeilBrown <neilb@suse.com> --- It is quite possible that you won't like the overloading of rpc_sleep_on_priority() to disable fair-scheduling and would prefer an extra arg to rpc_init_priority_wait_queue(). I can do it that way if you like. NeilBrown include/linux/sunrpc/sched.h | 1 + net/sunrpc/sched.c | 12 +++++++++--- 2 files changed, 10 insertions(+), 3 deletions(-)