blk-mq: Fix two causes of IO stalls found in reboot testing
Subsystems (2)
v2 cover blk-mq: Fix two causes of IO stalls found in reboot testing
While doing reboot testing, I found that occasionally my device would trigger the hung task detector. Many tasks were stuck waiting for the a blkdev mutex, but at least one task in the system was always sitting waiting for IO to complete (and holding the blkdev mutex). One example of a task that was just waiting for its IO to complete on one reboot:
udevd D 0 2177 306 0x00400209
Call trace: __switch_to+0x15c/0x17c __schedule+0x6e0/0x928 schedule+0x8c/0xbc schedule_timeout+0x9c/0xfc io_schedule_timeout+0x24/0x48 do_wait_for_common+0xd0/0x160 wait_for_completion_io_timeout+0x54/0x74 blk_execute_rq+0x9c/0xd8 __scsi_execute+0x104/0x198 scsi_test_unit_ready+0xa0/0x154 sd_check_events+0xb4/0x164 disk_check_events+0x58/0x154 disk_clear_events+0x74/0x110 check_disk_change+0x28/0x6c sd_open+0x5c/0x130 __blkdev_get+0x20c/0x3d4 blkdev_get+0x74/0x170 blkdev_open+0x94/0xa8 do_dentry_open+0x268/0x3a0 vfs_open+0x34/0x40 path_openat+0x39c/0xdf4 do_filp_open+0x90/0x10c do_sys_open+0x150/0x3c8 ...
I've reproduced this on two systems: one boots from an internal UFS disk and one from eMMC. Each has a card reader attached via USB with an SD card plugged in. On the USB-attached SD card is a disk with 12 partitions (a Chrome OS test image), if it matters. The system doesn't do much with the USB disk other than probe it (it's plugged in my system to help me recover).
From digging, I believe that there are two separate but related issues. Both issues relate to the SCSI code saying that there is no budget.
I have done testing with only one or the other of the two patches in this series and found that I could still encounter hung tasks if only one of the two patches was applied. This deserves a bit of explanation. To me, it's fairly obvious that the first fix wouldn't fix the problems talked about in the second patch. However, it's less obvious why the second patch doesn't fix the problems in blk_mq_dispatch_rq_list(). It turns out that it _almost_ does (problems become much more rare), but I did manage to get a single trace where the "kick" scheduled by the second patch happened really quickly. The scheduled kick then ran and found nothing to do. This happened in parallel to a task running in blk_mq_dispatch_rq_list() which hadn't gotten around to splicing the list back into hctx->dispatch. This is why we need both fixes or a heavier hammer where we always kick whenever two threads request budget at the same time.
Most of my testing has been atop Chrome OS 5.4's kernel tree which currently has v5.4.28 merged in. The Chrome OS 5.4 tree also has a patch by Salman Qazi, namely ("block: Limit number of items taken from the I/O scheduler in one go"). Reverting that patch didn't make the hung tasks go away, so I kept it in for most of my testing.
I have also done some testing on mainline Linux (git describe says I'm on v5.6-rc7-227-gf3e69428b5e2) even without Salman's patch. I found that I could reproduce the problems there and that traces looked about the same as I saw on the downstream branch. These patches were also confirmed to fix the problems on mainline.
Chrome OS is currently setup to use the BFQ scheduler and I found that I couldn't reproduce the problems without BFQ. As discussed in the second patch this is believed to be because BFQ sometimes returns "true" from has_work() but then NULL from dispatch_request().
I'll insert my usual caveat that I'm sending patches to code that I know very little about. If I'm making a total bozo patch here, please help me figure out how I should fix the problems I found in a better way.
If you want to see a total ridiculous amount of chatter where I stumbled around a whole bunch trying to figure out what was wrong and how to fix it, feel free to read <https://crbug.com/1061950>. I promise it will make your eyes glaze over right away if this cover letter didn't already do that. Specifically comment 79 in that bug includes a link to my ugly prototype of making BFQ's has_work() more exact (I only managed it by actually defining _both_ an exact and inexact function to avoid circular locking problems when it was called directly from blk_mq_hctx_has_pending()). Comment 79 also has more thoughts about alternatives considered.
I don't know if these fixes represent a regression of some sort or are new. As per above I could only reproduce with BFQ enabled which makes it nearly impossible to go too far back with this. I haven't listed any "Fixes" tags here, but if someone felt it was appropriate to backport this to some stable trees that seems like it'd be nice. Presumably at least 5.4 stable would make sense.
Thanks to Salman Qazi, Paolo Valente, and Guenter Roeck who spent a bunch of time helping me trawl through some of this code and reviewing early versions of this patch.
Changes in v2:
- Replace ("scsi: core: Fix stall...") w/ ("blk-mq: Rerun dispatch...")
Douglas Anderson (2):
blk-mq: In blk_mq_dispatch_rq_list() "no budget" is a reason to kick
blk-mq: Rerun dispatching in the case of budget contention
block/blk-mq-sched.c | 26 ++++++++++++++++++++++++-- block/blk-mq.c | 14 +++++++++++--- include/linux/blkdev.h | 2 ++ 3 files changed, 37 insertions(+), 5 deletions(-)
Signature · 1 line
2.26.0.rc2.310.g2932bb562d-goog
PATCH v2 1/2 blk-mq: In blk_mq_dispatch_rq_list() "no budget" is a reason to kick
In blk_mq_dispatch_rq_list(), if blk_mq_sched_needs_restart() returns true and the driver returns BLK_STS_RESOURCE then we'll kick the queue. However, there's another case where we might need to kick it. If we were unable to get budget we can be in much the same state as when the driver returns BLK_STS_RESOURCE, so we should treat it the same.
---
Changes in v2: None
block/blk-mq.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-)
block/blk-mq.c +8 -3
@@ -1189,6 +1189,7 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list,bool no_tag = false;int errors, queued;blk_status_t ret =BLK_STS_OK ; +bool no_budget_avail = false;if (list_empty (list))return false;
@@ -1205,8 +1206,10 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list, rq = list_first_entry(list, struct request, queuelist); hctx = rq->mq_hctx; -if (!got_budget && !blk_mq_get_dispatch_budget (hctx)) +if (!got_budget && !blk_mq_get_dispatch_budget (hctx)) { + no_budget_avail = true;break ; + }if (!blk_mq_get_driver_tag (rq)) {/*
@@ -1311,13 +1314,15 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list,* * If driver returns BLK_STS_RESOURCE and SCHED_RESTART * bit is set, run queue after a delay to avoid IO stalls -* that could otherwise occur if the queue is idle. +* that could otherwise occur if the queue is idle. We'll do +* similar if we couldn't get budget and SCHED_RESTART is set. */ needs_restart =blk_mq_sched_needs_restart (hctx);if (!needs_restart || (no_tag &&list_empty_careful (&hctx->dispatch_wait.entry)))blk_mq_run_hw_queue (hctx, true); -else if (needs_restart && (ret ==BLK_STS_RESOURCE )) +else if (needs_restart && (ret ==BLK_STS_RESOURCE || + no_budget_avail))blk_mq_delay_run_hw_queue (hctx,BLK_MQ_RESOURCE_DELAY );blk_mq_update_dispatch_busy (hctx, true);
Signature · 1 line
2.26.0.rc2.310.g2932bb562d-goog
Re: [PATCH v2 1/2] blk-mq: In blk_mq_dispatch_rq_list() "no budget" is a reason to kick
Quoted 6 lines · Douglas Anderson
In
blk_mq_dispatch_rq_list(), ifblk_mq_sched_needs_restart()returns true and the driver returnsBLK_STS_RESOURCEthen we'll kick the queue. However, there's another case where we might need to kick it. If we were unable to get budget we can be in much the same state as when the driver returnsBLK_STS_RESOURCE, so we should treat it the same.
The situation is really similar, especially the restart handler may not see the request which may not be added to hctx->dispatch, so
Thanks,
Ming
PATCH v2 2/2 blk-mq: Rerun dispatching in the case of budget contention
In reply to Douglas AndersonIt is possible for two threads to be running blk_mq_do_dispatch_sched() at the same time with the same "hctx". This is because there can be more than one caller to __blk_mq_run_hw_queue() with the same "hctx" and hctx_lock() doesn't prevent more than one thread from entering.
If more than one thread is running blk_mq_do_dispatch_sched() at the same time with the same "hctx", they may have contention acquiring budget. The blk_mq_get_dispatch_budget() can eventually translate into scsi_mq_get_budget(). If the device's "queue_depth" is 1 (not uncommon) then only one of the two threads will be the one to increment "device_busy" to 1 and get the budget.
The losing thread will break out of blk_mq_do_dispatch_sched() and will stop dispatching requests. The assumption is that when more budget is available later (when existing transactions finish) the queue will be kicked again, perhaps in scsi_end_request().
The winning thread now has budget and can go on to call dispatch_request(). If dispatch_request() returns NULL here then we have a potential problem. Specifically we'll now call blk_mq_put_dispatch_budget() which translates into scsi_mq_put_budget(). That will mark the device as no longer busy but doesn't do anything to kick the queue. This violates the assumption that the queue would be kicked when more budget was available.
Pictorially:
Thread A Thread B
================================= ================================== blk_mq_get_dispatch_budget() => 1dispatch_request() => NULL
blk_mq_get_dispatch_budget() => 0
// because Thread A marked
// "device_busy" in scsi_deviceblk_mq_put_dispatch_budget()
The above case was observed in reboot tests and caused a task to hang forever waiting for IO to complete. Traces showed that in fact two tasks were running blk_mq_do_dispatch_sched() at the same time with the same "hctx". The task that got the budget did in fact see dispatch_request() return NULL. Both tasks returned and the system went on for several minutes (until the hung task delay kicked in) without the given "hctx" showing up again in traces.
Let's attempt to fix this problem by detecting if there was contention for the budget in the case where we put the budget without dispatching anything. If we saw contention we kick all hctx's associated with the queue where there was contention. We do this without any locking by adding a double-check for budget and accepting a small amount of faux contention if the 2nd check gives us budget but then we don't dispatch anything (we'll look like we contended with ourselves).
A few extra notes:
- This whole thing is only a problem due to the inexact nature of has_work(). Specifically if has_work() always guaranteed that a "true" return meant that dispatch_request() would return non-NULL then we wouldn't have this problem. That's because we only grab the budget if has_work() returned true. If we had the non-NULL guarantee then at least one of the threads would actually dispatch work and when the work was done then queues would be kicked normally.
- One specific I/O scheduler that trips this problem quite a bit is BFQ which definitely returns "true" for has_work() in cases where it wouldn't dispatch. Making BFQ's has_work() more exact requires that has_work() becomes a much heavier function, including figuring out how to acquire spinlocks in has_work() without tripping circular lock dependencies. This is prototyped but it's unclear if it's really the way to go when the problem can be solved with a relatively lightweight contention detection mechanism.
- Because this problem only trips with inexact has_work() it's believed that blk_mq_do_dispatch_ctx() does not need a similar change.
---
Changes in v2:
- Replace ("scsi: core: Fix stall...") w/ ("blk-mq: Rerun dispatch...")
block/blk-mq-sched.c | 26 ++++++++++++++++++++++++-- block/blk-mq.c | 3 +++ include/linux/blkdev.h | 2 ++ 3 files changed, 29 insertions(+), 2 deletions(-)
block/blk-mq-sched.c +24 -2
@@ -97,12 +97,34 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)if (e->type->ops.has_work && !e->type->ops.has_work (hctx))break ; -if (!blk_mq_get_dispatch_budget (hctx)) -break ; + +if (!blk_mq_get_dispatch_budget (hctx)) { +/* +* We didn't get budget so set contention. If +* someone else had the budget but didn't dispatch +* they'll kick everything. NOTE: we check one +* extra time _after_ setting contention to fully +* close the race. If we don't actually dispatch +* we'll detext faux contention (with ourselves) +* but that should be rare. +*/ +atomic_set (&q->budget_contention,1 ); + +if (!blk_mq_get_dispatch_budget (hctx)) +break ; + } rq = e->type->ops.dispatch_request (hctx);if (!rq) {blk_mq_put_dispatch_budget (hctx); + +/* +* We've released the budget but us holding it might +* have prevented someone else from dispatching. +* Detect that case and run all queues again. +*/ +if (atomic_read (&q->budget_contention)) +blk_mq_run_hw_queues (q, true);break ; }
block/blk-mq.c +3 -0
@@ -1528,6 +1528,9 @@ void blk_mq_run_hw_queues(struct request_queue *q, bool async)struct blk_mq_hw_ctx *hctx;int i; +/* We're running the queues, so clear the contention detector */ +atomic_set (&q->budget_contention,0 ); +queue_for_each_hw_ctx (q, hctx, i) {if (blk_mq_hctx_stopped (hctx))continue ;
include/linux/blkdev.h +2 -0
@@ -583,6 +583,8 @@ struct request_queue {#define BLK_MAX_WRITE_HINTS 5u64 write_hints[BLK_MAX_WRITE_HINTS ]; + +atomic_t budget_contention; };#define QUEUE_FLAG_STOPPED 0/* queue is stopped */
Signature · 1 line
2.26.0.rc2.310.g2932bb562d-goog
Re: [PATCH v2 2/2] blk-mq: Rerun dispatching in the case of budget contention
Quoted 13 lines · Douglas Anderson
It is possible for two threads to be running
blk_mq_do_dispatch_sched()at the same time with the same "hctx". This is because there can be more than one caller to__blk_mq_run_hw_queue()with the same "hctx" andhctx_lock()doesn't prevent more than one thread from entering.If more than one thread is running
blk_mq_do_dispatch_sched()at the same time with the same "hctx", they may have contention acquiring budget. Theblk_mq_get_dispatch_budget()can eventually translate intoscsi_mq_get_budget(). If the device's "queue_depth" is 1 (not uncommon) then only one of the two threads will be the one to increment "device_busy" to 1 and get the budget.
Quoted 8 lines
The losing thread will break out of
blk_mq_do_dispatch_sched()and will stop dispatching requests. The assumption is that when more budget is available later (when existing transactions finish) the queue will be kicked again, perhaps inscsi_end_request().The winning thread now has budget and can go on to call
dispatch_request(). Ifdispatch_request()returns NULL here then we have a potential problem. Specifically we'll now call
As I mentioned before, it is a BFQ specific issue, it tells blk-mq that it has work to do, and now the budget is assigned to the only winning thread, however, dispatch_request() still returns NULL.
Quoted 4 lines
blk_mq_put_dispatch_budget()which translates intoscsi_mq_put_budget(). That will mark the device as no longer busy but doesn't do anything to kick the queue. This violates the assumption that the queue would be kicked when more budget was available.
The queue is still kicked in by BFQ in its idle timer, however that timer doesn't make forward progress.
Without this idle timer, it is simply a BFQ issue.
Quoted 11 lines
Pictorially:
Thread A Thread B
================================= ==================================blk_mq_get_dispatch_budget()=> 1dispatch_request()=> NULLblk_mq_get_dispatch_budget() => 0 // because Thread A marked // "device_busy" in scsi_device
blk_mq_put_dispatch_budget()
What if there is only thread A? You need to mention that thread B is from BFQ.
Quoted 37 lines
The above case was observed in reboot tests and caused a task to hang forever waiting for IO to complete. Traces showed that in fact two tasks were running
blk_mq_do_dispatch_sched()at the same time with the same "hctx". The task that got the budget did in fact seedispatch_request()return NULL. Both tasks returned and the system went on for several minutes (until the hung task delay kicked in) without the given "hctx" showing up again in traces.Let's attempt to fix this problem by detecting if there was contention for the budget in the case where we put the budget without dispatching anything. If we saw contention we kick all hctx's associated with the queue where there was contention. We do this without any locking by adding a double-check for budget and accepting a small amount of faux contention if the 2nd check gives us budget but then we don't dispatch anything (we'll look like we contended with ourselves).
A few extra notes:
- This whole thing is only a problem due to the inexact nature of
has_work(). Specifically ifhas_work()always guaranteed that a "true" return meant thatdispatch_request()would return non-NULL then we wouldn't have this problem. That's because we only grab the budget ifhas_work()returned true. If we had the non-NULL guarantee then at least one of the threads would actually dispatch work and when the work was done then queues would be kicked normally.- One specific I/O scheduler that trips this problem quite a bit is BFQ which definitely returns "true" for
has_work()in cases where it wouldn't dispatch. Making BFQ'shas_work()more exact requires thathas_work()becomes a much heavier function, including figuring out how to acquire spinlocks inhas_work()without tripping circular lock dependencies. This is prototyped but it's unclear if it's really the way to go when the problem can be solved with a relatively lightweight contention detection mechanism.
Quoted 3 lines
- Because this problem only trips with inexact
has_work()it's believed thatblk_mq_do_dispatch_ctx()does not need a similar change.
Right, I prefer to fix it in BFQ, given it isn't a generic issue, not worth of generic solution.
Quoted 26 lines
Signed-off-by Douglas Anderson <[email protected]>---
Changes in v2:
- Replace ("scsi: core: Fix stall...") w/ ("blk-mq: Rerun dispatch...")block/blk-mq-sched.c | 26 ++++++++++++++++++++++++-- block/blk-mq.c | 3 +++ include/linux/blkdev.h | 2 ++ 3 files changed, 29 insertions(+), 2 deletions(-)block/blk-mq-sched.c +4
-2@@ -97,12 +97,34 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)if (e->type->ops.has_work && !e->type->ops.has_work (hctx))break ; -if (!blk_mq_get_dispatch_budget (hctx)) -break ; + +if (!blk_mq_get_dispatch_budget (hctx)) { +/* +* We didn't get budget so set contention. If
block/blk-mq-sched.c static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) · line 113
Quoted 10 lines
block/blk-mq-sched.c +10
-0+* someone else had the budget but didn't dispatch +* they'll kick everything. NOTE: we check one +* extra time _after_ setting contention to fully +* close the race. If we don't actually dispatch +* we'll detext faux contention (with ourselves) +* but that should be rare. +*/ +atomic_set (&q->budget_contention,1 ); + +if (!blk_mq_get_dispatch_budget (hctx))
scsi_mq_get_budget() implies a smp_mb(), so the barrier can order between blk_mq_get_dispatch_budget() and atomic_set(&q->budget_contention, 0|1).
block/blk-mq-sched.c static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) · line 126
Quoted 13 lines
block/blk-mq-sched.c +9
-0+break ; + } rq = e->type->ops.dispatch_request (hctx);if (!rq) {blk_mq_put_dispatch_budget (hctx); + +/* +* We've released the budget but us holding it might +* have prevented someone else from dispatching. +* Detect that case and run all queues again. +*/ +if (atomic_read (&q->budget_contention))
scsi_mq_put_budget() doesn't imply smp_mb(), so one smp_mb__after_atomic() is needed between the above two op.
Quoted 9 lines
block/blk-mq-sched.c +1
-0+blk_mq_run_hw_queues (q, true);break ; }block/blk-mq.c +0
-0
block/blk-mq.c void blk_mq_run_hw_queues(struct request_queue *q, bool async) · line 1533
Quoted 6 lines
block/blk-mq.c +3
-0@@ -1528,6 +1528,9 @@ void blk_mq_run_hw_queues(struct request_queue *q, bool async)struct blk_mq_hw_ctx *hctx;int i; +/* We're running the queues, so clear the contention detector */ +atomic_set (&q->budget_contention,0 ); +
You add extra cost in fast path.
include/linux/blkdev.h struct request_queue { · line 587
Quoted 13 lines
block/blk-mq.c +0
-0queue_for_each_hw_ctx (q, hctx, i) {if (blk_mq_hctx_stopped (hctx))continue ;include/linux/blkdev.h +2
-0@@ -583,6 +583,8 @@ struct request_queue {#define BLK_MAX_WRITE_HINTS 5u64 write_hints[BLK_MAX_WRITE_HINTS ]; + +atomic_t budget_contention;
It needn't to be a atomic variable, and simple 'unsigned' int should be fine, what matters is that the order between R/W this flag and get/put budget.
thanks,
Ming
Re: [PATCH v2 2/2] blk-mq: Rerun dispatching in the case of budget contention
Hi,
Quoted 20 lines · Ming Lei
Quoted 18 lines · Douglas Anderson
It is possible for two threads to be running
blk_mq_do_dispatch_sched()at the same time with the same "hctx". This is because there can be more than one caller to__blk_mq_run_hw_queue()with the same "hctx" andhctx_lock()doesn't prevent more than one thread from entering.If more than one thread is running
blk_mq_do_dispatch_sched()at the same time with the same "hctx", they may have contention acquiring budget. Theblk_mq_get_dispatch_budget()can eventually translate intoscsi_mq_get_budget(). If the device's "queue_depth" is 1 (not uncommon) then only one of the two threads will be the one to increment "device_busy" to 1 and get the budget.The losing thread will break out of
blk_mq_do_dispatch_sched()and will stop dispatching requests. The assumption is that when more budget is available later (when existing transactions finish) the queue will be kicked again, perhaps inscsi_end_request().
Quoted 7 lines
Quoted 3 lines
The winning thread now has budget and can go on to call
dispatch_request(). Ifdispatch_request()returns NULL here then we have a potential problem. Specifically we'll now callAs I mentioned before, it is a BFQ specific issue, it tells blk-mq that it has work to do, and now the budget is assigned to the only winning thread, however,
dispatch_request()still returns NULL.
Correct that it only happens with BFQ, but whether it's a BFQ bug or not just depends on how you define the has_work() API. If has_work() is allowed to be in-exact then it's either a blk-mq bug or a SCSI bug depending on how you cut it. If has_work() must be exact then it is certainly a BFQ bug. If has_work() doesn't need to be exact then it's not a BFQ bug. I believe that a sane API could be defined either way. Either has_work() can be defined as a lightweight hint to trigger heavier code or it can be defined as something exact. It's really up to blk-mq to say how they define it.
From his response on the SCSI patch [1], it sounded like Jens was OK with has_work() being a lightweight hint as long as BFQ ensures that the queues run later. ...but, as my investigation found, I believe that BFQ _does_ try to ensure that the queue is run at a later time by calling blk_mq_run_hw_queues(). The irony is that due to the race we're talking about here, blk_mq_run_hw_queues() isn't guaranteed to be reliable if has_work() is inexact. :( One way to address this is to make blk_mq_run_hw_queues() reliable even if has_work() is inexact.
...so Jens: care to clarify how you'd like has_work() to be defined?
Quoted 6 lines
Quoted 3 lines
- Because this problem only trips with inexact
has_work()it's believed thatblk_mq_do_dispatch_ctx()does not need a similar change.Right, I prefer to fix it in BFQ, given it isn't a generic issue, not worth of generic solution.
Just to confirm: it sounds like you are saying that BFQ is not a first class citizen here because not everyone uses BFQ. Is that really the best policy? Couldn't I say that SCSI shouldn't be a first class citizen because not everyone uses SCSI? In such a case I could argue that we should speed up the blk-mq layer by removing all the budget code since SCSI is the only user. I'm not actually suggesting it, only pointing out that sometimes we need to make allowances in the code.
Quoted 15 lines
Quoted 15 lines
block/blk-mq-sched.c +9
-2@@ -97,12 +97,34 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)if (e->type->ops.has_work && !e->type->ops.has_work (hctx))break ; -if (!blk_mq_get_dispatch_budget (hctx)) -break ; + +if (!blk_mq_get_dispatch_budget (hctx)) { +/* +* We didn't get budget so set contention. If +* someone else had the budget but didn't dispatch +* they'll kick everything. NOTE: we check one +* extra time _after_ setting contention to fully +* close the race. If we don't actually dispatch +* we'll detext faux contention (with ourselves)
Quoted 8 lines
block/blk-mq-sched.c static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) · line 113
Quoted 5 lines
block/blk-mq-sched.c +5
-0+* but that should be rare. +*/ +atomic_set (&q->budget_contention,1 ); + +if (!blk_mq_get_dispatch_budget (hctx))
scsi_mq_get_budget()implies asmp_mb(), so the barrier can order betweenblk_mq_get_dispatch_budget()andatomic_set(&q->budget_contention, 0|1).
I always struggle when working with memory barriers. I think you are saying that this section of code is OK, though, presumably because the SCSI code does "atomic_inc_return" which implies this barrier.
Do you happen to know if it's documented that anyone who implements get_budget() for "struct blk_mq_ops" will have a memory barrier here? I know SCSI is the only existing user, but we'd want to keep it generic.
Quoted 14 lines
block/blk-mq-sched.c static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) · line 126
Quoted 13 lines
block/blk-mq-sched.c +9
-0+break ; + } rq = e->type->ops.dispatch_request (hctx);if (!rq) {blk_mq_put_dispatch_budget (hctx); + +/* +* We've released the budget but us holding it might +* have prevented someone else from dispatching. +* Detect that case and run all queues again. +*/ +if (atomic_read (&q->budget_contention))
Quoted 2 lines
scsi_mq_put_budget()doesn't implysmp_mb(), so onesmp_mb__after_atomic()is needed between the above two op.
OK, thanks. I will add that we decide to move forward with this patch. Again I'd wonder if this type of thing should be documented.
Quoted 12 lines
Quoted 12 lines
block/blk-mq-sched.c +1
-0+blk_mq_run_hw_queues (q, true);break ; }block/blk-mq.c +0
-0@@ -1528,6 +1528,9 @@ void blk_mq_run_hw_queues(struct request_queue *q, bool async)struct blk_mq_hw_ctx *hctx;int i;
Quoted 5 lines
block/blk-mq.c void blk_mq_run_hw_queues(struct request_queue *q, bool async) · line 1533
Quoted 3 lines
block/blk-mq.c +3
-0+/* We're running the queues, so clear the contention detector */ +atomic_set (&q->budget_contention,0 ); +You add extra cost in fast path.
Yes, but it is a fairly minor cost added. It's called once in a place where we're unlikely to be looping and where we're about to do a lot heavier operations (perhaps in a loop). This doesn't feel like a dealbreaker.
Quoted 9 lines
Quoted 9 lines
block/blk-mq.c +0
-0queue_for_each_hw_ctx (q, hctx, i) {if (blk_mq_hctx_stopped (hctx))continue ;include/linux/blkdev.h +0
-0@@ -583,6 +583,8 @@ struct request_queue {
Quoted 8 lines
include/linux/blkdev.h struct request_queue { · line 587
Quoted 4 lines
include/linux/blkdev.h +2
-0#define BLK_MAX_WRITE_HINTS 5u64 write_hints[BLK_MAX_WRITE_HINTS ]; + +atomic_t budget_contention;It needn't to be a atomic variable, and simple 'unsigned' int should be fine, what matters is that the order between R/W this flag and get/put budget.
Apparently I'm the only one who feels atomic_t is more documenting here. I will cede to the will of the majority here and change to 'unsigned int' if we decide to move forward with this patch.
-Doug
Re: [PATCH v2 2/2] blk-mq: Rerun dispatching in the case of budget contention
Hi,
Quoted 20 lines · Doug Anderson
Correct that it only happens with BFQ, but whether it's a BFQ bug or not just depends on how you define the
has_work()API. Ifhas_work()is allowed to be in-exact then it's either a blk-mq bug or a SCSI bug depending on how you cut it. Ifhas_work()must be exact then it is certainly a BFQ bug. Ifhas_work()doesn't need to be exact then it's not a BFQ bug. I believe that a sane API could be defined either way. Eitherhas_work()can be defined as a lightweight hint to trigger heavier code or it can be defined as something exact. It's really up to blk-mq to say how they define it.From his response on the SCSI patch [1], it sounded like Jens was OK with
has_work()being a lightweight hint as long as BFQ ensures that the queues run later. ...but, as my investigation found, I believe that BFQ _does_ try to ensure that the queue is run at a later time by callingblk_mq_run_hw_queues(). The irony is that due to the race we're talking about here,blk_mq_run_hw_queues()isn't guaranteed to be reliable ifhas_work()is inexact. :( One way to address this is to makeblk_mq_run_hw_queues()reliable even ifhas_work()is inexact.
Quoted 1 line
...so Jens: care to clarify how you'd like
has_work()to be defined?
Sorry to reply so quickly after my prior reply, but I might have found an extreme corner case where we can still run into the same race even if has_work() is exact. This is all theoretical from code analysis. Maybe you can poke a hole in my scenario or tell me it's so implausible that we don't care, but it seems like it's theoretically possible. For this example I'll assume a budget of 1 (AKA only one thread can get budget for a given queue):
* Threads A and B both run has_work() at the same time with the same "hctx". has_work() is exact but there's no lock, so it's OK if Thread A and B both get back true.
* Thread B gets interrupted for a long time right after it decides that there is work. Maybe its CPU gets an interrupt and the interrupt handler is slow.
* Thread A runs, get budget, dispatches work.
* Thread A's work finishes and budget is released.
* Thread B finally runs again and gets budget.
* Since Thread A already took care of the work and no new work has come in, Thread B will get NULL from dispatch_request(). I believe this is specifically why dispatch_request() is allowed to return NULL in the first place if has_work() must be exact.
* Thread B will now be holding the budget and is about to call put_budget(), but hasn't called it yet.
* Thread B gets interrupted for a long time (again). Dang interrupts.
* Now Thread C (with a different "hctx" but the same queue) comes along and runs blk_mq_do_dispatch_sched().
* Thread C won't do anything because it can't get budget.
* Finally Thread B will run again and put the budget without kicking any queues.
Now we have a potential I/O stall because nobody will ever kick the queues.
I think the above example could happen even on non-BFQ systems and I think it would also be fixed by an approach like the one in my patch.
-Doug
Re: [PATCH v2 2/2] blk-mq: Rerun dispatching in the case of budget contention
Quoted 65 lines · Doug Anderson
Hi,
Quoted 20 lines · Doug Anderson
Correct that it only happens with BFQ, but whether it's a BFQ bug or not just depends on how you define the
has_work()API. Ifhas_work()is allowed to be in-exact then it's either a blk-mq bug or a SCSI bug depending on how you cut it. Ifhas_work()must be exact then it is certainly a BFQ bug. Ifhas_work()doesn't need to be exact then it's not a BFQ bug. I believe that a sane API could be defined either way. Eitherhas_work()can be defined as a lightweight hint to trigger heavier code or it can be defined as something exact. It's really up to blk-mq to say how they define it.From his response on the SCSI patch [1], it sounded like Jens was OK with
has_work()being a lightweight hint as long as BFQ ensures that the queues run later. ...but, as my investigation found, I believe that BFQ _does_ try to ensure that the queue is run at a later time by callingblk_mq_run_hw_queues(). The irony is that due to the race we're talking about here,blk_mq_run_hw_queues()isn't guaranteed to be reliable ifhas_work()is inexact. :( One way to address this is to makeblk_mq_run_hw_queues()reliable even ifhas_work()is inexact.Quoted 1 line
...so Jens: care to clarify how you'd like
has_work()to be defined?Sorry to reply so quickly after my prior reply, but I might have found an extreme corner case where we can still run into the same race even if
has_work()is exact. This is all theoretical from code analysis. Maybe you can poke a hole in my scenario or tell me it's so implausible that we don't care, but it seems like it's theoretically possible. For this example I'll assume a budget of 1 (AKA only one thread can get budget for a given queue):* Threads A and B both run
has_work()at the same time with the same "hctx".has_work()is exact but there's no lock, so it's OK if Thread A and B both get back true.* Thread B gets interrupted for a long time right after it decides that there is work. Maybe its CPU gets an interrupt and the interrupt handler is slow.
* Thread A runs, get budget, dispatches work.
* Thread A's work finishes and budget is released.
* Thread B finally runs again and gets budget.
* Since Thread A already took care of the work and no new work has come in, Thread B will get NULL from
dispatch_request(). I believe this is specifically whydispatch_request()is allowed to return NULL in the first place ifhas_work()must be exact.* Thread B will now be holding the budget and is about to call
put_budget(), but hasn't called it yet.* Thread B gets interrupted for a long time (again). Dang interrupts.
* Now Thread C (with a different "hctx" but the same queue) comes along and runs
blk_mq_do_dispatch_sched().* Thread C won't do anything because it can't get budget.
* Finally Thread B will run again and put the budget without kicking any queues.
Quoted 6 lines
Now we have a potential I/O stall because nobody will ever kick the queues.
I think the above example could happen even on non-BFQ systems and I think it would also be fixed by an approach like the one in my patch.
OK, looks it isn't specific on BFQ any more.
Follows another candidate approach for this issue, given it is so hard to trigger, we can make it more reliable by rerun queue when has_work() returns true after ops->dispath_request() returns NULL.
block/blk-mq-sched.c +4 -0
@@ -80,6 +80,7 @@ void blk_mq_sched_restart(struct blk_mq_hw_ctx *hctx)blk_mq_run_hw_queue (hctx, true); } +#define BLK_MQ_BUDGET_DELAY 3/* ms units */ /* * Only SCSI implements .get_budget and .put_budget, and SCSI restarts * its queue by itself in its completion handler, so we don't need to
@@ -103,6 +104,9 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) rq = e->type->ops.dispatch_request (hctx);if (!rq) {blk_mq_put_dispatch_budget (hctx); + +if (e->type->ops.has_work && e->type->ops.has_work (hctx)) +blk_mq_delay_run_hw_queue (hctx,BLK_MQ_BUDGET_DELAY );break ; }
Thanks,
Ming
Re: [PATCH v2 2/2] blk-mq: Rerun dispatching in the case of budget contention
Hi,
Quoted 20 lines · Ming Lei
OK, looks it isn't specific on BFQ any more.
Follows another candidate approach for this issue, given it is so hard to trigger, we can make it more reliable by rerun queue when
has_work()returns true afterops->dispath_request()returns NULL.block/blk-mq-sched.c +1
-0@@ -80,6 +80,7 @@ void blk_mq_sched_restart(struct blk_mq_hw_ctx *hctx)blk_mq_run_hw_queue (hctx, true); } +#define BLK_MQ_BUDGET_DELAY 3/* ms units */ /* * Only SCSI implements .get_budget and .put_budget, and SCSI restarts * its queue by itself in its completion handler, so we don't need to
block/blk-mq-sched.c static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) · line 109
Quoted 6 lines
block/blk-mq-sched.c +3
-0@@ -103,6 +104,9 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) rq = e->type->ops.dispatch_request (hctx);if (!rq) {blk_mq_put_dispatch_budget (hctx); + +if (e->type->ops.has_work && e->type->ops.has_work (hctx)) +blk_mq_delay_run_hw_queue (hctx,BLK_MQ_BUDGET_DELAY );
I agree that your patch should solve the race. With the current BFQ's has_work() it's a bit of a disaster though. It will essentially put blk-mq into a busy-wait loop (with a 3 ms delay between each poll) while BFQ's has_work() says "true" but BFQ doesn't dispatch anything.
...so I guess the question that still needs to be answered: does has_work() need to be exact? If so then we need the patch you propose plus one to BFQ. If not, we should continue along the lines of my patch.
-Doug
Re: [PATCH v2 2/2] blk-mq: Rerun dispatching in the case of budget contention
Quoted 39 lines
Il giorno 5 apr 2020, alle ore 16:00, Doug Anderson <[email protected]> ha scritto:
Hi,
Quoted 20 lines · Ming Lei
OK, looks it isn't specific on BFQ any more.
Follows another candidate approach for this issue, given it is so hard to trigger, we can make it more reliable by rerun queue when
has_work()returns true afterops->dispath_request()returns NULL.block/blk-mq-sched.c +1
-0@@ -80,6 +80,7 @@ void blk_mq_sched_restart(struct blk_mq_hw_ctx *hctx)blk_mq_run_hw_queue (hctx, true); } +#define BLK_MQ_BUDGET_DELAY 3/* ms units */ /* * Only SCSI implements .get_budget and .put_budget, and SCSI restarts * its queue by itself in its completion handler, so we don't need to block/blk-mq-sched.c static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) · line 109
Quoted 6 lines
block/blk-mq-sched.c +3
-0@@ -103,6 +104,9 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) rq = e->type->ops.dispatch_request (hctx);if (!rq) {blk_mq_put_dispatch_budget (hctx); + +if (e->type->ops.has_work && e->type->ops.has_work (hctx)) +blk_mq_delay_run_hw_queue (hctx,BLK_MQ_BUDGET_DELAY );I agree that your patch should solve the race. With the current BFQ's
has_work()it's a bit of a disaster though. It will essentially put blk-mq into a busy-wait loop (with a 3 ms delay between each poll) while BFQ'shas_work()says "true" but BFQ doesn't dispatch anything.
Quoted 5 lines
...so I guess the question that still needs to be answered: does
has_work()need to be exact? If so then we need the patch you propose plus one to BFQ. If not, we should continue along the lines of my patch.
Some more comments. BFQ's I/O plugging lasts 9 ms by default. So, with this last Ming's patch, BFQ may happen to be polled every 3ms, for at most three times.
On the opposite end, making bfq_has_work plugging aware costs more complexity, and possibly one more lock. While avoiding the above occasional polling, this may imply a lot of overhead or CPU stalls on every dispatch.
Paolo
Quoted 1 line
-Doug
Re: [PATCH v2 2/2] blk-mq: Rerun dispatching in the case of budget contention
Hi,
Quoted 44 lines · Paolo Valente
Quoted 38 lines
Il giorno 5 apr 2020, alle ore 16:00, Doug Anderson <[email protected]> ha scritto:
Hi,
Quoted 20 lines · Ming Lei
OK, looks it isn't specific on BFQ any more.
Follows another candidate approach for this issue, given it is so hard to trigger, we can make it more reliable by rerun queue when
has_work()returns true afterops->dispath_request()returns NULL.block/blk-mq-sched.c +1
-0@@ -80,6 +80,7 @@ void blk_mq_sched_restart(struct blk_mq_hw_ctx *hctx)blk_mq_run_hw_queue (hctx, true); } +#define BLK_MQ_BUDGET_DELAY 3/* ms units */ /* * Only SCSI implements .get_budget and .put_budget, and SCSI restarts * its queue by itself in its completion handler, so we don't need to block/blk-mq-sched.c static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) · line 109
Quoted 6 lines
block/blk-mq-sched.c +3
-0@@ -103,6 +104,9 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) rq = e->type->ops.dispatch_request (hctx);if (!rq) {blk_mq_put_dispatch_budget (hctx); + +if (e->type->ops.has_work && e->type->ops.has_work (hctx)) +blk_mq_delay_run_hw_queue (hctx,BLK_MQ_BUDGET_DELAY );I agree that your patch should solve the race. With the current BFQ's
has_work()it's a bit of a disaster though. It will essentially put blk-mq into a busy-wait loop (with a 3 ms delay between each poll) while BFQ'shas_work()says "true" but BFQ doesn't dispatch anything.Quoted 5 lines
...so I guess the question that still needs to be answered: does
has_work()need to be exact? If so then we need the patch you propose plus one to BFQ. If not, we should continue along the lines of my patch.
Quoted 3 lines
Some more comments. BFQ's I/O plugging lasts 9 ms by default. So, with this last Ming's patch, BFQ may happen to be polled every 3ms, for at most three times.
Ah! I did not know this. OK, then Ming's patch seems like it should work. If nothing else it should fix the problem. If this ends up making BFQ chew up too much CPU time then presumably someone will notice and BFQ's has_work() can be improved.
Ming: how do you want to proceed? Do you want to formally post the patch? Do you want me to post a v3 of my series where I place patch #2 with your patch? Do you want authorship (which implies adding your Signed-off-by)?
Quoted 4 lines
On the opposite end, making
bfq_has_workplugging aware costs more complexity, and possibly one more lock. While avoiding the above occasional polling, this may imply a lot of overhead or CPU stalls on every dispatch.
I still think it would be interesting to run performance tests with my proof-of-concept solution for has_work(). Even if it's not ideal, knowing whether performance increased, decreased, or stayed the same would give information about how much more effort should be put into this.
-Doug
Re: [PATCH v2 2/2] blk-mq: Rerun dispatching in the case of budget contention
Quoted 71 lines
Il giorno 5 apr 2020, alle ore 18:16, Doug Anderson <[email protected]> ha scritto:
Hi,
Quoted 44 lines · Paolo Valente
Quoted 38 lines
Il giorno 5 apr 2020, alle ore 16:00, Doug Anderson <[email protected]> ha scritto:
Hi,
Quoted 20 lines · Ming Lei
OK, looks it isn't specific on BFQ any more.
Follows another candidate approach for this issue, given it is so hard to trigger, we can make it more reliable by rerun queue when
has_work()returns true afterops->dispath_request()returns NULL.block/blk-mq-sched.c +1
-0@@ -80,6 +80,7 @@ void blk_mq_sched_restart(struct blk_mq_hw_ctx *hctx)blk_mq_run_hw_queue (hctx, true); } +#define BLK_MQ_BUDGET_DELAY 3/* ms units */ /* * Only SCSI implements .get_budget and .put_budget, and SCSI restarts * its queue by itself in its completion handler, so we don't need to block/blk-mq-sched.c static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) · line 109
Quoted 6 lines
block/blk-mq-sched.c +3
-0@@ -103,6 +104,9 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) rq = e->type->ops.dispatch_request (hctx);if (!rq) {blk_mq_put_dispatch_budget (hctx); + +if (e->type->ops.has_work && e->type->ops.has_work (hctx)) +blk_mq_delay_run_hw_queue (hctx,BLK_MQ_BUDGET_DELAY );I agree that your patch should solve the race. With the current BFQ's
has_work()it's a bit of a disaster though. It will essentially put blk-mq into a busy-wait loop (with a 3 ms delay between each poll) while BFQ'shas_work()says "true" but BFQ doesn't dispatch anything.Quoted 5 lines
...so I guess the question that still needs to be answered: does
has_work()need to be exact? If so then we need the patch you propose plus one to BFQ. If not, we should continue along the lines of my patch.Quoted 3 lines
Some more comments. BFQ's I/O plugging lasts 9 ms by default. So, with this last Ming's patch, BFQ may happen to be polled every 3ms, for at most three times.
Ah! I did not know this. OK, then Ming's patch seems like it should work. If nothing else it should fix the problem. If this ends up making BFQ chew up too much CPU time then presumably someone will notice and BFQ's
has_work()can be improved.Ming: how do you want to proceed? Do you want to formally post the patch? Do you want me to post a v3 of my series where I place patch #2 with your patch? Do you want authorship (which implies adding your Signed-off-by)?
Quoted 4 lines
On the opposite end, making
bfq_has_workplugging aware costs more complexity, and possibly one more lock. While avoiding the above occasional polling, this may imply a lot of overhead or CPU stalls on every dispatch.
Quoted 6 lines
I still think it would be interesting to run performance tests with my proof-of-concept solution for
has_work(). Even if it's not ideal, knowing whether performance increased, decreased, or stayed the same would give information about how much more effort should be put into this.
Why not? It is however hard to hope that we add only negligible overhead and CPU stalls if we move from one lock-protected section per I/O-request dispatch, to two or more lock-protected sections per request (has_work may be invoked several times per request).
At any rate, if useful, one of the scripts in my S benchmark suite can also measure max IOPS (when limited only by I/O processing) [1]. The script is for Linux distros; I don't know whether it works in your environments of interest, Doug.
Paolo
[1] https://github.com/Algodev-github/S/blob/master/throughput-sync/throughput-sync.sh
Quoted 1 line
-Doug
Re: [PATCH v2 2/2] blk-mq: Rerun dispatching in the case of budget contention
In reply to Ming LeiHi,
block/blk-mq-sched.c static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) · line 109
Quoted 8 lines · Ming Lei
block/blk-mq-sched.c +3
-0@@ -103,6 +104,9 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) rq = e->type->ops.dispatch_request (hctx);if (!rq) {blk_mq_put_dispatch_budget (hctx); + +if (e->type->ops.has_work && e->type->ops.has_work (hctx)) +blk_mq_delay_run_hw_queue (hctx,BLK_MQ_BUDGET_DELAY );
To really close the race, don't we need to run all the queues associated with the hctx? I haven't traced it through, but I've been assuming that the multiple "hctx"s associated with the same queue will have the same budget associated with them and thus they can block each other out.
-Doug
Re: [PATCH v2 2/2] blk-mq: Rerun dispatching in the case of budget contention
Quoted 4 lines · Doug Anderson
Hi,
Quoted 1 line · Ming Lei
Quoted 13 lines
block/blk-mq-sched.c static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) · line 109
Quoted 7 lines
block/blk-mq-sched.c +3
-0@@ -103,6 +104,9 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) rq = e->type->ops.dispatch_request (hctx);if (!rq) {blk_mq_put_dispatch_budget (hctx); + +if (e->type->ops.has_work && e->type->ops.has_work (hctx)) +blk_mq_delay_run_hw_queue (hctx,BLK_MQ_BUDGET_DELAY );To really close the race, don't we need to run all the queues associated with the hctx? I haven't traced it through, but I've been assuming that the multiple "hctx"s associated with the same queue will have the same budget associated with them and thus they can block each other out.
Yeah, we should run all hctxs which share the same budget space.
Also, in theory, we don't have to add the delay, however BFQ may plug the dispatch for 9 ms, so looks delay run queue is required.
thanks,
Ming
Re: [PATCH v2 2/2] blk-mq: Rerun dispatching in the case of budget contention
Hi,
Quoted 20 lines · Ming Lei
Quoted 4 lines · Doug Anderson
Hi,
Quoted 1 line · Ming Lei
Quoted 13 lines
block/blk-mq-sched.c static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) · line 109
Quoted 7 lines
block/blk-mq-sched.c +3
-0@@ -103,6 +104,9 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) rq = e->type->ops.dispatch_request (hctx);if (!rq) {blk_mq_put_dispatch_budget (hctx); + +if (e->type->ops.has_work && e->type->ops.has_work (hctx)) +blk_mq_delay_run_hw_queue (hctx,BLK_MQ_BUDGET_DELAY );To really close the race, don't we need to run all the queues associated with the hctx? I haven't traced it through, but I've been assuming that the multiple "hctx"s associated with the same queue will have the same budget associated with them and thus they can block each other out.
Quoted 4 lines
Yeah, we should run all hctxs which share the same budget space.
Also, in theory, we don't have to add the delay, however BFQ may plug the dispatch for 9 ms, so looks delay run queue is required.
I have posted up v3. A few notes:
* Since we should run all "hctxs" I took out the check for has_work() before kicking the queue.
* As far as I can tell the theoretical race happens for _anyone_ who puts budget, so I added it to blk_mq_put_dispatch_budget(). Feel free to shout at me in response to v3 if you believe I'm wrong about that.
Thanks for all your reviews and suggestions so far!
-Doug