blk-mq: improvement CPU hotplug
Subsystems (3)
v6 cover blk-mq: improvement CPU hotplug
Hi,
Thomas mentioned:
"
That was the constraint of managed interrupts from the very beginning:
The driver/subsystem has to quiesce the interrupt line and the associated queue _before_ it gets shutdown in CPU unplug and not fiddle with it until it's restarted by the core when the CPU is plugged in again.
"
But no drivers or blk-mq do that before one hctx becomes inactive(all CPUs for one hctx are offline), and even it is worse, blk-mq stills tries to run hw queue after hctx is dead, see blk_mq_hctx_notify_dead().
This patchset tries to address the issue by two stages:
1) add one new cpuhp state of CPUHP_AP_BLK_MQ_ONLINE
- mark the hctx as internal stopped, and drain all in-flight requests if the hctx is going to be dead.
2) re-submit IO in the state of CPUHP_BLK_MQ_DEAD after the hctx becomes dead
- steal bios from the request, and resubmit them via generic_make_request(), then these IO will be mapped to other live hctx for dispatch
Please comment & review, thanks!
https://github.com/ming1/linux/commits/v5.6-blk-mq-improve-cpu-hotplug
V6:
- simplify getting driver tag, so that we can drain in-flight requests correctly without using synchronize_rcu()
- handle re-submission of flush & passthrough request correctly
V5:
- rename BLK_MQ_S_INTERNAL_STOPPED as BLK_MQ_S_INACTIVE
- re-factor code for re-submit requests in cpu dead hotplug handler
- address requeue corner case
V4:
- resubmit IOs in dispatch list in case that this hctx is dead
V3:
- re-organize patch 2 & 3 a bit for addressing Hannes's comment
- fix patch 4 for avoiding potential deadlock, as found by Hannes
V2:
- patch4 & patch 5 in V1 have been merged to block tree, so remove them
- address comments from John Garry and Minwoo
Ming Lei (8):
blk-mq: assign rq->tag in blk_mq_get_driver_tag
blk-mq: add new state of BLK_MQ_S_INACTIVE
blk-mq: prepare for draining IO when hctx's all CPUs are offline
blk-mq: stop to handle IO and drain IO before hctx becomes inactive
block: add blk_end_flush_machinery
blk-mq: re-submit IO in case that hctx is inactive
blk-mq: handle requests dispatched from IO scheduler in case of
inactive hctx
block: deactivate hctx when the hctx is actually inactive
block/blk-flush.c | 143 +++++++++++++--- block/blk-mq-debugfs.c | 2 + block/blk-mq-tag.c | 2 +- block/blk-mq-tag.h | 2 + block/blk-mq.c | 331 ++++++++++++++++++++++++++++++------- block/blk-mq.h | 24 +-- block/blk.h | 9 +- drivers/block/loop.c | 2 +- drivers/md/dm-rq.c | 2 +- include/linux/blk-mq.h | 6 + include/linux/cpuhotplug.h | 1 + 11 files changed, 420 insertions(+), 104 deletions(-)
Signature · 1 line
2.25.2
PATCH v6 1/8 blk-mq: assign rq->tag in blk_mq_get_driver_tag
Especially for none elevator, rq->tag is assigned after the request is allocated, so there isn't any way to figure out if one request is in being dispatched. Also the code path wrt. driver tag becomes a bit difference between none and io scheduler.
When one hctx becomes inactive, we have to prevent any request from being dispatched to LLD. And get driver tag provides one perfect chance to do that. Meantime we can drain any such requests by checking if rq->tag is assigned.
So only assign rq->tag until blk_mq_get_driver_tag() is called.
This way also simplifies code of dealing with driver tag a lot.
--- block/blk-flush.c | 18 +++--------------- block/blk-mq.c | 41 +++++++++++++++++++++-------------------- block/blk-mq.h | 21 ++++++++++----------- block/blk.h | 5 ----- 4 files changed, 34 insertions(+), 51 deletions(-)
block/blk-flush.c +3 -15
@@ -239,13 +239,8 @@ static void flush_end_io(struct request *flush_rq, blk_status_t error) error = fq->rq_status; hctx = flush_rq->mq_hctx; -if (!q->elevator) { -blk_mq_tag_set_rq (hctx, flush_rq->tag, fq->orig_rq); - flush_rq->tag =-1 ; - }else { -blk_mq_put_driver_tag (flush_rq); - flush_rq->internal_tag =-1 ; - } + flush_rq->internal_tag =-1 ; +blk_mq_put_driver_tag (flush_rq); running = &fq->flush_queue[fq->flush_running_idx];BUG_ON (fq->flush_pending_idx == fq->flush_running_idx);
@@ -320,14 +315,7 @@ static void blk_kick_flush(struct request_queue *q, struct blk_flush_queue *fq, flush_rq->mq_ctx = first_rq->mq_ctx; flush_rq->mq_hctx = first_rq->mq_hctx; -if (!q->elevator) { - fq->orig_rq = first_rq; - flush_rq->tag = first_rq->tag; -blk_mq_tag_set_rq (flush_rq->mq_hctx, first_rq->tag, flush_rq); - }else { - flush_rq->internal_tag = first_rq->internal_tag; - } - + flush_rq->internal_tag = first_rq->internal_tag; flush_rq->cmd_flags =REQ_OP_FLUSH |REQ_PREFLUSH ; flush_rq->cmd_flags |= (flags &REQ_DRV ) | (flags &REQ_FAILFAST_MASK ); flush_rq->rq_flags |=RQF_FLUSH_SEQ ;
block/blk-mq.c +21 -20
74 lines not on this page · raw message
block/blk-mq.h +10 -11
@@ -197,26 +197,25 @@ static inline bool blk_mq_get_dispatch_budget(struct blk_mq_hw_ctx *hctx)return true; } -static inline void __blk_mq_put_driver_tag (struct blk_mq_hw_ctx *hctx, -struct request *rq) +static inline void blk_mq_put_driver_tag (struct request *rq) { -blk_mq_put_tag (hctx->tags, rq->mq_ctx, rq->tag); +struct blk_mq_hw_ctx *hctx = rq->mq_hctx; +int tag = rq->tag; + +if (tag <0 ) +return ; + rq->tag =-1 ; +if (rq->q->elevator) +blk_mq_put_tag (hctx->tags, rq->mq_ctx, tag); +if (rq->rq_flags &RQF_MQ_INFLIGHT ) { rq->rq_flags &= ~RQF_MQ_INFLIGHT ;atomic_dec (&hctx->nr_active); } } -static inline void blk_mq_put_driver_tag (struct request *rq) -{ -if (rq->tag ==-1 || rq->internal_tag ==-1 ) -return ; - -__blk_mq_put_driver_tag (rq->mq_hctx, rq); -} -static inline void blk_mq_clear_mq_map (struct blk_mq_queue_map *qmap) {int cpu;
block/blk.h +0 -5
@@ -25,11 +25,6 @@ struct blk_flush_queue {struct list_head flush_data_in_flight;struct request *flush_rq; -/* -* flush_rq shares tag with this rq, both can't be active -* at the same time -*/ -struct request *orig_rq;struct lock_class_key key;spinlock_t mq_flush_lock; };
Signature · 1 line
2.25.2
Re: [PATCH V6 1/8] blk-mq: assign rq->tag in blk_mq_get_driver_tag
Quoted 7 lines · Ming Lei
block/blk-mq.c +2
-1@@ -472,14 +462,18 @@ static void __blk_mq_free_request(struct request *rq)struct request_queue *q = rq->q;struct blk_mq_ctx *ctx = rq->mq_ctx;struct blk_mq_hw_ctx *hctx = rq->mq_hctx; -const int sched_tag = rq->internal_tag; +const int tag = rq->internal_tag; +bool has_sched = !!hctx->sched_tags;
block/blk-mq.c static void __blk_mq_free_request(struct request *rq) · line 476
Quoted 10 lines
block/blk-mq.c +6
-0blk_pm_mark_last_busy (rq); rq->mq_hctx =NULL ; +if (!has_sched) +blk_mq_put_tag (hctx->tags, ctx, tag); +else if (rq->tag >=0 )blk_mq_put_tag (hctx->tags, ctx, rq->tag); + +if (has_sched) +blk_mq_put_tag (hctx->sched_tags, ctx, tag);
This looks weird to me. Why not simply:
if (hctx->sched_tags) {
if (rq->tag >= 0)
blk_mq_put_tag(hctx->tags, ctx, rq->tag);
blk_mq_put_tag(hctx->sched_tags, ctx, rq->internal_tag);
} else {
blk_mq_put_tag(hctx->tags, ctx, rq->internal_tag);
}Review bool blk_mq_get_driver_tag(struct request *rq)
Quoted 1 line
(quoted diff) +0
-0@@ -1037,14 +1031,21 @@ bool blk_mq_get_driver_tag(struct request *rq)
FYI, it seems like blk_mq_get_driver_tag can be marked static.
Otherwise this looks pretty sensible to me.
Re: [PATCH V6 1/8] blk-mq: assign rq->tag in blk_mq_get_driver_tag
Quoted 21 lines · Christoph Hellwig
Quoted 7 lines · Ming Lei
block/blk-mq.c +2
-1@@ -472,14 +462,18 @@ static void __blk_mq_free_request(struct request *rq)struct request_queue *q = rq->q;struct blk_mq_ctx *ctx = rq->mq_ctx;struct blk_mq_hw_ctx *hctx = rq->mq_hctx; -const int sched_tag = rq->internal_tag; +const int tag = rq->internal_tag; +bool has_sched = !!hctx->sched_tags;block/blk-mq.c static void __blk_mq_free_request(struct request *rq) · line 476
Quoted 10 lines
block/blk-mq.c +6
-0blk_pm_mark_last_busy (rq); rq->mq_hctx =NULL ; +if (!has_sched) +blk_mq_put_tag (hctx->tags, ctx, tag); +else if (rq->tag >=0 )blk_mq_put_tag (hctx->tags, ctx, rq->tag); + +if (has_sched) +blk_mq_put_tag (hctx->sched_tags, ctx, tag);This looks weird to me. Why not simply:
Quoted 7 lines
if (hctx->sched_tags) {
if (rq->tag >= 0)
blk_mq_put_tag(hctx->tags, ctx, rq->tag);
blk_mq_put_tag(hctx->sched_tags, ctx, rq->internal_tag);
} else {
blk_mq_put_tag(hctx->tags, ctx, rq->internal_tag);
}Nice!
Quoted 7 lines
Review bool blk_mq_get_driver_tag(struct request *rq)
Quoted 3 lines
(quoted diff) +0
-0@@ -1037,14 +1031,21 @@ bool blk_mq_get_driver_tag(struct request *rq)FYI, it seems like
blk_mq_get_driver_tagcan be marked static.Otherwise this looks pretty sensible to me.
Indeed, just forgot to do that.
Thanks,
Ming
PATCH v6 2/8 blk-mq: add new state of BLK_MQ_S_INACTIVE
In reply to Ming LeiAdd a new hw queue state of BLK_MQ_S_INACTIVE, which is set when all CPUs of this hctx are offline.
Prepares for stopping hw queue when all CPUs of this hctx become offline.
--- block/blk-mq-debugfs.c | 1 + block/blk-mq.h | 3 ++- include/linux/blk-mq.h | 3 +++ 3 files changed, 6 insertions(+), 1 deletion(-)
block/blk-mq-debugfs.c +1 -0
@@ -213,6 +213,7 @@ static const char *const hctx_state_name[] = {HCTX_STATE_NAME (STOPPED ),HCTX_STATE_NAME (TAG_ACTIVE ),HCTX_STATE_NAME (SCHED_RESTART ), + HCTX_STATE_NAME(INACTIVE), };#undef HCTX_STATE_NAME
block/blk-mq.h +2 -1
@@ -168,7 +168,8 @@ static inline struct blk_mq_tags *blk_mq_tags_from_data(struct blk_mq_alloc_datastatic inline bool blk_mq_hctx_stopped (struct blk_mq_hw_ctx *hctx) { -return test_bit (BLK_MQ_S_STOPPED , &hctx->state); +return test_bit (BLK_MQ_S_STOPPED , &hctx->state) || +test_bit (BLK_MQ_S_INACTIVE , &hctx->state); }static inline bool blk_mq_hw_queue_mapped (struct blk_mq_hw_ctx *hctx)
include/linux/blk-mq.h +3 -0
@@ -397,6 +397,9 @@ enum {BLK_MQ_S_TAG_ACTIVE =1 ,BLK_MQ_S_SCHED_RESTART =2 , +/* hw queue is inactive after all its CPUs become offline */ +BLK_MQ_S_INACTIVE =3 , +BLK_MQ_MAX_DEPTH =10240 , BLK_MQ_CPU_WORK_BATCH = 8,
Signature · 1 line
2.25.2
Re: [PATCH V6 2/8] blk-mq: add new state of BLK_MQ_S_INACTIVE
Quoted 4 lines · Ming Lei
Add a new hw queue state of
BLK_MQ_S_INACTIVE, which is set when all CPUs of this hctx are offline.Prepares for stopping hw queue when all CPUs of this hctx become offline.
It isn't set yet, is it? So maybe ".. will be set..". Or just merge it into the patch that sets it, which might be easier to follow.
PATCH v6 3/8 blk-mq: prepare for draining IO when hctx's all CPUs are offline
In reply to Ming LeiMost of blk-mq drivers depend on managed IRQ's auto-affinity to setup up queue mapping. Thomas mentioned the following point[1]:
"
That was the constraint of managed interrupts from the very beginning:
The driver/subsystem has to quiesce the interrupt line and the associated queue _before_ it gets shutdown in CPU unplug and not fiddle with it until it's restarted by the core when the CPU is plugged in again.
"
However, current blk-mq implementation doesn't quiesce hw queue before the last CPU in the hctx is shutdown. Even worse, CPUHP_BLK_MQ_DEAD is one cpuhp state handled after the CPU is down, so there isn't any chance to quiesce hctx for blk-mq wrt. CPU hotplug.
Add new cpuhp state of CPUHP_AP_BLK_MQ_ONLINE for blk-mq to stop queues and wait for completion of in-flight requests.
We will stop hw queue and wait for completion of in-flight requests when one hctx is becoming dead in the following patch. This way may cause dead-lock for some stacking blk-mq drivers, such as dm-rq and loop.
Add blk-mq flag of BLK_MQ_F_NO_MANAGED_IRQ and mark it for dm-rq and loop, so we needn't to wait for completion of in-flight requests from dm-rq & loop, then the potential dead-lock can be avoided.
[1] https://lore.kernel.org/linux-block/[email protected]/
--- block/blk-mq-debugfs.c | 1 + block/blk-mq.c | 13 +++++++++++++ drivers/block/loop.c | 2 +- drivers/md/dm-rq.c | 2 +- include/linux/blk-mq.h | 3 +++ include/linux/cpuhotplug.h | 1 + 6 files changed, 20 insertions(+), 2 deletions(-)
block/blk-mq-debugfs.c +1 -0
@@ -240,6 +240,7 @@ static const char *const hctx_flag_name[] = {HCTX_FLAG_NAME (TAG_SHARED ),HCTX_FLAG_NAME (BLOCKING ),HCTX_FLAG_NAME (NO_SCHED ), + HCTX_FLAG_NAME(NO_MANAGED_IRQ), };#undef HCTX_FLAG_NAME
block/blk-mq.c +13 -0
@@ -2249,6 +2249,11 @@ int blk_mq_alloc_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,return -ENOMEM ; } +static int blk_mq_hctx_notify_online (unsigned int cpu,struct hlist_node *node) +{ +return 0 ; +} +/* * 'cpu' is going away. splice any existing rq_list entries from this * software queue to the hw queue dispatch list, and ensure that it
@@ -2285,6 +2290,9 @@ static int blk_mq_hctx_notify_dead(unsigned int cpu, struct hlist_node *node)static void blk_mq_remove_cpuhp (struct blk_mq_hw_ctx *hctx) { +if (!(hctx->flags &BLK_MQ_F_NO_MANAGED_IRQ )) +cpuhp_state_remove_instance_nocalls (CPUHP_AP_BLK_MQ_ONLINE , + &hctx->cpuhp_online);cpuhp_state_remove_instance_nocalls (CPUHP_BLK_MQ_DEAD , &hctx->cpuhp_dead); }
@@ -2344,6 +2352,9 @@ static int blk_mq_init_hctx(struct request_queue *q, { hctx->queue_num = hctx_idx; +if (!(hctx->flags &BLK_MQ_F_NO_MANAGED_IRQ )) +cpuhp_state_add_instance_nocalls (CPUHP_AP_BLK_MQ_ONLINE , + &hctx->cpuhp_online);cpuhp_state_add_instance_nocalls (CPUHP_BLK_MQ_DEAD , &hctx->cpuhp_dead); hctx->tags = set->tags[hctx_idx];
@@ -3588,6 +3599,8 @@ static int __init blk_mq_init(void) {cpuhp_setup_state_multi (CPUHP_BLK_MQ_DEAD ,"block/mq:dead" ,NULL , blk_mq_hctx_notify_dead); +cpuhp_setup_state_multi (CPUHP_AP_BLK_MQ_ONLINE ,"block/mq:online" , +NULL , blk_mq_hctx_notify_online);return 0 ; }subsys_initcall (blk_mq_init);
drivers/block/loop.c +1 -1
@@ -2012,7 +2012,7 @@ static int loop_add(struct loop_device **l, int i) lo->tag_set.queue_depth =128 ; lo->tag_set.numa_node =NUMA_NO_NODE ; lo->tag_set.cmd_size =sizeof (struct loop_cmd ); - lo->tag_set.flags =BLK_MQ_F_SHOULD_MERGE ; + lo->tag_set.flags =BLK_MQ_F_SHOULD_MERGE |BLK_MQ_F_NO_MANAGED_IRQ ; lo->tag_set.driver_data = lo; err =blk_mq_alloc_tag_set (&lo->tag_set);
drivers/md/dm-rq.c +1 -1
@@ -547,7 +547,7 @@ int dm_mq_init_request_queue(struct mapped_device *md, struct dm_table *t) md->tag_set->ops = &dm_mq_ops; md->tag_set->queue_depth =dm_get_blk_mq_queue_depth (); md->tag_set->numa_node = md->numa_node_id; - md->tag_set->flags =BLK_MQ_F_SHOULD_MERGE ; + md->tag_set->flags =BLK_MQ_F_SHOULD_MERGE |BLK_MQ_F_NO_MANAGED_IRQ ; md->tag_set->nr_hw_queues =dm_get_blk_mq_nr_hw_queues (); md->tag_set->driver_data = md;
include/linux/blk-mq.h +3 -0
@@ -140,6 +140,8 @@ struct blk_mq_hw_ctx {*/ atomic_t nr_active; +/** @cpuhp_online: List to store request if CPU is going to die */ +struct hlist_node cpuhp_online;/** @cpuhp_dead: List to store request if some CPU die. */ struct hlist_node cpuhp_dead;/** @kobj: Kernel object for sysfs. */
@@ -388,6 +390,7 @@ struct blk_mq_ops {enum {BLK_MQ_F_SHOULD_MERGE =1 <<0 ,BLK_MQ_F_TAG_SHARED =1 <<1 , +BLK_MQ_F_NO_MANAGED_IRQ =1 <<2 ,BLK_MQ_F_BLOCKING =1 <<5 ,BLK_MQ_F_NO_SCHED =1 <<6 ,BLK_MQ_F_ALLOC_POLICY_START_BIT =8 ,
include/linux/cpuhotplug.h +1 -0
@@ -151,6 +151,7 @@ enum cpuhp_state {CPUHP_AP_SMPBOOT_THREADS ,CPUHP_AP_X86_VDSO_VMA_ONLINE ,CPUHP_AP_IRQ_AFFINITY_ONLINE , +CPUHP_AP_BLK_MQ_ONLINE ,CPUHP_AP_ARM_MVEBU_SYNC_CLOCKS ,CPUHP_AP_X86_INTEL_EPB_ONLINE , CPUHP_AP_PERF_ONLINE,
Signature · 1 line
2.25.2
PATCH v6 4/8 blk-mq: stop to handle IO and drain IO before hctx becomes inactive
In reply to Ming LeiBefore one CPU becomes offline, check if it is the last online CPU of hctx. If yes, mark this hctx as inactive, meantime wait for completion of all in-flight IOs originated from this hctx.
This way guarantees that there isn't any inflight IO before shutdowning the managed IRQ line.
--- block/blk-mq-tag.c | 2 +- block/blk-mq-tag.h | 2 ++ block/blk-mq.c | 67 ++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 70 insertions(+), 1 deletion(-)
block/blk-mq-tag.c +1 -1
@@ -317,7 +317,7 @@ static void bt_tags_for_each(struct blk_mq_tags *tags, struct sbitmap_queue *bt,* true to continue iterating tags, false to stop. * @priv: Will be passed as second argument to @fn. */ -static void blk_mq_all_tag_busy_iter (struct blk_mq_tags *tags, +void blk_mq_all_tag_busy_iter (struct blk_mq_tags *tags,busy_tag_iter_fn *fn,void *priv) {if (tags->nr_reserved_tags)
block/blk-mq-tag.h +2 -0
@@ -34,6 +34,8 @@ extern int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,extern void blk_mq_tag_wakeup_all (struct blk_mq_tags *tags,bool );void blk_mq_queue_tag_busy_iter (struct request_queue *q,busy_iter_fn *fn,void *priv); +void blk_mq_all_tag_busy_iter (struct blk_mq_tags *tags, +busy_tag_iter_fn *fn,void *priv);static inline struct sbq_wait_state *bt_wait_ptr (struct sbitmap_queue *bt,struct blk_mq_hw_ctx *hctx)
block/blk-mq.c +67 -0
87 lines not on this page · raw message
Signature · 1 line
2.25.2
PATCH v6 5/8 block: add blk_end_flush_machinery
In reply to Ming LeiFlush requests aren't same with normal FS request:
1) on dedicated per-hctx flush rq is pre-allocated for sending flush request
2) flush requests are issued to hardware via one machinary so that flush merge can be applied
We can't simply re-submit flush rqs via blk_steal_bios(), so add blk_end_flush_machinery to collect flush requests which needs to be resubmitted:
- if one flush command without DATA is enough, send one flush, complete this kind of requests
- otherwise, add the request into a list and let caller re-submit it.
--- block/blk-flush.c | 125 +++++++++++++++++++++++++++++++++++++++++++--- block/blk.h | 4 ++ 2 files changed, 122 insertions(+), 7 deletions(-)
block/blk-flush.c +118 -7
@@ -173,10 +173,11 @@ static void blk_flush_complete_seq(struct request *rq,unsigned int cmd_flags;BUG_ON (rq->flush.seq & seq); - rq->flush.seq |= seq; +if (!error) + rq->flush.seq |= seq; cmd_flags = rq->cmd_flags; -if (likely (!error)) +if (likely (!error && !fq->flush_queue_terminating)) seq =blk_flush_cur_seq (rq);else seq =REQ_FSEQ_DONE ;
@@ -203,9 +204,15 @@ static void blk_flush_complete_seq(struct request *rq,* normal completion and end it. */ BUG_ON (!list_empty (&rq->queuelist)); -list_del_init (&rq->flush.list); -blk_flush_restore_request (rq); -blk_mq_end_request (rq, error); + +/* Terminating code will end the request from flush queue */ +if (likely (!fq->flush_queue_terminating)) { +list_del_init (&rq->flush.list); +blk_flush_restore_request (rq); +blk_mq_end_request (rq, error); + }else { +list_move_tail (&rq->flush.list, pending); + }break ;default :
@@ -282,7 +289,8 @@ static void blk_kick_flush(struct request_queue *q, struct blk_flush_queue *fq,struct request *flush_rq = fq->flush_rq;/* C1 described at the top of this file */ -if (fq->flush_pending_idx != fq->flush_running_idx ||list_empty (pending)) +if (fq->flush_pending_idx != fq->flush_running_idx || +list_empty (pending) || fq->flush_queue_terminating)return ;/* C2 and C3
@@ -334,7 +342,7 @@ static void mq_flush_data_end_io(struct request *rq, blk_status_t error)struct blk_flush_queue *fq =blk_get_flush_queue (q, ctx);if (q->elevator) { -WARN_ON (rq->tag <0 ); +WARN_ON (rq->tag <0 && !fq->flush_queue_terminating);blk_mq_put_driver_tag (rq); }
@@ -515,3 +523,106 @@ void blk_free_flush_queue(struct blk_flush_queue *fq)kfree (fq->flush_rq);kfree (fq); } + +static void __blk_end_queued_flush (struct blk_flush_queue *fq, +unsigned int queue_idx,struct list_head *resubmit_list, +struct list_head *flush_list) +{ +struct list_head *queue = &fq->flush_queue[queue_idx]; +struct request *rq, *nxt; + +list_for_each_entry_safe (rq, nxt, queue, flush.list) { +unsigned int seq =blk_flush_cur_seq (rq); + +list_del_init (&rq->flush.list); +blk_flush_restore_request (rq); +if (!blk_rq_sectors (rq) || seq ==REQ_FSEQ_POSTFLUSH ) +list_add_tail (&rq->queuelist, flush_list); +else +list_add_tail (&rq->queuelist, resubmit_list); + } +} + +static void blk_end_queued_flush (struct blk_flush_queue *fq, +struct list_head *resubmit_list,struct list_head *flush_list) +{ +unsigned long flags; + +spin_lock_irqsave (&fq->mq_flush_lock, flags); +__blk_end_queued_flush (fq,0 , resubmit_list, flush_list); +__blk_end_queued_flush (fq,1 , resubmit_list, flush_list); +spin_unlock_irqrestore (&fq->mq_flush_lock, flags); +} + +/* complete requests which just requires one flush command */ +static void blk_complete_flush_requests (struct blk_flush_queue *fq, +struct list_head *flush_list) +{ +struct block_device *bdev; +struct request *rq; +int error = -ENXIO ; + +if (list_empty (flush_list)) +return ; + + rq = list_first_entry(flush_list, struct request, queuelist); + +/* Send flush via one active hctx so we can move on */ + bdev =bdget_disk (rq->rq_disk,0 ); +if (bdev) { + error =blkdev_issue_flush (bdev,GFP_KERNEL ,NULL ); +bdput (bdev); + } + +while (!list_empty (flush_list)) { + rq = list_first_entry(flush_list, struct request, queuelist); +list_del_init (&rq->queuelist); +blk_mq_end_request (rq, error); + } +} + +/* +* Called when this hctx is inactive and all CPUs of this hctx is dead. +* +* Terminate this hw queue's flush machinery, and try to complete flush +* IO requests if possible, such as any flush IO without data, or flush +* data IO in POSTFLUSH stage. Otherwise, add the flush IOs into @list +* and let caller to re-submit them. +* +* No need to hold the flush queue lock given no any flush activity can +* reach this queue now. +*/ +void blk_end_flush_machinery (struct blk_mq_hw_ctx *hctx, +struct list_head *in,struct list_head *out) +{ +LIST_HEAD (resubmit_list); +LIST_HEAD (flush_list); +struct blk_flush_queue *fq = hctx->fq; +struct request *rq, *nxt; +unsigned long flags; + +spin_lock_irqsave (&fq->mq_flush_lock, flags); + fq->flush_queue_terminating =1 ; +spin_unlock_irqrestore (&fq->mq_flush_lock, flags); + +/* End inflight flush requests */ +list_for_each_entry_safe (rq, nxt, in, queuelist) { +WARN_ON (!(rq->rq_flags &RQF_FLUSH_SEQ )); +list_del_init (&rq->queuelist); + rq->end_io (rq,BLK_STS_AGAIN ); + } + +/* End queued requests */ +blk_end_queued_flush (fq, &resubmit_list, &flush_list); + +/* Send flush and complete requests which just need one flush req */ +blk_complete_flush_requests (fq, &flush_list); + +spin_lock_irqsave (&fq->mq_flush_lock, flags); +/* reset flush queue so that it is ready to work next time */ + fq->flush_pending_idx = fq->flush_running_idx =0 ; + fq->flush_queue_terminating =0 ; +spin_unlock_irqrestore (&fq->mq_flush_lock, flags); + +list_splice_init (&resubmit_list, out); +}
block/blk.h +4 -0
@@ -19,6 +19,7 @@ struct blk_flush_queue { unsigned int flush_queue_delayed:1; unsigned int flush_pending_idx:1; unsigned int flush_running_idx:1; + unsigned int flush_queue_terminating:1;blk_status_t rq_status;unsigned long flush_pending_since;struct list_head flush_queue[2 ];
@@ -349,4 +350,7 @@ void blk_queue_free_zone_bitmaps(struct request_queue *q);static inline void blk_queue_free_zone_bitmaps (struct request_queue *q) {}#endif +void blk_end_flush_machinery (struct blk_mq_hw_ctx *hctx, +struct list_head *in,struct list_head *out); +#endif /* BLK_INTERNAL_H */
Signature · 1 line
2.25.2
PATCH v6 6/8 blk-mq: re-submit IO in case that hctx is inactive
In reply to Ming LeiWhen all CPUs in one hctx are offline and this hctx becomes inactive, we shouldn't run this hw queue for completing request any more.
So steal bios from the request, and resubmit them, and finally free the request in blk_mq_hctx_notify_dead().
--- block/blk-mq.c | 131 +++++++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 121 insertions(+), 10 deletions(-)
block/blk-mq.c +121 -10
@@ -2319,10 +2319,98 @@ static int blk_mq_hctx_notify_online(unsigned int cpu, struct hlist_node *node)return 0 ; } +static void blk_mq_resubmit_end_io (struct request *rq,blk_status_t error) +{ +struct request *orig_rq = rq->end_io_data; + +blk_mq_cleanup_rq (orig_rq); +blk_mq_end_request (orig_rq, error); + +blk_put_request (rq); +} + +static void blk_mq_resubmit_passthrough_io (struct request *rq) +{ +struct request *nrq; +unsigned int flags =0 , cmd_flags =0 ; +struct blk_mq_hw_ctx *hctx = rq->mq_hctx; +struct blk_mq_tags *tags = rq->q->elevator ? hctx->sched_tags : + hctx->tags; +bool reserved =blk_mq_tag_is_reserved (tags, rq->internal_tag); + +if (rq->rq_flags &RQF_PREEMPT ) + flags |=BLK_MQ_REQ_PREEMPT ; +if (reserved) + flags |=BLK_MQ_REQ_RESERVED ; + +/* avoid allocation failure & IO merge */ + cmd_flags = (rq->cmd_flags & ~REQ_NOWAIT ) |REQ_NOMERGE ; + + nrq =blk_get_request (rq->q, cmd_flags, flags); +if (!nrq) +return ; + + nrq->__sector =blk_rq_pos (rq); + nrq->__data_len =blk_rq_bytes (rq); +if (rq->rq_flags &RQF_SPECIAL_PAYLOAD ) { + nrq->rq_flags |=RQF_SPECIAL_PAYLOAD ; + nrq->special_vec = rq->special_vec; + } +#if defined(CONFIG_BLK_DEV_INTEGRITY ) + nrq->nr_integrity_segments = rq->nr_integrity_segments; +#endif + nrq->nr_phys_segments = rq->nr_phys_segments; + nrq->ioprio = rq->ioprio; + nrq->extra_len = rq->extra_len; + nrq->rq_disk = rq->rq_disk; + nrq->part = rq->part; + nrq->write_hint = rq->write_hint; + nrq->timeout = rq->timeout; + +memcpy (blk_mq_rq_to_pdu (nrq),blk_mq_rq_to_pdu (rq), + rq->q->tag_set->cmd_size); + + nrq->end_io = blk_mq_resubmit_end_io; + nrq->end_io_data = rq; + nrq->bio = rq->bio; + nrq->biotail = rq->biotail; + +blk_account_io_start (nrq, true); +blk_mq_sched_insert_request (nrq, true, true, true); +} + +static void blk_mq_resubmit_fs_io (struct request *rq) +{ +struct bio_list list; +struct bio *bio; + +bio_list_init (&list); +blk_steal_bios (&list, rq); + +while (true) { + bio =bio_list_pop (&list); +if (!bio) +break ; + +generic_make_request (bio); + } + +blk_mq_cleanup_rq (rq); +blk_mq_end_request (rq,0 ); +} + +static void blk_mq_resubmit_io (struct request *rq) +{ +if (rq->end_io ||blk_rq_is_passthrough (rq)) +blk_mq_resubmit_passthrough_io (rq); +else +blk_mq_resubmit_fs_io (rq); +} +/* -* 'cpu' is going away. splice any existing rq_list entries from this -* software queue to the hw queue dispatch list, and ensure that it -* gets run. +* 'cpu' has gone away. If this hctx is inactive, we can't dispatch request +* to the hctx any more, so steal bios from requests of this hctx, and +* re-submit them to the request queue, and free these requests finally. */ static int blk_mq_hctx_notify_dead (unsigned int cpu,struct hlist_node *node) {
@@ -2342,16 +2430,39 @@ static int blk_mq_hctx_notify_dead(unsigned int cpu, struct hlist_node *node) }spin_unlock (&ctx->lock); -clear_bit (BLK_MQ_S_INACTIVE , &hctx->state); +if (!test_bit (BLK_MQ_S_INACTIVE , &hctx->state)) { +if (!list_empty (&tmp)) { +spin_lock (&hctx->lock); +list_splice_tail_init (&tmp, &hctx->dispatch); +spin_unlock (&hctx->lock); +blk_mq_run_hw_queue (hctx, true); + } + }else { +LIST_HEAD (flush_in); +LIST_HEAD (flush_out); +struct request *rq, *nxt; -if (list_empty (&tmp)) -return 0 ; +/* requests in dispatch list have to be re-submitted too */ +spin_lock (&hctx->lock); +list_splice_tail_init (&hctx->dispatch, &tmp); +spin_unlock (&hctx->lock); -spin_lock (&hctx->lock); -list_splice_tail_init (&tmp, &hctx->dispatch); -spin_unlock (&hctx->lock); +/* blk_end_flush_machinery will cover flush request */ +list_for_each_entry_safe (rq, nxt, &tmp, queuelist) { +if (rq->rq_flags &RQF_FLUSH_SEQ ) +list_move (&rq->queuelist, &flush_in); + } +blk_end_flush_machinery (hctx, &flush_in, &flush_out); +list_splice_tail (&flush_out, &tmp); + +while (!list_empty (&tmp)) { + rq = list_first_entry(&tmp, struct request, queuelist); +list_del_init (&rq->queuelist); +blk_mq_resubmit_io (rq); + } +clear_bit (BLK_MQ_S_INACTIVE , &hctx->state); + } -blk_mq_run_hw_queue (hctx, true);return 0 ; }
Signature · 1 line
2.25.2
PATCH v6 7/8 blk-mq: handle requests dispatched from IO scheduler in case of inactive hctx
In reply to Ming LeiIf one hctx becomes inactive when its CPUs are all offline, all in-queue requests aimed at this hctx have to be re-submitted.
Re-submit requests from both sw queue or scheduler queue when the hctx is found as inactive.
--- block/blk-mq.c | 103 ++++++++++++++++++++++++++++++------------------- 1 file changed, 64 insertions(+), 39 deletions(-)
block/blk-mq.c +64 -39
@@ -2407,6 +2407,52 @@ static void blk_mq_resubmit_io(struct request *rq)blk_mq_resubmit_fs_io (rq); } +static void blk_mq_hctx_deactivate (struct blk_mq_hw_ctx *hctx) +{ +LIST_HEAD (sched_tmp); +LIST_HEAD (re_submit); +LIST_HEAD (flush_in); +LIST_HEAD (flush_out); +struct request *rq, *nxt; +struct elevator_queue *e = hctx->queue->elevator; + +if (!e) { +blk_mq_flush_busy_ctxs (hctx, &re_submit); + }else { +while ((rq = e->type->ops.dispatch_request (hctx))) { +if (rq->mq_hctx != hctx) +list_add (&rq->queuelist, &sched_tmp); +else +list_add (&rq->queuelist, &re_submit); + } + } +while (!list_empty (&sched_tmp)) { + rq = list_entry(sched_tmp.next, struct request, + queuelist); +list_del_init (&rq->queuelist); +blk_mq_sched_insert_request (rq, true, true, true); + } + +/* requests in dispatch list have to be re-submitted too */ +spin_lock (&hctx->lock); +list_splice_tail_init (&hctx->dispatch, &re_submit); +spin_unlock (&hctx->lock); + +/* blk_end_flush_machinery will cover flush request */ +list_for_each_entry_safe (rq, nxt, &re_submit, queuelist) { +if (rq->rq_flags &RQF_FLUSH_SEQ ) +list_move (&rq->queuelist, &flush_in); + } +blk_end_flush_machinery (hctx, &flush_in, &flush_out); +list_splice_tail (&flush_out, &re_submit); + +while (!list_empty (&re_submit)) { + rq = list_first_entry(&re_submit, struct request, queuelist); +list_del_init (&rq->queuelist); +blk_mq_resubmit_io (rq); + } +} +/* * 'cpu' has gone away. If this hctx is inactive, we can't dispatch request * to the hctx any more, so steal bios from requests of this hctx, and
@@ -2414,53 +2460,32 @@ static void blk_mq_resubmit_io(struct request *rq)*/ static int blk_mq_hctx_notify_dead (unsigned int cpu,struct hlist_node *node) { -struct blk_mq_hw_ctx *hctx; -struct blk_mq_ctx *ctx; -LIST_HEAD (tmp); -enum hctx_type type; - - hctx = hlist_entry_safe(node, struct blk_mq_hw_ctx, cpuhp_dead); - ctx =__blk_mq_get_ctx (hctx->queue, cpu); - type = hctx->type; +struct blk_mq_hw_ctx *hctx =hlist_entry_safe (node, + struct blk_mq_hw_ctx, cpuhp_dead); -spin_lock (&ctx->lock); -if (!list_empty (&ctx->rq_lists[type])) { -list_splice_init (&ctx->rq_lists[type], &tmp); -blk_mq_hctx_clear_pending (hctx, ctx); - } -spin_unlock (&ctx->lock); +if (test_bit (BLK_MQ_S_INACTIVE , &hctx->state)) { +WARN_ON_ONCE (blk_mq_tags_inflight_rqs (hctx) >0 ); +blk_mq_hctx_deactivate (hctx); +clear_bit (BLK_MQ_S_INACTIVE , &hctx->state); + }else if (!hctx->queue->elevator) { +struct blk_mq_ctx *ctx =__blk_mq_get_ctx (hctx->queue, cpu); +enum hctx_type type = hctx->type; +LIST_HEAD (tmp); + +spin_lock (&ctx->lock); +if (!list_empty (&ctx->rq_lists[type])) { +list_splice_init (&ctx->rq_lists[type], &tmp); +blk_mq_hctx_clear_pending (hctx, ctx); + } +spin_unlock (&ctx->lock); -if (!test_bit (BLK_MQ_S_INACTIVE , &hctx->state)) {if (!list_empty (&tmp)) {spin_lock (&hctx->lock);list_splice_tail_init (&tmp, &hctx->dispatch);spin_unlock (&hctx->lock); -blk_mq_run_hw_queue (hctx, true); - } - }else { -LIST_HEAD (flush_in); -LIST_HEAD (flush_out); -struct request *rq, *nxt; - -/* requests in dispatch list have to be re-submitted too */ -spin_lock (&hctx->lock); -list_splice_tail_init (&hctx->dispatch, &tmp); -spin_unlock (&hctx->lock); - -/* blk_end_flush_machinery will cover flush request */ -list_for_each_entry_safe (rq, nxt, &tmp, queuelist) { -if (rq->rq_flags &RQF_FLUSH_SEQ ) -list_move (&rq->queuelist, &flush_in); - } -blk_end_flush_machinery (hctx, &flush_in, &flush_out); -list_splice_tail (&flush_out, &tmp); -while (!list_empty (&tmp)) { - rq = list_first_entry(&tmp, struct request, queuelist); -list_del_init (&rq->queuelist); -blk_mq_resubmit_io (rq); +blk_mq_run_hw_queue (hctx, true); } -clear_bit (BLK_MQ_S_INACTIVE , &hctx->state); }return 0 ;
Signature · 1 line
2.25.2
PATCH v6 8/8 block: deactivate hctx when the hctx is actually inactive
In reply to Ming LeiRun queue on dead CPU still may be triggered in some corner case, such as one request is requeued after CPU hotplug is handled.
So handle this corner case during run queue.
--- block/blk-mq.c | 30 ++++++++++-------------------- 1 file changed, 10 insertions(+), 20 deletions(-)
block/blk-mq.c +10 -20
@@ -43,6 +43,8 @@static void blk_mq_poll_stats_start (struct request_queue *q);static void blk_mq_poll_stats_fn (struct blk_stat_callback *cb); +static void blk_mq_hctx_deactivate (struct blk_mq_hw_ctx *hctx); +static int blk_mq_poll_stats_bkt (const struct request *rq) {int ddir, sectors, bucket;
@@ -1352,28 +1354,16 @@ static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)int srcu_idx;/* -* We should be running this queue from one of the CPUs that -* are mapped to it. -* -* There are at least two related races now between setting -* hctx->next_cpu from blk_mq_hctx_next_cpu() and running -* __blk_mq_run_hw_queue(): -* -* - hctx->next_cpu is found offline in blk_mq_hctx_next_cpu(), -* but later it becomes online, then this warning is harmless -* at all -* -* - hctx->next_cpu is found online in blk_mq_hctx_next_cpu(), -* but later it becomes offline, then the warning can't be -* triggered, and we depend on blk-mq timeout handler to -* handle dispatched requests to this hctx +* BLK_MQ_S_INACTIVE may not deal with some requeue corner case: +* one request is requeued after cpu unplug is handled, so check +* if the hctx is actually inactive. If yes, deactive it and +* re-submit all requests in the queue. */ if (!cpumask_test_cpu (raw_smp_processor_id (), hctx->cpumask) && -cpu_online (hctx->next_cpu)) { -printk (KERN_WARNING "run queue from wrong CPU %d, hctx %s\n" , -raw_smp_processor_id (), -cpumask_empty (hctx->cpumask) ?"inactive" :"active" ); -dump_stack (); +cpumask_next_and (-1 , hctx->cpumask, cpu_online_mask) >= + nr_cpu_ids) { +blk_mq_hctx_deactivate (hctx); +return ; }/*
Signature · 1 line
2.25.2
Hi Ming,
Quoted 22 lines · Ming Lei
Hi,
Thomas mentioned:
"
That was the constraint of managed interrupts from the very beginning:The driver/subsystem has to quiesce the interrupt line and the associated queue _before_ it gets shutdown in CPU unplug and not fiddle with it until it's restarted by the core when the CPU is plugged in again.
"But no drivers or blk-mq do that before one hctx becomes inactive(all CPUs for one hctx are offline), and even it is worse, blk-mq stills tries to run hw queue after hctx is dead, see
blk_mq_hctx_notify_dead().This patchset tries to address the issue by two stages:
1) add one new cpuhp state of
CPUHP_AP_BLK_MQ_ONLINE- mark the hctx as internal stopped, and drain all in-flight requests if the hctx is going to be dead.
Quoted 6 lines
2) re-submit IO in the state of
CPUHP_BLK_MQ_DEADafter the hctx becomes dead- steal bios from the request, and resubmit them via
generic_make_request(), then these IO will be mapped to other live hctx for dispatchPlease comment & review, thanks!
FWIW, I've stress test this series by running the stress-cpu-hotplug with a fio workload in the background. Nothing exploded, all just worked fine.
Thanks,
Daniel
Quoted 27 lines · Daniel Wagner
Hi Ming,
Quoted 24 lines · Ming Lei
Hi,
Thomas mentioned:
"
That was the constraint of managed interrupts from the very beginning:The driver/subsystem has to quiesce the interrupt line and the associated queue _before_ it gets shutdown in CPU unplug and not fiddle with it until it's restarted by the core when the CPU is plugged in again.
"But no drivers or blk-mq do that before one hctx becomes inactive(all CPUs for one hctx are offline), and even it is worse, blk-mq stills tries to run hw queue after hctx is dead, see
blk_mq_hctx_notify_dead().This patchset tries to address the issue by two stages:
1) add one new cpuhp state of
CPUHP_AP_BLK_MQ_ONLINE- mark the hctx as internal stopped, and drain all in-flight requests if the hctx is going to be dead.
2) re-submit IO in the state of
CPUHP_BLK_MQ_DEADafter the hctx becomes dead
Quoted 8 lines
Quoted 4 lines
- steal bios from the request, and resubmit them via
generic_make_request(), then these IO will be mapped to other live hctx for dispatchPlease comment & review, thanks!
FWIW, I've stress test this series by running the stress-cpu-hotplug with a fio workload in the background. Nothing exploded, all just worked fine.
Hi Daniel,
Is stress-cpu-hotplug an ltp test? or from Steven Rostedt - I saw some threads where he mentioned some script?
Will the fio processes migrate back onto cpus which have been onlined again?
What is the block driver NVMe?
Thanks,
john
Hi John,
Quoted 2 lines · John Garry
Is stress-cpu-hotplug an ltp test? or from Steven Rostedt - I saw some threads where he mentioned some script?
My bad. It's the script from Steven, which toggle binary the cpus on/off:
[...]
2432 disabling cpu16 disabling cpu17 disabling cpu2
2433 disabling cpu1 enabling cpu16 enabling cpu17 enabling cpu2
2434 disabling cpu10 disabling cpu16 disabling cpu17 disabling cpu2 2435 enabling cpu1 enabling cpu10 enabling cpu16 enabling cpu17 enabling cpu2 2436 disabling cpu11 disabling cpu16 disabling cpu17 disabling cpu2 2437 disabling cpu1 enabling cpu11 enabling cpu16 enabling cpu17 enabling cpu2 2438 disabling cpu10 disabling cpu11 disabling cpu16 disabling cpu17 disabling cpu2 [..]
Quoted 1 line
Will the fio processes migrate back onto cpus which have been onlined again?
Hmm, good question. I've tried to assign them to a specific CPU via --cpus_allowed_policy=split and --cpus_allowed.
fio --rw=randwrite --name=test --size=50M --iodepth=32 --direct=1 \
--bs=4k --numjobs=40 --time_based --runtime=1h --ioengine=libaio \
--group_reporting --cpus_allowed_policy=split --cpus_allowed=0-40Though I haven't verified what happens when the CPU get's back online.
Quoted 1 line
What is the block driver NVMe?
I've used a qla2xxx device. Hannes asked my to retry it with a megasas device.
Thanks,
Daniel
Quoted 8 lines · Daniel Wagner
Hi John,
Quoted 2 lines · John Garry
Is stress-cpu-hotplug an ltp test? or from Steven Rostedt - I saw some threads where he mentioned some script?
My bad. It's the script from Steven, which toggle binary the cpus on/off:
Quoted 10 lines
[...]
2432 disabling cpu16 disabling cpu17 disabling cpu2
2433 disabling cpu1 enabling cpu16 enabling cpu17 enabling cpu2
2434 disabling cpu10 disabling cpu16 disabling cpu17 disabling cpu2 2435 enabling cpu1 enabling cpu10 enabling cpu16 enabling cpu17 enabling cpu2 2436 disabling cpu11 disabling cpu16 disabling cpu17 disabling cpu2 2437 disabling cpu1 enabling cpu11 enabling cpu16 enabling cpu17 enabling cpu2 2438 disabling cpu10 disabling cpu11 disabling cpu16 disabling cpu17 disabling cpu2 [..]
ok, but to really test this you need to ensure that all the cpus for a managed interrupt affinity mask are offlined together for some period of time greater than the IO timeout. Otherwise the hw queue's managed interrupt would not be shut down, and you're not verifying that the queues are fully drained.
Quoted 10 lines
Quoted 1 line
Will the fio processes migrate back onto cpus which have been onlined again?
Hmm, good question. I've tried to assign them to a specific CPU via --cpus_allowed_policy=split and --cpus_allowed.
fio --rw=randwrite --name=test --size=50M --iodepth=32 --direct=1 \ --bs=4k --numjobs=40 --time_based --runtime=1h --ioengine=libaio \ --group_reporting --cpus_allowed_policy=split --cpus_allowed=0-40Though I haven't verified what happens when the CPU get's back online.
Maybe this will work since you're offlining patterns of cpus and the fio processes have to migrate somewhere. But see above.
Quoted 6 lines
Quoted 2 lines
What is the block driver NVMe?
I've used a qla2xxx device. Hannes asked my to retry it with a megasas device.
Thanks,
John
Quoted 6 lines · John Garry
On 08/04/2020 14:10, Daniel Wagner wrote:
ok, but to really test this you need to ensure that all the cpus for a managed interrupt affinity mask are offlined together for some period of time greater than the IO timeout. Otherwise the hw queue's managed interrupt would not be shut down, and you're not verifying that the queues are fully drained.
Not sure if I understand you correctly: Are you saying that the IRQ related resources are not freed/moved from the offlining CPU?
Quoted 5 lines
Quoted 5 lines
Quoted 1 line
Will the fio processes migrate back onto cpus which have been onlined again?
Hmm, good question. I've tried to assign them to a specific CPU via --cpus_allowed_policy=split and --cpus_allowed.
Quoted 8 lines
Quoted 5 lines
fio --rw=randwrite --name=test --size=50M --iodepth=32 --direct=1 \ --bs=4k --numjobs=40 --time_based --runtime=1h --ioengine=libaio \ --group_reporting --cpus_allowed_policy=split --cpus_allowed=0-40Though I haven't verified what happens when the CPU get's back online.
Maybe this will work since you're offlining patterns of cpus and the fio processes have to migrate somewhere. But see above.
At least after the initial setup a fio thread will be migrated away from the offlining CPU.
A quick test shows, that the affinity mask for a fio thread will be cleared when the CPU goes offline. There seems to be a discussion going on about the cpu hotplug and the affinity mask:
https://lore.kernel.org/lkml/[email protected]
TL;DR: it can be scheduled back if affinity is tweaked via e.g. taskset, it won't if it's via cpusets
Thanks,
Daniel
Quoted 8 lines · Daniel Wagner
Quoted 6 lines · John Garry
On 08/04/2020 14:10, Daniel Wagner wrote:
ok, but to really test this you need to ensure that all the cpus for a managed interrupt affinity mask are offlined together for some period of time greater than the IO timeout. Otherwise the hw queue's managed interrupt would not be shut down, and you're not verifying that the queues are fully drained.
Hi Daniel,
Quoted 2 lines
Not sure if I understand you correctly: Are you saying that the IRQ related resources are not freed/moved from the offlining CPU?
This series tries to drain the hw queue when all cpus in the queue (IRQ) affinity mask are being offlined. This is because when all the cpus are offlined, the managed IRQ for that hw queue is shutdown - so there are no cpus remaining online to service the completion interrupt for in-flight IOs. The cover letter may explain this better.
Quoted 18 lines
Quoted 6 lines
Quoted 6 lines
Quoted 2 lines
Will the fio processes migrate back onto cpus which have been onlined again?
Hmm, good question. I've tried to assign them to a specific CPU via --cpus_allowed_policy=split and --cpus_allowed.
Quoted 8 lines
Quoted 5 lines
fio --rw=randwrite --name=test --size=50M --iodepth=32 --direct=1 \ --bs=4k --numjobs=40 --time_based --runtime=1h --ioengine=libaio \ --group_reporting --cpus_allowed_policy=split --cpus_allowed=0-40Though I haven't verified what happens when the CPU get's back online.
Maybe this will work since you're offlining patterns of cpus and the fio processes have to migrate somewhere. But see above.
At least after the initial setup a fio thread will be migrated away from the offlining CPU.
Quoted 8 lines
A quick test shows, that the affinity mask for a fio thread will be cleared when the CPU goes offline. There seems to be a discussion going on about the cpu hotplug and the affinity mask:
https://lore.kernel.org/lkml/[email protected]
TL;DR: it can be scheduled back if affinity is tweaked via e.g. taskset, it won't if it's via cpusets
I just avoid any of this in my test by looping in a sequence of onlining all cpus, start fio for short period, and then offline cpus.
BTW, you mentioned earlier that you would test megaraid_sas. As things stand, I don't think that series will help there as that driver still just exposes a single HW queue to blk-mq. I think qla2xxx driver does expose >1 queues, i.e. it sets Scsi_Host.nr_hq_queues, so may be a better option.
Cheers,
John
Hi Daniel,
Quoted 27 lines · Daniel Wagner
Hi Ming,
Quoted 24 lines · Ming Lei
Hi,
Thomas mentioned:
"
That was the constraint of managed interrupts from the very beginning:The driver/subsystem has to quiesce the interrupt line and the associated queue _before_ it gets shutdown in CPU unplug and not fiddle with it until it's restarted by the core when the CPU is plugged in again.
"But no drivers or blk-mq do that before one hctx becomes inactive(all CPUs for one hctx are offline), and even it is worse, blk-mq stills tries to run hw queue after hctx is dead, see
blk_mq_hctx_notify_dead().This patchset tries to address the issue by two stages:
1) add one new cpuhp state of
CPUHP_AP_BLK_MQ_ONLINE- mark the hctx as internal stopped, and drain all in-flight requests if the hctx is going to be dead.
2) re-submit IO in the state of
CPUHP_BLK_MQ_DEADafter the hctx becomes dead
Quoted 8 lines
Quoted 4 lines
- steal bios from the request, and resubmit them via
generic_make_request(), then these IO will be mapped to other live hctx for dispatchPlease comment & review, thanks!
FWIW, I've stress test this series by running the stress-cpu-hotplug with a fio workload in the background. Nothing exploded, all just worked fine.
Thanks for your test!
Especially this patch changes flush & passthrough IO handling during CPU hotplug, if possible, please include the two kinds of background IO when running cpu hotplug test.
BTW, I verified the patches by running 'dbench -s 64' & concurrent NVMe user IO during cpu hotplug, looks it works fine.
Also there is one known performance drop issue reported by John, which has been addressed in the following link:
https://github.com/ming1/linux/commit/1cfbe1b2f7fd7085bc86e09c6443a20e89142975
Thanks,
Ming