Re: [RFC PATCH 3/7] blk-mq: stop to handle IO before hctx's all CPUs become offline

[John Garry <john.garry@huawei.com>]

在 12/07/2019 10:47, Ming Lei 写道:
> Most 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.
> 
> [1] https://lore.kernel.org/linux-block/alpine.DEB.2.21.1904051331270.1802@nanos.tec.linutronix.de/
> 
> Cc: Bart Van Assche <bvanassche@acm.org>
> Cc: Hannes Reinecke <hare@suse.com>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Keith Busch <keith.busch@intel.com>
> Signed-off-by: Ming Lei <ming.lei@redhat.com>
> ---
>   block/blk-mq-tag.c         |  2 +-
>   block/blk-mq-tag.h         |  2 ++
>   block/blk-mq.c             | 67 ++++++++++++++++++++++++++++++++++++++
>   include/linux/blk-mq.h     |  1 +
>   include/linux/cpuhotplug.h |  1 +
>   5 files changed, 72 insertions(+), 1 deletion(-)
> 
> diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
> index da19f0bc8876..bcefb213ad69 100644
> --- a/block/blk-mq-tag.c
> +++ b/block/blk-mq-tag.c
> @@ -324,7 +324,7 @@ 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)
> diff --git a/block/blk-mq-tag.h b/block/blk-mq-tag.h
> index 61deab0b5a5a..321fd6f440e6 100644
> --- a/block/blk-mq-tag.h
> +++ b/block/blk-mq-tag.h
> @@ -35,6 +35,8 @@ 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)
> diff --git a/block/blk-mq.c b/block/blk-mq.c
> index e5ef40c603ca..028c5d78e409 100644
> --- a/block/blk-mq.c
> +++ b/block/blk-mq.c
> @@ -2205,6 +2205,64 @@ int blk_mq_alloc_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
>   	return -ENOMEM;
>   }
>   
> +static bool blk_mq_count_inflight_rq(struct request *rq, void *data,
> +				     bool reserved)
> +{
> +	unsigned *count = data;
> +
> +	if ((blk_mq_rq_state(rq) == MQ_RQ_IN_FLIGHT))
> +		(*count)++;
> +
> +	return true;
> +}
> +
> +unsigned blk_mq_tags_inflight_rqs(struct blk_mq_tags *tags)
> +{
> +	unsigned count = 0;
> +
> +	blk_mq_all_tag_busy_iter(tags, blk_mq_count_inflight_rq, &count);
> +
> +	return count;
> +}
> +
> +static void blk_mq_drain_inflight_rqs(struct blk_mq_hw_ctx *hctx)
> +{
> +	while (1) {
> +		if (!blk_mq_tags_inflight_rqs(hctx->tags))
> +			break;
> +		msleep(5);
> +	}
> +}
> +
> +static int blk_mq_hctx_notify_online(unsigned int cpu, struct hlist_node *node)
> +{
> +	struct blk_mq_hw_ctx *hctx = hlist_entry_safe(node,
> +			struct blk_mq_hw_ctx, cpuhp_online);
> +	unsigned prev_cpu = -1;
> +
> +	if (hctx->flags & BLK_MQ_F_NO_MANAGED_IRQ)
> +		return 0;

is it possible to make this check at registration, such that we don't 
even register for when (hctx->flags & BLK_MQ_F_NO_MANAGED_IRQ)?

> +
> +	while (true) {
> +		unsigned other_cpu = cpumask_next_and(prev_cpu, hctx->cpumask,

nit: to me name next_cpu seems better

> +				cpu_online_mask);
> +
> +		if (other_cpu >= nr_cpu_ids)
> +			break;
> +
> +		/* return if there is other online CPU on this hctx */
> +		if (other_cpu != cpu)
> +			return 0;
> +
> +		prev_cpu = other_cpu;
> +	}
> +
> +	set_bit(BLK_MQ_S_INTERNAL_STOPPED, &hctx->state);
> +	blk_mq_drain_inflight_rqs(hctx);
> +
> +	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
> @@ -2221,6 +2279,9 @@ static int blk_mq_hctx_notify_dead(unsigned int cpu, struct hlist_node *node)
>   	ctx = __blk_mq_get_ctx(hctx->queue, cpu);
>   	type = hctx->type;
>   
> +	if (test_bit(BLK_MQ_S_INTERNAL_STOPPED, &hctx->state))
> +		clear_bit(BLK_MQ_S_INTERNAL_STOPPED, &hctx->state);

can you just always clear the bit without the test?

> +
>   	spin_lock(&ctx->lock);
>   	if (!list_empty(&ctx->rq_lists[type])) {
>   		list_splice_init(&ctx->rq_lists[type], &tmp);
> @@ -2243,6 +2304,8 @@ static void blk_mq_remove_cpuhp(struct blk_mq_hw_ctx *hctx)
>   {
>   	cpuhp_state_remove_instance_nocalls(CPUHP_BLK_MQ_DEAD,
>   					    &hctx->cpuhp_dead);
> +	cpuhp_state_remove_instance_nocalls(CPUHP_AP_BLK_MQ_ONLINE,
> +					    &hctx->cpuhp_online);
>   }
>   
>   /* hctx->ctxs will be freed in queue's release handler */
> @@ -2301,6 +2364,8 @@ static int blk_mq_init_hctx(struct request_queue *q,
>   	hctx->queue_num = hctx_idx;
>   
>   	cpuhp_state_add_instance_nocalls(CPUHP_BLK_MQ_DEAD, &hctx->cpuhp_dead);
> +	cpuhp_state_add_instance_nocalls(CPUHP_AP_BLK_MQ_ONLINE,
> +			&hctx->cpuhp_online);
>   
>   	hctx->tags = set->tags[hctx_idx];
>   
> @@ -3536,6 +3601,8 @@ 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);
> diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
> index 911cdc6479dc..dc86bdac08f4 100644
> --- a/include/linux/blk-mq.h
> +++ b/include/linux/blk-mq.h
> @@ -59,6 +59,7 @@ struct blk_mq_hw_ctx {
>   	atomic_t		nr_active;
>   
>   	struct hlist_node	cpuhp_dead;
> +	struct hlist_node	cpuhp_online;

I'd reorder with cpuhp_dead

>   	struct kobject		kobj;
>   
>   	unsigned long		poll_considered;
> diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h
> index 87c211adf49e..5177f7bbcb88 100644
> --- a/include/linux/cpuhotplug.h
> +++ b/include/linux/cpuhotplug.h
> @@ -147,6 +147,7 @@ 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,
>