ping...
On 2021/9/16 22:10, qiulaibin wrote:
> When multiple hctx share one tagset. The wake_batch is calculated during initialization by queue_depth. But when multiple hctx share one tagset. The queue depth assigned to each user may be smaller than wakup_batch. This may cause the waiting queue to fail to wakup and leads to Hang.
>
> Fix this by recalculating wake_batch when inc or dec active_queues.
>
> Fixes: 0d2602ca30e41 ("blk-mq: improve support for shared tags maps")
> Signed-off-by: Laibin Qiu <[email protected]>
> ---
> block/blk-mq-tag.c | 44 +++++++++++++++++++++++++++++++++++++++--
> include/linux/sbitmap.h | 8 ++++++++
> lib/sbitmap.c | 3 ++-
> 3 files changed, 52 insertions(+), 3 deletions(-)
>
> diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c index 86f87346232a..d02f5ac0004c 100644
> --- a/block/blk-mq-tag.c
> +++ b/block/blk-mq-tag.c
> @@ -16,6 +16,27 @@
> #include "blk-mq-sched.h"
> #include "blk-mq-tag.h"
>
> +static void bt_update_wake_batch(struct sbitmap_queue *bt, unsigned int
> +users) {
> + unsigned int depth;
> +
> + depth = max((bt->sb.depth + users - 1) / users, 4U);
> + sbitmap_queue_update_wake_batch(bt, depth); }
> +
> +/*
> + * Recalculate wakeup batch when tag is shared by hctx.
> + */
> +static void blk_mq_update_wake_batch(struct sbitmap_queue *bitmap_tags,
> + struct sbitmap_queue *breserved_tags, unsigned int users) {
> + if (!users)
> + return;
> +
> + bt_update_wake_batch(bitmap_tags, users);
> + bt_update_wake_batch(breserved_tags, users); }
> +
> /*
> * If a previously inactive queue goes active, bump the active user count.
> * We need to do this before try to allocate driver tag, then even if fail @@ -24,17 +45,29 @@
> */
> bool __blk_mq_tag_busy(struct blk_mq_hw_ctx *hctx) {
> + unsigned int users;
> +
> if (blk_mq_is_sbitmap_shared(hctx->flags)) {
> struct request_queue *q = hctx->queue;
> struct blk_mq_tag_set *set = q->tag_set;
>
> if (!test_bit(QUEUE_FLAG_HCTX_ACTIVE, &q->queue_flags) &&
> - !test_and_set_bit(QUEUE_FLAG_HCTX_ACTIVE, &q->queue_flags))
> + !test_and_set_bit(QUEUE_FLAG_HCTX_ACTIVE, &q->queue_flags)) {
> atomic_inc(&set->active_queues_shared_sbitmap);
> +
> + users = atomic_read(&set->active_queues_shared_sbitmap);
> + blk_mq_update_wake_batch(&set->__bitmap_tags,
> + &set->__breserved_tags, users);
> + }
> } else {
> if (!test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state) &&
> - !test_and_set_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
> + !test_and_set_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state)) {
> atomic_inc(&hctx->tags->active_queues);
> +
> + users = atomic_read(&hctx->tags->active_queues);
> + blk_mq_update_wake_batch(&hctx->tags->__bitmap_tags,
> + &hctx->tags->__breserved_tags, users);
> + }
> }
>
> return true;
> @@ -59,16 +92,23 @@ void __blk_mq_tag_idle(struct blk_mq_hw_ctx *hctx)
> struct blk_mq_tags *tags = hctx->tags;
> struct request_queue *q = hctx->queue;
> struct blk_mq_tag_set *set = q->tag_set;
> + unsigned int users;
>
> if (blk_mq_is_sbitmap_shared(hctx->flags)) {
> if (!test_and_clear_bit(QUEUE_FLAG_HCTX_ACTIVE,
> &q->queue_flags))
> return;
> atomic_dec(&set->active_queues_shared_sbitmap);
> + users = atomic_read(&set->active_queues_shared_sbitmap);
> + blk_mq_update_wake_batch(&set->__bitmap_tags,
> + &set->__breserved_tags, users);
> } else {
> if (!test_and_clear_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
> return;
> atomic_dec(&tags->active_queues);
> + users = atomic_read(&tags->active_queues);
> + blk_mq_update_wake_batch(&tags->__bitmap_tags,
> + &tags->__breserved_tags, users);
> }
>
> blk_mq_tag_wakeup_all(tags, false);
> diff --git a/include/linux/sbitmap.h b/include/linux/sbitmap.h index 2713e689ad66..d49e4f054bfe 100644
> --- a/include/linux/sbitmap.h
> +++ b/include/linux/sbitmap.h
> @@ -406,6 +406,14 @@ static inline void sbitmap_queue_free(struct sbitmap_queue *sbq)
> sbitmap_free(&sbq->sb);
> }
>
> +/**
> + * sbitmap_queue_update_wake_batch() - Recalucate wake batch.
> + * @sbq: Bitmap queue.
> + * @depth: New number of queue depth.
> + */
> +void sbitmap_queue_update_wake_batch(struct sbitmap_queue *sbq,
> + unsigned int depth);
> +
> /**
> * sbitmap_queue_resize() - Resize a &struct sbitmap_queue.
> * @sbq: Bitmap queue to resize.
> diff --git a/lib/sbitmap.c b/lib/sbitmap.c index b25db9be938a..bbe1d663763f 100644
> --- a/lib/sbitmap.c
> +++ b/lib/sbitmap.c
> @@ -457,7 +457,7 @@ int sbitmap_queue_init_node(struct sbitmap_queue *sbq, unsigned int depth, } EXPORT_SYMBOL_GPL(sbitmap_queue_init_node);
>
> -static void sbitmap_queue_update_wake_batch(struct sbitmap_queue *sbq,
> +void sbitmap_queue_update_wake_batch(struct sbitmap_queue *sbq,
> unsigned int depth)
> {
> unsigned int wake_batch = sbq_calc_wake_batch(sbq, depth); @@ -475,6 +475,7 @@ static void sbitmap_queue_update_wake_batch(struct sbitmap_queue *sbq,
> atomic_set(&sbq->ws[i].wait_cnt, 1);
> }
> }
> +EXPORT_SYMBOL_GPL(sbitmap_queue_update_wake_batch);
>
> void sbitmap_queue_resize(struct sbitmap_queue *sbq, unsigned int depth) {
> --
> 2.22.0
>
> .
Hi Ming,
On 2021/09/26 20:48, Ming Lei wrote:
> Hi Laibin,
>
> On Mon, Sep 13, 2021 at 04:12:48PM +0800, Laibin Qiu wrote:
>> When multiple hctx share one tagset. The wake_batch is calculated
>> during initialization by queue_depth. But when multiple hctx share one
>> tagset. The queue depth assigned to each user may be smaller than
>> wakeup_batch. This may cause the waiting queue to fail to wakup and
>> leads to Hang.
> In case of shared tags, there might be more than one hctx which
allocates tag from single tags, and each hctx is limited to allocate at
most:
>
> hctx_max_depth = max((bt->sb.depth + users - 1) / users, 4U);
>
> and
>
> users = atomic_read(&hctx->tags->active_queues)
>
> See hctx_may_queue().
>
> tag idle detection is lazy, and may be delayed for 30sec, so there
could be just one real active hctx(queue) but all others are actually
idle and still accounted as active because of the lazy idle detection.
Then if wake_batch is > hctx_max_depth, driver tag allocation may wait
forever on this real active hctx.
>
> Correct me if my understanding is wrong.
Your understanding is right. When we add lots of users for one shared
tag, it will make wake_batch > hctx_max_dept. So driver tag allocation
may wait forever on this real active hctx.
>> Fix this by recalculating wake_batch when inc or dec active_queues.
>>
>> Fixes: 0d2602ca30e41 ("blk-mq: improve support for shared tags maps")
>> Signed-off-by: Laibin Qiu <[email protected]>
>> ---
>> block/blk-mq-tag.c | 44 +++++++++++++++++++++++++++++++++++++++--
>> include/linux/sbitmap.h | 8 ++++++++
>> lib/sbitmap.c | 3 ++-
>> 3 files changed, 52 insertions(+), 3 deletions(-)
>>
>> diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c index
>> 86f87346232a..d02f5ac0004c 100644
>> --- a/block/blk-mq-tag.c
>> +++ b/block/blk-mq-tag.c
>> @@ -16,6 +16,27 @@
>> #include "blk-mq-sched.h"
>> #include "blk-mq-tag.h"
>> +static void bt_update_wake_batch(struct sbitmap_queue *bt, unsigned
>> +int users) {
>> + unsigned int depth;
>> +
>> + depth = max((bt->sb.depth + users - 1) / users, 4U);
>> + sbitmap_queue_update_wake_batch(bt, depth);
> Use the hctx's max queue depth could reduce wake_batch a lot, then
performance may be degraded.
>
> Just wondering why not set sbq->wake_batch as hctx_max_depth if
> sbq->wake_batch is < hctx_max_depth?
>
__blk_mq_tag_busy() will add Users and __blk_mq_tag_idle() will decrease
Users. Only changes in Users will affect each user's max depth. So we
recalculate the matching wake_batch by
sbitmap_queue_update_wake_batch(). sbitmap_queue_update_wake_batch()
will calculate wake_batch by incoming depth. The value of
sbq->wake_batch will only be changed when the calculated wake_batch changes.
static void sbitmap_queue_update_wake_batch(struct sbitmap_queue *sbq,
unsigned int depth)
{
unsigned int wake_batch = sbq_calc_wake_batch(sbq, depth);
^^^^^^^^^^^^^^^^^
int i;
if (sbq->wake_batch != wake_batch) {
^^^^^^^^^^^^^^^^^^
WRITE_ONCE(sbq->wake_batch, wake_batch);
/*
* Pairs with the memory barrier in sbitmap_queue_wake_up()
* to ensure that the batch size is updated before the wait
* counts.
*/
smp_mb();
for (i = 0; i < SBQ_WAIT_QUEUES; i++)
atomic_set(&sbq->ws[i].wait_cnt, 1);
}
}
>
> Thanks,
> Ming
Thanks
Laibin