| Message ID | 20210125121340.70459-6-jefflexu@linux.alibaba.com (mailing list archive) |
|---|---|
| State | Changes Requested, archived |
| Delegated to: | Mike Snitzer |
| Headers | show |
| Series | dm: support IO polling for bio-based dm device | expand |
On Mon, Jan 25 2021 at 7:13am -0500, Jeffle Xu <jefflexu@linux.alibaba.com> wrote: > Introduce QUEUE_FLAG_POLL_CAP flag representing if the request queue > capable of polling or not. > > Signed-off-by: Jeffle Xu <jefflexu@linux.alibaba.com> Why are you adding QUEUE_FLAG_POLL_CAP? Doesn't seem as though DM or anything else actually needs it. Mike -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel
On 1/28/21 1:13 AM, Mike Snitzer wrote: > On Mon, Jan 25 2021 at 7:13am -0500, > Jeffle Xu <jefflexu@linux.alibaba.com> wrote: > >> Introduce QUEUE_FLAG_POLL_CAP flag representing if the request queue >> capable of polling or not. >> >> Signed-off-by: Jeffle Xu <jefflexu@linux.alibaba.com> > > Why are you adding QUEUE_FLAG_POLL_CAP? Doesn't seem as though DM or > anything else actually needs it. Users can switch on/off polling on device via '/sys/block/<dev>/queue/io_poll' at runtime. The requisite for turning on polling is that the device is **capable** of polling. For mq devices, the requisite is that there's polling hw queue for the device, i.e., ``` q->tag_set->nr_maps > HCTX_TYPE_POLL && q->tag_set->map[HCTX_TYPE_POLL].nr_queues ``` But for dm devices, we need to check if all the underlying devices support polling or not. Without this newly added queue flag, we need to check again every time users want to turn on polling via 'io_poll', and thus the dm layer need to export one interface to block layer, checking if all the underlying target devices support polling or not, maybe just like the iopoll() method we did in patch 3. Something like, ``` struct block_device_operations { + bool (*support_iopoll)(struct request_queue *q); ``` The newly added queue flag 'QUEUE_FLAG_POLL_CAP' is just used as a cache representing if the device **capable** of polling, while the original queue flag 'QUEUE_FLAG_POLL' representing if polling is turned on for this device **currently**. But indeed we are short of queue flag resource. Adding a new queue flag may not be the best resolution. Any inspiration?
diff --git a/block/blk-mq.c b/block/blk-mq.c index d058b9cbdf76..10f06337d8dc 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -3203,7 +3203,7 @@ struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set, q->queue_flags |= QUEUE_FLAG_MQ_DEFAULT; if (set->nr_maps > HCTX_TYPE_POLL && set->map[HCTX_TYPE_POLL].nr_queues) - blk_queue_flag_set(QUEUE_FLAG_POLL, q); + q->queue_flags |= QUEUE_FLAG_POLL_MASK; q->sg_reserved_size = INT_MAX; diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index b513f1683af0..65693efb7c76 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -420,8 +420,7 @@ static ssize_t queue_poll_store(struct request_queue *q, const char *page, unsigned long poll_on; ssize_t ret; - if (!q->tag_set || q->tag_set->nr_maps <= HCTX_TYPE_POLL || - !q->tag_set->map[HCTX_TYPE_POLL].nr_queues) + if (!blk_queue_poll_cap(q)) return -EINVAL; ret = queue_var_store(&poll_on, page, count); diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index bc540df197cb..095b486de02f 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -621,11 +621,16 @@ struct request_queue { #define QUEUE_FLAG_RQ_ALLOC_TIME 27 /* record rq->alloc_time_ns */ #define QUEUE_FLAG_HCTX_ACTIVE 28 /* at least one blk-mq hctx is active */ #define QUEUE_FLAG_NOWAIT 29 /* device supports NOWAIT */ +#define QUEUE_FLAG_POLL_CAP 30 /* capable of IO polling */ #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ (1 << QUEUE_FLAG_SAME_COMP) | \ (1 << QUEUE_FLAG_NOWAIT)) +#define QUEUE_FLAG_POLL_MASK ((1 << QUEUE_FLAG_POLL) | \ + (1<< QUEUE_FLAG_POLL_CAP)) + + void blk_queue_flag_set(unsigned int flag, struct request_queue *q); void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); @@ -667,6 +672,7 @@ bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags) #define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags) #define blk_queue_nowait(q) test_bit(QUEUE_FLAG_NOWAIT, &(q)->queue_flags) +#define blk_queue_poll_cap(q) test_bit(QUEUE_FLAG_POLL_CAP, &(q)->queue_flags) extern void blk_set_pm_only(struct request_queue *q); extern void blk_clear_pm_only(struct request_queue *q);
Introduce QUEUE_FLAG_POLL_CAP flag representing if the request queue capable of polling or not. Signed-off-by: Jeffle Xu <jefflexu@linux.alibaba.com> --- block/blk-mq.c | 2 +- block/blk-sysfs.c | 3 +-- include/linux/blkdev.h | 6 ++++++ 3 files changed, 8 insertions(+), 3 deletions(-)