* [PATCH] io_uring: don't iterate ctx list to update sq_thread_idle
@ 2021-03-15 8:56 Hao Xu
2021-03-15 10:49 ` Pavel Begunkov
0 siblings, 1 reply; 3+ messages in thread
From: Hao Xu @ 2021-03-15 8:56 UTC (permalink / raw)
To: Jens Axboe; +Cc: io-uring, Pavel Begunkov, Joseph Qi
sqd->sq_thread_idle can be updated by a simple max(), rather than
iterating the whole ctx list to get the max one.
Signed-off-by: Hao Xu <[email protected]>
---
fs/io_uring.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/fs/io_uring.c b/fs/io_uring.c
index a4bce17af506..17697b9890e3 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -7871,7 +7871,7 @@ static int io_sq_offload_create(struct io_ring_ctx *ctx,
ret = -ENXIO;
} else {
list_add(&ctx->sqd_list, &sqd->ctx_list);
- io_sqd_update_thread_idle(sqd);
+ sqd->sq_thread_idle = max(sqd->sq_thread_idle, ctx->sq_thread_idle);
}
io_sq_thread_unpark(sqd);
--
1.8.3.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH] io_uring: don't iterate ctx list to update sq_thread_idle
2021-03-15 8:56 [PATCH] io_uring: don't iterate ctx list to update sq_thread_idle Hao Xu
@ 2021-03-15 10:49 ` Pavel Begunkov
2021-03-16 2:41 ` Hao Xu
0 siblings, 1 reply; 3+ messages in thread
From: Pavel Begunkov @ 2021-03-15 10:49 UTC (permalink / raw)
To: Hao Xu, Jens Axboe; +Cc: io-uring, Joseph Qi
On 15/03/2021 08:56, Hao Xu wrote:
> sqd->sq_thread_idle can be updated by a simple max(), rather than
> iterating the whole ctx list to get the max one.
I left it as a more fool proof option. Do you have a performance issue
with it? E.g. dozens rings sharing a single SQPOLL and constantly adding
new ones.
>
> Signed-off-by: Hao Xu <[email protected]>
> ---
> fs/io_uring.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index a4bce17af506..17697b9890e3 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -7871,7 +7871,7 @@ static int io_sq_offload_create(struct io_ring_ctx *ctx,
> ret = -ENXIO;
> } else {
> list_add(&ctx->sqd_list, &sqd->ctx_list);
> - io_sqd_update_thread_idle(sqd);
> + sqd->sq_thread_idle = max(sqd->sq_thread_idle, ctx->sq_thread_idle);
> }
> io_sq_thread_unpark(sqd);
>
>
--
Pavel Begunkov
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] io_uring: don't iterate ctx list to update sq_thread_idle
2021-03-15 10:49 ` Pavel Begunkov
@ 2021-03-16 2:41 ` Hao Xu
0 siblings, 0 replies; 3+ messages in thread
From: Hao Xu @ 2021-03-16 2:41 UTC (permalink / raw)
To: Pavel Begunkov, Jens Axboe; +Cc: io-uring, Joseph Qi
在 2021/3/15 下午6:49, Pavel Begunkov 写道:
> On 15/03/2021 08:56, Hao Xu wrote:
>> sqd->sq_thread_idle can be updated by a simple max(), rather than
>> iterating the whole ctx list to get the max one.
>
> I left it as a more fool proof option. Do you have a performance issue
> with it? E.g. dozens rings sharing a single SQPOLL and constantly adding
> new ones.
Currently no performance cases, just happen to see this and make
the code change.
>
>>
>> Signed-off-by: Hao Xu <[email protected]>
>> ---
>> fs/io_uring.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>> index a4bce17af506..17697b9890e3 100644
>> --- a/fs/io_uring.c
>> +++ b/fs/io_uring.c
>> @@ -7871,7 +7871,7 @@ static int io_sq_offload_create(struct io_ring_ctx *ctx,
>> ret = -ENXIO;
>> } else {
>> list_add(&ctx->sqd_list, &sqd->ctx_list);
>> - io_sqd_update_thread_idle(sqd);
>> + sqd->sq_thread_idle = max(sqd->sq_thread_idle, ctx->sq_thread_idle);
>> }
>> io_sq_thread_unpark(sqd);
>>
>>
>
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2021-03-16 2:42 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-03-15 8:56 [PATCH] io_uring: don't iterate ctx list to update sq_thread_idle Hao Xu
2021-03-15 10:49 ` Pavel Begunkov
2021-03-16 2:41 ` Hao Xu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox