public inbox for [email protected]
 help / color / mirror / Atom feed
From: Pavel Begunkov <[email protected]>
To: Hao Xu <[email protected]>, Jens Axboe <[email protected]>
Cc: [email protected], Joseph Qi <[email protected]>
Subject: Re: [PATCH 4/6] io_uring: let fast poll support multishot
Date: Sat, 11 Sep 2021 11:49:55 +0100	[thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>

On 9/9/21 9:29 AM, Hao Xu wrote:
> 在 2021/9/9 下午3:01, Hao Xu 写道:
>> 在 2021/9/8 下午8:03, Pavel Begunkov 写道:
>>> On 9/8/21 12:21 PM, Hao Xu wrote:
>>>> 在 2021/9/7 下午2:48, Hao Xu 写道:
>>>>> 在 2021/9/7 上午3:04, Pavel Begunkov 写道:
>>>>>> On 9/3/21 12:00 PM, Hao Xu wrote:
>>>>>>> For operations like accept, multishot is a useful feature, since we can
>>>>>>> reduce a number of accept sqe. Let's integrate it to fast poll, it may
>>>>>>> be good for other operations in the future.
>>>>>>>
>>>>>>> Signed-off-by: Hao Xu <[email protected]>
>>>>>>> ---
>>>>>>>    fs/io_uring.c | 15 ++++++++++++---
>>>>>>>    1 file changed, 12 insertions(+), 3 deletions(-)
>>>>>>>
>>>>>>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>>>>>>> index d6df60c4cdb9..dae7044e0c24 100644
>>>>>>> --- a/fs/io_uring.c
>>>>>>> +++ b/fs/io_uring.c
>>>>>>> @@ -5277,8 +5277,15 @@ static void io_async_task_func(struct io_kiocb *req, bool *locked)
>>>>>>>            return;
>>>>>>>        }
>>>>>>> -    hash_del(&req->hash_node);
>>>>>>> -    io_poll_remove_double(req);
>>>>>>> +    if (READ_ONCE(apoll->poll.canceled))
>>>>>>> +        apoll->poll.events |= EPOLLONESHOT;
>>>>>>> +    if (apoll->poll.events & EPOLLONESHOT) {
>>>>>>> +        hash_del(&req->hash_node);
>>>>>>> +        io_poll_remove_double(req);
>>>>>>> +    } else {
>>>>>>> +        add_wait_queue(apoll->poll.head, &apoll->poll.wait);
>>>>>>
>>>>>> It looks like it does both io_req_task_submit() and adding back
>>>>>> to the wq, so io_issue_sqe() may be called in parallel with
>>>>>> io_async_task_func(). If so, there will be tons of all kind of
>>>>>> races.
>>>>> IMHO, io_async_task_func() is called in original context one by
>>>>> one(except PF_EXITING is set, it is also called in system-wq), so
>>>>> shouldn't be parallel case there.
>>>> ping...
>>>
>>> fwiw, the case we're talking about:
>>>
>>> CPU0                            | CPU1
>>> io_async_task_func()            |
>>> -> add_wait_queue();            |
>>> -> io_req_task_submit();        |
>>>                 /* no tw run happened in between */
>>>                                  | io_async_task_func()
>>>                                  | --> io_req_task_submit()
>>>
>>> We called io_req_task_submit() twice without running tw in-between,
>>> both of the calls use the same req->io_task_work.node field in the
>>> request for accounting, and so the second call will screw
>>> tctx->task_list and not only by not considering that
>>> req->io_task_work.node is already taken/enqueued.
>>>
>>> io_req_task_work_add() {
>>>          wq_list_add_tail(&req->io_task_work.node, &tctx->task_list);
>>> }
>>>
>> I guess you mean io_req_task_work_add() called by async_wake() two times:
>> io_async_task_func()
>> -> add_wait_queue()
>>                              async_wake()
>>                              ->io_req_task_work_add()
>>                              this one mess up the running task_work list
>>                              since req->io_task_work.node is in use.
>>
>> It seems the current poll_add + multishot logic has this issue too, I'll
>> give it a shot(simply clean req->io_task_work.node before running
>> req->io_task_work.func should work)
> Similar issue for double wait entry since we didn't remove double entry
> in interrupt handler:

Yep, sounds like that. Polling needs reworking, and not only
because of this one.


> async_wake() --> io_req_task_work_add()
> io_poll_double_wake()-->async_wake()-->io_req_task_work_add()
> 
>>>>>>
>>>>>>> +    }
>>>>>>> +
>>>>>>>        spin_unlock(&ctx->completion_lock);
>>>>>>>        if (!READ_ONCE(apoll->poll.canceled))
>>>>>>> @@ -5366,7 +5373,7 @@ static int io_arm_poll_handler(struct io_kiocb *req)
>>>>>>>        struct io_ring_ctx *ctx = req->ctx;
>>>>>>>        struct async_poll *apoll;
>>>>>>>        struct io_poll_table ipt;
>>>>>>> -    __poll_t ret, mask = EPOLLONESHOT | POLLERR | POLLPRI;
>>>>>>> +    __poll_t ret, mask = POLLERR | POLLPRI;
>>>>>>>        int rw;
>>>>>>>        if (!req->file || !file_can_poll(req->file))
>>>>>>> @@ -5388,6 +5395,8 @@ static int io_arm_poll_handler(struct io_kiocb *req)
>>>>>>>            rw = WRITE;
>>>>>>>            mask |= POLLOUT | POLLWRNORM;
>>>>>>>        }
>>>>>>> +    if (!(req->flags & REQ_F_APOLL_MULTISHOT))
>>>>>>> +        mask |= EPOLLONESHOT;
>>>>>>>        /* if we can't nonblock try, then no point in arming a poll handler */
>>>>>>>        if (!io_file_supports_nowait(req, rw))
>>>>>>>
>>>>>>
>>>>
>>>
> 

-- 
Pavel Begunkov

  reply	other threads:[~2021-09-11 10:50 UTC|newest]

Thread overview: 46+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-09-03 11:00 [RFC 0/6] fast poll multishot mode Hao Xu
2021-09-03 11:00 ` [PATCH 1/6] io_uring: enhance flush completion logic Hao Xu
2021-09-03 11:42   ` Pavel Begunkov
2021-09-03 12:08     ` Hao Xu
2021-09-03 12:27       ` Pavel Begunkov
2021-09-03 13:38         ` Hao Xu
2021-09-17 18:49           ` Hao Xu
2021-09-03 11:00 ` [PATCH 2/6] io_uring: add IORING_ACCEPT_MULTISHOT for accept Hao Xu
2021-09-03 11:00 ` [PATCH 3/6] io_uring: add REQ_F_APOLL_MULTISHOT for requests Hao Xu
2021-09-03 11:00 ` [PATCH 4/6] io_uring: let fast poll support multishot Hao Xu
2021-09-06 15:56   ` Pavel Begunkov
2021-09-06 17:40     ` Hao Xu
2021-09-06 19:09       ` Pavel Begunkov
2021-09-07  6:38         ` Hao Xu
2021-09-06 19:04   ` Pavel Begunkov
2021-09-07  6:48     ` Hao Xu
2021-09-08 11:21       ` Hao Xu
2021-09-08 12:03         ` Pavel Begunkov
2021-09-08 13:13           ` Pavel Begunkov
2021-09-09  7:01           ` Hao Xu
2021-09-09  8:29             ` Hao Xu
2021-09-11 10:49               ` Pavel Begunkov [this message]
2021-09-11 20:19                 ` Hao Xu
2021-09-03 11:00 ` [PATCH 5/6] io_uring: implement multishot mode for accept Hao Xu
2021-09-04 22:39   ` Pavel Begunkov
2021-09-04 22:40     ` Pavel Begunkov
2021-09-06 15:34       ` Pavel Begunkov
2021-09-03 11:00 ` [PATCH 6/6] io_uring: enable " Hao Xu
2021-09-03 16:29   ` Jens Axboe
2021-09-04 15:34     ` Hao Xu
2021-09-04 18:40       ` Jens Axboe
2021-09-04 22:46         ` Pavel Begunkov
2021-09-05  7:29           ` Hao Xu
2021-09-05 19:44           ` Jens Axboe
2021-09-06  8:26             ` Hao Xu
2021-09-06  8:28               ` Hao Xu
2021-09-06 13:24               ` Jens Axboe
2021-09-06 12:35             ` Hao Xu
2021-09-06 13:31               ` Jens Axboe
2021-09-06 15:00                 ` Hao Xu
2021-09-06 15:32               ` Pavel Begunkov
2021-09-06 16:42                 ` Jens Axboe
2021-09-04 22:43   ` Pavel Begunkov
2021-09-05  6:25     ` Hao Xu
2021-09-05  8:27       ` Pavel Begunkov
2021-09-03 11:02 ` [RFC 0/6] fast poll multishot mode Hao Xu

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox