public inbox for [email protected]
 help / color / mirror / Atom feed
From: Jens Axboe <[email protected]>
To: David Wei <[email protected]>, [email protected]
Cc: Pavel Begunkov <[email protected]>
Subject: Re: [RFC PATCH v1] io_uring: only account cqring wait time as iowait if enabled for a ring
Date: Fri, 23 Feb 2024 07:31:40 -0700	[thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>

On 2/22/24 10:40 PM, David Wei wrote:
> diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h
> index bd7071aeec5d..57318fc01379 100644
> --- a/include/linux/io_uring_types.h
> +++ b/include/linux/io_uring_types.h
> @@ -425,6 +425,9 @@ struct io_ring_ctx {
>  	DECLARE_HASHTABLE(napi_ht, 4);
>  #endif
>  
> +	/* iowait accounting */
> +	bool				iowait_enabled;
> +

Since this is just a single bit, you should put it in the top section
where we have other single bits for hot / read-mostly data. This avoids
needing something many cache lines away for the hotter wait path, and it
avoids growing the struct as there's still plenty of space there for
this.

> diff --git a/io_uring/register.c b/io_uring/register.c
> index 99c37775f974..7cbc08544c4c 100644
> --- a/io_uring/register.c
> +++ b/io_uring/register.c
> @@ -387,6 +387,12 @@ static __cold int io_register_iowq_max_workers(struct io_ring_ctx *ctx,
>  	return ret;
>  }
>  
> +static int io_register_iowait(struct io_ring_ctx *ctx)
> +{
> +	ctx->iowait_enabled = true;
> +	return 0;
> +}
> +
>  static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
>  			       void __user *arg, unsigned nr_args)
>  	__releases(ctx->uring_lock)
> @@ -563,6 +569,12 @@ static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
>  			break;
>  		ret = io_unregister_napi(ctx, arg);
>  		break;
> +	case IORING_REGISTER_IOWAIT:
> +		ret = -EINVAL;
> +		if (arg || nr_args)
> +			break;
> +		ret = io_register_iowait(ctx);
> +		break;


This only allows you to set it, not clear it. I think we want to make it
pass in the value, and pass back the previous. Something ala:

static int io_register_iowait(struct io_ring_ctx *ctx, int val)
{
	int was_enabled = ctx->iowait_enabled;

	if (val)
		ctx->iowait_enabled = 1;
	else
		ctx->iowait_enabled = 0;
	return was_enabled;
}

and then:

	case IORING_REGISTER_IOWAIT:
		ret = -EINVAL;
		if (arg)
			break;
		ret = io_register_iowait(ctx, nr_args);
		break;

I'd also add a:

Fixes: 8a796565cec3 ("io_uring: Use io_schedule* in cqring wait")

and mark it for stable, so we at least attempt to make it something that
can be depended on.

-- 
Jens Axboe


  reply	other threads:[~2024-02-23 14:31 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-02-23  5:40 [RFC PATCH v1] io_uring: only account cqring wait time as iowait if enabled for a ring David Wei
2024-02-23 14:31 ` Jens Axboe [this message]
2024-02-23 17:17   ` David Wei

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] \
    /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