public inbox for [email protected]
 help / color / mirror / Atom feed
From: Jens Axboe <[email protected]>
To: Artyom Pavlov <[email protected]>, [email protected]
Subject: Re: Sending CQE to a different ring
Date: Wed, 9 Mar 2022 18:55:57 -0700	[thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>

On 3/9/22 6:36 PM, Jens Axboe wrote:
> On 3/9/22 4:49 PM, Artyom Pavlov wrote:
>> Greetings!
>>
>> A common approach for multi-threaded servers is to have a number of
>> threads equal to a number of cores and launch a separate ring in each
>> one. AFAIK currently if we want to send an event to a different ring,
>> we have to write-lock this ring, create SQE, and update the index
>> ring. Alternatively, we could use some kind of user-space message
>> passing.
>>
>> Such approaches are somewhat inefficient and I think it can be solved
>> elegantly by updating the io_uring_sqe type to allow accepting fd of a
>> ring to which CQE must be sent by kernel. It can be done by
>> introducing an IOSQE_ flag and using one of currently unused padding
>> u64s.
>>
>> Such feature could be useful for load balancing and message passing
>> between threads which would ride on top of io-uring, i.e. you could
>> send NOP with user_data pointing to a message payload.
> 
> So what you want is a NOP with 'fd' set to the fd of another ring, and
> that nop posts a CQE on that other ring? I don't think we'd need IOSQE
> flags for that, we just need a NOP that supports that. I see a few ways
> of going about that:
> 
> 1) Add a new 'NOP' that takes an fd, and validates that that fd is an
>    io_uring instance. It can then grab the completion lock on that ring
>    and post an empty CQE.
> 
> 2) We add a FEAT flag saying NOP supports taking an 'fd' argument, where
>    'fd' is another ring. Posting CQE same as above.
> 
> 3) We add a specific opcode for this. Basically the same as #2, but
>    maybe with a more descriptive name than NOP.
> 
> Might make sense to pair that with a CQE flag or something like that, as
> there's no specific user_data that could be used as it doesn't match an
> existing SQE that has been issued. IORING_CQE_F_WAKEUP for example.
> Would be applicable to all the above cases.
> 
> I kind of like #3 the best. Add a IORING_OP_RING_WAKEUP command, require
> that sqe->fd point to a ring (could even be the ring itself, doesn't
> matter). And add IORING_CQE_F_WAKEUP as a specific flag for that.

Something like the below, totally untested. The request will complete on
the original ring with either 0, for success, or -EOVERFLOW if the
target ring was already in an overflow state. If the fd specified isn't
an io_uring context, then the request will complete with -EBADFD.

If you have any way of testing this, please do. I'll write a basic
functionality test for it as well, but not until tomorrow.

Maybe we want to include in cqe->res who the waker was? We can stuff the
pid/tid in there, for example.

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 2e04f718319d..b21f85a48224 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -1105,6 +1105,9 @@ static const struct io_op_def io_op_defs[] = {
 	[IORING_OP_MKDIRAT] = {},
 	[IORING_OP_SYMLINKAT] = {},
 	[IORING_OP_LINKAT] = {},
+	[IORING_OP_WAKEUP_RING] = {
+		.needs_file		= 1,
+	},
 };
 
 /* requests with any of those set should undergo io_disarm_next() */
@@ -4235,6 +4238,44 @@ static int io_nop(struct io_kiocb *req, unsigned int issue_flags)
 	return 0;
 }
 
+static int io_wakeup_ring_prep(struct io_kiocb *req,
+			       const struct io_uring_sqe *sqe)
+{
+	if (unlikely(sqe->addr || sqe->ioprio || sqe->buf_index || sqe->off ||
+		     sqe->len || sqe->rw_flags || sqe->splice_fd_in ||
+		     sqe->buf_index || sqe->personality))
+		return -EINVAL;
+
+	if (req->file->f_op != &io_uring_fops)
+		return -EBADFD;
+
+	return 0;
+}
+
+static int io_wakeup_ring(struct io_kiocb *req, unsigned int issue_flags)
+{
+	struct io_uring_cqe *cqe;
+	struct io_ring_ctx *ctx;
+	int ret = 0;
+
+	ctx = req->file->private_data;
+	spin_lock(&ctx->completion_lock);
+	cqe = io_get_cqe(ctx);
+	if (cqe) {
+		WRITE_ONCE(cqe->user_data, 0);
+		WRITE_ONCE(cqe->res, 0);
+		WRITE_ONCE(cqe->flags, IORING_CQE_F_WAKEUP);
+	} else {
+		ret = -EOVERFLOW;
+	}
+	io_commit_cqring(ctx);
+	spin_unlock(&ctx->completion_lock);
+	io_cqring_ev_posted(ctx);
+
+	__io_req_complete(req, issue_flags, ret, 0);
+	return 0;
+}
+
 static int io_fsync_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 {
 	struct io_ring_ctx *ctx = req->ctx;
@@ -6568,6 +6609,8 @@ static int io_req_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 		return io_symlinkat_prep(req, sqe);
 	case IORING_OP_LINKAT:
 		return io_linkat_prep(req, sqe);
+	case IORING_OP_WAKEUP_RING:
+		return io_wakeup_ring_prep(req, sqe);
 	}
 
 	printk_once(KERN_WARNING "io_uring: unhandled opcode %d\n",
@@ -6851,6 +6894,9 @@ static int io_issue_sqe(struct io_kiocb *req, unsigned int issue_flags)
 	case IORING_OP_LINKAT:
 		ret = io_linkat(req, issue_flags);
 		break;
+	case IORING_OP_WAKEUP_RING:
+		ret = io_wakeup_ring(req, issue_flags);
+		break;
 	default:
 		ret = -EINVAL;
 		break;
diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index 787f491f0d2a..088232133594 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -143,6 +143,7 @@ enum {
 	IORING_OP_MKDIRAT,
 	IORING_OP_SYMLINKAT,
 	IORING_OP_LINKAT,
+	IORING_OP_WAKEUP_RING,
 
 	/* this goes last, obviously */
 	IORING_OP_LAST,
@@ -199,9 +200,11 @@ struct io_uring_cqe {
  *
  * IORING_CQE_F_BUFFER	If set, the upper 16 bits are the buffer ID
  * IORING_CQE_F_MORE	If set, parent SQE will generate more CQE entries
+ * IORING_CQE_F_WAKEUP	Wakeup request CQE, no link to an SQE
  */
 #define IORING_CQE_F_BUFFER		(1U << 0)
 #define IORING_CQE_F_MORE		(1U << 1)
+#define IORING_CQE_F_WAKEUP		(1U << 2)
 
 enum {
 	IORING_CQE_BUFFER_SHIFT		= 16,

-- 
Jens Axboe



  reply	other threads:[~2022-03-10  1:56 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-03-09 23:49 Sending CQE to a different ring Artyom Pavlov
2022-03-10  1:36 ` Jens Axboe
2022-03-10  1:55   ` Jens Axboe [this message]
2022-03-10  2:33     ` Jens Axboe
     [not found]   ` <[email protected]>
2022-03-10  3:00     ` Jens Axboe
     [not found]       ` <[email protected]>
     [not found]         ` <[email protected]>
     [not found]           ` <[email protected]>
2022-03-10 15:36             ` Artyom Pavlov
2022-03-10  3:06     ` Jens Axboe

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] \
    /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