netdev - Re: IORING_SEND_NOTIF_REPORT_USAGE (was Re: IORING_CQE_F

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 21 Oct 2022 16:03:54 +0200
From:   Stefan Metzmacher <metze@...ba.org>
To:     Pavel Begunkov <asml.silence@...il.com>,
        io-uring <io-uring@...r.kernel.org>, Jens Axboe <axboe@...nel.dk>
Cc:     Jakub Kicinski <kuba@...nel.org>, netdev <netdev@...r.kernel.org>,
        Dylan Yudaken <dylany@...com>
Subject: Re: IORING_SEND_NOTIF_REPORT_USAGE (was Re: IORING_CQE_F_COPIED)

Am 21.10.22 um 13:09 schrieb Pavel Begunkov:
> On 10/21/22 10:36, Stefan Metzmacher wrote:
>> Hi Pavel,
> [...]
>>> Right, I'm just tired of back porting patches by hand :)
>>
>> ok, I just assumed it would be 6.1 only.
> 
> I'm fine with 6.1 only, it'd make things easier. I thought from
> your first postings you wanted it 6.0. Then we don't need to care
> about the placing of the copied/used flags.
> 
>>>> Otherwise we could have IORING_CQE_F_COPIED by default without opt-in
>>>> flag...
>>
>> Do you still want an opt-in flag to get IORING_CQE_F_COPIED?
>> If so what name do you want it to be?
> 
> Ala a IORING_SEND_* flag? Yes please.
> 
> *_REPORT_USAGE was fine but I'd make it IORING_SEND_ZC_REPORT_USAGE.
> And can be extended if there is more info needed in the future.
> 
> And I don't mind using a bit in cqe->res, makes cflags less polluted.

So no worries about the delayed/skip sendmsg completion anymore?

Should I define it like this, ok?

#define IORING_NOTIF_USAGE_ZC_COPIED    (1U << 31)

See the full patch below...

metze

diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index d69ae7eba773..32e1f2a55b70 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -296,10 +296,24 @@ enum io_uring_op {
   *
   * IORING_RECVSEND_FIXED_BUF	Use registered buffers, the index is stored in
   *				the buf_index field.
+
+ * IORING_SEND_NOTIF_REPORT_USAGE
+ *				If SEND[MSG]_ZC should report
+ *				the zerocopy usage in cqe.res
+ *				for the IORING_CQE_F_NOTIF cqe.
+ *				IORING_NOTIF_USAGE_ZC_COPIED if data was copied
+ *				(at least partially).
   */
  #define IORING_RECVSEND_POLL_FIRST	(1U << 0)
  #define IORING_RECV_MULTISHOT		(1U << 1)
  #define IORING_RECVSEND_FIXED_BUF	(1U << 2)
+#define IORING_SEND_ZC_REPORT_USAGE	(1U << 3)
+
+/*
+ * cqe.res for IORING_CQE_F_NOTIF if
+ * IORING_SEND_ZC_REPORT_USAGE was requested
+ */
+#define IORING_NOTIF_USAGE_ZC_COPIED    (1U << 31)

  /*
   * accept flags stored in sqe->ioprio
diff --git a/io_uring/net.c b/io_uring/net.c
index 56078f47efe7..1aa3b50b3e82 100644
--- a/io_uring/net.c
+++ b/io_uring/net.c
@@ -939,7 +939,8 @@ int io_send_zc_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)

  	zc->flags = READ_ONCE(sqe->ioprio);
  	if (zc->flags & ~(IORING_RECVSEND_POLL_FIRST |
-			  IORING_RECVSEND_FIXED_BUF))
+			  IORING_RECVSEND_FIXED_BUF |
+			  IORING_SEND_ZC_REPORT_USAGE))
  		return -EINVAL;
  	notif = zc->notif = io_alloc_notif(ctx);
  	if (!notif)
@@ -957,6 +958,9 @@ int io_send_zc_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
  		req->imu = READ_ONCE(ctx->user_bufs[idx]);
  		io_req_set_rsrc_node(notif, ctx, 0);
  	}
+	if (zc->flags & IORING_SEND_ZC_REPORT_USAGE) {
+		io_notif_to_data(notif)->zc_report = true;
+	}

  	if (req->opcode == IORING_OP_SEND_ZC) {
  		if (READ_ONCE(sqe->__pad3[0]))
diff --git a/io_uring/notif.c b/io_uring/notif.c
index e37c6569d82e..4bfef10161fa 100644
--- a/io_uring/notif.c
+++ b/io_uring/notif.c
@@ -18,6 +18,10 @@ static void __io_notif_complete_tw(struct io_kiocb *notif, bool *locked)
  		__io_unaccount_mem(ctx->user, nd->account_pages);
  		nd->account_pages = 0;
  	}
+
+	if (nd->zc_report && (nd->zc_copied || !nd->zc_used))
+		notif->cqe.res |= IORING_NOTIF_USAGE_ZC_COPIED;
+
  	io_req_task_complete(notif, locked);
  }

@@ -28,6 +32,13 @@ static void io_uring_tx_zerocopy_callback(struct sk_buff *skb,
  	struct io_notif_data *nd = container_of(uarg, struct io_notif_data, uarg);
  	struct io_kiocb *notif = cmd_to_io_kiocb(nd);

+	if (nd->zc_report) {
+		if (success && !nd->zc_used && skb)
+			WRITE_ONCE(nd->zc_used, true);
+		else if (!success && !nd->zc_copied)
+			WRITE_ONCE(nd->zc_copied, true);
+	}
+
  	if (refcount_dec_and_test(&uarg->refcnt)) {
  		notif->io_task_work.func = __io_notif_complete_tw;
  		io_req_task_work_add(notif);
@@ -55,6 +66,7 @@ struct io_kiocb *io_alloc_notif(struct io_ring_ctx *ctx)
  	nd->account_pages = 0;
  	nd->uarg.flags = SKBFL_ZEROCOPY_FRAG | SKBFL_DONT_ORPHAN;
  	nd->uarg.callback = io_uring_tx_zerocopy_callback;
+	nd->zc_report = nd->zc_used = nd->zc_copied = false;
  	refcount_set(&nd->uarg.refcnt, 1);
  	return notif;
  }
diff --git a/io_uring/notif.h b/io_uring/notif.h
index e4fbcae0f3fd..6be2e5ae8581 100644
--- a/io_uring/notif.h
+++ b/io_uring/notif.h
@@ -15,6 +15,9 @@ struct io_notif_data {
  	struct file		*file;
  	struct ubuf_info	uarg;
  	unsigned long		account_pages;
+	bool			zc_report;
+	bool			zc_used;
+	bool			zc_copied;
  };

  void io_notif_flush(struct io_kiocb *notif);