From: Stefano Brivio <sbrivio@redhat.com>
To: David Gibson <david@gibson.dropbear.id.au>
Cc: passt-dev@passt.top
Subject: Re: [PATCH 09/11] udp: Consolidate datagram batching
Date: Fri, 5 Jul 2024 11:10:45 +0200 [thread overview]
Message-ID: <20240705111045.3ddfa560@elisabeth> (raw)
In-Reply-To: <20240704045835.1149746-10-david@gibson.dropbear.id.au>
On Thu, 4 Jul 2024 14:58:33 +1000
David Gibson <david@gibson.dropbear.id.au> wrote:
> When we receive datagrams on a socket, we need to split them into batches
> depending on how they need to be forwarded (either via a specific splice
> socket, or via tap). The logic to do this, is somewhat awkwardly split
> between udp_buf_sock_handler() itself, udp_splice_send() and
> udp_tap_send().
>
> Move all the batching logic into udp_buf_sock_handler(), leaving
> udp_splice_send() to just send the prepared batch. udp_tap_send() reduces
> to just a call to tap_send_frames() so open-code that call in
> udp_buf_sock_handler().
>
> This will allow separating the batching logic from the rest of the datagram
> forwarding logic, which we'll need for upcoming flow table support.
>
> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> ---
> udp.c | 128 ++++++++++++++++++----------------------------------------
> 1 file changed, 39 insertions(+), 89 deletions(-)
>
> diff --git a/udp.c b/udp.c
> index 8ed59639..a317e986 100644
> --- a/udp.c
> +++ b/udp.c
> @@ -501,42 +501,29 @@ static void udp_splice_prepare(struct mmsghdr *mmh, unsigned idx)
> }
>
> /**
> - * udp_splice_send() - Send datagrams from socket to socket
> + * udp_splice_send() - Send a batch of datagrams from socket to socket
> * @c: Execution context
> - * @start: Index of first datagram in udp[46]_l2_buf
> - * @n: Total number of datagrams in udp[46]_l2_buf pool
> - * @dst: Datagrams will be sent to this port (on destination side)
> + * @start: Index of batch's first datagram in udp[46]_l2_buf
> + * @n: Number of datagrams in batch
> + * @src: Source port for datagram (target side)
> + * @dst: Destination port for datagrams (target side)
> * @ref: epoll reference for origin socket
> * @now: Timestamp
> - *
> - * This consumes as many datagrams as are sendable via a single socket. It
> - * requires that udp_meta[@start].splicesrc is initialised, and will initialise
> - * udp_meta[].splicesrc for each datagram it consumes *and one more* (if
> - * present).
> - *
> - * Return: Number of datagrams forwarded
> */
> -static unsigned udp_splice_send(const struct ctx *c, size_t start, size_t n,
> - in_port_t dst, union epoll_ref ref,
> - const struct timespec *now)
> +static void udp_splice_send(const struct ctx *c, size_t start, size_t n,
> + in_port_t src, in_port_t dst,
> + union epoll_ref ref,
> + const struct timespec *now)
> {
> - in_port_t src = udp_meta[start].splicesrc;
> - struct mmsghdr *mmh_recv;
> - unsigned int i = start;
> int s;
>
> - ASSERT(udp_meta[start].splicesrc >= 0);
> - ASSERT(ref.type == EPOLL_TYPE_UDP);
> -
> if (ref.udp.v6) {
> - mmh_recv = udp6_mh_recv;
> udp_spliceto.sa6 = (struct sockaddr_in6) {
> .sin6_family = AF_INET6,
> .sin6_addr = in6addr_loopback,
> .sin6_port = htons(dst),
> };
> } else {
> - mmh_recv = udp4_mh_recv;
> udp_spliceto.sa4 = (struct sockaddr_in) {
> .sin_family = AF_INET,
> .sin_addr = in4addr_loopback,
> @@ -544,15 +531,6 @@ static unsigned udp_splice_send(const struct ctx *c, size_t start, size_t n,
> };
> }
>
> - do {
> - udp_splice_prepare(mmh_recv, i);
> -
> - if (++i >= n)
> - break;
> -
> - udp_meta[i].splicesrc = udp_mmh_splice_port(ref, &mmh_recv[i]);
> - } while (udp_meta[i].splicesrc == src);
> -
> if (ref.udp.pif == PIF_SPLICE) {
> src += c->udp.fwd_in.rdelta[src];
> s = udp_splice_init[ref.udp.v6][src].sock;
> @@ -560,7 +538,7 @@ static unsigned udp_splice_send(const struct ctx *c, size_t start, size_t n,
> s = udp_splice_new(c, ref.udp.v6, src, false);
>
> if (s < 0)
> - goto out;
> + return;
>
> udp_splice_ns[ref.udp.v6][dst].ts = now->tv_sec;
> udp_splice_init[ref.udp.v6][src].ts = now->tv_sec;
> @@ -577,15 +555,13 @@ static unsigned udp_splice_send(const struct ctx *c, size_t start, size_t n,
> s = arg.s;
> }
> if (s < 0)
> - goto out;
> + return;
>
> udp_splice_init[ref.udp.v6][dst].ts = now->tv_sec;
> udp_splice_ns[ref.udp.v6][src].ts = now->tv_sec;
> }
>
> - sendmmsg(s, udp_mh_splice + start, i - start, MSG_NOSIGNAL);
> -out:
> - return i - start;
> + sendmmsg(s, udp_mh_splice + start, n, MSG_NOSIGNAL);
> }
>
> /**
> @@ -725,7 +701,7 @@ static size_t udp_update_hdr6(const struct ctx *c,
> * @v6: Prepare for IPv6?
> * @now: Current timestamp
> */
> -static void udp_tap_prepare(const struct ctx *c, struct mmsghdr *mmh,
> +static void udp_tap_prepare(const struct ctx *c, const struct mmsghdr *mmh,
> unsigned idx, in_port_t dstport, bool v6,
> const struct timespec *now)
> {
> @@ -752,49 +728,6 @@ static void udp_tap_prepare(const struct ctx *c, struct mmsghdr *mmh,
> (*tap_iov)[UDP_IOV_PAYLOAD].iov_len = l4len;
> }
>
> -/**
> - * udp_tap_send() - Prepare UDP datagrams and send to tap interface
> - * @c: Execution context
> - * @start: Index of first datagram in udp[46]_l2_buf pool
> - * @n: Total number of datagrams in udp[46]_l2_buf pool
> - * @dstport: Destination port number on destination side
> - * @ref: Epoll reference for origin socket
> - * @now: Current timestamp
> - *
> - * This consumes as many frames as are sendable via tap. It requires that
> - * udp_meta[@start].splicesrc is initialised, and will initialise
> - * udp_meta[].splicesrc for each frame it consumes *and one more* (if present).
> - *
> - * Return: Number of frames sent via tap
> - */
> -static unsigned udp_tap_send(const struct ctx *c, size_t start, size_t n,
> - in_port_t dstport, union epoll_ref ref,
> - const struct timespec *now)
> -{
> - struct mmsghdr *mmh_recv;
> - size_t i = start;
> -
> - ASSERT(udp_meta[start].splicesrc == -1);
> - ASSERT(ref.type == EPOLL_TYPE_UDP);
> -
> - if (ref.udp.v6)
> - mmh_recv = udp6_mh_recv;
> - else
> - mmh_recv = udp4_mh_recv;
> -
> - do {
> - udp_tap_prepare(c, mmh_recv, i, dstport, ref.udp.v6, now);
> -
> - if (++i >= n)
> - break;
> -
> - udp_meta[i].splicesrc = udp_mmh_splice_port(ref, &mmh_recv[i]);
> - } while (udp_meta[i].splicesrc == -1);
> -
> - tap_send_frames(c, &udp_l2_iov[start][0], UDP_NUM_IOVS, i - start);
> - return i - start;
> -}
> -
> /**
> * udp_sock_recv() - Receive datagrams from a socket
> * @c: Execution context
> @@ -842,7 +775,7 @@ void udp_buf_sock_handler(const struct ctx *c, union epoll_ref ref, uint32_t eve
> {
> struct mmsghdr *mmh_recv = ref.udp.v6 ? udp6_mh_recv : udp4_mh_recv;
> in_port_t dstport = ref.udp.port;
> - int n, m, i;
> + int n, i;
>
> if ((n = udp_sock_recv(c, ref.fd, events, mmh_recv)) <= 0)
> return;
> @@ -852,19 +785,36 @@ void udp_buf_sock_handler(const struct ctx *c, union epoll_ref ref, uint32_t eve
> else if (ref.udp.pif == PIF_HOST)
> dstport += c->udp.fwd_in.f.delta[dstport];
>
> - /* We divide things into batches based on how we need to send them,
> + /* We divide datagrams into batches based on how we need to send them,
> * determined by udp_meta[i].splicesrc. To avoid either two passes
> * through the array, or recalculating splicesrc for a single entry, we
> - * have to populate it one entry *ahead* of the loop counter (if
> - * present). So we fill in entry 0 before the loop, then udp_*_send()
> - * populate one entry past where they consume.
> + * have to populate it one entry *ahead* of the loop counter.
> */
> udp_meta[0].splicesrc = udp_mmh_splice_port(ref, mmh_recv);
> - for (i = 0; i < n; i += m) {
> - if (udp_meta[i].splicesrc >= 0)
> - m = udp_splice_send(c, i, n, dstport, ref, now);
> + for (i = 0; i < n; ) {
> + int batchsrc = udp_meta[i].splicesrc;
> + int batchstart = i;
> +
> + do {
> + if (batchsrc >= 0)
> + udp_splice_prepare(mmh_recv, i);
> + else
> + udp_tap_prepare(c, mmh_recv, i, dstport,
> + ref.udp.v6, now);
> +
> + if (++i >= n)
> + break;
> +
> + udp_meta[i].splicesrc = udp_mmh_splice_port(ref,
> + &mmh_recv[i]);
> + } while (udp_meta[i].splicesrc == batchsrc);
> +
> + if (batchsrc >= 0)
> + udp_splice_send(c, batchstart, i - batchstart,
> + batchsrc, dstport, ref, now);
> else
> - m = udp_tap_send(c, i, n, dstport, ref, now);
> + tap_send_frames(c, &udp_l2_iov[batchstart][0],
> + UDP_NUM_IOVS, i - batchstart);
> }
The logic looks correct to me, but the nested loop makes it a bit hard
to grasp.
I'm wondering if we shouldn't rather have a single loop, always
preparing the datagrams, noting down the previous
udp_meta[i].splicesrc and the first index of a batch, and starting a new
batch (sending the previous one) once the current udp_meta[i].splicesrc
doesn't match the previous value.
I tried to sketch this quickly but failed for the moment to come up
with anything vaguely elegant, so I'm fine with either version.
Nits: curly brackets around multiple lines.
--
Stefano
next prev parent reply other threads:[~2024-07-05 9:11 UTC|newest]
Thread overview: 24+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-07-04 4:58 [PATCH 00/11] Preliminaries for UDP flow support David Gibson
2024-07-04 4:58 ` [PATCH 01/11] util: sock_l4() determine protocol from epoll type rather than the reverse David Gibson
2024-07-04 21:19 ` Stefano Brivio
2024-07-04 23:51 ` David Gibson
2024-07-04 4:58 ` [PATCH 02/11] flow: Add flow_sidx_valid() helper David Gibson
2024-07-04 4:58 ` [PATCH 03/11] udp: Pass full epoll reference through more of sock handler path David Gibson
2024-07-04 21:20 ` Stefano Brivio
2024-07-04 23:54 ` David Gibson
2024-07-04 4:58 ` [PATCH 04/11] udp: Rename IOV and mmsghdr arrays David Gibson
2024-07-04 21:20 ` Stefano Brivio
2024-07-05 0:00 ` David Gibson
2024-07-04 4:58 ` [PATCH 05/11] udp: Unify udp[46]_mh_splice David Gibson
2024-07-04 4:58 ` [PATCH 06/11] udp: Unify udp[46]_l2_iov David Gibson
2024-07-04 4:58 ` [PATCH 07/11] udp: Don't repeatedly initialise udp[46]_eth_hdr David Gibson
2024-07-04 4:58 ` [PATCH 08/11] udp: Move some more of sock_handler tasks into sub-functions David Gibson
2024-07-04 4:58 ` [PATCH 09/11] udp: Consolidate datagram batching David Gibson
2024-07-05 9:10 ` Stefano Brivio [this message]
2024-07-05 9:36 ` David Gibson
2024-07-04 4:58 ` [PATCH 10/11] contrib: Add program to document and test assumptions about SO_REUSEADDR David Gibson
2024-07-04 21:21 ` Stefano Brivio
2024-07-05 0:06 ` David Gibson
2024-07-05 8:33 ` Stefano Brivio
2024-07-05 9:49 ` David Gibson
2024-07-04 4:58 ` [PATCH 11/11] contrib: Test behaviour of zero length datagram recv()s David Gibson
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 \
--in-reply-to=20240705111045.3ddfa560@elisabeth \
--to=sbrivio@redhat.com \
--cc=david@gibson.dropbear.id.au \
--cc=passt-dev@passt.top \
/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.
Code repositories for project(s) associated with this public inbox
https://passt.top/passt
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for IMAP folder(s).