From: Adrien Mazarguil <adrien.mazarguil@6wind.com>
To: Yongseok Koh <yskoh@mellanox.com>
Cc: Shahaf Shuler <shahafs@mellanox.com>,
Nelio Laranjeiro <nelio.laranjeiro@6wind.com>,
dev@dpdk.org
Subject: Re: [dpdk-dev] [PATCH 1/6] net/mlx5: lay groundwork for switch offloads
Date: Thu, 12 Jul 2018 12:46:35 +0200 [thread overview]
Message-ID: <20180712104635.GS5211@6wind.com> (raw)
In-Reply-To: <20180712001652.GC69686@yongseok-MBP.local>
On Wed, Jul 11, 2018 at 05:17:09PM -0700, Yongseok Koh wrote:
> On Wed, Jun 27, 2018 at 08:08:10PM +0200, Adrien Mazarguil wrote:
> > With mlx5, unlike normal flow rules implemented through Verbs for traffic
> > emitted and received by the application, those targeting different logical
> > ports of the device (VF representors for instance) are offloaded at the
> > switch level and must be configured through Netlink (TC interface).
> >
> > This patch adds preliminary support to manage such flow rules through the
> > flow API (rte_flow).
> >
> > Instead of rewriting tons of Netlink helpers and as previously suggested by
> > Stephen [1], this patch introduces a new dependency to libmnl [2]
> > (LGPL-2.1) when compiling mlx5.
> >
> > [1] https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fmails.dpdk.org%2Farchives%2Fdev%2F2018-March%2F092676.html&data=02%7C01%7Cyskoh%40mellanox.com%7C1250093eca0c4ad6d9f008d5dc58fbb4%7Ca652971c7d2e4d9ba6a4d149256f461b%7C0%7C0%7C636657197116524482&sdata=JrAyzK1s3JG5CnuquNcA7XRN4d2WYtHUi1KXyloGdvA%3D&reserved=0
> > [2] https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fnetfilter.org%2Fprojects%2Flibmnl%2F&data=02%7C01%7Cyskoh%40mellanox.com%7C1250093eca0c4ad6d9f008d5dc58fbb4%7Ca652971c7d2e4d9ba6a4d149256f461b%7C0%7C0%7C636657197116524482&sdata=yLYa0NzsTyE62BHDCZDoDah31snt6w4Coq47pY913Oo%3D&reserved=0
> >
> > Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
<snip>
> > diff --git a/drivers/net/mlx5/mlx5_nl_flow.c b/drivers/net/mlx5/mlx5_nl_flow.c
> > new file mode 100644
> > index 000000000..7a8683b03
> > --- /dev/null
> > +++ b/drivers/net/mlx5/mlx5_nl_flow.c
> > @@ -0,0 +1,139 @@
> > +/* SPDX-License-Identifier: BSD-3-Clause
> > + * Copyright 2018 6WIND S.A.
> > + * Copyright 2018 Mellanox Technologies, Ltd
> > + */
> > +
> > +#include <errno.h>
> > +#include <libmnl/libmnl.h>
> > +#include <linux/netlink.h>
> > +#include <linux/pkt_sched.h>
> > +#include <linux/rtnetlink.h>
> > +#include <stdalign.h>
> > +#include <stddef.h>
> > +#include <stdint.h>
> > +#include <stdlib.h>
> > +#include <sys/socket.h>
> > +
> > +#include <rte_errno.h>
> > +#include <rte_flow.h>
> > +
> > +#include "mlx5.h"
> > +
> > +/**
> > + * Send Netlink message with acknowledgment.
> > + *
> > + * @param nl
> > + * Libmnl socket to use.
> > + * @param nlh
> > + * Message to send. This function always raises the NLM_F_ACK flag before
> > + * sending.
> > + *
> > + * @return
> > + * 0 on success, a negative errno value otherwise and rte_errno is set.
> > + */
> > +static int
> > +mlx5_nl_flow_nl_ack(struct mnl_socket *nl, struct nlmsghdr *nlh)
> > +{
> > + alignas(struct nlmsghdr)
> > + uint8_t ans[MNL_SOCKET_BUFFER_SIZE];
>
> There are total 3 of this buffer. On a certain host having large pagesize, this
> can be 8kB * 3 = 24kB. This is not a gigantic buffer but as all the functions
> here are sequentially accessed, how about having just one global buffer instead?
All right it's not ideal, I opted for simplicity though. This is a generic
ack function. When NETLINK_CAP_ACK is not supported (note: this was made
optional for v2, some systems do not support it), an ack consumes a bit more
space than the original message, which may itself be huge, and failure to
receive acks is deemed fatal.
Its callers are mlx5_nl_flow_init(), called once per device during
initialization, and mlx5_nl_flow_create/destroy(), called for each
created/removed flow rule.
These last two are called often but do not put their own buffer on the
stack, they reuse previously generated messages from the heap.
So to improve stack consumption a bit, what I can do is size this buffer
according to nlh->nlmsg_len + extra room for ack header, yet still allocate
it locally since it would be a pain otherwise. Callers may not want their
own buffers to be overwritten with useless acks.
--
Adrien Mazarguil
6WIND
next prev parent reply other threads:[~2018-07-12 10:46 UTC|newest]
Thread overview: 33+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-06-27 18:08 [dpdk-dev] [PATCH 0/6] net/mlx5: add support for switch flow rules Adrien Mazarguil
2018-06-27 18:08 ` [dpdk-dev] [PATCH 1/6] net/mlx5: lay groundwork for switch offloads Adrien Mazarguil
2018-07-12 0:17 ` Yongseok Koh
2018-07-12 10:46 ` Adrien Mazarguil [this message]
2018-07-12 17:33 ` Yongseok Koh
2018-06-27 18:08 ` [dpdk-dev] [PATCH 2/6] net/mlx5: add framework for switch flow rules Adrien Mazarguil
2018-07-12 0:59 ` Yongseok Koh
2018-07-12 10:46 ` Adrien Mazarguil
2018-07-12 18:25 ` Yongseok Koh
2018-06-27 18:08 ` [dpdk-dev] [PATCH 3/6] net/mlx5: add fate actions to " Adrien Mazarguil
2018-07-12 1:00 ` Yongseok Koh
2018-06-27 18:08 ` [dpdk-dev] [PATCH 4/6] net/mlx5: add L2-L4 pattern items " Adrien Mazarguil
2018-07-12 1:02 ` Yongseok Koh
2018-06-27 18:08 ` [dpdk-dev] [PATCH 5/6] net/mlx5: add VLAN item and actions " Adrien Mazarguil
2018-07-12 1:10 ` Yongseok Koh
2018-07-12 10:47 ` Adrien Mazarguil
2018-07-12 18:49 ` Yongseok Koh
2018-06-27 18:08 ` [dpdk-dev] [PATCH 6/6] net/mlx5: add port ID pattern item " Adrien Mazarguil
2018-07-12 1:13 ` Yongseok Koh
2018-06-28 9:05 ` [dpdk-dev] [PATCH 0/6] net/mlx5: add support for " Nélio Laranjeiro
2018-07-13 9:40 ` [dpdk-dev] [PATCH v2 " Adrien Mazarguil
2018-07-13 9:40 ` [dpdk-dev] [PATCH v2 1/6] net/mlx5: lay groundwork for switch offloads Adrien Mazarguil
2018-07-14 1:29 ` Yongseok Koh
2018-07-23 21:40 ` Ferruh Yigit
2018-07-24 0:50 ` Stephen Hemminger
2018-07-24 4:35 ` Shahaf Shuler
2018-07-24 19:33 ` Stephen Hemminger
2018-07-13 9:40 ` [dpdk-dev] [PATCH v2 2/6] net/mlx5: add framework for switch flow rules Adrien Mazarguil
2018-07-13 9:40 ` [dpdk-dev] [PATCH v2 3/6] net/mlx5: add fate actions to " Adrien Mazarguil
2018-07-13 9:40 ` [dpdk-dev] [PATCH v2 4/6] net/mlx5: add L2-L4 pattern items " Adrien Mazarguil
2018-07-13 9:40 ` [dpdk-dev] [PATCH v2 5/6] net/mlx5: add VLAN item and actions " Adrien Mazarguil
2018-07-13 9:40 ` [dpdk-dev] [PATCH v2 6/6] net/mlx5: add port ID pattern item " Adrien Mazarguil
2018-07-22 11:21 ` [dpdk-dev] [PATCH v2 0/6] net/mlx5: add support for " Shahaf Shuler
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=20180712104635.GS5211@6wind.com \
--to=adrien.mazarguil@6wind.com \
--cc=dev@dpdk.org \
--cc=nelio.laranjeiro@6wind.com \
--cc=shahafs@mellanox.com \
--cc=yskoh@mellanox.com \
/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;
as well as URLs for NNTP newsgroup(s).