All of lore.kernel.org
 help / color / mirror / Atom feed
From: Magnus Karlsson <magnus.karlsson@gmail.com>
To: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Cc: "Network Development" <netdev@vger.kernel.org>,
	"Michael S. Tsirkin" <mst@redhat.com>,
	"Jason Wang" <jasowang@redhat.com>,
	"David S. Miller" <davem@davemloft.net>,
	"Jakub Kicinski" <kuba@kernel.org>,
	"Björn Töpel" <bjorn@kernel.org>,
	"Magnus Karlsson" <magnus.karlsson@intel.com>,
	"Jonathan Lemon" <jonathan.lemon@gmail.com>,
	"Alexei Starovoitov" <ast@kernel.org>,
	"Daniel Borkmann" <daniel@iogearbox.net>,
	"Jesper Dangaard Brouer" <hawk@kernel.org>,
	"John Fastabend" <john.fastabend@gmail.com>,
	virtualization@lists.linux-foundation.org,
	bpf <bpf@vger.kernel.org>,
	"dust . li" <dust.li@linux.alibaba.com>
Subject: Re: [PATCH net-next v4 08/10] virtio-net: xsk zero copy xmit setup
Date: Wed, 14 Apr 2021 09:36:15 +0200	[thread overview]
Message-ID: <CAJ8uoz2LzrvZUsDfFuKiFkyRwdWtEk8AF9y7Nb6RKzB7pO3YDw@mail.gmail.com> (raw)
In-Reply-To: <20210413031523.73507-9-xuanzhuo@linux.alibaba.com>

On Tue, Apr 13, 2021 at 9:58 AM Xuan Zhuo <xuanzhuo@linux.alibaba.com> wrote:
>
> xsk is a high-performance packet receiving and sending technology.
>
> This patch implements the binding and unbinding operations of xsk and
> the virtio-net queue for xsk zero copy xmit.
>
> The xsk zero copy xmit depends on tx napi. So if tx napi is not true,
> an error will be reported. And the entire operation is under the
> protection of rtnl_lock.
>
> If xsk is active, it will prevent ethtool from modifying tx napi.
>
> Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
> Reviewed-by: Dust Li <dust.li@linux.alibaba.com>
> ---
>  drivers/net/virtio_net.c | 78 +++++++++++++++++++++++++++++++++++++++-
>  1 file changed, 77 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index f52a25091322..8242a9e9f17d 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -22,6 +22,7 @@
>  #include <net/route.h>
>  #include <net/xdp.h>
>  #include <net/net_failover.h>
> +#include <net/xdp_sock_drv.h>
>
>  static int napi_weight = NAPI_POLL_WEIGHT;
>  module_param(napi_weight, int, 0444);
> @@ -133,6 +134,11 @@ struct send_queue {
>         struct virtnet_sq_stats stats;
>
>         struct napi_struct napi;
> +
> +       struct {
> +               /* xsk pool */
> +               struct xsk_buff_pool __rcu *pool;
> +       } xsk;
>  };
>
>  /* Internal representation of a receive virtqueue */
> @@ -2249,8 +2255,19 @@ static int virtnet_set_coalesce(struct net_device *dev,
>         if (napi_weight ^ vi->sq[0].napi.weight) {
>                 if (dev->flags & IFF_UP)
>                         return -EBUSY;
> -               for (i = 0; i < vi->max_queue_pairs; i++)
> +               for (i = 0; i < vi->max_queue_pairs; i++) {
> +                       /* xsk xmit depend on the tx napi. So if xsk is active,
> +                        * prevent modifications to tx napi.
> +                        */
> +                       rcu_read_lock();
> +                       if (rcu_dereference(vi->sq[i].xsk.pool)) {
> +                               rcu_read_unlock();
> +                               continue;
> +                       }
> +                       rcu_read_unlock();
> +
>                         vi->sq[i].napi.weight = napi_weight;
> +               }
>         }
>
>         return 0;
> @@ -2518,11 +2535,70 @@ static int virtnet_xdp_set(struct net_device *dev, struct bpf_prog *prog,
>         return err;
>  }
>
> +static int virtnet_xsk_pool_enable(struct net_device *dev,
> +                                  struct xsk_buff_pool *pool,
> +                                  u16 qid)
> +{
> +       struct virtnet_info *vi = netdev_priv(dev);
> +       struct send_queue *sq;
> +
> +       if (qid >= vi->curr_queue_pairs)
> +               return -EINVAL;

Your implementation is the first implementation that only supports
zerocopy for one out of Rx and Tx, and this will currently confuse the
control plane in some situations since it assumes that both Rx and Tx
are enabled by a call to this NDO. For example: user space creates an
xsk socket with both an Rx and a Tx ring, then calls bind with the
XDP_ZEROCOPY flag. In this case, the call should fail if the device is
virtio-net since it only supports zerocopy for Tx. But it should
succeed if the user only created a Tx ring since that makes it a
Tx-only socket which can be supported.

So you need to introduce a new interface in xdp_sock_drv.h that can be
used to ask if this socket has Rx enabled and if so fail the call (at
least one of them has to be enabled, otherwise the bind call would
fail before this ndo is called). Then the logic above will act on that
and try to fall back to copy mode (if allowed). Such an interface
(with an added "is_tx_enabled") might in the future be useful for
physical NIC drivers too if they would like to save on resources for
Tx-only and Rx-only sockets. Currently, they all just assume every
socket is Rx and Tx.

Thanks: Magnus

> +
> +       sq = &vi->sq[qid];
> +
> +       /* xsk zerocopy depend on the tx napi.
> +        *
> +        * xsk zerocopy xmit is driven by the tx interrupt. When the device is
> +        * not busy, napi will be called continuously to send data. When the
> +        * device is busy, wait for the notification interrupt after the
> +        * hardware has finished processing the data, and continue to send data
> +        * in napi.
> +        */
> +       if (!sq->napi.weight)
> +               return -EPERM;
> +
> +       rcu_read_lock();
> +       /* Here is already protected by rtnl_lock, so rcu_assign_pointer is
> +        * safe.
> +        */
> +       rcu_assign_pointer(sq->xsk.pool, pool);
> +       rcu_read_unlock();
> +
> +       return 0;
> +}
> +
> +static int virtnet_xsk_pool_disable(struct net_device *dev, u16 qid)
> +{
> +       struct virtnet_info *vi = netdev_priv(dev);
> +       struct send_queue *sq;
> +
> +       if (qid >= vi->curr_queue_pairs)
> +               return -EINVAL;
> +
> +       sq = &vi->sq[qid];
> +
> +       /* Here is already protected by rtnl_lock, so rcu_assign_pointer is
> +        * safe.
> +        */
> +       rcu_assign_pointer(sq->xsk.pool, NULL);
> +
> +       synchronize_net(); /* Sync with the XSK wakeup and with NAPI. */
> +
> +       return 0;
> +}
> +
>  static int virtnet_xdp(struct net_device *dev, struct netdev_bpf *xdp)
>  {
>         switch (xdp->command) {
>         case XDP_SETUP_PROG:
>                 return virtnet_xdp_set(dev, xdp->prog, xdp->extack);
> +       case XDP_SETUP_XSK_POOL:
> +               if (xdp->xsk.pool)
> +                       return virtnet_xsk_pool_enable(dev, xdp->xsk.pool,
> +                                                      xdp->xsk.queue_id);
> +               else
> +                       return virtnet_xsk_pool_disable(dev, xdp->xsk.queue_id);
>         default:
>                 return -EINVAL;
>         }
> --
> 2.31.0
>

  parent reply	other threads:[~2021-04-14  7:36 UTC|newest]

Thread overview: 31+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-04-13  3:15 [PATCH net-next v4 00/10] virtio-net support xdp socket zero copy xmit Xuan Zhuo
2021-04-13  3:15 ` [PATCH net-next v4 01/10] netdevice: priv_flags extend to 64bit Xuan Zhuo
2021-04-13  5:19   ` kernel test robot
2021-04-13  5:19     ` kernel test robot
2021-04-13  6:30   ` kernel test robot
2021-04-13  6:30     ` kernel test robot
2021-04-13 18:26   ` Jakub Kicinski
2021-04-13  3:15 ` [PATCH net-next v4 02/10] netdevice: add priv_flags IFF_NOT_USE_DMA_ADDR Xuan Zhuo
2021-04-13  3:15 ` [PATCH net-next v4 03/10] virtio-net: " Xuan Zhuo
2021-04-13  3:15 ` [PATCH net-next v4 04/10] xsk: support get page by addr Xuan Zhuo
2021-04-14  6:59   ` Magnus Karlsson
2021-04-13  3:15 ` [PATCH net-next v4 05/10] xsk: XDP_SETUP_XSK_POOL support option IFF_NOT_USE_DMA_ADDR Xuan Zhuo
2021-04-14  7:00   ` Magnus Karlsson
2021-04-13  3:15 ` [PATCH net-next v4 06/10] virtio-net: unify the code for recycling the xmit ptr Xuan Zhuo
2021-04-14  3:32   ` Jason Wang
2021-04-14  3:32     ` Jason Wang
2021-04-13  3:15 ` [PATCH net-next v4 07/10] virtio-net: virtnet_poll_tx support budget check Xuan Zhuo
2021-04-14  3:34   ` Jason Wang
2021-04-14  3:34     ` Jason Wang
2021-04-13  3:15 ` [PATCH net-next v4 08/10] virtio-net: xsk zero copy xmit setup Xuan Zhuo
2021-04-14  4:01   ` Jason Wang
2021-04-14  4:01     ` Jason Wang
2021-04-14  7:36   ` Magnus Karlsson [this message]
2021-04-14  7:51     ` Jason Wang
2021-04-14  7:51       ` Jason Wang
2021-04-13  3:15 ` [PATCH net-next v4 09/10] virtio-net: xsk zero copy xmit implement wakeup and xmit Xuan Zhuo
2021-04-14  5:46   ` Jason Wang
2021-04-14  5:46     ` Jason Wang
2021-04-13  3:15 ` [PATCH net-next v4 10/10] virtio-net: xsk zero copy xmit kick by threshold Xuan Zhuo
2021-04-14  5:51   ` Jason Wang
2021-04-14  5:51     ` Jason Wang

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=CAJ8uoz2LzrvZUsDfFuKiFkyRwdWtEk8AF9y7Nb6RKzB7pO3YDw@mail.gmail.com \
    --to=magnus.karlsson@gmail.com \
    --cc=ast@kernel.org \
    --cc=bjorn@kernel.org \
    --cc=bpf@vger.kernel.org \
    --cc=daniel@iogearbox.net \
    --cc=davem@davemloft.net \
    --cc=dust.li@linux.alibaba.com \
    --cc=hawk@kernel.org \
    --cc=jasowang@redhat.com \
    --cc=john.fastabend@gmail.com \
    --cc=jonathan.lemon@gmail.com \
    --cc=kuba@kernel.org \
    --cc=magnus.karlsson@intel.com \
    --cc=mst@redhat.com \
    --cc=netdev@vger.kernel.org \
    --cc=virtualization@lists.linux-foundation.org \
    --cc=xuanzhuo@linux.alibaba.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.