Re: [RFC Optimizing veth xsk performance 05/10] veth: use send queue tx napi to xmit xsk tx desc

[Simon Horman <horms@kernel.org>]

On Thu, Aug 03, 2023 at 10:04:31PM +0800, huangjie.albert wrote:

Please include a patch description.

> Signed-off-by: huangjie.albert <huangjie.albert@bytedance.com>

Please consider formatting this as:

	... Albert Huang <huangjie.albert@bytedance.com>

> ---
>  drivers/net/veth.c | 265 ++++++++++++++++++++++++++++++++++++++++++++-
>  1 file changed, 264 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/veth.c b/drivers/net/veth.c
> index 63c3ebe4c5d0..944761807ca4 100644
> --- a/drivers/net/veth.c
> +++ b/drivers/net/veth.c
> @@ -27,6 +27,8 @@
>  #include <linux/bpf_trace.h>
>  #include <linux/net_tstamp.h>
>  #include <net/page_pool.h>
> +#include <net/xdp_sock_drv.h>
> +#include <net/xdp.h>
>  
>  #define DRV_NAME	"veth"
>  #define DRV_VERSION	"1.0"

> @@ -1061,6 +1063,176 @@ static int veth_poll(struct napi_struct *napi, int budget)
>  	return done;
>  }
>  
> +static int veth_xsk_tx_xmit(struct veth_sq *sq, struct xsk_buff_pool *xsk_pool, int budget)
> +{
> +	struct veth_priv *priv, *peer_priv;
> +	struct net_device *dev, *peer_dev;
> +	struct veth_rq *peer_rq;
> +	struct veth_stats peer_stats = {};
> +	struct veth_stats stats = {};
> +	struct veth_xdp_tx_bq bq;
> +	struct xdp_desc desc;
> +	void *xdpf;
> +	int done = 0;

Please try to use reverse xmas tree ordering - longest line to shortest -
for local variable declarations in new Networking code.

https://github.com/ecree-solarflare/xmastree is your friend here.

> +
> +	bq.count = 0;
> +	dev = sq->dev;
> +	priv = netdev_priv(dev);
> +	peer_dev = priv->peer;

Sparse seems a bit unhappy about this.

  .../veth.c:1081:18: warning: incorrect type in assignment (different address spaces)
  .../veth.c:1081:18:    expected struct net_device *peer_dev
  .../veth.c:1081:18:    got struct net_device [noderef] __rcu *peer

Looking over existing code in this file, perhaps this is appropriate:

	peer_dev = rtnl_dereference(priv->peer);

Likewise in a few other places in this patch.

...