netdev - Re: [PATCH net-next 3/3] virtio-net: reduce the CPU consumption of dim worker

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]

Message-ID: <20240116195646.GF588419@kernel.org>
Date: Tue, 16 Jan 2024 19:56:46 +0000
From: Simon Horman <horms@...nel.org>
To: Heng Qi <hengqi@...ux.alibaba.com>
Cc: netdev@...r.kernel.org, virtualization@...ts.linux.dev,
	Jason Wang <jasowang@...hat.com>,
	"Michael S. Tsirkin" <mst@...hat.com>,
	Paolo Abeni <pabeni@...hat.com>, Jakub Kicinski <kuba@...nel.org>,
	Eric Dumazet <edumazet@...gle.com>,
	"David S. Miller" <davem@...emloft.net>,
	Xuan Zhuo <xuanzhuo@...ux.alibaba.com>
Subject: Re: [PATCH net-next 3/3] virtio-net: reduce the CPU consumption of
 dim worker

On Tue, Jan 16, 2024 at 09:11:33PM +0800, Heng Qi wrote:
> Accumulate multiple request commands to kick the device once,
> and obtain the processing results of the corresponding commands
> asynchronously. The batch command method is used to optimize the
> CPU overhead of the DIM worker caused by the guest being busy
> waiting for the command response result.
> 
> On an 8-queue device, without this patch, the guest cpu overhead
> due to waiting for cvq could be 10+% and above. With this patch,
> the corresponding overhead is basically invisible.
> 
> Signed-off-by: Heng Qi <hengqi@...ux.alibaba.com>

...

> @@ -3520,22 +3568,99 @@ static int virtnet_send_notf_coal_vq_cmds(struct virtnet_info *vi,
>  	return 0;
>  }
>  
> +static bool virtnet_add_dim_command(struct virtnet_info *vi,
> +				    struct virtnet_batch_coal *ctrl)
> +{
> +	struct scatterlist *sgs[4], hdr, stat, out;
> +	unsigned out_num = 0;
> +	int ret;
> +
> +	/* Caller should know better */
> +	BUG_ON(!virtio_has_feature(vi->vdev, VIRTIO_NET_F_CTRL_VQ));
> +
> +	ctrl->hdr.class = VIRTIO_NET_CTRL_NOTF_COAL;
> +	ctrl->hdr.cmd = VIRTIO_NET_CTRL_NOTF_COAL_VQS_SET;
> +
> +	/* Add header */
> +	sg_init_one(&hdr, &ctrl->hdr, sizeof(ctrl->hdr));
> +	sgs[out_num++] = &hdr;
> +
> +	/* Add body */
> +	sg_init_one(&out, &ctrl->num_entries, sizeof(ctrl->num_entries) +
> +		    ctrl->num_entries * sizeof(struct virtnet_coal_entry));

Hi Heng Qi,

I am a bit confused.
With this series applied on top of net-next
struct virtnet_coal_entry doesn't seem to exist.


> +	sgs[out_num++] = &out;
> +
> +	/* Add return status. */
> +	ctrl->status = VIRTIO_NET_OK;
> +	sg_init_one(&stat, &ctrl->status, sizeof(ctrl->status));
> +	sgs[out_num] = &stat;
> +
> +	BUG_ON(out_num + 1 > ARRAY_SIZE(sgs));
> +	ret = virtqueue_add_sgs(vi->cvq, sgs, out_num, 1, ctrl, GFP_ATOMIC);
> +	if (ret < 0) {
> +		dev_warn(&vi->vdev->dev, "Failed to add sgs for command vq: %d\n.", ret);
> +		return false;
> +	}
> +
> +	virtqueue_kick(vi->cvq);
> +
> +	ctrl->usable = false;
> +	vi->cvq_cmd_nums++;
> +
> +	return true;
> +}

...