netdev - Re: [PATCH bpf-next 4/6] selftests/xsk: implement set_hw_ring_size function to configure interface ring size

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: <ZfSI8UftKDGWTgUC@google.com>
Date: Fri, 15 Mar 2024 10:44:17 -0700
From: Stanislav Fomichev <sdf@...gle.com>
To: Tushar Vyavahare <tushar.vyavahare@...el.com>
Cc: bpf@...r.kernel.org, netdev@...r.kernel.org, bjorn@...nel.org, 
	magnus.karlsson@...el.com, maciej.fijalkowski@...el.com, 
	jonathan.lemon@...il.com, davem@...emloft.net, kuba@...nel.org, 
	pabeni@...hat.com, ast@...nel.org, daniel@...earbox.net, 
	tirthendu.sarkar@...el.com
Subject: Re: [PATCH bpf-next 4/6] selftests/xsk: implement set_hw_ring_size
 function to configure interface ring size

On 03/15, Tushar Vyavahare wrote:
> Introduce a new function called set_hw_ring_size that allows for the
> dynamic configuration of the ring size within the interface.
> 
> Signed-off-by: Tushar Vyavahare <tushar.vyavahare@...el.com>
> ---
>  tools/testing/selftests/bpf/xskxceiver.c | 35 ++++++++++++++++++++++++
>  1 file changed, 35 insertions(+)
> 
> diff --git a/tools/testing/selftests/bpf/xskxceiver.c b/tools/testing/selftests/bpf/xskxceiver.c
> index 32005bfb9c9f..aafa78307586 100644
> --- a/tools/testing/selftests/bpf/xskxceiver.c
> +++ b/tools/testing/selftests/bpf/xskxceiver.c
> @@ -441,6 +441,41 @@ static int get_hw_ring_size(struct ifobject *ifobj)
>  	return 0;
>  }
>  
> +static int set_hw_ring_size(struct ifobject *ifobj, u32 tx, u32 rx)

Hmm, now that we have set/get, should we put them into
network_helpers.c? These seem pretty generic if you accept
iface_name + ethtool_ringparam in the api.

> +{
> +	struct ethtool_ringparam ring_param = {0};
> +	struct ifreq ifr = {0};
> +	int sockfd, ret;
> +	u32 ctr = 0;
> +
> +	sockfd = socket(AF_INET, SOCK_DGRAM, 0);
> +	if (sockfd < 0)
> +		return errno;
> +
> +	memcpy(ifr.ifr_name, ifobj->ifname, sizeof(ifr.ifr_name));
> +
> +	ring_param.tx_pending = tx;
> +	ring_param.rx_pending = rx;
> +
> +	ring_param.cmd = ETHTOOL_SRINGPARAM;
> +	ifr.ifr_data = (char *)&ring_param;
> +

[..]

> +	while (ctr++ < SOCK_RECONF_CTR) {

Is it to retry EINTR? Retrying something else doesn't make sense
probably? So maybe do only errno==EINTR cases? Will make it more
generic and not dependent on SOCK_RECONF_CTR.


> +		ret = ioctl(sockfd, SIOCETHTOOL, &ifr);
> +		if (!ret)
> +			break;
> +		/* Retry if it fails */
> +		if (ctr >= SOCK_RECONF_CTR) {
> +			close(sockfd);
> +			return errno;
> +		}

[..]

> +		usleep(USLEEP_MAX);

Same here. Not sure what's the purpose of sleep? Alternatively, maybe
clarify in the commit description what particular error case we want
to retry.