netdev - Re: [PATCH v2] pktgen: create packet use IPv6 source address between src6_min and src6

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] [day] [month] [year] [list]

Message-ID: <20200113045136.4d59dded@cakuba>
Date:   Mon, 13 Jan 2020 04:51:36 -0800
From:   Jakub Kicinski <kubakici@...pl>
To:     Niu Xilei <niu_xilei@....com>
Cc:     davem@...emloft.net, tglx@...utronix.de, fw@...len.de,
        peterz@...radead.org, steffen.klassert@...unet.com,
        bigeasy@...utronix.de, jonathan.lemon@...il.com, pabeni@...hat.com,
        anshuman.khandual@....com, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] pktgen: create packet use IPv6 source address
 between src6_min and src6_max.

The subject line could use some rewording, I think.

On Mon, 13 Jan 2020 19:21:02 +0800, Niu Xilei wrote:
> Pktgen can use only one IPv6 source address from output device, or src6 command
> setting. In pressure test we need create lots of session more than 65536.If
> IPSRC_RND flag is set, generate random source address between src6_min and src6_max.
> 
> Signed-off-by: Niu Xilei <niu_xilei@....com>
> 
> Changes since v1:
>  - only create IPv6 source address over least significant 64 bit range

> +/* generate ipv6 source addr */
> +static inline void set_src_in6_addr(struct pktgen_dev *pkt_dev)

Please just use "static" instead of "static inline". The compiler will
be clever enough to decide the inlining.

> +{
> +	__be64 min6_h, min6_l, max6_h, max6_l, addr_l, *t;
> +	u64 min6, max6, rand, i;
> +	struct in6_addr addr6;
> +
> +	memcpy(&min6_h, pkt_dev->min_in6_saddr.s6_addr, 8);
> +	memcpy(&min6_l, pkt_dev->min_in6_saddr.s6_addr + 8, 8);
> +	memcpy(&max6_h, pkt_dev->max_in6_saddr.s6_addr, 8);
> +	memcpy(&max6_l, pkt_dev->max_in6_saddr.s6_addr + 8, 8);
> +
> +	/* only generate source address in least significant 64 bits range
> +	 * most significant 64 bits must be equal
> +	 */
> +	if (max6_h != min6_h)
> +		return;
> +
> +	addr6 = pkt_dev->min_in6_saddr;
> +	t = (__be64 *)addr6.s6_addr + 1;
> +	min6 = be64_to_cpu(min6_l);
> +	max6 = be64_to_cpu(max6_l);
> +
> +	if (min6 < max6) {

Since this code is executed on every packet, would it be possible to
pre-compute the decision if the IPv6 address is to be generated or not?
We have 4 memcpy()s and 2 byte swaps, and the conditions never change,
so it could be computed at setup time, right?

> +		if (pkt_dev->flags & F_IPSRC_RND) {
> +			do {
> +				prandom_bytes(&rand, sizeof(rand));
> +				rand = rand % (max6 - min6) + min6;
> +				addr_l = cpu_to_be64(rand);
> +				memcpy(t, &addr_l, 8);
> +			} while (ipv6_addr_loopback(&addr6) ||
> +				 ipv6_addr_v4mapped(&addr6) ||
> +				 ipv6_addr_is_multicast(&addr6));
> +		} else {
> +			addr6 = pkt_dev->cur_in6_saddr;
> +			i = be64_to_cpu(*t);
> +			if (++i > max6)
> +				i = min6;
> +			addr_l = cpu_to_be64(i);
> +			memcpy(t, &addr_l, 8);
> +		}
> +	}
> +	pkt_dev->cur_in6_saddr = addr6;
> +}