netdev - Re: [PATCH RFC net-next] pktgen: introduce 'rx' mode

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: <1430350786.3711.77.camel@edumazet-glaptop2.roam.corp.google.com>
Date:	Wed, 29 Apr 2015 16:39:46 -0700
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	Alexei Starovoitov <ast@...mgrid.com>
Cc:	"David S. Miller" <davem@...emloft.net>,
	Eric Dumazet <edumazet@...gle.com>,
	Daniel Borkmann <daniel@...earbox.net>,
	Thomas Graf <tgraf@...g.ch>,
	Jamal Hadi Salim <jhs@...atatu.com>,
	John Fastabend <john.r.fastabend@...el.com>,
	netdev@...r.kernel.org
Subject: Re: [PATCH RFC net-next] pktgen: introduce 'rx' mode

On Wed, 2015-04-29 at 16:28 -0700, Alexei Starovoitov wrote:
> On 4/29/15 3:56 PM, Eric Dumazet wrote:
> >
> > So pktgen in RX mode MUST deliver skb with skb->users = 1, there is no
> > way around it.
> 
> if I only knew how to do it...
> The cost of continuously allocating skbs is way higher than
> netif_receive_skb itself. Such benchmarking tool would measure the
> speed of skb alloc/free instead of speed of netif_receive_skb.
> Are you suggesting to pre-allocate 10s of millions of skbs and
> then feed them in one go? The profile will be dominated by
> cache misses in the first few lines of __netif_receive_skb_core()
> where it accesses skb->dev,data,head. Doesn't sound too useful either.
> Other thoughts?

The code I copy pasted from your patch is buggy. Not the whole thing.

You have to replace it by something smarter.

This should not be hard really.

Zap the 'burst/clone' thing, this is not going to work for RX.
TX was okay, not RX. 

You could for instance do :

atomic_inc(&skb->users);
netif_receive_skb(skb);
if (atomic_read(skb->users) != 1) {
	/* This is too bad, I can not recycle this skb because it is still used */
	consume_skb(skb);
	/* allocate a fresh new skb */
	skb = ...
} else {
	/* Yeah ! Lets celebrate, cost of reusing this skb was one atomic op */
}





--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html