netdev - Re: tbench wrt. loopback TSO

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]

Date:	Wed, 5 Nov 2008 14:25:57 +0200 (EET)
From:	"Ilpo Järvinen" <ilpo.jarvinen@...sinki.fi>
To:	David Miller <davem@...emloft.net>
cc:	zbr@...emap.net, Netdev <netdev@...r.kernel.org>, efault@....de,
	mingo@...e.hu, a.p.zijlstra@...llo.nl,
	Herbert Xu <herbert@...dor.apana.org.au>
Subject: Re: tbench wrt. loopback TSO

On Wed, 5 Nov 2008, David Miller wrote:

> From: "Ilpo Järvinen" <ilpo.jarvinen@...sinki.fi>
> Date: Wed, 5 Nov 2008 14:04:14 +0200 (EET)
> 
> > On Wed, 5 Nov 2008, David Miller wrote:
> > 
> > > From: Evgeniy Polyakov <zbr@...emap.net>
> > > Date: Wed, 5 Nov 2008 14:49:00 +0300
> > > 
> > > > But what if we just remove that trimming at all? This may result in
> > > > a not fully filled tso frame, but who cares if we copy data from
> > > > userspace anyway, will add another potentially small copy at sending
> > > > time?
> > > 
> > > Ok, the only part I care about is whether taking away this modulus
> > > will result in a sub-MSS sized packet on the wire, which is
> > > undesirable.
> > >
> > > The send logic will prevent this, right?
> > 
> > I don't see what currently would do that. We'd need to resegment skbs 
> > there then...
> 
> Look at tcp_mss_split_point() in tcp_output.c
> 
> I think it ends up doing exactly this.
> 
> 'needed' is set to min(skb->len, window)
> 
> and the return value is "needed - (needed % mss_now)"
> 
> The caller splits up the TSO segment as needed to satisfy this
> returned limit.
> 
> That should effectively chop off the sub-mss part.

Yes yes, I know as I wrote that part :-). But how does it prevent sub-MSS 
on wire?

Answer: It is intented for different case and doesn't prevent sub-MSS 
segments but solves some window limitation issues! It just makes sure
we won't add yet another sub-mss segment because of splitting we anyway 
are forced to.

In case of curr_mss != write_time_mss, we'll be sending those < mss 
segments per each skb, normally this happens over a single rtt so it's not 
that bad problem (and btw, that modulo won't even be executed if you have 
enough window but the splitting happens on a layer below, except for the 
last segment of course where we'd do nagle).

The problem is that we'd need to _resegment with the next skb_ since the 
mss boundary and skb boundary would basically constantly be running 
out-of-sync. That won't get done currently by anything.

-- 
 i.