[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 10 Oct 2014 16:30:50 +0000
From: David Laight <David.Laight@...LAB.COM>
To: 'Eric Dumazet' <eric.dumazet@...il.com>
CC: "'alexander.h.duyck@...hat.com'" <alexander.h.duyck@...hat.com>,
"David Miller" <davem@...emloft.net>,
"alexander.duyck@...il.com" <alexander.duyck@...il.com>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: RE: [PATCH] flow-dissector: Fix alignment issue in
__skb_flow_get_ports
From: Eric Dumazet [mailto:eric.dumazet@...il.com]
> Sent: 10 October 2014 16:34
> To: David Laight
> Cc: 'alexander.h.duyck@...hat.com'; David Miller; alexander.duyck@...il.com; netdev@...r.kernel.org
> Subject: Re: [PATCH] flow-dissector: Fix alignment issue in __skb_flow_get_ports
>
> On Fri, 2014-10-10 at 14:57 +0000, David Laight wrote:
>
> > I think there is code to copy the IP and TCP headers to aligned memory
> > before they are parsed.
>
> There is no such thing. You are here on netdev list, please read the
> code before doing such claims.
I did say 'I think'...
I must be thinking of some similar code somewhere else.
Possibly just the code that ensures the header isn't fragmented.
> > > The problem is the igb / ixgbe / fm10k hardware doesn't have a means of
> > > inserting padding from its side...
> >
> > Shoot the hardware engineers.
> >
> > You aren't going to get the performance you expect from a 10Ge card
> > unless the rx buffers are 'correctly' aligned.
>
> That is simply not true on current x86 cpus. They simply dont care at
> all.
I was referring to using them on sparc64, not x86.
I know that current intel x86 cpu have support for misaligned 'rep movsd',
but I thought there was still a small cost (maybe one clock) for
single word transfers.
So maybe they care 'just a little bit'.
> You cannot blame Intel for other arches.
True, but this does mean that you don't really want to use these adapters
on a system that can't to unaligned accesses.
David
Powered by blists - more mailing lists