lkml.org 
[lkml]   [2000]   [Feb]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: Gigabit Linux Server Bottlenecks
Jes.Sorensen@cern.ch said:
> >>>>> "Alexey" == kuznet <kuznet@ms2.inr.ac.ru> writes:
>
> >> There is no point in using it for transmit since the Linux TCP
> >> stack currently doesn't support this.
>
> Alexey> Hmm... Are you sure? Until now I thought exaclty the same with
> Alexey> opposite sign. 8) Namely: there are no reasons to awake the
> Alexey> feature in TCP, because no drivers support this.
>
> Well in order to use it and really gain on it for transmit, you want
> scatter/gather skbuffs, something which I really want to look at
> (provided I can find the time). If you are doing the data copy anyway,
> there checksum is almost free.

Agreed. Back in the early 2.2 days I changed the kernel in a generic
way not to checksum outgoing TCP/UDP packets if it knew the card could
handle it, and experimented using a Packet Engines Hamachi gigE. The
changes to the stack are fairly small, and the gains are near
negligible (max 5% faster) since a checksum while copying is indeed
free.

The curious might see http://dancer.ca.sandia.gov/~pw, but I'd much
rather see the card DMA the unchecksummed payload directly from
userspace.

-- Pete

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:56    [W:0.130 / U:0.600 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site