lkml.org 
[lkml]   [2002]   [May]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: khttpd rotten?
From
   From: Andi Kleen <ak@muc.de>
Date: 10 May 2002 12:20:17 +0200

It is basically impossible without hardware support/major stack changes to
do it securely for networking RX. The problem is that to put data directly
from the NIC into an user space address range the NIC already needs
to demultiplex the sockets before it starts to DMA.

Sun's Cassini gigabit chips have the infrastructure. It's a very
simple onboard header parser (it's not a real CPU, just some thick
ASIC logic). You write small header parsing scripts plus provide it
with a socket cache table, it accumulates data into pages from the
streams for you.

Of course all of this is moot until we can actually write a driver
for one of these cards with such features (I'm working on it for
Cassini and things look good right now).

For userspace you would need to put it into an page aligned buffer and
change the page tables of the user space process, but especially on SMP
that is usually slower than just copying. Managing the stuff in a shared
memory segment is also likely not secure

Alexey already did hacks to make this work on xmit, we have all of the
infrastructure really due to the direct I/O bits.

[There is a special hack in the stack to do it for network sniffing,
but it is not general purpose enough for real protocols]

RX TCP is very doable with the correct hardware. I already have the
implementation in my head given something like Cassini HW.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:26    [W:0.064 / U:2.232 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site