Jesse Brandeburg wrote:
On 11/9/05, Ben Greear <[EMAIL PROTECTED]> wrote:
I am frusterated that the linux kernel seems to be
unstable for high levels of TCP traffic for very common
hardware (e1000).
Is anyone doing tests that involve high levels of bi-directional
TCP traffic using TSO? If so, please let me know your results.
It's always possible I am screwing something up..
Ben, as for your test I think something is messed up in your patches,
as no one else seems to be reporting your freezes.
I will run some tests on standard kernels if I can still reproduce on
2.6.14. If it is bugs in my patches, it's subtle..since turning off
TSO fixes it.
as one of e1000's maintainers I can clearly state we and our customers
want TSO to work correctly. I appreciate all of David and the
community's work to get this feature finally figured out. Its close
but definitely doesn't work completely.
As for our testing lab's test passes, they are all run in the default
config, which means that TSO is on for all of our qualification tests.
The problem in the past (and still some now) is that TSO doesn't get
tested as much as we think it might be because it almost always gets
turned off due to SACKs and/or connection timeouts on long lived
connections (more than 10 seconds :-) ). I for one would love to
figure out why when e1000 shows no drops at all from hw perspective at
either end of a multi-stream test, and yet the stack for some reason
is dropping packets somewhere in the middle of the stack.
When you run your tests, do you ever do tests with sending significant traffic
both directions on a single socket?
Shall I follow up with some data? I can reproduce this at will with
netperf and a couple of scripts.
Ben i dont want to steal your thread, so lets start another one to
talk about TSO and drops if someone will follow up with me.
I think my part of this thread is on hold untill I manage to try
out 2.6.14, so please feel free to take it over :)
Ben
--
Ben Greear <[EMAIL PROTECTED]>
Candela Technologies Inc http://www.candelatech.com
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
