Michael Chan wrote:
On Mon, 2006-02-27 at 17:32 +0100, Dag Bakke wrote:
I have been playing with a couple of laptops with GbE interfaces, and
iperf. (Two Dell Latitude D600 connected back-to-back.)
[snip]
Is there anything I can do to increase the bandwidth further, or is this
as far as the
On Mon, Feb 27, Jesse Brandeburg wrote:
http://ozlabs.org/pipermail/linuxppc-dev/2006-February/021308.html
Olaf I reproduced this problem on a 630 and this patch fixes it for me,
please verify. Due to the new load ucode command the cb list must be
allocated before calling
Hello Jesse,
spddplx += (mii_reg 0x100)
-? FULL_DUPLEX :
-HALF_DUPLEX;
+? DUPLEX_FULL :
+
The following changes since commit 051d3cbd96909b2fe6b5038e7bbe77f41356db05:
David S. Miller:
[TG3]: Fix Sun tg3 variant detection.
are found in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-2.6.git
upstream-fixes
Pete Zaitcev:
This request is in addition to the request from 2/17, which does not
yet seem to be in the 'upstream' branch of netdev-2.6.
---
The following changes since commit acfaf10be5c19f7dceb9d7372039dc45af66b100:
John W. Linville:
Merge branch 'upstream-fixes'
are found in the git repository
As stated in a comment, the ipw2200 driver uses several routines that were borrowed from
ieee80211_geo.c. As ipw2200 requires ieee80211, these routines are duplicated. The attached patch,
which is sent as an attachment to preserve whitespace, converts ipw2200.c to use the ieee80211
versions,
On Tuesday 28 February 2006 02:40, John W. Linville wrote:
On Sun, Feb 12, 2006 at 12:00:11AM +0100, Ivo van Doorn wrote:
ieee80211_rx has been renamed __ieee80211_rx.
Use DRV_NAME as much as possible instead of a seperate name string.
Add new USB device ID.
Signed-off-by Ivo van Doorn
On Tue, Feb 28, 2006 at 09:21:39AM +0100, Dag Bakke wrote:
Got both. Max throughput increased from ~605 Mbps with carefully tuned
options to iperf, to ~650 Mbps with less carefully tuned options to iperf.
One other suggestion that might be worth testing: try comparing SMP and
UP kernels. I've
The patch below reworks socket.c to use fget_light() in place of fget()
for the various networking syscalls. This is of particular value with
SMP kernels running on the P4 as the memory barriers the atomic ops on
the file counter causes result in the CPU having to wait for quite a few
memory
What does your quick and dirty test program use for socket buffer sizes
and/or send sizes? What does the performance look like with a netperf
TCP_STREAM test of various socket buffer and send sizes?
The comparison might help find the extent to which application behaviour
matters.
rick
On Tue, 28 Feb 2006 09:48:28 -0600, Larry Finger [EMAIL PROTECTED] wrote:
As stated in a comment, the ipw2200 driver uses several routines that were
borrowed
from ieee80211_geo.c.
Signed-Off-By: Larry Finger [EMAIL PROTECTED]
Looks good to me, though I only made a cursory glance to verify
This trivial patch can go in the netdev can for 2.6.17.
It lets skge driver contribute to random entropy poll.
Signed-off-by: Stephen Hemminger [EMAIL PROTECTED]
--- sky2.orig/drivers/net/skge.c2006-02-27 10:00:48.0 -0800
+++ sky2/drivers/net/skge.c 2006-02-28
On Tue, 28 Feb 2006, Olaf Hering wrote:
On Mon, Feb 27, Jesse Brandeburg wrote:
http://ozlabs.org/pipermail/linuxppc-dev/2006-February/021308.html
Olaf I reproduced this problem on a 630 and this patch fixes it for me,
please verify. Due to the new load ucode command the cb list must be
On Tue, 28 Feb 2006 13:56:57 -0500
Jeff Garzik [EMAIL PROTECTED] wrote:
Stephen Hemminger wrote:
This trivial patch can go in the netdev can for 2.6.17.
It lets skge driver contribute to random entropy poll.
Signed-off-by: Stephen Hemminger [EMAIL PROTECTED]
---
The scope element in the ipv6_saddr_score struct used in
ipv6_dev_get_saddr() is an unsigned integer, but __ipv6_addr_src_scope()
returns a signed integer (and can return -1).
Signed-off-by: Brian Haley [EMAIL PROTECTED]
diff --git a/net/ipv6/addrconf.c b/net/ipv6/addrconf.c
---
From: Luiz Fernando Capitulino [EMAIL PROTECTED]
Date: Sat, 25 Feb 2006 16:07:34 -0300
As you've asked, I'm sending again my patches for pktgen:
[PATCH 1/6] pktgen: Lindent run.
[PATCH 2/6] pktgen: Ports thread list to Kernel list implementation.
[PATCH 3/6] pktgen: Fix kernel_thread()
On 2/28/06, Stephen Hemminger [EMAIL PROTECTED] wrote:
On Tue, 28 Feb 2006 14:29:30 -0500
Jeff Garzik [EMAIL PROTECTED] wrote:
Oh yeah, another aspect: cards with hardware interrupt mitigation
should NOT use SA_SAMPLE_RANDOM, for obvious reasons... Anything that
has the potential to have
Jesse Brandeburg wrote:
A recent patch attempted to enable more efficient memory usage by using
only 2kB descriptors for jumbo frames. The method used to implement
this has since been commented upon as illegal and in recent kernels
even causes a BUG when receiving ip fragments while using
Rick Jones wrote:
What does your quick and dirty test program use for socket buffer
sizes and/or send sizes? What does the performance look like with a
netperf TCP_STREAM test of various socket buffer and send sizes?
For my test program I used a 8K send and receive buffers. The socket
John Zielinski wrote:
Rick Jones wrote:
What does your quick and dirty test program use for socket buffer
sizes and/or send sizes? What does the performance look like with a
netperf TCP_STREAM test of various socket buffer and send sizes?
For my test program I used a 8K send and receive
Rick Jones wrote:
8KB socket buffers, 8KB writes to the socket or both?
8KB write to the socket with an 8KB read on the other end. TCP socket
buffers at default. I'm using netperf now since the numbers were about
the same and netperf has more options.
netperf -t TCP_STREAM -H remote --
Rick Jones wrote:
And if you add the test-specific -D option?
No difference. I have TCP_NODELAY in my Samba config and copying files
from the server is painfully slow. That's what got me started on all
these tests.
BTW, did someone already suggest disabling TSO if it happens to be
22 matches
Mail list logo
