Rick Jones wrote:
then the qdisc could/should place a cap on the size of a 'TSO' based on
the bitrate (and perhaps input as to how much time any one "burst" of
data should be allowed to consume on the network) and pass that up the
stack? right now you seem to be proposing what is effectively
Denys Fedoryshchenko wrote:
Hi
I took risk and installed ESFQ on my main backbone QoS. I found it highly
useful, and very need in setup's where is more than 128 flows passing and
especially where is nat available.
I agree it will be good when it's in.
Here is results with overloaded class
Jesper Dangaard Brouer wrote:
commit 07a74a2613440fc1a68d0faa7235ed7027532d78
Author: Jesper Dangaard Brouer <[EMAIL PROTECTED]>
Date: Tue Sep 11 16:59:58 2007 +0200
[IPROUTE2]: Overhead calculation is now done in the kernel.
The only current user is HTB. HTB overhead argument is
Patrick McHardy wrote:
My classifier uses jhash,
Ahh that's OK - I thought it still used the old sfq hash, which collided
alot with a low number of consecutive addresses.
Andy.
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
Mo
Patrick McHardy wrote:
One good thing about ESFQ is the more flexible flow classification, but
I don't like the concept of having a set of selectable hash functions
very much.
These patches change SFQ to allow attaching external classifiers and add
a new "flow" classifier that allows to classify
Adrian Bunk wrote:
Considering that NET_CLS_POLICE has been marked as obsolete for more
than one year, would a patch to remove it be acceptable?
cu
Adrian
People still request the "2.4 policer" on LARTC as unlike its
replacement it hooks after prerouting/denat so they can police on local
a
I Recently built a 2.6.17.7 and wanted to turn off CONFIG_NET_ESTIMATOR
but can't using menuconfig.
Is it on by default now, or is it a config issue?
I wanted it off to play with chains of policers and unless I
misunderstand it uses Hz, and is inaccurate when Hz=250 with its'
minimum time
jamal wrote:
About two more or so to complete these..
cheers,
jamal
+tc qdisc add dev lo eth0 ?
Andy.
-
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
Russell Stuart wrote:
On Tue, 2006-07-18 at 22:46 +0100, Andy Furniss wrote:
FWIW I think it may be possible to do it Patricks' way, as if I read it
properly he will end up with the ATM cell train length which gets
shifted by cell_log and looked up as before. The ATM length will be in
Russell Stuart wrote:
On Sat, 2006-06-24 at 10:13 -0400, jamal wrote:
And yes, I was arguing that the tc scheme you describe would not be so
bad either if the cost of making a generic change is expensive.
Patrick seems to have a simple way to compensate generically for link
layer fragmenta
jamal wrote:
I have taken linux-kernel off the list.
Russell's site is inaccessible to me (I actually think this is related
to some DNS issues i may be having) and your masters is too long to
spend 2 minutes and glean it; so heres a question or two for you:
- Have you tried to do a long-lived s
John Heffner wrote:
Andy Furniss wrote:
I've been doing some testing of my new wan connection and noticed that
when I specify a window with ip route it still changes after a while.
Looks like this is occurring in
net/ipv4/tcp_input.c:tcp_rcv_space_adjust().
The problem really is
I've been doing some testing of my new wan connection and noticed that
when I specify a window with ip route it still changes after a while.
Using 2.6.16.11 and latest iproute2.
ip ro del default
ip ro add default via 192.168.0.1 window 28000
ip ro ls
192.168.0.0/24 dev eth0 proto kernel sc
Roberto Nibali wrote:
I had the distinct pleasure of partly get involved with debugging
network stalls related to Linux clients (2.6.x kernel) and a Packeteer.
Dare I suggest that it could be something as trivial as it looks like
window scaling defaults to off on SunOS 2.5.1 and it's on on
Mark Butler wrote:
Andy Furniss wrote:
Mark Butler wrote:
There is no problem manipulating the TCP window per se. The problem
is advertising a window and then shrinking it faster than it is
naturally reduced by incoming data, essentially granting credit to
transmit x bytes, and then
Mark Butler wrote:
There is no problem manipulating the TCP window per se. The problem is
advertising a window and then shrinking it faster than it is naturally
reduced by incoming data, essentially granting credit to transmit x
bytes, and then revoking that credit. The net result is the peer
David S. Miller wrote:
From: "Michal Piotrowski" <[EMAIL PROTECTED]>
Date: Sun, 12 Mar 2006 02:51:40 +0100
I have noticed this warnings
TCP: Treason uncloaked! Peer 82.113.55.2:11759/50967 shrinks window
148470938:148470943. Repaired.
TCP: Treason uncloaked! Peer 82.113.55.2:11759/50967 shrink
Patrick McHardy wrote:
tc qdisc add dev ppp0 handle 1:0 root htb
tc class add dev ppp0 classid 1:1 htb rate 220kbit
tc filter add dev ppp0 protocol ip u32 match u32 0 0 classid 1:1
gives a 3 pkt queue I think each class gets 3 if there are more classes
That is because you use it on a ppp dev
Patrick McHardy wrote:
Andy Furniss wrote:
What do you think about making HTB hysteresis 0 and possibly set HZ to
1000 when HTB is selected?
I'm not qualified to judge about HTB hysteresis, but we can't change HZ,
most people just enable everything which would mean basically
Patrick McHardy wrote:
Sorry for the repost Dave, I accidentally used the @oss.sgi.com address
in my previous posting.
-
It seems to be a common mistake to use jiffies as clocksource,
which gives very bad results in most cases. This patch changes
the default to gettimeofday.
What do y
Andy Furniss wrote:
I'll
get everything upto date and try the new device tomorrow.
Just tried the example script with ifb on 2.6.15 and it works OK, will
try more complicated tests & iptables soon.
Andy.
-
To unsubscribe from this list: send the line "unsubscribe netdev&quo
jamal wrote:
On Fri, 2005-30-12 at 09:30 -0500, jamal wrote:
Ok, attached is what i tried - the script is not very realistic
but demoes that shaping will happen in the real device.
Ahh yes - I should have tried with the default class rather than
classifying. This behavior is fine and works
jamal wrote:
On Thu, 2005-29-12 at 17:02 +, Andy Furniss wrote:
Could you make it so that you can double queue the traffic - I couldn't
do this with dummy/htb (but didn't try that hard). I've read of imq
being used like this to enforce linkwide shaping policy above pe
jamal wrote:
I will think of something. But it seems like after talking to
other people as well, the consensus is agreeing with you. I will create
a new device instead of abusing dummy ;->
Could you make it so that you can double queue the traffic - I couldn't
do this with dummy/htb (but
Daniel J Blueman wrote:
Has anyone been able to get the RED (random early detection) qdisc
working lately?
I can't get anything going through it to be dropped or marked; the
'marked', 'early', 'pdrop' and 'other' fields remain at 0 [1]. In my
example script [2], I get the 3072Kbits/s transfer in
Andy Furniss wrote:
Hmm - I did some testing on old kernels, though I know I sorted it to
2ms I haven't got that kernel anymore and setting a vanilla 2.4.26 to hz
500 made the forwarded traffic cycle with 5ms not 2ms, local doesn't
cycle but isn't constant and gives 5ms vari
Andy Furniss wrote:
You are comparing two completely different ->systems<-. Please stop being
focused on HZ, ok ? HZ does not rule the priority of the in-kernel tasks.
OK - It was just that changing HZ sorted it on 2.4.
Hmm - I did some testing on old kernels, though I know I sorted
Francois Romieu wrote:
Andy Furniss <[EMAIL PROTECTED]> :
[...]
They will be the same as connexent haven't touched the drivers for
years, Patrick ported them to 2.6 and a diff doesn't show any difference
involving HZ that I can see.
You are comparing two completely dif
Francois Romieu wrote:
Andy Furniss <[EMAIL PROTECTED]> :
[ping latencies fluctuates with conexant adsl driver]
Does anyone know what could be causing this.
You will have a hard time figuring it and many people (me included)
will keep a good distance with an issue which includes >
I long ago noticed that if I ping somewhere on the internet for any
length of time I see a pattern - the ping times will slowly rise then
fall by 10ms and repeat the process.
I never asked before because my dsl modem taints the kernel - but I
think testing I did on 2.4s could mean that it's no
30 matches
Mail list logo
