netdev
[Top] [All Lists]

Re: [PATCH] PKT_SCHED: Provide compat policer stats in action policer

To: hadi@xxxxxxxxxx
Subject: Re: [PATCH] PKT_SCHED: Provide compat policer stats in action policer
From: Patrick McHardy <kaber@xxxxxxxxx>
Date: Tue, 21 Dec 2004 11:11:48 +0100
Cc: Thomas Graf <tgraf@xxxxxxx>, "David S. Miller" <davem@xxxxxxxxxxxxx>, netdev@xxxxxxxxxxx
In-reply-to: <1103552830.1049.355.camel@jzny.localdomain>
References: <20041215130128.GK8493@postel.suug.ch> <1103119774.1077.74.camel@jzny.localdomain> <41C05B60.6030504@trash.net> <1103484249.1046.143.camel@jzny.localdomain> <41C6A6CC.1050105@trash.net> <1103552830.1049.355.camel@jzny.localdomain>
Sender: netdev-bounce@xxxxxxxxxxx
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040413 Debian/1.6-5
jamal wrote:

I havent looked closely at tcng although Werner has showed it to me a
few times (may be under influence). We need to pick one or other test
setup. I dont care if its what I have, tcng or what Thomas has.
I just stared quickly at what Thomas has and realize its not really
automated. In my case it is easier because i can click on the proverbial
one-button and run 20 tests (including a subset of the policer ones)
and even capturing tcpdumps. I have attached a sample testcase. They are harder to create and require the environment i have.
But once you create them, you should be saying "go" - go do something
and come back and get results.
Whatever we end up having, my preference would be something along those
lines,

tcsim has one major advantage, you can test the actual scheduling algorithms for their behaviour under very controlled conditions. It gave me a lot more confidence when replacing the HFSC lists by rbtrees. But, as Thomas notes, it does all its tests in userspace, which might not be ideal for things besides scheduling algorithms. So a combination of both seems to be best.

Regards
Patrick

<Prev in Thread] Current Thread [Next in Thread>