netdev
[Top] [All Lists]

Re: [PATCH]Add new IPv6 MIBs counters support through netlink

To: xma@xxxxxxxxxx
Subject: Re: [PATCH]Add new IPv6 MIBs counters support through netlink
From: YOSHIFUJI Hideaki / 吉藤英明 <yoshfuji@xxxxxxxxxxxxxx>
Date: Thu, 27 May 2004 08:58:12 +0900 (JST)
Cc: davem@xxxxxxxxxx, mashirle@xxxxxxxxxx, netdev@xxxxxxxxxxx, yoshfuji@xxxxxxxxxxxxxx
In-reply-to: <OFA97F4753.DC8C93AD-ON87256EA0.0080DAAB-07256EA0.00814B08@us.ibm.com>
Organization: USAGI Project
References: <20040527.080828.06774575.yoshfuji@linux-ipv6.org> <OFA97F4753.DC8C93AD-ON87256EA0.0080DAAB-07256EA0.00814B08@us.ibm.com>
Sender: netdev-bounce@xxxxxxxxxxx
In article 
<OFA97F4753.DC8C93AD-ON87256EA0.0080DAAB-07256EA0.00814B08@xxxxxxxxxx> (at Wed, 
26 May 2004 16:32:04 -0700), Shirley Ma <xma@xxxxxxxxxx> says:

> > Without adding support for u64 counter, it does not make sense to add
> > new interface. You can add new counter to the end of the structures.
> 
> It's not good to add new counters to the end of the old structure. The mib 
> 
> number and offset will be wrong.

I don't think so.
But, even if so, we can add only missing stats.

> >This also contains per-interface statistics support.
> (I rember that you said that) we should add idev into an route and avoid
> get/put idev.
> 
> Only a few get/put idev there, not too many. And in some places there are 
> no 
> route, no we can't use idev in route.

okay.

> > I strongly do not want to poeple to change their code twice to support
> draft-ietf-ipv6-rfc2011-update-XX.txt.
> 
> All the counters not just IP will be eventually 64 bits. We can fix it 
> later. And 
> if we implement 64 counters here, there is performance issue on some 
> architectures which don't support atmoic 64 for 32 bit platform.

I strongly do not want to change interface twice to support these new
statistics.

--yoshfuji @ driving traffic jam

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