netdev
[Top] [All Lists]

Re: [PATCH 1/4] Prefix List against 2.5.73

To: kuznet@xxxxxxxxxxxxx
Subject: Re: [PATCH 1/4] Prefix List against 2.5.73
From: YOSHIFUJI Hideaki / 吉藤英明 <yoshfuji@xxxxxxxxxxxxxx>
Date: Wed, 16 Jul 2003 17:39:26 +0900 (JST)
Cc: krkumar@xxxxxxxxxx, davem@xxxxxxxxxx, netdev@xxxxxxxxxxx, yoshfuji@xxxxxxxxxxxxxx, linux-net@xxxxxxxxxxxxxxx
In-reply-to: <200307160021.EAA10195@dub.inr.ac.ru>
Organization: USAGI Project
References: <3F14492C.30708@us.ibm.com> <200307160021.EAA10195@dub.inr.ac.ru>
Sender: netdev-bounce@xxxxxxxxxxx
Hello.

In article <200307160021.EAA10195@xxxxxxxxxxxxx> (at Wed, 16 Jul 2003 04:21:33 
+0400 (MSD)), kuznet@xxxxxxxxxxxxx says:

> Select yourself: either IFA_IFFLAGS or translated flags in ifa_flags.
> I prefer the second way just because it is too unpleasant to add
> a new attribute for sake of two bits with no visible candidates
> to use remaining ones.

Well, I dislike ifa_flags because 
 - it is conceptually wrong to combine them.
   e.g. even if all autoconf addresses expired, flags lasts and 
   we should report it to userspace.
 - ifa_flags is extremely expensive resource.
   There are only 8 bits. Use it only for addresses.

My suggestion is:
 - create L3 per-interface RTM, say, RTM_xxxIFACE.
 - provide inet_device / inet6_dev things via this RTM.
    e.g. per-interface statistics, flags etc.

-- 
Hideaki YOSHIFUJI @ USAGI Project <yoshfuji@xxxxxxxxxxxxxx>
GPG FP: 9022 65EB 1ECF 3AD1 0BDF  80D8 4807 F894 E062 0EEA

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