netdev
[Top] [All Lists]

Re: Source Specific Query of MLDv2 [PATCH]

To: YOSHIFUJI Hideaki / 吉藤英明 <yoshfuji@xxxxxxxxxxxxxx>
Subject: Re: Source Specific Query of MLDv2 [PATCH]
From: David Stevens <dlstevens@xxxxxxxxxx>
Date: Fri, 6 Feb 2004 22:40:42 -0800
Cc: davem@xxxxxxxxxx, hibi665@xxxxxxx, netdev@xxxxxxxxxxx, "Hideaki YOSHIFUJI" <yoshfuji@xxxxxxxxxxxxxxxxxxxxxxxxx>, yoshfuji@xxxxxxxxxxxxxx
In-reply-to: <20040207.142519.117436499.yoshfuji@xxxxxxxxxxxxxx>
Sender: netdev-bounce@xxxxxxxxxxx

Good idea; I'll incorporate all the comments, retest and repost.

> So... hmm... Let's check dev->flags ^ (IFF_PROMISC|IFF_ALLMULTI) like this.

I think you mean (dev->flags & (IFF_PROMISC|IFF_ALLMULTI)) here, right?

> I don't think it is so costy for usual operation.
> If you think it is, please introduce hash table for mc_list in idev;
> it should have been "heavy." :-)
> (You can do it later, of course.)

This certainly could be done and I had this in mind for the future when
I designed the interface filter data structures (made sure everything

for the intersection and union is directly available given only the
ifmcaddr). I should do some testing to see how many MCA's or source
filters are needed to make a hash worthwhile, and maybe how many before
it falls over. :-)
thanks!
+-DLS

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