Date: Sat, 01 Mar 1997 17:56:24 -0800 From: David Greenman <dg@root.com> To: "David S. Miller" <davem@jenolan.rutgers.edu> Cc: netdev@roxanne.nuclecu.unam.mx, hackers@FreeBSD.ORG Subject: Re: ok, final sockhash changes, new diff Message-ID: <199703020156.RAA00329@root.com> In-Reply-To: Your message of "Sat, 01 Mar 1997 20:02:55 EST." <199703020102.UAA09468@jenolan.caipgeneral>
next in thread | previous in thread | raw e-mail | index | archive | help
> Reply-To: dg@root.com
> Date: Sat, 01 Mar 1997 16:59:21 -0800
>
> Hmmm. It seems that it might be better to add in the laddr if it
> contains additional variable information, but I don't see how not
> doing so would be a degenerate case when having a lot of IP
> aliases. The faddr, lport, and fport are still just as variable as
> in the non-lots-of-aliases case, so the hash distribution should be
> the same.
>
>Good point, but alas there was a reason I considered it useful to add
>in the laddr to the hash, give me some time and I'll remember what the
>reason exactly was (it happens to cost nothing anyways ;-).
The reason just occured to me: In the case of a server with googols of
IP aliases, the {faddr, lport, fport} are all the same (faddr and fport
are wildcards) for the listening sockets of a given service (say http)...so
the only thing unique is the laddr, which if not included in the hash, will
cause all of the listening sockets to be hashed to the same bucket.
-DG
David Greenman
Core-team/Principal Architect, The FreeBSD Project
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199703020156.RAA00329>