lkml.org 
[lkml]   [1998]   [Jan]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: missing aliased interfaces in /proc/net/dev
Date
Hello!

> > Now it is no device by itself (IMHO no problem), but there are no
> > longer statistics avail (IMHO a bad change) and they do not show up in
>
> Alan> Use ipfwadm
>
> How ? ipfwadm (resp /proc/net/ip_*) does not tell me what aliased
> interfaces are in use, if i don't know.
>
> And to collect statistics via ipfwadm rules needs priviledges not
> avail to any userspace program, but peviously by reading /proc files
> this was possible.

Mmm... But statistics about aliases never were available, right?
So that we losed nothing, and lines sort of "statistics is not available"
would not help. And if someone wants to know list of IP interfaces,
he calls SIOCGIFCONF.

Generally, philosophy behind removing aliases is:
we never were able to segregate reliably traffic related to
different aliased interfaces, criteria were obscure and
in 2.0 they were even confusing (even more, they were incorrect
f.e. arp on aliased devices did some strange things)

Besides that, it turns out that kernel need not know
this information, so that the thing that looks like aliases now
is implemented in the frontend of IP API. Actually, core logic
do not know anything about them, and turning of CONFIG_*_ALIASES
only disables old style access.

So that, the statement sort of "packet A arrived on alias B"
is non-sense. Packets arrive on physical devices and kernel even never
tries to determine on what "alias" they arrived.
The same thing is true about output packets.

From the other hand: the statement sort of "IP packet A arrived on
device B and has source address D and destination address E (and protocol...
and ports... etc. etc.)" is always reasonable question.
And it is the thing that IP accounting rules do.

> (Sorry, to lengthen this thread),

I am sorry too, my message is long and, however, I feel it is obscure :-)

I'd write it into a README, but it seems to be not urgent. Actually,
this approach is not new at all. It is parallel to Linux IPv6 and
it looks similar to sanitized 4.4BSD.

Alexey

\
 
 \ /
  Last update: 2005-03-22 13:41    [W:0.046 / U:1.268 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site