netdev - Re: [PATCH] net-netlink: Add a new attribute to expose TCLASS values via netlink

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Wed, 9 Nov 2011 17:50:27 -0800
From:	Maciej Żenczykowski <zenczykowski@...il.com>
To:	David Miller <davem@...emloft.net>
Cc:	netdev@...r.kernel.org, muralira@...gle.com, shemminger@...tta.com,
	eric.dumazet@...il.com
Subject: Re: [PATCH] net-netlink: Add a new attribute to expose TCLASS values
 via netlink

> I can't see how an ipv6 mapped socket can even set the inet->tos value.
>
> As far as I can see, only net/ipv4/ip_sockglue.c:ip_setsockopt() provides
> the interface to change inet->tos.

Yes, but an ipv6 socket is permitted to setsockopt SOL_IP in addition
to SOL_IPV6.
As such, one can simply setsockopt(ipv6_socket, SOL_IP, IP_TOS, value).
This is just as well, because inet->tos is what determines the ipv4 tos value
of any ipv4 packets generated by an ipv6 socket, ie. when connect(ed/ing)
to an ipv4 mapped ipv6 address.

> And ipv6 sockets, of any type, are provided no such vector by which to
> get at those interfaces.

> So inet->tos is always left at it's default value for ipv6 mapped sockets,
> and therefore I see no reason to report TCLASS vs. TOS separately.

> In fact, what I would suggest is to do something about the lack of
> ability to set inet->tos, and the best way to do that seems to be to
> simply propagate the npinfo->tclass setting into inet->tos.  Performaing
> any munging if necessary.

> I'm not applying this patch.

If people are in agreement that inet->tos vs ipv6->tclass being
separately settable is not desirable,
then I'm willing to go through and remove the tclass field (and send
out a patch to do that).

However, this is _QUITE DEFINITELY_ a user visible change in API and
application visible behaviour.

It is however an annoying corner case, so perhaps would be best fixed
by getting rid of it?
OTOH, I'm not sure if people aren't perhaps relying on the ability to
separately set ipv4 dscp vs ipv6 tclass due to some internal network
constraints...

Personally, I'd prefer the simplification of not having to deal with
IP_TOS vs IPV6_TCLASS discrepancies on ipv6 dual-stack sockets.

- Maciej
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html