netdev - Re: [RFC davem] revert: net: Make skb->skb

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]

Message-Id: <20130112.133630.257139657732337147.davem@davemloft.net>
Date:	Sat, 12 Jan 2013 13:36:30 -0800 (PST)
From:	David Miller <davem@...emloft.net>
To:	socketcan@...tkopp.net
Cc:	netdev@...r.kernel.org
Subject: Re: [RFC davem] revert: net: Make skb->skb_iif always track
 skb->dev

From: David Miller <davem@...emloft.net>
Date: Sat, 12 Jan 2013 13:23:16 -0800 (PST)

> From: Oliver Hartkopp <socketcan@...tkopp.net>
> Date: Sat, 12 Jan 2013 14:48:14 +0100
> 
>> To me it is not clear why skb_iff is needed anyway as the value should
>> always be available via skb->dev->ifindex, right?
> 
> But all the code uses skb_iif, in particular the ipv4 routing
> lookups use that as the key.
> 
> It absolutely must follow whatever is skb->dev, it is a hard
> invariant.
> 
> I am not reverting this change.

More information, because I can't believe how idiotic and
ignorant people are being able this issue.

skb->dev->ifindex IS NOT the same as skb->skb_iif

Why don't you put a test into tcp_recvmsg() for packets being removed
from the socket's receive queue and see if skb->dev->ifindex is the
same as skb->skb_iif

Surprise, skb->dev is going to be NULL at that point.

Why?

Because on packet receive we don't take references on devices we hook
into skb->dev, therefore we cannot let that pointer escape the
software interrupt packet input paths.

Therefore, as a bug trap, TCP input will set skb->dev to NULL.

The only valid way to figure out the final demuxed device the packet
arrived on, is therefore, via skb->skb_iif.

As per your problem with CAN, that's also rediculous.  You have an SKB
control block in skb->cb[] that you can put whatever values with
whatever semantics you want.

Use it.

I'm not discussing this any further.

--
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