netdev - Re: [PATCH] Fix ARM BUILD_BUG_ON() errors with batman-adv

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: <201312021858.48074.sw@simonwunderlich.de>
Date:	Mon, 2 Dec 2013 18:58:47 +0100
From:	Simon Wunderlich <sw@...onwunderlich.de>
To:	David Miller <davem@...emloft.net>
Cc:	linux@....linux.org.uk, mareklindner@...mailbox.ch,
	antonio@...hcoding.com, b.a.t.m.a.n@...ts.open-mesh.org,
	netdev@...r.kernel.org
Subject: Re: [PATCH] Fix ARM BUILD_BUG_ON() errors with batman-adv

David,

> From: Russell King - ARM Linux <linux@....linux.org.uk>
> Date: Sat, 30 Nov 2013 19:15:53 +0000
> 
> > so there should be no undesired side effect from this packing.
> 
> There is a huge side effect from ever using the packed attribute, in
> that the compiler can assume absolutely nothing about the alignment of
> any object or sub-object of the type you apply this attribute to.
> 
> Even if it is "obvious" that some members will be aligned, the
> compiler cannot take advantage of this assumption because this
> attribute also means that an array of such elements might have
> arbitrary alignment.  So you when you get a pointer to one of these
> objects, the compiler has to assume the worst possible case.
> 
> This means using 4 byte loads to load a 32-bit quantity, always,
> unconditionally, no matter what.
> 
> That's why we should do whatever is necessary to align things properly
> by hand, and use packed only as the last possible and least desirable
> resort.

so far we have two feasible options how to handle batadv_header:

 1) Russells patch
 2) unrolling these 3 bytes of header into every packet structure

I understand your concerns for __packed in generally, but in this case it 
might not be that bad:

 * batadv_header is only used by embedding it in other structs (like 
batadv_unicast_packet, batadv_broadcast_packet ...) which are well aligned - I 
guess there should be no performance problem when accessing through the parent 
structure (like unicast_packet->header.version)
 * there is one exception in batadv_interface_rx() where we use this 
batadv_header directly, but we can easily replace that by using e.g. 
batadv_bcast_packet instead.
 * It's easier to read and maintain than copying these 3 bytes in every other 
structure.

Therefore I'd suggest to pick Russels patch, I can send a fix for 
batadv_interface_rx() later, too.

Thanks,
    Simon
--
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