linux-kernel - Re: [PATCH] net_sched: force endianness annotation

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: <eb4449ae-70db-c487-9c47-301225734943@solarflare.com>
Date:   Mon, 29 Apr 2019 12:11:18 +0100
From:   Edward Cree <ecree@...arflare.com>
To:     Nicholas Mc Guire <der.herr@...r.at>
CC:     Nicholas Mc Guire <hofrat@...dl.org>,
        Jamal Hadi Salim <jhs@...atatu.com>,
        Cong Wang <xiyou.wangcong@...il.com>,
        Jiri Pirko <jiri@...nulli.us>,
        "David S. Miller" <davem@...emloft.net>, <netdev@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] net_sched: force endianness annotation

On 29/04/2019 11:44, Nicholas Mc Guire wrote:
> be16_to_cpu((__force __be16)val) should be a NOP on big-endian as well
Yes.  But it's semiotically wrong to call be16_to_cpu() on a cpu-endian
 value; if the existing behaviour is desired, it ought to be implemented
 differently.
> The problem with using swab16 is that it is impating the binary significantly
> so I'm not sure if the change is really side-effect free
It's not; it changes the behaviour.  That's why I brought up the question
 of the intended behaviour — it's unclear whether the current (no-op on BE)
 behaviour is correct or whether it's a bug in the original code.
Better to leave the sparse error in place — drawing future developers'
 attention to something being possibly wrong here — than to mask it with a
 synthetic 'fix' which we don't even know if it's correct or not.

> but I just am unsure if
> -                   val = be16_to_cpu(val);
> +                   val = swab16(val);
> is actually equivalent.
If you're not sure about such things, maybe you shouldn't be touching
 endianness-related code.  swab is *not* a no-op, either on BE or LE.

-Ed