netdev - Re: mlxsw: spectrum: Adjust headroom buffers for 8x ports

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] [day] [month] [year] [list]

Message-ID: <20200617175219.GB296888@splinter>
Date:   Wed, 17 Jun 2020 20:52:19 +0300
From:   Ido Schimmel <idosch@...lanox.com>
To:     Colin Ian King <colin.king@...onical.com>
Cc:     Jiri Pirko <jiri@...lanox.com>,
        "David S. Miller" <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>,
        Petr Machata <petrm@...lanox.com>, netdev@...r.kernel.org,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: mlxsw: spectrum: Adjust headroom buffers for 8x ports

On Wed, Jun 17, 2020 at 06:15:35PM +0100, Colin Ian King wrote:
> Hi
> 
> Static analysis with Coverity has detected an issue that relies on the
> machine endianness to work. The commit in question is:
> 
> commit 60833d54d56c21e7538296eb2e00e104768fd047
> Author: Ido Schimmel <idosch@...lanox.com>
> Date:   Tue Jun 16 10:14:58 2020 +0300
> 
>     mlxsw: spectrum: Adjust headroom buffers for 8x ports
> 
> in line:
>     mlxsw_sp_port_headroom_8x_adjust(mlxsw_sp_port, (u16 *) &buffsize);
> 
> 
> The cast of the u32 buffsize to (u16 *) to scale buffsize in the call to
> to mlxsw_sp_port_headroom_8x_adjust() will behave differently on big
> endian architectures to that of little endian architectures.  I'm not
> sure if this is intentional or not.

Not intentional. The hardware interface is quite weird. The buffer size
can be configured via two registers. One takes size as 24 bits and the
second as 16 bits. Either way, the max size is much lower than 2^16-1.

> One solution is to either make buffsize a u16, but I am concerned this
> may be incorrect as the buffsize is assigned from the call
> mlxsw_sp_span_buffsize_get() and this returns a u32 so we may have
> overflow issues. Probably better to make
> mlxsw_sp_port_headroom_8x_adjust handle u32 integers and to return the
> adjusted value rather than modifying it by pass-by-reference.

Thanks for the report, will fix.

> 
> Colin
>