netdev - Re: [PATCH iwl-net] ice: health.c: fix compilation on gcc 7.5

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: <4950e61a-189f-4749-baf7-454ca2cce250@kernel.org>
Date: Thu, 6 Feb 2025 08:53:55 +0100
From: Jiri Slaby <jirislaby@...nel.org>
To: David Laight <david.laight.linux@...il.com>,
 Simon Horman <horms@...nel.org>
Cc: Przemek Kitszel <przemyslaw.kitszel@...el.com>,
 intel-wired-lan@...ts.osuosl.org, Tony Nguyen <anthony.l.nguyen@...el.com>,
 netdev@...r.kernel.org, Konrad Knitter <konrad.knitter@...el.com>,
 Michal Swiatkowski <michal.swiatkowski@...ux.intel.com>,
 Qiuxu Zhuo <qiuxu.zhuo@...el.com>,
 Linus Torvalds <torvalds@...ux-foundation.org>, Kees Cook <kees@...nel.org>,
 Nick Desaulniers <nick.desaulniers@...il.com>
Subject: Re: [PATCH iwl-net] ice: health.c: fix compilation on gcc 7.5

On 05. 02. 25, 23:56, David Laight wrote:
> On Wed, 5 Feb 2025 20:45:46 +0000
> Simon Horman <horms@...nel.org> wrote:
> 
>> + Jiri
>>
>> On Wed, Feb 05, 2025 at 11:42:12AM +0100, Przemek Kitszel wrote:
>>> GCC 7 is not as good as GCC 8+ in telling what is a compile-time const,
>>> and thus could be used for static storage. So we could not use variables
>>> for that, no matter how much "const" keyword is sprinkled around.
>>>
>>> Excerpt from the report:
>>> My GCC is: gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0.
>>>
>>>    CC [M]  drivers/net/ethernet/stmicro/stmmac/stmmac_mdio.o
>>> drivers/net/ethernet/intel/ice/devlink/health.c:35:3: error: initializer element is not constant
>>>     ice_common_port_solutions, {ice_port_number_label}},
>>>     ^~~~~~~~~~~~~~~~~~~~~~~~~
>>> drivers/net/ethernet/intel/ice/devlink/health.c:35:3: note: (near initialization for 'ice_health_status_lookup[0].solution')
>>> drivers/net/ethernet/intel/ice/devlink/health.c:35:31: error: initializer element is not constant
>>>     ice_common_port_solutions, {ice_port_number_label}},
>>>                                 ^~~~~~~~~~~~~~~~~~~~~
>>> drivers/net/ethernet/intel/ice/devlink/health.c:35:31: note: (near initialization for 'ice_health_status_lookup[0].data_label[0]')
>>> drivers/net/ethernet/intel/ice/devlink/health.c:37:46: error: initializer element is not constant
>>>     "Change or replace the module or cable.", {ice_port_number_label}},
>>>                                                ^~~~~~~~~~~~~~~~~~~~~
>>> drivers/net/ethernet/intel/ice/devlink/health.c:37:46: note: (near initialization for 'ice_health_status_lookup[1].data_label[0]')
>>> drivers/net/ethernet/intel/ice/devlink/health.c:39:3: error: initializer element is not constant
>>>     ice_common_port_solutions, {ice_port_number_label}},
>>>     ^~~~~~~~~~~~~~~~~~~~~~~~~
>>>
>>> Fixes: 85d6164ec56d ("ice: add fw and port health reporters")
>>> Reported-by: Qiuxu Zhuo <qiuxu.zhuo@...el.com>
>>> Closes: https://lore.kernel.org/netdev/CY8PR11MB7134BF7A46D71E50D25FA7A989F72@CY8PR11MB7134.namprd11.prod.outlook.com
>>> Signed-off-by: Przemek Kitszel <przemyslaw.kitszel@...el.com>
>>> ---
>>> I would really like to bump min gcc to 8.5 (RH 8 family),
>>> instead of supporting old Ubuntu. However SLES 15 is also stuck with gcc 7.5 :(
>>>
>>> CC: Linus Torvalds <torvalds@...ux-foundation.org>
>>> CC: Kees Cook <kees@...nel.org>
>>> CC: Nick Desaulniers <nick.desaulniers@...il.com>
>>
>> Hi Prezemek,
>>
>> I ran into a similar problem not so long ago and I'm wondering if
>> the following, based on a suggestion by Jiri Slaby, resolves your
>> problem.
> 
> I'm sure I remember from somewhere that although the variables are
> 'static const' they have to be real variables because they can still
> be patched.

Not sure what you mean -- using macros, they placed the strings into 
.rodata anyway.

Mind the difference between:
const char *X;
char *const X;
const char *const X;

They are all different and allow different things (X++, (*X)++, nothing).

Possibly, the problem above could be fixed by the third variant too, 
IMO. But is ineffective (having generated an unused pointer).

thanks,
-- 
js
suse labs