Received-SPF: pass (google.com: domain of linux-kernel+bounces-27631-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Date: Tue, 16 Jan 2024 09:40:55 -0800
From: Brian Norris <briannorris@chromium.org>
To: kernel test robot <lkp@intel.com>,
	=?iso-8859-1?Q?N=EDcolas_F=2E_R=2E_A=2E?= Prado <nfraprado@collabora.com>
Cc: =?iso-8859-1?Q?N=EDcolas_F=2E_R=2E_A=2E?= Prado <nfraprado@collabora.com>,
	Tzung-Bi Shih <tzungbi@kernel.org>, llvm@lists.linux.dev,
	oe-kbuild-all@lists.linux.dev, kernel@collabora.com,
	AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>,
	Julius Werner <jwerner@chromium.org>,
	chrome-platform@lists.linux.dev, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 3/4] firmware: google: cbmem: Add to module device table
Message-ID: <Zaa_pwuPOMCQV4GD@google.com>
References: <20240111151226.842603-4-nfraprado@collabora.com>
 <202401151013.Xioj5wZo-lkp@intel.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <202401151013.Xioj5wZo-lkp@intel.com>

Hi Nicolas,

On Mon, Jan 15, 2024 at 10:53:48AM +0800, kernel test robot wrote:
> All warnings (new ones prefixed by >>):
> 
> >> drivers/firmware/google/cbmem.c:118:40: warning: unused variable 'cbmem_ids' [-Wunused-const-variable]
>      118 | static const struct coreboot_device_id cbmem_ids[] = {
>          |                                        ^~~~~~~~~
>    1 warning generated.
> 
> 
> vim +/cbmem_ids +118 drivers/firmware/google/cbmem.c
> 
>    117	
>  > 118	static const struct coreboot_device_id cbmem_ids[] = {
>    119		{ .tag = LB_TAG_CBMEM_ENTRY },
>    120		{ /* sentinel */ }
>    121	};
>    122	MODULE_DEVICE_TABLE(coreboot, cbmem_ids);
>    123	

I was wondering why we have a seemingly unique "unused variable" failure
mode in comparison to other similarly-structured device/bus drivers, and
I realized that's because we're not relying on the same structure for
both MODULE_DEVICE_TABLE (struct coreboot_device_id) and for the driver
definition (struct coreboot_driver -> 'u32 tag'). Thus, this structure
is only used for #define MODULE builds, and otherwise not used.

Rather than wrapping these definitions in "#ifdef MODULE", perhaps we
can settle on a single field, and replace `struct coreboot_driver::tag`
with an instance of `struct coreboot_device_id`? That would normally be
a breaking change that would require changing all drivers at the same
time as the bus (or else some kind of intermediate transition state),
but considering there are only 4 driver implementations and they all
live under the same maintainer tree, that seems like it should still be
OK (IMO).

If it makes the series more readable/incremental, perhaps the switchover
can be the last patch in the series, and there can remain some
duplication (and potential -Wunused-const-variable issues) for the
middle of the series.

Brian