linux-kernel - Re: [PATCH v2] platform/x86: serial-multi-instantiate: Don't require both I2C and SPI

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: <65362768-f64b-42d9-8f74-e4f5190af95c@redhat.com>
Date: Tue, 27 Aug 2024 11:05:25 +0200
From: Hans de Goede <hdegoede@...hat.com>
To: Andy Shevchenko <andy.shevchenko@...il.com>,
 Richard Fitzgerald <rf@...nsource.cirrus.com>
Cc: ilpo.jarvinen@...ux.intel.com, platform-driver-x86@...r.kernel.org,
 linux-kernel@...r.kernel.org, patches@...nsource.cirrus.com
Subject: Re: [PATCH v2] platform/x86: serial-multi-instantiate: Don't require
 both I2C and SPI

Hi Andy,

On 8/26/24 10:10 PM, Andy Shevchenko wrote:
> Wed, Aug 14, 2024 at 02:29:39PM +0100, Richard Fitzgerald kirjoitti:
>> Change the Kconfig dependency so that it doesn't require both I2C and SPI
>> subsystems to be built. Make a few small changes to the code so that the
>> code for a bus is only called if the bus is being built.
>>
>> When SPI support was added to serial-multi-instantiate it created a
>> dependency that both CONFIG_I2C and CONFIG_SPI must be enabled.
>> Typically they are, but there's no reason why this should be a
>> requirement. A specific kernel build could have only I2C devices
>> or only SPI devices. It should be possible to use serial-multi-instantiate
>> if only I2C or only SPI is enabled.
>>
>> The dependency formula used is:
>>
>>   depends on (I2C && !SPI) || (!I2C && SPI) || (I2C && SPI)
>>
>> The advantage of this approach is that if I2C=m or SPI=m then
>> SERIAL_MULTI_INSTANTIATE is limited to n/m.
> 
> ...
> 
>>  static void smi_devs_unregister(struct smi *smi)
>>  {
>> +#if IS_REACHABLE(CONFIG_I2C)
> 
> There is no explanation why ugly ifdeffery is used here, while normal
> conditionals elsewhere.

Note that this has already been merged as is, as you've figured
out yourself the reason to use #ifdef here is because
there is no i2c_unregister_device() prototype declared when
CONFIG_I2C=n

> 
>>  	while (smi->i2c_num--)
>>  		i2c_unregister_device(smi->i2c_devs[smi->i2c_num]);
>> +#endif
>>  
>> -	while (smi->spi_num--)
>> -		spi_unregister_device(smi->spi_devs[smi->spi_num]);
>> +	if (IS_REACHABLE(CONFIG_SPI)) {
>> +		while (smi->spi_num--)
>> +			spi_unregister_device(smi->spi_devs[smi->spi_num]);
>> +	}
>>  }
> 
> There are ways to solve this:
> 1) add a stub for I2C=n for i2c_unregister_device();

Yes that would be an option to clean this up a bit as a follow-up
patch series.

Note no need for a stub, just move the declaration out of
the #if IS_ENABLED(CONFIG_I2C) block, using if (IS_REACHABLE)
only requires a prototype.

> 2) resplit this driver to have several built modules:
>    core, I2C parts, SPI parts.
> 

Regards,

Hans