linux-kernel - Re: [PATCH 2/2] hwmon: chipcap2: disable sensor if request ready irq fails

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 for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <d90b6e77-be5f-4f47-8f9f-6fb926a0690f@gmail.com>
Date: Tue, 13 Aug 2024 00:04:29 +0200
From: Javier Carrasco <javier.carrasco.cruz@...il.com>
To: Guenter Roeck <linux@...ck-us.net>, Jean Delvare <jdelvare@...e.com>
Cc: linux-hwmon@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 2/2] hwmon: chipcap2: disable sensor if request ready irq
 fails

On 12/08/2024 23:26, Guenter Roeck wrote:
> On 8/12/24 13:48, Javier Carrasco wrote:
>> On 12/08/2024 22:08, Guenter Roeck wrote:
>>> On 8/12/24 12:59, Javier Carrasco wrote:
>>>> On 12/08/2024 18:49, Guenter Roeck wrote:
>>>>> On 8/12/24 08:43, Javier Carrasco wrote:
>>>>>> This check is carried out after getting the regulator, and the device
>>>>>> can be disabled if an error occurs.
>>>>>>
>>>>>
>>>>> I do not see a possible path for a call to cc2_enable() at this point,
>>>>> meaning the regulator won't ever be enabled. Please provide a better
>>>>> explanation why this patch would be necessary.
>>>>>
>>>>> Guenter
>>>>>
>>>>
>>>> Hi Guenter,
>>>>
>>>> this patch enforces the state where the dedicated regulator is
>>>> disabled,
>>>> no matter what the history of the regulator was. If a previous
>>>> regulator_disable() failed, it would still be desirable that the
>>>> regulator gets disabled the next time the driver is probed (i.e. a new
>>>> attempt to disable it on failure).
>>>> cc2_disable() checks first if the regulator is enabled to avoid any
>>>> imbalance.
>>>>
>>>
>>> That is very theoretic. Sorry, I am not going to accept this patch.
>>>
>>> Guenter
>>>
>>
>> I get your point, but given that this device requires a dedicated
>> regulator, I believe it makes sense that it tries to disable it whenever
>> possible if it's not going to be used. I think that makes more sense
>> that just returning an error value without even making sure that de
>> regulator was disabled, doesn't it?
>>
> 
> No, it doesn't make any sense whatsoever. What are you planning to do,
> clutter the kernel with code to disable regulators if instantiating a
> device
> fails for whatever reason and it turns out that a regulator which should
> not have been enabled to start with turns out to be enabled anyway ?
> 
>> Of course this is not a killer feature, and I don't want to make you
>> waste much time with it. But I think the dedicated regulator should be
>> shut down in all error paths, whatever status it had before.
>>
> 
> I strongly disagree. This can only mess up the kernel all over the place.
> Maybe you can convince other maintainers to accept such code, but please
> refrain from doing that in my scope of responsibility. If the regulator
> subsystem has the habit of leaving regulators enabled even after they
> have been released, that problem should be fixed in the regulator subsystem
> and not be worked around in individual drivers.
> 
> Guenter
> 

In that case the current behavior is wrong in the opposite direction,
and disabling the regulator before any call to cc2_enable() follows the
same invalid assumption.

I will remove the call to cc2_disable() in the probe function, which was
there exactly to ensure that the regulator gets disabled under all
circumstances. All error paths will just return, letting the regulator
untouched if it was not actively enabled.

Best regards,
Javier Carrasco