HW LED triggers again

HW LED triggers again

[Marek Behun]

Hi Jacek,

I want to open the discussions about HW LED triggers again.
The last time (which was almost a year ago, sorry for that) I proposed
an API which used the same sysfs trigger file as for regular trigger
setting, but the HW triggers were prefixed with "hw:" (and each LED
classdev can have different ones).

You wrote:

> I wonder what will be the gain of having hw triggers incorporated
> into LED trigger mechanism, if they are meant not be generic
> by design? Only the LED class driver exposing a hw trigger
> will know how to set it up, and will define protocol via which
> the settings will be passed from sysfs to the trigger (const char*
> parameter in the hw_trigger_set() op).
> 
> And it has to be that way because hardware triggers are hardware
> specific. LED class driver will have to create trigger specific
> sysfs files regardless of whether they are to be shown on
> trigger avtivation, or will persist for the whole LED class device
> lifetime.
> 
> Maybe I'm missing some vital details from the previous discussions,
> but this is what's come to my mind now, after analyzing the proposed
> design.
> 
> The question is: what problem we solve by exposing non-generic
> hw trigger, whose implementation will be in the driver anyway,
> instead of just bypassing the trigger mechanism and exposing
> the required interface directly?

I would still like to go this way, so my answer to this questions is:
- IMO this is simpler for users and existing scripts
- the idea is that it should no be possible to set a software trigger
  and a hardware trigger at the same time (this would just end up in
  more complications), and introducing special hw_trigger file or
  something could make users think that you can

Marek

Re: HW LED triggers again

[Pavel Machek]

[-- Attachment #1: Type: text/plain, Size: 985 bytes --]

Hi!

> I want to open the discussions about HW LED triggers again.
> The last time (which was almost a year ago, sorry for that) I proposed
> an API which used the same sysfs trigger file as for regular trigger
> setting, but the HW triggers were prefixed with "hw:" (and each LED
> classdev can have different ones).
> 
> You wrote:
...
> I would still like to go this way, so my answer to this questions is:
> - IMO this is simpler for users and existing scripts
> - the idea is that it should no be possible to set a software trigger
>   and a hardware trigger at the same time (this would just end up in
>   more complications), and introducing special hw_trigger file or
>   something could make users think that you can

As usual, devil is in the details, but so far, I like the proposal.

Best regards,
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

Re: HW LED triggers again

[Jacek Anaszewski]

Hi Marek,

On 3/20/20 8:43 PM, Marek Behun wrote:
> Hi Jacek,
> 
> I want to open the discussions about HW LED triggers again.
> The last time (which was almost a year ago, sorry for that) I proposed
> an API which used the same sysfs trigger file as for regular trigger
> setting, but the HW triggers were prefixed with "hw:" (and each LED
> classdev can have different ones).
> 
> You wrote:
> 
>> I wonder what will be the gain of having hw triggers incorporated
>> into LED trigger mechanism, if they are meant not be generic
>> by design? Only the LED class driver exposing a hw trigger
>> will know how to set it up, and will define protocol via which
>> the settings will be passed from sysfs to the trigger (const char*
>> parameter in the hw_trigger_set() op).
>>
>> And it has to be that way because hardware triggers are hardware
>> specific. LED class driver will have to create trigger specific
>> sysfs files regardless of whether they are to be shown on
>> trigger avtivation, or will persist for the whole LED class device
>> lifetime.
>>
>> Maybe I'm missing some vital details from the previous discussions,
>> but this is what's come to my mind now, after analyzing the proposed
>> design.
>>
>> The question is: what problem we solve by exposing non-generic
>> hw trigger, whose implementation will be in the driver anyway,
>> instead of just bypassing the trigger mechanism and exposing
>> the required interface directly?
> 
> I would still like to go this way, so my answer to this questions is:
> - IMO this is simpler for users and existing scripts
> - the idea is that it should no be possible to set a software trigger
>   and a hardware trigger at the same time (this would just end up in
>   more complications), and introducing special hw_trigger file or
>   something could make users think that you can

OK, that seems like a decent justification. If you had provided it at
that time then maybe we would have had generic hw trigger mechanism
merged a year ago :-).

-- 
Best regards,
Jacek Anaszewski