Nouveau - Oct 2025 - [PATCH 5/7] gpu: nova-core: Add support for managing GSP falcon interrupts

On Tue Oct 21, 2025 at 7:35 AM JST, John Hubbard wrote:
> Alex, this ".alter" method is misnamed, IMHO. Because for
registers,
> The One True Way (or so I claim, haha) is to have the following methods:
>
>     .read
>     .modify, also known as RMW (read-modify-write)
>     .write
>
> "alter" never shows up in this naming scheme. I'm going to
claim that
> this is a bit jarring for old hardware/kernel programmers.
>
> But it's not too late: these are only used in a very few places, and
entirely
> within nova-core, too.
>
> Can I *please* send a patch to rename "alter" to
"modify", perhaps?
Oh yes, although I was just thinking that this should be renamed to
`update` for consistency with regmap.

On 10/22/2025 2:48 AM, Alexandre Courbot wrote:
> On Tue Oct 21, 2025 at 7:35 AM JST, John Hubbard wrote:
>> Alex, this ".alter" method is misnamed, IMHO. Because for
registers,
>> The One True Way (or so I claim, haha) is to have the following
methods:
>>
>>     .read
>>     .modify, also known as RMW (read-modify-write)
>>     .write
>>
>> "alter" never shows up in this naming scheme. I'm going
to claim that
>> this is a bit jarring for old hardware/kernel programmers.
>>
>> But it's not too late: these are only used in a very few places,
and entirely
>> within nova-core, too.
>>
>> Can I *please* send a patch to rename "alter" to
"modify", perhaps?
> 
> Oh yes, although I was just thinking that this should be renamed to
> `update` for consistency with regmap.
> 
Either update or modify would be Ok with me. Update does make it sound more like
a total write though for some reason. Perhaps update_fields ?

thanks,

 - Joel