Re: [PATCH v4 0/9] gpu: nova-core: expose the logging buffers via debugfs

["Danilo Krummrich" <dakr@kernel.org>]

On Wed Jan 14, 2026 at 12:26 AM CET, Timur Tabi wrote:
> On Wed, 2026-01-14 at 00:11 +0100, Danilo Krummrich wrote:
>> On Tue Jan 13, 2026 at 11:53 PM CET, Timur Tabi wrote:
> Sorry, I couldn't find these messages in my inbox.  I will address them in v5.

Sounds good, thanks.

>> >   gpu: nova-core: use pin projection in method boot()
>> >   rust: debugfs: implement Directory trait for Dir
>> >   rust: debugfs: wrap Entry in an enum to prep for LookupDir
>> >   rust: debugfs: add LookupDir
>> 
>> https://lore.kernel.org/all/DF18RFX3IHVP.3GYNJIYAFFJU6@kernel.org/
>> 
>> Especially this one is a concern, I don't want to add this infrastructure as a
>> workaround until we land the feature Gary works on.
>> 
>> As mentioned in this reply, I think that debugfs_lookup() rarely is the correct
>> solution and more often indicates some kind of (design) issue, like it does in
>> this case.
>
> Well, I'm at a loss how to proceed, then.  I had a global variable, which is the approach that
> Nouveau uses, it was recommended instead that I use debugfs_lookup().  So I implemented that, but
> that is also rejected.

Yeah, the global is not a good approach for a final solution. But as temporary
workaround it's probably fine.

Maybe Gary has an alternative idea for a temporary workaround since he's working
on the proper solution to safely access module fields.

But again, I think a simple global with a FIXME comment should be fine.

> Danilo, you are the one who has been pushing for this code.  Please note that the logging buffers
> can only be parsed by Nvidia employees anyway, and we can just use this patchset in house to debug
> GSP-RM as needed.  So if you want to wait until Gary's work is ready (I don't know anything about
> what he's doing), please let me know now and I will just keep these patches internally.

I can't remember pushing for it, but I do think it would be good to land.