Re: [PATCH v3 5/7] gpu: nova-core: transition gen_bootloader to TLV images

[John Hubbard <jhubbard@nvidia.com>]

On 7/9/26 2:27 PM, Timur Tabi wrote:
> On Thu, 2026-07-09 at 11:24 +0900, Alexandre Courbot wrote:
>>>>
>>>> Also, since `dev_dbg` can be compiled out, the check performed on the
>>>> `VERS` tag existing is not guaranteed to be performed. We should
>>>> probably extract it out of the `dev_dbg` statement. This is true across
>>>> the series; which makes me think it might also benefit from a helper to
>>>> factor out the request_t.v/Tlv::new/dev_dbg common sequence.
>>>
>>> We don't care about the VERS tag if we're not going to print it out, so we probably should
>>> not
>>> extract it otherwise.
>>
>> The spec mentions that this tag is present in all TLV files, so its
>> absence means the file is invalid and should trigger an error.
> 
> If a tree falls in the woods and no one is around to hear it, does it make a sound?
> 
> I see your point, I'm just not sure I agree with it.  By that logic, we should test for VERS in
> the constructor.  My concern is that maybe one day we will have a TLV that doesn't have a VERS
> tag because it's not attached to a versioned firmware image, it's just some misc data.  
> 
> I need to think about this a little more.  What do you think about testing for VERS in the
> constructor?  We could move the dev_dbg! into the constructor, but then we'd have to pass `dev`
> as a parameter.
> 
> I'm going to post a v4 today in the meantime.

If you have a spec, then it's either a spec or it's just a waste of time.
Assuming that it's a spec, then if it says "shall have a VERS tag", then
by gosh we should fail out if that tag is missing.

Otherwise, write the spec differently.

This is not a fuzzy sort of thing.

thanks,
-- 
John Hubbard