fwget: Add device IDs and firmwares to pci_video_amd & pci_video_intel

[levelad]

Add GPU names as comments to the latest GPU device IDs as the kernel modules are segmented now.

Names taken from:

https://devicehunt.com/view/type/pci/vendor/1002/device/744C https://devicehunt.com/view/type/pci/vendor/1002/device/15BF https://devicehunt.com/view/type/pci/vendor/1002/device/15C8 https://devicehunt.com/view/type/pci/vendor/1002/device/164E

More references:

https://www.techpowerup.com/gpu-specs/amd-navi-31.g998 https://www.techpowerup.com/gpu-specs/amd-phoenix.g1024 https://www.techpowerup.com/gpu-specs/amd-raphael.g1021

[github-actions[bot]]

Thank you for taking the time to contribute to FreeBSD! There is an issue that needs to be fixed:

• Missing Signed-off-by lines (bc0f7b939c59d0ee918d795040aa4f27713eae63)

Please review CONTRIBUTING.md, then update and push your branch again.

[bsdimp]

Can you rebase this, since there's a conflict? I like the '*' -> '?'. However, the commit messages need to be more specific as to what they do and why.

[levelad]

Conflict with Bugzilla PR: 287441 is resolved. I cannot edit commit messages on GitHub. Should I do a new commit without any changes and add the message there?

The Cannon Lake issue in pci_video_intel needs to be reviewed by evadot.

[bsdimp]

OK. I'll hold off until @evadot can comment. Usually, you can just do a git rebase -i with the 'reword' option for each of the commits to update them, then do a force push and github will update everything.

[levelad]

@bsdimp iGPU for Cannon Lake was factory disabled, see https://en.wikipedia.org/wiki/Cannon_Lake_(microprocessor).

So I don't get why a firmware is loaded for it, should I just remove it completely as it interferes with Apollo Lake (Broxton)? Or anybody else knows why it is here? @evadot?

Broxton was also a cancelled but the firmware seems to be used for Apollo Lake, see https://en.wikipedia.org/wiki/List_of_Intel_CPU_microarchitectures.

[levelad]

@bsdimp I resolved a 2nd conflict now with Differential Revision D52312 Maybe it would be best if Bjoern A. Zeeb looked over my changes before a new conflict arises. He doesn't seem to be on GitHub so I can't mention him here. Some of his changes in his DR were already in my commit.

I just wanted to contribute a litte improvement to this code and I really would like to close this before I get stuck in this conflict nightmare. :)

[bzfbd]

Given I was asked in email to look:

• Can the amd changes be split up into two changes: (1) changes to firmware and IDs and (2) all the asterisk to question mark bits. That way the actual changes are visible. Note: I do not like any of the globbing but I for 0xXXXn I still prefer the * than the ?; it reads better, sticks out more; less confusing with numbers and the PCI IDs can never be longer anyway.
• The "Cannon Lake Issue" is not new; I had the same finding when I extracted them from the driver doing a first cut of the lua match script: https://reviews.freebsd.org/D52325#inline-318188
• I wonder how your expanded IDs match what I have in the above lua review where I manually kept the globs from the shell script and added the more specifics from the driver (at that time) which likely is not the full set either...

[levelad]

I put all the changes into the commit descriptions. I completely removed the Cannon Lake firmware now.

I checked the existing IDs with devicehunt and PCILookup and expanded them when they did also cover non-GPU devices. Some IDs still have some non-GPU devices so your explicit ID definitions could prevent unnecessary firmware loading.