Hi Ville,

Thanks for reviewing this.

> Which source OUI does the Apple driver write there?

When I first tried to get PSR working, I also tried writing Apple's
00:10:fa from i915 before link training. It made no difference. I then
traced macOS and found the extra writes in this patch. With those
added, PSR worked, leaving i915's Intel source OUI unchanged.

After your question, I read back 0x300-0x302 on the MacBookPro16,2
under macOS 26.6.1 and got 00 00 00 twice. The same read method
returned the expected sink identity at 0x400. I haven't captured boot,
so I can't say whether a different source OUI was set earlier.

> Maybe these should include "SOURCE" vs. "SINK" in the name as well.

Agreed. I'll use APPLE_SOURCE_PSR_SETUP for 0x321 and APPLE_SINK_PSR_*
for the other registers, and update the comment.

> This is some other device on these systems? Can't just match on the
GPU subsys vendor/device + sink OUI alone?

Yes, 106b:1801 is the separate T2 Bridge Controller. We only check its
presence to restrict the quirk to T2 systems and retain NO_PSR on older
untested Macs. The current subsystem device ID is a wildcard.

Specific GPU subsystem vendor/device pairs plus the sink OUI could
replace this. I chose T2 detection to avoid a per-model list; the five
tested models are only a subset of what matches. If you prefer explicit
subsystem IDs, I can collect and verify those for the tested machines.

Best regards,
Alex

Reply via email to