On Sun, Aug 23, 2026 at 12:00:39PM +0000, [email protected] wrote:
> This isn't a bug introduced by this patch, but since the stutter latency
> leak is explicitly left unpatched for DCN 3.01, will it remain vulnerable
> to MPO flip_done timeouts?

Yes. dcn301_fpu.c has the same calculate_wm_set_for_vlevel() helper and
only restores dram_clock_change_latency_us. DCN 3.01 also still calls
dcn30_internal_validate_bw() with allow_self_refresh_only true, and
watermarks run only for DC_VALIDATE_MODE_AND_PROGRAMMING, so the check
vs commit_tail mismatch is the same shape as on DCN 2.1.

I left it out of this patch on purpose: this is the DCN 2.1 regression
(360 Hz eDP dropped after b96150a70696), I have no DCN 3.01 hardware,
and I would rather not touch Van Gogh watermarks untested. Happy to
send a follow-up if someone with DCN 3.01 can try it.

v2 is otherwise unchanged. 1920x1080@360 is in daily use here on
Cezanne with this applied. A human look would be very welcome.

Thanks,
Julien

Reply via email to