[REGRESSION] Please revert 1191285ecb42 from 7.1.y
Sergio Callegari <[email protected]>
| Newsgroups | org.kernel.vger.stable |
|---|---|
| Message-ID | <[email protected]> |
Hi again, after having looked a bit more thoroughly into the matter, and having gone through the relevant commits that introduced and subsequently modified the AMD backlight brightness code paths, I would like to suggest reverting commit 1191285ecb4233e5bf907da7e9394d7999055566 from the 7.1 stable series. On affected AMD systems this change introduces a user-visible brightness regression: values near the newly advertised maximum can wrap around to the minimum brightness. On an ASUS G14, for example, brightness 64532 gives the expected maximum level, while 64533 immediately drops actual_brightness to 0; max_brightness is advertised as 65535. The previous max - min limit therefore appears to have been significant to the existing conversion path. While commit 1191285 fixes an inconsistency between actual_brightness and max_brightness, it exposes a range that the current implementation does not handle correctly. For stable kernels, reverting the commit seems to me the safest option: it restores the previous behaviour and avoids introducing a broader redesign into stable. Longer term, I think the AMD backlight code would benefit from a proper rework of the brightness coordinate systems. In particular, it seems cleaner to expose a fixed logical DC range (e.g. 0..65535), map that symmetrically onto the firmware/hardware [min, max] range, and keep the firmware custom-curve 0..255 scale as a separate conversion where required. That larger change could then be reviewed and tested upstream independently of the stable regression fix. Thanks, Sergio