Skylake no longer resumes

[trombonehero]

It's been awhile since I've been able to test drm-next, but after updating to f7745ad (from, er, something at least a month old), my Skylake system no longer resumes. Pressing the power button doesn't cause an ACPI shutdown, so something seems to be firmly wedged.

[trombonehero]

I attempted to do a git bisect with just buildkernel/installkernel, but that doesn't seem to cause the i915 drm bits to initialize properly: there is no /dev/drm directory. Do I need a full buildworld buildkernel installkernel installworld on every iteration of the bisection?

[hselasky]

drm-next only patch the kernel

[trombonehero]

Whenever I build and install an older kernel (even one as recent as dfcc5fb), I get the following output from kldload i915kms:

info: [drm] Initialized drm 1.1.0 20060810

Also, the mode doesn't actually switch and no /dev/dri appears. Running kldload i915kms on a recent (world-matching?) kernel gives me 688 lines of output together with a mode switch and all of the expected-normal behaviour (except for a lack of suspend/resume). Is there something on the filesystem that i915kms.ko needs? Could it have something to do with the version of LLVM that's used to build the module?

[trombonehero]

The patch at https://lists.freebsd.org/pipermail/freebsd-current/2017-May/065917.html causes my machine to resume again. There are visual artifacts that make the display unusable, but that seems like a separate bug, as those were around in 390597e too (whereas others in Issue #145 report that commit as a "last known good" version).