Who is 'development' here?
Pruning ...
A pretty solid clue for a cluey developer, one would think.
Assuming you use kern.vty=vt, have you tried toggling
sysctl kern.vt.suspendswitch
?
Default is 1, ie switch to vt0 before suspending, but some systems may prefer not to.
The equivalent if kern.vty=sc is
sysctl hw.syscons.sc_no_suspend_vtswitch
which, in opposite sense, has default 0. Setting this to 1 was essential on my Thinkpad T23 for successful resume.
This may or may not have anything to do with your issue.
Sounds like it needs a more committed developer.
Good luck.
"Development" is the response I got from the developer on the bugzilla, to quote what he said:-
"It's not a mesa problem, it's due to a rewriting of the vt intergration, and the last rewrite didn't fixed it. SO yeah that's something that someone (tm) should work on but since I don't have the problem with wayland on suspend/resume and I never switch VT I don't have the motivation right now." You can read the whole thread on the linked bugzilla.
So basically he says he uses wayland, never switches VT, and wayland doesn't do a VT switch on suspend, so he doesn't see the problem. Although it seems to me that if you did do a VT switch when using wayland, the gpu hardware acceleration that the wayland compositor depends on would be off after the VT switch; and never using VT's is a major limitation.
Anyway, EXCELLENT SUGGESTION smithi on changing the kern.vt.suspendswitch , that works a treat! I actually woke up this morning wondering if there was some way to suppress the VT switch on suspend-resume, but you'd already beaten me to it
I've tested it and it works; with kern.vt_suspendswitch set to 0, xdriinfo reports "crocus" following resume, and glxgears shows full performance. So that at least fixes the problem on suspend/resume, or appears to. Two heads are better than one
I've added this workaround to the bugzilla.
The issue of losing hardware gpu accel after a manual VT switch is still present, but at least now the system is stable to suspend-resume cycles.