Re: suspend-to-RAM intel-x86 issues and tests

[Riccardo Mottola <riccardo.mottola%libero.it@localhost>]

Hi,

David Brownlee wrote:
> Again, it might be interesting to see if the regression is in
> suspend/resume with DRM or more general:)

Identifying DRM as the only cause could be useful. Identifying a common 
cause among the computers also.

Said that, let me restrict to the only computers which are already 
runnign NetBSD and on which testing a current kernel is easy.

I repeat the test, with and without DRM and with a kernel now updated to 
30 Sept.

I just login into console and put to sleep, no X, no wireless, no USB, 
no external devices.

* ThinkPad R51 - Pentium-M - Intel Video
- Standard boot (enables DRM console): I see green messages of going to 
sleep, flushing caches, but no prper sleep status, fan remains up, no 
sleep LED, black screen
- disable i915drmkms : slightly different, does not turn off screen and 
sleep LED comes up, then reboots!
- disable i915drmkms, fxp: no change

* ThinkPad T30 - Pentium 4M - ATI Video Radeon M7
- Standard boot (does not enable DRM console [1]): message about kernel 
"fatal protection fault in supervisor, sleep LED comes up, system reboots
- disable radeondrmkmsfb, radeon: no change (as expected... DRM wasn't 
running anyway)
- disable radeondrmkmsfb, radeon, fxp: no change

* ThinkPad T43 - Pentium M - ATI Video Radeon M300
- Standard boot (enables DRM console): goes to sleep properly! awakens 
too! Although the video recovery time is very long... and worked the 
first time yes, not the second :) If video does not recover, no console, 
network...
- disable radeondrmkmsfb, radeon: goes to sleep and comes back much 
quicker and also reliably. Network comes up again, as a proof


My summary would be that Radeon DRM is in better shape than Intel.
The former however causes issues in waking up.
There are other issues besides drm too thouhg!


Is there a boot option or sysctl setting to have persistent dmesg, so at 
reboot I can try to read the last messages which might be helpful?

Can I drop to kernel debugger= That could help getting a trace for the 
T30 perhaps.

Riccardo

[1] any trick to enable it? it is convenient :)