Old suspend/resume issue likely root cause

[Eric McCorkle]

On 10/17/15 11:14, Glen Barber wrote:
> On Sat, Oct 17, 2015 at 11:09:03AM -0400, Eric McCorkle wrote:
>> A while back (in May or June or so), I was attempting to figure out why my
>> laptop screen stayed blank on suspend/resume.  I posted some dmesg outputs
>> and theorized that it might be something about the devices coming back up in
>> the wrong order.  John Baldwin suggested I try doing
>> hw.pci.do_power_suspend=0.
>>
>> I got tangled up with other projects and had to set it aside.  However, I
>> now finally have more information:
>>
>>
>> * hw.pci.power_suspend=0 does cause the screen to come back up properly.
>> The network (wireless via iwm driver) needs to be re-initialized, but
>> everything else seems to work properly.
>>
>> * Without hw.pci.power_suspend=0, the issue persists even when booting from
>> EFI, so we can rule out some kind of legacy BIOS issue.  The issue manifests
>> even with the experimental i915 driver, so we can rule out something with
>> the efifb or VGA framebuffer drivers.
>>
>>
>> Based on this, it looks pretty likely that the pci bus is the culprit. Any
>> suggestions on where to look?
>
> What graphics chipset do you have?  I noticed with dumbbell's i915kms
> update branch [1], suspend/resume now works for me with Haswell
> graphics, whereas previously I would experience the same blank screen on
> resume you report.

Intel i7-4700.  I tried suspend with the i915kms driver active, hoping 
that would do the trick, but it didn't work.


> (And I also see the iwm(4) reinitialization issue, as well.)

The workaround there is just to take netif, wpa_supplicant, dhclient, 
and rtsold down in rc.suspend and bring them back up in rc.resume.

> [1] https://github.com/freebsd/freebsd-base-graphics.git branch
>      drm-i915-update-38
>
> Glen
>