suspend/resume panic in ACPI..

Marc van Kempen marc at
Tue Aug 24 07:07:09 PDT 2004

Søren Schmidt wrote:

> Nate Lawson wrote:
>> Søren Schmidt wrote:
>>> Nate Lawson wrote:
>>>> Try the attached patch.
>>> OK, that makes the box suspend alright, but neither that nor any 
>>> other box or laptop I've got can get out of suspend without locking up.
>>> If I revert just /sys/dev/acpica and /sys/i386/acpica to aug-01 both 
>>> suspend and resume works (sortof, it does tend to panic later on).
>>> It starts failing around aug-05 and then detoriates from that on to 
>>> the current state of locking up hard on resume.
>>> It also fails in the exact same way on non-ATA systems, so I think 
>>> I'm out of the loop this time :)
>> This may be fixed now in both current and releng_5.  A logic 
>> inversion bug (= / !=) was causing pci irq link programming on resume 
>> to fail.  It didn't impact many people since a lot of systems have 
>> chipsets that keep power to irq links while suspended.
>> Please test again.
> There is no change, the systems all lock up hard on resume, on the two 
> laptops (ASUS & Acer) the backlight doesn't even come on anymore....
> Again reverting /sys/dev/acpica and /sys/i386/acpica back to aug 1st 
> make things work (well almost, but that might not be ACPI's fault)...
> This is a showstopper bug for 5.3 IMHO...
I don't know if I'm just lucky now and who is responsible, but after 
updating to RELENG_5 about an hour ago, my suspend problems seem to be 
fixed. The only thing that doesn't work (and I can't remember now if it 
has worked in before) is that using dri in XF86Config locks the display 
(not the machine, ctrl-alt-del will reboot it cleanly). Uncommenting dri 
makes suspend work ok.

This is with XFree86-4.3.0 on a thinkpad r40.

Cheers and thanks,

More information about the freebsd-current mailing list