Re: Acer TM 661 does not wake up from STR

[Nate Lawson <nate-Y6VGUYTwhu0@public.gmane.org>]

Mirko Werneke wrote:
> I've a really big problem with my Acer Travelmate 661 laptop: Once 
> suspended to ram, it won't wake up again! The laptop seems to suspend 
> corretly (no error messages while suspending; Power LED switches from 
> green to orange). But after a key press to wake it up the only reaction 
> the laptop shows is switching the power led to green again. Nothing more 
> happens. I've to press the power button for 4 seconds to switch the 
> laptop completely off. After that i can reboot normally.
> 
> Besides from playing around withh several parameters ("noapic", 
> "pci=routeirq", "acpi_sleep=s3_bios"/"s3_mode") i tried with help from 
> Stefan Dösinger the following:
> 
> I compiled a minimal kernel (without usb/wlan drivers, etc.) with a 
> little patch in wakeup.S.
> I inserted after .code16 the line "lcall $0xffff,$0". This should - 
> according to Stefan - immediatly call the BIOS reset code after the 
> kernel gets control back from BIOS. But the situation did not change (no 
> "automatic" reboot after resume). So we came to the conclusion, that the 
> kernel does not get the control back when the system wakes up.
> So i think that there is a problem with the BIOS or the kernel does 
> something wrong while suspending.

So the BIOS is not calling the wake vector.  I've been wondering about 
several deviations from the spec in the sleep path (hwsleep.c).

I think _GTS should be run immediately before writing the sleep value in 
EnterSleepState (not Prep).  Section 7.3.3:  "OSPM will set the sleep 
enable (SLP_EN) bit in the PM1 control register immediately following 
the execution of the _GTS control method without performing any other 
physical I/O or allowing any interrupt servicing."

I'm suspicious of writing the SLP_TYP and SLP_EN on separate passes. 
The sample code in the spec (7.3.4) writes this in one pass.  If there 
are systems that require multi-pass write, we should leave this as an 
option, disabled by default, since I think most systems do this correctly.

We should check the return value of \_WAK and at least print the result 
code if non-zero. (7.3.5)

One non-standard thing is that I think LeaveSleepState should read 
values from \\_S0 and if this method is present, write those to slp reg 
A/B in the resume path.

-- 
Nate


-------------------------------------------------------
SF.Net email is sponsored by Shop4tech.com-Lowest price on Blank Media
100pk Sonic DVD-R 4x for only $29 -100pk Sonic DVD+R for only $33
Save 50% off Retail on Ink & Toner - Free Shipping and Free Gift.
http://www.shop4tech.com/z/Inkjet_Cartridges/9_108_r285