On the new Intel Q45 chipset (Eaglelake) a simple echo -n mem > /sys/power/state means a system reset when hitting the power button to try to wake up. I assume that not wakeup fails but going to sleep fails because of the following: If I disable the PWRB wakeup by echo PWRB > /proc/acpi/wakeup, I still can "wake up" the box by pressing the button. (wakeup here means to a rebooted system). So I presume that the box is not in S3 but simply shut down. Otherwise it should refuse to wake up. This is on a plain Q45 box, no additional hardware.
Created attachment 19626 [details] lspci
Created attachment 19627 [details] dmesg
Will you please attach the output of acpidump? Thanks.
Created attachment 19628 [details] try the debug patch in which rsdt is used instead of XSDT Will you please try the debug patch and see whether the issue still exists?(It is noted that the boot option of "acpi=rsdt" should be added). It is great if you can confirm whether the suspend/resume can work well on windows Vista. Thanks.
we already have two patches for this issue. you can either get the latest acpi test tree source code or wait until 2.6.29-rc1 released.
BTW: in order to get the acpi test tree, please look at http://www.lesswatts.org/projects/acpi/git.php Or you can apply these two patches on the latest vanilla kernel tree. http://marc.info/?l=linux-acpi&m=122647871228157&w=2 http://marc.info/?l=linux-acpi&m=122647870428153&w=2
Thank you Zhang Rui, the last two patches solves the wake issue! Thanks to both of you for the very quick response. At the end my buying decision was heavily influenced by the premium commitment to Linux support by Intel.
Created attachment 19633 [details] acpidump acpidump. Just for reference.
the patches in comment #6 are dropped because they are conflict with another patch. I'll refresh the patch and attach them here soon.
Created attachment 19714 [details] patch-01: install fadt.facs to the global root table list
Created attachment 19715 [details] patch-02: set the 32bit waking vector in fadt.facs
Fruhwirth, it would be great if you can try these two patches on top of len's acpica tree. in order to get this tree, you can: $<cd to your linux git tree> $git remote add linux-acpi-2.6 git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux-acpi-2.6.git $git remote update $git branch -r linux-acpi-2.6/acpica linux-acpi-2.6/linus linux-acpi-2.6/release linux-acpi-2.6/suspend linux-acpi-2.6/test origin/HEAD origin/master $git checkout linux-acpi-2.6/acpica
Thank you. I tested the ACPI branch with the two patches applied. The box successfully wakes up from sleep now instead of resetting. The only minor problem is that is does not recover the console video (that's some sort of GMA4500 variant in there iirc). I have not yet toyed with the s3bios option, however it's only a minor issue as when switching back to X video is restored. I'm using dual DVI output.
Re: patches We're expecting these to come to Linux via ACPICA shortly. Re: video I believe that all the Intel graphics adapters should resume if CONFIG_DRM_I915=y
can you try if S3 works under windows vista? From our guess, S3 should not work under vista too.
I don't have Vista. Also the patch for acpi works for me, and S3 resume is no problem. So, please close this bug. The only problem for me is that console mode doesn't resume properly, but console mode has problems in general when X is running under intel's 2.4.2 drivers.
the patch in comment #10 and #11 won't be shipped upstream as it may bring some side effects. But I think you can also use acpi=rsdt to workaround this problem.
I upgraded to 2.6.29 now. It appears that the patch was dropped, as you said. However, the workaround acpi=rsdt works fine. Thanks.