Dell Inspiron 8000/with ATI Mobility M4, after returning from Suspend to Disk, X runs extremely slow (slow redraws). In VC1 when switching runlevels we get: "MTRR not used" --------------/var/log/messages------------ Feb 20 09:58:33 localhost netfs: Mounting other filesystems: succeeded Feb 20 09:58:34 localhost apmd[699]: Normal Resume after 00:02:31 (100% 5:45) AC power Feb 20 10:00:12 localhost login(pam_unix)[991]: session opened for user root by LOGIN(uid=0) Feb 20 10:00:12 localhost -- root[991]: ROOT LOGIN ON tty1 Feb 20 10:00:15 localhost gdm(pam_unix)[1007]: session closed for user root Feb 20 10:00:15 localhost gnome-name-server[1126]: input condition is: 0x11, exiting Feb 20 10:00:15 localhost kernel: [drm:drm_release] *ERROR* Process 1006 dead, freeing lock for context 1 Feb 20 10:00:15 localhost kernel: mtrr: MTRR 3 not used Feb 20 10:00:15 localhost kernel: mtrr: MTRR 3 not used Feb 20 10:00:15 localhost kernel: mtrr: reg 3 not used ------------ Adding "alias agpgart off" to /etc/modules.conf resolves the issue.
This is either a problem with your BIOS or a kernel bug. I tend to believe it's the former, no problems whatsoever here with a Gericom 3xC.
I agree with bero, this looks very much like a BIOS bug to me; it looks like the BIOS is not restoring agp state correctly. Please check with your APM BIOS implementors on this one.
The system will now properly resume from suspend to ram and suspend to disk without video slowdown. However, after a resume when trying to run gears, the system will lock up.
gears is the DRI screensaver, right ? comparing "lspci -vxxx" before and after the suspend would be interesting; I wouldn't be surprised if the bios doesn't restore AGP state exactly
When I looked at this before the 8000 seemed to be leaving the AGP state trashed across a power management event and also not restoring the MTRR registers. The current tree has agp save/restore hooks if we know what needs to be saved and restored...
I8000 laptop here even without X doesn't survive a suspend at all. The screen comes back but the kernel appears completely dead - the 'apm -s' command doesn't return and keyboard/mouse/network are dead. Probably dying in an IRQ storm? Ironically, it does actually manage to survive a suspend-to-disk. And even more strangely, once it's survived a suspend-to-disk the APM suspend also starts to mostly work, until the next reboot. I say 'mostly' because the IRQ of the built-in Ethernet seems to go AWOL after the APM suspend - although it was working OK after the suspend-to-disk; only subsequent APM suspend breaks it. This is with the latest (A12) BIOS.