Suspend and AGP in 2.6.0-test9

From: Maciej Freudenheim
Date: Mon Nov 03 2003 - 07:36:09 EST


2.6.0-test9 is the first kernel with *almost* working suspend to ram
and suspend to disk.

Well, almost, because when i run X with AGP support enabled i can't
resume - either it hangs forever after displaying "Waiting for DMAs to
settle down..." message or it reboots right after displaying it.
When i force using of PCI bus for acceleration, then suspend (to ram as
well as to disk) with resume works perfectly - but of course, with disabled
agp everything is significantly slower.

So, my question is - is it known (and not fixable :) bug or it's
something weird and shouldn't happen ? As fair as I googled for similar
problems I have found that people usually have problems with DRI, it looks
like agp works ok for most of them :) However, on my laptop disabling
DRI doesn't help.

I have Dell Lattitude D600 with:
- ATI Technologies Inc Radeon R250 Lf [Radeon Mobility 9000 M9]
- Intel Corp. 82855PM Processor to AGP Controllerntel Corp. 82855PM
- Latest BIOS (A06)

Here is full output from dmesg, lspci and my current kernel config:

I've tried both 2.6.0-test9 and 2.6.0-test9-mm1 with:
- XFree's radeon drivers (from XFree86's cvs) -
As i already wrote, everything work ok untill i load intel-agp
module (or don't force "BusType" to "PCI" in XF86Config) - enabling/
disabling DRI doesn't seem to affect anything.

- ATI's fglrx drivers (ver. 3.2.8) - the same, with enabled
"UseInternalAGPGART" i can't resume. When disabled and without loaded
intel-agp module i can.

I'm suspending by writing to /proc/acpi/sleep - writing to
/sys/power/state returns nice "call trace" screen.
On older versions of kernel (<= 2.6.0-test8*) suspend just doesn't work
- /proc/acpi/sleep behaves like /dev/null :)

If anyone know sollution for it *please* let me know. :>

BTW, standby mode doesn't work at all (and never had on 2.6.0-test* for
me, with 2.4 it is ok) - it goes to sleep, but when resuming it hangs
after displaying "PM: Finishing up."

Maciej Freudenheim.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at