Most recent kernel where this bug did not occur: no later kernel tested
Distribution: Debian Etch/Sid
Hardware Environment: IBM ThinkPad T23, Pentium 3, 1.13 GHz, 384 MB RAM
Software Environment: Kernel 2.6.15.4 with sws2 patches
Problem Description:

Suspend to RAM hangs hard on resume. Once I got a text display back (without 
any error message), but the other tries I simply got a black screen with three 
white pixels somewhere near the center of the display. The machine seems to be 
locked up completely.

When I tell grub not to use vga=0x305 as kernel boot parameter the resume works 
okay.

Steps to reproduce:
1) Have a similar hardware (probably), my T23 has a SuperSavage chip, but 
savagefb is not compiled into the kernel due to conflict with savage DRI 
driver.

2) use 2.6.15.4 + sws2-2.2 (I doubt it comes from there, but I can try without 
if needed and I find time) + my kernel configuration (attached to this bug)

3) use vga=0x305 as kernel boot parameter

4) try suspend to RAM... either 

a) by hibernate with ram.conf or
b) by simple script which disabled ALSA, ISDN, unload ehci and then calls: echo 
mem > /sys/power/state
c) Not tested, but I think sufficient should be just doing: echo mem 
> /sys/power/state

Earlier on I thought that resume hang would be a regression of 2.6.15.4 (see 
bug #6091), but I am not sure about this any longer. Maybe this problem 
happened with older kernels as well. I did not use the vga parameter for a long 
time and probably, I do not remember exactly anymore, this was the reason. I 
can try with some older kernel if need be, I at least have 2.6.11-2.6.14 as 
self compiled debian packages here, some with and some without sws2 patches.

Closing bug #6091 in favor of this one.

Regards,
Martin Steigerwald