Bug 4298

Distribution: Fedora Core 3

Hardware Environment:
AMD Sempron 3100+ (32-bit K8) desktop with 1.5GB RAM; also reproduced on a
Toshiba Portege 3500 laptop with 512MB RAM during 2.6.11-rc, although I had no
clue what was causing the bug at that point

Software Environment:
.config is similar to what ships in 2.6.10-2.6.11 "rawhide"/"fc-devel" kernels;
this means ACPI is enabled

Problem Description:
swsusp normally works, but if I enable CONFIG_DEBUG_PAGEALLOC, it breaks --
after "PM: snapshotting memory", swsusp never gets to the "critical section" and
it kind of bounces back from the suspend attempt, for lack of a better way for
me to describe the effect.

I first hit this when trying to hack a vendor kernel into supporting swsusp
(guess which one ;) -- sometimes things worked and sometimes they didn't. I
later reproduced it on mainline 2.6.11-rc1 or -rc2 -- I don't remember which --
but it seemed to go away for -rc4. Then I hit the problem again on 2.6.11-rc5
and 2.6.11 final.

Unfortunately it took me several weeks (i.e. until today) to figure out that I
was being sloppy with my .configs and it was actually changes in those that were
to blame for my troubles. I just finished narrowing it down to
CONFIG_DEBUG_PAGEALLOC.

Steps to reproduce:
1. Using either a pristine mainline tarball (with vanilla patches if you want a
-rc or -bk kernel), or a hacked Fedora Core 3/4 kernel SRPM, compile a kernel
with swsusp support. (Compared to the standard Fedora Core .config, set
CONFIG_SOFTWARE_SUSPEND=y and CONFIG_PM_STD_PARTITION="".)
2. Install this new kernel.
3. If needed, edit /etc/grub.conf to make the new kernel the default. Also make
sure that the new kernel has a "resume=" parameter 
4. Reboot into the new kernel.
5. Log in as root (or log in as a normal user and become root).
6. echo 4 > /proc/acpi/sleep