2.6.0-test8 on Sony VAIO R505EL and PowerMac 8500 (summary)

Here's a summary of my experience with 2.6.0-test8 on a Sony VAIO R505EL and PowerMac 8500. Aside from problems with software-suspend/suspend-to- disk, my biggest issue on the VAIO R505EL (i386) is that the display only seems to work cleanly at 1024x768x8 bits (or less). It gets past that with X kludgery, but software suspend is definitely limited to 8 bit color. Hopefully once these things get resolved, i won't be running Windows XP much of the time.

PowerMac 8500 (PPC) has more serious problems, its console display is broken and it currently cannot come up multi-user, making it difficult to examine the problem. Eventually, i'll hook up a serial console and at least look at that problem. Since 2.4.21 works fine for me, this isn't a major issue for me personally. Given our winter power and summer heat, i wish software suspend worked for the PPC...

Attached is a summary of the current issues for me, please write if you would like additional details on any of these, especially those not yet reported on.

			         -- JM

P.S. If you respond to anything below, you might want to change the Subject line so people don't have to follow multiple bugs in this thread.

-------------------------------------------------------------------------------
Outstanding issues for Sony VAIO R505EL

ohci1394/sbp2 fails after 'modprobe ohci1394' with slab corruption (on
test7-bk3). Bug report filed. Partial fix available (which still loses on 'rmmod sbp' or 'rmmod ohci1394').

'modprobe parport_pc' allocates IO ports 378-37a, 37b-37f but
'rmmod parport_pc' fails to release 37b-37f. Bug report filed and patch suggested.

Inserting infrared card served by 'serial_cs' allocates IO ports 2f8-2ff but
on ejection, claims it's trying to free non-existent resource 2f8-2ff (maybe it's trying to free it twice?). Appears to be benign, details upon request.

Software suspend (old flavor) does seem to work under limited circumstances
but might be described as being somewhat fragile.

Getting started with new flavor of suspend-to-disk was difficult, i couldn't
find where documention on mounting /sys. My /etc/fstab now has:
"/dev/zero /sys sysfs defaults". Seems kludgey and
not well documented.

suspend-to-disk (both flavors) seems not to check for valid swap partition
before attempting to suspend. (And then even 'sysrq-S/sysrq-U' won't avoid an 'fsck').

suspend-to-disk (new flavor) does nothing meaningful for 'standby' except
that it blinks the screen very briefly. Provided details to existing bug report.

suspend-to-disk with 'mem'/S3 (e.g. both flavors) puts machine goes into
a coma, which requires a hard reset to revive.

New flavor of suspend-to-disk with 'echo -n disk > /sys/power/state' stops
OK, but goes into a coma when the 'power' button is pressed. After a hard reset, 'power' button then resumes normally.

suspend-to-disk works for X server utilizing 'vesafb', but double-faults
and reboots upon resuming if i run X in native mode (drm/i830). Thus i'm stuck with 256 colors if i'm running on battery power (e.g. may need to suspend). Other(s) have already filed reports on this, so supplied additional comments/information.

'vesafb' only works at depth 8 (except for 640x480 works at depth 16). It
can work at higher depths with X, but that uses kludges beyond the scope of this mailing list.

'uhci-hcd' gets confused after suspend-to-disk. Attempting to 'rmmod' it
after suspending gets an oops. Workaround is to 'rmmod uhci-hcd' in the hibernate script, and i also have a private patch which may fix this (more testing needed). Bug report filed on this.

New flavor of suspend-to-disk under 'test7-bk7' would sometimes get
backtraces going to sleep, but they most seem to be harmless. Not yet checked yet under -test8.

Outstanding issues for PowerMac 8500

'drivers/block/swim3.c' won't compile, seems to be missing a couple of

	#includes's (and still gets a couple of warnings).  Bug report filed,
	temporary patch suggested.

Many other modules get compilation warnings. Bug report filed.

PPC gets one 'oops' per SCSI disk during boot up. Bug report filed.

Contrary to earlier reports, RAID5 seems to work fine on PPC (on -test7 at

	least).

Console comes up in an unusable video mode, specifically, sync is good, but
it appears to have wrong number of bits per scanline.

X eventually resets it to good video mode on test7-bk7, but X doesn't work
with an ADB keyboard (and it even doesn't get that far under -test8)

Currently can only test over network (and as noted, test7-bk7 was last
kernel i tried that even got that far.

May no longer fit on floppy, somewhat dangerous given i can't boot from CD
(not even MacOS) and have RAID5 root/usr.

===============================================================================

Regarding the pmac 8500, I'd appreciate some tests with the kernel at bk://ppc.bkbits.net/linuxppc-2.5-benh

Also, please CC me issues relative to PowerMacs in general, or at least CC them to linuxppc-dev, as I may miss them in the lkml traffic.

Real patch on the way. Right now, Linus tree is frozen regarding my driver updates though, and it's ok as far as I'm concerned, I think we'll have pmac support fully merged with 2.6.0.

CC me please, and test my tree as it's where the patches will come from ultimately ;)

Known problem. Seem to be a HW bug triggered by the slab debug code, I need to spend more time figuring out exactly what's going on in the chip.

Good.

There are several problems with controlfb and the new fbcon that I haven't fully figured out yet. Though it works for me when I boot (but strangely fails when I switch back to VT from X).

Ah ? Works fine here.

I fixed vmlinux.coff so it boots from network at least, quik should work with small kernels, but I posted a quik patch that may help with larger kernels. I plan to add yaboot support for oldworld sooner or later that will deal with that issue as well.