(old)Puppy Linux Discussion Forum

So, eowens2
you are using a patched 2.6.25-olpc kernel from os802 or a newer one?
Does your modified system runs fine in full install?
Can you zip and post your init? I need to do some debugging with my attempts with uPup .
I got an olpc.fth that appears to pass the arguments OK and the system looks like it boots (by the camera, mic, and stick light changes) but does not get me into console.
BTW the olpc.fth looks like this
======
\ boot script
" ro root=/dev/sda1 rootfstype=vfat console=tty0 fbcon=font:SUN12x22 video=fb:1200x900 pmedia=usbflash pdev1=sda1" to boot-file
" u:\vmlinuz" to boot-device
" u:\initrd.gz" to ramdisk
setup-smbios
unfreeze
dcon-unfreeze
visible
boot
=====
Which one are you using?

Sorry only questions and no answers.

mavrothal wrote about making a comment on Barry's blog:

How do you do that . It does not have any "register" or something in the site. Is it "cryptic"? or just an e-mail request?

If I remember correctly, if you click on "X Comments" or "No comments" at the bottom of a topic it should lead you to setting up a password and username.
There have been, and still may be, problems if you use a "bad" browser. I don't think IE works but not sure which do.

Vague I know but just trying to throw you a help line.
Dave

I also got opcode error when booting Quirky004, but it finished booting.

ttuuxxx, "pmedia=usbflash" should be there to be sure.

As to Mavrothal's issues with Barry's blog, he used Safari browser, so now he knows that he has to use Firefox or Opera.

I suggested a super-simple olpc.fth:

http://bkhome.org/blog/?viewDetailed=01338

...from my brief reading, it seems that just a 2-line olpc.fth is enough.

Regarding invalid op-code, that sure is weird. The kernel is compiled for a i486. Well, I guess I can re-compile it with the instruction set dumbed-down even more.

Mavrothal,

I am using a patched 2.6.25.15 kernel from os 802.

I have not intentionally tried to boot a full HD install. I think I unintentionally did so before I got pup_412.sfs into /mnt/tmpfs in RAM.

A few times I got thru switch_root and knocked on the door of Xorg (never got a window, though). I wish I had better notes from those attempts!

I am using the vanilla multiboot from the OLPC forums, nothing special. I do not try to pass any parameters from boot to the kernel.

Tuuxxx, thanks for the UserID/password info.

BarryK wrote:I suggested a super-simple olpc.fth:

http://bkhome.org/blog/?viewDetailed=01338

...from my brief reading, it seems that just a 2-line olpc.fth is enough.

Regarding invalid op-code, that sure is weird. The kernel is compiled for a i486. Well, I guess I can re-compile it with the instruction set dumbed-down even more.

Hi Barry here is the official Olpc linux-2.6.25 kernel configs. It also has the .config <--hidden file
That should give you a better idea, well I hope it helps
ttuuxxx

ttuuxxx,
I haven't looked at any "official" olpc kernel .config file yet. The 2.6.30.5 kernel has a lot of olpc support built in, so that .config you posted is not entirely relevant, though the cpu architecture settings will be.

Anyway, I have already recompiled the kernel. Here is my latest one-file Quirky, with the kernel configured for a Geode GX1 CPU, which should fix the invalid-opcode problem...

http://bkhome.org/blog/?viewDetailed=01339

raffy, davesurrey, Thanks!
I can now post in Barry's blog using Opera.
(Firefox is a heavy "mission-critical" browser on my XO )

Here is an 2.6.31 DOTconfig that has Aufs2, Squashfs and hfsplus ( ) activated as the only difference from the official OLPC 2.6.31 XO-1 kernel.
Used it to compile with OLPC sources from OLPC Git and the official Aufs2 standalone sources and patches from Aufs Git
Runs both Fedora 11 and uPup451 (full install) on the XO1 (with an OLPC initrd) fine.

ttuuxxx and others,
After discussion on my blog:

http://bkhome.org/blog/?viewDetailed=01339

...only the XO-1 has a Geode CPU, the XO 1.5 has a Via C7-M CPU.

So as my most recent one-file Quirky is compiled specifically for a Geode GX1 CPU, it might not work on the XO 1.5. Heck, there might even be a problem with the XO-1 as it has a Geode LX CPU and I'm only assuming that it is backwards compatible with the GX1.

I think Raffy has an XO-1, so can try the latest Quirky.

Anyway, I'll take a look at what mavrothal has posted and do another build.

ttuuxxx and others,
After discussion on my blog:

http://bkhome.org/blog/?viewDetailed=01339

...only the XO-1 has a Geode CPU, the XO 1.5 has a Via C7-M CPU.

So as my most recent one-file Quirky is compiled specifically for a Geode GX1 CPU, it might not work on the XO 1.5. Heck, there might even be a problem with the XO-1 as it has a Geode LX CPU and I'm only assuming that it is backwards compatible with the GX1.

I think Raffy has an XO-1, so can try the latest Quirky.

Anyway, I'll take a look at what mavrothal has posted and do another build.

Any geode can run using a kernel compiled for 'i586'. However, the geode lx will suffer from severe slowdowns and should use a kernel compiled for 'i486'. Using an 'i486' kernel will ensure compatibility with the XO 1.5

Sorry, I have a Geode miniPC, actually a GX2. But what differs among the newer Geodes is the companion chip, with CS5536 for newer boards, the old ones having CS5535.

There's no problem booting those Geodes with the usual proprietary BIOSes.

XO is unique because it used OpenFirmware. You'll have to read about Mitch Bradley to find out what it means.

So,dia anyone had better luck than me with the latest OLPX XO1 quirky?

I also tried Barry's 3-file version of OLPC-quirky
As I describe in Barry's blog (5th comment) despite my best efforts it still can not find the sfs files.
Please take a look and help if you can
C'mon ttuuxxx. Do something

mavrothal wrote:I also tried Barry's 3-file version of OLPC-quirky
As I describe in Barry's blog (5th comment) despite my best efforts it still can not find the sfs files.
Please take a look and help if you can
C'mon ttuuxxx. Do something

I haven't reported anything in the past 2 days because, neither his first or second attempt booted after 50 or so tries. But I haven't tried his 3 file version yet. That might be more promising, I'll try that either later on tonight or early tomorrow.
ttuuxxx

I tried the 3-file version, with the same results as mavrothal - no console messages. That means that the kernel is not right for the machine.

As I have mentioned a couple of times so far, trying to use a kernel derived from upstream sources is very unlikely to work. Some OLPC stuff has been accepted upstream, but I think there are still some loose ends.

The XO-1 does not have a PC-compatible text-mode framebuffer - no EGA character/attribute mapped display at 0xb8000. You must use fbcon, and in addition to that, there is also some Geode GX/LX magic for getting the resolution set right, some stuff for the "DCON" display swizzler chip, some GPIO stuff, and some special PCI configuration stuff to work around the fact that the Geode chipset doesn't really have PCI configuration registers. Conventional BIOS emulates those configuration registers with some ultra-gnarly System Management Mode magic. The code for that was proprietary, not to mention uncompilable since the toolchain it uses is a commercial product that is no long available. So instead of turning on System Management Mode, I wrote a Linux kernel PCI driver addition that does the PCI config spoofing (30 lines of code and some tables).

The reason I described all that is to convince you that you really, really want to start with the OLPC kernel source and fold in aufs and squashfs and whatever.

Apparently mavrothal has done just that, but from what I can tell, he doesn't know how to replace the modules in initrd.gz. If aufs and squashfs are modules instead of being hard-linked, the init isn't going to work if vmlinuz mismatches the modules in initrd.gz .

If I can get a copy of that kernel plus modules, I'll try to get this working.

Mitch Bradley

Mitch, thank you for stopping by
I certainly do no do initramfs, but in my latest attempt (5th post) the initrd modules do come form the same build as vmlinuz and load just fine (as far as the dmesg shows).
The problem appears to be that usb_storage module has unknown symbols for the
usb_storage_usb_ids
usb_usual_ignore_device
usb_usual_clear_present
usb_usual_check_type
usb_usual_set_present.
This seems to be the case also with the original Barry's version in a Starbox SB-800 (geode LX800) machine (above link, 4th post). So is not Kernel/module specific (init script?... )
I also tried with the infopup script included in the initrd and indeed can not see any device or partition, like it booted from thin air.
Any ideas for that?
Please keep close

PS. I'll upload last night's kernel rmps and srpm somewhere and post the link here shortly.

mavrothal wrote: The problem appears to be that usb_storage module has unknown symbols for the
usb_storage_usb_ids
...

If you set CONFIG_USB_LIBUSUAL you might get those symbols.

wmb@firmworks.com wrote:

mavrothal wrote: The problem appears to be that usb_storage module has unknown symbols for the
usb_storage_usb_ids
...

If you set CONFIG_USB_LIBUSUAL you might get those symbols.

Hmm, it seems that I have misread the sense of the ifeq in usb/storage/Makefile . That symbol should be present if CONFIG_USB_LIBUSUAL is not set. If it is set, the symbol should be in the module usb_libusual. Perhaps you need to modprobe usb_libusual ?