Linux Development

Hi All…

I have just started work at improving the netboot images for the Cobalt port. Here you’ll find my latest netboot images. What has changed, is that instead of distributing the raw kernel, these are now distributed as a tarball (uncompressed — bzipping them actually made them larger) complete with CoLo, default.colo startup script, and symbolic links for most MIPS-based Cobalt systems.

To use. On the server, download the .tar file (and optionally run it through gpg, sha1sum and md5sum to verify you downloaded it right). Then unpack it on your server by running: tar -C / -xvf nfsroot-...cobalt.tar.

Then, when booting the Cobalt server… hold down left & right arrow buttons. You should see “Net Booting” appear on the display, followed shortly after by a prompt asking if you want to enable the Console. “Console ON” should be already selected, just press ENTER here.

You’ll then be asked where to boot from. Scroll down two places, and you should see “Network Boot (NFS)”. Select it and press ENTER. You should see “Loading Gentoo/MIPS” on the LCD panel, and the serial console should display the bootup progress. You’ll finally arrive at the “Press ENTER to enable this console” prompt that you should all be familiar with.

If, however, you need kernel parameters for your machine (such as Cobalt 2700 users, who need to disable serial console and start telnet) — select the Boot Shell method, and boot in the traditional way. The kernel image is named kernel.gz.

I’ll be updating the Handbook shortly to reflect this. I have an open bug covering the revamp of documentation. My draft handbook is viewable in the usual place, and feedback is encouraged. A lot has changed in the Linux/MIPS world since the docs were last overhauled, and so I aim to bring the docs up-to-date.

I shall keep you all posted as the work progresses.

Yep, after 2 power failures, weeks of anticipation, and much hair pulling on my part (good thing though, I need a haircut anyways), the full set of stages is finally built and uploaded. You can find them on my devspace in the usual place.

These haven’t undergone much in the way of testing… so it would be greatly appreciated if some brave souls could give these a try and report back on how they go. All going well, we should be able to get these pushed out to mirrors by the end of the month. 🙂

I’ll also start knocking up some new netboot images shortly. The new images will differ from the old ones, in that you will no longer have to download and extract CoLo into /nfsroot. I’m planning to produce images which can just be untarred directly into /nfsroot, and contain all the necessary files for boot. It’d also be nice to rid ourselves with the past time of mucking with the shell. All going well here, I’ll be re-writing that section of the handbook, and things should become significantly more simpler.

But we’ll see — for now I’m breathing a sigh of relief, but soon I’ll be back into the slog that is end-of-semester exams, so I don’t have long just now. Likely, this will be a summer holiday project. (Yes, for those of you playing along in the Northern Hemisphere…temperatures are warming up down here in Brisbane.)

As always, I’ll keep you posted on further notices, and I’m usually online if people wish to contact me.
Regards,
Stuart Longland

Well, ain’t it lovely… I’m busy fixing a few things up in a local CVS repository…. I hit cvs commit, as you do… then all a sudden the lights go out, my screen goes blank, and I’m left in the dark — totally silent except two computer UPS’s screaming about the loss of power. As for my stage 3 tarball compilation on my Qube2 … up in smoke.

So I get up, and crash around in the dark looking for my torch so I can check the meter box (it hasn’t been unusual for the earth leakage to trip out without reason). No sooner do I get just 3 meters, the lights come back on. Okay, fine, it was late, so I just switched the rest of my computer off (namely the monitor, desklamp and tuner) and headded for bed. Then the power fails again. Lovely.

I found a torch and went looking at the meter box… Hrmm, not our end, and looking around, I see other people looking around in the darkness. Unfortunately in this case, the outage lasts longer than my UPS’s can supply power — so we go offline.

The power was restored at around 1:00AM, but unfortunately this means my compile for the Cobalt stages is now put back a whole couple of days, as the build must now start again. I’ve now pieced my network back together (namely a couple of IRC bots were offline, and my IPv6 tunnel was nonfunctional) and so I’m back online again.

Hopefully there won’t be any further interruptions, and this build can progress right to the end. 🙂

Well I’ve fixed it. After a lot of arguments with my Qube2, I finally have a stage 1 tarball for 2005.1, with stage 2 and 3 tarballs on the way.

I’ll be uploading the stage 1 tarball shortly (tomorrow in fact, as QUT has a slightly faster pipe, and it costs me nothing to upload there), and will upload the stage 2 and 3 ones later as they are built.

I’ll also be uploading binpkg’s, for those who wish to just upgrade their existing Cobalts to the latest Gentoo release.

So yeah, Gentoo/MIPS Cobalt 2005.1 comming soon, to a mirror near you. 🙂

I’m really beginning to dispise glibc, if I wasn’t doing so already.

For those trying to update their boxes from an existing 2005.0 installation, you should be fine now. glibc-2.3.5 was recently bumped to stable, as has binutils-2.16.1.

However, for whatever reason, catalyst doesn’t like me, or glibc. Thus stages are still in hiatus at the moment while I sort the issues out.

Hi All,

Those who are trying to update their boxes, may have struck problems building the latest versions of glibc. It appears, a bug in the binutils package, is causing the following error message to be spat out, and thus causes portage to bomb out:

/var/tmp/portage/glibc-2.3.4.20050125-r1/temp/ccyZTFe5.s: Assembler messages:
/var/tmp/portage/glibc-2.3.4.20050125-r1/temp/ccyZTFe5.s:131: Error: operation combines symbols in different segments

The build usually falls apart from there. This problem affects both 2.3.4 and 2.3.5. Currently, progress on building 2005.1 stages is halted due to this issue.

At the moment, I’m trialling binutils-2.16.1 (as opposed to 2.x.9x.x.x series, which have known issues), and there are plans to potentially mark it stable on mips. glibc-2.3.5 is a definate candidate for a bump to stable.

So, for those who have hit this problem… try upgrading using the following commands:

# ACCEPT_KEYWORDS="~mips" emerge =binutils-2.16.1
# source /etc/profile
# ACCEPT_KEYWORDS="~mips" emerge =glibc-2.3.5

If you’re paranoid, do the following, first, before attempting an upgrade:

# quickpkg /var/db/pkg/sys-libs/glibc* /var/db/pkg/sys-devel/binutils*

then store the resulting packages somewhere handy, as you’ll need them if you have to recover your system.

We hope to have the situation rectified in the very near future.

Regards,
Stuart Longland.

Hi All,

For those wondering where the 2005.1 Cobalt stages are, read on.

I have just started updating my Qube’s portage tree. Once that is done, I’ll start making a seed stage (a crude stage 3 image) before the build commences in earnest.

As with last time, I’ll upload the binary packages during the stage3 build phase as they are compiled, these will go into my devspace.

Please note, the changes to CoLo with respect to menus. This release will have the newer CoLo loader, and the syntax has changed slightly. See my previous post on the topic for details… I’ll be updating the documentation as soon as I have arcload nutted out.

I’ll keep you posted. 🙂

Hi All,

I will shortly (i.e. probably this weekend) will move CoLo 1.16 into
stable (mips, currently in ~mips), and eventually remove CoLo 1.13 from
the tree.

In the newer release, the syntax for your CoLo scripts does change
slightly. In particular, the “menu” command (covered in the handbook)
has been changed to the “select” command. It differs, in that rather
than spitting out a text label, it spits out a numerical ID that you use
with the goto command. You’ll therefore need to update your scripts.

Below is an example of the old ‘menu’ command and the new ‘select’ command.

The old ‘menu’ command:

#:CoLo:#
 
lcd "Mounting hda1"
mount hda1
menu "Which Kernel?" 50 Working working New new
lcd "Loading Linux" {menu-option}
load /vmlinux.gz.{menu-option}
lcd "Booting..."
execute root=/dev/hda5 ro console=ttyS0,115200
boot

The equivalent script, using ‘select’:

#:CoLo:#

lcd "Mounting hda1"
mount hda1
select "Which Kernel?" 50 Working New

goto {menu-option}
var image-name vmlinux.gz.working
goto 3f
@var image-name vmlinux.gz.working
goto 2f
@var image-name vmlinux.gz.new

@lcd "Loading Linux" {image-name}
load /{image-name}
lcd "Booting..."
execute root=/dev/hda5 ro console=ttyS0,115200
boot

It is advised that you have a look at the documentation that’s
distributed with CoLo. I’ll advise you all once the update is complete.

Regards,
Stuart Longland