Older blog entries for spot (starting at number 88)


Auroraconda gets to stage 2! There are still some quirks in stage 1, but man, it was pretty to see it roll into Stage 2 today. Mind you, it didn't get too far, but in the words of Jeremy Katz, "Stage 2 is a lot easier to fix".

Doing a lot of work on the rescuecd code. I know know a lot more about mkisofs on sparc and silo now. For the time being, I'm going to be content to have it boot into a bash shell with all the utilities and kernel modules present.

Kernel is pretty much frozen at 0.96. I know as soon as I say this, it will need fixing in some other area.

Dave Miller confirmed my assumptions about the SunFire 280R, the UltraSPARC III+ CPUs aren't supported by Linux yet.

The loss of my laptop is limiting the amount of work I can do from home. And when I say limiting, I mean, I'm not doing ANY work from home. I already can't wait for its return from Dell. Of course, Tadpole could offer to send me a SPARCbook to fill the void. Yeah. Thats not gonna happen.

Today was rough. Velociraptor, my Dell Inspiron 4000, refused to boot beyond its BIOS today. I called Dell to have it RMAd, to be informed that TODAY was the last day on my warranty. They decided that I was still covered today, but that to be safe, I really should purchase an extended warranty, which i did, for a little over $200. That was a quite unexpected expense. :/


I bought two Sparcstation LXs over the weekend, and yesterday I managed to coax a very minimal install of Red Hat Linux 6.2 onto the first one... and today, when I booted it, I discovered that its ethernet port is dead. Hopefully the other one will have a working ethernet port.

My manager convinced Red Hat to purchase (on loan) a Sunfire 280R, for some sort of traveling demo. He's letting me play with it for a week or so, before he needs to blank it and install Slowlaris. Red Hat Linux 6.2 doesn't boot on it, and auroraconda isn't quite ready yet, so I'm probably going to install SuSE and manually bootstrap Aurora over top of it.

SPARCs currently on my alternate desk at work:

(1) Ultra 2 SMP (1) Ultra 10 (2) Sparcstation LX (1) Sparcstation 2 (1) AXi

That SunFire makes more noise than an s/390.

Poked at the rescue cd code, I'm thinking that would be a nifty LinuxWorld giveaway. It will probably need the further auroraconda fixes as well. Anaconda needs to be reminded that, no, sparc doesn't want to use pcmcia. :) If tadpole sends me a SPARCbook, i'll reconsider.


Looked at the kernel spec file for a long time last night at work, trying to think of a sane way to apply a few patches to an "enterprise" kernel, and ONLY to an enterprise kernel. Unfortunately, the only way of doing so would also required those patches to apply to the kernel-source bundle, which will make dealing with all the other kernels unpleasant with in some situations. I don't think I'll be able to do that, short of doing one of the following: 1) have the spec generate a kernel-enterprise-source package in addition to kernel-enterprise 2) write a new spec just for kernel-enterprise & kernel-enterprise-source packages.

And since neither of those ideas is anything that I'd really prefer to do, I think this is what I'm going to do. The src.rpm will grab the patches, they just aren't going to be applied. I'll test the kernel with these patches after Build 0.3 releases, and if it holds up, I'll put them into the mainstream kernel RPM for the next Build.

Which patches are these? The RML realtime & Ingo's O1 for sparc64.

I think this is a ok decision anyways, since cache_decay_ticks isn't calculating sane values with the current algorithm + patchset, and I'd have to force it to 1 (which isn't end of the world bad, but its not optimal) to keep SMP actions from getting slowed down.

But lemme tell you, cleaning up patches so that they merge against Aurora is fun. Fun fun fun. First thing I'm going to do after Build 0.3 is start working on a 2.4.19 based kernel (not 2.4.18pre6, or any of the pres for that matter). Then again, it will probably be as bad as it is now, since most of my patches will appear in 2.4.19 anyways. Ah well, I can dream can't I?

In unrelated news, I'm going to the beach for a week, starting Saturday. This means, no Aurora, no emails, no nothing for a whole week. I'm going to try to enjoy it.

6 Jun 2002 (updated 6 Jun 2002 at 11:59 UTC) »
thomasvs: you're welcome. ;)

and now, the brief Aurora update:

I'm still waiting for Tom Duffy to send me his o(1) patches, so that I can go ahead and build 0.96 (it would just be 0.95-enterprise, but I'm going to put the sun4c forkbomb patch into the main kernel at the same time). if they dont come in by the end of the day today, i'm just going to disable the enterprise build in the spec and build 0.96...because...

Jeremy seems to have resolved one of the major auroraconda issues last night, everywhere i turn, theres a note from him saying "i think i fixed auroraconda". thats a good sign. ;)

I still need to respin the 6.2 ISO after inflating the initrd such that it takes up 2.88M of boot image total, and see if it will boot. If it does, isos are much much easier, if it doesn't, we're going to have to do some serious serious black magic to get the installer in 1.44M. Which in and of itself isn't necessarily so bad. We'd need that to happen anyways if we want boot floppies. Right now, I'd settle for bootable isos.

And if I can have something for a few people to test before I go to the beach on Saturday (i'll be gone for a week), that would be excellent. But, one thing at a time.

2 Jun 2002 (updated 2 Jun 2002 at 17:13 UTC) »

Some of the Toshibas (usually the Satellites, but also, occasionally the Tecras) refuse to boot off cdrom media with bootimages larger than 1.44M (7.0 had a 1.44M boot image, 7.1+ has 2.88M boot images).

Reference: http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=47288


The original bug poster for the sun4c problem submitted a patch that seems to resolve it. I'm still confused about using a sun4c as a high traffic mailserver, but hey, whatever floats your boat. I'll add the patch and respin the kernel on Monday.

Sun is now offering a SunBlade 100 for the show instead of the Ultra 10. They offered me a whole list of stuff, and I picked the SunBlade (so did the Debian team). It will give me a chance to try to fix some of the X issues, and hopefully, we'll have a working installer by August. *cough*

We're in the middle of the migration, and the mailing lists are down. Just be patient, they'll come back up, I promise.


Thank you to all the people who offered up hosting space and servers. Aurora is completely covered at this point, its just a race against the clock to move all the files off linuxpower.org. Shouldn't be a problem.

Spent all my time today working on my prtdiag script. About all I can do for Aurora now is sit and wait for the installer work to be done. Well, that and poke the people who are working on it with sticks.

I know Xconfigurator needs some significant rewrites to deal with the Blades, but I really need one of them in a functional state (and sitting, attached to a monitor, in front of me) to do the work. Thus, this won't happen anytime soon.

I'm not terribly concerned with sun4c being vulnerable to forkbombs. The fact that it boots without eating you alive is miracle enough. If you're seriously trying to use it as a high yield mailserver, then you're sick enough to fix the code yourself.

I'm going to add mdadm to Aurora, even though Red Hat hasn't done so yet. Its proved its worth in migrating sparc 2.2 RAID to sparc 2.4 RAID.

Oh, and the Carolina Hurricanes are in the Stanley Cup Finals. How bout them apples? :)

Hey! Red Hat patented something that they plan on always giving away, so that no one would ever patent it and not give it away *cough*Sistina*cough*. Lets all grab our pitchforks and torches and be idiots in the streets! There'll be looting!

And now, on to the world of real-life relevance... well, not quite, but close:


Unless something changes drastically between now and release, 0.95sparc will be the kernel for Build 0.3. And yes, I know I haven't put it up yet. linuxpower.org is going away at the end of the month, and it seems to be down altogether now. I haven't really switched hosting over yet. I should do that though.

Installed the Build 0.3 tree on the SMP Ultra 2. It seemed very happy, until I ran Xconfigurator, and it blew up in my face. So, I spent most of yesterday's freetime debugging huge chunks of the code, trying to find where it breaks. I ended up rewriting a section of the SBUS probing so that it works correctly, and making it always be nodri (I'm not aware of any DRM for sparc video cards, although, in theory, i think the ati DRM might work). The new Xconfigurator generates a valid config for 24 bit 1024x768 on the Creator 3d, although, the "valid modes" list was wrong. Have to look at that code at some point. Also had to edit hwdata, since it had invalid entries. I'm sure there are a lot more in there, but I don't have all the hardware to test.

Booted a test installer kernel image today, it got to the tui "Language Selection" screen and sig11'd when I chose English. Perhaps it wanted Redneck. Well, its a start.

Sun has offered to loan me an Ultra 10 + flat panel monitor for LWCE in San Francisco during the second week of August. I'm tempted to go, if any of the Aurora readers/volunteers feel like hanging out in the booth, let me know via email: (tcallawa@redhat.com). Don't officially have a booth yet, but it can't hurt to go ahead and see if theres any interest, or if it would just be me.

I've been a little distracted lately. My gf was hospitalized over the weekend for a blood clot between her lungs, so I haven't been devoting much time to Aurora. Build 0.3 is essentially done. I'm really just kicking the tires, hoping the auroraconda workers will have something bootable in time. Anyways, shes back home now, recovering, so heres the little update.


Rebuilt the kernel a few more times. Adjusted the patchset according to Dave M's recommendations, added the sun4c patch. Also stripped down the sparc64 boot kernel even more. Its the primary target for auroraconda. sparc32 will come afterwards.

Booted build 0.3 on the ss2. The hardware is very unhappy, I think it simply is flaky hardware. Kernel boots, but it has unpredictable behavior. The hard drive it shipped with was DOA, so I'm using one of my spares for testing. X started, as did Gnome 1.4, but Nautilus refused to do anything except complain that it wasn't going to run (after 10 minutes of grinding). If you're using an ss2 as an X terminal, you need to spend the 100 bucks and get a new computer. No offense.

ErikLevy: Congrats. :)


Was going to announce this yesterday, but I got really really busy.

This is a screenshot of KDE3 running on Aurora Build 0.3 with Konqueror & Kword (and the Mosfet Liquid Theme) on my Ultra 10. What I originally thought were lockups were just KDE being extremely slow to start up. Guess they need more than a gb of RAM and a half gb of swap to do the job....funny how GNOME doesn't. However, they run fine once they're started.

Screenshot of KDE3 on Aurora SPARC

Oh yeah, and I caught another bug in Xconfigurator and fixed it yesterday. Here's hoping it works for other machines besides my u10. :)


Got fed up with having to do X by hand, turned green, tore my shirt off, and fixed Xconfigurator/mouseconfig. spot smash!

This is excellent, I hadn't really planned on having them working. Then again, I hadn't planned on the fix being so simple. Having a working XF86Config-4 really helped. Dunno where else it works, but it works on an Ati Mach64 on the Ultra 10. Going to borrow Peter's Creator to check it out.

Kernel boots on sparc32 UP & SMP (SMP still has known issues, spinlocks and swapon segfaulting), boots on sparc64 UP & SMP (the sunblades seem to have specific evil issues that I'm still trying to resolve).

Build 0.3 tree is very very close to complete. When I freeze the kernel, I'll hand it to Jeremy and say "Make it (Auroraconda) go!".

79 older entries...

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!