| twb | Why is lucid's running the same old syslinux as jaunty? | 10:01 |
|---|---|---|
| twb | I wanted to use the "isohybrid" feature that allows me to make an install image that's both a bootable ISO 9660 image and a bootable USB hard disk image, but this feature was introduced around 3.70 -- which even Debian Lenny has :-/ | 10:02 |
| cjwatson | because it was really hard work to merge and I didn't have time to take on the QA load | 10:04 |
| cjwatson | due to gfxboot, our syslinux is heavily patched. hopefully that delta should drop with the update, but it was going to be hard work to make sure of that | 10:05 |
| twb | Yeah, #syslinux thought it was probably gfxboot's fault | 10:05 |
| twb | Apparently gfxboot is becoming a proper module, so that issue should disappear | 10:06 |
| twb | I guess I'll just have to work around it until 12.04 | 10:06 |
| applejack | hello, how would I go about installing 10.04 on a blue/white g3? the cmd64x driver doesn't seem to be included, and the ata-generic driver won't see the harddrive | 10:19 |
| twb | applejack: Ubuntu doesn't support the POWER architecture. | 10:19 |
| applejack | hmm,... the cd boots tho | 10:20 |
| twb | I think there are some unofficial ports | 10:20 |
| applejack | let me rephrase then,. how do I stop the livecd from loading the ata-generic module? | 10:20 |
| twb | You could try playing games in "expert" mode by bouncing into a shell and setting up blacklists | 10:21 |
| twb | Or modifying the CD, of course. | 10:21 |
| applejack | yea,. well, apart from doing that | 10:21 |
| twb | If that unit supports the so-called "target ata", just plugging it into a modern machine and doing the install that way would be easiest | 10:22 |
| twb | (Or simply ripping the disk out and putting it into a modern machine while you install to it.) | 10:22 |
| applejack | yea,. done that | 10:22 |
| applejack | but yaboot is broken | 10:23 |
| applejack | it seems that _has _ to be installed on hte hardware | 10:23 |
| twb | Shrug. I haven't dealt with newworld hardware for years. | 10:23 |
| twb | If it were me, I'd first try fixing yaboot's config | 10:23 |
| applejack | hmm | 10:24 |
| cjwatson | ata-generic shouldn't actually claim the device unless there's nothing better around to drive it | 10:25 |
| cjwatson | (unless you boot with all_generic_ide) | 10:25 |
| applejack | cjwatson: that's the problem,... the cmd64x driver is not on the livecd, so there is nothing better | 10:25 |
| applejack | and it's needed for the cd I suppose | 10:26 |
| cjwatson | right. please file a bug on the Ubuntu kernel asking them to turn on cmd64x for powerpc, so that at least this gets fixed in the future | 10:26 |
| cjwatson | it's simply not built for powerpc at all, as far as I can see, it's not just that it isn't on the live CD | 10:26 |
| cjwatson | which makes a workaround kind of hard | 10:26 |
| applejack | ubuntu-10.04-desktop-powerpc.iso | 10:27 |
| cjwatson | it'd make it a customisation job - or you could plead with the kernel guys to get this into a stable update for lucid, so that we can include it in a point release or something | 10:27 |
| ev | michaelforrest2: ready? | 11:04 |
| michaelforrest2 | ev: yep - wanna come in here? | 11:09 |
| michaelforrest2 | (TV ROOM) | 11:09 |
| ev | surely | 11:09 |
| twb | OK, WTF. I just went through a netboot install of lucid, doing my usual partman configuration of /boot ext2 on RAID1 on sd[ab]0 and root ext4 on LVM on RAID1 on sd[ab]1. | 12:46 |
| twb | When I go to boot the system, I find that /boot won't assemble, /dev/sd[ab] themselves are md1, and /cat/proc/partitions is reporting partitions sd[ab]{,0,1} and md1{,p1,p2} | 12:47 |
| twb | I could do without this kind of surprise at 10PM on a Friday :-/ | 12:48 |
| applejack | sd[ab]0 ?!?? | 12:49 |
| twb | Er, sorry, off-by-one error | 12:50 |
| twb | Lemme transcribe it exactly | 12:50 |
| applejack | does it assemble if you do it manually? | 12:51 |
| twb | mdadm -A /dev/md0 --scan ? | 12:52 |
| twb | I'm also a bit grumpy about plymouth being installed on a server when the only thing I picked in tasksel was libvirt. | 12:52 |
| twb | The entries in /proc/partitions are sd[ab], md1, sd[ab][12], md1p[12], and the LVM LVs. | 12:53 |
| twb | mdadm -A /dev/md0 --scan causes it to assemble md0 with one drive -- md1p1 | 12:54 |
| applejack | mdadm --examine --scan ? | 12:55 |
| twb | I'm guessing somehow nobody bothered to test the new partitions-on-md-devices code in this particular configuration. | 12:56 |
| applejack | you need to make your root md device NOT use the first 512 byte of the drive | 12:58 |
| applejack | I forget the option to create it that way | 12:58 |
| cjwatson | --zero-superblock might be relevant | 12:59 |
| twb | Well, I didn't make it use the first 512 bytes | 12:59 |
| twb | That's where the MBR should be | 12:59 |
| cjwatson | plymouth is mandatory for all Ubuntu installs, and it's not just for splash screens, I'm not going to go into that discussion yet again | 12:59 |
| applejack | bbiab | 12:59 |
| twb | I mean: I just asked partman to lay out normal partitions. | 12:59 |
| soren | twb: plymouth is in ubuntu-minimal. It's got nothing to do with libvirt. | 13:00 |
| soren | twb: Oh, cjwatson just pointed that out. Never mind. | 13:00 |
| twb | sda was previously a SCO 5 disk; sdb was almost certainly blank. I *think* I told partman to make a new disk label, but I might misremember. | 13:00 |
| twb | It appears that openssh-server is installed, but not running. | 13:02 |
| twb | Ah, it'll be because I dropped into a recovery shell when /boot failed to mount, so I never reached the "filesystems" event that ssh.conf was waiting for. | 13:03 |
| twb | Incidentally, when using qemu -curses, is there an -append option I can add to prevent lucid from loading vga16fb? qemu -curses can render the 80x25 VGA console, but not framebuffers. | 13:09 |
| twb | (Contrariwise, I can happily report that everything Just Worked when I installed lucid on a headless/serial-only Soekris net5501 router recently.) | 13:11 |
| twb | http://pastebin.com/DrhNVj7Q is the output from the current md1p1-wacky host | 13:14 |
| cjwatson | blacklist=vga16fb should do it | 13:22 |
| twb | Thanks. | 13:22 |
| twb | I was trying all sorts of things like video=false and vga=normal to no avail | 13:22 |
| cjwatson | those ought to blacklist vga16fb too | 13:22 |
| cjwatson | which implies blacklist=vga16fb mightn't help, but you can try it ... | 13:23 |
| twb | I will when I roll my next VM | 13:23 |
| soren | It works for me, fwiw. | 13:27 |
| soren | All of those do, actually. | 13:28 |
| soren | video=false, vga=normal, and blacklist=vga16fb. | 13:28 |
| twb | Hum. | 13:28 |
| twb | soren: in qemu? What version of qemu? | 13:28 |
| soren | Whatever's in Lucid. | 13:28 |
| cjwatson | I suspect that something after the initramfs is loading vga16fb | 13:28 |
| soren | cjwatson: Ah, point. | 13:28 |
| soren | I only tested up to and including initramfs. | 13:29 |
| cjwatson | the handling for those options is specific to the initramfs | 13:29 |
| cjwatson | I don't know what, though | 13:29 |
| twb | I vaguely remember looking at the module config shortly before lucid was released | 13:29 |
| twb | IIRC it said something like "make sure we never ever load a non-KMS fb" followed immediately by "except we should always always load vga16fb" | 13:29 |
| soren | twb: With blacklist=vga16fb, I can boot all the way to a login prompt with -curses. | 13:33 |
| soren | (in maverick, at least) | 13:33 |
| twb | OK, I just reproduced it in the netboot kernel before I get as far as installation. | 13:38 |
| twb | qemu-system-x86_64 -curses -kernel linux -initrd initrd.gz -append 'blacklist=vga16fb fb=false video=false vga=normal' | 13:38 |
| twb | Using qemu0.11.1-2.ubuntu1 and mirror.internode.on.net/pub/ubuntu/ubuntu/dists/lucid/main/installer-amd64/current/images/netboot/ubuntu-installer/amd64/{linux,initrd.gz} | 13:39 |
| twb | I'll grab the archive.ubuntu.com kernel and ramdisk to be doubly sure | 13:40 |
| twb | Yeah, same cksums for archive.ubuntu.com/ubuntu/dists/lucid/main/installer-amd64/current/images/netboot/ubuntu-installer/amd64/{linux,initrd.gz} | 13:42 |
| soren | ...and that doesn't work for you? | 13:44 |
| twb | It gets as far as [ 4.676325] fb0: VGA16 VGA frame buffer device | 13:45 |
| twb | Then I get "640 x 480 Graphic mode" from qemu, and can't see anything | 13:45 |
| twb | (Actually there are a couple of messages about e.g. virtio shortly after that kmsg, but I didn't capture them.) | 13:46 |
| soren | twb: Ah, that's the installer initrd. It might be different that what's in an installed system. | 13:46 |
| soren | Well, it /is/ different, of course, but in a way that matters for this use case. | 13:46 |
| twb | I don't really care about an installed system because 1) it'll be pre-seeded to include openssh-server; and 2) I can do arbitrarily filthy things, "install vga16fb true" in /etc/modprobe.d/kludges.conf | 13:47 |
| twb | But in the initrd, it means I *must* preseed at least enough to get the sshd udeb installed, or use -nographic (serial) or -vnc :0. | 13:48 |
| twb | soren: I *can* confirm that blacklist=vga16fb DTRT in a lucid live CD generated by live-helper. | 13:56 |
| CIA-4 | grub-installer: cjwatson * r855 ubuntu/ (debian/changelog grub-installer): Install grub-efi on i386/efi and amd64/efi subarchitectures. | 15:20 |
| CIA-4 | partman-basicfilesystems: cjwatson * r587 ubuntu/ (debian/changelog fstab.d/basic): Automatically mount the first method=efi filesystem we see on /boot/efi. | 15:37 |
| CIA-4 | partman-basicfilesystems: cjwatson * r588 ubuntu/debian/changelog: releasing version 63ubuntu6 | 15:39 |
| CIA-4 | partman-auto: cjwatson * r323 ubuntu/ (9 files in 3 dirs): | 15:41 |
| CIA-4 | partman-auto: Add i386/efi (and amd64/efi, kfreebsd-i386/efi, and kfreebsd-amd64/efi) | 15:41 |
| CIA-4 | partman-auto: recipes that create an EFI System Partition. | 15:41 |
| CIA-4 | partman-auto: cjwatson * r324 ubuntu/debian/changelog: releasing version 91ubuntu2 | 15:42 |
| cjwatson | ev: do you know how the foundations-m-gsoc-usbcreator-* work items for alpha-2 are going? | 15:46 |
| ev | cjwatson: I've moved them to A3. I don't think they'll be ready for A2. | 15:59 |
| cjwatson | ok, thanks | 16:04 |
| ev | michaelforrest: the plan is to get rid of the hostname box entirely, correct? | 17:21 |
| twb | FYI, the Australia/Melbourne vs. Australia/Victoria timezone detector thing is working now. | 18:01 |
| twb | Looks like I can't remove the sda5_crypt that guided partition created via partman -- but I *can* just spawn a shell and run "cryptsetup remove sda5_crypt". | 18:21 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!