A boot of Gentoo 2004.2 PPC live CD in fact scrambles the root password while yet leaving Ctrl-D defined to mean EOF i.e. exit to a login prompt for the unknowable password. Therefore I eventually had to cycle power with my HDD mounted. Now I can't write my HDD in Gentoo, though this one HDD is all the HDD's of this Mac. Only there could I record what I see in Gentoo. For me this is a Blocker bug. Booting a Mac OS X Installer CD to run Disk Utility assures me my HDD is actually ok, but Gentoo will not agree. Gentoo tells me to run fsck.hfsplus, but Gentoo doesn't include fsck.hfsplus. Gentoo does include hfsck, but that doesn't help either. Reproducible: Always Steps to Reproduce: 1. Boot Gentoo 2004.2 PPC live CD. 2. mkdir /mnt/hda ; mount -w /dev/hda /mnt/hda 3. Inadvertently Ctrl-D to logout. 4. Power off. Actual Results: Transcribed by hand, what I see includes: # hfsck Usage: hfsck [-v] [-n] [-a] device-path [partition-no] # hfsck /dev/hda *** Checking volume MDB *** Checking volume structure *** Checking extents overflow B*-tree *** Checking catalog B*-tree # # mount /dev/hda /mnt/hda # cd /mnt/hda # touch t touch: cannot touch `t': Read-only file system # # mount | grep hda /dev/hda on /mnt/hda type hfsplus (rw) # tail -1 /var/log/everything/current [kernel] HFS+-fs warning: Filesystem was not cleanly unmounted, running fsck.hfsplus is recommended. mounting read-only. # tune2fs -l /dev/hda tune2fs 1.34 (25-Jul-2003) tune2fs: Bad magic number in super-block while trying to open /dev/hda Couldn't find valid filesystem superblock. # man hfsck -bash: man: command not found # # man hfsck also not found in Google # # uname -msr Linux 2.6.7-gentoo-r11-G4-SMP ppc # Expected Results: 1) See that Apple Disk Utility has verified my boot HDD, and let me write it. 2) Give me a way to say get out of my way, I do myself believe I have verified my boot HDD, a way less expensive than having to erase and rewrite all of my only HDD. 3) Define a key like magic sysrq on x86 Linux keyboards that means sync and shutdown, and recommend that every newbie begin by learning to use it. (Keystrokes can take learning e.g. Ctrl+Alt+fn+F2 virtual terminal works here in that sequence but not in every possible sequence.)
1) Sorry I'm new to bugs.gentoo.org, looks like I've caused multiple e-mail notifications as I discovered the syntax to get the this-bug-blocks list correct here. 2) This bug, by preventing me from quoting procedure and results, also blocks me from reporting here helpfully the reproducible bugs I sketch at http://linux-pel.blog-city.com/read/782855.htm 3) Bugs found and not reported lately have included: --- Incidentally, dmesg captures nothing useful, being instead full of lines such as: evbug.c: Event. Dev: adb2:2.c3/input, Type: 1, Code: 28, Value: 1 evbug.c: Event. Dev: adb2:2.c3/input, Type: 0, Code: 0, Value: 1 --- Incidentally, sound works nowhere I've tried. Transcribed by hand from a Titanium Powerbook, the KDE "Informational - artsmessage" is: Sound server informational message: Error while initializing the sound driver: device /dev/dsp can't be opened (No such file or directory) The sound server will continue, using the null output device. --- Seems like my iLamp (800 MHz PowerPC G4, 256 MB SDRAM) doesn't boot by default, instead I must explicitly specify the purported default, boot: G4 Also once I saw I lost keyboard control, being left at a blood red full screen, if I try X without first reconfiguring and launching xfs. With correct setup, the red screen flashes by but does not remain. ---
Knoppix live CD, by contrast, logs root into every vt (virtual console) before turning off the root password.
> you mean the KDE/Gnome LiveCD .. > not the Minimal or Universal LiveCD ... ? Yes I mean the kde-gnome-universal-2004.2.iso of http://www.metadistribution.org/gentoo/2004.2/
I'm going to assign this to PPC to see if they are planning on supporting it.
Mac OS X Disk Utility lets me insist on "Repair" even though "Verify" reports nothing wrong. But even a forced "Repair" there does not persuade Gentoo to mount -w -t hfsplus.
I think I remember: I saw mount claim r/w, though dmesg and touch told me r/o ... serendipitously I now see that same symptom in x86 if I "SysRq : Emergency Remount R/O". Maybe learning to overcome that there would help.
Kindly offline I'm reminded that a FireWire Mac can be made a target of another Mac and then fsck'ed by that other Mac, even from Mac OS X. Trying fsck at the Terminal command line might make a difference. I can't try that option myself because my laptop's FireWire port died, and fsck'ing the mounted boot disc makes no sense, of course: $ uname -msr Darwin 7.5.0 Power Macintosh $ $ fsck_hfs /Volumes/Macintosh\ HD/ /Volumes/Macintosh HD/ is not a character device CONTINUE? [yn] n $
Kindly offline I received the suggestion: Boot the Mac with Command-S held down to try single user mode. That suggestion let me run the Mac OS X fsck -f -y. Same result as before, sorry. Again the Apple fsck assures me my boot disk is fine, but the Gentoo mount of /dev/hda doesn't let touch work, and /var/log/everything/current explains: [kernel] HFS+-fs warning: Filesystem was not cleanly unmounted, running fsck.hfsplus is recommended. mounting read-only.
Livecd team: has the CTRL-D issue been fixed in catalyst? On to the hd issue: We have included hfsplus stuff in the emerge system profile. It will be available in stage3 and also on the 2004.3 livecd. 2004.3 will also allow one to repair hfsplus volumes, besides mounting, resizing and creating them. Event debugging is disabled in 2004.3 Regarding the sound: could you do an 'lsmod' and send me the output? I'm not sure if I compiled sound support as module, but it should have been automatically loaded. Note that the G5 currently doesn't support sound. Default boot option: I'll report the "yaboot does not boot G4 or G4-SMP (depending on the livecd version) by default after some time om the iLamp 800mhz" upstream. It works correctly on all machines here. It might be an openfirmware issue that can be solved by resetting the firmware. Gentoo cannot rw mount my hfsplus hd: unmounting happens on reboot. An OS X hd being not cleanly unmounted is most likely caused by an OS X app (or daemon like LoginWindow) not shutting down properly in Mac OS X, preventing the HD from being cleanly unmounted. Do you run anything in the background on windows. If you have firewire disks/lot of external hardware attached, can you try booting into OS X without the external devices attached and check whether a reboot solves the problem in Gentoo?
livecd@ is not responsible for catalyst and do not have write access to the sources, just the actual release media that we create with catalyst. If you need something fixed in catalyst, you should file a bug and assign it to zhen@gentoo.org instead.
> > sound works nowhere I've tried. > > > > Transcribed by hand from a Titanium Powerbook, > > the KDE "Informational - artsmessage" is: > > > > Sound server informational message: > > Error while initializing the sound driver: > > device /dev/dsp can't be opened (No such file or directory) > > The sound server will continue, using the null output device. > > > > Regarding the sound: > > could you do an 'lsmod' and send me the output? Empty, sorry. Transcribed by hand: # lsmod Module Size Used by # > not sure if I compiled sound support as module, but > it should have been automatically loaded. > Note that the G5 currently doesn't support sound. Transcribed by hand: # grep -i sound /var/log/everything # find /lib/modules/`uname -r` -iname '*.*o' | grep sound | wc -l 18 # uname -r 2.6.7-gentoo-r11-G4-SMP # P.S. In case I can help pursue this further, I will mention, these logs reminded me, but only slowly, the procedure to reproduce was to launch KDE, which in 2004.2 meant: boot with Gentoo 2004.2 live CD inserted and C key held down Xautoconfig cd /etc/conf.d vi xfs +/XFS_PORT ./xfs start vi ~/.xinitrc startx Those two vi set XFS_PORT="7100" and provided the three line ~/.xinitrc: #!/bin/bash -x export LD_LIBRARY_PATH=/opt/Ximian-OpenOffice/program /usr/kde/3.2/bin/startkde
> If you need something fixed in catalyst, > you should file a bug and assign it to zhen@gentoo.org instead. Just now I created bug 65097 "Livecd: inadvertent ctrl D powers off with HDD mounted" and then I set its cc to pvdabeel@gentoo.org, wolf31o2@gentoo.org, zhen@gentoo.org I hope that helps.
> > Seems like my iLamp (800 MHz PowerPC G4, 256 MB SDRAM) > > doesn't boot by default, > > instead I must explicitly specify the purported default, boot: G4 Sorry I wasn't clear. I believe I meant to say if I answer the boot prompt with a strike of the return key, then Gentoo does not boot. If instead I specify G4 explicitly, then it does boot. That's different from what you understood me to say, yes? I quote: > Default boot option: I'll report the "yaboot does not boot > G4 or G4-SMP (depending on the livecd version) by > default after some time om the iLamp 800mhz" upstream. > It works correctly on all machines here. It might be an > openfirmware issue that can be solved by resetting the > firmware. Can I help further? I could file a separate bug, I could reproduce this result to confirm the details I think I remember, I could borrow another iLamp to try ... I'm comfortable enough with Open Firmware to have reverse engineered how to pass SCSI thru it, but I don't know which "reset" you mean to suggest I try.
> Gentoo cannot rw mount my hfsplus hd: unmounting happens on reboot. Help, this English I don't confidently understand, sorry. Do you mean to now be speaking of an HDD of your own, rather than mine? Are you reporting that Gentoo refuses to mount your HDD rw? > An OS X hd being not cleanly unmounted is most likely caused by an OS X app > (or daemon like LoginWindow) not shutting down properly in Mac OS X, > preventing the HD from being cleanly unmounted. Yes. But my unclean unmount was caused by me powering off after failing to guess the scrambled root password for which I was prompted after I inadvertently struck Ctrl-D while in the root boot shell before launching X. I notice that booting from a Mac OS X install and running Mac OS X Disk Utility, and booting the HDD single user and running Mac OS X fsck, both report my HDD clean. Only Gentoo claims my HDD remains unclean. > Do you run anything in the background on windows. "Help, this English I don't confidently understand, sorry." > If you have firewire disks/lot of external hardware attached, Sometimes yes, thanks for asking. Sorry I knew this already: to simplify these experiments, my Gentoo boots before now have all been with no drives attached. Today I thought I'd try a USB Flash, to see if I can log results there rather than to my ro HDD. > can you try booting into OS X without the external devices attached and Have done, above, sorry I did not include that detail in my report. > check whether a reboot solves the problem in Gentoo? Does not, sorry.
Let's split this up in a few bugs. We already have a bug open about the Font server on the kde/gnome livecd. Comment 13: What version did this happen with. If it happens with the 2004.2 livecd, then it's a cosmetic bug that will be fixed in 2004.3 :-) If I understand it correctly, this bug is about several things: 1. Ctrl-D on the livecd exits the bash shell, causing you to be promted for a password you don't know. This is a multi-arch bug, I've cc'ed zhen, who will take care of this. 2. As a consequence of being locked out you couldn't reboot the machine cleanly. Should be translated to: "PPC hasn't got a ctrl-alt-delete shortcut to reboot a machine". I'll open a new bug about that. 3. As a consequence to rebooting the machine uncleanly, gentoo refuses to rw mount your hfs filesystem. This should probably be fixed on 2004.3 if we include a newer version of the complete hfsplus toolkit.
*** Bug 61185 has been marked as a duplicate of this bug. ***
> Let's split this up in a few bugs. Yes please. Sorry I mentioned so many bugs here. I was then only trying to show how Gentoo refusing to write my HDD in August meant I could not then easily share here those other interesting bugs I was seeing. All that's over now. The paid work that kept me at the Titanium Powerbook ended 23rd of September. Now between jobs I will play mostly with the iLamp and iMac. Sorry To: p.lavarre@ieee.org was a black hole for much of the 23rd & 24th. > We already have a bug open about the Font > server on the kde/gnome livecd. Yes, bug 61185 "Xautoconfig configures for Fontserver, rather than fontpath". > Comment 13: What version did this happen > with. 2004.2 livecd > If it happens with the 2004.2 livecd, then > it's a cosmetic bug that will be fixed in > 2004.3 :-) That ":-)" grin I cannot confidently interpret in this context, sorry. With the 2004.2 live CD, both my iLamp and my iMac say G4 if I press Return, both get as far as a screen that begins: Welcome to Linux, kernel 2.6.7-gentoo-r11-G4-SMP The iMac proceeds to boot. But the iLamp soon dies, after printing 30 lines which end with: arch: exit openpic: enter openpic: timer openpic: ipi openpic: external openpic: spurious openpic: exit smp_core99_probe core99_setup_cpu 0 done The iLamp works if I explicitly type G 4 rather than pressing Return. Should I believe 2004.3 will fix that? Tell me that we want a separate bug created for this issue, and I will. > Comment 13: What version did this happen > with ... > If it happens with the 2004.2 livecd, ... Eh? By now is there already a newer .iso available for me to try? > If I understand it correctly, > this bug is about several things: By now I see maybe half a dozen bugs in this one bug, sorry. All were blocked by unwritable HDD, but by now I have erased that HDD and started over. > 1. Ctrl-D on the livecd exits the bash shell, > causing you to be promted for a password you > don't know. > > This is a multi-arch bug, I've cc'ed zhen, who > will take care of this. Thank you. Yes prompted for an unknowable password, after one inadvertent keystroke. This multi-arch claim I trust but I cannot myself confirm by observation, since I have only Mac PPC at home. Search tells me no such new bug yet, other than my own attempt to separate this issue: bug 65097 "Livecd: inadvertent ctrl D powers off with HDD mounted" > 2. As a consequence of being locked out you > couldn't reboot the machine cleanly. > > Should be translated to: "PPC hasn't got a > ctrl-alt-delete shortcut to reboot a > machine". I'll open a new bug about that. Thank you. Yes no shortcut to reboot. I trust already you understand, this "translation" adds a conclusion that was not mine i.e. the idea of wanting a ctrl-alt-delete for PPC. I'm glad to have a solution of any kind, but myself I prefer the solution of the Ctrl-Alt-F$n shortcuts of Knoppix, which let me choose reboot or less drastic means of recovery, by switching from one vt (virtual console) to another, into all of which root has automagically logged in. > 3. As a consequence to rebooting the machine > uncleanly, gentoo refuses to rw mount your > hfs filesystem. > > This should probably be fixed on 2004.3 if we > include a newer version of the complete > hfsplus toolkit. Yes. Tell me we want a separate bug created for what remains of this, and I will. I think what remains is that gentoo HFS+ flags volume not unmounted cleanly in a way that Mac OS X repair of HFS+ does not clear, ouch. > ... In the comments above, I see we've also fixed the bug: 4. mesg captures nothing useful, being instead full of evbug > ... Above already you saw I'm unsure if we want me to separately report the bug: 5. boot: Return falsely says it means boot: G4 on iLamp > ... I'll reply again here with concise cross-references after I detail and report separately: 6. sound works nowhere I've tried 7. X without reconfiguring locks up
> *** Bug 61185 has been marked as a duplicate of this bug. *** The title of bug 61185 was "Xautoconfig configures for Fontserver, rather than fontpath". Comment six of that bug kindly points to the fix: bug 53258 "Patch for Xautoconfig - adds support for X.org"
In summary, until I return ... 1) Yes I have by now reproduced the Titanium sound issue on my iLamp with further detail, so I can file that separately. 2) Yes please "include hfs+". Just now I again lost write privilege for a second boot drive to "[kernel] HFS+-fs warning: Filesystem was not cleanly unmounted", this time by mistyping the path I gave to my script for working around bug 53258 "Patch for Xautoconfig - adds support for X.org": --- #!/bin/bash -x f=$1 cd /etc/conf.d mv -i xfs xfs~ cp $f/etc.conf.d.xfs xfs cp $f/home.xinitrc ~/.xinitrc Xautoconfig cd /etc/init.d ./xfs status ./xfs start startx & --- With the path mistyped, that script boiled down to: Xautoconfig cd /etc/init.d ./xfs status ./xfs start startx & which crashed the console display of the 2004.2 livecd console of my iLamp, same as I saw before.
HFS+ is included in the betas All shortcuts are assigned The xorg.conf patch will be in cvs. I'll write a message when the beta cds are online.
http://www.metadistribution.org/gentoo/2004.3/ offers .tar.bz2 dated 2004-09-28 for "development" "beta" "testing" but no live CD .iso, as yet.
I am using hfs+ and use ctrl-alt-del many time to reboot system on 2004.3. I'll try to get my hand on 2005.0 today, and I am quite sure it will work. ctrl-d can be tested too.
I'm about to rewrite Xautoconfig to work with X.Org. See bug #84468 Ported hfsplus-tools from OS X are currently in testing phase. See bug #46342. It's not included in 2005.0 yet. The ctrl-d bug has been removed. You can't logout on a LiveCD. I can't test for ctrl-alt-del on my iBook, as I don't have a del-key ;-) This needs a rewrite of the i386 keymaps, to bind del on shift-backspace or something like this. But I see it as resolved, as you can't logout from any of the four consoles on the LiveCD and you should be able to type in reboot. And why are all the splitted bugs of this big bug marked as a duplicate? That makes it even harder to track all the different issues...
Everything seem to work now. Closing this bug.
