[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: PPC64 Image Installation Error



Perhaps I spoke too soon...

After reinstalling the system due to an unresolvable networking error (A perfectly stable connection would drop without reason, something that did not happen in Ubuntu 16.04, at least after network-manager was installed. The only fix was to deactivate and reactivate the wired connection in nmtui, which itself as a part of network-manager had to be manually downloaded from another OS in individual package format as it did not come out of the box.), it seems I am not able to fully boot into it via the yaboot from Ubuntu 16.04, which works well and recognizes both systems without any problems, boots into either without issue, but then Debian throws a BusyBox initramfs terminal halfway through booting, citing: "Gave up waiting for root device.", "Missing modules (cat /proc/modules; ls /dev)", "ALERT! /dev/disk/by-uuid/<long string of numbers> does not exist. Dropping to a shell!". Absolutely everything is partitioned and configured exactly the same as before, when I had it working. The only thing different, however, is perhaps having chosen the option to install LXQt in favor of GNOME, which was the environment installed when it was able to boot without issue. The same blockage occurs when LXDE is chosen, as well. I wish it would allow you to install your system without needing to choose a desktop environment, which just lengthens installation time by a great amount.

From my experiences alone in the last few days, the ppc64 versions really put the "unstable" in "Sid".

In any case, please advise. Thank you.


From: Noah Wolfe <n.wolfe4@outlook.com>
Sent: Friday, February 1, 2019 2:08 PM
To: Rick Thomas
Cc: PowerPC List Debian; John Paul Adrian Glaubitz
Subject: Re: PPC64 Image Installation Error
 

Thanks, Rick.

I got around this issue by simply making the /boot partition (sda2) 1 GB instead of 1 MB, and the / partition (sda3) doubled to 80 GB. Using this same method with a reinstall of Ubuntu Server 16.04 (which was giving me the same error when it was installed in partitions 6, 7, 8, and 9, below Sid) I am now, as a result, able to dual boot between it and Debian.

So it really goes to show that sometimes the best solutions in life are usually the easiest ones.

But I still found the 01-27 build to be unable to install the kernel due to an apparent "lack of space" in whatever configuration I put it in (save for the aforementioned method), so that still needs to be fixed.

And, a lot of ppc64 packages are experiencing issues unseen in the powerpc equivalents. Xfwm4 puts out graphical glitches and artifacts when any windows are moved, or when the cursor moves over any windows, and GNOME 3 hasn't displayed any color besides pinkish hues since Jessie, if I'm remembering correctly.

Anyway, thank you for your time.

N

On Jan 31, 2019 5:16 AM, Rick Thomas <rbthomas@pobox.com> wrote:
>
> Hi Noah,
>
> Is it possible for you to temporarily replace the hard drive in your machine with one you don’t mind experimenting on?
>
> If so, after a clean installation onto the fresh hard disk, what you want to finish up your installation with is something that looks kind of like this:
>
> > root@msmini:~# mac-fdisk -l /dev/sda
> > /dev/sda
> >         #                    type name                  length   base      ( size )  system
> > /dev/sda1     Apple_partition_map Apple                     63 @ 1         ( 31.5k)  Partition map
> > /dev/sda2         Apple_Bootstrap untitled                1954 @ 64        (977.0k)  NewWorld bootblock
> > /dev/sda3         Apple_UNIX_SVR2 untitled              500001 @ 2018      (244.1M)  Linux native
> > /dev/sda4               Linux_LVM untitled           468360060 @ 502019    (223.3G)  Unknown
> > /dev/sda5              Apple_Free Extra                     49 @ 468862079 ( 24.5k)  Free space
> >
> > Block size=512, Number of Blocks=468862128
> > DeviceType=0x0, DeviceId=0x0
> >
> > root@msmini:~# lvs
> >   LV     VG        Attr       LSize   Pool Origin Data%  Meta%  Move Log Cpy%Sync Convert
> >   root   msmini-vg -wi-ao---- <45.61g                                                   
> >   swap_1 msmini-vg -wi-ao---- 976.00m                                                   
> >
> > root@msmini:~# ls -l /dev/mapper
> > total 0
> > crw------- 1 root root 10, 236 Jan 31 00:30 control
> > lrwxrwxrwx 1 root root       7 Jan 31 00:30 msmini--vg-root -> ../dm-0
> > lrwxrwxrwx 1 root root       7 Jan 31 00:30 msmini--vg-swap_1 -> ../dm-1
> >
> > root@msmini:~# df -HTP | grep -v tmpfs
> > Filesystem                  Type      Size  Used Avail Use% Mounted on
> > /dev/mapper/msmini--vg-root ext4       48G  1.3G   45G   3% /
> > /dev/sda3                   ext2      248M   51M  185M  22% /boot
> >
> > root@msmini:~# cat /proc/swaps
> > Filename                              Type            Size    Used    Priority
> > /dev/dm-1                               partition     999420  0       -2
>
>
>
> Here at installer partition time I chose to use the entire disk and set up the LVM (logical volume manager) rather than have explicit partitions for root and swap.  You could do it the other way  (whole disk, without LVM) if you want.
>
> Explanation:
>
> The “root” on dm-0 (/dev/mapper/msmini—vg-root) has a standard Linux ext4 filesystem.
>
> The “swap” on dm-1 (/dev/mapper/msmini-vg-swap_1) is formatted as a standard Linux swap partition.
>
> The /boot partitions on sda3 is formatted as Linux ext2 (not ext3 or ext4).  It will hold, among other things, the kernel and init-ram-disk image.  It needs to be an explicit partition because the yaboot boot-loader doesn’t understand Linux LVM.  It needs to be formatted ext2 because yaboot doesn’t understand the more advanced features of ext3 or ext4.
>
> The “Apple_Bootstrap — NewWorld bootblock” on sda2 is formatted as an Apple HFS partition.  It will hold the yaboot bootloader and a some configuration stuff that yaboot needs.  It will not be mounted when Linux is running after the boot process is finished.  This is necessary because Apple’s “open firmware” (equivalent of PC BIOS) likes to have it’s first level boot-loader in the first HFS partition on the disk. This thing is crafted out of whole cloth by the Linux “ybin” program based on /etc/yaboot.conf.  It should not be modified in any way except by ybin.
>
> Sizes:
>
> For the most part, the sizing of these partitions will be taken care of automatically by the installer partitioning step, however, for completeness:
>     The Apple_Bootstrap partition should be approximately 1 MB (historical note: this is designed to be slightly larger than an 880 KB
> double density Apple Boot floppy)
>     The /boot partition should be big enough to hold a few copies of the Linux kernel and their respective initrd images.  A couple hundred MB to a GB is usually enough, but there’s no harm in making it bigger if you want.
>     Your swap partition should (opinions differ, I’m giving you mine) be between 1.0 and 10.0 times the size of your system RAM.  The Mac mini I’m using for this example has 1024 MiB of RAM.
>     The size of your root partition is mostly up to you, though 7-10 GB is a good minimum if you choose to have a separate /home partition.  If you decide to keep /home merged with the root filesystem (as I did for this example) you need to take that minimum and add enough to cover your anticipated needs for user file space in /home.
>
> Hope this helps!
> Rick
>
> ==================================================================
>
> > On Jan 29, 2019, at 5:44 AM, Noah Wolfe <n.wolfe4@outlook.com> wrote:
> >
> > Hi Rick,
> >
> > I partitioned the first disk (1 TB)  in my machine like so:
> >
> > #1 32 KB Apple Partition Map
> > #2 8 MB HFS NewWorld Boot Partition (mounted at /boot) - Designated to
> > Ubuntu Server 16.04, which used up 313 KB, leaving 7.32 MB unused.
> > #3 40 GB Ext4 System Partition (mounted at /) - Designated to Ubuntu
> > Server 16.04
> > #4 4.8 GB Swap Partition (4.5 GB installed) - Designated to Ubuntu
> > Server 16.04
> > #5 455 GB Ext4 Home Partition (mounted at /home) - Designated to Ubuntu
> > Server 16.04
> >
> > #6 8 MB Ext2 Boot Partition (to be mounted at /boot) - Designated to
> > Debian 10, which the installer left with 298 KB used, leaving 7.34 MB
> > unused. (Which is why I didn't jump to it being a space issue.)
> > #7 40 GB Ext4 System Partition (to be mounted at /) - Designated to
> > Debian 10
> > #8 4.8 GB Swap Partition (ditto) - Designated to Debian 10
> > #9 455 GB Ext4 Home Partition (to be mounted at /home) - Designated to
> > Debian 10
> >
> > This was actually also the first time I tried to dual-boot Linux
> > systems on PPC, so I now feel like I'm very obviously at fault for
> > this.
> >
> > But we're always learning. I've been on this train for just under a
> > year and not only do I thank myself every day, I also learn something
> > new every day. Or at least, every other day, perhaps. Just as well.Thanks, all.
> >
> > N
> >
> > On Tue, 2019-01-29 at 03:41 -0800, Rick Thomas wrote:
> >>
> >>>
> >>> On Jan 28, 2019, at 11:19 PM, Noah Wolfe <n.wolfe4@outlook.com>
> >>> wrote:
> >>>
> >>> Hello.
> >>>
> >>> I'm running a Late 2005 Power Mac G5 (11,2), and although the
> >>> regular
> >>> powerpc installer images may be fixed from most problems at this
> >>> point
> >>> (going off of other people's posted experiences), the newest
> >>> netinstall
> >>> ppc64 images from 2019-01-27
> >>> (http://cdimage.debian.org/cdimage/ports/2
> >>> 019-01-27/) are still unable to install the linux-image-powerpc64
> >>> package during normal (non-expert) installation, and when the
> >>> installer
> >>> gives you the notification of this, it quits the rest of the
> >>> installation, leaving an incomplete base system.
> >>>
> >>> The notification window itself displays "Unable to install the
> >>> selected
> >>> kernel" in red text at the top, with the description "An error was
> >>> returned while trying to install the selected kernel into the
> >>> target
> >>> system." in regular black text. It specifies with "Kernel package:
> >>> 'linux-image-powerpc64'." below, and adds "Check /var/log/syslog or
> >>> see
> >>> virtual console 4 for the details." at the end. It is accompanied
> >>> by a
> >>> blue backdrop instead of the usual red.
> >>>
> >>> Upon further inspection into virtual console 4 (Ctrl+Alt+F4), the
> >>> last
> >>> remaining written installer activity is displayed, a particular
> >>> point
> >>> of interest after the usual "Unpacking linux-image-4.19.0-2-
> >>> powerpc64
> >>> (4.19.16-1) ...", being: "dpkg: error processing archive
> >>> /media/cdrom//pool-ppc64/main/l/linux/linux-image-4.19.0-2-
> >>> powerpc64_4.19.16-1_ppc64.deb (--unpack)", "Cannot copy extracted
> >>> data
> >>> for './boot/vmlinux-4.19.0-2-powerpc64' to '/boot/vmlinux-4.19.0-2-
> >>> powerpc64.dpkg-new': failed to write (No space left on device)",
> >>> "dpkg-
> >>> deb: error: paste subprocess was killed by signal (Broken pipe)",
> >>> "W:
> >>> Last kernel image has been removed, so removing the default
> >>> symlinks".
> >>>
> >>> It then goes on to normally select, prepare to unpack, and unpack
> >>> linux-image-powerpc64, only to stop with "Errors were encountered
> >>> while
> >>> processing: /media/cdrom//pool-ppc64/main/l/linux/linux-image-
> >>> 4.19.0-2-
> >>> powerpc64_4.19.16-1_ppc64.deb", "error: exiting on error base-
> >>> installer/kernel/failed-install", "WARNING **: Configuring
> >>> 'bootstrap-
> >>> base' failed with error code 30", "WARNING **: Menu item
> >>> 'bootstrap-
> >>> base' failed.", ending there.
> >>>
> >>> Debian 7.11, 8.11, and the May 2018 Sid netinstall cd, all install
> >>> onto
> >>> this machine (given the same partition tables) without a hitch,
> >>> along
> >>> with Ubuntu Server 16.04.5. Though admittedly they were all of
> >>> regular
> >>> powerpc decent and not ppc64 as seen here. I had also tried the
> >>> Debian
> >>> 9 installer port for ppc64
> >>> (http://cdimage.debian.org/cdimage/ports/9.0
> >>> /ppc64/iso-cd/) and got the same result displayed here, so it
> >>> appears
> >>> this may be a longstanding issue among the ppc64 builds, at least
> >>> from
> >>> my point of view.
> >>>
> >>> Putting that aside, maybe something was overlooked in the ppc64
> >>> images
> >>> specifically when building the new batches? To my knowledge, much
> >>> of it
> >>> should be synonymous with the regular powerpc builds, save for a
> >>> 64-bit
> >>> userspace over a 32-bit one, but of course, I'm not a developer.
> >>>
> >>> In any case, all is much appreciated.
> >>>
> >>> N
> >> Hi Noah,
> >>
> >> It sounds like you’re running out of free space in the partition that
> >> contains the /boot directory.
> >>
> >> Can you describe the steps you used to partition the disk?  In
> >> particular, if you are re-using the /boot partition so as not to
> >> interfere with your existing installation, how much free space is
> >> there in it.  E.g. on your existing (working) setup do
> >>     df -HTP /boot
> >>
> >> hope it helps!
> >> Rick
> >>
>


Reply to: