| View previous topic :: View next topic |
| Author |
Message |
mpcww
Tux's lil' helper
Joined: 08 Apr 2003
Posts: 115
|
 Posted: Thu Aug 01, 2013 3:02 pm Post subject: |
 |
|
| keenblade wrote: |
I have everything /boot /usr / /var and others as being separate partitions. Using grub2 and systemd working great. Also using kdm as login manager. I don't use genkernel. Just compile kernel myself. Also I use dracut to create the initramfs.
What is your latest situation? |
Well, first I checked my grub-installation and was surprised by finding no installation at all.
How did the machine boot ?
/boot/grub and the stage files existed on sdc5 !
I reinstalled grub. Now I get this information
| Code: |
arthur ~ # eix -I grub
Installed versions: [b] 0.97-r12[/b](16:11:37 01.08.2013)(ncurses -custom-cflags -netboot -static)
|
| Quote: |
grub> root (hd2,4)
Filesystem type is ext2fs, partition type 0x83
grub> find /boot/grub/stage1
(hd2,4)
grub> setup (hd2)
Checking if "/boot/grub/stage1" exists... yes
Checking if "/boot/grub/stage2" exists... yes
Checking if "/boot/grub/e2fs_stage1_5" exists... yes
Running "embed /boot/grub/e2fs_stage1_5 (hd2)"... 20 sectors are embedded.
succeeded
Running "install /boot/grub/stage1 (hd2) (hd2)1+20 p (hd2,4)/boot/grub/stage2 /boot/grub/menu.lst"... succeeded
Done.
|
I just wonder why I have to use (hd0,...) instead of (hd2,..) in my menu.lst to work ?
Changes in menu.lst are displayed in the boot menu, so I am pretty sure to run this installation
(I added a new entry for gentoo with systemd 2 days ago).
I use genkernel and that's it. |
|
| Back to top |
|
 |
medspec
n00b
Joined: 31 Jul 2013
Posts: 14
|
| Posted: Thu Aug 01, 2013 5:11 pm Post subject: |
|
|
fixed it..
black screen was the nvidia driver being loaded and i forgot to enable the gdm.service.. still bit new to that service stuff
So..
- baked my own kernel
- lost the initramfs
- emerged the nvidia-driver again
- remember to enable gdm: "systemctl enable gdm.service"
- Boots up FReakishly fast now.. systemd is fast.. losing the initramfs also helps off course 
gdm back, wifi back, power state back.. i am loving it !! 
looks like i will be baking my own kernels from now on... |
|
| Back to top |
|
|
mpcww
Tux's lil' helper
Joined: 08 Apr 2003
Posts: 115
|
| Posted: Thu Aug 01, 2013 6:34 pm Post subject: |
|
|
@medspec
Great !
So, the genkernel and intramfs seems+ to have a problem with systemd .
Meanwhile I tried out grub2 as alternative following this description http://wiki.gentoo.org/wiki/GRUB2_Quick_Start
Parallel installation worked and I used grub2-mkconfig to fill the grub.cfg .
I didn't switch to systemd and wanted to test grub2 with the conditions grub 97 is already working with.
After a reboot grub2 gave me a kind of well know error:
"Cannot open root device sdc6 or unknown block (0,0): error -6
And after here are the available options came a kernel panic.
With this said I don't think the boot loader is the reason.
I will try to compile a kernel without genkernel as well. |
|
| Back to top |
|
|
medspec
n00b
Joined: 31 Jul 2013
Posts: 14
|
| Posted: Thu Aug 01, 2013 7:16 pm Post subject: |
|
|
yeah i had that error..
went away after i build the kernel with disk controller IN kernel
good luck |
|
| Back to top |
|
|
medspec
n00b
Joined: 31 Jul 2013
Posts: 14
|
|
| Back to top |
|
|
mpcww
Tux's lil' helper
Joined: 08 Apr 2003
Posts: 115
|
| Posted: Fri Aug 02, 2013 4:21 pm Post subject: |
|
|
Yes, that would have been the second stop in going the grub2-way.
- Checking that grub2 runs with the regular genkernel-made version
- Switching to systemd by adding this configuration change to grub2
But it didn't work already with step 1 ...
Today I tried out to identify the modules reponsibles for the disk controller and I'm going to compile it into the kernel ...[/list] |
|
| Back to top |
|
|
stareagle
n00b
Joined: 04 Jul 2006
Posts: 29
|
| Posted: Fri Aug 02, 2013 8:38 pm Post subject: |
|
|
Hello everyone,
I also have the problem with the block device. If I add | Code: | | real_init=/usr/lib/systemd/systemd |
to my kernel boot line, I get the this message:
| Quote: |
The filesystem mounted at /dev/sda3 does not appear to be a valid /, try again.
Could not find the root block device in .
|
This only happens if I use systemd as init system. With OpenRC (default) the system boots normally.
My partition layout looks like this:
| Code: |
/dev/sda1 boot
/dev/sda2 swap
/dev/sda3 /
/dev/vg1/home /home
/dev/vg1/usr /usr
/dev/vg1/var /var
/dev/vg1/opt /opt
/dev/vg1/tmp /tmp
/dev/vg1/srv /srv
|
All drivers from the kernel which are necessary for the system are integrated into the kernel. Anyone with ideas about this issue. Otherwise I would setup a virtual machine tomorrow a test the hole thing on this machine.
Best regards
Stareagle |
|
| Back to top |
|
|
mpcww
Tux's lil' helper
Joined: 08 Apr 2003
Posts: 115
|
| Posted: Fri Aug 02, 2013 9:20 pm Post subject: |
|
|
| stareagle wrote: |
This only happens if I use systemd as init system. With OpenRC (default) the system boots normally.
All drivers from the kernel which are necessary for the system are integrated into the kernel. Anyone with ideas about this issue. Otherwise I would setup a virtual machine tomorrow a test the hole thing on this machine.
|
Do you mean all drivers are not compiled as modules ?
I ask because, medpec reporte to be successful by switching from genkernel to a "make;make modules" -Kernel.
I started to compile modules I considered for disk-controller relelvant.
So the failure point and message slightly differred.
The bold ones from the lspci-command are build in:
| Code: |
Kernel driver in use: agpgart-intel
Kernel driver in use: i915
Kernel driver in use: uhci_hcd
Kernel modules: uhci_hcd
Kernel driver in use: uhci_hcd
Kernel modules: uhci_hcd
Kernel driver in use: uhci_hcd
Kernel modules: uhci_hcd
Kernel driver in use: ehci-pci
Kernel modules: ehci_pci
Kernel driver in use: snd_hda_intel
Kernel modules: snd_hda_intel
Kernel driver in use: pcieport
Kernel driver in use: pcieport
Kernel driver in use: uhci_hcd
Kernel modules: uhci_hcd
Kernel driver in use: uhci_hcd
Kernel modules: uhci_hcd
Kernel driver in use: uhci_hcd
Kernel modules: uhci_hcd
Kernel driver in use: ehci-pci
Kernel modules: ehci_pci
[b] Kernel driver in use: ata_piix
Kernel modules: ata_piix, pata_acpi
[/b] Kernel driver in use: i801_smbus
Kernel modules: i2c_i801
[b] Kernel driver in use: ata_piix
Kernel modules: ata_piix, pata_acpi
[/b] Kernel driver in use: r8169
Kernel modules: r8169
[b] Kernel driver in use: pata_it8213
Kernel modules: pata_it8213, pata_acpi
[/b] |
But I still need to identify more modules to built into the kernel. |
|
| Back to top |
|
|
mpcww
Tux's lil' helper
Joined: 08 Apr 2003
Posts: 115
|
| Posted: Sat Aug 03, 2013 11:27 am Post subject: |
|
|
Finally I compiled a kernel who skipped this "Cannot open root device".
But I need to change
| Code: |
root (hd0,4)
kernel /bzImagel root=/dev/sdc6 udev video=1920x1200-32
|
to
| Code: | root (hd0,4)
kernel /bzImagel root=/dev/sdb6 udev video=1920x1200-32
|
The way disk and partitions are discovered and named has changed when leaving initramfs and modules
create by genkernel.
The 'bzImage' offers
- sda with partition 5-9
- sdb with partition 5-9
- sdc with paritition 5
- sdd with partition 5-6
while the 'genkernel' finds this order
- sda wit partition 5-6 (ata-WDC_WD5000AAKB)
- sdb with partition 5-9 (ata-SAMSUNG_HD103SJ)
- sdc with partition 5-9 (ata-OCZ-VERTEX3_OCZ)
- sdd with partition 5 (ata-WDC_WD10EALX)
Now I don't want to rename everything in the fstab.
Is there a way to determine the oder and naming of the disks in order to provide the same naming for both "kernel-compile-types" ?
Found these finding-rootfs-during-boot-recipes:
http://www.ibm.com/developerworks/linux/library/l-boot-rootfs/#4tips |
|
| Back to top |
|
|
mpcww
Tux's lil' helper
Joined: 08 Apr 2003
Posts: 115
|
| Posted: Sat Aug 03, 2013 12:43 pm Post subject: |
|
|
Labeling the partitions and using those labels in the fstab and the 'root=' kernel parameter of grub's menu.lst didn't help unfortunately, at least it didn't get worse.
So:
- kernel compiled with genkernel uses Labels and boots
- switching to systemd with real_init... does't find root fs even with real_root=LABEL=MYROOTLABEL
- the non-genkernel version doesn't boot either (think I 'm going to try the standard .config for it) and then forget that way as a solution
conclusion:
- I' going to keep gnome 3.8, but not using gdm and systemd . |
|
| Back to top |
|
|
stareagle
n00b
Joined: 04 Jul 2006
Posts: 29
|
| Posted: Sun Aug 04, 2013 7:22 am Post subject: |
|
|
Hello everyone,
in the last days I've done some research about my problems (on a VM with a new installation containing only a very basic system and systemd). The problem with the root partition looks like a problem with genkernel. The problem is this block in the linuxrc file (the init system used by the initramfs created by genkernel, which is a shell script):
| Code: |
if [ -d ${NEW_ROOT}/dev -a -x "${NEW_ROOT}${REAL_INIT:-/sbin/init}" ] || [ "${REAL_ROOT}" = "/dev/nfs" ]
then
break
else
bad_msg "The filesystem mounted at ${REAL_ROOT} does not appear to be a valid /, try again"
got_good_root=0
REAL_ROOT=''
fi
|
It is located around line 645. You can find this file in the /usr/share/genkernel/defaults directory from which it is included into the initramfs. The problem is the condition in the if. The variable ${NEW_ROOT} points to the temporary mount point of the real root, which is /newroot normally. The last part can be ignored here, because it only applies when booting using NFS.
The first part of the condition checks if the ${NEW_ROOT}/dev exists. The -a parameter is simply an "and" (the same as "&&" in other languages). The second part of the condition checks if the file ${NEW_ROOT}${REAL_INIT:-/sbin/init} is exists and is executable. The problematic part is ${REAL_INIT:-/sbin/init}. This meaning on this expression is: Use the value of $REAL_INIT if the variable is set, otherwise use /sbin/init. When using systemd as init (using the init or real_init kernel parameter), the value of REAL_INIT is /usr/lib/systemd/systemd. That sounds fine.
But: If /usr is a separate partition, for example in an lvm, it is not mounted when this if block is executed. The mounting happens many lines below in the script. So, the test fails... I think this is a bug in genkernel, and I will create a ticket in the Gentoo Bugzilla for this.
What should work is to put the root partition itself into an lvm, as far as I understand the code the linuxrc script. But I will check this in next days.
@keenblade: I also tried dracut (version 30 and 31), but had no success. The system boots with the initramfs created by dracut, but /usr is not mounted. Do you may have any hints for that problem?
Best regards
Stareagle |
|
| Back to top |
|
|
mpcww
Tux's lil' helper
Joined: 08 Apr 2003
Posts: 115
|
| Posted: Sun Aug 04, 2013 3:18 pm Post subject: |
|
|
| stareagle wrote: |
But: If /usr is a separate partition, for example in an lvm, it is not mounted when this if block is executed. The mounting happens many lines below in the script. So, the test fails... I think this is a bug in genkernel, and I will create a ticket in the Gentoo Bugzilla for this.
What should work is to put the root partition itself into an lvm, as far as I understand the code the linuxrc script. But I will check this in next days.
|
To confirm your analysis I just copied the "/", "/boot" and "/usr" partition onto just one.
After changing real_root to that new partition systemd booted the genkernel/initramfs .
Please, raise a bug and publish the number here, so we can vote for it.
Another issue with the new udev I observe during the last days is:
booting with openrc gives me a stop at "waiting for uevents to proceed" with a following "udevadm timed out ..."
Now booting with systemd I find a lot of
"usb n-n: device descriptor read/64, error -110" messages in the log file. |
|
| Back to top |
|
|
stareagle
n00b
Joined: 04 Jul 2006
Posts: 29
|
|
| Back to top |
|
|
james.h.bates
n00b
Joined: 11 Jul 2013
Posts: 12
|
| Posted: Mon Aug 05, 2013 1:30 am Post subject: |
|
|
| vasettoo wrote: |
...
I've already got these two in package.accept_keywords and the block persists. I finally managed to lift it by putting the packages depending on after 'emerge -cav libsoup' and the update went flawlessly. Still can't figure out how to add online accounts - after each entry the window dies and nothing is added. Issue with nm-applet which is inactive and showing no connections but networkmanager is doing its job properly - I have internet connection on eth and wlan after login. |
To get online accounts working, see https://bugs.gentoo.org/show_bug.cgi?id=475954. You must upgrade json-glib to at least dev-libs/json-glib-0.16.0-r1, which means adding that to package.keywords (or accept_keywords). |
|
| Back to top |
|
|
vasettoo
Tux's lil' helper
Joined: 11 Nov 2012
Posts: 97
|
| Posted: Mon Aug 05, 2013 11:44 am Post subject: |
|
|
| Thanks to james.h.bates on the hint with json-glib keyword I was able to complete the migration to systemd with Gnome 3.8 and have fully usable and working desktop environment. My issue with nm-applet has been solved with putting back systemd use flag missing from my make.conf. |
|
| Back to top |
|
|
keenblade
Veteran
Joined: 03 Oct 2004
Posts: 1087
|
| Posted: Mon Aug 05, 2013 2:26 pm Post subject: |
|
|
| stareagle wrote: |
@keenblade: I also tried dracut (version 30 and 31), but had no success. The system boots with the initramfs created by dracut, but /usr is not mounted. Do you may have any hints for that problem?
|
Have you added the modules "usrmount and systemd" to the dracut config?
_________________
Anyway it's all the same at the end...
Need help to get it working: "x-fi surround 5.1" |
|
| Back to top |
|
|
|
Desktop Environments
