mauricev

Yes, it boots the machine! So whatever makes grub work automatically gets blocked somehow by not having /dev/sda1 mounted at install time.

No, it's not blank. I just changed the name of the kernel to "kernel", so I wouldn't have to worry about mistyping it manually in the grub command line.

It does? grub.conf was indeed a typo.

With /dev/sda1 mounted or not, I get the same output from boot-update Copyright 2009-2017 Funtoo Technologies [use option "-l" for license info, "-h" for help] * Generating config for grub... DEFAULT > Funtoo Linux genkernel - kernel * NOTE: Detected MBR boot. Configuring for Legacy MBR booting. * WARN: Intel system detected - please emerge sys-firmware/intel-microcode and sys-apps/iucode-tool and run boot-update again; boot-update will then patch your system with the latest Intel CPU and chipset microcode patches at boot-time, protecting you against important vulnerabilities and errata. * Completed successfully with warnings. When /dev/sda1 is not mounted, boot-update enters this in the system log Feb 04 00:32:16 [kernel] [ 231.913202] EXT4-fs (sda1): mounting ext2 file system using the ext4 subsystem Feb 04 00:32:16 [kernel] [ 231.928341] EXT4-fs (sda1): mounted filesystem without journal. Opts: (null) Grub is at 2.02_beta3-r2. boot-update is 1.8.2. At the failed grub prompt, configfile grub/grub.conf results in no output. It just gives me another blank prompt. It's not that it can't find the kernel; it can't find grub.conf

It does find the kernel if I enter the info manually at the grub prompt. But what does that tell us? If it's not finding it automatically, it's effectively the same as not finding it unless you consider it normal for users to have to manually enter the kernel information at every startup. Running boot-update with /dev/sda1 not mounted has no effect on grub failing. It still fails.

I confirmed the original issue is a bug. When "intel" is supplied to VIDEO_CARDS for X11, it inappropriately installs an unnecessary file, 20-intel.conf in xorg.conf.d. The presence of this file overrides the normal device detection in X11 and forces it to think the video card is intel. If there is not an actual Intel video card, X11 will throw up its hands and fail.

ls -alF /boot shows nothing when /dev/sda1 is not mounted. grub.cfg and the kernels reside on /dev/sda1, not in the empty /boot directory. When /dev/sda1 is not mounted and grub is installed, it spews bogus errors about lvm but no other errors. However, the machine loses the ability to boot because it now can’t find the kernel. bcowan seems to be claiming this behavior is impossible. But you’re claiming that it’s normal? That /dev/sda1 must be mounted when grub is installed or am I misunderstanding? Please disregard my mention of any custom script. I'm talking exclusively about installing grub.

I just tested it again and when /boot is not mounted, grub installation throws a fit and makes the computer non-bootable.

I have made some progress. consolekit in /etc/init.d has to be started. It's no longer crashing.

I tried to setup a Funtoo installation with cinnamon. I changed to the desktop flavor and added it as a mixin. It didn't complain, but emerge didn't find anything more to emerge. So I manually emerged cinnamon. When I started X, I got a basic X and not cinnamon. So I changed .xinitrc to the following exec ck-launch-session dbus-launch cinnamon-session When I ran startx again, it started to run and then promptly crashed. The logged output is Feb 02 12:33:52 [ck-launch-session] error connecting to ConsoleKit Feb 02 12:33:52 [cinnamon-session] WARNING: t+0.00378s: Could not connect to ConsoleKit: Failed to connect to socket /var/run/dbus/system_bus_socket: No such file or directory_ Feb 02 12:33:53 [kernel] [ 32.643783] csd-xrandr[24011]: segfault at 0 ip 00007fc98568a9d9 sp 00007fff44441e50 error 4 in libupower-glib.so.3.0.1[7fc985678000+26000] Feb 02 12:33:53 [kernel] [ 32.645367] csd-power[24010]: segfault at 0 ip 00007f812bfcd9d9 sp 00007ffdce11c8d0 error 4 in libupower-glib.so.3.0.1[7f812bfbb000+26000] Feb 02 12:33:53 [cinnamon-session] WARNING: t+0.47240s: Application 'cinnamon-settings-daemon-xrandr.desktop' killed by signal 11_ Feb 02 12:33:53 [cinnamon-session] WARNING: t+0.95457s: Application 'cinnamon-settings-daemon-power.desktop' killed by signal 11_ Feb 02 12:33:54 [kernel] [ 33.875096] csd-xrandr[24115]: segfault at 0 ip 00007fece8eb69d9 sp 00007ffcb40b5a50 error 4 in libupower-glib.so.3.0.1[7fece8ea4000+26000] Feb 02 12:33:54 [cinnamon-session] WARNING: t+1.70045s: Application 'cinnamon-settings-daemon-xrandr.desktop' killed by signal 11_ Feb 02 12:33:54 [kernel] [ 33.891378] csd-power[24116]: segfault at 0 ip 00007f154f4b59d9 sp 00007ffcb77fc6e0 error 4 in libupower-glib.so.3.0.1[7f154f4a3000+26000] Feb 02 12:33:54 [cinnamon-session] WARNING: t+1.71240s: Application 'cinnamon-settings-daemon-power.desktop' killed by signal 11_ Feb 02 12:34:00 [cinnamon-session] WARNING: t+7.43181s: Detected that screensaver has left the bus_

I don't know where that fstab came from, but somehow that fixed the other fstab and now it started mounting /boot. Very weird indeed.

I actually changed it to device name /dev/funtoo/root / ext4 noatime 0 1 /dev/sda1 /boot ext2 noauto,noatime 1 2 /dev/funtoo/swap none swap sw 0 0 /dev/funtoo/log /var/log ext4 noatime 0 1 /dev/funtoo/root / ext4 noatime 0 1 /dev/sda1 /boot ext2 noauto,noatime 1 2 /dev/funtoo/swap none swap sw 0 0 /dev/funtoo/log /var/log ext4 noatime 0 1 And that has no effect. /boot doesn't get mounted.

Not exactly. There are two scenarios, running the grub installation and /boot mounted and one where it’s not mounted. With /boot not mounted 1) grub will report bogus LVM warnings and no actual error. 2) At the subsequent startup, grub will not find the kernel and get stuck at the grub command prompt. With /boot mounted (The rc logs represent this scenario.) 1) grub will install fine with no warnings, and the machine will boot up fine 2) /boot mount entry in /etc/fstab gets ignored and /boot is never mounted (no output from mount |grep boot) To reproduce all this, did you create a disk formatted to follow what you see in blkid? I believe this exact configuration is critical to reproducing these bugs.

Here it is again with verbose logging turned on. I don't see any reference to attempting to mount /boot specifically. rc.log

This log is from the boot where grub is working and /boot is not being mounted. rc.log

It seems to make sense that in order to run boot-update, /boot must be mounted. How else would save the grub.cfg file? But does grub installation require access to the grub.cfg file at the precise time it's being installed? If there the answer is no, then there is a bug in the installation of grub. However, either way, it shouldn't be complaining about LVM. If the answer is yes, then the only bug is that utilizing LVM blocks the mount of /boot in /etc/fstab at boot time. Here is the contents of grub.cfg set timeout=3 insmod part_msdos insmod ext2 set root=(hostdisk//dev/sda,msdos1) search --no-floppy --fs-uuid --set 2149e6d9-3d79-492d-9d6f-db6b952852a5 if loadfont /grub/unifont.pf2; then set gfxmode=text insmod all_video terminal_output gfxterm fi set menu_color_normal=cyan/blue set menu_color_highlight=blue/cyan menuentry "Funtoo Linux genkernel - kernel-debian-sources-x86_64-4.8.15-1" { insmod part_msdos insmod ext2 set root=(hostdisk//dev/sda,msdos1) search --no-floppy --fs-uuid --set 2149e6d9-3d79-492d-9d6f-db6b952852a5 linux /kernel-debian-sources-x86_64-4.8.15-1 dolvm rootfstype=ext4 resume=swap:/dev/mapper/funtoo-swap quiet real _root=/dev/funtoo/root rootfstype=ext4 initrd /initramfs-debian-sources-x86_64-4.8.15-1 set gfxpayload=keep } set default=0

There is no error as you can see. It just seems that grub can't find the kernel on /boot.

/dev/sda1: LABEL="boot" UUID="2149e6d9-3d79-492d-9d6f-db6b952852a5" TYPE="ext2" PARTUUID="4f090ca0-01" /dev/sda2: UUID="awsCeO-ZbWl-OO0i-zxsF-E22P-9WZb-08aoG1" TYPE="LVM2_member" PARTUUID="4f090ca0-02" /dev/mapper/funtoo-swap: LABEL="swap" UUID="1eca2d56-9c4b-4673-94a3-a9c76cca4967" TYPE="swap" /dev/mapper/funtoo-log: LABEL="log" UUID="92c90ca6-216f-4b77-9338-07f17073377b" TYPE="ext4" /dev/mapper/funtoo-root: LABEL="root" UUID="1fb6792e-fc44-49cb-aa57-9fcd42ee1979" TYPE="ext4

I don't understand what you mean that /boot is never mounted at boot time. I have a bunch of funtoo machines and when they startup, /boot is mounted. Only on the machine with lvm does it not mount. The original problem was that on the lvm machine, the /etc/fstab entry for mounting /boot is being ignored. Installing grub with /boot not mounted results in those LVM errors and grub being unable to load the kernel. Manually mounting /boot seems to fix this problem.

I'm not sure what problem your answer is trying to address, 1) grub tripping up when it's installed and /boot isn't mounted, or 2) why /boot doesn't mount at boot time. On other systems without LVM, there isn't either problem. You certainly don't need the --disklabel argument or an actual root passed.

LABEL=root / ext4 noatime 0 1 LABEL=boot /boot ext2 noatime 1 2 LABEL=swap none swap sw 0 0 LABEL=log /var/log ext4 noatime 0 2

/etc/fstab is configured to mount /dev/sda1 on /boot, but it's ignored. Why is it ignored? Anyway, when /boot is not mounted, these errors occur and grub ends up being confused about where the kernel/initramfs are.

I run a script regularly on my computer that prepares all the configuration. It also reinstalls grub. But recently, it occasionally has a problem with lvm and that causes the machine, which is virtual, to get stuck at the grub screen unable to find the kernel. When I install grub /usr/sbin/grub-install --target=i386-pc --no-floppy /dev/sda And it happens to respond Installing for i386-pc platform. /run/lvm/lvmetad.socket: connect failed: No such file or directory WARNING: Failed to connect to lvmetad. Falling back to internal scanning. /run/lvm/lvmetad.socket: connect failed: No such file or directory WARNING: Failed to connect to lvmetad. Falling back to internal scanning. /run/lvm/lvmetad.socket: connect failed: No such file or directory WARNING: Failed to connect to lvmetad. Falling back to internal scanning. Installation finished. No error reported. Grub will get stuck and not be able to find the kernel. Boot.conf looks like this boot { generate grub default "Funtoo Linux genkernel" timeout 3 } "Funtoo Linux genkernel" { kernel kernel-debian-sources-x86_64-4.8.15-1 initrd initramfs-debian-sources-x86_64-4.8.15-1 params += dolvm real_root=auto rootfstype=ext4 resume=swap:/dev/mapper/funtoo-swap quiet }

Reverting to earlier kernel fixes the emerge issues, but the problem of the original post here still remains. X won't run on a freshly installed 32-bit installation of funtoo.

I started a fresh install under virtualbox and errors are piling up. First, Messages for package x11-drivers/xf86-video-virtualbox-5.1.30: * You need to edit the file /etc/X11/xorg.conf and set: * * Driver "vboxvideo" * * in the Graphics device section (Section "Device") I've never seen that before. This file doesn't even exist. Second, it complains ERROR: app-emulation/virtualbox-guest-additions-5.1.30::xorg-kit failed /var/tmp/portage/app-emulation/virtualbox-guest-additions-5.1.30/work/vboxvideo/vbox_drv.c:285:5: error: unknown field ‘ set_busid’ specified in initializer .set_busid = drm_pci_set_busid, ^ /var/tmp/portage/app-emulation/virtualbox-guest-additions-5.1.30/work/vboxvideo/vbox_drv.c:285:18: error: ‘drm_pci _set_busid’ undeclared here (not in a function) .set_busid = drm_pci_set_busid, ^ /var/tmp/portage/app-emulation/virtualbox-guest-additions-5.1.30/work/vboxvideo/vbox_drv.c: In function ‘vbox_init’: /var/tmp/portage/app-emulation/virtualbox-guest-additions-5.1.30/work/vboxvideo/vbox_drv.c:329:12: error: implicit decla ration of function ‘drm_pci_init’ [-Werror=implicit-function-declaration] return drm_pci_init(&driver, &vbox_pci_driver); ^ /var/tmp/portage/app-emulation/virtualbox-guest-additions-5.1.30/work/vboxvideo/vbox_drv.c: In function ‘vbox_exit’: /var/tmp/portage/app-emulation/virtualbox-guest-additions-5.1.30/work/vboxvideo/vbox_drv.c:333:5: error: implicit declar ation of function ‘drm_pci_exit’ [-Werror=implicit-function-declaration] drm_pci_exit(&driver, &vbox_pci_driver); ^