[Bug 1047385] New: Installer doesn't set up boot loader to boot a windows partition
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385 Bug ID: 1047385 Summary: Installer doesn't set up boot loader to boot a windows partition Classification: openSUSE Product: openSUSE Distribution Version: Leap 42.3 Hardware: x86-64 OS: SUSE Other Status: NEW Severity: Normal Priority: P5 - None Component: Installation Assignee: yast2-maintainers@suse.de Reporter: tjcw@physics.org QA Contact: jsrain@suse.com Found By: --- Blocker: --- I have a laptop with an internal disk and an external disk. The internal disk is set up with Windows 10 and Leap 42.2 . The external disk is for testing, and has Leap 42.3 and Fedora. When I update the Leap 42.3 system, the installer sets up the boot menu for Leap 42.3, Leap 42.2, and Fedora; but the Windows installation is missing. If I boot Leap 42.3, bring up YAST, go into bootloader, and select OK, the bootloader is re-set-up and at the next boot all 4 systems are offered including Windows. So the problem is only with the installer's setup of the bootloader. Does the installer keep a log of what it has done ? Tell me which file and I will upload it as an attachment. I am currently at Leap 42.3 build 0308. My laptop has secure (EFI) boot. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c12
Michael Chang
2017-07-08T11:12:30.818676+00:00 install os-prober: debug: running /usr/lib/os-probes/50mounted-tests on /dev/sda2 2017-07-08T11:12:31.200487+00:00 install 50mounted-tests: debug: mounted using GRUB fat filesystem driver 2017-07-08T11:12:31.204085+00:00 install 50mounted-tests: debug: running subtest /usr/lib/os-probes/mounted/05efi 2017-07-08T11:12:31.214430+00:00 install 05efi: debug: /dev/sda2 is a FAT partition (mounted by GRUB) 2017-07-08T11:12:31.358373+00:00 install 05efi: debug: /dev/sda2 container is '' 2017-07-08T11:12:31.365021+00:00 install 05efi: debug: /dev/sda2 partition scheme is 2017-07-08T11:12:31.368233+00:00 install 05efi: debug: /dev/sda2 partition type is 2017-07-08T11:12:31.371530+00:00 install 05efi: debug: /dev/sda2 is not a ESP partition: exiting
The partition scheme and type is retrieved from udev database 'udevadm info -q property -n /dev/sda2' of the two records ID_PART_ENTRY_TYPE ID_PART_ENTRY_SCHEME And both has no value during installation, in turn having the check result to be false one. Meanwhile I tested Tumbleweed on virtual machine and did not reproduced the problem, the Windows can be properly probed. Perhaps the problem is 42.3 specific ? Anyway lets needinfo udev maintainer to see if any better idea. Hi Franck Do you have any better idea ? The offending script in os-prober invoking "udevadm info" to retrieve the properties can be found here : https://anonscm.debian.org/git/d-i/os-prober.git/tree/os-probes/mounted/x86/... Thanks. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c13
--- Comment #13 from Chris Ward
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c14
--- Comment #14 from Michael Chang
I installed tumbleweed and got the same symptoms as before -- no Windows on the boot menu. The same workaround as before worked; go into Yast bootloader and regenerate the grub2 configuration on a running system.
That is strange, it seems udev is somehow misbehaving in installer environment and fully work on installed system.
The difference between my system and your virtual system is that I boot with GRUB-EFI (secure boot yes, trusted boot no) whereas you boot with 'legacy' GRUB. Could this account for the different behaviour ?
No. I was also booted the same setup with yours, that is GRUB-EFI with Secure Boot "ON".
Is it worth asking on the factory mailing list for some more testing in this area ?
I just downloaded and tested Leap 42.3 Build 0329, the problem was not reproducible. So maybe you're right to ask on factory to know is it a common problem or specific to certain hardware or setup .. Thanks. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c15
--- Comment #15 from Michael Chang
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c16
--- Comment #16 from Chris Ward
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c17
--- Comment #17 from Chris Ward
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c18
Franck Bui
Do you have any better idea ? The offending script in os-prober invoking "udevadm info" to retrieve the properties can be found here :
Hi Michael, Sorry for the delay, I was OFF until yesterday. udev relies on blkid library to retrieve the ID_PART_ENTRY_* values. Therefore if they are missing from the udevadm output then they are probably also missing from the data returned by the blkid library. To verify that the following commands can be run: $ blkid --probe /dev/sda2 -o udev and also: $ udevadm test-builtin blkid /block/sda/sda2 -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c19
Chris Ward
To verify that the following commands can be run:
$ blkid --probe /dev/sda2 -o udev
and also:
$ udevadm test-builtin blkid /block/sda/sda2
When I'm in the installer environment, presumably these commands need to be run as if on the target system. I can mount the target system's root disk, but you need to tell me what sequence of 'chroot' command and setting of PATH and LD_LIBRARY_PATH I need to get the right environment for these commands. To get a shell for issuing these commands, I run the installer for upgrade until it gives its countdown for 10 seconds to reboot; use the mouse to stop the reboot; and then 'Ctrl-Alt-Fn' to get a shell. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c20
Franck Bui
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c21
--- Comment #21 from Michael Chang
To get a shell for issuing these commands, I run the installer for upgrade until it gives its countdown for 10 seconds to reboot; use the mouse to stop the reboot; and then 'Ctrl-Alt-Fn' to get a shell.
The target system is mounted on /mnt during the installation, so for testing those commands you have to
chroot /mnt
But installer might have un-mounted /mnt in countdown, so you probably need to
mount
/mnt mount --bind /sys /mnt/sys mount --bind /proc /mnt/proc mount --bind /dev /mnt/dev chroot /mnt
Or you could hit <ABORT> at the very end of install (before countdown) to avoid the hassle of mount root partition and use simple chroot to obtain a shell to test the commands. (You could continue and cancel abort if you like after the test). -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c22
--- Comment #22 from Michael Chang
To verify that the following commands can be run:
$ blkid --probe /dev/sda2 -o udev
My blkid does not have --probe, is it equivalent to -p ? -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c23
--- Comment #23 from Michael Chang
Hi Chris, these commands need to be run where os-prober is.
@Franck, Does that mean we have to patch os-prober and use these commands (blkid) as fallback or replace existing command (udevadm info ..) to probe the ESP ? If so, what's the rationale you think that they might work better ? Thanks. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c24
--- Comment #24 from Chris Ward
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c25
--- Comment #25 from Chris Ward
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c26
--- Comment #26 from Franck Bui
Does that mean we have to patch os-prober and use these commands (blkid) as fallback or replace existing command (udevadm info ..) to probe the ESP ? If so, what's the rationale you think that they might work better ?
Hi Michael, no I wasn't suggesting to use blkid instead of udevadm. I just want to make sure that it is libblkid which returns empty values for ID_PART_ENTRY_* vars. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c27
Michael Chang
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c28
--- Comment #28 from Michael Chang
(In reply to Michael Chang from comment #23)
Hi Michael, no I wasn't suggesting to use blkid instead of udevadm. I just want to make sure that it is libblkid which returns empty values for ID_PART_ENTRY_* vars.
Chris seems to take the result of comment#24 "during" installation, therefore libblkid seems to work correctly ... @Chris Could you please help to confirm how did you test for results of comment#24 .. ? Thanks. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c29
--- Comment #29 from Franck Bui
'blkid --probe' gave a usage error message. [...] ID_PART_ENTRY_SCHEME=gpt ID_PART_ENTRY_TYPE=c12a7328-f81f-11d2-ba4b-00a0c93ec93b
'udevadm test-builtin blkid /block/sda/sda2' gave (to stdout) [...] ID_PART_ENTRY_SCHEME=gpt ID_PART_ENTRY_TYPE=c12a7328-f81f-11d2-ba4b-00a0c93ec93b
Hmm here the variables are defined... Are you sure those commands are executed in the right environment ? I'm really not sure and at least the output of "os-prober" in comment #16 seems to suggest the contrary:
Jul 19 19:58:04 linux-p2r7 05efi[3087]: debug: /dev/sda2 partition scheme is gpt Jul 19 19:58:04 linux-p2r7 05efi[3088]: debug: /dev/sda2 partition type is c12a7328-f81f-11d2-ba4b-00a0c93ec93b
Here os-prober detected correctly the type and the scheme used by sda2 which is different from the output shown in comment #12. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c30
--- Comment #30 from Michael Chang
(In reply to Chris Ward from comment #24)
Here os-prober detected correctly the type and the scheme used by sda2 which is different from the output shown in comment #12.
To be clear, the comment#12 is quoted from /var/log/messages file attached in comment#4. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c33
Josef Reidinger
Hi Josef,
I couldn't find "Include Add-On Products from Separate Media" options any longer in openSUSE Tumbleweed and Leap 42.3 installation. Is the function removed ?
it is removed from UI. You can still use it as boot parameter addon=<URL> or as steffen said you can use dud which affect also insts-sys -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c34
--- Comment #34 from Michael Chang
Michael, create a driver update and use it:
mkdud -c foo.dud -d leap42.3 foo.rpm
Then boot with 'dud=
'.
Thanks to remind me of dud, yes indeed it should be a good alternative. However, while testing the dud, I encountered integrity check failure due to key not available, which is understandable. But even passing insecure=1 to the installer still can't get rid of the error, which is a bit surprised as it seems to work for me in the past. I'll attach the screen shots for reference, not sure if this is a bug or of my own mistake. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c35
--- Comment #35 from Michael Chang
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385
http://bugzilla.opensuse.org/show_bug.cgi?id=1047385#c36
--- Comment #36 from Michael Chang
participants (1)
-
bugzilla_noreply@novell.com