| Summary: | Boot fails due to extra entry in /etc/fstab when choosing "Use existing partitions" and not checking in one of the parts | ||
|---|---|---|---|
| Product: | Mageia | Reporter: | Charles Edwards <cae> |
| Component: | Installer | Assignee: | Thierry Vignaud <thierry.vignaud> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | Normal | CC: | marja11, pterjan, tmb, zen25000 |
| Version: | Cauldron | Keywords: | NEEDINFO |
| Target Milestone: | --- | ||
| Hardware: | x86_64 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Source RPM: | drakx-installer-stage2 | CVE: | |
| Status comment: | |||
| Attachments: | bug report from non-bootable install | ||
|
Description
Charles Edwards
2016-06-10 04:49:13 CEST
If this is an installer bug, then /root/drakx/report.bug.xz is needed. Please be as kind as to attach it ;-) Keywords:
(none) =>
NEEDINFO and /boot/grub2/grub.cfg, too :-) CC:
(none) =>
zen25000 grub.cfg is already in report.bug.xz... (The error also appears when boot doesn't fail, so removing it from the summary) Summary:
Boot fails if / and /usr are on separate btrfs partitions (error: sparse file not found) =>
Boot fails if / and /usr are on separate btrfs partitions (In reply to Marja van Waes from comment #4) > (The error also appears when boot doesn't fail, so removing it from the > summary) Yes, that message may be ignored - it's not an issue. I do not have any of the requested reports available. There have been at least 4 other test installs done on that drive since that failed boot. The original problem with the boot failing was sent to the dev list by Thomas Spuhler. I simply tested his issue and then made the bug report. On my latest test install to that drive all partitions were created and formatted using gparted. / /usr swap /home all were formatted as primary and btrfs. This install had no problem re-booting to a working desktop. I'll try the install again this weekend, creating the partitions during the install and get reports to attach in case it still fails on re-boot. Created attachment 7967 [details]
bug report from non-bootable install
I ran the install again on the working bootable drive partitioned by gparted but on this occasion I allowed the installer to format both / sdb1, and /usr sdb2.
On boot after the install the boot again only makes it to a dracut prompt.
I think I now know what is causing this but have no idea on a fix.
When running the installer probe finds /dev/sdb1 UUID=A and /dev/sdb2 UUID=B.
When the partition are formatted they are changed: /dev/sdb1 UUID=aa and /dev/sdb2 UUID=bb.
There is some important file or something that does get updated with the /dev/sdb2 UUID change.
Boot will now fail as it searches for UUID=B which no longer exist.
OK it looks like there's a bug when using "Use existing partitions" and not checking in one of the existing partitions Status:
NEW =>
ASSIGNED
Thierry Vignaud
2016-06-11 09:57:32 CEST
Source RPM:
drakx-installer-stage2, grub2 =>
drakx-installer-stage2 Did some more test installs this morning. All test were done using nonfree boot.is from Jun 11 5:42:00am Stage2 was version 17.36 and using existing partitions. *Test 1 Allowed installer to format / and /usr After installation compelted 1st boot still get spurious UUID error and boot fails. *Test 2 Allowed installer to format / and /usr Here I re-booted and re-started the installation Set mount points for / /usr /home DID NOT format any partition and completed the install. On 1st boot system boots to a working desktop. Can you still reproduce as of drakx 17.41? Yes it's still happens. With btrfs / and /usr on mbr install to gpt disk a UUID is missing during boot. I will test again tomorrow using msdos disk and also test using EFI install and see how they perform when / and /usr already exist. Also need to see if it occurs with all fs types are only btrfs. Tried the MBR install on gpt drive with / and /usr both ext4. There was no issue during the install and after there was No issue booting to a working desktop. At this point it is my opinion the this "bug" will only occur if partitions are btrfs. If there is a listing, as such, I would have no problem if the this were set as "Later". (In reply to Charles Edwards from comment #7) The thing is for now we cannot keep the UUID when re-formating btrfs: - btrfstune needs to answer "y" - mkfs.btrfs fails with non unique UUID (as it detected the fs we're formatting has the same...) CC:
(none) =>
pterjan, tmb Though I've a patch for that commit 3d24a1eb51fd57b0025d997002202cf6b3ce5678
Author: Thierry Vignaud <thierry.vignaud@...>
Date: Fri Jun 24 13:00:14 2016 +0200
keep UUID when formating btrfs (mga#18673)
---
Commit Link:
http://gitweb.mageia.org/software/drakx/commit/?id=3d24a1eb51fd57b0025d997002202cf6b3ce5678
Can you confirm its fixed with drakx v17.45 when it lands on your favorite mirror? You can do a net install using install/images/boot-nonfree.iso You can check stage2 is up to date on your mirror by checking install/stage2/VERSION eg: http://distrib-coffee.ipsl.jussieu.fr/pub/linux/Mageia/distrib/cauldron/x86_64/install/images/boot-nonfree.iso http://distrib-coffee.ipsl.jussieu.fr/pub/linux/Mageia/distrib/cauldron/x86_64/install/stage2/VERSION Status:
ASSIGNED =>
RESOLVED I can confirm that the issue is now resolved. / and /usr btrfs and install to existing partition. Resulting system boots to a working desktop. |