[Bug 959641] New: When installing on UEFI, yast proposes /boot partition unnecessarily for LVM+Encryption proposal.
http://bugzilla.suse.com/show_bug.cgi?id=959641 Bug ID: 959641 Summary: When installing on UEFI, yast proposes /boot partition unnecessarily for LVM+Encryption proposal. Classification: openSUSE Product: openSUSE Tumbleweed Version: 2015* Hardware: Other OS: Other Status: NEW Severity: Major Priority: P5 - None Component: Installation Assignee: yast2-maintainers@suse.de Reporter: rbrown@suse.com QA Contact: jsrain@suse.com CC: kukuk@suse.com Found By: --- Blocker: --- Created attachment 659793 --> http://bugzilla.suse.com/attachment.cgi?id=659793&action=edit y2logs Background information: http://lists.opensuse.org/opensuse-factory/2015-12/msg00071.html Most relevant post: http://lists.opensuse.org/opensuse-factory/2015-12/msg00257.html GOAL: To install Leap, Tumbleweed or SLE on a machine with LVM and Encryption and still have the 'Boot to Snapshot' feature operational. This should work fine on any disk with a GPT partition table. STEPS TO REPRODUCE: - Start with a blank disk with a GPT partition table (parted /dev/sda mklabel gpt) - Boot Tumbleweed, Leap 42.1, or SLE 12 SP1 installation media - On the Partition proposal screen, click 'Edit Partition Proposal' - Select "Create LVM-based Partition Proposal" and "Encrypt Volume Group" - Ensure that 'btrfs' is picked as the root filesystem and 'Enabled Snapshots' is selected - Click OK EXPECTED RESULTS: Proposal should include include an LVM Volume Group which contains / (root), /home and swap partitions There should not be a separate /boot partition outside the LVM Volume Group ACTUAL RESULTS: If the system boots to BIOS: Proposal *MATCHES* Expected Results. You end up with a proposal that includes a small Grub/BIOS partition, a large LVM partition, and an LVM Volume Group which contains /, /home, and swap There is no separate /boot partition outside of the LVM Volume Group If the system boots to UEFI (or UEFI Secure Boot): Proposal *DOES NOT MATCH* Expected Results Proposal consists of a small (156MB) /boot/efi partition, a slightly larger (400MB) /boot partition, and then the remainder of the disk is an LVM Volume Group which contains /, /home, and swap CONSEQUENCE: On UEFI systems, the LVM + Encryption partition proposals presented by YaST result in a system which has a separate /boot partition and therefore a broken 'Boot to Snapshot' Feature YaST should not do this, as there does not seem to be any technical reason to enforce a separate /boot partition when using LVM+Encryption. y2logs from a Leap 42.1 UEFI installation are attached -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c1
Ancor Gonzalez Sosa
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c2
Thorsten Kukuk
Arvin, is this realistically fixable with the current code base?
It works for GPT tables with legacy BIOS today. So the code is there and works. We only need to adjust the check for UEFI and GPT tables. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c3
Arvin Schnell
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c4
--- Comment #4 from Thorsten Kukuk
https://bugzilla.suse.com/show_bug.cgi?id=887888#c37 shows that the extra /boot partition was needed.
No, the opposite: this comment is from me and documents, what for grub2 needs to be changed that it works without extra /boot partition. And snwint fixed it afterwards in perl-Bootloader. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c5
--- Comment #5 from Arvin Schnell
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c6
Richard Brown
https://bugzilla.suse.com/show_bug.cgi?id=887888#c37 shows that the extra /boot partition was needed. If the bootloader is now per default capable to handle the situation please make a feature request to change the proposal. Then it is documented and gets tested.
As confirmed by Thorsten in Comment #4, the details from the previous bug ( https://bugzilla.suse.com/show_bug.cgi?id=887888#c37 ) show that a /boot partition is not needed This is further confirmed by the fact that this works perfectly fine in the Legacy BIOS proposal already provided by YaST YaST supports LVM+Encryption with BTRFS without a /boot partition on Legacy BIOS, and it should do so on UEFI also. I therefore consider this a bug and not worthy of a feature request until a time where someone explains why /boot is needed in UEFI (but clearly doesn't in any other case) (In reply to Arvin Schnell from comment #5)
Before the change it *was* needed. And nobody included in the bug took care that the proposal was changed.
It looks to me like someone took care of the Legacy BIOS proposal but not the UEFI proposal, leading to this bug.. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c7
Lukas Ocilka
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c8
Arvin Schnell
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c10
--- Comment #10 from Stefan Hundhammer
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c11
Lukas Ocilka
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c12
--- Comment #12 from Stefan Hundhammer
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c13
Jiri Srain
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c14
Stefan Hundhammer
We should get this unified, the question is whether it can still be safely done for 42.3. Stefan, what is the risk for current proposal code?
IMHO the risk is to break it completely for all kinds of scenarios. This is exactly the most sore point of that old storage code: There are multiple entry points (which one is used in any given scenario is not clear at all - we'd have to find out by trial and error), there are no small-scale unit tests, just the full-blown complete installation tests that OpenQA does. Since all this will be obsolete in couple of months anyway with the advent of storage-ng, IMHO it is not worthwhile to try to optimize for this one use case. In storage-ng, we have unit tests on various levels with good coverage of possible installation scenarios; we don't have to rely on OpenQA alone. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c15
--- Comment #15 from Stefan Hundhammer
We should get this unified, the question is whether it can still be safely done for 42.3. Stefan, what is the risk for current proposal code?
IMHO the risk is to break it completely for all kinds of scenarios. This is exactly the most sore point of that old storage code: There are multiple entry points (which one is used in any given scenario is not clear at all - we'd have to find out by trial and error), there are no small-scale unit tests, just the full-blown complete installation tests that OpenQA does. Since all this will be obsolete in couple of months anyway with the advent of storage-ng, IMHO it is not worthwhile to try to optimize for this one use case. In storage-ng, we have unit tests on various levels with good coverage of possible installation scenarios; we don't have to rely on OpenQA alone. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c16
--- Comment #16 from Jiri Srain
http://bugzilla.suse.com/show_bug.cgi?id=959641
Stefan Hundhammer
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c17
Stefan Hundhammer
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c18
Richard Brown
http://bugzilla.suse.com/show_bug.cgi?id=959641
http://bugzilla.suse.com/show_bug.cgi?id=959641#c19
Jiri Srain
participants (1)
-
bugzilla_noreply@novell.com