Repository navigation
feat(installer): adopt Flatcar's EFI-SYSTEM/USR-A/USR-B/ROOT partition layout - #147
kylerankin wants to merge 5 commits into
Conversation
Replace the installer's target-disk repart.d drop-ins (10-esp.conf, 20-root-a.conf, 30-var.conf) with Flatcar's GPT layout, adopting the upstream type GUIDs verbatim (projectbluefin#134): EFI-SYSTEM ESP/vfat bootctl install + the OS UKI USR-A 5dfbf5f4-... CopyBlocks= the /usr DDI (read-only) USR-B 5dfbf5f4-... empty A/B rollback slot OEM 0fc63daf-... ext4, filesystem label OEM (stage-2 boot dep) ROOT 3884dd41-... writable state, grows to fill the disk BIOS-BOOT stays dropped (UEFI-only, hard rule 5); OEM-CONFIG stays dropped (nothing in the boot path references it). Verity stays off: stage 1 guards the veritysetup block behind [ "${usrhash}" != "" ], so this layout-only change is a bootable intermediate state and verity lands in the /usr DDI ticket. The installer media still stamps bluefin-installer-data, so USR-A's CopyBlocks= source is unchanged. Update the DDI and installer element docs, the fallback bootloader label (bluefin-server-root-a -> ROOT), and the docs/skills target-layout tables. tests/unit/test_repart_layout.py is rewritten for the new layout: it pins the four upstream type GUIDs verbatim, asserts USR-B is provisioned and empty, OEM is labelled ext4, and ROOT grows; the sysupdate label-match is moved to an xfail since wiring 50-root.transfer to USR-A/USR-B is a follow-on ticket. Signed-off-by: kylerankin <kylerankin@users.noreply.github.com>
|
Closing as a duplicate of #164, which is the copy to keep. They are the same change, not two related ones. I diffed the two patches against each other: both are 483-line diffs touching an identical list of 14 files, and they differ in exactly one substantive respect — a single test function name in Everything else is byte-identical. #164 is the one to keep on CI grounds: its Nothing is lost by closing here. If you prefer this branch's test name — and for what it is worth Please continue the discussion on #164. I have left a review there covering a blocking issue that applies equally to both branches, so it is worth reading before either is revived. |
Summary
Adopts Flatcar's GPT partition layout for the installer's target disk, per
issue #134 (design: phase 3 of the Flatcar base migration). Replaces the
old
10-esp.conf/20-root-a.conf/30-var.confdrop-ins with fivethat use upstream type GUIDs verbatim:
EFI-SYSTEMbootctl install+ the OS UKIUSR-A5dfbf5f4-…CopyBlocks=the/usrDDI (read-only)USR-B5dfbf5f4-…OEM0fc63daf-…Format=ext4, filesystem labelOEMROOT3884dd41-…Format=,GrowFileSystem=yesBIOS-BOOTstays dropped (UEFI-only, hard rule 5).OEM-CONFIGstaysdropped (nothing in the boot path references it).
OEMis kept — the bootproof showed stage 2 waits on
dev-disk-by-label-OEM.device, so a diskwithout it stalls to an emergency shell.
Verity stays off in this ticket: stage 1 already guards the
veritysetup block behind
[ "${usrhash}" != "" ], so this layout-onlychange is a bootable intermediate state. Verity lands in the
/usrDDIticket.
Changes
files/installer/repart.d/: rewrite the drop-ins to the new layout.tests/unit/test_repart_layout.py: rewritten for the new layout — pinsthe four upstream type GUIDs verbatim, asserts
USR-Bis provisioned andempty,
OEMis labelled ext4, andROOTgrows. The sysupdatelabel-match moves to an
xfailsince wiring50-root.transfertoUSR-A/USR-Bis a follow-on.bootloader partition label (
bluefin-server-root-a→ROOT),installer-repart.bst layout comment, and the
docs/skillstarget-layouttables.
Verification
python3 -m pytest tests/unit/test_repart_layout.py: 24 passed, 1 xfailed.check-release-versionandcheck-k0s-version: OK.just validate'sbst showsteps need buildstream/network and were notrunnable in this environment; the graph is unchanged in shape.
— hive: backend=pi model=lemonade/Ornith-1.5-35B-A3B-GGUF-Q6_K
🐝 Hive Agent:
contributor| SHA:dd744dd