fix(os): restore NVMe /data persistence (ext magic instead of BusyBox blkid) - #44
Open
keskad wants to merge 4 commits into
Open
fix(os): restore NVMe /data persistence (ext magic instead of BusyBox blkid)#44keskad wants to merge 4 commits into
keskad wants to merge 4 commits into
Conversation
/data never stayed on the NVMe: three independent bugs kept it on the SD card, so flashing an image wiped the SQLite database along with it. The image ships BusyBox blkid, which ignores -o value -s TYPE, so the type check never returned ext4 and both early-boot and prepare-nvme skipped the disk. Detect the ext superblock magic at offset 0x438 instead, in both. SD p3 and NVMe p1 both carried LABEL=bigfred-data, and BusyBox findfs scans /dev with an unsorted readdir(), so /data could land on either between boots. Select by device path only (marked NVMe partition first, then /dev/mmcblk0p3) and relabel the SD partition bigfred-sdboot so the collision cannot return. The migration marker was written after a non-atomic `cp -a` whose status was discarded, so a partial copy left a populated tree with no marker, which later boots read as "not empty, skip forever". Copy excluding ephemeral run/ and logs/, verify the result, then write the marker and sync. A non-empty ext4 volume without a marker is now reported as unmigrated-dirty for `adopt` or `reinit --yes` rather than silently ignored, and formatting only happens on a just-created partition or a confirmed-empty ext volume. Binary overrides lived in three layouts across seven mechanisms (two wrappers, resolve_bin, and four init scripts with their own /data/opt paths). Replace all of them with /data/bin first on PATH: init scripts exec bare names, the package installs real binaries to /usr/bin, and the UI Update tab installs to /data/bin. A changed image commit resets /data/bin so stale overrides do not outlive the image they were built against. SSH host keys move to /data/etc/dropbear so the fingerprint survives reboot and OS flash once /data itself persists. They are generated by the dropbear service rather than early-boot, because dropbearkey reads /dev/urandom without blocking and the CRNG may still be uninitialized that early. prepare-nvme grows status/adopt/reinit for operators, and reinit refuses to wipe a disk that still backs any mount: early-boot binds /etc/shadow, /root/.ssh and /etc/dropbear off /data, and those keep the filesystem live after `umount /data`. The UI exposes GET /api/v1/storage and shows a banner when an NVMe is present but /data is still on the SD. Co-authored-by: Cursor <cursoragent@cursor.com>
Bump MICRONET_REF after micronet CI published the artifact with null=auto Ethernet selection (first physical iface with carrier after PHY poll). Co-authored-by: Cursor <cursoragent@cursor.com>
Carrier-based iface selection is on micronet main; no sha pin needed. Co-authored-by: Cursor <cursoragent@cursor.com>
Restore the NVMe /data persistence fix (fb11934): detect ext superblock magic instead of BusyBox blkid -o/-s TYPE, which never returned ext4 and skipped mounting /dev/nvme0n1p1.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Restores
fb11934(fix/nvme-data-persistence) onto currentmain. That commit never landed (reset offmainon 2026-08-24, recovered onto the branch, no PR).BusyBox
blkidin the image ignores-o value -s TYPE, so bothearly-boot.shandprepare-nvmenever sawext4and skipped mounting NVMe:What this brings back
0xEF53at offset0x438in Rust (is_ext_family) and inearly-boot.sh(dd/od), instead ofblkid -o/-s./databy device path (marked NVMe first, then/dev/mmcblk0p3); relabel SDbigfred-sdbootso BusyBoxfindfscannot pick the wrong volume.status/adopt/reinitonprepare-nvme; UIGET /api/v1/storagebanner when NVMe is present but/datais still on SD./data/binfirst on PATH; SSH host keys under/data/etc/dropbear.scripts/check-os.shfails the build ifblkid -o/blkid -sreappear.MICRONET_REFstays onmain(the two follow-up pin/unpin commits on the source branch cancel out).Local
scripts/check-os.sh: OK.