Retry a format that failed, instead of skipping it for ever (#168) - #169
Merged
Merged
Conversation
A board sat unusable across five reboots and a fresh flash: storage FAILED, no images, mount refusing /dev/sda2 for a bad superblock, ssh-keygen-boot unable to restore host keys so sshd never started, and - with no readable options.cfg - the board on its own hotspot instead of the WiFi it was configured for. It read as a bad flash. It was a partition with no filesystem: correctly sized, and blank from its first byte. expand-usb asked whether the *partition* existed to decide whether there was anything to do, so the one thing that could fix the drive was the step being skipped. It now asks whether there is a filesystem - lsblk reports an empty FSTYPE, and lsblk is in the image - and formats a partition that has none without re-partitioning it. Why the format failed at all: on the first boot of a freshly written stick, ssh-keygen-boot started at 11:29:21 and failed at 11:29:22. mkfs on this stick takes ~31s by the measurement in the comment below, and that boot's log reached "Creating ext4 filesystem" - so mkfs was invoked and returned an error at once rather than being cut short. The device is fine: the same mkfs by hand afterwards succeeded in two seconds and brought the board straight back. That points at mke2fs opening the device O_EXCL while udev is still probing a partition that appeared a moment earlier - the same race as #145 and #147 - so there is a settle before the format now. It points, rather than proves, and that is the third fix here: mkfs wrote its reason to stderr, stderr went to the journal, and this rootfs is a ramdisk, so the explanation died with the reboot that followed. It goes to the reflash log now. set -e already stopped the script; nothing recorded why. The tests are static checks on the script text, like flash-cleanup.bats and for the same reason - this script hardcodes /dev/sda with no seams, and a real run needs root and losetup. All five defect tests fail against the old script. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
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.
Fixes #168.
A board sat unusable across five reboots and a fresh flash -
storage: FAILED, mount refusing/dev/sda2for a bad superblock,ssh-keygen-bootunable to restore host keys so sshd never started, and with no readableoptions.cfgthe board on its own hotspot instead of the configured WiFi. It read as a bad flash. It was a partition with no filesystem: correctly sized, blank from its first byte.Three changes
The guard.
expand-usbasked whether the partition existed to decide whether there was work to do, so oncesda2had been created a failed format was never retried - the one step that could fix the drive was the one being skipped. It now asks whether there is a filesystem (lsblk -no FSTYPE, andlsblkis in the image), and formats a partition that has none without re-partitioning it.A settle before mkfs. On the first boot of a freshly written stick,
ssh-keygen-bootstarted at 11:29:21 and failed at 11:29:22.mkfson this stick takes ~31s by this script's own measured comment, and that boot's log reached "Creating ext4 filesystem" - so mkfs was invoked and returned an error immediately, rather than being cut short. The device is fine: the samemkfsby hand afterwards succeeded in two seconds and brought the board straight back. That points atmke2fsopening the deviceO_EXCLwhile udev is still probing a partition that appeared a moment earlier - the same race as #145 and #147.mkfs's output into the log. It points rather than proves, and that is the third fix: mkfs wrote its reason to stderr, stderr went to the journal, and this rootfs is a ramdisk, so the explanation died with the reboot.
set -ealready stopped the script; nothing recorded why.Tests
Static checks on the script text, like
flash-cleanup.batsand for the same reason: this script hardcodes/dev/sdawith no seams, and a real run needs root and losetup. All five defect tests fail against the old script and pass against this one.make test: Go, bats (99 including the six new), and 144 vitest all pass.Note
The failure is intermittent - plenty of boots format cleanly - but its result was not: one unlucky second left a board with no storage, no ssh and no WiFi config until someone formatted the partition by hand over the serial console.
🤖 Generated with Claude Code