From a076deb55a4db8ee326e6c0b72275632b7392587 Mon Sep 17 00:00:00 2001
From: Elias Bakken <elias@iagent.no>
Date: Mon, 10 Aug 2026 21:11:45 +0200
Subject: [PATCH] Fix create-recore-config to actually provision the eMMC
 config

Three bugs, all reported in #85:

1. Formatted a nonexistent partition node - mkfs.ext4 targeted
   /dev/mmcblk2boot0p1, but the kernel doesn't create child partition
   nodes for eMMC boot partitions. mount-config and every other reader
   loop-mount the raw device directly at offset 17408 instead - the
   filesystem now gets created the same way, via a loop device at that
   offset, not via fdisk/a partition node that never existed.

2. metadata_csum (default-on in e2fsprogs 1.47) left no room for the
   dir-leaf checksum once the filesystem picked a 1K block size at
   this tiny size, so reads failed with "Bad message". Disabled via
   -O ^metadata_csum,^64bit.

3. The create path wrote through mount-config, which always mounts
   read-only (it's meant for readers) - the echo > serial_number would
   fail even with a valid filesystem. The create path now mounts the
   loop device directly, read-write, instead.

Also fixed a related bug in get-recore-serial-number: it called plain
umount instead of unmount-config to tear down the transient mount
mount-config sets up via systemd-mount. Plain umount unmounts the
filesystem but doesn't deregister the transient systemd unit, so a
later mount-config call can fail with "Unit mnt-config.mount was
already loaded".

Live-tested on real hardware: extracted and backed up the board's
existing (already-provisioned) config partition first, ran the fixed
create-recore-config against the same serial, and confirmed
get-recore-revision/get-recore-serial-number both work and the
downloaded calibration.json is byte-identical to the original.

Closes #85
---
 bin/prod/create-recore-config     | 38 +++++++++++++++++++++++--------
 bin/prod/get-recore-serial-number |  2 +-
 test/bats/getters.bats            |  2 +-
 3 files changed, 30 insertions(+), 12 deletions(-)

diff --git a/bin/prod/create-recore-config b/bin/prod/create-recore-config
index 2dfcfce..953968f 100755
--- a/bin/prod/create-recore-config
+++ b/bin/prod/create-recore-config
@@ -1,6 +1,9 @@
 #!/bin/bash
 
+set -euo pipefail
+
 SERIAL_NR=$1
+OFFSET=17408
 
 info() {
     echo "[info] $1" >> /var/log/reflash.log
@@ -52,23 +55,38 @@ fi
 info "Calibration file valid"
 
 DEV=`lsblk -n -o NAME | grep 'mmcblk[0-2]$'`
+DEV_BOOT="/dev/${DEV}boot0"
 
 echo 0 > "/sys/block/${DEV}boot0/force_ro"
 
-info "Creating boot partition"
-dd if=/dev/zero of=/dev/${DEV}boot0
-partprobe
-printf "g\nn\n\n\n\nw\n" | fdisk /dev/${DEV}boot0
-partprobe
-mkfs.ext4 -E nodiscard /dev/${DEV}boot0p1
+# The kernel doesn't create a child partition node for eMMC boot
+# partitions (there's no ${DEV_BOOT}p1), and mount-config/other readers
+# loop-mount the raw device directly at a byte offset - so the
+# filesystem has to be created the same way, not via fdisk/mkfs on a
+# nonexistent partition node.
+info "Creating config filesystem"
+# Zeroing the whole device this way always ends in "No space left on
+# device" once dd hits the end of the block device - that's the
+# expected way it finishes, not a real failure.
+dd if=/dev/zero of="$DEV_BOOT" bs=1M || true
+LOOP=$(losetup -f --show -o "$OFFSET" "$DEV_BOOT")
+# metadata_csum (default-on in e2fsprogs 1.47) leaves no room for the
+# dir-leaf checksum once the filesystem picks a 1K block size, which it
+# does at this size - reads then fail with "Bad message". Disable it.
+mkfs.ext4 -F -q -O ^metadata_csum,^64bit -E nodiscard "$LOOP"
 
-mount-config
+# mount-config always mounts read-only (it's meant for readers); the
+# create path needs to write, so mount the loop device directly rather
+# than going through mount-config/unmount-config here.
+mkdir -p /mnt/config
+mount "$LOOP" /mnt/config
 echo "$SNR" > /mnt/config/serial_number
 
 info "Downloading calibration file ${FILE}"
-wget $URL
-mv ${FILE} /mnt/config
+wget -O "/mnt/config/${FILE}" "$URL"
 
-unmount-config
+sync
+umount /mnt/config
+losetup -d "$LOOP"
 echo 1 > "/sys/block/${DEV}boot0/force_ro"
 info "Done"
diff --git a/bin/prod/get-recore-serial-number b/bin/prod/get-recore-serial-number
index dfb4913..407dc90 100755
--- a/bin/prod/get-recore-serial-number
+++ b/bin/prod/get-recore-serial-number
@@ -6,4 +6,4 @@ CONFIG_DIR="${REFLASH_CONFIG_DIR:-/mnt/config}"
 
 mount-config
 cat "$CONFIG_DIR/serial_number"
-umount "$CONFIG_DIR/"
+unmount-config
diff --git a/test/bats/getters.bats b/test/bats/getters.bats
index ea11b4e..77c3766 100644
--- a/test/bats/getters.bats
+++ b/test/bats/getters.bats
@@ -22,7 +22,7 @@ teardown() { teardown_sandbox; }
   mkdir -p "$REFLASH_CONFIG_DIR"
   echo "RC-0001-XYZ" > "$REFLASH_CONFIG_DIR/serial_number"
   stub_silent mount-config
-  stub_silent umount
+  stub_silent unmount-config
   run "$PROD_BIN/get-recore-serial-number"
   [ "$status" -eq 0 ]
   [ "$output" = "RC-0001-XYZ" ]