Skip to content

device-oriole: record native boot and RTC status finding - #200

Closed
Gr-linkk wants to merge 5 commits into
derekross:mainfrom
Gr-linkk:codex/oriole-rtc-evidence
Closed

Gr-linkk wants to merge 5 commits into
derekross:mainfrom
Gr-linkk:codex/oriole-rtc-evidence

Conversation

@Gr-linkk

@Gr-linkk Gr-linkk commented Oct 31, 2026 •

Copy link
Copy Markdown
Contributor

Requested tooling and metadata evidence (2026-10-31)

The vbmeta_b and devinfo writes were done without project approval. NOTES now says that explicitly. No further devinfo or vbmeta writes will be made without the maintainer's explicit approval; this follow-up is an offline audit.

Installer and push: the recovered October 28 project snapshot is based on 1155a7c352edf6313fb4e2ff02f054c9cb1902b6. Its tools/install-phone is byte-identical to that base. Its bin/omamobile differs: push adds the optional PUSH_MODULES_POLICY=require-match refusal, follows image symlinks for readback length, and uses systemd's reboot when PID 1 is systemd. The complete diff of both files is below, including the other builder/Fastboot changes in bin/omamobile. This compares the recovered source snapshot to its pinned upstream base, not today's main, and does not establish the exact script bytes used for every earlier deployment.

Snapshot SHA-256:

  • tools/install-phone: bd194ef43768a2aa94e34fa7fbb762fdc22a38e713d02b0bda1ed62008a10d07 (also the base file's hash).
  • bin/omamobile: 302c5e717fee5d720d6a4edff1cead8efa8c31cab996a72228f5bb98d5e852e3.

The recovered Oriole descriptor has BOOT_IMAGES="boot vendor_boot dtbo" and UFS_RW="boot_b vendor_boot_b dtbo_b userdata". The devinfo and vbmeta_b changes were separate bring-up operations.

What made devinfo writable: recovered source for the separate mark-b-success-37.py gate explicitly used blockdev --setrw on the containing boot LUN and then the devinfo partition. It restored the partition and disk to read-only with an EXIT trap, a 20-second systemd undo timer and a 120-second watchdog. It did not stop the watcher or alter UFS_RW: the historical init watcher skipped nodes already marked in /run/ufs-ro/, so it did not continuously reassert their read-only flags. The opening was an explicit block-device RO change, not an installer operation or device-node permission change.

The helper performs a 128-byte header write, with only bit 1 at 0x35 changed, followed by full 8192-byte verification. The recovered gate execution result at 2026-10-18 09:04:51 UTC reports that one-bit change, unchanged A/opaque bytes, both nodes restored RO, and refreshed backups verified.

Requested before/after evidence: the original gate's devinfo-before-b-success.img, devinfo-after-b-success.img and write logs are listed in the private backup inventory, but they are not recovered locally and the backup host is unreachable. I cannot yet supply the original 0x30–0x37 byte pair or contemporaneous retry count. The next Fastboot controller completed at 09:06:19 UTC after asserting current-slot=b, slot-successful:b=yes, and slot-unbootable:b=no, with no set-active. A native audit at 09:11:29 UTC reported B success persisted and A unchanged. Those are recovered executed assertions/results, not a recovered raw Fastboot transcript. The notes state this evidence gap explicitly; no new metadata write was attempted to reconstruct it.

Validation for this documentation-only follow-up: tools/pr-check upstream/main and the whitespace check passed. The runtime tools and packages are unchanged by this PR; the historical diff below is evidence, not code being proposed for installation.

Complete local tools diff against 1155a7c (install-phone has no diff)
diff --git a/bin/omamobile b/bin/omamobile
index 9e1a483..37a6b18 100755
--- a/bin/omamobile
+++ b/bin/omamobile
@@ -41,6 +41,7 @@ Usage: [DEVICE=<codename>] omamobile <command>
   flash            Flash the experiment images to slot B and boot them
   push             Write the images into slot B from the running phone over SSH,
                    verify them, and reboot into them (no fastboot)
+                   PUSH_MODULES_POLICY=require-match refuses a modules reinstall
   kernel-shell     Open a shell in the build environment, in the kernel tree
 USAGE
 }
@@ -119,12 +120,13 @@ in_container() {
 }
 
 cmd_capture() {
-  local name="${1:-$DEVICE-$(date +%Y%m%d-%H%M)}" dir
+  local name="${1:-$DEVICE-$(date +%Y%m%d-%H%M)}" dir serial
   dir="$REFERENCE_DIR/$name"
   mkdir -p "$dir"
   if fastboot devices | grep -q .; then
+    serial=$(fastboot_serial) || exit 1
     say "Capturing bootloader variables into $dir"
-    fastboot getvar all > "$dir/fastboot-getvar-all.txt" 2>&1
+    fastboot -s "$serial" getvar all > "$dir/fastboot-getvar-all.txt" 2>&1
   elif adb get-state >/dev/null 2>&1; then
     say "Capturing Android state into $dir"
     local root=""
@@ -149,17 +151,20 @@ cmd_capture() {
 }
 
 cmd_kernel() {
+  local fragment="${KERNEL_CONFIG:-$ROOT/kernel/configs/omamobile.config}"
+  local targets="${KERNEL_TARGETS:-Image.gz dtbs modules}"
   [[ -d "$KERNEL_DIR" ]] || die "no kernel tree at $KERNEL_DIR"
   [[ -n "$CLANG_DIR" ]] || die "no clang found; run 'lineage-ref sync' first or set CLANG_DIR"
+  [[ -f "$fragment" ]] || die "no kernel config fragment at $fragment"
   say "Building $KERNEL_DIR into $KERNEL_OUT"
   mkdir -p "$KERNEL_OUT"
   in_container "
     set -e
     make ARCH=arm64 LLVM=1 O=$KERNEL_OUT defconfig
-    scripts/kconfig/merge_config.sh -m -O $KERNEL_OUT $KERNEL_OUT/.config $ROOT/kernel/configs/omamobile.config
+    scripts/kconfig/merge_config.sh -m -O $KERNEL_OUT $KERNEL_OUT/.config $fragment
     make ARCH=arm64 LLVM=1 O=$KERNEL_OUT olddefconfig
-    make ARCH=arm64 LLVM=1 O=$KERNEL_OUT -j$JOBS Image.gz dtbs modules
-    ls -lh $KERNEL_OUT/arch/arm64/boot/Image.gz
+    make ARCH=arm64 LLVM=1 O=$KERNEL_OUT W=1 -j$JOBS $targets
+    ls -lh $KERNEL_OUT/arch/arm64/boot/Image
     ls $KERNEL_OUT/arch/arm64/boot/dts/exynos/google/
   "
 }
@@ -209,8 +214,10 @@ cmd_initramfs() {
   local dl="$WORK/downloads/alpine" stage="$KERNEL_OUT/initramfs" ver=3.24.2
   load_device
   local init="$ROOT/device/$DEVICE/$DEV_INITRAMFS_INIT"
-  # The init's own storage-guard allowlist must be the descriptor's
-  [[ "$(sed -n 's/^UFS_RW="\(.*\)"$/\1/p' "$init")" == "$DEV_UFS_RW" ]] \
+  # A device may use a generated allowlist; keep the legacy exact-match check.
+  local init_allowlist
+  init_allowlist=$(sed -n 's/^UFS_RW="\(.*\)"$/\1/p' "$init")
+  [[ "$init_allowlist" == '@UFS_RW@' || "$init_allowlist" == "$DEV_UFS_RW" ]] \
     || die "UFS_RW in $init differs from device/$DEVICE/device.conf"
   local tarball="alpine-minirootfs-$ver-aarch64.tar.gz"
   local sum=9bf70a7f18ea44094cbb5f70c58f9af129c8214745743db0e68e5502cc2ce773
@@ -234,6 +241,13 @@ cmd_initramfs() {
     [[ -x $stage/$f ]] || die "the initramfs has no /$f"
   done
   install -m 755 "$init" "$stage/init"
+  if [[ "$init_allowlist" == '@UFS_RW@' ]]; then
+    awk -v allow="$DEV_UFS_RW" '
+      $0 == "UFS_RW=\"@UFS_RW@\"" { print "UFS_RW=\"" allow "\""; next }
+      { print }
+    ' "$init" > "$stage/init"
+    chmod 755 "$stage/init"
+  fi
 
   # Boot logo, drawn by init straight onto the framebuffer until Plymouth
   # takes over: Omarchy's Plymouth screen (logo centred on #1a1b26, at scale
@@ -262,12 +276,25 @@ cmd_initramfs() {
   "$CLANG_DIR/bin/clang" --target=aarch64-linux-musl --sysroot="$sysroot" -fuse-ld=lld -static -nostdlib -O2 \
     -o "$stage/usr/local/bin/reboot-to" "$sysroot/usr/lib/crt1.o" "$sysroot/usr/lib/crti.o" \
     "$ROOT/packages/$DEV_PACKAGE/src/reboot-to.c" -L"$sysroot/usr/lib" -lc "$builtins" "$sysroot/usr/lib/crtn.o"
+  # BusyBox insmod can retry failed initialization through a second syscall.
+  "$CLANG_DIR/bin/clang" --target=aarch64-linux-musl --sysroot="$sysroot" -fuse-ld=lld -static -nostdlib -O2 \
+    -o "$stage/usr/local/bin/module-load-once" "$sysroot/usr/lib/crt1.o" "$sysroot/usr/lib/crti.o" \
+    "$ROOT/tools/module-load-once.c" -L"$sysroot/usr/lib" -lc "$builtins" "$sysroot/usr/lib/crtn.o"
   # A/B slot metadata in devinfo (Apache-2.0 port of Google's boot HAL). Not called by init yet.
   "$CLANG_DIR/bin/clang" --target=aarch64-linux-musl --sysroot="$sysroot" -fuse-ld=lld -static -nostdlib -O2 \
     -o "$stage/usr/local/bin/omamobile-bootctl" "$sysroot/usr/lib/crt1.o" "$sysroot/usr/lib/crti.o" \
     "$ROOT/packages/$DEV_PACKAGE/src/omamobile-bootctl.c" -L"$sysroot/usr/lib" -lc "$builtins" "$sysroot/usr/lib/crtn.o"
+  # Dropbear's non-login shell may search only /usr/bin:/bin.
+  ln -s ../local/bin/reboot-to "$stage/usr/bin/reboot-to"
+  ln -s ../local/bin/module-load-once "$stage/usr/bin/module-load-once"
+  ln -s ../local/bin/omamobile-bootctl "$stage/usr/bin/omamobile-bootctl"
   mkdir -p "$stage/root/.ssh"
-  cat "$HOME"/.ssh/*.pub > "$stage/root/.ssh/authorized_keys"
+  if [[ -n ${INITRAMFS_AUTHORIZED_KEYS:-} ]]; then
+    [[ -s $INITRAMFS_AUTHORIZED_KEYS ]] || die "missing SSH public key file: $INITRAMFS_AUTHORIZED_KEYS"
+    cat "$INITRAMFS_AUTHORIZED_KEYS" > "$stage/root/.ssh/authorized_keys"
+  else
+    cat "$HOME"/.ssh/*.pub > "$stage/root/.ssh/authorized_keys"
+  fi
   chmod 700 "$stage/root/.ssh" && chmod 600 "$stage/root/.ssh/authorized_keys"
 
   (cd "$stage" && find . | cpio -o -H newc --owner=0:0 --quiet) | lz4 -l -9 -q > "$KERNEL_OUT/initramfs.cpio.lz4"
@@ -337,15 +364,15 @@ cmd_images() {
   say "Boot images (header v4)"
   : | cpio -o -H newc --quiet > "$img/empty.cpio"
   local cmdline="$DEV_VENDOR_CMDLINE"
-  # The generic ramdisk (empty) and the OS version go in init_boot where the
-  # device has one, and in boot otherwise.
+  # The generic ramdisk goes in init_boot where present, otherwise in boot.
+  # The GKI boot header's version field stays zero; AVB carries version data.
   if has_image init_boot; then
     python3 "$tools/mkbootimg.py" --header_version 4 --kernel "$img/Image.lz4" --cmdline '' --output "$img/boot.img"
     python3 "$tools/mkbootimg.py" --header_version 4 --ramdisk "$img/empty.cpio" \
       --os_version "$DEV_OS_VERSION" --os_patch_level "$DEV_OS_PATCH_LEVEL" --output "$img/init_boot.img"
   else
     python3 "$tools/mkbootimg.py" --header_version 4 --kernel "$img/Image.lz4" --ramdisk "$img/empty.cpio" --cmdline '' \
-      --os_version "$DEV_OS_VERSION" --os_patch_level "$DEV_OS_PATCH_LEVEL" --output "$img/boot.img"
+      --output "$img/boot.img"
   fi
   # The DTB and our initramfs go in vendor_kernel_boot where the device has
   # one, leaving vendor_boot empty, and in vendor_boot otherwise.
@@ -395,6 +422,23 @@ cmd_kexec() {
 
 in_fastboot() { timeout 3 fastboot devices 2>/dev/null | grep -q .; }
 
+# Pin every Fastboot check/write to the same enumerated phone.
+# Listing devices is the only operation that intentionally has no -s.
+fastboot_serial() {
+  local serials n
+  serials=$(timeout 3 fastboot devices 2>/dev/null | awk '$2 == "fastboot" {print $1}') \
+    || die "could not enumerate phones in fastboot"
+  n=$(grep -c . <<<"$serials" || true)
+  if [[ -n ${FASTBOOT_SERIAL:-} ]]; then
+    n=$(grep -cxF -- "$FASTBOOT_SERIAL" <<<"$serials" || true)
+    ((n == 1)) || die "FASTBOOT_SERIAL must identify exactly one phone in fastboot"
+    printf '%s\n' "$FASTBOOT_SERIAL"
+    return
+  fi
+  ((n == 1)) || die "$n phones in fastboot: unplug the others, or set FASTBOOT_SERIAL to the one to flash"
+  printf '%s\n' "$serials"
+}
+
 cmd_fastboot() {
   local ssh="$ROOT/tools/phone-ssh" i
   in_fastboot && { say "Already in fastboot"; return 0; }
@@ -427,13 +471,15 @@ cmd_flash() {
   load_device
   images_for_device "$img"
   in_fastboot || die "phone not in fastboot (try: omamobile fastboot)"
+  local serial
+  serial=$(fastboot_serial) || exit 1
   for p in $DEV_BOOT_IMAGES; do
     slot_b_allowed "$p" || die "${p}_b is not on device/$DEVICE's storage-guard allowlist; refusing"
     [[ -f "$img/$p.img" ]] || die "missing $img/$p.img (run: omamobile images)"
-    timeout 60 fastboot flash "${p}_b" "$img/$p.img" 2>&1 | grep -E 'OKAY|FAIL' | tail -1 | sed "s/^/$p: /"
+    timeout 60 fastboot -s "$serial" flash "${p}_b" "$img/$p.img" 2>&1 | grep -E 'OKAY|FAIL' | tail -1 | sed "s/^/$p: /"
   done
-  timeout 20 fastboot --set-active=b >/dev/null 2>&1
-  timeout 20 fastboot reboot >/dev/null 2>&1
+  timeout 20 fastboot -s "$serial" --set-active=b >/dev/null 2>&1
+  timeout 20 fastboot -s "$serial" reboot >/dev/null 2>&1
   say "Booting slot B; waiting for SSH"
   local i
   for i in $(seq 1 45); do
@@ -444,16 +490,26 @@ cmd_flash() {
 }
 
 cmd_push() {
-  local img="$KERNEL_OUT/images" ssh="$ROOT/tools/phone-ssh" p dev want got rel
+  local img="$KERNEL_OUT/images" ssh="$ROOT/tools/phone-ssh" p dev want got rel init_system
+  local modules_policy="${PUSH_MODULES_POLICY:-auto}"
+  case "$modules_policy" in
+    auto|require-match) ;;
+    *) die "PUSH_MODULES_POLICY must be auto or require-match" ;;
+  esac
   load_device
   images_for_device "$img"
   "$ssh" true || die "phone not reachable at 172.16.42.1"
+  init_system=$("$ssh" 'cat /proc/1/comm')
   # A kernel whose modules aren't installed boots without its drivers (the GPU
   # among them): install them first. A config change can keep the release
   # name, so compare the tarball's checksum, not just the directory.
   rel=$(kernel_release)
   want=$(sha256sum < "$KERNEL_OUT/modules-$rel.tar.zst" | cut -d' ' -f1)
-  [[ "$("$ssh" "cat /usr/lib/modules/$rel/.omamobile-tarball 2>/dev/null")" == "$want" ]] || cmd_modules_push
+  if [[ "$("$ssh" "cat /usr/lib/modules/$rel/.omamobile-tarball 2>/dev/null")" != "$want" ]]; then
+    [[ $modules_policy != require-match ]] \
+      || die "module archive mismatch: require-match preserves installed modules; no boot images written"
+    cmd_modules_push
+  fi
   for p in $DEV_BOOT_IMAGES; do
     slot_b_allowed "$p" || die "${p}_b is not on device/$DEVICE's storage-guard allowlist; refusing"
     [[ -f "$img/$p.img" ]] || die "missing $img/$p.img (run: omamobile images)"
@@ -463,12 +519,18 @@ cmd_push() {
     [[ "$("$ssh" "cat /sys/block/sda/${dev#/dev/}/ro")" == 0 ]] || die "$dev (${p}_b) is read-only; refusing"
     "$ssh" "dd of=$dev bs=1M conv=fsync 2>/dev/null" < "$img/$p.img"
     want=$(sha256sum < "$img/$p.img" | cut -d' ' -f1)
-    got=$("$ssh" "head -c $(stat -c %s "$img/$p.img") $dev | sha256sum" | cut -d' ' -f1)
+    got=$("$ssh" "head -c $(stat -Lc %s "$img/$p.img") $dev | sha256sum" | cut -d' ' -f1)
     [[ "$want" == "$got" ]] || die "verify failed for ${p}_b ($dev)"
     say "${p}_b ($dev) written and verified"
   done
   say "Rebooting into the new images"
-  "$ssh" 'sync; (sleep 1; reboot-to omamobile) >/dev/null 2>&1 &' || true
+  if [[ $init_system == systemd ]]; then
+    # Native services own shutdown and their persistent interruption markers.
+    # A direct restart syscall bypasses ExecStop and leaves those markers set.
+    "$ssh" 'sync; systemctl reboot --no-block' || true
+  else
+    "$ssh" 'sync; (sleep 1; reboot-to omamobile) >/dev/null 2>&1 &' || true
+  fi
   sleep 15
   local i
   for i in $(seq 1 45); do

Review follow-up (2026-10-31)

The new NOTES section records the exact kernel/mobile bases, detached native kernel worktree and unpublished source changes, local package snapshots, shell backport, installation method and recovery limits. Every observed matrix cell now links to "unpublished kernel/packages".

The keyring concern was confirmed: the recovered local image's omamobile-keyring 20261013.oriole1-1 contains the upstream public key plus the local package-signing public key, and trusts the latter. Both keyring files match the package mtree; the additional key verifies the archived local device package. This was a packaged trust change, not merely a build-container key. The exact rebuilt recipe has not been recovered; the payload audit does not claim to reconstruct its build command. The note explicitly identifies it as outside contributor-build policy and describes the existing --local workflow with an unchanged upstream keyring as the correction for the next image. No modified keyring, signing material, image or package is included in this PR, and no phone trust was changed during this audit.

Historical records show the matching stock build booted on B then A with the expected bootloader before Linux installation. Per-slot bootloader-version queries were unsupported; this is not a fresh partition-version audit. The notes also disclose that the B verification and B-success writes were done without project approval, the historical record of preserved A images/slot record, and Linux userdata's separate Android restoration requirement.

After merging current main: tools/pr-check upstream/main and tests/all pass. The package sources are unchanged from the same main revision whose tests/packages passed during the companion #195 follow-up. Existing local environment skips are recorded below. This revision is an offline artifact/documentation audit, not a new hardware qualification.

The Pixel 6 notes and device matrix still say Linux has not booted. A
read-only smoke check on a native test image now confirms the boot, storage
guard, USB and session. Update those entries with the observed hardware
states, keeping the unpublished integration stack distinct from public
images.

Document the source-confirmed S2MPG10 alarm-status register discrepancy
that a separate kernel change addresses. The test phone already contains
the correction; the smoke check does not test a pending alarm or wake.

Checks: tools/pr-check upstream/main and tests/all passed. The suite
skipped Docker provisioning, AOSP image-layout checks, overlay compilation
and root-image smoke because their prerequisites were unavailable. The 45
overlays and the full image-layout test, including AVB test footers, passed
separately with a recovered dtc and Android 16 AOSP tools. The tests use
fake devices and images; no package was changed.

The smoke check recorded no failed units, oops, BUG or WARN and 48 dmesg
error lines. Sound, sensors and cameras were skipped.

Pixel 6 smoke report
{
  "device": "oriole",
  "repo": "review-readonly",
  "kernel": "7.3.0-rc6-oriole-hsi1-late-dev0-g4bd35d34abb9-dirty",
  "packages": [
    "hyprland 0.56.2-7",
    "libcamera 0.7.2-4",
    "mesa 1:26.2.4-1",
    "omacam 0.1.9-1",
    "omamobile-apps 0.2.0-1",
    "omamobile-backlight-oriole 0.1.0-8",
    "omamobile-battery-counters-oriole 0.2.0-1",
    "omamobile-bluetooth-oriole 0.1.2-2",
    "omamobile-bluez 5.87.omamobile1-1",
    "omamobile-charger-cache-oriole 0.1.0-8",
    "omamobile-device-oriole 0.1.9-2",
    "omamobile-firstboot 0.1.1-1",
    "omamobile-gpu-oriole 0.1.0-3",
    "omamobile-keyring 20261013.oriole1-1",
    "omamobile-mirror 0.2.1-1",
    "omamobile-nfc-oriole 0.1.0-9",
    "omamobile-omarchy 0.1.6-1",
    "omamobile-owl-widgets 0.1.1-1",
    "omamobile-shell 0.2.12-1",
    "omamobile-torch-oriole 0.1.1-2",
    "omamobile-wifi-oriole 0.1.8-3",
    "omarchy 4.0.4-2",
    "quickshell 0.3.2.r7.g2b4501a-1"
  ],
  "checks": [
    {
      "name": "boot",
      "status": "ok",
      "required": true,
      "detail": "kernel 7.3.0-rc6-oriole-hsi1-late-dev0-g4bd35d34abb9-dirty, slot b, up 851s"
    },
    {
      "name": "storage guard",
      "status": "ok",
      "required": true,
      "detail": "49 partitions; only userdata and slot B's boot images writable"
    },
    {
      "name": "root",
      "status": "ok",
      "required": true,
      "detail": "ext4 on userdata, 97G free"
    },
    {
      "name": "kernel health",
      "status": "ok",
      "required": true,
      "detail": "no oops or BUG; 0 WARN, 48 error lines in dmesg"
    },
    {
      "name": "systemd units",
      "status": "ok",
      "required": false,
      "detail": "none failed"
    },
    {
      "name": "USB gadget",
      "status": "ok",
      "required": true,
      "detail": "bound to 11110000.usb"
    },
    {
      "name": "display",
      "status": "ok",
      "required": false,
      "detail": "card0-Unknown-1 connected, 1080x2400"
    },
    {
      "name": "GPU",
      "status": "ok",
      "required": false,
      "detail": "render node on panfrost"
    },
    {
      "name": "session",
      "status": "ok",
      "required": true,
      "detail": "Hyprland and the Omarchy shell running"
    },
    {
      "name": "lock",
      "status": "ok",
      "required": false,
      "detail": "PAM file present (the shell didn't answer lock status)"
    },
    {
      "name": "touch",
      "status": "ok",
      "required": false,
      "detail": "ST FTM5 touchscreen (oriole staged)"
    },
    {
      "name": "buttons",
      "status": "ok",
      "required": false,
      "detail": "gpio-keys"
    },
    {
      "name": "Wi-Fi",
      "status": "ok",
      "required": false,
      "detail": "wlP1p1s0 on brcmfmac, link up"
    },
    {
      "name": "Bluetooth",
      "status": "ok",
      "required": false,
      "detail": "hci0, powered true"
    },
    {
      "name": "battery",
      "status": "ok",
      "required": false,
      "detail": "78% Charging, 34 °C"
    },
    {
      "name": "sound cards",
      "status": "skip",
      "required": false,
      "detail": "none"
    },
    {
      "name": "sensors",
      "status": "skip",
      "required": false,
      "detail": "no sensor proxy"
    },
    {
      "name": "cameras",
      "status": "skip",
      "required": false,
      "detail": "no libcamera tools"
    },
    {
      "name": "thermal",
      "status": "ok",
      "required": false,
      "detail": "hottest zone 34 °C"
    }
  ]
}

The Pixel 6 notes still say Linux has not booted, although a native
integration image passed the read-only phone smoke check. Record that
result and its limits, and bring the matrix into line with the hardware
states the check observed.

The S2MPG10 alarm-status register differs from S2MPS14. Record the source
evidence and distinguish the published kernel from the test phone's
corrected integration kernel. No live pending-alarm result is claimed.

Co-Authored-By: Codex <noreply@openai.com>
@derekross

Copy link
Copy Markdown
Owner

Congratulations on the first native boot! Before this lands: please record where the kernel and the oriole packages come from (branch, commit, and how slot B was written) in NOTES, and mark the matrix cells "unpublished kernel/packages" with a link to the notes, as raven's row does. The smoke report lists omamobile-device-oriole 0.1.9, *-oriole packages, shell 0.2.12 and a rebuilt omamobile-keyring 20261013.oriole1, none of which are in our repositories. Please also tell me how that keyring was built (a local key must never go into the keyring package) and whether slot A still has stock with matching bootloaders (CONTRIBUTING §3.1).

Answer the review's source, keyring and recovery questions without
turning the private integration stack into a public support claim.
Record the recovered kernel and package bases, local shell backport,
keyring payload audit and limits of the historical slot evidence.
Mark the observed matrix cells as unpublished kernel/packages.

The recovered root image's packaged keyring and trusted list contain
a local signing key, verified against its mtree and device-package
signature. Document the existing contributor workflow as the next
image's correction; no phone trust or partition state changes here.

Assisted-by: OpenAI GPT-6 Astra
The image and mtree establish that the local key was packaged. The rebuilt recipe remains in an unavailable backup; do not imply its exact command was reconstructed.

Assisted-by: OpenAI GPT-6 Astra
@derekross

Copy link
Copy Markdown
Owner

Thank you for the candour; this is exactly what the notes are for. Before I merge:

  1. Was your local install-phone/push changed from main's? If so, please post a diff.
  2. What made devinfo writable for the 0x35 write? Do you have the before/after bytes 0x30-0x37, next to fastboot getvar slot-successful/unbootable/retry-count?
  3. Please change "owner-authorized" to say these writes were done without project approval.

No further devinfo or vbmeta writes without my go, please.

Record the recovered local deployment-tool differences and the explicit block-device read-only changes used for B success. State that the metadata writes lacked project approval and further writes await the maintainer's approval.

Distinguish the recovered header-write and executed slot assertions from the original byte pair and retry count, which remain unavailable in the offline backup.

Assisted-by: OpenAI GPT-6 Astra
derekross added a commit that referenced this pull request Oct 31, 2026
… rounds

L-348 to L-361: shared drivers judged against every int-line user (#150),
board pins against every stock user (#151), ownership flags under the
consumer's lock (#152), /chosen children and Google's panel constants
(#154), same-day notes PRs on one device (#208/#209), merge simulations for
DEVICES rows, ours included (#165), stacked PRs that merged main (#115/#117),
titles behind renumbered versions, a contributor merge that dropped the PR
(#99), smoke package lists as provenance and "owner-authorized" writes
(#200), promised fixes and named tags checked on the remote (#153/#157/#158),
and file swaps into old packages (#193). C-23 (usbc-deadman's device scope,
kernel #107) and C-24 (BNEP in a lynx-only fragment, #98).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@derekross

Copy link
Copy Markdown
Owner

Thank you for answering everything so plainly, including the parts that can't be recovered. Going through my questions:

  • Your local install-phone and push: answered. install-phone matches the base, and the push changes (the module policy, the readback length, the systemd reboot) don't touch what gets written or where.
  • What made devinfo writable: answered. It was a separate script doing blockdev --setrw on the boot LUN and the partition, with an undo timer, and the notes record it. The before/after bytes at 0x30-0x37 aren't available, and the notes say so rather than reconstructing them, which is what I'd want. If the backup host comes back, please add those eight bytes and the slot variables in a small follow-up.
  • "Owner-authorized": gone. Both writes are recorded as done without project approval.
  • From my first comment: the provenance section, the "unpublished kernel/packages" links on every cell, the keyring answer and the slot A record are all there. Thanks for saying straight out that the local key went into the keyring package. For the next image, the --local flow keeps it out.

One thing I should have said last time. Slot B is marked on your phone, and we're not marking it again until #252 is settled. So please use push for new kernels, not flash. flash runs fastboot --set-active=b, which on the 9a clears B's mark, and from then on every boot spends one of B's three tries. When they run out, the phone falls back to slot A, and on the 9a and on one 8 Pro the Android there wiped userdata on its first boot. If you ever do need fastboot, check fastboot getvar slot-retry-count:b before each reboot. Note too that with B marked, the bootloader won't fall back on its own after a broken pushed image, so the way back is fastboot by the buttons.

Two small edits I'll make when I merge, unless you'd rather do them yourself:

  1. The boot cell in the Pixel 6 row gets a short qualifier, as the 9a's row has for its mark: "slot B marked successful and vbmeta_b verification turned off, both without project approval; see notes". Anyone reading the row should know that this boot relied on a non-stock vbmeta_b, and that a boot with a stock one hasn't been shown yet.
  2. The RTC bullet links the fix, rtc: s5m: fix S2MPG10 alarm pending readback linux-omarchy-mobile#149. I checked the finding: on the published kernel the S2MPG10 case reads S2MPS14_REG_ST2, which is offset 0x08, INT3M in the S2MPG10's map, while STATUS2 is 0x0d.

I'll write the squash message myself, since the branch has a merge from main in it. Thanks again. This is the kind of record that lets the next Pixel 6 owner start from solid ground.

@derekross derekross closed this in b068c7b Nov 1, 2026
gitColinX added a commit to gitColinX/omarchy-mobile that referenced this pull request Nov 13, 2026
…s files

`-d husky` staged the four Wi-Fi files only, with Bluetooth and the rest left "for drivers". Kernel derekross#199 (Bluetooth) and derekross#200 (haptics) are on `omamobile` now, so an installed Pixel 8 Pro had the drivers and not the files: cs40l26-vibra stopped at "No cs40l26.wmfw" and btbcm at "firmware Patch file not found" (a scan found nothing). husky's map gains stock's `BCM.hcd` as `brcm/BCM.google,zuma-husky.hcd` and the five CS40L26 files the 7 Pro and the 9a take. This `BCM.hcd` names the module and board ("G5SN_V17_HK3 REL ..."), not the chip, so husky's Bluetooth check looks for that; the haptics checks run for husky too. The test's husky fixture has the new files and a case refuses another chip's patch.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
derekross pushed a commit that referenced this pull request Nov 13, 2026
…s files (#722)

`-d husky` staged Wi-Fi only; the Bluetooth and haptics drivers are on `omamobile` now (kernel #199, #200), so an installed Pixel 8 Pro had the drivers without their files. husky's map gains stock's `BCM.hcd` as `brcm/BCM.google,zuma-husky.hcd` and the five CS40L26 files under the 7 Pro's names; its Bluetooth check looks for the module and board string (`G5SN_V<n>_HK3`), since this patch names no chip, and the haptics checks run for husky too. Tested on the author's own factory image (CP3A.261005.005, ten files) and phone. `Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>` (the commit's trailer, kept).
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
… rounds

L-348 to L-361: shared drivers judged against every int-line user (derekross#150),
board pins against every stock user (derekross#151), ownership flags under the
consumer's lock (derekross#152), /chosen children and Google's panel constants
(derekross#154), same-day notes PRs on one device (derekross#208/derekross#209), merge simulations for
DEVICES rows, ours included (derekross#165), stacked PRs that merged main (derekross#115/derekross#117),
titles behind renumbered versions, a contributor merge that dropped the PR
(derekross#99), smoke package lists as provenance and "owner-authorized" writes
(derekross#200), promised fixes and named tags checked on the remote (derekross#153/derekross#157/derekross#158),
and file swaps into old packages (derekross#193). C-23 (usbc-deadman's device scope,
kernel derekross#107) and C-24 (BNEP in a lynx-only fragment, derekross#98).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
…nd the RTC alarm finding (derekross#200)

Pixel 6 notes: the native boot on one phone, its build provenance and recovery limits, and the RTC alarm finding (the S2MPG10 alarm callback reads INT3M at 0x08 instead of STATUS2 at 0x0d, fixed by linux-omarchy-mobile#149). The DEVICES row records that slot B was marked successful and vbmeta_b verification turned off without project approval (Gr-linkk).

Closes derekross#200

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants