feat(rootfs): native armhf on aarch64 host via runtime-disable of qemu-arm

[iav]

Summary

Continues #9284 (arm64-compat-vdso extension + custom_kernel_make_params hook).
That PR enabled COMPAT_VDSO in arm64 kernels; this PR uses the resulting
32-bit-capable kernels to run 32-bit ARM ELF directly instead of through
qemu-arm emulation in chroot work — typically ~12× faster than qemu-arm-static
on Cortex-A53/A55/A72/A73/A76.

How

Disable qemu-arm in /proc/sys/fs/binfmt_misc/ for the duration of the
build; kernel binfmt_elf then handles 32-bit ARM ELFs natively via
CONFIG_COMPAT. Activation is delayed until AFTER mmdebstrap, whose
cross-arch path needs a working qemu-arm registration to populate the
chroot before libc/ld-linux-armhf.so.3 exist.

Killswitch: NATIVE_ARMHF_ON_ARM64=no (synonyms never, disabled).

Concurrency — kernel flock on the binfmt entry

Concurrent armbian builds on the same host kernel are coordinated purely
via kernel BSD flock(2) on /proc/sys/fs/binfmt_misc/qemu-arm itself.
No userspace state, no per-builder lockfiles, no /run/lock directory.

• Each builder holds LOCK_SH on the binfmt entry via a long-lived fd.
  Kernel BSD-flock counter is the refcount; kernel releases the fd on
  process exit (crash-safe).
• First-arrival idempotently echo 0 to disable; subsequent arrivals
  observe 0 and proceed without writing.
• On exit, release LOCK_SH; last-out detects via LOCK_EX-LOCK_NB on
  a fresh fd, succeeds iff zero other LOCK_SH holders. Last-out
  re-enables qemu-arm.

Trade-offs

1. Prior qemu-arm state is not recorded across builds. Last-out
   unconditionally re-enables. An admin's pre-existing disabled
   policy is not preserved across the build window.
2. No defense against an external agent toggling qemu-arm mid-build.
   Don't, or use NATIVE_ARMHF_ON_ARM64=no on builds that should not
   interfere.
3. Cleanup ordering invariant — BSD flock is per-OFD; a forked subshell
   inheriting our SH-fd shares the same lock entry. We rely on
   add_cleanup_handler running registrations in order and the umount
   teardown handlers (registered first) killing the docker container
   before our restore handler runs. Documented in the trap handler's
   docstring; if a future refactor breaks ordering, POSIX F_SETLK or
   explicit descendant-kill are the documented escape hatches.

The earlier revision of this work coordinated via a userspace
owner-flock directory (~250 lines, depended on /run/lock being
host-bind-mounted into build containers — which Armbian docker does
not do). Replaced with the kernel-flock approach above; available in
PR history if needed.

Empirical

./compile.sh build BOARD=helios4 BRANCH=edge BUILD_MINIMAL=yes RELEASE=noble
on droid (RK3328) and m1 (RK3568, Cortex-A55):

Scenario	Real time	Speedup
baseline, rootfs cache-miss	60:35	1×
this PR, rootfs cache-miss	19:27	3.12×
this PR, rootfs cache-hit	6:38	~9× over expected baseline cache-hit

mmdebstrap itself (~10 min of every cache-miss build) cannot be
accelerated — speedup applies to subsequent chroot operations.

Test plan

Phase 1 — feature on iron (single builder):

• helios4 cache-miss + cache-hit, full image
• cleanup re-enables qemu-arm on exit
• DOCKER_PRIVILEGED=no (CAP_SYS_ADMIN unconditional)
• killswitch (NATIVE_ARMHF_ON_ARM64=no/never/disabled) — silent return
• x86_64 host regression — silent return
• aarch64 native (no-container) PREFER_DOCKER=no smoke

Phase 2 — concurrency, all on Hetzner CAX21 (kernel 6.8.0-90, binfmt_misc fs):

• flock primitives on /proc/sys/fs/binfmt_misc/qemu-arm work as expected
• single-tree build: setup + 5 phase markers + last-out restore
• two independent worktrees concurrent: both enter native path, kernel
  SH-counter holds qemu-arm disabled, first-finished suppresses restore,
  last-finished restores
• NATIVE_ARMHF_ON_ARM64=no opt-out with concurrent native builder —
  refused-fast guard rejects (avoids clobbering)
• SIGINT mid-chroot — cleanup ordering invariant holds; last-out fires,
  qemu-arm restored, no stale state

Summary by CodeRabbit

• New Features

  • Allow native execution of 32-bit ARM binaries on compatible 64-bit hosts to speed and simplify ARM image builds.
  • Skip copying emulation binaries into chroots when native native-ARM handling is active.

• Bug Fixes

  • Improve coordination and recovery around host binary-format handling to avoid interference between concurrent builds and ignore non-fatal setup failures.

[coderabbitai]

📝 Walkthrough

Walkthrough

Adds coordinated host binfmt_misc handling to prefer kernel-native 32-bit ARM ELF on aarch64 hosts, conditional qemu deploy/undeploy behavior, and invokes the setup during rootfs creation and image build; setup failures are ignored.

Changes

Native armhf via binfmt_elf Mode

Layer / File(s)	Summary
Core Observation lib/functions/rootfs/qemu-static.sh	Adds _native_armhf_observe_qemu_arm_state() to report qemu-arm binfmt_misc presence and enabled/disabled state.
Core Setup & Coordination lib/functions/rootfs/qemu-static.sh	Adds _native_armhf_setup_binfmt_elf() that acquires a control flock, may disable qemu-arm, runs arch-test armhf, and exports ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF=yes.
Cleanup & Owner Restore lib/functions/rootfs/qemu-static.sh	Adds trap_handler_native_armhf_restore_qemu_arm() that restores qemu-arm only when the last owner exits; otherwise it releases ownership.
Killswitch Release Handler lib/functions/rootfs/qemu-static.sh	Adds trap_handler_native_armhf_release_emul_lock() that releases the shared emulation lock fd without restoring binfmt state.
qemu Deploy/Undeploy Wiring lib/functions/rootfs/qemu-static.sh	deploy_qemu_binary_to_chroot() skips copying qemu when ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF=yes; undeploy_qemu_binary_from_chroot() treats missing chroot qemu as a no-op only in native mode.
Prepare Guards & Verification lib/functions/rootfs/qemu-static.sh	Skip enabling qemu-arm for non-armhf targets on aarch64 hosts; attempt native-armhf setup early for aarch64→armhf; add preflight guard refusing to re-enable an observably-disabled registration; change verification to arch-test armhf and tweak heredoc formatting.
Rootfs Creation Integration lib/functions/rootfs/rootfs-create.sh	Calls `_native_armhf_setup_binfmt_elf
Image Build Integration lib/functions/main/rootfs-image.sh	Calls `_native_armhf_setup_binfmt_elf

Sequence Diagram(s)

sequenceDiagram
    participant Build as Build Flow
    participant Setup as _native_armhf_setup_binfmt_elf()
    participant Binfmt as /proc/sys/fs/binfmt_misc
    participant Trap as Cleanup Handler
    participant Deploy as deploy_qemu_binary_to_chroot()

    Build->>Setup: invoke (post-rootfs extraction)
    Setup->>Binfmt: read qemu-arm entry/state
    alt qemu-arm enabled
        Setup->>Binfmt: write disable qemu-arm
        Setup->>Trap: register restore handler
        Setup-->>Build: export ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF=yes
    else qemu-arm absent_or_disabled
        Setup-->>Build: export ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF=yes
    end
    Build->>Deploy: proceed to deploy step
    Deploy->>Deploy: check ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF
    alt native mode set
        Deploy-->>Build: early return (skip qemu copy)
    else
        Deploy->>Build: copy qemu-arm-static into chroot
    end
    Build-->>Trap: on exit/signal
    Trap->>Binfmt: write restore qemu-arm state

Loading

Estimated Code Review Effort

🎯 4 (Complex) | ⏱️ ~45 minutes

Poem

🐰 I nibbled at binfmt's quiet seam,
Turned qemu off so armhf could dream,
Flag raised, trap set neat,
Chroots wake on native feet,
Builds hop forward, light and trim.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 70.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and specifically describes the main change: enabling native armhf execution on aarch64 hosts by disabling qemu-arm binfmt_misc, which is the primary feature introduced across all modified files.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/native-armhf-on-arm64

Tip

💬 Introducing Slack Agent: The best way for teams to turn conversations into code.

Slack Agent is built on CodeRabbit's deep understanding of your code, so your team can collaborate across the entire SDLC without losing context.

• Generate code and open pull requests
• Plan features and break down work
• Investigate incidents and troubleshoot customer tickets together
• Automate recurring tasks and respond to alerts with triggers
• Summarize progress and report instantly

Built for teams:

• Shared memory across your entire org—no repeating context
• Per-thread sandboxes to safely plan and execute work
• Governance built-in—scoped access, auditability, and budget controls

One agent for your entire SDLC. Right inside Slack.

👉 Get started

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@lib/functions/rootfs/qemu-static.sh`:
- Around line 20-25: The early return in the qemu-static check (when
ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF == "yes") prevents the matching undeploy
from running later, leaking host qemu into cached rootfs; instead of returning
from the function, set a clear marker variable (e.g.,
ARMBIAN_QEMU_DEPLOY_SKIPPED=yes or ARMBIAN_QEMU_DEPLOYED=no) after calling
display_alert and allow the function to exit normally so the caller
(rootfs-create.sh) can always run the undeploy path and check that marker;
update the undeploy logic to look for that same marker
(ARMBIAN_QEMU_DEPLOY_SKIPPED or ARMBIAN_QEMU_DEPLOYED) so it only skips removal
when deploy truly never ran.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: eba7fdae-710a-42e6-bcb9-0d41993767df

📥 Commits

Reviewing files that changed from the base of the PR and between 3ce7dec and 502287f.

📒 Files selected for processing (4)

• lib/functions/main/rootfs-image.sh
• lib/functions/rootfs/qemu-static.sh
• lib/functions/rootfs/rootfs-create.sh
• lib/library-functions.sh

[coderabbitai]

Actionable comments posted: 1

♻️ Duplicate comments (1)

lib/functions/rootfs/qemu-static.sh (1)

20-25: ⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Deploy/undeploy asymmetry can leak host qemu-arm-static into the rootfs cache.

The flag check is symmetric inside this file, but at the call sites (rootfs-create.sh, rootfs-image.sh per the summary), deploy_qemu_binary_to_chroot runs before _native_armhf_setup_binfmt_elf flips ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF=yes, while undeploy_qemu_binary_from_chroot runs after. In that ordering the binary gets copied into the chroot, but the matching removal is suppressed by the new early-return at Line 56. The host qemu-arm-static then persists into the rootfs cache and into final armhf images on subsequent cache-hit builds.

Track per-(caller, chroot_target) deployment state and only skip undeploy when deploy actually skipped, rather than gating both on the flag alone.

🛠️ Suggested fix

+declare -gA armbian_qemu_deployed_to_chroot=()
+
 function deploy_qemu_binary_to_chroot() {
 	declare chroot_target="${1}" caller="${2}"
@@
 	if [[ "${ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF:-no}" == "yes" ]]; then
 		display_alert "Native armhf via binfmt_elf" "skipping qemu binary deployment during ${caller}" "info"
 		return 0
 	fi
@@
 	display_alert "Deploying qemu-user-static binary to chroot" "${QEMU_BINARY} during ${caller}" "info"
 	run_host_command_logged cp -pv "${src_host}" "${dst_target}"
+	armbian_qemu_deployed_to_chroot["${caller}:${chroot_target}"]=yes
 
 	return 0
 }
@@
 function undeploy_qemu_binary_from_chroot() {
@@
-	if [[ "${ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF:-no}" == "yes" ]]; then
+	if [[ "${ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF:-no}" == "yes" \
+		&& "${armbian_qemu_deployed_to_chroot["${caller}:${chroot_target}"]:-no}" != "yes" ]]; then
 		display_alert "Native armhf via binfmt_elf" "no qemu binary to remove during ${caller}" "debug"
 		return 0
 	fi

Also applies to: 55-59

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@lib/functions/rootfs/qemu-static.sh` around lines 20 - 25, The early-return
based solely on ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF in
deploy_qemu_binary_to_chroot/undeploy_qemu_binary_from_chroot causes asymmetry
across separate invocations; change the logic so that when
deploy_qemu_binary_to_chroot skips copying due to the flag it records that
decision with a per-(caller,chroot_target) marker (e.g., a temp file named with
${caller} and the chroot target inside the rootfs cache or a shared temp dir),
and then have undeploy_qemu_binary_from_chroot check that marker and only skip
removal when the deploy previously skipped for that same (caller,chroot_target);
ensure you still clean up the marker when an actual copy was performed and when
removal runs so state cannot leak into later builds (references: functions
deploy_qemu_binary_to_chroot, undeploy_qemu_binary_from_chroot and variable
ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF).

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@lib/functions/rootfs/qemu-static.sh`:
- Around line 169-188: Move the add_cleanup_handler call so it runs before
writing "0" to /proc/sys/fs/binfmt_misc/qemu-arm (register the trap before the
destructive action) and update trap_handler_native_armhf_restore_qemu_arm to not
early-return based on ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF; instead key the
cleanup on the actual binfmt_misc state (check [[ -e
/proc/sys/fs/binfmt_misc/qemu-arm ]] and attempt echo 1 >
/proc/sys/fs/binfmt_misc/qemu-arm || true), so a partially completed setup or an
interrupt between the disable and ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF assignment
will still re-enable qemu-arm.

---

Duplicate comments:
In `@lib/functions/rootfs/qemu-static.sh`:
- Around line 20-25: The early-return based solely on
ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF in
deploy_qemu_binary_to_chroot/undeploy_qemu_binary_from_chroot causes asymmetry
across separate invocations; change the logic so that when
deploy_qemu_binary_to_chroot skips copying due to the flag it records that
decision with a per-(caller,chroot_target) marker (e.g., a temp file named with
${caller} and the chroot target inside the rootfs cache or a shared temp dir),
and then have undeploy_qemu_binary_from_chroot check that marker and only skip
removal when the deploy previously skipped for that same (caller,chroot_target);
ensure you still clean up the marker when an actual copy was performed and when
removal runs so state cannot leak into later builds (references: functions
deploy_qemu_binary_to_chroot, undeploy_qemu_binary_from_chroot and variable
ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF).

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: e697b537-3deb-4c61-94a1-02ef9c76d299

📥 Commits

Reviewing files that changed from the base of the PR and between 502287f and 6f7903b.

📒 Files selected for processing (3)

• lib/functions/main/rootfs-image.sh
• lib/functions/rootfs/qemu-static.sh
• lib/functions/rootfs/rootfs-create.sh

🚧 Files skipped from review as they are similar to previous changes (2)

• lib/functions/main/rootfs-image.sh
• lib/functions/rootfs/rootfs-create.sh

[coderabbitai]

🧹 Nitpick comments (1)

lib/functions/rootfs/qemu-static.sh (1)

131-134: 💤 Low value

Optional: case-insensitive killswitch matching.

NATIVE_ARMHF_ON_ARM64=NO (or Disabled, False, etc.) would slip past the killswitch and proceed to detection. This is a common ergonomic gotcha when users export env vars from CI configs. Low priority since the function self-aborts on the subsequent guards anyway, but a one-liner makes the contract explicit.

♻️ Optional refactor

-	# Killswitch — accept no/never/disabled as synonyms; bail before any detection runs.
-	case "${NATIVE_ARMHF_ON_ARM64:-auto}" in
-		no | never | disabled) return 1 ;;
-	esac
+	# Killswitch — accept no/never/disabled (any case) as synonyms; bail before any detection runs.
+	case "${NATIVE_ARMHF_ON_ARM64:-auto,,}" in
+		no | never | disabled) return 1 ;;
+	esac

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@lib/functions/rootfs/qemu-static.sh` around lines 131 - 134, The killswitch
case for NATIVE_ARMHF_ON_ARM64 is currently case-sensitive and will miss values
like "NO" or "Disabled"; update the case statement around the
NATIVE_ARMHF_ON_ARM64 check to perform case-insensitive matching (e.g., convert
the value to lowercase with parameter expansion like
"${NATIVE_ARMHF_ON_ARM64,,}" before the case or expand the patterns to bracketed
forms such as
[Nn][Oo]|[Nn][Ee][Vv][Ee][Rr]|[Dd][Ii][Ss][Aa][Bb][Ll][Ee][Dd]|[Ff][Aa][Ll][Ss][Ee]
to accept common synonyms) so that the return 1 kill path in that branch
reliably triggers for uppercase/mixed-case inputs.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@lib/functions/rootfs/qemu-static.sh`:
- Around line 131-134: The killswitch case for NATIVE_ARMHF_ON_ARM64 is
currently case-sensitive and will miss values like "NO" or "Disabled"; update
the case statement around the NATIVE_ARMHF_ON_ARM64 check to perform
case-insensitive matching (e.g., convert the value to lowercase with parameter
expansion like "${NATIVE_ARMHF_ON_ARM64,,}" before the case or expand the
patterns to bracketed forms such as
[Nn][Oo]|[Nn][Ee][Vv][Ee][Rr]|[Dd][Ii][Ss][Aa][Bb][Ll][Ee][Dd]|[Ff][Aa][Ll][Ss][Ee]
to accept common synonyms) so that the return 1 kill path in that branch
reliably triggers for uppercase/mixed-case inputs.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: ff5855c0-1c00-4b4b-8c4e-dc619c12d9a0

📥 Commits

Reviewing files that changed from the base of the PR and between 6f7903b and d990278.

📒 Files selected for processing (3)

• lib/functions/main/rootfs-image.sh
• lib/functions/rootfs/qemu-static.sh
• lib/functions/rootfs/rootfs-create.sh

🚧 Files skipped from review as they are similar to previous changes (2)

• lib/functions/main/rootfs-image.sh
• lib/functions/rootfs/rootfs-create.sh

[iav]

Re: nitpick on qemu-static.sh:131-134 (case-insensitive killswitch matching) — declined. Project convention for boolean toggles is literal lowercase comparison ([[ "${VAR}" == "yes" ]], [[ "${VAR:-no}" == "yes" ]]). Examples: cli-artifact.sh:56 (ARTIFACT_USE_CACHE), utils-cli.sh:247 (PREFER_DOCKER:-yes), bsp/armbian-bsp-cli-deb.sh:239 (SHOW_DEBUG). No ${var,,} normalization on boolean flags anywhere in lib/functions/. Keeping the killswitch in line with that convention.

[iav]

Who use arm64 builders — try it!

[igorpecovnik]

aarch64 host (CIX). Always fails:

./compile.sh build BOARD=helios4 BRANCH=edge BUILD_MINIMAL=yes KERNEL_CONFIGURE=no RELEASE=noble SHARE=yes

[🐳|🌱] git: Fetch from remote completed, rev-parsing... [ 'mmdebstrap-debian-devel' 'main' 'FETCH_HEAD' ]
[🐳|🌱] mmdebstrap version [ '1.5.7 (2025-04-08)' for /armbian/cache/sources/mmdebstrap-debian-devel/mmdebstrap ]
[🐳|🌱] Installing base system with 5 packages [ Stage 1/1 ]
[🐳|🌱] fetch_distro_keyring(noble) [ cache found, skipping ]
[🐳|🌱] Deploying qemu-user-static binary to chroot [ qemu-arm-static during rootfs ]
[🐳|🔨]   '/usr/bin/qemu-arm-static' -> '/armbian/.tmp/rootfs-09e5dde9-e782-4c83-a704-7d2314fd45c0/usr/bin/qemu-arm-static'
[🐳|🔨]   I: automatically chosen mode: root
[🐳|🔨]   I: armhf cannot be executed natively, but transparently using qemu-user binfmt emulation
[🐳|🔨]   I: automatically chosen format: directory
[🐳|🔨]   I: skipping check/empty as requested
[🐳|🔨]   I: running special hook: copy-in /armbian/cache/keyrings/ubuntu/usr /armbian/cache/keyrings/ubuntu/etc /
[🐳|🔨]   I: running --setup-hook in shell: sh -c 'mkdir -p /armbian/cache/aptcache/noble-armhf/archives "$1"/var/cache/apt/archives/' exec /armbian/.tmp/rootfs-09e5dde9-e782-4c83-a704-7d2314fd45c0
[🐳|🔨]   I: running special hook: sync-in /armbian/cache/aptcache/noble-armhf/archives /var/cache/apt/archives/
[🐳|🔨]   I: running apt-get update...
[🐳|🔨]   I: downloading packages with apt...
[🐳|🔨]   I: extracting archives...
[🐳|🔨]   I: installing essential packages...
[🐳|🔨]   I: installing remaining packages inside the chroot...
[🐳|🔨]   I: running special hook: sync-out /var/cache/apt/archives/ /armbian/cache/aptcache/noble-armhf/archives
[🐳|🔨]   I: cleaning package lists and apt cache...
[🐳|🔨]   I: success in 338.3367 seconds
[🐳|🌱] Native armhf via binfmt_elf [ kernel 6.18.8-current-arm64, aarch64 host with COMPAT_VDSO; qemu-arm disabled, kernel binfmt_elf takes over ]
[🐳|🌱] Cleaning up after mmdebstrap [ mmdebstrap cleanup ]
[🐳|🌱] Diverting [ initctl/start-stop-daemon ]
[🐳|🔨]   chroot: failed to run command '/usr/bin/env': Exec format error

[iav]

@igorpecovnik thanks for the report.

Root cause: arch-test armhf returns 0 on your CIX host because qemu-arm in binfmt_misc interprets its tiny test stub — false positive on hosts whose kernel was built without COMPAT_VDSO. After my code disabled qemu-arm, the kernel tried to execute armhf ELF natively and failed with Exec format error because the host kernel doesn't have the compat-vdso layer.

Fix: arch-test moved to AFTER the binfmt_misc disable, so it reflects what the chroot exec will actually face. On detection failure we restore qemu-arm immediately and fall back to qemu-static emulation transparently.

After this fix lands on your builder, your build will work as before — slowly, via qemu-arm-static, no regression. To get the speedup, build your builder host's kernel with the arm64-compat-vdso extension enabled (extensions/arm64-compat-vdso, PR #9284); then native armhf via binfmt_elf will activate automatically.

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@lib/functions/rootfs/qemu-static.sh`:
- Around line 324-331: The else branch that runs when live_count > 1 must not
assume qemu-arm is already disabled; before registering the join behavior
(trap_handler_native_armhf_restore_qemu_arm, display_alert and setting
ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF), read the actual qemu-arm state (the same
state/sysfs/binfmt path used when writing "0"/"1") and verify it is "0"
(disabled); if it is not "0", fail fast (error out) rather than joining, so the
subsequent arch-test armhf and ARMBIAN_NATIVE_ARMHF_VIA_BINFMT_ELF logic do not
incorrectly skip deploying qemu-arm-static. Ensure you reference the same state
file/variable used by the disable code path and keep the cleanup/restore handler
only for true joiners.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 6b9577ae-7b84-40f0-972d-8e8b8d70ef1c

📥 Commits

Reviewing files that changed from the base of the PR and between df7aaf9 and 3d4c8cd.

📒 Files selected for processing (1)

• lib/functions/rootfs/qemu-static.sh

[coderabbitai]

🧹 Nitpick comments (1)

lib/functions/main/rootfs-image.sh (1)

19-19: ⚡ Quick win

Make the non-fatal fallback explicit instead of silently swallowing setup failure.

At Line 19, || true preserves behavior, but it hides when native armhf activation fails. Please wrap this in an explicit if ! ...; then display_alert ...; fi so operators can see when the build falls back to qemu-static.

Suggested change

-	_native_armhf_setup_binfmt_elf || true
+	if ! _native_armhf_setup_binfmt_elf; then
+		display_alert "Native armhf via binfmt_elf unavailable" "Falling back to qemu-static path" "warn"
+	fi

Based on learnings: In the Armbian build framework, scripts run with set -e, so masked failures should be explicitly handled and not silently ignored.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@lib/functions/main/rootfs-image.sh` at line 19, The call to
_native_armhf_setup_binfmt_elf currently uses "|| true" which silently swallows
failures; replace that pattern with an explicit conditional: invoke
_native_armhf_setup_binfmt_elf and if it returns non-zero run a visible fallback
notification (e.g. if ! _native_armhf_setup_binfmt_elf; then display_alert
"native armhf binfmt activation failed; falling back to qemu-static"; fi) so
operators see when the script falls back to qemu-static; keep the existing
non-fatal behavior but surface the failure via display_alert (or the project’s
alert/logging helper).

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@lib/functions/main/rootfs-image.sh`:
- Line 19: The call to _native_armhf_setup_binfmt_elf currently uses "|| true"
which silently swallows failures; replace that pattern with an explicit
conditional: invoke _native_armhf_setup_binfmt_elf and if it returns non-zero
run a visible fallback notification (e.g. if ! _native_armhf_setup_binfmt_elf;
then display_alert "native armhf binfmt activation failed; falling back to
qemu-static"; fi) so operators see when the script falls back to qemu-static;
keep the existing non-fatal behavior but surface the failure via display_alert
(or the project’s alert/logging helper).

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: c0143a21-0438-46b0-9277-ecb11ae8e453

📥 Commits

Reviewing files that changed from the base of the PR and between 4998f52 and 152d4f6.

📒 Files selected for processing (3)

• lib/functions/main/rootfs-image.sh
• lib/functions/rootfs/qemu-static.sh
• lib/functions/rootfs/rootfs-create.sh

🚧 Files skipped from review as they are similar to previous changes (2)

• lib/functions/rootfs/rootfs-create.sh
• lib/functions/rootfs/qemu-static.sh

[iav]

@coderabbitai full review please.

Branch force-pushed and squashed into a single commit (7b782ceac). The userspace owner-flock coordination layer (~250 lines) is replaced with kernel flock(2) directly on /proc/sys/fs/binfmt_misc/qemu-arm. Earlier review threads on prior revisions are obsolete; please look at the current revision fresh.

[coderabbitai]

@iav, understood — I'll do a fresh full review of the current squashed commit (7b782ceac), treating the kernel flock(2)-on-/proc/sys/fs/binfmt_misc/qemu-arm approach as the canonical implementation and ignoring all prior revision threads.

✅ Actions performed

Full review triggered.

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@lib/functions/rootfs/qemu-static.sh`:
- Around line 329-341: The early-continue when calling
_native_armhf_setup_binfmt_elf lets the "killswitch" path
(NATIVE_ARMHF_ON_ARM64=no) bypass the later disabled-state guard and skip
prepare_host_binfmt_qemu_cross_arm64_host_armhf_target; change the logic so the
caller only continues when _native_armhf_setup_binfmt_elf reports an actual
active-native success (explicit success exit code), and when the function
returns due to the killswitch (use a distinct exit code or status), fall through
instead of continuing so the later
prepare_host_binfmt_qemu_cross_arm64_host_armhf_target and the disabled-state
guard still run; apply the same fix to the analogous block around
prepare_host_binfmt_qemu_cross_arm64_host_armhf_target (lines 356-371) and
ensure _native_armhf_setup_binfmt_elf and its callers agree on exit codes
semantics.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 6912ee78-6640-4866-9e0d-f7c2e50ad718

📥 Commits

Reviewing files that changed from the base of the PR and between 38c583f and 7b782ce.

📒 Files selected for processing (3)

• lib/functions/main/rootfs-image.sh
• lib/functions/rootfs/qemu-static.sh
• lib/functions/rootfs/rootfs-create.sh

[iav]

@coderabbitai review

[coderabbitai]

✅ Actions performed

Review triggered.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

[iav]

@coderabbitai review

[coderabbitai]

✅ Actions performed

Review triggered.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.