Adversarial review of 9db25ba60 found four residual defects in the
auto-update asset install, plus a doc line it left contradicting itself.
- install_auto_update_assets copied the bundled helper into the staged
mktemp file with an unchecked cp, and both call sites invoke the
function under `if !`, which suppresses errexit for its whole body. A
failed copy (ENOSPC, EIO) fell through to
configure_auto_update_script_repo, whose awk emits a lone GITHUB_REPO=
line for empty input, so a shebang-less one-line stub replaced the
working helper with a "script" that only ever exits 0 - silently
disabling unattended updates. Check the cp, and refuse the swap unless
the staged helper is non-empty and starts with #!.
- Both units were rendered with a bare truncating `cat > "$unit"` whose
status was never checked, and the function's last statement is
safe_systemctl daemon-reload, which returns 0 by design. A failing
write truncated a working unit and still reported success. Render each
unit to ${path}.tmp and commit it with a checked rename, so a failure
leaves the installed unit byte-identical.
- The widened ReadWritePaths could not reach deployed boxes: the unit
that grants the write access is itself the file that has to be
rewritten, and on an existing install the sandbox running the
installer excludes /etc/systemd/system and /usr/local/bin (EROFS). The
Go update pipeline cannot carry it either - pulse.service runs as
User=pulse with its own ProtectSystem=strict over the install and
config dirs only. So probe each destination directory up front and,
when one is blocked, re-exec this already-signature-verified installer
through systemd-run with a new internal --repair-auto-update-units
entry point: PID 1 forks the transient unit, so it starts in the host
mount namespace instead of inheriting the sandbox. The installer is
copied into the install dir first because the calling unit's
PrivateTmp=yes hides its /tmp copy from PID 1. The escape needs root
and systemd-run, and never recurses.
- Keep the ReadWritePaths entries as directory grants: every write now
commits with a rename from a sibling staging file, and rename needs
write access on the containing directory, so the file-level entries
systemd would otherwise accept cannot work. Document the tradeoff in
the unit and the subsystem contract instead.
The deployment-installability contract still claimed the update sandbox
leaves "only the install dir, config dir and /tmp" writable, which the
paragraph the same file gained in 9db25ba60 contradicts; the same stale
rationale had been copied into two test comments.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Residual auto-update defects found while triaging #1643 and #1637
(the primary regression was fixed in 806cbe83d):
- The generated pulse-update.timer carried both OnCalendar=daily and
OnCalendar=02:00, so with RandomizedDelaySec=4h every box attempted
two updates per day (00:00-04:00 and 02:00-06:00 windows). Keep the
single documented 02:00 schedule.
- The generated pulse-update.service sandbox (ProtectSystem=strict)
excluded the helper and unit directories from ReadWritePaths, so the
unattended path could never refresh /usr/local/bin/pulse-auto-update.sh
or rewrite the units - updater fixes only reached boxes via manual
installs. Grant the sandbox write access to both directories on
purpose.
- Because the unattended path replaces the helper bash is currently
executing, stage the new helper next to its destination and swap it
in with an atomic rename only after repo configuration succeeds. A
failed download or configure now leaves the previously working helper
in place instead of rm -f'ing it out from under the enabled timer's
ExecStart.
- Delete scripts/systemd/pulse-update.{service,timer}: orphaned
reference copies that had drifted from the units install.sh actually
generates and were referenced by nothing.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two bugs from #1630 that combined to take installs down silently:
1. perform_update()'s install-failed rollback branch restored the backup
but never restarted pulse.service. Since the generated
pulse-update.service gates on ExecCondition=systemctl is-active,
every later timer run was then skipped and the install stayed down
until manual intervention. Restart is now guaranteed by a
service_was_active-guarded restart in that branch plus an
ensure_service_restarted RETURN trap so no exit path can miss it
(re-fix of #1323, originally c0b3a0e66, lost in 778a2577b and only
partially restored in 672e81985).
2. install.sh aborted under errexit when writing the /bin/update helper
on a read-only filesystem - after the new binary was installed and
the service stopped, landing in bug 1's no-restart branch. The stock
pulse-update.service uses ProtectSystem=strict, so /bin and
/usr/local/bin are read-only on stock unattended updates; transient
read-only remounts hit the same path. The helper write, PATH
appends, and the /usr/local/bin/pulse symlink are now idempotent and
non-fatal with a warning (install_binary_symlink).
Contract: deployment-installability now pins fail-closed service
availability for unattended updates and non-fatal writes outside the
hardened unit's writable set, with proofs in pulse_auto_update_test.go
and root_install_sh_test.go plus shell regression coverage in
scripts/tests/test-pulse-auto-update.sh (installer-exits-nonzero path)
and scripts/tests/test-install-update-resilience.sh (read-only helper
and symlink paths, verified under set -e, root-safe via ENOTDIR).
Fixes#1630
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Customer report against v6.0.5 (2026-07-13):
- resolve_target_release fell back to hardcoded v4.5.1 when the GitHub API
and the /releases/latest redirect both failed, silently downgrading a
v5.0.17 install two major versions. Now: fail with --version guidance,
and refuse any auto-resolved target older than the installed version.
- backup_existing wrote /etc/pulse.backup.<ts>, which is read-only under
pulse-update.service (ProtectSystem=strict, ReadWritePaths=/opt/pulse
/etc/pulse /tmp), so every unattended update on systemd/LXC failed at
the backup step. Now: fall back to $INSTALL_DIR/config-backups when the
config parent is not writable.
The installer's unit hardens with NoNewPrivileges=true, which strips the
ping binary's setuid/file capabilities, so ICMP availability probes could
never work on a systemd install. Grant the capability ambiently instead,
document the systemctl edit override for existing units, and pin the
hardening block with an install test and contract clause.
Addresses #1554 (discussion)
Two installer bugs reported after the v5 to v6 upgrades:
- The server installer (install.sh) guarded its "am I being sourced?" check
with a bare ${BASH_SOURCE[0]}. When piped to bash (curl ... | bash) there is
no source file, so under `set -u` the run aborted with "BASH_SOURCE[0]:
unbound variable" before the installer body. Default the lookup and only
early-return on a genuine source. Regression of the v5 fix in #1396. (#1526)
- The agent installer (scripts/install.sh) compared the downloaded binary's
version ("v6.0.4") against the server /api/version value ("6.0.4") verbatim,
raising a spurious mismatch warning on matching versions. Strip a leading "v"
from both sides before comparing so only a genuine difference warns. (#1527)
Refs #1526#1527
Two GA-blocking bugs in the auto-update install path:
The pulse-update.service unit was rendered through an unquoted heredoc
containing $${PULSE_SERVICE_NAME}, which bash expands to the installer's
PID, so the unit shipped with ExecCondition checking a service like
"24757{PULSE_SERVICE_NAME}". The condition always failed and systemd
silently skipped every scheduled run, for fresh installs and upgrades
alike. The heredoc now renders the detected service name directly:
ExecCondition=/bin/sh -c 'systemctl is-active --quiet pulse'.
Upgrades also never refreshed the updater assets: a v5 box with
auto-updates enabled keeps pulse-update.timer and autoUpdateEnabled=true,
so the update/reinstall flows never re-ran setup_auto_updates and the
v5.1-pinned /usr/local/bin/pulse-auto-update.sh stayed in place, logging
"Already running latest version" forever instead of selecting v6
releases. The asset-install half of setup_auto_updates is now a shared
install_auto_update_assets(), and a new refresh_auto_updates() rewrites
the helper script and units unconditionally whenever the timer already
exists, without touching system.json or the timer's enabled state. All
five install flows (update, reinstall, --version, --source, fresh) are
wired.
Tests now render the real unit and execute the rendered ExecCondition
against a recording systemctl stub instead of asserting source-text
fragments, pin refresh_auto_updates behavior (stale helper replaced,
system.json and enablement untouched), and pin the call-site wiring.
The deployment-installability contract records both invariants.
Note: rewritten message only — the original local commit carried a
parallel agent's licensing commit message due to a shared-index race;
the tree is byte-identical to the hook-verified original.
The install-time auto-register path (auto_register_pve_node) parsed the API
token secret out of pveum's box-drawing table output with a fragile awk
column-split. The web-setup render path was already hardened to request
'pveum ... --output-format json' first and parse the value field, but this
secondary install.sh path was never ported.
auto_register_pve_node now requests --output-format json first (falling back
to the bare --privsep 1 form only when an older pveum rejects the JSON flag,
which keeps the secure-installer contract pin on that form satisfied) and
extracts the secret via a new extract_pve_token_value helper: JSON value-field
parse first, then a locale-independent box-drawing table fallback (normalizes
the column separator to a plain pipe byte-wise before splitting, so it works
regardless of host locale). This mirrors the hardened render path and removes
the silent-failure / mis-parse risk when pveum table formatting drifts.
Functional + contract tests in root_install_sh_test.go; deployment-installability
contract documents the deterministic extraction. The host-agent path
(internal/hostagent/proxmox_setup.go setupPVEToken) carries an agent-lifecycle
token-permission proof obligation and is left for a governed lane.
A Proxmox host upgraded from v5 may still carry the legacy pulse-sensor-proxy
footprint (binary, systemd units, runtime/state dirs, dedicated service user,
and managed SSH keys in root's authorized_keys). install.sh --uninstall removed
everything for the Pulse server itself but left that legacy footprint behind,
so a 'complete uninstall' was not complete -- most notably it left SSH key
entries in /root/.ssh/authorized_keys.
uninstall_pulse now calls cleanup_local_sensor_proxy, which removes the LOCAL
footprint only: stop/disable the units, remove the binary/units/runtime/state
dirs, strip the '# pulse-managed-key' / '# pulse-proxy-key' entries, and remove
the service user/group. It is presence-gated (silent no-op when no proxy was
installed). The aggressive cluster-wide authorized_keys removal and
pulse-monitor@pam API-user deletion stay behind the explicit standalone
scripts/uninstall-sensor-proxy.sh, which we print a pointer to.
Functional + contract tests in scripts/installtests/root_install_sh_test.go;
deployment-installability contract documents the new uninstall removal scope.
Finishes the deferred half of the v5->v6 parity fix for #1323 (the
pulse-auto-update.sh half landed in 672e81985). The interactive install.sh
update/reinstall flow stopped a running Pulse then called start_pulse, which
tolerates a silent start failure (common on unprivileged LXC) by printing a note
and returning 0 — leaving Pulse stopped under an "installation completed!" message.
- stop_pulse_for_update records whether Pulse was running before the update.
- start_pulse, only when Pulse was running before (PULSE_WAS_ACTIVE), no longer
accepts a silent start failure: it verifies the service became active
(wait_for_service_active, 20s), retries one explicit start, and surfaces a clear
error + diagnostics if it still will not come up. Fresh installs are unchanged
(the flag stays false, so the reassuring container note is kept).
- Wired into all three update/reinstall sites; added a BASH_SOURCE guard so the
installer's functions can be unit-tested without running the installer.
Scope: fixes the #1323 'restart silently failed' case. Does NOT add a binary
rollback (download_pulse deletes bin/pulse.old right after the swap) — a
bad-release rollback is a separate concern.
Test: scripts/tests/test-install-update-resilience.sh (sources install.sh, stubs
systemctl, asserts was-active capture + retry + clear error). Go installtests +
bash -n confirm the guard does not change installer execution.
The running server serves the unified AGENT installer at its /install.sh
endpoint, which the "Install on Linux" wizard fetches with --url/--token-file.
For published releases handleDownloadInstallScriptCommon only serves the local
/opt/pulse/scripts/install.sh when its .sig and .sshsig sidecars sit next to it;
otherwise it proxies the top-level GitHub install.sh asset. Since 49412357a that
asset is the SERVER installer (LXC/systemd quickstart, --version), so the proxy
now hands the agent wizard a server installer that rejects --url with
"Unknown option".
deploy_agent_scripts copied only the .sh/.ps1 files, never the sidecars, so every
published-release LXC/systemd install always hit the proxy and served the wrong
script. The Docker image already deploys both sidecars (Dockerfile); LXC/systemd
did not. This was latent across rc.1-rc.5 (the proxied asset was the agent
installer then) and surfaced in rc.6 once the asset flipped to the server
installer.
Deploy the .sig/.sshsig sidecars alongside each script so the endpoint serves
the local agent installer and never falls back to the server-installer asset.
Add a behavioral guard that runs deploy_agent_scripts and asserts the sidecars
land next to the served script, and document the served-endpoint sidecar
requirement in the deployment-installability contract.
install.sh refuses to run inside Docker (correct behavior for end
users), but the install-sh-smoke gate runs the documented systemd
install path inside a privileged systemd-in-Docker container — the
one legitimate bypass case. Added PULSE_INSTALL_ALLOW_DOCKER=1
escape hatch in check_docker_environment() and set it on the docker
exec in the smoke workflow. Takes effect on the next RC; v6.0.0-rc.6's
published install.sh predates this and is signed-frozen, so its
post-publish smoke gate will remain red until the next prerelease.
The --archive PATH flag (and the Proxmox LXC bootstrap which
propagates --archive into the container) reached install_pulse_archive
without going through download_release_archive, which is the only
place that previously verified the cryptographic signature on a
Pulse release tarball. That made --archive a clean signature-
verification bypass: anyone who could swap a local tarball
between download and install (shared /tmp on a Proxmox host, a
maintainer testing a swapped local file, etc.) would get
arbitrary root code execution at extract time.
install.sh:
- install_pulse_archive now requires ${archive_path}.sshsig and
verifies it via the existing verify_release_signature helper
before tar runs. Missing or invalid signature is a hard
fail-closed.
- download_release_archive now keeps the .sshsig alongside the
archive at ${archive_path}.sshsig instead of stashing it in a
mktemp file that got deleted post-verify. install_pulse_archive
re-verifies it; no behavior change for the curl|bash path beyond
the extra integrity check at extract time.
- Proxmox LXC bootstrap now pct push'es ${container_archive_source}.sshsig
alongside the archive into the container so the in-container
install_pulse_archive run can verify. Missing sidecar fails the
bootstrap (cleanup_on_error tears down the half-provisioned
container).
- Cleanup paths that remove the temp archive on success/failure
now also remove the sidecar.
- --help advertises the .sshsig requirement.
install-sh-smoke.yml:
- Smoke test now pulls ${tarball}.sshsig alongside the tarball
before running install.sh --archive inside the test container.
Trade-off: maintainer flows that did `install.sh --archive
/path/to/local.tar.gz` against an unsigned local tarball will now
fail with a clear "Required signature sidecar not found" error.
Maintainers can either pull the corresponding .sshsig from the
release alongside the tarball, or skip --archive in favour of the
canonical download path.
install.sh:
- find_pulse_binary_in_dir now rejects symlinks, so a malicious
tarball that smuggled a symlink at the expected binary path
cannot redirect the cp to $INSTALL_DIR/bin/pulse onto an
arbitrary file outside the extract dir.
- tar -xzf calls in install_pulse_archive now pass
--no-same-owner --no-overwrite-dir, defending against archive
entries that try to chown extracted files or rewrite existing
directory metadata.
Demo deploy workflows:
- deploy-demo-server.yml routes inputs.target through env: so the
workflow_dispatch input cannot break out of the case statement,
even though type: choice constrains it client-side.
- update-demo-server.yml routes inputs.tag, inputs.target, and
github.event.release.tag_name through env: for the same reason.
All defense-in-depth: download_release_archive still signature-
verifies the tarball, and the demo workflows are gated to write-
access dispatchers. The argument-vs-env distinction matters once
the trigger surface or trust boundary changes.
Move the guest-agent file-read of /proc/meminfo earlier in the memory
fallback chain so it runs before RRD, giving real-time MemAvailable that
correctly excludes reclaimable buff/cache on Linux VMs. Also add
VM.GuestAgent.FileRead permission for PVE 9 and fix install.sh to use
comma-separated privilege strings.
The auto-update flow stops the Pulse service before applying updates.
If the update fails, the rollback path restored files but never
restarted the service. Since the main unit was explicitly stopped
(not crashed), systemd's Restart=always didn't rescue it.
Add restart-on-failure guards to both pulse-auto-update.sh and
install.sh so Pulse is always restarted after a failed update attempt.
The backup_existing function copied the entire config directory
(including metrics.db at ~2.5GB) on every upgrade with no cleanup.
On small VMs this filled the disk within a few releases.
The upgrade only swaps the binary; config files are not modified,
so the backup served no practical purpose.
The v4 installer added mount entries for /run/pulse-sensor-proxy to LXC
container configs. After upgrading to v5 and rebooting, /run (tmpfs) is
wiped and the container fails to start. The installer now detects and
removes these stale mp<N> and lxc.mount.entry references automatically
when run on a PVE host, and the upgrade docs include manual fix steps.
On SELinux-enforcing systems (Fedora, RHEL, CentOS), binaries installed to
non-standard locations need proper security contexts for systemd to execute
them. Without this, systemd fails with 'Permission denied' even when the
binary has correct Unix permissions.
Changes:
- Add restore_selinux_contexts() function to both install scripts
- Uses restorecon (preferred) or chcon (fallback) to set bin_t context
- Only runs when SELinux is detected and enforcing
- Called after binary installation, before systemd service start
Alpine uses apk/OpenRC instead of apt/systemd, which the Pulse
LXC installation flow requires. This prevents failed installations.
- Remove Alpine download option from advanced mode
- Add note that Pulse requires Debian-based templates
- Add validation when user selects from template list to catch
Alpine/Gentoo/Arch/Void and fall back to Debian 12 with warning
Related to #915
courtmanr@gmail.com