From b3dbe5459ca2c62ea87b3cf250f5aa191852176b Mon Sep 17 00:00:00 2001
From: Kevin Wang <wy721@qq.com>
Date: Thu, 24 Sep 2026 21:59:49 -0700
Subject: [PATCH] fix(os/mkosi): mask systemd's TPM SRK setup units

On a measured-UKI boot (GCP), systemd-tpm2-setup.service tries to write
the SRK public key to /var/lib/systemd/tpm2-srk-public-key.pem. It runs
before sysinit.target, but the writable /var overlays come from
dstack-volatile-binds.service, which the unit does not order after, so
it hits the read-only root, fails, and leaves the system degraded.

Nothing in dstack uses systemd's SRK: the TPM key provider creates its
own primary key at 0x81000100 through tpm2-tss, and no image component
uses systemd-creds or systemd-cryptenroll with a TPM. The Yocto image
builds systemd without TPM support, so these units never existed there.
Mask both SRK units instead of giving them a writable path, which would
only keep an unused object persisted in the vTPM.

Signed-off-by: Kevin Wang <wy721@qq.com>
---
 .../etc/systemd/system/systemd-tpm2-setup-early.service  | 1 +
 .../etc/systemd/system/systemd-tpm2-setup.service        | 1 +
 os/mkosi/tests/acceptance.sh                             | 9 +++++++++
 3 files changed, 11 insertions(+)
 create mode 120000 os/mkosi/mkosi.skeleton/etc/systemd/system/systemd-tpm2-setup-early.service
 create mode 120000 os/mkosi/mkosi.skeleton/etc/systemd/system/systemd-tpm2-setup.service

diff --git a/os/mkosi/mkosi.skeleton/etc/systemd/system/systemd-tpm2-setup-early.service b/os/mkosi/mkosi.skeleton/etc/systemd/system/systemd-tpm2-setup-early.service
new file mode 120000
index 000000000..dc1dc0cde
--- /dev/null
+++ b/os/mkosi/mkosi.skeleton/etc/systemd/system/systemd-tpm2-setup-early.service
@@ -0,0 +1 @@
+/dev/null
\ No newline at end of file
diff --git a/os/mkosi/mkosi.skeleton/etc/systemd/system/systemd-tpm2-setup.service b/os/mkosi/mkosi.skeleton/etc/systemd/system/systemd-tpm2-setup.service
new file mode 120000
index 000000000..dc1dc0cde
--- /dev/null
+++ b/os/mkosi/mkosi.skeleton/etc/systemd/system/systemd-tpm2-setup.service
@@ -0,0 +1 @@
+/dev/null
\ No newline at end of file
diff --git a/os/mkosi/tests/acceptance.sh b/os/mkosi/tests/acceptance.sh
index 44fb408ef..046115165 100755
--- a/os/mkosi/tests/acceptance.sh
+++ b/os/mkosi/tests/acceptance.sh
@@ -120,6 +120,15 @@ grep -q '^WatchdogSec=' "$gw_unit" || {
   echo 'dstack-gateway-checker.service must set WatchdogSec'; exit 1; }
 test ! -e "$D/../common/rootfs/wg-checker.sh"
 test ! -e "$D/../common/rootfs/wg-checker.service"
+# systemd-tpm2-setup writes the SRK public key to /var/lib/systemd before the
+# /var overlays exist, so on a measured-UKI boot it fails on the read-only root
+# and leaves the system degraded. Nothing in dstack uses systemd's SRK (the TPM
+# key provider has its own primary key), and the Yocto systemd has no TPM
+# support, so both SRK units are masked instead of given a writable path.
+for unit in systemd-tpm2-setup.service systemd-tpm2-setup-early.service; do
+  [[ $(readlink "$D/mkosi.skeleton/etc/systemd/system/$unit") == /dev/null ]] || {
+    echo "$unit must be masked in the image skeleton"; exit 1; }
+done
 # systemd enables any unit that matches no preset rule, so the enable list is
 # only meaningful with a terminal disable. Without it, every package pulled in
 # by Packages= would start at boot with no diff to 80-dstack.preset.