max / alloy

Give vmtest a machine with no keyboard, and stop it re-minting the TPM The USB gate's bench tests need hardware no machine in this tree can be. alloy-usb-gate suspends enforcement when the count of ID_INPUT_KEYBOARD devices reaches zero, and every machine here has an i8042 keyboard soldered to it that usbguard has no jurisdiction over, so the count never reaches zero on real hardware. q35 carries an i8042 property, so KEYBOARD=none is a machine with no keyboard at all and KEYBOARD=usb is one whose only keyboard is on an xHCI and can be pulled from the monitor with device_del kbd0. That is the second and third bench tests, neither of which was runnable before. Also stop the harness swapping the TPM out underneath an installed disk. swtpm_setup carries --overwrite and ran whenever the swtpm daemon was not up, rather than only on a first run, so a disk that unlocks from the TPM lost its sealed key and fell back to asking for a passphrase. It reports "Failed to unseal secret using TPM2: State not recoverable", which reads as a broken enrolment rather than as changed hardware. That matters more now: a machine with no keyboard cannot answer a passphrase prompt at all, so the TPM path is the only way an encrypted target boots in that mode.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WFBzMprSmNCfvdj2cGZyka

Author: Max Johnson <me@maxj.phd> · 2026-09-08 19:33 UTC

Signed with PGP, not checked

Commit: 26f7d65897ef3d459535c807b5b07f30379f8dda

Parent: 05d3a28

2 files changed, +83 insertions, -3 deletions

M build/vmtest/README.md +41

			@@ -47,6 +47,21 @@
47	47		TARGET_BUS=nvme ./run-vm.sh live # a target that reports `tran: nvme`
48	48		python3 install_preseeded.py # install from a medium with an answer sheet
49	49
	50	+	`KEYBOARD` decides what keyboards the machine has, and exists because the USB
	51	+	gate's bench tests need hardware no machine here can be. `usr/bin/alloy-usb-gate`
	52	+	suspends enforcement whenever the count of `ID_INPUT_KEYBOARD` devices reaches
	53	+	zero, and every physical machine in this tree has an i8042 keyboard soldered to
	54	+	it that USBGuard has no jurisdiction over, so the count never reaches zero on
	55	+	real hardware. `ps2` is q35's own i8042 controller and the default. `none` is
	56	+	`i8042=off`: no keyboard of any kind, which is the machine the gate is supposed
	57	+	to open for at boot. `usb` is `i8042=off` plus a USB keyboard on an xHCI
	58	+	controller, so the only keyboard is one USBGuard can take away; its `id=kbd0` is
	59	+	what `qmp.py cmd device_del '{"id": "kbd0"}'` unplugs, which is the third bench
	60	+	test and cannot be done from inside the guest.
	61	+
	62	+	KEYBOARD=none ./run-vm.sh installed # zero keyboards
	63	+	KEYBOARD=usb ./run-vm.sh installed # one, and it is removable
	64	+
50	65		`TARGET_BUS` decides how the target disk is attached, and it is not a
51	66		performance knob. A recipe whose disk rule is `single-internal-nvme` cannot be
52	67		exercised against a virtio disk, whose lsblk `tran` is null, and a rule that
			@@ -114,6 +129,32 @@
114	129		then written once in its own syntax with nothing to quote through nushell. Use
115	130		`sh -c '...'` only when stdin is carrying something else.
116	131
	132	+	**`run-vm.sh` re-creates the software TPM whenever swtpm is not running, and
	133	+	that silently breaks an installed disk that unlocks with one.** The setup call
	134	+	carries `--overwrite`, and it fires whenever `state/tpm/swtpm-sock` is absent —
	135	+	which is every time the swtpm daemon has been stopped, not only on a first run.
	136	+	A new TPM has a new seed, so the sealed key in the LUKS header can no longer be
	137	+	loaded, and the guest asks for the passphrase instead. What it prints is
	138	+	`Failed to unseal secret using TPM2: State not recoverable` with
	139	+	`Esys_Load() ... ErrorCode (0x0000018b)` above it, in the journal of the boot
	140	+	that prompted. It reads like a broken enrolment rather than a harness that
	141	+	changed the hardware underneath it. Delete `state/tpm/` deliberately to simulate
	142	+	a cleared TPM; do not let it happen by accident.
	143	+
	144	+	Re-enrolling a TPM2 keyslot takes two commands, not one.
	145	+	`systemd-cryptenroll --wipe-slot=tpm2 --tpm2-device=auto <dev>` answers
	146	+	`This PCR set is already enrolled, executing no operation` and changes nothing,
	147	+	because it compares against the existing enrolment before it wipes. Wipe and
	148	+	enrol separately, and read the output: the pair that works prints `Wiped slot N`
	149	+	and then `New TPM2 token enrolled as key slot N`.
	150	+
	151	+	A machine with `KEYBOARD=none` cannot answer a LUKS passphrase prompt.
	152	+	There is no input device for `vm.py type` to reach, so an encrypted target is
	153	+	unreachable in that mode and the run stalls at the prompt with no way forward.
	154	+	Bench-testing the gate against an encrypted disk therefore depends on the TPM
	155	+	path working. `install_drive.py` leaves encryption off by default, and a target
	156	+	installed that way is the one to use for keyboard-shaped tests.
	157	+
117	158		The medium's ssh installer needs privilege it does not get on its own.
118	159		`alloy install` does not escalate, and an ssh session lands unprivileged, so the
119	160		wizard draws and answers every step and then fails on the first command that

M build/vmtest/run-vm.sh +42 -3

			@@ -24,8 +24,19 @@
24	24		# would; delete that directory to simulate a cleared TPM.
25	25		mkdir -p "$SCRATCH/tpm"
26	26		if [ ! -S "$SCRATCH/tpm/swtpm-sock" ]; then
27		-	swtpm_setup --tpm2 --tpmstate "$SCRATCH/tpm" --createek --create-ek-cert \
28		-	--create-platform-cert --lock-nvram --overwrite >/dev/null 2>&1 \|\| true
	27	+	# Only mint a TPM when there is not one already. The setup call carries
	28	+	# `--overwrite`, and this branch is taken every time the swtpm daemon has been
	29	+	# stopped rather than only on a first run, so running it unconditionally gives
	30	+	# the machine a new seed behind its own back. An installed disk that unlocks
	31	+	# from the TPM then cannot load its sealed key and falls back to asking for the
	32	+	# passphrase, reporting `Failed to unseal secret using TPM2: State not
	33	+	# recoverable` — which reads as a broken enrolment rather than as hardware the
	34	+	# harness swapped out. Deleting state/tpm/ is still how you clear a TPM on
	35	+	# purpose; this only stops it happening by accident.
	36	+	if [ ! -f "$SCRATCH/tpm/tpm2-00.permall" ]; then
	37	+	swtpm_setup --tpm2 --tpmstate "$SCRATCH/tpm" --createek --create-ek-cert \
	38	+	--create-platform-cert --lock-nvram --overwrite >/dev/null 2>&1 \|\| true
	39	+	fi
29	40		swtpm socket --tpm2 --tpmstate dir="$SCRATCH/tpm" \
30	41		--ctrl type=unixio,path="$SCRATCH/tpm/swtpm-sock" \
31	42		--flags startup-clear --daemon
			@@ -48,8 +59,36 @@
48	59		*) echo "unknown TARGET_BUS '$TARGET_BUS'; expected virtio or nvme" >&2; exit 1 ;;
49	60		esac
50	61
	62	+	# What keyboards the machine has, which is hardware the gate's bench tests need
	63	+	# and no machine here can be. `usr/bin/alloy-usb-gate` suspends enforcement
	64	+	# whenever the count of ID_INPUT_KEYBOARD devices reaches zero, so exercising it
	65	+	# means a machine that can have zero, and every physical machine here has an
	66	+	# i8042 keyboard soldered to it that usbguard has no jurisdiction over.
	67	+	#
	68	+	# ps2 q35's own i8042 controller. The default, and what every scenario
	69	+	# written before this one assumes.
	70	+	# none i8042=off. Zero keyboards of any kind, which is the machine the gate
	71	+	# is supposed to open for at boot.
	72	+	# usb i8042=off plus a USB keyboard on an xHCI controller, so the only
	73	+	# keyboard on the machine is one usbguard CAN take away. `id=kbd0` is
	74	+	# load-bearing: unplugging it mid-session is `device_del kbd0` over the
	75	+	# monitor, which is the third bench test and cannot be done from inside
	76	+	# the guest.
	77	+	#
	78	+	# q35 carries no USB controller unless one is asked for, so the `usb` case adds
	79	+	# an xHCI rather than relying on a default that is not there.
	80	+	KEYBOARD="${KEYBOARD:-ps2}"
	81	+	case "$KEYBOARD" in
	82	+	ps2) MACHINE="q35" ; KBD_ARGS=() ;;
	83	+	none) MACHINE="q35,i8042=off"; KBD_ARGS=() ;;
	84	+	usb) MACHINE="q35,i8042=off"
	85	+	KBD_ARGS=(-device qemu-xhci,id=xhci -device usb-kbd,bus=xhci.0,id=kbd0) ;;
	86	+	*) echo "unknown KEYBOARD '$KEYBOARD'; expected ps2, none or usb" >&2; exit 1 ;;
	87	+	esac
	88	+
51	89		args=(
52		-	-enable-kvm -machine q35 -cpu host -m 4608 -smp 4
	90	+	-enable-kvm -machine "$MACHINE" -cpu host -m 4608 -smp 4
	91	+	"${KBD_ARGS[@]}"
53	92		-drive if=pflash,format=raw,unit=0,readonly=on,file=/usr/share/OVMF/OVMF_CODE_4M.fd
54	93		-drive "if=pflash,format=raw,unit=1,file=$VARS"
55	94		"${TARGET_ARGS[@]}"