Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
275 commits
Select commit Hold shift + click to select a range
299c757
orchestrator: Return RestoreOutcome from Recovery::restore
chrysh Sep 21, 2026
608cc20
orchestrator: Platform::execute must not block
chrysh Sep 22, 2026
46f8930
ast10x0/harness: reap orphaned pi_test_runner processes holding the U…
JesseMelon Sep 18, 2026
a561da8
util: IPC abstraction
cfrantz Jun 6, 2026
9de4d59
util/ipc: extend IpcChannel with async transact/cancel
rusty1968 Sep 22, 2026
24607fd
util/ipc: address review feedback, split roles, add AsyncTransaction
rusty1968 Sep 22, 2026
12f3b87
util/ipc: sort BUILD.bazel deps alphabetically
rusty1968 Sep 22, 2026
a770101
target/ast10x0: add QEMU test exercising util_ipc AsyncTransaction
rusty1968 Sep 22, 2026
d5aa611
util/ipc: add IpcInitiator::as_raw, stop leaking H's internals
rusty1968 Sep 22, 2026
0e7294e
util/ipc: Completion/StartError wrap Buffers, relocate signal doc
rusty1968 Sep 22, 2026
33aefb0
util/ipc: impl From<StartError> for Error, drop map_err at call sites
rusty1968 Sep 22, 2026
f202c83
target/ast10x0: fix CI format check on the new util_ipc test
rusty1968 Sep 22, 2026
7f8d126
util/ipc: suppress semgrep unsafe-usage on async_transact_start call
rusty1968 Sep 22, 2026
60d99cf
util/ipc: fix cancel() losing buffers on syscall error, add IpcHandle…
rusty1968 Sep 23, 2026
297d14a
target/ast10x0: add drop-cancel/double-start/try_recv-early test cases
rusty1968 Sep 23, 2026
d98fa17
config: reject sub-millisecond checkpoint timeouts
chrysh Sep 22, 2026
8d1db78
walk: take DeviceConfig instead of a raw checkpoint slice
chrysh Sep 22, 2026
ef0607a
orchestrator: gate Isolable/Cascading immediately on DeviceFatal boot…
chrysh Sep 22, 2026
0f08af0
services: flash client/server
cfrantz Jun 7, 2026
6510ef1
util/error: add ErrorCode::check_status
rusty1968 Aug 19, 2026
14793c3
target/ast10x0: allow intra-page program starts in SpiNorFlash
rusty1968 Aug 19, 2026
82a693c
util/error: add AST10x0 flash error module
rusty1968 Aug 19, 2026
5d7f2cb
target/ast10x0: add FMC flash service backend
rusty1968 Aug 19, 2026
0c52f8b
target/ast10x0: add flash service server image scaffolding
rusty1968 Aug 19, 2026
b2c3ad5
target/ast10x0: add flash service client test and system image
rusty1968 Aug 19, 2026
03784c0
target/ast10x0: move FMC pinmux from flash backend to kernel target init
rusty1968 Aug 19, 2026
0a43530
ast10x0: run flash service test under QEMU via fmc-model=w25q64 + see…
rusty1968 Aug 20, 2026
10e3d17
ast10x0: make flash test non-destructive with sector backup/restore (…
rusty1968 Aug 20, 2026
c8432e9
util/sfdp: add JESD216 SFDP parser with datasheet-derived known-good …
JesseMelon Aug 25, 2026
1f61553
hal/flash: add runtime geometry accessors to FlashDriver
JesseMelon Sep 8, 2026
7509f5c
util/sfdp: harden parser; add ConstParamTy and public decode_geometry
JesseMelon Sep 1, 2026
595c263
ast10x0/smc: config passed at init; derive per-CS geometry from SFDP;…
JesseMelon Sep 1, 2026
8e2d4aa
ast10x0/smc: port tests to SFDP-derived config; add sfdp discovery test
JesseMelon Sep 1, 2026
3a6176a
ast10x0: flash service backend and QEMU FMC model on discovered geometry
JesseMelon Sep 8, 2026
99b835f
ast10x0/tests/flash: flash service test with UART sentinel completion
JesseMelon Sep 1, 2026
ba367e8
ast10x0/smc: report geometry through GeometrySource so Pinned folds t…
JesseMelon Sep 3, 2026
a15d690
util/sfdp: rev-check only the BFP parameter header
JesseMelon Sep 10, 2026
62150dc
orchestrator: capabilities: fix SocReader's fault encoding, rename it
rusty1968 Sep 17, 2026
39b3814
orchestrator: config: add a doctested example alongside the ignored one
rusty1968 Sep 17, 2026
0779a17
orchestrator: runtime itest: fix timing/overflow, cover device faults
rusty1968 Sep 17, 2026
ea92d3e
orchestrator: Wire the Updatable seam through the board
chrysh Aug 25, 2026
e848d23
orchestrator: Connect the update frontend to the SM
chrysh Aug 25, 2026
3139ef2
docs: add orchestrator runtime concept page and board composition
rusty1968 Aug 27, 2026
aae0178
orchestrator: Add the update intake seam for an update source
chrysh Aug 31, 2026
43f06e0
orchestrator: Note that the intake phase order is not fixed
chrysh Sep 1, 2026
83304cb
orchestrator: add IncrementalVerifier capability trait
chrysh Sep 8, 2026
739b2ad
orchestrator: Add the per-device image layout to the device-table schema
chrysh Sep 16, 2026
7689e23
i2c: Make ArmedDma borrow the controller for the armed window
JesseMelon Sep 17, 2026
95d1f06
pigweed: patch the SysTick wrap double-count that ran the clock at 2x
JesseMelon Sep 18, 2026
36a9b3f
mctp/transport-i2c: Report the payload MTU, not the packet MTU, to th…
JesseMelon Sep 22, 2026
ad6f1ec
i2c: Derive the payload ceiling from the IPC buffer so a full MCTP fr…
JesseMelon Sep 22, 2026
4a8e436
orchestrator: Add GpioResetControl, a ResetControl over a GPIO output…
JesseMelon Sep 17, 2026
be4ef3a
orchestrator: Add GpioReadyMonitor, a boot-status reader over one own…
JesseMelon Sep 17, 2026
35113d9
ast10x0: Declare the mock BMC as a managed device the RoT resets and …
JesseMelon Sep 17, 2026
36a4026
ast10x0/harness: timestamp UART lines and stream the slave from flash…
JesseMelon Sep 18, 2026
b0c12e4
ast10x0/tests/pldm: two-card firmware update streaming the image into…
JesseMelon Sep 18, 2026
74b70cd
util/region: Mint MMIO regions as move-only tokens generated from sys…
JesseMelon Sep 23, 2026
6b7b6fb
ast10x0/gpio: Take the granted region as a type parameter so a missin…
JesseMelon Sep 23, 2026
e66bf05
ast10x0/tests: Take MMIO regions as tokens, whole-aperture for kernel…
JesseMelon Sep 23, 2026
0966032
ast10x0/board: Mark bmc_host_test host-only so firmware sweeps skip it
JesseMelon Sep 23, 2026
8500b31
orchestrator: Clear the pending update job on a deferred or aborted r…
chrysh Sep 24, 2026
aad6177
config: make the golden image optional
chrysh Sep 24, 2026
2b432f8
orchestrator: Assert on Sink overflow in tests
chrysh Sep 23, 2026
bb47064
orchestrator: Add the SelfUpdate capability trait
chrysh Sep 21, 2026
dc9c7e4
orchestrator: Scope TrialBoot to downstream devices as DeviceTrialBoot
chrysh Sep 22, 2026
d80b42c
orchestrator: Say the trial was judged, not that it judged itself
chrysh Sep 24, 2026
adf331d
util/ipc: Rework the AsyncTransaction API
chrysh Sep 25, 2026
f81db36
target/ast10x0: Let util crates see the target_ast10x0 constraint
chrysh Sep 25, 2026
8387af8
util/service: Add the IPC service seam traits
chrysh Sep 23, 2026
1b54ed3
util/ipc: Add AsyncChannelTransport, the kernel side of the service seam
chrysh Sep 25, 2026
438c6a0
ast10x0: Wire SGPIOM pin 42 to the bl1 boot checkpoint
chrysh Sep 17, 2026
0f689fc
pldm/firmware_device: Return Completed when the FD leaves update mode
JesseMelon Sep 24, 2026
dc1e556
ast10x0/tests/pldm: Run the full DSP0267 sequence, inventory through …
JesseMelon Sep 24, 2026
f475cd3
hal: pin-token capability vocabulary and compile-time field-mux MMIO …
JesseMelon Aug 17, 2026
80f40dc
hal: GPIO pin capability over the pin-token model
JesseMelon Aug 17, 2026
d6093e8
hal: I2C bus capability and opt-in command-word kit
JesseMelon Aug 17, 2026
d99181b
ast10x0/scu: pin universe and compile-time-coalesced SCU routing
JesseMelon Aug 17, 2026
2bd13f9
ast10x0/gpio: port to pin-token model with data-driven register map
JesseMelon Aug 17, 2026
aad2aae
ast10x0/i2c: port controller onto pin-token authority and command-wor…
JesseMelon Aug 17, 2026
c059e6b
ast10x0/board: wire I2C buses from pin tokens and route once
JesseMelon Aug 17, 2026
2db9da7
ast10x0/tests: port GPIO and I2C tests to pin-token APIs
JesseMelon Aug 17, 2026
f547233
hal, ast10x0: Type GPIO direction and add buffer-mode I2C open_bus
JesseMelon Sep 17, 2026
05d9ab0
hal, ast10x0/gpio: Bind GPIO pins through a granted region block
JesseMelon Sep 25, 2026
893cd7d
util/io: add Storage trait with split-phase writes
chrysh Sep 24, 2026
b033706
docs: add IPC service stack diagram
chrysh Sep 26, 2026
4050578
fixup! docs: add IPC service stack diagram
chrysh Sep 26, 2026
f3f1d7c
util/service: Add the generic loopback transport
chrysh Sep 26, 2026
c225a95
pldm: Add IPC API crate (wire format, types, transport seam)
chrysh Sep 22, 2026
442e740
pldm: Add IPC server crate with dispatch and loopback transport
chrysh Sep 22, 2026
fb67d3c
pldm: Move the IPC server onto the shared service seam
chrysh Sep 25, 2026
488dac3
driver: compose Recovery into BoardCapabilities
chrysh Sep 24, 2026
1559903
ast10x0/tests: add recovery scenarios to the orchestrator runtime test
chrysh Sep 25, 2026
4bf0639
tests/runtime: rename terminal to final_event, tighten assertions
chrysh Sep 26, 2026
4bddf50
Replace map_err with From impls and ? operator
chrysh Sep 26, 2026
0a899e0
util/service: Say that an error from poll discards the response
chrysh Sep 26, 2026
11b7825
pldm: Answer MalformedRequest for a frame that does not decode
chrysh Sep 26, 2026
3198e81
util/service: Name the test target for the crate, cover ResponseCode:…
chrysh Sep 26, 2026
9e3896d
pldm: Rename ipc-api and ipc-server to api and server
chrysh Sep 28, 2026
6d711ae
capabilities: Add Progress, adopt it in StageProgress and PollOutcome
chrysh Sep 27, 2026
195d778
pldm: Name the FD IPC server and handler for the channel they serve
chrysh Sep 29, 2026
50a0499
pldm: Say that DenyActivate revokes a stored grant
chrysh Sep 30, 2026
227cab3
pldm: Add the IPC client crate, the orchestrator's side of the channel
chrysh Sep 27, 2026
31ce90b
pldm: Drop the unused PldmIpcError
chrysh Sep 29, 2026
574eacc
sm: Merge StageUpdate + AuthenticateUpdate into AuthenticateStageUpdate
chrysh Sep 30, 2026
4376bc1
pldm: Name the FD ops as commands, not grants
chrysh Sep 30, 2026
1ae6a9b
pldm: Carry a reason on RejectOffer
chrysh Sep 30, 2026
80bf1df
pldm: Add the always-perform update gate
chrysh Sep 25, 2026
bba04e8
orchestrator: record and discard the update job
chrysh Sep 30, 2026
cd194ea
config: declare where eRoT-owned records live
chrysh Sep 30, 2026
71a3b49
config: split the schema into modules
chrysh Sep 30, 2026
0c2dddf
util/io: Add a byte-range read seam and a window over it
chrysh Sep 30, 2026
bcf87f0
util/io: Split the crate root into modules
chrysh Sep 30, 2026
c6effb2
orchestrator: Add the staging region and the prepare executor
chrysh Sep 30, 2026
28dfa03
orchestrator: Add the update pump
chrysh Sep 30, 2026
ddda11e
sm: Bound the commit window to the component that was activated
chrysh Sep 27, 2026
0e75d12
driver: Refuse an update request the platform cannot answer
chrysh Sep 27, 2026
a91a50d
sm: Split lib.rs into role modules
chrysh Oct 1, 2026
62768a1
ci: Run on pull requests against ocp-global-demo-wip
chrysh Oct 2, 2026
6fa6a1a
util: Collapse the nested if in Delayed::poll
chrysh Oct 2, 2026
f6cf60b
services: Fix the remaining clippy errors
chrysh Oct 2, 2026
251a16a
docs: Drop the generics from the Loopback participant
chrysh Oct 2, 2026
69f74f1
docs: Add developer-facing trust boundaries page
chrysh Oct 1, 2026
0c729e9
config: Declare a component once, in the device table
chrysh Oct 1, 2026
3b8ef26
capabilities: Rename arm_trial to set_trial_pending
chrysh Oct 3, 2026
2304c8e
services: Say fail secure, not fail closed
chrysh Oct 3, 2026
9845449
target: Give the QEMU runtime test's device its attrs
chrysh Oct 3, 2026
e04b826
driver: Settle the eRoT's last self-update at boot
chrysh Sep 27, 2026
99f146e
driver: Advance the eRoT's own floor when the update agent asks
chrysh Oct 4, 2026
79ddd63
orchestrator: Say start, not arm, outside the watchdogs
chrysh Oct 4, 2026
4fdddb1
driver: Bring the spare slot up to date after a commit
chrysh Sep 27, 2026
993219b
driver: Leave a self-managed component's slots to the device
chrysh Oct 5, 2026
f63b72e
driver: Collapse the floor-commit if, which clippy rejects
chrysh Oct 5, 2026
0c98dce
sm: Walk the platform again after an update activates
chrysh Oct 4, 2026
0f71c5f
driver: Wire IncrementalVerifier into the update pump
chrysh Oct 5, 2026
e9fd1d9
driver: Test the verify path with a chunked stub verifier
chrysh Oct 5, 2026
bd062c9
driver: Report a verifier fault
chrysh Oct 5, 2026
35b1236
driver: Move the update pump into its own file
chrysh Oct 5, 2026
9a1c8fd
capabilities: Add the Measurer trait and Measurement newtype
chrysh Oct 6, 2026
9982f92
driver: Measure a component's image while verifying it
chrysh Oct 5, 2026
e86f765
sm: End an update the requester withdrew
chrysh Oct 6, 2026
d5fcaf5
harness: Seed a CS image from a file
chrysh Oct 3, 2026
8ba0756
tools: Add the flash image builder
chrysh Oct 3, 2026
6f11f2f
services/test: Add the mock BMC boot model
chrysh Oct 3, 2026
ca7e35f
services/mctp: Add the loopback transport
chrysh Oct 3, 2026
81da269
target/ast10x0: Add the mock BMC integration test
chrysh Oct 3, 2026
ca78963
target/ast10x0: Add the PLDM update integration test
chrysh Oct 4, 2026
01d4aa6
target/ast10x0: Run the real orchestrator in the boot scenario
chrysh Oct 5, 2026
d490421
target/ast10x0: Put the orchestrator behind the PLDM update
chrysh Oct 5, 2026
b1dda20
target/ast10x0: Give the test boards an update verifier
chrysh Oct 5, 2026
508aa74
target/ast10x0: Stop the PLDM update test failing at discovery
chrysh Oct 5, 2026
a26114f
target/ast10x0: Describe how to run the integration tests
chrysh Oct 5, 2026
a717d68
target/ast10x0: Make the negative scenarios test targets
chrysh Oct 5, 2026
9deba1b
target/ast10x0: Stage the update into flash on CS1
chrysh Oct 5, 2026
fcec9bf
harness: Check what the guest left in CS1
chrysh Oct 5, 2026
382457c
target/ast10x0: Say what the flash check proves in the README
chrysh Oct 5, 2026
7687d2b
target/ast10x0: Command the firmware device over the shipped IPC seam
chrysh Oct 5, 2026
f0290ec
target/ast10x0: Fix and explain the memory map comments
chrysh Oct 8, 2026
35cf1bb
target/ast10x0: Fix review findings in the PLDM update tests
chrysh Oct 8, 2026
ea24f01
target/ast10x0: Write down what the integration tests do not cover
chrysh Oct 5, 2026
5fb39f3
target/ast10x0: Add the full update scenario
chrysh Oct 8, 2026
5eb7a51
target/ast10x0: Give the full update scenario its negatives
chrysh Oct 6, 2026
763dce6
target/ast10x0: Document untested parts of the DSP0267 Type 5 flow
chrysh Oct 6, 2026
6aa5bfd
target/ast10x0: Reword test comments to stand on their own
chrysh Oct 8, 2026
bb1d675
target/ast10x0: Fix the two builds the merges broke
chrysh Oct 8, 2026
b9d8582
crypto/api: add the crypto IPC wire protocol
chrysh Oct 1, 2026
bd2fda0
crypto/client: add the crypto IPC client
chrysh Oct 1, 2026
644b863
crypto: tighten the wire protocol API
chrysh Oct 10, 2026
3e6d69c
crypto_client: drop the duplicated transport paragraph
chrysh Oct 10, 2026
690c72e
docs: what the orchestrator cannot check during an update
chrysh Sep 29, 2026
5e73617
target/ast10x0: Discover the terminus before offering an update
chrysh Oct 8, 2026
69b32db
target/ast10x0: Error on one data request in the update scenario
chrysh Oct 8, 2026
4ec9c03
target/ast10x0: Withdraw an update mid-transfer
chrysh Oct 9, 2026
f604054
target/ast10x0: Offer an update to a RoT that supervises nothing
chrysh Oct 9, 2026
0e2ff6c
Rename the mock BMC to the mock downstream device
chrysh Oct 10, 2026
08763db
pldm: Let the firmware device wait out its retry timer
chrysh Oct 8, 2026
559ad14
target/ast10x0: Ask for one chunk twice in the update scenario
chrysh Oct 8, 2026
4519639
target/ast10x0: Drop the commit timeout from the gap list
chrysh Oct 12, 2026
e86ec87
third_party: Bump ruzstd to 0.9.1
chrysh Oct 13, 2026
8d11468
capabilities: Implement SelfUpdate for ()
chrysh Oct 10, 2026
810e7de
driver: Give the board the eRoT's own update session
chrysh Oct 10, 2026
a91fd3c
orchestrator: Add the eRoT itself as a component kind
chrysh Oct 12, 2026
2dd3014
sm: Pass over the eRoT in the chain walk
chrysh Oct 13, 2026
3577376
target/ast10x0: Put the eRoT at chain position 0
chrysh Oct 13, 2026
4428f35
config: Require the eRoT at chain position 0
chrysh Oct 13, 2026
1327d39
target/ast10x0: Count the device table instead of restating it
chrysh Oct 13, 2026
59b4096
driver: Give XorVerifier named constructors
chrysh Oct 13, 2026
b9e4b7d
driver: Tell a running trial from a floor advance that is waiting
chrysh Oct 13, 2026
71d080c
orchestrator: Confirm the eRoT's own trial once the chain is up
chrysh Oct 13, 2026
ba94edd
driver: Settle the eRoT's own session from the board
chrysh Oct 13, 2026
a939d83
sm: Say where the eRoT's anti-rollback floor lives
chrysh Oct 13, 2026
5aba1c7
driver: Settle the eRoT's own session at the start of a boot
chrysh Oct 13, 2026
3af925e
target/ast10x0: Give the boot board the eRoT's own session
chrysh Oct 13, 2026
c3df811
target/ast10x0: Boot the trial image of the eRoT's own update
chrysh Oct 13, 2026
bffec2c
docs: Record the self-trial verdict in the machine model
chrysh Oct 13, 2026
de7e09b
driver: Give each component one wiring struct
chrysh Oct 13, 2026
f4a1cbc
driver: Make each wiring name its own device table entry
chrysh Oct 13, 2026
5240ab8
docs: How the eRoT's own firmware is verified
chrysh Oct 13, 2026
f6f1970
sm: Move the floor advance to the update agent's ask
chrysh Oct 13, 2026
4ffd92f
target/ast10x0: Ask for the floor advance in the update scenarios
chrysh Oct 13, 2026
33b0e3e
docs: Record the floor advance waiting for the agent
chrysh Oct 13, 2026
655211b
capabilities: Report the SVN an authenticated candidate carries
chrysh Oct 14, 2026
6d74209
capabilities: One shape for what an authenticated pass learned
chrysh Oct 15, 2026
10a9cfe
capabilities: Let an authenticated pass say it read no SVN
chrysh Oct 15, 2026
31cda3c
target/ast10x0: One copy of the scenario board stubs
chrysh Oct 15, 2026
3b49ee1
third_party: Bump pldm-lib to the UpdateSecurityRevision responder
chrysh Oct 10, 2026
2b9fab1
target/ast10x0: Commit the security revision after an update
chrysh Oct 10, 2026
e5dfc7c
target/ast10x0: Answer the security revision request from the machine
chrysh Oct 15, 2026
1e1eb0a
target/ast10x0: One clock helper for the test apps
chrysh Oct 15, 2026
218e12e
target/ast10x0: One copy of what both firmware devices answer alike
chrysh Oct 15, 2026
47fea9a
target/ast10x0: List every scenario's single-run command
chrysh Oct 15, 2026
fde78ab
target/ast10x0: Record the events and resets no scenario reaches
chrysh Oct 15, 2026
aa205ca
target/ast10x0: Ask again when the security revision times out
chrysh Oct 19, 2026
afb6ac3
services/pldm: Say that the listener gap between calls stays
chrysh Oct 19, 2026
7120ec6
target/ast10x0: Rename cc to completion_code in the update agents
chrysh Oct 19, 2026
6172304
target/ast10x0: Refuse a second commit of the same revision
chrysh Oct 15, 2026
d7aa3de
target/ast10x0: Tell a lost answer from a refusal
chrysh Oct 19, 2026
d7c07c8
target/ast10x0: Run the measurement path
chrysh Oct 19, 2026
5c5bdfd
pldm_update: One helper for the agent's request-and-check
chrysh Oct 20, 2026
e89b51f
target/ast10x0: Stop counting scenarios in the itest docs
chrysh Oct 20, 2026
2082af0
capabilities: Add the SelfReset capability
chrysh Oct 19, 2026
9fb3069
target/ast10x0: Recover a device whose image the RoT refuses
chrysh Oct 15, 2026
0479edf
driver: Activate the eRoT's own update
chrysh Oct 19, 2026
72e14b0
driver: Move the eRoT's own update into its own file
chrysh Oct 20, 2026
1df08fb
orchestrator: Confirm or revert the trial image of a device
chrysh Oct 20, 2026
165bf47
sm: Refuse an update for another component while a trial is open
chrysh Oct 20, 2026
8444d69
sm: Say component instead of sibling in the recovery tests
chrysh Oct 20, 2026
715c28d
capabilities: Say pending marker instead of disarm
chrysh Oct 20, 2026
94126d9
target/ast10x0: Cast the trial counts pw_log asks for
chrysh Oct 20, 2026
aed181c
driver: Refuse what the eRoT cannot self-update
chrysh Oct 19, 2026
52f281e
orchestrator: Name the eRoT update struct for what it holds
chrysh Oct 21, 2026
17963ed
sm: Say which refusals reach the effect trace
chrysh Oct 21, 2026
12248d5
target/ast10x0: Say why the slot re-sync cannot run here
chrysh Oct 19, 2026
fa3a7d3
docs: Call the recovery effect RecoverComponent
chrysh Oct 20, 2026
8af3063
orchestrator: Retire the update intake seam
chrysh Oct 19, 2026
d82eada
orchestrator: Commit the eRoT's own floor through the same effect
chrysh Oct 15, 2026
aa8a23b
target/ast10x0: Record what a failed spare-slot re-sync leaves behind
chrysh Oct 21, 2026
5ea041c
orchestrator: State the rationale instead of citing the CSA
chrysh Oct 21, 2026
a7745ad
target/ast10x0: Stage and activate the eRoT's own update
chrysh Oct 15, 2026
c7b3af1
target/ast10x0: Run recovery preempting an update
chrysh Oct 19, 2026
f8b6c6a
target/ast10x0: Fix the wording in the preemption scenario
chrysh Oct 22, 2026
1f2aa56
target/ast10x0: Sweep the full_update comments for style
chrysh Oct 22, 2026
bf9a770
target/ast10x0: Write SVN floor, not floor
chrysh Oct 22, 2026
7b30c72
target/ast10x0: Say what the scenario constants guard
chrysh Oct 22, 2026
faf6f26
sm: Clear awaiting_floor_advance only when the id matches
chrysh Oct 21, 2026
0e2a0c1
sm: Test the request, not the event that records it
chrysh Oct 21, 2026
132b534
pldm: capitalize Update Agent and Firmware Device in prose
chrysh Oct 22, 2026
d6c5bea
util/service: move the traits out of lib.rs into transport and dispatch
chrysh Oct 23, 2026
02ffb78
util/io: rename ByteWindow::new to try_new
chrysh Oct 23, 2026
18523db
sm: import the model items rot.rs uses by name
chrysh Oct 23, 2026
f7272a6
services: Add the OCP recovery register layouts
chrysh Oct 28, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
7 changes: 7 additions & 0 deletions .bazelrc
Original file line number Diff line number Diff line change
Expand Up @@ -104,6 +104,13 @@ test:stress_k_ast1060_evb --run_under="//target/ast10x0/harness:test_runner \
--timeout 0 "
test:stress_k_ast1060_evb --test_env=AST1060_EVB_PI_HOST

# ── VeeR RISC-V kernel target ──
# Platform setup for building //target/veer/* artifacts.
common:k_veer --platforms=//target/veer:veer
build:k_veer --build_tag_filters=-do_not_build,-kernel_doc_test
# Usage: VCK190_FPGA_HOST=<host> bazel test --config=k_veer //target/veer/unittest_runner:fpga_test
test:k_veer --test_env=VCK190_FPGA_HOST

# Site-local overrides. Kept last so developer machine-specific settings (e.g. a
# local QEMU source override via --override_repository) take precedence. This
# file is not checked in; see third_party/qemu/setup.md.
Expand Down
2 changes: 1 addition & 1 deletion .github/workflows/ci.yml
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@ on:
push:
branches: [ main, develop ]
pull_request:
branches: [ main, earlgrey-hwe ]
branches: [ main, earlgrey-hwe, ocp-global-demo-wip ]

env:
CARGO_TERM_COLOR: always
Expand Down
9 changes: 2 additions & 7 deletions MODULE.bazel
Original file line number Diff line number Diff line change
Expand Up @@ -22,14 +22,9 @@ local_path_override(

git_override(
module_name = "pigweed",
commit = "f540ae6c33424230ed2d388599efa89b6baacf80",
commit = "294d431a761dc44fe07710c15d5ac1603f7ef7bf",
patch_args = ["-p1"],
patches = [
# Program the VeeR external-interrupt redirect table (MEIVT) in
# veer_pic early_init; without it the first external interrupt
# vectors through an unprogrammed table to address 0.
"//third_party/pigweed:veer_pic_meivt.patch",
],
patches = ["//patches:pigweed_cortex_m_systick_double_count.patch"],
remote = "https://pigweed.googlesource.com/pigweed/pigweed",
)

Expand Down
1,450 changes: 512 additions & 938 deletions MODULE.bazel.lock

Large diffs are not rendered by default.

3 changes: 3 additions & 0 deletions docs/src/SUMMARY.md
Original file line number Diff line number Diff line change
Expand Up @@ -32,6 +32,9 @@
* [Design](./design/README.md)
* [Pigweed Integration Overview](./design/pigweed-overview.md)
* [pw_kernel IPC](./design/pw-kernel-ipc.md)
* [IPC service stack](./design/ipc-service-stack.md)
* [Trust Boundaries](./design/trust-boundaries.md)
* [eRoT Self-Verification](./design/erot-self-verification.md)
* [Orchestrator](./design/orchestrator/orchestrator-overview.md)
* [Verification Model](./design/orchestrator/orchestrator-model.md)
* [State Machine](./design/orchestrator/orchestrator-machine.md)
Expand Down
138 changes: 138 additions & 0 deletions docs/src/design/erot-self-verification.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,138 @@
# How the eRoT's own firmware is verified

The orchestrator verifies every component it supervises: it reads the
firmware, verifies it at rest, releases the reset line and watches the boot
walk. It does none of that for the eRoT it runs on.

That is not a gap. The orchestrator is the thing that got booted, so it cannot
be the thing that decides whether booting it was safe. By the time any of this
code runs, the answer is already settled.

## Two different trust relationships

Governing other components is the orchestrator's job. It walks the chain,
reading and verifying each component before releasing it, and the
[state machine](orchestrator/orchestrator-machine.md) owns every step.

Self-governance happens one boot layer down, before the orchestrator exists.
The eRoT ROM verifies the bootloader, the bootloader verifies and measures the
runtime firmware, and only then does this code start. The orchestrator
receives a verdict it did not compute.

```mermaid
flowchart TB
subgraph below["Below the orchestrator, before it runs"]
direction TB
rom["eRoT ROM<br/>immutable, the hardware root"]
bl["Measuring bootloader<br/>verified by the ROM"]
rt["eRoT runtime firmware<br/>verified and measured by the bootloader"]
rom -->|verifies| bl
bl -->|verifies, measures| rt
end

rt -->|"Event::PowerGood(PowerOnResult)"| sm

subgraph sm["Orchestrator state machine"]
direction TB
por["PowerOnReset"]
pre["PreSupervision"]
locked["Locked"]
por -->|Provisioned| pre
por -->|"Unprovisioned<br/>cannot act as a RoT"| locked
por -->|"SelfVerificationFailed<br/>INV11: latches at once"| locked
end

pre --> walk["Chain walk over the components the eRoT supervises:<br/>ReadFirmware, VerifyFirmware, ReleaseReset<br/>(steps past ComponentId(0), the eRoT itself)"]
```

## What the orchestrator does with the verdict

The board hands it in as `PowerOnResult` inside `Event::PowerGood`, the first
event the machine ever sees. There are three answers:

| `PowerOnResult` | meaning | where it goes |
|---|---|---|
| `Provisioned` | self-verified, has an identity | `PreSupervision` |
| `Unprovisioned` | self-verified, but cannot act as a RoT | `Locked` |
| `SelfVerificationFailed` | the layer below said no | `Locked` |

`SelfVerificationFailed` latches `Locked` immediately, without passing through
`PreSupervision`. That is invariant INV11, pinned by
`self_verification_failure_latches_immediately`.

Producing the value is platform code's job. Nothing in this repository
constructs a `PowerOnResult` outside tests: the board's bring-up reads what
the ROM and the bootloader recorded (a status register, say, or the fact that
a fallback image is running) and names one of the three. Where the ROM
refuses to start an image that failed verification, `SelfVerificationFailed`
never arrives, because nothing is running to send it.

Refusing to run is the only move available. The orchestrator cannot re-verify
itself, so it cannot recover from a bad verdict, and a platform whose own RoT
is suspect has nothing left to supervise with.

## Why the walk skips chain position 0

The eRoT holds `ComponentId(0)` in every device table, so the orchestrator can
address it the way it addresses anything else. The walk steps past it: no
`ReadFirmware`, no `VerifyFirmware`, no `ReleaseReset`. There is nothing to
read or verify at rest, because that already happened, and nothing to release,
because the eRoT is what drives the reset lines.

The test `the_walk_passes_over_the_erot` pins both halves: a power-on walk
emits none of those three effects for the eRoT, and still reaches the
component behind it.

## The eRoT's own reset

A self-update activates a new image and resets the eRoT into it. The driver
does the sequence: it records the session with the SVN the verifier read off
the candidate, marks the trial pending, then asks the board's `SelfReset` to
reset.

The `SelfUpdate` session is durable: it survives the reset, and the next boot
reads it. NIST SP 800-193 §4.1.2(2) requires a mutable Root of Trust for
Update to stay operational or recoverable across the reboot its own update
causes, including after a power loss in the middle of a flash write, and the
session is what the eRoT recovers from.

The reset belongs to the platform. The eRoT's device table entry carries no
reset signal, because nothing outside the eRoT can reset it.

## Where the SVN comes from

A managed component's floor takes the SVN read back after the component
reboots, not the SVN read off the candidate. The two can differ: the
candidate is what the agent offered, the read-back is what runs. The
floor takes the read-back because the orchestrator verified it.

The eRoT has only the candidate's SVN. It could read its own flash, but
then the running image would vouch for itself, so the orchestrator never
verifies itself. The ROM and the bootloader verify it. They report a
verdict with no version: `PowerOnResult` is `Provisioned`, `Unprovisioned`
or `SelfVerificationFailed`.

`PollOutcome::Authenticated` carries the candidate's SVN. The activation
records it in the session. Confirming the trial does not move the floor.
The commit is a separate step, and it takes the SVN from the session.

An eRoT image carries its SVN in the signed header, and the ROM compares
that SVN against the anti-rollback counter before booting. An eRoT image
without an SVN cannot be compared, so the eRoT cannot install it. That is
why `SelfUpdate::prepare` takes an `Svn` and not an `Option<Svn>`: with no
SVN there is no session and no trial to mark, so the update is rejected.

A managed component needs no session, because the walk judges it after it
reboots. The image installs even with no SVN, and the anti-rollback floor
for that component stays where it was, which `Report::SvnFloorHeld`
records.

## Not covered here

The ROM and bootloader are platform firmware, outside this repository. How a
given platform verifies each layer, and where it keeps the keys to do so, is
that platform's business. This page only says what the orchestrator may assume
about the result and what it does when the result is bad.

Device identity and attestation key derivation are specified in
[Attestation](../specification/services/attestation.md).
129 changes: 129 additions & 0 deletions docs/src/design/ipc-service-stack.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,129 @@
# IPC service stack

How a service request travels from a client, through the kernel, to a
server and back. Two paths: the production kernel channel path and the
in-process loopback for host tests.

## Production path (kernel channel)

```mermaid
sequenceDiagram
participant C as Client<br/>(event loop)
participant CT as AsyncChannelTransport<br/>(util/ipc)
participant AT as AsyncTransaction<br/>(util/ipc)
participant K as pw_kernel<br/>(channel)
participant S as Server process
participant D as dispatch()<br/>(pldm/server)
participant H as FdIpcHandler

Note over C,H: Request

C->>CT: start(req: &[u8])
CT->>CT: copy req into 'static send buf
CT->>AT: start(send, send_len, recv)
AT->>K: channel_async_transact(send_ptr, recv_ptr)
K-->>C: returns immediately

Note over C: event loop parks on<br/>WaitGroup (READABLE)

K->>S: delivers request bytes
S->>D: dispatch(request, response)
D->>H: accept_offer() / query_status() / ...
H-->>D: Ok(()) or Err(ResponseCode)
D->>D: encode response frame
D-->>S: response length
S->>K: channel_respond(response)

Note over C,H: Response

K-->>C: READABLE signal fires
C->>CT: poll(resp: &mut [u8])
CT->>AT: try_recv()
AT->>K: channel_async_transact_complete()
K-->>AT: Ok(len)
AT-->>CT: Completion { len, send, recv }
CT->>CT: copy recv[..len] into resp
CT-->>C: Ok(Some(len))
C->>C: decode resp[..len]
```

## Loopback path (host tests)

A loopback answers on the first poll, so it cannot test a client's
not-ready path. `Delayed` wraps any transport and returns `Ok(None)` for a
set number of polls before forwarding, which is how a host test reaches that
path.

```mermaid
sequenceDiagram
participant C as Client<br/>(test code)
participant L as Loopback<br/>(util/service)
participant D as dispatch()<br/>(pldm/server)
participant H as FdIpcHandler

C->>L: start(req: &[u8])
L->>D: dispatch(req, held_buf)
D->>H: accept_offer() / query_status() / ...
H-->>D: Ok(()) or Err(ResponseCode)
D-->>L: response length
Note over L: response stored in held_buf,<br/>ready immediately

C->>L: poll(resp: &mut [u8])
L->>L: copy held_buf[..len] into resp
L-->>C: Ok(Some(len))
C->>C: decode resp[..len]
```

## Layer map

```mermaid
graph TD
subgraph "Client process"
CL[Service client code]
CT[AsyncChannelTransport]
AT[AsyncTransaction]
end

subgraph "Kernel"
CH[pw_kernel channel]
end

subgraph "Server process"
SV[Server main loop]
DS[service dispatch]
FH["handler (e.g. FdIpcHandler)"]
end

subgraph "Traits (util/service)"
TR["AsyncTransport trait"]
DI["Dispatch trait"]
end

CL -->|"start/poll/cancel"| CT
CT -->|"implements"| TR
CT -->|"owns 'static bufs"| AT
AT -->|"unsafe syscalls"| CH
CH -->|"delivers frames"| SV
SV -->|"request bytes"| DS
DS -->|"implements"| DI
DS -->|"decoded opcodes"| FH

style TR fill:#f0f0f0,stroke:#666
style DI fill:#f0f0f0,stroke:#666
```

## Buffer copies

The async path copies twice per direction:

| Step | What | Where |
|------|------|-------|
| 1 | req slice into 'static send buf | `AsyncChannelTransport::start()` |
| 2 | send buf to server process | kernel internal |
| 3 | server response to 'static recv buf | kernel internal |
| 4 | recv buf into caller's resp slice | `AsyncChannelTransport::poll()` |

Copies 1 and 4 exist because the kernel needs `'static` buffers (it holds
raw pointers for the duration of the transaction), but callers work with
ordinary stack slices. The loopback skips steps 2 and 3 since everything
runs in one process.
Loading
Loading