The Supported Commands mask is no longer set by DualModeController as
the list of implemented commands. This allows for setting unimplemented
commands as supported for feature discovery when testing new profiles.

The Supported Command mask value is unmodified in this CL, only adding
some commands that were implemented but missing from the HCI command
mapping:

- HCI Set Event Mask Page 2
- HCI Add SCO Connection
- HCI LE Remote Connection Parameter Request Reply
- HCI LE Remote Connection Parameter Request Negative Reply

The ability to configure supported commands from the JSON config file
is temporarily disabled as unimplemented commands now raise FATAL
errors and configured values are likely to be incorrect...

Test: m root-canal && atest pts-bot
Bug: 266087760
Change-Id: I9c8960ba65d885f3606a89d2dcc0781aa599c71f

There existed an implicit assumption that width, value and
enum_id, tag_id were mutually exclusive in the Field::Fixed
case.
This change makes the assumption explicit by modiyfing
the AST to add specialized cases.

Test: cargo build && cargo test
Change-Id: Ib51f61ad2cd39f8686ad9d94a6699cf2b3b318e7

Test: cargo build && cargo test
Change-Id: Ibede334254e6faa7920139a68747792e87ed05a3

Use the CommandComplete builder to simplify the code.

Test: m root-canal
Change-Id: I59e07c72761744bd77a142af21fc67e8c47066f1

The address is directly derived from the device identifier.
This change modifies TestModel to reuse identifiers from
disconnected devices.

The option is enabled by default and disabled by passing the
parameter --disable_address_reuse

Test: m root-canal && manual check
Bug: 253682053
Change-Id: Ie97763c1eae46df75d548a54b273cf8d363c7bb9

IncomingAclPacket took an illegal reference to
the end index of a vector.

Bug: 268037653
Test: m root-canal + manual validation
Change-Id: I8b431453256877f9cd173671e4eb7c46a29d57a8

DualModeController now distinguishes between
- supported commands: commands enabled in the supported commands mask
- implemented commands: commands for which an implementation is provided

Command are handled following this logic:
- implemented and supported command: registered handler invoked
- supported but not implemented command: raise an assert
- unsupported commands: send command complete which ERR_UNKNOWN_COMMAND

Test: m root-canal
Bug: 266087760
Change-Id: Ie930d368bd5c20d67fbc3a5f49f68763e12bfb50

Prevents "FATAL   | assertion 'HasHandle(handle)' failed - Unknown handle 47"

Tag: #compatibility
Test: ninja and nimble
Bug: 265583200

Change-Id: Ic4622ff5a17d21e86f4e3aea92e9a6007633fb49

Validate the set of supported commands according to the
constraints described in the section Vol 4, Part E § 3 Overview of
commands and events of the specification.

The check is non-fatal but displays instead the commands that
should be or should not be implemented following these rules.

Test: m root-canal
Change-Id: I12751149f77ba4d22b30280edbd6da1ade5ef7e9

model/controller/vendor_commands

The name vendor is reserved for external dependencies

Test: m root-canal
Change-Id: I005f57dbb8fdbc26cca9fc9129d8263a33607b19

- Replace ASSERT => LOG_ALWAYS_FATAL
- Replace include of os/log.h => log.h
- Iterate loop by reference

Test: m root-canal
Change-Id: I0b9978fcff2b4b9e6a50373d9f3f661fbc53ce06

Bug: 264630489
Test: atest --host rootcanal_ll_test
Change-Id: Ie674d0d7616fa82b2ecb60f574a729f2f68afd04

This is a zero-copy Rust backend that supports most of PDL (excluding
groups, checksums, custom types), BUT only little-endian packets.

This is needed for Rust-GATT (scoped for U). It is not a permanent
solution, I expect to unify / replace this with the official Rust PDL
once it is ready, but until that point arrives I don't want to remain
blocked.

Ignore-AOSP-First: Security bug
Bug: 255880936
Test: unit

Change-Id: I7d2305c20288d576aa1a5731b0d7e54727366947

Test: make a change to RootCanal and verify that the test is run.
Bug: 205758693
Change-Id: I27c4df211ef80f689819561a6b1ba7e2c34adb6d

Bug: 205758693
Test: atest VtsHalBluetoothTargetTest
Change-Id: I88830395a5fb31b2e6cf3be28375422ceddc75fa

Test: atest pdl_tests pdl_rust_generator_tests_{le,be}
Change-Id: Ie09c2fa09694f940e8584d56df3beb5019f91cf3

Test: atest pdl_tests pdl_rust_generator_tests_{le,be}
Change-Id: I6366b88f2b4683baf21e31519ccd981b6e81a380

The auto-generated “Packet” suffix now seems wrong, so I removed it.
This means that both “packet Foo” and “struct Foo” will generate the
following Rust types:

    pub struct Foo
    pub struct FooBuilder
    struct FooData

One potential problem with this: if someone has a “packet Packet”,
this now collides with “trait Packet” from our preamble. We might need
to correct for this when we integrate the generated code with clients.

Test: atest pdl_tests pdl_rust_generator_tests_{le,be}
Change-Id: I19354010e46ef1b5738fee9f7fb13ff7778d1b03

We will need access to both of them from the parser next.

Test: atest pdl_tests pdl_rust_generator_tests_{le,be}
Change-Id: I7cfd79f2ef811765cd7b3653b8ed1b0ddb14fc70

Being able to serialize data with Serde allows us to test enums using
the canonical test vectors: we can take the output of PDL and
serialize it to JSON, which we then compare with the canonical test
vectors.

While this was motivated by the need for testing, Serde support can be
useful in general so I think this is an overall nice change. The
support is behind an off-by-default feature flag, just in case people
don’t want the generated code.

Test: atest pdl_tests pdl_rust_generator_tests_{le,be}
Change-Id: I6a19a939df7b449d350dbcda8a7daa2cb96733a2

This adds support for parsing and serializing enum fields.

The canonical test vectors work.

Tests have been added to the canonical PDL files: they demonstrate
that we can generate valid Rust code. A test was added to show that we
need “#[repr(u64)]” for an enum with a maximum discriminant.

Test: atest pdl_tests pdl_rust_generator_tests_{le,be}
Change-Id: Id65bf831ddde8ade9780f89096842ef9516cf4c3

This CL reverts the code reading scalar fields to match the logic in
the Python backend. This approach will make it simpler to update the
two backends going forward.

Test: atest pdl_tests pdl_rust_generator_tests_{le,be}
Change-Id: I49ddb3dae8eab1bb00cf730eda50a051d326ce2c

The packet idc is not removed from the packet before being
sent to the command handler.

Bug: 261261502
Test: atest rootcanal_test_host
Change-Id: I9aa66e0f765c185405e002ae21d35ad784c9f9d8

The recovery state is entered when an invalid Idc
is received. In recovery state, all bytes are dropped
until a valid HCI Reset command is received.
The reasoning is that this issue occurs when the emulator
is rebooted, leaving bytes in the VirtIO console.
The next valid command should be HCI Reset command then.

Test: atest rootcanal_test_host
Bug: 261261502
Change-Id: I6e5482d9b2642fa789e6ed415ce3833663097bff

Bug: 264630489
Test: m root-canal
Change-Id: I1fff099890032e51aaaa087176df09ffb8ce3e68

The tasks registered within LinkLayerController capture
the pointer to the Controller instance for member access.
These tasks can outlive the Controller instance if the
controller is deleted (e.g. through the test channel, or
because the TCP connection is lost). In this case
attempting to execute the task callback would cause a
use-after-free invalid pointer access.

This change bypasses the issue by capturing a weak pointer
to the instance to check if it is live before running the
task callback.

Test: m root-canal && launch_cvd
Change-Id: I5384fc056fc4fbc0095b99ef7113fd4f706803ba

Test: m root-canal
Change-Id: I939c193680d8dd27e65b3cbc7e6d7ca878c8d5c5

Test: NA
Bug: 256013170
Change-Id: I02b2d11b0dd3ca0631e7a6514d43771b8a25306a

- LL/DDI/ADV/BV-15-C
- LL/DDI/ADV/BV-16-C
- LL/DDI/ADV/BV-17-C
- LL/DDI/ADV/BV-18-C
- LL/DDI/ADV/BV-19-C
- LL/DDI/ADV/BV-20-C
- LL/DDI/ADV/BV-21-C

Bug: 253523072
Test: atest rootcanal_ll_test
Change-Id: Id447dd5acf84cc2895ae2825cfa63e6db1f59e62

Test: manual
Bug: 264469240

Change-Id: I14115e04c959c3d3e4c15a075c997c2a7fb2f667

This reverts commit 9d76f0b7.

Reason for revert: breaks aosp_arm64-userdebug_tidy

Bug: 264648327
Change-Id: I7de1704c412ddb236456e9e6c44f074c704c0896

Bug: 260823636
Tag: #gd-refactor
Test: atest bluetooth_with_facades.test bt_topshim_facade.test pdl_tests
Change-Id: Id583d26459c05e3aaa8295c6f752923ec7d8fe93

Bug: 260823636
Tag: #gd-refactor
Test: atest pdl_tests
Change-Id: Ie9e9592718735f365b0780583b0a4915fe2a50bb

Test: ALLOW_LOCAL_TIDY_TRUE=1 m root-canal
Change-Id: Id840d82bc421e8ca6ad30e39fa9e8f501f6f73f1

Apply some modifications to improve the flakiness of tests
CanUnsubscribeTaskFromWithinTask and UnsubScribeWaitsUntilCompletion

Test: atest --host --rerun-until-failure 1000 -t AsyncManagerSocketTest rootcanal_test_host
Bug: 263327055
Change-Id: I43e41749cfd34a1e9ce2e5fe6e6eaf70f8951d84

Test: atest --host rootcanal_ll_test
Bug: 253523072
Change-Id: I56d5f58325af1e553c3416e8ac7f6ad6b7154ca1

This reverts commit 6e4f16af.

Reason for revert: Source of build breakage fixed in another CL

Test: NA
Change-Id: I0fba54ad0bbbbff2c7258f44719d88ed1c5f5e74

This reverts commit 5d597ba8.

Reason for revert: build breakage

Bug: 263527121
Change-Id: Ie2d6acd832a3fd17500dbd7b98348e176d4cb076

Test: build: m lib_rootcanal_python3
Change-Id: I9facfe33987d64397d25ee7c93cfaf45df860e95

size_t is unsigned, so we use %zu

Change-Id: Ic309393787b435c288df3bc29217db169e12b448
Test: Manual
Bug: 261188171