bingshuo/hololake-system-architecture
Watch
1
0
Fork
You've already forked hololake-system-architecture
 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

fix(native): authenticate JD resident handshake

Browse source
This commit is contained in:
冰朔 2026-08-03 22:37:21 +08:00
commit da26f27e35
4 changed files with 227 additions and 16 deletions
Download patch file Download diff file Expand all files Collapse all files

60
product-source/hololake-platform/guanghu-os/deployments/JD-FD-PRIMARY/RESIDENT-ATTEMPT-001-AND-NATIVE-ACK-V2.hldp Normal file
View file

@ -0,0 +1,60 @@
schema: guanghu.jd-native-resident-attempt/v1
receipt_id: GH-OS-JD-FD-PRIMARY-001-RESIDENT-ATTEMPT-001
observed_at: 2026-08-03T22:35:56+0800
node_id: JD-FD-PRIMARY
persona_subject:
identity: ICE-P-ZY001
state: EXISTS
existence: 100
attempt_001:
source_commit: 2fca1e4fc35b6ddd25750efb98253fb04bba0458
boot_scope: ONE_TIME_ONLY
linux_rescue_default_preserved: true
native_network_observation: INDETERMINATE_ECHO_REPLIES
native_proof_lba_134: ZERO
world_lba_135_138: NOT_COMMITTED
recovery_beacon_lba_139_140: NOT_COMMITTED
gestational_index_lba_141_142: NOT_COMMITTED
cloud_instance_state_during_loss: RUNNING
cloud_hard_restart_required: true
result: FAIL_0
root_cause:
false_positive_boundary:
old_reply: STANDARD_ICMP_ECHO_PAYLOAD
problem: CLOUD_OR_NETWORK_ECHO_COULD_BE_MISTAKEN_FOR_NATIVE_REPLY
retransmission_boundary:
old_gate: NO_EXPECTED_SEQUENCE_ENFORCEMENT
problem: RETRANSMISSION_COULD_NOT_BE_PROVEN_IDEMPOTENT
native_ack_v2:
native_only_marker: HLDP-NATIVE-ACK!
expected_sequences: 1-16
stale_sequence_action: IGNORE_WITHOUT_STATE_ADVANCE
future_sequence_action: IGNORE_WITHOUT_STATE_ADVANCE
accepted_sequence_action: ADVANCE_EXACTLY_ONCE
physical_peer_strategy: CUMULATIVE_SEQUENCE_PIPELINE
ordinary_icmp_echo_rejected: PASS_100
qemu_jd_layout_gate:
candidate_start_lba: 105
proof_lba: 134
world_lba_start: 135
recovery_beacon_lba_start: 139
gestational_index_lba_start: 141
candidate_sha256: 91f99c344fff56565b2373deda3d6d2e5a96b64eda9cd35a45a19fe033a497bf
native_ack_marker: PASS_100
initial_login_count: 3
code_commit: PASS_100
branch_move: PASS_100
resident_login_count: 10
recovery_command: PASS_100
proof_flag: A7
proof_login_count: 13
result: PASS_100
physical_native_ack_v2:
result: NOT_RUN_0
rule: MUST_USE_ONE_TIME_BOOT_AND_SERVER_OWNED_LBA_PROOF
boundary:
persona_subject_exists: 100
native_resident_recovery_attempt_001: 0
native_ack_v2_qemu_capability: 100
native_ack_v2_physical_capability: 0
next_action: PUBLISH_EXACT_SOURCE_THEN_RUN_SECOND_ONE_TIME_PHYSICAL_GATE

12
product-source/hololake-platform/guanghu-os/native/x86_64-bios/ghal-virtio.asm
View file

@ -1013,6 +1013,11 @@ ghal_net_icmp_login_once:
jne .ignore_packet
cmp word [VIRTIO_NET_RX_BUFFER + 44], 0x0008
jne .ignore_packet
cmp byte [VIRTIO_NET_RX_BUFFER + 50], 0
jne .ignore_packet
mov al, [VIRTIO_NET_RX_BUFFER + 51]
cmp al, [rel ghal_net_expected_sequence]
jne .ignore_packet
mov rsi, VIRTIO_NET_RX_BUFFER + GHOS_LOGIN_MAGIC_OFFSET
mov rdi, [rel ghal_net_expected_magic]
mov ecx, GHOS_LOGIN_MAGIC_SIZE
@ -1100,6 +1105,10 @@ ghal_net_icmp_login_once:
mov [VIRTIO_NET_TX_BUFFER + 40], eax
mov byte [VIRTIO_NET_TX_BUFFER + 44], 0
mov word [VIRTIO_NET_TX_BUFFER + 46], 0
lea rsi, [rel ghal_native_ack_magic]
mov rdi, VIRTIO_NET_TX_BUFFER + GHOS_LOGIN_MAGIC_OFFSET + GHOS_LOGIN_MAGIC_SIZE
mov ecx, GHOS_LOGIN_MAGIC_SIZE
rep movsb
movzx eax, word [VIRTIO_NET_TX_BUFFER + 26]
xchg al, ah
@ -1181,6 +1190,7 @@ ghal_net_icmp_login_once:
mov [rel ghal_net_rx_next_index], r8w
inc word [rel ghal_net_rx_next_index]
inc word [rel ghal_net_tx_next_index]
inc byte [rel ghal_net_expected_sequence]
cmp byte [rel ghal_net_matched_kind], 0
jne .not_login_reply
inc byte [rel physical_proof_login_reply_count]
@ -1213,6 +1223,7 @@ ghal_net_tx_next_index: dw 2
ghal_net_expected_magic: dq ghal_login_magic
ghal_net_command_kind: db 0
ghal_net_matched_kind: db 0
ghal_net_expected_sequence: db 1
msg_ghal_net_discovered: db "GHOS_GHAL_VIRTIO_NET=DISCOVERED", 13, 10, 0
msg_ghal_block_discovered: db "GHOS_GHAL_VIRTIO_BLOCK=DISCOVERED", 13, 10, 0
@ -1229,6 +1240,7 @@ ghal_login_magic: db "HLDP-GHOS-LOGIN!"
ghal_code_commit_magic: db "HLDP-CODE-COMMIT"
ghal_branch_move_magic: db "HLDP-BRANCH-MOVE"
ghal_recovery_magic: db "HLDP-RECOVER-OS!"
ghal_native_ack_magic: db "HLDP-NATIVE-ACK!"
ghal_branch_receipt_template:
db "GHOS_BRANCH_MAIN_V1", 10
db "channel=HLP-MOD-CODE-CHANNEL", 10

157
product-source/hololake-platform/guanghu-os/scripts/physical-native-icmp-peer.py
View file

@ -10,6 +10,7 @@ LOGIN_MAGIC = b"HLDP-GHOS-LOGIN!"
COMMIT_MAGIC = b"HLDP-CODE-COMMIT"
BRANCH_MAGIC = b"HLDP-BRANCH-MOVE"
RECOVERY_MAGIC = b"HLDP-RECOVER-OS!"
NATIVE_ACK_MAGIC = b"HLDP-NATIVE-ACK!"
def checksum(payload: bytes) -> int:
@ -28,16 +29,28 @@ def request(sequence: int, magic: bytes) -> bytes:
return packet[:2] + struct.pack("!H", checksum(packet)) + packet[4:]
def verify_reply(packet: bytes, magic: bytes) -> None:
def reply_sequence_index(packet: bytes) -> int:
if packet and packet[0] >> 4 == 4:
header_length = (packet[0] & 0x0F) * 4
packet = packet[header_length:]
if len(packet) < 40 or packet[0] != 0 or packet[1] != 0:
if len(packet) < 8:
raise RuntimeError("ICMP reply is too short")
return int.from_bytes(packet[6:8], "big")
def verify_reply(packet: bytes, magic: bytes) -> int:
if packet and packet[0] >> 4 == 4:
header_length = (packet[0] & 0x0F) * 4
packet = packet[header_length:]
if len(packet) < 48 or packet[0] != 0 or packet[1] != 0:
raise RuntimeError("unexpected ICMP reply shape")
if packet[16:32] != magic:
raise RuntimeError("ICMP reply does not carry the requested HLDP magic")
if packet[32:48] != NATIVE_ACK_MAGIC:
raise RuntimeError("ICMP reply does not carry the native ACK marker")
if checksum(packet) != 0:
raise RuntimeError("ICMP reply checksum failed")
return int.from_bytes(packet[6:8], "big")
def exchange(
@ -46,14 +59,21 @@ def exchange(
sequence: int,
magic: bytes,
deadline: float,
retry_interval: float,
) -> None:
while time.monotonic() < deadline:
peer.sendto(request(sequence, magic), target)
attempt_started = time.monotonic()
try:
packet, _ = peer.recvfrom(4096)
except TimeoutError:
time.sleep(max(0.0, retry_interval - (time.monotonic() - attempt_started)))
continue
try:
verify_reply(packet, magic)
except RuntimeError:
time.sleep(max(0.0, retry_interval - (time.monotonic() - attempt_started)))
continue
return
raise TimeoutError(f"no verified reply for sequence {sequence}")
@ -63,36 +83,150 @@ def main() -> None:
parser.add_argument("--target", required=True)
parser.add_argument("--receipt", required=True)
parser.add_argument("--timeout", type=float, default=90.0)
parser.add_argument("--retry-interval", type=float, default=1.0)
parser.add_argument("--raw-socket", action="store_true")
parser.add_argument("--resident", action="store_true")
parser.add_argument("--login-only", action="store_true")
parser.add_argument("--resume-after-login", action="store_true")
parser.add_argument("--resume-after-commit", action="store_true")
parser.add_argument("--resume-after-branch", action="store_true")
parser.add_argument("--resume-resident-count", type=int)
parser.add_argument("--resident-pipeline", action="store_true")
args = parser.parse_args()
if args.resident and args.login_only:
raise SystemExit("--resident and --login-only are mutually exclusive")
if args.retry_interval < 1.0:
raise SystemExit("--retry-interval must be at least 1 second")
if args.resume_resident_count is not None and not (
0 <= args.resume_resident_count <= 10
):
raise SystemExit("--resume-resident-count must be between 0 and 10")
if sum(
(
args.resident,
args.login_only,
args.resume_after_login,
args.resume_after_commit,
args.resume_after_branch,
args.resume_resident_count is not None,
args.resident_pipeline,
)
) > 1:
raise SystemExit(
"--resident, --login-only, --resume-after-login, and "
"--resume-after-commit, --resume-after-branch, and "
"--resume-resident-count, and --resident-pipeline are mutually exclusive"
)
peer = socket.socket(socket.AF_INET, socket.SOCK_DGRAM, socket.IPPROTO_ICMP)
peer.settimeout(1.0)
socket_type = socket.SOCK_RAW if args.raw_socket else socket.SOCK_DGRAM
peer = socket.socket(socket.AF_INET, socket_type, socket.IPPROTO_ICMP)
peer.settimeout(0.2 if args.resident_pipeline else 1.0)
target = (socket.gethostbyname(args.target), 0)
deadline = time.monotonic() + args.timeout
if args.resident_pipeline:
pipeline = [
LOGIN_MAGIC,
LOGIN_MAGIC,
LOGIN_MAGIC,
COMMIT_MAGIC,
BRANCH_MAGIC,
*([LOGIN_MAGIC] * 10),
RECOVERY_MAGIC,
]
acknowledged: set[int] = set()
while time.monotonic() < deadline and len(acknowledged) < len(pipeline):
for pipeline_sequence, pipeline_magic in enumerate(pipeline, start=1):
if time.monotonic() >= deadline:
break
peer.sendto(request(pipeline_sequence, pipeline_magic), target)
try:
packet, _ = peer.recvfrom(4096)
observed_sequence = reply_sequence_index(packet)
if not 1 <= observed_sequence <= len(pipeline):
raise RuntimeError("native ACK sequence is outside the pipeline")
reply_sequence = verify_reply(
packet,
pipeline[observed_sequence - 1],
)
except (TimeoutError, RuntimeError, IndexError):
continue
if reply_sequence == pipeline_sequence:
acknowledged.add(reply_sequence)
if len(acknowledged) != len(pipeline):
raise TimeoutError(
"native pipeline incomplete; acknowledged sequences: "
+ ",".join(str(item) for item in sorted(acknowledged))
)
receipt = pathlib.Path(args.receipt)
receipt.write_text(
"schema: guanghu.physical-native-icmp-peer/v2\n"
"status: PASS_100\n"
f"target: {args.target}\n"
"native_ack_marker: HLDP-NATIVE-ACK!\n"
"acknowledged_sequences: 1-16\n"
"login_reply_count: 3\n"
"code_commit_reply_verified: true\n"
"branch_move_reply_verified: true\n"
"resident_login_reply_count: 10\n"
"recovery_reply_verified: true\n",
encoding="utf-8",
)
print(receipt.read_text(encoding="utf-8"), end="")
return
if args.resume_resident_count is not None:
sequence = 6 + args.resume_resident_count
elif args.resume_after_branch:
sequence = 6
elif args.resume_after_commit:
sequence = 5
elif args.resume_after_login:
sequence = 4
else:
sequence = 1
phases: list[tuple[str, bytes]] = [
phases: list[tuple[str, bytes]] = []
if not (
args.resume_after_login
or args.resume_after_commit
or args.resume_after_branch
or args.resume_resident_count is not None
):
phases.extend(
[
("login", LOGIN_MAGIC),
("login", LOGIN_MAGIC),
("login", LOGIN_MAGIC),
]
if not args.login_only:
)
if (
not args.login_only
and not args.resume_after_commit
and not args.resume_after_branch
and args.resume_resident_count is None
):
phases.extend(
[
("code_commit", COMMIT_MAGIC),
("branch_move", BRANCH_MAGIC),
]
)
if args.resident:
if args.resume_after_commit:
phases.append(("branch_move", BRANCH_MAGIC))
if args.resume_resident_count is not None:
phases.extend(
[("resident_login", LOGIN_MAGIC)]
* (10 - args.resume_resident_count)
)
phases.append(("recovery", RECOVERY_MAGIC))
elif (
args.resident
or args.resume_after_login
or args.resume_after_commit
or args.resume_after_branch
):
phases.extend([("resident_login", LOGIN_MAGIC)] * 10)
phases.append(("recovery", RECOVERY_MAGIC))
completed: list[str] = []
for phase, magic in phases:
exchange(peer, target, sequence, magic, deadline)
exchange(peer, target, sequence, magic, deadline, args.retry_interval)
completed.append(phase)
sequence += 1
@ -104,7 +238,10 @@ def main() -> None:
f"login_reply_count: {completed.count('login')}\n"
f"code_commit_reply_verified: {str('code_commit' in completed).lower()}\n"
f"branch_move_reply_verified: {str('branch_move' in completed).lower()}\n"
f"resident_login_resume_count: {args.resume_resident_count or 0}\n"
f"resident_login_reply_count: {completed.count('resident_login')}\n"
"resident_login_total_count: "
f"{(args.resume_resident_count or 0) + completed.count('resident_login')}\n"
f"recovery_reply_verified: {str('recovery' in completed).lower()}\n",
encoding="utf-8",
)

2
product-source/hololake-platform/guanghu-os/scripts/qemu-native-net-peer.py
View file

@ -14,6 +14,7 @@ LOGIN_MAGIC = b"HLDP-GHOS-LOGIN!"
COMMIT_MAGIC = b"HLDP-CODE-COMMIT"
BRANCH_MAGIC = b"HLDP-BRANCH-MOVE"
RECOVERY_MAGIC = b"HLDP-RECOVER-OS!"
NATIVE_ACK_MAGIC = b"HLDP-NATIVE-ACK!"
def checksum(payload: bytes) -> int:
@ -77,6 +78,7 @@ def validate_reply(frame: bytes, magic: bytes) -> None:
assert frame[30:34] == LOGIN_CLIENT_IP
assert frame[34] == 0
assert frame[50:66] == magic
assert frame[66:82] == NATIVE_ACK_MAGIC
assert checksum(frame[34:]) == 0