feat(deckies): generic file drops on fleet + MazeNET deckies
Extracts the docker-exec-with-base64-stdin pattern out of canary/planter and orchestrator/drivers/ssh into a shared decnet.decky_io package. Both consumers now delegate; the canary planter test still proves the contract end-to-end. Adds POST/DELETE /api/v1/deckies/files for arbitrary file drops. Container resolution is shared with the canary path: topology_id absent means fleet (<name>-ssh), present routes through resolve_decky_container which picks <name>-ssh when the topology decky exposes ssh, else the topology base container decnet_t_<id8>_<name>. Path validation rejects relative paths and '..' traversal at the request model layer. Bad base64 → 400; unknown topology → 404; decky not in topology → 422; docker exec failure → 409.
This commit is contained in:
@@ -20,11 +20,8 @@ shape but speaks bytes-via-base64 over the wire.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import asyncio
|
||||
import base64
|
||||
import os
|
||||
import shlex
|
||||
import time
|
||||
from datetime import datetime, timedelta, timezone
|
||||
from secrets import token_urlsafe
|
||||
from typing import Any, Iterable, Optional
|
||||
|
||||
@@ -34,13 +31,16 @@ from decnet.bus.factory import get_bus
|
||||
from decnet.canary.base import CanaryArtifact, CanaryContext
|
||||
from decnet.canary.factory import get_generator
|
||||
from decnet.canary.paths import default_path_for
|
||||
from decnet.decky_io import (
|
||||
delete_file_from_container,
|
||||
resolve_topology_container,
|
||||
write_file_to_container,
|
||||
)
|
||||
from decnet.logging import get_logger
|
||||
from decnet.web.db.repository import BaseRepository
|
||||
|
||||
log = get_logger("canary.planter")
|
||||
|
||||
_DOCKER = "docker"
|
||||
_TIMEOUT = 8.0
|
||||
# Container suffix — matches the orchestrator SSH driver's convention
|
||||
# (``<decky_name>-ssh``). Canary placement always happens through the
|
||||
# ssh container because every decky has one and it carries the most
|
||||
@@ -52,77 +52,16 @@ def _container_for(decky_name: str) -> str:
|
||||
return f"{decky_name}{_SSH_CONTAINER_SUFFIX}"
|
||||
|
||||
|
||||
def resolve_topology_container(
|
||||
topology_id: str, decky_name: str, services: Iterable[str],
|
||||
) -> str:
|
||||
"""Container name to docker-exec into for a MazeNET decky.
|
||||
|
||||
The ssh service container (when present) wins because it carries the
|
||||
most realistic filesystem layout — same rationale as the fleet path.
|
||||
Otherwise we target the base container, whose name is set by
|
||||
:func:`decnet.topology.compose._container_name`.
|
||||
"""
|
||||
if "ssh" in set(services):
|
||||
return f"{decky_name}{_SSH_CONTAINER_SUFFIX}"
|
||||
return f"decnet_t_{topology_id[:8]}_{decky_name}"
|
||||
|
||||
|
||||
def _dirname(path: str) -> str:
|
||||
idx = path.rfind("/")
|
||||
if idx <= 0:
|
||||
return "/"
|
||||
return path[:idx]
|
||||
|
||||
|
||||
async def _run(
|
||||
argv: list[str], *, stdin_bytes: Optional[bytes] = None,
|
||||
) -> tuple[int, str, str]:
|
||||
try:
|
||||
proc = await asyncio.create_subprocess_exec(
|
||||
*argv,
|
||||
stdin=asyncio.subprocess.PIPE if stdin_bytes is not None else None,
|
||||
stdout=asyncio.subprocess.PIPE,
|
||||
stderr=asyncio.subprocess.PIPE,
|
||||
)
|
||||
except FileNotFoundError as exc:
|
||||
return 127, "", f"argv[0] not found: {exc}"
|
||||
try:
|
||||
stdout, stderr = await asyncio.wait_for(
|
||||
proc.communicate(input=stdin_bytes), timeout=_TIMEOUT,
|
||||
)
|
||||
except asyncio.TimeoutError:
|
||||
try:
|
||||
proc.kill()
|
||||
except ProcessLookupError:
|
||||
pass
|
||||
return 124, "", "timeout"
|
||||
return (
|
||||
proc.returncode if proc.returncode is not None else -1,
|
||||
stdout.decode("utf-8", "replace"),
|
||||
stderr.decode("utf-8", "replace"),
|
||||
)
|
||||
|
||||
|
||||
def _build_plant_command(artifact: CanaryArtifact) -> tuple[str, bytes]:
|
||||
"""Compose the ``sh -c`` script + stdin payload for one artifact.
|
||||
|
||||
Binary safety: we base64-encode on the host and stream the result
|
||||
over stdin to ``base64 -d`` inside the container, so the bytes
|
||||
never touch the argv (kernel ARG_MAX would reject anything larger
|
||||
than ~128KB-2MB depending on the host). Both ``base64`` (coreutils)
|
||||
and ``touch -d @<unix_ts>`` are present on every Linux base image
|
||||
we ship, so there's no per-distro branching.
|
||||
"""
|
||||
encoded = base64.b64encode(artifact.content)
|
||||
mtime = int(time.time() + artifact.mtime_offset)
|
||||
mode_str = oct(artifact.mode)[2:]
|
||||
parts = [
|
||||
f"mkdir -p {shlex.quote(_dirname(artifact.path))}",
|
||||
f"base64 -d > {shlex.quote(artifact.path)}",
|
||||
f"chmod {mode_str} {shlex.quote(artifact.path)}",
|
||||
f"touch -d @{mtime} {shlex.quote(artifact.path)}",
|
||||
]
|
||||
return " && ".join(parts), encoded
|
||||
# resolve_topology_container is re-exported from decky_io for back-compat
|
||||
# with callers (tests, deploy hook) that imported it from this module
|
||||
# before the decky_io extraction.
|
||||
__all__ = [
|
||||
"plant",
|
||||
"revoke",
|
||||
"resolve_topology_container",
|
||||
"seed_baseline",
|
||||
"seed_baseline_topology",
|
||||
]
|
||||
|
||||
|
||||
async def _publish(
|
||||
@@ -173,14 +112,12 @@ async def plant(
|
||||
await repo.update_canary_token_state(token_uuid, "failed", err)
|
||||
return False, err
|
||||
|
||||
sh_cmd, stdin_payload = _build_plant_command(artifact)
|
||||
target_container = container or _container_for(decky_name)
|
||||
# ``-i`` keeps stdin attached so base64 -d inside the container can
|
||||
# consume the encoded payload streamed from the host.
|
||||
argv = [_DOCKER, "exec", "-i", target_container, "sh", "-c", sh_cmd]
|
||||
rc, _stdout, stderr = await _run(argv, stdin_bytes=stdin_payload)
|
||||
success = rc == 0
|
||||
error = None if success else (stderr.strip()[:256] or f"rc={rc}")
|
||||
mtime = datetime.now(timezone.utc) + timedelta(seconds=artifact.mtime_offset)
|
||||
success, error = await write_file_to_container(
|
||||
target_container, artifact.path, artifact.content,
|
||||
mode=artifact.mode, mtime=mtime,
|
||||
)
|
||||
|
||||
if repo is not None:
|
||||
if success:
|
||||
@@ -199,8 +136,8 @@ async def plant(
|
||||
|
||||
if not success:
|
||||
log.warning(
|
||||
"canary.plant failed decky=%s token=%s rc=%d stderr=%r",
|
||||
decky_name, token_uuid, rc, stderr[:120],
|
||||
"canary.plant failed decky=%s token=%s container=%s err=%r",
|
||||
decky_name, token_uuid, target_container, error,
|
||||
)
|
||||
return success, error
|
||||
|
||||
@@ -221,12 +158,10 @@ async def revoke(
|
||||
the file is gone after the call (whether we deleted it or it was
|
||||
already missing); only docker / container-down errors return False.
|
||||
"""
|
||||
sh_cmd = f"rm -f {shlex.quote(placement_path)}"
|
||||
target_container = container or _container_for(decky_name)
|
||||
argv = [_DOCKER, "exec", target_container, "sh", "-c", sh_cmd]
|
||||
rc, _stdout, stderr = await _run(argv)
|
||||
success = rc == 0
|
||||
error = None if success else (stderr.strip()[:256] or f"rc={rc}")
|
||||
success, error = await delete_file_from_container(
|
||||
target_container, placement_path,
|
||||
)
|
||||
|
||||
if repo is not None:
|
||||
await repo.update_canary_token_state(token_uuid, "revoked", error if not success else None)
|
||||
|
||||
39
decnet/decky_io/__init__.py
Normal file
39
decnet/decky_io/__init__.py
Normal file
@@ -0,0 +1,39 @@
|
||||
"""Shared primitives for writing/deleting files inside running deckies.
|
||||
|
||||
The canary planter and the orchestrator SSH driver both need to drop
|
||||
bytes into a decky container's filesystem, then sometimes unlink them.
|
||||
The ARG_MAX-safe ``base64 -d``-via-stdin trick lived in two places
|
||||
before this module existed.
|
||||
|
||||
Public API:
|
||||
|
||||
* :func:`write_file_to_container` — write bytes at a path, set mode,
|
||||
optionally backdate mtime.
|
||||
* :func:`delete_file_from_container` — best-effort ``rm -f``.
|
||||
* :func:`resolve_topology_container` — pick the right docker container
|
||||
for a MazeNET decky based on its services list.
|
||||
* :func:`resolve_decky_container` — async helper that takes
|
||||
``(decky_name, topology_id?)``, hydrates the topology when needed,
|
||||
and returns the docker container name.
|
||||
|
||||
Container resolution conventions are documented in
|
||||
:mod:`decnet.topology.compose`; we mirror them here without taking
|
||||
a runtime dependency on the compose generator.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
from .resolve import (
|
||||
resolve_decky_container,
|
||||
resolve_topology_container,
|
||||
)
|
||||
from .write import (
|
||||
delete_file_from_container,
|
||||
write_file_to_container,
|
||||
)
|
||||
|
||||
__all__ = [
|
||||
"delete_file_from_container",
|
||||
"resolve_decky_container",
|
||||
"resolve_topology_container",
|
||||
"write_file_to_container",
|
||||
]
|
||||
72
decnet/decky_io/resolve.py
Normal file
72
decnet/decky_io/resolve.py
Normal file
@@ -0,0 +1,72 @@
|
||||
"""Decky-name → docker container name resolution.
|
||||
|
||||
Two scopes:
|
||||
|
||||
* **Fleet**: every fleet decky has a ``ssh`` service container named
|
||||
``<decky_name>-ssh`` (see :mod:`decnet.services.ssh`). We always
|
||||
target it because it carries the most realistic filesystem layout.
|
||||
* **MazeNET (topology)**: same ``<name>-ssh`` convention when the
|
||||
decky exposes the ssh service; otherwise the decky's base container
|
||||
named ``decnet_t_<topology_id8>_<decky_name>`` (matches
|
||||
:func:`decnet.topology.compose._container_name`).
|
||||
|
||||
Keeping resolution centralised here means new ``docker exec`` callers
|
||||
(file drops, future bulk planters, etc.) never need to learn the
|
||||
naming conventions — they just call :func:`resolve_decky_container`.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
from typing import Any, Iterable, Optional
|
||||
|
||||
_SSH_CONTAINER_SUFFIX = "-ssh"
|
||||
|
||||
|
||||
def resolve_topology_container(
|
||||
topology_id: str, decky_name: str, services: Iterable[str],
|
||||
) -> str:
|
||||
"""Container name for a MazeNET decky.
|
||||
|
||||
See module docstring for the convention. Pure function — no I/O.
|
||||
"""
|
||||
if "ssh" in set(services):
|
||||
return f"{decky_name}{_SSH_CONTAINER_SUFFIX}"
|
||||
return f"decnet_t_{topology_id[:8]}_{decky_name}"
|
||||
|
||||
|
||||
async def resolve_decky_container(
|
||||
repo: Any,
|
||||
decky_name: str,
|
||||
*,
|
||||
topology_id: Optional[str] = None,
|
||||
) -> str:
|
||||
"""Resolve the docker container name for *decky_name*.
|
||||
|
||||
Fleet path (``topology_id is None``): returns ``<decky_name>-ssh``
|
||||
unconditionally. No DB lookup — the caller is responsible for
|
||||
knowing the decky exists; if it doesn't, the subsequent
|
||||
``docker exec`` returns a clear error.
|
||||
|
||||
Topology path: hydrates the topology, looks up the decky's services
|
||||
list, delegates to :func:`resolve_topology_container`.
|
||||
|
||||
Raises:
|
||||
LookupError — when ``topology_id`` is set but the topology or
|
||||
its named decky doesn't exist. Callers translate this into
|
||||
404/422 at the API layer.
|
||||
"""
|
||||
if topology_id is None:
|
||||
return f"{decky_name}{_SSH_CONTAINER_SUFFIX}"
|
||||
|
||||
from decnet.topology.persistence import hydrate
|
||||
hydrated = await hydrate(repo, topology_id)
|
||||
if hydrated is None:
|
||||
raise LookupError(f"topology {topology_id!r} not found")
|
||||
for decky in hydrated["deckies"]:
|
||||
cfg = decky.get("decky_config") or {}
|
||||
name = cfg.get("name") or decky.get("name")
|
||||
if name == decky_name:
|
||||
services = decky.get("services") or []
|
||||
return resolve_topology_container(topology_id, decky_name, services)
|
||||
raise LookupError(
|
||||
f"decky {decky_name!r} is not in topology {topology_id!r}"
|
||||
)
|
||||
124
decnet/decky_io/write.py
Normal file
124
decnet/decky_io/write.py
Normal file
@@ -0,0 +1,124 @@
|
||||
"""``docker exec``-driven file write/delete inside a decky container.
|
||||
|
||||
The write path streams a base64-encoded payload over stdin to
|
||||
``base64 -d`` inside the container, so binary content of any size up
|
||||
to docker's stream limits is safe — interpolating bytes into argv
|
||||
would trip ARG_MAX (~128 KB on most kernels) for any non-trivial blob.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import asyncio
|
||||
import base64
|
||||
import shlex
|
||||
from datetime import datetime, timezone
|
||||
from typing import Optional
|
||||
|
||||
from decnet.logging import get_logger
|
||||
|
||||
log = get_logger("decky_io.write")
|
||||
|
||||
_DOCKER = "docker"
|
||||
_DEFAULT_TIMEOUT = 8.0
|
||||
|
||||
|
||||
def _dirname(path: str) -> str:
|
||||
idx = path.rfind("/")
|
||||
if idx <= 0:
|
||||
return "/"
|
||||
return path[:idx]
|
||||
|
||||
|
||||
async def _run(
|
||||
argv: list[str],
|
||||
*,
|
||||
stdin_bytes: Optional[bytes] = None,
|
||||
timeout: float = _DEFAULT_TIMEOUT,
|
||||
) -> tuple[int, str, str]:
|
||||
try:
|
||||
proc = await asyncio.create_subprocess_exec(
|
||||
*argv,
|
||||
stdin=asyncio.subprocess.PIPE if stdin_bytes is not None else None,
|
||||
stdout=asyncio.subprocess.PIPE,
|
||||
stderr=asyncio.subprocess.PIPE,
|
||||
)
|
||||
except FileNotFoundError as exc:
|
||||
return 127, "", f"argv[0] not found: {exc}"
|
||||
try:
|
||||
stdout, stderr = await asyncio.wait_for(
|
||||
proc.communicate(input=stdin_bytes), timeout=timeout,
|
||||
)
|
||||
except asyncio.TimeoutError:
|
||||
try:
|
||||
proc.kill()
|
||||
except ProcessLookupError:
|
||||
pass
|
||||
return 124, "", "timeout"
|
||||
return (
|
||||
proc.returncode if proc.returncode is not None else -1,
|
||||
stdout.decode("utf-8", "replace"),
|
||||
stderr.decode("utf-8", "replace"),
|
||||
)
|
||||
|
||||
|
||||
async def write_file_to_container(
|
||||
container: str,
|
||||
path: str,
|
||||
content: bytes,
|
||||
*,
|
||||
mode: int = 0o644,
|
||||
mtime: Optional[datetime] = None,
|
||||
timeout: float = _DEFAULT_TIMEOUT,
|
||||
) -> tuple[bool, Optional[str]]:
|
||||
"""Write *content* to *path* inside *container* via ``docker exec``.
|
||||
|
||||
The directory above *path* is created if missing; *mode* is applied
|
||||
after the write; when *mtime* is provided the file is backdated via
|
||||
``touch -d`` (UTC ISO 8601).
|
||||
|
||||
Returns ``(success, error_or_none)``. ``error`` is the trimmed
|
||||
docker stderr on rc != 0, or a short "rc=<n>" if stderr was empty.
|
||||
"""
|
||||
if not path:
|
||||
return False, "empty path"
|
||||
|
||||
encoded = base64.b64encode(content)
|
||||
parts = [
|
||||
f"mkdir -p {shlex.quote(_dirname(path))}",
|
||||
f"base64 -d > {shlex.quote(path)}",
|
||||
f"chmod {mode:o} {shlex.quote(path)}",
|
||||
]
|
||||
if mtime is not None:
|
||||
ts = mtime.astimezone(timezone.utc).strftime("%Y-%m-%d %H:%M:%S UTC")
|
||||
parts.append(f"touch -d {shlex.quote(ts)} {shlex.quote(path)}")
|
||||
sh_cmd = " && ".join(parts)
|
||||
argv = [_DOCKER, "exec", "-i", container, "sh", "-c", sh_cmd]
|
||||
rc, _stdout, stderr = await _run(argv, stdin_bytes=encoded, timeout=timeout)
|
||||
success = rc == 0
|
||||
if success:
|
||||
return True, None
|
||||
err = stderr.strip()[:256] or f"rc={rc}"
|
||||
log.warning(
|
||||
"decky_io.write failed container=%s path=%s rc=%d stderr=%r",
|
||||
container, path, rc, stderr[:120],
|
||||
)
|
||||
return False, err
|
||||
|
||||
|
||||
async def delete_file_from_container(
|
||||
container: str,
|
||||
path: str,
|
||||
*,
|
||||
timeout: float = _DEFAULT_TIMEOUT,
|
||||
) -> tuple[bool, Optional[str]]:
|
||||
"""Best-effort ``rm -f`` of *path* inside *container*.
|
||||
|
||||
Returns ``(success, error_or_none)``. ``rm -f`` returns rc=0 even
|
||||
when the file is already gone, so a True result here means "the
|
||||
file is not present after this call", regardless of who unlinked it.
|
||||
"""
|
||||
sh_cmd = f"rm -f {shlex.quote(path)}"
|
||||
argv = [_DOCKER, "exec", container, "sh", "-c", sh_cmd]
|
||||
rc, _stdout, stderr = await _run(argv, timeout=timeout)
|
||||
if rc == 0:
|
||||
return True, None
|
||||
return False, stderr.strip()[:256] or f"rc={rc}"
|
||||
@@ -18,11 +18,8 @@ or IP can't escape into a shell.
|
||||
from __future__ import annotations
|
||||
|
||||
import asyncio
|
||||
import shlex
|
||||
from typing import Any
|
||||
|
||||
import base64
|
||||
from datetime import datetime, timezone
|
||||
from datetime import datetime
|
||||
|
||||
from decnet.logging import get_logger
|
||||
from decnet.orchestrator.drivers.base import ActivityDriver, ActivityResult
|
||||
@@ -226,36 +223,24 @@ class SSHDriver(ActivityDriver):
|
||||
) -> ActivityResult:
|
||||
"""Write *content* to *path* inside *decky_name*'s ssh container.
|
||||
|
||||
Streams base64 via stdin (mirrors :mod:`decnet.canary.planter`'s
|
||||
ARG_MAX-safe write — see commit c17b9e0). Sets file mode and,
|
||||
when *mtime* is provided, ``touch -d`` to backdate the file so
|
||||
it doesn't all stamp at wall-clock-now (the realism failure
|
||||
this migration is fixing).
|
||||
Delegates to :func:`decnet.decky_io.write_file_to_container`,
|
||||
which carries the ARG_MAX-safe base64-via-stdin trick. Sets
|
||||
file mode and, when *mtime* is provided, ``touch -d`` to
|
||||
backdate the file (otherwise everything stamps at wall-clock-now
|
||||
— the realism failure this path was originally fixing).
|
||||
"""
|
||||
from decnet.decky_io import write_file_to_container
|
||||
|
||||
container = _container_for(decky_name)
|
||||
b64 = base64.b64encode(content).decode("ascii")
|
||||
# touch -d accepts ISO 8601; we always emit UTC so the
|
||||
# container's local TZ doesn't drift the mtime.
|
||||
if mtime is not None:
|
||||
ts = mtime.astimezone(timezone.utc).strftime("%Y-%m-%d %H:%M:%S UTC")
|
||||
touch_cmd = f"touch -d {shlex.quote(ts)} {shlex.quote(path)}"
|
||||
else:
|
||||
touch_cmd = f"touch {shlex.quote(path)}"
|
||||
sh_cmd = (
|
||||
f"mkdir -p {shlex.quote(_dirname(path))} && "
|
||||
f"base64 -d > {shlex.quote(path)} && "
|
||||
f"chmod {mode:o} {shlex.quote(path)} && "
|
||||
f"{touch_cmd}"
|
||||
success, error = await write_file_to_container(
|
||||
container, path, content, mode=mode, mtime=mtime, timeout=_TIMEOUT,
|
||||
)
|
||||
argv = [_DOCKER, "exec", "-i", container, "sh", "-c", sh_cmd]
|
||||
rc, _stdout, stderr = await _run_with_stdin(argv, b64.encode("ascii"))
|
||||
success = rc == 0
|
||||
payload: dict[str, Any] = {
|
||||
"dst_decky": decky_name,
|
||||
"path": path,
|
||||
"bytes": len(content),
|
||||
"rc": rc,
|
||||
"stderr": stderr.strip()[:256] if not success else None,
|
||||
"rc": 0 if success else 1,
|
||||
"stderr": error if not success else None,
|
||||
}
|
||||
return ActivityResult(success=success, payload=payload)
|
||||
|
||||
@@ -283,11 +268,3 @@ class SSHDriver(ActivityDriver):
|
||||
)
|
||||
|
||||
|
||||
def _dirname(path: str) -> str:
|
||||
"""Pure-string dirname. We can't trust ``os.path.dirname`` on the
|
||||
host to share the destination container's separator semantics, but
|
||||
deckies are POSIX so a plain ``rfind('/')`` suffices."""
|
||||
idx = path.rfind("/")
|
||||
if idx <= 0:
|
||||
return "/"
|
||||
return path[:idx]
|
||||
|
||||
@@ -63,6 +63,10 @@ from .deploy import (
|
||||
MutateIntervalRequest,
|
||||
PurgeResponse,
|
||||
)
|
||||
from .decky import (
|
||||
DeckyFileDeleteRequest,
|
||||
DeckyFileDropRequest,
|
||||
)
|
||||
from .fleet import (
|
||||
LOCAL_HOST_SENTINEL,
|
||||
FleetDecky,
|
||||
@@ -222,6 +226,8 @@ __all__ = [
|
||||
"PurgeResponse",
|
||||
# fleet
|
||||
"LOCAL_HOST_SENTINEL",
|
||||
"DeckyFileDeleteRequest",
|
||||
"DeckyFileDropRequest",
|
||||
"FleetDecky",
|
||||
# health
|
||||
"ComponentHealth",
|
||||
|
||||
61
decnet/web/db/models/decky.py
Normal file
61
decnet/web/db/models/decky.py
Normal file
@@ -0,0 +1,61 @@
|
||||
"""DTOs for cross-cutting decky operations (file drops, etc.).
|
||||
|
||||
These don't bind to a single table — fleet deckies and MazeNET
|
||||
(topology) deckies share the request shape, with ``topology_id``
|
||||
discriminating. Following ``feedback_models_single_source`` we put
|
||||
the request/response shapes alongside the rest of the API contracts
|
||||
under ``decnet.web.db.models``.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
from typing import Optional
|
||||
|
||||
from pydantic import BaseModel, Field as PydanticField, field_validator
|
||||
|
||||
|
||||
class DeckyFileDropRequest(BaseModel):
|
||||
"""Drop arbitrary bytes at an absolute path inside a decky container.
|
||||
|
||||
``content_b64`` is the base64-encoded payload. Binary-safe.
|
||||
|
||||
``mode`` defaults to ``0o644`` (octal int). ``mtime_offset`` is a
|
||||
seconds offset from now applied via ``touch -d`` so realistic-aged
|
||||
files don't all stamp at wall-clock-now.
|
||||
"""
|
||||
decky_name: str = PydanticField(..., min_length=1)
|
||||
topology_id: Optional[str] = None
|
||||
path: str = PydanticField(..., min_length=1)
|
||||
content_b64: str
|
||||
mode: int = 0o644
|
||||
mtime_offset: int = 0
|
||||
|
||||
@field_validator("path")
|
||||
@classmethod
|
||||
def _abs_no_traversal(cls, v: str) -> str:
|
||||
if not v.startswith("/"):
|
||||
raise ValueError("path must be absolute (start with '/')")
|
||||
# Defense in depth: even though we run as root inside the
|
||||
# container, ``..`` segments make the on-disk location depend
|
||||
# on the cwd at exec-time and surprise both operators and the
|
||||
# auditor reading the placement_path field later.
|
||||
for seg in v.split("/"):
|
||||
if seg == "..":
|
||||
raise ValueError("path must not contain '..' segments")
|
||||
return v
|
||||
|
||||
|
||||
class DeckyFileDeleteRequest(BaseModel):
|
||||
"""Best-effort ``rm -f`` of an absolute path inside a decky container."""
|
||||
decky_name: str = PydanticField(..., min_length=1)
|
||||
topology_id: Optional[str] = None
|
||||
path: str = PydanticField(..., min_length=1)
|
||||
|
||||
@field_validator("path")
|
||||
@classmethod
|
||||
def _abs_no_traversal(cls, v: str) -> str:
|
||||
if not v.startswith("/"):
|
||||
raise ValueError("path must be absolute (start with '/')")
|
||||
for seg in v.split("/"):
|
||||
if seg == "..":
|
||||
raise ValueError("path must not contain '..' segments")
|
||||
return v
|
||||
@@ -50,6 +50,7 @@ from .swarm_mgmt import swarm_mgmt_router
|
||||
from .system import system_router
|
||||
from .topology import topology_router
|
||||
from .canary import canary_router
|
||||
from .deckies import deckies_router
|
||||
from .webhooks import webhooks_router
|
||||
|
||||
api_router = APIRouter(
|
||||
@@ -156,6 +157,7 @@ api_router.include_router(topology_router)
|
||||
# Canary tokens — operator-facing CRUD (worker hosts the
|
||||
# attacker-facing surface separately via `decnet canary`).
|
||||
api_router.include_router(canary_router)
|
||||
api_router.include_router(deckies_router)
|
||||
|
||||
# External webhook subscriptions (SIEM/SOAR egress)
|
||||
api_router.include_router(webhooks_router)
|
||||
|
||||
@@ -66,26 +66,20 @@ async def _resolve_topology_target(
|
||||
) -> str:
|
||||
"""Validate (topology_id, decky_name) and return the docker container.
|
||||
|
||||
404 if the topology doesn't exist; 422 if the named decky isn't in it.
|
||||
Hoisted into ``decky_io/resolve.py`` in workstream 2 so the file-drop
|
||||
endpoint can share it; for now it's local to the canary router.
|
||||
Delegates to :func:`decnet.decky_io.resolve_decky_container` and
|
||||
translates its ``LookupError`` into HTTP 404/422 — 404 when the
|
||||
topology itself is missing, 422 when the named decky isn't in it.
|
||||
"""
|
||||
from decnet.topology.persistence import hydrate
|
||||
hydrated = await hydrate(repo, topology_id)
|
||||
if hydrated is None:
|
||||
raise HTTPException(status_code=404, detail="topology not found")
|
||||
for decky in hydrated["deckies"]:
|
||||
cfg = decky.get("decky_config") or {}
|
||||
name = cfg.get("name") or decky.get("name")
|
||||
if name == decky_name:
|
||||
services = decky.get("services") or []
|
||||
return planter.resolve_topology_container(
|
||||
topology_id, decky_name, services,
|
||||
)
|
||||
raise HTTPException(
|
||||
status_code=422,
|
||||
detail=f"decky {decky_name!r} is not in topology {topology_id!r}",
|
||||
)
|
||||
from decnet.decky_io import resolve_decky_container
|
||||
try:
|
||||
return await resolve_decky_container(
|
||||
repo, decky_name, topology_id=topology_id,
|
||||
)
|
||||
except LookupError as exc:
|
||||
msg = str(exc)
|
||||
if "topology" in msg and "not found" in msg:
|
||||
raise HTTPException(status_code=404, detail=msg) from exc
|
||||
raise HTTPException(status_code=422, detail=msg) from exc
|
||||
|
||||
|
||||
def _trigger_row_to_response(row: dict[str, Any]) -> CanaryTriggerResponse:
|
||||
|
||||
21
decnet/web/router/deckies/__init__.py
Normal file
21
decnet/web/router/deckies/__init__.py
Normal file
@@ -0,0 +1,21 @@
|
||||
"""Cross-cutting decky operation endpoints.
|
||||
|
||||
These routes apply to both fleet and MazeNET (topology) deckies; the
|
||||
MazeNET case is selected by passing ``topology_id`` in the request body.
|
||||
|
||||
Compare with:
|
||||
|
||||
* :mod:`decnet.web.router.fleet` — fleet-only CRUD (deploy, mutate,
|
||||
list).
|
||||
* :mod:`decnet.web.router.topology` — topology-only CRUD.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
from fastapi import APIRouter
|
||||
|
||||
from .api_file_drop import router as file_drop_router
|
||||
|
||||
deckies_router = APIRouter()
|
||||
deckies_router.include_router(file_drop_router)
|
||||
|
||||
__all__ = ["deckies_router"]
|
||||
126
decnet/web/router/deckies/api_file_drop.py
Normal file
126
decnet/web/router/deckies/api_file_drop.py
Normal file
@@ -0,0 +1,126 @@
|
||||
"""POST/DELETE /api/v1/deckies/files — generic file drops on deckies.
|
||||
|
||||
Wraps :func:`decnet.decky_io.write_file_to_container` /
|
||||
:func:`decnet.decky_io.delete_file_from_container` so admins can drop
|
||||
arbitrary bytes at arbitrary paths inside a running decky container —
|
||||
fleet OR MazeNET — without going through the canary surface.
|
||||
|
||||
Auth: ``require_admin`` everywhere (matches every other write op on
|
||||
deckies; see :mod:`decnet.web.router.fleet.api_mutate_decky`).
|
||||
|
||||
Container resolution mirrors the canary path: ``topology_id`` absent
|
||||
means fleet (``<decky>-ssh``), present routes through
|
||||
:func:`decnet.decky_io.resolve_decky_container` for the MazeNET
|
||||
``<decky>-ssh`` / ``decnet_t_<id8>_<decky>`` distinction.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import base64
|
||||
from datetime import datetime, timedelta, timezone
|
||||
|
||||
from fastapi import APIRouter, Depends, HTTPException
|
||||
|
||||
from decnet.decky_io import (
|
||||
delete_file_from_container,
|
||||
resolve_decky_container,
|
||||
write_file_to_container,
|
||||
)
|
||||
from decnet.logging import get_logger
|
||||
from decnet.web.db.models import (
|
||||
DeckyFileDeleteRequest,
|
||||
DeckyFileDropRequest,
|
||||
MessageResponse,
|
||||
)
|
||||
from decnet.web.dependencies import repo, require_admin
|
||||
|
||||
log = get_logger("api.deckies.files")
|
||||
|
||||
router = APIRouter(prefix="/deckies/files", tags=["Deckies"])
|
||||
|
||||
|
||||
async def _resolve_container_or_4xx(
|
||||
decky_name: str, topology_id: str | None,
|
||||
) -> str:
|
||||
"""Resolve to a docker container, mapping LookupError → 404/422."""
|
||||
try:
|
||||
return await resolve_decky_container(
|
||||
repo, decky_name, topology_id=topology_id,
|
||||
)
|
||||
except LookupError as exc:
|
||||
msg = str(exc)
|
||||
if topology_id and "topology" in msg and "not found" in msg:
|
||||
raise HTTPException(status_code=404, detail=msg) from exc
|
||||
raise HTTPException(status_code=422, detail=msg) from exc
|
||||
|
||||
|
||||
@router.post(
|
||||
"",
|
||||
response_model=MessageResponse,
|
||||
status_code=201,
|
||||
responses={
|
||||
400: {"description": "Invalid request body (bad base64, etc.)"},
|
||||
401: {"description": "Could not validate credentials"},
|
||||
403: {"description": "Insufficient permissions"},
|
||||
404: {"description": "Topology not found"},
|
||||
409: {"description": "docker exec failed (container down or path unwritable)"},
|
||||
422: {"description": "Path validation failed or decky not in topology"},
|
||||
},
|
||||
)
|
||||
async def api_drop_file(
|
||||
req: DeckyFileDropRequest,
|
||||
admin: dict = Depends(require_admin),
|
||||
) -> MessageResponse:
|
||||
try:
|
||||
content = base64.b64decode(req.content_b64, validate=True)
|
||||
except (ValueError, TypeError) as exc:
|
||||
raise HTTPException(
|
||||
status_code=400, detail=f"content_b64 is not valid base64: {exc}",
|
||||
) from exc
|
||||
|
||||
container = await _resolve_container_or_4xx(req.decky_name, req.topology_id)
|
||||
mtime = (
|
||||
datetime.now(timezone.utc) + timedelta(seconds=req.mtime_offset)
|
||||
if req.mtime_offset
|
||||
else None
|
||||
)
|
||||
success, error = await write_file_to_container(
|
||||
container, req.path, content, mode=req.mode, mtime=mtime,
|
||||
)
|
||||
if not success:
|
||||
raise HTTPException(status_code=409, detail=error or "docker exec failed")
|
||||
log.info(
|
||||
"decky.file.drop decky=%s topology=%s container=%s path=%s bytes=%d by=%s",
|
||||
req.decky_name, req.topology_id, container, req.path,
|
||||
len(content), admin.get("uuid", "unknown"),
|
||||
)
|
||||
return MessageResponse(message="ok")
|
||||
|
||||
|
||||
@router.delete(
|
||||
"",
|
||||
response_model=MessageResponse,
|
||||
responses={
|
||||
401: {"description": "Could not validate credentials"},
|
||||
403: {"description": "Insufficient permissions"},
|
||||
404: {"description": "Topology not found"},
|
||||
422: {"description": "Path validation failed or decky not in topology"},
|
||||
},
|
||||
)
|
||||
async def api_delete_file(
|
||||
req: DeckyFileDeleteRequest,
|
||||
admin: dict = Depends(require_admin),
|
||||
) -> MessageResponse:
|
||||
container = await _resolve_container_or_4xx(req.decky_name, req.topology_id)
|
||||
success, error = await delete_file_from_container(container, req.path)
|
||||
# ``rm -f`` returns 0 even when the file is already gone, so a
|
||||
# False here means the docker exec itself failed. Don't 404 — the
|
||||
# caller asked us to ensure absence and we couldn't reach the
|
||||
# container. Surface it as 409.
|
||||
if not success:
|
||||
raise HTTPException(status_code=409, detail=error or "docker exec failed")
|
||||
log.info(
|
||||
"decky.file.delete decky=%s topology=%s container=%s path=%s by=%s",
|
||||
req.decky_name, req.topology_id, container, req.path,
|
||||
admin.get("uuid", "unknown"),
|
||||
)
|
||||
return MessageResponse(message="ok")
|
||||
Reference in New Issue
Block a user