Files
shopdb-flask/shopdb/utils/authz.py
cproudlock 534f179ee6 Make main the lean product a site receives
main was 16 days and 259 commits behind the line the product is actually built
on, at contract 0.30.0 against 0.35.0, with no per-PC credential code at all. It
was neither current nor curated - just stale. This brings it forward and, at the
same time, makes it mean something: main is what a receiving site installs, and
nothing else.

WHAT IT CARRIES. Twelve plugins - backups, computers, endpoints, guides,
knowledgebase, machines, network, notifications, printers, slides, tools,
warranty - plus applications, which is manifest-less core and always ships. The
only manifest dependency inside that set is endpoints on computers, and both are
in it, so the closure is self-contained.

WHAT IT DOES NOT. employees, geenforce, measuringtools, printedparts and usb
stay on the site line. So do the Windows installer tree, the VM rehearsal
helpers, the agent tooling config, one site's data loader, the internal audit
and findings records, and the planner spreadsheet. 1,135 tracked files against
1,438.

A REMOVAL THAT BREAKS THE APP IS NOT A REMOVAL, and two nearly shipped:

deploy/site-profile-universal.json still named all seventeen plugins. It is the
file docs/DEPLOY.md tells a site to provision from, so a fresh install would
have failed at `flask plugin apply-profile` with PluginError - the first thing a
receiving site does, broken by the branch meant to serve them. It now names the
twelve that exist, and its comment says the two must agree.

agent/ was never refreshed. Whoever populated this branch copied shopdb/,
plugins/, frontend/ and tests/ from the site line and left the agent tree at its
September state: 22 C# and PowerShell sources missing, including AgentCredentials,
EnrollmentService, AgentHttpFactory and the fleet-alert client, plus one file the
site line had deleted still present. The old tree is internally consistent, so it
compiles, and nothing in CI builds it - every gate was green over a server that
speaks enrollment and per-PC credentials to a bundled agent source that cannot.
Resynced whole.

DECISIONS THIS ENCODES, rather than defers:

- The plugin-authoring lab builds printedparts, whose manifest depends on
  employees, so it could not stay. The lab remains on the site line and main
  ships no build-along tutorial. That is a real loss for an adopting developer
  and is recorded here rather than discovered later.
- Recognition, Training and Recertification notifications are no longer seeded.
  Each splits one notification into a card per employee and renders a photo,
  which needs a plugin this branch does not bundle. Seeding them would offer an
  author a type that cannot resolve anybody. A site that installs employees gets
  them from that plugin's own seed.
- Windows/IIS is documented as the legacy path throughout, and the README leads
  with Docker, because that is what a receiving site gets.

The API description is regenerated for THIS branch - 347 paths and 483
operations, down from 415 and 573 - so the published contract describes what is
here rather than what the site line serves. docs/CONFIG.md lost the two
measuring-tool label keys the code no longer declares, which two lanes had
decided opposite ways.

Suite: 3124 passed, 6 skipped. Naming and style gates pass. The docs gates run
rather than skip, and every index line count is derived from the file it names.
2026-09-19 11:24:40 -04:00

291 lines
12 KiB
Python

"""Authorization decorators for role and permission gating.
Authentication (is the caller logged in?) is handled by Flask-JWT-Extended's
@jwt_required. Authorization (is the caller ALLOWED to do this?) is handled
here. The two are separate concerns; a route needs both on any state-changing
action.
These decorators call verify_jwt_in_request() themselves, so they work whether
or not a separate @jwt_required is also present. The admin role bypasses every
permission check (see User.haspermission), so an admin never needs individual
permissions granted.
THE GUEST ROLE
Somebody with no token is not nobody: they hold the `guest` role. If that role
has been granted the permission a route asks for, the request proceeds; if it
has not - and by default it holds nothing at all - the answer is 401, exactly
as before.
This replaced a growing pile of per-feature booleans (`securesend_allow_guests`,
one per kiosk, one nearly written for guides). Every one of them answered the
same question in its own vocabulary, in its own settings page, checked by its
own hand-written branch inside a route. What a signed-out visitor may see is a
permissions question, and this app already administers permissions per role.
So: a site opens a door by ticking a box in the role grid, and a plugin needs
no code at all to participate. A feature FLAG is still a flag - whether a kiosk
screen exists at this site is not the same question as who may read it.
Usage:
@assets_bp.route('/<int:asset_id>', methods=['DELETE'])
@require_permission('assets.delete')
def delete_asset(asset_id):
...
A separate @jwt_required() above these is redundant - they verify the request
themselves - and it is actively wrong now: it refuses a request with no token
before the guest role can be consulted, so a permission granted to guests would
appear to do nothing.
"""
from functools import wraps
from flask_jwt_extended import verify_jwt_in_request, current_user, get_jwt
from shopdb.utils.responses import error_response, ErrorCodes
def currentuser():
"""The signed-in user, or None.
`current_user` is a LocalProxy: the proxy object exists whether or not
anybody is behind it, so `current_user is None` is False even for a request
with no token. Under a required JWT that never showed, because the request
was refused before reaching here. With optional verification it shows
immediately, as "'NoneType' object has no attribute 'haspermission'".
"""
try:
return current_user._get_current_object()
except Exception:
return None
def guestholds(permission_name: str) -> bool:
"""Whether the guest role has been granted this permission.
Read every time, not cached. Revoking guest access is what somebody does the
moment they realise what they opened up, and it has to bite on the very next
request - a cache keyed to the app context outlives a request whenever one
is pushed around several (which is exactly how the test client behaves, and
how the revoke test caught it). One small indexed query on the rare path is
worth more than that risk.
A site with no guest role yet - an older database, seeded before this
existed - answers no, which is the behaviour it already had.
"""
from shopdb.core.models import Role
try:
role = Role.guest()
except Exception:
# A database that cannot answer must not turn into an open door.
return False
if role is None:
return False
return any(p.name == permission_name for p in role.permissions)
def caller_holds(permission_name: str) -> bool:
"""Whether THIS caller holds a permission, without refusing the request.
require_permission answers the same question by returning 401/403. This
answers it in the body, for the case where a route serves everyone but
should only OFFER something to the callers who can use it.
The endpoints services panel is the example that prompted it: it emitted
Start/Stop buttons for every reader, because the emission never consulted
the caller. Clicking them 403s, so nothing was exploitable, but showing a
read-only viewer controls over a shop-floor PC is the wrong signal - and it
gets worse the moment a site grants that panel to guests.
Guests are included on purpose: a signed-out visitor holds whatever the
guest role was granted, so the answer must come from the same place
require_permission gets it.
Scoped personal tokens are included for the same reason, and this is the
half that was missing: the answer has to be about the CALLER, and a scoped
token is a narrower caller than its owner. Reading the owner's permissions
let a token scoped to endpoints.command queue a remote restart, because the
elevated check asked the owner rather than the token - the one bypass
require_permission exists to prevent, reintroduced by a route that did its
own check.
"""
user = currentuser()
if user is None:
return guestholds(permission_name)
try:
if not user.haspermission(permission_name):
return False
patscopes = get_jwt().get('patscopes')
except Exception:
return False
if patscopes is not None:
return permission_name in patscopes
return True
def caller_is_role(rolename: str) -> bool:
"""Whether THIS caller holds a role, without refusing the request.
The body-answering counterpart to require_role, for a route that serves
everyone but treats one class of caller differently - break-glass override
being the case that prompted it.
A scoped personal token is ALWAYS refused here, matching require_role: scopes
gate individual permissions, not roles. This is the invariant that keeps a
leaked scoped agent token from inheriting override authority just because the
token's owner happens to be an admin. A scoped PAT is a narrower caller than
its owner, including when that owner is an admin.
Guests hold no roles, so an unauthenticated caller is always False.
"""
user = currentuser()
if user is None:
return False
try:
if get_jwt().get('patscopes') is not None:
return False
return user.hasrole(rolename)
except Exception:
return False
def require_permission(permission_name: str):
"""Gate a route behind a single permission. Admin role bypasses.
A scoped personal API token (patscopes claim present) does NOT bypass: it
grants only the permissions in its scope list, intersected with what the
owner actually holds. See shopdb/utils/apitoken_auth.py.
"""
def decorator(view_func):
@wraps(view_func)
def wrapper(*args, **kwargs):
# Optional, so a request with no token reaches the guest check
# below rather than being refused by the JWT layer before this
# decorator can have an opinion. A request that DOES carry a token
# is still verified, and a bad one still fails here.
verify_jwt_in_request(optional=True)
user = currentuser()
if user is None:
if guestholds(permission_name):
return view_func(*args, **kwargs)
return error_response(
ErrorCodes.UNAUTHORIZED,
'Authentication required',
http_code=401
)
patscopes = get_jwt().get('patscopes')
if patscopes is not None:
# Scoped PAT: allowed only when this permission is in the scope
# list AND the owner actually holds it. haspermission still
# returns True for an admin owner (who legitimately holds
# everything), so the scope list is the real limiter - the
# admin bypass is suspended.
allowed = (permission_name in patscopes
and user.haspermission(permission_name))
if not allowed:
return error_response(
ErrorCodes.FORBIDDEN,
'This API token is not scoped for this action',
http_code=403
)
return view_func(*args, **kwargs)
if not user.haspermission(permission_name):
return error_response(
ErrorCodes.FORBIDDEN,
'You do not have permission to perform this action',
http_code=403
)
return view_func(*args, **kwargs)
# Named so a test can sweep the URL map and prove every read is gated;
# a decorator is invisible from the outside otherwise.
wrapper._requires_permission = (permission_name,)
return wrapper
return decorator
def require_any_permission(*permission_names):
"""Gate a route behind ANY ONE of several permissions. Admin bypasses.
For a read that two different rights should reach. Slides is the case: a
GET was gated on the MANAGE permission, because managing was the only thing
anybody did there. Splitting a read permission out would have locked out
every existing manager, so a read now accepts either, and the read
permission is the one a site can hand to a guest.
A scoped PAT must carry one of the names, and its owner must hold that same
one - the intersection rule of require_permission, applied per name.
"""
def decorator(view_func):
@wraps(view_func)
def wrapper(*args, **kwargs):
verify_jwt_in_request(optional=True)
user = currentuser()
if user is None:
if any(guestholds(name) for name in permission_names):
return view_func(*args, **kwargs)
return error_response(
ErrorCodes.UNAUTHORIZED,
'Authentication required',
http_code=401
)
patscopes = get_jwt().get('patscopes')
if patscopes is not None:
allowed = any(name in patscopes and user.haspermission(name)
for name in permission_names)
if not allowed:
return error_response(
ErrorCodes.FORBIDDEN,
'This API token is not scoped for this action',
http_code=403
)
return view_func(*args, **kwargs)
if not any(user.haspermission(name) for name in permission_names):
return error_response(
ErrorCodes.FORBIDDEN,
'You do not have permission to perform this action',
http_code=403
)
return view_func(*args, **kwargs)
wrapper._requires_permission = tuple(permission_names)
return wrapper
return decorator
def require_role(rolename: str):
"""Gate a route behind a single role (e.g. 'admin').
A scoped personal API token (patscopes claim present) is ALWAYS denied here:
scopes gate individual permissions, not roles, so role-gated admin surfaces
require an unscoped token. See shopdb/utils/apitoken_auth.py.
"""
def decorator(view_func):
@wraps(view_func)
def wrapper(*args, **kwargs):
verify_jwt_in_request()
user = currentuser()
if user is None:
return error_response(
ErrorCodes.UNAUTHORIZED,
'Authentication required',
http_code=401
)
if get_jwt().get('patscopes') is not None:
return error_response(
ErrorCodes.FORBIDDEN,
'Scoped API tokens cannot access role-gated endpoints',
http_code=403
)
if not user.hasrole(rolename):
return error_response(
ErrorCodes.FORBIDDEN,
f'{rolename.capitalize()} access required',
http_code=403
)
return view_func(*args, **kwargs)
# A role gate is a gate too, as far as the read sweep is concerned.
wrapper._requires_permission = ('role:' + rolename,)
return wrapper
return decorator