zulip/zerver/lib/user_groups.py

1026 lines
37 KiB
Python
Raw Normal View History

from collections import defaultdict
from collections.abc import Collection, Iterable, Iterator, Mapping
user_groups: Make locks required for updating user group memberships. **Background** User groups are expected to comply with the DAG constraint for the many-to-many inter-group membership. The check for this constraint has to be performed recursively so that we can find all direct and indirect subgroups of the user group to be added. This kind of check is vulnerable to phantom reads which is possible at the default read committed isolation level because we cannot guarantee that the check is still valid when we are adding the subgroups to the user group. **Solution** To avoid having another transaction concurrently update one of the to-be-subgroup after the recursive check is done, and before the subgroup is added, we use SELECT FOR UPDATE to lock the user group rows. The lock needs to be acquired before a group membership change is about to occur before any check has been conducted. Suppose that we are adding subgroup B to supergroup A, the locking protocol is specified as follows: 1. Acquire a lock for B and all its direct and indirect subgroups. 2. Acquire a lock for A. For the removal of user groups, we acquire a lock for the user group to be removed with all its direct and indirect subgroups. This is the special case A=B, which is still complaint with the protocol. **Error handling** We currently rely on Postgres' deadlock detection to abort transactions and show an error for the users. In the future, we might need some recovery mechanism or at least better error handling. **Notes** An important note is that we need to reuse the recursive CTE query that finds the direct and indirect subgroups when applying the lock on the rows. And the lock needs to be acquired the same way for the addition and removal of direct subgroups. User membership change (as opposed to user group membership) is not affected. Read-only queries aren't either. The locks only protect critical regions where the user group dependency graph might violate the DAG constraint, where users are not participating. **Testing** We implement a transaction test case targeting some typical scenarios when an internal server error is expected to happen (this means that the user group view makes the correct decision to abort the transaction when something goes wrong with locks). To achieve this, we add a development view intended only for unit tests. It has a global BARRIER that can be shared across threads, so that we can synchronize them to consistently reproduce certain potential race conditions prevented by the database locks. The transaction test case lanuches pairs of threads initiating possibly conflicting requests at the same time. The tests are set up such that exactly N of them are expected to succeed with a certain error message (while we don't know each one). **Security notes** get_recursive_subgroups_for_groups will no longer fetch user groups from other realms. As a result, trying to add/remove a subgroup from another realm results in a UserGroup not found error response. We also implement subgroup-specific checks in has_user_group_access to keep permission managing in a single place. Do note that the API currently don't have a way to violate that check because we are only checking the realm ID now.
2023-06-17 04:39:52 +02:00
from contextlib import contextmanager
from dataclasses import asdict, dataclass
from typing import Any, TypedDict
from django.conf import settings
from django.db import connection, transaction
from django.db.models import F, Q, QuerySet
from django.utils.timezone import now as timezone_now
from django.utils.translation import gettext as _
from django_cte import With
from psycopg2.sql import SQL, Literal
from zerver.lib.exceptions import (
CannotDeactivateGroupInUseError,
JsonableError,
PreviousSettingValueMismatchedError,
SystemGroupRequiredError,
)
from zerver.lib.timestamp import datetime_to_timestamp
from zerver.lib.types import GroupPermissionSetting, ServerSupportedPermissionSettings
from zerver.models import (
GroupGroupMembership,
NamedUserGroup,
Realm,
RealmAuditLog,
Stream,
UserGroup,
UserGroupMembership,
UserProfile,
)
from zerver.models.groups import SystemGroups
from zerver.models.realm_audit_logs import AuditLogEventType
@dataclass
class AnonymousSettingGroupDict:
direct_members: list[int]
direct_subgroups: list[int]
@dataclass
class GroupSettingChangeRequest:
new: int | AnonymousSettingGroupDict
old: int | AnonymousSettingGroupDict | None = None
class UserGroupDict(TypedDict):
id: int
name: str
description: str
members: list[int]
direct_subgroup_ids: list[int]
creator_id: int | None
date_created: int | None
is_system_group: bool
can_manage_group: int | AnonymousSettingGroupDict
can_mention_group: int | AnonymousSettingGroupDict
deactivated: bool
user_groups: Make locks required for updating user group memberships. **Background** User groups are expected to comply with the DAG constraint for the many-to-many inter-group membership. The check for this constraint has to be performed recursively so that we can find all direct and indirect subgroups of the user group to be added. This kind of check is vulnerable to phantom reads which is possible at the default read committed isolation level because we cannot guarantee that the check is still valid when we are adding the subgroups to the user group. **Solution** To avoid having another transaction concurrently update one of the to-be-subgroup after the recursive check is done, and before the subgroup is added, we use SELECT FOR UPDATE to lock the user group rows. The lock needs to be acquired before a group membership change is about to occur before any check has been conducted. Suppose that we are adding subgroup B to supergroup A, the locking protocol is specified as follows: 1. Acquire a lock for B and all its direct and indirect subgroups. 2. Acquire a lock for A. For the removal of user groups, we acquire a lock for the user group to be removed with all its direct and indirect subgroups. This is the special case A=B, which is still complaint with the protocol. **Error handling** We currently rely on Postgres' deadlock detection to abort transactions and show an error for the users. In the future, we might need some recovery mechanism or at least better error handling. **Notes** An important note is that we need to reuse the recursive CTE query that finds the direct and indirect subgroups when applying the lock on the rows. And the lock needs to be acquired the same way for the addition and removal of direct subgroups. User membership change (as opposed to user group membership) is not affected. Read-only queries aren't either. The locks only protect critical regions where the user group dependency graph might violate the DAG constraint, where users are not participating. **Testing** We implement a transaction test case targeting some typical scenarios when an internal server error is expected to happen (this means that the user group view makes the correct decision to abort the transaction when something goes wrong with locks). To achieve this, we add a development view intended only for unit tests. It has a global BARRIER that can be shared across threads, so that we can synchronize them to consistently reproduce certain potential race conditions prevented by the database locks. The transaction test case lanuches pairs of threads initiating possibly conflicting requests at the same time. The tests are set up such that exactly N of them are expected to succeed with a certain error message (while we don't know each one). **Security notes** get_recursive_subgroups_for_groups will no longer fetch user groups from other realms. As a result, trying to add/remove a subgroup from another realm results in a UserGroup not found error response. We also implement subgroup-specific checks in has_user_group_access to keep permission managing in a single place. Do note that the API currently don't have a way to violate that check because we are only checking the realm ID now.
2023-06-17 04:39:52 +02:00
@dataclass
class LockedUserGroupContext:
"""User groups in this dataclass are guaranteeed to be locked until the
end of the current transaction.
supergroup is the user group to have subgroups added or removed;
direct_subgroups are user groups that are recursively queried for subgroups;
recursive_subgroups include direct_subgroups and their descendants.
"""
supergroup: NamedUserGroup
direct_subgroups: list[NamedUserGroup]
recursive_subgroups: list[NamedUserGroup]
user_groups: Make locks required for updating user group memberships. **Background** User groups are expected to comply with the DAG constraint for the many-to-many inter-group membership. The check for this constraint has to be performed recursively so that we can find all direct and indirect subgroups of the user group to be added. This kind of check is vulnerable to phantom reads which is possible at the default read committed isolation level because we cannot guarantee that the check is still valid when we are adding the subgroups to the user group. **Solution** To avoid having another transaction concurrently update one of the to-be-subgroup after the recursive check is done, and before the subgroup is added, we use SELECT FOR UPDATE to lock the user group rows. The lock needs to be acquired before a group membership change is about to occur before any check has been conducted. Suppose that we are adding subgroup B to supergroup A, the locking protocol is specified as follows: 1. Acquire a lock for B and all its direct and indirect subgroups. 2. Acquire a lock for A. For the removal of user groups, we acquire a lock for the user group to be removed with all its direct and indirect subgroups. This is the special case A=B, which is still complaint with the protocol. **Error handling** We currently rely on Postgres' deadlock detection to abort transactions and show an error for the users. In the future, we might need some recovery mechanism or at least better error handling. **Notes** An important note is that we need to reuse the recursive CTE query that finds the direct and indirect subgroups when applying the lock on the rows. And the lock needs to be acquired the same way for the addition and removal of direct subgroups. User membership change (as opposed to user group membership) is not affected. Read-only queries aren't either. The locks only protect critical regions where the user group dependency graph might violate the DAG constraint, where users are not participating. **Testing** We implement a transaction test case targeting some typical scenarios when an internal server error is expected to happen (this means that the user group view makes the correct decision to abort the transaction when something goes wrong with locks). To achieve this, we add a development view intended only for unit tests. It has a global BARRIER that can be shared across threads, so that we can synchronize them to consistently reproduce certain potential race conditions prevented by the database locks. The transaction test case lanuches pairs of threads initiating possibly conflicting requests at the same time. The tests are set up such that exactly N of them are expected to succeed with a certain error message (while we don't know each one). **Security notes** get_recursive_subgroups_for_groups will no longer fetch user groups from other realms. As a result, trying to add/remove a subgroup from another realm results in a UserGroup not found error response. We also implement subgroup-specific checks in has_user_group_access to keep permission managing in a single place. Do note that the API currently don't have a way to violate that check because we are only checking the realm ID now.
2023-06-17 04:39:52 +02:00
def has_user_group_access_for_subgroup(
user_group: NamedUserGroup,
user_profile: UserProfile,
*,
allow_deactivated: bool = False,
) -> bool:
"""Minimal access control checks for whether the given group
is visible to the given user for use as a subgroup.
In the future, if groups whose existence is not visible to the
entire organization are added, this may grow more complex.
"""
user_groups: Make locks required for updating user group memberships. **Background** User groups are expected to comply with the DAG constraint for the many-to-many inter-group membership. The check for this constraint has to be performed recursively so that we can find all direct and indirect subgroups of the user group to be added. This kind of check is vulnerable to phantom reads which is possible at the default read committed isolation level because we cannot guarantee that the check is still valid when we are adding the subgroups to the user group. **Solution** To avoid having another transaction concurrently update one of the to-be-subgroup after the recursive check is done, and before the subgroup is added, we use SELECT FOR UPDATE to lock the user group rows. The lock needs to be acquired before a group membership change is about to occur before any check has been conducted. Suppose that we are adding subgroup B to supergroup A, the locking protocol is specified as follows: 1. Acquire a lock for B and all its direct and indirect subgroups. 2. Acquire a lock for A. For the removal of user groups, we acquire a lock for the user group to be removed with all its direct and indirect subgroups. This is the special case A=B, which is still complaint with the protocol. **Error handling** We currently rely on Postgres' deadlock detection to abort transactions and show an error for the users. In the future, we might need some recovery mechanism or at least better error handling. **Notes** An important note is that we need to reuse the recursive CTE query that finds the direct and indirect subgroups when applying the lock on the rows. And the lock needs to be acquired the same way for the addition and removal of direct subgroups. User membership change (as opposed to user group membership) is not affected. Read-only queries aren't either. The locks only protect critical regions where the user group dependency graph might violate the DAG constraint, where users are not participating. **Testing** We implement a transaction test case targeting some typical scenarios when an internal server error is expected to happen (this means that the user group view makes the correct decision to abort the transaction when something goes wrong with locks). To achieve this, we add a development view intended only for unit tests. It has a global BARRIER that can be shared across threads, so that we can synchronize them to consistently reproduce certain potential race conditions prevented by the database locks. The transaction test case lanuches pairs of threads initiating possibly conflicting requests at the same time. The tests are set up such that exactly N of them are expected to succeed with a certain error message (while we don't know each one). **Security notes** get_recursive_subgroups_for_groups will no longer fetch user groups from other realms. As a result, trying to add/remove a subgroup from another realm results in a UserGroup not found error response. We also implement subgroup-specific checks in has_user_group_access to keep permission managing in a single place. Do note that the API currently don't have a way to violate that check because we are only checking the realm ID now.
2023-06-17 04:39:52 +02:00
if user_group.realm_id != user_profile.realm_id:
return False
if not allow_deactivated and user_group.deactivated:
raise JsonableError(_("User group is deactivated."))
return True
def get_user_group_by_id_in_realm(
user_group_id: int,
realm: Realm,
*,
for_read: bool,
for_setting: bool = False,
allow_deactivated: bool = False,
) -> NamedUserGroup:
"""
Internal function for accessing a single user group from client
code. Locks the group if for_read is False.
Notably does not do any access control checks, beyond only fetching
groups from the provided realm.
"""
try:
if for_read:
user_group = NamedUserGroup.objects.get(id=user_group_id, realm=realm)
user_groups: Make locks required for updating user group memberships. **Background** User groups are expected to comply with the DAG constraint for the many-to-many inter-group membership. The check for this constraint has to be performed recursively so that we can find all direct and indirect subgroups of the user group to be added. This kind of check is vulnerable to phantom reads which is possible at the default read committed isolation level because we cannot guarantee that the check is still valid when we are adding the subgroups to the user group. **Solution** To avoid having another transaction concurrently update one of the to-be-subgroup after the recursive check is done, and before the subgroup is added, we use SELECT FOR UPDATE to lock the user group rows. The lock needs to be acquired before a group membership change is about to occur before any check has been conducted. Suppose that we are adding subgroup B to supergroup A, the locking protocol is specified as follows: 1. Acquire a lock for B and all its direct and indirect subgroups. 2. Acquire a lock for A. For the removal of user groups, we acquire a lock for the user group to be removed with all its direct and indirect subgroups. This is the special case A=B, which is still complaint with the protocol. **Error handling** We currently rely on Postgres' deadlock detection to abort transactions and show an error for the users. In the future, we might need some recovery mechanism or at least better error handling. **Notes** An important note is that we need to reuse the recursive CTE query that finds the direct and indirect subgroups when applying the lock on the rows. And the lock needs to be acquired the same way for the addition and removal of direct subgroups. User membership change (as opposed to user group membership) is not affected. Read-only queries aren't either. The locks only protect critical regions where the user group dependency graph might violate the DAG constraint, where users are not participating. **Testing** We implement a transaction test case targeting some typical scenarios when an internal server error is expected to happen (this means that the user group view makes the correct decision to abort the transaction when something goes wrong with locks). To achieve this, we add a development view intended only for unit tests. It has a global BARRIER that can be shared across threads, so that we can synchronize them to consistently reproduce certain potential race conditions prevented by the database locks. The transaction test case lanuches pairs of threads initiating possibly conflicting requests at the same time. The tests are set up such that exactly N of them are expected to succeed with a certain error message (while we don't know each one). **Security notes** get_recursive_subgroups_for_groups will no longer fetch user groups from other realms. As a result, trying to add/remove a subgroup from another realm results in a UserGroup not found error response. We also implement subgroup-specific checks in has_user_group_access to keep permission managing in a single place. Do note that the API currently don't have a way to violate that check because we are only checking the realm ID now.
2023-06-17 04:39:52 +02:00
else:
user_group = NamedUserGroup.objects.select_for_update().get(
id=user_group_id, realm=realm
user_groups: Make locks required for updating user group memberships. **Background** User groups are expected to comply with the DAG constraint for the many-to-many inter-group membership. The check for this constraint has to be performed recursively so that we can find all direct and indirect subgroups of the user group to be added. This kind of check is vulnerable to phantom reads which is possible at the default read committed isolation level because we cannot guarantee that the check is still valid when we are adding the subgroups to the user group. **Solution** To avoid having another transaction concurrently update one of the to-be-subgroup after the recursive check is done, and before the subgroup is added, we use SELECT FOR UPDATE to lock the user group rows. The lock needs to be acquired before a group membership change is about to occur before any check has been conducted. Suppose that we are adding subgroup B to supergroup A, the locking protocol is specified as follows: 1. Acquire a lock for B and all its direct and indirect subgroups. 2. Acquire a lock for A. For the removal of user groups, we acquire a lock for the user group to be removed with all its direct and indirect subgroups. This is the special case A=B, which is still complaint with the protocol. **Error handling** We currently rely on Postgres' deadlock detection to abort transactions and show an error for the users. In the future, we might need some recovery mechanism or at least better error handling. **Notes** An important note is that we need to reuse the recursive CTE query that finds the direct and indirect subgroups when applying the lock on the rows. And the lock needs to be acquired the same way for the addition and removal of direct subgroups. User membership change (as opposed to user group membership) is not affected. Read-only queries aren't either. The locks only protect critical regions where the user group dependency graph might violate the DAG constraint, where users are not participating. **Testing** We implement a transaction test case targeting some typical scenarios when an internal server error is expected to happen (this means that the user group view makes the correct decision to abort the transaction when something goes wrong with locks). To achieve this, we add a development view intended only for unit tests. It has a global BARRIER that can be shared across threads, so that we can synchronize them to consistently reproduce certain potential race conditions prevented by the database locks. The transaction test case lanuches pairs of threads initiating possibly conflicting requests at the same time. The tests are set up such that exactly N of them are expected to succeed with a certain error message (while we don't know each one). **Security notes** get_recursive_subgroups_for_groups will no longer fetch user groups from other realms. As a result, trying to add/remove a subgroup from another realm results in a UserGroup not found error response. We also implement subgroup-specific checks in has_user_group_access to keep permission managing in a single place. Do note that the API currently don't have a way to violate that check because we are only checking the realm ID now.
2023-06-17 04:39:52 +02:00
)
if not allow_deactivated and user_group.deactivated:
raise JsonableError(_("User group is deactivated."))
return user_group
except NamedUserGroup.DoesNotExist:
raise JsonableError(_("Invalid user group"))
def access_user_group_to_read_membership(user_group_id: int, realm: Realm) -> NamedUserGroup:
return get_user_group_by_id_in_realm(user_group_id, realm, for_read=True)
def check_permission_for_managing_all_groups(
user_group: UserGroup, user_profile: UserProfile
) -> bool:
"""
Given a user and a group in the same realm, checks if the user
can manage the group through the legacy can_manage_all_groups
permission, which is a permission that requires either certain roles
or membership in the group itself to be used.
"""
# This is a temporary exception and this should be removed as soon
# as `group_creator` is set as a default for `can_manage_group`
# property of user groups. See this topic for more details:
# https://chat.zulip.org/#narrow/stream/3-backend/topic/Group.20creation.20-.20who.20can.20change.20the.20setting.2E/near/1943861
if user_group.creator and user_group.creator.id == user_profile.id:
return True
can_manage_all_groups = user_profile.can_manage_all_groups()
if can_manage_all_groups:
if user_profile.is_realm_admin or user_profile.is_moderator:
return True
return is_user_in_group(user_group, user_profile)
return False
def access_user_group_for_update(
user_group_id: int,
user_profile: UserProfile,
*,
permission_setting: str,
allow_deactivated: bool = False,
) -> NamedUserGroup:
"""
Main entry point that views code should call when planning to modify
a given user group on behalf of a given user.
The permission_setting parameter indicates what permission to check;
different features will be used when editing the membership vs.
security-sensitive settings on a group.
"""
user_group = get_user_group_by_id_in_realm(
user_group_id, user_profile.realm, for_read=False, allow_deactivated=allow_deactivated
)
if user_group.is_system_group:
raise JsonableError(_("Insufficient permission"))
assert permission_setting in NamedUserGroup.GROUP_PERMISSION_SETTINGS
if permission_setting == "can_manage_group" and check_permission_for_managing_all_groups(
user_group, user_profile
):
return user_group
user_has_permission = user_has_permission_for_group_setting(
getattr(user_group, permission_setting),
user_profile,
NamedUserGroup.GROUP_PERMISSION_SETTINGS[permission_setting],
)
if not user_has_permission:
raise JsonableError(_("Insufficient permission"))
return user_group
def access_user_group_for_deactivation(
user_group_id: int, user_profile: UserProfile
) -> NamedUserGroup:
"""
Main security check / access function for whether the acting
user has permission to deactivate a given user group.
"""
user_group = access_user_group_for_update(
user_group_id, user_profile, permission_setting="can_manage_group"
)
objections: list[dict[str, Any]] = []
supergroup_ids = (
user_group.direct_supergroups.exclude(named_user_group=None)
.filter(named_user_group__deactivated=False)
.values_list("id", flat=True)
)
if supergroup_ids:
objections.append(dict(type="subgroup", supergroup_ids=list(supergroup_ids)))
anonymous_supergroup_ids = user_group.direct_supergroups.filter(
named_user_group=None
).values_list("id", flat=True)
# We check both the cases - whether the group is being directly used
# as the value of a setting or as a subgroup of an anonymous group
# used for a setting.
setting_group_ids_using_deactivating_user_group = {
*set(anonymous_supergroup_ids),
user_group.id,
}
stream_setting_query = Q()
for setting_name in Stream.stream_permission_group_settings:
stream_setting_query |= Q(
**{f"{setting_name}__in": setting_group_ids_using_deactivating_user_group}
)
for stream in Stream.objects.filter(realm_id=user_group.realm_id, deactivated=False).filter(
stream_setting_query
):
objection_settings = [
setting_name
for setting_name in Stream.stream_permission_group_settings
if getattr(stream, setting_name + "_id")
in setting_group_ids_using_deactivating_user_group
]
if len(objection_settings) > 0:
objections.append(
dict(type="channel", channel_id=stream.id, settings=objection_settings)
)
group_setting_query = Q()
for setting_name in NamedUserGroup.GROUP_PERMISSION_SETTINGS:
group_setting_query |= Q(
**{f"{setting_name}__in": setting_group_ids_using_deactivating_user_group}
)
for group in NamedUserGroup.objects.filter(
realm_id=user_group.realm_id, deactivated=False
).filter(group_setting_query):
objection_settings = []
for setting_name in NamedUserGroup.GROUP_PERMISSION_SETTINGS:
if (
getattr(group, setting_name + "_id")
in setting_group_ids_using_deactivating_user_group
):
objection_settings.append(setting_name)
if len(objection_settings) > 0:
objections.append(
dict(type="user_group", group_id=group.id, settings=objection_settings)
)
objection_settings = []
realm = user_group.realm
for setting_name in Realm.REALM_PERMISSION_GROUP_SETTINGS:
if getattr(realm, setting_name + "_id") in setting_group_ids_using_deactivating_user_group:
objection_settings.append(setting_name)
if objection_settings:
objections.append(dict(type="realm", settings=objection_settings))
if len(objections) > 0:
raise CannotDeactivateGroupInUseError(objections)
return user_group
user_groups: Make locks required for updating user group memberships. **Background** User groups are expected to comply with the DAG constraint for the many-to-many inter-group membership. The check for this constraint has to be performed recursively so that we can find all direct and indirect subgroups of the user group to be added. This kind of check is vulnerable to phantom reads which is possible at the default read committed isolation level because we cannot guarantee that the check is still valid when we are adding the subgroups to the user group. **Solution** To avoid having another transaction concurrently update one of the to-be-subgroup after the recursive check is done, and before the subgroup is added, we use SELECT FOR UPDATE to lock the user group rows. The lock needs to be acquired before a group membership change is about to occur before any check has been conducted. Suppose that we are adding subgroup B to supergroup A, the locking protocol is specified as follows: 1. Acquire a lock for B and all its direct and indirect subgroups. 2. Acquire a lock for A. For the removal of user groups, we acquire a lock for the user group to be removed with all its direct and indirect subgroups. This is the special case A=B, which is still complaint with the protocol. **Error handling** We currently rely on Postgres' deadlock detection to abort transactions and show an error for the users. In the future, we might need some recovery mechanism or at least better error handling. **Notes** An important note is that we need to reuse the recursive CTE query that finds the direct and indirect subgroups when applying the lock on the rows. And the lock needs to be acquired the same way for the addition and removal of direct subgroups. User membership change (as opposed to user group membership) is not affected. Read-only queries aren't either. The locks only protect critical regions where the user group dependency graph might violate the DAG constraint, where users are not participating. **Testing** We implement a transaction test case targeting some typical scenarios when an internal server error is expected to happen (this means that the user group view makes the correct decision to abort the transaction when something goes wrong with locks). To achieve this, we add a development view intended only for unit tests. It has a global BARRIER that can be shared across threads, so that we can synchronize them to consistently reproduce certain potential race conditions prevented by the database locks. The transaction test case lanuches pairs of threads initiating possibly conflicting requests at the same time. The tests are set up such that exactly N of them are expected to succeed with a certain error message (while we don't know each one). **Security notes** get_recursive_subgroups_for_groups will no longer fetch user groups from other realms. As a result, trying to add/remove a subgroup from another realm results in a UserGroup not found error response. We also implement subgroup-specific checks in has_user_group_access to keep permission managing in a single place. Do note that the API currently don't have a way to violate that check because we are only checking the realm ID now.
2023-06-17 04:39:52 +02:00
@contextmanager
def lock_subgroups_with_respect_to_supergroup(
potential_subgroup_ids: Collection[int], potential_supergroup_id: int, acting_user: UserProfile
) -> Iterator[LockedUserGroupContext]:
"""This locks the user groups with the given potential_subgroup_ids, as well
as their indirect subgroups, followed by the potential supergroup. It
ensures that we lock the user groups in a consistent order topologically to
avoid unnecessary deadlocks on non-conflicting queries.
user_groups: Make locks required for updating user group memberships. **Background** User groups are expected to comply with the DAG constraint for the many-to-many inter-group membership. The check for this constraint has to be performed recursively so that we can find all direct and indirect subgroups of the user group to be added. This kind of check is vulnerable to phantom reads which is possible at the default read committed isolation level because we cannot guarantee that the check is still valid when we are adding the subgroups to the user group. **Solution** To avoid having another transaction concurrently update one of the to-be-subgroup after the recursive check is done, and before the subgroup is added, we use SELECT FOR UPDATE to lock the user group rows. The lock needs to be acquired before a group membership change is about to occur before any check has been conducted. Suppose that we are adding subgroup B to supergroup A, the locking protocol is specified as follows: 1. Acquire a lock for B and all its direct and indirect subgroups. 2. Acquire a lock for A. For the removal of user groups, we acquire a lock for the user group to be removed with all its direct and indirect subgroups. This is the special case A=B, which is still complaint with the protocol. **Error handling** We currently rely on Postgres' deadlock detection to abort transactions and show an error for the users. In the future, we might need some recovery mechanism or at least better error handling. **Notes** An important note is that we need to reuse the recursive CTE query that finds the direct and indirect subgroups when applying the lock on the rows. And the lock needs to be acquired the same way for the addition and removal of direct subgroups. User membership change (as opposed to user group membership) is not affected. Read-only queries aren't either. The locks only protect critical regions where the user group dependency graph might violate the DAG constraint, where users are not participating. **Testing** We implement a transaction test case targeting some typical scenarios when an internal server error is expected to happen (this means that the user group view makes the correct decision to abort the transaction when something goes wrong with locks). To achieve this, we add a development view intended only for unit tests. It has a global BARRIER that can be shared across threads, so that we can synchronize them to consistently reproduce certain potential race conditions prevented by the database locks. The transaction test case lanuches pairs of threads initiating possibly conflicting requests at the same time. The tests are set up such that exactly N of them are expected to succeed with a certain error message (while we don't know each one). **Security notes** get_recursive_subgroups_for_groups will no longer fetch user groups from other realms. As a result, trying to add/remove a subgroup from another realm results in a UserGroup not found error response. We also implement subgroup-specific checks in has_user_group_access to keep permission managing in a single place. Do note that the API currently don't have a way to violate that check because we are only checking the realm ID now.
2023-06-17 04:39:52 +02:00
Regardless of whether the user groups returned are used, always call this
helper before making changes to subgroup memberships. This avoids
introducing cycles among user groups when there is a race condition in
which one of these subgroups become an ancestor of the parent user group in
another transaction.
Note that it only does a permission check on the potential supergroup,
not the potential subgroups or their recursive subgroups.
"""
with transaction.atomic(savepoint=False):
# Calling list with the QuerySet forces its evaluation putting a lock on
# the queried rows.
recursive_subgroups = list(
get_recursive_subgroups_for_groups(
potential_subgroup_ids, acting_user.realm
).select_for_update(nowait=True)
)
user_groups: Make locks required for updating user group memberships. **Background** User groups are expected to comply with the DAG constraint for the many-to-many inter-group membership. The check for this constraint has to be performed recursively so that we can find all direct and indirect subgroups of the user group to be added. This kind of check is vulnerable to phantom reads which is possible at the default read committed isolation level because we cannot guarantee that the check is still valid when we are adding the subgroups to the user group. **Solution** To avoid having another transaction concurrently update one of the to-be-subgroup after the recursive check is done, and before the subgroup is added, we use SELECT FOR UPDATE to lock the user group rows. The lock needs to be acquired before a group membership change is about to occur before any check has been conducted. Suppose that we are adding subgroup B to supergroup A, the locking protocol is specified as follows: 1. Acquire a lock for B and all its direct and indirect subgroups. 2. Acquire a lock for A. For the removal of user groups, we acquire a lock for the user group to be removed with all its direct and indirect subgroups. This is the special case A=B, which is still complaint with the protocol. **Error handling** We currently rely on Postgres' deadlock detection to abort transactions and show an error for the users. In the future, we might need some recovery mechanism or at least better error handling. **Notes** An important note is that we need to reuse the recursive CTE query that finds the direct and indirect subgroups when applying the lock on the rows. And the lock needs to be acquired the same way for the addition and removal of direct subgroups. User membership change (as opposed to user group membership) is not affected. Read-only queries aren't either. The locks only protect critical regions where the user group dependency graph might violate the DAG constraint, where users are not participating. **Testing** We implement a transaction test case targeting some typical scenarios when an internal server error is expected to happen (this means that the user group view makes the correct decision to abort the transaction when something goes wrong with locks). To achieve this, we add a development view intended only for unit tests. It has a global BARRIER that can be shared across threads, so that we can synchronize them to consistently reproduce certain potential race conditions prevented by the database locks. The transaction test case lanuches pairs of threads initiating possibly conflicting requests at the same time. The tests are set up such that exactly N of them are expected to succeed with a certain error message (while we don't know each one). **Security notes** get_recursive_subgroups_for_groups will no longer fetch user groups from other realms. As a result, trying to add/remove a subgroup from another realm results in a UserGroup not found error response. We also implement subgroup-specific checks in has_user_group_access to keep permission managing in a single place. Do note that the API currently don't have a way to violate that check because we are only checking the realm ID now.
2023-06-17 04:39:52 +02:00
# TODO: This select_for_update query is subject to deadlocking, and
# better error handling is needed. We may use
# select_for_update(nowait=True) and release the locks held by ending
# the transaction with a JsonableError by handling the DatabaseError.
# But at the current scale of concurrent requests, we rely on
# Postgres's deadlock detection when it occurs.
potential_supergroup = access_user_group_for_update(
potential_supergroup_id, acting_user, permission_setting="can_manage_group"
user_groups: Make locks required for updating user group memberships. **Background** User groups are expected to comply with the DAG constraint for the many-to-many inter-group membership. The check for this constraint has to be performed recursively so that we can find all direct and indirect subgroups of the user group to be added. This kind of check is vulnerable to phantom reads which is possible at the default read committed isolation level because we cannot guarantee that the check is still valid when we are adding the subgroups to the user group. **Solution** To avoid having another transaction concurrently update one of the to-be-subgroup after the recursive check is done, and before the subgroup is added, we use SELECT FOR UPDATE to lock the user group rows. The lock needs to be acquired before a group membership change is about to occur before any check has been conducted. Suppose that we are adding subgroup B to supergroup A, the locking protocol is specified as follows: 1. Acquire a lock for B and all its direct and indirect subgroups. 2. Acquire a lock for A. For the removal of user groups, we acquire a lock for the user group to be removed with all its direct and indirect subgroups. This is the special case A=B, which is still complaint with the protocol. **Error handling** We currently rely on Postgres' deadlock detection to abort transactions and show an error for the users. In the future, we might need some recovery mechanism or at least better error handling. **Notes** An important note is that we need to reuse the recursive CTE query that finds the direct and indirect subgroups when applying the lock on the rows. And the lock needs to be acquired the same way for the addition and removal of direct subgroups. User membership change (as opposed to user group membership) is not affected. Read-only queries aren't either. The locks only protect critical regions where the user group dependency graph might violate the DAG constraint, where users are not participating. **Testing** We implement a transaction test case targeting some typical scenarios when an internal server error is expected to happen (this means that the user group view makes the correct decision to abort the transaction when something goes wrong with locks). To achieve this, we add a development view intended only for unit tests. It has a global BARRIER that can be shared across threads, so that we can synchronize them to consistently reproduce certain potential race conditions prevented by the database locks. The transaction test case lanuches pairs of threads initiating possibly conflicting requests at the same time. The tests are set up such that exactly N of them are expected to succeed with a certain error message (while we don't know each one). **Security notes** get_recursive_subgroups_for_groups will no longer fetch user groups from other realms. As a result, trying to add/remove a subgroup from another realm results in a UserGroup not found error response. We also implement subgroup-specific checks in has_user_group_access to keep permission managing in a single place. Do note that the API currently don't have a way to violate that check because we are only checking the realm ID now.
2023-06-17 04:39:52 +02:00
)
# We avoid making a separate query for user_group_ids because the
# recursive query already returns those user groups.
potential_subgroups = [
user_group
for user_group in recursive_subgroups
if user_group.id in potential_subgroup_ids
]
# We expect that the passed user_group_ids each corresponds to an
# existing user group.
group_ids_found = [group.id for group in potential_subgroups]
group_ids_not_found = [
group_id for group_id in potential_subgroup_ids if group_id not in group_ids_found
]
if group_ids_not_found:
raise JsonableError(
_("Invalid user group ID: {group_id}").format(group_id=group_ids_not_found[0])
)
user_groups: Make locks required for updating user group memberships. **Background** User groups are expected to comply with the DAG constraint for the many-to-many inter-group membership. The check for this constraint has to be performed recursively so that we can find all direct and indirect subgroups of the user group to be added. This kind of check is vulnerable to phantom reads which is possible at the default read committed isolation level because we cannot guarantee that the check is still valid when we are adding the subgroups to the user group. **Solution** To avoid having another transaction concurrently update one of the to-be-subgroup after the recursive check is done, and before the subgroup is added, we use SELECT FOR UPDATE to lock the user group rows. The lock needs to be acquired before a group membership change is about to occur before any check has been conducted. Suppose that we are adding subgroup B to supergroup A, the locking protocol is specified as follows: 1. Acquire a lock for B and all its direct and indirect subgroups. 2. Acquire a lock for A. For the removal of user groups, we acquire a lock for the user group to be removed with all its direct and indirect subgroups. This is the special case A=B, which is still complaint with the protocol. **Error handling** We currently rely on Postgres' deadlock detection to abort transactions and show an error for the users. In the future, we might need some recovery mechanism or at least better error handling. **Notes** An important note is that we need to reuse the recursive CTE query that finds the direct and indirect subgroups when applying the lock on the rows. And the lock needs to be acquired the same way for the addition and removal of direct subgroups. User membership change (as opposed to user group membership) is not affected. Read-only queries aren't either. The locks only protect critical regions where the user group dependency graph might violate the DAG constraint, where users are not participating. **Testing** We implement a transaction test case targeting some typical scenarios when an internal server error is expected to happen (this means that the user group view makes the correct decision to abort the transaction when something goes wrong with locks). To achieve this, we add a development view intended only for unit tests. It has a global BARRIER that can be shared across threads, so that we can synchronize them to consistently reproduce certain potential race conditions prevented by the database locks. The transaction test case lanuches pairs of threads initiating possibly conflicting requests at the same time. The tests are set up such that exactly N of them are expected to succeed with a certain error message (while we don't know each one). **Security notes** get_recursive_subgroups_for_groups will no longer fetch user groups from other realms. As a result, trying to add/remove a subgroup from another realm results in a UserGroup not found error response. We also implement subgroup-specific checks in has_user_group_access to keep permission managing in a single place. Do note that the API currently don't have a way to violate that check because we are only checking the realm ID now.
2023-06-17 04:39:52 +02:00
for subgroup in potential_subgroups:
# At this time, we only do a check on the realm ID of the subgroup and
# whether the group is deactivated or not. Realm ID error would be caught
# above and in case the user group is deactivated the error will be raised
# in has_user_group_access_for_subgroup itself, so there is no coverage here.
if not has_user_group_access_for_subgroup(subgroup, acting_user):
user_groups: Make locks required for updating user group memberships. **Background** User groups are expected to comply with the DAG constraint for the many-to-many inter-group membership. The check for this constraint has to be performed recursively so that we can find all direct and indirect subgroups of the user group to be added. This kind of check is vulnerable to phantom reads which is possible at the default read committed isolation level because we cannot guarantee that the check is still valid when we are adding the subgroups to the user group. **Solution** To avoid having another transaction concurrently update one of the to-be-subgroup after the recursive check is done, and before the subgroup is added, we use SELECT FOR UPDATE to lock the user group rows. The lock needs to be acquired before a group membership change is about to occur before any check has been conducted. Suppose that we are adding subgroup B to supergroup A, the locking protocol is specified as follows: 1. Acquire a lock for B and all its direct and indirect subgroups. 2. Acquire a lock for A. For the removal of user groups, we acquire a lock for the user group to be removed with all its direct and indirect subgroups. This is the special case A=B, which is still complaint with the protocol. **Error handling** We currently rely on Postgres' deadlock detection to abort transactions and show an error for the users. In the future, we might need some recovery mechanism or at least better error handling. **Notes** An important note is that we need to reuse the recursive CTE query that finds the direct and indirect subgroups when applying the lock on the rows. And the lock needs to be acquired the same way for the addition and removal of direct subgroups. User membership change (as opposed to user group membership) is not affected. Read-only queries aren't either. The locks only protect critical regions where the user group dependency graph might violate the DAG constraint, where users are not participating. **Testing** We implement a transaction test case targeting some typical scenarios when an internal server error is expected to happen (this means that the user group view makes the correct decision to abort the transaction when something goes wrong with locks). To achieve this, we add a development view intended only for unit tests. It has a global BARRIER that can be shared across threads, so that we can synchronize them to consistently reproduce certain potential race conditions prevented by the database locks. The transaction test case lanuches pairs of threads initiating possibly conflicting requests at the same time. The tests are set up such that exactly N of them are expected to succeed with a certain error message (while we don't know each one). **Security notes** get_recursive_subgroups_for_groups will no longer fetch user groups from other realms. As a result, trying to add/remove a subgroup from another realm results in a UserGroup not found error response. We also implement subgroup-specific checks in has_user_group_access to keep permission managing in a single place. Do note that the API currently don't have a way to violate that check because we are only checking the realm ID now.
2023-06-17 04:39:52 +02:00
raise JsonableError(_("Insufficient permission")) # nocoverage
yield LockedUserGroupContext(
direct_subgroups=potential_subgroups,
recursive_subgroups=recursive_subgroups,
supergroup=potential_supergroup,
)
def check_setting_configuration_for_system_groups(
setting_group: NamedUserGroup,
setting_name: str,
permission_configuration: GroupPermissionSetting,
) -> None:
if setting_name != "can_mention_group" and (
not settings.ALLOW_GROUP_VALUED_SETTINGS and not setting_group.is_system_group
):
raise SystemGroupRequiredError(setting_name)
if permission_configuration.require_system_group and not setting_group.is_system_group:
raise SystemGroupRequiredError(setting_name)
if (
not permission_configuration.allow_internet_group
and setting_group.name == SystemGroups.EVERYONE_ON_INTERNET
):
raise JsonableError(
_("'{setting_name}' setting cannot be set to 'role:internet' group.").format(
setting_name=setting_name
)
)
if (
not permission_configuration.allow_owners_group
and setting_group.name == SystemGroups.OWNERS
):
raise JsonableError(
_("'{setting_name}' setting cannot be set to 'role:owners' group.").format(
setting_name=setting_name
)
)
if (
not permission_configuration.allow_nobody_group
and setting_group.name == SystemGroups.NOBODY
):
raise JsonableError(
_("'{setting_name}' setting cannot be set to 'role:nobody' group.").format(
setting_name=setting_name
)
)
if (
not permission_configuration.allow_everyone_group
and setting_group.name == SystemGroups.EVERYONE
):
raise JsonableError(
_("'{setting_name}' setting cannot be set to 'role:everyone' group.").format(
setting_name=setting_name
)
)
if (
permission_configuration.allowed_system_groups
and setting_group.name not in permission_configuration.allowed_system_groups
):
raise JsonableError(
_("'{setting_name}' setting cannot be set to '{group_name}' group.").format(
setting_name=setting_name, group_name=setting_group.name
)
)
def update_or_create_user_group_for_setting(
user_profile: UserProfile,
direct_members: list[int],
direct_subgroups: list[int],
current_setting_value: UserGroup | None,
) -> UserGroup:
realm = user_profile.realm
if current_setting_value is not None and not hasattr(current_setting_value, "named_user_group"):
# We do not create a new group if the setting was already set
# to an anonymous group. The memberships of existing group
# itself are updated.
user_group = current_setting_value
else:
user_group = UserGroup.objects.create(realm=realm)
from zerver.lib.users import user_ids_to_users
member_users = user_ids_to_users(direct_members, realm)
user_group.direct_members.set(member_users)
potential_subgroups = NamedUserGroup.objects.select_for_update().filter(
realm=realm, id__in=direct_subgroups
)
group_ids_found = [group.id for group in potential_subgroups]
group_ids_not_found = [
group_id for group_id in direct_subgroups if group_id not in group_ids_found
]
if group_ids_not_found:
raise JsonableError(
_("Invalid user group ID: {group_id}").format(group_id=group_ids_not_found[0])
)
for subgroup in potential_subgroups:
# At this time, we only do a check on the realm ID of the subgroup and
# whether the group is deactivated or not. Realm ID error would be caught
# above and in case the user group is deactivated the error will be raised
# in has_user_group_access_for_subgroup itself, so there is no coverage here.
if not has_user_group_access_for_subgroup(subgroup, user_profile):
raise JsonableError(_("Insufficient permission")) # nocoverage
user_group.direct_subgroups.set(group_ids_found)
return user_group
def access_user_group_for_setting(
setting_user_group: int | AnonymousSettingGroupDict,
user_profile: UserProfile,
*,
setting_name: str,
permission_configuration: GroupPermissionSetting,
current_setting_value: UserGroup | None = None,
) -> UserGroup:
"""Given a permission setting and specification of what value it
should have (setting_user_group), returns either a Named or
anonymous `UserGroup` with the requested membership.
"""
if isinstance(setting_user_group, int):
named_user_group = get_user_group_by_id_in_realm(
setting_user_group, user_profile.realm, for_read=False, for_setting=True
)
check_setting_configuration_for_system_groups(
named_user_group, setting_name, permission_configuration
)
return named_user_group.usergroup_ptr
if permission_configuration.require_system_group:
raise SystemGroupRequiredError(setting_name)
user_group = update_or_create_user_group_for_setting(
user_profile,
setting_user_group.direct_members,
setting_user_group.direct_subgroups,
current_setting_value,
)
return user_group
def check_user_group_name(group_name: str) -> str:
if group_name.strip() == "":
raise JsonableError(_("User group name can't be empty!"))
if len(group_name) > NamedUserGroup.MAX_NAME_LENGTH:
raise JsonableError(
_("User group name cannot exceed {max_length} characters.").format(
max_length=NamedUserGroup.MAX_NAME_LENGTH
)
)
for invalid_prefix in NamedUserGroup.INVALID_NAME_PREFIXES:
if group_name.startswith(invalid_prefix):
raise JsonableError(
_("User group name cannot start with '{prefix}'.").format(prefix=invalid_prefix)
)
return group_name
def get_group_setting_value_for_api(
setting_value_group: UserGroup,
) -> int | AnonymousSettingGroupDict:
if hasattr(setting_value_group, "named_user_group"):
return setting_value_group.id
return AnonymousSettingGroupDict(
direct_members=[member.id for member in setting_value_group.direct_members.all()],
direct_subgroups=[subgroup.id for subgroup in setting_value_group.direct_subgroups.all()],
)
def get_setting_value_for_user_group_object(
setting_value_group: UserGroup,
direct_members_dict: dict[int, list[int]],
direct_subgroups_dict: dict[int, list[int]],
) -> int | AnonymousSettingGroupDict:
if hasattr(setting_value_group, "named_user_group"):
return setting_value_group.id
direct_members = []
if setting_value_group.id in direct_members_dict:
direct_members = direct_members_dict[setting_value_group.id]
direct_subgroups = []
if setting_value_group.id in direct_subgroups_dict:
direct_subgroups = direct_subgroups_dict[setting_value_group.id]
return AnonymousSettingGroupDict(
direct_members=direct_members,
direct_subgroups=direct_subgroups,
)
def user_groups_in_realm_serialized(
realm: Realm, *, include_deactivated_groups: bool
) -> list[UserGroupDict]:
"""This function is used in do_events_register code path so this code
should be performant. We need to do 2 database queries because
Django's ORM doesn't properly support the left join between
UserGroup and UserGroupMembership that we need.
"""
realm_groups = NamedUserGroup.objects.select_related(
"can_manage_group",
"can_manage_group__named_user_group",
"can_mention_group",
"can_mention_group__named_user_group",
).filter(realm=realm)
if not include_deactivated_groups:
realm_groups = realm_groups.filter(deactivated=False)
membership = UserGroupMembership.objects.filter(user_group__realm=realm).values_list(
"user_group_id", "user_profile_id"
)
group_membership = GroupGroupMembership.objects.filter(subgroup__realm=realm).values_list(
"subgroup_id", "supergroup_id"
)
group_members = defaultdict(list)
for user_group_id, user_profile_id in membership:
group_members[user_group_id].append(user_profile_id)
group_subgroups = defaultdict(list)
for subgroup_id, supergroup_id in group_membership:
group_subgroups[supergroup_id].append(subgroup_id)
group_dicts: dict[int, UserGroupDict] = {}
for user_group in realm_groups:
direct_member_ids = []
if user_group.id in group_members:
direct_member_ids = group_members[user_group.id]
direct_subgroup_ids = []
if user_group.id in group_subgroups:
direct_subgroup_ids = group_subgroups[user_group.id]
creator_id = user_group.creator_id
date_created = (
datetime_to_timestamp(user_group.date_created)
if user_group.date_created is not None
else None
)
group_dicts[user_group.id] = dict(
id=user_group.id,
name=user_group.name,
creator_id=creator_id,
date_created=date_created,
description=user_group.description,
members=direct_member_ids,
direct_subgroup_ids=direct_subgroup_ids,
is_system_group=user_group.is_system_group,
can_manage_group=get_setting_value_for_user_group_object(
user_group.can_manage_group, group_members, group_subgroups
),
can_mention_group=get_setting_value_for_user_group_object(
user_group.can_mention_group, group_members, group_subgroups
),
deactivated=user_group.deactivated,
)
for group_dict in group_dicts.values():
group_dict["members"] = sorted(group_dict["members"])
group_dict["direct_subgroup_ids"] = sorted(group_dict["direct_subgroup_ids"])
return sorted(group_dicts.values(), key=lambda group_dict: group_dict["id"])
def get_direct_user_groups(user_profile: UserProfile) -> list[UserGroup]:
return list(user_profile.direct_groups.all())
def get_user_group_direct_member_ids(
user_group: UserGroup,
) -> QuerySet[UserGroupMembership, int]:
return UserGroupMembership.objects.filter(user_group=user_group).values_list(
"user_profile_id", flat=True
)
def get_user_group_direct_members(user_group: UserGroup) -> QuerySet[UserProfile]:
return user_group.direct_members.all()
def get_direct_memberships_of_users(user_group: UserGroup, members: list[UserProfile]) -> list[int]:
return list(
UserGroupMembership.objects.filter(
user_group=user_group, user_profile__in=members
).values_list("user_profile_id", flat=True)
)
# These recursive lookups use standard PostgreSQL common table
# expression (CTE) queries. These queries use the django-cte library,
# because upstream Django does not yet support CTE.
#
# https://www.postgresql.org/docs/current/queries-with.html
# https://pypi.org/project/django-cte/
# https://code.djangoproject.com/ticket/28919
def get_recursive_subgroups(user_group: UserGroup) -> QuerySet[UserGroup]:
cte = With.recursive(
lambda cte: UserGroup.objects.filter(id=user_group.id)
.values(group_id=F("id"))
.union(
cte.join(NamedUserGroup, direct_supergroups=cte.col.group_id).values(group_id=F("id"))
)
)
return cte.join(UserGroup, id=cte.col.group_id).with_cte(cte)
def get_recursive_strict_subgroups(user_group: UserGroup) -> QuerySet[NamedUserGroup]:
# Same as get_recursive_subgroups but does not include the
# user_group passed.
direct_subgroup_ids = user_group.direct_subgroups.all().values("id")
cte = With.recursive(
lambda cte: NamedUserGroup.objects.filter(id__in=direct_subgroup_ids)
.values(group_id=F("id"))
.union(
cte.join(NamedUserGroup, direct_supergroups=cte.col.group_id).values(group_id=F("id"))
)
)
return cte.join(NamedUserGroup, id=cte.col.group_id).with_cte(cte)
def get_recursive_group_members(user_group: UserGroup) -> QuerySet[UserProfile]:
return UserProfile.objects.filter(direct_groups__in=get_recursive_subgroups(user_group))
def get_recursive_membership_groups(user_profile: UserProfile) -> QuerySet[UserGroup]:
cte = With.recursive(
lambda cte: user_profile.direct_groups.values(group_id=F("id")).union(
cte.join(UserGroup, direct_subgroups=cte.col.group_id).values(group_id=F("id"))
)
)
return cte.join(UserGroup, id=cte.col.group_id).with_cte(cte)
def user_has_permission_for_group_setting(
user_group: UserGroup,
user: UserProfile,
setting_config: GroupPermissionSetting,
*,
direct_member_only: bool = False,
) -> bool:
if not setting_config.allow_everyone_group and user.is_guest:
return False
return is_user_in_group(user_group, user, direct_member_only=direct_member_only)
def is_user_in_group(
user_group: UserGroup, user: UserProfile, *, direct_member_only: bool = False
) -> bool:
if direct_member_only:
return get_user_group_direct_members(user_group=user_group).filter(id=user.id).exists()
return get_recursive_group_members(user_group=user_group).filter(id=user.id).exists()
def is_any_user_in_group(
user_group: UserGroup, user_ids: Iterable[int], *, direct_member_only: bool = False
) -> bool:
if direct_member_only:
return get_user_group_direct_members(user_group=user_group).filter(id__in=user_ids).exists()
return get_recursive_group_members(user_group=user_group).filter(id__in=user_ids).exists()
def get_user_group_member_ids(
user_group: UserGroup, *, direct_member_only: bool = False
) -> list[int]:
if direct_member_only:
member_ids: Iterable[int] = get_user_group_direct_member_ids(user_group)
else:
member_ids = get_recursive_group_members(user_group).values_list("id", flat=True)
return list(member_ids)
def get_subgroup_ids(user_group: UserGroup, *, direct_subgroup_only: bool = False) -> list[int]:
if direct_subgroup_only:
subgroup_ids = user_group.direct_subgroups.all().values_list("id", flat=True)
else:
subgroup_ids = get_recursive_strict_subgroups(user_group).values_list("id", flat=True)
return list(subgroup_ids)
user_groups: Make locks required for updating user group memberships. **Background** User groups are expected to comply with the DAG constraint for the many-to-many inter-group membership. The check for this constraint has to be performed recursively so that we can find all direct and indirect subgroups of the user group to be added. This kind of check is vulnerable to phantom reads which is possible at the default read committed isolation level because we cannot guarantee that the check is still valid when we are adding the subgroups to the user group. **Solution** To avoid having another transaction concurrently update one of the to-be-subgroup after the recursive check is done, and before the subgroup is added, we use SELECT FOR UPDATE to lock the user group rows. The lock needs to be acquired before a group membership change is about to occur before any check has been conducted. Suppose that we are adding subgroup B to supergroup A, the locking protocol is specified as follows: 1. Acquire a lock for B and all its direct and indirect subgroups. 2. Acquire a lock for A. For the removal of user groups, we acquire a lock for the user group to be removed with all its direct and indirect subgroups. This is the special case A=B, which is still complaint with the protocol. **Error handling** We currently rely on Postgres' deadlock detection to abort transactions and show an error for the users. In the future, we might need some recovery mechanism or at least better error handling. **Notes** An important note is that we need to reuse the recursive CTE query that finds the direct and indirect subgroups when applying the lock on the rows. And the lock needs to be acquired the same way for the addition and removal of direct subgroups. User membership change (as opposed to user group membership) is not affected. Read-only queries aren't either. The locks only protect critical regions where the user group dependency graph might violate the DAG constraint, where users are not participating. **Testing** We implement a transaction test case targeting some typical scenarios when an internal server error is expected to happen (this means that the user group view makes the correct decision to abort the transaction when something goes wrong with locks). To achieve this, we add a development view intended only for unit tests. It has a global BARRIER that can be shared across threads, so that we can synchronize them to consistently reproduce certain potential race conditions prevented by the database locks. The transaction test case lanuches pairs of threads initiating possibly conflicting requests at the same time. The tests are set up such that exactly N of them are expected to succeed with a certain error message (while we don't know each one). **Security notes** get_recursive_subgroups_for_groups will no longer fetch user groups from other realms. As a result, trying to add/remove a subgroup from another realm results in a UserGroup not found error response. We also implement subgroup-specific checks in has_user_group_access to keep permission managing in a single place. Do note that the API currently don't have a way to violate that check because we are only checking the realm ID now.
2023-06-17 04:39:52 +02:00
def get_recursive_subgroups_for_groups(
user_group_ids: Iterable[int], realm: Realm
) -> QuerySet[NamedUserGroup]:
cte = With.recursive(
lambda cte: NamedUserGroup.objects.filter(id__in=user_group_ids, realm=realm)
.values(group_id=F("id"))
.union(
cte.join(NamedUserGroup, direct_supergroups=cte.col.group_id).values(group_id=F("id"))
)
)
recursive_subgroups = cte.join(NamedUserGroup, id=cte.col.group_id).with_cte(cte)
return recursive_subgroups
def get_role_based_system_groups_dict(realm: Realm) -> dict[str, NamedUserGroup]:
system_groups = NamedUserGroup.objects.filter(realm=realm, is_system_group=True).select_related(
"usergroup_ptr"
)
system_groups_name_dict = {}
for group in system_groups:
system_groups_name_dict[group.name] = group
return system_groups_name_dict
def set_defaults_for_group_settings(
user_group: NamedUserGroup,
group_settings_map: Mapping[str, UserGroup],
system_groups_name_dict: dict[str, NamedUserGroup],
) -> NamedUserGroup:
for setting_name, permission_config in NamedUserGroup.GROUP_PERMISSION_SETTINGS.items():
if setting_name in group_settings_map:
# We skip the settings for which a value is passed
# in user group creation API request.
continue
if user_group.is_system_group and permission_config.default_for_system_groups is not None:
default_group_name = permission_config.default_for_system_groups
else:
default_group_name = permission_config.default_group_name
default_group = system_groups_name_dict[default_group_name].usergroup_ptr
setattr(user_group, setting_name, default_group)
return user_group
def bulk_create_system_user_groups(groups: list[dict[str, str]], realm: Realm) -> None:
# This value will be used to set the temporary initial value for different
# settings since we can only set them to the correct values after the groups
# are created.
initial_group_setting_value = -1
rows = [SQL("({})").format(Literal(realm.id))] * len(groups)
query = SQL(
"""
INSERT INTO zerver_usergroup (realm_id)
VALUES {rows}
RETURNING id
"""
).format(rows=SQL(", ").join(rows))
with connection.cursor() as cursor:
cursor.execute(query)
user_group_ids = [id for (id,) in cursor.fetchall()]
rows = [
SQL("({},{},{},{},{},{},{},{})").format(
Literal(user_group_ids[idx]),
Literal(realm.id),
Literal(group["name"]),
Literal(group["description"]),
Literal(True),
Literal(initial_group_setting_value),
Literal(initial_group_setting_value),
Literal(False),
)
for idx, group in enumerate(groups)
]
query = SQL(
"""
INSERT INTO zerver_namedusergroup (usergroup_ptr_id, realm_id, name, description, is_system_group, can_manage_group_id, can_mention_group_id, deactivated)
VALUES {rows}
"""
).format(rows=SQL(", ").join(rows))
with connection.cursor() as cursor:
cursor.execute(query)
@transaction.atomic(savepoint=False)
def create_system_user_groups_for_realm(realm: Realm) -> dict[int, NamedUserGroup]:
"""Any changes to this function likely require a migration to adjust
existing realms. See e.g. migration 0382_create_role_based_system_groups.py,
which is a copy of this function from when we introduced system groups.
"""
role_system_groups_dict: dict[int, NamedUserGroup] = {}
system_groups_info_list: list[dict[str, str]] = []
nobody_group_info = {
"name": SystemGroups.NOBODY,
"description": "Nobody",
}
full_members_group_info = {
"name": SystemGroups.FULL_MEMBERS,
"description": "Members of this organization, not including new accounts and guests",
}
everyone_on_internet_group_info = {
"name": SystemGroups.EVERYONE_ON_INTERNET,
"description": "Everyone on the Internet",
}
system_groups_info_list = [
nobody_group_info,
NamedUserGroup.SYSTEM_USER_GROUP_ROLE_MAP[UserProfile.ROLE_REALM_OWNER],
NamedUserGroup.SYSTEM_USER_GROUP_ROLE_MAP[UserProfile.ROLE_REALM_ADMINISTRATOR],
NamedUserGroup.SYSTEM_USER_GROUP_ROLE_MAP[UserProfile.ROLE_MODERATOR],
full_members_group_info,
NamedUserGroup.SYSTEM_USER_GROUP_ROLE_MAP[UserProfile.ROLE_MEMBER],
NamedUserGroup.SYSTEM_USER_GROUP_ROLE_MAP[UserProfile.ROLE_GUEST],
everyone_on_internet_group_info,
]
bulk_create_system_user_groups(system_groups_info_list, realm)
system_groups_name_dict: dict[str, NamedUserGroup] = get_role_based_system_groups_dict(realm)
for role in NamedUserGroup.SYSTEM_USER_GROUP_ROLE_MAP:
group_name = NamedUserGroup.SYSTEM_USER_GROUP_ROLE_MAP[role]["name"]
role_system_groups_dict[role] = system_groups_name_dict[group_name]
# Order of this list here is important to create correct GroupGroupMembership objects
# Note that because we do not create user memberships here, no audit log entries for
# user memberships are populated either.
system_user_groups_list = [
system_groups_name_dict[SystemGroups.NOBODY],
system_groups_name_dict[SystemGroups.OWNERS],
system_groups_name_dict[SystemGroups.ADMINISTRATORS],
system_groups_name_dict[SystemGroups.MODERATORS],
system_groups_name_dict[SystemGroups.FULL_MEMBERS],
system_groups_name_dict[SystemGroups.MEMBERS],
system_groups_name_dict[SystemGroups.EVERYONE],
system_groups_name_dict[SystemGroups.EVERYONE_ON_INTERNET],
]
creation_time = timezone_now()
realmauditlog_objects = [
RealmAuditLog(
realm=realm,
acting_user=None,
event_type=AuditLogEventType.USER_GROUP_CREATED,
event_time=creation_time,
modified_user_group=user_group,
)
for user_group in system_user_groups_list
]
groups_with_updated_settings = []
for group in system_user_groups_list:
user_group = set_defaults_for_group_settings(group, {}, system_groups_name_dict)
groups_with_updated_settings.append(user_group)
NamedUserGroup.objects.bulk_update(
groups_with_updated_settings, ["can_manage_group", "can_mention_group"]
)
subgroup_objects: list[GroupGroupMembership] = []
# "Nobody" system group is not a subgroup of any user group, since it is already empty.
subgroup, remaining_groups = system_user_groups_list[1], system_user_groups_list[2:]
for supergroup in remaining_groups:
subgroup_objects.append(GroupGroupMembership(subgroup=subgroup, supergroup=supergroup))
now = timezone_now()
realmauditlog_objects.extend(
[
RealmAuditLog(
realm=realm,
modified_user_group=supergroup,
event_type=AuditLogEventType.USER_GROUP_DIRECT_SUBGROUP_MEMBERSHIP_ADDED,
event_time=now,
acting_user=None,
extra_data={"subgroup_ids": [subgroup.id]},
),
RealmAuditLog(
realm=realm,
modified_user_group=subgroup,
event_type=AuditLogEventType.USER_GROUP_DIRECT_SUPERGROUP_MEMBERSHIP_ADDED,
event_time=now,
acting_user=None,
extra_data={"supergroup_ids": [supergroup.id]},
),
]
)
subgroup = supergroup
GroupGroupMembership.objects.bulk_create(subgroup_objects)
RealmAuditLog.objects.bulk_create(realmauditlog_objects)
return role_system_groups_dict
def get_system_user_group_for_user(user_profile: UserProfile) -> NamedUserGroup:
system_user_group_name = NamedUserGroup.SYSTEM_USER_GROUP_ROLE_MAP[user_profile.role]["name"]
system_user_group = NamedUserGroup.objects.get(
name=system_user_group_name, realm=user_profile.realm, is_system_group=True
)
return system_user_group
def get_server_supported_permission_settings() -> ServerSupportedPermissionSettings:
return ServerSupportedPermissionSettings(
realm=Realm.REALM_PERMISSION_GROUP_SETTINGS,
stream=Stream.stream_permission_group_settings,
group=NamedUserGroup.GROUP_PERMISSION_SETTINGS,
)
def parse_group_setting_value(
setting_value: int | AnonymousSettingGroupDict,
setting_name: str,
) -> int | AnonymousSettingGroupDict:
if isinstance(setting_value, int):
return setting_value
if len(setting_value.direct_members) == 0 and len(setting_value.direct_subgroups) == 1:
return setting_value.direct_subgroups[0]
if not settings.ALLOW_GROUP_VALUED_SETTINGS:
raise SystemGroupRequiredError(setting_name)
return setting_value
def are_both_group_setting_values_equal(
first_setting_value: int | AnonymousSettingGroupDict,
second_setting_value: int | AnonymousSettingGroupDict,
) -> bool:
if isinstance(first_setting_value, int) and isinstance(second_setting_value, int):
return first_setting_value == second_setting_value
if isinstance(first_setting_value, AnonymousSettingGroupDict) and isinstance(
second_setting_value, AnonymousSettingGroupDict
):
return set(first_setting_value.direct_members) == set(
second_setting_value.direct_members
) and set(first_setting_value.direct_subgroups) == set(
second_setting_value.direct_subgroups
)
return False
def validate_group_setting_value_change(
current_setting_api_value: int | AnonymousSettingGroupDict,
new_setting_value: int | AnonymousSettingGroupDict,
expected_current_setting_value: int | AnonymousSettingGroupDict | None,
) -> bool:
if expected_current_setting_value is not None and not are_both_group_setting_values_equal(
expected_current_setting_value,
current_setting_api_value,
):
# This check is here to help prevent races, by refusing to
# change a setting where the client (and thus the UI presented
# to user) showed a different existing state.
raise PreviousSettingValueMismatchedError
return not are_both_group_setting_values_equal(current_setting_api_value, new_setting_value)
def get_group_setting_value_for_audit_log_data(
setting_value: int | AnonymousSettingGroupDict,
) -> int | dict[str, list[int]]:
if isinstance(setting_value, int):
return setting_value
return asdict(setting_value)