| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
|
|
Add dependency `scopeguard` for cleaning up nftables ruleset.
|
|
`err_derive` is unmaintained and will probably stop working with rust
edition 2024. `thiserror` is almost a drop-in replacement. This commit
simply replaces all occurences of `derive(err_derive::Error)` with
`derive(thiserror::Error)` and fixes the attributes, but the Error and
Display impls should be identical.
|
|
|
|
|
|
|
|
Since `test_custom_access_methods_ui` may fail after messing with the
test runner environment, we want to fail fast and abort the entire test
run if the test fails before managing to clean up.
|
|
|
|
Add Shadowsocks & SOCKS5 (remote) access method tests. Simply try to
access the Mullvad API using these custom access methods.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add `locked_down` field to disconnected tunnel state.
|
|
|
|
|
|
state
|
|
|
|
Re-write some code in the test framework to prefer the type safe wrapper
around the Mullvad app gRPC client instead of its auto-generated dito.
`ManagementServiceClient` is automatically generated from the protobuf
definitions found in `management_interface.proto`, and contains some
very crude types. The `MullvadProxyClient` is a type-safe wrapper around
`ManagementServiceClient` which performs conversions & validation of the
data types from the gRPC server (the daemon) to their respective
mappings in the `talpid-*` and `mullvad-*` crates. These types are more
ergonomic to work with, and since we already have the conversions in
place we should prefer those.
|
|
|
|
feature is enabled
Move the logic for using overridden API endpoints for API calls from
`mullvad-api::rest` to `mullvad_daemon::api`. This is in line with how
the interaction between the two crates work for a normal release build,
i.e. when the `api-override` feature is disabled.
This commit also removes references to `force_direct_connection` in the
Android code. The flag does not exist in the `mullvad-*` rust crates
anymore, so it would be erroneous to try to serialize/deserialize the
value from the Android client.
|
|
|
|
|
|
|
|
Previously, the `mullvad-api` would tell the `mullvad-daemon` that it
wanted a new API endpoint by calling a certain callback
(`ApiEndpointUpdateCallback`), which would asynchronously resolve a new
API endpoint and tell the daemon to punch an appropriate hole in the
firewall for that particular endpoint before the `mullvad-api` crate
would consume it.
The logic of the callback can be moved inside `AccessModeSelector`,
which simplifies the contract between `mullvad-daemon` and `mullvad-api`
somewhat.
|
|
|
|
This PR has a couple of different purposes
- Allow users to use socks5 local proxies with the CLI without
having to be root nor use split-tunneling. This only works for
OpenVPN.
- Unify the types used by different proxy parts of the codebase,
such as the Access Methods as well as some already existing
OpenVPN proxy code.
This PR changes the firewall on all desktop platforms as well as changes
the routing table slightly on MacOS and Windows.
On Linux the firewall code is modified to apply the appropriate firewall
marks to all packages that go to a remote endpoint corresponding to the
remote part of a local socks5 proxy. The firewall marks will allow the
routing to be done without having to modify the routing table.
On MacOS and Windows the routing table is modified to allow packages to
go to that same endpoint to pass outside the VPN tunnel, it will
additionally punch a hole in the firewall.
The PR also migrates the settings file from version 7 to version 8 in order
to properly and neatly unify Proxy related types.
Finally it provides some slight extensions to the gRPC interface in
order to allow for control over the custom proxy settings.
|
|
|
|
Make the daemon send two tunnel state updates, one with out IP being
empty, and another with it being filled when am.i.mullvad.net responds.
Update CLI for this change. Other front ends are left out.
|
|
|
|
Flatten the inner loop of `start_packet_monitor_for_interface` by using
`tokio::select!` instead of composing `futures::select`
|
|
- Import qualified `network_monitor`
- Add `obtain_guest_ip`
- Fix docs
|
|
|
|
|
|
Installing the app twice (or more times) should have the same observable
effect as installing it once. If the app is upgraded and the user is in
a connected state before starting the upgrade, no traffic should leak
during the installation process and the app should start up in a
connected state preserving the relay constraints.
|
|
The function `set_mullvad_daemon_service_state(on: bool) -> Result<(),
test_rpc::Error>`, which would conditionally start or stop the Mullvad
daemon in the test runner, has been superseded by two separate functions which
accomplish the same thing: `start_mullvad_daemon` & `stop_mullvad_daemon`.
|
|
Add regression test which checks that the daemon successfully recovers
from a corrupt target state cache. If the target state cache is corrupt,
the daemon will default to the `Connecting` target state on startup.
|
|
- Implement RPC for writing to a file in a test runner / guest VM.
- Implement RPC for getting app cache directory
- Implement RPC for restarting the app in a test runner / guest vm
- Implement RPC for starting the app in a test runner / guest vm
- Implement RPC for stopping the app in a test runner / guest vm
- Implement `find_cache_traces` on Window & macOS
|
|
test_bridge)
|
|
|
|
We have seen `test_bridge` fail due to high latency if slow + far-away
servers were selected. Hopefully delegating the task of picking
appropriate entry & exit relays will help mitigate this.
|
|
Prefer types from `mullvad-types` over equivavelent types defined in
`mullvad-management-interface`.
|
|
|
|
|
|
|
|
|
