Skip to content

linux-test-project/usfstl

Repository files navigation

USFSTL

Introduction

This is the User Space Firmware Simulation Testing Library ("USFSTL") -- a simulation (and to some extent unit) testing framework (mostly) intended for firmware.

The library also integrates with User-Mode-Linux and vhost-user to allow simulating devices and testing them against a real (virtio) driver running in Linux. If you came here just for this, take a look at the vlab/ directory and the code under control/ first.

License

This library/code is licensed under BSD-3-Clause, see the LICENSE file.

Contributions

We welcome contributions to this code, please take a look at the CONTRIBUTING.md file.

Capabilities

The library (and associated tools) offer the following:

Job Scheduler

There's support for scheduling "jobs", which are really just function callbacks to occur at a specific point in time. This might be useful by itself, but is mostly used together with the contexts to provide a cooperative multi-threading implementation (tasks).

The job scheduler can also be used together with User-Mode-Linux, we have a global scheduler for the "time-travel=ext:..." protocol that Linux supports. This tool (under control/) also supports virtio-net devices using the vhost-user capability.

Simulation-capable vhost-user support

With a few extensions to the vhost-user protocol that are implemented here and in the User-Mode-Linux virtio driver, the protocol becomes capable of doing time-independent simulation.

Contexts

Contexts are execution contexts, provided either by threads (pthread, available on Windows and Linux), fibers (Windows) or ucontexts (Linux). This allows having different pieces of code execute semi-concurrently. In this, context switching would have to be implemented manually.

Tasks / Semaphores

This is a full cooperative multi-threading implementation that allows switching between tasks, locking (semaphores), relative priorities, and (through interaction between the scheduler and tasks) grouping tasks and blocking groups for execution, e.g. to implement critical sections.

Function stubbing

This allows skipping or replacing any inner function calls that may be made, in order to test.

Static function calling

This allows calling a function that's marked static in the tested code (assuming you make the compiler not inline it.)

Assert mechanism

There are various assert macros that fail a test and print out a more or less verbose message, including stack dump.

Multi-process testing / RPC framework

The RPC framework is geared towards simulation and will automatically propagate the current time on RPC calls, etc.

Logging framework

A very basic logging framework exists that can take advantage of the RPC mechanisms to have central logging when multiple binaries are involved in a test.

Test runner

Multiple tests can be linked into the same binary (though they currently need to have the same tested files, outside of which function calls are not permitted), and the test runner will run and summarize the tests, this is provide as a main() function in the USFSTL library that gets linked with the test cases.

Global data handling

Global data is saved at the beginning and restored after each test run to ensure a clean state for the next test.

Security

We do not expect this library to be used in any context other than testing, and as such don't expect that there will be any security issues.

However, should you nevertheless find a security issue, please report it privately to johannes@sipsolutions.net, rather than filing a github ticket.