diff --git a/test/README b/test/README
new file mode 100644
index 0000000..dfd83d6
--- /dev/null
+++ b/test/README
@@ -0,0 +1,82 @@
+Testing in U-Boot
+=================
+
+U-Boot has a large amount of code. This file describes how this code is
+tested and what tests you should write when adding a new feature.
+
+
+Sandbox
+-------
+U-Boot can be built as a user-space application (e.g. for Linux). This
+allows test to be executed without needing target hardware. The 'sandbox'
+target provides this feature and it is widely used in tests.
+
+
+Pytest Suite
+------------
+
+Many tests are available using the pytest suite, in test/py. This can run
+either on sandbox or on real hardware. It relies on the U-Boot console to
+inject test commands and check the result. It is slower to run than C code,
+but provides the ability to unify lots of test and summarise their results.
+
+You can run the tests on sandbox with:
+
+	./test/py/test.py --bd sandbox --build
+
+This will produce HTML output in build-sandbox/test-log.html
+
+See test/py/README.md for more information about the pytest suite.
+
+
+tbot
+----
+
+Tbot provides a way to execute tests on target hardware. It is intended for
+trying out both U-Boot and Linux (and potentially other software) on a
+number of boards automatically. It can be used to create a continuous test
+environment. See tools/tbot/README for more information.
+
+
+Ad-hoc tests
+------------
+
+There are several ad-hoc tests which run outside the pytest environment:
+
+   test/fs	- File system test (shell script)
+   test/image	- FIT and lagacy image tests (shell script and Python)
+   test/stdint	- A test that stdint.h can be used in U-Boot (shell script)
+   trace	- Test for the tracing feature (shell script)
+   vboot	- Test for verified boot (shell script)
+
+The above should be converted to run as part of the pytest suite.
+
+
+When to write tests
+-------------------
+
+If you add code to U-Boot without a test you are taking a risk. Even if you
+perform thorough manual testing at the time of submission, it may break when
+future changes are made to U-Boot. It may even break when applied to mainline,
+if other changes interact with it. A good mindset is that untested code
+probably doesn't work and should be deleted.
+
+You can assume that the Pytest suite will be run before patches are accepted
+to mainline, so this provides protection against future breakage.
+
+On the other hand there is quite a bit of code that is not covered with tests,
+or is covered sparingly. So here are some suggestions:
+
+- If you are adding a new uclass, add a sandbox driver and a test that uses it
+- If you are modifying code covered by an existing test, add a new test case
+  to cover your changes
+- If the code you are modifying has not tests, consider writing one. Even a
+  very basic test is useful, and may be picked up and enhanced by others. It
+  is much easier to add onto a test - writing a new large test can seem
+  daunting to most contributors.
+
+
+Future work
+-----------
+
+Converting existing shell scripts into pytest tests.