Skip to content

Latest commit

 

History

History
153 lines (115 loc) · 7.62 KB

README.md

File metadata and controls

153 lines (115 loc) · 7.62 KB
Raw

bpftrace Tests

There are three different types of tests: unit, runtime, and tool parsing.

Every contribution should (1) not break the existing tests and (2) introduce new tests if relevant.

Unit tests

Unit tests for individual components (semantic analyser, codegen, etc.) are based on the GoogleTest framework. These tests can be run with the bpftrace_test executable. Tests can be selected with the --gtest_filter flag or the GTEST_FILTER environment variable, see --help for more information. These are located in tests/*.cpp and are executed by <builddir>/tests/bpftrace_test.

Codegen tests

The codegen tests verify that the optimized IR matches our expectations.

The tests are defined as C++ files in the tests/codegen directory and look like:

TEST(codegen, call_avg)
{
  test("kprobe:f { @x = avg(pid) }", NAME);
}

The test function does all the heavy lifting and is defined in tests/codegen/common.h. It compiles the specified program (first argument) and compares it (string compare) with the expected result, a file named by the second argument. The NAME macro holds the test name, which is call_avg in this case.

These tests run as part of the normal suite of unit tests if you are running LLVM 12. If not, you need to install 'nix' and run these tests via this script: ./tests/codegen-tests.sh.

Updating

LLVM 12

If you are running LLVM 12 or want to only update specific tests with --gtest_filter run <builddir>/tests/bpftrace_test with BPFTRACE_UPDATE_TESTS=1 and the test helper will update the IR instead of running the tests.

Not LLVM 12

Run ./tests/codegen-tests.sh -u. This updates all the codegen tests.

Runtime tests

Runtime tests will call the bpftrace executable. These are located in tests/runtime and are managed by a custom framework.

  • Run: sudo make runtime_tests inside your build directory or sudo <builddir>/tests/runtime-tests.sh
  • Use the TEST_FILTER environment variable (or the --filter arg when running runtime-tests.sh) to only run a subset of the tests e.g. TEST_FILTER="uprobe.*" sudo make runtime-tests
  • There are environment variables to override paths for the bpftrace executables, if necessary. See runtime-tests.sh for details.

Runtime tests are grouped into "suites". A suite is usually a single file. The name of the file is the name of the suite.

Runtime test directives

Each runtime testcase consists of multiple directives. In no particular order:

  • NAME: Name of the test case. This field is required.
  • RUN: Run the command in a shell. See "Runtime variables" below for available placeholders. This XOR the PROG field is required
  • PROG: Run the provided bpftrace program. This directive is preferred over RUN unless you must pass flags or create a shell pipeline. This XOR the RUN field is required
  • EXPECT: The expected output. Python regular expressions are supported. One or more EXPECT and EXPECT_NONE or a single EXPECT_FILE or EXPECT_JSON is required.
  • EXPECT_NONE: The negation of EXPECT, which also supports Python regular expressions.
  • EXPECT_FILE: A file containing the expected output, matched as plain text after stripping initial and final empty lines
  • EXPECT_JSON: A json file containing the expected output, matched after converting the output and the file to a dict (thus ignoring field order).
  • TIMEOUT: The timeout for the testcase (in seconds). This field is required.
  • BEFORE: Run the command in a shell before running bpftrace. The command will be terminated after testcase is over. Can be used multiple times, commands will run in parallel.
  • AFTER: Run the command in a shell after running bpftrace (after the probes are attached). The command will be terminated after the testcase is over.
  • CLEANUP: Run the command in a shell after test is over. This holds any cleanup command to free resources after test completes.
  • MIN_KERNEL: Skip the test unless the host's kernel version is >= the provided kernel version. Try not to use this directive as kernel versions may be misleading (backported kernel features, for example)
  • REQUIRES: Run a command in a shell. If it succeeds, run the testcase. Else, skip the testcase.
  • ENV: Run bpftrace invocation with additional environment variables. Must be in format NAME=VALUE. Supports multiple values separated by spaces.
  • ARCH: Only run testcase on provided architectures. Supports | to logical OR multiple arches.
  • REQUIRES_FEATURE: Only run testcase if the following bpftrace feature is built in. See bpftrace --info and runtime/engine/runner.py for more details. Also supports negative features (by prefixing ! before feature).
  • WILL_FAIL: Mark that this test case will exit uncleanly (ie exit code != 0)
  • NEW_PIDNS: This will execute the BEFORE, the bpftrace (RUN or PROG), and the AFTER commands in a new pid namespace that mounts proc. At least one BEFORE is required.
  • SKIP_IF_ENV_HAS: Skip test case if specified environment variable is found and matches value provided. Accepted format is KEY=VALUE. Only a single key/value pair per test is accepted.

If you need to run a test program to probe (eg, uprobe/USDT), you can use the BEFORE clause. The test scripts will wait for the test program to have a pid.

The BEFORE clause will block up to the TIMEOUT waiting for a PID matching the basename of the last space-separated token. For instance, if the BEFORE clause is ./testprogs/usdt_test, it will wait for a processed called usdt_test. If it is ./testprogs/mountns_wrapper usdt_test it will also wait for a process called usdt_test. This approach is invalidated if a test program requires arguments in the future, but so far test programs are simple and separate minimal programs to test tracing functionality, and argument passing hasn't been required. If test programs need arguments, a more sophisticated approach will be necessary.

Runtime variables

Runtime variables are placeholders that the runtime test engine will fill out before running the test. These exist b/c the values of the variables are generally not known until test time. The following runtime variables are available for the RUN directive:

  • {{BPFTRACE}}: Path to bpftrace executable
  • {{BEFORE_PID}}: Process ID of the process in BEFORE directive

Test programs

You can add test programs for your runtime tests by placing a .c file corresponding to your test program in tests/testprogs.

You can add test libraries for your runtime tests by placing a .c file corresponding to your test library in tests/testlibs.

The test file tests/testprogs/my_test.c will result in an executable that you can call and probe in your runtime test at ./testprogs/my_test

This is intended to be useful for testing uprobes and USDT probes, or using uprobes to verify some other behavior in bpftrace. It can also be used to tightly control what code paths are triggered in the system.

Tool parsing tests

Tools parsing tests run every tool in the tools/ directory and ensure that the tools shipped with bpftrace are valid and can run. The validity of tools outputs is not checked at the moment.

Tests can be executed by: sudo <builddir>/tests/tools-parsing-test.sh

Flags and variables

The following environment variables can be set to modify the behaviour of the test suite

  • BPFTRACE_EXECUTABLE: location of the bpftrace executable, if left unset the script attempts to autodetect it.
  • TOOLS_TEST_DISABLE: comma separated list of tools to skip, e.g. vfscount.bt,swapin.bt
  • TOOLS_TEST_OLDVERSION: tests the tools/old version of these tools instead.