testing-overview.rst (8513B)
1.. SPDX-License-Identifier: GPL-2.0 2 3==================== 4Kernel Testing Guide 5==================== 6 7 8There are a number of different tools for testing the Linux kernel, so knowing 9when to use each of them can be a challenge. This document provides a rough 10overview of their differences, and how they fit together. 11 12 13Writing and Running Tests 14========================= 15 16The bulk of kernel tests are written using either the kselftest or KUnit 17frameworks. These both provide infrastructure to help make running tests and 18groups of tests easier, as well as providing helpers to aid in writing new 19tests. 20 21If you're looking to verify the behaviour of the Kernel — particularly specific 22parts of the kernel — then you'll want to use KUnit or kselftest. 23 24 25The Difference Between KUnit and kselftest 26------------------------------------------ 27 28KUnit (Documentation/dev-tools/kunit/index.rst) is an entirely in-kernel system 29for "white box" testing: because test code is part of the kernel, it can access 30internal structures and functions which aren't exposed to userspace. 31 32KUnit tests therefore are best written against small, self-contained parts 33of the kernel, which can be tested in isolation. This aligns well with the 34concept of 'unit' testing. 35 36For example, a KUnit test might test an individual kernel function (or even a 37single codepath through a function, such as an error handling case), rather 38than a feature as a whole. 39 40This also makes KUnit tests very fast to build and run, allowing them to be 41run frequently as part of the development process. 42 43There is a KUnit test style guide which may give further pointers in 44Documentation/dev-tools/kunit/style.rst 45 46 47kselftest (Documentation/dev-tools/kselftest.rst), on the other hand, is 48largely implemented in userspace, and tests are normal userspace scripts or 49programs. 50 51This makes it easier to write more complicated tests, or tests which need to 52manipulate the overall system state more (e.g., spawning processes, etc.). 53However, it's not possible to call kernel functions directly from kselftest. 54This means that only kernel functionality which is exposed to userspace somehow 55(e.g. by a syscall, device, filesystem, etc.) can be tested with kselftest. To 56work around this, some tests include a companion kernel module which exposes 57more information or functionality. If a test runs mostly or entirely within the 58kernel, however, KUnit may be the more appropriate tool. 59 60kselftest is therefore suited well to tests of whole features, as these will 61expose an interface to userspace, which can be tested, but not implementation 62details. This aligns well with 'system' or 'end-to-end' testing. 63 64For example, all new system calls should be accompanied by kselftest tests. 65 66Code Coverage Tools 67=================== 68 69The Linux Kernel supports two different code coverage measurement tools. These 70can be used to verify that a test is executing particular functions or lines 71of code. This is useful for determining how much of the kernel is being tested, 72and for finding corner-cases which are not covered by the appropriate test. 73 74Documentation/dev-tools/gcov.rst is GCC's coverage testing tool, which can be 75used with the kernel to get global or per-module coverage. Unlike KCOV, it 76does not record per-task coverage. Coverage data can be read from debugfs, 77and interpreted using the usual gcov tooling. 78 79Documentation/dev-tools/kcov.rst is a feature which can be built in to the 80kernel to allow capturing coverage on a per-task level. It's therefore useful 81for fuzzing and other situations where information about code executed during, 82for example, a single syscall is useful. 83 84 85Dynamic Analysis Tools 86====================== 87 88The kernel also supports a number of dynamic analysis tools, which attempt to 89detect classes of issues when they occur in a running kernel. These typically 90each look for a different class of bugs, such as invalid memory accesses, 91concurrency issues such as data races, or other undefined behaviour like 92integer overflows. 93 94Some of these tools are listed below: 95 96* kmemleak detects possible memory leaks. See 97 Documentation/dev-tools/kmemleak.rst 98* KASAN detects invalid memory accesses such as out-of-bounds and 99 use-after-free errors. See Documentation/dev-tools/kasan.rst 100* UBSAN detects behaviour that is undefined by the C standard, like integer 101 overflows. See Documentation/dev-tools/ubsan.rst 102* KCSAN detects data races. See Documentation/dev-tools/kcsan.rst 103* KFENCE is a low-overhead detector of memory issues, which is much faster than 104 KASAN and can be used in production. See Documentation/dev-tools/kfence.rst 105* lockdep is a locking correctness validator. See 106 Documentation/locking/lockdep-design.rst 107* There are several other pieces of debug instrumentation in the kernel, many 108 of which can be found in lib/Kconfig.debug 109 110These tools tend to test the kernel as a whole, and do not "pass" like 111kselftest or KUnit tests. They can be combined with KUnit or kselftest by 112running tests on a kernel with these tools enabled: you can then be sure 113that none of these errors are occurring during the test. 114 115Some of these tools integrate with KUnit or kselftest and will 116automatically fail tests if an issue is detected. 117 118Static Analysis Tools 119===================== 120 121In addition to testing a running kernel, one can also analyze kernel source code 122directly (**at compile time**) using **static analysis** tools. The tools 123commonly used in the kernel allow one to inspect the whole source tree or just 124specific files within it. They make it easier to detect and fix problems during 125the development process. 126 127Sparse can help test the kernel by performing type-checking, lock checking, 128value range checking, in addition to reporting various errors and warnings while 129examining the code. See the Documentation/dev-tools/sparse.rst documentation 130page for details on how to use it. 131 132Smatch extends Sparse and provides additional checks for programming logic 133mistakes such as missing breaks in switch statements, unused return values on 134error checking, forgetting to set an error code in the return of an error path, 135etc. Smatch also has tests against more serious issues such as integer 136overflows, null pointer dereferences, and memory leaks. See the project page at 137http://smatch.sourceforge.net/. 138 139Coccinelle is another static analyzer at our disposal. Coccinelle is often used 140to aid refactoring and collateral evolution of source code, but it can also help 141to avoid certain bugs that occur in common code patterns. The types of tests 142available include API tests, tests for correct usage of kernel iterators, checks 143for the soundness of free operations, analysis of locking behavior, and further 144tests known to help keep consistent kernel usage. See the 145Documentation/dev-tools/coccinelle.rst documentation page for details. 146 147Beware, though, that static analysis tools suffer from **false positives**. 148Errors and warns need to be evaluated carefully before attempting to fix them. 149 150When to use Sparse and Smatch 151----------------------------- 152 153Sparse does type checking, such as verifying that annotated variables do not 154cause endianness bugs, detecting places that use ``__user`` pointers improperly, 155and analyzing the compatibility of symbol initializers. 156 157Smatch does flow analysis and, if allowed to build the function database, it 158also does cross function analysis. Smatch tries to answer questions like where 159is this buffer allocated? How big is it? Can this index be controlled by the 160user? Is this variable larger than that variable? 161 162It's generally easier to write checks in Smatch than it is to write checks in 163Sparse. Nevertheless, there are some overlaps between Sparse and Smatch checks. 164 165Strong points of Smatch and Coccinelle 166-------------------------------------- 167 168Coccinelle is probably the easiest for writing checks. It works before the 169pre-processor so it's easier to check for bugs in macros using Coccinelle. 170Coccinelle also creates patches for you, which no other tool does. 171 172For example, with Coccinelle you can do a mass conversion from 173``kmalloc(x * size, GFP_KERNEL)`` to ``kmalloc_array(x, size, GFP_KERNEL)``, and 174that's really useful. If you just created a Smatch warning and try to push the 175work of converting on to the maintainers they would be annoyed. You'd have to 176argue about each warning if can really overflow or not. 177 178Coccinelle does no analysis of variable values, which is the strong point of 179Smatch. On the other hand, Coccinelle allows you to do simple things in a simple 180way.