Hi,

strscpy_pad() patch set now with added test shenanigans.

This version adds 5 initial patches to the set and splits the single patch from v2 into two separate patches (6 and 7).

While doing the testing for strscpy_pad() it was noticed that there is duplication in how test modules are being fed to kselftest and also in the test modules themselves.

This set makes an attempt at adding a framework to kselftest for writing kernel test modules. It also adds a script for use in creating script test runners for kselftest. My macro-foo is not great, all criticism and suggestions very much appreciated. The design is based on test modules lib/test_printf.c, lib/test_bitmap.c, lib/test_xarray.c.

Shua, I'm by no means a kselftest expert, if this approach does not fit in with your general direction please say so.

Kees, I put the strscpy_pad() addition patch separate so if this goes in through Shua's tree (and if it goes in at all) its a single patch to grab if we want to start playing around with strscpy_pad().

Patch 1 fixes module unload for lib/test_printf in preparation for the rest of the series.

Patch 2 Adds a shell script that can be used to create shell script test runners.

Patch 3 Converts current shell script runners in tools/testing/selftests/lib/ to use the script introduced in patch 2.

Patch 4 Adds the test framework by way of a header file (inc. documentation)

Patch 5 Converts a couple of current test modules to make some use of the newly added test framework.

Patch 6 Adds strscpy_pad()

Patch 7 Adds test module for strscpy_pad() using the new framework and script.

If you are a testing geek and you would like to play with this; if you are already running a kernel built recently from your tree you may want to just apply the first 5 patches then you don't need to build/boot a new kernel, just config and build the lib/ test modules (test_printf etc.) and then:

sudo make TARGETS=lib kselftest

Late in the development of this I found that a bunch of boiler plate had to be added to the script to handle running tests with:

make O=/path/to/kout kselftest

The reason is that during the build we are in the output directory but the script is in the source directory. I get the feeling that a better understanding of how the kernel build process works would provide a better solution to this. The current solution is disappointing since removing duplication and boiler plate was the point of the whole exercise. I'd love a better way to solve this?

One final interesting note: there are 36 test modules in lib/ only 3 of them are run by kselftest from tools/testing/selfests/lib?

Thanks for looking at this, Tobin.

Tobin C. Harding (7): lib/test_printf: Add empty module_exit function kselftest: Add test runner creation script kselftest/lib: Use new shell runner to define tests kselftest: Add test module framework header lib: Use new kselftest header lib/string: Add strscpy_pad() function lib: Add test module for strscpy_pad

Documentation/dev-tools/kselftest.rst | 108 ++++++++++++- include/linux/string.h | 4 + lib/Kconfig.debug | 3 + lib/Makefile | 1 + lib/string.c | 47 +++++- lib/test_bitmap.c | 20 +-- lib/test_printf.c | 17 +-- lib/test_strscpy.c | 150 +++++++++++++++++++ tools/testing/selftests/kselftest_module.h | 48 ++++++ tools/testing/selftests/kselftest_module.sh | 75 ++++++++++ tools/testing/selftests/lib/Makefile | 2 +- tools/testing/selftests/lib/bitmap.sh | 25 ++-- tools/testing/selftests/lib/config | 1 + tools/testing/selftests/lib/prime_numbers.sh | 23 ++- tools/testing/selftests/lib/printf.sh | 25 ++-- tools/testing/selftests/lib/strscpy.sh | 17 +++ 16 files changed, 490 insertions(+), 76 deletions(-) create mode 100644 lib/test_strscpy.c create mode 100644 tools/testing/selftests/kselftest_module.h create mode 100755 tools/testing/selftests/kselftest_module.sh create mode 100755 tools/testing/selftests/lib/strscpy.sh