]> git.ipfire.org Git - thirdparty/linux.git/commit
selftests/resctrl: Remove requirement on cache miss rate
authorReinette Chatre <reinette.chatre@intel.com>
Sat, 4 Apr 2026 01:56:07 +0000 (18:56 -0700)
committerShuah Khan <skhan@linuxfoundation.org>
Tue, 5 May 2026 00:40:02 +0000 (18:40 -0600)
commitca2e4f4ba15a00218bff3b89ff57b20d0164de0a
treec738a8e572eee29c1a980d52d4327135ec2b307a
parentd2ec0e8e2d024b67763670421678f723c817a901
selftests/resctrl: Remove requirement on cache miss rate

As the CAT test reads the same buffer into different sized cache portions
it compares the number of cache misses against an expected percentage
based on the size of the cache portion.

Systems and test conditions vary. The CAT test is a test of resctrl
subsystem health and not a test of the hardware architecture so it is not
required to place requirements on the size of the difference in cache
misses, just that the number of cache misses when reading a buffer
increase as the cache portion used for the buffer decreases.

Remove additional constraint on how big the difference between cache
misses should be as the cache portion size changes. Only test that the
cache misses increase as the cache portion size decreases. This remains
a good sanity check of resctrl subsystem health while reducing impact
of hardware architectural differences and the various conditions under
which the test may run.

Increase the size difference between cache portions to additionally avoid
any consequences resulting from smaller increments.

Link: https://lore.kernel.org/r/6de4da5486354c0f25fef0d194956470cb744041.1775266384.git.reinette.chatre@intel.com
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Tested-by: Chen Yu <yu.c.chen@intel.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
tools/testing/selftests/resctrl/cat_test.c