It's possible that memory allocation for 'filtered' will fail, but for the copy of the suite to succeed. In this case, the copy could be leaked.
Properly free 'copy' in the error case for the allocation of 'filtered' failing.
Note that there may also have been a similar issue in kunit_filter_subsuites, before it was removed in "kunit: flatten kunit_suite*** to kunit_suite** in .kunit_test_suites".
This was reported by clang-analyzer via the kernel test robot, here: https://lore.kernel.org/all/c8073b8e-7b9e-0830-4177-87c12f16349c@intel.com/
And by smatch via Dan Carpenter and the kernel test robot: https://lore.kernel.org/all/202207101328.ASjx88yj-lkp@intel.com/
Fixes: a02353f49162 ("kunit: bail out of test filtering logic quicker if OOM") Reported-by: kernel test robot yujie.liu@intel.com Reported-by: kernel test robot lkp@intel.com Reported-by: Dan Carpenter dan.carpenter@oracle.com Reviewed-by: Daniel Latypov dlatypov@google.com Reviewed-by: Brendan Higgins brendanhiggins@google.com Signed-off-by: David Gow davidgow@google.com ---
Thanks everyone! No actual code changes in v2, just fixes to the description.
Changes since v1: https://lore.kernel.org/all/20220712095627.1770312-1-davidgow@google.com/ - Fix a mistake in the commit description where we noted the allocation for 'copy' could fail, instead of 'filtered'. (Thanks Yujie!) - Noted in the description that smatch also found this (Thanks Dan!) - Added the extra Reported-by and Reviewed-by tags.
--- lib/kunit/executor.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/lib/kunit/executor.c b/lib/kunit/executor.c index 6c489d6c5e5d..5e223327196a 100644 --- a/lib/kunit/executor.c +++ b/lib/kunit/executor.c @@ -74,8 +74,10 @@ kunit_filter_tests(const struct kunit_suite *const suite, const char *test_glob) return ERR_PTR(-ENOMEM);
filtered = kcalloc(n + 1, sizeof(*filtered), GFP_KERNEL); - if (!filtered) + if (!filtered) { + kfree(copy); return ERR_PTR(-ENOMEM); + }
n = 0; kunit_suite_for_each_test_case(suite, test_case) {