selftests/bpf: make arg_parsing.c more robust to crashes
Subsystems (3)
Andrii Nakryiko
maintainerPATCH 1/1 selftests/bpf: make arg_parsing.c more robust to crashes
We started getting a crash in BPF CI, which seems to originate from test_parse_test_list_file() test and is happening at this line:
ASSERT_OK(strcmp("test_with_spaces", set.tests[0].name), "test 0 name");One way we can crash there is if set.cnt zero, which is checked for with ASSERT_EQ() above, but we proceed after this regardless of the outcome. Instead of crashing, we should bail out with test failure early.
Similarly, if parse_test_list_file() fails, we shouldn't be even looking at set, so bail even earlier if ASSERT_OK() fails.
--- tools/testing/selftests/bpf/prog_tests/arg_parsing.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)
tools/testing/selftests/bpf/prog_tests/arg_parsing.c +6 -3
@@ -146,9 +146,12 @@ static void test_parse_test_list_file(void)init_test_filter_set (&set); -ASSERT_OK (parse_test_list_file (tmpfile, &set, true),"parse file" ); +if (!ASSERT_OK (parse_test_list_file (tmpfile, &set, true),"parse file" )) + goto out_fclose; + +if (!ASSERT_EQ (set.cnt,4 ,"test count" )) + goto out_free_set; -ASSERT_EQ (set.cnt,4 ,"test count" );ASSERT_OK (strcmp ("test_with_spaces" , set.tests[0 ].name),"test 0 name" );ASSERT_EQ (set.tests[0 ].subtest_cnt,0 ,"test 0 subtest count" );ASSERT_OK (strcmp ("testA" , set.tests[1 ].name),"test 1 name" );
@@ -158,8 +161,8 @@ static void test_parse_test_list_file(void)ASSERT_OK (strcmp ("testB" , set.tests[2 ].name),"test 2 name" );ASSERT_OK (strcmp ("testC_no_eof_newline" , set.tests[3 ].name),"test 3 name" ); +out_free_set:free_test_filter_set (&set); - out_fclose:fclose (fp); out_remove:
Signature · 1 line
2.47.3
Quoted 9 lines · Andrii Nakryiko
We started getting a crash in BPF CI, which seems to originate from
test_parse_test_list_file()test and is happening at this line:ASSERT_OK(strcmp("test_with_spaces", set.tests[0].name), "test 0 name");One way we can crash there is if set.cnt zero, which is checked for with
ASSERT_EQ()above, but we proceed after this regardless of the outcome. Instead of crashing, we should bail out with test failure early.
Quoted 5 lines
Similarly, if
parse_test_list_file()fails, we shouldn't be even looking at set, so bail even earlier ifASSERT_OK()fails.Fixes 64276f01dce8 ("selftests/bpf: Test_progs can read test lists from file")Signed-off-by Andrii Nakryiko <[email protected]>
This patch in combination with another arg_parsing.c fix [1] mitigates recent BPF CI failures:
https://github.com/kernel-patches/vmtest/actions/runs/18510075579?pr=404
[1] https://lore.kernel.org/bpf/[email protected]/
Quoted 33 lines
--- tools/testing/selftests/bpf/prog_tests/arg_parsing.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)tools/testing/selftests/bpf/prog_tests/arg_parsing.c +6
-3@@ -146,9 +146,12 @@ static void test_parse_test_list_file(void)init_test_filter_set (&set); -ASSERT_OK (parse_test_list_file (tmpfile, &set, true),"parse file" ); +if (!ASSERT_OK (parse_test_list_file (tmpfile, &set, true),"parse file" )) + goto out_fclose; + +if (!ASSERT_EQ (set.cnt,4 ,"test count" )) + goto out_free_set; -ASSERT_EQ (set.cnt,4 ,"test count" );ASSERT_OK (strcmp ("test_with_spaces" , set.tests[0 ].name),"test 0 name" );ASSERT_EQ (set.tests[0 ].subtest_cnt,0 ,"test 0 subtest count" );ASSERT_OK (strcmp ("testA" , set.tests[1 ].name),"test 1 name" );@@ -158,8 +161,8 @@ static void test_parse_test_list_file(void)ASSERT_OK (strcmp ("testB" , set.tests[2 ].name),"test 2 name" );ASSERT_OK (strcmp ("testC_no_eof_newline" , set.tests[3 ].name),"test 3 name" ); +out_free_set:free_test_filter_set (&set); - out_fclose:fclose (fp); out_remove:
This patch was applied to bpf/bpf.git (master) by Alexei Starovoitov <[email protected]>:
Hello:
This patch was applied to bpf/bpf.git (master)
by Alexei Starovoitov <[email protected]>:
Quoted 10 lines · On Tue, 14 Oct 2025 13:20:37 -0700 you wrote:
We started getting a crash in BPF CI, which seems to originate from
test_parse_test_list_file()test and is happening at this line:ASSERT_OK(strcmp("test_with_spaces", set.tests[0].name), "test 0 name");One way we can crash there is if set.cnt zero, which is checked for with
ASSERT_EQ()above, but we proceed after this regardless of the outcome. Instead of crashing, we should bail out with test failure early.[...]
Here is the summary with links:
- [bpf-next] selftests/bpf: make arg_parsing.c more robust to crashes
https://git.kernel.org/bpf/bpf/c/e603a342cf7e
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html