Re: [PATCH bpf-next 3/4] selftests/bpf: Fix possible ksyms test failure with LTO kernel

[Yonghong Song <yonghong.song@linux.dev>]

On 3/4/24 9:37 PM, Alexei Starovoitov wrote:
> On Sat, Mar 2, 2024 at 8:50 AM Yonghong Song <yonghong.song@linux.dev> wrote:
>> In my locally build clang LTO kernel (enabling CONFIG_LTO and
>> CONFIG_LTO_CLANG_THIN), ksyms test failed like:
>>    test_ksyms:PASS:kallsyms_fopen 0 nsec
>>    test_ksyms:FAIL:ksym_find symbol 'bpf_link_fops' not found
>>    #118     ksyms:FAIL
>>
>> The reason is that 'bpf_link_fops' is renamed to
>>    bpf_link_fops.llvm.8325593422554671469
>> Due to cross-file inlining, the static variable 'bpf_link_fops'
>> in syscall.c is used by a function in another file. To avoid
>> potential duplicated names, the llvm added suffix '.llvm.<hash>'.
>>
>> To fix the failure, we can skip this test with LTO kernel
>> if the symbol 'bpf_link_fops' is not found in kallsyms.
>>
>> After this patch, with the same LTO kernel:
>>    #118     ksyms:SKIP
>>
>> Signed-off-by: Yonghong Song <yonghong.song@linux.dev>
>> ---
>>   tools/testing/selftests/bpf/prog_tests/ksyms.c | 6 +++++-
>>   1 file changed, 5 insertions(+), 1 deletion(-)
>>
>> diff --git a/tools/testing/selftests/bpf/prog_tests/ksyms.c b/tools/testing/selftests/bpf/prog_tests/ksyms.c
>> index e081f8bf3f17..cd81f190c5d7 100644
>> --- a/tools/testing/selftests/bpf/prog_tests/ksyms.c
>> +++ b/tools/testing/selftests/bpf/prog_tests/ksyms.c
>> @@ -21,7 +21,11 @@ void test_ksyms(void)
>>                  return;
>>          }
>>          if (err == -ENOENT) {
>> -               ASSERT_TRUE(false, "ksym_find for bpf_link_fops");
>> +               /* bpf_link_fops might be renamed to bpf_link_fops.llvm.<hash> in LTO kernel. */
>> +               if (check_lto_kernel() == 1)
>> +                       test__skip();
>> +               else
>> +                       ASSERT_TRUE(false, "ksym_find for bpf_link_fops");
> I'm afraid LTO breakage is bigger than this.
> pid_iter program as part of bpftool is using bpf_link_fops too.
> I suspect that part of bpftool feature is broken on LTO kernel.
> We need to find a solution instead of 'skip' a selftest.

Thanks for pointing this out! Let me do some investigation on
how to resolve this.