On 6/2/26 07:18, Alexei Starovoitov wrote:
> On Wed, Feb 4, 2026 at 7:54 PM Leon Hwang <[email protected]> wrote:
>>
>>
>>
>> On 5/2/26 04:14, Alexei Starovoitov wrote:
>>> On Mon, Feb 2, 2026 at 6:43 AM Leon Hwang <[email protected]> wrote:
>>>>
>>>> +
>>>> +#define BPF_LOG_FIXED 8
>>>> +
>>>> +static void test_map_create(enum bpf_map_type map_type, const char
>>>> *map_name,
>>>> + struct bpf_map_create_opts *opts, const char
>>>> *exp_msg)
>>>> +{
>>>> + const int key_size = 4, value_size = 4, max_entries = 1;
>>>> + char log_buf[128];
>>>> + int fd;
>>>> + LIBBPF_OPTS(bpf_log_opts, log_opts);
>>>> +
>>>> + log_buf[0] = '\0';
>>>> + log_opts.log_buf = log_buf;
>>>> + log_opts.log_size = sizeof(log_buf);
>>>> + log_opts.log_level = BPF_LOG_FIXED;
>>>
>>> Why? Which part of the test needs the log with this flag?
>>
>> BPF_LOG_FIXED looks odd here.
>>
>> This test sets 'log_level = BPF_LOG_FIXED' to match the behavior of
>> bpf_vlog_init() as initialized by bpf_log_attr_create_vlog() in
>> patch #7. BPF_LOG_FIXED is intended to be the default log_level
>> there.
>
> I don't think you answered my question.
> bpf_vlog_init() is using whatever log_level user space provided.
> Why do you pass BPF_LOG_FIXED ?
>
The intention behind passing BPF_LOG_FIXED was to ensure the log used
the buffer in a fixed mode, since the allocated buffer was large enough
to hold the full log message from the kernel. It was not intended to
test against log_level itself.
After reviewing commit 121664093803 ("bpf: Switch BPF verifier log to be
a rotating log by default"), I realized that BPF_LOG_FIXED was
introduced specifically to disable the rotating log behavior. In this
test case, that distinction is not relevant, so BPF_LOG_FIXED is indeed
unnecessary.
I tested with 'log_level is 0' and 'log_level is non-zero'. The tests
fail when log_level is 0, and pass when log_level is non-zero. So I will
switch to using 'log_level = 1' in the next revision.
Thanks,
Leon
