Re: [PATCH bpf-next v2 2/2] selftests/bpf: Fix xdp_pull_data failure with 64K page

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

On 1/22/26 1:18 PM, Amery Hung wrote:
> On Thu, Jan 22, 2026 at 8:56 AM Yonghong Song <yonghong.song@linux.dev> wrote:
>> If the argument 'pull_len' of run_test() is 'PULL_MAX' or
>> 'PULL_MAX | PULL_PLUS_ONE', the eventual pull_len size
>> will close to the page size. On arm64 systems with 64K pages,
>> the pull_len size will be close to 64K. But the existing buffer
>> will be close to 9000 which is not enough to pull.
>>
>> For those failed run_tests(), make buff size to
>>    2 * pg_sz + (pg_sz / 2)
>> This way, there will be enough buffer space to pull
>> regardless of page size.
>>
>> Tested-by: Alan Maguire <alan.maguire@oracle.com>
>> Cc: Amery Hung <ameryhung@gmail.com>
>> Signed-off-by: Yonghong Song <yonghong.song@linux.dev>
>> ---
>>   .../testing/selftests/bpf/prog_tests/xdp_pull_data.c | 12 +++++++-----
>>   1 file changed, 7 insertions(+), 5 deletions(-)
>>
>> Changelog:
>>   v1 -> v2:
>>     - v1: https://lore.kernel.org/bpf/20260120210930.2544950-1-yonghong.song@linux.dev/
>>     - Use pg_sz to simplify buff_len calculation.
>>
>> diff --git a/tools/testing/selftests/bpf/prog_tests/xdp_pull_data.c b/tools/testing/selftests/bpf/prog_tests/xdp_pull_data.c
>> index efa350d04ec5..35f0eb5b43f3 100644
>> --- a/tools/testing/selftests/bpf/prog_tests/xdp_pull_data.c
>> +++ b/tools/testing/selftests/bpf/prog_tests/xdp_pull_data.c
>> @@ -114,12 +114,14 @@ static void test_xdp_pull_data_basic(void)
>>   {
>>          u32 pg_sz, max_meta_len, max_data_len;
>>          struct test_xdp_pull_data *skel;
>> +       int buff_len;
>>
>>          skel = test_xdp_pull_data__open_and_load();
>>          if (!ASSERT_OK_PTR(skel, "test_xdp_pull_data__open_and_load"))
>>                  return;
>>
>>          pg_sz = sysconf(_SC_PAGE_SIZE);
>> +       buff_len = 2 * pg_sz + pg_sz / 2;
> Thanks for fixing the test. Just curious if there is an issue with 1.5
> pg_sz as suggested by Alan?

Although existing 'buff_len = 2 * pg_sz + pg_sz / 2' works fine for the test,
but let us use the Alan's suggestion (buff_len = pg_sz + pg_sz / 2)
since the buffer size is tighter and potentially better for test.

>
>>          if (find_xdp_sizes(skel, pg_sz))
>>                  goto out;

[...]