Re: [PATCH bpf-next v2] selftests/bpf: Use snake_case for subtest descriptions in jit_insns_arena.c

[Tiezhu Yang <yangtiezhu@loongson.cn>]

On 2026/10/19 下午3:18, Tiezhu Yang wrote:
> On 2026/10/19 下午2:47, Daniel Borkmann wrote:
>> On 10/19/26 6:01 AM, Tiezhu Yang wrote:
>>> Normalize two inline assembly test descriptions in jit_insns_arena.c
>>> from space-separated strings to snake_case.
>>>
>>> It is done for consistency with other neighboring subtests registered
>>> in test_jit_insns() (such as alu_imm, alu_reg, etc.). This also makes
>>> it convenient to execute these subtests directly from the command line
>>> without being forced to use shell quoting.
>>>
>>> Otherwise, unquoted shell word-splitting causes test_progs to reject
>>> the extra arguments before the filter code runs:
>>>
>>> $ sudo ./test_progs -t jit_insns/arena access with zero offset
>>> Usage: test_progs [OPTION...]
>>> Try `test_progs --help' or `test_progs --usage' for more information.
>>>
>>> $ sudo ./test_progs -t jit_insns/cast to a user address
>>> Usage: test_progs [OPTION...]
>>> Try `test_progs --help' or `test_progs --usage' for more information.
>>>
>>> Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
>> Do these not work already?
>>
>>    ./test_progs -t jit_insns/arena_zero_off
>>    ./test_progs -t jit_insns/cast_to_user
> 
> Yes, they work.
> 
> I misunderstood that the name is "cast to a user address" in the output
> rather than "cast_to_user" before, there is difference between the
> description and the name.
> 
> Something like this:
> 
> $ sudo ./test_progs -t jit_insns/cast_to_user
> #191/2   jit_insns/cast to a user address:OK
> #191     jit_insns:OK
> #192     jit_insns_blinded:OK
> Summary: 2/1 PASSED, 0 SKIPPED, 0/0 FAILED
> 
> Anyway, if this patch is not necessary, please ignore it.

By the way, IMO, it is better to remove the __description strings,
so that the test output style matches the actual program names like
other neighboring subtests. This prevents developers from being
misled into copying space-separated strings from the standard output.

What do you think about something like this?

diff --git a/tools/testing/selftests/bpf/progs/jit_insns_arena.c 
b/tools/testing/selftests/bpf/progs/jit_insns_arena.c
index b396d5aefb5a..95736eae2a58 100644
--- a/tools/testing/selftests/bpf/progs/jit_insns_arena.c
+++ b/tools/testing/selftests/bpf/progs/jit_insns_arena.c
@@ -34,7 +34,6 @@ struct {

  #if defined(__BPF_FEATURE_ADDR_SPACE_CAST)
  SEC("syscall")
-__description("arena access with zero offset")
  __success __retval(0)
  int arena_zero_off(void *ctx)
  {
@@ -82,7 +81,6 @@ int arena_zero_off(void *ctx)
  }

  SEC("syscall")
-__description("cast to a user address")
  __success __retval(0)
  int cast_to_user(void *ctx)
  {

The output is something like this:

$ sudo ./test_progs -t jit_insns/arena_zero_off | tail -3
#191/1   jit_insns/arena_zero_off:FAIL
#191     jit_insns:FAIL
Summary: 1/0 PASSED, 0 SKIPPED, 1/1 FAILED

$ sudo ./test_progs -t jit_insns/cast_to_user
#191/2   jit_insns/cast_to_user:OK
#191     jit_insns:OK
#192     jit_insns_blinded:OK
Summary: 2/1 PASSED, 0 SKIPPED, 0/0 FAILED

The main goal here is that if a test fails, developers can immediately
recognize the exact, actionable test name from the standard output and
rerun it without any confusion.

Thanks,
Tiezhu