Subject: Re: [PATCH bpf-next 3/3] selftests/bpf: Check
 __ARCH_WANT_SET_GET_RLIMIT before syscall(__NR_getrlimit)
To:     Mykola Lysenko <mykolal@meta.com>
References: <1677066908-15224-1-git-send-email-yangtiezhu@loongson.cn>
 <1677066908-15224-4-git-send-email-yangtiezhu@loongson.cn>
 <CAADnVQLLborN3ABxRPUhSL5jQ1XcWNM9DBfjaEbvnF9qdE_CJA@mail.gmail.com>
 <D90B1D7B-8B75-4790-8D09-1106240B2377@fb.com>
Cc:     Alexei Starovoitov <ast@kernel.org>,
        Daniel Borkmann <daniel@iogearbox.net>,
        Andrii Nakryiko <andrii@kernel.org>, bpf <bpf@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>,
        Alexei Starovoitov <alexei.starovoitov@gmail.com>
From:   Tiezhu Yang <yangtiezhu@loongson.cn>
Message-ID: <f8d8f114-2fa7-f6da-7c99-b72fcf8b106f@loongson.cn>
Date:   Thu, 23 Feb 2023 10:48:48 +0800
User-Agent: Mozilla/5.0 (X11; Linux mips64; rv:45.0) Gecko/20100101
 Thunderbird/45.4.0
MIME-Version: 1.0
In-Reply-To: <D90B1D7B-8B75-4790-8D09-1106240B2377@fb.com>
Content-Type: text/plain; charset=windows-1252; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk


On 02/23/2023 04:03 AM, Mykola Lysenko wrote:
> Hi Tiezhu,
>
> You can run BPF CI tests on your patch before sending it out by following these instructions:
> https://docs.kernel.org/bpf/bpf_devel_QA.html#q-how-do-i-run-bpf-ci-on-my-changes-before-sending-them-out-for-review

OK, thank you.

After commit 80d7da1cac62 ("asm-generic: Drop getrlimit and setrlimit
syscalls from default list"), new architectures won't need to include
getrlimit and setrlimit, they are superseded with prlimit64.

In order to maintain compatibility for the new architectures, such as
LoongArch which does not define __NR_getrlimit, it is better to use
__NR_prlimit64 instead of __NR_getrlimit in user_ringbuf.c to fix the
build error.

diff --git a/tools/testing/selftests/bpf/prog_tests/user_ringbuf.c 
b/tools/testing/selftests/bpf/prog_tests/user_ringbuf.c
index 3a13e10..e51721d 100644
--- a/tools/testing/selftests/bpf/prog_tests/user_ringbuf.c
+++ b/tools/testing/selftests/bpf/prog_tests/user_ringbuf.c
@@ -590,7 +590,7 @@ static void *kick_kernel_cb(void *arg)
         /* Kick the kernel, causing it to drain the ring buffer and 
then wake
          * up the test thread waiting on epoll.
          */
-       syscall(__NR_getrlimit);
+       syscall(__NR_prlimit64);

         return NULL;
  }

I will test it and then send v2. If you have more suggestions,
please let me know.

Thanks,
Tiezhu

>
> Thanks,
> Mykola
>
>> On Feb 22, 2023, at 10:06 AM, Alexei Starovoitov <alexei.starovoitov@gmail.com> wrote:
>>
>> !-------------------------------------------------------------------|
>>  This Message Is From an External Sender
>>
>> |-------------------------------------------------------------------!
>>
>> On Wed, Feb 22, 2023 at 3:55 AM Tiezhu Yang <yangtiezhu@loongson.cn> wrote:
>>>
>>> __NR_getrlimit is defined only if __ARCH_WANT_SET_GET_RLIMIT is defined:
>>>
>>>  #ifdef __ARCH_WANT_SET_GET_RLIMIT
>>>  /* getrlimit and setrlimit are superseded with prlimit64 */
>>>  #define __NR_getrlimit 163
>>>  ...
>>>  #endif
>>>
>>> Some archs do not define __ARCH_WANT_SET_GET_RLIMIT, it should check
>>> __ARCH_WANT_SET_GET_RLIMIT before syscall(__NR_getrlimit) to fix the
>>> following build error:
>>>
>>>    TEST-OBJ [test_progs] user_ringbuf.test.o
>>>  tools/testing/selftests/bpf/prog_tests/user_ringbuf.c: In function 'kick_kernel_cb':
>>>  tools/testing/selftests/bpf/prog_tests/user_ringbuf.c:593:17: error: '__NR_getrlimit' undeclared (first use in this function)
>>>    593 |         syscall(__NR_getrlimit);
>>>        |                 ^~~~~~~~~~~~~~
>>>  tools/testing/selftests/bpf/prog_tests/user_ringbuf.c:593:17: note: each undeclared identifier is reported only once for each function it appears in
>>>  make: *** [Makefile:573: tools/testing/selftests/bpf/user_ringbuf.test.o] Error 1
>>>  make: Leaving directory 'tools/testing/selftests/bpf'
>>>
>>> Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
>>> ---
>>> tools/testing/selftests/bpf/prog_tests/user_ringbuf.c | 2 ++
>>> 1 file changed, 2 insertions(+)
>>>
>>> diff --git a/tools/testing/selftests/bpf/prog_tests/user_ringbuf.c b/tools/testing/selftests/bpf/prog_tests/user_ringbuf.c
>>> index 3a13e10..0550307 100644
>>> --- a/tools/testing/selftests/bpf/prog_tests/user_ringbuf.c
>>> +++ b/tools/testing/selftests/bpf/prog_tests/user_ringbuf.c
>>> @@ -590,7 +590,9 @@ static void *kick_kernel_cb(void *arg)
>>>        /* Kick the kernel, causing it to drain the ring buffer and then wake
>>>         * up the test thread waiting on epoll.
>>>         */
>>> +#ifdef __ARCH_WANT_SET_GET_RLIMIT
>>>        syscall(__NR_getrlimit);
>>> +#endif
>>
>> This is clearly breaks user_ringbuf test on x86:
>> https://github.com/kernel-patches/bpf/actions/runs/4242660318/jobs/7374845859
>>
>> Please do not send patches that make selftest compile on your favorite arch.
>> Make sure the patches work correctly on other archs too.