Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <378731e7-eec5-44ca-eded-3277792b061e@efficios.com>
Date:   Tue, 17 Jan 2023 15:14:52 -0500
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.6.0
Subject: Re: [tip:sched/core 7/28] rseq.c:139:37: error: 'AT_RSEQ_ALIGN'
 undeclared; did you mean 'R_SH_ALIGN'?
Content-Language: en-US
To:     Shuah Khan <skhan@linuxfoundation.org>,
        Ingo Molnar <mingo@kernel.org>
Cc:     kernel test robot <lkp@intel.com>, Shuah Khan <shuah@kernel.org>,
        oe-kbuild-all@lists.linux.dev, linux-kernel@vger.kernel.org,
        x86@kernel.org, Peter Zijlstra <peterz@infradead.org>,
        linux-kselftest@vger.kernel.org
References: <202301170348.7WLKH1pl-lkp@intel.com>
 <bfa719c3-bd1f-5fc4-40ab-6dc6822b7628@efficios.com>
 <4449d8b5-b7a5-0f09-5b42-7b70ba00f8f6@linuxfoundation.org>
 <Y8ZlKOsN1wGk9tTc@gmail.com>
 <37625f31-6ac2-1f90-d864-e4644820bba3@efficios.com>
 <1eed08d1-c100-6ca5-63f3-73487970b08e@linuxfoundation.org>
From:   Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
In-Reply-To: <1eed08d1-c100-6ca5-63f3-73487970b08e@linuxfoundation.org>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 2023-01-17 14:11, Shuah Khan wrote:
> On 1/17/23 10:44, Mathieu Desnoyers wrote:
>> On 2023-01-17 04:06, Ingo Molnar wrote:
>>>
>>> * Shuah Khan <skhan@linuxfoundation.org> wrote:
>>>
>>>> On 1/16/23 13:18, Mathieu Desnoyers wrote:
>>>>> On 2023-01-16 14:40, kernel test robot wrote:
>>>>>> tree:   
>>>>>> https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git 
>>>>>> sched/core
>>>>>> head:   79ba1e607d68178db7d3fe4f6a4aa38f06805e7b
>>>>>> commit: 03f5c0272d1b59343144e199becc911dae52c37e [7/28] 
>>>>>> selftests/rseq: Use ELF auxiliary vector for extensible rseq
>>>>>> compiler: gcc-11 (Debian 11.3.0-8) 11.3.0
>>>>>> reproduce:
>>>>>>           # 
>>>>>> https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/commit/?id=03f5c0272d1b59343144e199becc911dae52c37e
>>>>>>           git remote add tip 
>>>>>> https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git
>>>>>>           git fetch --no-tags tip sched/core
>>>>>>           git checkout 03f5c0272d1b59343144e199becc911dae52c37e
>>>>>>           make O=/tmp/kselftest headers
>>>>>>           make O=/tmp/kselftest -C tools/testing/selftests
>>>>>>
>>>>>> If you fix the issue, kindly add following tag where applicable
>>>>>> | Reported-by: kernel test robot <lkp@intel.com>
>>>>>
>>>>> In order to fix this, I need to change -I../../../../usr/include/ 
>>>>> for $(KHDR_INCLUDES) in tools/testing/selftests/rseq/Makefile
>>>>>
>>>>> I can find 25 odd uses of the same pattern in the kernel selftests. 
>>>>> Should I fix them all in one go ?
>>>>
>>>> kselftest build depends on headers installed in the root directory.
>>
>> By "root directory", do you mean kernel sources root directory or 
>> build output root directory ?
>>
>>>> The main makefile enforces this dependency.
>>
>> How ? I figure that tools/testing/selftests/lib.mk overrides 
>> KHDR_INCLUDES if it is not defined yet:
>>
>> ifeq ($(KHDR_INCLUDES),)
>> KHDR_INCLUDES := -isystem $(top_srcdir)/usr/include
>> endif
>>
>> and selftests makefiles include ../lib.mk.
>>
>> This KHDR_INCLUDES can be modified by O=... when built from the kernel 
>> top level, thus using tools/testing/selftests/Makefile:
>>
>> ifneq ($(KBUILD_OUTPUT),)
>> [...]
>>    KHDR_INCLUDES := -isystem ${abs_objtree}/usr/include
>> else
>> [...]
>>    KHDR_INCLUDES := -isystem ${abs_srctree}/usr/include
>> endif
>>
>> But it's up to the individual selftests to actually use 
>> $(KHDR_INCLUDES). In many cases, they hardcode 
>> -I../../../../usr/include/ which is bogus when the build root (O=...) 
>> differs from the source root.
>>
>>   If this test is being
>>>> built without installing headers by itself, I think the scripts that
>>>> build individual tests have to makes sure headers are installed first.
>>
>> The headers were previously built by "make O=/tmp/kselftest headers", 
>> as it should, it's just that the selftest makefile uses a hardcoded 
>> path that is relative to the source directory, and it appears that 
>> this pattern is repeated all across the selftests.
>>
> 
> selftests Makefile used to install headers and there has been a recent
> change to have mani Makefile (root) to install it. As a result individual
> test builds (running make in the test directory) requires header install
> now.
> 
> I think the hard-coded includes are a problem and we have to fix them for
> all cases i.e make O=, individual test builds.
> 
> If you are still up for it, please send patch.

Sure.

Now that I dig a bit more, we could try using "-nostdinc" to tell the 
compiler not to search the standard system directories, but I've noticed 
that a few selftests depend on userspace library header files found in 
the build environment.

I'm not sure how we should approach this.

Thoughts ?

Thanks,

Mathieu

> 
> thanks,
> -- Shuah
> 

-- 
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com