[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <dae27584-faf7-f132-1272-cb21248e5fa9@linuxfoundation.org>
Date: Fri, 10 Feb 2023 17:15:41 -0700
From: Shuah Khan <skhan@...uxfoundation.org>
To: Mathieu Desnoyers <mathieu.desnoyers@...icios.com>,
Shuah Khan <shuah@...nel.org>, linux-kselftest@...r.kernel.org,
Steven Rostedt <rostedt@...dmis.org>,
Christoph Hellwig <hch@....de>
Cc: linux-kernel@...r.kernel.org, bpf <bpf@...r.kernel.org>,
Networking <netdev@...r.kernel.org>,
Ingo Molnar <mingo@...hat.com>, linuxppc-dev@...ts.ozlabs.org
Subject: Re: [PATCH 00/34] selftests: Fix incorrect kernel headers search path
On 2/3/23 18:06, Shuah Khan wrote:
> On 2/1/23 19:07, Shuah Khan wrote:
>> Hi Mathieu,
>>
>> On 1/30/23 15:29, Shuah Khan wrote:
>>> On 1/27/23 06:57, Mathieu Desnoyers wrote:
>>>> Hi,
>>>>
>>>> This series fixes incorrect kernel header search path in kernel
>>>> selftests.
>>>>
>>>> Near the end of the series, a few changes are not tagged as "Fixes"
>>>> because the current behavior is to rely on the kernel sources uapi files
>>>> rather than on the installed kernel header files. Nevertheless, those
>>>> are updated for consistency.
>>>>
>>>> There are situations where "../../../../include/" was added to -I search
>>>> path, which is bogus for userspace tests and caused issues with types.h.
>>>> Those are removed.
>>>>
>>
>> Thanks again for taking care of this. I did out of tree build testing on
>> x86 on linux-kselftest next with these patches below. I haven't seen
>> any problems introduced by the patch set.
>>
>>>> selftests: dma: Fix incorrect kernel headers search path
>> This one needs a change and I will send a patch on top of yours.
>> Even with that this test depends on unexported header from the
>> repo and won't build out of tree. This is not related to your
>> change.
>>
>>>> selftests: mount_setattr: Fix incorrect kernel headers search path
>> This one fails to build with our without patch - an existing error.
>>
>> I have to do cross-build tests on arm64 and other arch patches still.
>> This will happen later this week.
>
> arm64, s390 patches look good.
>
I am seeing problem with selftests/dma and selfttests/user_events.
1. selftests: dma: Fix incorrect kernel headers search path
dma test no longer builds. This test depends on linux/map_benchmark.h
which is not included in uapi
The order of include directorries -isystem followed by installed kernel
headers, breaks the test build with the change to use KHDR_INCLUDES
I am going to revert this patch for now and figure a longer term fix.
The problem is the dependency on a non-uapi file: linux/map_benchmark.h
Fixes: 8ddde07a3d28 ("dma-mapping: benchmark: extract a common
header file for map_benchmark definition") change added this
dependency on including linux/map_benchmark.h
Christoph, Do you see this map_benchmark.h as part of uapi?
2. selftests: user_events: Fix incorrect kernel headers search path
This one depends on linux/user_events.h which has bee removed from
uapi in this commit:
commit 5cfff569cab8bf544bab62c911c5d6efd5af5e05
Author: Steven Rostedt (Google) <rostedt@...dmis.org>
Date: Fri Apr 1 14:39:03 2022 -0400
tracing: Move user_events.h temporarily out of include/uapi
This isn't a regression from 6.2 - this test stopped building once
user_events.h has been removed from uapi. I will add a note that
this test depends on a non-uapi header and can't be built at the
moment.
thanks,
-- Shuah
Powered by blists - more mailing lists