On 2/11/22 1:30 AM, Sherry Yang wrote:
> seccomp_bpf failed on tests 47 global.user_notification_filter_empty
> and 48 global.user_notification_filter_empty_threaded when it's
> tested on updated kernel but with old kernel headers. Because old
> kernel headers don't have definition of macro __NR_clone3 which is
> required for these two tests. Since under selftests/, we can install
> headers once for all tests (the default INSTALL_HDR_PATH is
> usr/include), fix it by adding usr/include to the list of directories
> to be searched. Use "-isystem" to indicate it's a system directory as
> the real kernel headers directories are.
>
> Signed-off-by: Sherry Yang <[email protected]>
> Tested-by: Sherry Yang <[email protected]>
> ---
> tools/testing/selftests/seccomp/Makefile | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/tools/testing/selftests/seccomp/Makefile b/tools/testing/selftests/seccomp/Makefile
> index 0ebfe8b0e147..585f7a0c10cb 100644
> --- a/tools/testing/selftests/seccomp/Makefile
> +++ b/tools/testing/selftests/seccomp/Makefile
> @@ -1,5 +1,5 @@
> # SPDX-License-Identifier: GPL-2.0
> -CFLAGS += -Wl,-no-as-needed -Wall
> +CFLAGS += -Wl,-no-as-needed -Wall -isystem ../../../../usr/include/
"../../../../usr/include/" directory doesn't have header files if
different output directory is used for kselftests build like "make -C
tools/tests/selftest O=build". Can you try adding recently added
variable, KHDR_INCLUDES here which makes this kind of headers inclusion
easy and correct for other build combinations as well?
On Fri, Feb 11, 2022 at 04:14:17AM +0500, Muhammad Usama Anjum wrote:
> On 2/11/22 1:30 AM, Sherry Yang wrote:
> > seccomp_bpf failed on tests 47 global.user_notification_filter_empty
> > and 48 global.user_notification_filter_empty_threaded when it's
> > tested on updated kernel but with old kernel headers. Because old
> > kernel headers don't have definition of macro __NR_clone3 which is
> > required for these two tests. Since under selftests/, we can install
> > headers once for all tests (the default INSTALL_HDR_PATH is
> > usr/include), fix it by adding usr/include to the list of directories
> > to be searched. Use "-isystem" to indicate it's a system directory as
> > the real kernel headers directories are.
> >
> > Signed-off-by: Sherry Yang <[email protected]>
> > Tested-by: Sherry Yang <[email protected]>
> > ---
> > tools/testing/selftests/seccomp/Makefile | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/tools/testing/selftests/seccomp/Makefile b/tools/testing/selftests/seccomp/Makefile
> > index 0ebfe8b0e147..585f7a0c10cb 100644
> > --- a/tools/testing/selftests/seccomp/Makefile
> > +++ b/tools/testing/selftests/seccomp/Makefile
> > @@ -1,5 +1,5 @@
> > # SPDX-License-Identifier: GPL-2.0
> > -CFLAGS += -Wl,-no-as-needed -Wall
> > +CFLAGS += -Wl,-no-as-needed -Wall -isystem ../../../../usr/include/
>
> "../../../../usr/include/" directory doesn't have header files if
> different output directory is used for kselftests build like "make -C
> tools/tests/selftest O=build". Can you try adding recently added
> variable, KHDR_INCLUDES here which makes this kind of headers inclusion
> easy and correct for other build combinations as well?
Ah, if that's true I think there are some other instances in the tree
that need fixing too.
--
Kees Cook
> On Feb 10, 2022, at 3:14 PM, Muhammad Usama Anjum <[email protected]> wrote:
>
> On 2/11/22 1:30 AM, Sherry Yang wrote:
>> seccomp_bpf failed on tests 47 global.user_notification_filter_empty
>> and 48 global.user_notification_filter_empty_threaded when it's
>> tested on updated kernel but with old kernel headers. Because old
>> kernel headers don't have definition of macro __NR_clone3 which is
>> required for these two tests. Since under selftests/, we can install
>> headers once for all tests (the default INSTALL_HDR_PATH is
>> usr/include), fix it by adding usr/include to the list of directories
>> to be searched. Use "-isystem" to indicate it's a system directory as
>> the real kernel headers directories are.
>>
>> Signed-off-by: Sherry Yang <[email protected]>
>> Tested-by: Sherry Yang <[email protected]>
>> ---
>> tools/testing/selftests/seccomp/Makefile | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/tools/testing/selftests/seccomp/Makefile b/tools/testing/selftests/seccomp/Makefile
>> index 0ebfe8b0e147..585f7a0c10cb 100644
>> --- a/tools/testing/selftests/seccomp/Makefile
>> +++ b/tools/testing/selftests/seccomp/Makefile
>> @@ -1,5 +1,5 @@
>> # SPDX-License-Identifier: GPL-2.0
>> -CFLAGS += -Wl,-no-as-needed -Wall
>> +CFLAGS += -Wl,-no-as-needed -Wall -isystem ../../../../usr/include/
>
> "../../../../usr/include/" directory doesn't have header files if
> different output directory is used for kselftests build like "make -C
> tools/tests/selftest O=build". Can you try adding recently added
> variable, KHDR_INCLUDES here which makes this kind of headers inclusion
> easy and correct for other build combinations as well?
>
>
Hi Muhammad,
I just pulled linux-next, and tried with KHDR_INCLUDES. It works. Very nice
work! I really appreciate you made headers inclusion compatible. However,
my case is a little more complicated. It will throw warnings with -I, using
-isystem can suppress these warnings, more details please refer to
https://lore.kernel.org/all/[email protected]/
According to this case, do you think will it be better to export header path
(KHDR_INCLUDES) without “-I”?
Thanks,
Sherry
>> "../../../../usr/include/" directory doesn't have header files if
>> different output directory is used for kselftests build like "make -C
>> tools/tests/selftest O=build". Can you try adding recently added
>> variable, KHDR_INCLUDES here which makes this kind of headers inclusion
>> easy and correct for other build combinations as well?
>>
>>
>
> Hi Muhammad,
>
> I just pulled linux-next, and tried with KHDR_INCLUDES. It works. Very nice
> work! I really appreciate you made headers inclusion compatible. However,
> my case is a little more complicated. It will throw warnings with -I, using
> -isystem can suppress these warnings, more details please refer to
> https://lore.kernel.org/all/[email protected]/
>
> According to this case, do you think will it be better to export header path
> (KHDR_INCLUDES) without “-I”?
Well said. I've thought about it and it seems like -isystem is better
than -I. I've sent a patch:
https://lore.kernel.org/linux-kselftest/[email protected]/
I'm looking forward to discussion on it.
Thanks,
Usama
On 2/14/22 9:12 PM, Muhammad Usama Anjum wrote:
>>> "../../../../usr/include/" directory doesn't have header files if
>>> different output directory is used for kselftests build like "make -C
>>> tools/tests/selftest O=build". Can you try adding recently added
>>> variable, KHDR_INCLUDES here which makes this kind of headers inclusion
>>> easy and correct for other build combinations as well?
>>>
>>>
>>
>> Hi Muhammad,
>>
>> I just pulled linux-next, and tried with KHDR_INCLUDES. It works. Very nice
>> work! I really appreciate you made headers inclusion compatible. However,
>> my case is a little more complicated. It will throw warnings with -I, using
>> -isystem can suppress these warnings, more details please refer to
>> https://lore.kernel.org/all/[email protected]/
>>
>> According to this case, do you think will it be better to export header path
>> (KHDR_INCLUDES) without “-I”?
> Well said. I've thought about it and it seems like -isystem is better
> than -I. I've sent a patch:
> https://lore.kernel.org/linux-kselftest/[email protected]/
> I'm looking forward to discussion on it.
The patch has been accepted. It should appear in linux-next soon. You
should be able to use KHDR_INCLUDES easily now.
Thanks,
Usama
On 2/15/22 11:17 AM, Muhammad Usama Anjum wrote:
>
> On 2/14/22 9:12 PM, Muhammad Usama Anjum wrote:
>>>> "../../../../usr/include/" directory doesn't have header files if
>>>> different output directory is used for kselftests build like "make -C
>>>> tools/tests/selftest O=build". Can you try adding recently added
>>>> variable, KHDR_INCLUDES here which makes this kind of headers inclusion
>>>> easy and correct for other build combinations as well?
>>>>
>>>>
>>>
>>> Hi Muhammad,
>>>
>>> I just pulled linux-next, and tried with KHDR_INCLUDES. It works. Very nice
>>> work! I really appreciate you made headers inclusion compatible. However,
>>> my case is a little more complicated. It will throw warnings with -I, using
>>> -isystem can suppress these warnings, more details please refer to
>>> https://lore.kernel.org/all/[email protected]/
>>>
>>> According to this case, do you think will it be better to export header path
>>> (KHDR_INCLUDES) without “-I”?
>> Well said. I've thought about it and it seems like -isystem is better
>> than -I. I've sent a patch:
>> https://lore.kernel.org/linux-kselftest/[email protected]/
>> I'm looking forward to discussion on it.
> The patch has been accepted. It should appear in linux-next soon. You
> should be able to use KHDR_INCLUDES easily now.
>
Sherry,
I pulled in your patch as a fix as is for 5.17-rc5.
Using KHDR_INCLUDES can be separate patch for next release.
This way the fix is going to be pulled for this release
without dependencies on other patches.
thanks,
-- Shuah