Re: [GIT PULL] Kselftest update for 4.11-rc1

[Mike Kravetz <mike.kravetz@oracle.com>]

On 02/26/2017 03:37 AM, Mike Rapoport wrote:
> On Sat, Feb 25, 2017 at 03:42:55PM -0800, Linus Torvalds wrote:
>> On Fri, Feb 24, 2017 at 1:09 PM, Shuah Khan <shuahkh@osg.samsung.com> wrote:
>>>
>>> Stephen saw a conflict during kselftest merge with next
>> [..]
>>
>> There was also a conflict with some of Andrew's new userfaultfd tests.
>>
>> I'm not at all sure I merged that correctly, although I tried to make
>> the resolution at least look sane. The odd thing about those new tests
>> is that it's all the same source file, except built with different
>> compile-time defines to be different test binaries.
>>
>> I didn't find any way to do that with any standard selftest makefile
>> magic, so I just did the build rules by hand.
>>
>> Somebody should actually verify that it does the right thing, please..
>> Adding the two Mike's that did those test additions to the participant
>> list.
> 
> The userfaultfd_* tests are built and they are working just fine.

Thanks Mike,  I also verified that the tests build/run as they should.

> Maybe it's worth adding userfaultfd_{hugetlb,shmem}.c which will define the
> required defines and include userfaultfd.c?
> Than we can get away with only specifying LDFLAGS += -lpthread in the
> Makefile.

I started the current scheme when adding support for hugetlbfs, which I
thought would be a 'one off'.  Did not know about the shmem work that was
in progress.

Another option is to only have a single userfaultfd executable and pass
an option (anon, hugetlb, shmem ...) that indicates the type of pages/mapping
to test.

In any case, the makefile modifications work so this is not urgent.  Perhaps
we can discuss with Andrea and figure out what would be the best scheme
moving forward.

-- 
Mike Kravetz