Re: [LTP] [PATCH v1 0/3] userfaultfd: Fix and remove compile-time TCONF handling

[Petr Vorel <pvorel@suse.cz>]

> > On 28.11.22 13:29, Martin Doucha wrote:
> > > On 28. 11. 22 12:57, Petr Vorel wrote:
> > > > Hi David,

> > > > If I remember correctly the reason for runtime check was ppc64le
> > > > missing userfaultfd minor fault support which needs to be check in runtime,
> > > > right? [1]. At least this is how I understand Martin's suggestion [2] to replace
> > > > compile time check with lapi. I'd state this reason at first commit message as
> > > > it's not obvious.

> > > The reason for runtime check is that the presence of the header file
> > > does not guarantee that the kernel supports UFFD API. The reason for
> > > LAPI is that we only care about actual kernel support, not build-time
> > > header files.
Also hope that people compile LTP with relevant headers (including LTP distro
package), but of course it's better to cover the case when kernel headers are
missing or for different kernel version.

> Thx, agree. BTW I was thinking about compile time check #ifdef
> UFFD_FEATURE_MINOR_SHMEM, but that's the same as ifdef HAVE_LINUX_USERFAULTFD_H.

> > Right. Petr, do you still want a commit message state?

> I can just add to commit message of the first commit the reason Martin reported:
> ppc64le on kernel 5.14 does not seem to support userfaultfd minor fault.

In the end I merged without any change in commit message. You described the
reason (runtime check) enough in the second commit. I was wrongly looking at
fist commit, sorry for bothering you.

Thank you both!

Kind regards,
Petr

> (no need to resent new version).

> Kind regards,
> Petr


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp