[LTP] [PATCH] configure.ac: Fix redefinition of mount_attr by checking sys/mount.h
Marcos Paulo de Souza
mpdesouza@suse.de
Tue Sep 27 15:10:03 CEST 2022
On Tue, Sep 27, 2022 at 09:09:46AM +0200, Petr Vorel wrote:
> Hi Marcos,
>
> ...
> > > > +++ b/configure.ac
> > > > @@ -71,6 +71,7 @@ AC_CHECK_HEADERS_ONCE([ \
> > > > sys/epoll.h \
> > > > sys/fanotify.h \
> > > > sys/inotify.h \
> > > > + sys/mount.h \
>
> > > Well, we don't add check for this, as sys/mount.h has been here since "ever", so
> > > it'd just make configure little bit slower for no gain.
>
> > > But you'd have to use somewhere compiler macro #ifdef HAVE_SYS_MOUNT_H, as this
> > > macro is generated by configure script. Adding header to the list does nothing.
>
> > It wasn't needed in my case, let me explain better below.
> Interesting, maybe I'm missing something.
>
> > > You obviously wanted to fix something, could you post the problem you're trying
> > > to solve + steps you did to ML?
>
> > I was seeing redefinition problems:
>
> > In file included from fsconfig01.c:9:
> > ../../../../include/lapi/fsmount.h:55:8: error: redefinition of 'struct mount_attr'
> > 55 | struct mount_attr {
> > | ^~~~~~~~~~
> > In file included from /usr/include/sys/mount.h:32,
> > from ../../../../include/lapi/fsmount.h:14:
> > /usr/include/linux/mount.h:129:8: note: originally defined here
> > 129 | struct mount_attr {
> > | ^~~~~~~~~~
> > ../../../../include/lapi/fsmount.h:110:19: error: conflicting types for 'mount_setattr'; have 'int(int, const char *, unsigned int, struct mount_attr *, size_t)' {aka 'int(int, const char *, unsigned int, struct mount_attr *, long unsigned int)'}
> > 110 | static inline int mount_setattr(int dirfd, const char *from_pathname, unsigned int flags,
> > | ^~~~~~~~~~~~~
> > In file included from fsconfig02.c:8:
> > ../../../../include/lapi/fsmount.h:55:8: error: redefinition of 'struct mount_attr'
> > 55 | struct mount_attr {
> > | ^~~~~~~~~~
> > /usr/include/sys/mount.h:316:12: note: previous declaration of 'mount_setattr' with type 'int(int, const char *, unsigned int, struct mount_attr *, size_t)' {aka 'int(int, const char *, unsigned int, struct mount_attr *, long unsigned int)'}
> > 316 | extern int mount_setattr (int __dfd, const char *__path, unsigned int __flags,
> > | ^~~~~~~~~~~~~
> > In file included from /usr/include/sys/mount.h:32,
> > from ../../../../include/lapi/fsmount.h:14:
> > /usr/include/linux/mount.h:129:8: note: originally defined here
> > 129 | struct mount_attr {
> > | ^~~~~~~~~~
> > ../../../../include/lapi/fsmount.h:110:19: error: conflicting types for 'mount_setattr'; have 'int(int, const char *, unsigned int, struct mount_attr *, size_t)' {aka 'int(int, const char *, unsigned int, struct mount_attr *, long unsigned int)'}
> > 110 | static inline int mount_setattr(int dirfd, const char *from_pathname, unsigned int flags,
> > | ^~~~~~~~~~~~~
> > /usr/include/sys/mount.h:316:12: note: previous declaration of 'mount_setattr' with type 'int(int, const char *, unsigned int, struct mount_attr *, size_t)' {aka 'int(int, const char *, unsigned int, struct mount_attr *, long unsigned int)'}
> > 316 | extern int mount_setattr (int __dfd, const char *__path, unsigned int __flags,
> > | ^~~~~~~~~~~~~
> > make[4]: *** [../../../../include/mk/rules.mk:45: fsconfig01] Error 1
> > make[4]: *** Waiting for unfinished jobs....
> > make[4]: *** [../../../../include/mk/rules.mk:45: fsconfig02] Error 1
> > make[3]: *** [../../../include/mk/generic_trunk_target.inc:108: all] Error 2
> > make[2]: *** [../../include/mk/generic_trunk_target.inc:108: all] Error 2
> > make[1]: *** [../include/mk/generic_trunk_target.inc:108: all] Error 2
> > make[1]: Leaving directory '/home/mpdesouza/git/ltp/testcases'
> > make: *** [Makefile:106: testcases-all] Error 2
>
> > In my TW system I found that the struct mount_attr was being defined in
> > sys/mount.h. Checking include/lapi/fsmount.h I found some ifdefs:
>
> > ...
> > #ifndef HAVE_STRUCT_MOUNT_ATTR
> > /*
> > * mount_setattr()
> > */
> > struct mount_attr {
> > uint64_t attr_set;
> > uint64_t attr_clr;
> > uint64_t propagation;
> > uint64_t userns_fd;
> > };
> > #endif
> > ...
>
> > #ifndef HAVE_MOUNT_SETATTR
> > static inline int mount_setattr(int dirfd, const char *from_pathname, unsigned int flags,
> > struct mount_attr *attr, size_t size)
> > {
> > return tst_syscall(__NR_mount_setattr, dirfd, from_pathname, flags,
> > attr, size);
> > }
> > #endif /* HAVE_MOUNT_SETATTR */
>
> > So, only adding the sys/mount.h entry to the headers list is enough to make
> > autoconf to find that sys/mount.h defines the same struct and function, fixing
> > the issue.
> Hm, I'm likely missing something :).
>
> $ rpm -q glibc
> glibc-2.36-5.1.x86_64
>
> $ grep -B2 '^struct mount_attr' /usr/include/sys/mount.h
> #ifndef MOUNT_ATTR_SIZE_VER0
> /* For mount_setattr. */
> struct mount_attr
>
> => I'm not sure who/where is supposed to define MOUNT_ATTR_SIZE_VER0.
> Is it in your config?
> Do you have different config than mine?
>
> grep MOUNT include/config.h
> #define HAVE_FSMOUNT 1
> #define HAVE_LINUX_MOUNT_H 1
> #define HAVE_MOUNT_SETATTR 1
> #define HAVE_MOVE_MOUNT 1
> #define HAVE_STRUCT_MOUNT_ATTR 1
> (i.e. HAVE_STRUCT_MOUNT_ATTR found with code in current master)
>
> I also tested fedora rawhide (glibc 2.36.9000), which we don't have in our CI:
> https://github.com/pevik/ltp/actions/runs/3133529596
> Works well.
>
> > > BTW I have no problem to build locally on the latest Tumbleweed. Also LTP
> > > package build for Tumbleweed is also fine:
> > > https://build.opensuse.org/package/binaries/benchmark:ltp:devel/ltp/openSUSE_Factory
> > > https://build.opensuse.org/package/binaries/benchmark:ltp:devel/ltp/openSUSE_Tumbleweed
>
> > Interesting... so is there anything wrong with my setup? Am I missing something?
>
> Hard to tell. Could you share your commands?
> I suppose it's the usual: use gcc and in the root directory:
> make autotools && ./configure && make -j`nproc`
My bad. I had this repository for some months already, I might have forgotten to
run "make autotools"... it now build as expected. Thanks and sorry about the
noise.
>
> Kind regards,
> Petr
More information about the ltp
mailing list