[LTP] [PATCH] fanotify24: Verify expected count/offset info in pre content events

[Amir Goldstein]

On Mon, Oct 20, 2025 at 10:22 PM Petr Vorel <pvorel@suse.cz> wrote:
>
> Hi Amir, all,
>
> > To test fix commit 28bba2c2935e2 ("fsnotify: Pass correct offset to
> > fsnotify_mmap_perm()"), diversify the offsets and count used for mmap()
> > write() and read() and verify that the FAN_PRE_ACCESS events report the
> > expected count/offset.
>
> > For the FAN_PRE_ACCESS events generated by execve(), we cannot
> > anticipate the exact ranges, so we set 0 count to skip this verification.
>
> > Also add prints of path of the fd passed with the event (not verified
> > against expected path).
>
> > Make sure that we take the expected error value for an operation
> > (e.g. read) from a matching event type (e.g. FAN_PRE_ACCESS).
>
> Thanks for the update.  LGTM, but it'd be great if some of kernel developers
> also had look into it. Few minor notes below.
>
> Reviewed-by: Petr Vorel <pvorel@suse.cz>
>
> I also restarted failing jobs in github actions:
> https://github.com/linux-test-project/ltp/actions/runs/18599812482
>
> Unfortunately due patchwork API limitation, the failing jobs aren't replaced
> with successful ones, instead the fixes are appended:
> https://patchwork.ozlabs.org/project/ltp/patch/20251017161639.2088158-1-amir73il@gmail.com/
>
> > Signed-off-by: Amir Goldstein <amir73il@gmail.com>
> > ---
> >  .../kernel/syscalls/fanotify/fanotify24.c     | 167 +++++++++++-------
> >  1 file changed, 107 insertions(+), 60 deletions(-)
>
> > diff --git a/testcases/kernel/syscalls/fanotify/fanotify24.c b/testcases/kernel/syscalls/fanotify/fanotify24.c
> > index 27f0663ce..8f2dee55b 100644
> > --- a/testcases/kernel/syscalls/fanotify/fanotify24.c
> > +++ b/testcases/kernel/syscalls/fanotify/fanotify24.c
> > @@ -9,6 +9,8 @@
> >  /*\
> >   * - Test fanotify pre-content events
> >   * - Test respond to permission/pre-content events with cutsom error code
> > + * - Test count/offset info bug that was fixed by commit
> > + *   28bba2c2935e2 "fsnotify: Pass correct offset to fsnotify_mmap_perm()"
> >   */
>
> >  #define _GNU_SOURCE
> > @@ -44,6 +46,8 @@
> >  #define FILE_EXEC_PATH MOUNT_PATH"/"TEST_APP
>
> >  static char fname[BUF_SIZE];
> > +static char symlnk[BUF_SIZE];
> > +static char fdpath[BUF_SIZE];
> >  static char buf[BUF_SIZE];
> >  static volatile int fd_notify;
> >  static size_t page_sz;
> > @@ -55,6 +59,8 @@ static char event_buf[EVENT_BUF_LEN];
> >  struct event {
> >       unsigned long long mask;
> >       unsigned int response;
> > +     unsigned long pgcount;
> > +     unsigned long pgoff;
> >  };
>
> >  static struct tcase {
> > @@ -68,11 +74,11 @@ static struct tcase {
> >               INIT_FANOTIFY_MARK_TYPE(INODE),
> >               FAN_OPEN_PERM | FAN_PRE_ACCESS,
> >               {
> > -                     {FAN_OPEN_PERM, FAN_ALLOW},
> > -                     {FAN_PRE_ACCESS, FAN_ALLOW},
> > -                     {FAN_PRE_ACCESS, FAN_ALLOW},
> > -                     {FAN_PRE_ACCESS, FAN_DENY_ERRNO(EIO)},
> > -                     {FAN_OPEN_PERM, FAN_DENY_ERRNO(EBUSY)}
> > +                     {FAN_OPEN_PERM, FAN_ALLOW,0 ,0},
> nit: space is usually before comma, not after. I'll fix it before merge.
> I also prefer to use designated initializers when there are more struct members
> and some of them are zero. But it's up to you, or I can change it later in a
> separate patch.
>

Please change it later.

Thanks!
Amir.