[LTP] [PATCH ltp] fs/ext4: enhance logic of ext4_nsec_timestamps_test

[Petr Vorel]

Hi,

> > > 2. Add "sleep 1" between mount and umount, to avoid umount failed due to
> > >    device busy (accessed by system service like systemd-udev)

> > NACK to this.

> > In a case that we need to wait for something like this we are supposed
> > to retry in a loop with exponential backoff, which is usually faster
> > than hardcoded sleep but also more robust.

> > There is a in-flight patch for adding TST_RETRY_FUNC function to the
> > shell library "[PATCH] tst_test.sh: achieve TST_RETRY_FUNC function in shell"
> > which should make this really easy but we would have to add it to the
> > old shell test library as well so that we can reuse it here.

> And I remebered that we already solved this by adding tst_umount
> function to the shell library. The systemd-udev is not first daemon to
> probe newly mounted filesystem, we used to have this kind of problems
> with gvfsd-trash as well.

> So fixing this should be as easy as replacing the umount with tst_umount
> function.
I'm using tst_umount in IMA tests and it works nice :).
BTW It's just that gvfsd-trash TINFO message is a bit too specific (in my case it was that
some processes were still using the mounted filesystem - the same issue as Yixin has).

Kind regards,
Petr