Re: [PATCH] fsstress: improve error message on check_cwd() error

[Filipe Manana <fdmanana@gmail.com>]

On Wed, Sep 22, 2021 at 9:07 PM Luis Chamberlain <mcgrof@kernel.org> wrote:
>
> On Wed, Sep 22, 2021 at 11:15:08AM +0100, Filipe Manana wrote:
> > On Tue, Sep 21, 2021 at 6:51 PM Luis Chamberlain <mcgrof@kernel.org> wrote:
> > >
> > > I ran into an error with generic/083 with xfs due to check_cwd() but
> > > why it failed is not clear because there are two types of
> > > failures:
> > >
> > >   o stat64() failed (likely -ENOMEM is my guess)
> > >   o the inode actually changed
> > >
> > > Thow a bone out to developers so that in case en error does happen
> > > they know which rabbit hole to go down on.
> > >
> > > Cc: Anthony Iliopoulos <ailiopoulos@suse.de>
> > > Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
> > > ---
> > >  ltp/fsstress.c | 15 ++++++++++++++-
> > >  1 file changed, 14 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/ltp/fsstress.c b/ltp/fsstress.c
> > > index b4ddf5e2..d2f09901 100644
> > > --- a/ltp/fsstress.c
> > > +++ b/ltp/fsstress.c
> > > @@ -9,6 +9,7 @@
> > >  #include <sys/uio.h>
> > >  #include <stddef.h>
> > >  #include <stdbool.h>
> > > +#include <string.h>
> > >  #include "global.h"
> > >
> > >  #ifdef HAVE_BTRFSUTIL_H
> > > @@ -943,9 +944,21 @@ check_cwd(void)
> > >  {
> > >  #ifdef DEBUG
> > >         struct stat64   statbuf;
> > > +       int ret;
> > > +
> > > +       ret = stat64(".", &statbuf);
> > > +       if (ret !=0) {
> > > +               fprintf(stderr, "fsstress: check_cwd stat64 failed with: %d (%s)\n",
> > > +                       ret, strerror(ret));
> >
> > The error is not in 'ret' but instead in 'errno', so it should print
> > errno and strerror(errno).
>
> Good call will fix.
>
> > > +               goto out;
> > > +       }
> > >
> > > -       if (stat64(".", &statbuf) == 0 && statbuf.st_ino == top_ino)
> > > +       if (statbuf.st_ino == top_ino)
> > >                 return;
> > > +
> > > +       fprintf(stderr, "fsstress: check_cwd statbuf.st_ino (%ld) != top_ino (%ld)\n",
> > > +               (long)statbuf.st_ino, (long)top_ino);
> >
> > ino_t is guaranteed to be an unsigned integer, and can be either 64 or
> > 32 bits wide [1].
> > Casting to unsigned 64 bits would be better imo.
>
> There are other similar mi-uses in the code, should I fix those to
> then?

If you want to, as a separate patch.
Noticing now that since the code is using struct stat64, using %llu
for the printfs without any casts should be fine [1].

Thanks.

[1] https://www.mkssoftware.com/docs/man5/struct_stat.5.asp

>
>   Luis
>


-- 
Filipe David Manana,

“Whether you think you can, or you think you can't — you're right.”