[LTP] [PATCH v2 2/2] mseal01: handle more possible errnos when pkey_alloc gets fail

[Li Wang]

On Mon, Aug 5, 2024 at 9:59 PM Petr Vorel <pvorel@suse.cz> wrote:

> Hi Li,
>
> ...
> >  static void test_pkey_mprotect(void)
> >  {
> > -     int ret;
> >       int pkey;
>
> > -     pkey = tst_syscall(__NR_pkey_alloc, 0, 0);
> > -     if (pkey == -1) {
> > -             if (errno == EINVAL)
> > -                     tst_brk(TCONF, "pku is not supported on this CPU");
> > +     check_pkey_support();
>
> > -             tst_brk(TBROK | TERRNO, "pkey_alloc() error");
> > -     }
> > +     pkey = ltp_pkey_alloc(0, 0);
> > +     if (pkey == -1)
> > +             tst_brk(TBROK | TERRNO, "pkey_alloc failed");
>
> > -     TST_EXP_FAIL(tst_syscall(__NR_pkey_mprotect,
> > +     TST_EXP_FAIL(ltp_pkey_mprotect(
> >               mem_addr, mem_size,
> >               PROT_NONE,
> >               pkey),
> >               EPERM);
>
> Reviewed-by: Petr Vorel <pvorel@suse.cz>
>
> But as I wrote in first patch, I would consider better to rewrite test to
> use
> .test_variants = 2 to test both raw syscall and libc wrapper. That would
> require
>

Sure, as I reply in 1/2 we can delay the work.



> to rewrite the lapi header. And it'd be then safer if pkey01.c would use
> always
> raw syscall variant, no need for not standard "ltp_pkey_mprotect()" name.
>

Good point, I will send a follow-up patch to resolve that.



>
> > -     ret = tst_syscall(__NR_pkey_free, pkey);
> > -     if (ret == -1)
> > +     if (ltp_pkey_free(pkey) == -1)
> >               tst_brk(TBROK | TERRNO, "pkey_free() error");
> >  }
>
> > @@ -150,7 +147,6 @@ static struct tst_test test = {
> >       .test = run,
> >       .tcnt = ARRAY_SIZE(tcases),
> >       .setup = setup,
> > -     .min_kver = "6.10",
> This change would be worth to mention in the changelog.
>

+1


> Kind regards,
> Petr
>
> >       .forks_child = 1,
> >  };
>
>

-- 
Regards,
Li Wang