[LTP] [PATCH 2/2] setitimer03: convert to new API

[Li Wang]

On Thu, Oct 20, 2022 at 6:40 PM Richard Palethorpe <rpalethorpe@suse.de>
wrote:

> Hello,
>
> Li Wang <liwang@redhat.com> writes:
>
> > On Thu, Oct 20, 2022 at 5:30 PM Li Wang <liwang@redhat.com> wrote:
> >
> >  Richard Palethorpe <rpalethorpe@suse.de> wrote:
> >
> >  > -static struct itimerval *value;
> >  > +static struct itimerval *value, *ovalue;
> >  > +
> >  > +static struct tcase {
> >  > +       int which;
> >  > +       struct itimerval **val;
> >  > +       struct itimerval **oval;
> >  > +       int exp_errno;
> >
> >  There is a whitespace error here (see checkpatch/make check)
> >
> >  yes, thanks.
> >
> >
> >
> >  > +} tcases[] = {
> >  > +       {ITIMER_REAL,    &value, &ovalue, EFAULT},
> >  > +       {ITIMER_VIRTUAL, &value, &ovalue, EFAULT},
> >  > +       {-ITIMER_PROF,   &value, &ovalue, EINVAL},
> >  > +};
> >
> >  Why do we need value and ovalue in the struct?
> >
> >  Becuase it does not allow parsing an invalid pointer address
> >  from a structure, we have to give a valid address which pointer
> >  to save an invalid address. Otherwise segement fault will
> >  be hit in execution.
> >
> > On the other side, it also does not allow to initializer element
> > is not constant in structure. So the two-level pointer is only the
> > way I can make all things comprised here.
>
> I'm not sure what you mean and I don't understand why we need this
> struct at all. The following works and produces better output:
>

Ok, I got your point.  Previously I was stuck in thinking about how to keep
the unified tcase struct for error testing.

Seems I should give up the fixed mindset.

static struct tcase {
     ...
} tcases[] = {
     ...
};



>
> @@ -20,17 +20,6 @@
>
>  static struct itimerval *value, *ovalue;
>
> -static struct tcase {
> -       int which;
> -       struct itimerval **val;
> -       struct itimerval **oval;
> -       int exp_errno;
> -} tcases[] = {
> -       {ITIMER_REAL,    &value, &ovalue, EFAULT},
> -       {ITIMER_VIRTUAL, &value, &ovalue, EFAULT},
> -       {-ITIMER_PROF,   &value, &ovalue, EINVAL},
> -};
> -
>  static int sys_setitimer(int which, void *new_value, void *old_value)
>  {
>         return tst_syscall(__NR_setitimer, which, new_value, old_value);
> @@ -38,13 +27,17 @@ static int sys_setitimer(int which, void *new_value,
> void *old_value)
>
>  static void verify_setitimer(unsigned int i)
>  {
> -        struct tcase *tc = &tcases[i];
> -
> -       if (tc->exp_errno == EFAULT)
> -               *(tc->oval) = (struct itimerval *)-1;
> -
> -       TST_EXP_FAIL(sys_setitimer(tc->which, *(tc->val), *(tc->oval)),
> -                       tc->exp_errno);
> +       switch (i) {
> +       case 0:
> +               TST_EXP_FAIL(sys_setitimer(ITIMER_REAL, value, (void
> *)-1), EFAULT);
> +               break;
> +       case 1:
> +               TST_EXP_FAIL(sys_setitimer(ITIMER_VIRTUAL, value, (void
> *)-1), EFAULT);
> +               break;
> +       case 2:
> +               TST_EXP_FAIL(sys_setitimer(-ITIMER_PROF, value, ovalue),
> EINVAL);
> +               break;
> +       }
>  }
>
>  static void setup(void)
> @@ -56,7 +49,7 @@ static void setup(void)
>  }
>
>  static struct tst_test test = {
> -        .tcnt = ARRAY_SIZE(tcases),
> +        .tcnt = 3,
>         .test = verify_setitimer,
>         .setup = setup,
>         .bufs = (struct tst_buffers[]) {
>
> This prints
>
> setitimer02.c:32: TPASS: sys_setitimer(ITIMER_REAL, value, (void *)-1) :
> EFAULT (14)
> setitimer02.c:35: TPASS: sys_setitimer(ITIMER_VIRTUAL, value, (void *)-1)
> : EFAULT (14)
> setitimer02.c:38: TPASS: sys_setitimer(-ITIMER_PROF, value, ovalue) :
> EINVAL (22)
>
> instead of
>
> setitimer02.c:46: TPASS: sys_setitimer(tc->which, *(tc->val), *(tc->oval))
> : EFAULT (14)
> setitimer02.c:46: TPASS: sys_setitimer(tc->which, *(tc->val), *(tc->oval))
> : EFAULT (14)
> setitimer02.c:46: TPASS: sys_setitimer(tc->which, *(tc->val), *(tc->oval))
> : EINVAL (22)
>
> The same values are passed to the syscall according to strace.
>
> --
> Thank you,
> Richard.
>
>

-- 
Regards,
Li Wang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20221020/275013ec/attachment-0001.htm>