On Tue, 2021-04-27 at 01:05 +0800, chensong_2...@189.cn wrote:
>
> 在 2021/4/23 下午8:48, Bezdeka, Florian 写道:
> > On Wed, 2021-04-21 at 16:38 +0800, Song Chen wrote:
> > >
> > >
> > >
> > >
> > >
> > > +static int test_sc_cobalt_clock_nanosleep64(void)+{
> > > + long ret;
> > > + int code = __xn_syscode(sc_cobalt_clock_nanosleep64);
> > > + struct xn_timespec64 next, rmt;
> > > + struct timespec now;
> > > +
> > > + /* Make sure we don't crash because of NULL pointers */
> > > + ret = syscall(code, NULL, NULL, NULL, NULL);
> > > + if (ret == -1 && errno == ENOSYS) {
> > > + smokey_note("clock_nanosleep64: skipped. (no kernel support)");
> > > + return 0; // Not implemented, nothing to test, success
> > > + }
> > > + if (!smokey_assert(ret == -1) || !smokey_assert(errno == EFAULT))
> > > + return errno;
> > > +
> > > + /* Provide a valid 64bit timespec*/
> > > + ret = clock_gettime(CLOCK_MONOTONIC, &now);
> > > + if (ret) {
> > > + smokey_warning("clock_gettime failed in clock_nanosleep64.");
> > > + return errno;
> > > + }
> > > +
> > > + /* Providing an invalid address has to deliver EFAULT */
> > > + ret = syscall(code, CLOCK_MONOTONIC, TIMER_ABSTIME,
> > > + (void *)0xdeadbeefUL, &rmt);
> > > + if (!smokey_assert(ret == -1) || !smokey_assert(errno == EFAULT))
> > > + return errno;
> > > +
> > > + next.tv_sec = now.tv_sec + 1;
> > > + next.tv_nsec = now.tv_nsec;
> > > +
> > > + ret = syscall(code, CLOCK_MONOTONIC, TIMER_ABSTIME,
> > > + &next, (void *)0xdeadbeefUL);
> > > + if (!smokey_assert(ret == -1) || !smokey_assert(errno == EFAULT))
> > > + return errno;
> > > +
> > > + next.tv_sec = next.tv_sec + 1;
> >
> > That looks like duplicate code. The last call to nanosleep was an error
> > handling check, so we would end up here adding 2 secs to "now".
>
> point taken
> >
> > > +
> > > + ret = syscall(code, CLOCK_MONOTONIC, TIMER_ABSTIME, &next, &rmt);
> > > + if (!smokey_assert(!ret))
> > > + return errno;
> > > +
> >
> > As long as nobody has a better idea we should make sure that we don't
> > come back to early. Checking for the upper boundary requires some magic
> > numbers as already discussed, so let's check at least the lower
> > boundary.
>
> point is not taken, could you please explain upper/lower boundary more
> specifically, thanks.
Lower boundary: We should not come back too early
Upper boundary: We should not come back too late
While the lower boundary is defined, the upper boundary depends on the
used architecture and/or platform.
I had something similar to [1] in mind. That's the test case for
nanosleep() in glibc. Checking the lower boundary by making sure we
don't come back too early.
Do you agree? Was that helpful?
[1] https://elixir.bootlin.com/glibc/latest/source/posix/tst-nanosleep.c
>
> Song
>
> >
> > > + return 0;
> > > +}
> > > +
> > > static int run_y2038(struct smokey_test *t, int argc, char *const
> > > argv[])
> > > {
> > > int ret;
> > > @@ -244,5 +290,9 @@ static int run_y2038(struct smokey_test *t, int argc,
> > > char *const argv[])
> > > if (ret)
> > > return ret;
> > >
> > >
> > >
> > >
> > > + ret = test_sc_cobalt_clock_nanosleep64();
> > > + if (ret)
> > > + return ret;
> > > +
> > > return 0;
> > > }
> >
