| Message ID | 20190926144507.15765-1-cascardo@canonical.com |
|---|---|
| State | Accepted |
| Headers | show |
| Series | timer_create01: accept ENOTSUP | expand |
Hi! > CLOCK_REALTIME_ALARM and CLOCK_BOOTTIME_ALARM require an RTC with alarm > support, which may not be present on a system. In that case, the kernel > will return EOPNOTSUPP, which is defined as ENOTSUP in userspace. > > As these clocks are already possibly unsupported, accept that as an error > code besides the EINVAL code. > > Note that previously the kernel would incorrectly return the ENOTSUPP code, > which is undefined in userspace. With these unfixed kernels, the test will > fail when those RTCs are not present. Can you please include the kernel commit id that fixed the kernel here? Otherwise this patch is fine. > Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com> > --- > testcases/kernel/syscalls/timer_create/timer_create01.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/testcases/kernel/syscalls/timer_create/timer_create01.c b/testcases/kernel/syscalls/timer_create/timer_create01.c > index 258b6444c362..1ae5b9b2d1c2 100644 > --- a/testcases/kernel/syscalls/timer_create/timer_create01.c > +++ b/testcases/kernel/syscalls/timer_create/timer_create01.c > @@ -77,7 +77,8 @@ static void run(unsigned int n) > &created_timer_id)); > > if (TST_RET != 0) { > - if (possibly_unsupported(clock) && TST_ERR == EINVAL) { > + if (possibly_unsupported(clock) && > + (TST_ERR == EINVAL || TST_ERR == ENOTSUP)) { > tst_res(TPASS | TTERRNO, > "%s unsupported, failed as expected", > get_clock_str(clock)); > -- > 2.20.1 > > > -- > Mailing list info: https://lists.linux.it/listinfo/ltp
On Fri, Sep 27, 2019 at 04:49:02PM +0200, Cyril Hrubis wrote: > Hi! > > CLOCK_REALTIME_ALARM and CLOCK_BOOTTIME_ALARM require an RTC with alarm > > support, which may not be present on a system. In that case, the kernel > > will return EOPNOTSUPP, which is defined as ENOTSUP in userspace. > > > > As these clocks are already possibly unsupported, accept that as an error > > code besides the EINVAL code. > > > > Note that previously the kernel would incorrectly return the ENOTSUPP code, > > which is undefined in userspace. With these unfixed kernels, the test will > > fail when those RTCs are not present. > > Can you please include the kernel commit id that fixed the kernel here? > > Otherwise this patch is fine. f18ddc13af981ce3c7b7f26925f099e7c6929aba ("alarmtimer: Use EOPNOTSUPP instead of ENOTSUPP") Do you want me to send a v2, or will you fixup the patch? Thanks a lot. Cascardo. > > > Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com> > > --- > > testcases/kernel/syscalls/timer_create/timer_create01.c | 3 ++- > > 1 file changed, 2 insertions(+), 1 deletion(-) > > > > diff --git a/testcases/kernel/syscalls/timer_create/timer_create01.c b/testcases/kernel/syscalls/timer_create/timer_create01.c > > index 258b6444c362..1ae5b9b2d1c2 100644 > > --- a/testcases/kernel/syscalls/timer_create/timer_create01.c > > +++ b/testcases/kernel/syscalls/timer_create/timer_create01.c > > @@ -77,7 +77,8 @@ static void run(unsigned int n) > > &created_timer_id)); > > > > if (TST_RET != 0) { > > - if (possibly_unsupported(clock) && TST_ERR == EINVAL) { > > + if (possibly_unsupported(clock) && > > + (TST_ERR == EINVAL || TST_ERR == ENOTSUP)) { > > tst_res(TPASS | TTERRNO, > > "%s unsupported, failed as expected", > > get_clock_str(clock)); > > -- > > 2.20.1 > > > > > > -- > > Mailing list info: https://lists.linux.it/listinfo/ltp > > -- > Cyril Hrubis > chrubis@suse.cz
Hi! > > > CLOCK_REALTIME_ALARM and CLOCK_BOOTTIME_ALARM require an RTC with alarm > > > support, which may not be present on a system. In that case, the kernel > > > will return EOPNOTSUPP, which is defined as ENOTSUP in userspace. > > > > > > As these clocks are already possibly unsupported, accept that as an error > > > code besides the EINVAL code. > > > > > > Note that previously the kernel would incorrectly return the ENOTSUPP code, > > > which is undefined in userspace. With these unfixed kernels, the test will > > > fail when those RTCs are not present. > > > > Can you please include the kernel commit id that fixed the kernel here? > > > > Otherwise this patch is fine. > > f18ddc13af981ce3c7b7f26925f099e7c6929aba ("alarmtimer: Use EOPNOTSUPP instead of ENOTSUPP") > > Do you want me to send a v2, or will you fixup the patch? No need, I will add that info, thanks.
diff --git a/testcases/kernel/syscalls/timer_create/timer_create01.c b/testcases/kernel/syscalls/timer_create/timer_create01.c index 258b6444c362..1ae5b9b2d1c2 100644 --- a/testcases/kernel/syscalls/timer_create/timer_create01.c +++ b/testcases/kernel/syscalls/timer_create/timer_create01.c @@ -77,7 +77,8 @@ static void run(unsigned int n) &created_timer_id)); if (TST_RET != 0) { - if (possibly_unsupported(clock) && TST_ERR == EINVAL) { + if (possibly_unsupported(clock) && + (TST_ERR == EINVAL || TST_ERR == ENOTSUP)) { tst_res(TPASS | TTERRNO, "%s unsupported, failed as expected", get_clock_str(clock));
CLOCK_REALTIME_ALARM and CLOCK_BOOTTIME_ALARM require an RTC with alarm support, which may not be present on a system. In that case, the kernel will return EOPNOTSUPP, which is defined as ENOTSUP in userspace. As these clocks are already possibly unsupported, accept that as an error code besides the EINVAL code. Note that previously the kernel would incorrectly return the ENOTSUPP code, which is undefined in userspace. With these unfixed kernels, the test will fail when those RTCs are not present. Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com> --- testcases/kernel/syscalls/timer_create/timer_create01.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)