From mboxrd@z Thu Jan  1 00:00:00 1970
From: Jan Stancek <jstancek@redhat.com>
Date: Sun, 16 Jun 2019 03:04:43 -0400 (EDT)
Subject: [LTP] [PATCH v3 3/3] syscalls/tgkill03: add new test
In-Reply-To: <CAEemH2cCvaPKa480Y3L5+V8ihRtwcTYZPo5ebU7Ss8tfRDXkuQ@mail.gmail.com>
References: <1552978181-27748-1-git-send-email-sumit.garg@linaro.org>
 <1552978181-27748-4-git-send-email-sumit.garg@linaro.org>
 <1142529550.28348566.1560640103965.JavaMail.zimbra@redhat.com>
 <CAEemH2cCvaPKa480Y3L5+V8ihRtwcTYZPo5ebU7Ss8tfRDXkuQ@mail.gmail.com>
Message-ID: <847341899.28355719.1560668683548.JavaMail.zimbra@redhat.com>
List-Id: <ltp.lists.linux.it>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: ltp@lists.linux.it



----- Original Message -----
> On Sun, Jun 16, 2019 at 7:08 AM Jan Stancek <jstancek@redhat.com> wrote:
> 
> >
> > ----- Original Message -----
> > > +static void setup(void)
> > > +{
> > > +     sigset_t sigusr1;
> > > +     pthread_t defunct_thread;
> > > +
> > > +     sigemptyset(&sigusr1);
> > > +     sigaddset(&sigusr1, SIGUSR1);
> > > +     pthread_sigmask(SIG_BLOCK, &sigusr1, NULL);
> > > +
> > > +     parent_tgid = getpid();
> > > +     parent_tid = sys_gettid();
> > > +
> > > +     SAFE_PTHREAD_CREATE(&child_thread, NULL, child_thread_func, NULL);
> > > +
> > > +     TST_CHECKPOINT_WAIT(0);
> >
> > I'm seeing reports of this test failing on s390x:
> >
> > st_test.c:1096: INFO: Timeout per run is 0h 05m 00s
> > tgkill03.c:92: PASS: Invalid tgid failed as expected: EINVAL
> > tgkill03.c:92: PASS: Invalid tid failed as expected: EINVAL
> > tgkill03.c:92: PASS: Invalid signal failed as expected: EINVAL
> > tgkill03.c:96: FAIL: Defunct tid should have failed with ESRCH: SUCCESS
> > tgkill03.c:92: PASS: Defunct tgid failed as expected: ESRCH
> > tgkill03.c:99: PASS: Valid tgkill call succeeded
> >
> > and I suspect this piece:
> >
> > > +
> > > +     SAFE_PTHREAD_CREATE(&defunct_thread, NULL, defunct_thread_func,
> > NULL);
> > > +
> > > +     SAFE_PTHREAD_JOIN(defunct_thread, NULL);
> > > +}
> >
> > glibc pthread_join() waits for CLONE_CHILD_CLEARTID to clear tid,
> > and then resumes. Which kernel does at:
> >   do_exit
> >     exit_mm
> >       mm_release
> >         put_user(0, tsk->clear_child_tid);
> >
> > so there's still work to be done after that, and I suspect tid is still
> > valid
> > while that happens.
> >
> > My first idea: wait until /proc/pid/task/<tid> disappears.
> >
> 
> The anaysis is probably right, but this idea doesn't work for me. Seems
> /proc/pid/task/<tid> is not the key point to confirm that tid has been
> clear.
> 
> I just have a try as below:
> 
> ===========
> # for i in `seq 1000`; do echo "i = $i" && ./tgkill03 || break; done
> ...
> i = 96
> tst_test.c:1112: INFO: Timeout per run is 0h 05m 00s
> tgkill03.c:106: FAIL: Defunct tid should have failed with ESRCH: SUCCESS
> 
> ===========
> --- a/testcases/kernel/syscalls/tgkill/tgkill03.c
> +++ b/testcases/kernel/syscalls/tgkill/tgkill03.c
> @@ -5,6 +5,7 @@
>   * Test simple tgkill() error cases.
>   */
> 
> +#include <stdio.h>
>  #include <pthread.h>
>  #include <pwd.h>
>  #include <sys/types.h>
> @@ -19,6 +20,7 @@ static pid_t parent_tgid;
>  static pid_t parent_tid;
>  static pid_t child_tid;
>  static pid_t defunct_tid;
> +char buf[1024];
> 
>  static const int invalid_pid = -1;
> 
> @@ -35,6 +37,8 @@ static void *defunct_thread_func(void *arg)
>  {
>         defunct_tid = sys_gettid();
> 
> +       sprintf(buf, "/proc/pid/task/%d", defunct_tid);

How about?
          sprintf(buf, "/proc/%d/task/%d", getpid(), defunct_tid);