Thank you for your feedback! I'm submitting a fixed version. It has slightly changed logic though. The results are: (before) # cat rt/tst-bz28213.test-result FAIL: rt/tst-bz28213 original exit status 1 # cat rt/tst-bz28213.out Didn't expect signal from child: got `Segmentation fault' (after) # cat rt/tst-bz28213.test-result PASS: rt/tst-bz28213 original exit status 0 # cat rt/tst-bz28213.out чт, 12 авг. 2021 г. в 09:19, Siddhesh Poyarekar : > > On 8/11/21 4:56 PM, Никита Попов via Libc-alpha wrote: > > Hello. I'm submitting a test case for recent bug 28213. It shows the > > following results. > > Before fix of bug 28213: > > # cat ./rt/tst-bz28213.out > > Child caused segmentation fault > > # cat ./rt/tst-bz28213.test-result > > FAIL: rt/tst-bz28213 > > original exit status 1 > > > > After fix of bug 28213: > > # cat ./rt/tst-bz28213.out > > # cat ./rt/tst-bz28213.test-result > > PASS: rt/tst-bz28213 > > original exit status 0 > > I managed to get libc cause segmentation fault. > > Looking forward to receiving feedback from you. > > Thank you for sharing this. My detailed review comments are inline. > Overall I think this can be done without forking a new process since the > test infrastructure already does that and is capable of reporting > unexpected signals. All you need to test is that the program runs to > completion and ensure that the helper thread is called, perhaps by > simply sleeping for a while. > > > From eb51dcf05befe6e839102e0cb622b1afdbf72a7a Mon Sep 17 00:00:00 2001 > > From: Nikita Popov > > Date: Wed, 11 Aug 2021 14:36:50 +0500 > > Subject: [PATCH] librt: add test (bug 28213) > > To: libc-alpha@sourceware.org > > > > This test implements following logic: > > 1) Create dummy message queue, register it with mq_notify (using NULL attributes), > > immediately close the queue. > > Helper thread should cause NULL pointer dereference by this moment. > > 2) Create another queue and try to send a message via the helper thread. > > Test is considered successful if the callback function receives the same message as was sent. > > > > Signed-off-by: Nikita Popov > > --- > > rt/Makefile | 1 + > > rt/tst-bz28213.c | 174 +++++++++++++++++++++++++++++++++++++++++++++++ > > 2 files changed, 175 insertions(+) > > create mode 100644 rt/tst-bz28213.c > > > > diff --git a/rt/Makefile b/rt/Makefile > > index 113cea03a5..910e775995 100644 > > --- a/rt/Makefile > > +++ b/rt/Makefile > > @@ -74,6 +74,7 @@ tests := tst-shm tst-timer tst-timer2 \ > > tst-aio7 tst-aio8 tst-aio9 tst-aio10 \ > > tst-mqueue1 tst-mqueue2 tst-mqueue3 tst-mqueue4 \ > > tst-mqueue5 tst-mqueue6 tst-mqueue7 tst-mqueue8 tst-mqueue9 \ > > + tst-bz28213 \ > > tst-timer3 tst-timer4 tst-timer5 \ > > tst-cpuclock2 tst-cputimer1 tst-cputimer2 tst-cputimer3 \ > > tst-shm-cancel \ > > diff --git a/rt/tst-bz28213.c b/rt/tst-bz28213.c > > new file mode 100644 > > index 0000000000..6c043a9017 > > --- /dev/null > > +++ b/rt/tst-bz28213.c > > @@ -0,0 +1,174 @@ > > +/* Bug 28213: test for NULL pointer dereference in mq_notify. > > + Copyright (C) 2018-2021 Free Software Foundation, Inc. > > Please make this copyright "The GNU Toolchain Authors" since you don't > have a copyright assignment with the FSF. > > > + This file is part of the GNU C Library. > > + > > + The GNU C Library is free software; you can redistribute it and/or > > + modify it under the terms of the GNU Lesser General Public > > + License as published by the Free Software Foundation; either > > + version 2.1 of the License, or (at your option) any later version. > > + > > + The GNU C Library is distributed in the hope that it will be useful, > > + but WITHOUT ANY WARRANTY; without even the implied warranty of > > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > > + Lesser General Public License for more details. > > + > > + You should have received a copy of the GNU Lesser General Public > > + License along with the GNU C Library; if not, see > > + . */ > > + > > +#include > > +#include > > +#include > > +#include > > +#include > > +#include > > +#include > > +#include > > +#include > > +#include > > +#include > > +#include > > + > > +static const char check_bz28213_name[] = "/bz28213_queue"; > > +static const char check_bz28213_msg[] = "dummy"; > > + > > +static void > > +check_bz28213_cb (union sigval sv) > > +{ > > + char buf[sizeof (check_bz28213_msg)]; > > + mqd_t m = sv.sival_int; > > + ssize_t n; > > + > > + if ((n = mq_receive (m, buf, sizeof (buf), NULL)) < 0L || > > + (size_t) n != sizeof (buf)) > > + exit (1); > > + > > + if (memcmp (buf, check_bz28213_msg, sizeof (buf)) != 0) > > + exit (1); > > + > > + exit (0); > > +} > > Instead of exiting with error, please use the TEST_VERIFY_EXIT macro in > support/check.h. > > > + > > +static void > > +check_bz28213 (void) > > +{ > > + mqd_t m; > > + struct sigevent sev; > > + struct mq_attr attr; > > + unsigned int i; > > + > > + /* First iteration should lead to undefined behavior due to NULL pointer dereference. > > + Second iteration tests whether helper thread still works. */ > > + for (i = 0U; i < 2U; i++) > > + { > > + memset (&attr, '\0', sizeof (attr)); > > + attr.mq_maxmsg = 1; > > + attr.mq_msgsize = sizeof (check_bz28213_msg); > > + > > + if ((m = mq_open (check_bz28213_name, > > + O_RDWR | O_CREAT | O_EXCL, > > + 0600, > > + &attr)) < 0) > > + exit (1); > > + > > + if (mq_unlink (check_bz28213_name) < 0) > > + exit (1); > > + > > + memset (&sev, '\0', sizeof (sev)); > > + sev.sigev_notify = SIGEV_THREAD; > > + sev.sigev_value.sival_int = m; > > + sev.sigev_notify_function = check_bz28213_cb; > > + > > + if (mq_notify (m, &sev) < 0) > > + exit (1); > > + > > + if (i && mq_send (m, check_bz28213_msg, sizeof (check_bz28213_msg), 1) < 0) > > + exit (1); > > Likewise for these exits, please use TEST_VERIFY_EXIT instead. > > > + > > + if (!i) > > + mq_close (m); > > + } > > + > > + alarm (10); > > + > > + while (1) > > + pause (); > > Instead you could define TIMEOUT to DEFAULT_TIMEOUT and then use a > sleep(DEFAULT_TIMEOUT / 2) to wait long enough (10 seconds ought to be > enough for anybody?) for the helper thread to run. Architecture > maintainers can later bump up TIMEOUT if they find that the test times > out on their architecture. > > > +} > > + > > +/* Skip entire testing if queues are not implemented */ > > +static int > > +check_mq_api (void) > > +{ > > + struct mq_attr attr; > > + mqd_t m; > > + int rc = 0; > > + > > + memset (&attr, '\0', sizeof (attr)); > > + attr.mq_maxmsg = 1; > > + attr.mq_msgsize = 1; > > + > > + if ((m = mq_open (check_bz28213_name, > > + O_RDWR | O_CREAT | O_EXCL, > > + 0600, > > + &attr)) < 0) > > + { > > + if (errno == ENOSYS) > > + { > > + rc = 1; > > + printf ("SKIP: not implemented\n"); > > + } > > + > > + return rc; > > + } > > + > > + mq_unlink (check_bz28213_name); > > + mq_close (m); > > + > > + return rc; > > +} > > + > > +static int > > +do_test (void) > > +{ > > + pid_t pid; > > + int status, rc; > > + > > + if (check_mq_api () == 1) > > + return 0; > > Please use FAIL_UNSUPPORTED() here. > > > + > > + if ((pid = xfork ()) == 0) > > + check_bz28213 (); > > + > > + if (TEMP_FAILURE_RETRY (waitpid (pid, &status, 0)) != pid) > > + { > > + kill (pid, SIGKILL); > > + printf ("waitpid failed\n"); > > + return 1; > > + } > > + > > + rc = 1; > > + if (WIFEXITED (status)) > > + { > > + int child_rc; > > + > > + if ((child_rc = WEXITSTATUS (status)) == 0) > > + rc = 0; > > + else > > + printf ("Child returned non-zero exit code [%d]\n", child_rc); > > + } > > + else if (WIFSIGNALED (status)) > > + { > > + int child_signo; > > + > > + if ((child_signo = WTERMSIG (status)) == SIGABRT) > > + printf ("Child timed out\n"); > > + else if (child_signo == SIGSEGV) > > + printf ("Child caused segmentation fault\n"); > > + else > > + printf ("Child terminated with signal %d\n", child_signo); > > + } > > Forking should not be necessary since the test driver already reports > unexpected signals. > > > + > > + return rc; > > Return 0 in the end. > > > +} > > + > > +#include > > -- > > 2.17.1 >