public inbox for ltp@lists.linux.it
 help / color / mirror / Atom feed
From: Cyril Hrubis <chrubis@suse.cz>
To: ltp@lists.linux.it
Subject: [LTP] [PATCH] Convert ptrace01.c to new library
Date: Tue, 8 Oct 2019 12:47:23 +0200	[thread overview]
Message-ID: <20191008104723.GA9170@rei.lan> (raw)
In-Reply-To: <20191001070111.31369-1-jcronenberg@suse.de>

Hi!
> This is just a port of ptrace01.c. I wanted to make sure what I'm doing is right before porting the rest. I tried to make as little changes as possible so the test should be pretty much the same.

Generally looks good, a few comments below.

Also the part before the --- should look like:

Ported ptrace01 to new library.

Signed-off-by: Jorik Cronenberg <jcronenberg@suse.de>

> ---

And the informal part about making minimal changes should be here, after
the ---, which is the part that is not include in the final commit.

>  testcases/kernel/syscalls/ptrace/ptrace01.c | 296 +++++++---------------------
>  1 file changed, 72 insertions(+), 224 deletions(-)
> 
> diff --git a/testcases/kernel/syscalls/ptrace/ptrace01.c b/testcases/kernel/syscalls/ptrace/ptrace01.c
> index 5a7ed5ffd..ab9ef35ee 100644
> --- a/testcases/kernel/syscalls/ptrace/ptrace01.c
> +++ b/testcases/kernel/syscalls/ptrace/ptrace01.c
> @@ -1,36 +1,13 @@
> -/*
> +// SPDX-License-Identifier: GPL-2.0-or-later
> +/* 
>   * Copyright (c) Wipro Technologies Ltd, 2002.  All Rights Reserved.
>   *
> - * This program is free software; you can redistribute it and/or modify it
> - * under the terms of version 2 of the GNU General Public License as
> - * published by the Free Software Foundation.

Being pedantic about the licences this is GPL-2.0 only and we cannot
change the original licence without consent from the author, so we have
to keep the code under GPL-2.0 and not the -or-later variant.

> + * Author:	Saji Kumar.V.R <saji.kumar@wipro.com>
>   *
> - * This program is distributed in the hope that it would be useful, but
> - * WITHOUT ANY WARRANTY; without even the implied warranty of
> - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
> + * Ported to new library:
> + * 10/2019	Jorik Cronenberg <jcronenberg@suse.de>
>   *
> - * You should have received a copy of the GNU General Public License along
> - * with this program; if not, write the Free Software Foundation, Inc.,
> - * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
> - *
> - */
> -/**********************************************************
> - *
> - *    TEST IDENTIFIER	: ptrace01
> - *
> - *    EXECUTED BY	: anyone
> - *
> - *    TEST TITLE	: functionality test for ptrace(2)
> - *
> - *    TEST CASE TOTAL	: 2
> - *
> - *    AUTHOR		: Saji Kumar.V.R <saji.kumar@wipro.com>
> - *
> - *    SIGNALS
> - * 	Uses SIGUSR1 to pause before test if option set.
> - * 	(See the parse_opts(3) man page).
> - *
> - *    DESCRIPTION
> + * DESCRIPTION
>   *	This test case tests the functionality of ptrace() for
>   *	PTRACE_TRACEME & PTRACE_KILL requests.
>   *	Here, we fork a child & the child does ptrace(PTRACE_TRACEME, ...).
> @@ -44,243 +21,114 @@
>   * 		In both cases, child should stop & notify parent on reception
>   * 		of SIGUSR2
>   *
> - * 	Setup:
> - * 	  Setup signal handling.
> - *	  Pause for SIGUSR1 if option specified.
> - *
> - * 	Test:
> - *	 Loop if the proper options are given.
> - *	 setup signal handler for SIGUSR2 signal
> - *	 fork a child
> - *
> - *	 CHILD:
> - *		setup signal handler for SIGUSR2 signal
> - *		call ptrace() with PTRACE_TRACEME request
> - *		send SIGUSR2 signal to self
> - *	 PARENT:
> - *		wait() for child.
> - *		if parent is notified when child gets a signal through wait(),
> - *		then
> - *			do  ptrace(PTRACE_KILL, ..) on child
> - * 			wait() for child to finish,
> - * 			if child exited abnormaly,
> - *				TEST passed
> - * 			else
> - * 				TEST failed
> - *		else
> - *			TEST failed
> - *
> - * 	Cleanup:
> - * 	  Print errno log and/or timing stats if options given
> - *
> - * USAGE:  <for command-line>
> - *  ptrace01 [-c n] [-e] [-i n] [-I x] [-P x] [-t] [-h] [-f] [-p]
> - *			where,  -c n : Run n copies concurrently.
> - *				-e   : Turn on errno logging.
> - *				-h   : Show help screen
> - *				-f   : Turn off functional testing
> - *				-i n : Execute test n times.
> - *				-I x : Execute test for x seconds.
> - *				-p   : Pause for SIGUSR1 before starting
> - *				-P x : Pause for x seconds between iterations.
> - *				-t   : Turn on syscall timing.
> - *
> - ****************************************************************/
> + */
>  
> +#include <stdlib.h>
>  #include <errno.h>
>  #include <signal.h>
>  #include <sys/wait.h>
> -
>  #include <config.h>
>  #include "ptrace.h"
> +#include "tst_test.h"
>  
> -#include "test.h"
> -
> -static void do_child(void);
> -static void setup(void);
> -static void cleanup(void);
> -static void child_handler();
> -static void parent_handler();
> +static int got_signal;

Variables changed from signal handlers has to be declared volatile as
well, otherwise the compiler may mis-optimize access to such variables.

> -static int got_signal = 0;
> -
> -char *TCID = "ptrace01";
> -static int i;			/* loop test case counter, shared with do_child */
> -
> -int TST_TOTAL = 2;
> -
> -int main(int ac, char **av)
> +void child_handler(void)
>  {
> -
> -	int lc;
> -	pid_t child_pid;
> -	int status;
> -	struct sigaction parent_act;
> -
> -	tst_parse_opts(ac, av, NULL, NULL);
> -#ifdef UCLINUX
> -	maybe_run_child(&do_child, "d", &i);
> -#endif
> -
> -	setup();
> -
> -	for (lc = 0; TEST_LOOPING(lc); lc++) {
> -
> -		tst_count = 0;
> -
> -		for (i = 0; i < TST_TOTAL; ++i) {
> -			got_signal = 0;
> -
> -			/* Setup signal handler for parent */
> -			if (i == 1) {
> -				parent_act.sa_handler = parent_handler;
> -				parent_act.sa_flags = SA_RESTART;
> -				sigemptyset(&parent_act.sa_mask);
> -
> -				if ((sigaction(SIGUSR2, &parent_act, NULL))
> -				    == -1) {
> -					tst_resm(TWARN, "sigaction() failed"
> -						 " in parent");
> -					continue;
> -				}
> -			}
> -
> -			switch (child_pid = FORK_OR_VFORK()) {
> -
> -			case -1:
> -				/* fork() failed */
> -				tst_resm(TFAIL, "fork() failed");
> -				continue;
> -
> -			case 0:
> -				/* Child */
> -#ifdef UCLINUX
> -				if (self_exec(av[0], "d", i) < 0) {
> -					tst_resm(TFAIL, "self_exec failed");
> -					continue;
> -				}
> -#else
> -				do_child();
> -#endif
> -
> -			default:
> -				/* Parent */
> -				if ((waitpid(child_pid, &status, 0)) < 0) {
> -					tst_resm(TFAIL, "waitpid() failed");
> -					continue;
> -				}
> -
> -				/*
> -				 * Check the exit status of child. If (it exits
> -				 * normally with exit value 1) OR (child came
> -				 * through signal handler), Test Failed
> -				 */
> -
> -				if (((WIFEXITED(status)) &&
> -				     (WEXITSTATUS(status))) ||
> -				    (got_signal == 1)) {
> -					tst_resm(TFAIL, "Test Failed");
> -					continue;
> -				} else {
> -					/* Kill child */
> -					if ((ptrace(PTRACE_KILL, child_pid,
> -						    0, 0)) == -1) {
> -						tst_resm(TFAIL, "Test Failed:"
> -							 " Parent was not able to kill"
> -							 " child");
> -						continue;
> -					}
> -				}
> -
> -				if ((waitpid(child_pid, &status, 0)) < 0) {
> -					tst_resm(TFAIL, "waitpid() failed");
> -					continue;
> -				}
> -
> -				if (WIFEXITED(status)) {
> -					/* Child exits normally */
> -					tst_resm(TFAIL, "Test failed");
> -				} else {
> -					tst_resm(TPASS, "Test Passed");
> -				}
> -
> -			}
> -		}
> +	if ((kill(getppid(), SIGUSR2)) == -1) {
> +		tst_res(TWARN, "kill() failed in child_handler()");
> +		exit(1);
>  	}
> +}

Can we please use SAFE_MACROS() here?

E.g. SAFE_SIGACTION(), SAFE_KILL(), ...

Since with these we don't have to check the return value, which
simplifies the code greatly.

> -	/* cleanup and exit */
> -	cleanup();
> -	tst_exit();
> -
> +void parent_handler(void)
> +{
> +	got_signal = 1;
>  }

Can we please change all the functions to static as well? I.e. the
child_handler, parent_handler as well as do_child?

> -/* do_child() */
> -void do_child(void)
> +void do_child(unsigned int i)
>  {
>  	struct sigaction child_act;
>  
> -	/* Setup signal handler for child */
> -	if (i == 0) {
> +	if (i == 0)
>  		child_act.sa_handler = SIG_IGN;
> -	} else {
> +	else
>  		child_act.sa_handler = child_handler;
> -	}
> +
>  	child_act.sa_flags = SA_RESTART;
>  	sigemptyset(&child_act.sa_mask);
>  
>  	if ((sigaction(SIGUSR2, &child_act, NULL)) == -1) {
> -		tst_resm(TWARN, "sigaction() failed in child");
> +		tst_res(TWARN, "sigaction() failed in child");
>  		exit(1);
>  	}
>  
>  	if ((ptrace(PTRACE_TRACEME, 0, 0, 0)) == -1) {
> -		tst_resm(TWARN, "ptrace() failed in child");
> +		tst_res(TWARN, "ptrace() failed in child");
>  		exit(1);
>  	}
> -	/* ensure that child bypasses signal handler */
>  	if ((kill(getpid(), SIGUSR2)) == -1) {
> -		tst_resm(TWARN, "kill() failed in child");
> +		tst_res(TWARN, "kill() failed in child");
>  		exit(1);
>  	}
>  	exit(1);
>  }
>  
> -/* setup() - performs all ONE TIME setup for this test */
> -void setup(void)
> +static void run(unsigned int i)
>  {
>  
> -	tst_sig(FORK, DEF_HANDLER, cleanup);
> +	pid_t child_pid;
> +	int status;
> +	struct sigaction parent_act;
>  
> -	TEST_PAUSE;
> +#ifdef UCLINUX
> +	maybe_run_child(&do_child, "d", &i);
> +#endif

We are staging out the uClinux support, feel free to delete everyting
between #ifdef UCLINUX and #endif.

> -}
> +	got_signal = 0;
>  
> -/*
> - *cleanup() -  performs all ONE TIME cleanup for this test at
> - *		completion or premature exit.
> - */
> -void cleanup(void)
> -{
> +	if (i == 1) {
> +		parent_act.sa_handler = parent_handler;
> +		parent_act.sa_flags = SA_RESTART;
> +		sigemptyset(&parent_act.sa_mask);
>  
> -}
> +		if ((sigaction(SIGUSR2, &parent_act, NULL))
> +		    == -1) {

I guess that the original code was over 80 lines which resulted into
such ugly line wrapping. Can we please keep conditions on a single line
unless they are over 80 lines?

> +			tst_res(TWARN, "sigaction() failed in parent");
> +		}
> +	}
>  
> -/*
> - * child_handler() - Signal handler for child
> - */
> -void child_handler(void)
> -{
> +	child_pid = SAFE_FORK();
>  
> -	if ((kill(getppid(), SIGUSR2)) == -1) {
> -		tst_resm(TWARN, "kill() failed in child_handler()");
> -		exit(1);
> -	}
> -}
> +	if (child_pid != 0) {
>  
> -/*
> - * parent_handler() - Signal handler for parent
> - */
> -void parent_handler(void)
> -{
> +		SAFE_WAITPID(child_pid, &status, 0);
>  
> -	got_signal = 1;
> +		if (((WIFEXITED(status)) &&
> +		     (WEXITSTATUS(status))) ||
> +		    (got_signal == 1)) {
> +			tst_res(TFAIL, "Test Failed");
> +		} else {
> +			if ((ptrace(PTRACE_KILL, child_pid,
> +				    0, 0)) == -1) {
> +				tst_res(TFAIL,
> +						"Test Failed: Parent was not able to kill child");
> +			}
> +		}

Can with print errno here if the ptrace has failed? Hint: Use the TERRNO
flag for tst_res().

Also the message could be shorter, we don't have to include the
"Test Failed" in it since the test library already prints that because
of the TFAIL flag, and we may as well be shorter and to the point with
something as "ptrace(PTRACE_KILL, child_pid, ...) failed", which carries
the same information.

> +		SAFE_WAITPID(child_pid, &status, 0);
> +
> +		if (WIFEXITED(status))
> +			tst_res(TFAIL, "Test failed");
> +		else
> +			tst_res(TPASS, "Test Passed");

We can do better in these messages, "Test failed" is nearly meaningless
since that information is carried in the TFAIL flag already.

Also we should be more pedantic and pass the test only and only if it
was killed. Looking at manual pages the PTRACE_KILL should send SIGKILL
signal to the process being traced, so we should pass the test only and
only if it was killed. Also we do have a tst_strstatus() in the test
library, so that we can print what happened to the child process in case
that it wasn't killed by SIGKILL.

> +	} else
> +		do_child(i);

This part would be more elegant as:


	if (!child_pid) {
		do_child();
		exit(1);
	}

	SAFE_WAITPID(...);

	...

>  }
> +
> +static struct tst_test test = {
> +	.test = run,
> +	.tcnt = 2,
> +	.forks_child = 1,
> +};
> -- 
> 2.16.4
> 
> 
> -- 
> Mailing list info: https://lists.linux.it/listinfo/ltp

-- 
Cyril Hrubis
chrubis@suse.cz

      parent reply	other threads:[~2019-10-08 10:47 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-10-01  7:01 [LTP] [PATCH] Convert ptrace01.c to new library Jorik Cronenberg
2019-10-01 20:14 ` Tim.Bird
2019-10-03 10:39   ` Cyril Hrubis
2019-10-08 10:47 ` Cyril Hrubis [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20191008104723.GA9170@rei.lan \
    --to=chrubis@suse.cz \
    --cc=ltp@lists.linux.it \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox