[LTP] [PATCH 2/3] syscalls/shmat0*: cleanup && convert to new API

Cyril Hrubis chrubis@suse.cz
Mon May 29 16:11:24 CEST 2017


Hi!
> --- a/testcases/kernel/syscalls/ipc/libnewipc/libnewipc.h
> +++ b/testcases/kernel/syscalls/ipc/libnewipc/libnewipc.h
> @@ -50,4 +50,7 @@ int get_used_queues(const char *file, const int lineno);
>  #define GET_USED_QUEUES() \
>  	get_used_queues(__FILE__, __LINE__)
>  
> +void *probe_free_addr(void);
> +#define PROBE_FREE_ADDR() probe_free_addr()

Defining macro with the same name as function is pointless if we are not
passing down __FILE__ and __LINE__ parameters as we do in SAFE_MACROS().

> diff --git a/testcases/kernel/syscalls/ipc/shmat/shmat01.c b/testcases/kernel/syscalls/ipc/shmat/shmat01.c
> index 6de1872..9b9390f 100644
> --- a/testcases/kernel/syscalls/ipc/shmat/shmat01.c
> +++ b/testcases/kernel/syscalls/ipc/shmat/shmat01.c
> @@ -17,225 +17,111 @@
>   */
>  
>  /*
> - * NAME
> - *	shmat01.c
> - *
>   * DESCRIPTION
> - *	shmat01 - test that shmat() works correctly
>   *
> - * ALGORITHM
> - *	create a shared memory resouce with read/write permissions
> - *	loop if that option was specified
> - *	call shmat() with the TEST() macro using three valid conditions
> - *	check the return code
> - *	  if failure, issue a FAIL message.
> - *	otherwise,
> - *	  if doing functionality testing
> - *		check for the correct conditions after the call
> - *		if correct,
> - *			issue a PASS message
> - *		otherwise
> - *			issue a FAIL message
> - *	call cleanup
> + * 1) shmat() chooses a suitable (unused) address when shmaddr is NULL.
> + * 2) shmat() attaches shm segment to the shmaddr when shmaddr is a
> + *    page-aligned address.
> + * 3) shmat() attaches shm segment to the address equal to shmaddr rounded
> + *    down to the nearest multiple of SHMLBA when shmaddr is not a
> + *    page-unaligned address and shmflg is set to SHM_RND.

"not a page-unaligned"  == "page-aligned" as far as I can tell

You should drop one of the negations and use either "not page-aligned"
or "page-unaligned".

> + * 4) shmat() attaches shm segment to the shmaddr for reading when shmflg
> + *    is set to SHM_RDONLY.
>   */
>  
> -#include "ipcshm.h"
> -#include "shmat_common.h"
> +#include <errno.h>
> +#include <sys/types.h>
> +#include <sys/ipc.h>
> +#include <sys/shm.h>
>  
> -#define CASE0		10
> -#define CASE1		20
> +#include "tst_test.h"
> +#include "tst_safe_sysv_ipc.h"
> +#include "libnewipc.h"
>  
> -char *TCID = "shmat01";
> -int TST_TOTAL = 4;
> +#define CONTENT	10
>  
> -int shm_id_1 = -1;
> -
> -/*
> - * By probing this address first, we can make
> - * non-aligned addresses from it for different
> - * architectures without explicitly code it.
> - */
> -void *base_addr;
> -void *addr;
> +static int shm_id = -1;
> +static key_t shm_key;
> +static void *null_addr;
> +static void *aligned_addr;
> +static void *unaligned_addr;
>  
>  static struct test_case_t {
> -	int *shmid;
> -	int offset;
> -	int flags;
> -	int getbase;
> -} *TC;
> -
> -static void check_functionality(int);
> -
> -int main(int argc, char *argv[])
> +	void **shmaddr;
> +	int flag;
> +} tcases[] = {
> +	{&null_addr, 0},
> +	{&aligned_addr, 0},
> +	{&unaligned_addr, SHM_RND},
> +	{&aligned_addr, SHM_RDONLY}
> +};
> +
> +static void verify_shmat(unsigned int n)
>  {
> -	int lc, i;
> -	void *attchaddr;
> -
> -	tst_parse_opts(argc, argv, NULL, NULL);
> -
> -	setup();
> -
> -	for (lc = 0; TEST_LOOPING(lc); lc++) {
> -
> -		tst_count = 0;
> -
> -		for (i = 0; i < TST_TOTAL; i++) {
> -
> -			if (TC[i].getbase) {
> -				base_addr = probe_free_addr();
> -				attchaddr = base_addr + TC[i].offset;
> -			} else {
> -				attchaddr = NULL;
> -			}
> -
> -			addr = shmat(*(TC[i].shmid), attchaddr, TC[i].flags);
> +	int *addr;
> +	struct shmid_ds buf;
>  
> -			TEST_ERRNO = errno;
> -			if (addr == (void *)-1) {
> -				tst_brkm(TFAIL | TTERRNO, cleanup,
> -					 "shmat call failed");
> -			} else {
> -				check_functionality(i);
> -			}
> +	struct test_case_t *tc = &tcases[n];
>  
> -			if (shmdt(addr) == -1)
> -				tst_brkm(TBROK, cleanup,
> -					 "Couldn't detach shared memory");
> -		}
> +	addr = shmat(shm_id, *tc->shmaddr, tc->flag);
> +	TEST_ERRNO = errno;

Just use TERRNO below instead of TTERRNO.

> +	if (addr == (void *)-1) {
> +		tst_res(TFAIL | TTERRNO, "shmat() failed");
> +		return;
>  	}
>  
> -	cleanup();
> -
> -	tst_exit();
> -}
> +	SAFE_SHMCTL(shm_id, IPC_STAT, &buf);
>  
> -/*
> - * check_functionality - check various conditions to make sure they
> - *			 are correct.
> - */
> -static void check_functionality(int i)
> -{
> -	void *orig_add;
> -	int *shared;
> -	int fail = 0;
> -	struct shmid_ds buf;
> -
> -	shared = (int *)addr;
> -
> -	/* stat the shared memory ID */
> -	if (shmctl(shm_id_1, IPC_STAT, &buf) == -1)
> -		tst_brkm(TBROK, cleanup, "couldn't stat shared memory");
> -
> -	/* check the number of attaches */
>  	if (buf.shm_nattch != 1) {
> -		tst_resm(TFAIL, "# of attaches is incorrect");
> +		tst_res(TFAIL, "number of attaches was incorrect");
>  		return;
>  	}
>  
> -	/* check the size of the segment */
>  	if (buf.shm_segsz != INT_SIZE) {
> -		tst_resm(TFAIL, "segment size is incorrect");
> +		tst_res(TFAIL, "segment size was incorrect");
>  		return;
>  	}
>  
> -	/* check for specific conditions depending on the type of attach */
> -	switch (i) {
> -	case 0:
> -	case 1:
> -		/*
> -		 * Check the functionality of shmat by simply "writing"
> -		 * a value to the shared memory space.
> -		 * If this fails the program will get a SIGSEGV, dump
> -		 * core and exit.
> -		 */
> -
> -		*shared = CASE0;
> -		break;
> -	case 2:
> -		/*
> -		 * Check the functionality of shmat by writing a value
> -		 * to the shared memory space and then checking that
> -		 * the original address given was rounded down as
> -		 * specified in the man page.
> -		 */
> -
> -		*shared = CASE1;
> -		orig_add = addr + ((unsigned long)TC[2].offset % SHMLBA);
> -		if (orig_add != base_addr + TC[2].offset) {
> -			tst_resm(TFAIL, "shared memory address is not "
> -				 "correct");
> -			fail = 1;
> -		}
> -		break;
> -	case 3:
> -		/*
> -		 * This time the shared memory is read only.  Read the value
> -		 * and check that it is equal to the value set in last case,
> -		 * because shared memory is persistent.
> -		 */
> -
> -		if (*shared != CASE1) {
> -			tst_resm(TFAIL, "shared memory value isn't correct");
> -			fail = 1;
> +	if ((tc->flag & SHM_RND) && addr != aligned_addr) {
> +		tst_res(TFAIL, "shared memory address was not correct");
> +		return;
> +	}

We do check here that the address was aligned in a case that we passed
unaligned one, but we do not check that the aligned addres we supplied
was used.

What about introducing a function to check the address, something as:

#define ALIGN_DOWN ((void*)(((intptr_t)in_addr / SHMLBA) * SHMLBA))

void *expected_addr(void *in_addr, void *out_addr)
{
	if (!in_addr)
		return out_addr;

	return ALIGN_DOWN(in_addr);
}

....
	if (expected_addr(*tc->addr, addr) != addr) {
		tst_res(TFAIL, "shmat() returned %p expected %p",
			addr, expected_addr(*tc->addr, addr));
		return;
	}
...


> +	if (tc->flag & SHM_RDONLY) {
> +		if (*addr != CONTENT) {
> +			tst_res(TFAIL, "shared memory value wasn't correct");
> +			return;
>  		}
> -		break;
> +	} else {
> +		*addr = CONTENT;
>  	}


Hmm, if we really wanted to test the RDONLY case we would run this part
of the test in a child, tried to access the address and parent would
expect that it gets killed with SIGSEGV.

Or we may setup a signal handler for SIGSEGV and check that flag was set
after we tried to access read only memory.

> -	if (!fail)
> -		tst_resm(TPASS, "conditions and functionality are correct");
> -}
> +	tst_res(TPASS, "shmat() succeeded");
>  
> -void setup(void)
> -{
> -	tst_sig(NOFORK, DEF_HANDLER, cleanup);
> -
> -	TEST_PAUSE;
> -
> -	TC = malloc(TST_TOTAL * sizeof(struct test_case_t));
> -	if (TC == NULL)
> -		tst_brkm(TFAIL | TERRNO, cleanup, "failed to allocate memory");
> -
> -	/* set NULL as attaching address*/
> -	TC[0].shmid = &shm_id_1;
> -	TC[0].offset = 0;
> -	TC[0].flags = 0;
> -	TC[0].getbase = 0;
> -
> -	/* a straight forward read/write attach */
> -	TC[1].shmid = &shm_id_1;
> -	TC[1].offset = 0;
> -	TC[1].flags = 0;
> -	TC[1].getbase = 1;
> -
> -	/* an attach using unaligned memory */
> -	TC[2].shmid = &shm_id_1;
> -	TC[2].offset = SHMLBA - 1;
> -	TC[2].flags = SHM_RND;
> -	TC[2].getbase = 1;
> -
> -	/* a read only attach */
> -	TC[3].shmid = &shm_id_1;
> -	TC[3].offset = 0;
> -	TC[3].flags = SHM_RDONLY;
> -	TC[3].getbase = 1;
> -
> -	tst_tmpdir();
> -
> -	shmkey = getipckey();
> -
> -	shm_id_1 = shmget(shmkey++, INT_SIZE, SHM_RW | IPC_CREAT | IPC_EXCL);
> -	if (shm_id_1 == -1)
> -		tst_brkm(TBROK, cleanup, "Failed to create shared memory "
> -			 "resource 1 in setup()");
> +	SAFE_SHMDT(addr);
>  }
>  
> -void cleanup(void)
> +static void setup(void)
>  {
> -	rm_shm(shm_id_1);
> +	aligned_addr = PROBE_FREE_ADDR();
> +	unaligned_addr = aligned_addr + SHMLBA - 1;
>  
> -	if (TC != NULL)
> -		free(TC);
> +	shm_key = GETIPCKEY();
>  
> -	tst_rmdir();
> +	shm_id = SAFE_SHMGET(shm_key, INT_SIZE, SHM_RW | IPC_CREAT | IPC_EXCL);
>  }
> +
> +static void cleanup(void)
> +{
> +	if (shm_id != -1)
> +		SAFE_SHMCTL(shm_id, IPC_RMID, NULL);
> +}
> +
> +static struct tst_test test = {
> +	.tid = "shmat01",
> +	.needs_root = 1,
> +	.setup = setup,
> +	.cleanup = cleanup,
> +	.test = verify_shmat,
> +	.tcnt = ARRAY_SIZE(tcases)
> +};
> diff --git a/testcases/kernel/syscalls/ipc/shmat/shmat02.c b/testcases/kernel/syscalls/ipc/shmat/shmat02.c
> index d6355f7..4302043 100644
> --- a/testcases/kernel/syscalls/ipc/shmat/shmat02.c
> +++ b/testcases/kernel/syscalls/ipc/shmat/shmat02.c
> @@ -1,189 +1,133 @@
>  /*
> + * Copyright (c) International Business Machines  Corp., 2001
>   *
> - *   Copyright (c) International Business Machines  Corp., 2001
> + * This program is free software;  you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; either version 2 of the License, or
> + * (at your option) any later version.
>   *
> - *   This program is free software;  you can redistribute it and/or modify
> - *   it under the terms of the GNU General Public License as published by
> - *   the Free Software Foundation; either version 2 of the License, or
> - *   (at your option) any later version.
> + * This program 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 General Public License for more details.
>   *
> - *   This program 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 General Public License for more details.
> - *
> - *   You should have received a copy of the GNU General Public License
> - *   along with this program;  if not, write to the Free Software
> - *   Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
> + * You should have received a copy of the GNU General Public License
> + * along with this program.
>   */
>  
>  /*
> - * NAME
> - *	shmat02.c
> - *
>   * DESCRIPTION
> - *	shmat02 - check for EINVAL and EACCES errors
> - *
> - * ALGORITHM
> - *	loop if that option was specified
> - *	  call shmat() using three invalid test cases
> - *	  check the errno value
> - *	    issue a PASS message if we get EINVAL or EACCES
> - *	  otherwise, the tests fails
> - *	    issue a FAIL message
> - *	call cleanup
> - *
> - * USAGE:  <for command-line>
> - *  shmat02 [-c n] [-e] [-i n] [-I x] [-P x] [-t]
> - *     where,  -c n : Run n copies concurrently.
> - *             -e   : Turn on errno logging.
> - *	       -i n : Execute test n times.
> - *	       -I x : Execute test for x seconds.
> - *	       -P x : Pause for x seconds between iterations.
> - *	       -t   : Turn on syscall timing.
> - *
> - * HISTORY
> - *	03/2001 - Written by Wayne Boyer
>   *
> - *      27/02/2008 Renaud Lottiaux (Renaud.Lottiaux@kerlabs.com)
> - *      - Fix concurrency issue. The second key used for this test could
> - *        conflict with the key from another task.
> - *
> - * RESTRICTIONS
> - *	Must be ran as non-root
> + * 1) shmat() fails and set errno to EINVAL when shmid is invalid.
> + * 2) shmat() fails and set errno to EINVAL when shmaddr is not page
> + *    aligned and SHM_RND is not given
> + * 3) shmat() fails and set errno to EACCES when the shm resource has
> + *    no read/write permission.
> + * 4) shmat() fails and set errno to EACCES when non-root user access
> + *    shm created by root.
>   */
>  
> -#include "ipcshm.h"
> +#include <errno.h>
> +#include <stdlib.h>
> +#include <sys/types.h>
> +#include <sys/ipc.h>
> +#include <sys/shm.h>
>  #include <pwd.h>
> -#include "shmat_common.h"
> -
> -char *TCID = "shmat02";
> -char nobody_uid[] = "nobody";
> -struct passwd *ltpuser;
>  
> -int shm_id_1 = -1;
> -int shm_id_2 = -1;
> -int shm_id_3 = -1;
> +#include "tst_test.h"
> +#include "tst_safe_sysv_ipc.h"
> +#include "libnewipc.h"
>  
> -void *base_addr;		/* By probing this address first, we can make
> -				 * non-aligned addresses from it for different
> -				 * architectures without explicitly code it.
> -				 */
> +static int shm_id1 = -1;
> +static int shm_id2 = -1;
> +static int shm_id3 = -1;
> +static void *aligned_addr;
> +static void *unaligned_addr;
> +static key_t shm_key1, shm_key2;
>  
> -void *addr;			/* for result of shmat-call */
> -
> -struct test_case_t {
> +static struct test_case_t {
>  	int *shmid;
> -	int offset;
> -	int error;
> +	void **shmaddr;
> +	int rw_flag;
> +	int exp_err;
> +	int exp_user;
> +} tcases[] = {
> +	{&shm_id1, &aligned_addr, 1, EINVAL, 0},
> +	{&shm_id2, &unaligned_addr, 1, EINVAL, 0},
> +	{&shm_id2, &aligned_addr, 1, EACCES, 1},
> +	{&shm_id3, &aligned_addr, 0, EACCES, 1}
>  };
>  
> -int TST_TOTAL = 3;
> -
> -static void setup_tc(int i, struct test_case_t *tc)
> +static void verify_shmat(struct test_case_t *tc)
>  {
> +	void *addr;
>  
> -	struct test_case_t TC[] = {
> -		/* EINVAL - the shared memory ID is not valid */
> -		{&shm_id_1, 0, EINVAL},
> -		/* EINVAL - the address is not page aligned and SHM_RND is not given */
> -		{&shm_id_2, SHMLBA - 1, EINVAL},
> -		/* EACCES - the shared memory resource has no read/write permission */
> -		{&shm_id_3, 0, EACCES}
> -	};
> +	if (!tc->rw_flag)
> +		shm_id3 = SAFE_SHMGET(shm_key2, INT_SIZE, IPC_CREAT | IPC_EXCL);

Is there a reason for not initializing shm_id3 in the setup once?

> -	if (i > TST_TOTAL || i < 0)
> +	addr = shmat(*tc->shmid, *tc->shmaddr, 0);
> +	TEST_ERRNO = errno;

Here as well, no need to save errno into TEST_ERRNO, you can just use
errno and TERRNO in the code below.

> +	if (addr != (void *)-1) {
> +		tst_res(TFAIL, "shmat() succeeded unexpectedly");
>  		return;
> +	}
> +
> +	if (TEST_ERRNO == tc->exp_err) {
> +		tst_res(TPASS | TTERRNO, "shmat() failed as expected");
> +	} else {
> +		tst_res(TFAIL | TTERRNO, "shmat() failed unexpectedly,"
> +			 "expected: %s", tst_strerrno(tc->exp_err));
> +	}
>  
> -	*tc = TC[i];
> +	if (shm_id3 != -1)
> +		SAFE_SHMCTL(shm_id3, IPC_RMID, NULL);
>  }
>  
> -int main(int ac, char **av)
> +static void do_shmat(unsigned int n)
>  {
> -	int lc;
> -	int i;
> -	struct test_case_t *tc;
> -
> -	tc = NULL;
> -
> -	tst_parse_opts(ac, av, NULL, NULL);
> -
> -	tc = malloc(sizeof(struct test_case_t));
> -	if (tc == NULL)
> -		tst_brkm(TBROK | TERRNO, cleanup, "malloc failed");
> -
> -	setup();
> -
> -	for (lc = 0; TEST_LOOPING(lc); lc++) {
> -		tst_count = 0;
> -
> -		for (i = 0; i < TST_TOTAL; i++) {
> -
> -			setup_tc(i, tc);
> -
> -			base_addr = probe_free_addr();
> -			errno = 0;
> -			addr = shmat(*(tc->shmid), base_addr + tc->offset, 0);
> -
> -			if (addr != (void *)-1) {
> -				tst_resm(TFAIL, "call succeeded unexpectedly");
> -				continue;
> -			}
> -
> -			if (errno == tc->error)
> -				tst_resm(TPASS | TERRNO,
> -					 "shmat failed as expected");
> -			else
> -				tst_resm(TFAIL,
> -					 "shmat failed unexpectedly; expected: "
> -					 "%d - %s", tc->error,
> -					 strerror(tc->error));
> +	pid_t pid;
> +	struct passwd *pw;
> +
> +	struct test_case_t *tc = &tcases[n];
> +
> +	if (tc->exp_user == 0) {
> +		verify_shmat(tc);
> +	} else {
> +		pid = SAFE_FORK();
> +		if (pid) {
> +			tst_reap_children();
> +		} else {
> +			pw = SAFE_GETPWNAM("nobody");

The getpwnam should be done once in the test setup.

> +			SAFE_SETUID(pw->pw_uid);
> +			verify_shmat(tc);
> +			exit(0);
>  		}
>  	}
> -
> -	cleanup();
> -
> -	tst_exit();
>  }
>  
> -void setup(void)
> +static void setup(void)
>  {
> -	key_t shmkey2;
> -
> -	tst_require_root();
> -	ltpuser = getpwnam(nobody_uid);
> -	if (ltpuser == NULL)
> -		tst_brkm(TBROK | TERRNO, NULL, "getpwnam failed");
> -	if (setuid(ltpuser->pw_uid) == -1)
> -		tst_brkm(TBROK | TERRNO, NULL, "setuid failed");
> -
> -	tst_sig(NOFORK, DEF_HANDLER, cleanup);
> +	aligned_addr = PROBE_FREE_ADDR();
> +	unaligned_addr = aligned_addr + SHMLBA - 1;
>  
> -	TEST_PAUSE;
> +	shm_key1 = GETIPCKEY();
> +	shm_key2 = GETIPCKEY();
>  
> -	tst_tmpdir();
> -
> -	shmkey = getipckey();
> -
> -	shm_id_2 = shmget(shmkey, INT_SIZE, SHM_RW | IPC_CREAT | IPC_EXCL);
> -	if (shm_id_2 == -1)
> -		tst_brkm(TBROK | TERRNO, cleanup, "shmget #1 failed");
> -
> -	/* Get an new IPC resource key. */
> -	shmkey2 = getipckey();
> -
> -	/* create a shared memory resource without read and write permissions */
> -	shm_id_3 = shmget(shmkey2, INT_SIZE, IPC_CREAT | IPC_EXCL);
> -	if (shm_id_3 == -1)
> -		tst_brkm(TBROK | TERRNO, cleanup, "shmget #2 failed");
> +	shm_id2 = SAFE_SHMGET(shm_key1, INT_SIZE, SHM_RW | IPC_CREAT | IPC_EXCL);
>  }
>  
> -void cleanup(void)
> +static void cleanup(void)
>  {
> -	/* if they exist, remove the shared memory resources */
> -	rm_shm(shm_id_2);
> -	rm_shm(shm_id_3);
> -
> -	tst_rmdir();
> -
> +	if (shm_id2 != -1)
> +		SAFE_SHMCTL(shm_id2, IPC_RMID, NULL);
>  }
> +
> +static struct tst_test test = {
> +	.tid = "shmat02",
> +	.needs_root = 1,
> +	.forks_child = 1,
> +	.test = do_shmat,
> +	.tcnt = ARRAY_SIZE(tcases),
> +	.setup = setup,
> +	.cleanup = cleanup
> +};

Otherwise it looks fine.

-- 
Cyril Hrubis
chrubis@suse.cz


More information about the ltp mailing list