open_posix_testsuite: Don't use clock_settime() to reset a CPU-time Clock

This case is trying to set CPU-time clock(ID is returned by clock_getcpuclockid()) of a specific process by calling clock_settime(). But Linux kernel doesn't allow to reset a CPU-time clock of a specific process. Kernel returns -EPERM. So test fails.

In "IEEE Std 1003.1, 2004 Edition", It's said as following:
If _POSIX_CPUTIME is defined, implementations shall support clock ID values obtained by invoking clock_getcpuclockid(), which represent the CPU-time clock of a given process. Implementations shall also support the special clockid_t value CLOCK_PROCESS_CPUTIME_ID, which represents the CPU-time clock of the calling process when invoking one of the clock_*() or timer_*() functions. For these clock IDs, the values returned by clock_gettime() and specified by clock_settime() represent the amount of execution time of the process associated with the clock. Changing the value of a CPU-time clock via clock_settime() shall have no effect on the behavior of the sporadic server scheduling policy.

So the behaviour when changing the value of a CPU-time clock via clock_settime() is operating system specific.

We may change the test progress to stop using clock_settime(). Now, We just make use of clock_gettime() instead to achieve the same test purpose.

Signed-off-by Gui Jianfeng <[email protected]>

---
 .../interfaces/clock_getcpuclockid/2-1.c           |   67 ++++++++++----------
 1 files changed, 33 insertions(+), 34 deletions(-)

testcases/open_posix_testsuite/conformance/interfaces/clock_getcpuclockid/2-1.c +25 -34

index edeb397..52240df 100644
@@ -23,48 +23,47 @@ int main(int argc, char *argv[])
         printf("_POSIX_CPUTIME unsupported\n");
         return PTS_UNSUPPORTED;
 #else
-	unsigned long time_to_set;	
 	clockid_t clockid_1, clockid_2;
 	struct timespec tp1, tp2;
+	int i;
 
 	if (sysconf(_SC_CPUTIME) == -1) {
 		printf("_POSIX_CPUTIME unsupported\n");
 		return PTS_UNSUPPORTED;
 	}
 
-	if (clock_getcpuclockid(getpid(), &clockid_1) != 0) {
-		printf("clock_getcpuclockid(getpid(), ) failed\n");
-		return PTS_FAIL;
-	}
-	
-	if (clock_getcpuclockid(0, &clockid_2) != 0) {
-		printf("clock_getcpuclockid(0, ) failed\n");
-		return PTS_FAIL;
-	}
+        for (i = 0; i < 10; i++) {
+                if (clock_getcpuclockid(getpid(), &clockid_1) != 0) {
+                        printf("clock_getcpuclockid(getpid(), ) failed\n");
+                        return PTS_FAIL;
+                }
 
-	/* Set clockid_1 as a random value from 1 sec to 10 sec */
-	srand((unsigned long)time(NULL));
-	time_to_set = rand() * 10.0 / RAND_MAX + 1;
-	tp1.tv_sec = time_to_set;
-	tp1.tv_nsec = 0;	 
-	if (clock_settime(clockid_1, &tp1) != 0) {
-		printf("clock_getcpuclockid() returned an invalid clockid_t: "
-			"%d\n", clockid_1);
-		return PTS_FAIL;
-	}
-	/* Get the time of clockid_2, should almost the same as clockid_1 */
-	if (clock_gettime(clockid_2, &tp2) != 0) {
-		printf("clock_getcpuclockid() returned an invalid clockid_t: "
-			"%d\n", clockid_2);
-		return PTS_FAIL;
-	}
-	if (tp1.tv_sec == tp2.tv_sec)	
-	{
-		printf("Test PASSED\n");
-		return PTS_PASS;
-	}
-	printf("Test FAILED\n");
-	return PTS_FAIL;
-	
+                if (clock_getcpuclockid(0, &clockid_2) != 0) {
+                        printf("clock_getcpuclockid(0, ) failed\n");
+                        return PTS_FAIL;
+                }
+
+                /* Get the time of clockid_1 */
+                if (clock_gettime(clockid_1, &tp1) != 0) {
+                        printf("clock_getcpuclockid() returned an invalid clockid_t: "
+                               "%d\n", clockid_1);
+                        return PTS_FAIL;
+                }
+
+                /* Get the time of clockid_2, should almost the same as clockid_1 */
+                if (clock_gettime(clockid_2, &tp2) != 0) {
+                        printf("clock_getcpuclockid() returned an invalid clockid_t: "
+                               "%d\n", clockid_2);
+                        return PTS_FAIL;
+                }
+

Quoted 18 lines · Gui Jianfeng

This case is trying to set CPU-time clock(ID is returned by clock_getcpuclockid()) of a specific process by calling clock_settime(). But Linux kernel doesn't allow to reset a CPU-time clock of a specific process. Kernel returns -EPERM. So test fails.

In "IEEE Std 1003.1, 2004 Edition", It's said as following:
If _POSIX_CPUTIME is defined, implementations shall support clock ID values obtained by invoking clock_getcpuclockid(), which represent the CPU-time clock of a given process. Implementations shall also support the special clockid_t value CLOCK_PROCESS_CPUTIME_ID, which represents the CPU-time clock of the calling process when invoking one of the clock_*() or timer_*() functions. For these clock IDs, the values returned by clock_gettime() and specified by clock_settime() represent the amount of execution time of the process associated with the clock. Changing the value of a CPU-time clock via clock_settime() shall have no effect on the behavior of the sporadic server scheduling policy.

So the behaviour when changing the value of a CPU-time clock via clock_settime() is operating system specific.

Quoted 4 lines

We may change the test progress to stop using clock_settime(). Now, We just make use of clock_gettime() instead to achieve the same test purpose.

Signed-off-by Gui Jianfeng <[email protected]>

Hi Garrett,

How do you think this fix?

Quoted 19 lines · Gui Jianfeng

Quoted 18 lines · Gui Jianfeng

This case is trying to set CPU-time clock(ID is returned by clock_getcpuclockid()) of a specific process by calling clock_settime(). But Linux kernel doesn't allow to reset a CPU-time clock of a specific process. Kernel returns -EPERM. So test fails.

In "IEEE Std 1003.1, 2004 Edition", It's said as following:
If _POSIX_CPUTIME is defined, implementations shall support clock ID values obtained by invoking clock_getcpuclockid(), which represent the CPU-time clock of a given process. Implementations shall also support the special clockid_t value CLOCK_PROCESS_CPUTIME_ID, which represents the CPU-time clock of the calling process when invoking one of the clock_*() or timer_*() functions. For these clock IDs, the values returned by clock_gettime() and specified by clock_settime() represent the amount of execution time of the process associated with the clock. Changing the value of a CPU-time clock via clock_settime() shall have no effect on the behavior of the sporadic server scheduling policy.

So the behaviour when changing the value of a CPU-time clock via clock_settime() is operating system specific.

Quoted 7 lines

Quoted 4 lines

We may change the test progress to stop using clock_settime(). Now, We just make use of clock_gettime() instead to achieve the same test purpose.

Signed-off-by Gui Jianfeng <[email protected]>

Hi Garrett,

How do you think this fix?

The tort is the same in POSIX 2008.1. Taking out the chaff...

a. If _POSIX_CPUTIME is defined, implementations shall support clock ID values obtained by invoking clock_getcpuclockid(), which represent the CPU-time clock of a given process.
b. Implementations shall also support the special clockid_t value CLOCK_PROCESS_CPUTIME_ID, which represents the CPU-time clock of the calling process when invoking one of the clock_*() or timer_*() functions.
c. For these clock IDs, the values returned by clock_gettime() and specified by clock_settime() represent the amount of execution time of the process associated with the clock.

Nowhere does it say that you can't set that clock. Try running the test as root to see what you can do... if the test fails (even as root), then Linux needs to be fixed to conform to POSIX.

Also, try running the other tests I just added to see whether or not the function call fails per POSIX's requirements.

Thanks!
-Garrett

Quoted 40 lines · Garrett Cooper

Quoted 19 lines · Gui Jianfeng

Quoted 18 lines · Gui Jianfeng

This case is trying to set CPU-time clock(ID is returned by clock_getcpuclockid()) of a specific process by calling clock_settime(). But Linux kernel doesn't allow to reset a CPU-time clock of a specific process. Kernel returns -EPERM. So test fails.

In "IEEE Std 1003.1, 2004 Edition", It's said as following:
If _POSIX_CPUTIME is defined, implementations shall support clock ID values obtained by invoking clock_getcpuclockid(), which represent the CPU-time clock of a given process. Implementations shall also support the special clockid_t value CLOCK_PROCESS_CPUTIME_ID, which represents the CPU-time clock of the calling process when invoking one of the clock_*() or timer_*() functions. For these clock IDs, the values returned by clock_gettime() and specified by clock_settime() represent the amount of execution time of the process associated with the clock. Changing the value of a CPU-time clock via clock_settime() shall have no effect on the behavior of the sporadic server scheduling policy.

So the behaviour when changing the value of a CPU-time clock via clock_settime() is operating system specific.

Quoted 7 lines

Quoted 4 lines

We may change the test progress to stop using clock_settime(). Now, We just make use of clock_gettime() instead to achieve the same test purpose.

Signed-off-by Gui Jianfeng <[email protected]>

Hi Garrett,

How do you think this fix?

The tort is the same in POSIX 2008.1. Taking out the chaff...

a. If _POSIX_CPUTIME is defined, implementations shall support clock ID values obtained by invoking clock_getcpuclockid(), which represent the CPU-time clock of a given process.
b. Implementations shall also support the special clockid_t value CLOCK_PROCESS_CPUTIME_ID, which represents the CPU-time clock of the calling process when invoking one of the clock_*() or timer_*() functions.
c. For these clock IDs, the values returned by clock_gettime() and specified by clock_settime() represent the amount of execution time of the process associated with the clock.

Quoted 3 lines

Nowhere does it say that you can't set that clock. Try running the test as root to see what you can do... if the test fails (even as root), then Linux needs to be fixed to conform to POSIX.

POSIX also says that:
Changing the value of a CPU-time clock via clock_settime() shall have no effect on the behavior of the sporadic server scheduling policy.

IMHO, this means clock_settime() implementation is OS specific, is it? So, even if Linux fails to set this clock, it's just fine. Am i missing something?

Gui

Quoted 6 lines

Also, try running the other tests I just added to see whether or not the function call fails per POSIX's requirements.

Thanks!
-Garrett

------------------------------------------------------------------------------
The Next 800 Companies to Lead America's Growth: New Video Whitepaper David G. Thomson, author of the best-selling book "Blueprint to a Billion" shares his insights and actions to help propel your business during the next growth cycle. Listen Now!
http://p.sf.net/sfu/SAP-dev2dev
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

The tort is the same in POSIX 2008.1. Taking out the chaff...

a. If _POSIX_CPUTIME is defined, implementations shall support clock ID values obtained by invoking clock_getcpuclockid(), which represent the CPU-time clock of a given process.
b. Implementations shall also support the special clockid_t value CLOCK_PROCESS_CPUTIME_ID, which represents the CPU-time clock of the calling process when invoking one of the clock_*() or timer_*() functions.
c. For these clock IDs, the values returned by clock_gettime() and specified by clock_settime() represent the amount of execution time of the process associated with the clock.

Nowhere does it say that you can't set that clock. Try running the test as root to see what you can do... if the test fails (even as root), then Linux needs to be fixed to conform to POSIX.

Also, try running the other tests I just added to see whether or not the function call fails per POSIX's requirements.

Quoted 51 lines

Quoted 51 lines

Quoted 41 lines

Quoted 13 lines

The tort is the same in POSIX 2008.1. Taking out the chaff...

a. If _POSIX_CPUTIME is defined, implementations shall support clock ID values obtained by invoking clock_getcpuclockid(), which represent the CPU-time clock of a given process.
b. Implementations shall also support the special clockid_t value CLOCK_PROCESS_CPUTIME_ID, which represents the CPU-time clock of the calling process when invoking one of the clock_*() or timer_*() functions.
c. For these clock IDs, the values returned by clock_gettime() and specified by clock_settime() represent the amount of execution time of the process associated with the clock.

Quoted 7 lines

Nowhere does it say that you can't set that clock. Try running the test as root to see what you can do... if the test fails (even as root), then Linux needs to be fixed to conform to POSIX.

Also, try running the other tests I just added to see whether or not the function call fails per POSIX's requirements.

From clock_settime POSIX documentation:

...

The effect of setting a clock via clock_settime() on armed per-process timers associated with that clock is implementation-dependent.

The appropriate privilege to set a particular clock is implementation-dependent.

...

From man clock_settime:

...

According to POSIX.1-2001, a process with "appropriate privileges" may set the CLOCK_PROCESS_CPUTIME_ID and CLOCK_THREAD_CPUTIME_ID clocks using clock_settime(). On Linux, these clocks are not settable (i.e., no process has "appropriate privileges").

Quoted 6 lines

...

This is pretty creative understanding of the standart, but I couldn't say it's wrong anyway.

I agree that privilege is multitiered and the implications of running as a particular user, whether or not SELinux is enabled, MAC allows it, etc are all bits to take into consideration.

Quoted 8 lines

Quoted 2 lines

Just running the program as root seems like a simple test to run to see whether or not the testcase passes.

FYI, I removed the clock_settime call and simplified the testcase
because it wasn't really buying much in the overall scheme of things to make sure that the time matched on the two clock values. If anything that should be tested with the clock_settime conformance testcase.

Hi Cooper,

I have a look of the code you modified. Consider the following piece:

         /* Get the time of clockid_1. */
         if (clockid_1 != clockid_2) {
                 printf("clock_getcpuclockid(0, ..) != "
                   "clock_getcpuclockid(%d, ..): (%d != %d)\n", clockid_1,
                    clockid_2);
                return PTS_FAIL;
         }

IMO, you can't do such comparision, it's problematic. In this case, clockid_1 and clockid_2 own different values, but they point to the same clock. So you can't modify like this.

Yes this assumption seems to be incorrect as I couldn't find in a place in POSIX documentation which explicitly states that two obtained clock_id that points to the same clock must be the same. On linux calling clock_getcpuclockid() with 0 and value from getpid() returns two different id which both points to the same clock.

This case is trying to set CPU-time clock(ID is returned by clock_getcpuclockid()) of a specific process by calling clock_settime(). But Linux kernel doesn't allow to reset a CPU-time clock of a specific process. Kernel returns -EPERM. So test fails.

In "IEEE Std 1003.1, 2004 Edition", It's said as following:
If _POSIX_CPUTIME is defined, implementations shall support clock ID values obtained by invoking clock_getcpuclockid(), which represent the CPU-time clock of a given process. Implementations shall also support the special clockid_t value CLOCK_PROCESS_CPUTIME_ID, which represents the CPU-time clock of the calling process when invoking one of the clock_*() or timer_*() functions. For these clock IDs, the values returned by clock_gettime() and specified by clock_settime() represent the amount of execution time of the process associated with the clock. Changing the value of a CPU-time clock via clock_settime() shall have no effect on the behavior of the sporadic server scheduling policy.

So the behaviour when changing the value of a CPU-time clock via clock_settime() is operating system specific.

We may change the test progress to stop using clock_settime(). Now, We just make use of clock_gettime() instead to achieve the same test purpose.

Signed-off-by Gui Jianfeng <[email protected]>

---
 .../interfaces/clock_getcpuclockid/2-1.c           |   67 ++++++++++----------
 1 files changed, 33 insertions(+), 34 deletions(-)

testcases/open_posix_testsuite/conformance/interfaces/clock_getcpuclockid/2-1.c +25 -34

index edeb397..52240df 100644
@@ -23,48 +23,47 @@ int main(int argc, char *argv[])
         printf("_POSIX_CPUTIME unsupported\n");
         return PTS_UNSUPPORTED;
 #else
-	unsigned long time_to_set;	
 	clockid_t clockid_1, clockid_2;
 	struct timespec tp1, tp2;
+	int i;
 
 	if (sysconf(_SC_CPUTIME) == -1) {
 		printf("_POSIX_CPUTIME unsupported\n");
 		return PTS_UNSUPPORTED;
 	}
 
-	if (clock_getcpuclockid(getpid(), &clockid_1) != 0) {
-		printf("clock_getcpuclockid(getpid(), ) failed\n");
-		return PTS_FAIL;
-	}
-	
-	if (clock_getcpuclockid(0, &clockid_2) != 0) {
-		printf("clock_getcpuclockid(0, ) failed\n");
-		return PTS_FAIL;
-	}
+        for (i = 0; i < 10; i++) {
+                if (clock_getcpuclockid(getpid(), &clockid_1) != 0) {
+                        printf("clock_getcpuclockid(getpid(), ) failed\n");
+                        return PTS_FAIL;
+                }
 
-	/* Set clockid_1 as a random value from 1 sec to 10 sec */
-	srand((unsigned long)time(NULL));
-	time_to_set = rand() * 10.0 / RAND_MAX + 1;
-	tp1.tv_sec = time_to_set;
-	tp1.tv_nsec = 0;	 
-	if (clock_settime(clockid_1, &tp1) != 0) {
-		printf("clock_getcpuclockid() returned an invalid clockid_t: "
-			"%d\n", clockid_1);
-		return PTS_FAIL;
-	}
-	/* Get the time of clockid_2, should almost the same as clockid_1 */
-	if (clock_gettime(clockid_2, &tp2) != 0) {
-		printf("clock_getcpuclockid() returned an invalid clockid_t: "
-			"%d\n", clockid_2);
-		return PTS_FAIL;
-	}
-	if (tp1.tv_sec == tp2.tv_sec)	
-	{
-		printf("Test PASSED\n");
-		return PTS_PASS;
-	}
-	printf("Test FAILED\n");
-	return PTS_FAIL;
-	
+                if (clock_getcpuclockid(0, &clockid_2) != 0) {
+                        printf("clock_getcpuclockid(0, ) failed\n");
+                        return PTS_FAIL;
+                }
+
+                /* Get the time of clockid_1 */
+                if (clock_gettime(clockid_1, &tp1) != 0) {
+                        printf("clock_getcpuclockid() returned an invalid clockid_t: "
+                               "%d\n", clockid_1);
+                        return PTS_FAIL;
+                }
+
+                /* Get the time of clockid_2, should almost the same as clockid_1 */
+                if (clock_gettime(clockid_2, &tp2) != 0) {
+                        printf("clock_getcpuclockid() returned an invalid clockid_t: "
+                               "%d\n", clockid_2);
+                        return PTS_FAIL;
+                }
+