George -
In the latest 2.5.54-bk1 high-res-timers patches, it appears that
nanosleep() is returning 0 (success) and not setting errno when an rqtp
argument is sent that specifies a nsec value < 0 or >= 1000 million. In
this instance, the POSIX System Interfaces doc states that errno is supposed
to be set to EINVAL, and nanosleep should return -1.
In the 2.5.50 high-res-timers patches, behavior was as expected (i.e.,
returned -1 and set errno=EINVAL). Unfortunately, I haven't looked at any
patches since then to know exactly which patch stopped behaving as expected.
A plain 2.5.54-bk1 kernel also behaves as expected (returns -1, sets
errno=EINVAL).
The tests I am using to reproduce this issue are part of the POSIX Test
Suite at http://posixtest.sf.net under
posixtestsuite/conformance/interfaces/nanosleep. 5-1.c (sending -1 nsec),
6-1.c (sending multiple nsec values < 0 and >= 1,000 million), and 10000-1.c
(sending other nsec values < 0 and >= 1,000 million) are failing. I've
included 5-1.c below.
Additional information is below:
kernel used = 2.5.54-bk1
HRT patches applied =
hrtimers-core-2.5.54-bk1-1.0.patch
hrtimers-hrposix-2.5.54-bk1-1.0.patch
hrtimers-i386-2.5.54-bk1-1.0.patch
hrtimers-posix-2.5.54-bk1-1.0.patch
hrtimers-support-2.5.52-1.0.patch
Thanks.
- Julie Fleischer
----
test 5-1.c below
(Output was: nanosleep() did not return -1 on failure)
/*
* Copyright (c) 2002, Intel Corporation. All rights reserved.
* Created by: julie.n.fleischer REMOVE-THIS AT intel DOT com
* This file is licensed under the GPL license. For the full content
* of this license, see the COPYING file at the top level of this
* source tree.
* Test that nanosleep() returns -1 on failure.
* Simulate failure condition by sending -1 as the nsec to sleep for.
*/
#include <stdio.h>
#include <time.h>
#define PTS_PASS 0
#define PTS_FAIL 1
#define PTS_UNRESOLVED 2
int main(int argc, char *argv[])
{
struct timespec tssleepfor, tsstorage;
int sleepnsec = -1;
tssleepfor.tv_sec=0;
tssleepfor.tv_nsec=sleepnsec;
if (nanosleep(&tssleepfor, &tsstorage) == -1) {
printf("Test PASSED\n");
return PTS_PASS;
} else {
printf("nanosleep() did not return -1 on failure\n");
return PTS_FAIL;
}
printf("This code should not be executed.\n");
return PTS_UNRESOLVED;
}
**These views are not necessarily those of my employer.**
"Fleischer, Julie N" wrote:
>
> George -
> In the latest 2.5.54-bk1 high-res-timers patches, it appears that
> nanosleep() is returning 0 (success) and not setting errno when an rqtp
> argument is sent that specifies a nsec value < 0 or >= 1000 million. In
> this instance, the POSIX System Interfaces doc states that errno is supposed
> to be set to EINVAL, and nanosleep should return -1.
Looks like I missed a line in the compatibility layer. I
found it and it will be fixed in the next release.
If you don't mind, I will add your test code to my
clock_nanosleep test code so this does not creep back in.
Thanks for the report.
-g
>
> In the 2.5.50 high-res-timers patches, behavior was as expected (i.e.,
> returned -1 and set errno=EINVAL). Unfortunately, I haven't looked at any
> patches since then to know exactly which patch stopped behaving as expected.
> A plain 2.5.54-bk1 kernel also behaves as expected (returns -1, sets
> errno=EINVAL).
>
> The tests I am using to reproduce this issue are part of the POSIX Test
> Suite at http://posixtest.sf.net under
> posixtestsuite/conformance/interfaces/nanosleep. 5-1.c (sending -1 nsec),
> 6-1.c (sending multiple nsec values < 0 and >= 1,000 million), and 10000-1.c
> (sending other nsec values < 0 and >= 1,000 million) are failing. I've
> included 5-1.c below.
>
> Additional information is below:
> kernel used = 2.5.54-bk1
> HRT patches applied =
> hrtimers-core-2.5.54-bk1-1.0.patch
> hrtimers-hrposix-2.5.54-bk1-1.0.patch
> hrtimers-i386-2.5.54-bk1-1.0.patch
> hrtimers-posix-2.5.54-bk1-1.0.patch
> hrtimers-support-2.5.52-1.0.patch
>
> Thanks.
> - Julie Fleischer
>
> ----
> test 5-1.c below
> (Output was: nanosleep() did not return -1 on failure)
>
> /*
> * Copyright (c) 2002, Intel Corporation. All rights reserved.
> * Created by: julie.n.fleischer REMOVE-THIS AT intel DOT com
> * This file is licensed under the GPL license. For the full content
> * of this license, see the COPYING file at the top level of this
> * source tree.
>
> * Test that nanosleep() returns -1 on failure.
> * Simulate failure condition by sending -1 as the nsec to sleep for.
> */
> #include <stdio.h>
> #include <time.h>
>
> #define PTS_PASS 0
> #define PTS_FAIL 1
> #define PTS_UNRESOLVED 2
>
> int main(int argc, char *argv[])
> {
> struct timespec tssleepfor, tsstorage;
> int sleepnsec = -1;
>
> tssleepfor.tv_sec=0;
> tssleepfor.tv_nsec=sleepnsec;
> if (nanosleep(&tssleepfor, &tsstorage) == -1) {
> printf("Test PASSED\n");
> return PTS_PASS;
> } else {
> printf("nanosleep() did not return -1 on failure\n");
> return PTS_FAIL;
> }
>
> printf("This code should not be executed.\n");
> return PTS_UNRESOLVED;
> }
>
> **These views are not necessarily those of my employer.**
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
--
George Anzinger [email protected]
High-res-timers:
http://sourceforge.net/projects/high-res-timers/
Preemption patch:
http://www.kernel.org/pub/linux/kernel/people/rml
> george anzinger wrote:
> If you don't mind, I will add your test code to my
> clock_nanosleep test code so this does not creep back in.
Sure. It's all GPL. The 6-1.c test case (attached below) probably has the
best test because it does multiple values, some < 0, some >= 1000 million.
One other thing I forgot to write in my previous report is that
clock_nanosleep() appeared to behave as expected, just nanosleep() showed
the issue of returning 0 on failure.
Thanks.
- Julie
----
test 6-1.c
/*
* Copyright (c) 2002, Intel Corporation. All rights reserved.
* Created by: julie.n.fleischer REMOVE-THIS AT intel DOT com
* This file is licensed under the GPL license. For the full content
* of this license, see the COPYING file at the top level of this
* source tree.
* Test that nanosleep() sets errno to EINVAL if rqtp contained a
* nanosecond value < 0 or >= 1,000 million
*/
#include <stdio.h>
#include <time.h>
#include <errno.h>
#define PTS_PASS 0
#define PTS_FAIL 1
#define PTS_UNRESOLVED 2
#define NUMTESTS 7
int main(int argc, char *argv[])
{
struct timespec tssleepfor, tsstorage;
int sleepnsec[NUMTESTS] = {-1, -5, -1000000000, 1000000000,
1000000001, 2000000000, 2000000000 };
int i;
int failure = 0;
tssleepfor.tv_sec=0;
for (i=0; i<NUMTESTS;i++) {
tssleepfor.tv_nsec=sleepnsec[i];
printf("sleep %d\n", sleepnsec[i]);
if (nanosleep(&tssleepfor, &tsstorage) == -1) {
if (EINVAL != errno) {
printf("errno != EINVAL\n");
failure = 1;
}
} else {
printf("nanosleep() did not return -1 on
failure\n");
return PTS_UNRESOLVED;
}
}
if (failure) {
printf("At least one test FAILED\n");
return PTS_FAIL;
} else {
printf("All tests PASSED\n");
return PTS_PASS;
}
}
**These views are not necessarily those of my employer.**
"Fleischer, Julie N" wrote:
>
> > george anzinger wrote:
> > If you don't mind, I will add your test code to my
> > clock_nanosleep test code so this does not creep back in.
>
> Sure. It's all GPL. The 6-1.c test case (attached below) probably has the
> best test because it does multiple values, some < 0, some >= 1000 million.
>
> One other thing I forgot to write in my previous report is that
> clock_nanosleep() appeared to behave as expected, just nanosleep() showed
> the issue of returning 0 on failure.
Yes, I found the problem. Thanks for the code and the heads
up.
-g
>
> Thanks.
> - Julie
>
> ----
> test 6-1.c
> /*
> * Copyright (c) 2002, Intel Corporation. All rights reserved.
> * Created by: julie.n.fleischer REMOVE-THIS AT intel DOT com
> * This file is licensed under the GPL license. For the full content
> * of this license, see the COPYING file at the top level of this
> * source tree.
>
> * Test that nanosleep() sets errno to EINVAL if rqtp contained a
> * nanosecond value < 0 or >= 1,000 million
> */
> #include <stdio.h>
> #include <time.h>
> #include <errno.h>
>
> #define PTS_PASS 0
> #define PTS_FAIL 1
> #define PTS_UNRESOLVED 2
>
> #define NUMTESTS 7
>
> int main(int argc, char *argv[])
> {
> struct timespec tssleepfor, tsstorage;
> int sleepnsec[NUMTESTS] = {-1, -5, -1000000000, 1000000000,
> 1000000001, 2000000000, 2000000000 };
> int i;
> int failure = 0;
>
> tssleepfor.tv_sec=0;
>
> for (i=0; i<NUMTESTS;i++) {
> tssleepfor.tv_nsec=sleepnsec[i];
> printf("sleep %d\n", sleepnsec[i]);
> if (nanosleep(&tssleepfor, &tsstorage) == -1) {
> if (EINVAL != errno) {
> printf("errno != EINVAL\n");
> failure = 1;
> }
> } else {
> printf("nanosleep() did not return -1 on
> failure\n");
> return PTS_UNRESOLVED;
> }
> }
>
> if (failure) {
> printf("At least one test FAILED\n");
> return PTS_FAIL;
> } else {
> printf("All tests PASSED\n");
> return PTS_PASS;
> }
> }
> **These views are not necessarily those of my employer.**
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
--
George Anzinger [email protected]
High-res-timers:
http://sourceforge.net/projects/high-res-timers/
Preemption patch:
http://www.kernel.org/pub/linux/kernel/people/rml