Skip to contentSkip to conversation

openposix: timer_*/speculative: Handle SIGSEGV on old glibc

PATCH In review Avinesh Kumar via ltp 4 participants 9 messages Sep 29 – Sep 30

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 openposix: timer_*/speculative: Handle SIGSEGV on old glibc

From: Avinesh Kumar <[email protected]>

timer_delete/speculative/5-1, timer_getoverrun/speculative/6-1, timer_gettime/speculative/6-1, and timer_settime/speculative/12-1 all pass a bogus BOGUSTIMERID/BOGUSTID value as an invalid timerid.

Commit d377d1cff1ec ("openposix: timer_*/speculative: Don't pass a stack pointer as timerid") fixed a SIGSEGV on i586, where current glibc's sign-bit check (timer_is_sigev_thread()) can misread a stack address as a tagged pointer. A small constant avoids that.

But older glibc (<= 2.31, before this check existed) unconditionally dereferences timerid as a pointer, so the same small constant now faults instead, on any architecture:

timer_delete_sp[13131]: segfault at 186a3 ip 00007f15973aed4e sp 00007ffefb52c5a0 error 4 in librt-2.31.so[7f15973aa000+8000]

POSIX defines no required behavior for an invalid timer ID. Install a handler to accept SIGSEGV.

Fixes d377d1cff1ec ("openposix: timer_*/speculative: Don't pass a stack pointer as timerid")
---
 .../interfaces/timer_delete/speculative/5-1.c | 18 ++++++++++++++++++
 .../timer_getoverrun/speculative/6-1.c        | 18 ++++++++++++++++++
 .../timer_gettime/speculative/6-1.c           | 18 ++++++++++++++++++
 .../timer_settime/speculative/12-1.c          | 19 +++++++++++++++++++
 4 files changed, 73 insertions(+)
testcases/open_posix_testsuite/conformance/interfaces/timer_delete/speculative/5-1.c +18 -0
index dce9277b5d6a..c7a94e1446f7 100644
@@ -13,13 +13,31 @@
 #include <time.h>
 #include <stdio.h>
 #include <errno.h>
+#include <signal.h>
+#include <unistd.h>
 #include "posixtest.h"
 
 #define BOGUSTIMERID 99999
 
+/*
+ * timerid is not a valid timer ID; POSIX defines no required behavior
+ * here (EINVAL is only a recommendation), so SIGSEGV is an equally
+ * valid outcome.
+ */
+static void sigsegv_handler(int signum PTS_ATTRIBUTE_UNUSED)
+{
+	PTS_WRITE_MSG("Got SIGSEGV when calling timer_delete() with an invalid timer ID\n");
+	PTS_WRITE_MSG("Test PASSED\n");
+	_exit(PTS_PASS);
+}
+
 int test_main(int argc PTS_ATTRIBUTE_UNUSED, char **argv PTS_ATTRIBUTE_UNUSED)
 {
 	timer_t tid = (timer_t)BOGUSTIMERID;
+	struct sigaction sa = { .sa_handler = sigsegv_handler };
+
+	sigfillset(&sa.sa_mask);
+	sigaction(SIGSEGV, &sa, NULL);
 
 	if (timer_delete(tid) == -1) {
 		if (errno == EINVAL) {
testcases/open_posix_testsuite/conformance/interfaces/timer_getoverrun/speculative/6-1.c +18 -0
index 3a8f1448d4d4..7861e99d4c8d 100644
@@ -13,13 +13,31 @@
 #include <time.h>
 #include <stdio.h>
 #include <errno.h>
+#include <signal.h>
+#include <unistd.h>
 #include "posixtest.h"
 
 #define BOGUSTID 9999
 
+/*
+ * timerid is not a valid timer ID; POSIX defines no required behavior
+ * here (EINVAL is only a recommendation), so SIGSEGV is an equally
+ * valid outcome.
+ */
+static void sigsegv_handler(int signum PTS_ATTRIBUTE_UNUSED)
+{
+	PTS_WRITE_MSG("Got SIGSEGV when calling timer_getoverrun() with an invalid timer ID\n");
+	PTS_WRITE_MSG("Test PASSED\n");
+	_exit(PTS_PASS);
+}
+
 int test_main(int argc PTS_ATTRIBUTE_UNUSED, char **argv PTS_ATTRIBUTE_UNUSED)
 {
 	timer_t tid = (timer_t)BOGUSTID;
+	struct sigaction sa = { .sa_handler = sigsegv_handler };
+
+	sigfillset(&sa.sa_mask);
+	sigaction(SIGSEGV, &sa, NULL);
 
 	if (timer_getoverrun(tid) == -1) {
 		if (EINVAL == errno) {
testcases/open_posix_testsuite/conformance/interfaces/timer_gettime/speculative/6-1.c +18 -0
index 586b0ed3a2b1..a3a9e080e26c 100644
@@ -13,14 +13,32 @@
 #include <time.h>
 #include <stdio.h>
 #include <errno.h>
+#include <signal.h>
+#include <unistd.h>
 #include "posixtest.h"
 
 #define BOGUSTID 9999
 
+/*
+ * timerid is not a valid timer ID; POSIX defines no required behavior
+ * here (EINVAL is only a recommendation), so SIGSEGV is an equally
+ * valid outcome.
+ */
+static void sigsegv_handler(int signum PTS_ATTRIBUTE_UNUSED)
+{
+	PTS_WRITE_MSG("Got SIGSEGV when calling timer_gettime() with an invalid timer ID\n");
+	PTS_WRITE_MSG("Test PASSED\n");
+	_exit(PTS_PASS);
+}
+
 int test_main(int argc PTS_ATTRIBUTE_UNUSED, char **argv PTS_ATTRIBUTE_UNUSED)
 {
 	struct itimerspec its;
 	timer_t tid = (timer_t)BOGUSTID;
+	struct sigaction sa = { .sa_handler = sigsegv_handler };
+
+	sigfillset(&sa.sa_mask);
+	sigaction(SIGSEGV, &sa, NULL);
 
 	if (timer_gettime(tid, &its) == -1) {
 		if (EINVAL == errno) {
testcases/open_posix_testsuite/conformance/interfaces/timer_settime/speculative/12-1.c +19 -0
index ac7f7bf24c39..953d585b8730 100644
@@ -12,14 +12,33 @@
 #include <time.h>
 #include <stdio.h>
 #include <errno.h>
+#include <signal.h>
+#include <unistd.h>
 #include "posixtest.h"
 
 #define BOGUSTID 9999
 
+/*
+ * timerid is not a valid timer ID; POSIX defines no required behavior
+ * here (EINVAL is only a recommendation), so SIGSEGV is an equally
+ * valid outcome.
+ */
+static void sigsegv_handler(int signum PTS_ATTRIBUTE_UNUSED)
+{
+	PTS_WRITE_MSG("Got SIGSEGV when calling timer_settime() with an invalid timer ID\n");
+	PTS_WRITE_MSG("Test PASSED\n");
+	_exit(PTS_PASS);
+}
+
 int test_main(int argc PTS_ATTRIBUTE_UNUSED, char **argv PTS_ATTRIBUTE_UNUSED)
 {
 	struct itimerspec its;
 	timer_t tid = (timer_t)BOGUSTID;
+	struct sigaction sa = { .sa_handler = sigsegv_handler };
+
+	sigfillset(&sa.sa_mask);
+	sigaction(SIGSEGV, &sa, NULL);
+
 	its.it_interval.tv_sec = 0;
 	its.it_interval.tv_nsec = 0;
 	its.it_value.tv_sec = 0;
Signature · 5 lines
2.55.0


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#1 lore raw

Hi Avinesh,

Quoted 25 lines · Avinesh Kumar

openposix: timer_*/speculative: Handle SIGSEGV on old glibc

testcases/open_posix_testsuite/conformance/interfaces/timer_delete/speculative/5-1.c +11 -0
index dce9277b5..c7a94e144 100644
@@ -13,13 +13,31 @@
 #include <time.h>
 #include <stdio.h>
 #include <errno.h>
+#include <signal.h>
+#include <unistd.h>
 #include "posixtest.h"
 
 #define BOGUSTIMERID 99999
 
+/*
+ * timerid is not a valid timer ID; POSIX defines no required behavior
+ * here (EINVAL is only a recommendation), so SIGSEGV is an equally
+ * valid outcome.
+ */
+static void sigsegv_handler(int signum PTS_ATTRIBUTE_UNUSED)
+{
+	PTS_WRITE_MSG("Got SIGSEGV when calling timer_delete() with an invalid timer ID\n");
+	PTS_WRITE_MSG("Test PASSED\n");

testcases/open_posix_testsuite/conformance/interfaces/timer_delete/speculative/5-1.c line 40

Quoted 10 lines
testcases/open_posix_testsuite/conformance/interfaces/timer_delete/speculative/5-1.c +7 -0
+	_exit(PTS_PASS);
+}
+
 int test_main(int argc PTS_ATTRIBUTE_UNUSED, char **argv PTS_ATTRIBUTE_UNUSED)
 {
 	timer_t tid = (timer_t)BOGUSTIMERID;
+	struct sigaction sa = { .sa_handler = sigsegv_handler };
+
+	sigfillset(&sa.sa_mask);
+	sigaction(SIGSEGV, &sa, NULL);

Catching SIGSEGV and exiting with PTS_PASS circumvents the check for errno=EINVAL and masks faults instead of verifying the interface. Use "static int tval = BOGUSTIMERID; timer_t tid = (timer_t)&tval;" without a signal handler instead; this provides a readable pointer for older glibc (<= 2.31) while keeping the MSB clear so glibc >= 2.34 and musl do not treat it as a SIGEV_THREAD timer.

Quoted 23 lines
testcases/open_posix_testsuite/conformance/interfaces/timer_getoverrun/speculative/6-1.c +11 -0
index 3a8f1448d..7861e99d4 100644
@@ -13,13 +13,31 @@
 #include <time.h>
 #include <stdio.h>
 #include <errno.h>
+#include <signal.h>
+#include <unistd.h>
 #include "posixtest.h"
 
 #define BOGUSTID 9999
 
+/*
+ * timerid is not a valid timer ID; POSIX defines no required behavior
+ * here (EINVAL is only a recommendation), so SIGSEGV is an equally
+ * valid outcome.
+ */
+static void sigsegv_handler(int signum PTS_ATTRIBUTE_UNUSED)
+{
+	PTS_WRITE_MSG("Got SIGSEGV when calling timer_getoverrun() with an invalid timer ID\n");
+	PTS_WRITE_MSG("Test PASSED\n");

testcases/open_posix_testsuite/conformance/interfaces/timer_getoverrun/speculative/6-1.c line 40

Quoted 10 lines
testcases/open_posix_testsuite/conformance/interfaces/timer_getoverrun/speculative/6-1.c +7 -0
+	_exit(PTS_PASS);
+}
+
 int test_main(int argc PTS_ATTRIBUTE_UNUSED, char **argv PTS_ATTRIBUTE_UNUSED)
 {
 	timer_t tid = (timer_t)BOGUSTID;
+	struct sigaction sa = { .sa_handler = sigsegv_handler };
+
+	sigfillset(&sa.sa_mask);
+	sigaction(SIGSEGV, &sa, NULL);

Same issue here; pass a pointer to a static int variable instead of handling SIGSEGV.

Quoted 24 lines
testcases/open_posix_testsuite/conformance/interfaces/timer_gettime/speculative/6-1.c +12 -0
index 586b0ed3a..a3a9e080e 100644
@@ -13,14 +13,32 @@
 #include <time.h>
 #include <stdio.h>
 #include <errno.h>
+#include <signal.h>
+#include <unistd.h>
 #include "posixtest.h"
 
 #define BOGUSTID 9999
 
+/*
+ * timerid is not a valid timer ID; POSIX defines no required behavior
+ * here (EINVAL is only a recommendation), so SIGSEGV is an equally
+ * valid outcome.
+ */
+static void sigsegv_handler(int signum PTS_ATTRIBUTE_UNUSED)
+{
+	PTS_WRITE_MSG("Got SIGSEGV when calling timer_gettime() with an invalid timer ID\n");
+	PTS_WRITE_MSG("Test PASSED\n");
+	_exit(PTS_PASS);

testcases/open_posix_testsuite/conformance/interfaces/timer_gettime/speculative/6-1.c line 41

Quoted 10 lines
testcases/open_posix_testsuite/conformance/interfaces/timer_gettime/speculative/6-1.c +6 -0
+}
+
 int test_main(int argc PTS_ATTRIBUTE_UNUSED, char **argv PTS_ATTRIBUTE_UNUSED)
 {
 	struct itimerspec its;
 	timer_t tid = (timer_t)BOGUSTID;
+	struct sigaction sa = { .sa_handler = sigsegv_handler };
+
+	sigfillset(&sa.sa_mask);
+	sigaction(SIGSEGV, &sa, NULL);

Same issue here; pass a pointer to a static int variable instead of handling SIGSEGV.

Quoted 24 lines
testcases/open_posix_testsuite/conformance/interfaces/timer_settime/speculative/12-1.c +12 -0
index ac7f7bf24..953d585b8 100644
@@ -12,14 +12,33 @@
 #include <time.h>
 #include <stdio.h>
 #include <errno.h>
+#include <signal.h>
+#include <unistd.h>
 #include "posixtest.h"
 
 #define BOGUSTID 9999
 
+/*
+ * timerid is not a valid timer ID; POSIX defines no required behavior
+ * here (EINVAL is only a recommendation), so SIGSEGV is an equally
+ * valid outcome.
+ */
+static void sigsegv_handler(int signum PTS_ATTRIBUTE_UNUSED)
+{
+	PTS_WRITE_MSG("Got SIGSEGV when calling timer_settime() with an invalid timer ID\n");
+	PTS_WRITE_MSG("Test PASSED\n");
+	_exit(PTS_PASS);

testcases/open_posix_testsuite/conformance/interfaces/timer_settime/speculative/12-1.c line 40

Quoted 10 lines
testcases/open_posix_testsuite/conformance/interfaces/timer_settime/speculative/12-1.c +6 -0
+}
+
 int test_main(int argc PTS_ATTRIBUTE_UNUSED, char **argv PTS_ATTRIBUTE_UNUSED)
 {
 	struct itimerspec its;
 	timer_t tid = (timer_t)BOGUSTID;
+	struct sigaction sa = { .sa_handler = sigsegv_handler };
+
+	sigfillset(&sa.sa_mask);
+	sigaction(SIGSEGV, &sa, NULL);

Same issue here; pass a pointer to a static int variable instead of handling SIGSEGV.

Verdict - Needs revision

---
Note:

The agent can sometimes produce false positives although often its findings are genuine. If you find issues with the review, please comment this email or ignore the suggestions.

Regards,
LTP AI Reviewer

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
#2 lore raw

PATCH v2 1/1 openposix: timer_*/speculative: fork() to isolate the SIGSEGV on old glibc

In reply to Avinesh Kumar via ltp

From: Avinesh Kumar <[email protected]>

Commit d377d1cff1ec ("openposix: timer_*/speculative: Don't pass a stack pointer as timerid") fixed a SIGSEGV on i586, where current glibc's sign-bit check (timer_is_sigev_thread()) can misread a stack address as a tagged pointer. A small constant avoids that.

But older glibc (<= 2.31, before this check existed) unconditionally dereferences timerid as a pointer, so the same small constant now faults instead:

timer_delete_sp[13131]: segfault at 186a3 ip 00007f15973aed4e sp 00007ffefb52c5a0 error 4 in librt-2.31.so[7f15973aa000+8000]

fork() a child to make the call, then inspect the child's exit status. POSIX defines no required behavior for an invalid timer ID (EINVAL is only a recommendation), so the child dying from any signal is treated as success.

Fixes d377d1cff1ec ("openposix: timer_*/speculative: Don't pass a stack pointer as timerid")
---
 .../interfaces/timer_delete/speculative/5-1.c | 50 +++++++++++-----
 .../timer_getoverrun/speculative/6-1.c        | 49 ++++++++++-----
 .../timer_gettime/speculative/6-1.c           | 52 +++++++++++-----
 .../timer_settime/speculative/12-1.c          | 60 +++++++++++++------
 4 files changed, 145 insertions(+), 66 deletions(-)
testcases/open_posix_testsuite/conformance/interfaces/timer_delete/speculative/5-1.c +34 -16
index dce9277b5d6a..d1857efead18 100644
@@ -5,34 +5,52 @@
  * of this license, see the COPYING file at the top level of this
  * source tree.
 
- * Test to see if timer_delete() returns -1 and sets errno==EINVAL if
- * timerid is not a valid timer ID or not.
- * Since this is a "may" requirement, either option is a PASS.
+ * Test to see if timer_delete() sets errno==EINVAL for an invalid
+ * timerid. Since this is a "may" requirement, succeeding or crashing
+ * are equally valid outcomes.
  */
 
 #include <time.h>
 #include <stdio.h>
 #include <errno.h>
+#include <unistd.h>
+#include <sys/wait.h>
 #include "posixtest.h"
 
 #define BOGUSTIMERID 99999
 
 int test_main(int argc PTS_ATTRIBUTE_UNUSED, char **argv PTS_ATTRIBUTE_UNUSED)
 {
-	timer_t tid = (timer_t)BOGUSTIMERID;
-
-	if (timer_delete(tid) == -1) {
-		if (errno == EINVAL) {
-			printf
-			    ("timer_delete() returned -1 and set errno=EINVAL\n");
-			return PTS_PASS;
-		} else {
-			printf
-			    ("timer_delete() returned -1, but didn't set errno!=EINVAL\n");
-			return PTS_FAIL;
+	pid_t pid;
+	int status;
+
+	pid = fork();
+	if (pid == -1) {
+		perror("fork");
+		return PTS_UNRESOLVED;
+	}
+
+	if (pid == 0) {
+		timer_t tid = (timer_t) BOGUSTIMERID;
+
+		if (timer_delete(tid) == -1) {
+			if (errno == EINVAL)
+				_exit(PTS_PASS);
+			_exit(PTS_FAIL);
 		}
+		_exit(PTS_PASS);
+	}
+
+	if (waitpid(pid, &status, 0) == -1) {
+		perror("waitpid");
+		return PTS_UNRESOLVED;
+	}
+
+	if (WIFSIGNALED(status) || (WIFEXITED(status) && WEXITSTATUS(status) == 0)) {
+		printf("Test PASSED\n");
+		return PTS_PASS;
 	}
 
-	printf("timer_delete() did not return -1\n");
-	return PTS_PASS;
+	printf("Test FAILED\n");
+	return PTS_FAIL;
 }
testcases/open_posix_testsuite/conformance/interfaces/timer_getoverrun/speculative/6-1.c +34 -15

67 lines not on this page · raw message

testcases/open_posix_testsuite/conformance/interfaces/timer_gettime/speculative/6-1.c +36 -16

69 lines not on this page · raw message

testcases/open_posix_testsuite/conformance/interfaces/timer_settime/speculative/12-1.c +41 -19

77 lines not on this page · raw message

Signature · 5 lines
2.55.0


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#3 lore raw

Re: [LTP] openposix: timer_*/speculative: fork() to isolate the SIGSEGV on old glibc

Hi Avinesh,

Quoted 5 lines · Avinesh Kumar

openposix: timer_*/speculative: fork() to isolate the SIGSEGV on old glibc

* Test to see if timer_delete() sets errno==EINVAL for an invalid
* timerid. Since this is a "may" requirement, succeeding or crashing
* are equally valid outcomes.

POSIX "may fail" requirements define that returning -1 with EINVAL or succeeding are valid outcomes; crashing is undefined behavior due to pointer dereference on older glibc, not a defined requirement outcome. Please update the comments across all four tests to clarify that returning -1 with EINVAL, succeeding, or crashing (on older glibc) are accepted.

testcases/open_posix_testsuite/conformance/interfaces/timer_gettime/speculative/6-1.c line 53

Quoted 4 lines
testcases/open_posix_testsuite/conformance/interfaces/timer_gettime/speculative/6-1.c +4 -0
+	if (WIFSIGNALED(status) || (WIFEXITED(status) && WEXITSTATUS(status) == 0)) {
+		printf("Test PASSED\n");
+		return PTS_PASS;
+	}

WIFSIGNALED(status) without checking WTERMSIG(status) treats any terminating signal as success, which masks abnormal process exits (such as SIGABRT or SIGSYS) as well as test cancellation (e.g. SIGINT or test-runner SIGTERM). Check for the expected fault explicitly with WIFSIGNALED(status) && (WTERMSIG(status) == SIGSEGV || WTERMSIG(status) == SIGBUS), and use PTS_PASS instead of literal 0 for WEXITSTATUS(status). This applies to all four tests.

testcases/open_posix_testsuite/conformance/interfaces/timer_delete/speculative/5-1.c line 55

Quoted 8 lines
testcases/open_posix_testsuite/conformance/interfaces/timer_delete/speculative/5-1.c +7 -0
+		if (timer_delete(tid) == -1) {
+			if (errno == EINVAL)
+				_exit(PTS_PASS);
+			_exit(PTS_FAIL);
+		}
 
+	printf("Test FAILED\n");
+	return PTS_FAIL;

Printing only "Test FAILED" removes diagnostic information when the timer call fails with an unexpected errno, violating Open POSIX result reporting rules. The child should print the unexpected errno before calling _exit(), or the parent should report the reason for failure before returning PTS_FAIL.

Verdict - Needs revision

---
Note:

The agent can sometimes produce false positives although often its findings are genuine. If you find issues with the review, please comment this email or ignore the suggestions.

Regards,
LTP AI Reviewer

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
#4 lore raw

Re: [LTP] openposix: timer_*/speculative: fork() to isolate the SIGSEGV on old glibc

Hi Avinesh,

unfortunately all agent complains sounds reasonable, would you mind to send v3?

Kind regards,
Petr

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
#5 lore raw

PATCH v3 1/1 openposix: timer_*/speculative: fork() to isolate the SIGSEGV on old glibc

From: Avinesh Kumar <[email protected]>

Commit d377d1cff1ec ("openposix: timer_*/speculative: Don't pass a stack pointer as timerid") fixed a SIGSEGV on i586, where current glibc's sign-bit check (timer_is_sigev_thread()) can misread a stack address as a tagged pointer. A small constant avoids that.

But older glibc (<= 2.31, before this check existed) unconditionally dereferences timerid as a pointer, so the same small constant now faults instead:

timer_delete_sp[13131]: segfault at 186a3 ip 00007f15973aed4e sp 00007ffefb52c5a0 error 4 in librt-2.31.so[7f15973aa000+8000]

fork() a child to make the call, then inspect the child's exit status. POSIX defines no required behavior for an invalid timer ID (EINVAL is only a recommendation), so the child being killed by SIGSEGV or SIGBUS is treated as success.

Fixes d377d1cff1ec ("openposix: timer_*/speculative: Don't pass a stack pointer as timerid")
---
 .../interfaces/timer_delete/speculative/5-1.c | 56 +++++++++++-----
 .../timer_getoverrun/speculative/6-1.c        | 55 +++++++++++-----
 .../timer_gettime/speculative/6-1.c           | 58 +++++++++++-----
 .../timer_settime/speculative/12-1.c          | 66 +++++++++++++------
 4 files changed, 169 insertions(+), 66 deletions(-)
testcases/open_posix_testsuite/conformance/interfaces/timer_delete/speculative/5-1.c +40 -16
index dce9277b5d6a..8e7d1b2e54dc 100644
@@ -5,34 +5,58 @@
  * of this license, see the COPYING file at the top level of this
  * source tree.
 
- * Test to see if timer_delete() returns -1 and sets errno==EINVAL if
- * timerid is not a valid timer ID or not.
- * Since this is a "may" requirement, either option is a PASS.
+ * Test to see if timer_delete() sets errno==EINVAL for an invalid
+ * timerid. Recent POSIX versions leave this behavior undefined, so
+ * returning -1 with EINVAL, succeeding, or crashing (older glibc
+ * dereferences timerid as a pointer) are all accepted outcomes.
  */
 
 #include <time.h>
 #include <stdio.h>
 #include <errno.h>
+#include <unistd.h>
+#include <signal.h>
+#include <string.h>
+#include <sys/wait.h>
 #include "posixtest.h"
 
 #define BOGUSTIMERID 99999
 
 int test_main(int argc PTS_ATTRIBUTE_UNUSED, char **argv PTS_ATTRIBUTE_UNUSED)
 {
-	timer_t tid = (timer_t)BOGUSTIMERID;
-
-	if (timer_delete(tid) == -1) {
-		if (errno == EINVAL) {
-			printf
-			    ("timer_delete() returned -1 and set errno=EINVAL\n");
-			return PTS_PASS;
-		} else {
-			printf
-			    ("timer_delete() returned -1, but didn't set errno!=EINVAL\n");
-			return PTS_FAIL;
+	pid_t pid;
+	int status;
+
+	pid = fork();
+	if (pid == -1) {
+		perror("fork");
+		return PTS_UNRESOLVED;
+	}
+
+	if (pid == 0) {
+		timer_t tid = (timer_t) BOGUSTIMERID;
+
+		if (timer_delete(tid) == -1) {
+			if (errno == EINVAL)
+				_exit(PTS_PASS);
+			printf("timer_delete() returned -1, but errno=%d (%s), not EINVAL\n", errno, strerror(errno));
+			fflush(stdout);
+			_exit(PTS_FAIL);
 		}
+		_exit(PTS_PASS);
+	}
+
+	if (waitpid(pid, &status, 0) == -1) {
+		perror("waitpid");
+		return PTS_UNRESOLVED;
+	}
+
+	if ((WIFSIGNALED(status) && (WTERMSIG(status) == SIGSEGV || WTERMSIG(status) == SIGBUS)) ||
+		(WIFEXITED(status) && WEXITSTATUS(status) == PTS_PASS)) {
+		printf("Test PASSED\n");
+		return PTS_PASS;
 	}
 
-	printf("timer_delete() did not return -1\n");
-	return PTS_PASS;
+	printf("Test FAILED\n");
+	return PTS_FAIL;
 }
testcases/open_posix_testsuite/conformance/interfaces/timer_getoverrun/speculative/6-1.c +40 -15

73 lines not on this page · raw message

testcases/open_posix_testsuite/conformance/interfaces/timer_gettime/speculative/6-1.c +42 -16

75 lines not on this page · raw message

testcases/open_posix_testsuite/conformance/interfaces/timer_settime/speculative/12-1.c +47 -19

83 lines not on this page · raw message

Signature · 5 lines
2.55.0


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by on v3 1/1
#6 lore raw

Re: [LTP] [PATCH v3] openposix: timer_*/speculative: fork() to isolate the SIGSEGV on old glibc

Hi Avinesh,

Quoted 31 lines

Commit d377d1cff1ec ("openposix: timer_*/speculative: Don't pass a stack pointer as timerid") fixed a SIGSEGV on i586, where current glibc's sign-bit check (timer_is_sigev_thread()) can misread a stack address as a tagged pointer. A small constant avoids that.

But older glibc (<= 2.31, before this check existed) unconditionally dereferences timerid as a pointer, so the same small constant now faults instead:

timer_delete_sp[13131]: segfault at 186a3 ip 00007f15973aed4e sp 00007ffefb52c5a0 error 4 in librt-2.31.so[7f15973aa000+8000]

fork() a child to make the call, then inspect the child's exit status. POSIX defines no required behavior for an invalid timer ID (EINVAL is only a recommendation), so the child being killed by SIGSEGV or SIGBUS is treated as success.

Fixes d377d1cff1ec ("openposix: timer_*/speculative: Don't pass a stack pointer as timerid")
---
 .../interfaces/timer_delete/speculative/5-1.c | 56 +++++++++++-----
 .../timer_getoverrun/speculative/6-1.c        | 55 +++++++++++-----
 .../timer_gettime/speculative/6-1.c           | 58 +++++++++++-----
 .../timer_settime/speculative/12-1.c          | 66 +++++++++++++------
 4 files changed, 169 insertions(+), 66 deletions(-)
testcases/open_posix_testsuite/conformance/interfaces/timer_delete/speculative/5-1.c +0 -0
index dce9277b5d6a..8e7d1b2e54dc 100644

testcases/open_posix_testsuite/conformance/interfaces/timer_delete/speculative/5-1.c line 9

Quoted 8 lines
testcases/open_posix_testsuite/conformance/interfaces/timer_delete/speculative/5-1.c +2 -3
@@ -5,34 +5,58 @@
  * of this license, see the COPYING file at the top level of this
  * source tree.
 
- * Test to see if timer_delete() returns -1 and sets errno==EINVAL if
- * timerid is not a valid timer ID or not.
- * Since this is a "may" requirement, either option is a PASS.
+ * Test to see if timer_delete() sets errno==EINVAL for an invalid
+ * timerid. Recent POSIX versions leave this behavior undefined, so

nit: I suggest before merge to mention the version:

"Recent POSIX versions leave" => "/Since POSIX 2016 Edition is"

(no need for v4). Why? In 5+ years we can hopefully delete those tests. Specifying the version saves somebody searching it again.

The code itself LGTM.

Kind regards,
Petr

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by
#7 lore raw

Re: [LTP] [PATCH v3] openposix: timer_*/speculative: fork() to isolate the SIGSEGV on old glibc

Hi!

Quoted 9 lines

nit: I suggest before merge to mention the version:

"Recent POSIX versions leave" => "/Since POSIX 2016 Edition is"

(no need for v4). Why? In 5+ years we can hopefully delete those tests. Specifying the version saves somebody searching it again.

The code itself LGTM.

Agree, with that minor change:

Also I find this safe enough to be applied before release as this change touches only a few speculative tests.

Signature · 5 lines
Cyril Hrubis
[email protected]

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#8 lore raw

Re: [LTP] [PATCH v3] openposix: timer_*/speculative: fork() to isolate the SIGSEGV on old glibc

Quoted 11 lines

Hi!

Quoted 9 lines

nit: I suggest before merge to mention the version:

"Recent POSIX versions leave" => "/Since POSIX 2016 Edition is"

(no need for v4). Why? In 5+ years we can hopefully delete those tests. Specifying the version saves somebody searching it again.

The code itself LGTM.

Quoted 6 lines

Agree, with that minor change:

Also I find this safe enough to be applied before release as this change touches only a few speculative tests.

+1, merged. Thank you all.

Kind regards,
Petr

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp