[PATCH] selftests/harness: Prevent infinite loop due to Assert in FIXTURE_TEARDOWN

[PATCH] selftests/harness: Prevent infinite loop due to Assert in FIXTURE_TEARDOWN

[Shengyu Li]

This patch addresses an issue in the selftests/harness where an assertion within FIXTURE_TEARDOWN could trigger an infinite loop. The problem arises because the teardown procedure is meant to execute once, but the presence of failing assertions (ASSERT_EQ(0, 1)) leads to repeated attempts to execute teardown due to the long jump mechanism used by the harness for handling assertions.

To resolve this, the patch ensures that the teardown process runs only once, regardless of assertion outcomes, preventing the infinite loop and allowing tests to fail.

Signed-off-by: Shengyu Li <shengyu.li.evgeny@gmail.com>
---
 tools/testing/selftests/kselftest_harness.h | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/kselftest_harness.h b/tools/testing/selftests/kselftest_harness.h
index 4fd735e48ee7..230d62884885 100644
--- a/tools/testing/selftests/kselftest_harness.h
+++ b/tools/testing/selftests/kselftest_harness.h
@@ -383,6 +383,7 @@
 		FIXTURE_DATA(fixture_name) self; \
 		pid_t child = 1; \
 		int status = 0; \
+		bool jmp = false; \
 		memset(&self, 0, sizeof(FIXTURE_DATA(fixture_name))); \
 		if (setjmp(_metadata->env) == 0) { \
 			/* Use the same _metadata. */ \
@@ -399,8 +400,10 @@
 				_metadata->exit_code = KSFT_FAIL; \
 			} \
 		} \
+		else \
+			jmp = true; \
 		if (child == 0) { \
-			if (_metadata->setup_completed && !_metadata->teardown_parent) \
+			if (_metadata->setup_completed && !_metadata->teardown_parent && !jmp) \
 				fixture_name##_teardown(_metadata, &self, variant->data); \
 			_exit(0); \
 		} \
-- 
2.25.1

[PATCH] selftests/harness: Prevent infinite loop due to Assert in FIXTURE_TEARDOWN

[Shengyu Li]

This patch addresses an issue in the selftests/harness where an assertion within FIXTURE_TEARDOWN could trigger an infinite loop. 
The problem arises because the teardown procedure is meant to execute once, 
but the presence of failing assertions (ASSERT_EQ(0, 1)) 
leads to repeated attempts to execute teardown due to the long jump mechanism used by the harness for handling assertions.

To resolve this, the patch ensures that the teardown process runs only once, regardless of assertion outcomes, 
preventing the infinite loop and allowing tests to fail.

A simple test demo, also this is related to the issue mentioned in this patch
https://patchwork.kernel.org/project/linux-kselftest/patch/e2ba3f8c-80e6-477d-9cea-1c9af820e0ed@alu.unizg.hr/

#include "kselftest_harness.h"

FIXTURE(f)
{
	int fd;
};

FIXTURE_SETUP(f)
{
	self->fd = 0;
}

FIXTURE_TEARDOWN(f)
{
	TH_LOG("TEARDOWN");
	ASSERT_EQ(0, 1);
	self->fd = -1;
}

TEST_F(f, open_close)
{
	ASSERT_NE(self->fd, 1);
}

TEST_HARNESS_MAIN


Signed-off-by: Shengyu Li <shengyu.li.evgeny@gmail.com>
---
 tools/testing/selftests/kselftest_harness.h | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/kselftest_harness.h b/tools/testing/selftests/kselftest_harness.h
index 4fd735e48ee7..230d62884885 100644
--- a/tools/testing/selftests/kselftest_harness.h
+++ b/tools/testing/selftests/kselftest_harness.h
@@ -383,6 +383,7 @@
 		FIXTURE_DATA(fixture_name) self; \
 		pid_t child = 1; \
 		int status = 0; \
+		bool jmp = false; \
 		memset(&self, 0, sizeof(FIXTURE_DATA(fixture_name))); \
 		if (setjmp(_metadata->env) == 0) { \
 			/* Use the same _metadata. */ \
@@ -399,8 +400,10 @@
 				_metadata->exit_code = KSFT_FAIL; \
 			} \
 		} \
+		else \
+			jmp = true; \
 		if (child == 0) { \
-			if (_metadata->setup_completed && !_metadata->teardown_parent) \
+			if (_metadata->setup_completed && !_metadata->teardown_parent && !jmp) \
 				fixture_name##_teardown(_metadata, &self, variant->data); \
 			_exit(0); \
 		} \
-- 
2.25.1

Re: [PATCH] selftests/harness: Prevent infinite loop due to Assert in FIXTURE_TEARDOWN

[Shuah Khan]

On 3/14/24 16:43, Shengyu Li wrote:
> This patch addresses an issue in the selftests/harness where an assertion within FIXTURE_TEARDOWN could trigger an infinite loop. The problem arises because the teardown procedure is meant to execute once, but the presence of failing assertions (ASSERT_EQ(0, 1)) leads to repeated attempts to execute teardown due to the long jump mechanism used by the harness for handling assertions.
> 
> To resolve this, the patch ensures that the teardown process runs only once, regardless of assertion outcomes, preventing the infinite loop and allowing tests to fail.
> 

Please make sure change log is formatted with 70-75 character
long lines.

Please include information on how you found this problem.

Send v2 with that change.

> Signed-off-by: Shengyu Li <shengyu.li.evgeny@gmail.com>
> ---
>   tools/testing/selftests/kselftest_harness.h | 5 ++++-
>   1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/tools/testing/selftests/kselftest_harness.h b/tools/testing/selftests/kselftest_harness.h
> index 4fd735e48ee7..230d62884885 100644
> --- a/tools/testing/selftests/kselftest_harness.h
> +++ b/tools/testing/selftests/kselftest_harness.h
> @@ -383,6 +383,7 @@
>   		FIXTURE_DATA(fixture_name) self; \
>   		pid_t child = 1; \
>   		int status = 0; \
> +		bool jmp = false; \
>   		memset(&self, 0, sizeof(FIXTURE_DATA(fixture_name))); \
>   		if (setjmp(_metadata->env) == 0) { \
>   			/* Use the same _metadata. */ \
> @@ -399,8 +400,10 @@
>   				_metadata->exit_code = KSFT_FAIL; \
>   			} \
>   		} \
> +		else \
> +			jmp = true; \
>   		if (child == 0) { \
> -			if (_metadata->setup_completed && !_metadata->teardown_parent) \
> +			if (_metadata->setup_completed && !_metadata->teardown_parent && !jmp) \
>   				fixture_name##_teardown(_metadata, &self, variant->data); \
>   			_exit(0); \
>   		} \

thanks,
-- Shuah

[PATCH v2] selftests/harness: Prevent infinite loop due to Assert in FIXTURE_TEARDOWN

[Shengyu Li]

This patch addresses an issue in the selftests/harness where an 
assertion within FIXTURE_TEARDOWN could trigger an infinite loop. 
The problem arises because the teardown procedure is meant to 
execute once, but the presence of failing assertions (ASSERT_EQ(0, 1)) 
leads to repeated attempts to execute teardown due to 
the long jump mechanism used by the harness for handling assertions.

To resolve this, the patch ensures that the teardown process 
runs only once, regardless of assertion outcomes, preventing 
the infinite loop and allowing tests to fail.

A simple test demo(test.c):
 #include "kselftest_harness.h"

FIXTURE(f)
{
	int fd;
};

FIXTURE_SETUP(f)
{
	self->fd = 0;
}

FIXTURE_TEARDOWN(f)
{
	TH_LOG("TEARDOWN");
	ASSERT_EQ(0, 1);
	self->fd = -1;
}

TEST_F(f, open_close)
{
	ASSERT_NE(self->fd, 1);
}

TEST_HARNESS_MAIN

will always output the following output due to a dead loop until timeout:
 # test.c:15:open_close:TEARDOWN
 # test.c:16:open_close:Expected 0 (0) == 1 (1)
 # test.c:15:open_close:TEARDOWN
 # test.c:16:open_close:Expected 0 (0) == 1 (1)
 ...

But here's what we should and expect to get:
 TAP version 13
 1..1
 # Starting 1 tests from 2 test cases.
 #  RUN           f.open_close ...
 # test.c:15:open_close:TEARDOWN
 # test.c:16:open_close:Expected 0 (0) == 1 (1)
 # open_close: Test terminated by assertion
 #          FAIL  f.open_close
 not ok 1 f.open_close
 # FAILED: 0 / 1 tests passed.
 # Totals: pass:0 fail:1 xfail:0 xpass:0 skip:0 error:0

also this is related to the issue mentioned in this patch
https://patchwork.kernel.org/project/linux-kselftest/patch/e2ba3f8c-80e6-477d-9cea-1c9af820e0ed@alu.unizg.hr/



Signed-off-by: Shengyu Li <shengyu.li.evgeny@gmail.com>
---
 tools/testing/selftests/kselftest_harness.h | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/kselftest_harness.h b/tools/testing/selftests/kselftest_harness.h
index 4fd735e48ee7..230d62884885 100644
--- a/tools/testing/selftests/kselftest_harness.h
+++ b/tools/testing/selftests/kselftest_harness.h
@@ -383,6 +383,7 @@
 		FIXTURE_DATA(fixture_name) self; \
 		pid_t child = 1; \
 		int status = 0; \
+		bool jmp = false; \
 		memset(&self, 0, sizeof(FIXTURE_DATA(fixture_name))); \
 		if (setjmp(_metadata->env) == 0) { \
 			/* Use the same _metadata. */ \
@@ -399,8 +400,10 @@
 				_metadata->exit_code = KSFT_FAIL; \
 			} \
 		} \
+		else \
+			jmp = true; \
 		if (child == 0) { \
-			if (_metadata->setup_completed && !_metadata->teardown_parent) \
+			if (_metadata->setup_completed && !_metadata->teardown_parent && !jmp) \
 				fixture_name##_teardown(_metadata, &self, variant->data); \
 			_exit(0); \
 		} \
-- 
2.25.1

Re: [PATCH v2] selftests/harness: Prevent infinite loop due to Assert in FIXTURE_TEARDOWN

[Shuah Khan]

On 3/26/24 15:13, Shengyu Li wrote:
> This patch addresses an issue in the selftests/harness where an
> assertion within FIXTURE_TEARDOWN could trigger an infinite loop.
> The problem arises because the teardown procedure is meant to
> execute once, but the presence of failing assertions (ASSERT_EQ(0, 1))
> leads to repeated attempts to execute teardown due to
> the long jump mechanism used by the harness for handling assertions.
> 
> To resolve this, the patch ensures that the teardown process
> runs only once, regardless of assertion outcomes, preventing
> the infinite loop and allowing tests to fail.
> 
> A simple test demo(test.c):
>   #include "kselftest_harness.h"
> 
> FIXTURE(f)
> {
> 	int fd;
> };
> 
> FIXTURE_SETUP(f)
> {
> 	self->fd = 0;
> }
> 
> FIXTURE_TEARDOWN(f)
> {
> 	TH_LOG("TEARDOWN");
> 	ASSERT_EQ(0, 1);
> 	self->fd = -1;
> }
> 
> TEST_F(f, open_close)
> {
> 	ASSERT_NE(self->fd, 1);
> }
> 
> TEST_HARNESS_MAIN
> 
> will always output the following output due to a dead loop until timeout:
>   # test.c:15:open_close:TEARDOWN
>   # test.c:16:open_close:Expected 0 (0) == 1 (1)
>   # test.c:15:open_close:TEARDOWN
>   # test.c:16:open_close:Expected 0 (0) == 1 (1)
>   ...
> 
> But here's what we should and expect to get:
>   TAP version 13
>   1..1
>   # Starting 1 tests from 2 test cases.
>   #  RUN           f.open_close ...
>   # test.c:15:open_close:TEARDOWN
>   # test.c:16:open_close:Expected 0 (0) == 1 (1)
>   # open_close: Test terminated by assertion
>   #          FAIL  f.open_close
>   not ok 1 f.open_close
>   # FAILED: 0 / 1 tests passed.
>   # Totals: pass:0 fail:1 xfail:0 xpass:0 skip:0 error:0
> 
> also this is related to the issue mentioned in this patch
> https://patchwork.kernel.org/project/linux-kselftest/patch/e2ba3f8c-80e6-477d-9cea-1c9af820e0ed@alu.unizg.hr/
> 

Adding people who participated in the above thread discussion.
Please try this patch out to see if this fixes the problem
with iommu test.

> 
> 
> Signed-off-by: Shengyu Li <shengyu.li.evgeny@gmail.com>

Than you for finding and fixing this. Applied now to linux-kselftes fixes
branch for the next rc.

thanks,
-- Shuah