[v5] Fix Kselftest's vfork() side effects

[PATCH v5 10/10] selftests/harness: Handle TEST_F()'s explicit exit codes

Posted by Mickaël Salaün 1 year, 9 months ago

If TEST_F() explicitly calls exit(code) with code different than 0, then
_metadata->exit_code is set to this code (e.g. KVM_ONE_VCPU_TEST()).  We
need to keep in mind that _metadata->exit_code can be KSFT_SKIP while
the process exit code is 0.

Initial patch written by Sean Christopherson [1].

Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Mark Brown <broonie@kernel.org>
Cc: Sean Christopherson <seanjc@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Will Drewry <wad@chromium.org>
Link: https://lore.kernel.org/r/ZjPelW6-AbtYvslu@google.com [1]
Fixes: 0710a1a73fb4 ("selftests/harness: Merge TEST_F_FORK() into TEST_F()")
Signed-off-by: Mickaël Salaün <mic@digikod.net>
Link: https://lore.kernel.org/r/20240503105820.300927-11-mic@digikod.net
---

Changes since v4:
* Check abort status when the grandchild exited.
* Keep the _exit(0) calls because _metadata->exit_code is always
  checked.
* Only set _metadata->exit_code to WEXITSTATUS() if it is not zero.

Changes since v3:
* New patch mainly from Sean Christopherson.
---
 tools/testing/selftests/kselftest_harness.h | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/kselftest_harness.h b/tools/testing/selftests/kselftest_harness.h
index eb25f7c11949..7612bf09c5f8 100644
--- a/tools/testing/selftests/kselftest_harness.h
+++ b/tools/testing/selftests/kselftest_harness.h
@@ -462,9 +462,13 @@ static inline pid_t clone3_vfork(void)
 		munmap(teardown, sizeof(*teardown)); \
 		if (self && fixture_name##_teardown_parent) \
 			munmap(self, sizeof(*self)); \
-		if (!WIFEXITED(status) && WIFSIGNALED(status)) \
+		if (WIFEXITED(status)) { \
+			if (WEXITSTATUS(status)) \
+				_metadata->exit_code = WEXITSTATUS(status); \
+		} else if (WIFSIGNALED(status)) { \
 			/* Forward signal to __wait_for_test(). */ \
 			kill(getpid(), WTERMSIG(status)); \
+		} \
 		__test_check_assert(_metadata); \
 	} \
 	static void __attribute__((constructor)) \
-- 
2.45.0

Re: [PATCH v5 10/10] selftests/harness: Handle TEST_F()'s explicit exit codes

Posted by Sean Christopherson 1 year, 9 months ago

On Fri, May 03, 2024, Mickaël Salaün wrote:
> If TEST_F() explicitly calls exit(code) with code different than 0, then
> _metadata->exit_code is set to this code (e.g. KVM_ONE_VCPU_TEST()).  We
> need to keep in mind that _metadata->exit_code can be KSFT_SKIP while
> the process exit code is 0.
> 
> Initial patch written by Sean Christopherson [1].

Heh, my pseudo patch barely has any relevance at this point.  How about replacing
that with:

  Reported-by: Sean Christopherson <seanjc@google.com>
  Closes: https://lore.kernel.org/r/ZjPelW6-AbtYvslu@google.com

> Cc: Jakub Kicinski <kuba@kernel.org>
> Cc: Kees Cook <keescook@chromium.org>
> Cc: Mark Brown <broonie@kernel.org>
> Cc: Sean Christopherson <seanjc@google.com>
> Cc: Shuah Khan <shuah@kernel.org>
> Cc: Will Drewry <wad@chromium.org>
> Link: https://lore.kernel.org/r/ZjPelW6-AbtYvslu@google.com [1]
> Fixes: 0710a1a73fb4 ("selftests/harness: Merge TEST_F_FORK() into TEST_F()")
> Signed-off-by: Mickaël Salaün <mic@digikod.net>
> Link: https://lore.kernel.org/r/20240503105820.300927-11-mic@digikod.net
> ---
> 
> Changes since v4:
> * Check abort status when the grandchild exited.
> * Keep the _exit(0) calls because _metadata->exit_code is always
>   checked.
> * Only set _metadata->exit_code to WEXITSTATUS() if it is not zero.
> 
> Changes since v3:
> * New patch mainly from Sean Christopherson.
> ---
>  tools/testing/selftests/kselftest_harness.h | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/tools/testing/selftests/kselftest_harness.h b/tools/testing/selftests/kselftest_harness.h
> index eb25f7c11949..7612bf09c5f8 100644
> --- a/tools/testing/selftests/kselftest_harness.h
> +++ b/tools/testing/selftests/kselftest_harness.h
> @@ -462,9 +462,13 @@ static inline pid_t clone3_vfork(void)
>  		munmap(teardown, sizeof(*teardown)); \
>  		if (self && fixture_name##_teardown_parent) \
>  			munmap(self, sizeof(*self)); \
> -		if (!WIFEXITED(status) && WIFSIGNALED(status)) \
> +		if (WIFEXITED(status)) { \
> +			if (WEXITSTATUS(status)) \
> +				_metadata->exit_code = WEXITSTATUS(status); \

Ah, IIUC, this works because __run_test() effectively forwards the exit_code?

	} else if (t->pid == 0) {
		setpgrp();
		t->fn(t, variant);
		_exit(t->exit_code);
	}

Tested-by: Sean Christopherson <seanjc@google.com>

> +		} else if (WIFSIGNALED(status)) { \
>  			/* Forward signal to __wait_for_test(). */ \
>  			kill(getpid(), WTERMSIG(status)); \
> +		} \
>  		__test_check_assert(_metadata); \
>  	} \
>  	static void __attribute__((constructor)) \
> -- 
> 2.45.0
>

Re: [PATCH v5 10/10] selftests/harness: Handle TEST_F()'s explicit exit codes

Posted by Mickaël Salaün 1 year, 9 months ago

On Fri, May 03, 2024 at 07:17:16AM GMT, Sean Christopherson wrote:
> On Fri, May 03, 2024, Mickaël Salaün wrote:
> > If TEST_F() explicitly calls exit(code) with code different than 0, then
> > _metadata->exit_code is set to this code (e.g. KVM_ONE_VCPU_TEST()).  We
> > need to keep in mind that _metadata->exit_code can be KSFT_SKIP while
> > the process exit code is 0.
> > 
> > Initial patch written by Sean Christopherson [1].
> 
> Heh, my pseudo patch barely has any relevance at this point.  How about replacing
> that with:
> 
>   Reported-by: Sean Christopherson <seanjc@google.com>
>   Closes: https://lore.kernel.org/r/ZjPelW6-AbtYvslu@google.com
> 
> > Cc: Jakub Kicinski <kuba@kernel.org>
> > Cc: Kees Cook <keescook@chromium.org>
> > Cc: Mark Brown <broonie@kernel.org>
> > Cc: Sean Christopherson <seanjc@google.com>
> > Cc: Shuah Khan <shuah@kernel.org>
> > Cc: Will Drewry <wad@chromium.org>
> > Link: https://lore.kernel.org/r/ZjPelW6-AbtYvslu@google.com [1]
> > Fixes: 0710a1a73fb4 ("selftests/harness: Merge TEST_F_FORK() into TEST_F()")
> > Signed-off-by: Mickaël Salaün <mic@digikod.net>
> > Link: https://lore.kernel.org/r/20240503105820.300927-11-mic@digikod.net
> > ---
> > 
> > Changes since v4:
> > * Check abort status when the grandchild exited.
> > * Keep the _exit(0) calls because _metadata->exit_code is always
> >   checked.
> > * Only set _metadata->exit_code to WEXITSTATUS() if it is not zero.
> > 
> > Changes since v3:
> > * New patch mainly from Sean Christopherson.
> > ---
> >  tools/testing/selftests/kselftest_harness.h | 6 +++++-
> >  1 file changed, 5 insertions(+), 1 deletion(-)
> > 
> > diff --git a/tools/testing/selftests/kselftest_harness.h b/tools/testing/selftests/kselftest_harness.h
> > index eb25f7c11949..7612bf09c5f8 100644
> > --- a/tools/testing/selftests/kselftest_harness.h
> > +++ b/tools/testing/selftests/kselftest_harness.h
> > @@ -462,9 +462,13 @@ static inline pid_t clone3_vfork(void)
> >  		munmap(teardown, sizeof(*teardown)); \
> >  		if (self && fixture_name##_teardown_parent) \
> >  			munmap(self, sizeof(*self)); \
> > -		if (!WIFEXITED(status) && WIFSIGNALED(status)) \
> > +		if (WIFEXITED(status)) { \
> > +			if (WEXITSTATUS(status)) \
> > +				_metadata->exit_code = WEXITSTATUS(status); \
> 
> Ah, IIUC, this works because __run_test() effectively forwards the exit_code?
> 
> 	} else if (t->pid == 0) {
> 		setpgrp();
> 		t->fn(t, variant);
> 		_exit(t->exit_code);
> 	}

Yes

> 
> Tested-by: Sean Christopherson <seanjc@google.com>

OK, I'll send a v6. We really need to get this into -next.

> 
> > +		} else if (WIFSIGNALED(status)) { \
> >  			/* Forward signal to __wait_for_test(). */ \
> >  			kill(getpid(), WTERMSIG(status)); \
> > +		} \
> >  		__test_check_assert(_metadata); \
> >  	} \
> >  	static void __attribute__((constructor)) \
> > -- 
> > 2.45.0
> > 
>