[LTP] [PATCH v4] Correctly check if PIDFD_INFO_EXIT is available

[LTP] [PATCH v4] Correctly check if PIDFD_INFO_EXIT is available

[Andrea Cervesato]

From: Andrea Cervesato <andrea.cervesato@suse.com>

When systems are not having the PIDFD_INFO_EXIT implementation,
ioctl_pidfd testing suite might fail with:

	ioctl_pidfd.h:32: TBROK: ioctl(...) failed: ENOTTY (25)

Fix the ioctl_pidfd_info_exit_supported() behavior, considering ENOTTY,
EINVAL and ESRCH errors like a sign for not having PIDFD_INFO_EXIT
implemented in the system.

Reviewed-by: Cyril Hrubis <chrubis@suse.cz>
Reviewed-by: Avinesh Kumar <akumar@suse.de>
Tested-by: Avinesh Kumar <akumar@suse.de>
Signed-off-by: Andrea Cervesato <andrea.cervesato@suse.com>
---
Changes in v4:
- use the right AND logic...
- Link to v3: https://lore.kernel.org/r/20250730-ioctl_pidfd_supported-v3-1-f40c047b29de@suse.com

Changes in v3:
- take in consideration also EINVAL and ESRCH errors
- Link to v2: https://lore.kernel.org/r/20250729-ioctl_pidfd_supported-v2-1-f9ef90134138@suse.com

Changes in v2:
- close pidfd when PIDFD_INFO_EXIT is supported
- Link to v1: https://lore.kernel.org/r/20250729-ioctl_pidfd_supported-v1-1-985626a0c46b@suse.com
---
 testcases/kernel/syscalls/ioctl/ioctl_pidfd.h | 24 ++++++++++++++++++------
 1 file changed, 18 insertions(+), 6 deletions(-)

diff --git a/testcases/kernel/syscalls/ioctl/ioctl_pidfd.h b/testcases/kernel/syscalls/ioctl/ioctl_pidfd.h
index 8249ac753cf7fb8a3b749d55c7f0c3b30482c114..811f969cdee8147dfa43bc2ab1124dfd9fd21e41 100644
--- a/testcases/kernel/syscalls/ioctl/ioctl_pidfd.h
+++ b/testcases/kernel/syscalls/ioctl/ioctl_pidfd.h
@@ -11,9 +11,10 @@
 
 static inline int ioctl_pidfd_info_exit_supported(void)
 {
-	int ret = 0;
+	int ret;
 	pid_t pid;
 	int pidfd;
+	int supported = 0;
 	struct pidfd_info info;
 
 	if (tst_kvercmp(6, 15, 0) >= 0)
@@ -29,13 +30,24 @@ static inline int ioctl_pidfd_info_exit_supported(void)
 	pidfd = SAFE_PIDFD_OPEN(pid, 0);
 	SAFE_WAITPID(pid, NULL, 0);
 
-	SAFE_IOCTL(pidfd, PIDFD_GET_INFO, &info);
-	SAFE_CLOSE(pidfd);
+	ret = ioctl(pidfd, PIDFD_GET_INFO, &info);
+	if (ret == -1) {
+		/* - ENOTTY: old kernels not implementing fs/pidfs.c:pidfd_ioctl
+		 * - EINVAL: until v6.13 kernel
+		 * - ESRCH: all kernels between v6.13 and v6.15
+		 */
+		if (errno != ENOTTY &&
+			errno != EINVAL &&
+			errno != ESRCH)
+			tst_brk(TBROK | TERRNO, "ioctl error");
+	} else {
+		if (info.mask & PIDFD_INFO_EXIT)
+			supported = 1;
+	}
 
-	if (info.mask & PIDFD_INFO_EXIT)
-		ret = 1;
+	SAFE_CLOSE(pidfd);
 
-	return ret;
+	return supported;
 }
 
 #endif

---
base-commit: 890d6d64a235548cec559ffaf2e9208b80e6902b
change-id: 20250729-ioctl_pidfd_supported-437d25a4c382

Best regards,
-- 
Andrea Cervesato <andrea.cervesato@suse.com>


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH v4] Correctly check if PIDFD_INFO_EXIT is available

[Cyril Hrubis]

Hi!
> When systems are not having the PIDFD_INFO_EXIT implementation,
> ioctl_pidfd testing suite might fail with:
> 
> 	ioctl_pidfd.h:32: TBROK: ioctl(...) failed: ENOTTY (25)
> 
> Fix the ioctl_pidfd_info_exit_supported() behavior, considering ENOTTY,
> EINVAL and ESRCH errors like a sign for not having PIDFD_INFO_EXIT
> implemented in the system.


>  	if (tst_kvercmp(6, 15, 0) >= 0)
> @@ -29,13 +30,24 @@ static inline int ioctl_pidfd_info_exit_supported(void)
>  	pidfd = SAFE_PIDFD_OPEN(pid, 0);
>  	SAFE_WAITPID(pid, NULL, 0);
>  
> -	SAFE_IOCTL(pidfd, PIDFD_GET_INFO, &info);
> -	SAFE_CLOSE(pidfd);
> +	ret = ioctl(pidfd, PIDFD_GET_INFO, &info);
> +	if (ret == -1) {
> +		/* - ENOTTY: old kernels not implementing fs/pidfs.c:pidfd_ioctl

Okay that's what we get if there is no ioctl() handler for the pidfs.

> +		 * - EINVAL: until v6.13 kernel

And this is supposedly before PIDFD_GET_INFO was added.

> +		 * - ESRCH: all kernels between v6.13 and v6.15

I suppose what happens between 6.13 and 6.15 is that PIDFD_GET_INFO is
acutally implemented but all the data about the process are gone once it
has been waited for. I was thinking that we could eliminate this branch
if we waited for the process after the ioctl() but we need the actual
process to exit since otherwise the PIFD_INFO_EXIT flag woudln't be set
in the result even if it's supported. Sigh.


So I guess that we cannot make this any simpler.

-- 
Cyril Hrubis
chrubis@suse.cz

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH v4] Correctly check if PIDFD_INFO_EXIT is available

[Andrea Cervesato via ltp]

Merged! Thanks.

- Andrea


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH v4] Correctly check if PIDFD_INFO_EXIT is available

[Petr Vorel]

Hi Andrea, all,

> When systems are not having the PIDFD_INFO_EXIT implementation,
> ioctl_pidfd testing suite might fail with:

> 	ioctl_pidfd.h:32: TBROK: ioctl(...) failed: ENOTTY (25)

> Fix the ioctl_pidfd_info_exit_supported() behavior, considering ENOTTY,
> EINVAL and ESRCH errors like a sign for not having PIDFD_INFO_EXIT
> implemented in the system.

Thanks for doing the investigation and fixing this!

...
> -	SAFE_IOCTL(pidfd, PIDFD_GET_INFO, &info);
> -	SAFE_CLOSE(pidfd);
> +	ret = ioctl(pidfd, PIDFD_GET_INFO, &info);
> +	if (ret == -1) {
> +		/* - ENOTTY: old kernels not implementing fs/pidfs.c:pidfd_ioctl
> +		 * - EINVAL: until v6.13 kernel
> +		 * - ESRCH: all kernels between v6.13 and v6.15
> +		 */
> +		if (errno != ENOTTY &&
> +			errno != EINVAL &&
> +			errno != ESRCH)
> +			tst_brk(TBROK | TERRNO, "ioctl error");
> +	} else {
> +		if (info.mask & PIDFD_INFO_EXIT)

nit: this could have been:

	} else if (info.mask & PIDFD_INFO_EXIT) {

Kind regards,
Petr

> +			supported = 1;
> +	}

> -	if (info.mask & PIDFD_INFO_EXIT)
> -		ret = 1;

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp