Skip to contentSkip to conversation

syscall02: add invalid syscall number test

PATCH In review Andrea Cervesato 4 participants 6 messages Jun 29 – Jun 30

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 syscall02: add invalid syscall number test

From: Andrea Cervesato <[email protected]>

Add a test verifying that syscall() returns ENOSYS when called with unimplemented syscall numbers. The chosen numbers stay above the syscall table and below the architecture-specific private ranges, so they are reliably invalid on every supported architecture.

---
 runtest/syscalls                              |  1 +
 testcases/kernel/syscalls/syscall/.gitignore  |  1 +
 testcases/kernel/syscalls/syscall/syscall02.c | 47 +++++++++++++++++++++++++++
 3 files changed, 49 insertions(+)
runtest/syscalls +1 -0
index a021c79da49d4157c33187caed2dd3f0483759e0..8a777e61da1af59ba694e0bc5b5d92bb1d63754a 100644
@@ -1678,6 +1678,7 @@ sync_file_range01 sync_file_range01
 sync_file_range02 sync_file_range02
 
 syscall01 syscall01
+syscall02 syscall02
 
 sysconf01 sysconf01
 
testcases/kernel/syscalls/syscall/.gitignore +1 -0
index 74a081e40cb564e874e0f81478d3d1a0e95b6df4..fed1d81079450a7442023447285cd112924f7a10 100644
@@ -1 +1,2 @@
 /syscall01
+/syscall02
testcases/kernel/syscalls/syscall/syscall02.c new file +47 -0

47 lines not on this page · raw message

---
base-commit: ad68429b9cfd55b233733e45e600ffafa1458ac5
change-id: 20260629-invalid_syscalls-0476c52d7d0c

Best regards,

Signature · 5 lines
Andrea Cervesato <[email protected]>


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

Hi Andrea,

Quoted 1 line · Andrea Cervesato

syscall02: add invalid syscall number test

Verdict - Reviewed

---
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
In reply to Andrea Cervesato
Quoted 64 lines · Andrea Cervesato

From: Andrea Cervesato <[email protected]>

Add a test verifying that syscall() returns ENOSYS when called with unimplemented syscall numbers. The chosen numbers stay above the syscall table and below the architecture-specific private ranges, so they are reliably invalid on every supported architecture.

---
 runtest/syscalls                              |  1 +
 testcases/kernel/syscalls/syscall/.gitignore  |  1 +
 testcases/kernel/syscalls/syscall/syscall02.c | 47 +++++++++++++++++++++++++++
 3 files changed, 49 insertions(+)
runtest/syscalls +1 -0
index a021c79da49d4157c33187caed2dd3f0483759e0..8a777e61da1af59ba694e0bc5b5d92bb1d63754a 100644
@@ -1678,6 +1678,7 @@ sync_file_range01 sync_file_range01
 sync_file_range02 sync_file_range02
 
 syscall01 syscall01
+syscall02 syscall02
 
 sysconf01 sysconf01
 
testcases/kernel/syscalls/syscall/.gitignore +1 -0
index 74a081e40cb564e874e0f81478d3d1a0e95b6df4..fed1d81079450a7442023447285cd112924f7a10 100644
@@ -1 +1,2 @@
 /syscall01
+/syscall02
testcases/kernel/syscalls/syscall/syscall02.c new file +25 -0
new file mode 100644
index 0000000000000000000000000000000000000000..3bd9fa8cdcca257e03b358749ee7a2b676d8720d
@@ -0,0 +1,47 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (c) 2026 Andrea Cervesato <[email protected]>
+ */
+
+/*\
+ * Verify that :manpage:`syscall(2)` fails with ENOSYS when invoked with
+ * invalid syscall numbers.
+ *
+ * The syscall number is dispatched as an unsigned value on Linux. A value is
+ * reliably unimplemented on every supported architecture when it is:
+ *
+ * - above the syscall table
+ * - below the architecture-specific private syscall ranges (e.g. ARM's
+ *   0x0f0000 base)
+ * - without the x32 ABI bit (0x40000000) set
+ *
+ * Such a number is routed to sys_ni_syscall and returns ENOSYS.
+ */
+
+#include <unistd.h>
+#include <sys/syscall.h>
+
+#include "tst_test.h"
+

testcases/kernel/syscalls/syscall/syscall02.c line 35

Quoted 10 lines
testcases/kernel/syscalls/syscall/syscall02.c new file +10 -0
+#define TC(x) {.scno = (long)(x), .desc = #x}
+
+static struct tcase {
+	long scno;
+	const char *desc;
+} tcases[] = {
+	TC(0x8000),
+	TC(0xf000),
+	TC(0xfffe),
+};

I suggest add two more cases check:
1) -1
2) __NR_syscalls +1 (Base arch/powerpc/kernel/syscall.c)

Quoted 24 lines
testcases/kernel/syscalls/syscall/syscall02.c new file +12 -1
+
+static void verify_syscall(unsigned int n)
+{
+	struct tcase *tc = &tcases[n];
+
+	TST_EXP_FAIL2(syscall(tc->scno), ENOSYS, "syscall(%s)", tc->desc);
+}
+
+static struct tst_test test = {
+	.test = verify_syscall,
+	.tcnt = ARRAY_SIZE(tcases),
+};
 
---

base-commit: ad68429b9cfd55b233733e45e600ffafa1458ac5
change-id: 20260629-invalid_syscalls-0476c52d7d0c

Best regards,

Signature · 5 lines
Andrea Cervesato <[email protected]>


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

Hi Andrea, Wei,

Quoted 4 lines
Quoted 4 lines

Add a test verifying that syscall() returns ENOSYS when called with unimplemented syscall numbers. The chosen numbers stay above the syscall table and below the architecture-specific private ranges, so they are reliably invalid on every supported architecture.

+1

...

Quoted 33 lines
Quoted 33 lines

+++ b/testcases/kernel/syscalls/syscall/syscall02.c

testcases/kernel/syscalls/syscall/syscall02.c new file +31 -0
@@ -0,0 +1,47 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (c) 2026 Andrea Cervesato <[email protected]>
+ */
+
+/*\
+ * Verify that :manpage:`syscall(2)` fails with ENOSYS when invoked with
+ * invalid syscall numbers.
+ *
+ * The syscall number is dispatched as an unsigned value on Linux. A value is
+ * reliably unimplemented on every supported architecture when it is:
+ *
+ * - above the syscall table
+ * - below the architecture-specific private syscall ranges (e.g. ARM's
+ *   0x0f0000 base)
+ * - without the x32 ABI bit (0x40000000) set
+ *
+ * Such a number is routed to sys_ni_syscall and returns ENOSYS.
+ */
+
+#include <unistd.h>
+#include <sys/syscall.h>
+
+#include "tst_test.h"
+
+#define TC(x) {.scno = (long)(x), .desc = #x}
+
+static struct tcase {
+	long scno;
+	const char *desc;
+} tcases[] = {
Quoted 8 lines

testcases/kernel/syscalls/syscall/syscall02.c line 35

Quoted 4 lines
testcases/kernel/syscalls/syscall/syscall02.c new file +4 -0
+	TC(0x8000),
+	TC(0xf000),
+	TC(0xfffe),
+};

I suggest add two more cases check:
1) -1
2) __NR_syscalls +1 (Base arch/powerpc/kernel/syscall.c)

+1. Something like this wouldn't harm (tested only on x86_64 and x86).

Kind regards,
Petr

+++ testcases/kernel/syscalls/syscall/syscall02.c
@@ -14,12 +14,15 @@
* - below the architecture-specific private syscall ranges (e.g. ARM's
* 0x0f0000 base)
* - without the x32 ABI bit (0x40000000) set
+ * - -1
+ * - __NR_syscalls + 1
*
* Such a number is routed to sys_ni_syscall and returns ENOSYS. */

#include <unistd.h>
#include <sys/syscall.h>
+#include <asm-generic/unistd.h>

#include "tst_test.h"

@@ -32,6 +35,8 @@ static struct tcase {
TC(0x8000),
TC(0xf000),
TC(0xfffe),
+ TC(-1),
+ TC(__NR_syscalls + 1),
};

static void verify_syscall(unsigned int n)

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

Hi Petr,

Quoted 4 lines
Quoted 3 lines

I suggest add two more cases check:
1) -1
2) __NR_syscalls +1 (Base arch/powerpc/kernel/syscall.c)

that was my initial idea, but different architectures are starting from a different value and I wanted to make it generic. I guess the right way would be to include syscalls.h and use __LTP__NR_INVALID_SYSCALL at this point. WDYT?

Regards,
--
Andrea Cervesato
SUSE QE Automation Engineer Linux
[email protected]

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

Hi Andrea,

Quoted 11 lines

Hi Petr,

Quoted 3 lines
Quoted 3 lines

I suggest add two more cases check:
1) -1
2) __NR_syscalls +1 (Base arch/powerpc/kernel/syscall.c)

that was my initial idea, but different architectures are starting from a different value and I wanted to make it generic. I guess the right way would be to include syscalls.h and use __LTP__NR_INVALID_SYSCALL at this point. WDYT?

Yes, __LTP__NR_INVALID_SYSCALL is better than -1 (for all archs we have #define __LTP__NR_INVALID_SYSCALL -1)

But I'd also use __NR_syscalls from <asm-generic/unistd.h> (the second suggested case), i.e. have arch dependent value based on UAPI headers.

Also, you could add a real description to the values to .desc.

Kind regards,
Petr

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