GCC doesn't like dereferencing a pointer set to 0x1 (when building
at -O2):

pkey_sighandler_tests.c:166:9: warning: array subscript 0 is outside array bounds of 'int[0]' [-Warray-bounds=]
  166 |         *(int *) (0x1) = 1;
      |         ^~~~~~~~~~~~~~
cc1: note: source object is likely at address zero

Using NULL instead seems to make it happy. This should make no
difference in practice (SIGSEGV with SEGV_MAPERR will be the outcome
regardless), we just need to update the expected si_addr.

Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
---
 tools/testing/selftests/mm/pkey_sighandler_tests.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/tools/testing/selftests/mm/pkey_sighandler_tests.c b/tools/testing/selftests/mm/pkey_sighandler_tests.c
index c593a426341c..e7b91794f184 100644
--- a/tools/testing/selftests/mm/pkey_sighandler_tests.c
+++ b/tools/testing/selftests/mm/pkey_sighandler_tests.c
@@ -163,7 +163,7 @@ static void *thread_segv_with_pkey0_disabled(void *ptr)
 	__write_pkey_reg(pkey_reg_restrictive_default());
 
 	/* Segfault (with SEGV_MAPERR) */
-	*(int *) (0x1) = 1;
+	*(int *)NULL = 1;
 	return NULL;
 }
 
@@ -179,7 +179,6 @@ static void *thread_segv_pkuerr_stack(void *ptr)
 static void *thread_segv_maperr_ptr(void *ptr)
 {
 	stack_t *stack = ptr;
-	int *bad = (int *)1;
 	u64 pkey_reg;
 
 	/*
@@ -195,7 +194,7 @@ static void *thread_segv_maperr_ptr(void *ptr)
 	__write_pkey_reg(pkey_reg);
 
 	/* Segfault */
-	*bad = 1;
+	*(int *)NULL = 1;
 	syscall_raw(SYS_exit, 0, 0, 0, 0, 0, 0);
 	return NULL;
 }
@@ -234,7 +233,7 @@ static void test_sigsegv_handler_with_pkey0_disabled(void)
 
 	ksft_test_result(siginfo.si_signo == SIGSEGV &&
 			 siginfo.si_code == SEGV_MAPERR &&
-			 siginfo.si_addr == (void *)1,
+			 siginfo.si_addr == NULL,
 			 "%s\n", __func__);
 }
 
@@ -349,7 +348,7 @@ static void test_sigsegv_handler_with_different_pkey_for_stack(void)
 
 	ksft_test_result(siginfo.si_signo == SIGSEGV &&
 			 siginfo.si_code == SEGV_MAPERR &&
-			 siginfo.si_addr == (void *)1,
+			 siginfo.si_addr == NULL,
 			 "%s\n", __func__);
 }
 
-- 
2.47.0

Dereferencing a null pointer on Clang is not a good idea - it will
entirely optimise out the dereference. Make the pointer volatile to
force the access (and fault).

Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202412140850.4TW4YBqc-lkp@intel.com/
Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
---

Hi Andrew,

Another fixup for an issue reported by the LKP CI (building with Clang
instead of GCC). This should be squashed into the patch I'm replying to:
"selftests/mm: Fix -Warray-bounds warnings in pkey_sighandler_tests"

Cheers,
- Kevin

Cc: aruna.ramakrishna@oracle.com
Cc: catalin.marinas@arm.com
Cc: dave.hansen@linux.intel.com
Cc: joey.gouly@arm.com
Cc: keith.lucas@oracle.com
Cc: ryan.roberts@arm.com
Cc: shuah@kernel.org
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-kselftest@vger.kernel.org
Cc: linux-mm@kvack.org
Cc: x86@kernel.org
---
 tools/testing/selftests/mm/pkey_sighandler_tests.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/mm/pkey_sighandler_tests.c b/tools/testing/selftests/mm/pkey_sighandler_tests.c
index 17bbfcd552c6..1ac8c8809880 100644
--- a/tools/testing/selftests/mm/pkey_sighandler_tests.c
+++ b/tools/testing/selftests/mm/pkey_sighandler_tests.c
@@ -161,7 +161,7 @@ static void *thread_segv_with_pkey0_disabled(void *ptr)
 	__write_pkey_reg(pkey_reg_restrictive_default());
 
 	/* Segfault (with SEGV_MAPERR) */
-	*(int *)NULL = 1;
+	*(volatile int *)NULL = 1;
 	return NULL;
 }
 
@@ -192,7 +192,7 @@ static void *thread_segv_maperr_ptr(void *ptr)
 	__write_pkey_reg(pkey_reg);
 
 	/* Segfault */
-	*(int *)NULL = 1;
+	*(volatile int *)NULL = 1;
 	syscall_raw(SYS_exit, 0, 0, 0, 0, 0, 0);
 	return NULL;
 }
-- 
2.47.0