Re: [PATCH v5] riscv: selftests: Fix warnings pointer masking test

Hi Charlie,

On 2024-12-06 11:52 AM, Charlie Jenkins wrote:

When compiling the pointer masking tests with -Wall this warning is present:

pointer_masking.c: In function ‘test_tagged_addr_abi_sysctl’: pointer_masking.c:203:9: warning: ignoring return value of ‘pwrite’ declared with attribute ‘warn_unused_result’ [-Wunused-result] 203 | pwrite(fd, &value, 1, 0); | ^~~~~~~~~~~~~~~~~~~~~~~~ pointer_masking.c:208:9: warning: ignoring return value of ‘pwrite’ declared with attribute ‘warn_unused_result’ [-Wunused-result] 208 | pwrite(fd, &value, 1, 0);

I came across this on riscv64-linux-gnu-gcc (Ubuntu 11.4.0-1ubuntu1~22.04).

Fix this by checking that the number of bytes written equal the expected number of bytes written.

Fixes: 7470b5afd150 ("riscv: selftests: Add a pointer masking test") Signed-off-by: Charlie Jenkins charlie@rivosinc.com

---

Changes in v5:

• No longer skip second pwrite if first one fails
• Use wrapper function instead of goto (Drew)
• Link to v4: https://lore.kernel.org/r/20241205-fix_warnings_pointer_masking_tests-v4-1-0...

Changes in v4:

• Skip sysctl_enabled test if first pwrite failed
• Link to v3: https://lore.kernel.org/r/20241205-fix_warnings_pointer_masking_tests-v3-1-5...

Changes in v3:

• Fix sysctl enabled test case (Drew/Alex)
• Move pwrite err condition into goto (Drew)
• Link to v2: https://lore.kernel.org/r/20241204-fix_warnings_pointer_masking_tests-v2-1-1...

Changes in v2:

• I had ret != 2 for testing, I changed it to be ret != 1.
• Link to v1: https://lore.kernel.org/r/20241204-fix_warnings_pointer_masking_tests-v1-1-e...

---

.../testing/selftests/riscv/abi/pointer_masking.c | 22 ++++++++++++++++------ 1 file changed, 16 insertions(+), 6 deletions(-)

diff --git a/tools/testing/selftests/riscv/abi/pointer_masking.c b/tools/testing/selftests/riscv/abi/pointer_masking.c index dee41b7ee3e3..50c4d1bc7570 100644 --- a/tools/testing/selftests/riscv/abi/pointer_masking.c +++ b/tools/testing/selftests/riscv/abi/pointer_masking.c @@ -185,8 +185,20 @@ static void test_fork_exec(void) } } +static bool pwrite_wrapper(int fd, void *buf, size_t count, const char *msg) +{

• int ret = pwrite(fd, buf, count, 0);
• if (ret != count) {

• ksft_perror(msg);
  

• return false;
  

• }
• return true;

+}

static void test_tagged_addr_abi_sysctl(void) {

• char *err_pwrite_msg = "failed to write to /proc/sys/abi/tagged_addr_disabled\n"; char value; int fd;

@@ -200,14 +212,12 @@ static void test_tagged_addr_abi_sysctl(void) } value = '1';

• pwrite(fd, &value, 1, 0);
• ksft_test_result(set_tagged_addr_ctrl(min_pmlen, true) == -EINVAL,

• 	 "sysctl disabled\n");
  

• if (!pwrite_wrapper(fd, &value, 1, "write '1'"))

• ksft_test_result_fail(err_pwrite_msg);
  

value = '0';

• pwrite(fd, &value, 1, 0);
• ksft_test_result(set_tagged_addr_ctrl(min_pmlen, true) == 0,

• 	 "sysctl enabled\n");
  

• if (!pwrite_wrapper(fd, &value, 1, "write '0'"))

• ksft_test_result_fail(err_pwrite_msg);
  

You've removed the ksft_test_result() calls, which contain the actual behavioral tests (set_tagged_addr_ctrl()), so the test no longer does anything but toggle the sysctl:

# Testing fork/exec behavior ok 55 dereference after fork ok 56 dereference after fork+exec # Testing tagged address ABI sysctl # Testing tagged address ABI ok 57 PMLEN=0 tagged address ABI ok 58 PMLEN=7 tagged address ABI ok 59 PMLEN=16 tagged address ABI # Planned tests != run tests (61 != 59) # Totals: pass:59 fail:0 xfail:0 xpass:0 skip:0 error:0

It does do the right thing if I `mount --bind /dev/full /proc/sys/abi/tagged_addr_disabled`:

# Testing fork/exec behavior ok 55 dereference after fork ok 56 dereference after fork+exec # Testing tagged address ABI sysctl # write '1': No space left on device (28) not ok 57 failed to write to /proc/sys/abi/tagged_addr_disabled # write '0': No space left on device (28) not ok 58 failed to write to /proc/sys/abi/tagged_addr_disabled # Testing tagged address ABI ok 59 PMLEN=0 tagged address ABI ok 60 PMLEN=7 tagged address ABI ok 61 PMLEN=16 tagged address ABI # Totals: pass:59 fail:2 xfail:0 xpass:0 skip:0 error:0

So I guess the "ksft_test_result(set_tagged_addr_ctrl..." needs to go in the else case of the pwrite_wrapper() call.

Regards, Samuel

set_tagged_addr_ctrl(0, false);

---

base-commit: 40384c840ea1944d7c5a392e8975ed088ecf0b37 change-id: 20241204-fix_warnings_pointer_masking_tests-3860e4f35429