[bug report] kselftest/arm64: Add utilities and a test to validate mte memory

[bug report] kselftest/arm64: Add utilities and a test to validate mte memory

[Dan Carpenter]

Hello Amit Daniel Kachhap,

The patch e9b60476bea0: "kselftest/arm64: Add utilities and a test to
validate mte memory" from Oct 2, 2020, leads to the following Smatch
static checker warning:

	./tools/testing/selftests/arm64/mte/mte_common_util.c:336 mte_default_setup()
	warn: bitwise AND condition is false here

./tools/testing/selftests/arm64/mte/mte_common_util.c
    316 int mte_default_setup(void)
    317 {
    318         unsigned long hwcaps2 = getauxval(AT_HWCAP2);
    319         unsigned long en = 0;
    320         int ret;
    321 
    322         if (!(hwcaps2 & HWCAP2_MTE)) {
    323                 ksft_print_msg("SKIP: MTE features unavailable\n");
    324                 return KSFT_SKIP;
    325         }
    326         /* Get current mte mode */
    327         ret = prctl(PR_GET_TAGGED_ADDR_CTRL, en, 0, 0, 0);
    328         if (ret < 0) {
    329                 ksft_print_msg("FAIL:prctl PR_GET_TAGGED_ADDR_CTRL with error =%d\n", ret);
    330                 return KSFT_FAIL;
    331         }
    332         if (ret & PR_MTE_TCF_SYNC)
    333                 mte_cur_mode = MTE_SYNC_ERR;
    334         else if (ret & PR_MTE_TCF_ASYNC)
    335                 mte_cur_mode = MTE_ASYNC_ERR;
--> 336         else if (ret & PR_MTE_TCF_NONE)

It looks like the intent was to make PR_MTE_TCF_NONE into an ifdef
configurable thing but that never happened?

    337                 mte_cur_mode = MTE_NONE_ERR;
    338 
    339         mte_cur_pstate_tco = mte_get_pstate_tco();
    340         /* Disable PSTATE.TCO */
    341         mte_disable_pstate_tco();
    342         return 0;
    343 }

regards,
dan carpenter

Re: [bug report] kselftest/arm64: Add utilities and a test to validate mte memory

[Amit Kachhap]

Hi,

On 7/18/22 20:17, Dan Carpenter wrote:
> Hello Amit Daniel Kachhap,
> 
> The patch e9b60476bea0: "kselftest/arm64: Add utilities and a test to
> validate mte memory" from Oct 2, 2020, leads to the following Smatch
> static checker warning:
> 
> 	./tools/testing/selftests/arm64/mte/mte_common_util.c:336 mte_default_setup()
> 	warn: bitwise AND condition is false here
> 
> ./tools/testing/selftests/arm64/mte/mte_common_util.c
>      316 int mte_default_setup(void)
>      317 {
>      318         unsigned long hwcaps2 = getauxval(AT_HWCAP2);
>      319         unsigned long en = 0;
>      320         int ret;
>      321
>      322         if (!(hwcaps2 & HWCAP2_MTE)) {
>      323                 ksft_print_msg("SKIP: MTE features unavailable\n");
>      324                 return KSFT_SKIP;
>      325         }
>      326         /* Get current mte mode */
>      327         ret = prctl(PR_GET_TAGGED_ADDR_CTRL, en, 0, 0, 0);
>      328         if (ret < 0) {
>      329                 ksft_print_msg("FAIL:prctl PR_GET_TAGGED_ADDR_CTRL with error =%d\n", ret);
>      330                 return KSFT_FAIL;
>      331         }
>      332         if (ret & PR_MTE_TCF_SYNC)
>      333                 mte_cur_mode = MTE_SYNC_ERR;
>      334         else if (ret & PR_MTE_TCF_ASYNC)
>      335                 mte_cur_mode = MTE_ASYNC_ERR;
> --> 336         else if (ret & PR_MTE_TCF_NONE)
> 
> It looks like the intent was to make PR_MTE_TCF_NONE into an ifdef
> configurable thing but that never happened?

Thanks for pointing this out. I will send the proper fix with
the Reported bys .

Regards,
Amit
> 
>      337                 mte_cur_mode = MTE_NONE_ERR;
>      338
>      339         mte_cur_pstate_tco = mte_get_pstate_tco();
>      340         /* Disable PSTATE.TCO */
>      341         mte_disable_pstate_tco();
>      342         return 0;
>      343 }
> 
> regards,
> dan carpenter