Re: [PATCH] selftests: breakpoints: check RTC wakeup alarm support before test

[Xinyu Zheng <zhengxinyu6@huawei.com>]

On 11/20/2025 6:15 AM, Shuah Khan wrote:
> On 11/11/25 23:15, Xinyu Zheng wrote:
>> If RTC wakeup alarm feature is unsupported, this testcase may cause
> 
> It may cause infinite suspend? Doesn't sound definitive.> infinite 
> suspend if there is no other wakeup source. To solve this
>> problem, set wakeup alarm up before we trigger suspend. In this case,
>> we can test if RTC support RTC_FEATURE_ALARM and efi_set_alarm function.
>>
> 
> For a "may cause" problem, this change> Signed-off-by: Xinyu Zheng 
> <zhengxinyu6@huawei.com>
>> --->   .../breakpoints/step_after_suspend_test.c     | 23 ++++++++++++ 
>> +++++++
>>   1 file changed, 23 insertions(+)
>>
>> diff --git a/tools/testing/selftests/breakpoints/ 
>> step_after_suspend_test.c b/tools/testing/selftests/breakpoints/ 
>> step_after_suspend_test.c
>> index 8d233ac95696..e738af896ce1 100644
>> --- a/tools/testing/selftests/breakpoints/step_after_suspend_test.c
>> +++ b/tools/testing/selftests/breakpoints/step_after_suspend_test.c
>> @@ -13,6 +13,8 @@
>>   #include <stdio.h>
>>   #include <string.h>
>>   #include <unistd.h>
>> +#include <linux/rtc.h>
>> +#include <sys/ioctl.h>
>>   #include <sys/ptrace.h>
>>   #include <sys/stat.h>
>>   #include <sys/timerfd.h>
>> @@ -159,10 +161,30 @@ void suspend(void)
>>       int count_before;
>>       int count_after;
>>       struct itimerspec spec = {};
>> +    char *rtc_file = "/dev/rtc0";
>> +    int rtc_fd;
>> +    struct rtc_wkalrm alarm = { 0 };
>> +    time_t secs;
>>       if (getuid() != 0)
>>           ksft_exit_skip("Please run the test as root - Exiting.\n");
>> +    rtc_fd = open(rtc_file, O_RDONLY);
>> +    if (rtc_fd < 0)
>> +        ksft_exit_fail_msg("open rtc0 failed\n");
>> +
>> +    err = ioctl(rtc_fd, RTC_RD_TIME, &alarm.time);
>> +    if (err < 0)
>> +        ksft_exit_fail_msg("get rtc time failed\n");
>> +
>> +    secs = timegm((struct tm *)&alarm.time) + 3;
>> +    gmtime_r(&secs, (struct tm *)&alarm.time);
>> +    alarm.enabled = 1;
>> +
>> +    err = ioctl(rtc_fd, RTC_WKALM_SET, &alarm);
>> +    if (err < 0)
>> +        ksft_exit_fail_msg("set wake alarm test failed, errno %d\n", 
>> errno);
> 
> Essentially with this change the test doesn't run unless
> RTC wake is supported. That sounds restrictive - is there
> another way do do this?

Hi, Shuah.

If RTC wake is a must in this testcase?

In my test, when EFI report supporting EFI_RT_SUPPORTED_SET_WAKEUP_TIME.
There will be a alarmtimer.0.auto under /sys/class/rtc/rtc0/，while 
suspending, I found alarmtimer will also been suspended and pass the 
wakeup job to RTC. When RTC calls efi_set_alarm(), this function returns 
EFI_UNSUPPORTED then suspend fails.

When EFI doesn't report supporting EFI_RT_SUPPORTED_SET_WAKEUP_TIME, 
there is no /sys/class/rtc/rtc0/alarmtimer.0.auto and alarmtimer is not 
a suspend target. Then suspend success, but never wakeup.

> 
>> +
>>       timerfd = timerfd_create(CLOCK_BOOTTIME_ALARM, 0);
>>       if (timerfd < 0)
>>           ksft_exit_fail_msg("timerfd_create() failed\n");
>> @@ -180,6 +202,7 @@ void suspend(void)
>>       if (count_after <= count_before)
>>           ksft_exit_fail_msg("Failed to enter Suspend state\n");
>> +    close(rtc_fd);
>>       close(timerfd);
>>   }
>>
> thanks,
> -- Shuah

-- 
Xinyu Zheng