06:30 <superprower> Greetings! I'm trying to debug one issue on A64-based board - I am unable to use CLOCK_REALTIME_ALARM clockid as an argument for clock_gettime syscall (return value is -1, EINVAL). There was a guess that it may be due to missing hardware RTC, but it is present, configured in device tree and I am able to see /dev/rtc0 device and there is valid /proc/driver/rtc entry and there is entry is /sys. Where should I look? I tried adding some printk's to

06:30 <superprower> SYSCALL_DEFINE at kernel/time/posix-timers.c, but it doesn't print in dmesg. How should I debug this? Thanks in advance.

06:32 <superprower> Kernel version 5.1-rc7, by the way. I've read some mailing lists, but they mostly discussed return value, not the condition that enabled this clockid.

08:08 <superprower> Maybe anyone can point me to the place where clock_gettime is implemented for aarch64? There is a arch/arm64/kernel/vdso/gettimeofday.S that has a jumptable, and if I understand correctly for CLOCK_REALTIME_ALARM it fallback's to __NR_clock_gettime which is supposedly made into __arm64_sys_clock_gettime by the macro, but that's the end of the trail for me, I can only grep this in System.map and object files. Don't know what exactly is executed.

10:09 <MoeIcenowy> superprower: I referred to man page

10:09 <MoeIcenowy> does your process have the necessary capability

10:36 <superprower> MoeIcenowy: I'll double check that, but I just checked on another aarch64 device with 'getcap' and binary that makes that sytem call doesn't have any capabilities whatsoever, yet it works.

11:20 <z3ntu> jaganteki: did you try dsi on the A64 yet with an 5.2-rc kernel?

11:25 <jaganteki> zentu: yes I have rebased.

11:25 <jaganteki> what is the issue?

13:41 <z3ntu> jaganteki: The "display blinking" thing I wrote about yesterday