code review: help implementing clock_montonic

Tue Jun 8 15:19:54 UTC 2021

 ping

Sent with ProtonMail Secure Email.

‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐

On Saturday, May 15th, 2021 at 9:22 PM, zack_on_the_speed_chanel <zack_on_the_speed_chanel at protonmail.ch> wrote:

> Use helper functions in rtems/timespec.h or score/timespec.h
>
> > (depending where you implement your code, in this case, you probably
> >
> > use the score services).
>
> I did it using rtems subtract function.
>
> rtems_timespec_subtract (now, expire, result);
>
> where now expire and result are all timespec
>
> _TOD_Get(now); or _Timecounter_Nanouptime(now );
>
> and expire is
>
> rtems_timespec_from_ticks (ptimer->Timer.expire,expire );
>
> also i'm getting this compiler warning I don't know if i have to resolve it, I get that my timespec variable may not be initialized.
>
> warning: 'result' may be used uninitialized in this function [-Wmaybe-uninitialized]
>
> > The monotonic clock cannot be set.
>
> Were you talking about the fact that the monotonic clock only can go up in time?
>
> Yes, settime is an important function to distinguish between MONOTONIC
>
> > > > and REALTIME cases. That is going to be tricky to think through.
>
> i don't understand what you mean since i have added a clock type field to the POSIX_TIMER_CONTROL
>
> Thanks
>
> Zack
>
> ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
>
> On Thursday, May 13, 2021 3:05 PM, Gedare Bloom gedare at rtems.org wrote:
>
> > On Wed, May 12, 2021 at 2:42 PM zack_on_the_speed_chanel
> >
> > zack_on_the_speed_chanel at protonmail.ch wrote:
> >
> > > > This ticket mostly references the need for a test. Have you tried to
> > > >
> > > > write a test for the missing functionality?
> > >
> > > How I made a test for it was to create the timer using timer_create() . It also said to use a previous test with timer_realtime and i tested it with sometime similar to PSXtimer02 test. I got a invalid argument error when I tried it. Joel told me i also have to impliment the monotonic clock
> > >
> > > > his math doesn't make sense to me. explain what you're trying to do?
> > >
> > > I wanted calculate the remaining time with the timespec. I assumed that the timespec contained the total time seconds+miliseconds. I knew that a cpu tick is an uint32. So i converted to timspec
> > >
> > >       remaining = (uint32_t) ( ptimer->Timer.expire - now );
> > >
> > >
> > > to something like
> > >
> > > expire->tv.nsec+expire->tv.sec - spec->tv.nsec+spec->tv.sec);
> >
> > This doesn't work. Example:
> >
> > now = 0s 999999999ns
> >
> > expire = 1s 0ns
> >
> > expire - now = 1ns
> >
> > expire.nsec + expire.sec - now.nsec + now.sec = -999999999 cast to
> >
> > uint32_t is 3294967297ns or so.
> >
> > Use helper functions in rtems/timespec.h or score/timespec.h
> >
> > (depending where you implement your code, in this case, you probably
> >
> > use the score services).
> >
> > > > > Yes, settime is an important function to distinguish between MONOTONIC
> > > > >
> > > > > and REALTIME cases. That is going to be tricky to think through.
> > >
> > > Where do you want to check the clock type. I though I only need to distinguish between both when getting the value of the clock. I don't exactly know what needs to be done me being really new to posix and operating system development.
> >
> > The monotonic clock cannot be set.
> >
> > > Zack
> > >
> > > ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
> > >
> > > On Tuesday, May 11, 2021 6:23 AM, Gedare Bloom gedare at rtems.org wrote:
> > >
> > > > On Fri, May 7, 2021 at 12:53 PM zack_on_the_speed_chanel
> > > >
> > > > zack_on_the_speed_chanel at protonmail.ch wrote:
> > > >
> > > > > hello,
> > > > >
> > > > > Currenttly i'm trying to implement the clock_monotonic which was part of ticket #3889. So far these are the changes are as follows
> > > >
> > > > This ticket mostly references the need for a test. Have you tried to
> > > >
> > > > write a test for the missing functionality?
> > > >
> > > > > ptimer->clock_type= &clock_id;
> > > > >
> > > > > in the timer create i added a field for the timer id and saved it when the timer was made. Then in getttime I used the appropriate timers to get the time.
> > > > >
> > > > > for example.
> > > > >
> > > > > if (ptimer->clock_type ==CLOCK_REALTIME) {
> > > > >
> > > > > Per_CPU_Control *cpu;
> > > > >
> > > > > struct timespec spec;
> > > > >
> > > > > struct timespec expire;
> > > > >
> > > > >     cpu = _POSIX_Timer_Acquire_critical( ptimer, &lock_context );
> > > > >
> > > >
> > > > Maybe move the clock_type check to here?
> > > >
> > > > >     _TOD_Get(spec);
> > > > >       _Timespec_From_ticks( expire, &ptimer->Timer.expire);
> > > > >
> > > > >
> > > > >     if (spec->tv.nsec+spec->tv.sec > expire->tv.nsec+expire->tv.sec ) {
> > > > >
> > > > >       remaining = (uint32_t) (expire->tv.nsec+expire->tv.sec  - spec->tv.nsec+spec->tv.sec);
> > > > >
> > > >
> > > > this math doesn't make sense to me. explain what you're trying to do?
> > > >
> > > > >     } else {
> > > > >       remaining = 0;
> > > > >     }
> > > > >
> > > > >     _Timespec_From_ticks( remaining, &value->it_value );
> > > > >     value->it_interval = ptimer->timer_data.it_interval;
> > > > >
> > > > >     _POSIX_Timer_Release( cpu, &lock_context );
> > > > >     return 0;
> > > > >
> > > > >
> > > > > }
> > > > >
> > > > > Here i made two separate cases for getting the time. Also joel told me to I feel that it is a bit redundant to call everything using timespec to calculating remaining time and then just use timespec_From_ticks to get the time. It was a recommendation from Joel. Did I make the correct change. Also I assume that Delete does not need any changes because It does not require any measuring of time. Is there anything i need to change for timer_settime?
> > > >
> > > > Yes, settime is an important function to distinguish between MONOTONIC
> > > >
> > > > and REALTIME cases. That is going to be tricky to think through.
> > > >
> > > > > Thanks
> > > > >
> > > > > Zack
> > > > >
> > > > > devel mailing list
> > > > >
> > > > > devel at rtems.org
> > > > >
> > > > > http://lists.rtems.org/mailman/listinfo/devel