[PATCH] Enhancement of the RMS manager for the overrun handling.
Kuan Hsun Chen
c0066c at gmail.com
Wed Dec 21 16:48:41 UTC 2016
Hello Gedare,
I take the latest version of RTEMS eae4541d7b8001aa18c6fc2d62b32385b0310125
to integrate my patch this time.
All you mentioned coding-convention issues are solved accordingly I think.
I will hold the copyright by myself.
rtems_rate_monotonic_postponed_num() this function is prepared for run-time
monitoring.
The user can call this function to check the number of postponed jobs in
run-time.
Therefore I name it with rtems prefix as the feature for users.
Should I add the description of this function in the documentation?
Best,
Kuan-Hsun
2016-12-20 23:10 GMT+01:00 Gedare Bloom <gedare at rtems.org>:
> On Fri, Dec 9, 2016 at 11:21 AM, Kuan-Hsun Chen <c0066c at gmail.com> wrote:
> > Three additional functions:
> > RM_Postponed_num, RM_Renew_deadline, and RM_Release_postponedjob.
> >
> > Four refined functions:
> > RM_Activate, RM_Block_while_expired, rtems_rate_monotonic_period,
> RM_Timeout.
> >
> > Rate_monotonic_Control contains one counter for counting the postponed
> jobs and one for recording the recent deadline.
> > ---
> > cpukit/rtems/include/rtems/rtems/ratemon.h | 42 ++++++--
> > cpukit/rtems/include/rtems/rtems/ratemonimpl.h | 25 +++--
> > cpukit/rtems/src/ratemonperiod.c | 144
> +++++++++++++++++++++----
> > cpukit/rtems/src/ratemontimeout.c | 13 ++-
> > 4 files changed, 183 insertions(+), 41 deletions(-)
> >
> > diff --git a/cpukit/rtems/include/rtems/rtems/ratemon.h
> b/cpukit/rtems/include/rtems/rtems/ratemon.h
> > index 50b8478..71a99dc 100644
> > --- a/cpukit/rtems/include/rtems/rtems/ratemon.h
> > +++ b/cpukit/rtems/include/rtems/rtems/ratemon.h
> > @@ -22,6 +22,7 @@
> >
> > /* COPYRIGHT (c) 1989-2009, 2016.
> > * On-Line Applications Research Corporation (OAR).
> > + * COPYRIGHT (c) 2016 Kuan-Hsun Chen, TU Dortmund University (TUDo).
> > *
> > * The license and distribution terms for this file may be
> > * found in the file LICENSE in this distribution or at
> > @@ -194,11 +195,6 @@ typedef struct {
> > /** This field is the object management portion of a Period instance.
> */
> > Objects_Control Object;
> >
> > - /**
> > - * @brief Protects the rate monotonic period state.
> > - */
> > - ISR_LOCK_MEMBER( Lock )
> > -
> Why are these removed?
>
> > /** This is the timer used to provide the unblocking mechanism. */
> > Watchdog_Control Timer;
> >
> > @@ -206,12 +202,6 @@ typedef struct {
> > rtems_rate_monotonic_period_states state;
> >
> > /**
> > - * @brief A priority node for use by the scheduler job release and
> cancel
> > - * operations.
> > - */
> > - Priority_Node Priority;
> > -
> > - /**
> Ditto.
>
> > * This field contains the length of the next period to be
> > * executed.
> > */
> > @@ -240,6 +230,19 @@ typedef struct {
> > * This field contains the statistics maintained for the period.
> > */
> > Rate_monotonic_Statistics Statistics;
> > +
> > + /**
> > + * This field contains the number of postponed jobs. When the
> watchdog timeout,
> > + * this variable will be increased immediately.
> > + */
> > + uint32_t postponed_jobs;
> > +
> > + /**
> > + * This field contains the tick of the latest deadline decided by
> the period
> > + * watchdog.
> > + */
> > + uint64_t latest_deadline;
> > +
> > } Rate_monotonic_Control;
> >
> > /**
> > @@ -386,6 +389,23 @@ void rtems_rate_monotonic_report_
> statistics_with_plugin(
> > void rtems_rate_monotonic_report_statistics( void );
> >
> > /**
> > + * @brief RTEMS Return the number of postponed jobs
> remove "RTEMS".
>
> > + *
> > + * This is a helper function to return the number of postponed jobs by
> this
> by -> in
>
> > + * given period. This number is only increased by the corresponding
> watchdog,
> Is it only in a given period, or is it postponed jobs generally, i.e.
> you could miss multiple periods hence jobs in multiple periods would
> be counted here?
>
> > + * and is decreased by RMS manager with the postponed job releasing.
> > + *
> > + * @param[in] id is the period id
> > + *
> > + * @retval This helper function returns the number of postponed
> > + * jobs with given period_id.
> > + *
> > + */
> > +uint32_t rtems_rate_monotonic_Postponed_num(
> See coding conventions for naming rules. this is better
> 'rtems_rate_monotonic_postponed_jobs(). Is this function needed in the
> public-facing API?
>
> > + rtems_id period_id
> > +);
> > +
> > +/**
> > * @brief RTEMS Rate Monotonic Period
> > *
> > * This routine implements the rtems_rate_monotonic_period directive.
> When
> > diff --git a/cpukit/rtems/include/rtems/rtems/ratemonimpl.h
> b/cpukit/rtems/include/rtems/rtems/ratemonimpl.h
> > index b6b3ffd..6cdaaeb 100644
> > --- a/cpukit/rtems/include/rtems/rtems/ratemonimpl.h
> > +++ b/cpukit/rtems/include/rtems/rtems/ratemonimpl.h
> > @@ -9,6 +9,7 @@
> > /* COPYRIGHT (c) 1989-2008.
> > * On-Line Applications Research Corporation (OAR).
> > * Copyright (c) 2016 embedded brains GmbH.
> > + * COPYRIGHT (c) 2016 Kuan-Hsun Chen, TU Dortmund University (TUDo).
> > *
> > * The license and distribution terms for this file may be
> > * found in the file LICENSE in this distribution or at
> > @@ -69,19 +70,19 @@ RTEMS_INLINE_ROUTINE Rate_monotonic_Control
> *_Rate_monotonic_Allocate( void )
> > }
> >
> > RTEMS_INLINE_ROUTINE void _Rate_monotonic_Acquire_critical(
> > - Rate_monotonic_Control *the_period,
> > - ISR_lock_Context *lock_context
> > + Thread_Control *the_thread,
> > + ISR_lock_Context *lock_context
> > )
> > {
> > - _ISR_lock_Acquire( &the_period->Lock, lock_context );
> > + _Thread_Wait_acquire_default_critical( the_thread, lock_context );
> > }
> At what version did you make your patch? These lock changes are weird.
>
> >
> > RTEMS_INLINE_ROUTINE void _Rate_monotonic_Release(
> > - Rate_monotonic_Control *the_period,
> > - ISR_lock_Context *lock_context
> > + Thread_Control *the_thread,
> > + ISR_lock_Context *lock_context
> > )
> > {
> > - _ISR_lock_Release_and_ISR_enable( &the_period->Lock, lock_context );
> > + _Thread_Wait_release_default( the_thread, lock_context );
> > }
> >
> > RTEMS_INLINE_ROUTINE Rate_monotonic_Control *_Rate_monotonic_Get(
> > @@ -116,6 +117,18 @@ bool _Rate_monotonic_Get_status(
> > Timestamp_Control *cpu_since_last_period
> > );
> >
> > +/**
> > + * @brief Renew the watchdog deadline
> > + *
> > + * This routine is prepared for the watchdog timeout to renew its
> deadline
> > + * without releasing jobs.
> > + */
> > +void _Rate_monotonic_Renew_deadline(
> > + Rate_monotonic_Control *the_period,
> > + Thread_Control *owner,
> > + ISR_lock_Context *lock_context
> > +);
> > +
> > void _Rate_monotonic_Restart(
> > Rate_monotonic_Control *the_period,
> > Thread_Control *owner,
> > diff --git a/cpukit/rtems/src/ratemonperiod.c b/cpukit/rtems/src/
> ratemonperiod.c
> > index 77bd996..26cee58 100644
> > --- a/cpukit/rtems/src/ratemonperiod.c
> > +++ b/cpukit/rtems/src/ratemonperiod.c
> > @@ -9,6 +9,7 @@
> > * COPYRIGHT (c) 1989-2010.
> > * On-Line Applications Research Corporation (OAR).
> > * Copyright (c) 2016 embedded brains GmbH.
> > + * COPYRIGHT (c) 2016 Kuan-Hsun Chen, TU Dortmund University (TUDo).
> > *
> > * The license and distribution terms for this file may be
> > * found in the file LICENSE in this distribution or at
> > @@ -63,6 +64,24 @@ bool _Rate_monotonic_Get_status(
> > return true;
> > }
> >
> > +static void _Rate_monotonic_Release_postponedjob(
> better; Release_postponed_job
>
> > + Rate_monotonic_Control *the_period,
> > + Thread_Control *owner,
> > + rtems_interval next_length,
> > + ISR_lock_Context *lock_context
> > +)
> > +{
> > + /* This function only releases the postponed jobs. */
> > + Per_CPU_Control *cpu_self;
> > + cpu_self = _Thread_Dispatch_disable_critical( lock_context );
> > + _Rate_monotonic_Release( owner, lock_context );
> > +
> > + the_period->postponed_jobs -=1;
> > + _Scheduler_Release_job( owner, the_period->latest_deadline );
> > +
> > + _Thread_Dispatch_enable( cpu_self );
> > +}
> > +
> > static void _Rate_monotonic_Release_job(
> > Rate_monotonic_Control *the_period,
> > Thread_Control *owner,
> > @@ -70,29 +89,49 @@ static void _Rate_monotonic_Release_job(
> > ISR_lock_Context *lock_context
> > )
> > {
> > - Per_CPU_Control *cpu_self;
> > - Thread_queue_Context queue_context;
> > - uint64_t deadline;
> > + Per_CPU_Control *cpu_self;
> > + uint64_t deadline;
> >
> > cpu_self = _Thread_Dispatch_disable_critical( lock_context );
> > + _Rate_monotonic_Release( owner, lock_context );
> >
> > + _ISR_lock_ISR_disable( lock_context );
> > deadline = _Watchdog_Per_CPU_insert_relative(
> > &the_period->Timer,
> > cpu_self,
> > next_length
> > );
> > - _Scheduler_Release_job(
> > - owner,
> > - &the_period->Priority,
> > - deadline,
> > - &queue_context
> > - );
> > + _ISR_lock_ISR_enable( lock_context );
> > +
> > + _Scheduler_Release_job( owner, deadline );
> >
> > - _Rate_monotonic_Release( the_period, lock_context );
> > - _Thread_Priority_update( &queue_context );
> > _Thread_Dispatch_enable( cpu_self );
> > }
> >
> > +void _Rate_monotonic_Renew_deadline(
> > + Rate_monotonic_Control *the_period,
> > + Thread_Control *owner,
> > + ISR_lock_Context *lock_context
> > +)
> > +{
> > + Per_CPU_Control *cpu_self;
> > + uint64_t deadline;
> > +
> > + cpu_self = _Thread_Dispatch_disable_critical( lock_context );
> > + _Rate_monotonic_Release( owner, lock_context );
> > +
> > + _ISR_lock_ISR_disable( lock_context );
> > + deadline = _Watchdog_Per_CPU_insert_relative(
> > + &the_period->Timer,
> > + cpu_self,
> > + the_period->next_length
> > + );
> > + the_period->latest_deadline = deadline;
> > + _ISR_lock_ISR_enable( lock_context );
> > + _Thread_Dispatch_enable( cpu_self );
> > +
> > +}
> > +
> > void _Rate_monotonic_Restart(
> > Rate_monotonic_Control *the_period,
> > Thread_Control *owner,
> > @@ -190,6 +229,10 @@ static rtems_status_code _Rate_monotonic_Activate(
> > ISR_lock_Context *lock_context
> > )
> > {
> > +
> > + /* Initialize the number of postponed job variable */
> don't need this comment or the extra blank newlines around this. Just
> init the variable.
> > + the_period->postponed_jobs = 0;
> > +
> > the_period->state = RATE_MONOTONIC_ACTIVE;
> > the_period->next_length = length;
> > _Rate_monotonic_Restart( the_period, executing, lock_context );
> > @@ -221,7 +264,7 @@ static rtems_status_code _Rate_monotonic_Block_while_
> active(
> > _Thread_Wait_flags_set( executing, RATE_MONOTONIC_INTEND_TO_BLOCK );
> >
> > cpu_self = _Thread_Dispatch_disable_critical( lock_context );
> > - _Rate_monotonic_Release( the_period, lock_context );
> > + _Rate_monotonic_Release( executing, lock_context );
> >
> > _Thread_Set_state( executing, STATES_WAITING_FOR_PERIOD );
> >
> > @@ -241,6 +284,11 @@ static rtems_status_code
> _Rate_monotonic_Block_while_active(
> > return RTEMS_SUCCESSFUL;
> > }
> >
> > +/*
> > + * There are two possible cases: one is that the previous deadline is
> missed,
> > + * The other is that the number of postponed jobs is not 0, but the
> current
> > + * deadline is still not expired, i.e., state = RATE_MONOTONIC_ACTIVE.
> > + */
> > static rtems_status_code _Rate_monotonic_Block_while_expired(
> > Rate_monotonic_Control *the_period,
> > rtems_interval length,
> > @@ -248,6 +296,12 @@ static rtems_status_code
> _Rate_monotonic_Block_while_expired(
> > ISR_lock_Context *lock_context
> > )
> > {
> > + /*
> > + * No matter the just finished jobs in time or not,
> > + * they are actually missing their deadlines already.
> > + */
> > + the_period->state = RATE_MONOTONIC_EXPIRED;
> > +
> > /*
> > * Update statistics from the concluding period
> > */
> > @@ -255,11 +309,27 @@ static rtems_status_code
> _Rate_monotonic_Block_while_expired(
> >
> > the_period->state = RATE_MONOTONIC_ACTIVE;
> > the_period->next_length = length;
> > -
> > - _Rate_monotonic_Release_job( the_period, executing, length,
> lock_context );
> > +
> > + _Rate_monotonic_Release_postponedjob( the_period, executing, length,
> lock_context );
> > return RTEMS_TIMEOUT;
> > }
> >
> > +uint32_t rtems_rate_monotonic_Postponed_num(
> Is this function used/needed to be used?
>
> > + rtems_id period_id
> > +)
> > +{
> > + Rate_monotonic_Control *the_period;
> > + ISR_lock_Context lock_context;
> > + Thread_Control *owner;
> > +
> > + the_period = _Rate_monotonic_Get( period_id, &lock_context );
> > + _Assert(the_period != NULL);
> > + uint32_t jobs = the_period->postponed_jobs;
> > + owner = the_period->owner;
> > + _Rate_monotonic_Release( owner, &lock_context );
> > + return jobs;
> > +}
> > +
> > rtems_status_code rtems_rate_monotonic_period(
> > rtems_id id,
> > rtems_interval length
> > @@ -282,22 +352,44 @@ rtems_status_code rtems_rate_monotonic_period(
> > return RTEMS_NOT_OWNER_OF_RESOURCE;
> > }
> >
> > - _Rate_monotonic_Acquire_critical( the_period, &lock_context );
> > + _Rate_monotonic_Acquire_critical( executing, &lock_context );
> >
> > state = the_period->state;
> >
> > if ( length == RTEMS_PERIOD_STATUS ) {
> > status = _Rate_monotonic_Get_status_for_state( state );
> > - _Rate_monotonic_Release( the_period, &lock_context );
> > + _Rate_monotonic_Release( executing, &lock_context );
> > } else {
> > switch ( state ) {
> > case RATE_MONOTONIC_ACTIVE:
> > - status = _Rate_monotonic_Block_while_active(
> > - the_period,
> > - length,
> > - executing,
> > - &lock_context
> > - );
> > +
> > + if(the_period->postponed_jobs > 0){
> fix whitespace
>
> > + /*
> > + * If the number of postponed jobs is not 0, it means the
> > + * previous postponed instance is finished without
> exceeding
> > + * the current period deadline.
> > + *
> align asterisks.
>
> > + * Do nothing on the watchdog deadline assignment but release
> the next
> > + * remaining postponed job.
> > + */
> > + status = _Rate_monotonic_Block_while_expired(
> > + the_period,
> > + length,
> > + executing,
> > + &lock_context
> > + );
> > + }else{
> fix ws
>
> > + /*
> > + * Normal case that no postponed jobs and no expiration, so
> wait for the period
> > + * and update the deadline of watchdog accordingly.
> > + */
> > + status = _Rate_monotonic_Block_while_active(
> > + the_period,
> > + length,
> > + executing,
> > + &lock_context
> > + );
> > + }
> > break;
> > case RATE_MONOTONIC_INACTIVE:
> > status = _Rate_monotonic_Activate(
> > @@ -308,6 +400,14 @@ rtems_status_code rtems_rate_monotonic_period(
> > );
> > break;
> > default:
> > + /*
> > + * As now this period was already TIMEOUT, there must be at
> least one
> > + * postponed job recorded by the watchdog. The one which
> exceeded
> > + * the previous deadline"s" was just finished.
> remove those double-quotes.
>
> > + *
> > + * Maybe there is more than one job postponed due to the
> preemption or
> > + * the previous finished job.
> > + */
> > _Assert( state == RATE_MONOTONIC_EXPIRED );
> > status = _Rate_monotonic_Block_while_expired(
> > the_period,
> > diff --git a/cpukit/rtems/src/ratemontimeout.c b/cpukit/rtems/src/
> ratemontimeout.c
> > index e514a31..be0a770 100644
> > --- a/cpukit/rtems/src/ratemontimeout.c
> > +++ b/cpukit/rtems/src/ratemontimeout.c
> > @@ -9,6 +9,8 @@
> > * COPYRIGHT (c) 1989-2009.
> > * On-Line Applications Research Corporation (OAR).
> > *
> > + * COPYRIGHT (c) 2016 Kuan-Hsun Chen, TU Dortmund University (TUDo).
> > + *
> > * The license and distribution terms for this file may be
> > * found in the file LICENSE in this distribution or at
> > * http://www.rtems.org/license/LICENSE.
> > @@ -31,7 +33,7 @@ void _Rate_monotonic_Timeout( Watchdog_Control
> *the_watchdog )
> > owner = the_period->owner;
> >
> > _ISR_lock_ISR_disable( &lock_context );
> > - _Rate_monotonic_Acquire_critical( the_period, &lock_context );
> > + _Rate_monotonic_Acquire_critical( owner, &lock_context );
> > wait_flags = _Thread_Wait_flags_get( owner );
> >
> > if (
> > @@ -62,7 +64,14 @@ void _Rate_monotonic_Timeout( Watchdog_Control
> *the_watchdog )
> > _Thread_Unblock( owner );
> > }
> > } else {
> > + /*
> > + * If the watchdog is timeout, it means there is an additional
> postponed
> > + * job in the next period but it is not available to release now:
> > + * Either the current task is still executed, or it is preemptive
> by the
> > + * other higher priority tasks.
> > + */
> > + the_period->postponed_jobs += 1;
> > the_period->state = RATE_MONOTONIC_EXPIRED;
> > - _Rate_monotonic_Release( the_period, &lock_context );
> > + _Rate_monotonic_Renew_deadline( the_period, owner, &lock_context );
> > }
> > }
> > --
> > 1.9.1
> >
> > _______________________________________________
> > devel mailing list
> > devel at rtems.org
> > http://lists.rtems.org/mailman/listinfo/devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/devel/attachments/20161221/cf71ce08/attachment-0001.html>
More information about the devel
mailing list