[ECOS] scheduler/printf questions

[ECOS] scheduler/printf questions

[Mathieu Lacage]

hi all,

I am currently trying to understand the basic architecture of the eCos
HAL (and yes, I read the docs which are excellent btw :).

I have two questions:
	- so that the scheduler can interrupt a task which reached the end of
its timeslice, there must be a timer somewhere to trigger an interrupt
which asks for a reschedule. So, where is this setup ? And where can I
find this interrupt handler's code ? I have tried to find it in the
ARM/Integrator example but failed miserably.

	- what are the semantics of a printf call ? I have tried to trace in
the package/langage libc source what happens but I still cannot figure
out where the data will be printed. I would expect there is a way to
configure this on different targets. For example, if the target has an
output serial port, I would expect the bytes to be written there. Is
this a safe assumption ? If so, where can such configuration options be
found ?

Any answer will be welcome :)

Mathieu

-- 
Mathieu Lacage <mathieu_lacage@realmagic.fr>
#p: +33 1 69 19 61 97

Re: [ECOS] scheduler/printf questions

[Andrew Lunn]

> I am currently trying to understand the basic architecture of the eCos
> HAL (and yes, I read the docs which are excellent btw :).

These questions deal with more than the HAL....

> 
> I have two questions:
> 	- so that the scheduler can interrupt a task which reached the end of
> its timeslice, there must be a timer somewhere to trigger an interrupt
> which asks for a reschedule. So, where is this setup ? And where can I
> find this interrupt handler's code ? I have tried to find it in the
> ARM/Integrator example but failed miserably.

Im not familiar with that target. I know the EBSA well....

hal_hardware_init() starts the hardware timer by calling
hal_clock_initialize with the correct timer period. 

Then in the kernel clock.cxx 

Cyg_RealTimeClock::Cyg_RealTimeClock()
    : Cyg_Clock(rtc_resolution),
      interrupt(CYGNUM_HAL_INTERRUPT_RTC, 1, (CYG_ADDRWORD)this, isr, dsr)

Installs the isr/dsr which are just bellow in the source code. The dsr
calls the scheduler to do a timeslice.

> 	- what are the semantics of a printf call ? I have tried to trace in
> the package/langage libc source what happens but I still cannot figure
> out where the data will be printed. I would expect there is a way to
> configure this on different targets. For example, if the target has an
> output serial port, I would expect the bytes to be written there. Is
> this a safe assumption ? If so, where can such configuration options be
> found ?

Its all configurable via CDL and the ecos.ecc file. There are a number
of virtual devices which map onto the physical drivers. eg /dev/tty0
is a virtual device which is normally mapped onto the physical device
/dev/ser0. Similarly /dev/termios0 is mapped to /dev/ser0. I think
normal printf is mapped to the console, which defaults to
/dev/ttydiag.  

        Andrew

Re: [ECOS] scheduler/printf questions

[Mathieu Lacage]

le ven 23-11-2001 à 16:33, Andrew Lunn a écrit :
> > I am currently trying to understand the basic architecture of the eCos
> > HAL (and yes, I read the docs which are excellent btw :).
> 
> These questions deal with more than the HAL....
> 
> > 
> > I have two questions:
> > 	- so that the scheduler can interrupt a task which reached the end of
> > its timeslice, there must be a timer somewhere to trigger an interrupt
> > which asks for a reschedule. So, where is this setup ? And where can I
> > find this interrupt handler's code ? I have tried to find it in the
> > ARM/Integrator example but failed miserably.
> 
> Im not familiar with that target. I know the EBSA well....
> 
> hal_hardware_init() starts the hardware timer by calling
> hal_clock_initialize with the correct timer period. 

this is not the case on the INTEGRATOR board. I guess it must be
somwhere else... Dunno where though :)

> 
> Then in the kernel clock.cxx 
> 
> Cyg_RealTimeClock::Cyg_RealTimeClock()
>     : Cyg_Clock(rtc_resolution),
>       interrupt(CYGNUM_HAL_INTERRUPT_RTC, 1, (CYG_ADDRWORD)this, isr, dsr)
> 
> Installs the isr/dsr which are just bellow in the source code. The dsr
> calls the scheduler to do a timeslice.

I see.

> 
> > 	- what are the semantics of a printf call ? I have tried to trace in
> > the package/langage libc source what happens but I still cannot figure
> > out where the data will be printed. I would expect there is a way to
> > configure this on different targets. For example, if the target has an
> > output serial port, I would expect the bytes to be written there. Is
> > this a safe assumption ? If so, where can such configuration options be
> > found ?
> 
> Its all configurable via CDL and the ecos.ecc file. There are a number
> of virtual devices which map onto the physical drivers. eg /dev/tty0
> is a virtual device which is normally mapped onto the physical device
> /dev/ser0. Similarly /dev/termios0 is mapped to /dev/ser0. I think
> normal printf is mapped to the console, which defaults to
> /dev/ttydiag.  

right.


Thanks for your answer,

Mathieu 

> 
>         Andrew
> 
-- 
Mathieu Lacage <mathieu_lacage@realmagic.fr>
#p: +33 1 69 19 61 97

Re: [ECOS] scheduler/printf questions

[Gary Thomas]

On Mon, 2001-11-26 at 00:58, Mathieu Lacage wrote:
> le ven 23-11-2001 à 16:33, Andrew Lunn a écrit :
> > > I am currently trying to understand the basic architecture of the eCos
> > > HAL (and yes, I read the docs which are excellent btw :).
> > 
> > These questions deal with more than the HAL....
> > 
> > > 
> > > I have two questions:
> > > 	- so that the scheduler can interrupt a task which reached the end of
> > > its timeslice, there must be a timer somewhere to trigger an interrupt
> > > which asks for a reschedule. So, where is this setup ? And where can I
> > > find this interrupt handler's code ? I have tried to find it in the
> > > ARM/Integrator example but failed miserably.
> > 
> > Im not familiar with that target. I know the EBSA well....
> > 
> > hal_hardware_init() starts the hardware timer by calling
> > hal_clock_initialize with the correct timer period. 
> 
> this is not the case on the INTEGRATOR board. I guess it must be
> somwhere else... Dunno where though :)
> 

It's actually done using the HAL_CLOCK_INITIALIZE() macro (which is what
the HAL actually uses, even if the Integrator HAL maps this onto the
function hal_clock_initialize()).

Look in: .../ecc/kernel/current/src/common/clock.cxx

> > 
> > Then in the kernel clock.cxx 
> > 
> > Cyg_RealTimeClock::Cyg_RealTimeClock()
> >     : Cyg_Clock(rtc_resolution),
> >       interrupt(CYGNUM_HAL_INTERRUPT_RTC, 1, (CYG_ADDRWORD)this, isr, dsr)
> > 
> > Installs the isr/dsr which are just bellow in the source code. The dsr
> > calls the scheduler to do a timeslice.
> 
> I see.
> 
> > 
> > > 	- what are the semantics of a printf call ? I have tried to trace in
> > > the package/langage libc source what happens but I still cannot figure
> > > out where the data will be printed. I would expect there is a way to
> > > configure this on different targets. For example, if the target has an
> > > output serial port, I would expect the bytes to be written there. Is
> > > this a safe assumption ? If so, where can such configuration options be
> > > found ?
> > 
> > Its all configurable via CDL and the ecos.ecc file. There are a number
> > of virtual devices which map onto the physical drivers. eg /dev/tty0
> > is a virtual device which is normally mapped onto the physical device
> > /dev/ser0. Similarly /dev/termios0 is mapped to /dev/ser0. I think
> > normal printf is mapped to the console, which defaults to
> > /dev/ttydiag.  
> 
> right.
> 
> 
> Thanks for your answer,
> 
> Mathieu 
> 
> > 
> >         Andrew
> > 
> -- 
> Mathieu Lacage <mathieu_lacage@realmagic.fr>
> #p: +33 1 69 19 61 97

Re: [ECOS] scheduler/printf questions

[Gary Thomas]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 2605 bytes --]

On Mon, 2001-11-26 at 00:58, Mathieu Lacage wrote:
> le ven 23-11-2001 à 16:33, Andrew Lunn a écrit :
> > > I am currently trying to understand the basic architecture of the eCos
> > > HAL (and yes, I read the docs which are excellent btw :).
> > 
> > These questions deal with more than the HAL....
> > 
> > > 
> > > I have two questions:
> > > 	- so that the scheduler can interrupt a task which reached the end of
> > > its timeslice, there must be a timer somewhere to trigger an interrupt
> > > which asks for a reschedule. So, where is this setup ? And where can I
> > > find this interrupt handler's code ? I have tried to find it in the
> > > ARM/Integrator example but failed miserably.
> > 
> > Im not familiar with that target. I know the EBSA well....
> > 
> > hal_hardware_init() starts the hardware timer by calling
> > hal_clock_initialize with the correct timer period. 
> 
> this is not the case on the INTEGRATOR board. I guess it must be
> somwhere else... Dunno where though :)
> 

It's actually done using the HAL_CLOCK_INITIALIZE() macro (which is what
the HAL actually uses, even if the Integrator HAL maps this onto the
function hal_clock_initialize()).

Look in: .../ecc/kernel/current/src/common/clock.cxx

> > 
> > Then in the kernel clock.cxx 
> > 
> > Cyg_RealTimeClock::Cyg_RealTimeClock()
> >     : Cyg_Clock(rtc_resolution),
> >       interrupt(CYGNUM_HAL_INTERRUPT_RTC, 1, (CYG_ADDRWORD)this, isr, dsr)
> > 
> > Installs the isr/dsr which are just bellow in the source code. The dsr
> > calls the scheduler to do a timeslice.
> 
> I see.
> 
> > 
> > > 	- what are the semantics of a printf call ? I have tried to trace in
> > > the package/langage libc source what happens but I still cannot figure
> > > out where the data will be printed. I would expect there is a way to
> > > configure this on different targets. For example, if the target has an
> > > output serial port, I would expect the bytes to be written there. Is
> > > this a safe assumption ? If so, where can such configuration options be
> > > found ?
> > 
> > Its all configurable via CDL and the ecos.ecc file. There are a number
> > of virtual devices which map onto the physical drivers. eg /dev/tty0
> > is a virtual device which is normally mapped onto the physical device
> > /dev/ser0. Similarly /dev/termios0 is mapped to /dev/ser0. I think
> > normal printf is mapped to the console, which defaults to
> > /dev/ttydiag.  
> 
> right.
> 
> 
> Thanks for your answer,
> 
> Mathieu 
> 
> > 
> >         Andrew
> > 
> -- 
> Mathieu Lacage <mathieu_lacage@realmagic.fr>
> #p: +33 1 69 19 61 97