[Xenomai-help] timer-handling adeos/xenomai on arm

[Xenomai-help] timer-handling adeos/xenomai on arm

[Schlägl Manfred jun.]

[-- Attachment #1.1: Type: text/plain, Size: 2161 bytes --]

Hi again!

I've the presumption, there is something wrong with my timer-handling.
Could you please take a look at my handling.

Thanks in advance!



__ipipe_mach_get_tsc should return the absolute timer-ticks since
bootup, right?


my current handling:

INIT:
 * ns_timer_reload=ticks_per_jiffy
 * timer is loaded with ns_timer_reload
 * timer is configured to start counting down from ns_timer_reload to
zero and trigger an interrupt; after triggering it reloads automatically
with ns_timer_reload.


IPIPE-FUNCTIONS:

#define ns_getticksoffset(timer) 	\
	(likely(ns_tscok) ? ns_timer_reload - TIMER_READ(timer) : 0)


/* called after each timer-interrupt */
/* never called, if interrupt is triggered manually! */
void __ipipe_mach_acktimer(void)
{
/* int -> ns_timer_reload ticks are done; timer has already reloaded
with ns_timer_reload and is counting */
	__ipipe_mach_tsc+=ns_timer_reload;
	ns_timer_lxlost+=ns_timer_reload;

	/* ack interrupt */
	TIMER_ACKINT(TIMER_USED);
}
EXPORT_SYMBOL(__ipipe_mach_acktimer);

/* get ticks since start */
unsigned long long __ipipe_mach_get_tsc(void)
{
	unsigned long long result;
	unsigned long flags;

	spin_lock_irqsave_hw(&timer_lock, flags);

/* ticks before last interrupt + ticks since last interrupt (ack) */
	result = __ipipe_mach_tsc + ns_getticksoffset(TIMER_USED); 

	spin_unlock_irqrestore_hw(&timer_lock, flags);

	return result;
}
EXPORT_SYMBOL(__ipipe_mach_get_tsc);

/* set new timer reload */
void __ipipe_mach_set_dec(unsigned long reload)
{
	unsigned long ticks;
	unsigned long flags;

	spin_lock_irqsave_hw(&timer_lock, flags);
	
/* ticks since last interrupt (since ack) */
	ticks = ns_getticksoffset(TIMER_USED); 
	__ipipe_mach_tsc += ticks;
	ns_timer_lxlost += ticks;

	/* reprogramm timer */
	set_dec(TIMER_USED,reload);

	spin_unlock_irqrestore_hw(&timer_lock, flags);
}

EXPORT_SYMBOL(__ipipe_mach_set_dec);


/* get timer value */
unsigned long __ipipe_mach_get_dec(void)
{
	return TIMER_READ(TIMER_USED);
}
EXPORT_SYMBOL(__ipipe_mach_get_dec);




whole source see attachment



 - Manfred

[-- Attachment #1.2: time.c --]
[-- Type: text/x-csrc, Size: 13173 bytes --]

/*
 *  linux/include/asm-arm/arch-ns9750/time.h
 *
 * Copyright (C) 2004 by FS Forth-Systeme GmbH.
 * All rights reserved.
 * Markus Pietrek <mpietrek@fsforth.de>
 * @References: [1] NS9750 Hardware Reference/December 2003
 *              [2] derived from arch-integrator/time.h
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 */
/***********************************************************************
 *  @History:
 *     2006/09/01 : Adeos/ipipe support by Schlaegl Manfred jun. 
 * 			<manfred.schlaegl@gmx.at>
 *     2004/09/02 : implemented gettimeoffset
 *     2004/08/20 : ported to kernel 2.6.8
 ***********************************************************************/

#include <linux/module.h>			/* EXPORT_SYMBOL */
#include <linux/ioport.h>			/* request_resource */
#include <linux/interrupt.h>		/* irqreturn_t */

#include <asm/io.h>					/* writel */
#include <asm/mach-types.h>

#include <asm/arch/ns9750_sys.h> 	/* NS_SYS */
#include <asm/arch/irqs.h>			/* IRQ_TIMER_0 */
#include <asm/arch/time.h> 			/* ns_time_init */

#include <asm/arch/timex.h> 		/* frequency */


/* start default debug-level */
#define DEBUG_ENABLED
#define DEFAULT_DEBUG 		0

/* timer used as system timer */
#define TIMER_USED			0

/* activate sequence lock while system-timer-isr (from integrator) */
#define SEQ_LOCK


/*************************************************************************************
# begin		DEFINITIONS
**************************************************************************************/

#define CAT(x,y)	x ## y
#define XCAT(x,y)	CAT(x,y)
#define NS_SYS_TIMER_CTRL_TLCS(tlcs) 	XCAT(NS_SYS_TIMER_CTRL_TLCS_,tlcs)

#if 0
/* timer frequency = full clock-frequency / TLCS */
#define	CFG_HZ		(CPU_CLK_FREQ/TIMER_FDIV)
#endif

/* isr */
static irqreturn_t ns9750_timer_interrupt( /*@unused@*/ int irq,
					   void* dev_id,
					   struct pt_regs* regs );


static struct irqaction ns_timer_irq = {
	.name		= "NS9750 Timer Tick",
	.flags		= SA_INTERRUPT | SA_TIMER,
	.handler	= ns9750_timer_interrupt
};


/* reload value for timer */
static unsigned int ns_timer_reload;

/* correct ticks */
static unsigned long ns_timer_delta_old;
static unsigned int ns_timer_jiffies_old;

/* lost ticks, saved if __ipipe_mach_set_sec while running timer */
static unsigned long ns_timer_lxlost;

/* timer ok - flag */
static int ns_tscok;



/*************************************************************************************
# begin		IPIPE-SPECIFIC VARIABLES
**************************************************************************************/
#ifdef CONFIG_IPIPE

#ifdef CONFIG_NO_IDLE_HZ
#error "dynamic tick timer not yet supported with IPIPE"
#endif /* CONFIG_NO_IDLE_HZ */

/* the timer interrupt */
int __ipipe_mach_timerint = IRQ_TIMER_0 + TIMER_USED;
EXPORT_SYMBOL(__ipipe_mach_timerint);

/* ticks-counter */
static unsigned long long __ipipe_mach_tsc;

/* timer lock */
static DEFINE_SPINLOCK(timer_lock);

/* timer stolen - flag */
int __ipipe_mach_timerstolen = 0;
EXPORT_SYMBOL(__ipipe_mach_timerstolen);

/*************************************************************************************
# end		IPIPE-SPECIFIC VARIABLES
**************************************************************************************/
#endif /* CONFIG_IPIPE */



/* timer-ticks per jiffy (init in ns_time_init) */
unsigned int __ipipe_mach_ticks_per_jiffy;
EXPORT_SYMBOL(__ipipe_mach_ticks_per_jiffy); 


void __tm_init_timer(void);


/*************************************************************************************
# begin		DEBUG-HANDLING
**************************************************************************************/

int DEBUG = DEFAULT_DEBUG;

#define MY_NAME "time.c"
/* debug messages */

#ifdef DEBUG_ENABLED
#define dbg_raw(level, fmt, arg...) 						\
	do {									\
		if (level<=DEBUG)						\
			printk (fmt, ## arg);					\
	} while (0)
#else
#define dbg_raw(level, fmt, arg...)
#endif

/* debug messages with module- and function-information */
#define dbg(level, fmt, arg...) 						\
	dbg_raw(level, "%s: %s:\t\t" fmt, MY_NAME , __FUNCTION__ , ## arg)	

/*************************************************************************************
# end		DEBUG-HANDLING
**************************************************************************************/


/*************************************************************************************
# begin		GENERAL TIMER-FUNCTIONS
**************************************************************************************/

/* read timer value */
#define TIMER_READ(timer)					\
	readl(NS_SYS_TIMER_READ(timer))

/* set timer reload */
#define TIMER_SET_RELOAD(timer, reload)		\
	writel( reload, NS_SYS_TIMER_COUNT(timer) )

/* set timer ctrl */
#define TIMER_SET_CTRL(timer, ctrl)			\
	writel( ctrl, NS_SYS_TIMER_CTRL(timer) )

/* disable timer (TEN and generate interrupt) */
#define TIMER_DISABLE(timer)				\
	rmwl( NS_SYS_TIMER_CTRL(timer), & ~NS_SYS_TIMER_CTRL_TEN & ~NS_SYS_TIMER_CTRL_INTS);

/* enable timer (TEN and generate interrupt) */
#define TIMER_ENABLE(timer)					\
	rmwl( NS_SYS_TIMER_CTRL(timer), | NS_SYS_TIMER_CTRL_TEN | NS_SYS_TIMER_CTRL_INTS)

/* clear timer */
#define TIMER_CLEAR(timer)					\
do {										\
	writel(0,NS_SYS_TIMER_CTRL(timer) );	\
	TIMER_SET_RELOAD(timer,0);				\
} while(0)

/* ack timer interrupt */
#define TIMER_ACKINT(timer)													\
do {																		\
	rmwl( NS_SYS_TIMER_CTRL(timer), | NS_SYS_TIMER_CTRL_INTC );				\
	rmwl( NS_SYS_TIMER_CTRL(timer), & ~NS_SYS_TIMER_CTRL_INTC );			\
}while (0)


/*************************************************************************************
# end		GENERAL TIMER-FUNCTIONS
**************************************************************************************/

/*************************************************************************************
# begin		SPECIFIC SYSTEM-TIMER-FUNCTIONS
**************************************************************************************/


/* init system-timer */
#define TIMER_INIT(timer)															\
	TIMER_SET_CTRL(timer,															\
		NS_SYS_TIMER_CTRL_TLCS(TIMER_FDIV) |	/* clockfreq / TIMER_FDIV */		\
		NS_SYS_TIMER_CTRL_REN |					/* periodic */						\
		NS_SYS_TIMER_CTRL_TDBG |				/* stop timer in debug-mode */		\
		NS_SYS_TIMER_CTRL_UDS |					/* down-counter */					\
		NS_SYS_TIMER_CTRL_TSZ 					/* 32 bit */						\
	)



/* reprogramm timer */
static inline void set_dec(int timer, unsigned long reload)
{
	/* disable timer */
	TIMER_DISABLE(timer);

	/* set new reload-value */
	TIMER_SET_RELOAD(timer,reload);

	/* activate timer */
	TIMER_ENABLE(timer);

	/* save reload */
	ns_timer_reload = reload;
}


/* get ticks-offset */
#define ns_getticksoffset(timer) 											\
	(likely(ns_tscok) ? ns_timer_reload - TIMER_READ(timer) : 0)


/*
 * Returns number of us since last clock interrupt.  Note that interrupts
 * will have been disabled by do_gettimeoffset()
 */
unsigned long ns9750_gettimeoffset(void)
{
	/* get ticks */
	unsigned long timer_delta = ns_getticksoffset(TIMER_USED);

#if 1 /* correction from original-implementation */
	if( ns_timer_jiffies_old == jiffies && ns_timer_delta_old > timer_delta ) {
		/* we are called by arm/kernel/time.c:do_gettimeofday with
		 * interrupts disabled.
		 * It may happen then that there is a timer reload after
		 * interrupts disable. Therefore jiffies is not incremented.
		 * So while jiffies has not been updated, return the old
		 * time. */
		timer_delta = ns_timer_delta_old;
	} else {
		ns_timer_delta_old = timer_delta;
		ns_timer_jiffies_old = jiffies;
	}
#endif

	/* calc */
	return ( (1000 * (ns_timer_lxlost + timer_delta)) / (__ipipe_mach_ticks_per_jiffy / 10) );
}
EXPORT_SYMBOL(ns9750_gettimeoffset);



/*
 * IRQ handler for the timer
 */
static irqreturn_t ns9750_timer_interrupt( /*@unused@*/ int irq,
					   void* dev_id,
					   struct pt_regs* regs )
{
#ifdef SEQ_LOCK
	write_seqlock(&xtime_lock);
#endif

#ifdef CONFIG_IPIPE
	/*
	 * - if Linux is running natively (no ipipe), ack the timer (reprogramming automatically)
	 * - if Linux is running under ipipe, but it still has the control over
	 *   the timer (no Xenomai for example), then do nothing (ipipe
	 *   has already acked it; reprogramming automatically)
	 * - if some other domain has taken over the timer, then do nothing
	 *   (ipipe has acked it, and the other domain has reprogramed it)
	 */
	/* save ticks since boot */
	if (!__ipipe_mach_timerstolen) {
		/* ticks since last int (already done in ack)*/
//		__ipipe_mach_tsc += ns_timer_reload;

	}
#else
	/* ack_timer */
	TIMER_ACKINT(TIMER_USED);
#endif

	/* clear lost ticks since last timer_tick */
	ns_timer_lxlost = 0;

	/* linux timer tick */
	timer_tick( regs );

#ifdef SEQ_LOCK
	write_sequnlock(&xtime_lock);
#endif

	return IRQ_HANDLED;
}

static struct resource timer_resource[ 3 ] = {
	{
		.name	= "system timer",
		.start	= (NS_SYS_TIMER_CTRL( TIMER_USED ) ),
		.end	= (NS_SYS_TIMER_CTRL( TIMER_USED ) ) + 3,
	},
	{
		.name	= "system timer",
		.start	= (NS_SYS_TIMER_READ( TIMER_USED ) ),
		.end	= (NS_SYS_TIMER_READ( TIMER_USED ) ) + 3,
	},
	{
		.name	= "system timer",
		.start	= (NS_SYS_TIMER_COUNT( TIMER_USED ) ),
		.end	= (NS_SYS_TIMER_COUNT( TIMER_USED ) ) + 3,
	}
};

/*
 * Set up timer interrupt, and return the current time in seconds.
 */
void __init ns_time_init( void )
{
	request_resource( &iomem_resource, &timer_resource[ 0 ] );
	request_resource( &iomem_resource, &timer_resource[ 1 ] );
	request_resource( &iomem_resource, &timer_resource[ 2 ] );

	__ipipe_mach_ticks_per_jiffy = LATCH;

	dbg(0,"\n"
        " system-clock:                %d Hz\n"
		" timer-clock:                 %d Hz   (=system-clock/%d) \n"
		" ticks_per_jiffy:             %d \n"
		" jiffis per usecs             %d      (HZ-value)\n"
		,
		CPU_CLK_FREQ,
		CLOCK_TICK_RATE, TIMER_FDIV, 
		__ipipe_mach_ticks_per_jiffy,
		HZ
	);

	/* clear timer*/
	TIMER_CLEAR(TIMER_USED);

	/* set timer-isr */
	setup_irq( IRQ_TIMER_0 + TIMER_USED, &ns_timer_irq );

	/* init timer */
	TIMER_INIT(TIMER_USED);

	/* reload timer */
	set_dec(TIMER_USED,__ipipe_mach_ticks_per_jiffy); /* a tick every HZ */

	/* ok */
	ns_tscok = 1;
}



/*************************************************************************************
# end		SPECIFIC SYSTEM-TIMER-FUNCTIONS
**************************************************************************************/




#ifdef CONFIG_IPIPE
/*************************************************************************************
# begin		IPIPE-FUNCTIONS
**************************************************************************************/

/* called after each timer-interrupt */
/* never called, if interrupt is triggered manually! */
void __ipipe_mach_acktimer(void)
{
	/* int -> ns_timer_reload ticks done */
	__ipipe_mach_tsc+=ns_timer_reload;
	ns_timer_lxlost+=ns_timer_reload;

	/* ack interrupt */
	TIMER_ACKINT(TIMER_USED);
}
EXPORT_SYMBOL(__ipipe_mach_acktimer);

/* get ticks since start */
unsigned long long __ipipe_mach_get_tsc(void)
{
	unsigned long long result;
	unsigned long flags;

	spin_lock_irqsave_hw(&timer_lock, flags);
	result = __ipipe_mach_tsc + ns_getticksoffset(TIMER_USED);
	spin_unlock_irqrestore_hw(&timer_lock, flags);

	return result;
}
EXPORT_SYMBOL(__ipipe_mach_get_tsc);

/* set new timer reload */
void __ipipe_mach_set_dec(unsigned long reload)
{
	unsigned long ticks;
	unsigned long flags;

	spin_lock_irqsave_hw(&timer_lock, flags);
	
	/* ticks since last int */
	ticks = ns_getticksoffset(TIMER_USED);
	__ipipe_mach_tsc += ticks;
	ns_timer_lxlost += ticks;

	/* reprogramm timer */
	set_dec(TIMER_USED,reload);

	spin_unlock_irqrestore_hw(&timer_lock, flags);
}

EXPORT_SYMBOL(__ipipe_mach_set_dec);


/* get timer value */
unsigned long __ipipe_mach_get_dec(void)
{
	return TIMER_READ(TIMER_USED);
}
EXPORT_SYMBOL(__ipipe_mach_get_dec);


/*************************************************************************************
# end		IPIPE-FUNCTIONS
**************************************************************************************/
#endif /* CONFIG_IPIPE */


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [Xenomai-help] timer-handling adeos/xenomai on arm

[Gilles Chanteperdrix]

Schlägl "Manfred jun." wrote:
 > Hi again!
 > 
 > I've the presumption, there is something wrong with my timer-handling.
 > Could you please take a look at my handling.
 > 
 > Thanks in advance!

The problem I see with your code is that you are updating
ns_timer_lxlost in __ipipe_mach_acktimer, the integrator architecture
code, which also uses a decrementer does not do that. Apart from that, I
see nothing wrong.

If there is a problem with using ipipe_trigger_irq on your architecture,
maybe we could let rthal_timer_program_shot pass null delays to
__ipipe_mach_set_dec.

-- 


					    Gilles Chanteperdrix.

Re: [Xenomai-help] timer-handling adeos/xenomai on arm

[Philippe Gerum]

On Sat, 2006-11-11 at 12:05 +0100, Gilles Chanteperdrix wrote:
> Schlägl "Manfred jun." wrote:
>  > Hi again!
>  > 
>  > I've the presumption, there is something wrong with my timer-handling.
>  > Could you please take a look at my handling.
>  > 
>  > Thanks in advance!
> 
> The problem I see with your code is that you are updating
> ns_timer_lxlost in __ipipe_mach_acktimer, the integrator architecture
> code, which also uses a decrementer does not do that. Apart from that, I
> see nothing wrong.
> 
> If there is a problem with using ipipe_trigger_irq on your architecture,
> maybe we could let rthal_timer_program_shot pass null delays to
> __ipipe_mach_set_dec.


Maybe the issue is not with ipipe_trigger_irq, which looks ok for ARM
too, but rather with the internal latency/timings of the Xenomai port
over this arch? The very difference between calling ipipe_trigger_irq
and programming the decrementer for a close shot, is that interrupt
delivery would happen immediately after the interrupt mask is enabled
again with the former method. OTOH, the latter would likely leave some
time left to the nucleus to get out of the current IRQ, thus preventing
recursion-induced lockups.

-- 
Philippe.

Re: [Xenomai-help] timer-handling adeos/xenomai on arm

[Gilles Chanteperdrix]

Philippe Gerum wrote:
 > On Sat, 2006-11-11 at 12:05 +0100, Gilles Chanteperdrix wrote:
 > > Schlägl "Manfred jun." wrote:
 > >  > Hi again!
 > >  > 
 > >  > I've the presumption, there is something wrong with my timer-handling.
 > >  > Could you please take a look at my handling.
 > >  > 
 > >  > Thanks in advance!
 > > 
 > > The problem I see with your code is that you are updating
 > > ns_timer_lxlost in __ipipe_mach_acktimer, the integrator architecture
 > > code, which also uses a decrementer does not do that. Apart from that, I
 > > see nothing wrong.
 > > 
 > > If there is a problem with using ipipe_trigger_irq on your architecture,
 > > maybe we could let rthal_timer_program_shot pass null delays to
 > > __ipipe_mach_set_dec.
 > 
 > 
 > Maybe the issue is not with ipipe_trigger_irq, which looks ok for ARM
 > too, but rather with the internal latency/timings of the Xenomai port
 > over this arch? The very difference between calling ipipe_trigger_irq
 > and programming the decrementer for a close shot, is that interrupt
 > delivery would happen immediately after the interrupt mask is enabled
 > again with the former method. OTOH, the latter would likely leave some
 > time left to the nucleus to get out of the current IRQ, thus preventing
 > recursion-induced lockups.

What I meant is that maybe in Manfred case, some global variable need to
be set in __ipipe_mach_set_dec and used in __ipipe_mach_acktimer and when
using ipipe_trigger_irq, __ipipe_mach_acktimer get called without
__ipipe_mach_set_dec.

That said, I probably misunderstood ns_timer_reload.

-- 


					    Gilles Chanteperdrix.

Re: [Xenomai-help] timer-handling adeos/xenomai on arm

[Sebastian Smolorz]

Gilles Chanteperdrix wrote:
> Schlägl "Manfred jun." wrote:
>  > Hi again!
>  >
>  > I've the presumption, there is something wrong with my timer-handling.
>  > Could you please take a look at my handling.
>  >
>  > Thanks in advance!
>
> The problem I see with your code is that you are updating
> ns_timer_lxlost in __ipipe_mach_acktimer, the integrator architecture  
> code, which also uses a decrementer does not do that. Apart from that, I
> see nothing wrong.

I see another one which has to do with the fact that __ipipe_mach_tsc is 
updated both in __ipipe_mach_get_tsc and __ipipe_mach_set_dec. This leads to 
double-added ticks because the latter funcion is called only once a period 
and the former even more than once. So Xenomai counts jiffies 
in /proc/xenomai/timer to fast. Manfred, can you confirm this?

Sebastian

Re: [Xenomai-help] timer-handling adeos/xenomai on arm

[Sebastian Smolorz]

Sebastian Smolorz wrote:
> Gilles Chanteperdrix wrote:
> > Schlägl "Manfred jun." wrote:
> >  > Hi again!
> >  >
> >  > I've the presumption, there is something wrong with my timer-handling.
> >  > Could you please take a look at my handling.
> >  >
> >  > Thanks in advance!
> >
> > The problem I see with your code is that you are updating
> > ns_timer_lxlost in __ipipe_mach_acktimer, the integrator architecture
> > code, which also uses a decrementer does not do that. Apart from that, I
> > see nothing wrong.
>
> I see another one which has to do with the fact that __ipipe_mach_tsc is
> updated both in __ipipe_mach_get_tsc and __ipipe_mach_set_dec. This leads
> to double-added ticks because the latter funcion is called only once a
> period and the former even more than once. So Xenomai counts jiffies
> in /proc/xenomai/timer to fast. Manfred, can you confirm this?

Forget this, my eyes weren't open this morning ... In  __ipipe_mach_get_tsc 
there is no update of __ipipe_mach_tsc, of course.

Sebastian

Re: [Xenomai-help] timer-handling adeos/xenomai on arm

[Gilles Chanteperdrix]

Sebastian Smolorz wrote:
 > Sebastian Smolorz wrote:
 > > Gilles Chanteperdrix wrote:
 > > > Schlägl "Manfred jun." wrote:
 > > >  > Hi again!
 > > >  >
 > > >  > I've the presumption, there is something wrong with my timer-handling.
 > > >  > Could you please take a look at my handling.
 > > >  >
 > > >  > Thanks in advance!
 > > >
 > > > The problem I see with your code is that you are updating
 > > > ns_timer_lxlost in __ipipe_mach_acktimer, the integrator architecture
 > > > code, which also uses a decrementer does not do that. Apart from that, I
 > > > see nothing wrong.
 > >
 > > I see another one which has to do with the fact that __ipipe_mach_tsc is
 > > updated both in __ipipe_mach_get_tsc and __ipipe_mach_set_dec. This leads
 > > to double-added ticks because the latter funcion is called only once a
 > > period and the former even more than once. So Xenomai counts jiffies
 > > in /proc/xenomai/timer to fast. Manfred, can you confirm this?
 > 
 > Forget this, my eyes weren't open this morning ... In  __ipipe_mach_get_tsc 
 > there is no update of __ipipe_mach_tsc, of course.

Right, but __ipipe_mach_tsc and ns_timer_lxlost get also updated in
__ipipe_mach_acktimer, this looks wrong, in the integrator
implementation, they are updated in the timer interrupt, only if the
timer is not stolen.

-- 


					    Gilles Chanteperdrix.

Re: [Xenomai-help] timer-handling adeos/xenomai on arm

[Sebastian Smolorz]

Gilles Chanteperdrix wrote:
> Sebastian Smolorz wrote:
>  > Sebastian Smolorz wrote:
>  > > Gilles Chanteperdrix wrote:
>  > > > Schlägl "Manfred jun." wrote:
>  > > >  > Hi again!
>  > > >  >
>  > > >  > I've the presumption, there is something wrong with my
>  > > >  > timer-handling. Could you please take a look at my handling.
>  > > >  >
>  > > >  > Thanks in advance!
>  > > >
>  > > > The problem I see with your code is that you are updating
>  > > > ns_timer_lxlost in __ipipe_mach_acktimer, the integrator
>  > > > architecture code, which also uses a decrementer does not do that.
>  > > > Apart from that, I see nothing wrong.
>  > >
>  > > I see another one which has to do with the fact that __ipipe_mach_tsc
>  > > is updated both in __ipipe_mach_get_tsc and __ipipe_mach_set_dec. This
>  > > leads to double-added ticks because the latter funcion is called only
>  > > once a period and the former even more than once. So Xenomai counts
>  > > jiffies in /proc/xenomai/timer to fast. Manfred, can you confirm this?
>  >
>  > Forget this, my eyes weren't open this morning ... In 
>  > __ipipe_mach_get_tsc there is no update of __ipipe_mach_tsc, of course.
>
> Right, but __ipipe_mach_tsc and ns_timer_lxlost get also updated in
> __ipipe_mach_acktimer, this looks wrong, in the integrator
> implementation, they are updated in the timer interrupt, only if the
> timer is not stolen.

OK, let's speculate a little bit. (Maybe all of the following turns out to be 
completely wrong, but it wouldn't be the first time for me today. ;-) 
Hopefully Manfred can give some comments which pieces of my speculation are 
right.)

The timer counts down to zero, causes an interrupt and then reloads 
automatically with ns_timer_reload. So, when __ipipe_mach_acktimer is called 
ns_timer_reload ticks have passed and can be added to __ipipe_mach_tsc and 
ns_timer_lxlost. Some ticks later, __ipipe_mach_set_dec is executed, the 
passed ticks are added to __ipipe_mach_tsc and ns_timer_lxlost and the timer 
is reprogrammed to the new value.

Could be right ...

Sebastian

Re: [Xenomai-help] timer-handling adeos/xenomai on arm

[Gilles Chanteperdrix]

Sebastian Smolorz wrote:
 > Gilles Chanteperdrix wrote:
 > > Sebastian Smolorz wrote:
 > >  > Sebastian Smolorz wrote:
 > >  > > Gilles Chanteperdrix wrote:
 > >  > > > Schlägl "Manfred jun." wrote:
 > >  > > >  > Hi again!
 > >  > > >  >
 > >  > > >  > I've the presumption, there is something wrong with my
 > >  > > >  > timer-handling. Could you please take a look at my handling.
 > >  > > >  >
 > >  > > >  > Thanks in advance!
 > >  > > >
 > >  > > > The problem I see with your code is that you are updating
 > >  > > > ns_timer_lxlost in __ipipe_mach_acktimer, the integrator
 > >  > > > architecture code, which also uses a decrementer does not do that.
 > >  > > > Apart from that, I see nothing wrong.
 > >  > >
 > >  > > I see another one which has to do with the fact that __ipipe_mach_tsc
 > >  > > is updated both in __ipipe_mach_get_tsc and __ipipe_mach_set_dec. This
 > >  > > leads to double-added ticks because the latter funcion is called only
 > >  > > once a period and the former even more than once. So Xenomai counts
 > >  > > jiffies in /proc/xenomai/timer to fast. Manfred, can you confirm this?
 > >  >
 > >  > Forget this, my eyes weren't open this morning ... In 
 > >  > __ipipe_mach_get_tsc there is no update of __ipipe_mach_tsc, of course.
 > >
 > > Right, but __ipipe_mach_tsc and ns_timer_lxlost get also updated in
 > > __ipipe_mach_acktimer, this looks wrong, in the integrator
 > > implementation, they are updated in the timer interrupt, only if the
 > > timer is not stolen.
 > 
 > OK, let's speculate a little bit. (Maybe all of the following turns out to be 
 > completely wrong, but it wouldn't be the first time for me today. ;-) 
 > Hopefully Manfred can give some comments which pieces of my speculation are 
 > right.)
 > 
 > The timer counts down to zero, causes an interrupt and then reloads 
 > automatically with ns_timer_reload. So, when __ipipe_mach_acktimer is called 
 > ns_timer_reload ticks have passed and can be added to __ipipe_mach_tsc and 
 > ns_timer_lxlost. Some ticks later, __ipipe_mach_set_dec is executed, the 
 > passed ticks are added to __ipipe_mach_tsc and ns_timer_lxlost and the timer 
 > is reprogrammed to the new value.

The problem is that when Xenomai manages the timer, it get reloaded with
values distinct from ns_timer_reload, it is even not reloaded at all
when the system is a bit late, so adding ns_timer_reload to
__ipipe_mach_tsc and ns_timer_lxlost in these cases seem wrong.

-- 


					    Gilles Chanteperdrix.

Re: [Xenomai-help] timer-handling adeos/xenomai on arm

[Sebastian Smolorz]

Gilles Chanteperdrix wrote:
> Sebastian Smolorz wrote:
>  > Gilles Chanteperdrix wrote:
>  > > Sebastian Smolorz wrote:
>  > >  > Sebastian Smolorz wrote:
>  > >  > > Gilles Chanteperdrix wrote:
>  > >  > > > Schlägl "Manfred jun." wrote:
>  > >  > > >  > Hi again!
>  > >  > > >  >
>  > >  > > >  > I've the presumption, there is something wrong with my
>  > >  > > >  > timer-handling. Could you please take a look at my handling.
>  > >  > > >  >
>  > >  > > >  > Thanks in advance!
>  > >  > > >
>  > >  > > > The problem I see with your code is that you are updating
>  > >  > > > ns_timer_lxlost in __ipipe_mach_acktimer, the integrator
>  > >  > > > architecture code, which also uses a decrementer does not do
>  > >  > > > that. Apart from that, I see nothing wrong.
>  > >  > >
>  > >  > > I see another one which has to do with the fact that
>  > >  > > __ipipe_mach_tsc is updated both in __ipipe_mach_get_tsc and
>  > >  > > __ipipe_mach_set_dec. This leads to double-added ticks because
>  > >  > > the latter funcion is called only once a period and the former
>  > >  > > even more than once. So Xenomai counts jiffies in
>  > >  > > /proc/xenomai/timer to fast. Manfred, can you confirm this?
>  > >  >
>  > >  > Forget this, my eyes weren't open this morning ... In
>  > >  > __ipipe_mach_get_tsc there is no update of __ipipe_mach_tsc, of
>  > >  > course.
>  > >
>  > > Right, but __ipipe_mach_tsc and ns_timer_lxlost get also updated in
>  > > __ipipe_mach_acktimer, this looks wrong, in the integrator
>  > > implementation, they are updated in the timer interrupt, only if the
>  > > timer is not stolen.
>  >
>  > OK, let's speculate a little bit. (Maybe all of the following turns out
>  > to be completely wrong, but it wouldn't be the first time for me today.
>  > ;-) Hopefully Manfred can give some comments which pieces of my
>  > speculation are right.)
>  >
>  > The timer counts down to zero, causes an interrupt and then reloads
>  > automatically with ns_timer_reload. So, when __ipipe_mach_acktimer is
>  > called ns_timer_reload ticks have passed and can be added to
>  > __ipipe_mach_tsc and ns_timer_lxlost. Some ticks later,
>  > __ipipe_mach_set_dec is executed, the passed ticks are added to
>  > __ipipe_mach_tsc and ns_timer_lxlost and the timer is reprogrammed to
>  > the new value.
>
> The problem is that when Xenomai manages the timer, it get reloaded with
> values distinct from ns_timer_reload,

But the reload value is assigned to ns_timer_reload in set_dec. So every time 
__ipipe_mach_acktimer is executed it adds the last set reload value which is 
the right one, isn't it?

> it is even not reloaded at all 
> when the system is a bit late,

Are you referring to the discussed ipipe_trigger_irq call when the delay is 
zero?

--
Sebastian

Re: [Xenomai-help] timer-handling adeos/xenomai on arm

[Gilles Chanteperdrix]

Sebastian Smolorz wrote:
> Gilles Chanteperdrix wrote:
> 
>>Sebastian Smolorz wrote:
>> > Gilles Chanteperdrix wrote:
>> > > Sebastian Smolorz wrote:
>> > >  > Sebastian Smolorz wrote:
>> > >  > > Gilles Chanteperdrix wrote:
>> > >  > > > Schlägl "Manfred jun." wrote:
>> > >  > > >  > Hi again!
>> > >  > > >  >
>> > >  > > >  > I've the presumption, there is something wrong with my
>> > >  > > >  > timer-handling. Could you please take a look at my handling.
>> > >  > > >  >
>> > >  > > >  > Thanks in advance!
>> > >  > > >
>> > >  > > > The problem I see with your code is that you are updating
>> > >  > > > ns_timer_lxlost in __ipipe_mach_acktimer, the integrator
>> > >  > > > architecture code, which also uses a decrementer does not do
>> > >  > > > that. Apart from that, I see nothing wrong.
>> > >  > >
>> > >  > > I see another one which has to do with the fact that
>> > >  > > __ipipe_mach_tsc is updated both in __ipipe_mach_get_tsc and
>> > >  > > __ipipe_mach_set_dec. This leads to double-added ticks because
>> > >  > > the latter funcion is called only once a period and the former
>> > >  > > even more than once. So Xenomai counts jiffies in
>> > >  > > /proc/xenomai/timer to fast. Manfred, can you confirm this?
>> > >  >
>> > >  > Forget this, my eyes weren't open this morning ... In
>> > >  > __ipipe_mach_get_tsc there is no update of __ipipe_mach_tsc, of
>> > >  > course.
>> > >
>> > > Right, but __ipipe_mach_tsc and ns_timer_lxlost get also updated in
>> > > __ipipe_mach_acktimer, this looks wrong, in the integrator
>> > > implementation, they are updated in the timer interrupt, only if the
>> > > timer is not stolen.
>> >
>> > OK, let's speculate a little bit. (Maybe all of the following turns out
>> > to be completely wrong, but it wouldn't be the first time for me today.
>> > ;-) Hopefully Manfred can give some comments which pieces of my
>> > speculation are right.)
>> >
>> > The timer counts down to zero, causes an interrupt and then reloads
>> > automatically with ns_timer_reload. So, when __ipipe_mach_acktimer is
>> > called ns_timer_reload ticks have passed and can be added to
>> > __ipipe_mach_tsc and ns_timer_lxlost. Some ticks later,
>> > __ipipe_mach_set_dec is executed, the passed ticks are added to
>> > __ipipe_mach_tsc and ns_timer_lxlost and the timer is reprogrammed to
>> > the new value.
>>
>>The problem is that when Xenomai manages the timer, it get reloaded with
>>values distinct from ns_timer_reload,
> 
> 
> But the reload value is assigned to ns_timer_reload in set_dec. So every time 
> __ipipe_mach_acktimer is executed it adds the last set reload value which is 
> the right one, isn't it?

Manfred sent its code in two versions: one inlined in the mail, the
other as an attachment. In the inlined version, ns_timer_reload is not
set in set_dec.

> 
> 
>>it is even not reloaded at all 
>>when the system is a bit late,
> 
> 
> Are you referring to the discussed ipipe_trigger_irq call when the delay is 
> zero?

Yes. And in this case set_dec is not called before acktimer, so
ns_timer_reload value is wrong.

-- 
                                                  Gilles Chanteperdrix

Re: [Xenomai-help] timer-handling adeos/xenomai on arm

[Sebastian Smolorz]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 3768 bytes --]

On Mon, 13 Nov 2006, Gilles Chanteperdrix wrote:

> Sebastian Smolorz wrote:
>> Gilles Chanteperdrix wrote:
>> 
>>> Sebastian Smolorz wrote:
>>> > Gilles Chanteperdrix wrote:
>>> > > Sebastian Smolorz wrote:
>>> > >  > Sebastian Smolorz wrote:
>>> > >  > > Gilles Chanteperdrix wrote:
>>> > >  > > > Schlägl "Manfred jun." wrote:
>>> > >  > > >  > Hi again!
>>> > >  > > >  >
>>> > >  > > >  > I've the presumption, there is something wrong with my
>>> > >  > > >  > timer-handling. Could you please take a look at my handling.
>>> > >  > > >  >
>>> > >  > > >  > Thanks in advance!
>>> > >  > > >
>>> > >  > > > The problem I see with your code is that you are updating
>>> > >  > > > ns_timer_lxlost in __ipipe_mach_acktimer, the integrator
>>> > >  > > > architecture code, which also uses a decrementer does not do
>>> > >  > > > that. Apart from that, I see nothing wrong.
>>> > >  > >
>>> > >  > > I see another one which has to do with the fact that
>>> > >  > > __ipipe_mach_tsc is updated both in __ipipe_mach_get_tsc and
>>> > >  > > __ipipe_mach_set_dec. This leads to double-added ticks because
>>> > >  > > the latter funcion is called only once a period and the former
>>> > >  > > even more than once. So Xenomai counts jiffies in
>>> > >  > > /proc/xenomai/timer to fast. Manfred, can you confirm this?
>>> > >  >
>>> > >  > Forget this, my eyes weren't open this morning ... In
>>> > >  > __ipipe_mach_get_tsc there is no update of __ipipe_mach_tsc, of
>>> > >  > course.
>>> > >
>>> > > Right, but __ipipe_mach_tsc and ns_timer_lxlost get also updated in
>>> > > __ipipe_mach_acktimer, this looks wrong, in the integrator
>>> > > implementation, they are updated in the timer interrupt, only if the
>>> > > timer is not stolen.
>>> >
>>> > OK, let's speculate a little bit. (Maybe all of the following turns out
>>> > to be completely wrong, but it wouldn't be the first time for me today.
>>> > ;-) Hopefully Manfred can give some comments which pieces of my
>>> > speculation are right.)
>>> >
>>> > The timer counts down to zero, causes an interrupt and then reloads
>>> > automatically with ns_timer_reload. So, when __ipipe_mach_acktimer is
>>> > called ns_timer_reload ticks have passed and can be added to
>>> > __ipipe_mach_tsc and ns_timer_lxlost. Some ticks later,
>>> > __ipipe_mach_set_dec is executed, the passed ticks are added to
>>> > __ipipe_mach_tsc and ns_timer_lxlost and the timer is reprogrammed to
>>> > the new value.
>>> 
>>> The problem is that when Xenomai manages the timer, it get reloaded with
>>> values distinct from ns_timer_reload,
>> 
>> 
>> But the reload value is assigned to ns_timer_reload in set_dec. So every 
>> time __ipipe_mach_acktimer is executed it adds the last set reload value 
>> which is the right one, isn't it?
>
> Manfred sent its code in two versions: one inlined in the mail, the
> other as an attachment. In the inlined version, ns_timer_reload is not
> set in set_dec.

Ok, I was talking about the attachment because set_dec was not there in 
the inlined code.

>
>> 
>> 
>>> it is even not reloaded at all when the system is a bit late,
>> 
>> 
>> Are you referring to the discussed ipipe_trigger_irq call when the delay is 
>> zero?
>
> Yes. And in this case set_dec is not called before acktimer,

Agreed. So the adding of ticks must be removed from __ipipe_mach_acktimer. 
Additionally, when the timer has reloaded itself these ticks must not be 
"forgotten". The integrator implementation takes care of it in 
integrator_getticksoffset(), see [1].

[1] 
http://www.rts.uni-hannover.de/xenomai/lxr/source/ksrc/arch/arm/patches/adeos-ipipe-2.6.15-arm-1.5-02.patch?v=SVN-trunk;a=arm#L5609

--
Sebastian