[PATCH] Convert alpha to use arch_gettimeoffset()

[PATCH] Convert alpha to use arch_gettimeoffset()

[john stultz]

This patch converts alpha to use GENERIC_TIME via the arch_getoffset()
infrastructure, reducing the amount of arch specific code we need to
maintain.

I suspect the alpha arch could even be further improved to provide and
rpcc() based clocksource, but not having the hardware, I don't feel
comfortable attempting the more complicated conversion (but I'd be glad
to help if anyone else is interested).

This patch applies on top of Linus' current -git tree (pre 2.6.31-rc1).

I've taken my best swing at converting this, but I'm not 100% confident
I got it right. It does cross compile, so that's hopefully a good sign.
Any assistance from arch maintainers or testers to get this merged would
be great.

thanks
-john

Signed-off-by: John Stultz <johnstul@us.ibm.com>
---

 Kconfig       |    8 ++++++
 kernel/time.c |   72 +++-------------------------------------------------------
 2 files changed, 12 insertions(+), 68 deletions(-)

diff --git a/arch/alpha/Kconfig b/arch/alpha/Kconfig
index 9fb8aae..4434481 100644
--- a/arch/alpha/Kconfig
+++ b/arch/alpha/Kconfig
@@ -45,6 +45,14 @@ config GENERIC_CALIBRATE_DELAY
 	bool
 	default y
 
+config GENERIC_TIME
+	bool
+	default y
+
+config ARCH_USES_GETTIMEOFFSET
+	bool
+	default y
+
 config ZONE_DMA
 	bool
 	default y
diff --git a/arch/alpha/kernel/time.c b/arch/alpha/kernel/time.c
index b04e2cb..8f7eb6d 100644
--- a/arch/alpha/kernel/time.c
+++ b/arch/alpha/kernel/time.c
@@ -408,28 +408,17 @@ time_init(void)
  * part.  So we can't do the "find absolute time in terms of cycles" thing
  * that the other ports do.
  */
-void
-do_gettimeofday(struct timeval *tv)
+u32 arch_gettimeoffset(void)
 {
-	unsigned long flags;
-	unsigned long sec, usec, seq;
 	unsigned long delta_cycles, delta_usec, partial_tick;
 
-	do {
-		seq = read_seqbegin_irqsave(&xtime_lock, flags);
-
-		delta_cycles = rpcc() - state.last_time;
-		sec = xtime.tv_sec;
-		usec = (xtime.tv_nsec / 1000);
-		partial_tick = state.partial_tick;
-
-	} while (read_seqretry_irqrestore(&xtime_lock, seq, flags));
-
 #ifdef CONFIG_SMP
 	/* Until and unless we figure out how to get cpu cycle counters
 	   in sync and keep them there, we can't use the rpcc tricks.  */
 	delta_usec = 0;
 #else
+	delta_cycles = rpcc() - state.last_time;
+	partial_tick = state.partial_tick;
 	/*
 	 * usec = cycles * ticks_per_cycle * 2**48 * 1e6 / (2**48 * ticks)
 	 *	= cycles * (s_t_p_c) * 1e6 / (2**48 * ticks)
@@ -448,62 +437,9 @@ do_gettimeofday(struct timeval *tv)
 	delta_usec = ((delta_usec / ((1UL << (FIX_SHIFT-6-1)) * HZ)) + 1) / 2;
 #endif
 
-	usec += delta_usec;
-	if (usec >= 1000000) {
-		sec += 1;
-		usec -= 1000000;
-	}
-
-	tv->tv_sec = sec;
-	tv->tv_usec = usec;
-}
-
-EXPORT_SYMBOL(do_gettimeofday);
-
-int
-do_settimeofday(struct timespec *tv)
-{
-	time_t wtm_sec, sec = tv->tv_sec;
-	long wtm_nsec, nsec = tv->tv_nsec;
-	unsigned long delta_nsec;
-
-	if ((unsigned long)tv->tv_nsec >= NSEC_PER_SEC)
-		return -EINVAL;
-
-	write_seqlock_irq(&xtime_lock);
-
-	/* The offset that is added into time in do_gettimeofday above
-	   must be subtracted out here to keep a coherent view of the
-	   time.  Without this, a full-tick error is possible.  */
-
-#ifdef CONFIG_SMP
-	delta_nsec = 0;
-#else
-	delta_nsec = rpcc() - state.last_time;
-	delta_nsec = (delta_nsec * state.scaled_ticks_per_cycle 
-		      + state.partial_tick) * 15625;
-	delta_nsec = ((delta_nsec / ((1UL << (FIX_SHIFT-6-1)) * HZ)) + 1) / 2;
-	delta_nsec *= 1000;
-#endif
-
-	nsec -= delta_nsec;
-
-	wtm_sec  = wall_to_monotonic.tv_sec + (xtime.tv_sec - sec);
-	wtm_nsec = wall_to_monotonic.tv_nsec + (xtime.tv_nsec - nsec);
-
-	set_normalized_timespec(&xtime, sec, nsec);
-	set_normalized_timespec(&wall_to_monotonic, wtm_sec, wtm_nsec);
-
-	ntp_clear();
-
-	write_sequnlock_irq(&xtime_lock);
-	clock_was_set();
-	return 0;
+	return delta_usec * 1000;
 }
 
-EXPORT_SYMBOL(do_settimeofday);
-
-
 /*
  * In order to set the CMOS clock precisely, set_rtc_mmss has to be
  * called 500 ms after the second nowtime has started, because when

Re: [PATCH] Convert alpha to use arch_gettimeoffset()

[Dialup Jon Norstog]

John,

Thank you so much!


jn

---------- Original Message -----------
From: john stultz <johnstul@us.ibm.com>
To: rth@twiddle.net, ink@jurassic.park.msu.ru
Cc: lkml <linux-kernel@vger.kernel.org>, linux-alpha@vger.kernel.org
Sent: Thu, 11 Jun 2009 15:10:10 -0700
Subject: [PATCH] Convert alpha to use arch_gettimeoffset()

> This patch converts alpha to use GENERIC_TIME via the 
> arch_getoffset() infrastructure, reducing the amount of arch 
> specific code we need to maintain.
> 
> I suspect the alpha arch could even be further improved to provide 
> and rpcc() based clocksource, but not having the hardware, I don't feel
> comfortable attempting the more complicated conversion (but I'd be glad
> to help if anyone else is interested).
> 
> This patch applies on top of Linus' current -git tree (pre 2.6.31-
> rc1).
> 
> I've taken my best swing at converting this, but I'm not 100% confident
> I got it right. It does cross compile, so that's hopefully a good sign.
> Any assistance from arch maintainers or testers to get this merged would
> be great.
> 
> thanks
> -john
> 
> Signed-off-by: John Stultz <johnstul@us.ibm.com>
> ---
> 
>  Kconfig       |    8 ++++++
>  kernel/time.c |   72 +++--------------------------------------------
> ----------- 2 files changed, 12 insertions(+), 68 deletions(-)
> 
> diff --git a/arch/alpha/Kconfig b/arch/alpha/Kconfig
> index 9fb8aae..4434481 100644
> --- a/arch/alpha/Kconfig
> +++ b/arch/alpha/Kconfig
> @@ -45,6 +45,14 @@ config GENERIC_CALIBRATE_DELAY
>  	bool
>  	default y
> 
> +config GENERIC_TIME
> +	bool
> +	default y
> +
> +config ARCH_USES_GETTIMEOFFSET
> +	bool
> +	default y
> +
>  config ZONE_DMA
>  	bool
>  	default y
> diff --git a/arch/alpha/kernel/time.c b/arch/alpha/kernel/time.c
> index b04e2cb..8f7eb6d 100644
> --- a/arch/alpha/kernel/time.c
> +++ b/arch/alpha/kernel/time.c
> @@ -408,28 +408,17 @@ time_init(void)
>   * part.  So we can't do the "find absolute time in terms of 
> cycles" thing  * that the other ports do.  */ -void -
> do_gettimeofday(struct timeval *tv) +u32 arch_gettimeoffset(void) { -
> 	unsigned long flags; -	unsigned long sec, usec, seq; 	unsigned long 
> delta_cycles, delta_usec, partial_tick;
> 
> -	do {
> -		seq = read_seqbegin_irqsave(&xtime_lock, flags);
> -
> -		delta_cycles = rpcc() - state.last_time;
> -		sec = xtime.tv_sec;
> -		usec = (xtime.tv_nsec / 1000);
> -		partial_tick = state.partial_tick;
> -
> -	} while (read_seqretry_irqrestore(&xtime_lock, seq, flags));
> -
>  #ifdef CONFIG_SMP
>  	/* Until and unless we figure out how to get cpu cycle counters
>  	   in sync and keep them there, we can't use the rpcc tricks.  */
>  	delta_usec = 0;
>  #else
> +	delta_cycles = rpcc() - state.last_time;
> +	partial_tick = state.partial_tick;
>  	/*
>  	 * usec = cycles * ticks_per_cycle * 2**48 * 1e6 / (2**48 * ticks)
>  	 *	= cycles * (s_t_p_c) * 1e6 / (2**48 * ticks)
> @@ -448,62 +437,9 @@ do_gettimeofday(struct timeval *tv)
>  	delta_usec = ((delta_usec / ((1UL << (FIX_SHIFT-6-1)) * HZ)) + 1)
>  / 2; #endif
> 
> -	usec += delta_usec;
> -	if (usec >= 1000000) {
> -		sec += 1;
> -		usec -= 1000000;
> -	}
> -
> -	tv->tv_sec = sec;
> -	tv->tv_usec = usec;
> -}
> -
> -EXPORT_SYMBOL(do_gettimeofday);
> -
> -int
> -do_settimeofday(struct timespec *tv)
> -{
> -	time_t wtm_sec, sec = tv->tv_sec;
> -	long wtm_nsec, nsec = tv->tv_nsec;
> -	unsigned long delta_nsec;
> -
> -	if ((unsigned long)tv->tv_nsec >= NSEC_PER_SEC)
> -		return -EINVAL;
> -
> -	write_seqlock_irq(&xtime_lock);
> -
> -	/* The offset that is added into time in do_gettimeofday above
> -	   must be subtracted out here to keep a coherent view of the
> -	   time.  Without this, a full-tick error is possible.  */
> -
> -#ifdef CONFIG_SMP
> -	delta_nsec = 0;
> -#else
> -	delta_nsec = rpcc() - state.last_time;
> -	delta_nsec = (delta_nsec * state.scaled_ticks_per_cycle 
> -		      + state.partial_tick) * 15625;
> -	delta_nsec = ((delta_nsec / ((1UL << (FIX_SHIFT-6-1)) * HZ)) + 1)
>  / 2; -	delta_nsec *= 1000; -#endif
> -
> -	nsec -= delta_nsec;
> -
> -	wtm_sec  = wall_to_monotonic.tv_sec + (xtime.tv_sec - sec);
> -	wtm_nsec = wall_to_monotonic.tv_nsec + (xtime.tv_nsec - nsec);
> -
> -	set_normalized_timespec(&xtime, sec, nsec);
> -	set_normalized_timespec(&wall_to_monotonic, wtm_sec, wtm_nsec);
> -
> -	ntp_clear();
> -
> -	write_sequnlock_irq(&xtime_lock);
> -	clock_was_set();
> -	return 0;
> +	return delta_usec * 1000;
>  }
> 
> -EXPORT_SYMBOL(do_settimeofday);
> -
> -
>  /*
>   * In order to set the CMOS clock precisely, set_rtc_mmss has to be
>   * called 500 ms after the second nowtime has started, because when
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-
> alpha" in the body of a message to majordomo@vger.kernel.org More 
> majordomo info at  http://vger.kernel.org/majordomo-info.html
> 
> -- 
> This message has been scanned for viruses and
> dangerous content by MailScanner, and is
> believed to be clean.
------- End of Original Message -------

[PATCH] Convert m32r to use arch_gettimeoffset()

[john stultz]

This patch converts m32r to use GENERIC_TIME via the arch_getoffset()
infrastructure, reducing the amount of arch specific code we need to
maintain.

I also noted that m32r doesn't seem to be taking the xtime write lock
before calling do_timer()!  That looks like a pretty bad bug to me. If
folks agree, let me know and I can move the lock grab to the correct
spot.

This patch applies on top of Linus' current -git tree (pre 2.6.31-rc1).

I've taken my best swing at converting this, but I'm not 100% confident
I got it right. It does cross compile, so that's hopefully a good sign.
Any assistance from arch maintainers or testers to get this merged would
be great.

thanks
-john

Signed-off-by: John Stultz <johnstul@us.ibm.com>
---
 Kconfig       |    6 ++++
 kernel/time.c |   74 ++--------------------------------------------------------
 2 files changed, 9 insertions(+), 71 deletions(-)

diff --git a/arch/m32r/Kconfig b/arch/m32r/Kconfig
index cabba33..c41234f 100644
--- a/arch/m32r/Kconfig
+++ b/arch/m32r/Kconfig
@@ -41,6 +41,12 @@ config HZ
 	int
 	default 100
 
+config GENERIC_TIME
+	def_bool y
+
+config ARCH_USES_GETTIMEOFFSET
+	def_bool y
+
 source "init/Kconfig"
 
 source "kernel/Kconfig.freezer"
diff --git a/arch/m32r/kernel/time.c b/arch/m32r/kernel/time.c
index cada3ba..ba61c4c 100644
--- a/arch/m32r/kernel/time.c
+++ b/arch/m32r/kernel/time.c
@@ -48,7 +48,7 @@ extern void smp_local_timer_interrupt(void);
 
 static unsigned long latch;
 
-static unsigned long do_gettimeoffset(void)
+u32 arch_gettimeoffset(void)
 {
 	unsigned long  elapsed_time = 0;  /* [us] */
 
@@ -93,79 +93,10 @@ static unsigned long do_gettimeoffset(void)
 #error no chip configuration
 #endif
 
-	return elapsed_time;
+	return elapsed_time * 1000;
 }
 
 /*
- * This version of gettimeofday has near microsecond resolution.
- */
-void do_gettimeofday(struct timeval *tv)
-{
-	unsigned long seq;
-	unsigned long usec, sec;
-	unsigned long max_ntp_tick = tick_usec - tickadj;
-
-	do {
-		seq = read_seqbegin(&xtime_lock);
-
-		usec = do_gettimeoffset();
-
-		/*
-		 * If time_adjust is negative then NTP is slowing the clock
-		 * so make sure not to go into next possible interval.
-		 * Better to lose some accuracy than have time go backwards..
-		 */
-		if (unlikely(time_adjust < 0))
-			usec = min(usec, max_ntp_tick);
-
-		sec = xtime.tv_sec;
-		usec += (xtime.tv_nsec / 1000);
-	} while (read_seqretry(&xtime_lock, seq));
-
-	while (usec >= 1000000) {
-		usec -= 1000000;
-		sec++;
-	}
-
-	tv->tv_sec = sec;
-	tv->tv_usec = usec;
-}
-
-EXPORT_SYMBOL(do_gettimeofday);
-
-int do_settimeofday(struct timespec *tv)
-{
-	time_t wtm_sec, sec = tv->tv_sec;
- 	long wtm_nsec, nsec = tv->tv_nsec;
-
-	if ((unsigned long)tv->tv_nsec >= NSEC_PER_SEC)
-		return -EINVAL;
-
-	write_seqlock_irq(&xtime_lock);
-	/*
-	 * This is revolting. We need to set "xtime" correctly. However, the
-	 * value in this location is the value at the most recent update of
-	 * wall time.  Discover what correction gettimeofday() would have
-	 * made, and then undo it!
-	 */
-	nsec -= do_gettimeoffset() * NSEC_PER_USEC;
-
-	wtm_sec = wall_to_monotonic.tv_sec + (xtime.tv_sec - sec);
-	wtm_nsec = wall_to_monotonic.tv_nsec + (xtime.tv_nsec - nsec);
-
-	set_normalized_timespec(&xtime, sec, nsec);
-	set_normalized_timespec(&wall_to_monotonic, wtm_sec, wtm_nsec);
-
-	ntp_clear();
-	write_sequnlock_irq(&xtime_lock);
-	clock_was_set();
-
-	return 0;
-}
-
-EXPORT_SYMBOL(do_settimeofday);
-
-/*
  * In order to set the CMOS clock precisely, set_rtc_mmss has to be
  * called 500 ms after the second nowtime has started, because when
  * nowtime is written into the registers of the CMOS clock, it will
@@ -192,6 +123,7 @@ static irqreturn_t timer_interrupt(int irq, void *dev_id)
 #ifndef CONFIG_SMP
 	profile_tick(CPU_PROFILING);
 #endif
+	/* XXX FIXME. Uh, the xtime_lock should be held here, no? */
 	do_timer(1);
 
 #ifndef CONFIG_SMP

Re: [PATCH] Convert alpha to use arch_gettimeoffset()

[Richard Henderson]

john stultz wrote:
> I suspect the alpha arch could even be further improved to provide and
> rpcc() based clocksource, but not having the hardware, I don't feel
> comfortable attempting the more complicated conversion (but I'd be glad
> to help if anyone else is interested).

Unfortunately this isn't possible.  The rpcc value is only 32-bits wide,
and so rolls over in approx 3 to 4 seconds.  Thus all the contortions to
try and use the truncated value for sub 1 second adjustments.

> +	delta_cycles = rpcc() - state.last_time;
> +	partial_tick = state.partial_tick;
>  	/*
>  	 * usec = cycles * ticks_per_cycle * 2**48 * 1e6 / (2**48 * ticks)
>  	 *	= cycles * (s_t_p_c) * 1e6 / (2**48 * ticks)
> @@ -448,62 +437,9 @@ do_gettimeofday(struct timeval *tv)
>  	delta_usec = ((delta_usec / ((1UL << (FIX_SHIFT-6-1)) * HZ)) + 1) / 2;
>  #endif
...
> +	return delta_usec * 1000;

If we're going to be computing nsec now, we might as well adjust
the computation here.

   nsec = cycles * ticks_per_cycle * 1e9 / HZ
        = cycles * s_t_p_c * 1e9 / (2**48 * HZ)
        = cycles * s_t_p_c * 1953125 / (2**39 * HZ)

   delta_nsec = delta_cycles*state.scaled_ticks_per_cycle + partial_tick;
   delta_nsec *= 1953125
   delta_nsec /= HZ << (FIX_SHIFT - 9 - 1);
   delta_nsec = (delta_nsec + 1) / 2;

As far as I can tell, the range of the dividend is about 7e17,
which still fits in an unsigned long.


r~

Re: [PATCH] Convert alpha to use arch_gettimeoffset()

[john stultz]

On Fri, 2009-06-12 at 00:03 -0700, Richard Henderson wrote:
> john stultz wrote:
> > I suspect the alpha arch could even be further improved to provide and
> > rpcc() based clocksource, but not having the hardware, I don't feel
> > comfortable attempting the more complicated conversion (but I'd be glad
> > to help if anyone else is interested).
> 
> Unfortunately this isn't possible.  The rpcc value is only 32-bits wide,
> and so rolls over in approx 3 to 4 seconds.  Thus all the contortions to
> try and use the truncated value for sub 1 second adjustments.

That's actually ok. The ACPI PM clocksource on x86 is only 24 bits wide
and wraps every 5 seconds. The generic timekeeping code handles the
wrapping and accumulation for you.


> > +	delta_cycles = rpcc() - state.last_time;
> > +	partial_tick = state.partial_tick;
> >  	/*
> >  	 * usec = cycles * ticks_per_cycle * 2**48 * 1e6 / (2**48 * ticks)
> >  	 *	= cycles * (s_t_p_c) * 1e6 / (2**48 * ticks)
> > @@ -448,62 +437,9 @@ do_gettimeofday(struct timeval *tv)
> >  	delta_usec = ((delta_usec / ((1UL << (FIX_SHIFT-6-1)) * HZ)) + 1) / 2;
> >  #endif
> ...
> > +	return delta_usec * 1000;
> 
> If we're going to be computing nsec now, we might as well adjust
> the computation here.
> 
>    nsec = cycles * ticks_per_cycle * 1e9 / HZ
>         = cycles * s_t_p_c * 1e9 / (2**48 * HZ)
>         = cycles * s_t_p_c * 1953125 / (2**39 * HZ)
> 
>    delta_nsec = delta_cycles*state.scaled_ticks_per_cycle + partial_tick;
>    delta_nsec *= 1953125
>    delta_nsec /= HZ << (FIX_SHIFT - 9 - 1);
>    delta_nsec = (delta_nsec + 1) / 2;
> 
> As far as I can tell, the range of the dividend is about 7e17,
> which still fits in an unsigned long.

That seems like a good optimization, although I'm hesitant to submit
anything but structural, but equivalent code changes since I don't have
the hardware to test with.

Might you create a patch that does this conversion and apply it on top
of mine? 

thanks
-john

Re: [PATCH] Convert alpha to use arch_gettimeoffset()

[Richard Henderson]

john stultz wrote:
> On Fri, 2009-06-12 at 00:03 -0700, Richard Henderson wrote:
>   
>> john stultz wrote:
>>     
>>> I suspect the alpha arch could even be further improved to provide and
>>> rpcc() based clocksource, but not having the hardware, I don't feel
>>> comfortable attempting the more complicated conversion (but I'd be glad
>>> to help if anyone else is interested).
>>>       
>> Unfortunately this isn't possible.  The rpcc value is only 32-bits wide,
>> and so rolls over in approx 3 to 4 seconds.  Thus all the contortions to
>> try and use the truncated value for sub 1 second adjustments.
>>     
>
> That's actually ok. The ACPI PM clocksource on x86 is only 24 bits wide
> and wraps every 5 seconds. The generic timekeeping code handles the
> wrapping and accumulation for you.
>   
Interesting.
> Might you create a patch that does this conversion and apply it on top
> of mine? 
Unfortunately I don't have working hardware anymore.  Perhaps Ivan will 
be able to help here.


r~

Re: [PATCH] Convert alpha to use arch_gettimeoffset()

[Ivan Kokshaysky]

On Tue, Jun 16, 2009 at 10:45:16AM -0700, Richard Henderson wrote:
> Unfortunately I don't have working hardware anymore.  Perhaps Ivan will be 
> able to help here.

I can give it a try next week (I'm away from hardware currently).

Ivan.

[RESEND][PATCH] Convert m32r to use arch_gettimeoffset()

[john stultz]

This patch converts m32r to use GENERIC_TIME via the arch_getoffset()
infrastructure, reducing the amount of arch specific code we need to
maintain.

I also noted that m32r doesn't seem to be taking the xtime write lock
before calling do_timer()!  That looks like a pretty bad bug to me. If
folks agree, let me know and I can move the lock grab to the correct
spot.

This patch applies on top of Linus' current -git tree (2.6.31-rc3).

I've taken my best swing at converting this, but I'm not 100% confident
I got it right. My cross-compiler is now out of date (gcc4.2) so I
wasn't able to  check if it compiled. Any assistance from arch
maintainers or testers to get this merged would be great.

thanks
-john

Signed-off-by: John Stultz <johnstul@us.ibm.com>
---

 Kconfig       |    6 ++++
 kernel/time.c |   74 ++--------------------------------------------------------
 2 files changed, 9 insertions(+), 71 deletions(-)

diff --git a/arch/m32r/Kconfig b/arch/m32r/Kconfig
index cabba33..c41234f 100644
--- a/arch/m32r/Kconfig
+++ b/arch/m32r/Kconfig
@@ -41,6 +41,12 @@ config HZ
 	int
 	default 100
 
+config GENERIC_TIME
+	def_bool y
+
+config ARCH_USES_GETTIMEOFFSET
+	def_bool y
+
 source "init/Kconfig"
 
 source "kernel/Kconfig.freezer"
diff --git a/arch/m32r/kernel/time.c b/arch/m32r/kernel/time.c
index cada3ba..ba61c4c 100644
--- a/arch/m32r/kernel/time.c
+++ b/arch/m32r/kernel/time.c
@@ -48,7 +48,7 @@ extern void smp_local_timer_interrupt(void);
 
 static unsigned long latch;
 
-static unsigned long do_gettimeoffset(void)
+u32 arch_gettimeoffset(void)
 {
 	unsigned long  elapsed_time = 0;  /* [us] */
 
@@ -93,79 +93,10 @@ static unsigned long do_gettimeoffset(void)
 #error no chip configuration
 #endif
 
-	return elapsed_time;
+	return elapsed_time * 1000;
 }
 
 /*
- * This version of gettimeofday has near microsecond resolution.
- */
-void do_gettimeofday(struct timeval *tv)
-{
-	unsigned long seq;
-	unsigned long usec, sec;
-	unsigned long max_ntp_tick = tick_usec - tickadj;
-
-	do {
-		seq = read_seqbegin(&xtime_lock);
-
-		usec = do_gettimeoffset();
-
-		/*
-		 * If time_adjust is negative then NTP is slowing the clock
-		 * so make sure not to go into next possible interval.
-		 * Better to lose some accuracy than have time go backwards..
-		 */
-		if (unlikely(time_adjust < 0))
-			usec = min(usec, max_ntp_tick);
-
-		sec = xtime.tv_sec;
-		usec += (xtime.tv_nsec / 1000);
-	} while (read_seqretry(&xtime_lock, seq));
-
-	while (usec >= 1000000) {
-		usec -= 1000000;
-		sec++;
-	}
-
-	tv->tv_sec = sec;
-	tv->tv_usec = usec;
-}
-
-EXPORT_SYMBOL(do_gettimeofday);
-
-int do_settimeofday(struct timespec *tv)
-{
-	time_t wtm_sec, sec = tv->tv_sec;
- 	long wtm_nsec, nsec = tv->tv_nsec;
-
-	if ((unsigned long)tv->tv_nsec >= NSEC_PER_SEC)
-		return -EINVAL;
-
-	write_seqlock_irq(&xtime_lock);
-	/*
-	 * This is revolting. We need to set "xtime" correctly. However, the
-	 * value in this location is the value at the most recent update of
-	 * wall time.  Discover what correction gettimeofday() would have
-	 * made, and then undo it!
-	 */
-	nsec -= do_gettimeoffset() * NSEC_PER_USEC;
-
-	wtm_sec = wall_to_monotonic.tv_sec + (xtime.tv_sec - sec);
-	wtm_nsec = wall_to_monotonic.tv_nsec + (xtime.tv_nsec - nsec);
-
-	set_normalized_timespec(&xtime, sec, nsec);
-	set_normalized_timespec(&wall_to_monotonic, wtm_sec, wtm_nsec);
-
-	ntp_clear();
-	write_sequnlock_irq(&xtime_lock);
-	clock_was_set();
-
-	return 0;
-}
-
-EXPORT_SYMBOL(do_settimeofday);
-
-/*
  * In order to set the CMOS clock precisely, set_rtc_mmss has to be
  * called 500 ms after the second nowtime has started, because when
  * nowtime is written into the registers of the CMOS clock, it will
@@ -192,6 +123,7 @@ static irqreturn_t timer_interrupt(int irq, void *dev_id)
 #ifndef CONFIG_SMP
 	profile_tick(CPU_PROFILING);
 #endif
+	/* XXX FIXME. Uh, the xtime_lock should be held here, no? */
 	do_timer(1);
 
 #ifndef CONFIG_SMP

[PATCH] Convert cris to use arch_gettimeoffset()

[john stultz]

This patch converts cris to use GENERIC_TIME via the arch_getoffset()
infrastructure, reducing the amount of arch specific code we need to
maintain.

This patch applies on top of Linus' current -git tree (2.6.31-rc3).

I've taken my best swing at converting this, but I'm not 100% confident
I got it right. My cross-compiler is now out of date (gcc4.2) so I
wasn't able to  check if it compiled. Any assistance from arch
maintainers or testers to get this merged would be great.

thanks
-john

Signed-off-by: John Stultz <johnstul@us.ibm.com>
---
 Kconfig       |    6 +++++
 kernel/time.c |   68 +---------------------------------------------------------
 2 files changed, 8 insertions(+), 66 deletions(-)

diff --git a/arch/cris/Kconfig b/arch/cris/Kconfig
index 7adac38..059eac6 100644
--- a/arch/cris/Kconfig
+++ b/arch/cris/Kconfig
@@ -20,6 +20,12 @@ config RWSEM_GENERIC_SPINLOCK
 config RWSEM_XCHGADD_ALGORITHM
 	bool
 
+config GENERIC_TIME
+	def_bool y
+
+config ARCH_USES_GETTIMEOFFSET
+	def_bool y
+
 config GENERIC_IOMAP
        bool
        default y
diff --git a/arch/cris/kernel/time.c b/arch/cris/kernel/time.c
index 074fe7d..a05dd31 100644
--- a/arch/cris/kernel/time.c
+++ b/arch/cris/kernel/time.c
@@ -42,75 +42,11 @@ unsigned long loops_per_usec;
 extern unsigned long do_slow_gettimeoffset(void);
 static unsigned long (*do_gettimeoffset)(void) = do_slow_gettimeoffset;
 
-/*
- * This version of gettimeofday has near microsecond resolution.
- *
- * Note: Division is quite slow on CRIS and do_gettimeofday is called
- *       rather often. Maybe we should do some kind of approximation here
- *       (a naive approximation would be to divide by 1024).
- */
-void do_gettimeofday(struct timeval *tv)
-{
-	unsigned long flags;
-	signed long usec, sec;
-	local_irq_save(flags);
-	usec = do_gettimeoffset();
-
-        /*
-	 * If time_adjust is negative then NTP is slowing the clock
-	 * so make sure not to go into next possible interval.
-	 * Better to lose some accuracy than have time go backwards..
-	 */
-	if (unlikely(time_adjust < 0) && usec > tickadj)
-		usec = tickadj;
-
-	sec = xtime.tv_sec;
-	usec += xtime.tv_nsec / 1000;
-	local_irq_restore(flags);
-
-	while (usec >= 1000000) {
-		usec -= 1000000;
-		sec++;
-	}
-
-	tv->tv_sec = sec;
-	tv->tv_usec = usec;
-}
-
-EXPORT_SYMBOL(do_gettimeofday);
-
-int do_settimeofday(struct timespec *tv)
+u32 arch_gettimeoffset(void)
 {
-	time_t wtm_sec, sec = tv->tv_sec;
-	long wtm_nsec, nsec = tv->tv_nsec;
-
-	if ((unsigned long)tv->tv_nsec >= NSEC_PER_SEC)
-		return -EINVAL;
-
-	write_seqlock_irq(&xtime_lock);
-	/*
-	 * This is revolting. We need to set "xtime" correctly. However, the
-	 * value in this location is the value at the most recent update of
-	 * wall time.  Discover what correction gettimeofday() would have
-	 * made, and then undo it!
-	 */
-	nsec -= do_gettimeoffset() * NSEC_PER_USEC;
-
-	wtm_sec  = wall_to_monotonic.tv_sec + (xtime.tv_sec - sec);
-	wtm_nsec = wall_to_monotonic.tv_nsec + (xtime.tv_nsec - nsec);
-
-	set_normalized_timespec(&xtime, sec, nsec);
-	set_normalized_timespec(&wall_to_monotonic, wtm_sec, wtm_nsec);
-
-	ntp_clear();
-	write_sequnlock_irq(&xtime_lock);
-	clock_was_set();
-	return 0;
+	return do_gettimeoffset() * 1000;
 }
 
-EXPORT_SYMBOL(do_settimeofday);
-
-
 /*
  * BUG: This routine does not handle hour overflow properly; it just
  *      sets the minutes. Usually you'll only notice that after reboot!

[PATCH] Convert m68k to use arch_gettimeoffset()

[john stultz]

This patch converts m68k to use GENERIC_TIME via the arch_getoffset()
infrastructure, reducing the amount of arch specific code we need to
maintain.

This patch applies on top of Linus' current -git tree (2.6.31-rc3).

I've taken my best swing at converting this, but I'm not 100% confident
I got it right. My cross-compiler is now out of date (gcc4.2) so I
wasn't able to  check if it compiled. Any assistance from arch
maintainers or testers to get this merged would be great.

thanks
-john

Signed-off-by: John Stultz <johnstul@us.ibm.com>
---
 Kconfig       |    6 ++++
 kernel/time.c |   70 +---------------------------------------------------------
 2 files changed, 8 insertions(+), 68 deletions(-)

diff --git a/arch/m68k/Kconfig b/arch/m68k/Kconfig
index fb87c08..29dd848 100644
--- a/arch/m68k/Kconfig
+++ b/arch/m68k/Kconfig
@@ -58,6 +58,12 @@ config HZ
 	int
 	default 100
 
+config GENERIC_TIME
+	def_bool y
+
+config ARCH_USES_GETTIMEOFFSET
+	def_bool y
+
 mainmenu "Linux/68k Kernel Configuration"
 
 source "init/Kconfig"
diff --git a/arch/m68k/kernel/time.c b/arch/m68k/kernel/time.c
index 54d9807..17dc2a3 100644
--- a/arch/m68k/kernel/time.c
+++ b/arch/m68k/kernel/time.c
@@ -91,77 +91,11 @@ void __init time_init(void)
 	mach_sched_init(timer_interrupt);
 }
 
-/*
- * This version of gettimeofday has near microsecond resolution.
- */
-void do_gettimeofday(struct timeval *tv)
+u32 arch_gettimeoffset(void)
 {
-	unsigned long flags;
-	unsigned long seq;
-	unsigned long usec, sec;
-	unsigned long max_ntp_tick = tick_usec - tickadj;
-
-	do {
-		seq = read_seqbegin_irqsave(&xtime_lock, flags);
-
-		usec = mach_gettimeoffset();
-
-		/*
-		 * If time_adjust is negative then NTP is slowing the clock
-		 * so make sure not to go into next possible interval.
-		 * Better to lose some accuracy than have time go backwards..
-		 */
-		if (unlikely(time_adjust < 0))
-			usec = min(usec, max_ntp_tick);
-
-		sec = xtime.tv_sec;
-		usec += xtime.tv_nsec/1000;
-	} while (read_seqretry_irqrestore(&xtime_lock, seq, flags));
-
-
-	while (usec >= 1000000) {
-		usec -= 1000000;
-		sec++;
-	}
-
-	tv->tv_sec = sec;
-	tv->tv_usec = usec;
-}
-
-EXPORT_SYMBOL(do_gettimeofday);
-
-int do_settimeofday(struct timespec *tv)
-{
-	time_t wtm_sec, sec = tv->tv_sec;
-	long wtm_nsec, nsec = tv->tv_nsec;
-
-	if ((unsigned long)tv->tv_nsec >= NSEC_PER_SEC)
-		return -EINVAL;
-
-	write_seqlock_irq(&xtime_lock);
-	/* This is revolting. We need to set the xtime.tv_nsec
-	 * correctly. However, the value in this location is
-	 * is value at the last tick.
-	 * Discover what correction gettimeofday
-	 * would have done, and then undo it!
-	 */
-	nsec -= 1000 * mach_gettimeoffset();
-
-	wtm_sec  = wall_to_monotonic.tv_sec + (xtime.tv_sec - sec);
-	wtm_nsec = wall_to_monotonic.tv_nsec + (xtime.tv_nsec - nsec);
-
-	set_normalized_timespec(&xtime, sec, nsec);
-	set_normalized_timespec(&wall_to_monotonic, wtm_sec, wtm_nsec);
-
-	ntp_clear();
-	write_sequnlock_irq(&xtime_lock);
-	clock_was_set();
-	return 0;
+	return mach_gettimeoffset() * 1000;
 }
 
-EXPORT_SYMBOL(do_settimeofday);
-
-
 static int __init rtc_init(void)
 {
 	struct platform_device *pdev;

[PATCH] Convert blackfin to use arch_gettimeoffset()

[john stultz]

This patch converts blackfin to use GENERIC_TIME via the arch_getoffset()
infrastructure, reducing the amount of arch specific code we need to
maintain.

This patch applies on top of Linus' current -git tree (2.6.31-rc3).

I've taken my best swing at converting this, but I'm not 100% confident
I got it right. My cross-compiler is now out of date (gcc4.2) so I
wasn't able to  check if it compiled. Any assistance from arch
maintainers or testers to get this merged would be great.

thanks
-john

Signed-off-by: John Stultz <johnstul@us.ibm.com>
---
 Kconfig       |   16 +++++++++++---
 kernel/time.c |   63 +---------------------------------------------------------
 2 files changed, 14 insertions(+), 65 deletions(-)

diff --git a/arch/blackfin/Kconfig b/arch/blackfin/Kconfig
index 7faa2f5..263b133 100644
--- a/arch/blackfin/Kconfig
+++ b/arch/blackfin/Kconfig
@@ -229,7 +229,7 @@ endchoice
 
 config SMP
 	depends on BF561
-	select GENERIC_TIME
+	select NO_ARCH_GETTIMEOFFSET
 	bool "Symmetric multi-processing support"
 	---help---
 	  This enables support for systems with more than one CPU,
@@ -612,12 +612,13 @@ comment "Kernel Timer/Scheduler"
 source kernel/Kconfig.hz
 
 config GENERIC_TIME
-	bool "Generic time"
-	default y
+	def_bool y
+
+config NO_ARCH_GETTIMEOFFSET
+	def_bool n
 
 config GENERIC_CLOCKEVENTS
 	bool "Generic clock events"
-	depends on GENERIC_TIME
 	default y
 
 choice
@@ -639,6 +640,7 @@ config CYCLES_CLOCKSOURCE
 	depends on GENERIC_CLOCKEVENTS
 	depends on !BFIN_SCRATCH_REG_CYCLES
 	depends on !SMP
+	select NO_ARCH_GETTIMEOFFSET
 	help
 	  If you say Y here, you will enable support for using the 'cycles'
 	  registers as a clock source.  Doing so means you will be unable to
@@ -649,9 +651,15 @@ config CYCLES_CLOCKSOURCE
 config GPTMR0_CLOCKSOURCE
 	bool "Use GPTimer0 as a clocksource (higher rating)"
 	select BFIN_GPTIMERS
+	select NO_ARCH_GETTIMEOFFSET
 	depends on GENERIC_CLOCKEVENTS
 	depends on !TICKSOURCE_GPTMR0
 
+config ARCH_USES_GETTIMEOFFSET
+	depends on !NO_ARCH_GETTIMEOFFSET
+	def_bool y
+
+
 source kernel/time/Kconfig
 
 comment "Misc"
diff --git a/arch/blackfin/kernel/time.c b/arch/blackfin/kernel/time.c
index adb54aa..e5069fe 100644
--- a/arch/blackfin/kernel/time.c
+++ b/arch/blackfin/kernel/time.c
@@ -81,11 +81,11 @@ time_sched_init(irqreturn_t(*timer_routine) (int, void *))
 #endif
 }
 
+#ifdef CONFIG_ARCH_USES_GETTIMEOFFSET
 /*
  * Should return useconds since last timer tick
  */
-#ifndef CONFIG_GENERIC_TIME
-static unsigned long gettimeoffset(void)
+u32 arch_gettimeoffset(void)
 {
 	unsigned long offset;
 	unsigned long clocks_per_jiffy;
@@ -184,65 +184,6 @@ void __init time_init(void)
 	time_sched_init(timer_interrupt);
 }
 
-#ifndef CONFIG_GENERIC_TIME
-void do_gettimeofday(struct timeval *tv)
-{
-	unsigned long flags;
-	unsigned long seq;
-	unsigned long usec, sec;
-
-	do {
-		seq = read_seqbegin_irqsave(&xtime_lock, flags);
-		usec = gettimeoffset();
-		sec = xtime.tv_sec;
-		usec += (xtime.tv_nsec / NSEC_PER_USEC);
-	}
-	while (read_seqretry_irqrestore(&xtime_lock, seq, flags));
-
-	while (usec >= USEC_PER_SEC) {
-		usec -= USEC_PER_SEC;
-		sec++;
-	}
-
-	tv->tv_sec = sec;
-	tv->tv_usec = usec;
-}
-EXPORT_SYMBOL(do_gettimeofday);
-
-int do_settimeofday(struct timespec *tv)
-{
-	time_t wtm_sec, sec = tv->tv_sec;
-	long wtm_nsec, nsec = tv->tv_nsec;
-
-	if ((unsigned long)tv->tv_nsec >= NSEC_PER_SEC)
-		return -EINVAL;
-
-	write_seqlock_irq(&xtime_lock);
-	/*
-	 * This is revolting. We need to set the xtime.tv_usec
-	 * correctly. However, the value in this location is
-	 * is value at the last tick.
-	 * Discover what correction gettimeofday
-	 * would have done, and then undo it!
-	 */
-	nsec -= (gettimeoffset() * NSEC_PER_USEC);
-
-	wtm_sec = wall_to_monotonic.tv_sec + (xtime.tv_sec - sec);
-	wtm_nsec = wall_to_monotonic.tv_nsec + (xtime.tv_nsec - nsec);
-
-	set_normalized_timespec(&xtime, sec, nsec);
-	set_normalized_timespec(&wall_to_monotonic, wtm_sec, wtm_nsec);
-
-	ntp_clear();
-
-	write_sequnlock_irq(&xtime_lock);
-	clock_was_set();
-
-	return 0;
-}
-EXPORT_SYMBOL(do_settimeofday);
-#endif /* !CONFIG_GENERIC_TIME */
-
 /*
  * Scheduler clock - returns current time in nanosec units.
  */

Re: [PATCH] Convert blackfin to use arch_gettimeoffset()

[Mike Frysinger]

On Thu, Jul 16, 2009 at 20:25, john stultz wrote:
> This patch converts blackfin to use GENERIC_TIME via the arch_getoffset()
> infrastructure, reducing the amount of arch specific code we need to
> maintain.
>
> This patch applies on top of Linus' current -git tree (2.6.31-rc3).
>
> I've taken my best swing at converting this, but I'm not 100% confident
> I got it right. My cross-compiler is now out of date (gcc4.2) so I
> wasn't able to  check if it compiled. Any assistance from arch
> maintainers or testers to get this merged would be great.

something seems to be off as the Blackfin defconfig now does:
  LD      .tmp_vmlinux1
kernel/built-in.o: In function `clocksource_forward_now':
kernel/time/timekeeping.c:82: undefined reference to `arch_gettimeoffset'
kernel/built-in.o: In function `getnstimeofday':
kernel/time/timekeeping.c:119: undefined reference to `arch_gettimeoffset'
make: *** [.tmp_vmlinux1] Error 1

most likely because the Blackfin time.c is the old stuff
(non-clocksource) while time-ts.c is the newer stuff (clocksource) and
this change only touches time.c
-mike

Re: [PATCH] Convert blackfin to use arch_gettimeoffset()

[john stultz]

On Thu, 2009-07-16 at 22:26 -0400, Mike Frysinger wrote:
> On Thu, Jul 16, 2009 at 20:25, john stultz wrote:
> > This patch converts blackfin to use GENERIC_TIME via the arch_getoffset()
> > infrastructure, reducing the amount of arch specific code we need to
> > maintain.
> >
> > This patch applies on top of Linus' current -git tree (2.6.31-rc3).
> >
> > I've taken my best swing at converting this, but I'm not 100% confident
> > I got it right. My cross-compiler is now out of date (gcc4.2) so I
> > wasn't able to  check if it compiled. Any assistance from arch
> > maintainers or testers to get this merged would be great.
> 
> something seems to be off as the Blackfin defconfig now does:
>   LD      .tmp_vmlinux1
> kernel/built-in.o: In function `clocksource_forward_now':
> kernel/time/timekeeping.c:82: undefined reference to `arch_gettimeoffset'
> kernel/built-in.o: In function `getnstimeofday':
> kernel/time/timekeeping.c:119: undefined reference to `arch_gettimeoffset'
> make: *** [.tmp_vmlinux1] Error 1
> 
> most likely because the Blackfin time.c is the old stuff
> (non-clocksource) while time-ts.c is the newer stuff (clocksource) and
> this change only touches time.c

Hrm.. Ok. I'll take a look. Thanks for testing and the pointer!

-john

Re: [PATCH] Convert blackfin to use arch_gettimeoffset()

[john stultz]

On Thu, 2009-07-16 at 22:26 -0400, Mike Frysinger wrote:
> On Thu, Jul 16, 2009 at 20:25, john stultz wrote:
> > This patch converts blackfin to use GENERIC_TIME via the arch_getoffset()
> > infrastructure, reducing the amount of arch specific code we need to
> > maintain.
> >
> > This patch applies on top of Linus' current -git tree (2.6.31-rc3).
> >
> > I've taken my best swing at converting this, but I'm not 100% confident
> > I got it right. My cross-compiler is now out of date (gcc4.2) so I
> > wasn't able to  check if it compiled. Any assistance from arch
> > maintainers or testers to get this merged would be great.
> 
> something seems to be off as the Blackfin defconfig now does:
>   LD      .tmp_vmlinux1
> kernel/built-in.o: In function `clocksource_forward_now':
> kernel/time/timekeeping.c:82: undefined reference to `arch_gettimeoffset'
> kernel/built-in.o: In function `getnstimeofday':
> kernel/time/timekeeping.c:119: undefined reference to `arch_gettimeoffset'
> make: *** [.tmp_vmlinux1] Error 1
> 
> most likely because the Blackfin time.c is the old stuff
> (non-clocksource) while time-ts.c is the newer stuff (clocksource) and
> this change only touches time.c

Ok. So I guess I missed that time.c isn't built if GENERIC_CLOCKEVENTS
isn't enabled, so ARCH_HAS_GETTIMEOFFSET needs to depend on !
GENERIC_CLOCKEVENTS.

Hopefully I've got it cleared up now. Let me know if I'm still causing
trouble.

thanks
-john



This patch converts blackfin to use GENERIC_TIME via the
arch_getoffset() infrastructure, reducing the amount of arch specific
code we need to maintain.

This patch applies on top of Linus' current -git tree (2.6.31-rc3).

I've taken my best swing at converting this, but I'm not 100% confident
I got it right. My cross-compiler is now out of date (gcc4.2) so I
wasn't able to  check if it compiled. Any assistance from arch
maintainers or testers to get this merged would be great.

Signed-off-by: John Stultz <johnstul@us.ibm.com>
---

diff --git a/arch/blackfin/Kconfig b/arch/blackfin/Kconfig
index 7faa2f5..c8f5c3d 100644
--- a/arch/blackfin/Kconfig
+++ b/arch/blackfin/Kconfig
@@ -229,7 +229,7 @@ endchoice
 
 config SMP
 	depends on BF561
-	select GENERIC_TIME
+	select GENERIC_CLOCKEVENTS
 	bool "Symmetric multi-processing support"
 	---help---
 	  This enables support for systems with more than one CPU,
@@ -612,12 +612,10 @@ comment "Kernel Timer/Scheduler"
 source kernel/Kconfig.hz
 
 config GENERIC_TIME
-	bool "Generic time"
-	default y
+	def_bool y
 
 config GENERIC_CLOCKEVENTS
 	bool "Generic clock events"
-	depends on GENERIC_TIME
 	default y
 
 choice
@@ -652,6 +650,10 @@ config GPTMR0_CLOCKSOURCE
 	depends on GENERIC_CLOCKEVENTS
 	depends on !TICKSOURCE_GPTMR0
 
+config ARCH_USES_GETTIMEOFFSET
+	depends on !GENERIC_CLOCKEVENTS
+	def_bool y
+
 source kernel/time/Kconfig
 
 comment "Misc"
diff --git a/arch/blackfin/kernel/time.c b/arch/blackfin/kernel/time.c
index adb54aa..e5069fe 100644
--- a/arch/blackfin/kernel/time.c
+++ b/arch/blackfin/kernel/time.c
@@ -81,11 +81,11 @@ time_sched_init(irqreturn_t(*timer_routine) (int, void *))
 #endif
 }
 
+#ifdef CONFIG_ARCH_USES_GETTIMEOFFSET
 /*
  * Should return useconds since last timer tick
  */
-#ifndef CONFIG_GENERIC_TIME
-static unsigned long gettimeoffset(void)
+u32 arch_gettimeoffset(void)
 {
 	unsigned long offset;
 	unsigned long clocks_per_jiffy;
@@ -184,65 +184,6 @@ void __init time_init(void)
 	time_sched_init(timer_interrupt);
 }
 
-#ifndef CONFIG_GENERIC_TIME
-void do_gettimeofday(struct timeval *tv)
-{
-	unsigned long flags;
-	unsigned long seq;
-	unsigned long usec, sec;
-
-	do {
-		seq = read_seqbegin_irqsave(&xtime_lock, flags);
-		usec = gettimeoffset();
-		sec = xtime.tv_sec;
-		usec += (xtime.tv_nsec / NSEC_PER_USEC);
-	}
-	while (read_seqretry_irqrestore(&xtime_lock, seq, flags));
-
-	while (usec >= USEC_PER_SEC) {
-		usec -= USEC_PER_SEC;
-		sec++;
-	}
-
-	tv->tv_sec = sec;
-	tv->tv_usec = usec;
-}
-EXPORT_SYMBOL(do_gettimeofday);
-
-int do_settimeofday(struct timespec *tv)
-{
-	time_t wtm_sec, sec = tv->tv_sec;
-	long wtm_nsec, nsec = tv->tv_nsec;
-
-	if ((unsigned long)tv->tv_nsec >= NSEC_PER_SEC)
-		return -EINVAL;
-
-	write_seqlock_irq(&xtime_lock);
-	/*
-	 * This is revolting. We need to set the xtime.tv_usec
-	 * correctly. However, the value in this location is
-	 * is value at the last tick.
-	 * Discover what correction gettimeofday
-	 * would have done, and then undo it!
-	 */
-	nsec -= (gettimeoffset() * NSEC_PER_USEC);
-
-	wtm_sec = wall_to_monotonic.tv_sec + (xtime.tv_sec - sec);
-	wtm_nsec = wall_to_monotonic.tv_nsec + (xtime.tv_nsec - nsec);
-
-	set_normalized_timespec(&xtime, sec, nsec);
-	set_normalized_timespec(&wall_to_monotonic, wtm_sec, wtm_nsec);
-
-	ntp_clear();
-
-	write_sequnlock_irq(&xtime_lock);
-	clock_was_set();
-
-	return 0;
-}
-EXPORT_SYMBOL(do_settimeofday);
-#endif /* !CONFIG_GENERIC_TIME */
-
 /*
  * Scheduler clock - returns current time in nanosec units.
  */

Re: [PATCH] Convert blackfin to use arch_gettimeoffset()

[Mike Frysinger]

On Fri, Jul 17, 2009 at 01:33, john stultz wrote:
> On Thu, 2009-07-16 at 22:26 -0400, Mike Frysinger wrote:
>> On Thu, Jul 16, 2009 at 20:25, john stultz wrote:
>> > This patch converts blackfin to use GENERIC_TIME via the arch_getoffset()
>> > infrastructure, reducing the amount of arch specific code we need to
>> > maintain.
>> >
>> > This patch applies on top of Linus' current -git tree (2.6.31-rc3).
>> >
>> > I've taken my best swing at converting this, but I'm not 100% confident
>> > I got it right. My cross-compiler is now out of date (gcc4.2) so I
>> > wasn't able to  check if it compiled. Any assistance from arch
>> > maintainers or testers to get this merged would be great.
>>
>> something seems to be off as the Blackfin defconfig now does:
>>   LD      .tmp_vmlinux1
>> kernel/built-in.o: In function `clocksource_forward_now':
>> kernel/time/timekeeping.c:82: undefined reference to `arch_gettimeoffset'
>> kernel/built-in.o: In function `getnstimeofday':
>> kernel/time/timekeeping.c:119: undefined reference to `arch_gettimeoffset'
>> make: *** [.tmp_vmlinux1] Error 1
>>
>> most likely because the Blackfin time.c is the old stuff
>> (non-clocksource) while time-ts.c is the newer stuff (clocksource) and
>> this change only touches time.c
>
> Ok. So I guess I missed that time.c isn't built if GENERIC_CLOCKEVENTS
> isn't enabled, so ARCH_HAS_GETTIMEOFFSET needs to depend on !
> GENERIC_CLOCKEVENTS.
>
> Hopefully I've got it cleared up now. Let me know if I'm still causing
> trouble.

this one seems to build & boot for GENERIC_CLOCKEVENTS set to y or n.  thanks.
-mike