Linus, all,
I've been busy with other things, so I've just been sitting on this for
a while. Anyway, I figured it was about time to resend.
This patch tries to cleanup the delay code by moving the timer-specific
implementations into the timer_ops struct. Thus, rather then doing:
if(x86_delay_tsc)
__rdtsc_delay(loops);
else if(x86_delay_cyclone)
__cyclone_delay(loops);
else if(whatever....
we just simply do:
if(timer)
timer->delay(loops);
Making it much easier to accommodate alternate time sources.
Please apply.
thanks
-john
diff -Nru a/arch/i386/kernel/timers/timer_cyclone.c b/arch/i386/kernel/timers/timer_cyclone.c
--- a/arch/i386/kernel/timers/timer_cyclone.c Tue Jan 7 17:11:03 2003
+++ b/arch/i386/kernel/timers/timer_cyclone.c Tue Jan 7 17:11:03 2003
@@ -150,7 +150,6 @@
}
-#if 0 /* XXX future work */
static void delay_cyclone(unsigned long loops)
{
unsigned long bclock, now;
@@ -162,12 +161,12 @@
now = cyclone_timer[0];
} while ((now-bclock) < loops);
}
-#endif
/************************************************************/
/* cyclone timer_opts struct */
struct timer_opts timer_cyclone = {
.init = init_cyclone,
.mark_offset = mark_offset_cyclone,
- .get_offset = get_offset_cyclone
+ .get_offset = get_offset_cyclone,
+ .delay = delay_cyclone,
};
diff -Nru a/arch/i386/kernel/timers/timer_pit.c b/arch/i386/kernel/timers/timer_pit.c
--- a/arch/i386/kernel/timers/timer_pit.c Tue Jan 7 17:11:03 2003
+++ b/arch/i386/kernel/timers/timer_pit.c Tue Jan 7 17:11:03 2003
@@ -27,6 +27,19 @@
/* nothing needed */
}
+static void delay_pit(unsigned long loops)
+{
+ int d0;
+ __asm__ __volatile__(
+ "\tjmp 1f\n"
+ ".align 16\n"
+ "1:\tjmp 2f\n"
+ ".align 16\n"
+ "2:\tdecl %0\n\tjns 2b"
+ :"=&a" (d0)
+ :"0" (loops));
+}
+
/* This function must be called with interrupts disabled
* It was inspired by Steve McCanne's microtime-i386 for BSD. -- jrs
@@ -129,4 +142,5 @@
.init = init_pit,
.mark_offset = mark_offset_pit,
.get_offset = get_offset_pit,
+ .delay = delay_pit,
};
diff -Nru a/arch/i386/kernel/timers/timer_tsc.c b/arch/i386/kernel/timers/timer_tsc.c
--- a/arch/i386/kernel/timers/timer_tsc.c Tue Jan 7 17:11:03 2003
+++ b/arch/i386/kernel/timers/timer_tsc.c Tue Jan 7 17:11:03 2003
@@ -16,7 +16,6 @@
int tsc_disable __initdata = 0;
-extern int x86_udelay_tsc;
extern spinlock_t i8253_lock;
static int use_tsc;
@@ -107,6 +106,17 @@
delay_at_last_interrupt = (count + LATCH/2) / LATCH;
}
+static void delay_tsc(unsigned long loops)
+{
+ unsigned long bclock, now;
+
+ rdtscl(bclock);
+ do
+ {
+ rep_nop();
+ rdtscl(now);
+ } while ((now-bclock) < loops);
+}
/* ------ Calibrate the TSC -------
* Return 2^32 * (1 / (TSC clocks per usec)) for do_fast_gettimeoffset().
@@ -272,8 +282,6 @@
* We could be more selective here I suspect
* and just enable this for the next intel chips ?
*/
- x86_udelay_tsc = 1;
-
/* report CPU clock rate in Hz.
* The formula is (10^6 * 2^32) / (2^32 * 1 / (clocks/us)) =
* clock/second. Our precision is about 100 ppm.
@@ -310,4 +318,5 @@
.init = init_tsc,
.mark_offset = mark_offset_tsc,
.get_offset = get_offset_tsc,
+ .delay = delay_tsc,
};
diff -Nru a/arch/i386/lib/delay.c b/arch/i386/lib/delay.c
--- a/arch/i386/lib/delay.c Tue Jan 7 17:11:03 2003
+++ b/arch/i386/lib/delay.c Tue Jan 7 17:11:03 2003
@@ -15,35 +15,18 @@
#include <linux/delay.h>
#include <asm/processor.h>
#include <asm/delay.h>
+#include <asm/timer.h>
#ifdef CONFIG_SMP
#include <asm/smp.h>
#endif
-int x86_udelay_tsc = 0; /* Delay via TSC */
+extern struct timer_opts* timer;
-
/*
- * Do a udelay using the TSC for any CPU that happens
- * to have one that we trust.
+ * Backup non-TSC based delay loop.
+ * Used until a timer is chosen.
*/
-
-static void __rdtsc_delay(unsigned long loops)
-{
- unsigned long bclock, now;
-
- rdtscl(bclock);
- do
- {
- rep_nop();
- rdtscl(now);
- } while ((now-bclock) < loops);
-}
-
-/*
- * Non TSC based delay loop for 386, 486, MediaGX
- */
-
static void __loop_delay(unsigned long loops)
{
int d0;
@@ -59,8 +42,8 @@
void __delay(unsigned long loops)
{
- if (x86_udelay_tsc)
- __rdtsc_delay(loops);
+ if(timer)
+ timer->delay(loops);
else
__loop_delay(loops);
}
diff -Nru a/include/asm-i386/timer.h b/include/asm-i386/timer.h
--- a/include/asm-i386/timer.h Tue Jan 7 17:11:03 2003
+++ b/include/asm-i386/timer.h Tue Jan 7 17:11:03 2003
@@ -14,6 +14,7 @@
int (*init)(void);
void (*mark_offset)(void);
unsigned long (*get_offset)(void);
+ void (*delay)(unsigned long);
};
#define TICK_SIZE (tick_nsec / 1000)
On 7 Jan 2003, john stultz wrote:
>
> if (timer)
> timer->delay(loops);
Why the "if (timer)"?
Wouldn't it be saner to initialize the timer to something that can at
least do estimated loops, and then just unconditionally do
timer->delay(..);
instead?
Linus
On Tue, 2003-01-07 at 19:29, Linus Torvalds wrote:
> Wouldn't it be saner to initialize the timer to something that can at
> least do estimated loops, and then just unconditionally do
>
> timer->delay(..);
>
> instead?
Basically a default timer_null structure? Sure that can be done (I think
others have suggested this as well, but I've just not had the time to
get around to it). I'll see if I can scratch that out quickly and
resend.
thanks for the feedback!
-john
Hi!
> Linus, all,
> I've been busy with other things, so I've just been sitting on this for
> a while. Anyway, I figured it was about time to resend.
>
> This patch tries to cleanup the delay code by moving the timer-specific
> implementations into the timer_ops struct. Thus, rather then doing:
>
> if(x86_delay_tsc)
> __rdtsc_delay(loops);
> else if(x86_delay_cyclone)
> __cyclone_delay(loops);
> else if(whatever....
>
> we just simply do:
>
> if(timer)
> timer->delay(loops);
>
> diff -Nru a/arch/i386/kernel/timers/timer_pit.c b/arch/i386/kernel/timers/timer_pit.c
> --- a/arch/i386/kernel/timers/timer_pit.c Tue Jan 7 17:11:03 2003
> +++ b/arch/i386/kernel/timers/timer_pit.c Tue Jan 7 17:11:03 2003
> @@ -27,6 +27,19 @@
> /* nothing needed */
> }
>
> +static void delay_pit(unsigned long loops)
> +{
> + int d0;
> + __asm__ __volatile__(
> + "\tjmp 1f\n"
> + ".align 16\n"
> + "1:\tjmp 2f\n"
> + ".align 16\n"
> + "2:\tdecl %0\n\tjns 2b"
> + :"=&a" (d0)
> + :"0" (loops));
> +}
> +
But... this is not using pit to do the delay, right? It is sensitive
to CPU clock changes, pit-delay should not be.
Pavel
--
Worst form of spam? Adding advertisment signatures ala sourceforge.net.
What goes next? Inserting advertisment *into* email?