Re: [MIPS] Fix and cleanup the MIPS part of the (ab)use of CLOCK_TICK_RATE.

Re: [MIPS] Fix and cleanup the MIPS part of the (ab)use of CLOCK_TICK_RATE.

[Atsushi Nemoto]

On Thu, 01 Nov 2007 15:46:41 +0000, linux-mips@linux-mips.org wrote:
> Author: Ralf Baechle <ralf@linux-mips.org> Thu Nov 1 15:45:37 2007 +0000
> Commit: 0a354a30fe552b78a4db0873c19d8936551cc158
> Gitweb: http://www.linux-mips.org/g/linux/0a354a30
> Branch: master
> 
> This is the clock rate of the i8253 PIT.  A MIPS system may not have
> a PIT by the symbol is used all over the kernel including some APIs.
> So keeping it defined to the number for the PIT is the only sane thing
> for now.

The CLOCK_TICK_RATE is used for ACTHZ, TICK_NSEC, etc.

At least for i8253-free platforms, It looks a value multiple of HZ
would be better for such constants, assuming we have dyntick or
accurate HZ clockevents.

How about something like this?

diff --git a/include/asm-mips/timex.h b/include/asm-mips/timex.h
index 5816ad1..e9622b6 100644
--- a/include/asm-mips/timex.h
+++ b/include/asm-mips/timex.h
@@ -18,7 +18,11 @@
  * So keeping it defined to the number for the PIT is the only sane thing
  * for now.
  */
+#ifdef CONFIG_I8253
 #define CLOCK_TICK_RATE 1193182
+#else
+#define CLOCK_TICK_RATE 1024000	/* multiple of HZ */
+#endif
 
 /*
  * Standard way to access the cycle counter.

Re: [MIPS] Fix and cleanup the MIPS part of the (ab)use of CLOCK_TICK_RATE.

[Ralf Baechle]

On Wed, Nov 07, 2007 at 11:10:36PM +0900, Atsushi Nemoto wrote:

> The CLOCK_TICK_RATE is used for ACTHZ, TICK_NSEC, etc.
> 
> At least for i8253-free platforms, It looks a value multiple of HZ
> would be better for such constants, assuming we have dyntick or
> accurate HZ clockevents.
> 
> How about something like this?
> 
> diff --git a/include/asm-mips/timex.h b/include/asm-mips/timex.h
> index 5816ad1..e9622b6 100644
> --- a/include/asm-mips/timex.h
> +++ b/include/asm-mips/timex.h
> @@ -18,7 +18,11 @@
>   * So keeping it defined to the number for the PIT is the only sane thing
>   * for now.
>   */
> +#ifdef CONFIG_I8253
>  #define CLOCK_TICK_RATE 1193182
> +#else
> +#define CLOCK_TICK_RATE 1024000	/* multiple of HZ */
> +#endif

kernel/time/ntp.c:#define CLOCK_TICK_OVERFLOW   (LATCH * HZ - CLOCK_TICK_RATE)
kernel/time/ntp.c:                                      (s64)CLOCK_TICK_RATE)

drivers/char/vt_ioctl.c:                        arg = CLOCK_TICK_RATE / arg;
drivers/char/vt_ioctl.c:                        count = CLOCK_TICK_RATE / count;

There is so much abuse of this variable, it's not even funny.  It really
deserve to be taken out and shot.  And that's just two cases.

  Ralf