[cron,job] v4l-dvb daily build 2.6.22 and up: ERRORS, 2.6.16-2.6.21: WARNINGS

Commit Message

Miroslav Å ustek April 8, 2009, 10:56 p.m. UTC

Hans Verkuil <hverkuil <at> xs4all.nl> writes:
> 
> Can someone take a look at these warnings and errors? Looking at the log
> these seem to be pretty easy to fix (compat stuff for the most part).
> 
> I don't have the time for this for several more days, so I'd appreciate it
> if someone could take a look at this for me.
> 
> Thanks,
> 
>         Hans
> 

Yes, it looks like my version of kernel luckily built, but other (older)
versions have problems.

Here: http://article.gmane.org/gmane.linux.drivers.video-input-infrastructure/4400
Marton Balint sent patch with <linux/div64.h> include and changed 'div_s64_rem'
to 'div_s64'. But kernels older than 2.26.x have neither 'div_s64' nor
'div_s64_rem'.

Better apply this patch instead that from Marton.
It uses 'do_div' which is much longer supported.

With 'do_div' results should be same but for completeness here are some
appointments:

 * 'do_div' does computation with unsigned numbers.
 * So we need to pass dividend as unsigned (its ABS) and then alter the sign of
the result (if dividend was negative).
 * But function 'int_goertzel' returns u32 so if we pass dividend as u64 (cast
from s64), then no sign change is needed before returning. */
 * Also, divisor is N*N, so it will be always non-negative.
 * I think that some implementations(platforms) of 'do_div' requires divisor to
be 32bit. That's why I rather use u32 for divisor (and not u64). (Here
http://article.gmane.org/gmane.linux.drivers.video-input-infrastructure/4337
Marton wrote that N is at most 576.)


----- FILE: cx88-dsp_64bit_math3.patch -----

cx88-dsp: again fixing 64bit math on 32bit kernels

From: Miroslav Sustek <[e-mail]>
Signed-off-by: Miroslav Sustek <[e-mail]>



--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to [e-mail]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff -r 77ebdc14cc24 linux/drivers/media/video/cx88/cx88-dsp.c
--- a/linux/drivers/media/video/cx88/cx88-dsp.c	Wed Apr 08 14:01:19 2009 -0300
+++ b/linux/drivers/media/video/cx88/cx88-dsp.c	Thu Apr 09 00:52:27 2009 +0200
@@ -22,6 +22,7 @@ 
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/jiffies.h>
+#include <linux/math64.h>
 
 #include "cx88.h"
 #include "cx88-reg.h"
@@ -101,8 +102,8 @@ 
 	s32 coeff = 2*int_cos(freq);
 	u32 i;
 
-	s64 tmp;
-	u32 remainder;
+	u64 tmp;
+	u32 divisor;
 
 	for (i = 0; i < N; i++) {
 		s32 s = x[i] + ((s64)coeff*s_prev/32768) - s_prev2;
@@ -115,7 +116,10 @@ 
 
 	/* XXX: N must be low enough so that N*N fits in s32.
 	 * Else we need two divisions. */
-	return (u32) div_s64_rem(tmp, N*N, &remainder);
+	divisor = N*N;
+	do_div(tmp, divisor);
+
+	return (u32) tmp;
 }
 
 static u32 freq_magnitude(s16 x[], u32 N, u32 freq)