[mpeg2] Re: Porting kfir for kernel 2.6

[Anssi Hannula <anssi.hannula@gmail.com>]

Klaus Schmidinger wrote:
> Anssi Hannula wrote:
>
> > Klaus Schmidinger wrote:
> >
> > > Anssi Hannula wrote:
> > >
> > > > Klaus Schmidinger wrote:
> > > >
> > > > > Anssi Hannula wrote:
> > > > >
> > > > > > ...
> > > > > > The easiest-to-solve possible problem is the parameters module 
> > > > > > sends to Kfir. I had to alter them sligtly because they made some 
> > > > > > floating-point arithmetics, which can't be used in kernel code.
> > > > > >
> > > > > > To see any differences in module behaviour, I attach two patches, 
> > > > > > one for 2.4 (against cvs) and an other one for 2.6 (rev3). They 
> > > > > > produce verbose output whenever parameters are downloaded to Kfir. 
> > > > > > Please test them both with the same module parameters (debug=1 
> > > > > > included), do cat /dev/video0 > x.x and get the debug output from 
> > > > > > syslog. Parameter may be reported multiple times in the log, post 
> > > > > > everything.
> > > > >
> > > > >
> > > > > Logs attached.
> > > >
> > > >
> > > > Sorry, of course just the critical line was linewrapped, and didn't 
> > > > match my regexp search-replace.
> > > > Do the same with these patches.
> > >
> > >
> > > Logs attached.
> >
> >
> > That line indeed appears to be a problem.
> > I attached a patch for use with 2.4, which prints some values to help 
> > making the line 2.6-compatible.
> >
> > Then let's hope this was the only reason why there is no data.
> When I try to insmod kfir.o I get
>
>   kfir.o: unresolved symbol __fixunsdfdi
>
> and before that the compiler tells me
>
>   kfir.c: In function `Kfir_DownLoadParams':
>   kfir.c:1089: warning: unsigned int format, different type arg (arg 3)
>   kfir.c:1091: warning: unsigned int format, different type arg (arg 3)
>   kfir.c:1093: warning: unsigned int format, different type arg (arg 3)
>   kfir.c:1094: warning: unsigned int format, double arg (arg 3)
>   kfir.c:1095: warning: unsigned int format, double arg (arg 3)
>   kfir.c:1096: warning: unsigned int format, double arg (arg 3)
>   kfir.c:1097: warning: unsigned int format, double arg (arg 3)
>   kfir.c:1098: warning: unsigned int format, double arg (arg 3)
>   kfir.c:1099: warning: unsigned int format, double arg (arg 3)
>
> So I've commented out the first 6 INFOPRINTK and defined
>
>   #define INFOPRINTD(x) printk("kfir: %s: %f\n", #x, x)
>
> and used that for the rest, but that just gave me
>
>   Jan 24 18:05:26 video kernel: kfir: Download parameters...kfir: 
> KFIR_ADR_FIRST_PICT_DEL parameter debug info:
>   Jan 24 18:05:26 video kernel: kfir: 
> ((((((double)Params->VBFBufferSize)*16384.0*2.0)/8.0)*90000.0)/((double)MulBitRate)): 
> %%f
>   Jan 24 18:05:26 video kernel: kfir: 
> ((double)Params->VBFBufferSize)*16384.0*2.0: %%f
>   Jan 24 18:05:26 video kernel: kfir: 
> (((double)Params->VBFBufferSize)*16384.0*2.0)/8.0: %%f
>   Jan 24 18:05:26 video kernel: kfir: (double)Params->VBFBufferSize: %%f
>   Jan 24 18:05:26 video kernel: kfir: (double)MulBitRate: %%f
>   Jan 24 18:05:26 video kernel: kfir: 
> (double)((((((double)Params->VBFBufferSize)*16384.0*2.0)/8.0)*90000.0)/((double)MulBitRate)): 
> %%f
>   Jan 24 18:05:26 video kernel: kfir: done
>
> I wonder why the '%' appears twice?!
Because printk doesn't support %f.

> Maybe you can fix these lines so that they actually work.
I downloaded the 2.4.29 so I could test compiling.

I changed the QWORDs on the INFOPRINTKs to WORDs. Now it doesn't have at 
least unresolved symbol __fixunsdfdi. The warnings don't matter, it 
probably will still print some data.
If there are now other unresolved symbols, remove the lines beginning 
"INFOPRINTK((WORD". If it doesn't help, remove every INFOPRINTK except 
"INFOPRINTK(MulBitRate);" and "INFOPRINTK(Params->VBFBufferSize);".

--
Anssi Hannula
--- kfir/driver/kfir.c	2005-01-05 02:29:17.000000000 +0200
+++ kfir-2.4-debug/driver/kfir.c	2005-01-24 20:06:18.120437400 +0200
@@ -1075,8 +1075,21 @@
 	kfir->Par[KFIR_ADR_C2/4] = (WORD)((1<<16)/1.4);
 
 	kfir->Par[KFIR_ADR_STD_PICT_DEL/4] = (WORD)(90000000UL/MulFrameRate);
-	kfir->Par[KFIR_ADR_FIRST_PICT_DEL/4] = 
-		(WORD) ((((((double)Params->VBFBufferSize)*16384.0*2.0)/8.0)*90000.0)/((double)MulBitRate));
+#define INFOPRINTK(x) printk("kfir: %s: %u\n", #x, x)
+	printk("kfir: KFIR_ADR_FIRST_PICT_DEL parameter debug info:\n");
+	INFOPRINTK(Params->VBFBufferSize);
+	INFOPRINTK((WORD)((double)Params->VBFBufferSize));
+	INFOPRINTK(MulBitRate);
+	INFOPRINTK((WORD)((double)MulBitRate));
+	INFOPRINTK((WORD) ((((((double)Params->VBFBufferSize)*16384.0*2.0)/8.0)*90000.0)/((double)MulBitRate)));
+	INFOPRINTK((WORD)((double)((((((double)Params->VBFBufferSize)*16384.0*2.0)/8.0)*90000.0)/((double)MulBitRate))));
+	INFOPRINTK(((((((double)Params->VBFBufferSize)*16384.0*2.0)/8.0)*90000.0)/((double)MulBitRate)));
+	INFOPRINTK(((double)Params->VBFBufferSize)*16384.0*2.0);
+	INFOPRINTK((((double)Params->VBFBufferSize)*16384.0*2.0)/8.0);
+	INFOPRINTK((double)Params->VBFBufferSize);
+	INFOPRINTK((double)MulBitRate);
+	INFOPRINTK((double)((((((double)Params->VBFBufferSize)*16384.0*2.0)/8.0)*90000.0)/((double)MulBitRate)));
+	kfir->Par[KFIR_ADR_FIRST_PICT_DEL/4] = (WORD) ((((((double)Params->VBFBufferSize)*16384.0*2.0)/8.0)*90000.0)/((double)MulBitRate));
 
 
 	Data=RMUL(MulBitRate*2,