[PATCH] driver/input/ff-memless.c

[PATCH] driver/input/ff-memless.c

[Gary Stein]

I was experiencing force feedback joystick issues in my Force 3D Pro and
when developing a new driver for the new Logitech G940 after I
recompiled my kernel.

I was able to track it down to a signed vs unsigned problem in
ff-memless.c and gcc 4.4.2.
The code in question did not seem to give a problem with gcc 4.2.4.

According to gitweb, a signed vs unsigned was changed on May 8th by
Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Input: ff-memless - fix signed to unsigned bit overflow

to change the gain from a signed to an unsigned int to fix a rumble
issue, but it appears to broke a the
constant force section of the code.

It is a simple, cast to fix and should not have any ill effects.

gary


--- linux-2.6.31.orig/drivers/input/ff-memless.c	2009-09-09
17:13:59.000000000 -0500
+++ linux-2.6.31.int/drivers/input/ff-memless.c	2009-12-07
15:09:22.000000000 -0500
@@ -239,8 +239,8 @@
 		level = fixp_new16(apply_envelope(state,
 					new->u.constant.level,
 					&new->u.constant.envelope));
-		x = fixp_mult(fixp_sin(i), level) * gain / 0xffff;
-		y = fixp_mult(-fixp_cos(i), level) * gain / 0xffff;
+		x = (int)(fixp_mult(fixp_sin(i), level) * gain) / 0xffff;
+		y = (int)(fixp_mult(-fixp_cos(i), level) * gain) / 0xffff;
 		/*
 		 * here we abuse ff_ramp to hold x and y of constant force
 		 * If in future any driver wants something else than x and y

Re: [PATCH] driver/input/ff-memless.c

[Dmitry Torokhov]

Hi Gary,

On Mon, Dec 07, 2009 at 02:37:30PM -0500, Gary Stein wrote:
> I was experiencing force feedback joystick issues in my Force 3D Pro and
> when developing a new driver for the new Logitech G940 after I
> recompiled my kernel.
> 
> I was able to track it down to a signed vs unsigned problem in
> ff-memless.c and gcc 4.4.2.
> The code in question did not seem to give a problem with gcc 4.2.4.
> 
> According to gitweb, a signed vs unsigned was changed on May 8th by
> Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
> Input: ff-memless - fix signed to unsigned bit overflow
> 
> to change the gain from a signed to an unsigned int to fix a rumble
> issue, but it appears to broke a the
> constant force section of the code.
> 
> It is a simple, cast to fix and should not have any ill effects.
> 

Could I please have your "Signed-off-by: ..."? Thanks.

> gary
> 
> 
> --- linux-2.6.31.orig/drivers/input/ff-memless.c	2009-09-09
> 17:13:59.000000000 -0500
> +++ linux-2.6.31.int/drivers/input/ff-memless.c	2009-12-07
> 15:09:22.000000000 -0500
> @@ -239,8 +239,8 @@
>  		level = fixp_new16(apply_envelope(state,
>  					new->u.constant.level,
>  					&new->u.constant.envelope));
> -		x = fixp_mult(fixp_sin(i), level) * gain / 0xffff;
> -		y = fixp_mult(-fixp_cos(i), level) * gain / 0xffff;
> +		x = (int)(fixp_mult(fixp_sin(i), level) * gain) / 0xffff;
> +		y = (int)(fixp_mult(-fixp_cos(i), level) * gain) / 0xffff;
>  		/*
>  		 * here we abuse ff_ramp to hold x and y of constant force
>  		 * If in future any driver wants something else than x and y

-- 
Dmitry