[Intel-gfx] [PATCH] drm/i915: Only expand COND once in wait_for()

Zanoni, Paulo R paulo.r.zanoni at intel.com
Tue Sep 13 21:05:12 UTC 2016


Em Ter, 2016-09-13 às 20:40 +0100, Chris Wilson escreveu:
> I was looking at some wait_for() timeouts on a slow system, with lots
> of
> debug enabled (KASAN, lockdep, mmio_debug). Thinking that we were
> mishandling the timeout, I tried to ensure that we loop at least once
> after first testing COND. However, the double test of COND either
> side
> of the timeout check makes that unlikely. But we can do an equivalent
> loop, that keeps the COND check after testing for timeout (required
> so
> that we are not preempted between testing COND and then testing for a
> timeout) without expanding COND twice.
> 
> The advantage of only expanding COND once is a dramatic reduction in
> code size:
> 
>    text	   data	    bss	    dec	    hex
> 1308733	   5184	   1152	1315069	 1410fd	
> before
> 1305341	   5184	   1152	1311677	 1403bd	
> after
> 
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> Cc: Ville Syrjälä <ville.syrjala at linux.intel.com>
> Cc: Daniel Vetter <daniel.vetter at ffwll.ch>
> ---
>  drivers/gpu/drm/i915/intel_drv.h | 13 ++++++++-----
>  1 file changed, 8 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/intel_drv.h
> b/drivers/gpu/drm/i915/intel_drv.h
> index cb99a2540863..597899d71df9 100644
> --- a/drivers/gpu/drm/i915/intel_drv.h
> +++ b/drivers/gpu/drm/i915/intel_drv.h
> @@ -52,13 +52,16 @@
>   */
>  #define _wait_for(COND, US, W) ({ \
>  	unsigned long timeout__ = jiffies + usecs_to_jiffies(US) +
> 1;	\
> -	int ret__ = 0;						
> 	\
> -	while (!(COND)) {						
> \
> -		if (time_after(jiffies, timeout__)) {		
> 	\
> -			if (!(COND))					
> \
> -				ret__ = -ETIMEDOUT;			
> \
> +	int ret__;

ret__ starts "uninitialized".

> 							\
> +	for (;;) {							
> \
> +		if (time_after(jiffies, timeout__))			
> \
> +			ret__ = -ETIMEDOUT;			

If we didn't hit the timeout, it's still "uninitialized".

> 	\
> +		if (COND) {						
> \
> +			ret__ = 0;				

If the condition was not met, it's still "uninitialized".

> 	\
>  			break;					
> 	\
>  		}							
> \
> +		if (ret__)				

But we read its "uninitialized" value here.

But why isn't the compiler complaining about this? Am I failing to see
something here?

If my analysis is correct, all you need to do is to keep ret__ being
initialized to zero. At least for clarity of the future code readers in
case it's expected to be auto-initialized to zero due to some weird
rule about compound statements or something.

With the ret__ initialization to zero:
Reviewed-by: Paulo Zanoni <paulo.r.zanoni at intel.com>

> 		\
> +			break;					
> 	\
>  		if ((W) && drm_can_sleep()) {			
> 	\
>  			usleep_range((W), (W)*2);			
> \
>  		} else {						
> \


More information about the Intel-gfx mailing list