Re: [PATCH] i2c: mv64xxx: remove unreachable signal case handling

[Nicholas Mc Guire <der.herr-kA1LtwSENNE@public.gmane.org>]

On Wed, 17 Jun 2015, Gregory CLEMENT wrote:

> Hi Wolfram, Nicholas,
> 
> On 17/06/2015 15:00, Wolfram Sang wrote:
> > On Thu, Jun 11, 2015 at 05:27:33PM +0200, Nicholas Mc Guire wrote:
> >> 'commit d295a86eab20 ("i2c: mv64xxx: work around signals causing I2C
> >> transactions to be aborted")' removed the wait_event_interruptible_timeout 
> >> to prevent half/mixed i2c messages from being sent/received but forgot to
> >> drop the signal received cases in the return handling. This just removes
> >> this dead code and simplifies the error message as "time_left" only can be 
> >> 0 here and thus it conveys no additional information.
> >>
> >> Signed-off-by: Nicholas Mc Guire <hofrat-Q945KHDl0DbYtjvyW6yDsg@public.gmane.org>
> >> ---
> >>
> >> Patch was compile tested with multi_v7_defconfig 
> >> (implies CONFIG_I2C_MV64XXX=y)
> >>
> >> Patch is against 4.1-rc7 (localversion-next is -next-20150611)
> > 
> > Hmm, IMO this patch is too intrusive to be applied without actual
> > testing.
> > 
> >>
> >>  drivers/i2c/busses/i2c-mv64xxx.c |   15 +++------------
> >>  1 file changed, 3 insertions(+), 12 deletions(-)
> >>
> >> diff --git a/drivers/i2c/busses/i2c-mv64xxx.c b/drivers/i2c/busses/i2c-mv64xxx.c
> >> index 30059c1..a4f8ece 100644
> >> --- a/drivers/i2c/busses/i2c-mv64xxx.c
> >> +++ b/drivers/i2c/busses/i2c-mv64xxx.c
> >> @@ -534,7 +534,6 @@ mv64xxx_i2c_wait_for_completion(struct mv64xxx_i2c_data *drv_data)
> >>  {
> >>  	long		time_left;
> >>  	unsigned long	flags;
> >> -	char		abort = 0;
> >>  
> >>  	time_left = wait_event_timeout(drv_data->waitq,
> >>  		!drv_data->block, drv_data->adapter.timeout);
> >> @@ -542,25 +541,17 @@ mv64xxx_i2c_wait_for_completion(struct mv64xxx_i2c_data *drv_data)
> >>  	spin_lock_irqsave(&drv_data->lock, flags);
> >>  	if (!time_left) { /* Timed out */
> >>  		drv_data->rc = -ETIMEDOUT;
> >> -		abort = 1;
> >> -	} else if (time_left < 0) { /* Interrupted/Error */
> >> -		drv_data->rc = time_left; /* errno value */
> >> -		abort = 1;
> >> -	}
> >> -
> >> -	if (abort && drv_data->block) {
> >>  		drv_data->aborting = 1;
> >>  		spin_unlock_irqrestore(&drv_data->lock, flags);
> >>  
> >>  		time_left = wait_event_timeout(drv_data->waitq,
> >>  			!drv_data->block, drv_data->adapter.timeout);
> >>  
> >> -		if ((time_left <= 0) && drv_data->block) {
> > 
> > I am especially unsure about the drv_data->block removal. Did you double
> > check if we can do this?
> > 

The consideration was 
  * wait_event_timeout was checkign !drv_data->block - so it it returned
    the condition held OR timeout
  * since it was a timeout here the condition was NOT met so either both
    are true or both are wrong 
I think the current logic only makes sense if one assumes that a signal case
is also possible.

> >> +		if (time_left == 0) {
> >>  			drv_data->state = MV64XXX_I2C_STATE_IDLE;
> >>  			dev_err(&drv_data->adapter.dev,
> >> -				"mv64xxx: I2C bus locked, block: %d, "
> >> -				"time_left: %d\n", drv_data->block,
> >> -				(int)time_left);
> >> +				"mv64xxx: I2C bus locked, block: %d\n",
> >> +				drv_data->block);
> > 
> > And if so, shouldn't that also be always 1 in the output here?
> > 

yes drv_data->block is 0 | 1 only - so that probably could be dropped 
as well

> >>  			mv64xxx_i2c_hw_init(drv_data);
> >>  		}
> >>  	} else
> > 
> > Maybe (not sure) it also helps to split the patch into everything
> > dealing with time_left as patch 1) and simplifying by drv_data->block
> > removal as patch2?
> 
> I agree. I would like to see 2 patches. The first one should be not controversial
> and could be applied whereas the second one will need a deeper review.
>
thanks - will refactor and split it into two parts
and see if I can get this tested somehow - no urgency
as its really only cleanup.

thx!
hofrat