Re: [PATCH v2] staging: comedi: s626: use comedi_timeout() on remaining loops

From: Chase Southwood
Date: Thu Apr 03 2014 - 18:05:26 EST


>On Thursday, April 3, 2014 3:38 AM, Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:

>>On Tue, Mar 25, 2014 at 10:43:58PM -0500, Chase Southwood wrote:
>>There were just a handful of more while loops in this file that needed
>>timeouts, and this patch takes care of them.  One new callback is
>>introduced, and all of the proper comedi_timeout() calls are then used.
>>
>>Signed-off-by: Chase Southwood <chase.southwood@xxxxxxxxx>
>>---
>>2: s626_i2c_handshake_eoc() can be used in s626_initialize() as noted by
>>Ian.  So, s626_initialize_eoc() has been removed, and its uses swapped
>>for s626_i2c_handshake_eoc().
>>
>> drivers/staging/comedi/drivers/s626.c | 34 ++++++++++++++++++++++++++--------
>> 1 file changed, 26 insertions(+), 8 deletions(-)
>>
>>diff --git a/drivers/staging/comedi/drivers/s626.c b/drivers/staging/comedi/drivers/s626.c
>>index 95fadf3..865cf93 100644
>>--- a/drivers/staging/comedi/drivers/s626.c
>>+++ b/drivers/staging/comedi/drivers/s626.c
>>@@ -295,10 +295,24 @@ static void s626_debi_replace(struct comedi_device *dev, unsigned int addr,
>
>> /* **************  EEPROM ACCESS FUNCTIONS  ************** */
>
>>+static int s626_i2c_handshake_eoc(struct comedi_device *dev,
>>+                struct comedi_subdevice *s,
>>+                struct comedi_insn *insn,
>>+                unsigned long context)
>>+{
>>+    unsigned int status;
>
>This should probably be bool.
>

Oh, whoops...yeah s626_mc_test() definitely returns bool...I'll fix this up.

>>+
>>+    status = s626_mc_test(dev, S626_MC2_UPLD_IIC, S626_P_MC2);
>>+    if (status)
>>+        return 0;
>>+    return -EBUSY;
>>+}
>>+
>> static uint32_t s626_i2c_handshake(struct comedi_device *dev, uint32_t val)
>> {
>>     struct s626_private *devpriv = dev->private;
>>     unsigned int ctrl;
>>+    uint32_t ret;
>
>This should be int.  I get really suspicious when people start using
>uint32_t types.  Why does it have to be 32 bits?  Unsigned is wrong as
>well.

Yeah...I originally did that to conform to the current return type of the function,
not sure how I didn't manage to see that trying to return a negative error code as
an unsigned int is clearly a bug.  Sorry, I'll fix this up as well.

Thanks for the review, Dan.

Chase
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/