Hello Leo,

On Mon, Jan 18, 2016 at 04:22:51PM +0800, Li Yang wrote:

On Fri, Jan 15, 2016 at 5:03 PM, Uwe Kleine-K?nig
[off-list ref] wrote:

quoted

On Fri, Jan 15, 2016 at 04:46:06PM +0800, Li Yang wrote:

quoted

On Fri, Jan 15, 2016 at 3:53 PM, Uwe Kleine-K?nig
[off-list ref] wrote:

quoted

quoted

-     i2c_imx_init_recovery_info(i2c_imx, pdev);
+     if (IS_ERR_OR_NULL(i2c_imx->pinctrl))

I'd prefer to set i2c_imx->pinctrl to NULL if it's not used and then use

        if (!i2c_imx->pinctrl)

here. (Or maybe make i2c_imx_init_recovery_info aware of this situation
to keep the caller simple?)

Setting not used pointer to NULL might be a good coding practice
generally.  But in this case neither the driver nor gpio framework
checks if the pinctrl is NULL before accessing it.  We still need the
driver to make sure pinctrl is not used in this situation.  The
benefit of the proposed change seems to be pretty limited.  :)

At least it gets rid of IS_ERR_OR_NULL which is used wrongly more often
than not.

It's true that this API is not used a lot.  Are you suggesting that it
is poorly designed and probably should be removed?  But I still don't
quite get the point why it tends to be wrongly used.  Do you consider
the use in this patch a wrongly used case?

See http://permalink.gmane.org/gmane.linux.ports.arm.omap/97874 for an
earlier discussion. Having said that, I think at the moment there is no
such bug in this case, still I agree with Russell that IS_ERR_OR_NULL
isn't nice and so would prefer not to add new instances.

Best regards
Uwe


-- 
Pengutronix e.K.                           | Uwe Kleine-K?nig            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |