On Thu, 05 Jan 2017, Bjorn Andersson wrote:

On Wed 04 Jan 23:49 PST 2017, Lee Jones wrote:

quoted

On Wed, 04 Jan 2017, Bjorn Andersson wrote:

quoted

On Wed 04 Jan 03:54 PST 2017, Lee Jones wrote:

quoted

On Mon, 26 Dec 2016, Bjorn Andersson wrote:

quoted

From: Bjorn Andersson <redacted>

Implement support for initialization of the lm3533 driver core and
probing child devices from Device Tree.

[..]

quoted

quoted

@@ -512,6 +514,11 @@ static int lm3533_device_init(struct lm3533 *lm3533)
 	lm3533_device_bl_init(lm3533);
 	lm3533_device_led_init(lm3533);
 
+	if (lm3533->dev->of_node) {
+		of_platform_populate(lm3533->dev->of_node, NULL, NULL,
+				     lm3533->dev);
+	}

I think it's save to call of_platform_populate(), even if !of_node.
It will just fail and return an error code, which you are ignoring
anyway.

I thought so too, but that's apparently how you trigger probing children
of the root node. So we're stuck with a conditional.

Ah, so this is to protect against the case where DT is present, but a
node for this device is not (or is disabled), so is left unprobed.
Then the bind is initiated via I2C?  Or something else?

In the event that a new lm3533 is spawned from sysfs we would not have
platform_data when entering lm3533_device_init() and just bail early.

Therefor, this issue would be limited to the odd case of lm3533 being
initiated from code (e.g. a board file) on a DT enabled system. In which
case it will create and probe new devices from the root of the DT.

Eewww, do we really want to support that?

-- 
Lee Jones
Linaro STMicroelectronics Landing Team Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog