| Message ID | 527CB3A1.7050808@koalo.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 11/08/2013 02:49 AM, Florian Meier wrote: > In order to find I2C devices in the device tree, the platform nodes > have to be known by the I2C core. Analogous to the i2c-omap driver > this requires setting the dev.of_node parameter of the adapter. (CCing the I2C maintainers...) > diff --git a/drivers/i2c/busses/i2c-bcm2835.c b/drivers/i2c/busses/i2c-bcm2835.c > @@ -299,6 +299,7 @@ static int bcm2835_i2c_probe(struct platform_device *pdev) > strlcpy(adap->name, "bcm2835 I2C adapter", sizeof(adap->name)); > adap->algo = &bcm2835_i2c_algo; > adap->dev.parent = &pdev->dev; > + adap->dev.of_node = pdev->dev.of_node; Ah, that makes sense. Thinking about it now, I'd only ever used i2cget etc. to access I2C devices, rather than instantiating drivers from DT. That all said, I wonder if the I2C core shouldn't do something like the following inside i2c_add_adapter(): if (!adap->dev.of_node && adap->dev.parent) adap->dev.of_node = adap->dev.parent->of_node; That would save every single I2C driver from having to set up this field manually.
On 11/08/2013 09:59 AM, Stephen Warren wrote: > On 11/08/2013 02:49 AM, Florian Meier wrote: >> In order to find I2C devices in the device tree, the platform nodes >> have to be known by the I2C core. Analogous to the i2c-omap driver >> this requires setting the dev.of_node parameter of the adapter. > > (CCing the I2C maintainers...) > >> diff --git a/drivers/i2c/busses/i2c-bcm2835.c b/drivers/i2c/busses/i2c-bcm2835.c > >> @@ -299,6 +299,7 @@ static int bcm2835_i2c_probe(struct platform_device *pdev) >> strlcpy(adap->name, "bcm2835 I2C adapter", sizeof(adap->name)); >> adap->algo = &bcm2835_i2c_algo; >> adap->dev.parent = &pdev->dev; >> + adap->dev.of_node = pdev->dev.of_node; > > Ah, that makes sense. Thinking about it now, I'd only ever used i2cget > etc. to access I2C devices, rather than instantiating drivers from DT. > > That all said, I wonder if the I2C core shouldn't do something like the > following inside i2c_add_adapter(): > > if (!adap->dev.of_node && adap->dev.parent) > adap->dev.of_node = adap->dev.parent->of_node; > > That would save every single I2C driver from having to set up this field > manually. BTW, this should probably be: Cc: stable@vger.kernel.org
On Fri, Nov 08, 2013 at 09:59:28AM -0700, Stephen Warren wrote: > On 11/08/2013 02:49 AM, Florian Meier wrote: > > In order to find I2C devices in the device tree, the platform nodes > > have to be known by the I2C core. Analogous to the i2c-omap driver > > this requires setting the dev.of_node parameter of the adapter. > > (CCing the I2C maintainers...) > > > diff --git a/drivers/i2c/busses/i2c-bcm2835.c b/drivers/i2c/busses/i2c-bcm2835.c > > > @@ -299,6 +299,7 @@ static int bcm2835_i2c_probe(struct platform_device *pdev) > > strlcpy(adap->name, "bcm2835 I2C adapter", sizeof(adap->name)); > > adap->algo = &bcm2835_i2c_algo; > > adap->dev.parent = &pdev->dev; > > + adap->dev.of_node = pdev->dev.of_node; > > Ah, that makes sense. Thinking about it now, I'd only ever used i2cget > etc. to access I2C devices, rather than instantiating drivers from DT. > > That all said, I wonder if the I2C core shouldn't do something like the > following inside i2c_add_adapter(): > > if (!adap->dev.of_node && adap->dev.parent) > adap->dev.of_node = adap->dev.parent->of_node; Should this not also have an of_node_get to increment the ref count on the node? Thanks, Charles
On Tue, Nov 26, 2013 at 01:05:53PM +0000, Charles Keepax wrote: > On Fri, Nov 08, 2013 at 09:59:28AM -0700, Stephen Warren wrote: > > That all said, I wonder if the I2C core shouldn't do something like the > > following inside i2c_add_adapter(): > > > > if (!adap->dev.of_node && adap->dev.parent) > > adap->dev.of_node = adap->dev.parent->of_node; > Should this not also have an of_node_get to increment the ref > count on the node? Given that it's pointing to the device that registered the I2C adaptor it should be able to assume that there's a reference already.
diff --git a/drivers/i2c/busses/i2c-bcm2835.c b/drivers/i2c/busses/i2c-bcm2835.c index ea4b08f..8beecfa 100644 --- a/drivers/i2c/busses/i2c-bcm2835.c +++ b/drivers/i2c/busses/i2c-bcm2835.c @@ -299,6 +299,7 @@ static int bcm2835_i2c_probe(struct platform_device *pdev) strlcpy(adap->name, "bcm2835 I2C adapter", sizeof(adap->name)); adap->algo = &bcm2835_i2c_algo; adap->dev.parent = &pdev->dev; + adap->dev.of_node = pdev->dev.of_node; bcm2835_i2c_writel(i2c_dev, BCM2835_I2C_C, 0);
In order to find I2C devices in the device tree, the platform nodes have to be known by the I2C core. Analogous to the i2c-omap driver this requires setting the dev.of_node parameter of the adapter. Signed-off-by: Florian Meier <florian.meier@koalo.de> --- I don't know if this is really necessary, but for me it does not work without this line. drivers/i2c/busses/i2c-bcm2835.c | 1 + 1 file changed, 1 insertion(+)