Hello,
On 04/10/2016 17:42, Maxime Ripard wrote:
Hi,
On Tue, Oct 04, 2016 at 02:15:16PM +0200, Thomas Petazzoni wrote:
Hello,
On Tue, 4 Oct 2016 11:46:16 +0200, Mylène Josserand wrote:
#include <sound/dmaengine_pcm.h>
#include <sound/pcm_params.h>
@@ -589,6 +590,7 @@ static int sun4i_i2s_probe(struct platform_device *pdev)
{
struct sun4i_i2s *i2s;
struct resource *res;
+ struct reset_control *reset_apb;
void __iomem *regs;
int irq, ret;
@@ -626,7 +628,19 @@ static int sun4i_i2s_probe(struct platform_device *pdev)
dev_err(&pdev->dev, "Can't get our mod clock\n");
return PTR_ERR(i2s->mod_clk);
}
-
+
+ reset_apb = devm_reset_control_get(&pdev->dev, "apb_reset");
I believe this is a change in the Device Tree binding, since you're
adding support for a new resource. Perhaps the Device Tree binding
documentation should be updated accordingly?
Indeed.
You have two solutions to do that:
- Either mark it as optional and use reset_control_get_optional
(because here, you broke the other SoCs that have that controller
but no reset line)
- Or introduce a new compatible, and make the reset property
mandatory for that new compatible.
I prefer the latter, since you get a stricter error check, and you
cannot end up in a situation where your driver probes but is
useless. But you'll find both in our drivers.
Okay, thank you for the hints!
+ }
+
+ ret = reset_control_deassert(reset_apb);
+ if (ret < 0) {
+ dev_err(&pdev->dev, "Can't deassert apb reset (%d)\n", ret);
+ return ret;
+ }
Do you need to re-assert the reset line in the ->remove() hook?
Even better, you can add it to the runtime_pm hooks! :)
I will have a look to runtime_pm and update it for a V2.
--
Mylène Josserand, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com
