On 8 November 2012 22:08, Dmitry Torokhov [off-list ref] wrote:

On Thu, Nov 08, 2012 at 07:10:47PM +0530, Viresh Kumar wrote:
It also breaks the error unwinding/removal of the driver as it frees
input device while IRQ handler is still active.

I have heard of this argument before, probably from you. :)
Just need clarification again. How will we get an interrupt when the controller
is stopped, unless we have a shared irq.

I will push the patch which implements devm_input_allocate_device() to
my 'next' branch in a few, please use it because it will make sure input
device will stick around long enough.

Will surely use that.

quoted

+     kbd->io_base = devm_ioremap(&pdev->dev, res->start, resource_size(res));

There is devm_request_and_ioremap() which you can use here.

Should have been done in V1 only.

--
viresh