On 08/02/17 09:43, Johan Hovold wrote:
On Tue, Feb 07, 2017 at 02:31:04PM +0000, Bryan O'Donoghue wrote:
On 07/02/17 14:19, Johan Hovold wrote:
On Mon, Jan 23, 2017 at 01:04:14PM +0100, Johan Hovold wrote:
Add a struct timer_list to struct gb_operation and use that to implement generic operation timeouts.
This simplifies the synchronous operation handling somewhat while also providing a generic timeout mechanism that drivers can use for asynchronous operations.
Signed-off-by: Johan Hovold johan@kernel.org
Greg,
I believe you can apply this one now. This is the right way to implement operation timeouts, and it is independent of Bryan's patches converting loopback to use the new interface, which can be applied on top when they are ready.
Thanks, Johan
My approach using a single timer which either times out or is cancelled upon completion is about as efficient as this can get, and therefore also allows the current synchronous-operation implementation to be built on top of this generic mechanism.
I'm just wondering what impact it has instead of wait_event_interruptible() more/less overhead (I suspect more) - and I reckoned you'd not be on for that change - so only made a change on the asynchronous path.
But whatever you're happier with yourself.