Re: Threaded interrupts for synaptic touchscreen in HTC dream

[David Brownell <david-b@pacbell.net>]

On Wednesday 22 July 2009, Thomas Gleixner wrote:
> main interrupt
>      hardirq handler wakes main thread handler
> 
> main thread handler
>     bus magic
>         subdevice1 "hardirq" handler wakes subdevice1 irq thread
>         subdevice2 "hardirq" handler wakes subdevice2 irq thread

Why force this wasteful/undesired "all subdevices must
have their own IRQ handling thread" policy?

As previously noted, a single thread typically suffices.
There's no need to waste a dozen (or so) pages of memory
for such purposes ... these events are (as noted most
recently by Mark) infrequent/rare and performance isn't
a functionality gatekeeper.  Plus ...


>     main thread handler waits for subdevice1/2 handlers to complete

... sharing that thread can eliminate that synchronization
problem, and simplify the whole process.


> subdevice1 thread handler
>     bus magic
>     ....
>     thread_fn returns
>     signal main thread handler via completion
> 
> subdevice2 thread handler
>     bus magic
>     ....
>     thread_fn returns
>     signal main thread handler via completion
> 
> main thread handler resumes
>     bus magic
>     main thread handler returns from thread_fn
>     unmask main interrupt