On Fri, Mar 10, 2017 at 09:28:08AM -0800, Ray Jui wrote:
> Hi Julia/Linus,
>
> On 3/9/2017 8:21 AM, Julia Cartwright wrote:
> > The bcm-kona gpio driver currently implements an irq_chip for handling
> > interrupts; due to how irq_chip handling is done, it's necessary for the
> > irq_chip methods to be invoked from hardirq context, even on a a
> > real-time kernel.  Because the spinlock_t type becomes a "sleeping"
> > spinlock w/ RT kernels, it is not suitable to be used with irq_chips.
> > 
> > A quick audit of the operations under the lock reveal that they do only
> > minimal, bounded work, and are therefore safe to do under a raw spinlock.
>
> This is new to me. But it seems like, for the vast majority cases, user
> can still continue to use spin_lock as it is without needing to worry
> about the underlying difference between standard or RT kernels.

If by "user" you mean, "driver developer", then yes.  For most driver
authors, the distinction between raw and non-raw spinlocks is
irrelevant, they can use spinlocks and everything will work out just
fine w/ mainline and on RT.

> But in certain cases, e.g., irq_chips, extra care needs to be done,
> i.e., swtching to use raw spin lock to make sure that it is not
> blocking in the case of RT.

Correct, on RT the goal is to push as much as possible into a
preemptible context, including driver interrupts, etc.  However, there
are still codepaths which necessarily need to be executed in hardirq
context, including anything necessary to support scheduling.  This
includes: interrupt-dispatching (irq_chips), timers, and the scheduler
itself, which is why this "core" code must use the raw spinlock
variants.

> Is such API use change well accepted by the open source community
> already?

In what way is this an API change?  The API isn't changing, what's
changing in this patch is to fix what is an irq_chip implementation
which is _currently broken_ on RT.

   Julia