This fixes up the simplified multi-evt handling when sparseirq support is
enabled. While vectors are redirected through the single unique masking
source, each one of the redirected vectors still requires its own backing
irq_desc, which needs to be manually allocated in the sparseirq case.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>

This patch changes the way in which "multi-evt" interrups are handled.
The intc_evt2irq_table and related intc_evt2irq() have been removed and
the "redirecting" handler is installed for the coupled interrupts.

Thanks to that the do_IRQ() function don't have to use another level
of indirection for all the interrupts...

Signed-off-by: Pawel Moll <pawel.moll@st.com>
Signed-off-by: Stuart Menefy <stuart.menefy@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>

It is possible for the CPU to re-enable it's interrupt block bit
before the write to the interrupt controller has actually masked out
the external interupt at the controller. We get around this by
reading back from the interrupt controller which will ensure the
write has happened.

Signed-off-by: Stuart Menefy <stuart.menefy@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>

Now that the slab allocators are available much earlier, this triggers a
the slab_is_available() warning when registering the interrupt
controller. Convert to kzalloc() with GFP_NOWAIT, as per the generic
changes.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>

Now that the dependent patches are merged, we are ready to enable
sparseirq support. This simply adds the Kconfig option, and then converts
from the _cpu to the _node allocation routines to follow the upstream
sparseirq API changes.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>

irq_to_desc_alloc_cpu() has been renamed to irq_to_desc_alloc_node() in
-next, but as we can not presently enable SPARSE_IRQ without the early
irq_desc alloc patch, protect it with an ifdef until the interface has
settled and we are ready to enable it system-wide.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>

This hooks in irq_to_desc_alloc_cpu() to the necessary code paths in the
intc and ipr controller registration paths. As these are the primary call
paths for all SH CPUs, this alone will make all CPUs sparse IRQ ready.

There is the added benefit now that each CPU contains specific IPR and
INTC tables, so only the vectors with interrupt sources backing them will
ever see an irq_desc instantiation. This effectively packs irq_desc
down to match the CPU, rather than padding NR_IRQS out to cover the valid
vector range.

Boards with extra sources will still have to fiddle with the nr_irqs
setting, but they can continue doing so through the machvec as before.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>

It's required for all modules loaded in the previous runtime
session because not initilized duing the kernel start-up.

Signed-off-by: Francesco Virlinzi <francesco.virlinzi@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>

Add set_irq_wake() support to intc using sysdev and suspend.

The intc controllers are put on a list at registration time
and registered as sysdev devices later on during the boot.

The sysdev class suspend callback is used to find irqs with
wakeup enabled belonging to our intc controller. Such irqs
are simply enabled so wakeup interrupts may reach the cpu.

Signed-off-by: Magnus Damm <damm@igel.co.jp>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>

Modify the intc code to install a disable callback. The current
solution without a disable callback results in use of the
generic default_disable() function. This function is a no-op
so suspend_device_irqs() will not disable any intc interrupts
at suspend time without this patch. Also, install enable and
shutdown callbacks while at it.

Signed-off-by: Magnus Damm <damm@igel.co.jp>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>