Add usecounting for wakeup and sleep dependencies.  In the current
situation, if several functions add dependencies on the same
clockdomains, when the first dependency removal function is called,
the dependency will be incorrectly removed from the hardware.

Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which
provide a fast and usecounting-consistent way to clear all hardware
clockdomain dependencies, since accesses to these registers can be
quite slow.  pm{2,3}4xx.c has been updated to use these new functions.
The original version of this patch did not touch these files, which
previously wrote directly to the wkdep registers, and thus confused
the usecounting code.  This problem was found by Kevin Hilman
<khilman@deeprootsystems.com>.

N.B.: This patch introduces one significant functional difference over
the previous pm34xx.c code: sleepdeps are now cleared during
clockdomain initialization, whereas previously they were left
untouched.  This has been tested by Kevin and confirmed to work.

The original version of this patch also did not take into
consideration that some clockdomains do not have sleep or wakeup
dependency sources, which caused NULL pointer dereferences.  This
problem was debugged and fixed by Kevin Hilman
<khilman@deeprootsystems.com>.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Signed-off-by: NKevin Hilman <khilman@deeprootsystems.com>
Cc: Jouni Högander <jouni.hogander@nokia.com>

OMAP clockdomains and powerdomains are currently defined statically,
only registered at boot, and never unregistered, so we can remove the
unregister function and the locking.   A variant of this was originally
suggested a while ago by Dmitry Baryshkov <dbaryshkov@gmail.com>.
This version of this patch contains an additional fix from Kevin Hilman
<khilman@deeprootsystems.com> to address one of the pwrdm_for_each_nolock()
users in mach-omap2/pm-debug.c.   Thanks Kevin.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Dmitry Baryshkov <dbaryshkov@gmail.com>
Cc: Kevin Hilman <khilman@deeprootsystems.com>

Move clockdomain wakeup dependency and sleep dependency data
structures from the powerdomain layer to the clockdomain layer, where
they belong.  These dependencies were originally placed in the
powerdomain layer due to unclear documentation; however, it is clear
now that these dependencies are between clockdomains.  For OMAP2/3,
this is not such a big problem, but for OMAP4 this needs to be fixed.

Thanks to Benoît Cousson <b-cousson@ti.com> for his advice on this
patch.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

Port the APIs to support the OMAP4 clockdomain framework.
Also take care of the compiling requirements for the same.
Signed-off-by: NAbhijit Pagare <abhijitpagare@ti.com>
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoit Cousson <b-cousson@ti.com>
Cc: Rajendra Nayak <rnayak@ti.com>

Here the APIs are modified to use absolute addresses instead of module offsets.
Signed-off-by: NAbhijit Pagare <abhijitpagare@ti.com>
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoit Cousson <b-cousson@ti.com>
Cc: Rajendra Nayak <rnayak@ti.com>

Here the ".clkstctrl_reg" field is added to the clockdomain stucture
as the module offsets for OMAP4 do not map one to one for powerdomains
and clockdomains as it used to for OMAP3. Hence we need to use absolute
addresses to access the control registers. Some of the clock domains have
modules falling in the address space of PRM partition. Hence  necessitating
the use of absolute adresses.
Signed-off-by: NAbhijit Pagare <abhijitpagare@ti.com>
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoit Cousson <b-cousson@ti.com>
Cc: Rajendra Nayak <rnayak@ti.com>

Module offsets were same for OMAP2 and OMAP3 while they differ for OMAP4.
Hence we need different macros for identifying platform specific offsets.
Signed-off-by: NAbhijit Pagare <abhijitpagare@ti.com>
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoit Cousson <b-cousson@ti.com>
Cc: Rajendra Nayak <rnayak@ti.com>

In OMAP2/3 some of the clock-domains which did not have control
facility were being falsely written to and read using the CM_CLKSTCTRL
register though it did not exist for them. One check is added to remove
this flaw.
Signed-off-by: NAbhijit Pagare <abhijitpagare@ti.com>
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoit Cousson <b-cousson@ti.com>
Cc: Rajendra Nayak <rnayak@ti.com>

Avoid cluttering the Kconfig space with debug options that are rarely
used.  These can now be enabled and disabled by patching the "#undef DEBUG"
in the source files with "#define DEBUG", conforming to the practice for
the rest of the linux-omap code.

Also, while we're here, some lines in plat-omap/Kconfig use sets of
leading spaces when those lines should start with tabs.  Convert most
of them to use tabs.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Tony Lindgren <tony@atomide.com>

Move the remaining headers under plat-omap/include/mach
to plat-omap/include/plat. Also search and replace the
files using these headers to include using the right path.

This was done with:

#!/bin/bash
mach_dir_old="arch/arm/plat-omap/include/mach"
plat_dir_new="arch/arm/plat-omap/include/plat"
headers=$(cd $mach_dir_old && ls *.h)
omap_dirs="arch/arm/*omap*/ \
drivers/video/omap \
sound/soc/omap"
other_files="drivers/leds/leds-ams-delta.c \
drivers/mfd/menelaus.c \
drivers/mfd/twl4030-core.c \
drivers/mtd/nand/ams-delta.c"

for header in $headers; do
	old="#include <mach\/$header"
	new="#include <plat\/$header"
	for dir in $omap_dirs; do
		find $dir -type f -name \*.[chS] | \
			xargs sed -i "s/$old/$new/"
	done
	find drivers/ -type f -name \*omap*.[chS] | \
		xargs sed -i "s/$old/$new/"
	for file in $other_files; do
		sed -i "s/$old/$new/" $file
	done
done

for header in $(ls $mach_dir_old/*.h); do
	git mv $header $plat_dir_new/
done
Signed-off-by: NTony Lindgren <tony@atomide.com>

There is a possible race condition in clockdomain
code handling hw supported idle transitions.

When multiple autodeps dependencies are being added
or removed, a transition of still remaining dependent
powerdomain can result in false readings of the
state counter. This is especially fatal for off mode
state counter, as it could result in a driver not
noticing a context loss.

Fixed by disabling hw supported state transitions
when autodeps are being changed.
Signed-off-by: NKalle Jokiniemi <kalle.jokiniemi@digia.com>
Signed-off-by: NPaul Walmsley <paul@pwsan.com>

Add some infrastructure to easily iterate over clock and power
domains.
Signed-off-by: NPeter 'p2' De Schrijver <peter.de-schrijver@nokia.com>
Signed-off-by: NKevin Hilman <khilman@deeprootsystems.com>

This patch modifies the clock, clockdomain and OMAP3 specific
powerdomain code to call the PM counter infrastructure whenever one or
more powerdomains might have changed state.
Signed-off-by: NPeter 'p2' De Schrijver <peter.de-schrijver@nokia.com>
Signed-off-by: NKevin Hilman <khilman@deeprootsystems.com>

This patch provides the infrastructure to count how many times a
powerdomain entered a given power state (on, inactive, retention,
off). A number of functions are provided which will be called by the
chip specific powerdomain and clockdomain code whenever a transition
might have happened.
Signed-off-by: NPeter 'p2' De Schrijver <peter.de-schrijver@nokia.com>
Signed-off-by: NKevin Hilman <khilman@deeprootsystems.com>

Use pr_err() for errors rather than pr_debug().  pr_debug() are
compiled away unless -DDEBUG is used.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Fix the clockdomain autodep code to respect omap_chip platform flags.

Resolves "Unable to handle kernel paging request at virtual address
5f75706d" panic during power management initialization on OMAP2.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Signed-off-by: NTony Lindgren <tony@atomide.com>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Enabling clock in a disabled power domain causes the power domain to be
turned on. However, the power transition is not always finished when
clk_enable() returns and this randomly crashes the kernel when an
interrupt happens right after the clk_enable, and the kernel tries to
read the irq status register for that domain.

Why the irq status register is inaccessible, I don't know. Also it
doesn't seem to be related to the module being not powered up, but to
the transition itself.

The same could perhaps happen after clk_disable also, but I have not
witnessed that.

The problem affects at least dss, cam and sgx clocks.

This change waits for the transition to be finished before returning
from omap2_clkdm_clk_enable().
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@nokia.com>
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Signed-off-by: NTony Lindgren <tony@atomide.com>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

struct clockdomain contains a struct powerdomain *pwrdm and const char
*pwrdm_name.  The pwrdm_name is only used at initialization to look up
the appropriate pwrdm pointer.  Combining these into a union saves
about 100 bytes on 3430SDP.  This patch should not cause any change in
kernel function.

Updated to gracefully handle autodeps that contain invalid powerdomains,
per Russell King's review comments.

Boot-tested on BeagleBoard ES2.1.

linux-omap source commit is 718fc6cd4db902aa2242a736cc3feb8744a4c71a.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Signed-off-by: NTony Lindgren <tony@atomide.com>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Add clkdm_get_pwrdm() to the clockdomain code.  It will return a
pointer to the powerdomain struct that the clockdomain is contained
within.  Used by the PM code.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Signed-off-by: NTony Lindgren <tony@atomide.com>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Thie patch adds code to the powerdomain layer to track the
clockdomains associated with each powerdomain.

It also modifies the clockdomain code to register clockdomains
with their corresponding powerdomain when the clockdomain is registered.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Signed-off-by: NTony Lindgren <tony@atomide.com>

This patch creates an interface to the clockdomain registers in the
PRM/CM modules on OMAP2/3.  This interface is intended to be used by
PM code, e.g., pm.c; not by device drivers directly.

The patch also adds clockdomain usecount tracking.  This is intended
to be called whenever the first clock in a clockdomain is enabled, or
when the last enabled clock in a clockdomain is disabled.  If the
clockdomain is in software-supervised mode, the code will force-wakeup
or force-sleep the clockdomain.  If the clockdomain is in
hardware-supervised mode, the first clock enable will add sleep and
wakeup dependencies on a user-selectable set of parent domains (usually
MPU & IVA2), and the disable will remove them.

Each clockdomain will be defined in later patches as static
structures.  The clockdomain structures are linked into a list at boot
by clkdm_register(), similar to the OMAP clock code.

The patch adds a Kconfig option, CONFIG_OMAP_DEBUG_CLOCKDOMAIN, which
when enabled will emit verbose debug messages via pr_debug().
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Signed-off-by: NTony Lindgren <tony@atomide.com>