This is in preparation for teaching async_synchronize_full() to sync all
pending async work, and not just on the async_running domain.  This
conversion is functionally equivalent, just embedding the existing list
in a new async_domain type.

The .registered attribute is used in a later patch to distinguish
between domains that want to be flushed by async_synchronize_full()
versus those that only expect async_synchronize_{full|cookie}_domain to
be used for flushing.

[jejb: add async.h to scsi_priv.h for struct async_domain]
Signed-off-by: NDan Williams <dan.j.williams@intel.com>
Acked-by: NArjan van de Ven <arjan@linux.intel.com>
Acked-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Tested-by: NEldad Zack <eldad@fogrefinery.com>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

REGULATOR_STATUS_UNDEFINED is to be returned by regulator, if any other state
doesn't really apply.
Signed-off-by: NKrystian Garbaciak <krystian.garbaciak@diasemi.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Case REGULATOR_STATUS_STANDBY -> REGULATOR_MODE_STANDBY.
Signed-off-by: NKrystian Garbaciak <krystian.garbaciak@diasemi.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Since DT doesn't provide an idiomatic mechanism for enabling full
constraints and since it's much more natural with DT to provide them
just assume that a DT enabled system has full constraints.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: NLiam Girdwood <lrg@ti.com>

With changes introduced by commit 222cc7b1 (regulator: core: Allow
multiple requests of a single supply mapping) on create_regulator,
regulator_put needs a corresponding update on sysfs entry removing.

Also regulator->dev still needs to get assigned in create_regulator,
otherwise, sysfs_remove_link call in regulator_put will get bypassed.
Reported-by: NFabio Estevam <festevam@gmail.com>
Tested-by: NDong Aisheng <dong.aisheng@linaro.org>
Signed-off-by: NShawn Guo <shawn.guo@linaro.org>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

It is very common for regulators to support having their enable signal
controlled by a GPIO. Since there are a bunch of fiddly things to get
right like handling the operations when the enable signal is tied to
a rail and it's just replicated code add support for this to the core.

Drivers should set ena_gpio in their config if they have a GPIO control,
using ena_gpio_flags to specify any flags (including GPIOF_OUT_INIT_ for
the initial state) and ena_gpio_invert if the GPIO is active low. The
core will then override any enable and disable operations the driver has
and instead control the specified GPIO.

This will in the future also allow us to further extend the core by
identifying when several enable signals have been tied together and
handling this properly.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Create new _regulator_do_enable() and _regulator_do_disable() operations
which deal with the mechanics of performing the enable and disable, partly
to cut down on the levels of indentation and partly to support some future
work.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Many regulators have a fixed specification for their enable time. Allow
this to be set in the regulator_desc as a number to save them having to
implement an explicit operation.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Lots of regulator drivers have checks in their map_voltage() functions
to verify that the result of the mapping is in the range originally
specified. Factor these out in the core and provide a bit of extra
defensiveness for other drivers by doing the check in the core.

Since we're now doing a list_voltage() earlier move the current mapping
back to a voltage out into the set_voltage() call to save redoing it.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Currently regulator_is_supported_voltage() works by enumerating the set
of voltages which can be set by the regulator but the checks we're doing
to impose constraints mean that if we can't vary the voltage we'll not
report any voltages as supported even though the regulator is actually
set at that voltage.

We could fix the voltage listing but this would mean that list_voltage()
could end up going to the hardware to get the current voltage which isn't
expected (it's supposed to be very cheap) so instead special case things
when we can't change the voltage and compare the requested range against
the current voltage.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

A recursive lockdep warning occurs if you call
regulator_set_optimum_mode() on a regulator with a supply because
there is no nesting annotation for the rdev->mutex. To avoid this
warning, get the supply's load before locking the regulator's
mutex to avoid grabbing the same class of lock twice.

=============================================
[ INFO: possible recursive locking detected ]
3.4.0 #3257 Tainted: G        W
---------------------------------------------
swapper/0/1 is trying to acquire lock:
 (&rdev->mutex){+.+.+.}, at: [<c036e9e0>] regulator_get_voltage+0x18/0x38

but task is already holding lock:
 (&rdev->mutex){+.+.+.}, at: [<c036ef38>] regulator_set_optimum_mode+0x24/0x224

other info that might help us debug this:
 Possible unsafe locking scenario:

       CPU0
       ----
  lock(&rdev->mutex);
  lock(&rdev->mutex);

 *** DEADLOCK ***

 May be due to missing lock nesting notation

3 locks held by swapper/0/1:
 #0:  (&__lockdep_no_validate__){......}, at: [<c03dbb48>] __driver_attach+0x40/0x8c
 #1:  (&__lockdep_no_validate__){......}, at: [<c03dbb58>] __driver_attach+0x50/0x8c
 #2:  (&rdev->mutex){+.+.+.}, at: [<c036ef38>] regulator_set_optimum_mode+0x24/0x224

stack backtrace:
[<c001521c>] (unwind_backtrace+0x0/0x12c) from [<c00cc4d4>] (validate_chain+0x760/0x1080)
[<c00cc4d4>] (validate_chain+0x760/0x1080) from [<c00cd744>] (__lock_acquire+0x950/0xa10)
[<c00cd744>] (__lock_acquire+0x950/0xa10) from [<c00cd990>] (lock_acquire+0x18c/0x1e8)
[<c00cd990>] (lock_acquire+0x18c/0x1e8) from [<c080c248>] (mutex_lock_nested+0x68/0x3c4)
[<c080c248>] (mutex_lock_nested+0x68/0x3c4) from [<c036e9e0>] (regulator_get_voltage+0x18/0x38)
[<c036e9e0>] (regulator_get_voltage+0x18/0x38) from [<c036efb8>] (regulator_set_optimum_mode+0xa4/0x224)
...
Signed-off-by: NStephen Boyd <sboyd@codeaurora.org>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

commit 222cc7b1 (regulator: core: Allow multiple requests of a single supply mapping)
removed the usage of get_device_regulator().

Remove the function definition too amd get rid of the following warning:

drivers/regulator/core.c:112:26: warning: 'get_device_regulator' defined but not used [-Wunused-function]
Signed-off-by: NFabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Sometimes it may be useful to allow a device to request a supply multiple
times, for example in order to allow framework management of some uses of
the supply with some additional driver specific management or in order to
allow multiple children of an MFD to work with the supply. Currently this
is not possible due to the creation of

Solve this by removing the requested_uA entry (we have no current users
of this feature anyway) and ignoring errors creating the symlink to the
consumer. We should do something nicer than this as this causes sysfs to
spew enormous warnings but it allows users to run for now.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: NLiam Girdwood <lrg@ti.com>

There is no need to consider waiting for the voltage to ramp if we
didn't manage to set it and looking at the return value is going to be
cheaper than is_enabled().
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

With this change, regulator_set_voltage_time_sel() can be more generic and not
limited to linear and table based mapping now.
One side-effect of this change is that list_voltage() must be implemented.
Signed-off-by: NAxel Lin <axel.lin@gmail.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

This change makes it possible to set ramp_delay with 0.xxx mV/uS without
truncation issue.
Signed-off-by: NAxel Lin <axel.lin@gmail.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

For table based mapping, we can calculate voltage difference by below equation:
abs(rdev->desc->volt_table[new_selector] - rdev->desc->volt_table[old_selector])

Thus we can make regulator_set_voltage_time_sel work for table based mapping.

regulator_set_voltage_time_sel() only supports linear or table based mapping.
In case it is misused, also warn if neither linear nor table based mapping
is used with regulator_set_voltage_time_sel().
Signed-off-by: NAxel Lin <axel.lin@gmail.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

For some hardwares ramp_delay for BUCKs is a configurable parameter which can
be configured through DT or board file.This patch adds ramp_delay to regulator
constraints and allow user to configure it for regulators which supports this
feature, through DT or board file. It will provide two ways of setting the
ramp_delay for a regulator:
	First, by setting it as constraints in board file(for configurable
regulators) and set_machine_constraints() will take care of setting it on
hardware by calling(the provided) .set_ramp_delay() operation(callback).
	Second, by setting it as data in regulator_desc(as fixed/default
ramp_delay rate) for a regulator in driver.

regulator_set_voltage_time_sel() will give preference to
constraints->ramp_delay while reading ramp_delay rate for regulator. Similarly
users should also take care accordingly while refering ramp_delay rate(in case
of implementing their private .set_voltage_time_sel() callbacks for different
regulators).

[Rewrote subject for 80 columns -- broonie]
Signed-off-by: NYadwinder Singh Brar <yadi.brar@samsung.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

minor optimization: move delay code to where delay is set and
thus where it is used vs in the main line path.
Signed-off-by: NPhilip Rakity <prakity@marvell.com>
Acked-by: NAxel Lin <axel.lin@gmail.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

The voltage being set should be passed to the call in handler
requesting the callback.  Currently this is not done.

The calling handler cannot call regulator_get_voltage() to get the
information since the mutex is held by the regulator and
deadlock occurs.

Without this change the receiver of the call in cannot know what
action to take.  This is used, for example, to set PAD voltages
when doing SD vccq voltage changes.

[Review and spelling fix in the commit log from Pankaj Jangra]
Signed-off-by: NPhilip Rakity <prakity@marvell.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Reported-by: NPankaj Jangra <jangra.pankaj9@gmail.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

rdev->desc->uV_step * abs(new_selector - old_selector) returns uV.
The unit of ramp_delay is mV/us.

Current code multiples 1000 at wrong place.
Signed-off-by: NAxel Lin <axel.lin@gmail.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

This patch adds regulator_set_voltage_time_sel(), to move into core, the
commonly used code by drivers to provide the .set_voltage_time_sel callback.
It will also allow us to configure different ramp delay for different
regulators easily.
Signed-off-by: NYadwinder Singh Brar <yadi.brar@samsung.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Fixed voltage is a kind of linear mapping where n_voltages is 1.
This change allows [list|map]_voltage_linear to be used for fixed
voltage.

For fixed voltage, n_voltages is 1 and the only valid selector is 0.
Thus we actually don't care the uV_step setting.
Signed-off-by: NAxel Lin <axel.lin@gmail.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Properly handle the case if the specified min_uV is less than the voltage given
by the lowest selector.
Signed-off-by: NAxel Lin <axel.lin@gmail.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Signed-off-by: NAxel Lin <axel.lin@gmail.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

If no regmap is explicitly specified then use dev_get_regmap() to obtain
one. The driver must explicitly enable any actual usage of the regmap
so there's no concern with unwanted usage.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: NLiam Girdwood <lrg@ti.com>

devres_release() will call the destructor for the resource as well as
freeing the devres data.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: NLiam Girdwood <lrg@ti.com>

If the regulator is not on, it won't take time setting new voltage.
So only call set_voltage_time_sel() to get the necessary delay when
the regulator is on.
Signed-off-by: NAxel Lin <axel.lin@gmail.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Some regulator hardware use table based mapping can set volt_table in
regulator_desc and use regulator_list_voltage_table() for their list_voltage
callback.
Signed-off-by: NAxel Lin <axel.lin@gmail.com>
Acked-by: NLiam Girdwood <lrg@ti.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: NLiam Girdwood <lrg@ti.com>

If we fail while registering a regulator make sure we release the supply
for the regulator if there is one.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: NLiam Girdwood <lrg@ti.com>
Cc: stable@vger.kernel.org

Integer division may truncate the result.
Use DIV_ROUND_UP to ensure simple linear voltage mappings falls within the
specified range.
Signed-off-by: NAxel Lin <axel.lin@gmail.com>
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>

A lot of regulator hardware maps selectors on to voltages with a simple
linear mapping function

    selector = base + (selector * step size)

Provide off the shelf list_voltage() and map_voltage() operations which
use new min_uV and uV_step members in the regulator_desc to implement
this function.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: NLiam Girdwood <lrg@ti.com>

In order to allow more drivers to factor things out into data allow
drivers to provide a mapping function to convert voltages into selectors.
This allows any driver to use set_voltage_sel(). The existing mapping
based on iterating over list_voltage() is provided as an operation which
can be assigned to the new map_voltage() function though for ease of
transition it is treated as the default.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: NLiam Girdwood <lrg@ti.com>

The core really wants a struct device to be supplied for regulators and
there's no reason this should be impossible so provide one so complain
if we didn't get one.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: NLiam Girdwood <lrg@ti.com>

It turns out that (quite surprisingly) devres_destroy() only undoes the
devres mapping, it doesn't destroy the underlying resource, meaning that
anything using devm_regulator_put() would leak. While we wait for the new
devres_release() which does what we want to get merged open code it in
devm_regulator_put().
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: NLiam Girdwood <lrg@ti.com>

If a regulator is always on for any reason then cache that when the
consumer is created and use it to optimise away the need to take locks
or recurse up the supply tree when consumers do enable or disable calls.
The scheduling of asynchronous work for bulk enables is also skipped.

We don't actually check if the device physically supports control on the
basis that constraints allowing status changes on physically always on
regulators are nonsensical anyway.

This is a very common pattern in hardware - it's normal to have some
power supplies that have either no software control or are critical to
system function - so many systems should be able to benefit.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: NLiam Girdwood <lrg@ti.com>