For many drivers which will support rich endianness of Devices
need define DT properties by itself with the binding support.

The endianness using regmap:
Index      Device     Properties if needs bytes-swap,
                      or just ignore it
-------------------------------------------------------------
1          BE         'big-endian'
2          LE         'little-endian'

The properties include all the register values and the buffers.
And these properties are very usful for the MMIO devices:

Such as: a memory-mapped device, on one SoC is in BE mode, while
in another SoC will be in LE mode, and the CPU will always in LE
mode.

For the first case, we must use cpu_to_be32/be32_to_cpu for
32-bit registers accessing, so the 'big-endian' property is needed.

For the second case, we can just ignore the bytes-swap
functions like cpu_to_le32/le32_to_cpu, so the 'little-endian'
property could be abscent.

And vice versa...
Signed-off-by: NXiubo Li <Li.Xiubo@freescale.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Signed-off-by: NMark Brown <broonie@linaro.org>

Add a new function regmap_get_device to obtain the underlying struct
device from a regmap.
Signed-off-by: NTuomas Tynkkynen <ttynkkynen@nvidia.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Commit 93258040
"regmap: mmio: Add support for 1/2/8 bytes wide register address."
broke regmap_mmio_write for uneven counts, for example 32-bit register
addresses with no padding and 8-byte values (count = 5).
Fix this by allowing all counts large enough to include some value.
This check was BUG_ON(count < 4) before the last change.
Signed-off-by: NPhilipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: NMark Brown <broonie@linaro.org>

Since we cannot make sure the 'chip->num_regs' will always be none zero
from the users, and then if 'chip->num_regs' equals to zero by mistake
or other reasons, the kzalloc() will return ZERO_SIZE_PTR, which equals
to ((void *)16).

So this patch fix this with just checking the 'chip->num_regs' before
calling kzalloc().

This also sorts the header files in alphabetical order at the same time.
Signed-off-by: NXiubo Li <Li.Xiubo@freescale.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

drivers/base/regmap/regmap.c: In function ‘_regmap_range_multi_paged_reg_write’:
drivers/base/regmap/regmap.c:1665: warning: ‘this_page’ may be used uninitialized in this function
Signed-off-by: NGeert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: NMark Brown <broonie@linaro.org>

Since we cannot make sure the 'len = pair_size * num_regs' will always
be none zero from the users, and then if 'num_regs' equals to zero by
mistake or other reasons, the kzalloc() will return ZERO_SIZE_PTR, which
equals to ((void *)16).

So this patch fix this with just doing the 'len' zero check before calling
kzalloc().
Signed-off-by: NXiubo Li <Li.Xiubo@freescale.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Some I2C adapters are only compatible with the SMBus protocol and do not
support standard I2C transfers.

Fallback to SMBus transfers if we encounter such kind of adapters.
The transfer type is chosen according to the val_bits field in the regmap
config.
Signed-off-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Some busses do not support sending/receiving multiple registers in one go.
Such kind of busses just unpack the registers that have been previously
packed by the regmap core or pack registers that will be later unpacked by
the core code.

Add reg_write and reg_read callbacks in order to optimize access through
this kind of busses.
Signed-off-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Change regcache_rbtree_node strcuture fields order to align the pointers on
64bits architectures.
Signed-off-by: NJean-Christophe PINCE <jean-christophe.pince@intel.com>
Signed-off-by: NDavid Cohen <david.a.cohen@linux.intel.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

'offset = *(u32 *)reg;'
This will be okey for 32/64-bits register device, but for 8/16-bits
register ones, the 'offset' value will overflow, for example:

The IMX2 Watchdog, whose registers and values are all 16-bits:

If the IO base virtual address is ctx->regs = 0x888c0000, and the now
doing the 0x00 register accessing:
Using 'offset = *(u32 *)reg' the offset value will possiblly be 0x77310000,
Using 'offset = *(u16 *)reg' the offset value will be 0x0000.

In the regmap_mmio_gather_write(), ctx->regs + 0x7731000 will be 0xffbd0000,
but actually it should be ctx->regs + 0x0000 = 0x888c0000.
Signed-off-by: NXiubo Li <Li.Xiubo@freescale.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Allow busses to request little endianness formatting and
parsing for 16- and 32-bit values. This will be useful to
support regmap-mmio.

For the following the scenarios using the regmap-mmio,
for example:

Index    CPU       Device     Endianess flag for values
----------------------------------------------------------
1        LE        LE         REGMAP_ENDIAN_DEFAULT/NATIVE
2        LE        BE         REGMAP_ENDIAN_BIG
3        BE        BE         REGMAP_ENDIAN_DEFAULT/NATIVE
4        BE        LE         REGMAP_ENDIAN_LITTLE

For one device driver, which will support all the cases above,
needs two boolean properties in DT node like: 'big-endian'
for case 2 and 'little-endian' for case 4, and for cases 1
and 3 they all will be absent.
Signed-off-by: NXiubo Li <Li.Xiubo@freescale.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

It need to add curly braces because the inner for "if" has
two statements.

coccicheck says:
 drivers/base/regmap/regmap.c:765:2-44:
code aligned with following code on line 766
Signed-off-by: NDaeseok Youn <daeseok.youn@gmail.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Fix the support for 1/2/8 bytes wide register address checking.
Signed-off-by: NXiubo Li <Li.Xiubo@freescale.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Since regmap core and mmio have already support for 1/2/8 bytes wide values,
so adds support for 1/2/8 bytes wide registers address.
Signed-off-by: NXiubo Li <Li.Xiubo@freescale.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Signed-off-by: NXiubo Li <Li.Xiubo@freescale.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

In the regcache_default_sync, if a register isn't writeable, then
_regmap_write will return an error and the rest of the sync will be
aborted.  Avoid this by checking if a register is writeable before
trying to sync it.
Signed-off-by: NDylan Reid <dgreid@chromium.org>
Signed-off-by: NMark Brown <broonie@linaro.org>

The default sync operation was still assuming a stride of one, fix it
to respect the reg_stride set in the map.
Signed-off-by: NDylan Reid <dgreid@chromium.org>
Signed-off-by: NMark Brown <broonie@linaro.org>

regmap deploys the spinlock for the protection when set up in fast_io
mode.  This may lead to sleep-in-atomic by memory allocation with
GFP_KERNEL in regmap_bulk_write().  This patch fixes it by moving the
allocation out of the lock.

[Fix excessively large locked region -- broonie]
Signed-off-by: NTakashi Iwai <tiwai@suse.de>
Signed-off-by: NMark Brown <broonie@linaro.org>

With fast_io we use mutexes to lock the I/O operations so we would need
to do GFP_ATOMIC allocations if we wanted to do allocations inside the
lock as we do currently. Since it is unlikely that we will want to register
a patch outside of init where concurrency shouldn't be an issue move the
allocation of the patch data outside the lock.
Reported-by: NTakashi Iwai <tiwai@suse.de>
Signed-off-by: NMark Brown <broonie@linaro.org>

After setting the 'data' pointer (wchich is returned to the caller for
freeing later) the regmap_add_irq_chip() could still fail for various
reasons (ENOMEM, regmap_read or regmap_write failure). In such case the
memory under 'data' was freed in error path and error value was returned
but the 'data' variable was not changed.

This could lead to errors if the caller passed such 'data' to
regmap_del_irq_chip().

The 'data' pointer should be changed atomically from the caller
perspective - set it only on regmap_add_irq_chip() success.
Signed-off-by: NKrzysztof Kozlowski <k.kozlowski@samsung.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

This is the implementation of regmap_multi_reg_write()

There is a new capability 'can_multi_write' that device drivers
must set in order to use this multi reg write mode.

This replaces the first definition, which just defined the API.
Signed-off-by: NAnthony Olech <anthony.olech.opensource@diasemi.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

    In some cases, we need regmap's format parse_val function
to do be/le translation according to the bus configuration.
For example, snd_soc_bytes_put() uses regmap to write/read values,
and use cpu_to_be() directly to covert MASK into big endian. This
is a defect, and should use regmap's format function to do it according
to bus configuration.
Signed-off-by: NNenghua Cao <nhcao@marvell.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Since we now have an internal version of regmap_multi_reg_write use this
to apply the register patch.
Signed-off-by: NCharles Keepax <ckeepax@opensource.wolfsonmicro.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Devices with more complex boot proceedures may occasionally apply the
register patch manual. regmap_multi_reg_write is a logical way to do so,
however the patch must be applied with cache bypass on, such that it
doesn't override any user settings. This patch adds a
regmap_multi_reg_write_bypassed function that applies a set of writes
with the bypass enabled.
Signed-off-by: NCharles Keepax <ckeepax@opensource.wolfsonmicro.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

There should be no need for the writes supplied to this function to be
edited by it so mark them as const.
Signed-off-by: NCharles Keepax <ckeepax@opensource.wolfsonmicro.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

   /drivers/base/regmap/regmap.c:717:6-33: WARNING:
Comparison to bool.

   More information about semantic patching is available at
http://coccinelle.lip6.fr/Signed-off-by: NNenghua Cao <nhcao@marvell.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Currently, we check the registers in the patch are aligned to the
register stride everytime we sync the cache and the first time the patch
is written out is unchecked.

This patch checks the register patch when we first register it so the
first writes are no longer unchecked and then doesn't check on
subsequent syncs as the patch will be unchanged.
Signed-off-by: NCharles Keepax <ckeepax@opensource.wolfsonmicro.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Since sometimes the 'config' parameter has no use, it should be NULL.
And make the code simplifier.
Signed-off-by: NXiubo Li <Li.Xiubo@freescale.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

Create special function regmap_attach_dev
which can be called separately out of regmap_init.
Signed-off-by: NMichal Simek <michal.simek@xilinx.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

SPMI states that a slave may contain two register spaces, the Base
register space is a 5-bit byte-addressable space accessed via the
Register Read/Write and Register Zero Write command sequences, and the
Extended register space: a 16-bit byte-addressable space accessed via
the Extended Read/Write and Extended Read/Write Long command sequences.

Provide support for accessing both of these spaces, taking advantage of
the more bandwidth-efficient commands ('Register 0 Write' vs 'Register
Write', and 'Extended Register Read/Write' vs 'Extended Register
Read/Write Long') when possible.
Signed-off-by: NJosh Cartwright <joshc@codeaurora.org>
Acked-by: NMark Brown <broonie@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

None of these files are actually using any __init type directives
and hence don't need to include <linux/init.h>.  Most are just a
left over from __devinit and __cpuinit removal, or simply due to
code getting copied from one driver to the next.

Cc: Len Brown <len.brown@intel.com>
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>
Acked-by: NPavel Machek <pavel@ucw.cz>
Acked-by: NRafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: NMark Brown <broonie@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Check if regs are readable.
Signed-off-by: NMichal Simek <michal.simek@xilinx.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

irqdomain now supports removal of domains on exit so we can properly clean
up on deletion of a regmap irqchip.
Signed-off-by: NMark Brown <broonie@linaro.org>

When a map covers a single register, max_register is equal
to 0, so the "registers" & "access" files were not created.
Now they will be, as register 0 must be readable for such
map to make sense.
Signed-off-by: NPawel Moll <pawel.moll@arm.com>
Signed-off-by: NMark Brown <broonie@linaro.org>

regcache_sync_block_raw_flush takes the address of the base register
and the address of one past the last register to write to.  "count" is
the number of registers in the range, not the number of bytes, it
should be (end addr - start addr) / stride. Without accounting for
strides greater than one, registers past the end might be synced or
the writeable_reg callback at the beginning of _regmap_raw_write will
fail and nothing will be written.
Signed-off-by: NDylan Reid <dgreid@chromium.org>
Signed-off-by: NMark Brown <broonie@linaro.org>

regmap_bulk_write() should decay to performing individual writes
if we're using a "no-bus" regmap. Unfortunately, it returns an
error because there is no map->bus pointer. Fix it.
Signed-off-by: NStephen Boyd <sboyd@codeaurora.org>
Signed-off-by: NMark Brown <broonie@linaro.org>

regmap_bulk_read() should decay to performing individual reads if
we're using a "no-bus" regmap. Unfortunately, it returns an
error because there is no map->bus pointer. Fix it.
Signed-off-by: NStephen Boyd <sboyd@codeaurora.org>
Signed-off-by: NMark Brown <broonie@linaro.org>

In some cases, clear interrupt register may be at address 0.
This patch allows to use such configurations by adding additional
configuration bit to indicate this.

[With doc fix from Levente Kurusa <levex@linux.com> -- broonie]
Signed-off-by: NAlexander Shiyan <shc_work@mail.ru>
Signed-off-by: NMark Brown <broonie@linaro.org>

clk_get() returns an error pointer, or a valid token to pass back to the
clock API. Hence, the result must be checked with IS_ERR(), not by
comparison against NULL.
Signed-off-by: NStephen Warren <swarren@nvidia.com>
Signed-off-by: NMark Brown <broonie@linaro.org>