regmap: Reorder 'struct regmap'

On a x86_64, with configured with allmodconfig, pahole states that the
regmap structure is:

/* size: 1048, cachelines: 17, members: 78 */
/* sum members: 1006, hol

regmap: Reorder 'struct regmap'

On a x86_64, with configured with allmodconfig, pahole states that the
regmap structure is:

/* size: 1048, cachelines: 17, members: 78 */
/* sum members: 1006, holes: 9, sum holes: 35 */
/* padding: 7 */
/* member types with holes: 2, total: 2 */
/* last cacheline: 24 bytes */

So, when such a struct is allocated, 2048 bytes are allocated, with most of
this space being wasted.

Move a few bools so that the size is reduced to 1024.

After this change, pahole gives:

/* size: 1024, cachelines: 16, members: 78 */
/* sum members: 1006, holes: 6, sum holes: 18 */
/* member types with holes: 2, total: 2 */

Signed-off-by: Christophe JAILLET <[email protected]>
Link: https://patch.msgid.link/f01f900d15633d5cda5f27763723acb307c0d22f.1737725820.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: maple: Provide lockdep (sub)class for maple tree's internal lock

In some cases when using the maple tree register cache, the lockdep
validator might complain about invalid deadlocks:

[7.131

regmap: maple: Provide lockdep (sub)class for maple tree's internal lock

In some cases when using the maple tree register cache, the lockdep
validator might complain about invalid deadlocks:

[7.131886] Possible interrupt unsafe locking scenario:

[7.131890] CPU0 CPU1
[7.131893] ---- ----
[7.131896] lock(&mt->ma_lock);
[7.131904] local_irq_disable();
[7.131907] lock(rockchip_drm_vop2:3114:(&vop2_regmap_config)->lock);
[7.131916] lock(&mt->ma_lock);
[7.131925] <Interrupt>
[7.131928] lock(rockchip_drm_vop2:3114:(&vop2_regmap_config)->lock);
[7.131936]
*** DEADLOCK ***

[7.131939] no locks held by swapper/0/0.
[7.131944]
the shortest dependencies between 2nd lock and 1st lock:
[7.131950] -> (&mt->ma_lock){+.+.}-{2:2} {
[7.131966] HARDIRQ-ON-W at:
[7.131973] lock_acquire+0x200/0x330
[7.131986] _raw_spin_lock+0x50/0x70
[7.131998] regcache_maple_write+0x68/0xe0
[7.132010] regcache_write+0x6c/0x90
[7.132019] _regmap_read+0x19c/0x1d0
[7.132029] _regmap_update_bits+0xc0/0x148
[7.132038] regmap_update_bits_base+0x6c/0xa8
[7.132048] rk8xx_probe+0x22c/0x3d8
[7.132057] rk8xx_spi_probe+0x74/0x88
[7.132065] spi_probe+0xa8/0xe0

[...]

[7.132675] }
[7.132678] ... key at: [<ffff800082943c20>] __key.0+0x0/0x10
[7.132691] ... acquired at:
[7.132695] _raw_spin_lock+0x50/0x70
[7.132704] regcache_maple_write+0x68/0xe0
[7.132714] regcache_write+0x6c/0x90
[7.132724] _regmap_read+0x19c/0x1d0
[7.132732] _regmap_update_bits+0xc0/0x148
[7.132741] regmap_field_update_bits_base+0x74/0xb8
[7.132751] vop2_plane_atomic_update+0x480/0x14d8 [rockchipdrm]
[7.132820] drm_atomic_helper_commit_planes+0x1a0/0x320 [drm_kms_helper]

[...]

[7.135112] -> (rockchip_drm_vop2:3114:(&vop2_regmap_config)->lock){-...}-{2:2} {
[7.135130] IN-HARDIRQ-W at:
[7.135136] lock_acquire+0x200/0x330
[7.135147] _raw_spin_lock_irqsave+0x6c/0x98
[7.135157] regmap_lock_spinlock+0x20/0x40
[7.135166] regmap_read+0x44/0x90
[7.135175] vop2_isr+0x90/0x290 [rockchipdrm]
[7.135225] __handle_irq_event_percpu+0x124/0x2d0

In the example above, the validator seems to get the scope of
dependencies wrong, since the regmap instance used in rk8xx-spi driver
has nothing to do with the instance from vop2.

Improve validation by sharing the regmap's lockdep class with the maple
tree's internal lock, while also providing a subclass for the latter.

Signed-off-by: Cristian Ciocaltea <[email protected]>
Link: https://patch.msgid.link/20241031-regmap-maple-lockdep-fix-v2-1-06a3710f3623@collabora.com
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: kunit: Create a struct device for the regmap

Use kunit_device_register() to create a real struct device for the
regmap instead of leaving it at NULL.

The main reason for this is that it all

regmap: kunit: Create a struct device for the regmap

Use kunit_device_register() to create a real struct device for the
regmap instead of leaving it at NULL.

The main reason for this is that it allows context data to be passed
into the readable_reg/writable_reg/volatile_reg functions by attaching
it to the struct device with dev_set_drvdata().

The gen_regmap() and gen_raw_regmap() functions are updated to take a
struct kunit * argument.

A new struct regmap_test_priv has been created to hold the struct device
created by kunit_device_register(). This allows the struct to be
extended in the future to hold more private data for the test suite.

Signed-off-by: Richard Fitzgerald <[email protected]>
Link: https://msgid.link/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: rework ->max_register handling

When regmap consists of single register, 'regmap' subsystem is unable to
understand whether ->max_register is set or not, because in both cases it
is equal to

regmap: rework ->max_register handling

When regmap consists of single register, 'regmap' subsystem is unable to
understand whether ->max_register is set or not, because in both cases it
is equal to zero. It leads to that the logic based on value of
->max_register doesn't work. For example using of REGCACHE_FLAT fails.

This patch introduces an extra parameter to regmap config, indicating
that zero value in ->max_register is authentic.

Signed-off-by: Jan Dakinevich <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: ram: support noinc semantics

Support noinc semantics in RAM backed regmaps, for testing purposes. Add
a new callback that selects registers which should have noinc behavior.
Bulk writes to a

regmap: ram: support noinc semantics

Support noinc semantics in RAM backed regmaps, for testing purposes. Add
a new callback that selects registers which should have noinc behavior.
Bulk writes to a noinc register will cause the last value in the buffer
to be assigned to the register, while bulk reads will copy the same
value repeatedly into the buffer.

This patch only adds support to regmap-raw-ram, since regmap-ram does
not support bulk operations.

Signed-off-by: Ben Wolsieffer <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: Add debugfs file for forcing field writes

`_regmap_update_bits()` checks if the current register value differs
from the new value, and only writes to the register if they differ. When
testin

regmap: Add debugfs file for forcing field writes

`_regmap_update_bits()` checks if the current register value differs
from the new value, and only writes to the register if they differ. When
testing hardware drivers, it might be desirable to always force a
register write, for example when writing to a `regmap_field`. This
enables and simplifies testing and verification of the hardware
interaction. For example, when using a hardware mock/simulation model,
one can then more easily verify that the driver makes the correct
expected register writes during certain events.

Add a bool variable `force_write_field` and a corresponding debugfs
entry to enable this. Since this feature could interfere with driver
operation, guard it with a macro.

Signed-off-by: Waqar Hameed <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: Don't check for changes in regcache_set_val()

The only user of regcache_set_val() ignores the return value so we may as
well not bother checking if the value we are trying to set is the same

regmap: Don't check for changes in regcache_set_val()

The only user of regcache_set_val() ignores the return value so we may as
well not bother checking if the value we are trying to set is the same as
the value already stored.

Signed-off-by: Mark Brown <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: maple: Implement block sync for the maple tree cache

For register maps where we can write multiple values in a single bus
operation it is generally much faster to do so. Improve the performa

regmap: maple: Implement block sync for the maple tree cache

For register maps where we can write multiple values in a single bus
operation it is generally much faster to do so. Improve the performance of
maple tree cache syncs on such devices by identifying blocks of adjacent
registers that need to be written out and combining them into a single
operation.

Combining writes does mean that we need to allocate a scratch buffer and
format the data into it but it is expected that for most cases where caches
are in use the cost of I/O will be much greater than the cost of doing the
allocation and format.

Signed-off-by: Mark Brown <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: Provide a ram backed regmap with raw support

Provide a simple, 16 bit only, RAM backed regmap which supports raw I/O for
use in testing.

Signed-off-by: Mark Brown <[email protected]>
Link:

regmap: Provide a ram backed regmap with raw support

Provide a simple, 16 bit only, RAM backed regmap which supports raw I/O for
use in testing.

Signed-off-by: Mark Brown <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: allow upshifting register addresses before performing operations

Similar to the existing reg_downshift mechanism, that is used to
translate register addresses on busses that have a smaller a

regmap: allow upshifting register addresses before performing operations

Similar to the existing reg_downshift mechanism, that is used to
translate register addresses on busses that have a smaller address
stride, it's also possible to want to upshift register addresses.

Such a case was encountered when network PHYs and PCS that usually sit
on a MDIO bus (16-bits register with a stride of 1) are integrated
directly as memory-mapped devices. Here, the same register layout
defined in 802.3 is used, but the register now have a larger stride.

Introduce a mechanism to also allow upshifting register addresses.
Re-purpose reg_downshift into a more generic, signed reg_shift, whose
sign indicates the direction of the shift. To avoid confusion, also
introduce macros to explicitly indicate if we want to downshift or
upshift.

For bisectability, change any use of reg_downshift to use reg_shift.

Signed-off-by: Maxime Chevallier <[email protected]>
Tested-by: Colin Foster <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: Add maple tree based register cache

The current state of the art for sparse register maps is the
rbtree cache. This works well for most applications but isn't
always ideal for sparser regis

regmap: Add maple tree based register cache

The current state of the art for sparse register maps is the
rbtree cache. This works well for most applications but isn't
always ideal for sparser register maps since the rbtree can get
deep, requiring a lot of walking. Fortunately the kernel has a
data structure intended to address this very problem, the maple
tree. Provide an initial implementation of a register cache
based on the maple tree to start taking advantage of it.

The entries stored in the maple tree are arrays of register
values, with the maple tree keys holding the register addresses.
We store data in host native format rather than device native
format as we do for rbtree, this will be a benefit for devices
where we don't marshal data within regmap and simplifies the code
but will result in additional CPU overhead when syncing the cache
on devices where we do marshal data in regmap.

This should work well for a lot of devices, though there's some
additional areas that could be looked at such as caching the
last accessed entry like we do for rbtree and trying to minimise
the maple tree level locking. We should also use bulk writes
rather than single register writes when resyncing the cache where
possible, even if we don't store in device native format.

Very small register maps may continue to to better with rbtree
longer term.

Signed-off-by: Mark Brown <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: Factor out single value register syncing

In order to support sparse caches that don't store data in raw format
factor out the parts of the raw block sync implementation that deal with
writin

regmap: Factor out single value register syncing

In order to support sparse caches that don't store data in raw format
factor out the parts of the raw block sync implementation that deal with
writing a single register via _regmap_write().

Signed-off-by: Mark Brown <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: Add RAM backed register map

Add a register map that is a simple array of memory, for use in
KUnit testing of the framework. This is not exposed in regmap.h
since I can't think of a non-test

regmap: Add RAM backed register map

Add a register map that is a simple array of memory, for use in
KUnit testing of the framework. This is not exposed in regmap.h
since I can't think of a non-test use case, it is purely for use
internally. To facilitate testing we track if registers have been
read or written to.

Signed-off-by: Mark Brown <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

show more ...

regmap: Removed compressed cache support

The compressed register cache support has assumptions that make it hard to
cover in testing, mainly that it requires raw registers defaults be
provided. Rath

regmap: Removed compressed cache support

The compressed register cache support has assumptions that make it hard to
cover in testing, mainly that it requires raw registers defaults be
provided. Rather than either address these assumptions or leave it untested
by the forthcoming KUnit tests let's remove it, the use case is quite thin
and there are no current users.

Signed-off-by: Mark Brown <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

show more ...

regmap: Add bulk read/write callbacks into regmap_config

Currently the regmap_config structure only allows the user to implement
single element register read/write using .reg_read/.reg_write callbac

regmap: Add bulk read/write callbacks into regmap_config

Currently the regmap_config structure only allows the user to implement
single element register read/write using .reg_read/.reg_write callbacks.
The regmap_bus already implements bulk counterparts of both, and is being
misused as a workaround for the missing bulk read/write callbacks in
regmap_config by a couple of drivers. To stop this misuse, add the bulk
read/write callbacks to regmap_config and call them from the regmap core
code.

Signed-off-by: Marek Vasut <[email protected]>
Cc: Jagan Teki <[email protected]>
Cc: Mark Brown <[email protected]>
Cc: Maxime Ripard <[email protected]>
Cc: Robert Foss <[email protected]>
Cc: Sam Ravnborg <[email protected]>
Cc: Thomas Zimmermann <[email protected]>
To: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: allow a defined reg_base to be added to every address

There's an inconsistency that arises when a register set can be accessed
internally via MMIO, or externally via SPI. The VSC7514 chip al

regmap: allow a defined reg_base to be added to every address

There's an inconsistency that arises when a register set can be accessed
internally via MMIO, or externally via SPI. The VSC7514 chip allows both
modes of operation. When internally accessed, the system utilizes __iomem,
devm_ioremap_resource, and devm_regmap_init_mmio.

For SPI it isn't possible to utilize memory-mapped IO. To properly operate,
the resource base must be added to the register before every operation.

Signed-off-by: Colin Foster <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: add configurable downshift for addresses

Add an additional reg_downshift to be applied to register addresses before
any register accesses. An example of a device that uses this is a VSC7514

regmap: add configurable downshift for addresses

Add an additional reg_downshift to be applied to register addresses before
any register accesses. An example of a device that uses this is a VSC7514
chip, which require each register address to be downshifted by two if the
access is performed over a SPI bus.

Signed-off-by: Colin Foster <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: teach regmap to use raw spinlocks if requested in the config

Some drivers might access regmap in a context where a raw spinlock is
held. An example is drivers/irqchip/irq-ls-extirq.c, which

regmap: teach regmap to use raw spinlocks if requested in the config

Some drivers might access regmap in a context where a raw spinlock is
held. An example is drivers/irqchip/irq-ls-extirq.c, which calls
regmap_update_bits() from struct irq_chip :: irq_set_type, which is a
method called by __irq_set_trigger() under the desc->lock raw spin lock.

Since desc->lock is a raw spin lock and the regmap internal lock for
mmio is a plain spinlock (which can become sleepable on RT), this is an
invalid locking scheme and we get a splat stating that this is a
"[ BUG: Invalid wait context ]".

It seems reasonable for regmap to have an option use a raw spinlock too,
so add that in the config such that drivers can request it.

Suggested-by: Mark Brown <[email protected]>
Signed-off-by: Vladimir Oltean <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: fix page selection for noinc writes

Non-incrementing writes can fail if register + length crosses page
border. However for non-incrementing writes we should not check for page
border crossin

regmap: fix page selection for noinc writes

Non-incrementing writes can fail if register + length crosses page
border. However for non-incrementing writes we should not check for page
border crossing. Fix this by passing additional flag to _regmap_raw_write
and passing length to _regmap_select_page basing on the flag.

Signed-off-by: Dmitry Baryshkov <[email protected]>
Fixes: cdf6b11daa77 ("regmap: Add regmap_noinc_write API")
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: debugfs: Fix handling of name string for debugfs init delays

In regmap_debugfs_init the initialisation of the debugfs is delayed
if the root node isn't ready yet. Most callers of regmap_debu

regmap: debugfs: Fix handling of name string for debugfs init delays

In regmap_debugfs_init the initialisation of the debugfs is delayed
if the root node isn't ready yet. Most callers of regmap_debugfs_init
pass the name from the regmap_config, which is considered temporary
ie. may be unallocated after the regmap_init call returns. This leads
to a potential use after free, where config->name has been freed by
the time it is used in regmap_debugfs_initcall.

This situation can be seen on Zynq, where the architecture init_irq
callback registers a syscon device, using a local variable for the
regmap_config. As init_irq is very early in the platform bring up the
regmap debugfs root isn't ready yet. Although this doesn't crash it
does result in the debugfs entry not having the correct name.

Regmap already sets map->name from config->name on the regmap_init
path and the fact that a separate field is used to pass the name
to regmap_debugfs_init appears to be an artifact of the debugfs
name being added before the map name. As such this patch updates
regmap_debugfs_init to use map->name, which is already duplicated from
the config avoiding the issue.

This does however leave two lose ends, both regmap_attach_dev and
regmap_reinit_cache can be called after a regmap is registered and
would have had the effect of applying a new name to the debugfs
entries. In both of these cases it was chosen to update the map
name. In the case of regmap_attach_dev there are 3 users that
currently use this function to update the name, thus doing so avoids
changes for those users and it seems reasonable that attaching
a device would want to set the name of the map. In the case of
regmap_reinit_cache the primary use-case appears to be devices that
need some register access to identify the device (for example devices
in the same family) and then update the cache to match the exact
hardware. Whilst no users do currently update the name here, given the
use-case it seemed reasonable the name might want to be updated once
the device is better identified.

Signed-off-by: Charles Keepax <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: Add can_sleep configuration option

Regmap can't sleep if spinlock is used for the locking protection.
This patch fixes regression caused by a previous commit that switched
regmap to use fsle

regmap: Add can_sleep configuration option

Regmap can't sleep if spinlock is used for the locking protection.
This patch fixes regression caused by a previous commit that switched
regmap to use fsleep() and this broke Amlogic S922X platform.

This patch adds new configuration option for regmap users, allowing to
specify whether regmap operations can sleep and assuming that sleep is
allowed if mutex is used for the regmap locking protection.

Reported-by: Marek Szyprowski <[email protected]>
Fixes: 2b32d2f7ce0a ("regmap: Use flexible sleep")
Signed-off-by: Dmitry Osipenko <[email protected]>
Tested-by: Marek Szyprowski <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: add proper SPDX identifiers on files that did not have them.

There were a few files in the regmap code that did not have SPDX
identifiers on them, so fix that up. At the same time, remove t

regmap: add proper SPDX identifiers on files that did not have them.

There were a few files in the regmap code that did not have SPDX
identifiers on them, so fix that up. At the same time, remove the "free
form" text that specified the license of the file, as that is impossible
for any tool to properly parse.

Also, as Mark loves // comment markers, convert all of the headers to be
the same to make things look consistent :)

Signed-off-by: Greg Kroah-Hartman <[email protected]>
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: Add regmap_noinc_write API

The regmap API had a noinc_read function added for instances where devices
supported returning data from an internal FIFO in a single read.

This commit adds the n

regmap: Add regmap_noinc_write API

The regmap API had a noinc_read function added for instances where devices
supported returning data from an internal FIFO in a single read.

This commit adds the noinc_write variant to allow writing to a non
incrementing register, this is used in devices such as the sx1301 for
loading firmware.

Signed-off-by: Ben Whitten <[email protected]>
Signed-off-by: Mark Brown <[email protected]>

show more ...

regmap: fix comment for regmap.use_single_write

Signed-off-by: David Frey <[email protected]>
Signed-off-by: Mark Brown <[email protected]>

regmap: Add regmap_noinc_read API

The regmap API usually assumes that bulk read operations will read a
range of registers but some I2C/SPI devices have certain registers for
which a such a read oper

regmap: Add regmap_noinc_read API

The regmap API usually assumes that bulk read operations will read a
range of registers but some I2C/SPI devices have certain registers for
which a such a read operation will return data from an internal FIFO
instead. Add an explicit API to support bulk read without range semantics.

Some linux drivers use regmap_bulk_read or regmap_raw_read for such
registers, for example mpu6050 or bmi150 from IIO. This only happens to
work because when caching is disabled a single regmap read op will map
to a single bus read op (as desired). This breaks if caching is enabled and
reg+1 happens to be a cacheable register.

Without regmap support refactoring a driver to enable regmap caching
requires separate I2C and SPI paths. This is exactly what regmap is
supposed to help avoid.

Suggested-by: Jonathan Cameron <[email protected]>
Signed-off-by: Crestez Dan Leonard <[email protected]>
Signed-off-by: Stefan Popa <[email protected]>
Signed-off-by: Mark Brown <[email protected]>

show more ...