Commit 5e3ca2b349 ("regulator: Try to resolve regulators supplies on
registration") added a call to regulator_resolve_supply() within
regulator_register() where the regulator_list_mutex is held. This causes
a deadlock to occur on the Tegra114 Dalmore board when the palmas PMIC
is registered because regulator_register_resolve_supply() calls
regulator_dev_lookup() which may try to acquire the regulator_list_mutex
again.
Fix this by releasing the mutex before calling
regulator_register_resolve_supply() and update the error exit path to
ensure the mutex is released on an error.
[Made commit message more legible -- broonie]
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Commit 6261b06de5 ("regulator: Defer lookup of supply to regulator_get")
moved the regulator supplies lookup logic from the regulators registration
to the regulators get time.
Unfortunately, that changed the behavior of the regulator core since now a
parent supply with a child regulator marked as always-on, won't be enabled
unless a client driver attempts to get the child regulator during boot.
This patch tries to resolve the parent supply for the already registered
regulators each time that a new regulator is registered. So the regulators
that have child regulators marked as always on will be enabled regardless
if a driver gets the child regulator or not.
That was the behavior before the mentioned commit, since parent supplies
were looked up at regulator registration time instead of during child get.
Since regulator_resolve_supply() checks for rdev->supply, most of the times
it will be a no-op. Errors aren't checked to keep the possible out of order
dependencies which was the motivation for the mentioned commit.
Also, the supply being available will be enforced on regulator get anyways
in case the resolve fails on regulators registration.
Fixes: 6261b06de5 ("regulator: Defer lookup of supply to regulator_get")
Suggested-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Cc: <stable@vger.kernel.org> # 4.1+
Add support to enable/disable active discharge of regulator via
machine constraints. This configuration is done when setting
machine constraint during regulator register and if regulator
driver implemented the callback ops.
Signed-off-by: Laxman Dewangan <ldewangan@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Add support to enable/disable active discharge of regulator via
machine constraints. This configuration is done when setting
machine constraint during regulator register and if regulator
driver implemented the callback ops.
Signed-off-by: Laxman Dewangan <ldewangan@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
This problem was introduced by:
commit daad134d66 ("regulator: core: Request GPIO before creating
sysfs entries")
The error path was not updated correctly after moving GPIO registration
code and in case regulator_ena_gpio_free failed, device_unregister() was
called even though device_register() was not yet called.
This problem breaks the boot at least on all Tegra 32-bit devices. It
will also crash each device that specifices GPIO that is unavaiable at
regulator_register call. Here's error log I've got when forced GPIO to
be invalid:
[ 1.116612] usb-otg-vbus-reg: Failed to request enable GPIO10: -22
[ 1.122794] Unable to handle kernel NULL pointer dereference at
virtual address 00000044
[ 1.130894] pgd = c0004000
[ 1.133598] [00000044] *pgd=00000000
[ 1.137205] Internal error: Oops: 5 [#1] SMP ARM
and here's backtrace from KDB:
Exception stack(0xef11fbd0 to 0xef11fc18)
fbc0: 00000000 c0738a14 00000000 00000000
fbe0: c0b2a0b0 00000000 00000000 c0738a14 c0b5fdf8 00000001 ef7f6074 ef11fc4c
fc00: ef11fc50 ef11fc20 c02a8344 c02a7f1c 60000013 ffffffff
[<c010cee0>] (__dabt_svc) from [<c02a7f1c>] (kernfs_find_ns+0x18/0xf8)
[<c02a7f1c>] (kernfs_find_ns) from [<c02a8344>] (kernfs_find_and_get_ns+0x40/0x58)
[<c02a8344>] (kernfs_find_and_get_ns) from [<c02ac4a4>] (sysfs_unmerge_group+0x28/0x68)
[<c02ac4a4>] (sysfs_unmerge_group) from [<c044389c>] (dpm_sysfs_remove+0x30/0x5c)
[<c044389c>] (dpm_sysfs_remove) from [<c0436ba8>] (device_del+0x48/0x1f4)
[<c0436ba8>] (device_del) from [<c0436d84>] (device_unregister+0x30/0x6c)
[<c0436d84>] (device_unregister) from [<c0403910>] (regulator_register+0x6d0/0xdac)
[<c0403910>] (regulator_register) from [<c04052d4>] (devm_regulator_register+0x50/0x84)
[<c04052d4>] (devm_regulator_register) from [<c0406298>] (reg_fixed_voltage_probe+0x25c/0x3c0)
[<c0406298>] (reg_fixed_voltage_probe) from [<c043d21c>] (platform_drv_probe+0x60/0xb0)
[<c043d21c>] (platform_drv_probe) from [<c043b078>] (driver_probe_device+0x24c/0x440)
[<c043b078>] (driver_probe_device) from [<c043b5e8>] (__device_attach_driver+0xc0/0x120)
[<c043b5e8>] (__device_attach_driver) from [<c043901c>] (bus_for_each_drv+0x6c/0x98)
[<c043901c>] (bus_for_each_drv) from [<c043ad20>] (__device_attach+0xac/0x138)
[<c043ad20>] (__device_attach) from [<c043b664>] (device_initial_probe+0x1c/0x20)
[<c043b664>] (device_initial_probe) from [<c043a074>] (bus_probe_device+0x94/0x9c)
[<c043a074>] (bus_probe_device) from [<c043a610>] (deferred_probe_work_func+0x80/0xcc)
[<c043a610>] (deferred_probe_work_func) from [<c01381d0>] (process_one_work+0x158/0x454)
[<c01381d0>] (process_one_work) from [<c013854c>] (worker_thread+0x38/0x510)
[<c013854c>] (worker_thread) from [<c013e154>] (kthread+0xe8/0x104)
[<c013e154>] (kthread) from [<c0108638>] (ret_from_fork+0x14/0x3c)
Signed-off-by: Krzysztof Adamski <krzysztof.adamski@tieto.com>
Reported-by: Jon Hunter <jonathanh@nvidia.com>
Acked-by: Jon Hunter <jonathanh@nvidia.com>
Tested-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
regulator_attr_is_visible (which is a .is_visible callback of
regulator_dev_group attribute_grpup) checks rdev->ena_pin to decide if
"status" file should be present in sysfs. This field is set at the end
of regulator_ena_gpio_request so it has to be called before
device_register() otherwise this test will always fail, causing "status"
file to not be visible.
Since regulator_attr_is_visible also tests for is_enabled() op, this
problem is only visible for regulators that does not define this
callback, like regulator-fixed.c.
Signed-off-by: Krzysztof Adamski <krzysztof.adamski@tieto.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
As we now free the constraints in regulator_dev_release we will still
call free on the constraints pointer even if we went down an error
path in regulator_register, because it is only allocated after the
device_register. As such we no longer need to free rdev->constraints
on the error paths, so this patch removes said frees.
Fixes: 29f5f4860a ("regulator: core: Move more deallocation into class unregister")
Signed-off-by: Charles Keepax <ckeepax@opensource.wolfsonmicro.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Originally queue_delayed_work() used to negative error codes or 0 and 1
on success depending if the work was queued or not. It caused a lot of
bugs where people treated all non-zero returns as failures so we changed
it to return bool instead in d4283e9378 ('workqueue: make queueing
functions return bool'). Now it never returns failure.
Checking for negative values causes a static checker warning since it is
impossible based on the bool type.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Commit fa731ac7ea ("regulator: core: avoid unused variable warning")
introduced a subtle change in how supplies are locked. Where previously
code was always locking the regulator of the current iteration, the new
implementation only locks the regulator if it has a supply. For any
given power tree that means that the root will never get locked.
On the other hand the regulator_unlock_supply() will still release all
the locks, which in turn causes the lock debugging code to warn about a
mutex being unlocked which wasn't locked.
Cc: Mark Brown <broonie@kernel.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Fixes: Fixes: fa731ac7ea ("regulator: core: avoid unused variable warning")
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
The latest workaround for the lockdep interface's not using the second
argument of mutex_lock_nested() changed the loop missed locking the last
regulator due to a thinko with the loop termination condition exiting
one regulator too soon.
Reported-by: Tyler Baker <tyler.baker@linaro.org>
Signed-off-by: Mark Brown <broonie@kernel.org>
As noticed by Geert Uytterhoeven, my patch to avoid a harmless build warning
in regulator_lock_supply() was total crap and introduced a real bug:
> [ BUG: bad unlock balance detected! ]
> kworker/u4:0/6 is trying to release lock (&rdev->mutex) at:
> [<c0247b84>] regulator_set_voltage+0x38/0x50
we still lock the regulator supplies, but not the actual regulators,
so we are missing a lock, and the unlock is unbalanced.
This rectifies it by first locking the regulator device itself before
using the same loop as before to lock its supplies.
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 716fec9d1965 ("[SUBMITTED] regulator: core: avoid unused variable warning")
Signed-off-by: Mark Brown <broonie@kernel.org>
The second argument of the mutex_lock_nested() helper is only
evaluated if CONFIG_DEBUG_LOCK_ALLOC is set. Otherwise we
get this build warning for the new regulator_lock_supply
function:
drivers/regulator/core.c: In function 'regulator_lock_supply':
drivers/regulator/core.c:142:6: warning: unused variable 'i' [-Wunused-variable]
To avoid the warning, this restructures the code to make it
both simpler and to move the 'i++' outside of the mutex_lock_nested
call, where it is now always used and the variable is not
flagged as unused.
We had some discussion about changing mutex_lock_nested to an
inline function, which would make the code do the right thing here,
but in the end decided against it, in order to guarantee that
mutex_lock_nested() does not introduced overhead without
CONFIG_DEBUG_LOCK_ALLOC.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 9f01cd4a91 ("regulator: core: introduce function to lock regulators and its supplies")
Link: http://permalink.gmane.org/gmane.linux.kernel/2068900
Signed-off-by: Mark Brown <broonie@kernel.org>
Make it possible to use the bulk API with optional supplies, by allowing
the consumer to marking supplies as optional in the regulator_bulk_data.
Signed-off-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Since we need to read voltages of parents as part of setting supply
voltages we need to be able to do get_voltage() internally without
taking locks so reorganize the locking to take locks on the full tree on
entry rather than as we recurse when called externally.
Reported-by: John Stultz <john.stultz@linaro.org>
Tested-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Mark Brown <broonie@kernel.org>
Until now changing the voltage of a regulator only ever effected the
regulator itself, but never its supplies. It's a common pattern though
to put LDO regulators behind switching regulators. The switching
regulators efficiently drop the input voltage but have a high ripple on
their output. The output is then cleaned up by the LDOs. For higher
energy efficiency the voltage drop at the LDOs should be minimized. For
this scenario we need to propagate the voltage change to the supply
regulators. Another scenario where voltage propagation is desired is
a regulator which only consists of a switch and thus cannot regulate
voltages itself. In this case we can pass setting voltages to the
supply.
This patch adds support for voltage propagation. We do voltage
propagation when the current regulator has a minimum dropout voltage
specified or if the current regulator lacks a get_voltage operation
(indicating it's a switch and not a regulator).
Changing the supply voltage must be done carefully. When we are
increasing the current regulators output we must first increase the
supply voltage and then the regulator itself. When we are decreasing the
current regulators voltage we must decrease the supply voltage after
changing the current regulators voltage.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
_regulator_call_set_voltage has code to translate a minimum/maximum
voltage pair into a selector. This code is useful for others aswell,
so create a regulator_map_voltage function.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
The unlocked version will be needed when we start propagating voltage
changes to the supply regulators.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
The unlocked version will be needed when we start propagating voltage
changes to the supply regulators.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
Each regulator_dev is locked with its own mutex. This is fine as long
as only one regulator_dev is locked, but makes lockdep unhappy when we
have to walk up the supply chain like it can happen in
regulator_get_voltage:
regulator_get_voltage ->
mutex_lock(®ulator->rdev->mutex) ->
_regulator_get_voltage(regulator->rdev) ->
regulator_get_voltage(rdev->supply) ->
mutex_lock(®ulator->rdev->mutex);
This causes lockdep to issue a possible deadlock warning.
There are at least two ways to work around this:
- We can always lock the whole supply chain using the functions
introduced with this patch.
- We could store the current voltage in struct regulator_rdev so
that we do not have to walk up the supply chain for the
_regulator_get_voltage case.
Anyway, regulator_lock_supply/regulator_unlock_supply will be needed
once we allow regulator_set_voltage to optimize the supply voltages.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
When resolving regulator-regulator supplies we ignore probe deferral
returns from regulator_dev_lookup() (such as are generated for DT when
we can see a supply is registered) and just fall back to the dummy
regulator if there are full constraints (as is the case for DT). This
means that probe deferral is broken for DT systems, fix that by paying
attention to -EPROBE_DEFER return codes like we do -ENODEV.
A further patch will simplify this further, this is a minimal fix for
the specific issue.
Fixes: 9f7e25edb1 (regulator: core: Handle full constraints systems when resolving supplies)
Reported-by: Sascha Hauer <s.hauer@pengutronix.de>
Tested-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonnie@kernel.org>
Cc: stable@vger.kernel.org
As we are already registering a device with regulator_class for each
regulator device, regulator_list is redundant and can be replaced with
calls to class_find_device() and class_for_each_device().
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
The ret pointer passed to regulator_dev_lookup is only filled with a
valid error code if regulator_dev_lookup returned NULL. Currently
regulator_resolve_supply checks this ret value before it checks if a
regulator was returned, this can result in valid regulator lookups being
ignored.
Fixes: 6261b06de5 ("regulator: Defer lookup of supply to regulator_get")
Signed-off-by: Charles Keepax <ckeepax@opensource.wolfsonmicro.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Cc: stable@vger.kernel.org
The same error print exists 4 times in the regulator core
<rdev>: operation not allowed
Unfortunately, seeing this in the dmesg is not very informative.
Add what type of operation is not allowed to the message so that
these errors are unique, hopefully pointing developers in the
right direction
<rdev>: drms operation not allowed
<rdev>: voltage operation not allowed
<rdev>: current operation not allowed
<rdev>: mode operation not allowed
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Mark Brown <broonie@kernel.org>
This has been a busy release for regmap. By far the biggest set of
changes here are those from Markus Pargmann which implement support for
block transfers in smbus devices. This required quite a bit of
refactoring but leaves us better able to handle odd restrictions that
controllers may have and with better performance on smbus.
Other new features include:
- Fix interactions with lockdep for nested regmaps (eg, when a device
using regmap is connected to a bus where the bus controller has a
separate regmap). Lockdep's default class identification is too
crude to work without help.
- Support for must write bitfield operations, useful for operations
which require writing a bit to trigger them from Kuniori Morimoto.
- Support for delaying during register patch application from Nariman
Poushin.
- Support for overriding cache state via the debugfs implementation
from Richard Fitzgerald.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAABAgAGBQJV6cZyAAoJECTWi3JdVIfQM3sH/RSygzRIOoOuvro0U3qd4+nM
qLzpuZNtuAP7xNc5yJZiixz1S6PqUNl+pK/u58s6x10GWDGsWZY6E0Lg94lYmfhA
26jqWSzrMHp42x+ZN7btLExzPTVnvRerrjtj4mqz6t4yun9SzqVymaZTUZ1hqPHB
qxSNHs3rHPiqiEWpQKJXb+5dazYYJCbOUrlivxJCk60+ns1N88cA71aY+5/zq5uy
36e0iYe3s+lv0lptedarFCf9p7o1E6EQzhvEMfyquGXtvq8fl7Qdeo7riRFQ8Iiy
sygCb3AYuZIiqnOC7+90xkr2Oq0iwdJUD91A9sl/SRiwgItG9jW03bWNHYIhQyk=
=CbGt
-----END PGP SIGNATURE-----
Merge tag 'regmap-v4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap
Pull regmap updates from Mark Brown:
"This has been a busy release for regmap.
By far the biggest set of changes here are those from Markus Pargmann
which implement support for block transfers in smbus devices. This
required quite a bit of refactoring but leaves us better able to
handle odd restrictions that controllers may have and with better
performance on smbus.
Other new features include:
- Fix interactions with lockdep for nested regmaps (eg, when a device
using regmap is connected to a bus where the bus controller has a
separate regmap). Lockdep's default class identification is too
crude to work without help.
- Support for must write bitfield operations, useful for operations
which require writing a bit to trigger them from Kuniori Morimoto.
- Support for delaying during register patch application from Nariman
Poushin.
- Support for overriding cache state via the debugfs implementation
from Richard Fitzgerald"
* tag 'regmap-v4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap: (25 commits)
regmap: fix a NULL pointer dereference in __regmap_init
regmap: Support bulk reads for devices without raw formatting
regmap-i2c: Add smbus i2c block support
regmap: Add raw_write/read checks for max_raw_write/read sizes
regmap: regmap max_raw_read/write getter functions
regmap: Introduce max_raw_read/write for regmap_bulk_read/write
regmap: Add missing comments about struct regmap_bus
regmap: No multi_write support if bus->write does not exist
regmap: Split use_single_rw internally into use_single_read/write
regmap: Fix regmap_bulk_write for bus writes
regmap: regmap_raw_read return error on !bus->read
regulator: core: Print at debug level on debugfs creation failure
regmap: Fix regmap_can_raw_write check
regmap: fix typos in regmap.c
regmap: Fix integertypes for register address and value
regmap: Move documentation to regmap.h
regmap: Use different lockdep class for each regmap init call
thermal: sti: Add parentheses around bridge->ops->regmap_init call
mfd: vexpress: Add parentheses around bridge->ops->regmap_init call
regmap: debugfs: Fix misuse of IS_ENABLED
...
We were checking rdev->supply for NULL after dereferencing it. Lets
check for rdev->supply along with _regulator_is_enabled() and call
regulator_enable() only if rdev->supply is not NULL.
Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Signed-off-by: Mark Brown <broonie@kernel.org>
When calling regulator_set_load, regulator_check_drms prints and returns
an error if the regulator device's flag REGULATOR_CHANGE_DRMS isn't set.
drms_uA_update, however, bails out without reporting an error.
Replace the error print with a debug level print so that we don't get
such prints when the underlying regulator doesn't support DRMS.
Signed-off-by: Archit Taneja <architt@codeaurora.org>
Signed-off-by: Mark Brown <broonie@kernel.org>
Failure to create a debugfs node is not an error, but we print a
warning upon failure to create the node. Downgrade this to a
debug printk so that we're quiet here. This allows multiple
drivers to request a CPU's regulator so that CPUfreq and AVSish
drivers can coexist.
The downside of this approach is that whoever gets to debugfs first
the others who come later to not have any debugfs attributes associated
with them.
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Mark Brown <broonie@kernel.org>
The regulator_list has exactly the same contents as the list that the
driver core maintains of regulator_class members so is redundant. As a
first step in converting over to use the class device list convert our
iteration in late_initcall() to use the class device iterator.
Signed-off-by: Mark Brown <broonie@kernel.org>
We really ought to be using the class dvice lifetime management features
more than we are rather than open coding them so take a step towards that
by moving some of the simplest deallocations to the dev_release() function.
Signed-off-by: Mark Brown <broonie@kernel.org>
When we release a regulator we need to remove references to it from the
rdev which means locking the rdev. Currently we also free resources
associated with the regulator inside the rdev lock but there is no need
to do this, we can reduce the region the lock is held by restricting it
to just actions that affect the rdev.
Signed-off-by: Mark Brown <broonie@kernel.org>
When removing a regulator we hold regulator_list_mutex in order to
ensure the regualtor doesn't become removed again. However we only need
to protect the list until we remove the regulator from the list so move
the unlock earlier to reduce the locked region.
Signed-off-by: Mark Brown <broonie@kernel.org>
Some regulators can automatically shut down when they detect an
over current event. Add an op (set_over_current_protection) and a
DT property + constraint to support this capability.
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Mark Brown <broonie@kernel.org>