commit cb3543cff9 upstream.
When updating the operating mode as part of regulator enable, the caller
has already locked the regulator tree and drms_uA_update() must not try
to do the same in order not to trigger a deadlock.
The lock inversion is reported by lockdep as:
======================================================
WARNING: possible circular locking dependency detected
6.1.0-next-20221215 #142 Not tainted
------------------------------------------------------
udevd/154 is trying to acquire lock:
ffffc11f123d7e50 (regulator_list_mutex){+.+.}-{3:3}, at: regulator_lock_dependent+0x54/0x280
but task is already holding lock:
ffff80000e4c36e8 (regulator_ww_class_acquire){+.+.}-{0:0}, at: regulator_enable+0x34/0x80
which lock already depends on the new lock.
...
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(regulator_ww_class_acquire);
lock(regulator_list_mutex);
lock(regulator_ww_class_acquire);
lock(regulator_list_mutex);
*** DEADLOCK ***
just before probe of a Qualcomm UFS controller (occasionally) deadlocks
when enabling one of its regulators.
Fixes: 9243a195be ("regulator: core: Change voltage setting path")
Fixes: f8702f9e4a ("regulator: core: Use ww_mutex for regulators locking")
Cc: stable@vger.kernel.org # 5.0
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Link: https://lore.kernel.org/r/20221215104646.19818-1-johan+linaro@kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 0591b14ce0 ]
I found a use_count leakage towards supply regulator of rdev with
boot-on option.
┌───────────────────┐ ┌───────────────────┐
│ regulator_dev A │ │ regulator_dev B │
│ (boot-on) │ │ (boot-on) │
│ use_count=0 │◀──supply──│ use_count=1 │
│ │ │ │
└───────────────────┘ └───────────────────┘
In case of rdev(A) configured with `regulator-boot-on', the use_count
of supplying regulator(B) will increment inside
regulator_enable(rdev->supply).
Thus, B will acts like always-on, and further balanced
regulator_enable/disable cannot actually disable it anymore.
However, B was also configured with `regulator-boot-on', we wish it
could be disabled afterwards.
Signed-off-by: Rui Zhang <zr.zhang@vivo.com>
Link: https://lore.kernel.org/r/20221201033806.2567812-1-zr.zhang@vivo.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit f2b41b748c ]
I got the the following report:
OF: ERROR: memory leak, expected refcount 1 instead of 2,
of_node_get()/of_node_put() unbalanced - destroy cset entry:
attach overlay node /i2c/pmic@62/regulators/exten
In of_get_regulator(), the node is returned from of_parse_phandle()
with refcount incremented, after using it, of_node_put() need be called.
Fixes: 69511a452e ("regulator: map consumer regulator based on device tree")
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Link: https://lore.kernel.org/r/20221115091508.900752-1-yangyingliang@huawei.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 1f386d6894 ]
I got a UAF report as following:
==================================================================
BUG: KASAN: use-after-free in __lock_acquire+0x935/0x2060
Read of size 8 at addr ffff88810e838220 by task python3/268
Call Trace:
<TASK>
dump_stack_lvl+0x67/0x83
print_report+0x178/0x4b0
kasan_report+0x90/0x190
__lock_acquire+0x935/0x2060
lock_acquire+0x156/0x400
_raw_spin_lock+0x2a/0x40
lockref_get+0x11/0x30
simple_recursive_removal+0x41/0x440
debugfs_remove.part.12+0x32/0x50
debugfs_remove+0x29/0x30
_regulator_put.cold.54+0x3e/0x27f
regulator_put+0x1f/0x30
release_nodes+0x6a/0xa0
devres_release_all+0xf8/0x150
Allocated by task 37:
kasan_save_stack+0x1c/0x40
kasan_set_track+0x21/0x30
__kasan_slab_alloc+0x5d/0x70
slab_post_alloc_hook+0x62/0x510
kmem_cache_alloc_lru+0x222/0x5a0
__d_alloc+0x31/0x440
d_alloc+0x30/0xf0
d_alloc_parallel+0xc4/0xd20
__lookup_slow+0x15e/0x2f0
lookup_one_len+0x13a/0x150
start_creating+0xea/0x190
debugfs_create_dir+0x1e/0x210
create_regulator+0x254/0x4e0
_regulator_get+0x2a1/0x467
_devm_regulator_get+0x5a/0xb0
regulator_virtual_probe+0xb9/0x1a0
Freed by task 30:
kasan_save_stack+0x1c/0x40
kasan_set_track+0x21/0x30
kasan_save_free_info+0x2a/0x50
__kasan_slab_free+0x102/0x190
kmem_cache_free+0xf6/0x600
rcu_core+0x54c/0x12b0
__do_softirq+0xf2/0x5e3
Last potentially related work creation:
kasan_save_stack+0x1c/0x40
__kasan_record_aux_stack+0x98/0xb0
call_rcu+0x42/0x700
dentry_free+0x6c/0xd0
__dentry_kill+0x23b/0x2d0
dput.part.31+0x431/0x780
simple_recursive_removal+0xa9/0x440
debugfs_remove.part.12+0x32/0x50
debugfs_remove+0x29/0x30
regulator_unregister+0xe3/0x230
release_nodes+0x6a/0xa0
==================================================================
Here is how happened:
processor A processor B
regulator_register()
rdev_init_debugfs()
rdev->debugfs = debugfs_create_dir()
devm_regulator_get()
rdev = regulator_dev_lookup()
create_regulator(rdev)
// using rdev->debugfs as parent
debugfs_create_dir(rdev->debugfs)
mfd_remove_devices_fn()
release_nodes()
regulator_unregister()
// free rdev->debugfs
debugfs_remove_recursive(rdev->debugfs)
release_nodes()
destroy_regulator()
debugfs_remove_recursive() <- causes UAF
In devm_regulator_get(), after getting rdev, the refcount
is get, so fix this by moving debugfs_remove_recursive()
to regulator_dev_release(), then it can be proctected by
the refcount, the 'rdev->debugfs' can not be freed until
the refcount is 0.
Fixes: 5de705194e ("regulator: Add basic per consumer debugfs")
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Link: https://lore.kernel.org/r/20221116033706.3595812-1-yangyingliang@huawei.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 5f4b204b6b ]
Here is a warning report about lack of registered release()
from kobject lib:
Device '(null)' does not have a release() function, it is broken and must be fixed.
WARNING: CPU: 0 PID: 48430 at drivers/base/core.c:2332 device_release+0x104/0x120
Call Trace:
kobject_put+0xdc/0x180
put_device+0x1b/0x30
regulator_register+0x651/0x1170
devm_regulator_register+0x4f/0xb0
When regulator_register() returns fail and directly goto `clean` symbol,
rdev->dev has not registered release() function yet (which is registered
by regulator_class in the following), so rdev needs to be freed manually.
If rdev->dev.of_node is not NULL, which means the of_node has gotten by
regulator_of_get_init_data(), it needs to call of_node_put() to avoid
refcount leak.
Otherwise, only calling put_device() would lead memory leak of rdev
in further:
unreferenced object 0xffff88810d0b1000 (size 2048):
comm "107-i2c-rtq6752", pid 48430, jiffies 4342258431 (age 1341.780s)
backtrace:
kmalloc_trace+0x22/0x110
regulator_register+0x184/0x1170
devm_regulator_register+0x4f/0xb0
When regulator_register() returns fail and goto `wash` symbol,
rdev->dev has registered release() function, so directly call
put_device() to cleanup everything.
Fixes: d3c731564e ("regulator: plug of_node leak in regulator_register()'s error path")
Signed-off-by: Zeng Heng <zengheng4@huawei.com>
Link: https://lore.kernel.org/r/20221116074339.1024240-1-zengheng4@huawei.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit c32f1ebfd2 ]
If regulator_enable() fails, enable_count is incremented still.
A consumer, assuming no matching regulator_disable() is necessary on
failure, will then get this error message upon regulator_put()
since enable_count is non-zero:
[ 1.277418] WARNING: CPU: 3 PID: 1 at drivers/regulator/core.c:2304 _regulator_put.part.0+0x168/0x170
The consumer could try to fix this in their driver by cleaning up on
error from regulator_enable() (i.e. call regulator_disable()), but that
results in the following since regulator_enable() failed and didn't
increment user_count:
[ 1.258112] unbalanced disables for vreg_l17c
[ 1.262606] WARNING: CPU: 4 PID: 1 at drivers/regulator/core.c:2899 _regulator_disable+0xd4/0x190
Fix this by decrementing enable_count upon failure to enable.
With this in place, just the reason for failure to enable is printed
as expected and developers can focus on the root cause of their issue
instead of thinking their usage of the regulator consumer api is
incorrect. For example, in my case:
[ 1.240426] vreg_l17c: invalid input voltage found
Fixes: 5451781dad ("regulator: core: Only count load for enabled consumers")
Signed-off-by: Andrew Halaney <ahalaney@redhat.com>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Brian Masney <bmasney@redhat.com>
Link: https://lore.kernel.org/r/20220819194336.382740-1-ahalaney@redhat.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit c3e3ca05da ]
Since the introduction of regulator->enable_count, a driver that did
an exclusive get on an already-enabled regulator would end up with
enable_count initialized to 0 but rdev->use_count initialized to 1.
With that starting point the regulator is effectively stuck enabled,
because if the driver attempted to disable it it would fail the
enable_count underflow check in _regulator_handle_consumer_disable().
The EXCLUSIVE_GET path in _regulator_get() now initializes
enable_count along with rdev->use_count so that the regulator can be
disabled without underflowing the former.
Signed-off-by: Zev Weiss <zev@bewilderbeest.net>
Fixes: 5451781dad ("regulator: core: Only count load for enabled consumers")
Link: https://lore.kernel.org/r/20220505043152.12933-1-zev@bewilderbeest.net
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 4e2a354e37 ]
The check done by regulator_late_cleanup() to detect whether a regulator
is on was inconsistent with the check done by _regulator_is_enabled().
While _regulator_is_enabled() takes the enable GPIO into account,
regulator_late_cleanup() was not doing that.
This resulted in a false positive, e.g. when a GPIO-controlled fixed
regulator was used, which was not enabled at boot time, e.g.
reg_disp_1v2: reg_disp_1v2 {
compatible = "regulator-fixed";
regulator-name = "display_1v2";
regulator-min-microvolt = <1200000>;
regulator-max-microvolt = <1200000>;
gpio = <&tlmm 148 0>;
enable-active-high;
};
Such regulator doesn't have an is_enabled() operation. Nevertheless
it's state can be determined based on the enable GPIO. The check in
regulator_late_cleanup() wrongly assumed that the regulator is on and
tried to disable it.
Signed-off-by: Oliver Barta <oliver.barta@aptiv.com>
Link: https://lore.kernel.org/r/20220208084645.8686-1-oliver.barta@aptiv.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit dbe954d8f1 ]
Sometimes regulator_get() gets called twice for the same supply on the
same device. This may happen e.g. when a framework / library is used
which uses the regulator; and the driver itself also needs to enable
the regulator in some cases where the framework will not enable it.
Commit ff268b56ce ("regulator: core: Don't spew backtraces on
duplicate sysfs") already takes care of the backtrace which would
trigger when creating a duplicate consumer symlink under
/sys/class/regulator/regulator.%d in this scenario.
Commit c33d442328 ("debugfs: make error message a bit more verbose")
causes a new error to get logged in this scenario:
[ 26.938425] debugfs: Directory 'wm5102-codec-MICVDD' with parent 'spi-WM510204:00-MICVDD' already present!
There is no _nowarn variant of debugfs_create_dir(), but we can detect
and avoid this problem by checking the return value of the earlier
sysfs_create_link_nowarn() call.
Add a check for the earlier sysfs_create_link_nowarn() failing with
-EEXIST and skip the debugfs_create_dir() call in that case, avoiding
this error getting logged.
Fixes: c33d442328 ("debugfs: make error message a bit more verbose")
Cc: Charles Keepax <ckeepax@opensource.cirrus.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20210122183250.370571-1-hdegoede@redhat.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 14a71d509a ]
With commit eaa7995c52 (regulator: core: avoid
regulator_resolve_supply() race condition) we started holding the rdev
lock while resolving supplies, an operation that requires holding the
regulator_list_mutex. This results in lockdep warnings since in other
places we take the list mutex then the mutex on an individual rdev.
Since the goal is to make sure that we don't call set_supply() twice
rather than a concern about the cost of resolution pull the rdev lock
and check for duplicate resolution down to immediately before we do the
set_supply() and drop it again once the allocation is done.
Fixes: eaa7995c52 (regulator: core: avoid regulator_resolve_supply() race condition)
Reported-by: Marek Szyprowski <m.szyprowski@samsung.com>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20210122132042.10306-1-broonie@kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit eaa7995c52 ]
The final step in regulator_register() is to call
regulator_resolve_supply() for each registered regulator
(including the one in the process of being registered). The
regulator_resolve_supply() function first checks if rdev->supply
is NULL, then it performs various steps to try to find the supply.
If successful, rdev->supply is set inside of set_supply().
This procedure can encounter a race condition if two concurrent
tasks call regulator_register() near to each other on separate CPUs
and one of the regulators has rdev->supply_name specified. There
is currently nothing guaranteeing atomicity between the rdev->supply
check and set steps. Thus, both tasks can observe rdev->supply==NULL
in their regulator_resolve_supply() calls. This then results in
both creating a struct regulator for the supply. One ends up
actually stored in rdev->supply and the other is lost (though still
present in the supply's consumer_list).
Here is a kernel log snippet showing the issue:
[ 12.421768] gpu_cc_gx_gdsc: supplied by pm8350_s5_level
[ 12.425854] gpu_cc_gx_gdsc: supplied by pm8350_s5_level
[ 12.429064] debugfs: Directory 'regulator.4-SUPPLY' with parent
'17a00000.rsc:rpmh-regulator-gfxlvl-pm8350_s5_level'
already present!
Avoid this race condition by holding the rdev->mutex lock inside
of regulator_resolve_supply() while checking and setting
rdev->supply.
Signed-off-by: David Collins <collinsd@codeaurora.org>
Link: https://lore.kernel.org/r/1610068562-4410-1-git-send-email-collinsd@codeaurora.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
If debugging is disabled, print_constraints() does not print the actual
constraints, but still performs some processing and string formatting,
only to throw away the result later.
Fix this by moving all constraint debug processing to a separate
function, and replacing it by a dummy when debugging is disabled.
This reduces kernel size by almost 800 bytes (on arm/arm64).
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Link: https://lore.kernel.org/r/20201005131546.22448-1-geert+renesas@glider.be
Signed-off-by: Mark Brown <broonie@kernel.org>
Some regulator drivers may be defining very long names: this is the
case with the qcom_smd and qcom_spmi regulators, where we need to
parse the regulator parents from DT.
For clarity, this is an example:
{ "l13a", QCOM_SMD_RPM_LDOA, 13, &pm660_ht_lvpldo,
"vdd_l8_l9_l10_l11_l12_l13_l14" },
pm660-regulators {
...
vdd_l8_l9_l10_l11_l12_l13_l14-supply = <&vreg_s4a_2p04>
...
};
Now, with a 32 characters limit, the function is trying to parse,
exactly, "vdd_l8_l9_l10_l11_l12_l13_l14-s" (32 chars) instead of
the right one, which is 37 chars long in this specific case.
... And this is not only the case with PM660/PM660L, but also with
PMA8084, PM8916, PM8950 and others that are not implemented yet.
The length of 64 chars was chosen based on the longest parsed property
name that I could find, which is in PM8916, and would be 53 characters
long.
At that point, rounding that to 64 looked like being the best idea.
Signed-off-by: AngeloGioacchino Del Regno <kholk11@gmail.com>
Link: https://lore.kernel.org/r/20200926125549.13191-2-kholk11@gmail.com
Signed-off-by: Mark Brown <broonie@kernel.org>
When creating a new regulator its supply cannot create the sysfs link
because the device is not yet published. Remove early supply resolving
since it will be done later anyway. This makes the following error
disappear and the symlinks get created instead.
DCDC_REG1: supplied by VSYS
VSYS: could not add device link regulator.3 err -2
Note: It doesn't fix the problem for bypassed regulators, though.
Fixes: 45389c4752 ("regulator: core: Add early supply resolution for regulators")
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Link: https://lore.kernel.org/r/ba09e0a8617ffeeb25cb4affffe6f3149319cef8.1601155770.git.mirq-linux@rere.qmqm.pl
Signed-off-by: Mark Brown <broonie@kernel.org>
Initial support for ROHM BD9576MUF and BD9573MUF PMICs.
These PMICs are primarily intended to be used to power the R-Car family
processors. BD9576MUF includes some additional safety features the
BD9573MUF does not have. This initial version of drivers does not
utilize these features and for now the SW behaviour is identical.
Please note that this version of drivers is only tested on BD9576MUF
but according to the data-sheets the relevant parts of registers should
be same so drivers should also work on BD9573MUF.
This patch series includes MFD, watchdog and regulator drivers with
basic functionality such as:
- Enabling and pinging the watchdog
- configuring watchog timeout / window from device-tree
- reading regulator states/voltages
- enabling/disabling VOUT1 (VD50) when control mode B is used.
This patch series does not bring interrupt support. BD9576MUF and BD9573MUF
are designed to keep the IRQ line low for whole duration of error
condition. IRQ can't be 'acked'. So proper IRQ support would require
some IRQ limiter implementation (delayed unmask?) in order to not hog
the CPU.
---
Matti Vaittinen (6):
dt_bindings: mfd: Add ROHM BD9576MUF and BD9573MUF PMICs
dt_bindings: regulator: Add ROHM BD9576MUF and BD9573MUF PMICs
mfd: Support ROHM BD9576MUF and BD9573MUF
wdt: Support wdt on ROHM BD9576MUF and BD9573MUF
regulator: Support ROHM BD9576MUF and BD9573MUF
MAINTAINERS: Add ROHM BD9576MUF and BD9573MUF drivers
.../bindings/mfd/rohm,bd9576-pmic.yaml | 129 +++++++
.../regulator/rohm,bd9576-regulator.yaml | 33 ++
MAINTAINERS | 4 +
drivers/mfd/Kconfig | 11 +
drivers/mfd/Makefile | 1 +
drivers/mfd/rohm-bd9576.c | 130 +++++++
drivers/regulator/Kconfig | 10 +
drivers/regulator/Makefile | 1 +
drivers/regulator/bd9576-regulator.c | 337 ++++++++++++++++++
drivers/watchdog/Kconfig | 13 +
drivers/watchdog/Makefile | 1 +
drivers/watchdog/bd9576_wdt.c | 295 +++++++++++++++
include/linux/mfd/rohm-bd957x.h | 61 ++++
include/linux/mfd/rohm-generic.h | 2 +
14 files changed, 1028 insertions(+)
create mode 100644 Documentation/devicetree/bindings/mfd/rohm,bd9576-pmic.yaml
create mode 100644 Documentation/devicetree/bindings/regulator/rohm,bd9576-regulator.yaml
create mode 100644 drivers/mfd/rohm-bd9576.c
create mode 100644 drivers/regulator/bd9576-regulator.c
create mode 100644 drivers/watchdog/bd9576_wdt.c
create mode 100644 include/linux/mfd/rohm-bd957x.h
base-commit: f4d51dffc6
--
2.21.0
--
Matti Vaittinen, Linux device drivers
ROHM Semiconductors, Finland SWDC
Kiviharjunlenkki 1E
90220 OULU
FINLAND
~~~ "I don't think so," said Rene Descartes. Just then he vanished ~~~
Simon says - in Latin please.
~~~ "non cogito me" dixit Rene Descarte, deinde evanescavit ~~~
Thanks to Simon Glass for the translation =]
Pull regulator fixes from Mark Brown:
"The biggest set of fixes here is those from Michał Mirosław fixing
some locking issues with coupled regulators that are triggered in
cases where a coupled regulator is used by a device involved in
fs_reclaim like eMMC storage.
These are relatively serious for the affected systems, though the
circumstances where they trigger are very rare"
* tag 'regulator-fix-v5.9-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
regulator: pwm: Fix machine constraints application
regulator: core: Fix slab-out-of-bounds in regulator_unlock_recursive()
regulator: remove superfluous lock in regulator_resolve_coupling()
regulator: cleanup regulator_ena_gpio_free()
regulator: plug of_node leak in regulator_register()'s error path
regulator: push allocation in set_consumer_device_supply() out of lock
regulator: push allocations in create_regulator() outside of lock
regulator: push allocation in regulator_ena_gpio_request() out of lock
regulator: push allocation in regulator_init_coupling() outside of lock
regulator: fix spelling mistake "Cant" -> "Can't"
regulator: cros-ec-regulator: Add NULL test for devm_kmemdup call
The recent commit 7d8196641e ("regulator: Remove pointer table
overallocation") changed the size of coupled_rdevs and now KASAN is able
to detect slab-out-of-bounds problem in regulator_unlock_recursive(),
which is a legit problem caused by a typo in the code. The recursive
unlock function uses n_coupled value of a parent regulator for unlocking
supply regulator, while supply's n_coupled should be used. In practice
problem may only affect platforms that use coupled regulators.
Cc: stable@vger.kernel.org # 5.0+
Fixes: f8702f9e4a ("regulator: core: Use ww_mutex for regulators locking")
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Link: https://lore.kernel.org/r/20200831204335.19489-1-digetx@gmail.com
Signed-off-by: Mark Brown <broonie@kernel.org>
For systems that have eg. eMMC storage using voltage regulator, memory
reclaim path might call back into regulator subsystem. This means we
have to make sure no allocations happen with a regulator or regulator
list locked.
After this series I see no more lockdep complaints on my test system,
but please review and test further.
First four patches move allocations out of locked regions, next three
came as a drive-by cleanups.
---
v2: fix bug in patch #4 spotted by kernel test robot
reworded commit #7 description
Michał Mirosław (7):
regulator: push allocation in regulator_init_coupling() outside of
lock
regulator: push allocation in regulator_ena_gpio_request() out of lock
regulator: push allocations in create_regulator() outside of lock
regulator: push allocation in set_consumer_device_supply() out of lock
regulator: plug of_node leak in regulator_register()'s error path
regulator: cleanup regulator_ena_gpio_free()
regulator: remove superfluous lock in regulator_resolve_coupling()
drivers/regulator/core.c | 164 +++++++++++++++++++++------------------
1 file changed, 87 insertions(+), 77 deletions(-)
--
2.20.1
I see it takes about 5us per regulator to grab the lock, check that this
regulator isn't going to do anything for suspend, and then release the
lock. When that is combined with PMICs that have dozens of regulators we
get into a state where we spend a few miliseconds doing a bunch of
locking operations synchronously to figure out that there's nothing to
do. Let's reorganize the code here a bit so that we don't grab the lock
until we're actually going to do something so that suspend is a little
faster.
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Cc: Matthias Kaehlcke <mka@chromium.org>
Cc: Douglas Anderson <dianders@chromium.org>
Link: https://lore.kernel.org/r/20200804070837.1084024-1-swboyd@chromium.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Move all allocations outside of the regulator_lock()ed section.
======================================================
WARNING: possible circular locking dependency detected
5.7.13+ #535 Not tainted
------------------------------------------------------
f2fs_discard-179:7/702 is trying to acquire lock:
c0e5d920 (regulator_list_mutex){+.+.}-{3:3}, at: regulator_lock_dependent+0x54/0x2c0
but task is already holding lock:
cb95b080 (&dcc->cmd_lock){+.+.}-{3:3}, at: __issue_discard_cmd+0xec/0x5f8
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
[...]
-> #3 (fs_reclaim){+.+.}-{0:0}:
fs_reclaim_acquire.part.11+0x40/0x50
fs_reclaim_acquire+0x24/0x28
__kmalloc_track_caller+0x54/0x218
kstrdup+0x40/0x5c
create_regulator+0xf4/0x368
regulator_resolve_supply+0x1a0/0x200
regulator_register+0x9c8/0x163c
[...]
other info that might help us debug this:
Chain exists of:
regulator_list_mutex --> &sit_i->sentry_lock --> &dcc->cmd_lock
[...]
Fixes: f8702f9e4a ("regulator: core: Use ww_mutex for regulators locking")
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/6eebc99b2474f4ffaa0405b15178ece0e7e4f608.1597195321.git.mirq-linux@rere.qmqm.pl
Signed-off-by: Mark Brown <broonie@kernel.org>
Move another allocation out of regulator_list_mutex-protected region, as
reclaim might want to take the same lock.
WARNING: possible circular locking dependency detected
5.7.13+ #534 Not tainted
------------------------------------------------------
kswapd0/383 is trying to acquire lock:
c0e5d920 (regulator_list_mutex){+.+.}-{3:3}, at: regulator_lock_dependent+0x54/0x2c0
but task is already holding lock:
c0e38518 (fs_reclaim){+.+.}-{0:0}, at: __fs_reclaim_acquire+0x0/0x50
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #1 (fs_reclaim){+.+.}-{0:0}:
fs_reclaim_acquire.part.11+0x40/0x50
fs_reclaim_acquire+0x24/0x28
kmem_cache_alloc_trace+0x40/0x1e8
regulator_register+0x384/0x1630
devm_regulator_register+0x50/0x84
reg_fixed_voltage_probe+0x248/0x35c
[...]
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(fs_reclaim);
lock(regulator_list_mutex);
lock(fs_reclaim);
lock(regulator_list_mutex);
*** DEADLOCK ***
[...]
2 locks held by kswapd0/383:
#0: c0e38518 (fs_reclaim){+.+.}-{0:0}, at: __fs_reclaim_acquire+0x0/0x50
#1: cb70e5e0 (hctx->srcu){....}-{0:0}, at: hctx_lock+0x60/0xb8
[...]
Fixes: 541d052d72 ("regulator: core: Only support passing enable GPIO descriptors")
[this commit only changes context]
Fixes: f8702f9e4a ("regulator: core: Use ww_mutex for regulators locking")
[this is when the regulator_list_mutex was introduced in reclaim locking path]
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Link: https://lore.kernel.org/r/41fe6a9670335721b48e8f5195038c3d67a3bf92.1597195321.git.mirq-linux@rere.qmqm.pl
Signed-off-by: Mark Brown <broonie@kernel.org>
Allocating memory with regulator_list_mutex held makes lockdep unhappy
when memory pressure makes the system do fs_reclaim on eg. eMMC using
a regulator. Push the lock inside regulator_init_coupling() after the
allocation.
======================================================
WARNING: possible circular locking dependency detected
5.7.13+ #533 Not tainted
------------------------------------------------------
kswapd0/383 is trying to acquire lock:
cca78ca4 (&sbi->write_io[i][j].io_rwsem){++++}-{3:3}, at: __submit_merged_write_cond+0x104/0x154
but task is already holding lock:
c0e38518 (fs_reclaim){+.+.}-{0:0}, at: __fs_reclaim_acquire+0x0/0x50
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #2 (fs_reclaim){+.+.}-{0:0}:
fs_reclaim_acquire.part.11+0x40/0x50
fs_reclaim_acquire+0x24/0x28
__kmalloc+0x54/0x218
regulator_register+0x860/0x1584
dummy_regulator_probe+0x60/0xa8
[...]
other info that might help us debug this:
Chain exists of:
&sbi->write_io[i][j].io_rwsem --> regulator_list_mutex --> fs_reclaim
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(fs_reclaim);
lock(regulator_list_mutex);
lock(fs_reclaim);
lock(&sbi->write_io[i][j].io_rwsem);
*** DEADLOCK ***
1 lock held by kswapd0/383:
#0: c0e38518 (fs_reclaim){+.+.}-{0:0}, at: __fs_reclaim_acquire+0x0/0x50
[...]
Fixes: d8ca7d184b ("regulator: core: Introduce API for regulators coupling customization")
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Reviewed-by: Dmitry Osipenko <digetx@gmail.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/1a889cf7f61c6429c9e6b34ddcdde99be77a26b6.1597195321.git.mirq-linux@rere.qmqm.pl
Signed-off-by: Mark Brown <broonie@kernel.org>
The change corrects registration and deregistration on error path
of a regulator, the problem was manifested by a reported memory
leak on deferred probe:
as3722-regulator as3722-regulator: regulator 13 register failed -517
# cat /sys/kernel/debug/kmemleak
unreferenced object 0xecc43740 (size 64):
comm "swapper/0", pid 1, jiffies 4294937640 (age 712.880s)
hex dump (first 32 bytes):
72 65 67 75 6c 61 74 6f 72 2e 32 34 00 5a 5a 5a regulator.24.ZZZ
5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZZZZZZZZZ
backtrace:
[<0c4c3d1c>] __kmalloc_track_caller+0x15c/0x2c0
[<40c0ad48>] kvasprintf+0x64/0xd4
[<109abd29>] kvasprintf_const+0x70/0x84
[<c4215946>] kobject_set_name_vargs+0x34/0xa8
[<62282ea2>] dev_set_name+0x40/0x64
[<a39b6757>] regulator_register+0x3a4/0x1344
[<16a9543f>] devm_regulator_register+0x4c/0x84
[<51a4c6a1>] as3722_regulator_probe+0x294/0x754
...
The memory leak problem was introduced as a side ef another fix in
regulator_register() error path, I believe that the proper fix is
to decouple device_register() function into its two compounds and
initialize a struct device before assigning any values to its fields
and then using it before actual registration of a device happens.
This lets to call put_device() safely after initialization, and, since
now a release callback is called, kfree(rdev->constraints) shall be
removed to exclude a double free condition.
Fixes: a3cde9534e ("regulator: core: fix regulator_register() error paths to properly release rdev")
Signed-off-by: Vladimir Zapolskiy <vz@mleia.com>
Cc: Wen Yang <wenyang@linux.alibaba.com>
Link: https://lore.kernel.org/r/20200724005013.23278-1-vz@mleia.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Part of the regulator_get() code is already factored out into
create_regulator(). This patch factors out some of the regulator_put()
code into destroy_regulator() so that create_regulator() has a
corresponding unwind function. Subsequent patches will use this
function.
Signed-off-by: Saravana Kannan <saravanak@google.com>
Link: https://lore.kernel.org/r/20200716042053.1927676-3-saravanak@google.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Some regulators might need to verify that they have indeed been enabled
after the enable() call is made and enable_time delay has passed.
This is implemented by repeatedly checking is_enabled() upto
poll_enabled_time, waiting for the already calculated enable delay in
each iteration.
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
Link: https://lore.kernel.org/r/20200622124110.20971-2-sumit.semwal@linaro.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Hi!
This patchset is another attempt to fix the regulator coupling on
Exynos5800/5422 SoCs. Here are links to the previous attempts:
https://lore.kernel.org/linux-samsung-soc/20191008101709.qVNy8eijBi0LynOteWFMnTg4GUwKG599n6OyYoX1Abs@z/https://lore.kernel.org/lkml/20191017102758.8104-1-m.szyprowski@samsung.com/https://lore.kernel.org/linux-pm/cover.1589528491.git.viresh.kumar@linaro.org/https://lore.kernel.org/linux-pm/20200528131130.17984-1-m.szyprowski@samsung.com/
The problem is with "vdd_int" regulator coupled with "vdd_arm" on Odroid
XU3/XU4 boards family. "vdd_arm" is handled by CPUfreq. "vdd_int" is
handled by devfreq. CPUfreq initialized quite early during boot and it
starts changing OPPs and "vdd_arm" value. Sometimes CPU activity during
boot goes down and some low-frequency OPPs are selected, what in turn
causes lowering "vdd_arm". This happens before devfreq applies its
requirements on "vdd_int". Regulator balancing code reduces "vdd_arm"
voltage value, what in turn causes lowering "vdd_int" value to the lowest
possible value. This is much below the operation point of the wcore bus,
which still runs at the highest frequency.
The issue was hard to notice because in the most cases the board managed
to boot properly, even when the regulator was set to lowest value allowed
by the regulator constraints. However, it caused some random issues,
which can be observed as "Unhandled prefetch abort" or low USB stability.
Adding more and more special cases to the generic code has been rejected,
so the only way to ensure the desired behavior on Exynos5800-based SoCs
is to make a custom regulator coupler driver.
Best regards,
Marek Szyprowski
Patch summary:
Marek Szyprowski (2):
regulator: extract voltage balancing code to separate function
soc: samsung: Add simple voltage coupler for Exynos5800
arch/arm/mach-exynos/Kconfig | 1 +
drivers/regulator/core.c | 49 ++++++++-------
drivers/soc/samsung/Kconfig | 3 +
drivers/soc/samsung/Makefile | 1 +
.../soc/samsung/exynos-regulator-coupler.c | 59 +++++++++++++++++++
include/linux/regulator/coupler.h | 8 +++
6 files changed, 101 insertions(+), 20 deletions(-)
create mode 100644 drivers/soc/samsung/exynos-regulator-coupler.c
--
2.17.1
base-commit: 8f3d9f3542
