linux-stable.git/drivers/regulator/core.c, branch linux-6.2.y

regulator: core: Avoid lockdep reports when resolving supplies

2023-05-11T14:11:03+00:00

[ Upstream commit cba6cfdc7c3f1516f0d08ddfb24e689af0932573 ]

An automated bot told me that there was a potential lockdep problem
with regulators. This was on the chromeos-5.15 kernel, but I see
nothing that would be different downstream compared to upstream. The
bot said:
  ============================================
  WARNING: possible recursive locking detected
  5.15.104-lockdep-17461-gc1e499ed6604 #1 Not tainted
  --------------------------------------------
  kworker/u16:4/115 is trying to acquire lock:
  ffffff8083110170 (regulator_ww_class_mutex){+.+.}-{3:3}, at: create_regulator+0x398/0x7ec

  but task is already holding lock:
  ffffff808378e170 (regulator_ww_class_mutex){+.+.}-{3:3}, at: ww_mutex_trylock+0x3c/0x7b8

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

         CPU0
         ----
    lock(regulator_ww_class_mutex);
    lock(regulator_ww_class_mutex);

   *** DEADLOCK ***

   May be due to missing lock nesting notation

  4 locks held by kworker/u16:4/115:
   #0: ffffff808006a948 ((wq_completion)events_unbound){+.+.}-{0:0}, at: process_one_work+0x520/0x1348
   #1: ffffffc00e0a7cc0 ((work_completion)(&entry->work)){+.+.}-{0:0}, at: process_one_work+0x55c/0x1348
   #2: ffffff80828a2260 (&dev->mutex){....}-{3:3}, at: __device_attach_async_helper+0xd0/0x2a4
   #3: ffffff808378e170 (regulator_ww_class_mutex){+.+.}-{3:3}, at: ww_mutex_trylock+0x3c/0x7b8

  stack backtrace:
  CPU: 2 PID: 115 Comm: kworker/u16:4 Not tainted 5.15.104-lockdep-17461-gc1e499ed6604 #1 9292e52fa83c0e23762b2b3aa1bacf5787a4d5da
  Hardware name: Google Quackingstick (rev0+) (DT)
  Workqueue: events_unbound async_run_entry_fn
  Call trace:
   dump_backtrace+0x0/0x4ec
   show_stack+0x34/0x50
   dump_stack_lvl+0xdc/0x11c
   dump_stack+0x1c/0x48
   __lock_acquire+0x16d4/0x6c74
   lock_acquire+0x208/0x750
   __mutex_lock_common+0x11c/0x11f8
   ww_mutex_lock+0xc0/0x440
   create_regulator+0x398/0x7ec
   regulator_resolve_supply+0x654/0x7c4
   regulator_register_resolve_supply+0x30/0x120
   class_for_each_device+0x1b8/0x230
   regulator_register+0x17a4/0x1f40
   devm_regulator_register+0x60/0xd0
   reg_fixed_voltage_probe+0x728/0xaec
   platform_probe+0x150/0x1c8
   really_probe+0x274/0xa20
   __driver_probe_device+0x1dc/0x3f4
   driver_probe_device+0x78/0x1c0
   __device_attach_driver+0x1ac/0x2c8
   bus_for_each_drv+0x11c/0x190
   __device_attach_async_helper+0x1e4/0x2a4
   async_run_entry_fn+0xa0/0x3ac
   process_one_work+0x638/0x1348
   worker_thread+0x4a8/0x9c4
   kthread+0x2e4/0x3a0
   ret_from_fork+0x10/0x20

The problem was first reported soon after we made many of the
regulators probe asynchronously, though nothing I've seen implies that
the problems couldn't have also happened even without that.

I haven't personally been able to reproduce the lockdep issue, but the
issue does look somewhat legitimate. Specifically, it looks like in
regulator_resolve_supply() we are holding a "rdev" lock while calling
set_supply() -> create_regulator() which grabs the lock of a
_different_ "rdev" (the one for our supply). This is not necessarily
safe from a lockdep perspective since there is no documented ordering
between these two locks.

In reality, we should always be locking a regulator before the
supplying regulator, so I don't expect there to be any real deadlocks
in practice. However, the regulator framework in general doesn't
express this to lockdep.

Let's fix the issue by simply grabbing the two locks involved in the
same way we grab multiple locks elsewhere in the regulator framework:
using the "wound/wait" mechanisms.

Fixes: eaa7995c529b ("regulator: core: avoid regulator_resolve_supply() race condition")
Signed-off-by: Douglas Anderson 
Link: https://lore.kernel.org/r/20230329143317.RFC.v2.2.I30d8e1ca10cfbe5403884cdd192253a2e063eb9e@changeid
Signed-off-by: Mark Brown 
Signed-off-by: Sasha Levin

regulator: core: Consistently set mutex_owner when using ww_mutex_lock_slow()

2023-05-11T14:11:03+00:00

[ Upstream commit b83a1772be854f87602de14726737d3e5b06e1f4 ]

When a codepath locks a rdev using ww_mutex_lock_slow() directly then
that codepath is responsible for incrementing the "ref_cnt" and also
setting the "mutex_owner" to "current".

The regulator core consistently got that right for "ref_cnt" but
didn't always get it right for "mutex_owner". Let's fix this.

It's unlikely that this truly matters because the "mutex_owner" is
only needed if we're going to do subsequent locking of the same
rdev. However, even though it's not truly needed it seems less
surprising if we consistently set "mutex_owner" properly.

Fixes: f8702f9e4aa7 ("regulator: core: Use ww_mutex for regulators locking")
Signed-off-by: Douglas Anderson 
Link: https://lore.kernel.org/r/20230329143317.RFC.v2.1.I4e9d433ea26360c06dd1381d091c82bb1a4ce843@changeid
Signed-off-by: Mark Brown 
Signed-off-by: Sasha Levin

regulator: core: Shorten off-on-delay-us for always-on/boot-on by time since booted

2023-05-11T14:10:58+00:00

[ Upstream commit 691c1fcda5351ed98a44610b7dccc0e3ee920020 ]

This is very close to a straight revert of commit 218320fec294
("regulator: core: Fix off-on-delay-us for always-on/boot-on
regulators"). We've identified that patch as causing a boot speed
regression on sc7180-trogdor boards. While boot speed certainly isn't
more important than making sure that power sequencing is correct,
looking closely at the original change it doesn't seem to have been
fully justified. It mentions "cycling issues" without describing
exactly what the issues were. That means it's possible that the
cycling issues were really a problem that should be fixed in a
different way.

Let's take a careful look at how we should handle regulators that have
an off-on-delay and that are boot-on or always-on. Linux currently
doesn't have any way to identify whether a GPIO regulator was already
on when the kernel booted. That means that when the kernel boots we
probe a regulator, see that it wants boot-on / always-on we, and then
turn the regulator on. We could be in one of two cases when we do
this:

a) The regulator might have been left on by the bootloader and we're
   ensuring that it stays on.
b) The regulator might have been left off by the bootloader and we're
   just now turning it on.

For case a) we definitely don't need any sort of delay. For case b) we
_might_ need some delay in case the bootloader turned the regulator
off _right_ before booting the kernel. To get the proper delay for
case b) then we can just assume a `last_off` of 0, which is what it
gets initialized to by default.

As per above, we can't tell whether we're in case a) or case b) so
we'll assume the longer delay (case b). This basically puts the code
to how it was before commit 218320fec294 ("regulator: core: Fix
off-on-delay-us for always-on/boot-on regulators"). However, we add
one important change: we make sure that the delay is actually honored
if `last_off` is 0. Though the original "cycling issues" cited were
vague, I'm hopeful that this important extra change will be enough to
fix the issues that the initial commit mentioned.

With this fix, I've confined that on a sc7180-trogdor board the delay
at boot goes down from 500 ms to ~250 ms. That's not as good as the 0
ms that we had prior to commit 218320fec294 ("regulator: core: Fix
off-on-delay-us for always-on/boot-on regulators"), but it's probably
safer because we don't know if the bootloader turned the regulator off
right before booting.

One note is that it's possible that we could be in a state that's not
a) or b) if there are other issues in the kernel. The only one I can
think of is related to pinctrl. If the pinctrl driver being used on a
board isn't careful about avoiding glitches when setting up a pin then
it's possible that setting up a pin could cause the regulator to "turn
off" briefly immediately before the regulator probes. If this is
indeed causing problems then the pinctrl driver should be fixed,
perhaps in a similar way to what was done in commit d21f4b7ffc22
("pinctrl: qcom: Avoid glitching lines when we first mux to output")

Fixes: 218320fec294 ("regulator: core: Fix off-on-delay-us for always-on/boot-on regulators")
Cc: Christian Kohlschütter 
Signed-off-by: Douglas Anderson 
Link: https://lore.kernel.org/r/20230313111806.1.I2eaad872be0932a805c239a7c7a102233fb0b03b@changeid
Signed-off-by: Mark Brown 
Signed-off-by: Sasha Levin

regulator: core: Use ktime_get_boottime() to determine how long a regulator was off

2023-03-10T08:29:47+00:00

commit 80d2c29e09e663761c2778167a625b25ffe01b6f upstream.

For regulators with 'off-on-delay-us' the regulator framework currently
uses ktime_get() to determine how long the regulator has been off
before re-enabling it (after a delay if needed). A problem with using
ktime_get() is that it doesn't account for the time the system is
suspended. As a result a regulator with a longer 'off-on-delay' (e.g.
500ms) that was switched off during suspend might still incurr in a
delay on resume before it is re-enabled, even though the regulator
might have been off for hours. ktime_get_boottime() accounts for
suspend time, use it instead of ktime_get().

Fixes: a8ce7bd89689 ("regulator: core: Fix off_on_delay handling")
Cc: stable@vger.kernel.org    # 5.13+
Signed-off-by: Matthias Kaehlcke 
Reviewed-by: Stephen Boyd 
Link: https://lore.kernel.org/r/20230223003301.v2.1.I9719661b8eb0a73b8c416f9c26cf5bd8c0563f99@changeid
Signed-off-by: Mark Brown 
Signed-off-by: Greg Kroah-Hartman

Merge tag 'regulator-fix-v6.2-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator

2022-12-23T22:38:00+00:00

Pull regulator fixes from Mark Brown:
 "Two core fixes here, one for a long standing race which some Qualcomm
  systems have started triggering with their UFS driver and another
  fixing a problem with supply lookup introduced by the fixes for devm
  related use after free issues that were introduced in this merge
  window"

* tag 'regulator-fix-v6.2-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
  regulator: core: fix deadlock on regulator enable
  regulator: core: Fix resolve supply lookup issue

regulator: core: fix deadlock on regulator enable

2022-12-15T11:56:26+00:00

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: 9243a195be7a ("regulator: core: Change voltage setting path")
Fixes: f8702f9e4aa7 ("regulator: core: Use ww_mutex for regulators locking")
Cc: stable@vger.kernel.org      # 5.0
Signed-off-by: Johan Hovold 
Link: https://lore.kernel.org/r/20221215104646.19818-1-johan+linaro@kernel.org
Signed-off-by: Mark Brown

regulator: core: Fix resolve supply lookup issue

2022-12-14T13:36:45+00:00

From Marek's log, the previous change modify the parent of rdev.
https://lore.kernel.org/all/58b92e75-f373-dae7-7031-8abd465bb874@samsung.com/

In 'regulator_resolve_supply', it uses the parent DT node of rdev as the
DT-lookup starting node. But the parent DT node may not exist. This will
cause the NULL supply issue.

This patch modify the parent of rdev back to the device that provides
from 'regulator_config' in 'regulator_register'.

Fixes: 8f3cbcd6b440 ("regulator: core: Use different devices for resource allocation and DT lookup")
Reported-by: Marek Szyprowski 
Tested-by: Marek Szyprowski 
Signed-off-by: ChiYuan Huang 
Link: https://lore.kernel.org/r/1670981831-12583-1-git-send-email-u0084500@gmail.com
Signed-off-by: Mark Brown

Merge tag 'regulator-v6.2' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator

2022-12-13T20:49:59+00:00

Pull regulator updates from Mark Brown:
 "Quite a quiet release for regulator, the diffstat is dominated by the
  I2C migration to probe_new() and the newly added MT6357 driver. We've
  just one framework addition and the rest is all new device support,
  fixes and cleanups.

  The framework addition is an API for requesting all regulators defined
  in DT, this isn't great practice but has reasonable applications when
  there is generic code handling devices on buses where the bus
  specification doesn't include power. The immediate application is MDIO
  but I believe there's others, it's another API that'll need an eye
  keeping on it for undesirable usage.

  Summary:

    - An API for requesting all regulators defined in DT

    - Conversion of lots of drivers to the I2C probe_new() API

    - Support for Mediatek MT6357, Qualcomm PM8550, PMR735a and Richtek
      RT6190"

* tag 'regulator-v6.2' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator: (56 commits)
  regulator: core: Use different devices for resource allocation and DT lookup
  dt-bindings: Add missing 'unevaluatedProperties' to regulator nodes
  regulator: qcom-labibb: Fix missing of_node_put() in qcom_labibb_regulator_probe()
  regulator: add mt6357 regulator
  regulator: dt-bindings: Add binding schema for mt6357 regulators
  regulator: core: fix resource leak in regulator_register()
  regulator: core: fix module refcount leak in set_supply()
  regulator: core: fix use_count leakage when handling boot-on
  regulator: rk808: Use dev_err_probe
  regulator: rk808: reduce 'struct rk808' usage
  regulator: Drop obsolete dependencies on COMPILE_TEST
  regulator: pv88080-regulator: Convert to i2c's .probe_new()
  regulator: pfuze100-regulator: Convert to i2c's .probe_new()
  regulator: isl6271a-regulator: Convert to i2c's .probe_new()
  regulator: fan53555: Convert to i2c's .probe_new()
  regulator: act8865-regulator: Convert to i2c's .probe_new()
  regulator: qcom-rpmh: Add support for PM8550 regulators
  regulator: dt-bindings: qcom,rpmh: Add compatible for PM8550
  regulator: tps65023-regulator: Convert to i2c's .probe_new()
  regulator: tps62360-regulator: Convert to i2c's .probe_new()
  ...

regulator: core: Use different devices for resource allocation and DT lookup

2022-12-08T13:02:37+00:00

Following by the below discussion, there's the potential UAF issue
between regulator and mfd.
https://lore.kernel.org/all/20221128143601.1698148-1-yangyingliang@huawei.com/

From the analysis of Yingliang

CPU A				|CPU B
mt6370_probe()			|
  devm_mfd_add_devices()	|
				|mt6370_regulator_probe()
				|  regulator_register()
				|    //allocate init_data and add it to devres
				|    regulator_of_get_init_data()
i2c_unregister_device()		|
  device_del()			|
    devres_release_all()	|
      // init_data is freed	|
      release_nodes()		|
				|  // using init_data causes UAF
				|  regulator_register()

It's common to use mfd core to create child device for the regulator.
In order to do the DT lookup for init data, the child that registered
the regulator would pass its parent as the parameter. And this causes
init data resource allocated to its parent, not itself. The issue happen
when parent device is going to release and regulator core is still doing
some operation of init data constraint for the regulator of child device.

To fix it, this patch expand 'regulator_register' API to use the
different devices for init data allocation and DT lookup.

Reported-by: Yang Yingliang 
Signed-off-by: ChiYuan Huang 
Link: https://lore.kernel.org/r/1670311341-32664-1-git-send-email-u0084500@gmail.com
Signed-off-by: Mark Brown

regulator: core: fix resource leak in regulator_register()

2022-12-02T11:55:58+00:00

I got some resource leak reports while doing fault injection test:

  OF: ERROR: memory leak, expected refcount 1 instead of 100,
  of_node_get()/of_node_put() unbalanced - destroy cset entry:
  attach overlay node /i2c/pmic@64/regulators/buck1

unreferenced object 0xffff88810deea000 (size 512):
  comm "490-i2c-rt5190a", pid 253, jiffies 4294859840 (age 5061.046s)
  hex dump (first 32 bytes):
    00 00 00 00 ad 4e ad de ff ff ff ff 00 00 00 00  .....N..........
    ff ff ff ff ff ff ff ff a0 1e 00 a1 ff ff ff ff  ................
  backtrace:
    [<00000000d78541e2>] kmalloc_trace+0x21/0x110
    [<00000000b343d153>] device_private_init+0x32/0xd0
    [<00000000be1f0c70>] device_add+0xb2d/0x1030
    [<00000000e3e6344d>] regulator_register+0xaf2/0x12a0
    [<00000000e2f5e754>] devm_regulator_register+0x57/0xb0
    [<000000008b898197>] rt5190a_probe+0x52a/0x861 [rt5190a_regulator]

unreferenced object 0xffff88810b617b80 (size 32):
  comm "490-i2c-rt5190a", pid 253, jiffies 4294859904 (age 5060.983s)
  hex dump (first 32 bytes):
    72 65 67 75 6c 61 74 6f 72 2e 32 38 36 38 2d 53  regulator.2868-S
    55 50 50 4c 59 00 ff ff 29 00 00 00 2b 00 00 00  UPPLY...)...+...
  backtrace:
    [<000000009da9280d>] __kmalloc_node_track_caller+0x44/0x1b0
    [<0000000025c6a4e5>] kstrdup+0x3a/0x70
    [<00000000790efb69>] create_regulator+0xc0/0x4e0
    [<0000000005ed203a>] regulator_resolve_supply+0x2d4/0x440
    [<0000000045796214>] regulator_register+0x10b3/0x12a0
    [<00000000e2f5e754>] devm_regulator_register+0x57/0xb0
    [<000000008b898197>] rt5190a_probe+0x52a/0x861 [rt5190a_regulator]

After calling regulator_resolve_supply(), the 'rdev->supply' is set
by set_supply(), after this set, in the error path, the resources
need be released, so call regulator_put() to avoid the leaks.

Fixes: aea6cb99703e ("regulator: resolve supply after creating regulator")
Fixes: 8a866d527ac0 ("regulator: core: Resolve supply name earlier to prevent double-init")
Signed-off-by: Yang Yingliang 
Link: https://lore.kernel.org/r/20221202025111.496402-1-yangyingliang@huawei.com
Signed-off-by: Mark Brown