Re: [PATCH v1 2/2] power: reset: spacemit-p1: Use atomic I2C transfers for power-off and restart

From: Aurelien Jarno

Date: Tue Nov 17 2026 - 00:34:20 EST


Hi Anand,

On 2026-11-16 13:06, Anand Moon wrote:
> Hi All,
>
> On Mon, 16 Nov 2026 at 11:33, Sebastian Reichel
> <sebastian.reichel@xxxxxxxxxxxxx> wrote:
> >
> > Hi,
> >
> > On Tue, Nov 10, 2026 at 11:44:27AM +0530, Anand Moon wrote:
> > > The SpacemiT P1 power-off and restart handlers use regmap_set_bits()
> > > to access the PMIC. However, these handlers run in an atomic notifier
> > > context (specifically within an rcu_read_lock() critical section), while
> > > the I2C-backed regmap uses a sleepable mutex.
> > >
> > > This triggers a "WARNING: suspicious RCU usage" warning due to an illegal
> > > context switch in the RCU read-side critical section during shutdown on the
> > > Banana Pi BPI-F3:
> > >
> > > [ 563.803475] WARNING: suspicious RCU usage
> > > [ 563.807813] 7.3.1-ARCH-RISC-V #1 Not tainted
> > > [ 563.816614] ./include/linux/rcupdate.h:387 Illegal context switch in RCU read-side critical section!
> > > ...
> > > [ 563.841445] locks held by shutdown/1: 2, last CPU#0:
> > > [ 563.846694] #0: ffffffff8404fb20 (system_transition_mutex){+.+.}-{4:4}, at: __do_sys_reboot+0x230/0x29e
> > > [ 563.856806] #1: ffffffff84167860 (rcu_read_lock){....}-{1:3}, at: atomic_notifier_call_chain+0x4a/0x1c2
> > >
> > > Replace the regmap operations with atomic I2C transfers to avoid sleeping
> > > in an atomic context. Preserve unrelated register bits through
> > > read-modify-write operations and check for atomic I2C support before
> > > registering the handlers.
> > >
> > > Obtain the I2C client from the parent device and verify that the adapter
> > > supports atomic transfers before registering the handlers.
> > >
> > > Tested on the Banana Pi BPI-F3 (SpacemiT K1).
> > >
> > > Assisted-by: LLM
> > > Cc: Troy Mitchell <troy.mitchell@xxxxxxxxxxxxxxxxxx>
> > > Cc: Aurelien Jarno <aurelien@xxxxxxxxxxx>
> > > Fixes: 28124cc0fb8c ("driver: reset: spacemit-p1: add driver for poweroff/reboot")
> > > Signed-off-by: Anand Moon <linux.amoon@xxxxxxxxx>
> > > ---
> >
> > regmap runs i2c_transfer itself and thus also uses the atomic I2C
> > transfers. The problem here is regmap itself. But instead of moving
> > away from regmap, the atomic use case should be supported properly
> > in regmap.
> >
>
> On RISC-V platform, poweroff and reboot are already handled by OpenSBI
> through the SBI System Reset (SRST) extension. Linux delegates these
> operations to firmware using SBI_SRST_RESET_TYPE_SHUTDOWN for poweroff
> and SBI_SRST_RESET_TYPE_COLD_REBOOT or SBI_SRST_RESET_TYPE_WARM_REBOOT
> for reboot, depending on the supported reset type.
>
> Therefore, there is no need for an additional Linux PMIC poweroff/reboot handler
> on this platform, as it duplicates functionality already provided by firmware.
>
> While supporting atomic operations in regmap may be useful more generally,
> it is not necessary to address this platform's shutdown and reboot requirements.
>
> I tested this on the SpacemiT K1 Banana Pi BPI-F3 after removing this PMIC
> poweroff/reboot driver.
>
> OpenSBI commit which handles this extension
>
> d9637d00bf47 ("lib: utils/reset: add SpacemiT P1 PMIC support")
> b10e18ec854a ("lib: utils/i2c: add minimal SpacemiT I2C driver")
>
> Similarly, on the StarFive VisionFive 2, poweroff is handled by
> OpenSBI through the
> AXP15060 PMIC, so there is no need to register an additional poweroff
> handler using
> devm_register_power_off_handler().
>
> OpenSBI commit which handles this extension
>
> 568ea49490d9 ("platform: starfive: add PMIC power ops in JH7110
> visionfive2 board")
>
> -----8<----------8<----------8<------
> diff --git a/drivers/mfd/axp20x.c b/drivers/mfd/axp20x.c
> index 50df66de24e06..069259f7d7c15 100644
> --- a/drivers/mfd/axp20x.c
> +++ b/drivers/mfd/axp20x.c
> @@ -1458,8 +1458,14 @@ int axp20x_device_probe(struct axp20x_dev *axp20x)
> return ret;
> }
>
> - if (axp20x->variant != AXP288_ID)
> - devm_register_power_off_handler(axp20x->dev,
> axp20x_power_off, axp20x);
> + /*
> + * The VisionFive 2 handles AXP15060 poweroff through OpenSBI.
> + * Avoid accessing the I2C PMIC from the atomic poweroff path.
> + */
> + if (axp20x->variant != AXP288_ID && !(axp20x->variant == AXP15060_ID &&
> + of_machine_is_compatible("starfive,jh7110")))
> + devm_register_power_off_handler(axp20x->dev,
> + axp20x_power_off, axp20x);
>
> dev_info(axp20x->dev, "AXP20X driver loaded\n");
>
> Please let me know if this works on your board?

Yes, that works, but I don't think this is the correct solution to
disable here. Also this won't scale for future boards which might use a
different existing PMIC driver with poweroff support and that will also
need to be patched.

As pointed out by Troy in an earlier mail, the real problem is the
priority at which the handler is registers, 192 for the reset and -256
for the poweroff one. I wonder if the SBI SRST extension should also
register the poweroff one at 192. It seems it's about replacing the call
to register_platform_power_off() by register_sys_off_handler(), which
supports using a priority value. And maybe SYS_OFF_PRIO_FIRMWARE = 224
should be used there instead of 192.

Regards
Aurelien

--
Aurelien Jarno GPG: 4096R/1DDD8C9B
aurelien@xxxxxxxxxxx http://aurel32.net