The Qualcomm PM8008 PMIC is a so called QPNP PMIC with seven LDO
regulators, a temperature alarm block and two GPIO pins (which are also
used for interrupt signalling and reset).
Unlike previous QPNP PMICs it uses an I2C rather than SPMI interface,
which has implications for how interrupts are handled.
A previous attempt by Qualcomm to upstream support for PM8008 stalled
two years ago at version 15 after a lot of back and forth discussion on
how best to describe this device in the devicetree. [1]
After reviewing the backstory on this and surveying the current SPMI
PMIC bindings and implementation, I opted for a new approach that does
not describe internal details like register offsets and interrupts in
the devicetree.
The original decision to include registers offsets and internal
interrupts for SPMI PMICs has led to a number of PMIC dtsi being created
to avoid copying lots of boiler plate declarations. This in turn causes
trouble when the PMIC USID address is configurable as the address is
included in every interrupt specifier.
The current SPMI bindings still do not describe the devices fully and
additional data is therefore already provided by drivers (e.g.
additional register blocks, supplies, additional interrupt specifiers).
The fact that PMICs which use two USIDs (addresses) are modelled as two
separate devices causes trouble, for example, when there are
dependencies between subfunctions. [2]
Subfunctions also do not necessarily map neatly onto the 256-register
block partitioning of the SPMI register space, something which has lead
to unresolved inconsistencies in how functions like PWM are described.
[3]
In short, it's a bit of a mess.
With the new style of bindings, by contrast, only essential information
that actually differs between machines would be included in the
devicetree. The bindings would also be mostly decoupled from the
implementation, which has started to leak out into the binding (e.g. how
the QPNP interrupts are handled). This also allows for extending the
implementation without having to update the binding, which is especially
important as Qualcomm does not publish any documentation (e.g. to later
enable regulator over-current protection).
Some PMICs support both I2C and SPMI interfaces (e.g. PM8010) and we
want to be able to reuse the same bindings regardless of the interface.
As a proof concept I have written a new pmc8280 driver for one of the
SPMI PMICs in the Lenovo ThinkPad X13s that uses the new style of
bindings and I've been using that one to control backlight and
peripheral regulators for a while now. Specifically, the gpio and
temperature-alarm blocks can be used with some minor updates to the
current drivers.
That work still needs a bit of polish before posting, but my working PoC
means that I'm confident enough that the new model will work and that we
can go ahead and merge regulator support for the PM8008.
This series is specifically needed for the camera sensors in the X13s,
for which camera subsystem (camss) support has now been merged for 6.10.
The first seven patches are preparatory and can possibly be merged
separately from the rest of the series. The next two patches drop the
broken GPIO support for PM8008 which had already been upstreamed. The
last five patches rework the binding and MFD driver, add support for the
regulators and enable the camera PMIC on the X13s.
Johan
[1] https://lore.kernel.org/all/[email protected]
[2] https://lore.kernel.org/lkml/[email protected]
[3] https://lore.kernel.org/r/[email protected]
Changes in v2
- use IRQ_TYPE_SENSE_MASK in regmap_irq table
- add post-reset delay
- reorder pinctrl binding and driver update
- split out binding cleanups
- use platform_device_id matching
- replace underscore in supply names with dash
- use more fine-grained includes in regulator driver
- rework regulator driver and update authorship
Johan Hovold (14):
dt-bindings: mfd: pm8008: add reset gpio
mfd: pm8008: fix regmap irq chip initialisation
mfd: pm8008: deassert reset on probe
mfd: pm8008: mark regmap structures as const
mfd: pm8008: use lower case hex notation
mfd: pm8008: rename irq chip
mfd: pm8008: drop unused driver data
pinctrl: qcom: spmi-gpio: drop broken pm8008 support
dt-bindings: pinctrl: qcom,pmic-gpio: drop pm8008
dt-bindings: mfd: pm8008: drop redundant descriptions
dt-bindings: mfd: pm8008: rework binding
mfd: pm8008: rework driver
regulator: add pm8008 pmic regulator driver
arm64: dts: qcom: sc8280xp-x13s: enable pm8008 camera pmic
.../devicetree/bindings/mfd/qcom,pm8008.yaml | 158 ++++++++------
.../bindings/pinctrl/qcom,pmic-gpio.yaml | 3 -
.../qcom/sc8280xp-lenovo-thinkpad-x13s.dts | 123 +++++++++++
drivers/mfd/Kconfig | 1 +
drivers/mfd/qcom-pm8008.c | 169 ++++++++++-----
drivers/pinctrl/qcom/pinctrl-spmi-gpio.c | 1 -
drivers/regulator/Kconfig | 7 +
drivers/regulator/Makefile | 1 +
drivers/regulator/qcom-pm8008-regulator.c | 198 ++++++++++++++++++
include/dt-bindings/mfd/qcom-pm8008.h | 19 --
10 files changed, 542 insertions(+), 138 deletions(-)
create mode 100644 drivers/regulator/qcom-pm8008-regulator.c
delete mode 100644 include/dt-bindings/mfd/qcom-pm8008.h
--
2.44.1
On Wed, 29 May 2024, Johan Hovold wrote:
> The Qualcomm PM8008 PMIC is a so called QPNP PMIC with seven LDO
> regulators, a temperature alarm block and two GPIO pins (which are also
> used for interrupt signalling and reset).
I don't see any issues with the MFD commits.
When you submit this, would you do me a favour and change the subject
lines to match that of the subsystem. I usually silently change them,
but this is a large set and it'll become annoying real quick.
`git log --oneline -- <subsystem>` is your friend.
Thanks.
--
Lee Jones [李琼斯]