| Message ID | 20220802095252.2486591-1-foss+kernel@0leil.net |
|---|---|
| Headers | show |
| Series | Making Rockchip IO domains dependency from other devices explicit | expand |
On Tue, Aug 2, 2022 at 11:53 AM Quentin Schulz <foss+kernel@0leil.net> wrote: > Some background on IO domains on Rockchip: > > On some Rockchip SoCs, some SoC pins are split in what are called IO > domains. > > An IO domain is supplied power externally, by regulators from a PMIC for > example. This external power supply is then used by the IO domain as > "supply" for the IO pins if they are outputs. > > Each IO domain can configure which voltage the IO pins will be operating > on (1.8V or 3.3V). > > There already exists an IO domain driver for Rockchip SoCs[1]. This > driver allows to explicit the relationship between the external power > supplies and IO domains[2]. This makes sure the regulators are enabled > by the Linux kernel so the IO domains are supplied with power and > correctly configured as per the supplied voltage. > This driver is a regulator consumer and does not offer any other > interface for device dependency. What makes me confused about the patch is the relationship, if any, between this "IO domain" and generic power domains (genpd) that has been worked on for ~10 years. I am worried that we are reinventing the world. While my intuitive feeling is that genpd power domains are only on-chip and not considering off-chip pins, I am not so sure that it warrants its own abstraction and want to know whether this can be retrofit into genpd rather than inventing this? Documentation/devicetree/bindings/power/power-domain.yaml include/linux/pm_domain.h Yours, Linus Walleij
Hi Linus, Am Montag, 22. August 2022, 10:38:11 CEST schrieb Linus Walleij: > On Tue, Aug 2, 2022 at 11:53 AM Quentin Schulz <foss+kernel@0leil.net> wrote: > > > Some background on IO domains on Rockchip: > > > > On some Rockchip SoCs, some SoC pins are split in what are called IO > > domains. > > > > An IO domain is supplied power externally, by regulators from a PMIC for > > example. This external power supply is then used by the IO domain as > > "supply" for the IO pins if they are outputs. > > > > Each IO domain can configure which voltage the IO pins will be operating > > on (1.8V or 3.3V). > > > > There already exists an IO domain driver for Rockchip SoCs[1]. This > > driver allows to explicit the relationship between the external power > > supplies and IO domains[2]. This makes sure the regulators are enabled > > by the Linux kernel so the IO domains are supplied with power and > > correctly configured as per the supplied voltage. > > This driver is a regulator consumer and does not offer any other > > interface for device dependency. > > What makes me confused about the patch is the relationship, if any, > between this "IO domain" and generic power domains (genpd) that has > been worked on for ~10 years. > > I am worried that we are reinventing the world. In a nutshell, the Rockchip io-domains handle the voltages of specific pin-groups. I.e. mostly it is just switching between 1.8V and 3.3V . The voltage itself is always set in a (i2c-)regulator but there is a separate step necessary to tell the soc this information. 3.3 -> 1.8: set regulator to 1.8, tell io-domain "we're at 1.8 now" 1.8 -> 3.3: tell io-domain "3.3", set regulator to 3.3. There is supposedly a soc-health-issue if you set the regulator to 3.3 while the io-domain thinks it's at 1.8 . So the io-domain driver right now, just attaches to the regulator, catches the voltage-change events and sets the io-domain setting accordingly. What Quentin is trying to solve is a probe-dependency issue that can happen when stuff is built into the kernel, the regulator has probed the regulator using driver has probed, but the io.domain driver hasn't, as that also only attached to the regulator as described above. Heiko > While my intuitive feeling is that genpd power domains are only on-chip > and not considering off-chip pins, I am not so sure that it warrants > its own abstraction and want to know whether this can be retrofit into > genpd rather than inventing this? > > Documentation/devicetree/bindings/power/power-domain.yaml > include/linux/pm_domain.h > > Yours, > Linus Walleij >
From: Quentin Schulz <quentin.schulz@theobroma-systems.com> This is a follow-up to the mail sent almost two months ago asking for guidance: https://lore.kernel.org/lkml/778790a4-1239-e9d9-0549-6760a8792ceb@theobroma-systems.com/ This is what I could come up with but I'm not too happy about it so feel free to give some ideas or other possible implementations that would have less downsides than this one. Some background on IO domains on Rockchip: On some Rockchip SoCs, some SoC pins are split in what are called IO domains. An IO domain is supplied power externally, by regulators from a PMIC for example. This external power supply is then used by the IO domain as "supply" for the IO pins if they are outputs. Each IO domain can configure which voltage the IO pins will be operating on (1.8V or 3.3V). There already exists an IO domain driver for Rockchip SoCs[1]. This driver allows to explicit the relationship between the external power supplies and IO domains[2]. This makes sure the regulators are enabled by the Linux kernel so the IO domains are supplied with power and correctly configured as per the supplied voltage. This driver is a regulator consumer and does not offer any other interface for device dependency. However, IO pins belonging to an IO domain need to have this IO domain correctly configured before they are being used otherwise they do not operate correctly (in our case, a pin configured as output clock was oscillating between 0 and 150mV instead of the expected 1V8). In order to make this dependency transparent to the consumer of those pins and not add Rockchip-specific code to third party drivers (a camera driver in our case), it is hooked into the pinctrl driver which is Rockchip-specific obviously. Unfortunately, the dependency is not about the ultimate presence of the io-domain devices but more that prior to being able to use pins, the io-domain devices need to be probed. However of_find_device_by_node does not provide this information, hence the check whether the platform_device actually has its drvdata structure set (it defaults to NULL until it is filled by the driver during the probe of the device). Moreover, this dependency needs to be explicit on a pinmux level and postponed after the probe of the pinctrl driver because a circular dependency is observed otherwise with the following: pinctrl device depends on the io-domain device which depends on regulators from a PMIC on i2c which requires the i2c bus pins to be muxed from the pinctrl device. Instead, the pinmux dependency on IO domain is checked in set_mux callback and returns EPROBE_DEFER if the IO domain device hasn't probed yet. I wanted to add the appropriate rockchip,io-domains DT property to existing pinmux DT node. However, *all* of PX30's belong to an IO domain, including the i2c bus on which the PMIC supplying the power to the IO domain is on. Since the PMIC can be virtually on any i2c bus, a specific pinmux cannot be omitted or ignored, therefore none are added and it's up to the board maintainers to add them themselves. This is also assumed only necessary for IO domain that are configured differently by the bootloader or by register defaults (3V3 for RK3399/PX30) than their expected values in Linux and whose supplied power is fixed (e.g. not expected to be able to change from 3V3 to 1V8). The hope is to not have to handle the io domain configuration in the bootloader as it is currently done, c.f. https://elixir.bootlin.com/u-boot/latest/source/board/theobroma-systems/puma_rk3399/puma-rk3399.c#L28 (we'll need an additional IO domain configuration for a camera soon). However, this still relies on IO domains defaults or bootloader configuration to be able to omit some IO domain<->pinmux relationships to avoid circular dependencies. I don't know how well this RFC implementation would work with suspend/resume. [1] drivers/soc/rockchip/io-domain.c [2] Documentation/devicetree/bindings/power/rockchip-io-domain.yaml Cheers, Quentin Quentin Schulz (1): pinctrl: rockchip: add support for per-pinmux io-domain dependency drivers/pinctrl/pinctrl-rockchip.c | 19 +++++++++++++++++++ drivers/pinctrl/pinctrl-rockchip.h | 1 + 2 files changed, 20 insertions(+)