Re: [PATCH v2 3/4] regulator: add SCMI driver
From: Cristian Marussi <cristian.marussi@arm.com>
Date: 2020-10-16 16:55:13
Also in:
lkml
On Fri, Oct 16, 2020 at 04:09:42PM +0200, Etienne Carriere wrote:
Hi Cristian, Some minor comments... Etienne On Thu, 15 Oct 2020 at 17:40, Cristian Marussi [off-list ref] wrote:quoted
Add a simple regulator based on SCMI Voltage Domain Protocol. Signed-off-by: Cristian Marussi <cristian.marussi@arm.com> ---- v1 --> v2 - removed duplicate regulator naming - removed redundant .get/set_voltage ops: only _sel variants implemented - removed condexpr on fail path to increase readability v0 --> v1 - fixed init_data constraint parsing - fixes for v5.8 (linear_range.h) - fixed commit message content and subject line format - factored out SCMI core specific changes to distinct patch - reworked Kconfig and Makefile to keep proper alphabetic order - fixed SPDX comment style - removed unneeded inline functions - reworked conditionals for legibility - fixed some return paths to properly report SCMI original errors codes - added some more descriptive error messages when fw returns invalid ranges - removed unneeded explicit devm_regulator_unregister from .remove() --- drivers/regulator/Kconfig | 9 + drivers/regulator/Makefile | 1 + drivers/regulator/scmi-regulator.c | 453 +++++++++++++++++++++++++++++ 3 files changed, 463 insertions(+) create mode 100644 drivers/regulator/scmi-regulator.cdiff --git a/drivers/regulator/Kconfig b/drivers/regulator/Kconfig index de17ef7e18f0..6d3a10cb9833 100644 --- a/drivers/regulator/Kconfig +++ b/drivers/regulator/Kconfig@@ -155,6 +155,15 @@ config REGULATOR_ARIZONA_MICSUPP and Wolfson Microelectronic Arizona codecs devices. +config REGULATOR_ARM_SCMI + tristate "SCMI based regulator driver" + depends on ARM_SCMI_PROTOCOL && OF + help + This adds the regulator driver support for ARM platforms using SCMI + protocol for device voltage management. + This driver uses SCMI Message Protocol driver to interact with the + firmware providing the device Voltage functionality. + config REGULATOR_AS3711 tristate "AS3711 PMIC" depends on MFD_AS3711diff --git a/drivers/regulator/Makefile b/drivers/regulator/Makefile index d8d3ecf526a8..0532a7393d5d 100644 --- a/drivers/regulator/Makefile +++ b/drivers/regulator/Makefile@@ -24,6 +24,7 @@ obj-$(CONFIG_REGULATOR_AD5398) += ad5398.o obj-$(CONFIG_REGULATOR_ANATOP) += anatop-regulator.o obj-$(CONFIG_REGULATOR_ARIZONA_LDO1) += arizona-ldo1.o obj-$(CONFIG_REGULATOR_ARIZONA_MICSUPP) += arizona-micsupp.o +obj-$(CONFIG_REGULATOR_ARM_SCMI) += scmi-regulator.o obj-$(CONFIG_REGULATOR_AS3711) += as3711-regulator.o obj-$(CONFIG_REGULATOR_AS3722) += as3722-regulator.o obj-$(CONFIG_REGULATOR_AXP20X) += axp20x-regulator.odiff --git a/drivers/regulator/scmi-regulator.c b/drivers/regulator/scmi-regulator.c new file mode 100644 index 000000000000..e4e7d0345723 --- /dev/null +++ b/drivers/regulator/scmi-regulator.c@@ -0,0 +1,453 @@ +// SPDX-License-Identifier: GPL-2.0 +// +// System Control and Management Interface (SCMI) based regulator driver +// +// Copyright (C) 2020 ARM Ltd. +// +// Implements a regulator driver on top of the SCMI Voltage Protocol. +// +// The ARM SCMI Protocol aims in general to hide as much as possible all the +// underlying operational details while providing an abstracted interface for +// its users to operate upon: as a consequence the resulting operational +// capabilities and configurability of this regulator device are much more +// limited than the ones usually available on a standard physical regulator. +// +// The supported SCMI regulator ops are restricted to the bare minimum: +// +// - 'status_ops': enable/disable/is_enabled +// - 'voltage_ops': get_voltage_sel/set_voltage_sel +// list_voltage/map_voltage +// +// Each SCMI regulator instance is associated, through the means of a proper DT +// entry description, to a specific SCMI Voltage Domain. + +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + +#include <linux/linear_range.h> +#include <linux/module.h> +#include <linux/of.h> +#include <linux/regulator/driver.h> +#include <linux/regulator/machine.h> +#include <linux/regulator/of_regulator.h> +#include <linux/scmi_protocol.h> +#include <linux/slab.h> +#include <linux/types.h> + +struct scmi_regulator { + u32 id; + struct scmi_device *sdev; + struct regulator_dev *rdev; + struct device_node *of_node; + struct regulator_desc desc; + struct regulator_config conf; +}; + +struct scmi_regulator_info { + int num_doms; + struct scmi_regulator **sregv; +}; + +static int scmi_reg_enable(struct regulator_dev *rdev) +{ + struct scmi_regulator *sreg = rdev_get_drvdata(rdev); + const struct scmi_handle *handle = sreg->sdev->handle; + + return handle->voltage_ops->config_set(handle, sreg->id, + SCMI_VOLTAGE_ARCH_STATE_ON); +} + +static int scmi_reg_disable(struct regulator_dev *rdev) +{ + struct scmi_regulator *sreg = rdev_get_drvdata(rdev); + const struct scmi_handle *handle = sreg->sdev->handle; + + return handle->voltage_ops->config_set(handle, sreg->id, + SCMI_VOLTAGE_ARCH_STATE_OFF); +} + +static int scmi_reg_is_enabled(struct regulator_dev *rdev) +{ + int ret; + u32 config; + struct scmi_regulator *sreg = rdev_get_drvdata(rdev); + const struct scmi_handle *handle = sreg->sdev->handle; + + ret = handle->voltage_ops->config_get(handle, sreg->id, + &config); + if (ret) { + dev_err(&sreg->sdev->dev, + "Error %d reading regulator %s status.\n", + ret, sreg->desc.name); + return 0; + } + + return config & SCMI_VOLTAGE_ARCH_STATE_ON; +} + +static int scmi_reg_get_voltage_sel(struct regulator_dev *rdev) +{ + int ret; + s32 volt_uV; + struct scmi_regulator *sreg = rdev_get_drvdata(rdev); + const struct scmi_handle *handle = sreg->sdev->handle; + + ret = handle->voltage_ops->level_get(handle, sreg->id, &volt_uV); + if (ret) + return ret; + + return sreg->desc.ops->map_voltage(rdev, volt_uV, volt_uV); +} + +static int scmi_reg_set_voltage_sel(struct regulator_dev *rdev, + unsigned int selector) +{ + int ret; + s32 volt_uV; + struct scmi_regulator *sreg = rdev_get_drvdata(rdev); + const struct scmi_handle *handle = sreg->sdev->handle; + + volt_uV = sreg->desc.ops->list_voltage(rdev, selector); + if (volt_uV <= 0) + return -EINVAL; + + ret = handle->voltage_ops->level_set(handle, sreg->id, 0x0, volt_uV); + if (ret) + return ret; + + return ret;return handle->voltage_ops->level_set(handle, sreg->id, 0x0, volt_uV);quoted
+} + +static const struct regulator_ops scmi_reg_fixed_ops = { + .enable = scmi_reg_enable, + .disable = scmi_reg_disable, + .is_enabled = scmi_reg_is_enabled, +}; + +static const struct regulator_ops scmi_reg_linear_ops = { + .enable = scmi_reg_enable, + .disable = scmi_reg_disable, + .is_enabled = scmi_reg_is_enabled, + .get_voltage_sel = scmi_reg_get_voltage_sel, + .set_voltage_sel = scmi_reg_set_voltage_sel, + .list_voltage = regulator_list_voltage_linear, + .map_voltage = regulator_map_voltage_linear, +}; + +static const struct regulator_ops scmi_reg_range_ops = { + .enable = scmi_reg_enable, + .disable = scmi_reg_disable, + .is_enabled = scmi_reg_is_enabled, + .get_voltage_sel = scmi_reg_get_voltage_sel, + .set_voltage_sel = scmi_reg_set_voltage_sel, + .list_voltage = regulator_list_voltage_linear_range, + .map_voltage = regulator_map_voltage_linear_range, +}; + +static const struct regulator_ops scmi_reg_discrete_ops = { + .enable = scmi_reg_enable, + .disable = scmi_reg_disable, + .is_enabled = scmi_reg_is_enabled, + .get_voltage_sel = scmi_reg_get_voltage_sel, + .set_voltage_sel = scmi_reg_set_voltage_sel, + .list_voltage = regulator_list_voltage_table, + .map_voltage = regulator_map_voltage_iterate, +}; + +static int +scmi_config_linear_regulator_mappings(struct scmi_regulator *sreg, + const struct scmi_voltage_info *vinfo) +{ + /* + * Note that SCMI voltage domains describable by linear ranges + * (segments) {low, high, step} are guaranteed to come in triplets by + * the SCMI Voltage Domain protocol support itself. + */ + if (vinfo->num_levels == 3) { + s32 delta_uV; + + delta_uV = (vinfo->levels_uv[SCMI_VOLTAGE_SEGMENT_HIGH] - + vinfo->levels_uv[SCMI_VOLTAGE_SEGMENT_LOW]); + /* Rule out buggy negative-intervals answers from fw */ + if (delta_uV < 0) { + dev_err(&sreg->sdev->dev, + "Invalid volt-range %d-%duV for domain %d\n", + vinfo->levels_uv[SCMI_VOLTAGE_SEGMENT_LOW], + vinfo->levels_uv[SCMI_VOLTAGE_SEGMENT_HIGH], + sreg->id); + return -EINVAL; + } + + if (!delta_uV) { + /* Just one fixed voltage exposed by SCMI */ + sreg->desc.fixed_uV = + vinfo->levels_uv[SCMI_VOLTAGE_SEGMENT_LOW]; + sreg->desc.n_voltages = 1; + sreg->desc.ops = &scmi_reg_fixed_ops; + } else { + /* One simple linear mapping. */ + sreg->desc.min_uV = + vinfo->levels_uv[SCMI_VOLTAGE_SEGMENT_LOW]; + sreg->desc.uV_step = + vinfo->levels_uv[SCMI_VOLTAGE_SEGMENT_STEP]; + sreg->desc.linear_min_sel = 0; + sreg->desc.n_voltages = delta_uV / sreg->desc.uV_step; + sreg->desc.ops = &scmi_reg_linear_ops; + } + } else { + /* Multiple linear mappings. */I don't see this multi-linear mapping in the v3.0 beta of the spec. Is it something planned?
No, it was indeed something included in previous spec iterations (multiple triplets) but now removed, as said in previous patch, I forgot to remove this redundant support. I'll do in V3.
quoted
+ int i, num_ranges, last_max = -1; + struct linear_range *lr; + + num_ranges = vinfo->num_levels / 3; + lr = devm_kcalloc(&sreg->sdev->dev, num_ranges, + sizeof(*lr), GFP_KERNEL); + if (!lr) + return -ENOMEM; + + sreg->desc.n_linear_ranges = num_ranges; + sreg->desc.linear_ranges = lr; + for (i = 0; num_ranges; num_ranges--, i += 3, lr++) { + s32 delta_uV; + + lr->min = + vinfo->levels_uv[i + SCMI_VOLTAGE_SEGMENT_LOW]; + lr->step = + vinfo->levels_uv[i + SCMI_VOLTAGE_SEGMENT_STEP]; + delta_uV = + vinfo->levels_uv[i + SCMI_VOLTAGE_SEGMENT_HIGH] - + lr->min; + if (delta_uV <= 0 || !(delta_uV / lr->step)) { + dev_err(&sreg->sdev->dev, + "Invalid volt-range %d-%duV for domain %d\n", + vinfo->levels_uv[SCMI_VOLTAGE_SEGMENT_LOW], + vinfo->levels_uv[SCMI_VOLTAGE_SEGMENT_HIGH], + sreg->id); + return -EINVAL; + } + lr->max_sel = delta_uV / lr->step - 1; + lr->min_sel = last_max + 1; + last_max = lr->max_sel; + } + sreg->desc.n_voltages = last_max + 1; + sreg->desc.ops = &scmi_reg_range_ops; + } + + return 0; +} + +static int +scmi_config_discrete_regulator_mappings(struct scmi_regulator *sreg, + const struct scmi_voltage_info *vinfo) +{ + /* Discrete non linear levels are mapped to volt_table */ + sreg->desc.n_voltages = vinfo->num_levels; + if (sreg->desc.n_voltages > 1) { + sreg->desc.volt_table = (const unsigned int *)vinfo->levels_uv; + sreg->desc.ops = &scmi_reg_discrete_ops; + } else { + sreg->desc.fixed_uV = vinfo->levels_uv[0]; + sreg->desc.ops = &scmi_reg_fixed_ops; + } + + return 0; +} + +static int scmi_regulator_common_init(struct scmi_regulator *sreg) +{ + int ret; + const struct scmi_handle *handle = sreg->sdev->handle; + struct device *dev = &sreg->sdev->dev; + const struct scmi_voltage_info *vinfo; + + vinfo = handle->voltage_ops->info_get(handle, sreg->id); + if (!vinfo) + return -ENODEV; + + if (!vinfo->num_levels) + return -EINVAL; + + /* + * Regulator framework does not fully support negative voltages + * so we discard any voltage domain reported as supporting negative + * voltages: as a consequence each levels_uv entry is guaranteed to + * be non-negative from here on. + */ + if (vinfo->negative_volts_allowed) { + dev_warn(dev, "Negative voltages NOT supported...skip %s\n", + sreg->of_node->full_name); + return -EOPNOTSUPP; + } + + sreg->desc.name = devm_kasprintf(dev, GFP_KERNEL, "%s", vinfo->name); + if (!sreg->desc.name) + return -ENOMEM; + + sreg->desc.id = sreg->id; + sreg->desc.type = REGULATOR_VOLTAGE; + sreg->desc.owner = THIS_MODULE; + sreg->desc.of_match = sreg->of_node->name; + sreg->desc.regulators_node = "regulators"; + if (vinfo->segmented) + ret = scmi_config_linear_regulator_mappings(sreg, vinfo); + else + ret = scmi_config_discrete_regulator_mappings(sreg, vinfo); + if (ret) + return ret; + + sreg->conf.dev = dev; + sreg->conf.driver_data = sreg; + + return 0; +} + +static int process_scmi_regulator_of_node(struct scmi_device *sdev, + struct device_node *np, + struct scmi_regulator_info *rinfo) +{ + u32 dom, ret; + + ret = of_property_read_u32(np, "reg", &dom); + if (ret) + return ret; + + if (dom >= rinfo->num_doms) + return -ENODEV; + + if (rinfo->sregv[dom]) { + dev_err(&sdev->dev, + "SCMI Voltage Domain %d already in use. Skipping: %s\n", + dom, np->full_name); + return -EINVAL; + } + + rinfo->sregv[dom] = devm_kzalloc(&sdev->dev, + sizeof(struct scmi_regulator), + GFP_KERNEL); + if (!rinfo->sregv[dom]) + return -ENOMEM; + + rinfo->sregv[dom]->id = dom; + rinfo->sregv[dom]->sdev = sdev; + /* get hold of good nodes */ + of_node_get(np); + rinfo->sregv[dom]->of_node = np; + dev_info(&sdev->dev, + "Found valid SCMI Regulator -- OF node [%d] -> %s\n", + dom, np->full_name); + + return ret;suggest return 0; here.
Ok.
quoted
+} + +static int scmi_regulator_probe(struct scmi_device *sdev) +{ + int d, ret, num_doms; + struct device_node *np, *child; + const struct scmi_handle *handle = sdev->handle; + struct scmi_regulator_info *rinfo; + + if (!handle || !handle->voltage_ops) + return -ENODEV; + + num_doms = handle->voltage_ops->num_domains_get(handle); + if (num_doms <= 0) { + if (!num_doms) { + dev_err(&sdev->dev, + "number of voltage domains invalid\n"); + num_doms = -EINVAL; + } else { + dev_err(&sdev->dev, + "failed to get voltage domains - err:%d\n", + num_doms); + } + + return num_doms; + } + + rinfo = devm_kzalloc(&sdev->dev, sizeof(*rinfo), GFP_KERNEL); + if (!rinfo) + return -ENOMEM; + + /* Allocate pointers' array for all possible domains */ + rinfo->sregv = devm_kcalloc(&sdev->dev, num_doms, + sizeof(rinfo->sregv), GFP_KERNEL); + if (!rinfo->sregv) + return -ENOMEM; + + rinfo->num_doms = num_doms; + /* + * Start collecting into rinfo->sregv possibly good SCMI Regulators as + * described by a well-formed DT entry and associated with an existing + * plausible SCMI Voltage Domain number, all belonging to this SCMI + * platform instance node (handle->dev->of_node). + */ + np = of_find_node_by_name(handle->dev->of_node, "regulators"); + for_each_child_of_node(np, child) { + ret = process_scmi_regulator_of_node(sdev, child, rinfo); + /* abort on any mem issue */ + if (ret == -ENOMEM) + return ret; + } + + /* + * Register a regulator for each valid regulator-DT-entry that we + * can successfully reach via SCMI + */ + for (d = 0; d < num_doms; d++) { + struct scmi_regulator *sreg = rinfo->sregv[d]; + + if (!sreg) + continue; + ret = scmi_regulator_common_init(sreg); + if (ret) + continue; + sreg->rdev = devm_regulator_register(&sdev->dev, &sreg->desc, + &sreg->conf); + if (IS_ERR(sreg->rdev)) { + sreg->rdev = NULL; + continue; + } + + dev_info(&sdev->dev, + "Regulator %s registered for domain [%d](%s)\n", + sreg->desc.name, sreg->id, sreg->desc.name);Maybe not display sreg->desc.name twice.
Yes does not make any sense indeed now that I have only one name o_O
quoted
+ } + + dev_set_drvdata(&sdev->dev, rinfo); + + return 0; +} + +static void scmi_regulator_remove(struct scmi_device *sdev) +{ + int d; + struct scmi_regulator_info *rinfo; + + rinfo = dev_get_drvdata(&sdev->dev); + if (!rinfo) + return; + + for (d = 0; d < rinfo->num_doms; d++) { + if (!rinfo->sregv[d]) + continue; + of_node_put(rinfo->sregv[d]->of_node); + } +} + +static const struct scmi_device_id scmi_regulator_id_table[] = { + { SCMI_PROTOCOL_VOLTAGE, "regulator" }, + { }, +}; +MODULE_DEVICE_TABLE(scmi, scmi_regulator_id_table); + +static struct scmi_driver scmi_drv = { + .name = "scmi-regulator", + .probe = scmi_regulator_probe, + .remove = scmi_regulator_remove, + .id_table = scmi_regulator_id_table, +}; + +module_scmi_driver(scmi_drv); + +MODULE_AUTHOR("Cristian Marussi [off-list ref]"); +MODULE_DESCRIPTION("ARM SCMI regulator driver"); +MODULE_LICENSE("GPL v2"); -- 2.17.1
Thanks Cristian _______________________________________________ linux-arm-kernel mailing list linux-arm-kernel@lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-arm-kernel