lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
Open Source and information security mailing list archives
| ||
|
Message-ID: <3ee22b0e-2010-1cf6-bca8-3c803d73d7f9@sholland.org> Date: Fri, 25 Nov 2022 21:03:24 -0600 From: Samuel Holland <samuel@...lland.org> To: Andre Przywara <andre.przywara@....com> Cc: Liam Girdwood <lgirdwood@...il.com>, Mark Brown <broonie@...nel.org>, Chen-Yu Tsai <wens@...e.org>, Jernej Skrabec <jernej.skrabec@...il.com>, Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>, Rob Herring <robh+dt@...nel.org>, Andrew Lunn <andrew@...n.ch>, Heiko Stuebner <heiko@...ech.de>, Maxime Ripard <mripard@...nel.org>, devicetree@...r.kernel.org, linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org, linux-sunxi@...ts.linux.dev Subject: Re: [PATCH v4 2/4] regulator: sun20i: Add Allwinner D1 LDOs driver Hi Andre, On 11/25/22 18:22, Andre Przywara wrote: > On Thu, 24 Nov 2022 22:01:10 -0600 > Samuel Holland <samuel@...lland.org> wrote: > > Hi Samuel, > >> D1 contains two pairs of LDOs, "analog" LDOs and "system" LDOs. They are >> similar and can share a driver, but only the system LDOs have a DT >> binding defined so far. >> >> The system LDOs have a single linear range. The voltage step is not an >> integer, so a custom .list_voltage is needed to get the rounding right. >> >> Signed-off-by: Samuel Holland <samuel@...lland.org> >> --- >> >> Changes in v4: >> - Drop the analog LDOs until the codec binding is ready >> >> Changes in v3: >> - Adjust control flow in sun20i_regulator_get_regmap() for clarity >> >> Changes in v2: >> - Use decimal numbers for .n_voltages instead of field widths >> - Get the regmap from the parent device instead of a property/phandle >> >> drivers/regulator/Kconfig | 8 ++ >> drivers/regulator/Makefile | 1 + >> drivers/regulator/sun20i-regulator.c | 150 +++++++++++++++++++++++++++ >> 3 files changed, 159 insertions(+) >> create mode 100644 drivers/regulator/sun20i-regulator.c >> >> diff --git a/drivers/regulator/Kconfig b/drivers/regulator/Kconfig >> index 070e4403c6c2..8480532114c1 100644 >> --- a/drivers/regulator/Kconfig >> +++ b/drivers/regulator/Kconfig >> @@ -1280,6 +1280,14 @@ config REGULATOR_STW481X_VMMC >> This driver supports the internal VMMC regulator in the STw481x >> PMIC chips. >> >> +config REGULATOR_SUN20I >> + tristate "Allwinner D1 internal LDOs" >> + depends on ARCH_SUNXI || COMPILE_TEST >> + select MFD_SYSCON >> + default ARCH_SUNXI >> + help >> + This driver supports the internal LDOs in the Allwinner D1 SoC. >> + >> config REGULATOR_SY7636A >> tristate "Silergy SY7636A voltage regulator" >> depends on MFD_SY7636A >> diff --git a/drivers/regulator/Makefile b/drivers/regulator/Makefile >> index 5962307e1130..8e9b5a21123d 100644 >> --- a/drivers/regulator/Makefile >> +++ b/drivers/regulator/Makefile >> @@ -150,6 +150,7 @@ obj-$(CONFIG_REGULATOR_STM32_VREFBUF) += stm32-vrefbuf.o >> obj-$(CONFIG_REGULATOR_STM32_PWR) += stm32-pwr.o >> obj-$(CONFIG_REGULATOR_STPMIC1) += stpmic1_regulator.o >> obj-$(CONFIG_REGULATOR_STW481X_VMMC) += stw481x-vmmc.o >> +obj-$(CONFIG_REGULATOR_SUN20I) += sun20i-regulator.o >> obj-$(CONFIG_REGULATOR_SY7636A) += sy7636a-regulator.o >> obj-$(CONFIG_REGULATOR_SY8106A) += sy8106a-regulator.o >> obj-$(CONFIG_REGULATOR_SY8824X) += sy8824x.o >> diff --git a/drivers/regulator/sun20i-regulator.c b/drivers/regulator/sun20i-regulator.c >> new file mode 100644 >> index 000000000000..031bcc3dee50 >> --- /dev/null >> +++ b/drivers/regulator/sun20i-regulator.c >> @@ -0,0 +1,150 @@ >> +// SPDX-License-Identifier: GPL-2.0-only >> +// >> +// Copyright (c) 2021-2022 Samuel Holland <samuel@...lland.org> >> +// >> + >> +#include <linux/mfd/syscon.h> >> +#include <linux/module.h> >> +#include <linux/of_device.h> >> +#include <linux/platform_device.h> >> +#include <linux/regmap.h> >> +#include <linux/regulator/driver.h> >> + >> +#define SUN20I_SYS_LDO_CTRL_REG 0x150 >> + >> +struct sun20i_regulator_data { >> + const struct regulator_desc *descs; >> + unsigned int ndescs; >> +}; >> + >> +/* regulator_list_voltage_linear() modified for the non-integral uV_step. */ >> +static int sun20i_d1_system_ldo_list_voltage(struct regulator_dev *rdev, >> + unsigned int selector) >> +{ >> + const struct regulator_desc *desc = rdev->desc; >> + unsigned int uV; >> + >> + if (selector >= desc->n_voltages) >> + return -EINVAL; >> + >> + uV = desc->min_uV + (desc->uV_step * selector); >> + >> + /* Produce correctly-rounded absolute voltages. */ >> + return uV + ((selector + 1 + (desc->min_uV % 4)) / 3); >> +} >> + >> +static const struct regulator_ops sun20i_d1_system_ldo_ops = { >> + .list_voltage = sun20i_d1_system_ldo_list_voltage, >> + .map_voltage = regulator_map_voltage_ascend, >> + .set_voltage_sel = regulator_set_voltage_sel_regmap, >> + .get_voltage_sel = regulator_get_voltage_sel_regmap, >> +}; >> + >> +static const struct regulator_desc sun20i_d1_system_ldo_descs[] = { >> + { >> + .name = "ldoa", >> + .supply_name = "ldo-in", >> + .of_match = "ldoa", >> + .ops = &sun20i_d1_system_ldo_ops, >> + .type = REGULATOR_VOLTAGE, >> + .owner = THIS_MODULE, >> + .n_voltages = 32, >> + .min_uV = 1600000, >> + .uV_step = 13333, /* repeating */ > > So while I see that those values are probably the closest we can with a > simple linear algorithm, they first two values seem to be slightly off > from those values in the manual: > sel diff algor manual > 0: 7 (1.600 - 1.593) > 1: 6 (1.613 - 1.607) > Oddly enough the rest of the values are spot on. > I don't know if this really matters, or if the LDOs are actually > accurate enough to that level of precision, or if it's a manual bug, or > if we really care at all, but it might warrant some comment, I guess? > I just got triggered by the min value not being the first value in the > list. Whoops, I did the math based on the default 1.8 V and after some spot checks assumed the range was linear. I should have noticed this. Conveniently, LDOB is unused on most boards, so I can measure the range directly. The INA219 I'm using gives me a 4 mV LSB in its most precise mode, which I'm averaging to get 2 mV precision. reg measure manual diff ============================ 0x00: 1.160 1.167 -7 0x08: 1.266 1.273 -7 0x09: 1.280 1.287 -7 0x0a: 1.292 1.300 -8 0x0b: 1.306 1.313 -7 0x0c: 1.320 1.327 -7 0x0d: 1.332 1.340 -8 0x0e: 1.346 1.353 -7 0x0f: 1.360 1.367 -7 0x10: 1.372 1.380 -8 0x18: 1.480 1.487 -7 0x1c: 1.532 1.540 -8 0x1d: 1.546 1.553 -7 0x1e: 1.560 1.567 -7 0x1f: 1.572 1.580 -8 0x20: 1.586 1.593 -7 << 0x21: 1.600 1.607 -7 << 0x22: 1.620 1.627 -7 << 0x23: 1.632 1.640 -8 << 0x24: 1.646 1.653 -7 0x25: 1.660 1.667 -7 0x26: 1.672 1.680 -8 0x27: 1.686 1.693 -7 0x28: 1.700 1.707 -7 0x30: 1.806 1.813 -7 0x38: 1.912 1.920 -8 0x3e: 1.992 2.000 -8 0x3f: 2.006 2.013 -7 So it looks like the LDOs are accurate, and the manual is quite correct. I'm willing to blame the constant offset on my crude test setup. >> + .vsel_reg = SUN20I_SYS_LDO_CTRL_REG, >> + .vsel_mask = GENMASK(7, 0), >> + }, >> + { >> + .name = "ldob", >> + .supply_name = "ldo-in", >> + .of_match = "ldob", >> + .ops = &sun20i_d1_system_ldo_ops, >> + .type = REGULATOR_VOLTAGE, >> + .owner = THIS_MODULE, >> + .n_voltages = 64, >> + .min_uV = 1166666, >> + .uV_step = 13333, /* repeating */ > > For LDOB it seems to be worse, as the second half is constantly off by > what looks like 6.666mV: > sel diff algor manual > ... > 32: 0 (1.593 - 1.593) > 33: 0 (1.607 - 1.607) > 34: -7 (1.620 - 1.627) > 35: -7 (1.633 - 1.64) > 36: -6 (1.647 - 1.653) > ... > 63: -6 (2.007 - 2.013) > The first half is correct, though. Closer inspection reveals that > everything with bit 5 set is exactly the same as LDOA. Maybe we can use > that to our advantage? Since I already have a custom .list_voltage implementation, I think the simplest solution is: if (uV >= 1620000) uV += 7000; or similar to get the rounding right. Regards, Samuel > Cheers, > Andre > >> + .vsel_reg = SUN20I_SYS_LDO_CTRL_REG, >> + .vsel_mask = GENMASK(15, 8), >> + }, >> +}; >> + >> +static const struct sun20i_regulator_data sun20i_d1_system_ldos = { >> + .descs = sun20i_d1_system_ldo_descs, >> + .ndescs = ARRAY_SIZE(sun20i_d1_system_ldo_descs), >> +}; >> + >> +static struct regmap *sun20i_regulator_get_regmap(struct device *dev) >> +{ >> + struct regmap *regmap; >> + >> + /* >> + * First try the syscon interface. The system control device is not >> + * compatible with "syscon", so fall back to getting the regmap from >> + * its platform device. This is ugly, but required for devicetree >> + * backward compatibility. >> + */ >> + regmap = syscon_node_to_regmap(dev->parent->of_node); >> + if (!IS_ERR(regmap)) >> + return regmap; >> + >> + regmap = dev_get_regmap(dev->parent, NULL); >> + if (regmap) >> + return regmap; >> + >> + return ERR_PTR(-EPROBE_DEFER); >> +} >> + >> +static int sun20i_regulator_probe(struct platform_device *pdev) >> +{ >> + const struct sun20i_regulator_data *data; >> + struct device *dev = &pdev->dev; >> + struct regulator_config config; >> + struct regmap *regmap; >> + >> + data = of_device_get_match_data(dev); >> + if (!data) >> + return -EINVAL; >> + >> + regmap = sun20i_regulator_get_regmap(dev); >> + if (IS_ERR(regmap)) >> + return dev_err_probe(dev, PTR_ERR(regmap), "Failed to get regmap\n"); >> + >> + config = (struct regulator_config) { >> + .dev = dev, >> + .regmap = regmap, >> + }; >> + >> + for (unsigned int i = 0; i < data->ndescs; ++i) { >> + const struct regulator_desc *desc = &data->descs[i]; >> + struct regulator_dev *rdev; >> + >> + rdev = devm_regulator_register(dev, desc, &config); >> + if (IS_ERR(rdev)) >> + return PTR_ERR(rdev); >> + } >> + >> + return 0; >> +} >> + >> +static const struct of_device_id sun20i_regulator_of_match[] = { >> + { >> + .compatible = "allwinner,sun20i-d1-system-ldos", >> + .data = &sun20i_d1_system_ldos, >> + }, >> + { }, >> +}; >> +MODULE_DEVICE_TABLE(of, sun20i_regulator_of_match); >> + >> +static struct platform_driver sun20i_regulator_driver = { >> + .probe = sun20i_regulator_probe, >> + .driver = { >> + .name = "sun20i-regulator", >> + .of_match_table = sun20i_regulator_of_match, >> + }, >> +}; >> +module_platform_driver(sun20i_regulator_driver); >> + >> +MODULE_AUTHOR("Samuel Holland <samuel@...lland.org>"); >> +MODULE_DESCRIPTION("Allwinner D1 internal LDO driver"); >> +MODULE_LICENSE("GPL"); >
Powered by blists - more mailing lists