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
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20220426123904.GC25550@tellis.lin.mbt.kalray.eu>
Date:   Tue, 26 Apr 2022 14:39:05 +0200
From:   Jules Maselbas <jmaselbas@...ray.eu>
To:     Sean Anderson <sean.anderson@...o.com>
Cc:     linux-usb@...r.kernel.org, linux-phy@...ts.infradead.org,
        Vinod Koul <vkoul@...nel.org>, Mark Brown <broonie@...nel.org>,
        linux-kernel@...r.kernel.org, Felipe Balbi <balbi@...nel.org>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Arnd Bergmann <arnd@...db.de>
Subject: Re: [RFC PATCH] phy: Convert usb-nop-xceiv to the generic phy
 subsystem

On Mon, Apr 25, 2022 at 01:01:10PM -0400, Sean Anderson wrote:
> Hi Jules,
> 
> On 4/21/22 12:21 PM, Jules Maselbas wrote:
> > I recently encountered some limitations when using the usb-nop-xceiv
> > driver using the usb-phy subsystem. The situation is: I have a board
> > with two external usb phys (ULPI) that only needs to be reset by the
> > USB controller. However both phys shares the same reset line, which
> > will get toggled twice by the USB driver, once by controller.
> > 
> > This is what I observed with the current Linux usb-nop-xceiv driver:
> >           _______                            ___      _________
> >    RESET:        \__________________________/   \____/
> >                 1^                         2^  3^   4^
> >  1. and 2. first controller toggle the reset
> >  3. and 4. second controller toggle the reset
> > 
> > After the second toggle the first driver didn't expected the reset and
> > causing issues, but I don't recall what exactly, it is maybe a driver
> > that simply fail to get probed.
> 
> Ah, a tale as old as GPIO resets... (c.f. MDIO)
I didn't get it :]

> 
> > The following is what I observed when using the "usb-nop-xceiv"
> > driver in Barebox, which is a bit different than the Linux one.
> > 
> >            _______                            _________________
> >     RESET:        \__________________________/
> >                  1^                         2^
> > 
> > In this case the reset is only toggled once, this is because the phy
> > driver count the number of time it has called init and power_on, which
> > is exactly what is done by the generic phy subsystem in Linux, but not
> > by the "legacy" usb-nop-xceiv usb-phy driver.
> 
> Can you provide a snippet of your device tree? I am not quite sure how
> you are describing your phys. Do you use the same device for multiple USBs?
Yes I use the same phy "device" for multiple USBs, like so:

/ {
	usb_phy0: usb-phy0 {
		#phy-cells = <0>;
		compatible = "usb-nop-xceiv";
		reset-gpios = <&gpio1_banka 21 GPIO_ACTIVE_LOW>;
	};
};
&usb0 {
	dr_mode = "peripheral";
	phys = <&usb_phy0>;
	phy-names = "usb2-phy";
};
&usb1 {
	phys = <&usb_phy0>;
	phy-names = "usb2-phy";
};

> 
> > This difference in behavior is what motivated to "port" the existing
> > usb-nop-xceiv driver to the generic phy subsystem. However this is far
> > from being complete... I've only tested the reset function for this nop
> > phy, and it lack a lot of features from the original driver to be a
> > valid replacement for it.
> > 
> > I would like to know if there is an interest in converting the current
> > usb-nop-xceiv driver to the generic phy subsystem?
> > 
> > I would like to keep the usb-nop-xceiv compatible as it used by Barebox
> > and it works OK. If a new compatible is added then device-tree nodes
> > shouldn't have both, the new and usb-nop-xceiv, compatible because of
> > the specific check in phy-core.c which will always return ENODEV.
> > 
> > Signed-off-by: Jules Maselbas <jmaselbas@...ray.eu>
> > ---
> >  drivers/phy/Makefile          |   1 +
> >  drivers/phy/phy-core.c        |   4 -
> >  drivers/phy/phy-usb-generic.c | 241 ++++++++++++++++++++++++++++++++++
> >  3 files changed, 242 insertions(+), 4 deletions(-)
> >  create mode 100644 drivers/phy/phy-usb-generic.c
> 
> Since this is a conversion, I would expect that the old driver would be
> deleted. In particular, the Kconfig will need to be adjusted so that
> users of the old driver will have CONFIG_GENERIC_PHY enabled (and some
> reimplementation of usb_phy_gen_create_phy). I think we would still need
Of course, if this is a conversion i would also remove the old driver,
but I didn't wanted to cluter the diff... this is very much an RFC,
asking for advices.

> to call usb_add_phy_dev in probe() as well.
Ok, I'll look into that.

> Though I suppose this is why this patch is RFC?
Yes, this is why.

> 
> > 
> > diff --git a/drivers/phy/Makefile b/drivers/phy/Makefile
> > index 6eb2916773c5..8eeecf6d6f74 100644
> > --- a/drivers/phy/Makefile
> > +++ b/drivers/phy/Makefile
> > @@ -4,6 +4,7 @@
> >  #
> >  
> >  obj-$(CONFIG_GENERIC_PHY)		+= phy-core.o
> > +obj-$(CONFIG_GENERIC_PHY)		+= phy-usb-generic.o
> >  obj-$(CONFIG_GENERIC_PHY_MIPI_DPHY)	+= phy-core-mipi-dphy.o
> >  obj-$(CONFIG_PHY_LPC18XX_USB_OTG)	+= phy-lpc18xx-usb-otg.o
> >  obj-$(CONFIG_PHY_XGENE)			+= phy-xgene.o
> > diff --git a/drivers/phy/phy-core.c b/drivers/phy/phy-core.c
> > index 71cb10826326..e09fd1cb4455 100644
> > --- a/drivers/phy/phy-core.c
> > +++ b/drivers/phy/phy-core.c
> > @@ -506,10 +506,6 @@ static struct phy *_of_phy_get(struct device_node *np, int index)
> >  	if (ret)
> >  		return ERR_PTR(-ENODEV);
> >  
> > -	/* This phy type handled by the usb-phy subsystem for now */
> > -	if (of_device_is_compatible(args.np, "usb-nop-xceiv"))
> > -		return ERR_PTR(-ENODEV);
> > -
> >  	mutex_lock(&phy_provider_mutex);
> >  	phy_provider = of_phy_provider_lookup(args.np);
> >  	if (IS_ERR(phy_provider) || !try_module_get(phy_provider->owner)) {
> > diff --git a/drivers/phy/phy-usb-generic.c b/drivers/phy/phy-usb-generic.c
> > new file mode 100644
> > index 000000000000..fdd05bbd98f2
> > --- /dev/null
> > +++ b/drivers/phy/phy-usb-generic.c
> > @@ -0,0 +1,241 @@
> > +// SPDX-License-Identifier: GPL-2.0-or-later
> > +/*
> > + * derived from driver/usb/phy/phy-generic.c
> > + *
> > + * Generic USB PHY driver for all USB "nop" transceiver which are mostly
> > + * autonomous. This driver is derived from the usb-nop-xceiv driver, but
> > + * this one use the new generic phy api.
> > + *
> > + * Copyright (c) 2022 Kalray Inc.
> > + * Author(s): Jules Maselbas
> > + */
> > +
> > +#include <linux/clk.h>
> > +#include <linux/err.h>
> > +#include <linux/module.h>
> > +#include <linux/of.h>
> > +#include <linux/phy/phy.h>
> > +#include <linux/platform_device.h>
> > +#include <linux/regulator/consumer.h>
> > +#include <linux/gpio/consumer.h>
> > +#include <linux/delay.h>
> > +#include <linux/usb/otg.h>
> > +
> > +struct phy_usb_generic {
> > +	struct device *dev;
> > +	struct clk *clk;
> > +	struct regulator *vcc;
> > +	struct gpio_desc *gpiod_reset;
> > +	unsigned long mA;
> > +	unsigned int vbus;
> 
> These are unused
Indeed

> > +	enum usb_dr_mode dr_mode;
> > +};
> > +
> > +static int phy_usb_generic_init(struct phy *phy)
> > +{
> > +	struct phy_usb_generic *priv = phy_get_drvdata(phy);
> > +
> > +	if (priv->clk)
> > +		clk_prepare(priv->clk);
> > +
> > +	return 0;
> > +}
> > +
> > +static int phy_usb_generic_exit(struct phy *phy)
> > +{
> > +	struct phy_usb_generic *priv = phy_get_drvdata(phy);
> > +
> > +	if (priv->clk)
> > +		clk_unprepare(priv->clk);
> > +
> > +	return 0;
> > +}
> > +
> > +static int phy_usb_generic_power_on(struct phy *phy)
> > +{
> > +	struct phy_usb_generic *priv = phy_get_drvdata(phy);
> > +	int ret;
> > +
> > +	if (priv->vcc) {
> > +		ret = regulator_enable(priv->vcc);
> > +		if (ret) {
> > +			dev_err(priv->dev, "Failed to enable power\n");
> > +			return ret;
> > +		}
> > +	}
> > +
> > +	if (priv->clk) {
> > +		ret = clk_enable(priv->clk);
> > +		if (ret) {
> > +			dev_err(priv->dev, "Failed to enable clock\n");
> > +			return ret;
> > +		}
> > +	}
> > +
> > +	if (priv->gpiod_reset) {
> > +		dev_dbg(priv->dev, "Reset toggle\n");
> > +		gpiod_set_value_cansleep(priv->gpiod_reset, 1);
> > +		usleep_range(10000, 20000);
> > +		gpiod_set_value_cansleep(priv->gpiod_reset, 0);
> > +	}
> > +
> > +	return 0;
> > +}
> > +
> > +static int phy_usb_generic_power_off(struct phy *phy)
> > +{
> > +	struct phy_usb_generic *priv = phy_get_drvdata(phy);
> > +	int ret;
> > +
> > +	if (priv->gpiod_reset)
> > +		gpiod_set_value_cansleep(priv->gpiod_reset, 1);
> > +
> > +	if (priv->clk)
> > +		clk_disable_unprepare(priv->clk);
> > +
> > +	if (priv->vcc) {
> > +		ret = regulator_disable(priv->vcc);
> > +		if (ret) {
> > +			dev_err(priv->dev, "Failed to disable power\n");
> > +			return ret;
> > +		}
> > +	}
> > +
> > +	return 0;
> > +}
> > +
> > +static int phy_usb_generic_set_mode(struct phy *phy,
> > +				    enum phy_mode mode, int submode)
> > +{
> > +	struct phy_usb_generic *priv = phy_get_drvdata(phy);
> > +	enum usb_dr_mode new_mode;
> > +	const char *s = "";
> > +
> > +	switch (mode) {
> > +	case PHY_MODE_USB_HOST:
> > +	case PHY_MODE_USB_HOST_LS:
> > +	case PHY_MODE_USB_HOST_HS:
> > +	case PHY_MODE_USB_HOST_FS:
> > +		new_mode = USB_DR_MODE_HOST;
> > +		s = "host";
> > +		break;
> > +	case PHY_MODE_USB_DEVICE:
> > +	case PHY_MODE_USB_DEVICE_LS:
> > +	case PHY_MODE_USB_DEVICE_HS:
> > +	case PHY_MODE_USB_DEVICE_FS:
> > +		new_mode = USB_DR_MODE_PERIPHERAL;
> > +		s = "peripheral";
> > +		break;
> > +	case PHY_MODE_USB_OTG:
> > +		new_mode = USB_DR_MODE_OTG;
> > +		s = "otg";
> > +		break;
> > +	default:
> > +		return -EINVAL;
> > +	}
> > +
> > +	if (new_mode != priv->dr_mode) {
> > +		dev_info(priv->dev, "Changing dr_mode to %s\n", s);
> > +		priv->dr_mode = new_mode;
> > +	}
> > +
> > +	return 0;
> > +}
> > +
> > +static const struct phy_ops phy_usb_generic_ops = {
> > +	.init = phy_usb_generic_init,
> > +	.exit = phy_usb_generic_exit,
> > +	.power_on = phy_usb_generic_power_on,
> > +	.power_off = phy_usb_generic_power_off,
> > +	.set_mode  = phy_usb_generic_set_mode,
> > +	.owner = THIS_MODULE,
> > +};
> > +
> > +static int phy_usb_generic_probe(struct platform_device *pdev)
> > +{
> > +	struct device *dev = &pdev->dev;
> > +	struct device_node *np = dev->of_node;
> > +	struct phy_usb_generic *priv;
> > +	struct phy_provider *provider;
> > +	struct phy *phy;
> > +	int err;
> > +	u32 clk_rate = 0;
> > +
> > +	priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
> > +	if (!priv)
> > +		return -ENOMEM;
> > +	priv->dev = &pdev->dev;
> > +	platform_set_drvdata(pdev, priv);
> > +
> > +	priv->gpiod_reset = devm_gpiod_get_optional(dev, "reset", GPIOD_ASIS);
> > +	if (IS_ERR(priv->gpiod_reset)) {
> > +		err = PTR_ERR(priv->gpiod_reset);
> > +		dev_err(dev, "Failed to get reset gpio: %d\n", err);
> 
> These should be dev_err_probe to avoid noise when we get -EPROBE_DEFER. Ditto
> for the rest of the dev_errs.
> 
Thanks, TIL dev_err_probe! I'll modify this

> > +		return err;
> > +	}
> > +
> > +	if (np && of_property_read_u32(np, "clock-frequency", &clk_rate))
> > +		clk_rate = 0;
> > +
> > +	priv->clk = devm_clk_get_optional(dev, "main_clk");
> > +	if (IS_ERR(priv->clk)) {
> > +		err = PTR_ERR(priv->clk);
> > +		dev_err(dev, "Can't get phy clock: %d\n", err);
> > +		return err;
> > +	}
> > +	if (priv->clk && clk_rate) {
> > +		err = clk_set_rate(priv->clk, clk_rate);
> > +		if (err) {
> > +			dev_err(dev, "Error setting clock rate\n");
> > +			return err;
> > +		}
> > +	}
> > +
> > +	priv->vcc = devm_regulator_get_optional(dev, "vcc");
> > +	if (IS_ERR(priv->vcc)) {
> > +		err = PTR_ERR(priv->vcc);
> > +		if (err == -ENODEV || err == -ENOENT) {
> > +			priv->vcc = NULL;
> > +		} else {
> > +			dev_err(dev, "Error getting vcc regulator: %d\n", err);
> > +			return err;
> > +		}
> > +	}
> > +
> > +	phy = devm_phy_create(dev, NULL, &phy_usb_generic_ops);
> > +	if (IS_ERR(phy)) {
> > +		err = PTR_ERR(phy);
> > +		dev_err(dev, "Failed to create PHY: %d\n", err);
> > +		return err;
> > +	}
> > +	phy_set_drvdata(phy, priv);
> > +
> > +	provider = devm_of_phy_provider_register(dev, of_phy_simple_xlate);
> > +	if (IS_ERR(provider)) {
> > +		err = PTR_ERR(provider);
> > +		dev_err(dev, "Failed to register PHY provider: %d\n", err);
> > +		return err;
> > +	}
> > +
> > +	return 0;
> > +}
> > +
> > +static const struct of_device_id phy_usb_nop_dt_ids[] = {
> > +	{ .compatible = "phy-usb-generic" },
> > +	{ .compatible = "usb-nop-xceiv" },
> > +	{ }
> > +};
> > +MODULE_DEVICE_TABLE(of, phy_usb_nop_dt_ids);
> > +
> > +static struct platform_driver phy_usb_generic_driver = {
> > +	.probe		= phy_usb_generic_probe,
> > +	.driver		= {
> > +		.name	= "phy_usb_generic",
> > +		.of_match_table = phy_usb_nop_dt_ids,
> > +	},
> > +};
> > +module_platform_driver(phy_usb_generic_driver);
> > +
> > +MODULE_AUTHOR("Kalray Inc");
> > +MODULE_DESCRIPTION("Generic USB PHY driver");
> > +MODULE_LICENSE("GPL");
> > 
> 
> The rest looks good (modulo missing features).
Yeah, I feel also there is a lot of missing features from the
usb-nop-xceiv driver. I doesn't need much more than a simple gpio
hence it is difficult for me to test porting other features.

Thanks,
-- Jules





Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ