Skip to contentSkip to conversation

mv-usb phy enhancement patches

PATCH New Chao Xie 6 participants 34 messages Feb 21, 2013 – Mar 7, 2013

Subsystems (5)
lore
Share & tools
lore

Chao Xie

author
#0 lore raw

v8 cover mv-usb phy enhancement patches

The patches are divied into 2 parts

1. PHY driver
To remove the callbacks in the platform data, a usb PHY driver for marvell udc/otg/ehci is written.
For device tree support, it is not good to pass the callback pointers by platform data. The PHY driver also removes the block.

usb: phy: mv_usb2: add PHY driver for marvell usb2 controller
usb: gadget: mv_udc: use PHY driver for udc
usb: ehci: ehci-mv: use PHY driver for ehci
usb: otg: mv_otg: use PHY driver for otg
Above patches are marvell usb PHY driver support.

arm: mmp2: change the defintion of usb devices
arm: pxa910: change the defintion of usb devices
arm: brownstone: add usb support for the board
arm: ttc_dkb: add usb support
arm: mmp: remove the usb phy setting
arm: mmp: remove usb devices from pxa168
Above patches are for SOC/board support for marvell usb PHY driver.

2. external chip support
The marvell usb controller can detect the vbus/idpin, but it need PHY and usb clocks to be enabled.
Based on measurement it will import 15mA current, and increase the power when the usb is not used.
Using a external chip to detect vbus/idpin changes will save the power.
In fact the marvell PMIC 88pm860x and 88pm80x can do it. The drivers are located at drivers/mfd.
So add a middle layer in the marvell usb PHY driver. PMIC call the APIs in middle driver to registers the callback for vbus/idpin detection/query
udc/otg/ehci driver will call the APIs to get vbus/idpin changes and query the states of the vbus/idpin.
usb: phy: mv_usb2_phy: add externel chip support
usb: gadget: mv_udc: add extern chip support
usb: ehci: ehci-mv: add extern chip support
usb: otg: mv_otg: add extern chip support
Above patches are the middle layer suppor for udc/otg/ehci

arm: mmp: add extern chip support for brownstone
arm: mmp: add extern chip support for ttc_dkb
Above patches are corresponding board file changes

V2->V1:
Change the Signed-off-by to be right email address

v3->v2
re-format the patches to new code base

v4->v3
1. make mv udc gadget driver depend on ARCH_PXA and ARCH_MMP
2. remove __devinit and __devexit
3. make the driver compiled successful if CONFIG_MV_USB2_PHY is not defined. The modified patches are
usb: gadget: mv_udc: make mv_udc depends on ARCH_MMP or ARCH_PXA
usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

v5->v4
make the struct mv_usb2_extern_chip member ->head to be struct atomic_notifier_head instead of struct atomic_notifier_head *.

v6->v5
the bug fix patches are merged.
Removed the dependcy of ARCH_MMP and ARCH_PXA, and make the driver can be compiled for x86.
The device tree support patches need modification, remove them from this patch series, and they will be submitted in another series.

v7->v6
Use usb_add_phy_dev and related APIs to add PHY drivers. Removed the device tree support in PHY driver. It will be added in another patch series.

v8->v7
bugs fix in phy driver.
directly use devm_usb_get_phy_dev return value for error return.

Chao Xie (16):
usb: phy: mv_usb2: add PHY driver for marvell usb2 controller
usb: gadget: mv_udc: use PHY driver for udc
usb: ehci: ehci-mv: use PHY driver for ehci
usb: otg: mv_otg: use PHY driver for otg
arm: mmp2: change the defintion of usb devices
arm: pxa910: change the defintion of usb devices
arm: brownstone: add usb support for the board
arm: ttc_dkb: add usb support
arm: mmp: remove the usb phy setting
arm: mmp: remove usb devices from pxa168
usb: phy: mv_usb2_phy: add externel chip support
usb: gadget: mv_udc: add extern chip support
usb: ehci: ehci-mv: add extern chip support
usb: otg: mv_otg: add extern chip support
arm: mmp: add extern chip support for brownstone
arm: mmp: add extern chip support for ttc_dkb

 arch/arm/mach-mmp/brownstone.c            |   66 +++++
 arch/arm/mach-mmp/devices.c               |  278 ------------------
 arch/arm/mach-mmp/include/mach/mmp2.h     |    4 +
 arch/arm/mach-mmp/include/mach/pxa910.h   |    7 +-
 arch/arm/mach-mmp/include/mach/regs-usb.h |  253 ----------------
 arch/arm/mach-mmp/mmp2.c                  |    4 +
 arch/arm/mach-mmp/pxa168.c                |   42 ---
 arch/arm/mach-mmp/pxa910.c                |    4 +
 arch/arm/mach-mmp/ttc_dkb.c               |   52 +++-
 drivers/usb/gadget/mv_udc.h               |    5 +-
 drivers/usb/gadget/mv_udc_core.c          |  101 ++++---
 drivers/usb/host/ehci-mv.c                |   64 ++---
 drivers/usb/otg/mv_otg.c                  |  122 ++++----
 drivers/usb/otg/mv_otg.h                  |    5 +-
 drivers/usb/phy/Kconfig                   |    7 +
 drivers/usb/phy/Makefile                  |    1 +
 drivers/usb/phy/mv_usb2_phy.c             |  451 +++++++++++++++++++++++++++++
 include/linux/platform_data/mv_usb.h      |   22 +-
 include/linux/usb/mv_usb2.h               |  121 ++++++++
 19 files changed, 856 insertions(+), 753 deletions(-)
 delete mode 100644 arch/arm/mach-mmp/include/mach/regs-usb.h
 create mode 100644 drivers/usb/phy/mv_usb2_phy.c
 create mode 100644 include/linux/usb/mv_usb2.h
Signature · 1 line
1.7.4.1
author
#1 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

The PHY is seperated from usb controller.
The usb controller used in marvell pxa168/pxa910/mmp2 are same, but PHY initialization may be different.
the usb controller can support udc/otg/ehci, and for each of the mode, it need PHY to initialized before use the controller. Direclty writing the phy driver will make the usb controller driver to be simple and portable.
The PHY driver will be used by marvell udc/otg/ehci.

---
 drivers/usb/phy/Kconfig              |    7 +
 drivers/usb/phy/Makefile             |    1 +
 drivers/usb/phy/mv_usb2_phy.c        |  402 ++++++++++++++++++++++++++++++++++
 include/linux/platform_data/mv_usb.h |    9 +-
 include/linux/usb/mv_usb2.h          |   32 +++
 5 files changed, 448 insertions(+), 3 deletions(-)
 create mode 100644 drivers/usb/phy/mv_usb2_phy.c
 create mode 100644 include/linux/usb/mv_usb2.h
drivers/usb/phy/Kconfig +7 -0
index 65217a5..5479760 100644
@@ -73,3 +73,10 @@ config SAMSUNG_USBPHY
 	help
 	  Enable this to support Samsung USB phy controller for samsung
 	  SoCs.
+
+config MV_USB2_PHY
+	tristate "Marvell USB 2.0 PHY Driver"
+	depends on USB || USB_GADGET
+	help
+	  Enable this to support Marvell USB 2.0 phy driver for Marvell
+	  SoC.
drivers/usb/phy/Makefile +1 -0
index b13faa1..3835316 100644
@@ -12,3 +12,4 @@ obj-$(CONFIG_MV_U3D_PHY)		+= mv_u3d_phy.o
 obj-$(CONFIG_USB_EHCI_TEGRA)	+= tegra_usb_phy.o
 obj-$(CONFIG_USB_RCAR_PHY)		+= rcar-phy.o
 obj-$(CONFIG_SAMSUNG_USBPHY)		+= samsung-usbphy.o
+obj-$(CONFIG_MV_USB2_PHY)		+= mv_usb2_phy.o
drivers/usb/phy/mv_usb2_phy.c new file +402 -0

402 lines not on this page · raw message

include/linux/platform_data/mv_usb.h +6 -3
index 944b01d..fd3d1b4 100644
@@ -47,9 +47,12 @@ struct mv_usb_platform_data {
 	/* Force a_bus_req to be asserted */
 	 unsigned int    otg_force_a_bus_req:1;
 
-	int	(*phy_init)(void __iomem *regbase);
-	void	(*phy_deinit)(void __iomem *regbase);
 	int	(*set_vbus)(unsigned int vbus);
-	int     (*private_init)(void __iomem *opregs, void __iomem *phyregs);
 };
+
+struct mv_usb_phy_platform_data {
+	unsigned int	clknum;
+	char		**clkname;
+};
+
 #endif
include/linux/usb/mv_usb2.h new file +32 -0
new file mode 100644
index 0000000..77250f5
@@ -0,0 +1,32 @@
+/*
+ * Copyright (C) 2013 Marvell Inc.
+ *
+ * This software is licensed under the terms of the GNU General Public
+ * License version 2, as published by the Free Software Foundation, and
+ * may be copied, distributed, and modified under those terms.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ */
+
+#ifndef __MV_USB2_H
+#define __MV_USB2_H
+
+#define MV_USB2_PHY_INDEX	0
+#define MV_USB2_OTG_PHY_INDEX	1
+
+struct mv_usb2_phy {
+	struct usb_phy		phy;
+	struct platform_device	*pdev;
+	struct mutex		phy_lock;
+	unsigned int		refcount;
+	unsigned int		type;
+	void __iomem		*base;
+	struct clk		**clks;
+	unsigned int		clks_num;
+};
+
+#endif
Signature · 1 line
1.7.4.1
#2 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

Quoted 25 lines · Chao Xie

The PHY is seperated from usb controller.
The usb controller used in marvell pxa168/pxa910/mmp2 are same, but PHY initialization may be different.
the usb controller can support udc/otg/ehci, and for each of the mode, it need PHY to initialized before use the controller. Direclty writing the phy driver will make the usb controller driver to be simple and portable.
The PHY driver will be used by marvell udc/otg/ehci.

---
 drivers/usb/phy/Kconfig              |    7 +
 drivers/usb/phy/Makefile             |    1 +
 drivers/usb/phy/mv_usb2_phy.c        |  402 ++++++++++++++++++++++++++++++++++
 include/linux/platform_data/mv_usb.h |    9 +-
 include/linux/usb/mv_usb2.h          |   32 +++
 5 files changed, 448 insertions(+), 3 deletions(-)
 create mode 100644 drivers/usb/phy/mv_usb2_phy.c
 create mode 100644 include/linux/usb/mv_usb2.h
drivers/usb/phy/Kconfig +0 -0
index 65217a5..5479760 100644

drivers/usb/phy/Kconfig config SAMSUNG_USBPHY · line 79

Quoted 7 lines
drivers/usb/phy/Kconfig +4 -0
@@ -73,3 +73,10 @@ config SAMSUNG_USBPHY
 	help
 	  Enable this to support Samsung USB phy controller for samsung
 	  SoCs.
+
+config MV_USB2_PHY
+	tristate "Marvell USB 2.0 PHY Driver"
+	depends on USB || USB_GADGET

NAK, PHYs don't depend on the gadget framework.

Quoted 57 lines
drivers/usb/phy/Kconfig +3 -0
+	help
+	  Enable this to support Marvell USB 2.0 phy driver for Marvell
+	  SoC.
drivers/usb/phy/Makefile +1 -0
index b13faa1..3835316 100644
@@ -12,3 +12,4 @@ obj-$(CONFIG_MV_U3D_PHY)		+= mv_u3d_phy.o
 obj-$(CONFIG_USB_EHCI_TEGRA)	+= tegra_usb_phy.o
 obj-$(CONFIG_USB_RCAR_PHY)		+= rcar-phy.o
 obj-$(CONFIG_SAMSUNG_USBPHY)		+= samsung-usbphy.o
+obj-$(CONFIG_MV_USB2_PHY)		+= mv_usb2_phy.o
drivers/usb/phy/mv_usb2_phy.c new file +39 -0
new file mode 100644
index 0000000..a81e5e4
@@ -0,0 +1,402 @@
+/*
+ * Copyright (C) 2013 Marvell Inc.
+ *
+ * Author:
+ *	Chao Xie <[email protected]>
+ *
+ * This software is licensed under the terms of the GNU General Public
+ * License version 2, as published by the Free Software Foundation, and
+ * may be copied, distributed, and modified under those terms.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ */
+
+#include <linux/resource.h>
+#include <linux/delay.h>
+#include <linux/slab.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/io.h>
+#include <linux/err.h>
+#include <linux/clk.h>
+#include <linux/export.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/platform_data/mv_usb.h>
+#include <linux/usb/phy.h>
+#include <linux/usb/mv_usb2.h>
+
+/* phy regs */
+#define UTMI_REVISION		0x0
+#define UTMI_CTRL		0x4
+#define UTMI_PLL		0x8
+#define UTMI_TX			0xc
+#define UTMI_RX			0x10
+#define UTMI_IVREF		0x14

drivers/usb/phy/mv_usb2_phy.c line 49

Quoted 10 lines
drivers/usb/phy/mv_usb2_phy.c new file +10 -0
+#define UTMI_T0			0x18
+#define UTMI_T1			0x1c
+#define UTMI_T2			0x20
+#define UTMI_T3			0x24
+#define UTMI_T4			0x28
+#define UTMI_T5			0x2c
+#define UTMI_RESERVE		0x30
+#define UTMI_USB_INT		0x34
+#define UTMI_DBG_CTL		0x38
+#define UTMI_OTG_ADDON		0x3c

prepend these with MV_USB_

Quoted 14 lines
drivers/usb/phy/mv_usb2_phy.c new file +14 -0
+enum mv_usb2_phy_type {
+	PXA168_USB,
+	PXA910_USB,
+	MMP2_USB,
+};
+
+static unsigned int u2o_get(void __iomem *base, unsigned int offset)
+{
+	return readl(base + offset);
+}
+
+static void u2o_set(void __iomem *base, unsigned int offset,
+		unsigned int value)
+{

drivers/usb/phy/mv_usb2_phy.c line 149

Quoted 10 lines
drivers/usb/phy/mv_usb2_phy.c new file +10 -0
+	u32 reg;
+
+	reg = readl(base + offset);
+	reg |= value;
+	writel(reg, base + offset);
+	/*
+	 * read after write. It will make sure writing takes effect.
+	 * It is suggested by PHY design engineer.
+	 */
+	readl(base + offset);

ewww... you really don't need (and *shouldn't* use) u2o_set() or u2o_clear(). They clearly prevent compiler from optimizing variable usage and could cause pressure on your interconnect. By writing and reading multiple times for no reason.

drivers/usb/phy/mv_usb2_phy.c line 184

Quoted 7 lines
drivers/usb/phy/mv_usb2_phy.c new file +7 -0
+static int _mv_usb2_phy_init(struct mv_usb2_phy *mv_phy)
+{
+	struct platform_device *pdev = mv_phy->pdev;
+	unsigned int loops = 0;
+	void __iomem *base = mv_phy->base;
+
+	dev_dbg(&pdev->dev, "phy init\n");

remove this debugging message.

drivers/usb/phy/mv_usb2_phy.c line 187

Quoted 2 lines
drivers/usb/phy/mv_usb2_phy.c new file +2 -0
+	/* Initialize the USB PHY power */
+	if (mv_phy->type == PXA910_USB) {

you _do_ have a REVISION register. Are you 100% certain that revision is the same on all your devices ? It seems to me that you should be doing proper revision detection instead of adding the hacky enumeration of yours.

drivers/usb/phy/mv_usb2_phy.c line 224

Quoted 4 lines
drivers/usb/phy/mv_usb2_phy.c new file +4 -0
+	/* UTMI_IVREF */
+	if (mv_phy->type == PXA168_USB)
+		/* fixing Microsoft Altair board interface with NEC hub issue -
+		 * Set UTMI_IVREF from 0x4a3 to 0x4bf */

wrong comment style. Run *ALL* your patches through checkpatch.pl --strict and sparse.

drivers/usb/phy/mv_usb2_phy.c line 228

Quoted 4 lines
drivers/usb/phy/mv_usb2_phy.c new file +4 -0
+		u2o_write(base, UTMI_IVREF, 0x4bf);
+
+	/* toggle VCOCAL_START bit of UTMI_PLL */
+	udelay(200);

why the udelay() calls ? Add a comment to code explaining them.

Quoted 9 lines
drivers/usb/phy/mv_usb2_phy.c new file +9 -0
+	u2o_set(base, UTMI_PLL, VCOCAL_START);
+	udelay(40);
+	u2o_clear(base, UTMI_PLL, VCOCAL_START);
+
+	/* toggle REG_RCAL_START bit of UTMI_TX */
+	udelay(400);
+	u2o_set(base, UTMI_TX, REG_RCAL_START);
+	udelay(40);
+	u2o_clear(base, UTMI_TX, REG_RCAL_START);

drivers/usb/phy/mv_usb2_phy.c line 247

Quoted 10 lines
drivers/usb/phy/mv_usb2_phy.c new file +10 -0
+	udelay(400);
+
+	/* Make sure PHY PLL is ready */
+	loops = 0;
+	while ((u2o_get(base, UTMI_PLL) & PLL_READY) == 0) {
+		mdelay(1);
+		loops++;
+		if (loops > 100) {
+			dev_warn(&pdev->dev, "calibrate timeout, UTMI_PLL %x\n",
+				u2o_get(base, UTMI_PLL));

if this fails, shouldn't you return an error code ?

drivers/usb/phy/mv_usb2_phy.c line 272

Quoted 12 lines
drivers/usb/phy/mv_usb2_phy.c new file +12 -0
+static int _mv_usb2_phy_shutdown(struct mv_usb2_phy *mv_phy)
+{
+	void __iomem *base = mv_phy->base;
+
+	if (mv_phy->type == PXA168_USB)
+		u2o_clear(base, UTMI_OTG_ADDON, UTMI_OTG_ADDON_OTG_ON);
+
+	u2o_clear(base, UTMI_CTRL, UTMI_CTRL_RXBUF_PDWN);
+	u2o_clear(base, UTMI_CTRL, UTMI_CTRL_TXBUF_PDWN);
+	u2o_clear(base, UTMI_CTRL, UTMI_CTRL_USB_CLK_EN);
+	u2o_clear(base, UTMI_CTRL, 1<<UTMI_CTRL_PWR_UP_SHIFT);
+	u2o_clear(base, UTMI_CTRL, 1<<UTMI_CTRL_PLL_PWR_UP_SHIFT);

NAK, this is stupid, read once, clear bits you want to clear and write only once.

drivers/usb/phy/mv_usb2_phy.c line 282

Quoted 9 lines
drivers/usb/phy/mv_usb2_phy.c new file +9 -0
+	return 0;
+}
+
+static int mv_usb2_phy_init(struct usb_phy *phy)
+{
+	struct mv_usb2_phy *mv_phy = container_of(phy, struct mv_usb2_phy, phy);
+	int i = 0;
+
+	mutex_lock(&mv_phy->phy_lock);

what's this mutex for ?

drivers/usb/phy/mv_usb2_phy.c line 283

Quoted 1 line
drivers/usb/phy/mv_usb2_phy.c new file +1 -0
+	if (mv_phy->refcount++ == 0) {

can this device really be used simultaneously by multiple devices ?

drivers/usb/phy/mv_usb2_phy.c line 340

Quoted 3 lines
drivers/usb/phy/mv_usb2_phy.c new file +3 -0
+	for (i = 0; i < mv_phy->clks_num; i++) {
+		mv_phy->clks[i] = devm_clk_get(&pdev->dev,
+						pdata->clkname[i]);

*NEVER* pass clock names via platform_data, this is utterly wrong.

drivers/usb/phy/mv_usb2_phy.c line 352

Quoted 12 lines
drivers/usb/phy/mv_usb2_phy.c new file +12 -0
+		if (IS_ERR(mv_phy->clks[i])) {
+			dev_err(&pdev->dev, "failed to get clock %s\n",
+				pdata->clkname[i]);
+			return PTR_ERR(mv_phy->clks[i]);
+		}
+	}
+	r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (r == NULL) {
+		dev_err(&pdev->dev, "no phy I/O memory resource defined\n");
+		return -ENODEV;
+	}
+	mv_phy->base = devm_ioremap(&pdev->dev, r->start, resource_size(r));

use devm_ioremap_resource() instead.

Quoted 4 lines
drivers/usb/phy/mv_usb2_phy.c new file +4 -0
+static struct platform_driver mv_usb2_phy_driver = {
+	.probe	= mv_usb2_phy_probe,
+	.remove = mv_usb2_phy_remove,
+	.driver = {

drivers/usb/phy/mv_usb2_phy.c line 402

Quoted 10 lines
drivers/usb/phy/mv_usb2_phy.c new file +10 -0
+		.name   = "pxa168-usb-phy",
+	},
+	.id_table = mv_usb2_phy_ids,
+};
+
+static int __init mv_usb2_phy_driver_init(void)
+{
+	return platform_driver_register(&mv_usb2_phy_driver);
+}
+arch_initcall(mv_usb2_phy_driver_init);

NAK, use module_platform_driver() like everybody else and handle registration ordering with -EPROBE_DEFER.

include/linux/platform_data/mv_usb.h struct mv_usb_platform_data { · line 53

Quoted 12 lines
include/linux/platform_data/mv_usb.h +0 -3
index 944b01d..fd3d1b4 100644
@@ -47,9 +47,12 @@ struct mv_usb_platform_data {
 	/* Force a_bus_req to be asserted */
 	 unsigned int    otg_force_a_bus_req:1;
 
-	int	(*phy_init)(void __iomem *regbase);
-	void	(*phy_deinit)(void __iomem *regbase);
 	int	(*set_vbus)(unsigned int vbus);
-	int     (*private_init)(void __iomem *opregs, void __iomem *phyregs);

should be part of a separate patch.

include/linux/platform_data/mv_usb.h struct mv_usb_platform_data { · line 56

Quoted 6 lines
include/linux/platform_data/mv_usb.h +5 -0
 };
+
+struct mv_usb_phy_platform_data {
+	unsigned int	clknum;
+	char		**clkname;
+};

NAK for this platform_data.

Signature · 8 lines
balbi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20130304/def7411d/attachment.sig>
#3 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

Quoted 21 lines · Felipe Balbi
Quoted 20 lines · Chao Xie

The PHY is seperated from usb controller.
The usb controller used in marvell pxa168/pxa910/mmp2 are same, but PHY initialization may be different.
the usb controller can support udc/otg/ehci, and for each of the mode, it need PHY to initialized before use the controller. Direclty writing the phy driver will make the usb controller driver to be simple and portable.
The PHY driver will be used by marvell udc/otg/ehci.

---
 drivers/usb/phy/Kconfig              |    7 +
 drivers/usb/phy/Makefile             |    1 +
 drivers/usb/phy/mv_usb2_phy.c        |  402 ++++++++++++++++++++++++++++++++++
 include/linux/platform_data/mv_usb.h |    9 +-
 include/linux/usb/mv_usb2.h          |   32 +++
 5 files changed, 448 insertions(+), 3 deletions(-)
 create mode 100644 drivers/usb/phy/mv_usb2_phy.c
 create mode 100644 include/linux/usb/mv_usb2.h
Quoted 14 lines

drivers/usb/phy/Kconfig config SAMSUNG_USBPHY · line 79

Quoted 12 lines
drivers/usb/phy/Kconfig +4 -0
index 65217a5..5479760 100644
@@ -73,3 +73,10 @@ config SAMSUNG_USBPHY
      help
        Enable this to support Samsung USB phy controller for samsung
        SoCs.
+
+config MV_USB2_PHY
+     tristate "Marvell USB 2.0 PHY Driver"
+     depends on USB || USB_GADGET

NAK, PHYs don't depend on the gadget framework.

Sure, i will remove it.

Quoted 37 lines
Quoted 37 lines
drivers/usb/phy/Kconfig +3 -0
+     help
+       Enable this to support Marvell USB 2.0 phy driver for Marvell
+       SoC.
drivers/usb/phy/Makefile +1 -0
index b13faa1..3835316 100644
@@ -12,3 +12,4 @@ obj-$(CONFIG_MV_U3D_PHY)            += mv_u3d_phy.o
 obj-$(CONFIG_USB_EHCI_TEGRA) += tegra_usb_phy.o
 obj-$(CONFIG_USB_RCAR_PHY)           += rcar-phy.o
 obj-$(CONFIG_SAMSUNG_USBPHY)         += samsung-usbphy.o
+obj-$(CONFIG_MV_USB2_PHY)            += mv_usb2_phy.o
drivers/usb/phy/mv_usb2_phy.c new file +17 -0
new file mode 100644
index 0000000..a81e5e4
@@ -0,0 +1,402 @@
+/*
+ * Copyright (C) 2013 Marvell Inc.
+ *
+ * Author:
+ *   Chao Xie <[email protected]>
+ *
+ * This software is licensed under the terms of the GNU General Public
+ * License version 2, as published by the Free Software Foundation, and
 
+#define UTMI_CTRL            0x4
+#define UTMI_PLL             0x8
+#define UTMI_TX                      0xc
+#define UTMI_RX                      0x10
+#define UTMI_IVREF           0x14
+#define UTMI_T0                      0x18
+#define UTMI_T1                      0x1c
+#define UTMI_T2                      0x20
+#define UTMI_T3                      0x24
Quoted 8 lines

drivers/usb/phy/mv_usb2_phy.c line 49

Quoted 6 lines
drivers/usb/phy/mv_usb2_phy.c new file +6 -0
+#define UTMI_T4                      0x28
+#define UTMI_T5                      0x2c
+#define UTMI_RESERVE         0x30
+#define UTMI_USB_INT         0x34
+#define UTMI_DBG_CTL         0x38
+#define UTMI_OTG_ADDON               0x3c

prepend these with MV_USB_

Fine.

Quoted 13 lines

drivers/usb/phy/mv_usb2_phy.c line 130

Quoted 6 lines
drivers/usb/phy/mv_usb2_phy.c new file +5 -0
+enum mv_usb2_phy_type {
+     PXA168_USB,
+     PXA910_USB,
+     MMP2_USB,
+};

ewww... you really don't need (and *shouldn't* use) u2o_set() or u2o_clear(). They clearly prevent compiler from optimizing variable usage and could cause pressure on your interconnect. By writing and reading multiple times for no reason.

The APIs defined here is for device operation. The device register read/write is not same as memory.
When a silicon comes, it may not be stable, and will have done some workaround epecially for the
device register read/write. to define the APIs for the register read/write will help to implement the workaround
without changing phy init code every time.
Now the only constrain is should read back the registers if you have writen to it.
It will low down the performance, but it does not matter. Because phy init will only done once when you initialize it.
I will think about reove the u2o_xxx APIs.

Quoted 10 lines

drivers/usb/phy/mv_usb2_phy.c line 184

Quoted 7 lines
drivers/usb/phy/mv_usb2_phy.c new file +7 -0
+static int _mv_usb2_phy_init(struct mv_usb2_phy *mv_phy)
+{
+     struct platform_device *pdev = mv_phy->pdev;
+     unsigned int loops = 0;
+     void __iomem *base = mv_phy->base;
+
+     dev_dbg(&pdev->dev, "phy init\n");

remove this debugging message.

Quoted 7 lines

drivers/usb/phy/mv_usb2_phy.c line 187

Quoted 2 lines
drivers/usb/phy/mv_usb2_phy.c new file +2 -0
+     /* Initialize the USB PHY power */
+     if (mv_phy->type == PXA910_USB) {

you _do_ have a REVISION register. Are you 100% certain that revision is the same on all your devices ? It seems to me that you should be doing proper revision detection instead of adding the hacky enumeration of yours.

We do not have revison registers and will do not want ot define #ifdef CPU_PXA910 or CPU_PXA_XXX
or
use is_cpu_pxa910 or cpu_pxa_xxx
because it is not acceptable. for example, if we have new SOC and it use the same PHY as pxa910
i have to change the USB driver code to support it. for example #ifdef CPU_PXA910 || CPU_XXX

So i have to depends on the device_id to do the work.

Quoted 9 lines

drivers/usb/phy/mv_usb2_phy.c line 224

Quoted 5 lines
drivers/usb/phy/mv_usb2_phy.c new file +4 -0
+     /* UTMI_IVREF */
+     if (mv_phy->type == PXA168_USB)
+             /* fixing Microsoft Altair board interface with NEC hub issue -
+              * Set UTMI_IVREF from 0x4a3 to 0x4bf */

wrong comment style. Run *ALL* your patches through checkpatch.pl --strict and sparse.

It seems that checkpatch.pl can not detect everything. I really use checpatch.pl for
every patch i sent to maillist.
sorry for that, i will fix it.

Quoted 20 lines

drivers/usb/phy/mv_usb2_phy.c line 228

Quoted 4 lines
drivers/usb/phy/mv_usb2_phy.c new file +4 -0
+             u2o_write(base, UTMI_IVREF, 0x4bf);
+
+     /* toggle VCOCAL_START bit of UTMI_PLL */
+     udelay(200);

why the udelay() calls ? Add a comment to code explaining them.

Quoted 13 lines
drivers/usb/phy/mv_usb2_phy.c new file +13 -0
+     u2o_set(base, UTMI_PLL, VCOCAL_START);
+     udelay(40);
+     u2o_clear(base, UTMI_PLL, VCOCAL_START);
+
+     /* toggle REG_RCAL_START bit of UTMI_TX */
+     udelay(400);
+     u2o_set(base, UTMI_TX, REG_RCAL_START);
+     udelay(40);
+     u2o_clear(base, UTMI_TX, REG_RCAL_START);
+     udelay(400);
+
+     /* Make sure PHY PLL is ready */
+     loops = 0;
Quoted 9 lines

drivers/usb/phy/mv_usb2_phy.c line 247

Quoted 6 lines
drivers/usb/phy/mv_usb2_phy.c new file +6 -0
+     while ((u2o_get(base, UTMI_PLL) & PLL_READY) == 0) {
+             mdelay(1);
+             loops++;
+             if (loops > 100) {
+                     dev_warn(&pdev->dev, "calibrate timeout, UTMI_PLL %x\n",
+                             u2o_get(base, UTMI_PLL));

if this fails, shouldn't you return an error code ?

yes. it should return the error code.

Quoted 7 lines
Quoted 7 lines
drivers/usb/phy/mv_usb2_phy.c new file +7 -0
+static int _mv_usb2_phy_shutdown(struct mv_usb2_phy *mv_phy)
+{
+     void __iomem *base = mv_phy->base;
+
+     if (mv_phy->type == PXA168_USB)
+             u2o_clear(base, UTMI_OTG_ADDON, UTMI_OTG_ADDON_OTG_ON);
+
Quoted 9 lines

drivers/usb/phy/mv_usb2_phy.c line 272

Quoted 5 lines
drivers/usb/phy/mv_usb2_phy.c new file +5 -0
+     u2o_clear(base, UTMI_CTRL, UTMI_CTRL_RXBUF_PDWN);
+     u2o_clear(base, UTMI_CTRL, UTMI_CTRL_TXBUF_PDWN);
+     u2o_clear(base, UTMI_CTRL, UTMI_CTRL_USB_CLK_EN);
+     u2o_clear(base, UTMI_CTRL, 1<<UTMI_CTRL_PWR_UP_SHIFT);
+     u2o_clear(base, UTMI_CTRL, 1<<UTMI_CTRL_PLL_PWR_UP_SHIFT);

NAK, this is stupid, read once, clear bits you want to clear and write only once.

I will check with silicon design engineer. becuase all the phy init code is delivered by them.
I can not tell that if there are any speciall reason they will do so because the device register
read/write is not same as normal memory.

Quoted 12 lines

drivers/usb/phy/mv_usb2_phy.c line 282

Quoted 9 lines
drivers/usb/phy/mv_usb2_phy.c new file +9 -0
+     return 0;
+}
+
+static int mv_usb2_phy_init(struct usb_phy *phy)
+{
+     struct mv_usb2_phy *mv_phy = container_of(phy, struct mv_usb2_phy, phy);
+     int i = 0;
+
+     mutex_lock(&mv_phy->phy_lock);

what's this mutex for ?

Quoted 4 lines
Quoted 1 line

+ if (mv_phy->refcount++ == 0) {

can this device really be used simultaneously by multiple devices ?

Sure, we have another EHCI device, it will share the PHY with this USB controller.
So we will use refcount and mutext to protect the phy init.

Quoted 6 lines

drivers/usb/phy/mv_usb2_phy.c line 340

Quoted 3 lines
drivers/usb/phy/mv_usb2_phy.c new file +3 -0
+     for (i = 0; i < mv_phy->clks_num; i++) {
+             mv_phy->clks[i] = devm_clk_get(&pdev->dev,
+                                             pdata->clkname[i]);

*NEVER* pass clock names via platform_data, this is utterly wrong.

without device tree support, the only way we can get the clock is the pdata. the use phy have mutiple clocks.
So what do you suggest to handle it?

Quoted 13 lines

drivers/usb/phy/mv_usb2_phy.c line 352

Quoted 12 lines
drivers/usb/phy/mv_usb2_phy.c new file +12 -0
+             if (IS_ERR(mv_phy->clks[i])) {
+                     dev_err(&pdev->dev, "failed to get clock %s\n",
+                             pdata->clkname[i]);
+                     return PTR_ERR(mv_phy->clks[i]);
+             }
+     }
+     r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+     if (r == NULL) {
+             dev_err(&pdev->dev, "no phy I/O memory resource defined\n");
+             return -ENODEV;
+     }
+     mv_phy->base = devm_ioremap(&pdev->dev, r->start, resource_size(r));
Quoted 2 lines

use devm_ioremap_resource() instead.

fine.

Quoted 9 lines
Quoted 9 lines
drivers/usb/phy/mv_usb2_phy.c new file +9 -0
+static struct platform_driver mv_usb2_phy_driver = {
+     .probe  = mv_usb2_phy_probe,
+     .remove = mv_usb2_phy_remove,
+     .driver = {
+             .name   = "pxa168-usb-phy",
+     },
+     .id_table = mv_usb2_phy_ids,
+};
+
Quoted 9 lines

drivers/usb/phy/mv_usb2_phy.c line 402

Quoted 5 lines
drivers/usb/phy/mv_usb2_phy.c new file +5 -0
+static int __init mv_usb2_phy_driver_init(void)
+{
+     return platform_driver_register(&mv_usb2_phy_driver);
+}
+arch_initcall(mv_usb2_phy_driver_init);

NAK, use module_platform_driver() like everybody else and handle registration ordering with -EPROBE_DEFER.

The reason i do not use module_platform_driver is the compiling sequence of each directory of driver/usb/
the phy is compiled after otg/ehci. So it means that it can not find the usb phy, and will register fail.

Quoted 27 lines

include/linux/platform_data/mv_usb.h struct mv_usb_platform_data { · line 53

Quoted 12 lines
include/linux/platform_data/mv_usb.h +0 -3
index 944b01d..fd3d1b4 100644
@@ -47,9 +47,12 @@ struct mv_usb_platform_data {
      /* Force a_bus_req to be asserted */
       unsigned int    otg_force_a_bus_req:1;
 
-     int     (*phy_init)(void __iomem *regbase);
-     void    (*phy_deinit)(void __iomem *regbase);
      int     (*set_vbus)(unsigned int vbus);
-     int     (*private_init)(void __iomem *opregs, void __iomem *phyregs);

should be part of a separate patch.

include/linux/platform_data/mv_usb.h struct mv_usb_platform_data { · line 56

Quoted 6 lines
include/linux/platform_data/mv_usb.h +5 -0
 };
+
+struct mv_usb_phy_platform_data {
+     unsigned int    clknum;
+     char            **clkname;
+};

NAK for this platform_data.

--
balbi

#4 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

Hi,

Quoted 13 lines · Chao Xie
Quoted 12 lines

drivers/usb/phy/mv_usb2_phy.c line 130

Quoted 5 lines
drivers/usb/phy/mv_usb2_phy.c new file +5 -0
+enum mv_usb2_phy_type {
+     PXA168_USB,
+     PXA910_USB,
+     MMP2_USB,
+};

ewww... you really don't need (and *shouldn't* use) u2o_set() or u2o_clear(). They clearly prevent compiler from optimizing variable usage and could cause pressure on your interconnect. By writing and reading multiple times for no reason.

Quoted 11 lines

The APIs defined here is for device operation. The device register read/write is not same as memory.
When a silicon comes, it may not be stable, and will have done some workaround epecially for the device register read/write. to define the APIs for the register read/write will help to implement the workaround without changing phy init code every time.
Now the only constrain is should read back the registers if you have writen to it.
It will low down the performance, but it does not matter. Because phy init will only done once when you initialize it.
I will think about reove the u2o_xxx APIs.

You didn't even understand what I meant. Seriously.

Anyway, details are as follows:

readl() and writel() always add a memory barrier around each operation.

This is good because it makes sure memory mapped I/O region is always ordered, but your current usage will post reads and writes on the interconnect over and over again for no reason. What you should do, with any register access is:

reg = readl();

reg &= ~BITS_TO_DISABLE;
reg |= BITS_TO_ENABLE;

writel(reg);

Whereas your current code does:

reg = readl();
reg &= ~BITS_TO_DISABLE;
writel(reg);

reg = readl();
reg |= BITS_TO_ENABLE;
writel(reg);

You do that for each bit, in some cases.

Quoted 18 lines
Quoted 10 lines

drivers/usb/phy/mv_usb2_phy.c line 184

Quoted 7 lines
drivers/usb/phy/mv_usb2_phy.c new file +7 -0
+static int _mv_usb2_phy_init(struct mv_usb2_phy *mv_phy)
+{
+     struct platform_device *pdev = mv_phy->pdev;
+     unsigned int loops = 0;
+     void __iomem *base = mv_phy->base;
+
+     dev_dbg(&pdev->dev, "phy init\n");

remove this debugging message.

Quoted 7 lines

drivers/usb/phy/mv_usb2_phy.c line 187

Quoted 2 lines
drivers/usb/phy/mv_usb2_phy.c new file +2 -0
+     /* Initialize the USB PHY power */
+     if (mv_phy->type == PXA910_USB) {

you _do_ have a REVISION register. Are you 100% certain that revision is the same on all your devices ? It seems to me that you should be doing proper revision detection instead of adding the hacky enumeration of yours.

Quoted 8 lines

We do not have revison registers and will do not want ot define #ifdef CPU_PXA910 or CPU_PXA_XXX
or
use is_cpu_pxa910 or cpu_pxa_xxx
because it is not acceptable. for example, if we have new SOC and it use the same PHY as pxa910
i have to change the USB driver code to support it. for example #ifdef CPU_PXA910 || CPU_XXX

what a load of crap.

*read your code!!!*

There is a UTMI_REVISION register defined at offset 0.

Quoted 7 lines

So i have to depends on the device_id to do the work.

Quoted 6 lines
Quoted 5 lines
drivers/usb/phy/mv_usb2_phy.c new file +4 -0
+     /* UTMI_IVREF */
+     if (mv_phy->type == PXA168_USB)
+             /* fixing Microsoft Altair board interface with NEC hub issue -
+              * Set UTMI_IVREF from 0x4a3 to 0x4bf */
Quoted 8 lines
Quoted 3 lines

wrong comment style. Run *ALL* your patches through checkpatch.pl --strict and sparse.

It seems that checkpatch.pl can not detect everything. I really use checpatch.pl for
every patch i sent to maillist.
sorry for that, i will fix it.

did you use --strict ?

Quoted 17 lines
Quoted 7 lines
Quoted 7 lines
drivers/usb/phy/mv_usb2_phy.c new file +7 -0
+static int _mv_usb2_phy_shutdown(struct mv_usb2_phy *mv_phy)
+{
+     void __iomem *base = mv_phy->base;
+
+     if (mv_phy->type == PXA168_USB)
+             u2o_clear(base, UTMI_OTG_ADDON, UTMI_OTG_ADDON_OTG_ON);
+
Quoted 9 lines

drivers/usb/phy/mv_usb2_phy.c line 272

Quoted 5 lines
drivers/usb/phy/mv_usb2_phy.c new file +5 -0
+     u2o_clear(base, UTMI_CTRL, UTMI_CTRL_RXBUF_PDWN);
+     u2o_clear(base, UTMI_CTRL, UTMI_CTRL_TXBUF_PDWN);
+     u2o_clear(base, UTMI_CTRL, UTMI_CTRL_USB_CLK_EN);
+     u2o_clear(base, UTMI_CTRL, 1<<UTMI_CTRL_PWR_UP_SHIFT);
+     u2o_clear(base, UTMI_CTRL, 1<<UTMI_CTRL_PLL_PWR_UP_SHIFT);

NAK, this is stupid, read once, clear bits you want to clear and write only once.

Quoted 2 lines

I will check with silicon design engineer. becuase all the phy init code is delivered by them.

right, but you need to be critical with any code you read. You can't simply blindly make it compile on linux and hope that it'll work efficiently.

I doubt your silicon validation evironment has support for the memory barriers which are added on each readl()/writel() calls.

Quoted 2 lines

I can not tell that if there are any speciall reason they will do so because the device register read/write is not same as normal memory.

seriously ? Read your documentation dude, there's nothing that will require you to enable one bit in the register, then flush the write, then write another bit, flush the write, write another bit, flush the write and so on.

Those cases are extremely rare (MUSB has only *ONE* such case with regards to DMA_MODE bit) and in 99% of the cases, you can write everything to a variable and post a single write to your interconnect.

Stop trying to come up with nonsensical excuses that "register read/write is not same as normal memory", of course it's not same as normal memory, and that's why we have standardized I/O functions.

Quoted 12 lines
Quoted 12 lines

drivers/usb/phy/mv_usb2_phy.c line 282

Quoted 9 lines
drivers/usb/phy/mv_usb2_phy.c new file +9 -0
+     return 0;
+}
+
+static int mv_usb2_phy_init(struct usb_phy *phy)
+{
+     struct mv_usb2_phy *mv_phy = container_of(phy, struct mv_usb2_phy, phy);
+     int i = 0;
+
+     mutex_lock(&mv_phy->phy_lock);

what's this mutex for ?

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

+ if (mv_phy->refcount++ == 0) {

can this device really be used simultaneously by multiple devices ?

Sure, we have another EHCI device, it will share the PHY with this USB controller.

what is "this USB controller", current driver is for the PHY only. Do you mean to say that PHY will be shared by EHCI and UDC ?

Quoted 1 line

So we will use refcount and mutext to protect the phy init.

in that case, it might be better to add a generic refcounting to the PHY layer as a whole, instead of cooking your own private solutions.

Quoted 9 lines
Quoted 6 lines

drivers/usb/phy/mv_usb2_phy.c line 340

Quoted 3 lines
drivers/usb/phy/mv_usb2_phy.c new file +3 -0
+     for (i = 0; i < mv_phy->clks_num; i++) {
+             mv_phy->clks[i] = devm_clk_get(&pdev->dev,
+                                             pdata->clkname[i]);

*NEVER* pass clock names via platform_data, this is utterly wrong.

without device tree support, the only way we can get the clock is the pdata. the use phy have mutiple clocks.
So what do you suggest to handle it?

clkdev

Quoted 19 lines
Quoted 9 lines
Quoted 9 lines
drivers/usb/phy/mv_usb2_phy.c new file +9 -0
+static struct platform_driver mv_usb2_phy_driver = {
+     .probe  = mv_usb2_phy_probe,
+     .remove = mv_usb2_phy_remove,
+     .driver = {
+             .name   = "pxa168-usb-phy",
+     },
+     .id_table = mv_usb2_phy_ids,
+};
+
Quoted 9 lines

drivers/usb/phy/mv_usb2_phy.c line 402

Quoted 5 lines
drivers/usb/phy/mv_usb2_phy.c new file +5 -0
+static int __init mv_usb2_phy_driver_init(void)
+{
+     return platform_driver_register(&mv_usb2_phy_driver);
+}
+arch_initcall(mv_usb2_phy_driver_init);

NAK, use module_platform_driver() like everybody else and handle registration ordering with -EPROBE_DEFER.

Quoted 4 lines

The reason i do not use module_platform_driver is the compiling sequence of each directory of driver/usb/
the phy is compiled after otg/ehci. So it means that it can not find the usb phy, and will register fail.

it doesn't matter, you should teach EHCI about -EPROBE_DEFER. If it can't grab the PHY, return -EPROBE_DEFER.

Signature · 8 lines
balbi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20130305/6bef20d9/attachment.sig>
maintainer
#5 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

Quoted 3 lines · Felipe Balbi

Anyway, details are as follows:

readl() and writel() always add a memory barrier around each operation.

Is that supposed to be true on all architectures or only on ARM?

Alan Stern

#6 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

Quoted 7 lines · Alan Stern
Quoted 3 lines · Felipe Balbi

Anyway, details are as follows:

readl() and writel() always add a memory barrier around each operation.

Is that supposed to be true on all architectures or only on ARM?

I believe it's true for every architecture which doesn't have strongly ordered memory accesses. ARM is just one example of that ;-)

In any case, his read->set 1 bit->write loop isn't good even for architectures with strongly ordered memory accesses. It will continue to post unnecessary reads and writes to the interconnect.

Signature · 8 lines
balbi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20130305/24df4eb1/attachment.sig>
#7 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

In reply to Felipe Balbi
Quoted 216 lines · Felipe Balbi

Hi,

Quoted 13 lines · Chao Xie
Quoted 12 lines

drivers/usb/phy/mv_usb2_phy.c line 130

Quoted 5 lines
drivers/usb/phy/mv_usb2_phy.c new file +5 -0
+enum mv_usb2_phy_type {
+     PXA168_USB,
+     PXA910_USB,
+     MMP2_USB,
+};

ewww... you really don't need (and *shouldn't* use) u2o_set() or u2o_clear(). They clearly prevent compiler from optimizing variable usage and could cause pressure on your interconnect. By writing and reading multiple times for no reason.

Quoted 11 lines

The APIs defined here is for device operation. The device register read/write is not same as memory.
When a silicon comes, it may not be stable, and will have done some workaround epecially for the device register read/write. to define the APIs for the register read/write will help to implement the workaround without changing phy init code every time.
Now the only constrain is should read back the registers if you have writen to it.
It will low down the performance, but it does not matter. Because phy init will only done once when you initialize it.
I will think about reove the u2o_xxx APIs.

You didn't even understand what I meant. Seriously.

Anyway, details are as follows:

readl() and writel() always add a memory barrier around each operation.

This is good because it makes sure memory mapped I/O region is always ordered, but your current usage will post reads and writes on the interconnect over and over again for no reason. What you should do, with any register access is:

reg = readl();

reg &= ~BITS_TO_DISABLE;
reg |= BITS_TO_ENABLE;

writel(reg);

Whereas your current code does:

reg = readl();
reg &= ~BITS_TO_DISABLE;
writel(reg);

reg = readl();
reg |= BITS_TO_ENABLE;
writel(reg);

You do that for each bit, in some cases.

Quoted 18 lines
Quoted 10 lines

drivers/usb/phy/mv_usb2_phy.c line 184

Quoted 7 lines
drivers/usb/phy/mv_usb2_phy.c new file +7 -0
+static int _mv_usb2_phy_init(struct mv_usb2_phy *mv_phy)
+{
+     struct platform_device *pdev = mv_phy->pdev;
+     unsigned int loops = 0;
+     void __iomem *base = mv_phy->base;
+
+     dev_dbg(&pdev->dev, "phy init\n");

remove this debugging message.

Quoted 7 lines

drivers/usb/phy/mv_usb2_phy.c line 187

Quoted 2 lines
drivers/usb/phy/mv_usb2_phy.c new file +2 -0
+     /* Initialize the USB PHY power */
+     if (mv_phy->type == PXA910_USB) {

you _do_ have a REVISION register. Are you 100% certain that revision is the same on all your devices ? It seems to me that you should be doing proper revision detection instead of adding the hacky enumeration of yours.

Quoted 8 lines

We do not have revison registers and will do not want ot define #ifdef CPU_PXA910 or CPU_PXA_XXX
or
use is_cpu_pxa910 or cpu_pxa_xxx
because it is not acceptable. for example, if we have new SOC and it use the same PHY as pxa910
i have to change the USB driver code to support it. for example #ifdef CPU_PXA910 || CPU_XXX

what a load of crap.

*read your code!!!*

There is a UTMI_REVISION register defined at offset 0.

Quoted 7 lines

So i have to depends on the device_id to do the work.

Quoted 6 lines
Quoted 5 lines
drivers/usb/phy/mv_usb2_phy.c new file +4 -0
+     /* UTMI_IVREF */
+     if (mv_phy->type == PXA168_USB)
+             /* fixing Microsoft Altair board interface with NEC hub issue -
+              * Set UTMI_IVREF from 0x4a3 to 0x4bf */
Quoted 8 lines
Quoted 3 lines

wrong comment style. Run *ALL* your patches through checkpatch.pl --strict and sparse.

It seems that checkpatch.pl can not detect everything. I really use checpatch.pl for
every patch i sent to maillist.
sorry for that, i will fix it.

did you use --strict ?

Quoted 17 lines
Quoted 7 lines
Quoted 7 lines
drivers/usb/phy/mv_usb2_phy.c new file +7 -0
+static int _mv_usb2_phy_shutdown(struct mv_usb2_phy *mv_phy)
+{
+     void __iomem *base = mv_phy->base;
+
+     if (mv_phy->type == PXA168_USB)
+             u2o_clear(base, UTMI_OTG_ADDON, UTMI_OTG_ADDON_OTG_ON);
+
Quoted 9 lines

drivers/usb/phy/mv_usb2_phy.c line 272

Quoted 5 lines
drivers/usb/phy/mv_usb2_phy.c new file +5 -0
+     u2o_clear(base, UTMI_CTRL, UTMI_CTRL_RXBUF_PDWN);
+     u2o_clear(base, UTMI_CTRL, UTMI_CTRL_TXBUF_PDWN);
+     u2o_clear(base, UTMI_CTRL, UTMI_CTRL_USB_CLK_EN);
+     u2o_clear(base, UTMI_CTRL, 1<<UTMI_CTRL_PWR_UP_SHIFT);
+     u2o_clear(base, UTMI_CTRL, 1<<UTMI_CTRL_PLL_PWR_UP_SHIFT);

NAK, this is stupid, read once, clear bits you want to clear and write only once.

Quoted 2 lines

I will check with silicon design engineer. becuase all the phy init code is delivered by them.

right, but you need to be critical with any code you read. You can't simply blindly make it compile on linux and hope that it'll work efficiently.

I doubt your silicon validation evironment has support for the memory barriers which are added on each readl()/writel() calls.

Quoted 2 lines

I can not tell that if there are any speciall reason they will do so because the device register read/write is not same as normal memory.

seriously ? Read your documentation dude, there's nothing that will require you to enable one bit in the register, then flush the write, then write another bit, flush the write, write another bit, flush the write and so on.

Those cases are extremely rare (MUSB has only *ONE* such case with regards to DMA_MODE bit) and in 99% of the cases, you can write everything to a variable and post a single write to your interconnect.

Stop trying to come up with nonsensical excuses that "register read/write is not same as normal memory", of course it's not same as normal memory, and that's why we have standardized I/O functions.

Quoted 12 lines
Quoted 12 lines

drivers/usb/phy/mv_usb2_phy.c line 282

Quoted 9 lines
drivers/usb/phy/mv_usb2_phy.c new file +9 -0
+     return 0;
+}
+
+static int mv_usb2_phy_init(struct usb_phy *phy)
+{
+     struct mv_usb2_phy *mv_phy = container_of(phy, struct mv_usb2_phy, phy);
+     int i = 0;
+
+     mutex_lock(&mv_phy->phy_lock);

what's this mutex for ?

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

+ if (mv_phy->refcount++ == 0) {

can this device really be used simultaneously by multiple devices ?

Sure, we have another EHCI device, it will share the PHY with this USB controller.

what is "this USB controller", current driver is for the PHY only. Do you mean to say that PHY will be shared by EHCI and UDC ?

Quoted 1 line

So we will use refcount and mutext to protect the phy init.

in that case, it might be better to add a generic refcounting to the PHY layer as a whole, instead of cooking your own private solutions.

Quoted 9 lines
Quoted 6 lines

drivers/usb/phy/mv_usb2_phy.c line 340

Quoted 3 lines
drivers/usb/phy/mv_usb2_phy.c new file +3 -0
+     for (i = 0; i < mv_phy->clks_num; i++) {
+             mv_phy->clks[i] = devm_clk_get(&pdev->dev,
+                                             pdata->clkname[i]);

*NEVER* pass clock names via platform_data, this is utterly wrong.

without device tree support, the only way we can get the clock is the pdata. the use phy have mutiple clocks.
So what do you suggest to handle it?

clkdev

Quoted 19 lines
Quoted 9 lines
Quoted 9 lines
drivers/usb/phy/mv_usb2_phy.c new file +9 -0
+static struct platform_driver mv_usb2_phy_driver = {
+     .probe  = mv_usb2_phy_probe,
+     .remove = mv_usb2_phy_remove,
+     .driver = {
+             .name   = "pxa168-usb-phy",
+     },
+     .id_table = mv_usb2_phy_ids,
+};
+
Quoted 9 lines

drivers/usb/phy/mv_usb2_phy.c line 402

Quoted 5 lines
drivers/usb/phy/mv_usb2_phy.c new file +5 -0
+static int __init mv_usb2_phy_driver_init(void)
+{
+     return platform_driver_register(&mv_usb2_phy_driver);
+}
+arch_initcall(mv_usb2_phy_driver_init);

NAK, use module_platform_driver() like everybody else and handle registration ordering with -EPROBE_DEFER.

Quoted 4 lines

The reason i do not use module_platform_driver is the compiling sequence of each directory of driver/usb/
the phy is compiled after otg/ehci. So it means that it can not find the usb phy, and will register fail.

Quoted 5 lines

it doesn't matter, you should teach EHCI about -EPROBE_DEFER. If it can't grab the PHY, return -EPROBE_DEFER.

--
balbi

Thanks for the comments, i summary the feedbacks and suggestions as below,
1. For the phy setting and u2o_xx APIs something, i will optimize it. If there are some special cases, i will add comments.
2. For the module_platform_driver, -EPROBE_DEFER is what i missed. I will check it and thanks for suggestion.
3. For the revison register. It exists in some SOCes(pxa168), but for some SOCes, the register dispears(pxa910, armada610). These SOCes are developed by different desgin teams, and it need to be enhanced, but for current products i have to use the device_id to detect the PHY controller.
4. For the mutex and refcount. The "USB controller" includes two blocks - the udc and ehci. In fact they will not be used at same time, but for some SOCes it duplicates the ehci block. It means that the SOCes have two or more ehci blocks. The added ehci blocks depend on the PHY to be initialized, and they can be used at same time as the "USB controller". That is the reason i add mutex and refcount for phy driver.
5. For the clock name. Can you describe it in details? For clkdev, if it want to find the clock, it still depends on the dev_name and con_id.

#8 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

Hi,

Quoted 5 lines · Chao Xie

3. For the revison register. It exists in some SOCes(pxa168), but for some SOCes, the register dispears(pxa910, armada610). These SOCes are developed by different desgin teams, and it need to be enhanced, but for current products i have to use the device_id to detect the PHY controller.

fair enough, make sure to add a comment to the driver about this so grumpy maintainers stop complaining ;-)

Quoted 7 lines

4. For the mutex and refcount. The "USB controller" includes two blocks - the udc and ehci. In fact they will not be used at same time, but for some SOCes it duplicates the ehci block. It means that the SOCes have two or more ehci blocks. The added ehci blocks depend on the PHY to be initialized, and they can be used at same time as the "USB controller". That is the reason i add mutex and refcount for phy driver.

alright, let's add refcounting to the generic PHY layer then.

Quoted 3 lines

5. For the clock name. Can you describe it in details? For clkdev, if it want to find the clock, it still depends on the dev_name and con_id.

right, the idea of clkdev is that you associate a clock provider with its consumer by means of dev_name. Then you can fetch the clock from driver with any name you want. Which means if you do you clkdev properly you could:

clk_get(dev, "clk1");
clk_get(dev, "clk2");
...

and so on.

good luck

Signature · 8 lines
balbi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20130306/c75c0f9c/attachment.sig>
#9 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

Quoted 31 lines · Felipe Balbi

Hi,

Quoted 5 lines · Chao Xie

3. For the revison register. It exists in some SOCes(pxa168), but for some SOCes, the register dispears(pxa910, armada610). These SOCes are developed by different desgin teams, and it need to be enhanced, but for current products i have to use the device_id to detect the PHY controller.

fair enough, make sure to add a comment to the driver about this so grumpy maintainers stop complaining ;-)

Quoted 7 lines

4. For the mutex and refcount. The "USB controller" includes two blocks - the udc and ehci. In fact they will not be used at same time, but for some SOCes it duplicates the ehci block. It means that the SOCes have two or more ehci blocks. The added ehci blocks depend on the PHY to be initialized, and they can be used at same time as the "USB controller". That is the reason i add mutex and refcount for phy driver.

alright, let's add refcounting to the generic PHY layer then.

Quoted 3 lines

5. For the clock name. Can you describe it in details? For clkdev, if it want to find the clock, it still depends on the dev_name and con_id.

right, the idea of clkdev is that you associate a clock provider with its consumer by means of dev_name. Then you can fetch the clock from driver with any name you want. Which means if you do you clkdev properly you could:

Quoted 6 lines

clk_get(dev, "clk1");
clk_get(dev, "clk2");
...

and so on.

It is same as what i think.
The clock numbers and names are depent of SOCes, i do not want to fix it in ther driver. So i use pdata to pass the clknum and clk names "clk1", "clk2" and so on.
in the driver, i use
devm_clk_get(&pdev->dev, pdata->clkname[i]);
Do you think it is acceptable?

When device tree for clock framework are supported, the things get easier, and it is my next step.

Quoted 4 lines

good luck

--
balbi

#10 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

Quoted 36 lines · Chao Xie
Quoted 35 lines · Felipe Balbi

Hi,

Quoted 5 lines · Chao Xie

3. For the revison register. It exists in some SOCes(pxa168), but for some SOCes, the register dispears(pxa910, armada610). These SOCes are developed by different desgin teams, and it need to be enhanced, but for current products i have to use the device_id to detect the PHY controller.

fair enough, make sure to add a comment to the driver about this so grumpy maintainers stop complaining ;-)

Quoted 7 lines

4. For the mutex and refcount. The "USB controller" includes two blocks - the udc and ehci. In fact they will not be used at same time, but for some SOCes it duplicates the ehci block. It means that the SOCes have two or more ehci blocks. The added ehci blocks depend on the PHY to be initialized, and they can be used at same time as the "USB controller". That is the reason i add mutex and refcount for phy driver.

alright, let's add refcounting to the generic PHY layer then.

Quoted 3 lines

5. For the clock name. Can you describe it in details? For clkdev, if it want to find the clock, it still depends on the dev_name and con_id.

right, the idea of clkdev is that you associate a clock provider with its consumer by means of dev_name. Then you can fetch the clock from driver with any name you want. Which means if you do you clkdev properly you could:

clk_get(dev, "clk1");
clk_get(dev, "clk2");
...

Quoted 5 lines
Quoted 2 lines

and so on.

It is same as what i think.
The clock numbers and names are depent of SOCes, i do not want to fix

no, they're not dependent on anything, not if you use clkdev correctly.

Quoted 5 lines

it in ther driver. So i use pdata to pass the clknum and clk names "clk1", "clk2" and so on.
in the driver, i use
devm_clk_get(&pdev->dev, pdata->clkname[i]);
Do you think it is acceptable?

no

Quoted 2 lines

When device tree for clock framework are supported, the things get easier, and it is my next step.

not strong enough argument to accept passing clock names via platform_data.

Signature · 8 lines
balbi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20130306/826acfb5/attachment.sig>
#11 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

Quoted 37 lines · Felipe Balbi
Quoted 36 lines · Chao Xie
Quoted 35 lines · Felipe Balbi

Hi,

Quoted 5 lines · Chao Xie

3. For the revison register. It exists in some SOCes(pxa168), but for some SOCes, the register dispears(pxa910, armada610). These SOCes are developed by different desgin teams, and it need to be enhanced, but for current products i have to use the device_id to detect the PHY controller.

fair enough, make sure to add a comment to the driver about this so grumpy maintainers stop complaining ;-)

Quoted 7 lines

4. For the mutex and refcount. The "USB controller" includes two blocks - the udc and ehci. In fact they will not be used at same time, but for some SOCes it duplicates the ehci block. It means that the SOCes have two or more ehci blocks. The added ehci blocks depend on the PHY to be initialized, and they can be used at same time as the "USB controller". That is the reason i add mutex and refcount for phy driver.

alright, let's add refcounting to the generic PHY layer then.

Quoted 3 lines

5. For the clock name. Can you describe it in details? For clkdev, if it want to find the clock, it still depends on the dev_name and con_id.

right, the idea of clkdev is that you associate a clock provider with its consumer by means of dev_name. Then you can fetch the clock from driver with any name you want. Which means if you do you clkdev properly you could:

clk_get(dev, "clk1");
clk_get(dev, "clk2");
...

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

and so on.

It is same as what i think.
The clock numbers and names are depent of SOCes, i do not want to fix

no, they're not dependent on anything, not if you use clkdev correctly.

So
1. Does it mean that all SOCes clock driver should define same names such as "clk1", "clk2"?
2. Does it mean that if driver failed at clk_get, the probing will not stop because this SOC may define this clock?

Quoted 16 lines
Quoted 5 lines

it in ther driver. So i use pdata to pass the clknum and clk names "clk1", "clk2" and so on.
in the driver, i use
devm_clk_get(&pdev->dev, pdata->clkname[i]);
Do you think it is acceptable?

no

Quoted 2 lines

When device tree for clock framework are supported, the things get easier, and it is my next step.

not strong enough argument to accept passing clock names via platform_data.

--
balbi

#12 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

Quoted 44 lines · Chao Xie
Quoted 43 lines · Felipe Balbi
Quoted 41 lines · Chao Xie
Quoted 31 lines · Felipe Balbi

Hi,

Quoted 5 lines · Chao Xie

3. For the revison register. It exists in some SOCes(pxa168), but for some SOCes, the register dispears(pxa910, armada610). These SOCes are developed by different desgin teams, and it need to be enhanced, but for current products i have to use the device_id to detect the PHY controller.

fair enough, make sure to add a comment to the driver about this so grumpy maintainers stop complaining ;-)

Quoted 7 lines

4. For the mutex and refcount. The "USB controller" includes two blocks - the udc and ehci. In fact they will not be used at same time, but for some SOCes it duplicates the ehci block. It means that the SOCes have two or more ehci blocks. The added ehci blocks depend on the PHY to be initialized, and they can be used at same time as the "USB controller". That is the reason i add mutex and refcount for phy driver.

alright, let's add refcounting to the generic PHY layer then.

Quoted 3 lines

5. For the clock name. Can you describe it in details? For clkdev, if it want to find the clock, it still depends on the dev_name and con_id.

right, the idea of clkdev is that you associate a clock provider with its consumer by means of dev_name. Then you can fetch the clock from driver with any name you want. Which means if you do you clkdev properly you could:

Quoted 6 lines

clk_get(dev, "clk1");
clk_get(dev, "clk2");
...

and so on.

It is same as what i think.
The clock numbers and names are depent of SOCes, i do not want to fix

Quoted 6 lines
Quoted 2 lines

no, they're not dependent on anything, not if you use clkdev correctly.

So
1. Does it mean that all SOCes clock driver should define same names such as "clk1", "clk2"?

clk1, clk2, clk3 were just examples. Ideally you would have more meaningfull names like "fck" for functional clock.

Quoted 2 lines

2. Does it mean that if driver failed at clk_get, the probing will not stop because this SOC may define this clock?

that depends on how you write your driver. If platform A has 3 clocks and platform B has only 2 you have two options:

a) make the third clock optional by not bailing out if clk_get() fails
b) define a dummy no-op clock node to satisfy the third clock.

Signature · 8 lines
balbi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20130306/0317a612/attachment-0001.sig>
#13 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

In reply to Chao Xie
Quoted 1 line · Chao Xie

The clock numbers and names are depent of SOCes,

No they aren't. The clock names used to describe them in your documentation may vary, but their _purpose_ for the sake of the device will be fixed - and you should name them appropriately from the _device_ point of view.

Not the SoC point of view. That way leads to total madness. We've proven this over the years that we've had the clk API and people have come up with trash implementations that do that crap. After many years of struggling, they've seen the light and fixed their shite up to work the way I originally intended, and... instantly benefited from it.

#14 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

Quoted 7 lines · Russell King - ARM Linux
Quoted 1 line · Chao Xie

The clock numbers and names are depent of SOCes,

No they aren't. The clock names used to describe them in your documentation may vary, but their _purpose_ for the sake of the device will be fixed - and you should name them appropriately from the _device_ point of view.

Quoted 6 lines

Not the SoC point of view. That way leads to total madness. We've proven this over the years that we've had the clk API and people have come up with trash implementations that do that crap. After many years of struggling, they've seen the light and fixed their shite up to work the way I originally intended, and... instantly benefited from it.

Yes, you are right, and you explained it very clearly.. I understand what you said.
Our SOC document is lack of this kind of inoformation. I will talk with the desgin engneer to make clear it, and make a new version of patches.
Thanks for your review and comments.

#15 lore raw

PATCH v8 1/16 usb: phy: mv_usb2: add PHY driver for marvell usb2 controller

In reply to Chao Xie
Quoted 11 lines · Chao Xie
Quoted 7 lines · Felipe Balbi

drivers/usb/phy/mv_usb2_phy.c line 340

Quoted 3 lines · Chao Xie
drivers/usb/phy/mv_usb2_phy.c new file +3 -0
+     for (i = 0; i < mv_phy->clks_num; i++) {
+             mv_phy->clks[i] = devm_clk_get(&pdev->dev,
+                                             pdata->clkname[i]);

*NEVER* pass clock names via platform_data, this is utterly wrong.

without device tree support, the only way we can get the clock is the pdata. the use phy have mutiple clocks.
So what do you suggest to handle it?

Then you don't understand the clk API at all.

Read the documentation in include/linux/clk.h for clk_get().

The first parameter is the device which you're interested in getting the clock for.

The second parameter defines the INPUT as a string to THAT DEVICE. It is specific to the device. It is NOT the system name of the clock.

So, if you have a function clock and an interface clock to a device, then use a name like "fck" for the function clock and "ick" for the interface clock.

Do _NOT_ make the mistake of using "global" clock names. People have done that many times in the past and got into horrid sticky problems - and ended up with _far_ more code than is really necessary if you do things the right way.

author
#16 lore raw

PATCH v8 2/16 usb: gadget: mv_udc: use PHY driver for udc

In reply to Chao Xie

Originaly, udc driver will call the callbacks in platform data for PHY initialization and shut down.
With PHY driver, it will call the APIs provided by PHY driver for PHY initialization and shut down. It removes the callbacks in platform data, and at same time it removes one block in the way of enabling device tree for udc driver.

---
 drivers/usb/gadget/mv_udc.h      |    2 +-
 drivers/usb/gadget/mv_udc_core.c |   45 ++++++++++++++-----------------------
 2 files changed, 18 insertions(+), 29 deletions(-)
drivers/usb/gadget/mv_udc.h +1 -1
index 9073436..f339df4 100644
@@ -180,7 +180,6 @@ struct mv_udc {
 
 	struct mv_cap_regs __iomem	*cap_regs;
 	struct mv_op_regs __iomem	*op_regs;
-	void __iomem                    *phy_regs;
 	unsigned int			max_eps;
 	struct mv_dqh			*ep_dqh;
 	size_t				ep_dqh_size;
@@ -217,6 +216,7 @@ struct mv_udc {
 	struct work_struct	vbus_work;
 	struct workqueue_struct *qwork;
 
+	struct usb_phy		*phy;
 	struct usb_phy		*transceiver;
 
 	struct mv_usb_platform_data     *pdata;
drivers/usb/gadget/mv_udc_core.c +17 -28

82 lines not on this page · raw message

Signature · 1 line
1.7.4.1
#17 lore raw

PATCH v8 2/16 usb: gadget: mv_udc: use PHY driver for udc

Quoted 47 lines · Chao Xie

Originaly, udc driver will call the callbacks in platform data for PHY initialization and shut down.
With PHY driver, it will call the APIs provided by PHY driver for PHY initialization and shut down. It removes the callbacks in platform data, and at same time it removes one block in the way of enabling device tree for udc driver.

---
 drivers/usb/gadget/mv_udc.h      |    2 +-
 drivers/usb/gadget/mv_udc_core.c |   45 ++++++++++++++-----------------------
 2 files changed, 18 insertions(+), 29 deletions(-)
drivers/usb/gadget/mv_udc.h +1 -1
index 9073436..f339df4 100644
@@ -180,7 +180,6 @@ struct mv_udc {
 
 	struct mv_cap_regs __iomem	*cap_regs;
 	struct mv_op_regs __iomem	*op_regs;
-	void __iomem                    *phy_regs;
 	unsigned int			max_eps;
 	struct mv_dqh			*ep_dqh;
 	size_t				ep_dqh_size;
@@ -217,6 +216,7 @@ struct mv_udc {
 	struct work_struct	vbus_work;
 	struct workqueue_struct *qwork;
 
+	struct usb_phy		*phy;
 	struct usb_phy		*transceiver;
 
 	struct mv_usb_platform_data     *pdata;
drivers/usb/gadget/mv_udc_core.c +1 -0
index c8cf959..4876d2f 100644
@@ -35,6 +35,7 @@
 #include <linux/platform_device.h>
 #include <linux/clk.h>
 #include <linux/platform_data/mv_usb.h>
+#include <linux/usb/mv_usb2.h>
 #include <asm/unaligned.h>
 
 #include "mv_udc.h"
@@ -1121,15 +1122,14 @@ static int mv_udc_enable_internal(struct mv_udc *udc)
 

drivers/usb/gadget/mv_udc_core.c static int mv_udc_enable_internal(struct mv_udc *udc) · line 1131

Quoted 10 lines
drivers/usb/gadget/mv_udc_core.c +0 -8
 	dev_dbg(&udc->dev->dev, "enable udc\n");
 	udc_clock_enable(udc);
-	if (udc->pdata->phy_init) {
-		retval = udc->pdata->phy_init(udc->phy_regs);
-		if (retval) {
-			dev_err(&udc->dev->dev,
-				"init phy error %d\n", retval);
-			udc_clock_disable(udc);
-			return retval;
-		}

dude, you really don't test your patches, do you ? Your previous patch removed ->phy_init(), ->phy_deinit() and ->private_init() from your platform_data.

Are you seriously telling me you didn't compile test your patches ?

Rewrite your series again and make sure that each and every patch compiles and works on its own. You shouldn't cause any regressions or build breaks or build warnings while converting this driver.

I won't review the rest of this series and I'm purging it from my TODO list.

cheers

Signature · 8 lines
balbi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20130304/be03f1ea/attachment.sig>
#18 lore raw

PATCH v8 2/16 usb: gadget: mv_udc: use PHY driver for udc

Quoted 63 lines · Felipe Balbi
Quoted 47 lines · Chao Xie

Originaly, udc driver will call the callbacks in platform data for PHY initialization and shut down.
With PHY driver, it will call the APIs provided by PHY driver for PHY initialization and shut down. It removes the callbacks in platform data, and at same time it removes one block in the way of enabling device tree for udc driver.

---
 drivers/usb/gadget/mv_udc.h      |    2 +-
 drivers/usb/gadget/mv_udc_core.c |   45 ++++++++++++++-----------------------
 2 files changed, 18 insertions(+), 29 deletions(-)
drivers/usb/gadget/mv_udc.h +1 -1
index 9073436..f339df4 100644
@@ -180,7 +180,6 @@ struct mv_udc {
 
      struct mv_cap_regs __iomem      *cap_regs;
      struct mv_op_regs __iomem       *op_regs;
-     void __iomem                    *phy_regs;
      unsigned int                    max_eps;
      struct mv_dqh                   *ep_dqh;
      size_t                          ep_dqh_size;
@@ -217,6 +216,7 @@ struct mv_udc {
      struct work_struct      vbus_work;
      struct workqueue_struct *qwork;
 
+     struct usb_phy          *phy;
      struct usb_phy          *transceiver;
 
      struct mv_usb_platform_data     *pdata;
drivers/usb/gadget/mv_udc_core.c +1 -0
index c8cf959..4876d2f 100644
@@ -35,6 +35,7 @@
 #include <linux/platform_device.h>
 #include <linux/clk.h>
 #include <linux/platform_data/mv_usb.h>
+#include <linux/usb/mv_usb2.h>
 #include <asm/unaligned.h>
 
 #include "mv_udc.h"
@@ -1121,15 +1122,14 @@ static int mv_udc_enable_internal(struct mv_udc *udc)
 

drivers/usb/gadget/mv_udc_core.c static int mv_udc_enable_internal(struct mv_udc *udc) · line 1131

Quoted 10 lines
drivers/usb/gadget/mv_udc_core.c +0 -8
      dev_dbg(&udc->dev->dev, "enable udc\n");
      udc_clock_enable(udc);
-     if (udc->pdata->phy_init) {
-             retval = udc->pdata->phy_init(udc->phy_regs);
-             if (retval) {
-                     dev_err(&udc->dev->dev,
-                             "init phy error %d\n", retval);
-                     udc_clock_disable(udc);
-                     return retval;
-             }

dude, you really don't test your patches, do you ? Your previous patch removed ->phy_init(), ->phy_deinit() and ->private_init() from your platform_data.

Quoted 9 lines

Are you seriously telling me you didn't compile test your patches ?

Rewrite your series again and make sure that each and every patch compiles and works on its own. You shouldn't cause any regressions or build breaks or build warnings while converting this driver.

I won't review the rest of this series and I'm purging it from my TODO list.

Thanks for your review.

The patches are seperated into two parts. for each part it includes some patches.
I have compiled and tested each part for our configurartion and x86 configuration, but
not for every patch. i will do it.

Quoted 4 lines

cheers

--
balbi

author
#19 lore raw

PATCH v8 3/16 usb: ehci: ehci-mv: use PHY driver for ehci

In reply to Chao Xie

Originaly, ehci driver will call the callbacks in platform data for PHY initialization and shut down.
With PHY driver, it will call the APIs provided by PHY driver for PHY initialization and shut down. It removes the callbacks in platform data, and at same time it removes one block in the way of enabling device tree for ehci driver.

---
 drivers/usb/host/ehci-mv.c |   48 ++++++++++++++-----------------------------
 1 files changed, 16 insertions(+), 32 deletions(-)
drivers/usb/host/ehci-mv.c +16 -32
index 3065809..3e89bd4 100644
@@ -15,17 +15,18 @@
 #include <linux/clk.h>
 #include <linux/err.h>
 #include <linux/usb/otg.h>
+#include <linux/usb/mv_usb2.h>
 #include <linux/platform_data/mv_usb.h>
 
 #define CAPLENGTH_MASK         (0xff)
 
 struct ehci_hcd_mv {
 	struct usb_hcd *hcd;
+	struct usb_phy *phy;
 
 	/* Which mode does this ehci running OTG/Host ? */
 	int mode;
 
-	void __iomem *phy_regs;
 	void __iomem *cap_regs;
 	void __iomem *op_regs;
 
@@ -56,22 +57,15 @@ static void ehci_clock_disable(struct ehci_hcd_mv *ehci_mv)
 
 static int mv_ehci_enable(struct ehci_hcd_mv *ehci_mv)
 {
-	int retval;
-
 	ehci_clock_enable(ehci_mv);
-	if (ehci_mv->pdata->phy_init) {
-		retval = ehci_mv->pdata->phy_init(ehci_mv->phy_regs);
-		if (retval)
-			return retval;
-	}
 
-	return 0;
+	return usb_phy_init(ehci_mv->phy);
 }
 
 static void mv_ehci_disable(struct ehci_hcd_mv *ehci_mv)
 {
-	if (ehci_mv->pdata->phy_deinit)
-		ehci_mv->pdata->phy_deinit(ehci_mv->phy_regs);
+	usb_phy_shutdown(ehci_mv->phy);
+
 	ehci_clock_disable(ehci_mv);
 }
 
@@ -184,22 +178,7 @@ static int mv_ehci_probe(struct platform_device *pdev)
 		}
 	}
 
-	r = platform_get_resource_byname(pdev, IORESOURCE_MEM, "phyregs");
-	if (r == NULL) {
-		dev_err(&pdev->dev, "no phy I/O memory resource defined\n");
-		retval = -ENODEV;
-		goto err_clear_drvdata;
-	}
-
-	ehci_mv->phy_regs = devm_ioremap(&pdev->dev, r->start,
-					 resource_size(r));
-	if (ehci_mv->phy_regs == 0) {
-		dev_err(&pdev->dev, "failed to map phy I/O memory\n");
-		retval = -EFAULT;
-		goto err_clear_drvdata;
-	}
-
-	r = platform_get_resource_byname(pdev, IORESOURCE_MEM, "capregs");
+	r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
 	if (!r) {
 		dev_err(&pdev->dev, "no I/O memory resource defined\n");
 		retval = -ENODEV;
@@ -214,6 +193,13 @@ static int mv_ehci_probe(struct platform_device *pdev)
 		goto err_clear_drvdata;
 	}
 
+	ehci_mv->phy = devm_usb_get_phy_dev(&pdev->dev, MV_USB2_PHY_INDEX);
+	if (IS_ERR_OR_NULL(ehci_mv->phy)) {
+		dev_err(&pdev->dev, "failed to get the outer phy\n");
+		retval = PTR_ERR(ehci_mv->phy);
+		goto err_clear_drvdata;
+	}
+
 	retval = mv_ehci_enable(ehci_mv);
 	if (retval) {
 		dev_err(&pdev->dev, "init phy error %d\n", retval);
@@ -241,11 +227,12 @@ static int mv_ehci_probe(struct platform_device *pdev)
 	ehci_mv->mode = pdata->mode;
 	if (ehci_mv->mode == MV_USB_MODE_OTG) {
 #ifdef CONFIG_USB_OTG_UTILS
-		ehci_mv->otg = devm_usb_get_phy(&pdev->dev, USB_PHY_TYPE_USB2);
+		ehci_mv->otg = devm_usb_get_phy_dev(&pdev->dev,
+						MV_USB2_OTG_PHY_INDEX);
 		if (IS_ERR_OR_NULL(ehci_mv->otg)) {
 			dev_err(&pdev->dev,
 				"unable to find transceiver\n");
-			retval = -ENODEV;
+			retval = PTR_ERR(ehci_mv->otg);
 			goto err_disable_clk;
 		}
 
@@ -275,9 +262,6 @@ static int mv_ehci_probe(struct platform_device *pdev)
 		}
 	}
 
-	if (pdata->private_init)
-		pdata->private_init(ehci_mv->op_regs, ehci_mv->phy_regs);
-
 	dev_info(&pdev->dev,
 		 "successful find EHCI device with regs 0x%p irq %d"
 		 " working in %s mode\n", hcd->regs, hcd->irq,
Signature · 1 line
1.7.4.1
author
#20 lore raw

PATCH v8 4/16 usb: otg: mv_otg: use PHY driver for otg

In reply to Chao Xie

Originaly, otg driver will call the callbacks in platform data for PHY initialization and shut down.
With PHY driver, it will call the APIs provided by PHY driver for PHY initialization and shut down. It removes the callbacks in platform data, and at same time it removes one block in the way of enabling device tree for otg driver.

---
 drivers/usb/otg/mv_otg.c |   50 ++++++++++++++++++---------------------------
 drivers/usb/otg/mv_otg.h |    2 +-
 2 files changed, 21 insertions(+), 31 deletions(-)
drivers/usb/otg/mv_otg.c +20 -30
index b6a9be3..7282b0d 100644
@@ -25,6 +25,7 @@
 #include <linux/usb/otg.h>
 #include <linux/usb/gadget.h>
 #include <linux/usb/hcd.h>
+#include <linux/usb/mv_usb2.h>
 #include <linux/platform_data/mv_usb.h>
 
 #include "mv_otg.h"
@@ -261,14 +262,12 @@ static int mv_otg_enable_internal(struct mv_otg *mvotg)
 	dev_dbg(&mvotg->pdev->dev, "otg enabled\n");
 
 	otg_clock_enable(mvotg);
-	if (mvotg->pdata->phy_init) {
-		retval = mvotg->pdata->phy_init(mvotg->phy_regs);
-		if (retval) {
-			dev_err(&mvotg->pdev->dev,
-				"init phy error %d\n", retval);
-			otg_clock_disable(mvotg);
-			return retval;
-		}
+	retval = usb_phy_init(mvotg->outer_phy);
+	if (retval) {
+		dev_err(&mvotg->pdev->dev,
+			"init phy error %d\n", retval);
+		otg_clock_disable(mvotg);
+		return retval;
 	}
 	mvotg->active = 1;
 
@@ -288,8 +287,7 @@ static void mv_otg_disable_internal(struct mv_otg *mvotg)
 {
 	if (mvotg->active) {
 		dev_dbg(&mvotg->pdev->dev, "otg disabled\n");
-		if (mvotg->pdata->phy_deinit)
-			mvotg->pdata->phy_deinit(mvotg->phy_regs);
+		usb_phy_shutdown(mvotg->outer_phy);
 		otg_clock_disable(mvotg);
 		mvotg->active = 0;
 	}
@@ -729,6 +727,7 @@ static int mv_otg_probe(struct platform_device *pdev)
 	/* OTG common part */
 	mvotg->pdev = pdev;
 	mvotg->phy.dev = &pdev->dev;
+	mvotg->phy.type = USB_PHY_TYPE_USB2;
 	mvotg->phy.otg = otg;
 	mvotg->phy.label = driver_name;
 	mvotg->phy.state = OTG_STATE_UNDEFINED;
@@ -741,23 +740,8 @@ static int mv_otg_probe(struct platform_device *pdev)
 	for (i = 0; i < OTG_TIMER_NUM; i++)
 		init_timer(&mvotg->otg_ctrl.timer[i]);
 
-	r = platform_get_resource_byname(mvotg->pdev,
-					 IORESOURCE_MEM, "phyregs");
-	if (r == NULL) {
-		dev_err(&pdev->dev, "no phy I/O memory resource defined\n");
-		retval = -ENODEV;
-		goto err_destroy_workqueue;
-	}
-
-	mvotg->phy_regs = devm_ioremap(&pdev->dev, r->start, resource_size(r));
-	if (mvotg->phy_regs == NULL) {
-		dev_err(&pdev->dev, "failed to map phy I/O memory\n");
-		retval = -EFAULT;
-		goto err_destroy_workqueue;
-	}
-
-	r = platform_get_resource_byname(mvotg->pdev,
-					 IORESOURCE_MEM, "capregs");
+	r = platform_get_resource(mvotg->pdev,
+					 IORESOURCE_MEM, 0);
 	if (r == NULL) {
 		dev_err(&pdev->dev, "no I/O memory resource defined\n");
 		retval = -ENODEV;
@@ -770,6 +754,12 @@ static int mv_otg_probe(struct platform_device *pdev)
 		retval = -EFAULT;
 		goto err_destroy_workqueue;
 	}
+	mvotg->outer_phy = devm_usb_get_phy_dev(&pdev->dev, MV_USB2_PHY_INDEX);
+	if (IS_ERR_OR_NULL(mvotg->outer_phy)) {
+		dev_err(&pdev->dev, "can not find outer phy\n");
+		retval = PTR_ERR(mvotg->outer_phy);
+		goto err_destroy_workqueue;
+	}
 
 	/* we will acces controller register, so enable the udc controller */
 	retval = mv_otg_enable_internal(mvotg);
@@ -830,7 +820,7 @@ static int mv_otg_probe(struct platform_device *pdev)
 		goto err_disable_clk;
 	}
 
-	retval = usb_add_phy(&mvotg->phy, USB_PHY_TYPE_USB2);
+	retval = usb_add_phy_dev(&mvotg->phy);
 	if (retval < 0) {
 		dev_err(&pdev->dev, "can't register transceiver, %d\n",
 			retval);
@@ -841,7 +831,7 @@ static int mv_otg_probe(struct platform_device *pdev)
 	if (retval < 0) {
 		dev_dbg(&pdev->dev,
 			"Can't register sysfs attr group: %d\n", retval);
-		goto err_remove_phy;
+		goto err_remove_otg_phy;
 	}
 
 	spin_lock_init(&mvotg->wq_lock);
@@ -856,7 +846,7 @@ static int mv_otg_probe(struct platform_device *pdev)
 
 	return 0;
 
-err_remove_phy:
+err_remove_otg_phy:
 	usb_remove_phy(&mvotg->phy);
 err_disable_clk:
 	mv_otg_disable_internal(mvotg);
drivers/usb/otg/mv_otg.h +1 -1
index 8a9e351..65e7fd1 100644
@@ -137,10 +137,10 @@ struct mv_otg_regs {
 
 struct mv_otg {
 	struct usb_phy phy;
+	struct usb_phy *outer_phy;
 	struct mv_otg_ctrl otg_ctrl;
 
 	/* base address */
-	void __iomem *phy_regs;
 	void __iomem *cap_regs;
 	struct mv_otg_regs __iomem *op_regs;
 
Signature · 1 line
1.7.4.1
author
#21 lore raw

PATCH v8 5/16 arm: mmp2: change the defintion of usb devices

In reply to Chao Xie

add the udc/otg/ehci devices for mmp2

---
 arch/arm/mach-mmp/include/mach/mmp2.h |    4 ++++
 arch/arm/mach-mmp/mmp2.c              |    4 ++++
 2 files changed, 8 insertions(+), 0 deletions(-)
arch/arm/mach-mmp/include/mach/mmp2.h +4 -0
index c4ca4d1..58e96b0 100644
@@ -31,6 +31,10 @@ extern struct pxa_device_desc mmp2_device_sdh2;
 extern struct pxa_device_desc mmp2_device_sdh3;
 extern struct pxa_device_desc mmp2_device_asram;
 extern struct pxa_device_desc mmp2_device_isram;
+extern struct pxa_device_desc mmp2_device_u2o;
+extern struct pxa_device_desc mmp2_device_u2ootg;
+extern struct pxa_device_desc mmp2_device_u2oehci;
+extern struct pxa_device_desc mmp2_device_u2ophy;
 
 extern struct platform_device mmp2_device_gpio;
 
arch/arm/mach-mmp/mmp2.c +4 -0
index 3a3768c..73edbfc 100644
@@ -153,6 +153,10 @@ MMP2_DEVICE(sdh3, "sdhci-pxav3", 3, MMC4, 0xd4281800, 0x120);
 MMP2_DEVICE(asram, "asram", -1, NONE, 0xe0000000, 0x4000);
 /* 0xd1000000 ~ 0xd101ffff is reserved for secure processor */
 MMP2_DEVICE(isram, "isram", -1, NONE, 0xd1020000, 0x18000);
+MMP2_DEVICE(u2ophy, "mmp2-usb-phy", -1, NONE, 0xd4207000, 0x1ff);
+MMP2_DEVICE(u2o, "mv-udc", -1, USB_OTG, 0xd4208100, 0x1ff);
+MMP2_DEVICE(u2ootg, "mv-otg", -1, USB_OTG, 0xd4208100, 0x1ff);
+MMP2_DEVICE(u2oehci, "mv-ehci", -1, USB_OTG, 0xd4208100, 0x1ff);
 
 struct resource mmp2_resource_gpio[] = {
 	{
Signature · 1 line
1.7.4.1
author
#22 lore raw

PATCH v8 6/16 arm: pxa910: change the defintion of usb devices

In reply to Chao Xie

because phy is seperated from the usb controller driver, we can use the common pxa_device_desc for device register.

---
 arch/arm/mach-mmp/include/mach/pxa910.h |    7 ++++---
 arch/arm/mach-mmp/pxa910.c              |    4 ++++
 2 files changed, 8 insertions(+), 3 deletions(-)
arch/arm/mach-mmp/include/mach/pxa910.h +4 -3
index 3b58a3b..26ea4fe 100644
@@ -20,9 +20,10 @@ extern struct pxa_device_desc pxa910_device_pwm2;
 extern struct pxa_device_desc pxa910_device_pwm3;
 extern struct pxa_device_desc pxa910_device_pwm4;
 extern struct pxa_device_desc pxa910_device_nand;
-extern struct platform_device pxa168_device_u2o;
-extern struct platform_device pxa168_device_u2ootg;
-extern struct platform_device pxa168_device_u2oehci;
+extern struct pxa_device_desc pxa910_device_u2o;
+extern struct pxa_device_desc pxa910_device_u2ootg;
+extern struct pxa_device_desc pxa910_device_u2oehci;
+extern struct pxa_device_desc pxa910_device_u2ophy;
 
 extern struct platform_device pxa910_device_gpio;
 extern struct platform_device pxa910_device_rtc;
arch/arm/mach-mmp/pxa910.c +4 -0
index 8b1e16f..65174f7 100644
@@ -138,6 +138,10 @@ PXA910_DEVICE(pwm2, "pxa910-pwm", 1, NONE, 0xd401a400, 0x10);
 PXA910_DEVICE(pwm3, "pxa910-pwm", 2, NONE, 0xd401a800, 0x10);
 PXA910_DEVICE(pwm4, "pxa910-pwm", 3, NONE, 0xd401ac00, 0x10);
 PXA910_DEVICE(nand, "pxa3xx-nand", -1, NAND, 0xd4283000, 0x80, 97, 99);
+PXA910_DEVICE(u2ophy, "pxa910-usb-phy", -1, NONE, 0xd4207000, 0x1ff);
+PXA910_DEVICE(u2o, "mv-udc", -1, USB1, 0xd4208100, 0x1ff);
+PXA910_DEVICE(u2ootg, "mv-otg", -1, USB1, 0xd4208100, 0x1ff);
+PXA910_DEVICE(u2oehci, "mv-ehci", -1, USB1, 0xd4208100, 0x1ff);
 
 struct resource pxa910_resource_gpio[] = {
 	{
Signature · 1 line
1.7.4.1
author
#23 lore raw

PATCH v8 7/16 arm: brownstone: add usb support for the board

In reply to Chao Xie

for brownstone board, add the udc/otg/ehci support

---
 arch/arm/mach-mmp/brownstone.c |   68 ++++++++++++++++++++++++++++++++++++++++
 1 files changed, 68 insertions(+), 0 deletions(-)
arch/arm/mach-mmp/brownstone.c +68 -0
index 5cb769c..f84613a 100644
@@ -18,6 +18,9 @@
 #include <linux/regulator/max8649.h>
 #include <linux/regulator/fixed.h>
 #include <linux/mfd/max8925.h>
+#include <linux/usb/phy.h>
+#include <linux/usb/mv_usb2.h>
+#include <linux/platform_data/mv_usb.h>
 
 #include <asm/mach-types.h>
 #include <asm/mach/arch.h>
@@ -195,6 +198,31 @@ static struct sram_platdata mmp2_isram_platdata = {
 	.granularity	= SRAM_GRANULARITY,
 };
 
+#ifdef CONFIG_USB_SUPPORT
+
+static char *mmp2_usb_clock_name[] = {
+	[0] = "usb_clk",
+};
+
+static struct mv_usb_phy_platform_data brownstone_usb_phy_pdata = {
+	.clknum		= 1,
+	.clkname	= mmp2_usb_clock_name,
+};
+
+#if defined(CONFIG_USB_MV_UDC) || defined(CONFIG_USB_EHCI_MV_U2O)
+
+static struct mv_usb_platform_data brownstone_usb_pdata = {
+	.clknum		= 1,
+	.clkname	= mmp2_usb_clock_name,
+	.vbus		= NULL,
+	.mode		= MV_USB_MODE_OTG,
+	.otg_force_a_bus_req = 1,
+	.set_vbus	= NULL,
+};
+#endif
+#endif
+
+
 static void __init brownstone_init(void)
 {
 	mfp_config(ARRAY_AND_SIZE(brownstone_pin_config));
@@ -211,6 +239,46 @@ static void __init brownstone_init(void)
 
 	/* enable 5v regulator */
 	platform_device_register(&brownstone_v_5vp_device);
+
+#ifdef CONFIG_USB_SUPPORT
+	pxa_register_device(&mmp2_device_u2ophy, &brownstone_usb_phy_pdata,
+				sizeof(brownstone_usb_phy_pdata));
+#endif
+
+#ifdef CONFIG_USB_MV_UDC
+	/* for usb2 phy */
+	usb_bind_phy(mmp2_device_u2o.drv_name, MV_USB2_PHY_INDEX,
+				mmp2_device_u2ophy.drv_name);
+#ifdef CONFIG_USB_MV_OTG
+	/* for usb2 otg phy */
+	usb_bind_phy(mmp2_device_u2o.drv_name, MV_USB2_OTG_PHY_INDEX,
+				mmp2_device_u2ootg.drv_name);
+#endif
+	pxa_register_device(&mmp2_device_u2o, &brownstone_usb_pdata,
+				sizeof(brownstone_usb_pdata));
+#endif
+
+#ifdef CONFIG_USB_EHCI_MV_U2O
+	/* for usb2 phy */
+	usb_bind_phy(mmp2_device_u2oehci.dev_name, MV_USB2_PHY_INDEX,
+				mmp2_device_u2ophy.dev_name);
+#ifdef CONFIG_USB_MV_OTG
+	/* for usb2 otg phy */
+	usb_bind_phy(mmp2_device_u2oehci.drv_name, MV_USB2_OTG_PHY_INDEX,
+				mmp2_device_u2ootg.drv_name);
+#endif
+	pxa_register_device(&mmp2_device_u2oehci, &brownstone_usb_pdata,
+				sizeof(brownstone_usb_pdata));
+#endif
+
+#ifdef CONFIG_USB_MV_OTG
+	/* for usb2 phy */
+	usb_bind_phy(mmp2_device_u2ootg.dev_name, MV_USB2_PHY_INDEX,
+				mmp2_device_u2ophy.dev_name);
+	pxa_register_device(&mmp2_device_u2ootg, &brownstone_usb_pdata,
+				sizeof(brownstone_usb_pdata));
+#endif
+
 }
 
 MACHINE_START(BROWNSTONE, "Brownstone Development Platform")
Signature · 1 line
1.7.4.1
author
#24 lore raw

PATCH v8 8/16 arm: ttc_dkb: add usb support

In reply to Chao Xie

for ttc_dkb board, add udc/otg/ehci support

---
 arch/arm/mach-mmp/ttc_dkb.c |   50 +++++++++++++++++++++++++++++++++---------
 1 files changed, 39 insertions(+), 11 deletions(-)
arch/arm/mach-mmp/ttc_dkb.c +39 -11
index ce55fd8..5622092 100644
@@ -18,6 +18,8 @@
 #include <linux/i2c/pca953x.h>
 #include <linux/gpio.h>
 #include <linux/mfd/88pm860x.h>
+#include <linux/usb/phy.h>
+#include <linux/usb/mv_usb2.h>
 #include <linux/platform_data/mv_usb.h>
 
 #include <asm/mach-types.h>
@@ -27,7 +29,6 @@
 #include <mach/mfp-pxa910.h>
 #include <mach/pxa910.h>
 #include <mach/irqs.h>
-#include <mach/regs-usb.h>
 
 #include "common.h"
 
@@ -158,20 +159,24 @@ static struct i2c_board_info ttc_dkb_i2c_info[] = {
 };
 
 #ifdef CONFIG_USB_SUPPORT
-#if defined(CONFIG_USB_MV_UDC) || defined(CONFIG_USB_EHCI_MV_U2O)
 
 static char *pxa910_usb_clock_name[] = {
-	[0] = "U2OCLK",
+	[0] = "usb_clk",
 };
 
+static struct mv_usb_phy_platform_data ttc_usb_phy_pdata = {
+	.clknum		= 1,
+	.clkname	= pxa910_usb_clock_name,
+};
+
+#if defined(CONFIG_USB_MV_UDC) || defined(CONFIG_USB_EHCI_MV_U2O)
+
 static struct mv_usb_platform_data ttc_usb_pdata = {
 	.clknum		= 1,
 	.clkname	= pxa910_usb_clock_name,
 	.vbus		= NULL,
 	.mode		= MV_USB_MODE_OTG,
 	.otg_force_a_bus_req = 1,
-	.phy_init	= pxa_usb_phy_init,
-	.phy_deinit	= pxa_usb_phy_deinit,
 	.set_vbus	= NULL,
 };
 #endif
@@ -198,19 +203,42 @@ static void __init ttc_dkb_init(void)
 	pxa910_add_twsi(0, NULL, ARRAY_AND_SIZE(ttc_dkb_i2c_info));
 	platform_add_devices(ARRAY_AND_SIZE(ttc_dkb_devices));
 
+#ifdef CONFIG_USB_SUPPORT
+	pxa_register_device(&pxa910_device_u2ophy, &ttc_usb_phy_pdata,
+				sizeof(ttc_usb_phy_pdata));
+#endif
+
 #ifdef CONFIG_USB_MV_UDC
-	pxa168_device_u2o.dev.platform_data = &ttc_usb_pdata;
-	platform_device_register(&pxa168_device_u2o);
+	/* for usb2 phy */
+	usb_bind_phy(pxa910_device_u2o.drv_name, MV_USB2_PHY_INDEX,
+				pxa910_device_u2ophy.drv_name);
+#ifdef CONFIG_USB_MV_OTG
+	/* for usb2 otg phy */
+	usb_bind_phy(pxa910_device_u2o.drv_name, MV_USB2_OTG_PHY_INDEX,
+				pxa910_device_u2ootg.drv_name);
+#endif
+	pxa_register_device(&pxa910_device_u2o, &ttc_usb_pdata,
+				sizeof(ttc_usb_pdata));
 #endif
 
 #ifdef CONFIG_USB_EHCI_MV_U2O
-	pxa168_device_u2oehci.dev.platform_data = &ttc_usb_pdata;
-	platform_device_register(&pxa168_device_u2oehci);
+	/* for usb2 phy */
+	usb_bind_phy(pxa910_device_u2oehci.drv_name, MV_USB2_PHY_INDEX,
+				pxa910_device_u2ophy.drv_name);
+#ifdef CONFIG_USB_MV_OTG
+	/* for usb2 otg phy */
+	usb_bind_phy(pxa910_device_u2oehci.drv_name, MV_USB2_OTG_PHY_INDEX,
+				pxa910_device_u2ootg.drv_name);
+#endif
+	pxa_register_device(&pxa910_device_u2oehci, &ttc_usb_pdata,
+				sizeof(ttc_usb_pdata));
 #endif
 
 #ifdef CONFIG_USB_MV_OTG
-	pxa168_device_u2ootg.dev.platform_data = &ttc_usb_pdata;
-	platform_device_register(&pxa168_device_u2ootg);
+	usb_bind_phy(pxa910_device_u2ootg.drv_name, MV_USB2_PHY_INDEX,
+				pxa910_device_u2ophy.drv_name);
+	pxa_register_device(&pxa910_device_u2ootg, &ttc_usb_pdata,
+				sizeof(ttc_usb_pdata));
 #endif
 }
 
Signature · 1 line
1.7.4.1
author
#25 lore raw

PATCH v8 9/16 arm: mmp: remove the usb phy setting

In reply to Chao Xie

phy setting are formatted into a phy driver at drivers/usb/phy, we do not need do the setting in SOC files.

---
 arch/arm/mach-mmp/devices.c               |  278 -----------------------------
 arch/arm/mach-mmp/include/mach/regs-usb.h |  253 --------------------------
 2 files changed, 0 insertions(+), 531 deletions(-)
 delete mode 100644 arch/arm/mach-mmp/include/mach/regs-usb.h
arch/arm/mach-mmp/devices.c +0 -278
index dd2d8b1..af341e5 100644
@@ -15,7 +15,6 @@
 #include <mach/irqs.h>
 #include <mach/devices.h>
 #include <mach/cputype.h>
-#include <mach/regs-usb.h>
 
 int __init pxa_register_device(struct pxa_device_desc *desc,
 				void *data, size_t size)
@@ -72,280 +71,3 @@ int __init pxa_register_device(struct pxa_device_desc *desc,
 	return platform_device_add(pdev);
 }
 
-#if defined(CONFIG_USB) || defined(CONFIG_USB_GADGET)
-
-/*****************************************************************************
- * The registers read/write routines
- *****************************************************************************/
-
-static unsigned int u2o_get(void __iomem *base, unsigned int offset)
-{
-	return readl_relaxed(base + offset);
-}
-
-static void u2o_set(void __iomem *base, unsigned int offset,
-		unsigned int value)
-{
-	u32 reg;
-
-	reg = readl_relaxed(base + offset);
-	reg |= value;
-	writel_relaxed(reg, base + offset);
-	readl_relaxed(base + offset);
-}
-
-static void u2o_clear(void __iomem *base, unsigned int offset,
-		unsigned int value)
-{
-	u32 reg;
-
-	reg = readl_relaxed(base + offset);
-	reg &= ~value;
-	writel_relaxed(reg, base + offset);
-	readl_relaxed(base + offset);
-}
-
-static void u2o_write(void __iomem *base, unsigned int offset,
-		unsigned int value)
-{
-	writel_relaxed(value, base + offset);
-	readl_relaxed(base + offset);
-}
-
-#if defined(CONFIG_USB_MV_UDC) || defined(CONFIG_USB_EHCI_MV)
-
-#if defined(CONFIG_CPU_PXA910) || defined(CONFIG_CPU_PXA168)
-
-static DEFINE_MUTEX(phy_lock);
-static int phy_init_cnt;
-
-static int usb_phy_init_internal(void __iomem *base)
-{
-	int loops;
-
-	pr_info("Init usb phy!!!\n");
-
-	/* Initialize the USB PHY power */
-	if (cpu_is_pxa910()) {
-		u2o_set(base, UTMI_CTRL, (1<<UTMI_CTRL_INPKT_DELAY_SOF_SHIFT)
-			| (1<<UTMI_CTRL_PU_REF_SHIFT));
-	}
-
-	u2o_set(base, UTMI_CTRL, 1<<UTMI_CTRL_PLL_PWR_UP_SHIFT);
-	u2o_set(base, UTMI_CTRL, 1<<UTMI_CTRL_PWR_UP_SHIFT);
-
-	/* UTMI_PLL settings */
-	u2o_clear(base, UTMI_PLL, UTMI_PLL_PLLVDD18_MASK
-		| UTMI_PLL_PLLVDD12_MASK | UTMI_PLL_PLLCALI12_MASK
-		| UTMI_PLL_FBDIV_MASK | UTMI_PLL_REFDIV_MASK
-		| UTMI_PLL_ICP_MASK | UTMI_PLL_KVCO_MASK);
-
-	u2o_set(base, UTMI_PLL, 0xee<<UTMI_PLL_FBDIV_SHIFT
-		| 0xb<<UTMI_PLL_REFDIV_SHIFT | 3<<UTMI_PLL_PLLVDD18_SHIFT
-		| 3<<UTMI_PLL_PLLVDD12_SHIFT | 3<<UTMI_PLL_PLLCALI12_SHIFT
-		| 1<<UTMI_PLL_ICP_SHIFT | 3<<UTMI_PLL_KVCO_SHIFT);
-
-	/* UTMI_TX */
-	u2o_clear(base, UTMI_TX, UTMI_TX_REG_EXT_FS_RCAL_EN_MASK
-		| UTMI_TX_TXVDD12_MASK | UTMI_TX_CK60_PHSEL_MASK
-		| UTMI_TX_IMPCAL_VTH_MASK | UTMI_TX_REG_EXT_FS_RCAL_MASK
-		| UTMI_TX_AMP_MASK);
-	u2o_set(base, UTMI_TX, 3<<UTMI_TX_TXVDD12_SHIFT
-		| 4<<UTMI_TX_CK60_PHSEL_SHIFT | 4<<UTMI_TX_IMPCAL_VTH_SHIFT
-		| 8<<UTMI_TX_REG_EXT_FS_RCAL_SHIFT | 3<<UTMI_TX_AMP_SHIFT);
-
-	/* UTMI_RX */
-	u2o_clear(base, UTMI_RX, UTMI_RX_SQ_THRESH_MASK
-		| UTMI_REG_SQ_LENGTH_MASK);
-	u2o_set(base, UTMI_RX, 7<<UTMI_RX_SQ_THRESH_SHIFT
-		| 2<<UTMI_REG_SQ_LENGTH_SHIFT);
-
-	/* UTMI_IVREF */
-	if (cpu_is_pxa168())
-		/* fixing Microsoft Altair board interface with NEC hub issue -
-		 * Set UTMI_IVREF from 0x4a3 to 0x4bf */
-		u2o_write(base, UTMI_IVREF, 0x4bf);
-
-	/* toggle VCOCAL_START bit of UTMI_PLL */
-	udelay(200);
-	u2o_set(base, UTMI_PLL, VCOCAL_START);
-	udelay(40);
-	u2o_clear(base, UTMI_PLL, VCOCAL_START);
-
-	/* toggle REG_RCAL_START bit of UTMI_TX */
-	udelay(400);
-	u2o_set(base, UTMI_TX, REG_RCAL_START);
-	udelay(40);
-	u2o_clear(base, UTMI_TX, REG_RCAL_START);
-	udelay(400);
-
-	/* Make sure PHY PLL is ready */
-	loops = 0;
-	while ((u2o_get(base, UTMI_PLL) & PLL_READY) == 0) {
-		mdelay(1);
-		loops++;
-		if (loops > 100) {
-			printk(KERN_WARNING "calibrate timeout, UTMI_PLL %x\n",
-				u2o_get(base, UTMI_PLL));
-			break;
-		}
-	}
-
-	if (cpu_is_pxa168()) {
-		u2o_set(base, UTMI_RESERVE, 1 << 5);
-		/* Turn on UTMI PHY OTG extension */
-		u2o_write(base, UTMI_OTG_ADDON, 1);
-	}
-
-	return 0;
-}
-
-static int usb_phy_deinit_internal(void __iomem *base)
-{
-	pr_info("Deinit usb phy!!!\n");
-
-	if (cpu_is_pxa168())
-		u2o_clear(base, UTMI_OTG_ADDON, UTMI_OTG_ADDON_OTG_ON);
-
-	u2o_clear(base, UTMI_CTRL, UTMI_CTRL_RXBUF_PDWN);
-	u2o_clear(base, UTMI_CTRL, UTMI_CTRL_TXBUF_PDWN);
-	u2o_clear(base, UTMI_CTRL, UTMI_CTRL_USB_CLK_EN);
-	u2o_clear(base, UTMI_CTRL, 1<<UTMI_CTRL_PWR_UP_SHIFT);
-	u2o_clear(base, UTMI_CTRL, 1<<UTMI_CTRL_PLL_PWR_UP_SHIFT);
-
-	return 0;
-}
-
-int pxa_usb_phy_init(void __iomem *phy_reg)
-{
-	mutex_lock(&phy_lock);
-	if (phy_init_cnt++ == 0)
-		usb_phy_init_internal(phy_reg);
-	mutex_unlock(&phy_lock);
-	return 0;
-}
-
-void pxa_usb_phy_deinit(void __iomem *phy_reg)
-{
-	WARN_ON(phy_init_cnt == 0);
-
-	mutex_lock(&phy_lock);
-	if (--phy_init_cnt == 0)
-		usb_phy_deinit_internal(phy_reg);
-	mutex_unlock(&phy_lock);
-}
-#endif
-#endif
-#endif
-
-#ifdef CONFIG_USB_SUPPORT
-static u64 usb_dma_mask = ~(u32)0;
-
-#ifdef CONFIG_USB_MV_UDC
-struct resource pxa168_u2o_resources[] = {
-	/* regbase */
-	[0] = {
-		.start	= PXA168_U2O_REGBASE + U2x_CAPREGS_OFFSET,
-		.end	= PXA168_U2O_REGBASE + USB_REG_RANGE,
-		.flags	= IORESOURCE_MEM,
-		.name	= "capregs",
-	},
-	/* phybase */
-	[1] = {
-		.start	= PXA168_U2O_PHYBASE,
-		.end	= PXA168_U2O_PHYBASE + USB_PHY_RANGE,
-		.flags	= IORESOURCE_MEM,
-		.name	= "phyregs",
-	},
-	[2] = {
-		.start	= IRQ_PXA168_USB1,
-		.end	= IRQ_PXA168_USB1,
-		.flags	= IORESOURCE_IRQ,
-	},
-};
-
-struct platform_device pxa168_device_u2o = {
-	.name		= "mv-udc",
-	.id		= -1,
-	.resource	= pxa168_u2o_resources,
-	.num_resources	= ARRAY_SIZE(pxa168_u2o_resources),
-	.dev		=  {
-		.dma_mask	= &usb_dma_mask,
-		.coherent_dma_mask = 0xffffffff,
-	}
-};
-#endif /* CONFIG_USB_MV_UDC */
-
-#ifdef CONFIG_USB_EHCI_MV_U2O
-struct resource pxa168_u2oehci_resources[] = {
-	/* regbase */
-	[0] = {
-		.start	= PXA168_U2O_REGBASE + U2x_CAPREGS_OFFSET,
-		.end	= PXA168_U2O_REGBASE + USB_REG_RANGE,
-		.flags	= IORESOURCE_MEM,
-		.name	= "capregs",
-	},
-	/* phybase */
-	[1] = {
-		.start	= PXA168_U2O_PHYBASE,
-		.end	= PXA168_U2O_PHYBASE + USB_PHY_RANGE,
-		.flags	= IORESOURCE_MEM,
-		.name	= "phyregs",
-	},
-	[2] = {
-		.start	= IRQ_PXA168_USB1,
-		.end	= IRQ_PXA168_USB1,
-		.flags	= IORESOURCE_IRQ,
-	},
-};
-
-struct platform_device pxa168_device_u2oehci = {
-	.name		= "pxa-u2oehci",
-	.id		= -1,
-	.dev		= {
-		.dma_mask		= &usb_dma_mask,
-		.coherent_dma_mask	= 0xffffffff,
-	},
-
-	.num_resources	= ARRAY_SIZE(pxa168_u2oehci_resources),
-	.resource	= pxa168_u2oehci_resources,
-};
-#endif
-
-#if defined(CONFIG_USB_MV_OTG)
-struct resource pxa168_u2ootg_resources[] = {
-	/* regbase */
-	[0] = {
-		.start	= PXA168_U2O_REGBASE + U2x_CAPREGS_OFFSET,
-		.end	= PXA168_U2O_REGBASE + USB_REG_RANGE,
-		.flags	= IORESOURCE_MEM,
-		.name	= "capregs",
-	},
-	/* phybase */
-	[1] = {
-		.start	= PXA168_U2O_PHYBASE,
-		.end	= PXA168_U2O_PHYBASE + USB_PHY_RANGE,
-		.flags	= IORESOURCE_MEM,
-		.name	= "phyregs",
-	},
-	[2] = {
-		.start	= IRQ_PXA168_USB1,
-		.end	= IRQ_PXA168_USB1,
-		.flags	= IORESOURCE_IRQ,
-	},
-};
-
-struct platform_device pxa168_device_u2ootg = {
-	.name		= "mv-otg",
-	.id		= -1,
-	.dev  = {
-		.dma_mask          = &usb_dma_mask,
-		.coherent_dma_mask = 0xffffffff,
-	},
-
-	.num_resources	= ARRAY_SIZE(pxa168_u2ootg_resources),
-	.resource      = pxa168_u2ootg_resources,
-};
-#endif /* CONFIG_USB_MV_OTG */
-
-#endif
arch/arm/mach-mmp/include/mach/regs-usb.h deleted +0 -253

253 lines not on this page · raw message

Signature · 1 line
1.7.4.1
author
#26 lore raw

PATCH v8 10/16 arm: mmp: remove usb devices from pxa168

In reply to Chao Xie
---
 arch/arm/mach-mmp/pxa168.c |   42 ------------------------------------------
 1 files changed, 0 insertions(+), 42 deletions(-)
arch/arm/mach-mmp/pxa168.c +0 -42
index b7f074f..dd3a68b 100644
@@ -28,7 +28,6 @@
 #include <mach/mfp.h>
 #include <linux/dma-mapping.h>
 #include <mach/pxa168.h>
-#include <mach/regs-usb.h>
 
 #include "common.h"
 #include "clock.h"
@@ -135,47 +134,6 @@ struct platform_device pxa168_device_gpio = {
 	.resource	= pxa168_resource_gpio,
 };
 
-struct resource pxa168_usb_host_resources[] = {
-	/* USB Host conroller register base */
-	[0] = {
-		.start	= PXA168_U2H_REGBASE + U2x_CAPREGS_OFFSET,
-		.end	= PXA168_U2H_REGBASE + USB_REG_RANGE,
-		.flags	= IORESOURCE_MEM,
-		.name	= "capregs",
-	},
-	/* USB PHY register base */
-	[1] = {
-		.start	= PXA168_U2H_PHYBASE,
-		.end	= PXA168_U2H_PHYBASE + USB_PHY_RANGE,
-		.flags	= IORESOURCE_MEM,
-		.name	= "phyregs",
-	},
-	[2] = {
-		.start	= IRQ_PXA168_USB2,
-		.end	= IRQ_PXA168_USB2,
-		.flags	= IORESOURCE_IRQ,
-	},
-};
-
-static u64 pxa168_usb_host_dmamask = DMA_BIT_MASK(32);
-struct platform_device pxa168_device_usb_host = {
-	.name = "pxa-sph",
-	.id   = -1,
-	.dev  = {
-		.dma_mask = &pxa168_usb_host_dmamask,
-		.coherent_dma_mask = DMA_BIT_MASK(32),
-	},
-
-	.num_resources = ARRAY_SIZE(pxa168_usb_host_resources),
-	.resource      = pxa168_usb_host_resources,
-};
-
-int __init pxa168_add_usb_host(struct mv_usb_platform_data *pdata)
-{
-	pxa168_device_usb_host.dev.platform_data = pdata;
-	return platform_device_register(&pxa168_device_usb_host);
-}
-
 void pxa168_restart(char mode, const char *cmd)
 {
 	soft_restart(0xffff0000);
Signature · 1 line
1.7.4.1
author
#27 lore raw

PATCH v8 11/16 usb: phy: mv_usb2_phy: add externel chip support

In reply to Chao Xie

For the vbus and idpin detection. It may be completed by some external chip, for example the pmic chip 88pm860x in driver/mfd can do it.
Although the usb controller can detect the vbus and id pin, but it need clock on and PHY enabled to detect the vbus/idpin. It will increase the power.
Using the external chip to detect vbus/idpin can save the power.

---
 drivers/usb/phy/mv_usb2_phy.c        |   49 +++++++++++++++++++
 include/linux/platform_data/mv_usb.h |   15 ++----
 include/linux/usb/mv_usb2.h          |   89 ++++++++++++++++++++++++++++++++++
 3 files changed, 143 insertions(+), 10 deletions(-)
drivers/usb/phy/mv_usb2_phy.c +49 -0
index a81e5e4..cbf207c 100644
@@ -129,6 +129,53 @@ enum mv_usb2_phy_type {
 	MMP2_USB,
 };
 
+int mv_usb2_register_notifier(struct mv_usb2_phy *phy,
+		struct notifier_block *nb)
+{
+	int ret;
+
+	if (!phy)
+		return -ENODEV;
+
+	ret = atomic_notifier_chain_register(&phy->extern_chip.head, nb);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+EXPORT_SYMBOL(mv_usb2_register_notifier);
+
+int mv_usb2_unregister_notifier(struct mv_usb2_phy *phy,
+		struct notifier_block *nb)
+{
+	int ret;
+
+	if (!phy)
+		return -ENODEV;
+
+	ret = atomic_notifier_chain_unregister(&phy->extern_chip.head, nb);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+EXPORT_SYMBOL(mv_usb2_unregister_notifier);
+
+int mv_usb2_notify(struct mv_usb2_phy *phy, unsigned long val, void *v)
+{
+	int ret;
+
+	if (!phy)
+		return -ENODEV;
+
+	ret = atomic_notifier_call_chain(&phy->extern_chip.head, val, v);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+EXPORT_SYMBOL(mv_usb2_notify);
+
 static unsigned int u2o_get(void __iomem *base, unsigned int offset)
 {
 	return readl(base + offset);
@@ -363,6 +410,8 @@ static int mv_usb2_phy_probe(struct platform_device *pdev)
 
 	usb_add_phy_dev(&mv_phy->phy);
 
+	ATOMIC_INIT_NOTIFIER_HEAD(&mv_phy->extern_chip.head);
+
 	platform_set_drvdata(pdev, mv_phy);
 
 	return 0;
include/linux/platform_data/mv_usb.h +5 -10
index fd3d1b4..dc25d60 100644
@@ -28,16 +28,13 @@ enum {
 	VBUS_HIGH	= 1 << 0,
 };
 
-struct mv_usb_addon_irq {
-	unsigned int	irq;
-	int		(*poll)(void);
-};
-
+#define MV_USB_HAS_VBUS_DETECTION	(1 << 0)
+#define MV_USB_HAS_IDPIN_DETECTION	(1 << 1)
 struct mv_usb_platform_data {
 	unsigned int		clknum;
 	char			**clkname;
-	struct mv_usb_addon_irq	*id;	/* Only valid for OTG. ID pin change*/
-	struct mv_usb_addon_irq	*vbus;	/* valid for OTG/UDC. VBUS change*/
+
+	unsigned int		extern_attr;
 
 	/* only valid for HCD. OTG or Host only*/
 	unsigned int		mode;
@@ -45,9 +42,7 @@ struct mv_usb_platform_data {
 	/* This flag is used for that needs id pin checked by otg */
 	unsigned int    disable_otg_clock_gating:1;
 	/* Force a_bus_req to be asserted */
-	 unsigned int    otg_force_a_bus_req:1;
-
-	int	(*set_vbus)(unsigned int vbus);
+	unsigned int    otg_force_a_bus_req:1;
 };
 
 struct mv_usb_phy_platform_data {
include/linux/usb/mv_usb2.h +89 -0

101 lines not on this page · raw message

Signature · 1 line
1.7.4.1
author
#28 lore raw

PATCH v8 12/16 usb: gadget: mv_udc: add extern chip support

In reply to Chao Xie

Because arch-mmp will make use of irq domain for irq allocation, the irqs allocated for PMIC is dynamical. The vbus/idpin irqs from PMIC can not be passed by platform data. Using the extern chip APIs provides by PHY driver can solve this problem.
Marvell usb PHY driver provides a middle layer.
The PMIC usb drivers will help to register the callbacks in the marvell usb PHY driver.
udc/otg/ehci driver will call the callbacks.
Then we do not need pass the information in platform data. It will remove another block in the way of enabling device tree for usb.

---
 drivers/usb/gadget/mv_udc.h      |    3 ++
 drivers/usb/gadget/mv_udc_core.c |   56 +++++++++++++++++++++++++-------------
 2 files changed, 40 insertions(+), 19 deletions(-)
drivers/usb/gadget/mv_udc.h +3 -0
index f339df4..5d44a6f 100644
@@ -178,6 +178,9 @@ struct mv_udc {
 	struct platform_device		*dev;
 	int				irq;
 
+	unsigned int			extern_attr;
+	struct notifier_block		notifier;
+
 	struct mv_cap_regs __iomem	*cap_regs;
 	struct mv_op_regs __iomem	*op_regs;
 	unsigned int			max_eps;
drivers/usb/gadget/mv_udc_core.c +37 -19

139 lines not on this page · raw message

Signature · 1 line
1.7.4.1
author
#29 lore raw

PATCH v8 13/16 usb: ehci: ehci-mv: add extern chip support

In reply to Chao Xie

It does the similar things as what we do for udc driver.

---
 drivers/usb/host/ehci-mv.c |   16 ++++++++++------
 1 files changed, 10 insertions(+), 6 deletions(-)
drivers/usb/host/ehci-mv.c +10 -6
index 3e89bd4..ae5d8dc 100644
@@ -138,6 +138,7 @@ static int mv_ehci_probe(struct platform_device *pdev)
 	struct ehci_hcd *ehci;
 	struct ehci_hcd_mv *ehci_mv;
 	struct resource *r;
+	struct mv_usb2_phy *mv_phy;
 	int clk_i, retval = -ENODEV;
 	u32 offset;
 	size_t size;
@@ -199,6 +200,7 @@ static int mv_ehci_probe(struct platform_device *pdev)
 		retval = PTR_ERR(ehci_mv->phy);
 		goto err_clear_drvdata;
 	}
+	mv_phy = container_of(ehci_mv->phy, struct mv_usb2_phy, phy);
 
 	retval = mv_ehci_enable(ehci_mv);
 	if (retval) {
@@ -251,8 +253,8 @@ static int mv_ehci_probe(struct platform_device *pdev)
 		goto err_disable_clk;
 #endif
 	} else {
-		if (pdata->set_vbus)
-			pdata->set_vbus(1);
+		if (mv_usb2_has_extern_call(mv_phy, vbus, set_vbus))
+			mv_usb2_extern_call(mv_phy, vbus, set_vbus, 1);
 
 		retval = usb_add_hcd(hcd, hcd->irq, IRQF_SHARED);
 		if (retval) {
@@ -270,8 +272,8 @@ static int mv_ehci_probe(struct platform_device *pdev)
 	return 0;
 
 err_set_vbus:
-	if (pdata->set_vbus)
-		pdata->set_vbus(0);
+	if (mv_usb2_has_extern_call(mv_phy, vbus, set_vbus))
+		mv_usb2_extern_call(mv_phy, vbus, set_vbus, 0);
 err_disable_clk:
 	mv_ehci_disable(ehci_mv);
 err_clear_drvdata:
@@ -286,6 +288,7 @@ static int mv_ehci_remove(struct platform_device *pdev)
 {
 	struct ehci_hcd_mv *ehci_mv = platform_get_drvdata(pdev);
 	struct usb_hcd *hcd = ehci_mv->hcd;
+	struct mv_usb2_phy *mv_phy;
 
 	if (hcd->rh_registered)
 		usb_remove_hcd(hcd);
@@ -293,9 +296,10 @@ static int mv_ehci_remove(struct platform_device *pdev)
 	if (!IS_ERR_OR_NULL(ehci_mv->otg))
 		otg_set_host(ehci_mv->otg->otg, NULL);
 
+	mv_phy = container_of(ehci_mv->phy, struct mv_usb2_phy, phy);
 	if (ehci_mv->mode == MV_USB_MODE_HOST) {
-		if (ehci_mv->pdata->set_vbus)
-			ehci_mv->pdata->set_vbus(0);
+		if (mv_usb2_has_extern_call(mv_phy, vbus, set_vbus))
+			mv_usb2_extern_call(mv_phy, vbus, set_vbus, 1);
 
 		mv_ehci_disable(ehci_mv);
 	}
Signature · 1 line
1.7.4.1
author
#30 lore raw

PATCH v8 14/16 usb: otg: mv_otg: add extern chip support

In reply to Chao Xie

It does the similar things as what we do for udc driver.

---
 drivers/usb/otg/mv_otg.c |   72 +++++++++++++++++++++++----------------------
 drivers/usb/otg/mv_otg.h |    3 ++
 2 files changed, 40 insertions(+), 35 deletions(-)
drivers/usb/otg/mv_otg.c +37 -35
index 7282b0d..9b3789a 100644
@@ -59,10 +59,13 @@ static char *state_string[] = {
 static int mv_otg_set_vbus(struct usb_otg *otg, bool on)
 {
 	struct mv_otg *mvotg = container_of(otg->phy, struct mv_otg, phy);
-	if (mvotg->pdata->set_vbus == NULL)
+	struct mv_usb2_phy *mv_phy = container_of(mvotg->outer_phy,
+					struct mv_usb2_phy, phy);
+
+	if (!mv_usb2_has_extern_call(mv_phy, vbus, set_vbus))
 		return -ENODEV;
 
-	return mvotg->pdata->set_vbus(on);
+	return mv_usb2_extern_call(mv_phy, vbus, set_vbus, on);
 }
 
 static int mv_otg_set_host(struct usb_otg *otg,
@@ -184,14 +187,14 @@ static void mv_otg_init_irq(struct mv_otg *mvotg)
 	mvotg->irq_status = OTGSC_INTSTS_A_SESSION_VALID
 	    | OTGSC_INTSTS_A_VBUS_VALID;
 
-	if (mvotg->pdata->vbus == NULL) {
+	if (!(mvotg->extern_attr & MV_USB_HAS_VBUS_DETECTION)) {
 		mvotg->irq_en |= OTGSC_INTR_B_SESSION_VALID
 		    | OTGSC_INTR_B_SESSION_END;
 		mvotg->irq_status |= OTGSC_INTSTS_B_SESSION_VALID
 		    | OTGSC_INTSTS_B_SESSION_END;
 	}
 
-	if (mvotg->pdata->id == NULL) {
+	if (!(mvotg->extern_attr & MV_USB_HAS_IDPIN_DETECTION)) {
 		mvotg->irq_en |= OTGSC_INTR_USB_ID;
 		mvotg->irq_status |= OTGSC_INTSTS_USB_ID;
 	}
@@ -303,11 +306,16 @@ static void mv_otg_update_inputs(struct mv_otg *mvotg)
 {
 	struct mv_otg_ctrl *otg_ctrl = &mvotg->otg_ctrl;
 	u32 otgsc;
+	unsigned int vbus, idpin;
+	struct mv_usb2_phy *mv_phy = container_of(mvotg->outer_phy,
+					struct mv_usb2_phy, phy);
 
 	otgsc = readl(&mvotg->op_regs->otgsc);
 
-	if (mvotg->pdata->vbus) {
-		if (mvotg->pdata->vbus->poll() == VBUS_HIGH) {
+	if (mvotg->extern_attr & MV_USB_HAS_VBUS_DETECTION) {
+		if (mv_usb2_extern_call(mv_phy, vbus, get_vbus, &vbus))
+			return;
+		if (vbus == VBUS_HIGH) {
 			otg_ctrl->b_sess_vld = 1;
 			otg_ctrl->b_sess_end = 0;
 		} else {
@@ -319,8 +327,11 @@ static void mv_otg_update_inputs(struct mv_otg *mvotg)
 		otg_ctrl->b_sess_end = !!(otgsc & OTGSC_STS_B_SESSION_END);
 	}
 
-	if (mvotg->pdata->id)
-		otg_ctrl->id = !!mvotg->pdata->id->poll();
+	if (mvotg->extern_attr & MV_USB_HAS_IDPIN_DETECTION) {
+		if (mv_usb2_extern_call(mv_phy, idpin, get_idpin, &idpin))
+			return;
+		otg_ctrl->id = !!idpin;
+	}
 	else
 		otg_ctrl->id = !!(otgsc & OTGSC_STS_USB_ID);
 
@@ -502,7 +513,7 @@ static irqreturn_t mv_otg_irq(int irq, void *dev)
 	 * if we have vbus, then the vbus detection for B-device
 	 * will be done by mv_otg_inputs_irq().
 	 */
-	if (mvotg->pdata->vbus)
+	if (mvotg->extern_attr & MV_USB_HAS_VBUS_DETECTION)
 		if ((otgsc & OTGSC_STS_USB_ID) &&
 		    !(otgsc & OTGSC_INTSTS_USB_ID))
 			return IRQ_NONE;
@@ -515,9 +526,10 @@ static irqreturn_t mv_otg_irq(int irq, void *dev)
 	return IRQ_HANDLED;
 }
 
-static irqreturn_t mv_otg_inputs_irq(int irq, void *dev)
+static int mv_otg_notifier_call(struct notifier_block *nb,
+				unsigned long val, void *v)
 {
-	struct mv_otg *mvotg = dev;
+	struct mv_otg *mvotg = container_of(nb, struct mv_otg, notifier);
 
 	/* The clock may disabled at this time */
 	if (!mvotg->active) {
@@ -527,7 +539,7 @@ static irqreturn_t mv_otg_inputs_irq(int irq, void *dev)
 
 	mv_otg_run_state_machine(mvotg, 0);
 
-	return IRQ_HANDLED;
+	return 0;
 }
 
 static ssize_t
@@ -660,9 +672,15 @@ static struct attribute_group inputs_attr_group = {
 int mv_otg_remove(struct platform_device *pdev)
 {
 	struct mv_otg *mvotg = platform_get_drvdata(pdev);
+	struct mv_usb2_phy *mv_phy = container_of(mvotg->outer_phy,
+					struct mv_usb2_phy, phy);
 
 	sysfs_remove_group(&mvotg->pdev->dev.kobj, &inputs_attr_group);
 
+	if ((mvotg->extern_attr & MV_USB_HAS_VBUS_DETECTION) ||
+		(mvotg->extern_attr & MV_USB_HAS_IDPIN_DETECTION))
+		mv_usb2_unregister_notifier(mv_phy, &mvotg->notifier);
+
 	if (mvotg->qwork) {
 		flush_workqueue(mvotg->qwork);
 		destroy_workqueue(mvotg->qwork);
@@ -682,6 +700,7 @@ static int mv_otg_probe(struct platform_device *pdev)
 	struct mv_otg *mvotg;
 	struct usb_otg *otg;
 	struct resource *r;
+	struct mv_usb2_phy *mv_phy;
 	int retval = 0, clk_i, i;
 	size_t size;
 
@@ -704,6 +723,7 @@ static int mv_otg_probe(struct platform_device *pdev)
 	platform_set_drvdata(pdev, mvotg);
 
 	mvotg->pdev = pdev;
+	mvotg->extern_attr = pdata->extern_attr;
 	mvotg->pdata = pdata;
 
 	mvotg->clknum = pdata->clknum;
@@ -760,6 +780,7 @@ static int mv_otg_probe(struct platform_device *pdev)
 		retval = PTR_ERR(mvotg->outer_phy);
 		goto err_destroy_workqueue;
 	}
+	mv_phy = container_of(mvotg->outer_phy, struct mv_usb2_phy, phy);
 
 	/* we will acces controller register, so enable the udc controller */
 	retval = mv_otg_enable_internal(mvotg);
@@ -772,29 +793,10 @@ static int mv_otg_probe(struct platform_device *pdev)
 		(struct mv_otg_regs __iomem *) ((unsigned long) mvotg->cap_regs
 			+ (readl(mvotg->cap_regs) & CAPLENGTH_MASK));
 
-	if (pdata->id) {
-		retval = devm_request_threaded_irq(&pdev->dev, pdata->id->irq,
-						NULL, mv_otg_inputs_irq,
-						IRQF_ONESHOT, "id", mvotg);
-		if (retval) {
-			dev_info(&pdev->dev,
-				 "Failed to request irq for ID\n");
-			pdata->id = NULL;
-		}
-	}
-
-	if (pdata->vbus) {
-		mvotg->clock_gating = 1;
-		retval = devm_request_threaded_irq(&pdev->dev, pdata->vbus->irq,
-						NULL, mv_otg_inputs_irq,
-						IRQF_ONESHOT, "vbus", mvotg);
-		if (retval) {
-			dev_info(&pdev->dev,
-				 "Failed to request irq for VBUS, "
-				 "disable clock gating\n");
-			mvotg->clock_gating = 0;
-			pdata->vbus = NULL;
-		}
+	if ((mvotg->extern_attr & MV_USB_HAS_VBUS_DETECTION) ||
+		(mvotg->extern_attr & MV_USB_HAS_IDPIN_DETECTION)) {
+		mvotg->notifier.notifier_call = mv_otg_notifier_call;
+		mv_usb2_register_notifier(mv_phy, &mvotg->notifier);
 	}
 
 	if (pdata->disable_otg_clock_gating)
drivers/usb/otg/mv_otg.h +3 -0
index 65e7fd1..2a9ecae 100644
@@ -149,6 +149,9 @@ struct mv_otg {
 	u32 irq_status;
 	u32 irq_en;
 
+	unsigned int extern_attr;
+	struct notifier_block notifier;
+
 	struct delayed_work work;
 	struct workqueue_struct *qwork;
 
Signature · 1 line
1.7.4.1
author
#31 lore raw

PATCH v8 15/16 arm: mmp: add extern chip support for brownstone

In reply to Chao Xie

Change the board support for usb as extern chip is supported in marvell usb PHY driver

---
 arch/arm/mach-mmp/brownstone.c |    2 --
 1 files changed, 0 insertions(+), 2 deletions(-)
arch/arm/mach-mmp/brownstone.c +0 -2
index f84613a..eb92419 100644
@@ -214,10 +214,8 @@ static struct mv_usb_phy_platform_data brownstone_usb_phy_pdata = {
 static struct mv_usb_platform_data brownstone_usb_pdata = {
 	.clknum		= 1,
 	.clkname	= mmp2_usb_clock_name,
-	.vbus		= NULL,
 	.mode		= MV_USB_MODE_OTG,
 	.otg_force_a_bus_req = 1,
-	.set_vbus	= NULL,
 };
 #endif
 #endif
Signature · 1 line
1.7.4.1
author
#32 lore raw

PATCH v8 16/16 arm: mmp: add extern chip support for ttc_dkb

In reply to Chao Xie

Change the board support for usb as extern chip is supported in marvell usb PHY driver.

---
 arch/arm/mach-mmp/ttc_dkb.c |    2 --
 1 files changed, 0 insertions(+), 2 deletions(-)
arch/arm/mach-mmp/ttc_dkb.c +0 -2
index 5622092..0ba89e9 100644
@@ -174,10 +174,8 @@ static struct mv_usb_phy_platform_data ttc_usb_phy_pdata = {
 static struct mv_usb_platform_data ttc_usb_pdata = {
 	.clknum		= 1,
 	.clkname	= pxa910_usb_clock_name,
-	.vbus		= NULL,
 	.mode		= MV_USB_MODE_OTG,
 	.otg_force_a_bus_req = 1,
-	.set_vbus	= NULL,
 };
 #endif
 #endif
Signature · 1 line
1.7.4.1
maintainer
#33 lore raw

v8 cover mv-usb phy enhancement patches

In reply to Chao Xie
Quoted 1 line · Chao Xie

The patches are divied into 2 parts

<snip>

You do realize that we are currently in the middle of the merge window, and don't have any time at all to look at these patches until 2 weeks from now, at the earliest?

greg k-h