Hi,
On Thu, 2016-08-25 at 10:32 +0200, Oliver Neukum wrote:
> On Thu, 2016-08-25 at 11:05 +0800, Chunfeng Yun wrote:
> > This patch adds support for the MediaTek USB3 controller
> > integrated into MT8173. It can be configured as Dual-Role
> > Device (DRD), Peripheral Only and Host Only (xHCI) modes.
> >
>
> > +/**
> > + * General Purpose Descriptor (GPD):
> > + * The format of TX GPD is a little different from RX one.
> > + * And the size of GPD is 16 bytes.
> > + *
> > + * @flag:
> > + * bit0: Hardware Own (HWO)
> > + * bit1: Buffer Descriptor Present (BDP), always 0, BD is not supported
> > + * bit2: Bypass (BPS), 1: HW skips this GPD if HWO = 1
> > + * bit7: Interrupt On Completion (IOC)
> > + * @chksum: This is used to validate the contents of this GPD;
> > + * If TXQ_CS_EN / RXQ_CS_EN bit is set, an interrupt is issued
> > + * when checksum validation fails;
> > + * Checksum value is calculated over the 16 bytes of the GPD by default;
> > + * @data_buf_len (RX ONLY): This value indicates the length of
> > + * the assigned data buffer
> > + * @next_gpd: Physical address of the next GPD
> > + * @buffer: Physical address of the data buffer
> > + * @buf_len:
> > + * (TX): This value indicates the length of the assigned data buffer
> > + * (RX): The total length of data received
> > + * @ext_len: reserved
> > + * @ext_flag:
> > + * bit5 (TX ONLY): Zero Length Packet (ZLP),
> > + */
> > +struct qmu_gpd {
> > + u8 flag;
> > + u8 chksum;
> > + u16 data_buf_len;
> > + u32 next_gpd;
> > + u32 buffer;
> > + u16 buf_len;
> > + u8 ext_len;
> > + u8 ext_flag;
> > +} __packed;
>
> It looks like this is shared with hardware in memory.
> But you leave the endianness of the bigger fields undeclared.
>
The driver only supports Little-Endian SoCs currently, because I have no
Big-Endian platform to test it.
> > +/**
> > +* dma: physical base address of GPD segment
> > +* start: virtual base address of GPD segment
> > +* end: the last GPD element
> > +* enqueue: the first empty GPD to use
> > +* dequeue: the first completed GPD serviced by ISR
> > +* NOTE: the size of GPD ring should be >= 2
> > +*/
> > +struct mtu3_gpd_ring {
> > + dma_addr_t dma;
> > + struct qmu_gpd *start;
> > + struct qmu_gpd *end;
> > + struct qmu_gpd *enqueue;
> > + struct qmu_gpd *dequeue;
> > +};
> > +
> > +/**
> > +* @vbus: vbus 5V used by host mode
> > +* @edev: external connector used to detect vbus and iddig changes
> > +* @vbus_nb: notifier for vbus detection
> > +* @vbus_nb: notifier for iddig(idpin) detection
> > +* @extcon_reg_dwork: delay work for extcon notifier register, waiting for
> > +* xHCI driver initialization, it's necessary for system bootup
> > +* as device.
> > +* @is_u3_drd: whether port0 supports usb3.0 dual-role device or not
> > +* @id_*: used to maually switch between host and device modes by idpin
> > +* @manual_drd_enabled: it's true when supports dual-role device by debugfs
> > +* to switch host/device modes depending on user input.
>
> Please use a common interface for this. The Intel people are introducing
> one.
Can you give me the link address of the patch, I didn't find it.
>
>
> > +#endif
> > diff --git a/drivers/usb/mtu3/mtu3_core.c b/drivers/usb/mtu3/mtu3_core.c
> > new file mode 100644
> > index 0000000..fdc11b6
> > --- /dev/null
> > +++ b/drivers/usb/mtu3/mtu3_core.c
> > @@ -0,0 +1,874 @@
> > +/*
> > + * mtu3_core.c - hardware access layer and gadget init/exit of
> > + * MediaTek usb3 Dual-Role Controller Driver
> > + *
> > + * Copyright (C) 2016 MediaTek Inc.
> > + *
> > + * Author: Chunfeng Yun <chunfeng....@mediatek.com>
> > + *
> > + * 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/kernel.h>
> > +#include <linux/module.h>
> > +#include <linux/of_address.h>
> > +#include <linux/of_irq.h>
> > +#include <linux/platform_device.h>
> > +
> > +#include "mtu3.h"
> > +
> > +static int ep_fifo_alloc(struct mtu3_ep *mep, u32 seg_size)
> > +{
> > + struct mtu3_fifo_info *fifo = mep->fifo;
> > + u32 num_bits = DIV_ROUND_UP(seg_size, MTU3_EP_FIFO_UNIT);
> > + u32 start_bit;
> > +
> > + /* ensure that @mep->fifo_seg_size is power of two */
> > + num_bits = roundup_pow_of_two(num_bits);
> > + if (num_bits > fifo->limit)
> > + return -EINVAL;
> > +
> > + mep->fifo_seg_size = num_bits * MTU3_EP_FIFO_UNIT;
> > + num_bits = num_bits * (mep->slot + 1);
> > + start_bit = bitmap_find_next_zero_area(fifo->bitmap,
> > + fifo->limit, 0, num_bits, 0);
> > + if (start_bit >= fifo->limit)
> > + return -EOVERFLOW;
> > +
> > + bitmap_set(fifo->bitmap, start_bit, num_bits);
> > + mep->fifo_size = num_bits * MTU3_EP_FIFO_UNIT;
> > + mep->fifo_addr = fifo->base + MTU3_EP_FIFO_UNIT * start_bit;
> > +
> > + dev_dbg(mep->mtu->dev, "%s fifo:%#x/%#x, start_bit: %d\n",
> > + __func__, mep->fifo_seg_size, mep->fifo_size, start_bit);
>
> If you use the "f" option to dynamic debugging it will give you the
> function name anyway. So why include __func__ ?
>
It's used when dynamic debugging is disable but defines DEBUG macro.
>
> > +/* enable/disable U3D SS function */
> > +static void mtu3_ss_func_set(struct mtu3 *mtu, bool enable)
>
> Inline maybe ?
>
> > +{
> > + /* If usb3_en==0, LTSSM will go to SS.Disable state */
> > + if (enable)
> > + mtu3_setbits(mtu->mac_base, U3D_USB3_CONFIG, USB3_EN);
> > + else
> > + mtu3_clrbits(mtu->mac_base, U3D_USB3_CONFIG, USB3_EN);
> > +
> > + dev_dbg(mtu->dev, "USB3_EN = %d\n", !!enable);
> > +}
> > +
>
> [..]
>
>
> > +int mtu3_config_ep(struct mtu3 *mtu, struct mtu3_ep *mep,
> > + int interval, int burst, int mult)
> > +{
> > + void __iomem *mbase = mtu->mac_base;
> > + int epnum = mep->epnum;
> > + u32 csr0, csr1, csr2;
> > + int fifo_sgsz, fifo_addr;
> > + int num_pkts;
> > +
> > + fifo_addr = ep_fifo_alloc(mep, mep->maxp);
> > + if (fifo_addr < 0) {
> > + dev_err(mtu->dev, "alloc ep fifo failed(%d)\n", mep->maxp);
> > + return -ENOMEM;
> > + }
> > + fifo_sgsz = ilog2(mep->fifo_seg_size);
> > + dev_dbg(mtu->dev, "%s fifosz: %x(%x/%x)\n", __func__, fifo_sgsz,
> > + mep->fifo_seg_size, mep->fifo_size);
> > +
> > + if (mep->is_in) {
> > + csr0 = TX_TXMAXPKTSZ(mep->maxp);
> > + csr0 |= TX_DMAREQEN;
> > +
> > + num_pkts = (burst + 1) * (mult + 1) - 1;
> > + csr1 = TX_SS_BURST(burst) | TX_SLOT(mep->slot);
> > + csr1 |= TX_MAX_PKT(num_pkts) | TX_MULT(mult);
> > +
> > + csr2 = TX_FIFOADDR(fifo_addr >> 4);
> > + csr2 |= TX_FIFOSEGSIZE(fifo_sgsz);
> > +
> > + switch (mep->type) {
> > + case USB_ENDPOINT_XFER_BULK:
> > + csr1 |= TX_TYPE(TYPE_BULK);
> > + break;
> > + case USB_ENDPOINT_XFER_ISOC:
> > + csr1 |= TX_TYPE(TYPE_ISO);
> > + csr2 |= TX_BINTERVAL(interval);
> > + break;
> > + case USB_ENDPOINT_XFER_INT:
> > + csr1 |= TX_TYPE(TYPE_INT);
> > + csr2 |= TX_BINTERVAL(interval);
> > + break;
>
> And if it is control?
The function is only called for non control EP.
I will add a comment for it.
>
> > + }
> > +
> > + /* Enable QMU Done interrupt */
> > + mtu3_setbits(mbase, U3D_QIESR0, QMU_TX_DONE_INT(epnum));
> > +
> > + mtu3_writel(mbase, MU3D_EP_TXCR0(epnum), csr0);
> > + mtu3_writel(mbase, MU3D_EP_TXCR1(epnum), csr1);
> > + mtu3_writel(mbase, MU3D_EP_TXCR2(epnum), csr2);
> > +
> > + dev_dbg(mtu->dev, "U3D_TX%d CSR0:%#x, CSR1:%#x, CSR2:%#x\n",
> > + epnum, mtu3_readl(mbase, MU3D_EP_TXCR0(epnum)),
> > + mtu3_readl(mbase, MU3D_EP_TXCR1(epnum)),
> > + mtu3_readl(mbase, MU3D_EP_TXCR2(epnum)));
> > + } else {
> > + csr0 = RX_RXMAXPKTSZ(mep->maxp);
> > + csr0 |= RX_DMAREQEN;
> > +
> > + num_pkts = (burst + 1) * (mult + 1) - 1;
> > + csr1 = RX_SS_BURST(burst) | RX_SLOT(mep->slot);
> > + csr1 |= RX_MAX_PKT(num_pkts) | RX_MULT(mult);
> > +
> > + csr2 = RX_FIFOADDR(fifo_addr >> 4);
> > + csr2 |= RX_FIFOSEGSIZE(fifo_sgsz);
> > +
> > + switch (mep->type) {
> > + case USB_ENDPOINT_XFER_BULK:
> > + csr1 |= RX_TYPE(TYPE_BULK);
> > + break;
> > + case USB_ENDPOINT_XFER_ISOC:
> > + csr1 |= RX_TYPE(TYPE_ISO);
> > + csr2 |= RX_BINTERVAL(interval);
> > + break;
> > + case USB_ENDPOINT_XFER_INT:
> > + csr1 |= RX_TYPE(TYPE_INT);
> > + csr2 |= RX_BINTERVAL(interval);
> > + break;
>
> Again: control endpoints?
Only for non-EP0.
>
> > + }
> > +
> > + /*Enable QMU Done interrupt */
> > + mtu3_setbits(mbase, U3D_QIESR0, QMU_RX_DONE_INT(epnum));
> > +
> > + mtu3_writel(mbase, MU3D_EP_RXCR0(epnum), csr0);
> > + mtu3_writel(mbase, MU3D_EP_RXCR1(epnum), csr1);
> > + mtu3_writel(mbase, MU3D_EP_RXCR2(epnum), csr2);
> > +
> > + dev_dbg(mtu->dev, "U3D_RX%d CSR0:%#x, CSR1:%#x, CSR2:%#x\n",
> > + epnum, mtu3_readl(mbase, MU3D_EP_RXCR0(epnum)),
> > + mtu3_readl(mbase, MU3D_EP_RXCR1(epnum)),
> > + mtu3_readl(mbase, MU3D_EP_RXCR2(epnum)));
> > + }
> > +
> > + dev_dbg(mtu->dev, "csr0:%#x, csr1:%#x, csr2:%#x\n", csr0, csr1, csr2);
> > + dev_dbg(mtu->dev, "%s: %s, fifo-addr:%#x, fifo-size:%#x(%#x/%#x)\n",
> > + __func__, mep->name, mep->fifo_addr, mep->fifo_size,
> > + fifo_sgsz, mep->fifo_seg_size);
> > +
> > + return 0;
> > +}
> > +
>
> [..]
> > +static void mtu3_set_speed(struct mtu3 *mtu)
> > +{
> > + void __iomem *mbase = mtu->mac_base;
> > +
> > + if (!mtu->is_u3_ip && (mtu->max_speed > USB_SPEED_HIGH))
> > + mtu->max_speed = USB_SPEED_HIGH;
>
> You are missing the checks for USB_SPEED_WIRELESS in general.
> Can you just ignore it?
Doesn't support USB_SPEED_WIRELESS, so ignore it.
>
> > +
> > + if (mtu->max_speed == USB_SPEED_FULL) {
> > + /* disable U3 SS function */
> > + mtu3_clrbits(mbase, U3D_USB3_CONFIG, USB3_EN);
> > + /* disable HS function */
> > + mtu3_clrbits(mbase, U3D_POWER_MANAGEMENT, HS_ENABLE);
> > + } else if (mtu->max_speed == USB_SPEED_HIGH) {
> > + mtu3_clrbits(mbase, U3D_USB3_CONFIG, USB3_EN);
> > + /* HS/FS detected by HW */
> > + mtu3_setbits(mbase, U3D_POWER_MANAGEMENT, HS_ENABLE);
> > + }
> > + dev_info(mtu->dev, "max_speed: %s\n",
> > + usb_speed_string(mtu->max_speed));
> > +}
>
> [..]
> > +static irqreturn_t mtu3_link_isr(struct mtu3 *mtu)
> > +{
> > + void __iomem *mbase = mtu->mac_base;
> > + enum usb_device_speed udev_speed;
> > + u32 maxpkt = 64;
> > + u32 link;
> > + u32 speed;
> > +
> > + link = mtu3_readl(mbase, U3D_DEV_LINK_INTR);
> > + link &= mtu3_readl(mbase, U3D_DEV_LINK_INTR_ENABLE);
> > + mtu3_writel(mbase, U3D_DEV_LINK_INTR, link); /* W1C */
> > + dev_dbg(mtu->dev, "=== LINK[%x] ===\n", link);
> > +
> > + if (!(link & SSUSB_DEV_SPEED_CHG_INTR))
> > + return IRQ_NONE;
>
> Shouldn't you do this check before you clear interrupts?
In fact, U3D_DEV_LINK_INTR has only one interrupt,
SSUSB_DEV_SPEED_CHG_INTR, others can be treated as spurious ones.
>
> > + speed = SSUSB_DEV_SPEED(mtu3_readl(mbase, U3D_DEVICE_CONF));
>
> Do you really want to read this out of the hardware on every interrupt?
Yes, the interrupt is triggered only when speed is changed,
and get the new speed here.
>
> > +
> > + switch (speed) {
> > + case MTU3_SPEED_FULL:
> > + udev_speed = USB_SPEED_FULL;
> > + /*BESLCK = 4 < BESLCK_U3 = 10 < BESLDCK = 15 */
> > + mtu3_writel(mbase, U3D_USB20_LPM_PARAMETER, LPM_BESLDCK(0xf)
> > + | LPM_BESLCK(4) | LPM_BESLCK_U3(0xa));
> > + mtu3_setbits(mbase, U3D_POWER_MANAGEMENT,
> > + LPM_BESL_STALL | LPM_BESLD_STALL);
> > + break;
> > + case MTU3_SPEED_HIGH:
> > + udev_speed = USB_SPEED_HIGH;
> > + /*BESLCK = 4 < BESLCK_U3 = 10 < BESLDCK = 15 */
> > + mtu3_writel(mbase, U3D_USB20_LPM_PARAMETER, LPM_BESLDCK(0xf)
> > + | LPM_BESLCK(4) | LPM_BESLCK_U3(0xa));
> > + mtu3_setbits(mbase, U3D_POWER_MANAGEMENT,
> > + LPM_BESL_STALL | LPM_BESLD_STALL);
> > + break;
> > + case MTU3_SPEED_SUPER:
> > + udev_speed = USB_SPEED_SUPER;
> > + maxpkt = 512;
> > + break;
> > + default:
> > + udev_speed = USB_SPEED_UNKNOWN;
> > + break;
> > + }
> > + dev_dbg(mtu->dev, "%s: %s\n", __func__, usb_speed_string(udev_speed));
> > +
> > + mtu->g.speed = udev_speed;
> > + mtu->g.ep0->maxpacket = maxpkt;
> > + mtu->ep0_state = MU3D_EP0_STATE_SETUP;
> > +
> > + if (udev_speed == USB_SPEED_UNKNOWN)
> > + mtu3_gadget_disconnect(mtu);
> > + else
> > + mtu3_ep0_setup(mtu);
> > +
> > + return IRQ_HANDLED;
> > +}
> > +
> > +static irqreturn_t mtu3_u3_ltssm_isr(struct mtu3 *mtu)
> > +{
> > + void __iomem *mbase = mtu->mac_base;
> > + u32 ltssm;
> > +
> > + ltssm = mtu3_readl(mbase, U3D_LTSSM_INTR);
> > + ltssm &= mtu3_readl(mbase, U3D_LTSSM_INTR_ENABLE);
> > + mtu3_writel(mbase, U3D_LTSSM_INTR, ltssm); /* W1C */
> > + dev_dbg(mtu->dev, "=== LTSSM[%x] ===\n", ltssm);
> > +
> > + if (ltssm & HOT_RST_INTR)
> > + mtu3_gadget_reset(mtu);
> > +
> > + if (ltssm & WARM_RST_INTR)
> > + mtu3_gadget_reset(mtu);
>
> You really would reset the gadget twice if that happens?
> And do the rest of the tests make sense in that case?
I will merge them into one as following:
if (ltssm & (HOT_RST_INTR | WARM_RST_INTR))
...
>
> > +
> > + if (ltssm & VBUS_FALL_INTR)
> > + mtu3_ss_func_set(mtu, false);
> > +
> > + if (ltssm & VBUS_RISE_INTR)
> > + mtu3_ss_func_set(mtu, true);
> > +
> > + if (ltssm & EXIT_U3_INTR)
> > + mtu3_gadget_resume(mtu);
> > +
> > + if (ltssm & ENTER_U3_INTR)
> > + mtu3_gadget_suspend(mtu);
> > +
> > + return IRQ_HANDLED;
> > +}
>
> [..]
>
> > +static int mtu3_hw_init(struct mtu3 *mtu)
> > +{
> > + int ret;
> > +
> > + mtu3_device_reset(mtu);
> > +
> > + ret = mtu3_device_enable(mtu);
> > + if (ret) {
> > + dev_err(mtu->dev, "device enable failed %d\n", ret);
> > + return ret;
> > + }
> > +
> > + ret = mtu3_mem_alloc(mtu);
> > + if (ret) {
> > + dev_err(mtu->dev, "mem alloc failed, aborting\n");
>
> 1. You can't leave the device enabled in that case.
It enable device mac and port, otherwise I can't initialize other
register.
and the function is not enable until udc_start() is called.
> 2. No need for a message. The kernel will already do that if kmalloc()
> fails under such circumstances.
remove it latter.
>
> > + return -ENOMEM;
> > + }
> > +
> > + mtu3_regs_init(mtu);
> > +
> > + return 0;
> > +}
>
> > diff --git a/drivers/usb/mtu3/mtu3_dr.c b/drivers/usb/mtu3/mtu3_dr.c
> > new file mode 100644
> > index 0000000..f560f93
> > --- /dev/null
> > +++ b/drivers/usb/mtu3/mtu3_dr.c
> > @@ -0,0 +1,375 @@
> > +/*
> > + * mtu3_dr.c - dual role switch and host glue layer
> > + *
> > + * Copyright (C) 2016 MediaTek Inc.
> > + *
> > + * Author: Chunfeng Yun <chunfeng....@mediatek.com>
> > + *
> > + * 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/debugfs.h>
> > +#include <linux/irq.h>
> > +#include <linux/kernel.h>
> > +#include <linux/of_device.h>
> > +#include <linux/pinctrl/consumer.h>
> > +#include <linux/seq_file.h>
> > +#include <linux/uaccess.h>
> > +
> > +#include "mtu3.h"
> > +#include "mtu3_dr.h"
> > +
> > +#define USB2_PORT 2
> > +#define USB3_PORT 3
> > +
[...]
> > +
> > +static ssize_t ssusb_mode_write(struct file *file,
> > + const char __user *ubuf, size_t count, loff_t *ppos)
> > +{
> > + struct seq_file *sf = file->private_data;
> > + struct ssusb_mtk *ssusb = sf->private;
> > + char buf[16];
> > +
> > + if (copy_from_user(&buf, ubuf, min_t(size_t, sizeof(buf) - 1, count)))
> > + return -EFAULT;
> > +
> > + if (!strncmp(buf, "host", 4) && !ssusb->is_host)
> > + ssusb_mode_manual_switch(ssusb, 1);
> > + else if (!strncmp(buf, "device", 6) && ssusb->is_host)
> > + ssusb_mode_manual_switch(ssusb, 0);
> > + else
> > + dev_err(ssusb->dev, "wrong or duplicated setting\n");
>
> No proper error returns
add it later.
> > +
> > + return count;
> > +}
> > +
> > +static const struct file_operations ssusb_mode_fops = {
> > + .open = ssusb_mode_open,
> > + .write = ssusb_mode_write,
> > + .read = seq_read,
> > + .llseek = seq_lseek,
> > + .release = single_release,
> > +};
> > +
[..]
> > +static void ssusb_debugfs_exit(struct ssusb_mtk *ssusb)
> > +{
> > + debugfs_remove_recursive(ssusb->dbgfs_root);
> > +}
> > +
> > +int ssusb_otg_switch_init(struct ssusb_mtk *ssusb)
> > +{
> > + struct otg_switch_mtk *otg_sx = &ssusb->otg_switch;
> > +
> > + INIT_DELAYED_WORK(&otg_sx->extcon_reg_dwork, extcon_register_dwork);
> > +
> > + if (otg_sx->manual_drd_enabled)
> > + ssusb_debugfs_init(ssusb);
> > +
> > + /* It is enough to delay 1s for waiting for host initialization */
> > + schedule_delayed_work(&otg_sx->extcon_reg_dwork, HZ);
>
> You need to handle this still pending when you are deregistered.
OK.
>
> > +
> > + return 0;
> > +}
> > +
> > +void ssusb_otg_switch_exit(struct ssusb_mtk *ssusb)
> > +{
> > + struct otg_switch_mtk *otg_sx = &ssusb->otg_switch;
> > +
> > + if (otg_sx->edev) {
> > + extcon_unregister_notifier(otg_sx->edev,
> > + EXTCON_USB, &otg_sx->vbus_nb);
> > + extcon_unregister_notifier(otg_sx->edev,
> > + EXTCON_USB_HOST, &otg_sx->id_nb);
> > + }
> > +
> > + if (otg_sx->manual_drd_enabled)
> > + ssusb_debugfs_exit(ssusb);
> > +}
>
> Could you break this up a bit? It is extensively long a patch?
I afraid I can't break the patch into small ones but also provide
complete functionality at the same time.
Sorry for inconvenience.
>
Thank you very much :-)
> Regards
> Oliver
>
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
