USB: Gadget driver for Samsung HS/OtG block

Driver support for the new high-speed/OtG block that is
in the newer line of Samsung SoC devices such as the
S3C64XX series.

This driver does not currntly have DMA support enabled due
to issues with buffer alignment which need to be sorted out.

Signed-off-by: Ben Dooks <ben@simtec.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

3 files changed, 3284 insertions, 0 deletions

diff --git a/drivers/usb/gadget/Kconfig b/drivers/usb/gadget/Kconfig
index 5de9b4f2171..924bb7a1cec 100644
--- a/drivers/usb/gadget/Kconfig
+++ b/drivers/usb/gadget/Kconfig
@@ -272,6 +272,20 @@ config USB_PXA27X
 	default USB_GADGET
 	select USB_GADGET_SELECTED
 
+config USB_GADGET_S3C_HSOTG
+	boolean "S3C HS/OtG USB Device controller"
+	depends on S3C_DEV_USB_HSOTG
+	select USB_GADGET_S3C_HSOTG_PIO
+	help
+	  The Samsung S3C64XX USB2.0 high-speed gadget controller
+	  integrated into the S3C64XX series SoC.
+
+config USB_S3C_HSOTG
+	tristate
+	depends on USB_GADGET_S3C_HSOTG
+	default USB_GADGET
+	select USB_GADGET_SELECTED
+
 config USB_GADGET_S3C2410
 	boolean "S3C2410 USB Device Controller"
 	depends on ARCH_S3C2410
diff --git a/drivers/usb/gadget/Makefile b/drivers/usb/gadget/Makefile
index c3fe06396b7..9be2fbd6f5c 100644
--- a/drivers/usb/gadget/Makefile
+++ b/drivers/usb/gadget/Makefile
@@ -25,6 +25,7 @@ endif
 obj-$(CONFIG_USB_M66592)	+= m66592-udc.o
 obj-$(CONFIG_USB_FSL_QE)	+= fsl_qe_udc.o
 obj-$(CONFIG_USB_CI13XXX)	+= ci13xxx_udc.o
+obj-$(CONFIG_USB_S3C_HSOTG)	+= s3c-hsotg.o
 
 #
 # USB gadget drivers
diff --git a/drivers/usb/gadget/s3c-hsotg.c b/drivers/usb/gadget/s3c-hsotg.c
new file mode 100644
index 00000000000..50c71aae2cc
--- /dev/null
+++ b/drivers/usb/gadget/s3c-hsotg.c
@@ -0,0 +1,3269 @@
+/* linux/drivers/usb/gadget/s3c-hsotg.c
+ *
+ * Copyright 2008 Openmoko, Inc.
+ * Copyright 2008 Simtec Electronics
+ *      Ben Dooks <ben@simtec.co.uk>
+ *      http://armlinux.simtec.co.uk/
+ *
+ * S3C USB2.0 High-speed / OtG driver
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+*/
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/spinlock.h>
+#include <linux/interrupt.h>
+#include <linux/platform_device.h>
+#include <linux/dma-mapping.h>
+#include <linux/debugfs.h>
+#include <linux/seq_file.h>
+#include <linux/delay.h>
+#include <linux/io.h>
+
+#include <linux/usb/ch9.h>
+#include <linux/usb/gadget.h>
+
+#include <mach/map.h>
+
+#include <plat/regs-usb-hsotg-phy.h>
+#include <plat/regs-usb-hsotg.h>
+#include <plat/regs-sys.h>
+#include <plat/udc-hs.h>
+
+#define DMA_ADDR_INVALID (~((dma_addr_t)0))
+
+/* EP0_MPS_LIMIT
+ *
+ * Unfortunately there seems to be a limit of the amount of data that can
+ * be transfered by IN transactions on EP0. This is either 127 bytes or 3
+ * packets (which practially means 1 packet and 63 bytes of data) when the
+ * MPS is set to 64.
+ *
+ * This means if we are wanting to move >127 bytes of data, we need to
+ * split the transactions up, but just doing one packet at a time does
+ * not work (this may be an implicit DATA0 PID on first packet of the
+ * transaction) and doing 2 packets is outside the controller's limits.
+ *
+ * If we try to lower the MPS size for EP0, then no transfers work properly
+ * for EP0, and the system will fail basic enumeration. As no cause for this
+ * has currently been found, we cannot support any large IN transfers for
+ * EP0.
+ */
+#define EP0_MPS_LIMIT	64
+
+struct s3c_hsotg;
+struct s3c_hsotg_req;
+
+/**
+ * struct s3c_hsotg_ep - driver endpoint definition.
+ * @ep: The gadget layer representation of the endpoint.
+ * @name: The driver generated name for the endpoint.
+ * @queue: Queue of requests for this endpoint.
+ * @parent: Reference back to the parent device structure.
+ * @req: The current request that the endpoint is processing. This is
+ *       used to indicate an request has been loaded onto the endpoint
+ *       and has yet to be completed (maybe due to data move, or simply
+ *	 awaiting an ack from the core all the data has been completed).
+ * @debugfs: File entry for debugfs file for this endpoint.
+ * @lock: State lock to protect contents of endpoint.
+ * @dir_in: Set to true if this endpoint is of the IN direction, which
+ *	    means that it is sending data to the Host.
+ * @index: The index for the endpoint registers.
+ * @name: The name array passed to the USB core.
+ * @halted: Set if the endpoint has been halted.
+ * @periodic: Set if this is a periodic ep, such as Interrupt
+ * @sent_zlp: Set if we've sent a zero-length packet.
+ * @total_data: The total number of data bytes done.
+ * @fifo_size: The size of the FIFO (for periodic IN endpoints)
+ * @fifo_load: The amount of data loaded into the FIFO (periodic IN)
+ * @last_load: The offset of data for the last start of request.
+ * @size_loaded: The last loaded size for DxEPTSIZE for periodic IN
+ *
+ * This is the driver's state for each registered enpoint, allowing it
+ * to keep track of transactions that need doing. Each endpoint has a
+ * lock to protect the state, to try and avoid using an overall lock
+ * for the host controller as much as possible.
+ *
+ * For periodic IN endpoints, we have fifo_size and fifo_load to try
+ * and keep track of the amount of data in the periodic FIFO for each
+ * of these as we don't have a status register that tells us how much
+ * is in each of them.
+ */
+struct s3c_hsotg_ep {
+	struct usb_ep		ep;
+	struct list_head	queue;
+	struct s3c_hsotg	*parent;
+	struct s3c_hsotg_req	*req;
+	struct dentry		*debugfs;
+
+	spinlock_t		lock;
+
+	unsigned long		total_data;
+	unsigned int		size_loaded;
+	unsigned int		last_load;
+	unsigned int		fifo_load;
+	unsigned short		fifo_size;
+
+	unsigned char		dir_in;
+	unsigned char		index;
+
+	unsigned int		halted:1;
+	unsigned int		periodic:1;
+	unsigned int		sent_zlp:1;
+
+	char			name[10];
+};
+
+#define S3C_HSOTG_EPS	(8+1)	/* limit to 9 for the moment */
+
+/**
+ * struct s3c_hsotg - driver state.
+ * @dev: The parent device supplied to the probe function
+ * @driver: USB gadget driver
+ * @plat: The platform specific configuration data.
+ * @regs: The memory area mapped for accessing registers.
+ * @regs_res: The resource that was allocated when claiming register space.
+ * @irq: The IRQ number we are using
+ * @debug_root: root directrory for debugfs.
+ * @debug_file: main status file for debugfs.
+ * @debug_fifo: FIFO status file for debugfs.
+ * @ep0_reply: Request used for ep0 reply.
+ * @ep0_buff: Buffer for EP0 reply data, if needed.
+ * @ctrl_buff: Buffer for EP0 control requests.
+ * @ctrl_req: Request for EP0 control packets.
+ * @eps: The endpoints being supplied to the gadget framework
+ */
+struct s3c_hsotg {
+	struct device		 *dev;
+	struct usb_gadget_driver *driver;
+	struct s3c_hsotg_plat	 *plat;
+
+	void __iomem		*regs;
+	struct resource		*regs_res;
+	int			irq;
+
+	struct dentry		*debug_root;
+	struct dentry		*debug_file;
+	struct dentry		*debug_fifo;
+
+	struct usb_request	*ep0_reply;
+	struct usb_request	*ctrl_req;
+	u8			ep0_buff[8];
+	u8			ctrl_buff[8];
+
+	struct usb_gadget	gadget;
+	struct s3c_hsotg_ep	eps[];
+};
+
+/**
+ * struct s3c_hsotg_req - data transfer request
+ * @req: The USB gadget request
+ * @queue: The list of requests for the endpoint this is queued for.
+ * @in_progress: Has already had size/packets written to core
+ * @mapped: DMA buffer for this request has been mapped via dma_map_single().
+ */
+struct s3c_hsotg_req {
+	struct usb_request	req;
+	struct list_head	queue;
+	unsigned char		in_progress;
+	unsigned char		mapped;
+};
+
+/* conversion functions */
+static inline struct s3c_hsotg_req *our_req(struct usb_request *req)
+{
+	return container_of(req, struct s3c_hsotg_req, req);
+}
+
+static inline struct s3c_hsotg_ep *our_ep(struct usb_ep *ep)
+{
+	return container_of(ep, struct s3c_hsotg_ep, ep);
+}
+
+static inline struct s3c_hsotg *to_hsotg(struct usb_gadget *gadget)
+{
+	return container_of(gadget, struct s3c_hsotg, gadget);
+}
+
+static inline void __orr32(void __iomem *ptr, u32 val)
+{
+	writel(readl(ptr) | val, ptr);
+}
+
+static inline void __bic32(void __iomem *ptr, u32 val)
+{
+	writel(readl(ptr) & ~val, ptr);
+}
+
+/* forward decleration of functions */
+static void s3c_hsotg_dump(struct s3c_hsotg *hsotg);
+
+/**
+ * using_dma - return the DMA status of the driver.
+ * @hsotg: The driver state.
+ *
+ * Return true if we're using DMA.
+ *
+ * Currently, we have the DMA support code worked into everywhere
+ * that needs it, but the AMBA DMA implementation in the hardware can
+ * only DMA from 32bit aligned addresses. This means that gadgets such
+ * as the CDC Ethernet cannot work as they often pass packets which are
+ * not 32bit aligned.
+ *
+ * Unfortunately the choice to use DMA or not is global to the controller
+ * and seems to be only settable when the controller is being put through
+ * a core reset. This means we either need to fix the gadgets to take
+ * account of DMA alignment, or add bounce buffers (yuerk).
+ *
+ * Until this issue is sorted out, we always return 'false'.
+ */
+static inline bool using_dma(struct s3c_hsotg *hsotg)
+{
+	return false;	/* support is not complete */
+}
+
+/**
+ * s3c_hsotg_en_gsint - enable one or more of the general interrupt
+ * @hsotg: The device state
+ * @ints: A bitmask of the interrupts to enable
+ */
+static void s3c_hsotg_en_gsint(struct s3c_hsotg *hsotg, u32 ints)
+{
+	u32 gsintmsk = readl(hsotg->regs + S3C_GINTMSK);
+	u32 new_gsintmsk;
+
+	new_gsintmsk = gsintmsk | ints;
+
+	if (new_gsintmsk != gsintmsk) {
+		dev_dbg(hsotg->dev, "gsintmsk now 0x%08x\n", new_gsintmsk);
+		writel(new_gsintmsk, hsotg->regs + S3C_GINTMSK);
+	}
+}
+
+/**
+ * s3c_hsotg_disable_gsint - disable one or more of the general interrupt
+ * @hsotg: The device state
+ * @ints: A bitmask of the interrupts to enable
+ */
+static void s3c_hsotg_disable_gsint(struct s3c_hsotg *hsotg, u32 ints)
+{
+	u32 gsintmsk = readl(hsotg->regs + S3C_GINTMSK);
+	u32 new_gsintmsk;
+
+	new_gsintmsk = gsintmsk & ~ints;
+
+	if (new_gsintmsk != gsintmsk)
+		writel(new_gsintmsk, hsotg->regs + S3C_GINTMSK);
+}
+
+/**
+ * s3c_hsotg_ctrl_epint - enable/disable an endpoint irq
+ * @hsotg: The device state
+ * @ep: The endpoint index
+ * @dir_in: True if direction is in.
+ * @en: The enable value, true to enable
+ *
+ * Set or clear the mask for an individual endpoint's interrupt
+ * request.
+ */
+static void s3c_hsotg_ctrl_epint(struct s3c_hsotg *hsotg,
+				 unsigned int ep, unsigned int dir_in,
+				 unsigned int en)
+{
+	unsigned long flags;
+	u32 bit = 1 << ep;
+	u32 daint;
+
+	if (!dir_in)
+		bit <<= 16;
+
+	local_irq_save(flags);
+	daint = readl(hsotg->regs + S3C_DAINTMSK);
+	if (en)
+		daint |= bit;
+	else
+		daint &= ~bit;
+	writel(daint, hsotg->regs + S3C_DAINTMSK);
+	local_irq_restore(flags);
+}
+
+/**
+ * s3c_hsotg_init_fifo - initialise non-periodic FIFOs
+ * @hsotg: The device instance.
+ */
+static void s3c_hsotg_init_fifo(struct s3c_hsotg *hsotg)
+{
+	/* the ryu 2.6.24 release ahs
+	   writel(0x1C0, hsotg->regs + S3C_GRXFSIZ);
+	   writel(S3C_GNPTXFSIZ_NPTxFStAddr(0x200) |
+		S3C_GNPTXFSIZ_NPTxFDep(0x1C0),
+		hsotg->regs + S3C_GNPTXFSIZ);
+	*/
+
+	/* set FIFO sizes to 2048/0x1C0 */
+
+	writel(2048, hsotg->regs + S3C_GRXFSIZ);
+	writel(S3C_GNPTXFSIZ_NPTxFStAddr(2048) |
+	       S3C_GNPTXFSIZ_NPTxFDep(0x1C0),
+	       hsotg->regs + S3C_GNPTXFSIZ);
+}
+
+/**
+ * @ep: USB endpoint to allocate request for.
+ * @flags: Allocation flags
+ *
+ * Allocate a new USB request structure appropriate for the specified endpoint
+ */
+struct usb_request *s3c_hsotg_ep_alloc_request(struct usb_ep *ep, gfp_t flags)
+{
+	struct s3c_hsotg_req *req;
+
+	req = kzalloc(sizeof(struct s3c_hsotg_req), flags);
+	if (!req)
+		return NULL;
+
+	INIT_LIST_HEAD(&req->queue);
+
+	req->req.dma = DMA_ADDR_INVALID;
+	return &req->req;
+}
+
+/**
+ * is_ep_periodic - return true if the endpoint is in periodic mode.
+ * @hs_ep: The endpoint to query.
+ *
+ * Returns true if the endpoint is in periodic mode, meaning it is being
+ * used for an Interrupt or ISO transfer.
+ */
+static inline int is_ep_periodic(struct s3c_hsotg_ep *hs_ep)
+{
+	return hs_ep->periodic;
+}
+
+/**
+ * s3c_hsotg_unmap_dma - unmap the DMA memory being used for the request
+ * @hsotg: The device state.
+ * @hs_ep: The endpoint for the request
+ * @hs_req: The request being processed.
+ *
+ * This is the reverse of s3c_hsotg_map_dma(), called for the completion
+ * of a request to ensure the buffer is ready for access by the caller.
+*/
+static void s3c_hsotg_unmap_dma(struct s3c_hsotg *hsotg,
+				struct s3c_hsotg_ep *hs_ep,
+				struct s3c_hsotg_req *hs_req)
+{
+	struct usb_request *req = &hs_req->req;
+	enum dma_data_direction dir;
+
+	dir = hs_ep->dir_in ? DMA_TO_DEVICE : DMA_FROM_DEVICE;
+
+	/* ignore this if we're not moving any data */
+	if (hs_req->req.length == 0)
+		return;
+
+	if (hs_req->mapped) {
+		/* we mapped this, so unmap and remove the dma */
+
+		dma_unmap_single(hsotg->dev, req->dma, req->length, dir);
+
+		req->dma = DMA_ADDR_INVALID;
+		hs_req->mapped = 0;
+	} else {
+		dma_sync_single(hsotg->dev, req->dma, req->length, dir);
+	}
+}
+
+/**
+ * s3c_hsotg_write_fifo - write packet Data to the TxFIFO
+ * @hsotg: The controller state.
+ * @hs_ep: The endpoint we're going to write for.
+ * @hs_req: The request to write data for.
+ *
+ * This is called when the TxFIFO has some space in it to hold a new
+ * transmission and we have something to give it. The actual setup of
+ * the data size is done elsewhere, so all we have to do is to actually
+ * write the data.
+ *
+ * The return value is zero if there is more space (or nothing was done)
+ * otherwise -ENOSPC is returned if the FIFO space was used up.
+ *
+ * This routine is only needed for PIO
+*/
+static int s3c_hsotg_write_fifo(struct s3c_hsotg *hsotg,
+				struct s3c_hsotg_ep *hs_ep,
+				struct s3c_hsotg_req *hs_req)
+{
+	bool periodic = is_ep_periodic(hs_ep);
+	u32 gnptxsts = readl(hsotg->regs + S3C_GNPTXSTS);
+	int buf_pos = hs_req->req.actual;
+	int to_write = hs_ep->size_loaded;
+	void *data;
+	int can_write;
+	int pkt_round;
+
+	to_write -= (buf_pos - hs_ep->last_load);
+
+	/* if there's nothing to write, get out early */
+	if (to_write == 0)
+		return 0;
+
+	if (periodic) {
+		u32 epsize = readl(hsotg->regs + S3C_DIEPTSIZ(hs_ep->index));
+		int size_left;
+		int size_done;
+
+		/* work out how much data was loaded so we can calculate
+		 * how much data is left in the fifo. */
+
+		size_left = S3C_DxEPTSIZ_XferSize_GET(epsize);
+
+		dev_dbg(hsotg->dev, "%s: left=%d, load=%d, fifo=%d, size %d\n",
+			__func__, size_left,
+			hs_ep->size_loaded, hs_ep->fifo_load, hs_ep->fifo_size);
+
+		/* how much of the data has moved */
+		size_done = hs_ep->size_loaded - size_left;
+
+		/* how much data is left in the fifo */
+		can_write = hs_ep->fifo_load - size_done;
+		dev_dbg(hsotg->dev, "%s: => can_write1=%d\n",
+			__func__, can_write);
+
+		can_write = hs_ep->fifo_size - can_write;
+		dev_dbg(hsotg->dev, "%s: => can_write2=%d\n",
+			__func__, can_write);
+
+		if (can_write <= 0) {
+			s3c_hsotg_en_gsint(hsotg, S3C_GINTSTS_PTxFEmp);
+			return -ENOSPC;
+		}
+	} else {
+		if (S3C_GNPTXSTS_NPTxQSpcAvail_GET(gnptxsts) == 0) {
+			dev_dbg(hsotg->dev,
+				"%s: no queue slots available (0x%08x)\n",
+				__func__, gnptxsts);
+
+			s3c_hsotg_en_gsint(hsotg, S3C_GINTSTS_NPTxFEmp);
+			return -ENOSPC;
+		}
+
+		can_write = S3C_GNPTXSTS_NPTxFSpcAvail_GET(gnptxsts);
+	}
+
+	dev_dbg(hsotg->dev, "%s: GNPTXSTS=%08x, can=%d, to=%d, mps %d\n",
+		 __func__, gnptxsts, can_write, to_write, hs_ep->ep.maxpacket);
+
+	/* limit to 512 bytes of data, it seems at least on the non-periodic
+	 * FIFO, requests of >512 cause the endpoint to get stuck with a
+	 * fragment of the end of the transfer in it.
+	 */
+	if (can_write > 512)
+		can_write = 512;
+
+	/* see if we can write data */
+
+	if (to_write > can_write) {
+		to_write = can_write;
+		pkt_round = to_write % hs_ep->ep.maxpacket;
+
+		/* Not sure, but we probably shouldn't be writing partial
+		 * packets into the FIFO, so round the write down to an
+		 * exact number of packets.
+		 *
+		 * Note, we do not currently check to see if we can ever
+		 * write a full packet or not to the FIFO.
+		 */
+
+		if (pkt_round)
+			to_write -= pkt_round;
+
+		/* enable correct FIFO interrupt to alert us when there
+		 * is more room left. */
+
+		s3c_hsotg_en_gsint(hsotg,
+				   periodic ? S3C_GINTSTS_PTxFEmp :
+				   S3C_GINTSTS_NPTxFEmp);
+	}
+
+	dev_dbg(hsotg->dev, "write %d/%d, can_write %d, done %d\n",
+		 to_write, hs_req->req.length, can_write, buf_pos);
+
+	if (to_write <= 0)
+		return -ENOSPC;
+
+	hs_req->req.actual = buf_pos + to_write;
+	hs_ep->total_data += to_write;
+
+	if (periodic)
+		hs_ep->fifo_load += to_write;
+
+	to_write = DIV_ROUND_UP(to_write, 4);
+	data = hs_req->req.buf + buf_pos;
+
+	writesl(hsotg->regs + S3C_EPFIFO(hs_ep->index), data, to_write);
+
+	return (to_write >= can_write) ? -ENOSPC : 0;
+}
+
+/**
+ * get_ep_limit - get the maximum data legnth for this endpoint
+ * @hs_ep: The endpoint
+ *
+ * Return the maximum data that can be queued in one go on a given endpoint
+ * so that transfers that are too long can be split.
+ */
+static unsigned get_ep_limit(struct s3c_hsotg_ep *hs_ep)
+{
+	int index = hs_ep->index;
+	unsigned maxsize;
+	unsigned maxpkt;
+
+	if (index != 0) {
+		maxsize = S3C_DxEPTSIZ_XferSize_LIMIT + 1;
+		maxpkt = S3C_DxEPTSIZ_PktCnt_LIMIT + 1;
+	} else {
+		if (hs_ep->dir_in) {
+			/* maxsize = S3C_DIEPTSIZ0_XferSize_LIMIT + 1; */
+			maxsize = 64+64+1;
+			maxpkt = S3C_DIEPTSIZ0_PktCnt_LIMIT + 1;
+		} else {
+			maxsize = 0x3f;
+			maxpkt = 2;
+		}
+	}
+
+	/* we made the constant loading easier above by using +1 */
+	maxpkt--;
+	maxsize--;
+
+	/* constrain by packet count if maxpkts*pktsize is greater
+	 * than the length register size. */
+
+	if ((maxpkt * hs_ep->ep.maxpacket) < maxsize)
+		maxsize = maxpkt * hs_ep->ep.maxpacket;
+
+	return maxsize;
+}
+
+/**
+ * s3c_hsotg_start_req - start a USB request from an endpoint's queue
+ * @hsotg: The controller state.
+ * @hs_ep: The endpoint to process a request for
+ * @hs_req: The request to start.
+ * @continuing: True if we are doing more for the current request.
+ *
+ * Start the given request running by setting the endpoint registers
+ * appropriately, and writing any data to the FIFOs.
+ */
+static void s3c_hsotg_start_req(struct s3c_hsotg *hsotg,
+				struct s3c_hsotg_ep *hs_ep,
+				struct s3c_hsotg_req *hs_req,
+				bool continuing)
+{
+	struct usb_request *ureq = &hs_req->req;
+	int index = hs_ep->index;
+	int dir_in = hs_ep->dir_in;
+	u32 epctrl_reg;
+	u32 epsize_reg;
+	u32 epsize;
+	u32 ctrl;
+	unsigned length;
+	unsigned packets;
+	unsigned maxreq;
+
+	if (index != 0) {
+		if (hs_ep->req && !continuing) {
+			dev_err(hsotg->dev, "%s: active request\n", __func__);
+			WARN_ON(1);
+			return;
+		} else if (hs_ep->req != hs_req && continuing) {
+			dev_err(hsotg->dev,
+				"%s: continue different req\n", __func__);
+			WARN_ON(1);
+			return;
+		}
+	}
+
+	epctrl_reg = dir_in ? S3C_DIEPCTL(index) : S3C_DOEPCTL(index);
+	epsize_reg = dir_in ? S3C_DIEPTSIZ(index) : S3C_DOEPTSIZ(index);
+
+	dev_dbg(hsotg->dev, "%s: DxEPCTL=0x%08x, ep %d, dir %s\n",
+		__func__, readl(hsotg->regs + epctrl_reg), index,
+		hs_ep->dir_in ? "in" : "out");
+
+	length = ureq->length - ureq->actual;
+
+	if (0)
+		dev_dbg(hsotg->dev,
+			"REQ buf %p len %d dma 0x%08x noi=%d zp=%d snok=%d\n",
+			ureq->buf, length, ureq->dma,
+			ureq->no_interrupt, ureq->zero, ureq->short_not_ok);
+
+	maxreq = get_ep_limit(hs_ep);
+	if (length > maxreq) {
+		int round = maxreq % hs_ep->ep.maxpacket;
+
+		dev_dbg(hsotg->dev, "%s: length %d, max-req %d, r %d\n",
+			__func__, length, maxreq, round);
+
+		/* round down to multiple of packets */
+		if (round)
+			maxreq -= round;
+
+		length = maxreq;
+	}
+
+	if (length)
+		packets = DIV_ROUND_UP(length, hs_ep->ep.maxpacket);
+	else
+		packets = 1;	/* send one packet if length is zero. */
+
+	if (dir_in && index != 0)
+		epsize = S3C_DxEPTSIZ_MC(1);
+	else
+		epsize = 0;
+
+	if (index != 0 && ureq->zero) {
+		/* test for the packets being exactly right for the
+		 * transfer */
+
+		if (length == (packets * hs_ep->ep.maxpacket))
+			packets++;
+	}
+
+	epsize |= S3C_DxEPTSIZ_PktCnt(packets);
+	epsize |= S3C_DxEPTSIZ_XferSize(length);
+
+	dev_dbg(hsotg->dev, "%s: %d@%d/%d, 0x%08x => 0x%08x\n",
+		__func__, packets, length, ureq->length, epsize, epsize_reg);
+
+	/* store the request as the current one we're doing */
+	hs_ep->req = hs_req;
+
+	/* write size / packets */
+	writel(epsize, hsotg->regs + epsize_reg);
+
+	ctrl = readl(hsotg->regs + epctrl_reg);
+
+	if (ctrl & S3C_DxEPCTL_Stall) {
+		dev_warn(hsotg->dev, "%s: ep%d is stalled\n", __func__, index);
+
+		/* not sure what we can do here, if it is EP0 then we should
+		 * get this cleared once the endpoint has transmitted the
+		 * STALL packet, otherwise it needs to be cleared by the
+		 * host.
+		 */
+	}
+
+	if (using_dma(hsotg)) {
+		unsigned int dma_reg;
+
+		/* write DMA address to control register, buffer already
+		 * synced by s3c_hsotg_ep_queue().  */
+
+		dma_reg = dir_in ? S3C_DIEPDMA(index) : S3C_DOEPDMA(index);
+		writel(ureq->dma, hsotg->regs + dma_reg);
+
+		dev_dbg(hsotg->dev, "%s: 0x%08x => 0x%08x\n",
+			__func__, ureq->dma, dma_reg);
+	}
+
+	ctrl |= S3C_DxEPCTL_EPEna;	/* ensure ep enabled */
+	ctrl |= S3C_DxEPCTL_USBActEp;
+	ctrl |= S3C_DxEPCTL_CNAK;	/* clear NAK set by core */
+
+	dev_dbg(hsotg->dev, "%s: DxEPCTL=0x%08x\n", __func__, ctrl);
+	writel(ctrl, hsotg->regs + epctrl_reg);
+
+	/* set these, it seems that DMA support increments past the end
+	 * of the packet buffer so we need to calculate the length from
+	 * this information. */
+	hs_ep->size_loaded = length;
+	hs_ep->last_load = ureq->actual;
+
+	if (dir_in && !using_dma(hsotg)) {
+		/* set these anyway, we may need them for non-periodic in */
+		hs_ep->fifo_load = 0;
+
+		s3c_hsotg_write_fifo(hsotg, hs_ep, hs_req);
+	}
+
+	/* clear the INTknTXFEmpMsk when we start request, more as a aide
+	 * to debugging to see what is going on. */
+	if (dir_in)
+		writel(S3C_DIEPMSK_INTknTXFEmpMsk,
+		       hsotg->regs + S3C_DIEPINT(index));
+
+	/* Note, trying to clear the NAK here causes problems with transmit
+	 * on the S3C6400 ending up with the TXFIFO becomming full. */
+
+	/* check ep is enabled */
+	if (!(readl(hsotg->regs + epctrl_reg) & S3C_DxEPCTL_EPEna))
+		dev_warn(hsotg->dev,
+			 "ep%d: failed to become enabled (DxEPCTL=0x%08x)?\n",
+			 index, readl(hsotg->regs + epctrl_reg));
+
+	dev_dbg(hsotg->dev, "%s: DxEPCTL=0x%08x\n",
+		__func__, readl(hsotg->regs + epctrl_reg));
+}
+
+/**
+ * s3c_hsotg_map_dma - map the DMA memory being used for the request
+ * @hsotg: The device state.
+ * @hs_ep: The endpoint the request is on.
+ * @req: The request being processed.
+ *
+ * We've been asked to queue a request, so ensure that the memory buffer
+ * is correctly setup for DMA. If we've been passed an extant DMA address
+ * then ensure the buffer has been synced to memory. If our buffer has no
+ * DMA memory, then we map the memory and mark our request to allow us to
+ * cleanup on completion.
+*/
+static int s3c_hsotg_map_dma(struct s3c_hsotg *hsotg,
+			     struct s3c_hsotg_ep *hs_ep,
+			     struct usb_request *req)
+{
+	enum dma_data_direction dir;
+	struct s3c_hsotg_req *hs_req = our_req(req);
+
+	dir = hs_ep->dir_in ? DMA_TO_DEVICE : DMA_FROM_DEVICE;
+
+	/* if the length is zero, ignore the DMA data */
+	if (hs_req->req.length == 0)
+		return 0;
+
+	if (req->dma == DMA_ADDR_INVALID) {
+		dma_addr_t dma;
+
+		dma = dma_map_single(hsotg->dev, req->buf, req->length, dir);
+
+		if (unlikely(dma_mapping_error(hsotg->dev, dma)))
+			goto dma_error;
+
+		if (dma & 3) {
+			dev_err(hsotg->dev, "%s: unaligned dma buffer\n",
+				__func__);
+
+			dma_unmap_single(hsotg->dev, dma, req->length, dir);
+			return -EINVAL;
+		}
+
+		hs_req->mapped = 1;
+		req->dma = dma;
+	} else {
+		dma_sync_single(hsotg->dev, req->dma, req->length, dir);
+		hs_req->mapped = 0;
+	}
+
+	return 0;
+
+dma_error:
+	dev_err(hsotg->dev, "%s: failed to map buffer %p, %d bytes\n",
+		__func__, req->buf, req->length);
+
+	return -EIO;
+}
+
+static int s3c_hsotg_ep_queue(struct usb_ep *ep, struct usb_request *req,
+			      gfp_t gfp_flags)
+{
+	struct s3c_hsotg_req *hs_req = our_req(req);
+	struct s3c_hsotg_ep *hs_ep = our_ep(ep);
+	struct s3c_hsotg *hs = hs_ep->parent;
+	unsigned long irqflags;
+	bool first;
+
+	dev_dbg(hs->dev, "%s: req %p: %d@%p, noi=%d, zero=%d, snok=%d\n",
+		ep->name, req, req->length, req->buf, req->no_interrupt,
+		req->zero, req->short_not_ok);
+
+	/* initialise status of the request */
+	INIT_LIST_HEAD(&hs_req->queue);
+	req->actual = 0;
+	req->status = -EINPROGRESS;
+
+	/* if we're using DMA, sync the buffers as necessary */
+	if (using_dma(hs)) {
+		int ret = s3c_hsotg_map_dma(hs, hs_ep, req);
+		if (ret)
+			return ret;
+	}
+
+	spin_lock_irqsave(&hs_ep->lock, irqflags);
+
+	first = list_empty(&hs_ep->queue);
+	list_add_tail(&hs_req->queue, &hs_ep->queue);
+
+	if (first)
+		s3c_hsotg_start_req(hs, hs_ep, hs_req, false);
+
+	spin_unlock_irqrestore(&hs_ep->lock, irqflags);
+
+	return 0;
+}
+
+static void s3c_hsotg_ep_free_request(struct usb_ep *ep,
+				      struct usb_request *req)
+{
+	struct s3c_hsotg_req *hs_req = our_req(req);
+
+	kfree(hs_req);
+}
+
+/**
+ * s3c_hsotg_complete_oursetup - setup completion callback
+ * @ep: The endpoint the request was on.
+ * @req: The request completed.
+ *
+ * Called on completion of any requests the driver itself
+ * submitted that need cleaning up.
+ */
+static void s3c_hsotg_complete_oursetup(struct usb_ep *ep,
+					struct usb_request *req)
+{
+	struct s3c_hsotg_ep *hs_ep = our_ep(ep);
+	struct s3c_hsotg *hsotg = hs_ep->parent;
+
+	dev_dbg(hsotg->dev, "%s: ep %p, req %p\n", __func__, ep, req);
+
+	s3c_hsotg_ep_free_request(ep, req);
+}
+
+/**
+ * ep_from_windex - convert control wIndex value to endpoint
+ * @hsotg: The driver state.
+ * @windex: The control request wIndex field (in host order).
+ *
+ * Convert the given wIndex into a pointer to an driver endpoint
+ * structure, or return NULL if it is not a valid endpoint.
+*/
+static struct s3c_hsotg_ep *ep_from_windex(struct s3c_hsotg *hsotg,
+					   u32 windex)
+{
+	struct s3c_hsotg_ep *ep = &hsotg->eps[windex & 0x7F];
+	int dir = (windex & USB_DIR_IN) ? 1 : 0;
+	int idx = windex & 0x7F;
+
+	if (windex >= 0x100)
+		return NULL;
+
+	if (idx > S3C_HSOTG_EPS)
+		return NULL;
+
+	if (idx && ep->dir_in != dir)
+		return NULL;
+
+	return ep;
+}
+
+/**
+ * s3c_hsotg_send_reply - send reply to control request
+ * @hsotg: The device state
+ * @ep: Endpoint 0
+ * @buff: Buffer for request
+ * @length: Length of reply.
+ *
+ * Create a request and queue it on the given endpoint. This is useful as
+ * an internal method of sending replies to certain control requests, etc.
+ */
+static int s3c_hsotg_send_reply(struct s3c_hsotg *hsotg,
+				struct s3c_hsotg_ep *ep,
+				void *buff,
+				int length)
+{
+	struct usb_request *req;
+	int ret;
+
+	dev_dbg(hsotg->dev, "%s: buff %p, len %d\n", __func__, buff, length);
+
+	req = s3c_hsotg_ep_alloc_request(&ep->ep, GFP_ATOMIC);
+	hsotg->ep0_reply = req;
+	if (!req) {
+		dev_warn(hsotg->dev, "%s: cannot alloc req\n", __func__);
+		return -ENOMEM;
+	}
+
+	req->buf = hsotg->ep0_buff;
+	req->length = length;
+	req->zero = 1; /* always do zero-length final transfer */
+	req->complete = s3c_hsotg_complete_oursetup;
+
+	if (length)
+		memcpy(req->buf, buff, length);
+	else
+		ep->sent_zlp = 1;
+
+	ret = s3c_hsotg_ep_queue(&ep->ep, req, GFP_ATOMIC);
+	if (ret) {
+		dev_warn(hsotg->dev, "%s: cannot queue req\n", __func__);
+		return ret;
+	}
+
+	return 0;
+}
+
+/**
+ * s3c_hsotg_process_req_status - process request GET_STATUS
+ * @hsotg: The device state
+ * @ctrl: USB control request
+ */
+static int s3c_hsotg_process_req_status(struct s3c_hsotg *hsotg,
+					struct usb_ctrlrequest *ctrl)
+{
+	struct s3c_hsotg_ep *ep0 = &hsotg->eps[0];
+	struct s3c_hsotg_ep *ep;
+	__le16 reply;
+	int ret;
+
+	dev_dbg(hsotg->dev, "%s: USB_REQ_GET_STATUS\n", __func__);
+
+	if (!ep0->dir_in) {
+		dev_warn(hsotg->dev, "%s: direction out?\n", __func__);
+		return -EINVAL;
+	}
+
+	switch (ctrl->bRequestType & USB_RECIP_MASK) {
+	case USB_RECIP_DEVICE:
+		reply = cpu_to_le16(0); /* bit 0 => self powered,
+					 * bit 1 => remote wakeup */
+		break;
+
+	case USB_RECIP_INTERFACE:
+		/* currently, the data result should be zero */
+		reply = cpu_to_le16(0);
+		break;
+
+	case USB_RECIP_ENDPOINT:
+		ep = ep_from_windex(hsotg, le16_to_cpu(ctrl->wIndex));
+		if (!ep)
+			return -ENOENT;
+
+		reply = cpu_to_le16(ep->halted ? 1 : 0);
+		break;
+
+	default:
+		return 0;
+	}
+
+	if (le16_to_cpu(ctrl->wLength) != 2)
+		return -EINVAL;
+
+	ret = s3c_hsotg_send_reply(hsotg, ep0, &reply, 2);
+	if (ret) {
+		dev_err(hsotg->dev, "%s: failed to send reply\n", __func__);
+		return ret;
+	}
+
+	return 1;
+}
+
+static int s3c_hsotg_ep_sethalt(struct usb_ep *ep, int value);
+
+/**
+ * s3c_hsotg_process_req_featire - process request {SET,CLEAR}_FEATURE
+ * @hsotg: The device state
+ * @ctrl: USB control request
+ */
+static int s3c_hsotg_process_req_feature(struct s3c_hsotg *hsotg,
+					 struct usb_ctrlrequest *ctrl)
+{
+	bool set = (ctrl->bRequest == USB_REQ_SET_FEATURE);
+	struct s3c_hsotg_ep *ep;
+
+	dev_dbg(hsotg->dev, "%s: %s_FEATURE\n",
+		__func__, set ? "SET" : "CLEAR");
+
+	if (ctrl->bRequestType == USB_RECIP_ENDPOINT) {
+		ep = ep_from_windex(hsotg, le16_to_cpu(ctrl->wIndex));
+		if (!ep) {
+			dev_dbg(hsotg->dev, "%s: no endpoint for 0x%04x\n",
+				__func__, le16_to_cpu(ctrl->wIndex));
+			return -ENOENT;
+		}
+
+		switch (le16_to_cpu(ctrl->wValue)) {
+		case USB_ENDPOINT_HALT:
+			s3c_hsotg_ep_sethalt(&ep->ep, set);
+			break;
+
+		default:
+			return -ENOENT;
+		}
+	} else
+		return -ENOENT;  /* currently only deal with endpoint */
+
+	return 1;
+}
+
+/**
+ * s3c_hsotg_process_control - process a control request
+ * @hsotg: The device state
+ * @ctrl: The control request received
+ *
+ * The controller has received the SETUP phase of a control request, and
+ * needs to work out what to do next (and whether to pass it on to the
+ * gadget driver).
+ */
+static void s3c_hsotg_process_control(struct s3c_hsotg *hsotg,
+				      struct usb_ctrlrequest *ctrl)
+{
+	struct s3c_hsotg_ep *ep0 = &hsotg->eps[0];
+	int ret = 0;
+	u32 dcfg;
+
+	ep0->sent_zlp = 0;
+
+	dev_dbg(hsotg->dev, "ctrl Req=%02x, Type=%02x, V=%04x, L=%04x\n",
+		 ctrl->bRequest, ctrl->bRequestType,
+		 ctrl->wValue, ctrl->wLength);
+
+	/* record the direction of the request, for later use when enquing
+	 * packets onto EP0. */
+
+	ep0->dir_in = (ctrl->bRequestType & USB_DIR_IN) ? 1 : 0;
+	dev_dbg(hsotg->dev, "ctrl: dir_in=%d\n", ep0->dir_in);
+
+	/* if we've no data with this request, then the last part of the
+	 * transaction is going to implicitly be IN. */
+	if (ctrl->wLength == 0)
+		ep0->dir_in = 1;
+
+	if ((ctrl->bRequestType & USB_TYPE_MASK) == USB_TYPE_STANDARD) {
+		switch (ctrl->bRequest) {
+		case USB_REQ_SET_ADDRESS:
+			dcfg = readl(hsotg->regs + S3C_DCFG);
+			dcfg &= ~S3C_DCFG_DevAddr_MASK;
+			dcfg |= ctrl->wValue << S3C_DCFG_DevAddr_SHIFT;
+			writel(dcfg, hsotg->regs + S3C_DCFG);
+
+			dev_info(hsotg->dev, "new address %d\n", ctrl->wValue);
+
+			ret = s3c_hsotg_send_reply(hsotg, ep0, NULL, 0);
+			return;
+
+		case USB_REQ_GET_STATUS:
+			ret = s3c_hsotg_process_req_status(hsotg, ctrl);
+			break;
+
+		case USB_REQ_CLEAR_FEATURE:
+		case USB_REQ_SET_FEATURE:
+			ret = s3c_hsotg_process_req_feature(hsotg, ctrl);
+			break;
+		}
+	}
+
+	/* as a fallback, try delivering it to the driver to deal with */
+
+	if (ret == 0 && hsotg->driver) {
+		ret = hsotg->driver->setup(&hsotg->gadget, ctrl);
+		if (ret < 0)
+			dev_dbg(hsotg->dev, "driver->setup() ret %d\n", ret);
+	}
+
+	if (ret > 0) {
+		if (!ep0->dir_in) {
+			/* need to generate zlp in reply or take data */
+			/* todo - deal with any data we might be sent? */
+			ret = s3c_hsotg_send_reply(hsotg, ep0, NULL, 0);
+		}