USB device DMA support on OMAP2

USB device DMA support on OMAP2

[Kyungmin Park]

The current omap udc dosen't support the DMA mode
and it has some problem at setup time on OMAP2 with previous patch file.
I found that the code assumes bulk out required the big data transfer.
But MODE SELECT(6) sent the only 24 bytes. it makes a problem.
So I implement the small packets handling for it.

It is tested with both linux and windows.

Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
---
diff --git a/drivers/usb/gadget/omap_udc.c b/drivers/usb/gadget/omap_udc.c
index 983337e..9719f08 100644
--- a/drivers/usb/gadget/omap_udc.c
+++ b/drivers/usb/gadget/omap_udc.c
@@ -4,6 +4,8 @@
  * Copyright (C) 2004 Texas Instruments, Inc.
  * Copyright (C) 2004-2005 David Brownell
  *
+ * OMAP2 & DMA support by Kyungmin Park <kyungmin.park@samsung.com>
+ *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
  * the Free Software Foundation; either version 2 of the License, or
@@ -60,11 +62,6 @@
 /* bulk DMA seems to be behaving for both IN and OUT */
 #define	USE_DMA
 
-/* FIXME: OMAP2 currently has some problem in DMA mode */
-#ifdef CONFIG_ARCH_OMAP2
-#undef USE_DMA
-#endif
-
 /* ISO too */
 #define	USE_ISO
 
@@ -73,6 +70,8 @@
 
 #define	DMA_ADDR_INVALID	(~(dma_addr_t)0)
 
+#define OMAP2_DMA_CH(ch)	(((ch) - 1) << 1)
+#define OMAP24XX_DMA(name, ch)	(OMAP24XX_DMA_##name + OMAP2_DMA_CH(ch))
 
 /*
  * The OMAP UDC needs _very_ early endpoint setup:  before enabling the
@@ -571,20 +570,25 @@ static void next_in_dma(struct omap_ep *ep, struct omap_req *req)
 	const int	sync_mode = cpu_is_omap15xx()
 				? OMAP_DMA_SYNC_FRAME
 				: OMAP_DMA_SYNC_ELEMENT;
+	int		dma_trigger = 0;
+
+	if (cpu_is_omap24xx())
+		dma_trigger = OMAP24XX_DMA(USB_W2FC_TX0, ep->dma_channel);
 
 	/* measure length in either bytes or packets */
 	if ((cpu_is_omap16xx() && length <= UDC_TXN_TSC)
+			|| (cpu_is_omap24xx() && length < ep->maxpacket)
 			|| (cpu_is_omap15xx() && length < ep->maxpacket)) {
 		txdma_ctrl = UDC_TXN_EOT | length;
 		omap_set_dma_transfer_params(ep->lch, OMAP_DMA_DATA_TYPE_S8,
-				length, 1, sync_mode, 0, 0);
+				length, 1, sync_mode, dma_trigger, 0);
 	} else {
 		length = min(length / ep->maxpacket,
 				(unsigned) UDC_TXN_TSC + 1);
 		txdma_ctrl = length;
 		omap_set_dma_transfer_params(ep->lch, OMAP_DMA_DATA_TYPE_S16,
 				ep->ep.maxpacket >> 1, length, sync_mode,
-				0, 0);
+				dma_trigger, 0);
 		length *= ep->maxpacket;
 	}
 	omap_set_dma_src_params(ep->lch, OMAP_DMA_PORT_EMIFF,
@@ -622,20 +626,31 @@ static void finish_in_dma(struct omap_ep *ep, struct omap_req *req, int status)
 
 static void next_out_dma(struct omap_ep *ep, struct omap_req *req)
 {
-	unsigned packets;
+	unsigned packets = req->req.length - req->req.actual;
+	int dma_trigger = 0;
+
+	if (cpu_is_omap24xx())
+		dma_trigger = OMAP24XX_DMA(USB_W2FC_RX0, ep->dma_channel);
 
 	/* NOTE:  we filtered out "short reads" before, so we know
 	 * the buffer has only whole numbers of packets.
+	 * except MODE SELECT(6) sent the 24 bytes data in OMAP24XX DMA mode
 	 */
-
-	/* set up this DMA transfer, enable the fifo, start */
-	packets = (req->req.length - req->req.actual) / ep->ep.maxpacket;
-	packets = min(packets, (unsigned)UDC_RXN_TC + 1);
-	req->dma_bytes = packets * ep->ep.maxpacket;
-	omap_set_dma_transfer_params(ep->lch, OMAP_DMA_DATA_TYPE_S16,
-			ep->ep.maxpacket >> 1, packets,
-			OMAP_DMA_SYNC_ELEMENT,
-			0, 0);
+	if (cpu_is_omap24xx() && packets < ep->maxpacket) {
+		omap_set_dma_transfer_params(ep->lch, OMAP_DMA_DATA_TYPE_S8,
+				packets, 1, OMAP_DMA_SYNC_ELEMENT,
+				dma_trigger, 0);
+		req->dma_bytes = packets;
+	} else {
+		/* set up this DMA transfer, enable the fifo, start */
+		packets /= ep->ep.maxpacket;
+		packets = min(packets, (unsigned)UDC_RXN_TC + 1);
+		req->dma_bytes = packets * ep->ep.maxpacket;
+		omap_set_dma_transfer_params(ep->lch, OMAP_DMA_DATA_TYPE_S16,
+				ep->ep.maxpacket >> 1, packets,
+				OMAP_DMA_SYNC_ELEMENT,
+				dma_trigger, 0);
+	}
 	omap_set_dma_dest_params(ep->lch, OMAP_DMA_PORT_EMIFF,
 		OMAP_DMA_AMODE_POST_INC, req->req.dma + req->req.actual,
 		0, 0);
@@ -743,6 +758,7 @@ static void dma_channel_claim(struct omap_ep *ep, unsigned channel)
 {
 	u16	reg;
 	int	status, restart, is_in;
+	int	dma_channel;
 
 	is_in = ep->bEndpointAddress & USB_DIR_IN;
 	if (is_in)
@@ -769,11 +785,15 @@ static void dma_channel_claim(struct omap_ep *ep, unsigned channel)
 	ep->dma_channel = channel;
 
 	if (is_in) {
-		status = omap_request_dma(OMAP_DMA_USB_W2FC_TX0 - 1 + channel,
+		if (cpu_is_omap24xx())
+			dma_channel = OMAP24XX_DMA(USB_W2FC_TX0, channel);
+		else
+			dma_channel = OMAP_DMA_USB_W2FC_TX0 - 1 + channel;
+		status = omap_request_dma(dma_channel,
 			ep->ep.name, dma_error, ep, &ep->lch);
 		if (status == 0) {
 			UDC_TXDMA_CFG_REG = reg;
-			/* EMIFF */
+			/* EMIFF or SDRC */
 			omap_set_dma_src_burst_mode(ep->lch,
 						OMAP_DMA_DATA_BURST_4);
 			omap_set_dma_src_data_pack(ep->lch, 1);
@@ -785,7 +805,12 @@ static void dma_channel_claim(struct omap_ep *ep, unsigned channel)
 				0, 0);
 		}
 	} else {
-		status = omap_request_dma(OMAP_DMA_USB_W2FC_RX0 - 1 + channel,
+		if (cpu_is_omap24xx())
+			dma_channel = OMAP24XX_DMA(USB_W2FC_RX0, channel);
+		else
+			dma_channel = OMAP_DMA_USB_W2FC_RX0 - 1 + channel;
+
+		status = omap_request_dma(dma_channel,
 			ep->ep.name, dma_error, ep, &ep->lch);
 		if (status == 0) {
 			UDC_RXDMA_CFG_REG = reg;
@@ -795,7 +820,7 @@ static void dma_channel_claim(struct omap_ep *ep, unsigned channel)
 				OMAP_DMA_AMODE_CONSTANT,
 				(unsigned long) io_v2p((u32)&UDC_DATA_DMA_REG),
 				0, 0);
-			/* EMIFF */
+			/* EMIFF or SDRC */
 			omap_set_dma_dest_burst_mode(ep->lch,
 						OMAP_DMA_DATA_BURST_4);
 			omap_set_dma_dest_data_pack(ep->lch, 1);
@@ -808,7 +833,7 @@ static void dma_channel_claim(struct omap_ep *ep, unsigned channel)
 		omap_disable_dma_irq(ep->lch, OMAP_DMA_BLOCK_IRQ);
 
 		/* channel type P: hw synch (fifo) */
-		if (!cpu_is_omap15xx())
+		if (cpu_class_is_omap1() && !cpu_is_omap15xx())
 			OMAP1_DMA_LCH_CTRL_REG(ep->lch) = 2;
 	}
 
@@ -926,11 +951,13 @@ omap_ep_queue(struct usb_ep *_ep, struct usb_request *_req, gfp_t gfp_flags)
 
 	/* this isn't bogus, but OMAP DMA isn't the only hardware to
 	 * have a hard time with partial packet reads...  reject it.
+	 * Except OMAP2 can handle the small packets.
 	 */
 	if (use_dma
 			&& ep->has_dma
 			&& ep->bEndpointAddress != 0
 			&& (ep->bEndpointAddress & USB_DIR_IN) == 0
+			&& !cpu_class_is_omap2()
 			&& (req->req.length % ep->ep.maxpacket) != 0) {
 		DBG("%s, no partial packet OUT reads\n", __FUNCTION__);
 		return -EMSGSIZE;

Re: USB device DMA support on OMAP2

[David Brownell]

> The current omap udc dosen't support the DMA mode

It most certainly does!!  I think you it has DMA issues on OMAP2.


> and it has some problem at setup time on OMAP2 with previous patch file.
> I found that the code assumes bulk out required the big data transfer.
> But MODE SELECT(6) sent the only 24 bytes. it makes a problem.
> So I implement the small packets handling for it.
>
> It is tested with both linux and windows.
>
> Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>

I'll have a look at it, thanks.

Is this aginst the linux-omap tree, or current kernel.org code?
(There _shouldn't_ be any differences, but I've not checked in
some time now and such issues do creep in from time to time.)

- Dave

Re: USB device DMA support on OMAP2

[Kyungmin Park]

On 11/5/07, David Brownell <david-b@pacbell.net> wrote:
> > The current omap udc dosen't support the DMA mode
>
> It most certainly does!!  I think you it has DMA issues on OMAP2.
>

Ah, Yes, OMAP1 supports the DMA mode. It means the OMAP2 DMA support.

>
> > and it has some problem at setup time on OMAP2 with previous patch file.
> > I found that the code assumes bulk out required the big data transfer.
> > But MODE SELECT(6) sent the only 24 bytes. it makes a problem.
> > So I implement the small packets handling for it.
> >
> > It is tested with both linux and windows.
> >
> > Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
>
> I'll have a look at it, thanks.
>
> Is this aginst the linux-omap tree, or current kernel.org code?
> (There _shouldn't_ be any differences, but I've not checked in
> some time now and such issues do creep in from time to time.)
>

There's some difference related with double buffering.
It's against the latest kernel code. I checked it.

Thank you,
Kyungmin Park