Re: [PATCH] USB: DWC3: Fix missed isoc IN transaction

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]


On 5/18/2012 7:00 PM, Felipe Balbi wrote:
Hi,

On Tue, May 15, 2012 at 07:49:52PM +0530, Pratyush Anand wrote:
If an IN transfer is missed on isoc endpoint, then driver must insure
that next ep_queue is properly handled.
This patch fixes this issue by starting a new transfer for next queued
request.
Only limitation is that, gadget will have to submit a request within 4
uf time, which dwc3 driver considers safe enough for staring a new
transfer.

Signed-off-by: Pratyush Anand<pratyush.anand@xxxxxx>
---
  drivers/usb/dwc3/core.h   |    3 +++
  drivers/usb/dwc3/gadget.c |   36 +++++++++++++++++++++++++++++++-----
  2 files changed, 34 insertions(+), 5 deletions(-)

diff --git a/drivers/usb/dwc3/core.h b/drivers/usb/dwc3/core.h
index 2a920f8..c3f61f6 100644
--- a/drivers/usb/dwc3/core.h
+++ b/drivers/usb/dwc3/core.h
@@ -352,6 +352,7 @@ struct dwc3_event_buffer {
   * @number: endpoint number (1 - 15)
   * @type: set to bmAttributes&  USB_ENDPOINT_XFERTYPE_MASK
   * @res_trans_idx: Resource transfer index
+ * @event: pointer to event struct received during missed isoc xferinprogress
   * @interval: the intervall on which the ISOC transfer is started
   * @name: a human readable name e.g. ep1out-bulk
   * @direction: true for TX, false for RX
@@ -376,6 +377,7 @@ struct dwc3_ep {
  #define DWC3_EP_WEDGE		(1<<  2)
  #define DWC3_EP_BUSY		(1<<  4)
  #define DWC3_EP_PENDING_REQUEST	(1<<  5)
+#define DWC3_EP_MISSED_ISOC	(1<<  6)

  	/* This last one is specific to EP0 */
  #define DWC3_EP0_DIR_IN		(1<<  31)
@@ -385,6 +387,7 @@ struct dwc3_ep {
  	u8			number;
  	u8			type;
  	u8			res_trans_idx;
+	const struct dwc3_event_depevt *event;

please don't do this... (see more below)

  	u32			interval;

  	char			name[20];
diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
index 9affe67..5a9f1f19 100644
--- a/drivers/usb/dwc3/gadget.c
+++ b/drivers/usb/dwc3/gadget.c
@@ -995,6 +995,9 @@ static int __dwc3_gadget_kick_transfer(struct dwc3_ep *dep, u16 cmd_param,
  	return 0;
  }

+static void dwc3_gadget_start_isoc(struct dwc3 *dwc,
+		struct dwc3_ep *dep, const struct dwc3_event_depevt *event);

you can make two versions of this function by re-factoring. One version
depends on event and another depends on direct arguments. Similar to
what we did for ep0 ;-)

All you need from the event to start the isochronous transfer is the
uframe number, and we have dwc3_gadget_get_frame() (which should be
re-factored too not to depend on struct usb_gadget *g, something like
below:

Ok, got it.
I did not notice dwc3_gadget_get_frame earlier.
Will do it.


diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
index b9cbde5..dbe9b2b 100644
--- a/drivers/usb/dwc3/gadget.c
+++ b/drivers/usb/dwc3/gadget.c
@@ -1288,15 +1288,22 @@ static const struct usb_ep_ops dwc3_gadget_ep_ops = {

  /* -------------------------------------------------------------------------- */

-static int dwc3_gadget_get_frame(struct usb_gadget *g)
+static int __dwc3_gadget_get_frame(struct dwc3 *dwc)
  {
-	struct dwc3		*dwc = gadget_to_dwc(g);
  	u32			reg;

  	reg = dwc3_readl(dwc->regs, DWC3_DSTS);
+
  	return DWC3_DSTS_SOFFN(reg);
  }

+static int dwc3_gadget_get_frame(struct usb_gadget *g)
+{
+	struct dwc3		*dwc = gadget_to_dwc(g);
+
+	return __dwc3_gadget_get_frame(dwc);
+}
+
  static int dwc3_gadget_wakeup(struct usb_gadget *g)
  {
  	struct dwc3		*dwc = gadget_to_dwc(g);
)

Then you can properly start the isochronous transfer whenever you miss
one. No requirement of doing so on the next 4 uframes.

@@ -1024,8 +1027,19 @@ static int __dwc3_gadget_ep_queue(struct dwc3_ep *dep, struct dwc3_request *req)

  	list_add_tail(&req->list,&dep->request_list);

-	if (usb_endpoint_xfer_isoc(dep->desc)&&  (dep->flags&  DWC3_EP_BUSY))
-		dep->flags |= DWC3_EP_PENDING_REQUEST;
+	if (usb_endpoint_xfer_isoc(dep->desc)) {
+		if (dep->flags&  DWC3_EP_BUSY) {
+			dep->flags |= DWC3_EP_PENDING_REQUEST;
+		} else if (dep->flags&  DWC3_EP_MISSED_ISOC) {
+			/*
+			 * FIXME: Only case which I still see failing is
+			 * if gadget delays submission for more than 4
+			 * uf time.
+			 */
+			dwc3_gadget_start_isoc(dwc, dep, dep->event);
+			dep->flags&= ~DWC3_EP_MISSED_ISOC;
+		}
+	}

  	/*
  	 * There are two special cases:
@@ -1590,6 +1604,7 @@ static int dwc3_cleanup_done_reqs(struct dwc3 *dwc, struct dwc3_ep *dep,
  	struct dwc3_trb		*trb;
  	unsigned int		count;
  	unsigned int		s_pkt = 0;
+	unsigned int		m_isoc = 0;

  	do {
  		req = next_request(&dep->req_queued);
@@ -1615,9 +1630,18 @@ static int dwc3_cleanup_done_reqs(struct dwc3 *dwc, struct dwc3_ep *dep,

  		if (dep->direction) {
  			if (count) {
-				dev_err(dwc->dev, "incomplete IN transfer %s\n",
-						dep->name);
-				status = -ECONNRESET;
+				if (usb_endpoint_xfer_isoc(dep->desc)) {

doesn't look quite right. The TRB itself has a MISSED_ISOC status. Can't
you use that instead ?

OK.. Seems better..


+					dev_dbg(dwc->dev, "incomplete IN \
+							transfer %s\n",
+							dep->name);
+					dep->flags |= DWC3_EP_MISSED_ISOC;
+					dep->event = event;
+				} else {
+					dev_err(dwc->dev, "incomplete IN \
+							transfer %s\n",
+							dep->name);
+					status = -ECONNRESET;
+				}
  			}
  		} else {
  			if (count&&  (event->status&  DEPEVT_STATUS_SHORT))
@@ -1635,6 +1659,8 @@ static int dwc3_cleanup_done_reqs(struct dwc3 *dwc, struct dwc3_ep *dep,
  		dwc3_gadget_giveback(dep, req, status);
  		if (s_pkt)
  			break;
+		if (m_isoc)

this is never assigned to anything other than zero. What gives ??


oops..I need to set this when it is a missed isoc, so that loops breaks correctly.

Regards
Pratyush

--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html


B and H Foto and Electronics Corp.

[Linux Media]     [Video for Linux]     [Linux Input]     [Linux Audio Users]     [Photo]     [Yosemite News]    [Yosemite Photos]    [Free Online Dating]     [Linux Kernel]     [Linux SCSI]     [Old Linux USB Devel Archive]     [More Archives]

Add to Google Powered by Linux