: omap-udc stressed full duplex usb communication.

Hello

I'm working on an Omap target:

CPU: ARM926EJ-S [41069263] revision 3 (ARMv5TEJ), cr=00053177
Machine: TI OMAP1710 HW20acq board
Memory policy: ECC disabled, Data cache writeback OMAP1710 revision 8 handled as 16xx id: 8b5f702f03330200

I use the usb as cdc_acm and the communication in the both ways is very stressed.
Some time the device receives a strange IRQ from EP0, which stop all data reception from USB.
The usb registers during this IRQ are:
IRQ_SRC : 0xB0
EPN_STAT : 0x01
EP_NUM : 0x11
STAT_FLG : 0x00

The TI inc. documentation explains that the IRQ_SRC is TX and RX in the same time. (the irq callback can manage only one). But EPN_STAT gives only the EP number of the TX request.
In the same documentation we can read p15-67:
11:8 EPn_RX_IT_SRC The receive endpoint interrupt source (non-ISO) bit
only concerns non-ISO enpoints...
000: No receive endpoint interrupt is pending

I send you my patch. I hope that it will be usefull.

index f81e4f0..863f692 100644
@@ -1948,7 +1948,7 @@ static irqreturn_t omap_udc_pio_irq(int irq, void

*_dev)

        u16             epn_stat, irq_src;
        irqreturn_t     status = IRQ_NONE;
        struct omap_ep  *ep;
-       int             epnum;
+       int             epnum, epnumrx, epnumtx;
        struct omap_udc *udc = _dev;
        struct omap_req *req;
        unsigned long   flags;
@@ -1957,9 +1957,11 @@ static irqreturn_t omap_udc_pio_irq(int irq, void
*_dev)
        epn_stat = omap_readw(UDC_EPN_STAT);
        irq_src = omap_readw(UDC_IRQ_SRC);

+ epnumrx = (epn_stat >> 8) & 0x0f;

+       epnumtx = epn_stat & 0x0f;
        /* handle OUT first, to avoid some wasteful NAKs */
-       if (irq_src & UDC_EPN_RX) {
-               epnum = (epn_stat >> 8) & 0x0f;
+       if ((irq_src & UDC_EPN_RX) && epnumrx) {
+               epnum = epnumrx;
                omap_writew(UDC_EPN_RX, UDC_IRQ_SRC);
                status = IRQ_HANDLED;
                ep = &udc->ep[epnum];
@@ -1994,8 +1996,8 @@ static irqreturn_t omap_udc_pio_irq(int irq, void
*_dev)
        }

        /* then IN transfers */
-       else if (irq_src & UDC_EPN_TX) {
-               epnum = epn_stat & 0x0f;
+       else if ((irq_src & UDC_EPN_TX) && epnumtx){
+               epnum = epnumtx;
                omap_writew(UDC_EPN_TX, UDC_IRQ_SRC);
                status = IRQ_HANDLED;
                ep = &udc->ep[16 + epnum];

This message, including attachments, is intended solely for the addressee indicated in this message and is strictly confidential or otherwise privileged. If you are not the intended recipient (or responsible for delivery of the message to such person) : - (1) please immediately (i) notify the sender by reply email and (ii) delete this message and attachments, - (2) any use, copy or dissemination of this transmission is strictly prohibited. If you or your employer does not consent to Internet email messages of this kind, please advise Myriad Group AG by reply e-mail immediately. Opinions, conclusions and other information expressed in this message are not given or endorsed by Myriad Group AG unless otherwise indicated by an authorized representative independent of this message.

Hello again,

This is another patch for the same file. I send two patches because the first one is really a bug correction for all target.
I tested this one with only mine and I'm not sure that there is not trouble with other targets.

The IRQ callback (omap_udc_pio_irq) starts a timer (ep->timer, PIO_OUT_TIMEOUT) which starts again itself. Then this callback is called repeatly and uses ressources all the time. When I read the code, I understand that this timer callback (pio_out_timer) does the same work that the IRQ callback, if the treatement is not complete.

On my target we have not enought ressources to manage the real IRQ and this timer. Then I start it only if the treatement is not done.

On my target the kernel receives all IRQ then the timer is useless. Perhaps some targets have some troubles to receive all IRQ????

index f81e4f0..9da4d98 100644
@@ -1919,6 +1919,7 @@ static void pio_out_timer(unsigned long _ep)
        struct omap_ep  *ep = (void *) _ep;
        unsigned long   flags;
        u16             stat_flg;
+       int             stat = -1;
 
        spin_lock_irqsave(&ep->udc->lock, flags);
        if (!list_empty(&ep->queue) && ep->ackwait) {

@@ -1932,14 +1933,14 @@ static void pio_out_timer(unsigned long _ep)
                        VDBG("%s: lose, %04x\n", ep->ep.name, stat_flg);
                        req = container_of(ep->queue.next,
                                        struct omap_req, queue);
-                       (void) read_fifo(ep, req);
+                       stat = read_fifo(ep, req);
                        omap_writew(ep->bEndpointAddress, UDC_EP_NUM);
                        omap_writew(UDC_SET_FIFO_EN, UDC_CTRL);
                        ep->ackwait = 1 + ep->double_buf;
                } else
                        deselect_ep();
        }
-       mod_timer(&ep->timer, PIO_OUT_TIMEOUT);
+       if (stat == -1) mod_timer(&ep->timer, PIO_OUT_TIMEOUT);
        spin_unlock_irqrestore(&ep->udc->lock, flags);
 }
 

@@ -1948,7 +1949,7 @@ static irqreturn_t omap_udc_pio_irq(int irq, void

*_dev)

        u16             epn_stat, irq_src;
        irqreturn_t     status = IRQ_NONE;
        struct omap_ep  *ep;
-       int             epnum;
+       int             epnum, epnumrx, epnumtx;
        struct omap_udc *udc = _dev;
        struct omap_req *req;
        unsigned long   flags;
@@ -1957,9 +1958,12 @@ static irqreturn_t omap_udc_pio_irq(int irq, void
*_dev)
        epn_stat = omap_readw(UDC_EPN_STAT);
        irq_src = omap_readw(UDC_IRQ_SRC);

+ epnumrx = (epn_stat >> 8) & 0x0f;

+       epnumtx = epn_stat & 0x0f;
        /* handle OUT first, to avoid some wasteful NAKs */

- if (irq_src & UDC_EPN_RX) {
- epnum = (epn_stat >> 8) & 0x0f;
+ if ((irq_src & UDC_EPN_RX) && epnumrx) {
+ int stat = -1;

+               epnum = epnumrx;
                omap_writew(UDC_EPN_RX, UDC_IRQ_SRC);
                status = IRQ_HANDLED;
                ep = &udc->ep[epnum];
@@ -1970,7 +1974,6 @@ static irqreturn_t omap_udc_pio_irq(int irq, void
*_dev)
                if (omap_readw(UDC_STAT_FLG) & UDC_ACK) {
                        ep->ackwait--;
                        if (!list_empty(&ep->queue)) {
-                               int stat;
                                req = container_of(ep->queue.next,
                                                struct omap_req, queue);
                                stat = read_fifo(ep, req);
@@ -1990,12 +1993,12 @@ static irqreturn_t omap_udc_pio_irq(int irq,
void *_dev)
                        omap_writew(UDC_SET_FIFO_EN, UDC_CTRL);
                        ep->ackwait = 1 + ep->double_buf;
                }
-               mod_timer(&ep->timer, PIO_OUT_TIMEOUT);
+               if (stat == -1) mod_timer(&ep->timer, PIO_OUT_TIMEOUT);
        }

        /* then IN transfers */
-       else if (irq_src & UDC_EPN_TX) {
-               epnum = epn_stat & 0x0f;
+       else if ((irq_src & UDC_EPN_TX) && epnumtx){
+               epnum = epnumtx;
                omap_writew(UDC_EPN_TX, UDC_IRQ_SRC);
                status = IRQ_HANDLED;
                ep = &udc->ep[16 + epnum];

This message, including attachments, is intended solely for the addressee indicated in this message and is strictly confidential or otherwise privileged. If you are not the intended recipient (or responsible for delivery of the message to such person) : - (1) please immediately (i) notify the sender by reply email and (ii) delete this message and attachments, - (2) any use, copy or dissemination of this transmission is strictly prohibited. If you or your employer does not consent to Internet email messages of this kind, please advise Myriad Group AG by reply e-mail immediately. Opinions, conclusions and other information expressed in this message are not given or endorsed by Myriad Group AG unless otherwise indicated by an authorized representative independent of this message.

Hi,

wow, 1710. That brings me back to good old times with the 1710 H3 board.

please read Documentation/CodingStyle and Documentation/SubmittingPatches and resend following those rules ?

and you should remove these kind of footers when communicating with a mailing list. It would prevent archiving the list.

balbi

DefectiveByDesign.org