usb, davinci: usb 2.0 problem on an am1808 based board

[Heiko Schocher]

Hello,

a while ago (see discussion here:
http://comments.gmane.org/gmane.linux.usb.general/54505), I found this
"USB timing Bug" on an am1808 based board:

Heiko Schocher wrote:
> Hello Sergei,
[...]
> Sergei Shtylyov wrote:
>> Hello.
>>
>> On 11.11.2011 11:19, Felipe Balbi wrote:
>>
>>>> I try to bring up usb 2.0 support on an am1808 based
>>>> board and Linux version 3.1.0-rc10 and I am facing
>>>> some issues:
[...]
>>>>   But if I add in drivers/usb/musb/da8xx.c
>>>>    da8xx_musb_interrupt() a 10ms delay befor the
>>>>    "if (status&  (DA8XX_INTR_DRVVBUS<<  DA8XX_INTR_USB_SHIFT)) {"
>>>>    line usb works fine!
>>>>    It also works without this timeout if I change the
>>>>    "dev_dbg(musb->controller, "USB IRQ %08x\n", status);"
>>>>    to
>>>>    "printk(musb->controller, "USB IRQ %08x\n", status);"
>>>>
>>>>    ->  some timing issue, but I have no idea why!
>>>>       (Maybe you have an idea?)
>>> unfortunately I don't know any details about DaVinci.
>>    AM1808 is not exactly DaVinci, to be precise...
>>
>>> Sergei, are you able to help on this question ?
>>    Maybe. :-)
> 
> I hope it ;-)

now I digged deeper and found this patch, which solves the
problem (without adding a 10 ms delay):

-----------------------------
diff --git a/drivers/usb/musb/da8xx.c b/drivers/usb/musb/da8xx.c
index d32aa4d..6a6b17b 100644
--- a/drivers/usb/musb/da8xx.c
+++ b/drivers/usb/musb/da8xx.c
@@ -297,6 +297,7 @@ static irqreturn_t da8xx_musb_interrupt(int irq, void *hci)
        unsigned long           flags;
        irqreturn_t             ret = IRQ_NONE;
        u32                     status;
+       int                     flag = 1;

        spin_lock_irqsave(&musb->lock, flags);

@@ -368,6 +363,7 @@ static irqreturn_t da8xx_musb_interrupt(int irq, void *hci)
                        musb->xceiv->default_a = 0;
                        musb->xceiv->state = OTG_STATE_B_IDLE;
                        portstate(musb->port1_status &= ~USB_PORT_STAT_POWER);
+                       flag = 0;
                }

                dev_dbg(musb->controller, "VBUS %s (%s)%s, devctl %02x\n",
@@ -378,7 +374,7 @@ static irqreturn_t da8xx_musb_interrupt(int irq, void *hci)
                ret = IRQ_HANDLED;
        }

-       if (musb->int_tx || musb->int_rx || musb->int_usb)
+       if (flag && ((musb->int_tx || musb->int_rx || musb->int_usb)))
                ret |= musb_interrupt(musb);

  eoi:
-----------------------------

If we get an DA8XX_INTR_DRVVBUS IRQ, in the else case

musb->xceiv->state = OTG_STATE_B_IDLE;

is set, but overwritten in the musb_interrupt() call, which results
in not starting the "Poll for ID change" timer ... as I am not an USB
expert, posting this intermediate result here, maybe someone have an
idea/explanation if this patch is the way to go, or what would be a
better solution?

bye,
Heiko
-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de