Improve busprobe()

[tmk]

The library can miss plugin event and fails to enumerate recognize and initialize a device occasionally.

Seems like MAX3421e CONDETIRQ(Connect/Disconnect Interrupt) interrupt doesn't arise for all plugin events and we have to check bus state regularly.

I tried to catch INT pin change using ISR() with minimum code but it still failed to detect device insertion event occasionally. I believe that we cannot rely solely on the CONDETIRQ for plugin detection.

Meanwhile, I never seen the chip failed in detection of unpluging.

Branch for this topic:
https://github.com/tmk/USB_Host_Shield_2.0/tree/upstream_fix_busprobe

This fix seems like promising.

[tmk]

And these issues are related probably. felis#425 felis#452

[tmk]

The problem and the fix were confirmed with this sketch.

You will see it fails and stays in usb_state: WAIT_FOR_DEVICE(12) occasionally even when device is plugged.
This is happened likely when you plug-in and unplug a device quickly several times.

In case a device is recognized and goes to usb_state RUNNING(90).

usb_state: 12
usb_state: 20
usb_state: 40
usb_state: 50
usb_state: 51
usb_state: 90
.....

In case it fails

usb_state: 12
.....................

test sketch

test.ino:

#include <Usb.h>
#include <usbhub.h>



USB     Usb;

// Driver registry
//USBHub  hub1(&Usb);
//USBHub  hub2(&Usb);



void setup()
{
  Serial.begin( 115200 );
#if !defined(__MIPSEL__)
  while (!Serial);
#endif

  Serial.println("Start");

  Usb.Init();
}


static uint8_t prev_state = 0;
static uint32_t next_timer = 0;

void loop()
{
  Usb.Task();

  // USB state
  if (prev_state != Usb.getUsbTaskState()) {
    prev_state = Usb.getUsbTaskState();
    Serial.print(F("usb_state: ")); Serial.print(Usb.getUsbTaskState(), HEX); Serial.println();
  }

  if (next_timer < millis()) {
    Serial.print('.');
    next_timer = millis() + 1500;
  }
}
// vim: tabstop=2 softtabstop=2 shiftwidth=2 expandtab