Some USB memory sticks not (fully) recognized

[Christian Eggers]

Hi Sascha,

On Monday, 10 August 2020, 21:26:43 CEST, Sascha Hauer wrote:
> Hi Christian,
> 
> On Mon, Aug 10, 2020 at 01:52:01PM +0200, Christian Eggers wrote:
> > On Thursday, 6 August 2020, 16:29:13 CEST, Christian Eggers wrote:
> > > I got a bug report that some newer USB memory sticks do not work with
> > > barebox. While with barebox-2020.01 these devices are not recognized at
> > > all, in barebox-2020.07 I get at least some warnings:
> > I checked again with an older release of barebox (2019-06). With this
> > version, the USB drive is detected:
> > [...]
> > Also mounting the drive works fine. So I guess that this problem may have
> > been introduced here:
> > 
> > b1d9837182 ("usb: Change power-on / scanning timeout handling")
> 
> Before guessing further could you verify that exactly this commit breaks
> your setup? There are a few more changes in the area that could be the
> culprit.
sorry, that was nonprofessional. In the meantime, I got some new findings:

- v2019.06 is the last release were the device is detected

- reverting b1d9837182 ("usb: Change power-on / scanning timeout handling")
doesn't change anything (so it's not related to this problem)

- the problem was definitely introduced in
6044d6c08e ("usb: host: ehci: Use to USBSTS to wait for transfer completion")

Debugging with and without 6044d6c08e showed that waiting for USBSTS:INT
is not sufficient as QT_TOKEN_STATUS_ACTIVE is still set after this. Turning
the dev_dbg() into an dev_err() clearly shows this:

> usb: USB: scanning bus for devices...
> usb1: Bus 001 Device 001: ID 0000:0000 EHCI Host Controller
> usb1-0: Bus 001 Device 002: ID 0424:4916 USB4916
> ERROR: imx-usb 2184200.usb at 2184200.of: dev=3, usbsts=0x40081, p[1]=0x18001205, p[2]=0x0
> ERROR: imx-usb 2184200.usb at 2184200.of: dev=3, usbsts=0x40081, p[1]=0x18001205, p[2]=0x0
> ERROR: usb1-0-1: unable to get descriptor, error 80000000
> ERROR: imx-usb 2184200.usb at 2184200.of: dev=3, usbsts=0x40081, p[1]=0x18001205, p[2]=0x0
> ERROR: imx-usb 2184200.usb at 2184200.of: dev=3, usbsts=0x40081, p[1]=0x18001205, p[2]=0x0
> ERROR: imx-usb 2184200.usb at 2184200.of: dev=3, usbsts=0x40081, p[1]=0x18001205, p[2]=0x0
> ERROR: imx-usb 2184200.usb at 2184200.of: dev=3, usbsts=0x40081, p[1]=0x18001205, p[2]=0x0
> ERROR: imx-usb 2184200.usb at 2184200.of: dev=3, usbsts=0x40081, p[1]=0x18001205, p[2]=0x0
> ERROR: imx-usb 2184200.usb at 2184200.of: dev=3, usbsts=0x40081, p[1]=0x18001205, p[2]=0x0
> ERROR: imx-usb 2184200.usb at 2184200.of: dev=3, usbsts=0x40081, p[1]=0x18001205, p[2]=0x0
> ERROR: imx-usb 2184200.usb at 2184200.of: dev=3, usbsts=0x40081, p[1]=0x18001205, p[2]=0x0
> ERROR: imx-usb 2184200.usb at 2184200.of: dev=3, usbsts=0x40081, p[1]=0x18001205, p[2]=0x0
> ERROR: imx-usb 2184200.usb at 2184200.of: dev=3, usbsts=0x40081, p[1]=0x18001205, p[2]=0x0
> ERROR: imx-usb 2184200.usb at 2184200.of: dev=3, usbsts=0x40081, p[1]=0x18001205, p[2]=0x0
> ERROR: imx-usb 2184200.usb at 2184200.of: dev=3, usbsts=0x40081, p[1]=0x18001205, p[2]=0x0
> usb1-0-1: Bus 001 Device 003: ID 090c:3267 
> usb1-0-6: Bus 001 Device 004: ID 0424:494a USB2 Controller Hub
> usb: 4 USB Device(s) found

It seems that I can take multiple USBINTs until QT_TOKEN_STATUS_ACTIVE is
cleared. The following snippet works fine for me:

	volatile struct qTD *vtd;
	...
	vtd = td;
	do {
		ret = handshake(&ehci->hcor->or_usbsts, STS_USBINT, STS_USBINT,
				timeout_ms * 1000);
		if (ret < 0) {
			dev_err(ehci->dev, "handshake failed: %d\n", ret);
			ehci_enable_async_schedule(ehci, false);
			ehci_writel(&qh->qt_token, 0);
			return -ETIMEDOUT;
		}
		token = hc32_to_cpu(vtd->qt_token);
	} while (token & QT_TOKEN_STATUS_ACTIVE);

There is probably no benefit left compared to the version prior 6044d6c08e, so
if there is not better way to do this, I would propose to revert 6044d6c08e.

regards
Christian