Disapearing pl2303 usb serial adapter on rpi2

[bob prohaska]

Hi Hans,

On Thu, Dec 21, 2017 at 05:29:05PM +0100, Hans Petter Selasky wrote:
> On 12/21/17 17:11, bob prohaska wrote:
> > On Wed, Dec 20, 2017 at 09:13:12PM +0200, Jukka A. Ukkonen wrote:
> >> On 12/20/17 19:02, bob prohaska wrote:
> >>> On an RPI2 at r326951 a pl2303 usb-serial adapter seems to be locking up
> >>> after a few hours. In the past this could be rectified by using usbconfig
> >>> to power cycle it, but now even that doesn't work. Apart from the serial
> >>> device file vanishing from /dev there seems to be no other ill effect.
> >>> Rebooting fixes the problem for minutes to hours.
> >>>
> >>> There is no explict error report in dmesg, but when usbconfig tries to power
> >>> cycle the adapter the console reports:
> >>>
> >>> uhub_explore_handle_re_enumerate: Could not unconfigure device (ignored)
> >>> usbd_req_re_enumerate: addr=9, set address failed! (USB_ERR_IOERROR, ignored)
> >>> usbd_setup_device_desc: getting device descriptor at addr 9 failed, USB_ERR_IOERROR
> >>> usbd_req_re_enumerate: addr=9, set address failed! (USB_ERR_STALLED, ignored)
> >>> usbd_setup_device_desc: getting device descriptor at addr 9 failed, USB_ERR_STALLED
> >>>
> >>>
> >>> A second RPI2 running r326747 is subject to the same lockup, but the adapter
> >>> can be unstuck using usbconfig to power cycle it.
> 
> Hi,
> 
> You might want to run "usbdump" with the appropriate arguments to 
> capture all USB traffic during such a cycle.
> 
I'll try it next time the pl2303 locks up.

> Further, you might want to only enable USB HUB debug messages to figure 
> out why the USB stack thinks the device is going away:
> 
> sysctl hw.usb.uhub.debug=17
> 
I've turned that on, and the console is spitting out a lot of output.
It'll make quite a haystack in which to find a needle 8-) Any hint
what to look for?

>  From what I know, the RPI series of cards have a small external USB HUB 
> on the board. All USB traffic goes through this High-Speed HUB. When you 
> are using LOW and FULL speed, all traffic is translated from High-Speed 
> USB into LOW and FULL speed USB.
>

Would changing the baudrate have any effect? A console certainly needn't
run at 115200 baud. For present purposes 9600 or less is plenty. 
> I don't rule out some kind of "bug" in this external piece of hardware.
>
Many have said pl2303 devices are unreliable, but they seem well-behaved
on a Mac and on an RPI3 running Raspbian. Whatever is wrong, it seems not
deterministic, but cumulative, with a touch of random. 

> You might also try to set (my wild guess):
> sysctl hw.usb.no_cs_fail=1
>
Not sure I understand what this does; is there a description somewhere? 

> Like already mentioned in this thread, the micro USB power input 
> connecter on the RPI is not supposed to handle more than 500mA, 
> regardless of how strong your power supply is. Then you need to subtract 
> the mA's used by the CPU and other chips like ethernet. I wouldn't be 
> surprised if ethernet functionality alone required 150mA at +5V.
> 
Some time ago I put a voltmeter on the Pi and the voltage stayed over
five volts, but of course short spikes weren't visible in such a test.
On average, however, the connector seemed to handle the load.

> USB LOW/HIGH/FULL speed all use some kind of pull-UP or pull-DOWN on the 

> data lines. This resistors are typically wired to +5V or GND (0V) on the 
> board and transients in the voltage may be seen as a disconnect by the 
> host or device.
> 
> Regarding the DTB's. Make sure the USB data pins are correctly 
> configured with regards to pullup/pulldown and not used as GPIOs.
> 
Wouldn't that sort of error lead to things not working at all? The
problems observed develop after hours or sometimes days.

Many thanks for your guidance!

bob prohaska