[Nut-upsuser] MGE Nova 1100 using newhidups on netBSD 3.0

[Arnaud Quette]

Hi Nevil,

2006/3/6, Nevil Thatcher <NevilTh at nat.com.au>:
> Hi Arnaud,
>
> I have been playing around with the blocking at 'waiting for Notifications'
>
> What I have found is that the process will block at the read() statement in
> usb_interrupt_read() function in libusb (bsd.c).
>
> It seems to sit here indefinitely (have left it overnight).
> By unplugging/replugging the USB cable causes an error return from the
> read() statement, thus the function returns and status's are then received
> via the usb_control_msgs's. Subsequent calls's to usb_interrupt_read then
> also fail with the message "Failedto read form interrupt endpoint"
>
> In the call to usb_interrupt_read there is a timeout argument (5 sec). I am
> not sure if this means that the read is supposed to timeout after 5 sec if
> no data but this is not happenening.

yes, that should be its behaviour. So there is something wrong with
the timeout management. When adding the O_NONBLOCK flag, does it
return approx 5 sec. later? If so that would mean that O_NONBLOCK is
needed.

> Waiting for notifications...
> usb_interrupt_read 4 129 0x7f7fffffeb10 20 5000
> usb_interrupt_read Calling read
> usb_interrupt_read result=-1, size=20, retrieved=0
> USB error: error reading from interrupt endpoint /dev/ugen1.01: Resource
> temporarily unavailable

there is however an issue.
You should report both to the netBSD kernel hackers (the ugen one(s)
in fact) and the libusb team if not already done.

Arnaud