Re: port-i386/57662: StarTech ICUSB23208FD 8-Port USB-to-Serial Adapter Hub fails on Alix with NetBSD/i386 9.3

[Alexander Schreiber <als%thangorodrim.ch@localhost>]

On Sat, Oct 28, 2023 at 09:29:09PM +0000, Taylor R Campbell wrote:
> > Date: Sat, 28 Oct 2023 22:26:06 +0200
> > From: Alexander Schreiber <als%thangorodrim.ch@localhost>
> > 
> > Both patches applied and things do look better. Now, instead of hanging
> > forever, after the second (so, at the third) connection attempt, cu
> > errors out with: "cu: link down" after a few seconds. This behavior then
> > sticks.
> > 
> > Relevant snippet from the trace:
> > 
> >   1699   1699 cu       0.000010616 CALL  open(0xbfb9a56e,6,0)
> >   1699   1699 cu       0.000011733 NAMI  "/dev/ttyU1"
> >   1699   1699 cu       11.991814027 RET   open -1 errno 5 Input/output error
> > 
> > vmstat usbhist dump at:
> > 
> > https://www.thangorodrim.ch/tmp/NetBSD10BETA_serial_vmstat-usbhist_patch-1
> 
> Great, thanks!  So we've fixed the logic to abort on timeout that I
> broke last year (which is good!), but now we have to figure out why
> the transfer is timing out at all.

Small update: I rebuilt the kernel (stripped out a lot of unneeded stuff
from the config to get build time down and enable DIAGNOSTICS) and redid
the test. Now, while cu hangs, I see matching complaints in dmesg:

[ 394.1174957] ehci_sync_hc: timed out
[ 400.1176959] ehci_sync_hc: timed out
[ 406.1178962] ehci_sync_hc: timed out

and then cu comes back with "cu: link down". I didn't see those kernel
messages before, so I guess that is courtesy of DIAGNOSTICS.

Kind regards,
           Alex.
-- 
"Opportunity is missed by most people because it is dressed in overalls and
 looks like work."                                      -- Thomas A. Edison