Subject: Re: ucom read error restart
To: Darrin B. Jewell <dbj@NetBSD.org>
From: Charles M. Hannum <firstname.lastname@example.org>
Date: 07/29/2004 05:08:55
On Wednesday 28 July 2004 20:43, Darrin B. Jewell wrote:
> Here's the dmesg output after setting uhcidebug to 0t15 in ddb and
> connecting with tip, hitting return, getting a shell prompt and
> then disconnecting with "~." in tip. It gets 1-2 CRCTO|STALLED errors per
> second on its interrupt endpoint and it gets one on its bulk input
> endpoint after a few seconds of being idle. This is at 115200 baud,
> but the same thing does occur at 9600 baud as well.
So, there are only two conditions in which uhci is documented to return both
of these bits set in a bulk in or interrupt descriptor:
1) The device is simply not responding to the transaction. This corresponds
to the DEVICE_NOT_RESPONDING condition code in ohci.
2) We get an actual CRC error from the device.
(No, there is no way to distinguish the two cases with uhci.)
I'm not sure I believe that either of these cases is happening. I'm more
inclined to suspect that we're seeing some sort of NAK timeout condition
that's either not documented or is very badly documented.
A similar trace with ohci would probably be enlightening.