Re: Recent USB mouse regression

On Sat, 8 Nov 2008 03:53:17 +0100, Alexey Shuvaev <shuvaev_at_physik.uni-wuerzburg.de> wrote:
> On Sat, Nov 08, 2008 at 04:24:22AM +0200, Giorgos Keramidas wrote:
>> Building a kernel & userland just before USB2 from svn change 184609
>> seems to have fixed this for now.  It seems that even when USB2 is not
>> loaded it affects a bit the way ums(4) works.
>>
>> What is the best way of troubleshooting this?
>
> I have seen reversed situation when even with usb2_input_ms kld loaded
> I am finding ums + usb (old stack) in kldstat after the system goes
> multiuser.  I would check kldstat after each event (today klds are
> loaded automatically!) and try operate in single user mode first.

> Could it be that something (moused) is trying to load all klds
> matching some symbols (provided interfaces)?

My kernel config file is based on GENERIC, so `ums' is not a module and
I don't load any usb2_xxx modules in my `loader.conf' yet.

Should I rebuild it as a module?  I can try that later...

I just started a buildworld with all the changes up the HEAD except for
the five commits that are related to usb2, i.e. I have everything else
up to svn /head_at_184743 but I have backed out locally the changes:

  ------------------------------------------------------------------------
  r184610 | alfred | 2008-11-04 04:31:03 +0200 (Tue, 04 Nov 2008) | 123 lines
  r184612 | alfred | 2008-11-04 05:42:01 +0200 (Tue, 04 Nov 2008) | 2 lines
  r184658 | rink   | 2008-11-04 21:28:19 +0200 (Tue, 04 Nov 2008) | 4 lines
  r184659 | rink   | 2008-11-04 22:57:39 +0200 (Tue, 04 Nov 2008) | 7 lines
  r184736 | imp    | 2008-11-06 19:26:12 +0200 (Thu, 06 Nov 2008) | 4 lines
  ------------------------------------------------------------------------

These are the commits that I could see affecting sys/usb2 so far, so it
may help find out that it is indeed usb2 that causes ums(4) to act oddly.

If this fails, I'll remerge the usb2 changes, and try building ums as a
module-only.