[ale] USB reset error in dmesg [SOLVED]

[Rich Faulkner]

Believe I've found the solution to this so thought I'd share it here...

I was able to narrow down a few factors in this:

1.  Issue dependent on v5 RHEL family (driver related)
2.  Issue effected by changing feature in BIOS (memory related)
3.  Issue centered around EHCI driver and visible as power loss in USB
optical mouse (power related)

This pointed me toward a BIOS feature for Intel Power Technology.
Disabling this feature allowed proper functionality of the device (in
this case the USB optical mouse).  Further digging into the feature
found that setting the Processor C State limit to [C0] provides
operational functionality w/o power-up/down of USB optical mouse.  No
further errors observed.

Waiting to hear results of field testing but I think this is it:  that
when low/high speed devices are connected through a USB 2.0 hub to the
EHCI hub with Energy Efficiency enabled; the EHCI starts split
transactions, but gets delayed while trying to access memory.  Because
of this, EHCI cannot complete the split transaction before the
transaction translator in the hub discards the data.  The hid-core
driver retires the transaction but it also fails and the USB device is
reset.  

It is notable that this behaviour is not seen using this BIOS and RHEL
6.x (CentOS or Scientific Linux).

Cheers!

RinL




On Sat, 2012-06-16 at 04:21 -0400, Rich Faulkner wrote:

> Attaching a PS/2 mouse and keyboard resolve the error but by only
> removing the device(s) that are causing the problem.  That is not a
> solution as all USB have to be working.  I have recommended that the
> OS be migrated to 6.x but don't know that this will be accepted.
> Another find was setting the BIOS to support multi-monitor mode (an
> interesting connection); the error is not induced.  Only in
> single-monitor mode does the USB mouse/keyboard go bugger.  So there
> is a BIOS connection here for resource allocation that interacts with
> EHCI driver (or USB driver) and the function of low-speed USB
> device(s).  
> 
> In the end, the mouse and keyboard are constantly powering-up and
> down.  Looks like they fail and reinitialize; then fail; then
> initialize.  Up, down, up down...sand-a-floor; side-to-side...you get
> my drift...
> 
> 
> On Fri, 2012-06-15 at 17:13 -0400, Jim Kinney wrote:
> 
> > Oh. good catch! I have not had the chance to look at "the piles"
> > here
> > 
> > So disconnecting the mouse will stop the failures?
> > 
> > On Fri, Jun 15, 2012 at 5:09 PM, Rich Faulkner <rfaulkner at tux86.org>
> > wrote:
> > 
> >         The plot thickens:  found the MS USB mouse is at issue...
> >         
> >         - The low speed USB device using ehci_hcd and address 3 is
> >         the Microsoft 3-Button Mouse as / class/ input/ input0 on
> >         usb-0000:00:1d.0-1.1
> >         
> >         - dmesg indicates:  PCI: cache line size of 32 is not
> >         supported by device 0000:00:1d.0
> >         
> >         - I can see the USB 2.0, and EHCI 1.00 driver started for
> >         this device as well
> >         
> >         Observation of the mouse shows power blinking-off
> >         periodically.  Keyboard (USB) acts like it has sticking keys
> >         periodically.  I suspect this will goof-up most anything
> >         attached to the USB bus.
> >         
> >         I have been trying some kernel parameters to grub.conf but
> >         no improvement thus far to the mouse.  (Only got rid of an
> >         issue concerning mtrr).  Another hint in this is that when
> >         you set the BIOS for multi-monitor support the problem goes
> >         away.  (Yes, I know we could set that and leave it but
> >         that's not the point.  Need to find the fix for this other
> >         than a workaround like 6.x or the BIOS setting).
> >         
> >         TIA.......Rich 
> >         
> >         
> >         
> >         
> >         > > Chipset is the Intel C206 (PCH).   Datasheet for the SHB
> >         > > is here:
> >         > > http://www.trentontechnology.com/downloads/datasheets/trenton_tsb7053_productdatasheet.pdf 
> >         
> >         
> >         
> >         
> >         > > >         I have come across a problem with RHEL 5.x
> >         > > >         family where I am getting the following output
> >         > > >         in dmesg:
> >         > > >         
> >         > > >         "usb 2-1.1: reset low speed USB device using
> >         > > >         ehci_hcd and address 3"
> >         > > >         
> >         > > >         My USB attached keyboard and mouse do not
> >         > > >         function properly (lag and inaccurate output).
> >         > > >         It looks like there is an issue in the system
> >         > > >         support of USB but it only seems to exist in
> >         > > >         5.x family.  Once tested in 6.x it all works.
> >         > > >         (Testing with both CentOS and Scientific Linux
> >         > > >         - live and installed samples)
> >         > > >         
> >         > > >         I am going through the debug output manually
> >         > > >         but thought I'd shoot this over here for a
> >         > > >         more experienced perspective from those with
> >         > > >         grayer beards than I...
> >         > > >         
> >         > > >         TIA........Rich 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.ale.org/pipermail/ale/attachments/20120618/422183d0/attachment.html