[Nut-upsuser] 2.6.31.4: Intel P55 Chipset BUG [usbhid-raw/devices/broken?] [tested 3 different UPS']

[Alan Stern]

On Sat, 31 Oct 2009, Justin Piszcz wrote:

> >> SUMMARY: P55 Chipset/USB has issues with UPS' (broken pipe on 3 diff UPS')
> >> WORKAROUND: Buy/use PCI USB card if you want to use nut/connect a UPS.
> >> NOTES: All other USB (non-UPS) devices seem to function without any
> >>         problems.
> >>
> >> On Thu, 29 Oct 2009, Justin Piszcz wrote:
> >>
> >>> Problem: When I migrated from a DG965WH -> DP55KG motherboard, there were
> >>> quite a bit of issues, host still does not reboot without special flags, e.g.
> >>> reboot=a, the NIC driver is broken with the in-kernel version (the one on
> >>> e1000.sourceforge) fixes that and finally my UPS USB has driver stale
> >>> problems, with two issues already relating to the HW/drivers themselves it
> >>> would not be unrealistic to speculate that there could be an issue with the
> >>> USB chipset/driver with this P55 motherboard.
> >>
> >> --
> >>
> >> Tried another APC 1500VA UPS, same problem occurs, different cable/port etc.
> >>
> >> Problem appears to be the Intel P55KG / USB Chipset in the way that it
> >> handles usbhid-raw devices as CD/USB sticks/kbd/mouse/etc do not appear to
> >> be affected.
> >>
> >> So far all other USB devices work fine, with the exception of UPS', I
> >> have tested two APC 1500VA (one an RS the other an XS), in addition I also
> >> have a 'BELKIN 1200VA UPS with AVR' from a long time ago, the same
> >> problem occurs.
> >
> > ...
> >
> >> A bit frustrated, I popped in a PCI card with 4 external USBs, and now no
> >> more errors:
> >
> > We might learn something by comparing two usbmon logs for the same UPS
> > device, one plugged into a USB port on the motherboard and the other
> > plugged into an add-on card.
> >
> > Alan Stern
> >
> 
> Hi Alan,
> 
> Got it--two runs:
> 
> 1. 5 minutes of usbmon output on via chipset [working]
> 2. 5 minutes of usbmon output on intel p55 usb chipset [ehci/not working]

To summarize the differences between the logs, the Intel controller 
shows occasional instances of failed transfers like this:

ffff8802138c8500 3487005687 S Ci:2:004:0 s 81 06 2100 0000 0009 9 <
ffff8802138c8500 3487006051 C Ci:2:004:0 -32 0

while the corresponding transfers worked with the VIA controller:

ffff88020c46dd40 1774844264 S Ci:4:002:0 s 81 06 2100 0000 0009 9 <
ffff88020c46dd40 1774855248 C Ci:4:002:0 0 9 = 09211001 21012237 04

and indeed they worked in the other Intel logs (the failures appeared
to be more or less at random).

This does indeed look like a low-level hardware problem, but I'm
hesitant to blame it on the chipset.  For example, the problem might
lie in the hub you've got between the UPS and the computer.  I know,
this doesn't explain why everything works okay with the VIA controller.
Probably the only way to tell for sure what's really happening is by 
using an expensive bus analyzer.

Have you tried bypassing that hub, and plugging the UPS directly into 
the computer?

Alan Stern