Charles, Thanks for your reply.
1. We have tested this with several Patsburgs, Powerware 5115s, and cables, so I don't think that it is a cable problem. 2. We have found that with Fedora 17 (Linux 3.4.3-1 kernel), that USBDEVFS_CONTROL failed messages mentioned below do not appear and bcmxcp_usb does not see timeouts. However, we do see the problems with Ubuntu 12.04 (Linux 3.2 kernel), so apparently the problem is either with the kernel or the USB driver and it was fixed somewhere between Linux 3.2 and 3.4.3. We are currently trying to identify a patch to Linux 2.6.39 to apply this fix. Rich Wrenn From: Charles Lepple [mailto:[email protected]] Sent: Saturday, June 30, 2012 8:41 PM To: Rich Wrenn Cc: [email protected] Subject: Re: [Nut-upsuser] Problems with Powerware 5115 on Patsburge USB On Sunday, June 24, 2012, Rich Wrenn wrote: I am seeing some problems when using the Powerware 5115 UPS when connected to the Patsburg USB controller and would like to know if anyone has a solution. The symptom is that PING messages sent to the bcmxcp_usb driver take as long as 25 seconds to complete. The configuration is a Romley patform (a.k.a., Sandy Bridge + Patsburg) running Debian 6 with 2.6.39 kernel with NUT 2.6.4 in which the Powerware is connected via USB 2.0 to the Patsburg. Of note is that the following messages are logged in syslog frequently. 2012-06-21T09:03:10.754413+00:00 (none) kernel: [ 8659.067243] usb 2-1.4: usbfs: USBDEVFS_CONTROL failed cmd bcmxcp_usb rqt 66 rq 13 len 4 ret -110 I believe that these indicate that the USB driver is timing out USB messages to the UPS. That sounds about right. Unfortunately, the root cause could conceivably be anything from bad silicon to dropped interrupts to a flaky cable or jack. The identical setup running on a Westmere + Tylersburg + ICH10R works fine but on that platform the USB connection is USB 1.1. When you say that it is USB 1.1 here, is the other system connecting at a higher speed? Even with USB 2.0 controllers, most HID products (keyboard, mouse, UPS) will generally only connect at 1.5 Mbit/sec, since the device-side circuitry for that is much simpler. The kernel messages from when the UPS is plugged in should mention the speed. I know it sounds like snake oil, but try swapping the USB cable for a known good one. If not, it might be a kernel driver issue. Sorry it's a vague answer, but a more detailed analysis would probably need to be done by the engineers who built the UPS. -- - Charles Lepple
_______________________________________________ Nut-upsuser mailing list [email protected] http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/nut-upsuser
