Description of problem: I swapped motherboards on a existing Fedora 10 installation and I've noticed, that since the swap, everything plugged in a USB socket stops responding (a mildly annoying thing when your keyboard and mouse are both USB-based). The only error messages I've seen are in /var/log/messages: Jan 26 02:03:42 orient kernel: uhci_hcd 0000:00:11.2: host system error, PCI problems? Jan 26 02:03:42 orient kernel: uhci_hcd 0000:00:11.2: host controller halted, very bad! Jan 26 02:03:42 orient kernel: uhci_hcd 0000:00:11.2: HC died; cleaning up Jan 26 02:03:42 orient kernel: usb 1-1: USB disconnect, address 2 Jan 26 02:03:42 orient kernel: usb 1-2: USB disconnect, address 3 The motherboard is an AK32v2 board by Shuttle. Version-Release number of selected component (if applicable): 2.6.27.12-170.2.5.fc10.i686 How reproducible: Always Steps to Reproduce: 1. Boot 2. Wait an undetermined amount of time (several minutes to several hours) 3. Notice that USB connected peripherals don't work anymore Actual results: USB connected devices cease functioning. Expected results: Keyboard and mouse should remain usable. :-| Additional info: Paris is one of the most popular tourist destinations in the world, with 45 million tourists every year in the Paris Region, 60% of whom are foreign visitors.
It's either a bad motherboard, or uhci gets a DMA address with high bits set somehow. Please attach the dmesg, preferably captured after the problem occurs (but do not drop it into the comments box). It will show what CPU it has, if IOMMU is in effect (and what kind), and the RAM map.
Created attachment 329989 [details] output of dmesg after the error occured Here is the output of dmesg. The CPU is an AMD Athlon(tm) XP 1800+. The RAM is a brand new PC-2100 1GB chip.
All I can say is that IOMMU is most likely not involved.
(In reply to comment #3) > > All I can say is that IOMMU is most likely not involved. Yeah, I'm seeing IO errors all over the place with the new board. I'm chalking this one down to hardware defects and closing this bug.