Ubuntu
linux package

  1. Bug #647043
  2. Comment #23

Comment 23 for bug 647043

Revision history for this message
Bjorn Helgaas (bjorn-helgaas) wrote :

Using "pci=nocrs" is a manual workaround, but not sufficient to resolve this issue.

There is an upstream bug report for this issue on a Dell 1546: https://bugzilla.kernel.org/show_bug.cgi?id=31602

Leann reported in comment #8 that the Dell 1536 in the cert lab works. That machine appears to have 2GB of memory, and we move the USB devices just after that:

  BIOS-e820: 0000000000100000 - 000000007fe71800 (usable)
  Memory: 2041068k/2095556k available (5708k kernel code, 452k absent, 54036k reserved, 5382k data, 908k init)
  pci_root PNP0A03:00: host bridge window [mem 0x80000000-0xcfffffff]
  pci 0000:00:12.0: BAR 0: assigned [mem 0x80000000-0x80000fff]

The machines of other reporters (and the upstream report) appear to have 3GB or more of memory, so the host bridge window [mem 0x80000000-0xcfffffff] is not available, and we put the USB devices in a different place:

  BIOS-e820: 0000000000100000 - 00000000c7e60800 (usable)
  Memory: 3977176k/4980736k available (5085k kernel code, 83944k reserved, 2432k data, 704k init, 3148168k highmem)
  pci_root PNP0A03:00: host bridge window [mem 0xd4000000-0xfebfffff]
  pci 0000:00:12.0: BAR 0: assigned [mem 0xd4000000-0xd4000fff]

I suspect there's some other unreported device at 0xd4000000 that's causing the trouble.