Re: fxp problems

On 07-Aug-2004 Wesley Morgan wrote:
> I've got a box running 5.2.1-R p8 that has several times caught
> itself in 
> a fit of errors on both fxp devices, alternating between the two. A
> little 
> searching through PR's shows it reported back in 2003 and possibly
> again 
> in 2004, but always on -stable. Some posts to -stable seem to think
> it's 
> an interaction with the PCI and SCSI layers, but I'm not sure if this
> applies to -current. Anyone have any experience with these type of
> errors? 
> I've got the kernel conf and dmesg attached.
> 
> 
> Aug  6 00:00:14 tiamat kernel: fxp1: device timeout
> Aug  6 00:00:14 tiamat kernel: fxp1: SCB timeout: 0xff 0xff 0xff
> 0xffff
> Aug  6 00:00:14 tiamat kernel: fxp1: DMA timeout
> Aug  6 00:00:14 tiamat kernel: fxp1: SCB timeout: 0xff 0xff 0xff
> 0xffff
> Aug  6 00:00:14 tiamat kernel: fxp1: DMA timeout
> Aug  6 00:00:14 tiamat kernel: fxp1: SCB timeout: 0xff 0xff 0xff
> 0xffff
> Aug  6 00:00:14 tiamat kernel: fxp1: DMA timeout
> Aug  6 00:00:14 tiamat kernel: fxp1: SCB timeout: 0xff 0xff 0xff
> 0xffff
> Aug  6 00:00:14 tiamat kernel: fxp1: SCB timeout: 0xff 0xff 0xff
> 0xffff
> 
> Aug  6 00:00:29 tiamat kernel: fxp0: device timeout
> Aug  6 00:00:29 tiamat kernel: fxp0: SCB timeout: 0xff 0xff 0xff
> 0xffff
> Aug  6 00:00:29 tiamat kernel: fxp0: DMA timeout
> Aug  6 00:00:29 tiamat kernel: fxp0: SCB timeout: 0xff 0xff 0xff
> 0xffff
> Aug  6 00:00:29 tiamat kernel: fxp0: DMA timeout
> Aug  6 00:00:29 tiamat kernel: fxp0: SCB timeout: 0xff 0xff 0xff
> 0xffff
> Aug  6 00:00:29 tiamat kernel: fxp0: DMA timeout
> Aug  6 00:00:29 tiamat kernel: fxp0: SCB timeout: 0xff 0xff 0xff
> 0xffff
> Aug  6 00:00:29 tiamat kernel: fxp0: SCB timeout: 0xff 0xff 0xff
> 0xffff
> Aug  6 00:00:29 tiamat kernel: fxp1: SCB timeout: 0xff 0xff 0xff
> 0xffff

I recently tried running an SMP kernel here, even though my box is UP,
as a temporary workaround for the broken state of UP kernel builds on
amd64.  When I did, I experienced the fxp timeouts.  When I finally got
a UP kernel to build again, the fxp problem went away.

Perhaps there's a clue there.  :-)

-- 
Conrad J. Sabatier <conrads_at_cox.net> -- "In Unix veritas"