USB 2.0 Bug

[Matt Patterson]

So this is one that I have been fighting for a while, I was hoping 
someone would have a complete solution. I have a dual athlon 1700 
(Thunder K7), with one of the Via USB 2.0 pci cards. I am almost certain 
I have no problems when using a single processor kernel, but when using 
an smp kernel, 686 smp or k7 smp I experience the problem of a complete 
kernel freeze. When I use the 2.6.8 kernels currently for warty I get a 
freeze almost the instant I start any major transfer to an external hard 
disk over usb 2.0.

I believe that was related to the inotify feature. So I tried creating 
my own vanilla 2.6.10 kernel, which appeared to solve the problem. Then 
I discovered the backports page and have been using the 2.6.10 backport 
from hoary (k7 smp). That appeared to solve the problem (no inotify 
pretty sure), but yesterday I was transferring several hundred gigs 
around and when I woke up this morning, frozen machine.

So I booted into a single processor kernel (majorly slow...) and 
completed the transfer (by the way, it failed on a transfers of tens of 
thousands of little files). Then I went looking for causes, and I found 
this in the kernel log and simliar messages in the syslog:

Mar 29 21:29:39 localhost kernel: scsi1: PCI error Interrupt at seqaddr 
= 0x8
Mar 29 21:29:39 localhost kernel: scsi1: Data Parity Error Detected 
during address or write data phase
Mar 29 21:29:39 localhost kernel: scsi0: PCI error Interrupt at seqaddr 
= 0x8
Mar 29 21:29:39 localhost kernel: scsi0: Data Parity Error Detected 
during address or write data phase
Mar 29 21:31:14 localhost kernel: scsi1: PCI error Interrupt at seqaddr 
= 0x9
Mar 29 21:31:14 localhost kernel: scsi1: Data Parity Error Detected 
during address or write data phase
Mar 29 21:31:14 localhost kernel: scsi0: PCI error Interrupt at seqaddr 
= 0x9
Mar 29 21:31:14 localhost kernel: scsi1: WARNING WARNING WARNING WARNING
Mar 29 21:31:14 localhost kernel: scsi1: Too many PCI parity errors 
observed as a target.
Mar 29 21:31:14 localhost kernel: scsi1: Some device on this bus is 
generating bad parity.
Mar 29 21:31:14 localhost kernel: scsi1: This is an error *observed by*, 
not *generated by*, this controller.
Mar 29 21:31:14 localhost kernel: scsi1: PCI parity error checking has 
been disabled.
Mar 29 21:31:14 localhost kernel: scsi1: WARNING WARNING WARNING WARNING
Mar 29 21:31:14 localhost kernel: scsi0: Data Parity Error Detected 
during address or write data phase
Mar 29 21:31:14 localhost kernel: scsi0: WARNING WARNING WARNING WARNING
Mar 29 21:31:14 localhost kernel: scsi0: Too many PCI parity errors 
observed as a target.
Mar 29 21:31:14 localhost kernel: scsi0: Some device on this bus is 
generating bad parity.
Mar 29 21:31:14 localhost kernel: scsi0: This is an error *observed by*, 
not *generated by*, this controller.
Mar 29 21:31:14 localhost kernel: scsi0: PCI parity error checking has 
been disabled.
Mar 29 21:31:14 localhost kernel: scsi0: WARNING WARNING WARNING WARNING
Mar 30 08:13:45 localhost kernel: klogd 1.4.1#14ubuntu4, log source = 
/proc/kmsg started.
Mar 30 08:13:46 localhost kernel: Inspecting /boot/System.map-2.6.8.1-5-386

Anyone got any ideas besides get rid of the usb card, which I think I 
will do real soon.

Thanks,
Matt