sata_sil + UDMA CRC errors

[Alexander Shaposhnikov <shaposh@isp.nsc.ru>]

Good time of the day.
I think this might be a bug in sata_sil driver.

I have ASUS K8N-DL board with onborad sil3114 controller. As support for
nVidia Nforce 2200 SATA is broken in current sata_nv driver, i have to
use sil3114. Kernel is 2.6.13-rc6 + "sata_sil Mod15Write
quirk workaround" patch of TeJun Heo. OS is Fedora4 x86_64

There are two SATA drives, Seagate 7200.7 160Gb and Samsung SP0812C
80Gb. 

Here is the problem: 
Any intensive read/write activity results in alot of UDMA CRC errors for both drives.
I am getting :
ata1: status=0x51 { DriveReady SeekComplete Error }
ata1: error=0x04 { DriveStatusError }
kernel messages. With the patch applied, error rate is significantly low
for Seagate drive; Samsung is still virtually unusable. 
I checked SMART attributes, and the count of  UDMA CRC errors is ~200
for Seagate and more than 1000 for Samsung. Every time i get new
"status=0x51" errors for any drive, number of UDMA CRC errors grow
accordingly. 

But there seems to be no problems under Windows. Even highest read/write
activity do not generate any new CRC errors. 

Sorry for bad english.

Best Regards,
Alexander Shaposhnikov