Kevin Shanahan wrote:

I have done about 15 re-compiles now with increasing numbers of printk's added and have got this far.

scsi -> libata to ata_queued_cmd

Buffer I/O error on device sdb2, logical block 7281
ata2: command 0x25 timeout, stat 0x59 host_stat 0x1
scsi1: ERROR on channel 0, id 0, lun 0, CDB: 0x28 00 00 03 29 f0 00 00 ec 00 
Current sdb: sense = 70  3
ASC=11 ASCQ= 4
Raw sense data:0x70 0x00 0x03 0x00 0x00 0x00 0x00 0x06 0x00 0x00 0x00 0x00 0x11 0x04 
end_request: I/O error, dev sdb, sector 207344

Here the libata layer passes the error back to scsi and everyone is happy. (uses ata_scsi_qc_complete)

Then always ends with the following:

About here, scsi passes a command down to libata, ata_queued_cmd which causes the next 3 messages.
libata never errors out here and completes the request so it all falls in a heap.
I'm going out to dinner, I'll see if I can trace it further later tonight.

ATA: abnormal status 0x59 on port 0xB807
ATA: abnormal status 0x59 on port 0xB807
ATA: abnormal status 0x59 on port 0xB807

Not much help, but closing in.

Regards,
Brad