Re: [PATCH 1/1] libata-dev: ata_check_atapi_dma() fix for ATA_FLAG_PIO_POLLING LLDDs

[Albert Lee <albertcc@tw.ibm.com>]

Jeff Garzik wrote:

> 
> Applied, though I think its an open question what happens with
> CDB-interrupt ATAPI devices want to do DMA.  Perhaps we just get an
> interrupt event that we clear, then life proceeds as normal.

Not sure. I have no such device at hand for test; currently relying
on the SFF-8020i spec (http://www.bswd.com/sff8020i.pdf) as reference.
According to p.31, we have to send out the CDB after receiving
the CDB interrupt. So, just clearing the irq looks not enough.

> 
> Does anybody actually have such a device anywhere?  :)
> 

Maybe Alan or Mark knows? 

We really need tester for such device.
How about the attached patch?

---
Albert

--- 07_followup/drivers/scsi/libata-core.c	2006-03-13 16:25:52.000000000 +0800
+++ 08_need_tester/drivers/scsi/libata-core.c	2006-03-14 12:45:09.000000000 +0800
@@ -1294,8 +1294,12 @@ static int ata_dev_configure(struct ata_
 		}
 		dev->cdb_len = (unsigned int) rc;
 
-		if (ata_id_cdb_intr(dev->id))
+		if (ata_id_cdb_intr(dev->id)) {
 			dev->flags |= ATA_DFLAG_CDB_INTR;
+			printk(KERN_INFO "This ATAPI device generates CDB intr.\n"
+			       "Please send an email to linux-ide@vger.kernel.org\n"
+			       "if you are interested to help testing such device.\n");
+		}
 
 		/* print device info to dmesg */
 		if (print_info)