[PATCH 4/4] libata: Check per device max cdb length for SAS

[PATCH 4/4] libata: Check per device max cdb length for SAS

[Brian King]

Since libata does not own any scsi_host struct for
SAS attached SATA devices, libata cannot rely on scsi
core checking the max_cmd_len. This patch adds additional
checking into the SAS queuecommand path to prevent
too large of a cdb from being issued to an ATA or ATAPI device.

Signed-off-by: Brian King <brking@us.ibm.com>
---

 libata-dev-bjking1/drivers/scsi/libata-scsi.c |   10 +++++++---
 1 files changed, 7 insertions(+), 3 deletions(-)

diff -puN drivers/scsi/libata-scsi.c~libata_sas_cdb_len drivers/scsi/libata-scsi.c
--- libata-dev/drivers/scsi/libata-scsi.c~libata_sas_cdb_len	2006-06-27 14:43:35.000000000 -0500
+++ libata-dev-bjking1/drivers/scsi/libata-scsi.c	2006-06-27 14:43:35.000000000 -0500
@@ -3207,9 +3207,13 @@ int ata_sas_queuecmd(struct scsi_cmnd *c
 {
 	ata_scsi_dump_cdb(ap, cmd);
 
-	if (likely(ata_scsi_dev_enabled(ap->device)))
-		__ata_scsi_queuecmd(cmd, done, ap->device);
-	else {
+	if (likely(ata_scsi_dev_enabled(ap->device))) {
+		if (unlikely(cmd->cmd_len > ap->device->cdb_len)) {
+			cmd->result = (DID_ABORT << 16);
+			done(cmd);
+		} else
+			__ata_scsi_queuecmd(cmd, done, ap->device);
+	} else {
 		cmd->result = (DID_BAD_TARGET << 16);
 		done(cmd);
 	}
_

Re: [PATCH 4/4] libata: Check per device max cdb length for SAS - drop

[Brian King]

Brian King wrote:
> Since libata does not own any scsi_host struct for
> SAS attached SATA devices, libata cannot rely on scsi
> core checking the max_cmd_len. This patch adds additional
> checking into the SAS queuecommand path to prevent
> too large of a cdb from being issued to an ATA or ATAPI device.

I'd like to drop this patch from the series. This patch would end up
preventing ATA passthrough from working to ATA/ATAPI devices that
support < 16 byte CDBs.


Brian