[PATCH] 2.4.23

[PATCH] 2.4.23

[Frank Borich]

Hi All,

My name is Frank Borich, representing Xyratex International
(http://www.xyratex.com).  We design, and manufacture Data Storage
Subsystems.  During integration testing, we discovered that a patch to
scsi_scan.c is required (2.4.x kernels only), in order to set the
BLIST_SPARSELUN flag for our subsystems.  This flag adds support for
non-contiguous lun numbering.  We are unable to use the already
implemented "device_list" to set flags for our subsystems.  This is due
to the fact that we sell only to OEMs, and vendor name will change.
Rather than submit a separate patch for each branded subsystem we can
look elsewhere in the SCSI inquiry data, and set the flag accordingly.
The logic below, is executed once the end of the device_list is reached.
When the end of the table is reached, a memcmp is issued to look for a
particular string located at offset 96.  If the string is found, the
flag BLIST_SPARSELUN is returned.  Below is a patch for the SCSI driver
(scsi_scan.c). 


diff -urN linux-2.4.23/drivers/scsi/scsi_scan.c
linux-2.4.23.changed/drivers/scsi/scsi_scan.c
--- linux-2.4.23/drivers/scsi/scsi_scan.c	2003-12-03
08:56:28.000000000 -0600
+++ linux-2.4.23.changed/drivers/scsi/scsi_scan.c	2003-12-03
09:02:15.000000000 -0600
@@ -292,8 +292,16 @@
 	int i = 0;
 	unsigned char *pnt;
 	for (i = 0; 1; i++) {
-		if (device_list[i].vendor == NULL)
+		if (device_list[i].vendor == NULL){
+			/* Set flags for Xyratex subsystems here (all
OEMs). 
+			Cannot use device_list because can have
different vendor names for same device */
+			pnt = &response_data[96];
+			while (*pnt && *pnt == ' ')
+				pnt++;
+			if (memcmp("Xyratex F7",pnt,10) == 0)
+			return BLIST_SPARSELUN;
 			return 0;
+		}
 		pnt = &response_data[8];
 		while (*pnt && *pnt == ' ')
 			pnt++;
Frank Borich
------------------------------------------------------------------
Systems Engineer
Direct 630 654 6089
Fax 630 325 2513
e-mail frank_borich@us.xyratex.com
 


Frank Borich
------------------------------------------------------------------
Systems Engineer
Direct 630 654 6089
Fax 630 325 2513
e-mail frank_borich@us.xyratex.com
 

Re: [PATCH] 2.4.23

[Christoph Hellwig]

On Wed, Dec 17, 2003 at 08:13:40AM -0800, Frank Borich wrote:
> Hi All,
> 
> My name is Frank Borich, representing Xyratex International
> (http://www.xyratex.com).  We design, and manufacture Data Storage
> Subsystems.  During integration testing, we discovered that a patch to
> scsi_scan.c is required (2.4.x kernels only), in order to set the
> BLIST_SPARSELUN flag for our subsystems.  This flag adds support for
> non-contiguous lun numbering.  We are unable to use the already
> implemented "device_list" to set flags for our subsystems.  This is due
> to the fact that we sell only to OEMs, and vendor name will change.
> Rather than submit a separate patch for each branded subsystem we can
> look elsewhere in the SCSI inquiry data, and set the flag accordingly.
> The logic below, is executed once the end of the device_list is reached.
> When the end of the table is reached, a memcmp is issued to look for a
> particular string located at offset 96.  If the string is found, the
> flag BLIST_SPARSELUN is returned.  Below is a patch for the SCSI driver
> (scsi_scan.c). 

I'm not too happy about the form of this patch. Could you cook up one
for 2.6.0 that adds a function pointer to poke at the inquiry data
to the blacklist?  We could backport it to 2.4 later.

Re: [PATCH] 2.4.23

[James Bottomley]

On Sun, 2003-12-28 at 14:10, Christoph Hellwig wrote:
> I'm not too happy about the form of this patch. Could you cook up one
> for 2.6.0 that adds a function pointer to poke at the inquiry data
> to the blacklist?  We could backport it to 2.4 later.

Actually, this can't really be done at all for 2.6 since we're extremely
careful not to get inquiry data beyond byte 56 and this patch is looking
into the vendor specific area at 96.

James

RE: [PATCH] 2.4.23

[Frank Borich]

Shall I generate a patch that uses a function pointer in the blacklist,
to poke at the inquiry data, for 2.4 ?  

Frank Borich
------------------------------------------------------------------
Systems Engineer
Direct 630 654 6089
Fax 630 325 2513
e-mail frank_borich@us.xyratex.com
 

> -----Original Message-----
> From: James Bottomley [mailto:James.Bottomley@steeleye.com]
> Sent: Sunday, December 28, 2003 2:46 PM
> To: Christoph Hellwig
> Cc: Frank Borich; SCSI Mailing List
> Subject: Re: [PATCH] 2.4.23
> 
> On Sun, 2003-12-28 at 14:10, Christoph Hellwig wrote:
> > I'm not too happy about the form of this patch. Could you cook up
one
> > for 2.6.0 that adds a function pointer to poke at the inquiry data
> > to the blacklist?  We could backport it to 2.4 later.
> 
> Actually, this can't really be done at all for 2.6 since we're
extremely
> careful not to get inquiry data beyond byte 56 and this patch is
looking
> into the vendor specific area at 96.
> 
> James
>