Skip to contentSkip to conversation

scsi: fix check of PQ and PDT bits for WLUNs

PATCH In review Martin Petermann 3 participants 4 messages Aug 12, 2008 – Aug 25, 2008 SCSI

lore
Share & tools
lore
#1 lore raw

PATCH 1/1 scsi: fix check of PQ and PDT bits for WLUNs

From: Martin Petermann <[email protected]>

With kernel version 2.6.19 a check was introduced not to create a generic SCSI device for devices that return PQ=1 and PDT=0x1f. For WLUNs (see SAM-3, p. 41ff) generic SCSI devices should be created unconditionally without looking at the PQ bit. I'm referring here to the thread "With kernel 2.6.19 no sg devices for devices that return PQ=1, PDT=0x1f" of the linux-scsi forum. This is my proposed fix.

---
 drivers/scsi/scsi_scan.c |    7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/scsi/scsi_scan.c +6 -1
@@ -75,6 +75,11 @@
 #define SCSI_SCAN_TARGET_PRESENT	1
 #define SCSI_SCAN_LUN_PRESENT		2
 
+#define SCSI_W_LUN_BASE 0xc100
+#define SCSI_W_LUN_REPORT_LUNS (SCSI_W_LUN_BASE + 1)
+#define SCSI_W_LUN_ACCESS_CONTROL (SCSI_W_LUN_BASE + 2)
+#define SCSI_W_LUN_TARGET_LOG_PAGE (SCSI_W_LUN_BASE + 3)
+
 static const char *scsi_null_device_strs = "nullnullnullnull";
 
 #define MAX_SCSI_LUNS	512
@@ -1080,7 +1085,7 @@ static int scsi_probe_and_add_lun(struct
 	 * PDT=1Fh none (no FDD connected to the requested logical unit)
 	 */
 	if (((result[0] >> 5) == 1 || starget->pdt_1f_for_no_lun) &&
-	     (result[0] & 0x1f) == 0x1f) {
+	    (result[0] & 0x1f) == 0x1f && (lun & 0xff00) != SCSI_W_LUN_BASE) {
 		SCSI_LOG_SCAN_BUS(3, printk(KERN_INFO
 					"scsi scan: peripheral device type"
 					" of 31, no device added\n"));
maintainer
#2 lore raw
Quoted 19 lines · Martin Petermann

plain text document attachment (scsi_scan.diff)
From: Martin Petermann <[email protected]>

With kernel version 2.6.19 a check was introduced not to create a generic SCSI device for devices that return PQ=1 and PDT=0x1f. For WLUNs (see SAM-3, p. 41ff) generic SCSI devices should be created unconditionally without looking at the PQ bit. I'm referring here to the thread "With kernel 2.6.19 no sg devices for devices that return PQ=1, PDT=0x1f" of the linux-scsi forum. This is my proposed fix.

---
 drivers/scsi/scsi_scan.c |    7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

drivers/scsi/scsi_scan.c line 81

Quoted 7 lines
drivers/scsi/scsi_scan.c +4 -0
@@ -75,6 +75,11 @@
 #define SCSI_SCAN_TARGET_PRESENT	1
 #define SCSI_SCAN_LUN_PRESENT		2
 
+#define SCSI_W_LUN_BASE 0xc100
+#define SCSI_W_LUN_REPORT_LUNS (SCSI_W_LUN_BASE + 1)
+#define SCSI_W_LUN_ACCESS_CONTROL (SCSI_W_LUN_BASE + 2)
+#define SCSI_W_LUN_TARGET_LOG_PAGE (SCSI_W_LUN_BASE + 3)

These need to be in a header file, I think, probably scsi.h

drivers/scsi/scsi_scan.c static int scsi_probe_and_add_lun(struct · line 1088

Quoted 9 lines
drivers/scsi/scsi_scan.c +1 -1
 static const char *scsi_null_device_strs = "nullnullnullnull";
 
 #define MAX_SCSI_LUNS	512
@@ -1080,7 +1085,7 @@ static int scsi_probe_and_add_lun(struct
 	 * PDT=1Fh none (no FDD connected to the requested logical unit)
 	 */
 	if (((result[0] >> 5) == 1 || starget->pdt_1f_for_no_lun) &&
-	     (result[0] & 0x1f) == 0x1f) {
+	    (result[0] & 0x1f) == 0x1f && (lun & 0xff00) != SCSI_W_LUN_BASE) {

And this should probably be wrapped nicely too.

drivers/scsi/scsi_scan.c static int scsi_probe_and_add_lun(struct · line 1091

Quoted 3 lines
drivers/scsi/scsi_scan.c +0 -0
 		SCSI_LOG_SCAN_BUS(3, printk(KERN_INFO
 					"scsi scan: peripheral device type"
 					" of 31, no device added\n"));

Does this look OK?

James

---

drivers/scsi/scsi_scan.c +2 -1
index 84b4879..34d0de6 100644
@@ -1080,7 +1080,8 @@ static int scsi_probe_and_add_lun(struct scsi_target *starget,
 	 * PDT=1Fh none (no FDD connected to the requested logical unit)
 	 */
 	if (((result[0] >> 5) == 1 || starget->pdt_1f_for_no_lun) &&
-	     (result[0] & 0x1f) == 0x1f) {
+	    (result[0] & 0x1f) == 0x1f &&
+	    !scsi_is_wlun(lun)) {
 		SCSI_LOG_SCAN_BUS(3, printk(KERN_INFO
 					"scsi scan: peripheral device type"
 					" of 31, no device added\n"));
include/scsi/scsi.h +14 -0
index 5c40cc5..192f871 100644
@@ -309,6 +309,20 @@ struct scsi_lun {
 };
 
 /*
+ * The Well Known LUNS (SAM-3) in our int representation of a LUN
+ */
+#define SCSI_W_LUN_BASE 0xc100
+#define SCSI_W_LUN_REPORT_LUNS (SCSI_W_LUN_BASE + 1)
+#define SCSI_W_LUN_ACCESS_CONTROL (SCSI_W_LUN_BASE + 2)
+#define SCSI_W_LUN_TARGET_LOG_PAGE (SCSI_W_LUN_BASE + 3)
+
+static inline int scsi_is_wlun(unsigned int lun)
+{
+	return (lun & 0xff00) == SCSI_W_LUN_BASE;
+}
+
+
+/*
  *  MESSAGE CODES
  */
 
#3 lore raw
Quoted 80 lines · James Bottomley
Quoted 19 lines · Martin Petermann

plain text document attachment (scsi_scan.diff)
From: Martin Petermann <[email protected]>

With kernel version 2.6.19 a check was introduced not to create a generic SCSI device for devices that return PQ=1 and PDT=0x1f. For WLUNs (see SAM-3, p. 41ff) generic SCSI devices should be created unconditionally without looking at the PQ bit. I'm referring here to the thread "With kernel 2.6.19 no sg devices for devices that return PQ=1, PDT=0x1f" of the linux-scsi forum. This is my proposed fix.

---
 drivers/scsi/scsi_scan.c |    7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

drivers/scsi/scsi_scan.c line 81

Quoted 7 lines
drivers/scsi/scsi_scan.c +4 -0
@@ -75,6 +75,11 @@
 #define SCSI_SCAN_TARGET_PRESENT	1
 #define SCSI_SCAN_LUN_PRESENT		2
 
+#define SCSI_W_LUN_BASE 0xc100
+#define SCSI_W_LUN_REPORT_LUNS (SCSI_W_LUN_BASE + 1)
+#define SCSI_W_LUN_ACCESS_CONTROL (SCSI_W_LUN_BASE + 2)
+#define SCSI_W_LUN_TARGET_LOG_PAGE (SCSI_W_LUN_BASE + 3)

These need to be in a header file, I think, probably scsi.h

drivers/scsi/scsi_scan.c static int scsi_probe_and_add_lun(struct · line 1088

Quoted 9 lines
drivers/scsi/scsi_scan.c +1 -1
 static const char *scsi_null_device_strs = "nullnullnullnull";
 
 #define MAX_SCSI_LUNS	512
@@ -1080,7 +1085,7 @@ static int scsi_probe_and_add_lun(struct
 	 * PDT=1Fh none (no FDD connected to the requested logical unit)
 	 */
 	if (((result[0] >> 5) == 1 || starget->pdt_1f_for_no_lun) &&
-	     (result[0] & 0x1f) == 0x1f) {
+	    (result[0] & 0x1f) == 0x1f && (lun & 0xff00) != SCSI_W_LUN_BASE) {

And this should probably be wrapped nicely too.

drivers/scsi/scsi_scan.c static int scsi_probe_and_add_lun(struct scsi_target *starget, · line 1087

Quoted 3 lines
drivers/scsi/scsi_scan.c +0 -0
 		SCSI_LOG_SCAN_BUS(3, printk(KERN_INFO
 					"scsi scan: peripheral device type"
 					" of 31, no device added\n"));

Does this look OK?

James

---

drivers/scsi/scsi_scan.c +2 -1
index 84b4879..34d0de6 100644
@@ -1080,7 +1080,8 @@ static int scsi_probe_and_add_lun(struct scsi_target *starget,
 	 * PDT=1Fh none (no FDD connected to the requested logical unit)
 	 */
 	if (((result[0] >> 5) == 1 || starget->pdt_1f_for_no_lun) &&
-	     (result[0] & 0x1f) == 0x1f) {
+	    (result[0] & 0x1f) == 0x1f &&
+	    !scsi_is_wlun(lun)) {
 		SCSI_LOG_SCAN_BUS(3, printk(KERN_INFO
 					"scsi scan: peripheral device type"
 					" of 31, no device added\n"));
include/scsi/scsi.h +6 -0
index 5c40cc5..192f871 100644
@@ -309,6 +309,20 @@ struct scsi_lun {
 };
 
 /*
+ * The Well Known LUNS (SAM-3) in our int representation of a LUN
+ */
+#define SCSI_W_LUN_BASE 0xc100
+#define SCSI_W_LUN_REPORT_LUNS (SCSI_W_LUN_BASE + 1)
+#define SCSI_W_LUN_ACCESS_CONTROL (SCSI_W_LUN_BASE + 2)
+#define SCSI_W_LUN_TARGET_LOG_PAGE (SCSI_W_LUN_BASE + 3)

include/scsi/scsi.h struct scsi_lun { · line 327

Quoted 10 lines
include/scsi/scsi.h +8 -0
+
+static inline int scsi_is_wlun(unsigned int lun)
+{
+	return (lun & 0xff00) == SCSI_W_LUN_BASE;
+}
+
+
+/*
  *  MESSAGE CODES
  */

Looks ok.

So ACK from my side, if required.

Cheers Swen