>From a50c8a50550ab3d41f716711e08ec6f116a45f28 Mon Sep 17 00:00:00 2001
From: Nicholas Bellinger <nab@xxxxxxxxxxxxxxx>
Date: Fri, 23 Jan 2009 17:34:14 -0800
Subject: [PATCH 6/6] [Target_Core_Mod/ConfigFS]: Add support for $STORAGE_OBJECT EVPD Unit Serial Number emulation
This patch support for writeable EVPD Unit Serial configfs attribute at:
/sys/kernel/config/target/core/$HBA/$STORAGE_OBJECT/wwn/evpd_unit_serial
It works with Linux/Block and Linux/FILEIO $STORAGE_OBJECTS, as well as through
target_core_mod/RAMDISK subsystem plugins.
This logic checks se_subsystem_dev_t->su_dev_flags & SDF_FIRMWARE_EVPD_UNIT_SERIAL
to see if SCSI firmware (or software emulation for non SCSI devices) is being
handled by Linux/SCSI for us. It does right by the firmware and if said device
reports EVPD INQUIRY response payload information see target_core_pscsi.c:
pscsi_transport_complete()
The changing of configfs evpd_unit_serial with this patch is limited to only while
said $STORAGE_OBJECT is *NOT* in use by $FABRIC_MOD exports. This is mentioned
in the comment in target_core_dev_wwn_store_attr_evpd_unit_serial():
<SNIP>
/*
* Check to see if any active $FABRIC_MOD exports exist. If they
* do exist, fail here as changing this information on the fly
* (underneath the initiator side OS dependent multipath code)
* could cause negative effects.
*/
if ((dev = su_dev->se_dev_ptr)) {
if (DEV_OBJ_API(dev)->check_count(&dev->dev_export_obj)) {
printk(KERN_ERR "Unable to set EVPD Unit Serial while"
" active %d $FABRIC_MOD exports exist\n",
DEV_OBJ_API(dev)->check_count(
&dev->dev_access_obj));
return(-EINVAL);
}
}
<SNIP>
This logic may end up allowing to be overwritten for those who
"really know what they are doing" and to allow RAMDISK and IBLOCK/FILEIO
exports for the EVPD information that has no assoicated underlying
device firmware. For now, this assoicated target_core_mod subsystem
plugin logic just disables emulation when firmware and Linux/SCSI LLDs
report EVPD to us upon $HBA/$STORAGE_OBJECT enablement.
Signed-off-by: Nicholas A. Bellinger <nab@xxxxxxxxxxxxxxx>
---
drivers/lio-core/target_core_configfs.c | 59 +++++++++++++++++++++++++++++-
1 files changed, 57 insertions(+), 2 deletions(-)
diff --git a/drivers/lio-core/target_core_configfs.c b/drivers/lio-core/target_core_configfs.c
index 5b1d275..1788fbb 100644
--- a/drivers/lio-core/target_core_configfs.c
+++ b/drivers/lio-core/target_core_configfs.c
@@ -38,6 +38,7 @@
#include <target_core_device.h>
#include <target_core_hba.h>
#include <target_core_plugin.h>
+#include <target_core_seobj.h>
#include <target_core_transport.h>
#ifdef SNMP_SUPPORT
@@ -536,7 +537,7 @@ static ssize_t target_core_dev_wwn_show_attr_evpd_unit_serial (
if (!(dev = se_dev->se_dev_ptr))
return(-ENODEV);
- return(sprintf(page, "T10 EVPD Unit Serial Number:: %s\n",
+ return(sprintf(page, "T10 EVPD Unit Serial Number: %s\n",
&t10_wwn->unit_serial[0]));
}
@@ -545,7 +546,61 @@ static ssize_t target_core_dev_wwn_store_attr_evpd_unit_serial (
const char *page,
size_t count)
{
- return(-ENOMEM);
+ se_subsystem_dev_t *su_dev = t10_wwn->t10_sub_dev;
+ se_device_t *dev;
+ unsigned char buf[INQUIRY_EVPD_SERIAL_LEN];
+
+ /*
+ * If Linux/SCSI subsystem_api_t plugin got a EVPD Unit Serial
+ * from the struct scsi_device level firmware, do not allow
+ * EVPD Unit Serial to be emulated.
+ *
+ * Note this struct scsi_device could also be emulating EVPD
+ * information from its drivers/scsi LLD. But for now we assume
+ * it is doing 'the right thing' wrt a world wide unique
+ * EVPD Unit Serial Number that OS dependent multipath can depend on.
+ */
+ if (su_dev->su_dev_flags & SDF_FIRMWARE_EVPD_UNIT_SERIAL) {
+ printk(KERN_ERR "Underlying SCSI device firmware provided EVPD"
+ " Unit Serial, ignoring request\n");
+ return(-EOPNOTSUPP);
+ }
+
+ if ((strlen(page) + 1) > INQUIRY_EVPD_SERIAL_LEN) {
+ printk(KERN_ERR "Emulated EVPD Unit Serial exceeds"
+ " INQUIRY_EVPD_SERIAL_LEN: %d\n", INQUIRY_EVPD_SERIAL_LEN);
+ return(-EOVERFLOW);
+ }
+ /*
+ * Check to see if any active $FABRIC_MOD exports exist. If they
+ * do exist, fail here as changing this information on the fly
+ * (underneath the initiator side OS dependent multipath code)
+ * could cause negative effects.
+ */
+ if ((dev = su_dev->se_dev_ptr)) {
+ if (DEV_OBJ_API(dev)->check_count(&dev->dev_export_obj)) {
+ printk(KERN_ERR "Unable to set EVPD Unit Serial while"
+ " active %d $FABRIC_MOD exports exist\n",
+ DEV_OBJ_API(dev)->check_count(
+ &dev->dev_access_obj));
+ return(-EINVAL);
+ }
+ }
+ /*
+ * This currently assumes ASCII encoding for emulated EVPD Unit Serial.
+ *
+ * Also, strip any newline added from the userspace
+ * echo $UUID > $TARGET/$HBA/$STORAGE_OBJECT/wwn/evpd_unit_serial
+ */
+ memset(buf, 0, INQUIRY_EVPD_SERIAL_LEN);
+ snprintf(buf, INQUIRY_EVPD_SERIAL_LEN, "%s", page);
+ snprintf(su_dev->t10_wwn.unit_serial, INQUIRY_EVPD_SERIAL_LEN,
+ "%s", strstrip(buf));
+ su_dev->su_dev_flags |= SDF_EMULATED_EVPD_UNIT_SERIAL;
+
+ printk("Target_Core_ConfigFS: Set emulated EVPD Unit Serial: %s\n",
+ su_dev->t10_wwn.unit_serial);
+ return(count);
}
SE_DEV_WWN_ATTR(evpd_unit_serial, S_IRUGO | S_IWUSR);
--
1.5.4.1
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
