[SCSI] FC Transport support for vports based on NPIV
authorJames Smart <James.Smart@Emulex.Com>
Fri, 27 Apr 2007 16:41:09 +0000 (12:41 -0400)
committerJames Bottomley <jejb@mulgrave.il.steeleye.com>
Wed, 16 May 2007 13:36:15 +0000 (09:36 -0400)
This patch provides support for FC virtual ports based on NPIV.
For information on the interfaces and design, please read the
Documentation/scsi/scsi_fc_transport.txt file enclosed within
the patch.

The RFC was originally posted here:
http://marc.info/?l=linux-scsi&m=117226959918393&w=2

Changes from the initial RFC:
- Bug fix: needed a transport_class_unregister() for the vport class
- Create a symlink to the vport in the shost device if it is not the
    parent of the vport.
- Made symbolic name writable so it can be set after creation
- Made the temporary fc_vport_identifiers struct private to the
transport.
- Deleted the vport_id field from the vport. I couldn't find any good
  use for it (and symname is a good replacement).
- Made the vport_state and vport_last_state "private" attributes.
  Added the fc_vport_set_state() helper function to manage state
  transitions
- Updated vport_create() to allow a vport to be created in a disabled
  state.
- Added INITIALIZING and FAILED vport states
- Added VPCERR_xxx defines for errors to be returned from vport_create()
- Created a Documentation/scsi/scsi_fc_transport.txt file that describes
  the interfaces and expected LLDD behaviors.

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Documentation/scsi/scsi_fc_transport.txt [new file with mode: 0644]
drivers/scsi/scsi_transport_fc.c
include/scsi/scsi_transport_fc.h

diff --git a/Documentation/scsi/scsi_fc_transport.txt b/Documentation/scsi/scsi_fc_transport.txt
new file mode 100644 (file)
index 0000000..ab057af
--- /dev/null
@@ -0,0 +1,450 @@
+                             SCSI FC Tansport
+                 =============================================
+
+Date:  4/12/2007
+Kernel Revisions for features:
+  rports : <<TBS>>
+  vports : 2.6.22 (? TBD)
+
+
+Introduction
+============
+This file documents the features and components of the SCSI FC Transport.
+It also provides documents the API between the transport and FC LLDDs.
+The FC transport can be found at:
+  drivers/scsi/scsi_transport_fc.c
+  include/scsi/scsi_transport_fc.h
+  include/scsi/scsi_netlink_fc.h
+
+This file is found at Documentation/scsi/scsi_fc_transport.txt
+
+
+FC Remote Ports (rports)
+========================================================================
+<< To Be Supplied >>
+
+
+FC Virtual Ports (vports)
+========================================================================
+
+Overview:
+-------------------------------
+
+  New FC standards have defined mechanisms which allows for a single physical
+  port to appear on as multiple communication ports. Using the N_Port Id
+  Virtualization (NPIV) mechanism, a point-to-point connection to a Fabric
+  can be assigned more than 1 N_Port_ID.  Each N_Port_ID appears as a
+  separate port to other endpoints on the fabric, even though it shares one
+  physical link to the switch for communication. Each N_Port_ID can have a
+  unique view of the fabric based on fabric zoning and array lun-masking
+  (just like a normal non-NPIV adapter).  Using the Virtual Fabric (VF)
+  mechanism, adding a fabric header to each frame allows the port to
+  interact with the Fabric Port to join multiple fabrics. The port will
+  obtain an N_Port_ID on each fabric it joins. Each fabric will have its
+  own unique view of endpoints and configuration parameters.  NPIV may be
+  used together with VF so that the port can obtain multiple N_Port_IDs
+  on each virtual fabric.
+
+  The FC transport is now recognizing a new object - a vport.  A vport is
+  an entity that has a world-wide unique World Wide Port Name (wwpn) and
+  World Wide Node Name (wwnn). The transport also allows for the FC4's to
+  be specified for the vport, with FCP_Initiator being the primary role
+  expected. Once instantiated by one of the above methods, it will have a
+  distinct N_Port_ID and view of fabric endpoints and storage entities.
+  The fc_host associated with the physical adapter will export the ability
+  to create vports. The transport will create the vport object within the
+  Linux device tree, and instruct the fc_host's driver to instantiate the
+  virtual port. Typically, the driver will create a new scsi_host instance
+  on the vport, resulting in a unique <H,C,T,L> namespace for the vport.
+  Thus, whether a FC port is based on a physical port or on a virtual port,
+  each will appear as a unique scsi_host with its own target and lun space.
+
+  Note: At this time, the transport is written to create only NPIV-based
+    vports. However, consideration was given to VF-based vports and it
+    should be a minor change to add support if needed.  The remaining
+    discussion will concentrate on NPIV.
+
+  Note: World Wide Name assignment (and uniqueness guarantees) are left
+    up to an administrative entity controling the vport. For example,
+    if vports are to be associated with virtual machines, a XEN mgmt
+    utility would be responsible for creating wwpn/wwnn's for the vport,
+    using it's own naming authority and OUI. (Note: it already does this
+    for virtual MAC addresses).
+
+
+Device Trees and Vport Objects:
+-------------------------------
+
+  Today, the device tree typically contains the scsi_host object,
+  with rports and scsi target objects underneath it. Currently the FC
+  transport creates the vport object and places it under the scsi_host
+  object corresponding to the physical adapter.  The LLDD will allocate
+  a new scsi_host for the vport and link it's object under the vport.
+  The remainder of the tree under the vports scsi_host is the same
+  as the non-NPIV case. The transport is written currently to easily
+  allow the parent of the vport to be something other than the scsi_host.
+  This could be used in the future to link the object onto a vm-specific
+  device tree. If the vport's parent is not the physical port's scsi_host,
+  a symbolic link to the vport object will be placed in the physical
+  port's scsi_host.
+
+  Here's what to expect in the device tree :
+   The typical Physical Port's Scsi_Host:
+     /sys/devices/.../host17/
+   and it has the typical decendent tree:
+     /sys/devices/.../host17/rport-17:0-0/target17:0:0/17:0:0:0:
+   and then the vport is created on the Physical Port:
+     /sys/devices/.../host17/vport-17:0-0
+   and the vport's Scsi_Host is then created:
+     /sys/devices/.../host17/vport-17:0-0/host18
+   and then the rest of the tree progresses, such as:
+     /sys/devices/.../host17/vport-17:0-0/host18/rport-18:0-0/target18:0:0/18:0:0:0:
+
+  Here's what to expect in the sysfs tree :
+   scsi_hosts:
+     /sys/class/scsi_host/host17                physical port's scsi_host
+     /sys/class/scsi_host/host18                vport's scsi_host
+   fc_hosts:
+     /sys/class/fc_host/host17                  physical port's fc_host
+     /sys/class/fc_host/host18                  vport's fc_host
+   fc_vports:
+     /sys/class/fc_vports/vport-17:0-0          the vport's fc_vport
+   fc_rports:
+     /sys/class/fc_remote_ports/rport-17:0-0    rport on the physical port
+     /sys/class/fc_remote_ports/rport-18:0-0    rport on the vport
+
+
+Vport Attributes:
+-------------------------------
+
+  The new fc_vport class object has the following attributes
+
+     node_name:                                                        Read_Only
+       The WWNN of the vport
+
+     port_name:                                                        Read_Only
+       The WWPN of the vport
+
+     roles:                                                    Read_Only
+       Indicates the FC4 roles enabled on the vport.
+
+     symbolic_name:                                            Read_Write
+       A string, appended to the driver's symbolic port name string, which
+       is registered with the switch to identify the vport. For example,
+       a hypervisor could set this string to "Xen Domain 2 VM 5 Vport 2",
+       and this set of identifiers can be seen on switch management screens
+       to identify the port.
+
+     vport_delete:                                             Write_Only
+       When written with a "1", will tear down the vport.
+
+     vport_disable:                                            Write_Only
+       When written with a "1", will transition the vport to a disabled.
+       state.  The vport will still be instantiated with the Linux kernel,
+       but it will not be active on the FC link.
+       When written with a "0", will enable the vport.
+
+     vport_last_state:                                         Read_Only
+       Indicates the previous state of the vport.  See the section below on
+       "Vport States".
+
+     vport_state:                                              Read_Only
+       Indicates the state of the vport.  See the section below on
+       "Vport States".
+
+     vport_type:                                               Read_Only
+       Reflects the FC mechanism used to create the virtual port.
+       Only NPIV is supported currently.
+
+
+  For the fc_host class object, the following attributes are added for vports:
+
+     max_npiv_vports:                                          Read_Only
+       Indicates the maximum number of NPIV-based vports that the
+       driver/adapter can support on the fc_host.
+
+     npiv_vports_inuse:                                                Read_Only
+       Indicates how many NPIV-based vports have been instantiated on the
+       fc_host.
+
+     vport_create:                                             Write_Only
+       A "simple" create interface to instantiate a vport on an fc_host.
+       A "<WWPN>:<WWNN>" string is written to the attribute. The transport
+       then instantiates the vport object and calls the LLDD to create the
+       vport with the role of FCP_Initiator.  Each WWN is specified as 16
+       hex characters and may *not* contain any prefixes (e.g. 0x, x, etc).
+
+     vport_delete:                                             Write_Only
+        A "simple" delete interface to teardown a vport. A "<WWPN>:<WWNN>"
+       string is written to the attribute. The transport will locate the
+       vport on the fc_host with the same WWNs and tear it down.  Each WWN
+       is specified as 16 hex characters and may *not* contain any prefixes
+       (e.g. 0x, x, etc).
+
+
+Vport States:
+-------------------------------
+
+  Vport instantiation consists of two parts:
+    - Creation with the kernel and LLDD. This means all transport and
+      driver data structures are built up, and device objects created.
+      This is equivalent to a driver "attach" on an adapter, which is
+      independent of the adapter's link state.
+    - Instantiation of the vport on the FC link via ELS traffic, etc.
+      This is equivalent to a "link up" and successfull link initialization.
+  Futher information can be found in the interfaces section below for
+  Vport Creation.
+
+  Once a vport has been instantiated with the kernel/LLDD, a vport state
+  can be reported via the sysfs attribute. The following states exist:
+
+    FC_VPORT_UNKNOWN           - Unknown
+      An temporary state, typically set only while the vport is being
+      instantiated with the kernel and LLDD.
+
+    FC_VPORT_ACTIVE            - Active
+      The vport has been successfully been created on the FC link.
+      It is fully functional.
+
+    FC_VPORT_DISABLED          - Disabled
+      The vport instantiated, but "disabled". The vport is not instantiated
+      on the FC link. This is equivalent to a physical port with the
+      link "down".
+
+    FC_VPORT_LINKDOWN          - Linkdown
+      The vport is not operational as the physical link is not operational.
+
+    FC_VPORT_INITIALIZING      - Initializing
+      The vport is in the process of instantiating on the FC link.
+      The LLDD will set this state just prior to starting the ELS traffic
+      to create the vport. This state will persist until the vport is
+      successfully created (state becomes FC_VPORT_ACTIVE) or it fails
+      (state is one of the values below).  As this state is transitory,
+      it will not be preserved in the "vport_last_state".
+
+    FC_VPORT_NO_FABRIC_SUPP    - No Fabric Support
+      The vport is not operational. One of the following conditions were
+      encountered:
+       - The FC topology is not Point-to-Point
+       - The FC port is not connected to an F_Port
+       - The F_Port has indicated that NPIV is not supported.
+
+    FC_VPORT_NO_FABRIC_RSCS    - No Fabric Resources
+      The vport is not operational. The Fabric failed FDISC with a status
+      indicating that it does not have sufficient resources to complete
+      the operation.
+
+    FC_VPORT_FABRIC_LOGOUT     - Fabric Logout
+      The vport is not operational. The Fabric has LOGO'd the N_Port_ID
+      associated with the vport.
+
+    FC_VPORT_FABRIC_REJ_WWN    - Fabric Rejected WWN
+      The vport is not operational. The Fabric failed FDISC with a status
+      indicating that the WWN's are not valid.
+
+    FC_VPORT_FAILED            - VPort Failed
+      The vport is not operational. This is a catchall for all other
+      error conditions.
+
+
+  The following state table indicates the different state transitions:
+
+    State              Event                           New State
+    --------------------------------------------------------------------
+     n/a               Initialization                  Unknown
+    Unknown:           Link Down                       Linkdown
+                       Link Up & Loop                  No Fabric Support
+                       Link Up & no Fabric             No Fabric Support
+                       Link Up & FLOGI response        No Fabric Support
+                         indicates no NPIV support
+                       Link Up & FDISC being sent      Initializing
+                       Disable request                 Disable
+    Linkdown:          Link Up                         Unknown
+    Initializing:      FDISC ACC                       Active
+                       FDISC LS_RJT w/ no resources    No Fabric Resources
+                       FDISC LS_RJT w/ invalid         Fabric Rejected WWN
+                         pname or invalid nport_id
+                       FDISC LS_RJT failed for         Vport Failed
+                         other reasons
+                       Link Down                       Linkdown
+                       Disable request                 Disable
+    Disable:           Enable request                  Unknown
+    Active:            LOGO received from fabric       Fabric Logout
+                       Link Down                       Linkdown
+                       Disable request                 Disable
+    Fabric Logout:     Link still up                   Unknown
+
+         The following 4 error states all have the same transitions:
+    No Fabric Support:
+    No Fabric Resources:
+    Fabric Rejected WWN:
+    Vport Failed:
+                       Disable request                 Disable
+                       Link goes down                  Linkdown
+
+
+Transport <-> LLDD Interfaces :
+-------------------------------
+
+Vport support by LLDD:
+
+  The LLDD indicates support for vports by supplying a vport_create()
+  function in the transport template.  The presense of this function will
+  cause the creation of the new attributes on the fc_host.  As part of
+  the physical port completing its initialization relative to the
+  transport, it should set the max_npiv_vports attribute to indicate the
+  maximum number of vports the driver and/or adapter supports.
+
+
+Vport Creation:
+
+  The LLDD vport_create() syntax is:
+
+      int vport_create(struct fc_vport *vport, bool disable)
+
+    where:
+      vport:   Is the newly allocated vport object
+      disable: If "true", the vport is to be created in a disabled stated.
+               If "false", the vport is to be enabled upon creation.
+
+  When a request is made to create a new vport (via sgio/netlink, or the
+  vport_create fc_host attribute), the transport will validate that the LLDD
+  can support another vport (e.g. max_npiv_vports > npiv_vports_inuse).
+  If not, the create request will be failed.  If space remains, the transport
+  will increment the vport count, create the vport object, and then call the
+  LLDD's vport_create() function with the newly allocated vport object.
+
+  As mentioned above, vport creation is divided into two parts:
+    - Creation with the kernel and LLDD. This means all transport and
+      driver data structures are built up, and device objects created.
+      This is equivalent to a driver "attach" on an adapter, which is
+      independent of the adapter's link state.
+    - Instantiation of the vport on the FC link via ELS traffic, etc.
+      This is equivalent to a "link up" and successfull link initialization.
+
+  The LLDD's vport_create() function will not synchronously wait for both
+  parts to be fully completed before returning. It must validate that the
+  infrastructure exists to support NPIV, and complete the first part of
+  vport creation (data structure build up) before returning.  We do not
+  hinge vport_create() on the link-side operation mainly because:
+    - The link may be down. It is not a failure if it is. It simply
+      means the vport is in an inoperable state until the link comes up.
+      This is consistent with the link bouncing post vport creation.
+    - The vport may be created in a disabled state.
+    - This is consistent with a model where:  the vport equates to a
+      FC adapter. The vport_create is synonymous with driver attachment
+      to the adapter, which is independent of link state.
+
+    Note: special error codes have been defined to delineate infrastructure
+      failure cases for quicker resolution.
+
+  The expected behavior for the LLDD's vport_create() function is:
+    - Validate Infrastructure:
+        - If the driver or adapter cannot support another vport, whether
+            due to improper firmware, (a lie about) max_npiv, or a lack of
+           some other resource - return VPCERR_UNSUPPORTED.
+        - If the driver validates the WWN's against those already active on
+            the adapter and detects an overlap - return VPCERR_BAD_WWN.
+        - If the driver detects the topology is loop, non-fabric, or the
+            FLOGI did not support NPIV - return VPCERR_NO_FABRIC_SUPP.
+    - Allocate data structures. If errors are encountered, such as out
+        of memory conditions, return the respective negative Exxx error code.
+    - If the role is FCP Initiator, the LLDD is to :
+        - Call scsi_host_alloc() to allocate a scsi_host for the vport.
+       - Call scsi_add_host(new_shost, &vport->dev) to start the scsi_host
+         and bind it as a child of the vport device.
+       - Initializes the fc_host attribute values.
+    - Kick of further vport state transitions based on the disable flag and
+        link state - and return success (zero).
+
+  LLDD Implementers Notes:
+  - It is suggested that there be a different fc_function_templates for
+    the physical port and the virtual port.  The physical port's template
+    would have the vport_create, vport_delete, and vport_disable functions,
+    while the vports would not.
+  - It is suggested that there be different scsi_host_templates
+    for the physical port and virtual port. Likely, there are driver
+    attributes, embedded into the scsi_host_template, that are applicable
+    for the physical port only (link speed, topology setting, etc). This
+    ensures that the attributes are applicable to the respective scsi_host.
+
+
+Vport Disable/Enable:
+
+  The LLDD vport_disable() syntax is:
+
+      int vport_disable(struct fc_vport *vport, bool disable)
+
+    where:
+      vport:   Is vport to to be enabled or disabled
+      disable: If "true", the vport is to be disabled.
+               If "false", the vport is to be enabled.
+
+  When a request is made to change the disabled state on a vport, the
+  transport will validate the request against the existing vport state.
+  If the request is to disable and the vport is already disabled, the
+  request will fail. Similarly, if the request is to enable, and the
+  vport is not in a disabled state, the request will fail.  If the request
+  is valid for the vport state, the transport will call the LLDD to
+  change the vport's state.
+
+  Within the LLDD, if a vport is disabled, it remains instantiated with
+  the kernel and LLDD, but it is not active or visible on the FC link in
+  any way. (see Vport Creation and the 2 part instantiation discussion).
+  The vport will remain in this state until it is deleted or re-enabled.
+  When enabling a vport, the LLDD reinstantiates the vport on the FC
+  link - essentially restarting the LLDD statemachine (see Vport States
+  above).
+
+
+Vport Deletion:
+
+  The LLDD vport_delete() syntax is:
+
+      int vport_delete(struct fc_vport *vport)
+
+    where:
+      vport:   Is vport to delete
+
+  When a request is made to delete a vport (via sgio/netlink, or via the
+  fc_host or fc_vport vport_delete attributes), the transport will call
+  the LLDD to terminate the vport on the FC link, and teardown all other
+  datastructures and references.  If the LLDD completes successfully,
+  the transport will teardown the vport objects and complete the vport
+  removal.  If the LLDD delete request fails, the vport object will remain,
+  but will be in an indeterminate state.
+
+  Within the LLDD, the normal code paths for a scsi_host teardown should
+  be followed. E.g. If the vport has a FCP Initiator role, the LLDD
+  will call fc_remove_host() for the vports scsi_host, followed by
+  scsi_remove_host() and scsi_host_put() for the vports scsi_host.
+
+
+Other:
+  fc_host port_type attribute:
+    There is a new fc_host port_type value - FC_PORTTYPE_NPIV. This value
+    must be set on all vport-based fc_hosts.  Normally, on a physical port,
+    the port_type attribute would be set to NPORT, NLPORT, etc based on the
+    topology type and existence of the fabric. As this is not applicable to
+    a vport, it makes more sense to report the FC mechanism used to create
+    the vport.
+
+  Driver unload:
+    FC drivers are required to call fc_remove_host() prior to calling
+    scsi_remove_host().  This allows the fc_host to tear down all remote
+    ports prior the scsi_host being torn down.  The fc_remove_host() call
+    was updated to remove all vports for the fc_host as well.
+
+
+Credits
+=======
+The following people have contributed to this document:
+
+
+
+
+
+
+James Smart
+james.smart@emulex.com
+
index b4d1ece46f789d92e6cc94acdb38410625852233..217651468115aafc7488cb2beb8b70db101a1e37 100644 (file)
@@ -19,7 +19,7 @@
  *
  *  ========
  *
- *  Copyright (C) 2004-2005   James Smart, Emulex Corporation
+ *  Copyright (C) 2004-2007   James Smart, Emulex Corporation
  *    Rewrite for host, target, device, and remote port attributes,
  *    statistics, and service functions...
  *
 
 static int fc_queue_work(struct Scsi_Host *, struct work_struct *);
 
+/*
+ * This is a temporary carrier for creating a vport. It will eventually
+ * be replaced  by a real message definition for sgio or netlink.
+ *
+ * fc_vport_identifiers: This set of data contains all elements
+ * to uniquely identify and instantiate a FC virtual port.
+ *
+ * Notes:
+ *   symbolic_name: The driver is to append the symbolic_name string data
+ *      to the symbolic_node_name data that it generates by default.
+ *      the resulting combination should then be registered with the switch.
+ *      It is expected that things like Xen may stuff a VM title into
+ *      this field.
+ */
+struct fc_vport_identifiers {
+       u64 node_name;
+       u64 port_name;
+       u32 roles;
+       bool disable;
+       enum fc_port_type vport_type;   /* only FC_PORTTYPE_NPIV allowed */
+       char symbolic_name[FC_VPORT_SYMBOLIC_NAMELEN];
+};
+
+static int fc_vport_create(struct Scsi_Host *shost, int channel,
+       struct device *pdev, struct fc_vport_identifiers  *ids,
+       struct fc_vport **vport);
+
 /*
  * Redefine so that we can have same named attributes in the
  * sdev/starget/host objects.
@@ -90,10 +117,14 @@ static struct {
        { FC_PORTTYPE_NLPORT,   "NLPort (fabric via loop)" },
        { FC_PORTTYPE_LPORT,    "LPort (private loop)" },
        { FC_PORTTYPE_PTP,      "Point-To-Point (direct nport connection" },
+       { FC_PORTTYPE_NPIV,             "NPIV VPORT" },
 };
 fc_enum_name_search(port_type, fc_port_type, fc_port_type_names)
 #define FC_PORTTYPE_MAX_NAMELEN                50
 
+/* Reuse fc_port_type enum function for vport_type */
+#define get_fc_vport_type_name get_fc_port_type_name
+
 
 /* Convert fc_host_event_code values to ascii string name */
 static const struct {
@@ -139,6 +170,29 @@ fc_enum_name_search(port_state, fc_port_state, fc_port_state_names)
 #define FC_PORTSTATE_MAX_NAMELEN       20
 
 
+/* Convert fc_vport_state values to ascii string name */
+static struct {
+       enum fc_vport_state     value;
+       char                    *name;
+} fc_vport_state_names[] = {
+       { FC_VPORT_UNKNOWN,             "Unknown" },
+       { FC_VPORT_ACTIVE,              "Active" },
+       { FC_VPORT_DISABLED,            "Disabled" },
+       { FC_VPORT_LINKDOWN,            "Linkdown" },
+       { FC_VPORT_INITIALIZING,        "Initializing" },
+       { FC_VPORT_NO_FABRIC_SUPP,      "No Fabric Support" },
+       { FC_VPORT_NO_FABRIC_RSCS,      "No Fabric Resources" },
+       { FC_VPORT_FABRIC_LOGOUT,       "Fabric Logout" },
+       { FC_VPORT_FABRIC_REJ_WWN,      "Fabric Rejected WWN" },
+       { FC_VPORT_FAILED,              "VPort Failed" },
+};
+fc_enum_name_search(vport_state, fc_vport_state, fc_vport_state_names)
+#define FC_VPORTSTATE_MAX_NAMELEN      24
+
+/* Reuse fc_vport_state enum function for vport_last_state */
+#define get_fc_vport_last_state_name get_fc_vport_state_name
+
+
 /* Convert fc_tgtid_binding_type values to ascii string name */
 static const struct {
        enum fc_tgtid_binding_type      value;
@@ -219,16 +273,16 @@ show_fc_fc4s (char *buf, u8 *fc4_list)
 }
 
 
-/* Convert FC_RPORT_ROLE bit values to ascii string name */
+/* Convert FC_PORT_ROLE bit values to ascii string name */
 static const struct {
        u32                     value;
        char                    *name;
-} fc_remote_port_role_names[] = {
-       { FC_RPORT_ROLE_FCP_TARGET,     "FCP Target" },
-       { FC_RPORT_ROLE_FCP_INITIATOR,  "FCP Initiator" },
-       { FC_RPORT_ROLE_IP_PORT,        "IP Port" },
+} fc_port_role_names[] = {
+       { FC_PORT_ROLE_FCP_TARGET,      "FCP Target" },
+       { FC_PORT_ROLE_FCP_INITIATOR,   "FCP Initiator" },
+       { FC_PORT_ROLE_IP_PORT,         "IP Port" },
 };
-fc_bitfield_name_search(remote_port_roles, fc_remote_port_role_names)
+fc_bitfield_name_search(port_roles, fc_port_role_names)
 
 /*
  * Define roles that are specific to port_id. Values are relative to ROLE_MASK.
@@ -252,7 +306,8 @@ static void fc_scsi_scan_rport(struct work_struct *work);
  */
 #define FC_STARGET_NUM_ATTRS   3
 #define FC_RPORT_NUM_ATTRS     10
-#define FC_HOST_NUM_ATTRS      17
+#define FC_VPORT_NUM_ATTRS     9
+#define FC_HOST_NUM_ATTRS      21
 
 struct fc_internal {
        struct scsi_transport_template t;
@@ -278,6 +333,10 @@ struct fc_internal {
        struct transport_container rport_attr_cont;
        struct class_device_attribute private_rport_attrs[FC_RPORT_NUM_ATTRS];
        struct class_device_attribute *rport_attrs[FC_RPORT_NUM_ATTRS + 1];
+
+       struct transport_container vport_attr_cont;
+       struct class_device_attribute private_vport_attrs[FC_VPORT_NUM_ATTRS];
+       struct class_device_attribute *vport_attrs[FC_VPORT_NUM_ATTRS + 1];
 };
 
 #define to_fc_internal(tmpl)   container_of(tmpl, struct fc_internal, t)
@@ -331,6 +390,7 @@ static int fc_host_setup(struct transport_container *tc, struct device *dev,
                sizeof(fc_host->supported_fc4s));
        fc_host->supported_speeds = FC_PORTSPEED_UNKNOWN;
        fc_host->maxframe_size = -1;
+       fc_host->max_npiv_vports = 0;
        memset(fc_host->serial_number, 0,
                sizeof(fc_host->serial_number));
 
@@ -348,8 +408,11 @@ static int fc_host_setup(struct transport_container *tc, struct device *dev,
 
        INIT_LIST_HEAD(&fc_host->rports);
        INIT_LIST_HEAD(&fc_host->rport_bindings);
+       INIT_LIST_HEAD(&fc_host->vports);
        fc_host->next_rport_number = 0;
        fc_host->next_target_id = 0;
+       fc_host->next_vport_number = 0;
+       fc_host->npiv_vports_inuse = 0;
 
        snprintf(fc_host->work_q_name, KOBJ_NAME_LEN, "fc_wq_%d",
                shost->host_no);
@@ -387,6 +450,16 @@ static DECLARE_TRANSPORT_CLASS(fc_rport_class,
                               NULL,
                               NULL);
 
+/*
+ * Setup and Remove actions for virtual ports are handled
+ * in the service functions below.
+ */
+static DECLARE_TRANSPORT_CLASS(fc_vport_class,
+                              "fc_vports",
+                              NULL,
+                              NULL,
+                              NULL);
+
 /*
  * Module Parameters
  */
@@ -583,6 +656,9 @@ static __init int fc_transport_init(void)
        atomic_set(&fc_event_seq, 0);
 
        error = transport_class_register(&fc_host_class);
+       if (error)
+               return error;
+       error = transport_class_register(&fc_vport_class);
        if (error)
                return error;
        error = transport_class_register(&fc_rport_class);
@@ -596,6 +672,7 @@ static void __exit fc_transport_exit(void)
        transport_class_unregister(&fc_transport_class);
        transport_class_unregister(&fc_rport_class);
        transport_class_unregister(&fc_host_class);
+       transport_class_unregister(&fc_vport_class);
 }
 
 /*
@@ -800,9 +877,9 @@ show_fc_rport_roles (struct class_device *cdev, char *buf)
                        return snprintf(buf, 30, "Unknown Fabric Entity\n");
                }
        } else {
-               if (rport->roles == FC_RPORT_ROLE_UNKNOWN)
+               if (rport->roles == FC_PORT_ROLE_UNKNOWN)
                        return snprintf(buf, 20, "unknown\n");
-               return get_fc_remote_port_roles_names(rport->roles, buf);
+               return get_fc_port_roles_names(rport->roles, buf);
        }
 }
 static FC_CLASS_DEVICE_ATTR(rport, roles, S_IRUGO,
@@ -857,7 +934,7 @@ static FC_CLASS_DEVICE_ATTR(rport, fast_io_fail_tmo, S_IRUGO | S_IWUSR,
 
 /*
  * Note: in the target show function we recognize when the remote
- *  port is in the hierarchy and do not allow the driver to get
+ *  port is in the heirarchy and do not allow the driver to get
  *  involved in sysfs functions. The driver only gets involved if
  *  it's the "old" style that doesn't use rports.
  */
@@ -911,6 +988,260 @@ fc_starget_rd_attr_cast(port_name, "0x%llx\n", 20, unsigned long long);
 fc_starget_rd_attr(port_id, "0x%06x\n", 20);
 
 
+/*
+ * FC Virtual Port Attribute Management
+ */
+
+#define fc_vport_show_function(field, format_string, sz, cast)         \
+static ssize_t                                                         \
+show_fc_vport_##field (struct class_device *cdev, char *buf)           \
+{                                                                      \
+       struct fc_vport *vport = transport_class_to_vport(cdev);        \
+       struct Scsi_Host *shost = vport_to_shost(vport);                \
+       struct fc_internal *i = to_fc_internal(shost->transportt);      \
+       if ((i->f->get_vport_##field) &&                                \
+           !(vport->flags & (FC_VPORT_DEL | FC_VPORT_CREATING)))       \
+               i->f->get_vport_##field(vport);                         \
+       return snprintf(buf, sz, format_string, cast vport->field);     \
+}
+
+#define fc_vport_store_function(field)                                 \
+static ssize_t                                                         \
+store_fc_vport_##field(struct class_device *cdev, const char *buf,     \
+                          size_t count)                                \
+{                                                                      \
+       int val;                                                        \
+       struct fc_vport *vport = transport_class_to_vport(cdev);        \
+       struct Scsi_Host *shost = vport_to_shost(vport);                \
+       struct fc_internal *i = to_fc_internal(shost->transportt);      \
+       char *cp;                                                       \
+       if (vport->flags & (FC_VPORT_DEL | FC_VPORT_CREATING))  \
+               return -EBUSY;                                          \
+       val = simple_strtoul(buf, &cp, 0);                              \
+       if (*cp && (*cp != '\n'))                                       \
+               return -EINVAL;                                         \
+       i->f->set_vport_##field(vport, val);                            \
+       return count;                                                   \
+}
+
+#define fc_vport_store_str_function(field, slen)                       \
+static ssize_t                                                         \
+store_fc_vport_##field(struct class_device *cdev, const char *buf,     \
+                          size_t count)                                \
+{                                                                      \
+       struct fc_vport *vport = transport_class_to_vport(cdev);        \
+       struct Scsi_Host *shost = vport_to_shost(vport);                \
+       struct fc_internal *i = to_fc_internal(shost->transportt);      \
+       unsigned int cnt=count;                                         \
+                                                                       \
+       /* count may include a LF at end of string */                   \
+       if (buf[cnt-1] == '\n')                                         \
+               cnt--;                                                  \
+       if (cnt > ((slen) - 1))                                         \
+               return -EINVAL;                                         \
+       memcpy(vport->field, buf, cnt);                                 \
+       i->f->set_vport_##field(vport);                                 \
+       return count;                                                   \
+}
+
+#define fc_vport_rd_attr(field, format_string, sz)                     \
+       fc_vport_show_function(field, format_string, sz, )              \
+static FC_CLASS_DEVICE_ATTR(vport, field, S_IRUGO,                     \
+                        show_fc_vport_##field, NULL)
+
+#define fc_vport_rd_attr_cast(field, format_string, sz, cast)          \
+       fc_vport_show_function(field, format_string, sz, (cast))        \
+static FC_CLASS_DEVICE_ATTR(vport, field, S_IRUGO,                     \
+                         show_fc_vport_##field, NULL)
+
+#define fc_vport_rw_attr(field, format_string, sz)                     \
+       fc_vport_show_function(field, format_string, sz, )              \
+       fc_vport_store_function(field)                                  \
+static FC_CLASS_DEVICE_ATTR(vport, field, S_IRUGO | S_IWUSR,           \
+                       show_fc_vport_##field,                          \
+                       store_fc_vport_##field)
+
+#define fc_private_vport_show_function(field, format_string, sz, cast) \
+static ssize_t                                                         \
+show_fc_vport_##field (struct class_device *cdev, char *buf)           \
+{                                                                      \
+       struct fc_vport *vport = transport_class_to_vport(cdev);        \
+       return snprintf(buf, sz, format_string, cast vport->field);     \
+}
+
+#define fc_private_vport_store_u32_function(field)                     \
+static ssize_t                                                         \
+store_fc_vport_##field(struct class_device *cdev, const char *buf,     \
+                          size_t count)                                \
+{                                                                      \
+       u32 val;                                                        \
+       struct fc_vport *vport = transport_class_to_vport(cdev);        \
+       char *cp;                                                       \
+       if (vport->flags & (FC_VPORT_DEL | FC_VPORT_CREATING))          \
+               return -EBUSY;                                          \
+       val = simple_strtoul(buf, &cp, 0);                              \
+       if (*cp && (*cp != '\n'))                                       \
+               return -EINVAL;                                         \
+       vport->field = val;                                             \
+       return count;                                                   \
+}
+
+
+#define fc_private_vport_rd_attr(field, format_string, sz)             \
+       fc_private_vport_show_function(field, format_string, sz, )      \
+static FC_CLASS_DEVICE_ATTR(vport, field, S_IRUGO,                     \
+                        show_fc_vport_##field, NULL)
+
+#define fc_private_vport_rd_attr_cast(field, format_string, sz, cast)  \
+       fc_private_vport_show_function(field, format_string, sz, (cast)) \
+static FC_CLASS_DEVICE_ATTR(vport, field, S_IRUGO,                     \
+                         show_fc_vport_##field, NULL)
+
+#define fc_private_vport_rw_u32_attr(field, format_string, sz)         \
+       fc_private_vport_show_function(field, format_string, sz, )      \
+       fc_private_vport_store_u32_function(field)                      \
+static FC_CLASS_DEVICE_ATTR(vport, field, S_IRUGO | S_IWUSR,           \
+                       show_fc_vport_##field,                          \
+                       store_fc_vport_##field)
+
+
+#define fc_private_vport_rd_enum_attr(title, maxlen)                   \
+static ssize_t                                                         \
+show_fc_vport_##title (struct class_device *cdev, char *buf)           \
+{                                                                      \
+       struct fc_vport *vport = transport_class_to_vport(cdev);        \
+       const char *name;                                               \
+       name = get_fc_##title##_name(vport->title);                     \
+       if (!name)                                                      \
+               return -EINVAL;                                         \
+       return snprintf(buf, maxlen, "%s\n", name);                     \
+}                                                                      \
+static FC_CLASS_DEVICE_ATTR(vport, title, S_IRUGO,                     \
+                       show_fc_vport_##title, NULL)
+
+
+#define SETUP_VPORT_ATTRIBUTE_RD(field)                                        \
+       i->private_vport_attrs[count] = class_device_attr_vport_##field; \
+       i->private_vport_attrs[count].attr.mode = S_IRUGO;              \
+       i->private_vport_attrs[count].store = NULL;                     \
+       i->vport_attrs[count] = &i->private_vport_attrs[count];         \
+       if (i->f->get_##field)                                          \
+               count++
+       /* NOTE: Above MACRO differs: checks function not show bit */
+
+#define SETUP_PRIVATE_VPORT_ATTRIBUTE_RD(field)                                \
+       i->private_vport_attrs[count] = class_device_attr_vport_##field; \
+       i->private_vport_attrs[count].attr.mode = S_IRUGO;              \
+       i->private_vport_attrs[count].store = NULL;                     \
+       i->vport_attrs[count] = &i->private_vport_attrs[count];         \
+       count++
+
+#define SETUP_VPORT_ATTRIBUTE_WR(field)                                        \
+       i->private_vport_attrs[count] = class_device_attr_vport_##field; \
+       i->vport_attrs[count] = &i->private_vport_attrs[count];         \
+       if (i->f->field)                                                \
+               count++
+       /* NOTE: Above MACRO differs: checks function */
+
+#define SETUP_VPORT_ATTRIBUTE_RW(field)                                        \
+       i->private_vport_attrs[count] = class_device_attr_vport_##field; \
+       if (!i->f->set_vport_##field) {                                 \
+               i->private_vport_attrs[count].attr.mode = S_IRUGO;      \
+               i->private_vport_attrs[count].store = NULL;             \
+       }                                                               \
+       i->vport_attrs[count] = &i->private_vport_attrs[count];         \
+       count++
+       /* NOTE: Above MACRO differs: does not check show bit */
+
+#define SETUP_PRIVATE_VPORT_ATTRIBUTE_RW(field)                                \
+{                                                                      \
+       i->private_vport_attrs[count] = class_device_attr_vport_##field; \
+       i->vport_attrs[count] = &i->private_vport_attrs[count];         \
+       count++;                                                        \
+}
+
+
+/* The FC Transport Virtual Port Attributes: */
+
+/* Fixed Virtual Port Attributes */
+
+/* Dynamic Virtual Port Attributes */
+
+/* Private Virtual Port Attributes */
+
+fc_private_vport_rd_enum_attr(vport_state, FC_VPORTSTATE_MAX_NAMELEN);
+fc_private_vport_rd_enum_attr(vport_last_state, FC_VPORTSTATE_MAX_NAMELEN);
+fc_private_vport_rd_attr_cast(node_name, "0x%llx\n", 20, unsigned long long);
+fc_private_vport_rd_attr_cast(port_name, "0x%llx\n", 20, unsigned long long);
+
+static ssize_t
+show_fc_vport_roles (struct class_device *cdev, char *buf)
+{
+       struct fc_vport *vport = transport_class_to_vport(cdev);
+
+       if (vport->roles == FC_PORT_ROLE_UNKNOWN)
+               return snprintf(buf, 20, "unknown\n");
+       return get_fc_port_roles_names(vport->roles, buf);
+}
+static FC_CLASS_DEVICE_ATTR(vport, roles, S_IRUGO, show_fc_vport_roles, NULL);
+
+fc_private_vport_rd_enum_attr(vport_type, FC_PORTTYPE_MAX_NAMELEN);
+
+fc_private_vport_show_function(symbolic_name, "%s\n",
+               FC_VPORT_SYMBOLIC_NAMELEN + 1, )
+fc_vport_store_str_function(symbolic_name, FC_VPORT_SYMBOLIC_NAMELEN)
+static FC_CLASS_DEVICE_ATTR(vport, symbolic_name, S_IRUGO | S_IWUSR,
+               show_fc_vport_symbolic_name, store_fc_vport_symbolic_name);
+
+static ssize_t
+store_fc_vport_delete(struct class_device *cdev, const char *buf,
+                          size_t count)
+{
+       struct fc_vport *vport = transport_class_to_vport(cdev);
+       int stat;
+
+       stat = fc_vport_terminate(vport);
+       if (stat)
+               return stat;
+
+       return count;
+}
+static FC_CLASS_DEVICE_ATTR(vport, vport_delete, S_IWUSR,
+                       NULL, store_fc_vport_delete);
+
+
+/*
+ * Enable/Disable vport
+ *  Write "1" to disable, write "0" to enable
+ */
+static ssize_t
+store_fc_vport_disable(struct class_device *cdev, const char *buf,
+                          size_t count)
+{
+       struct fc_vport *vport = transport_class_to_vport(cdev);
+       struct Scsi_Host *shost = vport_to_shost(vport);
+       struct fc_internal *i = to_fc_internal(shost->transportt);
+       int stat;
+
+       if (vport->flags & (FC_VPORT_DEL | FC_VPORT_CREATING))
+               return -EBUSY;
+
+       if (*buf == '0') {
+               if (vport->vport_state != FC_VPORT_DISABLED)
+                       return -EALREADY;
+       } else if (*buf == '1') {
+               if (vport->vport_state == FC_VPORT_DISABLED)
+                       return -EALREADY;
+       } else
+               return -EINVAL;
+
+       stat = i->f->vport_disable(vport, ((*buf == '0') ? false : true));
+       return stat ? stat : count;
+}
+static FC_CLASS_DEVICE_ATTR(vport, vport_disable, S_IWUSR,
+                       NULL, store_fc_vport_disable);
+
+
 /*
  * Host Attribute Management
  */
@@ -1003,6 +1334,13 @@ static FC_CLASS_DEVICE_ATTR(host, title, S_IRUGO, show_fc_host_##title, NULL)
        if (i->f->show_host_##field)                                    \
                count++
 
+#define SETUP_HOST_ATTRIBUTE_RD_NS(field)                              \
+       i->private_host_attrs[count] = class_device_attr_host_##field;  \
+       i->private_host_attrs[count].attr.mode = S_IRUGO;               \
+       i->private_host_attrs[count].store = NULL;                      \
+       i->host_attrs[count] = &i->private_host_attrs[count];           \
+       count++
+
 #define SETUP_HOST_ATTRIBUTE_RW(field)                                 \
        i->private_host_attrs[count] = class_device_attr_host_##field;  \
        if (!i->f->set_host_##field) {                                  \
@@ -1090,6 +1428,7 @@ fc_private_host_rd_attr_cast(port_name, "0x%llx\n", 20, unsigned long long);
 fc_private_host_rd_attr_cast(permanent_port_name, "0x%llx\n", 20,
                             unsigned long long);
 fc_private_host_rd_attr(maxframe_size, "%u bytes\n", 20);
+fc_private_host_rd_attr(max_npiv_vports, "%u\n", 20);
 fc_private_host_rd_attr(serial_number, "%s\n", (FC_SERIAL_NUMBER_SIZE +1));
 
 
@@ -1210,6 +1549,9 @@ store_fc_private_host_issue_lip(struct class_device *cdev,
 static FC_CLASS_DEVICE_ATTR(host, issue_lip, S_IWUSR, NULL,
                        store_fc_private_host_issue_lip);
 
+fc_private_host_rd_attr(npiv_vports_inuse, "%u\n", 20);
+
+
 /*
  * Host Statistics Management
  */
@@ -1285,7 +1627,6 @@ fc_reset_statistics(struct class_device *cdev, const char *buf,
 static FC_CLASS_DEVICE_ATTR(host, reset_statistics, S_IWUSR, NULL,
                                fc_reset_statistics);
 
-
 static struct attribute *fc_statistics_attrs[] = {
        &class_device_attr_host_seconds_since_last_reset.attr,
        &class_device_attr_host_tx_frames.attr,
@@ -1316,6 +1657,142 @@ static struct attribute_group fc_statistics_group = {
        .attrs = fc_statistics_attrs,
 };
 
+
+/* Host Vport Attributes */
+
+static int
+fc_parse_wwn(const char *ns, u64 *nm)
+{
+       unsigned int i, j;
+       u8 wwn[8];
+
+       memset(wwn, 0, sizeof(wwn));
+
+       /* Validate and store the new name */
+       for (i=0, j=0; i < 16; i++) {
+               if ((*ns >= 'a') && (*ns <= 'f'))
+                       j = ((j << 4) | ((*ns++ -'a') + 10));
+               else if ((*ns >= 'A') && (*ns <= 'F'))
+                       j = ((j << 4) | ((*ns++ -'A') + 10));
+               else if ((*ns >= '0') && (*ns <= '9'))
+                       j = ((j << 4) | (*ns++ -'0'));
+               else
+                       return -EINVAL;
+               if (i % 2) {
+                       wwn[i/2] = j & 0xff;
+                       j = 0;
+               }
+       }
+
+       *nm = wwn_to_u64(wwn);
+
+       return 0;
+}
+
+
+/*
+ * "Short-cut" sysfs variable to create a new vport on a FC Host.
+ * Input is a string of the form "<WWPN>:<WWNN>". Other attributes
+ * will default to a NPIV-based FCP_Initiator; The WWNs are specified
+ * as hex characters, and may *not* contain any prefixes (e.g. 0x, x, etc)
+ */
+static ssize_t
+store_fc_host_vport_create(struct class_device *cdev, const char *buf,
+                          size_t count)
+{
+       struct Scsi_Host *shost = transport_class_to_shost(cdev);
+       struct fc_vport_identifiers vid;
+       struct fc_vport *vport;
+       unsigned int cnt=count;
+       int stat;
+
+       memset(&vid, 0, sizeof(vid));
+
+       /* count may include a LF at end of string */
+       if (buf[cnt-1] == '\n')
+               cnt--;
+
+       /* validate we have enough characters for WWPN */
+       if ((cnt != (16+1+16)) || (buf[16] != ':'))
+               return -EINVAL;
+
+       stat = fc_parse_wwn(&buf[0], &vid.port_name);
+       if (stat)
+               return stat;
+
+       stat = fc_parse_wwn(&buf[17], &vid.node_name);
+       if (stat)
+               return stat;
+
+       vid.roles = FC_PORT_ROLE_FCP_INITIATOR;
+       vid.vport_type = FC_PORTTYPE_NPIV;
+       /* vid.symbolic_name is already zero/NULL's */
+       vid.disable = false;            /* always enabled */
+
+       /* we only allow support on Channel 0 !!! */
+       stat = fc_vport_create(shost, 0, &shost->shost_gendev, &vid, &vport);
+       return stat ? stat : count;
+}
+static FC_CLASS_DEVICE_ATTR(host, vport_create, S_IWUSR, NULL,
+                       store_fc_host_vport_create);
+
+
+/*
+ * "Short-cut" sysfs variable to delete a vport on a FC Host.
+ * Vport is identified by a string containing "<WWPN>:<WWNN>".
+ * The WWNs are specified as hex characters, and may *not* contain
+ * any prefixes (e.g. 0x, x, etc)
+ */
+static ssize_t
+store_fc_host_vport_delete(struct class_device *cdev, const char *buf,
+                          size_t count)
+{
+       struct Scsi_Host *shost = transport_class_to_shost(cdev);
+       struct fc_host_attrs *fc_host = shost_to_fc_host(shost);
+       struct fc_vport *vport;
+       u64 wwpn, wwnn;
+       unsigned long flags;
+       unsigned int cnt=count;
+       int stat, match;
+
+       /* count may include a LF at end of string */
+       if (buf[cnt-1] == '\n')
+               cnt--;
+
+       /* validate we have enough characters for WWPN */
+       if ((cnt != (16+1+16)) || (buf[16] != ':'))
+               return -EINVAL;
+
+       stat = fc_parse_wwn(&buf[0], &wwpn);
+       if (stat)
+               return stat;
+
+       stat = fc_parse_wwn(&buf[17], &wwnn);
+       if (stat)
+               return stat;
+
+       spin_lock_irqsave(shost->host_lock, flags);
+       match = 0;
+       /* we only allow support on Channel 0 !!! */
+       list_for_each_entry(vport, &fc_host->vports, peers) {
+               if ((vport->channel == 0) &&
+                   (vport->port_name == wwpn) && (vport->node_name == wwnn)) {
+                       match = 1;
+                       break;
+               }
+       }
+       spin_unlock_irqrestore(shost->host_lock, flags);
+
+       if (!match)
+               return -ENODEV;
+
+       stat = fc_vport_terminate(vport);
+       return stat ? stat : count;
+}
+static FC_CLASS_DEVICE_ATTR(host, vport_delete, S_IWUSR, NULL,
+                       store_fc_host_vport_delete);
+
+
 static int fc_host_match(struct attribute_container *cont,
                          struct device *dev)
 {
@@ -1387,6 +1864,40 @@ static int fc_rport_match(struct attribute_container *cont,
 }
 
 
+static void fc_vport_dev_release(struct device *dev)
+{
+       struct fc_vport *vport = dev_to_vport(dev);
+       put_device(dev->parent);                /* release kobj parent */
+       kfree(vport);
+}
+
+int scsi_is_fc_vport(const struct device *dev)
+{
+       return dev->release == fc_vport_dev_release;
+}
+EXPORT_SYMBOL(scsi_is_fc_vport);
+
+static int fc_vport_match(struct attribute_container *cont,
+                           struct device *dev)
+{
+       struct fc_vport *vport;
+       struct Scsi_Host *shost;
+       struct fc_internal *i;
+
+       if (!scsi_is_fc_vport(dev))
+               return 0;
+       vport = dev_to_vport(dev);
+
+       shost = vport_to_shost(vport);
+       if (!shost->transportt  || shost->transportt->host_attrs.ac.class
+           != &fc_host_class.class)
+               return 0;
+
+       i = to_fc_internal(shost->transportt);
+       return &i->vport_attr_cont.ac == cont;
+}
+
+
 /**
  * fc_timed_out - FC Transport I/O timeout intercept handler
  *
@@ -1472,6 +1983,11 @@ fc_attach_transport(struct fc_function_template *ft)
        i->rport_attr_cont.ac.match = fc_rport_match;
        transport_container_register(&i->rport_attr_cont);
 
+       i->vport_attr_cont.ac.attrs = &i->vport_attrs[0];
+       i->vport_attr_cont.ac.class = &fc_vport_class.class;
+       i->vport_attr_cont.ac.match = fc_vport_match;
+       transport_container_register(&i->vport_attr_cont);
+
        i->f = ft;
 
        /* Transport uses the shost workq for scsi scanning */
@@ -1505,6 +2021,10 @@ fc_attach_transport(struct fc_function_template *ft)
        SETUP_HOST_ATTRIBUTE_RD(supported_fc4s);
        SETUP_HOST_ATTRIBUTE_RD(supported_speeds);
        SETUP_HOST_ATTRIBUTE_RD(maxframe_size);
+       if (ft->vport_create) {
+               SETUP_HOST_ATTRIBUTE_RD_NS(max_npiv_vports);
+               SETUP_HOST_ATTRIBUTE_RD_NS(npiv_vports_inuse);
+       }
        SETUP_HOST_ATTRIBUTE_RD(serial_number);
 
        SETUP_HOST_ATTRIBUTE_RD(port_id);
@@ -1520,6 +2040,10 @@ fc_attach_transport(struct fc_function_template *ft)
        SETUP_PRIVATE_HOST_ATTRIBUTE_RW(tgtid_bind_type);
        if (ft->issue_fc_host_lip)
                SETUP_PRIVATE_HOST_ATTRIBUTE_RW(issue_lip);
+       if (ft->vport_create)
+               SETUP_PRIVATE_HOST_ATTRIBUTE_RW(vport_create);
+       if (ft->vport_delete)
+               SETUP_PRIVATE_HOST_ATTRIBUTE_RW(vport_delete);
 
        BUG_ON(count > FC_HOST_NUM_ATTRS);
 
@@ -1545,6 +2069,24 @@ fc_attach_transport(struct fc_function_template *ft)
 
        i->rport_attrs[count] = NULL;
 
+       /*
+        * Setup Virtual Port Attributes.
+        */
+       count=0;
+       SETUP_PRIVATE_VPORT_ATTRIBUTE_RD(vport_state);
+       SETUP_PRIVATE_VPORT_ATTRIBUTE_RD(vport_last_state);
+       SETUP_PRIVATE_VPORT_ATTRIBUTE_RD(node_name);
+       SETUP_PRIVATE_VPORT_ATTRIBUTE_RD(port_name);
+       SETUP_PRIVATE_VPORT_ATTRIBUTE_RD(roles);
+       SETUP_PRIVATE_VPORT_ATTRIBUTE_RD(vport_type);
+       SETUP_VPORT_ATTRIBUTE_RW(symbolic_name);
+       SETUP_VPORT_ATTRIBUTE_WR(vport_delete);
+       SETUP_VPORT_ATTRIBUTE_WR(vport_disable);
+
+       BUG_ON(count > FC_VPORT_NUM_ATTRS);
+
+       i->vport_attrs[count] = NULL;
+
        return &i->t;
 }
 EXPORT_SYMBOL(fc_attach_transport);
@@ -1556,6 +2098,7 @@ void fc_release_transport(struct scsi_transport_template *t)
        transport_container_unregister(&i->t.target_attrs);
        transport_container_unregister(&i->t.host_attrs);
        transport_container_unregister(&i->rport_attr_cont);
+       transport_container_unregister(&i->vport_attr_cont);
 
        kfree(i);
 }
@@ -1667,9 +2210,28 @@ fc_flush_devloss(struct Scsi_Host *shost)
 void
 fc_remove_host(struct Scsi_Host *shost)
 {
-       struct fc_rport *rport, *next_rport;
+       struct fc_vport *vport = NULL, *next_vport = NULL;
+       struct fc_rport *rport = NULL, *next_rport = NULL;
        struct workqueue_struct *work_q;
        struct fc_host_attrs *fc_host = shost_to_fc_host(shost);
+       unsigned long flags;
+       int stat;
+
+       spin_lock_irqsave(shost->host_lock, flags);
+
+       /* Remove any vports */
+       list_for_each_entry_safe(vport, next_vport, &fc_host->vports, peers) {
+               spin_unlock_irqrestore(shost->host_lock, flags);
+               /* this must be called synchronously */
+               stat = fc_vport_terminate(vport);
+               spin_lock_irqsave(shost->host_lock, flags);
+               if (stat)
+                       dev_printk(KERN_ERR, vport->dev.parent,
+                               "%s: %s could not be deleted created via "
+                               "shost%d channel %d\n", __FUNCTION__,
+                               vport->dev.bus_id, vport->shost->host_no,
+                               vport->channel);
+       }
 
        /* Remove any remote ports */
        list_for_each_entry_safe(rport, next_rport,
@@ -1686,6 +2248,8 @@ fc_remove_host(struct Scsi_Host *shost)
                fc_queue_work(shost, &rport->rport_delete_work);
        }
 
+       spin_unlock_irqrestore(shost->host_lock, flags);
+
        /* flush all scan work items */
        scsi_flush_work(shost);
 
@@ -1844,7 +2408,7 @@ fc_rport_create(struct Scsi_Host *shost, int channel,
        spin_lock_irqsave(shost->host_lock, flags);
 
        rport->number = fc_host->next_rport_number++;
-       if (rport->roles & FC_RPORT_ROLE_FCP_TARGET)
+       if (rport->roles & FC_PORT_ROLE_FCP_TARGET)
                rport->scsi_target_id = fc_host->next_target_id++;
        else
                rport->scsi_target_id = -1;
@@ -1869,7 +2433,7 @@ fc_rport_create(struct Scsi_Host *shost, int channel,
        transport_add_device(dev);
        transport_configure_device(dev);
 
-       if (rport->roles & FC_RPORT_ROLE_FCP_TARGET) {
+       if (rport->roles & FC_PORT_ROLE_FCP_TARGET) {
                /* initiate a scan of the target */
                rport->flags |= FC_RPORT_SCAN_PENDING;
                scsi_queue_work(shost, &rport->scan_work);
@@ -2003,7 +2567,7 @@ fc_remote_port_add(struct Scsi_Host *shost, int channel,
 
                                /* was a target, not in roles */
                                if ((rport->scsi_target_id != -1) &&
-                                   (!(ids->roles & FC_RPORT_ROLE_FCP_TARGET)))
+                                   (!(ids->roles & FC_PORT_ROLE_FCP_TARGET)))
                                        return rport;
 
                                /*
@@ -2086,7 +2650,7 @@ fc_remote_port_add(struct Scsi_Host *shost, int channel,
                                memset(rport->dd_data, 0,
                                                fci->f->dd_fcrport_size);
 
-                       if (rport->roles & FC_RPORT_ROLE_FCP_TARGET) {
+                       if (rport->roles & FC_PORT_ROLE_FCP_TARGET) {
                                /* initiate a scan of the target */
                                rport->flags |= FC_RPORT_SCAN_PENDING;
                                scsi_queue_work(shost, &rport->scan_work);
@@ -2243,11 +2807,11 @@ fc_remote_port_rolechg(struct fc_rport  *rport, u32 roles)
        int create = 0;
 
        spin_lock_irqsave(shost->host_lock, flags);
-       if (roles & FC_RPORT_ROLE_FCP_TARGET) {
+       if (roles & FC_PORT_ROLE_FCP_TARGET) {
                if (rport->scsi_target_id == -1) {
                        rport->scsi_target_id = fc_host->next_target_id++;
                        create = 1;
-               } else if (!(rport->roles & FC_RPORT_ROLE_FCP_TARGET))
+               } else if (!(rport->roles & FC_PORT_ROLE_FCP_TARGET))
                        create = 1;
        }
 
@@ -2317,7 +2881,7 @@ fc_timeout_deleted_rport(struct work_struct *work)
         */
        if ((rport->port_state == FC_PORTSTATE_ONLINE) &&
            (rport->scsi_target_id != -1) &&
-           !(rport->roles & FC_RPORT_ROLE_FCP_TARGET)) {
+           !(rport->roles & FC_PORT_ROLE_FCP_TARGET)) {
                dev_printk(KERN_ERR, &rport->dev,
                        "blocked FC remote port time out: no longer"
                        " a FCP target, removing starget\n");
@@ -2367,7 +2931,7 @@ fc_timeout_deleted_rport(struct work_struct *work)
         */
        rport->maxframe_size = -1;
        rport->supported_classes = FC_COS_UNSPECIFIED;
-       rport->roles = FC_RPORT_ROLE_UNKNOWN;
+       rport->roles = FC_PORT_ROLE_UNKNOWN;
        rport->port_state = FC_PORTSTATE_NOTPRESENT;
 
        /* remove the identifiers that aren't used in the consisting binding */
@@ -2436,7 +3000,7 @@ fc_scsi_scan_rport(struct work_struct *work)
        unsigned long flags;
 
        if ((rport->port_state == FC_PORTSTATE_ONLINE) &&
-           (rport->roles & FC_RPORT_ROLE_FCP_TARGET)) {
+           (rport->roles & FC_PORT_ROLE_FCP_TARGET)) {
                scsi_scan_target(&rport->dev, rport->channel,
                        rport->scsi_target_id, SCAN_WILD_CARD, 1);
        }
@@ -2447,6 +3011,203 @@ fc_scsi_scan_rport(struct work_struct *work)
 }
 
 
+/**
+ * fc_vport_create - allocates and creates a FC virtual port.
+ * @shost:     scsi host the virtual port is connected to.
+ * @channel:   Channel on shost port connected to.
+ * @pdev:      parent device for vport
+ * @ids:       The world wide names, FC4 port roles, etc for
+ *              the virtual port.
+ * @ret_vport: The pointer to the created vport.
+ *
+ * Allocates and creates the vport structure, calls the parent host
+ * to instantiate the vport, the completes w/ class and sysfs creation.
+ *
+ * Notes:
+ *     This routine assumes no locks are held on entry.
+ **/
+static int
+fc_vport_create(struct Scsi_Host *shost, int channel, struct device *pdev,
+       struct fc_vport_identifiers  *ids, struct fc_vport **ret_vport)
+{
+       struct fc_host_attrs *fc_host = shost_to_fc_host(shost);
+       struct fc_internal *fci = to_fc_internal(shost->transportt);
+       struct fc_vport *vport;
+       struct device *dev;
+       unsigned long flags;
+       size_t size;
+       int error;
+
+       *ret_vport = NULL;
+
+       if ( ! fci->f->vport_create)
+               return -ENOENT;
+
+       size = (sizeof(struct fc_vport) + fci->f->dd_fcvport_size);
+       vport = kzalloc(size, GFP_KERNEL);
+       if (unlikely(!vport)) {
+               printk(KERN_ERR "%s: allocation failure\n", __FUNCTION__);
+               return -ENOMEM;
+       }
+
+       vport->vport_state = FC_VPORT_UNKNOWN;
+       vport->vport_last_state = FC_VPORT_UNKNOWN;
+       vport->node_name = ids->node_name;
+       vport->port_name = ids->port_name;
+       vport->roles = ids->roles;
+       vport->vport_type = ids->vport_type;
+       if (fci->f->dd_fcvport_size)
+               vport->dd_data = &vport[1];
+       vport->shost = shost;
+       vport->channel = channel;
+       vport->flags = FC_VPORT_CREATING;
+
+       spin_lock_irqsave(shost->host_lock, flags);
+
+       if (fc_host->npiv_vports_inuse >= fc_host->max_npiv_vports) {
+               spin_unlock_irqrestore(shost->host_lock, flags);
+               kfree(vport);
+               return -ENOSPC;
+       }
+       fc_host->npiv_vports_inuse++;
+       vport->number = fc_host->next_vport_number++;
+       list_add_tail(&vport->peers, &fc_host->vports);
+       get_device(&shost->shost_gendev);       /* for fc_host->vport list */
+
+       spin_unlock_irqrestore(shost->host_lock, flags);
+
+       dev = &vport->dev;
+       device_initialize(dev);                 /* takes self reference */
+       dev->parent = get_device(pdev);         /* takes parent reference */
+       dev->release = fc_vport_dev_release;
+       sprintf(dev->bus_id, "vport-%d:%d-%d",
+               shost->host_no, channel, vport->number);
+       transport_setup_device(dev);
+
+       error = device_add(dev);
+       if (error) {
+               printk(KERN_ERR "FC Virtual Port device_add failed\n");
+               goto delete_vport;
+       }
+       transport_add_device(dev);
+       transport_configure_device(dev);
+
+       error = fci->f->vport_create(vport, ids->disable);
+       if (error) {
+               printk(KERN_ERR "FC Virtual Port LLDD Create failed\n");
+               goto delete_vport_all;
+       }
+
+       /*
+        * if the parent isn't the physical adapter's Scsi_Host, ensure
+        * the Scsi_Host at least contains ia symlink to the vport.
+        */
+       if (pdev != &shost->shost_gendev) {
+               error = sysfs_create_link(&shost->shost_gendev.kobj,
+                                &dev->kobj, dev->bus_id);
+               if (error)
+                       printk(KERN_ERR
+                               "%s: Cannot create vport symlinks for "
+                               "%s, err=%d\n",
+                               __FUNCTION__, dev->bus_id, error);
+       }
+       spin_lock_irqsave(shost->host_lock, flags);
+       vport->flags &= ~FC_VPORT_CREATING;
+       spin_unlock_irqrestore(shost->host_lock, flags);
+
+       dev_printk(KERN_NOTICE, pdev,
+                       "%s created via shost%d channel %d\n", dev->bus_id,
+                       shost->host_no, channel);
+
+       *ret_vport = vport;
+
+       return 0;
+
+delete_vport_all:
+       transport_remove_device(dev);
+       device_del(dev);
+delete_vport:
+       transport_destroy_device(dev);
+       spin_lock_irqsave(shost->host_lock, flags);
+       list_del(&vport->peers);
+       put_device(&shost->shost_gendev);       /* for fc_host->vport list */
+       fc_host->npiv_vports_inuse--;
+       spin_unlock_irqrestore(shost->host_lock, flags);
+       put_device(dev->parent);
+       kfree(vport);
+
+       return error;
+}
+
+
+/**
+ * fc_vport_terminate - Admin App or LLDD requests termination of a vport
+ * @vport:     fc_vport to be terminated
+ *
+ * Calls the LLDD vport_delete() function, then deallocates and removes
+ * the vport from the shost and object tree.
+ *
+ * Notes:
+ *     This routine assumes no locks are held on entry.
+ **/
+int
+fc_vport_terminate(struct fc_vport *vport)
+{
+       struct Scsi_Host *shost = vport_to_shost(vport);
+       struct fc_host_attrs *fc_host = shost_to_fc_host(shost);
+       struct fc_internal *i = to_fc_internal(shost->transportt);
+       struct device *dev = &vport->dev;
+       unsigned long flags;
+       int stat;
+
+       spin_lock_irqsave(shost->host_lock, flags);
+       if (vport->flags & FC_VPORT_CREATING) {
+               spin_unlock_irqrestore(shost->host_lock, flags);
+               return -EBUSY;
+       }
+       if (vport->flags & (FC_VPORT_DEL)) {
+               spin_unlock_irqrestore(shost->host_lock, flags);
+               return -EALREADY;
+       }
+       vport->flags |= FC_VPORT_DELETING;
+       spin_unlock_irqrestore(shost->host_lock, flags);
+
+       if (i->f->vport_delete)
+               stat = i->f->vport_delete(vport);
+       else
+               stat = -ENOENT;
+
+       spin_lock_irqsave(shost->host_lock, flags);
+       vport->flags &= ~FC_VPORT_DELETING;
+       if (!stat) {
+               vport->flags |= FC_VPORT_DELETED;
+               list_del(&vport->peers);
+               fc_host->npiv_vports_inuse--;
+               put_device(&shost->shost_gendev);  /* for fc_host->vport list */
+       }
+       spin_unlock_irqrestore(shost->host_lock, flags);
+
+       if (stat)
+               return stat;
+
+       if (dev->parent != &shost->shost_gendev)
+               sysfs_remove_link(&shost->shost_gendev.kobj, dev->bus_id);
+       transport_remove_device(dev);
+       device_del(dev);
+       transport_destroy_device(dev);
+
+       /*
+        * Removing our self-reference should mean our
+        * release function gets called, which will drop the remaining
+        * parent reference and free the data structure.
+        */
+       put_device(dev);                        /* for self-reference */
+
+       return 0; /* SUCCESS */
+}
+EXPORT_SYMBOL(fc_vport_terminate);
+
+
 MODULE_AUTHOR("Martin Hicks");
 MODULE_DESCRIPTION("FC Transport Attributes");
 MODULE_LICENSE("GPL");
index 1e797308640a2801456efb131d9cb135415ef8af..81ea7b4bf81ee99746520d4c353e3dc9330bbcfc 100644 (file)
@@ -19,7 +19,7 @@
  *
  *  ========
  *
- *  Copyright (C) 2004-2005   James Smart, Emulex Corporation
+ *  Copyright (C) 2004-2007   James Smart, Emulex Corporation
  *    Rewrite for host, target, device, and remote port attributes,
  *    statistics, and service functions...
  *
@@ -62,8 +62,10 @@ enum fc_port_type {
        FC_PORTTYPE_NLPORT,             /* (Public) Loop w/ FLPort */
        FC_PORTTYPE_LPORT,              /* (Private) Loop w/o FLPort */
        FC_PORTTYPE_PTP,                /* Point to Point w/ another NPort */
+       FC_PORTTYPE_NPIV,               /* VPORT based on NPIV */
 };
 
+
 /*
  * fc_port_state: If you alter this, you also need to alter scsi_transport_fc.c
  * (for the ascii descriptions).
@@ -83,6 +85,25 @@ enum fc_port_state {
 };
 
 
+/*
+ * fc_vport_state: If you alter this, you also need to alter
+ * scsi_transport_fc.c (for the ascii descriptions).
+ */
+enum fc_vport_state {
+       FC_VPORT_UNKNOWN,
+       FC_VPORT_ACTIVE,
+       FC_VPORT_DISABLED,
+       FC_VPORT_LINKDOWN,
+       FC_VPORT_INITIALIZING,
+       FC_VPORT_NO_FABRIC_SUPP,
+       FC_VPORT_NO_FABRIC_RSCS,
+       FC_VPORT_FABRIC_LOGOUT,
+       FC_VPORT_FABRIC_REJ_WWN,
+       FC_VPORT_FAILED,
+};
+
+
+
 /* 
  * FC Classes of Service
  * Note: values are not enumerated, as they can be "or'd" together
@@ -124,17 +145,114 @@ enum fc_tgtid_binding_type  {
 };
 
 /*
- * FC Remote Port Roles
+ * FC Port Roles
  * Note: values are not enumerated, as they can be "or'd" together
  * for reporting (e.g. report roles). If you alter this list,
  * you also need to alter scsi_transport_fc.c (for the ascii descriptions).
  */
-#define FC_RPORT_ROLE_UNKNOWN                  0x00
-#define FC_RPORT_ROLE_FCP_TARGET               0x01
-#define FC_RPORT_ROLE_FCP_INITIATOR            0x02
-#define FC_RPORT_ROLE_IP_PORT                  0x04
+#define FC_PORT_ROLE_UNKNOWN                   0x00
+#define FC_PORT_ROLE_FCP_TARGET                        0x01
+#define FC_PORT_ROLE_FCP_INITIATOR             0x02
+#define FC_PORT_ROLE_IP_PORT                   0x04
+
+/* The following are for compatibility */
+#define FC_RPORT_ROLE_UNKNOWN                  FC_PORT_ROLE_UNKNOWN
+#define FC_RPORT_ROLE_FCP_TARGET               FC_PORT_ROLE_FCP_TARGET
+#define FC_RPORT_ROLE_FCP_INITIATOR            FC_PORT_ROLE_FCP_INITIATOR
+#define FC_RPORT_ROLE_IP_PORT                  FC_PORT_ROLE_IP_PORT
+
+
+/* Macro for use in defining Virtual Port attributes */
+#define FC_VPORT_ATTR(_name,_mode,_show,_store)                                \
+struct class_device_attribute class_device_attr_vport_##_name =        \
+       __ATTR(_name,_mode,_show,_store)
 
 
+/*
+ * FC Virtual Port Attributes
+ *
+ * This structure exists for each FC port is a virtual FC port. Virtual
+ * ports share the physical link with the Physical port. Each virtual
+ * ports has a unique presense on the SAN, and may be instantiated via
+ * NPIV, Virtual Fabrics, or via additional ALPAs. As the vport is a
+ * unique presense, each vport has it's own view of the fabric,
+ * authentication priviledge, and priorities.
+ *
+ * A virtual port may support 1 or more FC4 roles. Typically it is a
+ * FCP Initiator. It could be a FCP Target, or exist sole for an IP over FC
+ * roles. FC port attributes for the vport will be reported on any
+ * fc_host class object allocated for an FCP Initiator.
+ *
+ * --
+ *
+ * Fixed attributes are not expected to change. The driver is
+ * expected to set these values after receiving the fc_vport structure
+ * via the vport_create() call from the transport.
+ * The transport fully manages all get functions w/o driver interaction.
+ *
+ * Dynamic attributes are expected to change. The driver participates
+ * in all get/set operations via functions provided by the driver.
+ *
+ * Private attributes are transport-managed values. They are fully
+ * managed by the transport w/o driver interaction.
+ */
+
+#define FC_VPORT_SYMBOLIC_NAMELEN              64
+struct fc_vport {
+       /* Fixed Attributes */
+
+       /* Dynamic Attributes */
+
+       /* Private (Transport-managed) Attributes */
+       enum fc_vport_state vport_state;
+       enum fc_vport_state vport_last_state;
+       u64 node_name;
+       u64 port_name;
+       u32 roles;
+       u32 vport_id;           /* Admin Identifier for the vport */
+       enum fc_port_type vport_type;
+       char symbolic_name[FC_VPORT_SYMBOLIC_NAMELEN];
+
+       /* exported data */
+       void *dd_data;                  /* Used for driver-specific storage */
+
+       /* internal data */
+       struct Scsi_Host *shost;        /* Physical Port Parent */
+       unsigned int channel;
+       u32 number;
+       u8 flags;
+       struct list_head peers;
+       struct device dev;
+} __attribute__((aligned(sizeof(unsigned long))));
+
+/* bit field values for struct fc_vport "flags" field: */
+#define FC_VPORT_CREATING              0x01
+#define FC_VPORT_DELETING              0x02
+#define FC_VPORT_DELETED               0x04
+#define FC_VPORT_DEL                   0x06    /* Any DELETE state */
+
+#define        dev_to_vport(d)                         \
+       container_of(d, struct fc_vport, dev)
+#define transport_class_to_vport(classdev)     \
+       dev_to_vport(classdev->dev)
+#define vport_to_shost(v)                      \
+       (v->shost)
+#define vport_to_shost_channel(v)              \
+       (v->channel)
+#define vport_to_parent(v)                     \
+       (v->dev.parent)
+
+
+/* Error return codes for vport_create() callback */
+#define VPCERR_UNSUPPORTED             -ENOSYS         /* no driver/adapter
+                                                          support */
+#define VPCERR_BAD_WWN                 -ENOTUNIQ       /* driver validation
+                                                          of WWNs failed */
+#define VPCERR_NO_FABRIC_SUPP          -EOPNOTSUPP     /* Fabric connection
+                                                          is loop or the
+                                                          Fabric Port does
+                                                          not support NPIV */
+
 /*
  * fc_rport_identifiers: This set of data contains all elements
  * to uniquely identify a remote FC port. The driver uses this data
@@ -149,6 +267,7 @@ struct fc_rport_identifiers {
        u32 roles;
 };
 
+
 /* Macro for use in defining Remote Port attributes */
 #define FC_RPORT_ATTR(_name,_mode,_show,_store)                                \
 struct class_device_attribute class_device_attr_rport_##_name =        \
@@ -343,6 +462,7 @@ struct fc_host_attrs {
        u8  supported_fc4s[FC_FC4_LIST_SIZE];
        u32 supported_speeds;
        u32 maxframe_size;
+       u16 max_npiv_vports;
        char serial_number[FC_SERIAL_NUMBER_SIZE];
 
        /* Dynamic Attributes */
@@ -361,8 +481,11 @@ struct fc_host_attrs {
        /* internal data */
        struct list_head rports;
        struct list_head rport_bindings;
+       struct list_head vports;
        u32 next_rport_number;
        u32 next_target_id;
+       u32 next_vport_number;
+       u16 npiv_vports_inuse;
 
        /* work queues for rport state manipulation */
        char work_q_name[KOBJ_NAME_LEN];
@@ -388,6 +511,8 @@ struct fc_host_attrs {
        (((struct fc_host_attrs *)(x)->shost_data)->supported_speeds)
 #define fc_host_maxframe_size(x)       \
        (((struct fc_host_attrs *)(x)->shost_data)->maxframe_size)
+#define fc_host_max_npiv_vports(x)     \
+       (((struct fc_host_attrs *)(x)->shost_data)->max_npiv_vports)
 #define fc_host_serial_number(x)       \
        (((struct fc_host_attrs *)(x)->shost_data)->serial_number)
 #define fc_host_port_id(x)     \
@@ -412,10 +537,16 @@ struct fc_host_attrs {
        (((struct fc_host_attrs *)(x)->shost_data)->rports)
 #define fc_host_rport_bindings(x) \
        (((struct fc_host_attrs *)(x)->shost_data)->rport_bindings)
+#define fc_host_vports(x) \
+       (((struct fc_host_attrs *)(x)->shost_data)->vports)
 #define fc_host_next_rport_number(x) \
        (((struct fc_host_attrs *)(x)->shost_data)->next_rport_number)
 #define fc_host_next_target_id(x) \
        (((struct fc_host_attrs *)(x)->shost_data)->next_target_id)
+#define fc_host_next_vport_number(x) \
+       (((struct fc_host_attrs *)(x)->shost_data)->next_vport_number)
+#define fc_host_npiv_vports_inuse(x)   \
+       (((struct fc_host_attrs *)(x)->shost_data)->npiv_vports_inuse)
 #define fc_host_work_q_name(x) \
        (((struct fc_host_attrs *)(x)->shost_data)->work_q_name)
 #define fc_host_work_q(x) \
@@ -452,8 +583,14 @@ struct fc_function_template {
        void    (*dev_loss_tmo_callbk)(struct fc_rport *);
        void    (*terminate_rport_io)(struct fc_rport *);
 
+       void    (*set_vport_symbolic_name)(struct fc_vport *);
+       int     (*vport_create)(struct fc_vport *, bool);
+       int     (*vport_disable)(struct fc_vport *, bool);
+       int     (*vport_delete)(struct fc_vport *);
+
        /* allocation lengths for host-specific data */
        u32                             dd_fcrport_size;
+       u32                             dd_fcvport_size;
 
        /* 
         * The driver sets these to tell the transport class it
@@ -512,7 +649,7 @@ fc_remote_port_chkready(struct fc_rport *rport)
 
        switch (rport->port_state) {
        case FC_PORTSTATE_ONLINE:
-               if (rport->roles & FC_RPORT_ROLE_FCP_TARGET)
+               if (rport->roles & FC_PORT_ROLE_FCP_TARGET)
                        result = 0;
                else if (rport->flags & FC_RPORT_DEVLOSS_PENDING)
                        result = DID_IMM_RETRY << 16;
@@ -549,6 +686,27 @@ static inline void u64_to_wwn(u64 inm, u8 *wwn)
        wwn[7] = inm & 0xff;
 }
 
+/**
+ * fc_vport_set_state() - called to set a vport's state. Saves the old state,
+ *   excepting the transitory states of initializing and sending the ELS
+ *   traffic to instantiate the vport on the link.
+ *
+ * Assumes the driver has surrounded this with the proper locking to ensure
+ * a coherent state change.
+ *
+ * @vport:     virtual port whose state is changing
+ * @new_state:  new state
+ **/
+static inline void
+fc_vport_set_state(struct fc_vport *vport, enum fc_vport_state new_state)
+{
+       if ((new_state != FC_VPORT_UNKNOWN) &&
+           (new_state != FC_VPORT_INITIALIZING))
+               vport->vport_last_state = vport->vport_state;
+       vport->vport_state = new_state;
+}
+
+
 struct scsi_transport_template *fc_attach_transport(
                        struct fc_function_template *);
 void fc_release_transport(struct scsi_transport_template *);
@@ -567,5 +725,6 @@ void fc_host_post_vendor_event(struct Scsi_Host *shost, u32 event_number,
         *   be sure to read the Vendor Type and ID formatting requirements
         *   specified in scsi_netlink.h
         */
+int fc_vport_terminate(struct fc_vport *vport);
 
 #endif /* SCSI_TRANSPORT_FC_H */