Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / target / sbp / sbp_target.c
diff --git a/kernel/drivers/target/sbp/sbp_target.c b/kernel/drivers/target/sbp/sbp_target.c
new file mode 100644 (file)
index 0000000..18b0f97
--- /dev/null
@@ -0,0 +1,2566 @@
+/*
+ * SBP2 target driver (SCSI over IEEE1394 in target mode)
+ *
+ * Copyright (C) 2011  Chris Boot <bootc@bootc.net>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+
+#define KMSG_COMPONENT "sbp_target"
+#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/string.h>
+#include <linux/configfs.h>
+#include <linux/ctype.h>
+#include <linux/firewire.h>
+#include <linux/firewire-constants.h>
+#include <scsi/scsi.h>
+#include <scsi/scsi_tcq.h>
+#include <target/target_core_base.h>
+#include <target/target_core_backend.h>
+#include <target/target_core_fabric.h>
+#include <target/target_core_fabric_configfs.h>
+#include <target/target_core_configfs.h>
+#include <target/configfs_macros.h>
+#include <asm/unaligned.h>
+
+#include "sbp_target.h"
+
+static const struct target_core_fabric_ops sbp_ops;
+
+/* FireWire address region for management and command block address handlers */
+static const struct fw_address_region sbp_register_region = {
+       .start  = CSR_REGISTER_BASE + 0x10000,
+       .end    = 0x1000000000000ULL,
+};
+
+static const u32 sbp_unit_directory_template[] = {
+       0x1200609e, /* unit_specifier_id: NCITS/T10 */
+       0x13010483, /* unit_sw_version: 1155D Rev 4 */
+       0x3800609e, /* command_set_specifier_id: NCITS/T10 */
+       0x390104d8, /* command_set: SPC-2 */
+       0x3b000000, /* command_set_revision: 0 */
+       0x3c000001, /* firmware_revision: 1 */
+};
+
+#define SESSION_MAINTENANCE_INTERVAL HZ
+
+static atomic_t login_id = ATOMIC_INIT(0);
+
+static void session_maintenance_work(struct work_struct *);
+static int sbp_run_transaction(struct fw_card *, int, int, int, int,
+               unsigned long long, void *, size_t);
+
+static int read_peer_guid(u64 *guid, const struct sbp_management_request *req)
+{
+       int ret;
+       __be32 high, low;
+
+       ret = sbp_run_transaction(req->card, TCODE_READ_QUADLET_REQUEST,
+                       req->node_addr, req->generation, req->speed,
+                       (CSR_REGISTER_BASE | CSR_CONFIG_ROM) + 3 * 4,
+                       &high, sizeof(high));
+       if (ret != RCODE_COMPLETE)
+               return ret;
+
+       ret = sbp_run_transaction(req->card, TCODE_READ_QUADLET_REQUEST,
+                       req->node_addr, req->generation, req->speed,
+                       (CSR_REGISTER_BASE | CSR_CONFIG_ROM) + 4 * 4,
+                       &low, sizeof(low));
+       if (ret != RCODE_COMPLETE)
+               return ret;
+
+       *guid = (u64)be32_to_cpu(high) << 32 | be32_to_cpu(low);
+
+       return RCODE_COMPLETE;
+}
+
+static struct sbp_session *sbp_session_find_by_guid(
+       struct sbp_tpg *tpg, u64 guid)
+{
+       struct se_session *se_sess;
+       struct sbp_session *sess, *found = NULL;
+
+       spin_lock_bh(&tpg->se_tpg.session_lock);
+       list_for_each_entry(se_sess, &tpg->se_tpg.tpg_sess_list, sess_list) {
+               sess = se_sess->fabric_sess_ptr;
+               if (sess->guid == guid)
+                       found = sess;
+       }
+       spin_unlock_bh(&tpg->se_tpg.session_lock);
+
+       return found;
+}
+
+static struct sbp_login_descriptor *sbp_login_find_by_lun(
+               struct sbp_session *session, struct se_lun *lun)
+{
+       struct sbp_login_descriptor *login, *found = NULL;
+
+       spin_lock_bh(&session->lock);
+       list_for_each_entry(login, &session->login_list, link) {
+               if (login->lun == lun)
+                       found = login;
+       }
+       spin_unlock_bh(&session->lock);
+
+       return found;
+}
+
+static int sbp_login_count_all_by_lun(
+               struct sbp_tpg *tpg,
+               struct se_lun *lun,
+               int exclusive)
+{
+       struct se_session *se_sess;
+       struct sbp_session *sess;
+       struct sbp_login_descriptor *login;
+       int count = 0;
+
+       spin_lock_bh(&tpg->se_tpg.session_lock);
+       list_for_each_entry(se_sess, &tpg->se_tpg.tpg_sess_list, sess_list) {
+               sess = se_sess->fabric_sess_ptr;
+
+               spin_lock_bh(&sess->lock);
+               list_for_each_entry(login, &sess->login_list, link) {
+                       if (login->lun != lun)
+                               continue;
+
+                       if (!exclusive || login->exclusive)
+                               count++;
+               }
+               spin_unlock_bh(&sess->lock);
+       }
+       spin_unlock_bh(&tpg->se_tpg.session_lock);
+
+       return count;
+}
+
+static struct sbp_login_descriptor *sbp_login_find_by_id(
+       struct sbp_tpg *tpg, int login_id)
+{
+       struct se_session *se_sess;
+       struct sbp_session *sess;
+       struct sbp_login_descriptor *login, *found = NULL;
+
+       spin_lock_bh(&tpg->se_tpg.session_lock);
+       list_for_each_entry(se_sess, &tpg->se_tpg.tpg_sess_list, sess_list) {
+               sess = se_sess->fabric_sess_ptr;
+
+               spin_lock_bh(&sess->lock);
+               list_for_each_entry(login, &sess->login_list, link) {
+                       if (login->login_id == login_id)
+                               found = login;
+               }
+               spin_unlock_bh(&sess->lock);
+       }
+       spin_unlock_bh(&tpg->se_tpg.session_lock);
+
+       return found;
+}
+
+static struct se_lun *sbp_get_lun_from_tpg(struct sbp_tpg *tpg, int lun)
+{
+       struct se_portal_group *se_tpg = &tpg->se_tpg;
+       struct se_lun *se_lun;
+
+       if (lun >= TRANSPORT_MAX_LUNS_PER_TPG)
+               return ERR_PTR(-EINVAL);
+
+       spin_lock(&se_tpg->tpg_lun_lock);
+       se_lun = se_tpg->tpg_lun_list[lun];
+
+       if (se_lun->lun_status != TRANSPORT_LUN_STATUS_ACTIVE)
+               se_lun = ERR_PTR(-ENODEV);
+
+       spin_unlock(&se_tpg->tpg_lun_lock);
+
+       return se_lun;
+}
+
+static struct sbp_session *sbp_session_create(
+               struct sbp_tpg *tpg,
+               u64 guid)
+{
+       struct sbp_session *sess;
+       int ret;
+       char guid_str[17];
+       struct se_node_acl *se_nacl;
+
+       sess = kmalloc(sizeof(*sess), GFP_KERNEL);
+       if (!sess) {
+               pr_err("failed to allocate session descriptor\n");
+               return ERR_PTR(-ENOMEM);
+       }
+
+       sess->se_sess = transport_init_session(TARGET_PROT_NORMAL);
+       if (IS_ERR(sess->se_sess)) {
+               pr_err("failed to init se_session\n");
+
+               ret = PTR_ERR(sess->se_sess);
+               kfree(sess);
+               return ERR_PTR(ret);
+       }
+
+       snprintf(guid_str, sizeof(guid_str), "%016llx", guid);
+
+       se_nacl = core_tpg_check_initiator_node_acl(&tpg->se_tpg, guid_str);
+       if (!se_nacl) {
+               pr_warn("Node ACL not found for %s\n", guid_str);
+
+               transport_free_session(sess->se_sess);
+               kfree(sess);
+
+               return ERR_PTR(-EPERM);
+       }
+
+       sess->se_sess->se_node_acl = se_nacl;
+
+       spin_lock_init(&sess->lock);
+       INIT_LIST_HEAD(&sess->login_list);
+       INIT_DELAYED_WORK(&sess->maint_work, session_maintenance_work);
+
+       sess->guid = guid;
+
+       transport_register_session(&tpg->se_tpg, se_nacl, sess->se_sess, sess);
+
+       return sess;
+}
+
+static void sbp_session_release(struct sbp_session *sess, bool cancel_work)
+{
+       spin_lock_bh(&sess->lock);
+       if (!list_empty(&sess->login_list)) {
+               spin_unlock_bh(&sess->lock);
+               return;
+       }
+       spin_unlock_bh(&sess->lock);
+
+       if (cancel_work)
+               cancel_delayed_work_sync(&sess->maint_work);
+
+       transport_deregister_session_configfs(sess->se_sess);
+       transport_deregister_session(sess->se_sess);
+
+       if (sess->card)
+               fw_card_put(sess->card);
+
+       kfree(sess);
+}
+
+static void sbp_target_agent_unregister(struct sbp_target_agent *);
+
+static void sbp_login_release(struct sbp_login_descriptor *login,
+       bool cancel_work)
+{
+       struct sbp_session *sess = login->sess;
+
+       /* FIXME: abort/wait on tasks */
+
+       sbp_target_agent_unregister(login->tgt_agt);
+
+       if (sess) {
+               spin_lock_bh(&sess->lock);
+               list_del(&login->link);
+               spin_unlock_bh(&sess->lock);
+
+               sbp_session_release(sess, cancel_work);
+       }
+
+       kfree(login);
+}
+
+static struct sbp_target_agent *sbp_target_agent_register(
+       struct sbp_login_descriptor *);
+
+static void sbp_management_request_login(
+       struct sbp_management_agent *agent, struct sbp_management_request *req,
+       int *status_data_size)
+{
+       struct sbp_tport *tport = agent->tport;
+       struct sbp_tpg *tpg = tport->tpg;
+       struct se_lun *se_lun;
+       int ret;
+       u64 guid;
+       struct sbp_session *sess;
+       struct sbp_login_descriptor *login;
+       struct sbp_login_response_block *response;
+       int login_response_len;
+
+       se_lun = sbp_get_lun_from_tpg(tpg,
+                       LOGIN_ORB_LUN(be32_to_cpu(req->orb.misc)));
+       if (IS_ERR(se_lun)) {
+               pr_notice("login to unknown LUN: %d\n",
+                       LOGIN_ORB_LUN(be32_to_cpu(req->orb.misc)));
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_LUN_NOTSUPP));
+               return;
+       }
+
+       ret = read_peer_guid(&guid, req);
+       if (ret != RCODE_COMPLETE) {
+               pr_warn("failed to read peer GUID: %d\n", ret);
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_TRANSPORT_FAILURE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_UNSPECIFIED_ERROR));
+               return;
+       }
+
+       pr_notice("mgt_agent LOGIN to LUN %d from %016llx\n",
+               se_lun->unpacked_lun, guid);
+
+       sess = sbp_session_find_by_guid(tpg, guid);
+       if (sess) {
+               login = sbp_login_find_by_lun(sess, se_lun);
+               if (login) {
+                       pr_notice("initiator already logged-in\n");
+
+                       /*
+                        * SBP-2 R4 says we should return access denied, but
+                        * that can confuse initiators. Instead we need to
+                        * treat this like a reconnect, but send the login
+                        * response block like a fresh login.
+                        *
+                        * This is required particularly in the case of Apple
+                        * devices booting off the FireWire target, where
+                        * the firmware has an active login to the target. When
+                        * the OS takes control of the session it issues its own
+                        * LOGIN rather than a RECONNECT. To avoid the machine
+                        * waiting until the reconnect_hold expires, we can skip
+                        * the ACCESS_DENIED errors to speed things up.
+                        */
+
+                       goto already_logged_in;
+               }
+       }
+
+       /*
+        * check exclusive bit in login request
+        * reject with access_denied if any logins present
+        */
+       if (LOGIN_ORB_EXCLUSIVE(be32_to_cpu(req->orb.misc)) &&
+                       sbp_login_count_all_by_lun(tpg, se_lun, 0)) {
+               pr_warn("refusing exclusive login with other active logins\n");
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_ACCESS_DENIED));
+               return;
+       }
+
+       /*
+        * check exclusive bit in any existing login descriptor
+        * reject with access_denied if any exclusive logins present
+        */
+       if (sbp_login_count_all_by_lun(tpg, se_lun, 1)) {
+               pr_warn("refusing login while another exclusive login present\n");
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_ACCESS_DENIED));
+               return;
+       }
+
+       /*
+        * check we haven't exceeded the number of allowed logins
+        * reject with resources_unavailable if we have
+        */
+       if (sbp_login_count_all_by_lun(tpg, se_lun, 0) >=
+                       tport->max_logins_per_lun) {
+               pr_warn("max number of logins reached\n");
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_RESOURCES_UNAVAIL));
+               return;
+       }
+
+       if (!sess) {
+               sess = sbp_session_create(tpg, guid);
+               if (IS_ERR(sess)) {
+                       switch (PTR_ERR(sess)) {
+                       case -EPERM:
+                               ret = SBP_STATUS_ACCESS_DENIED;
+                               break;
+                       default:
+                               ret = SBP_STATUS_RESOURCES_UNAVAIL;
+                               break;
+                       }
+
+                       req->status.status = cpu_to_be32(
+                               STATUS_BLOCK_RESP(
+                                       STATUS_RESP_REQUEST_COMPLETE) |
+                               STATUS_BLOCK_SBP_STATUS(ret));
+                       return;
+               }
+
+               sess->node_id = req->node_addr;
+               sess->card = fw_card_get(req->card);
+               sess->generation = req->generation;
+               sess->speed = req->speed;
+
+               schedule_delayed_work(&sess->maint_work,
+                               SESSION_MAINTENANCE_INTERVAL);
+       }
+
+       /* only take the latest reconnect_hold into account */
+       sess->reconnect_hold = min(
+               1 << LOGIN_ORB_RECONNECT(be32_to_cpu(req->orb.misc)),
+               tport->max_reconnect_timeout) - 1;
+
+       login = kmalloc(sizeof(*login), GFP_KERNEL);
+       if (!login) {
+               pr_err("failed to allocate login descriptor\n");
+
+               sbp_session_release(sess, true);
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_RESOURCES_UNAVAIL));
+               return;
+       }
+
+       login->sess = sess;
+       login->lun = se_lun;
+       login->status_fifo_addr = sbp2_pointer_to_addr(&req->orb.status_fifo);
+       login->exclusive = LOGIN_ORB_EXCLUSIVE(be32_to_cpu(req->orb.misc));
+       login->login_id = atomic_inc_return(&login_id);
+
+       login->tgt_agt = sbp_target_agent_register(login);
+       if (IS_ERR(login->tgt_agt)) {
+               ret = PTR_ERR(login->tgt_agt);
+               pr_err("failed to map command block handler: %d\n", ret);
+
+               sbp_session_release(sess, true);
+               kfree(login);
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_RESOURCES_UNAVAIL));
+               return;
+       }
+
+       spin_lock_bh(&sess->lock);
+       list_add_tail(&login->link, &sess->login_list);
+       spin_unlock_bh(&sess->lock);
+
+already_logged_in:
+       response = kzalloc(sizeof(*response), GFP_KERNEL);
+       if (!response) {
+               pr_err("failed to allocate login response block\n");
+
+               sbp_login_release(login, true);
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_RESOURCES_UNAVAIL));
+               return;
+       }
+
+       login_response_len = clamp_val(
+                       LOGIN_ORB_RESPONSE_LENGTH(be32_to_cpu(req->orb.length)),
+                       12, sizeof(*response));
+       response->misc = cpu_to_be32(
+               ((login_response_len & 0xffff) << 16) |
+               (login->login_id & 0xffff));
+       response->reconnect_hold = cpu_to_be32(sess->reconnect_hold & 0xffff);
+       addr_to_sbp2_pointer(login->tgt_agt->handler.offset,
+               &response->command_block_agent);
+
+       ret = sbp_run_transaction(sess->card, TCODE_WRITE_BLOCK_REQUEST,
+               sess->node_id, sess->generation, sess->speed,
+               sbp2_pointer_to_addr(&req->orb.ptr2), response,
+               login_response_len);
+       if (ret != RCODE_COMPLETE) {
+               pr_debug("failed to write login response block: %x\n", ret);
+
+               kfree(response);
+               sbp_login_release(login, true);
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_TRANSPORT_FAILURE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_UNSPECIFIED_ERROR));
+               return;
+       }
+
+       kfree(response);
+
+       req->status.status = cpu_to_be32(
+               STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+               STATUS_BLOCK_SBP_STATUS(SBP_STATUS_OK));
+}
+
+static void sbp_management_request_query_logins(
+       struct sbp_management_agent *agent, struct sbp_management_request *req,
+       int *status_data_size)
+{
+       pr_notice("QUERY LOGINS not implemented\n");
+       /* FIXME: implement */
+
+       req->status.status = cpu_to_be32(
+               STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+               STATUS_BLOCK_SBP_STATUS(SBP_STATUS_REQ_TYPE_NOTSUPP));
+}
+
+static void sbp_management_request_reconnect(
+       struct sbp_management_agent *agent, struct sbp_management_request *req,
+       int *status_data_size)
+{
+       struct sbp_tport *tport = agent->tport;
+       struct sbp_tpg *tpg = tport->tpg;
+       int ret;
+       u64 guid;
+       struct sbp_login_descriptor *login;
+
+       ret = read_peer_guid(&guid, req);
+       if (ret != RCODE_COMPLETE) {
+               pr_warn("failed to read peer GUID: %d\n", ret);
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_TRANSPORT_FAILURE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_UNSPECIFIED_ERROR));
+               return;
+       }
+
+       pr_notice("mgt_agent RECONNECT from %016llx\n", guid);
+
+       login = sbp_login_find_by_id(tpg,
+               RECONNECT_ORB_LOGIN_ID(be32_to_cpu(req->orb.misc)));
+
+       if (!login) {
+               pr_err("mgt_agent RECONNECT unknown login ID\n");
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_ACCESS_DENIED));
+               return;
+       }
+
+       if (login->sess->guid != guid) {
+               pr_err("mgt_agent RECONNECT login GUID doesn't match\n");
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_ACCESS_DENIED));
+               return;
+       }
+
+       spin_lock_bh(&login->sess->lock);
+       if (login->sess->card)
+               fw_card_put(login->sess->card);
+
+       /* update the node details */
+       login->sess->generation = req->generation;
+       login->sess->node_id = req->node_addr;
+       login->sess->card = fw_card_get(req->card);
+       login->sess->speed = req->speed;
+       spin_unlock_bh(&login->sess->lock);
+
+       req->status.status = cpu_to_be32(
+               STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+               STATUS_BLOCK_SBP_STATUS(SBP_STATUS_OK));
+}
+
+static void sbp_management_request_logout(
+       struct sbp_management_agent *agent, struct sbp_management_request *req,
+       int *status_data_size)
+{
+       struct sbp_tport *tport = agent->tport;
+       struct sbp_tpg *tpg = tport->tpg;
+       int id;
+       struct sbp_login_descriptor *login;
+
+       id = LOGOUT_ORB_LOGIN_ID(be32_to_cpu(req->orb.misc));
+
+       login = sbp_login_find_by_id(tpg, id);
+       if (!login) {
+               pr_warn("cannot find login: %d\n", id);
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_LOGIN_ID_UNKNOWN));
+               return;
+       }
+
+       pr_info("mgt_agent LOGOUT from LUN %d session %d\n",
+               login->lun->unpacked_lun, login->login_id);
+
+       if (req->node_addr != login->sess->node_id) {
+               pr_warn("logout from different node ID\n");
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_ACCESS_DENIED));
+               return;
+       }
+
+       sbp_login_release(login, true);
+
+       req->status.status = cpu_to_be32(
+               STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+               STATUS_BLOCK_SBP_STATUS(SBP_STATUS_OK));
+}
+
+static void session_check_for_reset(struct sbp_session *sess)
+{
+       bool card_valid = false;
+
+       spin_lock_bh(&sess->lock);
+
+       if (sess->card) {
+               spin_lock_irq(&sess->card->lock);
+               card_valid = (sess->card->local_node != NULL);
+               spin_unlock_irq(&sess->card->lock);
+
+               if (!card_valid) {
+                       fw_card_put(sess->card);
+                       sess->card = NULL;
+               }
+       }
+
+       if (!card_valid || (sess->generation != sess->card->generation)) {
+               pr_info("Waiting for reconnect from node: %016llx\n",
+                               sess->guid);
+
+               sess->node_id = -1;
+               sess->reconnect_expires = get_jiffies_64() +
+                       ((sess->reconnect_hold + 1) * HZ);
+       }
+
+       spin_unlock_bh(&sess->lock);
+}
+
+static void session_reconnect_expired(struct sbp_session *sess)
+{
+       struct sbp_login_descriptor *login, *temp;
+       LIST_HEAD(login_list);
+
+       pr_info("Reconnect timer expired for node: %016llx\n", sess->guid);
+
+       spin_lock_bh(&sess->lock);
+       list_for_each_entry_safe(login, temp, &sess->login_list, link) {
+               login->sess = NULL;
+               list_move_tail(&login->link, &login_list);
+       }
+       spin_unlock_bh(&sess->lock);
+
+       list_for_each_entry_safe(login, temp, &login_list, link) {
+               list_del(&login->link);
+               sbp_login_release(login, false);
+       }
+
+       sbp_session_release(sess, false);
+}
+
+static void session_maintenance_work(struct work_struct *work)
+{
+       struct sbp_session *sess = container_of(work, struct sbp_session,
+                       maint_work.work);
+
+       /* could be called while tearing down the session */
+       spin_lock_bh(&sess->lock);
+       if (list_empty(&sess->login_list)) {
+               spin_unlock_bh(&sess->lock);
+               return;
+       }
+       spin_unlock_bh(&sess->lock);
+
+       if (sess->node_id != -1) {
+               /* check for bus reset and make node_id invalid */
+               session_check_for_reset(sess);
+
+               schedule_delayed_work(&sess->maint_work,
+                               SESSION_MAINTENANCE_INTERVAL);
+       } else if (!time_after64(get_jiffies_64(), sess->reconnect_expires)) {
+               /* still waiting for reconnect */
+               schedule_delayed_work(&sess->maint_work,
+                               SESSION_MAINTENANCE_INTERVAL);
+       } else {
+               /* reconnect timeout has expired */
+               session_reconnect_expired(sess);
+       }
+}
+
+static int tgt_agent_rw_agent_state(struct fw_card *card, int tcode, void *data,
+               struct sbp_target_agent *agent)
+{
+       int state;
+
+       switch (tcode) {
+       case TCODE_READ_QUADLET_REQUEST:
+               pr_debug("tgt_agent AGENT_STATE READ\n");
+
+               spin_lock_bh(&agent->lock);
+               state = agent->state;
+               spin_unlock_bh(&agent->lock);
+
+               *(__be32 *)data = cpu_to_be32(state);
+
+               return RCODE_COMPLETE;
+
+       case TCODE_WRITE_QUADLET_REQUEST:
+               /* ignored */
+               return RCODE_COMPLETE;
+
+       default:
+               return RCODE_TYPE_ERROR;
+       }
+}
+
+static int tgt_agent_rw_agent_reset(struct fw_card *card, int tcode, void *data,
+               struct sbp_target_agent *agent)
+{
+       switch (tcode) {
+       case TCODE_WRITE_QUADLET_REQUEST:
+               pr_debug("tgt_agent AGENT_RESET\n");
+               spin_lock_bh(&agent->lock);
+               agent->state = AGENT_STATE_RESET;
+               spin_unlock_bh(&agent->lock);
+               return RCODE_COMPLETE;
+
+       default:
+               return RCODE_TYPE_ERROR;
+       }
+}
+
+static int tgt_agent_rw_orb_pointer(struct fw_card *card, int tcode, void *data,
+               struct sbp_target_agent *agent)
+{
+       struct sbp2_pointer *ptr = data;
+
+       switch (tcode) {
+       case TCODE_WRITE_BLOCK_REQUEST:
+               spin_lock_bh(&agent->lock);
+               if (agent->state != AGENT_STATE_SUSPENDED &&
+                               agent->state != AGENT_STATE_RESET) {
+                       spin_unlock_bh(&agent->lock);
+                       pr_notice("Ignoring ORB_POINTER write while active.\n");
+                       return RCODE_CONFLICT_ERROR;
+               }
+               agent->state = AGENT_STATE_ACTIVE;
+               spin_unlock_bh(&agent->lock);
+
+               agent->orb_pointer = sbp2_pointer_to_addr(ptr);
+               agent->doorbell = false;
+
+               pr_debug("tgt_agent ORB_POINTER write: 0x%llx\n",
+                               agent->orb_pointer);
+
+               queue_work(system_unbound_wq, &agent->work);
+
+               return RCODE_COMPLETE;
+
+       case TCODE_READ_BLOCK_REQUEST:
+               pr_debug("tgt_agent ORB_POINTER READ\n");
+               spin_lock_bh(&agent->lock);
+               addr_to_sbp2_pointer(agent->orb_pointer, ptr);
+               spin_unlock_bh(&agent->lock);
+               return RCODE_COMPLETE;
+
+       default:
+               return RCODE_TYPE_ERROR;
+       }
+}
+
+static int tgt_agent_rw_doorbell(struct fw_card *card, int tcode, void *data,
+               struct sbp_target_agent *agent)
+{
+       switch (tcode) {
+       case TCODE_WRITE_QUADLET_REQUEST:
+               spin_lock_bh(&agent->lock);
+               if (agent->state != AGENT_STATE_SUSPENDED) {
+                       spin_unlock_bh(&agent->lock);
+                       pr_debug("Ignoring DOORBELL while active.\n");
+                       return RCODE_CONFLICT_ERROR;
+               }
+               agent->state = AGENT_STATE_ACTIVE;
+               spin_unlock_bh(&agent->lock);
+
+               agent->doorbell = true;
+
+               pr_debug("tgt_agent DOORBELL\n");
+
+               queue_work(system_unbound_wq, &agent->work);
+
+               return RCODE_COMPLETE;
+
+       case TCODE_READ_QUADLET_REQUEST:
+               return RCODE_COMPLETE;
+
+       default:
+               return RCODE_TYPE_ERROR;
+       }
+}
+
+static int tgt_agent_rw_unsolicited_status_enable(struct fw_card *card,
+               int tcode, void *data, struct sbp_target_agent *agent)
+{
+       switch (tcode) {
+       case TCODE_WRITE_QUADLET_REQUEST:
+               pr_debug("tgt_agent UNSOLICITED_STATUS_ENABLE\n");
+               /* ignored as we don't send unsolicited status */
+               return RCODE_COMPLETE;
+
+       case TCODE_READ_QUADLET_REQUEST:
+               return RCODE_COMPLETE;
+
+       default:
+               return RCODE_TYPE_ERROR;
+       }
+}
+
+static void tgt_agent_rw(struct fw_card *card, struct fw_request *request,
+               int tcode, int destination, int source, int generation,
+               unsigned long long offset, void *data, size_t length,
+               void *callback_data)
+{
+       struct sbp_target_agent *agent = callback_data;
+       struct sbp_session *sess = agent->login->sess;
+       int sess_gen, sess_node, rcode;
+
+       spin_lock_bh(&sess->lock);
+       sess_gen = sess->generation;
+       sess_node = sess->node_id;
+       spin_unlock_bh(&sess->lock);
+
+       if (generation != sess_gen) {
+               pr_notice("ignoring request with wrong generation\n");
+               rcode = RCODE_TYPE_ERROR;
+               goto out;
+       }
+
+       if (source != sess_node) {
+               pr_notice("ignoring request from foreign node (%x != %x)\n",
+                               source, sess_node);
+               rcode = RCODE_TYPE_ERROR;
+               goto out;
+       }
+
+       /* turn offset into the offset from the start of the block */
+       offset -= agent->handler.offset;
+
+       if (offset == 0x00 && length == 4) {
+               /* AGENT_STATE */
+               rcode = tgt_agent_rw_agent_state(card, tcode, data, agent);
+       } else if (offset == 0x04 && length == 4) {
+               /* AGENT_RESET */
+               rcode = tgt_agent_rw_agent_reset(card, tcode, data, agent);
+       } else if (offset == 0x08 && length == 8) {
+               /* ORB_POINTER */
+               rcode = tgt_agent_rw_orb_pointer(card, tcode, data, agent);
+       } else if (offset == 0x10 && length == 4) {
+               /* DOORBELL */
+               rcode = tgt_agent_rw_doorbell(card, tcode, data, agent);
+       } else if (offset == 0x14 && length == 4) {
+               /* UNSOLICITED_STATUS_ENABLE */
+               rcode = tgt_agent_rw_unsolicited_status_enable(card, tcode,
+                               data, agent);
+       } else {
+               rcode = RCODE_ADDRESS_ERROR;
+       }
+
+out:
+       fw_send_response(card, request, rcode);
+}
+
+static void sbp_handle_command(struct sbp_target_request *);
+static int sbp_send_status(struct sbp_target_request *);
+static void sbp_free_request(struct sbp_target_request *);
+
+static void tgt_agent_process_work(struct work_struct *work)
+{
+       struct sbp_target_request *req =
+               container_of(work, struct sbp_target_request, work);
+
+       pr_debug("tgt_orb ptr:0x%llx next_ORB:0x%llx data_descriptor:0x%llx misc:0x%x\n",
+                       req->orb_pointer,
+                       sbp2_pointer_to_addr(&req->orb.next_orb),
+                       sbp2_pointer_to_addr(&req->orb.data_descriptor),
+                       be32_to_cpu(req->orb.misc));
+
+       if (req->orb_pointer >> 32)
+               pr_debug("ORB with high bits set\n");
+
+       switch (ORB_REQUEST_FORMAT(be32_to_cpu(req->orb.misc))) {
+               case 0:/* Format specified by this standard */
+                       sbp_handle_command(req);
+                       return;
+               case 1: /* Reserved for future standardization */
+               case 2: /* Vendor-dependent */
+                       req->status.status |= cpu_to_be32(
+                                       STATUS_BLOCK_RESP(
+                                               STATUS_RESP_REQUEST_COMPLETE) |
+                                       STATUS_BLOCK_DEAD(0) |
+                                       STATUS_BLOCK_LEN(1) |
+                                       STATUS_BLOCK_SBP_STATUS(
+                                               SBP_STATUS_REQ_TYPE_NOTSUPP));
+                       sbp_send_status(req);
+                       sbp_free_request(req);
+                       return;
+               case 3: /* Dummy ORB */
+                       req->status.status |= cpu_to_be32(
+                                       STATUS_BLOCK_RESP(
+                                               STATUS_RESP_REQUEST_COMPLETE) |
+                                       STATUS_BLOCK_DEAD(0) |
+                                       STATUS_BLOCK_LEN(1) |
+                                       STATUS_BLOCK_SBP_STATUS(
+                                               SBP_STATUS_DUMMY_ORB_COMPLETE));
+                       sbp_send_status(req);
+                       sbp_free_request(req);
+                       return;
+               default:
+                       BUG();
+       }
+}
+
+/* used to double-check we haven't been issued an AGENT_RESET */
+static inline bool tgt_agent_check_active(struct sbp_target_agent *agent)
+{
+       bool active;
+
+       spin_lock_bh(&agent->lock);
+       active = (agent->state == AGENT_STATE_ACTIVE);
+       spin_unlock_bh(&agent->lock);
+
+       return active;
+}
+
+static void tgt_agent_fetch_work(struct work_struct *work)
+{
+       struct sbp_target_agent *agent =
+               container_of(work, struct sbp_target_agent, work);
+       struct sbp_session *sess = agent->login->sess;
+       struct sbp_target_request *req;
+       int ret;
+       bool doorbell = agent->doorbell;
+       u64 next_orb = agent->orb_pointer;
+
+       while (next_orb && tgt_agent_check_active(agent)) {
+               req = kzalloc(sizeof(*req), GFP_KERNEL);
+               if (!req) {
+                       spin_lock_bh(&agent->lock);
+                       agent->state = AGENT_STATE_DEAD;
+                       spin_unlock_bh(&agent->lock);
+                       return;
+               }
+
+               req->login = agent->login;
+               req->orb_pointer = next_orb;
+
+               req->status.status = cpu_to_be32(STATUS_BLOCK_ORB_OFFSET_HIGH(
+                                       req->orb_pointer >> 32));
+               req->status.orb_low = cpu_to_be32(
+                               req->orb_pointer & 0xfffffffc);
+
+               /* read in the ORB */
+               ret = sbp_run_transaction(sess->card, TCODE_READ_BLOCK_REQUEST,
+                               sess->node_id, sess->generation, sess->speed,
+                               req->orb_pointer, &req->orb, sizeof(req->orb));
+               if (ret != RCODE_COMPLETE) {
+                       pr_debug("tgt_orb fetch failed: %x\n", ret);
+                       req->status.status |= cpu_to_be32(
+                                       STATUS_BLOCK_SRC(
+                                               STATUS_SRC_ORB_FINISHED) |
+                                       STATUS_BLOCK_RESP(
+                                               STATUS_RESP_TRANSPORT_FAILURE) |
+                                       STATUS_BLOCK_DEAD(1) |
+                                       STATUS_BLOCK_LEN(1) |
+                                       STATUS_BLOCK_SBP_STATUS(
+                                               SBP_STATUS_UNSPECIFIED_ERROR));
+                       spin_lock_bh(&agent->lock);
+                       agent->state = AGENT_STATE_DEAD;
+                       spin_unlock_bh(&agent->lock);
+
+                       sbp_send_status(req);
+                       sbp_free_request(req);
+                       return;
+               }
+
+               /* check the next_ORB field */
+               if (be32_to_cpu(req->orb.next_orb.high) & 0x80000000) {
+                       next_orb = 0;
+                       req->status.status |= cpu_to_be32(STATUS_BLOCK_SRC(
+                                               STATUS_SRC_ORB_FINISHED));
+               } else {
+                       next_orb = sbp2_pointer_to_addr(&req->orb.next_orb);
+                       req->status.status |= cpu_to_be32(STATUS_BLOCK_SRC(
+                                               STATUS_SRC_ORB_CONTINUING));
+               }
+
+               if (tgt_agent_check_active(agent) && !doorbell) {
+                       INIT_WORK(&req->work, tgt_agent_process_work);
+                       queue_work(system_unbound_wq, &req->work);
+               } else {
+                       /* don't process this request, just check next_ORB */
+                       sbp_free_request(req);
+               }
+
+               spin_lock_bh(&agent->lock);
+               doorbell = agent->doorbell = false;
+
+               /* check if we should carry on processing */
+               if (next_orb)
+                       agent->orb_pointer = next_orb;
+               else
+                       agent->state = AGENT_STATE_SUSPENDED;
+
+               spin_unlock_bh(&agent->lock);
+       };
+}
+
+static struct sbp_target_agent *sbp_target_agent_register(
+               struct sbp_login_descriptor *login)
+{
+       struct sbp_target_agent *agent;
+       int ret;
+
+       agent = kmalloc(sizeof(*agent), GFP_KERNEL);
+       if (!agent)
+               return ERR_PTR(-ENOMEM);
+
+       spin_lock_init(&agent->lock);
+
+       agent->handler.length = 0x20;
+       agent->handler.address_callback = tgt_agent_rw;
+       agent->handler.callback_data = agent;
+
+       agent->login = login;
+       agent->state = AGENT_STATE_RESET;
+       INIT_WORK(&agent->work, tgt_agent_fetch_work);
+       agent->orb_pointer = 0;
+       agent->doorbell = false;
+
+       ret = fw_core_add_address_handler(&agent->handler,
+                       &sbp_register_region);
+       if (ret < 0) {
+               kfree(agent);
+               return ERR_PTR(ret);
+       }
+
+       return agent;
+}
+
+static void sbp_target_agent_unregister(struct sbp_target_agent *agent)
+{
+       fw_core_remove_address_handler(&agent->handler);
+       cancel_work_sync(&agent->work);
+       kfree(agent);
+}
+
+/*
+ * Simple wrapper around fw_run_transaction that retries the transaction several
+ * times in case of failure, with an exponential backoff.
+ */
+static int sbp_run_transaction(struct fw_card *card, int tcode, int destination_id,
+               int generation, int speed, unsigned long long offset,
+               void *payload, size_t length)
+{
+       int attempt, ret, delay;
+
+       for (attempt = 1; attempt <= 5; attempt++) {
+               ret = fw_run_transaction(card, tcode, destination_id,
+                               generation, speed, offset, payload, length);
+
+               switch (ret) {
+               case RCODE_COMPLETE:
+               case RCODE_TYPE_ERROR:
+               case RCODE_ADDRESS_ERROR:
+               case RCODE_GENERATION:
+                       return ret;
+
+               default:
+                       delay = 5 * attempt * attempt;
+                       usleep_range(delay, delay * 2);
+               }
+       }
+
+       return ret;
+}
+
+/*
+ * Wrapper around sbp_run_transaction that gets the card, destination,
+ * generation and speed out of the request's session.
+ */
+static int sbp_run_request_transaction(struct sbp_target_request *req,
+               int tcode, unsigned long long offset, void *payload,
+               size_t length)
+{
+       struct sbp_login_descriptor *login = req->login;
+       struct sbp_session *sess = login->sess;
+       struct fw_card *card;
+       int node_id, generation, speed, ret;
+
+       spin_lock_bh(&sess->lock);
+       card = fw_card_get(sess->card);
+       node_id = sess->node_id;
+       generation = sess->generation;
+       speed = sess->speed;
+       spin_unlock_bh(&sess->lock);
+
+       ret = sbp_run_transaction(card, tcode, node_id, generation, speed,
+                       offset, payload, length);
+
+       fw_card_put(card);
+
+       return ret;
+}
+
+static int sbp_fetch_command(struct sbp_target_request *req)
+{
+       int ret, cmd_len, copy_len;
+
+       cmd_len = scsi_command_size(req->orb.command_block);
+
+       req->cmd_buf = kmalloc(cmd_len, GFP_KERNEL);
+       if (!req->cmd_buf)
+               return -ENOMEM;
+
+       memcpy(req->cmd_buf, req->orb.command_block,
+               min_t(int, cmd_len, sizeof(req->orb.command_block)));
+
+       if (cmd_len > sizeof(req->orb.command_block)) {
+               pr_debug("sbp_fetch_command: filling in long command\n");
+               copy_len = cmd_len - sizeof(req->orb.command_block);
+
+               ret = sbp_run_request_transaction(req,
+                               TCODE_READ_BLOCK_REQUEST,
+                               req->orb_pointer + sizeof(req->orb),
+                               req->cmd_buf + sizeof(req->orb.command_block),
+                               copy_len);
+               if (ret != RCODE_COMPLETE)
+                       return -EIO;
+       }
+
+       return 0;
+}
+
+static int sbp_fetch_page_table(struct sbp_target_request *req)
+{
+       int pg_tbl_sz, ret;
+       struct sbp_page_table_entry *pg_tbl;
+
+       if (!CMDBLK_ORB_PG_TBL_PRESENT(be32_to_cpu(req->orb.misc)))
+               return 0;
+
+       pg_tbl_sz = CMDBLK_ORB_DATA_SIZE(be32_to_cpu(req->orb.misc)) *
+               sizeof(struct sbp_page_table_entry);
+
+       pg_tbl = kmalloc(pg_tbl_sz, GFP_KERNEL);
+       if (!pg_tbl)
+               return -ENOMEM;
+
+       ret = sbp_run_request_transaction(req, TCODE_READ_BLOCK_REQUEST,
+                       sbp2_pointer_to_addr(&req->orb.data_descriptor),
+                       pg_tbl, pg_tbl_sz);
+       if (ret != RCODE_COMPLETE) {
+               kfree(pg_tbl);
+               return -EIO;
+       }
+
+       req->pg_tbl = pg_tbl;
+       return 0;
+}
+
+static void sbp_calc_data_length_direction(struct sbp_target_request *req,
+       u32 *data_len, enum dma_data_direction *data_dir)
+{
+       int data_size, direction, idx;
+
+       data_size = CMDBLK_ORB_DATA_SIZE(be32_to_cpu(req->orb.misc));
+       direction = CMDBLK_ORB_DIRECTION(be32_to_cpu(req->orb.misc));
+
+       if (!data_size) {
+               *data_len = 0;
+               *data_dir = DMA_NONE;
+               return;
+       }
+
+       *data_dir = direction ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
+
+       if (req->pg_tbl) {
+               *data_len = 0;
+               for (idx = 0; idx < data_size; idx++) {
+                       *data_len += be16_to_cpu(
+                                       req->pg_tbl[idx].segment_length);
+               }
+       } else {
+               *data_len = data_size;
+       }
+}
+
+static void sbp_handle_command(struct sbp_target_request *req)
+{
+       struct sbp_login_descriptor *login = req->login;
+       struct sbp_session *sess = login->sess;
+       int ret, unpacked_lun;
+       u32 data_length;
+       enum dma_data_direction data_dir;
+
+       ret = sbp_fetch_command(req);
+       if (ret) {
+               pr_debug("sbp_handle_command: fetch command failed: %d\n", ret);
+               goto err;
+       }
+
+       ret = sbp_fetch_page_table(req);
+       if (ret) {
+               pr_debug("sbp_handle_command: fetch page table failed: %d\n",
+                       ret);
+               goto err;
+       }
+
+       unpacked_lun = req->login->lun->unpacked_lun;
+       sbp_calc_data_length_direction(req, &data_length, &data_dir);
+
+       pr_debug("sbp_handle_command ORB:0x%llx unpacked_lun:%d data_len:%d data_dir:%d\n",
+                       req->orb_pointer, unpacked_lun, data_length, data_dir);
+
+       if (target_submit_cmd(&req->se_cmd, sess->se_sess, req->cmd_buf,
+                             req->sense_buf, unpacked_lun, data_length,
+                             TCM_SIMPLE_TAG, data_dir, 0))
+               goto err;
+
+       return;
+
+err:
+       req->status.status |= cpu_to_be32(
+               STATUS_BLOCK_RESP(STATUS_RESP_TRANSPORT_FAILURE) |
+               STATUS_BLOCK_DEAD(0) |
+               STATUS_BLOCK_LEN(1) |
+               STATUS_BLOCK_SBP_STATUS(SBP_STATUS_UNSPECIFIED_ERROR));
+       sbp_send_status(req);
+       sbp_free_request(req);
+}
+
+/*
+ * DMA_TO_DEVICE = read from initiator (SCSI WRITE)
+ * DMA_FROM_DEVICE = write to initiator (SCSI READ)
+ */
+static int sbp_rw_data(struct sbp_target_request *req)
+{
+       struct sbp_session *sess = req->login->sess;
+       int tcode, sg_miter_flags, max_payload, pg_size, speed, node_id,
+               generation, num_pte, length, tfr_length,
+               rcode = RCODE_COMPLETE;
+       struct sbp_page_table_entry *pte;
+       unsigned long long offset;
+       struct fw_card *card;
+       struct sg_mapping_iter iter;
+
+       if (req->se_cmd.data_direction == DMA_FROM_DEVICE) {
+               tcode = TCODE_WRITE_BLOCK_REQUEST;
+               sg_miter_flags = SG_MITER_FROM_SG;
+       } else {
+               tcode = TCODE_READ_BLOCK_REQUEST;
+               sg_miter_flags = SG_MITER_TO_SG;
+       }
+
+       max_payload = 4 << CMDBLK_ORB_MAX_PAYLOAD(be32_to_cpu(req->orb.misc));
+       speed = CMDBLK_ORB_SPEED(be32_to_cpu(req->orb.misc));
+
+       pg_size = CMDBLK_ORB_PG_SIZE(be32_to_cpu(req->orb.misc));
+       if (pg_size) {
+               pr_err("sbp_run_transaction: page size ignored\n");
+               pg_size = 0x100 << pg_size;
+       }
+
+       spin_lock_bh(&sess->lock);
+       card = fw_card_get(sess->card);
+       node_id = sess->node_id;
+       generation = sess->generation;
+       spin_unlock_bh(&sess->lock);
+
+       if (req->pg_tbl) {
+               pte = req->pg_tbl;
+               num_pte = CMDBLK_ORB_DATA_SIZE(be32_to_cpu(req->orb.misc));
+
+               offset = 0;
+               length = 0;
+       } else {
+               pte = NULL;
+               num_pte = 0;
+
+               offset = sbp2_pointer_to_addr(&req->orb.data_descriptor);
+               length = req->se_cmd.data_length;
+       }
+
+       sg_miter_start(&iter, req->se_cmd.t_data_sg, req->se_cmd.t_data_nents,
+               sg_miter_flags);
+
+       while (length || num_pte) {
+               if (!length) {
+                       offset = (u64)be16_to_cpu(pte->segment_base_hi) << 32 |
+                               be32_to_cpu(pte->segment_base_lo);
+                       length = be16_to_cpu(pte->segment_length);
+
+                       pte++;
+                       num_pte--;
+               }
+
+               sg_miter_next(&iter);
+
+               tfr_length = min3(length, max_payload, (int)iter.length);
+
+               /* FIXME: take page_size into account */
+
+               rcode = sbp_run_transaction(card, tcode, node_id,
+                               generation, speed,
+                               offset, iter.addr, tfr_length);
+
+               if (rcode != RCODE_COMPLETE)
+                       break;
+
+               length -= tfr_length;
+               offset += tfr_length;
+               iter.consumed = tfr_length;
+       }
+
+       sg_miter_stop(&iter);
+       fw_card_put(card);
+
+       if (rcode == RCODE_COMPLETE) {
+               WARN_ON(length != 0);
+               return 0;
+       } else {
+               return -EIO;
+       }
+}
+
+static int sbp_send_status(struct sbp_target_request *req)
+{
+       int ret, length;
+       struct sbp_login_descriptor *login = req->login;
+
+       length = (((be32_to_cpu(req->status.status) >> 24) & 0x07) + 1) * 4;
+
+       ret = sbp_run_request_transaction(req, TCODE_WRITE_BLOCK_REQUEST,
+                       login->status_fifo_addr, &req->status, length);
+       if (ret != RCODE_COMPLETE) {
+               pr_debug("sbp_send_status: write failed: 0x%x\n", ret);
+               return -EIO;
+       }
+
+       pr_debug("sbp_send_status: status write complete for ORB: 0x%llx\n",
+                       req->orb_pointer);
+
+       return 0;
+}
+
+static void sbp_sense_mangle(struct sbp_target_request *req)
+{
+       struct se_cmd *se_cmd = &req->se_cmd;
+       u8 *sense = req->sense_buf;
+       u8 *status = req->status.data;
+
+       WARN_ON(se_cmd->scsi_sense_length < 18);
+
+       switch (sense[0] & 0x7f) {              /* sfmt */
+       case 0x70: /* current, fixed */
+               status[0] = 0 << 6;
+               break;
+       case 0x71: /* deferred, fixed */
+               status[0] = 1 << 6;
+               break;
+       case 0x72: /* current, descriptor */
+       case 0x73: /* deferred, descriptor */
+       default:
+               /*
+                * TODO: SBP-3 specifies what we should do with descriptor
+                * format sense data
+                */
+               pr_err("sbp_send_sense: unknown sense format: 0x%x\n",
+                       sense[0]);
+               req->status.status |= cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_DEAD(0) |
+                       STATUS_BLOCK_LEN(1) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_REQUEST_ABORTED));
+               return;
+       }
+
+       status[0] |= se_cmd->scsi_status & 0x3f;/* status */
+       status[1] =
+               (sense[0] & 0x80) |             /* valid */
+               ((sense[2] & 0xe0) >> 1) |      /* mark, eom, ili */
+               (sense[2] & 0x0f);              /* sense_key */
+       status[2] = se_cmd->scsi_asc;           /* sense_code */
+       status[3] = se_cmd->scsi_ascq;          /* sense_qualifier */
+
+       /* information */
+       status[4] = sense[3];
+       status[5] = sense[4];
+       status[6] = sense[5];
+       status[7] = sense[6];
+
+       /* CDB-dependent */
+       status[8] = sense[8];
+       status[9] = sense[9];
+       status[10] = sense[10];
+       status[11] = sense[11];
+
+       /* fru */
+       status[12] = sense[14];
+
+       /* sense_key-dependent */
+       status[13] = sense[15];
+       status[14] = sense[16];
+       status[15] = sense[17];
+
+       req->status.status |= cpu_to_be32(
+               STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+               STATUS_BLOCK_DEAD(0) |
+               STATUS_BLOCK_LEN(5) |
+               STATUS_BLOCK_SBP_STATUS(SBP_STATUS_OK));
+}
+
+static int sbp_send_sense(struct sbp_target_request *req)
+{
+       struct se_cmd *se_cmd = &req->se_cmd;
+
+       if (se_cmd->scsi_sense_length) {
+               sbp_sense_mangle(req);
+       } else {
+               req->status.status |= cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_DEAD(0) |
+                       STATUS_BLOCK_LEN(1) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_OK));
+       }
+
+       return sbp_send_status(req);
+}
+
+static void sbp_free_request(struct sbp_target_request *req)
+{
+       kfree(req->pg_tbl);
+       kfree(req->cmd_buf);
+       kfree(req);
+}
+
+static void sbp_mgt_agent_process(struct work_struct *work)
+{
+       struct sbp_management_agent *agent =
+               container_of(work, struct sbp_management_agent, work);
+       struct sbp_management_request *req = agent->request;
+       int ret;
+       int status_data_len = 0;
+
+       /* fetch the ORB from the initiator */
+       ret = sbp_run_transaction(req->card, TCODE_READ_BLOCK_REQUEST,
+               req->node_addr, req->generation, req->speed,
+               agent->orb_offset, &req->orb, sizeof(req->orb));
+       if (ret != RCODE_COMPLETE) {
+               pr_debug("mgt_orb fetch failed: %x\n", ret);
+               goto out;
+       }
+
+       pr_debug("mgt_orb ptr1:0x%llx ptr2:0x%llx misc:0x%x len:0x%x status_fifo:0x%llx\n",
+               sbp2_pointer_to_addr(&req->orb.ptr1),
+               sbp2_pointer_to_addr(&req->orb.ptr2),
+               be32_to_cpu(req->orb.misc), be32_to_cpu(req->orb.length),
+               sbp2_pointer_to_addr(&req->orb.status_fifo));
+
+       if (!ORB_NOTIFY(be32_to_cpu(req->orb.misc)) ||
+               ORB_REQUEST_FORMAT(be32_to_cpu(req->orb.misc)) != 0) {
+               pr_err("mgt_orb bad request\n");
+               goto out;
+       }
+
+       switch (MANAGEMENT_ORB_FUNCTION(be32_to_cpu(req->orb.misc))) {
+       case MANAGEMENT_ORB_FUNCTION_LOGIN:
+               sbp_management_request_login(agent, req, &status_data_len);
+               break;
+
+       case MANAGEMENT_ORB_FUNCTION_QUERY_LOGINS:
+               sbp_management_request_query_logins(agent, req,
+                               &status_data_len);
+               break;
+
+       case MANAGEMENT_ORB_FUNCTION_RECONNECT:
+               sbp_management_request_reconnect(agent, req, &status_data_len);
+               break;
+
+       case MANAGEMENT_ORB_FUNCTION_SET_PASSWORD:
+               pr_notice("SET PASSWORD not implemented\n");
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_REQ_TYPE_NOTSUPP));
+
+               break;
+
+       case MANAGEMENT_ORB_FUNCTION_LOGOUT:
+               sbp_management_request_logout(agent, req, &status_data_len);
+               break;
+
+       case MANAGEMENT_ORB_FUNCTION_ABORT_TASK:
+               pr_notice("ABORT TASK not implemented\n");
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_REQ_TYPE_NOTSUPP));
+
+               break;
+
+       case MANAGEMENT_ORB_FUNCTION_ABORT_TASK_SET:
+               pr_notice("ABORT TASK SET not implemented\n");
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_REQ_TYPE_NOTSUPP));
+
+               break;
+
+       case MANAGEMENT_ORB_FUNCTION_LOGICAL_UNIT_RESET:
+               pr_notice("LOGICAL UNIT RESET not implemented\n");
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_REQ_TYPE_NOTSUPP));
+
+               break;
+
+       case MANAGEMENT_ORB_FUNCTION_TARGET_RESET:
+               pr_notice("TARGET RESET not implemented\n");
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_REQ_TYPE_NOTSUPP));
+
+               break;
+
+       default:
+               pr_notice("unknown management function 0x%x\n",
+                       MANAGEMENT_ORB_FUNCTION(be32_to_cpu(req->orb.misc)));
+
+               req->status.status = cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_REQUEST_COMPLETE) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_REQ_TYPE_NOTSUPP));
+
+               break;
+       }
+
+       req->status.status |= cpu_to_be32(
+               STATUS_BLOCK_SRC(1) | /* Response to ORB, next_ORB absent */
+               STATUS_BLOCK_LEN(DIV_ROUND_UP(status_data_len, 4) + 1) |
+               STATUS_BLOCK_ORB_OFFSET_HIGH(agent->orb_offset >> 32));
+       req->status.orb_low = cpu_to_be32(agent->orb_offset);
+
+       /* write the status block back to the initiator */
+       ret = sbp_run_transaction(req->card, TCODE_WRITE_BLOCK_REQUEST,
+               req->node_addr, req->generation, req->speed,
+               sbp2_pointer_to_addr(&req->orb.status_fifo),
+               &req->status, 8 + status_data_len);
+       if (ret != RCODE_COMPLETE) {
+               pr_debug("mgt_orb status write failed: %x\n", ret);
+               goto out;
+       }
+
+out:
+       fw_card_put(req->card);
+       kfree(req);
+
+       spin_lock_bh(&agent->lock);
+       agent->state = MANAGEMENT_AGENT_STATE_IDLE;
+       spin_unlock_bh(&agent->lock);
+}
+
+static void sbp_mgt_agent_rw(struct fw_card *card,
+       struct fw_request *request, int tcode, int destination, int source,
+       int generation, unsigned long long offset, void *data, size_t length,
+       void *callback_data)
+{
+       struct sbp_management_agent *agent = callback_data;
+       struct sbp2_pointer *ptr = data;
+       int rcode = RCODE_ADDRESS_ERROR;
+
+       if (!agent->tport->enable)
+               goto out;
+
+       if ((offset != agent->handler.offset) || (length != 8))
+               goto out;
+
+       if (tcode == TCODE_WRITE_BLOCK_REQUEST) {
+               struct sbp_management_request *req;
+               int prev_state;
+
+               spin_lock_bh(&agent->lock);
+               prev_state = agent->state;
+               agent->state = MANAGEMENT_AGENT_STATE_BUSY;
+               spin_unlock_bh(&agent->lock);
+
+               if (prev_state == MANAGEMENT_AGENT_STATE_BUSY) {
+                       pr_notice("ignoring management request while busy\n");
+                       rcode = RCODE_CONFLICT_ERROR;
+                       goto out;
+               }
+
+               req = kzalloc(sizeof(*req), GFP_ATOMIC);
+               if (!req) {
+                       rcode = RCODE_CONFLICT_ERROR;
+                       goto out;
+               }
+
+               req->card = fw_card_get(card);
+               req->generation = generation;
+               req->node_addr = source;
+               req->speed = fw_get_request_speed(request);
+
+               agent->orb_offset = sbp2_pointer_to_addr(ptr);
+               agent->request = req;
+
+               queue_work(system_unbound_wq, &agent->work);
+               rcode = RCODE_COMPLETE;
+       } else if (tcode == TCODE_READ_BLOCK_REQUEST) {
+               addr_to_sbp2_pointer(agent->orb_offset, ptr);
+               rcode = RCODE_COMPLETE;
+       } else {
+               rcode = RCODE_TYPE_ERROR;
+       }
+
+out:
+       fw_send_response(card, request, rcode);
+}
+
+static struct sbp_management_agent *sbp_management_agent_register(
+               struct sbp_tport *tport)
+{
+       int ret;
+       struct sbp_management_agent *agent;
+
+       agent = kmalloc(sizeof(*agent), GFP_KERNEL);
+       if (!agent)
+               return ERR_PTR(-ENOMEM);
+
+       spin_lock_init(&agent->lock);
+       agent->tport = tport;
+       agent->handler.length = 0x08;
+       agent->handler.address_callback = sbp_mgt_agent_rw;
+       agent->handler.callback_data = agent;
+       agent->state = MANAGEMENT_AGENT_STATE_IDLE;
+       INIT_WORK(&agent->work, sbp_mgt_agent_process);
+       agent->orb_offset = 0;
+       agent->request = NULL;
+
+       ret = fw_core_add_address_handler(&agent->handler,
+                       &sbp_register_region);
+       if (ret < 0) {
+               kfree(agent);
+               return ERR_PTR(ret);
+       }
+
+       return agent;
+}
+
+static void sbp_management_agent_unregister(struct sbp_management_agent *agent)
+{
+       fw_core_remove_address_handler(&agent->handler);
+       cancel_work_sync(&agent->work);
+       kfree(agent);
+}
+
+static int sbp_check_true(struct se_portal_group *se_tpg)
+{
+       return 1;
+}
+
+static int sbp_check_false(struct se_portal_group *se_tpg)
+{
+       return 0;
+}
+
+static char *sbp_get_fabric_name(void)
+{
+       return "sbp";
+}
+
+static char *sbp_get_fabric_wwn(struct se_portal_group *se_tpg)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+       struct sbp_tport *tport = tpg->tport;
+
+       return &tport->tport_name[0];
+}
+
+static u16 sbp_get_tag(struct se_portal_group *se_tpg)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+       return tpg->tport_tpgt;
+}
+
+static u32 sbp_get_default_depth(struct se_portal_group *se_tpg)
+{
+       return 1;
+}
+
+static struct se_node_acl *sbp_alloc_fabric_acl(struct se_portal_group *se_tpg)
+{
+       struct sbp_nacl *nacl;
+
+       nacl = kzalloc(sizeof(struct sbp_nacl), GFP_KERNEL);
+       if (!nacl) {
+               pr_err("Unable to allocate struct sbp_nacl\n");
+               return NULL;
+       }
+
+       return &nacl->se_node_acl;
+}
+
+static void sbp_release_fabric_acl(
+       struct se_portal_group *se_tpg,
+       struct se_node_acl *se_nacl)
+{
+       struct sbp_nacl *nacl =
+               container_of(se_nacl, struct sbp_nacl, se_node_acl);
+       kfree(nacl);
+}
+
+static u32 sbp_tpg_get_inst_index(struct se_portal_group *se_tpg)
+{
+       return 1;
+}
+
+static void sbp_release_cmd(struct se_cmd *se_cmd)
+{
+       struct sbp_target_request *req = container_of(se_cmd,
+                       struct sbp_target_request, se_cmd);
+
+       sbp_free_request(req);
+}
+
+static int sbp_shutdown_session(struct se_session *se_sess)
+{
+       return 0;
+}
+
+static void sbp_close_session(struct se_session *se_sess)
+{
+       return;
+}
+
+static u32 sbp_sess_get_index(struct se_session *se_sess)
+{
+       return 0;
+}
+
+static int sbp_write_pending(struct se_cmd *se_cmd)
+{
+       struct sbp_target_request *req = container_of(se_cmd,
+                       struct sbp_target_request, se_cmd);
+       int ret;
+
+       ret = sbp_rw_data(req);
+       if (ret) {
+               req->status.status |= cpu_to_be32(
+                       STATUS_BLOCK_RESP(
+                               STATUS_RESP_TRANSPORT_FAILURE) |
+                       STATUS_BLOCK_DEAD(0) |
+                       STATUS_BLOCK_LEN(1) |
+                       STATUS_BLOCK_SBP_STATUS(
+                               SBP_STATUS_UNSPECIFIED_ERROR));
+               sbp_send_status(req);
+               return ret;
+       }
+
+       target_execute_cmd(se_cmd);
+       return 0;
+}
+
+static int sbp_write_pending_status(struct se_cmd *se_cmd)
+{
+       return 0;
+}
+
+static void sbp_set_default_node_attrs(struct se_node_acl *nacl)
+{
+       return;
+}
+
+static u32 sbp_get_task_tag(struct se_cmd *se_cmd)
+{
+       struct sbp_target_request *req = container_of(se_cmd,
+                       struct sbp_target_request, se_cmd);
+
+       /* only used for printk until we do TMRs */
+       return (u32)req->orb_pointer;
+}
+
+static int sbp_get_cmd_state(struct se_cmd *se_cmd)
+{
+       return 0;
+}
+
+static int sbp_queue_data_in(struct se_cmd *se_cmd)
+{
+       struct sbp_target_request *req = container_of(se_cmd,
+                       struct sbp_target_request, se_cmd);
+       int ret;
+
+       ret = sbp_rw_data(req);
+       if (ret) {
+               req->status.status |= cpu_to_be32(
+                       STATUS_BLOCK_RESP(STATUS_RESP_TRANSPORT_FAILURE) |
+                       STATUS_BLOCK_DEAD(0) |
+                       STATUS_BLOCK_LEN(1) |
+                       STATUS_BLOCK_SBP_STATUS(SBP_STATUS_UNSPECIFIED_ERROR));
+               sbp_send_status(req);
+               return ret;
+       }
+
+       return sbp_send_sense(req);
+}
+
+/*
+ * Called after command (no data transfer) or after the write (to device)
+ * operation is completed
+ */
+static int sbp_queue_status(struct se_cmd *se_cmd)
+{
+       struct sbp_target_request *req = container_of(se_cmd,
+                       struct sbp_target_request, se_cmd);
+
+       return sbp_send_sense(req);
+}
+
+static void sbp_queue_tm_rsp(struct se_cmd *se_cmd)
+{
+}
+
+static void sbp_aborted_task(struct se_cmd *se_cmd)
+{
+       return;
+}
+
+static int sbp_check_stop_free(struct se_cmd *se_cmd)
+{
+       struct sbp_target_request *req = container_of(se_cmd,
+                       struct sbp_target_request, se_cmd);
+
+       transport_generic_free_cmd(&req->se_cmd, 0);
+       return 1;
+}
+
+/*
+ * Handlers for Serial Bus Protocol 2/3 (SBP-2 / SBP-3)
+ */
+static u8 sbp_get_fabric_proto_ident(struct se_portal_group *se_tpg)
+{
+       /*
+        * Return a IEEE 1394 SCSI Protocol identifier for loopback operations
+        * This is defined in section 7.5.1 Table 362 in spc4r17
+        */
+       return SCSI_PROTOCOL_SBP;
+}
+
+static u32 sbp_get_pr_transport_id(
+       struct se_portal_group *se_tpg,
+       struct se_node_acl *se_nacl,
+       struct t10_pr_registration *pr_reg,
+       int *format_code,
+       unsigned char *buf)
+{
+       int ret;
+
+       /*
+        * Set PROTOCOL IDENTIFIER to 3h for SBP
+        */
+       buf[0] = SCSI_PROTOCOL_SBP;
+       /*
+        * From spc4r17, 7.5.4.4 TransportID for initiator ports using SCSI
+        * over IEEE 1394
+        */
+       ret = hex2bin(&buf[8], se_nacl->initiatorname, 8);
+       if (ret < 0)
+               pr_debug("sbp transport_id: invalid hex string\n");
+
+       /*
+        * The IEEE 1394 Transport ID is a hardcoded 24-byte length
+        */
+       return 24;
+}
+
+static u32 sbp_get_pr_transport_id_len(
+       struct se_portal_group *se_tpg,
+       struct se_node_acl *se_nacl,
+       struct t10_pr_registration *pr_reg,
+       int *format_code)
+{
+       *format_code = 0;
+       /*
+        * From spc4r17, 7.5.4.4 TransportID for initiator ports using SCSI
+        * over IEEE 1394
+        *
+        * The SBP Transport ID is a hardcoded 24-byte length
+        */
+       return 24;
+}
+
+/*
+ * Used for handling SCSI fabric dependent TransportIDs in SPC-3 and above
+ * Persistent Reservation SPEC_I_PT=1 and PROUT REGISTER_AND_MOVE operations.
+ */
+static char *sbp_parse_pr_out_transport_id(
+       struct se_portal_group *se_tpg,
+       const char *buf,
+       u32 *out_tid_len,
+       char **port_nexus_ptr)
+{
+       /*
+        * Assume the FORMAT CODE 00b from spc4r17, 7.5.4.4 TransportID
+        * for initiator ports using SCSI over SBP Serial SCSI Protocol
+        *
+        * The TransportID for a IEEE 1394 Initiator Port is of fixed size of
+        * 24 bytes, and IEEE 1394 does not contain a I_T nexus identifier,
+        * so we return the **port_nexus_ptr set to NULL.
+        */
+       *port_nexus_ptr = NULL;
+       *out_tid_len = 24;
+
+       return (char *)&buf[8];
+}
+
+static int sbp_count_se_tpg_luns(struct se_portal_group *tpg)
+{
+       int i, count = 0;
+
+       spin_lock(&tpg->tpg_lun_lock);
+       for (i = 0; i < TRANSPORT_MAX_LUNS_PER_TPG; i++) {
+               struct se_lun *se_lun = tpg->tpg_lun_list[i];
+
+               if (se_lun->lun_status == TRANSPORT_LUN_STATUS_FREE)
+                       continue;
+
+               count++;
+       }
+       spin_unlock(&tpg->tpg_lun_lock);
+
+       return count;
+}
+
+static int sbp_update_unit_directory(struct sbp_tport *tport)
+{
+       int num_luns, num_entries, idx = 0, mgt_agt_addr, ret, i;
+       u32 *data;
+
+       if (tport->unit_directory.data) {
+               fw_core_remove_descriptor(&tport->unit_directory);
+               kfree(tport->unit_directory.data);
+               tport->unit_directory.data = NULL;
+       }
+
+       if (!tport->enable || !tport->tpg)
+               return 0;
+
+       num_luns = sbp_count_se_tpg_luns(&tport->tpg->se_tpg);
+
+       /*
+        * Number of entries in the final unit directory:
+        *  - all of those in the template
+        *  - management_agent
+        *  - unit_characteristics
+        *  - reconnect_timeout
+        *  - unit unique ID
+        *  - one for each LUN
+        *
+        *  MUST NOT include leaf or sub-directory entries
+        */
+       num_entries = ARRAY_SIZE(sbp_unit_directory_template) + 4 + num_luns;
+
+       if (tport->directory_id != -1)
+               num_entries++;
+
+       /* allocate num_entries + 4 for the header and unique ID leaf */
+       data = kcalloc((num_entries + 4), sizeof(u32), GFP_KERNEL);
+       if (!data)
+               return -ENOMEM;
+
+       /* directory_length */
+       data[idx++] = num_entries << 16;
+
+       /* directory_id */
+       if (tport->directory_id != -1)
+               data[idx++] = (CSR_DIRECTORY_ID << 24) | tport->directory_id;
+
+       /* unit directory template */
+       memcpy(&data[idx], sbp_unit_directory_template,
+                       sizeof(sbp_unit_directory_template));
+       idx += ARRAY_SIZE(sbp_unit_directory_template);
+
+       /* management_agent */
+       mgt_agt_addr = (tport->mgt_agt->handler.offset - CSR_REGISTER_BASE) / 4;
+       data[idx++] = 0x54000000 | (mgt_agt_addr & 0x00ffffff);
+
+       /* unit_characteristics */
+       data[idx++] = 0x3a000000 |
+               (((tport->mgt_orb_timeout * 2) << 8) & 0xff00) |
+               SBP_ORB_FETCH_SIZE;
+
+       /* reconnect_timeout */
+       data[idx++] = 0x3d000000 | (tport->max_reconnect_timeout & 0xffff);
+
+       /* unit unique ID (leaf is just after LUNs) */
+       data[idx++] = 0x8d000000 | (num_luns + 1);
+
+       spin_lock(&tport->tpg->se_tpg.tpg_lun_lock);
+       for (i = 0; i < TRANSPORT_MAX_LUNS_PER_TPG; i++) {
+               struct se_lun *se_lun = tport->tpg->se_tpg.tpg_lun_list[i];
+               struct se_device *dev;
+               int type;
+
+               if (se_lun->lun_status == TRANSPORT_LUN_STATUS_FREE)
+                       continue;
+
+               spin_unlock(&tport->tpg->se_tpg.tpg_lun_lock);
+
+               dev = se_lun->lun_se_dev;
+               type = dev->transport->get_device_type(dev);
+
+               /* logical_unit_number */
+               data[idx++] = 0x14000000 |
+                       ((type << 16) & 0x1f0000) |
+                       (se_lun->unpacked_lun & 0xffff);
+
+               spin_lock(&tport->tpg->se_tpg.tpg_lun_lock);
+       }
+       spin_unlock(&tport->tpg->se_tpg.tpg_lun_lock);
+
+       /* unit unique ID leaf */
+       data[idx++] = 2 << 16;
+       data[idx++] = tport->guid >> 32;
+       data[idx++] = tport->guid;
+
+       tport->unit_directory.length = idx;
+       tport->unit_directory.key = (CSR_DIRECTORY | CSR_UNIT) << 24;
+       tport->unit_directory.data = data;
+
+       ret = fw_core_add_descriptor(&tport->unit_directory);
+       if (ret < 0) {
+               kfree(tport->unit_directory.data);
+               tport->unit_directory.data = NULL;
+       }
+
+       return ret;
+}
+
+static ssize_t sbp_parse_wwn(const char *name, u64 *wwn)
+{
+       const char *cp;
+       char c, nibble;
+       int pos = 0, err;
+
+       *wwn = 0;
+       for (cp = name; cp < &name[SBP_NAMELEN - 1]; cp++) {
+               c = *cp;
+               if (c == '\n' && cp[1] == '\0')
+                       continue;
+               if (c == '\0') {
+                       err = 2;
+                       if (pos != 16)
+                               goto fail;
+                       return cp - name;
+               }
+               err = 3;
+               if (isdigit(c))
+                       nibble = c - '0';
+               else if (isxdigit(c))
+                       nibble = tolower(c) - 'a' + 10;
+               else
+                       goto fail;
+               *wwn = (*wwn << 4) | nibble;
+               pos++;
+       }
+       err = 4;
+fail:
+       printk(KERN_INFO "err %u len %zu pos %u\n",
+                       err, cp - name, pos);
+       return -1;
+}
+
+static ssize_t sbp_format_wwn(char *buf, size_t len, u64 wwn)
+{
+       return snprintf(buf, len, "%016llx", wwn);
+}
+
+static struct se_node_acl *sbp_make_nodeacl(
+               struct se_portal_group *se_tpg,
+               struct config_group *group,
+               const char *name)
+{
+       struct se_node_acl *se_nacl, *se_nacl_new;
+       struct sbp_nacl *nacl;
+       u64 guid = 0;
+       u32 nexus_depth = 1;
+
+       if (sbp_parse_wwn(name, &guid) < 0)
+               return ERR_PTR(-EINVAL);
+
+       se_nacl_new = sbp_alloc_fabric_acl(se_tpg);
+       if (!se_nacl_new)
+               return ERR_PTR(-ENOMEM);
+
+       /*
+        * se_nacl_new may be released by core_tpg_add_initiator_node_acl()
+        * when converting a NodeACL from demo mode -> explict
+        */
+       se_nacl = core_tpg_add_initiator_node_acl(se_tpg, se_nacl_new,
+                       name, nexus_depth);
+       if (IS_ERR(se_nacl)) {
+               sbp_release_fabric_acl(se_tpg, se_nacl_new);
+               return se_nacl;
+       }
+
+       nacl = container_of(se_nacl, struct sbp_nacl, se_node_acl);
+       nacl->guid = guid;
+       sbp_format_wwn(nacl->iport_name, SBP_NAMELEN, guid);
+
+       return se_nacl;
+}
+
+static void sbp_drop_nodeacl(struct se_node_acl *se_acl)
+{
+       struct sbp_nacl *nacl =
+               container_of(se_acl, struct sbp_nacl, se_node_acl);
+
+       core_tpg_del_initiator_node_acl(se_acl->se_tpg, se_acl, 1);
+       kfree(nacl);
+}
+
+static int sbp_post_link_lun(
+               struct se_portal_group *se_tpg,
+               struct se_lun *se_lun)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+
+       return sbp_update_unit_directory(tpg->tport);
+}
+
+static void sbp_pre_unlink_lun(
+               struct se_portal_group *se_tpg,
+               struct se_lun *se_lun)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+       struct sbp_tport *tport = tpg->tport;
+       int ret;
+
+       if (sbp_count_se_tpg_luns(&tpg->se_tpg) == 0)
+               tport->enable = 0;
+
+       ret = sbp_update_unit_directory(tport);
+       if (ret < 0)
+               pr_err("unlink LUN: failed to update unit directory\n");
+}
+
+static struct se_portal_group *sbp_make_tpg(
+               struct se_wwn *wwn,
+               struct config_group *group,
+               const char *name)
+{
+       struct sbp_tport *tport =
+               container_of(wwn, struct sbp_tport, tport_wwn);
+
+       struct sbp_tpg *tpg;
+       unsigned long tpgt;
+       int ret;
+
+       if (strstr(name, "tpgt_") != name)
+               return ERR_PTR(-EINVAL);
+       if (kstrtoul(name + 5, 10, &tpgt) || tpgt > UINT_MAX)
+               return ERR_PTR(-EINVAL);
+
+       if (tport->tpg) {
+               pr_err("Only one TPG per Unit is possible.\n");
+               return ERR_PTR(-EBUSY);
+       }
+
+       tpg = kzalloc(sizeof(*tpg), GFP_KERNEL);
+       if (!tpg) {
+               pr_err("Unable to allocate struct sbp_tpg\n");
+               return ERR_PTR(-ENOMEM);
+       }
+
+       tpg->tport = tport;
+       tpg->tport_tpgt = tpgt;
+       tport->tpg = tpg;
+
+       /* default attribute values */
+       tport->enable = 0;
+       tport->directory_id = -1;
+       tport->mgt_orb_timeout = 15;
+       tport->max_reconnect_timeout = 5;
+       tport->max_logins_per_lun = 1;
+
+       tport->mgt_agt = sbp_management_agent_register(tport);
+       if (IS_ERR(tport->mgt_agt)) {
+               ret = PTR_ERR(tport->mgt_agt);
+               goto out_free_tpg;
+       }
+
+       ret = core_tpg_register(&sbp_ops, wwn, &tpg->se_tpg, tpg,
+                       TRANSPORT_TPG_TYPE_NORMAL);
+       if (ret < 0)
+               goto out_unreg_mgt_agt;
+
+       return &tpg->se_tpg;
+
+out_unreg_mgt_agt:
+       sbp_management_agent_unregister(tport->mgt_agt);
+out_free_tpg:
+       tport->tpg = NULL;
+       kfree(tpg);
+       return ERR_PTR(ret);
+}
+
+static void sbp_drop_tpg(struct se_portal_group *se_tpg)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+       struct sbp_tport *tport = tpg->tport;
+
+       core_tpg_deregister(se_tpg);
+       sbp_management_agent_unregister(tport->mgt_agt);
+       tport->tpg = NULL;
+       kfree(tpg);
+}
+
+static struct se_wwn *sbp_make_tport(
+               struct target_fabric_configfs *tf,
+               struct config_group *group,
+               const char *name)
+{
+       struct sbp_tport *tport;
+       u64 guid = 0;
+
+       if (sbp_parse_wwn(name, &guid) < 0)
+               return ERR_PTR(-EINVAL);
+
+       tport = kzalloc(sizeof(*tport), GFP_KERNEL);
+       if (!tport) {
+               pr_err("Unable to allocate struct sbp_tport\n");
+               return ERR_PTR(-ENOMEM);
+       }
+
+       tport->guid = guid;
+       sbp_format_wwn(tport->tport_name, SBP_NAMELEN, guid);
+
+       return &tport->tport_wwn;
+}
+
+static void sbp_drop_tport(struct se_wwn *wwn)
+{
+       struct sbp_tport *tport =
+               container_of(wwn, struct sbp_tport, tport_wwn);
+
+       kfree(tport);
+}
+
+static ssize_t sbp_wwn_show_attr_version(
+               struct target_fabric_configfs *tf,
+               char *page)
+{
+       return sprintf(page, "FireWire SBP fabric module %s\n", SBP_VERSION);
+}
+
+TF_WWN_ATTR_RO(sbp, version);
+
+static struct configfs_attribute *sbp_wwn_attrs[] = {
+       &sbp_wwn_version.attr,
+       NULL,
+};
+
+static ssize_t sbp_tpg_show_directory_id(
+               struct se_portal_group *se_tpg,
+               char *page)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+       struct sbp_tport *tport = tpg->tport;
+
+       if (tport->directory_id == -1)
+               return sprintf(page, "implicit\n");
+       else
+               return sprintf(page, "%06x\n", tport->directory_id);
+}
+
+static ssize_t sbp_tpg_store_directory_id(
+               struct se_portal_group *se_tpg,
+               const char *page,
+               size_t count)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+       struct sbp_tport *tport = tpg->tport;
+       unsigned long val;
+
+       if (tport->enable) {
+               pr_err("Cannot change the directory_id on an active target.\n");
+               return -EBUSY;
+       }
+
+       if (strstr(page, "implicit") == page) {
+               tport->directory_id = -1;
+       } else {
+               if (kstrtoul(page, 16, &val) < 0)
+                       return -EINVAL;
+               if (val > 0xffffff)
+                       return -EINVAL;
+
+               tport->directory_id = val;
+       }
+
+       return count;
+}
+
+static ssize_t sbp_tpg_show_enable(
+               struct se_portal_group *se_tpg,
+               char *page)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+       struct sbp_tport *tport = tpg->tport;
+       return sprintf(page, "%d\n", tport->enable);
+}
+
+static ssize_t sbp_tpg_store_enable(
+               struct se_portal_group *se_tpg,
+               const char *page,
+               size_t count)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+       struct sbp_tport *tport = tpg->tport;
+       unsigned long val;
+       int ret;
+
+       if (kstrtoul(page, 0, &val) < 0)
+               return -EINVAL;
+       if ((val != 0) && (val != 1))
+               return -EINVAL;
+
+       if (tport->enable == val)
+               return count;
+
+       if (val) {
+               if (sbp_count_se_tpg_luns(&tpg->se_tpg) == 0) {
+                       pr_err("Cannot enable a target with no LUNs!\n");
+                       return -EINVAL;
+               }
+       } else {
+               /* XXX: force-shutdown sessions instead? */
+               spin_lock_bh(&se_tpg->session_lock);
+               if (!list_empty(&se_tpg->tpg_sess_list)) {
+                       spin_unlock_bh(&se_tpg->session_lock);
+                       return -EBUSY;
+               }
+               spin_unlock_bh(&se_tpg->session_lock);
+       }
+
+       tport->enable = val;
+
+       ret = sbp_update_unit_directory(tport);
+       if (ret < 0) {
+               pr_err("Could not update Config ROM\n");
+               return ret;
+       }
+
+       return count;
+}
+
+TF_TPG_BASE_ATTR(sbp, directory_id, S_IRUGO | S_IWUSR);
+TF_TPG_BASE_ATTR(sbp, enable, S_IRUGO | S_IWUSR);
+
+static struct configfs_attribute *sbp_tpg_base_attrs[] = {
+       &sbp_tpg_directory_id.attr,
+       &sbp_tpg_enable.attr,
+       NULL,
+};
+
+static ssize_t sbp_tpg_attrib_show_mgt_orb_timeout(
+               struct se_portal_group *se_tpg,
+               char *page)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+       struct sbp_tport *tport = tpg->tport;
+       return sprintf(page, "%d\n", tport->mgt_orb_timeout);
+}
+
+static ssize_t sbp_tpg_attrib_store_mgt_orb_timeout(
+               struct se_portal_group *se_tpg,
+               const char *page,
+               size_t count)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+       struct sbp_tport *tport = tpg->tport;
+       unsigned long val;
+       int ret;
+
+       if (kstrtoul(page, 0, &val) < 0)
+               return -EINVAL;
+       if ((val < 1) || (val > 127))
+               return -EINVAL;
+
+       if (tport->mgt_orb_timeout == val)
+               return count;
+
+       tport->mgt_orb_timeout = val;
+
+       ret = sbp_update_unit_directory(tport);
+       if (ret < 0)
+               return ret;
+
+       return count;
+}
+
+static ssize_t sbp_tpg_attrib_show_max_reconnect_timeout(
+               struct se_portal_group *se_tpg,
+               char *page)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+       struct sbp_tport *tport = tpg->tport;
+       return sprintf(page, "%d\n", tport->max_reconnect_timeout);
+}
+
+static ssize_t sbp_tpg_attrib_store_max_reconnect_timeout(
+               struct se_portal_group *se_tpg,
+               const char *page,
+               size_t count)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+       struct sbp_tport *tport = tpg->tport;
+       unsigned long val;
+       int ret;
+
+       if (kstrtoul(page, 0, &val) < 0)
+               return -EINVAL;
+       if ((val < 1) || (val > 32767))
+               return -EINVAL;
+
+       if (tport->max_reconnect_timeout == val)
+               return count;
+
+       tport->max_reconnect_timeout = val;
+
+       ret = sbp_update_unit_directory(tport);
+       if (ret < 0)
+               return ret;
+
+       return count;
+}
+
+static ssize_t sbp_tpg_attrib_show_max_logins_per_lun(
+               struct se_portal_group *se_tpg,
+               char *page)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+       struct sbp_tport *tport = tpg->tport;
+       return sprintf(page, "%d\n", tport->max_logins_per_lun);
+}
+
+static ssize_t sbp_tpg_attrib_store_max_logins_per_lun(
+               struct se_portal_group *se_tpg,
+               const char *page,
+               size_t count)
+{
+       struct sbp_tpg *tpg = container_of(se_tpg, struct sbp_tpg, se_tpg);
+       struct sbp_tport *tport = tpg->tport;
+       unsigned long val;
+
+       if (kstrtoul(page, 0, &val) < 0)
+               return -EINVAL;
+       if ((val < 1) || (val > 127))
+               return -EINVAL;
+
+       /* XXX: also check against current count? */
+
+       tport->max_logins_per_lun = val;
+
+       return count;
+}
+
+TF_TPG_ATTRIB_ATTR(sbp, mgt_orb_timeout, S_IRUGO | S_IWUSR);
+TF_TPG_ATTRIB_ATTR(sbp, max_reconnect_timeout, S_IRUGO | S_IWUSR);
+TF_TPG_ATTRIB_ATTR(sbp, max_logins_per_lun, S_IRUGO | S_IWUSR);
+
+static struct configfs_attribute *sbp_tpg_attrib_attrs[] = {
+       &sbp_tpg_attrib_mgt_orb_timeout.attr,
+       &sbp_tpg_attrib_max_reconnect_timeout.attr,
+       &sbp_tpg_attrib_max_logins_per_lun.attr,
+       NULL,
+};
+
+static const struct target_core_fabric_ops sbp_ops = {
+       .module                         = THIS_MODULE,
+       .name                           = "sbp",
+       .get_fabric_name                = sbp_get_fabric_name,
+       .get_fabric_proto_ident         = sbp_get_fabric_proto_ident,
+       .tpg_get_wwn                    = sbp_get_fabric_wwn,
+       .tpg_get_tag                    = sbp_get_tag,
+       .tpg_get_default_depth          = sbp_get_default_depth,
+       .tpg_get_pr_transport_id        = sbp_get_pr_transport_id,
+       .tpg_get_pr_transport_id_len    = sbp_get_pr_transport_id_len,
+       .tpg_parse_pr_out_transport_id  = sbp_parse_pr_out_transport_id,
+       .tpg_check_demo_mode            = sbp_check_true,
+       .tpg_check_demo_mode_cache      = sbp_check_true,
+       .tpg_check_demo_mode_write_protect = sbp_check_false,
+       .tpg_check_prod_mode_write_protect = sbp_check_false,
+       .tpg_alloc_fabric_acl           = sbp_alloc_fabric_acl,
+       .tpg_release_fabric_acl         = sbp_release_fabric_acl,
+       .tpg_get_inst_index             = sbp_tpg_get_inst_index,
+       .release_cmd                    = sbp_release_cmd,
+       .shutdown_session               = sbp_shutdown_session,
+       .close_session                  = sbp_close_session,
+       .sess_get_index                 = sbp_sess_get_index,
+       .write_pending                  = sbp_write_pending,
+       .write_pending_status           = sbp_write_pending_status,
+       .set_default_node_attributes    = sbp_set_default_node_attrs,
+       .get_task_tag                   = sbp_get_task_tag,
+       .get_cmd_state                  = sbp_get_cmd_state,
+       .queue_data_in                  = sbp_queue_data_in,
+       .queue_status                   = sbp_queue_status,
+       .queue_tm_rsp                   = sbp_queue_tm_rsp,
+       .aborted_task                   = sbp_aborted_task,
+       .check_stop_free                = sbp_check_stop_free,
+
+       .fabric_make_wwn                = sbp_make_tport,
+       .fabric_drop_wwn                = sbp_drop_tport,
+       .fabric_make_tpg                = sbp_make_tpg,
+       .fabric_drop_tpg                = sbp_drop_tpg,
+       .fabric_post_link               = sbp_post_link_lun,
+       .fabric_pre_unlink              = sbp_pre_unlink_lun,
+       .fabric_make_np                 = NULL,
+       .fabric_drop_np                 = NULL,
+       .fabric_make_nodeacl            = sbp_make_nodeacl,
+       .fabric_drop_nodeacl            = sbp_drop_nodeacl,
+
+       .tfc_wwn_attrs                  = sbp_wwn_attrs,
+       .tfc_tpg_base_attrs             = sbp_tpg_base_attrs,
+       .tfc_tpg_attrib_attrs           = sbp_tpg_attrib_attrs,
+};
+
+static int __init sbp_init(void)
+{
+       return target_register_template(&sbp_ops);
+};
+
+static void __exit sbp_exit(void)
+{
+       target_unregister_template(&sbp_ops);
+};
+
+MODULE_DESCRIPTION("FireWire SBP fabric driver");
+MODULE_LICENSE("GPL");
+module_init(sbp_init);
+module_exit(sbp_exit);