]> www.pilppa.org Git - linux-2.6-omap-h63xx.git/commitdiff
dm mpath: rdac
authorChandra Seetharaman <sekharan@us.ibm.com>
Thu, 12 Jul 2007 16:30:05 +0000 (17:30 +0100)
committerLinus Torvalds <torvalds@woody.linux-foundation.org>
Thu, 12 Jul 2007 22:01:23 +0000 (15:01 -0700)
This patch supports LSI/Engenio devices in RDAC mode. Like dm-emc
it requires userspace support. In your multipath.conf file you must have:

path_checker            rdac
hardware_handler        "1 rdac"
prio_callout "/sbin/mpath_prio_tpc /dev/%n"

And you also then must have a updated multipath tools release which
has rdac support.

Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
drivers/md/Kconfig
drivers/md/Makefile
drivers/md/dm-mpath-rdac.c [new file with mode: 0644]

index 7df934d69134436cc2e1f5affb7f681dc0ab1671..466909f38d9897c11d74c9193005e084066daeaf 100644 (file)
@@ -262,6 +262,12 @@ config DM_MULTIPATH_EMC
        ---help---
          Multipath support for EMC CX/AX series hardware.
 
+config DM_MULTIPATH_RDAC
+       tristate "LSI/Engenio RDAC multipath support (EXPERIMENTAL)"
+       depends on DM_MULTIPATH && BLK_DEV_DM && EXPERIMENTAL
+       ---help---
+         Multipath support for LSI/Engenio RDAC.
+
 config DM_DELAY
        tristate "I/O delaying target (EXPERIMENTAL)"
        depends on BLK_DEV_DM && EXPERIMENTAL
index 38754084eac741462bc677f8c6911b662c719623..2c45d7683ae9b530328e76afa3a7e36f4e47e04a 100644 (file)
@@ -7,6 +7,7 @@ dm-mod-objs     := dm.o dm-table.o dm-target.o dm-linear.o dm-stripe.o \
 dm-multipath-objs := dm-hw-handler.o dm-path-selector.o dm-mpath.o
 dm-snapshot-objs := dm-snap.o dm-exception-store.o
 dm-mirror-objs := dm-log.o dm-raid1.o
+dm-rdac-objs   := dm-mpath-rdac.o
 md-mod-objs     := md.o bitmap.o
 raid456-objs   := raid5.o raid6algos.o raid6recov.o raid6tables.o \
                   raid6int1.o raid6int2.o raid6int4.o \
@@ -34,6 +35,7 @@ obj-$(CONFIG_DM_CRYPT)                += dm-crypt.o
 obj-$(CONFIG_DM_DELAY)         += dm-delay.o
 obj-$(CONFIG_DM_MULTIPATH)     += dm-multipath.o dm-round-robin.o
 obj-$(CONFIG_DM_MULTIPATH_EMC) += dm-emc.o
+obj-$(CONFIG_DM_MULTIPATH_RDAC)        += dm-rdac.o
 obj-$(CONFIG_DM_SNAPSHOT)      += dm-snapshot.o
 obj-$(CONFIG_DM_MIRROR)                += dm-mirror.o
 obj-$(CONFIG_DM_ZERO)          += dm-zero.o
diff --git a/drivers/md/dm-mpath-rdac.c b/drivers/md/dm-mpath-rdac.c
new file mode 100644 (file)
index 0000000..8b776b8
--- /dev/null
@@ -0,0 +1,700 @@
+/*
+ * Engenio/LSI RDAC DM HW handler
+ *
+ * Copyright (C) 2005 Mike Christie. All rights reserved.
+ * Copyright (C) Chandra Seetharaman, IBM Corp. 2007
+ *
+ * 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.
+ *
+ */
+#include <scsi/scsi.h>
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_eh.h>
+
+#define DM_MSG_PREFIX "multipath rdac"
+
+#include "dm.h"
+#include "dm-hw-handler.h"
+
+#define RDAC_DM_HWH_NAME "rdac"
+#define RDAC_DM_HWH_VER "0.4"
+
+/*
+ * LSI mode page stuff
+ *
+ * These struct definitions and the forming of the
+ * mode page were taken from the LSI RDAC 2.4 GPL'd
+ * driver, and then converted to Linux conventions.
+ */
+#define RDAC_QUIESCENCE_TIME 20;
+/*
+ * Page Codes
+ */
+#define RDAC_PAGE_CODE_REDUNDANT_CONTROLLER 0x2c
+
+/*
+ * Controller modes definitions
+ */
+#define RDAC_MODE_TRANSFER_ALL_LUNS            0x01
+#define RDAC_MODE_TRANSFER_SPECIFIED_LUNS      0x02
+
+/*
+ * RDAC Options field
+ */
+#define RDAC_FORCED_QUIESENCE 0x02
+
+#define RDAC_FAILOVER_TIMEOUT (60 * HZ)
+
+struct rdac_mode_6_hdr {
+       u8      data_len;
+       u8      medium_type;
+       u8      device_params;
+       u8      block_desc_len;
+};
+
+struct rdac_mode_10_hdr {
+       u16     data_len;
+       u8      medium_type;
+       u8      device_params;
+       u16     reserved;
+       u16     block_desc_len;
+};
+
+struct rdac_mode_common {
+       u8      controller_serial[16];
+       u8      alt_controller_serial[16];
+       u8      rdac_mode[2];
+       u8      alt_rdac_mode[2];
+       u8      quiescence_timeout;
+       u8      rdac_options;
+};
+
+struct rdac_pg_legacy {
+       struct rdac_mode_6_hdr hdr;
+       u8      page_code;
+       u8      page_len;
+       struct rdac_mode_common common;
+#define MODE6_MAX_LUN  32
+       u8      lun_table[MODE6_MAX_LUN];
+       u8      reserved2[32];
+       u8      reserved3;
+       u8      reserved4;
+};
+
+struct rdac_pg_expanded {
+       struct rdac_mode_10_hdr hdr;
+       u8      page_code;
+       u8      subpage_code;
+       u8      page_len[2];
+       struct rdac_mode_common common;
+       u8      lun_table[256];
+       u8      reserved3;
+       u8      reserved4;
+};
+
+struct c9_inquiry {
+       u8      peripheral_info;
+       u8      page_code;      /* 0xC9 */
+       u8      reserved1;
+       u8      page_len;
+       u8      page_id[4];     /* "vace" */
+       u8      avte_cvp;
+       u8      path_prio;
+       u8      reserved2[38];
+};
+
+#define SUBSYS_ID_LEN  16
+#define SLOT_ID_LEN    2
+
+struct c4_inquiry {
+       u8      peripheral_info;
+       u8      page_code;      /* 0xC4 */
+       u8      reserved1;
+       u8      page_len;
+       u8      page_id[4];     /* "subs" */
+       u8      subsys_id[SUBSYS_ID_LEN];
+       u8      revision[4];
+       u8      slot_id[SLOT_ID_LEN];
+       u8      reserved[2];
+};
+
+struct rdac_controller {
+       u8                      subsys_id[SUBSYS_ID_LEN];
+       u8                      slot_id[SLOT_ID_LEN];
+       int                     use_10_ms;
+       struct kref             kref;
+       struct list_head        node; /* list of all controllers */
+       spinlock_t              lock;
+       int                     submitted;
+       struct list_head        cmd_list; /* list of commands to be submitted */
+       union                   {
+               struct rdac_pg_legacy legacy;
+               struct rdac_pg_expanded expanded;
+       } mode_select;
+};
+struct c8_inquiry {
+       u8      peripheral_info;
+       u8      page_code; /* 0xC8 */
+       u8      reserved1;
+       u8      page_len;
+       u8      page_id[4]; /* "edid" */
+       u8      reserved2[3];
+       u8      vol_uniq_id_len;
+       u8      vol_uniq_id[16];
+       u8      vol_user_label_len;
+       u8      vol_user_label[60];
+       u8      array_uniq_id_len;
+       u8      array_unique_id[16];
+       u8      array_user_label_len;
+       u8      array_user_label[60];
+       u8      lun[8];
+};
+
+struct c2_inquiry {
+       u8      peripheral_info;
+       u8      page_code;      /* 0xC2 */
+       u8      reserved1;
+       u8      page_len;
+       u8      page_id[4];     /* "swr4" */
+       u8      sw_version[3];
+       u8      sw_date[3];
+       u8      features_enabled;
+       u8      max_lun_supported;
+       u8      partitions[239]; /* Total allocation length should be 0xFF */
+};
+
+struct rdac_handler {
+       struct list_head        entry; /* list waiting to submit MODE SELECT */
+       unsigned                timeout;
+       struct rdac_controller  *ctlr;
+#define UNINITIALIZED_LUN      (1 << 8)
+       unsigned                lun;
+       unsigned char           sense[SCSI_SENSE_BUFFERSIZE];
+       struct dm_path          *path;
+       struct work_struct      work;
+#define        SEND_C2_INQUIRY         1
+#define        SEND_C4_INQUIRY         2
+#define        SEND_C8_INQUIRY         3
+#define        SEND_C9_INQUIRY         4
+#define        SEND_MODE_SELECT        5
+       int                     cmd_to_send;
+       union                   {
+               struct c2_inquiry c2;
+               struct c4_inquiry c4;
+               struct c8_inquiry c8;
+               struct c9_inquiry c9;
+       } inq;
+};
+
+static LIST_HEAD(ctlr_list);
+static DEFINE_SPINLOCK(list_lock);
+static struct workqueue_struct *rdac_wkqd;
+
+static inline int had_failures(struct request *req, int error)
+{
+       return (error || host_byte(req->errors) != DID_OK ||
+                       msg_byte(req->errors) != COMMAND_COMPLETE);
+}
+
+static void rdac_resubmit_all(struct rdac_handler *h)
+{
+       struct rdac_controller *ctlr = h->ctlr;
+       struct rdac_handler *tmp, *h1;
+
+       spin_lock(&ctlr->lock);
+       list_for_each_entry_safe(h1, tmp, &ctlr->cmd_list, entry) {
+               h1->cmd_to_send = SEND_C9_INQUIRY;
+               queue_work(rdac_wkqd, &h1->work);
+               list_del(&h1->entry);
+       }
+       ctlr->submitted = 0;
+       spin_unlock(&ctlr->lock);
+}
+
+static void mode_select_endio(struct request *req, int error)
+{
+       struct rdac_handler *h = req->end_io_data;
+       struct scsi_sense_hdr sense_hdr;
+       int sense = 0, fail = 0;
+
+       if (had_failures(req, error)) {
+               fail = 1;
+               goto failed;
+       }
+
+       if (status_byte(req->errors) == CHECK_CONDITION) {
+               scsi_normalize_sense(req->sense, SCSI_SENSE_BUFFERSIZE,
+                               &sense_hdr);
+               sense = (sense_hdr.sense_key << 16) | (sense_hdr.asc << 8) |
+                               sense_hdr.ascq;
+               /* If it is retryable failure, submit the c9 inquiry again */
+               if (sense == 0x59136 || sense == 0x68b02 || sense == 0xb8b02 ||
+                   sense == 0x62900) {
+                       /* 0x59136    - Command lock contention
+                        * 0x[6b]8b02 - Quiesense in progress or achieved
+                        * 0x62900    - Power On, Reset, or Bus Device Reset
+                        */
+                       h->cmd_to_send = SEND_C9_INQUIRY;
+                       queue_work(rdac_wkqd, &h->work);
+                       goto done;
+               }
+               if (sense)
+                       DMINFO("MODE_SELECT failed on %s with sense 0x%x",
+                                               h->path->dev->name, sense);
+       }
+failed:
+       if (fail || sense)
+               dm_pg_init_complete(h->path, MP_FAIL_PATH);
+       else
+               dm_pg_init_complete(h->path, 0);
+
+done:
+       rdac_resubmit_all(h);
+       __blk_put_request(req->q, req);
+}
+
+static struct request *get_rdac_req(struct rdac_handler *h,
+                       void *buffer, unsigned buflen, int rw)
+{
+       struct request *rq;
+       struct request_queue *q = bdev_get_queue(h->path->dev->bdev);
+
+       rq = blk_get_request(q, rw, GFP_KERNEL);
+
+       if (!rq) {
+               DMINFO("get_rdac_req: blk_get_request failed");
+               return NULL;
+       }
+
+       if (buflen && blk_rq_map_kern(q, rq, buffer, buflen, GFP_KERNEL)) {
+               blk_put_request(rq);
+               DMINFO("get_rdac_req: blk_rq_map_kern failed");
+               return NULL;
+       }
+
+       memset(&rq->cmd, 0, BLK_MAX_CDB);
+       rq->sense = h->sense;
+       memset(rq->sense, 0, SCSI_SENSE_BUFFERSIZE);
+       rq->sense_len = 0;
+
+       rq->end_io_data = h;
+       rq->timeout = h->timeout;
+       rq->cmd_type = REQ_TYPE_BLOCK_PC;
+       rq->cmd_flags = REQ_FAILFAST | REQ_NOMERGE;
+       return rq;
+}
+
+static struct request *rdac_failover_get(struct rdac_handler *h)
+{
+       struct request *rq;
+       struct rdac_mode_common *common;
+       unsigned data_size;
+
+       if (h->ctlr->use_10_ms) {
+               struct rdac_pg_expanded *rdac_pg;
+
+               data_size = sizeof(struct rdac_pg_expanded);
+               rdac_pg = &h->ctlr->mode_select.expanded;
+               memset(rdac_pg, 0, data_size);
+               common = &rdac_pg->common;
+               rdac_pg->page_code = RDAC_PAGE_CODE_REDUNDANT_CONTROLLER + 0x40;
+               rdac_pg->subpage_code = 0x1;
+               rdac_pg->page_len[0] = 0x01;
+               rdac_pg->page_len[1] = 0x28;
+               rdac_pg->lun_table[h->lun] = 0x81;
+       } else {
+               struct rdac_pg_legacy *rdac_pg;
+
+               data_size = sizeof(struct rdac_pg_legacy);
+               rdac_pg = &h->ctlr->mode_select.legacy;
+               memset(rdac_pg, 0, data_size);
+               common = &rdac_pg->common;
+               rdac_pg->page_code = RDAC_PAGE_CODE_REDUNDANT_CONTROLLER;
+               rdac_pg->page_len = 0x68;
+               rdac_pg->lun_table[h->lun] = 0x81;
+       }
+       common->rdac_mode[1] = RDAC_MODE_TRANSFER_SPECIFIED_LUNS;
+       common->quiescence_timeout = RDAC_QUIESCENCE_TIME;
+       common->rdac_options = RDAC_FORCED_QUIESENCE;
+
+       /* get request for block layer packet command */
+       rq = get_rdac_req(h, &h->ctlr->mode_select, data_size, WRITE);
+       if (!rq) {
+               DMERR("rdac_failover_get: no rq");
+               return NULL;
+       }
+
+       /* Prepare the command. */
+       if (h->ctlr->use_10_ms) {
+               rq->cmd[0] = MODE_SELECT_10;
+               rq->cmd[7] = data_size >> 8;
+               rq->cmd[8] = data_size & 0xff;
+       } else {
+               rq->cmd[0] = MODE_SELECT;
+               rq->cmd[4] = data_size;
+       }
+       rq->cmd_len = COMMAND_SIZE(rq->cmd[0]);
+
+       return rq;
+}
+
+/* Acquires h->ctlr->lock */
+static void submit_mode_select(struct rdac_handler *h)
+{
+       struct request *rq;
+       struct request_queue *q = bdev_get_queue(h->path->dev->bdev);
+
+       spin_lock(&h->ctlr->lock);
+       if (h->ctlr->submitted) {
+               list_add(&h->entry, &h->ctlr->cmd_list);
+               goto drop_lock;
+       }
+
+       if (!q) {
+               DMINFO("submit_mode_select: no queue");
+               goto fail_path;
+       }
+
+       rq = rdac_failover_get(h);
+       if (!rq) {
+               DMERR("submit_mode_select: no rq");
+               goto fail_path;
+       }
+
+       DMINFO("queueing MODE_SELECT command on %s", h->path->dev->name);
+
+       blk_execute_rq_nowait(q, NULL, rq, 1, mode_select_endio);
+       h->ctlr->submitted = 1;
+       goto drop_lock;
+fail_path:
+       dm_pg_init_complete(h->path, MP_FAIL_PATH);
+drop_lock:
+       spin_unlock(&h->ctlr->lock);
+}
+
+static void release_ctlr(struct kref *kref)
+{
+       struct rdac_controller *ctlr;
+       ctlr = container_of(kref, struct rdac_controller, kref);
+
+       spin_lock(&list_lock);
+       list_del(&ctlr->node);
+       spin_unlock(&list_lock);
+       kfree(ctlr);
+}
+
+static struct rdac_controller *get_controller(u8 *subsys_id, u8 *slot_id)
+{
+       struct rdac_controller *ctlr, *tmp;
+
+       spin_lock(&list_lock);
+
+       list_for_each_entry(tmp, &ctlr_list, node) {
+               if ((memcmp(tmp->subsys_id, subsys_id, SUBSYS_ID_LEN) == 0) &&
+                         (memcmp(tmp->slot_id, slot_id, SLOT_ID_LEN) == 0)) {
+                       kref_get(&tmp->kref);
+                       spin_unlock(&list_lock);
+                       return tmp;
+               }
+       }
+       ctlr = kmalloc(sizeof(*ctlr), GFP_ATOMIC);
+       if (!ctlr)
+               goto done;
+
+       /* initialize fields of controller */
+       memcpy(ctlr->subsys_id, subsys_id, SUBSYS_ID_LEN);
+       memcpy(ctlr->slot_id, slot_id, SLOT_ID_LEN);
+       kref_init(&ctlr->kref);
+       spin_lock_init(&ctlr->lock);
+       ctlr->submitted = 0;
+       ctlr->use_10_ms = -1;
+       INIT_LIST_HEAD(&ctlr->cmd_list);
+       list_add(&ctlr->node, &ctlr_list);
+done:
+       spin_unlock(&list_lock);
+       return ctlr;
+}
+
+static void c4_endio(struct request *req, int error)
+{
+       struct rdac_handler *h = req->end_io_data;
+       struct c4_inquiry *sp;
+
+       if (had_failures(req, error)) {
+               dm_pg_init_complete(h->path, MP_FAIL_PATH);
+               goto done;
+       }
+
+       sp = &h->inq.c4;
+
+       h->ctlr = get_controller(sp->subsys_id, sp->slot_id);
+
+       if (h->ctlr) {
+               h->cmd_to_send = SEND_C9_INQUIRY;
+               queue_work(rdac_wkqd, &h->work);
+       } else
+               dm_pg_init_complete(h->path, MP_FAIL_PATH);
+done:
+       __blk_put_request(req->q, req);
+}
+
+static void c2_endio(struct request *req, int error)
+{
+       struct rdac_handler *h = req->end_io_data;
+       struct c2_inquiry *sp;
+
+       if (had_failures(req, error)) {
+               dm_pg_init_complete(h->path, MP_FAIL_PATH);
+               goto done;
+       }
+
+       sp = &h->inq.c2;
+
+       /* If more than MODE6_MAX_LUN luns are supported, use mode select 10 */
+       if (sp->max_lun_supported >= MODE6_MAX_LUN)
+               h->ctlr->use_10_ms = 1;
+       else
+               h->ctlr->use_10_ms = 0;
+
+       h->cmd_to_send = SEND_MODE_SELECT;
+       queue_work(rdac_wkqd, &h->work);
+done:
+       __blk_put_request(req->q, req);
+}
+
+static void c9_endio(struct request *req, int error)
+{
+       struct rdac_handler *h = req->end_io_data;
+       struct c9_inquiry *sp;
+
+       if (had_failures(req, error)) {
+               dm_pg_init_complete(h->path, MP_FAIL_PATH);
+               goto done;
+       }
+
+       /* We need to look at the sense keys here to take clear action.
+        * For now simple logic: If the host is in AVT mode or if controller
+        * owns the lun, return dm_pg_init_complete(), otherwise submit
+        * MODE SELECT.
+        */
+       sp = &h->inq.c9;
+
+       /* If in AVT mode, return success */
+       if ((sp->avte_cvp >> 7) == 0x1) {
+               dm_pg_init_complete(h->path, 0);
+               goto done;
+       }
+
+       /* If the controller on this path owns the LUN, return success */
+       if (sp->avte_cvp & 0x1) {
+               dm_pg_init_complete(h->path, 0);
+               goto done;
+       }
+
+       if (h->ctlr) {
+               if (h->ctlr->use_10_ms == -1)
+                       h->cmd_to_send = SEND_C2_INQUIRY;
+               else
+                       h->cmd_to_send = SEND_MODE_SELECT;
+       } else
+               h->cmd_to_send = SEND_C4_INQUIRY;
+       queue_work(rdac_wkqd, &h->work);
+done:
+       __blk_put_request(req->q, req);
+}
+
+static void c8_endio(struct request *req, int error)
+{
+       struct rdac_handler *h = req->end_io_data;
+       struct c8_inquiry *sp;
+
+       if (had_failures(req, error)) {
+               dm_pg_init_complete(h->path, MP_FAIL_PATH);
+               goto done;
+       }
+
+       /* We need to look at the sense keys here to take clear action.
+        * For now simple logic: Get the lun from the inquiry page.
+        */
+       sp = &h->inq.c8;
+       h->lun = sp->lun[7]; /* currently it uses only one byte */
+       h->cmd_to_send = SEND_C9_INQUIRY;
+       queue_work(rdac_wkqd, &h->work);
+done:
+       __blk_put_request(req->q, req);
+}
+
+static void submit_inquiry(struct rdac_handler *h, int page_code,
+               unsigned int len, rq_end_io_fn endio)
+{
+       struct request *rq;
+       struct request_queue *q = bdev_get_queue(h->path->dev->bdev);
+
+       if (!q)
+               goto fail_path;
+
+       rq = get_rdac_req(h, &h->inq, len, READ);
+       if (!rq)
+               goto fail_path;
+
+       /* Prepare the command. */
+       rq->cmd[0] = INQUIRY;
+       rq->cmd[1] = 1;
+       rq->cmd[2] = page_code;
+       rq->cmd[4] = len;
+       rq->cmd_len = COMMAND_SIZE(INQUIRY);
+       blk_execute_rq_nowait(q, NULL, rq, 1, endio);
+       return;
+
+fail_path:
+       dm_pg_init_complete(h->path, MP_FAIL_PATH);
+}
+
+static void service_wkq(struct work_struct *work)
+{
+       struct rdac_handler *h = container_of(work, struct rdac_handler, work);
+
+       switch (h->cmd_to_send) {
+       case SEND_C2_INQUIRY:
+               submit_inquiry(h, 0xC2, sizeof(struct c2_inquiry), c2_endio);
+               break;
+       case SEND_C4_INQUIRY:
+               submit_inquiry(h, 0xC4, sizeof(struct c4_inquiry), c4_endio);
+               break;
+       case SEND_C8_INQUIRY:
+               submit_inquiry(h, 0xC8, sizeof(struct c8_inquiry), c8_endio);
+               break;
+       case SEND_C9_INQUIRY:
+               submit_inquiry(h, 0xC9, sizeof(struct c9_inquiry), c9_endio);
+               break;
+       case SEND_MODE_SELECT:
+               submit_mode_select(h);
+               break;
+       default:
+               BUG();
+       }
+}
+/*
+ * only support subpage2c until we confirm that this is just a matter of
+ * of updating firmware or not, and RDAC (basic AVT works already) for now
+ * but we can add these in in when we get time and testers
+ */
+static int rdac_create(struct hw_handler *hwh, unsigned argc, char **argv)
+{
+       struct rdac_handler *h;
+       unsigned timeout;
+
+       if (argc == 0) {
+               /* No arguments: use defaults */
+               timeout = RDAC_FAILOVER_TIMEOUT;
+       } else if (argc != 1) {
+               DMWARN("incorrect number of arguments");
+               return -EINVAL;
+       } else {
+               if (sscanf(argv[1], "%u", &timeout) != 1) {
+                       DMWARN("invalid timeout value");
+                       return -EINVAL;
+               }
+       }
+
+       h = kzalloc(sizeof(*h), GFP_KERNEL);
+       if (!h)
+               return -ENOMEM;
+
+       hwh->context = h;
+       h->timeout = timeout;
+       h->lun = UNINITIALIZED_LUN;
+       INIT_WORK(&h->work, service_wkq);
+       DMWARN("using RDAC command with timeout %u", h->timeout);
+
+       return 0;
+}
+
+static void rdac_destroy(struct hw_handler *hwh)
+{
+       struct rdac_handler *h = hwh->context;
+
+       if (h->ctlr)
+               kref_put(&h->ctlr->kref, release_ctlr);
+       kfree(h);
+       hwh->context = NULL;
+}
+
+static unsigned rdac_error(struct hw_handler *hwh, struct bio *bio)
+{
+       /* Try default handler */
+       return dm_scsi_err_handler(hwh, bio);
+}
+
+static void rdac_pg_init(struct hw_handler *hwh, unsigned bypassed,
+                       struct dm_path *path)
+{
+       struct rdac_handler *h = hwh->context;
+
+       h->path = path;
+       switch (h->lun) {
+       case UNINITIALIZED_LUN:
+               submit_inquiry(h, 0xC8, sizeof(struct c8_inquiry), c8_endio);
+               break;
+       default:
+               submit_inquiry(h, 0xC9, sizeof(struct c9_inquiry), c9_endio);
+       }
+}
+
+static struct hw_handler_type rdac_handler = {
+       .name = RDAC_DM_HWH_NAME,
+       .module = THIS_MODULE,
+       .create = rdac_create,
+       .destroy = rdac_destroy,
+       .pg_init = rdac_pg_init,
+       .error = rdac_error,
+};
+
+static int __init rdac_init(void)
+{
+       int r = dm_register_hw_handler(&rdac_handler);
+
+       if (r < 0) {
+               DMERR("%s: register failed %d", RDAC_DM_HWH_NAME, r);
+               return r;
+       }
+
+       rdac_wkqd = create_singlethread_workqueue("rdac_wkqd");
+       if (!rdac_wkqd) {
+               DMERR("Failed to create workqueue rdac_wkqd.");
+               dm_unregister_hw_handler(&rdac_handler);
+               return -ENOMEM;
+       }
+
+       DMINFO("%s: version %s loaded", RDAC_DM_HWH_NAME, RDAC_DM_HWH_VER);
+       return 0;
+}
+
+static void __exit rdac_exit(void)
+{
+       int r = dm_unregister_hw_handler(&rdac_handler);
+
+       destroy_workqueue(rdac_wkqd);
+       if (r < 0)
+               DMERR("%s: unregister failed %d", RDAC_DM_HWH_NAME, r);
+}
+
+module_init(rdac_init);
+module_exit(rdac_exit);
+
+MODULE_DESCRIPTION("DM Multipath LSI/Engenio RDAC support");
+MODULE_AUTHOR("Mike Christie, Chandra Seetharaman");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(RDAC_DM_HWH_VER);