summaryrefslogtreecommitdiff
path: root/drivers/infiniband/ulp/iser
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/infiniband/ulp/iser')
-rw-r--r--drivers/infiniband/ulp/iser/Kconfig12
-rw-r--r--drivers/infiniband/ulp/iser/Makefile4
-rw-r--r--drivers/infiniband/ulp/iser/iscsi_iser.c787
-rw-r--r--drivers/infiniband/ulp/iser/iscsi_iser.h370
-rw-r--r--drivers/infiniband/ulp/iser/iser_initiator.c570
-rw-r--r--drivers/infiniband/ulp/iser/iser_memory.c422
-rw-r--r--drivers/infiniband/ulp/iser/iser_verbs.c868
7 files changed, 3033 insertions, 0 deletions
diff --git a/drivers/infiniband/ulp/iser/Kconfig b/drivers/infiniband/ulp/iser/Kconfig
new file mode 100644
index 00000000..d00af71a
--- /dev/null
+++ b/drivers/infiniband/ulp/iser/Kconfig
@@ -0,0 +1,12 @@
+config INFINIBAND_ISER
+ tristate "iSCSI Extensions for RDMA (iSER)"
+ depends on SCSI && INET && INFINIBAND_ADDR_TRANS
+ select SCSI_ISCSI_ATTRS
+ ---help---
+ Support for the iSCSI Extensions for RDMA (iSER) Protocol
+ over InfiniBand. This allows you to access storage devices
+ that speak iSCSI over iSER over InfiniBand.
+
+ The iSER protocol is defined by IETF.
+ See <http://www.ietf.org/rfc/rfc5046.txt>
+ and <http://members.infinibandta.org/kwspub/spec/Annex_iSER.PDF>
diff --git a/drivers/infiniband/ulp/iser/Makefile b/drivers/infiniband/ulp/iser/Makefile
new file mode 100644
index 00000000..fe6cd15f
--- /dev/null
+++ b/drivers/infiniband/ulp/iser/Makefile
@@ -0,0 +1,4 @@
+obj-$(CONFIG_INFINIBAND_ISER) += ib_iser.o
+
+ib_iser-y := iser_verbs.o iser_initiator.o iser_memory.o \
+ iscsi_iser.o
diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c
new file mode 100644
index 00000000..db43b311
--- /dev/null
+++ b/drivers/infiniband/ulp/iser/iscsi_iser.c
@@ -0,0 +1,787 @@
+/*
+ * iSCSI Initiator over iSER Data-Path
+ *
+ * Copyright (C) 2004 Dmitry Yusupov
+ * Copyright (C) 2004 Alex Aizman
+ * Copyright (C) 2005 Mike Christie
+ * Copyright (c) 2005, 2006 Voltaire, Inc. All rights reserved.
+ * maintained by openib-general@openib.org
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * Credits:
+ * Christoph Hellwig
+ * FUJITA Tomonori
+ * Arne Redlich
+ * Zhenyu Wang
+ * Modified by:
+ * Erez Zilber
+ */
+
+#include <linux/types.h>
+#include <linux/list.h>
+#include <linux/hardirq.h>
+#include <linux/kfifo.h>
+#include <linux/blkdev.h>
+#include <linux/init.h>
+#include <linux/ioctl.h>
+#include <linux/cdev.h>
+#include <linux/in.h>
+#include <linux/net.h>
+#include <linux/scatterlist.h>
+#include <linux/delay.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+
+#include <net/sock.h>
+
+#include <asm/uaccess.h>
+
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_device.h>
+#include <scsi/scsi_eh.h>
+#include <scsi/scsi_tcq.h>
+#include <scsi/scsi_host.h>
+#include <scsi/scsi.h>
+#include <scsi/scsi_transport_iscsi.h>
+
+#include "iscsi_iser.h"
+
+static struct scsi_host_template iscsi_iser_sht;
+static struct iscsi_transport iscsi_iser_transport;
+static struct scsi_transport_template *iscsi_iser_scsi_transport;
+
+static unsigned int iscsi_max_lun = 512;
+module_param_named(max_lun, iscsi_max_lun, uint, S_IRUGO);
+
+int iser_debug_level = 0;
+
+MODULE_DESCRIPTION("iSER (iSCSI Extensions for RDMA) Datamover "
+ "v" DRV_VER " (" DRV_DATE ")");
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_AUTHOR("Alex Nezhinsky, Dan Bar Dov, Or Gerlitz");
+
+module_param_named(debug_level, iser_debug_level, int, 0644);
+MODULE_PARM_DESC(debug_level, "Enable debug tracing if > 0 (default:disabled)");
+
+struct iser_global ig;
+
+void
+iscsi_iser_recv(struct iscsi_conn *conn,
+ struct iscsi_hdr *hdr, char *rx_data, int rx_data_len)
+{
+ int rc = 0;
+ int datalen;
+ int ahslen;
+
+ /* verify PDU length */
+ datalen = ntoh24(hdr->dlength);
+ if (datalen > rx_data_len || (datalen + 4) < rx_data_len) {
+ iser_err("wrong datalen %d (hdr), %d (IB)\n",
+ datalen, rx_data_len);
+ rc = ISCSI_ERR_DATALEN;
+ goto error;
+ }
+
+ if (datalen != rx_data_len)
+ iser_dbg("aligned datalen (%d) hdr, %d (IB)\n",
+ datalen, rx_data_len);
+
+ /* read AHS */
+ ahslen = hdr->hlength * 4;
+
+ rc = iscsi_complete_pdu(conn, hdr, rx_data, rx_data_len);
+ if (rc && rc != ISCSI_ERR_NO_SCSI_CMD)
+ goto error;
+
+ return;
+error:
+ iscsi_conn_failure(conn, rc);
+}
+
+static int iscsi_iser_pdu_alloc(struct iscsi_task *task, uint8_t opcode)
+{
+ struct iscsi_iser_task *iser_task = task->dd_data;
+
+ task->hdr = (struct iscsi_hdr *)&iser_task->desc.iscsi_header;
+ task->hdr_max = sizeof(iser_task->desc.iscsi_header);
+ return 0;
+}
+
+int iser_initialize_task_headers(struct iscsi_task *task,
+ struct iser_tx_desc *tx_desc)
+{
+ struct iscsi_iser_conn *iser_conn = task->conn->dd_data;
+ struct iser_device *device = iser_conn->ib_conn->device;
+ struct iscsi_iser_task *iser_task = task->dd_data;
+ u64 dma_addr;
+
+ dma_addr = ib_dma_map_single(device->ib_device, (void *)tx_desc,
+ ISER_HEADERS_LEN, DMA_TO_DEVICE);
+ if (ib_dma_mapping_error(device->ib_device, dma_addr))
+ return -ENOMEM;
+
+ tx_desc->dma_addr = dma_addr;
+ tx_desc->tx_sg[0].addr = tx_desc->dma_addr;
+ tx_desc->tx_sg[0].length = ISER_HEADERS_LEN;
+ tx_desc->tx_sg[0].lkey = device->mr->lkey;
+
+ iser_task->iser_conn = iser_conn;
+ return 0;
+}
+/**
+ * iscsi_iser_task_init - Initialize task
+ * @task: iscsi task
+ *
+ * Initialize the task for the scsi command or mgmt command.
+ */
+static int
+iscsi_iser_task_init(struct iscsi_task *task)
+{
+ struct iscsi_iser_task *iser_task = task->dd_data;
+
+ if (iser_initialize_task_headers(task, &iser_task->desc))
+ return -ENOMEM;
+
+ /* mgmt task */
+ if (!task->sc)
+ return 0;
+
+ iser_task->command_sent = 0;
+ iser_task_rdma_init(iser_task);
+ return 0;
+}
+
+/**
+ * iscsi_iser_mtask_xmit - xmit management(immediate) task
+ * @conn: iscsi connection
+ * @task: task management task
+ *
+ * Notes:
+ * The function can return -EAGAIN in which case caller must
+ * call it again later, or recover. '0' return code means successful
+ * xmit.
+ *
+ **/
+static int
+iscsi_iser_mtask_xmit(struct iscsi_conn *conn, struct iscsi_task *task)
+{
+ int error = 0;
+
+ iser_dbg("mtask xmit [cid %d itt 0x%x]\n", conn->id, task->itt);
+
+ error = iser_send_control(conn, task);
+
+ /* since iser xmits control with zero copy, tasks can not be recycled
+ * right after sending them.
+ * The recycling scheme is based on whether a response is expected
+ * - if yes, the task is recycled at iscsi_complete_pdu
+ * - if no, the task is recycled at iser_snd_completion
+ */
+ return error;
+}
+
+static int
+iscsi_iser_task_xmit_unsol_data(struct iscsi_conn *conn,
+ struct iscsi_task *task)
+{
+ struct iscsi_r2t_info *r2t = &task->unsol_r2t;
+ struct iscsi_data hdr;
+ int error = 0;
+
+ /* Send data-out PDUs while there's still unsolicited data to send */
+ while (iscsi_task_has_unsol_data(task)) {
+ iscsi_prep_data_out_pdu(task, r2t, &hdr);
+ iser_dbg("Sending data-out: itt 0x%x, data count %d\n",
+ hdr.itt, r2t->data_count);
+
+ /* the buffer description has been passed with the command */
+ /* Send the command */
+ error = iser_send_data_out(conn, task, &hdr);
+ if (error) {
+ r2t->datasn--;
+ goto iscsi_iser_task_xmit_unsol_data_exit;
+ }
+ r2t->sent += r2t->data_count;
+ iser_dbg("Need to send %d more as data-out PDUs\n",
+ r2t->data_length - r2t->sent);
+ }
+
+iscsi_iser_task_xmit_unsol_data_exit:
+ return error;
+}
+
+static int
+iscsi_iser_task_xmit(struct iscsi_task *task)
+{
+ struct iscsi_conn *conn = task->conn;
+ struct iscsi_iser_task *iser_task = task->dd_data;
+ int error = 0;
+
+ if (!task->sc)
+ return iscsi_iser_mtask_xmit(conn, task);
+
+ if (task->sc->sc_data_direction == DMA_TO_DEVICE) {
+ BUG_ON(scsi_bufflen(task->sc) == 0);
+
+ iser_dbg("cmd [itt %x total %d imm %d unsol_data %d\n",
+ task->itt, scsi_bufflen(task->sc),
+ task->imm_count, task->unsol_r2t.data_length);
+ }
+
+ iser_dbg("ctask xmit [cid %d itt 0x%x]\n",
+ conn->id, task->itt);
+
+ /* Send the cmd PDU */
+ if (!iser_task->command_sent) {
+ error = iser_send_command(conn, task);
+ if (error)
+ goto iscsi_iser_task_xmit_exit;
+ iser_task->command_sent = 1;
+ }
+
+ /* Send unsolicited data-out PDU(s) if necessary */
+ if (iscsi_task_has_unsol_data(task))
+ error = iscsi_iser_task_xmit_unsol_data(conn, task);
+
+ iscsi_iser_task_xmit_exit:
+ return error;
+}
+
+static void iscsi_iser_cleanup_task(struct iscsi_task *task)
+{
+ struct iscsi_iser_task *iser_task = task->dd_data;
+ struct iser_tx_desc *tx_desc = &iser_task->desc;
+
+ struct iscsi_iser_conn *iser_conn = task->conn->dd_data;
+ struct iser_device *device = iser_conn->ib_conn->device;
+
+ ib_dma_unmap_single(device->ib_device,
+ tx_desc->dma_addr, ISER_HEADERS_LEN, DMA_TO_DEVICE);
+
+ /* mgmt tasks do not need special cleanup */
+ if (!task->sc)
+ return;
+
+ if (iser_task->status == ISER_TASK_STATUS_STARTED) {
+ iser_task->status = ISER_TASK_STATUS_COMPLETED;
+ iser_task_rdma_finalize(iser_task);
+ }
+}
+
+static struct iscsi_cls_conn *
+iscsi_iser_conn_create(struct iscsi_cls_session *cls_session, uint32_t conn_idx)
+{
+ struct iscsi_conn *conn;
+ struct iscsi_cls_conn *cls_conn;
+ struct iscsi_iser_conn *iser_conn;
+
+ cls_conn = iscsi_conn_setup(cls_session, sizeof(*iser_conn), conn_idx);
+ if (!cls_conn)
+ return NULL;
+ conn = cls_conn->dd_data;
+
+ /*
+ * due to issues with the login code re iser sematics
+ * this not set in iscsi_conn_setup - FIXME
+ */
+ conn->max_recv_dlength = ISER_RECV_DATA_SEG_LEN;
+
+ iser_conn = conn->dd_data;
+ conn->dd_data = iser_conn;
+ iser_conn->iscsi_conn = conn;
+
+ return cls_conn;
+}
+
+static void
+iscsi_iser_conn_destroy(struct iscsi_cls_conn *cls_conn)
+{
+ struct iscsi_conn *conn = cls_conn->dd_data;
+ struct iscsi_iser_conn *iser_conn = conn->dd_data;
+ struct iser_conn *ib_conn = iser_conn->ib_conn;
+
+ iscsi_conn_teardown(cls_conn);
+ /*
+ * Userspace will normally call the stop callback and
+ * already have freed the ib_conn, but if it goofed up then
+ * we free it here.
+ */
+ if (ib_conn) {
+ ib_conn->iser_conn = NULL;
+ iser_conn_put(ib_conn, 1); /* deref iscsi/ib conn unbinding */
+ }
+}
+
+static int
+iscsi_iser_conn_bind(struct iscsi_cls_session *cls_session,
+ struct iscsi_cls_conn *cls_conn, uint64_t transport_eph,
+ int is_leading)
+{
+ struct iscsi_conn *conn = cls_conn->dd_data;
+ struct iscsi_iser_conn *iser_conn;
+ struct iser_conn *ib_conn;
+ struct iscsi_endpoint *ep;
+ int error;
+
+ error = iscsi_conn_bind(cls_session, cls_conn, is_leading);
+ if (error)
+ return error;
+
+ /* the transport ep handle comes from user space so it must be
+ * verified against the global ib connections list */
+ ep = iscsi_lookup_endpoint(transport_eph);
+ if (!ep) {
+ iser_err("can't bind eph %llx\n",
+ (unsigned long long)transport_eph);
+ return -EINVAL;
+ }
+ ib_conn = ep->dd_data;
+
+ if (iser_alloc_rx_descriptors(ib_conn))
+ return -ENOMEM;
+
+ /* binds the iSER connection retrieved from the previously
+ * connected ep_handle to the iSCSI layer connection. exchanges
+ * connection pointers */
+ iser_err("binding iscsi/iser conn %p %p to ib_conn %p\n",
+ conn, conn->dd_data, ib_conn);
+ iser_conn = conn->dd_data;
+ ib_conn->iser_conn = iser_conn;
+ iser_conn->ib_conn = ib_conn;
+ iser_conn_get(ib_conn); /* ref iscsi/ib conn binding */
+ return 0;
+}
+
+static void
+iscsi_iser_conn_stop(struct iscsi_cls_conn *cls_conn, int flag)
+{
+ struct iscsi_conn *conn = cls_conn->dd_data;
+ struct iscsi_iser_conn *iser_conn = conn->dd_data;
+ struct iser_conn *ib_conn = iser_conn->ib_conn;
+
+ /*
+ * Userspace may have goofed up and not bound the connection or
+ * might have only partially setup the connection.
+ */
+ if (ib_conn) {
+ iscsi_conn_stop(cls_conn, flag);
+ /*
+ * There is no unbind event so the stop callback
+ * must release the ref from the bind.
+ */
+ iser_conn_put(ib_conn, 1); /* deref iscsi/ib conn unbinding */
+ }
+ iser_conn->ib_conn = NULL;
+}
+
+static void iscsi_iser_session_destroy(struct iscsi_cls_session *cls_session)
+{
+ struct Scsi_Host *shost = iscsi_session_to_shost(cls_session);
+
+ iscsi_session_teardown(cls_session);
+ iscsi_host_remove(shost);
+ iscsi_host_free(shost);
+}
+
+static struct iscsi_cls_session *
+iscsi_iser_session_create(struct iscsi_endpoint *ep,
+ uint16_t cmds_max, uint16_t qdepth,
+ uint32_t initial_cmdsn)
+{
+ struct iscsi_cls_session *cls_session;
+ struct iscsi_session *session;
+ struct Scsi_Host *shost;
+ struct iser_conn *ib_conn;
+
+ shost = iscsi_host_alloc(&iscsi_iser_sht, 0, 0);
+ if (!shost)
+ return NULL;
+ shost->transportt = iscsi_iser_scsi_transport;
+ shost->max_lun = iscsi_max_lun;
+ shost->max_id = 0;
+ shost->max_channel = 0;
+ shost->max_cmd_len = 16;
+
+ /*
+ * older userspace tools (before 2.0-870) did not pass us
+ * the leading conn's ep so this will be NULL;
+ */
+ if (ep)
+ ib_conn = ep->dd_data;
+
+ if (iscsi_host_add(shost,
+ ep ? ib_conn->device->ib_device->dma_device : NULL))
+ goto free_host;
+
+ /*
+ * we do not support setting can_queue cmd_per_lun from userspace yet
+ * because we preallocate so many resources
+ */
+ cls_session = iscsi_session_setup(&iscsi_iser_transport, shost,
+ ISCSI_DEF_XMIT_CMDS_MAX, 0,
+ sizeof(struct iscsi_iser_task),
+ initial_cmdsn, 0);
+ if (!cls_session)
+ goto remove_host;
+ session = cls_session->dd_data;
+
+ shost->can_queue = session->scsi_cmds_max;
+ return cls_session;
+
+remove_host:
+ iscsi_host_remove(shost);
+free_host:
+ iscsi_host_free(shost);
+ return NULL;
+}
+
+static int
+iscsi_iser_set_param(struct iscsi_cls_conn *cls_conn,
+ enum iscsi_param param, char *buf, int buflen)
+{
+ int value;
+
+ switch (param) {
+ case ISCSI_PARAM_MAX_RECV_DLENGTH:
+ /* TBD */
+ break;
+ case ISCSI_PARAM_HDRDGST_EN:
+ sscanf(buf, "%d", &value);
+ if (value) {
+ printk(KERN_ERR "DataDigest wasn't negotiated to None");
+ return -EPROTO;
+ }
+ break;
+ case ISCSI_PARAM_DATADGST_EN:
+ sscanf(buf, "%d", &value);
+ if (value) {
+ printk(KERN_ERR "DataDigest wasn't negotiated to None");
+ return -EPROTO;
+ }
+ break;
+ case ISCSI_PARAM_IFMARKER_EN:
+ sscanf(buf, "%d", &value);
+ if (value) {
+ printk(KERN_ERR "IFMarker wasn't negotiated to No");
+ return -EPROTO;
+ }
+ break;
+ case ISCSI_PARAM_OFMARKER_EN:
+ sscanf(buf, "%d", &value);
+ if (value) {
+ printk(KERN_ERR "OFMarker wasn't negotiated to No");
+ return -EPROTO;
+ }
+ break;
+ default:
+ return iscsi_set_param(cls_conn, param, buf, buflen);
+ }
+
+ return 0;
+}
+
+static void
+iscsi_iser_conn_get_stats(struct iscsi_cls_conn *cls_conn, struct iscsi_stats *stats)
+{
+ struct iscsi_conn *conn = cls_conn->dd_data;
+
+ stats->txdata_octets = conn->txdata_octets;
+ stats->rxdata_octets = conn->rxdata_octets;
+ stats->scsicmd_pdus = conn->scsicmd_pdus_cnt;
+ stats->dataout_pdus = conn->dataout_pdus_cnt;
+ stats->scsirsp_pdus = conn->scsirsp_pdus_cnt;
+ stats->datain_pdus = conn->datain_pdus_cnt; /* always 0 */
+ stats->r2t_pdus = conn->r2t_pdus_cnt; /* always 0 */
+ stats->tmfcmd_pdus = conn->tmfcmd_pdus_cnt;
+ stats->tmfrsp_pdus = conn->tmfrsp_pdus_cnt;
+ stats->custom_length = 4;
+ strcpy(stats->custom[0].desc, "qp_tx_queue_full");
+ stats->custom[0].value = 0; /* TB iser_conn->qp_tx_queue_full; */
+ strcpy(stats->custom[1].desc, "fmr_map_not_avail");
+ stats->custom[1].value = 0; /* TB iser_conn->fmr_map_not_avail */;
+ strcpy(stats->custom[2].desc, "eh_abort_cnt");
+ stats->custom[2].value = conn->eh_abort_cnt;
+ strcpy(stats->custom[3].desc, "fmr_unalign_cnt");
+ stats->custom[3].value = conn->fmr_unalign_cnt;
+}
+
+static int iscsi_iser_get_ep_param(struct iscsi_endpoint *ep,
+ enum iscsi_param param, char *buf)
+{
+ struct iser_conn *ib_conn = ep->dd_data;
+ int len;
+
+ switch (param) {
+ case ISCSI_PARAM_CONN_PORT:
+ case ISCSI_PARAM_CONN_ADDRESS:
+ if (!ib_conn || !ib_conn->cma_id)
+ return -ENOTCONN;
+
+ return iscsi_conn_get_addr_param((struct sockaddr_storage *)
+ &ib_conn->cma_id->route.addr.dst_addr,
+ param, buf);
+ break;
+ default:
+ return -ENOSYS;
+ }
+
+ return len;
+}
+
+static struct iscsi_endpoint *
+iscsi_iser_ep_connect(struct Scsi_Host *shost, struct sockaddr *dst_addr,
+ int non_blocking)
+{
+ int err;
+ struct iser_conn *ib_conn;
+ struct iscsi_endpoint *ep;
+
+ ep = iscsi_create_endpoint(sizeof(*ib_conn));
+ if (!ep)
+ return ERR_PTR(-ENOMEM);
+
+ ib_conn = ep->dd_data;
+ ib_conn->ep = ep;
+ iser_conn_init(ib_conn);
+
+ err = iser_connect(ib_conn, NULL, (struct sockaddr_in *)dst_addr,
+ non_blocking);
+ if (err) {
+ iscsi_destroy_endpoint(ep);
+ return ERR_PTR(err);
+ }
+ return ep;
+}
+
+static int
+iscsi_iser_ep_poll(struct iscsi_endpoint *ep, int timeout_ms)
+{
+ struct iser_conn *ib_conn;
+ int rc;
+
+ ib_conn = ep->dd_data;
+ rc = wait_event_interruptible_timeout(ib_conn->wait,
+ ib_conn->state == ISER_CONN_UP,
+ msecs_to_jiffies(timeout_ms));
+
+ /* if conn establishment failed, return error code to iscsi */
+ if (!rc &&
+ (ib_conn->state == ISER_CONN_TERMINATING ||
+ ib_conn->state == ISER_CONN_DOWN))
+ rc = -1;
+
+ iser_err("ib conn %p rc = %d\n", ib_conn, rc);
+
+ if (rc > 0)
+ return 1; /* success, this is the equivalent of POLLOUT */
+ else if (!rc)
+ return 0; /* timeout */
+ else
+ return rc; /* signal */
+}
+
+static void
+iscsi_iser_ep_disconnect(struct iscsi_endpoint *ep)
+{
+ struct iser_conn *ib_conn;
+
+ ib_conn = ep->dd_data;
+ if (ib_conn->iser_conn)
+ /*
+ * Must suspend xmit path if the ep is bound to the
+ * iscsi_conn, so we know we are not accessing the ib_conn
+ * when we free it.
+ *
+ * This may not be bound if the ep poll failed.
+ */
+ iscsi_suspend_tx(ib_conn->iser_conn->iscsi_conn);
+
+
+ iser_err("ib conn %p state %d\n",ib_conn, ib_conn->state);
+ iser_conn_terminate(ib_conn);
+}
+
+static umode_t iser_attr_is_visible(int param_type, int param)
+{
+ switch (param_type) {
+ case ISCSI_HOST_PARAM:
+ switch (param) {
+ case ISCSI_HOST_PARAM_NETDEV_NAME:
+ case ISCSI_HOST_PARAM_HWADDRESS:
+ case ISCSI_HOST_PARAM_INITIATOR_NAME:
+ return S_IRUGO;
+ default:
+ return 0;
+ }
+ case ISCSI_PARAM:
+ switch (param) {
+ case ISCSI_PARAM_MAX_RECV_DLENGTH:
+ case ISCSI_PARAM_MAX_XMIT_DLENGTH:
+ case ISCSI_PARAM_HDRDGST_EN:
+ case ISCSI_PARAM_DATADGST_EN:
+ case ISCSI_PARAM_CONN_ADDRESS:
+ case ISCSI_PARAM_CONN_PORT:
+ case ISCSI_PARAM_EXP_STATSN:
+ case ISCSI_PARAM_PERSISTENT_ADDRESS:
+ case ISCSI_PARAM_PERSISTENT_PORT:
+ case ISCSI_PARAM_PING_TMO:
+ case ISCSI_PARAM_RECV_TMO:
+ case ISCSI_PARAM_INITIAL_R2T_EN:
+ case ISCSI_PARAM_MAX_R2T:
+ case ISCSI_PARAM_IMM_DATA_EN:
+ case ISCSI_PARAM_FIRST_BURST:
+ case ISCSI_PARAM_MAX_BURST:
+ case ISCSI_PARAM_PDU_INORDER_EN:
+ case ISCSI_PARAM_DATASEQ_INORDER_EN:
+ case ISCSI_PARAM_TARGET_NAME:
+ case ISCSI_PARAM_TPGT:
+ case ISCSI_PARAM_USERNAME:
+ case ISCSI_PARAM_PASSWORD:
+ case ISCSI_PARAM_USERNAME_IN:
+ case ISCSI_PARAM_PASSWORD_IN:
+ case ISCSI_PARAM_FAST_ABORT:
+ case ISCSI_PARAM_ABORT_TMO:
+ case ISCSI_PARAM_LU_RESET_TMO:
+ case ISCSI_PARAM_TGT_RESET_TMO:
+ case ISCSI_PARAM_IFACE_NAME:
+ case ISCSI_PARAM_INITIATOR_NAME:
+ return S_IRUGO;
+ default:
+ return 0;
+ }
+ }
+
+ return 0;
+}
+
+static struct scsi_host_template iscsi_iser_sht = {
+ .module = THIS_MODULE,
+ .name = "iSCSI Initiator over iSER, v." DRV_VER,
+ .queuecommand = iscsi_queuecommand,
+ .change_queue_depth = iscsi_change_queue_depth,
+ .sg_tablesize = ISCSI_ISER_SG_TABLESIZE,
+ .max_sectors = 1024,
+ .cmd_per_lun = ISER_DEF_CMD_PER_LUN,
+ .eh_abort_handler = iscsi_eh_abort,
+ .eh_device_reset_handler= iscsi_eh_device_reset,
+ .eh_target_reset_handler = iscsi_eh_recover_target,
+ .target_alloc = iscsi_target_alloc,
+ .use_clustering = DISABLE_CLUSTERING,
+ .proc_name = "iscsi_iser",
+ .this_id = -1,
+};
+
+static struct iscsi_transport iscsi_iser_transport = {
+ .owner = THIS_MODULE,
+ .name = "iser",
+ .caps = CAP_RECOVERY_L0 | CAP_MULTI_R2T,
+ /* session management */
+ .create_session = iscsi_iser_session_create,
+ .destroy_session = iscsi_iser_session_destroy,
+ /* connection management */
+ .create_conn = iscsi_iser_conn_create,
+ .bind_conn = iscsi_iser_conn_bind,
+ .destroy_conn = iscsi_iser_conn_destroy,
+ .attr_is_visible = iser_attr_is_visible,
+ .set_param = iscsi_iser_set_param,
+ .get_conn_param = iscsi_conn_get_param,
+ .get_ep_param = iscsi_iser_get_ep_param,
+ .get_session_param = iscsi_session_get_param,
+ .start_conn = iscsi_conn_start,
+ .stop_conn = iscsi_iser_conn_stop,
+ /* iscsi host params */
+ .get_host_param = iscsi_host_get_param,
+ .set_host_param = iscsi_host_set_param,
+ /* IO */
+ .send_pdu = iscsi_conn_send_pdu,
+ .get_stats = iscsi_iser_conn_get_stats,
+ .init_task = iscsi_iser_task_init,
+ .xmit_task = iscsi_iser_task_xmit,
+ .cleanup_task = iscsi_iser_cleanup_task,
+ .alloc_pdu = iscsi_iser_pdu_alloc,
+ /* recovery */
+ .session_recovery_timedout = iscsi_session_recovery_timedout,
+
+ .ep_connect = iscsi_iser_ep_connect,
+ .ep_poll = iscsi_iser_ep_poll,
+ .ep_disconnect = iscsi_iser_ep_disconnect
+};
+
+static int __init iser_init(void)
+{
+ int err;
+
+ iser_dbg("Starting iSER datamover...\n");
+
+ if (iscsi_max_lun < 1) {
+ printk(KERN_ERR "Invalid max_lun value of %u\n", iscsi_max_lun);
+ return -EINVAL;
+ }
+
+ memset(&ig, 0, sizeof(struct iser_global));
+
+ ig.desc_cache = kmem_cache_create("iser_descriptors",
+ sizeof(struct iser_tx_desc),
+ 0, SLAB_HWCACHE_ALIGN,
+ NULL);
+ if (ig.desc_cache == NULL)
+ return -ENOMEM;
+
+ /* device init is called only after the first addr resolution */
+ mutex_init(&ig.device_list_mutex);
+ INIT_LIST_HEAD(&ig.device_list);
+ mutex_init(&ig.connlist_mutex);
+ INIT_LIST_HEAD(&ig.connlist);
+
+ iscsi_iser_scsi_transport = iscsi_register_transport(
+ &iscsi_iser_transport);
+ if (!iscsi_iser_scsi_transport) {
+ iser_err("iscsi_register_transport failed\n");
+ err = -EINVAL;
+ goto register_transport_failure;
+ }
+
+ return 0;
+
+register_transport_failure:
+ kmem_cache_destroy(ig.desc_cache);
+
+ return err;
+}
+
+static void __exit iser_exit(void)
+{
+ iser_dbg("Removing iSER datamover...\n");
+ iscsi_unregister_transport(&iscsi_iser_transport);
+ kmem_cache_destroy(ig.desc_cache);
+}
+
+module_init(iser_init);
+module_exit(iser_exit);
diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h
new file mode 100644
index 00000000..296be431
--- /dev/null
+++ b/drivers/infiniband/ulp/iser/iscsi_iser.h
@@ -0,0 +1,370 @@
+/*
+ * iSER transport for the Open iSCSI Initiator & iSER transport internals
+ *
+ * Copyright (C) 2004 Dmitry Yusupov
+ * Copyright (C) 2004 Alex Aizman
+ * Copyright (C) 2005 Mike Christie
+ * based on code maintained by open-iscsi@googlegroups.com
+ *
+ * Copyright (c) 2004, 2005, 2006 Voltaire, Inc. All rights reserved.
+ * Copyright (c) 2005, 2006 Cisco Systems. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef __ISCSI_ISER_H__
+#define __ISCSI_ISER_H__
+
+#include <linux/types.h>
+#include <linux/net.h>
+#include <scsi/libiscsi.h>
+#include <scsi/scsi_transport_iscsi.h>
+
+#include <linux/interrupt.h>
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/dma-mapping.h>
+#include <linux/mutex.h>
+#include <linux/mempool.h>
+#include <linux/uio.h>
+
+#include <linux/socket.h>
+#include <linux/in.h>
+#include <linux/in6.h>
+
+#include <rdma/ib_verbs.h>
+#include <rdma/ib_fmr_pool.h>
+#include <rdma/rdma_cm.h>
+
+#define DRV_NAME "iser"
+#define PFX DRV_NAME ": "
+#define DRV_VER "0.1"
+#define DRV_DATE "May 7th, 2006"
+
+#define iser_dbg(fmt, arg...) \
+ do { \
+ if (iser_debug_level > 1) \
+ printk(KERN_DEBUG PFX "%s:" fmt,\
+ __func__ , ## arg); \
+ } while (0)
+
+#define iser_warn(fmt, arg...) \
+ do { \
+ if (iser_debug_level > 0) \
+ printk(KERN_DEBUG PFX "%s:" fmt,\
+ __func__ , ## arg); \
+ } while (0)
+
+#define iser_err(fmt, arg...) \
+ do { \
+ printk(KERN_ERR PFX "%s:" fmt, \
+ __func__ , ## arg); \
+ } while (0)
+
+#define SHIFT_4K 12
+#define SIZE_4K (1ULL << SHIFT_4K)
+#define MASK_4K (~(SIZE_4K-1))
+
+ /* support up to 512KB in one RDMA */
+#define ISCSI_ISER_SG_TABLESIZE (0x80000 >> SHIFT_4K)
+#define ISER_DEF_CMD_PER_LUN 128
+
+/* QP settings */
+/* Maximal bounds on received asynchronous PDUs */
+#define ISER_MAX_RX_MISC_PDUS 4 /* NOOP_IN(2) , ASYNC_EVENT(2) */
+
+#define ISER_MAX_TX_MISC_PDUS 6 /* NOOP_OUT(2), TEXT(1), *
+ * SCSI_TMFUNC(2), LOGOUT(1) */
+
+#define ISER_QP_MAX_RECV_DTOS (ISCSI_DEF_XMIT_CMDS_MAX)
+
+#define ISER_MIN_POSTED_RX (ISCSI_DEF_XMIT_CMDS_MAX >> 2)
+
+/* the max TX (send) WR supported by the iSER QP is defined by *
+ * max_send_wr = T * (1 + D) + C ; D is how many inflight dataouts we expect *
+ * to have at max for SCSI command. The tx posting & completion handling code *
+ * supports -EAGAIN scheme where tx is suspended till the QP has room for more *
+ * send WR. D=8 comes from 64K/8K */
+
+#define ISER_INFLIGHT_DATAOUTS 8
+
+#define ISER_QP_MAX_REQ_DTOS (ISCSI_DEF_XMIT_CMDS_MAX * \
+ (1 + ISER_INFLIGHT_DATAOUTS) + \
+ ISER_MAX_TX_MISC_PDUS + \
+ ISER_MAX_RX_MISC_PDUS)
+
+#define ISER_VER 0x10
+#define ISER_WSV 0x08
+#define ISER_RSV 0x04
+
+struct iser_hdr {
+ u8 flags;
+ u8 rsvd[3];
+ __be32 write_stag; /* write rkey */
+ __be64 write_va;
+ __be32 read_stag; /* read rkey */
+ __be64 read_va;
+} __attribute__((packed));
+
+/* Constant PDU lengths calculations */
+#define ISER_HEADERS_LEN (sizeof(struct iser_hdr) + sizeof(struct iscsi_hdr))
+
+#define ISER_RECV_DATA_SEG_LEN 128
+#define ISER_RX_PAYLOAD_SIZE (ISER_HEADERS_LEN + ISER_RECV_DATA_SEG_LEN)
+#define ISER_RX_LOGIN_SIZE (ISER_HEADERS_LEN + ISCSI_DEF_MAX_RECV_SEG_LEN)
+
+/* Length of an object name string */
+#define ISER_OBJECT_NAME_SIZE 64
+
+enum iser_ib_conn_state {
+ ISER_CONN_INIT, /* descriptor allocd, no conn */
+ ISER_CONN_PENDING, /* in the process of being established */
+ ISER_CONN_UP, /* up and running */
+ ISER_CONN_TERMINATING, /* in the process of being terminated */
+ ISER_CONN_DOWN, /* shut down */
+ ISER_CONN_STATES_NUM
+};
+
+enum iser_task_status {
+ ISER_TASK_STATUS_INIT = 0,
+ ISER_TASK_STATUS_STARTED,
+ ISER_TASK_STATUS_COMPLETED
+};
+
+enum iser_data_dir {
+ ISER_DIR_IN = 0, /* to initiator */
+ ISER_DIR_OUT, /* from initiator */
+ ISER_DIRS_NUM
+};
+
+struct iser_data_buf {
+ void *buf; /* pointer to the sg list */
+ unsigned int size; /* num entries of this sg */
+ unsigned long data_len; /* total data len */
+ unsigned int dma_nents; /* returned by dma_map_sg */
+ char *copy_buf; /* allocated copy buf for SGs unaligned *
+ * for rdma which are copied */
+ struct scatterlist sg_single; /* SG-ified clone of a non SG SC or *
+ * unaligned SG */
+ };
+
+/* fwd declarations */
+struct iser_device;
+struct iscsi_iser_conn;
+struct iscsi_iser_task;
+struct iscsi_endpoint;
+
+struct iser_mem_reg {
+ u32 lkey;
+ u32 rkey;
+ u64 va;
+ u64 len;
+ void *mem_h;
+ int is_fmr;
+};
+
+struct iser_regd_buf {
+ struct iser_mem_reg reg; /* memory registration info */
+ void *virt_addr;
+ struct iser_device *device; /* device->device for dma_unmap */
+ enum dma_data_direction direction; /* direction for dma_unmap */
+ unsigned int data_size;
+};
+
+enum iser_desc_type {
+ ISCSI_TX_CONTROL ,
+ ISCSI_TX_SCSI_COMMAND,
+ ISCSI_TX_DATAOUT
+};
+
+struct iser_tx_desc {
+ struct iser_hdr iser_header;
+ struct iscsi_hdr iscsi_header;
+ enum iser_desc_type type;
+ u64 dma_addr;
+ /* sg[0] points to iser/iscsi headers, sg[1] optionally points to either
+ of immediate data, unsolicited data-out or control (login,text) */
+ struct ib_sge tx_sg[2];
+ int num_sge;
+};
+
+#define ISER_RX_PAD_SIZE (256 - (ISER_RX_PAYLOAD_SIZE + \
+ sizeof(u64) + sizeof(struct ib_sge)))
+struct iser_rx_desc {
+ struct iser_hdr iser_header;
+ struct iscsi_hdr iscsi_header;
+ char data[ISER_RECV_DATA_SEG_LEN];
+ u64 dma_addr;
+ struct ib_sge rx_sg;
+ char pad[ISER_RX_PAD_SIZE];
+} __attribute__((packed));
+
+struct iser_device {
+ struct ib_device *ib_device;
+ struct ib_pd *pd;
+ struct ib_cq *rx_cq;
+ struct ib_cq *tx_cq;
+ struct ib_mr *mr;
+ struct tasklet_struct cq_tasklet;
+ struct ib_event_handler event_handler;
+ struct list_head ig_list; /* entry in ig devices list */
+ int refcount;
+};
+
+struct iser_conn {
+ struct iscsi_iser_conn *iser_conn; /* iser conn for upcalls */
+ struct iscsi_endpoint *ep;
+ enum iser_ib_conn_state state; /* rdma connection state */
+ atomic_t refcount;
+ spinlock_t lock; /* used for state changes */
+ struct iser_device *device; /* device context */
+ struct rdma_cm_id *cma_id; /* CMA ID */
+ struct ib_qp *qp; /* QP */
+ struct ib_fmr_pool *fmr_pool; /* pool of IB FMRs */
+ wait_queue_head_t wait; /* waitq for conn/disconn */
+ int post_recv_buf_count; /* posted rx count */
+ atomic_t post_send_buf_count; /* posted tx count */
+ char name[ISER_OBJECT_NAME_SIZE];
+ struct iser_page_vec *page_vec; /* represents SG to fmr maps*
+ * maps serialized as tx is*/
+ struct list_head conn_list; /* entry in ig conn list */
+
+ char *login_buf;
+ char *login_req_buf, *login_resp_buf;
+ u64 login_req_dma, login_resp_dma;
+ unsigned int rx_desc_head;
+ struct iser_rx_desc *rx_descs;
+ struct ib_recv_wr rx_wr[ISER_MIN_POSTED_RX];
+};
+
+struct iscsi_iser_conn {
+ struct iscsi_conn *iscsi_conn;/* ptr to iscsi conn */
+ struct iser_conn *ib_conn; /* iSER IB conn */
+};
+
+struct iscsi_iser_task {
+ struct iser_tx_desc desc;
+ struct iscsi_iser_conn *iser_conn;
+ enum iser_task_status status;
+ int command_sent; /* set if command sent */
+ int dir[ISER_DIRS_NUM]; /* set if dir use*/
+ struct iser_regd_buf rdma_regd[ISER_DIRS_NUM];/* regd rdma buf */
+ struct iser_data_buf data[ISER_DIRS_NUM]; /* orig. data des*/
+ struct iser_data_buf data_copy[ISER_DIRS_NUM];/* contig. copy */
+};
+
+struct iser_page_vec {
+ u64 *pages;
+ int length;
+ int offset;
+ int data_size;
+};
+
+struct iser_global {
+ struct mutex device_list_mutex;/* */
+ struct list_head device_list; /* all iSER devices */
+ struct mutex connlist_mutex;
+ struct list_head connlist; /* all iSER IB connections */
+
+ struct kmem_cache *desc_cache;
+};
+
+extern struct iser_global ig;
+extern int iser_debug_level;
+
+/* allocate connection resources needed for rdma functionality */
+int iser_conn_set_full_featured_mode(struct iscsi_conn *conn);
+
+int iser_send_control(struct iscsi_conn *conn,
+ struct iscsi_task *task);
+
+int iser_send_command(struct iscsi_conn *conn,
+ struct iscsi_task *task);
+
+int iser_send_data_out(struct iscsi_conn *conn,
+ struct iscsi_task *task,
+ struct iscsi_data *hdr);
+
+void iscsi_iser_recv(struct iscsi_conn *conn,
+ struct iscsi_hdr *hdr,
+ char *rx_data,
+ int rx_data_len);
+
+void iser_conn_init(struct iser_conn *ib_conn);
+
+void iser_conn_get(struct iser_conn *ib_conn);
+
+int iser_conn_put(struct iser_conn *ib_conn, int destroy_cma_id_allowed);
+
+void iser_conn_terminate(struct iser_conn *ib_conn);
+
+void iser_rcv_completion(struct iser_rx_desc *desc,
+ unsigned long dto_xfer_len,
+ struct iser_conn *ib_conn);
+
+void iser_snd_completion(struct iser_tx_desc *desc, struct iser_conn *ib_conn);
+
+void iser_task_rdma_init(struct iscsi_iser_task *task);
+
+void iser_task_rdma_finalize(struct iscsi_iser_task *task);
+
+void iser_free_rx_descriptors(struct iser_conn *ib_conn);
+
+void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_task *task,
+ enum iser_data_dir cmd_dir);
+
+int iser_reg_rdma_mem(struct iscsi_iser_task *task,
+ enum iser_data_dir cmd_dir);
+
+int iser_connect(struct iser_conn *ib_conn,
+ struct sockaddr_in *src_addr,
+ struct sockaddr_in *dst_addr,
+ int non_blocking);
+
+int iser_reg_page_vec(struct iser_conn *ib_conn,
+ struct iser_page_vec *page_vec,
+ struct iser_mem_reg *mem_reg);
+
+void iser_unreg_mem(struct iser_mem_reg *mem_reg);
+
+int iser_post_recvl(struct iser_conn *ib_conn);
+int iser_post_recvm(struct iser_conn *ib_conn, int count);
+int iser_post_send(struct iser_conn *ib_conn, struct iser_tx_desc *tx_desc);
+
+int iser_dma_map_task_data(struct iscsi_iser_task *iser_task,
+ struct iser_data_buf *data,
+ enum iser_data_dir iser_dir,
+ enum dma_data_direction dma_dir);
+
+void iser_dma_unmap_task_data(struct iscsi_iser_task *iser_task);
+int iser_initialize_task_headers(struct iscsi_task *task,
+ struct iser_tx_desc *tx_desc);
+int iser_alloc_rx_descriptors(struct iser_conn *ib_conn);
+#endif
diff --git a/drivers/infiniband/ulp/iser/iser_initiator.c b/drivers/infiniband/ulp/iser/iser_initiator.c
new file mode 100644
index 00000000..a00ccd1c
--- /dev/null
+++ b/drivers/infiniband/ulp/iser/iser_initiator.c
@@ -0,0 +1,570 @@
+/*
+ * Copyright (c) 2004, 2005, 2006 Voltaire, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/mm.h>
+#include <linux/scatterlist.h>
+#include <linux/kfifo.h>
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_host.h>
+
+#include "iscsi_iser.h"
+
+/* Register user buffer memory and initialize passive rdma
+ * dto descriptor. Total data size is stored in
+ * iser_task->data[ISER_DIR_IN].data_len
+ */
+static int iser_prepare_read_cmd(struct iscsi_task *task,
+ unsigned int edtl)
+
+{
+ struct iscsi_iser_task *iser_task = task->dd_data;
+ struct iser_regd_buf *regd_buf;
+ int err;
+ struct iser_hdr *hdr = &iser_task->desc.iser_header;
+ struct iser_data_buf *buf_in = &iser_task->data[ISER_DIR_IN];
+
+ err = iser_dma_map_task_data(iser_task,
+ buf_in,
+ ISER_DIR_IN,
+ DMA_FROM_DEVICE);
+ if (err)
+ return err;
+
+ if (edtl > iser_task->data[ISER_DIR_IN].data_len) {
+ iser_err("Total data length: %ld, less than EDTL: "
+ "%d, in READ cmd BHS itt: %d, conn: 0x%p\n",
+ iser_task->data[ISER_DIR_IN].data_len, edtl,
+ task->itt, iser_task->iser_conn);
+ return -EINVAL;
+ }
+
+ err = iser_reg_rdma_mem(iser_task,ISER_DIR_IN);
+ if (err) {
+ iser_err("Failed to set up Data-IN RDMA\n");
+ return err;
+ }
+ regd_buf = &iser_task->rdma_regd[ISER_DIR_IN];
+
+ hdr->flags |= ISER_RSV;
+ hdr->read_stag = cpu_to_be32(regd_buf->reg.rkey);
+ hdr->read_va = cpu_to_be64(regd_buf->reg.va);
+
+ iser_dbg("Cmd itt:%d READ tags RKEY:%#.4X VA:%#llX\n",
+ task->itt, regd_buf->reg.rkey,
+ (unsigned long long)regd_buf->reg.va);
+
+ return 0;
+}
+
+/* Register user buffer memory and initialize passive rdma
+ * dto descriptor. Total data size is stored in
+ * task->data[ISER_DIR_OUT].data_len
+ */
+static int
+iser_prepare_write_cmd(struct iscsi_task *task,
+ unsigned int imm_sz,
+ unsigned int unsol_sz,
+ unsigned int edtl)
+{
+ struct iscsi_iser_task *iser_task = task->dd_data;
+ struct iser_regd_buf *regd_buf;
+ int err;
+ struct iser_hdr *hdr = &iser_task->desc.iser_header;
+ struct iser_data_buf *buf_out = &iser_task->data[ISER_DIR_OUT];
+ struct ib_sge *tx_dsg = &iser_task->desc.tx_sg[1];
+
+ err = iser_dma_map_task_data(iser_task,
+ buf_out,
+ ISER_DIR_OUT,
+ DMA_TO_DEVICE);
+ if (err)
+ return err;
+
+ if (edtl > iser_task->data[ISER_DIR_OUT].data_len) {
+ iser_err("Total data length: %ld, less than EDTL: %d, "
+ "in WRITE cmd BHS itt: %d, conn: 0x%p\n",
+ iser_task->data[ISER_DIR_OUT].data_len,
+ edtl, task->itt, task->conn);
+ return -EINVAL;
+ }
+
+ err = iser_reg_rdma_mem(iser_task,ISER_DIR_OUT);
+ if (err != 0) {
+ iser_err("Failed to register write cmd RDMA mem\n");
+ return err;
+ }
+
+ regd_buf = &iser_task->rdma_regd[ISER_DIR_OUT];
+
+ if (unsol_sz < edtl) {
+ hdr->flags |= ISER_WSV;
+ hdr->write_stag = cpu_to_be32(regd_buf->reg.rkey);
+ hdr->write_va = cpu_to_be64(regd_buf->reg.va + unsol_sz);
+
+ iser_dbg("Cmd itt:%d, WRITE tags, RKEY:%#.4X "
+ "VA:%#llX + unsol:%d\n",
+ task->itt, regd_buf->reg.rkey,
+ (unsigned long long)regd_buf->reg.va, unsol_sz);
+ }
+
+ if (imm_sz > 0) {
+ iser_dbg("Cmd itt:%d, WRITE, adding imm.data sz: %d\n",
+ task->itt, imm_sz);
+ tx_dsg->addr = regd_buf->reg.va;
+ tx_dsg->length = imm_sz;
+ tx_dsg->lkey = regd_buf->reg.lkey;
+ iser_task->desc.num_sge = 2;
+ }
+
+ return 0;
+}
+
+/* creates a new tx descriptor and adds header regd buffer */
+static void iser_create_send_desc(struct iser_conn *ib_conn,
+ struct iser_tx_desc *tx_desc)
+{
+ struct iser_device *device = ib_conn->device;
+
+ ib_dma_sync_single_for_cpu(device->ib_device,
+ tx_desc->dma_addr, ISER_HEADERS_LEN, DMA_TO_DEVICE);
+
+ memset(&tx_desc->iser_header, 0, sizeof(struct iser_hdr));
+ tx_desc->iser_header.flags = ISER_VER;
+
+ tx_desc->num_sge = 1;
+
+ if (tx_desc->tx_sg[0].lkey != device->mr->lkey) {
+ tx_desc->tx_sg[0].lkey = device->mr->lkey;
+ iser_dbg("sdesc %p lkey mismatch, fixing\n", tx_desc);
+ }
+}
+
+
+int iser_alloc_rx_descriptors(struct iser_conn *ib_conn)
+{
+ int i, j;
+ u64 dma_addr;
+ struct iser_rx_desc *rx_desc;
+ struct ib_sge *rx_sg;
+ struct iser_device *device = ib_conn->device;
+
+ ib_conn->rx_descs = kmalloc(ISER_QP_MAX_RECV_DTOS *
+ sizeof(struct iser_rx_desc), GFP_KERNEL);
+ if (!ib_conn->rx_descs)
+ goto rx_desc_alloc_fail;
+
+ rx_desc = ib_conn->rx_descs;
+
+ for (i = 0; i < ISER_QP_MAX_RECV_DTOS; i++, rx_desc++) {
+ dma_addr = ib_dma_map_single(device->ib_device, (void *)rx_desc,
+ ISER_RX_PAYLOAD_SIZE, DMA_FROM_DEVICE);
+ if (ib_dma_mapping_error(device->ib_device, dma_addr))
+ goto rx_desc_dma_map_failed;
+
+ rx_desc->dma_addr = dma_addr;
+
+ rx_sg = &rx_desc->rx_sg;
+ rx_sg->addr = rx_desc->dma_addr;
+ rx_sg->length = ISER_RX_PAYLOAD_SIZE;
+ rx_sg->lkey = device->mr->lkey;
+ }
+
+ ib_conn->rx_desc_head = 0;
+ return 0;
+
+rx_desc_dma_map_failed:
+ rx_desc = ib_conn->rx_descs;
+ for (j = 0; j < i; j++, rx_desc++)
+ ib_dma_unmap_single(device->ib_device, rx_desc->dma_addr,
+ ISER_RX_PAYLOAD_SIZE, DMA_FROM_DEVICE);
+ kfree(ib_conn->rx_descs);
+ ib_conn->rx_descs = NULL;
+rx_desc_alloc_fail:
+ iser_err("failed allocating rx descriptors / data buffers\n");
+ return -ENOMEM;
+}
+
+void iser_free_rx_descriptors(struct iser_conn *ib_conn)
+{
+ int i;
+ struct iser_rx_desc *rx_desc;
+ struct iser_device *device = ib_conn->device;
+
+ if (!ib_conn->rx_descs)
+ return;
+
+ rx_desc = ib_conn->rx_descs;
+ for (i = 0; i < ISER_QP_MAX_RECV_DTOS; i++, rx_desc++)
+ ib_dma_unmap_single(device->ib_device, rx_desc->dma_addr,
+ ISER_RX_PAYLOAD_SIZE, DMA_FROM_DEVICE);
+ kfree(ib_conn->rx_descs);
+}
+
+static int iser_post_rx_bufs(struct iscsi_conn *conn, struct iscsi_hdr *req)
+{
+ struct iscsi_iser_conn *iser_conn = conn->dd_data;
+
+ iser_dbg("req op %x flags %x\n", req->opcode, req->flags);
+ /* check if this is the last login - going to full feature phase */
+ if ((req->flags & ISCSI_FULL_FEATURE_PHASE) != ISCSI_FULL_FEATURE_PHASE)
+ return 0;
+
+ /*
+ * Check that there is one posted recv buffer (for the last login
+ * response) and no posted send buffers left - they must have been
+ * consumed during previous login phases.
+ */
+ WARN_ON(iser_conn->ib_conn->post_recv_buf_count != 1);
+ WARN_ON(atomic_read(&iser_conn->ib_conn->post_send_buf_count) != 0);
+
+ iser_dbg("Initially post: %d\n", ISER_MIN_POSTED_RX);
+ /* Initial post receive buffers */
+ if (iser_post_recvm(iser_conn->ib_conn, ISER_MIN_POSTED_RX))
+ return -ENOMEM;
+
+ return 0;
+}
+
+/**
+ * iser_send_command - send command PDU
+ */
+int iser_send_command(struct iscsi_conn *conn,
+ struct iscsi_task *task)
+{
+ struct iscsi_iser_conn *iser_conn = conn->dd_data;
+ struct iscsi_iser_task *iser_task = task->dd_data;
+ unsigned long edtl;
+ int err;
+ struct iser_data_buf *data_buf;
+ struct iscsi_scsi_req *hdr = (struct iscsi_scsi_req *)task->hdr;
+ struct scsi_cmnd *sc = task->sc;
+ struct iser_tx_desc *tx_desc = &iser_task->desc;
+
+ edtl = ntohl(hdr->data_length);
+
+ /* build the tx desc regd header and add it to the tx desc dto */
+ tx_desc->type = ISCSI_TX_SCSI_COMMAND;
+ iser_create_send_desc(iser_conn->ib_conn, tx_desc);
+
+ if (hdr->flags & ISCSI_FLAG_CMD_READ)
+ data_buf = &iser_task->data[ISER_DIR_IN];
+ else
+ data_buf = &iser_task->data[ISER_DIR_OUT];
+
+ if (scsi_sg_count(sc)) { /* using a scatter list */
+ data_buf->buf = scsi_sglist(sc);
+ data_buf->size = scsi_sg_count(sc);
+ }
+
+ data_buf->data_len = scsi_bufflen(sc);
+
+ if (hdr->flags & ISCSI_FLAG_CMD_READ) {
+ err = iser_prepare_read_cmd(task, edtl);
+ if (err)
+ goto send_command_error;
+ }
+ if (hdr->flags & ISCSI_FLAG_CMD_WRITE) {
+ err = iser_prepare_write_cmd(task,
+ task->imm_count,
+ task->imm_count +
+ task->unsol_r2t.data_length,
+ edtl);
+ if (err)
+ goto send_command_error;
+ }
+
+ iser_task->status = ISER_TASK_STATUS_STARTED;
+
+ err = iser_post_send(iser_conn->ib_conn, tx_desc);
+ if (!err)
+ return 0;
+
+send_command_error:
+ iser_err("conn %p failed task->itt %d err %d\n",conn, task->itt, err);
+ return err;
+}
+
+/**
+ * iser_send_data_out - send data out PDU
+ */
+int iser_send_data_out(struct iscsi_conn *conn,
+ struct iscsi_task *task,
+ struct iscsi_data *hdr)
+{
+ struct iscsi_iser_conn *iser_conn = conn->dd_data;
+ struct iscsi_iser_task *iser_task = task->dd_data;
+ struct iser_tx_desc *tx_desc = NULL;
+ struct iser_regd_buf *regd_buf;
+ unsigned long buf_offset;
+ unsigned long data_seg_len;
+ uint32_t itt;
+ int err = 0;
+ struct ib_sge *tx_dsg;
+
+ itt = (__force uint32_t)hdr->itt;
+ data_seg_len = ntoh24(hdr->dlength);
+ buf_offset = ntohl(hdr->offset);
+
+ iser_dbg("%s itt %d dseg_len %d offset %d\n",
+ __func__,(int)itt,(int)data_seg_len,(int)buf_offset);
+
+ tx_desc = kmem_cache_zalloc(ig.desc_cache, GFP_ATOMIC);
+ if (tx_desc == NULL) {
+ iser_err("Failed to alloc desc for post dataout\n");
+ return -ENOMEM;
+ }
+
+ tx_desc->type = ISCSI_TX_DATAOUT;
+ tx_desc->iser_header.flags = ISER_VER;
+ memcpy(&tx_desc->iscsi_header, hdr, sizeof(struct iscsi_hdr));
+
+ /* build the tx desc */
+ iser_initialize_task_headers(task, tx_desc);
+
+ regd_buf = &iser_task->rdma_regd[ISER_DIR_OUT];
+ tx_dsg = &tx_desc->tx_sg[1];
+ tx_dsg->addr = regd_buf->reg.va + buf_offset;
+ tx_dsg->length = data_seg_len;
+ tx_dsg->lkey = regd_buf->reg.lkey;
+ tx_desc->num_sge = 2;
+
+ if (buf_offset + data_seg_len > iser_task->data[ISER_DIR_OUT].data_len) {
+ iser_err("Offset:%ld & DSL:%ld in Data-Out "
+ "inconsistent with total len:%ld, itt:%d\n",
+ buf_offset, data_seg_len,
+ iser_task->data[ISER_DIR_OUT].data_len, itt);
+ err = -EINVAL;
+ goto send_data_out_error;
+ }
+ iser_dbg("data-out itt: %d, offset: %ld, sz: %ld\n",
+ itt, buf_offset, data_seg_len);
+
+
+ err = iser_post_send(iser_conn->ib_conn, tx_desc);
+ if (!err)
+ return 0;
+
+send_data_out_error:
+ kmem_cache_free(ig.desc_cache, tx_desc);
+ iser_err("conn %p failed err %d\n",conn, err);
+ return err;
+}
+
+int iser_send_control(struct iscsi_conn *conn,
+ struct iscsi_task *task)
+{
+ struct iscsi_iser_conn *iser_conn = conn->dd_data;
+ struct iscsi_iser_task *iser_task = task->dd_data;
+ struct iser_tx_desc *mdesc = &iser_task->desc;
+ unsigned long data_seg_len;
+ int err = 0;
+ struct iser_device *device;
+ struct iser_conn *ib_conn = iser_conn->ib_conn;
+
+ /* build the tx desc regd header and add it to the tx desc dto */
+ mdesc->type = ISCSI_TX_CONTROL;
+ iser_create_send_desc(iser_conn->ib_conn, mdesc);
+
+ device = iser_conn->ib_conn->device;
+
+ data_seg_len = ntoh24(task->hdr->dlength);
+
+ if (data_seg_len > 0) {
+ struct ib_sge *tx_dsg = &mdesc->tx_sg[1];
+ if (task != conn->login_task) {
+ iser_err("data present on non login task!!!\n");
+ goto send_control_error;
+ }
+
+ ib_dma_sync_single_for_cpu(device->ib_device,
+ ib_conn->login_req_dma, task->data_count,
+ DMA_TO_DEVICE);
+
+ memcpy(iser_conn->ib_conn->login_req_buf, task->data,
+ task->data_count);
+
+ ib_dma_sync_single_for_device(device->ib_device,
+ ib_conn->login_req_dma, task->data_count,
+ DMA_TO_DEVICE);
+
+ tx_dsg->addr = iser_conn->ib_conn->login_req_dma;
+ tx_dsg->length = task->data_count;
+ tx_dsg->lkey = device->mr->lkey;
+ mdesc->num_sge = 2;
+ }
+
+ if (task == conn->login_task) {
+ err = iser_post_recvl(iser_conn->ib_conn);
+ if (err)
+ goto send_control_error;
+ err = iser_post_rx_bufs(conn, task->hdr);
+ if (err)
+ goto send_control_error;
+ }
+
+ err = iser_post_send(iser_conn->ib_conn, mdesc);
+ if (!err)
+ return 0;
+
+send_control_error:
+ iser_err("conn %p failed err %d\n",conn, err);
+ return err;
+}
+
+/**
+ * iser_rcv_dto_completion - recv DTO completion
+ */
+void iser_rcv_completion(struct iser_rx_desc *rx_desc,
+ unsigned long rx_xfer_len,
+ struct iser_conn *ib_conn)
+{
+ struct iscsi_iser_conn *conn = ib_conn->iser_conn;
+ struct iscsi_hdr *hdr;
+ u64 rx_dma;
+ int rx_buflen, outstanding, count, err;
+
+ /* differentiate between login to all other PDUs */
+ if ((char *)rx_desc == ib_conn->login_resp_buf) {
+ rx_dma = ib_conn->login_resp_dma;
+ rx_buflen = ISER_RX_LOGIN_SIZE;
+ } else {
+ rx_dma = rx_desc->dma_addr;
+ rx_buflen = ISER_RX_PAYLOAD_SIZE;
+ }
+
+ ib_dma_sync_single_for_cpu(ib_conn->device->ib_device, rx_dma,
+ rx_buflen, DMA_FROM_DEVICE);
+
+ hdr = &rx_desc->iscsi_header;
+
+ iser_dbg("op 0x%x itt 0x%x dlen %d\n", hdr->opcode,
+ hdr->itt, (int)(rx_xfer_len - ISER_HEADERS_LEN));
+
+ iscsi_iser_recv(conn->iscsi_conn, hdr,
+ rx_desc->data, rx_xfer_len - ISER_HEADERS_LEN);
+
+ ib_dma_sync_single_for_device(ib_conn->device->ib_device, rx_dma,
+ rx_buflen, DMA_FROM_DEVICE);
+
+ /* decrementing conn->post_recv_buf_count only --after-- freeing the *
+ * task eliminates the need to worry on tasks which are completed in *
+ * parallel to the execution of iser_conn_term. So the code that waits *
+ * for the posted rx bufs refcount to become zero handles everything */
+ conn->ib_conn->post_recv_buf_count--;
+
+ if (rx_dma == ib_conn->login_resp_dma)
+ return;
+
+ outstanding = ib_conn->post_recv_buf_count;
+ if (outstanding + ISER_MIN_POSTED_RX <= ISER_QP_MAX_RECV_DTOS) {
+ count = min(ISER_QP_MAX_RECV_DTOS - outstanding,
+ ISER_MIN_POSTED_RX);
+ err = iser_post_recvm(ib_conn, count);
+ if (err)
+ iser_err("posting %d rx bufs err %d\n", count, err);
+ }
+}
+
+void iser_snd_completion(struct iser_tx_desc *tx_desc,
+ struct iser_conn *ib_conn)
+{
+ struct iscsi_task *task;
+ struct iser_device *device = ib_conn->device;
+
+ if (tx_desc->type == ISCSI_TX_DATAOUT) {
+ ib_dma_unmap_single(device->ib_device, tx_desc->dma_addr,
+ ISER_HEADERS_LEN, DMA_TO_DEVICE);
+ kmem_cache_free(ig.desc_cache, tx_desc);
+ }
+
+ atomic_dec(&ib_conn->post_send_buf_count);
+
+ if (tx_desc->type == ISCSI_TX_CONTROL) {
+ /* this arithmetic is legal by libiscsi dd_data allocation */
+ task = (void *) ((long)(void *)tx_desc -
+ sizeof(struct iscsi_task));
+ if (task->hdr->itt == RESERVED_ITT)
+ iscsi_put_task(task);
+ }
+}
+
+void iser_task_rdma_init(struct iscsi_iser_task *iser_task)
+
+{
+ iser_task->status = ISER_TASK_STATUS_INIT;
+
+ iser_task->dir[ISER_DIR_IN] = 0;
+ iser_task->dir[ISER_DIR_OUT] = 0;
+
+ iser_task->data[ISER_DIR_IN].data_len = 0;
+ iser_task->data[ISER_DIR_OUT].data_len = 0;
+
+ memset(&iser_task->rdma_regd[ISER_DIR_IN], 0,
+ sizeof(struct iser_regd_buf));
+ memset(&iser_task->rdma_regd[ISER_DIR_OUT], 0,
+ sizeof(struct iser_regd_buf));
+}
+
+void iser_task_rdma_finalize(struct iscsi_iser_task *iser_task)
+{
+ int is_rdma_aligned = 1;
+ struct iser_regd_buf *regd;
+
+ /* if we were reading, copy back to unaligned sglist,
+ * anyway dma_unmap and free the copy
+ */
+ if (iser_task->data_copy[ISER_DIR_IN].copy_buf != NULL) {
+ is_rdma_aligned = 0;
+ iser_finalize_rdma_unaligned_sg(iser_task, ISER_DIR_IN);
+ }
+ if (iser_task->data_copy[ISER_DIR_OUT].copy_buf != NULL) {
+ is_rdma_aligned = 0;
+ iser_finalize_rdma_unaligned_sg(iser_task, ISER_DIR_OUT);
+ }
+
+ if (iser_task->dir[ISER_DIR_IN]) {
+ regd = &iser_task->rdma_regd[ISER_DIR_IN];
+ if (regd->reg.is_fmr)
+ iser_unreg_mem(&regd->reg);
+ }
+
+ if (iser_task->dir[ISER_DIR_OUT]) {
+ regd = &iser_task->rdma_regd[ISER_DIR_OUT];
+ if (regd->reg.is_fmr)
+ iser_unreg_mem(&regd->reg);
+ }
+
+ /* if the data was unaligned, it was already unmapped and then copied */
+ if (is_rdma_aligned)
+ iser_dma_unmap_task_data(iser_task);
+}
diff --git a/drivers/infiniband/ulp/iser/iser_memory.c b/drivers/infiniband/ulp/iser/iser_memory.c
new file mode 100644
index 00000000..2033a928
--- /dev/null
+++ b/drivers/infiniband/ulp/iser/iser_memory.c
@@ -0,0 +1,422 @@
+/*
+ * Copyright (c) 2004, 2005, 2006 Voltaire, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/mm.h>
+#include <linux/highmem.h>
+#include <linux/scatterlist.h>
+
+#include "iscsi_iser.h"
+
+#define ISER_KMALLOC_THRESHOLD 0x20000 /* 128K - kmalloc limit */
+
+/**
+ * iser_start_rdma_unaligned_sg
+ */
+static int iser_start_rdma_unaligned_sg(struct iscsi_iser_task *iser_task,
+ enum iser_data_dir cmd_dir)
+{
+ int dma_nents;
+ struct ib_device *dev;
+ char *mem = NULL;
+ struct iser_data_buf *data = &iser_task->data[cmd_dir];
+ unsigned long cmd_data_len = data->data_len;
+
+ if (cmd_data_len > ISER_KMALLOC_THRESHOLD)
+ mem = (void *)__get_free_pages(GFP_ATOMIC,
+ ilog2(roundup_pow_of_two(cmd_data_len)) - PAGE_SHIFT);
+ else
+ mem = kmalloc(cmd_data_len, GFP_ATOMIC);
+
+ if (mem == NULL) {
+ iser_err("Failed to allocate mem size %d %d for copying sglist\n",
+ data->size,(int)cmd_data_len);
+ return -ENOMEM;
+ }
+
+ if (cmd_dir == ISER_DIR_OUT) {
+ /* copy the unaligned sg the buffer which is used for RDMA */
+ struct scatterlist *sgl = (struct scatterlist *)data->buf;
+ struct scatterlist *sg;
+ int i;
+ char *p, *from;
+
+ p = mem;
+ for_each_sg(sgl, sg, data->size, i) {
+ from = kmap_atomic(sg_page(sg));
+ memcpy(p,
+ from + sg->offset,
+ sg->length);
+ kunmap_atomic(from);
+ p += sg->length;
+ }
+ }
+
+ sg_init_one(&iser_task->data_copy[cmd_dir].sg_single, mem, cmd_data_len);
+ iser_task->data_copy[cmd_dir].buf =
+ &iser_task->data_copy[cmd_dir].sg_single;
+ iser_task->data_copy[cmd_dir].size = 1;
+
+ iser_task->data_copy[cmd_dir].copy_buf = mem;
+
+ dev = iser_task->iser_conn->ib_conn->device->ib_device;
+ dma_nents = ib_dma_map_sg(dev,
+ &iser_task->data_copy[cmd_dir].sg_single,
+ 1,
+ (cmd_dir == ISER_DIR_OUT) ?
+ DMA_TO_DEVICE : DMA_FROM_DEVICE);
+ BUG_ON(dma_nents == 0);
+
+ iser_task->data_copy[cmd_dir].dma_nents = dma_nents;
+ return 0;
+}
+
+/**
+ * iser_finalize_rdma_unaligned_sg
+ */
+void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_task *iser_task,
+ enum iser_data_dir cmd_dir)
+{
+ struct ib_device *dev;
+ struct iser_data_buf *mem_copy;
+ unsigned long cmd_data_len;
+
+ dev = iser_task->iser_conn->ib_conn->device->ib_device;
+ mem_copy = &iser_task->data_copy[cmd_dir];
+
+ ib_dma_unmap_sg(dev, &mem_copy->sg_single, 1,
+ (cmd_dir == ISER_DIR_OUT) ?
+ DMA_TO_DEVICE : DMA_FROM_DEVICE);
+
+ if (cmd_dir == ISER_DIR_IN) {
+ char *mem;
+ struct scatterlist *sgl, *sg;
+ unsigned char *p, *to;
+ unsigned int sg_size;
+ int i;
+
+ /* copy back read RDMA to unaligned sg */
+ mem = mem_copy->copy_buf;
+
+ sgl = (struct scatterlist *)iser_task->data[ISER_DIR_IN].buf;
+ sg_size = iser_task->data[ISER_DIR_IN].size;
+
+ p = mem;
+ for_each_sg(sgl, sg, sg_size, i) {
+ to = kmap_atomic(sg_page(sg));
+ memcpy(to + sg->offset,
+ p,
+ sg->length);
+ kunmap_atomic(to);
+ p += sg->length;
+ }
+ }
+
+ cmd_data_len = iser_task->data[cmd_dir].data_len;
+
+ if (cmd_data_len > ISER_KMALLOC_THRESHOLD)
+ free_pages((unsigned long)mem_copy->copy_buf,
+ ilog2(roundup_pow_of_two(cmd_data_len)) - PAGE_SHIFT);
+ else
+ kfree(mem_copy->copy_buf);
+
+ mem_copy->copy_buf = NULL;
+}
+
+#define IS_4K_ALIGNED(addr) ((((unsigned long)addr) & ~MASK_4K) == 0)
+
+/**
+ * iser_sg_to_page_vec - Translates scatterlist entries to physical addresses
+ * and returns the length of resulting physical address array (may be less than
+ * the original due to possible compaction).
+ *
+ * we build a "page vec" under the assumption that the SG meets the RDMA
+ * alignment requirements. Other then the first and last SG elements, all
+ * the "internal" elements can be compacted into a list whose elements are
+ * dma addresses of physical pages. The code supports also the weird case
+ * where --few fragments of the same page-- are present in the SG as
+ * consecutive elements. Also, it handles one entry SG.
+ */
+
+static int iser_sg_to_page_vec(struct iser_data_buf *data,
+ struct iser_page_vec *page_vec,
+ struct ib_device *ibdev)
+{
+ struct scatterlist *sg, *sgl = (struct scatterlist *)data->buf;
+ u64 start_addr, end_addr, page, chunk_start = 0;
+ unsigned long total_sz = 0;
+ unsigned int dma_len;
+ int i, new_chunk, cur_page, last_ent = data->dma_nents - 1;
+
+ /* compute the offset of first element */
+ page_vec->offset = (u64) sgl[0].offset & ~MASK_4K;
+
+ new_chunk = 1;
+ cur_page = 0;
+ for_each_sg(sgl, sg, data->dma_nents, i) {
+ start_addr = ib_sg_dma_address(ibdev, sg);
+ if (new_chunk)
+ chunk_start = start_addr;
+ dma_len = ib_sg_dma_len(ibdev, sg);
+ end_addr = start_addr + dma_len;
+ total_sz += dma_len;
+
+ /* collect page fragments until aligned or end of SG list */
+ if (!IS_4K_ALIGNED(end_addr) && i < last_ent) {
+ new_chunk = 0;
+ continue;
+ }
+ new_chunk = 1;
+
+ /* address of the first page in the contiguous chunk;
+ masking relevant for the very first SG entry,
+ which might be unaligned */
+ page = chunk_start & MASK_4K;
+ do {
+ page_vec->pages[cur_page++] = page;
+ page += SIZE_4K;
+ } while (page < end_addr);
+ }
+
+ page_vec->data_size = total_sz;
+ iser_dbg("page_vec->data_size:%d cur_page %d\n", page_vec->data_size,cur_page);
+ return cur_page;
+}
+
+
+/**
+ * iser_data_buf_aligned_len - Tries to determine the maximal correctly aligned
+ * for RDMA sub-list of a scatter-gather list of memory buffers, and returns
+ * the number of entries which are aligned correctly. Supports the case where
+ * consecutive SG elements are actually fragments of the same physcial page.
+ */
+static int iser_data_buf_aligned_len(struct iser_data_buf *data,
+ struct ib_device *ibdev)
+{
+ struct scatterlist *sgl, *sg, *next_sg = NULL;
+ u64 start_addr, end_addr;
+ int i, ret_len, start_check = 0;
+
+ if (data->dma_nents == 1)
+ return 1;
+
+ sgl = (struct scatterlist *)data->buf;
+ start_addr = ib_sg_dma_address(ibdev, sgl);
+
+ for_each_sg(sgl, sg, data->dma_nents, i) {
+ if (start_check && !IS_4K_ALIGNED(start_addr))
+ break;
+
+ next_sg = sg_next(sg);
+ if (!next_sg)
+ break;
+
+ end_addr = start_addr + ib_sg_dma_len(ibdev, sg);
+ start_addr = ib_sg_dma_address(ibdev, next_sg);
+
+ if (end_addr == start_addr) {
+ start_check = 0;
+ continue;
+ } else
+ start_check = 1;
+
+ if (!IS_4K_ALIGNED(end_addr))
+ break;
+ }
+ ret_len = (next_sg) ? i : i+1;
+ iser_dbg("Found %d aligned entries out of %d in sg:0x%p\n",
+ ret_len, data->dma_nents, data);
+ return ret_len;
+}
+
+static void iser_data_buf_dump(struct iser_data_buf *data,
+ struct ib_device *ibdev)
+{
+ struct scatterlist *sgl = (struct scatterlist *)data->buf;
+ struct scatterlist *sg;
+ int i;
+
+ if (iser_debug_level == 0)
+ return;
+
+ for_each_sg(sgl, sg, data->dma_nents, i)
+ iser_warn("sg[%d] dma_addr:0x%lX page:0x%p "
+ "off:0x%x sz:0x%x dma_len:0x%x\n",
+ i, (unsigned long)ib_sg_dma_address(ibdev, sg),
+ sg_page(sg), sg->offset,
+ sg->length, ib_sg_dma_len(ibdev, sg));
+}
+
+static void iser_dump_page_vec(struct iser_page_vec *page_vec)
+{
+ int i;
+
+ iser_err("page vec length %d data size %d\n",
+ page_vec->length, page_vec->data_size);
+ for (i = 0; i < page_vec->length; i++)
+ iser_err("%d %lx\n",i,(unsigned long)page_vec->pages[i]);
+}
+
+static void iser_page_vec_build(struct iser_data_buf *data,
+ struct iser_page_vec *page_vec,
+ struct ib_device *ibdev)
+{
+ int page_vec_len = 0;
+
+ page_vec->length = 0;
+ page_vec->offset = 0;
+
+ iser_dbg("Translating sg sz: %d\n", data->dma_nents);
+ page_vec_len = iser_sg_to_page_vec(data, page_vec, ibdev);
+ iser_dbg("sg len %d page_vec_len %d\n", data->dma_nents,page_vec_len);
+
+ page_vec->length = page_vec_len;
+
+ if (page_vec_len * SIZE_4K < page_vec->data_size) {
+ iser_err("page_vec too short to hold this SG\n");
+ iser_data_buf_dump(data, ibdev);
+ iser_dump_page_vec(page_vec);
+ BUG();
+ }
+}
+
+int iser_dma_map_task_data(struct iscsi_iser_task *iser_task,
+ struct iser_data_buf *data,
+ enum iser_data_dir iser_dir,
+ enum dma_data_direction dma_dir)
+{
+ struct ib_device *dev;
+
+ iser_task->dir[iser_dir] = 1;
+ dev = iser_task->iser_conn->ib_conn->device->ib_device;
+
+ data->dma_nents = ib_dma_map_sg(dev, data->buf, data->size, dma_dir);
+ if (data->dma_nents == 0) {
+ iser_err("dma_map_sg failed!!!\n");
+ return -EINVAL;
+ }
+ return 0;
+}
+
+void iser_dma_unmap_task_data(struct iscsi_iser_task *iser_task)
+{
+ struct ib_device *dev;
+ struct iser_data_buf *data;
+
+ dev = iser_task->iser_conn->ib_conn->device->ib_device;
+
+ if (iser_task->dir[ISER_DIR_IN]) {
+ data = &iser_task->data[ISER_DIR_IN];
+ ib_dma_unmap_sg(dev, data->buf, data->size, DMA_FROM_DEVICE);
+ }
+
+ if (iser_task->dir[ISER_DIR_OUT]) {
+ data = &iser_task->data[ISER_DIR_OUT];
+ ib_dma_unmap_sg(dev, data->buf, data->size, DMA_TO_DEVICE);
+ }
+}
+
+/**
+ * iser_reg_rdma_mem - Registers memory intended for RDMA,
+ * obtaining rkey and va
+ *
+ * returns 0 on success, errno code on failure
+ */
+int iser_reg_rdma_mem(struct iscsi_iser_task *iser_task,
+ enum iser_data_dir cmd_dir)
+{
+ struct iscsi_conn *iscsi_conn = iser_task->iser_conn->iscsi_conn;
+ struct iser_conn *ib_conn = iser_task->iser_conn->ib_conn;
+ struct iser_device *device = ib_conn->device;
+ struct ib_device *ibdev = device->ib_device;
+ struct iser_data_buf *mem = &iser_task->data[cmd_dir];
+ struct iser_regd_buf *regd_buf;
+ int aligned_len;
+ int err;
+ int i;
+ struct scatterlist *sg;
+
+ regd_buf = &iser_task->rdma_regd[cmd_dir];
+
+ aligned_len = iser_data_buf_aligned_len(mem, ibdev);
+ if (aligned_len != mem->dma_nents) {
+ iscsi_conn->fmr_unalign_cnt++;
+ iser_warn("rdma alignment violation %d/%d aligned\n",
+ aligned_len, mem->size);
+ iser_data_buf_dump(mem, ibdev);
+
+ /* unmap the command data before accessing it */
+ iser_dma_unmap_task_data(iser_task);
+
+ /* allocate copy buf, if we are writing, copy the */
+ /* unaligned scatterlist, dma map the copy */
+ if (iser_start_rdma_unaligned_sg(iser_task, cmd_dir) != 0)
+ return -ENOMEM;
+ mem = &iser_task->data_copy[cmd_dir];
+ }
+
+ /* if there a single dma entry, FMR is not needed */
+ if (mem->dma_nents == 1) {
+ sg = (struct scatterlist *)mem->buf;
+
+ regd_buf->reg.lkey = device->mr->lkey;
+ regd_buf->reg.rkey = device->mr->rkey;
+ regd_buf->reg.len = ib_sg_dma_len(ibdev, &sg[0]);
+ regd_buf->reg.va = ib_sg_dma_address(ibdev, &sg[0]);
+ regd_buf->reg.is_fmr = 0;
+
+ iser_dbg("PHYSICAL Mem.register: lkey: 0x%08X rkey: 0x%08X "
+ "va: 0x%08lX sz: %ld]\n",
+ (unsigned int)regd_buf->reg.lkey,
+ (unsigned int)regd_buf->reg.rkey,
+ (unsigned long)regd_buf->reg.va,
+ (unsigned long)regd_buf->reg.len);
+ } else { /* use FMR for multiple dma entries */
+ iser_page_vec_build(mem, ib_conn->page_vec, ibdev);
+ err = iser_reg_page_vec(ib_conn, ib_conn->page_vec, &regd_buf->reg);
+ if (err) {
+ iser_data_buf_dump(mem, ibdev);
+ iser_err("mem->dma_nents = %d (dlength = 0x%x)\n",
+ mem->dma_nents,
+ ntoh24(iser_task->desc.iscsi_header.dlength));
+ iser_err("page_vec: data_size = 0x%x, length = %d, offset = 0x%x\n",
+ ib_conn->page_vec->data_size, ib_conn->page_vec->length,
+ ib_conn->page_vec->offset);
+ for (i=0 ; i<ib_conn->page_vec->length ; i++)
+ iser_err("page_vec[%d] = 0x%llx\n", i,
+ (unsigned long long) ib_conn->page_vec->pages[i]);
+ return err;
+ }
+ }
+ return 0;
+}
diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c
new file mode 100644
index 00000000..14224ba4
--- /dev/null
+++ b/drivers/infiniband/ulp/iser/iser_verbs.c
@@ -0,0 +1,868 @@
+/*
+ * Copyright (c) 2004, 2005, 2006 Voltaire, Inc. All rights reserved.
+ * Copyright (c) 2005, 2006 Cisco Systems. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+
+#include "iscsi_iser.h"
+
+#define ISCSI_ISER_MAX_CONN 8
+#define ISER_MAX_RX_CQ_LEN (ISER_QP_MAX_RECV_DTOS * ISCSI_ISER_MAX_CONN)
+#define ISER_MAX_TX_CQ_LEN (ISER_QP_MAX_REQ_DTOS * ISCSI_ISER_MAX_CONN)
+
+static void iser_cq_tasklet_fn(unsigned long data);
+static void iser_cq_callback(struct ib_cq *cq, void *cq_context);
+
+static void iser_cq_event_callback(struct ib_event *cause, void *context)
+{
+ iser_err("got cq event %d \n", cause->event);
+}
+
+static void iser_qp_event_callback(struct ib_event *cause, void *context)
+{
+ iser_err("got qp event %d\n",cause->event);
+}
+
+static void iser_event_handler(struct ib_event_handler *handler,
+ struct ib_event *event)
+{
+ iser_err("async event %d on device %s port %d\n", event->event,
+ event->device->name, event->element.port_num);
+}
+
+/**
+ * iser_create_device_ib_res - creates Protection Domain (PD), Completion
+ * Queue (CQ), DMA Memory Region (DMA MR) with the device associated with
+ * the adapator.
+ *
+ * returns 0 on success, -1 on failure
+ */
+static int iser_create_device_ib_res(struct iser_device *device)
+{
+ device->pd = ib_alloc_pd(device->ib_device);
+ if (IS_ERR(device->pd))
+ goto pd_err;
+
+ device->rx_cq = ib_create_cq(device->ib_device,
+ iser_cq_callback,
+ iser_cq_event_callback,
+ (void *)device,
+ ISER_MAX_RX_CQ_LEN, 0);
+ if (IS_ERR(device->rx_cq))
+ goto rx_cq_err;
+
+ device->tx_cq = ib_create_cq(device->ib_device,
+ NULL, iser_cq_event_callback,
+ (void *)device,
+ ISER_MAX_TX_CQ_LEN, 0);
+
+ if (IS_ERR(device->tx_cq))
+ goto tx_cq_err;
+
+ if (ib_req_notify_cq(device->rx_cq, IB_CQ_NEXT_COMP))
+ goto cq_arm_err;
+
+ tasklet_init(&device->cq_tasklet,
+ iser_cq_tasklet_fn,
+ (unsigned long)device);
+
+ device->mr = ib_get_dma_mr(device->pd, IB_ACCESS_LOCAL_WRITE |
+ IB_ACCESS_REMOTE_WRITE |
+ IB_ACCESS_REMOTE_READ);
+ if (IS_ERR(device->mr))
+ goto dma_mr_err;
+
+ INIT_IB_EVENT_HANDLER(&device->event_handler, device->ib_device,
+ iser_event_handler);
+ if (ib_register_event_handler(&device->event_handler))
+ goto handler_err;
+
+ return 0;
+
+handler_err:
+ ib_dereg_mr(device->mr);
+dma_mr_err:
+ tasklet_kill(&device->cq_tasklet);
+cq_arm_err:
+ ib_destroy_cq(device->tx_cq);
+tx_cq_err:
+ ib_destroy_cq(device->rx_cq);
+rx_cq_err:
+ ib_dealloc_pd(device->pd);
+pd_err:
+ iser_err("failed to allocate an IB resource\n");
+ return -1;
+}
+
+/**
+ * iser_free_device_ib_res - destroy/dealloc/dereg the DMA MR,
+ * CQ and PD created with the device associated with the adapator.
+ */
+static void iser_free_device_ib_res(struct iser_device *device)
+{
+ BUG_ON(device->mr == NULL);
+
+ tasklet_kill(&device->cq_tasklet);
+ (void)ib_unregister_event_handler(&device->event_handler);
+ (void)ib_dereg_mr(device->mr);
+ (void)ib_destroy_cq(device->tx_cq);
+ (void)ib_destroy_cq(device->rx_cq);
+ (void)ib_dealloc_pd(device->pd);
+
+ device->mr = NULL;
+ device->tx_cq = NULL;
+ device->rx_cq = NULL;
+ device->pd = NULL;
+}
+
+/**
+ * iser_create_ib_conn_res - Creates FMR pool and Queue-Pair (QP)
+ *
+ * returns 0 on success, -1 on failure
+ */
+static int iser_create_ib_conn_res(struct iser_conn *ib_conn)
+{
+ struct iser_device *device;
+ struct ib_qp_init_attr init_attr;
+ int req_err, resp_err, ret = -ENOMEM;
+ struct ib_fmr_pool_param params;
+
+ BUG_ON(ib_conn->device == NULL);
+
+ device = ib_conn->device;
+
+ ib_conn->login_buf = kmalloc(ISCSI_DEF_MAX_RECV_SEG_LEN +
+ ISER_RX_LOGIN_SIZE, GFP_KERNEL);
+ if (!ib_conn->login_buf)
+ goto out_err;
+
+ ib_conn->login_req_buf = ib_conn->login_buf;
+ ib_conn->login_resp_buf = ib_conn->login_buf + ISCSI_DEF_MAX_RECV_SEG_LEN;
+
+ ib_conn->login_req_dma = ib_dma_map_single(ib_conn->device->ib_device,
+ (void *)ib_conn->login_req_buf,
+ ISCSI_DEF_MAX_RECV_SEG_LEN, DMA_TO_DEVICE);
+
+ ib_conn->login_resp_dma = ib_dma_map_single(ib_conn->device->ib_device,
+ (void *)ib_conn->login_resp_buf,
+ ISER_RX_LOGIN_SIZE, DMA_FROM_DEVICE);
+
+ req_err = ib_dma_mapping_error(device->ib_device, ib_conn->login_req_dma);
+ resp_err = ib_dma_mapping_error(device->ib_device, ib_conn->login_resp_dma);
+
+ if (req_err || resp_err) {
+ if (req_err)
+ ib_conn->login_req_dma = 0;
+ if (resp_err)
+ ib_conn->login_resp_dma = 0;
+ goto out_err;
+ }
+
+ ib_conn->page_vec = kmalloc(sizeof(struct iser_page_vec) +
+ (sizeof(u64) * (ISCSI_ISER_SG_TABLESIZE +1)),
+ GFP_KERNEL);
+ if (!ib_conn->page_vec)
+ goto out_err;
+
+ ib_conn->page_vec->pages = (u64 *) (ib_conn->page_vec + 1);
+
+ params.page_shift = SHIFT_4K;
+ /* when the first/last SG element are not start/end *
+ * page aligned, the map whould be of N+1 pages */
+ params.max_pages_per_fmr = ISCSI_ISER_SG_TABLESIZE + 1;
+ /* make the pool size twice the max number of SCSI commands *
+ * the ML is expected to queue, watermark for unmap at 50% */
+ params.pool_size = ISCSI_DEF_XMIT_CMDS_MAX * 2;
+ params.dirty_watermark = ISCSI_DEF_XMIT_CMDS_MAX;
+ params.cache = 0;
+ params.flush_function = NULL;
+ params.access = (IB_ACCESS_LOCAL_WRITE |
+ IB_ACCESS_REMOTE_WRITE |
+ IB_ACCESS_REMOTE_READ);
+
+ ib_conn->fmr_pool = ib_create_fmr_pool(device->pd, &params);
+ if (IS_ERR(ib_conn->fmr_pool)) {
+ ret = PTR_ERR(ib_conn->fmr_pool);
+ ib_conn->fmr_pool = NULL;
+ goto out_err;
+ }
+
+ memset(&init_attr, 0, sizeof init_attr);
+
+ init_attr.event_handler = iser_qp_event_callback;
+ init_attr.qp_context = (void *)ib_conn;
+ init_attr.send_cq = device->tx_cq;
+ init_attr.recv_cq = device->rx_cq;
+ init_attr.cap.max_send_wr = ISER_QP_MAX_REQ_DTOS;
+ init_attr.cap.max_recv_wr = ISER_QP_MAX_RECV_DTOS;
+ init_attr.cap.max_send_sge = 2;
+ init_attr.cap.max_recv_sge = 1;
+ init_attr.sq_sig_type = IB_SIGNAL_REQ_WR;
+ init_attr.qp_type = IB_QPT_RC;
+
+ ret = rdma_create_qp(ib_conn->cma_id, device->pd, &init_attr);
+ if (ret)
+ goto out_err;
+
+ ib_conn->qp = ib_conn->cma_id->qp;
+ iser_err("setting conn %p cma_id %p: fmr_pool %p qp %p\n",
+ ib_conn, ib_conn->cma_id,
+ ib_conn->fmr_pool, ib_conn->cma_id->qp);
+ return ret;
+
+out_err:
+ iser_err("unable to alloc mem or create resource, err %d\n", ret);
+ return ret;
+}
+
+/**
+ * releases the FMR pool, QP and CMA ID objects, returns 0 on success,
+ * -1 on failure
+ */
+static int iser_free_ib_conn_res(struct iser_conn *ib_conn, int can_destroy_id)
+{
+ BUG_ON(ib_conn == NULL);
+
+ iser_err("freeing conn %p cma_id %p fmr pool %p qp %p\n",
+ ib_conn, ib_conn->cma_id,
+ ib_conn->fmr_pool, ib_conn->qp);
+
+ /* qp is created only once both addr & route are resolved */
+ if (ib_conn->fmr_pool != NULL)
+ ib_destroy_fmr_pool(ib_conn->fmr_pool);
+
+ if (ib_conn->qp != NULL)
+ rdma_destroy_qp(ib_conn->cma_id);
+
+ /* if cma handler context, the caller acts s.t the cma destroy the id */
+ if (ib_conn->cma_id != NULL && can_destroy_id)
+ rdma_destroy_id(ib_conn->cma_id);
+
+ ib_conn->fmr_pool = NULL;
+ ib_conn->qp = NULL;
+ ib_conn->cma_id = NULL;
+ kfree(ib_conn->page_vec);
+
+ if (ib_conn->login_buf) {
+ if (ib_conn->login_req_dma)
+ ib_dma_unmap_single(ib_conn->device->ib_device,
+ ib_conn->login_req_dma,
+ ISCSI_DEF_MAX_RECV_SEG_LEN, DMA_TO_DEVICE);
+ if (ib_conn->login_resp_dma)
+ ib_dma_unmap_single(ib_conn->device->ib_device,
+ ib_conn->login_resp_dma,
+ ISER_RX_LOGIN_SIZE, DMA_FROM_DEVICE);
+ kfree(ib_conn->login_buf);
+ }
+
+ return 0;
+}
+
+/**
+ * based on the resolved device node GUID see if there already allocated
+ * device for this device. If there's no such, create one.
+ */
+static
+struct iser_device *iser_device_find_by_ib_device(struct rdma_cm_id *cma_id)
+{
+ struct iser_device *device;
+
+ mutex_lock(&ig.device_list_mutex);
+
+ list_for_each_entry(device, &ig.device_list, ig_list)
+ /* find if there's a match using the node GUID */
+ if (device->ib_device->node_guid == cma_id->device->node_guid)
+ goto inc_refcnt;
+
+ device = kzalloc(sizeof *device, GFP_KERNEL);
+ if (device == NULL)
+ goto out;
+
+ /* assign this device to the device */
+ device->ib_device = cma_id->device;
+ /* init the device and link it into ig device list */
+ if (iser_create_device_ib_res(device)) {
+ kfree(device);
+ device = NULL;
+ goto out;
+ }
+ list_add(&device->ig_list, &ig.device_list);
+
+inc_refcnt:
+ device->refcount++;
+out:
+ mutex_unlock(&ig.device_list_mutex);
+ return device;
+}
+
+/* if there's no demand for this device, release it */
+static void iser_device_try_release(struct iser_device *device)
+{
+ mutex_lock(&ig.device_list_mutex);
+ device->refcount--;
+ iser_err("device %p refcount %d\n",device,device->refcount);
+ if (!device->refcount) {
+ iser_free_device_ib_res(device);
+ list_del(&device->ig_list);
+ kfree(device);
+ }
+ mutex_unlock(&ig.device_list_mutex);
+}
+
+static int iser_conn_state_comp_exch(struct iser_conn *ib_conn,
+ enum iser_ib_conn_state comp,
+ enum iser_ib_conn_state exch)
+{
+ int ret;
+
+ spin_lock_bh(&ib_conn->lock);
+ if ((ret = (ib_conn->state == comp)))
+ ib_conn->state = exch;
+ spin_unlock_bh(&ib_conn->lock);
+ return ret;
+}
+
+/**
+ * Frees all conn objects and deallocs conn descriptor
+ */
+static void iser_conn_release(struct iser_conn *ib_conn, int can_destroy_id)
+{
+ struct iser_device *device = ib_conn->device;
+
+ BUG_ON(ib_conn->state != ISER_CONN_DOWN);
+
+ mutex_lock(&ig.connlist_mutex);
+ list_del(&ib_conn->conn_list);
+ mutex_unlock(&ig.connlist_mutex);
+ iser_free_rx_descriptors(ib_conn);
+ iser_free_ib_conn_res(ib_conn, can_destroy_id);
+ ib_conn->device = NULL;
+ /* on EVENT_ADDR_ERROR there's no device yet for this conn */
+ if (device != NULL)
+ iser_device_try_release(device);
+ iscsi_destroy_endpoint(ib_conn->ep);
+}
+
+void iser_conn_get(struct iser_conn *ib_conn)
+{
+ atomic_inc(&ib_conn->refcount);
+}
+
+int iser_conn_put(struct iser_conn *ib_conn, int can_destroy_id)
+{
+ if (atomic_dec_and_test(&ib_conn->refcount)) {
+ iser_conn_release(ib_conn, can_destroy_id);
+ return 1;
+ }
+ return 0;
+}
+
+/**
+ * triggers start of the disconnect procedures and wait for them to be done
+ */
+void iser_conn_terminate(struct iser_conn *ib_conn)
+{
+ int err = 0;
+
+ /* change the ib conn state only if the conn is UP, however always call
+ * rdma_disconnect since this is the only way to cause the CMA to change
+ * the QP state to ERROR
+ */
+
+ iser_conn_state_comp_exch(ib_conn, ISER_CONN_UP, ISER_CONN_TERMINATING);
+ err = rdma_disconnect(ib_conn->cma_id);
+ if (err)
+ iser_err("Failed to disconnect, conn: 0x%p err %d\n",
+ ib_conn,err);
+
+ wait_event_interruptible(ib_conn->wait,
+ ib_conn->state == ISER_CONN_DOWN);
+
+ iser_conn_put(ib_conn, 1); /* deref ib conn deallocate */
+}
+
+static int iser_connect_error(struct rdma_cm_id *cma_id)
+{
+ struct iser_conn *ib_conn;
+ ib_conn = (struct iser_conn *)cma_id->context;
+
+ ib_conn->state = ISER_CONN_DOWN;
+ wake_up_interruptible(&ib_conn->wait);
+ return iser_conn_put(ib_conn, 0); /* deref ib conn's cma id */
+}
+
+static int iser_addr_handler(struct rdma_cm_id *cma_id)
+{
+ struct iser_device *device;
+ struct iser_conn *ib_conn;
+ int ret;
+
+ device = iser_device_find_by_ib_device(cma_id);
+ if (!device) {
+ iser_err("device lookup/creation failed\n");
+ return iser_connect_error(cma_id);
+ }
+
+ ib_conn = (struct iser_conn *)cma_id->context;
+ ib_conn->device = device;
+
+ ret = rdma_resolve_route(cma_id, 1000);
+ if (ret) {
+ iser_err("resolve route failed: %d\n", ret);
+ return iser_connect_error(cma_id);
+ }
+
+ return 0;
+}
+
+static int iser_route_handler(struct rdma_cm_id *cma_id)
+{
+ struct rdma_conn_param conn_param;
+ int ret;
+
+ ret = iser_create_ib_conn_res((struct iser_conn *)cma_id->context);
+ if (ret)
+ goto failure;
+
+ memset(&conn_param, 0, sizeof conn_param);
+ conn_param.responder_resources = 4;
+ conn_param.initiator_depth = 1;
+ conn_param.retry_count = 7;
+ conn_param.rnr_retry_count = 6;
+
+ ret = rdma_connect(cma_id, &conn_param);
+ if (ret) {
+ iser_err("failure connecting: %d\n", ret);
+ goto failure;
+ }
+
+ return 0;
+failure:
+ return iser_connect_error(cma_id);
+}
+
+static void iser_connected_handler(struct rdma_cm_id *cma_id)
+{
+ struct iser_conn *ib_conn;
+
+ ib_conn = (struct iser_conn *)cma_id->context;
+ ib_conn->state = ISER_CONN_UP;
+ wake_up_interruptible(&ib_conn->wait);
+}
+
+static int iser_disconnected_handler(struct rdma_cm_id *cma_id)
+{
+ struct iser_conn *ib_conn;
+ int ret;
+
+ ib_conn = (struct iser_conn *)cma_id->context;
+
+ /* getting here when the state is UP means that the conn is being *
+ * terminated asynchronously from the iSCSI layer's perspective. */
+ if (iser_conn_state_comp_exch(ib_conn, ISER_CONN_UP,
+ ISER_CONN_TERMINATING))
+ iscsi_conn_failure(ib_conn->iser_conn->iscsi_conn,
+ ISCSI_ERR_CONN_FAILED);
+
+ /* Complete the termination process if no posts are pending */
+ if (ib_conn->post_recv_buf_count == 0 &&
+ (atomic_read(&ib_conn->post_send_buf_count) == 0)) {
+ ib_conn->state = ISER_CONN_DOWN;
+ wake_up_interruptible(&ib_conn->wait);
+ }
+
+ ret = iser_conn_put(ib_conn, 0); /* deref ib conn's cma id */
+ return ret;
+}
+
+static int iser_cma_handler(struct rdma_cm_id *cma_id, struct rdma_cm_event *event)
+{
+ int ret = 0;
+
+ iser_err("event %d status %d conn %p id %p\n",
+ event->event, event->status, cma_id->context, cma_id);
+
+ switch (event->event) {
+ case RDMA_CM_EVENT_ADDR_RESOLVED:
+ ret = iser_addr_handler(cma_id);
+ break;
+ case RDMA_CM_EVENT_ROUTE_RESOLVED:
+ ret = iser_route_handler(cma_id);
+ break;
+ case RDMA_CM_EVENT_ESTABLISHED:
+ iser_connected_handler(cma_id);
+ break;
+ case RDMA_CM_EVENT_ADDR_ERROR:
+ case RDMA_CM_EVENT_ROUTE_ERROR:
+ case RDMA_CM_EVENT_CONNECT_ERROR:
+ case RDMA_CM_EVENT_UNREACHABLE:
+ case RDMA_CM_EVENT_REJECTED:
+ ret = iser_connect_error(cma_id);
+ break;
+ case RDMA_CM_EVENT_DISCONNECTED:
+ case RDMA_CM_EVENT_DEVICE_REMOVAL:
+ case RDMA_CM_EVENT_ADDR_CHANGE:
+ ret = iser_disconnected_handler(cma_id);
+ break;
+ default:
+ iser_err("Unexpected RDMA CM event (%d)\n", event->event);
+ break;
+ }
+ return ret;
+}
+
+void iser_conn_init(struct iser_conn *ib_conn)
+{
+ ib_conn->state = ISER_CONN_INIT;
+ init_waitqueue_head(&ib_conn->wait);
+ ib_conn->post_recv_buf_count = 0;
+ atomic_set(&ib_conn->post_send_buf_count, 0);
+ atomic_set(&ib_conn->refcount, 1); /* ref ib conn allocation */
+ INIT_LIST_HEAD(&ib_conn->conn_list);
+ spin_lock_init(&ib_conn->lock);
+}
+
+ /**
+ * starts the process of connecting to the target
+ * sleeps until the connection is established or rejected
+ */
+int iser_connect(struct iser_conn *ib_conn,
+ struct sockaddr_in *src_addr,
+ struct sockaddr_in *dst_addr,
+ int non_blocking)
+{
+ struct sockaddr *src, *dst;
+ int err = 0;
+
+ sprintf(ib_conn->name, "%pI4:%d",
+ &dst_addr->sin_addr.s_addr, dst_addr->sin_port);
+
+ /* the device is known only --after-- address resolution */
+ ib_conn->device = NULL;
+
+ iser_err("connecting to: %pI4, port 0x%x\n",
+ &dst_addr->sin_addr, dst_addr->sin_port);
+
+ ib_conn->state = ISER_CONN_PENDING;
+
+ iser_conn_get(ib_conn); /* ref ib conn's cma id */
+ ib_conn->cma_id = rdma_create_id(iser_cma_handler,
+ (void *)ib_conn,
+ RDMA_PS_TCP, IB_QPT_RC);
+ if (IS_ERR(ib_conn->cma_id)) {
+ err = PTR_ERR(ib_conn->cma_id);
+ iser_err("rdma_create_id failed: %d\n", err);
+ goto id_failure;
+ }
+
+ src = (struct sockaddr *)src_addr;
+ dst = (struct sockaddr *)dst_addr;
+ err = rdma_resolve_addr(ib_conn->cma_id, src, dst, 1000);
+ if (err) {
+ iser_err("rdma_resolve_addr failed: %d\n", err);
+ goto addr_failure;
+ }
+
+ if (!non_blocking) {
+ wait_event_interruptible(ib_conn->wait,
+ (ib_conn->state != ISER_CONN_PENDING));
+
+ if (ib_conn->state != ISER_CONN_UP) {
+ err = -EIO;
+ goto connect_failure;
+ }
+ }
+
+ mutex_lock(&ig.connlist_mutex);
+ list_add(&ib_conn->conn_list, &ig.connlist);
+ mutex_unlock(&ig.connlist_mutex);
+ return 0;
+
+id_failure:
+ ib_conn->cma_id = NULL;
+addr_failure:
+ ib_conn->state = ISER_CONN_DOWN;
+connect_failure:
+ iser_conn_release(ib_conn, 1);
+ return err;
+}
+
+/**
+ * iser_reg_page_vec - Register physical memory
+ *
+ * returns: 0 on success, errno code on failure
+ */
+int iser_reg_page_vec(struct iser_conn *ib_conn,
+ struct iser_page_vec *page_vec,
+ struct iser_mem_reg *mem_reg)
+{
+ struct ib_pool_fmr *mem;
+ u64 io_addr;
+ u64 *page_list;
+ int status;
+
+ page_list = page_vec->pages;
+ io_addr = page_list[0];
+
+ mem = ib_fmr_pool_map_phys(ib_conn->fmr_pool,
+ page_list,
+ page_vec->length,
+ io_addr);
+
+ if (IS_ERR(mem)) {
+ status = (int)PTR_ERR(mem);
+ iser_err("ib_fmr_pool_map_phys failed: %d\n", status);
+ return status;
+ }
+
+ mem_reg->lkey = mem->fmr->lkey;
+ mem_reg->rkey = mem->fmr->rkey;
+ mem_reg->len = page_vec->length * SIZE_4K;
+ mem_reg->va = io_addr;
+ mem_reg->is_fmr = 1;
+ mem_reg->mem_h = (void *)mem;
+
+ mem_reg->va += page_vec->offset;
+ mem_reg->len = page_vec->data_size;
+
+ iser_dbg("PHYSICAL Mem.register, [PHYS p_array: 0x%p, sz: %d, "
+ "entry[0]: (0x%08lx,%ld)] -> "
+ "[lkey: 0x%08X mem_h: 0x%p va: 0x%08lX sz: %ld]\n",
+ page_vec, page_vec->length,
+ (unsigned long)page_vec->pages[0],
+ (unsigned long)page_vec->data_size,
+ (unsigned int)mem_reg->lkey, mem_reg->mem_h,
+ (unsigned long)mem_reg->va, (unsigned long)mem_reg->len);
+ return 0;
+}
+
+/**
+ * Unregister (previosuly registered) memory.
+ */
+void iser_unreg_mem(struct iser_mem_reg *reg)
+{
+ int ret;
+
+ iser_dbg("PHYSICAL Mem.Unregister mem_h %p\n",reg->mem_h);
+
+ ret = ib_fmr_pool_unmap((struct ib_pool_fmr *)reg->mem_h);
+ if (ret)
+ iser_err("ib_fmr_pool_unmap failed %d\n", ret);
+
+ reg->mem_h = NULL;
+}
+
+int iser_post_recvl(struct iser_conn *ib_conn)
+{
+ struct ib_recv_wr rx_wr, *rx_wr_failed;
+ struct ib_sge sge;
+ int ib_ret;
+
+ sge.addr = ib_conn->login_resp_dma;
+ sge.length = ISER_RX_LOGIN_SIZE;
+ sge.lkey = ib_conn->device->mr->lkey;
+
+ rx_wr.wr_id = (unsigned long)ib_conn->login_resp_buf;
+ rx_wr.sg_list = &sge;
+ rx_wr.num_sge = 1;
+ rx_wr.next = NULL;
+
+ ib_conn->post_recv_buf_count++;
+ ib_ret = ib_post_recv(ib_conn->qp, &rx_wr, &rx_wr_failed);
+ if (ib_ret) {
+ iser_err("ib_post_recv failed ret=%d\n", ib_ret);
+ ib_conn->post_recv_buf_count--;
+ }
+ return ib_ret;
+}
+
+int iser_post_recvm(struct iser_conn *ib_conn, int count)
+{
+ struct ib_recv_wr *rx_wr, *rx_wr_failed;
+ int i, ib_ret;
+ unsigned int my_rx_head = ib_conn->rx_desc_head;
+ struct iser_rx_desc *rx_desc;
+
+ for (rx_wr = ib_conn->rx_wr, i = 0; i < count; i++, rx_wr++) {
+ rx_desc = &ib_conn->rx_descs[my_rx_head];
+ rx_wr->wr_id = (unsigned long)rx_desc;
+ rx_wr->sg_list = &rx_desc->rx_sg;
+ rx_wr->num_sge = 1;
+ rx_wr->next = rx_wr + 1;
+ my_rx_head = (my_rx_head + 1) & (ISER_QP_MAX_RECV_DTOS - 1);
+ }
+
+ rx_wr--;
+ rx_wr->next = NULL; /* mark end of work requests list */
+
+ ib_conn->post_recv_buf_count += count;
+ ib_ret = ib_post_recv(ib_conn->qp, ib_conn->rx_wr, &rx_wr_failed);
+ if (ib_ret) {
+ iser_err("ib_post_recv failed ret=%d\n", ib_ret);
+ ib_conn->post_recv_buf_count -= count;
+ } else
+ ib_conn->rx_desc_head = my_rx_head;
+ return ib_ret;
+}
+
+
+/**
+ * iser_start_send - Initiate a Send DTO operation
+ *
+ * returns 0 on success, -1 on failure
+ */
+int iser_post_send(struct iser_conn *ib_conn, struct iser_tx_desc *tx_desc)
+{
+ int ib_ret;
+ struct ib_send_wr send_wr, *send_wr_failed;
+
+ ib_dma_sync_single_for_device(ib_conn->device->ib_device,
+ tx_desc->dma_addr, ISER_HEADERS_LEN, DMA_TO_DEVICE);
+
+ send_wr.next = NULL;
+ send_wr.wr_id = (unsigned long)tx_desc;
+ send_wr.sg_list = tx_desc->tx_sg;
+ send_wr.num_sge = tx_desc->num_sge;
+ send_wr.opcode = IB_WR_SEND;
+ send_wr.send_flags = IB_SEND_SIGNALED;
+
+ atomic_inc(&ib_conn->post_send_buf_count);
+
+ ib_ret = ib_post_send(ib_conn->qp, &send_wr, &send_wr_failed);
+ if (ib_ret) {
+ iser_err("ib_post_send failed, ret:%d\n", ib_ret);
+ atomic_dec(&ib_conn->post_send_buf_count);
+ }
+ return ib_ret;
+}
+
+static void iser_handle_comp_error(struct iser_tx_desc *desc,
+ struct iser_conn *ib_conn)
+{
+ if (desc && desc->type == ISCSI_TX_DATAOUT)
+ kmem_cache_free(ig.desc_cache, desc);
+
+ if (ib_conn->post_recv_buf_count == 0 &&
+ atomic_read(&ib_conn->post_send_buf_count) == 0) {
+ /* getting here when the state is UP means that the conn is *
+ * being terminated asynchronously from the iSCSI layer's *
+ * perspective. */
+ if (iser_conn_state_comp_exch(ib_conn, ISER_CONN_UP,
+ ISER_CONN_TERMINATING))
+ iscsi_conn_failure(ib_conn->iser_conn->iscsi_conn,
+ ISCSI_ERR_CONN_FAILED);
+
+ /* no more non completed posts to the QP, complete the
+ * termination process w.o worrying on disconnect event */
+ ib_conn->state = ISER_CONN_DOWN;
+ wake_up_interruptible(&ib_conn->wait);
+ }
+}
+
+static int iser_drain_tx_cq(struct iser_device *device)
+{
+ struct ib_cq *cq = device->tx_cq;
+ struct ib_wc wc;
+ struct iser_tx_desc *tx_desc;
+ struct iser_conn *ib_conn;
+ int completed_tx = 0;
+
+ while (ib_poll_cq(cq, 1, &wc) == 1) {
+ tx_desc = (struct iser_tx_desc *) (unsigned long) wc.wr_id;
+ ib_conn = wc.qp->qp_context;
+ if (wc.status == IB_WC_SUCCESS) {
+ if (wc.opcode == IB_WC_SEND)
+ iser_snd_completion(tx_desc, ib_conn);
+ else
+ iser_err("expected opcode %d got %d\n",
+ IB_WC_SEND, wc.opcode);
+ } else {
+ iser_err("tx id %llx status %d vend_err %x\n",
+ wc.wr_id, wc.status, wc.vendor_err);
+ atomic_dec(&ib_conn->post_send_buf_count);
+ iser_handle_comp_error(tx_desc, ib_conn);
+ }
+ completed_tx++;
+ }
+ return completed_tx;
+}
+
+
+static void iser_cq_tasklet_fn(unsigned long data)
+{
+ struct iser_device *device = (struct iser_device *)data;
+ struct ib_cq *cq = device->rx_cq;
+ struct ib_wc wc;
+ struct iser_rx_desc *desc;
+ unsigned long xfer_len;
+ struct iser_conn *ib_conn;
+ int completed_tx, completed_rx;
+ completed_tx = completed_rx = 0;
+
+ while (ib_poll_cq(cq, 1, &wc) == 1) {
+ desc = (struct iser_rx_desc *) (unsigned long) wc.wr_id;
+ BUG_ON(desc == NULL);
+ ib_conn = wc.qp->qp_context;
+ if (wc.status == IB_WC_SUCCESS) {
+ if (wc.opcode == IB_WC_RECV) {
+ xfer_len = (unsigned long)wc.byte_len;
+ iser_rcv_completion(desc, xfer_len, ib_conn);
+ } else
+ iser_err("expected opcode %d got %d\n",
+ IB_WC_RECV, wc.opcode);
+ } else {
+ if (wc.status != IB_WC_WR_FLUSH_ERR)
+ iser_err("rx id %llx status %d vend_err %x\n",
+ wc.wr_id, wc.status, wc.vendor_err);
+ ib_conn->post_recv_buf_count--;
+ iser_handle_comp_error(NULL, ib_conn);
+ }
+ completed_rx++;
+ if (!(completed_rx & 63))
+ completed_tx += iser_drain_tx_cq(device);
+ }
+ /* #warning "it is assumed here that arming CQ only once its empty" *
+ * " would not cause interrupts to be missed" */
+ ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
+
+ completed_tx += iser_drain_tx_cq(device);
+ iser_dbg("got %d rx %d tx completions\n", completed_rx, completed_tx);
+}
+
+static void iser_cq_callback(struct ib_cq *cq, void *cq_context)
+{
+ struct iser_device *device = (struct iser_device *)cq_context;
+
+ tasklet_schedule(&device->cq_tasklet);
+}