Merge HEAD from ../scsi-iscsi-2.6

5 files changed, 5106 insertions, 248 deletions

diff --git a/drivers/scsi/Kconfig b/drivers/scsi/Kconfig
index 2d21265e650..1c60c568e91 100644
--- a/drivers/scsi/Kconfig
+++ b/drivers/scsi/Kconfig
@@ -229,12 +229,35 @@ config SCSI_FC_ATTRS
 
 config SCSI_ISCSI_ATTRS
 	tristate "iSCSI Transport Attributes"
-	depends on SCSI
+	depends on SCSI && NET
 	help
 	  If you wish to export transport-specific information about
 	  each attached iSCSI device to sysfs, say Y.
 	  Otherwise, say N.
 
+config ISCSI_TCP
+	tristate "iSCSI Initiator over TCP/IP"
+	depends on SCSI && INET && SCSI_ISCSI_ATTRS
+	select CRYPTO
+	select CRYPTO_MD5
+	select CRYPTO_CRC32C
+	help
+	 The iSCSI Driver provides a host with the ability to access storage
+	 through an IP network. The driver uses the iSCSI protocol to transport
+	 SCSI requests and responses over a TCP/IP network between the host
+	 (the "initiator") and "targets".  Architecturally, the iSCSI driver
+	 combines with the host's TCP/IP stack, network drivers, and Network
+	 Interface Card (NIC) to provide the same functions as a SCSI or a
+	 Fibre Channel (FC) adapter driver with a Host Bus Adapter (HBA).
+
+	 To compile this driver as a module, choose M here: the
+	 module will be called iscsi_tcp.
+
+	 The userspace component needed to initialize the driver, documentation,
+	 and sample configuration files can be found here:
+
+	 http://linux-iscsi.sf.net
+
 endmenu
 
 menu "SCSI low-level drivers"
diff --git a/drivers/scsi/Makefile b/drivers/scsi/Makefile
index 4b4fd94c267..2349f902b48 100644
--- a/drivers/scsi/Makefile
+++ b/drivers/scsi/Makefile
@@ -32,6 +32,7 @@ obj-$(CONFIG_SCSI_SPI_ATTRS)	+= scsi_transport_spi.o
 obj-$(CONFIG_SCSI_FC_ATTRS) 	+= scsi_transport_fc.o
 obj-$(CONFIG_SCSI_ISCSI_ATTRS)	+= scsi_transport_iscsi.o
 
+obj-$(CONFIG_ISCSI_TCP) 	+= iscsi_tcp.o
 obj-$(CONFIG_SCSI_AMIGA7XX)	+= amiga7xx.o	53c7xx.o
 obj-$(CONFIG_A3000_SCSI)	+= a3000.o	wd33c93.o
 obj-$(CONFIG_A2091_SCSI)	+= a2091.o	wd33c93.o
diff --git a/drivers/scsi/iscsi_tcp.c b/drivers/scsi/iscsi_tcp.c
new file mode 100644
index 00000000000..810e5e59658
--- /dev/null
+++ b/drivers/scsi/iscsi_tcp.c
@@ -0,0 +1,3627 @@
+/*
+ * iSCSI Initiator over TCP/IP Data-Path
+ *
+ * Copyright (C) 2004 Dmitry Yusupov
+ * Copyright (C) 2004 Alex Aizman
+ * Copyright (C) 2005 Mike Christie
+ * maintained by open-iscsi@googlegroups.com
+ *
+ * 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.
+ *
+ * See the file COPYING included with this distribution for more details.
+ *
+ * Credits:
+ *	Christoph Hellwig
+ *	FUJITA Tomonori
+ *	Arne Redlich
+ *	Zhenyu Wang
+ */
+
+#include <linux/types.h>
+#include <linux/list.h>
+#include <linux/inet.h>
+#include <linux/blkdev.h>
+#include <linux/crypto.h>
+#include <linux/delay.h>
+#include <linux/kfifo.h>
+#include <linux/scatterlist.h>
+#include <net/tcp.h>
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_device.h>
+#include <scsi/scsi_eh.h>
+#include <scsi/scsi_request.h>
+#include <scsi/scsi_tcq.h>
+#include <scsi/scsi_host.h>
+#include <scsi/scsi.h>
+#include <scsi/scsi_transport_iscsi.h>
+
+#include "iscsi_tcp.h"
+
+MODULE_AUTHOR("Dmitry Yusupov <dmitry_yus@yahoo.com>, "
+	      "Alex Aizman <itn780@yahoo.com>");
+MODULE_DESCRIPTION("iSCSI/TCP data-path");
+MODULE_LICENSE("GPL");
+
+/* #define DEBUG_TCP */
+/* #define DEBUG_SCSI */
+#define DEBUG_ASSERT
+
+#ifdef DEBUG_TCP
+#define debug_tcp(fmt...) printk(KERN_DEBUG "tcp: " fmt)
+#else
+#define debug_tcp(fmt...)
+#endif
+
+#ifdef DEBUG_SCSI
+#define debug_scsi(fmt...) printk(KERN_DEBUG "scsi: " fmt)
+#else
+#define debug_scsi(fmt...)
+#endif
+
+#ifndef DEBUG_ASSERT
+#ifdef BUG_ON
+#undef BUG_ON
+#endif
+#define BUG_ON(expr)
+#endif
+
+#define INVALID_SN_DELTA	0xffff
+
+static unsigned int iscsi_max_lun = 512;
+module_param_named(max_lun, iscsi_max_lun, uint, S_IRUGO);
+
+/* global data */
+static kmem_cache_t *taskcache;
+
+static inline void
+iscsi_buf_init_virt(struct iscsi_buf *ibuf, char *vbuf, int size)
+{
+	sg_init_one(&ibuf->sg, (u8 *)vbuf, size);
+	ibuf->sent = 0;
+}
+
+static inline void
+iscsi_buf_init_iov(struct iscsi_buf *ibuf, char *vbuf, int size)
+{
+	ibuf->sg.page = (void*)vbuf;
+	ibuf->sg.offset = (unsigned int)-1;
+	ibuf->sg.length = size;
+	ibuf->sent = 0;
+}
+
+static inline void*
+iscsi_buf_iov_base(struct iscsi_buf *ibuf)
+{
+	return (char*)ibuf->sg.page + ibuf->sent;
+}
+
+static inline void
+iscsi_buf_init_sg(struct iscsi_buf *ibuf, struct scatterlist *sg)
+{
+	/*
+	 * Fastpath: sg element fits into single page
+	 */
+	if (sg->length + sg->offset <= PAGE_SIZE && page_count(sg->page) >= 2) {
+		ibuf->sg.page = sg->page;
+		ibuf->sg.offset = sg->offset;
+		ibuf->sg.length = sg->length;
+	} else
+		iscsi_buf_init_iov(ibuf, page_address(sg->page), sg->length);
+	ibuf->sent = 0;
+}
+
+static inline int
+iscsi_buf_left(struct iscsi_buf *ibuf)
+{
+	int rc;
+
+	rc = ibuf->sg.length - ibuf->sent;
+	BUG_ON(rc < 0);
+	return rc;
+}
+
+static inline void
+iscsi_buf_init_hdr(struct iscsi_conn *conn, struct iscsi_buf *ibuf,
+		   char *vbuf, u8 *crc)
+{
+	iscsi_buf_init_virt(ibuf, vbuf, sizeof(struct iscsi_hdr));
+	if (conn->hdrdgst_en) {
+		crypto_digest_digest(conn->tx_tfm, &ibuf->sg, 1, crc);
+		ibuf->sg.length += sizeof(uint32_t);
+	}
+}
+
+static void
+iscsi_conn_failure(struct iscsi_conn *conn, enum iscsi_err err)
+{
+	struct iscsi_session *session = conn->session;
+	unsigned long flags;
+
+	spin_lock_irqsave(&session->lock, flags);
+	if (session->conn_cnt == 1 || session->leadconn == conn)
+		session->state = ISCSI_STATE_FAILED;
+	spin_unlock_irqrestore(&session->lock, flags);
+	set_bit(SUSPEND_BIT, &conn->suspend_tx);
+	set_bit(SUSPEND_BIT, &conn->suspend_rx);
+	iscsi_conn_error(iscsi_handle(conn), err);
+}
+
+static inline int
+iscsi_check_assign_cmdsn(struct iscsi_session *session, struct iscsi_nopin *hdr)
+{
+	uint32_t max_cmdsn = be32_to_cpu(hdr->max_cmdsn);
+	uint32_t exp_cmdsn = be32_to_cpu(hdr->exp_cmdsn);
+
+	if (max_cmdsn < exp_cmdsn -1 &&
+	    max_cmdsn > exp_cmdsn - INVALID_SN_DELTA)
+		return ISCSI_ERR_MAX_CMDSN;
+	if (max_cmdsn > session->max_cmdsn ||
+	    max_cmdsn < session->max_cmdsn - INVALID_SN_DELTA)
+		session->max_cmdsn = max_cmdsn;
+	if (exp_cmdsn > session->exp_cmdsn ||
+	    exp_cmdsn < session->exp_cmdsn - INVALID_SN_DELTA)
+		session->exp_cmdsn = exp_cmdsn;
+
+	return 0;
+}
+
+static inline int
+iscsi_hdr_extract(struct iscsi_conn *conn)
+{
+	struct sk_buff *skb = conn->in.skb;
+
+	if (conn->in.copy >= conn->hdr_size &&
+	    conn->in_progress == IN_PROGRESS_WAIT_HEADER) {
+		/*
+		 * Zero-copy PDU Header: using connection context
+		 * to store header pointer.
+		 */
+		if (skb_shinfo(skb)->frag_list == NULL &&
+		    !skb_shinfo(skb)->nr_frags)
+			conn->in.hdr = (struct iscsi_hdr *)
+				((char*)skb->data + conn->in.offset);
+		else {
+			/* ignoring return code since we checked
+			 * in.copy before */
+			skb_copy_bits(skb, conn->in.offset,
+				&conn->hdr, conn->hdr_size);
+			conn->in.hdr = &conn->hdr;
+		}
+		conn->in.offset += conn->hdr_size;
+		conn->in.copy -= conn->hdr_size;
+	} else {
+		int hdr_remains;
+		int copylen;
+
+		/*
+		 * PDU header scattered across SKB's,
+		 * copying it... This'll happen quite rarely.
+		 */
+
+		if (conn->in_progress == IN_PROGRESS_WAIT_HEADER)
+			conn->in.hdr_offset = 0;
+
+		hdr_remains = conn->hdr_size - conn->in.hdr_offset;
+		BUG_ON(hdr_remains <= 0);
+
+		copylen = min(conn->in.copy, hdr_remains);
+		skb_copy_bits(skb, conn->in.offset,
+			(char*)&conn->hdr + conn->in.hdr_offset, copylen);
+
+		debug_tcp("PDU gather offset %d bytes %d in.offset %d "
+		       "in.copy %d\n", conn->in.hdr_offset, copylen,
+		       conn->in.offset, conn->in.copy);
+
+		conn->in.offset += copylen;
+		conn->in.copy -= copylen;
+		if (copylen < hdr_remains)  {
+			conn->in_progress = IN_PROGRESS_HEADER_GATHER;
+			conn->in.hdr_offset += copylen;
+		        return -EAGAIN;
+		}
+		conn->in.hdr = &conn->hdr;
+		conn->discontiguous_hdr_cnt++;
+	        conn->in_progress = IN_PROGRESS_WAIT_HEADER;
+	}
+
+	return 0;
+}
+
+static inline void
+iscsi_ctask_cleanup(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
+{
+	struct scsi_cmnd *sc = ctask->sc;
+	struct iscsi_session *session = conn->session;
+
+	spin_lock(&session->lock);
+	if (unlikely(!sc)) {
+		spin_unlock(&session->lock);
+		return;
+	}
+	if (sc->sc_data_direction == DMA_TO_DEVICE) {
+		struct iscsi_data_task *dtask, *n;
+		/* WRITE: cleanup Data-Out's if any */
+		spin_lock(&conn->lock);
+		list_for_each_entry_safe(dtask, n, &ctask->dataqueue, item) {
+			list_del(&dtask->item);
+			mempool_free(dtask, ctask->datapool);
+		}
+		spin_unlock(&conn->lock);
+	}
+	ctask->xmstate = XMSTATE_IDLE;
+	ctask->r2t = NULL;
+	ctask->sc = NULL;
+	__kfifo_put(session->cmdpool.queue, (void*)&ctask, sizeof(void*));
+	spin_unlock(&session->lock);
+}
+
+/**
+ * iscsi_cmd_rsp - SCSI Command Response processing
+ * @conn: iscsi connection
+ * @ctask: scsi command task
+ **/
+static int
+iscsi_cmd_rsp(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
+{
+	int rc;
+	struct iscsi_cmd_rsp *rhdr = (struct iscsi_cmd_rsp *)conn->in.hdr;
+	struct iscsi_session *session = conn->session;
+	struct scsi_cmnd *sc = ctask->sc;
+
+	rc = iscsi_check_assign_cmdsn(session, (struct iscsi_nopin*)rhdr);
+	if (rc) {
+		sc->result = (DID_ERROR << 16);
+		goto out;
+	}
+
+	conn->exp_statsn = be32_to_cpu(rhdr->statsn) + 1;
+
+	sc->result = (DID_OK << 16) | rhdr->cmd_status;
+
+	if (rhdr->response != ISCSI_STATUS_CMD_COMPLETED) {
+		sc->result = (DID_ERROR << 16);
+		goto out;
+	}
+
+	if (rhdr->cmd_status == SAM_STAT_CHECK_CONDITION && conn->senselen) {
+		int sensecopy = min(conn->senselen, SCSI_SENSE_BUFFERSIZE);
+
+		memcpy(sc->sense_buffer, conn->data + 2, sensecopy);
+		debug_scsi("copied %d bytes of sense\n", sensecopy);
+	}
+
+	if (sc->sc_data_direction == DMA_TO_DEVICE)
+		goto out;
+
+	if (rhdr->flags & ISCSI_FLAG_CMD_UNDERFLOW) {
+		int res_count = be32_to_cpu(rhdr->residual_count);
+
+		if (res_count > 0 && res_count <= sc->request_bufflen)
+			sc->resid = res_count;
+		else
+			sc->result = (DID_BAD_TARGET << 16) | rhdr->cmd_status;
+	} else if (rhdr->flags & ISCSI_FLAG_CMD_BIDI_UNDERFLOW)
+		sc->result = (DID_BAD_TARGET << 16) | rhdr->cmd_status;
+	else if (rhdr->flags & ISCSI_FLAG_CMD_OVERFLOW)
+		sc->resid = be32_to_cpu(rhdr->residual_count);
+
+out:
+	debug_scsi("done [sc %lx res %d itt 0x%x]\n",
+		   (long)sc, sc->result, ctask->itt);
+	conn->scsirsp_pdus_cnt++;
+	iscsi_ctask_cleanup(conn, ctask);
+	sc->scsi_done(sc);
+	return rc;
+}
+
+/**
+ * iscsi_data_rsp - SCSI Data-In Response processing
+ * @conn: iscsi connection
+ * @ctask: scsi command task
+ **/
+static int
+iscsi_data_rsp(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
+{
+	int rc;
+	struct iscsi_data_rsp *rhdr = (struct iscsi_data_rsp *)conn->in.hdr;
+	struct iscsi_session *session = conn->session;
+	int datasn = be32_to_cpu(rhdr->datasn);
+
+	rc = iscsi_check_assign_cmdsn(session, (struct iscsi_nopin*)rhdr);
+	if (rc)
+		return rc;
+	/*
+	 * setup Data-In byte counter (gets decremented..)
+	 */
+	ctask->data_count = conn->in.datalen;
+
+	if (conn->in.datalen == 0)
+		return 0;
+
+	if (ctask->datasn != datasn)
+		return ISCSI_ERR_DATASN;
+
+	ctask->datasn++;
+
+	ctask->data_offset = be32_to_cpu(rhdr->offset);
+	if (ctask->data_offset + conn->in.datalen > ctask->total_length)
+		return ISCSI_ERR_DATA_OFFSET;
+
+	if (rhdr->flags & ISCSI_FLAG_DATA_STATUS) {
+		struct scsi_cmnd *sc = ctask->sc;
+
+		conn->exp_statsn = be32_to_cpu(rhdr->statsn) + 1;
+		if (rhdr->flags & ISCSI_FLAG_CMD_UNDERFLOW) {
+			int res_count = be32_to_cpu(rhdr->residual_count);
+
+			if (res_count > 0 &&
+			    res_count <= sc->request_bufflen) {
+				sc->resid = res_count;
+				sc->result = (DID_OK << 16) | rhdr->cmd_status;
+			} else
+				sc->result = (DID_BAD_TARGET << 16) |
+					rhdr->cmd_status;
+		} else if (rhdr->flags & ISCSI_FLAG_CMD_BIDI_UNDERFLOW)
+			sc->result = (DID_BAD_TARGET << 16) | rhdr->cmd_status;
+		else if (rhdr->flags & ISCSI_FLAG_CMD_OVERFLOW) {
+			sc->resid = be32_to_cpu(rhdr->residual_count);
+			sc->result = (DID_OK << 16) | rhdr->cmd_status;
+		} else
+			sc->result = (DID_OK << 16) | rhdr->cmd_status;
+	}
+
+	conn->datain_pdus_cnt++;
+	return 0;
+}
+
+/**
+ * iscsi_solicit_data_init - initialize first Data-Out
+ * @conn: iscsi connection
+ * @ctask: scsi command task
+ * @r2t: R2T info
+ *
+ * Notes:
+ *	Initialize first Data-Out within this R2T sequence and finds
+ *	proper data_offset within this SCSI command.
+ *
+ *	This function is called with connection lock taken.
+ **/
+static void
+iscsi_solicit_data_init(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask,
+			struct iscsi_r2t_info *r2t)
+{
+	struct iscsi_data *hdr;
+	struct iscsi_data_task *dtask;
+	struct scsi_cmnd *sc = ctask->sc;
+
+	dtask = mempool_alloc(ctask->datapool, GFP_ATOMIC);
+	BUG_ON(!dtask);
+	hdr = &dtask->hdr;
+	memset(hdr, 0, sizeof(struct iscsi_data));
+	hdr->ttt = r2t->ttt;
+	hdr->datasn = cpu_to_be32(r2t->solicit_datasn);
+	r2t->solicit_datasn++;
+	hdr->opcode = ISCSI_OP_SCSI_DATA_OUT;
+	memcpy(hdr->lun, ctask->hdr.lun, sizeof(hdr->lun));
+	hdr->itt = ctask->hdr.itt;
+	hdr->exp_statsn = r2t->exp_statsn;
+	hdr->offset = cpu_to_be32(r2t->data_offset);
+	if (r2t->data_length > conn->max_xmit_dlength) {
+		hton24(hdr->dlength, conn->max_xmit_dlength);
+		r2t->data_count = conn->max_xmit_dlength;
+		hdr->flags = 0;
+	} else {
+		hton24(hdr->dlength, r2t->data_length);
+		r2t->data_count = r2t->data_length;
+		hdr->flags = ISCSI_FLAG_CMD_FINAL;
+	}
+	conn->dataout_pdus_cnt++;
+
+	r2t->sent = 0;
+
+	iscsi_buf_init_hdr(conn, &r2t->headbuf, (char*)hdr,
+			   (u8 *)dtask->hdrext);
+
+	r2t->dtask = dtask;
+
+	if (sc->use_sg) {
+		int i, sg_count = 0;
+		struct scatterlist *sg = sc->request_buffer;
+
+		r2t->sg = NULL;
+		for (i = 0; i < sc->use_sg; i++, sg += 1) {
+			/* FIXME: prefetch ? */
+			if (sg_count + sg->length > r2t->data_offset) {
+				int page_offset;
+
+				/* sg page found! */
+
+				/* offset within this page */
+				page_offset = r2t->data_offset - sg_count;
+
+				/* fill in this buffer */
+				iscsi_buf_init_sg(&r2t->sendbuf, sg);
+				r2t->sendbuf.sg.offset += page_offset;
+				r2t->sendbuf.sg.length -= page_offset;
+
+				/* xmit logic will continue with next one */
+				r2t->sg = sg + 1;
+				break;
+			}
+			sg_count += sg->length;
+		}
+		BUG_ON(r2t->sg == NULL);
+	} else
+		iscsi_buf_init_iov(&ctask->sendbuf,
+			    (char*)sc->request_buffer + r2t->data_offset,
+			    r2t->data_count);
+
+	list_add(&dtask->item, &ctask->dataqueue);
+}
+
+/**
+ * iscsi_r2t_rsp - iSCSI R2T Response processing
+ * @conn: iscsi connection
+ * @ctask: scsi command task
+ **/
+static int
+iscsi_r2t_rsp(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
+{
+	struct iscsi_r2t_info *r2t;
+	struct iscsi_session *session = conn->session;
+	struct iscsi_r2t_rsp *rhdr = (struct iscsi_r2t_rsp *)conn->in.hdr;
+	int r2tsn = be32_to_cpu(rhdr->r2tsn);
+	int rc;
+
+	if (conn->in.ahslen)
+		return ISCSI_ERR_AHSLEN;
+
+	if (conn->in.datalen)
+		return ISCSI_ERR_DATALEN;
+
+	if (ctask->exp_r2tsn && ctask->exp_r2tsn != r2tsn)
+		return ISCSI_ERR_R2TSN;
+
+	rc = iscsi_check_assign_cmdsn(session, (struct iscsi_nopin*)rhdr);
+	if (rc)
+		return rc;
+
+	/* FIXME: use R2TSN to detect missing R2T */
+
+	/* fill-in new R2T associated with the task */
+	spin_lock(&session->lock);
+	if (!ctask->sc || ctask->mtask ||
+	     session->state != ISCSI_STATE_LOGGED_IN) {
+		printk(KERN_INFO "iscsi_tcp: dropping R2T itt %d in "
+		       "recovery...\n", ctask->itt);
+		spin_unlock(&session->lock);
+		return 0;
+	}
+	rc = __kfifo_get(ctask->r2tpool.queue, (void*)&r2t, sizeof(void*));
+	BUG_ON(!rc);
+
+	r2t->exp_statsn = rhdr->statsn;
+	r2t->data_length = be32_to_cpu(rhdr->data_length);
+	if (r2t->data_length == 0 ||
+	    r2t->data_length > session->max_burst) {
+		spin_unlock(&session->lock);
+		return ISCSI_ERR_DATALEN;
+	}
+
+	r2t->data_offset = be32_to_cpu(rhdr->data_offset);
+	if (r2t->data_offset + r2t->data_length > ctask->total_length) {
+		spin_unlock(&session->lock);
+		return ISCSI_ERR_DATALEN;
+	}
+
+	r2t->ttt = rhdr->ttt; /* no flip */
+	r2t->solicit_datasn = 0;
+
+	iscsi_solicit_data_init(conn, ctask, r2t);
+
+	ctask->exp_r2tsn = r2tsn + 1;
+	ctask->xmstate |= XMSTATE_SOL_HDR;
+	__kfifo_put(ctask->r2tqueue, (void*)&r2t, sizeof(void*));
+	__kfifo_put(conn->writequeue, (void*)&ctask, sizeof(void*));
+
+	schedule_work(&conn->xmitwork);
+	conn->r2t_pdus_cnt++;
+	spin_unlock(&session->lock);
+
+	return 0;
+}
+
+static int
+iscsi_hdr_recv(struct iscsi_conn *conn)
+{
+	int rc = 0;
+	struct iscsi_hdr *hdr;
+	struct iscsi_cmd_task *ctask;
+	struct iscsi_session *session = conn->session;
+	uint32_t cdgst, rdgst = 0;
+
+	hdr = conn->in.hdr;
+
+	/* verify PDU length */
+	conn->in.datalen = ntoh24(hdr->dlength);
+	if (conn->in.datalen > conn->max_recv_dlength) {
+		printk(KERN_ERR "iscsi_tcp: datalen %d > %d\n",
+		       conn->in.datalen, conn->max_recv_dlength);
+		return ISCSI_ERR_DATALEN;
+	}
+	conn->data_copied = 0;
+
+	/* read AHS */
+	conn->in.ahslen = hdr->hlength*(4*sizeof(__u16));
+	conn->in.offset += conn->in.ahslen;
+	conn->in.copy -= conn->in.ahslen;
+	if (conn->in.copy < 0) {
+		printk(KERN_ERR "iscsi_tcp: can't handle AHS with length "
+		       "%d bytes\n", conn->in.ahslen);
+		return ISCSI_ERR_AHSLEN;
+	}
+
+	/* calculate read padding */
+	conn->in.padding = conn->in.datalen & (ISCSI_PAD_LEN-1);
+	if (conn->in.padding) {
+		conn->in.padding = ISCSI_PAD_LEN - conn->in.padding;
+		debug_scsi("read padding %d bytes\n", conn->in.padding);
+	}
+
+	if (conn->hdrdgst_en) {
+		struct scatterlist sg;
+
+		sg_init_one(&sg, (u8 *)hdr,
+			    sizeof(struct iscsi_hdr) + conn->in.ahslen);
+		crypto_digest_digest(conn->rx_tfm, &sg, 1, (u8 *)&cdgst);
+		rdgst = *(uint32_t*)((char*)hdr + sizeof(struct iscsi_hdr) +
+				     conn->in.ahslen);
+	}
+
+	/* save opcode for later */
+	conn->in.opcode = hdr->opcode;
+
+	/* verify itt (itt encoding: age+cid+itt) */
+	if (hdr->itt != cpu_to_be32(ISCSI_RESERVED_TAG)) {
+		if ((hdr->itt & AGE_MASK) !=
+				(session->age << AGE_SHIFT)) {
+			printk(KERN_ERR "iscsi_tcp: received itt %x expected "
+				"session age (%x)\n", hdr->itt,
+				session->age & AGE_MASK);
+			return ISCSI_ERR_BAD_ITT;
+		}
+
+		if ((hdr->itt & CID_MASK) != (conn->id << CID_SHIFT)) {
+			printk(KERN_ERR "iscsi_tcp: received itt %x, expected "
+				"CID (%x)\n", hdr->itt, conn->id);
+			return ISCSI_ERR_BAD_ITT;
+		}
+		conn->in.itt = hdr->itt & ITT_MASK;
+	} else
+		conn->in.itt = hdr->itt;
+
+	debug_tcp("opcode 0x%x offset %d copy %d ahslen %d datalen %d\n",
+		  hdr->opcode, conn->in.offset, conn->in.copy,
+		  conn->in.ahslen, conn->in.datalen);
+
+	if (conn->in.itt < session->cmds_max) {
+		if (conn->hdrdgst_en && cdgst != rdgst) {
+			printk(KERN_ERR "iscsi_tcp: itt %x: hdrdgst error "
+			       "recv 0x%x calc 0x%x\n", conn->in.itt, rdgst,
+			       cdgst);
+			return ISCSI_ERR_HDR_DGST;
+		}
+
+		ctask = (struct iscsi_cmd_task *)session->cmds[conn->in.itt];
+
+		if (!ctask->sc) {
+			printk(KERN_INFO "iscsi_tcp: dropping ctask with "
+			       "itt 0x%x\n", ctask->itt);
+			conn->in.datalen = 0; /* force drop */
+			return 0;
+		}
+
+		if (ctask->sc->SCp.phase != session->age) {
+			printk(KERN_ERR "iscsi_tcp: ctask's session age %d, "
+				"expected %d\n", ctask->sc->SCp.phase,
+				session->age);
+			return ISCSI_ERR_SESSION_FAILED;
+		}
+
+		conn->in.ctask = ctask;
+
+		debug_scsi("rsp [op 0x%x cid %d sc %lx itt 0x%x len %d]\n",
+			   hdr->opcode, conn->id, (long)ctask->sc,
+			   ctask->itt, conn->in.datalen);
+
+		switch(conn->in.opcode) {
+		case ISCSI_OP_SCSI_CMD_RSP:
+			BUG_ON((void*)ctask != ctask->sc->SCp.ptr);
+			if (ctask->hdr.flags & ISCSI_FLAG_CMD_WRITE)
+				rc = iscsi_cmd_rsp(conn, ctask);
+			else if (!conn->in.datalen)
+				rc = iscsi_cmd_rsp(conn, ctask);
+			else
+				/*
+				 * got sense or response data; copying PDU
+				 * Header to the connection's header
+				 * placeholder
+				 */
+				memcpy(&conn->hdr, hdr,
+				       sizeof(struct iscsi_hdr));
+			break;
+		case ISCSI_OP_SCSI_DATA_IN:
+			BUG_ON((void*)ctask != ctask->sc->SCp.ptr);
+			/* save flags for non-exceptional status */
+			conn->in.flags = hdr->flags;
+			/* save cmd_status for sense data */
+			conn->in.cmd_status =
+				((struct iscsi_data_rsp*)hdr)->cmd_status;
+			rc = iscsi_data_rsp(conn, ctask);
+			break;
+		case ISCSI_OP_R2T:
+			BUG_ON((void*)ctask != ctask->sc->SCp.ptr);
+			if (ctask->hdr.flags & ISCSI_FLAG_CMD_WRITE &&
+			    ctask->sc->sc_data_direction == DMA_TO_DEVICE)
+				rc = iscsi_r2t_rsp(conn, ctask);
+			else
+				rc = ISCSI_ERR_PROTO;
+			break;
+		case ISCSI_OP_NOOP_IN:
+		case ISCSI_OP_TEXT_RSP:
+		case ISCSI_OP_LOGOUT_RSP:
+		case ISCSI_OP_ASYNC_EVENT:
+		case ISCSI_OP_REJECT:
+			rc = iscsi_check_assign_cmdsn(session,
+						 (struct iscsi_nopin*)hdr);
+			if (rc)
+				break;
+
+			/* update ExpStatSN */
+			conn->exp_statsn = be32_to_cpu(hdr->statsn) + 1;
+			if (!conn->in.datalen) {
+				struct iscsi_mgmt_task *mtask;
+
+				rc = iscsi_recv_pdu(iscsi_handle(conn), hdr,
+						    NULL, 0);
+				mtask = (struct iscsi_mgmt_task *)
+					session->mgmt_cmds[conn->in.itt -
+						ISCSI_MGMT_ITT_OFFSET];
+				if (conn->login_mtask != mtask) {
+					spin_lock(&session->lock);
+					__kfifo_put(session->mgmtpool.queue,
+					    (void*)&mtask, sizeof(void*));
+					spin_unlock(&session->lock);
+				}
+			}
+			break;
+		default:
+			rc = ISCSI_ERR_BAD_OPCODE;
+			break;
+		}
+	} else if (conn->in.itt >= ISCSI_MGMT_ITT_OFFSET &&
+		   conn->in.itt < ISCSI_MGMT_ITT_OFFSET +
+					session->mgmtpool_max) {
+		struct iscsi_mgmt_task *mtask = (struct iscsi_mgmt_task *)
+					session->mgmt_cmds[conn->in.itt -
+						ISCSI_MGMT_ITT_OFFSET];
+
+		debug_scsi("immrsp [op 0x%x cid %d itt 0x%x len %d]\n",
+			   conn->in.opcode, conn->id, mtask->itt,
+			   conn->in.datalen);
+
+		switch(conn->in.opcode) {
+		case ISCSI_OP_LOGIN_RSP:
+		case ISCSI_OP_TEXT_RSP:
+			rc = iscsi_check_assign_cmdsn(session,
+						 (struct iscsi_nopin*)hdr);
+			if (rc)
+				break;
+
+			if (!conn->in.datalen) {
+				rc = iscsi_recv_pdu(iscsi_handle(conn), hdr,
+						    NULL, 0);
+				if (conn->login_mtask != mtask) {
+					spin_lock(&session->lock);
+					__kfifo_put(session->mgmtpool.queue,
+					    (void*)&mtask, sizeof(void*));
+					spin_unlock(&session->lock);
+				}
+			}
+			break;
+		case ISCSI_OP_SCSI_TMFUNC_RSP:
+			rc = iscsi_check_assign_cmdsn(session,
+						 (struct iscsi_nopin*)hdr);
+			if (rc)
+				break;
+
+			if (conn->in.datalen || conn->in.ahslen) {
+				rc = ISCSI_ERR_PROTO;
+				break;
+			}
+			conn->tmfrsp_pdus_cnt++;
+			spin_lock(&session->lock);
+			if (conn->tmabort_state == TMABORT_INITIAL) {
+				__kfifo_put(session->mgmtpool.queue,
+						(void*)&mtask, sizeof(void*));
+				conn->tmabort_state =
+					((struct iscsi_tm_rsp *)hdr)->
+					response == SCSI_TCP_TM_RESP_COMPLETE ?
+						TMABORT_SUCCESS:TMABORT_FAILED;
+				/* unblock eh_abort() */
+				wake_up(&conn->ehwait);
+			}
+			spin_unlock(&session->lock);
+			break;
+		default:
+			rc = ISCSI_ERR_BAD_OPCODE;
+			break;
+		}
+	} else if (conn->in.itt == ISCSI_RESERVED_TAG) {
+		if (conn->in.opcode == ISCSI_OP_NOOP_IN && !conn->in.datalen) {
+			rc = iscsi_check_assign_cmdsn(session,
+						 (struct iscsi_nopin*)hdr);
+			if (!rc)
+				rc = iscsi_recv_pdu(iscsi_handle(conn),
+						    hdr, NULL, 0);
+		}
+		else
+			rc = ISCSI_ERR_BAD_OPCODE;
+	} else
+		rc = ISCSI_ERR_BAD_ITT;
+
+	return rc;
+}
+
+/**
+ * iscsi_ctask_copy - copy skb bits to the destanation cmd task
+ * @conn: iscsi connection
+ * @ctask: scsi command task
+ * @buf: buffer to copy to
+ * @buf_size: size of buffer
+ * @offset: offset within the buffer
+ *
+ * Notes:
+ *	The function calls skb_copy_bits() and updates per-connection and
+ *	per-cmd byte counters.
+ *
+ *	Read counters (in bytes):
+ *
+ *	conn->in.offset		offset within in progress SKB
+ *	conn->in.copy		left to copy from in progress SKB
+ *				including padding
+ *	conn->in.copied		copied already from in progress SKB
+ *	conn->data_copied	copied already from in progress buffer
+ *	ctask->sent		total bytes sent up to the MidLayer
+ *	ctask->data_count	left to copy from in progress Data-In
+ *	buf_left		left to copy from in progress buffer
+ **/
+static inline int
+iscsi_ctask_copy(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask,
+		void *buf, int buf_size, int offset)
+{
+	int buf_left = buf_size - (conn->data_copied + offset);
+	int size = min(conn->in.copy, buf_left);
+	int rc;
+
+	size = min(size, ctask->data_count);
+
+	debug_tcp("ctask_copy %d bytes at offset %d copied %d\n",
+	       size, conn->in.offset, conn->in.copied);
+
+	BUG_ON(size <= 0);
+	BUG_ON(ctask->sent + size > ctask->total_length);
+
+	rc = skb_copy_bits(conn->in.skb, conn->in.offset,
+			   (char*)buf + (offset + conn->data_copied), size);
+	/* must fit into skb->len */
+	BUG_ON(rc);
+
+	conn->in.offset += size;
+	conn->in.copy -= size;
+	conn->in.copied += size;
+	conn->data_copied += size;
+	ctask->sent += size;
+	ctask->data_count -= size;
+
+	BUG_ON(conn->in.copy < 0);
+	BUG_ON(ctask->data_count < 0);
+
+	if (buf_size != (conn->data_copied + offset)) {
+		if (!ctask->data_count) {
+			BUG_ON(buf_size - conn->data_copied < 0);
+			/* done with this PDU */
+			return buf_size - conn->data_copied;
+		}
+		return -EAGAIN;
+	}
+
+	/* done with this buffer or with both - PDU and buffer */
+	conn->data_copied = 0;
+	return 0;
+}
+
+/**
+ * iscsi_tcp_copy - copy skb bits to the destanation buffer
+ * @conn: iscsi connection
+ * @buf: buffer to copy to
+ * @buf_size: number of bytes to copy
+ *
+ * Notes:
+ *	The function calls skb_copy_bits() and updates per-connection
+ *	byte counters.
+ **/
+static inline int
+iscsi_tcp_copy(struct iscsi_conn *conn, void *buf, int buf_size)
+{
+	int buf_left = buf_size - conn->data_copied;
+	int size = min(conn->in.copy, buf_left);
+	int rc;
+
+	debug_tcp("tcp_copy %d bytes at offset %d copied %d\n",
+	       size, conn->in.offset, conn->data_copied);
+	BUG_ON(size <= 0);
+
+	rc = skb_copy_bits(conn->in.skb, conn->in.offset,
+			   (char*)buf + conn->data_copied, size);
+	BUG_ON(rc);
+
+	conn->in.offset += size;
+	conn->in.copy -= size;
+	conn->in.copied += size;
+	conn->data_copied += size;
+
+	if (buf_size != conn->data_copied)
+		return -EAGAIN;
+
+	return 0;
+}
+
+static inline void
+partial_sg_digest_update(struct iscsi_conn *conn, struct scatterlist *sg,
+			 int offset, int length)
+{
+	struct scatterlist temp;
+
+	memcpy(&temp, sg, sizeof(struct scatterlist));
+	temp.offset = offset;
+	temp.length = length;
+	crypto_digest_update(conn->data_rx_tfm, &temp, 1);
+}
+
+static int iscsi_scsi_data_in(struct iscsi_conn *conn)
+{
+	struct iscsi_cmd_task *ctask = conn->in.ctask;
+	struct scsi_cmnd *sc = ctask->sc;
+	struct scatterlist tmp, *sg;
+	int i, offset, rc = 0;
+
+	BUG_ON((void*)ctask != sc->SCp.ptr);
+
+	/*
+	 * copying Data-In into the Scsi_Cmnd
+	 */
+	if (!sc->use_sg) {
+		i = ctask->data_count;
+		rc = iscsi_ctask_copy(conn, ctask, sc->request_buffer,
+				      sc->request_bufflen, ctask->data_offset);
+		if (rc == -EAGAIN)
+			return rc;
+		if (conn->datadgst_en) {
+			sg_init_one(&tmp, sc->request_buffer, i);
+			crypto_digest_update(conn->data_rx_tfm, &tmp, 1);
+		}
+		rc = 0;
+		goto done;
+	}
+
+	offset = ctask->data_offset;
+	sg = sc->request_buffer;
+
+	if (ctask->data_offset)
+		for (i = 0; i < ctask->sg_count; i++)
+			offset -= sg[i].length;
+	/* we've passed through partial sg*/
+	if (offset < 0)
+		offset = 0;
+
+	for (i = ctask->sg_count; i < sc->use_sg; i++) {
+		char *dest;
+
+		dest = kmap_atomic(sg[i].page, KM_SOFTIRQ0);
+		rc = iscsi_ctask_copy(conn, ctask, dest + sg[i].offset,
+				      sg[i].length, offset);
+		kunmap_atomic(dest, KM_SOFTIRQ0);
+		if (rc == -EAGAIN)
+			/* continue with the next SKB/PDU */
+			return rc;
+		if (!rc) {
+			if (conn->datadgst_en) {
+				if (!offset)
+					crypto_digest_update(conn->data_rx_tfm,
+							     &sg[i], 1);
+				else
+					partial_sg_digest_update(conn, &sg[i],
+							sg[i].offset + offset,
+							sg[i].length - offset);
+			}
+			offset = 0;
+			ctask->sg_count++;
+		}
+
+		if (!ctask->data_count) {
+			if (rc && conn->datadgst_en)
+				/*
+				 * data-in is complete, but buffer not...
+				 */
+				partial_sg_digest_update(conn, &sg[i],
+						sg[i].offset, sg[i].length-rc);
+			rc = 0;
+			break;
+		}
+
+		if (!conn->in.copy)
+			return -EAGAIN;
+	}
+	BUG_ON(ctask->data_count);
+
+done:
+	/* check for non-exceptional status */
+	if (conn->in.flags & ISCSI_FLAG_DATA_STATUS) {
+		debug_scsi("done [sc %lx res %d itt 0x%x]\n",
+			   (long)sc, sc->result, ctask->itt);
+		conn->scsirsp_pdus_cnt++;
+		iscsi_ctask_cleanup(conn, ctask);
+		sc->scsi_done(sc);
+	}
+
+	return rc;
+}
+
+static int
+iscsi_data_recv(struct iscsi_conn *conn)
+{
+	struct iscsi_session *session = conn->session;
+	int rc = 0;
+
+	switch(conn->in.opcode) {
+	case ISCSI_OP_SCSI_DATA_IN:
+		rc = iscsi_scsi_data_in(conn);
+		break;
+	case ISCSI_OP_SCSI_CMD_RSP: {
+		/*
+		 * SCSI Sense Data:
+		 * copying the entire Data Segment.
+		 */
+		if (iscsi_tcp_copy(conn, conn->data, conn->in.datalen)) {
+			rc = -EAGAIN;
+			goto exit;
+		}
+
+		/*
+		 * check for sense
+		 */
+		conn->in.hdr = &conn->hdr;
+		conn->senselen = (conn->data[0] << 8) | conn->data[1];
+		rc = iscsi_cmd_rsp(conn, conn->in.ctask);
+	}
+	break;
+	case ISCSI_OP_TEXT_RSP:
+	case ISCSI_OP_LOGIN_RSP:
+	case ISCSI_OP_NOOP_IN: {
+		struct iscsi_mgmt_task *mtask = NULL;
+
+		if (conn->in.itt != ISCSI_RESERVED_TAG)
+			mtask = (struct iscsi_mgmt_task *)
+				session->mgmt_cmds[conn->in.itt -
+					ISCSI_MGMT_ITT_OFFSET];
+
+		/*
+		 * Collect data segment to the connection's data
+		 * placeholder
+		 */
+		if (iscsi_tcp_copy(conn, conn->data, conn->in.datalen)) {
+			rc = -EAGAIN;
+			goto exit;
+		}
+
+		rc = iscsi_recv_pdu(iscsi_handle(conn), conn->in.hdr,
+				    conn->data, conn->in.datalen);
+
+		if (mtask && conn->login_mtask != mtask) {
+			spin_lock(&session->lock);
+			__kfifo_put(session->mgmtpool.queue, (void*)&mtask,
+				    sizeof(void*));
+			spin_unlock(&session->lock);
+		}
+	}
+	break;
+	default:
+		BUG_ON(1);
+	}
+exit:
+	return rc;
+}
+
+/**
+ * iscsi_tcp_data_recv - TCP receive in sendfile fashion
+ * @rd_desc: read descriptor
+ * @skb: socket buffer
+ * @offset: offset in skb
+ * @len: skb->len - offset
+ **/
+static int
+iscsi_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb,
+		unsigned int offset, size_t len)
+{
+	int rc;
+	struct iscsi_conn *conn = rd_desc->arg.data;
+	int start = skb_headlen(skb);
+	int processed;
+	char pad[ISCSI_PAD_LEN];
+	struct scatterlist sg;
+
+	/*
+	 * Save current SKB and its offset in the corresponding
+	 * connection context.
+	 */
+	conn->in.copy = start - offset;
+	conn->in.offset = offset;
+	conn->in.skb = skb;
+	conn->in.len = conn->in.copy;
+	BUG_ON(conn->in.copy <= 0);
+	debug_tcp("in %d bytes\n", conn->in.copy);
+
+more:
+	conn->in.copied = 0;
+	rc = 0;
+
+	if (unlikely(conn->suspend_rx)) {
+		debug_tcp("conn %d Rx suspended!\n", conn->id);
+		return 0;
+	}
+
+	if (conn->in_progress == IN_PROGRESS_WAIT_HEADER ||
+	    conn->in_progress == IN_PROGRESS_HEADER_GATHER) {
+		rc = iscsi_hdr_extract(conn);
+		if (rc) {
+		       if (rc == -EAGAIN)
+				goto nomore;
+		       else {
+				iscsi_conn_failure(conn, rc);
+				return 0;
+		       }
+		}
+
+		/*
+		 * Verify and process incoming PDU header.
+		 */
+		rc = iscsi_hdr_recv(conn);
+		if (!rc && conn->in.datalen) {
+			if (conn->datadgst_en &&
+			    conn->in.opcode == ISCSI_OP_SCSI_DATA_IN) {
+				BUG_ON(!conn->data_rx_tfm);
+				crypto_digest_init(conn->data_rx_tfm);
+			}
+			conn->in_progress = IN_PROGRESS_DATA_RECV;
+		} else if (rc) {
+			iscsi_conn_failure(conn, rc);
+			return 0;
+		}
+	}
+
+	if (conn->in_progress == IN_PROGRESS_DDIGEST_RECV) {
+		debug_tcp("extra data_recv offset %d copy %d\n",
+			  conn->in.offset, conn->in.copy);
+		if (conn->in.opcode == ISCSI_OP_SCSI_DATA_IN) {
+			uint32_t recv_digest;
+			skb_copy_bits(conn->in.skb, conn->in.offset,
+				      &recv_digest, 4);
+			conn->in.offset += 4;
+			conn->in.copy -= 4;
+			if (recv_digest != conn->in.datadgst) {
+				debug_tcp("iscsi_tcp: data digest error!"
+					  "0x%x != 0x%x\n", recv_digest,
+					  conn->in.datadgst);
+				iscsi_conn_failure(conn, ISCSI_ERR_DATA_DGST);
+				return 0;
+			} else {
+				debug_tcp("iscsi_tcp: data digest match!"
+					  "0x%x == 0x%x\n", recv_digest,
+					  conn->in.datadgst);
+				conn->in_progress = IN_PROGRESS_WAIT_HEADER;
+			}
+		}
+	}
+
+	if (conn->in_progress == IN_PROGRESS_DATA_RECV && conn->in.copy) {
+
+		debug_tcp("data_recv offset %d copy %d\n",
+		       conn->in.offset, conn->in.copy);
+
+		rc = iscsi_data_recv(conn);
+		if (rc) {
+			if (rc == -EAGAIN) {
+				rd_desc->count = conn->in.datalen -
+						conn->in.ctask->data_count;
+				goto again;
+			}
+			iscsi_conn_failure(conn, rc);
+			return 0;
+		}
+		conn->in.copy -= conn->in.padding;
+		conn->in.offset += conn->in.padding;
+		if (conn->datadgst_en &&
+		    conn->in.opcode == ISCSI_OP_SCSI_DATA_IN) {
+			if (conn->in.padding) {
+				debug_tcp("padding -> %d\n", conn->in.padding);
+				memset(pad, 0, conn->in.padding);
+				sg_init_one(&sg, pad, conn->in.padding);
+				crypto_digest_update(conn->data_rx_tfm, &sg, 1);
+			}
+			crypto_digest_final(conn->data_rx_tfm,
+					    (u8 *) & conn->in.datadgst);
+			debug_tcp("rx digest 0x%x\n", conn->in.datadgst);
+			conn->in_progress = IN_PROGRESS_DDIGEST_RECV;
+		} else
+			conn->in_progress = IN_PROGRESS_WAIT_HEADER;
+	}
+
+	debug_tcp("f, processed %d from out of %d padding %d\n",
+	       conn->in.offset - offset, (int)len, conn->in.padding);
+	BUG_ON(conn->in.offset - offset > len);
+
+	if (conn->in.offset - offset != len) {
+		debug_tcp("continue to process %d bytes\n",
+		       (int)len - (conn->in.offset - offset));
+		goto more;
+	}
+
+nomore:
+	processed = conn->in.offset - offset;
+	BUG_ON(processed == 0);
+	return processed;
+
+again:
+	processed = conn->in.offset - offset;
+	debug_tcp("c, processed %d from out of %d rd_desc_cnt %d\n",
+	          processed, (int)len, (int)rd_desc->count);
+	BUG_ON(processed == 0);
+	BUG_ON(processed > len);
+
+	conn->rxdata_octets += processed;
+	return processed;
+}
+
+static void
+iscsi_tcp_data_ready(struct sock *sk, int flag)
+{
+	struct iscsi_conn *conn = sk->sk_user_data;
+	read_descriptor_t rd_desc;
+
+	read_lock(&sk->sk_callback_lock);
+
+	/* use rd_desc to pass 'conn' to iscsi_tcp_data_recv */
+	rd_desc.arg.data = conn;
+	rd_desc.count = 0;
+	tcp_read_sock(sk, &rd_desc, iscsi_tcp_data_recv);
+
+	read_unlock(&sk->sk_callback_lock);
+}
+
+static void
+iscsi_tcp_state_change(struct sock *sk)
+{
+	struct iscsi_conn *conn;
+	struct iscsi_session *session;
+	void (*old_state_change)(struct sock *);
+
+	read_lock(&sk->sk_callback_lock);
+
+	conn = (struct iscsi_conn*)sk->sk_user_data;
+	session = conn->session;
+
+	if (sk->sk_state == TCP_CLOSE_WAIT ||
+	    sk->sk_state == TCP_CLOSE) {
+		debug_tcp("iscsi_tcp_state_change: TCP_CLOSE|TCP_CLOSE_WAIT\n");
+		iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
+	}
+
+	old_state_change = conn->old_state_change;
+
+	read_unlock(&sk->sk_callback_lock);
+
+	old_state_change(sk);
+}
+
+/**
+ * iscsi_write_space - Called when more output buffer space is available
+ * @sk: socket space is available for
+ **/
+static void
+iscsi_write_space(struct sock *sk)
+{
+	struct iscsi_conn *conn = (struct iscsi_conn*)sk->sk_user_data;
+	conn->old_write_space(sk);
+	debug_tcp("iscsi_write_space: cid %d\n", conn->id);
+	clear_bit(SUSPEND_BIT, &conn->suspend_tx);
+	schedule_work(&conn->xmitwork);
+}
+
+static void
+iscsi_conn_set_callbacks(struct iscsi_conn *conn)
+{
+	struct sock *sk = conn->sock->sk;
+
+	/* assign new callbacks */
+	write_lock_bh(&sk->sk_callback_lock);
+	sk->sk_user_data = conn;
+	conn->old_data_ready = sk->sk_data_ready;
+	conn->old_state_change = sk->sk_state_change;
+	conn->old_write_space = sk->sk_write_space;
+	sk->sk_data_ready = iscsi_tcp_data_ready;
+	sk->sk_state_change = iscsi_tcp_state_change;
+	sk->sk_write_space = iscsi_write_space;
+	write_unlock_bh(&sk->sk_callback_lock);
+}
+
+static void
+iscsi_conn_restore_callbacks(struct iscsi_conn *conn)
+{
+	struct sock *sk = conn->sock->sk;
+
+	/* restore socket callbacks, see also: iscsi_conn_set_callbacks() */
+	write_lock_bh(&sk->sk_callback_lock);
+	sk->sk_user_data    = NULL;
+	sk->sk_data_ready   = conn->old_data_ready;
+	sk->sk_state_change = conn->old_state_change;
+	sk->sk_write_space  = conn->old_write_space;
+	sk->sk_no_check	 = 0;
+	write_unlock_bh(&sk->sk_callback_lock);
+}
+
+/**
+ * iscsi_send - generic send routine
+ * @sk: kernel's socket
+ * @buf: buffer to write from
+ * @size: actual size to write
+ * @flags: socket's flags
+ *
+ * Notes:
+ *	depending on buffer will use tcp_sendpage() or tcp_sendmsg().
+ *	buf->sg.offset == -1 tells us that buffer is non S/G and forces
+ *	to use tcp_sendmsg().
+ */
+static inline int
+iscsi_send(struct socket *sk, struct iscsi_buf *buf, int size, int flags)
+{
+	int res;
+
+	if ((int)buf->sg.offset >= 0) {
+		int offset = buf->sg.offset + buf->sent;
+
+		/* tcp_sendpage */
+		res = sk->ops->sendpage(sk, buf->sg.page, offset, size, flags);
+	} else {
+		struct msghdr msg;
+
+		buf->iov.iov_base = iscsi_buf_iov_base(buf);
+		buf->iov.iov_len = size;
+
+		memset(&msg, 0, sizeof(struct msghdr));
+
+		/* tcp_sendmsg */
+		res = kernel_sendmsg(sk, &msg, &buf->iov, 1, size);
+	}
+
+	return res;
+}
+
+/**
+ * iscsi_sendhdr - send PDU Header via tcp_sendpage()
+ * @conn: iscsi connection
+ * @buf: buffer to write from
+ * @datalen: lenght of data to be sent after the header
+ *
+ * Notes:
+ *	(Tx, Fast Path)
+ **/
+static inline int
+iscsi_sendhdr(struct iscsi_conn *conn, struct iscsi_buf *buf, int datalen)
+{
+	struct socket *sk = conn->sock;
+	int flags = 0; /* MSG_DONTWAIT; */
+	int res, size;
+
+	size = buf->sg.length - buf->sent;
+	BUG_ON(buf->sent + size > buf->sg.length);
+	if (buf->sent + size != buf->sg.length || datalen)
+		flags |= MSG_MORE;
+
+	res = iscsi_send(sk, buf, size, flags);
+	debug_tcp("sendhdr %d bytes, sent %d res %d\n", size, buf->sent, res);
+	if (res >= 0) {
+		conn->txdata_octets += res;
+		buf->sent += res;
+		if (size != res)
+			return -EAGAIN;
+		return 0;
+	} else if (res == -EAGAIN) {
+		conn->sendpage_failures_cnt++;
+		set_bit(SUSPEND_BIT, &conn->suspend_tx);
+	} else if (res == -EPIPE)
+		iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
+
+	return res;
+}
+
+/**
+ * iscsi_sendpage - send one page of iSCSI Data-Out.
+ * @conn: iscsi connection
+ * @buf: buffer to write from
+ * @count: remaining data
+ * @sent: number of bytes sent
+ *
+ * Notes:
+ *	(Tx, Fast Path)
+ **/
+static inline int
+iscsi_sendpage(struct iscsi_conn *conn, struct iscsi_buf *buf,
+	       int *count, int *sent)
+{
+	struct socket *sk = conn->sock;
+	int flags = 0; /* MSG_DONTWAIT; */
+	int res, size;
+
+	size = buf->sg.length - buf->sent;
+	BUG_ON(buf->sent + size > buf->sg.length);
+	if (size > *count)
+		size = *count;
+	if (buf->sent + size != buf->sg.length)
+		flags |= MSG_MORE;
+
+	res = iscsi_send(sk, buf, size, flags);
+	debug_tcp("sendpage: %d bytes, sent %d left %d sent %d res %d\n",
+		  size, buf->sent, *count, *sent, res);
+	if (res >= 0) {
+		conn->txdata_octets += res;
+		buf->sent += res;
+		*count -= res;
+		*sent += res;
+		if (size != res)
+			return -EAGAIN;
+		return 0;
+	} else if (res == -EAGAIN) {
+		conn->sendpage_failures_cnt++;
+		set_bit(SUSPEND_BIT, &conn->suspend_tx);
+	} else if (res == -EPIPE)
+		iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
+
+	return res;
+}
+
+static inline void
+iscsi_data_digest_init(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
+{
+	BUG_ON(!conn->data_tx_tfm);
+	crypto_digest_init(conn->data_tx_tfm);
+	ctask->digest_count = 4;
+}
+
+static inline void
+iscsi_buf_data_digest_update(struct iscsi_conn *conn, struct iscsi_buf *buf)
+{
+	struct scatterlist sg;
+
+	if (buf->sg.offset != -1)
+		crypto_digest_update(conn->data_tx_tfm, &buf->sg, 1);
+	else {
+		sg_init_one(&sg, (char *)buf->sg.page, buf->sg.length);
+		crypto_digest_update(conn->data_tx_tfm, &sg, 1);
+	}
+}
+
+static inline int
+iscsi_digest_final_send(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask,
+			struct iscsi_buf *buf, uint32_t *digest, int final)
+{
+	int rc = 0;
+	int sent = 0;
+
+	if (final)
+		crypto_digest_final(conn->data_tx_tfm, (u8*)digest);
+
+	iscsi_buf_init_virt(buf, (char*)digest, 4);
+	rc = iscsi_sendpage(conn, buf, &ctask->digest_count, &sent);
+	if (rc) {
+		ctask->datadigest = *digest;
+		ctask->xmstate |= XMSTATE_DATA_DIGEST;
+	} else
+		ctask->digest_count = 4;
+	return rc;
+}
+
+/**
+ * iscsi_solicit_data_cont - initialize next Data-Out
+ * @conn: iscsi connection
+ * @ctask: scsi command task
+ * @r2t: R2T info
+ * @left: bytes left to transfer
+ *
+ * Notes:
+ *	Initialize next Data-Out within this R2T sequence and continue
+ *	to process next Scatter-Gather element(if any) of this SCSI command.
+ *
+ *	Called under connection lock.
+ **/
+static void
+iscsi_solicit_data_cont(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask,
+			struct iscsi_r2t_info *r2t, int left)
+{
+	struct iscsi_data *hdr;
+	struct iscsi_data_task *dtask;
+	struct scsi_cmnd *sc = ctask->sc;
+	int new_offset;
+
+	dtask = mempool_alloc(ctask->datapool, GFP_ATOMIC);
+	BUG_ON(!dtask);
+	hdr = &dtask->hdr;
+	memset(hdr, 0, sizeof(struct iscsi_data));
+	hdr->ttt = r2t->ttt;
+	hdr->datasn = cpu_to_be32(r2t->solicit_datasn);
+	r2t->solicit_datasn++;
+	hdr->opcode = ISCSI_OP_SCSI_DATA_OUT;
+	memcpy(hdr->lun, ctask->hdr.lun, sizeof(hdr->lun));
+	hdr->itt = ctask->hdr.itt;
+	hdr->exp_statsn = r2t->exp_statsn;
+	new_offset = r2t->data_offset + r2t->sent;
+	hdr->offset = cpu_to_be32(new_offset);
+	if (left > conn->max_xmit_dlength) {
+		hton24(hdr->dlength, conn->max_xmit_dlength);
+		r2t->data_count = conn->max_xmit_dlength;
+	} else {
+		hton24(hdr->dlength, left);
+		r2t->data_count = left;
+		hdr->flags = ISCSI_FLAG_CMD_FINAL;
+	}
+	conn->dataout_pdus_cnt++;
+
+	iscsi_buf_init_hdr(conn, &r2t->headbuf, (char*)hdr,
+			   (u8 *)dtask->hdrext);
+
+	r2t->dtask = dtask;
+
+	if (sc->use_sg && !iscsi_buf_left(&r2t->sendbuf)) {
+		BUG_ON(ctask->bad_sg == r2t->sg);
+		iscsi_buf_init_sg(&r2t->sendbuf, r2t->sg);
+		r2t->sg += 1;
+	} else
+		iscsi_buf_init_iov(&ctask->sendbuf,
+			    (char*)sc->request_buffer + new_offset,
+			    r2t->data_count);
+
+	list_add(&dtask->item, &ctask->dataqueue);
+}
+
+static void
+iscsi_unsolicit_data_init(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
+{
+	struct iscsi_data *hdr;
+	struct iscsi_data_task *dtask;
+
+	dtask = mempool_alloc(ctask->datapool, GFP_ATOMIC);
+	BUG_ON(!dtask);
+	hdr = &dtask->hdr;
+	memset(hdr, 0, sizeof(struct iscsi_data));
+	hdr->ttt = cpu_to_be32(ISCSI_RESERVED_TAG);
+	hdr->datasn = cpu_to_be32(ctask->unsol_datasn);
+	ctask->unsol_datasn++;
+	hdr->opcode = ISCSI_OP_SCSI_DATA_OUT;
+	memcpy(hdr->lun, ctask->hdr.lun, sizeof(hdr->lun));
+	hdr->itt = ctask->hdr.itt;
+	hdr->exp_statsn = cpu_to_be32(conn->exp_statsn);
+	hdr->offset = cpu_to_be32(ctask->total_length -
+				  ctask->r2t_data_count -
+				  ctask->unsol_count);
+	if (ctask->unsol_count > conn->max_xmit_dlength) {
+		hton24(hdr->dlength, conn->max_xmit_dlength);
+		ctask->data_count = conn->max_xmit_dlength;
+		hdr->flags = 0;
+	} else {
+		hton24(hdr->dlength, ctask->unsol_count);
+		ctask->data_count = ctask->unsol_count;
+		hdr->flags = ISCSI_FLAG_CMD_FINAL;
+	}
+
+	iscsi_buf_init_hdr(conn, &ctask->headbuf, (char*)hdr,
+			   (u8 *)dtask->hdrext);
+
+	list_add(&dtask->item, &ctask->dataqueue);
+
+	ctask->dtask = dtask;
+}
+
+/**
+ * iscsi_cmd_init - Initialize iSCSI SCSI_READ or SCSI_WRITE commands
+ * @conn: iscsi connection
+ * @ctask: scsi command task
+ * @sc: scsi command
+ **/
+static void
+iscsi_cmd_init(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask,
+		struct scsi_cmnd *sc)
+{
+	struct iscsi_session *session = conn->session;
+
+	BUG_ON(__kfifo_len(ctask->r2tqueue));
+
+	ctask->sc = sc;
+	ctask->conn = conn;
+	ctask->hdr.opcode = ISCSI_OP_SCSI_CMD;
+	ctask->hdr.flags = ISCSI_ATTR_SIMPLE;
+	int_to_scsilun(sc->device->lun, (struct scsi_lun *)ctask->hdr.lun);
+	ctask->hdr.itt = ctask->itt | (conn->id << CID_SHIFT) |
+			 (session->age << AGE_SHIFT);
+	ctask->hdr.data_length = cpu_to_be32(sc->request_bufflen);
+	ctask->hdr.cmdsn = cpu_to_be32(session->cmdsn); session->cmdsn++;
+	ctask->hdr.exp_statsn = cpu_to_be32(conn->exp_statsn);
+	memcpy(ctask->hdr.cdb, sc->cmnd, sc->cmd_len);
+	memset(&ctask->hdr.cdb[sc->cmd_len], 0, MAX_COMMAND_SIZE - sc->cmd_len);
+
+	ctask->mtask = NULL;
+	ctask->sent = 0;
+	ctask->sg_count = 0;
+
+	ctask->total_length = sc->request_bufflen;
+
+	if (sc->sc_data_direction == DMA_TO_DEVICE) {
+		ctask->exp_r2tsn = 0;
+		ctask->hdr.flags |= ISCSI_FLAG_CMD_WRITE;
+		BUG_ON(ctask->total_length == 0);
+		if (sc->use_sg) {
+			struct scatterlist *sg = sc->request_buffer;
+
+			iscsi_buf_init_sg(&ctask->sendbuf,
+					  &sg[ctask->sg_count++]);
+			ctask->sg = sg;
+			ctask->bad_sg = sg + sc->use_sg;
+		} else {
+			iscsi_buf_init_iov(&ctask->sendbuf, sc->request_buffer,
+					sc->request_bufflen);
+		}
+
+		/*
+		 * Write counters:
+		 *
+		 *	imm_count	bytes to be sent right after
+		 *			SCSI PDU Header
+		 *
+		 *	unsol_count	bytes(as Data-Out) to be sent
+		 *			without	R2T ack right after
+		 *			immediate data
+		 *
+		 *	r2t_data_count	bytes to be sent via R2T ack's
+		 *
+		 *      pad_count       bytes to be sent as zero-padding
+		 */
+		ctask->imm_count = 0;
+		ctask->unsol_count = 0;
+		ctask->unsol_datasn = 0;
+		ctask->xmstate = XMSTATE_W_HDR;
+		/* calculate write padding */
+		ctask->pad_count = ctask->total_length & (ISCSI_PAD_LEN-1);
+		if (ctask->pad_count) {
+			ctask->pad_count = ISCSI_PAD_LEN - ctask->pad_count;
+			debug_scsi("write padding %d bytes\n",
+				ctask->pad_count);
+			ctask->xmstate |= XMSTATE_W_PAD;
+		}
+		if (session->imm_data_en) {
+			if (ctask->total_length >= session->first_burst)
+				ctask->imm_count = min(session->first_burst,
+							conn->max_xmit_dlength);
+			else
+				ctask->imm_count = min(ctask->total_length,
+							conn->max_xmit_dlength);
+			hton24(ctask->hdr.dlength, ctask->imm_count);
+			ctask->xmstate |= XMSTATE_IMM_DATA;
+		} else
+			zero_data(ctask->hdr.dlength);
+
+		if (!session->initial_r2t_en)
+			ctask->unsol_count = min(session->first_burst,
+				ctask->total_length) - ctask->imm_count;
+		if (!ctask->unsol_count)
+			/* No unsolicit Data-Out's */
+			ctask->hdr.flags |= ISCSI_FLAG_CMD_FINAL;
+		else
+			ctask->xmstate |= XMSTATE_UNS_HDR | XMSTATE_UNS_INIT;
+
+		ctask->r2t_data_count = ctask->total_length -
+				    ctask->imm_count -
+				    ctask->unsol_count;
+
+		debug_scsi("cmd [itt %x total %d imm %d imm_data %d "
+			   "r2t_data %d]\n",
+			   ctask->itt, ctask->total_length, ctask->imm_count,
+			   ctask->unsol_count, ctask->r2t_data_count);
+	} else {
+		ctask->hdr.flags |= ISCSI_FLAG_CMD_FINAL;
+		if (sc->sc_data_direction == DMA_FROM_DEVICE)
+			ctask->hdr.flags |= ISCSI_FLAG_CMD_READ;
+		ctask->datasn = 0;
+		ctask->xmstate = XMSTATE_R_HDR;
+		zero_data(ctask->hdr.dlength);
+	}
+
+	iscsi_buf_init_hdr(conn, &ctask->headbuf, (char*)&ctask->hdr,
+			    (u8 *)ctask->hdrext);
+	conn->scsicmd_pdus_cnt++;
+}
+
+/**
+ * iscsi_mtask_xmit - xmit management(immediate) task
+ * @conn: iscsi connection
+ * @mtask: 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.
+ *
+ *	Management xmit state machine consists of two states:
+ *		IN_PROGRESS_IMM_HEAD - PDU Header xmit in progress
+ *		IN_PROGRESS_IMM_DATA - PDU Data xmit in progress
+ **/
+static int
+iscsi_mtask_xmit(struct iscsi_conn *conn, struct iscsi_mgmt_task *mtask)
+{
+
+	debug_scsi("mtask deq [cid %d state %x itt 0x%x]\n",
+		conn->id, mtask->xmstate, mtask->itt);
+
+	if (mtask->xmstate & XMSTATE_IMM_HDR) {
+		mtask->xmstate &= ~XMSTATE_IMM_HDR;
+		if (mtask->data_count)
+			mtask->xmstate |= XMSTATE_IMM_DATA;
+		if (iscsi_sendhdr(conn, &mtask->headbuf, mtask->data_count)) {
+			mtask->xmstate |= XMSTATE_IMM_HDR;
+			if (mtask->data_count)
+				mtask->xmstate &= ~XMSTATE_IMM_DATA;
+			return -EAGAIN;
+		}
+	}
+
+	if (mtask->xmstate & XMSTATE_IMM_DATA) {
+		BUG_ON(!mtask->data_count);
+		mtask->xmstate &= ~XMSTATE_IMM_DATA;
+		/* FIXME: implement.
+		 * Virtual buffer could be spreaded across multiple pages...
+		 */
+		do {
+			if (iscsi_sendpage(conn, &mtask->sendbuf,
+				   &mtask->data_count, &mtask->sent)) {
+				mtask->xmstate |= XMSTATE_IMM_DATA;
+				return -EAGAIN;
+			}
+		} while (mtask->data_count);
+	}
+
+	BUG_ON(mtask->xmstate != XMSTATE_IDLE);
+	return 0;
+}
+
+static inline int
+handle_xmstate_r_hdr(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
+{
+	ctask->xmstate &= ~XMSTATE_R_HDR;
+	if (!iscsi_sendhdr(conn, &ctask->headbuf, 0)) {
+		BUG_ON(ctask->xmstate != XMSTATE_IDLE);
+		return 0; /* wait for Data-In */
+	}
+	ctask->xmstate |= XMSTATE_R_HDR;
+	return -EAGAIN;
+}
+
+static inline int
+handle_xmstate_w_hdr(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
+{
+	ctask->xmstate &= ~XMSTATE_W_HDR;
+	if (iscsi_sendhdr(conn, &ctask->headbuf, ctask->imm_count)) {
+		ctask->xmstate |= XMSTATE_W_HDR;
+		return -EAGAIN;
+	}
+	return 0;
+}
+
+static inline int
+handle_xmstate_data_digest(struct iscsi_conn *conn,
+			   struct iscsi_cmd_task *ctask)
+{
+	ctask->xmstate &= ~XMSTATE_DATA_DIGEST;
+	debug_tcp("resent data digest 0x%x\n", ctask->datadigest);
+	if (iscsi_digest_final_send(conn, ctask, &ctask->immbuf,
+				    &ctask->datadigest, 0)) {
+		ctask->xmstate |= XMSTATE_DATA_DIGEST;
+		debug_tcp("resent data digest 0x%x fail!\n",
+			  ctask->datadigest);
+		return -EAGAIN;
+	}
+	return 0;
+}
+
+static inline int
+handle_xmstate_imm_data(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
+{
+	BUG_ON(!ctask->imm_count);
+	ctask->xmstate &= ~XMSTATE_IMM_DATA;
+
+	if (conn->datadgst_en) {
+		iscsi_data_digest_init(conn, ctask);
+		ctask->immdigest = 0;
+	}
+
+	for (;;) {
+		if (iscsi_sendpage(conn, &ctask->sendbuf, &ctask->imm_count,
+				   &ctask->sent)) {
+			ctask->xmstate |= XMSTATE_IMM_DATA;
+			if (conn->datadgst_en) {
+				crypto_digest_final(conn->data_tx_tfm,
+						(u8*)&ctask->immdigest);
+				debug_tcp("tx imm sendpage fail 0x%x\n",
+					  ctask->datadigest);
+			}
+			return -EAGAIN;
+		}
+		if (conn->datadgst_en)
+			iscsi_buf_data_digest_update(conn, &ctask->sendbuf);
+
+		if (!ctask->imm_count)
+			break;
+		iscsi_buf_init_sg(&ctask->sendbuf,
+				  &ctask->sg[ctask->sg_count++]);
+	}
+
+	if (conn->datadgst_en && !(ctask->xmstate & XMSTATE_W_PAD)) {
+		if (iscsi_digest_final_send(conn, ctask, &ctask->immbuf,
+				            &ctask->immdigest, 1)) {
+			debug_tcp("sending imm digest 0x%x fail!\n",
+				  ctask->immdigest);
+			return -EAGAIN;
+		}
+		debug_tcp("sending imm digest 0x%x\n", ctask->immdigest);
+	}
+
+	return 0;
+}
+
+static inline int
+handle_xmstate_uns_hdr(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
+{
+	struct iscsi_data_task *dtask;
+
+	ctask->xmstate |= XMSTATE_UNS_DATA;
+	if (ctask->xmstate & XMSTATE_UNS_INIT) {
+		iscsi_unsolicit_data_init(conn, ctask);
+		BUG_ON(!ctask->dtask);
+		dtask = ctask->dtask;
+
+		ctask->xmstate &= ~XMSTATE_UNS_INIT;
+	}
+	if (iscsi_sendhdr(conn, &ctask->headbuf, ctask->data_count)) {
+		ctask->xmstate &= ~XMSTATE_UNS_DATA;
+		ctask->xmstate |= XMSTATE_UNS_HDR;
+		return -EAGAIN;
+	}
+
+	debug_scsi("uns dout [itt 0x%x dlen %d sent %d]\n",
+		   ctask->itt, ctask->unsol_count, ctask->sent);
+	return 0;
+}
+
+static inline int
+handle_xmstate_uns_data(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
+{
+	struct iscsi_data_task *dtask = ctask->dtask;
+
+	BUG_ON(!ctask->data_count);
+	ctask->xmstate &= ~XMSTATE_UNS_DATA;
+
+	if (conn->datadgst_en) {
+		iscsi_data_digest_init(conn, ctask);
+		dtask->digest = 0;
+	}
+
+	for (;;) {
+		int start = ctask->sent;
+
+		if (iscsi_sendpage(conn, &ctask->sendbuf, &ctask->data_count,
+				   &ctask->sent)) {
+			ctask->unsol_count -= ctask->sent - start;
+			ctask->xmstate |= XMSTATE_UNS_DATA;
+			/* will continue with this ctask later.. */
+			if (conn->datadgst_en) {
+				crypto_digest_final(conn->data_tx_tfm,
+						(u8 *)&dtask->digest);
+				debug_tcp("tx uns data fail 0x%x\n",
+					  dtask->digest);
+			}
+			return -EAGAIN;
+		}
+
+		BUG_ON(ctask->sent > ctask->total_length);
+		ctask->unsol_count -= ctask->sent - start;
+
+		/*
+		 * XXX:we may run here with un-initial sendbuf.
+		 * so pass it
+		 */
+		if (conn->datadgst_en && ctask->sent - start > 0)
+			iscsi_buf_data_digest_update(conn, &ctask->sendbuf);
+
+		if (!ctask->data_count)
+			break;
+		iscsi_buf_init_sg(&ctask->sendbuf,
+				  &ctask->sg[ctask->sg_count++]);
+	}
+	BUG_ON(ctask->unsol_count < 0);
+
+	/*
+	 * Done with the Data-Out. Next, check if we need
+	 * to send another unsolicited Data-Out.
+	 */
+	if (ctask->unsol_count) {
+		if (conn->datadgst_en) {
+			if (iscsi_digest_final_send(conn, ctask,
+						    &dtask->digestbuf,
+						    &dtask->digest, 1)) {
+				debug_tcp("send uns digest 0x%x fail\n",
+					  dtask->digest);
+				return -EAGAIN;
+			}
+			debug_tcp("sending uns digest 0x%x, more uns\n",
+				  dtask->digest);
+		}
+		ctask->xmstate |= XMSTATE_UNS_INIT;
+		return 1;
+	}
+
+	if (conn->datadgst_en && !(ctask->xmstate & XMSTATE_W_PAD)) {
+		if (iscsi_digest_final_send(conn, ctask,
+					    &dtask->digestbuf,
+					    &dtask->digest, 1)) {
+			debug_tcp("send last uns digest 0x%x fail\n",
+				   dtask->digest);
+			return -EAGAIN;
+		}
+		debug_tcp("sending uns digest 0x%x\n",dtask->digest);
+	}
+
+	return 0;
+}
+
+static inline int
+handle_xmstate_sol_data(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
+{
+	struct iscsi_session *session = conn->session;
+	struct iscsi_r2t_info *r2t = ctask->r2t;
+	struct iscsi_data_task *dtask = r2t->dtask;
+	int left;
+
+	ctask->xmstate &= ~XMSTATE_SOL_DATA;
+	ctask->dtask = dtask;
+
+	if (conn->datadgst_en) {
+		iscsi_data_digest_init(conn, ctask);
+		dtask->digest = 0;
+	}
+solicit_again:
+	/*
+	 * send Data-Out whitnin this R2T sequence.
+	 */
+	if (!r2t->data_count)
+		goto data_out_done;
+
+	if (iscsi_sendpage(conn, &r2t->sendbuf, &r2t->data_count, &r2t->sent)) {
+		ctask->xmstate |= XMSTATE_SOL_DATA;
+		/* will continue with this ctask later.. */
+		if (conn->datadgst_en) {
+			crypto_digest_final(conn->data_tx_tfm,
+					  (u8 *)&dtask->digest);
+			debug_tcp("r2t data send fail 0x%x\n", dtask->digest);
+		}
+		return -EAGAIN;
+	}
+
+	BUG_ON(r2t->data_count < 0);
+	if (conn->datadgst_en)
+		iscsi_buf_data_digest_update(conn, &r2t->sendbuf);
+
+	if (r2t->data_count) {
+		BUG_ON(ctask->sc->use_sg == 0);
+		if (!iscsi_buf_left(&r2t->sendbuf)) {
+			BUG_ON(ctask->bad_sg == r2t->sg);
+			iscsi_buf_init_sg(&r2t->sendbuf, r2t->sg);
+			r2t->sg += 1;
+		}
+		goto solicit_again;
+	}
+
+data_out_done:
+	/*
+	 * Done with this Data-Out. Next, check if we have
+	 * to send another Data-Out for this R2T.
+	 */
+	BUG_ON(r2t->data_length - r2t->sent < 0);
+	left = r2t->data_length - r2t->sent;
+	if (left) {
+		if (conn->datadgst_en) {
+			if (iscsi_digest_final_send(conn, ctask,
+						    &dtask->digestbuf,
+						    &dtask->digest, 1)) {
+				debug_tcp("send r2t data digest 0x%x"
+					  "fail\n", dtask->digest);
+				return -EAGAIN;
+			}
+			debug_tcp("r2t data send digest 0x%x\n",
+				  dtask->digest);
+		}
+		iscsi_solicit_data_cont(conn, ctask, r2t, left);
+		ctask->xmstate |= XMSTATE_SOL_DATA;
+		ctask->xmstate &= ~XMSTATE_SOL_HDR;
+		return 1;
+	}
+
+	/*
+	 * Done with this R2T. Check if there are more
+	 * outstanding R2Ts ready to be processed.
+	 */
+	BUG_ON(ctask->r2t_data_count - r2t->data_length < 0);
+	if (conn->datadgst_en) {
+		if (iscsi_digest_final_send(conn, ctask, &dtask->digestbuf,
+					    &dtask->digest, 1)) {
+			debug_tcp("send last r2t data digest 0x%x"
+				  "fail\n", dtask->digest);
+			return -EAGAIN;
+		}
+		debug_tcp("r2t done dout digest 0x%x\n", dtask->digest);
+	}
+
+	ctask->r2t_data_count -= r2t->data_length;
+	ctask->r2t = NULL;
+	spin_lock_bh(&session->lock);
+	__kfifo_put(ctask->r2tpool.queue, (void*)&r2t, sizeof(void*));
+	spin_unlock_bh(&session->lock);
+	if (__kfifo_get(ctask->r2tqueue, (void*)&r2t, sizeof(void*))) {
+		ctask->r2t = r2t;
+		ctask->xmstate |= XMSTATE_SOL_DATA;
+		ctask->xmstate &= ~XMSTATE_SOL_HDR;
+		return 1;
+	}
+
+	return 0;
+}
+
+static inline int
+handle_xmstate_w_pad(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
+{
+	struct iscsi_data_task *dtask = ctask->dtask;
+	int sent;
+
+	ctask->xmstate &= ~XMSTATE_W_PAD;
+	iscsi_buf_init_virt(&ctask->sendbuf, (char*)&ctask->pad,
+			    ctask->pad_count);
+	if (iscsi_sendpage(conn, &ctask->sendbuf, &ctask->pad_count, &sent)) {
+		ctask->xmstate |= XMSTATE_W_PAD;
+		return -EAGAIN;
+	}
+
+	if (conn->datadgst_en) {
+		iscsi_buf_data_digest_update(conn, &ctask->sendbuf);
+		/* imm data? */
+		if (!dtask) {
+			if (iscsi_digest_final_send(conn, ctask, &ctask->immbuf,
+						    &ctask->immdigest, 1)) {
+				debug_tcp("send padding digest 0x%x"
+					  "fail!\n", ctask->immdigest);
+				return -EAGAIN;
+			}
+			debug_tcp("done with padding, digest 0x%x\n",
+				  ctask->datadigest);
+		} else {
+			if (iscsi_digest_final_send(conn, ctask,
+						    &dtask->digestbuf,
+						    &dtask->digest, 1)) {
+				debug_tcp("send padding digest 0x%x"
+				          "fail\n", dtask->digest);
+				return -EAGAIN;
+			}
+			debug_tcp("done with padding, digest 0x%x\n",
+				  dtask->digest);
+		}
+	}
+
+	return 0;
+}
+
+static int
+iscsi_ctask_xmit(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
+{
+	int rc = 0;
+
+	debug_scsi("ctask deq [cid %d xmstate %x itt 0x%x]\n",
+		conn->id, ctask->xmstate, ctask->itt);
+
+	/*
+	 * serialize with TMF AbortTask
+	 */
+	if (ctask->mtask)
+		return rc;
+
+	if (ctask->xmstate & XMSTATE_R_HDR) {
+		rc = handle_xmstate_r_hdr(conn, ctask);
+		return rc;
+	}
+
+	if (ctask->xmstate & XMSTATE_W_HDR) {
+		rc = handle_xmstate_w_hdr(conn, ctask);
+		if (rc)
+			return rc;
+	}
+
+	/* XXX: for data digest xmit recover */
+	if (ctask->xmstate & XMSTATE_DATA_DIGEST) {
+		rc = handle_xmstate_data_digest(conn, ctask);
+		if (rc)
+			return rc;
+	}
+
+	if (ctask->xmstate & XMSTATE_IMM_DATA) {
+		rc = handle_xmstate_imm_data(conn, ctask);
+		if (rc)
+			return rc;
+	}
+
+	if (ctask->xmstate & XMSTATE_UNS_HDR) {
+		BUG_ON(!ctask->unsol_count);
+		ctask->xmstate &= ~XMSTATE_UNS_HDR;
+unsolicit_head_again:
+		rc = handle_xmstate_uns_hdr(conn, ctask);
+		if (rc)
+			return rc;
+	}
+
+	if (ctask->xmstate & XMSTATE_UNS_DATA) {
+		rc = handle_xmstate_uns_data(conn, ctask);
+		if (rc == 1)
+			goto unsolicit_head_again;
+		else if (rc)
+			return rc;
+		goto done;
+	}
+
+	if (ctask->xmstate & XMSTATE_SOL_HDR) {
+		struct iscsi_r2t_info *r2t;
+
+		ctask->xmstate &= ~XMSTATE_SOL_HDR;
+		ctask->xmstate |= XMSTATE_SOL_DATA;
+		if (!ctask->r2t)
+			__kfifo_get(ctask->r2tqueue, (void*)&ctask->r2t,
+				    sizeof(void*));
+solicit_head_again:
+		r2t = ctask->r2t;
+
+		if (iscsi_sendhdr(conn, &r2t->headbuf, r2t->data_count)) {
+			ctask->xmstate &= ~XMSTATE_SOL_DATA;
+			ctask->xmstate |= XMSTATE_SOL_HDR;
+			return -EAGAIN;
+		}
+
+		debug_scsi("sol dout [dsn %d itt 0x%x dlen %d sent %d]\n",
+			r2t->solicit_datasn - 1, ctask->itt, r2t->data_count,
+			r2t->sent);
+	}
+
+	if (ctask->xmstate & XMSTATE_SOL_DATA) {
+		rc = handle_xmstate_sol_data(conn, ctask);
+		if (rc == 1)
+			goto solicit_head_again;
+		if (rc)
+			return rc;
+	}
+
+done:
+	/*
+	 * Last thing to check is whether we need to send write
+	 * padding. Note that we check for xmstate equality, not just the bit.
+	 */
+	if (ctask->xmstate == XMSTATE_W_PAD)
+		rc = handle_xmstate_w_pad(conn, ctask);
+
+	return rc;
+}
+
+/**
+ * iscsi_data_xmit - xmit any command into the scheduled connection
+ * @conn: iscsi connection
+ *
+ * Notes:
+ *	The function can return -EAGAIN in which case the caller must
+ *	re-schedule it again later or recover. '0' return code means
+ *	successful xmit.
+ **/
+static int
+iscsi_data_xmit(struct iscsi_conn *conn)
+{
+	if (unlikely(conn->suspend_tx)) {
+		debug_tcp("conn %d Tx suspended!\n", conn->id);
+		return 0;
+	}
+
+	/*
+	 * Transmit in the following order:
+	 *
+	 * 1) un-finished xmit (ctask or mtask)
+	 * 2) immediate control PDUs
+	 * 3) write data
+	 * 4) SCSI commands
+	 * 5) non-immediate control PDUs
+	 *
+	 * No need to lock around __kfifo_get as long as
+	 * there's one producer and one consumer.
+	 */
+
+	BUG_ON(conn->ctask && conn->mtask);
+
+	if (conn->ctask) {
+		if (iscsi_ctask_xmit(conn, conn->ctask))
+			goto again;
+		/* done with this in-progress ctask */
+		conn->ctask = NULL;
+	}
+	if (conn->mtask) {
+	        if (iscsi_mtask_xmit(conn, conn->mtask))
+		        goto again;
+		/* done with this in-progress mtask */
+		conn->mtask = NULL;
+	}
+
+	/* process immediate first */
+        if (unlikely(__kfifo_len(conn->immqueue))) {
+		struct iscsi_session *session = conn->session;
+	        while (__kfifo_get(conn->immqueue, (void*)&conn->mtask,
+			           sizeof(void*))) {
+		        if (iscsi_mtask_xmit(conn, conn->mtask))
+			        goto again;
+
+		        if (conn->mtask->hdr.itt ==
+					cpu_to_be32(ISCSI_RESERVED_TAG)) {
+			        spin_lock_bh(&session->lock);
+			        __kfifo_put(session->mgmtpool.queue,
+					    (void*)&conn->mtask, sizeof(void*));
+			        spin_unlock_bh(&session->lock);
+		        }
+	        }
+		/* done with this mtask */
+		conn->mtask = NULL;
+	}
+
+	/* process write queue */
+	while (__kfifo_get(conn->writequeue, (void*)&conn->ctask,
+			   sizeof(void*))) {
+		if (iscsi_ctask_xmit(conn, conn->ctask))
+			goto again;
+	}
+
+	/* process command queue */
+	while (__kfifo_get(conn->xmitqueue, (void*)&conn->ctask,
+			   sizeof(void*))) {
+		if (iscsi_ctask_xmit(conn, conn->ctask))
+			goto again;
+	}
+	/* done with this ctask */
+	conn->ctask = NULL;
+
+	/* process the rest control plane PDUs, if any */
+        if (unlikely(__kfifo_len(conn->mgmtqueue))) {
+		struct iscsi_session *session = conn->session;
+
+	        while (__kfifo_get(conn->mgmtqueue, (void*)&conn->mtask,
+			           sizeof(void*))) {
+		        if (iscsi_mtask_xmit(conn, conn->mtask))
+			        goto again;
+
+		        if (conn->mtask->hdr.itt ==
+					cpu_to_be32(ISCSI_RESERVED_TAG)) {
+			        spin_lock_bh(&session->lock);
+			        __kfifo_put(session->mgmtpool.queue,
+					    (void*)&conn->mtask,
+				            sizeof(void*));
+			        spin_unlock_bh(&session->lock);
+		        }
+	        }
+		/* done with this mtask */
+		conn->mtask = NULL;
+	}
+
+	return 0;
+
+again:
+	if (unlikely(conn->suspend_tx))
+		return 0;
+
+	return -EAGAIN;
+}
+
+static void
+iscsi_xmitworker(void *data)
+{
+	struct iscsi_conn *conn = data;
+
+	/*
+	 * serialize Xmit worker on a per-connection basis.
+	 */
+	down(&conn->xmitsema);
+	if (iscsi_data_xmit(conn))
+		schedule_work(&conn->xmitwork);
+	up(&conn->xmitsema);
+}
+
+#define FAILURE_BAD_HOST		1
+#define FAILURE_SESSION_FAILED		2
+#define FAILURE_SESSION_FREED		3
+#define FAILURE_WINDOW_CLOSED		4
+#define FAILURE_SESSION_TERMINATE	5
+
+static int
+iscsi_queuecommand(struct scsi_cmnd *sc, void (*done)(struct scsi_cmnd *))
+{
+	struct Scsi_Host *host;
+	int reason = 0;
+	struct iscsi_session *session;
+	struct iscsi_conn *conn = NULL;
+	struct iscsi_cmd_task *ctask = NULL;
+
+	sc->scsi_done = done;
+	sc->result = 0;
+
+	host = sc->device->host;
+	session = iscsi_hostdata(host->hostdata);
+	BUG_ON(host != session->host);
+
+	spin_lock(&session->lock);
+
+	if (session->state != ISCSI_STATE_LOGGED_IN) {
+		if (session->state == ISCSI_STATE_FAILED) {
+			reason = FAILURE_SESSION_FAILED;
+			goto reject;
+		} else if (session->state == ISCSI_STATE_TERMINATE) {
+			reason = FAILURE_SESSION_TERMINATE;
+			goto fault;
+		}
+		reason = FAILURE_SESSION_FREED;
+		goto fault;
+	}
+
+	/*
+	 * Check for iSCSI window and take care of CmdSN wrap-around
+	 */
+	if ((int)(session->max_cmdsn - session->cmdsn) < 0) {
+		reason = FAILURE_WINDOW_CLOSED;
+		goto reject;
+	}
+
+	conn = session->leadconn;
+
+	__kfifo_get(session->cmdpool.queue, (void*)&ctask, sizeof(void*));
+	BUG_ON(ctask->sc);
+
+	sc->SCp.phase = session->age;
+	sc->SCp.ptr = (char*)ctask;
+	iscsi_cmd_init(conn, ctask, sc);
+
+	__kfifo_put(conn->xmitqueue, (void*)&ctask, sizeof(void*));
+	debug_scsi(
+	       "ctask enq [%s cid %d sc %lx itt 0x%x len %d cmdsn %d win %d]\n",
+		sc->sc_data_direction == DMA_TO_DEVICE ? "write" : "read",
+		conn->id, (long)sc, ctask->itt, sc->request_bufflen,
+		session->cmdsn, session->max_cmdsn - session->exp_cmdsn + 1);
+	spin_unlock(&session->lock);
+
+        if (!in_interrupt() && !down_trylock(&conn->xmitsema)) {
+		spin_unlock_irq(host->host_lock);
+		if (iscsi_data_xmit(conn))
+			schedule_work(&conn->xmitwork);
+		up(&conn->xmitsema);
+		spin_lock_irq(host->host_lock);
+	} else
+		schedule_work(&conn->xmitwork);
+
+	return 0;
+
+reject:
+	spin_unlock(&session->lock);
+	debug_scsi("cmd 0x%x rejected (%d)\n", sc->cmnd[0], reason);
+	return SCSI_MLQUEUE_HOST_BUSY;
+
+fault:
+	spin_unlock(&session->lock);
+	printk(KERN_ERR "iscsi_tcp: cmd 0x%x is not queued (%d)\n",
+	       sc->cmnd[0], reason);
+	sc->sense_buffer[0] = 0x70;
+	sc->sense_buffer[2] = NOT_READY;
+	sc->sense_buffer[7] = 0x6;
+	sc->sense_buffer[12] = 0x08;
+	sc->sense_buffer[13] = 0x00;
+	sc->result = (DID_NO_CONNECT << 16);
+	sc->resid = sc->request_bufflen;
+	sc->scsi_done(sc);
+	return 0;
+}
+
+static int
+iscsi_pool_init(struct iscsi_queue *q, int max, void ***items, int item_size)
+{
+	int i;
+
+	*items = kmalloc(max * sizeof(void*), GFP_KERNEL);
+	if (*items == NULL)
+		return -ENOMEM;
+
+	q->max = max;
+	q->pool = kmalloc(max * sizeof(void*), GFP_KERNEL);
+	if (q->pool == NULL) {
+		kfree(*items);
+		return -ENOMEM;
+	}
+
+	q->queue = kfifo_init((void*)q->pool, max * sizeof(void*),
+			      GFP_KERNEL, NULL);
+	if (q->queue == ERR_PTR(-ENOMEM)) {
+		kfree(q->pool);
+		kfree(*items);
+		return -ENOMEM;
+	}
+
+	for (i = 0; i < max; i++) {
+		q->pool[i] = kmalloc(item_size, GFP_KERNEL);
+		if (q->pool[i] == NULL) {
+			int j;
+
+			for (j = 0; j < i; j++)
+				kfree(q->pool[j]);
+
+			kfifo_free(q->queue);
+			kfree(q->pool);
+			kfree(*items);
+			return -ENOMEM;
+		}
+		memset(q->pool[i], 0, item_size);
+		(*items)[i] = q->pool[i];
+		__kfifo_put(q->queue, (void*)&q->pool[i], sizeof(void*));
+	}
+	return 0;
+}
+
+static void
+iscsi_pool_free(struct iscsi_queue *q, void **items)
+{
+	int i;
+
+	for (i = 0; i < q->max; i++)
+		kfree(items[i]);
+	kfree(q->pool);
+	kfree(items);
+}
+
+static iscsi_connh_t
+iscsi_conn_create(iscsi_sessionh_t sessionh, uint32_t conn_idx)
+{
+	struct iscsi_session *session = iscsi_ptr(sessionh);
+	struct iscsi_conn *conn = NULL;
+
+	conn = kmalloc(sizeof(struct iscsi_conn), GFP_KERNEL);
+	if (conn == NULL)
+		goto conn_alloc_fail;
+	memset(conn, 0, sizeof(struct iscsi_conn));
+
+	conn->c_stage = ISCSI_CONN_INITIAL_STAGE;
+	conn->in_progress = IN_PROGRESS_WAIT_HEADER;
+	conn->id = conn_idx;
+	conn->exp_statsn = 0;
+	conn->tmabort_state = TMABORT_INITIAL;
+
+	/* initial operational parameters */
+	conn->hdr_size = sizeof(struct iscsi_hdr);
+	conn->data_size = DEFAULT_MAX_RECV_DATA_SEGMENT_LENGTH;
+	conn->max_recv_dlength = DEFAULT_MAX_RECV_DATA_SEGMENT_LENGTH;
+
+	spin_lock_init(&conn->lock);
+
+	/* initialize general xmit PDU commands queue */
+	conn->xmitqueue = kfifo_alloc(session->cmds_max * sizeof(void*),
+					GFP_KERNEL, NULL);
+	if (conn->xmitqueue == ERR_PTR(-ENOMEM))
+		goto xmitqueue_alloc_fail;
+
+	/* initialize write response PDU commands queue */
+	conn->writequeue = kfifo_alloc(session->cmds_max * sizeof(void*),
+					GFP_KERNEL, NULL);
+	if (conn->writequeue == ERR_PTR(-ENOMEM))
+		goto writequeue_alloc_fail;
+
+	/* initialize general immediate & non-immediate PDU commands queue */
+	conn->immqueue = kfifo_alloc(session->mgmtpool_max * sizeof(void*),
+			                GFP_KERNEL, NULL);
+	if (conn->immqueue == ERR_PTR(-ENOMEM))
+		goto immqueue_alloc_fail;
+
+	conn->mgmtqueue = kfifo_alloc(session->mgmtpool_max * sizeof(void*),
+			                GFP_KERNEL, NULL);
+	if (conn->mgmtqueue == ERR_PTR(-ENOMEM))
+		goto mgmtqueue_alloc_fail;
+
+	INIT_WORK(&conn->xmitwork, iscsi_xmitworker, conn);
+
+	/* allocate login_mtask used for the login/text sequences */
+	spin_lock_bh(&session->lock);
+	if (!__kfifo_get(session->mgmtpool.queue,
+                         (void*)&conn->login_mtask,
+			 sizeof(void*))) {
+		spin_unlock_bh(&session->lock);
+		goto login_mtask_alloc_fail;
+	}
+	spin_unlock_bh(&session->lock);
+
+	/* allocate initial PDU receive place holder */
+	if (conn->data_size <= PAGE_SIZE)
+		conn->data = kmalloc(conn->data_size, GFP_KERNEL);
+	else
+		conn->data = (void*)__get_free_pages(GFP_KERNEL,
+					get_order(conn->data_size));
+	if (!conn->data)
+		goto max_recv_dlenght_alloc_fail;
+
+	init_timer(&conn->tmabort_timer);
+	init_MUTEX(&conn->xmitsema);
+	init_waitqueue_head(&conn->ehwait);
+
+	return iscsi_handle(conn);
+
+max_recv_dlenght_alloc_fail:
+	spin_lock_bh(&session->lock);
+	__kfifo_put(session->mgmtpool.queue, (void*)&conn->login_mtask,
+		    sizeof(void*));
+	spin_unlock_bh(&session->lock);
+login_mtask_alloc_fail:
+	kfifo_free(conn->mgmtqueue);
+mgmtqueue_alloc_fail:
+	kfifo_free(conn->immqueue);
+immqueue_alloc_fail:
+	kfifo_free(conn->writequeue);
+writequeue_alloc_fail:
+	kfifo_free(conn->xmitqueue);
+xmitqueue_alloc_fail:
+	kfree(conn);
+conn_alloc_fail:
+	return iscsi_handle(NULL);
+}
+
+static void
+iscsi_conn_destroy(iscsi_connh_t connh)
+{
+	struct iscsi_conn *conn = iscsi_ptr(connh);
+	struct iscsi_session *session = conn->session;
+
+	down(&conn->xmitsema);
+	set_bit(SUSPEND_BIT, &conn->suspend_tx);
+	if (conn->c_stage == ISCSI_CONN_INITIAL_STAGE && conn->sock) {
+		struct sock *sk = conn->sock->sk;
+
+		/*
+		 * conn_start() has never been called!
+		 * need to cleanup the socket.
+		 */
+		write_lock_bh(&sk->sk_callback_lock);
+		set_bit(SUSPEND_BIT, &conn->suspend_rx);
+		write_unlock_bh(&sk->sk_callback_lock);
+
+		sock_hold(conn->sock->sk);
+		iscsi_conn_restore_callbacks(conn);
+		sock_put(conn->sock->sk);
+		sock_release(conn->sock);
+		conn->sock = NULL;
+	}
+
+	spin_lock_bh(&session->lock);
+	conn->c_stage = ISCSI_CONN_CLEANUP_WAIT;
+	if (session->leadconn == conn) {
+		/*
+		 * leading connection? then give up on recovery.
+		 */
+		session->state = ISCSI_STATE_TERMINATE;
+		wake_up(&conn->ehwait);
+	}
+	spin_unlock_bh(&session->lock);
+
+	up(&conn->xmitsema);
+
+	/*
+	 * Block until all in-progress commands for this connection
+	 * time out or fail.
+	 */
+	for (;;) {
+		spin_lock_bh(&conn->lock);
+		if (!session->host->host_busy) { /* OK for ERL == 0 */
+			spin_unlock_bh(&conn->lock);
+			break;
+		}
+		spin_unlock_bh(&conn->lock);
+		msleep_interruptible(500);
+		printk("conn_destroy(): host_busy %d host_failed %d\n",
+			session->host->host_busy, session->host->host_failed);
+		/*
+		 * force eh_abort() to unblock
+		 */
+		wake_up(&conn->ehwait);
+	}
+
+	/* now free crypto */
+	if (conn->hdrdgst_en || conn->datadgst_en) {
+		if (conn->tx_tfm)
+			crypto_free_tfm(conn->tx_tfm);
+		if (conn->rx_tfm)
+			crypto_free_tfm(conn->rx_tfm);
+		if (conn->data_tx_tfm)
+			crypto_free_tfm(conn->data_tx_tfm);
+		if (conn->data_rx_tfm)
+			crypto_free_tfm(conn->data_rx_tfm);
+	}
+
+	/* free conn->data, size = MaxRecvDataSegmentLength */
+	if (conn->data_size <= PAGE_SIZE)
+		kfree(conn->data);
+	else
+		free_pages((unsigned long)conn->data,
+					get_order(conn->data_size));
+
+	spin_lock_bh(&session->lock);
+	__kfifo_put(session->mgmtpool.queue, (void*)&conn->login_mtask,
+		    sizeof(void*));
+	list_del(&conn->item);
+	if (list_empty(&session->connections))
+		session->leadconn = NULL;
+	if (session->leadconn && session->leadconn == conn)
+		session->leadconn = container_of(session->connections.next,
+			struct iscsi_conn, item);
+
+	if (session->leadconn == NULL)
+		/* none connections exits.. reset sequencing */
+		session->cmdsn = session->max_cmdsn = session->exp_cmdsn = 1;
+	spin_unlock_bh(&session->lock);
+
+	kfifo_free(conn->xmitqueue);
+	kfifo_free(conn->writequeue);
+	kfifo_free(conn->immqueue);
+	kfifo_free(conn->mgmtqueue);
+	kfree(conn);
+}
+
+static int
+iscsi_conn_bind(iscsi_sessionh_t sessionh, iscsi_connh_t connh,
+		uint32_t transport_fd, int is_leading)
+{
+	struct iscsi_session *session = iscsi_ptr(sessionh);
+	struct iscsi_conn *tmp = ERR_PTR(-EEXIST), *conn = iscsi_ptr(connh);
+	struct sock *sk;
+	struct socket *sock;
+	int err;
+
+	/* lookup for existing socket */
+	sock = sockfd_lookup(transport_fd, &err);
+	if (!sock) {
+		printk(KERN_ERR "iscsi_tcp: sockfd_lookup failed %d\n", err);
+		return -EEXIST;
+	}
+
+	/* lookup for existing connection */
+	spin_lock_bh(&session->lock);
+	list_for_each_entry(tmp, &session->connections, item) {
+		if (tmp == conn) {
+			if (conn->c_stage != ISCSI_CONN_STOPPED ||
+			    conn->stop_stage == STOP_CONN_TERM) {
+				printk(KERN_ERR "iscsi_tcp: can't bind "
+				       "non-stopped connection (%d:%d)\n",
+				       conn->c_stage, conn->stop_stage);
+				spin_unlock_bh(&session->lock);
+				return -EIO;
+			}
+			break;
+		}
+	}
+	if (tmp != conn) {
+		/* bind new iSCSI connection to session */
+		conn->session = session;
+
+		list_add(&conn->item, &session->connections);
+	}
+	spin_unlock_bh(&session->lock);
+
+	if (conn->stop_stage != STOP_CONN_SUSPEND) {
+		/* bind iSCSI connection and socket */
+		conn->sock = sock;
+
+		/* setup Socket parameters */
+		sk = sock->sk;
+		sk->sk_reuse = 1;
+		sk->sk_sndtimeo = 15 * HZ; /* FIXME: make it configurable */
+		sk->sk_allocation = GFP_ATOMIC;
+
+		/* FIXME: disable Nagle's algorithm */
+
+		/*
+		 * Intercept TCP callbacks for sendfile like receive
+		 * processing.
+		 */
+		iscsi_conn_set_callbacks(conn);
+
+		/*
+		 * set receive state machine into initial state
+		 */
+		conn->in_progress = IN_PROGRESS_WAIT_HEADER;
+	}
+
+	if (is_leading)
+		session->leadconn = conn;
+
+	/*
+	 * Unblock xmitworker(), Login Phase will pass through.
+	 */
+	clear_bit(SUSPEND_BIT, &conn->suspend_rx);
+	clear_bit(SUSPEND_BIT, &conn->suspend_tx);
+
+	return 0;
+}
+
+static int
+iscsi_conn_start(iscsi_connh_t connh)
+{
+	struct iscsi_conn *conn = iscsi_ptr(connh);
+	struct iscsi_session *session = conn->session;
+	struct sock *sk;
+
+	/* FF phase warming up... */
+
+	if (session == NULL) {
+		printk(KERN_ERR "iscsi_tcp: can't start unbound connection\n");
+		return -EPERM;
+	}
+
+	sk = conn->sock->sk;
+
+	write_lock_bh(&sk->sk_callback_lock);
+	spin_lock_bh(&session->lock);
+	conn->c_stage = ISCSI_CONN_STARTED;
+	session->state = ISCSI_STATE_LOGGED_IN;
+
+	switch(conn->stop_stage) {
+	case STOP_CONN_RECOVER:
+		/*
+		 * unblock eh_abort() if it is blocked. re-try all
+		 * commands after successful recovery
+		 */
+		session->conn_cnt++;
+		conn->stop_stage = 0;
+		conn->tmabort_state = TMABORT_INITIAL;
+		session->age++;
+		wake_up(&conn->ehwait);
+		break;
+	case STOP_CONN_TERM:
+		session->conn_cnt++;
+		conn->stop_stage = 0;
+		break;
+	case STOP_CONN_SUSPEND:
+		conn->stop_stage = 0;
+		clear_bit(SUSPEND_BIT, &conn->suspend_rx);
+		clear_bit(SUSPEND_BIT, &conn->suspend_tx);
+		break;
+	default:
+		break;
+	}
+	spin_unlock_bh(&session->lock);
+	write_unlock_bh(&sk->sk_callback_lock);
+
+	return 0;
+}
+
+static void
+iscsi_conn_stop(iscsi_connh_t connh, int flag)
+{
+	struct iscsi_conn *conn = iscsi_ptr(connh);
+	struct iscsi_session *session = conn->session;
+	struct sock *sk;
+	unsigned long flags;
+
+	BUG_ON(!conn->sock);
+	sk = conn->sock->sk;
+	write_lock_bh(&sk->sk_callback_lock);
+	set_bit(SUSPEND_BIT, &conn->suspend_rx);
+	write_unlock_bh(&sk->sk_callback_lock);
+
+	down(&conn->xmitsema);
+
+	spin_lock_irqsave(session->host->host_lock, flags);
+	spin_lock(&session->lock);
+	conn->stop_stage = flag;
+	conn->c_stage = ISCSI_CONN_STOPPED;
+	set_bit(SUSPEND_BIT, &conn->suspend_tx);
+
+	if (flag != STOP_CONN_SUSPEND)
+		session->conn_cnt--;
+
+	if (session->conn_cnt == 0 || session->leadconn == conn)
+		session->state = ISCSI_STATE_FAILED;
+
+	spin_unlock(&session->lock);
+	spin_unlock_irqrestore(session->host->host_lock, flags);
+
+	if (flag == STOP_CONN_TERM || flag == STOP_CONN_RECOVER) {
+		struct iscsi_cmd_task *ctask;
+		struct iscsi_mgmt_task *mtask;
+
+		/*
+		 * Socket must go now.
+		 */
+		sock_hold(conn->sock->sk);
+		iscsi_conn_restore_callbacks(conn);
+		sock_put(conn->sock->sk);
+
+		/*
+		 * flush xmit queues.
+		 */
+		spin_lock_bh(&session->lock);
+		while (__kfifo_get(conn->writequeue, (void*)&ctask,
+			    sizeof(void*)) ||
+			__kfifo_get(conn->xmitqueue, (void*)&ctask,
+			    sizeof(void*))) {
+			struct iscsi_r2t_info *r2t;
+
+			/*
+			 * flush ctask's r2t queues
+			 */
+			while (__kfifo_get(ctask->r2tqueue, (void*)&r2t,
+				sizeof(void*)))
+				__kfifo_put(ctask->r2tpool.queue, (void*)&r2t,
+					    sizeof(void*));
+
+			spin_unlock_bh(&session->lock);
+			local_bh_disable();
+			iscsi_ctask_cleanup(conn, ctask);
+			local_bh_enable();
+			spin_lock_bh(&session->lock);
+		}
+		conn->ctask = NULL;
+		while (__kfifo_get(conn->immqueue, (void*)&mtask,
+			   sizeof(void*)) ||
+			__kfifo_get(conn->mgmtqueue, (void*)&mtask,
+			   sizeof(void*))) {
+			__kfifo_put(session->mgmtpool.queue,
+				    (void*)&mtask, sizeof(void*));
+		}
+		conn->mtask = NULL;
+		spin_unlock_bh(&session->lock);
+
+		/*
+		 * release socket only after we stopped data_xmit()
+		 * activity and flushed all outstandings
+		 */
+		sock_release(conn->sock);
+		conn->sock = NULL;
+
+		/*
+		 * for connection level recovery we should not calculate
+		 * header digest. conn->hdr_size used for optimization
+		 * in hdr_extract() and will be re-negotiated at
+		 * set_param() time.
+		 */
+		if (flag == STOP_CONN_RECOVER)
+			conn->hdr_size = sizeof(struct iscsi_hdr);
+	}
+	up(&conn->xmitsema);
+}
+
+static int
+iscsi_conn_send_generic(struct iscsi_conn *conn, struct iscsi_hdr *hdr,
+			char *data, uint32_t data_size)
+{
+	struct iscsi_session *session = conn->session;
+	struct iscsi_nopout *nop = (struct iscsi_nopout *)hdr;
+	struct iscsi_mgmt_task *mtask;
+
+	spin_lock_bh(&session->lock);
+	if (session->state == ISCSI_STATE_TERMINATE) {
+		spin_unlock_bh(&session->lock);
+		return -EPERM;
+	}
+	if (hdr->opcode == (ISCSI_OP_LOGIN | ISCSI_OP_IMMEDIATE) ||
+	    hdr->opcode == (ISCSI_OP_TEXT | ISCSI_OP_IMMEDIATE))
+		/*
+		 * Login and Text are sent serially, in
+		 * request-followed-by-response sequence.
+		 * Same mtask can be used. Same ITT must be used.
+		 * Note that login_mtask is preallocated at conn_create().
+		 */
+		mtask = conn->login_mtask;
+	else {
+	        BUG_ON(conn->c_stage == ISCSI_CONN_INITIAL_STAGE);
+	        BUG_ON(conn->c_stage == ISCSI_CONN_STOPPED);
+
+		if (!__kfifo_get(session->mgmtpool.queue,
+				 (void*)&mtask, sizeof(void*))) {
+			spin_unlock_bh(&session->lock);
+			return -ENOSPC;
+		}
+	}
+
+	/*
+	 * pre-format CmdSN and ExpStatSN for outgoing PDU.
+	 */
+	if (hdr->itt != cpu_to_be32(ISCSI_RESERVED_TAG)) {
+		hdr->itt = mtask->itt | (conn->id << CID_SHIFT) |
+			   (session->age << AGE_SHIFT);
+		nop->cmdsn = cpu_to_be32(session->cmdsn);
+		if (conn->c_stage == ISCSI_CONN_STARTED &&
+		    !(hdr->opcode & ISCSI_OP_IMMEDIATE))
+			session->cmdsn++;
+	} else
+		/* do not advance CmdSN */
+		nop->cmdsn = cpu_to_be32(session->cmdsn);
+
+	nop->exp_statsn = cpu_to_be32(conn->exp_statsn);
+
+	memcpy(&mtask->hdr, hdr, sizeof(struct iscsi_hdr));
+
+	if (conn->c_stage == ISCSI_CONN_INITIAL_STAGE ||
+	    conn->stop_stage == STOP_CONN_RECOVER)
+		iscsi_buf_init_virt(&mtask->headbuf, (char*)&mtask->hdr,
+				    sizeof(struct iscsi_hdr));
+	else
+		/* this will update header digest */
+		iscsi_buf_init_hdr(conn, &mtask->headbuf, (char*)&mtask->hdr,
+				    (u8 *)mtask->hdrext);
+
+	spin_unlock_bh(&session->lock);
+
+	if (data_size) {
+		memcpy(mtask->data, data, data_size);
+		mtask->data_count = data_size;
+	} else
+		mtask->data_count = 0;
+
+	mtask->xmstate = XMSTATE_IMM_HDR;
+
+	if (mtask->data_count) {
+		iscsi_buf_init_iov(&mtask->sendbuf, (char*)mtask->data,
+				    mtask->data_count);
+	}
+
+	debug_scsi("mgmtpdu [op 0x%x hdr->itt 0x%x datalen %d]\n",
+		   hdr->opcode, hdr->itt, data_size);
+
+	/*
+	 * since send_pdu() could be called at least from two contexts,
+	 * we need to serialize __kfifo_put, so we don't have to take
+	 * additional lock on fast data-path
+	 */
+        if (hdr->opcode & ISCSI_OP_IMMEDIATE)
+	        __kfifo_put(conn->immqueue, (void*)&mtask, sizeof(void*));
+	else
+	        __kfifo_put(conn->mgmtqueue, (void*)&mtask, sizeof(void*));
+
+	schedule_work(&conn->xmitwork);
+
+	return 0;
+}
+
+static int
+iscsi_eh_host_reset(struct scsi_cmnd *sc)
+{
+	struct iscsi_cmd_task *ctask = (struct iscsi_cmd_task *)sc->SCp.ptr;
+	struct iscsi_conn *conn = ctask->conn;
+	struct iscsi_session *session = conn->session;
+
+	spin_lock_bh(&session->lock);
+	if (session->state == ISCSI_STATE_TERMINATE) {
+		debug_scsi("failing host reset: session terminated "
+			   "[CID %d age %d]", conn->id, session->age);
+		spin_unlock_bh(&session->lock);
+		return FAILED;
+	}
+	spin_unlock_bh(&session->lock);
+
+	debug_scsi("failing connection CID %d due to SCSI host reset "
+		   "[itt 0x%x age %d]", conn->id, ctask->itt,
+		   session->age);
+	iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
+
+	return SUCCESS;
+}
+
+static void
+iscsi_tmabort_timedout(unsigned long data)
+{
+	struct iscsi_cmd_task *ctask = (struct iscsi_cmd_task *)data;
+	struct iscsi_conn *conn = ctask->conn;
+	struct iscsi_session *session = conn->session;
+
+	spin_lock(&session->lock);
+	if (conn->tmabort_state == TMABORT_INITIAL) {
+		__kfifo_put(session->mgmtpool.queue,
+				(void*)&ctask->mtask, sizeof(void*));
+		conn->tmabort_state = TMABORT_TIMEDOUT;
+		debug_scsi("tmabort timedout [sc %lx itt 0x%x]\n",
+			(long)ctask->sc, ctask->itt);
+		/* unblock eh_abort() */
+		wake_up(&conn->ehwait);
+	}
+	spin_unlock(&session->lock);
+}
+
+static int
+iscsi_eh_abort(struct scsi_cmnd *sc)
+{
+	int rc;
+	struct iscsi_cmd_task *ctask = (struct iscsi_cmd_task *)sc->SCp.ptr;
+	struct iscsi_conn *conn = ctask->conn;
+	struct iscsi_session *session = conn->session;
+
+	conn->eh_abort_cnt++;
+	debug_scsi("aborting [sc %lx itt 0x%x]\n", (long)sc, ctask->itt);
+
+	/*
+	 * two cases for ERL=0 here:
+	 *
+	 * 1) connection-level failure;
+	 * 2) recovery due protocol error;
+	 */
+	down(&conn->xmitsema);
+	spin_lock_bh(&session->lock);
+	if (session->state != ISCSI_STATE_LOGGED_IN) {
+		if (session->state == ISCSI_STATE_TERMINATE) {
+			spin_unlock_bh(&session->lock);
+			up(&conn->xmitsema);
+			goto failed;
+		}
+		spin_unlock_bh(&session->lock);
+	} else {
+		struct iscsi_tm *hdr = &conn->tmhdr;
+
+		/*
+		 * Still LOGGED_IN...
+		 */
+
+		if (!ctask->sc || sc->SCp.phase != session->age) {
+			/*
+			 * 1) ctask completed before time out. But session
+			 *    is still ok => Happy Retry.
+			 * 2) session was re-open during time out of ctask.
+			 */
+			spin_unlock_bh(&session->lock);
+			up(&conn->xmitsema);
+			goto success;
+		}
+		conn->tmabort_state = TMABORT_INITIAL;
+		spin_unlock_bh(&session->lock);
+
+		/*
+		 * ctask timed out but session is OK
+		 * ERL=0 requires task mgmt abort to be issued on each
+		 * failed command. requests must be serialized.
+		 */
+		memset(hdr, 0, sizeof(struct iscsi_tm));
+		hdr->opcode = ISCSI_OP_SCSI_TMFUNC | ISCSI_OP_IMMEDIATE;
+		hdr->flags = ISCSI_TM_FUNC_ABORT_TASK;
+		hdr->flags |= ISCSI_FLAG_CMD_FINAL;
+		memcpy(hdr->lun, ctask->hdr.lun, sizeof(hdr->lun));
+		hdr->rtt = ctask->hdr.itt;
+		hdr->refcmdsn = ctask->hdr.cmdsn;
+
+		rc = iscsi_conn_send_generic(conn, (struct iscsi_hdr *)hdr,
+					     NULL, 0);
+		if (rc) {
+			iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
+			debug_scsi("abort sent failure [itt 0x%x]", ctask->itt);
+		} else {
+			struct iscsi_r2t_info *r2t;
+
+			/*
+			 * TMF abort vs. TMF response race logic
+			 */
+			spin_lock_bh(&session->lock);
+			ctask->mtask = (struct iscsi_mgmt_task *)
+				session->mgmt_cmds[(hdr->itt & ITT_MASK) -
+							ISCSI_MGMT_ITT_OFFSET];
+			/*
+			 * have to flush r2tqueue to avoid r2t leaks
+			 */
+			while (__kfifo_get(ctask->r2tqueue, (void*)&r2t,
+				sizeof(void*))) {
+				__kfifo_put(ctask->r2tpool.queue, (void*)&r2t,
+					sizeof(void*));
+			}
+			if (conn->tmabort_state == TMABORT_INITIAL) {
+				conn->tmfcmd_pdus_cnt++;
+				conn->tmabort_timer.expires = 3*HZ + jiffies;
+				conn->tmabort_timer.function =
+						iscsi_tmabort_timedout;
+				conn->tmabort_timer.data = (unsigned long)ctask;
+				add_timer(&conn->tmabort_timer);
+				debug_scsi("abort sent [itt 0x%x]", ctask->itt);
+			} else {
+				if (!ctask->sc ||
+				    conn->tmabort_state == TMABORT_SUCCESS) {
+					conn->tmabort_state = TMABORT_INITIAL;
+					spin_unlock_bh(&session->lock);
+					up(&conn->xmitsema);
+					goto success;
+				}
+				conn->tmabort_state = TMABORT_INITIAL;
+				iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
+			}
+			spin_unlock_bh(&session->lock);
+		}
+	}
+	up(&conn->xmitsema);
+
+
+	/*
+	 * block eh thread until:
+	 *
+	 * 1) abort response;
+	 * 2) abort timeout;
+	 * 3) session re-opened;
+	 * 4) session terminated;
+	 */
+	for (;;) {
+		int p_state = session->state;
+
+		rc = wait_event_interruptible(conn->ehwait,
+			(p_state == ISCSI_STATE_LOGGED_IN ?
+			 (session->state == ISCSI_STATE_TERMINATE ||
+			  conn->tmabort_state != TMABORT_INITIAL) :
+			 (session->state == ISCSI_STATE_TERMINATE ||
+			  session->state == ISCSI_STATE_LOGGED_IN)));
+		if (rc) {
+			/* shutdown.. */
+			session->state = ISCSI_STATE_TERMINATE;
+			goto failed;
+		}
+
+		if (signal_pending(current))
+			flush_signals(current);
+
+		if (session->state == ISCSI_STATE_TERMINATE)
+			goto failed;
+
+		spin_lock_bh(&session->lock);
+		if (sc->SCp.phase == session->age &&
+		   (conn->tmabort_state == TMABORT_TIMEDOUT ||
+		    conn->tmabort_state == TMABORT_FAILED)) {
+			conn->tmabort_state = TMABORT_INITIAL;
+			if (!ctask->sc) {
+				/*
+				 * ctask completed before tmf abort response or
+				 * time out.
+				 * But session is still ok => Happy Retry.
+				 */
+				spin_unlock_bh(&session->lock);
+				break;
+			}
+			spin_unlock_bh(&session->lock);
+			iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
+			continue;
+		}
+		spin_unlock_bh(&session->lock);
+		break;
+	}
+
+success:
+	debug_scsi("abort success [sc %lx itt 0x%x]\n", (long)sc, ctask->itt);
+	rc = SUCCESS;
+	goto exit;
+
+failed:
+	debug_scsi("abort failed [sc %lx itt 0x%x]\n", (long)sc, ctask->itt);
+	rc = FAILED;
+
+exit:
+	del_timer_sync(&conn->tmabort_timer);
+
+	down(&conn->xmitsema);
+	if (conn->sock) {
+		struct sock *sk = conn->sock->sk;
+
+		write_lock_bh(&sk->sk_callback_lock);
+		iscsi_ctask_cleanup(conn, ctask);
+		write_unlock_bh(&sk->sk_callback_lock);
+	}
+	up(&conn->xmitsema);
+	return rc;
+}
+
+static int
+iscsi_r2tpool_alloc(struct iscsi_session *session)
+{
+	int i;
+	int cmd_i;
+
+	/*
+	 * initialize per-task: R2T pool and xmit queue
+	 */
+	for (cmd_i = 0; cmd_i < session->cmds_max; cmd_i++) {
+	        struct iscsi_cmd_task *ctask = session->cmds[cmd_i];
+
+		/*
+		 * pre-allocated x4 as much r2ts to handle race when
+		 * target acks DataOut faster than we data_xmit() queues
+		 * could replenish r2tqueue.
+		 */
+
+		/* R2T pool */
+		if (iscsi_pool_init(&ctask->r2tpool, session->max_r2t * 4,
+			(void***)&ctask->r2ts, sizeof(struct iscsi_r2t_info))) {
+			goto r2t_alloc_fail;
+		}
+
+		/* R2T xmit queue */
+		ctask->r2tqueue = kfifo_alloc(
+		      session->max_r2t * 4 * sizeof(void*), GFP_KERNEL, NULL);
+		if (ctask->r2tqueue == ERR_PTR(-ENOMEM)) {
+			iscsi_pool_free(&ctask->r2tpool, (void**)ctask->r2ts);
+			goto r2t_alloc_fail;
+		}
+
+		/*
+		 * number of
+		 * Data-Out PDU's within R2T-sequence can be quite big;
+		 * using mempool
+		 */
+		ctask->datapool = mempool_create(ISCSI_DTASK_DEFAULT_MAX,
+			 mempool_alloc_slab, mempool_free_slab, taskcache);
+		if (ctask->datapool == NULL) {
+			kfifo_free(ctask->r2tqueue);
+			iscsi_pool_free(&ctask->r2tpool, (void**)ctask->r2ts);
+			goto r2t_alloc_fail;
+		}
+		INIT_LIST_HEAD(&ctask->dataqueue);
+	}
+
+	return 0;
+
+r2t_alloc_fail:
+	for (i = 0; i < cmd_i; i++) {
+		mempool_destroy(session->cmds[i]->datapool);
+		kfifo_free(session->cmds[i]->r2tqueue);
+		iscsi_pool_free(&session->cmds[i]->r2tpool,
+				(void**)session->cmds[i]->r2ts);
+	}
+	return -ENOMEM;
+}
+
+static void
+iscsi_r2tpool_free(struct iscsi_session *session)
+{
+	int i;
+
+	for (i = 0; i < session->cmds_max; i++) {
+		mempool_destroy(session->cmds[i]->datapool);
+		kfifo_free(session->cmds[i]->r2tqueue);
+		iscsi_pool_free(&session->cmds[i]->r2tpool,
+				(void**)session->cmds[i]->r2ts);
+	}
+}
+
+static struct scsi_host_template iscsi_sht = {
+	.name			= "iSCSI Initiator over TCP/IP, v."
+				  ISCSI_VERSION_STR,
+        .queuecommand           = iscsi_queuecommand,
+	.can_queue		= ISCSI_XMIT_CMDS_MAX - 1,
+	.sg_tablesize		= ISCSI_SG_TABLESIZE,
+	.cmd_per_lun		= ISCSI_CMD_PER_LUN,
+        .eh_abort_handler       = iscsi_eh_abort,
+        .eh_host_reset_handler	= iscsi_eh_host_reset,
+        .use_clustering         = DISABLE_CLUSTERING,
+	.proc_name		= "iscsi_tcp",
+	.this_id		= -1,
+};
+
+static iscsi_sessionh_t
+iscsi_session_create(uint32_t initial_cmdsn, struct Scsi_Host *host)
+{
+	int cmd_i;
+	struct iscsi_session *session;
+
+	session = iscsi_hostdata(host->hostdata);
+	memset(session, 0, sizeof(struct iscsi_session));
+
+	session->host = host;
+	session->id = host->host_no;
+	session->state = ISCSI_STATE_LOGGED_IN;
+	session->mgmtpool_max = ISCSI_MGMT_CMDS_MAX;
+	session->cmds_max = ISCSI_XMIT_CMDS_MAX;
+	session->cmdsn = initial_cmdsn;
+	session->exp_cmdsn = initial_cmdsn + 1;
+	session->max_cmdsn = initial_cmdsn + 1;
+	session->max_r2t = 1;
+
+	/* initialize SCSI PDU commands pool */
+	if (iscsi_pool_init(&session->cmdpool, session->cmds_max,
+		(void***)&session->cmds, sizeof(struct iscsi_cmd_task)))
+		goto cmdpool_alloc_fail;
+
+	/* pre-format cmds pool with ITT */
+	for (cmd_i = 0; cmd_i < session->cmds_max; cmd_i++)
+		session->cmds[cmd_i]->itt = cmd_i;
+
+	spin_lock_init(&session->lock);
+	INIT_LIST_HEAD(&session->connections);
+
+	/* initialize immediate command pool */
+	if (iscsi_pool_init(&session->mgmtpool, session->mgmtpool_max,
+		(void***)&session->mgmt_cmds, sizeof(struct iscsi_mgmt_task)))
+		goto mgmtpool_alloc_fail;
+
+
+	/* pre-format immediate cmds pool with ITT */
+	for (cmd_i = 0; cmd_i < session->mgmtpool_max; cmd_i++) {
+		session->mgmt_cmds[cmd_i]->itt = ISCSI_MGMT_ITT_OFFSET + cmd_i;
+		session->mgmt_cmds[cmd_i]->data = kmalloc(
+			DEFAULT_MAX_RECV_DATA_SEGMENT_LENGTH, GFP_KERNEL);
+		if (!session->mgmt_cmds[cmd_i]->data) {
+			int j;
+
+			for (j = 0; j < cmd_i; j++)
+				kfree(session->mgmt_cmds[j]->data);
+			goto immdata_alloc_fail;
+		}
+	}
+
+	if (iscsi_r2tpool_alloc(session))
+		goto r2tpool_alloc_fail;
+
+	return iscsi_handle(session);
+
+r2tpool_alloc_fail:
+	for (cmd_i = 0; cmd_i < session->mgmtpool_max; cmd_i++)
+		kfree(session->mgmt_cmds[cmd_i]->data);
+	iscsi_pool_free(&session->mgmtpool, (void**)session->mgmt_cmds);
+immdata_alloc_fail:
+mgmtpool_alloc_fail:
+	iscsi_pool_free(&session->cmdpool, (void**)session->cmds);
+cmdpool_alloc_fail:
+	return iscsi_handle(NULL);
+}
+
+static void
+iscsi_session_destroy(iscsi_sessionh_t sessionh)
+{
+	int cmd_i;
+	struct iscsi_data_task *dtask, *n;
+	struct iscsi_session *session = iscsi_ptr(sessionh);
+
+	for (cmd_i = 0; cmd_i < session->cmds_max; cmd_i++) {
+		struct iscsi_cmd_task *ctask = session->cmds[cmd_i];
+		list_for_each_entry_safe(dtask, n, &ctask->dataqueue, item) {
+			list_del(&dtask->item);
+			mempool_free(dtask, ctask->datapool);
+		}
+	}
+
+	for (cmd_i = 0; cmd_i < session->mgmtpool_max; cmd_i++)
+		kfree(session->mgmt_cmds[cmd_i]->data);
+
+	iscsi_r2tpool_free(session);
+	iscsi_pool_free(&session->mgmtpool, (void**)session->mgmt_cmds);
+	iscsi_pool_free(&session->cmdpool, (void**)session->cmds);
+}
+
+static int
+iscsi_conn_set_param(iscsi_connh_t connh, enum iscsi_param param,
+		     uint32_t value)
+{
+	struct iscsi_conn *conn = iscsi_ptr(connh);
+	struct iscsi_session *session = conn->session;
+
+	spin_lock_bh(&session->lock);
+	if (conn->c_stage != ISCSI_CONN_INITIAL_STAGE &&
+	    conn->stop_stage != STOP_CONN_RECOVER) {
+		printk(KERN_ERR "iscsi_tcp: can not change parameter [%d]\n",
+		       param);
+		spin_unlock_bh(&session->lock);
+		return 0;
+	}
+	spin_unlock_bh(&session->lock);
+
+	switch(param) {
+	case ISCSI_PARAM_MAX_RECV_DLENGTH: {
+		char *saveptr = conn->data;
+		int flags = GFP_KERNEL;
+
+		if (conn->data_size >= value) {
+			conn->max_recv_dlength = value;
+			break;
+		}
+
+		spin_lock_bh(&session->lock);
+		if (conn->stop_stage == STOP_CONN_RECOVER)
+			flags = GFP_ATOMIC;
+		spin_unlock_bh(&session->lock);
+
+		if (value <= PAGE_SIZE)
+			conn->data = kmalloc(value, flags);
+		else
+			conn->data = (void*)__get_free_pages(flags,
+							     get_order(value));
+		if (conn->data == NULL) {
+			conn->data = saveptr;
+			return -ENOMEM;
+		}
+		if (conn->data_size <= PAGE_SIZE)
+			kfree(saveptr);
+		else
+			free_pages((unsigned long)saveptr,
+				   get_order(conn->data_size));
+		conn->max_recv_dlength = value;
+		conn->data_size = value;
+		}
+		break;
+	case ISCSI_PARAM_MAX_XMIT_DLENGTH:
+		conn->max_xmit_dlength =  value;
+		break;
+	case ISCSI_PARAM_HDRDGST_EN:
+		conn->hdrdgst_en = value;
+		conn->hdr_size = sizeof(struct iscsi_hdr);
+		if (conn->hdrdgst_en) {
+			conn->hdr_size += sizeof(__u32);
+			if (!conn->tx_tfm)
+				conn->tx_tfm = crypto_alloc_tfm("crc32c", 0);
+			if (!conn->tx_tfm)
+				return -ENOMEM;
+			if (!conn->rx_tfm)
+				conn->rx_tfm = crypto_alloc_tfm("crc32c", 0);
+			if (!conn->rx_tfm) {
+				crypto_free_tfm(conn->tx_tfm);
+				return -ENOMEM;
+			}
+		} else {
+			if (conn->tx_tfm)
+				crypto_free_tfm(conn->tx_tfm);
+			if (conn->rx_tfm)
+				crypto_free_tfm(conn->rx_tfm);
+		}
+		break;
+	case ISCSI_PARAM_DATADGST_EN:
+		conn->datadgst_en = value;
+		if (conn->datadgst_en) {
+			if (!conn->data_tx_tfm)
+				conn->data_tx_tfm =
+				    crypto_alloc_tfm("crc32c", 0);
+			if (!conn->data_tx_tfm)
+				return -ENOMEM;
+			if (!conn->data_rx_tfm)
+				conn->data_rx_tfm =
+				    crypto_alloc_tfm("crc32c", 0);
+			if (!conn->data_rx_tfm) {
+				crypto_free_tfm(conn->data_tx_tfm);
+				return -ENOMEM;
+			}
+		} else {
+			if (conn->data_tx_tfm)
+				crypto_free_tfm(conn->data_tx_tfm);
+			if (conn->data_rx_tfm)
+				crypto_free_tfm(conn->data_rx_tfm);
+		}
+		break;
+	case ISCSI_PARAM_INITIAL_R2T_EN:
+		session->initial_r2t_en = value;
+		break;
+	case ISCSI_PARAM_MAX_R2T:
+		if (session->max_r2t == roundup_pow_of_two(value))
+			break;
+		iscsi_r2tpool_free(session);
+		session->max_r2t = value;
+		if (session->max_r2t & (session->max_r2t - 1))
+			session->max_r2t = roundup_pow_of_two(session->max_r2t);
+		if (iscsi_r2tpool_alloc(session))
+			return -ENOMEM;
+		break;
+	case ISCSI_PARAM_IMM_DATA_EN:
+		session->imm_data_en = value;
+		break;
+	case ISCSI_PARAM_FIRST_BURST:
+		session->first_burst = value;
+		break;
+	case ISCSI_PARAM_MAX_BURST:
+		session->max_burst = value;
+		break;
+	case ISCSI_PARAM_PDU_INORDER_EN:
+		session->pdu_inorder_en = value;
+		break;
+	case ISCSI_PARAM_DATASEQ_INORDER_EN:
+		session->dataseq_inorder_en = value;
+		break;
+	case ISCSI_PARAM_ERL:
+		session->erl = value;
+		break;
+	case ISCSI_PARAM_IFMARKER_EN:
+		BUG_ON(value);
+		session->ifmarker_en = value;
+		break;
+	case ISCSI_PARAM_OFMARKER_EN:
+		BUG_ON(value);
+		session->ofmarker_en = value;
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int
+iscsi_conn_get_param(iscsi_connh_t connh, enum iscsi_param param,
+		     uint32_t *value)
+{
+	struct iscsi_conn *conn = iscsi_ptr(connh);
+	struct iscsi_session *session = conn->session;
+
+	switch(param) {
+	case ISCSI_PARAM_MAX_RECV_DLENGTH:
+		*value = conn->max_recv_dlength;
+		break;
+	case ISCSI_PARAM_MAX_XMIT_DLENGTH:
+		*value = conn->max_xmit_dlength;
+		break;
+	case ISCSI_PARAM_HDRDGST_EN:
+		*value = conn->hdrdgst_en;
+		break;
+	case ISCSI_PARAM_DATADGST_EN:
+		*value = conn->datadgst_en;
+		break;
+	case ISCSI_PARAM_INITIAL_R2T_EN:
+		*value = session->initial_r2t_en;
+		break;
+	case ISCSI_PARAM_MAX_R2T:
+		*value = session->max_r2t;
+		break;
+	case ISCSI_PARAM_IMM_DATA_EN:
+		*value = session->imm_data_en;
+		break;
+	case ISCSI_PARAM_FIRST_BURST:
+		*value = session->first_burst;
+		break;
+	case ISCSI_PARAM_MAX_BURST:
+		*value = session->max_burst;
+		break;
+	case ISCSI_PARAM_PDU_INORDER_EN:
+		*value = session->pdu_inorder_en;
+		break;
+	case ISCSI_PARAM_DATASEQ_INORDER_EN:
+		*value = session->dataseq_inorder_en;
+		break;
+	case ISCSI_PARAM_ERL:
+		*value = session->erl;
+		break;
+	case ISCSI_PARAM_IFMARKER_EN:
+		*value = session->ifmarker_en;
+		break;
+	case ISCSI_PARAM_OFMARKER_EN:
+		*value = session->ofmarker_en;
+		break;
+	default:
+		return ISCSI_ERR_PARAM_NOT_FOUND;
+	}
+
+	return 0;
+}
+
+static void
+iscsi_conn_get_stats(iscsi_connh_t connh, struct iscsi_stats *stats)
+{
+	struct iscsi_conn *conn = iscsi_ptr(connh);
+
+	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;
+	stats->r2t_pdus = conn->r2t_pdus_cnt;
+	stats->tmfcmd_pdus = conn->tmfcmd_pdus_cnt;
+	stats->tmfrsp_pdus = conn->tmfrsp_pdus_cnt;
+	stats->custom_length = 3;
+	strcpy(stats->custom[0].desc, "tx_sendpage_failures");
+	stats->custom[0].value = conn->sendpage_failures_cnt;
+	strcpy(stats->custom[1].desc, "rx_discontiguous_hdr");
+	stats->custom[1].value = conn->discontiguous_hdr_cnt;
+	strcpy(stats->custom[2].desc, "eh_abort_cnt");
+	stats->custom[2].value = conn->eh_abort_cnt;
+}
+
+static int
+iscsi_conn_send_pdu(iscsi_connh_t connh, struct iscsi_hdr *hdr, char *data,
+		    uint32_t data_size)
+{
+	struct iscsi_conn *conn = iscsi_ptr(connh);
+	int rc;
+
+	down(&conn->xmitsema);
+	rc = iscsi_conn_send_generic(conn, hdr, data, data_size);
+	up(&conn->xmitsema);
+
+	return rc;
+}
+
+static struct iscsi_transport iscsi_tcp_transport = {
+	.owner			= THIS_MODULE,
+	.name			= "tcp",
+	.caps			= CAP_RECOVERY_L0 | CAP_MULTI_R2T | CAP_HDRDGST
+				  | CAP_DATADGST,
+	.host_template		= &iscsi_sht,
+	.hostdata_size		= sizeof(struct iscsi_session),
+	.max_conn		= 1,
+	.max_cmd_len		= ISCSI_TCP_MAX_CMD_LEN,
+	.create_session		= iscsi_session_create,
+	.destroy_session	= iscsi_session_destroy,
+	.create_conn		= iscsi_conn_create,
+	.bind_conn		= iscsi_conn_bind,
+	.destroy_conn		= iscsi_conn_destroy,
+	.set_param		= iscsi_conn_set_param,
+	.get_param		= iscsi_conn_get_param,
+	.start_conn		= iscsi_conn_start,
+	.stop_conn		= iscsi_conn_stop,
+	.send_pdu		= iscsi_conn_send_pdu,
+	.get_stats		= iscsi_conn_get_stats,
+};
+
+static int __init
+iscsi_tcp_init(void)
+{
+	int error;
+
+	if (iscsi_max_lun < 1) {
+		printk(KERN_ERR "Invalid max_lun value of %u\n", iscsi_max_lun);
+		return -EINVAL;
+	}
+	iscsi_tcp_transport.max_lun = iscsi_max_lun;
+
+	taskcache = kmem_cache_create("iscsi_taskcache",
+			sizeof(struct iscsi_data_task), 0,
+			SLAB_HWCACHE_ALIGN | SLAB_NO_REAP, NULL, NULL);
+	if (!taskcache)
+		return -ENOMEM;
+
+	error = iscsi_register_transport(&iscsi_tcp_transport);
+	if (error)
+		kmem_cache_destroy(taskcache);
+
+	return error;
+}
+
+static void __exit
+iscsi_tcp_exit(void)
+{
+	iscsi_unregister_transport(&iscsi_tcp_transport);
+	kmem_cache_destroy(taskcache);
+}
+
+module_init(iscsi_tcp_init);
+module_exit(iscsi_tcp_exit);
diff --git a/drivers/scsi/iscsi_tcp.h b/drivers/scsi/iscsi_tcp.h
new file mode 100644
index 00000000000..d23ae68fae0
--- /dev/null
+++ b/drivers/scsi/iscsi_tcp.h
@@ -0,0 +1,322 @@
+/*
+ * iSCSI Initiator TCP Transport
+ * Copyright (C) 2004 Dmitry Yusupov
+ * Copyright (C) 2004 Alex Aizman
+ * Copyright (C) 2005 Mike Christie
+ * maintained by open-iscsi@googlegroups.com
+ *
+ * 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.
+ *
+ * See the file COPYING included with this distribution for more details.
+ */
+
+#ifndef ISCSI_TCP_H
+#define ISCSI_TCP_H
+
+/* Session's states */
+#define ISCSI_STATE_FREE		1
+#define ISCSI_STATE_LOGGED_IN		2
+#define ISCSI_STATE_FAILED		3
+#define ISCSI_STATE_TERMINATE		4
+
+/* Connection's states */
+#define ISCSI_CONN_INITIAL_STAGE	0
+#define ISCSI_CONN_STARTED		1
+#define ISCSI_CONN_STOPPED		2
+#define ISCSI_CONN_CLEANUP_WAIT		3
+
+/* Connection suspend "bit" */
+#define SUSPEND_BIT			1
+
+/* Socket's Receive state machine */
+#define IN_PROGRESS_WAIT_HEADER		0x0
+#define IN_PROGRESS_HEADER_GATHER	0x1
+#define IN_PROGRESS_DATA_RECV		0x2
+#define IN_PROGRESS_DDIGEST_RECV	0x3
+
+/* Task Mgmt states */
+#define	TMABORT_INITIAL			0x0
+#define	TMABORT_SUCCESS			0x1
+#define	TMABORT_FAILED			0x2
+#define	TMABORT_TIMEDOUT		0x3
+
+/* xmit state machine */
+#define	XMSTATE_IDLE			0x0
+#define	XMSTATE_R_HDR			0x1
+#define	XMSTATE_W_HDR			0x2
+#define	XMSTATE_IMM_HDR			0x4
+#define	XMSTATE_IMM_DATA		0x8
+#define	XMSTATE_UNS_INIT		0x10
+#define	XMSTATE_UNS_HDR			0x20
+#define	XMSTATE_UNS_DATA		0x40
+#define	XMSTATE_SOL_HDR			0x80
+#define	XMSTATE_SOL_DATA		0x100
+#define	XMSTATE_W_PAD			0x200
+#define XMSTATE_DATA_DIGEST		0x400
+
+#define ISCSI_CONN_MAX			1
+#define ISCSI_CONN_RCVBUF_MIN		262144
+#define ISCSI_CONN_SNDBUF_MIN		262144
+#define ISCSI_PAD_LEN			4
+#define ISCSI_R2T_MAX			16
+#define ISCSI_XMIT_CMDS_MAX		128	/* must be power of 2 */
+#define ISCSI_MGMT_CMDS_MAX		32	/* must be power of 2 */
+#define ISCSI_MGMT_ITT_OFFSET		0xa00
+#define ISCSI_SG_TABLESIZE		SG_ALL
+#define ISCSI_CMD_PER_LUN		128
+#define ISCSI_TCP_MAX_CMD_LEN		16
+
+#define ITT_MASK			(0xfff)
+#define CID_SHIFT			12
+#define CID_MASK			(0xffff<<CID_SHIFT)
+#define AGE_SHIFT			28
+#define AGE_MASK			(0xf<<AGE_SHIFT)
+
+struct iscsi_queue {
+	struct kfifo		*queue;		/* FIFO Queue */
+	void			**pool;		/* Pool of elements */
+	int			max;		/* Max number of elements */
+};
+
+struct iscsi_session;
+struct iscsi_cmd_task;
+struct iscsi_mgmt_task;
+
+/* Socket connection recieve helper */
+struct iscsi_tcp_recv {
+	struct iscsi_hdr	*hdr;
+	struct sk_buff		*skb;
+	int			offset;
+	int			len;
+	int			hdr_offset;
+	int			copy;
+	int			copied;
+	int			padding;
+	struct iscsi_cmd_task	*ctask;		/* current cmd in progress */
+
+	/* copied and flipped values */
+	int			opcode;
+	int			flags;
+	int			cmd_status;
+	int			ahslen;
+	int			datalen;
+	uint32_t		itt;
+	int			datadgst;
+};
+
+struct iscsi_conn {
+	struct iscsi_hdr	hdr;		/* header placeholder */
+	char			hdrext[4*sizeof(__u16) +
+				    sizeof(__u32)];
+	int			data_copied;
+	char			*data;		/* data placeholder */
+	struct socket           *sock;          /* TCP socket */
+	int			data_size;	/* actual recv_dlength */
+	int			stop_stage;	/* conn_stop() flag: *
+						 * stop to recover,  *
+						 * stop to terminate */
+	/* iSCSI connection-wide sequencing */
+	uint32_t		exp_statsn;
+	int			hdr_size;	/* PDU header size */
+	unsigned long		suspend_rx;	/* suspend Rx */
+
+	struct crypto_tfm	*rx_tfm;	/* CRC32C (Rx) */
+	struct crypto_tfm	*data_rx_tfm;	/* CRC32C (Rx) for data */
+
+	/* control data */
+	int			senselen;	/* scsi sense length */
+	int			id;		/* CID */
+	struct iscsi_tcp_recv	in;		/* TCP receive context */
+	struct iscsi_session	*session;	/* parent session */
+	struct list_head	item;		/* maintains list of conns */
+	int			in_progress;	/* connection state machine */
+	int			c_stage;	/* connection state */
+	struct iscsi_mgmt_task	*login_mtask;	/* mtask used for login/text */
+	struct iscsi_mgmt_task	*mtask;		/* xmit mtask in progress */
+	struct iscsi_cmd_task	*ctask;		/* xmit ctask in progress */
+	spinlock_t		lock;		/* FIXME: to be removed */
+
+	/* old values for socket callbacks */
+	void			(*old_data_ready)(struct sock *, int);
+	void			(*old_state_change)(struct sock *);
+	void			(*old_write_space)(struct sock *);
+
+	/* xmit */
+	struct crypto_tfm	*tx_tfm;	/* CRC32C (Tx) */
+	struct crypto_tfm	*data_tx_tfm;	/* CRC32C (Tx) for data */
+	struct kfifo		*writequeue;	/* write cmds for Data-Outs */
+	struct kfifo		*immqueue;	/* immediate xmit queue */
+	struct kfifo		*mgmtqueue;	/* mgmt (control) xmit queue */
+	struct kfifo		*xmitqueue;	/* data-path cmd queue */
+	struct work_struct	xmitwork;	/* per-conn. xmit workqueue */
+	struct semaphore	xmitsema;	/* serializes connection xmit,
+						 * access to kfifos:	  *
+						 * xmitqueue, writequeue, *
+						 * immqueue, mgmtqueue    */
+	unsigned long		suspend_tx;	/* suspend Tx */
+
+	/* abort */
+	wait_queue_head_t	ehwait;		/* used in eh_abort()     */
+	struct iscsi_tm		tmhdr;
+	struct timer_list	tmabort_timer;  /* abort timer */
+	int			tmabort_state;  /* see TMABORT_INITIAL, etc.*/
+
+	/* negotiated params */
+	int			max_recv_dlength;
+	int			max_xmit_dlength;
+	int			hdrdgst_en;
+	int			datadgst_en;
+
+	/* MIB-statistics */
+	uint64_t		txdata_octets;
+	uint64_t		rxdata_octets;
+	uint32_t		scsicmd_pdus_cnt;
+	uint32_t		dataout_pdus_cnt;
+	uint32_t		scsirsp_pdus_cnt;
+	uint32_t		datain_pdus_cnt;
+	uint32_t		r2t_pdus_cnt;
+	uint32_t		tmfcmd_pdus_cnt;
+	int32_t			tmfrsp_pdus_cnt;
+
+	/* custom statistics */
+	uint32_t		sendpage_failures_cnt;
+	uint32_t		discontiguous_hdr_cnt;
+	uint32_t		eh_abort_cnt;
+};
+
+struct iscsi_session {
+	/* iSCSI session-wide sequencing */
+	uint32_t		cmdsn;
+	uint32_t		exp_cmdsn;
+	uint32_t		max_cmdsn;
+
+	/* configuration */
+	int			initial_r2t_en;
+	int			max_r2t;
+	int			imm_data_en;
+	int			first_burst;
+	int			max_burst;
+	int			time2wait;
+	int			time2retain;
+	int			pdu_inorder_en;
+	int			dataseq_inorder_en;
+	int			erl;
+	int			ifmarker_en;
+	int			ofmarker_en;
+
+	/* control data */
+	struct Scsi_Host	*host;
+	int			id;
+	struct iscsi_conn	*leadconn;	/* leading connection */
+	spinlock_t		lock;		/* protects session state, *
+						 * sequence numbers,       *
+						 * session resources:      *
+						 * - cmdpool,		   *
+						 * - mgmtpool,		   *
+						 * - r2tpool		   */
+	int			state;		/* session state           */
+	struct list_head	item;
+	void			*auth_client;
+	int			conn_cnt;
+	int			age;		/* counts session re-opens */
+
+	struct list_head	connections;	/* list of connections */
+	int			cmds_max;	/* size of cmds array */
+	struct iscsi_cmd_task	**cmds;		/* Original Cmds arr */
+	struct iscsi_queue	cmdpool;	/* PDU's pool */
+	int			mgmtpool_max;	/* size of mgmt array */
+	struct iscsi_mgmt_task	**mgmt_cmds;	/* Original mgmt arr */
+	struct iscsi_queue	mgmtpool;	/* Mgmt PDU's pool */
+};
+
+struct iscsi_buf {
+	struct scatterlist	sg;
+	struct kvec		iov;
+	unsigned int		sent;
+};
+
+struct iscsi_data_task {
+	struct iscsi_data	hdr;			/* PDU */
+	char			hdrext[sizeof(__u32)];	/* Header-Digest */
+	struct list_head	item;			/* data queue item */
+	struct iscsi_buf	digestbuf;		/* digest buffer */
+	uint32_t		digest;			/* data digest */
+};
+#define ISCSI_DTASK_DEFAULT_MAX	ISCSI_SG_TABLESIZE * PAGE_SIZE / 512
+
+struct iscsi_mgmt_task {
+	struct iscsi_hdr	hdr;		/* mgmt. PDU */
+	char			hdrext[sizeof(__u32)];	/* Header-Digest */
+	char			*data;		/* mgmt payload */
+	int			xmstate;	/* mgmt xmit progress */
+	int			data_count;	/* counts data to be sent */
+	struct iscsi_buf	headbuf;	/* header buffer */
+	struct iscsi_buf	sendbuf;	/* in progress buffer */
+	int			sent;
+	uint32_t		itt;		/* this ITT */
+};
+
+struct iscsi_r2t_info {
+	__be32			ttt;		/* copied from R2T */
+	__be32			exp_statsn;	/* copied from R2T */
+	uint32_t		data_length;	/* copied from R2T */
+	uint32_t		data_offset;	/* copied from R2T */
+	struct iscsi_buf	headbuf;	/* Data-Out Header Buffer */
+	struct iscsi_buf	sendbuf;	/* Data-Out in progress buffer*/
+	int			sent;		/* R2T sequence progress */
+	int			data_count;	/* DATA-Out payload progress */
+	struct scatterlist	*sg;		/* per-R2T SG list */
+	int			solicit_datasn;
+	struct iscsi_data_task   *dtask;        /* which data task */
+};
+
+struct iscsi_cmd_task {
+	struct iscsi_cmd	hdr;			/* iSCSI PDU header */
+	char			hdrext[4*sizeof(__u16)+	/* AHS */
+				    sizeof(__u32)];	/* HeaderDigest */
+	char			pad[ISCSI_PAD_LEN];
+	int			itt;			/* this ITT */
+	int			datasn;			/* DataSN */
+	struct iscsi_buf	headbuf;		/* header buf (xmit) */
+	struct iscsi_buf	sendbuf;		/* in progress buffer*/
+	int			sent;
+	struct scatterlist	*sg;			/* per-cmd SG list  */
+	struct scatterlist	*bad_sg;		/* assert statement */
+	int			sg_count;		/* SG's to process  */
+	uint32_t		unsol_datasn;
+	uint32_t		exp_r2tsn;
+	int			xmstate;		/* xmit xtate machine */
+	int			imm_count;		/* imm-data (bytes)   */
+	int			unsol_count;		/* unsolicited (bytes)*/
+	int			r2t_data_count;		/* R2T Data-Out bytes */
+	int			data_count;		/* remaining Data-Out */
+	int			pad_count;		/* padded bytes */
+	struct scsi_cmnd	*sc;			/* associated SCSI cmd*/
+	int			total_length;
+	int			data_offset;
+	struct iscsi_conn	*conn;			/* used connection    */
+	struct iscsi_mgmt_task	*mtask;			/* tmf mtask in progr */
+
+	struct iscsi_r2t_info	*r2t;			/* in progress R2T    */
+	struct iscsi_queue	r2tpool;
+	struct kfifo		*r2tqueue;
+	struct iscsi_r2t_info	**r2ts;
+	struct list_head	dataqueue;		/* Data-Out dataqueue */
+	mempool_t		*datapool;
+	uint32_t		datadigest;		/* for recover digest */
+	int			digest_count;
+	uint32_t		immdigest;		/* for imm data */
+	struct iscsi_buf	immbuf;			/* for imm data digest */
+	struct iscsi_data_task   *dtask;		/* data task in progress*/
+	int			digest_offset;		/* for partial buff digest */
+};
+
+#endif /* ISCSI_H */
diff --git a/drivers/scsi/scsi_transport_iscsi.c b/drivers/scsi/scsi_transport_iscsi.c
index 8bb8222ea58..f189ce7541f 100644
--- a/drivers/scsi/scsi_transport_iscsi.c
+++ b/drivers/scsi/scsi_transport_iscsi.c
@@ -1,8 +1,10 @@
-/* 
+/*
  * iSCSI transport class definitions
  *
  * Copyright (C) IBM Corporation, 2004
- * Copyright (C) Mike Christie, 2004
+ * Copyright (C) Mike Christie, 2004 - 2005
+ * Copyright (C) Dmitry Yusupov, 2004 - 2005
+ * Copyright (C) Alex Aizman, 2004 - 2005
  *
  * 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
@@ -19,370 +21,1253 @@
  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
  */
 #include <linux/module.h>
+#include <linux/mempool.h>
+#include <net/tcp.h>
 #include <scsi/scsi.h>
 #include <scsi/scsi_host.h>
 #include <scsi/scsi_device.h>
 #include <scsi/scsi_transport.h>
 #include <scsi/scsi_transport_iscsi.h>
+#include <scsi/iscsi_if.h>
 
-#define ISCSI_SESSION_ATTRS 20
-#define ISCSI_HOST_ATTRS 2
+#define ISCSI_SESSION_ATTRS 8
+#define ISCSI_CONN_ATTRS 6
 
 struct iscsi_internal {
 	struct scsi_transport_template t;
-	struct iscsi_function_template *fnt;
+	struct iscsi_transport *iscsi_transport;
+	struct list_head list;
+	/*
+	 * List of sessions for this transport
+	 */
+	struct list_head sessions;
+	/*
+	 * lock to serialize access to the sessions list which must
+	 * be taken after the rx_queue_sema
+	 */
+	spinlock_t session_lock;
+	/*
+	 * based on transport capabilities, at register time we set these
+	 * bits to tell the transport class it wants attributes displayed
+	 * in sysfs or that it can support different iSCSI Data-Path
+	 * capabilities
+	 */
+	uint32_t param_mask;
+
+	struct class_device cdev;
 	/*
 	 * We do not have any private or other attrs.
 	 */
+	struct transport_container conn_cont;
+	struct class_device_attribute *conn_attrs[ISCSI_CONN_ATTRS + 1];
+	struct transport_container session_cont;
 	struct class_device_attribute *session_attrs[ISCSI_SESSION_ATTRS + 1];
-	struct class_device_attribute *host_attrs[ISCSI_HOST_ATTRS + 1];
 };
 
-#define to_iscsi_internal(tmpl) container_of(tmpl, struct iscsi_internal, t)
+/*
+ * list of registered transports and lock that must
+ * be held while accessing list. The iscsi_transport_lock must
+ * be acquired after the rx_queue_sema.
+ */
+static LIST_HEAD(iscsi_transports);
+static DEFINE_SPINLOCK(iscsi_transport_lock);
+
+#define to_iscsi_internal(tmpl) \
+	container_of(tmpl, struct iscsi_internal, t)
+
+#define cdev_to_iscsi_internal(_cdev) \
+	container_of(_cdev, struct iscsi_internal, cdev)
+
+static void iscsi_transport_release(struct class_device *cdev)
+{
+	struct iscsi_internal *priv = cdev_to_iscsi_internal(cdev);
+	kfree(priv);
+}
 
-static DECLARE_TRANSPORT_CLASS(iscsi_transport_class,
-			       "iscsi_transport",
+/*
+ * iscsi_transport_class represents the iscsi_transports that are
+ * registered.
+ */
+static struct class iscsi_transport_class = {
+	.name = "iscsi_transport",
+	.release = iscsi_transport_release,
+};
+
+static ssize_t
+show_transport_handle(struct class_device *cdev, char *buf)
+{
+	struct iscsi_internal *priv = cdev_to_iscsi_internal(cdev);
+	return sprintf(buf, "%llu", (unsigned long long)iscsi_handle(priv->iscsi_transport));
+}
+static CLASS_DEVICE_ATTR(handle, S_IRUGO, show_transport_handle, NULL);
+
+#define show_transport_attr(name, format)				\
+static ssize_t								\
+show_transport_##name(struct class_device *cdev, char *buf)		\
+{									\
+	struct iscsi_internal *priv = cdev_to_iscsi_internal(cdev);	\
+	return sprintf(buf, format"\n", priv->iscsi_transport->name);	\
+}									\
+static CLASS_DEVICE_ATTR(name, S_IRUGO, show_transport_##name, NULL);
+
+show_transport_attr(caps, "0x%x");
+show_transport_attr(max_lun, "%d");
+show_transport_attr(max_conn, "%d");
+show_transport_attr(max_cmd_len, "%d");
+
+static struct attribute *iscsi_transport_attrs[] = {
+	&class_device_attr_handle.attr,
+	&class_device_attr_caps.attr,
+	&class_device_attr_max_lun.attr,
+	&class_device_attr_max_conn.attr,
+	&class_device_attr_max_cmd_len.attr,
+	NULL,
+};
+
+static struct attribute_group iscsi_transport_group = {
+	.attrs = iscsi_transport_attrs,
+};
+
+static DECLARE_TRANSPORT_CLASS(iscsi_session_class,
+			       "iscsi_session",
 			       NULL,
 			       NULL,
 			       NULL);
 
-static DECLARE_TRANSPORT_CLASS(iscsi_host_class,
-			       "iscsi_host",
+static DECLARE_TRANSPORT_CLASS(iscsi_connection_class,
+			       "iscsi_connection",
 			       NULL,
 			       NULL,
 			       NULL);
+
+static struct sock *nls;
+static int daemon_pid;
+static DECLARE_MUTEX(rx_queue_sema);
+
+struct mempool_zone {
+	mempool_t *pool;
+	atomic_t allocated;
+	int size;
+	int hiwat;
+	struct list_head freequeue;
+	spinlock_t freelock;
+};
+
+static struct mempool_zone z_reply;
+
 /*
- * iSCSI target and session attrs
+ * Z_MAX_* - actual mempool size allocated at the mempool_zone_init() time
+ * Z_HIWAT_* - zone's high watermark when if_error bit will be set to -ENOMEM
+ *             so daemon will notice OOM on NETLINK tranposrt level and will
+ *             be able to predict or change operational behavior
  */
-#define iscsi_session_show_fn(field, format)				\
-									\
-static ssize_t								\
-show_session_##field(struct class_device *cdev, char *buf)		\
-{									\
-	struct scsi_target *starget = transport_class_to_starget(cdev);	\
-	struct Scsi_Host *shost = dev_to_shost(starget->dev.parent);	\
-	struct iscsi_internal *i = to_iscsi_internal(shost->transportt); \
-									\
-	if (i->fnt->get_##field)					\
-		i->fnt->get_##field(starget);				\
-	return snprintf(buf, 20, format"\n", iscsi_##field(starget));	\
+#define Z_MAX_REPLY	8
+#define Z_HIWAT_REPLY	6
+#define Z_MAX_PDU	8
+#define Z_HIWAT_PDU	6
+#define Z_MAX_ERROR	16
+#define Z_HIWAT_ERROR	12
+
+struct iscsi_if_conn {
+	struct list_head conn_list;	/* item in connlist */
+	struct list_head session_list;	/* item in session->connections */
+	iscsi_connh_t connh;
+	int active;			/* must be accessed with the connlock */
+	struct Scsi_Host *host;		/* originated shost */
+	struct device dev;		/* sysfs transport/container device */
+	struct iscsi_transport *transport;
+	struct mempool_zone z_error;
+	struct mempool_zone z_pdu;
+	struct list_head freequeue;
+};
+
+#define iscsi_dev_to_if_conn(_dev) \
+	container_of(_dev, struct iscsi_if_conn, dev)
+
+#define iscsi_cdev_to_if_conn(_cdev) \
+	iscsi_dev_to_if_conn(_cdev->dev)
+
+static LIST_HEAD(connlist);
+static DEFINE_SPINLOCK(connlock);
+
+struct iscsi_if_session {
+	struct list_head list;	/* item in session_list */
+	struct list_head connections;
+	iscsi_sessionh_t sessionh;
+	struct iscsi_transport *transport;
+	struct device dev;	/* sysfs transport/container device */
+};
+
+#define iscsi_dev_to_if_session(_dev) \
+	container_of(_dev, struct iscsi_if_session, dev)
+
+#define iscsi_cdev_to_if_session(_cdev) \
+	iscsi_dev_to_if_session(_cdev->dev)
+
+#define iscsi_if_session_to_shost(_session) \
+	dev_to_shost(_session->dev.parent)
+
+static struct iscsi_if_conn*
+iscsi_if_find_conn(uint64_t key)
+{
+	unsigned long flags;
+	struct iscsi_if_conn *conn;
+
+	spin_lock_irqsave(&connlock, flags);
+	list_for_each_entry(conn, &connlist, conn_list)
+		if (conn->connh == key) {
+			spin_unlock_irqrestore(&connlock, flags);
+			return conn;
+		}
+	spin_unlock_irqrestore(&connlock, flags);
+	return NULL;
 }
 
-#define iscsi_session_rd_attr(field, format)				\
-	iscsi_session_show_fn(field, format)				\
-static CLASS_DEVICE_ATTR(field, S_IRUGO, show_session_##field, NULL);
+static struct iscsi_internal *
+iscsi_if_transport_lookup(struct iscsi_transport *tt)
+{
+	struct iscsi_internal *priv;
+	unsigned long flags;
+
+	spin_lock_irqsave(&iscsi_transport_lock, flags);
+	list_for_each_entry(priv, &iscsi_transports, list) {
+		if (tt == priv->iscsi_transport) {
+			spin_unlock_irqrestore(&iscsi_transport_lock, flags);
+			return priv;
+		}
+	}
+	spin_unlock_irqrestore(&iscsi_transport_lock, flags);
+	return NULL;
+}
 
-iscsi_session_rd_attr(tpgt, "%hu");
-iscsi_session_rd_attr(tsih, "%2x");
-iscsi_session_rd_attr(max_recv_data_segment_len, "%u");
-iscsi_session_rd_attr(max_burst_len, "%u");
-iscsi_session_rd_attr(first_burst_len, "%u");
-iscsi_session_rd_attr(def_time2wait, "%hu");
-iscsi_session_rd_attr(def_time2retain, "%hu");
-iscsi_session_rd_attr(max_outstanding_r2t, "%hu");
-iscsi_session_rd_attr(erl, "%d");
+static inline struct list_head *skb_to_lh(struct sk_buff *skb)
+{
+	return (struct list_head *)&skb->cb;
+}
 
+static void*
+mempool_zone_alloc_skb(unsigned int gfp_mask, void *pool_data)
+{
+	struct mempool_zone *zone = pool_data;
 
-#define iscsi_session_show_bool_fn(field)				\
-									\
-static ssize_t								\
-show_session_bool_##field(struct class_device *cdev, char *buf)		\
-{									\
-	struct scsi_target *starget = transport_class_to_starget(cdev);	\
-	struct Scsi_Host *shost = dev_to_shost(starget->dev.parent);	\
-	struct iscsi_internal *i = to_iscsi_internal(shost->transportt); \
-									\
-	if (i->fnt->get_##field)					\
-		i->fnt->get_##field(starget);				\
-									\
-	if (iscsi_##field(starget))					\
-		return sprintf(buf, "Yes\n");				\
-	return sprintf(buf, "No\n");					\
+	return alloc_skb(zone->size, gfp_mask);
 }
 
-#define iscsi_session_rd_bool_attr(field)				\
-	iscsi_session_show_bool_fn(field)				\
-static CLASS_DEVICE_ATTR(field, S_IRUGO, show_session_bool_##field, NULL);
+static void
+mempool_zone_free_skb(void *element, void *pool_data)
+{
+	kfree_skb(element);
+}
 
-iscsi_session_rd_bool_attr(initial_r2t);
-iscsi_session_rd_bool_attr(immediate_data);
-iscsi_session_rd_bool_attr(data_pdu_in_order);
-iscsi_session_rd_bool_attr(data_sequence_in_order);
+static void
+mempool_zone_complete(struct mempool_zone *zone)
+{
+	unsigned long flags;
+	struct list_head *lh, *n;
 
-#define iscsi_session_show_digest_fn(field)				\
-									\
-static ssize_t								\
-show_##field(struct class_device *cdev, char *buf)			\
-{									\
-	struct scsi_target *starget = transport_class_to_starget(cdev);	\
-	struct Scsi_Host *shost = dev_to_shost(starget->dev.parent);	\
-	struct iscsi_internal *i = to_iscsi_internal(shost->transportt); \
-									\
-	if (i->fnt->get_##field)					\
-		i->fnt->get_##field(starget);				\
-									\
-	if (iscsi_##field(starget))					\
-		return sprintf(buf, "CRC32C\n");			\
-	return sprintf(buf, "None\n");					\
+	spin_lock_irqsave(&zone->freelock, flags);
+	list_for_each_safe(lh, n, &zone->freequeue) {
+		struct sk_buff *skb = (struct sk_buff *)((char *)lh -
+				offsetof(struct sk_buff, cb));
+		if (!skb_shared(skb)) {
+			list_del(skb_to_lh(skb));
+			mempool_free(skb, zone->pool);
+			atomic_dec(&zone->allocated);
+		}
+	}
+	spin_unlock_irqrestore(&zone->freelock, flags);
 }
 
-#define iscsi_session_rd_digest_attr(field)				\
-	iscsi_session_show_digest_fn(field)				\
-static CLASS_DEVICE_ATTR(field, S_IRUGO, show_##field, NULL);
+static int
+mempool_zone_init(struct mempool_zone *zp, unsigned max, unsigned size,
+		unsigned hiwat)
+{
+	zp->pool = mempool_create(max, mempool_zone_alloc_skb,
+				  mempool_zone_free_skb, zp);
+	if (!zp->pool)
+		return -ENOMEM;
 
-iscsi_session_rd_digest_attr(header_digest);
-iscsi_session_rd_digest_attr(data_digest);
+	zp->size = size;
+	zp->hiwat = hiwat;
 
-static ssize_t
-show_port(struct class_device *cdev, char *buf)
+	INIT_LIST_HEAD(&zp->freequeue);
+	spin_lock_init(&zp->freelock);
+	atomic_set(&zp->allocated, 0);
+
+	return 0;
+}
+
+
+static struct sk_buff*
+mempool_zone_get_skb(struct mempool_zone *zone)
 {
-	struct scsi_target *starget = transport_class_to_starget(cdev);
-	struct Scsi_Host *shost = dev_to_shost(starget->dev.parent);
-	struct iscsi_internal *i = to_iscsi_internal(shost->transportt);
+	struct sk_buff *skb;
+
+	skb = mempool_alloc(zone->pool, GFP_ATOMIC);
+	if (skb)
+		atomic_inc(&zone->allocated);
+	return skb;
+}
 
-	if (i->fnt->get_port)
-		i->fnt->get_port(starget);
+static int
+iscsi_unicast_skb(struct mempool_zone *zone, struct sk_buff *skb)
+{
+	unsigned long flags;
+	int rc;
 
-	return snprintf(buf, 20, "%hu\n", ntohs(iscsi_port(starget)));
+	skb_get(skb);
+	rc = netlink_unicast(nls, skb, daemon_pid, MSG_DONTWAIT);
+	if (rc < 0) {
+		mempool_free(skb, zone->pool);
+		printk(KERN_ERR "iscsi: can not unicast skb (%d)\n", rc);
+		return rc;
+	}
+
+	spin_lock_irqsave(&zone->freelock, flags);
+	list_add(skb_to_lh(skb), &zone->freequeue);
+	spin_unlock_irqrestore(&zone->freelock, flags);
+
+	return 0;
 }
-static CLASS_DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
 
-static ssize_t
-show_ip_address(struct class_device *cdev, char *buf)
+int iscsi_recv_pdu(iscsi_connh_t connh, struct iscsi_hdr *hdr,
+		   char *data, uint32_t data_size)
 {
-	struct scsi_target *starget = transport_class_to_starget(cdev);
-	struct Scsi_Host *shost = dev_to_shost(starget->dev.parent);
-	struct iscsi_internal *i = to_iscsi_internal(shost->transportt);
+	struct nlmsghdr	*nlh;
+	struct sk_buff *skb;
+	struct iscsi_uevent *ev;
+	struct iscsi_if_conn *conn;
+	char *pdu;
+	int len = NLMSG_SPACE(sizeof(*ev) + sizeof(struct iscsi_hdr) +
+			      data_size);
+
+	conn = iscsi_if_find_conn(connh);
+	BUG_ON(!conn);
+
+	mempool_zone_complete(&conn->z_pdu);
+
+	skb = mempool_zone_get_skb(&conn->z_pdu);
+	if (!skb) {
+		iscsi_conn_error(connh, ISCSI_ERR_CONN_FAILED);
+		printk(KERN_ERR "iscsi%d: can not deliver control PDU: OOM\n",
+		       conn->host->host_no);
+		return -ENOMEM;
+	}
 
-	if (i->fnt->get_ip_address)
-		i->fnt->get_ip_address(starget);
+	nlh = __nlmsg_put(skb, daemon_pid, 0, 0, (len - sizeof(*nlh)), 0);
+	ev = NLMSG_DATA(nlh);
+	memset(ev, 0, sizeof(*ev));
+	ev->transport_handle = iscsi_handle(conn->transport);
+	ev->type = ISCSI_KEVENT_RECV_PDU;
+	if (atomic_read(&conn->z_pdu.allocated) >= conn->z_pdu.hiwat)
+		ev->iferror = -ENOMEM;
+	ev->r.recv_req.conn_handle = connh;
+	pdu = (char*)ev + sizeof(*ev);
+	memcpy(pdu, hdr, sizeof(struct iscsi_hdr));
+	memcpy(pdu + sizeof(struct iscsi_hdr), data, data_size);
 
-	if (iscsi_addr_type(starget) == AF_INET)
-		return sprintf(buf, "%u.%u.%u.%u\n",
-			       NIPQUAD(iscsi_sin_addr(starget)));
-	else if(iscsi_addr_type(starget) == AF_INET6)
-		return sprintf(buf, "%04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x\n",
-			       NIP6(iscsi_sin6_addr(starget)));
-	return -EINVAL;
+	return iscsi_unicast_skb(&conn->z_pdu, skb);
 }
-static CLASS_DEVICE_ATTR(ip_address, S_IRUGO, show_ip_address, NULL);
+EXPORT_SYMBOL_GPL(iscsi_recv_pdu);
 
-static ssize_t
-show_isid(struct class_device *cdev, char *buf)
+void iscsi_conn_error(iscsi_connh_t connh, enum iscsi_err error)
 {
-	struct scsi_target *starget = transport_class_to_starget(cdev);
-	struct Scsi_Host *shost = dev_to_shost(starget->dev.parent);
-	struct iscsi_internal *i = to_iscsi_internal(shost->transportt);
+	struct nlmsghdr	*nlh;
+	struct sk_buff	*skb;
+	struct iscsi_uevent *ev;
+	struct iscsi_if_conn *conn;
+	int len = NLMSG_SPACE(sizeof(*ev));
+
+	conn = iscsi_if_find_conn(connh);
+	BUG_ON(!conn);
+
+	mempool_zone_complete(&conn->z_error);
+
+	skb = mempool_zone_get_skb(&conn->z_error);
+	if (!skb) {
+		printk(KERN_ERR "iscsi%d: gracefully ignored conn error (%d)\n",
+		       conn->host->host_no, error);
+		return;
+	}
+
+	nlh = __nlmsg_put(skb, daemon_pid, 0, 0, (len - sizeof(*nlh)), 0);
+	ev = NLMSG_DATA(nlh);
+	ev->transport_handle = iscsi_handle(conn->transport);
+	ev->type = ISCSI_KEVENT_CONN_ERROR;
+	if (atomic_read(&conn->z_error.allocated) >= conn->z_error.hiwat)
+		ev->iferror = -ENOMEM;
+	ev->r.connerror.error = error;
+	ev->r.connerror.conn_handle = connh;
 
-	if (i->fnt->get_isid)
-		i->fnt->get_isid(starget);
+	iscsi_unicast_skb(&conn->z_error, skb);
 
-	return sprintf(buf, "%02x%02x%02x%02x%02x%02x\n",
-		       iscsi_isid(starget)[0], iscsi_isid(starget)[1],
-		       iscsi_isid(starget)[2], iscsi_isid(starget)[3],
-		       iscsi_isid(starget)[4], iscsi_isid(starget)[5]);
+	printk(KERN_INFO "iscsi%d: detected conn error (%d)\n",
+	       conn->host->host_no, error);
+}
+EXPORT_SYMBOL_GPL(iscsi_conn_error);
+
+static int
+iscsi_if_send_reply(int pid, int seq, int type, int done, int multi,
+		      void *payload, int size)
+{
+	struct sk_buff	*skb;
+	struct nlmsghdr	*nlh;
+	int len = NLMSG_SPACE(size);
+	int flags = multi ? NLM_F_MULTI : 0;
+	int t = done ? NLMSG_DONE : type;
+
+	mempool_zone_complete(&z_reply);
+
+	skb = mempool_zone_get_skb(&z_reply);
+	/*
+	 * FIXME:
+	 * user is supposed to react on iferror == -ENOMEM;
+	 * see iscsi_if_rx().
+	 */
+	BUG_ON(!skb);
+
+	nlh = __nlmsg_put(skb, pid, seq, t, (len - sizeof(*nlh)), 0);
+	nlh->nlmsg_flags = flags;
+	memcpy(NLMSG_DATA(nlh), payload, size);
+	return iscsi_unicast_skb(&z_reply, skb);
 }
-static CLASS_DEVICE_ATTR(isid, S_IRUGO, show_isid, NULL);
 
 /*
- * This is used for iSCSI names. Normally, we follow
- * the transport class convention of having the lld
- * set the field, but in these cases the value is
- * too large.
+ * iSCSI Session's hostdata organization:
+ *
+ *    *------------------* <== host->hostdata
+ *    | transport        |
+ *    |------------------| <== iscsi_hostdata(host->hostdata)
+ *    | transport's data |
+ *    |------------------| <== hostdata_session(host->hostdata)
+ *    | interface's data |
+ *    *------------------*
  */
-#define iscsi_session_show_str_fn(field)				\
-									\
-static ssize_t								\
-show_session_str_##field(struct class_device *cdev, char *buf)		\
-{									\
-	ssize_t ret = 0;						\
-	struct scsi_target *starget = transport_class_to_starget(cdev);	\
-	struct Scsi_Host *shost = dev_to_shost(starget->dev.parent);	\
-	struct iscsi_internal *i = to_iscsi_internal(shost->transportt); \
-									\
-	if (i->fnt->get_##field)					\
-		ret = i->fnt->get_##field(starget, buf, PAGE_SIZE);	\
-	return ret;							\
+
+#define hostdata_privsize(_t)	(sizeof(unsigned long) + _t->hostdata_size + \
+				 _t->hostdata_size % sizeof(unsigned long) + \
+				 sizeof(struct iscsi_if_session))
+
+#define hostdata_session(_hostdata) ((void*)_hostdata + sizeof(unsigned long) + \
+			((struct iscsi_transport *) \
+			 iscsi_ptr(*(uint64_t *)_hostdata))->hostdata_size)
+
+static void iscsi_if_session_dev_release(struct device *dev)
+{
+	struct iscsi_if_session *session = iscsi_dev_to_if_session(dev);
+	struct iscsi_transport *transport = session->transport;
+	struct Scsi_Host *shost = iscsi_if_session_to_shost(session);
+	struct iscsi_if_conn *conn, *tmp;
+	unsigned long flags;
+
+	/* now free connections */
+	spin_lock_irqsave(&connlock, flags);
+	list_for_each_entry_safe(conn, tmp, &session->connections,
+				 session_list) {
+		list_del(&conn->session_list);
+		mempool_destroy(conn->z_pdu.pool);
+		mempool_destroy(conn->z_error.pool);
+		kfree(conn);
+	}
+	spin_unlock_irqrestore(&connlock, flags);
+	scsi_host_put(shost);
+	module_put(transport->owner);
+}
+
+static int
+iscsi_if_create_session(struct iscsi_internal *priv, struct iscsi_uevent *ev)
+{
+	struct iscsi_transport *transport = priv->iscsi_transport;
+	struct iscsi_if_session *session;
+	struct Scsi_Host *shost;
+	unsigned long flags;
+	int error;
+
+	if (!try_module_get(transport->owner))
+		return -EPERM;
+
+	shost = scsi_host_alloc(transport->host_template,
+				hostdata_privsize(transport));
+	if (!shost) {
+		ev->r.c_session_ret.session_handle = iscsi_handle(NULL);
+		printk(KERN_ERR "iscsi: can not allocate SCSI host for "
+		       "session\n");
+		error = -ENOMEM;
+		goto out_module_put;
+	}
+	shost->max_id = 1;
+	shost->max_channel = 0;
+	shost->max_lun = transport->max_lun;
+	shost->max_cmd_len = transport->max_cmd_len;
+	shost->transportt = &priv->t;
+
+	/* store struct iscsi_transport in hostdata */
+	*(uint64_t*)shost->hostdata = ev->transport_handle;
+
+	ev->r.c_session_ret.session_handle = transport->create_session(
+					ev->u.c_session.initial_cmdsn, shost);
+	if (ev->r.c_session_ret.session_handle == iscsi_handle(NULL)) {
+		error = 0;
+		goto out_host_put;
+	}
+
+	/* host_no becomes assigned SID */
+	ev->r.c_session_ret.sid = shost->host_no;
+	/* initialize session */
+	session = hostdata_session(shost->hostdata);
+	INIT_LIST_HEAD(&session->connections);
+	INIT_LIST_HEAD(&session->list);
+	session->sessionh = ev->r.c_session_ret.session_handle;
+	session->transport = transport;
+
+	error = scsi_add_host(shost, NULL);
+	if (error)
+		goto out_destroy_session;
+
+	/*
+	 * this is released in the dev's release function)
+	 */
+	scsi_host_get(shost);
+	snprintf(session->dev.bus_id, BUS_ID_SIZE, "session%u", shost->host_no);
+	session->dev.parent = &shost->shost_gendev;
+	session->dev.release = iscsi_if_session_dev_release;
+	error = device_register(&session->dev);
+	if (error) {
+		printk(KERN_ERR "iscsi: could not register session%d's dev\n",
+		       shost->host_no);
+		goto out_remove_host;
+	}
+	transport_register_device(&session->dev);
+
+	/* add this session to the list of active sessions */
+	spin_lock_irqsave(&priv->session_lock, flags);
+	list_add(&session->list, &priv->sessions);
+	spin_unlock_irqrestore(&priv->session_lock, flags);
+
+	return 0;
+
+out_remove_host:
+	scsi_remove_host(shost);
+out_destroy_session:
+	transport->destroy_session(ev->r.c_session_ret.session_handle);
+	ev->r.c_session_ret.session_handle = iscsi_handle(NULL);
+out_host_put:
+	scsi_host_put(shost);
+out_module_put:
+	module_put(transport->owner);
+	return error;
+}
+
+static int
+iscsi_if_destroy_session(struct iscsi_internal *priv, struct iscsi_uevent *ev)
+{
+	struct iscsi_transport *transport = priv->iscsi_transport;
+	struct Scsi_Host *shost;
+	struct iscsi_if_session *session;
+	unsigned long flags;
+	struct iscsi_if_conn *conn;
+	int error = 0;
+
+	shost = scsi_host_lookup(ev->u.d_session.sid);
+	if (shost == ERR_PTR(-ENXIO))
+		return -EEXIST;
+	session = hostdata_session(shost->hostdata);
+
+	/* check if we have active connections */
+	spin_lock_irqsave(&connlock, flags);
+	list_for_each_entry(conn, &session->connections, session_list) {
+		if (conn->active) {
+			printk(KERN_ERR "iscsi%d: can not destroy session: "
+			       "has active connection (%p)\n",
+			       shost->host_no, iscsi_ptr(conn->connh));
+			spin_unlock_irqrestore(&connlock, flags);
+			error = EIO;
+			goto out_release_ref;
+		}
+	}
+	spin_unlock_irqrestore(&connlock, flags);
+
+	scsi_remove_host(shost);
+	transport->destroy_session(ev->u.d_session.session_handle);
+	transport_unregister_device(&session->dev);
+	device_unregister(&session->dev);
+
+	/* remove this session from the list of active sessions */
+	spin_lock_irqsave(&priv->session_lock, flags);
+	list_del(&session->list);
+	spin_unlock_irqrestore(&priv->session_lock, flags);
+
+	/* ref from host alloc */
+	scsi_host_put(shost);
+out_release_ref:
+	/* ref from host lookup */
+	scsi_host_put(shost);
+	return error;
+}
+
+static void iscsi_if_conn_dev_release(struct device *dev)
+{
+	struct iscsi_if_conn *conn = iscsi_dev_to_if_conn(dev);
+	struct Scsi_Host *shost = conn->host;
+
+	scsi_host_put(shost);
+}
+
+static int
+iscsi_if_create_conn(struct iscsi_transport *transport, struct iscsi_uevent *ev)
+{
+	struct iscsi_if_session *session;
+	struct Scsi_Host *shost;
+	struct iscsi_if_conn *conn;
+	unsigned long flags;
+	int error;
+
+	shost = scsi_host_lookup(ev->u.c_conn.sid);
+	if (shost == ERR_PTR(-ENXIO))
+		return -EEXIST;
+	session = hostdata_session(shost->hostdata);
+
+	conn = kmalloc(sizeof(struct iscsi_if_conn), GFP_KERNEL);
+	if (!conn) {
+		error = -ENOMEM;
+		goto out_release_ref;
+	}
+	memset(conn, 0, sizeof(struct iscsi_if_conn));
+	INIT_LIST_HEAD(&conn->session_list);
+	INIT_LIST_HEAD(&conn->conn_list);
+	conn->host = shost;
+	conn->transport = transport;
+
+	error = mempool_zone_init(&conn->z_pdu, Z_MAX_PDU,
+			NLMSG_SPACE(sizeof(struct iscsi_uevent) +
+				    sizeof(struct iscsi_hdr) +
+				    DEFAULT_MAX_RECV_DATA_SEGMENT_LENGTH),
+			Z_HIWAT_PDU);
+	if (error) {
+		printk(KERN_ERR "iscsi%d: can not allocate pdu zone for new "
+		       "conn\n", shost->host_no);
+		goto out_free_conn;
+	}
+	error = mempool_zone_init(&conn->z_error, Z_MAX_ERROR,
+			NLMSG_SPACE(sizeof(struct iscsi_uevent)),
+			Z_HIWAT_ERROR);
+	if (error) {
+		printk(KERN_ERR "iscsi%d: can not allocate error zone for "
+		       "new conn\n", shost->host_no);
+		goto out_free_pdu_pool;
+	}
+
+	ev->r.handle = transport->create_conn(ev->u.c_conn.session_handle,
+					ev->u.c_conn.cid);
+	if (!ev->r.handle) {
+		error = -ENODEV;
+		goto out_free_error_pool;
+	}
+
+	conn->connh = ev->r.handle;
+
+	/*
+	 * this is released in the dev's release function
+	 */
+	if (!scsi_host_get(shost))
+		goto out_destroy_conn;
+	snprintf(conn->dev.bus_id, BUS_ID_SIZE, "connection%d:%u",
+		 shost->host_no, ev->u.c_conn.cid);
+	conn->dev.parent = &session->dev;
+	conn->dev.release = iscsi_if_conn_dev_release;
+	error = device_register(&conn->dev);
+	if (error) {
+		printk(KERN_ERR "iscsi%d: could not register connections%u "
+		       "dev\n", shost->host_no, ev->u.c_conn.cid);
+		goto out_release_parent_ref;
+	}
+	transport_register_device(&conn->dev);
+
+	spin_lock_irqsave(&connlock, flags);
+	list_add(&conn->conn_list, &connlist);
+	list_add(&conn->session_list, &session->connections);
+	conn->active = 1;
+	spin_unlock_irqrestore(&connlock, flags);
+
+	scsi_host_put(shost);
+	return 0;
+
+out_release_parent_ref:
+	scsi_host_put(shost);
+out_destroy_conn:
+	transport->destroy_conn(ev->r.handle);
+out_free_error_pool:
+	mempool_destroy(conn->z_error.pool);
+out_free_pdu_pool:
+	mempool_destroy(conn->z_pdu.pool);
+out_free_conn:
+	kfree(conn);
+out_release_ref:
+	scsi_host_put(shost);
+	return error;
 }
 
-#define iscsi_session_rd_str_attr(field)				\
-	iscsi_session_show_str_fn(field)				\
-static CLASS_DEVICE_ATTR(field, S_IRUGO, show_session_str_##field, NULL);
+static int
+iscsi_if_destroy_conn(struct iscsi_transport *transport, struct iscsi_uevent *ev)
+{
+	unsigned long flags;
+	struct iscsi_if_conn *conn;
+
+	conn = iscsi_if_find_conn(ev->u.d_conn.conn_handle);
+	if (!conn)
+		return -EEXIST;
+
+	transport->destroy_conn(ev->u.d_conn.conn_handle);
+
+	spin_lock_irqsave(&connlock, flags);
+	conn->active = 0;
+	list_del(&conn->conn_list);
+	spin_unlock_irqrestore(&connlock, flags);
+
+	transport_unregister_device(&conn->dev);
+	device_unregister(&conn->dev);
+	return 0;
+}
+
+static int
+iscsi_if_get_stats(struct iscsi_transport *transport, struct sk_buff *skb,
+		   struct nlmsghdr *nlh)
+{
+	struct iscsi_uevent *ev = NLMSG_DATA(nlh);
+	struct iscsi_stats *stats;
+	struct sk_buff *skbstat;
+	struct iscsi_if_conn *conn;
+	struct nlmsghdr	*nlhstat;
+	struct iscsi_uevent *evstat;
+	int len = NLMSG_SPACE(sizeof(*ev) +
+			      sizeof(struct iscsi_stats) +
+			      sizeof(struct iscsi_stats_custom) *
+			      ISCSI_STATS_CUSTOM_MAX);
+	int err = 0;
+
+	conn = iscsi_if_find_conn(ev->u.get_stats.conn_handle);
+	if (!conn)
+		return -EEXIST;
+
+	do {
+		int actual_size;
+
+		mempool_zone_complete(&conn->z_pdu);
+
+		skbstat = mempool_zone_get_skb(&conn->z_pdu);
+		if (!skbstat) {
+			printk(KERN_ERR "iscsi%d: can not deliver stats: OOM\n",
+			       conn->host->host_no);
+			return -ENOMEM;
+		}
+
+		nlhstat = __nlmsg_put(skbstat, daemon_pid, 0, 0,
+				      (len - sizeof(*nlhstat)), 0);
+		evstat = NLMSG_DATA(nlhstat);
+		memset(evstat, 0, sizeof(*evstat));
+		evstat->transport_handle = iscsi_handle(conn->transport);
+		evstat->type = nlh->nlmsg_type;
+		if (atomic_read(&conn->z_pdu.allocated) >= conn->z_pdu.hiwat)
+			evstat->iferror = -ENOMEM;
+		evstat->u.get_stats.conn_handle =
+			ev->u.get_stats.conn_handle;
+		stats = (struct iscsi_stats *)
+			((char*)evstat + sizeof(*evstat));
+		memset(stats, 0, sizeof(*stats));
+
+		transport->get_stats(ev->u.get_stats.conn_handle, stats);
+		actual_size = NLMSG_SPACE(sizeof(struct iscsi_uevent) +
+					  sizeof(struct iscsi_stats) +
+					  sizeof(struct iscsi_stats_custom) *
+					  stats->custom_length);
+		actual_size -= sizeof(*nlhstat);
+		actual_size = NLMSG_LENGTH(actual_size);
+		skb_trim(skb, NLMSG_ALIGN(actual_size));
+		nlhstat->nlmsg_len = actual_size;
+
+		err = iscsi_unicast_skb(&conn->z_pdu, skbstat);
+	} while (err < 0 && err != -ECONNREFUSED);
+
+	return err;
+}
+
+static int
+iscsi_if_recv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
+{
+	int err = 0;
+	struct iscsi_uevent *ev = NLMSG_DATA(nlh);
+	struct iscsi_transport *transport = NULL;
+	struct iscsi_internal *priv;
+
+	if (NETLINK_CREDS(skb)->uid)
+		return -EPERM;
+
+	priv = iscsi_if_transport_lookup(iscsi_ptr(ev->transport_handle));
+	if (!priv)
+		return -EINVAL;
+	transport = priv->iscsi_transport;
+
+	daemon_pid = NETLINK_CREDS(skb)->pid;
+
+	switch (nlh->nlmsg_type) {
+	case ISCSI_UEVENT_CREATE_SESSION:
+		err = iscsi_if_create_session(priv, ev);
+		break;
+	case ISCSI_UEVENT_DESTROY_SESSION:
+		err = iscsi_if_destroy_session(priv, ev);
+		break;
+	case ISCSI_UEVENT_CREATE_CONN:
+		err = iscsi_if_create_conn(transport, ev);
+		break;
+	case ISCSI_UEVENT_DESTROY_CONN:
+		err = iscsi_if_destroy_conn(transport, ev);
+		break;
+	case ISCSI_UEVENT_BIND_CONN:
+		if (!iscsi_if_find_conn(ev->u.b_conn.conn_handle))
+			return -EEXIST;
+		ev->r.retcode = transport->bind_conn(
+			ev->u.b_conn.session_handle,
+			ev->u.b_conn.conn_handle,
+			ev->u.b_conn.transport_fd,
+			ev->u.b_conn.is_leading);
+		break;
+	case ISCSI_UEVENT_SET_PARAM:
+		if (!iscsi_if_find_conn(ev->u.set_param.conn_handle))
+			return -EEXIST;
+		ev->r.retcode = transport->set_param(
+			ev->u.set_param.conn_handle,
+			ev->u.set_param.param, ev->u.set_param.value);
+		break;
+	case ISCSI_UEVENT_START_CONN:
+		if (!iscsi_if_find_conn(ev->u.start_conn.conn_handle))
+			return -EEXIST;
+		ev->r.retcode = transport->start_conn(
+			ev->u.start_conn.conn_handle);
+		break;
+	case ISCSI_UEVENT_STOP_CONN:
+		if (!iscsi_if_find_conn(ev->u.stop_conn.conn_handle))
+			return -EEXIST;
+		transport->stop_conn(ev->u.stop_conn.conn_handle,
+			ev->u.stop_conn.flag);
+		break;
+	case ISCSI_UEVENT_SEND_PDU:
+		if (!iscsi_if_find_conn(ev->u.send_pdu.conn_handle))
+			return -EEXIST;
+		ev->r.retcode = transport->send_pdu(
+		       ev->u.send_pdu.conn_handle,
+		       (struct iscsi_hdr*)((char*)ev + sizeof(*ev)),
+		       (char*)ev + sizeof(*ev) + ev->u.send_pdu.hdr_size,
+			ev->u.send_pdu.data_size);
+		break;
+	case ISCSI_UEVENT_GET_STATS:
+		err = iscsi_if_get_stats(transport, skb, nlh);
+		break;
+	default:
+		err = -EINVAL;
+		break;
+	}
+
+	return err;
+}
+
+/* Get message from skb (based on rtnetlink_rcv_skb).  Each message is
+ * processed by iscsi_if_recv_msg.  Malformed skbs with wrong length are
+ * discarded silently.  */
+static void
+iscsi_if_rx(struct sock *sk, int len)
+{
+	struct sk_buff *skb;
 
-iscsi_session_rd_str_attr(target_name);
-iscsi_session_rd_str_attr(target_alias);
+	down(&rx_queue_sema);
+	while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
+		while (skb->len >= NLMSG_SPACE(0)) {
+			int err;
+			uint32_t rlen;
+			struct nlmsghdr	*nlh;
+			struct iscsi_uevent *ev;
+
+			nlh = (struct nlmsghdr *)skb->data;
+			if (nlh->nlmsg_len < sizeof(*nlh) ||
+			    skb->len < nlh->nlmsg_len) {
+				break;
+			}
+			ev = NLMSG_DATA(nlh);
+			rlen = NLMSG_ALIGN(nlh->nlmsg_len);
+			if (rlen > skb->len)
+				rlen = skb->len;
+			err = iscsi_if_recv_msg(skb, nlh);
+			if (err) {
+				ev->type = ISCSI_KEVENT_IF_ERROR;
+				ev->iferror = err;
+			}
+			do {
+				/*
+				 * special case for GET_STATS:
+				 * on success - sending reply and stats from
+				 * inside of if_recv_msg(),
+				 * on error - fall through.
+				 */
+				if (ev->type == ISCSI_UEVENT_GET_STATS && !err)
+					break;
+				err = iscsi_if_send_reply(
+					NETLINK_CREDS(skb)->pid, nlh->nlmsg_seq,
+					nlh->nlmsg_type, 0, 0, ev, sizeof(*ev));
+				if (atomic_read(&z_reply.allocated) >=
+						z_reply.hiwat)
+					ev->iferror = -ENOMEM;
+			} while (err < 0 && err != -ECONNREFUSED);
+			skb_pull(skb, rlen);
+		}
+		kfree_skb(skb);
+	}
+	up(&rx_queue_sema);
+}
 
 /*
- * iSCSI host attrs
+ * iSCSI connection attrs
  */
+#define iscsi_conn_int_attr_show(param, format)				\
+static ssize_t								\
+show_conn_int_param_##param(struct class_device *cdev, char *buf)	\
+{									\
+	uint32_t value = 0;						\
+	struct iscsi_if_conn *conn = iscsi_cdev_to_if_conn(cdev);	\
+	struct iscsi_internal *priv;					\
+									\
+	priv = to_iscsi_internal(conn->host->transportt);		\
+	if (priv->param_mask & (1 << param))				\
+		priv->iscsi_transport->get_param(conn->connh, param, &value); \
+	return snprintf(buf, 20, format"\n", value);			\
+}
+
+#define iscsi_conn_int_attr(field, param, format)			\
+	iscsi_conn_int_attr_show(param, format)				\
+static CLASS_DEVICE_ATTR(field, S_IRUGO, show_conn_int_param_##param, NULL);
+
+iscsi_conn_int_attr(max_recv_dlength, ISCSI_PARAM_MAX_RECV_DLENGTH, "%u");
+iscsi_conn_int_attr(max_xmit_dlength, ISCSI_PARAM_MAX_XMIT_DLENGTH, "%u");
+iscsi_conn_int_attr(header_digest, ISCSI_PARAM_HDRDGST_EN, "%d");
+iscsi_conn_int_attr(data_digest, ISCSI_PARAM_DATADGST_EN, "%d");
+iscsi_conn_int_attr(ifmarker, ISCSI_PARAM_IFMARKER_EN, "%d");
+iscsi_conn_int_attr(ofmarker, ISCSI_PARAM_OFMARKER_EN, "%d");
 
 /*
- * Again, this is used for iSCSI names. Normally, we follow
- * the transport class convention of having the lld set
- * the field, but in these cases the value is too large.
+ * iSCSI session attrs
  */
-#define iscsi_host_show_str_fn(field)					\
-									\
+#define iscsi_session_int_attr_show(param, format)			\
 static ssize_t								\
-show_host_str_##field(struct class_device *cdev, char *buf)		\
+show_session_int_param_##param(struct class_device *cdev, char *buf)	\
 {									\
-	int ret = 0;							\
-	struct Scsi_Host *shost = transport_class_to_shost(cdev);	\
-	struct iscsi_internal *i = to_iscsi_internal(shost->transportt); \
+	uint32_t value = 0;						\
+	struct iscsi_if_session *session = iscsi_cdev_to_if_session(cdev); \
+	struct Scsi_Host *shost = iscsi_if_session_to_shost(session);	\
+	struct iscsi_internal *priv = to_iscsi_internal(shost->transportt); \
+	struct iscsi_if_conn *conn = NULL;				\
+	unsigned long  flags;						\
+									\
+	spin_lock_irqsave(&connlock, flags);				\
+	if (!list_empty(&session->connections))				\
+		conn = list_entry(session->connections.next,		\
+				  struct iscsi_if_conn, session_list);	\
+	spin_unlock_irqrestore(&connlock, flags);			\
 									\
-	if (i->fnt->get_##field)					\
-		ret = i->fnt->get_##field(shost, buf, PAGE_SIZE);	\
-	return ret;							\
+	if (conn && (priv->param_mask & (1 << param)))			\
+		priv->iscsi_transport->get_param(conn->connh, param, &value);\
+	return snprintf(buf, 20, format"\n", value);			\
 }
 
-#define iscsi_host_rd_str_attr(field)					\
-	iscsi_host_show_str_fn(field)					\
-static CLASS_DEVICE_ATTR(field, S_IRUGO, show_host_str_##field, NULL);
+#define iscsi_session_int_attr(field, param, format)			\
+	iscsi_session_int_attr_show(param, format)			\
+static CLASS_DEVICE_ATTR(field, S_IRUGO, show_session_int_param_##param, NULL);
 
-iscsi_host_rd_str_attr(initiator_name);
-iscsi_host_rd_str_attr(initiator_alias);
+iscsi_session_int_attr(initial_r2t, ISCSI_PARAM_INITIAL_R2T_EN, "%d");
+iscsi_session_int_attr(max_outstanding_r2t, ISCSI_PARAM_MAX_R2T, "%hu");
+iscsi_session_int_attr(immediate_data, ISCSI_PARAM_IMM_DATA_EN, "%d");
+iscsi_session_int_attr(first_burst_len, ISCSI_PARAM_FIRST_BURST, "%u");
+iscsi_session_int_attr(max_burst_len, ISCSI_PARAM_MAX_BURST, "%u");
+iscsi_session_int_attr(data_pdu_in_order, ISCSI_PARAM_PDU_INORDER_EN, "%d");
+iscsi_session_int_attr(data_seq_in_order, ISCSI_PARAM_DATASEQ_INORDER_EN, "%d");
+iscsi_session_int_attr(erl, ISCSI_PARAM_ERL, "%d");
 
-#define SETUP_SESSION_RD_ATTR(field)					\
-	if (i->fnt->show_##field) {					\
-		i->session_attrs[count] = &class_device_attr_##field;	\
+#define SETUP_SESSION_RD_ATTR(field, param)				\
+	if (priv->param_mask & (1 << param)) {				\
+		priv->session_attrs[count] = &class_device_attr_##field;\
 		count++;						\
 	}
 
-#define SETUP_HOST_RD_ATTR(field)					\
-	if (i->fnt->show_##field) {					\
-		i->host_attrs[count] = &class_device_attr_##field;	\
+#define SETUP_CONN_RD_ATTR(field, param)				\
+	if (priv->param_mask & (1 << param)) {				\
+		priv->conn_attrs[count] = &class_device_attr_##field;	\
 		count++;						\
 	}
 
-static int iscsi_host_match(struct attribute_container *cont,
-			  struct device *dev)
+static int iscsi_is_session_dev(const struct device *dev)
+{
+	return dev->release == iscsi_if_session_dev_release;
+}
+
+static int iscsi_session_match(struct attribute_container *cont,
+			   struct device *dev)
 {
+	struct iscsi_if_session *session;
 	struct Scsi_Host *shost;
-	struct iscsi_internal *i;
+	struct iscsi_internal *priv;
+
+	if (!iscsi_is_session_dev(dev))
+		return 0;
 
-	if (!scsi_is_host_device(dev))
+	session = iscsi_dev_to_if_session(dev);
+	shost = iscsi_if_session_to_shost(session);
+	if (!shost->transportt)
 		return 0;
 
-	shost = dev_to_shost(dev);
-	if (!shost->transportt  || shost->transportt->host_attrs.ac.class
-	    != &iscsi_host_class.class)
+	priv = to_iscsi_internal(shost->transportt);
+	if (priv->session_cont.ac.class != &iscsi_session_class.class)
 		return 0;
 
-	i = to_iscsi_internal(shost->transportt);
-	
-	return &i->t.host_attrs.ac == cont;
+	return &priv->session_cont.ac == cont;
 }
 
-static int iscsi_target_match(struct attribute_container *cont,
-			    struct device *dev)
+static int iscsi_is_conn_dev(const struct device *dev)
 {
+	return dev->release == iscsi_if_conn_dev_release;
+}
+
+static int iscsi_conn_match(struct attribute_container *cont,
+			   struct device *dev)
+{
+	struct iscsi_if_conn *conn;
 	struct Scsi_Host *shost;
-	struct iscsi_internal *i;
+	struct iscsi_internal *priv;
 
-	if (!scsi_is_target_device(dev))
+	if (!iscsi_is_conn_dev(dev))
 		return 0;
 
-	shost = dev_to_shost(dev->parent);
-	if (!shost->transportt  || shost->transportt->host_attrs.ac.class
-	    != &iscsi_host_class.class)
+	conn = iscsi_dev_to_if_conn(dev);
+	shost = conn->host;
+	if (!shost->transportt)
 		return 0;
 
-	i = to_iscsi_internal(shost->transportt);
-	
-	return &i->t.target_attrs.ac == cont;
-}
-
-struct scsi_transport_template *
-iscsi_attach_transport(struct iscsi_function_template *fnt)
-{
-	struct iscsi_internal *i = kmalloc(sizeof(struct iscsi_internal),
-					   GFP_KERNEL);
-	int count = 0;
-
-	if (unlikely(!i))
-		return NULL;
-
-	memset(i, 0, sizeof(struct iscsi_internal));
-	i->fnt = fnt;
-
-	i->t.target_attrs.ac.attrs = &i->session_attrs[0];
-	i->t.target_attrs.ac.class = &iscsi_transport_class.class;
-	i->t.target_attrs.ac.match = iscsi_target_match;
-	transport_container_register(&i->t.target_attrs);
-	i->t.target_size = sizeof(struct iscsi_class_session);
-
-	SETUP_SESSION_RD_ATTR(tsih);
-	SETUP_SESSION_RD_ATTR(isid);
-	SETUP_SESSION_RD_ATTR(header_digest);
-	SETUP_SESSION_RD_ATTR(data_digest);
-	SETUP_SESSION_RD_ATTR(target_name);
-	SETUP_SESSION_RD_ATTR(target_alias);
-	SETUP_SESSION_RD_ATTR(port);
-	SETUP_SESSION_RD_ATTR(tpgt);
-	SETUP_SESSION_RD_ATTR(ip_address);
-	SETUP_SESSION_RD_ATTR(initial_r2t);
-	SETUP_SESSION_RD_ATTR(immediate_data);
-	SETUP_SESSION_RD_ATTR(max_recv_data_segment_len);
-	SETUP_SESSION_RD_ATTR(max_burst_len);
-	SETUP_SESSION_RD_ATTR(first_burst_len);
-	SETUP_SESSION_RD_ATTR(def_time2wait);
-	SETUP_SESSION_RD_ATTR(def_time2retain);
-	SETUP_SESSION_RD_ATTR(max_outstanding_r2t);
-	SETUP_SESSION_RD_ATTR(data_pdu_in_order);
-	SETUP_SESSION_RD_ATTR(data_sequence_in_order);
-	SETUP_SESSION_RD_ATTR(erl);
+	priv = to_iscsi_internal(shost->transportt);
+	if (priv->conn_cont.ac.class != &iscsi_connection_class.class)
+		return 0;
 
-	BUG_ON(count > ISCSI_SESSION_ATTRS);
-	i->session_attrs[count] = NULL;
+	return &priv->conn_cont.ac == cont;
+}
+
+int iscsi_register_transport(struct iscsi_transport *tt)
+{
+	struct iscsi_internal *priv;
+	unsigned long flags;
+	int count = 0, err;
+
+	BUG_ON(!tt);
+
+	priv = iscsi_if_transport_lookup(tt);
+	if (priv)
+		return -EEXIST;
+
+	priv = kmalloc(sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		return -ENOMEM;
+	memset(priv, 0, sizeof(*priv));
+	INIT_LIST_HEAD(&priv->list);
+	INIT_LIST_HEAD(&priv->sessions);
+	spin_lock_init(&priv->session_lock);
+	priv->iscsi_transport = tt;
+
+	priv->cdev.class = &iscsi_transport_class;
+	snprintf(priv->cdev.class_id, BUS_ID_SIZE, "%s", tt->name);
+	err = class_device_register(&priv->cdev);
+	if (err)
+		goto free_priv;
+
+	err = sysfs_create_group(&priv->cdev.kobj, &iscsi_transport_group);
+	if (err)
+		goto unregister_cdev;
+
+	/* setup parameters mask */
+	priv->param_mask = 0xFFFFFFFF;
+	if (!(tt->caps & CAP_MULTI_R2T))
+		priv->param_mask &= ~(1 << ISCSI_PARAM_MAX_R2T);
+	if (!(tt->caps & CAP_HDRDGST))
+		priv->param_mask &= ~(1 << ISCSI_PARAM_HDRDGST_EN);
+	if (!(tt->caps & CAP_DATADGST))
+		priv->param_mask &= ~(1 << ISCSI_PARAM_DATADGST_EN);
+	if (!(tt->caps & CAP_MARKERS)) {
+		priv->param_mask &= ~(1 << ISCSI_PARAM_IFMARKER_EN);
+		priv->param_mask &= ~(1 << ISCSI_PARAM_OFMARKER_EN);
+	}
 
-	i->t.host_attrs.ac.attrs = &i->host_attrs[0];
-	i->t.host_attrs.ac.class = &iscsi_host_class.class;
-	i->t.host_attrs.ac.match = iscsi_host_match;
-	transport_container_register(&i->t.host_attrs);
-	i->t.host_size = 0;
+	/* connection parameters */
+	priv->conn_cont.ac.attrs = &priv->conn_attrs[0];
+	priv->conn_cont.ac.class = &iscsi_connection_class.class;
+	priv->conn_cont.ac.match = iscsi_conn_match;
+	transport_container_register(&priv->conn_cont);
 
+	SETUP_CONN_RD_ATTR(max_recv_dlength, ISCSI_PARAM_MAX_RECV_DLENGTH);
+	SETUP_CONN_RD_ATTR(max_xmit_dlength, ISCSI_PARAM_MAX_XMIT_DLENGTH);
+	SETUP_CONN_RD_ATTR(header_digest, ISCSI_PARAM_HDRDGST_EN);
+	SETUP_CONN_RD_ATTR(data_digest, ISCSI_PARAM_DATADGST_EN);
+	SETUP_CONN_RD_ATTR(ifmarker, ISCSI_PARAM_IFMARKER_EN);
+	SETUP_CONN_RD_ATTR(ofmarker, ISCSI_PARAM_OFMARKER_EN);
+
+	BUG_ON(count > ISCSI_CONN_ATTRS);
+	priv->conn_attrs[count] = NULL;
 	count = 0;
-	SETUP_HOST_RD_ATTR(initiator_name);
-	SETUP_HOST_RD_ATTR(initiator_alias);
 
-	BUG_ON(count > ISCSI_HOST_ATTRS);
-	i->host_attrs[count] = NULL;
+	/* session parameters */
+	priv->session_cont.ac.attrs = &priv->session_attrs[0];
+	priv->session_cont.ac.class = &iscsi_session_class.class;
+	priv->session_cont.ac.match = iscsi_session_match;
+	transport_container_register(&priv->session_cont);
+
+	SETUP_SESSION_RD_ATTR(initial_r2t, ISCSI_PARAM_INITIAL_R2T_EN);
+	SETUP_SESSION_RD_ATTR(max_outstanding_r2t, ISCSI_PARAM_MAX_R2T);
+	SETUP_SESSION_RD_ATTR(immediate_data, ISCSI_PARAM_IMM_DATA_EN);
+	SETUP_SESSION_RD_ATTR(first_burst_len, ISCSI_PARAM_FIRST_BURST);
+	SETUP_SESSION_RD_ATTR(max_burst_len, ISCSI_PARAM_MAX_BURST);
+	SETUP_SESSION_RD_ATTR(data_pdu_in_order, ISCSI_PARAM_PDU_INORDER_EN);
+	SETUP_SESSION_RD_ATTR(data_seq_in_order,ISCSI_PARAM_DATASEQ_INORDER_EN)
+	SETUP_SESSION_RD_ATTR(erl, ISCSI_PARAM_ERL);
+
+	BUG_ON(count > ISCSI_SESSION_ATTRS);
+	priv->session_attrs[count] = NULL;
+
+	spin_lock_irqsave(&iscsi_transport_lock, flags);
+	list_add(&priv->list, &iscsi_transports);
+	spin_unlock_irqrestore(&iscsi_transport_lock, flags);
+
+	printk(KERN_NOTICE "iscsi: registered transport (%s)\n", tt->name);
+	return 0;
 
-	return &i->t;
+unregister_cdev:
+	class_device_unregister(&priv->cdev);
+free_priv:
+	kfree(priv);
+	return err;
 }
+EXPORT_SYMBOL_GPL(iscsi_register_transport);
+
+int iscsi_unregister_transport(struct iscsi_transport *tt)
+{
+	struct iscsi_internal *priv;
+	unsigned long flags;
+
+	BUG_ON(!tt);
+
+	down(&rx_queue_sema);
+
+	priv = iscsi_if_transport_lookup(tt);
+	BUG_ON (!priv);
+
+	spin_lock_irqsave(&priv->session_lock, flags);
+	if (!list_empty(&priv->sessions)) {
+		spin_unlock_irqrestore(&priv->session_lock, flags);
+		up(&rx_queue_sema);
+		return -EPERM;
+	}
+	spin_unlock_irqrestore(&priv->session_lock, flags);
+
+	spin_lock_irqsave(&iscsi_transport_lock, flags);
+	list_del(&priv->list);
+	spin_unlock_irqrestore(&iscsi_transport_lock, flags);
+
+	transport_container_unregister(&priv->conn_cont);
+	transport_container_unregister(&priv->session_cont);
+
+	sysfs_remove_group(&priv->cdev.kobj, &iscsi_transport_group);
+	class_device_unregister(&priv->cdev);
+	up(&rx_queue_sema);
 
-EXPORT_SYMBOL(iscsi_attach_transport);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(iscsi_unregister_transport);
 
-void iscsi_release_transport(struct scsi_transport_template *t)
+static int
+iscsi_rcv_nl_event(struct notifier_block *this, unsigned long event, void *ptr)
 {
-	struct iscsi_internal *i = to_iscsi_internal(t);
+	struct netlink_notify *n = ptr;
+
+	if (event == NETLINK_URELEASE &&
+	    n->protocol == NETLINK_ISCSI && n->pid) {
+		struct iscsi_if_conn *conn;
+		unsigned long flags;
+
+		mempool_zone_complete(&z_reply);
+		spin_lock_irqsave(&connlock, flags);
+		list_for_each_entry(conn, &connlist, conn_list) {
+			mempool_zone_complete(&conn->z_error);
+			mempool_zone_complete(&conn->z_pdu);
+		}
+		spin_unlock_irqrestore(&connlock, flags);
+	}
 
-	transport_container_unregister(&i->t.target_attrs);
-	transport_container_unregister(&i->t.host_attrs);
-  
-	kfree(i);
+	return NOTIFY_DONE;
 }
 
-EXPORT_SYMBOL(iscsi_release_transport);
+static struct notifier_block iscsi_nl_notifier = {
+	.notifier_call	= iscsi_rcv_nl_event,
+};
 
 static __init int iscsi_transport_init(void)
 {
-	int err = transport_class_register(&iscsi_transport_class);
+	int err;
 
+	err = class_register(&iscsi_transport_class);
 	if (err)
 		return err;
-	return transport_class_register(&iscsi_host_class);
+
+	err = transport_class_register(&iscsi_connection_class);
+	if (err)
+		goto unregister_transport_class;
+
+	err = transport_class_register(&iscsi_session_class);
+	if (err)
+		goto unregister_conn_class;
+
+	err = netlink_register_notifier(&iscsi_nl_notifier);
+	if (err)
+		goto unregister_session_class;
+
+	nls = netlink_kernel_create(NETLINK_ISCSI, iscsi_if_rx);
+	if (!nls) {
+		err = -ENOBUFS;
+		goto unregister_notifier;
+	}
+
+	err = mempool_zone_init(&z_reply, Z_MAX_REPLY,
+		NLMSG_SPACE(sizeof(struct iscsi_uevent)), Z_HIWAT_REPLY);
+	if (!err)
+		return 0;
+
+	sock_release(nls->sk_socket);
+unregister_notifier:
+	netlink_unregister_notifier(&iscsi_nl_notifier);
+unregister_session_class:
+	transport_class_unregister(&iscsi_session_class);
+unregister_conn_class:
+	transport_class_unregister(&iscsi_connection_class);
+unregister_transport_class:
+	class_unregister(&iscsi_transport_class);
+	return err;
 }
 
 static void __exit iscsi_transport_exit(void)
 {
-	transport_class_unregister(&iscsi_host_class);
-	transport_class_unregister(&iscsi_transport_class);
+	mempool_destroy(z_reply.pool);
+	sock_release(nls->sk_socket);
+	netlink_unregister_notifier(&iscsi_nl_notifier);
+	transport_class_unregister(&iscsi_connection_class);
+	transport_class_unregister(&iscsi_session_class);
+	class_unregister(&iscsi_transport_class);
 }
 
 module_init(iscsi_transport_init);
 module_exit(iscsi_transport_exit);
 
-MODULE_AUTHOR("Mike Christie");
-MODULE_DESCRIPTION("iSCSI Transport Attributes");
+MODULE_AUTHOR("Mike Christie <michaelc@cs.wisc.edu>, "
+	      "Dmitry Yusupov <dmitry_yus@yahoo.com>, "
+	      "Alex Aizman <itn780@yahoo.com>");
+MODULE_DESCRIPTION("iSCSI Transport Interface");
 MODULE_LICENSE("GPL");