[SCSI] iscsi_tcp: add iscsi_tcp prefix to iscsi_tcp functions

[safe/jmp/linux-2.6] / drivers / scsi / iscsi_tcp.c

/*
 * iSCSI Initiator over TCP/IP Data-Path
 *
 * Copyright (C) 2004 Dmitry Yusupov
 * Copyright (C) 2004 Alex Aizman
 * Copyright (C) 2005 - 2006 Mike Christie
 * Copyright (C) 2006 Red Hat, Inc.  All rights reserved.
 * 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/file.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_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");
#undef DEBUG_TCP
#define DEBUG_ASSERT

#ifdef DEBUG_TCP
#define debug_tcp(fmt...) printk(KERN_INFO "tcp: " fmt)
#else
#define debug_tcp(fmt...)
#endif

static struct scsi_transport_template *iscsi_tcp_scsi_transport;
static struct scsi_host_template iscsi_sht;
static struct iscsi_transport iscsi_tcp_transport;

static unsigned int iscsi_max_lun = 512;
module_param_named(max_lun, iscsi_max_lun, uint, S_IRUGO);

static int iscsi_tcp_hdr_recv_done(struct iscsi_tcp_conn *tcp_conn,
                                   struct iscsi_segment *segment);

/*
 * Scatterlist handling: inside the iscsi_segment, we
 * remember an index into the scatterlist, and set data/size
 * to the current scatterlist entry. For highmem pages, we
 * kmap as needed.
 *
 * Note that the page is unmapped when we return from
 * TCP's data_ready handler, so we may end up mapping and
 * unmapping the same page repeatedly. The whole reason
 * for this is that we shouldn't keep the page mapped
 * outside the softirq.
 */

/**
 * iscsi_tcp_segment_init_sg - init indicated scatterlist entry
 * @segment: the buffer object
 * @sg: scatterlist
 * @offset: byte offset into that sg entry
 *
 * This function sets up the segment so that subsequent
 * data is copied to the indicated sg entry, at the given
 * offset.
 */
static inline void
iscsi_tcp_segment_init_sg(struct iscsi_segment *segment,
                          struct scatterlist *sg, unsigned int offset)
{
        segment->sg = sg;
        segment->sg_offset = offset;
        segment->size = min(sg->length - offset,
                            segment->total_size - segment->total_copied);
        segment->data = NULL;
}

/**
 * iscsi_tcp_segment_map - map the current S/G page
 * @segment: iscsi_segment
 * @recv: 1 if called from recv path
 *
 * We only need to possibly kmap data if scatter lists are being used,
 * because the iscsi passthrough and internal IO paths will never use high
 * mem pages.
 */
static inline void
iscsi_tcp_segment_map(struct iscsi_segment *segment, int recv)
{
        struct scatterlist *sg;

        if (segment->data != NULL || !segment->sg)
                return;

        sg = segment->sg;
        BUG_ON(segment->sg_mapped);
        BUG_ON(sg->length == 0);

        /*
         * If the page count is greater than one it is ok to send
         * to the network layer's zero copy send path. If not we
         * have to go the slow sendmsg path. We always map for the
         * recv path.
         */
        if (page_count(sg_page(sg)) >= 1 && !recv)
                return;

        debug_tcp("iscsi_tcp_segment_map %s %p\n", recv ? "recv" : "xmit",
                  segment);
        segment->sg_mapped = kmap_atomic(sg_page(sg), KM_SOFTIRQ0);
        segment->data = segment->sg_mapped + sg->offset + segment->sg_offset;
}

static inline void
iscsi_tcp_segment_unmap(struct iscsi_segment *segment)
{
        debug_tcp("iscsi_tcp_segment_unmap %p\n", segment);

        if (segment->sg_mapped) {
                debug_tcp("iscsi_tcp_segment_unmap valid\n");
                kunmap_atomic(segment->sg_mapped, KM_SOFTIRQ0);
                segment->sg_mapped = NULL;
                segment->data = NULL;
        }
}

/*
 * Splice the digest buffer into the buffer
 */
static inline void
iscsi_tcp_segment_splice_digest(struct iscsi_segment *segment, void *digest)
{
        segment->data = digest;
        segment->digest_len = ISCSI_DIGEST_SIZE;
        segment->total_size += ISCSI_DIGEST_SIZE;
        segment->size = ISCSI_DIGEST_SIZE;
        segment->copied = 0;
        segment->sg = NULL;
        segment->hash = NULL;
}

/**
 * iscsi_tcp_segment_done - check whether the segment is complete
 * @segment: iscsi segment to check
 * @recv: set to one of this is called from the recv path
 * @copied: number of bytes copied
 *
 * Check if we're done receiving this segment. If the receive
 * buffer is full but we expect more data, move on to the
 * next entry in the scatterlist.
 *
 * If the amount of data we received isn't a multiple of 4,
 * we will transparently receive the pad bytes, too.
 *
 * This function must be re-entrant.
 */
static inline int
iscsi_tcp_segment_done(struct iscsi_segment *segment, int recv, unsigned copied)
{
        static unsigned char padbuf[ISCSI_PAD_LEN];
        struct scatterlist sg;
        unsigned int pad;

        debug_tcp("copied %u %u size %u %s\n", segment->copied, copied,
                  segment->size, recv ? "recv" : "xmit");
        if (segment->hash && copied) {
                /*
                 * If a segment is kmapd we must unmap it before sending
                 * to the crypto layer since that will try to kmap it again.
                 */
                iscsi_tcp_segment_unmap(segment);

                if (!segment->data) {
                        sg_init_table(&sg, 1);
                        sg_set_page(&sg, sg_page(segment->sg), copied,
                                    segment->copied + segment->sg_offset +
                                                        segment->sg->offset);
                } else
                        sg_init_one(&sg, segment->data + segment->copied,
                                    copied);
                crypto_hash_update(segment->hash, &sg, copied);
        }

        segment->copied += copied;
        if (segment->copied < segment->size) {
                iscsi_tcp_segment_map(segment, recv);
                return 0;
        }

        segment->total_copied += segment->copied;
        segment->copied = 0;
        segment->size = 0;

        /* Unmap the current scatterlist page, if there is one. */
        iscsi_tcp_segment_unmap(segment);

        /* Do we have more scatterlist entries? */
        debug_tcp("total copied %u total size %u\n", segment->total_copied,
                   segment->total_size);
        if (segment->total_copied < segment->total_size) {
                /* Proceed to the next entry in the scatterlist. */
                iscsi_tcp_segment_init_sg(segment, sg_next(segment->sg),
                                          0);
                iscsi_tcp_segment_map(segment, recv);
                BUG_ON(segment->size == 0);
                return 0;
        }

        /* Do we need to handle padding? */
        pad = iscsi_padding(segment->total_copied);
        if (pad != 0) {
                debug_tcp("consume %d pad bytes\n", pad);
                segment->total_size += pad;
                segment->size = pad;
                segment->data = padbuf;
                return 0;
        }

        /*
         * Set us up for transferring the data digest. hdr digest
         * is completely handled in hdr done function.
         */
        if (segment->hash) {
                crypto_hash_final(segment->hash, segment->digest);
                iscsi_tcp_segment_splice_digest(segment,
                                 recv ? segment->recv_digest : segment->digest);
                return 0;
        }

        return 1;
}

/**
 * iscsi_tcp_xmit_segment - transmit segment
 * @tcp_conn: the iSCSI TCP connection
 * @segment: the buffer to transmnit
 *
 * This function transmits as much of the buffer as
 * the network layer will accept, and returns the number of
 * bytes transmitted.
 *
 * If CRC hashing is enabled, the function will compute the
 * hash as it goes. When the entire segment has been transmitted,
 * it will retrieve the hash value and send it as well.
 */
static int
iscsi_tcp_xmit_segment(struct iscsi_tcp_conn *tcp_conn,
                       struct iscsi_segment *segment)
{
        struct socket *sk = tcp_conn->sock;
        unsigned int copied = 0;
        int r = 0;

        while (!iscsi_tcp_segment_done(segment, 0, r)) {
                struct scatterlist *sg;
                unsigned int offset, copy;
                int flags = 0;

                r = 0;
                offset = segment->copied;
                copy = segment->size - offset;

                if (segment->total_copied + segment->size < segment->total_size)
                        flags |= MSG_MORE;

                /* Use sendpage if we can; else fall back to sendmsg */
                if (!segment->data) {
                        sg = segment->sg;
                        offset += segment->sg_offset + sg->offset;
                        r = tcp_conn->sendpage(sk, sg_page(sg), offset, copy,
                                               flags);
                } else {
                        struct msghdr msg = { .msg_flags = flags };
                        struct kvec iov = {
                                .iov_base = segment->data + offset,
                                .iov_len = copy
                        };

                        r = kernel_sendmsg(sk, &msg, &iov, 1, copy);
                }

                if (r < 0) {
                        iscsi_tcp_segment_unmap(segment);
                        if (copied || r == -EAGAIN)
                                break;
                        return r;
                }
                copied += r;
        }
        return copied;
}

/**
 * iscsi_tcp_segment_recv - copy data to segment
 * @tcp_conn: the iSCSI TCP connection
 * @segment: the buffer to copy to
 * @ptr: data pointer
 * @len: amount of data available
 *
 * This function copies up to @len bytes to the
 * given buffer, and returns the number of bytes
 * consumed, which can actually be less than @len.
 *
 * If hash digest is enabled, the function will update the
 * hash while copying.
 * Combining these two operations doesn't buy us a lot (yet),
 * but in the future we could implement combined copy+crc,
 * just way we do for network layer checksums.
 */
static int
iscsi_tcp_segment_recv(struct iscsi_tcp_conn *tcp_conn,
                       struct iscsi_segment *segment, const void *ptr,
                       unsigned int len)
{
        unsigned int copy = 0, copied = 0;

        while (!iscsi_tcp_segment_done(segment, 1, copy)) {
                if (copied == len) {
                        debug_tcp("iscsi_tcp_segment_recv copied %d bytes\n",
                                  len);
                        break;
                }

                copy = min(len - copied, segment->size - segment->copied);
                debug_tcp("iscsi_tcp_segment_recv copying %d\n", copy);
                memcpy(segment->data + segment->copied, ptr + copied, copy);
                copied += copy;
        }
        return copied;
}

static inline void
iscsi_tcp_dgst_header(struct hash_desc *hash, const void *hdr, size_t hdrlen,
                      unsigned char digest[ISCSI_DIGEST_SIZE])
{
        struct scatterlist sg;

        sg_init_one(&sg, hdr, hdrlen);
        crypto_hash_digest(hash, &sg, hdrlen, digest);
}

static inline int
iscsi_tcp_dgst_verify(struct iscsi_tcp_conn *tcp_conn,
                      struct iscsi_segment *segment)
{
        if (!segment->digest_len)
                return 1;

        if (memcmp(segment->recv_digest, segment->digest,
                   segment->digest_len)) {
                debug_scsi("digest mismatch\n");
                return 0;
        }

        return 1;
}

/*
 * Helper function to set up segment buffer
 */
static inline void
__iscsi_segment_init(struct iscsi_segment *segment, size_t size,
                     iscsi_segment_done_fn_t *done, struct hash_desc *hash)
{
        memset(segment, 0, sizeof(*segment));
        segment->total_size = size;
        segment->done = done;

        if (hash) {
                segment->hash = hash;
                crypto_hash_init(hash);
        }
}

static inline void
iscsi_segment_init_linear(struct iscsi_segment *segment, void *data,
                          size_t size, iscsi_segment_done_fn_t *done,
                          struct hash_desc *hash)
{
        __iscsi_segment_init(segment, size, done, hash);
        segment->data = data;
        segment->size = size;
}

static inline int
iscsi_segment_seek_sg(struct iscsi_segment *segment,
                      struct scatterlist *sg_list, unsigned int sg_count,
                      unsigned int offset, size_t size,
                      iscsi_segment_done_fn_t *done, struct hash_desc *hash)
{
        struct scatterlist *sg;
        unsigned int i;

        debug_scsi("iscsi_segment_seek_sg offset %u size %llu\n",
                  offset, size);
        __iscsi_segment_init(segment, size, done, hash);
        for_each_sg(sg_list, sg, sg_count, i) {
                debug_scsi("sg %d, len %u offset %u\n", i, sg->length,
                           sg->offset);
                if (offset < sg->length) {
                        iscsi_tcp_segment_init_sg(segment, sg, offset);
                        return 0;
                }
                offset -= sg->length;
        }

        return ISCSI_ERR_DATA_OFFSET;
}

/**
 * iscsi_tcp_hdr_recv_prep - prep segment for hdr reception
 * @tcp_conn: iscsi connection to prep for
 *
 * This function always passes NULL for the hash argument, because when this
 * function is called we do not yet know the final size of the header and want
 * to delay the digest processing until we know that.
 */
static void
iscsi_tcp_hdr_recv_prep(struct iscsi_tcp_conn *tcp_conn)
{
        debug_tcp("iscsi_tcp_hdr_recv_prep(%p%s)\n", tcp_conn,
                  tcp_conn->iscsi_conn->hdrdgst_en ? ", digest enabled" : "");
        iscsi_segment_init_linear(&tcp_conn->in.segment,
                                tcp_conn->in.hdr_buf, sizeof(struct iscsi_hdr),
                                iscsi_tcp_hdr_recv_done, NULL);
}

/*
 * Handle incoming reply to any other type of command
 */
static int
iscsi_tcp_data_recv_done(struct iscsi_tcp_conn *tcp_conn,
                         struct iscsi_segment *segment)
{
        struct iscsi_conn *conn = tcp_conn->iscsi_conn;
        int rc = 0;

        if (!iscsi_tcp_dgst_verify(tcp_conn, segment))
                return ISCSI_ERR_DATA_DGST;

        rc = iscsi_complete_pdu(conn, tcp_conn->in.hdr,
                        conn->data, tcp_conn->in.datalen);
        if (rc)
                return rc;

        iscsi_tcp_hdr_recv_prep(tcp_conn);
        return 0;
}

static void
iscsi_tcp_data_recv_prep(struct iscsi_tcp_conn *tcp_conn)
{
        struct iscsi_conn *conn = tcp_conn->iscsi_conn;
        struct hash_desc *rx_hash = NULL;

        if (conn->datadgst_en &
            !(conn->session->tt->caps & CAP_DIGEST_OFFLOAD))
                rx_hash = &tcp_conn->rx_hash;

        iscsi_segment_init_linear(&tcp_conn->in.segment,
                                conn->data, tcp_conn->in.datalen,
                                iscsi_tcp_data_recv_done, rx_hash);
}

/*
 * must be called with session lock
 */
static void iscsi_tcp_cleanup_task(struct iscsi_task *task)
{
        struct iscsi_tcp_task *tcp_task = task->dd_data;
        struct iscsi_r2t_info *r2t;

        /* nothing to do for mgmt or pending tasks */
        if (!task->sc || task->state == ISCSI_TASK_PENDING)
                return;

        /* flush task's r2t queues */
        while (__kfifo_get(tcp_task->r2tqueue, (void*)&r2t, sizeof(void*))) {
                __kfifo_put(tcp_task->r2tpool.queue, (void*)&r2t,
                            sizeof(void*));
                debug_scsi("iscsi_tcp_cleanup_task pending r2t dropped\n");
        }

        r2t = tcp_task->r2t;
        if (r2t != NULL) {
                __kfifo_put(tcp_task->r2tpool.queue, (void*)&r2t,
                            sizeof(void*));
                tcp_task->r2t = NULL;
        }
}

/**
 * iscsi_tcp_data_in - SCSI Data-In Response processing
 * @conn: iscsi connection
 * @task: scsi command task
 */
static int iscsi_tcp_data_in(struct iscsi_conn *conn, struct iscsi_task *task)
{
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
        struct iscsi_tcp_task *tcp_task = task->dd_data;
        struct iscsi_data_rsp *rhdr = (struct iscsi_data_rsp *)tcp_conn->in.hdr;
        int datasn = be32_to_cpu(rhdr->datasn);
        unsigned total_in_length = scsi_in(task->sc)->length;

        iscsi_update_cmdsn(conn->session, (struct iscsi_nopin*)rhdr);
        if (tcp_conn->in.datalen == 0)
                return 0;

        if (tcp_task->exp_datasn != datasn) {
                debug_tcp("%s: task->exp_datasn(%d) != rhdr->datasn(%d)\n",
                          __func__, tcp_task->exp_datasn, datasn);
                return ISCSI_ERR_DATASN;
        }

        tcp_task->exp_datasn++;

        tcp_task->data_offset = be32_to_cpu(rhdr->offset);
        if (tcp_task->data_offset + tcp_conn->in.datalen > total_in_length) {
                debug_tcp("%s: data_offset(%d) + data_len(%d) > total_length_in(%d)\n",
                          __func__, tcp_task->data_offset,
                          tcp_conn->in.datalen, total_in_length);
                return ISCSI_ERR_DATA_OFFSET;
        }

        conn->datain_pdus_cnt++;
        return 0;
}

/**
 * iscsi_tcp_r2t_rsp - iSCSI R2T Response processing
 * @conn: iscsi connection
 * @task: scsi command task
 */
static int iscsi_tcp_r2t_rsp(struct iscsi_conn *conn, struct iscsi_task *task)
{
        struct iscsi_session *session = conn->session;
        struct iscsi_tcp_task *tcp_task = task->dd_data;
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
        struct iscsi_r2t_rsp *rhdr = (struct iscsi_r2t_rsp *)tcp_conn->in.hdr;
        struct iscsi_r2t_info *r2t;
        int r2tsn = be32_to_cpu(rhdr->r2tsn);
        int rc;

        if (tcp_conn->in.datalen) {
                iscsi_conn_printk(KERN_ERR, conn,
                                  "invalid R2t with datalen %d\n",
                                  tcp_conn->in.datalen);
                return ISCSI_ERR_DATALEN;
        }

        if (tcp_task->exp_datasn != r2tsn){
                debug_tcp("%s: task->exp_datasn(%d) != rhdr->r2tsn(%d)\n",
                          __func__, tcp_task->exp_datasn, r2tsn);
                return ISCSI_ERR_R2TSN;
        }

        /* fill-in new R2T associated with the task */
        iscsi_update_cmdsn(session, (struct iscsi_nopin*)rhdr);

        if (!task->sc || session->state != ISCSI_STATE_LOGGED_IN) {
                iscsi_conn_printk(KERN_INFO, conn,
                                  "dropping R2T itt %d in recovery.\n",
                                  task->itt);
                return 0;
        }

        rc = __kfifo_get(tcp_task->r2tpool.queue, (void*)&r2t, sizeof(void*));
        if (!rc) {
                iscsi_conn_printk(KERN_ERR, conn, "Could not allocate R2T. "
                                  "Target has sent more R2Ts than it "
                                  "negotiated for or driver has has leaked.\n");
                return ISCSI_ERR_PROTO;
        }

        r2t->exp_statsn = rhdr->statsn;
        r2t->data_length = be32_to_cpu(rhdr->data_length);
        if (r2t->data_length == 0) {
                iscsi_conn_printk(KERN_ERR, conn,
                                  "invalid R2T with zero data len\n");
                __kfifo_put(tcp_task->r2tpool.queue, (void*)&r2t,
                            sizeof(void*));
                return ISCSI_ERR_DATALEN;
        }

        if (r2t->data_length > session->max_burst)
                debug_scsi("invalid R2T with data len %u and max burst %u."
                           "Attempting to execute request.\n",
                            r2t->data_length, session->max_burst);

        r2t->data_offset = be32_to_cpu(rhdr->data_offset);
        if (r2t->data_offset + r2t->data_length > scsi_out(task->sc)->length) {
                iscsi_conn_printk(KERN_ERR, conn,
                                  "invalid R2T with data len %u at offset %u "
                                  "and total length %d\n", r2t->data_length,
                                  r2t->data_offset, scsi_out(task->sc)->length);
                __kfifo_put(tcp_task->r2tpool.queue, (void*)&r2t,
                            sizeof(void*));
                return ISCSI_ERR_DATALEN;
        }

        r2t->ttt = rhdr->ttt; /* no flip */
        r2t->datasn = 0;
        r2t->sent = 0;

        tcp_task->exp_datasn = r2tsn + 1;
        __kfifo_put(tcp_task->r2tqueue, (void*)&r2t, sizeof(void*));
        conn->r2t_pdus_cnt++;

        iscsi_requeue_task(task);
        return 0;
}

/*
 * Handle incoming reply to DataIn command
 */
static int
iscsi_tcp_process_data_in(struct iscsi_tcp_conn *tcp_conn,
                          struct iscsi_segment *segment)
{
        struct iscsi_conn *conn = tcp_conn->iscsi_conn;
        struct iscsi_hdr *hdr = tcp_conn->in.hdr;
        int rc;

        if (!iscsi_tcp_dgst_verify(tcp_conn, segment))
                return ISCSI_ERR_DATA_DGST;

        /* check for non-exceptional status */
        if (hdr->flags & ISCSI_FLAG_DATA_STATUS) {
                rc = iscsi_complete_pdu(conn, tcp_conn->in.hdr, NULL, 0);
                if (rc)
                        return rc;
        }

        iscsi_tcp_hdr_recv_prep(tcp_conn);
        return 0;
}

/**
 * iscsi_tcp_hdr_dissect - process PDU header
 * @conn: iSCSI connection
 * @hdr: PDU header
 *
 * This function analyzes the header of the PDU received,
 * and performs several sanity checks. If the PDU is accompanied
 * by data, the receive buffer is set up to copy the incoming data
 * to the correct location.
 */
static int
iscsi_tcp_hdr_dissect(struct iscsi_conn *conn, struct iscsi_hdr *hdr)
{
        int rc = 0, opcode, ahslen;
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
        struct iscsi_task *task;

        /* verify PDU length */
        tcp_conn->in.datalen = ntoh24(hdr->dlength);
        if (tcp_conn->in.datalen > conn->max_recv_dlength) {
                iscsi_conn_printk(KERN_ERR, conn,
                                  "iscsi_tcp: datalen %d > %d\n",
                                  tcp_conn->in.datalen, conn->max_recv_dlength);
                return ISCSI_ERR_DATALEN;
        }

        /* Additional header segments. So far, we don't
         * process additional headers.
         */
        ahslen = hdr->hlength << 2;

        opcode = hdr->opcode & ISCSI_OPCODE_MASK;
        /* verify itt (itt encoding: age+cid+itt) */
        rc = iscsi_verify_itt(conn, hdr->itt);
        if (rc)
                return rc;

        debug_tcp("opcode 0x%x ahslen %d datalen %d\n",
                  opcode, ahslen, tcp_conn->in.datalen);

        switch(opcode) {
        case ISCSI_OP_SCSI_DATA_IN:
                spin_lock(&conn->session->lock);
                task = iscsi_itt_to_ctask(conn, hdr->itt);
                if (!task)
                        rc = ISCSI_ERR_BAD_ITT;
                else
                        rc = iscsi_tcp_data_in(conn, task);
                if (rc) {
                        spin_unlock(&conn->session->lock);
                        break;
                }

                if (tcp_conn->in.datalen) {
                        struct iscsi_tcp_task *tcp_task = task->dd_data;
                        struct hash_desc *rx_hash = NULL;
                        struct scsi_data_buffer *sdb = scsi_in(task->sc);

                        /*
                         * Setup copy of Data-In into the Scsi_Cmnd
                         * Scatterlist case:
                         * We set up the iscsi_segment to point to the next
                         * scatterlist entry to copy to. As we go along,
                         * we move on to the next scatterlist entry and
                         * update the digest per-entry.
                         */
                        if (conn->datadgst_en &&
                            !(conn->session->tt->caps & CAP_DIGEST_OFFLOAD))
                                rx_hash = &tcp_conn->rx_hash;

                        debug_tcp("iscsi_tcp_begin_data_in(%p, offset=%d, "
                                  "datalen=%d)\n", tcp_conn,
                                  tcp_task->data_offset,
                                  tcp_conn->in.datalen);
                        rc = iscsi_segment_seek_sg(&tcp_conn->in.segment,
                                                   sdb->table.sgl,
                                                   sdb->table.nents,
                                                   tcp_task->data_offset,
                                                   tcp_conn->in.datalen,
                                                   iscsi_tcp_process_data_in,
                                                   rx_hash);
                        spin_unlock(&conn->session->lock);
                        return rc;
                }
                rc = __iscsi_complete_pdu(conn, hdr, NULL, 0);
                spin_unlock(&conn->session->lock);
                break;
        case ISCSI_OP_SCSI_CMD_RSP:
                if (tcp_conn->in.datalen) {
                        iscsi_tcp_data_recv_prep(tcp_conn);
                        return 0;
                }
                rc = iscsi_complete_pdu(conn, hdr, NULL, 0);
                break;
        case ISCSI_OP_R2T:
                spin_lock(&conn->session->lock);
                task = iscsi_itt_to_ctask(conn, hdr->itt);
                if (!task)
                        rc = ISCSI_ERR_BAD_ITT;
                else if (ahslen)
                        rc = ISCSI_ERR_AHSLEN;
                else if (task->sc->sc_data_direction == DMA_TO_DEVICE)
                        rc = iscsi_tcp_r2t_rsp(conn, task);
                else
                        rc = ISCSI_ERR_PROTO;
                spin_unlock(&conn->session->lock);
                break;
        case ISCSI_OP_LOGIN_RSP:
        case ISCSI_OP_TEXT_RSP:
        case ISCSI_OP_REJECT:
        case ISCSI_OP_ASYNC_EVENT:
                /*
                 * It is possible that we could get a PDU with a buffer larger
                 * than 8K, but there are no targets that currently do this.
                 * For now we fail until we find a vendor that needs it
                 */
                if (ISCSI_DEF_MAX_RECV_SEG_LEN < tcp_conn->in.datalen) {
                        iscsi_conn_printk(KERN_ERR, conn,
                                          "iscsi_tcp: received buffer of "
                                          "len %u but conn buffer is only %u "
                                          "(opcode %0x)\n",
                                          tcp_conn->in.datalen,
                                          ISCSI_DEF_MAX_RECV_SEG_LEN, opcode);
                        rc = ISCSI_ERR_PROTO;
                        break;
                }

                /* If there's data coming in with the response,
                 * receive it to the connection's buffer.
                 */
                if (tcp_conn->in.datalen) {
                        iscsi_tcp_data_recv_prep(tcp_conn);
                        return 0;
                }
        /* fall through */
        case ISCSI_OP_LOGOUT_RSP:
        case ISCSI_OP_NOOP_IN:
        case ISCSI_OP_SCSI_TMFUNC_RSP:
                rc = iscsi_complete_pdu(conn, hdr, NULL, 0);
                break;
        default:
                rc = ISCSI_ERR_BAD_OPCODE;
                break;
        }

        if (rc == 0) {
                /* Anything that comes with data should have
                 * been handled above. */
                if (tcp_conn->in.datalen)
                        return ISCSI_ERR_PROTO;
                iscsi_tcp_hdr_recv_prep(tcp_conn);
        }

        return rc;
}

/**
 * iscsi_tcp_hdr_recv_done - process PDU header
 *
 * This is the callback invoked when the PDU header has
 * been received. If the header is followed by additional
 * header segments, we go back for more data.
 */
static int
iscsi_tcp_hdr_recv_done(struct iscsi_tcp_conn *tcp_conn,
                        struct iscsi_segment *segment)
{
        struct iscsi_conn *conn = tcp_conn->iscsi_conn;
        struct iscsi_hdr *hdr;

        /* Check if there are additional header segments
         * *prior* to computing the digest, because we
         * may need to go back to the caller for more.
         */
        hdr = (struct iscsi_hdr *) tcp_conn->in.hdr_buf;
        if (segment->copied == sizeof(struct iscsi_hdr) && hdr->hlength) {
                /* Bump the header length - the caller will
                 * just loop around and get the AHS for us, and
                 * call again. */
                unsigned int ahslen = hdr->hlength << 2;

                /* Make sure we don't overflow */
                if (sizeof(*hdr) + ahslen > sizeof(tcp_conn->in.hdr_buf))
                        return ISCSI_ERR_AHSLEN;

                segment->total_size += ahslen;
                segment->size += ahslen;
                return 0;
        }

        /* We're done processing the header. See if we're doing
         * header digests; if so, set up the recv_digest buffer
         * and go back for more. */
        if (conn->hdrdgst_en) {
                if (segment->digest_len == 0) {
                        /*
                         * Even if we offload the digest processing we
                         * splice it in so we can increment the skb/segment
                         * counters in preparation for the data segment.
                         */
                        iscsi_tcp_segment_splice_digest(segment,
                                                        segment->recv_digest);
                        return 0;
                }

                if (!(conn->session->tt->caps & CAP_DIGEST_OFFLOAD)) {
                        iscsi_tcp_dgst_header(&tcp_conn->rx_hash, hdr,
                                segment->total_copied - ISCSI_DIGEST_SIZE,
                                segment->digest);

                        if (!iscsi_tcp_dgst_verify(tcp_conn, segment))
                                return ISCSI_ERR_HDR_DGST;
                }
        }

        tcp_conn->in.hdr = hdr;
        return iscsi_tcp_hdr_dissect(conn, hdr);
}

inline int iscsi_tcp_recv_segment_is_hdr(struct iscsi_tcp_conn *tcp_conn)
{
        return tcp_conn->in.segment.done == iscsi_tcp_hdr_recv_done;
}

enum {
        ISCSI_TCP_SEGMENT_DONE,         /* curr seg has been processed */
        ISCSI_TCP_SKB_DONE,             /* skb is out of data */
        ISCSI_TCP_CONN_ERR,             /* iscsi layer has fired a conn err */
        ISCSI_TCP_SUSPENDED,            /* conn is suspended */
};

/**
 * iscsi_tcp_recv_skb - Process skb
 * @conn: iscsi connection
 * @skb: network buffer with header and/or data segment
 * @offset: offset in skb
 * @offload: bool indicating if transfer was offloaded
 */
int iscsi_tcp_recv_skb(struct iscsi_conn *conn, struct sk_buff *skb,
                       unsigned int offset, bool offloaded, int *status)
{
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
        struct iscsi_segment *segment = &tcp_conn->in.segment;
        struct skb_seq_state seq;
        unsigned int consumed = 0;
        int rc = 0;

        debug_tcp("in %d bytes\n", skb->len - offset);

        if (unlikely(conn->suspend_rx)) {
                debug_tcp("conn %d Rx suspended!\n", conn->id);
                *status = ISCSI_TCP_SUSPENDED;
                return 0;
        }

        if (offloaded) {
                segment->total_copied = segment->total_size;
                goto segment_done;
        }

        skb_prepare_seq_read(skb, offset, skb->len, &seq);
        while (1) {
                unsigned int avail;
                const u8 *ptr;

                avail = skb_seq_read(consumed, &ptr, &seq);
                if (avail == 0) {
                        debug_tcp("no more data avail. Consumed %d\n",
                                  consumed);
                        *status = ISCSI_TCP_SKB_DONE;
                        skb_abort_seq_read(&seq);
                        goto skb_done;
                }
                BUG_ON(segment->copied >= segment->size);

                debug_tcp("skb %p ptr=%p avail=%u\n", skb, ptr, avail);
                rc = iscsi_tcp_segment_recv(tcp_conn, segment, ptr, avail);
                BUG_ON(rc == 0);
                consumed += rc;

                if (segment->total_copied >= segment->total_size) {
                        skb_abort_seq_read(&seq);
                        goto segment_done;
                }
        }

segment_done:
        *status = ISCSI_TCP_SEGMENT_DONE;
        debug_tcp("segment done\n");
        rc = segment->done(tcp_conn, segment);
        if (rc != 0) {
                *status = ISCSI_TCP_CONN_ERR;
                debug_tcp("Error receiving PDU, errno=%d\n", rc);
                iscsi_conn_failure(conn, rc);
                return 0;
        }
        /* The done() functions sets up the next segment. */

skb_done:
        conn->rxdata_octets += consumed;
        return consumed;
}
EXPORT_SYMBOL_GPL(iscsi_tcp_recv_skb);

/**
 * iscsi_tcp_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_recv(read_descriptor_t *rd_desc, struct sk_buff *skb,
               unsigned int offset, size_t len)
{
        struct iscsi_conn *conn = rd_desc->arg.data;
        unsigned int consumed, total_consumed = 0;
        int status;

        debug_tcp("in %d bytes\n", skb->len - offset);

        do {
                status = 0;
                consumed = iscsi_tcp_recv_skb(conn, skb, offset, 0, &status);
                offset += consumed;
                total_consumed += consumed;
        } while (consumed != 0 && status != ISCSI_TCP_SKB_DONE);

        debug_tcp("read %d bytes status %d\n", skb->len - offset, status);
        return total_consumed;
}

static void
iscsi_tcp_data_ready(struct sock *sk, int flag)
{
        struct iscsi_conn *conn = sk->sk_user_data;
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
        read_descriptor_t rd_desc;

        read_lock(&sk->sk_callback_lock);

        /*
         * Use rd_desc to pass 'conn' to iscsi_tcp_recv.
         * We set count to 1 because we want the network layer to
         * hand us all the skbs that are available. iscsi_tcp_recv
         * handled pdus that cross buffers or pdus that still need data.
         */
        rd_desc.arg.data = conn;
        rd_desc.count = 1;
        tcp_read_sock(sk, &rd_desc, iscsi_tcp_recv);

        read_unlock(&sk->sk_callback_lock);

        /* If we had to (atomically) map a highmem page,
         * unmap it now. */
        iscsi_tcp_segment_unmap(&tcp_conn->in.segment);
}

static void
iscsi_tcp_state_change(struct sock *sk)
{
        struct iscsi_tcp_conn *tcp_conn;
        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) &&
            !atomic_read(&sk->sk_rmem_alloc)) {
                debug_tcp("iscsi_tcp_state_change: TCP_CLOSE|TCP_CLOSE_WAIT\n");
                iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
        }

        tcp_conn = conn->dd_data;
        old_state_change = tcp_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_tcp_write_space(struct sock *sk)
{
        struct iscsi_conn *conn = (struct iscsi_conn*)sk->sk_user_data;
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;

        tcp_conn->old_write_space(sk);
        debug_tcp("iscsi_write_space: cid %d\n", conn->id);
        scsi_queue_work(conn->session->host, &conn->xmitwork);
}

static void iscsi_tcp_conn_set_callbacks(struct iscsi_conn *conn)
{
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
        struct sock *sk = tcp_conn->sock->sk;

        /* assign new callbacks */
        write_lock_bh(&sk->sk_callback_lock);
        sk->sk_user_data = conn;
        tcp_conn->old_data_ready = sk->sk_data_ready;
        tcp_conn->old_state_change = sk->sk_state_change;
        tcp_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_tcp_write_space;
        write_unlock_bh(&sk->sk_callback_lock);
}

static void iscsi_conn_restore_callbacks(struct iscsi_tcp_conn *tcp_conn)
{
        struct sock *sk = tcp_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   = tcp_conn->old_data_ready;
        sk->sk_state_change = tcp_conn->old_state_change;
        sk->sk_write_space  = tcp_conn->old_write_space;
        sk->sk_no_check  = 0;
        write_unlock_bh(&sk->sk_callback_lock);
}

/**
 * iscsi_tcp_xmit - TCP transmit
 **/
static int iscsi_tcp_xmit(struct iscsi_conn *conn)
{
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
        struct iscsi_segment *segment = &tcp_conn->out.segment;
        unsigned int consumed = 0;
        int rc = 0;

        while (1) {
                rc = iscsi_tcp_xmit_segment(tcp_conn, segment);
                if (rc < 0) {
                        rc = ISCSI_ERR_XMIT_FAILED;
                        goto error;
                }
                if (rc == 0)
                        break;

                consumed += rc;

                if (segment->total_copied >= segment->total_size) {
                        if (segment->done != NULL) {
                                rc = segment->done(tcp_conn, segment);
                                if (rc != 0)
                                        goto error;
                        }
                }
        }

        debug_tcp("xmit %d bytes\n", consumed);

        conn->txdata_octets += consumed;
        return consumed;

error:
        /* Transmit error. We could initiate error recovery
         * here. */
        debug_tcp("Error sending PDU, errno=%d\n", rc);
        iscsi_conn_failure(conn, rc);
        return -EIO;
}

/**
 * iscsi_tcp_xmit_qlen - return the number of bytes queued for xmit
 */
static inline int
iscsi_tcp_xmit_qlen(struct iscsi_conn *conn)
{
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
        struct iscsi_segment *segment = &tcp_conn->out.segment;

        return segment->total_copied - segment->total_size;
}

static int iscsi_tcp_flush(struct iscsi_task *task)
{
        struct iscsi_conn *conn = task->conn;
        int rc;

        while (iscsi_tcp_xmit_qlen(conn)) {
                rc = iscsi_tcp_xmit(conn);
                if (rc == 0)
                        return -EAGAIN;
                if (rc < 0)
                        return rc;
        }

        return 0;
}

/*
 * This is called when we're done sending the header.
 * Simply copy the data_segment to the send segment, and return.
 */
static int
iscsi_tcp_send_hdr_done(struct iscsi_tcp_conn *tcp_conn,
                        struct iscsi_segment *segment)
{
        tcp_conn->out.segment = tcp_conn->out.data_segment;
        debug_tcp("Header done. Next segment size %u total_size %u\n",
                  tcp_conn->out.segment.size, tcp_conn->out.segment.total_size);
        return 0;
}

static void
iscsi_tcp_send_hdr_prep(struct iscsi_conn *conn, void *hdr, size_t hdrlen)
{
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;

        debug_tcp("%s(%p%s)\n", __func__, tcp_conn,
                        conn->hdrdgst_en? ", digest enabled" : "");

        /* Clear the data segment - needs to be filled in by the
         * caller using iscsi_tcp_send_data_prep() */
        memset(&tcp_conn->out.data_segment, 0, sizeof(struct iscsi_segment));

        /* If header digest is enabled, compute the CRC and
         * place the digest into the same buffer. We make
         * sure that both iscsi_tcp_task and mtask have
         * sufficient room.
         */
        if (conn->hdrdgst_en) {
                iscsi_tcp_dgst_header(&tcp_conn->tx_hash, hdr, hdrlen,
                                      hdr + hdrlen);
                hdrlen += ISCSI_DIGEST_SIZE;
        }

        /* Remember header pointer for later, when we need
         * to decide whether there's a payload to go along
         * with the header. */
        tcp_conn->out.hdr = hdr;

        iscsi_segment_init_linear(&tcp_conn->out.segment, hdr, hdrlen,
                                iscsi_tcp_send_hdr_done, NULL);
}

/*
 * Prepare the send buffer for the payload data.
 * Padding and checksumming will all be taken care
 * of by the iscsi_segment routines.
 */
static int
iscsi_tcp_send_data_prep(struct iscsi_conn *conn, struct scatterlist *sg,
                         unsigned int count, unsigned int offset,
                         unsigned int len)
{
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
        struct hash_desc *tx_hash = NULL;
        unsigned int hdr_spec_len;

        debug_tcp("%s(%p, offset=%d, datalen=%d%s)\n", __func__,
                        tcp_conn, offset, len,
                        conn->datadgst_en? ", digest enabled" : "");

        /* Make sure the datalen matches what the caller
           said he would send. */
        hdr_spec_len = ntoh24(tcp_conn->out.hdr->dlength);
        WARN_ON(iscsi_padded(len) != iscsi_padded(hdr_spec_len));

        if (conn->datadgst_en)
                tx_hash = &tcp_conn->tx_hash;

        return iscsi_segment_seek_sg(&tcp_conn->out.data_segment,
                                   sg, count, offset, len,
                                   NULL, tx_hash);
}

static void
iscsi_tcp_send_linear_data_prepare(struct iscsi_conn *conn, void *data,
                                   size_t len)
{
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
        struct hash_desc *tx_hash = NULL;
        unsigned int hdr_spec_len;

        debug_tcp("%s(%p, datalen=%d%s)\n", __func__, tcp_conn, len,
                  conn->datadgst_en? ", digest enabled" : "");

        /* Make sure the datalen matches what the caller
           said he would send. */
        hdr_spec_len = ntoh24(tcp_conn->out.hdr->dlength);
        WARN_ON(iscsi_padded(len) != iscsi_padded(hdr_spec_len));

        if (conn->datadgst_en)
                tx_hash = &tcp_conn->tx_hash;

        iscsi_segment_init_linear(&tcp_conn->out.data_segment,
                                data, len, NULL, tx_hash);
}

static int iscsi_tcp_pdu_init(struct iscsi_task *task,
                              unsigned int offset, unsigned int count)
{
        struct iscsi_conn *conn = task->conn;
        int err = 0;

        iscsi_tcp_send_hdr_prep(conn, task->hdr, task->hdr_len);

        if (!count)
                return 0;

        if (!task->sc)
                iscsi_tcp_send_linear_data_prepare(conn, task->data, count);
        else {
                struct scsi_data_buffer *sdb = scsi_out(task->sc);

                err = iscsi_tcp_send_data_prep(conn, sdb->table.sgl,
                                               sdb->table.nents, offset, count);
        }

        if (err) {
                iscsi_conn_failure(conn, err);
                return -EIO;
        }
        return 0;
}

static int iscsi_tcp_pdu_alloc(struct iscsi_task *task)
{
        struct iscsi_tcp_task *tcp_task = task->dd_data;

        task->hdr = &tcp_task->hdr.hdrbuf;
        task->hdr_max = sizeof(tcp_task->hdr) - ISCSI_DIGEST_SIZE;
        return 0;
}

/**
 * iscsi_tcp_task - Initialize iSCSI SCSI_READ or SCSI_WRITE commands
 * @conn: iscsi connection
 * @task: scsi command task
 * @sc: scsi command
 **/
static int iscsi_tcp_task_init(struct iscsi_task *task)
{
        struct iscsi_tcp_task *tcp_task = task->dd_data;
        struct iscsi_conn *conn = task->conn;
        struct scsi_cmnd *sc = task->sc;
        int err;

        if (!sc) {
                /*
                 * mgmt tasks do not have a scatterlist since they come
                 * in from the iscsi interface.
                 */
                debug_scsi("mtask deq [cid %d itt 0x%x]\n", conn->id,
                           task->itt);

                return conn->session->tt->init_pdu(task, 0, task->data_count);
        }

        BUG_ON(__kfifo_len(tcp_task->r2tqueue));
        tcp_task->exp_datasn = 0;

        /* Prepare PDU, optionally w/ immediate data */
        debug_scsi("task deq [cid %d itt 0x%x imm %d unsol %d]\n",
                    conn->id, task->itt, task->imm_count,
                    task->unsol_r2t.data_length);

        err = conn->session->tt->init_pdu(task, 0, task->imm_count);
        if (err)
                return err;
        task->imm_count = 0;
        return 0;
}

static struct iscsi_r2t_info *iscsi_tcp_get_curr_r2t(struct iscsi_task *task)
{
        struct iscsi_session *session = task->conn->session;
        struct iscsi_tcp_task *tcp_task = task->dd_data;
        struct iscsi_r2t_info *r2t = NULL;

        if (iscsi_task_has_unsol_data(task))
                r2t = &task->unsol_r2t;
        else {
                spin_lock_bh(&session->lock);
                if (tcp_task->r2t) {
                        r2t = tcp_task->r2t;
                        /* Continue with this R2T? */
                        if (r2t->data_length <= r2t->sent) {
                                debug_scsi("  done with r2t %p\n", r2t);
                                __kfifo_put(tcp_task->r2tpool.queue,
                                            (void *)&tcp_task->r2t,
                                            sizeof(void *));
                                tcp_task->r2t = r2t = NULL;
                        }
                }

                if (r2t == NULL) {
                        __kfifo_get(tcp_task->r2tqueue,
                                    (void *)&tcp_task->r2t, sizeof(void *));
                        r2t = tcp_task->r2t;
                }
                spin_unlock_bh(&session->lock);
        }

        return r2t;
}

/*
 * iscsi_tcp_task_xmit - xmit normal PDU task
 * @task: iscsi command task
 *
 * We're expected to return 0 when everything was transmitted succesfully,
 * -EAGAIN if there's still data in the queue, or != 0 for any other kind
 * of error.
 */
static int iscsi_tcp_task_xmit(struct iscsi_task *task)
{
        struct iscsi_conn *conn = task->conn;
        struct iscsi_session *session = conn->session;
        struct iscsi_r2t_info *r2t;
        int rc = 0;

flush:
        /* Flush any pending data first. */
        rc = session->tt->xmit_pdu(task);
        if (rc < 0)
                return rc;

        /* mgmt command */
        if (!task->sc) {
                if (task->hdr->itt == RESERVED_ITT)
                        iscsi_put_task(task);
                return 0;
        }

        /* Are we done already? */
        if (task->sc->sc_data_direction != DMA_TO_DEVICE)
                return 0;

        r2t = iscsi_tcp_get_curr_r2t(task);
        if (r2t == NULL) {
                /* Waiting for more R2Ts to arrive. */
                debug_tcp("no R2Ts yet\n");
                return 0;
        }

        rc = conn->session->tt->alloc_pdu(task);
        if (rc)
                return rc;
        iscsi_prep_data_out_pdu(task, r2t, (struct iscsi_data *) task->hdr);

        debug_scsi("sol dout %p [dsn %d itt 0x%x doff %d dlen %d]\n",
                   r2t, r2t->datasn - 1, task->hdr->itt,
                   r2t->data_offset + r2t->sent, r2t->data_count);

        rc = conn->session->tt->init_pdu(task, r2t->data_offset + r2t->sent,
                                         r2t->data_count);
        if (rc)
                return rc;
        r2t->sent += r2t->data_count;
        goto flush;
}

static struct iscsi_cls_conn *
iscsi_tcp_conn_create(struct iscsi_cls_session *cls_session, uint32_t conn_idx)
{
        struct iscsi_conn *conn;
        struct iscsi_cls_conn *cls_conn;
        struct iscsi_tcp_conn *tcp_conn;

        cls_conn = iscsi_conn_setup(cls_session, sizeof(*tcp_conn), conn_idx);
        if (!cls_conn)
                return NULL;
        conn = cls_conn->dd_data;
        /*
         * due to strange issues with iser these are not set
         * in iscsi_conn_setup
         */
        conn->max_recv_dlength = ISCSI_DEF_MAX_RECV_SEG_LEN;

        tcp_conn = conn->dd_data;
        tcp_conn->iscsi_conn = conn;

        tcp_conn->tx_hash.tfm = crypto_alloc_hash("crc32c", 0,
                                                  CRYPTO_ALG_ASYNC);
        tcp_conn->tx_hash.flags = 0;
        if (IS_ERR(tcp_conn->tx_hash.tfm))
                goto free_conn;

        tcp_conn->rx_hash.tfm = crypto_alloc_hash("crc32c", 0,
                                                  CRYPTO_ALG_ASYNC);
        tcp_conn->rx_hash.flags = 0;
        if (IS_ERR(tcp_conn->rx_hash.tfm))
                goto free_tx_tfm;

        return cls_conn;

free_tx_tfm:
        crypto_free_hash(tcp_conn->tx_hash.tfm);
free_conn:
        iscsi_conn_printk(KERN_ERR, conn,
                          "Could not create connection due to crc32c "
                          "loading error. Make sure the crc32c "
                          "module is built as a module or into the "
                          "kernel\n");
        iscsi_conn_teardown(cls_conn);
        return NULL;
}

static void
iscsi_tcp_release_conn(struct iscsi_conn *conn)
{
        struct iscsi_session *session = conn->session;
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
        struct socket *sock = tcp_conn->sock;

        if (!sock)
                return;

        sock_hold(sock->sk);
        iscsi_conn_restore_callbacks(tcp_conn);
        sock_put(sock->sk);

        spin_lock_bh(&session->lock);
        tcp_conn->sock = NULL;
        spin_unlock_bh(&session->lock);
        sockfd_put(sock);
}

static void
iscsi_tcp_conn_destroy(struct iscsi_cls_conn *cls_conn)
{
        struct iscsi_conn *conn = cls_conn->dd_data;
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;

        iscsi_tcp_release_conn(conn);

        if (tcp_conn->tx_hash.tfm)
                crypto_free_hash(tcp_conn->tx_hash.tfm);
        if (tcp_conn->rx_hash.tfm)
                crypto_free_hash(tcp_conn->rx_hash.tfm);

        iscsi_conn_teardown(cls_conn);
}

static void
iscsi_tcp_conn_stop(struct iscsi_cls_conn *cls_conn, int flag)
{
        struct iscsi_conn *conn = cls_conn->dd_data;
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;

        /* userspace may have goofed up and not bound us */
        if (!tcp_conn->sock)
                return;
        /*
         * Make sure our recv side is stopped.
         * Older tools called conn stop before ep_disconnect
         * so IO could still be coming in.
         */
        write_lock_bh(&tcp_conn->sock->sk->sk_callback_lock);
        set_bit(ISCSI_SUSPEND_BIT, &conn->suspend_rx);
        write_unlock_bh(&tcp_conn->sock->sk->sk_callback_lock);

        iscsi_conn_stop(cls_conn, flag);
        iscsi_tcp_release_conn(conn);
}

static int iscsi_tcp_get_addr(struct iscsi_conn *conn, struct socket *sock,
                              char *buf, int *port,
                              int (*getname)(struct socket *, struct sockaddr *,
                                        int *addrlen))
{
        struct sockaddr_storage *addr;
        struct sockaddr_in6 *sin6;
        struct sockaddr_in *sin;
        int rc = 0, len;

        addr = kmalloc(sizeof(*addr), GFP_KERNEL);
        if (!addr)
                return -ENOMEM;

        if (getname(sock, (struct sockaddr *) addr, &len)) {
                rc = -ENODEV;
                goto free_addr;
        }

        switch (addr->ss_family) {
        case AF_INET:
                sin = (struct sockaddr_in *)addr;
                spin_lock_bh(&conn->session->lock);
                sprintf(buf, "%pI4", &sin->sin_addr.s_addr);
                *port = be16_to_cpu(sin->sin_port);
                spin_unlock_bh(&conn->session->lock);
                break;
        case AF_INET6:
                sin6 = (struct sockaddr_in6 *)addr;
                spin_lock_bh(&conn->session->lock);
                sprintf(buf, "%pI6", &sin6->sin6_addr);
                *port = be16_to_cpu(sin6->sin6_port);
                spin_unlock_bh(&conn->session->lock);
                break;
        }
free_addr:
        kfree(addr);
        return rc;
}

static int
iscsi_tcp_conn_bind(struct iscsi_cls_session *cls_session,
                    struct iscsi_cls_conn *cls_conn, uint64_t transport_eph,
                    int is_leading)
{
        struct Scsi_Host *shost = iscsi_session_to_shost(cls_session);
        struct iscsi_host *ihost = shost_priv(shost);
        struct iscsi_conn *conn = cls_conn->dd_data;
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
        struct sock *sk;
        struct socket *sock;
        int err;

        /* lookup for existing socket */
        sock = sockfd_lookup((int)transport_eph, &err);
        if (!sock) {
                iscsi_conn_printk(KERN_ERR, conn,
                                  "sockfd_lookup failed %d\n", err);
                return -EEXIST;
        }
        /*
         * copy these values now because if we drop the session
         * userspace may still want to query the values since we will
         * be using them for the reconnect
         */
        err = iscsi_tcp_get_addr(conn, sock, conn->portal_address,
                                 &conn->portal_port, kernel_getpeername);
        if (err)
                goto free_socket;

        err = iscsi_tcp_get_addr(conn, sock, ihost->local_address,
                                &ihost->local_port, kernel_getsockname);
        if (err)
                goto free_socket;

        err = iscsi_conn_bind(cls_session, cls_conn, is_leading);
        if (err)
                goto free_socket;

        /* bind iSCSI connection and socket */
        tcp_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;

        iscsi_tcp_conn_set_callbacks(conn);
        tcp_conn->sendpage = tcp_conn->sock->ops->sendpage;
        /*
         * set receive state machine into initial state
         */
        iscsi_tcp_hdr_recv_prep(tcp_conn);
        return 0;

free_socket:
        sockfd_put(sock);
        return err;
}

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_task *task = session->cmds[cmd_i];
                struct iscsi_tcp_task *tcp_task = task->dd_data;

                /*
                 * pre-allocated x2 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(&tcp_task->r2tpool,
                                    session->max_r2t * 2, NULL,
                                    sizeof(struct iscsi_r2t_info))) {
                        goto r2t_alloc_fail;
                }

                /* R2T xmit queue */
                tcp_task->r2tqueue = kfifo_alloc(
                      session->max_r2t * 4 * sizeof(void*), GFP_KERNEL, NULL);
                if (tcp_task->r2tqueue == ERR_PTR(-ENOMEM)) {
                        iscsi_pool_free(&tcp_task->r2tpool);
                        goto r2t_alloc_fail;
                }
        }

        return 0;

r2t_alloc_fail:
        for (i = 0; i < cmd_i; i++) {
                struct iscsi_task *task = session->cmds[i];
                struct iscsi_tcp_task *tcp_task = task->dd_data;

                kfifo_free(tcp_task->r2tqueue);
                iscsi_pool_free(&tcp_task->r2tpool);
        }
        return -ENOMEM;
}

static void
iscsi_r2tpool_free(struct iscsi_session *session)
{
        int i;

        for (i = 0; i < session->cmds_max; i++) {
                struct iscsi_task *task = session->cmds[i];
                struct iscsi_tcp_task *tcp_task = task->dd_data;

                kfifo_free(tcp_task->r2tqueue);
                iscsi_pool_free(&tcp_task->r2tpool);
        }
}

static int
iscsi_conn_set_param(struct iscsi_cls_conn *cls_conn, enum iscsi_param param,
                     char *buf, int buflen)
{
        struct iscsi_conn *conn = cls_conn->dd_data;
        struct iscsi_session *session = conn->session;
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
        int value;

        switch(param) {
        case ISCSI_PARAM_HDRDGST_EN:
                iscsi_set_param(cls_conn, param, buf, buflen);
                break;
        case ISCSI_PARAM_DATADGST_EN:
                iscsi_set_param(cls_conn, param, buf, buflen);
                tcp_conn->sendpage = conn->datadgst_en ?
                        sock_no_sendpage : tcp_conn->sock->ops->sendpage;
                break;
        case ISCSI_PARAM_MAX_R2T:
                sscanf(buf, "%d", &value);
                if (value <= 0 || !is_power_of_2(value))
                        return -EINVAL;
                if (session->max_r2t == value)
                        break;
                iscsi_r2tpool_free(session);
                iscsi_set_param(cls_conn, param, buf, buflen);
                if (iscsi_r2tpool_alloc(session))
                        return -ENOMEM;
                break;
        default:
                return iscsi_set_param(cls_conn, param, buf, buflen);
        }

        return 0;
}

static int
iscsi_tcp_conn_get_param(struct iscsi_cls_conn *cls_conn,
                         enum iscsi_param param, char *buf)
{
        struct iscsi_conn *conn = cls_conn->dd_data;
        int len;

        switch(param) {
        case ISCSI_PARAM_CONN_PORT:
                spin_lock_bh(&conn->session->lock);
                len = sprintf(buf, "%hu\n", conn->portal_port);
                spin_unlock_bh(&conn->session->lock);
                break;
        case ISCSI_PARAM_CONN_ADDRESS:
                spin_lock_bh(&conn->session->lock);
                len = sprintf(buf, "%s\n", conn->portal_address);
                spin_unlock_bh(&conn->session->lock);
                break;
        default:
                return iscsi_conn_get_param(cls_conn, param, buf);
        }

        return len;
}

static void
iscsi_conn_get_stats(struct iscsi_cls_conn *cls_conn, struct iscsi_stats *stats)
{
        struct iscsi_conn *conn = cls_conn->dd_data;
        struct iscsi_tcp_conn *tcp_conn = conn->dd_data;

        stats->txdata_octets = conn->txdata_octets;
        stats->rxdata_octets = conn->rxdata_octets;
        stats->scsicmd_pdus = conn->scsicmd_pdus_cnt;
        stats->dataout_pdus = conn->dataout_pdus_cnt;
        stats->scsirsp_pdus = conn->scsirsp_pdus_cnt;
        stats->datain_pdus = conn->datain_pdus_cnt;
        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 = tcp_conn->sendpage_failures_cnt;
        strcpy(stats->custom[1].desc, "rx_discontiguous_hdr");
        stats->custom[1].value = tcp_conn->discontiguous_hdr_cnt;
        strcpy(stats->custom[2].desc, "eh_abort_cnt");
        stats->custom[2].value = conn->eh_abort_cnt;
}

static struct iscsi_cls_session *
iscsi_tcp_session_create(struct iscsi_endpoint *ep, uint16_t cmds_max,
                         uint16_t qdepth, uint32_t initial_cmdsn,
                         uint32_t *hostno)
{
        struct iscsi_cls_session *cls_session;
        struct iscsi_session *session;
        struct Scsi_Host *shost;

        if (ep) {
                printk(KERN_ERR "iscsi_tcp: invalid ep %p.\n", ep);
                return NULL;
        }

        shost = iscsi_host_alloc(&iscsi_sht, 0, qdepth);
        if (!shost)
                return NULL;
        shost->transportt = iscsi_tcp_scsi_transport;
        shost->max_lun = iscsi_max_lun;
        shost->max_id = 0;
        shost->max_channel = 0;
        shost->max_cmd_len = SCSI_MAX_VARLEN_CDB_SIZE;

        if (iscsi_host_add(shost, NULL))
                goto free_host;
        *hostno = shost->host_no;

        cls_session = iscsi_session_setup(&iscsi_tcp_transport, shost, cmds_max,
                                          sizeof(struct iscsi_tcp_task),
                                          initial_cmdsn, 0);
        if (!cls_session)
                goto remove_host;
        session = cls_session->dd_data;

        shost->can_queue = session->scsi_cmds_max;
        if (iscsi_r2tpool_alloc(session))
                goto remove_session;
        return cls_session;

remove_session:
        iscsi_session_teardown(cls_session);
remove_host:
        iscsi_host_remove(shost);
free_host:
        iscsi_host_free(shost);
        return NULL;
}

static void iscsi_tcp_session_destroy(struct iscsi_cls_session *cls_session)
{
        struct Scsi_Host *shost = iscsi_session_to_shost(cls_session);

        iscsi_r2tpool_free(cls_session->dd_data);
        iscsi_session_teardown(cls_session);

        iscsi_host_remove(shost);
        iscsi_host_free(shost);
}

static int iscsi_tcp_slave_configure(struct scsi_device *sdev)
{
        blk_queue_bounce_limit(sdev->request_queue, BLK_BOUNCE_ANY);
        blk_queue_dma_alignment(sdev->request_queue, 0);
        return 0;
}

static struct scsi_host_template iscsi_sht = {
        .module                 = THIS_MODULE,
        .name                   = "iSCSI Initiator over TCP/IP",
        .queuecommand           = iscsi_queuecommand,
        .change_queue_depth     = iscsi_change_queue_depth,
        .can_queue              = ISCSI_DEF_XMIT_CMDS_MAX - 1,
        .sg_tablesize           = 4096,
        .max_sectors            = 0xFFFF,
        .cmd_per_lun            = ISCSI_DEF_CMD_PER_LUN,
        .eh_abort_handler       = iscsi_eh_abort,
        .eh_device_reset_handler= iscsi_eh_device_reset,
        .eh_target_reset_handler= iscsi_eh_target_reset,
        .use_clustering         = DISABLE_CLUSTERING,
        .slave_configure        = iscsi_tcp_slave_configure,
        .proc_name              = "iscsi_tcp",
        .this_id                = -1,
};

static struct iscsi_transport iscsi_tcp_transport = {
        .owner                  = THIS_MODULE,
        .name                   = "tcp",
        .caps                   = CAP_RECOVERY_L0 | CAP_MULTI_R2T | CAP_HDRDGST
                                  | CAP_DATADGST,
        .param_mask             = ISCSI_MAX_RECV_DLENGTH |
                                  ISCSI_MAX_XMIT_DLENGTH |
                                  ISCSI_HDRDGST_EN |
                                  ISCSI_DATADGST_EN |
                                  ISCSI_INITIAL_R2T_EN |
                                  ISCSI_MAX_R2T |
                                  ISCSI_IMM_DATA_EN |
                                  ISCSI_FIRST_BURST |
                                  ISCSI_MAX_BURST |
                                  ISCSI_PDU_INORDER_EN |
                                  ISCSI_DATASEQ_INORDER_EN |
                                  ISCSI_ERL |
                                  ISCSI_CONN_PORT |
                                  ISCSI_CONN_ADDRESS |
                                  ISCSI_EXP_STATSN |
                                  ISCSI_PERSISTENT_PORT |
                                  ISCSI_PERSISTENT_ADDRESS |
                                  ISCSI_TARGET_NAME | ISCSI_TPGT |
                                  ISCSI_USERNAME | ISCSI_PASSWORD |
                                  ISCSI_USERNAME_IN | ISCSI_PASSWORD_IN |
                                  ISCSI_FAST_ABORT | ISCSI_ABORT_TMO |
                                  ISCSI_LU_RESET_TMO |
                                  ISCSI_PING_TMO | ISCSI_RECV_TMO |
                                  ISCSI_IFACE_NAME | ISCSI_INITIATOR_NAME,
        .host_param_mask        = ISCSI_HOST_HWADDRESS | ISCSI_HOST_IPADDRESS |
                                  ISCSI_HOST_INITIATOR_NAME |
                                  ISCSI_HOST_NETDEV_NAME,
        /* session management */
        .create_session         = iscsi_tcp_session_create,
        .destroy_session        = iscsi_tcp_session_destroy,
        /* connection management */
        .create_conn            = iscsi_tcp_conn_create,
        .bind_conn              = iscsi_tcp_conn_bind,
        .destroy_conn           = iscsi_tcp_conn_destroy,
        .set_param              = iscsi_conn_set_param,
        .get_conn_param         = iscsi_tcp_conn_get_param,
        .get_session_param      = iscsi_session_get_param,
        .start_conn             = iscsi_conn_start,
        .stop_conn              = iscsi_tcp_conn_stop,
        /* iscsi host params */
        .get_host_param         = iscsi_host_get_param,
        .set_host_param         = iscsi_host_set_param,
        /* IO */
        .send_pdu               = iscsi_conn_send_pdu,
        .get_stats              = iscsi_conn_get_stats,
        /* iscsi task/cmd helpers */
        .init_task              = iscsi_tcp_task_init,
        .xmit_task              = iscsi_tcp_task_xmit,
        .cleanup_task           = iscsi_tcp_cleanup_task,
        /* low level pdu helpers */
        .xmit_pdu               = iscsi_tcp_flush,
        .init_pdu               = iscsi_tcp_pdu_init,
        .alloc_pdu              = iscsi_tcp_pdu_alloc,
        /* recovery */
        .session_recovery_timedout = iscsi_session_recovery_timedout,
};

static int __init
iscsi_tcp_init(void)
{
        if (iscsi_max_lun < 1) {
                printk(KERN_ERR "iscsi_tcp: Invalid max_lun value of %u\n",
                       iscsi_max_lun);
                return -EINVAL;
        }

        iscsi_tcp_scsi_transport = iscsi_register_transport(
                                                        &iscsi_tcp_transport);
        if (!iscsi_tcp_scsi_transport)
                return -ENODEV;

        return 0;
}

static void __exit
iscsi_tcp_exit(void)
{
        iscsi_unregister_transport(&iscsi_tcp_transport);
}

module_init(iscsi_tcp_init);
module_exit(iscsi_tcp_exit);