[SCSI] iscsi_tcp: split module into lib and lld
Mike Christie [Tue, 2 Dec 2008 06:32:11 +0000 (00:32 -0600)]
As explained in the previous mails, cxgb3i needs iscsi_tcp's
r2t/data_out and data_in procesing so this just moves functions
that both drivers want to use to a new module libiscsi_tcp. The
next patch will hook iscsi_tcp in.

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

drivers/scsi/libiscsi_tcp.c [new file with mode: 0644]
include/scsi/libiscsi_tcp.h [new file with mode: 0644]

diff --git a/drivers/scsi/libiscsi_tcp.c b/drivers/scsi/libiscsi_tcp.c
new file mode 100644 (file)
index 0000000..e865089
--- /dev/null
@@ -0,0 +1,1160 @@
+/*
+ * iSCSI over TCP/IP Data-Path lib
+ *
+ * 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("Mike Christie <michaelc@cs.wisc.edu>, "
+             "Dmitry Yusupov <dmitry_yus@yahoo.com>, "
+             "Alex Aizman <itn780@yahoo.com>");
+MODULE_DESCRIPTION("iSCSI/TCP data-path");
+MODULE_LICENSE("GPL");
+#undef DEBUG_TCP
+
+#ifdef DEBUG_TCP
+#define debug_tcp(fmt...) printk(KERN_INFO "tcp: " fmt)
+#else
+#define debug_tcp(fmt...)
+#endif
+
+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 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;
+}
+
+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;
+       }
+}
+EXPORT_SYMBOL_GPL(iscsi_tcp_segment_unmap);
+
+/*
+ * 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.
+ */
+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;
+}
+EXPORT_SYMBOL_GPL(iscsi_tcp_segment_done);
+
+/**
+ * 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;
+}
+
+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);
+}
+EXPORT_SYMBOL_GPL(iscsi_tcp_dgst_header);
+
+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);
+       }
+}
+
+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;
+}
+EXPORT_SYMBOL_GPL(iscsi_segment_init_linear);
+
+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;
+}
+EXPORT_SYMBOL_GPL(iscsi_segment_seek_sg);
+
+/**
+ * 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.
+ */
+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);
+}
+EXPORT_SYMBOL_GPL(iscsi_tcp_hdr_recv_prep);
+
+/*
+ * 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);
+}
+
+/**
+ * iscsi_tcp_cleanup_task - free tcp_task resources
+ * @task: iscsi task
+ *
+ * must be called with session lock
+ */
+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;
+       }
+}
+EXPORT_SYMBOL_GPL(iscsi_tcp_cleanup_task);
+
+/**
+ * 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);
+}
+
+/**
+ * iscsi_tcp_recv_segment_is_hdr - tests if we are reading in a header
+ * @tcp_conn: iscsi tcp conn
+ *
+ * returns non zero if we are currently processing or setup to process
+ * a header.
+ */
+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;
+}
+EXPORT_SYMBOL_GPL(iscsi_tcp_recv_segment_is_hdr);
+
+/**
+ * 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
+ *
+ * Will return status of transfer in status. And will return
+ * number of bytes copied.
+ */
+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_task_init - Initialize iSCSI SCSI_READ or SCSI_WRITE commands
+ * @conn: iscsi connection
+ * @task: scsi command task
+ * @sc: scsi command
+ */
+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;
+}
+EXPORT_SYMBOL_GPL(iscsi_tcp_task_init);
+
+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.
+ */
+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;
+}
+EXPORT_SYMBOL_GPL(iscsi_tcp_task_xmit);
+
+struct iscsi_cls_conn *
+iscsi_tcp_conn_setup(struct iscsi_cls_session *cls_session, int dd_data_size,
+                     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->dd_data = kzalloc(dd_data_size, GFP_KERNEL);
+       if (!tcp_conn->dd_data) {
+               iscsi_conn_teardown(cls_conn);
+               return NULL;
+       }
+       return cls_conn;
+}
+EXPORT_SYMBOL_GPL(iscsi_tcp_conn_setup);
+
+void iscsi_tcp_conn_teardown(struct iscsi_cls_conn *cls_conn)
+{
+       struct iscsi_conn *conn = cls_conn->dd_data;
+       struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
+
+       kfree(tcp_conn->dd_data);
+       iscsi_conn_teardown(cls_conn);
+}
+EXPORT_SYMBOL_GPL(iscsi_tcp_conn_teardown);
+
+int iscsi_tcp_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;
+}
+EXPORT_SYMBOL_GPL(iscsi_tcp_r2tpool_alloc);
+
+void iscsi_tcp_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);
+       }
+}
+EXPORT_SYMBOL_GPL(iscsi_tcp_r2tpool_free);
+
+void iscsi_tcp_conn_get_stats(struct iscsi_cls_conn *cls_conn,
+                             struct iscsi_stats *stats)
+{
+       struct iscsi_conn *conn = cls_conn->dd_data;
+
+       stats->txdata_octets = conn->txdata_octets;
+       stats->rxdata_octets = conn->rxdata_octets;
+       stats->scsicmd_pdus = conn->scsicmd_pdus_cnt;
+       stats->dataout_pdus = conn->dataout_pdus_cnt;
+       stats->scsirsp_pdus = conn->scsirsp_pdus_cnt;
+       stats->datain_pdus = conn->datain_pdus_cnt;
+       stats->r2t_pdus = conn->r2t_pdus_cnt;
+       stats->tmfcmd_pdus = conn->tmfcmd_pdus_cnt;
+       stats->tmfrsp_pdus = conn->tmfrsp_pdus_cnt;
+}
+EXPORT_SYMBOL_GPL(iscsi_tcp_conn_get_stats);
diff --git a/include/scsi/libiscsi_tcp.h b/include/scsi/libiscsi_tcp.h
new file mode 100644 (file)
index 0000000..e6bf8ef
--- /dev/null
@@ -0,0 +1,131 @@
+/*
+ * iSCSI over TCP/IP Data-Path lib
+ *
+ * Copyright (C) 2008 Mike Christie
+ * Copyright (C) 2008 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.
+ */
+
+#ifndef LIBISCSI_TCP_H
+#define LIBISCSI_TCP_H
+
+#include <scsi/libiscsi.h>
+
+struct iscsi_tcp_conn;
+struct iscsi_segment;
+struct sk_buff;
+struct hash_desc;
+
+typedef int iscsi_segment_done_fn_t(struct iscsi_tcp_conn *,
+                                   struct iscsi_segment *);
+
+struct iscsi_segment {
+       unsigned char           *data;
+       unsigned int            size;
+       unsigned int            copied;
+       unsigned int            total_size;
+       unsigned int            total_copied;
+
+       struct hash_desc        *hash;
+       unsigned char           recv_digest[ISCSI_DIGEST_SIZE];
+       unsigned char           digest[ISCSI_DIGEST_SIZE];
+       unsigned int            digest_len;
+
+       struct scatterlist      *sg;
+       void                    *sg_mapped;
+       unsigned int            sg_offset;
+
+       iscsi_segment_done_fn_t *done;
+};
+
+/* Socket connection recieve helper */
+struct iscsi_tcp_recv {
+       struct iscsi_hdr        *hdr;
+       struct iscsi_segment    segment;
+
+       /* Allocate buffer for BHS + AHS */
+       uint32_t                hdr_buf[64];
+
+       /* copied and flipped values */
+       int                     datalen;
+};
+
+struct iscsi_tcp_conn {
+       struct iscsi_conn       *iscsi_conn;
+       void                    *dd_data;
+       int                     stop_stage;     /* conn_stop() flag: *
+                                                * stop to recover,  *
+                                                * stop to terminate */
+       /* control data */
+       struct iscsi_tcp_recv   in;             /* TCP receive context */
+       /* CRC32C (Rx) LLD should set this is they do not offload */
+       struct hash_desc        *rx_hash;
+};
+
+struct iscsi_tcp_task {
+       uint32_t                exp_datasn;     /* expected target's R2TSN/DataSN */
+       int                     data_offset;
+       struct iscsi_r2t_info   *r2t;           /* in progress solict R2T */
+       struct iscsi_pool       r2tpool;
+       struct kfifo            *r2tqueue;
+       void                    *dd_data;
+};
+
+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 */
+};
+
+extern void iscsi_tcp_hdr_recv_prep(struct iscsi_tcp_conn *tcp_conn);
+extern int iscsi_tcp_recv_skb(struct iscsi_conn *conn, struct sk_buff *skb,
+                             unsigned int offset, bool offloaded, int *status);
+extern void iscsi_tcp_cleanup_task(struct iscsi_task *task);
+extern int iscsi_tcp_task_init(struct iscsi_task *task);
+extern int iscsi_tcp_task_xmit(struct iscsi_task *task);
+
+/* segment helpers */
+extern int iscsi_tcp_recv_segment_is_hdr(struct iscsi_tcp_conn *tcp_conn);
+extern int iscsi_tcp_segment_done(struct iscsi_segment *segment, int recv,
+                                 unsigned copied);
+extern void iscsi_tcp_segment_unmap(struct iscsi_segment *segment);
+
+extern void iscsi_segment_init_linear(struct iscsi_segment *segment,
+                                     void *data, size_t size,
+                                     iscsi_segment_done_fn_t *done,
+                                     struct hash_desc *hash);
+extern 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);
+
+/* digest helpers */
+extern void iscsi_tcp_dgst_header(struct hash_desc *hash, const void *hdr,
+                                 size_t hdrlen,
+                                 unsigned char digest[ISCSI_DIGEST_SIZE]);
+extern struct iscsi_cls_conn *
+iscsi_tcp_conn_setup(struct iscsi_cls_session *cls_session, int dd_data_size,
+                    uint32_t conn_idx);
+extern void iscsi_tcp_conn_teardown(struct iscsi_cls_conn *cls_conn);
+
+/* misc helpers */
+extern int iscsi_tcp_r2tpool_alloc(struct iscsi_session *session);
+extern void iscsi_tcp_r2tpool_free(struct iscsi_session *session);
+
+extern void iscsi_tcp_conn_get_stats(struct iscsi_cls_conn *cls_conn,
+                                    struct iscsi_stats *stats);
+#endif /* LIBISCSI_TCP_H */