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

#ifndef DEBUG_ASSERT
#ifdef BUG_ON
#undef BUG_ON
#endif
#define BUG_ON(expr)
#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)
		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_ctask(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
{
	struct iscsi_tcp_cmd_task *tcp_ctask = ctask->dd_data;
	struct iscsi_r2t_info *r2t;

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

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

/**
 * iscsi_data_rsp - SCSI Data-In Response processing
 * @conn: iscsi connection
 * @ctask: scsi command task
 **/
static int
iscsi_data_rsp(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
{
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
	struct iscsi_tcp_cmd_task *tcp_ctask = ctask->dd_data;
	struct iscsi_data_rsp *rhdr = (struct iscsi_data_rsp *)tcp_conn->in.hdr;
	struct iscsi_session *session = conn->session;
	struct scsi_cmnd *sc = ctask->sc;
	int datasn = be32_to_cpu(rhdr->datasn);
	unsigned total_in_length = scsi_in(sc)->length;

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

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

	tcp_ctask->exp_datasn++;

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

	if (rhdr->flags & ISCSI_FLAG_DATA_STATUS) {
		sc->result = (DID_OK << 16) | rhdr->cmd_status;
		conn->exp_statsn = be32_to_cpu(rhdr->statsn) + 1;
		if (rhdr->flags & (ISCSI_FLAG_DATA_UNDERFLOW |
		                   ISCSI_FLAG_DATA_OVERFLOW)) {
			int res_count = be32_to_cpu(rhdr->residual_count);

			if (res_count > 0 &&
			    (rhdr->flags & ISCSI_FLAG_CMD_OVERFLOW ||
			     res_count <= total_in_length))
				scsi_in(sc)->resid = res_count;
			else
				sc->result = (DID_BAD_TARGET << 16) |
					rhdr->cmd_status;
		}
	}

	conn->datain_pdus_cnt++;
	return 0;
}

/**
 * iscsi_solicit_data_init - initialize first Data-Out
 * @conn: iscsi connection
 * @ctask: scsi command task
 * @r2t: R2T info
 *
 * Notes:
 *	Initialize first Data-Out within this R2T sequence and finds
 *	proper data_offset within this SCSI command.
 *
 *	This function is called with connection lock taken.
 **/
static void
iscsi_solicit_data_init(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask,
			struct iscsi_r2t_info *r2t)
{
	struct iscsi_data *hdr;

	hdr = &r2t->dtask.hdr;
	memset(hdr, 0, sizeof(struct iscsi_data));
	hdr->ttt = r2t->ttt;
	hdr->datasn = cpu_to_be32(r2t->solicit_datasn);
	r2t->solicit_datasn++;
	hdr->opcode = ISCSI_OP_SCSI_DATA_OUT;
	memcpy(hdr->lun, ctask->hdr->lun, sizeof(hdr->lun));
	hdr->itt = ctask->hdr->itt;
	hdr->exp_statsn = r2t->exp_statsn;
	hdr->offset = cpu_to_be32(r2t->data_offset);
	if (r2t->data_length > conn->max_xmit_dlength) {
		hton24(hdr->dlength, conn->max_xmit_dlength);
		r2t->data_count = conn->max_xmit_dlength;
		hdr->flags = 0;
	} else {
		hton24(hdr->dlength, r2t->data_length);
		r2t->data_count = r2t->data_length;
		hdr->flags = ISCSI_FLAG_CMD_FINAL;
	}
	conn->dataout_pdus_cnt++;

	r2t->sent = 0;
}

/**
 * iscsi_r2t_rsp - iSCSI R2T Response processing
 * @conn: iscsi connection
 * @ctask: scsi command task
 **/
static int
iscsi_r2t_rsp(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask)
{
	struct iscsi_r2t_info *r2t;
	struct iscsi_session *session = conn->session;
	struct iscsi_tcp_cmd_task *tcp_ctask = ctask->dd_data;
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
	struct iscsi_r2t_rsp *rhdr = (struct iscsi_r2t_rsp *)tcp_conn->in.hdr;
	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_ctask->exp_datasn != r2tsn){
		debug_tcp("%s: ctask->exp_datasn(%d) != rhdr->r2tsn(%d)\n",
		          __FUNCTION__, tcp_ctask->exp_datasn, r2tsn);
		return ISCSI_ERR_R2TSN;
	}

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

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

	rc = __kfifo_get(tcp_ctask->r2tpool.queue, (void*)&r2t, sizeof(void*));
	BUG_ON(!rc);

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

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

	iscsi_solicit_data_init(conn, ctask, r2t);

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

	iscsi_requeue_ctask(ctask);
	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_session *session = conn->session;
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
	struct iscsi_cmd_task *ctask;
	uint32_t itt;

	/* 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:
		ctask = session->cmds[itt];
		spin_lock(&conn->session->lock);
		rc = iscsi_data_rsp(conn, ctask);
		spin_unlock(&conn->session->lock);
		if (rc)
			return rc;
		if (tcp_conn->in.datalen) {
			struct iscsi_tcp_cmd_task *tcp_ctask = ctask->dd_data;
			struct hash_desc *rx_hash = NULL;
			struct scsi_data_buffer *sdb = scsi_in(ctask->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)
				rx_hash = &tcp_conn->rx_hash;

			debug_tcp("iscsi_tcp_begin_data_in(%p, offset=%d, "
				  "datalen=%d)\n", tcp_conn,
				  tcp_ctask->data_offset,
				  tcp_conn->in.datalen);
			return iscsi_segment_seek_sg(&tcp_conn->in.segment,
						     sdb->table.sgl,
						     sdb->table.nents,
						     tcp_ctask->data_offset,
						     tcp_conn->in.datalen,
						     iscsi_tcp_process_data_in,
						     rx_hash);
		}
		/* fall through */
	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:
		ctask = session->cmds[itt];
		if (ahslen)
			rc = ISCSI_ERR_AHSLEN;
		else if (ctask->sc->sc_data_direction == DMA_TO_DEVICE) {
			spin_lock(&session->lock);
			rc = iscsi_r2t_rsp(conn, ctask);
			spin_unlock(&session->lock);
		} else
			rc = ISCSI_ERR_PROTO;
		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) {
			iscsi_tcp_segment_splice_digest(segment,
							segment->recv_digest);
			return 0;
		}
		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 - 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;
	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);
		return 0;
	}

	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);
			break;
		}
		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) {
			debug_tcp("segment done\n");
			rc = segment->done(tcp_conn, segment);
			if (rc != 0) {
				skb_abort_seq_read(&seq);
				goto error;
			}

			/* The done() functions sets up the
			 * next segment. */
		}
	}
	skb_abort_seq_read(&seq);
	conn->rxdata_octets += consumed;
	return consumed;

error:
	debug_tcp("Error receiving PDU, errno=%d\n", rc);
	iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
	return 0;
}

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);