Skip to content
Snippets Groups Projects
iscsi_tcp.c 60.1 KiB
Newer Older
/*
 * 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 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_chunk *chunk);

static inline void
iscsi_buf_init_iov(struct iscsi_buf *ibuf, char *vbuf, int size)
{
	ibuf->sg.page = virt_to_page(vbuf);
	ibuf->sg.offset = offset_in_page(vbuf);
	ibuf->sg.length = size;
	ibuf->use_sendmsg = 1;
}

static inline void
iscsi_buf_init_sg(struct iscsi_buf *ibuf, struct scatterlist *sg)
{
	ibuf->sg.page = sg->page;
	ibuf->sg.offset = sg->offset;
	ibuf->sg.length = sg->length;
	/*
	 * Fastpath: sg element fits into single page
	 */
	if (sg->length + sg->offset <= PAGE_SIZE && !PageSlab(sg->page))
		ibuf->use_sendmsg = 0;
	else
		ibuf->use_sendmsg = 1;
	ibuf->sent = 0;
}

static inline int
iscsi_buf_left(struct iscsi_buf *ibuf)
{
	int rc;

	rc = ibuf->sg.length - ibuf->sent;
	BUG_ON(rc < 0);
	return rc;
}

static inline void
iscsi_hdr_digest(struct iscsi_conn *conn, struct iscsi_buf *buf,
		 u8* crc)
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
	crypto_hash_digest(&tcp_conn->tx_hash, &buf->sg, buf->sg.length, crc);
	buf->sg.length += ISCSI_DIGEST_SIZE;
/*
 * Scatterlist handling: inside the iscsi_chunk, 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_chunk_init_sg - init indicated scatterlist entry
 * @chunk: the buffer object
 * @idx: index into scatterlist
 * @offset: byte offset into that sg entry
 *
 * This function sets up the chunk so that subsequent
 * data is copied to the indicated sg entry, at the given
 * offset.
 */
static inline void
iscsi_tcp_chunk_init_sg(struct iscsi_chunk *chunk,
			unsigned int idx, unsigned int offset)
{
	struct scatterlist *sg;

	BUG_ON(chunk->sg == NULL);

	sg = &chunk->sg[idx];
	chunk->sg_index = idx;
	chunk->sg_offset = offset;
	chunk->size = min(sg->length - offset, chunk->total_size);
	chunk->data = NULL;
}

/**
 * iscsi_tcp_chunk_map - map the current S/G page
 * @chunk: iscsi chunk
 *
 * 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_chunk_map(struct iscsi_chunk *chunk)
{
	struct scatterlist *sg;

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

	sg = &chunk->sg[chunk->sg_index];
	BUG_ON(chunk->sg_mapped);
	BUG_ON(sg->length == 0);
	chunk->sg_mapped = kmap_atomic(sg->page, KM_SOFTIRQ0);
	chunk->data = chunk->sg_mapped + sg->offset + chunk->sg_offset;
}

static inline void
iscsi_tcp_chunk_unmap(struct iscsi_chunk *chunk)
{
	if (chunk->sg_mapped) {
		kunmap_atomic(chunk->sg_mapped, KM_SOFTIRQ0);
		chunk->sg_mapped = NULL;
		chunk->data = NULL;
	}
}

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

/**
 * iscsi_tcp_chunk_done - check whether the chunk is complete
 * @chunk: iscsi chunk to check
 *
 * Check if we're done receiving this chunk. 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.
 */
iscsi_tcp_chunk_done(struct iscsi_chunk *chunk)
	static unsigned char padbuf[ISCSI_PAD_LEN];
	unsigned int pad;

	if (chunk->copied < chunk->size) {
		iscsi_tcp_chunk_map(chunk);
		return 0;
	}
	chunk->total_copied += chunk->copied;
	chunk->copied = 0;
	chunk->size = 0;
	/* Unmap the current scatterlist page, if there is one. */
	iscsi_tcp_chunk_unmap(chunk);

	/* Do we have more scatterlist entries? */
	if (chunk->total_copied < chunk->total_size) {
		/* Proceed to the next entry in the scatterlist. */
		iscsi_tcp_chunk_init_sg(chunk, chunk->sg_index + 1, 0);
		iscsi_tcp_chunk_map(chunk);
		BUG_ON(chunk->size == 0);
		return 0;
	}

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

	/*
	 * Set us up for receiving the data digest. hdr digest
	 * is completely handled in hdr done function.
	 */
	if (chunk->hash) {
		if (chunk->digest_len == 0) {
			crypto_hash_final(chunk->hash, chunk->digest);
			iscsi_tcp_chunk_splice_digest(chunk,
						      chunk->recv_digest);
			return 0;
/**
 * iscsi_tcp_chunk_recv - copy data to chunk
 * @tcp_conn: the iSCSI TCP connection
 * @chunk: 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_chunk_recv(struct iscsi_tcp_conn *tcp_conn,
		     struct iscsi_chunk *chunk, const void *ptr,
		     unsigned int len)
{
	struct scatterlist sg;
	unsigned int copy, copied = 0;
	while (!iscsi_tcp_chunk_done(chunk)) {
		if (copied == len)
			goto out;
		copy = min(len - copied, chunk->size - chunk->copied);
		memcpy(chunk->data + chunk->copied, ptr + copied, copy);
		if (chunk->hash) {
			sg_init_one(&sg, ptr + copied, copy);
			crypto_hash_update(chunk->hash, &sg, copy);
		}
		chunk->copied += copy;
		copied += copy;
	}

out:
	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_chunk *chunk)
{
	if (!chunk->digest_len)
		return 1;

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

	return 1;
}

/*
 * Helper function to set up chunk buffer
 */
static inline void
__iscsi_chunk_init(struct iscsi_chunk *chunk, size_t size,
		   iscsi_chunk_done_fn_t *done, struct hash_desc *hash)
{
	memset(chunk, 0, sizeof(*chunk));
	chunk->total_size = size;
	chunk->done = done;

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

static inline void
iscsi_chunk_init_linear(struct iscsi_chunk *chunk, void *data, size_t size,
			iscsi_chunk_done_fn_t *done, struct hash_desc *hash)
{
	__iscsi_chunk_init(chunk, size, done, hash);
	chunk->data = data;
	chunk->size = size;
}

static inline int
iscsi_chunk_seek_sg(struct iscsi_chunk *chunk,
		    struct scatterlist *sg, unsigned int sg_count,
		    unsigned int offset, size_t size,
		    iscsi_chunk_done_fn_t *done, struct hash_desc *hash)
{
	unsigned int i;

	__iscsi_chunk_init(chunk, size, done, hash);
	for (i = 0; i < sg_count; ++i) {
		if (offset < sg[i].length) {
			chunk->sg = sg;
			chunk->sg_count = sg_count;
			iscsi_tcp_chunk_init_sg(chunk, i, offset);
			return 0;
		offset -= sg[i].length;
	return ISCSI_ERR_DATA_OFFSET;
}

/**
 * iscsi_tcp_hdr_recv_prep - prep chunk 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_chunk_init_linear(&tcp_conn->in.chunk,
				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_chunk *chunk)
{
	struct iscsi_conn *conn = tcp_conn->iscsi_conn;
	int rc = 0;

	if (!iscsi_tcp_dgst_verify(tcp_conn, chunk))
		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);
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_chunk_init_linear(&tcp_conn->in.chunk,
				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;
	struct scsi_cmnd *sc;
	/* 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");
	}

	sc = ctask->sc;
	if (unlikely(!sc))
	tcp_ctask->xmstate = XMSTATE_IDLE;
	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;
	int datasn = be32_to_cpu(rhdr->datasn);

	iscsi_update_cmdsn(session, (struct iscsi_nopin*)rhdr);
	/*
	 * setup Data-In byte counter (gets decremented..)
	 */
	ctask->data_count = tcp_conn->in.datalen;
	if (tcp_conn->in.datalen == 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->data_offset = be32_to_cpu(rhdr->offset);
	if (tcp_ctask->data_offset + tcp_conn->in.datalen > scsi_bufflen(sc)) {
		debug_tcp("%s: data_offset(%d) + data_len(%d) > total_length_in(%d)\n",
		          __FUNCTION__, tcp_ctask->data_offset,
		          tcp_conn->in.datalen, scsi_bufflen(sc));
		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 <= scsi_bufflen(sc)))
				scsi_set_resid(sc, res_count);
				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;
	struct scsi_cmnd *sc = ctask->sc;
	int i, sg_count = 0;
	struct scatterlist *sg;
	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_buf_init_iov(&r2t->headbuf, (char*)hdr,
			   sizeof(struct iscsi_hdr));
	sg = scsi_sglist(sc);
	r2t->sg = NULL;
	for (i = 0; i < scsi_sg_count(sc); i++, sg += 1) {
		/* FIXME: prefetch ? */
		if (sg_count + sg->length > r2t->data_offset) {
			int page_offset;
			/* offset within this page */
			page_offset = r2t->data_offset - sg_count;
			/* fill in this buffer */
			iscsi_buf_init_sg(&r2t->sendbuf, sg);
			r2t->sendbuf.sg.offset += page_offset;
			r2t->sendbuf.sg.length -= page_offset;
			/* xmit logic will continue with next one */
			r2t->sg = sg + 1;
			break;
}

/**
 * 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) {
		printk(KERN_ERR "iscsi_tcp: 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 */
	spin_lock(&session->lock);
	iscsi_update_cmdsn(session, (struct iscsi_nopin*)rhdr);

	if (!ctask->sc || session->state != ISCSI_STATE_LOGGED_IN) {
		printk(KERN_INFO "iscsi_tcp: dropping R2T itt %d in "
		       "recovery...\n", ctask->itt);
		spin_unlock(&session->lock);
		return 0;
	}
	rc = __kfifo_get(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) {
		printk(KERN_ERR "iscsi_tcp: invalid R2T with zero data len\n");
		spin_unlock(&session->lock);
		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_bufflen(ctask->sc)) {
		spin_unlock(&session->lock);
		printk(KERN_ERR "iscsi_tcp: invalid R2T with data len %u at "
		       "offset %u and total length %d\n", r2t->data_length,
		       r2t->data_offset, scsi_bufflen(ctask->sc));
		return ISCSI_ERR_DATALEN;
	}

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

	iscsi_solicit_data_init(conn, ctask, r2t);

	__kfifo_put(tcp_ctask->r2tqueue, (void*)&r2t, sizeof(void*));
	tcp_ctask->xmstate |= XMSTATE_SOL_HDR_INIT;

	iscsi_requeue_ctask(ctask);
	spin_unlock(&session->lock);

	return 0;
}

/*
 * Handle incoming reply to DataIn command
 */
static int
iscsi_tcp_process_data_in(struct iscsi_tcp_conn *tcp_conn,
			  struct iscsi_chunk *chunk)
{
	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, chunk))
		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.
 */
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;
	tcp_conn->in.datalen = ntoh24(hdr->dlength);
	if (tcp_conn->in.datalen > conn->max_recv_dlength) {
		printk(KERN_ERR "iscsi_tcp: datalen %d > %d\n",
		       tcp_conn->in.datalen, conn->max_recv_dlength);
	/* 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 == ISCSI_ERR_NO_SCSI_CMD) {
		/* XXX: what does this do? */
		tcp_conn->in.datalen = 0; /* force drop */
		return 0;
	} else 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];
		rc = iscsi_data_rsp(conn, ctask);
		if (tcp_conn->in.datalen) {
			struct iscsi_tcp_cmd_task *tcp_ctask = ctask->dd_data;
			struct hash_desc *rx_hash = NULL;

			/*
			 * Setup copy of Data-In into the Scsi_Cmnd
			 * Scatterlist case:
			 * We set up the iscsi_chunk 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_chunk_seek_sg(&tcp_conn->in.chunk,
						scsi_sglist(ctask->sc),
						scsi_sg_count(ctask->sc),
						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)
			rc = iscsi_r2t_rsp(conn, ctask);
		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) {
			printk(KERN_ERR "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);
		/* 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;
partial_sg_digest_update(struct hash_desc *desc, struct scatterlist *sg,
	sg_init_table(&temp, 1);
	sg_set_page(&temp, sg_page(sg), length, offset);
	crypto_hash_update(desc, &temp, length);
/**
 * 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_chunk *chunk)
	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 (chunk->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;

		chunk->total_size += ahslen;
		chunk->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 (chunk->digest_len == 0) {
			iscsi_tcp_chunk_splice_digest(chunk,
						      chunk->recv_digest);
			return 0;
		iscsi_tcp_dgst_header(&tcp_conn->rx_hash, hdr,
				      chunk->total_copied - ISCSI_DIGEST_SIZE,
				      chunk->digest);
		if (!iscsi_tcp_dgst_verify(tcp_conn, chunk))
			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_chunk *chunk = &tcp_conn->in.chunk;
	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)
			break;
		BUG_ON(chunk->copied >= chunk->size);

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

		if (chunk->total_copied >= chunk->total_size) {
			rc = chunk->done(tcp_conn, chunk);
			if (rc != 0) {
				skb_abort_seq_read(&seq);
				goto error;
			/* The done() functions sets up the
			 * next chunk. */
	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);

	 * 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_chunk_unmap(&tcp_conn->in.chunk);