iscsi_tcp.c 26.0 KB
Newer Older
1 2 3 4 5
/*
 * iSCSI Initiator over TCP/IP Data-Path
 *
 * Copyright (C) 2004 Dmitry Yusupov
 * Copyright (C) 2004 Alex Aizman
6 7
 * Copyright (C) 2005 - 2006 Mike Christie
 * Copyright (C) 2006 Red Hat, Inc.  All rights reserved.
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
 * 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/inet.h>
M
Mike Christie 已提交
31
#include <linux/file.h>
32 33 34 35 36 37 38
#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>
39
#include <scsi/scsi_device.h>
40 41 42 43 44 45
#include <scsi/scsi_host.h>
#include <scsi/scsi.h>
#include <scsi/scsi_transport_iscsi.h>

#include "iscsi_tcp.h"

46 47
MODULE_AUTHOR("Mike Christie <michaelc@cs.wisc.edu>, "
	      "Dmitry Yusupov <dmitry_yus@yahoo.com>, "
48 49 50 51
	      "Alex Aizman <itn780@yahoo.com>");
MODULE_DESCRIPTION("iSCSI/TCP data-path");
MODULE_LICENSE("GPL");

52 53 54
static struct scsi_transport_template *iscsi_sw_tcp_scsi_transport;
static struct scsi_host_template iscsi_sw_tcp_sht;
static struct iscsi_transport iscsi_sw_tcp_transport;
55

56 57 58
static unsigned int iscsi_max_lun = 512;
module_param_named(max_lun, iscsi_max_lun, uint, S_IRUGO);

59 60 61 62 63 64 65 66 67 68 69 70 71 72 73
static int iscsi_sw_tcp_dbg;
module_param_named(debug_iscsi_tcp, iscsi_sw_tcp_dbg, int,
		   S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(debug_iscsi_tcp, "Turn on debugging for iscsi_tcp module "
		 "Set to 1 to turn on, and zero to turn off. Default is off.");

#define ISCSI_SW_TCP_DBG(_conn, dbg_fmt, arg...)		\
	do {							\
		if (iscsi_sw_tcp_dbg)				\
			iscsi_conn_printk(KERN_INFO, _conn,	\
					     "%s " dbg_fmt,	\
					     __func__, ##arg);	\
	} while (0);


O
Olaf Kirch 已提交
74
/**
75
 * iscsi_sw_tcp_recv - TCP receive in sendfile fashion
76 77 78 79
 * @rd_desc: read descriptor
 * @skb: socket buffer
 * @offset: offset in skb
 * @len: skb->len - offset
80 81 82
 */
static int iscsi_sw_tcp_recv(read_descriptor_t *rd_desc, struct sk_buff *skb,
			     unsigned int offset, size_t len)
83 84 85 86 87
{
	struct iscsi_conn *conn = rd_desc->arg.data;
	unsigned int consumed, total_consumed = 0;
	int status;

88
	ISCSI_SW_TCP_DBG(conn, "in %d bytes\n", skb->len - offset);
89 90 91 92 93 94 95 96

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

97 98
	ISCSI_SW_TCP_DBG(conn, "read %d bytes status %d\n",
			 skb->len - offset, status);
99
	return total_consumed;
100 101
}

102
static void iscsi_sw_tcp_data_ready(struct sock *sk, int flag)
103 104
{
	struct iscsi_conn *conn = sk->sk_user_data;
O
Olaf Kirch 已提交
105
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
106 107 108 109
	read_descriptor_t rd_desc;

	read_lock(&sk->sk_callback_lock);

110
	/*
O
Olaf Kirch 已提交
111
	 * Use rd_desc to pass 'conn' to iscsi_tcp_recv.
112
	 * We set count to 1 because we want the network layer to
O
Olaf Kirch 已提交
113
	 * hand us all the skbs that are available. iscsi_tcp_recv
114 115
	 * handled pdus that cross buffers or pdus that still need data.
	 */
116
	rd_desc.arg.data = conn;
117
	rd_desc.count = 1;
118
	tcp_read_sock(sk, &rd_desc, iscsi_sw_tcp_recv);
119 120

	read_unlock(&sk->sk_callback_lock);
O
Olaf Kirch 已提交
121 122 123

	/* If we had to (atomically) map a highmem page,
	 * unmap it now. */
124
	iscsi_tcp_segment_unmap(&tcp_conn->in.segment);
125 126
}

127
static void iscsi_sw_tcp_state_change(struct sock *sk)
128
{
129
	struct iscsi_tcp_conn *tcp_conn;
130
	struct iscsi_sw_tcp_conn *tcp_sw_conn;
131 132 133 134 135 136 137 138 139
	struct iscsi_conn *conn;
	struct iscsi_session *session;
	void (*old_state_change)(struct sock *);

	read_lock(&sk->sk_callback_lock);

	conn = (struct iscsi_conn*)sk->sk_user_data;
	session = conn->session;

M
Mike Christie 已提交
140 141 142
	if ((sk->sk_state == TCP_CLOSE_WAIT ||
	     sk->sk_state == TCP_CLOSE) &&
	    !atomic_read(&sk->sk_rmem_alloc)) {
143 144
		ISCSI_SW_TCP_DBG(conn, "iscsi_tcp_state_change: "
				 "TCP_CLOSE|TCP_CLOSE_WAIT\n");
145 146 147
		iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
	}

148
	tcp_conn = conn->dd_data;
149 150
	tcp_sw_conn = tcp_conn->dd_data;
	old_state_change = tcp_sw_conn->old_state_change;
151 152 153 154 155 156 157 158 159 160

	read_unlock(&sk->sk_callback_lock);

	old_state_change(sk);
}

/**
 * iscsi_write_space - Called when more output buffer space is available
 * @sk: socket space is available for
 **/
161
static void iscsi_sw_tcp_write_space(struct sock *sk)
162 163
{
	struct iscsi_conn *conn = (struct iscsi_conn*)sk->sk_user_data;
164
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
165
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;
166

167
	tcp_sw_conn->old_write_space(sk);
168
	ISCSI_SW_TCP_DBG(conn, "iscsi_write_space\n");
169
	iscsi_conn_queue_work(conn);
170 171
}

172
static void iscsi_sw_tcp_conn_set_callbacks(struct iscsi_conn *conn)
173
{
174
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
175 176
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;
	struct sock *sk = tcp_sw_conn->sock->sk;
177 178 179 180

	/* assign new callbacks */
	write_lock_bh(&sk->sk_callback_lock);
	sk->sk_user_data = conn;
181 182 183 184 185 186
	tcp_sw_conn->old_data_ready = sk->sk_data_ready;
	tcp_sw_conn->old_state_change = sk->sk_state_change;
	tcp_sw_conn->old_write_space = sk->sk_write_space;
	sk->sk_data_ready = iscsi_sw_tcp_data_ready;
	sk->sk_state_change = iscsi_sw_tcp_state_change;
	sk->sk_write_space = iscsi_sw_tcp_write_space;
187 188 189
	write_unlock_bh(&sk->sk_callback_lock);
}

190 191
static void
iscsi_sw_tcp_conn_restore_callbacks(struct iscsi_sw_tcp_conn *tcp_sw_conn)
192
{
193
	struct sock *sk = tcp_sw_conn->sock->sk;
194 195 196 197

	/* restore socket callbacks, see also: iscsi_conn_set_callbacks() */
	write_lock_bh(&sk->sk_callback_lock);
	sk->sk_user_data    = NULL;
198 199 200
	sk->sk_data_ready   = tcp_sw_conn->old_data_ready;
	sk->sk_state_change = tcp_sw_conn->old_state_change;
	sk->sk_write_space  = tcp_sw_conn->old_write_space;
201 202 203 204 205
	sk->sk_no_check	 = 0;
	write_unlock_bh(&sk->sk_callback_lock);
}

/**
206
 * iscsi_sw_tcp_xmit_segment - transmit segment
207
 * @tcp_conn: the iSCSI TCP connection
208 209 210 211 212 213 214 215 216 217
 * @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.
 */
218
static int iscsi_sw_tcp_xmit_segment(struct iscsi_tcp_conn *tcp_conn,
219 220
				     struct iscsi_segment *segment)
{
221
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;
222 223 224 225
	struct socket *sk = tcp_sw_conn->sock;
	unsigned int copied = 0;
	int r = 0;

226
	while (!iscsi_tcp_segment_done(tcp_conn, segment, 0, r)) {
227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266
		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_sw_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_sw_tcp_xmit - TCP transmit
267
 **/
268
static int iscsi_sw_tcp_xmit(struct iscsi_conn *conn)
269
{
270
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
271 272
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;
	struct iscsi_segment *segment = &tcp_sw_conn->out.segment;
273 274
	unsigned int consumed = 0;
	int rc = 0;
275

276
	while (1) {
277
		rc = iscsi_sw_tcp_xmit_segment(tcp_conn, segment);
278 279
		if (rc < 0) {
			rc = ISCSI_ERR_XMIT_FAILED;
280
			goto error;
281
		}
282 283 284 285 286 287 288 289
		if (rc == 0)
			break;

		consumed += rc;

		if (segment->total_copied >= segment->total_size) {
			if (segment->done != NULL) {
				rc = segment->done(tcp_conn, segment);
290
				if (rc != 0)
291 292 293
					goto error;
			}
		}
294 295
	}

296
	ISCSI_SW_TCP_DBG(conn, "xmit %d bytes\n", consumed);
297 298 299 300 301 302 303

	conn->txdata_octets += consumed;
	return consumed;

error:
	/* Transmit error. We could initiate error recovery
	 * here. */
304
	ISCSI_SW_TCP_DBG(conn, "Error sending PDU, errno=%d\n", rc);
305 306
	iscsi_conn_failure(conn, rc);
	return -EIO;
307 308 309
}

/**
310 311
 * iscsi_tcp_xmit_qlen - return the number of bytes queued for xmit
 */
312
static inline int iscsi_sw_tcp_xmit_qlen(struct iscsi_conn *conn)
313
{
314
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
315 316
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;
	struct iscsi_segment *segment = &tcp_sw_conn->out.segment;
317

318
	return segment->total_copied - segment->total_size;
319 320
}

321
static int iscsi_sw_tcp_pdu_xmit(struct iscsi_task *task)
322
{
323
	struct iscsi_conn *conn = task->conn;
324 325
	int rc;

326 327
	while (iscsi_sw_tcp_xmit_qlen(conn)) {
		rc = iscsi_sw_tcp_xmit(conn);
328
		if (rc == 0)
329
			return -EAGAIN;
330 331
		if (rc < 0)
			return rc;
332
	}
333

334
	return 0;
335 336
}

337 338 339 340
/*
 * This is called when we're done sending the header.
 * Simply copy the data_segment to the send segment, and return.
 */
341 342
static int iscsi_sw_tcp_send_hdr_done(struct iscsi_tcp_conn *tcp_conn,
				      struct iscsi_segment *segment)
343
{
344 345 346
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;

	tcp_sw_conn->out.segment = tcp_sw_conn->out.data_segment;
347 348 349 350
	ISCSI_SW_TCP_DBG(tcp_conn->iscsi_conn,
			 "Header done. Next segment size %u total_size %u\n",
			 tcp_sw_conn->out.segment.size,
			 tcp_sw_conn->out.segment.total_size);
351 352 353
	return 0;
}

354 355
static void iscsi_sw_tcp_send_hdr_prep(struct iscsi_conn *conn, void *hdr,
				       size_t hdrlen)
356 357
{
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
358
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;
359

360 361
	ISCSI_SW_TCP_DBG(conn, "%s\n", conn->hdrdgst_en ?
			 "digest enabled" : "digest disabled");
362 363 364

	/* Clear the data segment - needs to be filled in by the
	 * caller using iscsi_tcp_send_data_prep() */
365 366
	memset(&tcp_sw_conn->out.data_segment, 0,
	       sizeof(struct iscsi_segment));
367 368 369

	/* If header digest is enabled, compute the CRC and
	 * place the digest into the same buffer. We make
370
	 * sure that both iscsi_tcp_task and mtask have
371 372 373
	 * sufficient room.
	 */
	if (conn->hdrdgst_en) {
374
		iscsi_tcp_dgst_header(&tcp_sw_conn->tx_hash, hdr, hdrlen,
375 376 377 378 379 380 381
				      hdr + hdrlen);
		hdrlen += ISCSI_DIGEST_SIZE;
	}

	/* Remember header pointer for later, when we need
	 * to decide whether there's a payload to go along
	 * with the header. */
382
	tcp_sw_conn->out.hdr = hdr;
383

384 385
	iscsi_segment_init_linear(&tcp_sw_conn->out.segment, hdr, hdrlen,
				  iscsi_sw_tcp_send_hdr_done, NULL);
386 387 388 389 390 391 392 393
}

/*
 * Prepare the send buffer for the payload data.
 * Padding and checksumming will all be taken care
 * of by the iscsi_segment routines.
 */
static int
394 395 396
iscsi_sw_tcp_send_data_prep(struct iscsi_conn *conn, struct scatterlist *sg,
			    unsigned int count, unsigned int offset,
			    unsigned int len)
397 398
{
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
399
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;
400 401 402
	struct hash_desc *tx_hash = NULL;
	unsigned int hdr_spec_len;

403 404 405
	ISCSI_SW_TCP_DBG(conn, "offset=%d, datalen=%d %s\n", offset, len,
			 conn->datadgst_en ?
			 "digest enabled" : "digest disabled");
406 407 408

	/* Make sure the datalen matches what the caller
	   said he would send. */
409
	hdr_spec_len = ntoh24(tcp_sw_conn->out.hdr->dlength);
410 411 412
	WARN_ON(iscsi_padded(len) != iscsi_padded(hdr_spec_len));

	if (conn->datadgst_en)
413
		tx_hash = &tcp_sw_conn->tx_hash;
414

415 416 417
	return iscsi_segment_seek_sg(&tcp_sw_conn->out.data_segment,
				     sg, count, offset, len,
				     NULL, tx_hash);
418 419 420
}

static void
421
iscsi_sw_tcp_send_linear_data_prep(struct iscsi_conn *conn, void *data,
422 423 424
				   size_t len)
{
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
425
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;
426 427 428
	struct hash_desc *tx_hash = NULL;
	unsigned int hdr_spec_len;

429 430
	ISCSI_SW_TCP_DBG(conn, "datalen=%zd %s\n", len, conn->datadgst_en ?
			 "digest enabled" : "digest disabled");
431 432 433

	/* Make sure the datalen matches what the caller
	   said he would send. */
434
	hdr_spec_len = ntoh24(tcp_sw_conn->out.hdr->dlength);
435 436 437
	WARN_ON(iscsi_padded(len) != iscsi_padded(hdr_spec_len));

	if (conn->datadgst_en)
438
		tx_hash = &tcp_sw_conn->tx_hash;
439

440
	iscsi_segment_init_linear(&tcp_sw_conn->out.data_segment,
441
				data, len, NULL, tx_hash);
442 443
}

444 445
static int iscsi_sw_tcp_pdu_init(struct iscsi_task *task,
				 unsigned int offset, unsigned int count)
446 447 448 449
{
	struct iscsi_conn *conn = task->conn;
	int err = 0;

450
	iscsi_sw_tcp_send_hdr_prep(conn, task->hdr, task->hdr_len);
451 452 453 454 455

	if (!count)
		return 0;

	if (!task->sc)
456
		iscsi_sw_tcp_send_linear_data_prep(conn, task->data, count);
457 458 459
	else {
		struct scsi_data_buffer *sdb = scsi_out(task->sc);

460 461 462
		err = iscsi_sw_tcp_send_data_prep(conn, sdb->table.sgl,
						  sdb->table.nents, offset,
						  count);
463 464 465
	}

	if (err) {
466
		/* got invalid offset/len */
467 468 469 470 471
		return -EIO;
	}
	return 0;
}

472
static int iscsi_sw_tcp_pdu_alloc(struct iscsi_task *task, uint8_t opcode)
473
{
474
	struct iscsi_tcp_task *tcp_task = task->dd_data;
475

476 477
	task->hdr = task->dd_data + sizeof(*tcp_task);
	task->hdr_max = sizeof(struct iscsi_sw_tcp_hdrbuf) - ISCSI_DIGEST_SIZE;
478
	return 0;
479 480
}

481
static struct iscsi_cls_conn *
482 483
iscsi_sw_tcp_conn_create(struct iscsi_cls_session *cls_session,
			 uint32_t conn_idx)
484
{
485 486 487
	struct iscsi_conn *conn;
	struct iscsi_cls_conn *cls_conn;
	struct iscsi_tcp_conn *tcp_conn;
488
	struct iscsi_sw_tcp_conn *tcp_sw_conn;
489

490 491
	cls_conn = iscsi_tcp_conn_setup(cls_session, sizeof(*tcp_sw_conn),
					conn_idx);
492 493 494
	if (!cls_conn)
		return NULL;
	conn = cls_conn->dd_data;
495
	tcp_conn = conn->dd_data;
496
	tcp_sw_conn = tcp_conn->dd_data;
497

498 499 500 501
	tcp_sw_conn->tx_hash.tfm = crypto_alloc_hash("crc32c", 0,
						     CRYPTO_ALG_ASYNC);
	tcp_sw_conn->tx_hash.flags = 0;
	if (IS_ERR(tcp_sw_conn->tx_hash.tfm))
502
		goto free_conn;
503

504 505 506 507
	tcp_sw_conn->rx_hash.tfm = crypto_alloc_hash("crc32c", 0,
						     CRYPTO_ALG_ASYNC);
	tcp_sw_conn->rx_hash.flags = 0;
	if (IS_ERR(tcp_sw_conn->rx_hash.tfm))
508
		goto free_tx_tfm;
509
	tcp_conn->rx_hash = &tcp_sw_conn->rx_hash;
510

511
	return cls_conn;
512

513
free_tx_tfm:
514
	crypto_free_hash(tcp_sw_conn->tx_hash.tfm);
515
free_conn:
516 517 518 519 520
	iscsi_conn_printk(KERN_ERR, conn,
			  "Could not create connection due to crc32c "
			  "loading error. Make sure the crc32c "
			  "module is built as a module or into the "
			  "kernel\n");
521
	iscsi_tcp_conn_teardown(cls_conn);
522
	return NULL;
523 524
}

525
static void iscsi_sw_tcp_release_conn(struct iscsi_conn *conn)
526
{
527
	struct iscsi_session *session = conn->session;
528
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
529 530
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;
	struct socket *sock = tcp_sw_conn->sock;
531

532
	if (!sock)
533 534
		return;

535
	sock_hold(sock->sk);
536
	iscsi_sw_tcp_conn_restore_callbacks(tcp_sw_conn);
537
	sock_put(sock->sk);
538

539
	spin_lock_bh(&session->lock);
540
	tcp_sw_conn->sock = NULL;
541 542
	spin_unlock_bh(&session->lock);
	sockfd_put(sock);
543 544
}

545
static void iscsi_sw_tcp_conn_destroy(struct iscsi_cls_conn *cls_conn)
546
{
547 548
	struct iscsi_conn *conn = cls_conn->dd_data;
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
549
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;
550

551
	iscsi_sw_tcp_release_conn(conn);
552

553 554 555 556
	if (tcp_sw_conn->tx_hash.tfm)
		crypto_free_hash(tcp_sw_conn->tx_hash.tfm);
	if (tcp_sw_conn->rx_hash.tfm)
		crypto_free_hash(tcp_sw_conn->rx_hash.tfm);
557

558
	iscsi_tcp_conn_teardown(cls_conn);
559
}
560

561
static void iscsi_sw_tcp_conn_stop(struct iscsi_cls_conn *cls_conn, int flag)
562 563
{
	struct iscsi_conn *conn = cls_conn->dd_data;
564
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
565
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;
566 567

	/* userspace may have goofed up and not bound us */
568
	if (!tcp_sw_conn->sock)
569 570 571 572 573 574
		return;
	/*
	 * Make sure our recv side is stopped.
	 * Older tools called conn stop before ep_disconnect
	 * so IO could still be coming in.
	 */
575
	write_lock_bh(&tcp_sw_conn->sock->sk->sk_callback_lock);
576
	set_bit(ISCSI_SUSPEND_BIT, &conn->suspend_rx);
577
	write_unlock_bh(&tcp_sw_conn->sock->sk->sk_callback_lock);
578 579

	iscsi_conn_stop(cls_conn, flag);
580
	iscsi_sw_tcp_release_conn(conn);
581 582
}

583 584 585 586 587
static int iscsi_sw_tcp_get_addr(struct iscsi_conn *conn, struct socket *sock,
				 char *buf, int *port,
				 int (*getname)(struct socket *,
						struct sockaddr *,
						int *addrlen))
588 589 590 591 592 593
{
	struct sockaddr_storage *addr;
	struct sockaddr_in6 *sin6;
	struct sockaddr_in *sin;
	int rc = 0, len;

594
	addr = kmalloc(sizeof(*addr), GFP_KERNEL);
595 596 597 598 599 600 601 602 603 604 605 606
	if (!addr)
		return -ENOMEM;

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

	switch (addr->ss_family) {
	case AF_INET:
		sin = (struct sockaddr_in *)addr;
		spin_lock_bh(&conn->session->lock);
H
Harvey Harrison 已提交
607
		sprintf(buf, "%pI4", &sin->sin_addr.s_addr);
608 609 610 611 612 613
		*port = be16_to_cpu(sin->sin_port);
		spin_unlock_bh(&conn->session->lock);
		break;
	case AF_INET6:
		sin6 = (struct sockaddr_in6 *)addr;
		spin_lock_bh(&conn->session->lock);
H
Harvey Harrison 已提交
614
		sprintf(buf, "%pI6", &sin6->sin6_addr);
615 616 617 618 619 620 621 622 623
		*port = be16_to_cpu(sin6->sin6_port);
		spin_unlock_bh(&conn->session->lock);
		break;
	}
free_addr:
	kfree(addr);
	return rc;
}

624
static int
625 626 627
iscsi_sw_tcp_conn_bind(struct iscsi_cls_session *cls_session,
		       struct iscsi_cls_conn *cls_conn, uint64_t transport_eph,
		       int is_leading)
628
{
629 630
	struct Scsi_Host *shost = iscsi_session_to_shost(cls_session);
	struct iscsi_host *ihost = shost_priv(shost);
631 632
	struct iscsi_conn *conn = cls_conn->dd_data;
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
633
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;
634 635 636
	struct sock *sk;
	struct socket *sock;
	int err;
637

638
	/* lookup for existing socket */
639
	sock = sockfd_lookup((int)transport_eph, &err);
640
	if (!sock) {
641 642
		iscsi_conn_printk(KERN_ERR, conn,
				  "sockfd_lookup failed %d\n", err);
643
		return -EEXIST;
644
	}
645 646 647 648 649
	/*
	 * copy these values now because if we drop the session
	 * userspace may still want to query the values since we will
	 * be using them for the reconnect
	 */
650 651
	err = iscsi_sw_tcp_get_addr(conn, sock, conn->portal_address,
				    &conn->portal_port, kernel_getpeername);
652 653 654
	if (err)
		goto free_socket;

655 656
	err = iscsi_sw_tcp_get_addr(conn, sock, ihost->local_address,
				    &ihost->local_port, kernel_getsockname);
657 658
	if (err)
		goto free_socket;
659

660 661
	err = iscsi_conn_bind(cls_session, cls_conn, is_leading);
	if (err)
662
		goto free_socket;
663

664
	/* bind iSCSI connection and socket */
665
	tcp_sw_conn->sock = sock;
666

667 668 669 670 671
	/* setup Socket parameters */
	sk = sock->sk;
	sk->sk_reuse = 1;
	sk->sk_sndtimeo = 15 * HZ; /* FIXME: make it configurable */
	sk->sk_allocation = GFP_ATOMIC;
672

673 674
	iscsi_sw_tcp_conn_set_callbacks(conn);
	tcp_sw_conn->sendpage = tcp_sw_conn->sock->ops->sendpage;
675 676 677
	/*
	 * set receive state machine into initial state
	 */
O
Olaf Kirch 已提交
678
	iscsi_tcp_hdr_recv_prep(tcp_conn);
679
	return 0;
680 681 682 683

free_socket:
	sockfd_put(sock);
	return err;
684 685
}

686 687 688
static int iscsi_sw_tcp_conn_set_param(struct iscsi_cls_conn *cls_conn,
				       enum iscsi_param param, char *buf,
				       int buflen)
689
{
690
	struct iscsi_conn *conn = cls_conn->dd_data;
691
	struct iscsi_session *session = conn->session;
692
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
693
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;
694
	int value;
695 696 697

	switch(param) {
	case ISCSI_PARAM_HDRDGST_EN:
698
		iscsi_set_param(cls_conn, param, buf, buflen);
699 700
		break;
	case ISCSI_PARAM_DATADGST_EN:
701
		iscsi_set_param(cls_conn, param, buf, buflen);
702 703
		tcp_sw_conn->sendpage = conn->datadgst_en ?
			sock_no_sendpage : tcp_sw_conn->sock->ops->sendpage;
704 705
		break;
	case ISCSI_PARAM_MAX_R2T:
706
		sscanf(buf, "%d", &value);
707 708 709
		if (value <= 0 || !is_power_of_2(value))
			return -EINVAL;
		if (session->max_r2t == value)
710
			break;
711
		iscsi_tcp_r2tpool_free(session);
712
		iscsi_set_param(cls_conn, param, buf, buflen);
713
		if (iscsi_tcp_r2tpool_alloc(session))
714 715 716
			return -ENOMEM;
		break;
	default:
717
		return iscsi_set_param(cls_conn, param, buf, buflen);
718 719 720 721 722
	}

	return 0;
}

723 724
static int iscsi_sw_tcp_conn_get_param(struct iscsi_cls_conn *cls_conn,
				       enum iscsi_param param, char *buf)
725
{
726
	struct iscsi_conn *conn = cls_conn->dd_data;
727
	int len;
728 729

	switch(param) {
730
	case ISCSI_PARAM_CONN_PORT:
731 732 733
		spin_lock_bh(&conn->session->lock);
		len = sprintf(buf, "%hu\n", conn->portal_port);
		spin_unlock_bh(&conn->session->lock);
734
		break;
735
	case ISCSI_PARAM_CONN_ADDRESS:
736 737 738
		spin_lock_bh(&conn->session->lock);
		len = sprintf(buf, "%s\n", conn->portal_address);
		spin_unlock_bh(&conn->session->lock);
739 740
		break;
	default:
741
		return iscsi_conn_get_param(cls_conn, param, buf);
742 743 744 745 746
	}

	return len;
}

747
static void
748 749
iscsi_sw_tcp_conn_get_stats(struct iscsi_cls_conn *cls_conn,
			    struct iscsi_stats *stats)
750
{
751
	struct iscsi_conn *conn = cls_conn->dd_data;
752
	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
753
	struct iscsi_sw_tcp_conn *tcp_sw_conn = tcp_conn->dd_data;
754 755 756

	stats->custom_length = 3;
	strcpy(stats->custom[0].desc, "tx_sendpage_failures");
757
	stats->custom[0].value = tcp_sw_conn->sendpage_failures_cnt;
758
	strcpy(stats->custom[1].desc, "rx_discontiguous_hdr");
759
	stats->custom[1].value = tcp_sw_conn->discontiguous_hdr_cnt;
760 761
	strcpy(stats->custom[2].desc, "eh_abort_cnt");
	stats->custom[2].value = conn->eh_abort_cnt;
762 763

	iscsi_tcp_conn_get_stats(cls_conn, stats);
764 765
}

766
static struct iscsi_cls_session *
767
iscsi_sw_tcp_session_create(struct iscsi_endpoint *ep, uint16_t cmds_max,
768
			    uint16_t qdepth, uint32_t initial_cmdsn)
769
{
770 771
	struct iscsi_cls_session *cls_session;
	struct iscsi_session *session;
772
	struct Scsi_Host *shost;
773

774 775
	if (ep) {
		printk(KERN_ERR "iscsi_tcp: invalid ep %p.\n", ep);
776 777 778
		return NULL;
	}

779
	shost = iscsi_host_alloc(&iscsi_sw_tcp_sht, 0, 1);
780
	if (!shost)
781
		return NULL;
782
	shost->transportt = iscsi_sw_tcp_scsi_transport;
783
	shost->cmd_per_lun = qdepth;
784 785 786
	shost->max_lun = iscsi_max_lun;
	shost->max_id = 0;
	shost->max_channel = 0;
787
	shost->max_cmd_len = SCSI_MAX_VARLEN_CDB_SIZE;
788

789
	if (iscsi_host_add(shost, NULL))
790 791
		goto free_host;

792 793 794 795
	cls_session = iscsi_session_setup(&iscsi_sw_tcp_transport, shost,
					  cmds_max,
					  sizeof(struct iscsi_tcp_task) +
					  sizeof(struct iscsi_sw_tcp_hdrbuf),
796
					  initial_cmdsn, 0);
797 798 799
	if (!cls_session)
		goto remove_host;
	session = cls_session->dd_data;
800

801
	shost->can_queue = session->scsi_cmds_max;
802
	if (iscsi_tcp_r2tpool_alloc(session))
803
		goto remove_session;
804 805
	return cls_session;

806
remove_session:
807
	iscsi_session_teardown(cls_session);
808
remove_host:
809
	iscsi_host_remove(shost);
810
free_host:
811
	iscsi_host_free(shost);
812 813 814
	return NULL;
}

815
static void iscsi_sw_tcp_session_destroy(struct iscsi_cls_session *cls_session)
816
{
817 818
	struct Scsi_Host *shost = iscsi_session_to_shost(cls_session);

819
	iscsi_tcp_r2tpool_free(cls_session->dd_data);
820
	iscsi_session_teardown(cls_session);
821

822 823
	iscsi_host_remove(shost);
	iscsi_host_free(shost);
824 825
}

P
Pete Wyckoff 已提交
826 827 828 829 830 831
static int iscsi_sw_tcp_slave_alloc(struct scsi_device *sdev)
{
	set_bit(QUEUE_FLAG_BIDI, &sdev->request_queue->queue_flags);
	return 0;
}

832
static int iscsi_sw_tcp_slave_configure(struct scsi_device *sdev)
833
{
834
	blk_queue_bounce_limit(sdev->request_queue, BLK_BOUNCE_ANY);
835 836 837 838
	blk_queue_dma_alignment(sdev->request_queue, 0);
	return 0;
}

839
static struct scsi_host_template iscsi_sw_tcp_sht = {
840
	.module			= THIS_MODULE,
841
	.name			= "iSCSI Initiator over TCP/IP",
842 843
	.queuecommand           = iscsi_queuecommand,
	.change_queue_depth	= iscsi_change_queue_depth,
844
	.can_queue		= ISCSI_DEF_XMIT_CMDS_MAX - 1,
845
	.sg_tablesize		= 4096,
846
	.max_sectors		= 0xFFFF,
847 848
	.cmd_per_lun		= ISCSI_DEF_CMD_PER_LUN,
	.eh_abort_handler       = iscsi_eh_abort,
849
	.eh_device_reset_handler= iscsi_eh_device_reset,
850
	.eh_target_reset_handler= iscsi_eh_target_reset,
851
	.use_clustering         = DISABLE_CLUSTERING,
P
Pete Wyckoff 已提交
852
	.slave_alloc            = iscsi_sw_tcp_slave_alloc,
853
	.slave_configure        = iscsi_sw_tcp_slave_configure,
854
	.target_alloc		= iscsi_target_alloc,
855 856 857 858
	.proc_name		= "iscsi_tcp",
	.this_id		= -1,
};

859
static struct iscsi_transport iscsi_sw_tcp_transport = {
860 861 862 863
	.owner			= THIS_MODULE,
	.name			= "tcp",
	.caps			= CAP_RECOVERY_L0 | CAP_MULTI_R2T | CAP_HDRDGST
				  | CAP_DATADGST,
864 865 866 867 868 869 870 871 872 873 874 875 876
	.param_mask		= ISCSI_MAX_RECV_DLENGTH |
				  ISCSI_MAX_XMIT_DLENGTH |
				  ISCSI_HDRDGST_EN |
				  ISCSI_DATADGST_EN |
				  ISCSI_INITIAL_R2T_EN |
				  ISCSI_MAX_R2T |
				  ISCSI_IMM_DATA_EN |
				  ISCSI_FIRST_BURST |
				  ISCSI_MAX_BURST |
				  ISCSI_PDU_INORDER_EN |
				  ISCSI_DATASEQ_INORDER_EN |
				  ISCSI_ERL |
				  ISCSI_CONN_PORT |
877
				  ISCSI_CONN_ADDRESS |
878 879 880
				  ISCSI_EXP_STATSN |
				  ISCSI_PERSISTENT_PORT |
				  ISCSI_PERSISTENT_ADDRESS |
881 882
				  ISCSI_TARGET_NAME | ISCSI_TPGT |
				  ISCSI_USERNAME | ISCSI_PASSWORD |
883
				  ISCSI_USERNAME_IN | ISCSI_PASSWORD_IN |
884 885
				  ISCSI_FAST_ABORT | ISCSI_ABORT_TMO |
				  ISCSI_LU_RESET_TMO |
886 887
				  ISCSI_PING_TMO | ISCSI_RECV_TMO |
				  ISCSI_IFACE_NAME | ISCSI_INITIATOR_NAME,
888
	.host_param_mask	= ISCSI_HOST_HWADDRESS | ISCSI_HOST_IPADDRESS |
889 890
				  ISCSI_HOST_INITIATOR_NAME |
				  ISCSI_HOST_NETDEV_NAME,
891
	/* session management */
892 893
	.create_session		= iscsi_sw_tcp_session_create,
	.destroy_session	= iscsi_sw_tcp_session_destroy,
894
	/* connection management */
895 896 897 898 899
	.create_conn		= iscsi_sw_tcp_conn_create,
	.bind_conn		= iscsi_sw_tcp_conn_bind,
	.destroy_conn		= iscsi_sw_tcp_conn_destroy,
	.set_param		= iscsi_sw_tcp_conn_set_param,
	.get_conn_param		= iscsi_sw_tcp_conn_get_param,
900
	.get_session_param	= iscsi_session_get_param,
901
	.start_conn		= iscsi_conn_start,
902
	.stop_conn		= iscsi_sw_tcp_conn_stop,
903
	/* iscsi host params */
904
	.get_host_param		= iscsi_host_get_param,
905
	.set_host_param		= iscsi_host_set_param,
906
	/* IO */
907
	.send_pdu		= iscsi_conn_send_pdu,
908
	.get_stats		= iscsi_sw_tcp_conn_get_stats,
909
	/* iscsi task/cmd helpers */
910 911 912
	.init_task		= iscsi_tcp_task_init,
	.xmit_task		= iscsi_tcp_task_xmit,
	.cleanup_task		= iscsi_tcp_cleanup_task,
913
	/* low level pdu helpers */
914 915 916
	.xmit_pdu		= iscsi_sw_tcp_pdu_xmit,
	.init_pdu		= iscsi_sw_tcp_pdu_init,
	.alloc_pdu		= iscsi_sw_tcp_pdu_alloc,
917
	/* recovery */
M
Mike Christie 已提交
918
	.session_recovery_timedout = iscsi_session_recovery_timedout,
919 920
};

921
static int __init iscsi_sw_tcp_init(void)
922 923
{
	if (iscsi_max_lun < 1) {
O
Or Gerlitz 已提交
924 925
		printk(KERN_ERR "iscsi_tcp: Invalid max_lun value of %u\n",
		       iscsi_max_lun);
926 927 928
		return -EINVAL;
	}

929 930 931
	iscsi_sw_tcp_scsi_transport = iscsi_register_transport(
						&iscsi_sw_tcp_transport);
	if (!iscsi_sw_tcp_scsi_transport)
932
		return -ENODEV;
933

934
	return 0;
935 936
}

937
static void __exit iscsi_sw_tcp_exit(void)
938
{
939
	iscsi_unregister_transport(&iscsi_sw_tcp_transport);
940 941
}

942 943
module_init(iscsi_sw_tcp_init);
module_exit(iscsi_sw_tcp_exit);