scsi.c 58.2 KB
Newer Older
1 2 3
/*******************************************************************************
 * Vhost kernel TCM fabric driver for virtio SCSI initiators
 *
4
 * (C) Copyright 2010-2013 Datera, Inc.
5 6 7 8
 * (C) Copyright 2010-2012 IBM Corp.
 *
 * Licensed to the Linux Foundation under the General Public License (GPL) version 2.
 *
9
 * Authors: Nicholas A. Bellinger <nab@daterainc.com>
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37
 *          Stefan Hajnoczi <stefanha@linux.vnet.ibm.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.
 *
 ****************************************************************************/

#include <linux/module.h>
#include <linux/moduleparam.h>
#include <generated/utsrelease.h>
#include <linux/utsname.h>
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/kthread.h>
#include <linux/types.h>
#include <linux/string.h>
#include <linux/configfs.h>
#include <linux/ctype.h>
#include <linux/compat.h>
#include <linux/eventfd.h>
#include <linux/fs.h>
38
#include <linux/vmalloc.h>
39 40
#include <linux/miscdevice.h>
#include <asm/unaligned.h>
41 42
#include <scsi/scsi_common.h>
#include <scsi/scsi_proto.h>
43 44 45 46
#include <target/target_core_base.h>
#include <target/target_core_fabric.h>
#include <linux/vhost.h>
#include <linux/virtio_scsi.h>
47
#include <linux/llist.h>
A
Asias He 已提交
48
#include <linux/bitmap.h>
49
#include <linux/percpu_ida.h>
50 51

#include "vhost.h"
M
Michael S. Tsirkin 已提交
52

53 54 55 56 57 58 59
#define VHOST_SCSI_VERSION  "v0.1"
#define VHOST_SCSI_NAMELEN 256
#define VHOST_SCSI_MAX_CDB_SIZE 32
#define VHOST_SCSI_DEFAULT_TAGS 256
#define VHOST_SCSI_PREALLOC_SGLS 2048
#define VHOST_SCSI_PREALLOC_UPAGES 2048
#define VHOST_SCSI_PREALLOC_PROT_SGLS 512
M
Michael S. Tsirkin 已提交
60 61 62 63 64 65 66 67

struct vhost_scsi_inflight {
	/* Wait for the flush operation to finish */
	struct completion comp;
	/* Refcount for the inflight reqs */
	struct kref kref;
};

68
struct vhost_scsi_cmd {
M
Michael S. Tsirkin 已提交
69 70 71 72
	/* Descriptor from vhost_get_vq_desc() for virt_queue segment */
	int tvc_vq_desc;
	/* virtio-scsi initiator task attribute */
	int tvc_task_attr;
73 74
	/* virtio-scsi response incoming iovecs */
	int tvc_in_iovs;
M
Michael S. Tsirkin 已提交
75 76 77 78 79 80 81 82
	/* virtio-scsi initiator data direction */
	enum dma_data_direction tvc_data_direction;
	/* Expected data transfer length from virtio-scsi header */
	u32 tvc_exp_data_len;
	/* The Tag from include/linux/virtio_scsi.h:struct virtio_scsi_cmd_req */
	u64 tvc_tag;
	/* The number of scatterlists associated with this cmd */
	u32 tvc_sgl_count;
83
	u32 tvc_prot_sgl_count;
84
	/* Saved unpacked SCSI LUN for vhost_scsi_submission_work() */
M
Michael S. Tsirkin 已提交
85 86 87
	u32 tvc_lun;
	/* Pointer to the SGL formatted memory from virtio-scsi */
	struct scatterlist *tvc_sgl;
88
	struct scatterlist *tvc_prot_sgl;
89
	struct page **tvc_upages;
90 91
	/* Pointer to response header iovec */
	struct iovec *tvc_resp_iov;
M
Michael S. Tsirkin 已提交
92 93 94 95 96
	/* Pointer to vhost_scsi for our device */
	struct vhost_scsi *tvc_vhost;
	/* Pointer to vhost_virtqueue for the cmd */
	struct vhost_virtqueue *tvc_vq;
	/* Pointer to vhost nexus memory */
97
	struct vhost_scsi_nexus *tvc_nexus;
M
Michael S. Tsirkin 已提交
98 99
	/* The TCM I/O descriptor that is accessed via container_of() */
	struct se_cmd tvc_se_cmd;
100
	/* work item used for cmwq dispatch to vhost_scsi_submission_work() */
M
Michael S. Tsirkin 已提交
101 102
	struct work_struct work;
	/* Copy of the incoming SCSI command descriptor block (CDB) */
103
	unsigned char tvc_cdb[VHOST_SCSI_MAX_CDB_SIZE];
M
Michael S. Tsirkin 已提交
104 105 106 107 108 109 110 111
	/* Sense buffer that will be mapped into outgoing status */
	unsigned char tvc_sense_buf[TRANSPORT_SENSE_BUFFER];
	/* Completed commands list, serviced from vhost worker thread */
	struct llist_node tvc_completion_list;
	/* Used to track inflight cmd */
	struct vhost_scsi_inflight *inflight;
};

112
struct vhost_scsi_nexus {
M
Michael S. Tsirkin 已提交
113 114 115 116
	/* Pointer to TCM session for I_T Nexus */
	struct se_session *tvn_se_sess;
};

117
struct vhost_scsi_tpg {
M
Michael S. Tsirkin 已提交
118 119 120 121 122 123
	/* Vhost port target portal group tag for TCM */
	u16 tport_tpgt;
	/* Used to track number of TPG Port/Lun Links wrt to explict I_T Nexus shutdown */
	int tv_tpg_port_count;
	/* Used for vhost_scsi device reference to tpg_nexus, protected by tv_tpg_mutex */
	int tv_tpg_vhost_count;
124 125
	/* Used for enabling T10-PI with legacy devices */
	int tv_fabric_prot_type;
126
	/* list for vhost_scsi_list */
M
Michael S. Tsirkin 已提交
127 128 129 130
	struct list_head tv_tpg_list;
	/* Used to protect access for tpg_nexus */
	struct mutex tv_tpg_mutex;
	/* Pointer to the TCM VHost I_T Nexus for this TPG endpoint */
131 132 133 134
	struct vhost_scsi_nexus *tpg_nexus;
	/* Pointer back to vhost_scsi_tport */
	struct vhost_scsi_tport *tport;
	/* Returned by vhost_scsi_make_tpg() */
M
Michael S. Tsirkin 已提交
135 136 137 138 139
	struct se_portal_group se_tpg;
	/* Pointer back to vhost_scsi, protected by tv_tpg_mutex */
	struct vhost_scsi *vhost_scsi;
};

140
struct vhost_scsi_tport {
M
Michael S. Tsirkin 已提交
141 142 143 144 145
	/* SCSI protocol the tport is providing */
	u8 tport_proto_id;
	/* Binary World Wide unique Port Name for Vhost Target port */
	u64 tport_wwpn;
	/* ASCII formatted WWPN for Vhost Target port */
146 147
	char tport_name[VHOST_SCSI_NAMELEN];
	/* Returned by vhost_scsi_make_tport() */
M
Michael S. Tsirkin 已提交
148 149 150
	struct se_wwn tport_wwn;
};

151
struct vhost_scsi_evt {
M
Michael S. Tsirkin 已提交
152 153 154 155 156
	/* event to be sent to guest */
	struct virtio_scsi_event event;
	/* event list, serviced from vhost worker thread */
	struct llist_node list;
};
157

158 159 160 161 162 163
enum {
	VHOST_SCSI_VQ_CTL = 0,
	VHOST_SCSI_VQ_EVT = 1,
	VHOST_SCSI_VQ_IO = 2,
};

164
/* Note: can't set VIRTIO_F_VERSION_1 yet, since that implies ANY_LAYOUT. */
165
enum {
166
	VHOST_SCSI_FEATURES = VHOST_FEATURES | (1ULL << VIRTIO_SCSI_F_HOTPLUG) |
167
					       (1ULL << VIRTIO_SCSI_F_T10_PI)
168 169
};

A
Asias He 已提交
170 171
#define VHOST_SCSI_MAX_TARGET	256
#define VHOST_SCSI_MAX_VQ	128
172
#define VHOST_SCSI_MAX_EVENT	128
A
Asias He 已提交
173

174 175
struct vhost_scsi_virtqueue {
	struct vhost_virtqueue vq;
176 177 178 179 180
	/*
	 * Reference counting for inflight reqs, used for flush operation. At
	 * each time, one reference tracks new commands submitted, while we
	 * wait for another one to reach 0.
	 */
181
	struct vhost_scsi_inflight inflights[2];
182 183 184 185
	/*
	 * Indicate current inflight in use, protected by vq->mutex.
	 * Writers must also take dev mutex and flush under it.
	 */
186
	int inflight_idx;
187 188
};

189
struct vhost_scsi {
A
Asias He 已提交
190
	/* Protected by vhost_scsi->dev.mutex */
191
	struct vhost_scsi_tpg **vs_tpg;
A
Asias He 已提交
192 193
	char vs_vhost_wwpn[TRANSPORT_IQN_LEN];

194
	struct vhost_dev dev;
195
	struct vhost_scsi_virtqueue vqs[VHOST_SCSI_MAX_VQ];
196 197

	struct vhost_work vs_completion_work; /* cmd completion work item */
198
	struct llist_head vs_completion_list; /* cmd completion queue */
199 200 201 202 203 204

	struct vhost_work vs_event_work; /* evt injection work item */
	struct llist_head vs_event_list; /* evt injection queue */

	bool vs_events_missed; /* any missed events, protected by vq->mutex */
	int vs_events_nr; /* num of pending events, protected by vq->mutex */
205 206
};

207
static struct workqueue_struct *vhost_scsi_workqueue;
208

209 210 211
/* Global spinlock to protect vhost_scsi TPG list for vhost IOCTL access */
static DEFINE_MUTEX(vhost_scsi_mutex);
static LIST_HEAD(vhost_scsi_list);
212

213
static int iov_num_pages(void __user *iov_base, size_t iov_len)
A
Asias He 已提交
214
{
215 216
	return (PAGE_ALIGN((unsigned long)iov_base + iov_len) -
	       ((unsigned long)iov_base & PAGE_MASK)) >> PAGE_SHIFT;
A
Asias He 已提交
217 218
}

219
static void vhost_scsi_done_inflight(struct kref *kref)
220 221 222 223 224 225 226
{
	struct vhost_scsi_inflight *inflight;

	inflight = container_of(kref, struct vhost_scsi_inflight, kref);
	complete(&inflight->comp);
}

227
static void vhost_scsi_init_inflight(struct vhost_scsi *vs,
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
				    struct vhost_scsi_inflight *old_inflight[])
{
	struct vhost_scsi_inflight *new_inflight;
	struct vhost_virtqueue *vq;
	int idx, i;

	for (i = 0; i < VHOST_SCSI_MAX_VQ; i++) {
		vq = &vs->vqs[i].vq;

		mutex_lock(&vq->mutex);

		/* store old infight */
		idx = vs->vqs[i].inflight_idx;
		if (old_inflight)
			old_inflight[i] = &vs->vqs[i].inflights[idx];

		/* setup new infight */
		vs->vqs[i].inflight_idx = idx ^ 1;
		new_inflight = &vs->vqs[i].inflights[idx ^ 1];
		kref_init(&new_inflight->kref);
		init_completion(&new_inflight->comp);

		mutex_unlock(&vq->mutex);
	}
}

static struct vhost_scsi_inflight *
255
vhost_scsi_get_inflight(struct vhost_virtqueue *vq)
256 257 258 259 260 261 262 263 264 265 266
{
	struct vhost_scsi_inflight *inflight;
	struct vhost_scsi_virtqueue *svq;

	svq = container_of(vq, struct vhost_scsi_virtqueue, vq);
	inflight = &svq->inflights[svq->inflight_idx];
	kref_get(&inflight->kref);

	return inflight;
}

267
static void vhost_scsi_put_inflight(struct vhost_scsi_inflight *inflight)
268
{
269
	kref_put(&inflight->kref, vhost_scsi_done_inflight);
270 271
}

272
static int vhost_scsi_check_true(struct se_portal_group *se_tpg)
273 274 275 276
{
	return 1;
}

277
static int vhost_scsi_check_false(struct se_portal_group *se_tpg)
278 279 280 281
{
	return 0;
}

282
static char *vhost_scsi_get_fabric_name(void)
283 284 285 286
{
	return "vhost";
}

287
static char *vhost_scsi_get_fabric_wwn(struct se_portal_group *se_tpg)
288
{
289 290 291
	struct vhost_scsi_tpg *tpg = container_of(se_tpg,
				struct vhost_scsi_tpg, se_tpg);
	struct vhost_scsi_tport *tport = tpg->tport;
292 293 294 295

	return &tport->tport_name[0];
}

296
static u16 vhost_scsi_get_tpgt(struct se_portal_group *se_tpg)
297
{
298 299
	struct vhost_scsi_tpg *tpg = container_of(se_tpg,
				struct vhost_scsi_tpg, se_tpg);
300 301 302
	return tpg->tport_tpgt;
}

303 304 305 306 307 308 309 310
static int vhost_scsi_check_prot_fabric_only(struct se_portal_group *se_tpg)
{
	struct vhost_scsi_tpg *tpg = container_of(se_tpg,
				struct vhost_scsi_tpg, se_tpg);

	return tpg->tv_fabric_prot_type;
}

311
static u32 vhost_scsi_tpg_get_inst_index(struct se_portal_group *se_tpg)
312 313 314 315
{
	return 1;
}

316
static void vhost_scsi_release_cmd(struct se_cmd *se_cmd)
317
{
318 319
	struct vhost_scsi_cmd *tv_cmd = container_of(se_cmd,
				struct vhost_scsi_cmd, tvc_se_cmd);
320
	struct se_session *se_sess = tv_cmd->tvc_nexus->tvn_se_sess;
321
	int i;
322 323 324 325

	if (tv_cmd->tvc_sgl_count) {
		for (i = 0; i < tv_cmd->tvc_sgl_count; i++)
			put_page(sg_page(&tv_cmd->tvc_sgl[i]));
326
	}
327 328 329 330
	if (tv_cmd->tvc_prot_sgl_count) {
		for (i = 0; i < tv_cmd->tvc_prot_sgl_count; i++)
			put_page(sg_page(&tv_cmd->tvc_prot_sgl[i]));
	}
331

332
	vhost_scsi_put_inflight(tv_cmd->inflight);
333
	percpu_ida_free(&se_sess->sess_tag_pool, se_cmd->map_tag);
334 335
}

336
static int vhost_scsi_shutdown_session(struct se_session *se_sess)
337 338 339 340
{
	return 0;
}

341
static void vhost_scsi_close_session(struct se_session *se_sess)
342 343 344 345
{
	return;
}

346
static u32 vhost_scsi_sess_get_index(struct se_session *se_sess)
347 348 349 350
{
	return 0;
}

351
static int vhost_scsi_write_pending(struct se_cmd *se_cmd)
352 353 354 355 356 357
{
	/* Go ahead and process the write immediately */
	target_execute_cmd(se_cmd);
	return 0;
}

358
static int vhost_scsi_write_pending_status(struct se_cmd *se_cmd)
359 360 361 362
{
	return 0;
}

363
static void vhost_scsi_set_default_node_attrs(struct se_node_acl *nacl)
364 365 366 367
{
	return;
}

368
static int vhost_scsi_get_cmd_state(struct se_cmd *se_cmd)
369 370 371 372
{
	return 0;
}

373
static void vhost_scsi_complete_cmd(struct vhost_scsi_cmd *cmd)
374
{
375
	struct vhost_scsi *vs = cmd->tvc_vhost;
376

377
	llist_add(&cmd->tvc_completion_list, &vs->vs_completion_list);
378 379 380

	vhost_work_queue(&vs->dev, &vs->vs_completion_work);
}
381

382
static int vhost_scsi_queue_data_in(struct se_cmd *se_cmd)
383
{
384 385
	struct vhost_scsi_cmd *cmd = container_of(se_cmd,
				struct vhost_scsi_cmd, tvc_se_cmd);
386
	vhost_scsi_complete_cmd(cmd);
387 388 389
	return 0;
}

390
static int vhost_scsi_queue_status(struct se_cmd *se_cmd)
391
{
392 393
	struct vhost_scsi_cmd *cmd = container_of(se_cmd,
				struct vhost_scsi_cmd, tvc_se_cmd);
394
	vhost_scsi_complete_cmd(cmd);
395 396 397
	return 0;
}

398
static void vhost_scsi_queue_tm_rsp(struct se_cmd *se_cmd)
399
{
400
	return;
401 402
}

403
static void vhost_scsi_aborted_task(struct se_cmd *se_cmd)
404 405 406 407
{
	return;
}

408
static void vhost_scsi_free_evt(struct vhost_scsi *vs, struct vhost_scsi_evt *evt)
409 410 411 412 413
{
	vs->vs_events_nr--;
	kfree(evt);
}

414 415
static struct vhost_scsi_evt *
vhost_scsi_allocate_evt(struct vhost_scsi *vs,
416
		       u32 event, u32 reason)
417
{
418
	struct vhost_virtqueue *vq = &vs->vqs[VHOST_SCSI_VQ_EVT].vq;
419
	struct vhost_scsi_evt *evt;
420 421 422 423 424 425 426 427

	if (vs->vs_events_nr > VHOST_SCSI_MAX_EVENT) {
		vs->vs_events_missed = true;
		return NULL;
	}

	evt = kzalloc(sizeof(*evt), GFP_KERNEL);
	if (!evt) {
428
		vq_err(vq, "Failed to allocate vhost_scsi_evt\n");
429 430 431 432
		vs->vs_events_missed = true;
		return NULL;
	}

433 434
	evt->event.event = cpu_to_vhost32(vq, event);
	evt->event.reason = cpu_to_vhost32(vq, reason);
435 436 437 438 439
	vs->vs_events_nr++;

	return evt;
}

440
static void vhost_scsi_free_cmd(struct vhost_scsi_cmd *cmd)
441
{
442
	struct se_cmd *se_cmd = &cmd->tvc_se_cmd;
443 444

	/* TODO locking against target/backend threads? */
445
	transport_generic_free_cmd(se_cmd, 0);
446

447
}
448

449 450
static int vhost_scsi_check_stop_free(struct se_cmd *se_cmd)
{
451
	return target_put_sess_cmd(se_cmd);
452 453
}

454
static void
455
vhost_scsi_do_evt_work(struct vhost_scsi *vs, struct vhost_scsi_evt *evt)
456
{
457
	struct vhost_virtqueue *vq = &vs->vqs[VHOST_SCSI_VQ_EVT].vq;
458 459 460 461 462 463 464 465 466 467 468 469
	struct virtio_scsi_event *event = &evt->event;
	struct virtio_scsi_event __user *eventp;
	unsigned out, in;
	int head, ret;

	if (!vq->private_data) {
		vs->vs_events_missed = true;
		return;
	}

again:
	vhost_disable_notify(&vs->dev, vq);
470
	head = vhost_get_vq_desc(vq, vq->iov,
471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491
			ARRAY_SIZE(vq->iov), &out, &in,
			NULL, NULL);
	if (head < 0) {
		vs->vs_events_missed = true;
		return;
	}
	if (head == vq->num) {
		if (vhost_enable_notify(&vs->dev, vq))
			goto again;
		vs->vs_events_missed = true;
		return;
	}

	if ((vq->iov[out].iov_len != sizeof(struct virtio_scsi_event))) {
		vq_err(vq, "Expecting virtio_scsi_event, got %zu bytes\n",
				vq->iov[out].iov_len);
		vs->vs_events_missed = true;
		return;
	}

	if (vs->vs_events_missed) {
492
		event->event |= cpu_to_vhost32(vq, VIRTIO_SCSI_T_EVENTS_MISSED);
493 494 495 496 497 498 499 500
		vs->vs_events_missed = false;
	}

	eventp = vq->iov[out].iov_base;
	ret = __copy_to_user(eventp, event, sizeof(*event));
	if (!ret)
		vhost_add_used_and_signal(&vs->dev, vq, head, 0);
	else
501
		vq_err(vq, "Faulted on vhost_scsi_send_event\n");
502 503
}

504
static void vhost_scsi_evt_work(struct vhost_work *work)
505 506 507
{
	struct vhost_scsi *vs = container_of(work, struct vhost_scsi,
					vs_event_work);
508
	struct vhost_virtqueue *vq = &vs->vqs[VHOST_SCSI_VQ_EVT].vq;
509
	struct vhost_scsi_evt *evt;
510 511 512 513 514
	struct llist_node *llnode;

	mutex_lock(&vq->mutex);
	llnode = llist_del_all(&vs->vs_event_list);
	while (llnode) {
515
		evt = llist_entry(llnode, struct vhost_scsi_evt, list);
516
		llnode = llist_next(llnode);
517 518
		vhost_scsi_do_evt_work(vs, evt);
		vhost_scsi_free_evt(vs, evt);
519 520 521 522
	}
	mutex_unlock(&vq->mutex);
}

523 524 525 526 527 528 529 530 531
/* Fill in status and signal that we are done processing this command
 *
 * This is scheduled in the vhost work queue so we are called with the owner
 * process mm and can access the vring.
 */
static void vhost_scsi_complete_cmd_work(struct vhost_work *work)
{
	struct vhost_scsi *vs = container_of(work, struct vhost_scsi,
					vs_completion_work);
A
Asias He 已提交
532
	DECLARE_BITMAP(signal, VHOST_SCSI_MAX_VQ);
533
	struct virtio_scsi_cmd_resp v_rsp;
534
	struct vhost_scsi_cmd *cmd;
535 536
	struct llist_node *llnode;
	struct se_cmd *se_cmd;
537
	struct iov_iter iov_iter;
A
Asias He 已提交
538
	int ret, vq;
539

A
Asias He 已提交
540
	bitmap_zero(signal, VHOST_SCSI_MAX_VQ);
541 542
	llnode = llist_del_all(&vs->vs_completion_list);
	while (llnode) {
543
		cmd = llist_entry(llnode, struct vhost_scsi_cmd,
544 545
				     tvc_completion_list);
		llnode = llist_next(llnode);
546
		se_cmd = &cmd->tvc_se_cmd;
547 548

		pr_debug("%s tv_cmd %p resid %u status %#02x\n", __func__,
549
			cmd, se_cmd->residual_count, se_cmd->scsi_status);
550 551

		memset(&v_rsp, 0, sizeof(v_rsp));
552
		v_rsp.resid = cpu_to_vhost32(cmd->tvc_vq, se_cmd->residual_count);
553 554
		/* TODO is status_qualifier field needed? */
		v_rsp.status = se_cmd->scsi_status;
555 556
		v_rsp.sense_len = cpu_to_vhost32(cmd->tvc_vq,
						 se_cmd->scsi_sense_length);
557
		memcpy(v_rsp.sense, cmd->tvc_sense_buf,
558
		       se_cmd->scsi_sense_length);
559 560 561 562 563

		iov_iter_init(&iov_iter, READ, cmd->tvc_resp_iov,
			      cmd->tvc_in_iovs, sizeof(v_rsp));
		ret = copy_to_iter(&v_rsp, sizeof(v_rsp), &iov_iter);
		if (likely(ret == sizeof(v_rsp))) {
564
			struct vhost_scsi_virtqueue *q;
565 566
			vhost_add_used(cmd->tvc_vq, cmd->tvc_vq_desc, 0);
			q = container_of(cmd->tvc_vq, struct vhost_scsi_virtqueue, vq);
567
			vq = q - vs->vqs;
A
Asias He 已提交
568 569
			__set_bit(vq, signal);
		} else
570 571
			pr_err("Faulted on virtio_scsi_cmd_resp\n");

572
		vhost_scsi_free_cmd(cmd);
573 574
	}

A
Asias He 已提交
575 576 577
	vq = -1;
	while ((vq = find_next_bit(signal, VHOST_SCSI_MAX_VQ, vq + 1))
		< VHOST_SCSI_MAX_VQ)
578
		vhost_signal(&vs->dev, &vs->vqs[vq].vq);
579 580
}

581 582
static struct vhost_scsi_cmd *
vhost_scsi_get_tag(struct vhost_virtqueue *vq, struct vhost_scsi_tpg *tpg,
583 584
		   unsigned char *cdb, u64 scsi_tag, u16 lun, u8 task_attr,
		   u32 exp_data_len, int data_direction)
585
{
586 587
	struct vhost_scsi_cmd *cmd;
	struct vhost_scsi_nexus *tv_nexus;
588
	struct se_session *se_sess;
589
	struct scatterlist *sg, *prot_sg;
590
	struct page **pages;
591
	int tag;
592

593
	tv_nexus = tpg->tpg_nexus;
594
	if (!tv_nexus) {
595
		pr_err("Unable to locate active struct vhost_scsi_nexus\n");
596 597
		return ERR_PTR(-EIO);
	}
598
	se_sess = tv_nexus->tvn_se_sess;
599

600
	tag = percpu_ida_alloc(&se_sess->sess_tag_pool, TASK_RUNNING);
601
	if (tag < 0) {
602
		pr_err("Unable to obtain tag for vhost_scsi_cmd\n");
603 604 605
		return ERR_PTR(-ENOMEM);
	}

606
	cmd = &((struct vhost_scsi_cmd *)se_sess->sess_cmd_map)[tag];
607
	sg = cmd->tvc_sgl;
608
	prot_sg = cmd->tvc_prot_sgl;
609
	pages = cmd->tvc_upages;
610
	memset(cmd, 0, sizeof(struct vhost_scsi_cmd));
611

612
	cmd->tvc_sgl = sg;
613
	cmd->tvc_prot_sgl = prot_sg;
614
	cmd->tvc_upages = pages;
615
	cmd->tvc_se_cmd.map_tag = tag;
616 617 618
	cmd->tvc_tag = scsi_tag;
	cmd->tvc_lun = lun;
	cmd->tvc_task_attr = task_attr;
619 620 621
	cmd->tvc_exp_data_len = exp_data_len;
	cmd->tvc_data_direction = data_direction;
	cmd->tvc_nexus = tv_nexus;
622
	cmd->inflight = vhost_scsi_get_inflight(vq);
623

624
	memcpy(cmd->tvc_cdb, cdb, VHOST_SCSI_MAX_CDB_SIZE);
625

626
	return cmd;
627 628 629 630 631 632 633
}

/*
 * Map a user memory range into a scatterlist
 *
 * Returns the number of scatterlist entries used or -errno on error.
 */
634
static int
635
vhost_scsi_map_to_sgl(struct vhost_scsi_cmd *cmd,
636 637
		      void __user *ptr,
		      size_t len,
638
		      struct scatterlist *sgl,
639
		      bool write)
640
{
641 642
	unsigned int npages = 0, offset, nbytes;
	unsigned int pages_nr = iov_num_pages(ptr, len);
643
	struct scatterlist *sg = sgl;
644
	struct page **pages = cmd->tvc_upages;
645
	int ret, i;
646

647
	if (pages_nr > VHOST_SCSI_PREALLOC_UPAGES) {
648
		pr_err("vhost_scsi_map_to_sgl() pages_nr: %u greater than"
649 650
		       " preallocated VHOST_SCSI_PREALLOC_UPAGES: %u\n",
			pages_nr, VHOST_SCSI_PREALLOC_UPAGES);
651 652 653
		return -ENOBUFS;
	}

654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669
	ret = get_user_pages_fast((unsigned long)ptr, pages_nr, write, pages);
	/* No pages were pinned */
	if (ret < 0)
		goto out;
	/* Less pages pinned than wanted */
	if (ret != pages_nr) {
		for (i = 0; i < ret; i++)
			put_page(pages[i]);
		ret = -EFAULT;
		goto out;
	}

	while (len > 0) {
		offset = (uintptr_t)ptr & ~PAGE_MASK;
		nbytes = min_t(unsigned int, PAGE_SIZE - offset, len);
		sg_set_page(sg, pages[npages], nbytes, offset);
670 671 672 673 674 675
		ptr += nbytes;
		len -= nbytes;
		sg++;
		npages++;
	}

676
out:
677 678 679
	return ret;
}

680
static int
681
vhost_scsi_calc_sgls(struct iov_iter *iter, size_t bytes, int max_sgls)
682
{
683
	int sgl_count = 0;
684

685 686 687 688 689
	if (!iter || !iter->iov) {
		pr_err("%s: iter->iov is NULL, but expected bytes: %zu"
		       " present\n", __func__, bytes);
		return -EINVAL;
	}
690

691 692 693 694 695
	sgl_count = iov_iter_npages(iter, 0xffff);
	if (sgl_count > max_sgls) {
		pr_err("%s: requested sgl_count: %d exceeds pre-allocated"
		       " max_sgls: %d\n", __func__, sgl_count, max_sgls);
		return -EINVAL;
696
	}
697 698
	return sgl_count;
}
699

700
static int
701 702 703
vhost_scsi_iov_to_sgl(struct vhost_scsi_cmd *cmd, bool write,
		      struct iov_iter *iter,
		      struct scatterlist *sg, int sg_count)
704 705 706
{
	size_t off = iter->iov_offset;
	int i, ret;
707

708 709 710
	for (i = 0; i < iter->nr_segs; i++) {
		void __user *base = iter->iov[i].iov_base + off;
		size_t len = iter->iov[i].iov_len - off;
711

712
		ret = vhost_scsi_map_to_sgl(cmd, base, len, sg, write);
713
		if (ret < 0) {
714 715 716 717 718
			for (i = 0; i < sg_count; i++) {
				struct page *page = sg_page(&sg[i]);
				if (page)
					put_page(page);
			}
719 720 721
			return ret;
		}
		sg += ret;
722
		off = 0;
723 724 725 726
	}
	return 0;
}

727
static int
728
vhost_scsi_mapal(struct vhost_scsi_cmd *cmd,
729 730 731 732 733 734 735 736
		 size_t prot_bytes, struct iov_iter *prot_iter,
		 size_t data_bytes, struct iov_iter *data_iter)
{
	int sgl_count, ret;
	bool write = (cmd->tvc_data_direction == DMA_FROM_DEVICE);

	if (prot_bytes) {
		sgl_count = vhost_scsi_calc_sgls(prot_iter, prot_bytes,
737
						 VHOST_SCSI_PREALLOC_PROT_SGLS);
738 739 740 741 742 743 744 745 746 747 748
		if (sgl_count < 0)
			return sgl_count;

		sg_init_table(cmd->tvc_prot_sgl, sgl_count);
		cmd->tvc_prot_sgl_count = sgl_count;
		pr_debug("%s prot_sg %p prot_sgl_count %u\n", __func__,
			 cmd->tvc_prot_sgl, cmd->tvc_prot_sgl_count);

		ret = vhost_scsi_iov_to_sgl(cmd, write, prot_iter,
					    cmd->tvc_prot_sgl,
					    cmd->tvc_prot_sgl_count);
749 750 751 752
		if (ret < 0) {
			cmd->tvc_prot_sgl_count = 0;
			return ret;
		}
753 754
	}
	sgl_count = vhost_scsi_calc_sgls(data_iter, data_bytes,
755
					 VHOST_SCSI_PREALLOC_SGLS);
756 757 758 759 760 761 762 763 764 765 766 767 768
	if (sgl_count < 0)
		return sgl_count;

	sg_init_table(cmd->tvc_sgl, sgl_count);
	cmd->tvc_sgl_count = sgl_count;
	pr_debug("%s data_sg %p data_sgl_count %u\n", __func__,
		  cmd->tvc_sgl, cmd->tvc_sgl_count);

	ret = vhost_scsi_iov_to_sgl(cmd, write, data_iter,
				    cmd->tvc_sgl, cmd->tvc_sgl_count);
	if (ret < 0) {
		cmd->tvc_sgl_count = 0;
		return ret;
769 770 771 772
	}
	return 0;
}

773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789
static int vhost_scsi_to_tcm_attr(int attr)
{
	switch (attr) {
	case VIRTIO_SCSI_S_SIMPLE:
		return TCM_SIMPLE_TAG;
	case VIRTIO_SCSI_S_ORDERED:
		return TCM_ORDERED_TAG;
	case VIRTIO_SCSI_S_HEAD:
		return TCM_HEAD_TAG;
	case VIRTIO_SCSI_S_ACA:
		return TCM_ACA_TAG;
	default:
		break;
	}
	return TCM_SIMPLE_TAG;
}

790
static void vhost_scsi_submission_work(struct work_struct *work)
791
{
792 793 794
	struct vhost_scsi_cmd *cmd =
		container_of(work, struct vhost_scsi_cmd, work);
	struct vhost_scsi_nexus *tv_nexus;
795
	struct se_cmd *se_cmd = &cmd->tvc_se_cmd;
796 797
	struct scatterlist *sg_ptr, *sg_prot_ptr = NULL;
	int rc;
798

799
	/* FIXME: BIDI operation */
800 801
	if (cmd->tvc_sgl_count) {
		sg_ptr = cmd->tvc_sgl;
802 803 804 805 806

		if (cmd->tvc_prot_sgl_count)
			sg_prot_ptr = cmd->tvc_prot_sgl;
		else
			se_cmd->prot_pto = true;
807 808 809
	} else {
		sg_ptr = NULL;
	}
810
	tv_nexus = cmd->tvc_nexus;
811

812
	se_cmd->tag = 0;
813
	rc = target_submit_cmd_map_sgls(se_cmd, tv_nexus->tvn_se_sess,
814 815
			cmd->tvc_cdb, &cmd->tvc_sense_buf[0],
			cmd->tvc_lun, cmd->tvc_exp_data_len,
816 817 818 819
			vhost_scsi_to_tcm_attr(cmd->tvc_task_attr),
			cmd->tvc_data_direction, TARGET_SCF_ACK_KREF,
			sg_ptr, cmd->tvc_sgl_count, NULL, 0, sg_prot_ptr,
			cmd->tvc_prot_sgl_count);
820 821
	if (rc < 0) {
		transport_send_check_condition_and_sense(se_cmd,
822
				TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE, 0);
823 824 825 826
		transport_generic_free_cmd(se_cmd, 0);
	}
}

827 828 829 830
static void
vhost_scsi_send_bad_target(struct vhost_scsi *vs,
			   struct vhost_virtqueue *vq,
			   int head, unsigned out)
831 832 833 834 835 836 837 838 839 840 841 842 843 844 845
{
	struct virtio_scsi_cmd_resp __user *resp;
	struct virtio_scsi_cmd_resp rsp;
	int ret;

	memset(&rsp, 0, sizeof(rsp));
	rsp.response = VIRTIO_SCSI_S_BAD_TARGET;
	resp = vq->iov[out].iov_base;
	ret = __copy_to_user(resp, &rsp, sizeof(rsp));
	if (!ret)
		vhost_add_used_and_signal(&vs->dev, vq, head, 0);
	else
		pr_err("Faulted on virtio_scsi_cmd_resp\n");
}

846 847
static void
vhost_scsi_handle_vq(struct vhost_scsi *vs, struct vhost_virtqueue *vq)
848
{
849
	struct vhost_scsi_tpg **vs_tpg, *tpg;
850
	struct virtio_scsi_cmd_req v_req;
851
	struct virtio_scsi_cmd_req_pi v_req_pi;
852
	struct vhost_scsi_cmd *cmd;
853
	struct iov_iter out_iter, in_iter, prot_iter, data_iter;
854
	u64 tag;
855 856 857 858 859
	u32 exp_data_len, data_direction;
	unsigned out, in;
	int head, ret, prot_bytes;
	size_t req_size, rsp_size = sizeof(struct virtio_scsi_cmd_resp);
	size_t out_size, in_size;
860 861
	u16 lun;
	u8 *target, *lunp, task_attr;
862
	bool t10_pi = vhost_has_feature(vq, VIRTIO_SCSI_F_T10_PI);
863
	void *req, *cdb;
864

865
	mutex_lock(&vq->mutex);
866 867 868 869
	/*
	 * We can handle the vq only after the endpoint is setup by calling the
	 * VHOST_SCSI_SET_ENDPOINT ioctl.
	 */
870
	vs_tpg = vq->private_data;
871
	if (!vs_tpg)
872
		goto out;
873 874 875 876

	vhost_disable_notify(&vs->dev, vq);

	for (;;) {
877
		head = vhost_get_vq_desc(vq, vq->iov,
878 879
					 ARRAY_SIZE(vq->iov), &out, &in,
					 NULL, NULL);
880
		pr_debug("vhost_get_vq_desc: head: %d, out: %u in: %u\n",
881
			 head, out, in);
882 883 884 885 886 887 888 889 890 891 892 893
		/* On error, stop handling until the next kick. */
		if (unlikely(head < 0))
			break;
		/* Nothing new?  Wait for eventfd to tell us they refilled. */
		if (head == vq->num) {
			if (unlikely(vhost_enable_notify(&vs->dev, vq))) {
				vhost_disable_notify(&vs->dev, vq);
				continue;
			}
			break;
		}
		/*
894 895
		 * Check for a sane response buffer so we can report early
		 * errors back to the guest.
896
		 */
897 898 899
		if (unlikely(vq->iov[out].iov_len < rsp_size)) {
			vq_err(vq, "Expecting at least virtio_scsi_cmd_resp"
				" size, got %zu bytes\n", vq->iov[out].iov_len);
900 901
			break;
		}
902 903 904 905 906
		/*
		 * Setup pointers and values based upon different virtio-scsi
		 * request header if T10_PI is enabled in KVM guest.
		 */
		if (t10_pi) {
907
			req = &v_req_pi;
908
			req_size = sizeof(v_req_pi);
909 910 911 912
			lunp = &v_req_pi.lun[0];
			target = &v_req_pi.lun[1];
		} else {
			req = &v_req;
913
			req_size = sizeof(v_req);
914 915 916
			lunp = &v_req.lun[0];
			target = &v_req.lun[1];
		}
917 918 919 920 921
		/*
		 * FIXME: Not correct for BIDI operation
		 */
		out_size = iov_length(vq->iov, out);
		in_size = iov_length(&vq->iov[out], in);
922

923 924 925 926 927 928 929 930 931 932 933
		/*
		 * Copy over the virtio-scsi request header, which for a
		 * ANY_LAYOUT enabled guest may span multiple iovecs, or a
		 * single iovec may contain both the header + outgoing
		 * WRITE payloads.
		 *
		 * copy_from_iter() will advance out_iter, so that it will
		 * point at the start of the outgoing WRITE payload, if
		 * DMA_TO_DEVICE is set.
		 */
		iov_iter_init(&out_iter, WRITE, vq->iov, out, out_size);
934

935 936 937
		ret = copy_from_iter(req, req_size, &out_iter);
		if (unlikely(ret != req_size)) {
			vq_err(vq, "Faulted on copy_from_iter\n");
938 939
			vhost_scsi_send_bad_target(vs, vq, head, out);
			continue;
940
		}
941
		/* virtio-scsi spec requires byte 0 of the lun to be 1 */
942
		if (unlikely(*lunp != 1)) {
943
			vq_err(vq, "Illegal virtio-scsi lun: %u\n", *lunp);
944 945 946 947
			vhost_scsi_send_bad_target(vs, vq, head, out);
			continue;
		}

948
		tpg = ACCESS_ONCE(vs_tpg[*target]);
949
		if (unlikely(!tpg)) {
950
			/* Target does not exist, fail the request */
951
			vhost_scsi_send_bad_target(vs, vq, head, out);
A
Asias He 已提交
952 953
			continue;
		}
954
		/*
955 956 957
		 * Determine data_direction by calculating the total outgoing
		 * iovec sizes + incoming iovec sizes vs. virtio-scsi request +
		 * response headers respectively.
958
		 *
959 960 961 962 963 964
		 * For DMA_TO_DEVICE this is out_iter, which is already pointing
		 * to the right place.
		 *
		 * For DMA_FROM_DEVICE, the iovec will be just past the end
		 * of the virtio-scsi response header in either the same
		 * or immediately following iovec.
965
		 *
966 967
		 * Any associated T10_PI bytes for the outgoing / incoming
		 * payloads are included in calculation of exp_data_len here.
968
		 */
969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990
		prot_bytes = 0;

		if (out_size > req_size) {
			data_direction = DMA_TO_DEVICE;
			exp_data_len = out_size - req_size;
			data_iter = out_iter;
		} else if (in_size > rsp_size) {
			data_direction = DMA_FROM_DEVICE;
			exp_data_len = in_size - rsp_size;

			iov_iter_init(&in_iter, READ, &vq->iov[out], in,
				      rsp_size + exp_data_len);
			iov_iter_advance(&in_iter, rsp_size);
			data_iter = in_iter;
		} else {
			data_direction = DMA_NONE;
			exp_data_len = 0;
		}
		/*
		 * If T10_PI header + payload is present, setup prot_iter values
		 * and recalculate data_iter for vhost_scsi_mapal() mapping to
		 * host scatterlists via get_user_pages_fast().
991
		 */
992
		if (t10_pi) {
993 994
			if (v_req_pi.pi_bytesout) {
				if (data_direction != DMA_TO_DEVICE) {
995 996
					vq_err(vq, "Received non zero pi_bytesout,"
						" but wrong data_direction\n");
997 998
					vhost_scsi_send_bad_target(vs, vq, head, out);
					continue;
999
				}
1000
				prot_bytes = vhost32_to_cpu(vq, v_req_pi.pi_bytesout);
1001 1002
			} else if (v_req_pi.pi_bytesin) {
				if (data_direction != DMA_FROM_DEVICE) {
1003 1004
					vq_err(vq, "Received non zero pi_bytesin,"
						" but wrong data_direction\n");
1005 1006
					vhost_scsi_send_bad_target(vs, vq, head, out);
					continue;
1007
				}
1008
				prot_bytes = vhost32_to_cpu(vq, v_req_pi.pi_bytesin);
1009
			}
1010 1011 1012 1013 1014 1015 1016
			/*
			 * Set prot_iter to data_iter, and advance past any
			 * preceeding prot_bytes that may be present.
			 *
			 * Also fix up the exp_data_len to reflect only the
			 * actual data payload length.
			 */
1017
			if (prot_bytes) {
1018 1019 1020
				exp_data_len -= prot_bytes;
				prot_iter = data_iter;
				iov_iter_advance(&data_iter, prot_bytes);
1021
			}
1022
			tag = vhost64_to_cpu(vq, v_req_pi.tag);
1023 1024 1025 1026
			task_attr = v_req_pi.task_attr;
			cdb = &v_req_pi.cdb[0];
			lun = ((v_req_pi.lun[2] << 8) | v_req_pi.lun[3]) & 0x3FFF;
		} else {
1027
			tag = vhost64_to_cpu(vq, v_req.tag);
1028 1029 1030 1031 1032
			task_attr = v_req.task_attr;
			cdb = &v_req.cdb[0];
			lun = ((v_req.lun[2] << 8) | v_req.lun[3]) & 0x3FFF;
		}
		/*
1033 1034 1035
		 * Check that the received CDB size does not exceeded our
		 * hardcoded max for vhost-scsi, then get a pre-allocated
		 * cmd descriptor for the new virtio-scsi tag.
1036 1037 1038
		 *
		 * TODO what if cdb was too small for varlen cdb header?
		 */
1039
		if (unlikely(scsi_command_size(cdb) > VHOST_SCSI_MAX_CDB_SIZE)) {
1040 1041
			vq_err(vq, "Received SCSI CDB with command_size: %d that"
				" exceeds SCSI_MAX_VARLEN_CDB_SIZE: %d\n",
1042
				scsi_command_size(cdb), VHOST_SCSI_MAX_CDB_SIZE);
1043 1044
			vhost_scsi_send_bad_target(vs, vq, head, out);
			continue;
1045 1046
		}
		cmd = vhost_scsi_get_tag(vq, tpg, cdb, tag, lun, task_attr,
1047 1048
					 exp_data_len + prot_bytes,
					 data_direction);
1049
		if (IS_ERR(cmd)) {
1050
			vq_err(vq, "vhost_scsi_get_tag failed %ld\n",
1051
			       PTR_ERR(cmd));
1052 1053
			vhost_scsi_send_bad_target(vs, vq, head, out);
			continue;
1054
		}
1055 1056
		cmd->tvc_vhost = vs;
		cmd->tvc_vq = vq;
1057 1058
		cmd->tvc_resp_iov = &vq->iov[out];
		cmd->tvc_in_iovs = in;
1059 1060

		pr_debug("vhost_scsi got command opcode: %#02x, lun: %d\n",
1061 1062 1063
			 cmd->tvc_cdb[0], cmd->tvc_lun);
		pr_debug("cmd: %p exp_data_len: %d, prot_bytes: %d data_direction:"
			 " %d\n", cmd, exp_data_len, prot_bytes, data_direction);
1064 1065

		if (data_direction != DMA_NONE) {
1066 1067 1068
			ret = vhost_scsi_mapal(cmd,
					       prot_bytes, &prot_iter,
					       exp_data_len, &data_iter);
1069 1070
			if (unlikely(ret)) {
				vq_err(vq, "Failed to map iov to sgl\n");
1071
				vhost_scsi_release_cmd(&cmd->tvc_se_cmd);
1072 1073
				vhost_scsi_send_bad_target(vs, vq, head, out);
				continue;
1074 1075 1076 1077 1078
			}
		}
		/*
		 * Save the descriptor from vhost_get_vq_desc() to be used to
		 * complete the virtio-scsi request in TCM callback context via
1079
		 * vhost_scsi_queue_data_in() and vhost_scsi_queue_status()
1080
		 */
1081
		cmd->tvc_vq_desc = head;
1082
		/*
1083 1084 1085 1086
		 * Dispatch cmd descriptor for cmwq execution in process
		 * context provided by vhost_scsi_workqueue.  This also ensures
		 * cmd is executed on the same kworker CPU as this vhost
		 * thread to gain positive L2 cache locality effects.
1087
		 */
1088 1089
		INIT_WORK(&cmd->work, vhost_scsi_submission_work);
		queue_work(vhost_scsi_workqueue, &cmd->work);
1090
	}
1091
out:
1092
	mutex_unlock(&vq->mutex);
1093 1094 1095 1096
}

static void vhost_scsi_ctl_handle_kick(struct vhost_work *work)
{
1097
	pr_debug("%s: The handling func for control queue.\n", __func__);
1098 1099
}

1100
static void
1101 1102
vhost_scsi_send_evt(struct vhost_scsi *vs,
		   struct vhost_scsi_tpg *tpg,
1103 1104 1105
		   struct se_lun *lun,
		   u32 event,
		   u32 reason)
1106
{
1107
	struct vhost_scsi_evt *evt;
1108

1109
	evt = vhost_scsi_allocate_evt(vs, event, reason);
1110 1111 1112 1113 1114 1115 1116 1117 1118 1119
	if (!evt)
		return;

	if (tpg && lun) {
		/* TODO: share lun setup code with virtio-scsi.ko */
		/*
		 * Note: evt->event is zeroed when we allocate it and
		 * lun[4-7] need to be zero according to virtio-scsi spec.
		 */
		evt->event.lun[0] = 0x01;
1120
		evt->event.lun[1] = tpg->tport_tpgt;
1121 1122 1123 1124 1125 1126 1127 1128 1129
		if (lun->unpacked_lun >= 256)
			evt->event.lun[2] = lun->unpacked_lun >> 8 | 0x40 ;
		evt->event.lun[3] = lun->unpacked_lun & 0xFF;
	}

	llist_add(&evt->list, &vs->vs_event_list);
	vhost_work_queue(&vs->dev, &vs->vs_event_work);
}

1130 1131
static void vhost_scsi_evt_handle_kick(struct vhost_work *work)
{
1132 1133 1134 1135 1136 1137 1138 1139 1140
	struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue,
						poll.work);
	struct vhost_scsi *vs = container_of(vq->dev, struct vhost_scsi, dev);

	mutex_lock(&vq->mutex);
	if (!vq->private_data)
		goto out;

	if (vs->vs_events_missed)
1141
		vhost_scsi_send_evt(vs, NULL, NULL, VIRTIO_SCSI_T_NO_EVENT, 0);
1142 1143
out:
	mutex_unlock(&vq->mutex);
1144 1145 1146 1147 1148 1149 1150 1151
}

static void vhost_scsi_handle_kick(struct vhost_work *work)
{
	struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue,
						poll.work);
	struct vhost_scsi *vs = container_of(vq->dev, struct vhost_scsi, dev);

A
Asias He 已提交
1152
	vhost_scsi_handle_vq(vs, vq);
1153 1154
}

1155 1156
static void vhost_scsi_flush_vq(struct vhost_scsi *vs, int index)
{
1157
	vhost_poll_flush(&vs->vqs[index].vq.poll);
1158 1159
}

1160
/* Callers must hold dev mutex */
1161 1162
static void vhost_scsi_flush(struct vhost_scsi *vs)
{
1163
	struct vhost_scsi_inflight *old_inflight[VHOST_SCSI_MAX_VQ];
1164 1165
	int i;

1166
	/* Init new inflight and remember the old inflight */
1167
	vhost_scsi_init_inflight(vs, old_inflight);
1168 1169 1170 1171 1172 1173 1174

	/*
	 * The inflight->kref was initialized to 1. We decrement it here to
	 * indicate the start of the flush operation so that it will reach 0
	 * when all the reqs are finished.
	 */
	for (i = 0; i < VHOST_SCSI_MAX_VQ; i++)
1175
		kref_put(&old_inflight[i]->kref, vhost_scsi_done_inflight);
1176 1177

	/* Flush both the vhost poll and vhost work */
1178 1179 1180
	for (i = 0; i < VHOST_SCSI_MAX_VQ; i++)
		vhost_scsi_flush_vq(vs, i);
	vhost_work_flush(&vs->dev, &vs->vs_completion_work);
1181
	vhost_work_flush(&vs->dev, &vs->vs_event_work);
1182 1183 1184 1185

	/* Wait for all reqs issued before the flush to be finished */
	for (i = 0; i < VHOST_SCSI_MAX_VQ; i++)
		wait_for_completion(&old_inflight[i]->comp);
1186 1187
}

1188 1189
/*
 * Called from vhost_scsi_ioctl() context to walk the list of available
1190
 * vhost_scsi_tpg with an active struct vhost_scsi_nexus
1191 1192
 *
 *  The lock nesting rule is:
1193
 *    vhost_scsi_mutex -> vs->dev.mutex -> tpg->tv_tpg_mutex -> vq->mutex
1194
 */
1195 1196 1197
static int
vhost_scsi_set_endpoint(struct vhost_scsi *vs,
			struct vhost_scsi_target *t)
1198
{
1199
	struct se_portal_group *se_tpg;
1200 1201 1202
	struct vhost_scsi_tport *tv_tport;
	struct vhost_scsi_tpg *tpg;
	struct vhost_scsi_tpg **vs_tpg;
1203 1204
	struct vhost_virtqueue *vq;
	int index, ret, i, len;
A
Asias He 已提交
1205
	bool match = false;
1206

1207
	mutex_lock(&vhost_scsi_mutex);
1208
	mutex_lock(&vs->dev.mutex);
1209

1210 1211 1212
	/* Verify that ring has been setup correctly. */
	for (index = 0; index < vs->dev.nvqs; ++index) {
		/* Verify that ring has been setup correctly. */
1213
		if (!vhost_vq_access_ok(&vs->vqs[index].vq)) {
1214 1215
			ret = -EFAULT;
			goto out;
1216 1217 1218
		}
	}

1219 1220 1221
	len = sizeof(vs_tpg[0]) * VHOST_SCSI_MAX_TARGET;
	vs_tpg = kzalloc(len, GFP_KERNEL);
	if (!vs_tpg) {
1222 1223
		ret = -ENOMEM;
		goto out;
1224 1225 1226 1227
	}
	if (vs->vs_tpg)
		memcpy(vs_tpg, vs->vs_tpg, len);

1228
	list_for_each_entry(tpg, &vhost_scsi_list, tv_tpg_list) {
1229 1230 1231
		mutex_lock(&tpg->tv_tpg_mutex);
		if (!tpg->tpg_nexus) {
			mutex_unlock(&tpg->tv_tpg_mutex);
1232 1233
			continue;
		}
1234 1235
		if (tpg->tv_tpg_vhost_count != 0) {
			mutex_unlock(&tpg->tv_tpg_mutex);
1236 1237
			continue;
		}
1238
		tv_tport = tpg->tport;
1239

A
Asias He 已提交
1240
		if (!strcmp(tv_tport->tport_name, t->vhost_wwpn)) {
1241
			if (vs->vs_tpg && vs->vs_tpg[tpg->tport_tpgt]) {
1242
				kfree(vs_tpg);
1243
				mutex_unlock(&tpg->tv_tpg_mutex);
1244 1245
				ret = -EEXIST;
				goto out;
1246
			}
1247 1248 1249 1250 1251 1252 1253
			/*
			 * In order to ensure individual vhost-scsi configfs
			 * groups cannot be removed while in use by vhost ioctl,
			 * go ahead and take an explicit se_tpg->tpg_group.cg_item
			 * dependency now.
			 */
			se_tpg = &tpg->se_tpg;
1254
			ret = target_depend_item(&se_tpg->tpg_group.cg_item);
1255 1256 1257 1258 1259 1260
			if (ret) {
				pr_warn("configfs_depend_item() failed: %d\n", ret);
				kfree(vs_tpg);
				mutex_unlock(&tpg->tv_tpg_mutex);
				goto out;
			}
1261 1262 1263
			tpg->tv_tpg_vhost_count++;
			tpg->vhost_scsi = vs;
			vs_tpg[tpg->tport_tpgt] = tpg;
1264
			smp_mb__after_atomic();
A
Asias He 已提交
1265
			match = true;
1266
		}
1267
		mutex_unlock(&tpg->tv_tpg_mutex);
1268
	}
A
Asias He 已提交
1269 1270 1271 1272

	if (match) {
		memcpy(vs->vs_vhost_wwpn, t->vhost_wwpn,
		       sizeof(vs->vs_vhost_wwpn));
1273
		for (i = 0; i < VHOST_SCSI_MAX_VQ; i++) {
1274
			vq = &vs->vqs[i].vq;
1275
			mutex_lock(&vq->mutex);
A
Asias He 已提交
1276
			vq->private_data = vs_tpg;
1277
			vhost_init_used(vq);
1278 1279
			mutex_unlock(&vq->mutex);
		}
A
Asias He 已提交
1280 1281 1282 1283 1284
		ret = 0;
	} else {
		ret = -EEXIST;
	}

1285 1286 1287 1288 1289 1290 1291 1292
	/*
	 * Act as synchronize_rcu to make sure access to
	 * old vs->vs_tpg is finished.
	 */
	vhost_scsi_flush(vs);
	kfree(vs->vs_tpg);
	vs->vs_tpg = vs_tpg;

1293
out:
A
Asias He 已提交
1294
	mutex_unlock(&vs->dev.mutex);
1295
	mutex_unlock(&vhost_scsi_mutex);
A
Asias He 已提交
1296
	return ret;
1297 1298
}

1299 1300 1301
static int
vhost_scsi_clear_endpoint(struct vhost_scsi *vs,
			  struct vhost_scsi_target *t)
1302
{
1303
	struct se_portal_group *se_tpg;
1304 1305
	struct vhost_scsi_tport *tv_tport;
	struct vhost_scsi_tpg *tpg;
1306 1307
	struct vhost_virtqueue *vq;
	bool match = false;
A
Asias He 已提交
1308 1309
	int index, ret, i;
	u8 target;
1310

1311
	mutex_lock(&vhost_scsi_mutex);
1312 1313 1314
	mutex_lock(&vs->dev.mutex);
	/* Verify that ring has been setup correctly. */
	for (index = 0; index < vs->dev.nvqs; ++index) {
1315
		if (!vhost_vq_access_ok(&vs->vqs[index].vq)) {
1316
			ret = -EFAULT;
1317
			goto err_dev;
1318 1319
		}
	}
1320 1321

	if (!vs->vs_tpg) {
1322 1323
		ret = 0;
		goto err_dev;
1324 1325
	}

A
Asias He 已提交
1326 1327
	for (i = 0; i < VHOST_SCSI_MAX_TARGET; i++) {
		target = i;
1328 1329
		tpg = vs->vs_tpg[target];
		if (!tpg)
A
Asias He 已提交
1330 1331
			continue;

1332 1333
		mutex_lock(&tpg->tv_tpg_mutex);
		tv_tport = tpg->tport;
A
Asias He 已提交
1334 1335
		if (!tv_tport) {
			ret = -ENODEV;
1336
			goto err_tpg;
A
Asias He 已提交
1337 1338 1339
		}

		if (strcmp(tv_tport->tport_name, t->vhost_wwpn)) {
1340
			pr_warn("tv_tport->tport_name: %s, tpg->tport_tpgt: %hu"
A
Asias He 已提交
1341
				" does not match t->vhost_wwpn: %s, t->vhost_tpgt: %hu\n",
1342
				tv_tport->tport_name, tpg->tport_tpgt,
A
Asias He 已提交
1343 1344
				t->vhost_wwpn, t->vhost_tpgt);
			ret = -EINVAL;
1345
			goto err_tpg;
A
Asias He 已提交
1346
		}
1347 1348
		tpg->tv_tpg_vhost_count--;
		tpg->vhost_scsi = NULL;
A
Asias He 已提交
1349
		vs->vs_tpg[target] = NULL;
1350
		match = true;
1351
		mutex_unlock(&tpg->tv_tpg_mutex);
1352 1353 1354 1355 1356
		/*
		 * Release se_tpg->tpg_group.cg_item configfs dependency now
		 * to allow vhost-scsi WWPN se_tpg->tpg_group shutdown to occur.
		 */
		se_tpg = &tpg->se_tpg;
1357
		target_undepend_item(&se_tpg->tpg_group.cg_item);
1358
	}
1359 1360
	if (match) {
		for (i = 0; i < VHOST_SCSI_MAX_VQ; i++) {
1361
			vq = &vs->vqs[i].vq;
1362
			mutex_lock(&vq->mutex);
A
Asias He 已提交
1363
			vq->private_data = NULL;
1364 1365 1366 1367 1368 1369 1370 1371 1372 1373
			mutex_unlock(&vq->mutex);
		}
	}
	/*
	 * Act as synchronize_rcu to make sure access to
	 * old vs->vs_tpg is finished.
	 */
	vhost_scsi_flush(vs);
	kfree(vs->vs_tpg);
	vs->vs_tpg = NULL;
1374
	WARN_ON(vs->vs_events_nr);
1375
	mutex_unlock(&vs->dev.mutex);
1376
	mutex_unlock(&vhost_scsi_mutex);
1377
	return 0;
1378

1379
err_tpg:
1380
	mutex_unlock(&tpg->tv_tpg_mutex);
1381
err_dev:
1382
	mutex_unlock(&vs->dev.mutex);
1383
	mutex_unlock(&vhost_scsi_mutex);
1384
	return ret;
1385 1386
}

1387 1388
static int vhost_scsi_set_features(struct vhost_scsi *vs, u64 features)
{
1389 1390 1391
	struct vhost_virtqueue *vq;
	int i;

1392 1393 1394 1395 1396 1397 1398 1399 1400
	if (features & ~VHOST_SCSI_FEATURES)
		return -EOPNOTSUPP;

	mutex_lock(&vs->dev.mutex);
	if ((features & (1 << VHOST_F_LOG_ALL)) &&
	    !vhost_log_access_ok(&vs->dev)) {
		mutex_unlock(&vs->dev.mutex);
		return -EFAULT;
	}
1401 1402 1403 1404 1405 1406 1407

	for (i = 0; i < VHOST_SCSI_MAX_VQ; i++) {
		vq = &vs->vqs[i].vq;
		mutex_lock(&vq->mutex);
		vq->acked_features = features;
		mutex_unlock(&vq->mutex);
	}
1408 1409 1410 1411
	mutex_unlock(&vs->dev.mutex);
	return 0;
}

1412 1413
static int vhost_scsi_open(struct inode *inode, struct file *f)
{
1414
	struct vhost_scsi *vs;
1415
	struct vhost_virtqueue **vqs;
1416
	int r = -ENOMEM, i;
1417

1418 1419 1420 1421 1422 1423
	vs = kzalloc(sizeof(*vs), GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!vs) {
		vs = vzalloc(sizeof(*vs));
		if (!vs)
			goto err_vs;
	}
1424

1425
	vqs = kmalloc(VHOST_SCSI_MAX_VQ * sizeof(*vqs), GFP_KERNEL);
1426 1427
	if (!vqs)
		goto err_vqs;
1428

1429
	vhost_work_init(&vs->vs_completion_work, vhost_scsi_complete_cmd_work);
1430
	vhost_work_init(&vs->vs_event_work, vhost_scsi_evt_work);
1431

1432 1433
	vs->vs_events_nr = 0;
	vs->vs_events_missed = false;
1434

1435 1436 1437 1438
	vqs[VHOST_SCSI_VQ_CTL] = &vs->vqs[VHOST_SCSI_VQ_CTL].vq;
	vqs[VHOST_SCSI_VQ_EVT] = &vs->vqs[VHOST_SCSI_VQ_EVT].vq;
	vs->vqs[VHOST_SCSI_VQ_CTL].vq.handle_kick = vhost_scsi_ctl_handle_kick;
	vs->vqs[VHOST_SCSI_VQ_EVT].vq.handle_kick = vhost_scsi_evt_handle_kick;
1439
	for (i = VHOST_SCSI_VQ_IO; i < VHOST_SCSI_MAX_VQ; i++) {
1440 1441
		vqs[i] = &vs->vqs[i].vq;
		vs->vqs[i].vq.handle_kick = vhost_scsi_handle_kick;
1442
	}
Z
Zhi Yong Wu 已提交
1443
	vhost_dev_init(&vs->dev, vqs, VHOST_SCSI_MAX_VQ);
1444

1445
	vhost_scsi_init_inflight(vs, NULL);
1446

1447
	f->private_data = vs;
1448
	return 0;
1449 1450

err_vqs:
1451
	kvfree(vs);
1452 1453
err_vs:
	return r;
1454 1455 1456 1457
}

static int vhost_scsi_release(struct inode *inode, struct file *f)
{
1458
	struct vhost_scsi *vs = f->private_data;
A
Asias He 已提交
1459
	struct vhost_scsi_target t;
1460

1461 1462 1463 1464 1465 1466
	mutex_lock(&vs->dev.mutex);
	memcpy(t.vhost_wwpn, vs->vs_vhost_wwpn, sizeof(t.vhost_wwpn));
	mutex_unlock(&vs->dev.mutex);
	vhost_scsi_clear_endpoint(vs, &t);
	vhost_dev_stop(&vs->dev);
	vhost_dev_cleanup(&vs->dev, false);
1467
	/* Jobs can re-queue themselves in evt kick handler. Do extra flush. */
1468 1469
	vhost_scsi_flush(vs);
	kfree(vs->dev.vqs);
1470
	kvfree(vs);
1471 1472 1473
	return 0;
}

1474 1475 1476 1477
static long
vhost_scsi_ioctl(struct file *f,
		 unsigned int ioctl,
		 unsigned long arg)
1478 1479 1480 1481 1482
{
	struct vhost_scsi *vs = f->private_data;
	struct vhost_scsi_target backend;
	void __user *argp = (void __user *)arg;
	u64 __user *featurep = argp;
1483 1484
	u32 __user *eventsp = argp;
	u32 events_missed;
1485
	u64 features;
1486
	int r, abi_version = VHOST_SCSI_ABI_VERSION;
1487
	struct vhost_virtqueue *vq = &vs->vqs[VHOST_SCSI_VQ_EVT].vq;
1488 1489 1490 1491 1492

	switch (ioctl) {
	case VHOST_SCSI_SET_ENDPOINT:
		if (copy_from_user(&backend, argp, sizeof backend))
			return -EFAULT;
1493 1494
		if (backend.reserved != 0)
			return -EOPNOTSUPP;
1495 1496 1497 1498 1499

		return vhost_scsi_set_endpoint(vs, &backend);
	case VHOST_SCSI_CLEAR_ENDPOINT:
		if (copy_from_user(&backend, argp, sizeof backend))
			return -EFAULT;
1500 1501
		if (backend.reserved != 0)
			return -EOPNOTSUPP;
1502 1503 1504

		return vhost_scsi_clear_endpoint(vs, &backend);
	case VHOST_SCSI_GET_ABI_VERSION:
1505
		if (copy_to_user(argp, &abi_version, sizeof abi_version))
1506 1507
			return -EFAULT;
		return 0;
1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521
	case VHOST_SCSI_SET_EVENTS_MISSED:
		if (get_user(events_missed, eventsp))
			return -EFAULT;
		mutex_lock(&vq->mutex);
		vs->vs_events_missed = events_missed;
		mutex_unlock(&vq->mutex);
		return 0;
	case VHOST_SCSI_GET_EVENTS_MISSED:
		mutex_lock(&vq->mutex);
		events_missed = vs->vs_events_missed;
		mutex_unlock(&vq->mutex);
		if (put_user(events_missed, eventsp))
			return -EFAULT;
		return 0;
1522
	case VHOST_GET_FEATURES:
1523
		features = VHOST_SCSI_FEATURES;
1524 1525 1526 1527 1528 1529 1530 1531 1532
		if (copy_to_user(featurep, &features, sizeof features))
			return -EFAULT;
		return 0;
	case VHOST_SET_FEATURES:
		if (copy_from_user(&features, featurep, sizeof features))
			return -EFAULT;
		return vhost_scsi_set_features(vs, features);
	default:
		mutex_lock(&vs->dev.mutex);
1533 1534 1535 1536
		r = vhost_dev_ioctl(&vs->dev, ioctl, argp);
		/* TODO: flush backend after dev ioctl. */
		if (r == -ENOIOCTLCMD)
			r = vhost_vring_ioctl(&vs->dev, ioctl, argp);
1537 1538 1539 1540 1541
		mutex_unlock(&vs->dev.mutex);
		return r;
	}
}

1542 1543 1544 1545 1546 1547 1548 1549
#ifdef CONFIG_COMPAT
static long vhost_scsi_compat_ioctl(struct file *f, unsigned int ioctl,
				unsigned long arg)
{
	return vhost_scsi_ioctl(f, ioctl, (unsigned long)compat_ptr(arg));
}
#endif

1550 1551 1552 1553
static const struct file_operations vhost_scsi_fops = {
	.owner          = THIS_MODULE,
	.release        = vhost_scsi_release,
	.unlocked_ioctl = vhost_scsi_ioctl,
1554 1555 1556
#ifdef CONFIG_COMPAT
	.compat_ioctl	= vhost_scsi_compat_ioctl,
#endif
1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571
	.open           = vhost_scsi_open,
	.llseek		= noop_llseek,
};

static struct miscdevice vhost_scsi_misc = {
	MISC_DYNAMIC_MINOR,
	"vhost-scsi",
	&vhost_scsi_fops,
};

static int __init vhost_scsi_register(void)
{
	return misc_register(&vhost_scsi_misc);
}

1572
static void vhost_scsi_deregister(void)
1573
{
1574
	misc_deregister(&vhost_scsi_misc);
1575 1576
}

1577
static char *vhost_scsi_dump_proto_id(struct vhost_scsi_tport *tport)
1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592
{
	switch (tport->tport_proto_id) {
	case SCSI_PROTOCOL_SAS:
		return "SAS";
	case SCSI_PROTOCOL_FCP:
		return "FCP";
	case SCSI_PROTOCOL_ISCSI:
		return "iSCSI";
	default:
		break;
	}

	return "Unknown";
}

1593
static void
1594
vhost_scsi_do_plug(struct vhost_scsi_tpg *tpg,
1595
		  struct se_lun *lun, bool plug)
1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611
{

	struct vhost_scsi *vs = tpg->vhost_scsi;
	struct vhost_virtqueue *vq;
	u32 reason;

	if (!vs)
		return;

	mutex_lock(&vs->dev.mutex);

	if (plug)
		reason = VIRTIO_SCSI_EVT_RESET_RESCAN;
	else
		reason = VIRTIO_SCSI_EVT_RESET_REMOVED;

1612
	vq = &vs->vqs[VHOST_SCSI_VQ_EVT].vq;
1613
	mutex_lock(&vq->mutex);
1614
	if (vhost_has_feature(vq, VIRTIO_SCSI_F_HOTPLUG))
1615
		vhost_scsi_send_evt(vs, tpg, lun,
1616
				   VIRTIO_SCSI_T_TRANSPORT_RESET, reason);
1617 1618 1619 1620
	mutex_unlock(&vq->mutex);
	mutex_unlock(&vs->dev.mutex);
}

1621
static void vhost_scsi_hotplug(struct vhost_scsi_tpg *tpg, struct se_lun *lun)
1622
{
1623
	vhost_scsi_do_plug(tpg, lun, true);
1624 1625
}

1626
static void vhost_scsi_hotunplug(struct vhost_scsi_tpg *tpg, struct se_lun *lun)
1627
{
1628
	vhost_scsi_do_plug(tpg, lun, false);
1629 1630
}

1631
static int vhost_scsi_port_link(struct se_portal_group *se_tpg,
1632
			       struct se_lun *lun)
1633
{
1634 1635
	struct vhost_scsi_tpg *tpg = container_of(se_tpg,
				struct vhost_scsi_tpg, se_tpg);
1636

1637
	mutex_lock(&vhost_scsi_mutex);
1638

1639 1640 1641
	mutex_lock(&tpg->tv_tpg_mutex);
	tpg->tv_tpg_port_count++;
	mutex_unlock(&tpg->tv_tpg_mutex);
1642

1643
	vhost_scsi_hotplug(tpg, lun);
1644

1645
	mutex_unlock(&vhost_scsi_mutex);
1646

1647 1648 1649
	return 0;
}

1650
static void vhost_scsi_port_unlink(struct se_portal_group *se_tpg,
1651
				  struct se_lun *lun)
1652
{
1653 1654
	struct vhost_scsi_tpg *tpg = container_of(se_tpg,
				struct vhost_scsi_tpg, se_tpg);
1655

1656
	mutex_lock(&vhost_scsi_mutex);
1657

1658 1659 1660
	mutex_lock(&tpg->tv_tpg_mutex);
	tpg->tv_tpg_port_count--;
	mutex_unlock(&tpg->tv_tpg_mutex);
1661

1662
	vhost_scsi_hotunplug(tpg, lun);
1663

1664
	mutex_unlock(&vhost_scsi_mutex);
1665 1666
}

1667
static void vhost_scsi_free_cmd_map_res(struct vhost_scsi_nexus *nexus,
1668 1669
				       struct se_session *se_sess)
{
1670
	struct vhost_scsi_cmd *tv_cmd;
1671 1672 1673 1674 1675
	unsigned int i;

	if (!se_sess->sess_cmd_map)
		return;

1676 1677
	for (i = 0; i < VHOST_SCSI_DEFAULT_TAGS; i++) {
		tv_cmd = &((struct vhost_scsi_cmd *)se_sess->sess_cmd_map)[i];
1678 1679

		kfree(tv_cmd->tvc_sgl);
1680
		kfree(tv_cmd->tvc_prot_sgl);
1681 1682 1683 1684
		kfree(tv_cmd->tvc_upages);
	}
}

1685 1686
static ssize_t vhost_scsi_tpg_attrib_fabric_prot_type_store(
		struct config_item *item, const char *page, size_t count)
1687
{
1688
	struct se_portal_group *se_tpg = attrib_to_tpg(item);
1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706
	struct vhost_scsi_tpg *tpg = container_of(se_tpg,
				struct vhost_scsi_tpg, se_tpg);
	unsigned long val;
	int ret = kstrtoul(page, 0, &val);

	if (ret) {
		pr_err("kstrtoul() returned %d for fabric_prot_type\n", ret);
		return ret;
	}
	if (val != 0 && val != 1 && val != 3) {
		pr_err("Invalid vhost_scsi fabric_prot_type: %lu\n", val);
		return -EINVAL;
	}
	tpg->tv_fabric_prot_type = val;

	return count;
}

1707 1708
static ssize_t vhost_scsi_tpg_attrib_fabric_prot_type_show(
		struct config_item *item, char *page)
1709
{
1710
	struct se_portal_group *se_tpg = attrib_to_tpg(item);
1711 1712 1713 1714 1715
	struct vhost_scsi_tpg *tpg = container_of(se_tpg,
				struct vhost_scsi_tpg, se_tpg);

	return sprintf(page, "%d\n", tpg->tv_fabric_prot_type);
}
1716 1717

CONFIGFS_ATTR(vhost_scsi_tpg_attrib_, fabric_prot_type);
1718 1719

static struct configfs_attribute *vhost_scsi_tpg_attrib_attrs[] = {
1720
	&vhost_scsi_tpg_attrib_attr_fabric_prot_type,
1721 1722 1723
	NULL,
};

1724
static int vhost_scsi_make_nexus(struct vhost_scsi_tpg *tpg,
1725
				const char *name)
1726 1727
{
	struct se_portal_group *se_tpg;
1728
	struct se_session *se_sess;
1729 1730
	struct vhost_scsi_nexus *tv_nexus;
	struct vhost_scsi_cmd *tv_cmd;
1731
	unsigned int i;
1732

1733 1734 1735 1736
	mutex_lock(&tpg->tv_tpg_mutex);
	if (tpg->tpg_nexus) {
		mutex_unlock(&tpg->tv_tpg_mutex);
		pr_debug("tpg->tpg_nexus already exists\n");
1737 1738
		return -EEXIST;
	}
1739
	se_tpg = &tpg->se_tpg;
1740

1741
	tv_nexus = kzalloc(sizeof(struct vhost_scsi_nexus), GFP_KERNEL);
1742
	if (!tv_nexus) {
1743
		mutex_unlock(&tpg->tv_tpg_mutex);
1744
		pr_err("Unable to allocate struct vhost_scsi_nexus\n");
1745 1746 1747
		return -ENOMEM;
	}
	/*
1748
	 *  Initialize the struct se_session pointer and setup tagpool
1749
	 *  for struct vhost_scsi_cmd descriptors
1750
	 */
1751
	tv_nexus->tvn_se_sess = transport_init_session_tags(
1752 1753
					VHOST_SCSI_DEFAULT_TAGS,
					sizeof(struct vhost_scsi_cmd),
1754
					TARGET_PROT_DIN_PASS | TARGET_PROT_DOUT_PASS);
1755
	if (IS_ERR(tv_nexus->tvn_se_sess)) {
1756
		mutex_unlock(&tpg->tv_tpg_mutex);
1757 1758 1759
		kfree(tv_nexus);
		return -ENOMEM;
	}
1760
	se_sess = tv_nexus->tvn_se_sess;
1761 1762
	for (i = 0; i < VHOST_SCSI_DEFAULT_TAGS; i++) {
		tv_cmd = &((struct vhost_scsi_cmd *)se_sess->sess_cmd_map)[i];
1763 1764

		tv_cmd->tvc_sgl = kzalloc(sizeof(struct scatterlist) *
1765
					VHOST_SCSI_PREALLOC_SGLS, GFP_KERNEL);
1766 1767 1768 1769 1770 1771 1772
		if (!tv_cmd->tvc_sgl) {
			mutex_unlock(&tpg->tv_tpg_mutex);
			pr_err("Unable to allocate tv_cmd->tvc_sgl\n");
			goto out;
		}

		tv_cmd->tvc_upages = kzalloc(sizeof(struct page *) *
1773
					VHOST_SCSI_PREALLOC_UPAGES, GFP_KERNEL);
1774 1775 1776 1777 1778
		if (!tv_cmd->tvc_upages) {
			mutex_unlock(&tpg->tv_tpg_mutex);
			pr_err("Unable to allocate tv_cmd->tvc_upages\n");
			goto out;
		}
1779 1780

		tv_cmd->tvc_prot_sgl = kzalloc(sizeof(struct scatterlist) *
1781
					VHOST_SCSI_PREALLOC_PROT_SGLS, GFP_KERNEL);
1782 1783 1784 1785 1786
		if (!tv_cmd->tvc_prot_sgl) {
			mutex_unlock(&tpg->tv_tpg_mutex);
			pr_err("Unable to allocate tv_cmd->tvc_prot_sgl\n");
			goto out;
		}
1787
	}
1788 1789
	/*
	 * Since we are running in 'demo mode' this call with generate a
1790
	 * struct se_node_acl for the vhost_scsi struct se_portal_group with
1791 1792 1793 1794 1795
	 * the SCSI Initiator port name of the passed configfs group 'name'.
	 */
	tv_nexus->tvn_se_sess->se_node_acl = core_tpg_check_initiator_node_acl(
				se_tpg, (unsigned char *)name);
	if (!tv_nexus->tvn_se_sess->se_node_acl) {
1796
		mutex_unlock(&tpg->tv_tpg_mutex);
1797 1798
		pr_debug("core_tpg_check_initiator_node_acl() failed"
				" for %s\n", name);
1799
		goto out;
1800 1801
	}
	/*
1802
	 * Now register the TCM vhost virtual I_T Nexus as active.
1803
	 */
1804
	transport_register_session(se_tpg, tv_nexus->tvn_se_sess->se_node_acl,
1805
			tv_nexus->tvn_se_sess, tv_nexus);
1806
	tpg->tpg_nexus = tv_nexus;
1807

1808
	mutex_unlock(&tpg->tv_tpg_mutex);
1809
	return 0;
1810 1811

out:
1812
	vhost_scsi_free_cmd_map_res(tv_nexus, se_sess);
1813 1814 1815
	transport_free_session(se_sess);
	kfree(tv_nexus);
	return -ENOMEM;
1816 1817
}

1818
static int vhost_scsi_drop_nexus(struct vhost_scsi_tpg *tpg)
1819 1820
{
	struct se_session *se_sess;
1821
	struct vhost_scsi_nexus *tv_nexus;
1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835

	mutex_lock(&tpg->tv_tpg_mutex);
	tv_nexus = tpg->tpg_nexus;
	if (!tv_nexus) {
		mutex_unlock(&tpg->tv_tpg_mutex);
		return -ENODEV;
	}

	se_sess = tv_nexus->tvn_se_sess;
	if (!se_sess) {
		mutex_unlock(&tpg->tv_tpg_mutex);
		return -ENODEV;
	}

1836
	if (tpg->tv_tpg_port_count != 0) {
1837
		mutex_unlock(&tpg->tv_tpg_mutex);
1838
		pr_err("Unable to remove TCM_vhost I_T Nexus with"
1839
			" active TPG port count: %d\n",
1840 1841
			tpg->tv_tpg_port_count);
		return -EBUSY;
1842 1843
	}

1844
	if (tpg->tv_tpg_vhost_count != 0) {
1845
		mutex_unlock(&tpg->tv_tpg_mutex);
1846
		pr_err("Unable to remove TCM_vhost I_T Nexus with"
1847
			" active TPG vhost count: %d\n",
1848 1849
			tpg->tv_tpg_vhost_count);
		return -EBUSY;
1850 1851
	}

1852
	pr_debug("TCM_vhost_ConfigFS: Removing I_T Nexus to emulated"
1853
		" %s Initiator Port: %s\n", vhost_scsi_dump_proto_id(tpg->tport),
1854
		tv_nexus->tvn_se_sess->se_node_acl->initiatorname);
1855

1856
	vhost_scsi_free_cmd_map_res(tv_nexus, se_sess);
1857
	/*
1858
	 * Release the SCSI I_T Nexus to the emulated vhost Target Port
1859 1860 1861 1862 1863 1864 1865 1866 1867
	 */
	transport_deregister_session(tv_nexus->tvn_se_sess);
	tpg->tpg_nexus = NULL;
	mutex_unlock(&tpg->tv_tpg_mutex);

	kfree(tv_nexus);
	return 0;
}

1868
static ssize_t vhost_scsi_tpg_nexus_show(struct config_item *item, char *page)
1869
{
1870
	struct se_portal_group *se_tpg = to_tpg(item);
1871 1872 1873
	struct vhost_scsi_tpg *tpg = container_of(se_tpg,
				struct vhost_scsi_tpg, se_tpg);
	struct vhost_scsi_nexus *tv_nexus;
1874 1875
	ssize_t ret;

1876 1877
	mutex_lock(&tpg->tv_tpg_mutex);
	tv_nexus = tpg->tpg_nexus;
1878
	if (!tv_nexus) {
1879
		mutex_unlock(&tpg->tv_tpg_mutex);
1880 1881 1882 1883
		return -ENODEV;
	}
	ret = snprintf(page, PAGE_SIZE, "%s\n",
			tv_nexus->tvn_se_sess->se_node_acl->initiatorname);
1884
	mutex_unlock(&tpg->tv_tpg_mutex);
1885 1886 1887 1888

	return ret;
}

1889 1890
static ssize_t vhost_scsi_tpg_nexus_store(struct config_item *item,
		const char *page, size_t count)
1891
{
1892
	struct se_portal_group *se_tpg = to_tpg(item);
1893 1894 1895 1896
	struct vhost_scsi_tpg *tpg = container_of(se_tpg,
				struct vhost_scsi_tpg, se_tpg);
	struct vhost_scsi_tport *tport_wwn = tpg->tport;
	unsigned char i_port[VHOST_SCSI_NAMELEN], *ptr, *port_ptr;
1897 1898 1899 1900 1901
	int ret;
	/*
	 * Shutdown the active I_T nexus if 'NULL' is passed..
	 */
	if (!strncmp(page, "NULL", 4)) {
1902
		ret = vhost_scsi_drop_nexus(tpg);
1903 1904 1905 1906
		return (!ret) ? count : ret;
	}
	/*
	 * Otherwise make sure the passed virtual Initiator port WWN matches
1907 1908
	 * the fabric protocol_id set in vhost_scsi_make_tport(), and call
	 * vhost_scsi_make_nexus().
1909
	 */
1910
	if (strlen(page) >= VHOST_SCSI_NAMELEN) {
1911
		pr_err("Emulated NAA Sas Address: %s, exceeds"
1912
				" max: %d\n", page, VHOST_SCSI_NAMELEN);
1913 1914
		return -EINVAL;
	}
1915
	snprintf(&i_port[0], VHOST_SCSI_NAMELEN, "%s", page);
1916 1917 1918 1919 1920 1921

	ptr = strstr(i_port, "naa.");
	if (ptr) {
		if (tport_wwn->tport_proto_id != SCSI_PROTOCOL_SAS) {
			pr_err("Passed SAS Initiator Port %s does not"
				" match target port protoid: %s\n", i_port,
1922
				vhost_scsi_dump_proto_id(tport_wwn));
1923 1924 1925 1926 1927 1928 1929 1930 1931 1932
			return -EINVAL;
		}
		port_ptr = &i_port[0];
		goto check_newline;
	}
	ptr = strstr(i_port, "fc.");
	if (ptr) {
		if (tport_wwn->tport_proto_id != SCSI_PROTOCOL_FCP) {
			pr_err("Passed FCP Initiator Port %s does not"
				" match target port protoid: %s\n", i_port,
1933
				vhost_scsi_dump_proto_id(tport_wwn));
1934 1935 1936 1937 1938 1939 1940 1941 1942 1943
			return -EINVAL;
		}
		port_ptr = &i_port[3]; /* Skip over "fc." */
		goto check_newline;
	}
	ptr = strstr(i_port, "iqn.");
	if (ptr) {
		if (tport_wwn->tport_proto_id != SCSI_PROTOCOL_ISCSI) {
			pr_err("Passed iSCSI Initiator Port %s does not"
				" match target port protoid: %s\n", i_port,
1944
				vhost_scsi_dump_proto_id(tport_wwn));
1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959
			return -EINVAL;
		}
		port_ptr = &i_port[0];
		goto check_newline;
	}
	pr_err("Unable to locate prefix for emulated Initiator Port:"
			" %s\n", i_port);
	return -EINVAL;
	/*
	 * Clear any trailing newline for the NAA WWN
	 */
check_newline:
	if (i_port[strlen(i_port)-1] == '\n')
		i_port[strlen(i_port)-1] = '\0';

1960
	ret = vhost_scsi_make_nexus(tpg, port_ptr);
1961 1962 1963 1964 1965 1966
	if (ret < 0)
		return ret;

	return count;
}

1967
CONFIGFS_ATTR(vhost_scsi_tpg_, nexus);
1968

1969
static struct configfs_attribute *vhost_scsi_tpg_attrs[] = {
1970
	&vhost_scsi_tpg_attr_nexus,
1971 1972 1973
	NULL,
};

1974
static struct se_portal_group *
1975
vhost_scsi_make_tpg(struct se_wwn *wwn,
1976 1977
		   struct config_group *group,
		   const char *name)
1978
{
1979 1980
	struct vhost_scsi_tport *tport = container_of(wwn,
			struct vhost_scsi_tport, tport_wwn);
1981

1982
	struct vhost_scsi_tpg *tpg;
1983
	u16 tpgt;
1984 1985 1986 1987
	int ret;

	if (strstr(name, "tpgt_") != name)
		return ERR_PTR(-EINVAL);
1988
	if (kstrtou16(name + 5, 10, &tpgt) || tpgt >= VHOST_SCSI_MAX_TARGET)
1989 1990
		return ERR_PTR(-EINVAL);

1991
	tpg = kzalloc(sizeof(struct vhost_scsi_tpg), GFP_KERNEL);
1992
	if (!tpg) {
1993
		pr_err("Unable to allocate struct vhost_scsi_tpg");
1994 1995 1996 1997 1998 1999 2000
		return ERR_PTR(-ENOMEM);
	}
	mutex_init(&tpg->tv_tpg_mutex);
	INIT_LIST_HEAD(&tpg->tv_tpg_list);
	tpg->tport = tport;
	tpg->tport_tpgt = tpgt;

2001
	ret = core_tpg_register(wwn, &tpg->se_tpg, tport->tport_proto_id);
2002 2003 2004 2005
	if (ret < 0) {
		kfree(tpg);
		return NULL;
	}
2006 2007 2008
	mutex_lock(&vhost_scsi_mutex);
	list_add_tail(&tpg->tv_tpg_list, &vhost_scsi_list);
	mutex_unlock(&vhost_scsi_mutex);
2009 2010 2011 2012

	return &tpg->se_tpg;
}

2013
static void vhost_scsi_drop_tpg(struct se_portal_group *se_tpg)
2014
{
2015 2016
	struct vhost_scsi_tpg *tpg = container_of(se_tpg,
				struct vhost_scsi_tpg, se_tpg);
2017

2018
	mutex_lock(&vhost_scsi_mutex);
2019
	list_del(&tpg->tv_tpg_list);
2020
	mutex_unlock(&vhost_scsi_mutex);
2021
	/*
2022
	 * Release the virtual I_T Nexus for this vhost TPG
2023
	 */
2024
	vhost_scsi_drop_nexus(tpg);
2025 2026 2027 2028 2029 2030 2031
	/*
	 * Deregister the se_tpg from TCM..
	 */
	core_tpg_deregister(se_tpg);
	kfree(tpg);
}

2032
static struct se_wwn *
2033
vhost_scsi_make_tport(struct target_fabric_configfs *tf,
2034 2035
		     struct config_group *group,
		     const char *name)
2036
{
2037
	struct vhost_scsi_tport *tport;
2038 2039 2040 2041
	char *ptr;
	u64 wwpn = 0;
	int off = 0;

2042
	/* if (vhost_scsi_parse_wwn(name, &wwpn, 1) < 0)
2043 2044
		return ERR_PTR(-EINVAL); */

2045
	tport = kzalloc(sizeof(struct vhost_scsi_tport), GFP_KERNEL);
2046
	if (!tport) {
2047
		pr_err("Unable to allocate struct vhost_scsi_tport");
2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077
		return ERR_PTR(-ENOMEM);
	}
	tport->tport_wwpn = wwpn;
	/*
	 * Determine the emulated Protocol Identifier and Target Port Name
	 * based on the incoming configfs directory name.
	 */
	ptr = strstr(name, "naa.");
	if (ptr) {
		tport->tport_proto_id = SCSI_PROTOCOL_SAS;
		goto check_len;
	}
	ptr = strstr(name, "fc.");
	if (ptr) {
		tport->tport_proto_id = SCSI_PROTOCOL_FCP;
		off = 3; /* Skip over "fc." */
		goto check_len;
	}
	ptr = strstr(name, "iqn.");
	if (ptr) {
		tport->tport_proto_id = SCSI_PROTOCOL_ISCSI;
		goto check_len;
	}

	pr_err("Unable to locate prefix for emulated Target Port:"
			" %s\n", name);
	kfree(tport);
	return ERR_PTR(-EINVAL);

check_len:
2078
	if (strlen(name) >= VHOST_SCSI_NAMELEN) {
2079
		pr_err("Emulated %s Address: %s, exceeds"
2080 2081
			" max: %d\n", name, vhost_scsi_dump_proto_id(tport),
			VHOST_SCSI_NAMELEN);
2082 2083 2084
		kfree(tport);
		return ERR_PTR(-EINVAL);
	}
2085
	snprintf(&tport->tport_name[0], VHOST_SCSI_NAMELEN, "%s", &name[off]);
2086 2087

	pr_debug("TCM_VHost_ConfigFS: Allocated emulated Target"
2088
		" %s Address: %s\n", vhost_scsi_dump_proto_id(tport), name);
2089 2090 2091 2092

	return &tport->tport_wwn;
}

2093
static void vhost_scsi_drop_tport(struct se_wwn *wwn)
2094
{
2095 2096
	struct vhost_scsi_tport *tport = container_of(wwn,
				struct vhost_scsi_tport, tport_wwn);
2097 2098

	pr_debug("TCM_VHost_ConfigFS: Deallocating emulated Target"
2099
		" %s Address: %s\n", vhost_scsi_dump_proto_id(tport),
2100 2101 2102 2103 2104
		tport->tport_name);

	kfree(tport);
}

2105
static ssize_t
2106
vhost_scsi_wwn_version_show(struct config_item *item, char *page)
2107 2108
{
	return sprintf(page, "TCM_VHOST fabric module %s on %s/%s"
2109
		"on "UTS_RELEASE"\n", VHOST_SCSI_VERSION, utsname()->sysname,
2110 2111 2112
		utsname()->machine);
}

2113
CONFIGFS_ATTR_RO(vhost_scsi_wwn_, version);
2114

2115
static struct configfs_attribute *vhost_scsi_wwn_attrs[] = {
2116
	&vhost_scsi_wwn_attr_version,
2117 2118 2119
	NULL,
};

2120
static struct target_core_fabric_ops vhost_scsi_ops = {
2121 2122
	.module				= THIS_MODULE,
	.name				= "vhost",
2123 2124 2125 2126 2127 2128 2129
	.get_fabric_name		= vhost_scsi_get_fabric_name,
	.tpg_get_wwn			= vhost_scsi_get_fabric_wwn,
	.tpg_get_tag			= vhost_scsi_get_tpgt,
	.tpg_check_demo_mode		= vhost_scsi_check_true,
	.tpg_check_demo_mode_cache	= vhost_scsi_check_true,
	.tpg_check_demo_mode_write_protect = vhost_scsi_check_false,
	.tpg_check_prod_mode_write_protect = vhost_scsi_check_false,
2130
	.tpg_check_prot_fabric_only	= vhost_scsi_check_prot_fabric_only,
2131 2132
	.tpg_get_inst_index		= vhost_scsi_tpg_get_inst_index,
	.release_cmd			= vhost_scsi_release_cmd,
2133
	.check_stop_free		= vhost_scsi_check_stop_free,
2134 2135 2136
	.shutdown_session		= vhost_scsi_shutdown_session,
	.close_session			= vhost_scsi_close_session,
	.sess_get_index			= vhost_scsi_sess_get_index,
2137
	.sess_get_initiator_sid		= NULL,
2138 2139 2140 2141 2142 2143 2144 2145
	.write_pending			= vhost_scsi_write_pending,
	.write_pending_status		= vhost_scsi_write_pending_status,
	.set_default_node_attributes	= vhost_scsi_set_default_node_attrs,
	.get_cmd_state			= vhost_scsi_get_cmd_state,
	.queue_data_in			= vhost_scsi_queue_data_in,
	.queue_status			= vhost_scsi_queue_status,
	.queue_tm_rsp			= vhost_scsi_queue_tm_rsp,
	.aborted_task			= vhost_scsi_aborted_task,
2146 2147 2148
	/*
	 * Setup callers for generic logic in target_core_fabric_configfs.c
	 */
2149 2150 2151 2152 2153 2154
	.fabric_make_wwn		= vhost_scsi_make_tport,
	.fabric_drop_wwn		= vhost_scsi_drop_tport,
	.fabric_make_tpg		= vhost_scsi_make_tpg,
	.fabric_drop_tpg		= vhost_scsi_drop_tpg,
	.fabric_post_link		= vhost_scsi_port_link,
	.fabric_pre_unlink		= vhost_scsi_port_unlink,
2155 2156 2157 2158

	.tfc_wwn_attrs			= vhost_scsi_wwn_attrs,
	.tfc_tpg_base_attrs		= vhost_scsi_tpg_attrs,
	.tfc_tpg_attrib_attrs		= vhost_scsi_tpg_attrib_attrs,
2159 2160
};

2161
static int __init vhost_scsi_init(void)
2162
{
2163
	int ret = -ENOMEM;
2164

2165
	pr_debug("TCM_VHOST fabric module %s on %s/%s"
2166
		" on "UTS_RELEASE"\n", VHOST_SCSI_VERSION, utsname()->sysname,
2167 2168
		utsname()->machine);

2169 2170 2171 2172
	/*
	 * Use our own dedicated workqueue for submitting I/O into
	 * target core to avoid contention within system_wq.
	 */
2173 2174
	vhost_scsi_workqueue = alloc_workqueue("vhost_scsi", 0, 0);
	if (!vhost_scsi_workqueue)
2175 2176 2177 2178 2179 2180
		goto out;

	ret = vhost_scsi_register();
	if (ret < 0)
		goto out_destroy_workqueue;

2181
	ret = target_register_template(&vhost_scsi_ops);
2182 2183 2184 2185 2186 2187 2188 2189
	if (ret < 0)
		goto out_vhost_scsi_deregister;

	return 0;

out_vhost_scsi_deregister:
	vhost_scsi_deregister();
out_destroy_workqueue:
2190
	destroy_workqueue(vhost_scsi_workqueue);
2191 2192 2193 2194
out:
	return ret;
};

2195
static void vhost_scsi_exit(void)
2196
{
2197
	target_unregister_template(&vhost_scsi_ops);
2198
	vhost_scsi_deregister();
2199
	destroy_workqueue(vhost_scsi_workqueue);
2200 2201
};

M
Michael S. Tsirkin 已提交
2202 2203
MODULE_DESCRIPTION("VHOST_SCSI series fabric driver");
MODULE_ALIAS("tcm_vhost");
2204
MODULE_LICENSE("GPL");
2205 2206
module_init(vhost_scsi_init);
module_exit(vhost_scsi_exit);