net.c 30.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14
/* Copyright (C) 2009 Red Hat, Inc.
 * Author: Michael S. Tsirkin <mst@redhat.com>
 *
 * This work is licensed under the terms of the GNU GPL, version 2.
 *
 * virtio-net server in host kernel.
 */

#include <linux/compat.h>
#include <linux/eventfd.h>
#include <linux/vhost.h>
#include <linux/virtio_net.h>
#include <linux/miscdevice.h>
#include <linux/module.h>
15
#include <linux/moduleparam.h>
16 17 18
#include <linux/mutex.h>
#include <linux/workqueue.h>
#include <linux/file.h>
19
#include <linux/slab.h>
20
#include <linux/vmalloc.h>
21 22 23 24 25

#include <linux/net.h>
#include <linux/if_packet.h>
#include <linux/if_arp.h>
#include <linux/if_tun.h>
26
#include <linux/if_macvlan.h>
27
#include <linux/if_vlan.h>
28 29 30 31 32

#include <net/sock.h>

#include "vhost.h"

33
static int experimental_zcopytx = 1;
34
module_param(experimental_zcopytx, int, 0444);
35 36
MODULE_PARM_DESC(experimental_zcopytx, "Enable Zero Copy TX;"
		                       " 1 -Enable; 0 - Disable");
37

38 39 40 41
/* Max number of bytes transferred before requeueing the job.
 * Using this limit prevents one virtqueue from starving others. */
#define VHOST_NET_WEIGHT 0x80000

42 43 44 45
/* MAX number of TX used buffers for outstanding zerocopy */
#define VHOST_MAX_PEND 128
#define VHOST_GOODCOPY_LEN 256

46 47 48 49 50
/*
 * For transmit, used buffer len is unused; we override it to track buffer
 * status internally; used for zerocopy tx only.
 */
/* Lower device DMA failed */
51
#define VHOST_DMA_FAILED_LEN	((__force __virtio32)3)
52
/* Lower device DMA done */
53
#define VHOST_DMA_DONE_LEN	((__force __virtio32)2)
54
/* Lower device DMA in progress */
55
#define VHOST_DMA_IN_PROGRESS	((__force __virtio32)1)
56
/* Buffer unused */
57
#define VHOST_DMA_CLEAR_LEN	((__force __virtio32)0)
58

59
#define VHOST_DMA_IS_DONE(len) ((__force u32)(len) >= (__force u32)VHOST_DMA_DONE_LEN)
60

61 62 63
enum {
	VHOST_NET_FEATURES = VHOST_FEATURES |
			 (1ULL << VHOST_NET_F_VIRTIO_NET_HDR) |
Jason Wang's avatar
Jason Wang committed
64 65
			 (1ULL << VIRTIO_NET_F_MRG_RXBUF) |
			 (1ULL << VIRTIO_F_IOMMU_PLATFORM)
66 67
};

68 69 70 71 72 73
enum {
	VHOST_NET_VQ_RX = 0,
	VHOST_NET_VQ_TX = 1,
	VHOST_NET_VQ_MAX = 2,
};

74
struct vhost_net_ubuf_ref {
75 76 77 78 79 80
	/* refcount follows semantics similar to kref:
	 *  0: object is released
	 *  1: no outstanding ubufs
	 * >1: outstanding ubufs
	 */
	atomic_t refcount;
81 82 83 84
	wait_queue_head_t wait;
	struct vhost_virtqueue *vq;
};

85 86
struct vhost_net_virtqueue {
	struct vhost_virtqueue vq;
87 88
	size_t vhost_hlen;
	size_t sock_hlen;
89 90 91 92 93 94 95 96 97
	/* vhost zerocopy support fields below: */
	/* last used idx for outstanding DMA zerocopy buffers */
	int upend_idx;
	/* first used idx for DMA done zerocopy buffers */
	int done_idx;
	/* an array of userspace buffers info */
	struct ubuf_info *ubuf_info;
	/* Reference counting for outstanding ubufs.
	 * Protected by vq mutex. Writers must also take device mutex. */
98
	struct vhost_net_ubuf_ref *ubufs;
99 100
};

101 102
struct vhost_net {
	struct vhost_dev dev;
103
	struct vhost_net_virtqueue vqs[VHOST_NET_VQ_MAX];
104
	struct vhost_poll poll[VHOST_NET_VQ_MAX];
105 106 107 108 109 110
	/* Number of TX recently submitted.
	 * Protected by tx vq lock. */
	unsigned tx_packets;
	/* Number of times zerocopy TX recently failed.
	 * Protected by tx vq lock. */
	unsigned tx_zcopy_err;
111 112
	/* Flush in progress. Protected by tx vq lock. */
	bool tx_flush;
113 114
};

115
static unsigned vhost_net_zcopy_mask __read_mostly;
116

117
static void vhost_net_enable_zcopy(int vq)
118
{
119
	vhost_net_zcopy_mask |= 0x1 << vq;
120 121
}

122 123
static struct vhost_net_ubuf_ref *
vhost_net_ubuf_alloc(struct vhost_virtqueue *vq, bool zcopy)
124
{
125
	struct vhost_net_ubuf_ref *ubufs;
126 127 128 129 130 131
	/* No zero copy backend? Nothing to count. */
	if (!zcopy)
		return NULL;
	ubufs = kmalloc(sizeof(*ubufs), GFP_KERNEL);
	if (!ubufs)
		return ERR_PTR(-ENOMEM);
132
	atomic_set(&ubufs->refcount, 1);
133 134 135 136 137
	init_waitqueue_head(&ubufs->wait);
	ubufs->vq = vq;
	return ubufs;
}

138
static int vhost_net_ubuf_put(struct vhost_net_ubuf_ref *ubufs)
139
{
140 141 142 143
	int r = atomic_sub_return(1, &ubufs->refcount);
	if (unlikely(!r))
		wake_up(&ubufs->wait);
	return r;
144 145
}

146
static void vhost_net_ubuf_put_and_wait(struct vhost_net_ubuf_ref *ubufs)
147
{
148 149
	vhost_net_ubuf_put(ubufs);
	wait_event(ubufs->wait, !atomic_read(&ubufs->refcount));
150 151 152 153 154
}

static void vhost_net_ubuf_put_wait_and_free(struct vhost_net_ubuf_ref *ubufs)
{
	vhost_net_ubuf_put_and_wait(ubufs);
155 156 157
	kfree(ubufs);
}

158 159 160 161
static void vhost_net_clear_ubuf_info(struct vhost_net *n)
{
	int i;

162 163 164
	for (i = 0; i < VHOST_NET_VQ_MAX; ++i) {
		kfree(n->vqs[i].ubuf_info);
		n->vqs[i].ubuf_info = NULL;
165 166 167
	}
}

Asias He's avatar
Asias He committed
168
static int vhost_net_set_ubuf_info(struct vhost_net *n)
169 170 171 172
{
	bool zcopy;
	int i;

173
	for (i = 0; i < VHOST_NET_VQ_MAX; ++i) {
174
		zcopy = vhost_net_zcopy_mask & (0x1 << i);
175 176 177 178 179 180 181 182 183 184
		if (!zcopy)
			continue;
		n->vqs[i].ubuf_info = kmalloc(sizeof(*n->vqs[i].ubuf_info) *
					      UIO_MAXIOV, GFP_KERNEL);
		if  (!n->vqs[i].ubuf_info)
			goto err;
	}
	return 0;

err:
185
	vhost_net_clear_ubuf_info(n);
186 187 188
	return -ENOMEM;
}

Asias He's avatar
Asias He committed
189
static void vhost_net_vq_reset(struct vhost_net *n)
190 191 192
{
	int i;

193 194
	vhost_net_clear_ubuf_info(n);

195 196 197 198
	for (i = 0; i < VHOST_NET_VQ_MAX; i++) {
		n->vqs[i].done_idx = 0;
		n->vqs[i].upend_idx = 0;
		n->vqs[i].ubufs = NULL;
199 200
		n->vqs[i].vhost_hlen = 0;
		n->vqs[i].sock_hlen = 0;
201 202 203 204
	}

}

205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220
static void vhost_net_tx_packet(struct vhost_net *net)
{
	++net->tx_packets;
	if (net->tx_packets < 1024)
		return;
	net->tx_packets = 0;
	net->tx_zcopy_err = 0;
}

static void vhost_net_tx_err(struct vhost_net *net)
{
	++net->tx_zcopy_err;
}

static bool vhost_net_tx_select_zcopy(struct vhost_net *net)
{
221 222 223 224 225
	/* TX flush waits for outstanding DMAs to be done.
	 * Don't start new DMAs.
	 */
	return !net->tx_flush &&
		net->tx_packets / 64 >= net->tx_zcopy_err;
226 227
}

228 229 230 231 232 233
static bool vhost_sock_zcopy(struct socket *sock)
{
	return unlikely(experimental_zcopytx) &&
		sock_flag(sock->sk, SOCK_ZEROCOPY);
}

234 235 236 237 238
/* In case of DMA done not in order in lower device driver for some reason.
 * upend_idx is used to track end of used idx, done_idx is used to track head
 * of used idx. Once lower device DMA done contiguously, we will signal KVM
 * guest used idx.
 */
239 240
static void vhost_zerocopy_signal_used(struct vhost_net *net,
				       struct vhost_virtqueue *vq)
241
{
242 243
	struct vhost_net_virtqueue *nvq =
		container_of(vq, struct vhost_net_virtqueue, vq);
244
	int i, add;
245 246
	int j = 0;

247
	for (i = nvq->done_idx; i != nvq->upend_idx; i = (i + 1) % UIO_MAXIOV) {
248 249
		if (vq->heads[i].len == VHOST_DMA_FAILED_LEN)
			vhost_net_tx_err(net);
250 251 252 253 254 255
		if (VHOST_DMA_IS_DONE(vq->heads[i].len)) {
			vq->heads[i].len = VHOST_DMA_CLEAR_LEN;
			++j;
		} else
			break;
	}
256 257 258 259 260 261 262
	while (j) {
		add = min(UIO_MAXIOV - nvq->done_idx, j);
		vhost_add_used_and_signal_n(vq->dev, vq,
					    &vq->heads[nvq->done_idx], add);
		nvq->done_idx = (nvq->done_idx + add) % UIO_MAXIOV;
		j -= add;
	}
263 264
}

265
static void vhost_zerocopy_callback(struct ubuf_info *ubuf, bool success)
266
{
267
	struct vhost_net_ubuf_ref *ubufs = ubuf->ctx;
268
	struct vhost_virtqueue *vq = ubufs->vq;
269
	int cnt;
270

271 272
	rcu_read_lock_bh();

273 274 275
	/* set len to mark this desc buffers done DMA */
	vq->heads[ubuf->desc].len = success ?
		VHOST_DMA_DONE_LEN : VHOST_DMA_FAILED_LEN;
276
	cnt = vhost_net_ubuf_put(ubufs);
277

278 279
	/*
	 * Trigger polling thread if guest stopped submitting new buffers:
280
	 * in this case, the refcount after decrement will eventually reach 1.
281 282 283 284
	 * We also trigger polling periodically after each 16 packets
	 * (the value 16 here is more or less arbitrary, it's tuned to trigger
	 * less than 10% of times).
	 */
285
	if (cnt <= 1 || !(cnt % 16))
286
		vhost_poll_queue(&vq->poll);
287 288

	rcu_read_unlock_bh();
289 290
}

291 292 293 294 295 296 297 298 299 300 301 302 303 304
static inline unsigned long busy_clock(void)
{
	return local_clock() >> 10;
}

static bool vhost_can_busy_poll(struct vhost_dev *dev,
				unsigned long endtime)
{
	return likely(!need_resched()) &&
	       likely(!time_after(busy_clock(), endtime)) &&
	       likely(!signal_pending(current)) &&
	       !vhost_has_work(dev);
}

305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330
static void vhost_net_disable_vq(struct vhost_net *n,
				 struct vhost_virtqueue *vq)
{
	struct vhost_net_virtqueue *nvq =
		container_of(vq, struct vhost_net_virtqueue, vq);
	struct vhost_poll *poll = n->poll + (nvq - n->vqs);
	if (!vq->private_data)
		return;
	vhost_poll_stop(poll);
}

static int vhost_net_enable_vq(struct vhost_net *n,
				struct vhost_virtqueue *vq)
{
	struct vhost_net_virtqueue *nvq =
		container_of(vq, struct vhost_net_virtqueue, vq);
	struct vhost_poll *poll = n->poll + (nvq - n->vqs);
	struct socket *sock;

	sock = vq->private_data;
	if (!sock)
		return 0;

	return vhost_poll_start(poll, sock->file);
}

331 332 333 334 335 336 337
static int vhost_net_tx_get_vq_desc(struct vhost_net *net,
				    struct vhost_virtqueue *vq,
				    struct iovec iov[], unsigned int iov_size,
				    unsigned int *out_num, unsigned int *in_num)
{
	unsigned long uninitialized_var(endtime);
	int r = vhost_get_vq_desc(vq, vq->iov, ARRAY_SIZE(vq->iov),
Jason Wang's avatar
Jason Wang committed
338
				  out_num, in_num, NULL, NULL);
339 340 341 342 343 344 345 346 347

	if (r == vq->num && vq->busyloop_timeout) {
		preempt_disable();
		endtime = busy_clock() + vq->busyloop_timeout;
		while (vhost_can_busy_poll(vq->dev, endtime) &&
		       vhost_vq_avail_empty(vq->dev, vq))
			cpu_relax_lowlatency();
		preempt_enable();
		r = vhost_get_vq_desc(vq, vq->iov, ARRAY_SIZE(vq->iov),
Jason Wang's avatar
Jason Wang committed
348
				      out_num, in_num, NULL, NULL);
349 350 351 352 353
	}

	return r;
}

354 355 356 357
/* Expects to be always run from workqueue - which acts as
 * read-size critical section for our kind of RCU. */
static void handle_tx(struct vhost_net *net)
{
358
	struct vhost_net_virtqueue *nvq = &net->vqs[VHOST_NET_VQ_TX];
359
	struct vhost_virtqueue *vq = &nvq->vq;
360
	unsigned out, in;
361
	int head;
362 363 364 365 366 367 368 369
	struct msghdr msg = {
		.msg_name = NULL,
		.msg_namelen = 0,
		.msg_control = NULL,
		.msg_controllen = 0,
		.msg_flags = MSG_DONTWAIT,
	};
	size_t len, total_len = 0;
370
	int err;
371
	size_t hdr_size;
Arnd Bergmann's avatar
Arnd Bergmann committed
372
	struct socket *sock;
373
	struct vhost_net_ubuf_ref *uninitialized_var(ubufs);
374
	bool zcopy, zcopy_used;
Arnd Bergmann's avatar
Arnd Bergmann committed
375

376 377
	mutex_lock(&vq->mutex);
	sock = vq->private_data;
378
	if (!sock)
379
		goto out;
380

Jason Wang's avatar
Jason Wang committed
381 382 383
	if (!vq_iotlb_prefetch(vq))
		goto out;

384
	vhost_disable_notify(&net->dev, vq);
385

386
	hdr_size = nvq->vhost_hlen;
387
	zcopy = nvq->ubufs;
388 389

	for (;;) {
390 391
		/* Release DMAs done buffers first */
		if (zcopy)
392
			vhost_zerocopy_signal_used(net, vq);
393

394 395 396 397 398 399 400
		/* If more outstanding DMAs, queue the work.
		 * Handle upend_idx wrap around
		 */
		if (unlikely((nvq->upend_idx + vq->num - VHOST_MAX_PEND)
			      % UIO_MAXIOV == nvq->done_idx))
			break;

401 402 403
		head = vhost_net_tx_get_vq_desc(net, vq, vq->iov,
						ARRAY_SIZE(vq->iov),
						&out, &in);
404
		/* On error, stop handling until the next kick. */
405
		if (unlikely(head < 0))
406
			break;
407 408
		/* Nothing new?  Wait for eventfd to tell us they refilled. */
		if (head == vq->num) {
409 410
			if (unlikely(vhost_enable_notify(&net->dev, vq))) {
				vhost_disable_notify(&net->dev, vq);
411 412 413 414 415 416 417 418 419 420 421
				continue;
			}
			break;
		}
		if (in) {
			vq_err(vq, "Unexpected descriptor format for TX: "
			       "out %d, int %d\n", out, in);
			break;
		}
		/* Skip header. TODO: support TSO. */
		len = iov_length(vq->iov, out);
Al Viro's avatar
Al Viro committed
422
		iov_iter_init(&msg.msg_iter, WRITE, vq->iov, out, len);
423
		iov_iter_advance(&msg.msg_iter, hdr_size);
424
		/* Sanity check */
Al Viro's avatar
Al Viro committed
425
		if (!msg_data_left(&msg)) {
426 427
			vq_err(vq, "Unexpected header len for TX: "
			       "%zd expected %zd\n",
428
			       len, hdr_size);
429 430
			break;
		}
Al Viro's avatar
Al Viro committed
431
		len = msg_data_left(&msg);
432 433 434 435 436

		zcopy_used = zcopy && len >= VHOST_GOODCOPY_LEN
				   && (nvq->upend_idx + 1) % UIO_MAXIOV !=
				      nvq->done_idx
				   && vhost_net_tx_select_zcopy(net);
437

438
		/* use msg_control to pass vhost zerocopy ubuf info to skb */
439
		if (zcopy_used) {
440 441 442
			struct ubuf_info *ubuf;
			ubuf = nvq->ubuf_info + nvq->upend_idx;

443
			vq->heads[nvq->upend_idx].id = cpu_to_vhost32(vq, head);
444 445 446 447 448 449 450
			vq->heads[nvq->upend_idx].len = VHOST_DMA_IN_PROGRESS;
			ubuf->callback = vhost_zerocopy_callback;
			ubuf->ctx = nvq->ubufs;
			ubuf->desc = nvq->upend_idx;
			msg.msg_control = ubuf;
			msg.msg_controllen = sizeof(ubuf);
			ubufs = nvq->ubufs;
451
			atomic_inc(&ubufs->refcount);
452
			nvq->upend_idx = (nvq->upend_idx + 1) % UIO_MAXIOV;
453
		} else {
454
			msg.msg_control = NULL;
455 456
			ubufs = NULL;
		}
457
		/* TODO: Check specific error and bomb out unless ENOBUFS? */
458
		err = sock->ops->sendmsg(sock, &msg, len);
459
		if (unlikely(err < 0)) {
460
			if (zcopy_used) {
461
				vhost_net_ubuf_put(ubufs);
462 463
				nvq->upend_idx = ((unsigned)nvq->upend_idx - 1)
					% UIO_MAXIOV;
464
			}
465
			vhost_discard_vq_desc(vq, 1);
466 467 468
			break;
		}
		if (err != len)
469 470
			pr_debug("Truncated TX packet: "
				 " len %d != %zd\n", err, len);
471
		if (!zcopy_used)
472
			vhost_add_used_and_signal(&net->dev, vq, head, 0);
473
		else
474
			vhost_zerocopy_signal_used(net, vq);
475
		total_len += len;
476
		vhost_net_tx_packet(net);
477 478 479 480 481
		if (unlikely(total_len >= VHOST_NET_WEIGHT)) {
			vhost_poll_queue(&vq->poll);
			break;
		}
	}
482
out:
483 484 485
	mutex_unlock(&vq->mutex);
}

486 487
static int peek_head_len(struct sock *sk)
{
488
	struct socket *sock = sk->sk_socket;
489 490
	struct sk_buff *head;
	int len = 0;
491
	unsigned long flags;
492

493 494 495
	if (sock->ops->peek_len)
		return sock->ops->peek_len(sock);

496
	spin_lock_irqsave(&sk->sk_receive_queue.lock, flags);
497
	head = skb_peek(&sk->sk_receive_queue);
498
	if (likely(head)) {
499
		len = head->len;
500
		if (skb_vlan_tag_present(head))
501 502 503
			len += VLAN_HLEN;
	}

504
	spin_unlock_irqrestore(&sk->sk_receive_queue.lock, flags);
505 506 507
	return len;
}

508 509 510 511 512 513 514 515 516 517
static int sk_has_rx_data(struct sock *sk)
{
	struct socket *sock = sk->sk_socket;

	if (sock->ops->peek_len)
		return sock->ops->peek_len(sock);

	return skb_queue_empty(&sk->sk_receive_queue);
}

518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533
static int vhost_net_rx_peek_head_len(struct vhost_net *net, struct sock *sk)
{
	struct vhost_net_virtqueue *nvq = &net->vqs[VHOST_NET_VQ_TX];
	struct vhost_virtqueue *vq = &nvq->vq;
	unsigned long uninitialized_var(endtime);
	int len = peek_head_len(sk);

	if (!len && vq->busyloop_timeout) {
		/* Both tx vq and rx socket were polled here */
		mutex_lock(&vq->mutex);
		vhost_disable_notify(&net->dev, vq);

		preempt_disable();
		endtime = busy_clock() + vq->busyloop_timeout;

		while (vhost_can_busy_poll(&net->dev, endtime) &&
534
		       !sk_has_rx_data(sk) &&
535 536 537 538 539 540 541 542 543 544 545 546 547 548 549
		       vhost_vq_avail_empty(&net->dev, vq))
			cpu_relax_lowlatency();

		preempt_enable();

		if (vhost_enable_notify(&net->dev, vq))
			vhost_poll_queue(&vq->poll);
		mutex_unlock(&vq->mutex);

		len = peek_head_len(sk);
	}

	return len;
}

550 551 552 553 554 555 556
/* This is a multi-buffer version of vhost_get_desc, that works if
 *	vq has read descriptors only.
 * @vq		- the relevant virtqueue
 * @datalen	- data length we'll be reading
 * @iovcount	- returned count of io vectors we fill
 * @log		- vhost log
 * @log_num	- log offset
557
 * @quota       - headcount quota, 1 for big buffer
558 559 560 561 562 563 564
 *	returns number of buffer heads allocated, negative on error
 */
static int get_rx_bufs(struct vhost_virtqueue *vq,
		       struct vring_used_elem *heads,
		       int datalen,
		       unsigned *iovcount,
		       struct vhost_log *log,
565 566
		       unsigned *log_num,
		       unsigned int quota)
567 568 569 570 571 572
{
	unsigned int out, in;
	int seg = 0;
	int headcount = 0;
	unsigned d;
	int r, nlogs = 0;
573 574 575 576
	/* len is always initialized before use since we are always called with
	 * datalen > 0.
	 */
	u32 uninitialized_var(len);
577

578
	while (datalen > 0 && headcount < quota) {
Jason Wang's avatar
Jason Wang committed
579
		if (unlikely(seg >= UIO_MAXIOV)) {
580 581 582
			r = -ENOBUFS;
			goto err;
		}
583
		r = vhost_get_vq_desc(vq, vq->iov + seg,
584 585
				      ARRAY_SIZE(vq->iov) - seg, &out,
				      &in, log, log_num);
586 587 588 589
		if (unlikely(r < 0))
			goto err;

		d = r;
590 591 592 593 594 595 596 597 598 599 600 601 602 603
		if (d == vq->num) {
			r = 0;
			goto err;
		}
		if (unlikely(out || in <= 0)) {
			vq_err(vq, "unexpected descriptor format for RX: "
				"out %d, in %d\n", out, in);
			r = -EINVAL;
			goto err;
		}
		if (unlikely(log)) {
			nlogs += *log_num;
			log += *log_num;
		}
604 605 606 607
		heads[headcount].id = cpu_to_vhost32(vq, d);
		len = iov_length(vq->iov + seg, in);
		heads[headcount].len = cpu_to_vhost32(vq, len);
		datalen -= len;
608 609 610
		++headcount;
		seg += in;
	}
611
	heads[headcount - 1].len = cpu_to_vhost32(vq, len + datalen);
612 613 614
	*iovcount = seg;
	if (unlikely(log))
		*log_num = nlogs;
615 616 617 618 619 620

	/* Detect overrun */
	if (unlikely(datalen > 0)) {
		r = UIO_MAXIOV + 1;
		goto err;
	}
621 622 623 624 625 626
	return headcount;
err:
	vhost_discard_vq_desc(vq, headcount);
	return r;
}

627 628
/* Expects to be always run from workqueue - which acts as
 * read-size critical section for our kind of RCU. */
629
static void handle_rx(struct vhost_net *net)
630
{
631 632
	struct vhost_net_virtqueue *nvq = &net->vqs[VHOST_NET_VQ_RX];
	struct vhost_virtqueue *vq = &nvq->vq;
633 634 635 636 637 638 639 640 641
	unsigned uninitialized_var(in), log;
	struct vhost_log *vq_log;
	struct msghdr msg = {
		.msg_name = NULL,
		.msg_namelen = 0,
		.msg_control = NULL, /* FIXME: get and handle RX aux data. */
		.msg_controllen = 0,
		.msg_flags = MSG_DONTWAIT,
	};
642 643 644
	struct virtio_net_hdr hdr = {
		.flags = 0,
		.gso_type = VIRTIO_NET_HDR_GSO_NONE
645 646
	};
	size_t total_len = 0;
647 648
	int err, mergeable;
	s16 headcount;
649 650
	size_t vhost_hlen, sock_hlen;
	size_t vhost_len, sock_len;
651
	struct socket *sock;
652
	struct iov_iter fixup;
653
	__virtio16 num_buffers;
654 655

	mutex_lock(&vq->mutex);
656 657 658
	sock = vq->private_data;
	if (!sock)
		goto out;
Jason Wang's avatar
Jason Wang committed
659 660 661 662

	if (!vq_iotlb_prefetch(vq))
		goto out;

663
	vhost_disable_notify(&net->dev, vq);
664
	vhost_net_disable_vq(net, vq);
665

666 667
	vhost_hlen = nvq->vhost_hlen;
	sock_hlen = nvq->sock_hlen;
668

669
	vq_log = unlikely(vhost_has_feature(vq, VHOST_F_LOG_ALL)) ?
670
		vq->log : NULL;
671
	mergeable = vhost_has_feature(vq, VIRTIO_NET_F_MRG_RXBUF);
672

673
	while ((sock_len = vhost_net_rx_peek_head_len(net, sock->sk))) {
674 675 676
		sock_len += sock_hlen;
		vhost_len = sock_len + vhost_hlen;
		headcount = get_rx_bufs(vq, vq->heads, vhost_len,
677 678
					&in, vq_log, &log,
					likely(mergeable) ? UIO_MAXIOV : 1);
679 680
		/* On error, stop handling until the next kick. */
		if (unlikely(headcount < 0))
681
			goto out;
682 683
		/* On overrun, truncate and discard */
		if (unlikely(headcount > UIO_MAXIOV)) {
Al Viro's avatar
Al Viro committed
684
			iov_iter_init(&msg.msg_iter, READ, vq->iov, 1, 1);
685
			err = sock->ops->recvmsg(sock, &msg,
686 687 688 689
						 1, MSG_DONTWAIT | MSG_TRUNC);
			pr_debug("Discarded rx packet: len %zd\n", sock_len);
			continue;
		}
690 691
		/* OK, now we need to know about added descriptors. */
		if (!headcount) {
692
			if (unlikely(vhost_enable_notify(&net->dev, vq))) {
693 694
				/* They have slipped one in as we were
				 * doing that: check again. */
695
				vhost_disable_notify(&net->dev, vq);
696 697 698 699
				continue;
			}
			/* Nothing new?  Wait for eventfd to tell us
			 * they refilled. */
700
			goto out;
701 702
		}
		/* We don't need to be notified again. */
703 704 705 706 707 708 709 710
		iov_iter_init(&msg.msg_iter, READ, vq->iov, in, vhost_len);
		fixup = msg.msg_iter;
		if (unlikely((vhost_hlen))) {
			/* We will supply the header ourselves
			 * TODO: support TSO.
			 */
			iov_iter_advance(&msg.msg_iter, vhost_hlen);
		}
711
		err = sock->ops->recvmsg(sock, &msg,
712 713 714 715 716 717 718 719 720 721
					 sock_len, MSG_DONTWAIT | MSG_TRUNC);
		/* Userspace might have consumed the packet meanwhile:
		 * it's not supposed to do this usually, but might be hard
		 * to prevent. Discard data we got (if any) and keep going. */
		if (unlikely(err != sock_len)) {
			pr_debug("Discarded rx packet: "
				 " len %d, expected %zd\n", err, sock_len);
			vhost_discard_vq_desc(vq, headcount);
			continue;
		}
722
		/* Supply virtio_net_hdr if VHOST_NET_F_VIRTIO_NET_HDR */
723 724 725 726 727
		if (unlikely(vhost_hlen)) {
			if (copy_to_iter(&hdr, sizeof(hdr),
					 &fixup) != sizeof(hdr)) {
				vq_err(vq, "Unable to write vnet_hdr "
				       "at addr %p\n", vq->iov->iov_base);
728
				goto out;
729 730 731 732 733 734
			}
		} else {
			/* Header came from socket; we'll need to patch
			 * ->num_buffers over if VIRTIO_NET_F_MRG_RXBUF
			 */
			iov_iter_advance(&fixup, sizeof(hdr));
735 736
		}
		/* TODO: Should check and handle checksum. */
737

738
		num_buffers = cpu_to_vhost16(vq, headcount);
739
		if (likely(mergeable) &&
740 741
		    copy_to_iter(&num_buffers, sizeof num_buffers,
				 &fixup) != sizeof num_buffers) {
742 743
			vq_err(vq, "Failed num_buffers write");
			vhost_discard_vq_desc(vq, headcount);
744
			goto out;
745 746 747 748 749 750 751 752
		}
		vhost_add_used_and_signal_n(&net->dev, vq, vq->heads,
					    headcount);
		if (unlikely(vq_log))
			vhost_log_write(vq, vq_log, log, vhost_len);
		total_len += vhost_len;
		if (unlikely(total_len >= VHOST_NET_WEIGHT)) {
			vhost_poll_queue(&vq->poll);
753
			goto out;
754 755
		}
	}
756
	vhost_net_enable_vq(net, vq);
757
out:
758 759 760
	mutex_unlock(&vq->mutex);
}

761
static void handle_tx_kick(struct vhost_work *work)
762
{
763 764 765 766
	struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue,
						  poll.work);
	struct vhost_net *net = container_of(vq->dev, struct vhost_net, dev);

767 768 769
	handle_tx(net);
}

770
static void handle_rx_kick(struct vhost_work *work)
771
{
772 773 774 775
	struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue,
						  poll.work);
	struct vhost_net *net = container_of(vq->dev, struct vhost_net, dev);

776 777 778
	handle_rx(net);
}

779
static void handle_tx_net(struct vhost_work *work)
780
{
781 782
	struct vhost_net *net = container_of(work, struct vhost_net,
					     poll[VHOST_NET_VQ_TX].work);
783 784 785
	handle_tx(net);
}

786
static void handle_rx_net(struct vhost_work *work)
787
{
788 789
	struct vhost_net *net = container_of(work, struct vhost_net,
					     poll[VHOST_NET_VQ_RX].work);
790 791 792 793 794
	handle_rx(net);
}

static int vhost_net_open(struct inode *inode, struct file *f)
{
795
	struct vhost_net *n;
796
	struct vhost_dev *dev;
797
	struct vhost_virtqueue **vqs;
Zhi Yong Wu's avatar
Zhi Yong Wu committed
798
	int i;
799

800 801 802 803 804 805
	n = kmalloc(sizeof *n, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!n) {
		n = vmalloc(sizeof *n);
		if (!n)
			return -ENOMEM;
	}
806 807
	vqs = kmalloc(VHOST_NET_VQ_MAX * sizeof(*vqs), GFP_KERNEL);
	if (!vqs) {
808
		kvfree(n);
809 810
		return -ENOMEM;
	}
811 812

	dev = &n->dev;
813 814 815 816
	vqs[VHOST_NET_VQ_TX] = &n->vqs[VHOST_NET_VQ_TX].vq;
	vqs[VHOST_NET_VQ_RX] = &n->vqs[VHOST_NET_VQ_RX].vq;
	n->vqs[VHOST_NET_VQ_TX].vq.handle_kick = handle_tx_kick;
	n->vqs[VHOST_NET_VQ_RX].vq.handle_kick = handle_rx_kick;
817 818 819 820 821
	for (i = 0; i < VHOST_NET_VQ_MAX; i++) {
		n->vqs[i].ubufs = NULL;
		n->vqs[i].ubuf_info = NULL;
		n->vqs[i].upend_idx = 0;
		n->vqs[i].done_idx = 0;
822 823
		n->vqs[i].vhost_hlen = 0;
		n->vqs[i].sock_hlen = 0;
824
	}
Zhi Yong Wu's avatar
Zhi Yong Wu committed
825
	vhost_dev_init(dev, vqs, VHOST_NET_VQ_MAX);
826

827 828
	vhost_poll_init(n->poll + VHOST_NET_VQ_TX, handle_tx_net, POLLOUT, dev);
	vhost_poll_init(n->poll + VHOST_NET_VQ_RX, handle_rx_net, POLLIN, dev);
829 830 831 832 833 834 835 836 837 838 839 840

	f->private_data = n;

	return 0;
}

static struct socket *vhost_net_stop_vq(struct vhost_net *n,
					struct vhost_virtqueue *vq)
{
	struct socket *sock;

	mutex_lock(&vq->mutex);
841
	sock = vq->private_data;
842
	vhost_net_disable_vq(n, vq);
843
	vq->private_data = NULL;
844 845 846 847 848 849 850
	mutex_unlock(&vq->mutex);
	return sock;
}

static void vhost_net_stop(struct vhost_net *n, struct socket **tx_sock,
			   struct socket **rx_sock)
{
851 852
	*tx_sock = vhost_net_stop_vq(n, &n->vqs[VHOST_NET_VQ_TX].vq);
	*rx_sock = vhost_net_stop_vq(n, &n->vqs[VHOST_NET_VQ_RX].vq);
853 854 855 856 857
}

static void vhost_net_flush_vq(struct vhost_net *n, int index)
{
	vhost_poll_flush(n->poll + index);
858
	vhost_poll_flush(&n->vqs[index].vq.poll);
859 860 861 862 863 864
}

static void vhost_net_flush(struct vhost_net *n)
{
	vhost_net_flush_vq(n, VHOST_NET_VQ_TX);
	vhost_net_flush_vq(n, VHOST_NET_VQ_RX);
865
	if (n->vqs[VHOST_NET_VQ_TX].ubufs) {
866
		mutex_lock(&n->vqs[VHOST_NET_VQ_TX].vq.mutex);
867
		n->tx_flush = true;
868
		mutex_unlock(&n->vqs[VHOST_NET_VQ_TX].vq.mutex);
869
		/* Wait for all lower device DMAs done. */
870
		vhost_net_ubuf_put_and_wait(n->vqs[VHOST_NET_VQ_TX].ubufs);
871
		mutex_lock(&n->vqs[VHOST_NET_VQ_TX].vq.mutex);
872
		n->tx_flush = false;
873
		atomic_set(&n->vqs[VHOST_NET_VQ_TX].ubufs->refcount, 1);
874
		mutex_unlock(&n->vqs[VHOST_NET_VQ_TX].vq.mutex);
875
	}
876 877 878 879 880 881 882 883 884 885
}

static int vhost_net_release(struct inode *inode, struct file *f)
{
	struct vhost_net *n = f->private_data;
	struct socket *tx_sock;
	struct socket *rx_sock;

	vhost_net_stop(n, &tx_sock, &rx_sock);
	vhost_net_flush(n);
886
	vhost_dev_stop(&n->dev);
887
	vhost_dev_cleanup(&n->dev, false);
888
	vhost_net_vq_reset(n);
889
	if (tx_sock)
890
		sockfd_put(tx_sock);
891
	if (rx_sock)
892
		sockfd_put(rx_sock);
893 894
	/* Make sure no callbacks are outstanding */
	synchronize_rcu_bh();
895 896 897
	/* We do an extra flush before freeing memory,
	 * since jobs can re-queue themselves. */
	vhost_net_flush(n);
898
	kfree(n->dev.vqs);
899
	kvfree(n);
900 901 902 903 904 905 906 907 908 909 910
	return 0;
}

static struct socket *get_raw_socket(int fd)
{
	struct {
		struct sockaddr_ll sa;
		char  buf[MAX_ADDR_LEN];
	} uaddr;
	int uaddr_len = sizeof uaddr, r;
	struct socket *sock = sockfd_lookup(fd, &r);
911

912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931
	if (!sock)
		return ERR_PTR(-ENOTSOCK);

	/* Parameter checking */
	if (sock->sk->sk_type != SOCK_RAW) {
		r = -ESOCKTNOSUPPORT;
		goto err;
	}

	r = sock->ops->getname(sock, (struct sockaddr *)&uaddr.sa,
			       &uaddr_len, 0);
	if (r)
		goto err;

	if (uaddr.sa.sll_family != AF_PACKET) {
		r = -EPFNOSUPPORT;
		goto err;
	}
	return sock;
err:
932
	sockfd_put(sock);
933 934 935
	return ERR_PTR(r);
}

936
static struct socket *get_tap_socket(int fd)
937 938 939
{
	struct file *file = fget(fd);
	struct socket *sock;
940

941 942 943
	if (!file)
		return ERR_PTR(-EBADF);
	sock = tun_get_socket(file);
944 945 946
	if (!IS_ERR(sock))
		return sock;
	sock = macvtap_get_socket(file);
947 948 949 950 951 952 953 954
	if (IS_ERR(sock))
		fput(file);
	return sock;
}

static struct socket *get_socket(int fd)
{
	struct socket *sock;
955

956 957 958 959 960 961
	/* special case to disable backend */
	if (fd == -1)
		return NULL;
	sock = get_raw_socket(fd);
	if (!IS_ERR(sock))
		return sock;
962
	sock = get_tap_socket(fd);
963 964 965 966 967 968 969 970 971
	if (!IS_ERR(sock))
		return sock;
	return ERR_PTR(-ENOTSOCK);
}

static long vhost_net_set_backend(struct vhost_net *n, unsigned index, int fd)
{
	struct socket *sock, *oldsock;
	struct vhost_virtqueue *vq;
972
	struct vhost_net_virtqueue *nvq;
973
	struct vhost_net_ubuf_ref *ubufs, *oldubufs = NULL;
974 975 976 977 978 979 980 981 982 983 984
	int r;

	mutex_lock(&n->dev.mutex);
	r = vhost_dev_check_owner(&n->dev);
	if (r)
		goto err;

	if (index >= VHOST_NET_VQ_MAX) {
		r = -ENOBUFS;
		goto err;
	}
985
	vq = &n->vqs[index].vq;
986
	nvq = &n->vqs[index];
987 988 989 990 991
	mutex_lock(&vq->mutex);

	/* Verify that ring has been setup correctly. */
	if (!vhost_vq_access_ok(vq)) {
		r = -EFAULT;
992
		goto err_vq;
993 994 995 996
	}
	sock = get_socket(fd);
	if (IS_ERR(sock)) {
		r = PTR_ERR(sock);
997
		goto err_vq;
998 999 1000
	}

	/* start polling new socket */
1001
	oldsock = vq->private_data;
1002
	if (sock != oldsock) {
1003 1004
		ubufs = vhost_net_ubuf_alloc(vq,
					     sock && vhost_sock_zcopy(sock));
1005 1006 1007 1008
		if (IS_ERR(ubufs)) {
			r = PTR_ERR(ubufs);
			goto err_ubufs;
		}
1009

1010
		vhost_net_disable_vq(n, vq);
1011
		vq->private_data = sock;
Greg Kurz's avatar
Greg Kurz committed
1012
		r = vhost_vq_init_access(vq);
1013
		if (r)
1014
			goto err_used;
1015 1016 1017
		r = vhost_net_enable_vq(n, vq);
		if (r)
			goto err_used;
1018

1019 1020
		oldubufs = nvq->ubufs;
		nvq->ubufs = ubufs;
1021 1022 1023

		n->tx_packets = 0;
		n->tx_zcopy_err = 0;
1024
		n->tx_flush = false;
Jeff Dike's avatar
Jeff Dike committed
1025
	}
1026

1027 1028
	mutex_unlock(&vq->mutex);

1029
	if (oldubufs) {
1030
		vhost_net_ubuf_put_wait_and_free(oldubufs);
1031
		mutex_lock(&vq->mutex);
1032
		vhost_zerocopy_signal_used(n, vq);
1033 1034
		mutex_unlock(&vq->mutex);
	}
1035

1036 1037
	if (oldsock) {
		vhost_net_flush_vq(n, index);
1038
		sockfd_put(oldsock);
1039
	}
1040

1041 1042 1043
	mutex_unlock(&n->dev.mutex);
	return 0;

1044
err_used:
1045
	vq->private_data = oldsock;
1046 1047
	vhost_net_enable_vq(n, vq);
	if (ubufs)
1048
		vhost_net_ubuf_put_wait_and_free(ubufs);
1049
err_ubufs:
1050
	sockfd_put(sock);
1051 1052
err_vq:
	mutex_unlock(&vq->mutex);
1053 1054 1055 1056 1057 1058 1059 1060 1061 1062
err:
	mutex_unlock(&n->dev.mutex);
	return r;
}

static long vhost_net_reset_owner(struct vhost_net *n)
{
	struct socket *tx_sock = NULL;
	struct socket *rx_sock = NULL;
	long err;
1063
	struct vhost_umem *umem;
1064

1065 1066 1067 1068
	mutex_lock(&n->dev.mutex);
	err = vhost_dev_check_owner(&n->dev);
	if (err)
		goto done;
1069 1070
	umem = vhost_dev_reset_owner_prepare();
	if (!umem) {
1071 1072 1073
		err = -ENOMEM;
		goto done;
	}
1074 1075
	vhost_net_stop(n, &tx_sock, &rx_sock);
	vhost_net_flush(n);
1076
	vhost_dev_reset_owner(&n->dev, umem);
1077
	vhost_net_vq_reset(n);
1078 1079 1080
done:
	mutex_unlock(&n->dev.mutex);
	if (tx_sock)
1081
		sockfd_put(tx_sock);
1082
	if (rx_sock)
1083
		sockfd_put(rx_sock);
1084 1085 1086 1087 1088
	return err;
}

static int vhost_net_set_features(struct vhost_net *n, u64 features)
{
1089
	size_t vhost_hlen, sock_hlen, hdr_len;
1090
	int i;
1091

1092 1093
	hdr_len = (features & ((1ULL << VIRTIO_NET_F_MRG_RXBUF) |
			       (1ULL << VIRTIO_F_VERSION_1))) ?
1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104
			sizeof(struct virtio_net_hdr_mrg_rxbuf) :
			sizeof(struct virtio_net_hdr);
	if (features & (1 << VHOST_NET_F_VIRTIO_NET_HDR)) {
		/* vhost provides vnet_hdr */
		vhost_hlen = hdr_len;
		sock_hlen = 0;
	} else {
		/* socket provides vnet_hdr */
		vhost_hlen = 0;
		sock_hlen = hdr_len;
	}
1105 1106
	mutex_lock(&n->dev.mutex);
	if ((features & (1 << VHOST_F_LOG_ALL)) &&
Jason Wang's avatar
Jason Wang committed
1107 1108 1109 1110 1111 1112
	    !vhost_log_access_ok(&n->dev))
		goto out_unlock;

	if ((features & (1ULL << VIRTIO_F_IOMMU_PLATFORM))) {
		if (vhost_init_device_iotlb(&n->dev, true))
			goto out_unlock;
1113
	}
Jason Wang's avatar
Jason Wang committed
1114

1115
	for (i = 0; i < VHOST_NET_VQ_MAX; ++i) {
1116
		mutex_lock(&n->vqs[i].vq.mutex);
1117
		n->vqs[i].vq.acked_features = features;
1118 1119
		n->vqs[i].vhost_hlen = vhost_hlen;
		n->vqs[i].sock_hlen = sock_hlen;
1120
		mutex_unlock(&n->vqs[i].vq.mutex);
1121