1/*
   2 * Back-end of the driver for virtual network devices. This portion of the
   3 * driver exports a 'unified' network-device interface that can be accessed
   4 * by any operating system that implements a compatible front end. A
   5 * reference front-end implementation can be found in:
   6 *  drivers/net/xen-netfront.c
   7 *
   8 * Copyright (c) 2002-2005, K A Fraser
   9 *
  10 * This program is free software; you can redistribute it and/or
  11 * modify it under the terms of the GNU General Public License version 2
  12 * as published by the Free Software Foundation; or, when distributed
  13 * separately from the Linux kernel or incorporated into other
  14 * software packages, subject to the following license:
  15 *
  16 * Permission is hereby granted, free of charge, to any person obtaining a copy
  17 * of this source file (the "Software"), to deal in the Software without
  18 * restriction, including without limitation the rights to use, copy, modify,
  19 * merge, publish, distribute, sublicense, and/or sell copies of the Software,
  20 * and to permit persons to whom the Software is furnished to do so, subject to
  21 * the following conditions:
  22 *
  23 * The above copyright notice and this permission notice shall be included in
  24 * all copies or substantial portions of the Software.
  25 *
  26 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  27 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  28 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
  29 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  30 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  31 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  32 * IN THE SOFTWARE.
  33 */
  34
  35#include "common.h"
  36
  37#include <linux/kthread.h>
  38#include <linux/if_vlan.h>
  39#include <linux/udp.h>
  40#include <linux/highmem.h>
  41#include <linux/skbuff_ref.h>
  42
  43#include <net/tcp.h>
  44
  45#include <xen/xen.h>
  46#include <xen/events.h>
  47#include <xen/interface/memory.h>
  48#include <xen/page.h>
  49
  50#include <asm/xen/hypercall.h>
  51
  52/* Provide an option to disable split event channels at load time as
  53 * event channels are limited resource. Split event channels are
  54 * enabled by default.
  55 */
  56bool separate_tx_rx_irq = true;
  57module_param(separate_tx_rx_irq, bool, 0644);
  58
  59/* The time that packets can stay on the guest Rx internal queue
  60 * before they are dropped.
  61 */
  62unsigned int rx_drain_timeout_msecs = 10000;
  63module_param(rx_drain_timeout_msecs, uint, 0444);
  64
  65/* The length of time before the frontend is considered unresponsive
  66 * because it isn't providing Rx slots.
  67 */
  68unsigned int rx_stall_timeout_msecs = 60000;
  69module_param(rx_stall_timeout_msecs, uint, 0444);
  70
  71#define MAX_QUEUES_DEFAULT 8
  72unsigned int xenvif_max_queues;
  73module_param_named(max_queues, xenvif_max_queues, uint, 0644);
  74MODULE_PARM_DESC(max_queues,
  75		 "Maximum number of queues per virtual interface");
  76
  77/*
  78 * This is the maximum slots a skb can have. If a guest sends a skb
  79 * which exceeds this limit it is considered malicious.
  80 */
  81#define FATAL_SKB_SLOTS_DEFAULT 20
  82static unsigned int fatal_skb_slots = FATAL_SKB_SLOTS_DEFAULT;
  83module_param(fatal_skb_slots, uint, 0444);
  84
  85/* The amount to copy out of the first guest Tx slot into the skb's
  86 * linear area.  If the first slot has more data, it will be mapped
  87 * and put into the first frag.
  88 *
  89 * This is sized to avoid pulling headers from the frags for most
  90 * TCP/IP packets.
  91 */
  92#define XEN_NETBACK_TX_COPY_LEN 128
  93
  94/* This is the maximum number of flows in the hash cache. */
  95#define XENVIF_HASH_CACHE_SIZE_DEFAULT 64
  96unsigned int xenvif_hash_cache_size = XENVIF_HASH_CACHE_SIZE_DEFAULT;
  97module_param_named(hash_cache_size, xenvif_hash_cache_size, uint, 0644);
  98MODULE_PARM_DESC(hash_cache_size, "Number of flows in the hash cache");
  99
 100/* The module parameter tells that we have to put data
 101 * for xen-netfront with the XDP_PACKET_HEADROOM offset
 102 * needed for XDP processing
 103 */
 104bool provides_xdp_headroom = true;
 105module_param(provides_xdp_headroom, bool, 0644);
 106
 107static void xenvif_idx_release(struct xenvif_queue *queue, u16 pending_idx,
 108			       s8 status);
 109
 110static void make_tx_response(struct xenvif_queue *queue,
 111			     const struct xen_netif_tx_request *txp,
 112			     unsigned int extra_count,
 113			     s8 status);
 114
 115static void xenvif_idx_unmap(struct xenvif_queue *queue, u16 pending_idx);
 116
 117static inline int tx_work_todo(struct xenvif_queue *queue);
 118
 119static inline unsigned long idx_to_pfn(struct xenvif_queue *queue,
 120				       u16 idx)
 121{
 122	return page_to_pfn(queue->mmap_pages[idx]);
 123}
 124
 125static inline unsigned long idx_to_kaddr(struct xenvif_queue *queue,
 126					 u16 idx)
 127{
 128	return (unsigned long)pfn_to_kaddr(idx_to_pfn(queue, idx));
 129}
 130
 131#define callback_param(vif, pending_idx) \
 132	(vif->pending_tx_info[pending_idx].callback_struct)
 133
 134/* Find the containing VIF's structure from a pointer in pending_tx_info array
 135 */
 136static inline struct xenvif_queue *ubuf_to_queue(const struct ubuf_info_msgzc *ubuf)
 137{
 138	u16 pending_idx = ubuf->desc;
 139	struct pending_tx_info *temp =
 140		container_of(ubuf, struct pending_tx_info, callback_struct);
 141	return container_of(temp - pending_idx,
 142			    struct xenvif_queue,
 143			    pending_tx_info[0]);
 144}
 145
 146static u16 frag_get_pending_idx(skb_frag_t *frag)
 147{
 148	return (u16)skb_frag_off(frag);
 149}
 150
 151static void frag_set_pending_idx(skb_frag_t *frag, u16 pending_idx)
 152{
 153	skb_frag_off_set(frag, pending_idx);
 154}
 155
 156static inline pending_ring_idx_t pending_index(unsigned i)
 157{
 158	return i & (MAX_PENDING_REQS-1);
 159}
 160
 161void xenvif_kick_thread(struct xenvif_queue *queue)
 162{
 163	wake_up(&queue->wq);
 164}
 165
 166void xenvif_napi_schedule_or_enable_events(struct xenvif_queue *queue)
 167{
 168	int more_to_do;
 169
 170	RING_FINAL_CHECK_FOR_REQUESTS(&queue->tx, more_to_do);
 171
 172	if (more_to_do)
 173		napi_schedule(&queue->napi);
 174	else if (atomic_fetch_andnot(NETBK_TX_EOI | NETBK_COMMON_EOI,
 175				     &queue->eoi_pending) &
 176		 (NETBK_TX_EOI | NETBK_COMMON_EOI))
 177		xen_irq_lateeoi(queue->tx_irq, 0);
 178}
 179
 180static void tx_add_credit(struct xenvif_queue *queue)
 181{
 182	unsigned long max_burst, max_credit;
 183
 184	/*
 185	 * Allow a burst big enough to transmit a jumbo packet of up to 128kB.
 186	 * Otherwise the interface can seize up due to insufficient credit.
 187	 */
 188	max_burst = max(131072UL, queue->credit_bytes);
 189
 190	/* Take care that adding a new chunk of credit doesn't wrap to zero. */
 191	max_credit = queue->remaining_credit + queue->credit_bytes;
 192	if (max_credit < queue->remaining_credit)
 193		max_credit = ULONG_MAX; /* wrapped: clamp to ULONG_MAX */
 194
 195	queue->remaining_credit = min(max_credit, max_burst);
 196	queue->rate_limited = false;
 197}
 198
 199void xenvif_tx_credit_callback(struct timer_list *t)
 200{
 201	struct xenvif_queue *queue = from_timer(queue, t, credit_timeout);
 202	tx_add_credit(queue);
 203	xenvif_napi_schedule_or_enable_events(queue);
 204}
 205
 206static void xenvif_tx_err(struct xenvif_queue *queue,
 207			  struct xen_netif_tx_request *txp,
 208			  unsigned int extra_count, RING_IDX end)
 209{
 210	RING_IDX cons = queue->tx.req_cons;
 
 211
 212	do {
 
 213		make_tx_response(queue, txp, extra_count, XEN_NETIF_RSP_ERROR);
 
 
 214		if (cons == end)
 215			break;
 216		RING_COPY_REQUEST(&queue->tx, cons++, txp);
 217		extra_count = 0; /* only the first frag can have extras */
 218	} while (1);
 219	queue->tx.req_cons = cons;
 220}
 221
 222static void xenvif_fatal_tx_err(struct xenvif *vif)
 223{
 224	netdev_err(vif->dev, "fatal error; disabling device\n");
 225	vif->disabled = true;
 226	/* Disable the vif from queue 0's kthread */
 227	if (vif->num_queues)
 228		xenvif_kick_thread(&vif->queues[0]);
 229}
 230
 231static int xenvif_count_requests(struct xenvif_queue *queue,
 232				 struct xen_netif_tx_request *first,
 233				 unsigned int extra_count,
 234				 struct xen_netif_tx_request *txp,
 235				 int work_to_do)
 236{
 237	RING_IDX cons = queue->tx.req_cons;
 238	int slots = 0;
 239	int drop_err = 0;
 240	int more_data;
 241
 242	if (!(first->flags & XEN_NETTXF_more_data))
 243		return 0;
 244
 245	do {
 246		struct xen_netif_tx_request dropped_tx = { 0 };
 247
 248		if (slots >= work_to_do) {
 249			netdev_err(queue->vif->dev,
 250				   "Asked for %d slots but exceeds this limit\n",
 251				   work_to_do);
 252			xenvif_fatal_tx_err(queue->vif);
 253			return -ENODATA;
 254		}
 255
 256		/* This guest is really using too many slots and
 257		 * considered malicious.
 258		 */
 259		if (unlikely(slots >= fatal_skb_slots)) {
 260			netdev_err(queue->vif->dev,
 261				   "Malicious frontend using %d slots, threshold %u\n",
 262				   slots, fatal_skb_slots);
 263			xenvif_fatal_tx_err(queue->vif);
 264			return -E2BIG;
 265		}
 266
 267		/* Xen network protocol had implicit dependency on
 268		 * MAX_SKB_FRAGS. XEN_NETBK_LEGACY_SLOTS_MAX is set to
 269		 * the historical MAX_SKB_FRAGS value 18 to honor the
 270		 * same behavior as before. Any packet using more than
 271		 * 18 slots but less than fatal_skb_slots slots is
 272		 * dropped
 273		 */
 274		if (!drop_err && slots >= XEN_NETBK_LEGACY_SLOTS_MAX) {
 275			if (net_ratelimit())
 276				netdev_dbg(queue->vif->dev,
 277					   "Too many slots (%d) exceeding limit (%d), dropping packet\n",
 278					   slots, XEN_NETBK_LEGACY_SLOTS_MAX);
 279			drop_err = -E2BIG;
 280		}
 281
 282		if (drop_err)
 283			txp = &dropped_tx;
 284
 285		RING_COPY_REQUEST(&queue->tx, cons + slots, txp);
 286
 287		/* If the guest submitted a frame >= 64 KiB then
 288		 * first->size overflowed and following slots will
 289		 * appear to be larger than the frame.
 290		 *
 291		 * This cannot be fatal error as there are buggy
 292		 * frontends that do this.
 293		 *
 294		 * Consume all slots and drop the packet.
 295		 */
 296		if (!drop_err && txp->size > first->size) {
 297			if (net_ratelimit())
 298				netdev_dbg(queue->vif->dev,
 299					   "Invalid tx request, slot size %u > remaining size %u\n",
 300					   txp->size, first->size);
 301			drop_err = -EIO;
 302		}
 303
 304		first->size -= txp->size;
 305		slots++;
 306
 307		if (unlikely((txp->offset + txp->size) > XEN_PAGE_SIZE)) {
 308			netdev_err(queue->vif->dev, "Cross page boundary, txp->offset: %u, size: %u\n",
 309				 txp->offset, txp->size);
 310			xenvif_fatal_tx_err(queue->vif);
 311			return -EINVAL;
 312		}
 313
 314		more_data = txp->flags & XEN_NETTXF_more_data;
 315
 316		if (!drop_err)
 317			txp++;
 318
 319	} while (more_data);
 320
 321	if (drop_err) {
 322		xenvif_tx_err(queue, first, extra_count, cons + slots);
 323		return drop_err;
 324	}
 325
 326	return slots;
 327}
 328
 329
 330struct xenvif_tx_cb {
 331	u16 copy_pending_idx[XEN_NETBK_LEGACY_SLOTS_MAX + 1];
 332	u8 copy_count;
 333	u32 split_mask;
 334};
 335
 336#define XENVIF_TX_CB(skb) ((struct xenvif_tx_cb *)(skb)->cb)
 337#define copy_pending_idx(skb, i) (XENVIF_TX_CB(skb)->copy_pending_idx[i])
 338#define copy_count(skb) (XENVIF_TX_CB(skb)->copy_count)
 339
 340static inline void xenvif_tx_create_map_op(struct xenvif_queue *queue,
 341					   u16 pending_idx,
 342					   struct xen_netif_tx_request *txp,
 343					   unsigned int extra_count,
 344					   struct gnttab_map_grant_ref *mop)
 345{
 346	queue->pages_to_map[mop-queue->tx_map_ops] = queue->mmap_pages[pending_idx];
 347	gnttab_set_map_op(mop, idx_to_kaddr(queue, pending_idx),
 348			  GNTMAP_host_map | GNTMAP_readonly,
 349			  txp->gref, queue->vif->domid);
 350
 351	memcpy(&queue->pending_tx_info[pending_idx].req, txp,
 352	       sizeof(*txp));
 353	queue->pending_tx_info[pending_idx].extra_count = extra_count;
 354}
 355
 356static inline struct sk_buff *xenvif_alloc_skb(unsigned int size)
 357{
 358	struct sk_buff *skb =
 359		alloc_skb(size + NET_SKB_PAD + NET_IP_ALIGN,
 360			  GFP_ATOMIC | __GFP_NOWARN);
 361
 362	BUILD_BUG_ON(sizeof(*XENVIF_TX_CB(skb)) > sizeof(skb->cb));
 363	if (unlikely(skb == NULL))
 364		return NULL;
 365
 366	/* Packets passed to netif_rx() must have some headroom. */
 367	skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN);
 368
 369	/* Initialize it here to avoid later surprises */
 370	skb_shinfo(skb)->destructor_arg = NULL;
 371
 372	return skb;
 373}
 374
 375static void xenvif_get_requests(struct xenvif_queue *queue,
 376				struct sk_buff *skb,
 377				struct xen_netif_tx_request *first,
 378				struct xen_netif_tx_request *txfrags,
 379			        unsigned *copy_ops,
 380			        unsigned *map_ops,
 381				unsigned int frag_overflow,
 382				struct sk_buff *nskb,
 383				unsigned int extra_count,
 384				unsigned int data_len)
 385{
 386	struct skb_shared_info *shinfo = skb_shinfo(skb);
 387	skb_frag_t *frags = shinfo->frags;
 388	u16 pending_idx;
 
 389	pending_ring_idx_t index;
 390	unsigned int nr_slots;
 391	struct gnttab_copy *cop = queue->tx_copy_ops + *copy_ops;
 392	struct gnttab_map_grant_ref *gop = queue->tx_map_ops + *map_ops;
 393	struct xen_netif_tx_request *txp = first;
 394
 395	nr_slots = shinfo->nr_frags + frag_overflow + 1;
 396
 397	copy_count(skb) = 0;
 398	XENVIF_TX_CB(skb)->split_mask = 0;
 399
 400	/* Create copy ops for exactly data_len bytes into the skb head. */
 401	__skb_put(skb, data_len);
 402	while (data_len > 0) {
 403		int amount = data_len > txp->size ? txp->size : data_len;
 404		bool split = false;
 405
 406		cop->source.u.ref = txp->gref;
 407		cop->source.domid = queue->vif->domid;
 408		cop->source.offset = txp->offset;
 409
 410		cop->dest.domid = DOMID_SELF;
 411		cop->dest.offset = (offset_in_page(skb->data +
 412						   skb_headlen(skb) -
 413						   data_len)) & ~XEN_PAGE_MASK;
 414		cop->dest.u.gmfn = virt_to_gfn(skb->data + skb_headlen(skb)
 415				               - data_len);
 416
 417		/* Don't cross local page boundary! */
 418		if (cop->dest.offset + amount > XEN_PAGE_SIZE) {
 419			amount = XEN_PAGE_SIZE - cop->dest.offset;
 420			XENVIF_TX_CB(skb)->split_mask |= 1U << copy_count(skb);
 421			split = true;
 422		}
 423
 424		cop->len = amount;
 425		cop->flags = GNTCOPY_source_gref;
 426
 427		index = pending_index(queue->pending_cons);
 428		pending_idx = queue->pending_ring[index];
 429		callback_param(queue, pending_idx).ctx = NULL;
 430		copy_pending_idx(skb, copy_count(skb)) = pending_idx;
 431		if (!split)
 432			copy_count(skb)++;
 433
 434		cop++;
 435		data_len -= amount;
 436
 437		if (amount == txp->size) {
 438			/* The copy op covered the full tx_request */
 439
 440			memcpy(&queue->pending_tx_info[pending_idx].req,
 441			       txp, sizeof(*txp));
 442			queue->pending_tx_info[pending_idx].extra_count =
 443				(txp == first) ? extra_count : 0;
 444
 445			if (txp == first)
 446				txp = txfrags;
 447			else
 448				txp++;
 449			queue->pending_cons++;
 450			nr_slots--;
 451		} else {
 452			/* The copy op partially covered the tx_request.
 453			 * The remainder will be mapped or copied in the next
 454			 * iteration.
 455			 */
 456			txp->offset += amount;
 457			txp->size -= amount;
 458		}
 459	}
 460
 461	for (shinfo->nr_frags = 0; nr_slots > 0 && shinfo->nr_frags < MAX_SKB_FRAGS;
 462	     nr_slots--) {
 463		if (unlikely(!txp->size)) {
 464			make_tx_response(queue, txp, 0, XEN_NETIF_RSP_OKAY);
 465			++txp;
 466			continue;
 467		}
 468
 
 
 469		index = pending_index(queue->pending_cons++);
 470		pending_idx = queue->pending_ring[index];
 471		xenvif_tx_create_map_op(queue, pending_idx, txp,
 472				        txp == first ? extra_count : 0, gop);
 473		frag_set_pending_idx(&frags[shinfo->nr_frags], pending_idx);
 474		++shinfo->nr_frags;
 475		++gop;
 476
 477		if (txp == first)
 478			txp = txfrags;
 479		else
 480			txp++;
 481	}
 482
 483	if (nr_slots > 0) {
 484
 485		shinfo = skb_shinfo(nskb);
 486		frags = shinfo->frags;
 487
 488		for (shinfo->nr_frags = 0; shinfo->nr_frags < nr_slots; ++txp) {
 489			if (unlikely(!txp->size)) {
 490				make_tx_response(queue, txp, 0,
 491						 XEN_NETIF_RSP_OKAY);
 492				continue;
 493			}
 494
 495			index = pending_index(queue->pending_cons++);
 496			pending_idx = queue->pending_ring[index];
 497			xenvif_tx_create_map_op(queue, pending_idx, txp, 0,
 498						gop);
 499			frag_set_pending_idx(&frags[shinfo->nr_frags],
 500					     pending_idx);
 501			++shinfo->nr_frags;
 502			++gop;
 503		}
 504
 505		if (shinfo->nr_frags) {
 506			skb_shinfo(skb)->frag_list = nskb;
 507			nskb = NULL;
 508		}
 509	}
 510
 511	if (nskb) {
 512		/* A frag_list skb was allocated but it is no longer needed
 513		 * because enough slots were converted to copy ops above or some
 514		 * were empty.
 515		 */
 516		kfree_skb(nskb);
 517	}
 518
 519	(*copy_ops) = cop - queue->tx_copy_ops;
 520	(*map_ops) = gop - queue->tx_map_ops;
 521}
 522
 523static inline void xenvif_grant_handle_set(struct xenvif_queue *queue,
 524					   u16 pending_idx,
 525					   grant_handle_t handle)
 526{
 527	if (unlikely(queue->grant_tx_handle[pending_idx] !=
 528		     NETBACK_INVALID_HANDLE)) {
 529		netdev_err(queue->vif->dev,
 530			   "Trying to overwrite active handle! pending_idx: 0x%x\n",
 531			   pending_idx);
 532		BUG();
 533	}
 534	queue->grant_tx_handle[pending_idx] = handle;
 535}
 536
 537static inline void xenvif_grant_handle_reset(struct xenvif_queue *queue,
 538					     u16 pending_idx)
 539{
 540	if (unlikely(queue->grant_tx_handle[pending_idx] ==
 541		     NETBACK_INVALID_HANDLE)) {
 542		netdev_err(queue->vif->dev,
 543			   "Trying to unmap invalid handle! pending_idx: 0x%x\n",
 544			   pending_idx);
 545		BUG();
 546	}
 547	queue->grant_tx_handle[pending_idx] = NETBACK_INVALID_HANDLE;
 548}
 549
 550static int xenvif_tx_check_gop(struct xenvif_queue *queue,
 551			       struct sk_buff *skb,
 552			       struct gnttab_map_grant_ref **gopp_map,
 553			       struct gnttab_copy **gopp_copy)
 554{
 555	struct gnttab_map_grant_ref *gop_map = *gopp_map;
 556	u16 pending_idx;
 557	/* This always points to the shinfo of the skb being checked, which
 558	 * could be either the first or the one on the frag_list
 559	 */
 560	struct skb_shared_info *shinfo = skb_shinfo(skb);
 561	/* If this is non-NULL, we are currently checking the frag_list skb, and
 562	 * this points to the shinfo of the first one
 563	 */
 564	struct skb_shared_info *first_shinfo = NULL;
 565	int nr_frags = shinfo->nr_frags;
 566	const bool sharedslot = nr_frags &&
 567				frag_get_pending_idx(&shinfo->frags[0]) ==
 568				    copy_pending_idx(skb, copy_count(skb) - 1);
 569	int i, err = 0;
 570
 571	for (i = 0; i < copy_count(skb); i++) {
 572		int newerr;
 573
 574		/* Check status of header. */
 575		pending_idx = copy_pending_idx(skb, i);
 576
 577		newerr = (*gopp_copy)->status;
 578
 579		/* Split copies need to be handled together. */
 580		if (XENVIF_TX_CB(skb)->split_mask & (1U << i)) {
 581			(*gopp_copy)++;
 582			if (!newerr)
 583				newerr = (*gopp_copy)->status;
 584		}
 585		if (likely(!newerr)) {
 586			/* The first frag might still have this slot mapped */
 587			if (i < copy_count(skb) - 1 || !sharedslot)
 588				xenvif_idx_release(queue, pending_idx,
 589						   XEN_NETIF_RSP_OKAY);
 590		} else {
 591			err = newerr;
 592			if (net_ratelimit())
 593				netdev_dbg(queue->vif->dev,
 594					   "Grant copy of header failed! status: %d pending_idx: %u ref: %u\n",
 595					   (*gopp_copy)->status,
 596					   pending_idx,
 597					   (*gopp_copy)->source.u.ref);
 598			/* The first frag might still have this slot mapped */
 599			if (i < copy_count(skb) - 1 || !sharedslot)
 600				xenvif_idx_release(queue, pending_idx,
 601						   XEN_NETIF_RSP_ERROR);
 602		}
 603		(*gopp_copy)++;
 604	}
 
 605
 606check_frags:
 607	for (i = 0; i < nr_frags; i++, gop_map++) {
 608		int j, newerr;
 609
 610		pending_idx = frag_get_pending_idx(&shinfo->frags[i]);
 611
 612		/* Check error status: if okay then remember grant handle. */
 613		newerr = gop_map->status;
 614
 615		if (likely(!newerr)) {
 616			xenvif_grant_handle_set(queue,
 617						pending_idx,
 618						gop_map->handle);
 619			/* Had a previous error? Invalidate this fragment. */
 620			if (unlikely(err)) {
 621				xenvif_idx_unmap(queue, pending_idx);
 622				/* If the mapping of the first frag was OK, but
 623				 * the header's copy failed, and they are
 624				 * sharing a slot, send an error
 625				 */
 626				if (i == 0 && !first_shinfo && sharedslot)
 627					xenvif_idx_release(queue, pending_idx,
 628							   XEN_NETIF_RSP_ERROR);
 629				else
 630					xenvif_idx_release(queue, pending_idx,
 631							   XEN_NETIF_RSP_OKAY);
 632			}
 633			continue;
 634		}
 635
 636		/* Error on this fragment: respond to client with an error. */
 637		if (net_ratelimit())
 638			netdev_dbg(queue->vif->dev,
 639				   "Grant map of %d. frag failed! status: %d pending_idx: %u ref: %u\n",
 640				   i,
 641				   gop_map->status,
 642				   pending_idx,
 643				   gop_map->ref);
 644
 645		xenvif_idx_release(queue, pending_idx, XEN_NETIF_RSP_ERROR);
 646
 647		/* Not the first error? Preceding frags already invalidated. */
 648		if (err)
 649			continue;
 650
 
 
 
 
 
 
 
 
 651		/* Invalidate preceding fragments of this skb. */
 652		for (j = 0; j < i; j++) {
 653			pending_idx = frag_get_pending_idx(&shinfo->frags[j]);
 654			xenvif_idx_unmap(queue, pending_idx);
 655			xenvif_idx_release(queue, pending_idx,
 656					   XEN_NETIF_RSP_OKAY);
 657		}
 658
 659		/* And if we found the error while checking the frag_list, unmap
 660		 * the first skb's frags
 661		 */
 662		if (first_shinfo) {
 663			for (j = 0; j < first_shinfo->nr_frags; j++) {
 664				pending_idx = frag_get_pending_idx(&first_shinfo->frags[j]);
 665				xenvif_idx_unmap(queue, pending_idx);
 666				xenvif_idx_release(queue, pending_idx,
 667						   XEN_NETIF_RSP_OKAY);
 668			}
 669		}
 670
 671		/* Remember the error: invalidate all subsequent fragments. */
 672		err = newerr;
 673	}
 674
 675	if (skb_has_frag_list(skb) && !first_shinfo) {
 676		first_shinfo = shinfo;
 677		shinfo = skb_shinfo(shinfo->frag_list);
 678		nr_frags = shinfo->nr_frags;
 679
 680		goto check_frags;
 681	}
 682
 683	*gopp_map = gop_map;
 684	return err;
 685}
 686
 687static void xenvif_fill_frags(struct xenvif_queue *queue, struct sk_buff *skb)
 688{
 689	struct skb_shared_info *shinfo = skb_shinfo(skb);
 690	int nr_frags = shinfo->nr_frags;
 691	int i;
 692	u16 prev_pending_idx = INVALID_PENDING_IDX;
 693
 694	for (i = 0; i < nr_frags; i++) {
 695		skb_frag_t *frag = shinfo->frags + i;
 696		struct xen_netif_tx_request *txp;
 697		struct page *page;
 698		u16 pending_idx;
 699
 700		pending_idx = frag_get_pending_idx(frag);
 701
 702		/* If this is not the first frag, chain it to the previous*/
 703		if (prev_pending_idx == INVALID_PENDING_IDX)
 704			skb_shinfo(skb)->destructor_arg =
 705				&callback_param(queue, pending_idx);
 706		else
 707			callback_param(queue, prev_pending_idx).ctx =
 708				&callback_param(queue, pending_idx);
 709
 710		callback_param(queue, pending_idx).ctx = NULL;
 711		prev_pending_idx = pending_idx;
 712
 713		txp = &queue->pending_tx_info[pending_idx].req;
 714		page = virt_to_page((void *)idx_to_kaddr(queue, pending_idx));
 715		__skb_fill_page_desc(skb, i, page, txp->offset, txp->size);
 716		skb->len += txp->size;
 717		skb->data_len += txp->size;
 718		skb->truesize += txp->size;
 719
 720		/* Take an extra reference to offset network stack's put_page */
 721		get_page(queue->mmap_pages[pending_idx]);
 722	}
 723}
 724
 725static int xenvif_get_extras(struct xenvif_queue *queue,
 726			     struct xen_netif_extra_info *extras,
 727			     unsigned int *extra_count,
 728			     int work_to_do)
 729{
 730	struct xen_netif_extra_info extra;
 731	RING_IDX cons = queue->tx.req_cons;
 732
 733	do {
 734		if (unlikely(work_to_do-- <= 0)) {
 735			netdev_err(queue->vif->dev, "Missing extra info\n");
 736			xenvif_fatal_tx_err(queue->vif);
 737			return -EBADR;
 738		}
 739
 740		RING_COPY_REQUEST(&queue->tx, cons, &extra);
 741
 742		queue->tx.req_cons = ++cons;
 743		(*extra_count)++;
 744
 745		if (unlikely(!extra.type ||
 746			     extra.type >= XEN_NETIF_EXTRA_TYPE_MAX)) {
 747			netdev_err(queue->vif->dev,
 748				   "Invalid extra type: %d\n", extra.type);
 749			xenvif_fatal_tx_err(queue->vif);
 750			return -EINVAL;
 751		}
 752
 753		memcpy(&extras[extra.type - 1], &extra, sizeof(extra));
 754	} while (extra.flags & XEN_NETIF_EXTRA_FLAG_MORE);
 755
 756	return work_to_do;
 757}
 758
 759static int xenvif_set_skb_gso(struct xenvif *vif,
 760			      struct sk_buff *skb,
 761			      struct xen_netif_extra_info *gso)
 762{
 763	if (!gso->u.gso.size) {
 764		netdev_err(vif->dev, "GSO size must not be zero.\n");
 765		xenvif_fatal_tx_err(vif);
 766		return -EINVAL;
 767	}
 768
 769	switch (gso->u.gso.type) {
 770	case XEN_NETIF_GSO_TYPE_TCPV4:
 771		skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
 772		break;
 773	case XEN_NETIF_GSO_TYPE_TCPV6:
 774		skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
 775		break;
 776	default:
 777		netdev_err(vif->dev, "Bad GSO type %d.\n", gso->u.gso.type);
 778		xenvif_fatal_tx_err(vif);
 779		return -EINVAL;
 780	}
 781
 782	skb_shinfo(skb)->gso_size = gso->u.gso.size;
 783	/* gso_segs will be calculated later */
 784
 785	return 0;
 786}
 787
 788static int checksum_setup(struct xenvif_queue *queue, struct sk_buff *skb)
 789{
 790	bool recalculate_partial_csum = false;
 791
 792	/* A GSO SKB must be CHECKSUM_PARTIAL. However some buggy
 793	 * peers can fail to set NETRXF_csum_blank when sending a GSO
 794	 * frame. In this case force the SKB to CHECKSUM_PARTIAL and
 795	 * recalculate the partial checksum.
 796	 */
 797	if (skb->ip_summed != CHECKSUM_PARTIAL && skb_is_gso(skb)) {
 798		queue->stats.rx_gso_checksum_fixup++;
 799		skb->ip_summed = CHECKSUM_PARTIAL;
 800		recalculate_partial_csum = true;
 801	}
 802
 803	/* A non-CHECKSUM_PARTIAL SKB does not require setup. */
 804	if (skb->ip_summed != CHECKSUM_PARTIAL)
 805		return 0;
 806
 807	return skb_checksum_setup(skb, recalculate_partial_csum);
 808}
 809
 810static bool tx_credit_exceeded(struct xenvif_queue *queue, unsigned size)
 811{
 812	u64 now = get_jiffies_64();
 813	u64 next_credit = queue->credit_window_start +
 814		msecs_to_jiffies(queue->credit_usec / 1000);
 815
 816	/* Timer could already be pending in rare cases. */
 817	if (timer_pending(&queue->credit_timeout)) {
 818		queue->rate_limited = true;
 819		return true;
 820	}
 821
 822	/* Passed the point where we can replenish credit? */
 823	if (time_after_eq64(now, next_credit)) {
 824		queue->credit_window_start = now;
 825		tx_add_credit(queue);
 826	}
 827
 828	/* Still too big to send right now? Set a callback. */
 829	if (size > queue->remaining_credit) {
 830		mod_timer(&queue->credit_timeout,
 831			  next_credit);
 832		queue->credit_window_start = next_credit;
 833		queue->rate_limited = true;
 834
 835		return true;
 836	}
 837
 838	return false;
 839}
 840
 841/* No locking is required in xenvif_mcast_add/del() as they are
 842 * only ever invoked from NAPI poll. An RCU list is used because
 843 * xenvif_mcast_match() is called asynchronously, during start_xmit.
 844 */
 845
 846static int xenvif_mcast_add(struct xenvif *vif, const u8 *addr)
 847{
 848	struct xenvif_mcast_addr *mcast;
 849
 850	if (vif->fe_mcast_count == XEN_NETBK_MCAST_MAX) {
 851		if (net_ratelimit())
 852			netdev_err(vif->dev,
 853				   "Too many multicast addresses\n");
 854		return -ENOSPC;
 855	}
 856
 857	mcast = kzalloc(sizeof(*mcast), GFP_ATOMIC);
 858	if (!mcast)
 859		return -ENOMEM;
 860
 861	ether_addr_copy(mcast->addr, addr);
 862	list_add_tail_rcu(&mcast->entry, &vif->fe_mcast_addr);
 863	vif->fe_mcast_count++;
 864
 865	return 0;
 866}
 867
 868static void xenvif_mcast_del(struct xenvif *vif, const u8 *addr)
 869{
 870	struct xenvif_mcast_addr *mcast;
 871
 872	list_for_each_entry_rcu(mcast, &vif->fe_mcast_addr, entry) {
 873		if (ether_addr_equal(addr, mcast->addr)) {
 874			--vif->fe_mcast_count;
 875			list_del_rcu(&mcast->entry);
 876			kfree_rcu(mcast, rcu);
 877			break;
 878		}
 879	}
 880}
 881
 882bool xenvif_mcast_match(struct xenvif *vif, const u8 *addr)
 883{
 884	struct xenvif_mcast_addr *mcast;
 885
 886	rcu_read_lock();
 887	list_for_each_entry_rcu(mcast, &vif->fe_mcast_addr, entry) {
 888		if (ether_addr_equal(addr, mcast->addr)) {
 889			rcu_read_unlock();
 890			return true;
 891		}
 892	}
 893	rcu_read_unlock();
 894
 895	return false;
 896}
 897
 898void xenvif_mcast_addr_list_free(struct xenvif *vif)
 899{
 900	/* No need for locking or RCU here. NAPI poll and TX queue
 901	 * are stopped.
 902	 */
 903	while (!list_empty(&vif->fe_mcast_addr)) {
 904		struct xenvif_mcast_addr *mcast;
 905
 906		mcast = list_first_entry(&vif->fe_mcast_addr,
 907					 struct xenvif_mcast_addr,
 908					 entry);
 909		--vif->fe_mcast_count;
 910		list_del(&mcast->entry);
 911		kfree(mcast);
 912	}
 913}
 914
 915static void xenvif_tx_build_gops(struct xenvif_queue *queue,
 916				     int budget,
 917				     unsigned *copy_ops,
 918				     unsigned *map_ops)
 919{
 
 920	struct sk_buff *skb, *nskb;
 921	int ret;
 922	unsigned int frag_overflow;
 923
 924	while (skb_queue_len(&queue->tx_queue) < budget) {
 925		struct xen_netif_tx_request txreq;
 926		struct xen_netif_tx_request txfrags[XEN_NETBK_LEGACY_SLOTS_MAX];
 927		struct xen_netif_extra_info extras[XEN_NETIF_EXTRA_TYPE_MAX-1];
 928		unsigned int extra_count;
 
 929		RING_IDX idx;
 930		int work_to_do;
 931		unsigned int data_len;
 
 932
 933		if (queue->tx.sring->req_prod - queue->tx.req_cons >
 934		    XEN_NETIF_TX_RING_SIZE) {
 935			netdev_err(queue->vif->dev,
 936				   "Impossible number of requests. "
 937				   "req_prod %d, req_cons %d, size %ld\n",
 938				   queue->tx.sring->req_prod, queue->tx.req_cons,
 939				   XEN_NETIF_TX_RING_SIZE);
 940			xenvif_fatal_tx_err(queue->vif);
 941			break;
 942		}
 943
 944		work_to_do = XEN_RING_NR_UNCONSUMED_REQUESTS(&queue->tx);
 945		if (!work_to_do)
 946			break;
 947
 948		idx = queue->tx.req_cons;
 949		rmb(); /* Ensure that we see the request before we copy it. */
 950		RING_COPY_REQUEST(&queue->tx, idx, &txreq);
 951
 952		/* Credit-based scheduling. */
 953		if (txreq.size > queue->remaining_credit &&
 954		    tx_credit_exceeded(queue, txreq.size))
 955			break;
 956
 957		queue->remaining_credit -= txreq.size;
 958
 959		work_to_do--;
 960		queue->tx.req_cons = ++idx;
 961
 962		memset(extras, 0, sizeof(extras));
 963		extra_count = 0;
 964		if (txreq.flags & XEN_NETTXF_extra_info) {
 965			work_to_do = xenvif_get_extras(queue, extras,
 966						       &extra_count,
 967						       work_to_do);
 968			idx = queue->tx.req_cons;
 969			if (unlikely(work_to_do < 0))
 970				break;
 971		}
 972
 973		if (extras[XEN_NETIF_EXTRA_TYPE_MCAST_ADD - 1].type) {
 974			struct xen_netif_extra_info *extra;
 975
 976			extra = &extras[XEN_NETIF_EXTRA_TYPE_MCAST_ADD - 1];
 977			ret = xenvif_mcast_add(queue->vif, extra->u.mcast.addr);
 978
 979			make_tx_response(queue, &txreq, extra_count,
 980					 (ret == 0) ?
 981					 XEN_NETIF_RSP_OKAY :
 982					 XEN_NETIF_RSP_ERROR);
 
 983			continue;
 984		}
 985
 986		if (extras[XEN_NETIF_EXTRA_TYPE_MCAST_DEL - 1].type) {
 987			struct xen_netif_extra_info *extra;
 988
 989			extra = &extras[XEN_NETIF_EXTRA_TYPE_MCAST_DEL - 1];
 990			xenvif_mcast_del(queue->vif, extra->u.mcast.addr);
 991
 992			make_tx_response(queue, &txreq, extra_count,
 993					 XEN_NETIF_RSP_OKAY);
 
 994			continue;
 995		}
 996
 997		data_len = (txreq.size > XEN_NETBACK_TX_COPY_LEN) ?
 998			XEN_NETBACK_TX_COPY_LEN : txreq.size;
 999
1000		ret = xenvif_count_requests(queue, &txreq, extra_count,
1001					    txfrags, work_to_do);
1002
1003		if (unlikely(ret < 0))
1004			break;
1005
1006		idx += ret;
1007
1008		if (unlikely(txreq.size < ETH_HLEN)) {
1009			netdev_dbg(queue->vif->dev,
1010				   "Bad packet size: %d\n", txreq.size);
1011			xenvif_tx_err(queue, &txreq, extra_count, idx);
1012			break;
1013		}
1014
1015		/* No crossing a page as the payload mustn't fragment. */
1016		if (unlikely((txreq.offset + txreq.size) > XEN_PAGE_SIZE)) {
1017			netdev_err(queue->vif->dev, "Cross page boundary, txreq.offset: %u, size: %u\n",
1018				   txreq.offset, txreq.size);
 
 
1019			xenvif_fatal_tx_err(queue->vif);
1020			break;
1021		}
1022
1023		if (ret >= XEN_NETBK_LEGACY_SLOTS_MAX - 1 && data_len < txreq.size)
1024			data_len = txreq.size;
 
 
 
 
1025
1026		skb = xenvif_alloc_skb(data_len);
1027		if (unlikely(skb == NULL)) {
1028			netdev_dbg(queue->vif->dev,
1029				   "Can't allocate a skb in start_xmit.\n");
1030			xenvif_tx_err(queue, &txreq, extra_count, idx);
1031			break;
1032		}
1033
1034		skb_shinfo(skb)->nr_frags = ret;
 
 
1035		/* At this point shinfo->nr_frags is in fact the number of
1036		 * slots, which can be as large as XEN_NETBK_LEGACY_SLOTS_MAX.
1037		 */
1038		frag_overflow = 0;
1039		nskb = NULL;
1040		if (skb_shinfo(skb)->nr_frags > MAX_SKB_FRAGS) {
1041			frag_overflow = skb_shinfo(skb)->nr_frags - MAX_SKB_FRAGS;
1042			BUG_ON(frag_overflow > MAX_SKB_FRAGS);
1043			skb_shinfo(skb)->nr_frags = MAX_SKB_FRAGS;
1044			nskb = xenvif_alloc_skb(0);
1045			if (unlikely(nskb == NULL)) {
1046				skb_shinfo(skb)->nr_frags = 0;
1047				kfree_skb(skb);
1048				xenvif_tx_err(queue, &txreq, extra_count, idx);
1049				if (net_ratelimit())
1050					netdev_err(queue->vif->dev,
1051						   "Can't allocate the frag_list skb.\n");
1052				break;
1053			}
1054		}
1055
1056		if (extras[XEN_NETIF_EXTRA_TYPE_GSO - 1].type) {
1057			struct xen_netif_extra_info *gso;
1058			gso = &extras[XEN_NETIF_EXTRA_TYPE_GSO - 1];
1059
1060			if (xenvif_set_skb_gso(queue->vif, skb, gso)) {
1061				/* Failure in xenvif_set_skb_gso is fatal. */
1062				skb_shinfo(skb)->nr_frags = 0;
1063				kfree_skb(skb);
1064				kfree_skb(nskb);
1065				break;
1066			}
1067		}
1068
1069		if (extras[XEN_NETIF_EXTRA_TYPE_HASH - 1].type) {
1070			struct xen_netif_extra_info *extra;
1071			enum pkt_hash_types type = PKT_HASH_TYPE_NONE;
1072
1073			extra = &extras[XEN_NETIF_EXTRA_TYPE_HASH - 1];
1074
1075			switch (extra->u.hash.type) {
1076			case _XEN_NETIF_CTRL_HASH_TYPE_IPV4:
1077			case _XEN_NETIF_CTRL_HASH_TYPE_IPV6:
1078				type = PKT_HASH_TYPE_L3;
1079				break;
1080
1081			case _XEN_NETIF_CTRL_HASH_TYPE_IPV4_TCP:
1082			case _XEN_NETIF_CTRL_HASH_TYPE_IPV6_TCP:
1083				type = PKT_HASH_TYPE_L4;
1084				break;
1085
1086			default:
1087				break;
1088			}
1089
1090			if (type != PKT_HASH_TYPE_NONE)
1091				skb_set_hash(skb,
1092					     *(u32 *)extra->u.hash.value,
1093					     type);
1094		}
1095
1096		xenvif_get_requests(queue, skb, &txreq, txfrags, copy_ops,
1097				    map_ops, frag_overflow, nskb, extra_count,
1098				    data_len);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1099
1100		__skb_queue_tail(&queue->tx_queue, skb);
1101
1102		queue->tx.req_cons = idx;
 
 
 
 
1103	}
1104
 
1105	return;
1106}
1107
1108/* Consolidate skb with a frag_list into a brand new one with local pages on
1109 * frags. Returns 0 or -ENOMEM if can't allocate new pages.
1110 */
1111static int xenvif_handle_frag_list(struct xenvif_queue *queue, struct sk_buff *skb)
1112{
1113	unsigned int offset = skb_headlen(skb);
1114	skb_frag_t frags[MAX_SKB_FRAGS];
1115	int i, f;
1116	struct ubuf_info *uarg;
1117	struct sk_buff *nskb = skb_shinfo(skb)->frag_list;
1118
1119	queue->stats.tx_zerocopy_sent += 2;
1120	queue->stats.tx_frag_overflow++;
1121
1122	xenvif_fill_frags(queue, nskb);
1123	/* Subtract frags size, we will correct it later */
1124	skb->truesize -= skb->data_len;
1125	skb->len += nskb->len;
1126	skb->data_len += nskb->len;
1127
1128	/* create a brand new frags array and coalesce there */
1129	for (i = 0; offset < skb->len; i++) {
1130		struct page *page;
1131		unsigned int len;
1132
1133		BUG_ON(i >= MAX_SKB_FRAGS);
1134		page = alloc_page(GFP_ATOMIC);
1135		if (!page) {
1136			int j;
1137			skb->truesize += skb->data_len;
1138			for (j = 0; j < i; j++)
1139				put_page(skb_frag_page(&frags[j]));
1140			return -ENOMEM;
1141		}
1142
1143		if (offset + PAGE_SIZE < skb->len)
1144			len = PAGE_SIZE;
1145		else
1146			len = skb->len - offset;
1147		if (skb_copy_bits(skb, offset, page_address(page), len))
1148			BUG();
1149
1150		offset += len;
1151		skb_frag_fill_page_desc(&frags[i], page, 0, len);
 
 
1152	}
1153
 
 
 
 
 
1154	/* Release all the original (foreign) frags. */
1155	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
1156		skb_frag_unref(skb, f);
1157	uarg = skb_shinfo(skb)->destructor_arg;
1158	/* increase inflight counter to offset decrement in callback */
1159	atomic_inc(&queue->inflight_packets);
1160	uarg->ops->complete(NULL, uarg, true);
1161	skb_shinfo(skb)->destructor_arg = NULL;
1162
1163	/* Fill the skb with the new (local) frags. */
1164	memcpy(skb_shinfo(skb)->frags, frags, i * sizeof(skb_frag_t));
1165	skb_shinfo(skb)->nr_frags = i;
1166	skb->truesize += i * PAGE_SIZE;
1167
1168	return 0;
1169}
1170
1171static int xenvif_tx_submit(struct xenvif_queue *queue)
1172{
1173	struct gnttab_map_grant_ref *gop_map = queue->tx_map_ops;
1174	struct gnttab_copy *gop_copy = queue->tx_copy_ops;
1175	struct sk_buff *skb;
1176	int work_done = 0;
1177
1178	while ((skb = __skb_dequeue(&queue->tx_queue)) != NULL) {
1179		struct xen_netif_tx_request *txp;
1180		u16 pending_idx;
 
1181
1182		pending_idx = copy_pending_idx(skb, 0);
1183		txp = &queue->pending_tx_info[pending_idx].req;
1184
1185		/* Check the remap error code. */
1186		if (unlikely(xenvif_tx_check_gop(queue, skb, &gop_map, &gop_copy))) {
1187			/* If there was an error, xenvif_tx_check_gop is
1188			 * expected to release all the frags which were mapped,
1189			 * so kfree_skb shouldn't do it again
1190			 */
1191			skb_shinfo(skb)->nr_frags = 0;
1192			if (skb_has_frag_list(skb)) {
1193				struct sk_buff *nskb =
1194						skb_shinfo(skb)->frag_list;
1195				skb_shinfo(nskb)->nr_frags = 0;
1196			}
1197			kfree_skb(skb);
1198			continue;
1199		}
1200
 
 
 
 
 
 
 
 
 
 
 
 
1201		if (txp->flags & XEN_NETTXF_csum_blank)
1202			skb->ip_summed = CHECKSUM_PARTIAL;
1203		else if (txp->flags & XEN_NETTXF_data_validated)
1204			skb->ip_summed = CHECKSUM_UNNECESSARY;
1205
1206		xenvif_fill_frags(queue, skb);
1207
1208		if (unlikely(skb_has_frag_list(skb))) {
1209			struct sk_buff *nskb = skb_shinfo(skb)->frag_list;
1210			xenvif_skb_zerocopy_prepare(queue, nskb);
1211			if (xenvif_handle_frag_list(queue, skb)) {
1212				if (net_ratelimit())
1213					netdev_err(queue->vif->dev,
1214						   "Not enough memory to consolidate frag_list!\n");
1215				xenvif_skb_zerocopy_prepare(queue, skb);
1216				kfree_skb(skb);
1217				continue;
1218			}
1219			/* Copied all the bits from the frag list -- free it. */
1220			skb_frag_list_init(skb);
1221			kfree_skb(nskb);
1222		}
1223
1224		skb->dev      = queue->vif->dev;
1225		skb->protocol = eth_type_trans(skb, skb->dev);
1226		skb_reset_network_header(skb);
1227
1228		if (checksum_setup(queue, skb)) {
1229			netdev_dbg(queue->vif->dev,
1230				   "Can't setup checksum in net_tx_action\n");
1231			/* We have to set this flag to trigger the callback */
1232			if (skb_shinfo(skb)->destructor_arg)
1233				xenvif_skb_zerocopy_prepare(queue, skb);
1234			kfree_skb(skb);
1235			continue;
1236		}
1237
1238		skb_probe_transport_header(skb);
1239
1240		/* If the packet is GSO then we will have just set up the
1241		 * transport header offset in checksum_setup so it's now
1242		 * straightforward to calculate gso_segs.
1243		 */
1244		if (skb_is_gso(skb)) {
1245			int mss, hdrlen;
1246
1247			/* GSO implies having the L4 header. */
1248			WARN_ON_ONCE(!skb_transport_header_was_set(skb));
1249			if (unlikely(!skb_transport_header_was_set(skb))) {
1250				kfree_skb(skb);
1251				continue;
1252			}
1253
1254			mss = skb_shinfo(skb)->gso_size;
1255			hdrlen = skb_tcp_all_headers(skb);
1256
1257			skb_shinfo(skb)->gso_segs =
1258				DIV_ROUND_UP(skb->len - hdrlen, mss);
1259		}
1260
1261		queue->stats.rx_bytes += skb->len;
1262		queue->stats.rx_packets++;
1263
1264		work_done++;
1265
1266		/* Set this flag right before netif_receive_skb, otherwise
1267		 * someone might think this packet already left netback, and
1268		 * do a skb_copy_ubufs while we are still in control of the
1269		 * skb. E.g. the __pskb_pull_tail earlier can do such thing.
1270		 */
1271		if (skb_shinfo(skb)->destructor_arg) {
1272			xenvif_skb_zerocopy_prepare(queue, skb);
1273			queue->stats.tx_zerocopy_sent++;
1274		}
1275
1276		netif_receive_skb(skb);
1277	}
1278
1279	return work_done;
1280}
1281
1282static void xenvif_zerocopy_callback(struct sk_buff *skb,
1283				     struct ubuf_info *ubuf_base,
1284				     bool zerocopy_success)
1285{
1286	unsigned long flags;
1287	pending_ring_idx_t index;
1288	struct ubuf_info_msgzc *ubuf = uarg_to_msgzc(ubuf_base);
1289	struct xenvif_queue *queue = ubuf_to_queue(ubuf);
1290
1291	/* This is the only place where we grab this lock, to protect callbacks
1292	 * from each other.
1293	 */
1294	spin_lock_irqsave(&queue->callback_lock, flags);
1295	do {
1296		u16 pending_idx = ubuf->desc;
1297		ubuf = (struct ubuf_info_msgzc *) ubuf->ctx;
1298		BUG_ON(queue->dealloc_prod - queue->dealloc_cons >=
1299			MAX_PENDING_REQS);
1300		index = pending_index(queue->dealloc_prod);
1301		queue->dealloc_ring[index] = pending_idx;
1302		/* Sync with xenvif_tx_dealloc_action:
1303		 * insert idx then incr producer.
1304		 */
1305		smp_wmb();
1306		queue->dealloc_prod++;
1307	} while (ubuf);
1308	spin_unlock_irqrestore(&queue->callback_lock, flags);
1309
1310	if (likely(zerocopy_success))
1311		queue->stats.tx_zerocopy_success++;
1312	else
1313		queue->stats.tx_zerocopy_fail++;
1314	xenvif_skb_zerocopy_complete(queue);
1315}
1316
1317const struct ubuf_info_ops xenvif_ubuf_ops = {
1318	.complete = xenvif_zerocopy_callback,
1319};
1320
1321static inline void xenvif_tx_dealloc_action(struct xenvif_queue *queue)
1322{
1323	struct gnttab_unmap_grant_ref *gop;
1324	pending_ring_idx_t dc, dp;
1325	u16 pending_idx, pending_idx_release[MAX_PENDING_REQS];
1326	unsigned int i = 0;
1327
1328	dc = queue->dealloc_cons;
1329	gop = queue->tx_unmap_ops;
1330
1331	/* Free up any grants we have finished using */
1332	do {
1333		dp = queue->dealloc_prod;
1334
1335		/* Ensure we see all indices enqueued by all
1336		 * xenvif_zerocopy_callback().
1337		 */
1338		smp_rmb();
1339
1340		while (dc != dp) {
1341			BUG_ON(gop - queue->tx_unmap_ops >= MAX_PENDING_REQS);
1342			pending_idx =
1343				queue->dealloc_ring[pending_index(dc++)];
1344
1345			pending_idx_release[gop - queue->tx_unmap_ops] =
1346				pending_idx;
1347			queue->pages_to_unmap[gop - queue->tx_unmap_ops] =
1348				queue->mmap_pages[pending_idx];
1349			gnttab_set_unmap_op(gop,
1350					    idx_to_kaddr(queue, pending_idx),
1351					    GNTMAP_host_map,
1352					    queue->grant_tx_handle[pending_idx]);
1353			xenvif_grant_handle_reset(queue, pending_idx);
1354			++gop;
1355		}
1356
1357	} while (dp != queue->dealloc_prod);
1358
1359	queue->dealloc_cons = dc;
1360
1361	if (gop - queue->tx_unmap_ops > 0) {
1362		int ret;
1363		ret = gnttab_unmap_refs(queue->tx_unmap_ops,
1364					NULL,
1365					queue->pages_to_unmap,
1366					gop - queue->tx_unmap_ops);
1367		if (ret) {
1368			netdev_err(queue->vif->dev, "Unmap fail: nr_ops %tu ret %d\n",
1369				   gop - queue->tx_unmap_ops, ret);
1370			for (i = 0; i < gop - queue->tx_unmap_ops; ++i) {
1371				if (gop[i].status != GNTST_okay)
1372					netdev_err(queue->vif->dev,
1373						   " host_addr: 0x%llx handle: 0x%x status: %d\n",
1374						   gop[i].host_addr,
1375						   gop[i].handle,
1376						   gop[i].status);
1377			}
1378			BUG();
1379		}
1380	}
1381
1382	for (i = 0; i < gop - queue->tx_unmap_ops; ++i)
1383		xenvif_idx_release(queue, pending_idx_release[i],
1384				   XEN_NETIF_RSP_OKAY);
1385}
1386
1387
1388/* Called after netfront has transmitted */
1389int xenvif_tx_action(struct xenvif_queue *queue, int budget)
1390{
1391	unsigned nr_mops = 0, nr_cops = 0;
1392	int work_done, ret;
1393
1394	if (unlikely(!tx_work_todo(queue)))
1395		return 0;
1396
1397	xenvif_tx_build_gops(queue, budget, &nr_cops, &nr_mops);
1398
1399	if (nr_cops == 0)
1400		return 0;
1401
1402	gnttab_batch_copy(queue->tx_copy_ops, nr_cops);
1403	if (nr_mops != 0) {
1404		ret = gnttab_map_refs(queue->tx_map_ops,
1405				      NULL,
1406				      queue->pages_to_map,
1407				      nr_mops);
1408		if (ret) {
1409			unsigned int i;
1410
1411			netdev_err(queue->vif->dev, "Map fail: nr %u ret %d\n",
1412				   nr_mops, ret);
1413			for (i = 0; i < nr_mops; ++i)
1414				WARN_ON_ONCE(queue->tx_map_ops[i].status ==
1415				             GNTST_okay);
1416		}
1417	}
1418
1419	work_done = xenvif_tx_submit(queue);
1420
1421	return work_done;
1422}
1423
1424static void _make_tx_response(struct xenvif_queue *queue,
1425			     const struct xen_netif_tx_request *txp,
1426			     unsigned int extra_count,
1427			     s8 status)
1428{
1429	RING_IDX i = queue->tx.rsp_prod_pvt;
1430	struct xen_netif_tx_response *resp;
1431
1432	resp = RING_GET_RESPONSE(&queue->tx, i);
1433	resp->id     = txp->id;
1434	resp->status = status;
1435
1436	while (extra_count-- != 0)
1437		RING_GET_RESPONSE(&queue->tx, ++i)->status = XEN_NETIF_RSP_NULL;
1438
1439	queue->tx.rsp_prod_pvt = ++i;
1440}
1441
1442static void push_tx_responses(struct xenvif_queue *queue)
1443{
1444	int notify;
1445
1446	RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&queue->tx, notify);
1447	if (notify)
1448		notify_remote_via_irq(queue->tx_irq);
1449}
1450
1451static void xenvif_idx_release(struct xenvif_queue *queue, u16 pending_idx,
1452			       s8 status)
1453{
1454	struct pending_tx_info *pending_tx_info;
1455	pending_ring_idx_t index;
1456	unsigned long flags;
1457
1458	pending_tx_info = &queue->pending_tx_info[pending_idx];
1459
1460	spin_lock_irqsave(&queue->response_lock, flags);
1461
1462	_make_tx_response(queue, &pending_tx_info->req,
1463			  pending_tx_info->extra_count, status);
1464
1465	/* Release the pending index before pusing the Tx response so
1466	 * its available before a new Tx request is pushed by the
1467	 * frontend.
1468	 */
1469	index = pending_index(queue->pending_prod++);
1470	queue->pending_ring[index] = pending_idx;
1471
1472	push_tx_responses(queue);
1473
1474	spin_unlock_irqrestore(&queue->response_lock, flags);
1475}
1476
 
1477static void make_tx_response(struct xenvif_queue *queue,
1478			     const struct xen_netif_tx_request *txp,
1479			     unsigned int extra_count,
1480			     s8 status)
1481{
1482	unsigned long flags;
 
1483
1484	spin_lock_irqsave(&queue->response_lock, flags);
 
 
1485
1486	_make_tx_response(queue, txp, extra_count, status);
1487	push_tx_responses(queue);
1488
1489	spin_unlock_irqrestore(&queue->response_lock, flags);
1490}
1491
1492static void xenvif_idx_unmap(struct xenvif_queue *queue, u16 pending_idx)
 
 
 
 
 
 
 
 
 
1493{
1494	int ret;
1495	struct gnttab_unmap_grant_ref tx_unmap_op;
1496
1497	gnttab_set_unmap_op(&tx_unmap_op,
1498			    idx_to_kaddr(queue, pending_idx),
1499			    GNTMAP_host_map,
1500			    queue->grant_tx_handle[pending_idx]);
1501	xenvif_grant_handle_reset(queue, pending_idx);
1502
1503	ret = gnttab_unmap_refs(&tx_unmap_op, NULL,
1504				&queue->mmap_pages[pending_idx], 1);
1505	if (ret) {
1506		netdev_err(queue->vif->dev,
1507			   "Unmap fail: ret: %d pending_idx: %d host_addr: %llx handle: 0x%x status: %d\n",
1508			   ret,
1509			   pending_idx,
1510			   tx_unmap_op.host_addr,
1511			   tx_unmap_op.handle,
1512			   tx_unmap_op.status);
1513		BUG();
1514	}
1515}
1516
1517static inline int tx_work_todo(struct xenvif_queue *queue)
1518{
1519	if (likely(RING_HAS_UNCONSUMED_REQUESTS(&queue->tx)))
1520		return 1;
1521
1522	return 0;
1523}
1524
1525static inline bool tx_dealloc_work_todo(struct xenvif_queue *queue)
1526{
1527	return queue->dealloc_cons != queue->dealloc_prod;
1528}
1529
1530void xenvif_unmap_frontend_data_rings(struct xenvif_queue *queue)
1531{
1532	if (queue->tx.sring)
1533		xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(queue->vif),
1534					queue->tx.sring);
1535	if (queue->rx.sring)
1536		xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(queue->vif),
1537					queue->rx.sring);
1538}
1539
1540int xenvif_map_frontend_data_rings(struct xenvif_queue *queue,
1541				   grant_ref_t tx_ring_ref,
1542				   grant_ref_t rx_ring_ref)
1543{
1544	void *addr;
1545	struct xen_netif_tx_sring *txs;
1546	struct xen_netif_rx_sring *rxs;
1547	RING_IDX rsp_prod, req_prod;
1548	int err;
1549
1550	err = xenbus_map_ring_valloc(xenvif_to_xenbus_device(queue->vif),
1551				     &tx_ring_ref, 1, &addr);
1552	if (err)
1553		goto err;
1554
1555	txs = (struct xen_netif_tx_sring *)addr;
1556	rsp_prod = READ_ONCE(txs->rsp_prod);
1557	req_prod = READ_ONCE(txs->req_prod);
1558
1559	BACK_RING_ATTACH(&queue->tx, txs, rsp_prod, XEN_PAGE_SIZE);
1560
1561	err = -EIO;
1562	if (req_prod - rsp_prod > RING_SIZE(&queue->tx))
1563		goto err;
1564
1565	err = xenbus_map_ring_valloc(xenvif_to_xenbus_device(queue->vif),
1566				     &rx_ring_ref, 1, &addr);
1567	if (err)
1568		goto err;
1569
1570	rxs = (struct xen_netif_rx_sring *)addr;
1571	rsp_prod = READ_ONCE(rxs->rsp_prod);
1572	req_prod = READ_ONCE(rxs->req_prod);
1573
1574	BACK_RING_ATTACH(&queue->rx, rxs, rsp_prod, XEN_PAGE_SIZE);
1575
1576	err = -EIO;
1577	if (req_prod - rsp_prod > RING_SIZE(&queue->rx))
1578		goto err;
1579
1580	return 0;
1581
1582err:
1583	xenvif_unmap_frontend_data_rings(queue);
1584	return err;
1585}
1586
1587static bool xenvif_dealloc_kthread_should_stop(struct xenvif_queue *queue)
1588{
1589	/* Dealloc thread must remain running until all inflight
1590	 * packets complete.
1591	 */
1592	return kthread_should_stop() &&
1593		!atomic_read(&queue->inflight_packets);
1594}
1595
1596int xenvif_dealloc_kthread(void *data)
1597{
1598	struct xenvif_queue *queue = data;
1599
1600	for (;;) {
1601		wait_event_interruptible(queue->dealloc_wq,
1602					 tx_dealloc_work_todo(queue) ||
1603					 xenvif_dealloc_kthread_should_stop(queue));
1604		if (xenvif_dealloc_kthread_should_stop(queue))
1605			break;
1606
1607		xenvif_tx_dealloc_action(queue);
1608		cond_resched();
1609	}
1610
1611	/* Unmap anything remaining*/
1612	if (tx_dealloc_work_todo(queue))
1613		xenvif_tx_dealloc_action(queue);
1614
1615	return 0;
1616}
1617
1618static void make_ctrl_response(struct xenvif *vif,
1619			       const struct xen_netif_ctrl_request *req,
1620			       u32 status, u32 data)
1621{
1622	RING_IDX idx = vif->ctrl.rsp_prod_pvt;
1623	struct xen_netif_ctrl_response rsp = {
1624		.id = req->id,
1625		.type = req->type,
1626		.status = status,
1627		.data = data,
1628	};
1629
1630	*RING_GET_RESPONSE(&vif->ctrl, idx) = rsp;
1631	vif->ctrl.rsp_prod_pvt = ++idx;
1632}
1633
1634static void push_ctrl_response(struct xenvif *vif)
1635{
1636	int notify;
1637
1638	RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&vif->ctrl, notify);
1639	if (notify)
1640		notify_remote_via_irq(vif->ctrl_irq);
1641}
1642
1643static void process_ctrl_request(struct xenvif *vif,
1644				 const struct xen_netif_ctrl_request *req)
1645{
1646	u32 status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED;
1647	u32 data = 0;
1648
1649	switch (req->type) {
1650	case XEN_NETIF_CTRL_TYPE_SET_HASH_ALGORITHM:
1651		status = xenvif_set_hash_alg(vif, req->data[0]);
1652		break;
1653
1654	case XEN_NETIF_CTRL_TYPE_GET_HASH_FLAGS:
1655		status = xenvif_get_hash_flags(vif, &data);
1656		break;
1657
1658	case XEN_NETIF_CTRL_TYPE_SET_HASH_FLAGS:
1659		status = xenvif_set_hash_flags(vif, req->data[0]);
1660		break;
1661
1662	case XEN_NETIF_CTRL_TYPE_SET_HASH_KEY:
1663		status = xenvif_set_hash_key(vif, req->data[0],
1664					     req->data[1]);
1665		break;
1666
1667	case XEN_NETIF_CTRL_TYPE_GET_HASH_MAPPING_SIZE:
1668		status = XEN_NETIF_CTRL_STATUS_SUCCESS;
1669		data = XEN_NETBK_MAX_HASH_MAPPING_SIZE;
1670		break;
1671
1672	case XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING_SIZE:
1673		status = xenvif_set_hash_mapping_size(vif,
1674						      req->data[0]);
1675		break;
1676
1677	case XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING:
1678		status = xenvif_set_hash_mapping(vif, req->data[0],
1679						 req->data[1],
1680						 req->data[2]);
1681		break;
1682
1683	default:
1684		break;
1685	}
1686
1687	make_ctrl_response(vif, req, status, data);
1688	push_ctrl_response(vif);
1689}
1690
1691static void xenvif_ctrl_action(struct xenvif *vif)
1692{
1693	for (;;) {
1694		RING_IDX req_prod, req_cons;
1695
1696		req_prod = vif->ctrl.sring->req_prod;
1697		req_cons = vif->ctrl.req_cons;
1698
1699		/* Make sure we can see requests before we process them. */
1700		rmb();
1701
1702		if (req_cons == req_prod)
1703			break;
1704
1705		while (req_cons != req_prod) {
1706			struct xen_netif_ctrl_request req;
1707
1708			RING_COPY_REQUEST(&vif->ctrl, req_cons, &req);
1709			req_cons++;
1710
1711			process_ctrl_request(vif, &req);
1712		}
1713
1714		vif->ctrl.req_cons = req_cons;
1715		vif->ctrl.sring->req_event = req_cons + 1;
1716	}
1717}
1718
1719static bool xenvif_ctrl_work_todo(struct xenvif *vif)
1720{
1721	if (likely(RING_HAS_UNCONSUMED_REQUESTS(&vif->ctrl)))
1722		return true;
1723
1724	return false;
1725}
1726
1727irqreturn_t xenvif_ctrl_irq_fn(int irq, void *data)
1728{
1729	struct xenvif *vif = data;
1730	unsigned int eoi_flag = XEN_EOI_FLAG_SPURIOUS;
1731
1732	while (xenvif_ctrl_work_todo(vif)) {
1733		xenvif_ctrl_action(vif);
1734		eoi_flag = 0;
1735	}
1736
1737	xen_irq_lateeoi(irq, eoi_flag);
1738
1739	return IRQ_HANDLED;
1740}
1741
1742static int __init netback_init(void)
1743{
1744	int rc = 0;
1745
1746	if (!xen_domain())
1747		return -ENODEV;
1748
1749	/* Allow as many queues as there are CPUs but max. 8 if user has not
1750	 * specified a value.
1751	 */
1752	if (xenvif_max_queues == 0)
1753		xenvif_max_queues = min_t(unsigned int, MAX_QUEUES_DEFAULT,
1754					  num_online_cpus());
1755
1756	if (fatal_skb_slots < XEN_NETBK_LEGACY_SLOTS_MAX) {
1757		pr_info("fatal_skb_slots too small (%d), bump it to XEN_NETBK_LEGACY_SLOTS_MAX (%d)\n",
1758			fatal_skb_slots, XEN_NETBK_LEGACY_SLOTS_MAX);
1759		fatal_skb_slots = XEN_NETBK_LEGACY_SLOTS_MAX;
1760	}
1761
1762	rc = xenvif_xenbus_init();
1763	if (rc)
1764		goto failed_init;
1765
1766#ifdef CONFIG_DEBUG_FS
1767	xen_netback_dbg_root = debugfs_create_dir("xen-netback", NULL);
 
 
 
1768#endif /* CONFIG_DEBUG_FS */
1769
1770	return 0;
1771
1772failed_init:
1773	return rc;
1774}
1775
1776module_init(netback_init);
1777
1778static void __exit netback_fini(void)
1779{
1780#ifdef CONFIG_DEBUG_FS
1781	debugfs_remove_recursive(xen_netback_dbg_root);
 
1782#endif /* CONFIG_DEBUG_FS */
1783	xenvif_xenbus_fini();
1784}
1785module_exit(netback_fini);
1786
1787MODULE_DESCRIPTION("Xen backend network device module");
1788MODULE_LICENSE("Dual BSD/GPL");
1789MODULE_ALIAS("xen-backend:vif");