drivers/net/xen-netback/netback.c

   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
  42 #include <net/tcp.h>
  43
  44 #include <xen/xen.h>
  45 #include <xen/events.h>
  46 #include <xen/interface/memory.h>
  47
  48 #include <asm/xen/hypercall.h>
  49 #include <asm/xen/page.h>
  50
  51 /* Provide an option to disable split event channels at load time as
  52  * event channels are limited resource. Split event channels are
  53  * enabled by default.
  54  */
  55 bool separate_tx_rx_irq = 1;
  56 module_param(separate_tx_rx_irq, bool, 0644);
  57
  58 /* The time that packets can stay on the guest Rx internal queue
  59  * before they are dropped.
  60  */
  61 unsigned int rx_drain_timeout_msecs = 10000;
  62 module_param(rx_drain_timeout_msecs, uint, 0444);
  63
  64 /* The length of time before the frontend is considered unresponsive
  65  * because it isn't providing Rx slots.
  66  */
  67 unsigned int rx_stall_timeout_msecs = 60000;
  68 module_param(rx_stall_timeout_msecs, uint, 0444);
  69
  70 unsigned int xenvif_max_queues;
  71 module_param_named(max_queues, xenvif_max_queues, uint, 0644);
  72 MODULE_PARM_DESC(max_queues,
  73                  "Maximum number of queues per virtual interface");
  74
  75 /*
  76  * This is the maximum slots a skb can have. If a guest sends a skb
  77  * which exceeds this limit it is considered malicious.
  78  */
  79 #define FATAL_SKB_SLOTS_DEFAULT 20
  80 static unsigned int fatal_skb_slots = FATAL_SKB_SLOTS_DEFAULT;
  81 module_param(fatal_skb_slots, uint, 0444);
  82
  83 /* The amount to copy out of the first guest Tx slot into the skb's
  84  * linear area.  If the first slot has more data, it will be mapped
  85  * and put into the first frag.
  86  *
  87  * This is sized to avoid pulling headers from the frags for most
  88  * TCP/IP packets.
  89  */
  90 #define XEN_NETBACK_TX_COPY_LEN 128
  91
  92
  93 static void xenvif_idx_release(struct xenvif_queue *queue, u16 pending_idx,
  94                                u8 status);
  95
  96 static void make_tx_response(struct xenvif_queue *queue,
  97                              struct xen_netif_tx_request *txp,
  98                              s8       st);
  99
 100 static inline int tx_work_todo(struct xenvif_queue *queue);
 101
 102 static struct xen_netif_rx_response *make_rx_response(struct xenvif_queue *queue,
 103                                              u16      id,
 104                                              s8       st,
 105                                              u16      offset,
 106                                              u16      size,
 107                                              u16      flags);
 108
 109 static inline unsigned long idx_to_pfn(struct xenvif_queue *queue,
 110                                        u16 idx)
 111 {
 112         return page_to_pfn(queue->mmap_pages[idx]);
 113 }
 114
 115 static inline unsigned long idx_to_kaddr(struct xenvif_queue *queue,
 116                                          u16 idx)
 117 {
 118         return (unsigned long)pfn_to_kaddr(idx_to_pfn(queue, idx));
 119 }
 120
 121 #define callback_param(vif, pending_idx) \
 122         (vif->pending_tx_info[pending_idx].callback_struct)
 123
 124 /* Find the containing VIF's structure from a pointer in pending_tx_info array
 125  */
 126 static inline struct xenvif_queue *ubuf_to_queue(const struct ubuf_info *ubuf)
 127 {
 128         u16 pending_idx = ubuf->desc;
 129         struct pending_tx_info *temp =
 130                 container_of(ubuf, struct pending_tx_info, callback_struct);
 131         return container_of(temp - pending_idx,
 132                             struct xenvif_queue,
 133                             pending_tx_info[0]);
 134 }
 135
 136 static u16 frag_get_pending_idx(skb_frag_t *frag)
 137 {
 138         return (u16)frag->page_offset;
 139 }
 140
 141 static void frag_set_pending_idx(skb_frag_t *frag, u16 pending_idx)
 142 {
 143         frag->page_offset = pending_idx;
 144 }
 145
 146 static inline pending_ring_idx_t pending_index(unsigned i)
 147 {
 148         return i & (MAX_PENDING_REQS-1);
 149 }
 150
 151 bool xenvif_rx_ring_slots_available(struct xenvif_queue *queue, int needed)
 152 {
 153         RING_IDX prod, cons;
 154
 155         do {
 156                 prod = queue->rx.sring->req_prod;
 157                 cons = queue->rx.req_cons;
 158
 159                 if (prod - cons >= needed)
 160                         return true;
 161
 162                 queue->rx.sring->req_event = prod + 1;
 163
 164                 /* Make sure event is visible before we check prod
 165                  * again.
 166                  */
 167                 mb();
 168         } while (queue->rx.sring->req_prod != prod);
 169
 170         return false;
 171 }
 172
 173 void xenvif_rx_queue_tail(struct xenvif_queue *queue, struct sk_buff *skb)
 174 {
 175         unsigned long flags;
 176
 177         spin_lock_irqsave(&queue->rx_queue.lock, flags);
 178
 179         __skb_queue_tail(&queue->rx_queue, skb);
 180
 181         queue->rx_queue_len += skb->len;
 182         if (queue->rx_queue_len > queue->rx_queue_max)
 183                 netif_tx_stop_queue(netdev_get_tx_queue(queue->vif->dev, queue->id));
 184
 185         spin_unlock_irqrestore(&queue->rx_queue.lock, flags);
 186 }
 187
 188 static struct sk_buff *xenvif_rx_dequeue(struct xenvif_queue *queue)
 189 {
 190         struct sk_buff *skb;
 191
 192         spin_lock_irq(&queue->rx_queue.lock);
 193
 194         skb = __skb_dequeue(&queue->rx_queue);
 195         if (skb)
 196                 queue->rx_queue_len -= skb->len;
 197
 198         spin_unlock_irq(&queue->rx_queue.lock);
 199
 200         return skb;
 201 }
 202
 203 static void xenvif_rx_queue_maybe_wake(struct xenvif_queue *queue)
 204 {
 205         spin_lock_irq(&queue->rx_queue.lock);
 206
 207         if (queue->rx_queue_len < queue->rx_queue_max)
 208                 netif_tx_wake_queue(netdev_get_tx_queue(queue->vif->dev, queue->id));
 209
 210         spin_unlock_irq(&queue->rx_queue.lock);
 211 }
 212
 213
 214 static void xenvif_rx_queue_purge(struct xenvif_queue *queue)
 215 {
 216         struct sk_buff *skb;
 217         while ((skb = xenvif_rx_dequeue(queue)) != NULL)
 218                 kfree_skb(skb);
 219 }
 220
 221 static void xenvif_rx_queue_drop_expired(struct xenvif_queue *queue)
 222 {
 223         struct sk_buff *skb;
 224
 225         for(;;) {
 226                 skb = skb_peek(&queue->rx_queue);
 227                 if (!skb)
 228                         break;
 229                 if (time_before(jiffies, XENVIF_RX_CB(skb)->expires))
 230                         break;
 231                 xenvif_rx_dequeue(queue);
 232                 kfree_skb(skb);
 233         }
 234 }
 235
 236 struct netrx_pending_operations {
 237         unsigned copy_prod, copy_cons;
 238         unsigned meta_prod, meta_cons;
 239         struct gnttab_copy *copy;
 240         struct xenvif_rx_meta *meta;
 241         int copy_off;
 242         grant_ref_t copy_gref;
 243 };
 244
 245 static struct xenvif_rx_meta *get_next_rx_buffer(struct xenvif_queue *queue,
 246                                                  struct netrx_pending_operations *npo)
 247 {
 248         struct xenvif_rx_meta *meta;
 249         struct xen_netif_rx_request *req;
 250
 251         req = RING_GET_REQUEST(&queue->rx, queue->rx.req_cons++);
 252
 253         meta = npo->meta + npo->meta_prod++;
 254         meta->gso_type = XEN_NETIF_GSO_TYPE_NONE;
 255         meta->gso_size = 0;
 256         meta->size = 0;
 257         meta->id = req->id;
 258
 259         npo->copy_off = 0;
 260         npo->copy_gref = req->gref;
 261
 262         return meta;
 263 }
 264
 265 /*
 266  * Set up the grant operations for this fragment. If it's a flipping
 267  * interface, we also set up the unmap request from here.
 268  */
 269 static void xenvif_gop_frag_copy(struct xenvif_queue *queue, struct sk_buff *skb,
 270                                  struct netrx_pending_operations *npo,
 271                                  struct page *page, unsigned long size,
 272                                  unsigned long offset, int *head)
 273 {
 274         struct gnttab_copy *copy_gop;
 275         struct xenvif_rx_meta *meta;
 276         unsigned long bytes;
 277         int gso_type = XEN_NETIF_GSO_TYPE_NONE;
 278
 279         /* Data must not cross a page boundary. */
 280         BUG_ON(size + offset > PAGE_SIZE<<compound_order(page));
 281
 282         meta = npo->meta + npo->meta_prod - 1;
 283
 284         /* Skip unused frames from start of page */
 285         page += offset >> PAGE_SHIFT;
 286         offset &= ~PAGE_MASK;
 287
 288         while (size > 0) {
 289                 struct xen_page_foreign *foreign;
 290
 291                 BUG_ON(offset >= PAGE_SIZE);
 292                 BUG_ON(npo->copy_off > MAX_BUFFER_OFFSET);
 293
 294                 if (npo->copy_off == MAX_BUFFER_OFFSET)
 295                         meta = get_next_rx_buffer(queue, npo);
 296
 297                 bytes = PAGE_SIZE - offset;
 298                 if (bytes > size)
 299                         bytes = size;
 300
 301                 if (npo->copy_off + bytes > MAX_BUFFER_OFFSET)
 302                         bytes = MAX_BUFFER_OFFSET - npo->copy_off;
 303
 304                 copy_gop = npo->copy + npo->copy_prod++;
 305                 copy_gop->flags = GNTCOPY_dest_gref;
 306                 copy_gop->len = bytes;
 307
 308                 foreign = xen_page_foreign(page);
 309                 if (foreign) {
 310                         copy_gop->source.domid = foreign->domid;
 311                         copy_gop->source.u.ref = foreign->gref;
 312                         copy_gop->flags |= GNTCOPY_source_gref;
 313                 } else {
 314                         copy_gop->source.domid = DOMID_SELF;
 315                         copy_gop->source.u.gmfn =
 316                                 virt_to_mfn(page_address(page));
 317                 }
 318                 copy_gop->source.offset = offset;
 319
 320                 copy_gop->dest.domid = queue->vif->domid;
 321                 copy_gop->dest.offset = npo->copy_off;
 322                 copy_gop->dest.u.ref = npo->copy_gref;
 323
 324                 npo->copy_off += bytes;
 325                 meta->size += bytes;
 326
 327                 offset += bytes;
 328                 size -= bytes;
 329
 330                 /* Next frame */
 331                 if (offset == PAGE_SIZE && size) {
 332                         BUG_ON(!PageCompound(page));
 333                         page++;
 334                         offset = 0;
 335                 }
 336
 337                 /* Leave a gap for the GSO descriptor. */
 338                 if (skb_is_gso(skb)) {
 339                         if (skb_shinfo(skb)->gso_type & SKB_GSO_TCPV4)
 340                                 gso_type = XEN_NETIF_GSO_TYPE_TCPV4;
 341                         else if (skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6)
 342                                 gso_type = XEN_NETIF_GSO_TYPE_TCPV6;
 343                 }
 344
 345                 if (*head && ((1 << gso_type) & queue->vif->gso_mask))
 346                         queue->rx.req_cons++;
 347
 348                 *head = 0; /* There must be something in this buffer now. */
 349
 350         }
 351 }
 352
 353 /*
 354  * Prepare an SKB to be transmitted to the frontend.
 355  *
 356  * This function is responsible for allocating grant operations, meta
 357  * structures, etc.
 358  *
 359  * It returns the number of meta structures consumed. The number of
 360  * ring slots used is always equal to the number of meta slots used
 361  * plus the number of GSO descriptors used. Currently, we use either
 362  * zero GSO descriptors (for non-GSO packets) or one descriptor (for
 363  * frontend-side LRO).
 364  */
 365 static int xenvif_gop_skb(struct sk_buff *skb,
 366                           struct netrx_pending_operations *npo,
 367                           struct xenvif_queue *queue)
 368 {
 369         struct xenvif *vif = netdev_priv(skb->dev);
 370         int nr_frags = skb_shinfo(skb)->nr_frags;
 371         int i;
 372         struct xen_netif_rx_request *req;
 373         struct xenvif_rx_meta *meta;
 374         unsigned char *data;
 375         int head = 1;
 376         int old_meta_prod;
 377         int gso_type;
 378
 379         old_meta_prod = npo->meta_prod;
 380
 381         gso_type = XEN_NETIF_GSO_TYPE_NONE;
 382         if (skb_is_gso(skb)) {
 383                 if (skb_shinfo(skb)->gso_type & SKB_GSO_TCPV4)
 384                         gso_type = XEN_NETIF_GSO_TYPE_TCPV4;
 385                 else if (skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6)
 386                         gso_type = XEN_NETIF_GSO_TYPE_TCPV6;
 387         }
 388
 389         /* Set up a GSO prefix descriptor, if necessary */
 390         if ((1 << gso_type) & vif->gso_prefix_mask) {
 391                 req = RING_GET_REQUEST(&queue->rx, queue->rx.req_cons++);
 392                 meta = npo->meta + npo->meta_prod++;
 393                 meta->gso_type = gso_type;
 394                 meta->gso_size = skb_shinfo(skb)->gso_size;
 395                 meta->size = 0;
 396                 meta->id = req->id;
 397         }
 398
 399         req = RING_GET_REQUEST(&queue->rx, queue->rx.req_cons++);
 400         meta = npo->meta + npo->meta_prod++;
 401
 402         if ((1 << gso_type) & vif->gso_mask) {
 403                 meta->gso_type = gso_type;
 404                 meta->gso_size = skb_shinfo(skb)->gso_size;
 405         } else {
 406                 meta->gso_type = XEN_NETIF_GSO_TYPE_NONE;
 407                 meta->gso_size = 0;
 408         }
 409
 410         meta->size = 0;
 411         meta->id = req->id;
 412         npo->copy_off = 0;
 413         npo->copy_gref = req->gref;
 414
 415         data = skb->data;
 416         while (data < skb_tail_pointer(skb)) {
 417                 unsigned int offset = offset_in_page(data);
 418                 unsigned int len = PAGE_SIZE - offset;
 419
 420                 if (data + len > skb_tail_pointer(skb))
 421                         len = skb_tail_pointer(skb) - data;
 422
 423                 xenvif_gop_frag_copy(queue, skb, npo,
 424                                      virt_to_page(data), len, offset, &head);
 425                 data += len;
 426         }
 427
 428         for (i = 0; i < nr_frags; i++) {
 429                 xenvif_gop_frag_copy(queue, skb, npo,
 430                                      skb_frag_page(&skb_shinfo(skb)->frags[i]),
 431                                      skb_frag_size(&skb_shinfo(skb)->frags[i]),
 432                                      skb_shinfo(skb)->frags[i].page_offset,
 433                                      &head);
 434         }
 435
 436         return npo->meta_prod - old_meta_prod;
 437 }
 438
 439 /*
 440  * This is a twin to xenvif_gop_skb.  Assume that xenvif_gop_skb was
 441  * used to set up the operations on the top of
 442  * netrx_pending_operations, which have since been done.  Check that
 443  * they didn't give any errors and advance over them.
 444  */
 445 static int xenvif_check_gop(struct xenvif *vif, int nr_meta_slots,
 446                             struct netrx_pending_operations *npo)
 447 {
 448         struct gnttab_copy     *copy_op;
 449         int status = XEN_NETIF_RSP_OKAY;
 450         int i;
 451
 452         for (i = 0; i < nr_meta_slots; i++) {
 453                 copy_op = npo->copy + npo->copy_cons++;
 454                 if (copy_op->status != GNTST_okay) {
 455                         netdev_dbg(vif->dev,
 456                                    "Bad status %d from copy to DOM%d.\n",
 457                                    copy_op->status, vif->domid);
 458                         status = XEN_NETIF_RSP_ERROR;
 459                 }
 460         }
 461
 462         return status;
 463 }
 464
 465 static void xenvif_add_frag_responses(struct xenvif_queue *queue, int status,
 466                                       struct xenvif_rx_meta *meta,
 467                                       int nr_meta_slots)
 468 {
 469         int i;
 470         unsigned long offset;
 471
 472         /* No fragments used */
 473         if (nr_meta_slots <= 1)
 474                 return;
 475
 476         nr_meta_slots--;
 477
 478         for (i = 0; i < nr_meta_slots; i++) {
 479                 int flags;
 480                 if (i == nr_meta_slots - 1)
 481                         flags = 0;
 482                 else
 483                         flags = XEN_NETRXF_more_data;
 484
 485                 offset = 0;
 486                 make_rx_response(queue, meta[i].id, status, offset,
 487                                  meta[i].size, flags);
 488         }
 489 }
 490
 491 void xenvif_kick_thread(struct xenvif_queue *queue)
 492 {
 493         wake_up(&queue->wq);
 494 }
 495
 496 static void xenvif_rx_action(struct xenvif_queue *queue)
 497 {
 498         s8 status;
 499         u16 flags;
 500         struct xen_netif_rx_response *resp;
 501         struct sk_buff_head rxq;
 502         struct sk_buff *skb;
 503         LIST_HEAD(notify);
 504         int ret;
 505         unsigned long offset;
 506         bool need_to_notify = false;
 507
 508         struct netrx_pending_operations npo = {
 509                 .copy  = queue->grant_copy_op,
 510                 .meta  = queue->meta,
 511         };
 512
 513         skb_queue_head_init(&rxq);
 514
 515         while (xenvif_rx_ring_slots_available(queue, XEN_NETBK_RX_SLOTS_MAX)
 516                && (skb = xenvif_rx_dequeue(queue)) != NULL) {
 517                 RING_IDX old_req_cons;
 518                 RING_IDX ring_slots_used;
 519
 520                 queue->last_rx_time = jiffies;
 521
 522                 old_req_cons = queue->rx.req_cons;
 523                 XENVIF_RX_CB(skb)->meta_slots_used = xenvif_gop_skb(skb, &npo, queue);
 524                 ring_slots_used = queue->rx.req_cons - old_req_cons;
 525
 526                 __skb_queue_tail(&rxq, skb);
 527         }
 528
 529         BUG_ON(npo.meta_prod > ARRAY_SIZE(queue->meta));
 530
 531         if (!npo.copy_prod)
 532                 goto done;
 533
 534         BUG_ON(npo.copy_prod > MAX_GRANT_COPY_OPS);
 535         gnttab_batch_copy(queue->grant_copy_op, npo.copy_prod);
 536
 537         while ((skb = __skb_dequeue(&rxq)) != NULL) {
 538
 539                 if ((1 << queue->meta[npo.meta_cons].gso_type) &
 540                     queue->vif->gso_prefix_mask) {
 541                         resp = RING_GET_RESPONSE(&queue->rx,
 542                                                  queue->rx.rsp_prod_pvt++);
 543
 544                         resp->flags = XEN_NETRXF_gso_prefix | XEN_NETRXF_more_data;
 545
 546                         resp->offset = queue->meta[npo.meta_cons].gso_size;
 547                         resp->id = queue->meta[npo.meta_cons].id;
 548                         resp->status = XENVIF_RX_CB(skb)->meta_slots_used;
 549
 550                         npo.meta_cons++;
 551                         XENVIF_RX_CB(skb)->meta_slots_used--;
 552                 }
 553
 554
 555                 queue->stats.tx_bytes += skb->len;
 556                 queue->stats.tx_packets++;
 557
 558                 status = xenvif_check_gop(queue->vif,
 559                                           XENVIF_RX_CB(skb)->meta_slots_used,
 560                                           &npo);
 561
 562                 if (XENVIF_RX_CB(skb)->meta_slots_used == 1)
 563                         flags = 0;
 564                 else
 565                         flags = XEN_NETRXF_more_data;
 566
 567                 if (skb->ip_summed == CHECKSUM_PARTIAL) /* local packet? */
 568                         flags |= XEN_NETRXF_csum_blank | XEN_NETRXF_data_validated;
 569                 else if (skb->ip_summed == CHECKSUM_UNNECESSARY)
 570                         /* remote but checksummed. */
 571                         flags |= XEN_NETRXF_data_validated;
 572
 573                 offset = 0;
 574                 resp = make_rx_response(queue, queue->meta[npo.meta_cons].id,
 575                                         status, offset,
 576                                         queue->meta[npo.meta_cons].size,
 577                                         flags);
 578
 579                 if ((1 << queue->meta[npo.meta_cons].gso_type) &
 580                     queue->vif->gso_mask) {
 581                         struct xen_netif_extra_info *gso =
 582                                 (struct xen_netif_extra_info *)
 583                                 RING_GET_RESPONSE(&queue->rx,
 584                                                   queue->rx.rsp_prod_pvt++);
 585
 586                         resp->flags |= XEN_NETRXF_extra_info;
 587
 588                         gso->u.gso.type = queue->meta[npo.meta_cons].gso_type;
 589                         gso->u.gso.size = queue->meta[npo.meta_cons].gso_size;
 590                         gso->u.gso.pad = 0;
 591                         gso->u.gso.features = 0;
 592
 593                         gso->type = XEN_NETIF_EXTRA_TYPE_GSO;
 594                         gso->flags = 0;
 595                 }
 596
 597                 xenvif_add_frag_responses(queue, status,
 598                                           queue->meta + npo.meta_cons + 1,
 599                                           XENVIF_RX_CB(skb)->meta_slots_used);
 600
 601                 RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&queue->rx, ret);
 602
 603                 need_to_notify |= !!ret;
 604
 605                 npo.meta_cons += XENVIF_RX_CB(skb)->meta_slots_used;
 606                 dev_kfree_skb(skb);
 607         }
 608
 609 done:
 610         if (need_to_notify)
 611                 notify_remote_via_irq(queue->rx_irq);
 612 }
 613
 614 void xenvif_napi_schedule_or_enable_events(struct xenvif_queue *queue)
 615 {
 616         int more_to_do;
 617
 618         RING_FINAL_CHECK_FOR_REQUESTS(&queue->tx, more_to_do);
 619
 620         if (more_to_do)
 621                 napi_schedule(&queue->napi);
 622 }
 623
 624 static void tx_add_credit(struct xenvif_queue *queue)
 625 {
 626         unsigned long max_burst, max_credit;
 627
 628         /*
 629          * Allow a burst big enough to transmit a jumbo packet of up to 128kB.
 630          * Otherwise the interface can seize up due to insufficient credit.
 631          */
 632         max_burst = RING_GET_REQUEST(&queue->tx, queue->tx.req_cons)->size;
 633         max_burst = min(max_burst, 131072UL);
 634         max_burst = max(max_burst, queue->credit_bytes);
 635
 636         /* Take care that adding a new chunk of credit doesn't wrap to zero. */
 637         max_credit = queue->remaining_credit + queue->credit_bytes;
 638         if (max_credit < queue->remaining_credit)
 639                 max_credit = ULONG_MAX; /* wrapped: clamp to ULONG_MAX */
 640
 641         queue->remaining_credit = min(max_credit, max_burst);
 642 }
 643
 644 static void tx_credit_callback(unsigned long data)
 645 {
 646         struct xenvif_queue *queue = (struct xenvif_queue *)data;
 647         tx_add_credit(queue);
 648         xenvif_napi_schedule_or_enable_events(queue);
 649 }
 650
 651 static void xenvif_tx_err(struct xenvif_queue *queue,
 652                           struct xen_netif_tx_request *txp, RING_IDX end)
 653 {
 654         RING_IDX cons = queue->tx.req_cons;
 655         unsigned long flags;
 656
 657         do {
 658                 int notify;
 659
 660                 spin_lock_irqsave(&queue->response_lock, flags);
 661                 make_tx_response(queue, txp, XEN_NETIF_RSP_ERROR);
 662                 RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&queue->tx, notify);
 663                 spin_unlock_irqrestore(&queue->response_lock, flags);
 664                 if (notify)
 665                         notify_remote_via_irq(queue->tx_irq);
 666
 667                 if (cons == end)
 668                         break;
 669                 txp = RING_GET_REQUEST(&queue->tx, cons++);
 670         } while (1);
 671         queue->tx.req_cons = cons;
 672 }
 673
 674 static void xenvif_fatal_tx_err(struct xenvif *vif)
 675 {
 676         netdev_err(vif->dev, "fatal error; disabling device\n");
 677         vif->disabled = true;
 678         /* Disable the vif from queue 0's kthread */
 679         if (vif->queues)
 680                 xenvif_kick_thread(&vif->queues[0]);
 681 }
 682
 683 static int xenvif_count_requests(struct xenvif_queue *queue,
 684                                  struct xen_netif_tx_request *first,
 685                                  struct xen_netif_tx_request *txp,
 686                                  int work_to_do)
 687 {
 688         RING_IDX cons = queue->tx.req_cons;
 689         int slots = 0;
 690         int drop_err = 0;
 691         int more_data;
 692
 693         if (!(first->flags & XEN_NETTXF_more_data))
 694                 return 0;
 695
 696         do {
 697                 struct xen_netif_tx_request dropped_tx = { 0 };
 698
 699                 if (slots >= work_to_do) {
 700                         netdev_err(queue->vif->dev,
 701                                    "Asked for %d slots but exceeds this limit\n",
 702                                    work_to_do);
 703                         xenvif_fatal_tx_err(queue->vif);
 704                         return -ENODATA;
 705                 }
 706
 707                 /* This guest is really using too many slots and
 708                  * considered malicious.
 709                  */
 710                 if (unlikely(slots >= fatal_skb_slots)) {
 711                         netdev_err(queue->vif->dev,
 712                                    "Malicious frontend using %d slots, threshold %u\n",
 713                                    slots, fatal_skb_slots);
 714                         xenvif_fatal_tx_err(queue->vif);
 715                         return -E2BIG;
 716                 }
 717
 718                 /* Xen network protocol had implicit dependency on
 719                  * MAX_SKB_FRAGS. XEN_NETBK_LEGACY_SLOTS_MAX is set to
 720                  * the historical MAX_SKB_FRAGS value 18 to honor the
 721                  * same behavior as before. Any packet using more than
 722                  * 18 slots but less than fatal_skb_slots slots is
 723                  * dropped
 724                  */
 725                 if (!drop_err && slots >= XEN_NETBK_LEGACY_SLOTS_MAX) {
 726                         if (net_ratelimit())
 727                                 netdev_dbg(queue->vif->dev,
 728                                            "Too many slots (%d) exceeding limit (%d), dropping packet\n",
 729                                            slots, XEN_NETBK_LEGACY_SLOTS_MAX);
 730                         drop_err = -E2BIG;
 731                 }
 732
 733                 if (drop_err)
 734                         txp = &dropped_tx;
 735
 736                 memcpy(txp, RING_GET_REQUEST(&queue->tx, cons + slots),
 737                        sizeof(*txp));
 738
 739                 /* If the guest submitted a frame >= 64 KiB then
 740                  * first->size overflowed and following slots will
 741                  * appear to be larger than the frame.
 742                  *
 743                  * This cannot be fatal error as there are buggy
 744                  * frontends that do this.
 745                  *
 746                  * Consume all slots and drop the packet.
 747                  */
 748                 if (!drop_err && txp->size > first->size) {
 749                         if (net_ratelimit())
 750                                 netdev_dbg(queue->vif->dev,
 751                                            "Invalid tx request, slot size %u > remaining size %u\n",
 752                                            txp->size, first->size);
 753                         drop_err = -EIO;
 754                 }
 755
 756                 first->size -= txp->size;
 757                 slots++;
 758
 759                 if (unlikely((txp->offset + txp->size) > PAGE_SIZE)) {
 760                         netdev_err(queue->vif->dev, "Cross page boundary, txp->offset: %x, size: %u\n",
 761                                  txp->offset, txp->size);
 762                         xenvif_fatal_tx_err(queue->vif);
 763                         return -EINVAL;
 764                 }
 765
 766                 more_data = txp->flags & XEN_NETTXF_more_data;
 767
 768                 if (!drop_err)
 769                         txp++;
 770
 771         } while (more_data);
 772
 773         if (drop_err) {
 774                 xenvif_tx_err(queue, first, cons + slots);
 775                 return drop_err;
 776         }
 777
 778         return slots;
 779 }
 780
 781
 782 struct xenvif_tx_cb {
 783         u16 pending_idx;
 784 };
 785
 786 #define XENVIF_TX_CB(skb) ((struct xenvif_tx_cb *)(skb)->cb)
 787
 788 static inline void xenvif_tx_create_map_op(struct xenvif_queue *queue,
 789                                           u16 pending_idx,
 790                                           struct xen_netif_tx_request *txp,
 791                                           struct gnttab_map_grant_ref *mop)
 792 {
 793         queue->pages_to_map[mop-queue->tx_map_ops] = queue->mmap_pages[pending_idx];
 794         gnttab_set_map_op(mop, idx_to_kaddr(queue, pending_idx),
 795                           GNTMAP_host_map | GNTMAP_readonly,
 796                           txp->gref, queue->vif->domid);
 797
 798         memcpy(&queue->pending_tx_info[pending_idx].req, txp,
 799                sizeof(*txp));
 800 }
 801
 802 static inline struct sk_buff *xenvif_alloc_skb(unsigned int size)
 803 {
 804         struct sk_buff *skb =
 805                 alloc_skb(size + NET_SKB_PAD + NET_IP_ALIGN,
 806                           GFP_ATOMIC | __GFP_NOWARN);
 807         if (unlikely(skb == NULL))
 808                 return NULL;
 809
 810         /* Packets passed to netif_rx() must have some headroom. */
 811         skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN);
 812
 813         /* Initialize it here to avoid later surprises */
 814         skb_shinfo(skb)->destructor_arg = NULL;
 815
 816         return skb;
 817 }
 818
 819 static struct gnttab_map_grant_ref *xenvif_get_requests(struct xenvif_queue *queue,
 820                                                         struct sk_buff *skb,
 821                                                         struct xen_netif_tx_request *txp,
 822                                                         struct gnttab_map_grant_ref *gop)
 823 {
 824         struct skb_shared_info *shinfo = skb_shinfo(skb);
 825         skb_frag_t *frags = shinfo->frags;
 826         u16 pending_idx = XENVIF_TX_CB(skb)->pending_idx;
 827         int start;
 828         pending_ring_idx_t index;
 829         unsigned int nr_slots, frag_overflow = 0;
 830
 831         /* At this point shinfo->nr_frags is in fact the number of
 832          * slots, which can be as large as XEN_NETBK_LEGACY_SLOTS_MAX.
 833          */
 834         if (shinfo->nr_frags > MAX_SKB_FRAGS) {
 835                 frag_overflow = shinfo->nr_frags - MAX_SKB_FRAGS;
 836                 BUG_ON(frag_overflow > MAX_SKB_FRAGS);
 837                 shinfo->nr_frags = MAX_SKB_FRAGS;
 838         }
 839         nr_slots = shinfo->nr_frags;
 840
 841         /* Skip first skb fragment if it is on same page as header fragment. */
 842         start = (frag_get_pending_idx(&shinfo->frags[0]) == pending_idx);
 843
 844         for (shinfo->nr_frags = start; shinfo->nr_frags < nr_slots;
 845              shinfo->nr_frags++, txp++, gop++) {
 846                 index = pending_index(queue->pending_cons++);
 847                 pending_idx = queue->pending_ring[index];
 848                 xenvif_tx_create_map_op(queue, pending_idx, txp, gop);
 849                 frag_set_pending_idx(&frags[shinfo->nr_frags], pending_idx);
 850         }
 851
 852         if (frag_overflow) {
 853                 struct sk_buff *nskb = xenvif_alloc_skb(0);
 854                 if (unlikely(nskb == NULL)) {
 855                         if (net_ratelimit())
 856                                 netdev_err(queue->vif->dev,
 857                                            "Can't allocate the frag_list skb.\n");
 858                         return NULL;
 859                 }
 860
 861                 shinfo = skb_shinfo(nskb);
 862                 frags = shinfo->frags;
 863
 864                 for (shinfo->nr_frags = 0; shinfo->nr_frags < frag_overflow;
 865                      shinfo->nr_frags++, txp++, gop++) {
 866                         index = pending_index(queue->pending_cons++);
 867                         pending_idx = queue->pending_ring[index];
 868                         xenvif_tx_create_map_op(queue, pending_idx, txp, gop);
 869                         frag_set_pending_idx(&frags[shinfo->nr_frags],
 870                                              pending_idx);
 871                 }
 872
 873                 skb_shinfo(skb)->frag_list = nskb;
 874         }
 875
 876         return gop;
 877 }
 878
 879 static inline void xenvif_grant_handle_set(struct xenvif_queue *queue,
 880                                            u16 pending_idx,
 881                                            grant_handle_t handle)
 882 {
 883         if (unlikely(queue->grant_tx_handle[pending_idx] !=
 884                      NETBACK_INVALID_HANDLE)) {
 885                 netdev_err(queue->vif->dev,
 886                            "Trying to overwrite active handle! pending_idx: %x\n",
 887                            pending_idx);
 888                 BUG();
 889         }
 890         queue->grant_tx_handle[pending_idx] = handle;
 891 }
 892
 893 static inline void xenvif_grant_handle_reset(struct xenvif_queue *queue,
 894                                              u16 pending_idx)
 895 {
 896         if (unlikely(queue->grant_tx_handle[pending_idx] ==
 897                      NETBACK_INVALID_HANDLE)) {
 898                 netdev_err(queue->vif->dev,
 899                            "Trying to unmap invalid handle! pending_idx: %x\n",
 900                            pending_idx);
 901                 BUG();
 902         }
 903         queue->grant_tx_handle[pending_idx] = NETBACK_INVALID_HANDLE;
 904 }
 905
 906 static int xenvif_tx_check_gop(struct xenvif_queue *queue,
 907                                struct sk_buff *skb,
 908                                struct gnttab_map_grant_ref **gopp_map,
 909                                struct gnttab_copy **gopp_copy)
 910 {
 911         struct gnttab_map_grant_ref *gop_map = *gopp_map;
 912         u16 pending_idx = XENVIF_TX_CB(skb)->pending_idx;
 913         /* This always points to the shinfo of the skb being checked, which
 914          * could be either the first or the one on the frag_list
 915          */
 916         struct skb_shared_info *shinfo = skb_shinfo(skb);
 917         /* If this is non-NULL, we are currently checking the frag_list skb, and
 918          * this points to the shinfo of the first one
 919          */
 920         struct skb_shared_info *first_shinfo = NULL;
 921         int nr_frags = shinfo->nr_frags;
 922         const bool sharedslot = nr_frags &&
 923                                 frag_get_pending_idx(&shinfo->frags[0]) == pending_idx;
 924         int i, err;
 925
 926         /* Check status of header. */
 927         err = (*gopp_copy)->status;
 928         if (unlikely(err)) {
 929                 if (net_ratelimit())
 930                         netdev_dbg(queue->vif->dev,
 931                                    "Grant copy of header failed! status: %d pending_idx: %u ref: %u\n",
 932                                    (*gopp_copy)->status,
 933                                    pending_idx,
 934                                    (*gopp_copy)->source.u.ref);
 935                 /* The first frag might still have this slot mapped */
 936                 if (!sharedslot)
 937                         xenvif_idx_release(queue, pending_idx,
 938                                            XEN_NETIF_RSP_ERROR);
 939         }
 940         (*gopp_copy)++;
 941
 942 check_frags:
 943         for (i = 0; i < nr_frags; i++, gop_map++) {
 944                 int j, newerr;
 945
 946                 pending_idx = frag_get_pending_idx(&shinfo->frags[i]);
 947
 948                 /* Check error status: if okay then remember grant handle. */
 949                 newerr = gop_map->status;
 950
 951                 if (likely(!newerr)) {
 952                         xenvif_grant_handle_set(queue,
 953                                                 pending_idx,
 954                                                 gop_map->handle);
 955                         /* Had a previous error? Invalidate this fragment. */
 956                         if (unlikely(err)) {
 957                                 xenvif_idx_unmap(queue, pending_idx);
 958                                 /* If the mapping of the first frag was OK, but
 959                                  * the header's copy failed, and they are
 960                                  * sharing a slot, send an error
 961                                  */
 962                                 if (i == 0 && sharedslot)
 963                                         xenvif_idx_release(queue, pending_idx,
 964                                                            XEN_NETIF_RSP_ERROR);
 965                                 else
 966                                         xenvif_idx_release(queue, pending_idx,
 967                                                            XEN_NETIF_RSP_OKAY);
 968                         }
 969                         continue;
 970                 }
 971
 972                 /* Error on this fragment: respond to client with an error. */
 973                 if (net_ratelimit())
 974                         netdev_dbg(queue->vif->dev,
 975                                    "Grant map of %d. frag failed! status: %d pending_idx: %u ref: %u\n",
 976                                    i,
 977                                    gop_map->status,
 978                                    pending_idx,
 979                                    gop_map->ref);
 980
 981                 xenvif_idx_release(queue, pending_idx, XEN_NETIF_RSP_ERROR);
 982
 983                 /* Not the first error? Preceding frags already invalidated. */
 984                 if (err)
 985                         continue;
 986
 987                 /* First error: if the header haven't shared a slot with the
 988                  * first frag, release it as well.
 989                  */
 990                 if (!sharedslot)
 991                         xenvif_idx_release(queue,
 992                                            XENVIF_TX_CB(skb)->pending_idx,
 993                                            XEN_NETIF_RSP_OKAY);
 994
 995                 /* Invalidate preceding fragments of this skb. */
 996                 for (j = 0; j < i; j++) {
 997                         pending_idx = frag_get_pending_idx(&shinfo->frags[j]);
 998                         xenvif_idx_unmap(queue, pending_idx);
 999                         xenvif_idx_release(queue, pending_idx,
1000                                            XEN_NETIF_RSP_OKAY);
1001                 }
1002
1003                 /* And if we found the error while checking the frag_list, unmap
1004                  * the first skb's frags
1005                  */
1006                 if (first_shinfo) {
1007                         for (j = 0; j < first_shinfo->nr_frags; j++) {
1008                                 pending_idx = frag_get_pending_idx(&first_shinfo->frags[j]);
1009                                 xenvif_idx_unmap(queue, pending_idx);
1010                                 xenvif_idx_release(queue, pending_idx,
1011                                                    XEN_NETIF_RSP_OKAY);
1012                         }
1013                 }
1014
1015                 /* Remember the error: invalidate all subsequent fragments. */
1016                 err = newerr;
1017         }
1018
1019         if (skb_has_frag_list(skb) && !first_shinfo) {
1020                 first_shinfo = skb_shinfo(skb);
1021                 shinfo = skb_shinfo(skb_shinfo(skb)->frag_list);
1022                 nr_frags = shinfo->nr_frags;
1023
1024                 goto check_frags;
1025         }
1026
1027         *gopp_map = gop_map;
1028         return err;
1029 }
1030
1031 static void xenvif_fill_frags(struct xenvif_queue *queue, struct sk_buff *skb)
1032 {
1033         struct skb_shared_info *shinfo = skb_shinfo(skb);
1034         int nr_frags = shinfo->nr_frags;
1035         int i;
1036         u16 prev_pending_idx = INVALID_PENDING_IDX;
1037
1038         for (i = 0; i < nr_frags; i++) {
1039                 skb_frag_t *frag = shinfo->frags + i;
1040                 struct xen_netif_tx_request *txp;
1041                 struct page *page;
1042                 u16 pending_idx;
1043
1044                 pending_idx = frag_get_pending_idx(frag);
1045
1046                 /* If this is not the first frag, chain it to the previous*/
1047                 if (prev_pending_idx == INVALID_PENDING_IDX)
1048                         skb_shinfo(skb)->destructor_arg =
1049                                 &callback_param(queue, pending_idx);
1050                 else
1051                         callback_param(queue, prev_pending_idx).ctx =
1052                                 &callback_param(queue, pending_idx);
1053
1054                 callback_param(queue, pending_idx).ctx = NULL;
1055                 prev_pending_idx = pending_idx;
1056
1057                 txp = &queue->pending_tx_info[pending_idx].req;
1058                 page = virt_to_page(idx_to_kaddr(queue, pending_idx));
1059                 __skb_fill_page_desc(skb, i, page, txp->offset, txp->size);
1060                 skb->len += txp->size;
1061                 skb->data_len += txp->size;
1062                 skb->truesize += txp->size;
1063
1064                 /* Take an extra reference to offset network stack's put_page */
1065                 get_page(queue->mmap_pages[pending_idx]);
1066         }
1067 }
1068
1069 static int xenvif_get_extras(struct xenvif_queue *queue,
1070                                 struct xen_netif_extra_info *extras,
1071                                 int work_to_do)
1072 {
1073         struct xen_netif_extra_info extra;
1074         RING_IDX cons = queue->tx.req_cons;
1075
1076         do {
1077                 if (unlikely(work_to_do-- <= 0)) {
1078                         netdev_err(queue->vif->dev, "Missing extra info\n");
1079                         xenvif_fatal_tx_err(queue->vif);
1080                         return -EBADR;
1081                 }
1082
1083                 memcpy(&extra, RING_GET_REQUEST(&queue->tx, cons),
1084                        sizeof(extra));
1085                 if (unlikely(!extra.type ||
1086                              extra.type >= XEN_NETIF_EXTRA_TYPE_MAX)) {
1087                         queue->tx.req_cons = ++cons;
1088                         netdev_err(queue->vif->dev,
1089                                    "Invalid extra type: %d\n", extra.type);
1090                         xenvif_fatal_tx_err(queue->vif);
1091                         return -EINVAL;
1092                 }
1093
1094                 memcpy(&extras[extra.type - 1], &extra, sizeof(extra));
1095                 queue->tx.req_cons = ++cons;
1096         } while (extra.flags & XEN_NETIF_EXTRA_FLAG_MORE);
1097
1098         return work_to_do;
1099 }
1100
1101 static int xenvif_set_skb_gso(struct xenvif *vif,
1102                               struct sk_buff *skb,
1103                               struct xen_netif_extra_info *gso)
1104 {
1105         if (!gso->u.gso.size) {
1106                 netdev_err(vif->dev, "GSO size must not be zero.\n");
1107                 xenvif_fatal_tx_err(vif);
1108                 return -EINVAL;
1109         }
1110
1111         switch (gso->u.gso.type) {
1112         case XEN_NETIF_GSO_TYPE_TCPV4:
1113                 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1114                 break;
1115         case XEN_NETIF_GSO_TYPE_TCPV6:
1116                 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
1117                 break;
1118         default:
1119                 netdev_err(vif->dev, "Bad GSO type %d.\n", gso->u.gso.type);
1120                 xenvif_fatal_tx_err(vif);
1121                 return -EINVAL;
1122         }
1123
1124         skb_shinfo(skb)->gso_size = gso->u.gso.size;
1125         /* gso_segs will be calculated later */
1126
1127         return 0;
1128 }
1129
1130 static int checksum_setup(struct xenvif_queue *queue, struct sk_buff *skb)
1131 {
1132         bool recalculate_partial_csum = false;
1133
1134         /* A GSO SKB must be CHECKSUM_PARTIAL. However some buggy
1135          * peers can fail to set NETRXF_csum_blank when sending a GSO
1136          * frame. In this case force the SKB to CHECKSUM_PARTIAL and
1137          * recalculate the partial checksum.
1138          */
1139         if (skb->ip_summed != CHECKSUM_PARTIAL && skb_is_gso(skb)) {
1140                 queue->stats.rx_gso_checksum_fixup++;
1141                 skb->ip_summed = CHECKSUM_PARTIAL;
1142                 recalculate_partial_csum = true;
1143         }
1144
1145         /* A non-CHECKSUM_PARTIAL SKB does not require setup. */
1146         if (skb->ip_summed != CHECKSUM_PARTIAL)
1147                 return 0;
1148
1149         return skb_checksum_setup(skb, recalculate_partial_csum);
1150 }
1151
1152 static bool tx_credit_exceeded(struct xenvif_queue *queue, unsigned size)
1153 {
1154         u64 now = get_jiffies_64();
1155         u64 next_credit = queue->credit_window_start +
1156                 msecs_to_jiffies(queue->credit_usec / 1000);
1157
1158         /* Timer could already be pending in rare cases. */
1159         if (timer_pending(&queue->credit_timeout))
1160                 return true;
1161
1162         /* Passed the point where we can replenish credit? */
1163         if (time_after_eq64(now, next_credit)) {
1164                 queue->credit_window_start = now;
1165                 tx_add_credit(queue);
1166         }
1167
1168         /* Still too big to send right now? Set a callback. */
1169         if (size > queue->remaining_credit) {
1170                 queue->credit_timeout.data     =
1171                         (unsigned long)queue;
1172                 queue->credit_timeout.function =
1173                         tx_credit_callback;
1174                 mod_timer(&queue->credit_timeout,
1175                           next_credit);
1176                 queue->credit_window_start = next_credit;
1177
1178                 return true;
1179         }
1180
1181         return false;
1182 }
1183
1184 static void xenvif_tx_build_gops(struct xenvif_queue *queue,
1185                                      int budget,
1186                                      unsigned *copy_ops,
1187                                      unsigned *map_ops)
1188 {
1189         struct gnttab_map_grant_ref *gop = queue->tx_map_ops, *request_gop;
1190         struct sk_buff *skb;
1191         int ret;
1192
1193         while (skb_queue_len(&queue->tx_queue) < budget) {
1194                 struct xen_netif_tx_request txreq;
1195                 struct xen_netif_tx_request txfrags[XEN_NETBK_LEGACY_SLOTS_MAX];
1196                 struct xen_netif_extra_info extras[XEN_NETIF_EXTRA_TYPE_MAX-1];
1197                 u16 pending_idx;
1198                 RING_IDX idx;
1199                 int work_to_do;
1200                 unsigned int data_len;
1201                 pending_ring_idx_t index;
1202
1203                 if (queue->tx.sring->req_prod - queue->tx.req_cons >
1204                     XEN_NETIF_TX_RING_SIZE) {
1205                         netdev_err(queue->vif->dev,
1206                                    "Impossible number of requests. "
1207                                    "req_prod %d, req_cons %d, size %ld\n",
1208                                    queue->tx.sring->req_prod, queue->tx.req_cons,
1209                                    XEN_NETIF_TX_RING_SIZE);
1210                         xenvif_fatal_tx_err(queue->vif);
1211                         break;
1212                 }
1213
1214                 work_to_do = RING_HAS_UNCONSUMED_REQUESTS(&queue->tx);
1215                 if (!work_to_do)
1216                         break;
1217
1218                 idx = queue->tx.req_cons;
1219                 rmb(); /* Ensure that we see the request before we copy it. */
1220                 memcpy(&txreq, RING_GET_REQUEST(&queue->tx, idx), sizeof(txreq));
1221
1222                 /* Credit-based scheduling. */
1223                 if (txreq.size > queue->remaining_credit &&
1224                     tx_credit_exceeded(queue, txreq.size))
1225                         break;
1226
1227                 queue->remaining_credit -= txreq.size;
1228
1229                 work_to_do--;
1230                 queue->tx.req_cons = ++idx;
1231
1232                 memset(extras, 0, sizeof(extras));
1233                 if (txreq.flags & XEN_NETTXF_extra_info) {
1234                         work_to_do = xenvif_get_extras(queue, extras,
1235                                                        work_to_do);
1236                         idx = queue->tx.req_cons;
1237                         if (unlikely(work_to_do < 0))
1238                                 break;
1239                 }
1240
1241                 ret = xenvif_count_requests(queue, &txreq, txfrags, work_to_do);
1242                 if (unlikely(ret < 0))
1243                         break;
1244
1245                 idx += ret;
1246
1247                 if (unlikely(txreq.size < ETH_HLEN)) {
1248                         netdev_dbg(queue->vif->dev,
1249                                    "Bad packet size: %d\n", txreq.size);
1250                         xenvif_tx_err(queue, &txreq, idx);
1251                         break;
1252                 }
1253
1254                 /* No crossing a page as the payload mustn't fragment. */
1255                 if (unlikely((txreq.offset + txreq.size) > PAGE_SIZE)) {
1256                         netdev_err(queue->vif->dev,
1257                                    "txreq.offset: %x, size: %u, end: %lu\n",
1258                                    txreq.offset, txreq.size,
1259                                    (txreq.offset&~PAGE_MASK) + txreq.size);
1260                         xenvif_fatal_tx_err(queue->vif);
1261                         break;
1262                 }
1263
1264                 index = pending_index(queue->pending_cons);
1265                 pending_idx = queue->pending_ring[index];
1266
1267                 data_len = (txreq.size > XEN_NETBACK_TX_COPY_LEN &&
1268                             ret < XEN_NETBK_LEGACY_SLOTS_MAX) ?
1269                         XEN_NETBACK_TX_COPY_LEN : txreq.size;
1270
1271                 skb = xenvif_alloc_skb(data_len);
1272                 if (unlikely(skb == NULL)) {
1273                         netdev_dbg(queue->vif->dev,
1274                                    "Can't allocate a skb in start_xmit.\n");
1275                         xenvif_tx_err(queue, &txreq, idx);
1276                         break;
1277                 }
1278
1279                 if (extras[XEN_NETIF_EXTRA_TYPE_GSO - 1].type) {
1280                         struct xen_netif_extra_info *gso;
1281                         gso = &extras[XEN_NETIF_EXTRA_TYPE_GSO - 1];
1282
1283                         if (xenvif_set_skb_gso(queue->vif, skb, gso)) {
1284                                 /* Failure in xenvif_set_skb_gso is fatal. */
1285                                 kfree_skb(skb);
1286                                 break;
1287                         }
1288                 }
1289
1290                 XENVIF_TX_CB(skb)->pending_idx = pending_idx;
1291
1292                 __skb_put(skb, data_len);
1293                 queue->tx_copy_ops[*copy_ops].source.u.ref = txreq.gref;
1294                 queue->tx_copy_ops[*copy_ops].source.domid = queue->vif->domid;
1295                 queue->tx_copy_ops[*copy_ops].source.offset = txreq.offset;
1296
1297                 queue->tx_copy_ops[*copy_ops].dest.u.gmfn =
1298                         virt_to_mfn(skb->data);
1299                 queue->tx_copy_ops[*copy_ops].dest.domid = DOMID_SELF;
1300                 queue->tx_copy_ops[*copy_ops].dest.offset =
1301                         offset_in_page(skb->data);
1302
1303                 queue->tx_copy_ops[*copy_ops].len = data_len;
1304                 queue->tx_copy_ops[*copy_ops].flags = GNTCOPY_source_gref;
1305
1306                 (*copy_ops)++;
1307
1308                 skb_shinfo(skb)->nr_frags = ret;
1309                 if (data_len < txreq.size) {
1310                         skb_shinfo(skb)->nr_frags++;
1311                         frag_set_pending_idx(&skb_shinfo(skb)->frags[0],
1312                                              pending_idx);
1313                         xenvif_tx_create_map_op(queue, pending_idx, &txreq, gop);
1314                         gop++;
1315                 } else {
1316                         frag_set_pending_idx(&skb_shinfo(skb)->frags[0],
1317                                              INVALID_PENDING_IDX);
1318                         memcpy(&queue->pending_tx_info[pending_idx].req, &txreq,
1319                                sizeof(txreq));
1320                 }
1321
1322                 queue->pending_cons++;
1323
1324                 request_gop = xenvif_get_requests(queue, skb, txfrags, gop);
1325                 if (request_gop == NULL) {
1326                         kfree_skb(skb);
1327                         xenvif_tx_err(queue, &txreq, idx);
1328                         break;
1329                 }
1330                 gop = request_gop;
1331
1332                 __skb_queue_tail(&queue->tx_queue, skb);
1333
1334                 queue->tx.req_cons = idx;
1335
1336                 if (((gop-queue->tx_map_ops) >= ARRAY_SIZE(queue->tx_map_ops)) ||
1337                     (*copy_ops >= ARRAY_SIZE(queue->tx_copy_ops)))
1338                         break;
1339         }
1340
1341         (*map_ops) = gop - queue->tx_map_ops;
1342         return;
1343 }
1344
1345 /* Consolidate skb with a frag_list into a brand new one with local pages on
1346  * frags. Returns 0 or -ENOMEM if can't allocate new pages.
1347  */
1348 static int xenvif_handle_frag_list(struct xenvif_queue *queue, struct sk_buff *skb)
1349 {
1350         unsigned int offset = skb_headlen(skb);
1351         skb_frag_t frags[MAX_SKB_FRAGS];
1352         int i;
1353         struct ubuf_info *uarg;
1354         struct sk_buff *nskb = skb_shinfo(skb)->frag_list;
1355
1356         queue->stats.tx_zerocopy_sent += 2;
1357         queue->stats.tx_frag_overflow++;
1358
1359         xenvif_fill_frags(queue, nskb);
1360         /* Subtract frags size, we will correct it later */
1361         skb->truesize -= skb->data_len;
1362         skb->len += nskb->len;
1363         skb->data_len += nskb->len;
1364
1365         /* create a brand new frags array and coalesce there */
1366         for (i = 0; offset < skb->len; i++) {
1367                 struct page *page;
1368                 unsigned int len;
1369
1370                 BUG_ON(i >= MAX_SKB_FRAGS);
1371                 page = alloc_page(GFP_ATOMIC);
1372                 if (!page) {
1373                         int j;
1374                         skb->truesize += skb->data_len;
1375                         for (j = 0; j < i; j++)
1376                                 put_page(frags[j].page.p);
1377                         return -ENOMEM;
1378                 }
1379
1380                 if (offset + PAGE_SIZE < skb->len)
1381                         len = PAGE_SIZE;
1382                 else
1383                         len = skb->len - offset;
1384                 if (skb_copy_bits(skb, offset, page_address(page), len))
1385                         BUG();
1386
1387                 offset += len;
1388                 frags[i].page.p = page;
1389                 frags[i].page_offset = 0;
1390                 skb_frag_size_set(&frags[i], len);
1391         }
1392         /* swap out with old one */
1393         memcpy(skb_shinfo(skb)->frags,
1394                frags,
1395                i * sizeof(skb_frag_t));
1396         skb_shinfo(skb)->nr_frags = i;
1397         skb->truesize += i * PAGE_SIZE;
1398
1399         /* remove traces of mapped pages and frag_list */
1400         skb_frag_list_init(skb);
1401         uarg = skb_shinfo(skb)->destructor_arg;
1402         /* increase inflight counter to offset decrement in callback */
1403         atomic_inc(&queue->inflight_packets);
1404         uarg->callback(uarg, true);
1405         skb_shinfo(skb)->destructor_arg = NULL;
1406
1407         xenvif_skb_zerocopy_prepare(queue, nskb);
1408         kfree_skb(nskb);
1409
1410         return 0;
1411 }
1412
1413 static int xenvif_tx_submit(struct xenvif_queue *queue)
1414 {
1415         struct gnttab_map_grant_ref *gop_map = queue->tx_map_ops;
1416         struct gnttab_copy *gop_copy = queue->tx_copy_ops;
1417         struct sk_buff *skb;
1418         int work_done = 0;
1419
1420         while ((skb = __skb_dequeue(&queue->tx_queue)) != NULL) {
1421                 struct xen_netif_tx_request *txp;
1422                 u16 pending_idx;
1423                 unsigned data_len;
1424
1425                 pending_idx = XENVIF_TX_CB(skb)->pending_idx;
1426                 txp = &queue->pending_tx_info[pending_idx].req;
1427
1428                 /* Check the remap error code. */
1429                 if (unlikely(xenvif_tx_check_gop(queue, skb, &gop_map, &gop_copy))) {
1430                         /* If there was an error, xenvif_tx_check_gop is
1431                          * expected to release all the frags which were mapped,
1432                          * so kfree_skb shouldn't do it again
1433                          */
1434                         skb_shinfo(skb)->nr_frags = 0;
1435                         if (skb_has_frag_list(skb)) {
1436                                 struct sk_buff *nskb =
1437                                                 skb_shinfo(skb)->frag_list;
1438                                 skb_shinfo(nskb)->nr_frags = 0;
1439                         }
1440                         kfree_skb(skb);
1441                         continue;
1442                 }
1443
1444                 data_len = skb->len;
1445                 callback_param(queue, pending_idx).ctx = NULL;
1446                 if (data_len < txp->size) {
1447                         /* Append the packet payload as a fragment. */
1448                         txp->offset += data_len;
1449                         txp->size -= data_len;
1450                 } else {
1451                         /* Schedule a response immediately. */
1452                         xenvif_idx_release(queue, pending_idx,
1453                                            XEN_NETIF_RSP_OKAY);
1454                 }
1455
1456                 if (txp->flags & XEN_NETTXF_csum_blank)
1457                         skb->ip_summed = CHECKSUM_PARTIAL;
1458                 else if (txp->flags & XEN_NETTXF_data_validated)
1459                         skb->ip_summed = CHECKSUM_UNNECESSARY;
1460
1461                 xenvif_fill_frags(queue, skb);
1462
1463                 if (unlikely(skb_has_frag_list(skb))) {
1464                         if (xenvif_handle_frag_list(queue, skb)) {
1465                                 if (net_ratelimit())
1466                                         netdev_err(queue->vif->dev,
1467                                                    "Not enough memory to consolidate frag_list!\n");
1468                                 xenvif_skb_zerocopy_prepare(queue, skb);
1469                                 kfree_skb(skb);
1470                                 continue;
1471                         }
1472                 }
1473
1474                 skb->dev      = queue->vif->dev;
1475                 skb->protocol = eth_type_trans(skb, skb->dev);
1476                 skb_reset_network_header(skb);
1477
1478                 if (checksum_setup(queue, skb)) {
1479                         netdev_dbg(queue->vif->dev,
1480                                    "Can't setup checksum in net_tx_action\n");
1481                         /* We have to set this flag to trigger the callback */
1482                         if (skb_shinfo(skb)->destructor_arg)
1483                                 xenvif_skb_zerocopy_prepare(queue, skb);
1484                         kfree_skb(skb);
1485                         continue;
1486                 }
1487
1488                 skb_probe_transport_header(skb, 0);
1489
1490                 /* If the packet is GSO then we will have just set up the
1491                  * transport header offset in checksum_setup so it's now
1492                  * straightforward to calculate gso_segs.
1493                  */
1494                 if (skb_is_gso(skb)) {
1495                         int mss = skb_shinfo(skb)->gso_size;
1496                         int hdrlen = skb_transport_header(skb) -
1497                                 skb_mac_header(skb) +
1498                                 tcp_hdrlen(skb);
1499
1500                         skb_shinfo(skb)->gso_segs =
1501                                 DIV_ROUND_UP(skb->len - hdrlen, mss);
1502                 }
1503
1504                 queue->stats.rx_bytes += skb->len;
1505                 queue->stats.rx_packets++;
1506
1507                 work_done++;
1508
1509                 /* Set this flag right before netif_receive_skb, otherwise
1510                  * someone might think this packet already left netback, and
1511                  * do a skb_copy_ubufs while we are still in control of the
1512                  * skb. E.g. the __pskb_pull_tail earlier can do such thing.
1513                  */
1514                 if (skb_shinfo(skb)->destructor_arg) {
1515                         xenvif_skb_zerocopy_prepare(queue, skb);
1516                         queue->stats.tx_zerocopy_sent++;
1517                 }
1518
1519                 netif_receive_skb(skb);
1520         }
1521
1522         return work_done;
1523 }
1524
1525 void xenvif_zerocopy_callback(struct ubuf_info *ubuf, bool zerocopy_success)
1526 {
1527         unsigned long flags;
1528         pending_ring_idx_t index;
1529         struct xenvif_queue *queue = ubuf_to_queue(ubuf);
1530
1531         /* This is the only place where we grab this lock, to protect callbacks
1532          * from each other.
1533          */
1534         spin_lock_irqsave(&queue->callback_lock, flags);
1535         do {
1536                 u16 pending_idx = ubuf->desc;
1537                 ubuf = (struct ubuf_info *) ubuf->ctx;
1538                 BUG_ON(queue->dealloc_prod - queue->dealloc_cons >=
1539                         MAX_PENDING_REQS);
1540                 index = pending_index(queue->dealloc_prod);
1541                 queue->dealloc_ring[index] = pending_idx;
1542                 /* Sync with xenvif_tx_dealloc_action:
1543                  * insert idx then incr producer.
1544                  */
1545                 smp_wmb();
1546                 queue->dealloc_prod++;
1547         } while (ubuf);
1548         wake_up(&queue->dealloc_wq);
1549         spin_unlock_irqrestore(&queue->callback_lock, flags);
1550
1551         if (likely(zerocopy_success))
1552                 queue->stats.tx_zerocopy_success++;
1553         else
1554                 queue->stats.tx_zerocopy_fail++;
1555         xenvif_skb_zerocopy_complete(queue);
1556 }
1557
1558 static inline void xenvif_tx_dealloc_action(struct xenvif_queue *queue)
1559 {
1560         struct gnttab_unmap_grant_ref *gop;
1561         pending_ring_idx_t dc, dp;
1562         u16 pending_idx, pending_idx_release[MAX_PENDING_REQS];
1563         unsigned int i = 0;
1564
1565         dc = queue->dealloc_cons;
1566         gop = queue->tx_unmap_ops;
1567
1568         /* Free up any grants we have finished using */
1569         do {
1570                 dp = queue->dealloc_prod;
1571
1572                 /* Ensure we see all indices enqueued by all
1573                  * xenvif_zerocopy_callback().
1574                  */
1575                 smp_rmb();
1576
1577                 while (dc != dp) {
1578                         BUG_ON(gop - queue->tx_unmap_ops > MAX_PENDING_REQS);
1579                         pending_idx =
1580                                 queue->dealloc_ring[pending_index(dc++)];
1581
1582                         pending_idx_release[gop-queue->tx_unmap_ops] =
1583                                 pending_idx;
1584                         queue->pages_to_unmap[gop-queue->tx_unmap_ops] =
1585                                 queue->mmap_pages[pending_idx];
1586                         gnttab_set_unmap_op(gop,
1587                                             idx_to_kaddr(queue, pending_idx),
1588                                             GNTMAP_host_map,
1589                                             queue->grant_tx_handle[pending_idx]);
1590                         xenvif_grant_handle_reset(queue, pending_idx);
1591                         ++gop;
1592                 }
1593
1594         } while (dp != queue->dealloc_prod);
1595
1596         queue->dealloc_cons = dc;
1597
1598         if (gop - queue->tx_unmap_ops > 0) {
1599                 int ret;
1600                 ret = gnttab_unmap_refs(queue->tx_unmap_ops,
1601                                         NULL,
1602                                         queue->pages_to_unmap,
1603                                         gop - queue->tx_unmap_ops);
1604                 if (ret) {
1605                         netdev_err(queue->vif->dev, "Unmap fail: nr_ops %tx ret %d\n",
1606                                    gop - queue->tx_unmap_ops, ret);
1607                         for (i = 0; i < gop - queue->tx_unmap_ops; ++i) {
1608                                 if (gop[i].status != GNTST_okay)
1609                                         netdev_err(queue->vif->dev,
1610                                                    " host_addr: %llx handle: %x status: %d\n",
1611                                                    gop[i].host_addr,
1612                                                    gop[i].handle,
1613                                                    gop[i].status);
1614                         }
1615                         BUG();
1616                 }
1617         }
1618
1619         for (i = 0; i < gop - queue->tx_unmap_ops; ++i)
1620                 xenvif_idx_release(queue, pending_idx_release[i],
1621                                    XEN_NETIF_RSP_OKAY);
1622 }
1623
1624
1625 /* Called after netfront has transmitted */
1626 int xenvif_tx_action(struct xenvif_queue *queue, int budget)
1627 {
1628         unsigned nr_mops, nr_cops = 0;
1629         int work_done, ret;
1630
1631         if (unlikely(!tx_work_todo(queue)))
1632                 return 0;
1633
1634         xenvif_tx_build_gops(queue, budget, &nr_cops, &nr_mops);
1635
1636         if (nr_cops == 0)
1637                 return 0;
1638
1639         gnttab_batch_copy(queue->tx_copy_ops, nr_cops);
1640         if (nr_mops != 0) {
1641                 ret = gnttab_map_refs(queue->tx_map_ops,
1642                                       NULL,
1643                                       queue->pages_to_map,
1644                                       nr_mops);
1645                 BUG_ON(ret);
1646         }
1647
1648         work_done = xenvif_tx_submit(queue);
1649
1650         return work_done;
1651 }
1652
1653 static void xenvif_idx_release(struct xenvif_queue *queue, u16 pending_idx,
1654                                u8 status)
1655 {
1656         struct pending_tx_info *pending_tx_info;
1657         pending_ring_idx_t index;
1658         int notify;
1659         unsigned long flags;
1660
1661         pending_tx_info = &queue->pending_tx_info[pending_idx];
1662
1663         spin_lock_irqsave(&queue->response_lock, flags);
1664
1665         make_tx_response(queue, &pending_tx_info->req, status);
1666
1667         /* Release the pending index before pusing the Tx response so
1668          * its available before a new Tx request is pushed by the
1669          * frontend.
1670          */
1671         index = pending_index(queue->pending_prod++);
1672         queue->pending_ring[index] = pending_idx;
1673
1674         RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&queue->tx, notify);
1675
1676         spin_unlock_irqrestore(&queue->response_lock, flags);
1677
1678         if (notify)
1679                 notify_remote_via_irq(queue->tx_irq);
1680 }
1681
1682
1683 static void make_tx_response(struct xenvif_queue *queue,
1684                              struct xen_netif_tx_request *txp,
1685                              s8       st)
1686 {
1687         RING_IDX i = queue->tx.rsp_prod_pvt;
1688         struct xen_netif_tx_response *resp;
1689
1690         resp = RING_GET_RESPONSE(&queue->tx, i);
1691         resp->id     = txp->id;
1692         resp->status = st;
1693
1694         if (txp->flags & XEN_NETTXF_extra_info)
1695                 RING_GET_RESPONSE(&queue->tx, ++i)->status = XEN_NETIF_RSP_NULL;
1696
1697         queue->tx.rsp_prod_pvt = ++i;
1698 }
1699
1700 static struct xen_netif_rx_response *make_rx_response(struct xenvif_queue *queue,
1701                                              u16      id,
1702                                              s8       st,
1703                                              u16      offset,
1704                                              u16      size,
1705                                              u16      flags)
1706 {
1707         RING_IDX i = queue->rx.rsp_prod_pvt;
1708         struct xen_netif_rx_response *resp;
1709
1710         resp = RING_GET_RESPONSE(&queue->rx, i);
1711         resp->offset     = offset;
1712         resp->flags      = flags;
1713         resp->id         = id;
1714         resp->status     = (s16)size;
1715         if (st < 0)
1716                 resp->status = (s16)st;
1717
1718         queue->rx.rsp_prod_pvt = ++i;
1719
1720         return resp;
1721 }
1722
1723 void xenvif_idx_unmap(struct xenvif_queue *queue, u16 pending_idx)
1724 {
1725         int ret;
1726         struct gnttab_unmap_grant_ref tx_unmap_op;
1727
1728         gnttab_set_unmap_op(&tx_unmap_op,
1729                             idx_to_kaddr(queue, pending_idx),
1730                             GNTMAP_host_map,
1731                             queue->grant_tx_handle[pending_idx]);
1732         xenvif_grant_handle_reset(queue, pending_idx);
1733
1734         ret = gnttab_unmap_refs(&tx_unmap_op, NULL,
1735                                 &queue->mmap_pages[pending_idx], 1);
1736         if (ret) {
1737                 netdev_err(queue->vif->dev,
1738                            "Unmap fail: ret: %d pending_idx: %d host_addr: %llx handle: %x status: %d\n",
1739                            ret,
1740                            pending_idx,
1741                            tx_unmap_op.host_addr,
1742                            tx_unmap_op.handle,
1743                            tx_unmap_op.status);
1744                 BUG();
1745         }
1746 }
1747
1748 static inline int tx_work_todo(struct xenvif_queue *queue)
1749 {
1750         if (likely(RING_HAS_UNCONSUMED_REQUESTS(&queue->tx)))
1751                 return 1;
1752
1753         return 0;
1754 }
1755
1756 static inline bool tx_dealloc_work_todo(struct xenvif_queue *queue)
1757 {
1758         return queue->dealloc_cons != queue->dealloc_prod;
1759 }
1760
1761 void xenvif_unmap_frontend_rings(struct xenvif_queue *queue)
1762 {
1763         if (queue->tx.sring)
1764                 xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(queue->vif),
1765                                         queue->tx.sring);
1766         if (queue->rx.sring)
1767                 xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(queue->vif),
1768                                         queue->rx.sring);
1769 }
1770
1771 int xenvif_map_frontend_rings(struct xenvif_queue *queue,
1772                               grant_ref_t tx_ring_ref,
1773                               grant_ref_t rx_ring_ref)
1774 {
1775         void *addr;
1776         struct xen_netif_tx_sring *txs;
1777         struct xen_netif_rx_sring *rxs;
1778
1779         int err = -ENOMEM;
1780
1781         err = xenbus_map_ring_valloc(xenvif_to_xenbus_device(queue->vif),
1782                                      tx_ring_ref, &addr);
1783         if (err)
1784                 goto err;
1785
1786         txs = (struct xen_netif_tx_sring *)addr;
1787         BACK_RING_INIT(&queue->tx, txs, PAGE_SIZE);
1788
1789         err = xenbus_map_ring_valloc(xenvif_to_xenbus_device(queue->vif),
1790                                      rx_ring_ref, &addr);
1791         if (err)
1792                 goto err;
1793
1794         rxs = (struct xen_netif_rx_sring *)addr;
1795         BACK_RING_INIT(&queue->rx, rxs, PAGE_SIZE);
1796
1797         return 0;
1798
1799 err:
1800         xenvif_unmap_frontend_rings(queue);
1801         return err;
1802 }
1803
1804 static void xenvif_queue_carrier_off(struct xenvif_queue *queue)
1805 {
1806         struct xenvif *vif = queue->vif;
1807
1808         queue->stalled = true;
1809
1810         /* At least one queue has stalled? Disable the carrier. */
1811         spin_lock(&vif->lock);
1812         if (vif->stalled_queues++ == 0) {
1813                 netdev_info(vif->dev, "Guest Rx stalled");
1814                 netif_carrier_off(vif->dev);
1815         }
1816         spin_unlock(&vif->lock);
1817 }
1818
1819 static void xenvif_queue_carrier_on(struct xenvif_queue *queue)
1820 {
1821         struct xenvif *vif = queue->vif;
1822
1823         queue->last_rx_time = jiffies; /* Reset Rx stall detection. */
1824         queue->stalled = false;
1825
1826         /* All queues are ready? Enable the carrier. */
1827         spin_lock(&vif->lock);
1828         if (--vif->stalled_queues == 0) {
1829                 netdev_info(vif->dev, "Guest Rx ready");
1830                 netif_carrier_on(vif->dev);
1831         }
1832         spin_unlock(&vif->lock);
1833 }
1834
1835 static bool xenvif_rx_queue_stalled(struct xenvif_queue *queue)
1836 {
1837         RING_IDX prod, cons;
1838
1839         prod = queue->rx.sring->req_prod;
1840         cons = queue->rx.req_cons;
1841
1842         return !queue->stalled
1843                 && prod - cons < XEN_NETBK_RX_SLOTS_MAX
1844                 && time_after(jiffies,
1845                               queue->last_rx_time + queue->vif->stall_timeout);
1846 }
1847
1848 static bool xenvif_rx_queue_ready(struct xenvif_queue *queue)
1849 {
1850         RING_IDX prod, cons;
1851
1852         prod = queue->rx.sring->req_prod;
1853         cons = queue->rx.req_cons;
1854
1855         return queue->stalled
1856                 && prod - cons >= XEN_NETBK_RX_SLOTS_MAX;
1857 }
1858
1859 static bool xenvif_have_rx_work(struct xenvif_queue *queue)
1860 {
1861         return (!skb_queue_empty(&queue->rx_queue)
1862                 && xenvif_rx_ring_slots_available(queue, XEN_NETBK_RX_SLOTS_MAX))
1863                 || (queue->vif->stall_timeout &&
1864                     (xenvif_rx_queue_stalled(queue)
1865                      || xenvif_rx_queue_ready(queue)))
1866                 || kthread_should_stop()
1867                 || queue->vif->disabled;
1868 }
1869
1870 static long xenvif_rx_queue_timeout(struct xenvif_queue *queue)
1871 {
1872         struct sk_buff *skb;
1873         long timeout;
1874
1875         skb = skb_peek(&queue->rx_queue);
1876         if (!skb)
1877                 return MAX_SCHEDULE_TIMEOUT;
1878
1879         timeout = XENVIF_RX_CB(skb)->expires - jiffies;
1880         return timeout < 0 ? 0 : timeout;
1881 }
1882
1883 /* Wait until the guest Rx thread has work.
1884  *
1885  * The timeout needs to be adjusted based on the current head of the
1886  * queue (and not just the head at the beginning).  In particular, if
1887  * the queue is initially empty an infinite timeout is used and this
1888  * needs to be reduced when a skb is queued.
1889  *
1890  * This cannot be done with wait_event_timeout() because it only
1891  * calculates the timeout once.
1892  */
1893 static void xenvif_wait_for_rx_work(struct xenvif_queue *queue)
1894 {
1895         DEFINE_WAIT(wait);
1896
1897         if (xenvif_have_rx_work(queue))
1898                 return;
1899
1900         for (;;) {
1901                 long ret;
1902
1903                 prepare_to_wait(&queue->wq, &wait, TASK_INTERRUPTIBLE);
1904                 if (xenvif_have_rx_work(queue))
1905                         break;
1906                 ret = schedule_timeout(xenvif_rx_queue_timeout(queue));
1907                 if (!ret)
1908                         break;
1909         }
1910         finish_wait(&queue->wq, &wait);
1911 }
1912
1913 int xenvif_kthread_guest_rx(void *data)
1914 {
1915         struct xenvif_queue *queue = data;
1916         struct xenvif *vif = queue->vif;
1917
1918         if (!vif->stall_timeout)
1919                 xenvif_queue_carrier_on(queue);
1920
1921         for (;;) {
1922                 xenvif_wait_for_rx_work(queue);
1923
1924                 if (kthread_should_stop())
1925                         break;
1926
1927                 /* This frontend is found to be rogue, disable it in
1928                  * kthread context. Currently this is only set when
1929                  * netback finds out frontend sends malformed packet,
1930                  * but we cannot disable the interface in softirq
1931                  * context so we defer it here, if this thread is
1932                  * associated with queue 0.
1933                  */
1934                 if (unlikely(vif->disabled && queue->id == 0)) {
1935                         xenvif_carrier_off(vif);
1936                         break;
1937                 }
1938
1939                 if (!skb_queue_empty(&queue->rx_queue))
1940                         xenvif_rx_action(queue);
1941
1942                 /* If the guest hasn't provided any Rx slots for a
1943                  * while it's probably not responsive, drop the
1944                  * carrier so packets are dropped earlier.
1945                  */
1946                 if (vif->stall_timeout) {
1947                         if (xenvif_rx_queue_stalled(queue))
1948                                 xenvif_queue_carrier_off(queue);
1949                         else if (xenvif_rx_queue_ready(queue))
1950                                 xenvif_queue_carrier_on(queue);
1951                 }
1952
1953                 /* Queued packets may have foreign pages from other
1954                  * domains.  These cannot be queued indefinitely as
1955                  * this would starve guests of grant refs and transmit
1956                  * slots.
1957                  */
1958                 xenvif_rx_queue_drop_expired(queue);
1959
1960                 xenvif_rx_queue_maybe_wake(queue);
1961
1962                 cond_resched();
1963         }
1964
1965         /* Bin any remaining skbs */
1966         xenvif_rx_queue_purge(queue);
1967
1968         return 0;
1969 }
1970
1971 static bool xenvif_dealloc_kthread_should_stop(struct xenvif_queue *queue)
1972 {
1973         /* Dealloc thread must remain running until all inflight
1974          * packets complete.
1975          */
1976         return kthread_should_stop() &&
1977                 !atomic_read(&queue->inflight_packets);
1978 }
1979
1980 int xenvif_dealloc_kthread(void *data)
1981 {
1982         struct xenvif_queue *queue = data;
1983
1984         for (;;) {
1985                 wait_event_interruptible(queue->dealloc_wq,
1986                                          tx_dealloc_work_todo(queue) ||
1987                                          xenvif_dealloc_kthread_should_stop(queue));
1988                 if (xenvif_dealloc_kthread_should_stop(queue))
1989                         break;
1990
1991                 xenvif_tx_dealloc_action(queue);
1992                 cond_resched();
1993         }
1994
1995         /* Unmap anything remaining*/
1996         if (tx_dealloc_work_todo(queue))
1997                 xenvif_tx_dealloc_action(queue);
1998
1999         return 0;
2000 }
2001
2002 static int __init netback_init(void)
2003 {
2004         int rc = 0;
2005
2006         if (!xen_domain())
2007                 return -ENODEV;
2008
2009         /* Allow as many queues as there are CPUs, by default */
2010         xenvif_max_queues = num_online_cpus();
2011
2012         if (fatal_skb_slots < XEN_NETBK_LEGACY_SLOTS_MAX) {
2013                 pr_info("fatal_skb_slots too small (%d), bump it to XEN_NETBK_LEGACY_SLOTS_MAX (%d)\n",
2014                         fatal_skb_slots, XEN_NETBK_LEGACY_SLOTS_MAX);
2015                 fatal_skb_slots = XEN_NETBK_LEGACY_SLOTS_MAX;
2016         }
2017
2018         rc = xenvif_xenbus_init();
2019         if (rc)
2020                 goto failed_init;
2021
2022 #ifdef CONFIG_DEBUG_FS
2023         xen_netback_dbg_root = debugfs_create_dir("xen-netback", NULL);
2024         if (IS_ERR_OR_NULL(xen_netback_dbg_root))
2025                 pr_warn("Init of debugfs returned %ld!\n",
2026                         PTR_ERR(xen_netback_dbg_root));
2027 #endif /* CONFIG_DEBUG_FS */
2028
2029         return 0;
2030
2031 failed_init:
2032         return rc;
2033 }
2034
2035 module_init(netback_init);
2036
2037 static void __exit netback_fini(void)
2038 {
2039 #ifdef CONFIG_DEBUG_FS
2040         if (!IS_ERR_OR_NULL(xen_netback_dbg_root))
2041                 debugfs_remove_recursive(xen_netback_dbg_root);
2042 #endif /* CONFIG_DEBUG_FS */
2043         xenvif_xenbus_fini();
2044 }
2045 module_exit(netback_fini);
2046
2047 MODULE_LICENSE("Dual BSD/GPL");
2048 MODULE_ALIAS("xen-backend:vif");