drivers/net/xen-netfront.c

   1 /*
   2  * Virtual network driver for conversing with remote driver backends.
   3  *
   4  * Copyright (c) 2002-2005, K A Fraser
   5  * Copyright (c) 2005, XenSource Ltd
   6  *
   7  * This program is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU General Public License version 2
   9  * as published by the Free Software Foundation; or, when distributed
  10  * separately from the Linux kernel or incorporated into other
  11  * software packages, subject to the following license:
  12  *
  13  * Permission is hereby granted, free of charge, to any person obtaining a copy
  14  * of this source file (the "Software"), to deal in the Software without
  15  * restriction, including without limitation the rights to use, copy, modify,
  16  * merge, publish, distribute, sublicense, and/or sell copies of the Software,
  17  * and to permit persons to whom the Software is furnished to do so, subject to
  18  * the following conditions:
  19  *
  20  * The above copyright notice and this permission notice shall be included in
  21  * all copies or substantial portions of the Software.
  22  *
  23  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  24  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  25  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
  26  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  27  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  28  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  29  * IN THE SOFTWARE.
  30  */
  31
  32 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  33
  34 #include <linux/module.h>
  35 #include <linux/kernel.h>
  36 #include <linux/netdevice.h>
  37 #include <linux/etherdevice.h>
  38 #include <linux/skbuff.h>
  39 #include <linux/ethtool.h>
  40 #include <linux/if_ether.h>
  41 #include <net/tcp.h>
  42 #include <linux/udp.h>
  43 #include <linux/moduleparam.h>
  44 #include <linux/mm.h>
  45 #include <linux/slab.h>
  46 #include <net/ip.h>
  47
  48 #include <asm/xen/page.h>
  49 #include <xen/xen.h>
  50 #include <xen/xenbus.h>
  51 #include <xen/events.h>
  52 #include <xen/page.h>
  53 #include <xen/platform_pci.h>
  54 #include <xen/grant_table.h>
  55
  56 #include <xen/interface/io/netif.h>
  57 #include <xen/interface/memory.h>
  58 #include <xen/interface/grant_table.h>
  59
  60 /* Module parameters */
  61 static unsigned int xennet_max_queues;
  62 module_param_named(max_queues, xennet_max_queues, uint, 0644);
  63 MODULE_PARM_DESC(max_queues,
  64                  "Maximum number of queues per virtual interface");
  65
  66 static const struct ethtool_ops xennet_ethtool_ops;
  67
  68 struct netfront_cb {
  69         int pull_to;
  70 };
  71
  72 #define NETFRONT_SKB_CB(skb)    ((struct netfront_cb *)((skb)->cb))
  73
  74 #define RX_COPY_THRESHOLD 256
  75
  76 #define GRANT_INVALID_REF       0
  77
  78 #define NET_TX_RING_SIZE __CONST_RING_SIZE(xen_netif_tx, PAGE_SIZE)
  79 #define NET_RX_RING_SIZE __CONST_RING_SIZE(xen_netif_rx, PAGE_SIZE)
  80
  81 /* Minimum number of Rx slots (includes slot for GSO metadata). */
  82 #define NET_RX_SLOTS_MIN (XEN_NETIF_NR_SLOTS_MIN + 1)
  83
  84 /* Queue name is interface name with "-qNNN" appended */
  85 #define QUEUE_NAME_SIZE (IFNAMSIZ + 6)
  86
  87 /* IRQ name is queue name with "-tx" or "-rx" appended */
  88 #define IRQ_NAME_SIZE (QUEUE_NAME_SIZE + 3)
  89
  90 struct netfront_stats {
  91         u64                     rx_packets;
  92         u64                     tx_packets;
  93         u64                     rx_bytes;
  94         u64                     tx_bytes;
  95         struct u64_stats_sync   syncp;
  96 };
  97
  98 struct netfront_info;
  99
 100 struct netfront_queue {
 101         unsigned int id; /* Queue ID, 0-based */
 102         char name[QUEUE_NAME_SIZE]; /* DEVNAME-qN */
 103         struct netfront_info *info;
 104
 105         struct napi_struct napi;
 106
 107         /* Split event channels support, tx_* == rx_* when using
 108          * single event channel.
 109          */
 110         unsigned int tx_evtchn, rx_evtchn;
 111         unsigned int tx_irq, rx_irq;
 112         /* Only used when split event channels support is enabled */
 113         char tx_irq_name[IRQ_NAME_SIZE]; /* DEVNAME-qN-tx */
 114         char rx_irq_name[IRQ_NAME_SIZE]; /* DEVNAME-qN-rx */
 115
 116         spinlock_t   tx_lock;
 117         struct xen_netif_tx_front_ring tx;
 118         int tx_ring_ref;
 119
 120         /*
 121          * {tx,rx}_skbs store outstanding skbuffs. Free tx_skb entries
 122          * are linked from tx_skb_freelist through skb_entry.link.
 123          *
 124          *  NB. Freelist index entries are always going to be less than
 125          *  PAGE_OFFSET, whereas pointers to skbs will always be equal or
 126          *  greater than PAGE_OFFSET: we use this property to distinguish
 127          *  them.
 128          */
 129         union skb_entry {
 130                 struct sk_buff *skb;
 131                 unsigned long link;
 132         } tx_skbs[NET_TX_RING_SIZE];
 133         grant_ref_t gref_tx_head;
 134         grant_ref_t grant_tx_ref[NET_TX_RING_SIZE];
 135         struct page *grant_tx_page[NET_TX_RING_SIZE];
 136         unsigned tx_skb_freelist;
 137
 138         spinlock_t   rx_lock ____cacheline_aligned_in_smp;
 139         struct xen_netif_rx_front_ring rx;
 140         int rx_ring_ref;
 141
 142         struct timer_list rx_refill_timer;
 143
 144         struct sk_buff *rx_skbs[NET_RX_RING_SIZE];
 145         grant_ref_t gref_rx_head;
 146         grant_ref_t grant_rx_ref[NET_RX_RING_SIZE];
 147
 148         unsigned long rx_pfn_array[NET_RX_RING_SIZE];
 149         struct multicall_entry rx_mcl[NET_RX_RING_SIZE+1];
 150         struct mmu_update rx_mmu[NET_RX_RING_SIZE];
 151 };
 152
 153 struct netfront_info {
 154         struct list_head list;
 155         struct net_device *netdev;
 156
 157         struct xenbus_device *xbdev;
 158
 159         /* Multi-queue support */
 160         struct netfront_queue *queues;
 161
 162         /* Statistics */
 163         struct netfront_stats __percpu *stats;
 164
 165         atomic_t rx_gso_checksum_fixup;
 166 };
 167
 168 struct netfront_rx_info {
 169         struct xen_netif_rx_response rx;
 170         struct xen_netif_extra_info extras[XEN_NETIF_EXTRA_TYPE_MAX - 1];
 171 };
 172
 173 static void skb_entry_set_link(union skb_entry *list, unsigned short id)
 174 {
 175         list->link = id;
 176 }
 177
 178 static int skb_entry_is_link(const union skb_entry *list)
 179 {
 180         BUILD_BUG_ON(sizeof(list->skb) != sizeof(list->link));
 181         return (unsigned long)list->skb < PAGE_OFFSET;
 182 }
 183
 184 /*
 185  * Access macros for acquiring freeing slots in tx_skbs[].
 186  */
 187
 188 static void add_id_to_freelist(unsigned *head, union skb_entry *list,
 189                                unsigned short id)
 190 {
 191         skb_entry_set_link(&list[id], *head);
 192         *head = id;
 193 }
 194
 195 static unsigned short get_id_from_freelist(unsigned *head,
 196                                            union skb_entry *list)
 197 {
 198         unsigned int id = *head;
 199         *head = list[id].link;
 200         return id;
 201 }
 202
 203 static int xennet_rxidx(RING_IDX idx)
 204 {
 205         return idx & (NET_RX_RING_SIZE - 1);
 206 }
 207
 208 static struct sk_buff *xennet_get_rx_skb(struct netfront_queue *queue,
 209                                          RING_IDX ri)
 210 {
 211         int i = xennet_rxidx(ri);
 212         struct sk_buff *skb = queue->rx_skbs[i];
 213         queue->rx_skbs[i] = NULL;
 214         return skb;
 215 }
 216
 217 static grant_ref_t xennet_get_rx_ref(struct netfront_queue *queue,
 218                                             RING_IDX ri)
 219 {
 220         int i = xennet_rxidx(ri);
 221         grant_ref_t ref = queue->grant_rx_ref[i];
 222         queue->grant_rx_ref[i] = GRANT_INVALID_REF;
 223         return ref;
 224 }
 225
 226 #ifdef CONFIG_SYSFS
 227 static int xennet_sysfs_addif(struct net_device *netdev);
 228 static void xennet_sysfs_delif(struct net_device *netdev);
 229 #else /* !CONFIG_SYSFS */
 230 #define xennet_sysfs_addif(dev) (0)
 231 #define xennet_sysfs_delif(dev) do { } while (0)
 232 #endif
 233
 234 static bool xennet_can_sg(struct net_device *dev)
 235 {
 236         return dev->features & NETIF_F_SG;
 237 }
 238
 239
 240 static void rx_refill_timeout(unsigned long data)
 241 {
 242         struct netfront_queue *queue = (struct netfront_queue *)data;
 243         napi_schedule(&queue->napi);
 244 }
 245
 246 static int netfront_tx_slot_available(struct netfront_queue *queue)
 247 {
 248         return (queue->tx.req_prod_pvt - queue->tx.rsp_cons) <
 249                 (NET_TX_RING_SIZE - MAX_SKB_FRAGS - 2);
 250 }
 251
 252 static void xennet_maybe_wake_tx(struct netfront_queue *queue)
 253 {
 254         struct net_device *dev = queue->info->netdev;
 255         struct netdev_queue *dev_queue = netdev_get_tx_queue(dev, queue->id);
 256
 257         if (unlikely(netif_tx_queue_stopped(dev_queue)) &&
 258             netfront_tx_slot_available(queue) &&
 259             likely(netif_running(dev)))
 260                 netif_tx_wake_queue(netdev_get_tx_queue(dev, queue->id));
 261 }
 262
 263
 264 static struct sk_buff *xennet_alloc_one_rx_buffer(struct netfront_queue *queue)
 265 {
 266         struct sk_buff *skb;
 267         struct page *page;
 268
 269         skb = __netdev_alloc_skb(queue->info->netdev,
 270                                  RX_COPY_THRESHOLD + NET_IP_ALIGN,
 271                                  GFP_ATOMIC | __GFP_NOWARN);
 272         if (unlikely(!skb))
 273                 return NULL;
 274
 275         page = alloc_page(GFP_ATOMIC | __GFP_NOWARN);
 276         if (!page) {
 277                 kfree_skb(skb);
 278                 return NULL;
 279         }
 280         skb_add_rx_frag(skb, 0, page, 0, 0, PAGE_SIZE);
 281
 282         /* Align ip header to a 16 bytes boundary */
 283         skb_reserve(skb, NET_IP_ALIGN);
 284         skb->dev = queue->info->netdev;
 285
 286         return skb;
 287 }
 288
 289
 290 static void xennet_alloc_rx_buffers(struct netfront_queue *queue)
 291 {
 292         RING_IDX req_prod = queue->rx.req_prod_pvt;
 293         int notify;
 294
 295         if (unlikely(!netif_carrier_ok(queue->info->netdev)))
 296                 return;
 297
 298         for (req_prod = queue->rx.req_prod_pvt;
 299              req_prod - queue->rx.rsp_cons < NET_RX_RING_SIZE;
 300              req_prod++) {
 301                 struct sk_buff *skb;
 302                 unsigned short id;
 303                 grant_ref_t ref;
 304                 unsigned long pfn;
 305                 struct xen_netif_rx_request *req;
 306
 307                 skb = xennet_alloc_one_rx_buffer(queue);
 308                 if (!skb)
 309                         break;
 310
 311                 id = xennet_rxidx(req_prod);
 312
 313                 BUG_ON(queue->rx_skbs[id]);
 314                 queue->rx_skbs[id] = skb;
 315
 316                 ref = gnttab_claim_grant_reference(&queue->gref_rx_head);
 317                 BUG_ON((signed short)ref < 0);
 318                 queue->grant_rx_ref[id] = ref;
 319
 320                 pfn = page_to_pfn(skb_frag_page(&skb_shinfo(skb)->frags[0]));
 321
 322                 req = RING_GET_REQUEST(&queue->rx, req_prod);
 323                 gnttab_grant_foreign_access_ref(ref,
 324                                                 queue->info->xbdev->otherend_id,
 325                                                 pfn_to_mfn(pfn),
 326                                                 0);
 327
 328                 req->id = id;
 329                 req->gref = ref;
 330         }
 331
 332         queue->rx.req_prod_pvt = req_prod;
 333
 334         /* Not enough requests? Try again later. */
 335         if (req_prod - queue->rx.rsp_cons < NET_RX_SLOTS_MIN) {
 336                 mod_timer(&queue->rx_refill_timer, jiffies + (HZ/10));
 337                 return;
 338         }
 339
 340         wmb();          /* barrier so backend seens requests */
 341
 342         RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&queue->rx, notify);
 343         if (notify)
 344                 notify_remote_via_irq(queue->rx_irq);
 345 }
 346
 347 static int xennet_open(struct net_device *dev)
 348 {
 349         struct netfront_info *np = netdev_priv(dev);
 350         unsigned int num_queues = dev->real_num_tx_queues;
 351         unsigned int i = 0;
 352         struct netfront_queue *queue = NULL;
 353
 354         for (i = 0; i < num_queues; ++i) {
 355                 queue = &np->queues[i];
 356                 napi_enable(&queue->napi);
 357
 358                 spin_lock_bh(&queue->rx_lock);
 359                 if (netif_carrier_ok(dev)) {
 360                         xennet_alloc_rx_buffers(queue);
 361                         queue->rx.sring->rsp_event = queue->rx.rsp_cons + 1;
 362                         if (RING_HAS_UNCONSUMED_RESPONSES(&queue->rx))
 363                                 napi_schedule(&queue->napi);
 364                 }
 365                 spin_unlock_bh(&queue->rx_lock);
 366         }
 367
 368         netif_tx_start_all_queues(dev);
 369
 370         return 0;
 371 }
 372
 373 static void xennet_tx_buf_gc(struct netfront_queue *queue)
 374 {
 375         RING_IDX cons, prod;
 376         unsigned short id;
 377         struct sk_buff *skb;
 378
 379         BUG_ON(!netif_carrier_ok(queue->info->netdev));
 380
 381         do {
 382                 prod = queue->tx.sring->rsp_prod;
 383                 rmb(); /* Ensure we see responses up to 'rp'. */
 384
 385                 for (cons = queue->tx.rsp_cons; cons != prod; cons++) {
 386                         struct xen_netif_tx_response *txrsp;
 387
 388                         txrsp = RING_GET_RESPONSE(&queue->tx, cons);
 389                         if (txrsp->status == XEN_NETIF_RSP_NULL)
 390                                 continue;
 391
 392                         id  = txrsp->id;
 393                         skb = queue->tx_skbs[id].skb;
 394                         if (unlikely(gnttab_query_foreign_access(
 395                                 queue->grant_tx_ref[id]) != 0)) {
 396                                 pr_alert("%s: warning -- grant still in use by backend domain\n",
 397                                          __func__);
 398                                 BUG();
 399                         }
 400                         gnttab_end_foreign_access_ref(
 401                                 queue->grant_tx_ref[id], GNTMAP_readonly);
 402                         gnttab_release_grant_reference(
 403                                 &queue->gref_tx_head, queue->grant_tx_ref[id]);
 404                         queue->grant_tx_ref[id] = GRANT_INVALID_REF;
 405                         queue->grant_tx_page[id] = NULL;
 406                         add_id_to_freelist(&queue->tx_skb_freelist, queue->tx_skbs, id);
 407                         dev_kfree_skb_irq(skb);
 408                 }
 409
 410                 queue->tx.rsp_cons = prod;
 411
 412                 /*
 413                  * Set a new event, then check for race with update of tx_cons.
 414                  * Note that it is essential to schedule a callback, no matter
 415                  * how few buffers are pending. Even if there is space in the
 416                  * transmit ring, higher layers may be blocked because too much
 417                  * data is outstanding: in such cases notification from Xen is
 418                  * likely to be the only kick that we'll get.
 419                  */
 420                 queue->tx.sring->rsp_event =
 421                         prod + ((queue->tx.sring->req_prod - prod) >> 1) + 1;
 422                 mb();           /* update shared area */
 423         } while ((cons == prod) && (prod != queue->tx.sring->rsp_prod));
 424
 425         xennet_maybe_wake_tx(queue);
 426 }
 427
 428 static void xennet_make_frags(struct sk_buff *skb, struct netfront_queue *queue,
 429                               struct xen_netif_tx_request *tx)
 430 {
 431         char *data = skb->data;
 432         unsigned long mfn;
 433         RING_IDX prod = queue->tx.req_prod_pvt;
 434         int frags = skb_shinfo(skb)->nr_frags;
 435         unsigned int offset = offset_in_page(data);
 436         unsigned int len = skb_headlen(skb);
 437         unsigned int id;
 438         grant_ref_t ref;
 439         int i;
 440
 441         /* While the header overlaps a page boundary (including being
 442            larger than a page), split it it into page-sized chunks. */
 443         while (len > PAGE_SIZE - offset) {
 444                 tx->size = PAGE_SIZE - offset;
 445                 tx->flags |= XEN_NETTXF_more_data;
 446                 len -= tx->size;
 447                 data += tx->size;
 448                 offset = 0;
 449
 450                 id = get_id_from_freelist(&queue->tx_skb_freelist, queue->tx_skbs);
 451                 queue->tx_skbs[id].skb = skb_get(skb);
 452                 tx = RING_GET_REQUEST(&queue->tx, prod++);
 453                 tx->id = id;
 454                 ref = gnttab_claim_grant_reference(&queue->gref_tx_head);
 455                 BUG_ON((signed short)ref < 0);
 456
 457                 mfn = virt_to_mfn(data);
 458                 gnttab_grant_foreign_access_ref(ref, queue->info->xbdev->otherend_id,
 459                                                 mfn, GNTMAP_readonly);
 460
 461                 queue->grant_tx_page[id] = virt_to_page(data);
 462                 tx->gref = queue->grant_tx_ref[id] = ref;
 463                 tx->offset = offset;
 464                 tx->size = len;
 465                 tx->flags = 0;
 466         }
 467
 468         /* Grant backend access to each skb fragment page. */
 469         for (i = 0; i < frags; i++) {
 470                 skb_frag_t *frag = skb_shinfo(skb)->frags + i;
 471                 struct page *page = skb_frag_page(frag);
 472
 473                 len = skb_frag_size(frag);
 474                 offset = frag->page_offset;
 475
 476                 /* Skip unused frames from start of page */
 477                 page += offset >> PAGE_SHIFT;
 478                 offset &= ~PAGE_MASK;
 479
 480                 while (len > 0) {
 481                         unsigned long bytes;
 482
 483                         bytes = PAGE_SIZE - offset;
 484                         if (bytes > len)
 485                                 bytes = len;
 486
 487                         tx->flags |= XEN_NETTXF_more_data;
 488
 489                         id = get_id_from_freelist(&queue->tx_skb_freelist,
 490                                                   queue->tx_skbs);
 491                         queue->tx_skbs[id].skb = skb_get(skb);
 492                         tx = RING_GET_REQUEST(&queue->tx, prod++);
 493                         tx->id = id;
 494                         ref = gnttab_claim_grant_reference(&queue->gref_tx_head);
 495                         BUG_ON((signed short)ref < 0);
 496
 497                         mfn = pfn_to_mfn(page_to_pfn(page));
 498                         gnttab_grant_foreign_access_ref(ref,
 499                                                         queue->info->xbdev->otherend_id,
 500                                                         mfn, GNTMAP_readonly);
 501
 502                         queue->grant_tx_page[id] = page;
 503                         tx->gref = queue->grant_tx_ref[id] = ref;
 504                         tx->offset = offset;
 505                         tx->size = bytes;
 506                         tx->flags = 0;
 507
 508                         offset += bytes;
 509                         len -= bytes;
 510
 511                         /* Next frame */
 512                         if (offset == PAGE_SIZE && len) {
 513                                 BUG_ON(!PageCompound(page));
 514                                 page++;
 515                                 offset = 0;
 516                         }
 517                 }
 518         }
 519
 520         queue->tx.req_prod_pvt = prod;
 521 }
 522
 523 /*
 524  * Count how many ring slots are required to send the frags of this
 525  * skb. Each frag might be a compound page.
 526  */
 527 static int xennet_count_skb_frag_slots(struct sk_buff *skb)
 528 {
 529         int i, frags = skb_shinfo(skb)->nr_frags;
 530         int pages = 0;
 531
 532         for (i = 0; i < frags; i++) {
 533                 skb_frag_t *frag = skb_shinfo(skb)->frags + i;
 534                 unsigned long size = skb_frag_size(frag);
 535                 unsigned long offset = frag->page_offset;
 536
 537                 /* Skip unused frames from start of page */
 538                 offset &= ~PAGE_MASK;
 539
 540                 pages += PFN_UP(offset + size);
 541         }
 542
 543         return pages;
 544 }
 545
 546 static u16 xennet_select_queue(struct net_device *dev, struct sk_buff *skb,
 547                                void *accel_priv, select_queue_fallback_t fallback)
 548 {
 549         unsigned int num_queues = dev->real_num_tx_queues;
 550         u32 hash;
 551         u16 queue_idx;
 552
 553         /* First, check if there is only one queue */
 554         if (num_queues == 1) {
 555                 queue_idx = 0;
 556         } else {
 557                 hash = skb_get_hash(skb);
 558                 queue_idx = hash % num_queues;
 559         }
 560
 561         return queue_idx;
 562 }
 563
 564 static int xennet_start_xmit(struct sk_buff *skb, struct net_device *dev)
 565 {
 566         unsigned short id;
 567         struct netfront_info *np = netdev_priv(dev);
 568         struct netfront_stats *stats = this_cpu_ptr(np->stats);
 569         struct xen_netif_tx_request *tx;
 570         char *data = skb->data;
 571         RING_IDX i;
 572         grant_ref_t ref;
 573         unsigned long mfn;
 574         int notify;
 575         int slots;
 576         unsigned int offset = offset_in_page(data);
 577         unsigned int len = skb_headlen(skb);
 578         unsigned long flags;
 579         struct netfront_queue *queue = NULL;
 580         unsigned int num_queues = dev->real_num_tx_queues;
 581         u16 queue_index;
 582
 583         /* Drop the packet if no queues are set up */
 584         if (num_queues < 1)
 585                 goto drop;
 586         /* Determine which queue to transmit this SKB on */
 587         queue_index = skb_get_queue_mapping(skb);
 588         queue = &np->queues[queue_index];
 589
 590         /* If skb->len is too big for wire format, drop skb and alert
 591          * user about misconfiguration.
 592          */
 593         if (unlikely(skb->len > XEN_NETIF_MAX_TX_SIZE)) {
 594                 net_alert_ratelimited(
 595                         "xennet: skb->len = %u, too big for wire format\n",
 596                         skb->len);
 597                 goto drop;
 598         }
 599
 600         slots = DIV_ROUND_UP(offset + len, PAGE_SIZE) +
 601                 xennet_count_skb_frag_slots(skb);
 602         if (unlikely(slots > MAX_SKB_FRAGS + 1)) {
 603                 net_dbg_ratelimited("xennet: skb rides the rocket: %d slots, %d bytes\n",
 604                                     slots, skb->len);
 605                 if (skb_linearize(skb))
 606                         goto drop;
 607                 data = skb->data;
 608                 offset = offset_in_page(data);
 609                 len = skb_headlen(skb);
 610         }
 611
 612         spin_lock_irqsave(&queue->tx_lock, flags);
 613
 614         if (unlikely(!netif_carrier_ok(dev) ||
 615                      (slots > 1 && !xennet_can_sg(dev)) ||
 616                      netif_needs_gso(dev, skb, netif_skb_features(skb)))) {
 617                 spin_unlock_irqrestore(&queue->tx_lock, flags);
 618                 goto drop;
 619         }
 620
 621         i = queue->tx.req_prod_pvt;
 622
 623         id = get_id_from_freelist(&queue->tx_skb_freelist, queue->tx_skbs);
 624         queue->tx_skbs[id].skb = skb;
 625
 626         tx = RING_GET_REQUEST(&queue->tx, i);
 627
 628         tx->id   = id;
 629         ref = gnttab_claim_grant_reference(&queue->gref_tx_head);
 630         BUG_ON((signed short)ref < 0);
 631         mfn = virt_to_mfn(data);
 632         gnttab_grant_foreign_access_ref(
 633                 ref, queue->info->xbdev->otherend_id, mfn, GNTMAP_readonly);
 634         queue->grant_tx_page[id] = virt_to_page(data);
 635         tx->gref = queue->grant_tx_ref[id] = ref;
 636         tx->offset = offset;
 637         tx->size = len;
 638
 639         tx->flags = 0;
 640         if (skb->ip_summed == CHECKSUM_PARTIAL)
 641                 /* local packet? */
 642                 tx->flags |= XEN_NETTXF_csum_blank | XEN_NETTXF_data_validated;
 643         else if (skb->ip_summed == CHECKSUM_UNNECESSARY)
 644                 /* remote but checksummed. */
 645                 tx->flags |= XEN_NETTXF_data_validated;
 646
 647         if (skb_shinfo(skb)->gso_size) {
 648                 struct xen_netif_extra_info *gso;
 649
 650                 gso = (struct xen_netif_extra_info *)
 651                         RING_GET_REQUEST(&queue->tx, ++i);
 652
 653                 tx->flags |= XEN_NETTXF_extra_info;
 654
 655                 gso->u.gso.size = skb_shinfo(skb)->gso_size;
 656                 gso->u.gso.type = (skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6) ?
 657                         XEN_NETIF_GSO_TYPE_TCPV6 :
 658                         XEN_NETIF_GSO_TYPE_TCPV4;
 659                 gso->u.gso.pad = 0;
 660                 gso->u.gso.features = 0;
 661
 662                 gso->type = XEN_NETIF_EXTRA_TYPE_GSO;
 663                 gso->flags = 0;
 664         }
 665
 666         queue->tx.req_prod_pvt = i + 1;
 667
 668         xennet_make_frags(skb, queue, tx);
 669         tx->size = skb->len;
 670
 671         RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&queue->tx, notify);
 672         if (notify)
 673                 notify_remote_via_irq(queue->tx_irq);
 674
 675         u64_stats_update_begin(&stats->syncp);
 676         stats->tx_bytes += skb->len;
 677         stats->tx_packets++;
 678         u64_stats_update_end(&stats->syncp);
 679
 680         /* Note: It is not safe to access skb after xennet_tx_buf_gc()! */
 681         xennet_tx_buf_gc(queue);
 682
 683         if (!netfront_tx_slot_available(queue))
 684                 netif_tx_stop_queue(netdev_get_tx_queue(dev, queue->id));
 685
 686         spin_unlock_irqrestore(&queue->tx_lock, flags);
 687
 688         return NETDEV_TX_OK;
 689
 690  drop:
 691         dev->stats.tx_dropped++;
 692         dev_kfree_skb_any(skb);
 693         return NETDEV_TX_OK;
 694 }
 695
 696 static int xennet_close(struct net_device *dev)
 697 {
 698         struct netfront_info *np = netdev_priv(dev);
 699         unsigned int num_queues = dev->real_num_tx_queues;
 700         unsigned int i;
 701         struct netfront_queue *queue;
 702         netif_tx_stop_all_queues(np->netdev);
 703         for (i = 0; i < num_queues; ++i) {
 704                 queue = &np->queues[i];
 705                 napi_disable(&queue->napi);
 706         }
 707         return 0;
 708 }
 709
 710 static void xennet_move_rx_slot(struct netfront_queue *queue, struct sk_buff *skb,
 711                                 grant_ref_t ref)
 712 {
 713         int new = xennet_rxidx(queue->rx.req_prod_pvt);
 714
 715         BUG_ON(queue->rx_skbs[new]);
 716         queue->rx_skbs[new] = skb;
 717         queue->grant_rx_ref[new] = ref;
 718         RING_GET_REQUEST(&queue->rx, queue->rx.req_prod_pvt)->id = new;
 719         RING_GET_REQUEST(&queue->rx, queue->rx.req_prod_pvt)->gref = ref;
 720         queue->rx.req_prod_pvt++;
 721 }
 722
 723 static int xennet_get_extras(struct netfront_queue *queue,
 724                              struct xen_netif_extra_info *extras,
 725                              RING_IDX rp)
 726
 727 {
 728         struct xen_netif_extra_info *extra;
 729         struct device *dev = &queue->info->netdev->dev;
 730         RING_IDX cons = queue->rx.rsp_cons;
 731         int err = 0;
 732
 733         do {
 734                 struct sk_buff *skb;
 735                 grant_ref_t ref;
 736
 737                 if (unlikely(cons + 1 == rp)) {
 738                         if (net_ratelimit())
 739                                 dev_warn(dev, "Missing extra info\n");
 740                         err = -EBADR;
 741                         break;
 742                 }
 743
 744                 extra = (struct xen_netif_extra_info *)
 745                         RING_GET_RESPONSE(&queue->rx, ++cons);
 746
 747                 if (unlikely(!extra->type ||
 748                              extra->type >= XEN_NETIF_EXTRA_TYPE_MAX)) {
 749                         if (net_ratelimit())
 750                                 dev_warn(dev, "Invalid extra type: %d\n",
 751                                         extra->type);
 752                         err = -EINVAL;
 753                 } else {
 754                         memcpy(&extras[extra->type - 1], extra,
 755                                sizeof(*extra));
 756                 }
 757
 758                 skb = xennet_get_rx_skb(queue, cons);
 759                 ref = xennet_get_rx_ref(queue, cons);
 760                 xennet_move_rx_slot(queue, skb, ref);
 761         } while (extra->flags & XEN_NETIF_EXTRA_FLAG_MORE);
 762
 763         queue->rx.rsp_cons = cons;
 764         return err;
 765 }
 766
 767 static int xennet_get_responses(struct netfront_queue *queue,
 768                                 struct netfront_rx_info *rinfo, RING_IDX rp,
 769                                 struct sk_buff_head *list)
 770 {
 771         struct xen_netif_rx_response *rx = &rinfo->rx;
 772         struct xen_netif_extra_info *extras = rinfo->extras;
 773         struct device *dev = &queue->info->netdev->dev;
 774         RING_IDX cons = queue->rx.rsp_cons;
 775         struct sk_buff *skb = xennet_get_rx_skb(queue, cons);
 776         grant_ref_t ref = xennet_get_rx_ref(queue, cons);
 777         int max = MAX_SKB_FRAGS + (rx->status <= RX_COPY_THRESHOLD);
 778         int slots = 1;
 779         int err = 0;
 780         unsigned long ret;
 781
 782         if (rx->flags & XEN_NETRXF_extra_info) {
 783                 err = xennet_get_extras(queue, extras, rp);
 784                 cons = queue->rx.rsp_cons;
 785         }
 786
 787         for (;;) {
 788                 if (unlikely(rx->status < 0 ||
 789                              rx->offset + rx->status > PAGE_SIZE)) {
 790                         if (net_ratelimit())
 791                                 dev_warn(dev, "rx->offset: %x, size: %u\n",
 792                                          rx->offset, rx->status);
 793                         xennet_move_rx_slot(queue, skb, ref);
 794                         err = -EINVAL;
 795                         goto next;
 796                 }
 797
 798                 /*
 799                  * This definitely indicates a bug, either in this driver or in
 800                  * the backend driver. In future this should flag the bad
 801                  * situation to the system controller to reboot the backend.
 802                  */
 803                 if (ref == GRANT_INVALID_REF) {
 804                         if (net_ratelimit())
 805                                 dev_warn(dev, "Bad rx response id %d.\n",
 806                                          rx->id);
 807                         err = -EINVAL;
 808                         goto next;
 809                 }
 810
 811                 ret = gnttab_end_foreign_access_ref(ref, 0);
 812                 BUG_ON(!ret);
 813
 814                 gnttab_release_grant_reference(&queue->gref_rx_head, ref);
 815
 816                 __skb_queue_tail(list, skb);
 817
 818 next:
 819                 if (!(rx->flags & XEN_NETRXF_more_data))
 820                         break;
 821
 822                 if (cons + slots == rp) {
 823                         if (net_ratelimit())
 824                                 dev_warn(dev, "Need more slots\n");
 825                         err = -ENOENT;
 826                         break;
 827                 }
 828
 829                 rx = RING_GET_RESPONSE(&queue->rx, cons + slots);
 830                 skb = xennet_get_rx_skb(queue, cons + slots);
 831                 ref = xennet_get_rx_ref(queue, cons + slots);
 832                 slots++;
 833         }
 834
 835         if (unlikely(slots > max)) {
 836                 if (net_ratelimit())
 837                         dev_warn(dev, "Too many slots\n");
 838                 err = -E2BIG;
 839         }
 840
 841         if (unlikely(err))
 842                 queue->rx.rsp_cons = cons + slots;
 843
 844         return err;
 845 }
 846
 847 static int xennet_set_skb_gso(struct sk_buff *skb,
 848                               struct xen_netif_extra_info *gso)
 849 {
 850         if (!gso->u.gso.size) {
 851                 if (net_ratelimit())
 852                         pr_warn("GSO size must not be zero\n");
 853                 return -EINVAL;
 854         }
 855
 856         if (gso->u.gso.type != XEN_NETIF_GSO_TYPE_TCPV4 &&
 857             gso->u.gso.type != XEN_NETIF_GSO_TYPE_TCPV6) {
 858                 if (net_ratelimit())
 859                         pr_warn("Bad GSO type %d\n", gso->u.gso.type);
 860                 return -EINVAL;
 861         }
 862
 863         skb_shinfo(skb)->gso_size = gso->u.gso.size;
 864         skb_shinfo(skb)->gso_type =
 865                 (gso->u.gso.type == XEN_NETIF_GSO_TYPE_TCPV4) ?
 866                 SKB_GSO_TCPV4 :
 867                 SKB_GSO_TCPV6;
 868
 869         /* Header must be checked, and gso_segs computed. */
 870         skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
 871         skb_shinfo(skb)->gso_segs = 0;
 872
 873         return 0;
 874 }
 875
 876 static RING_IDX xennet_fill_frags(struct netfront_queue *queue,
 877                                   struct sk_buff *skb,
 878                                   struct sk_buff_head *list)
 879 {
 880         struct skb_shared_info *shinfo = skb_shinfo(skb);
 881         RING_IDX cons = queue->rx.rsp_cons;
 882         struct sk_buff *nskb;
 883
 884         while ((nskb = __skb_dequeue(list))) {
 885                 struct xen_netif_rx_response *rx =
 886                         RING_GET_RESPONSE(&queue->rx, ++cons);
 887                 skb_frag_t *nfrag = &skb_shinfo(nskb)->frags[0];
 888
 889                 if (shinfo->nr_frags == MAX_SKB_FRAGS) {
 890                         unsigned int pull_to = NETFRONT_SKB_CB(skb)->pull_to;
 891
 892                         BUG_ON(pull_to <= skb_headlen(skb));
 893                         __pskb_pull_tail(skb, pull_to - skb_headlen(skb));
 894                 }
 895                 BUG_ON(shinfo->nr_frags >= MAX_SKB_FRAGS);
 896
 897                 skb_add_rx_frag(skb, shinfo->nr_frags, skb_frag_page(nfrag),
 898                                 rx->offset, rx->status, PAGE_SIZE);
 899
 900                 skb_shinfo(nskb)->nr_frags = 0;
 901                 kfree_skb(nskb);
 902         }
 903
 904         return cons;
 905 }
 906
 907 static int checksum_setup(struct net_device *dev, struct sk_buff *skb)
 908 {
 909         bool recalculate_partial_csum = false;
 910
 911         /*
 912          * A GSO SKB must be CHECKSUM_PARTIAL. However some buggy
 913          * peers can fail to set NETRXF_csum_blank when sending a GSO
 914          * frame. In this case force the SKB to CHECKSUM_PARTIAL and
 915          * recalculate the partial checksum.
 916          */
 917         if (skb->ip_summed != CHECKSUM_PARTIAL && skb_is_gso(skb)) {
 918                 struct netfront_info *np = netdev_priv(dev);
 919                 atomic_inc(&np->rx_gso_checksum_fixup);
 920                 skb->ip_summed = CHECKSUM_PARTIAL;
 921                 recalculate_partial_csum = true;
 922         }
 923
 924         /* A non-CHECKSUM_PARTIAL SKB does not require setup. */
 925         if (skb->ip_summed != CHECKSUM_PARTIAL)
 926                 return 0;
 927
 928         return skb_checksum_setup(skb, recalculate_partial_csum);
 929 }
 930
 931 static int handle_incoming_queue(struct netfront_queue *queue,
 932                                  struct sk_buff_head *rxq)
 933 {
 934         struct netfront_stats *stats = this_cpu_ptr(queue->info->stats);
 935         int packets_dropped = 0;
 936         struct sk_buff *skb;
 937
 938         while ((skb = __skb_dequeue(rxq)) != NULL) {
 939                 int pull_to = NETFRONT_SKB_CB(skb)->pull_to;
 940
 941                 if (pull_to > skb_headlen(skb))
 942                         __pskb_pull_tail(skb, pull_to - skb_headlen(skb));
 943
 944                 /* Ethernet work: Delayed to here as it peeks the header. */
 945                 skb->protocol = eth_type_trans(skb, queue->info->netdev);
 946                 skb_reset_network_header(skb);
 947
 948                 if (checksum_setup(queue->info->netdev, skb)) {
 949                         kfree_skb(skb);
 950                         packets_dropped++;
 951                         queue->info->netdev->stats.rx_errors++;
 952                         continue;
 953                 }
 954
 955                 u64_stats_update_begin(&stats->syncp);
 956                 stats->rx_packets++;
 957                 stats->rx_bytes += skb->len;
 958                 u64_stats_update_end(&stats->syncp);
 959
 960                 /* Pass it up. */
 961                 napi_gro_receive(&queue->napi, skb);
 962         }
 963
 964         return packets_dropped;
 965 }
 966
 967 static int xennet_poll(struct napi_struct *napi, int budget)
 968 {
 969         struct netfront_queue *queue = container_of(napi, struct netfront_queue, napi);
 970         struct net_device *dev = queue->info->netdev;
 971         struct sk_buff *skb;
 972         struct netfront_rx_info rinfo;
 973         struct xen_netif_rx_response *rx = &rinfo.rx;
 974         struct xen_netif_extra_info *extras = rinfo.extras;
 975         RING_IDX i, rp;
 976         int work_done;
 977         struct sk_buff_head rxq;
 978         struct sk_buff_head errq;
 979         struct sk_buff_head tmpq;
 980         int err;
 981
 982         spin_lock(&queue->rx_lock);
 983
 984         skb_queue_head_init(&rxq);
 985         skb_queue_head_init(&errq);
 986         skb_queue_head_init(&tmpq);
 987
 988         rp = queue->rx.sring->rsp_prod;
 989         rmb(); /* Ensure we see queued responses up to 'rp'. */
 990
 991         i = queue->rx.rsp_cons;
 992         work_done = 0;
 993         while ((i != rp) && (work_done < budget)) {
 994                 memcpy(rx, RING_GET_RESPONSE(&queue->rx, i), sizeof(*rx));
 995                 memset(extras, 0, sizeof(rinfo.extras));
 996
 997                 err = xennet_get_responses(queue, &rinfo, rp, &tmpq);
 998
 999                 if (unlikely(err)) {
1000 err:
1001                         while ((skb = __skb_dequeue(&tmpq)))
1002                                 __skb_queue_tail(&errq, skb);
1003                         dev->stats.rx_errors++;
1004                         i = queue->rx.rsp_cons;
1005                         continue;
1006                 }
1007
1008                 skb = __skb_dequeue(&tmpq);
1009
1010                 if (extras[XEN_NETIF_EXTRA_TYPE_GSO - 1].type) {
1011                         struct xen_netif_extra_info *gso;
1012                         gso = &extras[XEN_NETIF_EXTRA_TYPE_GSO - 1];
1013
1014                         if (unlikely(xennet_set_skb_gso(skb, gso))) {
1015                                 __skb_queue_head(&tmpq, skb);
1016                                 queue->rx.rsp_cons += skb_queue_len(&tmpq);
1017                                 goto err;
1018                         }
1019                 }
1020
1021                 NETFRONT_SKB_CB(skb)->pull_to = rx->status;
1022                 if (NETFRONT_SKB_CB(skb)->pull_to > RX_COPY_THRESHOLD)
1023                         NETFRONT_SKB_CB(skb)->pull_to = RX_COPY_THRESHOLD;
1024
1025                 skb_shinfo(skb)->frags[0].page_offset = rx->offset;
1026                 skb_frag_size_set(&skb_shinfo(skb)->frags[0], rx->status);
1027                 skb->data_len = rx->status;
1028                 skb->len += rx->status;
1029
1030                 i = xennet_fill_frags(queue, skb, &tmpq);
1031
1032                 if (rx->flags & XEN_NETRXF_csum_blank)
1033                         skb->ip_summed = CHECKSUM_PARTIAL;
1034                 else if (rx->flags & XEN_NETRXF_data_validated)
1035                         skb->ip_summed = CHECKSUM_UNNECESSARY;
1036
1037                 __skb_queue_tail(&rxq, skb);
1038
1039                 queue->rx.rsp_cons = ++i;
1040                 work_done++;
1041         }
1042
1043         __skb_queue_purge(&errq);
1044
1045         work_done -= handle_incoming_queue(queue, &rxq);
1046
1047         xennet_alloc_rx_buffers(queue);
1048
1049         if (work_done < budget) {
1050                 int more_to_do = 0;
1051
1052                 napi_complete(napi);
1053
1054                 RING_FINAL_CHECK_FOR_RESPONSES(&queue->rx, more_to_do);
1055                 if (more_to_do)
1056                         napi_schedule(napi);
1057         }
1058
1059         spin_unlock(&queue->rx_lock);
1060
1061         return work_done;
1062 }
1063
1064 static int xennet_change_mtu(struct net_device *dev, int mtu)
1065 {
1066         int max = xennet_can_sg(dev) ?
1067                 XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER : ETH_DATA_LEN;
1068
1069         if (mtu > max)
1070                 return -EINVAL;
1071         dev->mtu = mtu;
1072         return 0;
1073 }
1074
1075 static struct rtnl_link_stats64 *xennet_get_stats64(struct net_device *dev,
1076                                                     struct rtnl_link_stats64 *tot)
1077 {
1078         struct netfront_info *np = netdev_priv(dev);
1079         int cpu;
1080
1081         for_each_possible_cpu(cpu) {
1082                 struct netfront_stats *stats = per_cpu_ptr(np->stats, cpu);
1083                 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
1084                 unsigned int start;
1085
1086                 do {
1087                         start = u64_stats_fetch_begin_irq(&stats->syncp);
1088
1089                         rx_packets = stats->rx_packets;
1090                         tx_packets = stats->tx_packets;
1091                         rx_bytes = stats->rx_bytes;
1092                         tx_bytes = stats->tx_bytes;
1093                 } while (u64_stats_fetch_retry_irq(&stats->syncp, start));
1094
1095                 tot->rx_packets += rx_packets;
1096                 tot->tx_packets += tx_packets;
1097                 tot->rx_bytes   += rx_bytes;
1098                 tot->tx_bytes   += tx_bytes;
1099         }
1100
1101         tot->rx_errors  = dev->stats.rx_errors;
1102         tot->tx_dropped = dev->stats.tx_dropped;
1103
1104         return tot;
1105 }
1106
1107 static void xennet_release_tx_bufs(struct netfront_queue *queue)
1108 {
1109         struct sk_buff *skb;
1110         int i;
1111
1112         for (i = 0; i < NET_TX_RING_SIZE; i++) {
1113                 /* Skip over entries which are actually freelist references */
1114                 if (skb_entry_is_link(&queue->tx_skbs[i]))
1115                         continue;
1116
1117                 skb = queue->tx_skbs[i].skb;
1118                 get_page(queue->grant_tx_page[i]);
1119                 gnttab_end_foreign_access(queue->grant_tx_ref[i],
1120                                           GNTMAP_readonly,
1121                                           (unsigned long)page_address(queue->grant_tx_page[i]));
1122                 queue->grant_tx_page[i] = NULL;
1123                 queue->grant_tx_ref[i] = GRANT_INVALID_REF;
1124                 add_id_to_freelist(&queue->tx_skb_freelist, queue->tx_skbs, i);
1125                 dev_kfree_skb_irq(skb);
1126         }
1127 }
1128
1129 static void xennet_release_rx_bufs(struct netfront_queue *queue)
1130 {
1131         int id, ref;
1132
1133         spin_lock_bh(&queue->rx_lock);
1134
1135         for (id = 0; id < NET_RX_RING_SIZE; id++) {
1136                 struct sk_buff *skb;
1137                 struct page *page;
1138
1139                 skb = queue->rx_skbs[id];
1140                 if (!skb)
1141                         continue;
1142
1143                 ref = queue->grant_rx_ref[id];
1144                 if (ref == GRANT_INVALID_REF)
1145                         continue;
1146
1147                 page = skb_frag_page(&skb_shinfo(skb)->frags[0]);
1148
1149                 /* gnttab_end_foreign_access() needs a page ref until
1150                  * foreign access is ended (which may be deferred).
1151                  */
1152                 get_page(page);
1153                 gnttab_end_foreign_access(ref, 0,
1154                                           (unsigned long)page_address(page));
1155                 queue->grant_rx_ref[id] = GRANT_INVALID_REF;
1156
1157                 kfree_skb(skb);
1158         }
1159
1160         spin_unlock_bh(&queue->rx_lock);
1161 }
1162
1163 static netdev_features_t xennet_fix_features(struct net_device *dev,
1164         netdev_features_t features)
1165 {
1166         struct netfront_info *np = netdev_priv(dev);
1167         int val;
1168
1169         if (features & NETIF_F_SG) {
1170                 if (xenbus_scanf(XBT_NIL, np->xbdev->otherend, "feature-sg",
1171                                  "%d", &val) < 0)
1172                         val = 0;
1173
1174                 if (!val)
1175                         features &= ~NETIF_F_SG;
1176         }
1177
1178         if (features & NETIF_F_IPV6_CSUM) {
1179                 if (xenbus_scanf(XBT_NIL, np->xbdev->otherend,
1180                                  "feature-ipv6-csum-offload", "%d", &val) < 0)
1181                         val = 0;
1182
1183                 if (!val)
1184                         features &= ~NETIF_F_IPV6_CSUM;
1185         }
1186
1187         if (features & NETIF_F_TSO) {
1188                 if (xenbus_scanf(XBT_NIL, np->xbdev->otherend,
1189                                  "feature-gso-tcpv4", "%d", &val) < 0)
1190                         val = 0;
1191
1192                 if (!val)
1193                         features &= ~NETIF_F_TSO;
1194         }
1195
1196         if (features & NETIF_F_TSO6) {
1197                 if (xenbus_scanf(XBT_NIL, np->xbdev->otherend,
1198                                  "feature-gso-tcpv6", "%d", &val) < 0)
1199                         val = 0;
1200
1201                 if (!val)
1202                         features &= ~NETIF_F_TSO6;
1203         }
1204
1205         return features;
1206 }
1207
1208 static int xennet_set_features(struct net_device *dev,
1209         netdev_features_t features)
1210 {
1211         if (!(features & NETIF_F_SG) && dev->mtu > ETH_DATA_LEN) {
1212                 netdev_info(dev, "Reducing MTU because no SG offload");
1213                 dev->mtu = ETH_DATA_LEN;
1214         }
1215
1216         return 0;
1217 }
1218
1219 static irqreturn_t xennet_tx_interrupt(int irq, void *dev_id)
1220 {
1221         struct netfront_queue *queue = dev_id;
1222         unsigned long flags;
1223
1224         spin_lock_irqsave(&queue->tx_lock, flags);
1225         xennet_tx_buf_gc(queue);
1226         spin_unlock_irqrestore(&queue->tx_lock, flags);
1227
1228         return IRQ_HANDLED;
1229 }
1230
1231 static irqreturn_t xennet_rx_interrupt(int irq, void *dev_id)
1232 {
1233         struct netfront_queue *queue = dev_id;
1234         struct net_device *dev = queue->info->netdev;
1235
1236         if (likely(netif_carrier_ok(dev) &&
1237                    RING_HAS_UNCONSUMED_RESPONSES(&queue->rx)))
1238                 napi_schedule(&queue->napi);
1239
1240         return IRQ_HANDLED;
1241 }
1242
1243 static irqreturn_t xennet_interrupt(int irq, void *dev_id)
1244 {
1245         xennet_tx_interrupt(irq, dev_id);
1246         xennet_rx_interrupt(irq, dev_id);
1247         return IRQ_HANDLED;
1248 }
1249
1250 #ifdef CONFIG_NET_POLL_CONTROLLER
1251 static void xennet_poll_controller(struct net_device *dev)
1252 {
1253         /* Poll each queue */
1254         struct netfront_info *info = netdev_priv(dev);
1255         unsigned int num_queues = dev->real_num_tx_queues;
1256         unsigned int i;
1257         for (i = 0; i < num_queues; ++i)
1258                 xennet_interrupt(0, &info->queues[i]);
1259 }
1260 #endif
1261
1262 static const struct net_device_ops xennet_netdev_ops = {
1263         .ndo_open            = xennet_open,
1264         .ndo_stop            = xennet_close,
1265         .ndo_start_xmit      = xennet_start_xmit,
1266         .ndo_change_mtu      = xennet_change_mtu,
1267         .ndo_get_stats64     = xennet_get_stats64,
1268         .ndo_set_mac_address = eth_mac_addr,
1269         .ndo_validate_addr   = eth_validate_addr,
1270         .ndo_fix_features    = xennet_fix_features,
1271         .ndo_set_features    = xennet_set_features,
1272         .ndo_select_queue    = xennet_select_queue,
1273 #ifdef CONFIG_NET_POLL_CONTROLLER
1274         .ndo_poll_controller = xennet_poll_controller,
1275 #endif
1276 };
1277
1278 static struct net_device *xennet_create_dev(struct xenbus_device *dev)
1279 {
1280         int err;
1281         struct net_device *netdev;
1282         struct netfront_info *np;
1283
1284         netdev = alloc_etherdev_mq(sizeof(struct netfront_info), xennet_max_queues);
1285         if (!netdev)
1286                 return ERR_PTR(-ENOMEM);
1287
1288         np                   = netdev_priv(netdev);
1289         np->xbdev            = dev;
1290
1291         /* No need to use rtnl_lock() before the call below as it
1292          * happens before register_netdev().
1293          */
1294         netif_set_real_num_tx_queues(netdev, 0);
1295         np->queues = NULL;
1296
1297         err = -ENOMEM;
1298         np->stats = netdev_alloc_pcpu_stats(struct netfront_stats);
1299         if (np->stats == NULL)
1300                 goto exit;
1301
1302         netdev->netdev_ops      = &xennet_netdev_ops;
1303
1304         netdev->features        = NETIF_F_IP_CSUM | NETIF_F_RXCSUM |
1305                                   NETIF_F_GSO_ROBUST;
1306         netdev->hw_features     = NETIF_F_SG |
1307                                   NETIF_F_IPV6_CSUM |
1308                                   NETIF_F_TSO | NETIF_F_TSO6;
1309
1310         /*
1311          * Assume that all hw features are available for now. This set
1312          * will be adjusted by the call to netdev_update_features() in
1313          * xennet_connect() which is the earliest point where we can
1314          * negotiate with the backend regarding supported features.
1315          */
1316         netdev->features |= netdev->hw_features;
1317
1318         netdev->ethtool_ops = &xennet_ethtool_ops;
1319         SET_NETDEV_DEV(netdev, &dev->dev);
1320
1321         netif_set_gso_max_size(netdev, XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER);
1322
1323         np->netdev = netdev;
1324
1325         netif_carrier_off(netdev);
1326
1327         return netdev;
1328
1329  exit:
1330         free_netdev(netdev);
1331         return ERR_PTR(err);
1332 }
1333
1334 /**
1335  * Entry point to this code when a new device is created.  Allocate the basic
1336  * structures and the ring buffers for communication with the backend, and
1337  * inform the backend of the appropriate details for those.
1338  */
1339 static int netfront_probe(struct xenbus_device *dev,
1340                           const struct xenbus_device_id *id)
1341 {
1342         int err;
1343         struct net_device *netdev;
1344         struct netfront_info *info;
1345
1346         netdev = xennet_create_dev(dev);
1347         if (IS_ERR(netdev)) {
1348                 err = PTR_ERR(netdev);
1349                 xenbus_dev_fatal(dev, err, "creating netdev");
1350                 return err;
1351         }
1352
1353         info = netdev_priv(netdev);
1354         dev_set_drvdata(&dev->dev, info);
1355
1356         err = register_netdev(info->netdev);
1357         if (err) {
1358                 pr_warn("%s: register_netdev err=%d\n", __func__, err);
1359                 goto fail;
1360         }
1361
1362         err = xennet_sysfs_addif(info->netdev);
1363         if (err) {
1364                 unregister_netdev(info->netdev);
1365                 pr_warn("%s: add sysfs failed err=%d\n", __func__, err);
1366                 goto fail;
1367         }
1368
1369         return 0;
1370
1371  fail:
1372         free_netdev(netdev);
1373         dev_set_drvdata(&dev->dev, NULL);
1374         return err;
1375 }
1376
1377 static void xennet_end_access(int ref, void *page)
1378 {
1379         /* This frees the page as a side-effect */
1380         if (ref != GRANT_INVALID_REF)
1381                 gnttab_end_foreign_access(ref, 0, (unsigned long)page);
1382 }
1383
1384 static void xennet_disconnect_backend(struct netfront_info *info)
1385 {
1386         unsigned int i = 0;
1387         unsigned int num_queues = info->netdev->real_num_tx_queues;
1388
1389         netif_carrier_off(info->netdev);
1390
1391         for (i = 0; i < num_queues; ++i) {
1392                 struct netfront_queue *queue = &info->queues[i];
1393
1394                 if (queue->tx_irq && (queue->tx_irq == queue->rx_irq))
1395                         unbind_from_irqhandler(queue->tx_irq, queue);
1396                 if (queue->tx_irq && (queue->tx_irq != queue->rx_irq)) {
1397                         unbind_from_irqhandler(queue->tx_irq, queue);
1398                         unbind_from_irqhandler(queue->rx_irq, queue);
1399                 }
1400                 queue->tx_evtchn = queue->rx_evtchn = 0;
1401                 queue->tx_irq = queue->rx_irq = 0;
1402
1403                 napi_synchronize(&queue->napi);
1404
1405                 xennet_release_tx_bufs(queue);
1406                 xennet_release_rx_bufs(queue);
1407                 gnttab_free_grant_references(queue->gref_tx_head);
1408                 gnttab_free_grant_references(queue->gref_rx_head);
1409
1410                 /* End access and free the pages */
1411                 xennet_end_access(queue->tx_ring_ref, queue->tx.sring);
1412                 xennet_end_access(queue->rx_ring_ref, queue->rx.sring);
1413
1414                 queue->tx_ring_ref = GRANT_INVALID_REF;
1415                 queue->rx_ring_ref = GRANT_INVALID_REF;
1416                 queue->tx.sring = NULL;
1417                 queue->rx.sring = NULL;
1418         }
1419 }
1420
1421 /**
1422  * We are reconnecting to the backend, due to a suspend/resume, or a backend
1423  * driver restart.  We tear down our netif structure and recreate it, but
1424  * leave the device-layer structures intact so that this is transparent to the
1425  * rest of the kernel.
1426  */
1427 static int netfront_resume(struct xenbus_device *dev)
1428 {
1429         struct netfront_info *info = dev_get_drvdata(&dev->dev);
1430
1431         dev_dbg(&dev->dev, "%s\n", dev->nodename);
1432
1433         xennet_disconnect_backend(info);
1434         return 0;
1435 }
1436
1437 static int xen_net_read_mac(struct xenbus_device *dev, u8 mac[])
1438 {
1439         char *s, *e, *macstr;
1440         int i;
1441
1442         macstr = s = xenbus_read(XBT_NIL, dev->nodename, "mac", NULL);
1443         if (IS_ERR(macstr))
1444                 return PTR_ERR(macstr);
1445
1446         for (i = 0; i < ETH_ALEN; i++) {
1447                 mac[i] = simple_strtoul(s, &e, 16);
1448                 if ((s == e) || (*e != ((i == ETH_ALEN-1) ? '\0' : ':'))) {
1449                         kfree(macstr);
1450                         return -ENOENT;
1451                 }
1452                 s = e+1;
1453         }
1454
1455         kfree(macstr);
1456         return 0;
1457 }
1458
1459 static int setup_netfront_single(struct netfront_queue *queue)
1460 {
1461         int err;
1462
1463         err = xenbus_alloc_evtchn(queue->info->xbdev, &queue->tx_evtchn);
1464         if (err < 0)
1465                 goto fail;
1466
1467         err = bind_evtchn_to_irqhandler(queue->tx_evtchn,
1468                                         xennet_interrupt,
1469                                         0, queue->info->netdev->name, queue);
1470         if (err < 0)
1471                 goto bind_fail;
1472         queue->rx_evtchn = queue->tx_evtchn;
1473         queue->rx_irq = queue->tx_irq = err;
1474
1475         return 0;
1476
1477 bind_fail:
1478         xenbus_free_evtchn(queue->info->xbdev, queue->tx_evtchn);
1479         queue->tx_evtchn = 0;
1480 fail:
1481         return err;
1482 }
1483
1484 static int setup_netfront_split(struct netfront_queue *queue)
1485 {
1486         int err;
1487
1488         err = xenbus_alloc_evtchn(queue->info->xbdev, &queue->tx_evtchn);
1489         if (err < 0)
1490                 goto fail;
1491         err = xenbus_alloc_evtchn(queue->info->xbdev, &queue->rx_evtchn);
1492         if (err < 0)
1493                 goto alloc_rx_evtchn_fail;
1494
1495         snprintf(queue->tx_irq_name, sizeof(queue->tx_irq_name),
1496                  "%s-tx", queue->name);
1497         err = bind_evtchn_to_irqhandler(queue->tx_evtchn,
1498                                         xennet_tx_interrupt,
1499                                         0, queue->tx_irq_name, queue);
1500         if (err < 0)
1501                 goto bind_tx_fail;
1502         queue->tx_irq = err;
1503
1504         snprintf(queue->rx_irq_name, sizeof(queue->rx_irq_name),
1505                  "%s-rx", queue->name);
1506         err = bind_evtchn_to_irqhandler(queue->rx_evtchn,
1507                                         xennet_rx_interrupt,
1508                                         0, queue->rx_irq_name, queue);
1509         if (err < 0)
1510                 goto bind_rx_fail;
1511         queue->rx_irq = err;
1512
1513         return 0;
1514
1515 bind_rx_fail:
1516         unbind_from_irqhandler(queue->tx_irq, queue);
1517         queue->tx_irq = 0;
1518 bind_tx_fail:
1519         xenbus_free_evtchn(queue->info->xbdev, queue->rx_evtchn);
1520         queue->rx_evtchn = 0;
1521 alloc_rx_evtchn_fail:
1522         xenbus_free_evtchn(queue->info->xbdev, queue->tx_evtchn);
1523         queue->tx_evtchn = 0;
1524 fail:
1525         return err;
1526 }
1527
1528 static int setup_netfront(struct xenbus_device *dev,
1529                         struct netfront_queue *queue, unsigned int feature_split_evtchn)
1530 {
1531         struct xen_netif_tx_sring *txs;
1532         struct xen_netif_rx_sring *rxs;
1533         int err;
1534
1535         queue->tx_ring_ref = GRANT_INVALID_REF;
1536         queue->rx_ring_ref = GRANT_INVALID_REF;
1537         queue->rx.sring = NULL;
1538         queue->tx.sring = NULL;
1539
1540         txs = (struct xen_netif_tx_sring *)get_zeroed_page(GFP_NOIO | __GFP_HIGH);
1541         if (!txs) {
1542                 err = -ENOMEM;
1543                 xenbus_dev_fatal(dev, err, "allocating tx ring page");
1544                 goto fail;
1545         }
1546         SHARED_RING_INIT(txs);
1547         FRONT_RING_INIT(&queue->tx, txs, PAGE_SIZE);
1548
1549         err = xenbus_grant_ring(dev, virt_to_mfn(txs));
1550         if (err < 0)
1551                 goto grant_tx_ring_fail;
1552         queue->tx_ring_ref = err;
1553
1554         rxs = (struct xen_netif_rx_sring *)get_zeroed_page(GFP_NOIO | __GFP_HIGH);
1555         if (!rxs) {
1556                 err = -ENOMEM;
1557                 xenbus_dev_fatal(dev, err, "allocating rx ring page");
1558                 goto alloc_rx_ring_fail;
1559         }
1560         SHARED_RING_INIT(rxs);
1561         FRONT_RING_INIT(&queue->rx, rxs, PAGE_SIZE);
1562
1563         err = xenbus_grant_ring(dev, virt_to_mfn(rxs));
1564         if (err < 0)
1565                 goto grant_rx_ring_fail;
1566         queue->rx_ring_ref = err;
1567
1568         if (feature_split_evtchn)
1569                 err = setup_netfront_split(queue);
1570         /* setup single event channel if
1571          *  a) feature-split-event-channels == 0
1572          *  b) feature-split-event-channels == 1 but failed to setup
1573          */
1574         if (!feature_split_evtchn || (feature_split_evtchn && err))
1575                 err = setup_netfront_single(queue);
1576
1577         if (err)
1578                 goto alloc_evtchn_fail;
1579
1580         return 0;
1581
1582         /* If we fail to setup netfront, it is safe to just revoke access to
1583          * granted pages because backend is not accessing it at this point.
1584          */
1585 alloc_evtchn_fail:
1586         gnttab_end_foreign_access_ref(queue->rx_ring_ref, 0);
1587 grant_rx_ring_fail:
1588         free_page((unsigned long)rxs);
1589 alloc_rx_ring_fail:
1590         gnttab_end_foreign_access_ref(queue->tx_ring_ref, 0);
1591 grant_tx_ring_fail:
1592         free_page((unsigned long)txs);
1593 fail:
1594         return err;
1595 }
1596
1597 /* Queue-specific initialisation
1598  * This used to be done in xennet_create_dev() but must now
1599  * be run per-queue.
1600  */
1601 static int xennet_init_queue(struct netfront_queue *queue)
1602 {
1603         unsigned short i;
1604         int err = 0;
1605
1606         spin_lock_init(&queue->tx_lock);
1607         spin_lock_init(&queue->rx_lock);
1608
1609         init_timer(&queue->rx_refill_timer);
1610         queue->rx_refill_timer.data = (unsigned long)queue;
1611         queue->rx_refill_timer.function = rx_refill_timeout;
1612
1613         snprintf(queue->name, sizeof(queue->name), "%s-q%u",
1614                  queue->info->netdev->name, queue->id);
1615
1616         /* Initialise tx_skbs as a free chain containing every entry. */
1617         queue->tx_skb_freelist = 0;
1618         for (i = 0; i < NET_TX_RING_SIZE; i++) {
1619                 skb_entry_set_link(&queue->tx_skbs[i], i+1);
1620                 queue->grant_tx_ref[i] = GRANT_INVALID_REF;
1621                 queue->grant_tx_page[i] = NULL;
1622         }
1623
1624         /* Clear out rx_skbs */
1625         for (i = 0; i < NET_RX_RING_SIZE; i++) {
1626                 queue->rx_skbs[i] = NULL;
1627                 queue->grant_rx_ref[i] = GRANT_INVALID_REF;
1628         }
1629
1630         /* A grant for every tx ring slot */
1631         if (gnttab_alloc_grant_references(NET_TX_RING_SIZE,
1632                                           &queue->gref_tx_head) < 0) {
1633                 pr_alert("can't alloc tx grant refs\n");
1634                 err = -ENOMEM;
1635                 goto exit;
1636         }
1637
1638         /* A grant for every rx ring slot */
1639         if (gnttab_alloc_grant_references(NET_RX_RING_SIZE,
1640                                           &queue->gref_rx_head) < 0) {
1641                 pr_alert("can't alloc rx grant refs\n");
1642                 err = -ENOMEM;
1643                 goto exit_free_tx;
1644         }
1645
1646         return 0;
1647
1648  exit_free_tx:
1649         gnttab_free_grant_references(queue->gref_tx_head);
1650  exit:
1651         return err;
1652 }
1653
1654 static int write_queue_xenstore_keys(struct netfront_queue *queue,
1655                            struct xenbus_transaction *xbt, int write_hierarchical)
1656 {
1657         /* Write the queue-specific keys into XenStore in the traditional
1658          * way for a single queue, or in a queue subkeys for multiple
1659          * queues.
1660          */
1661         struct xenbus_device *dev = queue->info->xbdev;
1662         int err;
1663         const char *message;
1664         char *path;
1665         size_t pathsize;
1666
1667         /* Choose the correct place to write the keys */
1668         if (write_hierarchical) {
1669                 pathsize = strlen(dev->nodename) + 10;
1670                 path = kzalloc(pathsize, GFP_KERNEL);
1671                 if (!path) {
1672                         err = -ENOMEM;
1673                         message = "out of memory while writing ring references";
1674                         goto error;
1675                 }
1676                 snprintf(path, pathsize, "%s/queue-%u",
1677                                 dev->nodename, queue->id);
1678         } else {
1679                 path = (char *)dev->nodename;
1680         }
1681
1682         /* Write ring references */
1683         err = xenbus_printf(*xbt, path, "tx-ring-ref", "%u",
1684                         queue->tx_ring_ref);
1685         if (err) {
1686                 message = "writing tx-ring-ref";
1687                 goto error;
1688         }
1689
1690         err = xenbus_printf(*xbt, path, "rx-ring-ref", "%u",
1691                         queue->rx_ring_ref);
1692         if (err) {
1693                 message = "writing rx-ring-ref";
1694                 goto error;
1695         }
1696
1697         /* Write event channels; taking into account both shared
1698          * and split event channel scenarios.
1699          */
1700         if (queue->tx_evtchn == queue->rx_evtchn) {
1701                 /* Shared event channel */
1702                 err = xenbus_printf(*xbt, path,
1703                                 "event-channel", "%u", queue->tx_evtchn);
1704                 if (err) {
1705                         message = "writing event-channel";
1706                         goto error;
1707                 }
1708         } else {
1709                 /* Split event channels */
1710                 err = xenbus_printf(*xbt, path,
1711                                 "event-channel-tx", "%u", queue->tx_evtchn);
1712                 if (err) {
1713                         message = "writing event-channel-tx";
1714                         goto error;
1715                 }
1716
1717                 err = xenbus_printf(*xbt, path,
1718                                 "event-channel-rx", "%u", queue->rx_evtchn);
1719                 if (err) {
1720                         message = "writing event-channel-rx";
1721                         goto error;
1722                 }
1723         }
1724
1725         if (write_hierarchical)
1726                 kfree(path);
1727         return 0;
1728
1729 error:
1730         if (write_hierarchical)
1731                 kfree(path);
1732         xenbus_dev_fatal(dev, err, "%s", message);
1733         return err;
1734 }
1735
1736 static void xennet_destroy_queues(struct netfront_info *info)
1737 {
1738         unsigned int i;
1739
1740         rtnl_lock();
1741
1742         for (i = 0; i < info->netdev->real_num_tx_queues; i++) {
1743                 struct netfront_queue *queue = &info->queues[i];
1744
1745                 if (netif_running(info->netdev))
1746                         napi_disable(&queue->napi);
1747                 netif_napi_del(&queue->napi);
1748         }
1749
1750         rtnl_unlock();
1751
1752         kfree(info->queues);
1753         info->queues = NULL;
1754 }
1755
1756 static int xennet_create_queues(struct netfront_info *info,
1757                                 unsigned int num_queues)
1758 {
1759         unsigned int i;
1760         int ret;
1761
1762         info->queues = kcalloc(num_queues, sizeof(struct netfront_queue),
1763                                GFP_KERNEL);
1764         if (!info->queues)
1765                 return -ENOMEM;
1766
1767         rtnl_lock();
1768
1769         for (i = 0; i < num_queues; i++) {
1770                 struct netfront_queue *queue = &info->queues[i];
1771
1772                 queue->id = i;
1773                 queue->info = info;
1774
1775                 ret = xennet_init_queue(queue);
1776                 if (ret < 0) {
1777                         dev_warn(&info->netdev->dev,
1778                                  "only created %d queues\n", i);
1779                         num_queues = i;
1780                         break;
1781                 }
1782
1783                 netif_napi_add(queue->info->netdev, &queue->napi,
1784                                xennet_poll, 64);
1785                 if (netif_running(info->netdev))
1786                         napi_enable(&queue->napi);
1787         }
1788
1789         netif_set_real_num_tx_queues(info->netdev, num_queues);
1790
1791         rtnl_unlock();
1792
1793         if (num_queues == 0) {
1794                 dev_err(&info->netdev->dev, "no queues\n");
1795                 return -EINVAL;
1796         }
1797         return 0;
1798 }
1799
1800 /* Common code used when first setting up, and when resuming. */
1801 static int talk_to_netback(struct xenbus_device *dev,
1802                            struct netfront_info *info)
1803 {
1804         const char *message;
1805         struct xenbus_transaction xbt;
1806         int err;
1807         unsigned int feature_split_evtchn;
1808         unsigned int i = 0;
1809         unsigned int max_queues = 0;
1810         struct netfront_queue *queue = NULL;
1811         unsigned int num_queues = 1;
1812
1813         info->netdev->irq = 0;
1814
1815         /* Check if backend supports multiple queues */
1816         err = xenbus_scanf(XBT_NIL, info->xbdev->otherend,
1817                            "multi-queue-max-queues", "%u", &max_queues);
1818         if (err < 0)
1819                 max_queues = 1;
1820         num_queues = min(max_queues, xennet_max_queues);
1821
1822         /* Check feature-split-event-channels */
1823         err = xenbus_scanf(XBT_NIL, info->xbdev->otherend,
1824                            "feature-split-event-channels", "%u",
1825                            &feature_split_evtchn);
1826         if (err < 0)
1827                 feature_split_evtchn = 0;
1828
1829         /* Read mac addr. */
1830         err = xen_net_read_mac(dev, info->netdev->dev_addr);
1831         if (err) {
1832                 xenbus_dev_fatal(dev, err, "parsing %s/mac", dev->nodename);
1833                 goto out;
1834         }
1835
1836         if (info->queues)
1837                 xennet_destroy_queues(info);
1838
1839         err = xennet_create_queues(info, num_queues);
1840         if (err < 0)
1841                 goto destroy_ring;
1842
1843         /* Create shared ring, alloc event channel -- for each queue */
1844         for (i = 0; i < num_queues; ++i) {
1845                 queue = &info->queues[i];
1846                 err = setup_netfront(dev, queue, feature_split_evtchn);
1847                 if (err) {
1848                         /* setup_netfront() will tidy up the current
1849                          * queue on error, but we need to clean up
1850                          * those already allocated.
1851                          */
1852                         if (i > 0) {
1853                                 rtnl_lock();
1854                                 netif_set_real_num_tx_queues(info->netdev, i);
1855                                 rtnl_unlock();
1856                                 goto destroy_ring;
1857                         } else {
1858                                 goto out;
1859                         }
1860                 }
1861         }
1862
1863 again:
1864         err = xenbus_transaction_start(&xbt);
1865         if (err) {
1866                 xenbus_dev_fatal(dev, err, "starting transaction");
1867                 goto destroy_ring;
1868         }
1869
1870         if (num_queues == 1) {
1871                 err = write_queue_xenstore_keys(&info->queues[0], &xbt, 0); /* flat */
1872                 if (err)
1873                         goto abort_transaction_no_dev_fatal;
1874         } else {
1875                 /* Write the number of queues */
1876                 err = xenbus_printf(xbt, dev->nodename, "multi-queue-num-queues",
1877                                     "%u", num_queues);
1878                 if (err) {
1879                         message = "writing multi-queue-num-queues";
1880                         goto abort_transaction_no_dev_fatal;
1881                 }
1882
1883                 /* Write the keys for each queue */
1884                 for (i = 0; i < num_queues; ++i) {
1885                         queue = &info->queues[i];
1886                         err = write_queue_xenstore_keys(queue, &xbt, 1); /* hierarchical */
1887                         if (err)
1888                                 goto abort_transaction_no_dev_fatal;
1889                 }
1890         }
1891
1892         /* The remaining keys are not queue-specific */
1893         err = xenbus_printf(xbt, dev->nodename, "request-rx-copy", "%u",
1894                             1);
1895         if (err) {
1896                 message = "writing request-rx-copy";
1897                 goto abort_transaction;
1898         }
1899
1900         err = xenbus_printf(xbt, dev->nodename, "feature-rx-notify", "%d", 1);
1901         if (err) {
1902                 message = "writing feature-rx-notify";
1903                 goto abort_transaction;
1904         }
1905
1906         err = xenbus_printf(xbt, dev->nodename, "feature-sg", "%d", 1);
1907         if (err) {
1908                 message = "writing feature-sg";
1909                 goto abort_transaction;
1910         }
1911
1912         err = xenbus_printf(xbt, dev->nodename, "feature-gso-tcpv4", "%d", 1);
1913         if (err) {
1914                 message = "writing feature-gso-tcpv4";
1915                 goto abort_transaction;
1916         }
1917
1918         err = xenbus_write(xbt, dev->nodename, "feature-gso-tcpv6", "1");
1919         if (err) {
1920                 message = "writing feature-gso-tcpv6";
1921                 goto abort_transaction;
1922         }
1923
1924         err = xenbus_write(xbt, dev->nodename, "feature-ipv6-csum-offload",
1925                            "1");
1926         if (err) {
1927                 message = "writing feature-ipv6-csum-offload";
1928                 goto abort_transaction;
1929         }
1930
1931         err = xenbus_transaction_end(xbt, 0);
1932         if (err) {
1933                 if (err == -EAGAIN)
1934                         goto again;
1935                 xenbus_dev_fatal(dev, err, "completing transaction");
1936                 goto destroy_ring;
1937         }
1938
1939         return 0;
1940
1941  abort_transaction:
1942         xenbus_dev_fatal(dev, err, "%s", message);
1943 abort_transaction_no_dev_fatal:
1944         xenbus_transaction_end(xbt, 1);
1945  destroy_ring:
1946         xennet_disconnect_backend(info);
1947         kfree(info->queues);
1948         info->queues = NULL;
1949         rtnl_lock();
1950         netif_set_real_num_tx_queues(info->netdev, 0);
1951         rtnl_unlock();
1952  out:
1953         return err;
1954 }
1955
1956 static int xennet_connect(struct net_device *dev)
1957 {
1958         struct netfront_info *np = netdev_priv(dev);
1959         unsigned int num_queues = 0;
1960         int err;
1961         unsigned int feature_rx_copy;
1962         unsigned int j = 0;
1963         struct netfront_queue *queue = NULL;
1964
1965         err = xenbus_scanf(XBT_NIL, np->xbdev->otherend,
1966                            "feature-rx-copy", "%u", &feature_rx_copy);
1967         if (err != 1)
1968                 feature_rx_copy = 0;
1969
1970         if (!feature_rx_copy) {
1971                 dev_info(&dev->dev,
1972                          "backend does not support copying receive path\n");
1973                 return -ENODEV;
1974         }
1975
1976         err = talk_to_netback(np->xbdev, np);
1977         if (err)
1978                 return err;
1979
1980         /* talk_to_netback() sets the correct number of queues */
1981         num_queues = dev->real_num_tx_queues;
1982
1983         rtnl_lock();
1984         netdev_update_features(dev);
1985         rtnl_unlock();
1986
1987         /*
1988          * All public and private state should now be sane.  Get
1989          * ready to start sending and receiving packets and give the driver
1990          * domain a kick because we've probably just requeued some
1991          * packets.
1992          */
1993         netif_carrier_on(np->netdev);
1994         for (j = 0; j < num_queues; ++j) {
1995                 queue = &np->queues[j];
1996
1997                 notify_remote_via_irq(queue->tx_irq);
1998                 if (queue->tx_irq != queue->rx_irq)
1999                         notify_remote_via_irq(queue->rx_irq);
2000
2001                 spin_lock_irq(&queue->tx_lock);
2002                 xennet_tx_buf_gc(queue);
2003                 spin_unlock_irq(&queue->tx_lock);
2004
2005                 spin_lock_bh(&queue->rx_lock);
2006                 xennet_alloc_rx_buffers(queue);
2007                 spin_unlock_bh(&queue->rx_lock);
2008         }
2009
2010         return 0;
2011 }
2012
2013 /**
2014  * Callback received when the backend's state changes.
2015  */
2016 static void netback_changed(struct xenbus_device *dev,
2017                             enum xenbus_state backend_state)
2018 {
2019         struct netfront_info *np = dev_get_drvdata(&dev->dev);
2020         struct net_device *netdev = np->netdev;
2021
2022         dev_dbg(&dev->dev, "%s\n", xenbus_strstate(backend_state));
2023
2024         switch (backend_state) {
2025         case XenbusStateInitialising:
2026         case XenbusStateInitialised:
2027         case XenbusStateReconfiguring:
2028         case XenbusStateReconfigured:
2029         case XenbusStateUnknown:
2030                 break;
2031
2032         case XenbusStateInitWait:
2033                 if (dev->state != XenbusStateInitialising)
2034                         break;
2035                 if (xennet_connect(netdev) != 0)
2036                         break;
2037                 xenbus_switch_state(dev, XenbusStateConnected);
2038                 break;
2039
2040         case XenbusStateConnected:
2041                 netdev_notify_peers(netdev);
2042                 break;
2043
2044         case XenbusStateClosed:
2045                 if (dev->state == XenbusStateClosed)
2046                         break;
2047                 /* Missed the backend's CLOSING state -- fallthrough */
2048         case XenbusStateClosing:
2049                 xenbus_frontend_closed(dev);
2050                 break;
2051         }
2052 }
2053
2054 static const struct xennet_stat {
2055         char name[ETH_GSTRING_LEN];
2056         u16 offset;
2057 } xennet_stats[] = {
2058         {
2059                 "rx_gso_checksum_fixup",
2060                 offsetof(struct netfront_info, rx_gso_checksum_fixup)
2061         },
2062 };
2063
2064 static int xennet_get_sset_count(struct net_device *dev, int string_set)
2065 {
2066         switch (string_set) {
2067         case ETH_SS_STATS:
2068                 return ARRAY_SIZE(xennet_stats);
2069         default:
2070                 return -EINVAL;
2071         }
2072 }
2073
2074 static void xennet_get_ethtool_stats(struct net_device *dev,
2075                                      struct ethtool_stats *stats, u64 * data)
2076 {
2077         void *np = netdev_priv(dev);
2078         int i;
2079
2080         for (i = 0; i < ARRAY_SIZE(xennet_stats); i++)
2081                 data[i] = atomic_read((atomic_t *)(np + xennet_stats[i].offset));
2082 }
2083
2084 static void xennet_get_strings(struct net_device *dev, u32 stringset, u8 * data)
2085 {
2086         int i;
2087
2088         switch (stringset) {
2089         case ETH_SS_STATS:
2090                 for (i = 0; i < ARRAY_SIZE(xennet_stats); i++)
2091                         memcpy(data + i * ETH_GSTRING_LEN,
2092                                xennet_stats[i].name, ETH_GSTRING_LEN);
2093                 break;
2094         }
2095 }
2096
2097 static const struct ethtool_ops xennet_ethtool_ops =
2098 {
2099         .get_link = ethtool_op_get_link,
2100
2101         .get_sset_count = xennet_get_sset_count,
2102         .get_ethtool_stats = xennet_get_ethtool_stats,
2103         .get_strings = xennet_get_strings,
2104 };
2105
2106 #ifdef CONFIG_SYSFS
2107 static ssize_t show_rxbuf(struct device *dev,
2108                           struct device_attribute *attr, char *buf)
2109 {
2110         return sprintf(buf, "%lu\n", NET_RX_RING_SIZE);
2111 }
2112
2113 static ssize_t store_rxbuf(struct device *dev,
2114                            struct device_attribute *attr,
2115                            const char *buf, size_t len)
2116 {
2117         char *endp;
2118         unsigned long target;
2119
2120         if (!capable(CAP_NET_ADMIN))
2121                 return -EPERM;
2122
2123         target = simple_strtoul(buf, &endp, 0);
2124         if (endp == buf)
2125                 return -EBADMSG;
2126
2127         /* rxbuf_min and rxbuf_max are no longer configurable. */
2128
2129         return len;
2130 }
2131
2132 static struct device_attribute xennet_attrs[] = {
2133         __ATTR(rxbuf_min, S_IRUGO|S_IWUSR, show_rxbuf, store_rxbuf),
2134         __ATTR(rxbuf_max, S_IRUGO|S_IWUSR, show_rxbuf, store_rxbuf),
2135         __ATTR(rxbuf_cur, S_IRUGO, show_rxbuf, NULL),
2136 };
2137
2138 static int xennet_sysfs_addif(struct net_device *netdev)
2139 {
2140         int i;
2141         int err;
2142
2143         for (i = 0; i < ARRAY_SIZE(xennet_attrs); i++) {
2144                 err = device_create_file(&netdev->dev,
2145                                            &xennet_attrs[i]);
2146                 if (err)
2147                         goto fail;
2148         }
2149         return 0;
2150
2151  fail:
2152         while (--i >= 0)
2153                 device_remove_file(&netdev->dev, &xennet_attrs[i]);
2154         return err;
2155 }
2156
2157 static void xennet_sysfs_delif(struct net_device *netdev)
2158 {
2159         int i;
2160
2161         for (i = 0; i < ARRAY_SIZE(xennet_attrs); i++)
2162                 device_remove_file(&netdev->dev, &xennet_attrs[i]);
2163 }
2164
2165 #endif /* CONFIG_SYSFS */
2166
2167 static int xennet_remove(struct xenbus_device *dev)
2168 {
2169         struct netfront_info *info = dev_get_drvdata(&dev->dev);
2170         unsigned int num_queues = info->netdev->real_num_tx_queues;
2171         struct netfront_queue *queue = NULL;
2172         unsigned int i = 0;
2173
2174         dev_dbg(&dev->dev, "%s\n", dev->nodename);
2175
2176         xennet_disconnect_backend(info);
2177
2178         xennet_sysfs_delif(info->netdev);
2179
2180         unregister_netdev(info->netdev);
2181
2182         for (i = 0; i < num_queues; ++i) {
2183                 queue = &info->queues[i];
2184                 del_timer_sync(&queue->rx_refill_timer);
2185         }
2186
2187         if (num_queues) {
2188                 kfree(info->queues);
2189                 info->queues = NULL;
2190         }
2191
2192         free_percpu(info->stats);
2193
2194         free_netdev(info->netdev);
2195
2196         return 0;
2197 }
2198
2199 static const struct xenbus_device_id netfront_ids[] = {
2200         { "vif" },
2201         { "" }
2202 };
2203
2204 static struct xenbus_driver netfront_driver = {
2205         .ids = netfront_ids,
2206         .probe = netfront_probe,
2207         .remove = xennet_remove,
2208         .resume = netfront_resume,
2209         .otherend_changed = netback_changed,
2210 };
2211
2212 static int __init netif_init(void)
2213 {
2214         if (!xen_domain())
2215                 return -ENODEV;
2216
2217         if (!xen_has_pv_nic_devices())
2218                 return -ENODEV;
2219
2220         pr_info("Initialising Xen virtual ethernet driver\n");
2221
2222         /* Allow as many queues as there are CPUs, by default */
2223         xennet_max_queues = num_online_cpus();
2224
2225         return xenbus_register_frontend(&netfront_driver);
2226 }
2227 module_init(netif_init);
2228
2229
2230 static void __exit netif_exit(void)
2231 {
2232         xenbus_unregister_driver(&netfront_driver);
2233 }
2234 module_exit(netif_exit);
2235
2236 MODULE_DESCRIPTION("Xen virtual network device frontend");
2237 MODULE_LICENSE("GPL");
2238 MODULE_ALIAS("xen:vif");
2239 MODULE_ALIAS("xennet");