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