Loading...
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 */
55bool separate_tx_rx_irq = true;
56module_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 */
61unsigned int rx_drain_timeout_msecs = 10000;
62module_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 */
67unsigned int rx_stall_timeout_msecs = 60000;
68module_param(rx_stall_timeout_msecs, uint, 0444);
69
70#define MAX_QUEUES_DEFAULT 8
71unsigned int xenvif_max_queues;
72module_param_named(max_queues, xenvif_max_queues, uint, 0644);
73MODULE_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
81static unsigned int fatal_skb_slots = FATAL_SKB_SLOTS_DEFAULT;
82module_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
95unsigned int xenvif_hash_cache_size = XENVIF_HASH_CACHE_SIZE_DEFAULT;
96module_param_named(hash_cache_size, xenvif_hash_cache_size, uint, 0644);
97MODULE_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 */
103bool provides_xdp_headroom = true;
104module_param(provides_xdp_headroom, bool, 0644);
105
106static void xenvif_idx_release(struct xenvif_queue *queue, u16 pending_idx,
107 u8 status);
108
109static void make_tx_response(struct xenvif_queue *queue,
110 struct xen_netif_tx_request *txp,
111 unsigned int extra_count,
112 s8 st);
113static void push_tx_responses(struct xenvif_queue *queue);
114
115static void xenvif_idx_unmap(struct xenvif_queue *queue, u16 pending_idx);
116
117static inline int tx_work_todo(struct xenvif_queue *queue);
118
119static inline unsigned long idx_to_pfn(struct xenvif_queue *queue,
120 u16 idx)
121{
122 return page_to_pfn(queue->mmap_pages[idx]);
123}
124
125static inline unsigned long idx_to_kaddr(struct xenvif_queue *queue,
126 u16 idx)
127{
128 return (unsigned long)pfn_to_kaddr(idx_to_pfn(queue, idx));
129}
130
131#define callback_param(vif, pending_idx) \
132 (vif->pending_tx_info[pending_idx].callback_struct)
133
134/* Find the containing VIF's structure from a pointer in pending_tx_info array
135 */
136static inline struct xenvif_queue *ubuf_to_queue(const struct ubuf_info_msgzc *ubuf)
137{
138 u16 pending_idx = ubuf->desc;
139 struct pending_tx_info *temp =
140 container_of(ubuf, struct pending_tx_info, callback_struct);
141 return container_of(temp - pending_idx,
142 struct xenvif_queue,
143 pending_tx_info[0]);
144}
145
146static u16 frag_get_pending_idx(skb_frag_t *frag)
147{
148 return (u16)skb_frag_off(frag);
149}
150
151static void frag_set_pending_idx(skb_frag_t *frag, u16 pending_idx)
152{
153 skb_frag_off_set(frag, pending_idx);
154}
155
156static inline pending_ring_idx_t pending_index(unsigned i)
157{
158 return i & (MAX_PENDING_REQS-1);
159}
160
161void xenvif_kick_thread(struct xenvif_queue *queue)
162{
163 wake_up(&queue->wq);
164}
165
166void xenvif_napi_schedule_or_enable_events(struct xenvif_queue *queue)
167{
168 int more_to_do;
169
170 RING_FINAL_CHECK_FOR_REQUESTS(&queue->tx, more_to_do);
171
172 if (more_to_do)
173 napi_schedule(&queue->napi);
174 else if (atomic_fetch_andnot(NETBK_TX_EOI | NETBK_COMMON_EOI,
175 &queue->eoi_pending) &
176 (NETBK_TX_EOI | NETBK_COMMON_EOI))
177 xen_irq_lateeoi(queue->tx_irq, 0);
178}
179
180static void tx_add_credit(struct xenvif_queue *queue)
181{
182 unsigned long max_burst, max_credit;
183
184 /*
185 * Allow a burst big enough to transmit a jumbo packet of up to 128kB.
186 * Otherwise the interface can seize up due to insufficient credit.
187 */
188 max_burst = max(131072UL, queue->credit_bytes);
189
190 /* Take care that adding a new chunk of credit doesn't wrap to zero. */
191 max_credit = queue->remaining_credit + queue->credit_bytes;
192 if (max_credit < queue->remaining_credit)
193 max_credit = ULONG_MAX; /* wrapped: clamp to ULONG_MAX */
194
195 queue->remaining_credit = min(max_credit, max_burst);
196 queue->rate_limited = false;
197}
198
199void xenvif_tx_credit_callback(struct timer_list *t)
200{
201 struct xenvif_queue *queue = from_timer(queue, t, credit_timeout);
202 tx_add_credit(queue);
203 xenvif_napi_schedule_or_enable_events(queue);
204}
205
206static void xenvif_tx_err(struct xenvif_queue *queue,
207 struct xen_netif_tx_request *txp,
208 unsigned int extra_count, RING_IDX end)
209{
210 RING_IDX cons = queue->tx.req_cons;
211 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
226static 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
235static 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
334struct xenvif_tx_cb {
335 u16 copy_pending_idx[XEN_NETBK_LEGACY_SLOTS_MAX + 1];
336 u8 copy_count;
337};
338
339#define XENVIF_TX_CB(skb) ((struct xenvif_tx_cb *)(skb)->cb)
340#define copy_pending_idx(skb, i) (XENVIF_TX_CB(skb)->copy_pending_idx[i])
341#define copy_count(skb) (XENVIF_TX_CB(skb)->copy_count)
342
343static inline void xenvif_tx_create_map_op(struct xenvif_queue *queue,
344 u16 pending_idx,
345 struct xen_netif_tx_request *txp,
346 unsigned int extra_count,
347 struct gnttab_map_grant_ref *mop)
348{
349 queue->pages_to_map[mop-queue->tx_map_ops] = queue->mmap_pages[pending_idx];
350 gnttab_set_map_op(mop, idx_to_kaddr(queue, pending_idx),
351 GNTMAP_host_map | GNTMAP_readonly,
352 txp->gref, queue->vif->domid);
353
354 memcpy(&queue->pending_tx_info[pending_idx].req, txp,
355 sizeof(*txp));
356 queue->pending_tx_info[pending_idx].extra_count = extra_count;
357}
358
359static inline struct sk_buff *xenvif_alloc_skb(unsigned int size)
360{
361 struct sk_buff *skb =
362 alloc_skb(size + NET_SKB_PAD + NET_IP_ALIGN,
363 GFP_ATOMIC | __GFP_NOWARN);
364 if (unlikely(skb == NULL))
365 return NULL;
366
367 /* Packets passed to netif_rx() must have some headroom. */
368 skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN);
369
370 /* Initialize it here to avoid later surprises */
371 skb_shinfo(skb)->destructor_arg = NULL;
372
373 return skb;
374}
375
376static void xenvif_get_requests(struct xenvif_queue *queue,
377 struct sk_buff *skb,
378 struct xen_netif_tx_request *first,
379 struct xen_netif_tx_request *txfrags,
380 unsigned *copy_ops,
381 unsigned *map_ops,
382 unsigned int frag_overflow,
383 struct sk_buff *nskb,
384 unsigned int extra_count,
385 unsigned int data_len)
386{
387 struct skb_shared_info *shinfo = skb_shinfo(skb);
388 skb_frag_t *frags = shinfo->frags;
389 u16 pending_idx;
390 pending_ring_idx_t index;
391 unsigned int nr_slots;
392 struct gnttab_copy *cop = queue->tx_copy_ops + *copy_ops;
393 struct gnttab_map_grant_ref *gop = queue->tx_map_ops + *map_ops;
394 struct xen_netif_tx_request *txp = first;
395
396 nr_slots = shinfo->nr_frags + 1;
397
398 copy_count(skb) = 0;
399
400 /* Create copy ops for exactly data_len bytes into the skb head. */
401 __skb_put(skb, data_len);
402 while (data_len > 0) {
403 int amount = data_len > txp->size ? txp->size : data_len;
404
405 cop->source.u.ref = txp->gref;
406 cop->source.domid = queue->vif->domid;
407 cop->source.offset = txp->offset;
408
409 cop->dest.domid = DOMID_SELF;
410 cop->dest.offset = (offset_in_page(skb->data +
411 skb_headlen(skb) -
412 data_len)) & ~XEN_PAGE_MASK;
413 cop->dest.u.gmfn = virt_to_gfn(skb->data + skb_headlen(skb)
414 - data_len);
415
416 cop->len = amount;
417 cop->flags = GNTCOPY_source_gref;
418
419 index = pending_index(queue->pending_cons);
420 pending_idx = queue->pending_ring[index];
421 callback_param(queue, pending_idx).ctx = NULL;
422 copy_pending_idx(skb, copy_count(skb)) = pending_idx;
423 copy_count(skb)++;
424
425 cop++;
426 data_len -= amount;
427
428 if (amount == txp->size) {
429 /* The copy op covered the full tx_request */
430
431 memcpy(&queue->pending_tx_info[pending_idx].req,
432 txp, sizeof(*txp));
433 queue->pending_tx_info[pending_idx].extra_count =
434 (txp == first) ? extra_count : 0;
435
436 if (txp == first)
437 txp = txfrags;
438 else
439 txp++;
440 queue->pending_cons++;
441 nr_slots--;
442 } else {
443 /* The copy op partially covered the tx_request.
444 * The remainder will be mapped.
445 */
446 txp->offset += amount;
447 txp->size -= amount;
448 }
449 }
450
451 for (shinfo->nr_frags = 0; shinfo->nr_frags < nr_slots;
452 shinfo->nr_frags++, gop++) {
453 index = pending_index(queue->pending_cons++);
454 pending_idx = queue->pending_ring[index];
455 xenvif_tx_create_map_op(queue, pending_idx, txp,
456 txp == first ? extra_count : 0, gop);
457 frag_set_pending_idx(&frags[shinfo->nr_frags], pending_idx);
458
459 if (txp == first)
460 txp = txfrags;
461 else
462 txp++;
463 }
464
465 if (frag_overflow) {
466
467 shinfo = skb_shinfo(nskb);
468 frags = shinfo->frags;
469
470 for (shinfo->nr_frags = 0; shinfo->nr_frags < frag_overflow;
471 shinfo->nr_frags++, txp++, gop++) {
472 index = pending_index(queue->pending_cons++);
473 pending_idx = queue->pending_ring[index];
474 xenvif_tx_create_map_op(queue, pending_idx, txp, 0,
475 gop);
476 frag_set_pending_idx(&frags[shinfo->nr_frags],
477 pending_idx);
478 }
479
480 skb_shinfo(skb)->frag_list = nskb;
481 }
482
483 (*copy_ops) = cop - queue->tx_copy_ops;
484 (*map_ops) = gop - queue->tx_map_ops;
485}
486
487static inline void xenvif_grant_handle_set(struct xenvif_queue *queue,
488 u16 pending_idx,
489 grant_handle_t handle)
490{
491 if (unlikely(queue->grant_tx_handle[pending_idx] !=
492 NETBACK_INVALID_HANDLE)) {
493 netdev_err(queue->vif->dev,
494 "Trying to overwrite active handle! pending_idx: 0x%x\n",
495 pending_idx);
496 BUG();
497 }
498 queue->grant_tx_handle[pending_idx] = handle;
499}
500
501static inline void xenvif_grant_handle_reset(struct xenvif_queue *queue,
502 u16 pending_idx)
503{
504 if (unlikely(queue->grant_tx_handle[pending_idx] ==
505 NETBACK_INVALID_HANDLE)) {
506 netdev_err(queue->vif->dev,
507 "Trying to unmap invalid handle! pending_idx: 0x%x\n",
508 pending_idx);
509 BUG();
510 }
511 queue->grant_tx_handle[pending_idx] = NETBACK_INVALID_HANDLE;
512}
513
514static int xenvif_tx_check_gop(struct xenvif_queue *queue,
515 struct sk_buff *skb,
516 struct gnttab_map_grant_ref **gopp_map,
517 struct gnttab_copy **gopp_copy)
518{
519 struct gnttab_map_grant_ref *gop_map = *gopp_map;
520 u16 pending_idx;
521 /* This always points to the shinfo of the skb being checked, which
522 * could be either the first or the one on the frag_list
523 */
524 struct skb_shared_info *shinfo = skb_shinfo(skb);
525 /* If this is non-NULL, we are currently checking the frag_list skb, and
526 * this points to the shinfo of the first one
527 */
528 struct skb_shared_info *first_shinfo = NULL;
529 int nr_frags = shinfo->nr_frags;
530 const bool sharedslot = nr_frags &&
531 frag_get_pending_idx(&shinfo->frags[0]) ==
532 copy_pending_idx(skb, copy_count(skb) - 1);
533 int i, err = 0;
534
535 for (i = 0; i < copy_count(skb); i++) {
536 int newerr;
537
538 /* Check status of header. */
539 pending_idx = copy_pending_idx(skb, i);
540
541 newerr = (*gopp_copy)->status;
542 if (likely(!newerr)) {
543 /* The first frag might still have this slot mapped */
544 if (i < copy_count(skb) - 1 || !sharedslot)
545 xenvif_idx_release(queue, pending_idx,
546 XEN_NETIF_RSP_OKAY);
547 } else {
548 err = newerr;
549 if (net_ratelimit())
550 netdev_dbg(queue->vif->dev,
551 "Grant copy of header failed! status: %d pending_idx: %u ref: %u\n",
552 (*gopp_copy)->status,
553 pending_idx,
554 (*gopp_copy)->source.u.ref);
555 /* The first frag might still have this slot mapped */
556 if (i < copy_count(skb) - 1 || !sharedslot)
557 xenvif_idx_release(queue, pending_idx,
558 XEN_NETIF_RSP_ERROR);
559 }
560 (*gopp_copy)++;
561 }
562
563check_frags:
564 for (i = 0; i < nr_frags; i++, gop_map++) {
565 int j, newerr;
566
567 pending_idx = frag_get_pending_idx(&shinfo->frags[i]);
568
569 /* Check error status: if okay then remember grant handle. */
570 newerr = gop_map->status;
571
572 if (likely(!newerr)) {
573 xenvif_grant_handle_set(queue,
574 pending_idx,
575 gop_map->handle);
576 /* Had a previous error? Invalidate this fragment. */
577 if (unlikely(err)) {
578 xenvif_idx_unmap(queue, pending_idx);
579 /* If the mapping of the first frag was OK, but
580 * the header's copy failed, and they are
581 * sharing a slot, send an error
582 */
583 if (i == 0 && !first_shinfo && sharedslot)
584 xenvif_idx_release(queue, pending_idx,
585 XEN_NETIF_RSP_ERROR);
586 else
587 xenvif_idx_release(queue, pending_idx,
588 XEN_NETIF_RSP_OKAY);
589 }
590 continue;
591 }
592
593 /* Error on this fragment: respond to client with an error. */
594 if (net_ratelimit())
595 netdev_dbg(queue->vif->dev,
596 "Grant map of %d. frag failed! status: %d pending_idx: %u ref: %u\n",
597 i,
598 gop_map->status,
599 pending_idx,
600 gop_map->ref);
601
602 xenvif_idx_release(queue, pending_idx, XEN_NETIF_RSP_ERROR);
603
604 /* Not the first error? Preceding frags already invalidated. */
605 if (err)
606 continue;
607
608 /* Invalidate preceding fragments of this skb. */
609 for (j = 0; j < i; j++) {
610 pending_idx = frag_get_pending_idx(&shinfo->frags[j]);
611 xenvif_idx_unmap(queue, pending_idx);
612 xenvif_idx_release(queue, pending_idx,
613 XEN_NETIF_RSP_OKAY);
614 }
615
616 /* And if we found the error while checking the frag_list, unmap
617 * the first skb's frags
618 */
619 if (first_shinfo) {
620 for (j = 0; j < first_shinfo->nr_frags; j++) {
621 pending_idx = frag_get_pending_idx(&first_shinfo->frags[j]);
622 xenvif_idx_unmap(queue, pending_idx);
623 xenvif_idx_release(queue, pending_idx,
624 XEN_NETIF_RSP_OKAY);
625 }
626 }
627
628 /* Remember the error: invalidate all subsequent fragments. */
629 err = newerr;
630 }
631
632 if (skb_has_frag_list(skb) && !first_shinfo) {
633 first_shinfo = shinfo;
634 shinfo = skb_shinfo(shinfo->frag_list);
635 nr_frags = shinfo->nr_frags;
636
637 goto check_frags;
638 }
639
640 *gopp_map = gop_map;
641 return err;
642}
643
644static void xenvif_fill_frags(struct xenvif_queue *queue, struct sk_buff *skb)
645{
646 struct skb_shared_info *shinfo = skb_shinfo(skb);
647 int nr_frags = shinfo->nr_frags;
648 int i;
649 u16 prev_pending_idx = INVALID_PENDING_IDX;
650
651 for (i = 0; i < nr_frags; i++) {
652 skb_frag_t *frag = shinfo->frags + i;
653 struct xen_netif_tx_request *txp;
654 struct page *page;
655 u16 pending_idx;
656
657 pending_idx = frag_get_pending_idx(frag);
658
659 /* If this is not the first frag, chain it to the previous*/
660 if (prev_pending_idx == INVALID_PENDING_IDX)
661 skb_shinfo(skb)->destructor_arg =
662 &callback_param(queue, pending_idx);
663 else
664 callback_param(queue, prev_pending_idx).ctx =
665 &callback_param(queue, pending_idx);
666
667 callback_param(queue, pending_idx).ctx = NULL;
668 prev_pending_idx = pending_idx;
669
670 txp = &queue->pending_tx_info[pending_idx].req;
671 page = virt_to_page(idx_to_kaddr(queue, pending_idx));
672 __skb_fill_page_desc(skb, i, page, txp->offset, txp->size);
673 skb->len += txp->size;
674 skb->data_len += txp->size;
675 skb->truesize += txp->size;
676
677 /* Take an extra reference to offset network stack's put_page */
678 get_page(queue->mmap_pages[pending_idx]);
679 }
680}
681
682static int xenvif_get_extras(struct xenvif_queue *queue,
683 struct xen_netif_extra_info *extras,
684 unsigned int *extra_count,
685 int work_to_do)
686{
687 struct xen_netif_extra_info extra;
688 RING_IDX cons = queue->tx.req_cons;
689
690 do {
691 if (unlikely(work_to_do-- <= 0)) {
692 netdev_err(queue->vif->dev, "Missing extra info\n");
693 xenvif_fatal_tx_err(queue->vif);
694 return -EBADR;
695 }
696
697 RING_COPY_REQUEST(&queue->tx, cons, &extra);
698
699 queue->tx.req_cons = ++cons;
700 (*extra_count)++;
701
702 if (unlikely(!extra.type ||
703 extra.type >= XEN_NETIF_EXTRA_TYPE_MAX)) {
704 netdev_err(queue->vif->dev,
705 "Invalid extra type: %d\n", extra.type);
706 xenvif_fatal_tx_err(queue->vif);
707 return -EINVAL;
708 }
709
710 memcpy(&extras[extra.type - 1], &extra, sizeof(extra));
711 } while (extra.flags & XEN_NETIF_EXTRA_FLAG_MORE);
712
713 return work_to_do;
714}
715
716static int xenvif_set_skb_gso(struct xenvif *vif,
717 struct sk_buff *skb,
718 struct xen_netif_extra_info *gso)
719{
720 if (!gso->u.gso.size) {
721 netdev_err(vif->dev, "GSO size must not be zero.\n");
722 xenvif_fatal_tx_err(vif);
723 return -EINVAL;
724 }
725
726 switch (gso->u.gso.type) {
727 case XEN_NETIF_GSO_TYPE_TCPV4:
728 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
729 break;
730 case XEN_NETIF_GSO_TYPE_TCPV6:
731 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
732 break;
733 default:
734 netdev_err(vif->dev, "Bad GSO type %d.\n", gso->u.gso.type);
735 xenvif_fatal_tx_err(vif);
736 return -EINVAL;
737 }
738
739 skb_shinfo(skb)->gso_size = gso->u.gso.size;
740 /* gso_segs will be calculated later */
741
742 return 0;
743}
744
745static int checksum_setup(struct xenvif_queue *queue, struct sk_buff *skb)
746{
747 bool recalculate_partial_csum = false;
748
749 /* A GSO SKB must be CHECKSUM_PARTIAL. However some buggy
750 * peers can fail to set NETRXF_csum_blank when sending a GSO
751 * frame. In this case force the SKB to CHECKSUM_PARTIAL and
752 * recalculate the partial checksum.
753 */
754 if (skb->ip_summed != CHECKSUM_PARTIAL && skb_is_gso(skb)) {
755 queue->stats.rx_gso_checksum_fixup++;
756 skb->ip_summed = CHECKSUM_PARTIAL;
757 recalculate_partial_csum = true;
758 }
759
760 /* A non-CHECKSUM_PARTIAL SKB does not require setup. */
761 if (skb->ip_summed != CHECKSUM_PARTIAL)
762 return 0;
763
764 return skb_checksum_setup(skb, recalculate_partial_csum);
765}
766
767static bool tx_credit_exceeded(struct xenvif_queue *queue, unsigned size)
768{
769 u64 now = get_jiffies_64();
770 u64 next_credit = queue->credit_window_start +
771 msecs_to_jiffies(queue->credit_usec / 1000);
772
773 /* Timer could already be pending in rare cases. */
774 if (timer_pending(&queue->credit_timeout)) {
775 queue->rate_limited = true;
776 return true;
777 }
778
779 /* Passed the point where we can replenish credit? */
780 if (time_after_eq64(now, next_credit)) {
781 queue->credit_window_start = now;
782 tx_add_credit(queue);
783 }
784
785 /* Still too big to send right now? Set a callback. */
786 if (size > queue->remaining_credit) {
787 mod_timer(&queue->credit_timeout,
788 next_credit);
789 queue->credit_window_start = next_credit;
790 queue->rate_limited = true;
791
792 return true;
793 }
794
795 return false;
796}
797
798/* No locking is required in xenvif_mcast_add/del() as they are
799 * only ever invoked from NAPI poll. An RCU list is used because
800 * xenvif_mcast_match() is called asynchronously, during start_xmit.
801 */
802
803static int xenvif_mcast_add(struct xenvif *vif, const u8 *addr)
804{
805 struct xenvif_mcast_addr *mcast;
806
807 if (vif->fe_mcast_count == XEN_NETBK_MCAST_MAX) {
808 if (net_ratelimit())
809 netdev_err(vif->dev,
810 "Too many multicast addresses\n");
811 return -ENOSPC;
812 }
813
814 mcast = kzalloc(sizeof(*mcast), GFP_ATOMIC);
815 if (!mcast)
816 return -ENOMEM;
817
818 ether_addr_copy(mcast->addr, addr);
819 list_add_tail_rcu(&mcast->entry, &vif->fe_mcast_addr);
820 vif->fe_mcast_count++;
821
822 return 0;
823}
824
825static void xenvif_mcast_del(struct xenvif *vif, const u8 *addr)
826{
827 struct xenvif_mcast_addr *mcast;
828
829 list_for_each_entry_rcu(mcast, &vif->fe_mcast_addr, entry) {
830 if (ether_addr_equal(addr, mcast->addr)) {
831 --vif->fe_mcast_count;
832 list_del_rcu(&mcast->entry);
833 kfree_rcu(mcast, rcu);
834 break;
835 }
836 }
837}
838
839bool xenvif_mcast_match(struct xenvif *vif, const u8 *addr)
840{
841 struct xenvif_mcast_addr *mcast;
842
843 rcu_read_lock();
844 list_for_each_entry_rcu(mcast, &vif->fe_mcast_addr, entry) {
845 if (ether_addr_equal(addr, mcast->addr)) {
846 rcu_read_unlock();
847 return true;
848 }
849 }
850 rcu_read_unlock();
851
852 return false;
853}
854
855void xenvif_mcast_addr_list_free(struct xenvif *vif)
856{
857 /* No need for locking or RCU here. NAPI poll and TX queue
858 * are stopped.
859 */
860 while (!list_empty(&vif->fe_mcast_addr)) {
861 struct xenvif_mcast_addr *mcast;
862
863 mcast = list_first_entry(&vif->fe_mcast_addr,
864 struct xenvif_mcast_addr,
865 entry);
866 --vif->fe_mcast_count;
867 list_del(&mcast->entry);
868 kfree(mcast);
869 }
870}
871
872static void xenvif_tx_build_gops(struct xenvif_queue *queue,
873 int budget,
874 unsigned *copy_ops,
875 unsigned *map_ops)
876{
877 struct sk_buff *skb, *nskb;
878 int ret;
879 unsigned int frag_overflow;
880
881 while (skb_queue_len(&queue->tx_queue) < budget) {
882 struct xen_netif_tx_request txreq;
883 struct xen_netif_tx_request txfrags[XEN_NETBK_LEGACY_SLOTS_MAX];
884 struct xen_netif_extra_info extras[XEN_NETIF_EXTRA_TYPE_MAX-1];
885 unsigned int extra_count;
886 u16 pending_idx;
887 RING_IDX idx;
888 int work_to_do;
889 unsigned int data_len;
890 pending_ring_idx_t index;
891
892 if (queue->tx.sring->req_prod - queue->tx.req_cons >
893 XEN_NETIF_TX_RING_SIZE) {
894 netdev_err(queue->vif->dev,
895 "Impossible number of requests. "
896 "req_prod %d, req_cons %d, size %ld\n",
897 queue->tx.sring->req_prod, queue->tx.req_cons,
898 XEN_NETIF_TX_RING_SIZE);
899 xenvif_fatal_tx_err(queue->vif);
900 break;
901 }
902
903 work_to_do = XEN_RING_NR_UNCONSUMED_REQUESTS(&queue->tx);
904 if (!work_to_do)
905 break;
906
907 idx = queue->tx.req_cons;
908 rmb(); /* Ensure that we see the request before we copy it. */
909 RING_COPY_REQUEST(&queue->tx, idx, &txreq);
910
911 /* Credit-based scheduling. */
912 if (txreq.size > queue->remaining_credit &&
913 tx_credit_exceeded(queue, txreq.size))
914 break;
915
916 queue->remaining_credit -= txreq.size;
917
918 work_to_do--;
919 queue->tx.req_cons = ++idx;
920
921 memset(extras, 0, sizeof(extras));
922 extra_count = 0;
923 if (txreq.flags & XEN_NETTXF_extra_info) {
924 work_to_do = xenvif_get_extras(queue, extras,
925 &extra_count,
926 work_to_do);
927 idx = queue->tx.req_cons;
928 if (unlikely(work_to_do < 0))
929 break;
930 }
931
932 if (extras[XEN_NETIF_EXTRA_TYPE_MCAST_ADD - 1].type) {
933 struct xen_netif_extra_info *extra;
934
935 extra = &extras[XEN_NETIF_EXTRA_TYPE_MCAST_ADD - 1];
936 ret = xenvif_mcast_add(queue->vif, extra->u.mcast.addr);
937
938 make_tx_response(queue, &txreq, extra_count,
939 (ret == 0) ?
940 XEN_NETIF_RSP_OKAY :
941 XEN_NETIF_RSP_ERROR);
942 push_tx_responses(queue);
943 continue;
944 }
945
946 if (extras[XEN_NETIF_EXTRA_TYPE_MCAST_DEL - 1].type) {
947 struct xen_netif_extra_info *extra;
948
949 extra = &extras[XEN_NETIF_EXTRA_TYPE_MCAST_DEL - 1];
950 xenvif_mcast_del(queue->vif, extra->u.mcast.addr);
951
952 make_tx_response(queue, &txreq, extra_count,
953 XEN_NETIF_RSP_OKAY);
954 push_tx_responses(queue);
955 continue;
956 }
957
958 data_len = (txreq.size > XEN_NETBACK_TX_COPY_LEN) ?
959 XEN_NETBACK_TX_COPY_LEN : txreq.size;
960
961 ret = xenvif_count_requests(queue, &txreq, extra_count,
962 txfrags, work_to_do);
963
964 if (unlikely(ret < 0))
965 break;
966
967 idx += ret;
968
969 if (unlikely(txreq.size < ETH_HLEN)) {
970 netdev_dbg(queue->vif->dev,
971 "Bad packet size: %d\n", txreq.size);
972 xenvif_tx_err(queue, &txreq, extra_count, idx);
973 break;
974 }
975
976 /* No crossing a page as the payload mustn't fragment. */
977 if (unlikely((txreq.offset + txreq.size) > XEN_PAGE_SIZE)) {
978 netdev_err(queue->vif->dev,
979 "txreq.offset: %u, size: %u, end: %lu\n",
980 txreq.offset, txreq.size,
981 (unsigned long)(txreq.offset&~XEN_PAGE_MASK) + txreq.size);
982 xenvif_fatal_tx_err(queue->vif);
983 break;
984 }
985
986 index = pending_index(queue->pending_cons);
987 pending_idx = queue->pending_ring[index];
988
989 if (ret >= XEN_NETBK_LEGACY_SLOTS_MAX - 1 && data_len < txreq.size)
990 data_len = txreq.size;
991
992 skb = xenvif_alloc_skb(data_len);
993 if (unlikely(skb == NULL)) {
994 netdev_dbg(queue->vif->dev,
995 "Can't allocate a skb in start_xmit.\n");
996 xenvif_tx_err(queue, &txreq, extra_count, idx);
997 break;
998 }
999
1000 skb_shinfo(skb)->nr_frags = ret;
1001 /* At this point shinfo->nr_frags is in fact the number of
1002 * slots, which can be as large as XEN_NETBK_LEGACY_SLOTS_MAX.
1003 */
1004 frag_overflow = 0;
1005 nskb = NULL;
1006 if (skb_shinfo(skb)->nr_frags > MAX_SKB_FRAGS) {
1007 frag_overflow = skb_shinfo(skb)->nr_frags - MAX_SKB_FRAGS;
1008 BUG_ON(frag_overflow > MAX_SKB_FRAGS);
1009 skb_shinfo(skb)->nr_frags = MAX_SKB_FRAGS;
1010 nskb = xenvif_alloc_skb(0);
1011 if (unlikely(nskb == NULL)) {
1012 skb_shinfo(skb)->nr_frags = 0;
1013 kfree_skb(skb);
1014 xenvif_tx_err(queue, &txreq, extra_count, idx);
1015 if (net_ratelimit())
1016 netdev_err(queue->vif->dev,
1017 "Can't allocate the frag_list skb.\n");
1018 break;
1019 }
1020 }
1021
1022 if (extras[XEN_NETIF_EXTRA_TYPE_GSO - 1].type) {
1023 struct xen_netif_extra_info *gso;
1024 gso = &extras[XEN_NETIF_EXTRA_TYPE_GSO - 1];
1025
1026 if (xenvif_set_skb_gso(queue->vif, skb, gso)) {
1027 /* Failure in xenvif_set_skb_gso is fatal. */
1028 skb_shinfo(skb)->nr_frags = 0;
1029 kfree_skb(skb);
1030 kfree_skb(nskb);
1031 break;
1032 }
1033 }
1034
1035 if (extras[XEN_NETIF_EXTRA_TYPE_HASH - 1].type) {
1036 struct xen_netif_extra_info *extra;
1037 enum pkt_hash_types type = PKT_HASH_TYPE_NONE;
1038
1039 extra = &extras[XEN_NETIF_EXTRA_TYPE_HASH - 1];
1040
1041 switch (extra->u.hash.type) {
1042 case _XEN_NETIF_CTRL_HASH_TYPE_IPV4:
1043 case _XEN_NETIF_CTRL_HASH_TYPE_IPV6:
1044 type = PKT_HASH_TYPE_L3;
1045 break;
1046
1047 case _XEN_NETIF_CTRL_HASH_TYPE_IPV4_TCP:
1048 case _XEN_NETIF_CTRL_HASH_TYPE_IPV6_TCP:
1049 type = PKT_HASH_TYPE_L4;
1050 break;
1051
1052 default:
1053 break;
1054 }
1055
1056 if (type != PKT_HASH_TYPE_NONE)
1057 skb_set_hash(skb,
1058 *(u32 *)extra->u.hash.value,
1059 type);
1060 }
1061
1062 xenvif_get_requests(queue, skb, &txreq, txfrags, copy_ops,
1063 map_ops, frag_overflow, nskb, extra_count,
1064 data_len);
1065
1066 __skb_queue_tail(&queue->tx_queue, skb);
1067
1068 queue->tx.req_cons = idx;
1069
1070 if ((*map_ops >= ARRAY_SIZE(queue->tx_map_ops)) ||
1071 (*copy_ops >= ARRAY_SIZE(queue->tx_copy_ops)))
1072 break;
1073 }
1074
1075 return;
1076}
1077
1078/* Consolidate skb with a frag_list into a brand new one with local pages on
1079 * frags. Returns 0 or -ENOMEM if can't allocate new pages.
1080 */
1081static int xenvif_handle_frag_list(struct xenvif_queue *queue, struct sk_buff *skb)
1082{
1083 unsigned int offset = skb_headlen(skb);
1084 skb_frag_t frags[MAX_SKB_FRAGS];
1085 int i, f;
1086 struct ubuf_info *uarg;
1087 struct sk_buff *nskb = skb_shinfo(skb)->frag_list;
1088
1089 queue->stats.tx_zerocopy_sent += 2;
1090 queue->stats.tx_frag_overflow++;
1091
1092 xenvif_fill_frags(queue, nskb);
1093 /* Subtract frags size, we will correct it later */
1094 skb->truesize -= skb->data_len;
1095 skb->len += nskb->len;
1096 skb->data_len += nskb->len;
1097
1098 /* create a brand new frags array and coalesce there */
1099 for (i = 0; offset < skb->len; i++) {
1100 struct page *page;
1101 unsigned int len;
1102
1103 BUG_ON(i >= MAX_SKB_FRAGS);
1104 page = alloc_page(GFP_ATOMIC);
1105 if (!page) {
1106 int j;
1107 skb->truesize += skb->data_len;
1108 for (j = 0; j < i; j++)
1109 put_page(skb_frag_page(&frags[j]));
1110 return -ENOMEM;
1111 }
1112
1113 if (offset + PAGE_SIZE < skb->len)
1114 len = PAGE_SIZE;
1115 else
1116 len = skb->len - offset;
1117 if (skb_copy_bits(skb, offset, page_address(page), len))
1118 BUG();
1119
1120 offset += len;
1121 __skb_frag_set_page(&frags[i], page);
1122 skb_frag_off_set(&frags[i], 0);
1123 skb_frag_size_set(&frags[i], len);
1124 }
1125
1126 /* Release all the original (foreign) frags. */
1127 for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
1128 skb_frag_unref(skb, f);
1129 uarg = skb_shinfo(skb)->destructor_arg;
1130 /* increase inflight counter to offset decrement in callback */
1131 atomic_inc(&queue->inflight_packets);
1132 uarg->callback(NULL, uarg, true);
1133 skb_shinfo(skb)->destructor_arg = NULL;
1134
1135 /* Fill the skb with the new (local) frags. */
1136 memcpy(skb_shinfo(skb)->frags, frags, i * sizeof(skb_frag_t));
1137 skb_shinfo(skb)->nr_frags = i;
1138 skb->truesize += i * PAGE_SIZE;
1139
1140 return 0;
1141}
1142
1143static int xenvif_tx_submit(struct xenvif_queue *queue)
1144{
1145 struct gnttab_map_grant_ref *gop_map = queue->tx_map_ops;
1146 struct gnttab_copy *gop_copy = queue->tx_copy_ops;
1147 struct sk_buff *skb;
1148 int work_done = 0;
1149
1150 while ((skb = __skb_dequeue(&queue->tx_queue)) != NULL) {
1151 struct xen_netif_tx_request *txp;
1152 u16 pending_idx;
1153
1154 pending_idx = copy_pending_idx(skb, 0);
1155 txp = &queue->pending_tx_info[pending_idx].req;
1156
1157 /* Check the remap error code. */
1158 if (unlikely(xenvif_tx_check_gop(queue, skb, &gop_map, &gop_copy))) {
1159 /* If there was an error, xenvif_tx_check_gop is
1160 * expected to release all the frags which were mapped,
1161 * so kfree_skb shouldn't do it again
1162 */
1163 skb_shinfo(skb)->nr_frags = 0;
1164 if (skb_has_frag_list(skb)) {
1165 struct sk_buff *nskb =
1166 skb_shinfo(skb)->frag_list;
1167 skb_shinfo(nskb)->nr_frags = 0;
1168 }
1169 kfree_skb(skb);
1170 continue;
1171 }
1172
1173 if (txp->flags & XEN_NETTXF_csum_blank)
1174 skb->ip_summed = CHECKSUM_PARTIAL;
1175 else if (txp->flags & XEN_NETTXF_data_validated)
1176 skb->ip_summed = CHECKSUM_UNNECESSARY;
1177
1178 xenvif_fill_frags(queue, skb);
1179
1180 if (unlikely(skb_has_frag_list(skb))) {
1181 struct sk_buff *nskb = skb_shinfo(skb)->frag_list;
1182 xenvif_skb_zerocopy_prepare(queue, nskb);
1183 if (xenvif_handle_frag_list(queue, skb)) {
1184 if (net_ratelimit())
1185 netdev_err(queue->vif->dev,
1186 "Not enough memory to consolidate frag_list!\n");
1187 xenvif_skb_zerocopy_prepare(queue, skb);
1188 kfree_skb(skb);
1189 continue;
1190 }
1191 /* Copied all the bits from the frag list -- free it. */
1192 skb_frag_list_init(skb);
1193 kfree_skb(nskb);
1194 }
1195
1196 skb->dev = queue->vif->dev;
1197 skb->protocol = eth_type_trans(skb, skb->dev);
1198 skb_reset_network_header(skb);
1199
1200 if (checksum_setup(queue, skb)) {
1201 netdev_dbg(queue->vif->dev,
1202 "Can't setup checksum in net_tx_action\n");
1203 /* We have to set this flag to trigger the callback */
1204 if (skb_shinfo(skb)->destructor_arg)
1205 xenvif_skb_zerocopy_prepare(queue, skb);
1206 kfree_skb(skb);
1207 continue;
1208 }
1209
1210 skb_probe_transport_header(skb);
1211
1212 /* If the packet is GSO then we will have just set up the
1213 * transport header offset in checksum_setup so it's now
1214 * straightforward to calculate gso_segs.
1215 */
1216 if (skb_is_gso(skb)) {
1217 int mss, hdrlen;
1218
1219 /* GSO implies having the L4 header. */
1220 WARN_ON_ONCE(!skb_transport_header_was_set(skb));
1221 if (unlikely(!skb_transport_header_was_set(skb))) {
1222 kfree_skb(skb);
1223 continue;
1224 }
1225
1226 mss = skb_shinfo(skb)->gso_size;
1227 hdrlen = skb_tcp_all_headers(skb);
1228
1229 skb_shinfo(skb)->gso_segs =
1230 DIV_ROUND_UP(skb->len - hdrlen, mss);
1231 }
1232
1233 queue->stats.rx_bytes += skb->len;
1234 queue->stats.rx_packets++;
1235
1236 work_done++;
1237
1238 /* Set this flag right before netif_receive_skb, otherwise
1239 * someone might think this packet already left netback, and
1240 * do a skb_copy_ubufs while we are still in control of the
1241 * skb. E.g. the __pskb_pull_tail earlier can do such thing.
1242 */
1243 if (skb_shinfo(skb)->destructor_arg) {
1244 xenvif_skb_zerocopy_prepare(queue, skb);
1245 queue->stats.tx_zerocopy_sent++;
1246 }
1247
1248 netif_receive_skb(skb);
1249 }
1250
1251 return work_done;
1252}
1253
1254void xenvif_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *ubuf_base,
1255 bool zerocopy_success)
1256{
1257 unsigned long flags;
1258 pending_ring_idx_t index;
1259 struct ubuf_info_msgzc *ubuf = uarg_to_msgzc(ubuf_base);
1260 struct xenvif_queue *queue = ubuf_to_queue(ubuf);
1261
1262 /* This is the only place where we grab this lock, to protect callbacks
1263 * from each other.
1264 */
1265 spin_lock_irqsave(&queue->callback_lock, flags);
1266 do {
1267 u16 pending_idx = ubuf->desc;
1268 ubuf = (struct ubuf_info_msgzc *) ubuf->ctx;
1269 BUG_ON(queue->dealloc_prod - queue->dealloc_cons >=
1270 MAX_PENDING_REQS);
1271 index = pending_index(queue->dealloc_prod);
1272 queue->dealloc_ring[index] = pending_idx;
1273 /* Sync with xenvif_tx_dealloc_action:
1274 * insert idx then incr producer.
1275 */
1276 smp_wmb();
1277 queue->dealloc_prod++;
1278 } while (ubuf);
1279 spin_unlock_irqrestore(&queue->callback_lock, flags);
1280
1281 if (likely(zerocopy_success))
1282 queue->stats.tx_zerocopy_success++;
1283 else
1284 queue->stats.tx_zerocopy_fail++;
1285 xenvif_skb_zerocopy_complete(queue);
1286}
1287
1288static inline void xenvif_tx_dealloc_action(struct xenvif_queue *queue)
1289{
1290 struct gnttab_unmap_grant_ref *gop;
1291 pending_ring_idx_t dc, dp;
1292 u16 pending_idx, pending_idx_release[MAX_PENDING_REQS];
1293 unsigned int i = 0;
1294
1295 dc = queue->dealloc_cons;
1296 gop = queue->tx_unmap_ops;
1297
1298 /* Free up any grants we have finished using */
1299 do {
1300 dp = queue->dealloc_prod;
1301
1302 /* Ensure we see all indices enqueued by all
1303 * xenvif_zerocopy_callback().
1304 */
1305 smp_rmb();
1306
1307 while (dc != dp) {
1308 BUG_ON(gop - queue->tx_unmap_ops >= MAX_PENDING_REQS);
1309 pending_idx =
1310 queue->dealloc_ring[pending_index(dc++)];
1311
1312 pending_idx_release[gop - queue->tx_unmap_ops] =
1313 pending_idx;
1314 queue->pages_to_unmap[gop - queue->tx_unmap_ops] =
1315 queue->mmap_pages[pending_idx];
1316 gnttab_set_unmap_op(gop,
1317 idx_to_kaddr(queue, pending_idx),
1318 GNTMAP_host_map,
1319 queue->grant_tx_handle[pending_idx]);
1320 xenvif_grant_handle_reset(queue, pending_idx);
1321 ++gop;
1322 }
1323
1324 } while (dp != queue->dealloc_prod);
1325
1326 queue->dealloc_cons = dc;
1327
1328 if (gop - queue->tx_unmap_ops > 0) {
1329 int ret;
1330 ret = gnttab_unmap_refs(queue->tx_unmap_ops,
1331 NULL,
1332 queue->pages_to_unmap,
1333 gop - queue->tx_unmap_ops);
1334 if (ret) {
1335 netdev_err(queue->vif->dev, "Unmap fail: nr_ops %tu ret %d\n",
1336 gop - queue->tx_unmap_ops, ret);
1337 for (i = 0; i < gop - queue->tx_unmap_ops; ++i) {
1338 if (gop[i].status != GNTST_okay)
1339 netdev_err(queue->vif->dev,
1340 " host_addr: 0x%llx handle: 0x%x status: %d\n",
1341 gop[i].host_addr,
1342 gop[i].handle,
1343 gop[i].status);
1344 }
1345 BUG();
1346 }
1347 }
1348
1349 for (i = 0; i < gop - queue->tx_unmap_ops; ++i)
1350 xenvif_idx_release(queue, pending_idx_release[i],
1351 XEN_NETIF_RSP_OKAY);
1352}
1353
1354
1355/* Called after netfront has transmitted */
1356int xenvif_tx_action(struct xenvif_queue *queue, int budget)
1357{
1358 unsigned nr_mops = 0, nr_cops = 0;
1359 int work_done, ret;
1360
1361 if (unlikely(!tx_work_todo(queue)))
1362 return 0;
1363
1364 xenvif_tx_build_gops(queue, budget, &nr_cops, &nr_mops);
1365
1366 if (nr_cops == 0)
1367 return 0;
1368
1369 gnttab_batch_copy(queue->tx_copy_ops, nr_cops);
1370 if (nr_mops != 0) {
1371 ret = gnttab_map_refs(queue->tx_map_ops,
1372 NULL,
1373 queue->pages_to_map,
1374 nr_mops);
1375 if (ret) {
1376 unsigned int i;
1377
1378 netdev_err(queue->vif->dev, "Map fail: nr %u ret %d\n",
1379 nr_mops, ret);
1380 for (i = 0; i < nr_mops; ++i)
1381 WARN_ON_ONCE(queue->tx_map_ops[i].status ==
1382 GNTST_okay);
1383 }
1384 }
1385
1386 work_done = xenvif_tx_submit(queue);
1387
1388 return work_done;
1389}
1390
1391static void xenvif_idx_release(struct xenvif_queue *queue, u16 pending_idx,
1392 u8 status)
1393{
1394 struct pending_tx_info *pending_tx_info;
1395 pending_ring_idx_t index;
1396 unsigned long flags;
1397
1398 pending_tx_info = &queue->pending_tx_info[pending_idx];
1399
1400 spin_lock_irqsave(&queue->response_lock, flags);
1401
1402 make_tx_response(queue, &pending_tx_info->req,
1403 pending_tx_info->extra_count, status);
1404
1405 /* Release the pending index before pusing the Tx response so
1406 * its available before a new Tx request is pushed by the
1407 * frontend.
1408 */
1409 index = pending_index(queue->pending_prod++);
1410 queue->pending_ring[index] = pending_idx;
1411
1412 push_tx_responses(queue);
1413
1414 spin_unlock_irqrestore(&queue->response_lock, flags);
1415}
1416
1417
1418static void make_tx_response(struct xenvif_queue *queue,
1419 struct xen_netif_tx_request *txp,
1420 unsigned int extra_count,
1421 s8 st)
1422{
1423 RING_IDX i = queue->tx.rsp_prod_pvt;
1424 struct xen_netif_tx_response *resp;
1425
1426 resp = RING_GET_RESPONSE(&queue->tx, i);
1427 resp->id = txp->id;
1428 resp->status = st;
1429
1430 while (extra_count-- != 0)
1431 RING_GET_RESPONSE(&queue->tx, ++i)->status = XEN_NETIF_RSP_NULL;
1432
1433 queue->tx.rsp_prod_pvt = ++i;
1434}
1435
1436static void push_tx_responses(struct xenvif_queue *queue)
1437{
1438 int notify;
1439
1440 RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&queue->tx, notify);
1441 if (notify)
1442 notify_remote_via_irq(queue->tx_irq);
1443}
1444
1445static void xenvif_idx_unmap(struct xenvif_queue *queue, u16 pending_idx)
1446{
1447 int ret;
1448 struct gnttab_unmap_grant_ref tx_unmap_op;
1449
1450 gnttab_set_unmap_op(&tx_unmap_op,
1451 idx_to_kaddr(queue, pending_idx),
1452 GNTMAP_host_map,
1453 queue->grant_tx_handle[pending_idx]);
1454 xenvif_grant_handle_reset(queue, pending_idx);
1455
1456 ret = gnttab_unmap_refs(&tx_unmap_op, NULL,
1457 &queue->mmap_pages[pending_idx], 1);
1458 if (ret) {
1459 netdev_err(queue->vif->dev,
1460 "Unmap fail: ret: %d pending_idx: %d host_addr: %llx handle: 0x%x status: %d\n",
1461 ret,
1462 pending_idx,
1463 tx_unmap_op.host_addr,
1464 tx_unmap_op.handle,
1465 tx_unmap_op.status);
1466 BUG();
1467 }
1468}
1469
1470static inline int tx_work_todo(struct xenvif_queue *queue)
1471{
1472 if (likely(RING_HAS_UNCONSUMED_REQUESTS(&queue->tx)))
1473 return 1;
1474
1475 return 0;
1476}
1477
1478static inline bool tx_dealloc_work_todo(struct xenvif_queue *queue)
1479{
1480 return queue->dealloc_cons != queue->dealloc_prod;
1481}
1482
1483void xenvif_unmap_frontend_data_rings(struct xenvif_queue *queue)
1484{
1485 if (queue->tx.sring)
1486 xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(queue->vif),
1487 queue->tx.sring);
1488 if (queue->rx.sring)
1489 xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(queue->vif),
1490 queue->rx.sring);
1491}
1492
1493int xenvif_map_frontend_data_rings(struct xenvif_queue *queue,
1494 grant_ref_t tx_ring_ref,
1495 grant_ref_t rx_ring_ref)
1496{
1497 void *addr;
1498 struct xen_netif_tx_sring *txs;
1499 struct xen_netif_rx_sring *rxs;
1500 RING_IDX rsp_prod, req_prod;
1501 int err;
1502
1503 err = xenbus_map_ring_valloc(xenvif_to_xenbus_device(queue->vif),
1504 &tx_ring_ref, 1, &addr);
1505 if (err)
1506 goto err;
1507
1508 txs = (struct xen_netif_tx_sring *)addr;
1509 rsp_prod = READ_ONCE(txs->rsp_prod);
1510 req_prod = READ_ONCE(txs->req_prod);
1511
1512 BACK_RING_ATTACH(&queue->tx, txs, rsp_prod, XEN_PAGE_SIZE);
1513
1514 err = -EIO;
1515 if (req_prod - rsp_prod > RING_SIZE(&queue->tx))
1516 goto err;
1517
1518 err = xenbus_map_ring_valloc(xenvif_to_xenbus_device(queue->vif),
1519 &rx_ring_ref, 1, &addr);
1520 if (err)
1521 goto err;
1522
1523 rxs = (struct xen_netif_rx_sring *)addr;
1524 rsp_prod = READ_ONCE(rxs->rsp_prod);
1525 req_prod = READ_ONCE(rxs->req_prod);
1526
1527 BACK_RING_ATTACH(&queue->rx, rxs, rsp_prod, XEN_PAGE_SIZE);
1528
1529 err = -EIO;
1530 if (req_prod - rsp_prod > RING_SIZE(&queue->rx))
1531 goto err;
1532
1533 return 0;
1534
1535err:
1536 xenvif_unmap_frontend_data_rings(queue);
1537 return err;
1538}
1539
1540static bool xenvif_dealloc_kthread_should_stop(struct xenvif_queue *queue)
1541{
1542 /* Dealloc thread must remain running until all inflight
1543 * packets complete.
1544 */
1545 return kthread_should_stop() &&
1546 !atomic_read(&queue->inflight_packets);
1547}
1548
1549int xenvif_dealloc_kthread(void *data)
1550{
1551 struct xenvif_queue *queue = data;
1552
1553 for (;;) {
1554 wait_event_interruptible(queue->dealloc_wq,
1555 tx_dealloc_work_todo(queue) ||
1556 xenvif_dealloc_kthread_should_stop(queue));
1557 if (xenvif_dealloc_kthread_should_stop(queue))
1558 break;
1559
1560 xenvif_tx_dealloc_action(queue);
1561 cond_resched();
1562 }
1563
1564 /* Unmap anything remaining*/
1565 if (tx_dealloc_work_todo(queue))
1566 xenvif_tx_dealloc_action(queue);
1567
1568 return 0;
1569}
1570
1571static void make_ctrl_response(struct xenvif *vif,
1572 const struct xen_netif_ctrl_request *req,
1573 u32 status, u32 data)
1574{
1575 RING_IDX idx = vif->ctrl.rsp_prod_pvt;
1576 struct xen_netif_ctrl_response rsp = {
1577 .id = req->id,
1578 .type = req->type,
1579 .status = status,
1580 .data = data,
1581 };
1582
1583 *RING_GET_RESPONSE(&vif->ctrl, idx) = rsp;
1584 vif->ctrl.rsp_prod_pvt = ++idx;
1585}
1586
1587static void push_ctrl_response(struct xenvif *vif)
1588{
1589 int notify;
1590
1591 RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&vif->ctrl, notify);
1592 if (notify)
1593 notify_remote_via_irq(vif->ctrl_irq);
1594}
1595
1596static void process_ctrl_request(struct xenvif *vif,
1597 const struct xen_netif_ctrl_request *req)
1598{
1599 u32 status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED;
1600 u32 data = 0;
1601
1602 switch (req->type) {
1603 case XEN_NETIF_CTRL_TYPE_SET_HASH_ALGORITHM:
1604 status = xenvif_set_hash_alg(vif, req->data[0]);
1605 break;
1606
1607 case XEN_NETIF_CTRL_TYPE_GET_HASH_FLAGS:
1608 status = xenvif_get_hash_flags(vif, &data);
1609 break;
1610
1611 case XEN_NETIF_CTRL_TYPE_SET_HASH_FLAGS:
1612 status = xenvif_set_hash_flags(vif, req->data[0]);
1613 break;
1614
1615 case XEN_NETIF_CTRL_TYPE_SET_HASH_KEY:
1616 status = xenvif_set_hash_key(vif, req->data[0],
1617 req->data[1]);
1618 break;
1619
1620 case XEN_NETIF_CTRL_TYPE_GET_HASH_MAPPING_SIZE:
1621 status = XEN_NETIF_CTRL_STATUS_SUCCESS;
1622 data = XEN_NETBK_MAX_HASH_MAPPING_SIZE;
1623 break;
1624
1625 case XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING_SIZE:
1626 status = xenvif_set_hash_mapping_size(vif,
1627 req->data[0]);
1628 break;
1629
1630 case XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING:
1631 status = xenvif_set_hash_mapping(vif, req->data[0],
1632 req->data[1],
1633 req->data[2]);
1634 break;
1635
1636 default:
1637 break;
1638 }
1639
1640 make_ctrl_response(vif, req, status, data);
1641 push_ctrl_response(vif);
1642}
1643
1644static void xenvif_ctrl_action(struct xenvif *vif)
1645{
1646 for (;;) {
1647 RING_IDX req_prod, req_cons;
1648
1649 req_prod = vif->ctrl.sring->req_prod;
1650 req_cons = vif->ctrl.req_cons;
1651
1652 /* Make sure we can see requests before we process them. */
1653 rmb();
1654
1655 if (req_cons == req_prod)
1656 break;
1657
1658 while (req_cons != req_prod) {
1659 struct xen_netif_ctrl_request req;
1660
1661 RING_COPY_REQUEST(&vif->ctrl, req_cons, &req);
1662 req_cons++;
1663
1664 process_ctrl_request(vif, &req);
1665 }
1666
1667 vif->ctrl.req_cons = req_cons;
1668 vif->ctrl.sring->req_event = req_cons + 1;
1669 }
1670}
1671
1672static bool xenvif_ctrl_work_todo(struct xenvif *vif)
1673{
1674 if (likely(RING_HAS_UNCONSUMED_REQUESTS(&vif->ctrl)))
1675 return true;
1676
1677 return false;
1678}
1679
1680irqreturn_t xenvif_ctrl_irq_fn(int irq, void *data)
1681{
1682 struct xenvif *vif = data;
1683 unsigned int eoi_flag = XEN_EOI_FLAG_SPURIOUS;
1684
1685 while (xenvif_ctrl_work_todo(vif)) {
1686 xenvif_ctrl_action(vif);
1687 eoi_flag = 0;
1688 }
1689
1690 xen_irq_lateeoi(irq, eoi_flag);
1691
1692 return IRQ_HANDLED;
1693}
1694
1695static int __init netback_init(void)
1696{
1697 int rc = 0;
1698
1699 if (!xen_domain())
1700 return -ENODEV;
1701
1702 /* Allow as many queues as there are CPUs but max. 8 if user has not
1703 * specified a value.
1704 */
1705 if (xenvif_max_queues == 0)
1706 xenvif_max_queues = min_t(unsigned int, MAX_QUEUES_DEFAULT,
1707 num_online_cpus());
1708
1709 if (fatal_skb_slots < XEN_NETBK_LEGACY_SLOTS_MAX) {
1710 pr_info("fatal_skb_slots too small (%d), bump it to XEN_NETBK_LEGACY_SLOTS_MAX (%d)\n",
1711 fatal_skb_slots, XEN_NETBK_LEGACY_SLOTS_MAX);
1712 fatal_skb_slots = XEN_NETBK_LEGACY_SLOTS_MAX;
1713 }
1714
1715 rc = xenvif_xenbus_init();
1716 if (rc)
1717 goto failed_init;
1718
1719#ifdef CONFIG_DEBUG_FS
1720 xen_netback_dbg_root = debugfs_create_dir("xen-netback", NULL);
1721#endif /* CONFIG_DEBUG_FS */
1722
1723 return 0;
1724
1725failed_init:
1726 return rc;
1727}
1728
1729module_init(netback_init);
1730
1731static void __exit netback_fini(void)
1732{
1733#ifdef CONFIG_DEBUG_FS
1734 debugfs_remove_recursive(xen_netback_dbg_root);
1735#endif /* CONFIG_DEBUG_FS */
1736 xenvif_xenbus_fini();
1737}
1738module_exit(netback_fini);
1739
1740MODULE_LICENSE("Dual BSD/GPL");
1741MODULE_ALIAS("xen-backend:vif");
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
41#include <net/tcp.h>
42
43#include <xen/events.h>
44#include <xen/interface/memory.h>
45
46#include <asm/xen/hypercall.h>
47#include <asm/xen/page.h>
48
49struct pending_tx_info {
50 struct xen_netif_tx_request req;
51 struct xenvif *vif;
52};
53typedef unsigned int pending_ring_idx_t;
54
55struct netbk_rx_meta {
56 int id;
57 int size;
58 int gso_size;
59};
60
61#define MAX_PENDING_REQS 256
62
63/* Discriminate from any valid pending_idx value. */
64#define INVALID_PENDING_IDX 0xFFFF
65
66#define MAX_BUFFER_OFFSET PAGE_SIZE
67
68/* extra field used in struct page */
69union page_ext {
70 struct {
71#if BITS_PER_LONG < 64
72#define IDX_WIDTH 8
73#define GROUP_WIDTH (BITS_PER_LONG - IDX_WIDTH)
74 unsigned int group:GROUP_WIDTH;
75 unsigned int idx:IDX_WIDTH;
76#else
77 unsigned int group, idx;
78#endif
79 } e;
80 void *mapping;
81};
82
83struct xen_netbk {
84 wait_queue_head_t wq;
85 struct task_struct *task;
86
87 struct sk_buff_head rx_queue;
88 struct sk_buff_head tx_queue;
89
90 struct timer_list net_timer;
91
92 struct page *mmap_pages[MAX_PENDING_REQS];
93
94 pending_ring_idx_t pending_prod;
95 pending_ring_idx_t pending_cons;
96 struct list_head net_schedule_list;
97
98 /* Protect the net_schedule_list in netif. */
99 spinlock_t net_schedule_list_lock;
100
101 atomic_t netfront_count;
102
103 struct pending_tx_info pending_tx_info[MAX_PENDING_REQS];
104 struct gnttab_copy tx_copy_ops[MAX_PENDING_REQS];
105
106 u16 pending_ring[MAX_PENDING_REQS];
107
108 /*
109 * Given MAX_BUFFER_OFFSET of 4096 the worst case is that each
110 * head/fragment page uses 2 copy operations because it
111 * straddles two buffers in the frontend.
112 */
113 struct gnttab_copy grant_copy_op[2*XEN_NETIF_RX_RING_SIZE];
114 struct netbk_rx_meta meta[2*XEN_NETIF_RX_RING_SIZE];
115};
116
117static struct xen_netbk *xen_netbk;
118static int xen_netbk_group_nr;
119
120void xen_netbk_add_xenvif(struct xenvif *vif)
121{
122 int i;
123 int min_netfront_count;
124 int min_group = 0;
125 struct xen_netbk *netbk;
126
127 min_netfront_count = atomic_read(&xen_netbk[0].netfront_count);
128 for (i = 0; i < xen_netbk_group_nr; i++) {
129 int netfront_count = atomic_read(&xen_netbk[i].netfront_count);
130 if (netfront_count < min_netfront_count) {
131 min_group = i;
132 min_netfront_count = netfront_count;
133 }
134 }
135
136 netbk = &xen_netbk[min_group];
137
138 vif->netbk = netbk;
139 atomic_inc(&netbk->netfront_count);
140}
141
142void xen_netbk_remove_xenvif(struct xenvif *vif)
143{
144 struct xen_netbk *netbk = vif->netbk;
145 vif->netbk = NULL;
146 atomic_dec(&netbk->netfront_count);
147}
148
149static void xen_netbk_idx_release(struct xen_netbk *netbk, u16 pending_idx);
150static void make_tx_response(struct xenvif *vif,
151 struct xen_netif_tx_request *txp,
152 s8 st);
153static struct xen_netif_rx_response *make_rx_response(struct xenvif *vif,
154 u16 id,
155 s8 st,
156 u16 offset,
157 u16 size,
158 u16 flags);
159
160static inline unsigned long idx_to_pfn(struct xen_netbk *netbk,
161 u16 idx)
162{
163 return page_to_pfn(netbk->mmap_pages[idx]);
164}
165
166static inline unsigned long idx_to_kaddr(struct xen_netbk *netbk,
167 u16 idx)
168{
169 return (unsigned long)pfn_to_kaddr(idx_to_pfn(netbk, idx));
170}
171
172/* extra field used in struct page */
173static inline void set_page_ext(struct page *pg, struct xen_netbk *netbk,
174 unsigned int idx)
175{
176 unsigned int group = netbk - xen_netbk;
177 union page_ext ext = { .e = { .group = group + 1, .idx = idx } };
178
179 BUILD_BUG_ON(sizeof(ext) > sizeof(ext.mapping));
180 pg->mapping = ext.mapping;
181}
182
183static int get_page_ext(struct page *pg,
184 unsigned int *pgroup, unsigned int *pidx)
185{
186 union page_ext ext = { .mapping = pg->mapping };
187 struct xen_netbk *netbk;
188 unsigned int group, idx;
189
190 group = ext.e.group - 1;
191
192 if (group < 0 || group >= xen_netbk_group_nr)
193 return 0;
194
195 netbk = &xen_netbk[group];
196
197 idx = ext.e.idx;
198
199 if ((idx < 0) || (idx >= MAX_PENDING_REQS))
200 return 0;
201
202 if (netbk->mmap_pages[idx] != pg)
203 return 0;
204
205 *pgroup = group;
206 *pidx = idx;
207
208 return 1;
209}
210
211/*
212 * This is the amount of packet we copy rather than map, so that the
213 * guest can't fiddle with the contents of the headers while we do
214 * packet processing on them (netfilter, routing, etc).
215 */
216#define PKT_PROT_LEN (ETH_HLEN + \
217 VLAN_HLEN + \
218 sizeof(struct iphdr) + MAX_IPOPTLEN + \
219 sizeof(struct tcphdr) + MAX_TCP_OPTION_SPACE)
220
221static u16 frag_get_pending_idx(skb_frag_t *frag)
222{
223 return (u16)frag->page_offset;
224}
225
226static void frag_set_pending_idx(skb_frag_t *frag, u16 pending_idx)
227{
228 frag->page_offset = pending_idx;
229}
230
231static inline pending_ring_idx_t pending_index(unsigned i)
232{
233 return i & (MAX_PENDING_REQS-1);
234}
235
236static inline pending_ring_idx_t nr_pending_reqs(struct xen_netbk *netbk)
237{
238 return MAX_PENDING_REQS -
239 netbk->pending_prod + netbk->pending_cons;
240}
241
242static void xen_netbk_kick_thread(struct xen_netbk *netbk)
243{
244 wake_up(&netbk->wq);
245}
246
247static int max_required_rx_slots(struct xenvif *vif)
248{
249 int max = DIV_ROUND_UP(vif->dev->mtu, PAGE_SIZE);
250
251 if (vif->can_sg || vif->gso || vif->gso_prefix)
252 max += MAX_SKB_FRAGS + 1; /* extra_info + frags */
253
254 return max;
255}
256
257int xen_netbk_rx_ring_full(struct xenvif *vif)
258{
259 RING_IDX peek = vif->rx_req_cons_peek;
260 RING_IDX needed = max_required_rx_slots(vif);
261
262 return ((vif->rx.sring->req_prod - peek) < needed) ||
263 ((vif->rx.rsp_prod_pvt + XEN_NETIF_RX_RING_SIZE - peek) < needed);
264}
265
266int xen_netbk_must_stop_queue(struct xenvif *vif)
267{
268 if (!xen_netbk_rx_ring_full(vif))
269 return 0;
270
271 vif->rx.sring->req_event = vif->rx_req_cons_peek +
272 max_required_rx_slots(vif);
273 mb(); /* request notification /then/ check the queue */
274
275 return xen_netbk_rx_ring_full(vif);
276}
277
278/*
279 * Returns true if we should start a new receive buffer instead of
280 * adding 'size' bytes to a buffer which currently contains 'offset'
281 * bytes.
282 */
283static bool start_new_rx_buffer(int offset, unsigned long size, int head)
284{
285 /* simple case: we have completely filled the current buffer. */
286 if (offset == MAX_BUFFER_OFFSET)
287 return true;
288
289 /*
290 * complex case: start a fresh buffer if the current frag
291 * would overflow the current buffer but only if:
292 * (i) this frag would fit completely in the next buffer
293 * and (ii) there is already some data in the current buffer
294 * and (iii) this is not the head buffer.
295 *
296 * Where:
297 * - (i) stops us splitting a frag into two copies
298 * unless the frag is too large for a single buffer.
299 * - (ii) stops us from leaving a buffer pointlessly empty.
300 * - (iii) stops us leaving the first buffer
301 * empty. Strictly speaking this is already covered
302 * by (ii) but is explicitly checked because
303 * netfront relies on the first buffer being
304 * non-empty and can crash otherwise.
305 *
306 * This means we will effectively linearise small
307 * frags but do not needlessly split large buffers
308 * into multiple copies tend to give large frags their
309 * own buffers as before.
310 */
311 if ((offset + size > MAX_BUFFER_OFFSET) &&
312 (size <= MAX_BUFFER_OFFSET) && offset && !head)
313 return true;
314
315 return false;
316}
317
318/*
319 * Figure out how many ring slots we're going to need to send @skb to
320 * the guest. This function is essentially a dry run of
321 * netbk_gop_frag_copy.
322 */
323unsigned int xen_netbk_count_skb_slots(struct xenvif *vif, struct sk_buff *skb)
324{
325 unsigned int count;
326 int i, copy_off;
327
328 count = DIV_ROUND_UP(skb_headlen(skb), PAGE_SIZE);
329
330 copy_off = skb_headlen(skb) % PAGE_SIZE;
331
332 if (skb_shinfo(skb)->gso_size)
333 count++;
334
335 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
336 unsigned long size = skb_frag_size(&skb_shinfo(skb)->frags[i]);
337 unsigned long bytes;
338 while (size > 0) {
339 BUG_ON(copy_off > MAX_BUFFER_OFFSET);
340
341 if (start_new_rx_buffer(copy_off, size, 0)) {
342 count++;
343 copy_off = 0;
344 }
345
346 bytes = size;
347 if (copy_off + bytes > MAX_BUFFER_OFFSET)
348 bytes = MAX_BUFFER_OFFSET - copy_off;
349
350 copy_off += bytes;
351 size -= bytes;
352 }
353 }
354 return count;
355}
356
357struct netrx_pending_operations {
358 unsigned copy_prod, copy_cons;
359 unsigned meta_prod, meta_cons;
360 struct gnttab_copy *copy;
361 struct netbk_rx_meta *meta;
362 int copy_off;
363 grant_ref_t copy_gref;
364};
365
366static struct netbk_rx_meta *get_next_rx_buffer(struct xenvif *vif,
367 struct netrx_pending_operations *npo)
368{
369 struct netbk_rx_meta *meta;
370 struct xen_netif_rx_request *req;
371
372 req = RING_GET_REQUEST(&vif->rx, vif->rx.req_cons++);
373
374 meta = npo->meta + npo->meta_prod++;
375 meta->gso_size = 0;
376 meta->size = 0;
377 meta->id = req->id;
378
379 npo->copy_off = 0;
380 npo->copy_gref = req->gref;
381
382 return meta;
383}
384
385/*
386 * Set up the grant operations for this fragment. If it's a flipping
387 * interface, we also set up the unmap request from here.
388 */
389static void netbk_gop_frag_copy(struct xenvif *vif, struct sk_buff *skb,
390 struct netrx_pending_operations *npo,
391 struct page *page, unsigned long size,
392 unsigned long offset, int *head)
393{
394 struct gnttab_copy *copy_gop;
395 struct netbk_rx_meta *meta;
396 /*
397 * These variables are used iff get_page_ext returns true,
398 * in which case they are guaranteed to be initialized.
399 */
400 unsigned int uninitialized_var(group), uninitialized_var(idx);
401 int foreign = get_page_ext(page, &group, &idx);
402 unsigned long bytes;
403
404 /* Data must not cross a page boundary. */
405 BUG_ON(size + offset > PAGE_SIZE);
406
407 meta = npo->meta + npo->meta_prod - 1;
408
409 while (size > 0) {
410 BUG_ON(npo->copy_off > MAX_BUFFER_OFFSET);
411
412 if (start_new_rx_buffer(npo->copy_off, size, *head)) {
413 /*
414 * Netfront requires there to be some data in the head
415 * buffer.
416 */
417 BUG_ON(*head);
418
419 meta = get_next_rx_buffer(vif, npo);
420 }
421
422 bytes = size;
423 if (npo->copy_off + bytes > MAX_BUFFER_OFFSET)
424 bytes = MAX_BUFFER_OFFSET - npo->copy_off;
425
426 copy_gop = npo->copy + npo->copy_prod++;
427 copy_gop->flags = GNTCOPY_dest_gref;
428 if (foreign) {
429 struct xen_netbk *netbk = &xen_netbk[group];
430 struct pending_tx_info *src_pend;
431
432 src_pend = &netbk->pending_tx_info[idx];
433
434 copy_gop->source.domid = src_pend->vif->domid;
435 copy_gop->source.u.ref = src_pend->req.gref;
436 copy_gop->flags |= GNTCOPY_source_gref;
437 } else {
438 void *vaddr = page_address(page);
439 copy_gop->source.domid = DOMID_SELF;
440 copy_gop->source.u.gmfn = virt_to_mfn(vaddr);
441 }
442 copy_gop->source.offset = offset;
443 copy_gop->dest.domid = vif->domid;
444
445 copy_gop->dest.offset = npo->copy_off;
446 copy_gop->dest.u.ref = npo->copy_gref;
447 copy_gop->len = bytes;
448
449 npo->copy_off += bytes;
450 meta->size += bytes;
451
452 offset += bytes;
453 size -= bytes;
454
455 /* Leave a gap for the GSO descriptor. */
456 if (*head && skb_shinfo(skb)->gso_size && !vif->gso_prefix)
457 vif->rx.req_cons++;
458
459 *head = 0; /* There must be something in this buffer now. */
460
461 }
462}
463
464/*
465 * Prepare an SKB to be transmitted to the frontend.
466 *
467 * This function is responsible for allocating grant operations, meta
468 * structures, etc.
469 *
470 * It returns the number of meta structures consumed. The number of
471 * ring slots used is always equal to the number of meta slots used
472 * plus the number of GSO descriptors used. Currently, we use either
473 * zero GSO descriptors (for non-GSO packets) or one descriptor (for
474 * frontend-side LRO).
475 */
476static int netbk_gop_skb(struct sk_buff *skb,
477 struct netrx_pending_operations *npo)
478{
479 struct xenvif *vif = netdev_priv(skb->dev);
480 int nr_frags = skb_shinfo(skb)->nr_frags;
481 int i;
482 struct xen_netif_rx_request *req;
483 struct netbk_rx_meta *meta;
484 unsigned char *data;
485 int head = 1;
486 int old_meta_prod;
487
488 old_meta_prod = npo->meta_prod;
489
490 /* Set up a GSO prefix descriptor, if necessary */
491 if (skb_shinfo(skb)->gso_size && vif->gso_prefix) {
492 req = RING_GET_REQUEST(&vif->rx, vif->rx.req_cons++);
493 meta = npo->meta + npo->meta_prod++;
494 meta->gso_size = skb_shinfo(skb)->gso_size;
495 meta->size = 0;
496 meta->id = req->id;
497 }
498
499 req = RING_GET_REQUEST(&vif->rx, vif->rx.req_cons++);
500 meta = npo->meta + npo->meta_prod++;
501
502 if (!vif->gso_prefix)
503 meta->gso_size = skb_shinfo(skb)->gso_size;
504 else
505 meta->gso_size = 0;
506
507 meta->size = 0;
508 meta->id = req->id;
509 npo->copy_off = 0;
510 npo->copy_gref = req->gref;
511
512 data = skb->data;
513 while (data < skb_tail_pointer(skb)) {
514 unsigned int offset = offset_in_page(data);
515 unsigned int len = PAGE_SIZE - offset;
516
517 if (data + len > skb_tail_pointer(skb))
518 len = skb_tail_pointer(skb) - data;
519
520 netbk_gop_frag_copy(vif, skb, npo,
521 virt_to_page(data), len, offset, &head);
522 data += len;
523 }
524
525 for (i = 0; i < nr_frags; i++) {
526 netbk_gop_frag_copy(vif, skb, npo,
527 skb_frag_page(&skb_shinfo(skb)->frags[i]),
528 skb_frag_size(&skb_shinfo(skb)->frags[i]),
529 skb_shinfo(skb)->frags[i].page_offset,
530 &head);
531 }
532
533 return npo->meta_prod - old_meta_prod;
534}
535
536/*
537 * This is a twin to netbk_gop_skb. Assume that netbk_gop_skb was
538 * used to set up the operations on the top of
539 * netrx_pending_operations, which have since been done. Check that
540 * they didn't give any errors and advance over them.
541 */
542static int netbk_check_gop(struct xenvif *vif, int nr_meta_slots,
543 struct netrx_pending_operations *npo)
544{
545 struct gnttab_copy *copy_op;
546 int status = XEN_NETIF_RSP_OKAY;
547 int i;
548
549 for (i = 0; i < nr_meta_slots; i++) {
550 copy_op = npo->copy + npo->copy_cons++;
551 if (copy_op->status != GNTST_okay) {
552 netdev_dbg(vif->dev,
553 "Bad status %d from copy to DOM%d.\n",
554 copy_op->status, vif->domid);
555 status = XEN_NETIF_RSP_ERROR;
556 }
557 }
558
559 return status;
560}
561
562static void netbk_add_frag_responses(struct xenvif *vif, int status,
563 struct netbk_rx_meta *meta,
564 int nr_meta_slots)
565{
566 int i;
567 unsigned long offset;
568
569 /* No fragments used */
570 if (nr_meta_slots <= 1)
571 return;
572
573 nr_meta_slots--;
574
575 for (i = 0; i < nr_meta_slots; i++) {
576 int flags;
577 if (i == nr_meta_slots - 1)
578 flags = 0;
579 else
580 flags = XEN_NETRXF_more_data;
581
582 offset = 0;
583 make_rx_response(vif, meta[i].id, status, offset,
584 meta[i].size, flags);
585 }
586}
587
588struct skb_cb_overlay {
589 int meta_slots_used;
590};
591
592static void xen_netbk_rx_action(struct xen_netbk *netbk)
593{
594 struct xenvif *vif = NULL, *tmp;
595 s8 status;
596 u16 irq, flags;
597 struct xen_netif_rx_response *resp;
598 struct sk_buff_head rxq;
599 struct sk_buff *skb;
600 LIST_HEAD(notify);
601 int ret;
602 int nr_frags;
603 int count;
604 unsigned long offset;
605 struct skb_cb_overlay *sco;
606
607 struct netrx_pending_operations npo = {
608 .copy = netbk->grant_copy_op,
609 .meta = netbk->meta,
610 };
611
612 skb_queue_head_init(&rxq);
613
614 count = 0;
615
616 while ((skb = skb_dequeue(&netbk->rx_queue)) != NULL) {
617 vif = netdev_priv(skb->dev);
618 nr_frags = skb_shinfo(skb)->nr_frags;
619
620 sco = (struct skb_cb_overlay *)skb->cb;
621 sco->meta_slots_used = netbk_gop_skb(skb, &npo);
622
623 count += nr_frags + 1;
624
625 __skb_queue_tail(&rxq, skb);
626
627 /* Filled the batch queue? */
628 if (count + MAX_SKB_FRAGS >= XEN_NETIF_RX_RING_SIZE)
629 break;
630 }
631
632 BUG_ON(npo.meta_prod > ARRAY_SIZE(netbk->meta));
633
634 if (!npo.copy_prod)
635 return;
636
637 BUG_ON(npo.copy_prod > ARRAY_SIZE(netbk->grant_copy_op));
638 ret = HYPERVISOR_grant_table_op(GNTTABOP_copy, &netbk->grant_copy_op,
639 npo.copy_prod);
640 BUG_ON(ret != 0);
641
642 while ((skb = __skb_dequeue(&rxq)) != NULL) {
643 sco = (struct skb_cb_overlay *)skb->cb;
644
645 vif = netdev_priv(skb->dev);
646
647 if (netbk->meta[npo.meta_cons].gso_size && vif->gso_prefix) {
648 resp = RING_GET_RESPONSE(&vif->rx,
649 vif->rx.rsp_prod_pvt++);
650
651 resp->flags = XEN_NETRXF_gso_prefix | XEN_NETRXF_more_data;
652
653 resp->offset = netbk->meta[npo.meta_cons].gso_size;
654 resp->id = netbk->meta[npo.meta_cons].id;
655 resp->status = sco->meta_slots_used;
656
657 npo.meta_cons++;
658 sco->meta_slots_used--;
659 }
660
661
662 vif->dev->stats.tx_bytes += skb->len;
663 vif->dev->stats.tx_packets++;
664
665 status = netbk_check_gop(vif, sco->meta_slots_used, &npo);
666
667 if (sco->meta_slots_used == 1)
668 flags = 0;
669 else
670 flags = XEN_NETRXF_more_data;
671
672 if (skb->ip_summed == CHECKSUM_PARTIAL) /* local packet? */
673 flags |= XEN_NETRXF_csum_blank | XEN_NETRXF_data_validated;
674 else if (skb->ip_summed == CHECKSUM_UNNECESSARY)
675 /* remote but checksummed. */
676 flags |= XEN_NETRXF_data_validated;
677
678 offset = 0;
679 resp = make_rx_response(vif, netbk->meta[npo.meta_cons].id,
680 status, offset,
681 netbk->meta[npo.meta_cons].size,
682 flags);
683
684 if (netbk->meta[npo.meta_cons].gso_size && !vif->gso_prefix) {
685 struct xen_netif_extra_info *gso =
686 (struct xen_netif_extra_info *)
687 RING_GET_RESPONSE(&vif->rx,
688 vif->rx.rsp_prod_pvt++);
689
690 resp->flags |= XEN_NETRXF_extra_info;
691
692 gso->u.gso.size = netbk->meta[npo.meta_cons].gso_size;
693 gso->u.gso.type = XEN_NETIF_GSO_TYPE_TCPV4;
694 gso->u.gso.pad = 0;
695 gso->u.gso.features = 0;
696
697 gso->type = XEN_NETIF_EXTRA_TYPE_GSO;
698 gso->flags = 0;
699 }
700
701 netbk_add_frag_responses(vif, status,
702 netbk->meta + npo.meta_cons + 1,
703 sco->meta_slots_used);
704
705 RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&vif->rx, ret);
706 irq = vif->irq;
707 if (ret && list_empty(&vif->notify_list))
708 list_add_tail(&vif->notify_list, ¬ify);
709
710 xenvif_notify_tx_completion(vif);
711
712 xenvif_put(vif);
713 npo.meta_cons += sco->meta_slots_used;
714 dev_kfree_skb(skb);
715 }
716
717 list_for_each_entry_safe(vif, tmp, ¬ify, notify_list) {
718 notify_remote_via_irq(vif->irq);
719 list_del_init(&vif->notify_list);
720 }
721
722 /* More work to do? */
723 if (!skb_queue_empty(&netbk->rx_queue) &&
724 !timer_pending(&netbk->net_timer))
725 xen_netbk_kick_thread(netbk);
726}
727
728void xen_netbk_queue_tx_skb(struct xenvif *vif, struct sk_buff *skb)
729{
730 struct xen_netbk *netbk = vif->netbk;
731
732 skb_queue_tail(&netbk->rx_queue, skb);
733
734 xen_netbk_kick_thread(netbk);
735}
736
737static void xen_netbk_alarm(unsigned long data)
738{
739 struct xen_netbk *netbk = (struct xen_netbk *)data;
740 xen_netbk_kick_thread(netbk);
741}
742
743static int __on_net_schedule_list(struct xenvif *vif)
744{
745 return !list_empty(&vif->schedule_list);
746}
747
748/* Must be called with net_schedule_list_lock held */
749static void remove_from_net_schedule_list(struct xenvif *vif)
750{
751 if (likely(__on_net_schedule_list(vif))) {
752 list_del_init(&vif->schedule_list);
753 xenvif_put(vif);
754 }
755}
756
757static struct xenvif *poll_net_schedule_list(struct xen_netbk *netbk)
758{
759 struct xenvif *vif = NULL;
760
761 spin_lock_irq(&netbk->net_schedule_list_lock);
762 if (list_empty(&netbk->net_schedule_list))
763 goto out;
764
765 vif = list_first_entry(&netbk->net_schedule_list,
766 struct xenvif, schedule_list);
767 if (!vif)
768 goto out;
769
770 xenvif_get(vif);
771
772 remove_from_net_schedule_list(vif);
773out:
774 spin_unlock_irq(&netbk->net_schedule_list_lock);
775 return vif;
776}
777
778void xen_netbk_schedule_xenvif(struct xenvif *vif)
779{
780 unsigned long flags;
781 struct xen_netbk *netbk = vif->netbk;
782
783 if (__on_net_schedule_list(vif))
784 goto kick;
785
786 spin_lock_irqsave(&netbk->net_schedule_list_lock, flags);
787 if (!__on_net_schedule_list(vif) &&
788 likely(xenvif_schedulable(vif))) {
789 list_add_tail(&vif->schedule_list, &netbk->net_schedule_list);
790 xenvif_get(vif);
791 }
792 spin_unlock_irqrestore(&netbk->net_schedule_list_lock, flags);
793
794kick:
795 smp_mb();
796 if ((nr_pending_reqs(netbk) < (MAX_PENDING_REQS/2)) &&
797 !list_empty(&netbk->net_schedule_list))
798 xen_netbk_kick_thread(netbk);
799}
800
801void xen_netbk_deschedule_xenvif(struct xenvif *vif)
802{
803 struct xen_netbk *netbk = vif->netbk;
804 spin_lock_irq(&netbk->net_schedule_list_lock);
805 remove_from_net_schedule_list(vif);
806 spin_unlock_irq(&netbk->net_schedule_list_lock);
807}
808
809void xen_netbk_check_rx_xenvif(struct xenvif *vif)
810{
811 int more_to_do;
812
813 RING_FINAL_CHECK_FOR_REQUESTS(&vif->tx, more_to_do);
814
815 if (more_to_do)
816 xen_netbk_schedule_xenvif(vif);
817}
818
819static void tx_add_credit(struct xenvif *vif)
820{
821 unsigned long max_burst, max_credit;
822
823 /*
824 * Allow a burst big enough to transmit a jumbo packet of up to 128kB.
825 * Otherwise the interface can seize up due to insufficient credit.
826 */
827 max_burst = RING_GET_REQUEST(&vif->tx, vif->tx.req_cons)->size;
828 max_burst = min(max_burst, 131072UL);
829 max_burst = max(max_burst, vif->credit_bytes);
830
831 /* Take care that adding a new chunk of credit doesn't wrap to zero. */
832 max_credit = vif->remaining_credit + vif->credit_bytes;
833 if (max_credit < vif->remaining_credit)
834 max_credit = ULONG_MAX; /* wrapped: clamp to ULONG_MAX */
835
836 vif->remaining_credit = min(max_credit, max_burst);
837}
838
839static void tx_credit_callback(unsigned long data)
840{
841 struct xenvif *vif = (struct xenvif *)data;
842 tx_add_credit(vif);
843 xen_netbk_check_rx_xenvif(vif);
844}
845
846static void netbk_tx_err(struct xenvif *vif,
847 struct xen_netif_tx_request *txp, RING_IDX end)
848{
849 RING_IDX cons = vif->tx.req_cons;
850
851 do {
852 make_tx_response(vif, txp, XEN_NETIF_RSP_ERROR);
853 if (cons >= end)
854 break;
855 txp = RING_GET_REQUEST(&vif->tx, cons++);
856 } while (1);
857 vif->tx.req_cons = cons;
858 xen_netbk_check_rx_xenvif(vif);
859 xenvif_put(vif);
860}
861
862static int netbk_count_requests(struct xenvif *vif,
863 struct xen_netif_tx_request *first,
864 struct xen_netif_tx_request *txp,
865 int work_to_do)
866{
867 RING_IDX cons = vif->tx.req_cons;
868 int frags = 0;
869
870 if (!(first->flags & XEN_NETTXF_more_data))
871 return 0;
872
873 do {
874 if (frags >= work_to_do) {
875 netdev_dbg(vif->dev, "Need more frags\n");
876 return -frags;
877 }
878
879 if (unlikely(frags >= MAX_SKB_FRAGS)) {
880 netdev_dbg(vif->dev, "Too many frags\n");
881 return -frags;
882 }
883
884 memcpy(txp, RING_GET_REQUEST(&vif->tx, cons + frags),
885 sizeof(*txp));
886 if (txp->size > first->size) {
887 netdev_dbg(vif->dev, "Frags galore\n");
888 return -frags;
889 }
890
891 first->size -= txp->size;
892 frags++;
893
894 if (unlikely((txp->offset + txp->size) > PAGE_SIZE)) {
895 netdev_dbg(vif->dev, "txp->offset: %x, size: %u\n",
896 txp->offset, txp->size);
897 return -frags;
898 }
899 } while ((txp++)->flags & XEN_NETTXF_more_data);
900 return frags;
901}
902
903static struct page *xen_netbk_alloc_page(struct xen_netbk *netbk,
904 struct sk_buff *skb,
905 u16 pending_idx)
906{
907 struct page *page;
908 page = alloc_page(GFP_KERNEL|__GFP_COLD);
909 if (!page)
910 return NULL;
911 set_page_ext(page, netbk, pending_idx);
912 netbk->mmap_pages[pending_idx] = page;
913 return page;
914}
915
916static struct gnttab_copy *xen_netbk_get_requests(struct xen_netbk *netbk,
917 struct xenvif *vif,
918 struct sk_buff *skb,
919 struct xen_netif_tx_request *txp,
920 struct gnttab_copy *gop)
921{
922 struct skb_shared_info *shinfo = skb_shinfo(skb);
923 skb_frag_t *frags = shinfo->frags;
924 u16 pending_idx = *((u16 *)skb->data);
925 int i, start;
926
927 /* Skip first skb fragment if it is on same page as header fragment. */
928 start = (frag_get_pending_idx(&shinfo->frags[0]) == pending_idx);
929
930 for (i = start; i < shinfo->nr_frags; i++, txp++) {
931 struct page *page;
932 pending_ring_idx_t index;
933 struct pending_tx_info *pending_tx_info =
934 netbk->pending_tx_info;
935
936 index = pending_index(netbk->pending_cons++);
937 pending_idx = netbk->pending_ring[index];
938 page = xen_netbk_alloc_page(netbk, skb, pending_idx);
939 if (!page)
940 return NULL;
941
942 gop->source.u.ref = txp->gref;
943 gop->source.domid = vif->domid;
944 gop->source.offset = txp->offset;
945
946 gop->dest.u.gmfn = virt_to_mfn(page_address(page));
947 gop->dest.domid = DOMID_SELF;
948 gop->dest.offset = txp->offset;
949
950 gop->len = txp->size;
951 gop->flags = GNTCOPY_source_gref;
952
953 gop++;
954
955 memcpy(&pending_tx_info[pending_idx].req, txp, sizeof(*txp));
956 xenvif_get(vif);
957 pending_tx_info[pending_idx].vif = vif;
958 frag_set_pending_idx(&frags[i], pending_idx);
959 }
960
961 return gop;
962}
963
964static int xen_netbk_tx_check_gop(struct xen_netbk *netbk,
965 struct sk_buff *skb,
966 struct gnttab_copy **gopp)
967{
968 struct gnttab_copy *gop = *gopp;
969 u16 pending_idx = *((u16 *)skb->data);
970 struct pending_tx_info *pending_tx_info = netbk->pending_tx_info;
971 struct xenvif *vif = pending_tx_info[pending_idx].vif;
972 struct xen_netif_tx_request *txp;
973 struct skb_shared_info *shinfo = skb_shinfo(skb);
974 int nr_frags = shinfo->nr_frags;
975 int i, err, start;
976
977 /* Check status of header. */
978 err = gop->status;
979 if (unlikely(err)) {
980 pending_ring_idx_t index;
981 index = pending_index(netbk->pending_prod++);
982 txp = &pending_tx_info[pending_idx].req;
983 make_tx_response(vif, txp, XEN_NETIF_RSP_ERROR);
984 netbk->pending_ring[index] = pending_idx;
985 xenvif_put(vif);
986 }
987
988 /* Skip first skb fragment if it is on same page as header fragment. */
989 start = (frag_get_pending_idx(&shinfo->frags[0]) == pending_idx);
990
991 for (i = start; i < nr_frags; i++) {
992 int j, newerr;
993 pending_ring_idx_t index;
994
995 pending_idx = frag_get_pending_idx(&shinfo->frags[i]);
996
997 /* Check error status: if okay then remember grant handle. */
998 newerr = (++gop)->status;
999 if (likely(!newerr)) {
1000 /* Had a previous error? Invalidate this fragment. */
1001 if (unlikely(err))
1002 xen_netbk_idx_release(netbk, pending_idx);
1003 continue;
1004 }
1005
1006 /* Error on this fragment: respond to client with an error. */
1007 txp = &netbk->pending_tx_info[pending_idx].req;
1008 make_tx_response(vif, txp, XEN_NETIF_RSP_ERROR);
1009 index = pending_index(netbk->pending_prod++);
1010 netbk->pending_ring[index] = pending_idx;
1011 xenvif_put(vif);
1012
1013 /* Not the first error? Preceding frags already invalidated. */
1014 if (err)
1015 continue;
1016
1017 /* First error: invalidate header and preceding fragments. */
1018 pending_idx = *((u16 *)skb->data);
1019 xen_netbk_idx_release(netbk, pending_idx);
1020 for (j = start; j < i; j++) {
1021 pending_idx = frag_get_pending_idx(&shinfo->frags[j]);
1022 xen_netbk_idx_release(netbk, pending_idx);
1023 }
1024
1025 /* Remember the error: invalidate all subsequent fragments. */
1026 err = newerr;
1027 }
1028
1029 *gopp = gop + 1;
1030 return err;
1031}
1032
1033static void xen_netbk_fill_frags(struct xen_netbk *netbk, struct sk_buff *skb)
1034{
1035 struct skb_shared_info *shinfo = skb_shinfo(skb);
1036 int nr_frags = shinfo->nr_frags;
1037 int i;
1038
1039 for (i = 0; i < nr_frags; i++) {
1040 skb_frag_t *frag = shinfo->frags + i;
1041 struct xen_netif_tx_request *txp;
1042 struct page *page;
1043 u16 pending_idx;
1044
1045 pending_idx = frag_get_pending_idx(frag);
1046
1047 txp = &netbk->pending_tx_info[pending_idx].req;
1048 page = virt_to_page(idx_to_kaddr(netbk, pending_idx));
1049 __skb_fill_page_desc(skb, i, page, txp->offset, txp->size);
1050 skb->len += txp->size;
1051 skb->data_len += txp->size;
1052 skb->truesize += txp->size;
1053
1054 /* Take an extra reference to offset xen_netbk_idx_release */
1055 get_page(netbk->mmap_pages[pending_idx]);
1056 xen_netbk_idx_release(netbk, pending_idx);
1057 }
1058}
1059
1060static int xen_netbk_get_extras(struct xenvif *vif,
1061 struct xen_netif_extra_info *extras,
1062 int work_to_do)
1063{
1064 struct xen_netif_extra_info extra;
1065 RING_IDX cons = vif->tx.req_cons;
1066
1067 do {
1068 if (unlikely(work_to_do-- <= 0)) {
1069 netdev_dbg(vif->dev, "Missing extra info\n");
1070 return -EBADR;
1071 }
1072
1073 memcpy(&extra, RING_GET_REQUEST(&vif->tx, cons),
1074 sizeof(extra));
1075 if (unlikely(!extra.type ||
1076 extra.type >= XEN_NETIF_EXTRA_TYPE_MAX)) {
1077 vif->tx.req_cons = ++cons;
1078 netdev_dbg(vif->dev,
1079 "Invalid extra type: %d\n", extra.type);
1080 return -EINVAL;
1081 }
1082
1083 memcpy(&extras[extra.type - 1], &extra, sizeof(extra));
1084 vif->tx.req_cons = ++cons;
1085 } while (extra.flags & XEN_NETIF_EXTRA_FLAG_MORE);
1086
1087 return work_to_do;
1088}
1089
1090static int netbk_set_skb_gso(struct xenvif *vif,
1091 struct sk_buff *skb,
1092 struct xen_netif_extra_info *gso)
1093{
1094 if (!gso->u.gso.size) {
1095 netdev_dbg(vif->dev, "GSO size must not be zero.\n");
1096 return -EINVAL;
1097 }
1098
1099 /* Currently only TCPv4 S.O. is supported. */
1100 if (gso->u.gso.type != XEN_NETIF_GSO_TYPE_TCPV4) {
1101 netdev_dbg(vif->dev, "Bad GSO type %d.\n", gso->u.gso.type);
1102 return -EINVAL;
1103 }
1104
1105 skb_shinfo(skb)->gso_size = gso->u.gso.size;
1106 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1107
1108 /* Header must be checked, and gso_segs computed. */
1109 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
1110 skb_shinfo(skb)->gso_segs = 0;
1111
1112 return 0;
1113}
1114
1115static int checksum_setup(struct xenvif *vif, struct sk_buff *skb)
1116{
1117 struct iphdr *iph;
1118 unsigned char *th;
1119 int err = -EPROTO;
1120 int recalculate_partial_csum = 0;
1121
1122 /*
1123 * A GSO SKB must be CHECKSUM_PARTIAL. However some buggy
1124 * peers can fail to set NETRXF_csum_blank when sending a GSO
1125 * frame. In this case force the SKB to CHECKSUM_PARTIAL and
1126 * recalculate the partial checksum.
1127 */
1128 if (skb->ip_summed != CHECKSUM_PARTIAL && skb_is_gso(skb)) {
1129 vif->rx_gso_checksum_fixup++;
1130 skb->ip_summed = CHECKSUM_PARTIAL;
1131 recalculate_partial_csum = 1;
1132 }
1133
1134 /* A non-CHECKSUM_PARTIAL SKB does not require setup. */
1135 if (skb->ip_summed != CHECKSUM_PARTIAL)
1136 return 0;
1137
1138 if (skb->protocol != htons(ETH_P_IP))
1139 goto out;
1140
1141 iph = (void *)skb->data;
1142 th = skb->data + 4 * iph->ihl;
1143 if (th >= skb_tail_pointer(skb))
1144 goto out;
1145
1146 skb->csum_start = th - skb->head;
1147 switch (iph->protocol) {
1148 case IPPROTO_TCP:
1149 skb->csum_offset = offsetof(struct tcphdr, check);
1150
1151 if (recalculate_partial_csum) {
1152 struct tcphdr *tcph = (struct tcphdr *)th;
1153 tcph->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
1154 skb->len - iph->ihl*4,
1155 IPPROTO_TCP, 0);
1156 }
1157 break;
1158 case IPPROTO_UDP:
1159 skb->csum_offset = offsetof(struct udphdr, check);
1160
1161 if (recalculate_partial_csum) {
1162 struct udphdr *udph = (struct udphdr *)th;
1163 udph->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
1164 skb->len - iph->ihl*4,
1165 IPPROTO_UDP, 0);
1166 }
1167 break;
1168 default:
1169 if (net_ratelimit())
1170 netdev_err(vif->dev,
1171 "Attempting to checksum a non-TCP/UDP packet, dropping a protocol %d packet\n",
1172 iph->protocol);
1173 goto out;
1174 }
1175
1176 if ((th + skb->csum_offset + 2) > skb_tail_pointer(skb))
1177 goto out;
1178
1179 err = 0;
1180
1181out:
1182 return err;
1183}
1184
1185static bool tx_credit_exceeded(struct xenvif *vif, unsigned size)
1186{
1187 unsigned long now = jiffies;
1188 unsigned long next_credit =
1189 vif->credit_timeout.expires +
1190 msecs_to_jiffies(vif->credit_usec / 1000);
1191
1192 /* Timer could already be pending in rare cases. */
1193 if (timer_pending(&vif->credit_timeout))
1194 return true;
1195
1196 /* Passed the point where we can replenish credit? */
1197 if (time_after_eq(now, next_credit)) {
1198 vif->credit_timeout.expires = now;
1199 tx_add_credit(vif);
1200 }
1201
1202 /* Still too big to send right now? Set a callback. */
1203 if (size > vif->remaining_credit) {
1204 vif->credit_timeout.data =
1205 (unsigned long)vif;
1206 vif->credit_timeout.function =
1207 tx_credit_callback;
1208 mod_timer(&vif->credit_timeout,
1209 next_credit);
1210
1211 return true;
1212 }
1213
1214 return false;
1215}
1216
1217static unsigned xen_netbk_tx_build_gops(struct xen_netbk *netbk)
1218{
1219 struct gnttab_copy *gop = netbk->tx_copy_ops, *request_gop;
1220 struct sk_buff *skb;
1221 int ret;
1222
1223 while (((nr_pending_reqs(netbk) + MAX_SKB_FRAGS) < MAX_PENDING_REQS) &&
1224 !list_empty(&netbk->net_schedule_list)) {
1225 struct xenvif *vif;
1226 struct xen_netif_tx_request txreq;
1227 struct xen_netif_tx_request txfrags[MAX_SKB_FRAGS];
1228 struct page *page;
1229 struct xen_netif_extra_info extras[XEN_NETIF_EXTRA_TYPE_MAX-1];
1230 u16 pending_idx;
1231 RING_IDX idx;
1232 int work_to_do;
1233 unsigned int data_len;
1234 pending_ring_idx_t index;
1235
1236 /* Get a netif from the list with work to do. */
1237 vif = poll_net_schedule_list(netbk);
1238 if (!vif)
1239 continue;
1240
1241 RING_FINAL_CHECK_FOR_REQUESTS(&vif->tx, work_to_do);
1242 if (!work_to_do) {
1243 xenvif_put(vif);
1244 continue;
1245 }
1246
1247 idx = vif->tx.req_cons;
1248 rmb(); /* Ensure that we see the request before we copy it. */
1249 memcpy(&txreq, RING_GET_REQUEST(&vif->tx, idx), sizeof(txreq));
1250
1251 /* Credit-based scheduling. */
1252 if (txreq.size > vif->remaining_credit &&
1253 tx_credit_exceeded(vif, txreq.size)) {
1254 xenvif_put(vif);
1255 continue;
1256 }
1257
1258 vif->remaining_credit -= txreq.size;
1259
1260 work_to_do--;
1261 vif->tx.req_cons = ++idx;
1262
1263 memset(extras, 0, sizeof(extras));
1264 if (txreq.flags & XEN_NETTXF_extra_info) {
1265 work_to_do = xen_netbk_get_extras(vif, extras,
1266 work_to_do);
1267 idx = vif->tx.req_cons;
1268 if (unlikely(work_to_do < 0)) {
1269 netbk_tx_err(vif, &txreq, idx);
1270 continue;
1271 }
1272 }
1273
1274 ret = netbk_count_requests(vif, &txreq, txfrags, work_to_do);
1275 if (unlikely(ret < 0)) {
1276 netbk_tx_err(vif, &txreq, idx - ret);
1277 continue;
1278 }
1279 idx += ret;
1280
1281 if (unlikely(txreq.size < ETH_HLEN)) {
1282 netdev_dbg(vif->dev,
1283 "Bad packet size: %d\n", txreq.size);
1284 netbk_tx_err(vif, &txreq, idx);
1285 continue;
1286 }
1287
1288 /* No crossing a page as the payload mustn't fragment. */
1289 if (unlikely((txreq.offset + txreq.size) > PAGE_SIZE)) {
1290 netdev_dbg(vif->dev,
1291 "txreq.offset: %x, size: %u, end: %lu\n",
1292 txreq.offset, txreq.size,
1293 (txreq.offset&~PAGE_MASK) + txreq.size);
1294 netbk_tx_err(vif, &txreq, idx);
1295 continue;
1296 }
1297
1298 index = pending_index(netbk->pending_cons);
1299 pending_idx = netbk->pending_ring[index];
1300
1301 data_len = (txreq.size > PKT_PROT_LEN &&
1302 ret < MAX_SKB_FRAGS) ?
1303 PKT_PROT_LEN : txreq.size;
1304
1305 skb = alloc_skb(data_len + NET_SKB_PAD + NET_IP_ALIGN,
1306 GFP_ATOMIC | __GFP_NOWARN);
1307 if (unlikely(skb == NULL)) {
1308 netdev_dbg(vif->dev,
1309 "Can't allocate a skb in start_xmit.\n");
1310 netbk_tx_err(vif, &txreq, idx);
1311 break;
1312 }
1313
1314 /* Packets passed to netif_rx() must have some headroom. */
1315 skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN);
1316
1317 if (extras[XEN_NETIF_EXTRA_TYPE_GSO - 1].type) {
1318 struct xen_netif_extra_info *gso;
1319 gso = &extras[XEN_NETIF_EXTRA_TYPE_GSO - 1];
1320
1321 if (netbk_set_skb_gso(vif, skb, gso)) {
1322 kfree_skb(skb);
1323 netbk_tx_err(vif, &txreq, idx);
1324 continue;
1325 }
1326 }
1327
1328 /* XXX could copy straight to head */
1329 page = xen_netbk_alloc_page(netbk, skb, pending_idx);
1330 if (!page) {
1331 kfree_skb(skb);
1332 netbk_tx_err(vif, &txreq, idx);
1333 continue;
1334 }
1335
1336 gop->source.u.ref = txreq.gref;
1337 gop->source.domid = vif->domid;
1338 gop->source.offset = txreq.offset;
1339
1340 gop->dest.u.gmfn = virt_to_mfn(page_address(page));
1341 gop->dest.domid = DOMID_SELF;
1342 gop->dest.offset = txreq.offset;
1343
1344 gop->len = txreq.size;
1345 gop->flags = GNTCOPY_source_gref;
1346
1347 gop++;
1348
1349 memcpy(&netbk->pending_tx_info[pending_idx].req,
1350 &txreq, sizeof(txreq));
1351 netbk->pending_tx_info[pending_idx].vif = vif;
1352 *((u16 *)skb->data) = pending_idx;
1353
1354 __skb_put(skb, data_len);
1355
1356 skb_shinfo(skb)->nr_frags = ret;
1357 if (data_len < txreq.size) {
1358 skb_shinfo(skb)->nr_frags++;
1359 frag_set_pending_idx(&skb_shinfo(skb)->frags[0],
1360 pending_idx);
1361 } else {
1362 frag_set_pending_idx(&skb_shinfo(skb)->frags[0],
1363 INVALID_PENDING_IDX);
1364 }
1365
1366 __skb_queue_tail(&netbk->tx_queue, skb);
1367
1368 netbk->pending_cons++;
1369
1370 request_gop = xen_netbk_get_requests(netbk, vif,
1371 skb, txfrags, gop);
1372 if (request_gop == NULL) {
1373 kfree_skb(skb);
1374 netbk_tx_err(vif, &txreq, idx);
1375 continue;
1376 }
1377 gop = request_gop;
1378
1379 vif->tx.req_cons = idx;
1380 xen_netbk_check_rx_xenvif(vif);
1381
1382 if ((gop-netbk->tx_copy_ops) >= ARRAY_SIZE(netbk->tx_copy_ops))
1383 break;
1384 }
1385
1386 return gop - netbk->tx_copy_ops;
1387}
1388
1389static void xen_netbk_tx_submit(struct xen_netbk *netbk)
1390{
1391 struct gnttab_copy *gop = netbk->tx_copy_ops;
1392 struct sk_buff *skb;
1393
1394 while ((skb = __skb_dequeue(&netbk->tx_queue)) != NULL) {
1395 struct xen_netif_tx_request *txp;
1396 struct xenvif *vif;
1397 u16 pending_idx;
1398 unsigned data_len;
1399
1400 pending_idx = *((u16 *)skb->data);
1401 vif = netbk->pending_tx_info[pending_idx].vif;
1402 txp = &netbk->pending_tx_info[pending_idx].req;
1403
1404 /* Check the remap error code. */
1405 if (unlikely(xen_netbk_tx_check_gop(netbk, skb, &gop))) {
1406 netdev_dbg(vif->dev, "netback grant failed.\n");
1407 skb_shinfo(skb)->nr_frags = 0;
1408 kfree_skb(skb);
1409 continue;
1410 }
1411
1412 data_len = skb->len;
1413 memcpy(skb->data,
1414 (void *)(idx_to_kaddr(netbk, pending_idx)|txp->offset),
1415 data_len);
1416 if (data_len < txp->size) {
1417 /* Append the packet payload as a fragment. */
1418 txp->offset += data_len;
1419 txp->size -= data_len;
1420 } else {
1421 /* Schedule a response immediately. */
1422 xen_netbk_idx_release(netbk, pending_idx);
1423 }
1424
1425 if (txp->flags & XEN_NETTXF_csum_blank)
1426 skb->ip_summed = CHECKSUM_PARTIAL;
1427 else if (txp->flags & XEN_NETTXF_data_validated)
1428 skb->ip_summed = CHECKSUM_UNNECESSARY;
1429
1430 xen_netbk_fill_frags(netbk, skb);
1431
1432 /*
1433 * If the initial fragment was < PKT_PROT_LEN then
1434 * pull through some bytes from the other fragments to
1435 * increase the linear region to PKT_PROT_LEN bytes.
1436 */
1437 if (skb_headlen(skb) < PKT_PROT_LEN && skb_is_nonlinear(skb)) {
1438 int target = min_t(int, skb->len, PKT_PROT_LEN);
1439 __pskb_pull_tail(skb, target - skb_headlen(skb));
1440 }
1441
1442 skb->dev = vif->dev;
1443 skb->protocol = eth_type_trans(skb, skb->dev);
1444
1445 if (checksum_setup(vif, skb)) {
1446 netdev_dbg(vif->dev,
1447 "Can't setup checksum in net_tx_action\n");
1448 kfree_skb(skb);
1449 continue;
1450 }
1451
1452 vif->dev->stats.rx_bytes += skb->len;
1453 vif->dev->stats.rx_packets++;
1454
1455 xenvif_receive_skb(vif, skb);
1456 }
1457}
1458
1459/* Called after netfront has transmitted */
1460static void xen_netbk_tx_action(struct xen_netbk *netbk)
1461{
1462 unsigned nr_gops;
1463 int ret;
1464
1465 nr_gops = xen_netbk_tx_build_gops(netbk);
1466
1467 if (nr_gops == 0)
1468 return;
1469 ret = HYPERVISOR_grant_table_op(GNTTABOP_copy,
1470 netbk->tx_copy_ops, nr_gops);
1471 BUG_ON(ret);
1472
1473 xen_netbk_tx_submit(netbk);
1474
1475}
1476
1477static void xen_netbk_idx_release(struct xen_netbk *netbk, u16 pending_idx)
1478{
1479 struct xenvif *vif;
1480 struct pending_tx_info *pending_tx_info;
1481 pending_ring_idx_t index;
1482
1483 /* Already complete? */
1484 if (netbk->mmap_pages[pending_idx] == NULL)
1485 return;
1486
1487 pending_tx_info = &netbk->pending_tx_info[pending_idx];
1488
1489 vif = pending_tx_info->vif;
1490
1491 make_tx_response(vif, &pending_tx_info->req, XEN_NETIF_RSP_OKAY);
1492
1493 index = pending_index(netbk->pending_prod++);
1494 netbk->pending_ring[index] = pending_idx;
1495
1496 xenvif_put(vif);
1497
1498 netbk->mmap_pages[pending_idx]->mapping = 0;
1499 put_page(netbk->mmap_pages[pending_idx]);
1500 netbk->mmap_pages[pending_idx] = NULL;
1501}
1502
1503static void make_tx_response(struct xenvif *vif,
1504 struct xen_netif_tx_request *txp,
1505 s8 st)
1506{
1507 RING_IDX i = vif->tx.rsp_prod_pvt;
1508 struct xen_netif_tx_response *resp;
1509 int notify;
1510
1511 resp = RING_GET_RESPONSE(&vif->tx, i);
1512 resp->id = txp->id;
1513 resp->status = st;
1514
1515 if (txp->flags & XEN_NETTXF_extra_info)
1516 RING_GET_RESPONSE(&vif->tx, ++i)->status = XEN_NETIF_RSP_NULL;
1517
1518 vif->tx.rsp_prod_pvt = ++i;
1519 RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&vif->tx, notify);
1520 if (notify)
1521 notify_remote_via_irq(vif->irq);
1522}
1523
1524static struct xen_netif_rx_response *make_rx_response(struct xenvif *vif,
1525 u16 id,
1526 s8 st,
1527 u16 offset,
1528 u16 size,
1529 u16 flags)
1530{
1531 RING_IDX i = vif->rx.rsp_prod_pvt;
1532 struct xen_netif_rx_response *resp;
1533
1534 resp = RING_GET_RESPONSE(&vif->rx, i);
1535 resp->offset = offset;
1536 resp->flags = flags;
1537 resp->id = id;
1538 resp->status = (s16)size;
1539 if (st < 0)
1540 resp->status = (s16)st;
1541
1542 vif->rx.rsp_prod_pvt = ++i;
1543
1544 return resp;
1545}
1546
1547static inline int rx_work_todo(struct xen_netbk *netbk)
1548{
1549 return !skb_queue_empty(&netbk->rx_queue);
1550}
1551
1552static inline int tx_work_todo(struct xen_netbk *netbk)
1553{
1554
1555 if (((nr_pending_reqs(netbk) + MAX_SKB_FRAGS) < MAX_PENDING_REQS) &&
1556 !list_empty(&netbk->net_schedule_list))
1557 return 1;
1558
1559 return 0;
1560}
1561
1562static int xen_netbk_kthread(void *data)
1563{
1564 struct xen_netbk *netbk = data;
1565 while (!kthread_should_stop()) {
1566 wait_event_interruptible(netbk->wq,
1567 rx_work_todo(netbk) ||
1568 tx_work_todo(netbk) ||
1569 kthread_should_stop());
1570 cond_resched();
1571
1572 if (kthread_should_stop())
1573 break;
1574
1575 if (rx_work_todo(netbk))
1576 xen_netbk_rx_action(netbk);
1577
1578 if (tx_work_todo(netbk))
1579 xen_netbk_tx_action(netbk);
1580 }
1581
1582 return 0;
1583}
1584
1585void xen_netbk_unmap_frontend_rings(struct xenvif *vif)
1586{
1587 if (vif->tx.sring)
1588 xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(vif),
1589 vif->tx.sring);
1590 if (vif->rx.sring)
1591 xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(vif),
1592 vif->rx.sring);
1593}
1594
1595int xen_netbk_map_frontend_rings(struct xenvif *vif,
1596 grant_ref_t tx_ring_ref,
1597 grant_ref_t rx_ring_ref)
1598{
1599 void *addr;
1600 struct xen_netif_tx_sring *txs;
1601 struct xen_netif_rx_sring *rxs;
1602
1603 int err = -ENOMEM;
1604
1605 err = xenbus_map_ring_valloc(xenvif_to_xenbus_device(vif),
1606 tx_ring_ref, &addr);
1607 if (err)
1608 goto err;
1609
1610 txs = (struct xen_netif_tx_sring *)addr;
1611 BACK_RING_INIT(&vif->tx, txs, PAGE_SIZE);
1612
1613 err = xenbus_map_ring_valloc(xenvif_to_xenbus_device(vif),
1614 rx_ring_ref, &addr);
1615 if (err)
1616 goto err;
1617
1618 rxs = (struct xen_netif_rx_sring *)addr;
1619 BACK_RING_INIT(&vif->rx, rxs, PAGE_SIZE);
1620
1621 vif->rx_req_cons_peek = 0;
1622
1623 return 0;
1624
1625err:
1626 xen_netbk_unmap_frontend_rings(vif);
1627 return err;
1628}
1629
1630static int __init netback_init(void)
1631{
1632 int i;
1633 int rc = 0;
1634 int group;
1635
1636 if (!xen_domain())
1637 return -ENODEV;
1638
1639 xen_netbk_group_nr = num_online_cpus();
1640 xen_netbk = vzalloc(sizeof(struct xen_netbk) * xen_netbk_group_nr);
1641 if (!xen_netbk)
1642 return -ENOMEM;
1643
1644 for (group = 0; group < xen_netbk_group_nr; group++) {
1645 struct xen_netbk *netbk = &xen_netbk[group];
1646 skb_queue_head_init(&netbk->rx_queue);
1647 skb_queue_head_init(&netbk->tx_queue);
1648
1649 init_timer(&netbk->net_timer);
1650 netbk->net_timer.data = (unsigned long)netbk;
1651 netbk->net_timer.function = xen_netbk_alarm;
1652
1653 netbk->pending_cons = 0;
1654 netbk->pending_prod = MAX_PENDING_REQS;
1655 for (i = 0; i < MAX_PENDING_REQS; i++)
1656 netbk->pending_ring[i] = i;
1657
1658 init_waitqueue_head(&netbk->wq);
1659 netbk->task = kthread_create(xen_netbk_kthread,
1660 (void *)netbk,
1661 "netback/%u", group);
1662
1663 if (IS_ERR(netbk->task)) {
1664 printk(KERN_ALERT "kthread_create() fails at netback\n");
1665 del_timer(&netbk->net_timer);
1666 rc = PTR_ERR(netbk->task);
1667 goto failed_init;
1668 }
1669
1670 kthread_bind(netbk->task, group);
1671
1672 INIT_LIST_HEAD(&netbk->net_schedule_list);
1673
1674 spin_lock_init(&netbk->net_schedule_list_lock);
1675
1676 atomic_set(&netbk->netfront_count, 0);
1677
1678 wake_up_process(netbk->task);
1679 }
1680
1681 rc = xenvif_xenbus_init();
1682 if (rc)
1683 goto failed_init;
1684
1685 return 0;
1686
1687failed_init:
1688 while (--group >= 0) {
1689 struct xen_netbk *netbk = &xen_netbk[group];
1690 for (i = 0; i < MAX_PENDING_REQS; i++) {
1691 if (netbk->mmap_pages[i])
1692 __free_page(netbk->mmap_pages[i]);
1693 }
1694 del_timer(&netbk->net_timer);
1695 kthread_stop(netbk->task);
1696 }
1697 vfree(xen_netbk);
1698 return rc;
1699
1700}
1701
1702module_init(netback_init);
1703
1704MODULE_LICENSE("Dual BSD/GPL");
1705MODULE_ALIAS("xen-backend:vif");