Commit | Line | Data |
---|---|---|
f942dc25 IC |
1 | /* |
2 | * This program is free software; you can redistribute it and/or | |
3 | * modify it under the terms of the GNU General Public License version 2 | |
4 | * as published by the Free Software Foundation; or, when distributed | |
5 | * separately from the Linux kernel or incorporated into other | |
6 | * software packages, subject to the following license: | |
7 | * | |
8 | * Permission is hereby granted, free of charge, to any person obtaining a copy | |
9 | * of this source file (the "Software"), to deal in the Software without | |
10 | * restriction, including without limitation the rights to use, copy, modify, | |
11 | * merge, publish, distribute, sublicense, and/or sell copies of the Software, | |
12 | * and to permit persons to whom the Software is furnished to do so, subject to | |
13 | * the following conditions: | |
14 | * | |
15 | * The above copyright notice and this permission notice shall be included in | |
16 | * all copies or substantial portions of the Software. | |
17 | * | |
18 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
19 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
20 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE | |
21 | * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
22 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING | |
23 | * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS | |
24 | * IN THE SOFTWARE. | |
25 | */ | |
26 | ||
27 | #ifndef __XEN_NETBACK__COMMON_H__ | |
28 | #define __XEN_NETBACK__COMMON_H__ | |
29 | ||
30 | #define pr_fmt(fmt) KBUILD_MODNAME ":%s: " fmt, __func__ | |
31 | ||
32 | #include <linux/module.h> | |
33 | #include <linux/interrupt.h> | |
34 | #include <linux/slab.h> | |
35 | #include <linux/ip.h> | |
36 | #include <linux/in.h> | |
37 | #include <linux/io.h> | |
38 | #include <linux/netdevice.h> | |
39 | #include <linux/etherdevice.h> | |
40 | #include <linux/wait.h> | |
41 | #include <linux/sched.h> | |
42 | ||
43 | #include <xen/interface/io/netif.h> | |
44 | #include <xen/interface/grant_table.h> | |
45 | #include <xen/grant_table.h> | |
46 | #include <xen/xenbus.h> | |
d0089e8a | 47 | #include <xen/page.h> |
f51de243 | 48 | #include <linux/debugfs.h> |
f942dc25 | 49 | |
b3f980bd WL |
50 | typedef unsigned int pending_ring_idx_t; |
51 | #define INVALID_PENDING_RING_IDX (~0U) | |
52 | ||
b3f980bd | 53 | struct pending_tx_info { |
62bad319 | 54 | struct xen_netif_tx_request req; /* tx request */ |
562abd39 | 55 | unsigned int extra_count; |
f53c3fe8 ZK |
56 | /* Callback data for released SKBs. The callback is always |
57 | * xenvif_zerocopy_callback, desc contains the pending_idx, which is | |
58 | * also an index in pending_tx_info array. It is initialized in | |
59 | * xenvif_alloc and it never changes. | |
60 | * skb_shinfo(skb)->destructor_arg points to the first mapped slot's | |
61 | * callback_struct in this array of struct pending_tx_info's, then ctx | |
62 | * to the next, or NULL if there is no more slot for this skb. | |
63 | * ubuf_to_vif is a helper which finds the struct xenvif from a pointer | |
64 | * to this field. | |
65 | */ | |
66 | struct ubuf_info callback_struct; | |
b3f980bd WL |
67 | }; |
68 | ||
d0089e8a JG |
69 | #define XEN_NETIF_TX_RING_SIZE __CONST_RING_SIZE(xen_netif_tx, XEN_PAGE_SIZE) |
70 | #define XEN_NETIF_RX_RING_SIZE __CONST_RING_SIZE(xen_netif_rx, XEN_PAGE_SIZE) | |
b3f980bd WL |
71 | |
72 | struct xenvif_rx_meta { | |
73 | int id; | |
74 | int size; | |
82cada22 | 75 | int gso_type; |
b3f980bd WL |
76 | int gso_size; |
77 | }; | |
78 | ||
82cada22 PD |
79 | #define GSO_BIT(type) \ |
80 | (1 << XEN_NETIF_GSO_TYPE_ ## type) | |
81 | ||
b3f980bd WL |
82 | /* Discriminate from any valid pending_idx value. */ |
83 | #define INVALID_PENDING_IDX 0xFFFF | |
84 | ||
d0089e8a | 85 | #define MAX_BUFFER_OFFSET XEN_PAGE_SIZE |
b3f980bd | 86 | |
869b9b19 | 87 | #define MAX_PENDING_REQS XEN_NETIF_TX_RING_SIZE |
f942dc25 | 88 | |
d0089e8a JG |
89 | /* The maximum number of frags is derived from the size of a grant (same |
90 | * as a Xen page size for now). | |
91 | */ | |
92 | #define MAX_XEN_SKB_FRAGS (65536 / XEN_PAGE_SIZE + 1) | |
93 | ||
ac3d5ac2 PD |
94 | /* It's possible for an skb to have a maximal number of frags |
95 | * but still be less than MAX_BUFFER_OFFSET in size. Thus the | |
d0089e8a | 96 | * worst-case number of copy operations is MAX_XEN_SKB_FRAGS per |
ac3d5ac2 PD |
97 | * ring slot. |
98 | */ | |
d0089e8a | 99 | #define MAX_GRANT_COPY_OPS (MAX_XEN_SKB_FRAGS * XEN_NETIF_RX_RING_SIZE) |
ac3d5ac2 | 100 | |
121fa4b7 ZK |
101 | #define NETBACK_INVALID_HANDLE -1 |
102 | ||
103 | /* To avoid confusion, we define XEN_NETBK_LEGACY_SLOTS_MAX indicating | |
104 | * the maximum slots a valid packet can use. Now this value is defined | |
105 | * to be XEN_NETIF_NR_SLOTS_MIN, which is supposed to be supported by | |
106 | * all backend. | |
107 | */ | |
108 | #define XEN_NETBK_LEGACY_SLOTS_MAX XEN_NETIF_NR_SLOTS_MIN | |
109 | ||
e9ce7cb6 WL |
110 | /* Queue name is interface name with "-qNNN" appended */ |
111 | #define QUEUE_NAME_SIZE (IFNAMSIZ + 5) | |
f942dc25 | 112 | |
e9ce7cb6 WL |
113 | /* IRQ name is queue name with "-tx" or "-rx" appended */ |
114 | #define IRQ_NAME_SIZE (QUEUE_NAME_SIZE + 3) | |
115 | ||
116 | struct xenvif; | |
117 | ||
118 | struct xenvif_stats { | |
119 | /* Stats fields to be updated per-queue. | |
120 | * A subset of struct net_device_stats that contains only the | |
121 | * fields that are updated in netback.c for each queue. | |
e9d8b2c2 | 122 | */ |
e9ce7cb6 WL |
123 | unsigned int rx_bytes; |
124 | unsigned int rx_packets; | |
125 | unsigned int tx_bytes; | |
126 | unsigned int tx_packets; | |
127 | ||
128 | /* Additional stats used by xenvif */ | |
129 | unsigned long rx_gso_checksum_fixup; | |
130 | unsigned long tx_zerocopy_sent; | |
131 | unsigned long tx_zerocopy_success; | |
132 | unsigned long tx_zerocopy_fail; | |
133 | unsigned long tx_frag_overflow; | |
134 | }; | |
135 | ||
136 | struct xenvif_queue { /* Per-queue data for xenvif */ | |
137 | unsigned int id; /* Queue ID, 0-based */ | |
138 | char name[QUEUE_NAME_SIZE]; /* DEVNAME-qN */ | |
139 | struct xenvif *vif; /* Parent VIF */ | |
e9d8b2c2 | 140 | |
b3f980bd WL |
141 | /* Use NAPI for guest TX */ |
142 | struct napi_struct napi; | |
143 | /* When feature-split-event-channels = 0, tx_irq = rx_irq. */ | |
144 | unsigned int tx_irq; | |
145 | /* Only used when feature-split-event-channels = 1 */ | |
e9ce7cb6 | 146 | char tx_irq_name[IRQ_NAME_SIZE]; /* DEVNAME-qN-tx */ |
b3f980bd WL |
147 | struct xen_netif_tx_back_ring tx; |
148 | struct sk_buff_head tx_queue; | |
149 | struct page *mmap_pages[MAX_PENDING_REQS]; | |
150 | pending_ring_idx_t pending_prod; | |
151 | pending_ring_idx_t pending_cons; | |
152 | u16 pending_ring[MAX_PENDING_REQS]; | |
153 | struct pending_tx_info pending_tx_info[MAX_PENDING_REQS]; | |
f53c3fe8 | 154 | grant_handle_t grant_tx_handle[MAX_PENDING_REQS]; |
b3f980bd | 155 | |
bdab8275 | 156 | struct gnttab_copy tx_copy_ops[MAX_PENDING_REQS]; |
f53c3fe8 ZK |
157 | struct gnttab_map_grant_ref tx_map_ops[MAX_PENDING_REQS]; |
158 | struct gnttab_unmap_grant_ref tx_unmap_ops[MAX_PENDING_REQS]; | |
159 | /* passed to gnttab_[un]map_refs with pages under (un)mapping */ | |
160 | struct page *pages_to_map[MAX_PENDING_REQS]; | |
161 | struct page *pages_to_unmap[MAX_PENDING_REQS]; | |
162 | ||
163 | /* This prevents zerocopy callbacks to race over dealloc_ring */ | |
164 | spinlock_t callback_lock; | |
165 | /* This prevents dealloc thread and NAPI instance to race over response | |
166 | * creation and pending_ring in xenvif_idx_release. In xenvif_tx_err | |
167 | * it only protect response creation | |
168 | */ | |
169 | spinlock_t response_lock; | |
170 | pending_ring_idx_t dealloc_prod; | |
171 | pending_ring_idx_t dealloc_cons; | |
172 | u16 dealloc_ring[MAX_PENDING_REQS]; | |
173 | struct task_struct *dealloc_task; | |
174 | wait_queue_head_t dealloc_wq; | |
a64bd934 | 175 | atomic_t inflight_packets; |
f942dc25 | 176 | |
b3f980bd WL |
177 | /* Use kthread for guest RX */ |
178 | struct task_struct *task; | |
179 | wait_queue_head_t wq; | |
e1f00a69 | 180 | /* When feature-split-event-channels = 0, tx_irq = rx_irq. */ |
e1f00a69 WL |
181 | unsigned int rx_irq; |
182 | /* Only used when feature-split-event-channels = 1 */ | |
e9ce7cb6 | 183 | char rx_irq_name[IRQ_NAME_SIZE]; /* DEVNAME-qN-rx */ |
b3f980bd WL |
184 | struct xen_netif_rx_back_ring rx; |
185 | struct sk_buff_head rx_queue; | |
09350788 | 186 | |
f48da8b1 DV |
187 | unsigned int rx_queue_max; |
188 | unsigned int rx_queue_len; | |
ecf08d2d DV |
189 | unsigned long last_rx_time; |
190 | bool stalled; | |
b3f980bd | 191 | |
a55d9766 | 192 | struct gnttab_copy grant_copy_op[MAX_GRANT_COPY_OPS]; |
f942dc25 | 193 | |
ac3d5ac2 PD |
194 | /* We create one meta structure per ring request we consume, so |
195 | * the maximum number is the same as the ring size. | |
196 | */ | |
197 | struct xenvif_rx_meta meta[XEN_NETIF_RX_RING_SIZE]; | |
b3f980bd | 198 | |
e9ce7cb6 WL |
199 | /* Transmit shaping: allow 'credit_bytes' every 'credit_usec'. */ |
200 | unsigned long credit_bytes; | |
201 | unsigned long credit_usec; | |
202 | unsigned long remaining_credit; | |
203 | struct timer_list credit_timeout; | |
204 | u64 credit_window_start; | |
205 | ||
206 | /* Statistics */ | |
207 | struct xenvif_stats stats; | |
208 | }; | |
209 | ||
3d1af1df ZK |
210 | enum state_bit_shift { |
211 | /* This bit marks that the vif is connected */ | |
f34a4cf9 | 212 | VIF_STATUS_CONNECTED, |
3d1af1df ZK |
213 | }; |
214 | ||
210c34dc PD |
215 | struct xenvif_mcast_addr { |
216 | struct list_head entry; | |
217 | struct rcu_head rcu; | |
218 | u8 addr[6]; | |
219 | }; | |
220 | ||
221 | #define XEN_NETBK_MCAST_MAX 64 | |
222 | ||
40d8abde PD |
223 | #define XEN_NETBK_MAX_HASH_KEY_SIZE 40 |
224 | #define XEN_NETBK_MAX_HASH_MAPPING_SIZE 128 | |
225 | #define XEN_NETBK_HASH_TAG_SIZE 40 | |
226 | ||
227 | struct xenvif_hash_cache_entry { | |
228 | struct list_head link; | |
229 | struct rcu_head rcu; | |
230 | u8 tag[XEN_NETBK_HASH_TAG_SIZE]; | |
231 | unsigned int len; | |
232 | u32 val; | |
233 | int seq; | |
234 | }; | |
235 | ||
236 | struct xenvif_hash_cache { | |
237 | spinlock_t lock; | |
238 | struct list_head list; | |
239 | unsigned int count; | |
240 | atomic_t seq; | |
241 | }; | |
242 | ||
243 | struct xenvif_hash { | |
244 | unsigned int alg; | |
245 | u32 flags; | |
246 | u8 key[XEN_NETBK_MAX_HASH_KEY_SIZE]; | |
247 | u32 mapping[XEN_NETBK_MAX_HASH_MAPPING_SIZE]; | |
248 | unsigned int size; | |
249 | struct xenvif_hash_cache cache; | |
250 | }; | |
251 | ||
e9ce7cb6 WL |
252 | struct xenvif { |
253 | /* Unique identifier for this interface. */ | |
254 | domid_t domid; | |
255 | unsigned int handle; | |
256 | ||
b3f980bd | 257 | u8 fe_dev_addr[6]; |
210c34dc PD |
258 | struct list_head fe_mcast_addr; |
259 | unsigned int fe_mcast_count; | |
f942dc25 | 260 | |
f942dc25 | 261 | /* Frontend feature information. */ |
82cada22 PD |
262 | int gso_mask; |
263 | int gso_prefix_mask; | |
264 | ||
f942dc25 | 265 | u8 can_sg:1; |
146c8a77 PD |
266 | u8 ip_csum:1; |
267 | u8 ipv6_csum:1; | |
210c34dc | 268 | u8 multicast_control:1; |
f942dc25 | 269 | |
e9ce7cb6 WL |
270 | /* Is this interface disabled? True when backend discovers |
271 | * frontend is rogue. | |
272 | */ | |
273 | bool disabled; | |
3d1af1df | 274 | unsigned long status; |
26c0e102 DV |
275 | unsigned long drain_timeout; |
276 | unsigned long stall_timeout; | |
f942dc25 | 277 | |
e9ce7cb6 WL |
278 | /* Queues */ |
279 | struct xenvif_queue *queues; | |
f7b50c4e | 280 | unsigned int num_queues; /* active queues, resource allocated */ |
ecf08d2d DV |
281 | unsigned int stalled_queues; |
282 | ||
40d8abde PD |
283 | struct xenvif_hash hash; |
284 | ||
edafc132 | 285 | struct xenbus_watch credit_watch; |
22fae97d | 286 | struct xenbus_watch mcast_ctrl_watch; |
edafc132 | 287 | |
ecf08d2d | 288 | spinlock_t lock; |
f942dc25 | 289 | |
f51de243 ZK |
290 | #ifdef CONFIG_DEBUG_FS |
291 | struct dentry *xenvif_dbg_root; | |
292 | #endif | |
293 | ||
4e15ee2c PD |
294 | struct xen_netif_ctrl_back_ring ctrl; |
295 | struct task_struct *ctrl_task; | |
296 | wait_queue_head_t ctrl_wq; | |
297 | unsigned int ctrl_irq; | |
298 | ||
f942dc25 | 299 | /* Miscellaneous private stuff. */ |
f942dc25 | 300 | struct net_device *dev; |
f942dc25 IC |
301 | }; |
302 | ||
f48da8b1 DV |
303 | struct xenvif_rx_cb { |
304 | unsigned long expires; | |
305 | int meta_slots_used; | |
f48da8b1 DV |
306 | }; |
307 | ||
308 | #define XENVIF_RX_CB(skb) ((struct xenvif_rx_cb *)(skb)->cb) | |
309 | ||
c9d63699 DV |
310 | static inline struct xenbus_device *xenvif_to_xenbus_device(struct xenvif *vif) |
311 | { | |
312 | return to_xenbus_device(vif->dev->dev.parent); | |
313 | } | |
314 | ||
edafc132 PI |
315 | void xenvif_tx_credit_callback(unsigned long data); |
316 | ||
f942dc25 IC |
317 | struct xenvif *xenvif_alloc(struct device *parent, |
318 | domid_t domid, | |
319 | unsigned int handle); | |
320 | ||
e9ce7cb6 | 321 | int xenvif_init_queue(struct xenvif_queue *queue); |
8d3d53b3 | 322 | void xenvif_deinit_queue(struct xenvif_queue *queue); |
e9ce7cb6 | 323 | |
4e15ee2c PD |
324 | int xenvif_connect_data(struct xenvif_queue *queue, |
325 | unsigned long tx_ring_ref, | |
326 | unsigned long rx_ring_ref, | |
327 | unsigned int tx_evtchn, | |
328 | unsigned int rx_evtchn); | |
329 | void xenvif_disconnect_data(struct xenvif *vif); | |
330 | int xenvif_connect_ctrl(struct xenvif *vif, grant_ref_t ring_ref, | |
331 | unsigned int evtchn); | |
332 | void xenvif_disconnect_ctrl(struct xenvif *vif); | |
279f438e | 333 | void xenvif_free(struct xenvif *vif); |
f942dc25 | 334 | |
f942dc25 | 335 | int xenvif_xenbus_init(void); |
b103f358 | 336 | void xenvif_xenbus_fini(void); |
f942dc25 IC |
337 | |
338 | int xenvif_schedulable(struct xenvif *vif); | |
339 | ||
e9ce7cb6 WL |
340 | int xenvif_queue_stopped(struct xenvif_queue *queue); |
341 | void xenvif_wake_queue(struct xenvif_queue *queue); | |
f942dc25 IC |
342 | |
343 | /* (Un)Map communication rings. */ | |
4e15ee2c PD |
344 | void xenvif_unmap_frontend_data_rings(struct xenvif_queue *queue); |
345 | int xenvif_map_frontend_data_rings(struct xenvif_queue *queue, | |
346 | grant_ref_t tx_ring_ref, | |
347 | grant_ref_t rx_ring_ref); | |
f942dc25 | 348 | |
f942dc25 | 349 | /* Check for SKBs from frontend and schedule backend processing */ |
e9ce7cb6 | 350 | void xenvif_napi_schedule_or_enable_events(struct xenvif_queue *queue); |
f942dc25 | 351 | |
48856286 IC |
352 | /* Prevent the device from generating any further traffic. */ |
353 | void xenvif_carrier_off(struct xenvif *vif); | |
354 | ||
e9ce7cb6 | 355 | int xenvif_tx_action(struct xenvif_queue *queue, int budget); |
b3f980bd | 356 | |
121fa4b7 | 357 | int xenvif_kthread_guest_rx(void *data); |
e9ce7cb6 | 358 | void xenvif_kick_thread(struct xenvif_queue *queue); |
ca2f09f2 | 359 | |
f53c3fe8 ZK |
360 | int xenvif_dealloc_kthread(void *data); |
361 | ||
4e15ee2c PD |
362 | int xenvif_ctrl_kthread(void *data); |
363 | ||
f48da8b1 DV |
364 | void xenvif_rx_queue_tail(struct xenvif_queue *queue, struct sk_buff *skb); |
365 | ||
e9ce7cb6 | 366 | void xenvif_carrier_on(struct xenvif *vif); |
b3f980bd | 367 | |
f53c3fe8 ZK |
368 | /* Callback from stack when TX packet can be released */ |
369 | void xenvif_zerocopy_callback(struct ubuf_info *ubuf, bool zerocopy_success); | |
370 | ||
371 | /* Unmap a pending page and release it back to the guest */ | |
e9ce7cb6 | 372 | void xenvif_idx_unmap(struct xenvif_queue *queue, u16 pending_idx); |
f53c3fe8 | 373 | |
e9ce7cb6 | 374 | static inline pending_ring_idx_t nr_pending_reqs(struct xenvif_queue *queue) |
121fa4b7 ZK |
375 | { |
376 | return MAX_PENDING_REQS - | |
e9ce7cb6 | 377 | queue->pending_prod + queue->pending_cons; |
121fa4b7 ZK |
378 | } |
379 | ||
f51de243 ZK |
380 | irqreturn_t xenvif_interrupt(int irq, void *dev_id); |
381 | ||
e1f00a69 WL |
382 | extern bool separate_tx_rx_irq; |
383 | ||
09350788 | 384 | extern unsigned int rx_drain_timeout_msecs; |
26c0e102 | 385 | extern unsigned int rx_stall_timeout_msecs; |
8d3d53b3 | 386 | extern unsigned int xenvif_max_queues; |
40d8abde | 387 | extern unsigned int xenvif_hash_cache_size; |
09350788 | 388 | |
f51de243 ZK |
389 | #ifdef CONFIG_DEBUG_FS |
390 | extern struct dentry *xen_netback_dbg_root; | |
391 | #endif | |
392 | ||
a64bd934 WL |
393 | void xenvif_skb_zerocopy_prepare(struct xenvif_queue *queue, |
394 | struct sk_buff *skb); | |
395 | void xenvif_skb_zerocopy_complete(struct xenvif_queue *queue); | |
396 | ||
210c34dc PD |
397 | /* Multicast control */ |
398 | bool xenvif_mcast_match(struct xenvif *vif, const u8 *addr); | |
399 | void xenvif_mcast_addr_list_free(struct xenvif *vif); | |
400 | ||
40d8abde PD |
401 | /* Hash */ |
402 | void xenvif_init_hash(struct xenvif *vif); | |
403 | void xenvif_deinit_hash(struct xenvif *vif); | |
404 | ||
405 | u32 xenvif_set_hash_alg(struct xenvif *vif, u32 alg); | |
406 | u32 xenvif_get_hash_flags(struct xenvif *vif, u32 *flags); | |
407 | u32 xenvif_set_hash_flags(struct xenvif *vif, u32 flags); | |
408 | u32 xenvif_set_hash_key(struct xenvif *vif, u32 gref, u32 len); | |
409 | u32 xenvif_set_hash_mapping_size(struct xenvif *vif, u32 size); | |
410 | u32 xenvif_set_hash_mapping(struct xenvif *vif, u32 gref, u32 len, | |
411 | u32 off); | |
412 | ||
413 | void xenvif_set_skb_hash(struct xenvif *vif, struct sk_buff *skb); | |
414 | ||
f942dc25 | 415 | #endif /* __XEN_NETBACK__COMMON_H__ */ |