common.h 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415
  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. #ifndef __XEN_NETBACK__COMMON_H__
  27. #define __XEN_NETBACK__COMMON_H__
  28. #define pr_fmt(fmt) KBUILD_MODNAME ":%s: " fmt, __func__
  29. #include <linux/module.h>
  30. #include <linux/interrupt.h>
  31. #include <linux/slab.h>
  32. #include <linux/ip.h>
  33. #include <linux/in.h>
  34. #include <linux/io.h>
  35. #include <linux/netdevice.h>
  36. #include <linux/etherdevice.h>
  37. #include <linux/wait.h>
  38. #include <linux/sched.h>
  39. #include <xen/interface/io/netif.h>
  40. #include <xen/interface/grant_table.h>
  41. #include <xen/grant_table.h>
  42. #include <xen/xenbus.h>
  43. #include <xen/page.h>
  44. #include <linux/debugfs.h>
  45. typedef unsigned int pending_ring_idx_t;
  46. #define INVALID_PENDING_RING_IDX (~0U)
  47. struct pending_tx_info {
  48. struct xen_netif_tx_request req; /* tx request */
  49. unsigned int extra_count;
  50. /* Callback data for released SKBs. The callback is always
  51. * xenvif_zerocopy_callback, desc contains the pending_idx, which is
  52. * also an index in pending_tx_info array. It is initialized in
  53. * xenvif_alloc and it never changes.
  54. * skb_shinfo(skb)->destructor_arg points to the first mapped slot's
  55. * callback_struct in this array of struct pending_tx_info's, then ctx
  56. * to the next, or NULL if there is no more slot for this skb.
  57. * ubuf_to_vif is a helper which finds the struct xenvif from a pointer
  58. * to this field.
  59. */
  60. struct ubuf_info callback_struct;
  61. };
  62. #define XEN_NETIF_TX_RING_SIZE __CONST_RING_SIZE(xen_netif_tx, XEN_PAGE_SIZE)
  63. #define XEN_NETIF_RX_RING_SIZE __CONST_RING_SIZE(xen_netif_rx, XEN_PAGE_SIZE)
  64. struct xenvif_rx_meta {
  65. int id;
  66. int size;
  67. int gso_type;
  68. int gso_size;
  69. };
  70. #define GSO_BIT(type) \
  71. (1 << XEN_NETIF_GSO_TYPE_ ## type)
  72. /* Discriminate from any valid pending_idx value. */
  73. #define INVALID_PENDING_IDX 0xFFFF
  74. #define MAX_BUFFER_OFFSET XEN_PAGE_SIZE
  75. #define MAX_PENDING_REQS XEN_NETIF_TX_RING_SIZE
  76. /* The maximum number of frags is derived from the size of a grant (same
  77. * as a Xen page size for now).
  78. */
  79. #define MAX_XEN_SKB_FRAGS (65536 / XEN_PAGE_SIZE + 1)
  80. #define NETBACK_INVALID_HANDLE -1
  81. /* To avoid confusion, we define XEN_NETBK_LEGACY_SLOTS_MAX indicating
  82. * the maximum slots a valid packet can use. Now this value is defined
  83. * to be XEN_NETIF_NR_SLOTS_MIN, which is supposed to be supported by
  84. * all backend.
  85. */
  86. #define XEN_NETBK_LEGACY_SLOTS_MAX XEN_NETIF_NR_SLOTS_MIN
  87. /* Queue name is interface name with "-qNNN" appended */
  88. #define QUEUE_NAME_SIZE (IFNAMSIZ + 5)
  89. /* IRQ name is queue name with "-tx" or "-rx" appended */
  90. #define IRQ_NAME_SIZE (QUEUE_NAME_SIZE + 3)
  91. struct xenvif;
  92. struct xenvif_stats {
  93. /* Stats fields to be updated per-queue.
  94. * A subset of struct net_device_stats that contains only the
  95. * fields that are updated in netback.c for each queue.
  96. */
  97. unsigned int rx_bytes;
  98. unsigned int rx_packets;
  99. unsigned int tx_bytes;
  100. unsigned int tx_packets;
  101. /* Additional stats used by xenvif */
  102. unsigned long rx_gso_checksum_fixup;
  103. unsigned long tx_zerocopy_sent;
  104. unsigned long tx_zerocopy_success;
  105. unsigned long tx_zerocopy_fail;
  106. unsigned long tx_frag_overflow;
  107. };
  108. #define COPY_BATCH_SIZE 64
  109. struct xenvif_copy_state {
  110. struct gnttab_copy op[COPY_BATCH_SIZE];
  111. RING_IDX idx[COPY_BATCH_SIZE];
  112. unsigned int num;
  113. struct sk_buff_head *completed;
  114. };
  115. struct xenvif_queue { /* Per-queue data for xenvif */
  116. unsigned int id; /* Queue ID, 0-based */
  117. char name[QUEUE_NAME_SIZE]; /* DEVNAME-qN */
  118. struct xenvif *vif; /* Parent VIF */
  119. /* Use NAPI for guest TX */
  120. struct napi_struct napi;
  121. /* When feature-split-event-channels = 0, tx_irq = rx_irq. */
  122. unsigned int tx_irq;
  123. /* Only used when feature-split-event-channels = 1 */
  124. char tx_irq_name[IRQ_NAME_SIZE]; /* DEVNAME-qN-tx */
  125. struct xen_netif_tx_back_ring tx;
  126. struct sk_buff_head tx_queue;
  127. struct page *mmap_pages[MAX_PENDING_REQS];
  128. pending_ring_idx_t pending_prod;
  129. pending_ring_idx_t pending_cons;
  130. u16 pending_ring[MAX_PENDING_REQS];
  131. struct pending_tx_info pending_tx_info[MAX_PENDING_REQS];
  132. grant_handle_t grant_tx_handle[MAX_PENDING_REQS];
  133. struct gnttab_copy tx_copy_ops[MAX_PENDING_REQS];
  134. struct gnttab_map_grant_ref tx_map_ops[MAX_PENDING_REQS];
  135. struct gnttab_unmap_grant_ref tx_unmap_ops[MAX_PENDING_REQS];
  136. /* passed to gnttab_[un]map_refs with pages under (un)mapping */
  137. struct page *pages_to_map[MAX_PENDING_REQS];
  138. struct page *pages_to_unmap[MAX_PENDING_REQS];
  139. /* This prevents zerocopy callbacks to race over dealloc_ring */
  140. spinlock_t callback_lock;
  141. /* This prevents dealloc thread and NAPI instance to race over response
  142. * creation and pending_ring in xenvif_idx_release. In xenvif_tx_err
  143. * it only protect response creation
  144. */
  145. spinlock_t response_lock;
  146. pending_ring_idx_t dealloc_prod;
  147. pending_ring_idx_t dealloc_cons;
  148. u16 dealloc_ring[MAX_PENDING_REQS];
  149. struct task_struct *dealloc_task;
  150. wait_queue_head_t dealloc_wq;
  151. atomic_t inflight_packets;
  152. /* Use kthread for guest RX */
  153. struct task_struct *task;
  154. wait_queue_head_t wq;
  155. /* When feature-split-event-channels = 0, tx_irq = rx_irq. */
  156. unsigned int rx_irq;
  157. /* Only used when feature-split-event-channels = 1 */
  158. char rx_irq_name[IRQ_NAME_SIZE]; /* DEVNAME-qN-rx */
  159. struct xen_netif_rx_back_ring rx;
  160. struct sk_buff_head rx_queue;
  161. unsigned int rx_queue_max;
  162. unsigned int rx_queue_len;
  163. unsigned long last_rx_time;
  164. bool stalled;
  165. struct xenvif_copy_state rx_copy;
  166. /* Transmit shaping: allow 'credit_bytes' every 'credit_usec'. */
  167. unsigned long credit_bytes;
  168. unsigned long credit_usec;
  169. unsigned long remaining_credit;
  170. struct timer_list credit_timeout;
  171. u64 credit_window_start;
  172. bool rate_limited;
  173. /* Statistics */
  174. struct xenvif_stats stats;
  175. };
  176. enum state_bit_shift {
  177. /* This bit marks that the vif is connected */
  178. VIF_STATUS_CONNECTED,
  179. };
  180. struct xenvif_mcast_addr {
  181. struct list_head entry;
  182. struct rcu_head rcu;
  183. u8 addr[6];
  184. };
  185. #define XEN_NETBK_MCAST_MAX 64
  186. #define XEN_NETBK_MAX_HASH_KEY_SIZE 40
  187. #define XEN_NETBK_MAX_HASH_MAPPING_SIZE 128
  188. #define XEN_NETBK_HASH_TAG_SIZE 40
  189. struct xenvif_hash_cache_entry {
  190. struct list_head link;
  191. struct rcu_head rcu;
  192. u8 tag[XEN_NETBK_HASH_TAG_SIZE];
  193. unsigned int len;
  194. u32 val;
  195. int seq;
  196. };
  197. struct xenvif_hash_cache {
  198. spinlock_t lock;
  199. struct list_head list;
  200. unsigned int count;
  201. atomic_t seq;
  202. };
  203. struct xenvif_hash {
  204. unsigned int alg;
  205. u32 flags;
  206. u8 key[XEN_NETBK_MAX_HASH_KEY_SIZE];
  207. u32 mapping[XEN_NETBK_MAX_HASH_MAPPING_SIZE];
  208. unsigned int size;
  209. struct xenvif_hash_cache cache;
  210. };
  211. struct xenvif {
  212. /* Unique identifier for this interface. */
  213. domid_t domid;
  214. unsigned int handle;
  215. u8 fe_dev_addr[6];
  216. struct list_head fe_mcast_addr;
  217. unsigned int fe_mcast_count;
  218. /* Frontend feature information. */
  219. int gso_mask;
  220. u8 can_sg:1;
  221. u8 ip_csum:1;
  222. u8 ipv6_csum:1;
  223. u8 multicast_control:1;
  224. /* Is this interface disabled? True when backend discovers
  225. * frontend is rogue.
  226. */
  227. bool disabled;
  228. unsigned long status;
  229. unsigned long drain_timeout;
  230. unsigned long stall_timeout;
  231. /* Queues */
  232. struct xenvif_queue *queues;
  233. unsigned int num_queues; /* active queues, resource allocated */
  234. unsigned int stalled_queues;
  235. struct xenvif_hash hash;
  236. struct xenbus_watch credit_watch;
  237. struct xenbus_watch mcast_ctrl_watch;
  238. spinlock_t lock;
  239. #ifdef CONFIG_DEBUG_FS
  240. struct dentry *xenvif_dbg_root;
  241. #endif
  242. struct xen_netif_ctrl_back_ring ctrl;
  243. unsigned int ctrl_irq;
  244. /* Miscellaneous private stuff. */
  245. struct net_device *dev;
  246. };
  247. struct xenvif_rx_cb {
  248. unsigned long expires;
  249. int meta_slots_used;
  250. };
  251. #define XENVIF_RX_CB(skb) ((struct xenvif_rx_cb *)(skb)->cb)
  252. static inline struct xenbus_device *xenvif_to_xenbus_device(struct xenvif *vif)
  253. {
  254. return to_xenbus_device(vif->dev->dev.parent);
  255. }
  256. void xenvif_tx_credit_callback(unsigned long data);
  257. struct xenvif *xenvif_alloc(struct device *parent,
  258. domid_t domid,
  259. unsigned int handle);
  260. int xenvif_init_queue(struct xenvif_queue *queue);
  261. void xenvif_deinit_queue(struct xenvif_queue *queue);
  262. int xenvif_connect_data(struct xenvif_queue *queue,
  263. unsigned long tx_ring_ref,
  264. unsigned long rx_ring_ref,
  265. unsigned int tx_evtchn,
  266. unsigned int rx_evtchn);
  267. void xenvif_disconnect_data(struct xenvif *vif);
  268. int xenvif_connect_ctrl(struct xenvif *vif, grant_ref_t ring_ref,
  269. unsigned int evtchn);
  270. void xenvif_disconnect_ctrl(struct xenvif *vif);
  271. void xenvif_free(struct xenvif *vif);
  272. int xenvif_xenbus_init(void);
  273. void xenvif_xenbus_fini(void);
  274. int xenvif_schedulable(struct xenvif *vif);
  275. int xenvif_queue_stopped(struct xenvif_queue *queue);
  276. void xenvif_wake_queue(struct xenvif_queue *queue);
  277. /* (Un)Map communication rings. */
  278. void xenvif_unmap_frontend_data_rings(struct xenvif_queue *queue);
  279. int xenvif_map_frontend_data_rings(struct xenvif_queue *queue,
  280. grant_ref_t tx_ring_ref,
  281. grant_ref_t rx_ring_ref);
  282. /* Check for SKBs from frontend and schedule backend processing */
  283. void xenvif_napi_schedule_or_enable_events(struct xenvif_queue *queue);
  284. /* Prevent the device from generating any further traffic. */
  285. void xenvif_carrier_off(struct xenvif *vif);
  286. int xenvif_tx_action(struct xenvif_queue *queue, int budget);
  287. int xenvif_kthread_guest_rx(void *data);
  288. void xenvif_kick_thread(struct xenvif_queue *queue);
  289. int xenvif_dealloc_kthread(void *data);
  290. irqreturn_t xenvif_ctrl_irq_fn(int irq, void *data);
  291. void xenvif_rx_action(struct xenvif_queue *queue);
  292. void xenvif_rx_queue_tail(struct xenvif_queue *queue, struct sk_buff *skb);
  293. void xenvif_carrier_on(struct xenvif *vif);
  294. /* Callback from stack when TX packet can be released */
  295. void xenvif_zerocopy_callback(struct ubuf_info *ubuf, bool zerocopy_success);
  296. /* Unmap a pending page and release it back to the guest */
  297. void xenvif_idx_unmap(struct xenvif_queue *queue, u16 pending_idx);
  298. static inline pending_ring_idx_t nr_pending_reqs(struct xenvif_queue *queue)
  299. {
  300. return MAX_PENDING_REQS -
  301. queue->pending_prod + queue->pending_cons;
  302. }
  303. irqreturn_t xenvif_interrupt(int irq, void *dev_id);
  304. extern bool separate_tx_rx_irq;
  305. extern unsigned int rx_drain_timeout_msecs;
  306. extern unsigned int rx_stall_timeout_msecs;
  307. extern unsigned int xenvif_max_queues;
  308. extern unsigned int xenvif_hash_cache_size;
  309. #ifdef CONFIG_DEBUG_FS
  310. extern struct dentry *xen_netback_dbg_root;
  311. #endif
  312. void xenvif_skb_zerocopy_prepare(struct xenvif_queue *queue,
  313. struct sk_buff *skb);
  314. void xenvif_skb_zerocopy_complete(struct xenvif_queue *queue);
  315. /* Multicast control */
  316. bool xenvif_mcast_match(struct xenvif *vif, const u8 *addr);
  317. void xenvif_mcast_addr_list_free(struct xenvif *vif);
  318. /* Hash */
  319. void xenvif_init_hash(struct xenvif *vif);
  320. void xenvif_deinit_hash(struct xenvif *vif);
  321. u32 xenvif_set_hash_alg(struct xenvif *vif, u32 alg);
  322. u32 xenvif_get_hash_flags(struct xenvif *vif, u32 *flags);
  323. u32 xenvif_set_hash_flags(struct xenvif *vif, u32 flags);
  324. u32 xenvif_set_hash_key(struct xenvif *vif, u32 gref, u32 len);
  325. u32 xenvif_set_hash_mapping_size(struct xenvif *vif, u32 size);
  326. u32 xenvif_set_hash_mapping(struct xenvif *vif, u32 gref, u32 len,
  327. u32 off);
  328. void xenvif_set_skb_hash(struct xenvif *vif, struct sk_buff *skb);
  329. #ifdef CONFIG_DEBUG_FS
  330. void xenvif_dump_hash_info(struct xenvif *vif, struct seq_file *m);
  331. #endif
  332. #endif /* __XEN_NETBACK__COMMON_H__ */