common.h 8.5 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262
  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. typedef unsigned int pending_ring_idx_t;
  44. #define INVALID_PENDING_RING_IDX (~0U)
  45. struct pending_tx_info {
  46. struct xen_netif_tx_request req; /* tx request */
  47. /* Callback data for released SKBs. The callback is always
  48. * xenvif_zerocopy_callback, desc contains the pending_idx, which is
  49. * also an index in pending_tx_info array. It is initialized in
  50. * xenvif_alloc and it never changes.
  51. * skb_shinfo(skb)->destructor_arg points to the first mapped slot's
  52. * callback_struct in this array of struct pending_tx_info's, then ctx
  53. * to the next, or NULL if there is no more slot for this skb.
  54. * ubuf_to_vif is a helper which finds the struct xenvif from a pointer
  55. * to this field.
  56. */
  57. struct ubuf_info callback_struct;
  58. };
  59. #define XEN_NETIF_TX_RING_SIZE __CONST_RING_SIZE(xen_netif_tx, PAGE_SIZE)
  60. #define XEN_NETIF_RX_RING_SIZE __CONST_RING_SIZE(xen_netif_rx, PAGE_SIZE)
  61. struct xenvif_rx_meta {
  62. int id;
  63. int size;
  64. int gso_type;
  65. int gso_size;
  66. };
  67. #define GSO_BIT(type) \
  68. (1 << XEN_NETIF_GSO_TYPE_ ## type)
  69. /* Discriminate from any valid pending_idx value. */
  70. #define INVALID_PENDING_IDX 0xFFFF
  71. #define MAX_BUFFER_OFFSET PAGE_SIZE
  72. #define MAX_PENDING_REQS XEN_NETIF_TX_RING_SIZE
  73. /* It's possible for an skb to have a maximal number of frags
  74. * but still be less than MAX_BUFFER_OFFSET in size. Thus the
  75. * worst-case number of copy operations is MAX_SKB_FRAGS per
  76. * ring slot.
  77. */
  78. #define MAX_GRANT_COPY_OPS (MAX_SKB_FRAGS * XEN_NETIF_RX_RING_SIZE)
  79. #define NETBACK_INVALID_HANDLE -1
  80. /* To avoid confusion, we define XEN_NETBK_LEGACY_SLOTS_MAX indicating
  81. * the maximum slots a valid packet can use. Now this value is defined
  82. * to be XEN_NETIF_NR_SLOTS_MIN, which is supposed to be supported by
  83. * all backend.
  84. */
  85. #define XEN_NETBK_LEGACY_SLOTS_MAX XEN_NETIF_NR_SLOTS_MIN
  86. struct xenvif {
  87. /* Unique identifier for this interface. */
  88. domid_t domid;
  89. unsigned int handle;
  90. /* Use NAPI for guest TX */
  91. struct napi_struct napi;
  92. /* When feature-split-event-channels = 0, tx_irq = rx_irq. */
  93. unsigned int tx_irq;
  94. /* Only used when feature-split-event-channels = 1 */
  95. char tx_irq_name[IFNAMSIZ+4]; /* DEVNAME-tx */
  96. struct xen_netif_tx_back_ring tx;
  97. struct sk_buff_head tx_queue;
  98. struct page *mmap_pages[MAX_PENDING_REQS];
  99. pending_ring_idx_t pending_prod;
  100. pending_ring_idx_t pending_cons;
  101. u16 pending_ring[MAX_PENDING_REQS];
  102. struct pending_tx_info pending_tx_info[MAX_PENDING_REQS];
  103. grant_handle_t grant_tx_handle[MAX_PENDING_REQS];
  104. struct gnttab_map_grant_ref tx_map_ops[MAX_PENDING_REQS];
  105. struct gnttab_unmap_grant_ref tx_unmap_ops[MAX_PENDING_REQS];
  106. /* passed to gnttab_[un]map_refs with pages under (un)mapping */
  107. struct page *pages_to_map[MAX_PENDING_REQS];
  108. struct page *pages_to_unmap[MAX_PENDING_REQS];
  109. /* This prevents zerocopy callbacks to race over dealloc_ring */
  110. spinlock_t callback_lock;
  111. /* This prevents dealloc thread and NAPI instance to race over response
  112. * creation and pending_ring in xenvif_idx_release. In xenvif_tx_err
  113. * it only protect response creation
  114. */
  115. spinlock_t response_lock;
  116. pending_ring_idx_t dealloc_prod;
  117. pending_ring_idx_t dealloc_cons;
  118. u16 dealloc_ring[MAX_PENDING_REQS];
  119. struct task_struct *dealloc_task;
  120. wait_queue_head_t dealloc_wq;
  121. /* Use kthread for guest RX */
  122. struct task_struct *task;
  123. wait_queue_head_t wq;
  124. /* When feature-split-event-channels = 0, tx_irq = rx_irq. */
  125. unsigned int rx_irq;
  126. /* Only used when feature-split-event-channels = 1 */
  127. char rx_irq_name[IFNAMSIZ+4]; /* DEVNAME-rx */
  128. struct xen_netif_rx_back_ring rx;
  129. struct sk_buff_head rx_queue;
  130. RING_IDX rx_last_skb_slots;
  131. bool rx_queue_purge;
  132. struct timer_list wake_queue;
  133. /* This array is allocated seperately as it is large */
  134. struct gnttab_copy *grant_copy_op;
  135. /* We create one meta structure per ring request we consume, so
  136. * the maximum number is the same as the ring size.
  137. */
  138. struct xenvif_rx_meta meta[XEN_NETIF_RX_RING_SIZE];
  139. u8 fe_dev_addr[6];
  140. /* Frontend feature information. */
  141. int gso_mask;
  142. int gso_prefix_mask;
  143. u8 can_sg:1;
  144. u8 ip_csum:1;
  145. u8 ipv6_csum:1;
  146. /* Internal feature information. */
  147. u8 can_queue:1; /* can queue packets for receiver? */
  148. /* Transmit shaping: allow 'credit_bytes' every 'credit_usec'. */
  149. unsigned long credit_bytes;
  150. unsigned long credit_usec;
  151. unsigned long remaining_credit;
  152. struct timer_list credit_timeout;
  153. u64 credit_window_start;
  154. /* Statistics */
  155. unsigned long rx_gso_checksum_fixup;
  156. unsigned long tx_zerocopy_sent;
  157. unsigned long tx_zerocopy_success;
  158. unsigned long tx_zerocopy_fail;
  159. unsigned long tx_frag_overflow;
  160. /* Miscellaneous private stuff. */
  161. struct net_device *dev;
  162. };
  163. static inline struct xenbus_device *xenvif_to_xenbus_device(struct xenvif *vif)
  164. {
  165. return to_xenbus_device(vif->dev->dev.parent);
  166. }
  167. struct xenvif *xenvif_alloc(struct device *parent,
  168. domid_t domid,
  169. unsigned int handle);
  170. int xenvif_connect(struct xenvif *vif, unsigned long tx_ring_ref,
  171. unsigned long rx_ring_ref, unsigned int tx_evtchn,
  172. unsigned int rx_evtchn);
  173. void xenvif_disconnect(struct xenvif *vif);
  174. void xenvif_free(struct xenvif *vif);
  175. int xenvif_xenbus_init(void);
  176. void xenvif_xenbus_fini(void);
  177. int xenvif_schedulable(struct xenvif *vif);
  178. int xenvif_must_stop_queue(struct xenvif *vif);
  179. /* (Un)Map communication rings. */
  180. void xenvif_unmap_frontend_rings(struct xenvif *vif);
  181. int xenvif_map_frontend_rings(struct xenvif *vif,
  182. grant_ref_t tx_ring_ref,
  183. grant_ref_t rx_ring_ref);
  184. /* Check for SKBs from frontend and schedule backend processing */
  185. void xenvif_check_rx_xenvif(struct xenvif *vif);
  186. /* Prevent the device from generating any further traffic. */
  187. void xenvif_carrier_off(struct xenvif *vif);
  188. int xenvif_tx_action(struct xenvif *vif, int budget);
  189. int xenvif_kthread_guest_rx(void *data);
  190. void xenvif_kick_thread(struct xenvif *vif);
  191. int xenvif_dealloc_kthread(void *data);
  192. /* Determine whether the needed number of slots (req) are available,
  193. * and set req_event if not.
  194. */
  195. bool xenvif_rx_ring_slots_available(struct xenvif *vif, int needed);
  196. void xenvif_stop_queue(struct xenvif *vif);
  197. /* Callback from stack when TX packet can be released */
  198. void xenvif_zerocopy_callback(struct ubuf_info *ubuf, bool zerocopy_success);
  199. /* Unmap a pending page and release it back to the guest */
  200. void xenvif_idx_unmap(struct xenvif *vif, u16 pending_idx);
  201. static inline pending_ring_idx_t nr_pending_reqs(struct xenvif *vif)
  202. {
  203. return MAX_PENDING_REQS -
  204. vif->pending_prod + vif->pending_cons;
  205. }
  206. /* Callback from stack when TX packet can be released */
  207. void xenvif_zerocopy_callback(struct ubuf_info *ubuf, bool zerocopy_success);
  208. extern bool separate_tx_rx_irq;
  209. extern unsigned int rx_drain_timeout_msecs;
  210. extern unsigned int rx_drain_timeout_jiffies;
  211. #endif /* __XEN_NETBACK__COMMON_H__ */