ll_poll.h 4.6 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200
  1. /*
  2. * Low Latency Sockets
  3. * Copyright(c) 2013 Intel Corporation.
  4. *
  5. * This program is free software; you can redistribute it and/or modify it
  6. * under the terms and conditions of the GNU General Public License,
  7. * version 2, as published by the Free Software Foundation.
  8. *
  9. * This program is distributed in the hope it will be useful, but WITHOUT
  10. * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  11. * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
  12. * more details.
  13. *
  14. * You should have received a copy of the GNU General Public License along with
  15. * this program; if not, write to the Free Software Foundation, Inc.,
  16. * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
  17. *
  18. * Author: Eliezer Tamir
  19. *
  20. * Contact Information:
  21. * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
  22. */
  23. #ifndef _LINUX_NET_LL_POLL_H
  24. #define _LINUX_NET_LL_POLL_H
  25. #include <linux/netdevice.h>
  26. #include <net/ip.h>
  27. #ifdef CONFIG_NET_LL_RX_POLL
  28. struct napi_struct;
  29. extern unsigned int sysctl_net_ll_read __read_mostly;
  30. extern unsigned int sysctl_net_ll_poll __read_mostly;
  31. /* return values from ndo_ll_poll */
  32. #define LL_FLUSH_FAILED -1
  33. #define LL_FLUSH_BUSY -2
  34. static inline unsigned int ll_get_flag(void)
  35. {
  36. return sysctl_net_ll_poll ? POLL_LL : 0;
  37. }
  38. /* a wrapper to make debug_smp_processor_id() happy
  39. * we can use sched_clock() because we don't care much about precision
  40. * we only care that the average is bounded
  41. */
  42. #ifdef CONFIG_DEBUG_PREEMPT
  43. static inline u64 ll_sched_clock(void)
  44. {
  45. u64 rc;
  46. preempt_disable_notrace();
  47. rc = sched_clock();
  48. preempt_enable_no_resched_notrace();
  49. return rc;
  50. }
  51. #else /* CONFIG_DEBUG_PREEMPT */
  52. static inline u64 ll_sched_clock(void)
  53. {
  54. return sched_clock();
  55. }
  56. #endif /* CONFIG_DEBUG_PREEMPT */
  57. /* we don't mind a ~2.5% imprecision so <<10 instead of *1000
  58. * sk->sk_ll_usec is a u_int so this can't overflow
  59. */
  60. static inline u64 ll_sk_run_time(struct sock *sk)
  61. {
  62. return (u64)ACCESS_ONCE(sk->sk_ll_usec) << 10;
  63. }
  64. /* in poll/select we use the global sysctl_net_ll_poll value
  65. * only call sched_clock() if enabled
  66. */
  67. static inline u64 ll_run_time(void)
  68. {
  69. return (u64)ACCESS_ONCE(sysctl_net_ll_poll) << 10;
  70. }
  71. /* if flag is not set we don't need to know the time */
  72. static inline u64 ll_start_time(unsigned int flag)
  73. {
  74. return flag ? ll_sched_clock() : 0;
  75. }
  76. static inline bool sk_valid_ll(struct sock *sk)
  77. {
  78. return sk->sk_ll_usec && sk->sk_napi_id &&
  79. !need_resched() && !signal_pending(current);
  80. }
  81. /* careful! time_in_range64 will evaluate now twice */
  82. static inline bool can_poll_ll(u64 start_time, u64 run_time)
  83. {
  84. u64 now = ll_sched_clock();
  85. return time_in_range64(now, start_time, start_time + run_time);
  86. }
  87. /* when used in sock_poll() nonblock is known at compile time to be true
  88. * so the loop and end_time will be optimized out
  89. */
  90. static inline bool sk_poll_ll(struct sock *sk, int nonblock)
  91. {
  92. u64 start_time = ll_start_time(!nonblock);
  93. u64 run_time = ll_sk_run_time(sk);
  94. const struct net_device_ops *ops;
  95. struct napi_struct *napi;
  96. int rc = false;
  97. /*
  98. * rcu read lock for napi hash
  99. * bh so we don't race with net_rx_action
  100. */
  101. rcu_read_lock_bh();
  102. napi = napi_by_id(sk->sk_napi_id);
  103. if (!napi)
  104. goto out;
  105. ops = napi->dev->netdev_ops;
  106. if (!ops->ndo_ll_poll)
  107. goto out;
  108. do {
  109. rc = ops->ndo_ll_poll(napi);
  110. if (rc == LL_FLUSH_FAILED)
  111. break; /* permanent failure */
  112. if (rc > 0)
  113. /* local bh are disabled so it is ok to use _BH */
  114. NET_ADD_STATS_BH(sock_net(sk),
  115. LINUX_MIB_LOWLATENCYRXPACKETS, rc);
  116. } while (!nonblock && skb_queue_empty(&sk->sk_receive_queue) &&
  117. can_poll_ll(start_time, run_time));
  118. rc = !skb_queue_empty(&sk->sk_receive_queue);
  119. out:
  120. rcu_read_unlock_bh();
  121. return rc;
  122. }
  123. /* used in the NIC receive handler to mark the skb */
  124. static inline void skb_mark_ll(struct sk_buff *skb, struct napi_struct *napi)
  125. {
  126. skb->napi_id = napi->napi_id;
  127. }
  128. /* used in the protocol hanlder to propagate the napi_id to the socket */
  129. static inline void sk_mark_ll(struct sock *sk, struct sk_buff *skb)
  130. {
  131. sk->sk_napi_id = skb->napi_id;
  132. }
  133. #else /* CONFIG_NET_LL_RX_POLL */
  134. static inline unsigned long ll_get_flag(void)
  135. {
  136. return 0;
  137. }
  138. static inline u64 sk_ll_end_time(struct sock *sk)
  139. {
  140. return 0;
  141. }
  142. static inline u64 ll_end_time(void)
  143. {
  144. return 0;
  145. }
  146. static inline bool sk_valid_ll(struct sock *sk)
  147. {
  148. return false;
  149. }
  150. static inline bool sk_poll_ll(struct sock *sk, int nonblock)
  151. {
  152. return false;
  153. }
  154. static inline void skb_mark_ll(struct sk_buff *skb, struct napi_struct *napi)
  155. {
  156. }
  157. static inline void sk_mark_ll(struct sock *sk, struct sk_buff *skb)
  158. {
  159. }
  160. static inline bool can_poll_ll(u64 end_time)
  161. {
  162. return false;
  163. }
  164. #endif /* CONFIG_NET_LL_RX_POLL */
  165. #endif /* _LINUX_NET_LL_POLL_H */