verbs.h 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451
  1. /*
  2. * Copyright(c) 2015 - 2017 Intel Corporation.
  3. *
  4. * This file is provided under a dual BSD/GPLv2 license. When using or
  5. * redistributing this file, you may do so under either license.
  6. *
  7. * GPL LICENSE SUMMARY
  8. *
  9. * This program is free software; you can redistribute it and/or modify
  10. * it under the terms of version 2 of the GNU General Public License as
  11. * published by the Free Software Foundation.
  12. *
  13. * This program is distributed in the hope that it will be useful, but
  14. * WITHOUT ANY WARRANTY; without even the implied warranty of
  15. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  16. * General Public License for more details.
  17. *
  18. * BSD LICENSE
  19. *
  20. * Redistribution and use in source and binary forms, with or without
  21. * modification, are permitted provided that the following conditions
  22. * are met:
  23. *
  24. * - Redistributions of source code must retain the above copyright
  25. * notice, this list of conditions and the following disclaimer.
  26. * - Redistributions in binary form must reproduce the above copyright
  27. * notice, this list of conditions and the following disclaimer in
  28. * the documentation and/or other materials provided with the
  29. * distribution.
  30. * - Neither the name of Intel Corporation nor the names of its
  31. * contributors may be used to endorse or promote products derived
  32. * from this software without specific prior written permission.
  33. *
  34. * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  35. * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  36. * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
  37. * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
  38. * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  39. * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
  40. * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  41. * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  42. * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  43. * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  44. * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  45. *
  46. */
  47. #ifndef HFI1_VERBS_H
  48. #define HFI1_VERBS_H
  49. #include <linux/types.h>
  50. #include <linux/seqlock.h>
  51. #include <linux/kernel.h>
  52. #include <linux/interrupt.h>
  53. #include <linux/kref.h>
  54. #include <linux/workqueue.h>
  55. #include <linux/kthread.h>
  56. #include <linux/completion.h>
  57. #include <linux/slab.h>
  58. #include <rdma/ib_pack.h>
  59. #include <rdma/ib_user_verbs.h>
  60. #include <rdma/ib_mad.h>
  61. #include <rdma/ib_hdrs.h>
  62. #include <rdma/rdma_vt.h>
  63. #include <rdma/rdmavt_qp.h>
  64. #include <rdma/rdmavt_cq.h>
  65. struct hfi1_ctxtdata;
  66. struct hfi1_pportdata;
  67. struct hfi1_devdata;
  68. struct hfi1_packet;
  69. #include "iowait.h"
  70. #define HFI1_MAX_RDMA_ATOMIC 16
  71. /*
  72. * Increment this value if any changes that break userspace ABI
  73. * compatibility are made.
  74. */
  75. #define HFI1_UVERBS_ABI_VERSION 2
  76. /* IB Performance Manager status values */
  77. #define IB_PMA_SAMPLE_STATUS_DONE 0x00
  78. #define IB_PMA_SAMPLE_STATUS_STARTED 0x01
  79. #define IB_PMA_SAMPLE_STATUS_RUNNING 0x02
  80. /* Mandatory IB performance counter select values. */
  81. #define IB_PMA_PORT_XMIT_DATA cpu_to_be16(0x0001)
  82. #define IB_PMA_PORT_RCV_DATA cpu_to_be16(0x0002)
  83. #define IB_PMA_PORT_XMIT_PKTS cpu_to_be16(0x0003)
  84. #define IB_PMA_PORT_RCV_PKTS cpu_to_be16(0x0004)
  85. #define IB_PMA_PORT_XMIT_WAIT cpu_to_be16(0x0005)
  86. #define HFI1_VENDOR_IPG cpu_to_be16(0xFFA0)
  87. #define IB_DEFAULT_GID_PREFIX cpu_to_be64(0xfe80000000000000ULL)
  88. #define OPA_BTH_MIG_REQ BIT(31)
  89. #define RC_OP(x) IB_OPCODE_RC_##x
  90. #define UC_OP(x) IB_OPCODE_UC_##x
  91. /* flags passed by hfi1_ib_rcv() */
  92. enum {
  93. HFI1_HAS_GRH = (1 << 0),
  94. };
  95. struct hfi1_16b_header {
  96. u32 lrh[4];
  97. union {
  98. struct {
  99. struct ib_grh grh;
  100. struct ib_other_headers oth;
  101. } l;
  102. struct ib_other_headers oth;
  103. } u;
  104. } __packed;
  105. struct hfi1_opa_header {
  106. union {
  107. struct ib_header ibh; /* 9B header */
  108. struct hfi1_16b_header opah; /* 16B header */
  109. };
  110. u8 hdr_type; /* 9B or 16B */
  111. } __packed;
  112. struct hfi1_ahg_info {
  113. u32 ahgdesc[2];
  114. u16 tx_flags;
  115. u8 ahgcount;
  116. u8 ahgidx;
  117. };
  118. struct hfi1_sdma_header {
  119. __le64 pbc;
  120. struct hfi1_opa_header hdr;
  121. } __packed;
  122. /*
  123. * hfi1 specific data structures that will be hidden from rvt after the queue
  124. * pair is made common
  125. */
  126. struct hfi1_qp_priv {
  127. struct hfi1_ahg_info *s_ahg; /* ahg info for next header */
  128. struct sdma_engine *s_sde; /* current sde */
  129. struct send_context *s_sendcontext; /* current sendcontext */
  130. u8 s_sc; /* SC[0..4] for next packet */
  131. struct iowait s_iowait;
  132. struct rvt_qp *owner;
  133. u8 hdr_type; /* 9B or 16B */
  134. };
  135. /*
  136. * This structure is used to hold commonly lookedup and computed values during
  137. * the send engine progress.
  138. */
  139. struct hfi1_pkt_state {
  140. struct hfi1_ibdev *dev;
  141. struct hfi1_ibport *ibp;
  142. struct hfi1_pportdata *ppd;
  143. struct verbs_txreq *s_txreq;
  144. unsigned long flags;
  145. unsigned long timeout;
  146. unsigned long timeout_int;
  147. int cpu;
  148. u8 opcode;
  149. bool in_thread;
  150. bool pkts_sent;
  151. };
  152. #define HFI1_PSN_CREDIT 16
  153. struct hfi1_opcode_stats {
  154. u64 n_packets; /* number of packets */
  155. u64 n_bytes; /* total number of bytes */
  156. };
  157. struct hfi1_opcode_stats_perctx {
  158. struct hfi1_opcode_stats stats[256];
  159. };
  160. static inline void inc_opstats(
  161. u32 tlen,
  162. struct hfi1_opcode_stats *stats)
  163. {
  164. #ifdef CONFIG_DEBUG_FS
  165. stats->n_bytes += tlen;
  166. stats->n_packets++;
  167. #endif
  168. }
  169. struct hfi1_ibport {
  170. struct rvt_qp __rcu *qp[2];
  171. struct rvt_ibport rvp;
  172. /* the first 16 entries are sl_to_vl for !OPA */
  173. u8 sl_to_sc[32];
  174. u8 sc_to_sl[32];
  175. };
  176. struct hfi1_ibdev {
  177. struct rvt_dev_info rdi; /* Must be first */
  178. /* QP numbers are shared by all IB ports */
  179. /* protect txwait list */
  180. seqlock_t txwait_lock ____cacheline_aligned_in_smp;
  181. struct list_head txwait; /* list for wait verbs_txreq */
  182. struct list_head memwait; /* list for wait kernel memory */
  183. struct kmem_cache *verbs_txreq_cache;
  184. u64 n_txwait;
  185. u64 n_kmem_wait;
  186. /* protect iowait lists */
  187. seqlock_t iowait_lock ____cacheline_aligned_in_smp;
  188. u64 n_piowait;
  189. u64 n_piodrain;
  190. struct timer_list mem_timer;
  191. #ifdef CONFIG_DEBUG_FS
  192. /* per HFI debugfs */
  193. struct dentry *hfi1_ibdev_dbg;
  194. /* per HFI symlinks to above */
  195. struct dentry *hfi1_ibdev_link;
  196. #ifdef CONFIG_FAULT_INJECTION
  197. struct fault_opcode *fault_opcode;
  198. struct fault_packet *fault_packet;
  199. bool fault_suppress_err;
  200. #endif
  201. #endif
  202. };
  203. static inline struct hfi1_ibdev *to_idev(struct ib_device *ibdev)
  204. {
  205. struct rvt_dev_info *rdi;
  206. rdi = container_of(ibdev, struct rvt_dev_info, ibdev);
  207. return container_of(rdi, struct hfi1_ibdev, rdi);
  208. }
  209. static inline struct rvt_qp *iowait_to_qp(struct iowait *s_iowait)
  210. {
  211. struct hfi1_qp_priv *priv;
  212. priv = container_of(s_iowait, struct hfi1_qp_priv, s_iowait);
  213. return priv->owner;
  214. }
  215. /*
  216. * Send if not busy or waiting for I/O and either
  217. * a RC response is pending or we can process send work requests.
  218. */
  219. static inline int hfi1_send_ok(struct rvt_qp *qp)
  220. {
  221. return !(qp->s_flags & (RVT_S_BUSY | RVT_S_ANY_WAIT_IO)) &&
  222. (qp->s_hdrwords || (qp->s_flags & RVT_S_RESP_PENDING) ||
  223. !(qp->s_flags & RVT_S_ANY_WAIT_SEND));
  224. }
  225. /*
  226. * This must be called with s_lock held.
  227. */
  228. void hfi1_bad_pkey(struct hfi1_ibport *ibp, u32 key, u32 sl,
  229. u32 qp1, u32 qp2, u32 lid1, u32 lid2);
  230. void hfi1_cap_mask_chg(struct rvt_dev_info *rdi, u8 port_num);
  231. void hfi1_sys_guid_chg(struct hfi1_ibport *ibp);
  232. void hfi1_node_desc_chg(struct hfi1_ibport *ibp);
  233. int hfi1_process_mad(struct ib_device *ibdev, int mad_flags, u8 port,
  234. const struct ib_wc *in_wc, const struct ib_grh *in_grh,
  235. const struct ib_mad_hdr *in_mad, size_t in_mad_size,
  236. struct ib_mad_hdr *out_mad, size_t *out_mad_size,
  237. u16 *out_mad_pkey_index);
  238. /*
  239. * The PSN_MASK and PSN_SHIFT allow for
  240. * 1) comparing two PSNs
  241. * 2) returning the PSN with any upper bits masked
  242. * 3) returning the difference between to PSNs
  243. *
  244. * The number of significant bits in the PSN must
  245. * necessarily be at least one bit less than
  246. * the container holding the PSN.
  247. */
  248. #define PSN_MASK 0x7FFFFFFF
  249. #define PSN_SHIFT 1
  250. #define PSN_MODIFY_MASK 0xFFFFFF
  251. /*
  252. * Compare two PSNs
  253. * Returns an integer <, ==, or > than zero.
  254. */
  255. static inline int cmp_psn(u32 a, u32 b)
  256. {
  257. return (((int)a) - ((int)b)) << PSN_SHIFT;
  258. }
  259. /*
  260. * Return masked PSN
  261. */
  262. static inline u32 mask_psn(u32 a)
  263. {
  264. return a & PSN_MASK;
  265. }
  266. /*
  267. * Return delta between two PSNs
  268. */
  269. static inline u32 delta_psn(u32 a, u32 b)
  270. {
  271. return (((int)a - (int)b) << PSN_SHIFT) >> PSN_SHIFT;
  272. }
  273. struct verbs_txreq;
  274. void hfi1_put_txreq(struct verbs_txreq *tx);
  275. int hfi1_verbs_send(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
  276. void hfi1_copy_sge(struct rvt_sge_state *ss, void *data, u32 length,
  277. bool release, bool copy_last);
  278. void hfi1_cnp_rcv(struct hfi1_packet *packet);
  279. void hfi1_uc_rcv(struct hfi1_packet *packet);
  280. void hfi1_rc_rcv(struct hfi1_packet *packet);
  281. void hfi1_rc_hdrerr(
  282. struct hfi1_ctxtdata *rcd,
  283. struct hfi1_packet *packet,
  284. struct rvt_qp *qp);
  285. u8 ah_to_sc(struct ib_device *ibdev, struct rdma_ah_attr *ah_attr);
  286. void hfi1_rc_send_complete(struct rvt_qp *qp, struct hfi1_opa_header *opah);
  287. void hfi1_ud_rcv(struct hfi1_packet *packet);
  288. int hfi1_lookup_pkey_idx(struct hfi1_ibport *ibp, u16 pkey);
  289. int hfi1_rvt_get_rwqe(struct rvt_qp *qp, int wr_id_only);
  290. void hfi1_migrate_qp(struct rvt_qp *qp);
  291. int hfi1_check_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
  292. int attr_mask, struct ib_udata *udata);
  293. void hfi1_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
  294. int attr_mask, struct ib_udata *udata);
  295. void hfi1_restart_rc(struct rvt_qp *qp, u32 psn, int wait);
  296. int hfi1_check_send_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe);
  297. extern const u32 rc_only_opcode;
  298. extern const u32 uc_only_opcode;
  299. int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct hfi1_packet *packet);
  300. u32 hfi1_make_grh(struct hfi1_ibport *ibp, struct ib_grh *hdr,
  301. const struct ib_global_route *grh, u32 hwords, u32 nwords);
  302. void hfi1_make_ruc_header(struct rvt_qp *qp, struct ib_other_headers *ohdr,
  303. u32 bth0, u32 bth2, int middle,
  304. struct hfi1_pkt_state *ps);
  305. void _hfi1_do_send(struct work_struct *work);
  306. void hfi1_do_send_from_rvt(struct rvt_qp *qp);
  307. void hfi1_do_send(struct rvt_qp *qp, bool in_thread);
  308. void hfi1_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
  309. enum ib_wc_status status);
  310. void hfi1_send_rc_ack(struct hfi1_ctxtdata *rcd, struct rvt_qp *qp,
  311. bool is_fecn);
  312. int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
  313. int hfi1_make_uc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
  314. int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
  315. int hfi1_register_ib_device(struct hfi1_devdata *);
  316. void hfi1_unregister_ib_device(struct hfi1_devdata *);
  317. void hfi1_ib_rcv(struct hfi1_packet *packet);
  318. void hfi1_16B_rcv(struct hfi1_packet *packet);
  319. unsigned hfi1_get_npkeys(struct hfi1_devdata *);
  320. int hfi1_verbs_send_dma(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
  321. u64 pbc);
  322. int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
  323. u64 pbc);
  324. int hfi1_wss_init(void);
  325. void hfi1_wss_exit(void);
  326. /* platform specific: return the lowest level cache (llc) size, in KiB */
  327. static inline int wss_llc_size(void)
  328. {
  329. /* assume that the boot CPU value is universal for all CPUs */
  330. return boot_cpu_data.x86_cache_size;
  331. }
  332. /* platform specific: cacheless copy */
  333. static inline void cacheless_memcpy(void *dst, void *src, size_t n)
  334. {
  335. /*
  336. * Use the only available X64 cacheless copy. Add a __user cast
  337. * to quiet sparse. The src agument is already in the kernel so
  338. * there are no security issues. The extra fault recovery machinery
  339. * is not invoked.
  340. */
  341. __copy_user_nocache(dst, (void __user *)src, n, 0);
  342. }
  343. extern const enum ib_wc_opcode ib_hfi1_wc_opcode[];
  344. extern const u8 hdr_len_by_opcode[];
  345. extern const int ib_rvt_state_ops[];
  346. extern __be64 ib_hfi1_sys_image_guid; /* in network order */
  347. extern unsigned int hfi1_max_cqes;
  348. extern unsigned int hfi1_max_cqs;
  349. extern unsigned int hfi1_max_qp_wrs;
  350. extern unsigned int hfi1_max_qps;
  351. extern unsigned int hfi1_max_sges;
  352. extern unsigned int hfi1_max_mcast_grps;
  353. extern unsigned int hfi1_max_mcast_qp_attached;
  354. extern unsigned int hfi1_max_srqs;
  355. extern unsigned int hfi1_max_srq_sges;
  356. extern unsigned int hfi1_max_srq_wrs;
  357. extern unsigned short piothreshold;
  358. extern const u32 ib_hfi1_rnr_table[];
  359. #endif /* HFI1_VERBS_H */