blob: a4d06502f06df61d34611ee09bc888af80fcc04c [file] [log] [blame]
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001/*
2 * Copyright(c) 2015 - 2018 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
48#ifndef HFI1_VERBS_H
49#define HFI1_VERBS_H
50
51#include <linux/types.h>
52#include <linux/seqlock.h>
53#include <linux/kernel.h>
54#include <linux/interrupt.h>
55#include <linux/kref.h>
56#include <linux/workqueue.h>
57#include <linux/kthread.h>
58#include <linux/completion.h>
59#include <linux/slab.h>
60#include <rdma/ib_pack.h>
61#include <rdma/ib_user_verbs.h>
62#include <rdma/ib_mad.h>
63#include <rdma/ib_hdrs.h>
64#include <rdma/rdma_vt.h>
65#include <rdma/rdmavt_qp.h>
66#include <rdma/rdmavt_cq.h>
67
68struct hfi1_ctxtdata;
69struct hfi1_pportdata;
70struct hfi1_devdata;
71struct hfi1_packet;
72
73#include "iowait.h"
74
75#define HFI1_MAX_RDMA_ATOMIC 16
76
77/*
78 * Increment this value if any changes that break userspace ABI
79 * compatibility are made.
80 */
81#define HFI1_UVERBS_ABI_VERSION 2
82
83/* IB Performance Manager status values */
84#define IB_PMA_SAMPLE_STATUS_DONE 0x00
85#define IB_PMA_SAMPLE_STATUS_STARTED 0x01
86#define IB_PMA_SAMPLE_STATUS_RUNNING 0x02
87
88/* Mandatory IB performance counter select values. */
89#define IB_PMA_PORT_XMIT_DATA cpu_to_be16(0x0001)
90#define IB_PMA_PORT_RCV_DATA cpu_to_be16(0x0002)
91#define IB_PMA_PORT_XMIT_PKTS cpu_to_be16(0x0003)
92#define IB_PMA_PORT_RCV_PKTS cpu_to_be16(0x0004)
93#define IB_PMA_PORT_XMIT_WAIT cpu_to_be16(0x0005)
94
95#define HFI1_VENDOR_IPG cpu_to_be16(0xFFA0)
96
97#define IB_DEFAULT_GID_PREFIX cpu_to_be64(0xfe80000000000000ULL)
98#define OPA_BTH_MIG_REQ BIT(31)
99
100#define RC_OP(x) IB_OPCODE_RC_##x
101#define UC_OP(x) IB_OPCODE_UC_##x
102
103/* flags passed by hfi1_ib_rcv() */
104enum {
105 HFI1_HAS_GRH = (1 << 0),
106};
107
108#define LRH_16B_BYTES (FIELD_SIZEOF(struct hfi1_16b_header, lrh))
109#define LRH_16B_DWORDS (LRH_16B_BYTES / sizeof(u32))
110#define LRH_9B_BYTES (FIELD_SIZEOF(struct ib_header, lrh))
111#define LRH_9B_DWORDS (LRH_9B_BYTES / sizeof(u32))
112
113/* 24Bits for qpn, upper 8Bits reserved */
114struct opa_16b_mgmt {
115 __be32 dest_qpn;
116 __be32 src_qpn;
117};
118
119struct hfi1_16b_header {
120 u32 lrh[4];
121 union {
122 struct {
123 struct ib_grh grh;
124 struct ib_other_headers oth;
125 } l;
126 struct ib_other_headers oth;
127 struct opa_16b_mgmt mgmt;
128 } u;
129} __packed;
130
131struct hfi1_opa_header {
132 union {
133 struct ib_header ibh; /* 9B header */
134 struct hfi1_16b_header opah; /* 16B header */
135 };
136 u8 hdr_type; /* 9B or 16B */
137} __packed;
138
139struct hfi1_ahg_info {
140 u32 ahgdesc[2];
141 u16 tx_flags;
142 u8 ahgcount;
143 u8 ahgidx;
144};
145
146struct hfi1_sdma_header {
147 __le64 pbc;
148 struct hfi1_opa_header hdr;
149} __packed;
150
151/*
152 * hfi1 specific data structures that will be hidden from rvt after the queue
153 * pair is made common
154 */
155struct hfi1_qp_priv {
156 struct hfi1_ahg_info *s_ahg; /* ahg info for next header */
157 struct sdma_engine *s_sde; /* current sde */
158 struct send_context *s_sendcontext; /* current sendcontext */
159 u8 s_sc; /* SC[0..4] for next packet */
160 struct iowait s_iowait;
161 struct rvt_qp *owner;
162 u8 hdr_type; /* 9B or 16B */
163};
164
165/*
166 * This structure is used to hold commonly lookedup and computed values during
167 * the send engine progress.
168 */
169struct hfi1_pkt_state {
170 struct hfi1_ibdev *dev;
171 struct hfi1_ibport *ibp;
172 struct hfi1_pportdata *ppd;
173 struct verbs_txreq *s_txreq;
174 unsigned long flags;
175 unsigned long timeout;
176 unsigned long timeout_int;
177 int cpu;
178 u8 opcode;
179 bool in_thread;
180 bool pkts_sent;
181};
182
183#define HFI1_PSN_CREDIT 16
184
185struct hfi1_opcode_stats {
186 u64 n_packets; /* number of packets */
187 u64 n_bytes; /* total number of bytes */
188};
189
190struct hfi1_opcode_stats_perctx {
191 struct hfi1_opcode_stats stats[256];
192};
193
194static inline void inc_opstats(
195 u32 tlen,
196 struct hfi1_opcode_stats *stats)
197{
198#ifdef CONFIG_DEBUG_FS
199 stats->n_bytes += tlen;
200 stats->n_packets++;
201#endif
202}
203
204struct hfi1_ibport {
205 struct rvt_qp __rcu *qp[2];
206 struct rvt_ibport rvp;
207
208 /* the first 16 entries are sl_to_vl for !OPA */
209 u8 sl_to_sc[32];
210 u8 sc_to_sl[32];
211};
212
213struct hfi1_ibdev {
214 struct rvt_dev_info rdi; /* Must be first */
215
216 /* QP numbers are shared by all IB ports */
217 /* protect txwait list */
218 seqlock_t txwait_lock ____cacheline_aligned_in_smp;
219 struct list_head txwait; /* list for wait verbs_txreq */
220 struct list_head memwait; /* list for wait kernel memory */
221 struct kmem_cache *verbs_txreq_cache;
222 u64 n_txwait;
223 u64 n_kmem_wait;
224
225 /* protect iowait lists */
226 seqlock_t iowait_lock ____cacheline_aligned_in_smp;
227 u64 n_piowait;
228 u64 n_piodrain;
229 struct timer_list mem_timer;
230
231#ifdef CONFIG_DEBUG_FS
232 /* per HFI debugfs */
233 struct dentry *hfi1_ibdev_dbg;
234 /* per HFI symlinks to above */
235 struct dentry *hfi1_ibdev_link;
236#ifdef CONFIG_FAULT_INJECTION
237 struct fault *fault;
238#endif
239#endif
240};
241
242static inline struct hfi1_ibdev *to_idev(struct ib_device *ibdev)
243{
244 struct rvt_dev_info *rdi;
245
246 rdi = container_of(ibdev, struct rvt_dev_info, ibdev);
247 return container_of(rdi, struct hfi1_ibdev, rdi);
248}
249
250static inline struct rvt_qp *iowait_to_qp(struct iowait *s_iowait)
251{
252 struct hfi1_qp_priv *priv;
253
254 priv = container_of(s_iowait, struct hfi1_qp_priv, s_iowait);
255 return priv->owner;
256}
257
258/*
259 * This must be called with s_lock held.
260 */
261void hfi1_bad_pkey(struct hfi1_ibport *ibp, u32 key, u32 sl,
262 u32 qp1, u32 qp2, u32 lid1, u32 lid2);
263void hfi1_cap_mask_chg(struct rvt_dev_info *rdi, u8 port_num);
264void hfi1_sys_guid_chg(struct hfi1_ibport *ibp);
265void hfi1_node_desc_chg(struct hfi1_ibport *ibp);
266int hfi1_process_mad(struct ib_device *ibdev, int mad_flags, u8 port,
267 const struct ib_wc *in_wc, const struct ib_grh *in_grh,
268 const struct ib_mad_hdr *in_mad, size_t in_mad_size,
269 struct ib_mad_hdr *out_mad, size_t *out_mad_size,
270 u16 *out_mad_pkey_index);
271
272/*
273 * The PSN_MASK and PSN_SHIFT allow for
274 * 1) comparing two PSNs
275 * 2) returning the PSN with any upper bits masked
276 * 3) returning the difference between to PSNs
277 *
278 * The number of significant bits in the PSN must
279 * necessarily be at least one bit less than
280 * the container holding the PSN.
281 */
282#define PSN_MASK 0x7FFFFFFF
283#define PSN_SHIFT 1
284#define PSN_MODIFY_MASK 0xFFFFFF
285
286/*
287 * Compare two PSNs
288 * Returns an integer <, ==, or > than zero.
289 */
290static inline int cmp_psn(u32 a, u32 b)
291{
292 return (((int)a) - ((int)b)) << PSN_SHIFT;
293}
294
295/*
296 * Return masked PSN
297 */
298static inline u32 mask_psn(u32 a)
299{
300 return a & PSN_MASK;
301}
302
303/*
304 * Return delta between two PSNs
305 */
306static inline u32 delta_psn(u32 a, u32 b)
307{
308 return (((int)a - (int)b) << PSN_SHIFT) >> PSN_SHIFT;
309}
310
311struct verbs_txreq;
312void hfi1_put_txreq(struct verbs_txreq *tx);
313
314int hfi1_verbs_send(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
315
316void hfi1_copy_sge(struct rvt_sge_state *ss, void *data, u32 length,
317 bool release, bool copy_last);
318
319void hfi1_cnp_rcv(struct hfi1_packet *packet);
320
321void hfi1_uc_rcv(struct hfi1_packet *packet);
322
323void hfi1_rc_rcv(struct hfi1_packet *packet);
324
325void hfi1_rc_hdrerr(
326 struct hfi1_ctxtdata *rcd,
327 struct hfi1_packet *packet,
328 struct rvt_qp *qp);
329
330u8 ah_to_sc(struct ib_device *ibdev, struct rdma_ah_attr *ah_attr);
331
332void hfi1_rc_send_complete(struct rvt_qp *qp, struct hfi1_opa_header *opah);
333
334void hfi1_ud_rcv(struct hfi1_packet *packet);
335
336int hfi1_lookup_pkey_idx(struct hfi1_ibport *ibp, u16 pkey);
337
338void hfi1_migrate_qp(struct rvt_qp *qp);
339
340int hfi1_check_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
341 int attr_mask, struct ib_udata *udata);
342
343void hfi1_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
344 int attr_mask, struct ib_udata *udata);
345void hfi1_restart_rc(struct rvt_qp *qp, u32 psn, int wait);
346int hfi1_check_send_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe);
347
348extern const u32 rc_only_opcode;
349extern const u32 uc_only_opcode;
350
351int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct hfi1_packet *packet);
352
353u32 hfi1_make_grh(struct hfi1_ibport *ibp, struct ib_grh *hdr,
354 const struct ib_global_route *grh, u32 hwords, u32 nwords);
355
356void hfi1_make_ruc_header(struct rvt_qp *qp, struct ib_other_headers *ohdr,
357 u32 bth0, u32 bth2, int middle,
358 struct hfi1_pkt_state *ps);
359
360void _hfi1_do_send(struct work_struct *work);
361
362void hfi1_do_send_from_rvt(struct rvt_qp *qp);
363
364void hfi1_do_send(struct rvt_qp *qp, bool in_thread);
365
366void hfi1_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
367 enum ib_wc_status status);
368
369void hfi1_send_rc_ack(struct hfi1_packet *packet, bool is_fecn);
370
371int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
372
373int hfi1_make_uc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
374
375int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
376
377int hfi1_register_ib_device(struct hfi1_devdata *);
378
379void hfi1_unregister_ib_device(struct hfi1_devdata *);
380
381void hfi1_ib_rcv(struct hfi1_packet *packet);
382
383void hfi1_16B_rcv(struct hfi1_packet *packet);
384
385unsigned hfi1_get_npkeys(struct hfi1_devdata *);
386
387int hfi1_verbs_send_dma(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
388 u64 pbc);
389
390int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
391 u64 pbc);
392
393int hfi1_wss_init(void);
394void hfi1_wss_exit(void);
395
396/* platform specific: return the lowest level cache (llc) size, in KiB */
397static inline int wss_llc_size(void)
398{
399 /* assume that the boot CPU value is universal for all CPUs */
400 return boot_cpu_data.x86_cache_size;
401}
402
403/* platform specific: cacheless copy */
404static inline void cacheless_memcpy(void *dst, void *src, size_t n)
405{
406 /*
407 * Use the only available X64 cacheless copy. Add a __user cast
408 * to quiet sparse. The src agument is already in the kernel so
409 * there are no security issues. The extra fault recovery machinery
410 * is not invoked.
411 */
412 __copy_user_nocache(dst, (void __user *)src, n, 0);
413}
414
415static inline bool opa_bth_is_migration(struct ib_other_headers *ohdr)
416{
417 return ohdr->bth[1] & cpu_to_be32(OPA_BTH_MIG_REQ);
418}
419
420extern const enum ib_wc_opcode ib_hfi1_wc_opcode[];
421
422extern const u8 hdr_len_by_opcode[];
423
424extern const int ib_rvt_state_ops[];
425
426extern __be64 ib_hfi1_sys_image_guid; /* in network order */
427
428extern unsigned int hfi1_max_cqes;
429
430extern unsigned int hfi1_max_cqs;
431
432extern unsigned int hfi1_max_qp_wrs;
433
434extern unsigned int hfi1_max_qps;
435
436extern unsigned int hfi1_max_sges;
437
438extern unsigned int hfi1_max_mcast_grps;
439
440extern unsigned int hfi1_max_mcast_qp_attached;
441
442extern unsigned int hfi1_max_srqs;
443
444extern unsigned int hfi1_max_srq_sges;
445
446extern unsigned int hfi1_max_srq_wrs;
447
448extern unsigned short piothreshold;
449
450extern const u32 ib_hfi1_rnr_table[];
451
452#endif /* HFI1_VERBS_H */