2 * Copyright(c) 2015 - 2017 Intel Corporation.
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.
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.
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.
20 * Redistribution and use in source and binary forms, with or without
21 * modification, are permitted provided that the following conditions
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
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.
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.
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>
69 struct hfi1_pportdata;
75 #define HFI1_MAX_RDMA_ATOMIC 16
76 #define HFI1_GUIDS_PER_PORT 5
79 * Increment this value if any changes that break userspace ABI
80 * compatibility are made.
82 #define HFI1_UVERBS_ABI_VERSION 2
84 /* IB Performance Manager status values */
85 #define IB_PMA_SAMPLE_STATUS_DONE 0x00
86 #define IB_PMA_SAMPLE_STATUS_STARTED 0x01
87 #define IB_PMA_SAMPLE_STATUS_RUNNING 0x02
89 /* Mandatory IB performance counter select values. */
90 #define IB_PMA_PORT_XMIT_DATA cpu_to_be16(0x0001)
91 #define IB_PMA_PORT_RCV_DATA cpu_to_be16(0x0002)
92 #define IB_PMA_PORT_XMIT_PKTS cpu_to_be16(0x0003)
93 #define IB_PMA_PORT_RCV_PKTS cpu_to_be16(0x0004)
94 #define IB_PMA_PORT_XMIT_WAIT cpu_to_be16(0x0005)
96 #define HFI1_VENDOR_IPG cpu_to_be16(0xFFA0)
98 #define IB_DEFAULT_GID_PREFIX cpu_to_be64(0xfe80000000000000ULL)
100 #define RC_OP(x) IB_OPCODE_RC_##x
101 #define UC_OP(x) IB_OPCODE_UC_##x
103 /* flags passed by hfi1_ib_rcv() */
105 HFI1_HAS_GRH = (1 << 0),
108 struct hfi1_ahg_info {
115 struct hfi1_sdma_header {
117 struct ib_header hdr;
121 * hfi1 specific data structures that will be hidden from rvt after the queue
122 * pair is made common
124 struct hfi1_qp_priv {
125 struct hfi1_ahg_info *s_ahg; /* ahg info for next header */
126 struct sdma_engine *s_sde; /* current sde */
127 struct send_context *s_sendcontext; /* current sendcontext */
128 u8 s_sc; /* SC[0..4] for next packet */
129 u8 r_adefered; /* number of acks defered */
130 struct iowait s_iowait;
131 struct timer_list s_rnr_timer;
132 struct rvt_qp *owner;
136 * This structure is used to hold commonly lookedup and computed values during
137 * the send engine progress.
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;
147 #define HFI1_PSN_CREDIT 16
149 struct hfi1_opcode_stats {
150 u64 n_packets; /* number of packets */
151 u64 n_bytes; /* total number of bytes */
154 struct hfi1_opcode_stats_perctx {
155 struct hfi1_opcode_stats stats[256];
158 static inline void inc_opstats(
160 struct hfi1_opcode_stats *stats)
162 #ifdef CONFIG_DEBUG_FS
163 stats->n_bytes += tlen;
169 struct rvt_qp __rcu *qp[2];
170 struct rvt_ibport rvp;
172 __be64 guids[HFI1_GUIDS_PER_PORT - 1]; /* writable GUIDs */
174 /* the first 16 entries are sl_to_vl for !OPA */
180 struct rvt_dev_info rdi; /* Must be first */
182 /* QP numbers are shared by all IB ports */
183 /* protect wait lists */
184 seqlock_t iowait_lock;
185 struct list_head txwait; /* list for wait verbs_txreq */
186 struct list_head memwait; /* list for wait kernel memory */
187 struct list_head txreq_free;
188 struct kmem_cache *verbs_txreq_cache;
189 struct timer_list mem_timer;
196 #ifdef CONFIG_DEBUG_FS
197 /* per HFI debugfs */
198 struct dentry *hfi1_ibdev_dbg;
199 /* per HFI symlinks to above */
200 struct dentry *hfi1_ibdev_link;
204 static inline struct hfi1_ibdev *to_idev(struct ib_device *ibdev)
206 struct rvt_dev_info *rdi;
208 rdi = container_of(ibdev, struct rvt_dev_info, ibdev);
209 return container_of(rdi, struct hfi1_ibdev, rdi);
212 static inline struct rvt_qp *iowait_to_qp(struct iowait *s_iowait)
214 struct hfi1_qp_priv *priv;
216 priv = container_of(s_iowait, struct hfi1_qp_priv, s_iowait);
221 * Send if not busy or waiting for I/O and either
222 * a RC response is pending or we can process send work requests.
224 static inline int hfi1_send_ok(struct rvt_qp *qp)
226 return !(qp->s_flags & (RVT_S_BUSY | RVT_S_ANY_WAIT_IO)) &&
227 (qp->s_hdrwords || (qp->s_flags & RVT_S_RESP_PENDING) ||
228 !(qp->s_flags & RVT_S_ANY_WAIT_SEND));
232 * This must be called with s_lock held.
234 void hfi1_bad_pqkey(struct hfi1_ibport *ibp, __be16 trap_num, u32 key, u32 sl,
235 u32 qp1, u32 qp2, u16 lid1, u16 lid2);
236 void hfi1_cap_mask_chg(struct rvt_dev_info *rdi, u8 port_num);
237 void hfi1_sys_guid_chg(struct hfi1_ibport *ibp);
238 void hfi1_node_desc_chg(struct hfi1_ibport *ibp);
239 int hfi1_process_mad(struct ib_device *ibdev, int mad_flags, u8 port,
240 const struct ib_wc *in_wc, const struct ib_grh *in_grh,
241 const struct ib_mad_hdr *in_mad, size_t in_mad_size,
242 struct ib_mad_hdr *out_mad, size_t *out_mad_size,
243 u16 *out_mad_pkey_index);
246 * The PSN_MASK and PSN_SHIFT allow for
247 * 1) comparing two PSNs
248 * 2) returning the PSN with any upper bits masked
249 * 3) returning the difference between to PSNs
251 * The number of significant bits in the PSN must
252 * necessarily be at least one bit less than
253 * the container holding the PSN.
255 #ifndef CONFIG_HFI1_VERBS_31BIT_PSN
256 #define PSN_MASK 0xFFFFFF
259 #define PSN_MASK 0x7FFFFFFF
262 #define PSN_MODIFY_MASK 0xFFFFFF
265 * Compare the lower 24 bits of the msn values.
266 * Returns an integer <, ==, or > than zero.
268 static inline int cmp_msn(u32 a, u32 b)
270 return (((int)a) - ((int)b)) << 8;
275 * Returns an integer <, ==, or > than zero.
277 static inline int cmp_psn(u32 a, u32 b)
279 return (((int)a) - ((int)b)) << PSN_SHIFT;
285 static inline u32 mask_psn(u32 a)
291 * Return delta between two PSNs
293 static inline u32 delta_psn(u32 a, u32 b)
295 return (((int)a - (int)b) << PSN_SHIFT) >> PSN_SHIFT;
299 void hfi1_put_txreq(struct verbs_txreq *tx);
301 int hfi1_verbs_send(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
303 void hfi1_copy_sge(struct rvt_sge_state *ss, void *data, u32 length,
304 int release, int copy_last);
306 void hfi1_skip_sge(struct rvt_sge_state *ss, u32 length, int release);
308 void hfi1_cnp_rcv(struct hfi1_packet *packet);
310 void hfi1_uc_rcv(struct hfi1_packet *packet);
312 void hfi1_rc_rcv(struct hfi1_packet *packet);
315 struct hfi1_ctxtdata *rcd,
316 struct ib_header *hdr,
320 u8 ah_to_sc(struct ib_device *ibdev, struct ib_ah_attr *ah_attr);
322 struct ib_ah *hfi1_create_qp0_ah(struct hfi1_ibport *ibp, u16 dlid);
324 void hfi1_rc_rnr_retry(unsigned long arg);
325 void hfi1_add_rnr_timer(struct rvt_qp *qp, u32 to);
326 void hfi1_rc_timeout(unsigned long arg);
327 void hfi1_del_timers_sync(struct rvt_qp *qp);
328 void hfi1_stop_rc_timers(struct rvt_qp *qp);
330 void hfi1_rc_send_complete(struct rvt_qp *qp, struct ib_header *hdr);
332 void hfi1_rc_error(struct rvt_qp *qp, enum ib_wc_status err);
334 void hfi1_ud_rcv(struct hfi1_packet *packet);
336 int hfi1_lookup_pkey_idx(struct hfi1_ibport *ibp, u16 pkey);
338 int hfi1_rvt_get_rwqe(struct rvt_qp *qp, int wr_id_only);
340 void hfi1_migrate_qp(struct rvt_qp *qp);
342 int hfi1_check_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
343 int attr_mask, struct ib_udata *udata);
345 void hfi1_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
346 int attr_mask, struct ib_udata *udata);
348 int hfi1_check_send_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe);
350 extern const u32 rc_only_opcode;
351 extern const u32 uc_only_opcode;
353 static inline u8 get_opcode(struct ib_header *h)
355 u16 lnh = be16_to_cpu(h->lrh[0]) & 3;
357 if (lnh == IB_LNH_IBA_LOCAL)
358 return be32_to_cpu(h->u.oth.bth[0]) >> 24;
360 return be32_to_cpu(h->u.l.oth.bth[0]) >> 24;
363 int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct ib_header *hdr,
364 int has_grh, struct rvt_qp *qp, u32 bth0);
366 u32 hfi1_make_grh(struct hfi1_ibport *ibp, struct ib_grh *hdr,
367 struct ib_global_route *grh, u32 hwords, u32 nwords);
369 void hfi1_make_ruc_header(struct rvt_qp *qp, struct ib_other_headers *ohdr,
370 u32 bth0, u32 bth2, int middle,
371 struct hfi1_pkt_state *ps);
373 void _hfi1_do_send(struct work_struct *work);
375 void hfi1_do_send_from_rvt(struct rvt_qp *qp);
377 void hfi1_do_send(struct rvt_qp *qp, bool in_thread);
379 void hfi1_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
380 enum ib_wc_status status);
382 void hfi1_send_rc_ack(struct hfi1_ctxtdata *, struct rvt_qp *qp, int is_fecn);
384 int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
386 int hfi1_make_uc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
388 int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
390 int hfi1_register_ib_device(struct hfi1_devdata *);
392 void hfi1_unregister_ib_device(struct hfi1_devdata *);
394 void hfi1_ib_rcv(struct hfi1_packet *packet);
396 unsigned hfi1_get_npkeys(struct hfi1_devdata *);
398 int hfi1_verbs_send_dma(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
401 int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
404 int hfi1_wss_init(void);
405 void hfi1_wss_exit(void);
407 /* platform specific: return the lowest level cache (llc) size, in KiB */
408 static inline int wss_llc_size(void)
410 /* assume that the boot CPU value is universal for all CPUs */
411 return boot_cpu_data.x86_cache_size;
414 /* platform specific: cacheless copy */
415 static inline void cacheless_memcpy(void *dst, void *src, size_t n)
418 * Use the only available X64 cacheless copy. Add a __user cast
419 * to quiet sparse. The src agument is already in the kernel so
420 * there are no security issues. The extra fault recovery machinery
423 __copy_user_nocache(dst, (void __user *)src, n, 0);
426 extern const enum ib_wc_opcode ib_hfi1_wc_opcode[];
428 extern const u8 hdr_len_by_opcode[];
430 extern const int ib_rvt_state_ops[];
432 extern __be64 ib_hfi1_sys_image_guid; /* in network order */
434 extern unsigned int hfi1_max_cqes;
436 extern unsigned int hfi1_max_cqs;
438 extern unsigned int hfi1_max_qp_wrs;
440 extern unsigned int hfi1_max_qps;
442 extern unsigned int hfi1_max_sges;
444 extern unsigned int hfi1_max_mcast_grps;
446 extern unsigned int hfi1_max_mcast_qp_attached;
448 extern unsigned int hfi1_max_srqs;
450 extern unsigned int hfi1_max_srq_sges;
452 extern unsigned int hfi1_max_srq_wrs;
454 extern unsigned short piothreshold;
456 extern const u32 ib_hfi1_rnr_table[];
458 #endif /* HFI1_VERBS_H */