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
78 * Increment this value if any changes that break userspace ABI
79 * compatibility are made.
81 #define HFI1_UVERBS_ABI_VERSION 2
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
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)
95 #define HFI1_VENDOR_IPG cpu_to_be16(0xFFA0)
97 #define IB_DEFAULT_GID_PREFIX cpu_to_be64(0xfe80000000000000ULL)
98 #define OPA_BTH_MIG_REQ BIT(31)
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 #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))
113 struct hfi1_16b_header
{
118 struct ib_other_headers oth
;
120 struct ib_other_headers oth
;
124 struct hfi1_opa_header
{
126 struct ib_header ibh
; /* 9B header */
127 struct hfi1_16b_header opah
; /* 16B header */
129 u8 hdr_type
; /* 9B or 16B */
132 struct hfi1_ahg_info
{
139 struct hfi1_sdma_header
{
141 struct hfi1_opa_header hdr
;
145 * hfi1 specific data structures that will be hidden from rvt after the queue
146 * pair is made common
148 struct hfi1_qp_priv
{
149 struct hfi1_ahg_info
*s_ahg
; /* ahg info for next header */
150 struct sdma_engine
*s_sde
; /* current sde */
151 struct send_context
*s_sendcontext
; /* current sendcontext */
152 u8 s_sc
; /* SC[0..4] for next packet */
153 struct iowait s_iowait
;
154 struct rvt_qp
*owner
;
155 u8 hdr_type
; /* 9B or 16B */
159 * This structure is used to hold commonly lookedup and computed values during
160 * the send engine progress.
162 struct hfi1_pkt_state
{
163 struct hfi1_ibdev
*dev
;
164 struct hfi1_ibport
*ibp
;
165 struct hfi1_pportdata
*ppd
;
166 struct verbs_txreq
*s_txreq
;
168 unsigned long timeout
;
169 unsigned long timeout_int
;
176 #define HFI1_PSN_CREDIT 16
178 struct hfi1_opcode_stats
{
179 u64 n_packets
; /* number of packets */
180 u64 n_bytes
; /* total number of bytes */
183 struct hfi1_opcode_stats_perctx
{
184 struct hfi1_opcode_stats stats
[256];
187 static inline void inc_opstats(
189 struct hfi1_opcode_stats
*stats
)
191 #ifdef CONFIG_DEBUG_FS
192 stats
->n_bytes
+= tlen
;
198 struct rvt_qp __rcu
*qp
[2];
199 struct rvt_ibport rvp
;
201 /* the first 16 entries are sl_to_vl for !OPA */
207 struct rvt_dev_info rdi
; /* Must be first */
209 /* QP numbers are shared by all IB ports */
210 /* protect txwait list */
211 seqlock_t txwait_lock ____cacheline_aligned_in_smp
;
212 struct list_head txwait
; /* list for wait verbs_txreq */
213 struct list_head memwait
; /* list for wait kernel memory */
214 struct kmem_cache
*verbs_txreq_cache
;
218 /* protect iowait lists */
219 seqlock_t iowait_lock ____cacheline_aligned_in_smp
;
222 struct timer_list mem_timer
;
224 #ifdef CONFIG_DEBUG_FS
225 /* per HFI debugfs */
226 struct dentry
*hfi1_ibdev_dbg
;
227 /* per HFI symlinks to above */
228 struct dentry
*hfi1_ibdev_link
;
229 #ifdef CONFIG_FAULT_INJECTION
230 struct fault_opcode
*fault_opcode
;
231 struct fault_packet
*fault_packet
;
232 bool fault_suppress_err
;
237 static inline struct hfi1_ibdev
*to_idev(struct ib_device
*ibdev
)
239 struct rvt_dev_info
*rdi
;
241 rdi
= container_of(ibdev
, struct rvt_dev_info
, ibdev
);
242 return container_of(rdi
, struct hfi1_ibdev
, rdi
);
245 static inline struct rvt_qp
*iowait_to_qp(struct iowait
*s_iowait
)
247 struct hfi1_qp_priv
*priv
;
249 priv
= container_of(s_iowait
, struct hfi1_qp_priv
, s_iowait
);
254 * This must be called with s_lock held.
256 void hfi1_bad_pkey(struct hfi1_ibport
*ibp
, u32 key
, u32 sl
,
257 u32 qp1
, u32 qp2
, u32 lid1
, u32 lid2
);
258 void hfi1_cap_mask_chg(struct rvt_dev_info
*rdi
, u8 port_num
);
259 void hfi1_sys_guid_chg(struct hfi1_ibport
*ibp
);
260 void hfi1_node_desc_chg(struct hfi1_ibport
*ibp
);
261 int hfi1_process_mad(struct ib_device
*ibdev
, int mad_flags
, u8 port
,
262 const struct ib_wc
*in_wc
, const struct ib_grh
*in_grh
,
263 const struct ib_mad_hdr
*in_mad
, size_t in_mad_size
,
264 struct ib_mad_hdr
*out_mad
, size_t *out_mad_size
,
265 u16
*out_mad_pkey_index
);
268 * The PSN_MASK and PSN_SHIFT allow for
269 * 1) comparing two PSNs
270 * 2) returning the PSN with any upper bits masked
271 * 3) returning the difference between to PSNs
273 * The number of significant bits in the PSN must
274 * necessarily be at least one bit less than
275 * the container holding the PSN.
277 #define PSN_MASK 0x7FFFFFFF
279 #define PSN_MODIFY_MASK 0xFFFFFF
283 * Returns an integer <, ==, or > than zero.
285 static inline int cmp_psn(u32 a
, u32 b
)
287 return (((int)a
) - ((int)b
)) << PSN_SHIFT
;
293 static inline u32
mask_psn(u32 a
)
299 * Return delta between two PSNs
301 static inline u32
delta_psn(u32 a
, u32 b
)
303 return (((int)a
- (int)b
) << PSN_SHIFT
) >> PSN_SHIFT
;
307 void hfi1_put_txreq(struct verbs_txreq
*tx
);
309 int hfi1_verbs_send(struct rvt_qp
*qp
, struct hfi1_pkt_state
*ps
);
311 void hfi1_copy_sge(struct rvt_sge_state
*ss
, void *data
, u32 length
,
312 bool release
, bool copy_last
);
314 void hfi1_cnp_rcv(struct hfi1_packet
*packet
);
316 void hfi1_uc_rcv(struct hfi1_packet
*packet
);
318 void hfi1_rc_rcv(struct hfi1_packet
*packet
);
321 struct hfi1_ctxtdata
*rcd
,
322 struct hfi1_packet
*packet
,
325 u8
ah_to_sc(struct ib_device
*ibdev
, struct rdma_ah_attr
*ah_attr
);
327 void hfi1_rc_send_complete(struct rvt_qp
*qp
, struct hfi1_opa_header
*opah
);
329 void hfi1_ud_rcv(struct hfi1_packet
*packet
);
331 int hfi1_lookup_pkey_idx(struct hfi1_ibport
*ibp
, u16 pkey
);
333 int hfi1_rvt_get_rwqe(struct rvt_qp
*qp
, int wr_id_only
);
335 void hfi1_migrate_qp(struct rvt_qp
*qp
);
337 int hfi1_check_modify_qp(struct rvt_qp
*qp
, struct ib_qp_attr
*attr
,
338 int attr_mask
, struct ib_udata
*udata
);
340 void hfi1_modify_qp(struct rvt_qp
*qp
, struct ib_qp_attr
*attr
,
341 int attr_mask
, struct ib_udata
*udata
);
342 void hfi1_restart_rc(struct rvt_qp
*qp
, u32 psn
, int wait
);
343 int hfi1_check_send_wqe(struct rvt_qp
*qp
, struct rvt_swqe
*wqe
);
345 extern const u32 rc_only_opcode
;
346 extern const u32 uc_only_opcode
;
348 int hfi1_ruc_check_hdr(struct hfi1_ibport
*ibp
, struct hfi1_packet
*packet
);
350 u32
hfi1_make_grh(struct hfi1_ibport
*ibp
, struct ib_grh
*hdr
,
351 const struct ib_global_route
*grh
, u32 hwords
, u32 nwords
);
353 void hfi1_make_ruc_header(struct rvt_qp
*qp
, struct ib_other_headers
*ohdr
,
354 u32 bth0
, u32 bth2
, int middle
,
355 struct hfi1_pkt_state
*ps
);
357 void _hfi1_do_send(struct work_struct
*work
);
359 void hfi1_do_send_from_rvt(struct rvt_qp
*qp
);
361 void hfi1_do_send(struct rvt_qp
*qp
, bool in_thread
);
363 void hfi1_send_complete(struct rvt_qp
*qp
, struct rvt_swqe
*wqe
,
364 enum ib_wc_status status
);
366 void hfi1_send_rc_ack(struct hfi1_packet
*packet
, bool is_fecn
);
368 int hfi1_make_rc_req(struct rvt_qp
*qp
, struct hfi1_pkt_state
*ps
);
370 int hfi1_make_uc_req(struct rvt_qp
*qp
, struct hfi1_pkt_state
*ps
);
372 int hfi1_make_ud_req(struct rvt_qp
*qp
, struct hfi1_pkt_state
*ps
);
374 int hfi1_register_ib_device(struct hfi1_devdata
*);
376 void hfi1_unregister_ib_device(struct hfi1_devdata
*);
378 void hfi1_ib_rcv(struct hfi1_packet
*packet
);
380 void hfi1_16B_rcv(struct hfi1_packet
*packet
);
382 unsigned hfi1_get_npkeys(struct hfi1_devdata
*);
384 int hfi1_verbs_send_dma(struct rvt_qp
*qp
, struct hfi1_pkt_state
*ps
,
387 int hfi1_verbs_send_pio(struct rvt_qp
*qp
, struct hfi1_pkt_state
*ps
,
390 int hfi1_wss_init(void);
391 void hfi1_wss_exit(void);
393 /* platform specific: return the lowest level cache (llc) size, in KiB */
394 static inline int wss_llc_size(void)
396 /* assume that the boot CPU value is universal for all CPUs */
397 return boot_cpu_data
.x86_cache_size
;
400 /* platform specific: cacheless copy */
401 static inline void cacheless_memcpy(void *dst
, void *src
, size_t n
)
404 * Use the only available X64 cacheless copy. Add a __user cast
405 * to quiet sparse. The src agument is already in the kernel so
406 * there are no security issues. The extra fault recovery machinery
409 __copy_user_nocache(dst
, (void __user
*)src
, n
, 0);
412 static inline bool opa_bth_is_migration(struct ib_other_headers
*ohdr
)
414 return ohdr
->bth
[1] & cpu_to_be32(OPA_BTH_MIG_REQ
);
417 extern const enum ib_wc_opcode ib_hfi1_wc_opcode
[];
419 extern const u8 hdr_len_by_opcode
[];
421 extern const int ib_rvt_state_ops
[];
423 extern __be64 ib_hfi1_sys_image_guid
; /* in network order */
425 extern unsigned int hfi1_max_cqes
;
427 extern unsigned int hfi1_max_cqs
;
429 extern unsigned int hfi1_max_qp_wrs
;
431 extern unsigned int hfi1_max_qps
;
433 extern unsigned int hfi1_max_sges
;
435 extern unsigned int hfi1_max_mcast_grps
;
437 extern unsigned int hfi1_max_mcast_qp_attached
;
439 extern unsigned int hfi1_max_srqs
;
441 extern unsigned int hfi1_max_srq_sges
;
443 extern unsigned int hfi1_max_srq_wrs
;
445 extern unsigned short piothreshold
;
447 extern const u32 ib_hfi1_rnr_table
[];
449 #endif /* HFI1_VERBS_H */