Line data Source code
1 : /* SPDX-License-Identifier: BSD-3-Clause
2 : * Copyright (C) 2016 Intel Corporation. All rights reserved.
3 : * Copyright (c) 2019-2021 Mellanox Technologies LTD. All rights reserved.
4 : * Copyright (c) 2021-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
5 : */
6 :
7 : #include "spdk/stdinc.h"
8 :
9 : #include "spdk/config.h"
10 : #include "spdk/thread.h"
11 : #include "spdk/likely.h"
12 : #include "spdk/nvmf_transport.h"
13 : #include "spdk/string.h"
14 : #include "spdk/trace.h"
15 : #include "spdk/tree.h"
16 : #include "spdk/util.h"
17 :
18 : #include "spdk_internal/assert.h"
19 : #include "spdk/log.h"
20 : #include "spdk_internal/rdma_provider.h"
21 : #include "spdk_internal/rdma_utils.h"
22 :
23 : #include "nvmf_internal.h"
24 : #include "transport.h"
25 :
26 : #include "spdk_internal/trace_defs.h"
27 :
28 : struct spdk_nvme_rdma_hooks g_nvmf_hooks = {};
29 : const struct spdk_nvmf_transport_ops spdk_nvmf_transport_rdma;
30 :
31 : /*
32 : RDMA Connection Resource Defaults
33 : */
34 : #define NVMF_DEFAULT_MSDBD 16
35 : #define NVMF_DEFAULT_TX_SGE SPDK_NVMF_MAX_SGL_ENTRIES
36 : #define NVMF_DEFAULT_RSP_SGE 1
37 : #define NVMF_DEFAULT_RX_SGE 2
38 :
39 : #define NVMF_RDMA_MAX_EVENTS_PER_POLL 32
40 :
41 : SPDK_STATIC_ASSERT(NVMF_DEFAULT_MSDBD <= SPDK_NVMF_MAX_SGL_ENTRIES,
42 : "MSDBD must not exceed SPDK_NVMF_MAX_SGL_ENTRIES");
43 :
44 : /* The RDMA completion queue size */
45 : #define DEFAULT_NVMF_RDMA_CQ_SIZE 4096
46 : #define MAX_WR_PER_QP(queue_depth) (queue_depth * 3 + 2)
47 :
48 : enum spdk_nvmf_rdma_request_state {
49 : /* The request is not currently in use */
50 : RDMA_REQUEST_STATE_FREE = 0,
51 :
52 : /* Initial state when request first received */
53 : RDMA_REQUEST_STATE_NEW,
54 :
55 : /* The request is queued until a data buffer is available. */
56 : RDMA_REQUEST_STATE_NEED_BUFFER,
57 :
58 : /* The request is waiting on RDMA queue depth availability
59 : * to transfer data from the host to the controller.
60 : */
61 : RDMA_REQUEST_STATE_DATA_TRANSFER_TO_CONTROLLER_PENDING,
62 :
63 : /* The request is currently transferring data from the host to the controller. */
64 : RDMA_REQUEST_STATE_TRANSFERRING_HOST_TO_CONTROLLER,
65 :
66 : /* The request is ready to execute at the block device */
67 : RDMA_REQUEST_STATE_READY_TO_EXECUTE,
68 :
69 : /* The request is currently executing at the block device */
70 : RDMA_REQUEST_STATE_EXECUTING,
71 :
72 : /* The request finished executing at the block device */
73 : RDMA_REQUEST_STATE_EXECUTED,
74 :
75 : /* The request is waiting on RDMA queue depth availability
76 : * to transfer data from the controller to the host.
77 : */
78 : RDMA_REQUEST_STATE_DATA_TRANSFER_TO_HOST_PENDING,
79 :
80 : /* The request is waiting on RDMA queue depth availability
81 : * to send response to the host.
82 : */
83 : RDMA_REQUEST_STATE_READY_TO_COMPLETE_PENDING,
84 :
85 : /* The request is ready to send a completion */
86 : RDMA_REQUEST_STATE_READY_TO_COMPLETE,
87 :
88 : /* The request is currently transferring data from the controller to the host. */
89 : RDMA_REQUEST_STATE_TRANSFERRING_CONTROLLER_TO_HOST,
90 :
91 : /* The request currently has an outstanding completion without an
92 : * associated data transfer.
93 : */
94 : RDMA_REQUEST_STATE_COMPLETING,
95 :
96 : /* The request completed and can be marked free. */
97 : RDMA_REQUEST_STATE_COMPLETED,
98 :
99 : /* Terminator */
100 : RDMA_REQUEST_NUM_STATES,
101 : };
102 :
103 : static void
104 0 : nvmf_trace(void)
105 : {
106 0 : spdk_trace_register_object(OBJECT_NVMF_RDMA_IO, 'r');
107 :
108 0 : struct spdk_trace_tpoint_opts opts[] = {
109 : {
110 : "RDMA_REQ_NEW", TRACE_RDMA_REQUEST_STATE_NEW,
111 : OWNER_TYPE_NONE, OBJECT_NVMF_RDMA_IO, 1,
112 : {
113 : { "qpair", SPDK_TRACE_ARG_TYPE_PTR, 8 },
114 : { "qd", SPDK_TRACE_ARG_TYPE_INT, 4 }
115 : }
116 : },
117 : {
118 : "RDMA_REQ_COMPLETED", TRACE_RDMA_REQUEST_STATE_COMPLETED,
119 : OWNER_TYPE_NONE, OBJECT_NVMF_RDMA_IO, 0,
120 : {
121 : { "qpair", SPDK_TRACE_ARG_TYPE_PTR, 8 },
122 : { "qd", SPDK_TRACE_ARG_TYPE_INT, 4 }
123 : }
124 : },
125 : };
126 :
127 0 : spdk_trace_register_description_ext(opts, SPDK_COUNTOF(opts));
128 0 : spdk_trace_register_description("RDMA_REQ_NEED_BUFFER", TRACE_RDMA_REQUEST_STATE_NEED_BUFFER,
129 : OWNER_TYPE_NONE, OBJECT_NVMF_RDMA_IO, 0,
130 : SPDK_TRACE_ARG_TYPE_PTR, "qpair");
131 0 : spdk_trace_register_description("RDMA_REQ_TX_PENDING_C2H",
132 : TRACE_RDMA_REQUEST_STATE_DATA_TRANSFER_TO_HOST_PENDING,
133 : OWNER_TYPE_NONE, OBJECT_NVMF_RDMA_IO, 0,
134 : SPDK_TRACE_ARG_TYPE_PTR, "qpair");
135 0 : spdk_trace_register_description("RDMA_REQ_TX_PENDING_H2C",
136 : TRACE_RDMA_REQUEST_STATE_DATA_TRANSFER_TO_CONTROLLER_PENDING,
137 : OWNER_TYPE_NONE, OBJECT_NVMF_RDMA_IO, 0,
138 : SPDK_TRACE_ARG_TYPE_PTR, "qpair");
139 0 : spdk_trace_register_description("RDMA_REQ_TX_H2C",
140 : TRACE_RDMA_REQUEST_STATE_TRANSFERRING_HOST_TO_CONTROLLER,
141 : OWNER_TYPE_NONE, OBJECT_NVMF_RDMA_IO, 0,
142 : SPDK_TRACE_ARG_TYPE_PTR, "qpair");
143 0 : spdk_trace_register_description("RDMA_REQ_RDY_TO_EXECUTE",
144 : TRACE_RDMA_REQUEST_STATE_READY_TO_EXECUTE,
145 : OWNER_TYPE_NONE, OBJECT_NVMF_RDMA_IO, 0,
146 : SPDK_TRACE_ARG_TYPE_PTR, "qpair");
147 0 : spdk_trace_register_description("RDMA_REQ_EXECUTING",
148 : TRACE_RDMA_REQUEST_STATE_EXECUTING,
149 : OWNER_TYPE_NONE, OBJECT_NVMF_RDMA_IO, 0,
150 : SPDK_TRACE_ARG_TYPE_PTR, "qpair");
151 0 : spdk_trace_register_description("RDMA_REQ_EXECUTED",
152 : TRACE_RDMA_REQUEST_STATE_EXECUTED,
153 : OWNER_TYPE_NONE, OBJECT_NVMF_RDMA_IO, 0,
154 : SPDK_TRACE_ARG_TYPE_PTR, "qpair");
155 0 : spdk_trace_register_description("RDMA_REQ_RDY2COMPL_PEND",
156 : TRACE_RDMA_REQUEST_STATE_READY_TO_COMPLETE_PENDING,
157 : OWNER_TYPE_NONE, OBJECT_NVMF_RDMA_IO, 0,
158 : SPDK_TRACE_ARG_TYPE_PTR, "qpair");
159 0 : spdk_trace_register_description("RDMA_REQ_RDY_TO_COMPL",
160 : TRACE_RDMA_REQUEST_STATE_READY_TO_COMPLETE,
161 : OWNER_TYPE_NONE, OBJECT_NVMF_RDMA_IO, 0,
162 : SPDK_TRACE_ARG_TYPE_PTR, "qpair");
163 0 : spdk_trace_register_description("RDMA_REQ_COMPLETING_C2H",
164 : TRACE_RDMA_REQUEST_STATE_TRANSFERRING_CONTROLLER_TO_HOST,
165 : OWNER_TYPE_NONE, OBJECT_NVMF_RDMA_IO, 0,
166 : SPDK_TRACE_ARG_TYPE_PTR, "qpair");
167 0 : spdk_trace_register_description("RDMA_REQ_COMPLETING",
168 : TRACE_RDMA_REQUEST_STATE_COMPLETING,
169 : OWNER_TYPE_NONE, OBJECT_NVMF_RDMA_IO, 0,
170 : SPDK_TRACE_ARG_TYPE_PTR, "qpair");
171 :
172 0 : spdk_trace_register_description("RDMA_QP_CREATE", TRACE_RDMA_QP_CREATE,
173 : OWNER_TYPE_NONE, OBJECT_NONE, 0,
174 : SPDK_TRACE_ARG_TYPE_INT, "");
175 0 : spdk_trace_register_description("RDMA_IBV_ASYNC_EVENT", TRACE_RDMA_IBV_ASYNC_EVENT,
176 : OWNER_TYPE_NONE, OBJECT_NONE, 0,
177 : SPDK_TRACE_ARG_TYPE_INT, "type");
178 0 : spdk_trace_register_description("RDMA_CM_ASYNC_EVENT", TRACE_RDMA_CM_ASYNC_EVENT,
179 : OWNER_TYPE_NONE, OBJECT_NONE, 0,
180 : SPDK_TRACE_ARG_TYPE_INT, "type");
181 0 : spdk_trace_register_description("RDMA_QP_DISCONNECT", TRACE_RDMA_QP_DISCONNECT,
182 : OWNER_TYPE_NONE, OBJECT_NONE, 0,
183 : SPDK_TRACE_ARG_TYPE_INT, "");
184 0 : spdk_trace_register_description("RDMA_QP_DESTROY", TRACE_RDMA_QP_DESTROY,
185 : OWNER_TYPE_NONE, OBJECT_NONE, 0,
186 : SPDK_TRACE_ARG_TYPE_INT, "");
187 :
188 0 : spdk_trace_tpoint_register_relation(TRACE_BDEV_IO_START, OBJECT_NVMF_RDMA_IO, 1);
189 0 : spdk_trace_tpoint_register_relation(TRACE_BDEV_IO_DONE, OBJECT_NVMF_RDMA_IO, 0);
190 0 : }
191 2 : SPDK_TRACE_REGISTER_FN(nvmf_trace, "nvmf_rdma", TRACE_GROUP_NVMF_RDMA)
192 :
193 : enum spdk_nvmf_rdma_wr_type {
194 : RDMA_WR_TYPE_RECV,
195 : RDMA_WR_TYPE_SEND,
196 : RDMA_WR_TYPE_DATA,
197 : };
198 :
199 : struct spdk_nvmf_rdma_wr {
200 : /* Uses enum spdk_nvmf_rdma_wr_type */
201 : uint8_t type;
202 : };
203 :
204 : /* This structure holds commands as they are received off the wire.
205 : * It must be dynamically paired with a full request object
206 : * (spdk_nvmf_rdma_request) to service a request. It is separate
207 : * from the request because RDMA does not appear to order
208 : * completions, so occasionally we'll get a new incoming
209 : * command when there aren't any free request objects.
210 : */
211 : struct spdk_nvmf_rdma_recv {
212 : struct ibv_recv_wr wr;
213 : struct ibv_sge sgl[NVMF_DEFAULT_RX_SGE];
214 :
215 : struct spdk_nvmf_rdma_qpair *qpair;
216 :
217 : /* In-capsule data buffer */
218 : uint8_t *buf;
219 :
220 : struct spdk_nvmf_rdma_wr rdma_wr;
221 : uint64_t receive_tsc;
222 :
223 : STAILQ_ENTRY(spdk_nvmf_rdma_recv) link;
224 : };
225 :
226 : struct spdk_nvmf_rdma_request_data {
227 : struct ibv_send_wr wr;
228 : struct ibv_sge sgl[SPDK_NVMF_MAX_SGL_ENTRIES];
229 : };
230 :
231 : struct spdk_nvmf_rdma_request {
232 : struct spdk_nvmf_request req;
233 :
234 : bool fused_failed;
235 :
236 : struct spdk_nvmf_rdma_wr data_wr;
237 : struct spdk_nvmf_rdma_wr rsp_wr;
238 :
239 : /* Uses enum spdk_nvmf_rdma_request_state */
240 : uint8_t state;
241 :
242 : /* Data offset in req.iov */
243 : uint32_t offset;
244 :
245 : struct spdk_nvmf_rdma_recv *recv;
246 :
247 : struct {
248 : struct ibv_send_wr wr;
249 : struct ibv_sge sgl[NVMF_DEFAULT_RSP_SGE];
250 : } rsp;
251 :
252 : uint16_t iovpos;
253 : uint16_t num_outstanding_data_wr;
254 : /* Used to split Write IO with multi SGL payload */
255 : uint16_t num_remaining_data_wr;
256 : uint64_t receive_tsc;
257 : struct spdk_nvmf_rdma_request *fused_pair;
258 : STAILQ_ENTRY(spdk_nvmf_rdma_request) state_link;
259 : struct ibv_send_wr *remaining_tranfer_in_wrs;
260 : struct ibv_send_wr *transfer_wr;
261 : struct spdk_nvmf_rdma_request_data data;
262 : };
263 :
264 : struct spdk_nvmf_rdma_resource_opts {
265 : struct spdk_nvmf_rdma_qpair *qpair;
266 : /* qp points either to an ibv_qp object or an ibv_srq object depending on the value of shared. */
267 : void *qp;
268 : struct spdk_rdma_utils_mem_map *map;
269 : uint32_t max_queue_depth;
270 : uint32_t in_capsule_data_size;
271 : bool shared;
272 : };
273 :
274 : struct spdk_nvmf_rdma_resources {
275 : /* Array of size "max_queue_depth" containing RDMA requests. */
276 : struct spdk_nvmf_rdma_request *reqs;
277 :
278 : /* Array of size "max_queue_depth" containing RDMA recvs. */
279 : struct spdk_nvmf_rdma_recv *recvs;
280 :
281 : /* Array of size "max_queue_depth" containing 64 byte capsules
282 : * used for receive.
283 : */
284 : union nvmf_h2c_msg *cmds;
285 :
286 : /* Array of size "max_queue_depth" containing 16 byte completions
287 : * to be sent back to the user.
288 : */
289 : union nvmf_c2h_msg *cpls;
290 :
291 : /* Array of size "max_queue_depth * InCapsuleDataSize" containing
292 : * buffers to be used for in capsule data.
293 : */
294 : void *bufs;
295 :
296 : /* Receives that are waiting for a request object */
297 : STAILQ_HEAD(, spdk_nvmf_rdma_recv) incoming_queue;
298 :
299 : /* Queue to track free requests */
300 : STAILQ_HEAD(, spdk_nvmf_rdma_request) free_queue;
301 : };
302 :
303 : typedef void (*spdk_nvmf_rdma_qpair_ibv_event)(struct spdk_nvmf_rdma_qpair *rqpair);
304 :
305 : typedef void (*spdk_poller_destroy_cb)(void *ctx);
306 :
307 : struct spdk_nvmf_rdma_ibv_event_ctx {
308 : struct spdk_nvmf_rdma_qpair *rqpair;
309 : spdk_nvmf_rdma_qpair_ibv_event cb_fn;
310 : /* Link to other ibv events associated with this qpair */
311 : STAILQ_ENTRY(spdk_nvmf_rdma_ibv_event_ctx) link;
312 : };
313 :
314 : struct spdk_nvmf_rdma_qpair {
315 : struct spdk_nvmf_qpair qpair;
316 :
317 : struct spdk_nvmf_rdma_device *device;
318 : struct spdk_nvmf_rdma_poller *poller;
319 :
320 : struct spdk_rdma_provider_qp *rdma_qp;
321 : struct rdma_cm_id *cm_id;
322 : struct spdk_rdma_provider_srq *srq;
323 : struct rdma_cm_id *listen_id;
324 :
325 : /* Cache the QP number to improve QP search by RB tree. */
326 : uint32_t qp_num;
327 :
328 : /* The maximum number of I/O outstanding on this connection at one time */
329 : uint16_t max_queue_depth;
330 :
331 : /* The maximum number of active RDMA READ and ATOMIC operations at one time */
332 : uint16_t max_read_depth;
333 :
334 : /* The maximum number of RDMA SEND operations at one time */
335 : uint32_t max_send_depth;
336 :
337 : /* The current number of outstanding WRs from this qpair's
338 : * recv queue. Should not exceed device->attr.max_queue_depth.
339 : */
340 : uint16_t current_recv_depth;
341 :
342 : /* The current number of active RDMA READ operations */
343 : uint16_t current_read_depth;
344 :
345 : /* The current number of posted WRs from this qpair's
346 : * send queue. Should not exceed max_send_depth.
347 : */
348 : uint32_t current_send_depth;
349 :
350 : /* The maximum number of SGEs per WR on the send queue */
351 : uint32_t max_send_sge;
352 :
353 : /* The maximum number of SGEs per WR on the recv queue */
354 : uint32_t max_recv_sge;
355 :
356 : struct spdk_nvmf_rdma_resources *resources;
357 :
358 : STAILQ_HEAD(, spdk_nvmf_rdma_request) pending_rdma_read_queue;
359 :
360 : STAILQ_HEAD(, spdk_nvmf_rdma_request) pending_rdma_write_queue;
361 :
362 : STAILQ_HEAD(, spdk_nvmf_rdma_request) pending_rdma_send_queue;
363 :
364 : /* Number of requests not in the free state */
365 : uint32_t qd;
366 :
367 : bool ibv_in_error_state;
368 :
369 : RB_ENTRY(spdk_nvmf_rdma_qpair) node;
370 :
371 : STAILQ_ENTRY(spdk_nvmf_rdma_qpair) recv_link;
372 :
373 : STAILQ_ENTRY(spdk_nvmf_rdma_qpair) send_link;
374 :
375 : /* Points to the a request that has fuse bits set to
376 : * SPDK_NVME_CMD_FUSE_FIRST, when the qpair is waiting
377 : * for the request that has SPDK_NVME_CMD_FUSE_SECOND.
378 : */
379 : struct spdk_nvmf_rdma_request *fused_first;
380 :
381 : /*
382 : * io_channel which is used to destroy qpair when it is removed from poll group
383 : */
384 : struct spdk_io_channel *destruct_channel;
385 :
386 : /* List of ibv async events */
387 : STAILQ_HEAD(, spdk_nvmf_rdma_ibv_event_ctx) ibv_events;
388 :
389 : /* Lets us know that we have received the last_wqe event. */
390 : bool last_wqe_reached;
391 :
392 : /* Indicate that nvmf_rdma_close_qpair is called */
393 : bool to_close;
394 : };
395 :
396 : struct spdk_nvmf_rdma_poller_stat {
397 : uint64_t completions;
398 : uint64_t polls;
399 : uint64_t idle_polls;
400 : uint64_t requests;
401 : uint64_t request_latency;
402 : uint64_t pending_free_request;
403 : uint64_t pending_rdma_read;
404 : uint64_t pending_rdma_write;
405 : uint64_t pending_rdma_send;
406 : struct spdk_rdma_provider_qp_stats qp_stats;
407 : };
408 :
409 : struct spdk_nvmf_rdma_poller {
410 : struct spdk_nvmf_rdma_device *device;
411 : struct spdk_nvmf_rdma_poll_group *group;
412 :
413 : int num_cqe;
414 : int required_num_wr;
415 : struct ibv_cq *cq;
416 :
417 : /* The maximum number of I/O outstanding on the shared receive queue at one time */
418 : uint16_t max_srq_depth;
419 : bool need_destroy;
420 :
421 : /* Shared receive queue */
422 : struct spdk_rdma_provider_srq *srq;
423 :
424 : struct spdk_nvmf_rdma_resources *resources;
425 : struct spdk_nvmf_rdma_poller_stat stat;
426 :
427 : spdk_poller_destroy_cb destroy_cb;
428 : void *destroy_cb_ctx;
429 :
430 : RB_HEAD(qpairs_tree, spdk_nvmf_rdma_qpair) qpairs;
431 :
432 : STAILQ_HEAD(, spdk_nvmf_rdma_qpair) qpairs_pending_recv;
433 :
434 : STAILQ_HEAD(, spdk_nvmf_rdma_qpair) qpairs_pending_send;
435 :
436 : TAILQ_ENTRY(spdk_nvmf_rdma_poller) link;
437 : };
438 :
439 : struct spdk_nvmf_rdma_poll_group_stat {
440 : uint64_t pending_data_buffer;
441 : };
442 :
443 : struct spdk_nvmf_rdma_poll_group {
444 : struct spdk_nvmf_transport_poll_group group;
445 : struct spdk_nvmf_rdma_poll_group_stat stat;
446 : TAILQ_HEAD(, spdk_nvmf_rdma_poller) pollers;
447 : TAILQ_ENTRY(spdk_nvmf_rdma_poll_group) link;
448 : };
449 :
450 : struct spdk_nvmf_rdma_conn_sched {
451 : struct spdk_nvmf_rdma_poll_group *next_admin_pg;
452 : struct spdk_nvmf_rdma_poll_group *next_io_pg;
453 : };
454 :
455 : /* Assuming rdma_cm uses just one protection domain per ibv_context. */
456 : struct spdk_nvmf_rdma_device {
457 : struct ibv_device_attr attr;
458 : struct ibv_context *context;
459 :
460 : struct spdk_rdma_utils_mem_map *map;
461 : struct ibv_pd *pd;
462 :
463 : int num_srq;
464 : bool need_destroy;
465 : bool ready_to_destroy;
466 : bool is_ready;
467 :
468 : TAILQ_ENTRY(spdk_nvmf_rdma_device) link;
469 : };
470 :
471 : struct spdk_nvmf_rdma_port {
472 : const struct spdk_nvme_transport_id *trid;
473 : struct rdma_cm_id *id;
474 : struct spdk_nvmf_rdma_device *device;
475 : TAILQ_ENTRY(spdk_nvmf_rdma_port) link;
476 : };
477 :
478 : struct rdma_transport_opts {
479 : int num_cqe;
480 : uint32_t max_srq_depth;
481 : bool no_srq;
482 : bool no_wr_batching;
483 : int acceptor_backlog;
484 : };
485 :
486 : struct spdk_nvmf_rdma_transport {
487 : struct spdk_nvmf_transport transport;
488 : struct rdma_transport_opts rdma_opts;
489 :
490 : struct spdk_nvmf_rdma_conn_sched conn_sched;
491 :
492 : struct rdma_event_channel *event_channel;
493 :
494 : struct spdk_mempool *data_wr_pool;
495 :
496 : struct spdk_poller *accept_poller;
497 :
498 : /* fields used to poll RDMA/IB events */
499 : nfds_t npoll_fds;
500 : struct pollfd *poll_fds;
501 :
502 : TAILQ_HEAD(, spdk_nvmf_rdma_device) devices;
503 : TAILQ_HEAD(, spdk_nvmf_rdma_port) ports;
504 : TAILQ_HEAD(, spdk_nvmf_rdma_poll_group) poll_groups;
505 :
506 : /* ports that are removed unexpectedly and need retry listen */
507 : TAILQ_HEAD(, spdk_nvmf_rdma_port) retry_ports;
508 : };
509 :
510 : struct poller_manage_ctx {
511 : struct spdk_nvmf_rdma_transport *rtransport;
512 : struct spdk_nvmf_rdma_poll_group *rgroup;
513 : struct spdk_nvmf_rdma_poller *rpoller;
514 : struct spdk_nvmf_rdma_device *device;
515 :
516 : struct spdk_thread *thread;
517 : volatile int *inflight_op_counter;
518 : };
519 :
520 : static const struct spdk_json_object_decoder rdma_transport_opts_decoder[] = {
521 : {
522 : "num_cqe", offsetof(struct rdma_transport_opts, num_cqe),
523 : spdk_json_decode_int32, true
524 : },
525 : {
526 : "max_srq_depth", offsetof(struct rdma_transport_opts, max_srq_depth),
527 : spdk_json_decode_uint32, true
528 : },
529 : {
530 : "no_srq", offsetof(struct rdma_transport_opts, no_srq),
531 : spdk_json_decode_bool, true
532 : },
533 : {
534 : "no_wr_batching", offsetof(struct rdma_transport_opts, no_wr_batching),
535 : spdk_json_decode_bool, true
536 : },
537 : {
538 : "acceptor_backlog", offsetof(struct rdma_transport_opts, acceptor_backlog),
539 : spdk_json_decode_int32, true
540 : },
541 : };
542 :
543 : static int
544 2 : nvmf_rdma_qpair_compare(struct spdk_nvmf_rdma_qpair *rqpair1, struct spdk_nvmf_rdma_qpair *rqpair2)
545 : {
546 2 : return rqpair1->qp_num < rqpair2->qp_num ? -1 : rqpair1->qp_num > rqpair2->qp_num;
547 : }
548 :
549 0 : RB_GENERATE_STATIC(qpairs_tree, spdk_nvmf_rdma_qpair, node, nvmf_rdma_qpair_compare);
550 :
551 : static bool nvmf_rdma_request_process(struct spdk_nvmf_rdma_transport *rtransport,
552 : struct spdk_nvmf_rdma_request *rdma_req);
553 :
554 : static void _poller_submit_sends(struct spdk_nvmf_rdma_transport *rtransport,
555 : struct spdk_nvmf_rdma_poller *rpoller);
556 :
557 : static void _poller_submit_recvs(struct spdk_nvmf_rdma_transport *rtransport,
558 : struct spdk_nvmf_rdma_poller *rpoller);
559 :
560 : static void _nvmf_rdma_remove_destroyed_device(void *c);
561 :
562 : static inline enum spdk_nvme_media_error_status_code
563 0 : nvmf_rdma_dif_error_to_compl_status(uint8_t err_type) {
564 : enum spdk_nvme_media_error_status_code result;
565 0 : switch (err_type)
566 : {
567 0 : case SPDK_DIF_REFTAG_ERROR:
568 0 : result = SPDK_NVME_SC_REFERENCE_TAG_CHECK_ERROR;
569 0 : break;
570 0 : case SPDK_DIF_APPTAG_ERROR:
571 0 : result = SPDK_NVME_SC_APPLICATION_TAG_CHECK_ERROR;
572 0 : break;
573 0 : case SPDK_DIF_GUARD_ERROR:
574 0 : result = SPDK_NVME_SC_GUARD_CHECK_ERROR;
575 0 : break;
576 0 : default:
577 0 : SPDK_UNREACHABLE();
578 : }
579 :
580 0 : return result;
581 : }
582 :
583 : /*
584 : * Return data_wrs to pool starting from \b data_wr
585 : * Request's own response and data WR are excluded
586 : */
587 : static void
588 7 : _nvmf_rdma_request_free_data(struct spdk_nvmf_rdma_request *rdma_req,
589 : struct ibv_send_wr *data_wr,
590 : struct spdk_mempool *pool)
591 : {
592 7 : struct spdk_nvmf_rdma_request_data *work_requests[SPDK_NVMF_MAX_SGL_ENTRIES];
593 : struct spdk_nvmf_rdma_request_data *nvmf_data;
594 : struct ibv_send_wr *next_send_wr;
595 7 : uint64_t req_wrid = (uint64_t)&rdma_req->data_wr;
596 7 : uint32_t num_wrs = 0;
597 :
598 15 : while (data_wr && data_wr->wr_id == req_wrid) {
599 8 : nvmf_data = SPDK_CONTAINEROF(data_wr, struct spdk_nvmf_rdma_request_data, wr);
600 8 : memset(nvmf_data->sgl, 0, sizeof(data_wr->sg_list[0]) * data_wr->num_sge);
601 8 : data_wr->num_sge = 0;
602 8 : next_send_wr = data_wr->next;
603 8 : if (data_wr != &rdma_req->data.wr) {
604 1 : data_wr->next = NULL;
605 1 : assert(num_wrs < SPDK_NVMF_MAX_SGL_ENTRIES);
606 1 : work_requests[num_wrs] = nvmf_data;
607 1 : num_wrs++;
608 : }
609 8 : data_wr = (!next_send_wr || next_send_wr == &rdma_req->rsp.wr) ? NULL : next_send_wr;
610 : }
611 :
612 7 : if (num_wrs) {
613 1 : spdk_mempool_put_bulk(pool, (void **) work_requests, num_wrs);
614 : }
615 7 : }
616 :
617 : static void
618 7 : nvmf_rdma_request_free_data(struct spdk_nvmf_rdma_request *rdma_req,
619 : struct spdk_nvmf_rdma_transport *rtransport)
620 : {
621 7 : rdma_req->num_outstanding_data_wr = 0;
622 :
623 7 : _nvmf_rdma_request_free_data(rdma_req, rdma_req->transfer_wr, rtransport->data_wr_pool);
624 :
625 7 : if (rdma_req->remaining_tranfer_in_wrs) {
626 0 : _nvmf_rdma_request_free_data(rdma_req, rdma_req->remaining_tranfer_in_wrs,
627 : rtransport->data_wr_pool);
628 0 : rdma_req->remaining_tranfer_in_wrs = NULL;
629 : }
630 :
631 7 : rdma_req->data.wr.next = NULL;
632 7 : rdma_req->rsp.wr.next = NULL;
633 7 : }
634 :
635 : static void
636 0 : nvmf_rdma_dump_request(struct spdk_nvmf_rdma_request *req)
637 : {
638 0 : SPDK_ERRLOG("\t\tRequest Data From Pool: %d\n", req->req.data_from_pool);
639 0 : if (req->req.cmd) {
640 0 : SPDK_ERRLOG("\t\tRequest opcode: %d\n", req->req.cmd->nvmf_cmd.opcode);
641 : }
642 0 : if (req->recv) {
643 0 : SPDK_ERRLOG("\t\tRequest recv wr_id%lu\n", req->recv->wr.wr_id);
644 : }
645 0 : }
646 :
647 : static void
648 0 : nvmf_rdma_dump_qpair_contents(struct spdk_nvmf_rdma_qpair *rqpair)
649 : {
650 : int i;
651 :
652 0 : SPDK_ERRLOG("Dumping contents of queue pair (QID %d)\n", rqpair->qpair.qid);
653 0 : for (i = 0; i < rqpair->max_queue_depth; i++) {
654 0 : if (rqpair->resources->reqs[i].state != RDMA_REQUEST_STATE_FREE) {
655 0 : nvmf_rdma_dump_request(&rqpair->resources->reqs[i]);
656 : }
657 : }
658 0 : }
659 :
660 : static void
661 1 : nvmf_rdma_resources_destroy(struct spdk_nvmf_rdma_resources *resources)
662 : {
663 1 : spdk_free(resources->cmds);
664 1 : spdk_free(resources->cpls);
665 1 : spdk_free(resources->bufs);
666 1 : spdk_free(resources->reqs);
667 1 : spdk_free(resources->recvs);
668 1 : free(resources);
669 1 : }
670 :
671 :
672 : static struct spdk_nvmf_rdma_resources *
673 1 : nvmf_rdma_resources_create(struct spdk_nvmf_rdma_resource_opts *opts)
674 : {
675 : struct spdk_nvmf_rdma_resources *resources;
676 : struct spdk_nvmf_rdma_request *rdma_req;
677 : struct spdk_nvmf_rdma_recv *rdma_recv;
678 1 : struct spdk_rdma_provider_qp *qp = NULL;
679 1 : struct spdk_rdma_provider_srq *srq = NULL;
680 1 : struct ibv_recv_wr *bad_wr = NULL;
681 1 : struct spdk_rdma_utils_memory_translation translation;
682 : uint32_t i;
683 1 : int rc = 0;
684 :
685 1 : resources = calloc(1, sizeof(struct spdk_nvmf_rdma_resources));
686 1 : if (!resources) {
687 0 : SPDK_ERRLOG("Unable to allocate resources for receive queue.\n");
688 0 : return NULL;
689 : }
690 :
691 1 : resources->reqs = spdk_zmalloc(opts->max_queue_depth * sizeof(*resources->reqs),
692 : 0x1000, NULL, SPDK_ENV_LCORE_ID_ANY, SPDK_MALLOC_DMA);
693 1 : resources->recvs = spdk_zmalloc(opts->max_queue_depth * sizeof(*resources->recvs),
694 : 0x1000, NULL, SPDK_ENV_LCORE_ID_ANY, SPDK_MALLOC_DMA);
695 1 : resources->cmds = spdk_zmalloc(opts->max_queue_depth * sizeof(*resources->cmds),
696 : 0x1000, NULL, SPDK_ENV_LCORE_ID_ANY, SPDK_MALLOC_DMA);
697 1 : resources->cpls = spdk_zmalloc(opts->max_queue_depth * sizeof(*resources->cpls),
698 : 0x1000, NULL, SPDK_ENV_LCORE_ID_ANY, SPDK_MALLOC_DMA);
699 :
700 1 : if (opts->in_capsule_data_size > 0) {
701 1 : resources->bufs = spdk_zmalloc(opts->max_queue_depth * opts->in_capsule_data_size,
702 : 0x1000, NULL, SPDK_ENV_LCORE_ID_ANY,
703 : SPDK_MALLOC_DMA);
704 : }
705 :
706 1 : if (!resources->reqs || !resources->recvs || !resources->cmds ||
707 1 : !resources->cpls || (opts->in_capsule_data_size && !resources->bufs)) {
708 0 : SPDK_ERRLOG("Unable to allocate sufficient memory for RDMA queue.\n");
709 0 : goto cleanup;
710 : }
711 :
712 1 : SPDK_DEBUGLOG(rdma, "Command Array: %p Length: %lx\n",
713 : resources->cmds, opts->max_queue_depth * sizeof(*resources->cmds));
714 1 : SPDK_DEBUGLOG(rdma, "Completion Array: %p Length: %lx\n",
715 : resources->cpls, opts->max_queue_depth * sizeof(*resources->cpls));
716 1 : if (resources->bufs) {
717 1 : SPDK_DEBUGLOG(rdma, "In Capsule Data Array: %p Length: %x\n",
718 : resources->bufs, opts->max_queue_depth *
719 : opts->in_capsule_data_size);
720 : }
721 :
722 : /* Initialize queues */
723 1 : STAILQ_INIT(&resources->incoming_queue);
724 1 : STAILQ_INIT(&resources->free_queue);
725 :
726 1 : if (opts->shared) {
727 1 : srq = (struct spdk_rdma_provider_srq *)opts->qp;
728 : } else {
729 0 : qp = (struct spdk_rdma_provider_qp *)opts->qp;
730 : }
731 :
732 129 : for (i = 0; i < opts->max_queue_depth; i++) {
733 128 : rdma_recv = &resources->recvs[i];
734 128 : rdma_recv->qpair = opts->qpair;
735 :
736 : /* Set up memory to receive commands */
737 128 : if (resources->bufs) {
738 128 : rdma_recv->buf = (void *)((uintptr_t)resources->bufs + (i *
739 128 : opts->in_capsule_data_size));
740 : }
741 :
742 128 : rdma_recv->rdma_wr.type = RDMA_WR_TYPE_RECV;
743 :
744 128 : rdma_recv->sgl[0].addr = (uintptr_t)&resources->cmds[i];
745 128 : rdma_recv->sgl[0].length = sizeof(resources->cmds[i]);
746 128 : rc = spdk_rdma_utils_get_translation(opts->map, &resources->cmds[i], sizeof(resources->cmds[i]),
747 : &translation);
748 128 : if (rc) {
749 0 : goto cleanup;
750 : }
751 128 : rdma_recv->sgl[0].lkey = spdk_rdma_utils_memory_translation_get_lkey(&translation);
752 128 : rdma_recv->wr.num_sge = 1;
753 :
754 128 : if (rdma_recv->buf) {
755 128 : rdma_recv->sgl[1].addr = (uintptr_t)rdma_recv->buf;
756 128 : rdma_recv->sgl[1].length = opts->in_capsule_data_size;
757 128 : rc = spdk_rdma_utils_get_translation(opts->map, rdma_recv->buf, opts->in_capsule_data_size,
758 : &translation);
759 128 : if (rc) {
760 0 : goto cleanup;
761 : }
762 128 : rdma_recv->sgl[1].lkey = spdk_rdma_utils_memory_translation_get_lkey(&translation);
763 128 : rdma_recv->wr.num_sge++;
764 : }
765 :
766 128 : rdma_recv->wr.wr_id = (uintptr_t)&rdma_recv->rdma_wr;
767 128 : rdma_recv->wr.sg_list = rdma_recv->sgl;
768 128 : if (srq) {
769 0 : spdk_rdma_provider_srq_queue_recv_wrs(srq, &rdma_recv->wr);
770 : } else {
771 128 : spdk_rdma_provider_qp_queue_recv_wrs(qp, &rdma_recv->wr);
772 : }
773 : }
774 :
775 129 : for (i = 0; i < opts->max_queue_depth; i++) {
776 128 : rdma_req = &resources->reqs[i];
777 :
778 128 : if (opts->qpair != NULL) {
779 128 : rdma_req->req.qpair = &opts->qpair->qpair;
780 : } else {
781 0 : rdma_req->req.qpair = NULL;
782 : }
783 128 : rdma_req->req.cmd = NULL;
784 128 : rdma_req->req.iovcnt = 0;
785 128 : rdma_req->req.stripped_data = NULL;
786 :
787 : /* Set up memory to send responses */
788 128 : rdma_req->req.rsp = &resources->cpls[i];
789 :
790 128 : rdma_req->rsp.sgl[0].addr = (uintptr_t)&resources->cpls[i];
791 128 : rdma_req->rsp.sgl[0].length = sizeof(resources->cpls[i]);
792 128 : rc = spdk_rdma_utils_get_translation(opts->map, &resources->cpls[i], sizeof(resources->cpls[i]),
793 : &translation);
794 128 : if (rc) {
795 0 : goto cleanup;
796 : }
797 128 : rdma_req->rsp.sgl[0].lkey = spdk_rdma_utils_memory_translation_get_lkey(&translation);
798 :
799 128 : rdma_req->rsp_wr.type = RDMA_WR_TYPE_SEND;
800 128 : rdma_req->rsp.wr.wr_id = (uintptr_t)&rdma_req->rsp_wr;
801 128 : rdma_req->rsp.wr.next = NULL;
802 128 : rdma_req->rsp.wr.opcode = IBV_WR_SEND;
803 128 : rdma_req->rsp.wr.send_flags = IBV_SEND_SIGNALED;
804 128 : rdma_req->rsp.wr.sg_list = rdma_req->rsp.sgl;
805 128 : rdma_req->rsp.wr.num_sge = SPDK_COUNTOF(rdma_req->rsp.sgl);
806 :
807 : /* Set up memory for data buffers */
808 128 : rdma_req->data_wr.type = RDMA_WR_TYPE_DATA;
809 128 : rdma_req->data.wr.wr_id = (uintptr_t)&rdma_req->data_wr;
810 128 : rdma_req->data.wr.next = NULL;
811 128 : rdma_req->data.wr.send_flags = IBV_SEND_SIGNALED;
812 128 : rdma_req->data.wr.sg_list = rdma_req->data.sgl;
813 128 : rdma_req->data.wr.num_sge = SPDK_COUNTOF(rdma_req->data.sgl);
814 :
815 : /* Initialize request state to FREE */
816 128 : rdma_req->state = RDMA_REQUEST_STATE_FREE;
817 128 : STAILQ_INSERT_TAIL(&resources->free_queue, rdma_req, state_link);
818 : }
819 :
820 1 : if (srq) {
821 0 : rc = spdk_rdma_provider_srq_flush_recv_wrs(srq, &bad_wr);
822 : } else {
823 1 : rc = spdk_rdma_provider_qp_flush_recv_wrs(qp, &bad_wr);
824 : }
825 :
826 1 : if (rc) {
827 0 : goto cleanup;
828 : }
829 :
830 1 : return resources;
831 :
832 0 : cleanup:
833 0 : nvmf_rdma_resources_destroy(resources);
834 0 : return NULL;
835 : }
836 :
837 : static void
838 0 : nvmf_rdma_qpair_clean_ibv_events(struct spdk_nvmf_rdma_qpair *rqpair)
839 : {
840 : struct spdk_nvmf_rdma_ibv_event_ctx *ctx, *tctx;
841 0 : STAILQ_FOREACH_SAFE(ctx, &rqpair->ibv_events, link, tctx) {
842 0 : ctx->rqpair = NULL;
843 : /* Memory allocated for ctx is freed in nvmf_rdma_qpair_process_ibv_event */
844 0 : STAILQ_REMOVE(&rqpair->ibv_events, ctx, spdk_nvmf_rdma_ibv_event_ctx, link);
845 : }
846 0 : }
847 :
848 : static void nvmf_rdma_poller_destroy(struct spdk_nvmf_rdma_poller *poller);
849 :
850 : static void
851 0 : nvmf_rdma_qpair_destroy(struct spdk_nvmf_rdma_qpair *rqpair)
852 : {
853 : struct spdk_nvmf_rdma_recv *rdma_recv, *recv_tmp;
854 0 : struct ibv_recv_wr *bad_recv_wr = NULL;
855 : int rc;
856 :
857 0 : spdk_trace_record(TRACE_RDMA_QP_DESTROY, 0, 0, (uintptr_t)rqpair);
858 :
859 0 : if (rqpair->qd != 0) {
860 0 : struct spdk_nvmf_qpair *qpair = &rqpair->qpair;
861 0 : struct spdk_nvmf_rdma_transport *rtransport = SPDK_CONTAINEROF(qpair->transport,
862 : struct spdk_nvmf_rdma_transport, transport);
863 : struct spdk_nvmf_rdma_request *req;
864 0 : uint32_t i, max_req_count = 0;
865 :
866 0 : SPDK_WARNLOG("Destroying qpair when queue depth is %d\n", rqpair->qd);
867 :
868 0 : if (rqpair->srq == NULL) {
869 0 : nvmf_rdma_dump_qpair_contents(rqpair);
870 0 : max_req_count = rqpair->max_queue_depth;
871 0 : } else if (rqpair->poller && rqpair->resources) {
872 0 : max_req_count = rqpair->poller->max_srq_depth;
873 : }
874 :
875 0 : SPDK_DEBUGLOG(rdma, "Release incomplete requests\n");
876 0 : for (i = 0; i < max_req_count; i++) {
877 0 : req = &rqpair->resources->reqs[i];
878 0 : if (req->req.qpair == qpair && req->state != RDMA_REQUEST_STATE_FREE) {
879 : /* nvmf_rdma_request_process checks qpair ibv and internal state
880 : * and completes a request */
881 0 : nvmf_rdma_request_process(rtransport, req);
882 : }
883 : }
884 0 : assert(rqpair->qd == 0);
885 : }
886 :
887 0 : if (rqpair->poller) {
888 0 : RB_REMOVE(qpairs_tree, &rqpair->poller->qpairs, rqpair);
889 :
890 0 : if (rqpair->srq != NULL && rqpair->resources != NULL) {
891 : /* Drop all received but unprocessed commands for this queue and return them to SRQ */
892 0 : STAILQ_FOREACH_SAFE(rdma_recv, &rqpair->resources->incoming_queue, link, recv_tmp) {
893 0 : if (rqpair == rdma_recv->qpair) {
894 0 : STAILQ_REMOVE(&rqpair->resources->incoming_queue, rdma_recv, spdk_nvmf_rdma_recv, link);
895 0 : spdk_rdma_provider_srq_queue_recv_wrs(rqpair->srq, &rdma_recv->wr);
896 0 : rc = spdk_rdma_provider_srq_flush_recv_wrs(rqpair->srq, &bad_recv_wr);
897 0 : if (rc) {
898 0 : SPDK_ERRLOG("Unable to re-post rx descriptor\n");
899 : }
900 : }
901 : }
902 : }
903 : }
904 :
905 0 : if (rqpair->cm_id) {
906 0 : if (rqpair->rdma_qp != NULL) {
907 0 : spdk_rdma_provider_qp_destroy(rqpair->rdma_qp);
908 0 : rqpair->rdma_qp = NULL;
909 : }
910 :
911 0 : if (rqpair->poller != NULL && rqpair->srq == NULL) {
912 0 : rqpair->poller->required_num_wr -= MAX_WR_PER_QP(rqpair->max_queue_depth);
913 : }
914 : }
915 :
916 0 : if (rqpair->srq == NULL && rqpair->resources != NULL) {
917 0 : nvmf_rdma_resources_destroy(rqpair->resources);
918 : }
919 :
920 0 : nvmf_rdma_qpair_clean_ibv_events(rqpair);
921 :
922 0 : if (rqpair->destruct_channel) {
923 0 : spdk_put_io_channel(rqpair->destruct_channel);
924 0 : rqpair->destruct_channel = NULL;
925 : }
926 :
927 0 : if (rqpair->poller && rqpair->poller->need_destroy && RB_EMPTY(&rqpair->poller->qpairs)) {
928 0 : nvmf_rdma_poller_destroy(rqpair->poller);
929 : }
930 :
931 : /* destroy cm_id last so cma device will not be freed before we destroy the cq. */
932 0 : if (rqpair->cm_id) {
933 0 : rdma_destroy_id(rqpair->cm_id);
934 : }
935 :
936 0 : free(rqpair);
937 0 : }
938 :
939 : static int
940 5 : nvmf_rdma_resize_cq(struct spdk_nvmf_rdma_qpair *rqpair, struct spdk_nvmf_rdma_device *device)
941 : {
942 : struct spdk_nvmf_rdma_poller *rpoller;
943 : int rc, num_cqe, required_num_wr;
944 :
945 : /* Enlarge CQ size dynamically */
946 5 : rpoller = rqpair->poller;
947 5 : required_num_wr = rpoller->required_num_wr + MAX_WR_PER_QP(rqpair->max_queue_depth);
948 5 : num_cqe = rpoller->num_cqe;
949 5 : if (num_cqe < required_num_wr) {
950 4 : num_cqe = spdk_max(num_cqe * 2, required_num_wr);
951 4 : num_cqe = spdk_min(num_cqe, device->attr.max_cqe);
952 : }
953 :
954 5 : if (rpoller->num_cqe != num_cqe) {
955 4 : if (device->context->device->transport_type == IBV_TRANSPORT_IWARP) {
956 1 : SPDK_ERRLOG("iWARP doesn't support CQ resize. Current capacity %u, required %u\n"
957 : "Using CQ of insufficient size may lead to CQ overrun\n", rpoller->num_cqe, num_cqe);
958 1 : return -1;
959 : }
960 3 : if (required_num_wr > device->attr.max_cqe) {
961 1 : SPDK_ERRLOG("RDMA CQE requirement (%d) exceeds device max_cqe limitation (%d)\n",
962 : required_num_wr, device->attr.max_cqe);
963 1 : return -1;
964 : }
965 :
966 2 : SPDK_DEBUGLOG(rdma, "Resize RDMA CQ from %d to %d\n", rpoller->num_cqe, num_cqe);
967 2 : rc = ibv_resize_cq(rpoller->cq, num_cqe);
968 2 : if (rc) {
969 1 : SPDK_ERRLOG("RDMA CQ resize failed: errno %d: %s\n", errno, spdk_strerror(errno));
970 1 : return -1;
971 : }
972 :
973 1 : rpoller->num_cqe = num_cqe;
974 : }
975 :
976 2 : rpoller->required_num_wr = required_num_wr;
977 2 : return 0;
978 : }
979 :
980 : static int
981 0 : nvmf_rdma_qpair_initialize(struct spdk_nvmf_qpair *qpair)
982 : {
983 : struct spdk_nvmf_rdma_qpair *rqpair;
984 : struct spdk_nvmf_rdma_transport *rtransport;
985 : struct spdk_nvmf_transport *transport;
986 0 : struct spdk_nvmf_rdma_resource_opts opts;
987 : struct spdk_nvmf_rdma_device *device;
988 0 : struct spdk_rdma_provider_qp_init_attr qp_init_attr = {};
989 :
990 0 : rqpair = SPDK_CONTAINEROF(qpair, struct spdk_nvmf_rdma_qpair, qpair);
991 0 : device = rqpair->device;
992 :
993 0 : qp_init_attr.qp_context = rqpair;
994 0 : qp_init_attr.pd = device->pd;
995 0 : qp_init_attr.send_cq = rqpair->poller->cq;
996 0 : qp_init_attr.recv_cq = rqpair->poller->cq;
997 :
998 0 : if (rqpair->srq) {
999 0 : qp_init_attr.srq = rqpair->srq->srq;
1000 : } else {
1001 0 : qp_init_attr.cap.max_recv_wr = rqpair->max_queue_depth;
1002 : }
1003 :
1004 : /* SEND, READ, and WRITE operations */
1005 0 : qp_init_attr.cap.max_send_wr = (uint32_t)rqpair->max_queue_depth * 2;
1006 0 : qp_init_attr.cap.max_send_sge = spdk_min((uint32_t)device->attr.max_sge, NVMF_DEFAULT_TX_SGE);
1007 0 : qp_init_attr.cap.max_recv_sge = spdk_min((uint32_t)device->attr.max_sge, NVMF_DEFAULT_RX_SGE);
1008 0 : qp_init_attr.stats = &rqpair->poller->stat.qp_stats;
1009 :
1010 0 : if (rqpair->srq == NULL && nvmf_rdma_resize_cq(rqpair, device) < 0) {
1011 0 : SPDK_ERRLOG("Failed to resize the completion queue. Cannot initialize qpair.\n");
1012 0 : goto error;
1013 : }
1014 :
1015 0 : rqpair->rdma_qp = spdk_rdma_provider_qp_create(rqpair->cm_id, &qp_init_attr);
1016 0 : if (!rqpair->rdma_qp) {
1017 0 : goto error;
1018 : }
1019 :
1020 0 : rqpair->qp_num = rqpair->rdma_qp->qp->qp_num;
1021 :
1022 0 : rqpair->max_send_depth = spdk_min((uint32_t)(rqpair->max_queue_depth * 2),
1023 : qp_init_attr.cap.max_send_wr);
1024 0 : rqpair->max_send_sge = spdk_min(NVMF_DEFAULT_TX_SGE, qp_init_attr.cap.max_send_sge);
1025 0 : rqpair->max_recv_sge = spdk_min(NVMF_DEFAULT_RX_SGE, qp_init_attr.cap.max_recv_sge);
1026 0 : spdk_trace_record(TRACE_RDMA_QP_CREATE, 0, 0, (uintptr_t)rqpair);
1027 0 : SPDK_DEBUGLOG(rdma, "New RDMA Connection: %p\n", qpair);
1028 :
1029 0 : if (rqpair->poller->srq == NULL) {
1030 0 : rtransport = SPDK_CONTAINEROF(qpair->transport, struct spdk_nvmf_rdma_transport, transport);
1031 0 : transport = &rtransport->transport;
1032 :
1033 0 : opts.qp = rqpair->rdma_qp;
1034 0 : opts.map = device->map;
1035 0 : opts.qpair = rqpair;
1036 0 : opts.shared = false;
1037 0 : opts.max_queue_depth = rqpair->max_queue_depth;
1038 0 : opts.in_capsule_data_size = transport->opts.in_capsule_data_size;
1039 :
1040 0 : rqpair->resources = nvmf_rdma_resources_create(&opts);
1041 :
1042 0 : if (!rqpair->resources) {
1043 0 : SPDK_ERRLOG("Unable to allocate resources for receive queue.\n");
1044 0 : rdma_destroy_qp(rqpair->cm_id);
1045 0 : goto error;
1046 : }
1047 : } else {
1048 0 : rqpair->resources = rqpair->poller->resources;
1049 : }
1050 :
1051 0 : rqpair->current_recv_depth = 0;
1052 0 : STAILQ_INIT(&rqpair->pending_rdma_read_queue);
1053 0 : STAILQ_INIT(&rqpair->pending_rdma_write_queue);
1054 0 : STAILQ_INIT(&rqpair->pending_rdma_send_queue);
1055 0 : rqpair->qpair.queue_depth = 0;
1056 :
1057 0 : return 0;
1058 :
1059 0 : error:
1060 0 : rdma_destroy_id(rqpair->cm_id);
1061 0 : rqpair->cm_id = NULL;
1062 0 : return -1;
1063 : }
1064 :
1065 : /* Append the given recv wr structure to the resource structs outstanding recvs list. */
1066 : /* This function accepts either a single wr or the first wr in a linked list. */
1067 : static void
1068 6 : nvmf_rdma_qpair_queue_recv_wrs(struct spdk_nvmf_rdma_qpair *rqpair, struct ibv_recv_wr *first)
1069 : {
1070 6 : struct spdk_nvmf_rdma_transport *rtransport = SPDK_CONTAINEROF(rqpair->qpair.transport,
1071 : struct spdk_nvmf_rdma_transport, transport);
1072 :
1073 6 : if (rqpair->srq != NULL) {
1074 0 : spdk_rdma_provider_srq_queue_recv_wrs(rqpair->srq, first);
1075 : } else {
1076 6 : if (spdk_rdma_provider_qp_queue_recv_wrs(rqpair->rdma_qp, first)) {
1077 6 : STAILQ_INSERT_TAIL(&rqpair->poller->qpairs_pending_recv, rqpair, recv_link);
1078 : }
1079 : }
1080 :
1081 6 : if (rtransport->rdma_opts.no_wr_batching) {
1082 0 : _poller_submit_recvs(rtransport, rqpair->poller);
1083 : }
1084 6 : }
1085 :
1086 : static inline void
1087 4 : request_transfer_in(struct spdk_nvmf_request *req)
1088 : {
1089 : struct spdk_nvmf_rdma_request *rdma_req;
1090 : struct spdk_nvmf_qpair *qpair;
1091 : struct spdk_nvmf_rdma_qpair *rqpair;
1092 : struct spdk_nvmf_rdma_transport *rtransport;
1093 :
1094 4 : qpair = req->qpair;
1095 4 : rdma_req = SPDK_CONTAINEROF(req, struct spdk_nvmf_rdma_request, req);
1096 4 : rqpair = SPDK_CONTAINEROF(qpair, struct spdk_nvmf_rdma_qpair, qpair);
1097 4 : rtransport = SPDK_CONTAINEROF(rqpair->qpair.transport,
1098 : struct spdk_nvmf_rdma_transport, transport);
1099 :
1100 4 : assert(req->xfer == SPDK_NVME_DATA_HOST_TO_CONTROLLER);
1101 4 : assert(rdma_req != NULL);
1102 :
1103 4 : if (spdk_rdma_provider_qp_queue_send_wrs(rqpair->rdma_qp, rdma_req->transfer_wr)) {
1104 4 : STAILQ_INSERT_TAIL(&rqpair->poller->qpairs_pending_send, rqpair, send_link);
1105 : }
1106 4 : if (rtransport->rdma_opts.no_wr_batching) {
1107 0 : _poller_submit_sends(rtransport, rqpair->poller);
1108 : }
1109 :
1110 4 : assert(rqpair->current_read_depth + rdma_req->num_outstanding_data_wr <= rqpair->max_read_depth);
1111 4 : rqpair->current_read_depth += rdma_req->num_outstanding_data_wr;
1112 4 : assert(rqpair->current_send_depth + rdma_req->num_outstanding_data_wr <= rqpair->max_send_depth);
1113 4 : rqpair->current_send_depth += rdma_req->num_outstanding_data_wr;
1114 4 : }
1115 :
1116 : static inline void
1117 0 : nvmf_rdma_request_reset_transfer_in(struct spdk_nvmf_rdma_request *rdma_req,
1118 : struct spdk_nvmf_rdma_transport *rtransport)
1119 : {
1120 : /* Put completed WRs back to pool and move transfer_wr pointer */
1121 0 : _nvmf_rdma_request_free_data(rdma_req, rdma_req->transfer_wr, rtransport->data_wr_pool);
1122 0 : rdma_req->transfer_wr = rdma_req->remaining_tranfer_in_wrs;
1123 0 : rdma_req->remaining_tranfer_in_wrs = NULL;
1124 0 : rdma_req->num_outstanding_data_wr = rdma_req->num_remaining_data_wr;
1125 0 : rdma_req->num_remaining_data_wr = 0;
1126 0 : }
1127 :
1128 : static inline int
1129 0 : request_prepare_transfer_in_part(struct spdk_nvmf_request *req, uint32_t num_reads_available)
1130 : {
1131 : struct spdk_nvmf_rdma_request *rdma_req;
1132 : struct ibv_send_wr *wr;
1133 : uint32_t i;
1134 :
1135 0 : rdma_req = SPDK_CONTAINEROF(req, struct spdk_nvmf_rdma_request, req);
1136 :
1137 0 : assert(req->xfer == SPDK_NVME_DATA_HOST_TO_CONTROLLER);
1138 0 : assert(rdma_req != NULL);
1139 0 : assert(num_reads_available > 0);
1140 0 : assert(rdma_req->num_outstanding_data_wr > num_reads_available);
1141 0 : wr = rdma_req->transfer_wr;
1142 :
1143 0 : for (i = 0; i < num_reads_available - 1; i++) {
1144 0 : wr = wr->next;
1145 : }
1146 :
1147 0 : rdma_req->remaining_tranfer_in_wrs = wr->next;
1148 0 : rdma_req->num_remaining_data_wr = rdma_req->num_outstanding_data_wr - num_reads_available;
1149 0 : rdma_req->num_outstanding_data_wr = num_reads_available;
1150 : /* Break chain of WRs to send only part. Once this portion completes, we continue sending RDMA_READs */
1151 0 : wr->next = NULL;
1152 :
1153 0 : return 0;
1154 : }
1155 :
1156 : static int
1157 6 : request_transfer_out(struct spdk_nvmf_request *req, int *data_posted)
1158 : {
1159 6 : int num_outstanding_data_wr = 0;
1160 : struct spdk_nvmf_rdma_request *rdma_req;
1161 : struct spdk_nvmf_qpair *qpair;
1162 : struct spdk_nvmf_rdma_qpair *rqpair;
1163 : struct spdk_nvme_cpl *rsp;
1164 6 : struct ibv_send_wr *first = NULL;
1165 : struct spdk_nvmf_rdma_transport *rtransport;
1166 :
1167 6 : *data_posted = 0;
1168 6 : qpair = req->qpair;
1169 6 : rsp = &req->rsp->nvme_cpl;
1170 6 : rdma_req = SPDK_CONTAINEROF(req, struct spdk_nvmf_rdma_request, req);
1171 6 : rqpair = SPDK_CONTAINEROF(qpair, struct spdk_nvmf_rdma_qpair, qpair);
1172 6 : rtransport = SPDK_CONTAINEROF(rqpair->qpair.transport,
1173 : struct spdk_nvmf_rdma_transport, transport);
1174 :
1175 : /* Advance our sq_head pointer */
1176 6 : if (qpair->sq_head == qpair->sq_head_max) {
1177 6 : qpair->sq_head = 0;
1178 : } else {
1179 0 : qpair->sq_head++;
1180 : }
1181 6 : rsp->sqhd = qpair->sq_head;
1182 :
1183 : /* queue the capsule for the recv buffer */
1184 6 : assert(rdma_req->recv != NULL);
1185 :
1186 6 : nvmf_rdma_qpair_queue_recv_wrs(rqpair, &rdma_req->recv->wr);
1187 :
1188 6 : rdma_req->recv = NULL;
1189 6 : assert(rqpair->current_recv_depth > 0);
1190 6 : rqpair->current_recv_depth--;
1191 :
1192 : /* Build the response which consists of optional
1193 : * RDMA WRITEs to transfer data, plus an RDMA SEND
1194 : * containing the response.
1195 : */
1196 6 : first = &rdma_req->rsp.wr;
1197 :
1198 6 : if (spdk_unlikely(rsp->status.sc != SPDK_NVME_SC_SUCCESS)) {
1199 : /* On failure, data was not read from the controller. So clear the
1200 : * number of outstanding data WRs to zero.
1201 : */
1202 1 : rdma_req->num_outstanding_data_wr = 0;
1203 5 : } else if (req->xfer == SPDK_NVME_DATA_CONTROLLER_TO_HOST) {
1204 1 : first = rdma_req->transfer_wr;
1205 1 : *data_posted = 1;
1206 1 : num_outstanding_data_wr = rdma_req->num_outstanding_data_wr;
1207 : }
1208 6 : if (spdk_rdma_provider_qp_queue_send_wrs(rqpair->rdma_qp, first)) {
1209 6 : STAILQ_INSERT_TAIL(&rqpair->poller->qpairs_pending_send, rqpair, send_link);
1210 : }
1211 6 : if (rtransport->rdma_opts.no_wr_batching) {
1212 0 : _poller_submit_sends(rtransport, rqpair->poller);
1213 : }
1214 :
1215 : /* +1 for the rsp wr */
1216 6 : assert(rqpair->current_send_depth + num_outstanding_data_wr + 1 <= rqpair->max_send_depth);
1217 6 : rqpair->current_send_depth += num_outstanding_data_wr + 1;
1218 :
1219 6 : return 0;
1220 : }
1221 :
1222 : static int
1223 0 : nvmf_rdma_event_accept(struct rdma_cm_id *id, struct spdk_nvmf_rdma_qpair *rqpair)
1224 : {
1225 0 : struct spdk_nvmf_rdma_accept_private_data accept_data;
1226 0 : struct rdma_conn_param ctrlr_event_data = {};
1227 : int rc;
1228 :
1229 0 : accept_data.recfmt = 0;
1230 0 : accept_data.crqsize = rqpair->max_queue_depth;
1231 :
1232 0 : ctrlr_event_data.private_data = &accept_data;
1233 0 : ctrlr_event_data.private_data_len = sizeof(accept_data);
1234 0 : if (id->ps == RDMA_PS_TCP) {
1235 0 : ctrlr_event_data.responder_resources = 0; /* We accept 0 reads from the host */
1236 0 : ctrlr_event_data.initiator_depth = rqpair->max_read_depth;
1237 : }
1238 :
1239 : /* Configure infinite retries for the initiator side qpair.
1240 : * We need to pass this value to the initiator to prevent the
1241 : * initiator side NIC from completing SEND requests back to the
1242 : * initiator with status rnr_retry_count_exceeded. */
1243 0 : ctrlr_event_data.rnr_retry_count = 0x7;
1244 :
1245 : /* When qpair is created without use of rdma cm API, an additional
1246 : * information must be provided to initiator in the connection response:
1247 : * whether qpair is using SRQ and its qp_num
1248 : * Fields below are ignored by rdma cm if qpair has been
1249 : * created using rdma cm API. */
1250 0 : ctrlr_event_data.srq = rqpair->srq ? 1 : 0;
1251 0 : ctrlr_event_data.qp_num = rqpair->qp_num;
1252 :
1253 0 : rc = spdk_rdma_provider_qp_accept(rqpair->rdma_qp, &ctrlr_event_data);
1254 0 : if (rc) {
1255 0 : SPDK_ERRLOG("Error %d on spdk_rdma_provider_qp_accept\n", errno);
1256 : } else {
1257 0 : SPDK_DEBUGLOG(rdma, "Sent back the accept\n");
1258 : }
1259 :
1260 0 : return rc;
1261 : }
1262 :
1263 : static void
1264 0 : nvmf_rdma_event_reject(struct rdma_cm_id *id, enum spdk_nvmf_rdma_transport_error error)
1265 : {
1266 0 : struct spdk_nvmf_rdma_reject_private_data rej_data;
1267 :
1268 0 : rej_data.recfmt = 0;
1269 0 : rej_data.sts = error;
1270 :
1271 0 : rdma_reject(id, &rej_data, sizeof(rej_data));
1272 0 : }
1273 :
1274 : static int
1275 0 : nvmf_rdma_connect(struct spdk_nvmf_transport *transport, struct rdma_cm_event *event)
1276 : {
1277 : struct spdk_nvmf_rdma_transport *rtransport;
1278 0 : struct spdk_nvmf_rdma_qpair *rqpair = NULL;
1279 : struct spdk_nvmf_rdma_port *port;
1280 0 : struct rdma_conn_param *rdma_param = NULL;
1281 0 : const struct spdk_nvmf_rdma_request_private_data *private_data = NULL;
1282 : uint16_t max_queue_depth;
1283 : uint16_t max_read_depth;
1284 :
1285 0 : rtransport = SPDK_CONTAINEROF(transport, struct spdk_nvmf_rdma_transport, transport);
1286 :
1287 0 : assert(event->id != NULL); /* Impossible. Can't even reject the connection. */
1288 0 : assert(event->id->verbs != NULL); /* Impossible. No way to handle this. */
1289 :
1290 0 : rdma_param = &event->param.conn;
1291 0 : if (rdma_param->private_data == NULL ||
1292 0 : rdma_param->private_data_len < sizeof(struct spdk_nvmf_rdma_request_private_data)) {
1293 0 : SPDK_ERRLOG("connect request: no private data provided\n");
1294 0 : nvmf_rdma_event_reject(event->id, SPDK_NVMF_RDMA_ERROR_INVALID_PRIVATE_DATA_LENGTH);
1295 0 : return -1;
1296 : }
1297 :
1298 0 : private_data = rdma_param->private_data;
1299 0 : if (private_data->recfmt != 0) {
1300 0 : SPDK_ERRLOG("Received RDMA private data with RECFMT != 0\n");
1301 0 : nvmf_rdma_event_reject(event->id, SPDK_NVMF_RDMA_ERROR_INVALID_RECFMT);
1302 0 : return -1;
1303 : }
1304 :
1305 0 : SPDK_DEBUGLOG(rdma, "Connect Recv on fabric intf name %s, dev_name %s\n",
1306 : event->id->verbs->device->name, event->id->verbs->device->dev_name);
1307 :
1308 0 : port = event->listen_id->context;
1309 0 : SPDK_DEBUGLOG(rdma, "Listen Id was %p with verbs %p. ListenAddr: %p\n",
1310 : event->listen_id, event->listen_id->verbs, port);
1311 :
1312 : /* Figure out the supported queue depth. This is a multi-step process
1313 : * that takes into account hardware maximums, host provided values,
1314 : * and our target's internal memory limits */
1315 :
1316 0 : SPDK_DEBUGLOG(rdma, "Calculating Queue Depth\n");
1317 :
1318 : /* Start with the maximum queue depth allowed by the target */
1319 0 : max_queue_depth = rtransport->transport.opts.max_queue_depth;
1320 0 : max_read_depth = rtransport->transport.opts.max_queue_depth;
1321 0 : SPDK_DEBUGLOG(rdma, "Target Max Queue Depth: %d\n",
1322 : rtransport->transport.opts.max_queue_depth);
1323 :
1324 : /* Next check the local NIC's hardware limitations */
1325 0 : SPDK_DEBUGLOG(rdma,
1326 : "Local NIC Max Send/Recv Queue Depth: %d Max Read/Write Queue Depth: %d\n",
1327 : port->device->attr.max_qp_wr, port->device->attr.max_qp_rd_atom);
1328 0 : max_queue_depth = spdk_min(max_queue_depth, port->device->attr.max_qp_wr);
1329 0 : max_read_depth = spdk_min(max_read_depth, port->device->attr.max_qp_init_rd_atom);
1330 :
1331 : /* Next check the remote NIC's hardware limitations */
1332 0 : SPDK_DEBUGLOG(rdma,
1333 : "Host (Initiator) NIC Max Incoming RDMA R/W operations: %d Max Outgoing RDMA R/W operations: %d\n",
1334 : rdma_param->initiator_depth, rdma_param->responder_resources);
1335 : /* from man3 rdma_get_cm_event
1336 : * responder_resources - Specifies the number of responder resources that is requested by the recipient.
1337 : * The responder_resources field must match the initiator depth specified by the remote node when running
1338 : * the rdma_connect and rdma_accept functions. */
1339 0 : if (rdma_param->responder_resources != 0) {
1340 0 : if (private_data->qid) {
1341 0 : SPDK_DEBUGLOG(rdma, "Host (Initiator) is not allowed to use RDMA operations,"
1342 : " responder_resources must be 0 but set to %u\n",
1343 : rdma_param->responder_resources);
1344 : } else {
1345 0 : SPDK_WARNLOG("Host (Initiator) is not allowed to use RDMA operations,"
1346 : " responder_resources must be 0 but set to %u\n",
1347 : rdma_param->responder_resources);
1348 : }
1349 : }
1350 : /* from man3 rdma_get_cm_event
1351 : * initiator_depth - Specifies the maximum number of outstanding RDMA read operations that the recipient holds.
1352 : * The initiator_depth field must match the responder resources specified by the remote node when running
1353 : * the rdma_connect and rdma_accept functions. */
1354 0 : if (rdma_param->initiator_depth == 0) {
1355 0 : SPDK_ERRLOG("Host (Initiator) doesn't support RDMA_READ or atomic operations\n");
1356 0 : nvmf_rdma_event_reject(event->id, SPDK_NVMF_RDMA_ERROR_INVALID_IRD);
1357 0 : return -1;
1358 : }
1359 0 : max_read_depth = spdk_min(max_read_depth, rdma_param->initiator_depth);
1360 :
1361 0 : SPDK_DEBUGLOG(rdma, "Host Receive Queue Size: %d\n", private_data->hrqsize);
1362 0 : SPDK_DEBUGLOG(rdma, "Host Send Queue Size: %d\n", private_data->hsqsize);
1363 0 : max_queue_depth = spdk_min(max_queue_depth, private_data->hrqsize);
1364 0 : max_queue_depth = spdk_min(max_queue_depth, private_data->hsqsize + 1);
1365 :
1366 0 : SPDK_DEBUGLOG(rdma, "Final Negotiated Queue Depth: %d R/W Depth: %d\n",
1367 : max_queue_depth, max_read_depth);
1368 :
1369 0 : rqpair = calloc(1, sizeof(struct spdk_nvmf_rdma_qpair));
1370 0 : if (rqpair == NULL) {
1371 0 : SPDK_ERRLOG("Could not allocate new connection.\n");
1372 0 : nvmf_rdma_event_reject(event->id, SPDK_NVMF_RDMA_ERROR_NO_RESOURCES);
1373 0 : return -1;
1374 : }
1375 :
1376 0 : rqpair->device = port->device;
1377 0 : rqpair->max_queue_depth = max_queue_depth;
1378 0 : rqpair->max_read_depth = max_read_depth;
1379 0 : rqpair->cm_id = event->id;
1380 0 : rqpair->listen_id = event->listen_id;
1381 0 : rqpair->qpair.transport = transport;
1382 0 : STAILQ_INIT(&rqpair->ibv_events);
1383 : /* use qid from the private data to determine the qpair type
1384 : qid will be set to the appropriate value when the controller is created */
1385 0 : rqpair->qpair.qid = private_data->qid;
1386 0 : rqpair->qpair.numa.id_valid = 1;
1387 0 : rqpair->qpair.numa.id = spdk_rdma_cm_id_get_numa_id(rqpair->cm_id);
1388 :
1389 0 : event->id->context = &rqpair->qpair;
1390 :
1391 0 : spdk_nvmf_tgt_new_qpair(transport->tgt, &rqpair->qpair);
1392 :
1393 0 : return 0;
1394 : }
1395 :
1396 : static inline void
1397 28 : nvmf_rdma_setup_wr(struct ibv_send_wr *wr, struct ibv_send_wr *next,
1398 : enum spdk_nvme_data_transfer xfer)
1399 : {
1400 28 : if (xfer == SPDK_NVME_DATA_CONTROLLER_TO_HOST) {
1401 24 : wr->opcode = IBV_WR_RDMA_WRITE;
1402 24 : wr->send_flags = 0;
1403 24 : wr->next = next;
1404 4 : } else if (xfer == SPDK_NVME_DATA_HOST_TO_CONTROLLER) {
1405 4 : wr->opcode = IBV_WR_RDMA_READ;
1406 4 : wr->send_flags = IBV_SEND_SIGNALED;
1407 4 : wr->next = NULL;
1408 : } else {
1409 0 : assert(0);
1410 : }
1411 28 : }
1412 :
1413 : static int
1414 6 : nvmf_request_alloc_wrs(struct spdk_nvmf_rdma_transport *rtransport,
1415 : struct spdk_nvmf_rdma_request *rdma_req,
1416 : uint32_t num_sgl_descriptors)
1417 : {
1418 6 : struct spdk_nvmf_rdma_request_data *work_requests[SPDK_NVMF_MAX_SGL_ENTRIES];
1419 : struct spdk_nvmf_rdma_request_data *current_data_wr;
1420 : uint32_t i;
1421 :
1422 6 : if (spdk_unlikely(num_sgl_descriptors > SPDK_NVMF_MAX_SGL_ENTRIES)) {
1423 0 : SPDK_ERRLOG("Requested too much entries (%u), the limit is %u\n",
1424 : num_sgl_descriptors, SPDK_NVMF_MAX_SGL_ENTRIES);
1425 0 : return -EINVAL;
1426 : }
1427 :
1428 6 : if (spdk_unlikely(spdk_mempool_get_bulk(rtransport->data_wr_pool, (void **)work_requests,
1429 : num_sgl_descriptors))) {
1430 0 : return -ENOMEM;
1431 : }
1432 :
1433 6 : current_data_wr = &rdma_req->data;
1434 :
1435 12 : for (i = 0; i < num_sgl_descriptors; i++) {
1436 6 : nvmf_rdma_setup_wr(¤t_data_wr->wr, &work_requests[i]->wr, rdma_req->req.xfer);
1437 6 : current_data_wr->wr.next = &work_requests[i]->wr;
1438 6 : current_data_wr = work_requests[i];
1439 6 : current_data_wr->wr.sg_list = current_data_wr->sgl;
1440 6 : current_data_wr->wr.wr_id = rdma_req->data.wr.wr_id;
1441 : }
1442 :
1443 6 : nvmf_rdma_setup_wr(¤t_data_wr->wr, &rdma_req->rsp.wr, rdma_req->req.xfer);
1444 :
1445 6 : return 0;
1446 : }
1447 :
1448 : static inline void
1449 16 : nvmf_rdma_setup_request(struct spdk_nvmf_rdma_request *rdma_req)
1450 : {
1451 16 : struct ibv_send_wr *wr = &rdma_req->data.wr;
1452 16 : struct spdk_nvme_sgl_descriptor *sgl = &rdma_req->req.cmd->nvme_cmd.dptr.sgl1;
1453 :
1454 16 : wr->wr.rdma.rkey = sgl->keyed.key;
1455 16 : wr->wr.rdma.remote_addr = sgl->address;
1456 16 : nvmf_rdma_setup_wr(wr, &rdma_req->rsp.wr, rdma_req->req.xfer);
1457 16 : }
1458 :
1459 : static inline void
1460 1 : nvmf_rdma_update_remote_addr(struct spdk_nvmf_rdma_request *rdma_req, uint32_t num_wrs)
1461 : {
1462 1 : struct ibv_send_wr *wr = &rdma_req->data.wr;
1463 1 : struct spdk_nvme_sgl_descriptor *sgl = &rdma_req->req.cmd->nvme_cmd.dptr.sgl1;
1464 : uint32_t i;
1465 : int j;
1466 1 : uint64_t remote_addr_offset = 0;
1467 :
1468 3 : for (i = 0; i < num_wrs; ++i) {
1469 2 : wr->wr.rdma.rkey = sgl->keyed.key;
1470 2 : wr->wr.rdma.remote_addr = sgl->address + remote_addr_offset;
1471 19 : for (j = 0; j < wr->num_sge; ++j) {
1472 17 : remote_addr_offset += wr->sg_list[j].length;
1473 : }
1474 2 : wr = wr->next;
1475 : }
1476 1 : }
1477 :
1478 : static int
1479 15 : nvmf_rdma_fill_wr_sgl(struct spdk_nvmf_rdma_device *device,
1480 : struct spdk_nvmf_rdma_request *rdma_req,
1481 : struct ibv_send_wr *wr,
1482 : uint32_t total_length)
1483 : {
1484 15 : struct spdk_rdma_utils_memory_translation mem_translation;
1485 : struct ibv_sge *sg_ele;
1486 : struct iovec *iov;
1487 : uint32_t lkey, remaining;
1488 : int rc;
1489 :
1490 15 : wr->num_sge = 0;
1491 :
1492 74 : while (total_length && wr->num_sge < SPDK_NVMF_MAX_SGL_ENTRIES) {
1493 59 : iov = &rdma_req->req.iov[rdma_req->iovpos];
1494 59 : rc = spdk_rdma_utils_get_translation(device->map, iov->iov_base, iov->iov_len, &mem_translation);
1495 59 : if (spdk_unlikely(rc)) {
1496 0 : return rc;
1497 : }
1498 :
1499 59 : lkey = spdk_rdma_utils_memory_translation_get_lkey(&mem_translation);
1500 59 : sg_ele = &wr->sg_list[wr->num_sge];
1501 59 : remaining = spdk_min((uint32_t)iov->iov_len - rdma_req->offset, total_length);
1502 :
1503 59 : sg_ele->lkey = lkey;
1504 59 : sg_ele->addr = (uintptr_t)iov->iov_base + rdma_req->offset;
1505 59 : sg_ele->length = remaining;
1506 59 : SPDK_DEBUGLOG(rdma, "sge[%d] %p addr 0x%"PRIx64", len %u\n", wr->num_sge, sg_ele, sg_ele->addr,
1507 : sg_ele->length);
1508 59 : rdma_req->offset += sg_ele->length;
1509 59 : total_length -= sg_ele->length;
1510 59 : wr->num_sge++;
1511 :
1512 59 : if (rdma_req->offset == iov->iov_len) {
1513 57 : rdma_req->offset = 0;
1514 57 : rdma_req->iovpos++;
1515 : }
1516 : }
1517 :
1518 15 : if (spdk_unlikely(total_length)) {
1519 0 : SPDK_ERRLOG("Not enough SG entries to hold data buffer\n");
1520 0 : return -EINVAL;
1521 : }
1522 :
1523 15 : return 0;
1524 : }
1525 :
1526 : static int
1527 10 : nvmf_rdma_fill_wr_sgl_with_dif(struct spdk_nvmf_rdma_device *device,
1528 : struct spdk_nvmf_rdma_request *rdma_req,
1529 : struct ibv_send_wr *wr,
1530 : uint32_t total_length,
1531 : uint32_t num_extra_wrs)
1532 : {
1533 10 : struct spdk_rdma_utils_memory_translation mem_translation;
1534 10 : struct spdk_dif_ctx *dif_ctx = &rdma_req->req.dif.dif_ctx;
1535 : struct ibv_sge *sg_ele;
1536 : struct iovec *iov;
1537 : struct iovec *rdma_iov;
1538 : uint32_t lkey, remaining;
1539 : uint32_t remaining_data_block, data_block_size, md_size;
1540 : uint32_t sge_len;
1541 : int rc;
1542 :
1543 10 : data_block_size = dif_ctx->block_size - dif_ctx->md_size;
1544 :
1545 10 : if (spdk_likely(!rdma_req->req.stripped_data)) {
1546 5 : rdma_iov = rdma_req->req.iov;
1547 5 : remaining_data_block = data_block_size;
1548 5 : md_size = dif_ctx->md_size;
1549 : } else {
1550 5 : rdma_iov = rdma_req->req.stripped_data->iov;
1551 5 : total_length = total_length / dif_ctx->block_size * data_block_size;
1552 5 : remaining_data_block = total_length;
1553 5 : md_size = 0;
1554 : }
1555 :
1556 10 : wr->num_sge = 0;
1557 :
1558 25 : while (total_length && (num_extra_wrs || wr->num_sge < SPDK_NVMF_MAX_SGL_ENTRIES)) {
1559 15 : iov = rdma_iov + rdma_req->iovpos;
1560 15 : rc = spdk_rdma_utils_get_translation(device->map, iov->iov_base, iov->iov_len, &mem_translation);
1561 15 : if (spdk_unlikely(rc)) {
1562 0 : return rc;
1563 : }
1564 :
1565 15 : lkey = spdk_rdma_utils_memory_translation_get_lkey(&mem_translation);
1566 15 : sg_ele = &wr->sg_list[wr->num_sge];
1567 15 : remaining = spdk_min((uint32_t)iov->iov_len - rdma_req->offset, total_length);
1568 :
1569 53 : while (remaining) {
1570 38 : if (wr->num_sge >= SPDK_NVMF_MAX_SGL_ENTRIES) {
1571 1 : if (num_extra_wrs > 0 && wr->next) {
1572 1 : wr = wr->next;
1573 1 : wr->num_sge = 0;
1574 1 : sg_ele = &wr->sg_list[wr->num_sge];
1575 1 : num_extra_wrs--;
1576 : } else {
1577 : break;
1578 : }
1579 : }
1580 38 : sg_ele->lkey = lkey;
1581 38 : sg_ele->addr = (uintptr_t)((char *)iov->iov_base + rdma_req->offset);
1582 38 : sge_len = spdk_min(remaining, remaining_data_block);
1583 38 : sg_ele->length = sge_len;
1584 38 : SPDK_DEBUGLOG(rdma, "sge[%d] %p addr 0x%"PRIx64", len %u\n", wr->num_sge, sg_ele,
1585 : sg_ele->addr, sg_ele->length);
1586 38 : remaining -= sge_len;
1587 38 : remaining_data_block -= sge_len;
1588 38 : rdma_req->offset += sge_len;
1589 38 : total_length -= sge_len;
1590 :
1591 38 : sg_ele++;
1592 38 : wr->num_sge++;
1593 :
1594 38 : if (remaining_data_block == 0) {
1595 : /* skip metadata */
1596 34 : rdma_req->offset += md_size;
1597 34 : total_length -= md_size;
1598 : /* Metadata that do not fit this IO buffer will be included in the next IO buffer */
1599 34 : remaining -= spdk_min(remaining, md_size);
1600 34 : remaining_data_block = data_block_size;
1601 : }
1602 :
1603 38 : if (remaining == 0) {
1604 : /* By subtracting the size of the last IOV from the offset, we ensure that we skip
1605 : the remaining metadata bits at the beginning of the next buffer */
1606 15 : rdma_req->offset -= spdk_min(iov->iov_len, rdma_req->offset);
1607 15 : rdma_req->iovpos++;
1608 : }
1609 : }
1610 : }
1611 :
1612 10 : if (spdk_unlikely(total_length)) {
1613 0 : SPDK_ERRLOG("Not enough SG entries to hold data buffer\n");
1614 0 : return -EINVAL;
1615 : }
1616 :
1617 10 : return 0;
1618 : }
1619 :
1620 : static inline uint32_t
1621 8 : nvmf_rdma_calc_num_wrs(uint32_t length, uint32_t io_unit_size, uint32_t block_size)
1622 : {
1623 : /* estimate the number of SG entries and WRs needed to process the request */
1624 8 : uint32_t num_sge = 0;
1625 : uint32_t i;
1626 8 : uint32_t num_buffers = SPDK_CEIL_DIV(length, io_unit_size);
1627 :
1628 23 : for (i = 0; i < num_buffers && length > 0; i++) {
1629 15 : uint32_t buffer_len = spdk_min(length, io_unit_size);
1630 15 : uint32_t num_sge_in_block = SPDK_CEIL_DIV(buffer_len, block_size);
1631 :
1632 15 : if (num_sge_in_block * block_size > buffer_len) {
1633 11 : ++num_sge_in_block;
1634 : }
1635 15 : num_sge += num_sge_in_block;
1636 15 : length -= buffer_len;
1637 : }
1638 8 : return SPDK_CEIL_DIV(num_sge, SPDK_NVMF_MAX_SGL_ENTRIES);
1639 : }
1640 :
1641 : static int
1642 16 : nvmf_rdma_request_fill_iovs(struct spdk_nvmf_rdma_transport *rtransport,
1643 : struct spdk_nvmf_rdma_device *device,
1644 : struct spdk_nvmf_rdma_request *rdma_req)
1645 : {
1646 : struct spdk_nvmf_rdma_qpair *rqpair;
1647 : struct spdk_nvmf_rdma_poll_group *rgroup;
1648 16 : struct spdk_nvmf_request *req = &rdma_req->req;
1649 16 : struct ibv_send_wr *wr = &rdma_req->data.wr;
1650 : int rc;
1651 16 : uint32_t num_wrs = 1;
1652 : uint32_t length;
1653 :
1654 16 : rqpair = SPDK_CONTAINEROF(req->qpair, struct spdk_nvmf_rdma_qpair, qpair);
1655 16 : rgroup = rqpair->poller->group;
1656 :
1657 : /* rdma wr specifics */
1658 16 : nvmf_rdma_setup_request(rdma_req);
1659 :
1660 16 : length = req->length;
1661 16 : if (spdk_unlikely(req->dif_enabled)) {
1662 8 : req->dif.orig_length = length;
1663 8 : length = spdk_dif_get_length_with_md(length, &req->dif.dif_ctx);
1664 8 : req->dif.elba_length = length;
1665 : }
1666 :
1667 16 : rc = spdk_nvmf_request_get_buffers(req, &rgroup->group, &rtransport->transport,
1668 : length);
1669 16 : if (spdk_unlikely(rc != 0)) {
1670 1 : return rc;
1671 : }
1672 :
1673 15 : assert(req->iovcnt <= rqpair->max_send_sge);
1674 :
1675 : /* When dif_insert_or_strip is true and the I/O data length is greater than one block,
1676 : * the stripped_buffers are got for DIF stripping. */
1677 15 : if (spdk_unlikely(req->dif_enabled && (req->xfer == SPDK_NVME_DATA_CONTROLLER_TO_HOST)
1678 : && (req->dif.elba_length > req->dif.dif_ctx.block_size))) {
1679 7 : rc = nvmf_request_get_stripped_buffers(req, &rgroup->group,
1680 : &rtransport->transport, req->dif.orig_length);
1681 7 : if (rc != 0) {
1682 4 : SPDK_INFOLOG(rdma, "Get stripped buffers fail %d, fallback to req.iov.\n", rc);
1683 : }
1684 : }
1685 :
1686 15 : rdma_req->iovpos = 0;
1687 :
1688 15 : if (spdk_unlikely(req->dif_enabled)) {
1689 8 : num_wrs = nvmf_rdma_calc_num_wrs(length, rtransport->transport.opts.io_unit_size,
1690 : req->dif.dif_ctx.block_size);
1691 8 : if (num_wrs > 1) {
1692 1 : rc = nvmf_request_alloc_wrs(rtransport, rdma_req, num_wrs - 1);
1693 1 : if (spdk_unlikely(rc != 0)) {
1694 0 : goto err_exit;
1695 : }
1696 : }
1697 :
1698 8 : rc = nvmf_rdma_fill_wr_sgl_with_dif(device, rdma_req, wr, length, num_wrs - 1);
1699 8 : if (spdk_unlikely(rc != 0)) {
1700 0 : goto err_exit;
1701 : }
1702 :
1703 8 : if (num_wrs > 1) {
1704 1 : nvmf_rdma_update_remote_addr(rdma_req, num_wrs);
1705 : }
1706 : } else {
1707 7 : rc = nvmf_rdma_fill_wr_sgl(device, rdma_req, wr, length);
1708 7 : if (spdk_unlikely(rc != 0)) {
1709 0 : goto err_exit;
1710 : }
1711 : }
1712 :
1713 : /* set the number of outstanding data WRs for this request. */
1714 15 : rdma_req->num_outstanding_data_wr = num_wrs;
1715 :
1716 15 : return rc;
1717 :
1718 0 : err_exit:
1719 0 : spdk_nvmf_request_free_buffers(req, &rgroup->group, &rtransport->transport);
1720 0 : nvmf_rdma_request_free_data(rdma_req, rtransport);
1721 0 : req->iovcnt = 0;
1722 0 : return rc;
1723 : }
1724 :
1725 : static int
1726 5 : nvmf_rdma_request_fill_iovs_multi_sgl(struct spdk_nvmf_rdma_transport *rtransport,
1727 : struct spdk_nvmf_rdma_device *device,
1728 : struct spdk_nvmf_rdma_request *rdma_req)
1729 : {
1730 : struct spdk_nvmf_rdma_qpair *rqpair;
1731 : struct spdk_nvmf_rdma_poll_group *rgroup;
1732 : struct ibv_send_wr *current_wr;
1733 5 : struct spdk_nvmf_request *req = &rdma_req->req;
1734 : struct spdk_nvme_sgl_descriptor *inline_segment, *desc;
1735 : uint32_t num_sgl_descriptors;
1736 5 : uint32_t lengths[SPDK_NVMF_MAX_SGL_ENTRIES], total_length = 0;
1737 : uint32_t i;
1738 : int rc;
1739 :
1740 5 : rqpair = SPDK_CONTAINEROF(rdma_req->req.qpair, struct spdk_nvmf_rdma_qpair, qpair);
1741 5 : rgroup = rqpair->poller->group;
1742 :
1743 5 : inline_segment = &req->cmd->nvme_cmd.dptr.sgl1;
1744 5 : assert(inline_segment->generic.type == SPDK_NVME_SGL_TYPE_LAST_SEGMENT);
1745 5 : assert(inline_segment->unkeyed.subtype == SPDK_NVME_SGL_SUBTYPE_OFFSET);
1746 :
1747 5 : num_sgl_descriptors = inline_segment->unkeyed.length / sizeof(struct spdk_nvme_sgl_descriptor);
1748 5 : assert(num_sgl_descriptors <= SPDK_NVMF_MAX_SGL_ENTRIES);
1749 :
1750 5 : desc = (struct spdk_nvme_sgl_descriptor *)rdma_req->recv->buf + inline_segment->address;
1751 15 : for (i = 0; i < num_sgl_descriptors; i++) {
1752 10 : if (spdk_likely(!req->dif_enabled)) {
1753 8 : lengths[i] = desc->keyed.length;
1754 : } else {
1755 2 : req->dif.orig_length += desc->keyed.length;
1756 2 : lengths[i] = spdk_dif_get_length_with_md(desc->keyed.length, &req->dif.dif_ctx);
1757 2 : req->dif.elba_length += lengths[i];
1758 : }
1759 10 : total_length += lengths[i];
1760 10 : desc++;
1761 : }
1762 :
1763 5 : if (spdk_unlikely(total_length > rtransport->transport.opts.max_io_size)) {
1764 0 : SPDK_ERRLOG("Multi SGL length 0x%x exceeds max io size 0x%x\n",
1765 : total_length, rtransport->transport.opts.max_io_size);
1766 0 : req->rsp->nvme_cpl.status.sc = SPDK_NVME_SC_DATA_SGL_LENGTH_INVALID;
1767 0 : return -EINVAL;
1768 : }
1769 :
1770 5 : rc = nvmf_request_alloc_wrs(rtransport, rdma_req, num_sgl_descriptors - 1);
1771 5 : if (spdk_unlikely(rc != 0)) {
1772 0 : return -ENOMEM;
1773 : }
1774 :
1775 5 : rc = spdk_nvmf_request_get_buffers(req, &rgroup->group, &rtransport->transport, total_length);
1776 5 : if (spdk_unlikely(rc != 0)) {
1777 0 : nvmf_rdma_request_free_data(rdma_req, rtransport);
1778 0 : return rc;
1779 : }
1780 :
1781 : /* When dif_insert_or_strip is true and the I/O data length is greater than one block,
1782 : * the stripped_buffers are got for DIF stripping. */
1783 5 : if (spdk_unlikely(req->dif_enabled && (req->xfer == SPDK_NVME_DATA_CONTROLLER_TO_HOST)
1784 : && (req->dif.elba_length > req->dif.dif_ctx.block_size))) {
1785 1 : rc = nvmf_request_get_stripped_buffers(req, &rgroup->group,
1786 : &rtransport->transport, req->dif.orig_length);
1787 1 : if (spdk_unlikely(rc != 0)) {
1788 0 : SPDK_INFOLOG(rdma, "Get stripped buffers fail %d, fallback to req.iov.\n", rc);
1789 : }
1790 : }
1791 :
1792 : /* The first WR must always be the embedded data WR. This is how we unwind them later. */
1793 5 : current_wr = &rdma_req->data.wr;
1794 5 : assert(current_wr != NULL);
1795 :
1796 5 : req->length = 0;
1797 5 : rdma_req->iovpos = 0;
1798 5 : desc = (struct spdk_nvme_sgl_descriptor *)rdma_req->recv->buf + inline_segment->address;
1799 15 : for (i = 0; i < num_sgl_descriptors; i++) {
1800 : /* The descriptors must be keyed data block descriptors with an address, not an offset. */
1801 10 : if (spdk_unlikely(desc->generic.type != SPDK_NVME_SGL_TYPE_KEYED_DATA_BLOCK ||
1802 : desc->keyed.subtype != SPDK_NVME_SGL_SUBTYPE_ADDRESS)) {
1803 0 : rc = -EINVAL;
1804 0 : goto err_exit;
1805 : }
1806 :
1807 10 : if (spdk_likely(!req->dif_enabled)) {
1808 8 : rc = nvmf_rdma_fill_wr_sgl(device, rdma_req, current_wr, lengths[i]);
1809 : } else {
1810 2 : rc = nvmf_rdma_fill_wr_sgl_with_dif(device, rdma_req, current_wr,
1811 : lengths[i], 0);
1812 : }
1813 10 : if (spdk_unlikely(rc != 0)) {
1814 0 : rc = -ENOMEM;
1815 0 : goto err_exit;
1816 : }
1817 :
1818 10 : req->length += desc->keyed.length;
1819 10 : current_wr->wr.rdma.rkey = desc->keyed.key;
1820 10 : current_wr->wr.rdma.remote_addr = desc->address;
1821 10 : current_wr = current_wr->next;
1822 10 : desc++;
1823 : }
1824 :
1825 : #ifdef SPDK_CONFIG_RDMA_SEND_WITH_INVAL
1826 : /* Go back to the last descriptor in the list. */
1827 5 : desc--;
1828 5 : if ((device->attr.device_cap_flags & IBV_DEVICE_MEM_MGT_EXTENSIONS) != 0) {
1829 0 : if (desc->keyed.subtype == SPDK_NVME_SGL_SUBTYPE_INVALIDATE_KEY) {
1830 0 : rdma_req->rsp.wr.opcode = IBV_WR_SEND_WITH_INV;
1831 0 : rdma_req->rsp.wr.imm_data = desc->keyed.key;
1832 : }
1833 : }
1834 : #endif
1835 :
1836 5 : rdma_req->num_outstanding_data_wr = num_sgl_descriptors;
1837 :
1838 5 : return 0;
1839 :
1840 0 : err_exit:
1841 0 : spdk_nvmf_request_free_buffers(req, &rgroup->group, &rtransport->transport);
1842 0 : nvmf_rdma_request_free_data(rdma_req, rtransport);
1843 0 : return rc;
1844 : }
1845 :
1846 : static int
1847 25 : nvmf_rdma_request_parse_sgl(struct spdk_nvmf_rdma_transport *rtransport,
1848 : struct spdk_nvmf_rdma_device *device,
1849 : struct spdk_nvmf_rdma_request *rdma_req)
1850 : {
1851 25 : struct spdk_nvmf_request *req = &rdma_req->req;
1852 : struct spdk_nvme_cpl *rsp;
1853 : struct spdk_nvme_sgl_descriptor *sgl;
1854 : int rc;
1855 : uint32_t length;
1856 :
1857 25 : rsp = &req->rsp->nvme_cpl;
1858 25 : sgl = &req->cmd->nvme_cmd.dptr.sgl1;
1859 :
1860 25 : if (sgl->generic.type == SPDK_NVME_SGL_TYPE_KEYED_DATA_BLOCK &&
1861 17 : (sgl->keyed.subtype == SPDK_NVME_SGL_SUBTYPE_ADDRESS ||
1862 0 : sgl->keyed.subtype == SPDK_NVME_SGL_SUBTYPE_INVALIDATE_KEY)) {
1863 :
1864 17 : length = sgl->keyed.length;
1865 17 : if (spdk_unlikely(length > rtransport->transport.opts.max_io_size)) {
1866 1 : SPDK_ERRLOG("SGL length 0x%x exceeds max io size 0x%x\n",
1867 : length, rtransport->transport.opts.max_io_size);
1868 1 : rsp->status.sc = SPDK_NVME_SC_DATA_SGL_LENGTH_INVALID;
1869 1 : return -1;
1870 : }
1871 : #ifdef SPDK_CONFIG_RDMA_SEND_WITH_INVAL
1872 16 : if ((device->attr.device_cap_flags & IBV_DEVICE_MEM_MGT_EXTENSIONS) != 0) {
1873 0 : if (sgl->keyed.subtype == SPDK_NVME_SGL_SUBTYPE_INVALIDATE_KEY) {
1874 0 : rdma_req->rsp.wr.opcode = IBV_WR_SEND_WITH_INV;
1875 0 : rdma_req->rsp.wr.imm_data = sgl->keyed.key;
1876 : }
1877 : }
1878 : #endif
1879 :
1880 : /* fill request length and populate iovs */
1881 16 : req->length = length;
1882 :
1883 16 : rc = nvmf_rdma_request_fill_iovs(rtransport, device, rdma_req);
1884 16 : if (spdk_unlikely(rc < 0)) {
1885 1 : if (rc == -EINVAL) {
1886 0 : SPDK_ERRLOG("SGL length exceeds the max I/O size\n");
1887 0 : rsp->status.sc = SPDK_NVME_SC_DATA_SGL_LENGTH_INVALID;
1888 0 : return -1;
1889 : }
1890 : /* No available buffers. Queue this request up. */
1891 1 : SPDK_DEBUGLOG(rdma, "No available large data buffers. Queueing request %p\n", rdma_req);
1892 1 : return 0;
1893 : }
1894 :
1895 15 : SPDK_DEBUGLOG(rdma, "Request %p took %d buffer/s from central pool\n", rdma_req,
1896 : req->iovcnt);
1897 :
1898 15 : return 0;
1899 8 : } else if (sgl->generic.type == SPDK_NVME_SGL_TYPE_DATA_BLOCK &&
1900 3 : sgl->unkeyed.subtype == SPDK_NVME_SGL_SUBTYPE_OFFSET) {
1901 3 : uint64_t offset = sgl->address;
1902 3 : uint32_t max_len = rtransport->transport.opts.in_capsule_data_size;
1903 :
1904 3 : SPDK_DEBUGLOG(nvmf, "In-capsule data: offset 0x%" PRIx64 ", length 0x%x\n",
1905 : offset, sgl->unkeyed.length);
1906 :
1907 3 : if (spdk_unlikely(offset > max_len)) {
1908 0 : SPDK_ERRLOG("In-capsule offset 0x%" PRIx64 " exceeds capsule length 0x%x\n",
1909 : offset, max_len);
1910 0 : rsp->status.sc = SPDK_NVME_SC_INVALID_SGL_OFFSET;
1911 0 : return -1;
1912 : }
1913 3 : max_len -= (uint32_t)offset;
1914 :
1915 3 : if (spdk_unlikely(sgl->unkeyed.length > max_len)) {
1916 2 : SPDK_ERRLOG("In-capsule data length 0x%x exceeds capsule length 0x%x\n",
1917 : sgl->unkeyed.length, max_len);
1918 2 : rsp->status.sc = SPDK_NVME_SC_DATA_SGL_LENGTH_INVALID;
1919 2 : return -1;
1920 : }
1921 :
1922 1 : rdma_req->num_outstanding_data_wr = 0;
1923 1 : req->data_from_pool = false;
1924 1 : req->length = sgl->unkeyed.length;
1925 :
1926 1 : req->iov[0].iov_base = rdma_req->recv->buf + offset;
1927 1 : req->iov[0].iov_len = req->length;
1928 1 : req->iovcnt = 1;
1929 :
1930 1 : return 0;
1931 5 : } else if (sgl->generic.type == SPDK_NVME_SGL_TYPE_LAST_SEGMENT &&
1932 5 : sgl->unkeyed.subtype == SPDK_NVME_SGL_SUBTYPE_OFFSET) {
1933 :
1934 5 : rc = nvmf_rdma_request_fill_iovs_multi_sgl(rtransport, device, rdma_req);
1935 5 : if (spdk_unlikely(rc == -ENOMEM)) {
1936 0 : SPDK_DEBUGLOG(rdma, "No available large data buffers. Queueing request %p\n", rdma_req);
1937 0 : return 0;
1938 5 : } else if (spdk_unlikely(rc == -EINVAL)) {
1939 0 : SPDK_ERRLOG("Multi SGL element request length exceeds the max I/O size\n");
1940 0 : rsp->status.sc = SPDK_NVME_SC_DATA_SGL_LENGTH_INVALID;
1941 0 : return -1;
1942 : }
1943 :
1944 5 : SPDK_DEBUGLOG(rdma, "Request %p took %d buffer/s from central pool\n", rdma_req,
1945 : req->iovcnt);
1946 :
1947 5 : return 0;
1948 : }
1949 :
1950 0 : SPDK_ERRLOG("Invalid NVMf I/O Command SGL: Type 0x%x, Subtype 0x%x\n",
1951 : sgl->generic.type, sgl->generic.subtype);
1952 0 : rsp->status.sc = SPDK_NVME_SC_SGL_DESCRIPTOR_TYPE_INVALID;
1953 0 : return -1;
1954 : }
1955 :
1956 : static void
1957 6 : _nvmf_rdma_request_free(struct spdk_nvmf_rdma_request *rdma_req,
1958 : struct spdk_nvmf_rdma_transport *rtransport)
1959 : {
1960 : struct spdk_nvmf_rdma_qpair *rqpair;
1961 : struct spdk_nvmf_rdma_poll_group *rgroup;
1962 :
1963 6 : rqpair = SPDK_CONTAINEROF(rdma_req->req.qpair, struct spdk_nvmf_rdma_qpair, qpair);
1964 6 : if (rdma_req->req.data_from_pool) {
1965 5 : rgroup = rqpair->poller->group;
1966 :
1967 5 : spdk_nvmf_request_free_buffers(&rdma_req->req, &rgroup->group, &rtransport->transport);
1968 : }
1969 6 : if (rdma_req->req.stripped_data) {
1970 0 : nvmf_request_free_stripped_buffers(&rdma_req->req,
1971 0 : &rqpair->poller->group->group,
1972 : &rtransport->transport);
1973 : }
1974 6 : nvmf_rdma_request_free_data(rdma_req, rtransport);
1975 6 : rdma_req->req.length = 0;
1976 6 : rdma_req->req.iovcnt = 0;
1977 6 : rdma_req->offset = 0;
1978 6 : rdma_req->req.dif_enabled = false;
1979 6 : rdma_req->fused_failed = false;
1980 6 : rdma_req->transfer_wr = NULL;
1981 6 : if (rdma_req->fused_pair) {
1982 : /* This req was part of a valid fused pair, but failed before it got to
1983 : * READ_TO_EXECUTE state. This means we need to fail the other request
1984 : * in the pair, because it is no longer part of a valid pair. If the pair
1985 : * already reached READY_TO_EXECUTE state, we need to kick it.
1986 : */
1987 0 : rdma_req->fused_pair->fused_failed = true;
1988 0 : if (rdma_req->fused_pair->state == RDMA_REQUEST_STATE_READY_TO_EXECUTE) {
1989 0 : nvmf_rdma_request_process(rtransport, rdma_req->fused_pair);
1990 : }
1991 0 : rdma_req->fused_pair = NULL;
1992 : }
1993 6 : memset(&rdma_req->req.dif, 0, sizeof(rdma_req->req.dif));
1994 6 : rqpair->qd--;
1995 :
1996 6 : STAILQ_INSERT_HEAD(&rqpair->resources->free_queue, rdma_req, state_link);
1997 6 : rqpair->qpair.queue_depth--;
1998 6 : rdma_req->state = RDMA_REQUEST_STATE_FREE;
1999 6 : }
2000 :
2001 : static void
2002 6 : nvmf_rdma_check_fused_ordering(struct spdk_nvmf_rdma_transport *rtransport,
2003 : struct spdk_nvmf_rdma_qpair *rqpair,
2004 : struct spdk_nvmf_rdma_request *rdma_req)
2005 : {
2006 : enum spdk_nvme_cmd_fuse last, next;
2007 :
2008 6 : last = rqpair->fused_first ? rqpair->fused_first->req.cmd->nvme_cmd.fuse : SPDK_NVME_CMD_FUSE_NONE;
2009 6 : next = rdma_req->req.cmd->nvme_cmd.fuse;
2010 :
2011 6 : assert(last != SPDK_NVME_CMD_FUSE_SECOND);
2012 :
2013 6 : if (spdk_likely(last == SPDK_NVME_CMD_FUSE_NONE && next == SPDK_NVME_CMD_FUSE_NONE)) {
2014 6 : return;
2015 : }
2016 :
2017 0 : if (last == SPDK_NVME_CMD_FUSE_FIRST) {
2018 0 : if (next == SPDK_NVME_CMD_FUSE_SECOND) {
2019 : /* This is a valid pair of fused commands. Point them at each other
2020 : * so they can be submitted consecutively once ready to be executed.
2021 : */
2022 0 : rqpair->fused_first->fused_pair = rdma_req;
2023 0 : rdma_req->fused_pair = rqpair->fused_first;
2024 0 : rqpair->fused_first = NULL;
2025 0 : return;
2026 : } else {
2027 : /* Mark the last req as failed since it wasn't followed by a SECOND. */
2028 0 : rqpair->fused_first->fused_failed = true;
2029 :
2030 : /* If the last req is in READY_TO_EXECUTE state, then call
2031 : * nvmf_rdma_request_process(), otherwise nothing else will kick it.
2032 : */
2033 0 : if (rqpair->fused_first->state == RDMA_REQUEST_STATE_READY_TO_EXECUTE) {
2034 0 : nvmf_rdma_request_process(rtransport, rqpair->fused_first);
2035 : }
2036 :
2037 0 : rqpair->fused_first = NULL;
2038 : }
2039 : }
2040 :
2041 0 : if (next == SPDK_NVME_CMD_FUSE_FIRST) {
2042 : /* Set rqpair->fused_first here so that we know to check that the next request
2043 : * is a SECOND (and to fail this one if it isn't).
2044 : */
2045 0 : rqpair->fused_first = rdma_req;
2046 0 : } else if (next == SPDK_NVME_CMD_FUSE_SECOND) {
2047 : /* Mark this req failed since it ia SECOND and the last one was not a FIRST. */
2048 0 : rdma_req->fused_failed = true;
2049 : }
2050 : }
2051 :
2052 : bool
2053 23 : nvmf_rdma_request_process(struct spdk_nvmf_rdma_transport *rtransport,
2054 : struct spdk_nvmf_rdma_request *rdma_req)
2055 : {
2056 : struct spdk_nvmf_rdma_qpair *rqpair;
2057 : struct spdk_nvmf_rdma_device *device;
2058 : struct spdk_nvmf_rdma_poll_group *rgroup;
2059 23 : struct spdk_nvme_cpl *rsp = &rdma_req->req.rsp->nvme_cpl;
2060 : int rc;
2061 : struct spdk_nvmf_rdma_recv *rdma_recv;
2062 : enum spdk_nvmf_rdma_request_state prev_state;
2063 23 : bool progress = false;
2064 23 : int data_posted;
2065 : uint32_t num_blocks, num_rdma_reads_available, qdepth;
2066 :
2067 23 : rqpair = SPDK_CONTAINEROF(rdma_req->req.qpair, struct spdk_nvmf_rdma_qpair, qpair);
2068 23 : device = rqpair->device;
2069 23 : rgroup = rqpair->poller->group;
2070 :
2071 23 : assert(rdma_req->state != RDMA_REQUEST_STATE_FREE);
2072 :
2073 : /* If the queue pair is in an error state, force the request to the completed state
2074 : * to release resources. */
2075 23 : if (spdk_unlikely(rqpair->ibv_in_error_state || !spdk_nvmf_qpair_is_active(&rqpair->qpair))) {
2076 0 : switch (rdma_req->state) {
2077 0 : case RDMA_REQUEST_STATE_NEED_BUFFER:
2078 0 : STAILQ_REMOVE(&rgroup->group.pending_buf_queue, &rdma_req->req, spdk_nvmf_request, buf_link);
2079 0 : break;
2080 0 : case RDMA_REQUEST_STATE_DATA_TRANSFER_TO_CONTROLLER_PENDING:
2081 0 : STAILQ_REMOVE(&rqpair->pending_rdma_read_queue, rdma_req, spdk_nvmf_rdma_request, state_link);
2082 0 : break;
2083 0 : case RDMA_REQUEST_STATE_TRANSFERRING_HOST_TO_CONTROLLER:
2084 0 : if (rdma_req->num_remaining_data_wr) {
2085 : /* Partially sent request is still in the pending_rdma_read_queue,
2086 : * remove it before completing */
2087 0 : rdma_req->num_remaining_data_wr = 0;
2088 0 : STAILQ_REMOVE(&rqpair->pending_rdma_read_queue, rdma_req, spdk_nvmf_rdma_request, state_link);
2089 : }
2090 0 : break;
2091 0 : case RDMA_REQUEST_STATE_DATA_TRANSFER_TO_HOST_PENDING:
2092 0 : STAILQ_REMOVE(&rqpair->pending_rdma_write_queue, rdma_req, spdk_nvmf_rdma_request, state_link);
2093 0 : break;
2094 0 : case RDMA_REQUEST_STATE_READY_TO_COMPLETE_PENDING:
2095 0 : STAILQ_REMOVE(&rqpair->pending_rdma_send_queue, rdma_req, spdk_nvmf_rdma_request, state_link);
2096 0 : break;
2097 0 : default:
2098 0 : break;
2099 : }
2100 0 : rdma_req->state = RDMA_REQUEST_STATE_COMPLETED;
2101 : }
2102 :
2103 : /* The loop here is to allow for several back-to-back state changes. */
2104 : do {
2105 66 : prev_state = rdma_req->state;
2106 :
2107 66 : SPDK_DEBUGLOG(rdma, "Request %p entering state %d\n", rdma_req, prev_state);
2108 :
2109 66 : switch (rdma_req->state) {
2110 6 : case RDMA_REQUEST_STATE_FREE:
2111 : /* Some external code must kick a request into RDMA_REQUEST_STATE_NEW
2112 : * to escape this state. */
2113 6 : break;
2114 6 : case RDMA_REQUEST_STATE_NEW:
2115 6 : spdk_trace_record(TRACE_RDMA_REQUEST_STATE_NEW, 0, 0,
2116 : (uintptr_t)rdma_req, (uintptr_t)rqpair, rqpair->qpair.queue_depth);
2117 6 : rdma_recv = rdma_req->recv;
2118 :
2119 : /* The first element of the SGL is the NVMe command */
2120 6 : rdma_req->req.cmd = (union nvmf_h2c_msg *)rdma_recv->sgl[0].addr;
2121 6 : memset(rdma_req->req.rsp, 0, sizeof(*rdma_req->req.rsp));
2122 6 : rdma_req->transfer_wr = &rdma_req->data.wr;
2123 :
2124 6 : if (spdk_unlikely(rqpair->ibv_in_error_state || !spdk_nvmf_qpair_is_active(&rqpair->qpair))) {
2125 0 : rdma_req->state = RDMA_REQUEST_STATE_COMPLETED;
2126 0 : break;
2127 : }
2128 :
2129 6 : if (spdk_unlikely(spdk_nvmf_request_get_dif_ctx(&rdma_req->req, &rdma_req->req.dif.dif_ctx))) {
2130 0 : rdma_req->req.dif_enabled = true;
2131 : }
2132 :
2133 6 : nvmf_rdma_check_fused_ordering(rtransport, rqpair, rdma_req);
2134 :
2135 : #ifdef SPDK_CONFIG_RDMA_SEND_WITH_INVAL
2136 6 : rdma_req->rsp.wr.opcode = IBV_WR_SEND;
2137 6 : rdma_req->rsp.wr.imm_data = 0;
2138 : #endif
2139 :
2140 : /* The next state transition depends on the data transfer needs of this request. */
2141 6 : rdma_req->req.xfer = spdk_nvmf_req_get_xfer(&rdma_req->req);
2142 :
2143 6 : if (spdk_unlikely(rdma_req->req.xfer == SPDK_NVME_DATA_BIDIRECTIONAL)) {
2144 1 : rsp->status.sct = SPDK_NVME_SCT_GENERIC;
2145 1 : rsp->status.sc = SPDK_NVME_SC_INVALID_OPCODE;
2146 1 : STAILQ_INSERT_TAIL(&rqpair->pending_rdma_send_queue, rdma_req, state_link);
2147 1 : rdma_req->state = RDMA_REQUEST_STATE_READY_TO_COMPLETE_PENDING;
2148 1 : SPDK_DEBUGLOG(rdma, "Request %p: invalid xfer type (BIDIRECTIONAL)\n", rdma_req);
2149 1 : break;
2150 : }
2151 :
2152 : /* If no data to transfer, ready to execute. */
2153 5 : if (rdma_req->req.xfer == SPDK_NVME_DATA_NONE) {
2154 0 : rdma_req->state = RDMA_REQUEST_STATE_READY_TO_EXECUTE;
2155 0 : break;
2156 : }
2157 :
2158 5 : rdma_req->state = RDMA_REQUEST_STATE_NEED_BUFFER;
2159 5 : STAILQ_INSERT_TAIL(&rgroup->group.pending_buf_queue, &rdma_req->req, buf_link);
2160 5 : break;
2161 5 : case RDMA_REQUEST_STATE_NEED_BUFFER:
2162 5 : spdk_trace_record(TRACE_RDMA_REQUEST_STATE_NEED_BUFFER, 0, 0,
2163 : (uintptr_t)rdma_req, (uintptr_t)rqpair);
2164 :
2165 5 : assert(rdma_req->req.xfer != SPDK_NVME_DATA_NONE);
2166 :
2167 5 : if (&rdma_req->req != STAILQ_FIRST(&rgroup->group.pending_buf_queue)) {
2168 : /* This request needs to wait in line to obtain a buffer */
2169 0 : break;
2170 : }
2171 :
2172 : /* Try to get a data buffer */
2173 5 : rc = nvmf_rdma_request_parse_sgl(rtransport, device, rdma_req);
2174 5 : if (spdk_unlikely(rc < 0)) {
2175 0 : STAILQ_REMOVE_HEAD(&rgroup->group.pending_buf_queue, buf_link);
2176 0 : STAILQ_INSERT_TAIL(&rqpair->pending_rdma_send_queue, rdma_req, state_link);
2177 0 : rdma_req->state = RDMA_REQUEST_STATE_READY_TO_COMPLETE_PENDING;
2178 0 : break;
2179 : }
2180 :
2181 5 : if (rdma_req->req.iovcnt == 0) {
2182 : /* No buffers available. */
2183 0 : rgroup->stat.pending_data_buffer++;
2184 0 : break;
2185 : }
2186 :
2187 5 : STAILQ_REMOVE_HEAD(&rgroup->group.pending_buf_queue, buf_link);
2188 :
2189 : /* If data is transferring from host to controller and the data didn't
2190 : * arrive using in capsule data, we need to do a transfer from the host.
2191 : */
2192 5 : if (rdma_req->req.xfer == SPDK_NVME_DATA_HOST_TO_CONTROLLER &&
2193 : rdma_req->req.data_from_pool) {
2194 4 : STAILQ_INSERT_TAIL(&rqpair->pending_rdma_read_queue, rdma_req, state_link);
2195 4 : rdma_req->state = RDMA_REQUEST_STATE_DATA_TRANSFER_TO_CONTROLLER_PENDING;
2196 4 : break;
2197 : }
2198 :
2199 1 : rdma_req->state = RDMA_REQUEST_STATE_READY_TO_EXECUTE;
2200 1 : break;
2201 4 : case RDMA_REQUEST_STATE_DATA_TRANSFER_TO_CONTROLLER_PENDING:
2202 4 : spdk_trace_record(TRACE_RDMA_REQUEST_STATE_DATA_TRANSFER_TO_CONTROLLER_PENDING, 0, 0,
2203 : (uintptr_t)rdma_req, (uintptr_t)rqpair);
2204 :
2205 4 : if (rdma_req != STAILQ_FIRST(&rqpair->pending_rdma_read_queue)) {
2206 : /* This request needs to wait in line to perform RDMA */
2207 0 : break;
2208 : }
2209 4 : assert(rqpair->max_send_depth >= rqpair->current_send_depth);
2210 4 : qdepth = rqpair->max_send_depth - rqpair->current_send_depth;
2211 4 : assert(rqpair->max_read_depth >= rqpair->current_read_depth);
2212 4 : num_rdma_reads_available = rqpair->max_read_depth - rqpair->current_read_depth;
2213 4 : if (rdma_req->num_outstanding_data_wr > qdepth ||
2214 4 : rdma_req->num_outstanding_data_wr > num_rdma_reads_available) {
2215 0 : if (num_rdma_reads_available && qdepth) {
2216 : /* Send as much as we can */
2217 0 : request_prepare_transfer_in_part(&rdma_req->req, spdk_min(num_rdma_reads_available, qdepth));
2218 : } else {
2219 : /* We can only have so many WRs outstanding. we have to wait until some finish. */
2220 0 : rqpair->poller->stat.pending_rdma_read++;
2221 0 : break;
2222 : }
2223 : }
2224 :
2225 : /* We have already verified that this request is the head of the queue. */
2226 4 : if (rdma_req->num_remaining_data_wr == 0) {
2227 4 : STAILQ_REMOVE_HEAD(&rqpair->pending_rdma_read_queue, state_link);
2228 : }
2229 :
2230 4 : request_transfer_in(&rdma_req->req);
2231 4 : rdma_req->state = RDMA_REQUEST_STATE_TRANSFERRING_HOST_TO_CONTROLLER;
2232 :
2233 4 : break;
2234 4 : case RDMA_REQUEST_STATE_TRANSFERRING_HOST_TO_CONTROLLER:
2235 4 : spdk_trace_record(TRACE_RDMA_REQUEST_STATE_TRANSFERRING_HOST_TO_CONTROLLER, 0, 0,
2236 : (uintptr_t)rdma_req, (uintptr_t)rqpair);
2237 : /* Some external code must kick a request into RDMA_REQUEST_STATE_READY_TO_EXECUTE
2238 : * to escape this state. */
2239 4 : break;
2240 5 : case RDMA_REQUEST_STATE_READY_TO_EXECUTE:
2241 5 : spdk_trace_record(TRACE_RDMA_REQUEST_STATE_READY_TO_EXECUTE, 0, 0,
2242 : (uintptr_t)rdma_req, (uintptr_t)rqpair);
2243 :
2244 5 : if (spdk_unlikely(rdma_req->req.dif_enabled)) {
2245 0 : if (rdma_req->req.xfer == SPDK_NVME_DATA_HOST_TO_CONTROLLER) {
2246 : /* generate DIF for write operation */
2247 0 : num_blocks = SPDK_CEIL_DIV(rdma_req->req.dif.elba_length, rdma_req->req.dif.dif_ctx.block_size);
2248 0 : assert(num_blocks > 0);
2249 :
2250 0 : rc = spdk_dif_generate(rdma_req->req.iov, rdma_req->req.iovcnt,
2251 0 : num_blocks, &rdma_req->req.dif.dif_ctx);
2252 0 : if (rc != 0) {
2253 0 : SPDK_ERRLOG("DIF generation failed\n");
2254 0 : rdma_req->state = RDMA_REQUEST_STATE_COMPLETED;
2255 0 : spdk_nvmf_qpair_disconnect(&rqpair->qpair);
2256 0 : break;
2257 : }
2258 : }
2259 :
2260 0 : assert(rdma_req->req.dif.elba_length >= rdma_req->req.length);
2261 : /* set extended length before IO operation */
2262 0 : rdma_req->req.length = rdma_req->req.dif.elba_length;
2263 : }
2264 :
2265 5 : if (rdma_req->req.cmd->nvme_cmd.fuse != SPDK_NVME_CMD_FUSE_NONE) {
2266 0 : if (rdma_req->fused_failed) {
2267 : /* This request failed FUSED semantics. Fail it immediately, without
2268 : * even sending it to the target layer.
2269 : */
2270 0 : rsp->status.sct = SPDK_NVME_SCT_GENERIC;
2271 0 : rsp->status.sc = SPDK_NVME_SC_ABORTED_MISSING_FUSED;
2272 0 : STAILQ_INSERT_TAIL(&rqpair->pending_rdma_send_queue, rdma_req, state_link);
2273 0 : rdma_req->state = RDMA_REQUEST_STATE_READY_TO_COMPLETE_PENDING;
2274 0 : break;
2275 : }
2276 :
2277 0 : if (rdma_req->fused_pair == NULL ||
2278 0 : rdma_req->fused_pair->state != RDMA_REQUEST_STATE_READY_TO_EXECUTE) {
2279 : /* This request is ready to execute, but either we don't know yet if it's
2280 : * valid - i.e. this is a FIRST but we haven't received the next
2281 : * request yet or the other request of this fused pair isn't ready to
2282 : * execute. So break here and this request will get processed later either
2283 : * when the other request is ready or we find that this request isn't valid.
2284 : */
2285 : break;
2286 : }
2287 : }
2288 :
2289 : /* If we get to this point, and this request is a fused command, we know that
2290 : * it is part of valid sequence (FIRST followed by a SECOND) and that both
2291 : * requests are READY_TO_EXECUTE. So call spdk_nvmf_request_exec() both on this
2292 : * request, and the other request of the fused pair, in the correct order.
2293 : * Also clear the ->fused_pair pointers on both requests, since after this point
2294 : * we no longer need to maintain the relationship between these two requests.
2295 : */
2296 5 : if (rdma_req->req.cmd->nvme_cmd.fuse == SPDK_NVME_CMD_FUSE_SECOND) {
2297 0 : assert(rdma_req->fused_pair != NULL);
2298 0 : assert(rdma_req->fused_pair->fused_pair != NULL);
2299 0 : rdma_req->fused_pair->state = RDMA_REQUEST_STATE_EXECUTING;
2300 0 : spdk_nvmf_request_exec(&rdma_req->fused_pair->req);
2301 0 : rdma_req->fused_pair->fused_pair = NULL;
2302 0 : rdma_req->fused_pair = NULL;
2303 : }
2304 5 : rdma_req->state = RDMA_REQUEST_STATE_EXECUTING;
2305 5 : spdk_nvmf_request_exec(&rdma_req->req);
2306 5 : if (rdma_req->req.cmd->nvme_cmd.fuse == SPDK_NVME_CMD_FUSE_FIRST) {
2307 0 : assert(rdma_req->fused_pair != NULL);
2308 0 : assert(rdma_req->fused_pair->fused_pair != NULL);
2309 0 : rdma_req->fused_pair->state = RDMA_REQUEST_STATE_EXECUTING;
2310 0 : spdk_nvmf_request_exec(&rdma_req->fused_pair->req);
2311 0 : rdma_req->fused_pair->fused_pair = NULL;
2312 0 : rdma_req->fused_pair = NULL;
2313 : }
2314 5 : break;
2315 5 : case RDMA_REQUEST_STATE_EXECUTING:
2316 5 : spdk_trace_record(TRACE_RDMA_REQUEST_STATE_EXECUTING, 0, 0,
2317 : (uintptr_t)rdma_req, (uintptr_t)rqpair);
2318 : /* Some external code must kick a request into RDMA_REQUEST_STATE_EXECUTED
2319 : * to escape this state. */
2320 5 : break;
2321 5 : case RDMA_REQUEST_STATE_EXECUTED:
2322 5 : spdk_trace_record(TRACE_RDMA_REQUEST_STATE_EXECUTED, 0, 0,
2323 : (uintptr_t)rdma_req, (uintptr_t)rqpair);
2324 5 : if (rsp->status.sc == SPDK_NVME_SC_SUCCESS &&
2325 5 : rdma_req->req.xfer == SPDK_NVME_DATA_CONTROLLER_TO_HOST) {
2326 1 : STAILQ_INSERT_TAIL(&rqpair->pending_rdma_write_queue, rdma_req, state_link);
2327 1 : rdma_req->state = RDMA_REQUEST_STATE_DATA_TRANSFER_TO_HOST_PENDING;
2328 : } else {
2329 4 : STAILQ_INSERT_TAIL(&rqpair->pending_rdma_send_queue, rdma_req, state_link);
2330 4 : rdma_req->state = RDMA_REQUEST_STATE_READY_TO_COMPLETE_PENDING;
2331 : }
2332 5 : if (spdk_unlikely(rdma_req->req.dif_enabled)) {
2333 : /* restore the original length */
2334 0 : rdma_req->req.length = rdma_req->req.dif.orig_length;
2335 :
2336 0 : if (rdma_req->req.xfer == SPDK_NVME_DATA_CONTROLLER_TO_HOST) {
2337 0 : struct spdk_dif_error error_blk;
2338 :
2339 0 : num_blocks = SPDK_CEIL_DIV(rdma_req->req.dif.elba_length, rdma_req->req.dif.dif_ctx.block_size);
2340 0 : if (!rdma_req->req.stripped_data) {
2341 0 : rc = spdk_dif_verify(rdma_req->req.iov, rdma_req->req.iovcnt, num_blocks,
2342 0 : &rdma_req->req.dif.dif_ctx, &error_blk);
2343 : } else {
2344 0 : rc = spdk_dif_verify_copy(rdma_req->req.stripped_data->iov,
2345 0 : rdma_req->req.stripped_data->iovcnt,
2346 0 : rdma_req->req.iov, rdma_req->req.iovcnt, num_blocks,
2347 0 : &rdma_req->req.dif.dif_ctx, &error_blk);
2348 : }
2349 0 : if (rc) {
2350 0 : struct spdk_nvme_cpl *rsp = &rdma_req->req.rsp->nvme_cpl;
2351 :
2352 0 : SPDK_ERRLOG("DIF error detected. type=%d, offset=%" PRIu32 "\n", error_blk.err_type,
2353 : error_blk.err_offset);
2354 0 : rsp->status.sct = SPDK_NVME_SCT_MEDIA_ERROR;
2355 0 : rsp->status.sc = nvmf_rdma_dif_error_to_compl_status(error_blk.err_type);
2356 0 : STAILQ_REMOVE(&rqpair->pending_rdma_write_queue, rdma_req, spdk_nvmf_rdma_request, state_link);
2357 0 : STAILQ_INSERT_TAIL(&rqpair->pending_rdma_send_queue, rdma_req, state_link);
2358 0 : rdma_req->state = RDMA_REQUEST_STATE_READY_TO_COMPLETE_PENDING;
2359 : }
2360 : }
2361 : }
2362 5 : break;
2363 1 : case RDMA_REQUEST_STATE_DATA_TRANSFER_TO_HOST_PENDING:
2364 1 : spdk_trace_record(TRACE_RDMA_REQUEST_STATE_DATA_TRANSFER_TO_HOST_PENDING, 0, 0,
2365 : (uintptr_t)rdma_req, (uintptr_t)rqpair);
2366 :
2367 1 : if (rdma_req != STAILQ_FIRST(&rqpair->pending_rdma_write_queue)) {
2368 : /* This request needs to wait in line to perform RDMA */
2369 0 : break;
2370 : }
2371 1 : if ((rqpair->current_send_depth + rdma_req->num_outstanding_data_wr + 1) >
2372 1 : rqpair->max_send_depth) {
2373 : /* We can only have so many WRs outstanding. we have to wait until some finish.
2374 : * +1 since each request has an additional wr in the resp. */
2375 0 : rqpair->poller->stat.pending_rdma_write++;
2376 0 : break;
2377 : }
2378 :
2379 : /* We have already verified that this request is the head of the queue. */
2380 1 : STAILQ_REMOVE_HEAD(&rqpair->pending_rdma_write_queue, state_link);
2381 :
2382 : /* The data transfer will be kicked off from
2383 : * RDMA_REQUEST_STATE_READY_TO_COMPLETE state.
2384 : * We verified that data + response fit into send queue, so we can go to the next state directly
2385 : */
2386 1 : rdma_req->state = RDMA_REQUEST_STATE_READY_TO_COMPLETE;
2387 1 : break;
2388 7 : case RDMA_REQUEST_STATE_READY_TO_COMPLETE_PENDING:
2389 7 : spdk_trace_record(TRACE_RDMA_REQUEST_STATE_READY_TO_COMPLETE_PENDING, 0, 0,
2390 : (uintptr_t)rdma_req, (uintptr_t)rqpair);
2391 :
2392 7 : if (rdma_req != STAILQ_FIRST(&rqpair->pending_rdma_send_queue)) {
2393 : /* This request needs to wait in line to send the completion */
2394 0 : break;
2395 : }
2396 :
2397 7 : assert(rqpair->current_send_depth <= rqpair->max_send_depth);
2398 7 : if (rqpair->current_send_depth == rqpair->max_send_depth) {
2399 : /* We can only have so many WRs outstanding. we have to wait until some finish */
2400 2 : rqpair->poller->stat.pending_rdma_send++;
2401 2 : break;
2402 : }
2403 :
2404 : /* We have already verified that this request is the head of the queue. */
2405 5 : STAILQ_REMOVE_HEAD(&rqpair->pending_rdma_send_queue, state_link);
2406 :
2407 : /* The response sending will be kicked off from
2408 : * RDMA_REQUEST_STATE_READY_TO_COMPLETE state.
2409 : */
2410 5 : rdma_req->state = RDMA_REQUEST_STATE_READY_TO_COMPLETE;
2411 5 : break;
2412 6 : case RDMA_REQUEST_STATE_READY_TO_COMPLETE:
2413 6 : spdk_trace_record(TRACE_RDMA_REQUEST_STATE_READY_TO_COMPLETE, 0, 0,
2414 : (uintptr_t)rdma_req, (uintptr_t)rqpair);
2415 6 : rc = request_transfer_out(&rdma_req->req, &data_posted);
2416 6 : assert(rc == 0); /* No good way to handle this currently */
2417 6 : if (spdk_unlikely(rc)) {
2418 0 : rdma_req->state = RDMA_REQUEST_STATE_COMPLETED;
2419 : } else {
2420 6 : rdma_req->state = data_posted ? RDMA_REQUEST_STATE_TRANSFERRING_CONTROLLER_TO_HOST :
2421 : RDMA_REQUEST_STATE_COMPLETING;
2422 : }
2423 6 : break;
2424 1 : case RDMA_REQUEST_STATE_TRANSFERRING_CONTROLLER_TO_HOST:
2425 1 : spdk_trace_record(TRACE_RDMA_REQUEST_STATE_TRANSFERRING_CONTROLLER_TO_HOST, 0, 0,
2426 : (uintptr_t)rdma_req, (uintptr_t)rqpair);
2427 : /* Some external code must kick a request into RDMA_REQUEST_STATE_COMPLETED
2428 : * to escape this state. */
2429 1 : break;
2430 5 : case RDMA_REQUEST_STATE_COMPLETING:
2431 5 : spdk_trace_record(TRACE_RDMA_REQUEST_STATE_COMPLETING, 0, 0,
2432 : (uintptr_t)rdma_req, (uintptr_t)rqpair);
2433 : /* Some external code must kick a request into RDMA_REQUEST_STATE_COMPLETED
2434 : * to escape this state. */
2435 5 : break;
2436 6 : case RDMA_REQUEST_STATE_COMPLETED:
2437 6 : spdk_trace_record(TRACE_RDMA_REQUEST_STATE_COMPLETED, 0, 0,
2438 : (uintptr_t)rdma_req, (uintptr_t)rqpair, rqpair->qpair.queue_depth);
2439 :
2440 6 : rqpair->poller->stat.request_latency += spdk_get_ticks() - rdma_req->receive_tsc;
2441 6 : _nvmf_rdma_request_free(rdma_req, rtransport);
2442 6 : break;
2443 0 : case RDMA_REQUEST_NUM_STATES:
2444 : default:
2445 0 : assert(0);
2446 : break;
2447 : }
2448 :
2449 66 : if (rdma_req->state != prev_state) {
2450 43 : progress = true;
2451 : }
2452 66 : } while (rdma_req->state != prev_state);
2453 :
2454 23 : return progress;
2455 : }
2456 :
2457 : /* Public API callbacks begin here */
2458 :
2459 : #define SPDK_NVMF_RDMA_DEFAULT_MAX_QUEUE_DEPTH 128
2460 : #define SPDK_NVMF_RDMA_DEFAULT_AQ_DEPTH 128
2461 : #define SPDK_NVMF_RDMA_DEFAULT_SRQ_DEPTH 4096
2462 : #define SPDK_NVMF_RDMA_DEFAULT_MAX_QPAIRS_PER_CTRLR 128
2463 : #define SPDK_NVMF_RDMA_DEFAULT_IN_CAPSULE_DATA_SIZE 4096
2464 : #define SPDK_NVMF_RDMA_DEFAULT_MAX_IO_SIZE 131072
2465 : #define SPDK_NVMF_RDMA_MIN_IO_BUFFER_SIZE (SPDK_NVMF_RDMA_DEFAULT_MAX_IO_SIZE / SPDK_NVMF_MAX_SGL_ENTRIES)
2466 : #define SPDK_NVMF_RDMA_DEFAULT_NUM_SHARED_BUFFERS 4095
2467 : #define SPDK_NVMF_RDMA_DEFAULT_BUFFER_CACHE_SIZE UINT32_MAX
2468 : #define SPDK_NVMF_RDMA_DEFAULT_NO_SRQ false
2469 : #define SPDK_NVMF_RDMA_DIF_INSERT_OR_STRIP false
2470 : #define SPDK_NVMF_RDMA_ACCEPTOR_BACKLOG 100
2471 : #define SPDK_NVMF_RDMA_DEFAULT_ABORT_TIMEOUT_SEC 1
2472 : #define SPDK_NVMF_RDMA_DEFAULT_NO_WR_BATCHING false
2473 : #define SPDK_NVMF_RDMA_DEFAULT_DATA_WR_POOL_SIZE 4095
2474 :
2475 : static void
2476 1 : nvmf_rdma_opts_init(struct spdk_nvmf_transport_opts *opts)
2477 : {
2478 1 : opts->max_queue_depth = SPDK_NVMF_RDMA_DEFAULT_MAX_QUEUE_DEPTH;
2479 1 : opts->max_qpairs_per_ctrlr = SPDK_NVMF_RDMA_DEFAULT_MAX_QPAIRS_PER_CTRLR;
2480 1 : opts->in_capsule_data_size = SPDK_NVMF_RDMA_DEFAULT_IN_CAPSULE_DATA_SIZE;
2481 1 : opts->max_io_size = SPDK_NVMF_RDMA_DEFAULT_MAX_IO_SIZE;
2482 1 : opts->io_unit_size = SPDK_NVMF_RDMA_MIN_IO_BUFFER_SIZE;
2483 1 : opts->max_aq_depth = SPDK_NVMF_RDMA_DEFAULT_AQ_DEPTH;
2484 1 : opts->num_shared_buffers = SPDK_NVMF_RDMA_DEFAULT_NUM_SHARED_BUFFERS;
2485 1 : opts->buf_cache_size = SPDK_NVMF_RDMA_DEFAULT_BUFFER_CACHE_SIZE;
2486 1 : opts->dif_insert_or_strip = SPDK_NVMF_RDMA_DIF_INSERT_OR_STRIP;
2487 1 : opts->abort_timeout_sec = SPDK_NVMF_RDMA_DEFAULT_ABORT_TIMEOUT_SEC;
2488 1 : opts->transport_specific = NULL;
2489 1 : opts->data_wr_pool_size = SPDK_NVMF_RDMA_DEFAULT_DATA_WR_POOL_SIZE;
2490 1 : }
2491 :
2492 : static int nvmf_rdma_destroy(struct spdk_nvmf_transport *transport,
2493 : spdk_nvmf_transport_destroy_done_cb cb_fn, void *cb_arg);
2494 :
2495 : static inline bool
2496 0 : nvmf_rdma_is_rxe_device(struct spdk_nvmf_rdma_device *device)
2497 : {
2498 0 : return device->attr.vendor_id == SPDK_RDMA_RXE_VENDOR_ID_OLD ||
2499 0 : device->attr.vendor_id == SPDK_RDMA_RXE_VENDOR_ID_NEW;
2500 : }
2501 :
2502 : static int nvmf_rdma_accept(void *ctx);
2503 : static bool nvmf_rdma_retry_listen_port(struct spdk_nvmf_rdma_transport *rtransport);
2504 : static void destroy_ib_device(struct spdk_nvmf_rdma_transport *rtransport,
2505 : struct spdk_nvmf_rdma_device *device);
2506 :
2507 : static int
2508 0 : create_ib_device(struct spdk_nvmf_rdma_transport *rtransport, struct ibv_context *context,
2509 : struct spdk_nvmf_rdma_device **new_device)
2510 : {
2511 : struct spdk_nvmf_rdma_device *device;
2512 0 : int flag = 0;
2513 0 : int rc = 0;
2514 :
2515 0 : device = calloc(1, sizeof(*device));
2516 0 : if (!device) {
2517 0 : SPDK_ERRLOG("Unable to allocate memory for RDMA devices.\n");
2518 0 : return -ENOMEM;
2519 : }
2520 0 : device->context = context;
2521 0 : rc = ibv_query_device(device->context, &device->attr);
2522 0 : if (rc < 0) {
2523 0 : SPDK_ERRLOG("Failed to query RDMA device attributes.\n");
2524 0 : free(device);
2525 0 : return rc;
2526 : }
2527 :
2528 : #ifdef SPDK_CONFIG_RDMA_SEND_WITH_INVAL
2529 0 : if ((device->attr.device_cap_flags & IBV_DEVICE_MEM_MGT_EXTENSIONS) == 0) {
2530 0 : SPDK_WARNLOG("The libibverbs on this system supports SEND_WITH_INVALIDATE,");
2531 0 : SPDK_WARNLOG("but the device with vendor ID %u does not.\n", device->attr.vendor_id);
2532 : }
2533 :
2534 : /**
2535 : * The vendor ID is assigned by the IEEE and an ID of 0 implies Soft-RoCE.
2536 : * The Soft-RoCE RXE driver does not currently support send with invalidate,
2537 : * but incorrectly reports that it does. There are changes making their way
2538 : * through the kernel now that will enable this feature. When they are merged,
2539 : * we can conditionally enable this feature.
2540 : *
2541 : * TODO: enable this for versions of the kernel rxe driver that support it.
2542 : */
2543 0 : if (nvmf_rdma_is_rxe_device(device)) {
2544 0 : device->attr.device_cap_flags &= ~(IBV_DEVICE_MEM_MGT_EXTENSIONS);
2545 : }
2546 : #endif
2547 :
2548 : /* set up device context async ev fd as NON_BLOCKING */
2549 0 : flag = fcntl(device->context->async_fd, F_GETFL);
2550 0 : rc = fcntl(device->context->async_fd, F_SETFL, flag | O_NONBLOCK);
2551 0 : if (rc < 0) {
2552 0 : SPDK_ERRLOG("Failed to set context async fd to NONBLOCK.\n");
2553 0 : free(device);
2554 0 : return rc;
2555 : }
2556 :
2557 0 : TAILQ_INSERT_TAIL(&rtransport->devices, device, link);
2558 0 : SPDK_DEBUGLOG(rdma, "New device %p is added to RDMA transport\n", device);
2559 :
2560 0 : if (g_nvmf_hooks.get_ibv_pd) {
2561 0 : device->pd = g_nvmf_hooks.get_ibv_pd(NULL, device->context);
2562 : } else {
2563 0 : device->pd = ibv_alloc_pd(device->context);
2564 : }
2565 :
2566 0 : if (!device->pd) {
2567 0 : SPDK_ERRLOG("Unable to allocate protection domain.\n");
2568 0 : destroy_ib_device(rtransport, device);
2569 0 : return -ENOMEM;
2570 : }
2571 :
2572 0 : assert(device->map == NULL);
2573 :
2574 0 : device->map = spdk_rdma_utils_create_mem_map(device->pd, &g_nvmf_hooks, IBV_ACCESS_LOCAL_WRITE);
2575 0 : if (!device->map) {
2576 0 : SPDK_ERRLOG("Unable to allocate memory map for listen address\n");
2577 0 : destroy_ib_device(rtransport, device);
2578 0 : return -ENOMEM;
2579 : }
2580 :
2581 0 : assert(device->map != NULL);
2582 0 : assert(device->pd != NULL);
2583 :
2584 0 : if (new_device) {
2585 0 : *new_device = device;
2586 : }
2587 0 : SPDK_NOTICELOG("Create IB device %s(%p/%p) succeed.\n", ibv_get_device_name(context->device),
2588 : device, context);
2589 :
2590 0 : return 0;
2591 : }
2592 :
2593 : static void
2594 0 : free_poll_fds(struct spdk_nvmf_rdma_transport *rtransport)
2595 : {
2596 0 : if (rtransport->poll_fds) {
2597 0 : free(rtransport->poll_fds);
2598 0 : rtransport->poll_fds = NULL;
2599 : }
2600 0 : rtransport->npoll_fds = 0;
2601 0 : }
2602 :
2603 : static int
2604 0 : generate_poll_fds(struct spdk_nvmf_rdma_transport *rtransport)
2605 : {
2606 : /* Set up poll descriptor array to monitor events from RDMA and IB
2607 : * in a single poll syscall
2608 : */
2609 0 : int device_count = 0;
2610 0 : int i = 0;
2611 : struct spdk_nvmf_rdma_device *device, *tmp;
2612 :
2613 0 : TAILQ_FOREACH_SAFE(device, &rtransport->devices, link, tmp) {
2614 0 : device_count++;
2615 : }
2616 :
2617 0 : rtransport->npoll_fds = device_count + 1;
2618 :
2619 0 : rtransport->poll_fds = calloc(rtransport->npoll_fds, sizeof(struct pollfd));
2620 0 : if (rtransport->poll_fds == NULL) {
2621 0 : SPDK_ERRLOG("poll_fds allocation failed\n");
2622 0 : return -ENOMEM;
2623 : }
2624 :
2625 0 : rtransport->poll_fds[i].fd = rtransport->event_channel->fd;
2626 0 : rtransport->poll_fds[i++].events = POLLIN;
2627 :
2628 0 : TAILQ_FOREACH_SAFE(device, &rtransport->devices, link, tmp) {
2629 0 : rtransport->poll_fds[i].fd = device->context->async_fd;
2630 0 : rtransport->poll_fds[i++].events = POLLIN;
2631 : }
2632 :
2633 0 : return 0;
2634 : }
2635 :
2636 : static struct spdk_nvmf_transport *
2637 0 : nvmf_rdma_create(struct spdk_nvmf_transport_opts *opts)
2638 : {
2639 : int rc;
2640 : struct spdk_nvmf_rdma_transport *rtransport;
2641 0 : struct spdk_nvmf_rdma_device *device;
2642 : struct ibv_context **contexts;
2643 : size_t data_wr_pool_size;
2644 : uint32_t i;
2645 : int flag;
2646 : uint32_t sge_count;
2647 : uint32_t min_shared_buffers;
2648 : uint32_t min_in_capsule_data_size;
2649 0 : int max_device_sge = SPDK_NVMF_MAX_SGL_ENTRIES;
2650 :
2651 0 : rtransport = calloc(1, sizeof(*rtransport));
2652 0 : if (!rtransport) {
2653 0 : return NULL;
2654 : }
2655 :
2656 0 : TAILQ_INIT(&rtransport->devices);
2657 0 : TAILQ_INIT(&rtransport->ports);
2658 0 : TAILQ_INIT(&rtransport->poll_groups);
2659 0 : TAILQ_INIT(&rtransport->retry_ports);
2660 :
2661 0 : rtransport->transport.ops = &spdk_nvmf_transport_rdma;
2662 0 : rtransport->rdma_opts.num_cqe = DEFAULT_NVMF_RDMA_CQ_SIZE;
2663 0 : rtransport->rdma_opts.max_srq_depth = SPDK_NVMF_RDMA_DEFAULT_SRQ_DEPTH;
2664 0 : rtransport->rdma_opts.no_srq = SPDK_NVMF_RDMA_DEFAULT_NO_SRQ;
2665 0 : rtransport->rdma_opts.acceptor_backlog = SPDK_NVMF_RDMA_ACCEPTOR_BACKLOG;
2666 0 : rtransport->rdma_opts.no_wr_batching = SPDK_NVMF_RDMA_DEFAULT_NO_WR_BATCHING;
2667 0 : if (opts->transport_specific != NULL &&
2668 0 : spdk_json_decode_object_relaxed(opts->transport_specific, rdma_transport_opts_decoder,
2669 : SPDK_COUNTOF(rdma_transport_opts_decoder),
2670 0 : &rtransport->rdma_opts)) {
2671 0 : SPDK_ERRLOG("spdk_json_decode_object_relaxed failed\n");
2672 0 : nvmf_rdma_destroy(&rtransport->transport, NULL, NULL);
2673 0 : return NULL;
2674 : }
2675 :
2676 0 : SPDK_INFOLOG(rdma, "*** RDMA Transport Init ***\n"
2677 : " Transport opts: max_ioq_depth=%d, max_io_size=%d,\n"
2678 : " max_io_qpairs_per_ctrlr=%d, io_unit_size=%d,\n"
2679 : " in_capsule_data_size=%d, max_aq_depth=%d,\n"
2680 : " num_shared_buffers=%d, num_cqe=%d, max_srq_depth=%d, no_srq=%d,"
2681 : " acceptor_backlog=%d, no_wr_batching=%d abort_timeout_sec=%d\n",
2682 : opts->max_queue_depth,
2683 : opts->max_io_size,
2684 : opts->max_qpairs_per_ctrlr - 1,
2685 : opts->io_unit_size,
2686 : opts->in_capsule_data_size,
2687 : opts->max_aq_depth,
2688 : opts->num_shared_buffers,
2689 : rtransport->rdma_opts.num_cqe,
2690 : rtransport->rdma_opts.max_srq_depth,
2691 : rtransport->rdma_opts.no_srq,
2692 : rtransport->rdma_opts.acceptor_backlog,
2693 : rtransport->rdma_opts.no_wr_batching,
2694 : opts->abort_timeout_sec);
2695 :
2696 : /* I/O unit size cannot be larger than max I/O size */
2697 0 : if (opts->io_unit_size > opts->max_io_size) {
2698 0 : opts->io_unit_size = opts->max_io_size;
2699 : }
2700 :
2701 0 : if (rtransport->rdma_opts.acceptor_backlog <= 0) {
2702 0 : SPDK_ERRLOG("The acceptor backlog cannot be less than 1, setting to the default value of (%d).\n",
2703 : SPDK_NVMF_RDMA_ACCEPTOR_BACKLOG);
2704 0 : rtransport->rdma_opts.acceptor_backlog = SPDK_NVMF_RDMA_ACCEPTOR_BACKLOG;
2705 : }
2706 :
2707 0 : if (opts->num_shared_buffers < (SPDK_NVMF_MAX_SGL_ENTRIES * 2)) {
2708 0 : SPDK_ERRLOG("The number of shared data buffers (%d) is less than"
2709 : "the minimum number required to guarantee that forward progress can be made (%d)\n",
2710 : opts->num_shared_buffers, (SPDK_NVMF_MAX_SGL_ENTRIES * 2));
2711 0 : nvmf_rdma_destroy(&rtransport->transport, NULL, NULL);
2712 0 : return NULL;
2713 : }
2714 :
2715 : /* If buf_cache_size == UINT32_MAX, we will dynamically pick a cache size later that we know will fit. */
2716 0 : if (opts->buf_cache_size < UINT32_MAX) {
2717 0 : min_shared_buffers = spdk_env_get_core_count() * opts->buf_cache_size;
2718 0 : if (min_shared_buffers > opts->num_shared_buffers) {
2719 0 : SPDK_ERRLOG("There are not enough buffers to satisfy"
2720 : "per-poll group caches for each thread. (%" PRIu32 ")"
2721 : "supplied. (%" PRIu32 ") required\n", opts->num_shared_buffers, min_shared_buffers);
2722 0 : SPDK_ERRLOG("Please specify a larger number of shared buffers\n");
2723 0 : nvmf_rdma_destroy(&rtransport->transport, NULL, NULL);
2724 0 : return NULL;
2725 : }
2726 : }
2727 :
2728 0 : sge_count = opts->max_io_size / opts->io_unit_size;
2729 0 : if (sge_count > NVMF_DEFAULT_TX_SGE) {
2730 0 : SPDK_ERRLOG("Unsupported IO Unit size specified, %d bytes\n", opts->io_unit_size);
2731 0 : nvmf_rdma_destroy(&rtransport->transport, NULL, NULL);
2732 0 : return NULL;
2733 : }
2734 :
2735 0 : min_in_capsule_data_size = sizeof(struct spdk_nvme_sgl_descriptor) * SPDK_NVMF_MAX_SGL_ENTRIES;
2736 0 : if (opts->in_capsule_data_size < min_in_capsule_data_size) {
2737 0 : SPDK_WARNLOG("In capsule data size is set to %u, this is minimum size required to support msdbd=16\n",
2738 : min_in_capsule_data_size);
2739 0 : opts->in_capsule_data_size = min_in_capsule_data_size;
2740 : }
2741 :
2742 0 : rtransport->event_channel = rdma_create_event_channel();
2743 0 : if (rtransport->event_channel == NULL) {
2744 0 : SPDK_ERRLOG("rdma_create_event_channel() failed, %s\n", spdk_strerror(errno));
2745 0 : nvmf_rdma_destroy(&rtransport->transport, NULL, NULL);
2746 0 : return NULL;
2747 : }
2748 :
2749 0 : flag = fcntl(rtransport->event_channel->fd, F_GETFL);
2750 0 : if (fcntl(rtransport->event_channel->fd, F_SETFL, flag | O_NONBLOCK) < 0) {
2751 0 : SPDK_ERRLOG("fcntl can't set nonblocking mode for socket, fd: %d (%s)\n",
2752 : rtransport->event_channel->fd, spdk_strerror(errno));
2753 0 : nvmf_rdma_destroy(&rtransport->transport, NULL, NULL);
2754 0 : return NULL;
2755 : }
2756 :
2757 0 : data_wr_pool_size = opts->data_wr_pool_size;
2758 0 : if (data_wr_pool_size < SPDK_NVMF_MAX_SGL_ENTRIES * 2 * spdk_env_get_core_count()) {
2759 0 : data_wr_pool_size = SPDK_NVMF_MAX_SGL_ENTRIES * 2 * spdk_env_get_core_count();
2760 0 : SPDK_NOTICELOG("data_wr_pool_size is changed to %zu to guarantee enough cache for handling "
2761 : "at least one IO in each core\n", data_wr_pool_size);
2762 : }
2763 0 : rtransport->data_wr_pool = spdk_mempool_create("spdk_nvmf_rdma_wr_data", data_wr_pool_size,
2764 : sizeof(struct spdk_nvmf_rdma_request_data), SPDK_MEMPOOL_DEFAULT_CACHE_SIZE,
2765 : SPDK_ENV_NUMA_ID_ANY);
2766 0 : if (!rtransport->data_wr_pool) {
2767 0 : if (spdk_mempool_lookup("spdk_nvmf_rdma_wr_data") != NULL) {
2768 0 : SPDK_ERRLOG("Unable to allocate work request pool for poll group: already exists\n");
2769 0 : SPDK_ERRLOG("Probably running in multiprocess environment, which is "
2770 : "unsupported by the nvmf library\n");
2771 : } else {
2772 0 : SPDK_ERRLOG("Unable to allocate work request pool for poll group\n");
2773 : }
2774 0 : nvmf_rdma_destroy(&rtransport->transport, NULL, NULL);
2775 0 : return NULL;
2776 : }
2777 :
2778 0 : contexts = rdma_get_devices(NULL);
2779 0 : if (contexts == NULL) {
2780 0 : SPDK_ERRLOG("rdma_get_devices() failed: %s (%d)\n", spdk_strerror(errno), errno);
2781 0 : nvmf_rdma_destroy(&rtransport->transport, NULL, NULL);
2782 0 : return NULL;
2783 : }
2784 :
2785 0 : i = 0;
2786 0 : rc = 0;
2787 0 : while (contexts[i] != NULL) {
2788 0 : rc = create_ib_device(rtransport, contexts[i], &device);
2789 0 : if (rc < 0) {
2790 0 : break;
2791 : }
2792 0 : i++;
2793 0 : max_device_sge = spdk_min(max_device_sge, device->attr.max_sge);
2794 0 : device->is_ready = true;
2795 : }
2796 0 : rdma_free_devices(contexts);
2797 :
2798 0 : if (opts->io_unit_size * max_device_sge < opts->max_io_size) {
2799 : /* divide and round up. */
2800 0 : opts->io_unit_size = (opts->max_io_size + max_device_sge - 1) / max_device_sge;
2801 :
2802 : /* round up to the nearest 4k. */
2803 0 : opts->io_unit_size = (opts->io_unit_size + NVMF_DATA_BUFFER_ALIGNMENT - 1) & ~NVMF_DATA_BUFFER_MASK;
2804 :
2805 0 : opts->io_unit_size = spdk_max(opts->io_unit_size, SPDK_NVMF_RDMA_MIN_IO_BUFFER_SIZE);
2806 0 : SPDK_NOTICELOG("Adjusting the io unit size to fit the device's maximum I/O size. New I/O unit size %u\n",
2807 : opts->io_unit_size);
2808 : }
2809 :
2810 0 : if (rc < 0) {
2811 0 : nvmf_rdma_destroy(&rtransport->transport, NULL, NULL);
2812 0 : return NULL;
2813 : }
2814 :
2815 0 : rc = generate_poll_fds(rtransport);
2816 0 : if (rc < 0) {
2817 0 : nvmf_rdma_destroy(&rtransport->transport, NULL, NULL);
2818 0 : return NULL;
2819 : }
2820 :
2821 0 : rtransport->accept_poller = SPDK_POLLER_REGISTER(nvmf_rdma_accept, &rtransport->transport,
2822 : opts->acceptor_poll_rate);
2823 0 : if (!rtransport->accept_poller) {
2824 0 : nvmf_rdma_destroy(&rtransport->transport, NULL, NULL);
2825 0 : return NULL;
2826 : }
2827 :
2828 0 : return &rtransport->transport;
2829 : }
2830 :
2831 : static void
2832 0 : destroy_ib_device(struct spdk_nvmf_rdma_transport *rtransport,
2833 : struct spdk_nvmf_rdma_device *device)
2834 : {
2835 0 : TAILQ_REMOVE(&rtransport->devices, device, link);
2836 0 : spdk_rdma_utils_free_mem_map(&device->map);
2837 0 : if (device->pd) {
2838 0 : if (!g_nvmf_hooks.get_ibv_pd) {
2839 0 : ibv_dealloc_pd(device->pd);
2840 : }
2841 : }
2842 0 : SPDK_DEBUGLOG(rdma, "IB device [%p] is destroyed.\n", device);
2843 0 : free(device);
2844 0 : }
2845 :
2846 : static void
2847 0 : nvmf_rdma_dump_opts(struct spdk_nvmf_transport *transport, struct spdk_json_write_ctx *w)
2848 : {
2849 : struct spdk_nvmf_rdma_transport *rtransport;
2850 0 : assert(w != NULL);
2851 :
2852 0 : rtransport = SPDK_CONTAINEROF(transport, struct spdk_nvmf_rdma_transport, transport);
2853 0 : spdk_json_write_named_uint32(w, "max_srq_depth", rtransport->rdma_opts.max_srq_depth);
2854 0 : spdk_json_write_named_bool(w, "no_srq", rtransport->rdma_opts.no_srq);
2855 0 : if (rtransport->rdma_opts.no_srq == true) {
2856 0 : spdk_json_write_named_int32(w, "num_cqe", rtransport->rdma_opts.num_cqe);
2857 : }
2858 0 : spdk_json_write_named_int32(w, "acceptor_backlog", rtransport->rdma_opts.acceptor_backlog);
2859 0 : spdk_json_write_named_bool(w, "no_wr_batching", rtransport->rdma_opts.no_wr_batching);
2860 0 : }
2861 :
2862 : static int
2863 0 : nvmf_rdma_destroy(struct spdk_nvmf_transport *transport,
2864 : spdk_nvmf_transport_destroy_done_cb cb_fn, void *cb_arg)
2865 : {
2866 : struct spdk_nvmf_rdma_transport *rtransport;
2867 : struct spdk_nvmf_rdma_port *port, *port_tmp;
2868 : struct spdk_nvmf_rdma_device *device, *device_tmp;
2869 :
2870 0 : rtransport = SPDK_CONTAINEROF(transport, struct spdk_nvmf_rdma_transport, transport);
2871 :
2872 0 : TAILQ_FOREACH_SAFE(port, &rtransport->retry_ports, link, port_tmp) {
2873 0 : TAILQ_REMOVE(&rtransport->retry_ports, port, link);
2874 0 : free(port);
2875 : }
2876 :
2877 0 : TAILQ_FOREACH_SAFE(port, &rtransport->ports, link, port_tmp) {
2878 0 : TAILQ_REMOVE(&rtransport->ports, port, link);
2879 0 : rdma_destroy_id(port->id);
2880 0 : free(port);
2881 : }
2882 :
2883 0 : free_poll_fds(rtransport);
2884 :
2885 0 : if (rtransport->event_channel != NULL) {
2886 0 : rdma_destroy_event_channel(rtransport->event_channel);
2887 : }
2888 :
2889 0 : TAILQ_FOREACH_SAFE(device, &rtransport->devices, link, device_tmp) {
2890 0 : destroy_ib_device(rtransport, device);
2891 : }
2892 :
2893 0 : if (rtransport->data_wr_pool != NULL) {
2894 0 : if (spdk_mempool_count(rtransport->data_wr_pool) != transport->opts.data_wr_pool_size) {
2895 0 : SPDK_ERRLOG("transport wr pool count is %zu but should be %u\n",
2896 : spdk_mempool_count(rtransport->data_wr_pool),
2897 : transport->opts.max_queue_depth * SPDK_NVMF_MAX_SGL_ENTRIES);
2898 : }
2899 : }
2900 :
2901 0 : spdk_mempool_free(rtransport->data_wr_pool);
2902 :
2903 0 : spdk_poller_unregister(&rtransport->accept_poller);
2904 0 : free(rtransport);
2905 :
2906 0 : if (cb_fn) {
2907 0 : cb_fn(cb_arg);
2908 : }
2909 0 : return 0;
2910 : }
2911 :
2912 : static int nvmf_rdma_trid_from_cm_id(struct rdma_cm_id *id,
2913 : struct spdk_nvme_transport_id *trid,
2914 : bool peer);
2915 :
2916 : static bool nvmf_rdma_rescan_devices(struct spdk_nvmf_rdma_transport *rtransport);
2917 :
2918 : static int
2919 0 : nvmf_rdma_listen(struct spdk_nvmf_transport *transport, const struct spdk_nvme_transport_id *trid,
2920 : struct spdk_nvmf_listen_opts *listen_opts)
2921 : {
2922 : struct spdk_nvmf_rdma_transport *rtransport;
2923 : struct spdk_nvmf_rdma_device *device;
2924 : struct spdk_nvmf_rdma_port *port, *tmp_port;
2925 0 : struct addrinfo *res;
2926 0 : struct addrinfo hints;
2927 : int family;
2928 : int rc;
2929 : long int port_val;
2930 0 : bool is_retry = false;
2931 :
2932 0 : if (!strlen(trid->trsvcid)) {
2933 0 : SPDK_ERRLOG("Service id is required\n");
2934 0 : return -EINVAL;
2935 : }
2936 :
2937 0 : rtransport = SPDK_CONTAINEROF(transport, struct spdk_nvmf_rdma_transport, transport);
2938 0 : assert(rtransport->event_channel != NULL);
2939 :
2940 0 : port = calloc(1, sizeof(*port));
2941 0 : if (!port) {
2942 0 : SPDK_ERRLOG("Port allocation failed\n");
2943 0 : return -ENOMEM;
2944 : }
2945 :
2946 0 : port->trid = trid;
2947 :
2948 0 : switch (trid->adrfam) {
2949 0 : case SPDK_NVMF_ADRFAM_IPV4:
2950 0 : family = AF_INET;
2951 0 : break;
2952 0 : case SPDK_NVMF_ADRFAM_IPV6:
2953 0 : family = AF_INET6;
2954 0 : break;
2955 0 : default:
2956 0 : SPDK_ERRLOG("Unhandled ADRFAM %d\n", trid->adrfam);
2957 0 : free(port);
2958 0 : return -EINVAL;
2959 : }
2960 :
2961 0 : memset(&hints, 0, sizeof(hints));
2962 0 : hints.ai_family = family;
2963 0 : hints.ai_flags = AI_NUMERICSERV;
2964 0 : hints.ai_socktype = SOCK_STREAM;
2965 0 : hints.ai_protocol = 0;
2966 :
2967 : /* Range check the trsvcid. Fail in 3 cases:
2968 : * < 0: means that spdk_strtol hit an error
2969 : * 0: this results in ephemeral port which we don't want
2970 : * > 65535: port too high
2971 : */
2972 0 : port_val = spdk_strtol(trid->trsvcid, 10);
2973 0 : if (port_val <= 0 || port_val > 65535) {
2974 0 : SPDK_ERRLOG("invalid trsvcid %s\n", trid->trsvcid);
2975 0 : free(port);
2976 0 : return -EINVAL;
2977 : }
2978 :
2979 0 : rc = getaddrinfo(trid->traddr, trid->trsvcid, &hints, &res);
2980 0 : if (rc) {
2981 0 : SPDK_ERRLOG("getaddrinfo failed: %s (%d)\n", gai_strerror(rc), rc);
2982 0 : free(port);
2983 0 : return -(abs(rc));
2984 : }
2985 :
2986 0 : rc = rdma_create_id(rtransport->event_channel, &port->id, port, RDMA_PS_TCP);
2987 0 : if (rc < 0) {
2988 0 : SPDK_ERRLOG("rdma_create_id() failed\n");
2989 0 : freeaddrinfo(res);
2990 0 : free(port);
2991 0 : return rc;
2992 : }
2993 :
2994 0 : rc = rdma_bind_addr(port->id, res->ai_addr);
2995 0 : freeaddrinfo(res);
2996 :
2997 0 : if (rc < 0) {
2998 0 : TAILQ_FOREACH(tmp_port, &rtransport->retry_ports, link) {
2999 0 : if (spdk_nvme_transport_id_compare(tmp_port->trid, trid) == 0) {
3000 0 : is_retry = true;
3001 0 : break;
3002 : }
3003 : }
3004 0 : if (!is_retry) {
3005 0 : SPDK_ERRLOG("rdma_bind_addr() failed\n");
3006 : }
3007 0 : rdma_destroy_id(port->id);
3008 0 : free(port);
3009 0 : return rc;
3010 : }
3011 :
3012 0 : if (!port->id->verbs) {
3013 0 : SPDK_ERRLOG("ibv_context is null\n");
3014 0 : rdma_destroy_id(port->id);
3015 0 : free(port);
3016 0 : return -1;
3017 : }
3018 :
3019 0 : rc = rdma_listen(port->id, rtransport->rdma_opts.acceptor_backlog);
3020 0 : if (rc < 0) {
3021 0 : SPDK_ERRLOG("rdma_listen() failed\n");
3022 0 : rdma_destroy_id(port->id);
3023 0 : free(port);
3024 0 : return rc;
3025 : }
3026 :
3027 0 : TAILQ_FOREACH(device, &rtransport->devices, link) {
3028 0 : if (device->context == port->id->verbs && device->is_ready) {
3029 0 : port->device = device;
3030 0 : break;
3031 : }
3032 : }
3033 0 : if (!port->device) {
3034 0 : SPDK_ERRLOG("Accepted a connection with verbs %p, but unable to find a corresponding device.\n",
3035 : port->id->verbs);
3036 0 : rdma_destroy_id(port->id);
3037 0 : free(port);
3038 0 : nvmf_rdma_rescan_devices(rtransport);
3039 0 : return -EINVAL;
3040 : }
3041 :
3042 0 : SPDK_NOTICELOG("*** NVMe/RDMA Target Listening on %s port %s ***\n",
3043 : trid->traddr, trid->trsvcid);
3044 :
3045 0 : TAILQ_INSERT_TAIL(&rtransport->ports, port, link);
3046 0 : return 0;
3047 : }
3048 :
3049 : static void
3050 0 : nvmf_rdma_stop_listen_ex(struct spdk_nvmf_transport *transport,
3051 : const struct spdk_nvme_transport_id *trid, bool need_retry)
3052 : {
3053 : struct spdk_nvmf_rdma_transport *rtransport;
3054 : struct spdk_nvmf_rdma_port *port, *tmp;
3055 :
3056 0 : rtransport = SPDK_CONTAINEROF(transport, struct spdk_nvmf_rdma_transport, transport);
3057 :
3058 0 : if (!need_retry) {
3059 0 : TAILQ_FOREACH_SAFE(port, &rtransport->retry_ports, link, tmp) {
3060 0 : if (spdk_nvme_transport_id_compare(port->trid, trid) == 0) {
3061 0 : TAILQ_REMOVE(&rtransport->retry_ports, port, link);
3062 0 : free(port);
3063 : }
3064 : }
3065 : }
3066 :
3067 0 : TAILQ_FOREACH_SAFE(port, &rtransport->ports, link, tmp) {
3068 0 : if (spdk_nvme_transport_id_compare(port->trid, trid) == 0) {
3069 0 : SPDK_DEBUGLOG(rdma, "Port %s:%s removed. need retry: %d\n",
3070 : port->trid->traddr, port->trid->trsvcid, need_retry);
3071 0 : TAILQ_REMOVE(&rtransport->ports, port, link);
3072 0 : rdma_destroy_id(port->id);
3073 0 : port->id = NULL;
3074 0 : port->device = NULL;
3075 0 : if (need_retry) {
3076 0 : TAILQ_INSERT_TAIL(&rtransport->retry_ports, port, link);
3077 : } else {
3078 0 : free(port);
3079 : }
3080 0 : break;
3081 : }
3082 : }
3083 0 : }
3084 :
3085 : static void
3086 0 : nvmf_rdma_stop_listen(struct spdk_nvmf_transport *transport,
3087 : const struct spdk_nvme_transport_id *trid)
3088 : {
3089 0 : nvmf_rdma_stop_listen_ex(transport, trid, false);
3090 0 : }
3091 :
3092 : static void _nvmf_rdma_register_poller_in_group(void *c);
3093 : static void _nvmf_rdma_remove_poller_in_group(void *c);
3094 :
3095 : static bool
3096 0 : nvmf_rdma_all_pollers_management_done(void *c)
3097 : {
3098 0 : struct poller_manage_ctx *ctx = c;
3099 : int counter;
3100 :
3101 0 : counter = __atomic_sub_fetch(ctx->inflight_op_counter, 1, __ATOMIC_SEQ_CST);
3102 0 : SPDK_DEBUGLOG(rdma, "nvmf_rdma_all_pollers_management_done called. counter: %d, poller: %p\n",
3103 : counter, ctx->rpoller);
3104 :
3105 0 : if (counter == 0) {
3106 0 : free((void *)ctx->inflight_op_counter);
3107 : }
3108 0 : free(ctx);
3109 :
3110 0 : return counter == 0;
3111 : }
3112 :
3113 : static int
3114 0 : nvmf_rdma_manage_poller(struct spdk_nvmf_rdma_transport *rtransport,
3115 : struct spdk_nvmf_rdma_device *device, bool *has_inflight, bool is_add)
3116 : {
3117 : struct spdk_nvmf_rdma_poll_group *rgroup;
3118 : struct spdk_nvmf_rdma_poller *rpoller;
3119 : struct spdk_nvmf_poll_group *poll_group;
3120 : struct poller_manage_ctx *ctx;
3121 : bool found;
3122 : int *inflight_counter;
3123 : spdk_msg_fn do_fn;
3124 :
3125 0 : *has_inflight = false;
3126 0 : do_fn = is_add ? _nvmf_rdma_register_poller_in_group : _nvmf_rdma_remove_poller_in_group;
3127 0 : inflight_counter = calloc(1, sizeof(int));
3128 0 : if (!inflight_counter) {
3129 0 : SPDK_ERRLOG("Failed to allocate inflight counter when removing pollers\n");
3130 0 : return -ENOMEM;
3131 : }
3132 :
3133 0 : TAILQ_FOREACH(rgroup, &rtransport->poll_groups, link) {
3134 0 : (*inflight_counter)++;
3135 : }
3136 :
3137 0 : TAILQ_FOREACH(rgroup, &rtransport->poll_groups, link) {
3138 0 : found = false;
3139 0 : TAILQ_FOREACH(rpoller, &rgroup->pollers, link) {
3140 0 : if (rpoller->device == device) {
3141 0 : found = true;
3142 0 : break;
3143 : }
3144 : }
3145 0 : if (found == is_add) {
3146 0 : __atomic_fetch_sub(inflight_counter, 1, __ATOMIC_SEQ_CST);
3147 0 : continue;
3148 : }
3149 :
3150 0 : ctx = calloc(1, sizeof(struct poller_manage_ctx));
3151 0 : if (!ctx) {
3152 0 : SPDK_ERRLOG("Failed to allocate poller_manage_ctx when removing pollers\n");
3153 0 : if (!*has_inflight) {
3154 0 : free(inflight_counter);
3155 : }
3156 0 : return -ENOMEM;
3157 : }
3158 :
3159 0 : ctx->rtransport = rtransport;
3160 0 : ctx->rgroup = rgroup;
3161 0 : ctx->rpoller = rpoller;
3162 0 : ctx->device = device;
3163 0 : ctx->thread = spdk_get_thread();
3164 0 : ctx->inflight_op_counter = inflight_counter;
3165 0 : *has_inflight = true;
3166 :
3167 0 : poll_group = rgroup->group.group;
3168 0 : if (poll_group->thread != spdk_get_thread()) {
3169 0 : spdk_thread_send_msg(poll_group->thread, do_fn, ctx);
3170 : } else {
3171 0 : do_fn(ctx);
3172 : }
3173 : }
3174 :
3175 0 : if (!*has_inflight) {
3176 0 : free(inflight_counter);
3177 : }
3178 :
3179 0 : return 0;
3180 : }
3181 :
3182 : static void nvmf_rdma_handle_device_removal(struct spdk_nvmf_rdma_transport *rtransport,
3183 : struct spdk_nvmf_rdma_device *device);
3184 :
3185 : static struct spdk_nvmf_rdma_device *
3186 0 : nvmf_rdma_find_ib_device(struct spdk_nvmf_rdma_transport *rtransport,
3187 : struct ibv_context *context)
3188 : {
3189 : struct spdk_nvmf_rdma_device *device, *tmp_device;
3190 :
3191 0 : TAILQ_FOREACH_SAFE(device, &rtransport->devices, link, tmp_device) {
3192 0 : if (device->need_destroy) {
3193 0 : continue;
3194 : }
3195 :
3196 0 : if (strcmp(device->context->device->dev_name, context->device->dev_name) == 0) {
3197 0 : return device;
3198 : }
3199 : }
3200 :
3201 0 : return NULL;
3202 : }
3203 :
3204 : static bool
3205 0 : nvmf_rdma_check_devices_context(struct spdk_nvmf_rdma_transport *rtransport,
3206 : struct ibv_context *context)
3207 : {
3208 0 : struct spdk_nvmf_rdma_device *old_device, *new_device;
3209 0 : int rc = 0;
3210 0 : bool has_inflight;
3211 :
3212 0 : old_device = nvmf_rdma_find_ib_device(rtransport, context);
3213 :
3214 0 : if (old_device) {
3215 0 : if (old_device->context != context && !old_device->need_destroy && old_device->is_ready) {
3216 : /* context may not have time to be cleaned when rescan. exactly one context
3217 : * is valid for a device so this context must be invalid and just remove it. */
3218 0 : SPDK_WARNLOG("Device %p has a invalid context %p\n", old_device, old_device->context);
3219 0 : old_device->need_destroy = true;
3220 0 : nvmf_rdma_handle_device_removal(rtransport, old_device);
3221 : }
3222 0 : return false;
3223 : }
3224 :
3225 0 : rc = create_ib_device(rtransport, context, &new_device);
3226 : /* TODO: update transport opts. */
3227 0 : if (rc < 0) {
3228 0 : SPDK_ERRLOG("Failed to create ib device for context: %s(%p)\n",
3229 : ibv_get_device_name(context->device), context);
3230 0 : return false;
3231 : }
3232 :
3233 0 : rc = nvmf_rdma_manage_poller(rtransport, new_device, &has_inflight, true);
3234 0 : if (rc < 0) {
3235 0 : SPDK_ERRLOG("Failed to add poller for device context: %s(%p)\n",
3236 : ibv_get_device_name(context->device), context);
3237 0 : return false;
3238 : }
3239 :
3240 0 : if (has_inflight) {
3241 0 : new_device->is_ready = true;
3242 : }
3243 :
3244 0 : return true;
3245 : }
3246 :
3247 : static bool
3248 0 : nvmf_rdma_rescan_devices(struct spdk_nvmf_rdma_transport *rtransport)
3249 : {
3250 : struct spdk_nvmf_rdma_device *device;
3251 0 : struct ibv_device **ibv_device_list = NULL;
3252 0 : struct ibv_context **contexts = NULL;
3253 0 : int i = 0;
3254 0 : int num_dev = 0;
3255 0 : bool new_create = false, has_new_device = false;
3256 0 : struct ibv_context *tmp_verbs = NULL;
3257 :
3258 : /* do not rescan when any device is destroying, or context may be freed when
3259 : * regenerating the poll fds.
3260 : */
3261 0 : TAILQ_FOREACH(device, &rtransport->devices, link) {
3262 0 : if (device->need_destroy) {
3263 0 : return false;
3264 : }
3265 : }
3266 :
3267 0 : ibv_device_list = ibv_get_device_list(&num_dev);
3268 :
3269 : /* There is a bug in librdmacm. If verbs init failed in rdma_get_devices, it'll be
3270 : * marked as dead verbs and never be init again. So we need to make sure the
3271 : * verbs is available before we call rdma_get_devices. */
3272 0 : if (num_dev >= 0) {
3273 0 : for (i = 0; i < num_dev; i++) {
3274 0 : tmp_verbs = ibv_open_device(ibv_device_list[i]);
3275 0 : if (!tmp_verbs) {
3276 0 : SPDK_WARNLOG("Failed to init ibv device %p, err %d. Skip rescan.\n", ibv_device_list[i], errno);
3277 0 : break;
3278 : }
3279 0 : if (nvmf_rdma_find_ib_device(rtransport, tmp_verbs) == NULL) {
3280 0 : SPDK_DEBUGLOG(rdma, "Find new verbs init ibv device %p(%s).\n", ibv_device_list[i],
3281 : tmp_verbs->device->dev_name);
3282 0 : has_new_device = true;
3283 : }
3284 0 : ibv_close_device(tmp_verbs);
3285 : }
3286 0 : ibv_free_device_list(ibv_device_list);
3287 0 : if (!tmp_verbs || !has_new_device) {
3288 0 : return false;
3289 : }
3290 : }
3291 :
3292 0 : contexts = rdma_get_devices(NULL);
3293 :
3294 0 : for (i = 0; contexts && contexts[i] != NULL; i++) {
3295 0 : new_create |= nvmf_rdma_check_devices_context(rtransport, contexts[i]);
3296 : }
3297 :
3298 0 : if (new_create) {
3299 0 : free_poll_fds(rtransport);
3300 0 : generate_poll_fds(rtransport);
3301 : }
3302 :
3303 0 : if (contexts) {
3304 0 : rdma_free_devices(contexts);
3305 : }
3306 :
3307 0 : return new_create;
3308 : }
3309 :
3310 : static bool
3311 0 : nvmf_rdma_retry_listen_port(struct spdk_nvmf_rdma_transport *rtransport)
3312 : {
3313 : struct spdk_nvmf_rdma_port *port, *tmp_port;
3314 0 : int rc = 0;
3315 0 : bool new_create = false;
3316 :
3317 0 : if (TAILQ_EMPTY(&rtransport->retry_ports)) {
3318 0 : return false;
3319 : }
3320 :
3321 0 : new_create = nvmf_rdma_rescan_devices(rtransport);
3322 :
3323 0 : TAILQ_FOREACH_SAFE(port, &rtransport->retry_ports, link, tmp_port) {
3324 0 : rc = nvmf_rdma_listen(&rtransport->transport, port->trid, NULL);
3325 :
3326 0 : TAILQ_REMOVE(&rtransport->retry_ports, port, link);
3327 0 : if (rc) {
3328 0 : if (new_create) {
3329 0 : SPDK_ERRLOG("Found new IB device but port %s:%s is still failed(%d) to listen.\n",
3330 : port->trid->traddr, port->trid->trsvcid, rc);
3331 : }
3332 0 : TAILQ_INSERT_TAIL(&rtransport->retry_ports, port, link);
3333 0 : break;
3334 : } else {
3335 0 : SPDK_NOTICELOG("Port %s:%s come back\n", port->trid->traddr, port->trid->trsvcid);
3336 0 : free(port);
3337 : }
3338 : }
3339 :
3340 0 : return true;
3341 : }
3342 :
3343 : static void
3344 0 : nvmf_rdma_qpair_process_pending(struct spdk_nvmf_rdma_transport *rtransport,
3345 : struct spdk_nvmf_rdma_qpair *rqpair, bool drain)
3346 : {
3347 : struct spdk_nvmf_request *req, *tmp;
3348 : struct spdk_nvmf_rdma_request *rdma_req, *req_tmp;
3349 : struct spdk_nvmf_rdma_resources *resources;
3350 :
3351 : /* First process requests which are waiting for response to be sent */
3352 0 : STAILQ_FOREACH_SAFE(rdma_req, &rqpair->pending_rdma_send_queue, state_link, req_tmp) {
3353 0 : if (nvmf_rdma_request_process(rtransport, rdma_req) == false && drain == false) {
3354 0 : break;
3355 : }
3356 : }
3357 :
3358 : /* We process I/O in the data transfer pending queue at the highest priority. */
3359 0 : STAILQ_FOREACH_SAFE(rdma_req, &rqpair->pending_rdma_read_queue, state_link, req_tmp) {
3360 0 : if (rdma_req->state != RDMA_REQUEST_STATE_DATA_TRANSFER_TO_CONTROLLER_PENDING) {
3361 : /* Requests in this queue might be in state RDMA_REQUEST_STATE_TRANSFERRING_HOST_TO_CONTROLLER,
3362 : * they are transmitting data over network but we keep them in the list to guarantee
3363 : * fair processing. */
3364 0 : continue;
3365 : }
3366 0 : if (nvmf_rdma_request_process(rtransport, rdma_req) == false && drain == false) {
3367 0 : break;
3368 : }
3369 : }
3370 :
3371 : /* Then RDMA writes since reads have stronger restrictions than writes */
3372 0 : STAILQ_FOREACH_SAFE(rdma_req, &rqpair->pending_rdma_write_queue, state_link, req_tmp) {
3373 0 : if (nvmf_rdma_request_process(rtransport, rdma_req) == false && drain == false) {
3374 0 : break;
3375 : }
3376 : }
3377 :
3378 : /* Then we handle request waiting on memory buffers. */
3379 0 : STAILQ_FOREACH_SAFE(req, &rqpair->poller->group->group.pending_buf_queue, buf_link, tmp) {
3380 0 : rdma_req = SPDK_CONTAINEROF(req, struct spdk_nvmf_rdma_request, req);
3381 0 : if (nvmf_rdma_request_process(rtransport, rdma_req) == false && drain == false) {
3382 0 : break;
3383 : }
3384 : }
3385 :
3386 0 : resources = rqpair->resources;
3387 0 : while (!STAILQ_EMPTY(&resources->free_queue) && !STAILQ_EMPTY(&resources->incoming_queue)) {
3388 0 : rdma_req = STAILQ_FIRST(&resources->free_queue);
3389 0 : STAILQ_REMOVE_HEAD(&resources->free_queue, state_link);
3390 0 : rdma_req->recv = STAILQ_FIRST(&resources->incoming_queue);
3391 0 : STAILQ_REMOVE_HEAD(&resources->incoming_queue, link);
3392 :
3393 0 : if (rqpair->srq != NULL) {
3394 0 : rdma_req->req.qpair = &rdma_req->recv->qpair->qpair;
3395 0 : rdma_req->recv->qpair->qd++;
3396 : } else {
3397 0 : rqpair->qd++;
3398 : }
3399 :
3400 0 : rdma_req->receive_tsc = rdma_req->recv->receive_tsc;
3401 0 : rdma_req->state = RDMA_REQUEST_STATE_NEW;
3402 0 : if (nvmf_rdma_request_process(rtransport, rdma_req) == false) {
3403 0 : break;
3404 : }
3405 : }
3406 0 : if (!STAILQ_EMPTY(&resources->incoming_queue) && STAILQ_EMPTY(&resources->free_queue)) {
3407 0 : rqpair->poller->stat.pending_free_request++;
3408 : }
3409 0 : }
3410 :
3411 : static void
3412 0 : nvmf_rdma_poller_process_pending_buf_queue(struct spdk_nvmf_rdma_transport *rtransport,
3413 : struct spdk_nvmf_rdma_poller *rpoller)
3414 : {
3415 : struct spdk_nvmf_request *req, *tmp;
3416 : struct spdk_nvmf_rdma_request *rdma_req;
3417 :
3418 0 : STAILQ_FOREACH_SAFE(req, &rpoller->group->group.pending_buf_queue, buf_link, tmp) {
3419 0 : rdma_req = SPDK_CONTAINEROF(req, struct spdk_nvmf_rdma_request, req);
3420 0 : if (nvmf_rdma_request_process(rtransport, rdma_req) == false) {
3421 0 : break;
3422 : }
3423 : }
3424 0 : }
3425 :
3426 : static inline bool
3427 0 : nvmf_rdma_can_ignore_last_wqe_reached(struct spdk_nvmf_rdma_device *device)
3428 : {
3429 : /* iWARP transport and SoftRoCE driver don't support LAST_WQE_REACHED ibv async event */
3430 0 : return nvmf_rdma_is_rxe_device(device) ||
3431 0 : device->context->device->transport_type == IBV_TRANSPORT_IWARP;
3432 : }
3433 :
3434 : static void
3435 0 : nvmf_rdma_destroy_drained_qpair(struct spdk_nvmf_rdma_qpair *rqpair)
3436 : {
3437 0 : struct spdk_nvmf_rdma_transport *rtransport = SPDK_CONTAINEROF(rqpair->qpair.transport,
3438 : struct spdk_nvmf_rdma_transport, transport);
3439 :
3440 0 : nvmf_rdma_qpair_process_pending(rtransport, rqpair, true);
3441 :
3442 : /* nvmf_rdma_close_qpair is not called */
3443 0 : if (!rqpair->to_close) {
3444 0 : return;
3445 : }
3446 :
3447 : /* device is already destroyed and we should force destroy this qpair. */
3448 0 : if (rqpair->poller && rqpair->poller->need_destroy) {
3449 0 : nvmf_rdma_qpair_destroy(rqpair);
3450 0 : return;
3451 : }
3452 :
3453 : /* In non SRQ path, we will reach rqpair->max_queue_depth. In SRQ path, we will get the last_wqe event. */
3454 0 : if (rqpair->current_send_depth != 0) {
3455 0 : return;
3456 : }
3457 :
3458 0 : if (rqpair->srq == NULL && rqpair->current_recv_depth != rqpair->max_queue_depth) {
3459 0 : return;
3460 : }
3461 :
3462 0 : if (rqpair->srq != NULL && rqpair->last_wqe_reached == false &&
3463 0 : !nvmf_rdma_can_ignore_last_wqe_reached(rqpair->device)) {
3464 0 : return;
3465 : }
3466 :
3467 0 : assert(rqpair->qpair.state == SPDK_NVMF_QPAIR_ERROR);
3468 :
3469 0 : nvmf_rdma_qpair_destroy(rqpair);
3470 : }
3471 :
3472 : static int
3473 0 : nvmf_rdma_disconnect(struct rdma_cm_event *evt, bool *event_acked)
3474 : {
3475 : struct spdk_nvmf_qpair *qpair;
3476 : struct spdk_nvmf_rdma_qpair *rqpair;
3477 :
3478 0 : if (evt->id == NULL) {
3479 0 : SPDK_ERRLOG("disconnect request: missing cm_id\n");
3480 0 : return -1;
3481 : }
3482 :
3483 0 : qpair = evt->id->context;
3484 0 : if (qpair == NULL) {
3485 0 : SPDK_ERRLOG("disconnect request: no active connection\n");
3486 0 : return -1;
3487 : }
3488 :
3489 0 : rdma_ack_cm_event(evt);
3490 0 : *event_acked = true;
3491 :
3492 0 : rqpair = SPDK_CONTAINEROF(qpair, struct spdk_nvmf_rdma_qpair, qpair);
3493 :
3494 0 : spdk_trace_record(TRACE_RDMA_QP_DISCONNECT, 0, 0, (uintptr_t)rqpair);
3495 :
3496 0 : spdk_nvmf_qpair_disconnect(&rqpair->qpair);
3497 :
3498 0 : return 0;
3499 : }
3500 :
3501 : #ifdef DEBUG
3502 : static const char *CM_EVENT_STR[] = {
3503 : "RDMA_CM_EVENT_ADDR_RESOLVED",
3504 : "RDMA_CM_EVENT_ADDR_ERROR",
3505 : "RDMA_CM_EVENT_ROUTE_RESOLVED",
3506 : "RDMA_CM_EVENT_ROUTE_ERROR",
3507 : "RDMA_CM_EVENT_CONNECT_REQUEST",
3508 : "RDMA_CM_EVENT_CONNECT_RESPONSE",
3509 : "RDMA_CM_EVENT_CONNECT_ERROR",
3510 : "RDMA_CM_EVENT_UNREACHABLE",
3511 : "RDMA_CM_EVENT_REJECTED",
3512 : "RDMA_CM_EVENT_ESTABLISHED",
3513 : "RDMA_CM_EVENT_DISCONNECTED",
3514 : "RDMA_CM_EVENT_DEVICE_REMOVAL",
3515 : "RDMA_CM_EVENT_MULTICAST_JOIN",
3516 : "RDMA_CM_EVENT_MULTICAST_ERROR",
3517 : "RDMA_CM_EVENT_ADDR_CHANGE",
3518 : "RDMA_CM_EVENT_TIMEWAIT_EXIT"
3519 : };
3520 : #endif /* DEBUG */
3521 :
3522 : static void
3523 0 : nvmf_rdma_disconnect_qpairs_on_port(struct spdk_nvmf_rdma_transport *rtransport,
3524 : struct spdk_nvmf_rdma_port *port)
3525 : {
3526 : struct spdk_nvmf_rdma_poll_group *rgroup;
3527 : struct spdk_nvmf_rdma_poller *rpoller;
3528 : struct spdk_nvmf_rdma_qpair *rqpair;
3529 :
3530 0 : TAILQ_FOREACH(rgroup, &rtransport->poll_groups, link) {
3531 0 : TAILQ_FOREACH(rpoller, &rgroup->pollers, link) {
3532 0 : RB_FOREACH(rqpair, qpairs_tree, &rpoller->qpairs) {
3533 0 : if (rqpair->listen_id == port->id) {
3534 0 : spdk_nvmf_qpair_disconnect(&rqpair->qpair);
3535 : }
3536 : }
3537 : }
3538 : }
3539 0 : }
3540 :
3541 : static bool
3542 0 : nvmf_rdma_handle_cm_event_addr_change(struct spdk_nvmf_transport *transport,
3543 : struct rdma_cm_event *event)
3544 : {
3545 : const struct spdk_nvme_transport_id *trid;
3546 : struct spdk_nvmf_rdma_port *port;
3547 : struct spdk_nvmf_rdma_transport *rtransport;
3548 0 : bool event_acked = false;
3549 :
3550 0 : rtransport = SPDK_CONTAINEROF(transport, struct spdk_nvmf_rdma_transport, transport);
3551 0 : TAILQ_FOREACH(port, &rtransport->ports, link) {
3552 0 : if (port->id == event->id) {
3553 0 : SPDK_ERRLOG("ADDR_CHANGE: IP %s:%s migrated\n", port->trid->traddr, port->trid->trsvcid);
3554 0 : rdma_ack_cm_event(event);
3555 0 : event_acked = true;
3556 0 : trid = port->trid;
3557 0 : break;
3558 : }
3559 : }
3560 :
3561 0 : if (event_acked) {
3562 0 : nvmf_rdma_disconnect_qpairs_on_port(rtransport, port);
3563 :
3564 0 : nvmf_rdma_stop_listen(transport, trid);
3565 0 : nvmf_rdma_listen(transport, trid, NULL);
3566 : }
3567 :
3568 0 : return event_acked;
3569 : }
3570 :
3571 : static void
3572 0 : nvmf_rdma_handle_device_removal(struct spdk_nvmf_rdma_transport *rtransport,
3573 : struct spdk_nvmf_rdma_device *device)
3574 : {
3575 : struct spdk_nvmf_rdma_port *port, *port_tmp;
3576 : int rc;
3577 0 : bool has_inflight;
3578 :
3579 0 : rc = nvmf_rdma_manage_poller(rtransport, device, &has_inflight, false);
3580 0 : if (rc) {
3581 0 : SPDK_ERRLOG("Failed to handle device removal, rc %d\n", rc);
3582 0 : return;
3583 : }
3584 :
3585 0 : if (!has_inflight) {
3586 : /* no pollers, destroy the device */
3587 0 : device->ready_to_destroy = true;
3588 0 : spdk_thread_send_msg(spdk_get_thread(), _nvmf_rdma_remove_destroyed_device, rtransport);
3589 : }
3590 :
3591 0 : TAILQ_FOREACH_SAFE(port, &rtransport->ports, link, port_tmp) {
3592 0 : if (port->device == device) {
3593 0 : SPDK_NOTICELOG("Port %s:%s on device %s is being removed.\n",
3594 : port->trid->traddr,
3595 : port->trid->trsvcid,
3596 : ibv_get_device_name(port->device->context->device));
3597 :
3598 : /* keep NVMF listener and only destroy structures of the
3599 : * RDMA transport. when the device comes back we can retry listening
3600 : * and the application's workflow will not be interrupted.
3601 : */
3602 0 : nvmf_rdma_stop_listen_ex(&rtransport->transport, port->trid, true);
3603 : }
3604 : }
3605 : }
3606 :
3607 : static void
3608 0 : nvmf_rdma_handle_cm_event_port_removal(struct spdk_nvmf_transport *transport,
3609 : struct rdma_cm_event *event)
3610 : {
3611 : struct spdk_nvmf_rdma_port *port, *tmp_port;
3612 : struct spdk_nvmf_rdma_transport *rtransport;
3613 :
3614 0 : port = event->id->context;
3615 0 : rtransport = SPDK_CONTAINEROF(transport, struct spdk_nvmf_rdma_transport, transport);
3616 :
3617 0 : rdma_ack_cm_event(event);
3618 :
3619 : /* if device removal happens during ctrl qpair disconnecting, it's possible that we receive
3620 : * an DEVICE_REMOVAL event on qpair but the id->qp is just NULL. So we should make sure that
3621 : * we are handling a port event here.
3622 : */
3623 0 : TAILQ_FOREACH(tmp_port, &rtransport->ports, link) {
3624 0 : if (port == tmp_port && port->device && !port->device->need_destroy) {
3625 0 : port->device->need_destroy = true;
3626 0 : nvmf_rdma_handle_device_removal(rtransport, port->device);
3627 : }
3628 : }
3629 0 : }
3630 :
3631 : static void
3632 0 : nvmf_process_cm_events(struct spdk_nvmf_transport *transport, uint32_t max_events)
3633 : {
3634 : struct spdk_nvmf_rdma_transport *rtransport;
3635 0 : struct rdma_cm_event *event;
3636 : uint32_t i;
3637 : int rc;
3638 0 : bool event_acked;
3639 :
3640 0 : rtransport = SPDK_CONTAINEROF(transport, struct spdk_nvmf_rdma_transport, transport);
3641 :
3642 0 : if (rtransport->event_channel == NULL) {
3643 0 : return;
3644 : }
3645 :
3646 0 : for (i = 0; i < max_events; i++) {
3647 0 : event_acked = false;
3648 0 : rc = rdma_get_cm_event(rtransport->event_channel, &event);
3649 0 : if (rc) {
3650 0 : if (errno != EAGAIN && errno != EWOULDBLOCK) {
3651 0 : SPDK_ERRLOG("Acceptor Event Error: %s\n", spdk_strerror(errno));
3652 : }
3653 0 : break;
3654 : }
3655 :
3656 0 : SPDK_DEBUGLOG(rdma, "Acceptor Event: %s\n", CM_EVENT_STR[event->event]);
3657 :
3658 0 : spdk_trace_record(TRACE_RDMA_CM_ASYNC_EVENT, 0, 0, 0, event->event);
3659 :
3660 0 : switch (event->event) {
3661 0 : case RDMA_CM_EVENT_ADDR_RESOLVED:
3662 : case RDMA_CM_EVENT_ADDR_ERROR:
3663 : case RDMA_CM_EVENT_ROUTE_RESOLVED:
3664 : case RDMA_CM_EVENT_ROUTE_ERROR:
3665 : /* No action required. The target never attempts to resolve routes. */
3666 0 : break;
3667 0 : case RDMA_CM_EVENT_CONNECT_REQUEST:
3668 0 : rc = nvmf_rdma_connect(transport, event);
3669 0 : if (rc < 0) {
3670 0 : SPDK_ERRLOG("Unable to process connect event. rc: %d\n", rc);
3671 0 : break;
3672 : }
3673 0 : break;
3674 0 : case RDMA_CM_EVENT_CONNECT_RESPONSE:
3675 : /* The target never initiates a new connection. So this will not occur. */
3676 0 : break;
3677 0 : case RDMA_CM_EVENT_CONNECT_ERROR:
3678 : /* Can this happen? The docs say it can, but not sure what causes it. */
3679 0 : break;
3680 0 : case RDMA_CM_EVENT_UNREACHABLE:
3681 : case RDMA_CM_EVENT_REJECTED:
3682 : /* These only occur on the client side. */
3683 0 : break;
3684 0 : case RDMA_CM_EVENT_ESTABLISHED:
3685 : /* TODO: Should we be waiting for this event anywhere? */
3686 0 : break;
3687 0 : case RDMA_CM_EVENT_DISCONNECTED:
3688 0 : rc = nvmf_rdma_disconnect(event, &event_acked);
3689 0 : if (rc < 0) {
3690 0 : SPDK_ERRLOG("Unable to process disconnect event. rc: %d\n", rc);
3691 0 : break;
3692 : }
3693 0 : break;
3694 0 : case RDMA_CM_EVENT_DEVICE_REMOVAL:
3695 : /* In case of device removal, kernel IB part triggers IBV_EVENT_DEVICE_FATAL
3696 : * which triggers RDMA_CM_EVENT_DEVICE_REMOVAL on all cma_id’s.
3697 : * Once these events are sent to SPDK, we should release all IB resources and
3698 : * don't make attempts to call any ibv_query/modify/create functions. We can only call
3699 : * ibv_destroy* functions to release user space memory allocated by IB. All kernel
3700 : * resources are already cleaned. */
3701 0 : if (event->id->qp) {
3702 : /* If rdma_cm event has a valid `qp` pointer then the event refers to the
3703 : * corresponding qpair. Otherwise the event refers to a listening device. */
3704 0 : rc = nvmf_rdma_disconnect(event, &event_acked);
3705 0 : if (rc < 0) {
3706 0 : SPDK_ERRLOG("Unable to process disconnect event. rc: %d\n", rc);
3707 0 : break;
3708 : }
3709 : } else {
3710 0 : nvmf_rdma_handle_cm_event_port_removal(transport, event);
3711 0 : event_acked = true;
3712 : }
3713 0 : break;
3714 0 : case RDMA_CM_EVENT_MULTICAST_JOIN:
3715 : case RDMA_CM_EVENT_MULTICAST_ERROR:
3716 : /* Multicast is not used */
3717 0 : break;
3718 0 : case RDMA_CM_EVENT_ADDR_CHANGE:
3719 0 : event_acked = nvmf_rdma_handle_cm_event_addr_change(transport, event);
3720 0 : break;
3721 0 : case RDMA_CM_EVENT_TIMEWAIT_EXIT:
3722 : /* For now, do nothing. The target never re-uses queue pairs. */
3723 0 : break;
3724 0 : default:
3725 0 : SPDK_ERRLOG("Unexpected Acceptor Event [%d]\n", event->event);
3726 0 : break;
3727 : }
3728 0 : if (!event_acked) {
3729 0 : rdma_ack_cm_event(event);
3730 : }
3731 : }
3732 : }
3733 :
3734 : static void
3735 0 : nvmf_rdma_handle_last_wqe_reached(struct spdk_nvmf_rdma_qpair *rqpair)
3736 : {
3737 0 : rqpair->last_wqe_reached = true;
3738 0 : nvmf_rdma_destroy_drained_qpair(rqpair);
3739 0 : }
3740 :
3741 : static void
3742 0 : nvmf_rdma_qpair_process_ibv_event(void *ctx)
3743 : {
3744 0 : struct spdk_nvmf_rdma_ibv_event_ctx *event_ctx = ctx;
3745 :
3746 0 : if (event_ctx->rqpair) {
3747 0 : STAILQ_REMOVE(&event_ctx->rqpair->ibv_events, event_ctx, spdk_nvmf_rdma_ibv_event_ctx, link);
3748 0 : if (event_ctx->cb_fn) {
3749 0 : event_ctx->cb_fn(event_ctx->rqpair);
3750 : }
3751 : }
3752 0 : free(event_ctx);
3753 0 : }
3754 :
3755 : static int
3756 0 : nvmf_rdma_send_qpair_async_event(struct spdk_nvmf_rdma_qpair *rqpair,
3757 : spdk_nvmf_rdma_qpair_ibv_event fn)
3758 : {
3759 : struct spdk_nvmf_rdma_ibv_event_ctx *ctx;
3760 0 : struct spdk_thread *thr = NULL;
3761 : int rc;
3762 :
3763 0 : if (rqpair->qpair.group) {
3764 0 : thr = rqpair->qpair.group->thread;
3765 0 : } else if (rqpair->destruct_channel) {
3766 0 : thr = spdk_io_channel_get_thread(rqpair->destruct_channel);
3767 : }
3768 :
3769 0 : if (!thr) {
3770 0 : SPDK_DEBUGLOG(rdma, "rqpair %p has no thread\n", rqpair);
3771 0 : return -EINVAL;
3772 : }
3773 :
3774 0 : ctx = calloc(1, sizeof(*ctx));
3775 0 : if (!ctx) {
3776 0 : return -ENOMEM;
3777 : }
3778 :
3779 0 : ctx->rqpair = rqpair;
3780 0 : ctx->cb_fn = fn;
3781 0 : STAILQ_INSERT_TAIL(&rqpair->ibv_events, ctx, link);
3782 :
3783 0 : rc = spdk_thread_send_msg(thr, nvmf_rdma_qpair_process_ibv_event, ctx);
3784 0 : if (rc) {
3785 0 : STAILQ_REMOVE(&rqpair->ibv_events, ctx, spdk_nvmf_rdma_ibv_event_ctx, link);
3786 0 : free(ctx);
3787 : }
3788 :
3789 0 : return rc;
3790 : }
3791 :
3792 : static int
3793 0 : nvmf_process_ib_event(struct spdk_nvmf_rdma_device *device)
3794 : {
3795 : int rc;
3796 0 : struct spdk_nvmf_rdma_qpair *rqpair = NULL;
3797 0 : struct ibv_async_event event;
3798 :
3799 0 : rc = ibv_get_async_event(device->context, &event);
3800 :
3801 0 : if (rc) {
3802 : /* In non-blocking mode -1 means there are no events available */
3803 0 : return rc;
3804 : }
3805 :
3806 0 : switch (event.event_type) {
3807 0 : case IBV_EVENT_QP_FATAL:
3808 : case IBV_EVENT_QP_LAST_WQE_REACHED:
3809 : case IBV_EVENT_QP_REQ_ERR:
3810 : case IBV_EVENT_QP_ACCESS_ERR:
3811 : case IBV_EVENT_COMM_EST:
3812 : case IBV_EVENT_PATH_MIG:
3813 : case IBV_EVENT_PATH_MIG_ERR:
3814 0 : rqpair = event.element.qp->qp_context;
3815 0 : if (!rqpair) {
3816 : /* Any QP event for NVMe-RDMA initiator may be returned. */
3817 0 : SPDK_NOTICELOG("Async QP event for unknown QP: %s\n",
3818 : ibv_event_type_str(event.event_type));
3819 0 : break;
3820 : }
3821 :
3822 0 : switch (event.event_type) {
3823 0 : case IBV_EVENT_QP_FATAL:
3824 0 : SPDK_ERRLOG("Fatal event received for rqpair %p\n", rqpair);
3825 0 : spdk_trace_record(TRACE_RDMA_IBV_ASYNC_EVENT, 0, 0,
3826 : (uintptr_t)rqpair, event.event_type);
3827 0 : rqpair->ibv_in_error_state = true;
3828 0 : spdk_nvmf_qpair_disconnect(&rqpair->qpair);
3829 0 : break;
3830 0 : case IBV_EVENT_QP_LAST_WQE_REACHED:
3831 : /* This event only occurs for shared receive queues. */
3832 0 : SPDK_DEBUGLOG(rdma, "Last WQE reached event received for rqpair %p\n", rqpair);
3833 0 : rc = nvmf_rdma_send_qpair_async_event(rqpair, nvmf_rdma_handle_last_wqe_reached);
3834 0 : if (rc) {
3835 0 : SPDK_WARNLOG("Failed to send LAST_WQE_REACHED event. rqpair %p, err %d\n", rqpair, rc);
3836 0 : rqpair->last_wqe_reached = true;
3837 : }
3838 0 : break;
3839 0 : case IBV_EVENT_QP_REQ_ERR:
3840 : case IBV_EVENT_QP_ACCESS_ERR:
3841 : case IBV_EVENT_COMM_EST:
3842 : case IBV_EVENT_PATH_MIG:
3843 : case IBV_EVENT_PATH_MIG_ERR:
3844 0 : SPDK_NOTICELOG("Async QP event: %s\n",
3845 : ibv_event_type_str(event.event_type));
3846 0 : spdk_trace_record(TRACE_RDMA_IBV_ASYNC_EVENT, 0, 0,
3847 : (uintptr_t)rqpair, event.event_type);
3848 0 : rqpair->ibv_in_error_state = true;
3849 0 : break;
3850 0 : default:
3851 0 : break;
3852 : }
3853 0 : break;
3854 0 : case IBV_EVENT_DEVICE_FATAL:
3855 0 : SPDK_ERRLOG("Device Fatal event[%s] received on %s. device: %p\n",
3856 : ibv_event_type_str(event.event_type), ibv_get_device_name(device->context->device), device);
3857 0 : device->need_destroy = true;
3858 0 : break;
3859 0 : case IBV_EVENT_CQ_ERR:
3860 : case IBV_EVENT_PORT_ACTIVE:
3861 : case IBV_EVENT_PORT_ERR:
3862 : case IBV_EVENT_LID_CHANGE:
3863 : case IBV_EVENT_PKEY_CHANGE:
3864 : case IBV_EVENT_SM_CHANGE:
3865 : case IBV_EVENT_SRQ_ERR:
3866 : case IBV_EVENT_SRQ_LIMIT_REACHED:
3867 : case IBV_EVENT_CLIENT_REREGISTER:
3868 : case IBV_EVENT_GID_CHANGE:
3869 : case IBV_EVENT_SQ_DRAINED:
3870 : default:
3871 0 : SPDK_NOTICELOG("Async event: %s\n",
3872 : ibv_event_type_str(event.event_type));
3873 0 : spdk_trace_record(TRACE_RDMA_IBV_ASYNC_EVENT, 0, 0, 0, event.event_type);
3874 0 : break;
3875 : }
3876 0 : ibv_ack_async_event(&event);
3877 :
3878 0 : return 0;
3879 : }
3880 :
3881 : static void
3882 0 : nvmf_process_ib_events(struct spdk_nvmf_rdma_device *device, uint32_t max_events)
3883 : {
3884 0 : int rc = 0;
3885 0 : uint32_t i = 0;
3886 :
3887 0 : for (i = 0; i < max_events; i++) {
3888 0 : rc = nvmf_process_ib_event(device);
3889 0 : if (rc) {
3890 0 : break;
3891 : }
3892 : }
3893 :
3894 0 : SPDK_DEBUGLOG(rdma, "Device %s: %u events processed\n", device->context->device->name, i);
3895 0 : }
3896 :
3897 : static int
3898 0 : nvmf_rdma_accept(void *ctx)
3899 : {
3900 0 : int nfds, i = 0;
3901 0 : struct spdk_nvmf_transport *transport = ctx;
3902 : struct spdk_nvmf_rdma_transport *rtransport;
3903 : struct spdk_nvmf_rdma_device *device, *tmp;
3904 : uint32_t count;
3905 : short revents;
3906 : bool do_retry;
3907 :
3908 0 : rtransport = SPDK_CONTAINEROF(transport, struct spdk_nvmf_rdma_transport, transport);
3909 0 : do_retry = nvmf_rdma_retry_listen_port(rtransport);
3910 :
3911 0 : count = nfds = poll(rtransport->poll_fds, rtransport->npoll_fds, 0);
3912 :
3913 0 : if (nfds <= 0) {
3914 0 : return do_retry ? SPDK_POLLER_BUSY : SPDK_POLLER_IDLE;
3915 : }
3916 :
3917 : /* The first poll descriptor is RDMA CM event */
3918 0 : if (rtransport->poll_fds[i++].revents & POLLIN) {
3919 0 : nvmf_process_cm_events(transport, NVMF_RDMA_MAX_EVENTS_PER_POLL);
3920 0 : nfds--;
3921 : }
3922 :
3923 0 : if (nfds == 0) {
3924 0 : return SPDK_POLLER_BUSY;
3925 : }
3926 :
3927 : /* Second and subsequent poll descriptors are IB async events */
3928 0 : TAILQ_FOREACH_SAFE(device, &rtransport->devices, link, tmp) {
3929 0 : revents = rtransport->poll_fds[i++].revents;
3930 0 : if (revents & POLLIN) {
3931 0 : if (spdk_likely(!device->need_destroy)) {
3932 0 : nvmf_process_ib_events(device, NVMF_RDMA_MAX_EVENTS_PER_POLL);
3933 0 : if (spdk_unlikely(device->need_destroy)) {
3934 0 : nvmf_rdma_handle_device_removal(rtransport, device);
3935 : }
3936 : }
3937 0 : nfds--;
3938 0 : } else if (revents & POLLNVAL || revents & POLLHUP) {
3939 0 : SPDK_ERRLOG("Receive unknown revent %x on device %p\n", (int)revents, device);
3940 0 : nfds--;
3941 : }
3942 : }
3943 : /* check all flagged fd's have been served */
3944 0 : assert(nfds == 0);
3945 :
3946 0 : return count > 0 ? SPDK_POLLER_BUSY : SPDK_POLLER_IDLE;
3947 : }
3948 :
3949 : static void
3950 0 : nvmf_rdma_cdata_init(struct spdk_nvmf_transport *transport, struct spdk_nvmf_subsystem *subsystem,
3951 : struct spdk_nvmf_ctrlr_data *cdata)
3952 : {
3953 0 : cdata->nvmf_specific.msdbd = NVMF_DEFAULT_MSDBD;
3954 :
3955 : /* Disable in-capsule data transfer for RDMA controller when dif_insert_or_strip is enabled
3956 : since in-capsule data only works with NVME drives that support SGL memory layout */
3957 0 : if (transport->opts.dif_insert_or_strip) {
3958 0 : cdata->nvmf_specific.ioccsz = sizeof(struct spdk_nvme_cmd) / 16;
3959 : }
3960 :
3961 0 : if (cdata->nvmf_specific.ioccsz > ((sizeof(struct spdk_nvme_cmd) + 0x1000) / 16)) {
3962 0 : SPDK_WARNLOG("RDMA is configured to support up to 16 SGL entries while in capsule"
3963 : " data is greater than 4KiB.\n");
3964 0 : SPDK_WARNLOG("When used in conjunction with the NVMe-oF initiator from the Linux "
3965 : "kernel between versions 5.4 and 5.12 data corruption may occur for "
3966 : "writes that are not a multiple of 4KiB in size.\n");
3967 : }
3968 0 : }
3969 :
3970 : static void
3971 0 : nvmf_rdma_discover(struct spdk_nvmf_transport *transport,
3972 : struct spdk_nvme_transport_id *trid,
3973 : struct spdk_nvmf_discovery_log_page_entry *entry)
3974 : {
3975 0 : entry->trtype = SPDK_NVMF_TRTYPE_RDMA;
3976 0 : entry->adrfam = trid->adrfam;
3977 0 : entry->treq.secure_channel = SPDK_NVMF_TREQ_SECURE_CHANNEL_NOT_REQUIRED;
3978 :
3979 0 : spdk_strcpy_pad(entry->trsvcid, trid->trsvcid, sizeof(entry->trsvcid), ' ');
3980 0 : spdk_strcpy_pad(entry->traddr, trid->traddr, sizeof(entry->traddr), ' ');
3981 :
3982 0 : entry->tsas.rdma.rdma_qptype = SPDK_NVMF_RDMA_QPTYPE_RELIABLE_CONNECTED;
3983 0 : entry->tsas.rdma.rdma_prtype = SPDK_NVMF_RDMA_PRTYPE_NONE;
3984 0 : entry->tsas.rdma.rdma_cms = SPDK_NVMF_RDMA_CMS_RDMA_CM;
3985 0 : }
3986 :
3987 : static int
3988 0 : nvmf_rdma_poller_create(struct spdk_nvmf_rdma_transport *rtransport,
3989 : struct spdk_nvmf_rdma_poll_group *rgroup, struct spdk_nvmf_rdma_device *device,
3990 : struct spdk_nvmf_rdma_poller **out_poller)
3991 : {
3992 : struct spdk_nvmf_rdma_poller *poller;
3993 0 : struct spdk_rdma_provider_srq_init_attr srq_init_attr;
3994 0 : struct spdk_nvmf_rdma_resource_opts opts;
3995 : int num_cqe;
3996 :
3997 0 : poller = calloc(1, sizeof(*poller));
3998 0 : if (!poller) {
3999 0 : SPDK_ERRLOG("Unable to allocate memory for new RDMA poller\n");
4000 0 : return -1;
4001 : }
4002 :
4003 0 : poller->device = device;
4004 0 : poller->group = rgroup;
4005 0 : *out_poller = poller;
4006 :
4007 0 : RB_INIT(&poller->qpairs);
4008 0 : STAILQ_INIT(&poller->qpairs_pending_send);
4009 0 : STAILQ_INIT(&poller->qpairs_pending_recv);
4010 :
4011 0 : TAILQ_INSERT_TAIL(&rgroup->pollers, poller, link);
4012 0 : SPDK_DEBUGLOG(rdma, "Create poller %p on device %p in poll group %p.\n", poller, device, rgroup);
4013 0 : if (rtransport->rdma_opts.no_srq == false && device->num_srq < device->attr.max_srq) {
4014 0 : if ((int)rtransport->rdma_opts.max_srq_depth > device->attr.max_srq_wr) {
4015 0 : SPDK_WARNLOG("Requested SRQ depth %u, max supported by dev %s is %d\n",
4016 : rtransport->rdma_opts.max_srq_depth, device->context->device->name, device->attr.max_srq_wr);
4017 : }
4018 0 : poller->max_srq_depth = spdk_min((int)rtransport->rdma_opts.max_srq_depth, device->attr.max_srq_wr);
4019 :
4020 0 : device->num_srq++;
4021 0 : memset(&srq_init_attr, 0, sizeof(srq_init_attr));
4022 0 : srq_init_attr.pd = device->pd;
4023 0 : srq_init_attr.stats = &poller->stat.qp_stats.recv;
4024 0 : srq_init_attr.srq_init_attr.attr.max_wr = poller->max_srq_depth;
4025 0 : srq_init_attr.srq_init_attr.attr.max_sge = spdk_min(device->attr.max_sge, NVMF_DEFAULT_RX_SGE);
4026 0 : poller->srq = spdk_rdma_provider_srq_create(&srq_init_attr);
4027 0 : if (!poller->srq) {
4028 0 : SPDK_ERRLOG("Unable to create shared receive queue, errno %d\n", errno);
4029 0 : return -1;
4030 : }
4031 :
4032 0 : opts.qp = poller->srq;
4033 0 : opts.map = device->map;
4034 0 : opts.qpair = NULL;
4035 0 : opts.shared = true;
4036 0 : opts.max_queue_depth = poller->max_srq_depth;
4037 0 : opts.in_capsule_data_size = rtransport->transport.opts.in_capsule_data_size;
4038 :
4039 0 : poller->resources = nvmf_rdma_resources_create(&opts);
4040 0 : if (!poller->resources) {
4041 0 : SPDK_ERRLOG("Unable to allocate resources for shared receive queue.\n");
4042 0 : return -1;
4043 : }
4044 : }
4045 :
4046 : /*
4047 : * When using an srq, we can limit the completion queue at startup.
4048 : * The following formula represents the calculation:
4049 : * num_cqe = num_recv + num_data_wr + num_send_wr.
4050 : * where num_recv=num_data_wr=and num_send_wr=poller->max_srq_depth
4051 : */
4052 0 : if (poller->srq) {
4053 0 : num_cqe = poller->max_srq_depth * 3;
4054 : } else {
4055 0 : num_cqe = rtransport->rdma_opts.num_cqe;
4056 : }
4057 :
4058 0 : poller->cq = ibv_create_cq(device->context, num_cqe, poller, NULL, 0);
4059 0 : if (!poller->cq) {
4060 0 : SPDK_ERRLOG("Unable to create completion queue\n");
4061 0 : return -1;
4062 : }
4063 0 : poller->num_cqe = num_cqe;
4064 0 : return 0;
4065 : }
4066 :
4067 : static void
4068 0 : _nvmf_rdma_register_poller_in_group(void *c)
4069 : {
4070 0 : struct spdk_nvmf_rdma_poller *poller;
4071 0 : struct poller_manage_ctx *ctx = c;
4072 : struct spdk_nvmf_rdma_device *device;
4073 : int rc;
4074 :
4075 0 : rc = nvmf_rdma_poller_create(ctx->rtransport, ctx->rgroup, ctx->device, &poller);
4076 0 : if (rc < 0 && poller) {
4077 0 : nvmf_rdma_poller_destroy(poller);
4078 : }
4079 :
4080 0 : device = ctx->device;
4081 0 : if (nvmf_rdma_all_pollers_management_done(ctx)) {
4082 0 : device->is_ready = true;
4083 : }
4084 0 : }
4085 :
4086 : static void nvmf_rdma_poll_group_destroy(struct spdk_nvmf_transport_poll_group *group);
4087 :
4088 : static struct spdk_nvmf_transport_poll_group *
4089 5 : nvmf_rdma_poll_group_create(struct spdk_nvmf_transport *transport,
4090 : struct spdk_nvmf_poll_group *group)
4091 : {
4092 : struct spdk_nvmf_rdma_transport *rtransport;
4093 : struct spdk_nvmf_rdma_poll_group *rgroup;
4094 5 : struct spdk_nvmf_rdma_poller *poller;
4095 : struct spdk_nvmf_rdma_device *device;
4096 : int rc;
4097 :
4098 5 : if (spdk_interrupt_mode_is_enabled()) {
4099 0 : SPDK_ERRLOG("RDMA transport does not support interrupt mode\n");
4100 0 : return NULL;
4101 : }
4102 :
4103 5 : rtransport = SPDK_CONTAINEROF(transport, struct spdk_nvmf_rdma_transport, transport);
4104 :
4105 5 : rgroup = calloc(1, sizeof(*rgroup));
4106 5 : if (!rgroup) {
4107 0 : return NULL;
4108 : }
4109 :
4110 5 : TAILQ_INIT(&rgroup->pollers);
4111 :
4112 5 : TAILQ_FOREACH(device, &rtransport->devices, link) {
4113 0 : rc = nvmf_rdma_poller_create(rtransport, rgroup, device, &poller);
4114 0 : if (rc < 0) {
4115 0 : nvmf_rdma_poll_group_destroy(&rgroup->group);
4116 0 : return NULL;
4117 : }
4118 : }
4119 :
4120 5 : TAILQ_INSERT_TAIL(&rtransport->poll_groups, rgroup, link);
4121 5 : if (rtransport->conn_sched.next_admin_pg == NULL) {
4122 1 : rtransport->conn_sched.next_admin_pg = rgroup;
4123 1 : rtransport->conn_sched.next_io_pg = rgroup;
4124 : }
4125 :
4126 5 : return &rgroup->group;
4127 : }
4128 :
4129 : static uint32_t
4130 12 : nvmf_poll_group_get_io_qpair_count(struct spdk_nvmf_poll_group *pg)
4131 : {
4132 : uint32_t count;
4133 :
4134 : /* Just assume that unassociated qpairs will eventually be io
4135 : * qpairs. This is close enough for the use cases for this
4136 : * function.
4137 : */
4138 12 : pthread_mutex_lock(&pg->mutex);
4139 12 : count = pg->stat.current_io_qpairs + pg->current_unassociated_qpairs;
4140 12 : pthread_mutex_unlock(&pg->mutex);
4141 :
4142 12 : return count;
4143 : }
4144 :
4145 : static struct spdk_nvmf_transport_poll_group *
4146 14 : nvmf_rdma_get_optimal_poll_group(struct spdk_nvmf_qpair *qpair)
4147 : {
4148 : struct spdk_nvmf_rdma_transport *rtransport;
4149 : struct spdk_nvmf_rdma_poll_group **pg;
4150 : struct spdk_nvmf_transport_poll_group *result;
4151 : uint32_t count;
4152 :
4153 14 : rtransport = SPDK_CONTAINEROF(qpair->transport, struct spdk_nvmf_rdma_transport, transport);
4154 :
4155 14 : if (TAILQ_EMPTY(&rtransport->poll_groups)) {
4156 2 : return NULL;
4157 : }
4158 :
4159 12 : if (qpair->qid == 0) {
4160 6 : pg = &rtransport->conn_sched.next_admin_pg;
4161 : } else {
4162 : struct spdk_nvmf_rdma_poll_group *pg_min, *pg_start, *pg_current;
4163 : uint32_t min_value;
4164 :
4165 6 : pg = &rtransport->conn_sched.next_io_pg;
4166 6 : pg_min = *pg;
4167 6 : pg_start = *pg;
4168 6 : pg_current = *pg;
4169 6 : min_value = nvmf_poll_group_get_io_qpair_count(pg_current->group.group);
4170 :
4171 : while (1) {
4172 6 : count = nvmf_poll_group_get_io_qpair_count(pg_current->group.group);
4173 :
4174 6 : if (count < min_value) {
4175 0 : min_value = count;
4176 0 : pg_min = pg_current;
4177 : }
4178 :
4179 6 : pg_current = TAILQ_NEXT(pg_current, link);
4180 6 : if (pg_current == NULL) {
4181 2 : pg_current = TAILQ_FIRST(&rtransport->poll_groups);
4182 : }
4183 :
4184 6 : if (pg_current == pg_start || min_value == 0) {
4185 : break;
4186 : }
4187 : }
4188 6 : *pg = pg_min;
4189 : }
4190 :
4191 12 : assert(*pg != NULL);
4192 :
4193 12 : result = &(*pg)->group;
4194 :
4195 12 : *pg = TAILQ_NEXT(*pg, link);
4196 12 : if (*pg == NULL) {
4197 4 : *pg = TAILQ_FIRST(&rtransport->poll_groups);
4198 : }
4199 :
4200 12 : return result;
4201 : }
4202 :
4203 : static void
4204 0 : nvmf_rdma_poller_destroy(struct spdk_nvmf_rdma_poller *poller)
4205 : {
4206 : struct spdk_nvmf_rdma_qpair *qpair, *tmp_qpair;
4207 : int rc;
4208 :
4209 0 : TAILQ_REMOVE(&poller->group->pollers, poller, link);
4210 0 : RB_FOREACH_SAFE(qpair, qpairs_tree, &poller->qpairs, tmp_qpair) {
4211 0 : nvmf_rdma_qpair_destroy(qpair);
4212 : }
4213 :
4214 0 : if (poller->srq) {
4215 0 : if (poller->resources) {
4216 0 : nvmf_rdma_resources_destroy(poller->resources);
4217 : }
4218 0 : spdk_rdma_provider_srq_destroy(poller->srq);
4219 0 : SPDK_DEBUGLOG(rdma, "Destroyed RDMA shared queue %p\n", poller->srq);
4220 : }
4221 :
4222 0 : if (poller->cq) {
4223 0 : rc = ibv_destroy_cq(poller->cq);
4224 0 : if (rc != 0) {
4225 0 : SPDK_ERRLOG("Destroy cq return %d, error: %s\n", rc, strerror(errno));
4226 : }
4227 : }
4228 :
4229 0 : if (poller->destroy_cb) {
4230 0 : poller->destroy_cb(poller->destroy_cb_ctx);
4231 0 : poller->destroy_cb = NULL;
4232 : }
4233 :
4234 0 : free(poller);
4235 0 : }
4236 :
4237 : static void
4238 5 : nvmf_rdma_poll_group_destroy(struct spdk_nvmf_transport_poll_group *group)
4239 : {
4240 : struct spdk_nvmf_rdma_poll_group *rgroup, *next_rgroup;
4241 : struct spdk_nvmf_rdma_poller *poller, *tmp;
4242 : struct spdk_nvmf_rdma_transport *rtransport;
4243 :
4244 5 : rgroup = SPDK_CONTAINEROF(group, struct spdk_nvmf_rdma_poll_group, group);
4245 5 : if (!rgroup) {
4246 0 : return;
4247 : }
4248 :
4249 5 : TAILQ_FOREACH_SAFE(poller, &rgroup->pollers, link, tmp) {
4250 0 : nvmf_rdma_poller_destroy(poller);
4251 : }
4252 :
4253 5 : if (rgroup->group.transport == NULL) {
4254 : /* Transport can be NULL when nvmf_rdma_poll_group_create()
4255 : * calls this function directly in a failure path. */
4256 0 : free(rgroup);
4257 0 : return;
4258 : }
4259 :
4260 5 : rtransport = SPDK_CONTAINEROF(rgroup->group.transport, struct spdk_nvmf_rdma_transport, transport);
4261 :
4262 5 : next_rgroup = TAILQ_NEXT(rgroup, link);
4263 5 : TAILQ_REMOVE(&rtransport->poll_groups, rgroup, link);
4264 5 : if (next_rgroup == NULL) {
4265 1 : next_rgroup = TAILQ_FIRST(&rtransport->poll_groups);
4266 : }
4267 5 : if (rtransport->conn_sched.next_admin_pg == rgroup) {
4268 5 : rtransport->conn_sched.next_admin_pg = next_rgroup;
4269 : }
4270 5 : if (rtransport->conn_sched.next_io_pg == rgroup) {
4271 5 : rtransport->conn_sched.next_io_pg = next_rgroup;
4272 : }
4273 :
4274 5 : free(rgroup);
4275 : }
4276 :
4277 : static void
4278 0 : nvmf_rdma_qpair_reject_connection(struct spdk_nvmf_rdma_qpair *rqpair)
4279 : {
4280 0 : if (rqpair->cm_id != NULL) {
4281 0 : nvmf_rdma_event_reject(rqpair->cm_id, SPDK_NVMF_RDMA_ERROR_NO_RESOURCES);
4282 : }
4283 0 : }
4284 :
4285 : static int
4286 0 : nvmf_rdma_poll_group_add(struct spdk_nvmf_transport_poll_group *group,
4287 : struct spdk_nvmf_qpair *qpair)
4288 : {
4289 : struct spdk_nvmf_rdma_poll_group *rgroup;
4290 : struct spdk_nvmf_rdma_qpair *rqpair;
4291 : struct spdk_nvmf_rdma_device *device;
4292 : struct spdk_nvmf_rdma_poller *poller;
4293 : int rc;
4294 :
4295 0 : rgroup = SPDK_CONTAINEROF(group, struct spdk_nvmf_rdma_poll_group, group);
4296 0 : rqpair = SPDK_CONTAINEROF(qpair, struct spdk_nvmf_rdma_qpair, qpair);
4297 :
4298 0 : device = rqpair->device;
4299 :
4300 0 : TAILQ_FOREACH(poller, &rgroup->pollers, link) {
4301 0 : if (poller->device == device) {
4302 0 : break;
4303 : }
4304 : }
4305 :
4306 0 : if (!poller) {
4307 0 : SPDK_ERRLOG("No poller found for device.\n");
4308 0 : return -1;
4309 : }
4310 :
4311 0 : if (poller->need_destroy) {
4312 0 : SPDK_ERRLOG("Poller is destroying.\n");
4313 0 : return -1;
4314 : }
4315 :
4316 0 : rqpair->poller = poller;
4317 0 : rqpair->srq = rqpair->poller->srq;
4318 :
4319 0 : rc = nvmf_rdma_qpair_initialize(qpair);
4320 0 : if (rc < 0) {
4321 0 : SPDK_ERRLOG("Failed to initialize nvmf_rdma_qpair with qpair=%p\n", qpair);
4322 0 : rqpair->poller = NULL;
4323 0 : rqpair->srq = NULL;
4324 0 : return -1;
4325 : }
4326 :
4327 0 : RB_INSERT(qpairs_tree, &poller->qpairs, rqpair);
4328 :
4329 0 : rc = nvmf_rdma_event_accept(rqpair->cm_id, rqpair);
4330 0 : if (rc) {
4331 : /* Try to reject, but we probably can't */
4332 0 : nvmf_rdma_qpair_reject_connection(rqpair);
4333 0 : return -1;
4334 : }
4335 :
4336 0 : return 0;
4337 : }
4338 :
4339 : static int
4340 0 : nvmf_rdma_poll_group_remove(struct spdk_nvmf_transport_poll_group *group,
4341 : struct spdk_nvmf_qpair *qpair)
4342 : {
4343 : struct spdk_nvmf_rdma_qpair *rqpair;
4344 :
4345 0 : rqpair = SPDK_CONTAINEROF(qpair, struct spdk_nvmf_rdma_qpair, qpair);
4346 0 : assert(group->transport->tgt != NULL);
4347 :
4348 0 : rqpair->destruct_channel = spdk_get_io_channel(group->transport->tgt);
4349 :
4350 0 : if (!rqpair->destruct_channel) {
4351 0 : SPDK_WARNLOG("failed to get io_channel, qpair %p\n", qpair);
4352 0 : return 0;
4353 : }
4354 :
4355 : /* Sanity check that we get io_channel on the correct thread */
4356 0 : if (qpair->group) {
4357 0 : assert(qpair->group->thread == spdk_io_channel_get_thread(rqpair->destruct_channel));
4358 : }
4359 :
4360 0 : return 0;
4361 : }
4362 :
4363 : static int
4364 0 : nvmf_rdma_request_free(struct spdk_nvmf_request *req)
4365 : {
4366 0 : struct spdk_nvmf_rdma_request *rdma_req = SPDK_CONTAINEROF(req, struct spdk_nvmf_rdma_request, req);
4367 0 : struct spdk_nvmf_rdma_transport *rtransport = SPDK_CONTAINEROF(req->qpair->transport,
4368 : struct spdk_nvmf_rdma_transport, transport);
4369 0 : struct spdk_nvmf_rdma_qpair *rqpair = SPDK_CONTAINEROF(rdma_req->req.qpair,
4370 : struct spdk_nvmf_rdma_qpair, qpair);
4371 :
4372 : /*
4373 : * AER requests are freed when a qpair is destroyed. The recv corresponding to that request
4374 : * needs to be returned to the shared receive queue or the poll group will eventually be
4375 : * starved of RECV structures.
4376 : */
4377 0 : if (rqpair->srq && rdma_req->recv) {
4378 : int rc;
4379 0 : struct ibv_recv_wr *bad_recv_wr;
4380 :
4381 0 : spdk_rdma_provider_srq_queue_recv_wrs(rqpair->srq, &rdma_req->recv->wr);
4382 0 : rc = spdk_rdma_provider_srq_flush_recv_wrs(rqpair->srq, &bad_recv_wr);
4383 0 : if (rc) {
4384 0 : SPDK_ERRLOG("Unable to re-post rx descriptor\n");
4385 : }
4386 : }
4387 :
4388 0 : _nvmf_rdma_request_free(rdma_req, rtransport);
4389 0 : return 0;
4390 : }
4391 :
4392 : static int
4393 0 : nvmf_rdma_request_complete(struct spdk_nvmf_request *req)
4394 : {
4395 0 : struct spdk_nvmf_rdma_transport *rtransport = SPDK_CONTAINEROF(req->qpair->transport,
4396 : struct spdk_nvmf_rdma_transport, transport);
4397 0 : struct spdk_nvmf_rdma_request *rdma_req = SPDK_CONTAINEROF(req,
4398 : struct spdk_nvmf_rdma_request, req);
4399 0 : struct spdk_nvmf_rdma_qpair *rqpair = SPDK_CONTAINEROF(rdma_req->req.qpair,
4400 : struct spdk_nvmf_rdma_qpair, qpair);
4401 :
4402 0 : if (spdk_unlikely(rqpair->ibv_in_error_state)) {
4403 : /* The connection is dead. Move the request directly to the completed state. */
4404 0 : rdma_req->state = RDMA_REQUEST_STATE_COMPLETED;
4405 : } else {
4406 : /* The connection is alive, so process the request as normal */
4407 0 : rdma_req->state = RDMA_REQUEST_STATE_EXECUTED;
4408 : }
4409 :
4410 0 : nvmf_rdma_request_process(rtransport, rdma_req);
4411 :
4412 0 : return 0;
4413 : }
4414 :
4415 : static void
4416 0 : nvmf_rdma_close_qpair(struct spdk_nvmf_qpair *qpair,
4417 : spdk_nvmf_transport_qpair_fini_cb cb_fn, void *cb_arg)
4418 : {
4419 0 : struct spdk_nvmf_rdma_qpair *rqpair = SPDK_CONTAINEROF(qpair, struct spdk_nvmf_rdma_qpair, qpair);
4420 :
4421 0 : rqpair->to_close = true;
4422 :
4423 : /* This happens only when the qpair is disconnected before
4424 : * it is added to the poll group. Since there is no poll group,
4425 : * the RDMA qp has not been initialized yet and the RDMA CM
4426 : * event has not yet been acknowledged, so we need to reject it.
4427 : */
4428 0 : if (rqpair->qpair.state == SPDK_NVMF_QPAIR_UNINITIALIZED) {
4429 0 : nvmf_rdma_qpair_reject_connection(rqpair);
4430 0 : nvmf_rdma_qpair_destroy(rqpair);
4431 0 : return;
4432 : }
4433 :
4434 0 : if (rqpair->rdma_qp) {
4435 0 : spdk_rdma_provider_qp_disconnect(rqpair->rdma_qp);
4436 : }
4437 :
4438 0 : nvmf_rdma_destroy_drained_qpair(rqpair);
4439 :
4440 0 : if (cb_fn) {
4441 0 : cb_fn(cb_arg);
4442 : }
4443 : }
4444 :
4445 : static struct spdk_nvmf_rdma_qpair *
4446 0 : get_rdma_qpair_from_wc(struct spdk_nvmf_rdma_poller *rpoller, struct ibv_wc *wc)
4447 : {
4448 0 : struct spdk_nvmf_rdma_qpair find;
4449 :
4450 0 : find.qp_num = wc->qp_num;
4451 :
4452 0 : return RB_FIND(qpairs_tree, &rpoller->qpairs, &find);
4453 : }
4454 :
4455 : #ifdef DEBUG
4456 : static int
4457 0 : nvmf_rdma_req_is_completing(struct spdk_nvmf_rdma_request *rdma_req)
4458 : {
4459 0 : return rdma_req->state == RDMA_REQUEST_STATE_TRANSFERRING_CONTROLLER_TO_HOST ||
4460 0 : rdma_req->state == RDMA_REQUEST_STATE_COMPLETING;
4461 : }
4462 : #endif
4463 :
4464 : static void
4465 0 : _poller_reset_failed_recvs(struct spdk_nvmf_rdma_poller *rpoller, struct ibv_recv_wr *bad_recv_wr,
4466 : int rc)
4467 : {
4468 : struct spdk_nvmf_rdma_recv *rdma_recv;
4469 : struct spdk_nvmf_rdma_wr *bad_rdma_wr;
4470 :
4471 0 : SPDK_ERRLOG("Failed to post a recv for the poller %p with errno %d\n", rpoller, -rc);
4472 0 : while (bad_recv_wr != NULL) {
4473 0 : bad_rdma_wr = (struct spdk_nvmf_rdma_wr *)bad_recv_wr->wr_id;
4474 0 : rdma_recv = SPDK_CONTAINEROF(bad_rdma_wr, struct spdk_nvmf_rdma_recv, rdma_wr);
4475 :
4476 0 : rdma_recv->qpair->current_recv_depth++;
4477 0 : bad_recv_wr = bad_recv_wr->next;
4478 0 : SPDK_ERRLOG("Failed to post a recv for the qpair %p with errno %d\n", rdma_recv->qpair, -rc);
4479 0 : spdk_nvmf_qpair_disconnect(&rdma_recv->qpair->qpair);
4480 : }
4481 0 : }
4482 :
4483 : static void
4484 0 : _qp_reset_failed_recvs(struct spdk_nvmf_rdma_qpair *rqpair, struct ibv_recv_wr *bad_recv_wr, int rc)
4485 : {
4486 0 : SPDK_ERRLOG("Failed to post a recv for the qpair %p with errno %d\n", rqpair, -rc);
4487 0 : while (bad_recv_wr != NULL) {
4488 0 : bad_recv_wr = bad_recv_wr->next;
4489 0 : rqpair->current_recv_depth++;
4490 : }
4491 0 : spdk_nvmf_qpair_disconnect(&rqpair->qpair);
4492 0 : }
4493 :
4494 : static void
4495 0 : _poller_submit_recvs(struct spdk_nvmf_rdma_transport *rtransport,
4496 : struct spdk_nvmf_rdma_poller *rpoller)
4497 : {
4498 : struct spdk_nvmf_rdma_qpair *rqpair;
4499 0 : struct ibv_recv_wr *bad_recv_wr;
4500 : int rc;
4501 :
4502 0 : if (rpoller->srq) {
4503 0 : rc = spdk_rdma_provider_srq_flush_recv_wrs(rpoller->srq, &bad_recv_wr);
4504 0 : if (spdk_unlikely(rc)) {
4505 0 : _poller_reset_failed_recvs(rpoller, bad_recv_wr, rc);
4506 : }
4507 : } else {
4508 0 : while (!STAILQ_EMPTY(&rpoller->qpairs_pending_recv)) {
4509 0 : rqpair = STAILQ_FIRST(&rpoller->qpairs_pending_recv);
4510 0 : rc = spdk_rdma_provider_qp_flush_recv_wrs(rqpair->rdma_qp, &bad_recv_wr);
4511 0 : if (spdk_unlikely(rc)) {
4512 0 : _qp_reset_failed_recvs(rqpair, bad_recv_wr, rc);
4513 : }
4514 0 : STAILQ_REMOVE_HEAD(&rpoller->qpairs_pending_recv, recv_link);
4515 : }
4516 : }
4517 0 : }
4518 :
4519 : static void
4520 0 : _qp_reset_failed_sends(struct spdk_nvmf_rdma_transport *rtransport,
4521 : struct spdk_nvmf_rdma_qpair *rqpair, struct ibv_send_wr *bad_wr, int rc)
4522 : {
4523 : struct spdk_nvmf_rdma_wr *bad_rdma_wr;
4524 0 : struct spdk_nvmf_rdma_request *prev_rdma_req = NULL, *cur_rdma_req = NULL;
4525 :
4526 0 : SPDK_ERRLOG("Failed to post a send for the qpair %p with errno %d\n", rqpair, -rc);
4527 0 : for (; bad_wr != NULL; bad_wr = bad_wr->next) {
4528 0 : bad_rdma_wr = (struct spdk_nvmf_rdma_wr *)bad_wr->wr_id;
4529 0 : assert(rqpair->current_send_depth > 0);
4530 0 : rqpair->current_send_depth--;
4531 0 : switch (bad_rdma_wr->type) {
4532 0 : case RDMA_WR_TYPE_DATA:
4533 0 : cur_rdma_req = SPDK_CONTAINEROF(bad_rdma_wr, struct spdk_nvmf_rdma_request, data_wr);
4534 0 : if (bad_wr->opcode == IBV_WR_RDMA_READ) {
4535 0 : assert(rqpair->current_read_depth > 0);
4536 0 : rqpair->current_read_depth--;
4537 : }
4538 0 : break;
4539 0 : case RDMA_WR_TYPE_SEND:
4540 0 : cur_rdma_req = SPDK_CONTAINEROF(bad_rdma_wr, struct spdk_nvmf_rdma_request, rsp_wr);
4541 0 : break;
4542 0 : default:
4543 0 : SPDK_ERRLOG("Found a RECV in the list of pending SEND requests for qpair %p\n", rqpair);
4544 0 : prev_rdma_req = cur_rdma_req;
4545 0 : continue;
4546 : }
4547 :
4548 0 : if (prev_rdma_req == cur_rdma_req) {
4549 : /* this request was handled by an earlier wr. i.e. we were performing an nvme read. */
4550 : /* We only have to check against prev_wr since each requests wrs are contiguous in this list. */
4551 0 : continue;
4552 : }
4553 :
4554 0 : switch (cur_rdma_req->state) {
4555 0 : case RDMA_REQUEST_STATE_TRANSFERRING_HOST_TO_CONTROLLER:
4556 0 : cur_rdma_req->req.rsp->nvme_cpl.status.sc = SPDK_NVME_SC_INTERNAL_DEVICE_ERROR;
4557 0 : STAILQ_INSERT_TAIL(&rqpair->pending_rdma_send_queue, cur_rdma_req, state_link);
4558 0 : cur_rdma_req->state = RDMA_REQUEST_STATE_READY_TO_COMPLETE_PENDING;
4559 0 : break;
4560 0 : case RDMA_REQUEST_STATE_TRANSFERRING_CONTROLLER_TO_HOST:
4561 : case RDMA_REQUEST_STATE_COMPLETING:
4562 0 : cur_rdma_req->state = RDMA_REQUEST_STATE_COMPLETED;
4563 0 : break;
4564 0 : default:
4565 0 : SPDK_ERRLOG("Found a request in a bad state %d when draining pending SEND requests for qpair %p\n",
4566 : cur_rdma_req->state, rqpair);
4567 0 : continue;
4568 : }
4569 :
4570 0 : nvmf_rdma_request_process(rtransport, cur_rdma_req);
4571 0 : prev_rdma_req = cur_rdma_req;
4572 : }
4573 :
4574 0 : if (spdk_nvmf_qpair_is_active(&rqpair->qpair)) {
4575 : /* Disconnect the connection. */
4576 0 : spdk_nvmf_qpair_disconnect(&rqpair->qpair);
4577 : }
4578 :
4579 0 : }
4580 :
4581 : static void
4582 0 : _poller_submit_sends(struct spdk_nvmf_rdma_transport *rtransport,
4583 : struct spdk_nvmf_rdma_poller *rpoller)
4584 : {
4585 : struct spdk_nvmf_rdma_qpair *rqpair;
4586 0 : struct ibv_send_wr *bad_wr = NULL;
4587 : int rc;
4588 :
4589 0 : while (!STAILQ_EMPTY(&rpoller->qpairs_pending_send)) {
4590 0 : rqpair = STAILQ_FIRST(&rpoller->qpairs_pending_send);
4591 0 : rc = spdk_rdma_provider_qp_flush_send_wrs(rqpair->rdma_qp, &bad_wr);
4592 :
4593 : /* bad wr always points to the first wr that failed. */
4594 0 : if (spdk_unlikely(rc)) {
4595 0 : _qp_reset_failed_sends(rtransport, rqpair, bad_wr, rc);
4596 : }
4597 0 : STAILQ_REMOVE_HEAD(&rpoller->qpairs_pending_send, send_link);
4598 : }
4599 0 : }
4600 :
4601 : static const char *
4602 0 : nvmf_rdma_wr_type_str(enum spdk_nvmf_rdma_wr_type wr_type)
4603 : {
4604 0 : switch (wr_type) {
4605 0 : case RDMA_WR_TYPE_RECV:
4606 0 : return "RECV";
4607 0 : case RDMA_WR_TYPE_SEND:
4608 0 : return "SEND";
4609 0 : case RDMA_WR_TYPE_DATA:
4610 0 : return "DATA";
4611 0 : default:
4612 0 : SPDK_ERRLOG("Unknown WR type %d\n", wr_type);
4613 0 : SPDK_UNREACHABLE();
4614 : }
4615 : }
4616 :
4617 : static inline void
4618 0 : nvmf_rdma_log_wc_status(struct spdk_nvmf_rdma_qpair *rqpair, struct ibv_wc *wc)
4619 : {
4620 0 : enum spdk_nvmf_rdma_wr_type wr_type = ((struct spdk_nvmf_rdma_wr *)wc->wr_id)->type;
4621 :
4622 0 : if (wc->status == IBV_WC_WR_FLUSH_ERR) {
4623 : /* If qpair is in ERR state, we will receive completions for all posted and not completed
4624 : * Work Requests with IBV_WC_WR_FLUSH_ERR status. Don't log an error in that case */
4625 0 : SPDK_DEBUGLOG(rdma,
4626 : "Error on CQ %p, (qp state %d, in_error %d) request 0x%lu, type %s, status: (%d): %s\n",
4627 : rqpair->poller->cq, rqpair->qpair.state, rqpair->ibv_in_error_state, wc->wr_id,
4628 : nvmf_rdma_wr_type_str(wr_type), wc->status, ibv_wc_status_str(wc->status));
4629 : } else {
4630 0 : SPDK_ERRLOG("Error on CQ %p, (qp state %d, in_error %d) request 0x%lu, type %s, status: (%d): %s\n",
4631 : rqpair->poller->cq, rqpair->qpair.state, rqpair->ibv_in_error_state, wc->wr_id,
4632 : nvmf_rdma_wr_type_str(wr_type), wc->status, ibv_wc_status_str(wc->status));
4633 : }
4634 0 : }
4635 :
4636 : static int
4637 0 : nvmf_rdma_poller_poll(struct spdk_nvmf_rdma_transport *rtransport,
4638 : struct spdk_nvmf_rdma_poller *rpoller)
4639 : {
4640 0 : struct ibv_wc wc[32];
4641 : struct spdk_nvmf_rdma_wr *rdma_wr;
4642 : struct spdk_nvmf_rdma_request *rdma_req;
4643 : struct spdk_nvmf_rdma_recv *rdma_recv;
4644 : struct spdk_nvmf_rdma_qpair *rqpair, *tmp_rqpair;
4645 : int reaped, i;
4646 0 : int count = 0;
4647 : int rc;
4648 0 : bool error = false;
4649 0 : uint64_t poll_tsc = spdk_get_ticks();
4650 :
4651 0 : if (spdk_unlikely(rpoller->need_destroy)) {
4652 : /* If qpair is closed before poller destroy, nvmf_rdma_destroy_drained_qpair may not
4653 : * be called because we cannot poll anything from cq. So we call that here to force
4654 : * destroy the qpair after to_close turning true.
4655 : */
4656 0 : RB_FOREACH_SAFE(rqpair, qpairs_tree, &rpoller->qpairs, tmp_rqpair) {
4657 0 : nvmf_rdma_destroy_drained_qpair(rqpair);
4658 : }
4659 0 : return 0;
4660 : }
4661 :
4662 : /* Poll for completing operations. */
4663 0 : reaped = ibv_poll_cq(rpoller->cq, 32, wc);
4664 0 : if (spdk_unlikely(reaped < 0)) {
4665 0 : SPDK_ERRLOG("Error polling CQ! (%d): %s\n",
4666 : errno, spdk_strerror(errno));
4667 0 : return -1;
4668 0 : } else if (reaped == 0) {
4669 0 : rpoller->stat.idle_polls++;
4670 : }
4671 :
4672 0 : rpoller->stat.polls++;
4673 0 : rpoller->stat.completions += reaped;
4674 :
4675 0 : for (i = 0; i < reaped; i++) {
4676 :
4677 0 : rdma_wr = (struct spdk_nvmf_rdma_wr *)wc[i].wr_id;
4678 :
4679 0 : switch (rdma_wr->type) {
4680 0 : case RDMA_WR_TYPE_SEND:
4681 0 : rdma_req = SPDK_CONTAINEROF(rdma_wr, struct spdk_nvmf_rdma_request, rsp_wr);
4682 0 : rqpair = SPDK_CONTAINEROF(rdma_req->req.qpair, struct spdk_nvmf_rdma_qpair, qpair);
4683 :
4684 0 : if (spdk_likely(!wc[i].status)) {
4685 0 : count++;
4686 0 : assert(wc[i].opcode == IBV_WC_SEND);
4687 0 : assert(nvmf_rdma_req_is_completing(rdma_req));
4688 : }
4689 :
4690 0 : rdma_req->state = RDMA_REQUEST_STATE_COMPLETED;
4691 : /* RDMA_WRITE operation completed. +1 since it was chained with rsp WR */
4692 0 : assert(rqpair->current_send_depth >= (uint32_t)rdma_req->num_outstanding_data_wr + 1);
4693 0 : rqpair->current_send_depth -= rdma_req->num_outstanding_data_wr + 1;
4694 0 : rdma_req->num_outstanding_data_wr = 0;
4695 :
4696 0 : nvmf_rdma_request_process(rtransport, rdma_req);
4697 0 : break;
4698 0 : case RDMA_WR_TYPE_RECV:
4699 : /* rdma_recv->qpair will be invalid if using an SRQ. In that case we have to get the qpair from the wc. */
4700 0 : rdma_recv = SPDK_CONTAINEROF(rdma_wr, struct spdk_nvmf_rdma_recv, rdma_wr);
4701 0 : if (rpoller->srq != NULL) {
4702 0 : rdma_recv->qpair = get_rdma_qpair_from_wc(rpoller, &wc[i]);
4703 : /* It is possible that there are still some completions for destroyed QP
4704 : * associated with SRQ. We just ignore these late completions and re-post
4705 : * receive WRs back to SRQ.
4706 : */
4707 0 : if (spdk_unlikely(NULL == rdma_recv->qpair)) {
4708 0 : struct ibv_recv_wr *bad_wr;
4709 :
4710 0 : rdma_recv->wr.next = NULL;
4711 0 : spdk_rdma_provider_srq_queue_recv_wrs(rpoller->srq, &rdma_recv->wr);
4712 0 : rc = spdk_rdma_provider_srq_flush_recv_wrs(rpoller->srq, &bad_wr);
4713 0 : if (rc) {
4714 0 : SPDK_ERRLOG("Failed to re-post recv WR to SRQ, err %d\n", rc);
4715 : }
4716 0 : continue;
4717 : }
4718 : }
4719 0 : rqpair = rdma_recv->qpair;
4720 :
4721 0 : assert(rqpair != NULL);
4722 0 : if (spdk_likely(!wc[i].status)) {
4723 0 : assert(wc[i].opcode == IBV_WC_RECV);
4724 0 : if (rqpair->current_recv_depth >= rqpair->max_queue_depth) {
4725 0 : spdk_nvmf_qpair_disconnect(&rqpair->qpair);
4726 0 : break;
4727 : }
4728 : }
4729 :
4730 0 : rdma_recv->wr.next = NULL;
4731 0 : rqpair->current_recv_depth++;
4732 0 : rdma_recv->receive_tsc = poll_tsc;
4733 0 : rpoller->stat.requests++;
4734 0 : STAILQ_INSERT_HEAD(&rqpair->resources->incoming_queue, rdma_recv, link);
4735 0 : rqpair->qpair.queue_depth++;
4736 0 : break;
4737 0 : case RDMA_WR_TYPE_DATA:
4738 0 : rdma_req = SPDK_CONTAINEROF(rdma_wr, struct spdk_nvmf_rdma_request, data_wr);
4739 0 : rqpair = SPDK_CONTAINEROF(rdma_req->req.qpair, struct spdk_nvmf_rdma_qpair, qpair);
4740 :
4741 0 : assert(rdma_req->num_outstanding_data_wr > 0);
4742 :
4743 0 : rqpair->current_send_depth--;
4744 0 : rdma_req->num_outstanding_data_wr--;
4745 0 : if (spdk_likely(!wc[i].status)) {
4746 0 : assert(wc[i].opcode == IBV_WC_RDMA_READ);
4747 0 : rqpair->current_read_depth--;
4748 : /* wait for all outstanding reads associated with the same rdma_req to complete before proceeding. */
4749 0 : if (rdma_req->num_outstanding_data_wr == 0) {
4750 0 : if (rdma_req->num_remaining_data_wr) {
4751 : /* Only part of RDMA_READ operations was submitted, process the rest */
4752 0 : nvmf_rdma_request_reset_transfer_in(rdma_req, rtransport);
4753 0 : rdma_req->state = RDMA_REQUEST_STATE_DATA_TRANSFER_TO_CONTROLLER_PENDING;
4754 0 : nvmf_rdma_request_process(rtransport, rdma_req);
4755 0 : break;
4756 : }
4757 0 : rdma_req->state = RDMA_REQUEST_STATE_READY_TO_EXECUTE;
4758 0 : nvmf_rdma_request_process(rtransport, rdma_req);
4759 : }
4760 : } else {
4761 : /* If the data transfer fails still force the queue into the error state,
4762 : * if we were performing an RDMA_READ, we need to force the request into a
4763 : * completed state since it wasn't linked to a send. However, in the RDMA_WRITE
4764 : * case, we should wait for the SEND to complete. */
4765 0 : if (rdma_req->data.wr.opcode == IBV_WR_RDMA_READ) {
4766 0 : rqpair->current_read_depth--;
4767 0 : if (rdma_req->num_outstanding_data_wr == 0) {
4768 0 : if (rdma_req->num_remaining_data_wr) {
4769 : /* Partially sent request is still in the pending_rdma_read_queue,
4770 : * remove it now before completing */
4771 0 : rdma_req->num_remaining_data_wr = 0;
4772 0 : STAILQ_REMOVE(&rqpair->pending_rdma_read_queue, rdma_req, spdk_nvmf_rdma_request, state_link);
4773 : }
4774 0 : rdma_req->state = RDMA_REQUEST_STATE_COMPLETED;
4775 0 : nvmf_rdma_request_process(rtransport, rdma_req);
4776 : }
4777 : }
4778 : }
4779 0 : break;
4780 0 : default:
4781 0 : SPDK_ERRLOG("Received an unknown opcode on the CQ: %d\n", wc[i].opcode);
4782 0 : continue;
4783 : }
4784 :
4785 : /* Handle error conditions */
4786 0 : if (spdk_unlikely(wc[i].status)) {
4787 0 : rqpair->ibv_in_error_state = true;
4788 0 : nvmf_rdma_log_wc_status(rqpair, &wc[i]);
4789 :
4790 0 : error = true;
4791 :
4792 0 : if (spdk_nvmf_qpair_is_active(&rqpair->qpair)) {
4793 : /* Disconnect the connection. */
4794 0 : spdk_nvmf_qpair_disconnect(&rqpair->qpair);
4795 : } else {
4796 0 : nvmf_rdma_destroy_drained_qpair(rqpair);
4797 : }
4798 0 : continue;
4799 : }
4800 :
4801 0 : nvmf_rdma_qpair_process_pending(rtransport, rqpair, false);
4802 :
4803 0 : if (spdk_unlikely(!spdk_nvmf_qpair_is_active(&rqpair->qpair))) {
4804 0 : nvmf_rdma_destroy_drained_qpair(rqpair);
4805 : }
4806 : }
4807 :
4808 0 : if (spdk_unlikely(error == true)) {
4809 0 : return -1;
4810 : }
4811 :
4812 0 : if (reaped == 0) {
4813 : /* In some cases we may not receive any CQE but we still may have pending IO requests waiting for
4814 : * a resource (e.g. a WR from the data_wr_pool).
4815 : * We need to start processing of such requests if no CQE reaped */
4816 0 : nvmf_rdma_poller_process_pending_buf_queue(rtransport, rpoller);
4817 : }
4818 :
4819 : /* submit outstanding work requests. */
4820 0 : _poller_submit_recvs(rtransport, rpoller);
4821 0 : _poller_submit_sends(rtransport, rpoller);
4822 :
4823 0 : return count;
4824 : }
4825 :
4826 : static void
4827 0 : _nvmf_rdma_remove_destroyed_device(void *c)
4828 : {
4829 0 : struct spdk_nvmf_rdma_transport *rtransport = c;
4830 : struct spdk_nvmf_rdma_device *device, *device_tmp;
4831 : int rc;
4832 :
4833 0 : TAILQ_FOREACH_SAFE(device, &rtransport->devices, link, device_tmp) {
4834 0 : if (device->ready_to_destroy) {
4835 0 : destroy_ib_device(rtransport, device);
4836 : }
4837 : }
4838 :
4839 0 : free_poll_fds(rtransport);
4840 0 : rc = generate_poll_fds(rtransport);
4841 : /* cannot handle fd allocation error here */
4842 0 : if (rc != 0) {
4843 0 : SPDK_ERRLOG("Failed to generate poll fds after remove ib device.\n");
4844 : }
4845 0 : }
4846 :
4847 : static void
4848 0 : _nvmf_rdma_remove_poller_in_group_cb(void *c)
4849 : {
4850 0 : struct poller_manage_ctx *ctx = c;
4851 0 : struct spdk_nvmf_rdma_transport *rtransport = ctx->rtransport;
4852 0 : struct spdk_nvmf_rdma_device *device = ctx->device;
4853 0 : struct spdk_thread *thread = ctx->thread;
4854 :
4855 0 : if (nvmf_rdma_all_pollers_management_done(c)) {
4856 : /* destroy device when last poller is destroyed */
4857 0 : device->ready_to_destroy = true;
4858 0 : spdk_thread_send_msg(thread, _nvmf_rdma_remove_destroyed_device, rtransport);
4859 : }
4860 0 : }
4861 :
4862 : static void
4863 0 : _nvmf_rdma_remove_poller_in_group(void *c)
4864 : {
4865 0 : struct poller_manage_ctx *ctx = c;
4866 :
4867 0 : ctx->rpoller->need_destroy = true;
4868 0 : ctx->rpoller->destroy_cb_ctx = ctx;
4869 0 : ctx->rpoller->destroy_cb = _nvmf_rdma_remove_poller_in_group_cb;
4870 :
4871 : /* qp will be disconnected after receiving a RDMA_CM_EVENT_DEVICE_REMOVAL event. */
4872 0 : if (RB_EMPTY(&ctx->rpoller->qpairs)) {
4873 0 : nvmf_rdma_poller_destroy(ctx->rpoller);
4874 : }
4875 0 : }
4876 :
4877 : static int
4878 0 : nvmf_rdma_poll_group_poll(struct spdk_nvmf_transport_poll_group *group)
4879 : {
4880 : struct spdk_nvmf_rdma_transport *rtransport;
4881 : struct spdk_nvmf_rdma_poll_group *rgroup;
4882 : struct spdk_nvmf_rdma_poller *rpoller, *tmp;
4883 0 : int count = 0, rc, rc2 = 0;
4884 :
4885 0 : rtransport = SPDK_CONTAINEROF(group->transport, struct spdk_nvmf_rdma_transport, transport);
4886 0 : rgroup = SPDK_CONTAINEROF(group, struct spdk_nvmf_rdma_poll_group, group);
4887 :
4888 0 : TAILQ_FOREACH_SAFE(rpoller, &rgroup->pollers, link, tmp) {
4889 0 : rc = nvmf_rdma_poller_poll(rtransport, rpoller);
4890 0 : if (spdk_unlikely(rc < 0)) {
4891 0 : if (rc2 == 0) {
4892 0 : rc2 = rc;
4893 : }
4894 0 : continue;
4895 : }
4896 0 : count += rc;
4897 : }
4898 :
4899 0 : return rc2 ? rc2 : count;
4900 : }
4901 :
4902 : static int
4903 0 : nvmf_rdma_trid_from_cm_id(struct rdma_cm_id *id,
4904 : struct spdk_nvme_transport_id *trid,
4905 : bool peer)
4906 : {
4907 : struct sockaddr *saddr;
4908 : uint16_t port;
4909 :
4910 0 : spdk_nvme_trid_populate_transport(trid, SPDK_NVME_TRANSPORT_RDMA);
4911 :
4912 0 : if (peer) {
4913 0 : saddr = rdma_get_peer_addr(id);
4914 : } else {
4915 0 : saddr = rdma_get_local_addr(id);
4916 : }
4917 0 : switch (saddr->sa_family) {
4918 0 : case AF_INET: {
4919 0 : struct sockaddr_in *saddr_in = (struct sockaddr_in *)saddr;
4920 :
4921 0 : trid->adrfam = SPDK_NVMF_ADRFAM_IPV4;
4922 0 : inet_ntop(AF_INET, &saddr_in->sin_addr,
4923 0 : trid->traddr, sizeof(trid->traddr));
4924 0 : if (peer) {
4925 0 : port = ntohs(rdma_get_dst_port(id));
4926 : } else {
4927 0 : port = ntohs(rdma_get_src_port(id));
4928 : }
4929 0 : snprintf(trid->trsvcid, sizeof(trid->trsvcid), "%u", port);
4930 0 : break;
4931 : }
4932 0 : case AF_INET6: {
4933 0 : struct sockaddr_in6 *saddr_in = (struct sockaddr_in6 *)saddr;
4934 0 : trid->adrfam = SPDK_NVMF_ADRFAM_IPV6;
4935 0 : inet_ntop(AF_INET6, &saddr_in->sin6_addr,
4936 0 : trid->traddr, sizeof(trid->traddr));
4937 0 : if (peer) {
4938 0 : port = ntohs(rdma_get_dst_port(id));
4939 : } else {
4940 0 : port = ntohs(rdma_get_src_port(id));
4941 : }
4942 0 : snprintf(trid->trsvcid, sizeof(trid->trsvcid), "%u", port);
4943 0 : break;
4944 : }
4945 0 : default:
4946 0 : return -1;
4947 :
4948 : }
4949 :
4950 0 : return 0;
4951 : }
4952 :
4953 : static int
4954 0 : nvmf_rdma_qpair_get_peer_trid(struct spdk_nvmf_qpair *qpair,
4955 : struct spdk_nvme_transport_id *trid)
4956 : {
4957 : struct spdk_nvmf_rdma_qpair *rqpair;
4958 :
4959 0 : rqpair = SPDK_CONTAINEROF(qpair, struct spdk_nvmf_rdma_qpair, qpair);
4960 :
4961 0 : return nvmf_rdma_trid_from_cm_id(rqpair->cm_id, trid, true);
4962 : }
4963 :
4964 : static int
4965 0 : nvmf_rdma_qpair_get_local_trid(struct spdk_nvmf_qpair *qpair,
4966 : struct spdk_nvme_transport_id *trid)
4967 : {
4968 : struct spdk_nvmf_rdma_qpair *rqpair;
4969 :
4970 0 : rqpair = SPDK_CONTAINEROF(qpair, struct spdk_nvmf_rdma_qpair, qpair);
4971 :
4972 0 : return nvmf_rdma_trid_from_cm_id(rqpair->cm_id, trid, false);
4973 : }
4974 :
4975 : static int
4976 0 : nvmf_rdma_qpair_get_listen_trid(struct spdk_nvmf_qpair *qpair,
4977 : struct spdk_nvme_transport_id *trid)
4978 : {
4979 : struct spdk_nvmf_rdma_qpair *rqpair;
4980 :
4981 0 : rqpair = SPDK_CONTAINEROF(qpair, struct spdk_nvmf_rdma_qpair, qpair);
4982 :
4983 0 : return nvmf_rdma_trid_from_cm_id(rqpair->listen_id, trid, false);
4984 : }
4985 :
4986 : void
4987 0 : spdk_nvmf_rdma_init_hooks(struct spdk_nvme_rdma_hooks *hooks)
4988 : {
4989 0 : g_nvmf_hooks = *hooks;
4990 0 : }
4991 :
4992 : static void
4993 0 : nvmf_rdma_request_set_abort_status(struct spdk_nvmf_request *req,
4994 : struct spdk_nvmf_rdma_request *rdma_req_to_abort,
4995 : struct spdk_nvmf_rdma_qpair *rqpair)
4996 : {
4997 0 : rdma_req_to_abort->req.rsp->nvme_cpl.status.sct = SPDK_NVME_SCT_GENERIC;
4998 0 : rdma_req_to_abort->req.rsp->nvme_cpl.status.sc = SPDK_NVME_SC_ABORTED_BY_REQUEST;
4999 :
5000 0 : STAILQ_INSERT_TAIL(&rqpair->pending_rdma_send_queue, rdma_req_to_abort, state_link);
5001 0 : rdma_req_to_abort->state = RDMA_REQUEST_STATE_READY_TO_COMPLETE_PENDING;
5002 :
5003 0 : req->rsp->nvme_cpl.cdw0 &= ~1U; /* Command was successfully aborted. */
5004 0 : }
5005 :
5006 : static int
5007 0 : _nvmf_rdma_qpair_abort_request(void *ctx)
5008 : {
5009 0 : struct spdk_nvmf_request *req = ctx;
5010 0 : struct spdk_nvmf_rdma_request *rdma_req_to_abort = SPDK_CONTAINEROF(
5011 : req->req_to_abort, struct spdk_nvmf_rdma_request, req);
5012 0 : struct spdk_nvmf_rdma_qpair *rqpair = SPDK_CONTAINEROF(req->req_to_abort->qpair,
5013 : struct spdk_nvmf_rdma_qpair, qpair);
5014 : int rc;
5015 :
5016 0 : spdk_poller_unregister(&req->poller);
5017 :
5018 0 : switch (rdma_req_to_abort->state) {
5019 0 : case RDMA_REQUEST_STATE_EXECUTING:
5020 0 : rc = nvmf_ctrlr_abort_request(req);
5021 0 : if (rc == SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS) {
5022 0 : return SPDK_POLLER_BUSY;
5023 : }
5024 0 : break;
5025 :
5026 0 : case RDMA_REQUEST_STATE_NEED_BUFFER:
5027 0 : STAILQ_REMOVE(&rqpair->poller->group->group.pending_buf_queue,
5028 : &rdma_req_to_abort->req, spdk_nvmf_request, buf_link);
5029 :
5030 0 : nvmf_rdma_request_set_abort_status(req, rdma_req_to_abort, rqpair);
5031 0 : break;
5032 :
5033 0 : case RDMA_REQUEST_STATE_DATA_TRANSFER_TO_CONTROLLER_PENDING:
5034 0 : STAILQ_REMOVE(&rqpair->pending_rdma_read_queue, rdma_req_to_abort,
5035 : spdk_nvmf_rdma_request, state_link);
5036 :
5037 0 : nvmf_rdma_request_set_abort_status(req, rdma_req_to_abort, rqpair);
5038 0 : break;
5039 :
5040 0 : case RDMA_REQUEST_STATE_DATA_TRANSFER_TO_HOST_PENDING:
5041 0 : STAILQ_REMOVE(&rqpair->pending_rdma_write_queue, rdma_req_to_abort,
5042 : spdk_nvmf_rdma_request, state_link);
5043 :
5044 0 : nvmf_rdma_request_set_abort_status(req, rdma_req_to_abort, rqpair);
5045 0 : break;
5046 :
5047 0 : case RDMA_REQUEST_STATE_READY_TO_COMPLETE_PENDING:
5048 : /* Remove req from the list here to re-use common function */
5049 0 : STAILQ_REMOVE(&rqpair->pending_rdma_send_queue, rdma_req_to_abort,
5050 : spdk_nvmf_rdma_request, state_link);
5051 :
5052 0 : nvmf_rdma_request_set_abort_status(req, rdma_req_to_abort, rqpair);
5053 0 : break;
5054 :
5055 0 : case RDMA_REQUEST_STATE_TRANSFERRING_HOST_TO_CONTROLLER:
5056 0 : if (spdk_get_ticks() < req->timeout_tsc) {
5057 0 : req->poller = SPDK_POLLER_REGISTER(_nvmf_rdma_qpair_abort_request, req, 0);
5058 0 : return SPDK_POLLER_BUSY;
5059 : }
5060 0 : break;
5061 :
5062 0 : default:
5063 0 : break;
5064 : }
5065 :
5066 0 : spdk_nvmf_request_complete(req);
5067 0 : return SPDK_POLLER_BUSY;
5068 : }
5069 :
5070 : static void
5071 0 : nvmf_rdma_qpair_abort_request(struct spdk_nvmf_qpair *qpair,
5072 : struct spdk_nvmf_request *req)
5073 : {
5074 : struct spdk_nvmf_rdma_qpair *rqpair;
5075 : struct spdk_nvmf_rdma_transport *rtransport;
5076 : struct spdk_nvmf_transport *transport;
5077 : uint16_t cid;
5078 : uint32_t i, max_req_count;
5079 0 : struct spdk_nvmf_rdma_request *rdma_req_to_abort = NULL, *rdma_req;
5080 :
5081 0 : rqpair = SPDK_CONTAINEROF(qpair, struct spdk_nvmf_rdma_qpair, qpair);
5082 0 : rtransport = SPDK_CONTAINEROF(qpair->transport, struct spdk_nvmf_rdma_transport, transport);
5083 0 : transport = &rtransport->transport;
5084 :
5085 0 : cid = req->cmd->nvme_cmd.cdw10_bits.abort.cid;
5086 0 : max_req_count = rqpair->srq == NULL ? rqpair->max_queue_depth : rqpair->poller->max_srq_depth;
5087 :
5088 0 : for (i = 0; i < max_req_count; i++) {
5089 0 : rdma_req = &rqpair->resources->reqs[i];
5090 : /* When SRQ == NULL, rqpair has its own requests and req.qpair pointer always points to the qpair
5091 : * When SRQ != NULL all rqpairs share common requests and qpair pointer is assigned when we start to
5092 : * process a request. So in both cases all requests which are not in FREE state have valid qpair ptr */
5093 0 : if (rdma_req->state != RDMA_REQUEST_STATE_FREE && rdma_req->req.cmd->nvme_cmd.cid == cid &&
5094 0 : rdma_req->req.qpair == qpair) {
5095 0 : rdma_req_to_abort = rdma_req;
5096 0 : break;
5097 : }
5098 : }
5099 :
5100 0 : if (rdma_req_to_abort == NULL) {
5101 0 : spdk_nvmf_request_complete(req);
5102 0 : return;
5103 : }
5104 :
5105 0 : req->req_to_abort = &rdma_req_to_abort->req;
5106 0 : req->timeout_tsc = spdk_get_ticks() +
5107 0 : transport->opts.abort_timeout_sec * spdk_get_ticks_hz();
5108 0 : req->poller = NULL;
5109 :
5110 0 : _nvmf_rdma_qpair_abort_request(req);
5111 : }
5112 :
5113 : static void
5114 0 : nvmf_rdma_poll_group_dump_stat(struct spdk_nvmf_transport_poll_group *group,
5115 : struct spdk_json_write_ctx *w)
5116 : {
5117 : struct spdk_nvmf_rdma_poll_group *rgroup;
5118 : struct spdk_nvmf_rdma_poller *rpoller;
5119 :
5120 0 : assert(w != NULL);
5121 :
5122 0 : rgroup = SPDK_CONTAINEROF(group, struct spdk_nvmf_rdma_poll_group, group);
5123 :
5124 0 : spdk_json_write_named_uint64(w, "pending_data_buffer", rgroup->stat.pending_data_buffer);
5125 :
5126 0 : spdk_json_write_named_array_begin(w, "devices");
5127 :
5128 0 : TAILQ_FOREACH(rpoller, &rgroup->pollers, link) {
5129 0 : spdk_json_write_object_begin(w);
5130 0 : spdk_json_write_named_string(w, "name",
5131 0 : ibv_get_device_name(rpoller->device->context->device));
5132 0 : spdk_json_write_named_uint64(w, "polls",
5133 : rpoller->stat.polls);
5134 0 : spdk_json_write_named_uint64(w, "idle_polls",
5135 : rpoller->stat.idle_polls);
5136 0 : spdk_json_write_named_uint64(w, "completions",
5137 : rpoller->stat.completions);
5138 0 : spdk_json_write_named_uint64(w, "requests",
5139 : rpoller->stat.requests);
5140 0 : spdk_json_write_named_uint64(w, "request_latency",
5141 : rpoller->stat.request_latency);
5142 0 : spdk_json_write_named_uint64(w, "pending_free_request",
5143 : rpoller->stat.pending_free_request);
5144 0 : spdk_json_write_named_uint64(w, "pending_rdma_read",
5145 : rpoller->stat.pending_rdma_read);
5146 0 : spdk_json_write_named_uint64(w, "pending_rdma_write",
5147 : rpoller->stat.pending_rdma_write);
5148 0 : spdk_json_write_named_uint64(w, "pending_rdma_send",
5149 : rpoller->stat.pending_rdma_send);
5150 0 : spdk_json_write_named_uint64(w, "total_send_wrs",
5151 : rpoller->stat.qp_stats.send.num_submitted_wrs);
5152 0 : spdk_json_write_named_uint64(w, "send_doorbell_updates",
5153 : rpoller->stat.qp_stats.send.doorbell_updates);
5154 0 : spdk_json_write_named_uint64(w, "total_recv_wrs",
5155 : rpoller->stat.qp_stats.recv.num_submitted_wrs);
5156 0 : spdk_json_write_named_uint64(w, "recv_doorbell_updates",
5157 : rpoller->stat.qp_stats.recv.doorbell_updates);
5158 0 : spdk_json_write_object_end(w);
5159 : }
5160 :
5161 0 : spdk_json_write_array_end(w);
5162 0 : }
5163 :
5164 : const struct spdk_nvmf_transport_ops spdk_nvmf_transport_rdma = {
5165 : .name = "RDMA",
5166 : .type = SPDK_NVME_TRANSPORT_RDMA,
5167 : .opts_init = nvmf_rdma_opts_init,
5168 : .create = nvmf_rdma_create,
5169 : .dump_opts = nvmf_rdma_dump_opts,
5170 : .destroy = nvmf_rdma_destroy,
5171 :
5172 : .listen = nvmf_rdma_listen,
5173 : .stop_listen = nvmf_rdma_stop_listen,
5174 : .cdata_init = nvmf_rdma_cdata_init,
5175 :
5176 : .listener_discover = nvmf_rdma_discover,
5177 :
5178 : .poll_group_create = nvmf_rdma_poll_group_create,
5179 : .get_optimal_poll_group = nvmf_rdma_get_optimal_poll_group,
5180 : .poll_group_destroy = nvmf_rdma_poll_group_destroy,
5181 : .poll_group_add = nvmf_rdma_poll_group_add,
5182 : .poll_group_remove = nvmf_rdma_poll_group_remove,
5183 : .poll_group_poll = nvmf_rdma_poll_group_poll,
5184 :
5185 : .req_free = nvmf_rdma_request_free,
5186 : .req_complete = nvmf_rdma_request_complete,
5187 :
5188 : .qpair_fini = nvmf_rdma_close_qpair,
5189 : .qpair_get_peer_trid = nvmf_rdma_qpair_get_peer_trid,
5190 : .qpair_get_local_trid = nvmf_rdma_qpair_get_local_trid,
5191 : .qpair_get_listen_trid = nvmf_rdma_qpair_get_listen_trid,
5192 : .qpair_abort_request = nvmf_rdma_qpair_abort_request,
5193 :
5194 : .poll_group_dump_stat = nvmf_rdma_poll_group_dump_stat,
5195 : };
5196 :
5197 2 : SPDK_NVMF_TRANSPORT_REGISTER(rdma, &spdk_nvmf_transport_rdma);
5198 2 : SPDK_LOG_REGISTER_COMPONENT(rdma)
|