1fi_endpoint(3) Libfabric v1.17.0 fi_endpoint(3)
2
6 fi_endpoint - Fabric endpoint operations
7 fi_endpoint / fi_endpoint2 / fi_scalable_ep / fi_passive_ep / fi_close
9 Allocate or close an endpoint.
10 fi_ep_bind
12 Associate an endpoint with hardware resources, such as event
13 queues, completion queues, counters, address vectors, or shared
14 transmit/receive contexts.
15 fi_scalable_ep_bind
17 Associate a scalable endpoint with an address vector
18 fi_pep_bind
20 Associate a passive endpoint with an event queue
21 fi_enable
23 Transitions an active endpoint into an enabled state.
24 fi_cancel
26 Cancel a pending asynchronous data transfer
27 fi_ep_alias
29 Create an alias to the endpoint
30 fi_control
32 Control endpoint operation.
33 fi_getopt / fi_setopt
35 Get or set endpoint options.
36 fi_rx_context / fi_tx_context / fi_srx_context / fi_stx_context
38 Open a transmit or receive context.
39 fi_tc_dscp_set / fi_tc_dscp_get
41 Convert between a DSCP value and a network traffic class
42 fi_rx_size_left / fi_tx_size_left (DEPRECATED)
44 Query the lower bound on how many RX/TX operations may be posted
45 without an operation returning -FI_EAGAIN. This functions have
46 been deprecated and will be removed in a future version of the
47 library.
48
50 #include <rdma/fabric.h>
51 #include <rdma/fi_endpoint.h>
53 int fi_endpoint(struct fid_domain *domain, struct fi_info *info,
55 struct fid_ep **ep, void *context);
56 int fi_endpoint2(struct fid_domain *domain, struct fi_info *info,
58 struct fid_ep **ep, uint64_t flags, void *context);
59 int fi_scalable_ep(struct fid_domain *domain, struct fi_info *info,
61 struct fid_ep **sep, void *context);
62 int fi_passive_ep(struct fi_fabric *fabric, struct fi_info *info,
64 struct fid_pep **pep, void *context);
65 int fi_tx_context(struct fid_ep *sep, int index,
67 struct fi_tx_attr *attr, struct fid_ep **tx_ep,
68 void *context);
69 int fi_rx_context(struct fid_ep *sep, int index,
71 struct fi_rx_attr *attr, struct fid_ep **rx_ep,
72 void *context);
73 int fi_stx_context(struct fid_domain *domain,
75 struct fi_tx_attr *attr, struct fid_stx **stx,
76 void *context);
77 int fi_srx_context(struct fid_domain *domain,
79 struct fi_rx_attr *attr, struct fid_ep **rx_ep,
80 void *context);
81 int fi_close(struct fid *ep);
83 int fi_ep_bind(struct fid_ep *ep, struct fid *fid, uint64_t flags);
85 int fi_scalable_ep_bind(struct fid_ep *sep, struct fid *fid, uint64_t flags);
87 int fi_pep_bind(struct fid_pep *pep, struct fid *fid, uint64_t flags);
89 int fi_enable(struct fid_ep *ep);
91 int fi_cancel(struct fid_ep *ep, void *context);
93 int fi_ep_alias(struct fid_ep *ep, struct fid_ep **alias_ep, uint64_t flags);
95 int fi_control(struct fid *ep, int command, void *arg);
97 int fi_getopt(struct fid *ep, int level, int optname,
99 void *optval, size_t *optlen);
100 int fi_setopt(struct fid *ep, int level, int optname,
102 const void *optval, size_t optlen);
103 uint32_t fi_tc_dscp_set(uint8_t dscp);
105 uint8_t fi_tc_dscp_get(uint32_t tclass);
107 DEPRECATED ssize_t fi_rx_size_left(struct fid_ep *ep);
109 DEPRECATED ssize_t fi_tx_size_left(struct fid_ep *ep);
111
113 fid On creation, specifies a fabric or access domain. On bind,
114 identifies the event queue, completion queue, counter, or ad‐
115 dress vector to bind to the endpoint. In other cases, it’s a
116 fabric identifier of an associated resource.
117 info Details about the fabric interface endpoint to be opened, ob‐
119 tained from fi_getinfo.
120 ep A fabric endpoint.
122 sep A scalable fabric endpoint.
124 pep A passive fabric endpoint.
126 context
128 Context associated with the endpoint or asynchronous operation.
129 index Index to retrieve a specific transmit/receive context.
131 attr Transmit or receive context attributes.
133 flags Additional flags to apply to the operation.
135 command
137 Command of control operation to perform on endpoint.
138 arg Optional control argument.
140 level Protocol level at which the desired option resides.
142 optname
144 The protocol option to read or set.
145 optval The option value that was read or to set.
147 optlen The size of the optval buffer.
149
151 Endpoints are transport level communication portals. There are two
152 types of endpoints: active and passive. Passive endpoints belong to a
153 fabric domain and are most often used to listen for incoming connection
154 requests. However, a passive endpoint may be used to reserve a fabric
155 address that can be granted to an active endpoint. Active endpoints
156 belong to access domains and can perform data transfers.
157 Active endpoints may be connection-oriented or connectionless, and may
159 provide data reliability. The data transfer interfaces – messages
160 (fi_msg), tagged messages (fi_tagged), RMA (fi_rma), and atomics
161 (fi_atomic) – are associated with active endpoints. In basic configu‐
162 rations, an active endpoint has transmit and receive queues. In gener‐
163 al, operations that generate traffic on the fabric are posted to the
164 transmit queue. This includes all RMA and atomic operations, along
165 with sent messages and sent tagged messages. Operations that post buf‐
166 fers for receiving incoming data are submitted to the receive queue.
167 Active endpoints are created in the disabled state. They must transi‐
169 tion into an enabled state before accepting data transfer operations,
170 including posting of receive buffers. The fi_enable call is used to
171 transition an active endpoint into an enabled state. The fi_connect
172 and fi_accept calls will also transition an endpoint into the enabled
173 state, if it is not already active.
174 In order to transition an endpoint into an enabled state, it must be
176 bound to one or more fabric resources. An endpoint that will generate
177 asynchronous completions, either through data transfer operations or
178 communication establishment events, must be bound to the appropriate
179 completion queues or event queues, respectively, before being enabled.
180 Additionally, endpoints that use manual progress must be associated
181 with relevant completion queues or event queues in order to drive
182 progress. For endpoints that are only used as the target of RMA or
183 atomic operations, this means binding the endpoint to a completion
184 queue associated with receive processing. Connectionless endpoints
185 must be bound to an address vector.
186 Once an endpoint has been activated, it may be associated with an ad‐
188 dress vector. Receive buffers may be posted to it and calls may be
189 made to connection establishment routines. Connectionless endpoints
190 may also perform data transfers.
191 The behavior of an endpoint may be adjusted by setting its control data
193 and protocol options. This allows the underlying provider to redirect
194 function calls to implementations optimized to meet the desired appli‐
195 cation behavior.
196 If an endpoint experiences a critical error, it will transition back
198 into a disabled state. Critical errors are reported through the event
199 queue associated with the EP. In certain cases, a disabled endpoint
200 may be re-enabled. The ability to transition back into an enabled
201 state is provider specific and depends on the type of error that the
202 endpoint experienced. When an endpoint is disabled as a result of a
203 critical error, all pending operations are discarded.
204 fi_endpoint / fi_passive_ep / fi_scalable_ep
206 fi_endpoint allocates a new active endpoint. fi_passive_ep allocates a
207 new passive endpoint. fi_scalable_ep allocates a scalable endpoint.
208 The properties and behavior of the endpoint are defined based on the
209 provided struct fi_info. See fi_getinfo for additional details on
210 fi_info. fi_info flags that control the operation of an endpoint are
211 defined below. See section SCALABLE ENDPOINTS.
212 If an active endpoint is allocated in order to accept a connection re‐
214 quest, the fi_info parameter must be the same as the fi_info structure
215 provided with the connection request (FI_CONNREQ) event.
216 An active endpoint may acquire the properties of a passive endpoint by
218 setting the fi_info handle field to the passive endpoint fabric de‐
219 scriptor. This is useful for applications that need to reserve the
220 fabric address of an endpoint prior to knowing if the endpoint will be
221 used on the active or passive side of a connection. For example, this
222 feature is useful for simulating socket semantics. Once an active end‐
223 point acquires the properties of a passive endpoint, the passive end‐
224 point is no longer bound to any fabric resources and must no longer be
225 used. The user is expected to close the passive endpoint after opening
226 the active endpoint in order to free up any lingering resources that
227 had been used.
228 fi_endpoint2
230 Similar to fi_endpoint, buf accepts an extra parameter flags. Mainly
231 used for opening endpoints that use peer transfer feature. See
232 fi_peer(3)
233 fi_close
235 Closes an endpoint and release all resources associated with it.
236 When closing a scalable endpoint, there must be no opened transmit con‐
238 texts, or receive contexts associated with the scalable endpoint. If
239 resources are still associated with the scalable endpoint when attempt‐
240 ing to close, the call will return -FI_EBUSY.
241 Outstanding operations posted to the endpoint when fi_close is called
243 will be discarded. Discarded operations will silently be dropped, with
244 no completions reported. Additionally, a provider may discard previ‐
245 ously completed operations from the associated completion queue(s).
246 The behavior to discard completed operations is provider specific.
247 fi_ep_bind
249 fi_ep_bind is used to associate an endpoint with other allocated re‐
250 sources, such as completion queues, counters, address vectors, event
251 queues, shared contexts, and memory regions. The type of objects that
252 must be bound with an endpoint depend on the endpoint type and its con‐
253 figuration.
254 Passive endpoints must be bound with an EQ that supports connection
256 management events. Connectionless endpoints must be bound to a single
257 address vector. If an endpoint is using a shared transmit and/or re‐
258 ceive context, the shared contexts must be bound to the endpoint. CQs,
259 counters, AV, and shared contexts must be bound to endpoints before
260 they are enabled either explicitly or implicitly.
261 An endpoint must be bound with CQs capable of reporting completions for
263 any asynchronous operation initiated on the endpoint. For example, if
264 the endpoint supports any outbound transfers (sends, RMA, atomics,
265 etc.), then it must be bound to a completion queue that can report
266 transmit completions. This is true even if the endpoint is configured
267 to suppress successful completions, in order that operations that com‐
268 plete in error may be reported to the user.
269 An active endpoint may direct asynchronous completions to different
271 CQs, based on the type of operation. This is specified using
272 fi_ep_bind flags. The following flags may be OR’ed together when bind‐
273 ing an endpoint to a completion domain CQ.
274 FI_RECV
276 Directs the notification of inbound data transfers to the speci‐
277 fied completion queue. This includes received messages. This
278 binding automatically includes FI_REMOTE_WRITE, if applicable to
279 the endpoint.
280 FI_SELECTIVE_COMPLETION
282 By default, data transfer operations write CQ completion entries
283 into the associated completion queue after they have successful‐
284 ly completed. Applications can use this bind flag to selective‐
285 ly enable when completions are generated. If FI_SELECTIVE_COM‐
286 PLETION is specified, data transfer operations will not generate
287 CQ entries for successful completions unless FI_COMPLETION is
288 set as an operational flag for the given operation. Operations
289 that fail asynchronously will still generate completions, even
290 if a completion is not requested. FI_SELECTIVE_COMPLETION must
291 be OR’ed with FI_TRANSMIT and/or FI_RECV flags.
292 When FI_SELECTIVE_COMPLETION is set, the user must determine when a re‐
294 quest that does NOT have FI_COMPLETION set has completed indirectly,
295 usually based on the completion of a subsequent operation or by using
296 completion counters. Use of this flag may improve performance by al‐
297 lowing the provider to avoid writing a CQ completion entry for every
298 operation.
299 See Notes section below for additional information on how this flag in‐
301 teracts with the FI_CONTEXT and FI_CONTEXT2 mode bits.
302 FI_TRANSMIT
304 Directs the completion of outbound data transfer requests to the
305 specified completion queue. This includes send message, RMA,
306 and atomic operations.
307 An endpoint may optionally be bound to a completion counter. Associat‐
309 ing an endpoint with a counter is in addition to binding the EP with a
310 CQ. When binding an endpoint to a counter, the following flags may be
311 specified.
312 FI_READ
314 Increments the specified counter whenever an RMA read, atomic
315 fetch, or atomic compare operation initiated from the endpoint
316 has completed successfully or in error.
317 FI_RECV
319 Increments the specified counter whenever a message is received
320 over the endpoint. Received messages include both tagged and
321 normal message operations.
322 FI_REMOTE_READ
324 Increments the specified counter whenever an RMA read, atomic
325 fetch, or atomic compare operation is initiated from a remote
326 endpoint that targets the given endpoint. Use of this flag re‐
327 quires that the endpoint be created using FI_RMA_EVENT.
328 FI_REMOTE_WRITE
330 Increments the specified counter whenever an RMA write or base
331 atomic operation is initiated from a remote endpoint that tar‐
332 gets the given endpoint. Use of this flag requires that the
333 endpoint be created using FI_RMA_EVENT.
334 FI_SEND
336 Increments the specified counter whenever a message transfer
337 initiated over the endpoint has completed successfully or in er‐
338 ror. Sent messages include both tagged and normal message oper‐
339 ations.
340 FI_WRITE
342 Increments the specified counter whenever an RMA write or base
343 atomic operation initiated from the endpoint has completed suc‐
344 cessfully or in error.
345 An endpoint may only be bound to a single CQ or counter for a given
347 type of operation. For example, a EP may not bind to two counters both
348 using FI_WRITE. Furthermore, providers may limit CQ and counter bind‐
349 ings to endpoints of the same endpoint type (DGRAM, MSG, RDM, etc.).
350 fi_scalable_ep_bind
352 fi_scalable_ep_bind is used to associate a scalable endpoint with an
353 address vector. See section on SCALABLE ENDPOINTS. A scalable end‐
354 point has a single transport level address and can support multiple
355 transmit and receive contexts. The transmit and receive contexts share
356 the transport-level address. Address vectors that are bound to scal‐
357 able endpoints are implicitly bound to any transmit or receive contexts
358 created using the scalable endpoint.
359 fi_enable
361 This call transitions the endpoint into an enabled state. An endpoint
362 must be enabled before it may be used to perform data transfers. En‐
363 abling an endpoint typically results in hardware resources being as‐
364 signed to it. Endpoints making use of completion queues, counters,
365 event queues, and/or address vectors must be bound to them before being
366 enabled.
367 Calling connect or accept on an endpoint will implicitly enable an end‐
369 point if it has not already been enabled.
370 fi_enable may also be used to re-enable an endpoint that has been dis‐
372 abled as a result of experiencing a critical error. Applications
373 should check the return value from fi_enable to see if a disabled end‐
374 point has successfully be re-enabled.
375 fi_cancel
377 fi_cancel attempts to cancel an outstanding asynchronous operation.
378 Canceling an operation causes the fabric provider to search for the op‐
379 eration and, if it is still pending, complete it as having been can‐
380 celed. An error queue entry will be available in the associated error
381 queue with error code FI_ECANCELED. On the other hand, if the opera‐
382 tion completed before the call to fi_cancel, then the completion status
383 of that operation will be available in the associated completion queue.
384 No specific entry related to fi_cancel itself will be posted.
385 Cancel uses the context parameter associated with an operation to iden‐
387 tify the request to cancel. Operations posted without a valid context
388 parameter – either no context parameter is specified or the context
389 value was ignored by the provider – cannot be canceled. If multiple
390 outstanding operations match the context parameter, only one will be
391 canceled. In this case, the operation which is canceled is provider
392 specific. The cancel operation is asynchronous, but will complete
393 within a bounded period of time.
394 fi_ep_alias
396 This call creates an alias to the specified endpoint. Conceptually, an
397 endpoint alias provides an alternate software path from the application
398 to the underlying provider hardware. An alias EP differs from its par‐
399 ent endpoint only by its default data transfer flags. For example, an
400 alias EP may be configured to use a different completion mode. By de‐
401 fault, an alias EP inherits the same data transfer flags as the parent
402 endpoint. An application can use fi_control to modify the alias EP op‐
403 erational flags.
404 When allocating an alias, an application may configure either the
406 transmit or receive operational flags. This avoids needing a separate
407 call to fi_control to set those flags. The flags passed to fi_ep_alias
408 must include FI_TRANSMIT or FI_RECV (not both) with other operational
409 flags OR’ed in. This will override the transmit or receive flags, re‐
410 spectively, for operations posted through the alias endpoint. All al‐
411 located aliases must be closed for the underlying endpoint to be re‐
412 leased.
413 fi_control
415 The control operation is used to adjust the default behavior of an end‐
416 point. It allows the underlying provider to redirect function calls to
417 implementations optimized to meet the desired application behavior. As
418 a result, calls to fi_ep_control must be serialized against all other
419 calls to an endpoint.
420 The base operation of an endpoint is selected during creation using
422 struct fi_info. The following control commands and arguments may be
423 assigned to an endpoint.
424 **FI_BACKLOG - int *value**
426 This option only applies to passive endpoints. It is used to
427 set the connection request backlog for listening endpoints.
428 **FI_GETOPSFLAG – uint64_t *flags**
430 Used to retrieve the current value of flags associated with the
431 data transfer operations initiated on the endpoint. The control
432 argument must include FI_TRANSMIT or FI_RECV (not both) flags to
433 indicate the type of data transfer flags to be returned. See
434 below for a list of control flags.
435 FI_GETWAIT – void **
437 This command allows the user to retrieve the file descriptor as‐
438 sociated with a socket endpoint. The fi_control arg parameter
439 should be an address where a pointer to the returned file de‐
440 scriptor will be written. See fi_eq.3 for addition details us‐
441 ing fi_control with FI_GETWAIT. The file descriptor may be used
442 for notification that the endpoint is ready to send or receive
443 data.
444 **FI_SETOPSFLAG – uint64_t *flags**
446 Used to change the data transfer operation flags associated with
447 an endpoint. The control argument must include FI_TRANSMIT or
448 FI_RECV (not both) to indicate the type of data transfer that
449 the flags should apply to, with other flags OR’ed in. The given
450 flags will override the previous transmit and receive attributes
451 that were set when the endpoint was created. Valid control
452 flags are defined below.
453 fi_getopt / fi_setopt
455 Endpoint protocol operations may be retrieved using fi_getopt or set
456 using fi_setopt. Applications specify the level that a desired option
457 exists, identify the option, and provide input/output buffers to get or
458 set the option. fi_setopt provides an application a way to adjust low-
459 level protocol and implementation specific details of an endpoint.
460 The following option levels and option names and parameters are de‐
462 fined.
463 FI_OPT_ENDPOINT • .RS 2
465 FI_OPT_BUFFERED_LIMIT - size_t
467 Defines the maximum size of a buffered message that will be re‐
468 ported to users as part of a receive completion when the
469 FI_BUFFERED_RECV mode is enabled on an endpoint.
470 fi_getopt() will return the currently configured threshold, or the
472 provider’s default threshold if one has not be set by the application.
473 fi_setopt() allows an application to configure the threshold. If the
474 provider cannot support the requested threshold, it will fail the
475 fi_setopt() call with FI_EMSGSIZE. Calling fi_setopt() with the
476 threshold set to SIZE_MAX will set the threshold to the maximum sup‐
477 ported by the provider. fi_getopt() can then be used to retrieve the
478 set size.
479 In most cases, the sending and receiving endpoints must be configured
481 to use the same threshold value, and the threshold must be set prior to
482 enabling the endpoint.
483 • .RS 2
484 FI_OPT_BUFFERED_MIN - size_t
486 Defines the minimum size of a buffered message that will be re‐
487 ported. Applications would set this to a size that’s big enough
488 to decide whether to discard or claim a buffered receive or when
489 to claim a buffered receive on getting a buffered receive com‐
490 pletion. The value is typically used by a provider when sending
491 a rendezvous protocol request where it would send at least
492 FI_OPT_BUFFERED_MIN bytes of application data along with it. A
493 smaller sized rendezvous protocol message usually results in
494 better latency for the overall transfer of a large message.
495 • .RS 2
496 FI_OPT_CM_DATA_SIZE - size_t
498 Defines the size of available space in CM messages for user-de‐
499 fined data. This value limits the amount of data that applica‐
500 tions can exchange between peer endpoints using the fi_connect,
501 fi_accept, and fi_reject operations. The size returned is de‐
502 pendent upon the properties of the endpoint, except in the case
503 of passive endpoints, in which the size reflects the maximum
504 size of the data that may be present as part of a connection re‐
505 quest event. This option is read only.
506 • .RS 2
507 FI_OPT_MIN_MULTI_RECV - size_t
509 Defines the minimum receive buffer space available when the re‐
510 ceive buffer is released by the provider (see FI_MULTI_RECV).
511 Modifying this value is only guaranteed to set the minimum buf‐
512 fer space needed on receives posted after the value has been
513 changed. It is recommended that applications that want to over‐
514 ride the default MIN_MULTI_RECV value set this option before en‐
515 abling the corresponding endpoint.
516 • .RS 2
517 FI_OPT_FI_HMEM_P2P - int
519 Defines how the provider should handle peer to peer FI_HMEM
520 transfers for this endpoint. By default, the provider will
521 chose whether to use peer to peer support based on the type of
522 transfer (FI_HMEM_P2P_ENABLED). Valid values defined in fi_end‐
523 point.h are:
524 • FI_HMEM_P2P_ENABLED: Peer to peer support may be used by the
526 provider to handle FI_HMEM transfers, and which transfers are
527 initiated using peer to peer is subject to the provider imple‐
528 mentation.
529 • FI_HMEM_P2P_REQUIRED: Peer to peer support must be used for
531 transfers, transfers that cannot be performed using p2p will
532 be reported as failing.
533 • FI_HMEM_P2P_PREFERRED: Peer to peer support should be used by
535 the provider for all transfers if available, but the provider
536 may choose to copy the data to initiate the transfer if peer
537 to peer support is unavailable.
538 • FI_HMEM_P2P_DISABLED: Peer to peer support should not be used.
540 fi_setopt() will return -FI_EOPNOTSUPP if the mode requested cannot be
541 supported by the provider. The FI_HMEM_DISABLE_P2P environment vari‐
542 able discussed in fi_mr(3) takes precedence over this setopt option.
543 • .RS 2
544 FI_OPT_XPU_TRIGGER - struct fi_trigger_xpu *
546 This option only applies to the fi_getopt() call. It is used to
547 query the maximum number of variables required to support XPU
548 triggered operations, along with the size of each variable.
549 The user provides a filled out struct fi_trigger_xpu on input. The
551 iface and device fields should reference an HMEM domain. If the
552 provider does not support XPU triggered operations from the given de‐
553 vice, fi_getopt() will return -FI_EOPNOTSUPP. On input, var should
554 reference an array of struct fi_trigger_var data structures, with count
555 set to the size of the referenced array. If count is 0, the var field
556 will be ignored, and the provider will return the number of fi_trig‐
557 ger_var structures needed. If count is > 0, the provider will set
558 count to the needed value, and for each fi_trigger_var available, set
559 the datatype and count of the variable used for the trigger.
560 fi_tc_dscp_set
562 This call converts a DSCP defined value into a libfabric traffic class
563 value. It should be used when assigning a DSCP value when setting the
564 tclass field in either domain or endpoint attributes
565 fi_tc_dscp_get
567 This call returns the DSCP value associated with the tclass field for
568 the domain or endpoint attributes.
569 fi_rx_size_left (DEPRECATED)
571 This function has been deprecated and will be removed in a future ver‐
572 sion of the library. It may not be supported by all providers.
573 The fi_rx_size_left call returns a lower bound on the number of receive
575 operations that may be posted to the given endpoint without that opera‐
576 tion returning -FI_EAGAIN. Depending on the specific details of the
577 subsequently posted receive operations (e.g., number of iov entries,
578 which receive function is called, etc.), it may be possible to post
579 more receive operations than originally indicated by fi_rx_size_left.
580 fi_tx_size_left (DEPRECATED)
582 This function has been deprecated and will be removed in a future ver‐
583 sion of the library. It may not be supported by all providers.
584 The fi_tx_size_left call returns a lower bound on the number of trans‐
586 mit operations that may be posted to the given endpoint without that
587 operation returning -FI_EAGAIN. Depending on the specific details of
588 the subsequently posted transmit operations (e.g., number of iov en‐
589 tries, which transmit function is called, etc.), it may be possible to
590 post more transmit operations than originally indicated by
591 fi_tx_size_left.
592
594 The fi_ep_attr structure defines the set of attributes associated with
595 an endpoint. Endpoint attributes may be further refined using the
596 transmit and receive context attributes as shown below.
597 struct fi_ep_attr {
599 enum fi_ep_type type;
600 uint32_t protocol;
601 uint32_t protocol_version;
602 size_t max_msg_size;
603 size_t msg_prefix_size;
604 size_t max_order_raw_size;
605 size_t max_order_war_size;
606 size_t max_order_waw_size;
607 uint64_t mem_tag_format;
608 size_t tx_ctx_cnt;
609 size_t rx_ctx_cnt;
610 size_t auth_key_size;
611 uint8_t *auth_key;
612 };
613 type - Endpoint Type
615 If specified, indicates the type of fabric interface communication de‐
616 sired. Supported types are:
617 FI_EP_DGRAM
619 Supports a connectionless, unreliable datagram communication.
620 Message boundaries are maintained, but the maximum message size
621 may be limited to the fabric MTU. Flow control is not guaran‐
622 teed.
623 FI_EP_MSG
625 Provides a reliable, connection-oriented data transfer service
626 with flow control that maintains message boundaries.
627 FI_EP_RDM
629 Reliable datagram message. Provides a reliable, connectionless
630 data transfer service with flow control that maintains message
631 boundaries.
632 FI_EP_SOCK_DGRAM
634 A connectionless, unreliable datagram endpoint with UDP socket-
635 like semantics. FI_EP_SOCK_DGRAM is most useful for applica‐
636 tions designed around using UDP sockets. See the SOCKET END‐
637 POINT section for additional details and restrictions that apply
638 to datagram socket endpoints.
639 FI_EP_SOCK_STREAM
641 Data streaming endpoint with TCP socket-like semantics. Pro‐
642 vides a reliable, connection-oriented data transfer service that
643 does not maintain message boundaries. FI_EP_SOCK_STREAM is most
644 useful for applications designed around using TCP sockets. See
645 the SOCKET ENDPOINT section for additional details and restric‐
646 tions that apply to stream endpoints.
647 FI_EP_UNSPEC
649 The type of endpoint is not specified. This is usually provided
650 as input, with other attributes of the endpoint or the provider
651 selecting the type.
652 Protocol
654 Specifies the low-level end to end protocol employed by the provider.
655 A matching protocol must be used by communicating endpoints to ensure
656 interoperability. The following protocol values are defined. Provider
657 specific protocols are also allowed. Provider specific protocols will
658 be indicated by having the upper bit of the protocol value set to one.
659 FI_PROTO_GNI
661 Protocol runs over Cray GNI low-level interface.
662 FI_PROTO_IB_RDM
664 Reliable-datagram protocol implemented over InfiniBand reliable-
665 connected queue pairs.
666 FI_PROTO_IB_UD
668 The protocol runs over Infiniband unreliable datagram queue
669 pairs.
670 FI_PROTO_IWARP
672 The protocol runs over the Internet wide area RDMA protocol
673 transport.
674 FI_PROTO_IWARP_RDM
676 Reliable-datagram protocol implemented over iWarp reliable-con‐
677 nected queue pairs.
678 FI_PROTO_NETWORKDIRECT
680 Protocol runs over Microsoft NetworkDirect service provider in‐
681 terface. This adds reliable-datagram semantics over the Net‐
682 workDirect connection- oriented endpoint semantics.
683 FI_PROTO_PSMX
685 The protocol is based on an Intel proprietary protocol known as
686 PSM, performance scaled messaging. PSMX is an extended version
687 of the PSM protocol to support the libfabric interfaces.
688 FI_PROTO_PSMX2
690 The protocol is based on an Intel proprietary protocol known as
691 PSM2, performance scaled messaging version 2. PSMX2 is an ex‐
692 tended version of the PSM2 protocol to support the libfabric in‐
693 terfaces.
694 FI_PROTO_PSMX3
696 The protocol is Intel’s protocol known as PSM3, performance
697 scaled messaging version 3. PSMX3 is implemented over RoCEv2
698 and verbs.
699 FI_PROTO_RDMA_CM_IB_RC
701 The protocol runs over Infiniband reliable-connected queue
702 pairs, using the RDMA CM protocol for connection establishment.
703 FI_PROTO_RXD
705 Reliable-datagram protocol implemented over datagram endpoints.
706 RXD is a libfabric utility component that adds RDM endpoint se‐
707 mantics over DGRAM endpoint semantics.
708 FI_PROTO_RXM
710 Reliable-datagram protocol implemented over message endpoints.
711 RXM is a libfabric utility component that adds RDM endpoint se‐
712 mantics over MSG endpoint semantics.
713 FI_PROTO_SOCK_TCP
715 The protocol is layered over TCP packets.
716 FI_PROTO_UDP
718 The protocol sends and receives UDP datagrams. For example, an
719 endpoint using FI_PROTO_UDP will be able to communicate with a
720 remote peer that is using Berkeley SOCK_DGRAM sockets using IP‐
721 PROTO_UDP.
722 FI_PROTO_UNSPEC
724 The protocol is not specified. This is usually provided as in‐
725 put, with other attributes of the socket or the provider select‐
726 ing the actual protocol.
727 protocol_version - Protocol Version
729 Identifies which version of the protocol is employed by the provider.
730 The protocol version allows providers to extend an existing protocol,
731 by adding support for additional features or functionality for example,
732 in a backward compatible manner. Providers that support different ver‐
733 sions of the same protocol should inter-operate, but only when using
734 the capabilities defined for the lesser version.
735 max_msg_size - Max Message Size
737 Defines the maximum size for an application data transfer as a single
738 operation.
739 msg_prefix_size - Message Prefix Size
741 Specifies the size of any required message prefix buffer space. This
742 field will be 0 unless the FI_MSG_PREFIX mode is enabled. If msg_pre‐
743 fix_size is > 0 the specified value will be a multiple of 8-bytes.
744 Max RMA Ordered Size
746 The maximum ordered size specifies the delivery order of transport data
747 into target memory for RMA and atomic operations. Data ordering is
748 separate, but dependent on message ordering (defined below). Data or‐
749 dering is unspecified where message order is not defined.
750 Data ordering refers to the access of the same target memory by subse‐
752 quent operations. When back to back RMA read or write operations ac‐
753 cess the same registered memory location, data ordering indicates
754 whether the second operation reads or writes the target memory after
755 the first operation has completed. For example, will an RMA read that
756 follows an RMA write read back the data that was written? Similarly,
757 will an RMA write that follows an RMA read update the target buffer af‐
758 ter the read has transferred the original data? Data ordering answers
759 these questions, even in the presence of errors, such as the need to
760 resend data because of lost or corrupted network traffic.
761 RMA ordering applies between two operations, and not within a single
763 data transfer. Therefore, ordering is defined per byte-addressable
764 memory location. I.e. ordering specifies whether location X is ac‐
765 cessed by the second operation after the first operation. Nothing is
766 implied about the completion of the first operation before the second
767 operation is initiated. For example, if the first operation updates
768 locations X and Y, but the second operation only accesses location X,
769 there are no guarantees defined relative to location Y and the second
770 operation.
771 In order to support large data transfers being broken into multiple
773 packets and sent using multiple paths through the fabric, data ordering
774 may be limited to transfers of a specific size or less. Providers
775 specify when data ordering is maintained through the following values.
776 Note that even if data ordering is not maintained, message ordering may
777 be.
778 max_order_raw_size
780 Read after write size. If set, an RMA or atomic read operation
781 issued after an RMA or atomic write operation, both of which are
782 smaller than the size, will be ordered. Where the target memory
783 locations overlap, the RMA or atomic read operation will see the
784 results of the previous RMA or atomic write.
785 max_order_war_size
787 Write after read size. If set, an RMA or atomic write operation
788 issued after an RMA or atomic read operation, both of which are
789 smaller than the size, will be ordered. The RMA or atomic read
790 operation will see the initial value of the target memory loca‐
791 tion before a subsequent RMA or atomic write updates the value.
792 max_order_waw_size
794 Write after write size. If set, an RMA or atomic write opera‐
795 tion issued after an RMA or atomic write operation, both of
796 which are smaller than the size, will be ordered. The target
797 memory location will reflect the results of the second RMA or
798 atomic write.
799 An order size value of 0 indicates that ordering is not guaranteed. A
801 value of -1 guarantees ordering for any data size.
802 mem_tag_format - Memory Tag Format
804 The memory tag format is a bit array used to convey the number of
805 tagged bits supported by a provider. Additionally, it may be used to
806 divide the bit array into separate fields. The mem_tag_format option‐
807 ally begins with a series of bits set to 0, to signify bits which are
808 ignored by the provider. Following the initial prefix of ignored bits,
809 the array will consist of alternating groups of bits set to all 1’s or
810 all 0’s. Each group of bits corresponds to a tagged field. The impli‐
811 cation of defining a tagged field is that when a mask is applied to the
812 tagged bit array, all bits belonging to a single field will either be
813 set to 1 or 0, collectively.
814 For example, a mem_tag_format of 0x30FF indicates support for 14 tagged
816 bits, separated into 3 fields. The first field consists of 2-bits, the
817 second field 4-bits, and the final field 8-bits. Valid masks for such
818 a tagged field would be a bitwise OR’ing of zero or more of the follow‐
819 ing values: 0x3000, 0x0F00, and 0x00FF. The provider may not validate
820 the mask provided by the application for performance reasons.
821 By identifying fields within a tag, a provider may be able to optimize
823 their search routines. An application which requests tag fields must
824 provide tag masks that either set all mask bits corresponding to a
825 field to all 0 or all 1. When negotiating tag fields, an application
826 can request a specific number of fields of a given size. A provider
827 must return a tag format that supports the requested number of fields,
828 with each field being at least the size requested, or fail the request.
829 A provider may increase the size of the fields. When reporting comple‐
830 tions (see FI_CQ_FORMAT_TAGGED), it is not guaranteed that the provider
831 would clear out any unsupported tag bits in the tag field of the com‐
832 pletion entry.
833 It is recommended that field sizes be ordered from smallest to largest.
835 A generic, unstructured tag and mask can be achieved by requesting a
836 bit array consisting of alternating 1’s and 0’s.
837 tx_ctx_cnt - Transmit Context Count
839 Number of transmit contexts to associate with the endpoint. If not
840 specified (0), 1 context will be assigned if the endpoint supports out‐
841 bound transfers. Transmit contexts are independent transmit queues
842 that may be separately configured. Each transmit context may be bound
843 to a separate CQ, and no ordering is defined between contexts. Addi‐
844 tionally, no synchronization is needed when accessing contexts in par‐
845 allel.
846 If the count is set to the value FI_SHARED_CONTEXT, the endpoint will
848 be configured to use a shared transmit context, if supported by the
849 provider. Providers that do not support shared transmit contexts will
850 fail the request.
851 See the scalable endpoint and shared contexts sections for additional
853 details.
854 rx_ctx_cnt - Receive Context Count
856 Number of receive contexts to associate with the endpoint. If not
857 specified, 1 context will be assigned if the endpoint supports inbound
858 transfers. Receive contexts are independent processing queues that may
859 be separately configured. Each receive context may be bound to a sepa‐
860 rate CQ, and no ordering is defined between contexts. Additionally, no
861 synchronization is needed when accessing contexts in parallel.
862 If the count is set to the value FI_SHARED_CONTEXT, the endpoint will
864 be configured to use a shared receive context, if supported by the
865 provider. Providers that do not support shared receive contexts will
866 fail the request.
867 See the scalable endpoint and shared contexts sections for additional
869 details.
870 auth_key_size - Authorization Key Length
872 The length of the authorization key in bytes. This field will be 0 if
873 authorization keys are not available or used. This field is ignored
874 unless the fabric is opened with API version 1.5 or greater.
875 auth_key - Authorization Key
877 If supported by the fabric, an authorization key (a.k.a. job key) to
878 associate with the endpoint. An authorization key is used to limit
879 communication between endpoints. Only peer endpoints that are pro‐
880 grammed to use the same authorization key may communicate. Authoriza‐
881 tion keys are often used to implement job keys, to ensure that process‐
882 es running in different jobs do not accidentally cross traffic. The
883 domain authorization key will be used if auth_key_size is set to 0.
884 This field is ignored unless the fabric is opened with API version 1.5
885 or greater.
886
888 Attributes specific to the transmit capabilities of an endpoint are
889 specified using struct fi_tx_attr.
890 struct fi_tx_attr {
892 uint64_t caps;
893 uint64_t mode;
894 uint64_t op_flags;
895 uint64_t msg_order;
896 uint64_t comp_order;
897 size_t inject_size;
898 size_t size;
899 size_t iov_limit;
900 size_t rma_iov_limit;
901 uint32_t tclass;
902 };
903 caps - Capabilities
905 The requested capabilities of the context. The capabilities must be a
906 subset of those requested of the associated endpoint. See the CAPABIL‐
907 ITIES section of fi_getinfo(3) for capability details. If the caps
908 field is 0 on input to fi_getinfo(3), the applicable capability bits
909 from the fi_info structure will be used.
910 The following capabilities apply to the transmit attributes: FI_MSG,
912 FI_RMA, FI_TAGGED, FI_ATOMIC, FI_READ, FI_WRITE, FI_SEND, FI_HMEM,
913 FI_TRIGGER, FI_FENCE, FI_MULTICAST, FI_RMA_PMEM, FI_NAMED_RX_CTX,
914 FI_COLLECTIVE, and FI_XPU.
915 Many applications will be able to ignore this field and rely solely on
917 the fi_info::caps field. Use of this field provides fine grained con‐
918 trol over the transmit capabilities associated with an endpoint. It is
919 useful when handling scalable endpoints, with multiple transmit con‐
920 texts, for example, and allows configuring a specific transmit context
921 with fewer capabilities than that supported by the endpoint or other
922 transmit contexts.
923 mode
925 The operational mode bits of the context. The mode bits will be a sub‐
926 set of those associated with the endpoint. See the MODE section of
927 fi_getinfo(3) for details. A mode value of 0 will be ignored on input
928 to fi_getinfo(3), with the mode value of the fi_info structure used in‐
929 stead. On return from fi_getinfo(3), the mode will be set only to
930 those constraints specific to transmit operations.
931 op_flags - Default transmit operation flags
933 Flags that control the operation of operations submitted against the
934 context. Applicable flags are listed in the Operation Flags section.
935 msg_order - Message Ordering
937 Message ordering refers to the order in which transport layer headers
938 (as viewed by the application) are identified and processed. Relaxed
939 message order enables data transfers to be sent and received out of or‐
940 der, which may improve performance by utilizing multiple paths through
941 the fabric from the initiating endpoint to a target endpoint. Message
942 order applies only between a single source and destination endpoint
943 pair. Ordering between different target endpoints is not defined.
944 Message order is determined using a set of ordering bits. Each set bit
946 indicates that ordering is maintained between data transfers of the
947 specified type. Message order is defined for [read | write | send] op‐
948 erations submitted by an application after [read | write | send] opera‐
949 tions.
950 Message ordering only applies to the end to end transmission of trans‐
952 port headers. Message ordering is necessary, but does not guarantee,
953 the order in which message data is sent or received by the transport
954 layer. Message ordering requires matching ordering semantics on the
955 receiving side of a data transfer operation in order to guarantee that
956 ordering is met.
957 FI_ORDER_ATOMIC_RAR
959 Atomic read after read. If set, atomic fetch operations are
960 transmitted in the order submitted relative to other atomic
961 fetch operations. If not set, atomic fetches may be transmitted
962 out of order from their submission.
963 FI_ORDER_ATOMIC_RAW
965 Atomic read after write. If set, atomic fetch operations are
966 transmitted in the order submitted relative to atomic update op‐
967 erations. If not set, atomic fetches may be transmitted ahead
968 of atomic updates.
969 FI_ORDER_ATOMIC_WAR
971 RMA write after read. If set, atomic update operations are
972 transmitted in the order submitted relative to atomic fetch op‐
973 erations. If not set, atomic updates may be transmitted ahead
974 of atomic fetches.
975 FI_ORDER_ATOMIC_WAW
977 RMA write after write. If set, atomic update operations are
978 transmitted in the order submitted relative to other atomic up‐
979 date operations. If not atomic updates may be transmitted out
980 of order from their submission.
981 FI_ORDER_NONE
983 No ordering is specified. This value may be used as input in
984 order to obtain the default message order supported by the
985 provider. FI_ORDER_NONE is an alias for the value 0.
986 FI_ORDER_RAR
988 Read after read. If set, RMA and atomic read operations are
989 transmitted in the order submitted relative to other RMA and
990 atomic read operations. If not set, RMA and atomic reads may be
991 transmitted out of order from their submission.
992 FI_ORDER_RAS
994 Read after send. If set, RMA and atomic read operations are
995 transmitted in the order submitted relative to message send op‐
996 erations, including tagged sends. If not set, RMA and atomic
997 reads may be transmitted ahead of sends.
998 FI_ORDER_RAW
1000 Read after write. If set, RMA and atomic read operations are
1001 transmitted in the order submitted relative to RMA and atomic
1002 write operations. If not set, RMA and atomic reads may be
1003 transmitted ahead of RMA and atomic writes.
1004 FI_ORDER_RMA_RAR
1006 RMA read after read. If set, RMA read operations are transmit‐
1007 ted in the order submitted relative to other RMA read opera‐
1008 tions. If not set, RMA reads may be transmitted out of order
1009 from their submission.
1010 FI_ORDER_RMA_RAW
1012 RMA read after write. If set, RMA read operations are transmit‐
1013 ted in the order submitted relative to RMA write operations. If
1014 not set, RMA reads may be transmitted ahead of RMA writes.
1015 FI_ORDER_RMA_WAR
1017 RMA write after read. If set, RMA write operations are trans‐
1018 mitted in the order submitted relative to RMA read operations.
1019 If not set, RMA writes may be transmitted ahead of RMA reads.
1020 FI_ORDER_RMA_WAW
1022 RMA write after write. If set, RMA write operations are trans‐
1023 mitted in the order submitted relative to other RMA write opera‐
1024 tions. If not set, RMA writes may be transmitted out of order
1025 from their submission.
1026 FI_ORDER_SAR
1028 Send after read. If set, message send operations, including
1029 tagged sends, are transmitted in order submitted relative to RMA
1030 and atomic read operations. If not set, message sends may be
1031 transmitted ahead of RMA and atomic reads.
1032 FI_ORDER_SAS
1034 Send after send. If set, message send operations, including
1035 tagged sends, are transmitted in the order submitted relative to
1036 other message send. If not set, message sends may be transmit‐
1037 ted out of order from their submission.
1038 FI_ORDER_SAW
1040 Send after write. If set, message send operations, including
1041 tagged sends, are transmitted in order submitted relative to RMA
1042 and atomic write operations. If not set, message sends may be
1043 transmitted ahead of RMA and atomic writes.
1044 FI_ORDER_WAR
1046 Write after read. If set, RMA and atomic write operations are
1047 transmitted in the order submitted relative to RMA and atomic
1048 read operations. If not set, RMA and atomic writes may be
1049 transmitted ahead of RMA and atomic reads.
1050 FI_ORDER_WAS
1052 Write after send. If set, RMA and atomic write operations are
1053 transmitted in the order submitted relative to message send op‐
1054 erations, including tagged sends. If not set, RMA and atomic
1055 writes may be transmitted ahead of sends.
1056 FI_ORDER_WAW
1058 Write after write. If set, RMA and atomic write operations are
1059 transmitted in the order submitted relative to other RMA and
1060 atomic write operations. If not set, RMA and atomic writes may
1061 be transmitted out of order from their submission.
1062 comp_order - Completion Ordering
1064 Completion ordering refers to the order in which completed requests are
1065 written into the completion queue. Completion ordering is similar to
1066 message order. Relaxed completion order may enable faster reporting of
1067 completed transfers, allow acknowledgments to be sent over different
1068 fabric paths, and support more sophisticated retry mechanisms. This
1069 can result in lower-latency completions, particularly when using con‐
1070 nectionless endpoints. Strict completion ordering may require that
1071 providers queue completed operations or limit available optimizations.
1072 For transmit requests, completion ordering depends on the endpoint com‐
1074 munication type. For unreliable communication, completion ordering ap‐
1075 plies to all data transfer requests submitted to an endpoint. For re‐
1076 liable communication, completion ordering only applies to requests that
1077 target a single destination endpoint. Completion ordering of requests
1078 that target different endpoints over a reliable transport is not de‐
1079 fined.
1080 Applications should specify the completion ordering that they support
1082 or require. Providers should return the completion order that they ac‐
1083 tually provide, with the constraint that the returned ordering is
1084 stricter than that specified by the application. Supported completion
1085 order values are:
1086 FI_ORDER_NONE
1088 No ordering is defined for completed operations. Requests sub‐
1089 mitted to the transmit context may complete in any order.
1090 FI_ORDER_STRICT
1092 Requests complete in the order in which they are submitted to
1093 the transmit context.
1094 inject_size
1096 The requested inject operation size (see the FI_INJECT flag) that the
1097 context will support. This is the maximum size data transfer that can
1098 be associated with an inject operation (such as fi_inject) or may be
1099 used with the FI_INJECT data transfer flag.
1100 size
1102 The size of the transmit context. The mapping of the size value to re‐
1103 sources is provider specific, but it is directly related to the number
1104 of command entries allocated for the endpoint. A smaller size value
1105 consumes fewer hardware and software resources, while a larger size al‐
1106 lows queuing more transmit requests.
1107 While the size attribute guides the size of underlying endpoint trans‐
1109 mit queue, there is not necessarily a one-to-one mapping between a
1110 transmit operation and a queue entry. A single transmit operation may
1111 consume multiple queue entries; for example, one per scatter-gather en‐
1112 try. Additionally, the size field is intended to guide the allocation
1113 of the endpoint’s transmit context. Specifically, for connectionless
1114 endpoints, there may be lower-level queues use to track communication
1115 on a per peer basis. The sizes of any lower-level queues may only be
1116 significantly smaller than the endpoint’s transmit size, in order to
1117 reduce resource utilization.
1118 iov_limit
1120 This is the maximum number of IO vectors (scatter-gather elements) that
1121 a single posted operation may reference.
1122 rma_iov_limit
1124 This is the maximum number of RMA IO vectors (scatter-gather elements)
1125 that an RMA or atomic operation may reference. The rma_iov_limit cor‐
1126 responds to the rma_iov_count values in RMA and atomic operations. See
1127 struct fi_msg_rma and struct fi_msg_atomic in fi_rma.3 and fi_atomic.3,
1128 for additional details. This limit applies to both the number of RMA
1129 IO vectors that may be specified when initiating an operation from the
1130 local endpoint, as well as the maximum number of IO vectors that may be
1131 carried in a single request from a remote endpoint.
1132 Traffic Class (tclass)
1134 Traffic classes can be a differentiated services code point (DSCP) val‐
1135 ue, one of the following defined labels, or a provider-specific defini‐
1136 tion. If tclass is unset or set to FI_TC_UNSPEC, the endpoint will use
1137 the default traffic class associated with the domain.
1138 FI_TC_BEST_EFFORT
1140 This is the default in the absence of any other local or fabric
1141 configuration. This class carries the traffic for a number of
1142 applications executing concurrently over the same network infra‐
1143 structure. Even though it is shared, network capacity and re‐
1144 source allocation are distributed fairly across the applica‐
1145 tions.
1146 FI_TC_BULK_DATA
1148 This class is intended for large data transfers associated with
1149 I/O and is present to separate sustained I/O transfers from oth‐
1150 er application inter-process communications.
1151 FI_TC_DEDICATED_ACCESS
1153 This class operates at the highest priority, except the manage‐
1154 ment class. It carries a high bandwidth allocation, minimum la‐
1155 tency targets, and the highest scheduling and arbitration prior‐
1156 ity.
1157 FI_TC_LOW_LATENCY
1159 This class supports low latency, low jitter data patterns typi‐
1160 cally caused by transactional data exchanges, barrier synchro‐
1161 nizations, and collective operations that are typical of HPC ap‐
1162 plications. This class often requires maximum tolerable laten‐
1163 cies that data transfers must achieve for correct or performance
1164 operations. Fulfillment of such requests in this class will
1165 typically require accompanying bandwidth and message size limi‐
1166 tations so as not to consume excessive bandwidth at high priori‐
1167 ty.
1168 FI_TC_NETWORK_CTRL
1170 This class is intended for traffic directly related to fabric
1171 (network) management, which is critical to the correct operation
1172 of the network. Its use is typically restricted to privileged
1173 network management applications.
1174 FI_TC_SCAVENGER
1176 This class is used for data that is desired but does not have
1177 strict delivery requirements, such as in-band network or appli‐
1178 cation level monitoring data. Use of this class indicates that
1179 the traffic is considered lower priority and should not inter‐
1180 fere with higher priority workflows.
1181 fi_tc_dscp_set / fi_tc_dscp_get
1183 DSCP values are supported via the DSCP get and set functions.
1184 The definitions for DSCP values are outside the scope of libfab‐
1185 ric. See the fi_tc_dscp_set and fi_tc_dscp_get function defini‐
1186 tions for details on their use.
1187
1189 Attributes specific to the receive capabilities of an endpoint are
1190 specified using struct fi_rx_attr.
1191 struct fi_rx_attr {
1193 uint64_t caps;
1194 uint64_t mode;
1195 uint64_t op_flags;
1196 uint64_t msg_order;
1197 uint64_t comp_order;
1198 size_t total_buffered_recv;
1199 size_t size;
1200 size_t iov_limit;
1201 };
1202 caps - Capabilities
1204 The requested capabilities of the context. The capabilities must be a
1205 subset of those requested of the associated endpoint. See the CAPABIL‐
1206 ITIES section if fi_getinfo(3) for capability details. If the caps
1207 field is 0 on input to fi_getinfo(3), the applicable capability bits
1208 from the fi_info structure will be used.
1209 The following capabilities apply to the receive attributes: FI_MSG,
1211 FI_RMA, FI_TAGGED, FI_ATOMIC, FI_REMOTE_READ, FI_REMOTE_WRITE, FI_RECV,
1212 FI_HMEM, FI_TRIGGER, FI_RMA_PMEM, FI_DIRECTED_RECV, FI_VARIABLE_MSG,
1213 FI_MULTI_RECV, FI_SOURCE, FI_RMA_EVENT, FI_SOURCE_ERR, FI_COLLECTIVE,
1214 and FI_XPU.
1215 Many applications will be able to ignore this field and rely solely on
1217 the fi_info::caps field. Use of this field provides fine grained con‐
1218 trol over the receive capabilities associated with an endpoint. It is
1219 useful when handling scalable endpoints, with multiple receive con‐
1220 texts, for example, and allows configuring a specific receive context
1221 with fewer capabilities than that supported by the endpoint or other
1222 receive contexts.
1223 mode
1225 The operational mode bits of the context. The mode bits will be a sub‐
1226 set of those associated with the endpoint. See the MODE section of
1227 fi_getinfo(3) for details. A mode value of 0 will be ignored on input
1228 to fi_getinfo(3), with the mode value of the fi_info structure used in‐
1229 stead. On return from fi_getinfo(3), the mode will be set only to
1230 those constraints specific to receive operations.
1231 op_flags - Default receive operation flags
1233 Flags that control the operation of operations submitted against the
1234 context. Applicable flags are listed in the Operation Flags section.
1235 msg_order - Message Ordering
1237 For a description of message ordering, see the msg_order field in the
1238 Transmit Context Attribute section. Receive context message ordering
1239 defines the order in which received transport message headers are pro‐
1240 cessed when received by an endpoint. When ordering is set, it indi‐
1241 cates that message headers will be processed in order, based on how the
1242 transmit side has identified the messages. Typically, this means that
1243 messages will be handled in order based on a message level sequence
1244 number.
1245 The following ordering flags, as defined for transmit ordering, also
1247 apply to the processing of received operations: FI_ORDER_NONE, FI_OR‐
1248 DER_RAR, FI_ORDER_RAW, FI_ORDER_RAS, FI_ORDER_WAR, FI_ORDER_WAW, FI_OR‐
1249 DER_WAS, FI_ORDER_SAR, FI_ORDER_SAW, FI_ORDER_SAS, FI_ORDER_RMA_RAR,
1250 FI_ORDER_RMA_RAW, FI_ORDER_RMA_WAR, FI_ORDER_RMA_WAW, FI_ORDER_ATOM‐
1251 IC_RAR, FI_ORDER_ATOMIC_RAW, FI_ORDER_ATOMIC_WAR, and FI_ORDER_ATOM‐
1252 IC_WAW.
1253 comp_order - Completion Ordering
1255 For a description of completion ordering, see the comp_order field in
1256 the Transmit Context Attribute section.
1257 FI_ORDER_DATA
1259 When set, this bit indicates that received data is written into
1260 memory in order. Data ordering applies to memory accessed as
1261 part of a single operation and between operations if message or‐
1262 dering is guaranteed.
1263 FI_ORDER_NONE
1265 No ordering is defined for completed operations. Receive opera‐
1266 tions may complete in any order, regardless of their submission
1267 order.
1268 FI_ORDER_STRICT
1270 Receive operations complete in the order in which they are pro‐
1271 cessed by the receive context, based on the receive side msg_or‐
1272 der attribute.
1273 total_buffered_recv
1275 This field is supported for backwards compatibility purposes. It is a
1276 hint to the provider of the total available space that may be needed to
1277 buffer messages that are received for which there is no matching re‐
1278 ceive operation. The provider may adjust or ignore this value. The
1279 allocation of internal network buffering among received message is
1280 provider specific. For instance, a provider may limit the size of mes‐
1281 sages which can be buffered or the amount of buffering allocated to a
1282 single message.
1283 If receive side buffering is disabled (total_buffered_recv = 0) and a
1285 message is received by an endpoint, then the behavior is dependent on
1286 whether resource management has been enabled (FI_RM_ENABLED has be set
1287 or not). See the Resource Management section of fi_domain.3 for fur‐
1288 ther clarification. It is recommended that applications enable re‐
1289 source management if they anticipate receiving unexpected messages,
1290 rather than modifying this value.
1291 size
1293 The size of the receive context. The mapping of the size value to re‐
1294 sources is provider specific, but it is directly related to the number
1295 of command entries allocated for the endpoint. A smaller size value
1296 consumes fewer hardware and software resources, while a larger size al‐
1297 lows queuing more transmit requests.
1298 While the size attribute guides the size of underlying endpoint receive
1300 queue, there is not necessarily a one-to-one mapping between a receive
1301 operation and a queue entry. A single receive operation may consume
1302 multiple queue entries; for example, one per scatter-gather entry. Ad‐
1303 ditionally, the size field is intended to guide the allocation of the
1304 endpoint’s receive context. Specifically, for connectionless end‐
1305 points, there may be lower-level queues use to track communication on a
1306 per peer basis. The sizes of any lower-level queues may only be sig‐
1307 nificantly smaller than the endpoint’s receive size, in order to reduce
1308 resource utilization.
1309 iov_limit
1311 This is the maximum number of IO vectors (scatter-gather elements) that
1312 a single posted operating may reference.
1313
1315 A scalable endpoint is a communication portal that supports multiple
1316 transmit and receive contexts. Scalable endpoints are loosely modeled
1317 after the networking concept of transmit/receive side scaling, also
1318 known as multi-queue. Support for scalable endpoints is domain specif‐
1319 ic. Scalable endpoints may improve the performance of multi-threaded
1320 and parallel applications, by allowing threads to access independent
1321 transmit and receive queues. A scalable endpoint has a single trans‐
1322 port level address, which can reduce the memory requirements needed to
1323 store remote addressing data, versus using standard endpoints. Scal‐
1324 able endpoints cannot be used directly for communication operations,
1325 and require the application to explicitly create transmit and receive
1326 contexts as described below.
1327 fi_tx_context
1329 Transmit contexts are independent transmit queues. Ordering and syn‐
1330 chronization between contexts are not defined. Conceptually a transmit
1331 context behaves similar to a send-only endpoint. A transmit context
1332 may be configured with fewer capabilities than the base endpoint and
1333 with different attributes (such as ordering requirements and inject
1334 size) than other contexts associated with the same scalable endpoint.
1335 Each transmit context has its own completion queue. The number of
1336 transmit contexts associated with an endpoint is specified during end‐
1337 point creation.
1338 The fi_tx_context call is used to retrieve a specific context, identi‐
1340 fied by an index (see above for details on transmit context at‐
1341 tributes). Providers may dynamically allocate contexts when fi_tx_con‐
1342 text is called, or may statically create all contexts when fi_endpoint
1343 is invoked. By default, a transmit context inherits the properties of
1344 its associated endpoint. However, applications may request context
1345 specific attributes through the attr parameter. Support for per trans‐
1346 mit context attributes is provider specific and not guaranteed.
1347 Providers will return the actual attributes assigned to the context
1348 through the attr parameter, if provided.
1349 fi_rx_context
1351 Receive contexts are independent receive queues for receiving incoming
1352 data. Ordering and synchronization between contexts are not guaran‐
1353 teed. Conceptually a receive context behaves similar to a receive-only
1354 endpoint. A receive context may be configured with fewer capabilities
1355 than the base endpoint and with different attributes (such as ordering
1356 requirements and inject size) than other contexts associated with the
1357 same scalable endpoint. Each receive context has its own completion
1358 queue. The number of receive contexts associated with an endpoint is
1359 specified during endpoint creation.
1360 Receive contexts are often associated with steering flows, that specify
1362 which incoming packets targeting a scalable endpoint to process. How‐
1363 ever, receive contexts may be targeted directly by the initiator, if
1364 supported by the underlying protocol. Such contexts are referred to as
1365 `named'. Support for named contexts must be indicated by setting the
1366 caps FI_NAMED_RX_CTX capability when the corresponding endpoint is cre‐
1367 ated. Support for named receive contexts is coordinated with address
1368 vectors. See fi_av(3) and fi_rx_addr(3).
1369 The fi_rx_context call is used to retrieve a specific context, identi‐
1371 fied by an index (see above for details on receive context attributes).
1372 Providers may dynamically allocate contexts when fi_rx_context is
1373 called, or may statically create all contexts when fi_endpoint is in‐
1374 voked. By default, a receive context inherits the properties of its
1375 associated endpoint. However, applications may request context specif‐
1376 ic attributes through the attr parameter. Support for per receive con‐
1377 text attributes is provider specific and not guaranteed. Providers
1378 will return the actual attributes assigned to the context through the
1379 attr parameter, if provided.
1380
1382 Shared contexts are transmit and receive contexts explicitly shared
1383 among one or more endpoints. A shareable context allows an application
1384 to use a single dedicated provider resource among multiple transport
1385 addressable endpoints. This can greatly reduce the resources needed to
1386 manage communication over multiple endpoints by multiplexing transmit
1387 and/or receive processing, with the potential cost of serializing ac‐
1388 cess across multiple endpoints. Support for shareable contexts is do‐
1389 main specific.
1390 Conceptually, shareable transmit contexts are transmit queues that may
1392 be accessed by many endpoints. The use of a shared transmit context is
1393 mostly opaque to an application. Applications must allocate and bind
1394 shared transmit contexts to endpoints, but operations are posted di‐
1395 rectly to the endpoint. Shared transmit contexts are not associated
1396 with completion queues or counters. Completed operations are posted to
1397 the CQs bound to the endpoint. An endpoint may only be associated with
1398 a single shared transmit context.
1399 Unlike shared transmit contexts, applications interact directly with
1401 shared receive contexts. Users post receive buffers directly to a
1402 shared receive context, with the buffers usable by any endpoint bound
1403 to the shared receive context. Shared receive contexts are not associ‐
1404 ated with completion queues or counters. Completed receive operations
1405 are posted to the CQs bound to the endpoint. An endpoint may only be
1406 associated with a single receive context, and all connectionless end‐
1407 points associated with a shared receive context must also share the
1408 same address vector.
1409 Endpoints associated with a shared transmit context may use dedicated
1411 receive contexts, and vice-versa. Or an endpoint may use shared trans‐
1412 mit and receive contexts. And there is no requirement that the same
1413 group of endpoints sharing a context of one type also share the context
1414 of an alternate type. Furthermore, an endpoint may use a shared con‐
1415 text of one type, but a scalable set of contexts of the alternate type.
1416 fi_stx_context
1418 This call is used to open a shareable transmit context (see above for
1419 details on the transmit context attributes). Endpoints associated with
1420 a shared transmit context must use a subset of the transmit context’s
1421 attributes. Note that this is the reverse of the requirement for
1422 transmit contexts for scalable endpoints.
1423 fi_srx_context
1425 This allocates a shareable receive context (see above for details on
1426 the receive context attributes). Endpoints associated with a shared
1427 receive context must use a subset of the receive context’s attributes.
1428 Note that this is the reverse of the requirement for receive contexts
1429 for scalable endpoints.
1430
1432 The following feature and description should be considered experimen‐
1433 tal. Until the experimental tag is removed, the interfaces, semantics,
1434 and data structures associated with socket endpoints may change between
1435 library versions.
1436 This section applies to endpoints of type FI_EP_SOCK_STREAM and
1438 FI_EP_SOCK_DGRAM, commonly referred to as socket endpoints.
1439 Socket endpoints are defined with semantics that allow them to more
1441 easily be adopted by developers familiar with the UNIX socket API, or
1442 by middleware that exposes the socket API, while still taking advantage
1443 of high-performance hardware features.
1444 The key difference between socket endpoints and other active endpoints
1446 are socket endpoints use synchronous data transfers. Buffers passed
1447 into send and receive operations revert to the control of the applica‐
1448 tion upon returning from the function call. As a result, no data
1449 transfer completions are reported to the application, and socket end‐
1450 points are not associated with completion queues or counters.
1451 Socket endpoints support a subset of message operations: fi_send,
1453 fi_sendv, fi_sendmsg, fi_recv, fi_recvv, fi_recvmsg, and fi_inject.
1454 Because data transfers are synchronous, the return value from send and
1455 receive operations indicate the number of bytes transferred on success,
1456 or a negative value on error, including -FI_EAGAIN if the endpoint can‐
1457 not send or receive any data because of full or empty queues, respec‐
1458 tively.
1459 Socket endpoints are associated with event queues and address vectors,
1461 and process connection management events asynchronously, similar to
1462 other endpoints. Unlike UNIX sockets, socket endpoint must still be
1463 declared as either active or passive.
1464 Socket endpoints behave like non-blocking sockets. In order to support
1466 select and poll semantics, active socket endpoints are associated with
1467 a file descriptor that is signaled whenever the endpoint is ready to
1468 send and/or receive data. The file descriptor may be retrieved using
1469 fi_control.
1470
1472 Operation flags are obtained by OR-ing the following flags together.
1473 Operation flags define the default flags applied to an endpoint’s data
1474 transfer operations, where a flags parameter is not available. Data
1475 transfer operations that take flags as input override the op_flags val‐
1476 ue of transmit or receive context attributes of an endpoint.
1477 FI_COMMIT_COMPLETE
1479 Indicates that a completion should not be generated (locally or
1480 at the peer) until the result of an operation have been made
1481 persistent. See fi_cq(3) for additional details on completion
1482 semantics.
1483 FI_COMPLETION
1485 Indicates that a completion queue entry should be written for
1486 data transfer operations. This flag only applies to operations
1487 issued on an endpoint that was bound to a completion queue with
1488 the FI_SELECTIVE_COMPLETION flag set, otherwise, it is ignored.
1489 See the fi_ep_bind section above for more detail.
1490 FI_DELIVERY_COMPLETE
1492 Indicates that a completion should be generated when the opera‐
1493 tion has been processed by the destination endpoint(s). See
1494 fi_cq(3) for additional details on completion semantics.
1495 FI_INJECT
1497 Indicates that all outbound data buffers should be returned to
1498 the user’s control immediately after a data transfer call re‐
1499 turns, even if the operation is handled asynchronously. This
1500 may require that the provider copy the data into a local buffer
1501 and transfer out of that buffer. A provider can limit the total
1502 amount of send data that may be buffered and/or the size of a
1503 single send that can use this flag. This limit is indicated us‐
1504 ing inject_size (see inject_size above).
1505 FI_INJECT_COMPLETE
1507 Indicates that a completion should be generated when the source
1508 buffer(s) may be reused. See fi_cq(3) for additional details on
1509 completion semantics.
1510 FI_MULTICAST
1512 Indicates that data transfers will target multicast addresses by
1513 default. Any fi_addr_t passed into a data transfer operation
1514 will be treated as a multicast address.
1515 FI_MULTI_RECV
1517 Applies to posted receive operations. This flag allows the user
1518 to post a single buffer that will receive multiple incoming mes‐
1519 sages. Received messages will be packed into the receive buffer
1520 until the buffer has been consumed. Use of this flag may cause
1521 a single posted receive operation to generate multiple comple‐
1522 tions as messages are placed into the buffer. The placement of
1523 received data into the buffer may be subjected to provider spe‐
1524 cific alignment restrictions. The buffer will be released by
1525 the provider when the available buffer space falls below the
1526 specified minimum (see FI_OPT_MIN_MULTI_RECV).
1527 FI_TRANSMIT_COMPLETE
1529 Indicates that a completion should be generated when the trans‐
1530 mit operation has completed relative to the local provider. See
1531 fi_cq(3) for additional details on completion semantics.
1532
1534 Users should call fi_close to release all resources allocated to the
1535 fabric endpoint.
1536 Endpoints allocated with the FI_CONTEXT or FI_CONTEXT2 mode bits set
1538 must typically provide struct fi_context(2) as their per operation con‐
1539 text parameter. (See fi_getinfo.3 for details.) However, when FI_SE‐
1540 LECTIVE_COMPLETION is enabled to suppress CQ completion entries, and an
1541 operation is initiated without the FI_COMPLETION flag set, then the
1542 context parameter is ignored. An application does not need to pass in
1543 a valid struct fi_context(2) into such data transfers.
1544 Operations that complete in error that are not associated with valid
1546 operational context will use the endpoint context in any error report‐
1547 ing structures.
1548 Although applications typically associate individual completions with
1550 either completion queues or counters, an endpoint can be attached to
1551 both a counter and completion queue. When combined with using selec‐
1552 tive completions, this allows an application to use counters to track
1553 successful completions, with a CQ used to report errors. Operations
1554 that complete with an error increment the error counter and generate a
1555 CQ completion event.
1556 As mentioned in fi_getinfo(3), the ep_attr structure can be used to
1558 query providers that support various endpoint attributes. fi_getinfo
1559 can return provider info structures that can support the minimal set of
1560 requirements (such that the application maintains correctness). Howev‐
1561 er, it can also return provider info structures that exceed application
1562 requirements. As an example, consider an application requesting
1563 msg_order as FI_ORDER_NONE. The resulting output from fi_getinfo may
1564 have all the ordering bits set. The application can reset the ordering
1565 bits it does not require before creating the endpoint. The provider is
1566 free to implement a stricter ordering than is required by the applica‐
1567 tion.
1568
1570 Returns 0 on success. On error, a negative value corresponding to fab‐
1571 ric errno is returned. For fi_cancel, a return value of 0 indicates
1572 that the cancel request was submitted for processing.
1573 Fabric errno values are defined in rdma/fi_errno.h.
1575
1577 -FI_EDOMAIN
1578 A resource domain was not bound to the endpoint or an attempt
1579 was made to bind multiple domains.
1580 -FI_ENOCQ
1582 The endpoint has not been configured with necessary event queue.
1583 -FI_EOPBADSTATE
1585 The endpoint’s state does not permit the requested operation.
1586
1588 fi_getinfo(3), fi_domain(3), fi_cq(3) fi_msg(3), fi_tagged(3),
1589 fi_rma(3) fi_peer(3)
1590
1592 OpenFabrics.
1593Libfabric Programmer’s Manual 2022-12-11 fi_endpoint(3)