1fi_tagged(3) Libfabric v1.18.1 fi_tagged(3)
2
6 fi_tagged - Tagged data transfer operations
7 fi_trecv / fi_trecvv / fi_trecvmsg
9 Post a buffer to receive an incoming message
10 fi_tsend / fi_tsendv / fi_tsendmsg / fi_tinject / fi_tsenddata
12 Initiate an operation to send a message
13
15 #include <rdma/fi_tagged.h>
16 ssize_t fi_trecv(struct fid_ep *ep, void *buf, size_t len, void *desc,
18 fi_addr_t src_addr, uint64_t tag, uint64_t ignore, void *context);
19 ssize_t fi_trecvv(struct fid_ep *ep, const struct iovec *iov, void **desc,
21 size_t count, fi_addr_t src_addr, uint64_t tag, uint64_t ignore,
22 void *context);
23 ssize_t fi_trecvmsg(struct fid_ep *ep, const struct fi_msg_tagged *msg,
25 uint64_t flags);
26 ssize_t fi_tsend(struct fid_ep *ep, const void *buf, size_t len,
28 void *desc, fi_addr_t dest_addr, uint64_t tag, void *context);
29 ssize_t fi_tsendv(struct fid_ep *ep, const struct iovec *iov,
31 void **desc, size_t count, fi_addr_t dest_addr, uint64_t tag,
32 void *context);
33 ssize_t fi_tsendmsg(struct fid_ep *ep, const struct fi_msg_tagged *msg,
35 uint64_t flags);
36 ssize_t fi_tinject(struct fid_ep *ep, const void *buf, size_t len,
38 fi_addr_t dest_addr, uint64_t tag);
39 ssize_t fi_tsenddata(struct fid_ep *ep, const void *buf, size_t len,
41 void *desc, uint64_t data, fi_addr_t dest_addr, uint64_t tag,
42 void *context);
43 ssize_t fi_tinjectdata(struct fid_ep *ep, const void *buf, size_t len,
45 uint64_t data, fi_addr_t dest_addr, uint64_t tag);
46
48 fid Fabric endpoint on which to initiate tagged communication opera‐
49 tion.
50 buf Data buffer to send or receive.
52 len Length of data buffer to send or receive.
54 iov Vectored data buffer.
56 count Count of vectored data entries.
58 tag Tag associated with the message.
60 ignore Mask of bits to ignore applied to the tag for receive opera‐
62 tions.
63 desc Memory descriptor associated with the data buffer. See
65 fi_mr(3).
66 data Remote CQ data to transfer with the sent data.
68 dest_addr
70 Destination address for connectionless transfers. Ignored for
71 connected endpoints.
72 src_addr
74 Source address to receive from for connectionless transfers.
75 Applies only to connectionless endpoints with the FI_DIRECT‐
76 ED_RECV capability enabled, otherwise this field is ignored. If
77 set to FI_ADDR_UNSPEC, any source address may match.
78 msg Message descriptor for send and receive operations.
80 flags Additional flags to apply for the send or receive operation.
82 context
84 User specified pointer to associate with the operation. This
85 parameter is ignored if the operation will not generate a suc‐
86 cessful completion, unless an op flag specifies the context pa‐
87 rameter be used for required input.
88
90 Tagged messages are data transfers which carry a key or tag with the
91 message buffer. The tag is used at the receiving endpoint to match the
92 incoming message with a corresponding receive buffer. Message tags
93 match when the receive buffer tag is the same as the send buffer tag
94 with the ignored bits masked out. This can be stated as:
95 send_tag & ~ignore == recv_tag & ~ignore
97 or
99 send_tag | ignore == recv_tag | ignore
101 In general, message tags are checked against receive buffers in the or‐
103 der in which messages have been posted to the endpoint. See the order‐
104 ing discussion below for more details.
105 The send functions – fi_tsend, fi_tsendv, fi_tsendmsg, fi_tinject, and
107 fi_tsenddata – are used to transmit a tagged message from one endpoint
108 to another endpoint. The main difference between send functions are
109 the number and type of parameters that they accept as input. Other‐
110 wise, they perform the same general function.
111 The receive functions – fi_trecv, fi_trecvv, fi_recvmsg – post a data
113 buffer to an endpoint to receive inbound tagged messages. Similar to
114 the send operations, receive operations operate asynchronously. Users
115 should not touch the posted data buffer(s) until the receive operation
116 has completed. Posted receive buffers are matched with inbound send
117 messages based on the tags associated with the send and receive buf‐
118 fers.
119 An endpoint must be enabled before an application can post send or re‐
121 ceive operations to it. For connected endpoints, receive buffers may
122 be posted prior to connect or accept being called on the endpoint.
123 This ensures that buffers are available to receive incoming data imme‐
124 diately after the connection has been established.
125 Completed message operations are reported to the user through one or
127 more event collectors associated with the endpoint. Users provide con‐
128 text which are associated with each operation, and is returned to the
129 user as part of the event completion. See fi_cq for completion event
130 details.
131 fi_tsend
133 The call fi_tsend transfers the data contained in the user-specified
134 data buffer to a remote endpoint, with message boundaries being main‐
135 tained. The local endpoint must be connected to a remote endpoint or
136 destination before fi_tsend is called. Unless the endpoint has been
137 configured differently, the data buffer passed into fi_tsend must not
138 be touched by the application until the fi_tsend call completes asyn‐
139 chronously.
140 fi_tsendv
142 The fi_tsendv call adds support for a scatter-gather list to fi_tsend.
143 The fi_sendv transfers the set of data buffers referenced by the iov
144 parameter to a remote endpoint as a single message.
145 fi_tsendmsg
147 The fi_tsendmsg call supports data transfers over both connected and
148 connectionless endpoints, with the ability to control the send opera‐
149 tion per call through the use of flags. The fi_tsendmsg function takes
150 a struct fi_msg_tagged as input.
151 struct fi_msg_tagged {
153 const struct iovec *msg_iov; /* scatter-gather array */
154 void *desc; /* data descriptor */
155 size_t iov_count;/* # elements in msg_iov */
156 fi_addr_t addr; /* optional endpoint address */
157 uint64_t tag; /* tag associated with message */
158 uint64_t ignore; /* mask applied to tag for receives */
159 void *context; /* user-defined context */
160 uint64_t data; /* optional immediate data */
161 };
162 fi_tinject
164 The tagged inject call is an optimized version of fi_tsend. It pro‐
165 vides similar completion semantics as fi_inject fi_msg(3).
166 fi_tsenddata
168 The tagged send data call is similar to fi_tsend, but allows for the
169 sending of remote CQ data (see FI_REMOTE_CQ_DATA flag) as part of the
170 transfer.
171 fi_tinjectdata
173 The tagged inject data call is similar to fi_tinject, but allows for
174 the sending of remote CQ data (see FI_REMOTE_CQ_DATA flag) as part of
175 the transfer.
176 fi_trecv
178 The fi_trecv call posts a data buffer to the receive queue of the cor‐
179 responding endpoint. Posted receives are searched in the order in
180 which they were posted in order to match sends. Message boundaries are
181 maintained. The order in which the receives complete is dependent on
182 the endpoint type and protocol.
183 fi_trecvv
185 The fi_trecvv call adds support for a scatter-gather list to fi_trecv.
186 The fi_trecvv posts the set of data buffers referenced by the iov pa‐
187 rameter to a receive incoming data.
188 fi_trecvmsg
190 The fi_trecvmsg call supports posting buffers over both connected and
191 connectionless endpoints, with the ability to control the receive oper‐
192 ation per call through the use of flags. The fi_trecvmsg function
193 takes a struct fi_msg_tagged as input.
194
196 The fi_trecvmsg and fi_tsendmsg calls allow the user to specify flags
197 which can change the default message handling of the endpoint. Flags
198 specified with fi_trecvmsg / fi_tsendmsg override most flags previously
199 configured with the endpoint, except where noted (see fi_endpoint).
200 The following list of flags are usable with fi_trecvmsg and/or
201 fi_tsendmsg.
202 FI_REMOTE_CQ_DATA
204 Applies to fi_tsendmsg and fi_tsenddata. Indicates that remote
205 CQ data is available and should be sent as part of the request.
206 See fi_getinfo for additional details on FI_REMOTE_CQ_DATA.
207 FI_COMPLETION
209 Indicates that a completion entry should be generated for the
210 specified operation. The endpoint must be bound to a completion
211 queue with FI_SELECTIVE_COMPLETION that corresponds to the spec‐
212 ified operation, or this flag is ignored.
213 FI_MORE
215 Indicates that the user has additional requests that will imme‐
216 diately be posted after the current call returns. Use of this
217 flag may improve performance by enabling the provider to opti‐
218 mize its access to the fabric hardware.
219 FI_INJECT
221 Applies to fi_tsendmsg. Indicates that the outbound data buffer
222 should be returned to user immediately after the send call re‐
223 turns, even if the operation is handled asynchronously. This
224 may require that the underlying provider implementation copy the
225 data into a local buffer and transfer out of that buffer. This
226 flag can only be used with messages smaller than inject_size.
227 FI_INJECT_COMPLETE
229 Applies to fi_tsendmsg. Indicates that a completion should be
230 generated when the source buffer(s) may be reused.
231 FI_TRANSMIT_COMPLETE
233 Applies to fi_tsendmsg. Indicates that a completion should not
234 be generated until the operation has been successfully transmit‐
235 ted and is no longer being tracked by the provider.
236 FI_MATCH_COMPLETE
238 Applies to fi_tsendmsg. Indicates that a completion should be
239 generated only after the message has either been matched with a
240 tagged buffer or was discarded by the target application.
241 FI_FENCE
243 Applies to transmits. Indicates that the requested operation,
244 also known as the fenced operation, and any operation posted af‐
245 ter the fenced operation will be deferred until all previous op‐
246 erations targeting the same peer endpoint have completed. Oper‐
247 ations posted after the fencing will see and/or replace the re‐
248 sults of any operations initiated prior to the fenced operation.
249 The ordering of operations starting at the posting of the fenced opera‐
251 tion (inclusive) to the posting of a subsequent fenced operation (ex‐
252 clusive) is controlled by the endpoint’s ordering semantics.
253 The following flags may be used with fi_trecvmsg.
255 FI_PEEK
257 The peek flag may be used to see if a specified message has ar‐
258 rived. A peek request is often useful on endpoints that have
259 provider allocated buffering enabled (see fi_rx_attr to‐
260 tal_buffered_recv). Unlike standard receive operations, a re‐
261 ceive operation with the FI_PEEK flag set does not remain queued
262 with the provider after the peek completes successfully. The
263 peek operation operates asynchronously, and the results of the
264 peek operation are available in the completion queue associated
265 with the endpoint. If no message is found matching the tags
266 specified in the peek request, then a completion queue error en‐
267 try with err field set to FI_ENOMSG will be available.
268 If a peek request locates a matching message, the operation will com‐
270 plete successfully. The returned completion data will indicate the
271 meta-data associated with the message, such as the message length, com‐
272 pletion flags, available CQ data, tag, and source address. The data
273 available is subject to the completion entry format (e.g. struct
274 fi_cq_tagged_entry).
275 An application may supply a buffer if it desires to receive data as a
277 part of the peek operation. In order to receive data as a part of the
278 peek operation, the buf and len fields must be available in the CQ for‐
279 mat. In particular, FI_CQ_FORMAT_CONTEXT and FI_CQ_FORMAT_MSG cannot
280 be used if peek operations desire to obtain a copy of the data. The
281 returned data is limited to the size of the input buffer(s) or the mes‐
282 sage size, if smaller. A provider indicates if data is available by
283 setting the buf field of the CQ entry to the user’s first input buffer.
284 If buf is NULL, no data was available to return. A provider may return
285 NULL even if the peek operation completes successfully. Note that the
286 CQ entry len field will reference the size of the message, not neces‐
287 sarily the size of the returned data.
288 FI_CLAIM
290 If this flag is used in conjunction with FI_PEEK, it indicates
291 if the peek request completes successfully – indicating that a
292 matching message was located – the message is claimed by caller.
293 Claimed messages can only be retrieved using a subsequent,
294 paired receive operation with the FI_CLAIM flag set. A receive
295 operation with the FI_CLAIM flag set, but FI_PEEK not set is
296 used to retrieve a previously claimed message.
297 In order to use the FI_CLAIM flag, an application must supply a struct
299 fi_context structure as the context for the receive operation, or a
300 struct fi_recv_context in the case of buffered receives. The same
301 fi_context structure used for an FI_PEEK + FI_CLAIM operation must be
302 used by the paired FI_CLAIM request.
303 This flag also applies to endpoints configured for FI_BUFFERED_RECV or
305 FI_VARIABLE_MSG. When set, it is used to retrieve a tagged message
306 that was buffered by the provider. See Buffered Tagged Receives sec‐
307 tion for details.
308 FI_DISCARD
310 This flag may be used in conjunction with either FI_PEEK or
311 FI_CLAIM. If this flag is used in conjunction with FI_PEEK, it
312 indicates if the peek request completes successfully – indicat‐
313 ing that a matching message was located – the message is dis‐
314 carded by the provider, as the data is not needed by the appli‐
315 cation. This flag may also be used in conjunction with FI_CLAIM
316 in order to discard a message previously claimed using an
317 FI_PEEK + FI_CLAIM request.
318 This flag also applies to endpoints configured for FI_BUFFERED_RECV or
320 FI_VARIABLE_MSG. When set, it indicates that the provider should free
321 a buffered messages. See Buffered Tagged Receives section for details.
322 If this flag is set, the input buffer(s) and length parameters are ig‐
324 nored.
325
327 See fi_msg(3) for an introduction to buffered receives. The handling
328 of buffered receives differs between fi_msg operations and fi_tagged.
329 Although the provider is responsible for allocating and managing net‐
330 work buffers, the application is responsible for identifying the tags
331 that will be used to match incoming messages. The provider handles
332 matching incoming receives to the application specified tags.
333 When FI_BUFFERED_RECV is enabled, the application posts the tags that
335 will be used for matching purposes. Tags are posted using fi_trecv,
336 fi_trecvv, and fi_trecvmsg; however, parameters related to the input
337 buffers are ignored (e.g. buf, len, iov, desc). When a provider re‐
338 ceives a message for which there is a matching tag, it will write an
339 entry to the completion queue associated with the receiving endpoint.
340 For discussion purposes, the completion queue is assumed to be config‐
342 ured for FI_CQ_FORMAT_TAGGED. The op_context field will point to a
343 struct fi_recv_context.
344 struct fi_recv_context {
346 struct fid_ep *ep;
347 void *context;
348 };
349 The `ep' field will be NULL. The `context' field will match the appli‐
351 cation context specified when posting the tag. Other fields are set as
352 defined in fi_msg(3).
353 After being notified that a buffered receive has arrived, applications
355 must either claim or discard the message as described in fi_msg(3).
356
358 Variable length messages are defined in fi_msg(3). The requirements
359 for handling variable length tagged messages is identical to those de‐
360 fined above for buffered tagged receives.
361
363 The tagged send and receive calls return 0 on success. On error, a
364 negative value corresponding to fabric errno is returned. Fabric er‐
365 rno values are defined in fi_errno.h.
366
368 -FI_EAGAIN
369 See fi_msg(3) for a detailed description of handling FI_EAGAIN.
370 -FI_EINVAL
372 Indicates that an invalid argument was supplied by the user.
373 -FI_EOTHER
375 Indicates that an unspecified error occurred.
376
378 fi_getinfo(3), fi_endpoint(3), fi_domain(3), fi_cq(3)
379
381 OpenFabrics.
382Libfabric Programmer’s Manual 2023-01-02 fi_tagged(3)