1fi_rma(3) Libfabric v1.8.0 fi_rma(3)
2
3
4
6 fi_rma - Remote memory access operations
7
8 fi_read / fi_readv / fi_readmsg
9 Initiates a read from remote memory
10
11 fi_write / fi_writev / fi_writemsg
12 fi_inject_write / fi_writedata : Initiate a write to remote memory
13
15 #include <rdma/fi_rma.h>
16
17 ssize_t fi_read(struct fid_ep *ep, void *buf, size_t len, void *desc,
18 fi_addr_t src_addr, uint64_t addr, uint64_t key, void *context);
19
20 ssize_t fi_readv(struct fid_ep *ep, const struct iovec *iov, void **desc,
21 size_t count, fi_addr_t src_addr, uint64_t addr, uint64_t key,
22 void *context);
23
24 ssize_t fi_readmsg(struct fid_ep *ep, const struct fi_msg_rma *msg,
25 uint64_t flags);
26
27 ssize_t fi_write(struct fid_ep *ep, const void *buf, size_t len,
28 void *desc, fi_addr_t dest_addr, uint64_t addr, uint64_t key,
29 void *context);
30
31 ssize_t fi_writev(struct fid_ep *ep, const struct iovec *iov, void **desc,
32 size_t count, fi_addr_t dest_addr, uint64_t addr, uint64_t key,
33 void *context);
34
35 ssize_t fi_writemsg(struct fid_ep *ep, const struct fi_msg_rma *msg,
36 uint64_t flags);
37
38 ssize_t fi_inject_write(struct fid_ep *ep, const void *buf, size_t len,
39 fi_addr_t dest_addr, uint64_t addr, uint64_t key);
40
41 ssize_t fi_writedata(struct fid_ep *ep, const void *buf, size_t len,
42 void *desc, uint64_t data, fi_addr_t dest_addr, uint64_t addr,
43 uint64_t key, void *context);
44
45 ssize_t fi_inject_writedata(struct fid_ep *ep, const void *buf, size_t len,
46 uint64_t data, fi_addr_t dest_addr, uint64_t addr, uint64_t key);
47
49 ep Fabric endpoint on which to initiate read or write operation.
50
51 buf Local data buffer to read into (read target) or write from
52 (write source)
53
54 len Length of data to read or write, specified in bytes. Valid
55 transfers are from 0 bytes up to the endpoint's max_msg_size.
56
57 iov Vectored data buffer.
58
59 count Count of vectored data entries.
60
61 addr Address of remote memory to access. This will be the virtual
62 address of the remote region in the case of FI_MR_BASIC, or the
63 offset from the starting address in the case of FI_MR_SCALABLE.
64
65 key Protection key associated with the remote memory.
66
67 desc Descriptor associated with the local data buffer
68
69 data Remote CQ data to transfer with the operation.
70
71 dest_addr
72 Destination address for connectionless write transfers. Ignored
73 for connected endpoints.
74
75 src_addr
76 Source address to read from for connectionless transfers. Ig‐
77 nored for connected endpoints.
78
79 msg Message descriptor for read and write operations.
80
81 flags Additional flags to apply for the read or write operation.
82
83 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 RMA (remote memory access) operations are used to transfer data direct‐
91 ly between a local data buffer and a remote data buffer. RMA transfers
92 occur on a byte level granularity, and no message boundaries are main‐
93 tained.
94
95 The write functions -- fi_write, fi_writev, fi_writemsg, fi_in‐
96 ject_write, and fi_writedata -- are used to transmit data into a remote
97 memory buffer. The main difference between write functions are the
98 number and type of parameters that they accept as input. Otherwise,
99 they perform the same general function.
100
101 The read functions -- fi_read, fi_readv, and fi_readmsg -- are used to
102 transfer data from a remote memory region into local data buffer(s).
103 Similar to the write operations, read operations operate asynchronous‐
104 ly. Users should not touch the posted data buffer(s) until the read
105 operation has completed.
106
107 Completed RMA operations are reported to the user through one or more
108 completion queues associated with the endpoint. Users provide context
109 which are associated with each operation, and is returned to the user
110 as part of the completion. See fi_cq for completion event details.
111
112 By default, the remote endpoint does not generate an event or notify
113 the user when a memory region has been accessed by an RMA read or write
114 operation. However, immediate data may be associated with an RMA write
115 operation. RMA writes with immediate data will generate a completion
116 entry at the remote endpoint, so that the immediate data may be deliv‐
117 ered.
118
119 fi_write
120 The call fi_write transfers the data contained in the user-specified
121 data buffer to a remote memory region. The local endpoint must be con‐
122 nected to a remote endpoint or destination before fi_write is called.
123 Unless the endpoint has been configured differently, the data buffer
124 passed into fi_write must not be touched by the application until the
125 fi_write call completes asynchronously.
126
127 fi_writev
128 The fi_writev call adds support for a scatter-gather list to fi_write.
129 The fi_writev transfers the set of data buffers referenced by the iov
130 parameter to the remote memory region.
131
132 fi_writemsg
133 The fi_writemsg call supports data transfers over both connected and
134 unconnected endpoints, with the ability to control the write operation
135 per call through the use of flags. The fi_writemsg function takes a
136 struct fi_msg_rma as input.
137
138 struct fi_msg_rma {
139 const struct iovec *msg_iov; /* local scatter-gather array */
140 void **desc; /* operation descriptor */
141 size_t iov_count; /* # elements in msg_iov */
142 fi_addr_t addr; /* optional endpoint address */
143 const struct fi_rma_iov *rma_iov;/* remote SGL */
144 size_t rma_iov_count;/* # elements in rma_iov */
145 void *context; /* user-defined context */
146 uint64_t data; /* optional immediate data */
147 };
148
149 struct fi_rma_iov {
150 uint64_t addr; /* target RMA address */
151 size_t len; /* size of target buffer */
152 uint64_t key; /* access key */
153 };
154
155 fi_inject_write
156 The write inject call is an optimized version of fi_write. It provides
157 similar completion semantics as fi_inject fi_msg(3).
158
159 fi_writedata
160 The write data call is similar to fi_write, but allows for the sending
161 of remote CQ data (see FI_REMOTE_CQ_DATA flag) as part of the transfer.
162
163 fi_inject_writedata
164 The inject write data call is similar to fi_inject_write, but allows
165 for the sending of remote CQ data (see FI_REMOTE_CQ_DATA flag) as part
166 of the transfer.
167
168 fi_read
169 The fi_read call requests that the remote endpoint transfer data from
170 the remote memory region into the local data buffer. The local end‐
171 point must be connected to a remote endpoint or destination before
172 fi_read is called.
173
174 fi_readv
175 The fi_readv call adds support for a scatter-gather list to fi_read.
176 The fi_readv transfers data from the remote memory region into the set
177 of data buffers referenced by the iov parameter.
178
179 fi_readmsg
180 The fi_readmsg call supports data transfers over both connected and un‐
181 connected endpoints, with the ability to control the read operation per
182 call through the use of flags. The fi_readmsg function takes a struct
183 fi_msg_rma as input.
184
186 The fi_readmsg and fi_writemsg calls allow the user to specify flags
187 which can change the default data transfer operation. Flags specified
188 with fi_readmsg / fi_writemsg override most flags previously configured
189 with the endpoint, except where noted (see fi_endpoint.3). The follow‐
190 ing list of flags are usable with fi_readmsg and/or fi_writemsg.
191
192 FI_REMOTE_CQ_DATA
193 Applies to fi_writemsg and fi_writedata. Indicates that remote
194 CQ data is available and should be sent as part of the request.
195 See fi_getinfo for additional details on FI_REMOTE_CQ_DATA.
196
197 FI_COMPLETION
198 Indicates that a completion entry should be generated for the
199 specified operation. The endpoint must be bound to a completion
200 queue with FI_SELECTIVE_COMPLETION that corresponds to the spec‐
201 ified operation, or this flag is ignored.
202
203 FI_MORE
204 Indicates that the user has additional requests that will imme‐
205 diately be posted after the current call returns. Use of this
206 flag may improve performance by enabling the provider to opti‐
207 mize its access to the fabric hardware.
208
209 FI_INJECT
210 Applies to fi_writemsg. Indicates that the outbound data buffer
211 should be returned to user immediately after the write call re‐
212 turns, even if the operation is handled asynchronously. This
213 may require that the underlying provider implementation copy the
214 data into a local buffer and transfer out of that buffer. This
215 flag can only be used with messages smaller than inject_size.
216
217 FI_INJECT_COMPLETE
218 Applies to fi_writemsg. Indicates that a completion should be
219 generated when the source buffer(s) may be reused.
220
221 FI_TRANSMIT_COMPLETE
222 Applies to fi_writemsg. Indicates that a completion should not
223 be generated until the operation has been successfully transmit‐
224 ted and is no longer being tracked by the provider.
225
226 FI_DELIVERY_COMPLETE
227 Applies to fi_writemsg. Indicates that a completion should be
228 generated when the operation has been processed by the destina‐
229 tion.
230
231 FI_COMMIT_COMPLETE
232 Applies to fi_writemsg when targeting persistent memory regions.
233 Indicates that a completion should be generated only after the
234 result of the operation has been made durable.
235
236 FI_FENCE
237 Applies to transmits. Indicates that the requested operation,
238 also known as the fenced operation, and any operation posted af‐
239 ter the fenced operation will be deferred until all previous op‐
240 erations targeting the same peer endpoint have completed. Oper‐
241 ations posted after the fencing will see and/or replace the re‐
242 sults of any operations initiated prior to the fenced operation.
243
244 The ordering of operations starting at the posting of the fenced opera‐
245 tion (inclusive) to the posting of a subsequent fenced operation (ex‐
246 clusive) is controlled by the endpoint's ordering semantics.
247
249 Returns 0 on success. On error, a negative value corresponding to fab‐
250 ric errno is returned. Fabric errno values are defined in rdma/fi_er‐
251 rno.h.
252
254 -FI_EAGAIN
255 See fi_msg(3) for a detailed description of handling FI_EAGAIN.
256
258 fi_getinfo(3), fi_endpoint(3), fi_domain(3), fi_cq(3)
259
261 OpenFabrics.
262
263
264
265Libfabric Programmer's Manual 2019-02-04 fi_rma(3)