1 ==================== 1 ====================
2 DMA Engine API Guide 2 DMA Engine API Guide
3 ==================== 3 ====================
4 4
5 Vinod Koul <vinod dot koul at intel.com> 5 Vinod Koul <vinod dot koul at intel.com>
6 6
7 .. note:: For DMA Engine usage in async_tx ple 7 .. note:: For DMA Engine usage in async_tx please see:
8 ``Documentation/crypto/async-tx-api. 8 ``Documentation/crypto/async-tx-api.rst``
9 9
10 10
11 Below is a guide to device driver writers on h 11 Below is a guide to device driver writers on how to use the Slave-DMA API of the
12 DMA Engine. This is applicable only for slave 12 DMA Engine. This is applicable only for slave DMA usage only.
13 13
14 DMA usage 14 DMA usage
15 ========= 15 =========
16 16
17 The slave DMA usage consists of following step 17 The slave DMA usage consists of following steps:
18 18
19 - Allocate a DMA slave channel 19 - Allocate a DMA slave channel
20 20
21 - Set slave and controller specific parameters 21 - Set slave and controller specific parameters
22 22
23 - Get a descriptor for transaction 23 - Get a descriptor for transaction
24 24
25 - Submit the transaction 25 - Submit the transaction
26 26
27 - Issue pending requests and wait for callback 27 - Issue pending requests and wait for callback notification
28 28
29 The details of these operations are: 29 The details of these operations are:
30 30
31 1. Allocate a DMA slave channel 31 1. Allocate a DMA slave channel
32 32
33 Channel allocation is slightly different in 33 Channel allocation is slightly different in the slave DMA context,
34 client drivers typically need a channel fro 34 client drivers typically need a channel from a particular DMA
35 controller only and even in some cases a sp 35 controller only and even in some cases a specific channel is desired.
36 To request a channel dma_request_chan() API 36 To request a channel dma_request_chan() API is used.
37 37
38 Interface: 38 Interface:
39 39
40 .. code-block:: c 40 .. code-block:: c
41 41
42 struct dma_chan *dma_request_chan(struct 42 struct dma_chan *dma_request_chan(struct device *dev, const char *name);
43 43
44 Which will find and return the ``name`` DMA 44 Which will find and return the ``name`` DMA channel associated with the 'dev'
45 device. The association is done via DT, ACP 45 device. The association is done via DT, ACPI or board file based
46 dma_slave_map matching table. 46 dma_slave_map matching table.
47 47
48 A channel allocated via this interface is e 48 A channel allocated via this interface is exclusive to the caller,
49 until dma_release_channel() is called. 49 until dma_release_channel() is called.
50 50
51 2. Set slave and controller specific parameter 51 2. Set slave and controller specific parameters
52 52
53 Next step is always to pass some specific i 53 Next step is always to pass some specific information to the DMA
54 driver. Most of the generic information whi 54 driver. Most of the generic information which a slave DMA can use
55 is in struct dma_slave_config. This allows 55 is in struct dma_slave_config. This allows the clients to specify
56 DMA direction, DMA addresses, bus widths, D 56 DMA direction, DMA addresses, bus widths, DMA burst lengths etc
57 for the peripheral. 57 for the peripheral.
58 58
59 If some DMA controllers have more parameter 59 If some DMA controllers have more parameters to be sent then they
60 should try to embed struct dma_slave_config 60 should try to embed struct dma_slave_config in their controller
61 specific structure. That gives flexibility 61 specific structure. That gives flexibility to client to pass more
62 parameters, if required. 62 parameters, if required.
63 63
64 Interface: 64 Interface:
65 65
66 .. code-block:: c 66 .. code-block:: c
67 67
68 int dmaengine_slave_config(struct dma_ch 68 int dmaengine_slave_config(struct dma_chan *chan,
69 struct dma_slave_confi 69 struct dma_slave_config *config)
70 70
71 Please see the dma_slave_config structure d 71 Please see the dma_slave_config structure definition in dmaengine.h
72 for a detailed explanation of the struct me 72 for a detailed explanation of the struct members. Please note
73 that the 'direction' member will be going a 73 that the 'direction' member will be going away as it duplicates the
74 direction given in the prepare call. 74 direction given in the prepare call.
75 75
76 3. Get a descriptor for transaction 76 3. Get a descriptor for transaction
77 77
78 For slave usage the various modes of slave t 78 For slave usage the various modes of slave transfers supported by the
79 DMA-engine are: 79 DMA-engine are:
80 80
81 - slave_sg: DMA a list of scatter gather buf 81 - slave_sg: DMA a list of scatter gather buffers from/to a peripheral
82 82
83 - peripheral_dma_vec: DMA an array of scatte 83 - peripheral_dma_vec: DMA an array of scatter gather buffers from/to a
84 peripheral. Similar to slave_sg, but uses 84 peripheral. Similar to slave_sg, but uses an array of dma_vec
85 structures instead of a scatterlist. 85 structures instead of a scatterlist.
86 86
87 - dma_cyclic: Perform a cyclic DMA operation 87 - dma_cyclic: Perform a cyclic DMA operation from/to a peripheral till the
88 operation is explicitly stopped. 88 operation is explicitly stopped.
89 89
90 - interleaved_dma: This is common to Slave a 90 - interleaved_dma: This is common to Slave as well as M2M clients. For slave
91 address of devices' fifo could be already 91 address of devices' fifo could be already known to the driver.
92 Various types of operations could be expre 92 Various types of operations could be expressed by setting
93 appropriate values to the 'dma_interleaved 93 appropriate values to the 'dma_interleaved_template' members. Cyclic
94 interleaved DMA transfers are also possibl 94 interleaved DMA transfers are also possible if supported by the channel by
95 setting the DMA_PREP_REPEAT transfer flag. 95 setting the DMA_PREP_REPEAT transfer flag.
96 96
97 A non-NULL return of this transfer API repre 97 A non-NULL return of this transfer API represents a "descriptor" for
98 the given transaction. 98 the given transaction.
99 99
100 Interface: 100 Interface:
101 101
102 .. code-block:: c 102 .. code-block:: c
103 103
104 struct dma_async_tx_descriptor *dmaengine 104 struct dma_async_tx_descriptor *dmaengine_prep_slave_sg(
105 struct dma_chan *chan, struct 105 struct dma_chan *chan, struct scatterlist *sgl,
106 unsigned int sg_len, enum dma_ 106 unsigned int sg_len, enum dma_data_direction direction,
107 unsigned long flags); 107 unsigned long flags);
108 108
109 struct dma_async_tx_descriptor *dmaengine 109 struct dma_async_tx_descriptor *dmaengine_prep_peripheral_dma_vec(
110 struct dma_chan *chan, const s 110 struct dma_chan *chan, const struct dma_vec *vecs,
111 size_t nents, enum dma_data_di 111 size_t nents, enum dma_data_direction direction,
112 unsigned long flags); 112 unsigned long flags);
113 113
114 struct dma_async_tx_descriptor *dmaengine 114 struct dma_async_tx_descriptor *dmaengine_prep_dma_cyclic(
115 struct dma_chan *chan, dma_add 115 struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
116 size_t period_len, enum dma_da 116 size_t period_len, enum dma_data_direction direction);
117 117
118 struct dma_async_tx_descriptor *dmaengine 118 struct dma_async_tx_descriptor *dmaengine_prep_interleaved_dma(
119 struct dma_chan *chan, struct 119 struct dma_chan *chan, struct dma_interleaved_template *xt,
120 unsigned long flags); 120 unsigned long flags);
121 121
122 The peripheral driver is expected to have ma 122 The peripheral driver is expected to have mapped the scatterlist for
123 the DMA operation prior to calling dmaengine 123 the DMA operation prior to calling dmaengine_prep_slave_sg(), and must
124 keep the scatterlist mapped until the DMA op 124 keep the scatterlist mapped until the DMA operation has completed.
125 The scatterlist must be mapped using the DMA 125 The scatterlist must be mapped using the DMA struct device.
126 If a mapping needs to be synchronized later, 126 If a mapping needs to be synchronized later, dma_sync_*_for_*() must be
127 called using the DMA struct device, too. 127 called using the DMA struct device, too.
128 So, normal setup should look like this: 128 So, normal setup should look like this:
129 129
130 .. code-block:: c 130 .. code-block:: c
131 131
132 struct device *dma_dev = dmaengine_get_dm 132 struct device *dma_dev = dmaengine_get_dma_device(chan);
133 133
134 nr_sg = dma_map_sg(dma_dev, sgl, sg_len); 134 nr_sg = dma_map_sg(dma_dev, sgl, sg_len);
135 if (nr_sg == 0) 135 if (nr_sg == 0)
136 /* error */ 136 /* error */
137 137
138 desc = dmaengine_prep_slave_sg(chan, s 138 desc = dmaengine_prep_slave_sg(chan, sgl, nr_sg, direction, flags);
139 139
140 Once a descriptor has been obtained, the cal 140 Once a descriptor has been obtained, the callback information can be
141 added and the descriptor must then be submit 141 added and the descriptor must then be submitted. Some DMA engine
142 drivers may hold a spinlock between a succes 142 drivers may hold a spinlock between a successful preparation and
143 submission so it is important that these two 143 submission so it is important that these two operations are closely
144 paired. 144 paired.
145 145
146 .. note:: 146 .. note::
147 147
148 Although the async_tx API specifies that 148 Although the async_tx API specifies that completion callback
149 routines cannot submit any new operations 149 routines cannot submit any new operations, this is not the
150 case for slave/cyclic DMA. 150 case for slave/cyclic DMA.
151 151
152 For slave DMA, the subsequent transaction 152 For slave DMA, the subsequent transaction may not be available
153 for submission prior to callback function 153 for submission prior to callback function being invoked, so
154 slave DMA callbacks are permitted to prep 154 slave DMA callbacks are permitted to prepare and submit a new
155 transaction. 155 transaction.
156 156
157 For cyclic DMA, a callback function may w 157 For cyclic DMA, a callback function may wish to terminate the
158 DMA via dmaengine_terminate_async(). 158 DMA via dmaengine_terminate_async().
159 159
160 Therefore, it is important that DMA engin 160 Therefore, it is important that DMA engine drivers drop any
161 locks before calling the callback functio 161 locks before calling the callback function which may cause a
162 deadlock. 162 deadlock.
163 163
164 Note that callbacks will always be invoke 164 Note that callbacks will always be invoked from the DMA
165 engines tasklet, never from interrupt con 165 engines tasklet, never from interrupt context.
166 166
167 **Optional: per descriptor metadata** 167 **Optional: per descriptor metadata**
168 168
169 DMAengine provides two ways for metadata sup 169 DMAengine provides two ways for metadata support.
170 170
171 DESC_METADATA_CLIENT 171 DESC_METADATA_CLIENT
172 172
173 The metadata buffer is allocated/provided 173 The metadata buffer is allocated/provided by the client driver and it is
174 attached to the descriptor. 174 attached to the descriptor.
175 175
176 .. code-block:: c 176 .. code-block:: c
177 177
178 int dmaengine_desc_attach_metadata(struct 178 int dmaengine_desc_attach_metadata(struct dma_async_tx_descriptor *desc,
179 void *data, 179 void *data, size_t len);
180 180
181 DESC_METADATA_ENGINE 181 DESC_METADATA_ENGINE
182 182
183 The metadata buffer is allocated/managed b 183 The metadata buffer is allocated/managed by the DMA driver. The client
184 driver can ask for the pointer, maximum si 184 driver can ask for the pointer, maximum size and the currently used size of
185 the metadata and can directly update or re 185 the metadata and can directly update or read it.
186 186
187 Because the DMA driver manages the memory 187 Because the DMA driver manages the memory area containing the metadata,
188 clients must make sure that they do not tr 188 clients must make sure that they do not try to access or get the pointer
189 after their transfer completion callback h 189 after their transfer completion callback has run for the descriptor.
190 If no completion callback has been defined 190 If no completion callback has been defined for the transfer, then the
191 metadata must not be accessed after issue_ 191 metadata must not be accessed after issue_pending.
192 In other words: if the aim is to read back 192 In other words: if the aim is to read back metadata after the transfer is
193 completed, then the client must use comple 193 completed, then the client must use completion callback.
194 194
195 .. code-block:: c 195 .. code-block:: c
196 196
197 void *dmaengine_desc_get_metadata_ptr(str 197 void *dmaengine_desc_get_metadata_ptr(struct dma_async_tx_descriptor *desc,
198 size_t *payload_len, size_t *m 198 size_t *payload_len, size_t *max_len);
199 199
200 int dmaengine_desc_set_metadata_len(struc 200 int dmaengine_desc_set_metadata_len(struct dma_async_tx_descriptor *desc,
201 size_t payload_len); 201 size_t payload_len);
202 202
203 Client drivers can query if a given mode is 203 Client drivers can query if a given mode is supported with:
204 204
205 .. code-block:: c 205 .. code-block:: c
206 206
207 bool dmaengine_is_metadata_mode_supported 207 bool dmaengine_is_metadata_mode_supported(struct dma_chan *chan,
208 enum dma_desc_metadata_mode mo 208 enum dma_desc_metadata_mode mode);
209 209
210 Depending on the used mode client drivers mu 210 Depending on the used mode client drivers must follow different flow.
211 211
212 DESC_METADATA_CLIENT 212 DESC_METADATA_CLIENT
213 213
214 - DMA_MEM_TO_DEV / DEV_MEM_TO_MEM: 214 - DMA_MEM_TO_DEV / DEV_MEM_TO_MEM:
215 215
216 1. prepare the descriptor (dmaengine_pre 216 1. prepare the descriptor (dmaengine_prep_*)
217 construct the metadata in the client' 217 construct the metadata in the client's buffer
218 2. use dmaengine_desc_attach_metadata() 218 2. use dmaengine_desc_attach_metadata() to attach the buffer to the
219 descriptor 219 descriptor
220 3. submit the transfer 220 3. submit the transfer
221 221
222 - DMA_DEV_TO_MEM: 222 - DMA_DEV_TO_MEM:
223 223
224 1. prepare the descriptor (dmaengine_pre 224 1. prepare the descriptor (dmaengine_prep_*)
225 2. use dmaengine_desc_attach_metadata() 225 2. use dmaengine_desc_attach_metadata() to attach the buffer to the
226 descriptor 226 descriptor
227 3. submit the transfer 227 3. submit the transfer
228 4. when the transfer is completed, the m 228 4. when the transfer is completed, the metadata should be available in the
229 attached buffer 229 attached buffer
230 230
231 DESC_METADATA_ENGINE 231 DESC_METADATA_ENGINE
232 232
233 - DMA_MEM_TO_DEV / DEV_MEM_TO_MEM: 233 - DMA_MEM_TO_DEV / DEV_MEM_TO_MEM:
234 234
235 1. prepare the descriptor (dmaengine_pre 235 1. prepare the descriptor (dmaengine_prep_*)
236 2. use dmaengine_desc_get_metadata_ptr() 236 2. use dmaengine_desc_get_metadata_ptr() to get the pointer to the
237 engine's metadata area 237 engine's metadata area
238 3. update the metadata at the pointer 238 3. update the metadata at the pointer
239 4. use dmaengine_desc_set_metadata_len() 239 4. use dmaengine_desc_set_metadata_len() to tell the DMA engine the
240 amount of data the client has placed 240 amount of data the client has placed into the metadata buffer
241 5. submit the transfer 241 5. submit the transfer
242 242
243 - DMA_DEV_TO_MEM: 243 - DMA_DEV_TO_MEM:
244 244
245 1. prepare the descriptor (dmaengine_pre 245 1. prepare the descriptor (dmaengine_prep_*)
246 2. submit the transfer 246 2. submit the transfer
247 3. on transfer completion, use dmaengine 247 3. on transfer completion, use dmaengine_desc_get_metadata_ptr() to get
248 the pointer to the engine's metadata 248 the pointer to the engine's metadata area
249 4. read out the metadata from the pointe 249 4. read out the metadata from the pointer
250 250
251 .. note:: 251 .. note::
252 252
253 When DESC_METADATA_ENGINE mode is used th 253 When DESC_METADATA_ENGINE mode is used the metadata area for the descriptor
254 is no longer valid after the transfer has 254 is no longer valid after the transfer has been completed (valid up to the
255 point when the completion callback return 255 point when the completion callback returns if used).
256 256
257 Mixed use of DESC_METADATA_CLIENT / DESC_ 257 Mixed use of DESC_METADATA_CLIENT / DESC_METADATA_ENGINE is not allowed,
258 client drivers must use either of the mod 258 client drivers must use either of the modes per descriptor.
259 259
260 4. Submit the transaction 260 4. Submit the transaction
261 261
262 Once the descriptor has been prepared and t 262 Once the descriptor has been prepared and the callback information
263 added, it must be placed on the DMA engine 263 added, it must be placed on the DMA engine drivers pending queue.
264 264
265 Interface: 265 Interface:
266 266
267 .. code-block:: c 267 .. code-block:: c
268 268
269 dma_cookie_t dmaengine_submit(struct dma 269 dma_cookie_t dmaengine_submit(struct dma_async_tx_descriptor *desc)
270 270
271 This returns a cookie can be used to check 271 This returns a cookie can be used to check the progress of DMA engine
272 activity via other DMA engine calls not cov 272 activity via other DMA engine calls not covered in this document.
273 273
274 dmaengine_submit() will not start the DMA o 274 dmaengine_submit() will not start the DMA operation, it merely adds
275 it to the pending queue. For this, see step 275 it to the pending queue. For this, see step 5, dma_async_issue_pending.
276 276
277 .. note:: 277 .. note::
278 278
279 After calling ``dmaengine_submit()`` the 279 After calling ``dmaengine_submit()`` the submitted transfer descriptor
280 (``struct dma_async_tx_descriptor``) bel 280 (``struct dma_async_tx_descriptor``) belongs to the DMA engine.
281 Consequently, the client must consider i 281 Consequently, the client must consider invalid the pointer to that
282 descriptor. 282 descriptor.
283 283
284 5. Issue pending DMA requests and wait for cal 284 5. Issue pending DMA requests and wait for callback notification
285 285
286 The transactions in the pending queue can b 286 The transactions in the pending queue can be activated by calling the
287 issue_pending API. If channel is idle then 287 issue_pending API. If channel is idle then the first transaction in
288 queue is started and subsequent ones queued 288 queue is started and subsequent ones queued up.
289 289
290 On completion of each DMA operation, the ne 290 On completion of each DMA operation, the next in queue is started and
291 a tasklet triggered. The tasklet will then 291 a tasklet triggered. The tasklet will then call the client driver
292 completion callback routine for notificatio 292 completion callback routine for notification, if set.
293 293
294 Interface: 294 Interface:
295 295
296 .. code-block:: c 296 .. code-block:: c
297 297
298 void dma_async_issue_pending(struct dma_ 298 void dma_async_issue_pending(struct dma_chan *chan);
299 299
300 Further APIs 300 Further APIs
301 ------------ 301 ------------
302 302
303 1. Terminate APIs 303 1. Terminate APIs
304 304
305 .. code-block:: c 305 .. code-block:: c
306 306
307 int dmaengine_terminate_sync(struct dma_ 307 int dmaengine_terminate_sync(struct dma_chan *chan)
308 int dmaengine_terminate_async(struct dma 308 int dmaengine_terminate_async(struct dma_chan *chan)
309 int dmaengine_terminate_all(struct dma_c 309 int dmaengine_terminate_all(struct dma_chan *chan) /* DEPRECATED */
310 310
311 This causes all activity for the DMA channe 311 This causes all activity for the DMA channel to be stopped, and may
312 discard data in the DMA FIFO which hasn't b 312 discard data in the DMA FIFO which hasn't been fully transferred.
313 No callback functions will be called for an 313 No callback functions will be called for any incomplete transfers.
314 314
315 Two variants of this function are available 315 Two variants of this function are available.
316 316
317 dmaengine_terminate_async() might not wait 317 dmaengine_terminate_async() might not wait until the DMA has been fully
318 stopped or until any running complete callb 318 stopped or until any running complete callbacks have finished. But it is
319 possible to call dmaengine_terminate_async( 319 possible to call dmaengine_terminate_async() from atomic context or from
320 within a complete callback. dmaengine_synch 320 within a complete callback. dmaengine_synchronize() must be called before it
321 is safe to free the memory accessed by the 321 is safe to free the memory accessed by the DMA transfer or free resources
322 accessed from within the complete callback. 322 accessed from within the complete callback.
323 323
324 dmaengine_terminate_sync() will wait for th 324 dmaengine_terminate_sync() will wait for the transfer and any running
325 complete callbacks to finish before it retu 325 complete callbacks to finish before it returns. But the function must not be
326 called from atomic context or from within a 326 called from atomic context or from within a complete callback.
327 327
328 dmaengine_terminate_all() is deprecated and 328 dmaengine_terminate_all() is deprecated and should not be used in new code.
329 329
330 2. Pause API 330 2. Pause API
331 331
332 .. code-block:: c 332 .. code-block:: c
333 333
334 int dmaengine_pause(struct dma_chan *cha 334 int dmaengine_pause(struct dma_chan *chan)
335 335
336 This pauses activity on the DMA channel wit 336 This pauses activity on the DMA channel without data loss.
337 337
338 3. Resume API 338 3. Resume API
339 339
340 .. code-block:: c 340 .. code-block:: c
341 341
342 int dmaengine_resume(struct dma_chan *c 342 int dmaengine_resume(struct dma_chan *chan)
343 343
344 Resume a previously paused DMA channel. It 344 Resume a previously paused DMA channel. It is invalid to resume a
345 channel which is not currently paused. 345 channel which is not currently paused.
346 346
347 4. Check Txn complete 347 4. Check Txn complete
348 348
349 .. code-block:: c 349 .. code-block:: c
350 350
351 enum dma_status dma_async_is_tx_complete 351 enum dma_status dma_async_is_tx_complete(struct dma_chan *chan,
352 dma_cookie_t cookie, dma_cooki 352 dma_cookie_t cookie, dma_cookie_t *last, dma_cookie_t *used)
353 353
354 This can be used to check the status of the 354 This can be used to check the status of the channel. Please see
355 the documentation in include/linux/dmaengin 355 the documentation in include/linux/dmaengine.h for a more complete
356 description of this API. 356 description of this API.
357 357
358 This can be used in conjunction with dma_as 358 This can be used in conjunction with dma_async_is_complete() and
359 the cookie returned from dmaengine_submit() 359 the cookie returned from dmaengine_submit() to check for
360 completion of a specific DMA transaction. 360 completion of a specific DMA transaction.
361 361
362 .. note:: 362 .. note::
363 363
364 Not all DMA engine drivers can return re 364 Not all DMA engine drivers can return reliable information for
365 a running DMA channel. It is recommended 365 a running DMA channel. It is recommended that DMA engine users
366 pause or stop (via dmaengine_terminate_a 366 pause or stop (via dmaengine_terminate_all()) the channel before
367 using this API. 367 using this API.
368 368
369 5. Synchronize termination API 369 5. Synchronize termination API
370 370
371 .. code-block:: c 371 .. code-block:: c
372 372
373 void dmaengine_synchronize(struct dma_ch 373 void dmaengine_synchronize(struct dma_chan *chan)
374 374
375 Synchronize the termination of the DMA chan 375 Synchronize the termination of the DMA channel to the current context.
376 376
377 This function should be used after dmaengin 377 This function should be used after dmaengine_terminate_async() to synchronize
378 the termination of the DMA channel to the c 378 the termination of the DMA channel to the current context. The function will
379 wait for the transfer and any running compl 379 wait for the transfer and any running complete callbacks to finish before it
380 returns. 380 returns.
381 381
382 If dmaengine_terminate_async() is used to s 382 If dmaengine_terminate_async() is used to stop the DMA channel this function
383 must be called before it is safe to free me 383 must be called before it is safe to free memory accessed by previously
384 submitted descriptors or to free any resour 384 submitted descriptors or to free any resources accessed within the complete
385 callback of previously submitted descriptor 385 callback of previously submitted descriptors.
386 386
387 The behavior of this function is undefined 387 The behavior of this function is undefined if dma_async_issue_pending() has
388 been called between dmaengine_terminate_asy 388 been called between dmaengine_terminate_async() and this function.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.