1====================
2DMA Engine API Guide
3====================
4
5Vinod Koul <vinod dot koul at intel.com>
6
7.. note:: For DMA Engine usage in async_tx please see:
8          ``Documentation/crypto/async-tx-api.txt``
9
10
11Below is a guide to device driver writers on how to use the Slave-DMA API of the
12DMA Engine. This is applicable only for slave DMA usage only.
13
14DMA usage
15=========
16
17The slave DMA usage consists of following steps:
18
19- Allocate a DMA slave channel
20
21- Set slave and controller specific parameters
22
23- Get a descriptor for transaction
24
25- Submit the transaction
26
27- Issue pending requests and wait for callback notification
28
29The details of these operations are:
30
311. Allocate a DMA slave channel
32
33   Channel allocation is slightly different in the slave DMA context,
34   client drivers typically need a channel from a particular DMA
35   controller only and even in some cases a specific channel is desired.
36   To request a channel dma_request_chan() API is used.
37
38   Interface:
39
40   .. code-block:: c
41
42      struct dma_chan *dma_request_chan(struct device *dev, const char *name);
43
44   Which will find and return the ``name`` DMA channel associated with the 'dev'
45   device. The association is done via DT, ACPI or board file based
46   dma_slave_map matching table.
47
48   A channel allocated via this interface is exclusive to the caller,
49   until dma_release_channel() is called.
50
512. Set slave and controller specific parameters
52
53   Next step is always to pass some specific information to the DMA
54   driver. Most of the generic information which a slave DMA can use
55   is in struct dma_slave_config. This allows the clients to specify
56   DMA direction, DMA addresses, bus widths, DMA burst lengths etc
57   for the peripheral.
58
59   If some DMA controllers have more parameters to be sent then they
60   should try to embed struct dma_slave_config in their controller
61   specific structure. That gives flexibility to client to pass more
62   parameters, if required.
63
64   Interface:
65
66   .. code-block:: c
67
68      int dmaengine_slave_config(struct dma_chan *chan,
69			struct dma_slave_config *config)
70
71   Please see the dma_slave_config structure definition in dmaengine.h
72   for a detailed explanation of the struct members. Please note
73   that the 'direction' member will be going away as it duplicates the
74   direction given in the prepare call.
75
763. Get a descriptor for transaction
77
78  For slave usage the various modes of slave transfers supported by the
79  DMA-engine are:
80
81  - slave_sg: DMA a list of scatter gather buffers from/to a peripheral
82
83  - dma_cyclic: Perform a cyclic DMA operation from/to a peripheral till the
84    operation is explicitly stopped.
85
86  - interleaved_dma: This is common to Slave as well as M2M clients. For slave
87    address of devices' fifo could be already known to the driver.
88    Various types of operations could be expressed by setting
89    appropriate values to the 'dma_interleaved_template' members.
90
91  A non-NULL return of this transfer API represents a "descriptor" for
92  the given transaction.
93
94  Interface:
95
96  .. code-block:: c
97
98     struct dma_async_tx_descriptor *dmaengine_prep_slave_sg(
99		struct dma_chan *chan, struct scatterlist *sgl,
100		unsigned int sg_len, enum dma_data_direction direction,
101		unsigned long flags);
102
103     struct dma_async_tx_descriptor *dmaengine_prep_dma_cyclic(
104		struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
105		size_t period_len, enum dma_data_direction direction);
106
107     struct dma_async_tx_descriptor *dmaengine_prep_interleaved_dma(
108		struct dma_chan *chan, struct dma_interleaved_template *xt,
109		unsigned long flags);
110
111  The peripheral driver is expected to have mapped the scatterlist for
112  the DMA operation prior to calling dmaengine_prep_slave_sg(), and must
113  keep the scatterlist mapped until the DMA operation has completed.
114  The scatterlist must be mapped using the DMA struct device.
115  If a mapping needs to be synchronized later, dma_sync_*_for_*() must be
116  called using the DMA struct device, too.
117  So, normal setup should look like this:
118
119  .. code-block:: c
120
121     nr_sg = dma_map_sg(chan->device->dev, sgl, sg_len);
122	if (nr_sg == 0)
123		/* error */
124
125	desc = dmaengine_prep_slave_sg(chan, sgl, nr_sg, direction, flags);
126
127  Once a descriptor has been obtained, the callback information can be
128  added and the descriptor must then be submitted. Some DMA engine
129  drivers may hold a spinlock between a successful preparation and
130  submission so it is important that these two operations are closely
131  paired.
132
133  .. note::
134
135     Although the async_tx API specifies that completion callback
136     routines cannot submit any new operations, this is not the
137     case for slave/cyclic DMA.
138
139     For slave DMA, the subsequent transaction may not be available
140     for submission prior to callback function being invoked, so
141     slave DMA callbacks are permitted to prepare and submit a new
142     transaction.
143
144     For cyclic DMA, a callback function may wish to terminate the
145     DMA via dmaengine_terminate_async().
146
147     Therefore, it is important that DMA engine drivers drop any
148     locks before calling the callback function which may cause a
149     deadlock.
150
151     Note that callbacks will always be invoked from the DMA
152     engines tasklet, never from interrupt context.
153
1544. Submit the transaction
155
156   Once the descriptor has been prepared and the callback information
157   added, it must be placed on the DMA engine drivers pending queue.
158
159   Interface:
160
161   .. code-block:: c
162
163      dma_cookie_t dmaengine_submit(struct dma_async_tx_descriptor *desc)
164
165   This returns a cookie can be used to check the progress of DMA engine
166   activity via other DMA engine calls not covered in this document.
167
168   dmaengine_submit() will not start the DMA operation, it merely adds
169   it to the pending queue. For this, see step 5, dma_async_issue_pending.
170
171   .. note::
172
173      After calling ``dmaengine_submit()`` the submitted transfer descriptor
174      (``struct dma_async_tx_descriptor``) belongs to the DMA engine.
175      Consequently, the client must consider invalid the pointer to that
176      descriptor.
177
1785. Issue pending DMA requests and wait for callback notification
179
180   The transactions in the pending queue can be activated by calling the
181   issue_pending API. If channel is idle then the first transaction in
182   queue is started and subsequent ones queued up.
183
184   On completion of each DMA operation, the next in queue is started and
185   a tasklet triggered. The tasklet will then call the client driver
186   completion callback routine for notification, if set.
187
188   Interface:
189
190   .. code-block:: c
191
192      void dma_async_issue_pending(struct dma_chan *chan);
193
194Further APIs:
195-------------
196
1971. Terminate APIs
198
199   .. code-block:: c
200
201      int dmaengine_terminate_sync(struct dma_chan *chan)
202      int dmaengine_terminate_async(struct dma_chan *chan)
203      int dmaengine_terminate_all(struct dma_chan *chan) /* DEPRECATED */
204
205   This causes all activity for the DMA channel to be stopped, and may
206   discard data in the DMA FIFO which hasn't been fully transferred.
207   No callback functions will be called for any incomplete transfers.
208
209   Two variants of this function are available.
210
211   dmaengine_terminate_async() might not wait until the DMA has been fully
212   stopped or until any running complete callbacks have finished. But it is
213   possible to call dmaengine_terminate_async() from atomic context or from
214   within a complete callback. dmaengine_synchronize() must be called before it
215   is safe to free the memory accessed by the DMA transfer or free resources
216   accessed from within the complete callback.
217
218   dmaengine_terminate_sync() will wait for the transfer and any running
219   complete callbacks to finish before it returns. But the function must not be
220   called from atomic context or from within a complete callback.
221
222   dmaengine_terminate_all() is deprecated and should not be used in new code.
223
2242. Pause API
225
226   .. code-block:: c
227
228      int dmaengine_pause(struct dma_chan *chan)
229
230   This pauses activity on the DMA channel without data loss.
231
2323. Resume API
233
234   .. code-block:: c
235
236       int dmaengine_resume(struct dma_chan *chan)
237
238   Resume a previously paused DMA channel. It is invalid to resume a
239   channel which is not currently paused.
240
2414. Check Txn complete
242
243   .. code-block:: c
244
245      enum dma_status dma_async_is_tx_complete(struct dma_chan *chan,
246		dma_cookie_t cookie, dma_cookie_t *last, dma_cookie_t *used)
247
248   This can be used to check the status of the channel. Please see
249   the documentation in include/linux/dmaengine.h for a more complete
250   description of this API.
251
252   This can be used in conjunction with dma_async_is_complete() and
253   the cookie returned from dmaengine_submit() to check for
254   completion of a specific DMA transaction.
255
256   .. note::
257
258      Not all DMA engine drivers can return reliable information for
259      a running DMA channel. It is recommended that DMA engine users
260      pause or stop (via dmaengine_terminate_all()) the channel before
261      using this API.
262
2635. Synchronize termination API
264
265   .. code-block:: c
266
267      void dmaengine_synchronize(struct dma_chan *chan)
268
269   Synchronize the termination of the DMA channel to the current context.
270
271   This function should be used after dmaengine_terminate_async() to synchronize
272   the termination of the DMA channel to the current context. The function will
273   wait for the transfer and any running complete callbacks to finish before it
274   returns.
275
276   If dmaengine_terminate_async() is used to stop the DMA channel this function
277   must be called before it is safe to free memory accessed by previously
278   submitted descriptors or to free any resources accessed within the complete
279   callback of previously submitted descriptors.
280
281   The behavior of this function is undefined if dma_async_issue_pending() has
282   been called between dmaengine_terminate_async() and this function.
283