1 DMA Buffer Sharing API Guide
2 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
5 <sumit dot semwal at linaro dot org>
6 <sumit dot semwal at ti dot com>
8 This document serves as a guide to device-driver writers on what is the dma-buf
9 buffer sharing API, how to use it for exporting and using shared buffers.
11 Any device driver which wishes to be a part of DMA buffer sharing, can do so as
12 either the 'exporter' of buffers, or the 'user' of buffers.
14 Say a driver A wants to use buffers created by driver B, then we call B as the
15 exporter, and A as buffer-user.
18 - implements and manages operations[1] for the buffer
19 - allows other users to share the buffer by using dma_buf sharing APIs,
20 - manages the details of buffer allocation,
21 - decides about the actual backing storage where this allocation happens,
22 - takes care of any migration of scatterlist - for all (shared) users of this
26 - is one of (many) sharing users of the buffer.
27 - doesn't need to worry about how the buffer is allocated, or where.
28 - needs a mechanism to get access to the scatterlist that makes up this buffer
29 in memory, mapped into its own address space, so it can access the same area
32 *IMPORTANT*: [see https://lkml.org/lkml/2011/12/20/211 for more details]
33 For this first version, A buffer shared using the dma_buf sharing API:
34 - *may* be exported to user space using "mmap" *ONLY* by exporter, outside of
36 - may be used *ONLY* by importers that do not need CPU access to the buffer.
38 The dma_buf buffer sharing API usage contains the following steps:
40 1. Exporter announces that it wishes to export a buffer
41 2. Userspace gets the file descriptor associated with the exported buffer, and
42 passes it around to potential buffer-users based on use case
43 3. Each buffer-user 'connects' itself to the buffer
44 4. When needed, buffer-user requests access to the buffer from exporter
45 5. When finished with its use, the buffer-user notifies end-of-DMA to exporter
46 6. when buffer-user is done using this buffer completely, it 'disconnects'
47 itself from the buffer.
50 1. Exporter's announcement of buffer export
52 The buffer exporter announces its wish to export a buffer. In this, it
53 connects its own private buffer data, provides implementation for operations
54 that can be performed on the exported dma_buf, and flags for the file
55 associated with this buffer.
58 struct dma_buf *dma_buf_export(void *priv, struct dma_buf_ops *ops,
59 size_t size, int flags)
61 If this succeeds, dma_buf_export allocates a dma_buf structure, and returns a
62 pointer to the same. It also associates an anonymous file with this buffer,
63 so it can be exported. On failure to allocate the dma_buf object, it returns
66 2. Userspace gets a handle to pass around to potential buffer-users
68 Userspace entity requests for a file-descriptor (fd) which is a handle to the
69 anonymous file associated with the buffer. It can then share the fd with other
70 drivers and/or processes.
73 int dma_buf_fd(struct dma_buf *dmabuf)
75 This API installs an fd for the anonymous file associated with this buffer;
76 returns either 'fd', or error.
78 3. Each buffer-user 'connects' itself to the buffer
80 Each buffer-user now gets a reference to the buffer, using the fd passed to
84 struct dma_buf *dma_buf_get(int fd)
86 This API will return a reference to the dma_buf, and increment refcount for
89 After this, the buffer-user needs to attach its device with the buffer, which
90 helps the exporter to know of device buffer constraints.
93 struct dma_buf_attachment *dma_buf_attach(struct dma_buf *dmabuf,
96 This API returns reference to an attachment structure, which is then used
97 for scatterlist operations. It will optionally call the 'attach' dma_buf
98 operation, if provided by the exporter.
100 The dma-buf sharing framework does the bookkeeping bits related to managing
101 the list of all attachments to a buffer.
103 Until this stage, the buffer-exporter has the option to choose not to actually
104 allocate the backing storage for this buffer, but wait for the first buffer-user
105 to request use of buffer for allocation.
108 4. When needed, buffer-user requests access to the buffer
110 Whenever a buffer-user wants to use the buffer for any DMA, it asks for
111 access to the buffer using dma_buf_map_attachment API. At least one attach to
112 the buffer must have happened before map_dma_buf can be called.
115 struct sg_table * dma_buf_map_attachment(struct dma_buf_attachment *,
116 enum dma_data_direction);
118 This is a wrapper to dma_buf->ops->map_dma_buf operation, which hides the
119 "dma_buf->ops->" indirection from the users of this interface.
121 In struct dma_buf_ops, map_dma_buf is defined as
122 struct sg_table * (*map_dma_buf)(struct dma_buf_attachment *,
123 enum dma_data_direction);
125 It is one of the buffer operations that must be implemented by the exporter.
126 It should return the sg_table containing scatterlist for this buffer, mapped
127 into caller's address space.
129 If this is being called for the first time, the exporter can now choose to
130 scan through the list of attachments for this buffer, collate the requirements
131 of the attached devices, and choose an appropriate backing storage for the
134 Based on enum dma_data_direction, it might be possible to have multiple users
135 accessing at the same time (for reading, maybe), or any other kind of sharing
136 that the exporter might wish to make available to buffer-users.
138 map_dma_buf() operation can return -EINTR if it is interrupted by a signal.
141 5. When finished, the buffer-user notifies end-of-DMA to exporter
143 Once the DMA for the current buffer-user is over, it signals 'end-of-DMA' to
144 the exporter using the dma_buf_unmap_attachment API.
147 void dma_buf_unmap_attachment(struct dma_buf_attachment *,
150 This is a wrapper to dma_buf->ops->unmap_dma_buf() operation, which hides the
151 "dma_buf->ops->" indirection from the users of this interface.
153 In struct dma_buf_ops, unmap_dma_buf is defined as
154 void (*unmap_dma_buf)(struct dma_buf_attachment *, struct sg_table *);
156 unmap_dma_buf signifies the end-of-DMA for the attachment provided. Like
157 map_dma_buf, this API also must be implemented by the exporter.
160 6. when buffer-user is done using this buffer, it 'disconnects' itself from the
163 After the buffer-user has no more interest in using this buffer, it should
164 disconnect itself from the buffer:
166 - it first detaches itself from the buffer.
169 void dma_buf_detach(struct dma_buf *dmabuf,
170 struct dma_buf_attachment *dmabuf_attach);
172 This API removes the attachment from the list in dmabuf, and optionally calls
173 dma_buf->ops->detach(), if provided by exporter, for any housekeeping bits.
175 - Then, the buffer-user returns the buffer reference to exporter.
178 void dma_buf_put(struct dma_buf *dmabuf);
180 This API then reduces the refcount for this buffer.
182 If, as a result of this call, the refcount becomes 0, the 'release' file
183 operation related to this fd is called. It calls the dmabuf->ops->release()
184 operation in turn, and frees the memory allocated for dmabuf when exported.
187 - Importance of attach-detach and {map,unmap}_dma_buf operation pairs
188 The attach-detach calls allow the exporter to figure out backing-storage
189 constraints for the currently-interested devices. This allows preferential
190 allocation, and/or migration of pages across different types of storage
191 available, if possible.
193 Bracketing of DMA access with {map,unmap}_dma_buf operations is essential
194 to allow just-in-time backing of storage, and migration mid-way through a
197 - Migration of backing storage if needed
199 - at least one map_dma_buf has happened,
200 - and the backing storage has been allocated for this buffer,
201 another new buffer-user intends to attach itself to this buffer, it might
202 be allowed, if possible for the exporter.
204 In case it is allowed by the exporter:
205 if the new buffer-user has stricter 'backing-storage constraints', and the
206 exporter can handle these constraints, the exporter can just stall on the
207 map_dma_buf until all outstanding access is completed (as signalled by
209 Once all users have finished accessing and have unmapped this buffer, the
210 exporter could potentially move the buffer to the stricter backing-storage,
211 and then allow further {map,unmap}_dma_buf operations from any buffer-user
212 from the migrated backing-storage.
214 If the exporter cannot fulfil the backing-storage constraints of the new
215 buffer-user device as requested, dma_buf_attach() would return an error to
216 denote non-compatibility of the new buffer-sharing request with the current
219 If the exporter chooses not to allow an attach() operation once a
220 map_dma_buf() API has been called, it simply returns an error.
223 [1] struct dma_buf_ops in include/linux/dma-buf.h
224 [2] All interfaces mentioned above defined in include/linux/dma-buf.h