1Buffer Sharing and Synchronization
2==================================
3
4The dma-buf subsystem provides the framework for sharing buffers for
5hardware (DMA) access across multiple device drivers and subsystems, and
6for synchronizing asynchronous hardware access.
7
8This is used, for example, by drm "prime" multi-GPU support, but is of
9course not limited to GPU use cases.
10
11The three main components of this are: (1) dma-buf, representing a
12sg_table and exposed to userspace as a file descriptor to allow passing
13between devices, (2) fence, which provides a mechanism to signal when
14one device has finished access, and (3) reservation, which manages the
15shared or exclusive fence(s) associated with the buffer.
16
17Shared DMA Buffers
18------------------
19
20This document serves as a guide to device-driver writers on what is the dma-buf
21buffer sharing API, how to use it for exporting and using shared buffers.
22
23Any device driver which wishes to be a part of DMA buffer sharing, can do so as
24either the 'exporter' of buffers, or the 'user' or 'importer' of buffers.
25
26Say a driver A wants to use buffers created by driver B, then we call B as the
27exporter, and A as buffer-user/importer.
28
29The exporter
30
31 - implements and manages operations in :c:type:`struct dma_buf_ops
32   <dma_buf_ops>` for the buffer,
33 - allows other users to share the buffer by using dma_buf sharing APIs,
34 - manages the details of buffer allocation, wrapped in a :c:type:`struct
35   dma_buf <dma_buf>`,
36 - decides about the actual backing storage where this allocation happens,
37 - and takes care of any migration of scatterlist - for all (shared) users of
38   this buffer.
39
40The buffer-user
41
42 - is one of (many) sharing users of the buffer.
43 - doesn't need to worry about how the buffer is allocated, or where.
44 - and needs a mechanism to get access to the scatterlist that makes up this
45   buffer in memory, mapped into its own address space, so it can access the
46   same area of memory. This interface is provided by :c:type:`struct
47   dma_buf_attachment <dma_buf_attachment>`.
48
49Any exporters or users of the dma-buf buffer sharing framework must have a
50'select DMA_SHARED_BUFFER' in their respective Kconfigs.
51
52Userspace Interface Notes
53~~~~~~~~~~~~~~~~~~~~~~~~~
54
55Mostly a DMA buffer file descriptor is simply an opaque object for userspace,
56and hence the generic interface exposed is very minimal. There's a few things to
57consider though:
58
59- Since kernel 3.12 the dma-buf FD supports the llseek system call, but only
60  with offset=0 and whence=SEEK_END|SEEK_SET. SEEK_SET is supported to allow
61  the usual size discover pattern size = SEEK_END(0); SEEK_SET(0). Every other
62  llseek operation will report -EINVAL.
63
64  If llseek on dma-buf FDs isn't support the kernel will report -ESPIPE for all
65  cases. Userspace can use this to detect support for discovering the dma-buf
66  size using llseek.
67
68- In order to avoid fd leaks on exec, the FD_CLOEXEC flag must be set
69  on the file descriptor.  This is not just a resource leak, but a
70  potential security hole.  It could give the newly exec'd application
71  access to buffers, via the leaked fd, to which it should otherwise
72  not be permitted access.
73
74  The problem with doing this via a separate fcntl() call, versus doing it
75  atomically when the fd is created, is that this is inherently racy in a
76  multi-threaded app[3].  The issue is made worse when it is library code
77  opening/creating the file descriptor, as the application may not even be
78  aware of the fd's.
79
80  To avoid this problem, userspace must have a way to request O_CLOEXEC
81  flag be set when the dma-buf fd is created.  So any API provided by
82  the exporting driver to create a dmabuf fd must provide a way to let
83  userspace control setting of O_CLOEXEC flag passed in to dma_buf_fd().
84
85- Memory mapping the contents of the DMA buffer is also supported. See the
86  discussion below on `CPU Access to DMA Buffer Objects`_ for the full details.
87
88- The DMA buffer FD is also pollable, see `Implicit Fence Poll Support`_ below for
89  details.
90
91Basic Operation and Device DMA Access
92~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
93
94.. kernel-doc:: drivers/dma-buf/dma-buf.c
95   :doc: dma buf device access
96
97CPU Access to DMA Buffer Objects
98~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
99
100.. kernel-doc:: drivers/dma-buf/dma-buf.c
101   :doc: cpu access
102
103Implicit Fence Poll Support
104~~~~~~~~~~~~~~~~~~~~~~~~~~~
105
106.. kernel-doc:: drivers/dma-buf/dma-buf.c
107   :doc: implicit fence polling
108
109Kernel Functions and Structures Reference
110~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
111
112.. kernel-doc:: drivers/dma-buf/dma-buf.c
113   :export:
114
115.. kernel-doc:: include/linux/dma-buf.h
116   :internal:
117
118Reservation Objects
119-------------------
120
121.. kernel-doc:: drivers/dma-buf/dma-resv.c
122   :doc: Reservation Object Overview
123
124.. kernel-doc:: drivers/dma-buf/dma-resv.c
125   :export:
126
127.. kernel-doc:: include/linux/dma-resv.h
128   :internal:
129
130DMA Fences
131----------
132
133.. kernel-doc:: drivers/dma-buf/dma-fence.c
134   :doc: DMA fences overview
135
136DMA Fence Cross-Driver Contract
137~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
138
139.. kernel-doc:: drivers/dma-buf/dma-fence.c
140   :doc: fence cross-driver contract
141
142DMA Fence Signalling Annotations
143~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
144
145.. kernel-doc:: drivers/dma-buf/dma-fence.c
146   :doc: fence signalling annotation
147
148DMA Fences Functions Reference
149~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
150
151.. kernel-doc:: drivers/dma-buf/dma-fence.c
152   :export:
153
154.. kernel-doc:: include/linux/dma-fence.h
155   :internal:
156
157Seqno Hardware Fences
158~~~~~~~~~~~~~~~~~~~~~
159
160.. kernel-doc:: include/linux/seqno-fence.h
161   :internal:
162
163DMA Fence Array
164~~~~~~~~~~~~~~~
165
166.. kernel-doc:: drivers/dma-buf/dma-fence-array.c
167   :export:
168
169.. kernel-doc:: include/linux/dma-fence-array.h
170   :internal:
171
172DMA Fence uABI/Sync File
173~~~~~~~~~~~~~~~~~~~~~~~~
174
175.. kernel-doc:: drivers/dma-buf/sync_file.c
176   :export:
177
178.. kernel-doc:: include/linux/sync_file.h
179   :internal:
180
181Indefinite DMA Fences
182~~~~~~~~~~~~~~~~~~~~~
183
184At various times &dma_fence with an indefinite time until dma_fence_wait()
185finishes have been proposed. Examples include:
186
187* Future fences, used in HWC1 to signal when a buffer isn't used by the display
188  any longer, and created with the screen update that makes the buffer visible.
189  The time this fence completes is entirely under userspace's control.
190
191* Proxy fences, proposed to handle &drm_syncobj for which the fence has not yet
192  been set. Used to asynchronously delay command submission.
193
194* Userspace fences or gpu futexes, fine-grained locking within a command buffer
195  that userspace uses for synchronization across engines or with the CPU, which
196  are then imported as a DMA fence for integration into existing winsys
197  protocols.
198
199* Long-running compute command buffers, while still using traditional end of
200  batch DMA fences for memory management instead of context preemption DMA
201  fences which get reattached when the compute job is rescheduled.
202
203Common to all these schemes is that userspace controls the dependencies of these
204fences and controls when they fire. Mixing indefinite fences with normal
205in-kernel DMA fences does not work, even when a fallback timeout is included to
206protect against malicious userspace:
207
208* Only the kernel knows about all DMA fence dependencies, userspace is not aware
209  of dependencies injected due to memory management or scheduler decisions.
210
211* Only userspace knows about all dependencies in indefinite fences and when
212  exactly they will complete, the kernel has no visibility.
213
214Furthermore the kernel has to be able to hold up userspace command submission
215for memory management needs, which means we must support indefinite fences being
216dependent upon DMA fences. If the kernel also support indefinite fences in the
217kernel like a DMA fence, like any of the above proposal would, there is the
218potential for deadlocks.
219
220.. kernel-render:: DOT
221   :alt: Indefinite Fencing Dependency Cycle
222   :caption: Indefinite Fencing Dependency Cycle
223
224   digraph "Fencing Cycle" {
225      node [shape=box bgcolor=grey style=filled]
226      kernel [label="Kernel DMA Fences"]
227      userspace [label="userspace controlled fences"]
228      kernel -> userspace [label="memory management"]
229      userspace -> kernel [label="Future fence, fence proxy, ..."]
230
231      { rank=same; kernel userspace }
232   }
233
234This means that the kernel might accidentally create deadlocks
235through memory management dependencies which userspace is unaware of, which
236randomly hangs workloads until the timeout kicks in. Workloads, which from
237userspace's perspective, do not contain a deadlock.  In such a mixed fencing
238architecture there is no single entity with knowledge of all dependencies.
239Thefore preventing such deadlocks from within the kernel is not possible.
240
241The only solution to avoid dependencies loops is by not allowing indefinite
242fences in the kernel. This means:
243
244* No future fences, proxy fences or userspace fences imported as DMA fences,
245  with or without a timeout.
246
247* No DMA fences that signal end of batchbuffer for command submission where
248  userspace is allowed to use userspace fencing or long running compute
249  workloads. This also means no implicit fencing for shared buffers in these
250  cases.
251