1 ==============
2 DMA attributes
3 ==============
4
5 This document describes the semantics of the DMA attributes that are
6 defined in linux/dma-mapping.h.
7
8 DMA_ATTR_WEAK_ORDERING
9 ----------------------
10
11 DMA_ATTR_WEAK_ORDERING specifies that reads and writes to the mapping
12 may be weakly ordered, that is that reads and writes may pass each other.
13
14 Since it is optional for platforms to implement DMA_ATTR_WEAK_ORDERING,
15 those that do not will simply ignore the attribute and exhibit default
16 behavior.
17
18 DMA_ATTR_WRITE_COMBINE
19 ----------------------
20
21 DMA_ATTR_WRITE_COMBINE specifies that writes to the mapping may be
22 buffered to improve performance.
23
24 Since it is optional for platforms to implement DMA_ATTR_WRITE_COMBINE,
25 those that do not will simply ignore the attribute and exhibit default
26 behavior.
27
28 DMA_ATTR_NO_KERNEL_MAPPING
29 --------------------------
30
31 DMA_ATTR_NO_KERNEL_MAPPING lets the platform to avoid creating a kernel
32 virtual mapping for the allocated buffer. On some architectures creating
33 such mapping is non-trivial task and consumes very limited resources
34 (like kernel virtual address space or dma consistent address space).
35 Buffers allocated with this attribute can be only passed to user space
36 by calling dma_mmap_attrs(). By using this API, you are guaranteeing
37 that you won't dereference the pointer returned by dma_alloc_attr(). You
38 can treat it as a cookie that must be passed to dma_mmap_attrs() and
39 dma_free_attrs(). Make sure that both of these also get this attribute
40 set on each call.
41
42 Since it is optional for platforms to implement
43 DMA_ATTR_NO_KERNEL_MAPPING, those that do not will simply ignore the
44 attribute and exhibit default behavior.
45
46 DMA_ATTR_SKIP_CPU_SYNC
47 ----------------------
48
49 By default dma_map_{single,page,sg} functions family transfer a given
50 buffer from CPU domain to device domain. Some advanced use cases might
51 require sharing a buffer between more than one device. This requires
52 having a mapping created separately for each device and is usually
53 performed by calling dma_map_{single,page,sg} function more than once
54 for the given buffer with device pointer to each device taking part in
55 the buffer sharing. The first call transfers a buffer from 'CPU' domain
56 to 'device' domain, what synchronizes CPU caches for the given region
57 (usually it means that the cache has been flushed or invalidated
58 depending on the dma direction). However, next calls to
59 dma_map_{single,page,sg}() for other devices will perform exactly the
60 same synchronization operation on the CPU cache. CPU cache synchronization
61 might be a time consuming operation, especially if the buffers are
62 large, so it is highly recommended to avoid it if possible.
63 DMA_ATTR_SKIP_CPU_SYNC allows platform code to skip synchronization of
64 the CPU cache for the given buffer assuming that it has been already
65 transferred to 'device' domain. This attribute can be also used for
66 dma_unmap_{single,page,sg} functions family to force buffer to stay in
67 device domain after releasing a mapping for it. Use this attribute with
68 care!
69
70 DMA_ATTR_FORCE_CONTIGUOUS
71 -------------------------
72
73 By default DMA-mapping subsystem is allowed to assemble the buffer
74 allocated by dma_alloc_attrs() function from individual pages if it can
75 be mapped as contiguous chunk into device dma address space. By
76 specifying this attribute the allocated buffer is forced to be contiguous
77 also in physical memory.
78
79 DMA_ATTR_ALLOC_SINGLE_PAGES
80 ---------------------------
81
82 This is a hint to the DMA-mapping subsystem that it's probably not worth
83 the time to try to allocate memory to in a way that gives better TLB
84 efficiency (AKA it's not worth trying to build the mapping out of larger
85 pages). You might want to specify this if:
86
87 - You know that the accesses to this memory won't thrash the TLB.
88 You might know that the accesses are likely to be sequential or
89 that they aren't sequential but it's unlikely you'll ping-pong
90 between many addresses that are likely to be in different physical
91 pages.
92 - You know that the penalty of TLB misses while accessing the
93 memory will be small enough to be inconsequential. If you are
94 doing a heavy operation like decryption or decompression this
95 might be the case.
96 - You know that the DMA mapping is fairly transitory. If you expect
97 the mapping to have a short lifetime then it may be worth it to
98 optimize allocation (avoid coming up with large pages) instead of
99 getting the slight performance win of larger pages.
100
101 Setting this hint doesn't guarantee that you won't get huge pages, but it
102 means that we won't try quite as hard to get them.
103
104 .. note:: At the moment DMA_ATTR_ALLOC_SINGLE_PAGES is only implemented on ARM,
105 though ARM64 patches will likely be posted soon.
106
107 DMA_ATTR_NO_WARN
108 ----------------
109
110 This tells the DMA-mapping subsystem to suppress allocation failure reports
111 (similarly to __GFP_NOWARN).
112
113 On some architectures allocation failures are reported with error messages
114 to the system logs. Although this can help to identify and debug problems,
115 drivers which handle failures (eg, retry later) have no problems with them,
116 and can actually flood the system logs with error messages that aren't any
117 problem at all, depending on the implementation of the retry mechanism.
118
119 So, this provides a way for drivers to avoid those error messages on calls
120 where allocation failures are not a problem, and shouldn't bother the logs.
121
122 .. note:: At the moment DMA_ATTR_NO_WARN is only implemented on PowerPC.
123
124 DMA_ATTR_PRIVILEGED
125 -------------------
126
127 Some advanced peripherals such as remote processors and GPUs perform
128 accesses to DMA buffers in both privileged "supervisor" and unprivileged
129 "user" modes. This attribute is used to indicate to the DMA-mapping
130 subsystem that the buffer is fully accessible at the elevated privilege
131 level (and ideally inaccessible or at least read-only at the
132 lesser-privileged levels).
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.