xref: /linux/Documentation/driver-api/io-mapping.rst (revision 4b132aacb0768ac1e652cf517097ea6f237214b9)
1========================
2The io_mapping functions
3========================
4
5API
6===
7
8The io_mapping functions in linux/io-mapping.h provide an abstraction for
9efficiently mapping small regions of an I/O device to the CPU. The initial
10usage is to support the large graphics aperture on 32-bit processors where
11ioremap_wc cannot be used to statically map the entire aperture to the CPU
12as it would consume too much of the kernel address space.
13
14A mapping object is created during driver initialization using::
15
16	struct io_mapping *io_mapping_create_wc(unsigned long base,
17						unsigned long size)
18
19'base' is the bus address of the region to be made
20mappable, while 'size' indicates how large a mapping region to
21enable. Both are in bytes.
22
23This _wc variant provides a mapping which may only be used with
24io_mapping_map_atomic_wc(), io_mapping_map_local_wc() or
25io_mapping_map_wc().
26
27With this mapping object, individual pages can be mapped either temporarily
28or long term, depending on the requirements. Of course, temporary maps are
29more efficient. They come in two flavours::
30
31	void *io_mapping_map_local_wc(struct io_mapping *mapping,
32				      unsigned long offset)
33
34	void *io_mapping_map_atomic_wc(struct io_mapping *mapping,
35				       unsigned long offset)
36
37'offset' is the offset within the defined mapping region.  Accessing
38addresses beyond the region specified in the creation function yields
39undefined results. Using an offset which is not page aligned yields an
40undefined result. The return value points to a single page in CPU address
41space.
42
43This _wc variant returns a write-combining map to the page and may only be
44used with mappings created by io_mapping_create_wc()
45
46Temporary mappings are only valid in the context of the caller. The mapping
47is not guaranteed to be globally visible.
48
49io_mapping_map_local_wc() has a side effect on X86 32bit as it disables
50migration to make the mapping code work. No caller can rely on this side
51effect.
52
53io_mapping_map_atomic_wc() has the side effect of disabling preemption and
54pagefaults. Don't use in new code. Use io_mapping_map_local_wc() instead.
55
56Nested mappings need to be undone in reverse order because the mapping
57code uses a stack for keeping track of them::
58
59 addr1 = io_mapping_map_local_wc(map1, offset1);
60 addr2 = io_mapping_map_local_wc(map2, offset2);
61 ...
62 io_mapping_unmap_local(addr2);
63 io_mapping_unmap_local(addr1);
64
65The mappings are released with::
66
67	void io_mapping_unmap_local(void *vaddr)
68	void io_mapping_unmap_atomic(void *vaddr)
69
70'vaddr' must be the value returned by the last io_mapping_map_local_wc() or
71io_mapping_map_atomic_wc() call. This unmaps the specified mapping and
72undoes the side effects of the mapping functions.
73
74If you need to sleep while holding a mapping, you can use the regular
75variant, although this may be significantly slower::
76
77	void *io_mapping_map_wc(struct io_mapping *mapping,
78				unsigned long offset)
79
80This works like io_mapping_map_atomic/local_wc() except it has no side
81effects and the pointer is globally visible.
82
83The mappings are released with::
84
85	void io_mapping_unmap(void *vaddr)
86
87Use for pages mapped with io_mapping_map_wc().
88
89At driver close time, the io_mapping object must be freed::
90
91	void io_mapping_free(struct io_mapping *mapping)
92