blob: b8e86460046e7ee3dba003c12de57ae2d6f7fb36 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001 Dynamic DMA mapping using the generic device
2 ============================================
3
4 James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
5
6This document describes the DMA API. For a more gentle introduction
7phrased in terms of the pci_ equivalents (and actual examples) see
8DMA-mapping.txt
9
10This API is split into two pieces. Part I describes the API and the
11corresponding pci_ API. Part II describes the extensions to the API
12for supporting non-consistent memory machines. Unless you know that
13your driver absolutely has to support non-consistent platforms (this
14is usually only legacy platforms) you should only use the API
15described in part I.
16
17Part I - pci_ and dma_ Equivalent API
18-------------------------------------
19
20To get the pci_ API, you must #include <linux/pci.h>
21To get the dma_ API, you must #include <linux/dma-mapping.h>
22
23
24Part Ia - Using large dma-coherent buffers
25------------------------------------------
26
27void *
28dma_alloc_coherent(struct device *dev, size_t size,
Randy Dunlapa12e2c62007-07-31 00:38:17 -070029 dma_addr_t *dma_handle, gfp_t flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -070030void *
31pci_alloc_consistent(struct pci_dev *dev, size_t size,
32 dma_addr_t *dma_handle)
33
34Consistent memory is memory for which a write by either the device or
35the processor can immediately be read by the processor or device
David Brownell21440d32006-04-01 10:21:52 -080036without having to worry about caching effects. (You may however need
37to make sure to flush the processor's write buffers before telling
38devices to read that memory.)
Linus Torvalds1da177e2005-04-16 15:20:36 -070039
40This routine allocates a region of <size> bytes of consistent memory.
Randy Dunlapa12e2c62007-07-31 00:38:17 -070041It also returns a <dma_handle> which may be cast to an unsigned
Linus Torvalds1da177e2005-04-16 15:20:36 -070042integer the same width as the bus and used as the physical address
43base of the region.
44
45Returns: a pointer to the allocated region (in the processor's virtual
46address space) or NULL if the allocation failed.
47
48Note: consistent memory can be expensive on some platforms, and the
49minimum allocation length may be as big as a page, so you should
50consolidate your requests for consistent memory as much as possible.
51The simplest way to do that is to use the dma_pool calls (see below).
52
53The flag parameter (dma_alloc_coherent only) allows the caller to
54specify the GFP_ flags (see kmalloc) for the allocation (the
Randy Dunlapa12e2c62007-07-31 00:38:17 -070055implementation may choose to ignore flags that affect the location of
Linus Torvalds1da177e2005-04-16 15:20:36 -070056the returned memory, like GFP_DMA). For pci_alloc_consistent, you
57must assume GFP_ATOMIC behaviour.
58
59void
Randy Dunlapa12e2c62007-07-31 00:38:17 -070060dma_free_coherent(struct device *dev, size_t size, void *cpu_addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -070061 dma_addr_t dma_handle)
62void
Randy Dunlapa12e2c62007-07-31 00:38:17 -070063pci_free_consistent(struct pci_dev *dev, size_t size, void *cpu_addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -070064 dma_addr_t dma_handle)
65
66Free the region of consistent memory you previously allocated. dev,
67size and dma_handle must all be the same as those passed into the
68consistent allocate. cpu_addr must be the virtual address returned by
Randy Dunlapa12e2c62007-07-31 00:38:17 -070069the consistent allocate.
Linus Torvalds1da177e2005-04-16 15:20:36 -070070
David Brownellaa248862007-08-10 13:10:27 -070071Note that unlike their sibling allocation calls, these routines
72may only be called with IRQs enabled.
73
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
75Part Ib - Using small dma-coherent buffers
76------------------------------------------
77
78To get this part of the dma_ API, you must #include <linux/dmapool.h>
79
80Many drivers need lots of small dma-coherent memory regions for DMA
81descriptors or I/O buffers. Rather than allocating in units of a page
82or more using dma_alloc_coherent(), you can use DMA pools. These work
Randy Dunlapa12e2c62007-07-31 00:38:17 -070083much like a struct kmem_cache, except that they use the dma-coherent allocator,
Linus Torvalds1da177e2005-04-16 15:20:36 -070084not __get_free_pages(). Also, they understand common hardware constraints
Randy Dunlapa12e2c62007-07-31 00:38:17 -070085for alignment, like queue heads needing to be aligned on N-byte boundaries.
Linus Torvalds1da177e2005-04-16 15:20:36 -070086
87
88 struct dma_pool *
89 dma_pool_create(const char *name, struct device *dev,
90 size_t size, size_t align, size_t alloc);
91
92 struct pci_pool *
93 pci_pool_create(const char *name, struct pci_device *dev,
94 size_t size, size_t align, size_t alloc);
95
96The pool create() routines initialize a pool of dma-coherent buffers
97for use with a given device. It must be called in a context which
98can sleep.
99
Christoph Lametere18b8902006-12-06 20:33:20 -0800100The "name" is for diagnostics (like a struct kmem_cache name); dev and size
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101are like what you'd pass to dma_alloc_coherent(). The device's hardware
102alignment requirement for this type of data is "align" (which is expressed
103in bytes, and must be a power of two). If your device has no boundary
104crossing restrictions, pass 0 for alloc; passing 4096 says memory allocated
105from this pool must not cross 4KByte boundaries.
106
107
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700108 void *dma_pool_alloc(struct dma_pool *pool, gfp_t gfp_flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109 dma_addr_t *dma_handle);
110
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700111 void *pci_pool_alloc(struct pci_pool *pool, gfp_t gfp_flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 dma_addr_t *dma_handle);
113
114This allocates memory from the pool; the returned memory will meet the size
115and alignment requirements specified at creation time. Pass GFP_ATOMIC to
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700116prevent blocking, or if it's permitted (not in_interrupt, not holding SMP locks),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117pass GFP_KERNEL to allow blocking. Like dma_alloc_coherent(), this returns
118two values: an address usable by the cpu, and the dma address usable by the
119pool's device.
120
121
122 void dma_pool_free(struct dma_pool *pool, void *vaddr,
123 dma_addr_t addr);
124
125 void pci_pool_free(struct pci_pool *pool, void *vaddr,
126 dma_addr_t addr);
127
128This puts memory back into the pool. The pool is what was passed to
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700129the pool allocation routine; the cpu (vaddr) and dma addresses are what
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130were returned when that routine allocated the memory being freed.
131
132
133 void dma_pool_destroy(struct dma_pool *pool);
134
135 void pci_pool_destroy(struct pci_pool *pool);
136
137The pool destroy() routines free the resources of the pool. They must be
138called in a context which can sleep. Make sure you've freed all allocated
139memory back to the pool before you destroy it.
140
141
142Part Ic - DMA addressing limitations
143------------------------------------
144
145int
146dma_supported(struct device *dev, u64 mask)
147int
Randy Dunlap02d15c42008-04-29 01:00:35 -0700148pci_dma_supported(struct pci_dev *hwdev, u64 mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149
150Checks to see if the device can support DMA to the memory described by
151mask.
152
153Returns: 1 if it can and 0 if it can't.
154
155Notes: This routine merely tests to see if the mask is possible. It
156won't change the current mask settings. It is more intended as an
157internal API for use by the platform than an external API for use by
158driver writers.
159
160int
161dma_set_mask(struct device *dev, u64 mask)
162int
163pci_set_dma_mask(struct pci_device *dev, u64 mask)
164
165Checks to see if the mask is possible and updates the device
166parameters if it is.
167
168Returns: 0 if successful and a negative error if not.
169
170u64
171dma_get_required_mask(struct device *dev)
172
173After setting the mask with dma_set_mask(), this API returns the
174actual mask (within that already set) that the platform actually
175requires to operate efficiently. Usually this means the returned mask
176is the minimum required to cover all of memory. Examining the
177required mask gives drivers with variable descriptor sizes the
178opportunity to use smaller descriptors as necessary.
179
180Requesting the required mask does not alter the current mask. If you
181wish to take advantage of it, you should issue another dma_set_mask()
182call to lower the mask again.
183
184
185Part Id - Streaming DMA mappings
186--------------------------------
187
188dma_addr_t
189dma_map_single(struct device *dev, void *cpu_addr, size_t size,
190 enum dma_data_direction direction)
191dma_addr_t
Randy Dunlap02d15c42008-04-29 01:00:35 -0700192pci_map_single(struct pci_dev *hwdev, void *cpu_addr, size_t size,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 int direction)
194
195Maps a piece of processor virtual memory so it can be accessed by the
196device and returns the physical handle of the memory.
197
198The direction for both api's may be converted freely by casting.
199However the dma_ API uses a strongly typed enumerator for its
200direction:
201
202DMA_NONE = PCI_DMA_NONE no direction (used for
203 debugging)
204DMA_TO_DEVICE = PCI_DMA_TODEVICE data is going from the
205 memory to the device
206DMA_FROM_DEVICE = PCI_DMA_FROMDEVICE data is coming from
207 the device to the
208 memory
209DMA_BIDIRECTIONAL = PCI_DMA_BIDIRECTIONAL direction isn't known
210
211Notes: Not all memory regions in a machine can be mapped by this
212API. Further, regions that appear to be physically contiguous in
213kernel virtual space may not be contiguous as physical memory. Since
214this API does not provide any scatter/gather capability, it will fail
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700215if the user tries to map a non-physically contiguous piece of memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216For this reason, it is recommended that memory mapped by this API be
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700217obtained only from sources which guarantee it to be physically contiguous
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218(like kmalloc).
219
220Further, the physical address of the memory must be within the
221dma_mask of the device (the dma_mask represents a bit mask of the
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700222addressable region for the device. I.e., if the physical address of
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223the memory anded with the dma_mask is still equal to the physical
224address, then the device can perform DMA to the memory). In order to
225ensure that the memory allocated by kmalloc is within the dma_mask,
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700226the driver may specify various platform-dependent flags to restrict
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227the physical memory range of the allocation (e.g. on x86, GFP_DMA
228guarantees to be within the first 16Mb of available physical memory,
229as required by ISA devices).
230
231Note also that the above constraints on physical contiguity and
232dma_mask may not apply if the platform has an IOMMU (a device which
233supplies a physical to virtual mapping between the I/O memory bus and
234the device). However, to be portable, device driver writers may *not*
235assume that such an IOMMU exists.
236
237Warnings: Memory coherency operates at a granularity called the cache
238line width. In order for memory mapped by this API to operate
239correctly, the mapped region must begin exactly on a cache line
240boundary and end exactly on one (to prevent two separately mapped
241regions from sharing a single cache line). Since the cache line size
242may not be known at compile time, the API will not enforce this
243requirement. Therefore, it is recommended that driver writers who
244don't take special care to determine the cache line size at run time
245only map virtual regions that begin and end on page boundaries (which
246are guaranteed also to be cache line boundaries).
247
248DMA_TO_DEVICE synchronisation must be done after the last modification
249of the memory region by the software and before it is handed off to
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700250the driver. Once this primitive is used, memory covered by this
251primitive should be treated as read-only by the device. If the device
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252may write to it at any point, it should be DMA_BIDIRECTIONAL (see
253below).
254
255DMA_FROM_DEVICE synchronisation must be done before the driver
256accesses data that may be changed by the device. This memory should
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700257be treated as read-only by the driver. If the driver needs to write
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258to it at any point, it should be DMA_BIDIRECTIONAL (see below).
259
260DMA_BIDIRECTIONAL requires special handling: it means that the driver
261isn't sure if the memory was modified before being handed off to the
262device and also isn't sure if the device will also modify it. Thus,
263you must always sync bidirectional memory twice: once before the
264memory is handed off to the device (to make sure all memory changes
265are flushed from the processor) and once before the data may be
266accessed after being used by the device (to make sure any processor
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700267cache lines are updated with data that the device may have changed).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268
269void
270dma_unmap_single(struct device *dev, dma_addr_t dma_addr, size_t size,
271 enum dma_data_direction direction)
272void
273pci_unmap_single(struct pci_dev *hwdev, dma_addr_t dma_addr,
274 size_t size, int direction)
275
276Unmaps the region previously mapped. All the parameters passed in
277must be identical to those passed in (and returned) by the mapping
278API.
279
280dma_addr_t
281dma_map_page(struct device *dev, struct page *page,
282 unsigned long offset, size_t size,
283 enum dma_data_direction direction)
284dma_addr_t
285pci_map_page(struct pci_dev *hwdev, struct page *page,
286 unsigned long offset, size_t size, int direction)
287void
288dma_unmap_page(struct device *dev, dma_addr_t dma_address, size_t size,
289 enum dma_data_direction direction)
290void
291pci_unmap_page(struct pci_dev *hwdev, dma_addr_t dma_address,
292 size_t size, int direction)
293
294API for mapping and unmapping for pages. All the notes and warnings
295for the other mapping APIs apply here. Also, although the <offset>
296and <size> parameters are provided to do partial page mapping, it is
297recommended that you never use these unless you really know what the
298cache width is.
299
300int
FUJITA Tomonori8d8bb392008-07-25 19:44:49 -0700301dma_mapping_error(struct device *dev, dma_addr_t dma_addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302
303int
FUJITA Tomonori8d8bb392008-07-25 19:44:49 -0700304pci_dma_mapping_error(struct pci_dev *hwdev, dma_addr_t dma_addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305
306In some circumstances dma_map_single and dma_map_page will fail to create
307a mapping. A driver can check for these errors by testing the returned
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700308dma address with dma_mapping_error(). A non-zero return value means the mapping
309could not be created and the driver should take appropriate action (e.g.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310reduce current DMA mapping usage or delay and try again later).
311
David Brownell21440d32006-04-01 10:21:52 -0800312 int
313 dma_map_sg(struct device *dev, struct scatterlist *sg,
314 int nents, enum dma_data_direction direction)
315 int
316 pci_map_sg(struct pci_dev *hwdev, struct scatterlist *sg,
317 int nents, int direction)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318
319Maps a scatter gather list from the block layer.
320
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700321Returns: the number of physical segments mapped (this may be shorter
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322than <nents> passed in if the block layer determines that some
323elements of the scatter/gather list are physically adjacent and thus
324may be mapped with a single entry).
325
326Please note that the sg cannot be mapped again if it has been mapped once.
327The mapping process is allowed to destroy information in the sg.
328
329As with the other mapping interfaces, dma_map_sg can fail. When it
330does, 0 is returned and a driver must take appropriate action. It is
331critical that the driver do something, in the case of a block driver
332aborting the request or even oopsing is better than doing nothing and
333corrupting the filesystem.
334
David Brownell21440d32006-04-01 10:21:52 -0800335With scatterlists, you use the resulting mapping like this:
336
337 int i, count = dma_map_sg(dev, sglist, nents, direction);
338 struct scatterlist *sg;
339
FUJITA Tomonori79eb0142008-09-18 09:35:28 -0700340 for_each_sg(sglist, sg, count, i) {
David Brownell21440d32006-04-01 10:21:52 -0800341 hw_address[i] = sg_dma_address(sg);
342 hw_len[i] = sg_dma_len(sg);
343 }
344
345where nents is the number of entries in the sglist.
346
347The implementation is free to merge several consecutive sglist entries
348into one (e.g. with an IOMMU, or if several pages just happen to be
349physically contiguous) and returns the actual number of sg entries it
350mapped them to. On failure 0, is returned.
351
352Then you should loop count times (note: this can be less than nents times)
353and use sg_dma_address() and sg_dma_len() macros where you previously
354accessed sg->address and sg->length as shown above.
355
356 void
357 dma_unmap_sg(struct device *dev, struct scatterlist *sg,
358 int nhwentries, enum dma_data_direction direction)
359 void
360 pci_unmap_sg(struct pci_dev *hwdev, struct scatterlist *sg,
361 int nents, int direction)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700363Unmap the previously mapped scatter/gather list. All the parameters
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364must be the same as those and passed in to the scatter/gather mapping
365API.
366
367Note: <nents> must be the number you passed in, *not* the number of
368physical entries returned.
369
370void
371dma_sync_single(struct device *dev, dma_addr_t dma_handle, size_t size,
372 enum dma_data_direction direction)
373void
374pci_dma_sync_single(struct pci_dev *hwdev, dma_addr_t dma_handle,
375 size_t size, int direction)
376void
377dma_sync_sg(struct device *dev, struct scatterlist *sg, int nelems,
378 enum dma_data_direction direction)
379void
380pci_dma_sync_sg(struct pci_dev *hwdev, struct scatterlist *sg,
381 int nelems, int direction)
382
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700383Synchronise a single contiguous or scatter/gather mapping. All the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384parameters must be the same as those passed into the single mapping
385API.
386
387Notes: You must do this:
388
389- Before reading values that have been written by DMA from the device
390 (use the DMA_FROM_DEVICE direction)
391- After writing values that will be written to the device using DMA
392 (use the DMA_TO_DEVICE) direction
393- before *and* after handing memory to the device if the memory is
394 DMA_BIDIRECTIONAL
395
396See also dma_map_single().
397
Arthur Kepnera75b0a22008-04-29 01:00:31 -0700398dma_addr_t
399dma_map_single_attrs(struct device *dev, void *cpu_addr, size_t size,
400 enum dma_data_direction dir,
401 struct dma_attrs *attrs)
402
403void
404dma_unmap_single_attrs(struct device *dev, dma_addr_t dma_addr,
405 size_t size, enum dma_data_direction dir,
406 struct dma_attrs *attrs)
407
408int
409dma_map_sg_attrs(struct device *dev, struct scatterlist *sgl,
410 int nents, enum dma_data_direction dir,
411 struct dma_attrs *attrs)
412
413void
414dma_unmap_sg_attrs(struct device *dev, struct scatterlist *sgl,
415 int nents, enum dma_data_direction dir,
416 struct dma_attrs *attrs)
417
418The four functions above are just like the counterpart functions
419without the _attrs suffixes, except that they pass an optional
420struct dma_attrs*.
421
422struct dma_attrs encapsulates a set of "dma attributes". For the
423definition of struct dma_attrs see linux/dma-attrs.h.
424
425The interpretation of dma attributes is architecture-specific, and
426each attribute should be documented in Documentation/DMA-attributes.txt.
427
428If struct dma_attrs* is NULL, the semantics of each of these
429functions is identical to those of the corresponding function
430without the _attrs suffix. As a result dma_map_single_attrs()
431can generally replace dma_map_single(), etc.
432
433As an example of the use of the *_attrs functions, here's how
434you could pass an attribute DMA_ATTR_FOO when mapping memory
435for DMA:
436
437#include <linux/dma-attrs.h>
438/* DMA_ATTR_FOO should be defined in linux/dma-attrs.h and
439 * documented in Documentation/DMA-attributes.txt */
440...
441
442 DEFINE_DMA_ATTRS(attrs);
443 dma_set_attr(DMA_ATTR_FOO, &attrs);
444 ....
445 n = dma_map_sg_attrs(dev, sg, nents, DMA_TO_DEVICE, &attr);
446 ....
447
448Architectures that care about DMA_ATTR_FOO would check for its
449presence in their implementations of the mapping and unmapping
450routines, e.g.:
451
452void whizco_dma_map_sg_attrs(struct device *dev, dma_addr_t dma_addr,
453 size_t size, enum dma_data_direction dir,
454 struct dma_attrs *attrs)
455{
456 ....
457 int foo = dma_get_attr(DMA_ATTR_FOO, attrs);
458 ....
459 if (foo)
460 /* twizzle the frobnozzle */
461 ....
462
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463
464Part II - Advanced dma_ usage
465-----------------------------
466
467Warning: These pieces of the DMA API have no PCI equivalent. They
468should also not be used in the majority of cases, since they cater for
469unlikely corner cases that don't belong in usual drivers.
470
471If you don't understand how cache line coherency works between a
472processor and an I/O device, you should not be using this part of the
473API at all.
474
475void *
476dma_alloc_noncoherent(struct device *dev, size_t size,
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700477 dma_addr_t *dma_handle, gfp_t flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478
479Identical to dma_alloc_coherent() except that the platform will
480choose to return either consistent or non-consistent memory as it sees
481fit. By using this API, you are guaranteeing to the platform that you
482have all the correct and necessary sync points for this memory in the
483driver should it choose to return non-consistent memory.
484
485Note: where the platform can return consistent memory, it will
486guarantee that the sync points become nops.
487
488Warning: Handling non-consistent memory is a real pain. You should
489only ever use this API if you positively know your driver will be
490required to work on one of the rare (usually non-PCI) architectures
491that simply cannot make consistent memory.
492
493void
494dma_free_noncoherent(struct device *dev, size_t size, void *cpu_addr,
495 dma_addr_t dma_handle)
496
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700497Free memory allocated by the nonconsistent API. All parameters must
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498be identical to those passed in (and returned by
499dma_alloc_noncoherent()).
500
501int
Ralf Baechlef67637e2006-12-06 20:38:54 -0800502dma_is_consistent(struct device *dev, dma_addr_t dma_handle)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700504Returns true if the device dev is performing consistent DMA on the memory
Ralf Baechlef67637e2006-12-06 20:38:54 -0800505area pointed to by the dma_handle.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506
507int
508dma_get_cache_alignment(void)
509
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700510Returns the processor cache alignment. This is the absolute minimum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511alignment *and* width that you must observe when either mapping
512memory or doing partial flushes.
513
514Notes: This API may return a number *larger* than the actual cache
515line, but it will guarantee that one or more cache lines fit exactly
516into the width returned by this call. It will also always be a power
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700517of two for easy alignment.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518
519void
520dma_sync_single_range(struct device *dev, dma_addr_t dma_handle,
521 unsigned long offset, size_t size,
522 enum dma_data_direction direction)
523
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700524Does a partial sync, starting at offset and continuing for size. You
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525must be careful to observe the cache alignment and width when doing
526anything like this. You must also be extra careful about accessing
527memory you intend to sync partially.
528
529void
Ralf Baechled3fa72e2006-12-06 20:38:56 -0800530dma_cache_sync(struct device *dev, void *vaddr, size_t size,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 enum dma_data_direction direction)
532
533Do a partial sync of memory that was allocated by
534dma_alloc_noncoherent(), starting at virtual address vaddr and
535continuing on for size. Again, you *must* observe the cache line
536boundaries when doing this.
537
538int
539dma_declare_coherent_memory(struct device *dev, dma_addr_t bus_addr,
540 dma_addr_t device_addr, size_t size, int
541 flags)
542
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543Declare region of memory to be handed out by dma_alloc_coherent when
544it's asked for coherent memory for this device.
545
546bus_addr is the physical address to which the memory is currently
547assigned in the bus responding region (this will be used by the
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700548platform to perform the mapping).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549
550device_addr is the physical address the device needs to be programmed
551with actually to address this memory (this will be handed out as the
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700552dma_addr_t in dma_alloc_coherent()).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553
554size is the size of the area (must be multiples of PAGE_SIZE).
555
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700556flags can be or'd together and are:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557
558DMA_MEMORY_MAP - request that the memory returned from
Matt LaPlante4ae0edc2006-11-30 04:58:40 +0100559dma_alloc_coherent() be directly writable.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560
561DMA_MEMORY_IO - request that the memory returned from
562dma_alloc_coherent() be addressable using read/write/memcpy_toio etc.
563
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700564One or both of these flags must be present.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565
566DMA_MEMORY_INCLUDES_CHILDREN - make the declared memory be allocated by
567dma_alloc_coherent of any child devices of this one (for memory residing
568on a bridge).
569
570DMA_MEMORY_EXCLUSIVE - only allocate memory from the declared regions.
571Do not allow dma_alloc_coherent() to fall back to system memory when
572it's out of memory in the declared region.
573
574The return value will be either DMA_MEMORY_MAP or DMA_MEMORY_IO and
575must correspond to a passed in flag (i.e. no returning DMA_MEMORY_IO
576if only DMA_MEMORY_MAP were passed in) for success or zero for
577failure.
578
579Note, for DMA_MEMORY_IO returns, all subsequent memory returned by
580dma_alloc_coherent() may no longer be accessed directly, but instead
581must be accessed using the correct bus functions. If your driver
582isn't prepared to handle this contingency, it should not specify
583DMA_MEMORY_IO in the input flags.
584
585As a simplification for the platforms, only *one* such region of
586memory may be declared per device.
587
588For reasons of efficiency, most platforms choose to track the declared
589region only at the granularity of a page. For smaller allocations,
590you should use the dma_pool() API.
591
592void
593dma_release_declared_memory(struct device *dev)
594
595Remove the memory region previously declared from the system. This
596API performs *no* in-use checking for this region and will return
597unconditionally having removed all the required structures. It is the
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700598driver's job to ensure that no parts of this memory region are
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599currently in use.
600
601void *
602dma_mark_declared_memory_occupied(struct device *dev,
603 dma_addr_t device_addr, size_t size)
604
605This is used to occupy specific regions of the declared space
606(dma_alloc_coherent() will hand out the first free region it finds).
607
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700608device_addr is the *device* address of the region requested.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609
Randy Dunlapa12e2c62007-07-31 00:38:17 -0700610size is the size (and should be a page-sized multiple).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611
612The return value will be either a pointer to the processor virtual
613address of the memory, or an error (via PTR_ERR()) if any part of the
614region is occupied.