dma-mapping: deprecate dma_zalloc_coherent
[linux-2.6-block.git] / Documentation / DMA-API.txt
CommitLineData
03158a70
MCC
1============================================
2Dynamic DMA mapping using the generic device
3============================================
1da177e4 4
03158a70 5:Author: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
1da177e4
LT
6
7This document describes the DMA API. For a more gentle introduction
77f2ea2f 8of the API (and actual examples), see Documentation/DMA-API-HOWTO.txt.
1da177e4 9
77f2ea2f
BH
10This API is split into two pieces. Part I describes the basic API.
11Part II describes extensions for supporting non-consistent memory
12machines. Unless you know that your driver absolutely has to support
13non-consistent platforms (this is usually only legacy platforms) you
14should only use the API described in part I.
1da177e4 15
03158a70
MCC
16Part I - dma_API
17----------------
1da177e4 18
03158a70 19To get the dma_API, you must #include <linux/dma-mapping.h>. This
77f2ea2f 20provides dma_addr_t and the interfaces described below.
1da177e4 21
3a9ad0b4
YL
22A dma_addr_t can hold any valid DMA address for the platform. It can be
23given to a device to use as a DMA source or target. A CPU cannot reference
24a dma_addr_t directly because there may be translation between its physical
25address space and the DMA address space.
1da177e4 26
77f2ea2f 27Part Ia - Using large DMA-coherent buffers
1da177e4
LT
28------------------------------------------
29
03158a70
MCC
30::
31
32 void *
33 dma_alloc_coherent(struct device *dev, size_t size,
34 dma_addr_t *dma_handle, gfp_t flag)
1da177e4
LT
35
36Consistent memory is memory for which a write by either the device or
37the processor can immediately be read by the processor or device
21440d31
DB
38without having to worry about caching effects. (You may however need
39to make sure to flush the processor's write buffers before telling
40devices to read that memory.)
1da177e4
LT
41
42This routine allocates a region of <size> bytes of consistent memory.
1da177e4 43
77f2ea2f 44It returns a pointer to the allocated region (in the processor's virtual
1da177e4
LT
45address space) or NULL if the allocation failed.
46
77f2ea2f 47It also returns a <dma_handle> which may be cast to an unsigned integer the
3a9ad0b4 48same width as the bus and given to the device as the DMA address base of
77f2ea2f
BH
49the region.
50
1da177e4
LT
51Note: consistent memory can be expensive on some platforms, and the
52minimum allocation length may be as big as a page, so you should
53consolidate your requests for consistent memory as much as possible.
54The simplest way to do that is to use the dma_pool calls (see below).
55
77f2ea2f 56The flag parameter (dma_alloc_coherent() only) allows the caller to
03158a70 57specify the ``GFP_`` flags (see kmalloc()) for the allocation (the
a12e2c6c 58implementation may choose to ignore flags that affect the location of
f5a69f4c 59the returned memory, like GFP_DMA).
1da177e4 60
03158a70
MCC
61::
62
63 void
64 dma_free_coherent(struct device *dev, size_t size, void *cpu_addr,
65 dma_addr_t dma_handle)
1da177e4 66
77f2ea2f
BH
67Free a region of consistent memory you previously allocated. dev,
68size and dma_handle must all be the same as those passed into
69dma_alloc_coherent(). cpu_addr must be the virtual address returned by
70the dma_alloc_coherent().
1da177e4 71
aa24886e
DB
72Note that unlike their sibling allocation calls, these routines
73may only be called with IRQs enabled.
74
1da177e4 75
77f2ea2f 76Part Ib - Using small DMA-coherent buffers
1da177e4
LT
77------------------------------------------
78
03158a70 79To get this part of the dma_API, you must #include <linux/dmapool.h>
1da177e4 80
77f2ea2f 81Many drivers need lots of small DMA-coherent memory regions for DMA
1da177e4
LT
82descriptors or I/O buffers. Rather than allocating in units of a page
83or more using dma_alloc_coherent(), you can use DMA pools. These work
77f2ea2f 84much like a struct kmem_cache, except that they use the DMA-coherent allocator,
1da177e4 85not __get_free_pages(). Also, they understand common hardware constraints
a12e2c6c 86for alignment, like queue heads needing to be aligned on N-byte boundaries.
1da177e4
LT
87
88
03158a70
MCC
89::
90
1da177e4
LT
91 struct dma_pool *
92 dma_pool_create(const char *name, struct device *dev,
93 size_t size, size_t align, size_t alloc);
94
77f2ea2f 95dma_pool_create() initializes a pool of DMA-coherent buffers
1da177e4
LT
96for use with a given device. It must be called in a context which
97can sleep.
98
e18b890b 99The "name" is for diagnostics (like a struct kmem_cache name); dev and size
1da177e4
LT
100are like what you'd pass to dma_alloc_coherent(). The device's hardware
101alignment requirement for this type of data is "align" (which is expressed
102in bytes, and must be a power of two). If your device has no boundary
103crossing restrictions, pass 0 for alloc; passing 4096 says memory allocated
104from this pool must not cross 4KByte boundaries.
105
03158a70 106::
1da177e4 107
03158a70
MCC
108 void *
109 dma_pool_zalloc(struct dma_pool *pool, gfp_t mem_flags,
110 dma_addr_t *handle)
ad82362b
SS
111
112Wraps dma_pool_alloc() and also zeroes the returned memory if the
113allocation attempt succeeded.
114
115
03158a70
MCC
116::
117
118 void *
119 dma_pool_alloc(struct dma_pool *pool, gfp_t gfp_flags,
120 dma_addr_t *dma_handle);
1da177e4 121
77f2ea2f
BH
122This allocates memory from the pool; the returned memory will meet the
123size and alignment requirements specified at creation time. Pass
124GFP_ATOMIC to prevent blocking, or if it's permitted (not
125in_interrupt, not holding SMP locks), pass GFP_KERNEL to allow
126blocking. Like dma_alloc_coherent(), this returns two values: an
f311a724 127address usable by the CPU, and the DMA address usable by the pool's
77f2ea2f 128device.
1da177e4 129
03158a70 130::
1da177e4 131
03158a70
MCC
132 void
133 dma_pool_free(struct dma_pool *pool, void *vaddr,
134 dma_addr_t addr);
1da177e4 135
1da177e4 136This puts memory back into the pool. The pool is what was passed to
f311a724 137dma_pool_alloc(); the CPU (vaddr) and DMA addresses are what
1da177e4
LT
138were returned when that routine allocated the memory being freed.
139
03158a70 140::
1da177e4 141
03158a70
MCC
142 void
143 dma_pool_destroy(struct dma_pool *pool);
1da177e4 144
77f2ea2f 145dma_pool_destroy() frees the resources of the pool. It must be
1da177e4
LT
146called in a context which can sleep. Make sure you've freed all allocated
147memory back to the pool before you destroy it.
148
149
150Part Ic - DMA addressing limitations
151------------------------------------
152
03158a70
MCC
153::
154
155 int
156 dma_set_mask_and_coherent(struct device *dev, u64 mask)
4aa806b7
RK
157
158Checks to see if the mask is possible and updates the device
159streaming and coherent DMA mask parameters if it is.
160
161Returns: 0 if successful and a negative error if not.
162
03158a70
MCC
163::
164
165 int
166 dma_set_mask(struct device *dev, u64 mask)
1da177e4
LT
167
168Checks to see if the mask is possible and updates the device
169parameters if it is.
170
171Returns: 0 if successful and a negative error if not.
172
03158a70
MCC
173::
174
175 int
176 dma_set_coherent_mask(struct device *dev, u64 mask)
6a1961f4
FT
177
178Checks to see if the mask is possible and updates the device
179parameters if it is.
180
181Returns: 0 if successful and a negative error if not.
182
03158a70
MCC
183::
184
185 u64
186 dma_get_required_mask(struct device *dev)
1da177e4 187
175add19
JK
188This API returns the mask that the platform requires to
189operate efficiently. Usually this means the returned mask
1da177e4
LT
190is the minimum required to cover all of memory. Examining the
191required mask gives drivers with variable descriptor sizes the
192opportunity to use smaller descriptors as necessary.
193
194Requesting the required mask does not alter the current mask. If you
175add19
JK
195wish to take advantage of it, you should issue a dma_set_mask()
196call to set the mask to the value returned.
1da177e4
LT
197
198
199Part Id - Streaming DMA mappings
200--------------------------------
201
03158a70
MCC
202::
203
204 dma_addr_t
205 dma_map_single(struct device *dev, void *cpu_addr, size_t size,
206 enum dma_data_direction direction)
1da177e4
LT
207
208Maps a piece of processor virtual memory so it can be accessed by the
3a9ad0b4 209device and returns the DMA address of the memory.
1da177e4 210
77f2ea2f 211The direction for both APIs may be converted freely by casting.
03158a70 212However the dma_API uses a strongly typed enumerator for its
1da177e4
LT
213direction:
214
03158a70 215======================= =============================================
f5a69f4c
FT
216DMA_NONE no direction (used for debugging)
217DMA_TO_DEVICE data is going from the memory to the device
218DMA_FROM_DEVICE data is coming from the device to the memory
219DMA_BIDIRECTIONAL direction isn't known
03158a70
MCC
220======================= =============================================
221
222.. note::
223
224 Not all memory regions in a machine can be mapped by this API.
225 Further, contiguous kernel virtual space may not be contiguous as
226 physical memory. Since this API does not provide any scatter/gather
227 capability, it will fail if the user tries to map a non-physically
228 contiguous piece of memory. For this reason, memory to be mapped by
229 this API should be obtained from sources which guarantee it to be
230 physically contiguous (like kmalloc).
231
232 Further, the DMA address of the memory must be within the
233 dma_mask of the device (the dma_mask is a bit mask of the
234 addressable region for the device, i.e., if the DMA address of
235 the memory ANDed with the dma_mask is still equal to the DMA
236 address, then the device can perform DMA to the memory). To
237 ensure that the memory allocated by kmalloc is within the dma_mask,
238 the driver may specify various platform-dependent flags to restrict
239 the DMA address range of the allocation (e.g., on x86, GFP_DMA
240 guarantees to be within the first 16MB of available DMA addresses,
241 as required by ISA devices).
242
243 Note also that the above constraints on physical contiguity and
244 dma_mask may not apply if the platform has an IOMMU (a device which
245 maps an I/O DMA address to a physical memory address). However, to be
246 portable, device driver writers may *not* assume that such an IOMMU
247 exists.
248
249.. warning::
250
251 Memory coherency operates at a granularity called the cache
252 line width. In order for memory mapped by this API to operate
253 correctly, the mapped region must begin exactly on a cache line
254 boundary and end exactly on one (to prevent two separately mapped
255 regions from sharing a single cache line). Since the cache line size
256 may not be known at compile time, the API will not enforce this
257 requirement. Therefore, it is recommended that driver writers who
258 don't take special care to determine the cache line size at run time
259 only map virtual regions that begin and end on page boundaries (which
260 are guaranteed also to be cache line boundaries).
261
262 DMA_TO_DEVICE synchronisation must be done after the last modification
263 of the memory region by the software and before it is handed off to
264 the device. Once this primitive is used, memory covered by this
265 primitive should be treated as read-only by the device. If the device
266 may write to it at any point, it should be DMA_BIDIRECTIONAL (see
267 below).
268
269 DMA_FROM_DEVICE synchronisation must be done before the driver
270 accesses data that may be changed by the device. This memory should
271 be treated as read-only by the driver. If the driver needs to write
272 to it at any point, it should be DMA_BIDIRECTIONAL (see below).
273
274 DMA_BIDIRECTIONAL requires special handling: it means that the driver
275 isn't sure if the memory was modified before being handed off to the
276 device and also isn't sure if the device will also modify it. Thus,
277 you must always sync bidirectional memory twice: once before the
278 memory is handed off to the device (to make sure all memory changes
279 are flushed from the processor) and once before the data may be
280 accessed after being used by the device (to make sure any processor
281 cache lines are updated with data that the device may have changed).
282
283::
1da177e4 284
03158a70
MCC
285 void
286 dma_unmap_single(struct device *dev, dma_addr_t dma_addr, size_t size,
287 enum dma_data_direction direction)
1da177e4
LT
288
289Unmaps the region previously mapped. All the parameters passed in
290must be identical to those passed in (and returned) by the mapping
291API.
292
03158a70
MCC
293::
294
295 dma_addr_t
296 dma_map_page(struct device *dev, struct page *page,
297 unsigned long offset, size_t size,
298 enum dma_data_direction direction)
299
300 void
301 dma_unmap_page(struct device *dev, dma_addr_t dma_address, size_t size,
302 enum dma_data_direction direction)
1da177e4
LT
303
304API for mapping and unmapping for pages. All the notes and warnings
305for the other mapping APIs apply here. Also, although the <offset>
306and <size> parameters are provided to do partial page mapping, it is
307recommended that you never use these unless you really know what the
308cache width is.
309
03158a70 310::
6f3d8796 311
03158a70
MCC
312 dma_addr_t
313 dma_map_resource(struct device *dev, phys_addr_t phys_addr, size_t size,
314 enum dma_data_direction dir, unsigned long attrs)
315
316 void
317 dma_unmap_resource(struct device *dev, dma_addr_t addr, size_t size,
318 enum dma_data_direction dir, unsigned long attrs)
6f3d8796
NS
319
320API for mapping and unmapping for MMIO resources. All the notes and
321warnings for the other mapping APIs apply here. The API should only be
322used to map device MMIO resources, mapping of RAM is not permitted.
323
03158a70
MCC
324::
325
326 int
327 dma_mapping_error(struct device *dev, dma_addr_t dma_addr)
1da177e4 328
6f3d8796
NS
329In some circumstances dma_map_single(), dma_map_page() and dma_map_resource()
330will fail to create a mapping. A driver can check for these errors by testing
331the returned DMA address with dma_mapping_error(). A non-zero return value
332means the mapping could not be created and the driver should take appropriate
333action (e.g. reduce current DMA mapping usage or delay and try again later).
1da177e4 334
03158a70
MCC
335::
336
21440d31
DB
337 int
338 dma_map_sg(struct device *dev, struct scatterlist *sg,
03158a70 339 int nents, enum dma_data_direction direction)
1da177e4 340
3a9ad0b4 341Returns: the number of DMA address segments mapped (this may be shorter
1d678f36
FT
342than <nents> passed in if some elements of the scatter/gather list are
343physically or virtually adjacent and an IOMMU maps them with a single
344entry).
1da177e4
LT
345
346Please note that the sg cannot be mapped again if it has been mapped once.
347The mapping process is allowed to destroy information in the sg.
348
77f2ea2f 349As with the other mapping interfaces, dma_map_sg() can fail. When it
1da177e4
LT
350does, 0 is returned and a driver must take appropriate action. It is
351critical that the driver do something, in the case of a block driver
352aborting the request or even oopsing is better than doing nothing and
353corrupting the filesystem.
354
03158a70 355With scatterlists, you use the resulting mapping like this::
21440d31
DB
356
357 int i, count = dma_map_sg(dev, sglist, nents, direction);
358 struct scatterlist *sg;
359
79eb0145 360 for_each_sg(sglist, sg, count, i) {
21440d31
DB
361 hw_address[i] = sg_dma_address(sg);
362 hw_len[i] = sg_dma_len(sg);
363 }
364
365where nents is the number of entries in the sglist.
366
367The implementation is free to merge several consecutive sglist entries
368into one (e.g. with an IOMMU, or if several pages just happen to be
369physically contiguous) and returns the actual number of sg entries it
370mapped them to. On failure 0, is returned.
371
372Then you should loop count times (note: this can be less than nents times)
373and use sg_dma_address() and sg_dma_len() macros where you previously
374accessed sg->address and sg->length as shown above.
375
03158a70
MCC
376::
377
21440d31
DB
378 void
379 dma_unmap_sg(struct device *dev, struct scatterlist *sg,
03158a70 380 int nents, enum dma_data_direction direction)
1da177e4 381
a12e2c6c 382Unmap the previously mapped scatter/gather list. All the parameters
1da177e4
LT
383must be the same as those and passed in to the scatter/gather mapping
384API.
385
386Note: <nents> must be the number you passed in, *not* the number of
3a9ad0b4 387DMA address entries returned.
1da177e4 388
03158a70
MCC
389::
390
391 void
392 dma_sync_single_for_cpu(struct device *dev, dma_addr_t dma_handle,
393 size_t size,
394 enum dma_data_direction direction)
395
396 void
397 dma_sync_single_for_device(struct device *dev, dma_addr_t dma_handle,
398 size_t size,
399 enum dma_data_direction direction)
400
401 void
402 dma_sync_sg_for_cpu(struct device *dev, struct scatterlist *sg,
403 int nents,
404 enum dma_data_direction direction)
405
406 void
407 dma_sync_sg_for_device(struct device *dev, struct scatterlist *sg,
408 int nents,
409 enum dma_data_direction direction)
9705ef7e 410
f311a724 411Synchronise a single contiguous or scatter/gather mapping for the CPU
9705ef7e
FT
412and device. With the sync_sg API, all the parameters must be the same
413as those passed into the single mapping API. With the sync_single API,
414you can use dma_handle and size parameters that aren't identical to
415those passed into the single mapping API to do a partial sync.
416
9705ef7e 417
03158a70
MCC
418.. note::
419
420 You must do this:
421
422 - Before reading values that have been written by DMA from the device
423 (use the DMA_FROM_DEVICE direction)
424 - After writing values that will be written to the device using DMA
425 (use the DMA_TO_DEVICE) direction
426 - before *and* after handing memory to the device if the memory is
427 DMA_BIDIRECTIONAL
9705ef7e
FT
428
429See also dma_map_single().
430
03158a70
MCC
431::
432
433 dma_addr_t
434 dma_map_single_attrs(struct device *dev, void *cpu_addr, size_t size,
435 enum dma_data_direction dir,
436 unsigned long attrs)
a75b0a2f 437
03158a70
MCC
438 void
439 dma_unmap_single_attrs(struct device *dev, dma_addr_t dma_addr,
440 size_t size, enum dma_data_direction dir,
441 unsigned long attrs)
a75b0a2f 442
03158a70
MCC
443 int
444 dma_map_sg_attrs(struct device *dev, struct scatterlist *sgl,
445 int nents, enum dma_data_direction dir,
446 unsigned long attrs)
a75b0a2f 447
03158a70
MCC
448 void
449 dma_unmap_sg_attrs(struct device *dev, struct scatterlist *sgl,
450 int nents, enum dma_data_direction dir,
451 unsigned long attrs)
a75b0a2f
AK
452
453The four functions above are just like the counterpart functions
454without the _attrs suffixes, except that they pass an optional
00085f1e 455dma_attrs.
a75b0a2f 456
77f2ea2f 457The interpretation of DMA attributes is architecture-specific, and
a75b0a2f
AK
458each attribute should be documented in Documentation/DMA-attributes.txt.
459
00085f1e
KK
460If dma_attrs are 0, the semantics of each of these functions
461is identical to those of the corresponding function
a75b0a2f
AK
462without the _attrs suffix. As a result dma_map_single_attrs()
463can generally replace dma_map_single(), etc.
464
03158a70 465As an example of the use of the ``*_attrs`` functions, here's how
a75b0a2f 466you could pass an attribute DMA_ATTR_FOO when mapping memory
03158a70 467for DMA::
a75b0a2f 468
03158a70
MCC
469 #include <linux/dma-mapping.h>
470 /* DMA_ATTR_FOO should be defined in linux/dma-mapping.h and
471 * documented in Documentation/DMA-attributes.txt */
472 ...
a75b0a2f 473
03158a70
MCC
474 unsigned long attr;
475 attr |= DMA_ATTR_FOO;
476 ....
477 n = dma_map_sg_attrs(dev, sg, nents, DMA_TO_DEVICE, attr);
478 ....
a75b0a2f
AK
479
480Architectures that care about DMA_ATTR_FOO would check for its
481presence in their implementations of the mapping and unmapping
03158a70
MCC
482routines, e.g.:::
483
484 void whizco_dma_map_sg_attrs(struct device *dev, dma_addr_t dma_addr,
485 size_t size, enum dma_data_direction dir,
486 unsigned long attrs)
487 {
488 ....
489 if (attrs & DMA_ATTR_FOO)
490 /* twizzle the frobnozzle */
491 ....
492 }
a75b0a2f 493
1da177e4 494
03158a70
MCC
495Part II - Advanced dma usage
496----------------------------
1da177e4 497
f5a69f4c
FT
498Warning: These pieces of the DMA API should not be used in the
499majority of cases, since they cater for unlikely corner cases that
500don't belong in usual drivers.
1da177e4
LT
501
502If you don't understand how cache line coherency works between a
503processor and an I/O device, you should not be using this part of the
504API at all.
505
03158a70
MCC
506::
507
508 void *
2fd523c5
CH
509 dma_alloc_attrs(struct device *dev, size_t size, dma_addr_t *dma_handle,
510 gfp_t flag, unsigned long attrs)
1da177e4 511
2fd523c5
CH
512Identical to dma_alloc_coherent() except that when the
513DMA_ATTR_NON_CONSISTENT flags is passed in the attrs argument, the
514platform will choose to return either consistent or non-consistent memory
515as it sees fit. By using this API, you are guaranteeing to the platform
516that you have all the correct and necessary sync points for this memory
517in the driver should it choose to return non-consistent memory.
1da177e4
LT
518
519Note: where the platform can return consistent memory, it will
520guarantee that the sync points become nops.
521
522Warning: Handling non-consistent memory is a real pain. You should
77f2ea2f 523only use this API if you positively know your driver will be
1da177e4
LT
524required to work on one of the rare (usually non-PCI) architectures
525that simply cannot make consistent memory.
526
03158a70
MCC
527::
528
529 void
2fd523c5
CH
530 dma_free_attrs(struct device *dev, size_t size, void *cpu_addr,
531 dma_addr_t dma_handle, unsigned long attrs)
1da177e4 532
2fd523c5
CH
533Free memory allocated by the dma_alloc_attrs(). All parameters common
534parameters must identical to those otherwise passed to dma_fre_coherent,
535and the attrs argument must be identical to the attrs passed to
536dma_alloc_attrs().
1da177e4 537
03158a70
MCC
538::
539
540 int
541 dma_get_cache_alignment(void)
1da177e4 542
a12e2c6c 543Returns the processor cache alignment. This is the absolute minimum
1da177e4
LT
544alignment *and* width that you must observe when either mapping
545memory or doing partial flushes.
546
03158a70 547.. note::
1da177e4 548
03158a70
MCC
549 This API may return a number *larger* than the actual cache
550 line, but it will guarantee that one or more cache lines fit exactly
551 into the width returned by this call. It will also always be a power
552 of two for easy alignment.
553
554::
555
556 void
557 dma_cache_sync(struct device *dev, void *vaddr, size_t size,
558 enum dma_data_direction direction)
1da177e4 559
2fd523c5
CH
560Do a partial sync of memory that was allocated by dma_alloc_attrs() with
561the DMA_ATTR_NON_CONSISTENT flag starting at virtual address vaddr and
1da177e4
LT
562continuing on for size. Again, you *must* observe the cache line
563boundaries when doing this.
564
03158a70
MCC
565::
566
567 int
568 dma_declare_coherent_memory(struct device *dev, phys_addr_t phys_addr,
569 dma_addr_t device_addr, size_t size, int
570 flags)
1da177e4 571
77f2ea2f 572Declare region of memory to be handed out by dma_alloc_coherent() when
1da177e4
LT
573it's asked for coherent memory for this device.
574
f311a724
BH
575phys_addr is the CPU physical address to which the memory is currently
576assigned (this will be ioremapped so the CPU can access the region).
1da177e4 577
3a9ad0b4 578device_addr is the DMA address the device needs to be programmed
88a984ba 579with to actually address this memory (this will be handed out as the
a12e2c6c 580dma_addr_t in dma_alloc_coherent()).
1da177e4
LT
581
582size is the size of the area (must be multiples of PAGE_SIZE).
583
77f2ea2f 584flags can be ORed together and are:
1da177e4 585
03158a70
MCC
586- DMA_MEMORY_EXCLUSIVE - only allocate memory from the declared regions.
587 Do not allow dma_alloc_coherent() to fall back to system memory when
588 it's out of memory in the declared region.
1da177e4 589
2436bdcd 590As a simplification for the platforms, only *one* such region of
1da177e4
LT
591memory may be declared per device.
592
593For reasons of efficiency, most platforms choose to track the declared
594region only at the granularity of a page. For smaller allocations,
595you should use the dma_pool() API.
596
03158a70
MCC
597::
598
599 void
600 dma_release_declared_memory(struct device *dev)
1da177e4
LT
601
602Remove the memory region previously declared from the system. This
603API performs *no* in-use checking for this region and will return
604unconditionally having removed all the required structures. It is the
a12e2c6c 605driver's job to ensure that no parts of this memory region are
1da177e4
LT
606currently in use.
607
03158a70
MCC
608::
609
610 void *
611 dma_mark_declared_memory_occupied(struct device *dev,
612 dma_addr_t device_addr, size_t size)
1da177e4
LT
613
614This is used to occupy specific regions of the declared space
615(dma_alloc_coherent() will hand out the first free region it finds).
616
a12e2c6c 617device_addr is the *device* address of the region requested.
1da177e4 618
a12e2c6c 619size is the size (and should be a page-sized multiple).
1da177e4
LT
620
621The return value will be either a pointer to the processor virtual
622address of the memory, or an error (via PTR_ERR()) if any part of the
623region is occupied.
187f9c3f
JR
624
625Part III - Debug drivers use of the DMA-API
626-------------------------------------------
627
77f2ea2f 628The DMA-API as described above has some constraints. DMA addresses must be
187f9c3f
JR
629released with the corresponding function with the same size for example. With
630the advent of hardware IOMMUs it becomes more and more important that drivers
631do not violate those constraints. In the worst case such a violation can
632result in data corruption up to destroyed filesystems.
633
634To debug drivers and find bugs in the usage of the DMA-API checking code can
635be compiled into the kernel which will tell the developer about those
636violations. If your architecture supports it you can select the "Enable
637debugging of DMA-API usage" option in your kernel configuration. Enabling this
638option has a performance impact. Do not enable it in production kernels.
639
640If you boot the resulting kernel will contain code which does some bookkeeping
641about what DMA memory was allocated for which device. If this code detects an
642error it prints a warning message with some details into your kernel log. An
03158a70
MCC
643example warning message may look like this::
644
645 WARNING: at /data2/repos/linux-2.6-iommu/lib/dma-debug.c:448
646 check_unmap+0x203/0x490()
647 Hardware name:
648 forcedeth 0000:00:08.0: DMA-API: device driver frees DMA memory with wrong
649 function [device address=0x00000000640444be] [size=66 bytes] [mapped as
650 single] [unmapped as page]
651 Modules linked in: nfsd exportfs bridge stp llc r8169
652 Pid: 0, comm: swapper Tainted: G W 2.6.28-dmatest-09289-g8bb99c0 #1
653 Call Trace:
654 <IRQ> [<ffffffff80240b22>] warn_slowpath+0xf2/0x130
655 [<ffffffff80647b70>] _spin_unlock+0x10/0x30
656 [<ffffffff80537e75>] usb_hcd_link_urb_to_ep+0x75/0xc0
657 [<ffffffff80647c22>] _spin_unlock_irqrestore+0x12/0x40
658 [<ffffffff8055347f>] ohci_urb_enqueue+0x19f/0x7c0
659 [<ffffffff80252f96>] queue_work+0x56/0x60
660 [<ffffffff80237e10>] enqueue_task_fair+0x20/0x50
661 [<ffffffff80539279>] usb_hcd_submit_urb+0x379/0xbc0
662 [<ffffffff803b78c3>] cpumask_next_and+0x23/0x40
663 [<ffffffff80235177>] find_busiest_group+0x207/0x8a0
664 [<ffffffff8064784f>] _spin_lock_irqsave+0x1f/0x50
665 [<ffffffff803c7ea3>] check_unmap+0x203/0x490
666 [<ffffffff803c8259>] debug_dma_unmap_page+0x49/0x50
667 [<ffffffff80485f26>] nv_tx_done_optimized+0xc6/0x2c0
668 [<ffffffff80486c13>] nv_nic_irq_optimized+0x73/0x2b0
669 [<ffffffff8026df84>] handle_IRQ_event+0x34/0x70
670 [<ffffffff8026ffe9>] handle_edge_irq+0xc9/0x150
671 [<ffffffff8020e3ab>] do_IRQ+0xcb/0x1c0
672 [<ffffffff8020c093>] ret_from_intr+0x0/0xa
673 <EOI> <4>---[ end trace f6435a98e2a38c0e ]---
187f9c3f
JR
674
675The driver developer can find the driver and the device including a stacktrace
676of the DMA-API call which caused this warning.
677
678Per default only the first error will result in a warning message. All other
679errors will only silently counted. This limitation exist to prevent the code
680from flooding your kernel log. To support debugging a device driver this can
681be disabled via debugfs. See the debugfs interface documentation below for
682details.
683
684The debugfs directory for the DMA-API debugging code is called dma-api/. In
685this directory the following files can currently be found:
686
03158a70
MCC
687=============================== ===============================================
688dma-api/all_errors This file contains a numeric value. If this
187f9c3f
JR
689 value is not equal to zero the debugging code
690 will print a warning for every error it finds
19f59460
ML
691 into the kernel log. Be careful with this
692 option, as it can easily flood your logs.
187f9c3f 693
03158a70 694dma-api/disabled This read-only file contains the character 'Y'
187f9c3f
JR
695 if the debugging code is disabled. This can
696 happen when it runs out of memory or if it was
697 disabled at boot time
698
03158a70 699dma-api/error_count This file is read-only and shows the total
187f9c3f
JR
700 numbers of errors found.
701
03158a70 702dma-api/num_errors The number in this file shows how many
187f9c3f
JR
703 warnings will be printed to the kernel log
704 before it stops. This number is initialized to
705 one at system boot and be set by writing into
706 this file
707
03158a70 708dma-api/min_free_entries This read-only file can be read to get the
187f9c3f
JR
709 minimum number of free dma_debug_entries the
710 allocator has ever seen. If this value goes
2b9d9ac0
RM
711 down to zero the code will attempt to increase
712 nr_total_entries to compensate.
187f9c3f 713
03158a70 714dma-api/num_free_entries The current number of free dma_debug_entries
187f9c3f
JR
715 in the allocator.
716
9f191555
RM
717dma-api/nr_total_entries The total number of dma_debug_entries in the
718 allocator, both free and used.
719
03158a70 720dma-api/driver-filter You can write a name of a driver into this file
016ea687
JR
721 to limit the debug output to requests from that
722 particular driver. Write an empty string to
723 that file to disable the filter and see
724 all errors again.
03158a70 725=============================== ===============================================
016ea687 726
187f9c3f
JR
727If you have this code compiled into your kernel it will be enabled by default.
728If you want to boot without the bookkeeping anyway you can provide
729'dma_debug=off' as a boot parameter. This will disable DMA-API debugging.
730Notice that you can not enable it again at runtime. You have to reboot to do
731so.
732
016ea687
JR
733If you want to see debug messages only for a special device driver you can
734specify the dma_debug_driver=<drivername> parameter. This will enable the
735driver filter at boot time. The debug code will only print errors for that
736driver afterwards. This filter can be disabled or changed later using debugfs.
737
187f9c3f 738When the code disables itself at runtime this is most likely because it ran
2b9d9ac0
RM
739out of dma_debug_entries and was unable to allocate more on-demand. 65536
740entries are preallocated at boot - if this is too low for you boot with
ad78dee0
RM
741'dma_debug_entries=<your_desired_number>' to overwrite the default. Note
742that the code allocates entries in batches, so the exact number of
743preallocated entries may be greater than the actual number requested. The
ceb51173
RM
744code will print to the kernel log each time it has dynamically allocated
745as many entries as were initially preallocated. This is to indicate that a
746larger preallocation size may be appropriate, or if it happens continually
747that a driver may be leaking mappings.
6c9c6d63 748
03158a70
MCC
749::
750
751 void
752 debug_dma_mapping_error(struct device *dev, dma_addr_t dma_addr);
6c9c6d63
SK
753
754dma-debug interface debug_dma_mapping_error() to debug drivers that fail
77f2ea2f 755to check DMA mapping errors on addresses returned by dma_map_single() and
6c9c6d63
SK
756dma_map_page() interfaces. This interface clears a flag set by
757debug_dma_map_page() to indicate that dma_mapping_error() has been called by
758the driver. When driver does unmap, debug_dma_unmap() checks the flag and if
759this flag is still set, prints warning message that includes call trace that
760leads up to the unmap. This interface can be called from dma_mapping_error()
77f2ea2f 761routines to enable DMA mapping error check debugging.