Root/
Source at commit ec7cab4cbb721bff91ec924ec691efd8daf36579 created 12 years 8 months ago. By Maarten ter Huurne, MIPS: JZ4740: A320: Updated quickstart documentation. | |
---|---|
1 | Dynamic DMA mapping using the generic device |
2 | ============================================ |
3 | |
4 | James E.J. Bottomley <James.Bottomley@HansenPartnership.com> |
5 | |
6 | This document describes the DMA API. For a more gentle introduction |
7 | of the API (and actual examples) see |
8 | Documentation/DMA-API-HOWTO.txt. |
9 | |
10 | This API is split into two pieces. Part I describes the API. Part II |
11 | describes the extensions to the API for supporting non-consistent |
12 | memory machines. Unless you know that your driver absolutely has to |
13 | support non-consistent platforms (this is usually only legacy |
14 | platforms) you should only use the API described in part I. |
15 | |
16 | Part I - dma_ API |
17 | ------------------------------------- |
18 | |
19 | To get the dma_ API, you must #include <linux/dma-mapping.h> |
20 | |
21 | |
22 | Part Ia - Using large dma-coherent buffers |
23 | ------------------------------------------ |
24 | |
25 | void * |
26 | dma_alloc_coherent(struct device *dev, size_t size, |
27 | dma_addr_t *dma_handle, gfp_t flag) |
28 | |
29 | Consistent memory is memory for which a write by either the device or |
30 | the processor can immediately be read by the processor or device |
31 | without having to worry about caching effects. (You may however need |
32 | to make sure to flush the processor's write buffers before telling |
33 | devices to read that memory.) |
34 | |
35 | This routine allocates a region of <size> bytes of consistent memory. |
36 | It also returns a <dma_handle> which may be cast to an unsigned |
37 | integer the same width as the bus and used as the physical address |
38 | base of the region. |
39 | |
40 | Returns: a pointer to the allocated region (in the processor's virtual |
41 | address space) or NULL if the allocation failed. |
42 | |
43 | Note: consistent memory can be expensive on some platforms, and the |
44 | minimum allocation length may be as big as a page, so you should |
45 | consolidate your requests for consistent memory as much as possible. |
46 | The simplest way to do that is to use the dma_pool calls (see below). |
47 | |
48 | The flag parameter (dma_alloc_coherent only) allows the caller to |
49 | specify the GFP_ flags (see kmalloc) for the allocation (the |
50 | implementation may choose to ignore flags that affect the location of |
51 | the returned memory, like GFP_DMA). |
52 | |
53 | void |
54 | dma_free_coherent(struct device *dev, size_t size, void *cpu_addr, |
55 | dma_addr_t dma_handle) |
56 | |
57 | Free the region of consistent memory you previously allocated. dev, |
58 | size and dma_handle must all be the same as those passed into the |
59 | consistent allocate. cpu_addr must be the virtual address returned by |
60 | the consistent allocate. |
61 | |
62 | Note that unlike their sibling allocation calls, these routines |
63 | may only be called with IRQs enabled. |
64 | |
65 | |
66 | Part Ib - Using small dma-coherent buffers |
67 | ------------------------------------------ |
68 | |
69 | To get this part of the dma_ API, you must #include <linux/dmapool.h> |
70 | |
71 | Many drivers need lots of small dma-coherent memory regions for DMA |
72 | descriptors or I/O buffers. Rather than allocating in units of a page |
73 | or more using dma_alloc_coherent(), you can use DMA pools. These work |
74 | much like a struct kmem_cache, except that they use the dma-coherent allocator, |
75 | not __get_free_pages(). Also, they understand common hardware constraints |
76 | for alignment, like queue heads needing to be aligned on N-byte boundaries. |
77 | |
78 | |
79 | struct dma_pool * |
80 | dma_pool_create(const char *name, struct device *dev, |
81 | size_t size, size_t align, size_t alloc); |
82 | |
83 | The pool create() routines initialize a pool of dma-coherent buffers |
84 | for use with a given device. It must be called in a context which |
85 | can sleep. |
86 | |
87 | The "name" is for diagnostics (like a struct kmem_cache name); dev and size |
88 | are like what you'd pass to dma_alloc_coherent(). The device's hardware |
89 | alignment requirement for this type of data is "align" (which is expressed |
90 | in bytes, and must be a power of two). If your device has no boundary |
91 | crossing restrictions, pass 0 for alloc; passing 4096 says memory allocated |
92 | from this pool must not cross 4KByte boundaries. |
93 | |
94 | |
95 | void *dma_pool_alloc(struct dma_pool *pool, gfp_t gfp_flags, |
96 | dma_addr_t *dma_handle); |
97 | |
98 | This allocates memory from the pool; the returned memory will meet the size |
99 | and alignment requirements specified at creation time. Pass GFP_ATOMIC to |
100 | prevent blocking, or if it's permitted (not in_interrupt, not holding SMP locks), |
101 | pass GFP_KERNEL to allow blocking. Like dma_alloc_coherent(), this returns |
102 | two values: an address usable by the cpu, and the dma address usable by the |
103 | pool's device. |
104 | |
105 | |
106 | void dma_pool_free(struct dma_pool *pool, void *vaddr, |
107 | dma_addr_t addr); |
108 | |
109 | This puts memory back into the pool. The pool is what was passed to |
110 | the pool allocation routine; the cpu (vaddr) and dma addresses are what |
111 | were returned when that routine allocated the memory being freed. |
112 | |
113 | |
114 | void dma_pool_destroy(struct dma_pool *pool); |
115 | |
116 | The pool destroy() routines free the resources of the pool. They must be |
117 | called in a context which can sleep. Make sure you've freed all allocated |
118 | memory back to the pool before you destroy it. |
119 | |
120 | |
121 | Part Ic - DMA addressing limitations |
122 | ------------------------------------ |
123 | |
124 | int |
125 | dma_supported(struct device *dev, u64 mask) |
126 | |
127 | Checks to see if the device can support DMA to the memory described by |
128 | mask. |
129 | |
130 | Returns: 1 if it can and 0 if it can't. |
131 | |
132 | Notes: This routine merely tests to see if the mask is possible. It |
133 | won't change the current mask settings. It is more intended as an |
134 | internal API for use by the platform than an external API for use by |
135 | driver writers. |
136 | |
137 | int |
138 | dma_set_mask(struct device *dev, u64 mask) |
139 | |
140 | Checks to see if the mask is possible and updates the device |
141 | parameters if it is. |
142 | |
143 | Returns: 0 if successful and a negative error if not. |
144 | |
145 | int |
146 | dma_set_coherent_mask(struct device *dev, u64 mask) |
147 | |
148 | Checks to see if the mask is possible and updates the device |
149 | parameters if it is. |
150 | |
151 | Returns: 0 if successful and a negative error if not. |
152 | |
153 | u64 |
154 | dma_get_required_mask(struct device *dev) |
155 | |
156 | This API returns the mask that the platform requires to |
157 | operate efficiently. Usually this means the returned mask |
158 | is the minimum required to cover all of memory. Examining the |
159 | required mask gives drivers with variable descriptor sizes the |
160 | opportunity to use smaller descriptors as necessary. |
161 | |
162 | Requesting the required mask does not alter the current mask. If you |
163 | wish to take advantage of it, you should issue a dma_set_mask() |
164 | call to set the mask to the value returned. |
165 | |
166 | |
167 | Part Id - Streaming DMA mappings |
168 | -------------------------------- |
169 | |
170 | dma_addr_t |
171 | dma_map_single(struct device *dev, void *cpu_addr, size_t size, |
172 | enum dma_data_direction direction) |
173 | |
174 | Maps a piece of processor virtual memory so it can be accessed by the |
175 | device and returns the physical handle of the memory. |
176 | |
177 | The direction for both api's may be converted freely by casting. |
178 | However the dma_ API uses a strongly typed enumerator for its |
179 | direction: |
180 | |
181 | DMA_NONE no direction (used for debugging) |
182 | DMA_TO_DEVICE data is going from the memory to the device |
183 | DMA_FROM_DEVICE data is coming from the device to the memory |
184 | DMA_BIDIRECTIONAL direction isn't known |
185 | |
186 | Notes: Not all memory regions in a machine can be mapped by this |
187 | API. Further, regions that appear to be physically contiguous in |
188 | kernel virtual space may not be contiguous as physical memory. Since |
189 | this API does not provide any scatter/gather capability, it will fail |
190 | if the user tries to map a non-physically contiguous piece of memory. |
191 | For this reason, it is recommended that memory mapped by this API be |
192 | obtained only from sources which guarantee it to be physically contiguous |
193 | (like kmalloc). |
194 | |
195 | Further, the physical address of the memory must be within the |
196 | dma_mask of the device (the dma_mask represents a bit mask of the |
197 | addressable region for the device. I.e., if the physical address of |
198 | the memory anded with the dma_mask is still equal to the physical |
199 | address, then the device can perform DMA to the memory). In order to |
200 | ensure that the memory allocated by kmalloc is within the dma_mask, |
201 | the driver may specify various platform-dependent flags to restrict |
202 | the physical memory range of the allocation (e.g. on x86, GFP_DMA |
203 | guarantees to be within the first 16Mb of available physical memory, |
204 | as required by ISA devices). |
205 | |
206 | Note also that the above constraints on physical contiguity and |
207 | dma_mask may not apply if the platform has an IOMMU (a device which |
208 | supplies a physical to virtual mapping between the I/O memory bus and |
209 | the device). However, to be portable, device driver writers may *not* |
210 | assume that such an IOMMU exists. |
211 | |
212 | Warnings: Memory coherency operates at a granularity called the cache |
213 | line width. In order for memory mapped by this API to operate |
214 | correctly, the mapped region must begin exactly on a cache line |
215 | boundary and end exactly on one (to prevent two separately mapped |
216 | regions from sharing a single cache line). Since the cache line size |
217 | may not be known at compile time, the API will not enforce this |
218 | requirement. Therefore, it is recommended that driver writers who |
219 | don't take special care to determine the cache line size at run time |
220 | only map virtual regions that begin and end on page boundaries (which |
221 | are guaranteed also to be cache line boundaries). |
222 | |
223 | DMA_TO_DEVICE synchronisation must be done after the last modification |
224 | of the memory region by the software and before it is handed off to |
225 | the driver. Once this primitive is used, memory covered by this |
226 | primitive should be treated as read-only by the device. If the device |
227 | may write to it at any point, it should be DMA_BIDIRECTIONAL (see |
228 | below). |
229 | |
230 | DMA_FROM_DEVICE synchronisation must be done before the driver |
231 | accesses data that may be changed by the device. This memory should |
232 | be treated as read-only by the driver. If the driver needs to write |
233 | to it at any point, it should be DMA_BIDIRECTIONAL (see below). |
234 | |
235 | DMA_BIDIRECTIONAL requires special handling: it means that the driver |
236 | isn't sure if the memory was modified before being handed off to the |
237 | device and also isn't sure if the device will also modify it. Thus, |
238 | you must always sync bidirectional memory twice: once before the |
239 | memory is handed off to the device (to make sure all memory changes |
240 | are flushed from the processor) and once before the data may be |
241 | accessed after being used by the device (to make sure any processor |
242 | cache lines are updated with data that the device may have changed). |
243 | |
244 | void |
245 | dma_unmap_single(struct device *dev, dma_addr_t dma_addr, size_t size, |
246 | enum dma_data_direction direction) |
247 | |
248 | Unmaps the region previously mapped. All the parameters passed in |
249 | must be identical to those passed in (and returned) by the mapping |
250 | API. |
251 | |
252 | dma_addr_t |
253 | dma_map_page(struct device *dev, struct page *page, |
254 | unsigned long offset, size_t size, |
255 | enum dma_data_direction direction) |
256 | void |
257 | dma_unmap_page(struct device *dev, dma_addr_t dma_address, size_t size, |
258 | enum dma_data_direction direction) |
259 | |
260 | API for mapping and unmapping for pages. All the notes and warnings |
261 | for the other mapping APIs apply here. Also, although the <offset> |
262 | and <size> parameters are provided to do partial page mapping, it is |
263 | recommended that you never use these unless you really know what the |
264 | cache width is. |
265 | |
266 | int |
267 | dma_mapping_error(struct device *dev, dma_addr_t dma_addr) |
268 | |
269 | In some circumstances dma_map_single and dma_map_page will fail to create |
270 | a mapping. A driver can check for these errors by testing the returned |
271 | dma address with dma_mapping_error(). A non-zero return value means the mapping |
272 | could not be created and the driver should take appropriate action (e.g. |
273 | reduce current DMA mapping usage or delay and try again later). |
274 | |
275 | int |
276 | dma_map_sg(struct device *dev, struct scatterlist *sg, |
277 | int nents, enum dma_data_direction direction) |
278 | |
279 | Returns: the number of physical segments mapped (this may be shorter |
280 | than <nents> passed in if some elements of the scatter/gather list are |
281 | physically or virtually adjacent and an IOMMU maps them with a single |
282 | entry). |
283 | |
284 | Please note that the sg cannot be mapped again if it has been mapped once. |
285 | The mapping process is allowed to destroy information in the sg. |
286 | |
287 | As with the other mapping interfaces, dma_map_sg can fail. When it |
288 | does, 0 is returned and a driver must take appropriate action. It is |
289 | critical that the driver do something, in the case of a block driver |
290 | aborting the request or even oopsing is better than doing nothing and |
291 | corrupting the filesystem. |
292 | |
293 | With scatterlists, you use the resulting mapping like this: |
294 | |
295 | int i, count = dma_map_sg(dev, sglist, nents, direction); |
296 | struct scatterlist *sg; |
297 | |
298 | for_each_sg(sglist, sg, count, i) { |
299 | hw_address[i] = sg_dma_address(sg); |
300 | hw_len[i] = sg_dma_len(sg); |
301 | } |
302 | |
303 | where nents is the number of entries in the sglist. |
304 | |
305 | The implementation is free to merge several consecutive sglist entries |
306 | into one (e.g. with an IOMMU, or if several pages just happen to be |
307 | physically contiguous) and returns the actual number of sg entries it |
308 | mapped them to. On failure 0, is returned. |
309 | |
310 | Then you should loop count times (note: this can be less than nents times) |
311 | and use sg_dma_address() and sg_dma_len() macros where you previously |
312 | accessed sg->address and sg->length as shown above. |
313 | |
314 | void |
315 | dma_unmap_sg(struct device *dev, struct scatterlist *sg, |
316 | int nhwentries, enum dma_data_direction direction) |
317 | |
318 | Unmap the previously mapped scatter/gather list. All the parameters |
319 | must be the same as those and passed in to the scatter/gather mapping |
320 | API. |
321 | |
322 | Note: <nents> must be the number you passed in, *not* the number of |
323 | physical entries returned. |
324 | |
325 | void |
326 | dma_sync_single_for_cpu(struct device *dev, dma_addr_t dma_handle, size_t size, |
327 | enum dma_data_direction direction) |
328 | void |
329 | dma_sync_single_for_device(struct device *dev, dma_addr_t dma_handle, size_t size, |
330 | enum dma_data_direction direction) |
331 | void |
332 | dma_sync_sg_for_cpu(struct device *dev, struct scatterlist *sg, int nelems, |
333 | enum dma_data_direction direction) |
334 | void |
335 | dma_sync_sg_for_device(struct device *dev, struct scatterlist *sg, int nelems, |
336 | enum dma_data_direction direction) |
337 | |
338 | Synchronise a single contiguous or scatter/gather mapping for the cpu |
339 | and device. With the sync_sg API, all the parameters must be the same |
340 | as those passed into the single mapping API. With the sync_single API, |
341 | you can use dma_handle and size parameters that aren't identical to |
342 | those passed into the single mapping API to do a partial sync. |
343 | |
344 | Notes: You must do this: |
345 | |
346 | - Before reading values that have been written by DMA from the device |
347 | (use the DMA_FROM_DEVICE direction) |
348 | - After writing values that will be written to the device using DMA |
349 | (use the DMA_TO_DEVICE) direction |
350 | - before *and* after handing memory to the device if the memory is |
351 | DMA_BIDIRECTIONAL |
352 | |
353 | See also dma_map_single(). |
354 | |
355 | dma_addr_t |
356 | dma_map_single_attrs(struct device *dev, void *cpu_addr, size_t size, |
357 | enum dma_data_direction dir, |
358 | struct dma_attrs *attrs) |
359 | |
360 | void |
361 | dma_unmap_single_attrs(struct device *dev, dma_addr_t dma_addr, |
362 | size_t size, enum dma_data_direction dir, |
363 | struct dma_attrs *attrs) |
364 | |
365 | int |
366 | dma_map_sg_attrs(struct device *dev, struct scatterlist *sgl, |
367 | int nents, enum dma_data_direction dir, |
368 | struct dma_attrs *attrs) |
369 | |
370 | void |
371 | dma_unmap_sg_attrs(struct device *dev, struct scatterlist *sgl, |
372 | int nents, enum dma_data_direction dir, |
373 | struct dma_attrs *attrs) |
374 | |
375 | The four functions above are just like the counterpart functions |
376 | without the _attrs suffixes, except that they pass an optional |
377 | struct dma_attrs*. |
378 | |
379 | struct dma_attrs encapsulates a set of "dma attributes". For the |
380 | definition of struct dma_attrs see linux/dma-attrs.h. |
381 | |
382 | The interpretation of dma attributes is architecture-specific, and |
383 | each attribute should be documented in Documentation/DMA-attributes.txt. |
384 | |
385 | If struct dma_attrs* is NULL, the semantics of each of these |
386 | functions is identical to those of the corresponding function |
387 | without the _attrs suffix. As a result dma_map_single_attrs() |
388 | can generally replace dma_map_single(), etc. |
389 | |
390 | As an example of the use of the *_attrs functions, here's how |
391 | you could pass an attribute DMA_ATTR_FOO when mapping memory |
392 | for DMA: |
393 | |
394 | #include <linux/dma-attrs.h> |
395 | /* DMA_ATTR_FOO should be defined in linux/dma-attrs.h and |
396 | * documented in Documentation/DMA-attributes.txt */ |
397 | ... |
398 | |
399 | DEFINE_DMA_ATTRS(attrs); |
400 | dma_set_attr(DMA_ATTR_FOO, &attrs); |
401 | .... |
402 | n = dma_map_sg_attrs(dev, sg, nents, DMA_TO_DEVICE, &attr); |
403 | .... |
404 | |
405 | Architectures that care about DMA_ATTR_FOO would check for its |
406 | presence in their implementations of the mapping and unmapping |
407 | routines, e.g.: |
408 | |
409 | void whizco_dma_map_sg_attrs(struct device *dev, dma_addr_t dma_addr, |
410 | size_t size, enum dma_data_direction dir, |
411 | struct dma_attrs *attrs) |
412 | { |
413 | .... |
414 | int foo = dma_get_attr(DMA_ATTR_FOO, attrs); |
415 | .... |
416 | if (foo) |
417 | /* twizzle the frobnozzle */ |
418 | .... |
419 | |
420 | |
421 | Part II - Advanced dma_ usage |
422 | ----------------------------- |
423 | |
424 | Warning: These pieces of the DMA API should not be used in the |
425 | majority of cases, since they cater for unlikely corner cases that |
426 | don't belong in usual drivers. |
427 | |
428 | If you don't understand how cache line coherency works between a |
429 | processor and an I/O device, you should not be using this part of the |
430 | API at all. |
431 | |
432 | void * |
433 | dma_alloc_noncoherent(struct device *dev, size_t size, |
434 | dma_addr_t *dma_handle, gfp_t flag) |
435 | |
436 | Identical to dma_alloc_coherent() except that the platform will |
437 | choose to return either consistent or non-consistent memory as it sees |
438 | fit. By using this API, you are guaranteeing to the platform that you |
439 | have all the correct and necessary sync points for this memory in the |
440 | driver should it choose to return non-consistent memory. |
441 | |
442 | Note: where the platform can return consistent memory, it will |
443 | guarantee that the sync points become nops. |
444 | |
445 | Warning: Handling non-consistent memory is a real pain. You should |
446 | only ever use this API if you positively know your driver will be |
447 | required to work on one of the rare (usually non-PCI) architectures |
448 | that simply cannot make consistent memory. |
449 | |
450 | void |
451 | dma_free_noncoherent(struct device *dev, size_t size, void *cpu_addr, |
452 | dma_addr_t dma_handle) |
453 | |
454 | Free memory allocated by the nonconsistent API. All parameters must |
455 | be identical to those passed in (and returned by |
456 | dma_alloc_noncoherent()). |
457 | |
458 | int |
459 | dma_get_cache_alignment(void) |
460 | |
461 | Returns the processor cache alignment. This is the absolute minimum |
462 | alignment *and* width that you must observe when either mapping |
463 | memory or doing partial flushes. |
464 | |
465 | Notes: This API may return a number *larger* than the actual cache |
466 | line, but it will guarantee that one or more cache lines fit exactly |
467 | into the width returned by this call. It will also always be a power |
468 | of two for easy alignment. |
469 | |
470 | void |
471 | dma_cache_sync(struct device *dev, void *vaddr, size_t size, |
472 | enum dma_data_direction direction) |
473 | |
474 | Do a partial sync of memory that was allocated by |
475 | dma_alloc_noncoherent(), starting at virtual address vaddr and |
476 | continuing on for size. Again, you *must* observe the cache line |
477 | boundaries when doing this. |
478 | |
479 | int |
480 | dma_declare_coherent_memory(struct device *dev, dma_addr_t bus_addr, |
481 | dma_addr_t device_addr, size_t size, int |
482 | flags) |
483 | |
484 | Declare region of memory to be handed out by dma_alloc_coherent when |
485 | it's asked for coherent memory for this device. |
486 | |
487 | bus_addr is the physical address to which the memory is currently |
488 | assigned in the bus responding region (this will be used by the |
489 | platform to perform the mapping). |
490 | |
491 | device_addr is the physical address the device needs to be programmed |
492 | with actually to address this memory (this will be handed out as the |
493 | dma_addr_t in dma_alloc_coherent()). |
494 | |
495 | size is the size of the area (must be multiples of PAGE_SIZE). |
496 | |
497 | flags can be or'd together and are: |
498 | |
499 | DMA_MEMORY_MAP - request that the memory returned from |
500 | dma_alloc_coherent() be directly writable. |
501 | |
502 | DMA_MEMORY_IO - request that the memory returned from |
503 | dma_alloc_coherent() be addressable using read/write/memcpy_toio etc. |
504 | |
505 | One or both of these flags must be present. |
506 | |
507 | DMA_MEMORY_INCLUDES_CHILDREN - make the declared memory be allocated by |
508 | dma_alloc_coherent of any child devices of this one (for memory residing |
509 | on a bridge). |
510 | |
511 | DMA_MEMORY_EXCLUSIVE - only allocate memory from the declared regions. |
512 | Do not allow dma_alloc_coherent() to fall back to system memory when |
513 | it's out of memory in the declared region. |
514 | |
515 | The return value will be either DMA_MEMORY_MAP or DMA_MEMORY_IO and |
516 | must correspond to a passed in flag (i.e. no returning DMA_MEMORY_IO |
517 | if only DMA_MEMORY_MAP were passed in) for success or zero for |
518 | failure. |
519 | |
520 | Note, for DMA_MEMORY_IO returns, all subsequent memory returned by |
521 | dma_alloc_coherent() may no longer be accessed directly, but instead |
522 | must be accessed using the correct bus functions. If your driver |
523 | isn't prepared to handle this contingency, it should not specify |
524 | DMA_MEMORY_IO in the input flags. |
525 | |
526 | As a simplification for the platforms, only *one* such region of |
527 | memory may be declared per device. |
528 | |
529 | For reasons of efficiency, most platforms choose to track the declared |
530 | region only at the granularity of a page. For smaller allocations, |
531 | you should use the dma_pool() API. |
532 | |
533 | void |
534 | dma_release_declared_memory(struct device *dev) |
535 | |
536 | Remove the memory region previously declared from the system. This |
537 | API performs *no* in-use checking for this region and will return |
538 | unconditionally having removed all the required structures. It is the |
539 | driver's job to ensure that no parts of this memory region are |
540 | currently in use. |
541 | |
542 | void * |
543 | dma_mark_declared_memory_occupied(struct device *dev, |
544 | dma_addr_t device_addr, size_t size) |
545 | |
546 | This is used to occupy specific regions of the declared space |
547 | (dma_alloc_coherent() will hand out the first free region it finds). |
548 | |
549 | device_addr is the *device* address of the region requested. |
550 | |
551 | size is the size (and should be a page-sized multiple). |
552 | |
553 | The return value will be either a pointer to the processor virtual |
554 | address of the memory, or an error (via PTR_ERR()) if any part of the |
555 | region is occupied. |
556 | |
557 | Part III - Debug drivers use of the DMA-API |
558 | ------------------------------------------- |
559 | |
560 | The DMA-API as described above as some constraints. DMA addresses must be |
561 | released with the corresponding function with the same size for example. With |
562 | the advent of hardware IOMMUs it becomes more and more important that drivers |
563 | do not violate those constraints. In the worst case such a violation can |
564 | result in data corruption up to destroyed filesystems. |
565 | |
566 | To debug drivers and find bugs in the usage of the DMA-API checking code can |
567 | be compiled into the kernel which will tell the developer about those |
568 | violations. If your architecture supports it you can select the "Enable |
569 | debugging of DMA-API usage" option in your kernel configuration. Enabling this |
570 | option has a performance impact. Do not enable it in production kernels. |
571 | |
572 | If you boot the resulting kernel will contain code which does some bookkeeping |
573 | about what DMA memory was allocated for which device. If this code detects an |
574 | error it prints a warning message with some details into your kernel log. An |
575 | example warning message may look like this: |
576 | |
577 | ------------[ cut here ]------------ |
578 | WARNING: at /data2/repos/linux-2.6-iommu/lib/dma-debug.c:448 |
579 | check_unmap+0x203/0x490() |
580 | Hardware name: |
581 | forcedeth 0000:00:08.0: DMA-API: device driver frees DMA memory with wrong |
582 | function [device address=0x00000000640444be] [size=66 bytes] [mapped as |
583 | single] [unmapped as page] |
584 | Modules linked in: nfsd exportfs bridge stp llc r8169 |
585 | Pid: 0, comm: swapper Tainted: G W 2.6.28-dmatest-09289-g8bb99c0 #1 |
586 | Call Trace: |
587 | <IRQ> [<ffffffff80240b22>] warn_slowpath+0xf2/0x130 |
588 | [<ffffffff80647b70>] _spin_unlock+0x10/0x30 |
589 | [<ffffffff80537e75>] usb_hcd_link_urb_to_ep+0x75/0xc0 |
590 | [<ffffffff80647c22>] _spin_unlock_irqrestore+0x12/0x40 |
591 | [<ffffffff8055347f>] ohci_urb_enqueue+0x19f/0x7c0 |
592 | [<ffffffff80252f96>] queue_work+0x56/0x60 |
593 | [<ffffffff80237e10>] enqueue_task_fair+0x20/0x50 |
594 | [<ffffffff80539279>] usb_hcd_submit_urb+0x379/0xbc0 |
595 | [<ffffffff803b78c3>] cpumask_next_and+0x23/0x40 |
596 | [<ffffffff80235177>] find_busiest_group+0x207/0x8a0 |
597 | [<ffffffff8064784f>] _spin_lock_irqsave+0x1f/0x50 |
598 | [<ffffffff803c7ea3>] check_unmap+0x203/0x490 |
599 | [<ffffffff803c8259>] debug_dma_unmap_page+0x49/0x50 |
600 | [<ffffffff80485f26>] nv_tx_done_optimized+0xc6/0x2c0 |
601 | [<ffffffff80486c13>] nv_nic_irq_optimized+0x73/0x2b0 |
602 | [<ffffffff8026df84>] handle_IRQ_event+0x34/0x70 |
603 | [<ffffffff8026ffe9>] handle_edge_irq+0xc9/0x150 |
604 | [<ffffffff8020e3ab>] do_IRQ+0xcb/0x1c0 |
605 | [<ffffffff8020c093>] ret_from_intr+0x0/0xa |
606 | <EOI> <4>---[ end trace f6435a98e2a38c0e ]--- |
607 | |
608 | The driver developer can find the driver and the device including a stacktrace |
609 | of the DMA-API call which caused this warning. |
610 | |
611 | Per default only the first error will result in a warning message. All other |
612 | errors will only silently counted. This limitation exist to prevent the code |
613 | from flooding your kernel log. To support debugging a device driver this can |
614 | be disabled via debugfs. See the debugfs interface documentation below for |
615 | details. |
616 | |
617 | The debugfs directory for the DMA-API debugging code is called dma-api/. In |
618 | this directory the following files can currently be found: |
619 | |
620 | dma-api/all_errors This file contains a numeric value. If this |
621 | value is not equal to zero the debugging code |
622 | will print a warning for every error it finds |
623 | into the kernel log. Be careful with this |
624 | option, as it can easily flood your logs. |
625 | |
626 | dma-api/disabled This read-only file contains the character 'Y' |
627 | if the debugging code is disabled. This can |
628 | happen when it runs out of memory or if it was |
629 | disabled at boot time |
630 | |
631 | dma-api/error_count This file is read-only and shows the total |
632 | numbers of errors found. |
633 | |
634 | dma-api/num_errors The number in this file shows how many |
635 | warnings will be printed to the kernel log |
636 | before it stops. This number is initialized to |
637 | one at system boot and be set by writing into |
638 | this file |
639 | |
640 | dma-api/min_free_entries |
641 | This read-only file can be read to get the |
642 | minimum number of free dma_debug_entries the |
643 | allocator has ever seen. If this value goes |
644 | down to zero the code will disable itself |
645 | because it is not longer reliable. |
646 | |
647 | dma-api/num_free_entries |
648 | The current number of free dma_debug_entries |
649 | in the allocator. |
650 | |
651 | dma-api/driver-filter |
652 | You can write a name of a driver into this file |
653 | to limit the debug output to requests from that |
654 | particular driver. Write an empty string to |
655 | that file to disable the filter and see |
656 | all errors again. |
657 | |
658 | If you have this code compiled into your kernel it will be enabled by default. |
659 | If you want to boot without the bookkeeping anyway you can provide |
660 | 'dma_debug=off' as a boot parameter. This will disable DMA-API debugging. |
661 | Notice that you can not enable it again at runtime. You have to reboot to do |
662 | so. |
663 | |
664 | If you want to see debug messages only for a special device driver you can |
665 | specify the dma_debug_driver=<drivername> parameter. This will enable the |
666 | driver filter at boot time. The debug code will only print errors for that |
667 | driver afterwards. This filter can be disabled or changed later using debugfs. |
668 | |
669 | When the code disables itself at runtime this is most likely because it ran |
670 | out of dma_debug_entries. These entries are preallocated at boot. The number |
671 | of preallocated entries is defined per architecture. If it is too low for you |
672 | boot with 'dma_debug_entries=<your_desired_number>' to overwrite the |
673 | architectural default. |
674 |
Branches:
ben-wpan
ben-wpan-stefan
javiroman/ks7010
jz-2.6.34
jz-2.6.34-rc5
jz-2.6.34-rc6
jz-2.6.34-rc7
jz-2.6.35
jz-2.6.36
jz-2.6.37
jz-2.6.38
jz-2.6.39
jz-3.0
jz-3.1
jz-3.11
jz-3.12
jz-3.13
jz-3.15
jz-3.16
jz-3.18-dt
jz-3.2
jz-3.3
jz-3.4
jz-3.5
jz-3.6
jz-3.6-rc2-pwm
jz-3.9
jz-3.9-clk
jz-3.9-rc8
jz47xx
jz47xx-2.6.38
master
Tags:
od-2011-09-04
od-2011-09-18
v2.6.34-rc5
v2.6.34-rc6
v2.6.34-rc7
v3.9