About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog

Documentation / DMA-API.txt

Custom Search

Based on kernel version 4.3. Page generated on 2015-11-02 12:48 EST.

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

Information is copyright its respective author. All material is available from the Linux Kernel Source distributed under a GPL License. This page is provided as a free service by mjmwired.net.