2 .\" Copyright (c) 2002, 2003, 2004 The DragonFly Project. All rights reserved.
4 .\" This code is derived from software contributed to The DragonFly Project
5 .\" by Hiten Pandya <hmp@backplane.com>.
7 .\" Redistribution and use in source and binary forms, with or without
8 .\" modification, are permitted provided that the following conditions
11 .\" 1. Redistributions of source code must retain the above copyright
12 .\" notice, this list of conditions and the following disclaimer.
13 .\" 2. Redistributions in binary form must reproduce the above copyright
14 .\" notice, this list of conditions and the following disclaimer in
15 .\" the documentation and/or other materials provided with the
17 .\" 3. Neither the name of The DragonFly Project nor the names of its
18 .\" contributors may be used to endorse or promote products derived
19 .\" from this software without specific, prior written permission.
21 .\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 .\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 .\" LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 .\" FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 .\" COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 .\" INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 .\" BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 .\" LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 .\" AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 .\" OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 .\" OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 .\" Copyright (c) 1996, 1997, 1998, 2001 The NetBSD Foundation, Inc.
35 .\" All rights reserved.
37 .\" This code is derived from software contributed to The NetBSD Foundation
38 .\" by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
39 .\" NASA Ames Research Center.
41 .\" Redistribution and use in source and binary forms, with or without
42 .\" modification, are permitted provided that the following conditions
44 .\" 1. Redistributions of source code must retain the above copyright
45 .\" notice, this list of conditions and the following disclaimer.
46 .\" 2. Redistributions in binary form must reproduce the above copyright
47 .\" notice, this list of conditions and the following disclaimer in the
48 .\" documentation and/or other materials provided with the distribution.
49 .\" 3. All advertising materials mentioning features or use of this software
50 .\" must display the following acknowledgment:
51 .\" This product includes software developed by the NetBSD
52 .\" Foundation, Inc. and its contributors.
53 .\" 4. Neither the name of The NetBSD Foundation nor the names of its
54 .\" contributors may be used to endorse or promote products derived
55 .\" from this software without specific prior written permission.
57 .\" THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
58 .\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
59 .\" TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
60 .\" PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
61 .\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
62 .\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
63 .\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
64 .\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
65 .\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
66 .\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
67 .\" POSSIBILITY OF SUCH DAMAGE.
69 .\" $FreeBSD: /repoman/r/ncvs/src/share/man/man9/bus_dma.9,v 1.7 2003/07/27 14:05:29 mux Exp $
70 .\" $NetBSD: bus_dma.9,v 1.25 2002/10/14 13:43:16 wiz Exp $
71 .\" $DragonFly: src/share/man/man9/bus_dma.9,v 1.7 2006/03/25 03:05:03 swildner Exp $
78 .Nm bus_dma_tag_create ,
79 .Nm bus_dma_tag_destroy ,
80 .Nm bus_dmamap_create ,
81 .Nm bus_dmamap_destroy ,
83 .Nm bus_dmamap_load_mbuf ,
84 .Nm bus_dmamap_load_uio ,
85 .Nm bus_dmamap_unload ,
87 .Nm bus_dmamem_alloc ,
89 .Nd Bus and Machine Independent DMA Mapping Interface
93 .Fn bus_dma_tag_create "bus_dma_tag_t parent" "bus_size_t alignment" \
94 "bus_size_t boundary" "bus_addr_t lowaddr" "bus_addr_t highaddr" \
95 "bus_dma_filter_t *filtfunc" "void *filtfuncarg" "bus_size_t maxsize" \
96 "int nsegments" "bus_size_t maxsegsz" "int flags" "bus_dma_tag_t *dmat"
98 .Fn bus_dma_tag_destroy "bus_dma_tag_t dmat"
100 .Fn bus_dmamap_create "bus_dma_tag_t dmat" "int flags" "bus_dmamap_t *mapp"
102 .Fn bus_dmamap_destroy "bus_dma_tag_t dmat" "bus_dmamap_t map"
104 .Fn bus_dmamap_load "bus_dma_tag_t dmat" "bus_dmamap_t map" "void *buf" \
105 "bus_size_t buflen" "bus_dmamap_callback_t *callback" "void *callback_arg" \
108 .Fn bus_dmamap_load_mbuf "bus_dma_tag_t dmat" "bus_dmamap_t map" \
109 "struct mbuf *mbuf" "bus_dmamap_callback2_t *callback" "void *callback_arg" \
112 .Fn bus_dmamap_load_uio "bus_dma_tag_t dmat" "bus_dmamap_t map" \
113 "struct uio *uio" "bus_dmamap_callback2_t *callback" "void *callback_arg" \
116 .Fn bus_dmamem_alloc "bus_dma_tag_t dmat" "void **vaddr" \
117 "int flags" "bus_dmamap_t *mapp"
119 .Fn bus_dmamap_unload "bus_dma_tag_t dmat" "bus_dmamap_t map"
121 .Fn bus_dmamap_sync "bus_dma_tag_t dmat" "bus_dmamap_t map" \
124 .Fn bus_dmamem_free "bus_dma_tag_t dmat" "void *vaddr" \
127 Direct Memory Access (DMA) is a method of transferring data
128 without involving the CPU, thus providing higher performance.
129 A DMA transaction can be achieved between device to memory,
130 device to device, or memory to memory.
134 API is a bus, device, and machine-independent (MI) interface to
136 It provides the client with flexibility and simplicity by
137 abstracting machine dependent issues like setting up
138 DMA mappings, handling cache issues, bus specific features
140 .Sh STRUCTURES AND TYPES
141 .Bl -tag -width compact
143 A machine-dependent (MD) opaque type that describes the
144 characteristics of DMA transactions.
145 DMA tags are organized into a hierarchy, with each child
146 tag inheriting the restrictions of its parent.
147 This allows all devices along the path of DMA transactions
148 to contribute to the constraints of those transactions.
149 .It Vt bus_dma_filter_t
150 Client specified address filter having the format:
151 .Bl -tag -width compact
153 .Fn "client_filter" "void *filtarg" "bus_addr_t testaddr"
156 Address filters can be specified during tag creation to allow
157 for devices who's DMA address restrictions cannot be specified
161 is client specified during tag creation to be passed to all
162 invocations of the callback.
165 argument contains a potential starting address of a DMA mapping.
166 The filter function operates on the set of addresses from
169 .Ql trunc_page(testaddr) + PAGE_SIZE - 1 ,
171 The filter function should return zero for any mapping in this range
172 that can be accommodated by the device and non-zero otherwise.
173 .It Vt bus_dma_segment_t
174 A machine-dependent type that describes individual
183 field contains the device visible address of the DMA segment, and
185 contains the length of the DMA segment.
186 Although the DMA segments returned by a mapping call will adhere to
187 all restrictions necessary for a successful DMA operation, some conversion
188 (e.g. a conversion from host byte order to the device's byte order) is
189 almost always required when presenting segment information to the device.
191 A machine-dependent opaque type describing an individual mapping.
192 Multiple DMA maps can be associated with one DMA tag.
193 .It Vt bus_dmamap_callback_t
194 Client specified callback for receiving mapping information resulting from
198 .Fn bus_dmamap_load .
199 Callbacks are of the format:
200 .Bl -tag -width compact
202 .Fn "client_callback" "void *callback_arg" "bus_dma_segment_t *segs" \
203 "int nseg" "int error"
208 is the callback argument passed to dmamap load functions.
213 parameters describe an array of
214 .Vt bus_dma_segment_t
215 structures that represent the mapping.
216 This array is only valid within the scope of the callback function.
217 The success or failure of the mapping is indicated by the
220 More information on the use of callbacks can be found in the
221 description of the individual dmamap load functions.
222 .It Vt bus_dmamap_callback2_t
223 Client specified callback for receiving mapping information resulting from
227 .Fn bus_dmamap_load_uio
229 .Fn bus_dmamap_load_mbuf .
231 Callback2s are of the format:
232 .Bl -tag -width compact
234 .Fn "client_callback2" "void *callback_arg" "bus_dma_segment_t *segs" \
235 "int nseg" "bus_size_t mapsize" "int error"
238 Callback2's behavior is the same as
239 .Vt bus_dmamap_callback_t
240 with the addition that the length of the data mapped is provided via
242 .It Vt bus_dmasync_op_t
243 Memory synchronization operation specifier.
244 Bus DMA requires explicit synchronization of memory with it's device
245 visible mapping in order to guarantee memory coherency.
248 allows the type of DMA operation that will be or has been performed
249 to be communicated to the system so that the correct coherency measures
251 All operations specified below are performed from the DMA engine's
253 .Bl -tag -width BUS_DMASYNC_POSTWRITE
254 .It Dv BUS_DMASYNC_PREREAD
255 Perform any synchronization required after an update of memory by the CPU
256 but prior to DMA read operations.
257 .It Dv BUS_DMASYNC_PREWRITE
258 Perform any synchronization required after an update of memory by the CPU
259 but prior to DMA write operations.
260 .It Dv BUS_DMASYNC_POSTREAD
261 Perform any synchronization required after DMA read operations, but prior
262 to CPU access of the memory.
263 .It Dv BUS_DMASYNC_POSTWRITE
264 Perform any synchronization required after DMA write operations, but prior
265 to CPU access of the memory.
270 .Bl -tag -width compact
271 .It Fn bus_dma_tag_create "parent" "alignment" "boundary" "lowaddr" \
272 "highaddr" "*filtfunc" "*filtfuncarg" "maxsize" "nsegments" "maxsegsz" \
274 Allocates a device specific DMA tag, and initializes it according to
275 the arguments provided:
276 .Bl -tag -width *filtfuncarg -compact
278 Indicates restrictions between the parent bridge, CPU memory, and the
280 May be NULL, if no DMA restrictions are to be inherited.
282 Alignment constraint, in bytes, of any mappings created using this tag.
283 The alignment must be a power of 2.
284 Hardware that can DMA starting at any address would specify
287 Hardware requiring DMA transfers to start on a multiple of 4K
291 Boundary constraint, in bytes, of the target DMA memory region.
292 The boundary indicates the set of addresses, all multiples of the
293 boundary argument, that cannot be crossed by a single
294 .Vt bus_dma_segment_t .
295 The boundary must be either a power of 2 or 0.
297 indicates that there are no boundary restrictions.
300 Bounds of the window of bus address space that
302 be directly accessed by the device.
303 The window contains all address greater than lowaddr and
304 less than or equal to highaddr.
305 For example, a device incapable of DMA above 4GB, would specify
307 .Dv BUS_SPACE_MAXADDR
309 .Dv BUS_SPACE_MAXADDR_32BIT .
310 Similarly a device that can only dma to addresses bellow 16MB would
311 specify a highaddr of
312 .Dv BUS_SPACE_MAXADDR
314 .Dv BUS_SPACE_MAXADDR_24BIT .
315 Some implementations requires that some region of device visible
316 address space, overlapping available host memory, be outside the
320 is used to bounce requests that would otherwise conflict with
321 the exclusion window.
323 Optional filter function (may be NULL) to be called for any attempt to
324 map memory into the window described by
328 A filter function is only required when the single window described
333 cannot adequately describe the constraints of the device.
334 The filter function will be called for every machine page
335 that overlaps the exclusion window.
337 Argument passed to all calls to the filter function for this tag.
340 Maximum size, in bytes, of the sum of all segment lengths in a given
341 DMA mapping associated with this tag.
343 Number of discontinuities (scatter/gather segments) allowed
344 in a DMA mapped region.
345 If there is no restriction,
346 .Dv BUS_SPACE_UNRESTRICTED
349 Maximum size, in bytes, of a segment in any DMA mapped region associated
354 .Bl -tag -width "BUS_DMA_ALLOCNOW" -compact
355 .It Dv BUS_DMA_ALLOCNOW
356 Allocate the resources necessary to guarantee that all map load
357 operations associated with this tag will not block.
358 If sufficient resources are not available,
363 Pointer to a bus_dma_tag_t where the resulting DMA tag will
369 if sufficient memory is not available for tag creation
370 or allocating mapping resources.
371 .It Fn bus_dma_tag_destroy "dmat"
372 Deallocate the DMA tag
375 .Fn bus_dma_tag_create .
379 if any DMA maps remain associated with
384 .It Fn bus_dmamap_create "dmat" "flags" "*mapp"
385 Allocates and initializes a DMA map.
386 Arguments are as follows:
387 .Bl -tag -width nsegments -compact
391 The value of this argument is currently undefined and should be
397 where the resulting DMA map will be stored.
402 if sufficient memory is not available for creating the
403 map or allocating mapping resources.
404 .It Fn bus_dmamap_destroy "dmat" "map"
405 Frees all resources associated with a given DMA map.
406 Arguments are as follows:
407 .Bl -tag -width dmat -compact
409 DMA tag used to allocate
412 The DMA map to destroy.
417 if a mapping is still active for
419 .It Fn bus_dmamap_load "dmat" "map" "buf" "buflen" "*callback" "..."
420 Creates a mapping in device visible address space of
424 associated with the DMA map
426 Arguments are as follows:
427 .Bl -tag -width buflen -compact
429 DMA tag used to allocate
432 A DMA map without a currently active mapping.
434 A kernel virtual address pointer to a contiguous (in KVA) buffer, to be
435 mapped into device visible address space.
437 The size of the buffer.
438 .It Fa callback Fa callback_arg
439 The callback function, and its argument.
441 The value of this argument is currently undefined, and should be
446 Return values to the caller are as follows:
447 .Bl -tag -width ".Er EINPROGRESS" -compact
449 The callback has been called and completed.
450 The status of the mapping has been delivered to the callback.
452 The mapping has been deferred for lack of resources.
453 The callback will be called as soon as resources are available.
454 Callbacks are serviced in FIFO order.
455 DMA maps created from DMA tags that are allocated with
458 flag will never return this status for a load operation.
460 The load request was invalid.
461 The callback has not, and will not be called.
462 This error value may indicate that
472 argument used to create the dma tag
476 When the callback is called, it is presented with an error value
477 indicating the disposition of the mapping.
478 Error may be one of the following:
479 .Bl -tag -width ".Er EINPROGRESS" -compact
481 The mapping was successful and the
483 callback argument contains an array of
484 .Vt bus_dma_segment_t
485 elements describing the mapping.
486 This array is only valid during the scope of the callback function.
488 A mapping could not be achieved within the segment constraints provided
489 in the tag even though the requested allocation size was less than maxsize.
491 .It Fn bus_dmamap_load_mbuf "dmat" "map" "mbuf" "callback2" "callback_arg" \
493 This is a variation of
495 which maps mbuf chains
499 argument is also passed to the callback routine, which
500 contains the mbuf chain's packet header length.
502 Mbuf chains are assumed to be in kernel virtual address space.
506 if the size of the mbuf chain exceeds the maximum limit of the
508 .It Fn bus_dmamap_load_uio "dmat" "map" "uio" "callback2" "callback_arg" "flags"
509 This is a variation of
511 which maps buffers pointed to by
516 argument is also passed to the callback routine, which contains the size of
525 then it is assumed that the buffer,
528 .Fa "uio->uio_td->td_proc" Ns 's
530 User space memory must be in-core and wired prior to attempting a map
532 .It Fn bus_dmamap_unload "dmat" "map"
534 Arguments are as follows:
535 .Bl -tag -width dmam -compact
537 DMA tag used to allocate
540 The DMA map that is to be unloaded.
543 .Fn bus_dmamap_unload
544 will not perform any implicit synchronization of DMA buffers.
545 This must be done explicitly by a call to
547 prior to unloading the map.
548 .It Fn bus_dmamap_sync "dmat" "map" "op"
549 Performs synchronization of a device visible mapping with the CPU visible
550 memory referenced by that mapping.
551 Arguments are as follows:
552 .Bl -tag -width dmat -compact
554 DMA tag used to allocate
557 The DMA mapping to be synchronized.
559 Type of synchronization operation to perform.
560 See the definition of
562 for a description of the acceptable values for
567 is the method used to ensure that CPU and device DMA access to shared
569 For example, the CPU might be used to setup the contents of a buffer
570 that is to be DMA'ed into a device.
571 To ensure that the data are visible via the device's mapping of that
572 memory, the buffer must be loaded and a dma sync operation of
573 .Dv BUS_DMASYNC_PREREAD
575 Additional sync operations must be performed after every CPU write
576 to this memory if additional DMA reads are to be performed.
577 Conversely, for the DMA write case, the buffer must be loaded,
578 and a dma sync operation of
579 .Dv BUS_DMASYNC_PREWRITE
581 The CPU will only be able to see the results of this DMA write
582 once the DMA has completed and a
583 .Dv BUS_DMASYNC_POSTWRITE
584 operation has been performed.
586 If DMA read and write operations are not preceded and followed by the
587 appropriate synchronization operations, behavior is undefined.
588 .It Fn bus_dmamem_alloc "dmat" "**vaddr" "flags" "mapp"
589 Allocates memory that is mapped into KVA at the address returned
592 that is permanently loaded into the newly created
596 Arguments are as follows:
597 .Bl -tag -width alignment -compact
599 DMA tag describing the constraints of the DMA mapping.
601 Pointer to a pointer that will hold the returned KVA mapping of
602 the allocated region.
604 Flags are defined as follows:
605 .Bl -tag -width BUS_DMA_NOWAIT -compact
606 .It Dv BUS_DMA_WAITOK
607 The routine can safely wait (sleep) for resources.
608 .It Dv BUS_DMA_NOWAIT
609 The routine is not allowed to wait for resources.
610 If resources are not available,
613 .It Dv BUS_DMA_COHERENT
614 Attempt to map this memory such that cache sync operations are
615 as cheap as possible.
616 This flag is typically set on memory that will be accessed by both
617 a CPU and a DMA engine, frequently.
618 Use of this flag does not remove the requirement of using
619 bus_dmamap_sync, but it may reduce the cost of performing
622 Causes the allocated memory to be set to all zeros.
625 Pointer to storage for the returned DMA map.
628 The size of memory to be allocated is
633 The current implementation of
635 will allocate all requests as a single segment.
637 Although no explicit loading is required to access the memory
638 referenced by the returned map, the synchronization requirements
645 if sufficient memory is not available for completing
647 .It Fn bus_dmamem_free "dmat" "*vaddr" "map"
648 Frees memory previously allocated by
649 .Fn bus_dmamem_alloc .
652 Arguments are as follows:
653 .Bl -tag -width vaddr -compact
657 Kernel virtual address of the memory.
659 DMA map to be invalidated.
663 Behavior is undefined if invalid arguments are passed to
664 any of the above functions.
665 If sufficient resources cannot be allocated for a given
670 routines that are not of type,
672 will return 0 on success or an error
673 code, as discussed above.
677 routines will succeed if provided with valid arguments.
685 .%A "Jason R. Thorpe"
686 .%T "A Machine-Independent DMA Framework for NetBSD"
687 .%J "Proceedings of the Summer 1998 USENIX Technical Conference"
688 .%Q "USENIX Association"
694 interface first appeared in
701 for use in the CAM SCSI subsystem.
702 The alterations to the original API were aimed to remove the need for
704 .Vt bus_dma_segment_t
707 while allowing callers to queue up on scarce resources.
711 interface was designed and implemented by
713 of the Numerical Aerospace Simulation Facility, NASA Ames Research Center.
714 Additional input on the
716 design was provided by
718 .An Chris Demetriou ,
726 This manual page was written by
729 .An Justin T. Gibbs .