1 .\" Copyright (c) 2002, 2003 Hiten M. Pandya.
2 .\" All rights reserved.
4 .\" Redistribution and use in source and binary forms, with or without
5 .\" modification, are permitted provided that the following conditions
7 .\" 1. Redistributions of source code must retain the above copyright
8 .\" notice, this list of conditions, and the following disclaimer,
9 .\" without modification, immediately at the beginning of the file.
10 .\" 2. The name of the author may not be used to endorse or promote products
11 .\" derived from this software without specific prior written permission.
13 .\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
14 .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15 .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16 .\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR, CONTRIBUTORS OR THE
17 .\" VOICES IN HITEN PANDYA'S HEAD BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
18 .\" SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
19 .\" TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
20 .\" PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
21 .\" LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
22 .\" NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
23 .\" SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25 .\" Copyright (c) 1996, 1997, 1998, 2001 The NetBSD Foundation, Inc.
26 .\" All rights reserved.
28 .\" This code is derived from software contributed to The NetBSD Foundation
29 .\" by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
30 .\" NASA Ames Research Center.
32 .\" Redistribution and use in source and binary forms, with or without
33 .\" modification, are permitted provided that the following conditions
35 .\" 1. Redistributions of source code must retain the above copyright
36 .\" notice, this list of conditions and the following disclaimer.
37 .\" 2. Redistributions in binary form must reproduce the above copyright
38 .\" notice, this list of conditions and the following disclaimer in the
39 .\" documentation and/or other materials provided with the distribution.
40 .\" 3. All advertising materials mentioning features or use of this software
41 .\" must display the following acknowledgment:
42 .\" This product includes software developed by the NetBSD
43 .\" Foundation, Inc. and its contributors.
44 .\" 4. Neither the name of The NetBSD Foundation nor the names of its
45 .\" contributors may be used to endorse or promote products derived
46 .\" from this software without specific prior written permission.
48 .\" THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
49 .\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
50 .\" TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
51 .\" PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
52 .\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
53 .\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
54 .\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
55 .\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
56 .\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
57 .\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
58 .\" POSSIBILITY OF SUCH DAMAGE.
61 .\" $NetBSD: bus_dma.9,v 1.25 2002/10/14 13:43:16 wiz Exp $
68 .Nm bus_dma_tag_create ,
69 .Nm bus_dma_tag_destroy ,
70 .Nm bus_dmamap_create ,
71 .Nm bus_dmamap_destroy ,
73 .Nm bus_dmamap_load_mbuf ,
74 .Nm bus_dmamap_load_mbuf_sg ,
75 .Nm bus_dmamap_load_uio ,
76 .Nm bus_dmamap_unload ,
78 .Nm bus_dmamem_alloc ,
80 .Nd Bus and Machine Independent DMA Mapping Interface
84 .Fn bus_dma_tag_create "bus_dma_tag_t parent" "bus_size_t alignment" \
85 "bus_size_t boundary" "bus_addr_t lowaddr" "bus_addr_t highaddr" \
86 "bus_dma_filter_t *filtfunc" "void *filtfuncarg" "bus_size_t maxsize" \
87 "int nsegments" "bus_size_t maxsegsz" "int flags" "bus_dma_lock_t *lockfunc" \
88 "void *lockfuncarg" "bus_dma_tag_t *dmat"
90 .Fn bus_dma_tag_destroy "bus_dma_tag_t dmat"
92 .Fn bus_dmamap_create "bus_dma_tag_t dmat" "int flags" "bus_dmamap_t *mapp"
94 .Fn bus_dmamap_destroy "bus_dma_tag_t dmat" "bus_dmamap_t map"
96 .Fn bus_dmamap_load "bus_dma_tag_t dmat" "bus_dmamap_t map" "void *buf" \
97 "bus_size_t buflen" "bus_dmamap_callback_t *callback" "void *callback_arg" \
100 .Fn bus_dmamap_load_mbuf "bus_dma_tag_t dmat" "bus_dmamap_t map" \
101 "struct mbuf *mbuf" "bus_dmamap_callback2_t *callback" "void *callback_arg" \
104 .Fn bus_dmamap_load_mbuf_sg "bus_dma_tag_t dmat" "bus_dmamap_t map" \
105 "struct mbuf *mbuf" "bus_dma_segment_t *segs" "int *nsegs" "int flags"
107 .Fn bus_dmamap_load_uio "bus_dma_tag_t dmat" "bus_dmamap_t map" \
108 "struct uio *uio" "bus_dmamap_callback2_t *callback" "void *callback_arg" \
111 .Fn bus_dmamap_unload "bus_dma_tag_t dmat" "bus_dmamap_t map"
113 .Fn bus_dmamap_sync "bus_dma_tag_t dmat" "bus_dmamap_t map" \
116 .Fn bus_dmamem_alloc "bus_dma_tag_t dmat" "void **vaddr" \
117 "int flags" "bus_dmamap_t *mapp"
119 .Fn bus_dmamem_free "bus_dma_tag_t dmat" "void *vaddr" \
122 Direct Memory Access (DMA) is a method of transferring data
123 without involving the CPU, thus providing higher performance.
124 A DMA transaction can be achieved between device to memory,
125 device to device, or memory to memory.
129 API is a bus, device, and machine-independent (MI) interface to
131 It provides the client with flexibility and simplicity by
132 abstracting machine dependent issues like setting up
133 DMA mappings, handling cache issues, bus specific features
135 .Sh STRUCTURES AND TYPES
136 .Bl -tag -width compact
138 A machine-dependent (MD) opaque type that describes the
139 characteristics of DMA transactions.
140 DMA tags are organized into a hierarchy, with each child
141 tag inheriting the restrictions of its parent.
142 This allows all devices along the path of DMA transactions
143 to contribute to the constraints of those transactions.
144 .It Vt bus_dma_filter_t
145 Client specified address filter having the format:
146 .Bl -tag -width compact
148 .Fn "client_filter" "void *filtarg" "bus_addr_t testaddr"
151 Address filters can be specified during tag creation to allow
152 for devices whose DMA address restrictions cannot be specified
156 is client specified during tag creation to be passed to all
157 invocations of the callback.
160 argument contains a potential starting address of a DMA mapping.
161 The filter function operates on the set of addresses from
164 .Ql trunc_page(testaddr) + PAGE_SIZE - 1 ,
166 The filter function should return zero for any mapping in this range
167 that can be accommodated by the device and non-zero otherwise.
168 .It Vt bus_dma_segment_t
169 A machine-dependent type that describes individual
178 field contains the device visible address of the DMA segment, and
180 contains the length of the DMA segment.
181 Although the DMA segments returned by a mapping call will adhere to
182 all restrictions necessary for a successful DMA operation, some conversion
183 (e.g.\& a conversion from host byte order to the device's byte order) is
184 almost always required when presenting segment information to the device.
186 A machine-dependent opaque type describing an individual mapping.
187 One map is used for each memory allocation that will be loaded.
188 Maps can be reused once they have been unloaded.
189 Multiple maps can be associated with one DMA tag.
190 While the value of the map may evaluate to NULL on some platforms under
191 certain conditions, it should never be assumed that it will be NULL in all
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 its 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 The operations are represented as bitfield flags that can be combined together,
252 though it only makes sense to combine PRE flags or POST flags, not both.
255 description below for more details on how to use these operations.
257 All operations specified below are performed from the host memory point of view,
258 where a read implies data coming from the device to the host memory, and a write
259 implies data going from the host memory to the device.
260 Alternately, the operations can be thought of in terms of driver operations,
261 where reading a network packet or storage sector corresponds to a read operation
264 .Bl -tag -width BUS_DMASYNC_POSTWRITE
265 .It Dv BUS_DMASYNC_PREREAD
266 Perform any synchronization required prior to an update of host memory by the
268 .It Dv BUS_DMASYNC_PREWRITE
269 Perform any synchronization required after an update of host memory by the CPU
270 and prior to DMA write operations.
271 .It Dv BUS_DMASYNC_POSTREAD
272 Perform any synchronization required after DMA read operations and prior to
273 CPU access to host memory.
274 .It Dv BUS_DMASYNC_POSTWRITE
275 Perform any synchronization required after DMA write operations.
277 .It Vt bus_dma_lock_t
278 Client specified lock/mutex manipulation method.
279 This will be called from
280 within busdma whenever a client lock needs to be manipulated.
281 In its current form, the function will be called immediately before
282 the callback for a dma load operation that has been deferred with
284 and immediately after with
286 If the load operation does not need to be deferred, then it
287 will not be called since the function loading the map should
288 be holding the appropriate locks.
289 This method is of the format:
290 .Bl -tag -width compact
292 .Fn "lockfunc" "void *lockfunc_arg" "bus_dma_lock_op_t op"
297 implementations are provided for convenience.
298 .Fn busdma_lock_mutex
299 performs standard mutex operations on the sleep mutex provided via the
302 .Fn bus_dma_tag_create .
304 will generate a system panic if it is called.
305 It is substituted into the tag when
308 .Fn bus_dma_tag_create .
309 .It Vt bus_dma_lock_op_t
310 Operations to be performed by the client-specified
312 .Bl -tag -width BUS_DMA_UNLOCK
314 Acquires and/or locks the client locking primitive.
315 .It Dv BUS_DMA_UNLOCK
316 Releases and/or unlocks the client locking primitive.
321 .Bl -tag -width compact
322 .It Fn bus_dma_tag_create "parent" "alignment" "boundary" "lowaddr" \
323 "highaddr" "*filtfunc" "*filtfuncarg" "maxsize" "nsegments" "maxsegsz" \
324 "flags" "lockfunc" "lockfuncarg" "*dmat"
325 Allocates a device specific DMA tag, and initializes it according to
326 the arguments provided:
327 .Bl -tag -width *filtfuncarg -compact
329 Indicates restrictions between the parent bridge, CPU memory, and the
331 May be NULL, if no DMA restrictions are to be inherited.
333 Alignment constraint, in bytes, of any mappings created using this tag.
334 The alignment must be a power of 2.
335 Hardware that can DMA starting at any address would specify
338 Hardware requiring DMA transfers to start on a multiple of 4K
342 Boundary constraint, in bytes, of the target DMA memory region.
343 The boundary indicates the set of addresses, all multiples of the
344 boundary argument, that cannot be crossed by a single
345 .Vt bus_dma_segment_t .
346 The boundary must be a power of 2 and must be no smaller than the
347 maximum segment size.
349 indicates that there are no boundary restrictions.
352 Bounds of the window of bus address space that
354 be directly accessed by the device.
355 The window contains all addresses greater than lowaddr and
356 less than or equal to highaddr.
357 For example, a device incapable of DMA above 4GB, would specify
359 .Dv BUS_SPACE_MAXADDR
361 .Dv BUS_SPACE_MAXADDR_32BIT .
362 Similarly a device that can only dma to addresses bellow 16MB would
363 specify a highaddr of
364 .Dv BUS_SPACE_MAXADDR
366 .Dv BUS_SPACE_MAXADDR_24BIT .
367 Some implementations requires that some region of device visible
368 address space, overlapping available host memory, be outside the
372 is used to bounce requests that would otherwise conflict with
373 the exclusion window.
375 Optional filter function (may be NULL) to be called for any attempt to
376 map memory into the window described by
380 A filter function is only required when the single window described
385 cannot adequately describe the constraints of the device.
386 The filter function will be called for every machine page
387 that overlaps the exclusion window.
389 Argument passed to all calls to the filter function for this tag.
392 Maximum size, in bytes, of the sum of all segment lengths in a given
393 DMA mapping associated with this tag.
395 Number of discontinuities (scatter/gather segments) allowed
396 in a DMA mapped region.
397 If there is no restriction,
398 .Dv BUS_SPACE_UNRESTRICTED
401 Maximum size, in bytes, of a segment in any DMA mapped region associated
406 .Bl -tag -width "BUS_DMA_ALLOCNOW" -compact
407 .It Dv BUS_DMA_ALLOCNOW
408 Pre-allocate enough resources to handle at least one map load operation on
410 If sufficient resources are not available,
413 This should not be used for tags that only describe buffers that will be
415 .Fn bus_dmamem_alloc .
416 Also, due to resource sharing with other tags, this flag does not guarantee
417 that resources will be allocated or reserved exclusively for this tag.
418 It should be treated only as a minor optimization.
421 Optional lock manipulation function (may be NULL) to be called when busdma
422 needs to manipulate a lock on behalf of the client.
423 If NULL is specified,
427 Optional argument to be passed to the function specified by
430 Pointer to a bus_dma_tag_t where the resulting DMA tag will
436 if sufficient memory is not available for tag creation
437 or allocating mapping resources.
438 .It Fn bus_dma_tag_destroy "dmat"
439 Deallocate the DMA tag
442 .Fn bus_dma_tag_create .
446 if any DMA maps remain associated with
451 .It Fn bus_dmamap_create "dmat" "flags" "*mapp"
452 Allocates and initializes a DMA map.
453 Arguments are as follows:
454 .Bl -tag -width nsegments -compact
458 The value of this argument is currently undefined and should be
464 where the resulting DMA map will be stored.
469 if sufficient memory is not available for creating the
470 map or allocating mapping resources.
471 .It Fn bus_dmamap_destroy "dmat" "map"
472 Frees all resources associated with a given DMA map.
473 Arguments are as follows:
474 .Bl -tag -width dmat -compact
476 DMA tag used to allocate
479 The DMA map to destroy.
484 if a mapping is still active for
486 .It Fn bus_dmamap_load "dmat" "map" "buf" "buflen" "*callback" \
487 "callback_arg" "flags"
488 Creates a mapping in device visible address space of
492 associated with the DMA map
494 This call will always return immediately and will not block for any reason.
495 Arguments are as follows:
496 .Bl -tag -width buflen -compact
498 DMA tag used to allocate
501 A DMA map without a currently active mapping.
503 A kernel virtual address pointer to a contiguous (in KVA) buffer, to be
504 mapped into device visible address space.
506 The size of the buffer.
507 .It Fa callback Fa callback_arg
508 The callback function, and its argument.
509 This function is called once sufficient mapping resources are available for
511 If resources are temporarily unavailable, this function will be deferred until
512 later, but the load operation will still return immediately to the caller.
513 Thus, callers should not assume that the callback will be called before the
514 load returns, and code should be structured appropriately to handle this.
515 See below for specific flags and error codes that control this behavior.
518 .Bl -tag -width BUS_DMA_NOWAIT -compact
519 .It Er BUS_DMA_NOWAIT
520 The load should not be deferred in case of insufficient mapping resources,
521 and instead should return immediately with an appropriate error.
525 Return values to the caller are as follows:
526 .Bl -tag -width EINPROGRESS -compact
528 The callback has been called and completed.
529 The status of the mapping has been delivered to the callback.
531 The mapping has been deferred for lack of resources.
532 The callback will be called as soon as resources are available.
533 Callbacks are serviced in FIFO order.
534 To ensure that ordering is guaranteed, all subsequent load requests will also
535 be deferred until all callbacks have been processed.
537 The load request has failed due to insufficient resources, and the caller
538 specifically used the
542 The load request was invalid.
543 The callback has been called and has been provided the same error.
544 This error value may indicate that
554 argument used to create the dma tag
558 When the callback is called, it is presented with an error value
559 indicating the disposition of the mapping.
560 Error may be one of the following:
561 .Bl -tag -width EINPROGRESS -compact
563 The mapping was successful and the
565 callback argument contains an array of
566 .Vt bus_dma_segment_t
567 elements describing the mapping.
568 This array is only valid during the scope of the callback function.
570 A mapping could not be achieved within the segment constraints provided
571 in the tag even though the requested allocation size was less than maxsize.
573 .It Fn bus_dmamap_load_mbuf "dmat" "map" "mbuf" "callback2" "callback_arg" \
575 This is a variation of
577 which maps mbuf chains
581 argument is also passed to the callback routine, which
582 contains the mbuf chain's packet header length.
585 flag is implied, thus no callback deferral will happen.
587 Mbuf chains are assumed to be in kernel virtual address space.
589 Beside the error values listed for
590 .Fn bus_dmamap_load ,
592 will be returned if the size of the mbuf chain exceeds the maximum limit of the
594 .It Fn bus_dmamap_load_mbuf_sg "dmat" "map" "mbuf" "segs" "nsegs" "flags"
596 .Fn bus_dmamap_load_mbuf
597 except that it returns immediately without calling a callback function.
598 It is provided for efficiency.
599 The scatter/gather segment array
601 is provided by the caller and filled in directly by the function.
604 argument is returned with the number of segments filled in.
605 Returns the same errors as
606 .Fn bus_dmamap_load_mbuf .
607 .It Fn bus_dmamap_load_uio "dmat" "map" "uio" "callback2" "callback_arg" "flags"
608 This is a variation of
610 which maps buffers pointed to by
615 argument is also passed to the callback routine, which contains the size of
621 flag is implied, thus no callback deferral will happen.
622 Returns the same errors as
623 .Fn bus_dmamap_load .
629 then it is assumed that the buffer,
632 .Fa "uio->uio_td->td_proc" Ns 's
634 User space memory must be in-core and wired prior to attempting a map
636 Pages may be locked using
638 .It Fn bus_dmamap_unload "dmat" "map"
640 Arguments are as follows:
641 .Bl -tag -width dmam -compact
643 DMA tag used to allocate
646 The DMA map that is to be unloaded.
649 .Fn bus_dmamap_unload
650 will not perform any implicit synchronization of DMA buffers.
651 This must be done explicitly by a call to
653 prior to unloading the map.
654 .It Fn bus_dmamap_sync "dmat" "map" "op"
655 Performs synchronization of a device visible mapping with the CPU visible
656 memory referenced by that mapping.
657 Arguments are as follows:
658 .Bl -tag -width dmat -compact
660 DMA tag used to allocate
663 The DMA mapping to be synchronized.
665 Type of synchronization operation to perform.
666 See the definition of
668 for a description of the acceptable values for
673 is the method used to ensure that CPU and device DMA access to shared
675 For example, the CPU might be used to setup the contents of a buffer
676 that is to be DMA'ed into a device.
677 To ensure that the data are visible via the device's mapping of that
678 memory, the buffer must be loaded and a dma sync operation of
679 .Dv BUS_DMASYNC_PREREAD
681 Additional sync operations must be performed after every CPU write
682 to this memory if additional DMA reads are to be performed.
683 Conversely, for the DMA write case, the buffer must be loaded,
684 and a dma sync operation of
685 .Dv BUS_DMASYNC_PREWRITE
687 The CPU will only be able to see the results of this DMA write
688 once the DMA has completed and a
689 .Dv BUS_DMASYNC_POSTWRITE
690 operation has been performed.
692 If DMA read and write operations are not preceded and followed by the
693 appropriate synchronization operations, behavior is undefined.
694 .It Fn bus_dmamem_alloc "dmat" "**vaddr" "flags" "*mapp"
695 Allocates memory that is mapped into KVA at the address returned
698 that is permanently loaded into the newly created
702 Arguments are as follows:
703 .Bl -tag -width alignment -compact
705 DMA tag describing the constraints of the DMA mapping.
707 Pointer to a pointer that will hold the returned KVA mapping of
708 the allocated region.
710 Flags are defined as follows:
711 .Bl -tag -width BUS_DMA_NOWAIT -compact
712 .It Dv BUS_DMA_WAITOK
713 The routine can safely wait (sleep) for resources.
714 .It Dv BUS_DMA_NOWAIT
715 The routine is not allowed to wait for resources.
716 If resources are not available,
719 .It Dv BUS_DMA_COHERENT
720 Attempt to map this memory such that cache sync operations are
721 as cheap as possible.
722 This flag is typically set on memory that will be accessed by both
723 a CPU and a DMA engine, frequently.
724 Use of this flag does not remove the requirement of using
725 bus_dmamap_sync, but it may reduce the cost of performing
728 Causes the allocated memory to be set to all zeros.
733 where the resulting DMA map will be stored.
736 The size of memory to be allocated is
741 The current implementation of
743 will allocate all requests as a single segment.
745 An initial load operation is required to obtain the bus address of the allocated
746 memory, and an unload operation is required before freeing the memory, as
748 .Fn bus_dmamem_free .
749 Maps are automatically handled by this function and should not be explicitly
750 allocated or destroyed.
752 Although an explicit load is not required for each access to the memory
753 referenced by the returned map, the synchronization requirements
756 section still apply and should be used to achieve portability on architecutures
757 without coherent buses.
761 if sufficient memory is not available for completing
763 .It Fn bus_dmamem_free "dmat" "*vaddr" "map"
764 Frees memory previously allocated by
765 .Fn bus_dmamem_alloc .
768 Arguments are as follows:
769 .Bl -tag -width vaddr -compact
773 Kernel virtual address of the memory.
775 DMA map to be invalidated.
779 Behavior is undefined if invalid arguments are passed to
780 any of the above functions.
781 If sufficient resources cannot be allocated for a given
786 routines that are not of type,
788 will return 0 on success or an error
789 code, as discussed above.
793 routines will succeed if provided with valid arguments.
802 .%A "Jason R. Thorpe"
803 .%T "A Machine-Independent DMA Framework for NetBSD"
804 .%J "Proceedings of the Summer 1998 USENIX Technical Conference"
805 .%Q "USENIX Association"
811 interface first appeared in
818 for use in the CAM SCSI subsystem.
819 The alterations to the original API were aimed to remove the need for
821 .Vt bus_dma_segment_t
824 while allowing callers to queue up on scarce resources.
828 interface was designed and implemented by
830 of the Numerical Aerospace Simulation Facility, NASA Ames Research Center.
831 Additional input on the
833 design was provided by
835 .An Chris Demetriou ,
847 benefits from the contributions of
848 .An Justin T. Gibbs ,
851 .An Matthew N. Dodd ,
854 .An Jake Burkholder ,
855 .An Takahashi Yoshihiro ,
859 This manual page was written by
862 .An Justin T. Gibbs .