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 indent
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 indent
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
192 on some platforms under certain conditions,
193 it should never be assumed that it will be
196 .It Vt bus_dmamap_callback_t
197 Client specified callback for receiving mapping information resulting from
201 .Fn bus_dmamap_load .
202 Callbacks are of the format:
203 .Bl -tag -width indent
205 .Fn "client_callback" "void *callback_arg" "bus_dma_segment_t *segs" \
206 "int nseg" "int error"
211 is the callback argument passed to dmamap load functions.
216 parameters describe an array of
217 .Vt bus_dma_segment_t
218 structures that represent the mapping.
219 This array is only valid within the scope of the callback function.
220 The success or failure of the mapping is indicated by the
223 More information on the use of callbacks can be found in the
224 description of the individual dmamap load functions.
225 .It Vt bus_dmamap_callback2_t
226 Client specified callback for receiving mapping information resulting from
230 .Fn bus_dmamap_load_uio
232 .Fn bus_dmamap_load_mbuf .
234 Callback2s are of the format:
235 .Bl -tag -width indent
237 .Fn "client_callback2" "void *callback_arg" "bus_dma_segment_t *segs" \
238 "int nseg" "bus_size_t mapsize" "int error"
241 Callback2's behavior is the same as
242 .Vt bus_dmamap_callback_t
243 with the addition that the length of the data mapped is provided via
245 .It Vt bus_dmasync_op_t
246 Memory synchronization operation specifier.
247 Bus DMA requires explicit synchronization of memory with its device
248 visible mapping in order to guarantee memory coherency.
251 allows the type of DMA operation that will be or has been performed
252 to be communicated to the system so that the correct coherency measures
254 The operations are represented as bitfield flags that can be combined together,
255 though it only makes sense to combine PRE flags or POST flags, not both.
258 description below for more details on how to use these operations.
260 All operations specified below are performed from the host memory point of view,
261 where a read implies data coming from the device to the host memory, and a write
262 implies data going from the host memory to the device.
263 Alternatively, the operations can be thought of in terms of driver operations,
264 where reading a network packet or storage sector corresponds to a read operation
267 .Bl -tag -width ".Dv BUS_DMASYNC_POSTWRITE"
268 .It Dv BUS_DMASYNC_PREREAD
269 Perform any synchronization required prior to an update of host memory by the
271 .It Dv BUS_DMASYNC_PREWRITE
272 Perform any synchronization required after an update of host memory by the CPU
273 and prior to device access to host memory.
274 .It Dv BUS_DMASYNC_POSTREAD
275 Perform any synchronization required after an update of host memory by the
276 device and prior to CPU access to host memory.
277 .It Dv BUS_DMASYNC_POSTWRITE
278 Perform any synchronization required after the device access to host memory.
280 .It Vt bus_dma_lock_t
281 Client specified lock/mutex manipulation method.
282 This will be called from
283 within busdma whenever a client lock needs to be manipulated.
284 In its current form, the function will be called immediately before
285 the callback for a dma load operation that has been deferred with
287 and immediately after with
289 If the load operation does not need to be deferred, then it
290 will not be called since the function loading the map should
291 be holding the appropriate locks.
292 This method is of the format:
293 .Bl -tag -width indent
295 .Fn "lockfunc" "void *lockfunc_arg" "bus_dma_lock_op_t op"
300 implementations are provided for convenience.
301 .Fn busdma_lock_mutex
302 performs standard mutex operations on the sleep mutex provided via the
305 .Fn bus_dma_tag_create .
307 will generate a system panic if it is called.
308 It is substituted into the tag when
313 .Fn bus_dma_tag_create .
314 .It Vt bus_dma_lock_op_t
315 Operations to be performed by the client-specified
317 .Bl -tag -width ".Dv BUS_DMA_UNLOCK"
319 Acquires and/or locks the client locking primitive.
320 .It Dv BUS_DMA_UNLOCK
321 Releases and/or unlocks the client locking primitive.
325 .Bl -tag -width indent
326 .It Fn bus_dma_tag_create "parent" "alignment" "boundary" "lowaddr" \
327 "highaddr" "*filtfunc" "*filtfuncarg" "maxsize" "nsegments" "maxsegsz" \
328 "flags" "lockfunc" "lockfuncarg" "*dmat"
329 Allocates a device specific DMA tag, and initializes it according to
330 the arguments provided:
331 .Bl -tag -width ".Fa filtfuncarg"
333 Indicates restrictions between the parent bridge, CPU memory, and the
335 May be NULL, if no DMA restrictions are to be inherited.
337 Alignment constraint, in bytes, of any mappings created using this tag.
338 The alignment must be a power of 2.
339 Hardware that can DMA starting at any address would specify
342 Hardware requiring DMA transfers to start on a multiple of 4K
346 Boundary constraint, in bytes, of the target DMA memory region.
347 The boundary indicates the set of addresses, all multiples of the
348 boundary argument, that cannot be crossed by a single
349 .Vt bus_dma_segment_t .
350 The boundary must be a power of 2 and must be no smaller than the
351 maximum segment size.
353 indicates that there are no boundary restrictions.
354 .It Fa lowaddr , highaddr
355 Bounds of the window of bus address space that
357 be directly accessed by the device.
358 The window contains all addresses greater than lowaddr and
359 less than or equal to highaddr.
360 For example, a device incapable of DMA above 4GB, would specify
362 .Dv BUS_SPACE_MAXADDR
364 .Dv BUS_SPACE_MAXADDR_32BIT .
365 Similarly a device that can only dma to addresses bellow 16MB would
366 specify a highaddr of
367 .Dv BUS_SPACE_MAXADDR
369 .Dv BUS_SPACE_MAXADDR_24BIT .
370 Some implementations requires that some region of device visible
371 address space, overlapping available host memory, be outside the
375 is used to bounce requests that would otherwise conflict with
376 the exclusion window.
378 Optional filter function (may be
380 to be called for any attempt to
381 map memory into the window described by
385 A filter function is only required when the single window described
390 cannot adequately describe the constraints of the device.
391 The filter function will be called for every machine page
392 that overlaps the exclusion window.
394 Argument passed to all calls to the filter function for this tag.
398 Maximum size, in bytes, of the sum of all segment lengths in a given
399 DMA mapping associated with this tag.
401 Number of discontinuities (scatter/gather segments) allowed
402 in a DMA mapped region.
403 If there is no restriction,
404 .Dv BUS_SPACE_UNRESTRICTED
407 Maximum size, in bytes, of a segment in any DMA mapped region associated
412 .Bl -tag -width ".Dv BUS_DMA_ALLOCNOW"
413 .It Dv BUS_DMA_ALLOCNOW
414 Pre-allocate enough resources to handle at least one map load operation on
416 If sufficient resources are not available,
419 This should not be used for tags that only describe buffers that will be
421 .Fn bus_dmamem_alloc .
422 Also, due to resource sharing with other tags, this flag does not guarantee
423 that resources will be allocated or reserved exclusively for this tag.
424 It should be treated only as a minor optimization.
427 Optional lock manipulation function (may be
429 to be called when busdma
430 needs to manipulate a lock on behalf of the client.
437 Optional argument to be passed to the function specified by
440 Pointer to a bus_dma_tag_t where the resulting DMA tag will
446 if sufficient memory is not available for tag creation
447 or allocating mapping resources.
448 .It Fn bus_dma_tag_destroy "dmat"
449 Deallocate the DMA tag
452 .Fn bus_dma_tag_create .
456 if any DMA maps remain associated with
461 .It Fn bus_dmamap_create "dmat" "flags" "*mapp"
462 Allocates and initializes a DMA map.
463 Arguments are as follows:
464 .Bl -tag -width ".Fa nsegments"
468 The value of this argument is currently undefined and should be
474 where the resulting DMA map will be stored.
479 if sufficient memory is not available for creating the
480 map or allocating mapping resources.
481 .It Fn bus_dmamap_destroy "dmat" "map"
482 Frees all resources associated with a given DMA map.
483 Arguments are as follows:
484 .Bl -tag -width ".Fa dmat"
486 DMA tag used to allocate
489 The DMA map to destroy.
494 if a mapping is still active for
496 .It Fn bus_dmamap_load "dmat" "map" "buf" "buflen" "*callback" \
497 "callback_arg" "flags"
498 Creates a mapping in device visible address space of
502 associated with the DMA map
504 This call will always return immediately and will not block for any reason.
505 Arguments are as follows:
506 .Bl -tag -width ".Fa buflen"
508 DMA tag used to allocate
511 A DMA map without a currently active mapping.
513 A kernel virtual address pointer to a contiguous (in KVA) buffer, to be
514 mapped into device visible address space.
516 The size of the buffer.
517 .It Fa callback Fa callback_arg
518 The callback function, and its argument.
519 This function is called once sufficient mapping resources are available for
521 If resources are temporarily unavailable, this function will be deferred until
522 later, but the load operation will still return immediately to the caller.
523 Thus, callers should not assume that the callback will be called before the
524 load returns, and code should be structured appropriately to handle this.
525 See below for specific flags and error codes that control this behavior.
528 .Bl -tag -width ".Dv BUS_DMA_NOWAIT"
529 .It Dv BUS_DMA_NOWAIT
530 The load should not be deferred in case of insufficient mapping resources,
531 and instead should return immediately with an appropriate error.
535 Return values to the caller are as follows:
536 .Bl -tag -width ".Er EINPROGRESS"
538 The callback has been called and completed.
539 The status of the mapping has been delivered to the callback.
541 The mapping has been deferred for lack of resources.
542 The callback will be called as soon as resources are available.
543 Callbacks are serviced in FIFO order.
544 To ensure that ordering is guaranteed, all subsequent load requests will also
545 be deferred until all callbacks have been processed.
547 The load request has failed due to insufficient resources, and the caller
548 specifically used the
552 The load request was invalid.
553 The callback has been called and has been provided the same error.
554 This error value may indicate that
564 argument used to create the dma tag
568 When the callback is called, it is presented with an error value
569 indicating the disposition of the mapping.
570 Error may be one of the following:
571 .Bl -tag -width ".Er EINPROGRESS"
573 The mapping was successful and the
575 callback argument contains an array of
576 .Vt bus_dma_segment_t
577 elements describing the mapping.
578 This array is only valid during the scope of the callback function.
580 A mapping could not be achieved within the segment constraints provided
581 in the tag even though the requested allocation size was less than maxsize.
583 .It Fn bus_dmamap_load_mbuf "dmat" "map" "mbuf" "callback2" "callback_arg" \
585 This is a variation of
587 which maps mbuf chains
591 argument is also passed to the callback routine, which
592 contains the mbuf chain's packet header length.
595 flag is implied, thus no callback deferral will happen.
597 Mbuf chains are assumed to be in kernel virtual address space.
599 Beside the error values listed for
600 .Fn bus_dmamap_load ,
602 will be returned if the size of the mbuf chain exceeds the maximum limit of the
604 .It Fn bus_dmamap_load_mbuf_sg "dmat" "map" "mbuf" "segs" "nsegs" "flags"
606 .Fn bus_dmamap_load_mbuf
607 except that it returns immediately without calling a callback function.
608 It is provided for efficiency.
609 The scatter/gather segment array
611 is provided by the caller and filled in directly by the function.
614 argument is returned with the number of segments filled in.
615 Returns the same errors as
616 .Fn bus_dmamap_load_mbuf .
617 .It Fn bus_dmamap_load_uio "dmat" "map" "uio" "callback2" "callback_arg" "flags"
618 This is a variation of
620 which maps buffers pointed to by
625 argument is also passed to the callback routine, which contains the size of
631 flag is implied, thus no callback deferral will happen.
632 Returns the same errors as
633 .Fn bus_dmamap_load .
639 then it is assumed that the buffer,
642 .Fa "uio->uio_td->td_proc" Ns 's
644 User space memory must be in-core and wired prior to attempting a map
646 Pages may be locked using
648 .It Fn bus_dmamap_unload "dmat" "map"
650 Arguments are as follows:
651 .Bl -tag -width ".Fa dmam"
653 DMA tag used to allocate
656 The DMA map that is to be unloaded.
659 .Fn bus_dmamap_unload
660 will not perform any implicit synchronization of DMA buffers.
661 This must be done explicitly by a call to
663 prior to unloading the map.
664 .It Fn bus_dmamap_sync "dmat" "map" "op"
665 Performs synchronization of a device visible mapping with the CPU visible
666 memory referenced by that mapping.
667 Arguments are as follows:
668 .Bl -tag -width ".Fa dmat"
670 DMA tag used to allocate
673 The DMA mapping to be synchronized.
675 Type of synchronization operation to perform.
676 See the definition of
678 for a description of the acceptable values for
685 is the method used to ensure that CPU's and device's direct
686 memory access (DMA) to shared
688 For example, the CPU might be used to set up the contents of a buffer
689 that is to be made available to a device.
690 To ensure that the data are visible via the device's mapping of that
691 memory, the buffer must be loaded and a DMA sync operation of
692 .Dv BUS_DMASYNC_PREWRITE
694 If later CPU modifies this buffer again, another
695 .Dv BUS_DMASYNC_PREWRITE
696 sync operation must be performed before an additional
697 access to this memory by a device.
698 Conversely, a device updates the memory that is to be read by a CPU.
699 In this case, the buffer must be loaded, and a DMA sync operation of
700 .Dv BUS_DMASYNC_PREREAD
702 The CPU will only be able to see the results of this memory update
703 once the DMA operation has completed and a
704 .Dv BUS_DMASYNC_POSTREAD
705 sync operation has been performed.
707 If read and write operations are not preceded and followed by the
708 appropriate synchronization operations, behavior is undefined.
709 .It Fn bus_dmamem_alloc "dmat" "**vaddr" "flags" "*mapp"
710 Allocates memory that is mapped into KVA at the address returned
713 that is permanently loaded into the newly created
717 Arguments are as follows:
718 .Bl -tag -width ".Fa alignment"
720 DMA tag describing the constraints of the DMA mapping.
722 Pointer to a pointer that will hold the returned KVA mapping of
723 the allocated region.
725 Flags are defined as follows:
726 .Bl -tag -width ".Dv BUS_DMA_NOWAIT"
727 .It Dv BUS_DMA_WAITOK
728 The routine can safely wait (sleep) for resources.
729 .It Dv BUS_DMA_NOWAIT
730 The routine is not allowed to wait for resources.
731 If resources are not available,
734 .It Dv BUS_DMA_COHERENT
735 Attempt to map this memory such that cache sync operations are
736 as cheap as possible.
737 This flag is typically set on memory that will be accessed by both
738 a CPU and a DMA engine, frequently.
739 Use of this flag does not remove the requirement of using
740 bus_dmamap_sync, but it may reduce the cost of performing
744 flag is currently implemented on sparc64 and arm.
746 Causes the allocated memory to be set to all zeros.
751 where the resulting DMA map will be stored.
754 The size of memory to be allocated is
759 The current implementation of
761 will allocate all requests as a single segment.
763 An initial load operation is required to obtain the bus address of the allocated
764 memory, and an unload operation is required before freeing the memory, as
766 .Fn bus_dmamem_free .
767 Maps are automatically handled by this function and should not be explicitly
768 allocated or destroyed.
770 Although an explicit load is not required for each access to the memory
771 referenced by the returned map, the synchronization requirements
774 section still apply and should be used to achieve portability on architectures
775 without coherent buses.
779 if sufficient memory is not available for completing
781 .It Fn bus_dmamem_free "dmat" "*vaddr" "map"
782 Frees memory previously allocated by
783 .Fn bus_dmamem_alloc .
786 Arguments are as follows:
787 .Bl -tag -width ".Fa vaddr"
791 Kernel virtual address of the memory.
793 DMA map to be invalidated.
797 Behavior is undefined if invalid arguments are passed to
798 any of the above functions.
799 If sufficient resources cannot be allocated for a given
804 routines that are not of type,
806 will return 0 on success or an error
807 code, as discussed above.
811 routines will succeed if provided with valid arguments.
813 Two locking protocols are used by
815 The first is a private global lock that is used to synchronize access to the
816 bounce buffer pool on the architectures that make use of them.
817 This lock is strictly a leaf lock that is only used internally to
819 and is not exposed to clients of the API.
821 The second protocol involves protecting various resources stored in the tag.
824 operations are done through requests from the driver that created the tag,
825 the most efficient way to protect the tag resources is through the lock that
829 acts on its own without being called by the driver, the lock primitive
830 specified in the tag is acquired and released automatically.
831 An example of this is when the
833 callback function is called from a deferred context instead of the driver
835 This means that certain
837 functions must always be called with the same lock held that is specified in the
839 These functions include:
841 .Bl -item -offset indent -compact
845 .Fn bus_dmamap_load_uio
847 .Fn bus_dmamap_load_mbuf
849 .Fn bus_dmamap_load_mbuf_sg
851 .Fn bus_dmamap_unload
856 There is one exception to this rule.
857 It is common practice to call some of these functions during driver start-up
858 without any locks held.
859 So long as there is a guarantee of no possible concurrent use of the tag by
860 different threads during this operation, it is safe to not hold a lock for
865 operations should not be called with the driver lock held, either because
866 they are already protected by an internal lock, or because they might sleep
867 due to memory or resource allocation.
868 The following functions must not be
869 called with any non-sleepable locks held:
871 .Bl -item -offset indent -compact
873 .Fn bus_dma_tag_create
875 .Fn bus_dmamap_create
880 All other functions do not have a locking protocol and can thus be
881 called with or without and system or driver locks held.
890 .%A "Jason R. Thorpe"
891 .%T "A Machine-Independent DMA Framework for NetBSD"
892 .%J "Proceedings of the Summer 1998 USENIX Technical Conference"
893 .%Q "USENIX Association"
899 interface first appeared in
906 for use in the CAM SCSI subsystem.
907 The alterations to the original API were aimed to remove the need for
909 .Vt bus_dma_segment_t
912 while allowing callers to queue up on scarce resources.
916 interface was designed and implemented by
918 of the Numerical Aerospace Simulation Facility, NASA Ames Research Center.
919 Additional input on the
921 design was provided by
923 .An Chris Demetriou ,
935 benefits from the contributions of
936 .An Justin T. Gibbs ,
939 .An Matthew N. Dodd ,
942 .An Jake Burkholder ,
943 .An Takahashi Yoshihiro ,
947 This manual page was written by
950 .An Justin T. Gibbs .