xref: /freebsd/share/man/man9/bus_dma.9 (revision a0409676120c1e558d0ade943019934e0f15118d)
1.\" Copyright (c) 2002, 2003 Hiten M. Pandya.
2.\" All rights reserved.
3.\"
4.\" Redistribution and use in source and binary forms, with or without
5.\" modification, are permitted provided that the following conditions
6.\" are met:
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.
12.\"
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.
24.\"
25.\" Copyright (c) 1996, 1997, 1998, 2001 The NetBSD Foundation, Inc.
26.\" All rights reserved.
27.\"
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.
31.\"
32.\" Redistribution and use in source and binary forms, with or without
33.\" modification, are permitted provided that the following conditions
34.\" are met:
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.\"
41.\" THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
42.\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
43.\" TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
44.\" PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
45.\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
46.\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
47.\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
48.\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
49.\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
50.\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
51.\" POSSIBILITY OF SUCH DAMAGE.
52.\"
53.\" $FreeBSD$
54.\" $NetBSD: bus_dma.9,v 1.25 2002/10/14 13:43:16 wiz Exp $
55.\"
56.Dd May 25, 2020
57.Dt BUS_DMA 9
58.Os
59.Sh NAME
60.Nm bus_dma ,
61.Nm bus_dma_tag_create ,
62.Nm bus_dma_tag_destroy ,
63.Nm bus_dma_template_init ,
64.Nm bus_dma_template_tag ,
65.Nm bus_dma_template_clone ,
66.Nm bus_dma_template_fill ,
67.Nm BUS_DMA_TEMPLATE_FILL ,
68.Nm bus_dmamap_create ,
69.Nm bus_dmamap_destroy ,
70.Nm bus_dmamap_load ,
71.Nm bus_dmamap_load_bio ,
72.Nm bus_dmamap_load_ccb ,
73.Nm bus_dmamap_load_crp ,
74.Nm bus_dmamap_load_crp_buffer ,
75.Nm bus_dmamap_load_mbuf ,
76.Nm bus_dmamap_load_mbuf_sg ,
77.Nm bus_dmamap_load_uio ,
78.Nm bus_dmamap_unload ,
79.Nm bus_dmamap_sync ,
80.Nm bus_dmamem_alloc ,
81.Nm bus_dmamem_free
82.Nd Bus and Machine Independent DMA Mapping Interface
83.Sh SYNOPSIS
84.In machine/bus.h
85.Ft int
86.Fn bus_dma_tag_create "bus_dma_tag_t parent" "bus_size_t alignment" \
87"bus_addr_t boundary" "bus_addr_t lowaddr" "bus_addr_t highaddr" \
88"bus_dma_filter_t *filtfunc" "void *filtfuncarg" "bus_size_t maxsize" \
89"int nsegments" "bus_size_t maxsegsz" "int flags" "bus_dma_lock_t *lockfunc" \
90"void *lockfuncarg" "bus_dma_tag_t *dmat"
91.Ft int
92.Fn bus_dma_tag_destroy "bus_dma_tag_t dmat"
93.Ft void
94.Fo bus_dma_template_init
95.Fa "bus_dma_template_t *template"
96.Fa "bus_dma_tag_t parent"
97.Fc
98.Ft int
99.Fo bus_dma_template_tag
100.Fa "bus_dma_template_t *template"
101.Fa "bus_dma_tag_t *dmat"
102.Fc
103.Ft void
104.Fo bus_dma_template_clone
105.Fa "bus_dma_template_t *template"
106.Fa "bus_dma_tag_t dmat"
107.Fc
108.Ft void
109.Fo bus_dma_template_fill
110.Fa "bus_dma_template_t *template"
111.Fa "bus_dma_param_t params[]"
112.Fa "u_int count"
113.Fc
114.Fo BUS_DMA_TEMPLATE_FILL
115.Fa "bus_dma_template_t *template"
116.Fa "bus_dma_param_t param ..."
117.Fc
118.Ft int
119.Fn bus_dmamap_create "bus_dma_tag_t dmat" "int flags" "bus_dmamap_t *mapp"
120.Ft int
121.Fn bus_dmamap_destroy "bus_dma_tag_t dmat" "bus_dmamap_t map"
122.Ft int
123.Fn bus_dmamap_load "bus_dma_tag_t dmat" "bus_dmamap_t map" "void *buf" \
124"bus_size_t buflen" "bus_dmamap_callback_t *callback" "void *callback_arg" \
125"int flags"
126.Ft int
127.Fn bus_dmamap_load_bio "bus_dma_tag_t dmat" "bus_dmamap_t map" \
128"struct bio *bio" "bus_dmamap_callback_t *callback" "void *callback_arg" \
129"int flags"
130.Ft int
131.Fn bus_dmamap_load_ccb "bus_dma_tag_t dmat" "bus_dmamap_t map" \
132"union ccb *ccb" "bus_dmamap_callback_t *callback" "void *callback_arg" \
133"int flags"
134.Ft int
135.Fn bus_dmamap_load_crp "bus_dma_tag_t dmat" "bus_dmamap_t map" \
136"struct crypto *crp" "bus_dmamap_callback_t *callback" "void *callback_arg" \
137"int flags"
138.Ft int
139.Fn bus_dmamap_load_crp_buffer "bus_dma_tag_t dmat" "bus_dmamap_t map" \
140"struct crypto_buffer *cb" "bus_dmamap_callback_t *callback" \
141"void *callback_arg" "int flags"
142.Ft int
143.Fn bus_dmamap_load_mbuf "bus_dma_tag_t dmat" "bus_dmamap_t map" \
144"struct mbuf *mbuf" "bus_dmamap_callback2_t *callback" "void *callback_arg" \
145"int flags"
146.Ft int
147.Fn bus_dmamap_load_mbuf_sg "bus_dma_tag_t dmat" "bus_dmamap_t map" \
148"struct mbuf *mbuf" "bus_dma_segment_t *segs" "int *nsegs" "int flags"
149.Ft int
150.Fn bus_dmamap_load_uio "bus_dma_tag_t dmat" "bus_dmamap_t map" \
151"struct uio *uio" "bus_dmamap_callback2_t *callback" "void *callback_arg" \
152"int flags"
153.Ft void
154.Fn bus_dmamap_unload "bus_dma_tag_t dmat" "bus_dmamap_t map"
155.Ft void
156.Fn bus_dmamap_sync "bus_dma_tag_t dmat" "bus_dmamap_t map" \
157"op"
158.Ft int
159.Fn bus_dmamem_alloc "bus_dma_tag_t dmat" "void **vaddr" \
160"int flags" "bus_dmamap_t *mapp"
161.Ft void
162.Fn bus_dmamem_free "bus_dma_tag_t dmat" "void *vaddr" \
163"bus_dmamap_t map"
164.Sh DESCRIPTION
165Direct Memory Access (DMA) is a method of transferring data
166without involving the CPU, thus providing higher performance.
167A DMA transaction can be achieved between device to memory,
168device to device, or memory to memory.
169.Pp
170The
171.Nm
172API is a bus, device, and machine-independent (MI) interface to
173DMA mechanisms.
174It provides the client with flexibility and simplicity by
175abstracting machine dependent issues like setting up
176DMA mappings, handling cache issues, bus specific features
177and limitations.
178.Sh OVERVIEW
179A tag structure
180.Vt ( bus_dma_tag_t )
181is used to describe the properties of a group of related DMA
182transactions.
183One way to view this is that a tag describes the limitations of a DMA engine.
184For example, if a DMA engine in a device is limited to 32-bit addresses,
185that limitation is specified by a parameter when creating the tag
186for that device.
187Similarly, a tag can be marked as requiring buffers whose addresses are
188aligned to a specific boundary.
189.Pp
190Some devices may require multiple tags to describe DMA
191transactions with differing properties.
192For example, a device might require 16-byte alignment of its descriptor ring
193while permitting arbitrary alignment of I/O buffers.
194In this case,
195the driver must create one tag for the descriptor ring and a separate tag for
196I/O buffers.
197If a device has restrictions that are common to all DMA transactions
198in addition to restrictions that differ between unrelated groups of
199transactions,
200the driver can first create a
201.Dq parent
202tag that decribes the common restrictions.
203The per-group tags can then inherit these restrictions from this
204.Dq parent
205tag rather than having to list them explicitly when creating the per-group tags.
206.Pp
207A mapping structure
208.Vt ( bus_dmamap_t )
209represents a mapping of a memory region for DMA.
210On systems with I/O MMUs,
211the mapping structure tracks any I/O MMU entries used by a request.
212For DMA requests that require bounce pages,
213the mapping tracks the bounce pages used.
214.Pp
215To prepare for one or more DMA transactions,
216a mapping must be bound to a memory region by calling one of the
217.Fn bus_dmamap_load
218functions.
219These functions configure the mapping which can include programming entries
220in an I/O MMU and/or allocating bounce pages.
221An output of these functions
222(either directly or indirectly by invoking a callback routine)
223is the list of scatter/gather address ranges a consumer can pass to a DMA
224engine to access the memory region.
225When a mapping is no longer needed,
226the mapping must be unloaded via
227.Fn bus_dmamap_unload .
228.Pp
229Before and after each DMA transaction,
230.Fn bus_dmamap_sync
231must be used to ensure that the correct data is used by the DMA engine and
232the CPU.
233If a mapping uses bounce pages,
234the sync operations copy data between the bounce pages and the memory region
235bound to the mapping.
236Sync operations also handle architecture-specific details such as CPU cache
237flushing and CPU memory operation ordering.
238.Sh STATIC VS DYNAMIC
239.Nm
240handles two types of DMA transactions: static and dynamic.
241Static transactions are used with a long-lived memory region that is reused
242for many transactions such as a descriptor ring.
243Dynamic transactions are used for transfers to or from transient buffers
244such as I/O buffers holding a network packet or disk block.
245Each transaction type uses a different subset of the
246.Nm
247API.
248.Ss Static Transactions
249Static transactions use memory regions allocated by
250.Nm .
251Each static memory region is allocated by calling
252.Fn bus_dmamem_alloc .
253This function requires a valid tag describing the properties of the
254DMA transactions to this region such as alignment or address restrictions.
255Multiple regions can share a single tag if they share the same restrictions.
256.Pp
257.Fn bus_dmamem_alloc
258allocates a memory region along with a mapping object.
259The associated tag, memory region, and mapping object must then be passed to
260.Fn bus_dmamap_load
261to bind the mapping to the allocated region and obtain the
262scatter/gather list.
263.Pp
264It is expected that
265.Fn bus_dmamem_alloc
266will attempt to allocate memory requiring less expensive sync operations
267(for example, implementations should not allocate regions requiring bounce
268pages),
269but sync operations should still be used.
270For example, a driver should use
271.Fn bus_dmamap_sync
272in an interrupt handler before reading descriptor ring entries written by the
273device prior to the interrupt.
274.Pp
275When a consumer is finished with a memory region,
276it should unload the mapping via
277.Fn bus_dmamap_unload
278and then release the memory region and mapping object via
279.Fn bus_dmamem_free .
280.Ss Dynamic Transactions
281Dynamic transactions map memory regions provided by other parts of the system.
282A tag must be created via
283.Fn bus_dma_tag_create
284to describe the DMA transactions to and from these memory regions,
285and a pool of mapping objects must be allocated via
286.Fn bus_dmamap_create
287to track the mappings of any in-flight transactions.
288.Pp
289When a consumer wishes to schedule a transaction for a memory region,
290the consumer must first obtain an unused mapping object from its pool
291of mapping objects.
292The memory region must be bound to the mapping object via one of the
293.Fn bus_dmamap_load
294functions.
295Before scheduling the transaction,
296the consumer should sync the memory region via
297.Fn bus_dmamap_sync
298with one or more of the
299.Dq PRE
300flags.
301After the transaction has completed,
302the consumer should sync the memory region via
303.Fn bus_dmamap_sync
304with one or more of the
305.Dq POST
306flags.
307The mapping can then be unloaded via
308.Fn bus_dmamap_unload ,
309and the mapping object can be returned to the pool of unused mapping objects.
310.Pp
311When a consumer is no longer scheduling DMA transactions,
312the mapping objects should be freed via
313.Fn bus_dmamap_destroy ,
314and the tag should be freed via
315.Fn bus_dma_tag_destroy .
316.Sh STRUCTURES AND TYPES
317.Bl -tag -width indent
318.It Vt bus_dma_tag_t
319A machine-dependent (MD) opaque type that describes the
320characteristics of a group of DMA transactions.
321DMA tags are organized into a hierarchy, with each child
322tag inheriting the restrictions of its parent.
323This allows all devices along the path of DMA transactions
324to contribute to the constraints of those transactions.
325.It Vt bus_dma_template_t
326A template is a structure for creating a
327.Fa bus_dma_tag_t
328from a set of defaults.
329Once initialized with
330.Fn bus_dma_template_init ,
331a driver can over-ride individual fields to suit its needs.
332The following fields start with the indicated default values:
333.Bd -literal
334	alignment	1
335	boundary	0
336	lowaddr		BUS_SPACE_MAXADDR
337	highaddr	BUS_SPACE_MAXADDR
338	maxsize		BUS_SPACE_MAXSIZE
339	nsegments	BUS_SPACE_UNRESTRICTED
340	maxsegsize	BUS_SPACE_MAXSIZE
341	flags		0
342	lockfunc	NULL
343	lockfuncarg	NULL
344.Ed
345.Pp
346Descriptions of each field are documented with
347.Fn bus_dma_tag_create .
348Note that the
349.Fa filtfunc
350and
351.Fa filtfuncarg
352attributes of the DMA tag are not supported with templates.
353.It Vt bus_dma_filter_t
354Client specified address filter having the format:
355.Bl -tag -width indent
356.It Ft int
357.Fn "client_filter" "void *filtarg" "bus_addr_t testaddr"
358.El
359.Pp
360Address filters can be specified during tag creation to allow
361for devices whose DMA address restrictions cannot be specified
362by a single window.
363The
364.Fa filtarg
365argument is specified by the client during tag creation to be passed to all
366invocations of the callback.
367The
368.Fa testaddr
369argument contains a potential starting address of a DMA mapping.
370The filter function operates on the set of addresses from
371.Fa testaddr
372to
373.Ql trunc_page(testaddr) + PAGE_SIZE - 1 ,
374inclusive.
375The filter function should return zero if any mapping in this range
376can be accommodated by the device and non-zero otherwise.
377.Pp
378.Em Note: The use of filters is deprecated.  Proper operation is not guaranteed.
379.It Vt bus_dma_segment_t
380A machine-dependent type that describes individual
381DMA segments.
382It contains the following fields:
383.Bd -literal
384	bus_addr_t	ds_addr;
385	bus_size_t	ds_len;
386.Ed
387.Pp
388The
389.Fa ds_addr
390field contains the device visible address of the DMA segment, and
391.Fa ds_len
392contains the length of the DMA segment.
393Although the DMA segments returned by a mapping call will adhere to
394all restrictions necessary for a successful DMA operation, some conversion
395(e.g.\& a conversion from host byte order to the device's byte order) is
396almost always required when presenting segment information to the device.
397.It Vt bus_dmamap_t
398A machine-dependent opaque type describing an individual mapping.
399One map is used for each memory allocation that will be loaded.
400Maps can be reused once they have been unloaded.
401Multiple maps can be associated with one DMA tag.
402While the value of the map may evaluate to
403.Dv NULL
404on some platforms under certain conditions,
405it should never be assumed that it will be
406.Dv NULL
407in all cases.
408.It Vt bus_dmamap_callback_t
409Client specified callback for receiving mapping information resulting from
410the load of a
411.Vt bus_dmamap_t
412via
413.Fn bus_dmamap_load ,
414.Fn bus_dmamap_load_bio ,
415.Fn bus_dmamap_load_ccb ,
416.Fn bus_dmamap_load_crp ,
417or
418.Fn bus_dmamap_load_crp_buffer .
419Callbacks are of the format:
420.Bl -tag -width indent
421.It Ft void
422.Fn "client_callback" "void *callback_arg" "bus_dma_segment_t *segs" \
423"int nseg" "int error"
424.El
425.Pp
426The
427.Fa callback_arg
428is the callback argument passed to dmamap load functions.
429The
430.Fa segs
431and
432.Fa nseg
433arguments describe an array of
434.Vt bus_dma_segment_t
435structures that represent the mapping.
436This array is only valid within the scope of the callback function.
437The success or failure of the mapping is indicated by the
438.Fa error
439argument.
440More information on the use of callbacks can be found in the
441description of the individual dmamap load functions.
442.It Vt bus_dmamap_callback2_t
443Client specified callback for receiving mapping information resulting from
444the load of a
445.Vt bus_dmamap_t
446via
447.Fn bus_dmamap_load_uio
448or
449.Fn bus_dmamap_load_mbuf .
450.Pp
451Callback2s are of the format:
452.Bl -tag -width indent
453.It Ft void
454.Fn "client_callback2" "void *callback_arg" "bus_dma_segment_t *segs" \
455"int nseg" "bus_size_t mapsize" "int error"
456.El
457.Pp
458Callback2's behavior is the same as
459.Vt bus_dmamap_callback_t
460with the addition that the length of the data mapped is provided via
461.Fa mapsize .
462.It Vt bus_dmasync_op_t
463Memory synchronization operation specifier.
464Bus DMA requires explicit synchronization of memory with its device
465visible mapping in order to guarantee memory coherency.
466The
467.Vt bus_dmasync_op_t
468allows the type of DMA operation that will be or has been performed
469to be communicated to the system so that the correct coherency measures
470are taken.
471The operations are represented as bitfield flags that can be combined together,
472though it only makes sense to combine PRE flags or POST flags, not both.
473See the
474.Fn bus_dmamap_sync
475description below for more details on how to use these operations.
476.Pp
477All operations specified below are performed from the host memory point of view,
478where a read implies data coming from the device to the host memory, and a write
479implies data going from the host memory to the device.
480Alternatively, the operations can be thought of in terms of driver operations,
481where reading a network packet or storage sector corresponds to a read operation
482in
483.Nm .
484.Bl -tag -width ".Dv BUS_DMASYNC_POSTWRITE"
485.It Dv BUS_DMASYNC_PREREAD
486Perform any synchronization required prior to an update of host memory by the
487device.
488.It Dv BUS_DMASYNC_PREWRITE
489Perform any synchronization required after an update of host memory by the CPU
490and prior to device access to host memory.
491.It Dv BUS_DMASYNC_POSTREAD
492Perform any synchronization required after an update of host memory by the
493device and prior to CPU access to host memory.
494.It Dv BUS_DMASYNC_POSTWRITE
495Perform any synchronization required after device access to host memory.
496.El
497.It Vt bus_dma_lock_t
498Client specified lock/mutex manipulation method.
499This will be called from
500within busdma whenever a client lock needs to be manipulated.
501In its current form, the function will be called immediately before
502the callback for a DMA load operation that has been deferred with
503.Dv BUS_DMA_LOCK
504and immediately after with
505.Dv BUS_DMA_UNLOCK .
506If the load operation does not need to be deferred, then it
507will not be called since the function loading the map should
508be holding the appropriate locks.
509This method is of the format:
510.Bl -tag -width indent
511.It Ft void
512.Fn "lockfunc" "void *lockfunc_arg" "bus_dma_lock_op_t op"
513.El
514.Pp
515The
516.Fa lockfuncarg
517argument is specified by the client during tag creation to be passed to all
518invocations of the callback.
519The
520.Fa op
521argument specifies the lock operation to perform.
522.Pp
523Two
524.Vt lockfunc
525implementations are provided for convenience.
526.Fn busdma_lock_mutex
527performs standard mutex operations on the sleep mutex provided via
528.Fa lockfuncarg .
529.Fn dflt_lock
530will generate a system panic if it is called.
531It is substituted into the tag when
532.Fa lockfunc
533is passed as
534.Dv NULL
535to
536.Fn bus_dma_tag_create
537and is useful for tags that should not be used with deferred load operations.
538.It Vt bus_dma_lock_op_t
539Operations to be performed by the client-specified
540.Fn lockfunc .
541.Bl -tag -width ".Dv BUS_DMA_UNLOCK"
542.It Dv BUS_DMA_LOCK
543Acquires and/or locks the client locking primitive.
544.It Dv BUS_DMA_UNLOCK
545Releases and/or unlocks the client locking primitive.
546.El
547.El
548.Sh FUNCTIONS
549.Bl -tag -width indent
550.It Fn bus_dma_tag_create "parent" "alignment" "boundary" "lowaddr" \
551"highaddr" "*filtfunc" "*filtfuncarg" "maxsize" "nsegments" "maxsegsz" \
552"flags" "lockfunc" "lockfuncarg" "*dmat"
553Allocates a DMA tag, and initializes it according to
554the arguments provided:
555.Bl -tag -width ".Fa filtfuncarg"
556.It Fa parent
557A parent tag from which to inherit restrictions.
558The restrictions passed in other arguments can only further tighten the
559restrictions inherited from the parent tag.
560.Pp
561All tags created by a device driver must inherit from the tag returned by
562.Fn bus_get_dma_tag
563to honor restrictions between the parent bridge, CPU memory, and the
564device.
565.It Fa alignment
566Alignment constraint, in bytes, of any mappings created using this tag.
567The alignment must be a power of 2.
568Hardware that can DMA starting at any address would specify
569.Em 1
570for byte alignment.
571Hardware requiring DMA transfers to start on a multiple of 4K
572would specify
573.Em 4096 .
574.It Fa boundary
575Boundary constraint, in bytes, of the target DMA memory region.
576The boundary indicates the set of addresses, all multiples of the
577boundary argument, that cannot be crossed by a single
578.Vt bus_dma_segment_t .
579The boundary must be a power of 2 and must be no smaller than the
580maximum segment size.
581.Ql 0
582indicates that there are no boundary restrictions.
583.It Fa lowaddr , highaddr
584Bounds of the window of bus address space that
585.Em cannot
586be directly accessed by the device.
587The window contains all addresses greater than
588.Fa lowaddr
589and less than or equal to
590.Fa highaddr .
591For example, a device incapable of DMA above 4GB, would specify a
592.Fa highaddr
593of
594.Dv BUS_SPACE_MAXADDR
595and a
596.Fa lowaddr
597of
598.Dv BUS_SPACE_MAXADDR_32BIT .
599Similarly a device that can only perform DMA to addresses below
60016MB would specify a
601.Fa highaddr
602of
603.Dv BUS_SPACE_MAXADDR
604and a
605.Fa lowaddr
606of
607.Dv BUS_SPACE_MAXADDR_24BIT .
608Some implementations require that some region of device visible
609address space, overlapping available host memory, be outside the
610window.
611This area of
612.Ql safe memory
613is used to bounce requests that would otherwise conflict with
614the exclusion window.
615.It Fa filtfunc
616Optional filter function (may be
617.Dv NULL )
618to be called for any attempt to
619map memory into the window described by
620.Fa lowaddr
621and
622.Fa highaddr .
623A filter function is only required when the single window described
624by
625.Fa lowaddr
626and
627.Fa highaddr
628cannot adequately describe the constraints of the device.
629The filter function will be called for every machine page
630that overlaps the exclusion window.
631.Pp
632.Em Note: The use of filters is deprecated.  Proper operation is not guaranteed.
633.It Fa filtfuncarg
634Argument passed to all calls to the filter function for this tag.
635May be
636.Dv NULL .
637.It Fa maxsize
638Maximum size, in bytes, of the sum of all segment lengths in a given
639DMA mapping associated with this tag.
640.It Fa nsegments
641Number of discontinuities (scatter/gather segments) allowed
642in a DMA mapped region.
643.It Fa maxsegsz
644Maximum size, in bytes, of a segment in any DMA mapped region associated
645with
646.Fa dmat .
647.It Fa flags
648Are as follows:
649.Bl -tag -width ".Dv BUS_DMA_ALLOCNOW"
650.It Dv BUS_DMA_ALLOCNOW
651Pre-allocate enough resources to handle at least one map load operation on
652this tag.
653If sufficient resources are not available,
654.Er ENOMEM
655is returned.
656This should not be used for tags that only describe buffers that will be
657allocated with
658.Fn bus_dmamem_alloc .
659Also, due to resource sharing with other tags, this flag does not guarantee
660that resources will be allocated or reserved exclusively for this tag.
661It should be treated only as a minor optimization.
662.It Dv BUS_DMA_COHERENT
663Indicate that the DMA engine and CPU are cache-coherent.
664Cached memory may be used to back allocations created by
665.Fn bus_dmamem_alloc .
666For
667.Fn bus_dma_tag_create ,
668the
669.Dv BUS_DMA_COHERENT
670flag is currently implemented on arm64.
671.El
672.It Fa lockfunc
673Optional lock manipulation function (may be
674.Dv NULL )
675to be called when busdma
676needs to manipulate a lock on behalf of the client.
677If
678.Dv NULL
679is specified,
680.Fn dflt_lock
681is used.
682.It Fa lockfuncarg
683Optional argument to be passed to the function specified by
684.Fa lockfunc .
685.It Fa dmat
686Pointer to a bus_dma_tag_t where the resulting DMA tag will
687be stored.
688.El
689.Pp
690Returns
691.Er ENOMEM
692if sufficient memory is not available for tag creation
693or allocating mapping resources.
694.It Fn bus_dma_tag_destroy "dmat"
695Deallocate the DMA tag
696.Fa dmat
697that was created by
698.Fn bus_dma_tag_create .
699.Pp
700Returns
701.Er EBUSY
702if any DMA maps remain associated with
703.Fa dmat
704or
705.Ql 0
706on success.
707.It Fn bus_dma_template_init "*template" "parent"
708Initializes a
709.Fa bus_dma_template_t
710structure.
711If the
712.Fa parent
713argument is non-NULL, this parent tag is associated with the template and
714will be compiled into the dma tag that is later created.
715The values of the parent are not copied into the template.
716During tag creation in
717.Fn bus_dma_tag_template ,
718any parameters from the parent tag that are more restrictive than what is
719in the provided template will overwrite what goes into the new tag.
720.It Fn bus_dma_template_tag "*template" "*dmat"
721Unpacks a template into a tag, and returns the tag via the
722.Fa dmat .
723All return values are identical to
724.Fn bus_dma_tag_create .
725The template is not modified by this function, and can be reused and/or
726freed upon return.
727.It Fn bus_dma_template_clone "*template" "dmat"
728Copies the fields from an existing tag to a template.
729The template does not need to be initialized first.
730All of its fields will be overwritten by the values contained in the tag.
731When paired with
732.Fn bus_dma_template_tag ,
733this function is useful for creating copies of tags.
734.It Fn bus_dma_template_fill "*template" "params[]" "count"
735Fills in the selected fields of the template with the keyed values from the
736.Fa params
737array.
738This is not meant to be called directly, use
739.Fn BUS_DMA_TEMPLATE_FILL
740instead.
741.It Fn BUS_DMA_TEMPLATE_FILL "*template" "param ..."
742Fills in the selected fields of the template with a variable number of
743key-value parameters.
744The macros listed below take an argument of the specified type and encapsulate
745it into a key-value structure that is directly usable as a parameter argument.
746Muliple parameters may be provided at once.
747.Bd -literal
748	BD_PARENT()	void *
749	BD_ALIGNMENT()	uintmax_t
750	BD_BOUNDARY()	uintmax_t
751	BD_LOWADDR()	vm_paddr_t
752	BD_HIGHADDR()	vm_paddr_t
753	BD_MAXSIZE()	uintmax_t
754	BD_NSEGMENTS()	uintmax_t
755	BD_MAXSEGSIZE()	uintmax_t
756	BD_FLAGS()	uintmax_t
757	BD_LOCKFUNC()	void *
758	BD_LOCKFUNCARG() void *
759.Ed
760.It Fn bus_dmamap_create "dmat" "flags" "*mapp"
761Allocates and initializes a DMA map.
762Arguments are as follows:
763.Bl -tag -width ".Fa nsegments"
764.It Fa dmat
765DMA tag.
766.It Fa flags
767Are as follows:
768.Bl -tag -width ".Dv BUS_DMA_COHERENT"
769.It Dv BUS_DMA_COHERENT
770Attempt to map the memory loaded with this map such that cache sync
771operations are as cheap as possible.
772This flag is typically set on maps when the memory loaded with these will
773be accessed by both a CPU and a DMA engine, frequently such as control data
774and as opposed to streamable data such as receive and transmit buffers.
775Use of this flag does not remove the requirement of using
776.Fn bus_dmamap_sync ,
777but it may reduce the cost of performing these operations.
778.El
779.It Fa mapp
780Pointer to a
781.Vt bus_dmamap_t
782where the resulting DMA map will be stored.
783.El
784.Pp
785Returns
786.Er ENOMEM
787if sufficient memory is not available for creating the
788map or allocating mapping resources.
789.It Fn bus_dmamap_destroy "dmat" "map"
790Frees all resources associated with a given DMA map.
791Arguments are as follows:
792.Bl -tag -width ".Fa dmat"
793.It Fa dmat
794DMA tag used to allocate
795.Fa map .
796.It Fa map
797The DMA map to destroy.
798.El
799.Pp
800Returns
801.Er EBUSY
802if a mapping is still active for
803.Fa map .
804.It Fn bus_dmamap_load "dmat" "map" "buf" "buflen" "*callback" \
805"callback_arg" "flags"
806Creates a mapping in device visible address space of
807.Fa buflen
808bytes of
809.Fa buf ,
810associated with the DMA map
811.Fa map .
812This call will always return immediately and will not block for any reason.
813Arguments are as follows:
814.Bl -tag -width ".Fa buflen"
815.It Fa dmat
816DMA tag used to allocate
817.Fa map .
818.It Fa map
819A DMA map without a currently active mapping.
820.It Fa buf
821A kernel virtual address pointer to a contiguous (in KVA) buffer, to be
822mapped into device visible address space.
823.It Fa buflen
824The size of the buffer.
825.It Fa callback Fa callback_arg
826The callback function, and its argument.
827This function is called once sufficient mapping resources are available for
828the DMA operation.
829If resources are temporarily unavailable, this function will be deferred until
830later, but the load operation will still return immediately to the caller.
831Thus, callers should not assume that the callback will be called before the
832load returns, and code should be structured appropriately to handle this.
833See below for specific flags and error codes that control this behavior.
834.It Fa flags
835Are as follows:
836.Bl -tag -width ".Dv BUS_DMA_NOWAIT"
837.It Dv BUS_DMA_NOWAIT
838The load should not be deferred in case of insufficient mapping resources,
839and instead should return immediately with an appropriate error.
840.It Dv BUS_DMA_NOCACHE
841The generated transactions to and from the virtual page are non-cacheable.
842.El
843.El
844.Pp
845Return values to the caller are as follows:
846.Bl -tag -width ".Er EINPROGRESS"
847.It 0
848The callback has been called and completed.
849The status of the mapping has been delivered to the callback.
850.It Er EINPROGRESS
851The mapping has been deferred for lack of resources.
852The callback will be called as soon as resources are available.
853Callbacks are serviced in FIFO order.
854.Pp
855Note that subsequent load operations for the same tag that do not require
856extra resources will still succeed.
857This may result in out-of-order processing of requests.
858If the caller requires the order of requests to be preserved,
859then the caller is required to stall subsequent requests until a pending
860request's callback is invoked.
861.It Er ENOMEM
862The load request has failed due to insufficient resources, and the caller
863specifically used the
864.Dv BUS_DMA_NOWAIT
865flag.
866.It Er EINVAL
867The load request was invalid.
868The callback has been called and has been provided the same error.
869This error value may indicate that
870.Fa dmat ,
871.Fa map ,
872.Fa buf ,
873or
874.Fa callback
875were invalid, or
876.Fa buflen
877was larger than the
878.Fa maxsize
879argument used to create the dma tag
880.Fa dmat .
881.El
882.Pp
883When the callback is called, it is presented with an error value
884indicating the disposition of the mapping.
885Error may be one of the following:
886.Bl -tag -width ".Er EINPROGRESS"
887.It 0
888The mapping was successful and the
889.Fa dm_segs
890callback argument contains an array of
891.Vt bus_dma_segment_t
892elements describing the mapping.
893This array is only valid during the scope of the callback function.
894.It Er EFBIG
895A mapping could not be achieved within the segment constraints provided
896in the tag even though the requested allocation size was less than maxsize.
897.El
898.It Fn bus_dmamap_load_bio "dmat" "map" "bio" "callback" "callback_arg" "flags"
899This is a variation of
900.Fn bus_dmamap_load
901which maps buffers pointed to by
902.Fa bio
903for DMA transfers.
904.Fa bio
905may point to either a mapped or unmapped buffer.
906.It Fn bus_dmamap_load_ccb "dmat" "map" "ccb" "callback" "callback_arg" "flags"
907This is a variation of
908.Fn bus_dmamap_load
909which maps data pointed to by
910.Fa ccb
911for DMA transfers.
912The data for
913.Fa ccb
914may be any of the following types:
915.Bl -tag -width ".Er CAM_DATA_SG_PADDR"
916.It CAM_DATA_VADDR
917The data is a single KVA buffer.
918.It CAM_DATA_PADDR
919The data is a single bus address range.
920.It CAM_DATA_SG
921The data is a scatter/gather list of KVA buffers.
922.It CAM_DATA_SG_PADDR
923The data is a scatter/gather list of bus address ranges.
924.It CAM_DATA_BIO
925The data is contained in a
926.Vt struct bio
927attached to the CCB.
928.El
929.Pp
930.Fn bus_dmamap_load_ccb
931supports the following CCB XPT function codes:
932.Pp
933.Bl -item -offset indent -compact
934.It
935XPT_ATA_IO
936.It
937XPT_CONT_TARGET_IO
938.It
939XPT_SCSI_IO
940.El
941.It Fn bus_dmamap_load_crp "dmat" "map" "crp" "callback" "callback_arg" "flags"
942This is a variation of
943.Fn bus_dmamap_load
944which maps the input buffer pointed to by
945.Fa crp
946for DMA transfers.
947The
948.Dv BUS_DMA_NOWAIT
949flag is implied, thus no callback deferral will happen.
950.It Fn bus_dmamap_load_crp_buffer "dmat" "map" "cb" "callback" "callback_arg" \
951"flags"
952This is a variation of
953.Fn bus_dmamap_load
954which maps the crypto data buffer pointed to by
955.Fa cb
956for DMA transfers.
957The
958.Dv BUS_DMA_NOWAIT
959flag is implied, thus no callback deferral will happen.
960.It Fn bus_dmamap_load_mbuf "dmat" "map" "mbuf" "callback2" "callback_arg" \
961"flags"
962This is a variation of
963.Fn bus_dmamap_load
964which maps mbuf chains
965for DMA transfers.
966A
967.Vt bus_size_t
968argument is also passed to the callback routine, which
969contains the mbuf chain's packet header length.
970The
971.Dv BUS_DMA_NOWAIT
972flag is implied, thus no callback deferral will happen.
973.Pp
974Mbuf chains are assumed to be in kernel virtual address space.
975.Pp
976Beside the error values listed for
977.Fn bus_dmamap_load ,
978.Er EINVAL
979will be returned if the size of the mbuf chain exceeds the maximum limit of the
980DMA tag.
981.It Fn bus_dmamap_load_mbuf_sg "dmat" "map" "mbuf" "segs" "nsegs" "flags"
982This is just like
983.Fn bus_dmamap_load_mbuf
984except that it returns immediately without calling a callback function.
985It is provided for efficiency.
986The scatter/gather segment array
987.Va segs
988is provided by the caller and filled in directly by the function.
989The
990.Va nsegs
991argument is returned with the number of segments filled in.
992Returns the same errors as
993.Fn bus_dmamap_load_mbuf .
994.It Fn bus_dmamap_load_uio "dmat" "map" "uio" "callback2" "callback_arg" "flags"
995This is a variation of
996.Fn bus_dmamap_load
997which maps buffers pointed to by
998.Fa uio
999for DMA transfers.
1000A
1001.Vt bus_size_t
1002argument is also passed to the callback routine, which contains the size of
1003.Fa uio ,
1004i.e.
1005.Fa uio->uio_resid .
1006The
1007.Dv BUS_DMA_NOWAIT
1008flag is implied, thus no callback deferral will happen.
1009Returns the same errors as
1010.Fn bus_dmamap_load .
1011.Pp
1012If
1013.Fa uio->uio_segflg
1014is
1015.Dv UIO_USERSPACE ,
1016then it is assumed that the buffer,
1017.Fa uio
1018is in
1019.Fa "uio->uio_td->td_proc" Ns 's
1020address space.
1021User space memory must be in-core and wired prior to attempting a map
1022load operation.
1023Pages may be locked using
1024.Xr vslock 9 .
1025.It Fn bus_dmamap_unload "dmat" "map"
1026Unloads a DMA map.
1027Arguments are as follows:
1028.Bl -tag -width ".Fa dmam"
1029.It Fa dmat
1030DMA tag used to allocate
1031.Fa map .
1032.It Fa map
1033The DMA map that is to be unloaded.
1034.El
1035.Pp
1036.Fn bus_dmamap_unload
1037will not perform any implicit synchronization of DMA buffers.
1038This must be done explicitly by a call to
1039.Fn bus_dmamap_sync
1040prior to unloading the map.
1041.It Fn bus_dmamap_sync "dmat" "map" "op"
1042Performs synchronization of a device visible mapping with the CPU visible
1043memory referenced by that mapping.
1044Arguments are as follows:
1045.Bl -tag -width ".Fa dmat"
1046.It Fa dmat
1047DMA tag used to allocate
1048.Fa map .
1049.It Fa map
1050The DMA mapping to be synchronized.
1051.It Fa op
1052Type of synchronization operation to perform.
1053See the definition of
1054.Vt bus_dmasync_op_t
1055for a description of the acceptable values for
1056.Fa op .
1057.El
1058.Pp
1059The
1060.Fn bus_dmamap_sync
1061function
1062is the method used to ensure that CPU's and device's direct
1063memory access (DMA) to shared
1064memory is coherent.
1065For example, the CPU might be used to set up the contents of a buffer
1066that is to be made available to a device.
1067To ensure that the data are visible via the device's mapping of that
1068memory, the buffer must be loaded and a DMA sync operation of
1069.Dv BUS_DMASYNC_PREWRITE
1070must be performed after the CPU has updated the buffer and before the device
1071access is initiated.
1072If the CPU modifies this buffer again later, another
1073.Dv BUS_DMASYNC_PREWRITE
1074sync operation must be performed before an additional device
1075access.
1076Conversely, suppose a device updates memory that is to be read by a CPU.
1077In this case, the buffer must be loaded, and a DMA sync operation of
1078.Dv BUS_DMASYNC_PREREAD
1079must be performed before the device access is initiated.
1080The CPU will only be able to see the results of this memory update
1081once the DMA operation has completed and a
1082.Dv BUS_DMASYNC_POSTREAD
1083sync operation has been performed.
1084.Pp
1085If read and write operations are not preceded and followed by the
1086appropriate synchronization operations, behavior is undefined.
1087.It Fn bus_dmamem_alloc "dmat" "**vaddr" "flags" "*mapp"
1088Allocates memory that is mapped into KVA at the address returned
1089in
1090.Fa vaddr
1091and that is permanently loaded into the newly created
1092.Vt bus_dmamap_t
1093returned via
1094.Fa mapp .
1095Arguments are as follows:
1096.Bl -tag -width ".Fa alignment"
1097.It Fa dmat
1098DMA tag describing the constraints of the DMA mapping.
1099.It Fa vaddr
1100Pointer to a pointer that will hold the returned KVA mapping of
1101the allocated region.
1102.It Fa flags
1103Flags are defined as follows:
1104.Bl -tag -width ".Dv BUS_DMA_NOWAIT"
1105.It Dv BUS_DMA_WAITOK
1106The routine can safely wait (sleep) for resources.
1107.It Dv BUS_DMA_NOWAIT
1108The routine is not allowed to wait for resources.
1109If resources are not available,
1110.Dv ENOMEM
1111is returned.
1112.It Dv BUS_DMA_COHERENT
1113Attempt to map this memory in a coherent fashion.
1114See
1115.Fn bus_dmamap_create
1116above for a description of this flag.
1117For
1118.Fn bus_dmamem_alloc ,
1119the
1120.Dv BUS_DMA_COHERENT
1121flag is currently implemented on arm and arm64.
1122.It Dv BUS_DMA_ZERO
1123Causes the allocated memory to be set to all zeros.
1124.It Dv BUS_DMA_NOCACHE
1125The allocated memory will not be cached in the processor caches.
1126All memory accesses appear on the bus and are executed
1127without reordering.
1128For
1129.Fn bus_dmamem_alloc ,
1130the
1131.Dv BUS_DMA_NOCACHE
1132flag is currently implemented on amd64 and i386 where it results in the
1133Strong Uncacheable PAT to be set for the allocated virtual address range.
1134.El
1135.It Fa mapp
1136Pointer to a
1137.Vt bus_dmamap_t
1138where the resulting DMA map will be stored.
1139.El
1140.Pp
1141The size of memory to be allocated is
1142.Fa maxsize
1143as specified in the call to
1144.Fn bus_dma_tag_create
1145for
1146.Fa dmat .
1147.Pp
1148The current implementation of
1149.Fn bus_dmamem_alloc
1150will allocate all requests as a single segment.
1151.Pp
1152An initial load operation is required to obtain the bus address of the allocated
1153memory, and an unload operation is required before freeing the memory, as
1154described below in
1155.Fn bus_dmamem_free .
1156Maps are automatically handled by this function and should not be explicitly
1157allocated or destroyed.
1158.Pp
1159Although an explicit load is not required for each access to the memory
1160referenced by the returned map, the synchronization requirements
1161as described in the
1162.Fn bus_dmamap_sync
1163section still apply and should be used to achieve portability on architectures
1164without coherent buses.
1165.Pp
1166Returns
1167.Er ENOMEM
1168if sufficient memory is not available for completing
1169the operation.
1170.It Fn bus_dmamem_free "dmat" "*vaddr" "map"
1171Frees memory previously allocated by
1172.Fn bus_dmamem_alloc .
1173Any mappings
1174will be invalidated.
1175Arguments are as follows:
1176.Bl -tag -width ".Fa vaddr"
1177.It Fa dmat
1178DMA tag.
1179.It Fa vaddr
1180Kernel virtual address of the memory.
1181.It Fa map
1182DMA map to be invalidated.
1183.El
1184.El
1185.Sh RETURN VALUES
1186Behavior is undefined if invalid arguments are passed to
1187any of the above functions.
1188If sufficient resources cannot be allocated for a given
1189transaction,
1190.Er ENOMEM
1191is returned.
1192All
1193routines that are not of type
1194.Vt void
1195will return 0 on success or an error
1196code on failure as discussed above.
1197.Pp
1198All
1199.Vt void
1200routines will succeed if provided with valid arguments.
1201.Sh LOCKING
1202Two locking protocols are used by
1203.Nm .
1204The first is a private global lock that is used to synchronize access to the
1205bounce buffer pool on the architectures that make use of them.
1206This lock is strictly a leaf lock that is only used internally to
1207.Nm
1208and is not exposed to clients of the API.
1209.Pp
1210The second protocol involves protecting various resources stored in the tag.
1211Since almost all
1212.Nm
1213operations are done through requests from the driver that created the tag,
1214the most efficient way to protect the tag resources is through the lock that
1215the driver uses.
1216In cases where
1217.Nm
1218acts on its own without being called by the driver, the lock primitive
1219specified in the tag is acquired and released automatically.
1220An example of this is when the
1221.Fn bus_dmamap_load
1222callback function is called from a deferred context instead of the driver
1223context.
1224This means that certain
1225.Nm
1226functions must always be called with the same lock held that is specified in the
1227tag.
1228These functions include:
1229.Pp
1230.Bl -item -offset indent -compact
1231.It
1232.Fn bus_dmamap_load
1233.It
1234.Fn bus_dmamap_load_bio
1235.It
1236.Fn bus_dmamap_load_ccb
1237.It
1238.Fn bus_dmamap_load_mbuf
1239.It
1240.Fn bus_dmamap_load_mbuf_sg
1241.It
1242.Fn bus_dmamap_load_uio
1243.It
1244.Fn bus_dmamap_unload
1245.It
1246.Fn bus_dmamap_sync
1247.El
1248.Pp
1249There is one exception to this rule.
1250It is common practice to call some of these functions during driver start-up
1251without any locks held.
1252So long as there is a guarantee of no possible concurrent use of the tag by
1253different threads during this operation, it is safe to not hold a lock for
1254these functions.
1255.Pp
1256Certain
1257.Nm
1258operations should not be called with the driver lock held, either because
1259they are already protected by an internal lock, or because they might sleep
1260due to memory or resource allocation.
1261The following functions must not be
1262called with any non-sleepable locks held:
1263.Pp
1264.Bl -item -offset indent -compact
1265.It
1266.Fn bus_dma_tag_create
1267.It
1268.Fn bus_dmamap_create
1269.It
1270.Fn bus_dmamem_alloc
1271.El
1272.Pp
1273All other functions do not have a locking protocol and can thus be
1274called with or without any system or driver locks held.
1275.Sh SEE ALSO
1276.Xr devclass 9 ,
1277.Xr device 9 ,
1278.Xr driver 9 ,
1279.Xr rman 9 ,
1280.Xr vslock 9
1281.Pp
1282.Rs
1283.%A "Jason R. Thorpe"
1284.%T "A Machine-Independent DMA Framework for NetBSD"
1285.%J "Proceedings of the Summer 1998 USENIX Technical Conference"
1286.%Q "USENIX Association"
1287.%D "June 1998"
1288.Re
1289.Sh HISTORY
1290The
1291.Nm
1292interface first appeared in
1293.Nx 1.3 .
1294.Pp
1295The
1296.Nm
1297API was adopted from
1298.Nx
1299for use in the CAM SCSI subsystem.
1300The alterations to the original API were aimed to remove the need for
1301a
1302.Vt bus_dma_segment_t
1303array stored in each
1304.Vt bus_dmamap_t
1305while allowing callers to queue up on scarce resources.
1306.Sh AUTHORS
1307The
1308.Nm
1309interface was designed and implemented by
1310.An Jason R. Thorpe
1311of the Numerical Aerospace Simulation Facility, NASA Ames Research Center.
1312Additional input on the
1313.Nm
1314design was provided by
1315.An -nosplit
1316.An Chris Demetriou ,
1317.An Charles Hannum ,
1318.An Ross Harvey ,
1319.An Matthew Jacob ,
1320.An Jonathan Stone ,
1321and
1322.An Matt Thomas .
1323.Pp
1324The
1325.Nm
1326interface in
1327.Fx
1328benefits from the contributions of
1329.An Justin T. Gibbs ,
1330.An Peter Wemm ,
1331.An Doug Rabson ,
1332.An Matthew N. Dodd ,
1333.An Sam Leffler ,
1334.An Maxime Henrion ,
1335.An Jake Burkholder ,
1336.An Takahashi Yoshihiro ,
1337.An Scott Long
1338and many others.
1339.Pp
1340This manual page was written by
1341.An Hiten M. Pandya
1342and
1343.An Justin T. Gibbs .
1344