freebsd-dev/share/man/man9/bus_map_resource.9

166 lines
4.8 KiB
Groff
Raw Normal View History

Add new bus methods for mapping resources. Add a pair of bus methods that can be used to "map" resources for direct CPU access using bus_space(9). bus_map_resource() creates a mapping and bus_unmap_resource() releases a previously created mapping. Mappings are described by 'struct resource_map' object. Pointers to these objects can be passed as the first argument to the bus_space wrapper API used for bus resources. Drivers that wish to map all of a resource using default settings (for example, using uncacheable memory attributes) do not need to change. However, drivers that wish to use non-default settings can now do so without jumping through hoops. First, an RF_UNMAPPED flag is added to request that a resource is not implicitly mapped with the default settings when it is activated. This permits other activation steps (such as enabling I/O or memory decoding in a device's PCI command register) to be taken without creating a mapping. Right now the AGP drivers don't set RF_ACTIVE to avoid using up a large amount of KVA to map the AGP aperture on 32-bit platforms. Once RF_UNMAPPED is supported on all platforms that support AGP this can be changed to using RF_UNMAPPED with RF_ACTIVE instead. Second, bus_map_resource accepts an optional structure that defines additional settings for a given mapping. For example, a driver can now request to map only a subset of a resource instead of the entire range. The AGP driver could also use this to only map the first page of the aperture (IIRC, it calls pmap_mapdev() directly to map the first page currently). I will also eventually change the PCI-PCI bridge driver to request mappings of the subset of the I/O window resource on its parent side to create mappings for child devices rather than passing child resources directly up to nexus to be mapped. This also permits bridges that do address translation to request suitable mappings from a resource on the "upper" side of the bus when mapping resources on the "lower" side of the bus. Another attribute that can be specified is an alternate memory attribute for memory-mapped resources. This can be used to request a Write-Combining mapping of a PCI BAR in an MI fashion. (Currently the drivers that do this call pmap_change_attr() directly for x86 only.) Note that this commit only adds the MI framework. Each platform needs to add support for handling RF_UNMAPPED and thew new bus_map/unmap_resource methods. Generally speaking, any drivers that are calling rman_set_bustag() and rman_set_bushandle() need to be updated. Discussed on: arch Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D5237
2016-05-20 17:57:47 +00:00
.\" -*- nroff -*-
.\"
.\" Copyright (c) 2016 John H. Baldwin <jhb@FreeBSD.org>
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\" notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\" notice, this list of conditions and the following disclaimer in the
.\" documentation and/or other materials provided with the distribution.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.\" $FreeBSD$
.\"
.Dd February 5, 2018
Add new bus methods for mapping resources. Add a pair of bus methods that can be used to "map" resources for direct CPU access using bus_space(9). bus_map_resource() creates a mapping and bus_unmap_resource() releases a previously created mapping. Mappings are described by 'struct resource_map' object. Pointers to these objects can be passed as the first argument to the bus_space wrapper API used for bus resources. Drivers that wish to map all of a resource using default settings (for example, using uncacheable memory attributes) do not need to change. However, drivers that wish to use non-default settings can now do so without jumping through hoops. First, an RF_UNMAPPED flag is added to request that a resource is not implicitly mapped with the default settings when it is activated. This permits other activation steps (such as enabling I/O or memory decoding in a device's PCI command register) to be taken without creating a mapping. Right now the AGP drivers don't set RF_ACTIVE to avoid using up a large amount of KVA to map the AGP aperture on 32-bit platforms. Once RF_UNMAPPED is supported on all platforms that support AGP this can be changed to using RF_UNMAPPED with RF_ACTIVE instead. Second, bus_map_resource accepts an optional structure that defines additional settings for a given mapping. For example, a driver can now request to map only a subset of a resource instead of the entire range. The AGP driver could also use this to only map the first page of the aperture (IIRC, it calls pmap_mapdev() directly to map the first page currently). I will also eventually change the PCI-PCI bridge driver to request mappings of the subset of the I/O window resource on its parent side to create mappings for child devices rather than passing child resources directly up to nexus to be mapped. This also permits bridges that do address translation to request suitable mappings from a resource on the "upper" side of the bus when mapping resources on the "lower" side of the bus. Another attribute that can be specified is an alternate memory attribute for memory-mapped resources. This can be used to request a Write-Combining mapping of a PCI BAR in an MI fashion. (Currently the drivers that do this call pmap_change_attr() directly for x86 only.) Note that this commit only adds the MI framework. Each platform needs to add support for handling RF_UNMAPPED and thew new bus_map/unmap_resource methods. Generally speaking, any drivers that are calling rman_set_bustag() and rman_set_bushandle() need to be updated. Discussed on: arch Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D5237
2016-05-20 17:57:47 +00:00
.Dt BUS_MAP_RESOURCE 9
.Os
.Sh NAME
.Nm bus_map_resource , bus_unmap_resource , resource_init_map_request
.Nd map or unmap an active resource
.Sh SYNOPSIS
.In sys/param.h
.In sys/bus.h
.Pp
.In machine/bus.h
.In sys/rman.h
.In machine/resource.h
.Ft int
.Fo bus_map_resource
.Fa "device_t dev" "int type" "struct resource *r"
.Fa "struct resource_map_request *args" "struct resource_map *map"
.Fc
.Ft int
.Fo bus_unmap_resource
.Fa "device_t dev" "int type" "struct resource *r" "struct resource_map *map"
.Fc
.Ft void
.Fn resource_init_map_request "struct resource_map_request *args"
.Sh DESCRIPTION
These functions create or destroy a mapping of a previously activated
resource.
Mappings permit CPU access to the resource via the
.Xr bus_space 9
API.
.Pp
The arguments are as follows:
.Bl -tag -width indent
.It Fa dev
The device that owns the resource.
.It Fa type
The type of resource to map.
It is one of:
.Pp
.Bl -tag -width ".Dv SYS_RES_MEMORY" -compact
.It Dv SYS_RES_IOPORT
for I/O ports
.It Dv SYS_RES_MEMORY
for I/O memory
.El
.It Fa r
A pointer to the
.Vt "struct resource"
returned by
.Xr bus_alloc_resource 9 .
.It Fa args
A set of optional properties to apply when creating a mapping.
This argument can be set to
.Dv NULL
to request a mapping of the entire resource with the default properties.
.It Fa map
The resource mapping to create or destroy.
.El
.Ss Resource Mappings
Resource mappings are described by a
.Vt "struct resource_map"
object.
This structure contains a
.Xr bus_space 9
tag and handle in the
.Va r_bustag
and
.Va r_bushandle
members that can be used for CPU access to the mapping.
The structure also contains a
.Va r_vaddr
member which contains the virtual address of the mapping if one exists.
.Pp
The wrapper API for
.Vt "struct resource"
objects described in
.Xr bus_activate_resource 9
can also be used with
.Vt "struct resource_map" .
For example,
a pointer to a mapping object can be passed as the first argument to
.Fn bus_read_4 .
This wrapper API is preferred over using the
.Va r_bustag
and
.Va r_bushandle
members directly.
.Ss Optional Mapping Properties
The
.Vt "struct resource_map_request"
object passed in
.Fa args
can be used to specify optional properties of a mapping.
The structure must be initialized by invoking
.Fn resource_init_map_request .
Properties are then specified by setting one or more of these members:
.Bl -tag -width indent
.It Va offset , length
These two members specify a region of the resource to map.
By default a mapping is created for the entire resource.
The
.Va offset
is relative to the start of the resource.
.It Va memattr
Specifies a memory attribute to use when mapping the resource.
By default memory mappings use the
.Dv VM_MEMATTR_UNCACHEABLE
attribute.
.El
.Sh RETURN VALUES
Zero is returned on success, otherwise an error is returned.
Add new bus methods for mapping resources. Add a pair of bus methods that can be used to "map" resources for direct CPU access using bus_space(9). bus_map_resource() creates a mapping and bus_unmap_resource() releases a previously created mapping. Mappings are described by 'struct resource_map' object. Pointers to these objects can be passed as the first argument to the bus_space wrapper API used for bus resources. Drivers that wish to map all of a resource using default settings (for example, using uncacheable memory attributes) do not need to change. However, drivers that wish to use non-default settings can now do so without jumping through hoops. First, an RF_UNMAPPED flag is added to request that a resource is not implicitly mapped with the default settings when it is activated. This permits other activation steps (such as enabling I/O or memory decoding in a device's PCI command register) to be taken without creating a mapping. Right now the AGP drivers don't set RF_ACTIVE to avoid using up a large amount of KVA to map the AGP aperture on 32-bit platforms. Once RF_UNMAPPED is supported on all platforms that support AGP this can be changed to using RF_UNMAPPED with RF_ACTIVE instead. Second, bus_map_resource accepts an optional structure that defines additional settings for a given mapping. For example, a driver can now request to map only a subset of a resource instead of the entire range. The AGP driver could also use this to only map the first page of the aperture (IIRC, it calls pmap_mapdev() directly to map the first page currently). I will also eventually change the PCI-PCI bridge driver to request mappings of the subset of the I/O window resource on its parent side to create mappings for child devices rather than passing child resources directly up to nexus to be mapped. This also permits bridges that do address translation to request suitable mappings from a resource on the "upper" side of the bus when mapping resources on the "lower" side of the bus. Another attribute that can be specified is an alternate memory attribute for memory-mapped resources. This can be used to request a Write-Combining mapping of a PCI BAR in an MI fashion. (Currently the drivers that do this call pmap_change_attr() directly for x86 only.) Note that this commit only adds the MI framework. Each platform needs to add support for handling RF_UNMAPPED and thew new bus_map/unmap_resource methods. Generally speaking, any drivers that are calling rman_set_bustag() and rman_set_bushandle() need to be updated. Discussed on: arch Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D5237
2016-05-20 17:57:47 +00:00
.Sh EXAMPLES
This maps a PCI memory BAR with the write-combining memory attribute and
reads the first 32-bit word:
.Bd -literal
struct resource *r;
struct resource_map map;
struct resource_map_request req;
Add new bus methods for mapping resources. Add a pair of bus methods that can be used to "map" resources for direct CPU access using bus_space(9). bus_map_resource() creates a mapping and bus_unmap_resource() releases a previously created mapping. Mappings are described by 'struct resource_map' object. Pointers to these objects can be passed as the first argument to the bus_space wrapper API used for bus resources. Drivers that wish to map all of a resource using default settings (for example, using uncacheable memory attributes) do not need to change. However, drivers that wish to use non-default settings can now do so without jumping through hoops. First, an RF_UNMAPPED flag is added to request that a resource is not implicitly mapped with the default settings when it is activated. This permits other activation steps (such as enabling I/O or memory decoding in a device's PCI command register) to be taken without creating a mapping. Right now the AGP drivers don't set RF_ACTIVE to avoid using up a large amount of KVA to map the AGP aperture on 32-bit platforms. Once RF_UNMAPPED is supported on all platforms that support AGP this can be changed to using RF_UNMAPPED with RF_ACTIVE instead. Second, bus_map_resource accepts an optional structure that defines additional settings for a given mapping. For example, a driver can now request to map only a subset of a resource instead of the entire range. The AGP driver could also use this to only map the first page of the aperture (IIRC, it calls pmap_mapdev() directly to map the first page currently). I will also eventually change the PCI-PCI bridge driver to request mappings of the subset of the I/O window resource on its parent side to create mappings for child devices rather than passing child resources directly up to nexus to be mapped. This also permits bridges that do address translation to request suitable mappings from a resource on the "upper" side of the bus when mapping resources on the "lower" side of the bus. Another attribute that can be specified is an alternate memory attribute for memory-mapped resources. This can be used to request a Write-Combining mapping of a PCI BAR in an MI fashion. (Currently the drivers that do this call pmap_change_attr() directly for x86 only.) Note that this commit only adds the MI framework. Each platform needs to add support for handling RF_UNMAPPED and thew new bus_map/unmap_resource methods. Generally speaking, any drivers that are calling rman_set_bustag() and rman_set_bushandle() need to be updated. Discussed on: arch Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D5237
2016-05-20 17:57:47 +00:00
uint32_t val;
int rid;
rid = PCIR_BAR(0);
r = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid, RF_ACTIVE |
RF_UNMAPPED);
resource_init_map_request(&req);
req.memattr = VM_MEMATTR_WRITE_COMBINING;
bus_map_resource(dev, SYS_RES_MEMORY, r, &req, &map);
Add new bus methods for mapping resources. Add a pair of bus methods that can be used to "map" resources for direct CPU access using bus_space(9). bus_map_resource() creates a mapping and bus_unmap_resource() releases a previously created mapping. Mappings are described by 'struct resource_map' object. Pointers to these objects can be passed as the first argument to the bus_space wrapper API used for bus resources. Drivers that wish to map all of a resource using default settings (for example, using uncacheable memory attributes) do not need to change. However, drivers that wish to use non-default settings can now do so without jumping through hoops. First, an RF_UNMAPPED flag is added to request that a resource is not implicitly mapped with the default settings when it is activated. This permits other activation steps (such as enabling I/O or memory decoding in a device's PCI command register) to be taken without creating a mapping. Right now the AGP drivers don't set RF_ACTIVE to avoid using up a large amount of KVA to map the AGP aperture on 32-bit platforms. Once RF_UNMAPPED is supported on all platforms that support AGP this can be changed to using RF_UNMAPPED with RF_ACTIVE instead. Second, bus_map_resource accepts an optional structure that defines additional settings for a given mapping. For example, a driver can now request to map only a subset of a resource instead of the entire range. The AGP driver could also use this to only map the first page of the aperture (IIRC, it calls pmap_mapdev() directly to map the first page currently). I will also eventually change the PCI-PCI bridge driver to request mappings of the subset of the I/O window resource on its parent side to create mappings for child devices rather than passing child resources directly up to nexus to be mapped. This also permits bridges that do address translation to request suitable mappings from a resource on the "upper" side of the bus when mapping resources on the "lower" side of the bus. Another attribute that can be specified is an alternate memory attribute for memory-mapped resources. This can be used to request a Write-Combining mapping of a PCI BAR in an MI fashion. (Currently the drivers that do this call pmap_change_attr() directly for x86 only.) Note that this commit only adds the MI framework. Each platform needs to add support for handling RF_UNMAPPED and thew new bus_map/unmap_resource methods. Generally speaking, any drivers that are calling rman_set_bustag() and rman_set_bushandle() need to be updated. Discussed on: arch Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D5237
2016-05-20 17:57:47 +00:00
val = bus_read_4(&map, 0);
.Ed
.Sh SEE ALSO
.Xr bus_activate_resource 9 ,
.Xr bus_alloc_resource 9 ,
.Xr bus_space 9 ,
.Xr device 9 ,
.Xr driver 9
.Sh AUTHORS
This manual page was written by
.An John Baldwin Aq Mt jhb@FreeBSD.org .