freebsd-skq/share/man/man9/pmap_enter.9
kib d5ea06ecc4 Document new required MI behaviour of pmap_enter(9) for CoW.
Reviewed by:	markj
MFC after:	3 days
Sponsored by:	The FreeBSD Foundation
Differential revision:	https://reviews.freebsd.org/D18568
2018-12-16 01:51:38 +00:00

173 lines
5.1 KiB
Groff

.\"
.\" Copyright (c) 2003 Bruce M Simpson <bms@spc.org>
.\" Copyright (c) 2014 The FreeBSD Foundation
.\" All rights reserved.
.\"
.\" 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 December 16, 2018
.Dt PMAP_ENTER 9
.Os
.Sh NAME
.Nm pmap_enter
.Nd insert a virtual page into a physical map
.Sh SYNOPSIS
.In sys/param.h
.In vm/vm.h
.In vm/pmap.h
.Ft int
.Fo pmap_enter
.Fa "pmap_t pmap" "vm_offset_t va" "vm_page_t m" "vm_prot_t prot"
.Fa "u_int flags" "int8_t psind"
.Fc
.Sh DESCRIPTION
The
.Fn pmap_enter
function creates a mapping in the physical map
.Fa pmap
from the virtual address
.Fa va
to the physical page
.Fa m
with the protection
.Fa prot .
Any previous mapping at the virtual address
.Fa va
is destroyed.
.Pp
The
.Fa flags
argument may have the following values:
.Bl -tag -width ".Dv PMAP_ENTER_NOSLEEP"
.It Dv VM_PROT_READ
A read access to the given virtual address triggered the call.
.It Dv VM_PROT_WRITE
A write access to the given virtual address triggered the call.
.It Dv VM_PROT_EXECUTE
An execute access to the given virtual address triggered the call.
.It Dv PMAP_ENTER_WIRED
The mapping should be marked as wired.
.It Dv PMAP_ENTER_NOSLEEP
This function may not sleep during creation of the mapping.
If the mapping cannot be created without sleeping, an appropriate
Mach VM error is returned.
.El
If the
.Dv PMAP_ENTER_NOSLEEP
flag is not specified, this function must create the requested mapping
before returning.
It may not fail.
In order to create the requested mapping, this function may destroy
any non-wired mapping in any pmap.
.Pp
The
.Fa psind
parameter specifies the page size that should be used by the mapping.
The supported page sizes are described by the global array
.Dv pagesizes[] .
The desired page size is specified by passing the index of the array
element that equals the desired page size.
.Pp
When the
.Fn pmap_enter
function destroys or updates a managed mapping, including an existing
mapping at virtual address
.Fa va ,
it updates the
.Ft vm_page
structure corresponding to the previously mapped physical page.
If the physical page was accessed through the managed mapping,
then the
.Ft vm_page
structure's
.Dv PGA_REFERENCED
aflag is set.
If the physical page was modified through the managed mapping, then the
.Fn vm_page_dirty
function is called on the
.Ft vm_page
structure.
.Pp
The
.Dv PGA_WRITEABLE
aflag must be set for the page
.Fa m
if the new mapping is managed and writeable.
It is advised to clear
.Dv PGA_WRITEABLE
for destroyed mappings if the implementation can ensure
that no other writeable managed mappings for the previously
mapped pages exist.
.Pp
If the request modifies an existing mapping to use a different physical
page, an implementation of
.Nm
must invalidate the previous mapping before installing the new one.
This ensures that all threads sharing the pmap keep a consistent
view of the mapping, which is necessary for the correct handling
of CoW (copy on write) faults.
.Pp
If the page
.Fa m
is managed, the page must be busied by the caller
or the owning object must be locked.
In the later case, the
.Dv PMAP_ENTER_NOSLEEP
must be specified by the caller.
.Pp
The
.Fn pmap_enter
function must handle the multiprocessor TLB consistency for the
given address.
.Sh NOTES
On arm and i386 architectures the existing implementation
of the
.Nm
function is incomplete, only value 0 for
.Fa psind
is supported.
Other supported architectures, except amd64, have
.Dv pagesizes[]
array of size 1.
.Sh RETURN VALUES
If successful, the
.Fn pmap_enter
function returns
.Er KERN_SUCCESS .
If the
.Dv PMAP_ENTER_NOSLEEP
flag was specified and the resources required for the mapping cannot
be acquired without sleeping,
.Dv KERN_RESOURCE_SHORTAGE
is returned.
.Sh SEE ALSO
.Xr pmap 9
.Sh AUTHORS
This manual page was first written by
.An Bruce M Simpson Aq Mt bms@spc.org
and then rewritten by
.An Alan Cox Aq Mt alc@FreeBSD.org
and
.An Konstantin Belousov Aq Mt kib@FreeBSD.org .