.\"     $NetBSD: vfsops.9,v 1.28 2007/01/21 15:42:36 hannken Exp $
.\"
.\" Copyright (c) 2001 The NetBSD Foundation, Inc.
.\" All rights reserved.
.\"
.\" This code is derived from software contributed to The NetBSD Foundation
.\" by Gregory McGarry.
.\"
.\" 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.
.\" 3. All advertising materials mentioning features or use of this software
.\"    must display the following acknowledgement:
.\"        This product includes software developed by the NetBSD
.\"        Foundation, Inc. and its contributors.
.\" 4. Neither the name of The NetBSD Foundation nor the names of its
.\"    contributors may be used to endorse or promote products derived
.\"    from this software without specific prior written permission.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. 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 FOUNDATION 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.
.\"
.Dd January 21, 2007
.Dt VFSOPS 9
.Os
.Sh NAME
.Nm vfsops ,
.Nm VFS_MOUNT ,
.Nm VFS_START ,
.Nm VFS_UNMOUNT ,
.Nm VFS_ROOT ,
.Nm VFS_QUOTACTL ,
.Nm VFS_STATVFS ,
.Nm VFS_SYNC ,
.Nm VFS_VGET ,
.Nm VFS_FHTOVP ,
.Nm VFS_VPTOFH ,
.Nm VFS_SNAPSHOT ,
.Nm VFS_SUSPENDCTL
.Nd kernel file system interface
.Sh SYNOPSIS
.In sys/param.h
.In sys/mount.h
.In sys/vnode.h
.Ft int
.Fn VFS_MOUNT "struct mount *mp" "const char *path" "void *data" \
"struct nameidata *ndp" "struct lwp *l"
.Ft int
.Fn VFS_START "struct mount *mp" "int flags" "struct lwp *l"
.Ft int
.Fn VFS_UNMOUNT "struct mount *mp" "int mntflags" "struct lwp *l"
.Ft int
.Fn VFS_ROOT "struct mount *mp" "struct vnode **vpp"
.Ft int
.Fn VFS_QUOTACTL "struct mount *mp" "int cmds" "uid_t uid" \
"caddr_t arg" "struct lwp *l"
.Ft int
.Fn VFS_STATVFS "struct mount *mp" "struct statvfs *sbp" "struct lwp *l"
.Ft int
.Fn VFS_SYNC "struct mount *mp" "int waitfor" "kauth_cred_t cred" \
"struct lwp *l"
.Ft int
.Fn VFS_VGET "struct mount *mp" "ino_t ino" "struct vnode **vpp"
.Ft int
.Fn VFS_FHTOVP "struct mount *mp" "struct fid *fhp" "struct vnode **vpp"
.Ft int
.Fn VFS_VPTOFH "struct vnode *vp" "struct fid *fhp"
.Ft int
.Fn VFS_SNAPSHOT "struct mount *mp" "struct vnode *vp" "struct timespec *ts"
.Ft int
.Fn VFS_SUSPENDCTL "struct mount *mp" "int cmd"
.Sh DESCRIPTION
In a similar fashion to the
.Xr vnode 9
interface, all operations that are done on a file system are conducted
through a single interface that allows the system to carry out
operations on a file system without knowing its construction or type.
.Pp
All supported file systems in the kernel have an entry in the
.Va vfs_list_initial
table.
This table is generated by
.Xr config 1
and is a
.Dv NULL Ns No -terminated
list of
.Em vfsops
structures.
The vfsops structure describes the operations that can be done to a
specific file system type.
The following table lists the elements of the vfsops vector, the
corresponding invocation macro, and a description of the element.
.Pp
.nf
.ta \w'int (*vfs_mountroot)()'u+2n +\w'VFS_QUOTACTL'u+2n +\w'Get the file system root vnode'u
\fIVector element\fP	\fIMacro\fP	\fIDescription\fP
.ta \w'int (*vfs_mountroot)()'u+2n +\w'VFS_QUOTACTL'u+2n +\w'Get the file system root vnode'u+6nC
.sp 5p
int  (*vfs_mount)()	VFS_MOUNT	Mount a file system
int  (*vfs_start)()	VFS_START	Make operational
int  (*vfs_unmount)()	VFS_UMOUNT	Unmount a file system
int  (*vfs_root)()	VFS_ROOT	Get the file system root vnode
int  (*vfs_quotactl)()	VFS_QUOTACTL	Query/modify space quotas
int  (*vfs_statvfs)()	VFS_STATVFS	Get file system statistics
int  (*vfs_sync)()	VFS_SYNC	Flush file system buffers
int  (*vfs_vget)()	VFS_VGET	Get vnode from file id
int  (*vfs_fhtovp)()	VFS_FHTOVP	NFS file handle to vnode lookup
int  (*vfs_vptofh)()	VFS_VPTOFH	Vnode to NFS file handle lookup
void (*vfs_init)()	-	Initialise file system
void (*vfs_reinit)()	-	Reinitialise file system
void (*vfs_done)()	-	Cleanup unmounted file system
int  (*vfs_mountroot)()	-	Mount the root file system
int  (*vfs_snapshot)()	VFS_SNAPSHOT	Take a snapshot
int  (*vfs_suspendctl)()	VFS_SUSPENDCTL	Suspend or resume
.fi
.Pp
Some additional non-function members of the vfsops structure are the
file system name
.Ns Em vfs_name
and a reference count
.Ns Em vfs_refcount  .
It is not mandatory for a file system type to support a particular
operation, but it must assign each member function pointer to a
suitable function to do the minimum required of it.
In most cases, such functions either do nothing or return an error
value to the effect that it is not supported.
.Em vfs_reinit ,
.Em vfs_mountroot ,
.Em vfs_fhtovp ,
and
.Em vfs_vptofh
may
be
.Dv NULL .
.Pp
At system boot, each file system with an entry in
.Va vfs_list_initial
is established and initialised.
Each initialised file system is recorded by the kernel in the list
.Va vfs_list
and the file system specific initialisation function
.Em vfs_init
in its vfsops vector is invoked.
When the file system is no longer needed
.Em vfs_done
is invoked to run file system specific cleanups and the file system is
removed from the kernel list.
.Pp
At system boot, the root filesystem is mounted by invoking the file
system type specific
.Em vfs_mountroot
function in the vfsops vector.
All filesystems that can be mounted as a root file system must define
this function.
It is responsible for initialising to list of mount structures for
all future mounted file systems.
.Pp
Kernel state which affects a specific file system type can be
queried and modified using the
.Xr sysctl 8
interface.
.Sh FUNCTIONS
.Bl -tag -width compact
.It Fn VFS_MOUNT "mp" "path" "data" "ndp" "l"
Mount a file system specified by the mount structure
.Fa mp
on the mount point described by
.Fa path .
The argument
.Fa data
contains file system type specific data and is read into the kernel
using
.Xr copyin 9 .
The argument
.Fa ndp
contains the result of a
.Xr namei 9
call on the pathname of the mount point and
.Fa p
is the calling process.
.Fn VFS_MOUNT
initialises the mount structure for the mounted file system.
This structure records mount-specific information for the file system and
records the list of vnodes associated with the file system.
This function is invoked both to mount new file systems and to change the
attributes of an existing file system.
If the MNT_UPDATE flag is set in
.Em mp-\*[Gt]mnt_flag
then the filesystem should update its internal state from the value of
.Em mp-\*[Gt]mnt_flag .
This can be used, for instance, to convert a read-only filesystem to
read-write.
If the MNT_UPDATE flag is not specified, then this is a newly
mounted filesystem.
.It Fn VFS_START "mp" "flags" "l"
Make the file system specified by the mount structure
.Fa mp
operational.
The argument
.Fa p
is the calling process.
The argument
.Fa flags
is a set of flags for controlling the operation of
.Fn VFS_START .
This function is invoked after
.Fn VFS_MOUNT
and before the first access to the file system.
.It Fn VFS_UNMOUNT "mp" "mntflags" "l"
Unmount a file system specified by the mount structure
.Fa mp .
The argument
.Fa p
is the calling process.
.Fn VFS_UNMOUNT
performs any file system type specific operations required before the
file system is unmounted, such are flushing buffers.
If MNT_FORCE is specified in the flags
.Fa mntflags
then open files are forcibly closed.
The function also deallocates space associated with data structure
that were allocated for the file system when it was mounted.
.It Fn VFS_ROOT "mp" "vpp"
Get the root vnode of the file system specified by the mount
structure
.Fa mp .
The vnode is returned in the address given by
.Fa vpp .
This function is used by the pathname translation algorithms when a
vnode that has been covered by a mounted file system is encountered.
While resolving the pathname, the pathname translation algorithm will
have to go through the directory tree in the file system associated
with that mount point and therefore requires the root vnode of the
file system.
.It Fn VFS_QUOTACTL "mp" "cmds" "uid" "arg" "l"
Query/modify user space quotas for the file system specified by the
mount structure
.Fa mp .
The argument specifies the control command to perform.
The userid is specified in
.Fa id ,
the calling process is
.Fa p
and
.Fa arg
allows command-specific data to be returned to the system call
interface.
.Fn VFS_QUOTACTL
is the file system type specific implementation of the
.Xr quotactl 2
system call.
.It Fn VFS_STATVFS "mp" "sbp" "l"
Get file system statistics for the file system specified by the mount
structure
.Fa mp .
The argument
.Fa p
is the calling process.
A statvfs structure filled with the statistics is returned in
.Fa sbp .
.Fn VFS_STATVFS
is the file system type specific implementation of the
.Xr statvfs 2
and
.Xr fstatvfs 2
system calls.
.It Fn VFS_SYNC "mp" "waitfor" "cred" "l"
Flush file system I/O buffers for the file system specified by the mount
structure
.Fa mp .
The
.Fa waitfor
argument indicates whether a partial flush or complete flush should be
performed.
The arguments
.Fa l
and
.Fa cred
specific the calling lwp and its credentials, respectively.
.Fn VFS_SYNC
does not provide any return value since the operation can never fail.
.It Fn VFS_VGET "mp" "ino" "vpp"
Get vnode for a file system type specific file id
.Fa ino
for the file system specified by the mount structure
.Fa mp .
The vnode is returned in the address specified
.Fa vpp .
The function is optional for file systems which have a unique id
number for every file in the file system.
It is used internally by the UFS file system and also by the NFSv3
server to implement the READDIRPLUS nfs call.
If the file system does not support this function, it should return
.Er EOPNOTSUPP .
.It Fn VFS_FHTOVP "mp" "fhp" "vpp"
Get the vnode for the NFS file specified by the file handle
.Fa fhp
in the file system specified by the mount structure
.Fa mp .
The locked vnode is returned in
.Fa vpp .
.Pp
A call to
.Fn VFS_FHTOVP
should generally be followed by a call to
.Fn nfs_check_export
to check if the file is accessable to the client.
.Pp
If this function's pointer is set to
.Dv NULL ,
the underlying file system will not be NFS exportable.
If it is defined, though,
.Fn VFS_VPTOFH
must be defined too.
.It Fn VFS_VPTOFH "vp" "fhp"
Get a unique NFS file handle for the file specified by the vnode
.Fa vp .
The file handle is returned in
.Fa fhp .
The contents of the file handle are defined by the file system and are
not examined by any other subsystems.
It should contain enough information to uniquely identify a file within
the file system as well as noticing when a file has been removed and
the file system resources have been recycled for a new file.
.Pp
If this function's pointer is set to
.Dv NULL ,
the underlying file system will not be NFS exportable.
If it is defined, though,
.Fn VFS_FHTOVP
must be defined too.
.It Fn VFS_SNAPSHOT "mp" "vp" "ts"
Take a snapshot of the file system specified by the mount structure
.Fa mp
and make it accessible through the locked vnode
.Fa vp .
If
.Fa ts
is not
.Dv NULL
it will receive the time this snapshot was taken.
If the file system does not support this function, it should return
.Er EOPNOTSUPP .
.It Fn VFS_SUSPENDCTL "mp" "cmd"
Suspend or resume all operations on this file system.
.Fa cmd
is either
.Dv SUSPEND_SUSPEND
to suspend or
.Dv SUSPEND_RESUME
to resume operations.
If the file system does not support this function, it should return
.Er EOPNOTSUPP .
.El
.Sh CODE REFERENCES
This section describes places within the
.Nx
source tree where actual code implementing or using the vfs
operations can be found.
All pathnames are relative to
.Pa /usr/src .
.Pp
The vfs operations are implemented within the files
.Pa sys/kern/vfs_subr.c
and
.Pa sys/kern/vfs_init.c .
.Sh SEE ALSO
.Xr intro 9 ,
.Xr namei 9 ,
.Xr vfs 9 ,
.Xr vfssubr 9 ,
.Xr vnode 9 ,
.Xr vnodeops 9
.Sh HISTORY
The vfs operations vector, its functions and the corresponding macros
appeared in
.Bx 4.3 .