NetBSD/sys/kern/vnode_if.src

600 lines
12 KiB
Plaintext
Raw Normal View History

# $NetBSD: vnode_if.src,v 1.36 2003/11/17 11:16:10 wiz Exp $
1994-06-08 15:28:29 +04:00
#
# Copyright (c) 1992, 1993
# The Regents of the University of California. 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.
# 3. All advertising materials mentioning features or use of this software
# must display the following acknowledgement:
# This product includes software developed by the University of
# California, Berkeley and its contributors.
# 4. Neither the name of the University 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 REGENTS 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 REGENTS 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.
#
1998-03-01 05:20:01 +03:00
# @(#)vnode_if.src 8.14 (Berkeley) 8/6/95
#
#
#
# Above each of the vop descriptors is a specification of the locking
# protocol used by each vop call. The first column is the name of
# the variable, the remaining three columns are in, out and error
# respectively. The "in" column defines the lock state on input,
# the "out" column defines the state on successful return, and the
1998-03-01 05:20:01 +03:00
# "error" column defines the locking state on error exit.
#
# The locking value can take the following values:
# L: locked.
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
# U: unlocked.
1998-03-01 05:20:01 +03:00
# -: not applicable. vnode does not yet (or no longer) exists.
# =: the same on input and output, may be either L or U.
# X: locked if not nil.
#
# For operations other than VOP_LOOKUP which require a component name
# parameter, the flags required for the initial namei() call are listed.
# Additional flags may be added to the namei() call, but these are required.
1998-03-01 05:20:01 +03:00
#
#
#% lookup dvp L ? ?
#% lookup vpp - L -
#
# XXX - the lookup locking protocol defies simple description and depends
# on the flags and operation fields in the (cnp) structure. Note
# especially that *vpp may equal dvp and both may be locked.
1994-06-08 15:28:29 +04:00
#
# More details:
# There are three types of lookups: ".", ".." (ISDOTDOT), and other.
# On successful lookup of ".", a reference is added to dvp, and it
# is returned in *vpp.
# To look up ISDOTDOT, dvp is unlocked, the ".." node is locked, and
# then dvp is relocked iff LOCKPARENT is set and this is the last
# component name (ISLASTCN set). This preserves the
# protocol of always locking nodes from root ("/") downward and
# prevents deadlock.
# Other lookups find the named node (creating the vnode if needed) and
# return it, locked, in *vpp.
# For non-"." lookups, if LOCKPARENT is not set or this was not the
# last component name, dvp is returned unlocked on a successful
# lookup.
# On failure, *vpp is NULL, and *dvp is left locked. If there was
# an error re-locking dvp (for instance in the ISDOTDOT case),
# the error is returned with PDIRUNLOCK set.
#
# *vpp is always locked on return if the operation succeeds.
# typically, if *vpp == dvp, you need to release twice, but unlock once.
#
# The PDIRUNLOCK flag is set when dvp is unlocked in the lookup routine.
# It signals the caller that dvp's lock state changed. It will
# be set on exit if either a successful lookup unlocked the
# parrent, or there was an error re-locking dvp in the ISDOTDOT case.
#
# See sys/sys/namei.h for a description of the SAVENAME and SAVESTART
# flags.
#
1994-06-08 15:28:29 +04:00
vop_lookup {
IN struct vnode *dvp;
INOUT struct vnode **vpp;
IN struct componentname *cnp;
};
1998-03-01 05:20:01 +03:00
#
#% create dvp L U U
#% create vpp - L -
#
#! create cnp CREATE, LOCKPARENT
#
1994-06-08 15:28:29 +04:00
vop_create {
IN WILLPUT struct vnode *dvp;
1994-06-08 15:28:29 +04:00
OUT struct vnode **vpp;
IN struct componentname *cnp;
IN struct vattr *vap;
};
1998-03-01 05:20:01 +03:00
#
#% mknod dvp L U U
#% mknod vpp - L -
1998-03-01 05:20:01 +03:00
#
#! mknod cnp CREATE, LOCKPARENT
#
1994-06-08 15:28:29 +04:00
vop_mknod {
IN WILLPUT struct vnode *dvp;
OUT struct vnode **vpp;
1994-06-08 15:28:29 +04:00
IN struct componentname *cnp;
IN struct vattr *vap;
};
1998-03-01 05:20:01 +03:00
#
#% open vp L L L
#
1994-06-08 15:28:29 +04:00
vop_open {
IN struct vnode *vp;
IN int mode;
IN struct ucred *cred;
IN struct proc *p;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% close vp L L L
1998-03-01 05:20:01 +03:00
#
1994-06-08 15:28:29 +04:00
vop_close {
IN struct vnode *vp;
IN int fflag;
IN struct ucred *cred;
IN struct proc *p;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% access vp L L L
#
1994-06-08 15:28:29 +04:00
vop_access {
IN struct vnode *vp;
IN int mode;
IN struct ucred *cred;
IN struct proc *p;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% getattr vp = = =
#
1994-06-08 15:28:29 +04:00
vop_getattr {
IN struct vnode *vp;
IN struct vattr *vap;
IN struct ucred *cred;
IN struct proc *p;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% setattr vp L L L
#
1994-06-08 15:28:29 +04:00
vop_setattr {
IN struct vnode *vp;
IN struct vattr *vap;
IN struct ucred *cred;
IN struct proc *p;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% read vp L L L
#
1994-06-08 15:28:29 +04:00
vop_read {
IN struct vnode *vp;
INOUT struct uio *uio;
IN int ioflag;
IN struct ucred *cred;
};
1998-03-01 05:20:01 +03:00
#
#% write vp L L L
#
1994-06-08 15:28:29 +04:00
vop_write {
IN struct vnode *vp;
INOUT struct uio *uio;
IN int ioflag;
IN struct ucred *cred;
};
1998-03-01 05:20:01 +03:00
#
#% ioctl vp U U U
#
1994-06-08 15:28:29 +04:00
vop_ioctl {
IN struct vnode *vp;
IN u_long command;
IN void *data;
1994-06-08 15:28:29 +04:00
IN int fflag;
IN struct ucred *cred;
IN struct proc *p;
1994-06-08 15:28:29 +04:00
};
1999-08-03 22:19:08 +04:00
#
#% fcntl vp L L L
#
vop_fcntl {
IN struct vnode *vp;
IN u_int command;
IN void *data;
1999-08-03 22:19:08 +04:00
IN int fflag;
IN struct ucred *cred;
IN struct proc *p;
1999-08-03 22:19:08 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% poll vp U U U
#
1996-09-07 16:40:22 +04:00
vop_poll {
1994-06-08 15:28:29 +04:00
IN struct vnode *vp;
1996-09-07 16:40:22 +04:00
IN int events;
IN struct proc *p;
1994-06-08 15:28:29 +04:00
};
#
#% kqfilter vp U U U
#
vop_kqfilter {
IN struct vnode *vp;
IN struct knote *kn;
};
1998-03-01 05:20:01 +03:00
#
#% revoke vp U U U
#
vop_revoke {
IN struct vnode *vp;
IN int flags;
};
#
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
#% mmap vp = = =
1998-03-01 05:20:01 +03:00
#
1994-06-08 15:28:29 +04:00
vop_mmap {
IN struct vnode *vp;
IN int fflags;
IN struct ucred *cred;
IN struct proc *p;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% fsync vp L L L
#
1994-06-08 15:28:29 +04:00
vop_fsync {
IN struct vnode *vp;
IN struct ucred *cred;
IN int flags;
IN off_t offlo
IN off_t offhi
IN struct proc *p;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
# Needs work: Is newoff right? What's it mean?
#
1994-06-08 15:28:29 +04:00
vop_seek {
IN struct vnode *vp;
IN off_t oldoff;
IN off_t newoff;
IN struct ucred *cred;
};
1998-03-01 05:20:01 +03:00
#
#% remove dvp L U U
#% remove vp L U U
#
#! remove cnp DELETE, LOCKPARENT | LOCKLEAF
#
1994-06-08 15:28:29 +04:00
vop_remove {
IN WILLPUT struct vnode *dvp;
IN WILLPUT struct vnode *vp;
1994-06-08 15:28:29 +04:00
IN struct componentname *cnp;
};
1998-03-01 05:20:01 +03:00
#
#% link vp U U U
#% link tdvp L U U
#
#! link cnp CREATE, LOCKPARENT
#
1994-06-08 15:28:29 +04:00
vop_link {
IN WILLPUT struct vnode *dvp;
IN struct vnode *vp;
1994-06-08 15:28:29 +04:00
IN struct componentname *cnp;
};
1998-03-01 05:20:01 +03:00
#
#% rename fdvp U U U
#% rename fvp U U U
#% rename tdvp L U U
#% rename tvp X U U
#
#! rename fcnp DELETE, WANTPARENT | SAVESTART
#! rename tcnp RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART
#
# XXX the vop_rename routines should REALLY NOT be depending on SAVESTART!
#
1994-06-08 15:28:29 +04:00
vop_rename {
IN WILLRELE struct vnode *fdvp;
IN WILLRELE struct vnode *fvp;
IN struct componentname *fcnp;
IN WILLPUT struct vnode *tdvp;
IN WILLPUT struct vnode *tvp;
1994-06-08 15:28:29 +04:00
IN struct componentname *tcnp;
};
1998-03-01 05:20:01 +03:00
#
#% mkdir dvp L U U
#% mkdir vpp - L -
#
#! mkdir cnp CREATE, LOCKPARENT
#
1994-06-08 15:28:29 +04:00
vop_mkdir {
IN WILLPUT struct vnode *dvp;
1994-06-08 15:28:29 +04:00
OUT struct vnode **vpp;
IN struct componentname *cnp;
IN struct vattr *vap;
};
1998-03-01 05:20:01 +03:00
#
#% rmdir dvp L U U
#% rmdir vp L U U
#
#! rmdir cnp DELETE, LOCKPARENT | LOCKLEAF
#
1994-06-08 15:28:29 +04:00
vop_rmdir {
IN WILLPUT struct vnode *dvp;
IN WILLPUT struct vnode *vp;
1994-06-08 15:28:29 +04:00
IN struct componentname *cnp;
};
1998-03-01 05:20:01 +03:00
#
#% symlink dvp L U U
#% symlink vpp - L -
1998-03-01 05:20:01 +03:00
#
#! symlink cnp CREATE, LOCKPARENT
#
1994-06-08 15:28:29 +04:00
vop_symlink {
IN WILLPUT struct vnode *dvp;
OUT struct vnode **vpp;
1994-06-08 15:28:29 +04:00
IN struct componentname *cnp;
IN struct vattr *vap;
IN char *target;
};
1998-03-01 05:20:01 +03:00
#
#% readdir vp L L L
#
1994-06-08 15:28:29 +04:00
vop_readdir {
IN struct vnode *vp;
INOUT struct uio *uio;
IN struct ucred *cred;
OUT int *eofflag;
1998-03-01 05:20:01 +03:00
OUT off_t **cookies;
IN int *ncookies;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% readlink vp L L L
#
1994-06-08 15:28:29 +04:00
vop_readlink {
IN struct vnode *vp;
INOUT struct uio *uio;
IN struct ucred *cred;
};
1998-03-01 05:20:01 +03:00
#
#% abortop dvp = = =
#
#! abortop cnp as appropriate.
#
1994-06-08 15:28:29 +04:00
vop_abortop {
IN struct vnode *dvp;
IN struct componentname *cnp;
};
1998-03-01 05:20:01 +03:00
#
#% inactive vp L U U
#
1994-06-08 15:28:29 +04:00
vop_inactive {
IN WILLUNLOCK struct vnode *vp;
IN struct proc *p;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% reclaim vp U U U
#
1994-06-08 15:28:29 +04:00
vop_reclaim {
IN struct vnode *vp;
IN struct proc *p;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% lock vp U L U
#
1994-06-08 15:28:29 +04:00
vop_lock {
IN struct vnode *vp;
1998-03-01 05:20:01 +03:00
IN int flags;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% unlock vp L U L
#
1994-06-08 15:28:29 +04:00
vop_unlock {
IN struct vnode *vp;
1998-03-01 05:20:01 +03:00
IN int flags;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% bmap vp L L L
#% bmap vpp - U -
#
1994-06-08 15:28:29 +04:00
vop_bmap {
IN struct vnode *vp;
IN daddr_t bn;
OUT struct vnode **vpp;
IN daddr_t *bnp;
OUT int *runp;
};
1998-03-01 05:20:01 +03:00
#
# Needs work: no vp?
#
1994-06-08 15:28:29 +04:00
#vop_strategy {
# IN struct buf *bp;
#};
1998-03-01 05:20:01 +03:00
#
#% print vp = = =
#
1994-06-08 15:28:29 +04:00
vop_print {
IN struct vnode *vp;
};
1998-03-01 05:20:01 +03:00
#
#% islocked vp = = =
#
1994-06-08 15:28:29 +04:00
vop_islocked {
IN struct vnode *vp;
};
1998-03-01 05:20:01 +03:00
#
#% pathconf vp L L L
#
1994-06-08 15:28:29 +04:00
vop_pathconf {
IN struct vnode *vp;
IN int name;
OUT register_t *retval;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% advlock vp U U U
#
1994-06-08 15:28:29 +04:00
vop_advlock {
IN struct vnode *vp;
IN void *id;
1994-06-08 15:28:29 +04:00
IN int op;
IN struct flock *fl;
IN int flags;
};
1998-03-01 05:20:01 +03:00
#
#% blkatoff vp L L L
#
1994-06-08 15:28:29 +04:00
vop_blkatoff {
IN struct vnode *vp;
IN off_t offset;
OUT char **res;
OUT struct buf **bpp;
};
1998-03-01 05:20:01 +03:00
#
#% valloc pvp L L L
#
1994-06-08 15:28:29 +04:00
vop_valloc {
IN struct vnode *pvp;
IN int mode;
IN struct ucred *cred;
OUT struct vnode **vpp;
};
#
#% balloc vp L L L
#
vop_balloc {
IN struct vnode *vp;
IN off_t startoffset;
IN int size;
IN struct ucred *cred;
IN int flags;
OUT struct buf **bpp;
};
1998-03-01 05:20:01 +03:00
#
#% reallocblks vp L L L
#
1994-06-08 15:28:29 +04:00
vop_reallocblks {
IN struct vnode *vp;
IN struct cluster_save *buflist;
};
1998-03-01 05:20:01 +03:00
#
#% vfree pvp L L L
#
1994-06-08 15:28:29 +04:00
vop_vfree {
IN struct vnode *pvp;
IN ino_t ino;
IN int mode;
};
1998-03-01 05:20:01 +03:00
#
#% truncate vp L L L
#
1994-06-08 15:28:29 +04:00
vop_truncate {
IN struct vnode *vp;
IN off_t length;
IN int flags;
IN struct ucred *cred;
IN struct proc *p;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% update vp L L L
#
1994-06-08 15:28:29 +04:00
vop_update {
IN struct vnode *vp;
IN struct timespec *access;
IN struct timespec *modify;
IN int flags;
1994-06-08 15:28:29 +04:00
};
1998-03-01 05:20:01 +03:00
#
#% lease vp = = =
#
vop_lease {
IN struct vnode *vp;
IN struct proc *p;
IN struct ucred *cred;
IN int flag;
};
1998-03-01 05:20:01 +03:00
#
#% whiteout dvp L L L
#% whiteout cnp - - -
#% whiteout flag - - -
#
#! whiteout cnp CREATE, LOCKPARENT
1998-03-01 05:20:01 +03:00
#
1994-12-14 16:03:11 +03:00
vop_whiteout {
IN struct vnode *dvp;
IN struct componentname *cnp;
IN int flags;
};
1998-03-01 05:20:01 +03:00
#
1994-06-08 15:28:29 +04:00
# Needs work: no vp?
1998-03-01 05:20:01 +03:00
#
1994-06-08 15:28:29 +04:00
#vop_bwrite {
# IN struct buf *bp;
#};
#
#% getpages vp L L L
#
vop_getpages {
IN struct vnode *vp;
IN voff_t offset;
IN struct vm_page **m;
IN int *count;
IN int centeridx;
IN vm_prot_t access_type;
IN int advice;
IN int flags;
};
#
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
#% putpages vp = = =
#
vop_putpages {
IN struct vnode *vp;
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
IN voff_t offlo;
IN voff_t offhi;
IN int flags;
};