These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
/*
|
1994-05-25 09:21:21 +00:00
|
|
|
* Copyright (c) 1991 Regents of the University of California.
|
|
|
|
* All rights reserved.
|
1994-05-24 10:09:53 +00:00
|
|
|
*
|
|
|
|
* This code is derived from software contributed to Berkeley by
|
|
|
|
* The Mach Operating System project at Carnegie-Mellon University.
|
|
|
|
*
|
|
|
|
* 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.
|
|
|
|
*
|
1994-05-25 09:21:21 +00:00
|
|
|
* from: @(#)vm_page.c 7.4 (Berkeley) 5/7/91
|
1998-10-25 17:44:59 +00:00
|
|
|
* $Id: vm_page.c,v 1.109 1998/10/21 14:46:41 dg Exp $
|
1994-05-25 09:21:21 +00:00
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
1994-05-24 10:09:53 +00:00
|
|
|
* Copyright (c) 1987, 1990 Carnegie-Mellon University.
|
|
|
|
* All rights reserved.
|
|
|
|
*
|
|
|
|
* Authors: Avadis Tevanian, Jr., Michael Wayne Young
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
*
|
1994-05-24 10:09:53 +00:00
|
|
|
* Permission to use, copy, modify and distribute this software and
|
|
|
|
* its documentation is hereby granted, provided that both the copyright
|
|
|
|
* notice and this permission notice appear in all copies of the
|
|
|
|
* software, derivative works or modified versions, and any portions
|
|
|
|
* thereof, and that both notices appear in supporting documentation.
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
*
|
|
|
|
* CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
|
|
|
|
* CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
|
1994-05-24 10:09:53 +00:00
|
|
|
* FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
*
|
1994-05-24 10:09:53 +00:00
|
|
|
* Carnegie Mellon requests users of this software to return to
|
|
|
|
*
|
|
|
|
* Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU
|
|
|
|
* School of Computer Science
|
|
|
|
* Carnegie Mellon University
|
|
|
|
* Pittsburgh PA 15213-3890
|
|
|
|
*
|
|
|
|
* any improvements or extensions that they make and grant Carnegie the
|
|
|
|
* rights to redistribute these changes.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Resident memory management module.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <sys/param.h>
|
|
|
|
#include <sys/systm.h>
|
1996-01-27 00:13:33 +00:00
|
|
|
#include <sys/malloc.h>
|
1994-05-25 09:21:21 +00:00
|
|
|
#include <sys/proc.h>
|
1995-12-07 12:48:31 +00:00
|
|
|
#include <sys/vmmeter.h>
|
1997-12-29 00:25:11 +00:00
|
|
|
#include <sys/vnode.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
#include <vm/vm.h>
|
1995-12-07 12:48:31 +00:00
|
|
|
#include <vm/vm_param.h>
|
|
|
|
#include <vm/vm_prot.h>
|
1997-02-10 02:22:35 +00:00
|
|
|
#include <sys/lock.h>
|
1995-03-16 18:17:34 +00:00
|
|
|
#include <vm/vm_kern.h>
|
1995-12-07 12:48:31 +00:00
|
|
|
#include <vm/vm_object.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <vm/vm_page.h>
|
|
|
|
#include <vm/vm_pageout.h>
|
1995-12-07 12:48:31 +00:00
|
|
|
#include <vm/vm_extern.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1996-09-14 11:54:59 +00:00
|
|
|
static void vm_page_queue_init __P((void));
|
1996-09-28 03:33:40 +00:00
|
|
|
static vm_page_t vm_page_select_free __P((vm_object_t object,
|
|
|
|
vm_pindex_t pindex, int prefqueue));
|
This mega-commit is meant to fix numerous interrelated problems. There
has been some bitrot and incorrect assumptions in the vfs_bio code. These
problems have manifest themselves worse on NFS type filesystems, but can
still affect local filesystems under certain circumstances. Most of
the problems have involved mmap consistancy, and as a side-effect broke
the vfs.ioopt code. This code might have been committed seperately, but
almost everything is interrelated.
1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that
are fully valid.
2) Rather than deactivating erroneously read initial (header) pages in
kern_exec, we now free them.
3) Fix the rundown of non-VMIO buffers that are in an inconsistent
(missing vp) state.
4) Fix the disassociation of pages from buffers in brelse. The previous
code had rotted and was faulty in a couple of important circumstances.
5) Remove a gratuitious buffer wakeup in vfs_vmio_release.
6) Remove a crufty and currently unused cluster mechanism for VBLK
files in vfs_bio_awrite. When the code is functional, I'll add back
a cleaner version.
7) The page busy count wakeups assocated with the buffer cache usage were
incorrectly cleaned up in a previous commit by me. Revert to the
original, correct version, but with a cleaner implementation.
8) The cluster read code now tries to keep data associated with buffers
more aggressively (without breaking the heuristics) when it is presumed
that the read data (buffers) will be soon needed.
9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The
delay loop waiting is not useful for filesystem locks, due to the
length of the time intervals.
10) Correct and clean-up spec_getpages.
11) Implement a fully functional nfs_getpages, nfs_putpages.
12) Fix nfs_write so that modifications are coherent with the NFS data on
the server disk (at least as well as NFS seems to allow.)
13) Properly support MS_INVALIDATE on NFS.
14) Properly pass down MS_INVALIDATE to lower levels of the VM code from
vm_map_clean.
15) Better support the notion of pages being busy but valid, so that
fewer in-transit waits occur. (use p->busy more for pageouts instead
of PG_BUSY.) Since the page is fully valid, it is still usable for
reads.
16) It is possible (in error) for cached pages to be busy. Make the
page allocation code handle that case correctly. (It should probably
be a printf or panic, but I want the system to handle coding errors
robustly. I'll probably add a printf.)
17) Correct the design and usage of vm_page_sleep. It didn't handle
consistancy problems very well, so make the design a little less
lofty. After vm_page_sleep, if it ever blocked, it is still important
to relookup the page (if the object generation count changed), and
verify it's status (always.)
18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up.
19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush.
20) Fix vm_pager_put_pages and it's descendents to support an int flag
instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
|
|
|
static vm_page_t vm_page_select_cache __P((vm_object_t, vm_pindex_t));
|
1996-09-08 20:44:49 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Associated with page of user-allocatable memory is a
|
|
|
|
* page structure.
|
|
|
|
*/
|
|
|
|
|
1995-12-14 09:55:16 +00:00
|
|
|
static struct pglist *vm_page_buckets; /* Array of buckets */
|
|
|
|
static int vm_page_bucket_count; /* How big is array? */
|
1995-11-20 12:20:02 +00:00
|
|
|
static int vm_page_hash_mask; /* Mask for hash function */
|
1998-02-05 03:32:49 +00:00
|
|
|
static volatile int vm_page_bucket_generation;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1996-11-17 02:38:31 +00:00
|
|
|
struct pglist vm_page_queue_free[PQ_L2_SIZE] = {0};
|
|
|
|
struct pglist vm_page_queue_zero[PQ_L2_SIZE] = {0};
|
|
|
|
struct pglist vm_page_queue_active = {0};
|
|
|
|
struct pglist vm_page_queue_inactive = {0};
|
|
|
|
struct pglist vm_page_queue_cache[PQ_L2_SIZE] = {0};
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1998-02-09 06:11:36 +00:00
|
|
|
static int no_queue=0;
|
1996-01-19 04:00:31 +00:00
|
|
|
|
1996-11-17 02:38:31 +00:00
|
|
|
struct vpgqueues vm_page_queues[PQ_COUNT] = {0};
|
1998-02-09 06:11:36 +00:00
|
|
|
static int pqcnt[PQ_COUNT] = {0};
|
1996-09-08 20:44:49 +00:00
|
|
|
|
|
|
|
static void
|
|
|
|
vm_page_queue_init(void) {
|
|
|
|
int i;
|
|
|
|
|
|
|
|
vm_page_queues[PQ_NONE].pl = NULL;
|
|
|
|
vm_page_queues[PQ_NONE].cnt = &no_queue;
|
|
|
|
for(i=0;i<PQ_L2_SIZE;i++) {
|
|
|
|
vm_page_queues[PQ_FREE+i].pl = &vm_page_queue_free[i];
|
|
|
|
vm_page_queues[PQ_FREE+i].cnt = &cnt.v_free_count;
|
|
|
|
}
|
|
|
|
for(i=0;i<PQ_L2_SIZE;i++) {
|
|
|
|
vm_page_queues[PQ_ZERO+i].pl = &vm_page_queue_zero[i];
|
|
|
|
vm_page_queues[PQ_ZERO+i].cnt = &cnt.v_free_count;
|
|
|
|
}
|
|
|
|
vm_page_queues[PQ_INACTIVE].pl = &vm_page_queue_inactive;
|
|
|
|
vm_page_queues[PQ_INACTIVE].cnt = &cnt.v_inactive_count;
|
|
|
|
|
|
|
|
vm_page_queues[PQ_ACTIVE].pl = &vm_page_queue_active;
|
|
|
|
vm_page_queues[PQ_ACTIVE].cnt = &cnt.v_active_count;
|
|
|
|
for(i=0;i<PQ_L2_SIZE;i++) {
|
|
|
|
vm_page_queues[PQ_CACHE+i].pl = &vm_page_queue_cache[i];
|
|
|
|
vm_page_queues[PQ_CACHE+i].cnt = &cnt.v_cache_count;
|
|
|
|
}
|
|
|
|
for(i=0;i<PQ_COUNT;i++) {
|
|
|
|
if (vm_page_queues[i].pl) {
|
|
|
|
TAILQ_INIT(vm_page_queues[i].pl);
|
|
|
|
} else if (i != 0) {
|
|
|
|
panic("vm_page_queue_init: queue %d is null", i);
|
|
|
|
}
|
|
|
|
vm_page_queues[i].lcnt = &pqcnt[i];
|
|
|
|
}
|
|
|
|
}
|
1996-01-19 04:00:31 +00:00
|
|
|
|
1996-11-17 02:38:31 +00:00
|
|
|
vm_page_t vm_page_array = 0;
|
1998-02-09 06:11:36 +00:00
|
|
|
static int vm_page_array_size = 0;
|
1996-11-17 02:38:31 +00:00
|
|
|
long first_page = 0;
|
1995-12-14 09:55:16 +00:00
|
|
|
static long last_page;
|
|
|
|
static vm_size_t page_mask;
|
|
|
|
static int page_shift;
|
1996-11-17 02:38:31 +00:00
|
|
|
int vm_page_zero_count = 0;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* map of contiguous valid DEV_BSIZE chunks in a page
|
|
|
|
* (this list is valid for page sizes upto 16*DEV_BSIZE)
|
|
|
|
*/
|
|
|
|
static u_short vm_page_dev_bsize_chunks[] = {
|
|
|
|
0x0, 0x1, 0x3, 0x7, 0xf, 0x1f, 0x3f, 0x7f, 0xff,
|
|
|
|
0x1ff, 0x3ff, 0x7ff, 0xfff, 0x1fff, 0x3fff, 0x7fff, 0xffff
|
|
|
|
};
|
|
|
|
|
1998-04-15 17:47:40 +00:00
|
|
|
static __inline int vm_page_hash __P((vm_object_t object, vm_pindex_t pindex));
|
1996-06-05 03:31:49 +00:00
|
|
|
static int vm_page_freechk_and_unqueue __P((vm_page_t m));
|
|
|
|
static void vm_page_free_wakeup __P((void));
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* vm_set_page_size:
|
|
|
|
*
|
|
|
|
* Sets the page size, perhaps based upon the memory
|
|
|
|
* size. Must be called before any use of page-size
|
|
|
|
* dependent functions.
|
|
|
|
*
|
|
|
|
* Sets page_shift and page_mask from cnt.v_page_size.
|
|
|
|
*/
|
1995-05-30 08:16:23 +00:00
|
|
|
void
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
vm_set_page_size()
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
|
|
|
|
|
|
|
if (cnt.v_page_size == 0)
|
|
|
|
cnt.v_page_size = DEFAULT_PAGE_SIZE;
|
|
|
|
page_mask = cnt.v_page_size - 1;
|
|
|
|
if ((page_mask & cnt.v_page_size) != 0)
|
|
|
|
panic("vm_set_page_size: page size not a power of two");
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
for (page_shift = 0;; page_shift++)
|
1994-05-24 10:09:53 +00:00
|
|
|
if ((1 << page_shift) == cnt.v_page_size)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* vm_page_startup:
|
|
|
|
*
|
|
|
|
* Initializes the resident memory module.
|
|
|
|
*
|
|
|
|
* Allocates memory for the page cells, and
|
|
|
|
* for the object/offset-to-page hash table headers.
|
|
|
|
* Each page cell is initialized and placed on the free list.
|
|
|
|
*/
|
1994-05-25 09:21:21 +00:00
|
|
|
|
|
|
|
vm_offset_t
|
|
|
|
vm_page_startup(starta, enda, vaddr)
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
register vm_offset_t starta;
|
|
|
|
vm_offset_t enda;
|
|
|
|
register vm_offset_t vaddr;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
register vm_offset_t mapped;
|
|
|
|
register vm_page_t m;
|
1994-05-25 09:21:21 +00:00
|
|
|
register struct pglist *bucket;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
vm_size_t npages, page_range;
|
|
|
|
register vm_offset_t new_start;
|
|
|
|
int i;
|
|
|
|
vm_offset_t pa;
|
1994-05-25 09:21:21 +00:00
|
|
|
int nblocks;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
vm_offset_t first_managed_page;
|
1994-05-25 09:21:21 +00:00
|
|
|
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
/* the biggest memory array is the second group of pages */
|
1994-05-25 09:21:21 +00:00
|
|
|
vm_offset_t start;
|
|
|
|
vm_offset_t biggestone, biggestsize;
|
|
|
|
|
|
|
|
vm_offset_t total;
|
|
|
|
|
|
|
|
total = 0;
|
|
|
|
biggestsize = 0;
|
|
|
|
biggestone = 0;
|
|
|
|
nblocks = 0;
|
|
|
|
vaddr = round_page(vaddr);
|
|
|
|
|
|
|
|
for (i = 0; phys_avail[i + 1]; i += 2) {
|
|
|
|
phys_avail[i] = round_page(phys_avail[i]);
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
phys_avail[i + 1] = trunc_page(phys_avail[i + 1]);
|
1994-05-25 09:21:21 +00:00
|
|
|
}
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
|
1994-05-25 09:21:21 +00:00
|
|
|
for (i = 0; phys_avail[i + 1]; i += 2) {
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
int size = phys_avail[i + 1] - phys_avail[i];
|
|
|
|
|
1994-05-25 09:21:21 +00:00
|
|
|
if (size > biggestsize) {
|
|
|
|
biggestone = i;
|
|
|
|
biggestsize = size;
|
|
|
|
}
|
|
|
|
++nblocks;
|
|
|
|
total += size;
|
|
|
|
}
|
|
|
|
|
|
|
|
start = phys_avail[biggestone];
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Initialize the queue headers for the free queue, the active queue
|
|
|
|
* and the inactive queue.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1996-09-08 20:44:49 +00:00
|
|
|
vm_page_queue_init();
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Allocate (and initialize) the hash table buckets.
|
1995-05-30 08:16:23 +00:00
|
|
|
*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* The number of buckets MUST BE a power of 2, and the actual value is
|
|
|
|
* the next power of 2 greater than the number of physical pages in
|
|
|
|
* the system.
|
1995-05-30 08:16:23 +00:00
|
|
|
*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Note: This computation can be tweaked if desired.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
vm_page_buckets = (struct pglist *) vaddr;
|
1994-05-25 09:21:21 +00:00
|
|
|
bucket = vm_page_buckets;
|
1994-05-24 10:09:53 +00:00
|
|
|
if (vm_page_bucket_count == 0) {
|
1996-05-18 03:38:05 +00:00
|
|
|
vm_page_bucket_count = 1;
|
1994-05-25 09:21:21 +00:00
|
|
|
while (vm_page_bucket_count < atop(total))
|
1994-05-24 10:09:53 +00:00
|
|
|
vm_page_bucket_count <<= 1;
|
|
|
|
}
|
|
|
|
vm_page_hash_mask = vm_page_bucket_count - 1;
|
|
|
|
|
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Validate these addresses.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1994-05-25 09:21:21 +00:00
|
|
|
new_start = start + vm_page_bucket_count * sizeof(struct pglist);
|
|
|
|
new_start = round_page(new_start);
|
1998-06-02 05:50:08 +00:00
|
|
|
mapped = round_page(vaddr);
|
1994-05-25 09:21:21 +00:00
|
|
|
vaddr = pmap_map(mapped, start, new_start,
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
VM_PROT_READ | VM_PROT_WRITE);
|
1994-05-25 09:21:21 +00:00
|
|
|
start = new_start;
|
1998-06-02 05:50:08 +00:00
|
|
|
vaddr = round_page(vaddr);
|
1994-05-25 09:21:21 +00:00
|
|
|
bzero((caddr_t) mapped, vaddr - mapped);
|
|
|
|
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
for (i = 0; i < vm_page_bucket_count; i++) {
|
1994-05-24 10:09:53 +00:00
|
|
|
TAILQ_INIT(bucket);
|
|
|
|
bucket++;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Compute the number of pages of memory that will be available for
|
|
|
|
* use (taking into account the overhead of a page structure per
|
|
|
|
* page).
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1994-05-25 09:21:21 +00:00
|
|
|
first_page = phys_avail[0] / PAGE_SIZE;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
last_page = phys_avail[(nblocks - 1) * 2 + 1] / PAGE_SIZE;
|
|
|
|
|
|
|
|
page_range = last_page - (phys_avail[0] / PAGE_SIZE);
|
|
|
|
npages = (total - (page_range * sizeof(struct vm_page)) -
|
|
|
|
(start - phys_avail[biggestone])) / PAGE_SIZE;
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Initialize the mem entry structures now, and put them in the free
|
|
|
|
* queue.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
1994-05-25 09:21:21 +00:00
|
|
|
vm_page_array = (vm_page_t) vaddr;
|
|
|
|
mapped = vaddr;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Validate these addresses.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
new_start = round_page(start + page_range * sizeof(struct vm_page));
|
1994-05-25 09:21:21 +00:00
|
|
|
mapped = pmap_map(mapped, start, new_start,
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
VM_PROT_READ | VM_PROT_WRITE);
|
1994-05-25 09:21:21 +00:00
|
|
|
start = new_start;
|
|
|
|
|
|
|
|
first_managed_page = start / PAGE_SIZE;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Clear all of the page structures
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
bzero((caddr_t) vm_page_array, page_range * sizeof(struct vm_page));
|
|
|
|
vm_page_array_size = page_range;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1994-05-25 09:21:21 +00:00
|
|
|
cnt.v_page_count = 0;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
cnt.v_free_count = 0;
|
1994-05-25 09:21:21 +00:00
|
|
|
for (i = 0; phys_avail[i + 1] && npages > 0; i += 2) {
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
if (i == biggestone)
|
1994-05-25 09:21:21 +00:00
|
|
|
pa = ptoa(first_managed_page);
|
|
|
|
else
|
|
|
|
pa = phys_avail[i];
|
|
|
|
while (pa < phys_avail[i + 1] && npages-- > 0) {
|
|
|
|
++cnt.v_page_count;
|
|
|
|
++cnt.v_free_count;
|
|
|
|
m = PHYS_TO_VM_PAGE(pa);
|
|
|
|
m->phys_addr = pa;
|
1996-09-08 20:44:49 +00:00
|
|
|
m->flags = 0;
|
|
|
|
m->pc = (pa >> PAGE_SHIFT) & PQ_L2_MASK;
|
1998-06-02 05:50:08 +00:00
|
|
|
m->queue = m->pc + PQ_FREE;
|
1996-09-08 20:44:49 +00:00
|
|
|
TAILQ_INSERT_TAIL(vm_page_queues[m->queue].pl, m, pageq);
|
|
|
|
++(*vm_page_queues[m->queue].lcnt);
|
1994-05-25 09:21:21 +00:00
|
|
|
pa += PAGE_SIZE;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
}
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
return (mapped);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* vm_page_hash:
|
|
|
|
*
|
|
|
|
* Distributes the object/offset key pair among hash buckets.
|
|
|
|
*
|
|
|
|
* NOTE: This macro depends on vm_page_bucket_count being a power of 2.
|
|
|
|
*/
|
1998-04-15 17:47:40 +00:00
|
|
|
static __inline int
|
1995-12-11 04:58:34 +00:00
|
|
|
vm_page_hash(object, pindex)
|
|
|
|
vm_object_t object;
|
|
|
|
vm_pindex_t pindex;
|
1994-05-25 09:21:21 +00:00
|
|
|
{
|
1998-07-15 04:17:55 +00:00
|
|
|
return ((((uintptr_t) object) >> 5) + (pindex >> 1)) & vm_page_hash_mask;
|
1994-05-25 09:21:21 +00:00
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* vm_page_insert: [ internal use only ]
|
|
|
|
*
|
|
|
|
* Inserts the given mem entry into the object/object-page
|
|
|
|
* table and object list.
|
|
|
|
*
|
1995-03-01 23:30:04 +00:00
|
|
|
* The object and page must be locked, and must be splhigh.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1996-09-08 20:44:49 +00:00
|
|
|
void
|
1996-01-19 04:00:31 +00:00
|
|
|
vm_page_insert(m, object, pindex)
|
|
|
|
register vm_page_t m;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
register vm_object_t object;
|
1995-12-11 04:58:34 +00:00
|
|
|
register vm_pindex_t pindex;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
register struct pglist *bucket;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1998-10-21 14:46:42 +00:00
|
|
|
if (m->object != NULL)
|
1994-05-24 10:09:53 +00:00
|
|
|
panic("vm_page_insert: already inserted");
|
|
|
|
|
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Record the object/offset pair in this page
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1996-01-19 04:00:31 +00:00
|
|
|
m->object = object;
|
|
|
|
m->pindex = pindex;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Insert it into the object_object/offset hash table
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1995-12-11 04:58:34 +00:00
|
|
|
bucket = &vm_page_buckets[vm_page_hash(object, pindex)];
|
1996-01-19 04:00:31 +00:00
|
|
|
TAILQ_INSERT_TAIL(bucket, m, hashq);
|
VM level code cleanups.
1) Start using TSM.
Struct procs continue to point to upages structure, after being freed.
Struct vmspace continues to point to pte object and kva space for kstack.
u_map is now superfluous.
2) vm_map's don't need to be reference counted. They always exist either
in the kernel or in a vmspace. The vmspaces are managed by reference
counts.
3) Remove the "wired" vm_map nonsense.
4) No need to keep a cache of kernel stack kva's.
5) Get rid of strange looking ++var, and change to var++.
6) Change more data structures to use our "zone" allocator. Added
struct proc, struct vmspace and struct vnode. This saves a significant
amount of kva space and physical memory. Additionally, this enables
TSM for the zone managed memory.
7) Keep ioopt disabled for now.
8) Remove the now bogus "single use" map concept.
9) Use generation counts or id's for data structures residing in TSM, where
it allows us to avoid unneeded restart overhead during traversals, where
blocking might occur.
10) Account better for memory deficits, so the pageout daemon will be able
to make enough memory available (experimental.)
11) Fix some vnode locking problems. (From Tor, I think.)
12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp.
(experimental.)
13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c
code. Use generation counts, get rid of unneded collpase operations,
and clean up the cluster code.
14) Make vm_zone more suitable for TSM.
This commit is partially as a result of discussions and contributions from
other people, including DG, Tor Egge, PHK, and probably others that I
have forgotten to attribute (so let me know, if I forgot.)
This is not the infamous, final cleanup of the vnode stuff, but a necessary
step. Vnode mgmt should be correct, but things might still change, and
there is still some missing stuff (like ioopt, and physical backing of
non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
|
|
|
vm_page_bucket_generation++;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Now link into the object's list of backed pages.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1996-01-19 04:00:31 +00:00
|
|
|
TAILQ_INSERT_TAIL(&object->memq, m, listq);
|
1996-09-28 03:33:40 +00:00
|
|
|
m->object->page_hint = m;
|
VM level code cleanups.
1) Start using TSM.
Struct procs continue to point to upages structure, after being freed.
Struct vmspace continues to point to pte object and kva space for kstack.
u_map is now superfluous.
2) vm_map's don't need to be reference counted. They always exist either
in the kernel or in a vmspace. The vmspaces are managed by reference
counts.
3) Remove the "wired" vm_map nonsense.
4) No need to keep a cache of kernel stack kva's.
5) Get rid of strange looking ++var, and change to var++.
6) Change more data structures to use our "zone" allocator. Added
struct proc, struct vmspace and struct vnode. This saves a significant
amount of kva space and physical memory. Additionally, this enables
TSM for the zone managed memory.
7) Keep ioopt disabled for now.
8) Remove the now bogus "single use" map concept.
9) Use generation counts or id's for data structures residing in TSM, where
it allows us to avoid unneeded restart overhead during traversals, where
blocking might occur.
10) Account better for memory deficits, so the pageout daemon will be able
to make enough memory available (experimental.)
11) Fix some vnode locking problems. (From Tor, I think.)
12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp.
(experimental.)
13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c
code. Use generation counts, get rid of unneded collpase operations,
and clean up the cluster code.
14) Make vm_zone more suitable for TSM.
This commit is partially as a result of discussions and contributions from
other people, including DG, Tor Egge, PHK, and probably others that I
have forgotten to attribute (so let me know, if I forgot.)
This is not the infamous, final cleanup of the vnode stuff, but a necessary
step. Vnode mgmt should be correct, but things might still change, and
there is still some missing stuff (like ioopt, and physical backing of
non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
|
|
|
m->object->generation++;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1998-01-31 11:56:53 +00:00
|
|
|
if (m->wire_count)
|
|
|
|
object->wire_count++;
|
|
|
|
|
|
|
|
if ((m->queue - m->pc) == PQ_CACHE)
|
|
|
|
object->cache_count++;
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* And show that the object has one more resident page.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
|
|
|
object->resident_page_count++;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* vm_page_remove: [ internal use only ]
|
|
|
|
* NOTE: used by device pager as well -wfj
|
|
|
|
*
|
|
|
|
* Removes the given mem entry from the object/offset-page
|
|
|
|
* table and the object page list.
|
|
|
|
*
|
1995-03-01 23:30:04 +00:00
|
|
|
* The object and page must be locked, and at splhigh.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1996-09-08 20:44:49 +00:00
|
|
|
void
|
1996-01-19 04:00:31 +00:00
|
|
|
vm_page_remove(m)
|
|
|
|
register vm_page_t m;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
register struct pglist *bucket;
|
1998-02-05 03:32:49 +00:00
|
|
|
vm_object_t object;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1998-10-21 14:46:42 +00:00
|
|
|
if (m->object == NULL)
|
1994-05-24 10:09:53 +00:00
|
|
|
return;
|
|
|
|
|
1998-02-05 03:32:49 +00:00
|
|
|
#if !defined(MAX_PERF)
|
1998-01-31 11:56:53 +00:00
|
|
|
if ((m->flags & PG_BUSY) == 0) {
|
|
|
|
panic("vm_page_remove: page not busy");
|
|
|
|
}
|
1998-02-05 03:32:49 +00:00
|
|
|
#endif
|
1998-01-31 11:56:53 +00:00
|
|
|
|
1998-09-04 08:06:57 +00:00
|
|
|
vm_page_flag_clear(m, PG_BUSY);
|
1998-01-31 11:56:53 +00:00
|
|
|
if (m->flags & PG_WANTED) {
|
1998-09-04 08:06:57 +00:00
|
|
|
vm_page_flag_clear(m, PG_WANTED);
|
1998-01-31 11:56:53 +00:00
|
|
|
wakeup(m);
|
|
|
|
}
|
|
|
|
|
1998-02-05 03:32:49 +00:00
|
|
|
object = m->object;
|
|
|
|
if (object->page_hint == m)
|
|
|
|
object->page_hint = NULL;
|
1996-09-28 03:33:40 +00:00
|
|
|
|
1998-01-31 11:56:53 +00:00
|
|
|
if (m->wire_count)
|
1998-02-05 03:32:49 +00:00
|
|
|
object->wire_count--;
|
1998-01-31 11:56:53 +00:00
|
|
|
|
|
|
|
if ((m->queue - m->pc) == PQ_CACHE)
|
1998-02-05 03:32:49 +00:00
|
|
|
object->cache_count--;
|
1998-01-31 11:56:53 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Remove from the object_object/offset hash table
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1996-01-19 04:00:31 +00:00
|
|
|
bucket = &vm_page_buckets[vm_page_hash(m->object, m->pindex)];
|
|
|
|
TAILQ_REMOVE(bucket, m, hashq);
|
VM level code cleanups.
1) Start using TSM.
Struct procs continue to point to upages structure, after being freed.
Struct vmspace continues to point to pte object and kva space for kstack.
u_map is now superfluous.
2) vm_map's don't need to be reference counted. They always exist either
in the kernel or in a vmspace. The vmspaces are managed by reference
counts.
3) Remove the "wired" vm_map nonsense.
4) No need to keep a cache of kernel stack kva's.
5) Get rid of strange looking ++var, and change to var++.
6) Change more data structures to use our "zone" allocator. Added
struct proc, struct vmspace and struct vnode. This saves a significant
amount of kva space and physical memory. Additionally, this enables
TSM for the zone managed memory.
7) Keep ioopt disabled for now.
8) Remove the now bogus "single use" map concept.
9) Use generation counts or id's for data structures residing in TSM, where
it allows us to avoid unneeded restart overhead during traversals, where
blocking might occur.
10) Account better for memory deficits, so the pageout daemon will be able
to make enough memory available (experimental.)
11) Fix some vnode locking problems. (From Tor, I think.)
12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp.
(experimental.)
13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c
code. Use generation counts, get rid of unneded collpase operations,
and clean up the cluster code.
14) Make vm_zone more suitable for TSM.
This commit is partially as a result of discussions and contributions from
other people, including DG, Tor Egge, PHK, and probably others that I
have forgotten to attribute (so let me know, if I forgot.)
This is not the infamous, final cleanup of the vnode stuff, but a necessary
step. Vnode mgmt should be correct, but things might still change, and
there is still some missing stuff (like ioopt, and physical backing of
non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
|
|
|
vm_page_bucket_generation++;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Now remove from the object's list of backed pages.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1998-02-05 03:32:49 +00:00
|
|
|
TAILQ_REMOVE(&object->memq, m, listq);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* And show that the object has one fewer resident page.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1998-02-05 03:32:49 +00:00
|
|
|
object->resident_page_count--;
|
|
|
|
object->generation++;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1998-10-21 14:46:42 +00:00
|
|
|
m->object = NULL;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* vm_page_lookup:
|
|
|
|
*
|
|
|
|
* Returns the page associated with the object/offset
|
|
|
|
* pair specified; if none is found, NULL is returned.
|
|
|
|
*
|
|
|
|
* The object must be locked. No side effects.
|
|
|
|
*/
|
|
|
|
|
1995-05-30 08:16:23 +00:00
|
|
|
vm_page_t
|
1995-12-11 04:58:34 +00:00
|
|
|
vm_page_lookup(object, pindex)
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
register vm_object_t object;
|
1995-12-11 04:58:34 +00:00
|
|
|
register vm_pindex_t pindex;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
1996-01-19 04:00:31 +00:00
|
|
|
register vm_page_t m;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
register struct pglist *bucket;
|
1998-02-05 03:32:49 +00:00
|
|
|
int generation;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Search the hash table for this object/offset pair
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1998-02-05 03:32:49 +00:00
|
|
|
if (object->page_hint && (object->page_hint->pindex == pindex) &&
|
|
|
|
(object->page_hint->object == object))
|
|
|
|
return object->page_hint;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1998-02-05 03:32:49 +00:00
|
|
|
retry:
|
|
|
|
generation = vm_page_bucket_generation;
|
|
|
|
bucket = &vm_page_buckets[vm_page_hash(object, pindex)];
|
1996-05-18 03:38:05 +00:00
|
|
|
for (m = TAILQ_FIRST(bucket); m != NULL; m = TAILQ_NEXT(m,hashq)) {
|
1996-01-19 04:00:31 +00:00
|
|
|
if ((m->object == object) && (m->pindex == pindex)) {
|
1998-02-05 03:32:49 +00:00
|
|
|
if (vm_page_bucket_generation != generation)
|
|
|
|
goto retry;
|
1996-09-28 03:33:40 +00:00
|
|
|
m->object->page_hint = m;
|
1996-01-19 04:00:31 +00:00
|
|
|
return (m);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
}
|
1998-02-05 03:32:49 +00:00
|
|
|
if (vm_page_bucket_generation != generation)
|
|
|
|
goto retry;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
return (NULL);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* vm_page_rename:
|
|
|
|
*
|
|
|
|
* Move the given memory entry from its
|
|
|
|
* current object to the specified target object/offset.
|
|
|
|
*
|
|
|
|
* The object must be locked.
|
|
|
|
*/
|
1995-05-30 08:16:23 +00:00
|
|
|
void
|
1996-01-19 04:00:31 +00:00
|
|
|
vm_page_rename(m, new_object, new_pindex)
|
|
|
|
register vm_page_t m;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
register vm_object_t new_object;
|
1995-12-11 04:58:34 +00:00
|
|
|
vm_pindex_t new_pindex;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
int s;
|
|
|
|
|
1996-05-18 03:38:05 +00:00
|
|
|
s = splvm();
|
1996-01-19 04:00:31 +00:00
|
|
|
vm_page_remove(m);
|
|
|
|
vm_page_insert(m, new_object, new_pindex);
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
splx(s);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
1996-07-30 03:08:57 +00:00
|
|
|
/*
|
|
|
|
* vm_page_unqueue without any wakeup
|
|
|
|
*/
|
1996-09-08 20:44:49 +00:00
|
|
|
void
|
1996-07-30 03:08:57 +00:00
|
|
|
vm_page_unqueue_nowakeup(m)
|
|
|
|
vm_page_t m;
|
|
|
|
{
|
|
|
|
int queue = m->queue;
|
1996-09-28 03:33:40 +00:00
|
|
|
struct vpgqueues *pq;
|
1996-07-30 03:08:57 +00:00
|
|
|
if (queue != PQ_NONE) {
|
1996-09-28 03:33:40 +00:00
|
|
|
pq = &vm_page_queues[queue];
|
1996-07-30 03:08:57 +00:00
|
|
|
m->queue = PQ_NONE;
|
1996-09-28 03:33:40 +00:00
|
|
|
TAILQ_REMOVE(pq->pl, m, pageq);
|
1998-01-31 11:56:53 +00:00
|
|
|
(*pq->cnt)--;
|
|
|
|
(*pq->lcnt)--;
|
|
|
|
if ((queue - m->pc) == PQ_CACHE) {
|
1998-02-05 03:32:49 +00:00
|
|
|
if (m->object)
|
|
|
|
m->object->cache_count--;
|
1998-01-31 11:56:53 +00:00
|
|
|
}
|
1996-07-30 03:08:57 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
1995-03-01 23:30:04 +00:00
|
|
|
/*
|
|
|
|
* vm_page_unqueue must be called at splhigh();
|
|
|
|
*/
|
1996-09-08 20:44:49 +00:00
|
|
|
void
|
1996-07-30 03:08:57 +00:00
|
|
|
vm_page_unqueue(m)
|
1996-06-05 03:31:49 +00:00
|
|
|
vm_page_t m;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
{
|
1996-01-19 04:00:31 +00:00
|
|
|
int queue = m->queue;
|
1996-09-28 03:33:40 +00:00
|
|
|
struct vpgqueues *pq;
|
1996-06-05 03:31:49 +00:00
|
|
|
if (queue != PQ_NONE) {
|
|
|
|
m->queue = PQ_NONE;
|
1996-09-28 03:33:40 +00:00
|
|
|
pq = &vm_page_queues[queue];
|
|
|
|
TAILQ_REMOVE(pq->pl, m, pageq);
|
1998-01-31 11:56:53 +00:00
|
|
|
(*pq->cnt)--;
|
|
|
|
(*pq->lcnt)--;
|
1997-05-01 14:36:01 +00:00
|
|
|
if ((queue - m->pc) == PQ_CACHE) {
|
1996-06-05 03:31:49 +00:00
|
|
|
if ((cnt.v_cache_count + cnt.v_free_count) <
|
|
|
|
(cnt.v_free_reserved + cnt.v_cache_min))
|
|
|
|
pagedaemon_wakeup();
|
1998-02-05 03:32:49 +00:00
|
|
|
if (m->object)
|
|
|
|
m->object->cache_count--;
|
1996-06-05 03:31:49 +00:00
|
|
|
}
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
1996-09-08 20:44:49 +00:00
|
|
|
/*
|
|
|
|
* Find a page on the specified queue with color optimization.
|
|
|
|
*/
|
|
|
|
vm_page_t
|
|
|
|
vm_page_list_find(basequeue, index)
|
|
|
|
int basequeue, index;
|
|
|
|
{
|
1996-09-28 03:33:40 +00:00
|
|
|
#if PQ_L2_SIZE > 1
|
|
|
|
|
1996-09-08 20:44:49 +00:00
|
|
|
int i,j;
|
|
|
|
vm_page_t m;
|
|
|
|
int hindex;
|
1997-11-06 08:35:50 +00:00
|
|
|
struct vpgqueues *pq;
|
|
|
|
|
|
|
|
pq = &vm_page_queues[basequeue];
|
|
|
|
|
|
|
|
m = TAILQ_FIRST(pq[index].pl);
|
|
|
|
if (m)
|
|
|
|
return m;
|
1996-09-08 20:44:49 +00:00
|
|
|
|
|
|
|
for(j = 0; j < PQ_L1_SIZE; j++) {
|
1997-11-06 08:35:50 +00:00
|
|
|
int ij;
|
|
|
|
for(i = (PQ_L2_SIZE / 2) - PQ_L1_SIZE;
|
|
|
|
(ij = i + j) > 0;
|
1996-09-28 03:33:40 +00:00
|
|
|
i -= PQ_L1_SIZE) {
|
1997-11-06 08:35:50 +00:00
|
|
|
|
|
|
|
hindex = index + ij;
|
|
|
|
if (hindex >= PQ_L2_SIZE)
|
|
|
|
hindex -= PQ_L2_SIZE;
|
|
|
|
if (m = TAILQ_FIRST(pq[hindex].pl))
|
1996-09-08 20:44:49 +00:00
|
|
|
return m;
|
|
|
|
|
1997-11-06 08:35:50 +00:00
|
|
|
hindex = index - ij;
|
|
|
|
if (hindex < 0)
|
|
|
|
hindex += PQ_L2_SIZE;
|
|
|
|
if (m = TAILQ_FIRST(pq[hindex].pl))
|
1996-09-08 20:44:49 +00:00
|
|
|
return m;
|
|
|
|
}
|
|
|
|
}
|
1997-11-06 08:35:50 +00:00
|
|
|
|
|
|
|
hindex = index + PQ_L2_SIZE / 2;
|
|
|
|
if (hindex >= PQ_L2_SIZE)
|
|
|
|
hindex -= PQ_L2_SIZE;
|
|
|
|
m = TAILQ_FIRST(pq[hindex].pl);
|
|
|
|
if (m)
|
|
|
|
return m;
|
|
|
|
|
1996-09-08 20:44:49 +00:00
|
|
|
return NULL;
|
|
|
|
#else
|
|
|
|
return TAILQ_FIRST(vm_page_queues[basequeue].pl);
|
|
|
|
#endif
|
|
|
|
|
|
|
|
}
|
|
|
|
|
1996-09-28 03:33:40 +00:00
|
|
|
/*
|
|
|
|
* Find a page on the specified queue with color optimization.
|
|
|
|
*/
|
|
|
|
vm_page_t
|
|
|
|
vm_page_select(object, pindex, basequeue)
|
|
|
|
vm_object_t object;
|
|
|
|
vm_pindex_t pindex;
|
|
|
|
int basequeue;
|
|
|
|
{
|
|
|
|
|
|
|
|
#if PQ_L2_SIZE > 1
|
|
|
|
int index;
|
|
|
|
index = (pindex + object->pg_color) & PQ_L2_MASK;
|
|
|
|
return vm_page_list_find(basequeue, index);
|
|
|
|
|
|
|
|
#else
|
|
|
|
return TAILQ_FIRST(vm_page_queues[basequeue].pl);
|
|
|
|
#endif
|
|
|
|
|
|
|
|
}
|
|
|
|
|
This mega-commit is meant to fix numerous interrelated problems. There
has been some bitrot and incorrect assumptions in the vfs_bio code. These
problems have manifest themselves worse on NFS type filesystems, but can
still affect local filesystems under certain circumstances. Most of
the problems have involved mmap consistancy, and as a side-effect broke
the vfs.ioopt code. This code might have been committed seperately, but
almost everything is interrelated.
1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that
are fully valid.
2) Rather than deactivating erroneously read initial (header) pages in
kern_exec, we now free them.
3) Fix the rundown of non-VMIO buffers that are in an inconsistent
(missing vp) state.
4) Fix the disassociation of pages from buffers in brelse. The previous
code had rotted and was faulty in a couple of important circumstances.
5) Remove a gratuitious buffer wakeup in vfs_vmio_release.
6) Remove a crufty and currently unused cluster mechanism for VBLK
files in vfs_bio_awrite. When the code is functional, I'll add back
a cleaner version.
7) The page busy count wakeups assocated with the buffer cache usage were
incorrectly cleaned up in a previous commit by me. Revert to the
original, correct version, but with a cleaner implementation.
8) The cluster read code now tries to keep data associated with buffers
more aggressively (without breaking the heuristics) when it is presumed
that the read data (buffers) will be soon needed.
9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The
delay loop waiting is not useful for filesystem locks, due to the
length of the time intervals.
10) Correct and clean-up spec_getpages.
11) Implement a fully functional nfs_getpages, nfs_putpages.
12) Fix nfs_write so that modifications are coherent with the NFS data on
the server disk (at least as well as NFS seems to allow.)
13) Properly support MS_INVALIDATE on NFS.
14) Properly pass down MS_INVALIDATE to lower levels of the VM code from
vm_map_clean.
15) Better support the notion of pages being busy but valid, so that
fewer in-transit waits occur. (use p->busy more for pageouts instead
of PG_BUSY.) Since the page is fully valid, it is still usable for
reads.
16) It is possible (in error) for cached pages to be busy. Make the
page allocation code handle that case correctly. (It should probably
be a printf or panic, but I want the system to handle coding errors
robustly. I'll probably add a printf.)
17) Correct the design and usage of vm_page_sleep. It didn't handle
consistancy problems very well, so make the design a little less
lofty. After vm_page_sleep, if it ever blocked, it is still important
to relookup the page (if the object generation count changed), and
verify it's status (always.)
18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up.
19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush.
20) Fix vm_pager_put_pages and it's descendents to support an int flag
instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
|
|
|
/*
|
|
|
|
* Find a page on the cache queue with color optimization. As pages
|
|
|
|
* might be found, but not applicable, they are deactivated. This
|
|
|
|
* keeps us from using potentially busy cached pages.
|
|
|
|
*/
|
|
|
|
vm_page_t
|
|
|
|
vm_page_select_cache(object, pindex)
|
|
|
|
vm_object_t object;
|
|
|
|
vm_pindex_t pindex;
|
|
|
|
{
|
|
|
|
vm_page_t m;
|
|
|
|
|
|
|
|
while (TRUE) {
|
|
|
|
#if PQ_L2_SIZE > 1
|
|
|
|
int index;
|
|
|
|
index = (pindex + object->pg_color) & PQ_L2_MASK;
|
|
|
|
m = vm_page_list_find(PQ_CACHE, index);
|
|
|
|
|
|
|
|
#else
|
|
|
|
m = TAILQ_FIRST(vm_page_queues[PQ_CACHE].pl);
|
|
|
|
#endif
|
|
|
|
if (m && ((m->flags & PG_BUSY) || m->busy ||
|
|
|
|
m->hold_count || m->wire_count)) {
|
|
|
|
vm_page_deactivate(m);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
return m;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
1996-09-08 20:44:49 +00:00
|
|
|
/*
|
|
|
|
* Find a free or zero page, with specified preference.
|
|
|
|
*/
|
1996-09-14 11:54:59 +00:00
|
|
|
static vm_page_t
|
1996-09-08 20:44:49 +00:00
|
|
|
vm_page_select_free(object, pindex, prefqueue)
|
|
|
|
vm_object_t object;
|
|
|
|
vm_pindex_t pindex;
|
|
|
|
int prefqueue;
|
|
|
|
{
|
1996-09-28 03:33:40 +00:00
|
|
|
#if PQ_L2_SIZE > 1
|
|
|
|
int i,j;
|
1996-09-08 20:44:49 +00:00
|
|
|
int index, hindex;
|
1996-09-28 03:33:40 +00:00
|
|
|
#endif
|
1997-11-06 08:35:50 +00:00
|
|
|
vm_page_t m, mh;
|
1996-09-28 17:53:18 +00:00
|
|
|
int oqueuediff;
|
1997-11-06 08:35:50 +00:00
|
|
|
struct vpgqueues *pq;
|
1996-09-08 20:44:49 +00:00
|
|
|
|
|
|
|
if (prefqueue == PQ_ZERO)
|
|
|
|
oqueuediff = PQ_FREE - PQ_ZERO;
|
|
|
|
else
|
|
|
|
oqueuediff = PQ_ZERO - PQ_FREE;
|
|
|
|
|
1997-11-06 08:35:50 +00:00
|
|
|
if (mh = object->page_hint) {
|
|
|
|
if (mh->pindex == (pindex - 1)) {
|
|
|
|
if ((mh->flags & PG_FICTITIOUS) == 0) {
|
|
|
|
if ((mh < &vm_page_array[cnt.v_page_count-1]) &&
|
|
|
|
(mh >= &vm_page_array[0])) {
|
1996-09-28 03:33:40 +00:00
|
|
|
int queue;
|
1997-11-06 08:35:50 +00:00
|
|
|
m = mh + 1;
|
|
|
|
if (VM_PAGE_TO_PHYS(m) == (VM_PAGE_TO_PHYS(mh) + PAGE_SIZE)) {
|
|
|
|
queue = m->queue - m->pc;
|
|
|
|
if (queue == PQ_FREE || queue == PQ_ZERO) {
|
|
|
|
return m;
|
|
|
|
}
|
1996-09-28 03:33:40 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
1997-11-06 08:35:50 +00:00
|
|
|
|
|
|
|
pq = &vm_page_queues[prefqueue];
|
1996-09-28 03:33:40 +00:00
|
|
|
|
1996-09-08 20:44:49 +00:00
|
|
|
#if PQ_L2_SIZE > 1
|
|
|
|
|
1997-11-06 08:35:50 +00:00
|
|
|
index = (pindex + object->pg_color) & PQ_L2_MASK;
|
|
|
|
|
|
|
|
if (m = TAILQ_FIRST(pq[index].pl))
|
|
|
|
return m;
|
|
|
|
if (m = TAILQ_FIRST(pq[index + oqueuediff].pl))
|
|
|
|
return m;
|
|
|
|
|
1996-09-08 20:44:49 +00:00
|
|
|
for(j = 0; j < PQ_L1_SIZE; j++) {
|
1997-11-06 08:35:50 +00:00
|
|
|
int ij;
|
|
|
|
for(i = (PQ_L2_SIZE / 2) - PQ_L1_SIZE;
|
|
|
|
(ij = i + j) >= 0;
|
1996-09-08 20:44:49 +00:00
|
|
|
i -= PQ_L1_SIZE) {
|
1996-09-28 03:33:40 +00:00
|
|
|
|
1997-11-06 08:35:50 +00:00
|
|
|
hindex = index + ij;
|
|
|
|
if (hindex >= PQ_L2_SIZE)
|
|
|
|
hindex -= PQ_L2_SIZE;
|
|
|
|
if (m = TAILQ_FIRST(pq[hindex].pl))
|
1996-09-28 03:33:40 +00:00
|
|
|
return m;
|
1997-11-06 08:35:50 +00:00
|
|
|
if (m = TAILQ_FIRST(pq[hindex + oqueuediff].pl))
|
1996-09-28 03:33:40 +00:00
|
|
|
return m;
|
|
|
|
|
1997-11-06 08:35:50 +00:00
|
|
|
hindex = index - ij;
|
|
|
|
if (hindex < 0)
|
|
|
|
hindex += PQ_L2_SIZE;
|
|
|
|
if (m = TAILQ_FIRST(pq[hindex].pl))
|
1996-09-28 03:33:40 +00:00
|
|
|
return m;
|
1997-11-06 08:35:50 +00:00
|
|
|
if (m = TAILQ_FIRST(pq[hindex + oqueuediff].pl))
|
1996-09-28 03:33:40 +00:00
|
|
|
return m;
|
1996-09-08 20:44:49 +00:00
|
|
|
}
|
|
|
|
}
|
1997-11-06 08:35:50 +00:00
|
|
|
|
|
|
|
hindex = index + PQ_L2_SIZE / 2;
|
|
|
|
if (hindex >= PQ_L2_SIZE)
|
|
|
|
hindex -= PQ_L2_SIZE;
|
|
|
|
if (m = TAILQ_FIRST(pq[hindex].pl))
|
|
|
|
return m;
|
|
|
|
if (m = TAILQ_FIRST(pq[hindex+oqueuediff].pl))
|
|
|
|
return m;
|
|
|
|
|
1996-09-08 20:44:49 +00:00
|
|
|
#else
|
1997-11-06 08:35:50 +00:00
|
|
|
if (m = TAILQ_FIRST(pq[0].pl))
|
1996-09-08 20:44:49 +00:00
|
|
|
return m;
|
|
|
|
else
|
1997-11-06 08:35:50 +00:00
|
|
|
return TAILQ_FIRST(pq[oqueuediff].pl);
|
1996-09-08 20:44:49 +00:00
|
|
|
#endif
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* vm_page_alloc:
|
|
|
|
*
|
|
|
|
* Allocate and return a memory cell associated
|
|
|
|
* with this VM object/offset pair.
|
|
|
|
*
|
1995-03-01 23:30:04 +00:00
|
|
|
* page_req classes:
|
|
|
|
* VM_ALLOC_NORMAL normal process request
|
|
|
|
* VM_ALLOC_SYSTEM system *really* needs a page
|
|
|
|
* VM_ALLOC_INTERRUPT interrupt time request
|
1995-09-03 20:40:43 +00:00
|
|
|
* VM_ALLOC_ZERO zero page
|
1995-01-24 10:14:09 +00:00
|
|
|
*
|
1994-05-24 10:09:53 +00:00
|
|
|
* Object must be locked.
|
|
|
|
*/
|
1994-05-25 09:21:21 +00:00
|
|
|
vm_page_t
|
1995-12-11 04:58:34 +00:00
|
|
|
vm_page_alloc(object, pindex, page_req)
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
vm_object_t object;
|
1995-12-11 04:58:34 +00:00
|
|
|
vm_pindex_t pindex;
|
1995-01-24 10:14:09 +00:00
|
|
|
int page_req;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
1996-01-19 04:00:31 +00:00
|
|
|
register vm_page_t m;
|
1996-09-28 03:33:40 +00:00
|
|
|
struct vpgqueues *pq;
|
1998-01-12 01:46:33 +00:00
|
|
|
vm_object_t oldobject;
|
1997-03-23 02:44:54 +00:00
|
|
|
int queue, qtype;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
int s;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1995-12-11 15:41:50 +00:00
|
|
|
#ifdef DIAGNOSTIC
|
1996-01-19 04:00:31 +00:00
|
|
|
m = vm_page_lookup(object, pindex);
|
|
|
|
if (m)
|
NOTE: libkvm, w, ps, 'top', and any other utility which depends on struct
proc or any VM system structure will have to be rebuilt!!!
Much needed overhaul of the VM system. Included in this first round of
changes:
1) Improved pager interfaces: init, alloc, dealloc, getpages, putpages,
haspage, and sync operations are supported. The haspage interface now
provides information about clusterability. All pager routines now take
struct vm_object's instead of "pagers".
2) Improved data structures. In the previous paradigm, there is constant
confusion caused by pagers being both a data structure ("allocate a
pager") and a collection of routines. The idea of a pager structure has
escentially been eliminated. Objects now have types, and this type is
used to index the appropriate pager. In most cases, items in the pager
structure were duplicated in the object data structure and thus were
unnecessary. In the few cases that remained, a un_pager structure union
was created in the object to contain these items.
3) Because of the cleanup of #1 & #2, a lot of unnecessary layering can now
be removed. For instance, vm_object_enter(), vm_object_lookup(),
vm_object_remove(), and the associated object hash list were some of the
things that were removed.
4) simple_lock's removed. Discussion with several people reveals that the
SMP locking primitives used in the VM system aren't likely the mechanism
that we'll be adopting. Even if it were, the locking that was in the code
was very inadequate and would have to be mostly re-done anyway. The
locking in a uni-processor kernel was a no-op but went a long way toward
making the code difficult to read and debug.
5) Places that attempted to kludge-up the fact that we don't have kernel
thread support have been fixed to reflect the reality that we are really
dealing with processes, not threads. The VM system didn't have complete
thread support, so the comments and mis-named routines were just wrong.
We now use tsleep and wakeup directly in the lock routines, for instance.
6) Where appropriate, the pagers have been improved, especially in the
pager_alloc routines. Most of the pager_allocs have been rewritten and
are now faster and easier to maintain.
7) The pagedaemon pageout clustering algorithm has been rewritten and
now tries harder to output an even number of pages before and after
the requested page. This is sort of the reverse of the ideal pagein
algorithm and should provide better overall performance.
8) Unnecessary (incorrect) casts to caddr_t in calls to tsleep & wakeup
have been removed. Some other unnecessary casts have also been removed.
9) Some almost useless debugging code removed.
10) Terminology of shadow objects vs. backing objects straightened out.
The fact that the vm_object data structure escentially had this
backwards really confused things. The use of "shadow" and "backing
object" throughout the code is now internally consistent and correct
in the Mach terminology.
11) Several minor bug fixes, including one in the vm daemon that caused
0 RSS objects to not get purged as intended.
12) A "default pager" has now been created which cleans up the transition
of objects to the "swap" type. The previous checks throughout the code
for swp->pg_data != NULL were really ugly. This change also provides
the rudiments for future backing of "anonymous" memory by something
other than the swap pager (via the vnode pager, for example), and it
allows the decision about which of these pagers to use to be made
dynamically (although will need some additional decision code to do
this, of course).
13) (dyson) MAP_COPY has been deprecated and the corresponding "copy
object" code has been removed. MAP_COPY was undocumented and non-
standard. It was furthermore broken in several ways which caused its
behavior to degrade to MAP_PRIVATE. Binaries that use MAP_COPY will
continue to work correctly, but via the slightly different semantics
of MAP_PRIVATE.
14) (dyson) Sharing maps have been removed. It's marginal usefulness in a
threads design can be worked around in other ways. Both #12 and #13
were done to simplify the code and improve readability and maintain-
ability. (As were most all of these changes)
TODO:
1) Rewrite most of the vnode pager to use VOP_GETPAGES/PUTPAGES. Doing
this will reduce the vnode pager to a mere fraction of its current size.
2) Rewrite vm_fault and the swap/vnode pagers to use the clustering
information provided by the new haspage pager interface. This will
substantially reduce the overhead by eliminating a large number of
VOP_BMAP() calls. The VOP_BMAP() filesystem interface should be
improved to provide both a "behind" and "ahead" indication of
contiguousness.
3) Implement the extended features of pager_haspage in swap_pager_haspage().
It currently just says 0 pages ahead/behind.
4) Re-implement the swap device (swstrategy) in a more elegant way, perhaps
via a much more general mechanism that could also be used for disk
striping of regular filesystems.
5) Do something to improve the architecture of vm_object_collapse(). The
fact that it makes calls into the swap pager and knows too much about
how the swap pager operates really bothers me. It also doesn't allow
for collapsing of non-swap pager objects ("unnamed" objects backed by
other pagers).
1995-07-13 08:48:48 +00:00
|
|
|
panic("vm_page_alloc: page already allocated");
|
1995-12-11 15:41:50 +00:00
|
|
|
#endif
|
NOTE: libkvm, w, ps, 'top', and any other utility which depends on struct
proc or any VM system structure will have to be rebuilt!!!
Much needed overhaul of the VM system. Included in this first round of
changes:
1) Improved pager interfaces: init, alloc, dealloc, getpages, putpages,
haspage, and sync operations are supported. The haspage interface now
provides information about clusterability. All pager routines now take
struct vm_object's instead of "pagers".
2) Improved data structures. In the previous paradigm, there is constant
confusion caused by pagers being both a data structure ("allocate a
pager") and a collection of routines. The idea of a pager structure has
escentially been eliminated. Objects now have types, and this type is
used to index the appropriate pager. In most cases, items in the pager
structure were duplicated in the object data structure and thus were
unnecessary. In the few cases that remained, a un_pager structure union
was created in the object to contain these items.
3) Because of the cleanup of #1 & #2, a lot of unnecessary layering can now
be removed. For instance, vm_object_enter(), vm_object_lookup(),
vm_object_remove(), and the associated object hash list were some of the
things that were removed.
4) simple_lock's removed. Discussion with several people reveals that the
SMP locking primitives used in the VM system aren't likely the mechanism
that we'll be adopting. Even if it were, the locking that was in the code
was very inadequate and would have to be mostly re-done anyway. The
locking in a uni-processor kernel was a no-op but went a long way toward
making the code difficult to read and debug.
5) Places that attempted to kludge-up the fact that we don't have kernel
thread support have been fixed to reflect the reality that we are really
dealing with processes, not threads. The VM system didn't have complete
thread support, so the comments and mis-named routines were just wrong.
We now use tsleep and wakeup directly in the lock routines, for instance.
6) Where appropriate, the pagers have been improved, especially in the
pager_alloc routines. Most of the pager_allocs have been rewritten and
are now faster and easier to maintain.
7) The pagedaemon pageout clustering algorithm has been rewritten and
now tries harder to output an even number of pages before and after
the requested page. This is sort of the reverse of the ideal pagein
algorithm and should provide better overall performance.
8) Unnecessary (incorrect) casts to caddr_t in calls to tsleep & wakeup
have been removed. Some other unnecessary casts have also been removed.
9) Some almost useless debugging code removed.
10) Terminology of shadow objects vs. backing objects straightened out.
The fact that the vm_object data structure escentially had this
backwards really confused things. The use of "shadow" and "backing
object" throughout the code is now internally consistent and correct
in the Mach terminology.
11) Several minor bug fixes, including one in the vm daemon that caused
0 RSS objects to not get purged as intended.
12) A "default pager" has now been created which cleans up the transition
of objects to the "swap" type. The previous checks throughout the code
for swp->pg_data != NULL were really ugly. This change also provides
the rudiments for future backing of "anonymous" memory by something
other than the swap pager (via the vnode pager, for example), and it
allows the decision about which of these pagers to use to be made
dynamically (although will need some additional decision code to do
this, of course).
13) (dyson) MAP_COPY has been deprecated and the corresponding "copy
object" code has been removed. MAP_COPY was undocumented and non-
standard. It was furthermore broken in several ways which caused its
behavior to degrade to MAP_PRIVATE. Binaries that use MAP_COPY will
continue to work correctly, but via the slightly different semantics
of MAP_PRIVATE.
14) (dyson) Sharing maps have been removed. It's marginal usefulness in a
threads design can be worked around in other ways. Both #12 and #13
were done to simplify the code and improve readability and maintain-
ability. (As were most all of these changes)
TODO:
1) Rewrite most of the vnode pager to use VOP_GETPAGES/PUTPAGES. Doing
this will reduce the vnode pager to a mere fraction of its current size.
2) Rewrite vm_fault and the swap/vnode pagers to use the clustering
information provided by the new haspage pager interface. This will
substantially reduce the overhead by eliminating a large number of
VOP_BMAP() calls. The VOP_BMAP() filesystem interface should be
improved to provide both a "behind" and "ahead" indication of
contiguousness.
3) Implement the extended features of pager_haspage in swap_pager_haspage().
It currently just says 0 pages ahead/behind.
4) Re-implement the swap device (swstrategy) in a more elegant way, perhaps
via a much more general mechanism that could also be used for disk
striping of regular filesystems.
5) Do something to improve the architecture of vm_object_collapse(). The
fact that it makes calls into the swap pager and knows too much about
how the swap pager operates really bothers me. It also doesn't allow
for collapsing of non-swap pager objects ("unnamed" objects backed by
other pagers).
1995-07-13 08:48:48 +00:00
|
|
|
|
1995-03-01 23:30:04 +00:00
|
|
|
if ((curproc == pageproc) && (page_req != VM_ALLOC_INTERRUPT)) {
|
|
|
|
page_req = VM_ALLOC_SYSTEM;
|
|
|
|
};
|
1995-05-30 08:16:23 +00:00
|
|
|
|
1996-05-18 03:38:05 +00:00
|
|
|
s = splvm();
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
|
1996-01-19 04:00:31 +00:00
|
|
|
switch (page_req) {
|
|
|
|
|
1995-03-01 23:30:04 +00:00
|
|
|
case VM_ALLOC_NORMAL:
|
|
|
|
if (cnt.v_free_count >= cnt.v_free_reserved) {
|
1996-09-08 20:44:49 +00:00
|
|
|
m = vm_page_select_free(object, pindex, PQ_FREE);
|
1996-09-28 03:33:40 +00:00
|
|
|
#if defined(DIAGNOSTIC)
|
1996-09-08 20:44:49 +00:00
|
|
|
if (m == NULL)
|
|
|
|
panic("vm_page_alloc(NORMAL): missing page on free queue\n");
|
1996-09-28 03:33:40 +00:00
|
|
|
#endif
|
1995-03-01 23:30:04 +00:00
|
|
|
} else {
|
This mega-commit is meant to fix numerous interrelated problems. There
has been some bitrot and incorrect assumptions in the vfs_bio code. These
problems have manifest themselves worse on NFS type filesystems, but can
still affect local filesystems under certain circumstances. Most of
the problems have involved mmap consistancy, and as a side-effect broke
the vfs.ioopt code. This code might have been committed seperately, but
almost everything is interrelated.
1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that
are fully valid.
2) Rather than deactivating erroneously read initial (header) pages in
kern_exec, we now free them.
3) Fix the rundown of non-VMIO buffers that are in an inconsistent
(missing vp) state.
4) Fix the disassociation of pages from buffers in brelse. The previous
code had rotted and was faulty in a couple of important circumstances.
5) Remove a gratuitious buffer wakeup in vfs_vmio_release.
6) Remove a crufty and currently unused cluster mechanism for VBLK
files in vfs_bio_awrite. When the code is functional, I'll add back
a cleaner version.
7) The page busy count wakeups assocated with the buffer cache usage were
incorrectly cleaned up in a previous commit by me. Revert to the
original, correct version, but with a cleaner implementation.
8) The cluster read code now tries to keep data associated with buffers
more aggressively (without breaking the heuristics) when it is presumed
that the read data (buffers) will be soon needed.
9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The
delay loop waiting is not useful for filesystem locks, due to the
length of the time intervals.
10) Correct and clean-up spec_getpages.
11) Implement a fully functional nfs_getpages, nfs_putpages.
12) Fix nfs_write so that modifications are coherent with the NFS data on
the server disk (at least as well as NFS seems to allow.)
13) Properly support MS_INVALIDATE on NFS.
14) Properly pass down MS_INVALIDATE to lower levels of the VM code from
vm_map_clean.
15) Better support the notion of pages being busy but valid, so that
fewer in-transit waits occur. (use p->busy more for pageouts instead
of PG_BUSY.) Since the page is fully valid, it is still usable for
reads.
16) It is possible (in error) for cached pages to be busy. Make the
page allocation code handle that case correctly. (It should probably
be a printf or panic, but I want the system to handle coding errors
robustly. I'll probably add a printf.)
17) Correct the design and usage of vm_page_sleep. It didn't handle
consistancy problems very well, so make the design a little less
lofty. After vm_page_sleep, if it ever blocked, it is still important
to relookup the page (if the object generation count changed), and
verify it's status (always.)
18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up.
19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush.
20) Fix vm_pager_put_pages and it's descendents to support an int flag
instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
|
|
|
m = vm_page_select_cache(object, pindex);
|
1996-01-19 04:00:31 +00:00
|
|
|
if (m == NULL) {
|
|
|
|
splx(s);
|
1996-05-31 00:38:04 +00:00
|
|
|
#if defined(DIAGNOSTIC)
|
|
|
|
if (cnt.v_cache_count > 0)
|
|
|
|
printf("vm_page_alloc(NORMAL): missing pages on cache queue: %d\n", cnt.v_cache_count);
|
|
|
|
#endif
|
VM level code cleanups.
1) Start using TSM.
Struct procs continue to point to upages structure, after being freed.
Struct vmspace continues to point to pte object and kva space for kstack.
u_map is now superfluous.
2) vm_map's don't need to be reference counted. They always exist either
in the kernel or in a vmspace. The vmspaces are managed by reference
counts.
3) Remove the "wired" vm_map nonsense.
4) No need to keep a cache of kernel stack kva's.
5) Get rid of strange looking ++var, and change to var++.
6) Change more data structures to use our "zone" allocator. Added
struct proc, struct vmspace and struct vnode. This saves a significant
amount of kva space and physical memory. Additionally, this enables
TSM for the zone managed memory.
7) Keep ioopt disabled for now.
8) Remove the now bogus "single use" map concept.
9) Use generation counts or id's for data structures residing in TSM, where
it allows us to avoid unneeded restart overhead during traversals, where
blocking might occur.
10) Account better for memory deficits, so the pageout daemon will be able
to make enough memory available (experimental.)
11) Fix some vnode locking problems. (From Tor, I think.)
12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp.
(experimental.)
13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c
code. Use generation counts, get rid of unneded collpase operations,
and clean up the cluster code.
14) Make vm_zone more suitable for TSM.
This commit is partially as a result of discussions and contributions from
other people, including DG, Tor Egge, PHK, and probably others that I
have forgotten to attribute (so let me know, if I forgot.)
This is not the infamous, final cleanup of the vnode stuff, but a necessary
step. Vnode mgmt should be correct, but things might still change, and
there is still some missing stuff (like ioopt, and physical backing of
non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
|
|
|
vm_pageout_deficit++;
|
1996-01-19 04:00:31 +00:00
|
|
|
pagedaemon_wakeup();
|
|
|
|
return (NULL);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
|
|
|
|
case VM_ALLOC_ZERO:
|
|
|
|
if (cnt.v_free_count >= cnt.v_free_reserved) {
|
1996-09-08 20:44:49 +00:00
|
|
|
m = vm_page_select_free(object, pindex, PQ_ZERO);
|
1996-09-28 03:33:40 +00:00
|
|
|
#if defined(DIAGNOSTIC)
|
1996-09-08 20:44:49 +00:00
|
|
|
if (m == NULL)
|
|
|
|
panic("vm_page_alloc(ZERO): missing page on free queue\n");
|
1996-09-28 03:33:40 +00:00
|
|
|
#endif
|
1996-01-19 04:00:31 +00:00
|
|
|
} else {
|
This mega-commit is meant to fix numerous interrelated problems. There
has been some bitrot and incorrect assumptions in the vfs_bio code. These
problems have manifest themselves worse on NFS type filesystems, but can
still affect local filesystems under certain circumstances. Most of
the problems have involved mmap consistancy, and as a side-effect broke
the vfs.ioopt code. This code might have been committed seperately, but
almost everything is interrelated.
1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that
are fully valid.
2) Rather than deactivating erroneously read initial (header) pages in
kern_exec, we now free them.
3) Fix the rundown of non-VMIO buffers that are in an inconsistent
(missing vp) state.
4) Fix the disassociation of pages from buffers in brelse. The previous
code had rotted and was faulty in a couple of important circumstances.
5) Remove a gratuitious buffer wakeup in vfs_vmio_release.
6) Remove a crufty and currently unused cluster mechanism for VBLK
files in vfs_bio_awrite. When the code is functional, I'll add back
a cleaner version.
7) The page busy count wakeups assocated with the buffer cache usage were
incorrectly cleaned up in a previous commit by me. Revert to the
original, correct version, but with a cleaner implementation.
8) The cluster read code now tries to keep data associated with buffers
more aggressively (without breaking the heuristics) when it is presumed
that the read data (buffers) will be soon needed.
9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The
delay loop waiting is not useful for filesystem locks, due to the
length of the time intervals.
10) Correct and clean-up spec_getpages.
11) Implement a fully functional nfs_getpages, nfs_putpages.
12) Fix nfs_write so that modifications are coherent with the NFS data on
the server disk (at least as well as NFS seems to allow.)
13) Properly support MS_INVALIDATE on NFS.
14) Properly pass down MS_INVALIDATE to lower levels of the VM code from
vm_map_clean.
15) Better support the notion of pages being busy but valid, so that
fewer in-transit waits occur. (use p->busy more for pageouts instead
of PG_BUSY.) Since the page is fully valid, it is still usable for
reads.
16) It is possible (in error) for cached pages to be busy. Make the
page allocation code handle that case correctly. (It should probably
be a printf or panic, but I want the system to handle coding errors
robustly. I'll probably add a printf.)
17) Correct the design and usage of vm_page_sleep. It didn't handle
consistancy problems very well, so make the design a little less
lofty. After vm_page_sleep, if it ever blocked, it is still important
to relookup the page (if the object generation count changed), and
verify it's status (always.)
18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up.
19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush.
20) Fix vm_pager_put_pages and it's descendents to support an int flag
instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
|
|
|
m = vm_page_select_cache(object, pindex);
|
1996-01-19 04:00:31 +00:00
|
|
|
if (m == NULL) {
|
1995-03-01 23:30:04 +00:00
|
|
|
splx(s);
|
1996-05-31 00:38:04 +00:00
|
|
|
#if defined(DIAGNOSTIC)
|
|
|
|
if (cnt.v_cache_count > 0)
|
|
|
|
printf("vm_page_alloc(ZERO): missing pages on cache queue: %d\n", cnt.v_cache_count);
|
|
|
|
#endif
|
VM level code cleanups.
1) Start using TSM.
Struct procs continue to point to upages structure, after being freed.
Struct vmspace continues to point to pte object and kva space for kstack.
u_map is now superfluous.
2) vm_map's don't need to be reference counted. They always exist either
in the kernel or in a vmspace. The vmspaces are managed by reference
counts.
3) Remove the "wired" vm_map nonsense.
4) No need to keep a cache of kernel stack kva's.
5) Get rid of strange looking ++var, and change to var++.
6) Change more data structures to use our "zone" allocator. Added
struct proc, struct vmspace and struct vnode. This saves a significant
amount of kva space and physical memory. Additionally, this enables
TSM for the zone managed memory.
7) Keep ioopt disabled for now.
8) Remove the now bogus "single use" map concept.
9) Use generation counts or id's for data structures residing in TSM, where
it allows us to avoid unneeded restart overhead during traversals, where
blocking might occur.
10) Account better for memory deficits, so the pageout daemon will be able
to make enough memory available (experimental.)
11) Fix some vnode locking problems. (From Tor, I think.)
12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp.
(experimental.)
13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c
code. Use generation counts, get rid of unneded collpase operations,
and clean up the cluster code.
14) Make vm_zone more suitable for TSM.
This commit is partially as a result of discussions and contributions from
other people, including DG, Tor Egge, PHK, and probably others that I
have forgotten to attribute (so let me know, if I forgot.)
This is not the infamous, final cleanup of the vnode stuff, but a necessary
step. Vnode mgmt should be correct, but things might still change, and
there is still some missing stuff (like ioopt, and physical backing of
non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
|
|
|
vm_pageout_deficit++;
|
1995-03-01 23:30:04 +00:00
|
|
|
pagedaemon_wakeup();
|
|
|
|
return (NULL);
|
1995-02-02 09:09:15 +00:00
|
|
|
}
|
1995-03-01 23:30:04 +00:00
|
|
|
}
|
|
|
|
break;
|
|
|
|
|
|
|
|
case VM_ALLOC_SYSTEM:
|
|
|
|
if ((cnt.v_free_count >= cnt.v_free_reserved) ||
|
|
|
|
((cnt.v_cache_count == 0) &&
|
|
|
|
(cnt.v_free_count >= cnt.v_interrupt_free_min))) {
|
1996-09-08 20:44:49 +00:00
|
|
|
m = vm_page_select_free(object, pindex, PQ_FREE);
|
1996-09-28 03:33:40 +00:00
|
|
|
#if defined(DIAGNOSTIC)
|
1996-09-08 20:44:49 +00:00
|
|
|
if (m == NULL)
|
|
|
|
panic("vm_page_alloc(SYSTEM): missing page on free queue\n");
|
1996-09-28 03:33:40 +00:00
|
|
|
#endif
|
1995-03-01 23:30:04 +00:00
|
|
|
} else {
|
This mega-commit is meant to fix numerous interrelated problems. There
has been some bitrot and incorrect assumptions in the vfs_bio code. These
problems have manifest themselves worse on NFS type filesystems, but can
still affect local filesystems under certain circumstances. Most of
the problems have involved mmap consistancy, and as a side-effect broke
the vfs.ioopt code. This code might have been committed seperately, but
almost everything is interrelated.
1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that
are fully valid.
2) Rather than deactivating erroneously read initial (header) pages in
kern_exec, we now free them.
3) Fix the rundown of non-VMIO buffers that are in an inconsistent
(missing vp) state.
4) Fix the disassociation of pages from buffers in brelse. The previous
code had rotted and was faulty in a couple of important circumstances.
5) Remove a gratuitious buffer wakeup in vfs_vmio_release.
6) Remove a crufty and currently unused cluster mechanism for VBLK
files in vfs_bio_awrite. When the code is functional, I'll add back
a cleaner version.
7) The page busy count wakeups assocated with the buffer cache usage were
incorrectly cleaned up in a previous commit by me. Revert to the
original, correct version, but with a cleaner implementation.
8) The cluster read code now tries to keep data associated with buffers
more aggressively (without breaking the heuristics) when it is presumed
that the read data (buffers) will be soon needed.
9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The
delay loop waiting is not useful for filesystem locks, due to the
length of the time intervals.
10) Correct and clean-up spec_getpages.
11) Implement a fully functional nfs_getpages, nfs_putpages.
12) Fix nfs_write so that modifications are coherent with the NFS data on
the server disk (at least as well as NFS seems to allow.)
13) Properly support MS_INVALIDATE on NFS.
14) Properly pass down MS_INVALIDATE to lower levels of the VM code from
vm_map_clean.
15) Better support the notion of pages being busy but valid, so that
fewer in-transit waits occur. (use p->busy more for pageouts instead
of PG_BUSY.) Since the page is fully valid, it is still usable for
reads.
16) It is possible (in error) for cached pages to be busy. Make the
page allocation code handle that case correctly. (It should probably
be a printf or panic, but I want the system to handle coding errors
robustly. I'll probably add a printf.)
17) Correct the design and usage of vm_page_sleep. It didn't handle
consistancy problems very well, so make the design a little less
lofty. After vm_page_sleep, if it ever blocked, it is still important
to relookup the page (if the object generation count changed), and
verify it's status (always.)
18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up.
19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush.
20) Fix vm_pager_put_pages and it's descendents to support an int flag
instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
|
|
|
m = vm_page_select_cache(object, pindex);
|
1996-01-19 04:00:31 +00:00
|
|
|
if (m == NULL) {
|
1995-02-02 09:09:15 +00:00
|
|
|
splx(s);
|
1996-05-31 00:38:04 +00:00
|
|
|
#if defined(DIAGNOSTIC)
|
|
|
|
if (cnt.v_cache_count > 0)
|
|
|
|
printf("vm_page_alloc(SYSTEM): missing pages on cache queue: %d\n", cnt.v_cache_count);
|
|
|
|
#endif
|
VM level code cleanups.
1) Start using TSM.
Struct procs continue to point to upages structure, after being freed.
Struct vmspace continues to point to pte object and kva space for kstack.
u_map is now superfluous.
2) vm_map's don't need to be reference counted. They always exist either
in the kernel or in a vmspace. The vmspaces are managed by reference
counts.
3) Remove the "wired" vm_map nonsense.
4) No need to keep a cache of kernel stack kva's.
5) Get rid of strange looking ++var, and change to var++.
6) Change more data structures to use our "zone" allocator. Added
struct proc, struct vmspace and struct vnode. This saves a significant
amount of kva space and physical memory. Additionally, this enables
TSM for the zone managed memory.
7) Keep ioopt disabled for now.
8) Remove the now bogus "single use" map concept.
9) Use generation counts or id's for data structures residing in TSM, where
it allows us to avoid unneeded restart overhead during traversals, where
blocking might occur.
10) Account better for memory deficits, so the pageout daemon will be able
to make enough memory available (experimental.)
11) Fix some vnode locking problems. (From Tor, I think.)
12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp.
(experimental.)
13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c
code. Use generation counts, get rid of unneded collpase operations,
and clean up the cluster code.
14) Make vm_zone more suitable for TSM.
This commit is partially as a result of discussions and contributions from
other people, including DG, Tor Egge, PHK, and probably others that I
have forgotten to attribute (so let me know, if I forgot.)
This is not the infamous, final cleanup of the vnode stuff, but a necessary
step. Vnode mgmt should be correct, but things might still change, and
there is still some missing stuff (like ioopt, and physical backing of
non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
|
|
|
vm_pageout_deficit++;
|
1995-03-01 23:30:04 +00:00
|
|
|
pagedaemon_wakeup();
|
1995-02-02 09:09:15 +00:00
|
|
|
return (NULL);
|
1995-01-24 10:14:09 +00:00
|
|
|
}
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
}
|
1995-03-01 23:30:04 +00:00
|
|
|
break;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1995-03-01 23:30:04 +00:00
|
|
|
case VM_ALLOC_INTERRUPT:
|
1995-09-03 20:40:43 +00:00
|
|
|
if (cnt.v_free_count > 0) {
|
1996-09-08 20:44:49 +00:00
|
|
|
m = vm_page_select_free(object, pindex, PQ_FREE);
|
1996-10-06 18:27:39 +00:00
|
|
|
#if defined(DIAGNOSTIC)
|
|
|
|
if (m == NULL)
|
|
|
|
panic("vm_page_alloc(INTERRUPT): missing page on free queue\n");
|
|
|
|
#endif
|
1995-03-01 23:30:04 +00:00
|
|
|
} else {
|
|
|
|
splx(s);
|
VM level code cleanups.
1) Start using TSM.
Struct procs continue to point to upages structure, after being freed.
Struct vmspace continues to point to pte object and kva space for kstack.
u_map is now superfluous.
2) vm_map's don't need to be reference counted. They always exist either
in the kernel or in a vmspace. The vmspaces are managed by reference
counts.
3) Remove the "wired" vm_map nonsense.
4) No need to keep a cache of kernel stack kva's.
5) Get rid of strange looking ++var, and change to var++.
6) Change more data structures to use our "zone" allocator. Added
struct proc, struct vmspace and struct vnode. This saves a significant
amount of kva space and physical memory. Additionally, this enables
TSM for the zone managed memory.
7) Keep ioopt disabled for now.
8) Remove the now bogus "single use" map concept.
9) Use generation counts or id's for data structures residing in TSM, where
it allows us to avoid unneeded restart overhead during traversals, where
blocking might occur.
10) Account better for memory deficits, so the pageout daemon will be able
to make enough memory available (experimental.)
11) Fix some vnode locking problems. (From Tor, I think.)
12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp.
(experimental.)
13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c
code. Use generation counts, get rid of unneded collpase operations,
and clean up the cluster code.
14) Make vm_zone more suitable for TSM.
This commit is partially as a result of discussions and contributions from
other people, including DG, Tor Egge, PHK, and probably others that I
have forgotten to attribute (so let me know, if I forgot.)
This is not the infamous, final cleanup of the vnode stuff, but a necessary
step. Vnode mgmt should be correct, but things might still change, and
there is still some missing stuff (like ioopt, and physical backing of
non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
|
|
|
vm_pageout_deficit++;
|
1995-03-01 23:30:04 +00:00
|
|
|
pagedaemon_wakeup();
|
1996-01-19 04:00:31 +00:00
|
|
|
return (NULL);
|
1995-03-01 23:30:04 +00:00
|
|
|
}
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
1998-02-05 03:32:49 +00:00
|
|
|
m = NULL;
|
|
|
|
#if !defined(MAX_PERF)
|
1995-03-01 23:30:04 +00:00
|
|
|
panic("vm_page_alloc: invalid allocation class");
|
1998-02-05 03:32:49 +00:00
|
|
|
#endif
|
1995-03-01 23:30:04 +00:00
|
|
|
}
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
|
1996-01-19 04:00:31 +00:00
|
|
|
queue = m->queue;
|
1997-03-23 02:44:54 +00:00
|
|
|
qtype = queue - m->pc;
|
|
|
|
if (qtype == PQ_ZERO)
|
1998-01-31 11:56:53 +00:00
|
|
|
vm_page_zero_count--;
|
1996-09-28 03:33:40 +00:00
|
|
|
pq = &vm_page_queues[queue];
|
|
|
|
TAILQ_REMOVE(pq->pl, m, pageq);
|
1998-01-31 11:56:53 +00:00
|
|
|
(*pq->cnt)--;
|
|
|
|
(*pq->lcnt)--;
|
1998-01-12 01:46:33 +00:00
|
|
|
oldobject = NULL;
|
1997-03-23 02:44:54 +00:00
|
|
|
if (qtype == PQ_ZERO) {
|
1998-03-01 04:18:54 +00:00
|
|
|
m->flags = PG_ZERO | PG_BUSY;
|
1997-03-23 02:44:54 +00:00
|
|
|
} else if (qtype == PQ_CACHE) {
|
1998-01-12 01:46:33 +00:00
|
|
|
oldobject = m->object;
|
1998-09-04 08:06:57 +00:00
|
|
|
vm_page_busy(m);
|
1996-01-19 04:00:31 +00:00
|
|
|
vm_page_remove(m);
|
|
|
|
m->flags = PG_BUSY;
|
|
|
|
} else {
|
|
|
|
m->flags = PG_BUSY;
|
|
|
|
}
|
|
|
|
m->wire_count = 0;
|
1996-05-18 03:38:05 +00:00
|
|
|
m->hold_count = 0;
|
This commit does a couple of things:
Re-enables the RSS limiting, and the routine is now tail-recursive,
making it much more safe (eliminates the possiblity of kernel stack
overflow.) Also, the RSS limiting is a little more intelligent about
finding the likely objects that are pushing the process over the limit.
Added some sysctls that help with VM system tuning.
New sysctl features:
1) Enable/disable lru pageout algorithm.
vm.pageout_algorithm = 0, default algorithm that works
well, especially using X windows and heavy
memory loading. Can have adverse effects,
sometimes slowing down program loading.
vm.pageout_algorithm = 1, close to true LRU. Works much
better than clock, etc. Does not work as well as
the default algorithm in general. Certain memory
"malloc" type benchmarks work a little better with
this setting.
Please give me feedback on the performance results
associated with these.
2) Enable/disable swapping.
vm.swapping_enabled = 1, default.
vm.swapping_enabled = 0, useful for cases where swapping
degrades performance.
The config option "NO_SWAPPING" is still operative, and
takes precedence over the sysctl. If "NO_SWAPPING" is
specified, the sysctl still exists, but "vm.swapping_enabled"
is hard-wired to "0".
Each of these can be changed "on the fly."
1996-06-26 05:39:27 +00:00
|
|
|
m->act_count = 0;
|
1996-01-19 04:00:31 +00:00
|
|
|
m->busy = 0;
|
|
|
|
m->valid = 0;
|
|
|
|
m->dirty = 0;
|
|
|
|
m->queue = PQ_NONE;
|
1995-01-10 09:19:52 +00:00
|
|
|
|
|
|
|
/* XXX before splx until vm_page_insert is safe */
|
1996-01-19 04:00:31 +00:00
|
|
|
vm_page_insert(m, object, pindex);
|
1995-01-10 09:19:52 +00:00
|
|
|
|
1995-03-01 23:30:04 +00:00
|
|
|
/*
|
|
|
|
* Don't wakeup too often - wakeup the pageout daemon when
|
|
|
|
* we would be nearly out of memory.
|
|
|
|
*/
|
1996-01-19 04:00:31 +00:00
|
|
|
if (((cnt.v_free_count + cnt.v_cache_count) <
|
1996-03-09 06:56:39 +00:00
|
|
|
(cnt.v_free_reserved + cnt.v_cache_min)) ||
|
1996-01-19 04:00:31 +00:00
|
|
|
(cnt.v_free_count < cnt.v_pageout_free_min))
|
1995-03-01 23:30:04 +00:00
|
|
|
pagedaemon_wakeup();
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1998-01-17 09:17:02 +00:00
|
|
|
if ((qtype == PQ_CACHE) &&
|
|
|
|
((page_req == VM_ALLOC_NORMAL) || (page_req == VM_ALLOC_ZERO)) &&
|
|
|
|
oldobject && (oldobject->type == OBJT_VNODE) &&
|
|
|
|
((oldobject->flags & OBJ_DEAD) == 0)) {
|
1998-01-12 01:46:33 +00:00
|
|
|
struct vnode *vp;
|
|
|
|
vp = (struct vnode *) oldobject->handle;
|
1998-01-17 09:17:02 +00:00
|
|
|
if (vp && VSHOULDFREE(vp)) {
|
|
|
|
if ((vp->v_flag & (VFREE|VTBFREE|VDOOMED)) == 0) {
|
|
|
|
TAILQ_INSERT_TAIL(&vnode_tobefree_list, vp, v_freelist);
|
|
|
|
vp->v_flag |= VTBFREE;
|
|
|
|
}
|
1998-01-12 01:46:33 +00:00
|
|
|
}
|
|
|
|
}
|
1998-01-17 09:17:02 +00:00
|
|
|
splx(s);
|
1998-01-12 01:46:33 +00:00
|
|
|
|
1996-01-19 04:00:31 +00:00
|
|
|
return (m);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
1996-11-28 23:15:07 +00:00
|
|
|
void
|
|
|
|
vm_wait()
|
|
|
|
{
|
|
|
|
int s;
|
|
|
|
|
|
|
|
s = splvm();
|
|
|
|
if (curproc == pageproc) {
|
|
|
|
vm_pageout_pages_needed = 1;
|
|
|
|
tsleep(&vm_pageout_pages_needed, PSWP, "vmwait", 0);
|
|
|
|
} else {
|
|
|
|
if (!vm_pages_needed) {
|
|
|
|
vm_pages_needed++;
|
|
|
|
wakeup(&vm_pages_needed);
|
|
|
|
}
|
|
|
|
tsleep(&cnt.v_free_count, PVM, "vmwait", 0);
|
|
|
|
}
|
|
|
|
splx(s);
|
|
|
|
}
|
|
|
|
|
1998-03-01 04:18:54 +00:00
|
|
|
int
|
|
|
|
vm_page_sleep(vm_page_t m, char *msg, char *busy) {
|
This mega-commit is meant to fix numerous interrelated problems. There
has been some bitrot and incorrect assumptions in the vfs_bio code. These
problems have manifest themselves worse on NFS type filesystems, but can
still affect local filesystems under certain circumstances. Most of
the problems have involved mmap consistancy, and as a side-effect broke
the vfs.ioopt code. This code might have been committed seperately, but
almost everything is interrelated.
1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that
are fully valid.
2) Rather than deactivating erroneously read initial (header) pages in
kern_exec, we now free them.
3) Fix the rundown of non-VMIO buffers that are in an inconsistent
(missing vp) state.
4) Fix the disassociation of pages from buffers in brelse. The previous
code had rotted and was faulty in a couple of important circumstances.
5) Remove a gratuitious buffer wakeup in vfs_vmio_release.
6) Remove a crufty and currently unused cluster mechanism for VBLK
files in vfs_bio_awrite. When the code is functional, I'll add back
a cleaner version.
7) The page busy count wakeups assocated with the buffer cache usage were
incorrectly cleaned up in a previous commit by me. Revert to the
original, correct version, but with a cleaner implementation.
8) The cluster read code now tries to keep data associated with buffers
more aggressively (without breaking the heuristics) when it is presumed
that the read data (buffers) will be soon needed.
9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The
delay loop waiting is not useful for filesystem locks, due to the
length of the time intervals.
10) Correct and clean-up spec_getpages.
11) Implement a fully functional nfs_getpages, nfs_putpages.
12) Fix nfs_write so that modifications are coherent with the NFS data on
the server disk (at least as well as NFS seems to allow.)
13) Properly support MS_INVALIDATE on NFS.
14) Properly pass down MS_INVALIDATE to lower levels of the VM code from
vm_map_clean.
15) Better support the notion of pages being busy but valid, so that
fewer in-transit waits occur. (use p->busy more for pageouts instead
of PG_BUSY.) Since the page is fully valid, it is still usable for
reads.
16) It is possible (in error) for cached pages to be busy. Make the
page allocation code handle that case correctly. (It should probably
be a printf or panic, but I want the system to handle coding errors
robustly. I'll probably add a printf.)
17) Correct the design and usage of vm_page_sleep. It didn't handle
consistancy problems very well, so make the design a little less
lofty. After vm_page_sleep, if it ever blocked, it is still important
to relookup the page (if the object generation count changed), and
verify it's status (always.)
18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up.
19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush.
20) Fix vm_pager_put_pages and it's descendents to support an int flag
instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
|
|
|
int slept = 0;
|
1998-03-01 04:18:54 +00:00
|
|
|
if ((busy && *busy) || (m->flags & PG_BUSY)) {
|
|
|
|
int s;
|
|
|
|
s = splvm();
|
|
|
|
if ((busy && *busy) || (m->flags & PG_BUSY)) {
|
1998-09-04 08:06:57 +00:00
|
|
|
vm_page_flag_set(m, PG_WANTED);
|
This mega-commit is meant to fix numerous interrelated problems. There
has been some bitrot and incorrect assumptions in the vfs_bio code. These
problems have manifest themselves worse on NFS type filesystems, but can
still affect local filesystems under certain circumstances. Most of
the problems have involved mmap consistancy, and as a side-effect broke
the vfs.ioopt code. This code might have been committed seperately, but
almost everything is interrelated.
1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that
are fully valid.
2) Rather than deactivating erroneously read initial (header) pages in
kern_exec, we now free them.
3) Fix the rundown of non-VMIO buffers that are in an inconsistent
(missing vp) state.
4) Fix the disassociation of pages from buffers in brelse. The previous
code had rotted and was faulty in a couple of important circumstances.
5) Remove a gratuitious buffer wakeup in vfs_vmio_release.
6) Remove a crufty and currently unused cluster mechanism for VBLK
files in vfs_bio_awrite. When the code is functional, I'll add back
a cleaner version.
7) The page busy count wakeups assocated with the buffer cache usage were
incorrectly cleaned up in a previous commit by me. Revert to the
original, correct version, but with a cleaner implementation.
8) The cluster read code now tries to keep data associated with buffers
more aggressively (without breaking the heuristics) when it is presumed
that the read data (buffers) will be soon needed.
9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The
delay loop waiting is not useful for filesystem locks, due to the
length of the time intervals.
10) Correct and clean-up spec_getpages.
11) Implement a fully functional nfs_getpages, nfs_putpages.
12) Fix nfs_write so that modifications are coherent with the NFS data on
the server disk (at least as well as NFS seems to allow.)
13) Properly support MS_INVALIDATE on NFS.
14) Properly pass down MS_INVALIDATE to lower levels of the VM code from
vm_map_clean.
15) Better support the notion of pages being busy but valid, so that
fewer in-transit waits occur. (use p->busy more for pageouts instead
of PG_BUSY.) Since the page is fully valid, it is still usable for
reads.
16) It is possible (in error) for cached pages to be busy. Make the
page allocation code handle that case correctly. (It should probably
be a printf or panic, but I want the system to handle coding errors
robustly. I'll probably add a printf.)
17) Correct the design and usage of vm_page_sleep. It didn't handle
consistancy problems very well, so make the design a little less
lofty. After vm_page_sleep, if it ever blocked, it is still important
to relookup the page (if the object generation count changed), and
verify it's status (always.)
18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up.
19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush.
20) Fix vm_pager_put_pages and it's descendents to support an int flag
instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
|
|
|
tsleep(m, PVM, msg, 0);
|
|
|
|
slept = 1;
|
1998-03-01 04:18:54 +00:00
|
|
|
}
|
|
|
|
splx(s);
|
|
|
|
}
|
This mega-commit is meant to fix numerous interrelated problems. There
has been some bitrot and incorrect assumptions in the vfs_bio code. These
problems have manifest themselves worse on NFS type filesystems, but can
still affect local filesystems under certain circumstances. Most of
the problems have involved mmap consistancy, and as a side-effect broke
the vfs.ioopt code. This code might have been committed seperately, but
almost everything is interrelated.
1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that
are fully valid.
2) Rather than deactivating erroneously read initial (header) pages in
kern_exec, we now free them.
3) Fix the rundown of non-VMIO buffers that are in an inconsistent
(missing vp) state.
4) Fix the disassociation of pages from buffers in brelse. The previous
code had rotted and was faulty in a couple of important circumstances.
5) Remove a gratuitious buffer wakeup in vfs_vmio_release.
6) Remove a crufty and currently unused cluster mechanism for VBLK
files in vfs_bio_awrite. When the code is functional, I'll add back
a cleaner version.
7) The page busy count wakeups assocated with the buffer cache usage were
incorrectly cleaned up in a previous commit by me. Revert to the
original, correct version, but with a cleaner implementation.
8) The cluster read code now tries to keep data associated with buffers
more aggressively (without breaking the heuristics) when it is presumed
that the read data (buffers) will be soon needed.
9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The
delay loop waiting is not useful for filesystem locks, due to the
length of the time intervals.
10) Correct and clean-up spec_getpages.
11) Implement a fully functional nfs_getpages, nfs_putpages.
12) Fix nfs_write so that modifications are coherent with the NFS data on
the server disk (at least as well as NFS seems to allow.)
13) Properly support MS_INVALIDATE on NFS.
14) Properly pass down MS_INVALIDATE to lower levels of the VM code from
vm_map_clean.
15) Better support the notion of pages being busy but valid, so that
fewer in-transit waits occur. (use p->busy more for pageouts instead
of PG_BUSY.) Since the page is fully valid, it is still usable for
reads.
16) It is possible (in error) for cached pages to be busy. Make the
page allocation code handle that case correctly. (It should probably
be a printf or panic, but I want the system to handle coding errors
robustly. I'll probably add a printf.)
17) Correct the design and usage of vm_page_sleep. It didn't handle
consistancy problems very well, so make the design a little less
lofty. After vm_page_sleep, if it ever blocked, it is still important
to relookup the page (if the object generation count changed), and
verify it's status (always.)
18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up.
19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush.
20) Fix vm_pager_put_pages and it's descendents to support an int flag
instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
|
|
|
return slept;
|
1998-03-01 04:18:54 +00:00
|
|
|
}
|
1996-11-28 23:15:07 +00:00
|
|
|
|
1996-01-27 00:13:33 +00:00
|
|
|
/*
|
1996-05-18 03:38:05 +00:00
|
|
|
* vm_page_activate:
|
|
|
|
*
|
|
|
|
* Put the specified page on the active list (if appropriate).
|
|
|
|
*
|
|
|
|
* The page queues must be locked.
|
1996-01-27 00:13:33 +00:00
|
|
|
*/
|
1996-05-18 03:38:05 +00:00
|
|
|
void
|
|
|
|
vm_page_activate(m)
|
|
|
|
register vm_page_t m;
|
1994-09-27 18:00:29 +00:00
|
|
|
{
|
1996-05-18 03:38:05 +00:00
|
|
|
int s;
|
1994-09-27 18:00:29 +00:00
|
|
|
|
1996-05-18 03:38:05 +00:00
|
|
|
s = splvm();
|
1998-01-31 11:56:53 +00:00
|
|
|
if (m->queue != PQ_ACTIVE) {
|
|
|
|
if ((m->queue - m->pc) == PQ_CACHE)
|
|
|
|
cnt.v_reactivated++;
|
1994-09-27 18:00:29 +00:00
|
|
|
|
1998-01-31 11:56:53 +00:00
|
|
|
vm_page_unqueue(m);
|
1995-02-20 13:58:14 +00:00
|
|
|
|
1998-01-31 11:56:53 +00:00
|
|
|
if (m->wire_count == 0) {
|
|
|
|
m->queue = PQ_ACTIVE;
|
|
|
|
++(*vm_page_queues[PQ_ACTIVE].lcnt);
|
|
|
|
TAILQ_INSERT_TAIL(&vm_page_queue_active, m, pageq);
|
|
|
|
if (m->act_count < ACT_INIT)
|
|
|
|
m->act_count = ACT_INIT;
|
|
|
|
cnt.v_active_count++;
|
|
|
|
}
|
|
|
|
} else {
|
This commit does a couple of things:
Re-enables the RSS limiting, and the routine is now tail-recursive,
making it much more safe (eliminates the possiblity of kernel stack
overflow.) Also, the RSS limiting is a little more intelligent about
finding the likely objects that are pushing the process over the limit.
Added some sysctls that help with VM system tuning.
New sysctl features:
1) Enable/disable lru pageout algorithm.
vm.pageout_algorithm = 0, default algorithm that works
well, especially using X windows and heavy
memory loading. Can have adverse effects,
sometimes slowing down program loading.
vm.pageout_algorithm = 1, close to true LRU. Works much
better than clock, etc. Does not work as well as
the default algorithm in general. Certain memory
"malloc" type benchmarks work a little better with
this setting.
Please give me feedback on the performance results
associated with these.
2) Enable/disable swapping.
vm.swapping_enabled = 1, default.
vm.swapping_enabled = 0, useful for cases where swapping
degrades performance.
The config option "NO_SWAPPING" is still operative, and
takes precedence over the sysctl. If "NO_SWAPPING" is
specified, the sysctl still exists, but "vm.swapping_enabled"
is hard-wired to "0".
Each of these can be changed "on the fly."
1996-06-26 05:39:27 +00:00
|
|
|
if (m->act_count < ACT_INIT)
|
|
|
|
m->act_count = ACT_INIT;
|
1994-09-27 18:00:29 +00:00
|
|
|
}
|
1998-01-31 11:56:53 +00:00
|
|
|
|
1994-09-27 18:00:29 +00:00
|
|
|
splx(s);
|
|
|
|
}
|
|
|
|
|
1996-06-05 03:31:49 +00:00
|
|
|
/*
|
|
|
|
* helper routine for vm_page_free and vm_page_free_zero
|
|
|
|
*/
|
1996-07-30 03:08:57 +00:00
|
|
|
static int
|
1996-06-05 03:31:49 +00:00
|
|
|
vm_page_freechk_and_unqueue(m)
|
|
|
|
vm_page_t m;
|
|
|
|
{
|
1998-01-17 09:17:02 +00:00
|
|
|
vm_object_t oldobject;
|
|
|
|
|
|
|
|
oldobject = m->object;
|
|
|
|
|
1998-01-12 01:46:33 +00:00
|
|
|
#if !defined(MAX_PERF)
|
1998-01-31 11:56:53 +00:00
|
|
|
if (m->busy || ((m->queue - m->pc) == PQ_FREE) ||
|
1996-06-05 03:31:49 +00:00
|
|
|
(m->hold_count != 0)) {
|
1998-07-11 07:46:16 +00:00
|
|
|
printf(
|
|
|
|
"vm_page_free: pindex(%lu), busy(%d), PG_BUSY(%d), hold(%d)\n",
|
|
|
|
(u_long)m->pindex, m->busy, (m->flags & PG_BUSY) ? 1 : 0,
|
|
|
|
m->hold_count);
|
1996-09-08 20:44:49 +00:00
|
|
|
if ((m->queue - m->pc) == PQ_FREE)
|
1996-06-05 03:31:49 +00:00
|
|
|
panic("vm_page_free: freeing free page");
|
|
|
|
else
|
|
|
|
panic("vm_page_free: freeing busy page");
|
|
|
|
}
|
1998-01-12 01:46:33 +00:00
|
|
|
#endif
|
1996-06-05 03:31:49 +00:00
|
|
|
|
1996-07-30 03:08:57 +00:00
|
|
|
vm_page_unqueue_nowakeup(m);
|
1998-01-31 11:56:53 +00:00
|
|
|
vm_page_remove(m);
|
|
|
|
|
1996-06-05 03:31:49 +00:00
|
|
|
if ((m->flags & PG_FICTITIOUS) != 0) {
|
|
|
|
return 0;
|
|
|
|
}
|
1998-01-31 11:56:53 +00:00
|
|
|
|
1998-03-01 04:18:54 +00:00
|
|
|
m->valid = 0;
|
|
|
|
|
1996-06-05 03:31:49 +00:00
|
|
|
if (m->wire_count != 0) {
|
1998-02-05 03:32:49 +00:00
|
|
|
#if !defined(MAX_PERF)
|
1996-06-05 03:31:49 +00:00
|
|
|
if (m->wire_count > 1) {
|
|
|
|
panic("vm_page_free: invalid wire count (%d), pindex: 0x%x",
|
|
|
|
m->wire_count, m->pindex);
|
|
|
|
}
|
1998-02-05 03:32:49 +00:00
|
|
|
#endif
|
1998-10-21 11:43:04 +00:00
|
|
|
printf("vm_page_free: freeing wired page\n");
|
1996-06-05 03:31:49 +00:00
|
|
|
m->wire_count = 0;
|
1998-02-05 03:32:49 +00:00
|
|
|
if (m->object)
|
|
|
|
m->object->wire_count--;
|
1996-06-12 06:52:12 +00:00
|
|
|
cnt.v_wire_count--;
|
1996-06-05 03:31:49 +00:00
|
|
|
}
|
|
|
|
|
1998-01-17 09:17:02 +00:00
|
|
|
if (oldobject && (oldobject->type == OBJT_VNODE) &&
|
|
|
|
((oldobject->flags & OBJ_DEAD) == 0)) {
|
|
|
|
struct vnode *vp;
|
|
|
|
vp = (struct vnode *) oldobject->handle;
|
|
|
|
if (vp && VSHOULDFREE(vp)) {
|
|
|
|
if ((vp->v_flag & (VTBFREE|VDOOMED|VFREE)) == 0) {
|
|
|
|
TAILQ_INSERT_TAIL(&vnode_tobefree_list, vp, v_freelist);
|
|
|
|
vp->v_flag |= VTBFREE;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
1998-07-26 18:15:20 +00:00
|
|
|
#ifdef __alpha__
|
|
|
|
pmap_page_is_free(m);
|
|
|
|
#endif
|
|
|
|
|
1996-06-05 03:31:49 +00:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* helper routine for vm_page_free and vm_page_free_zero
|
|
|
|
*/
|
1996-07-30 03:08:57 +00:00
|
|
|
static __inline void
|
1996-06-05 03:31:49 +00:00
|
|
|
vm_page_free_wakeup()
|
|
|
|
{
|
|
|
|
|
|
|
|
/*
|
|
|
|
* if pageout daemon needs pages, then tell it that there are
|
|
|
|
* some free.
|
|
|
|
*/
|
|
|
|
if (vm_pageout_pages_needed) {
|
|
|
|
wakeup(&vm_pageout_pages_needed);
|
|
|
|
vm_pageout_pages_needed = 0;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* wakeup processes that are waiting on memory if we hit a
|
|
|
|
* high water mark. And wakeup scheduler process if we have
|
|
|
|
* lots of memory. this process will swapin processes.
|
|
|
|
*/
|
1996-06-17 03:35:40 +00:00
|
|
|
if (vm_pages_needed &&
|
|
|
|
((cnt.v_free_count + cnt.v_cache_count) >= cnt.v_free_min)) {
|
1996-06-05 03:31:49 +00:00
|
|
|
wakeup(&cnt.v_free_count);
|
1996-06-17 03:35:40 +00:00
|
|
|
vm_pages_needed = 0;
|
1996-06-05 03:31:49 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* vm_page_free:
|
|
|
|
*
|
|
|
|
* Returns the given page to the free list,
|
|
|
|
* disassociating it with any VM object.
|
|
|
|
*
|
|
|
|
* Object and page must be locked prior to entry.
|
|
|
|
*/
|
1995-05-30 08:16:23 +00:00
|
|
|
void
|
1996-01-19 04:00:31 +00:00
|
|
|
vm_page_free(m)
|
|
|
|
register vm_page_t m;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
1994-05-25 09:21:21 +00:00
|
|
|
int s;
|
1996-09-28 03:33:40 +00:00
|
|
|
struct vpgqueues *pq;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
|
1996-05-18 03:38:05 +00:00
|
|
|
s = splvm();
|
1996-01-19 04:00:31 +00:00
|
|
|
|
1996-06-05 03:31:49 +00:00
|
|
|
cnt.v_tfree++;
|
1996-05-18 03:38:05 +00:00
|
|
|
|
1996-06-05 03:31:49 +00:00
|
|
|
if (!vm_page_freechk_and_unqueue(m)) {
|
1994-05-25 09:21:21 +00:00
|
|
|
splx(s);
|
1996-06-05 03:31:49 +00:00
|
|
|
return;
|
|
|
|
}
|
1994-05-25 09:21:21 +00:00
|
|
|
|
1996-09-08 20:44:49 +00:00
|
|
|
m->queue = PQ_FREE + m->pc;
|
1996-09-28 03:33:40 +00:00
|
|
|
pq = &vm_page_queues[m->queue];
|
|
|
|
++(*pq->lcnt);
|
|
|
|
++(*pq->cnt);
|
1996-06-05 03:31:49 +00:00
|
|
|
/*
|
|
|
|
* If the pageout process is grabbing the page, it is likely
|
|
|
|
* that the page is NOT in the cache. It is more likely that
|
|
|
|
* the page will be partially in the cache if it is being
|
|
|
|
* explicitly freed.
|
|
|
|
*/
|
|
|
|
if (curproc == pageproc) {
|
1996-09-28 03:33:40 +00:00
|
|
|
TAILQ_INSERT_TAIL(pq->pl, m, pageq);
|
1994-05-25 09:21:21 +00:00
|
|
|
} else {
|
1996-09-28 03:33:40 +00:00
|
|
|
TAILQ_INSERT_HEAD(pq->pl, m, pageq);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
1998-01-17 09:17:02 +00:00
|
|
|
|
1996-06-05 03:31:49 +00:00
|
|
|
vm_page_free_wakeup();
|
1996-06-16 20:37:31 +00:00
|
|
|
splx(s);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
1996-06-05 03:31:49 +00:00
|
|
|
void
|
|
|
|
vm_page_free_zero(m)
|
|
|
|
register vm_page_t m;
|
|
|
|
{
|
|
|
|
int s;
|
1996-09-28 03:33:40 +00:00
|
|
|
struct vpgqueues *pq;
|
1996-06-05 03:31:49 +00:00
|
|
|
|
|
|
|
s = splvm();
|
|
|
|
|
|
|
|
cnt.v_tfree++;
|
|
|
|
|
|
|
|
if (!vm_page_freechk_and_unqueue(m)) {
|
|
|
|
splx(s);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
1996-09-08 20:44:49 +00:00
|
|
|
m->queue = PQ_ZERO + m->pc;
|
1996-09-28 03:33:40 +00:00
|
|
|
pq = &vm_page_queues[m->queue];
|
|
|
|
++(*pq->lcnt);
|
|
|
|
++(*pq->cnt);
|
1996-06-05 03:31:49 +00:00
|
|
|
|
1996-09-28 03:33:40 +00:00
|
|
|
TAILQ_INSERT_HEAD(pq->pl, m, pageq);
|
1996-06-05 03:31:49 +00:00
|
|
|
++vm_page_zero_count;
|
|
|
|
vm_page_free_wakeup();
|
1996-06-16 20:37:31 +00:00
|
|
|
splx(s);
|
1996-06-05 03:31:49 +00:00
|
|
|
}
|
1994-05-25 09:21:21 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* vm_page_wire:
|
|
|
|
*
|
|
|
|
* Mark this page as wired down by yet
|
|
|
|
* another map, removing it from paging queues
|
|
|
|
* as necessary.
|
|
|
|
*
|
|
|
|
* The page queues must be locked.
|
|
|
|
*/
|
1995-05-30 08:16:23 +00:00
|
|
|
void
|
1996-01-19 04:00:31 +00:00
|
|
|
vm_page_wire(m)
|
|
|
|
register vm_page_t m;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
1994-05-25 09:21:21 +00:00
|
|
|
int s;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1996-01-19 04:00:31 +00:00
|
|
|
if (m->wire_count == 0) {
|
1996-05-18 03:38:05 +00:00
|
|
|
s = splvm();
|
1996-07-30 03:08:57 +00:00
|
|
|
vm_page_unqueue(m);
|
1995-03-01 23:30:04 +00:00
|
|
|
splx(s);
|
1994-05-24 10:09:53 +00:00
|
|
|
cnt.v_wire_count++;
|
1998-01-31 20:30:18 +00:00
|
|
|
if (m->object)
|
|
|
|
m->object->wire_count++;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
1998-01-31 11:56:53 +00:00
|
|
|
(*vm_page_queues[PQ_NONE].lcnt)++;
|
1996-01-19 04:00:31 +00:00
|
|
|
m->wire_count++;
|
1998-09-04 08:06:57 +00:00
|
|
|
vm_page_flag_set(m, PG_MAPPED);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* vm_page_unwire:
|
|
|
|
*
|
|
|
|
* Release one wiring of this page, potentially
|
|
|
|
* enabling it to be paged again.
|
|
|
|
*
|
|
|
|
* The page queues must be locked.
|
|
|
|
*/
|
1995-05-30 08:16:23 +00:00
|
|
|
void
|
1996-01-19 04:00:31 +00:00
|
|
|
vm_page_unwire(m)
|
|
|
|
register vm_page_t m;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
1994-05-25 09:21:21 +00:00
|
|
|
int s;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
|
1996-05-18 03:38:05 +00:00
|
|
|
s = splvm();
|
1994-05-25 09:21:21 +00:00
|
|
|
|
1998-01-31 11:56:53 +00:00
|
|
|
if (m->wire_count > 0) {
|
1996-01-19 04:00:31 +00:00
|
|
|
m->wire_count--;
|
1998-01-31 11:56:53 +00:00
|
|
|
if (m->wire_count == 0) {
|
1998-01-31 20:30:18 +00:00
|
|
|
if (m->object)
|
|
|
|
m->object->wire_count--;
|
1998-01-31 11:56:53 +00:00
|
|
|
cnt.v_wire_count--;
|
|
|
|
TAILQ_INSERT_TAIL(&vm_page_queue_active, m, pageq);
|
|
|
|
m->queue = PQ_ACTIVE;
|
|
|
|
(*vm_page_queues[PQ_ACTIVE].lcnt)++;
|
|
|
|
cnt.v_active_count++;
|
|
|
|
}
|
|
|
|
} else {
|
1998-02-05 03:32:49 +00:00
|
|
|
#if !defined(MAX_PERF)
|
1998-01-31 11:56:53 +00:00
|
|
|
panic("vm_page_unwire: invalid wire count: %d\n", m->wire_count);
|
1998-02-05 03:32:49 +00:00
|
|
|
#endif
|
1995-03-25 08:47:35 +00:00
|
|
|
}
|
|
|
|
splx(s);
|
|
|
|
}
|
|
|
|
|
1996-05-18 03:38:05 +00:00
|
|
|
|
1994-05-25 09:21:21 +00:00
|
|
|
/*
|
|
|
|
* vm_page_deactivate:
|
|
|
|
*
|
|
|
|
* Returns the given page to the inactive list,
|
|
|
|
* indicating that no physical maps have access
|
|
|
|
* to this page. [Used by the physical mapping system.]
|
|
|
|
*
|
|
|
|
* The page queues must be locked.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
vm_page_deactivate(m)
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
register vm_page_t m;
|
1994-05-25 09:21:21 +00:00
|
|
|
{
|
1996-05-31 00:38:04 +00:00
|
|
|
int s;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
|
1994-05-25 09:21:21 +00:00
|
|
|
/*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* Only move active pages -- ignore locked or already inactive ones.
|
1995-05-30 08:16:23 +00:00
|
|
|
*
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
* XXX: sometimes we get pages which aren't wired down or on any queue -
|
|
|
|
* we need to put them on the inactive queue also, otherwise we lose
|
|
|
|
* track of them. Paul Mackerras (paulus@cs.anu.edu.au) 9-Jan-93.
|
1994-05-25 09:21:21 +00:00
|
|
|
*/
|
1996-01-19 04:00:31 +00:00
|
|
|
if (m->queue == PQ_INACTIVE)
|
|
|
|
return;
|
1994-05-25 09:21:21 +00:00
|
|
|
|
1996-05-31 00:38:04 +00:00
|
|
|
s = splvm();
|
1998-03-08 06:27:30 +00:00
|
|
|
if (m->wire_count == 0) {
|
1996-09-08 20:44:49 +00:00
|
|
|
if ((m->queue - m->pc) == PQ_CACHE)
|
1995-03-25 08:47:35 +00:00
|
|
|
cnt.v_reactivated++;
|
1996-07-30 03:08:57 +00:00
|
|
|
vm_page_unqueue(m);
|
1994-05-25 09:21:21 +00:00
|
|
|
TAILQ_INSERT_TAIL(&vm_page_queue_inactive, m, pageq);
|
1996-01-19 04:00:31 +00:00
|
|
|
m->queue = PQ_INACTIVE;
|
1996-09-08 20:44:49 +00:00
|
|
|
++(*vm_page_queues[PQ_INACTIVE].lcnt);
|
1994-05-25 09:21:21 +00:00
|
|
|
cnt.v_inactive_count++;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
}
|
1996-05-31 00:38:04 +00:00
|
|
|
splx(s);
|
1994-05-25 09:21:21 +00:00
|
|
|
}
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* vm_page_cache
|
|
|
|
*
|
|
|
|
* Put the specified page onto the page cache queue (if appropriate).
|
|
|
|
*/
|
1995-05-30 08:16:23 +00:00
|
|
|
void
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
vm_page_cache(m)
|
|
|
|
register vm_page_t m;
|
|
|
|
{
|
|
|
|
int s;
|
|
|
|
|
1998-02-05 03:32:49 +00:00
|
|
|
#if !defined(MAX_PERF)
|
1996-03-28 04:53:28 +00:00
|
|
|
if ((m->flags & PG_BUSY) || m->busy || m->wire_count) {
|
|
|
|
printf("vm_page_cache: attempting to cache busy page\n");
|
1996-01-19 04:00:31 +00:00
|
|
|
return;
|
1996-03-28 04:53:28 +00:00
|
|
|
}
|
1998-02-05 03:32:49 +00:00
|
|
|
#endif
|
1996-09-08 20:44:49 +00:00
|
|
|
if ((m->queue - m->pc) == PQ_CACHE)
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
return;
|
|
|
|
|
1996-01-19 04:00:31 +00:00
|
|
|
vm_page_protect(m, VM_PROT_NONE);
|
1998-02-05 03:32:49 +00:00
|
|
|
#if !defined(MAX_PERF)
|
1996-05-24 05:20:15 +00:00
|
|
|
if (m->dirty != 0) {
|
|
|
|
panic("vm_page_cache: caching a dirty page, pindex: %d", m->pindex);
|
|
|
|
}
|
1998-02-05 03:32:49 +00:00
|
|
|
#endif
|
1996-05-18 03:38:05 +00:00
|
|
|
s = splvm();
|
1996-07-30 03:08:57 +00:00
|
|
|
vm_page_unqueue_nowakeup(m);
|
1996-09-08 20:44:49 +00:00
|
|
|
m->queue = PQ_CACHE + m->pc;
|
1998-01-31 11:56:53 +00:00
|
|
|
(*vm_page_queues[m->queue].lcnt)++;
|
1996-09-08 20:44:49 +00:00
|
|
|
TAILQ_INSERT_TAIL(vm_page_queues[m->queue].pl, m, pageq);
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
cnt.v_cache_count++;
|
1998-01-31 11:56:53 +00:00
|
|
|
m->object->cache_count++;
|
1996-06-16 20:37:31 +00:00
|
|
|
vm_page_free_wakeup();
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
splx(s);
|
|
|
|
}
|
|
|
|
|
1998-02-05 03:32:49 +00:00
|
|
|
/*
|
|
|
|
* Grab a page, waiting until we are waken up due to the page
|
|
|
|
* changing state. We keep on waiting, if the page continues
|
|
|
|
* to be in the object. If the page doesn't exist, allocate it.
|
|
|
|
*/
|
|
|
|
vm_page_t
|
|
|
|
vm_page_grab(object, pindex, allocflags)
|
|
|
|
vm_object_t object;
|
|
|
|
vm_pindex_t pindex;
|
|
|
|
int allocflags;
|
|
|
|
{
|
|
|
|
|
|
|
|
vm_page_t m;
|
|
|
|
int s, generation;
|
|
|
|
|
|
|
|
retrylookup:
|
|
|
|
if ((m = vm_page_lookup(object, pindex)) != NULL) {
|
|
|
|
if (m->busy || (m->flags & PG_BUSY)) {
|
|
|
|
generation = object->generation;
|
|
|
|
|
|
|
|
s = splvm();
|
|
|
|
while ((object->generation == generation) &&
|
|
|
|
(m->busy || (m->flags & PG_BUSY))) {
|
1998-09-04 08:06:57 +00:00
|
|
|
vm_page_flag_set(m, PG_WANTED | PG_REFERENCED);
|
1998-02-05 03:32:49 +00:00
|
|
|
tsleep(m, PVM, "pgrbwt", 0);
|
|
|
|
if ((allocflags & VM_ALLOC_RETRY) == 0) {
|
|
|
|
splx(s);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
splx(s);
|
|
|
|
goto retrylookup;
|
|
|
|
} else {
|
1998-09-04 08:06:57 +00:00
|
|
|
vm_page_busy(m);
|
1998-02-05 03:32:49 +00:00
|
|
|
return m;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
m = vm_page_alloc(object, pindex, allocflags & ~VM_ALLOC_RETRY);
|
|
|
|
if (m == NULL) {
|
|
|
|
VM_WAIT;
|
|
|
|
if ((allocflags & VM_ALLOC_RETRY) == 0)
|
|
|
|
return NULL;
|
|
|
|
goto retrylookup;
|
|
|
|
}
|
|
|
|
|
|
|
|
return m;
|
|
|
|
}
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* mapping function for valid bits or for dirty bits in
|
|
|
|
* a page
|
|
|
|
*/
|
1998-04-15 17:47:40 +00:00
|
|
|
__inline int
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
vm_page_bits(int base, int size)
|
|
|
|
{
|
|
|
|
u_short chunk;
|
|
|
|
|
1995-02-22 10:16:21 +00:00
|
|
|
if ((base == 0) && (size >= PAGE_SIZE))
|
1995-02-20 14:00:50 +00:00
|
|
|
return VM_PAGE_BITS_ALL;
|
Some VM improvements, including elimination of alot of Sig-11
problems. Tor Egge and others have helped with various VM bugs
lately, but don't blame him -- blame me!!!
pmap.c:
1) Create an object for kernel page table allocations. This
fixes a bogus allocation method previously used for such, by
grabbing pages from the kernel object, using bogus pindexes.
(This was a code cleanup, and perhaps a minor system stability
issue.)
pmap.c:
2) Pre-set the modify and accessed bits when prudent. This will
decrease bus traffic under certain circumstances.
vfs_bio.c, vfs_cluster.c:
3) Rather than calculating the beginning virtual byte offset
multiple times, stick the offset into the buffer header, so
that the calculated offset can be reused. (Long long multiplies
are often expensive, and this is a probably unmeasurable performance
improvement, and code cleanup.)
vfs_bio.c:
4) Handle write recursion more intelligently (but not perfectly) so
that it is less likely to cause a system panic, and is also
much more robust.
vfs_bio.c:
5) getblk incorrectly wrote out blocks that are incorrectly sized.
The problem is fixed, and writes blocks out ONLY when B_DELWRI
is true.
vfs_bio.c:
6) Check that already constituted buffers have fully valid pages. If
not, then make sure that the B_CACHE bit is not set. (This was
a major source of Sig-11 type problems.)
vfs_bio.c:
7) Fix a potential system deadlock due to an incorrectly specified
sleep priority while waiting for a buffer write operation. The
change that I made opens the system up to serious problems, and
we need to examine the issue of process sleep priorities.
vfs_cluster.c, vfs_bio.c:
8) Make clustered reads work more correctly (and more completely)
when buffers are already constituted, but not fully valid.
(This was another system reliability issue.)
vfs_subr.c, ffs_inode.c:
9) Create a vtruncbuf function, which is used by filesystems that
can truncate files. The vinvalbuf forced a file sync type operation,
while vtruncbuf only invalidates the buffers past the new end of file,
and also invalidates the appropriate pages. (This was a system reliabiliy
and performance issue.)
10) Modify FFS to use vtruncbuf.
vm_object.c:
11) Make the object rundown mechanism for OBJT_VNODE type objects work
more correctly. Included in that fix, create pager entries for
the OBJT_DEAD pager type, so that paging requests that might slip
in during race conditions are properly handled. (This was a system
reliability issue.)
vm_page.c:
12) Make some of the page validation routines be a little less picky
about arguments passed to them. Also, support page invalidation
change the object generation count so that we handle generation
counts a little more robustly.
vm_pageout.c:
13) Further reduce pageout daemon activity when the system doesn't
need help from it. There should be no additional performance
decrease even when the pageout daemon is running. (This was
a significant performance issue.)
vnode_pager.c:
14) Teach the vnode pager to handle race conditions during vnode
deallocations.
1998-03-16 01:56:03 +00:00
|
|
|
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
size = (size + DEV_BSIZE - 1) & ~(DEV_BSIZE - 1);
|
1998-05-02 03:02:13 +00:00
|
|
|
base &= PAGE_MASK;
|
Some VM improvements, including elimination of alot of Sig-11
problems. Tor Egge and others have helped with various VM bugs
lately, but don't blame him -- blame me!!!
pmap.c:
1) Create an object for kernel page table allocations. This
fixes a bogus allocation method previously used for such, by
grabbing pages from the kernel object, using bogus pindexes.
(This was a code cleanup, and perhaps a minor system stability
issue.)
pmap.c:
2) Pre-set the modify and accessed bits when prudent. This will
decrease bus traffic under certain circumstances.
vfs_bio.c, vfs_cluster.c:
3) Rather than calculating the beginning virtual byte offset
multiple times, stick the offset into the buffer header, so
that the calculated offset can be reused. (Long long multiplies
are often expensive, and this is a probably unmeasurable performance
improvement, and code cleanup.)
vfs_bio.c:
4) Handle write recursion more intelligently (but not perfectly) so
that it is less likely to cause a system panic, and is also
much more robust.
vfs_bio.c:
5) getblk incorrectly wrote out blocks that are incorrectly sized.
The problem is fixed, and writes blocks out ONLY when B_DELWRI
is true.
vfs_bio.c:
6) Check that already constituted buffers have fully valid pages. If
not, then make sure that the B_CACHE bit is not set. (This was
a major source of Sig-11 type problems.)
vfs_bio.c:
7) Fix a potential system deadlock due to an incorrectly specified
sleep priority while waiting for a buffer write operation. The
change that I made opens the system up to serious problems, and
we need to examine the issue of process sleep priorities.
vfs_cluster.c, vfs_bio.c:
8) Make clustered reads work more correctly (and more completely)
when buffers are already constituted, but not fully valid.
(This was another system reliability issue.)
vfs_subr.c, ffs_inode.c:
9) Create a vtruncbuf function, which is used by filesystems that
can truncate files. The vinvalbuf forced a file sync type operation,
while vtruncbuf only invalidates the buffers past the new end of file,
and also invalidates the appropriate pages. (This was a system reliabiliy
and performance issue.)
10) Modify FFS to use vtruncbuf.
vm_object.c:
11) Make the object rundown mechanism for OBJT_VNODE type objects work
more correctly. Included in that fix, create pager entries for
the OBJT_DEAD pager type, so that paging requests that might slip
in during race conditions are properly handled. (This was a system
reliability issue.)
vm_page.c:
12) Make some of the page validation routines be a little less picky
about arguments passed to them. Also, support page invalidation
change the object generation count so that we handle generation
counts a little more robustly.
vm_pageout.c:
13) Further reduce pageout daemon activity when the system doesn't
need help from it. There should be no additional performance
decrease even when the pageout daemon is running. (This was
a significant performance issue.)
vnode_pager.c:
14) Teach the vnode pager to handle race conditions during vnode
deallocations.
1998-03-16 01:56:03 +00:00
|
|
|
if (size > PAGE_SIZE - base) {
|
|
|
|
size = PAGE_SIZE - base;
|
|
|
|
}
|
|
|
|
|
1998-05-02 03:02:13 +00:00
|
|
|
base = base / DEV_BSIZE;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
chunk = vm_page_dev_bsize_chunks[size / DEV_BSIZE];
|
|
|
|
return (chunk << base) & VM_PAGE_BITS_ALL;
|
|
|
|
}
|
|
|
|
|
1995-09-03 19:57:25 +00:00
|
|
|
/*
|
|
|
|
* set a page valid and clean
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
vm_page_set_validclean(m, base, size)
|
|
|
|
vm_page_t m;
|
|
|
|
int base;
|
|
|
|
int size;
|
|
|
|
{
|
|
|
|
int pagebits = vm_page_bits(base, size);
|
|
|
|
m->valid |= pagebits;
|
|
|
|
m->dirty &= ~pagebits;
|
|
|
|
if( base == 0 && size == PAGE_SIZE)
|
1996-07-30 03:08:57 +00:00
|
|
|
pmap_clear_modify(VM_PAGE_TO_PHYS(m));
|
1995-09-03 19:57:25 +00:00
|
|
|
}
|
|
|
|
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
/*
|
|
|
|
* set a page (partially) invalid
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
vm_page_set_invalid(m, base, size)
|
|
|
|
vm_page_t m;
|
|
|
|
int base;
|
|
|
|
int size;
|
|
|
|
{
|
|
|
|
int bits;
|
|
|
|
|
|
|
|
m->valid &= ~(bits = vm_page_bits(base, size));
|
|
|
|
if (m->valid == 0)
|
|
|
|
m->dirty &= ~bits;
|
Some VM improvements, including elimination of alot of Sig-11
problems. Tor Egge and others have helped with various VM bugs
lately, but don't blame him -- blame me!!!
pmap.c:
1) Create an object for kernel page table allocations. This
fixes a bogus allocation method previously used for such, by
grabbing pages from the kernel object, using bogus pindexes.
(This was a code cleanup, and perhaps a minor system stability
issue.)
pmap.c:
2) Pre-set the modify and accessed bits when prudent. This will
decrease bus traffic under certain circumstances.
vfs_bio.c, vfs_cluster.c:
3) Rather than calculating the beginning virtual byte offset
multiple times, stick the offset into the buffer header, so
that the calculated offset can be reused. (Long long multiplies
are often expensive, and this is a probably unmeasurable performance
improvement, and code cleanup.)
vfs_bio.c:
4) Handle write recursion more intelligently (but not perfectly) so
that it is less likely to cause a system panic, and is also
much more robust.
vfs_bio.c:
5) getblk incorrectly wrote out blocks that are incorrectly sized.
The problem is fixed, and writes blocks out ONLY when B_DELWRI
is true.
vfs_bio.c:
6) Check that already constituted buffers have fully valid pages. If
not, then make sure that the B_CACHE bit is not set. (This was
a major source of Sig-11 type problems.)
vfs_bio.c:
7) Fix a potential system deadlock due to an incorrectly specified
sleep priority while waiting for a buffer write operation. The
change that I made opens the system up to serious problems, and
we need to examine the issue of process sleep priorities.
vfs_cluster.c, vfs_bio.c:
8) Make clustered reads work more correctly (and more completely)
when buffers are already constituted, but not fully valid.
(This was another system reliability issue.)
vfs_subr.c, ffs_inode.c:
9) Create a vtruncbuf function, which is used by filesystems that
can truncate files. The vinvalbuf forced a file sync type operation,
while vtruncbuf only invalidates the buffers past the new end of file,
and also invalidates the appropriate pages. (This was a system reliabiliy
and performance issue.)
10) Modify FFS to use vtruncbuf.
vm_object.c:
11) Make the object rundown mechanism for OBJT_VNODE type objects work
more correctly. Included in that fix, create pager entries for
the OBJT_DEAD pager type, so that paging requests that might slip
in during race conditions are properly handled. (This was a system
reliability issue.)
vm_page.c:
12) Make some of the page validation routines be a little less picky
about arguments passed to them. Also, support page invalidation
change the object generation count so that we handle generation
counts a little more robustly.
vm_pageout.c:
13) Further reduce pageout daemon activity when the system doesn't
need help from it. There should be no additional performance
decrease even when the pageout daemon is running. (This was
a significant performance issue.)
vnode_pager.c:
14) Teach the vnode pager to handle race conditions during vnode
deallocations.
1998-03-16 01:56:03 +00:00
|
|
|
m->object->generation++;
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* is (partial) page valid?
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
vm_page_is_valid(m, base, size)
|
|
|
|
vm_page_t m;
|
|
|
|
int base;
|
|
|
|
int size;
|
|
|
|
{
|
1995-04-09 06:03:56 +00:00
|
|
|
int bits = vm_page_bits(base, size);
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
|
1995-04-09 06:03:56 +00:00
|
|
|
if (m->valid && ((m->valid & bits) == bits))
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
return 1;
|
|
|
|
else
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
vm_page_test_dirty(m)
|
|
|
|
vm_page_t m;
|
|
|
|
{
|
1996-07-30 03:08:57 +00:00
|
|
|
if ((m->dirty != VM_PAGE_BITS_ALL) &&
|
|
|
|
pmap_is_modified(VM_PAGE_TO_PHYS(m))) {
|
|
|
|
m->dirty = VM_PAGE_BITS_ALL;
|
|
|
|
}
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
}
|
|
|
|
|
1996-05-18 03:38:05 +00:00
|
|
|
/*
|
|
|
|
* This interface is for merging with malloc() someday.
|
|
|
|
* Even if we never implement compaction so that contiguous allocation
|
|
|
|
* works after initialization time, malloc()'s data structures are good
|
|
|
|
* for statistics and for allocations of less than a page.
|
|
|
|
*/
|
|
|
|
void *
|
1997-02-13 19:37:40 +00:00
|
|
|
contigmalloc1(size, type, flags, low, high, alignment, boundary, map)
|
1996-05-18 03:38:05 +00:00
|
|
|
unsigned long size; /* should be size_t here and for malloc() */
|
1997-10-10 18:18:47 +00:00
|
|
|
struct malloc_type *type;
|
1996-05-18 03:38:05 +00:00
|
|
|
int flags;
|
|
|
|
unsigned long low;
|
|
|
|
unsigned long high;
|
|
|
|
unsigned long alignment;
|
|
|
|
unsigned long boundary;
|
1997-02-13 19:37:40 +00:00
|
|
|
vm_map_t map;
|
1996-05-18 03:38:05 +00:00
|
|
|
{
|
|
|
|
int i, s, start;
|
|
|
|
vm_offset_t addr, phys, tmp_addr;
|
1996-11-05 04:19:08 +00:00
|
|
|
int pass;
|
1996-05-18 03:38:05 +00:00
|
|
|
vm_page_t pga = vm_page_array;
|
|
|
|
|
|
|
|
size = round_page(size);
|
1998-02-05 03:32:49 +00:00
|
|
|
#if !defined(MAX_PERF)
|
1996-05-18 03:38:05 +00:00
|
|
|
if (size == 0)
|
1997-09-13 15:04:52 +00:00
|
|
|
panic("contigmalloc1: size must not be 0");
|
1996-05-18 03:38:05 +00:00
|
|
|
if ((alignment & (alignment - 1)) != 0)
|
1997-09-13 15:04:52 +00:00
|
|
|
panic("contigmalloc1: alignment must be a power of 2");
|
1996-05-18 03:38:05 +00:00
|
|
|
if ((boundary & (boundary - 1)) != 0)
|
1997-09-13 15:04:52 +00:00
|
|
|
panic("contigmalloc1: boundary must be a power of 2");
|
1998-02-05 03:32:49 +00:00
|
|
|
#endif
|
1996-05-18 03:38:05 +00:00
|
|
|
|
|
|
|
start = 0;
|
1996-11-05 04:19:08 +00:00
|
|
|
for (pass = 0; pass <= 1; pass++) {
|
|
|
|
s = splvm();
|
1996-05-18 03:38:05 +00:00
|
|
|
again:
|
1996-11-05 04:19:08 +00:00
|
|
|
/*
|
|
|
|
* Find first page in array that is free, within range, aligned, and
|
|
|
|
* such that the boundary won't be crossed.
|
|
|
|
*/
|
|
|
|
for (i = start; i < cnt.v_page_count; i++) {
|
|
|
|
int pqtype;
|
|
|
|
phys = VM_PAGE_TO_PHYS(&pga[i]);
|
|
|
|
pqtype = pga[i].queue - pga[i].pc;
|
|
|
|
if (((pqtype == PQ_ZERO) || (pqtype == PQ_FREE) || (pqtype == PQ_CACHE)) &&
|
|
|
|
(phys >= low) && (phys < high) &&
|
|
|
|
((phys & (alignment - 1)) == 0) &&
|
|
|
|
(((phys ^ (phys + size - 1)) & ~(boundary - 1)) == 0))
|
|
|
|
break;
|
|
|
|
}
|
1996-05-18 03:38:05 +00:00
|
|
|
|
1996-11-05 04:19:08 +00:00
|
|
|
/*
|
|
|
|
* If the above failed or we will exceed the upper bound, fail.
|
|
|
|
*/
|
|
|
|
if ((i == cnt.v_page_count) ||
|
|
|
|
((VM_PAGE_TO_PHYS(&pga[i]) + size) > high)) {
|
|
|
|
vm_page_t m, next;
|
1996-05-18 03:38:05 +00:00
|
|
|
|
1996-11-05 04:19:08 +00:00
|
|
|
again1:
|
|
|
|
for (m = TAILQ_FIRST(&vm_page_queue_inactive);
|
|
|
|
m != NULL;
|
|
|
|
m = next) {
|
|
|
|
|
|
|
|
if (m->queue != PQ_INACTIVE) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
next = TAILQ_NEXT(m, pageq);
|
1998-03-01 04:18:54 +00:00
|
|
|
if (vm_page_sleep(m, "vpctw0", &m->busy))
|
1996-11-05 04:19:08 +00:00
|
|
|
goto again1;
|
|
|
|
vm_page_test_dirty(m);
|
|
|
|
if (m->dirty) {
|
|
|
|
if (m->object->type == OBJT_VNODE) {
|
1997-12-29 00:25:11 +00:00
|
|
|
vn_lock(m->object->handle, LK_EXCLUSIVE | LK_RETRY, curproc);
|
This mega-commit is meant to fix numerous interrelated problems. There
has been some bitrot and incorrect assumptions in the vfs_bio code. These
problems have manifest themselves worse on NFS type filesystems, but can
still affect local filesystems under certain circumstances. Most of
the problems have involved mmap consistancy, and as a side-effect broke
the vfs.ioopt code. This code might have been committed seperately, but
almost everything is interrelated.
1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that
are fully valid.
2) Rather than deactivating erroneously read initial (header) pages in
kern_exec, we now free them.
3) Fix the rundown of non-VMIO buffers that are in an inconsistent
(missing vp) state.
4) Fix the disassociation of pages from buffers in brelse. The previous
code had rotted and was faulty in a couple of important circumstances.
5) Remove a gratuitious buffer wakeup in vfs_vmio_release.
6) Remove a crufty and currently unused cluster mechanism for VBLK
files in vfs_bio_awrite. When the code is functional, I'll add back
a cleaner version.
7) The page busy count wakeups assocated with the buffer cache usage were
incorrectly cleaned up in a previous commit by me. Revert to the
original, correct version, but with a cleaner implementation.
8) The cluster read code now tries to keep data associated with buffers
more aggressively (without breaking the heuristics) when it is presumed
that the read data (buffers) will be soon needed.
9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The
delay loop waiting is not useful for filesystem locks, due to the
length of the time intervals.
10) Correct and clean-up spec_getpages.
11) Implement a fully functional nfs_getpages, nfs_putpages.
12) Fix nfs_write so that modifications are coherent with the NFS data on
the server disk (at least as well as NFS seems to allow.)
13) Properly support MS_INVALIDATE on NFS.
14) Properly pass down MS_INVALIDATE to lower levels of the VM code from
vm_map_clean.
15) Better support the notion of pages being busy but valid, so that
fewer in-transit waits occur. (use p->busy more for pageouts instead
of PG_BUSY.) Since the page is fully valid, it is still usable for
reads.
16) It is possible (in error) for cached pages to be busy. Make the
page allocation code handle that case correctly. (It should probably
be a printf or panic, but I want the system to handle coding errors
robustly. I'll probably add a printf.)
17) Correct the design and usage of vm_page_sleep. It didn't handle
consistancy problems very well, so make the design a little less
lofty. After vm_page_sleep, if it ever blocked, it is still important
to relookup the page (if the object generation count changed), and
verify it's status (always.)
18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up.
19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush.
20) Fix vm_pager_put_pages and it's descendents to support an int flag
instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
|
|
|
vm_object_page_clean(m->object, 0, 0, OBJPC_SYNC);
|
1997-12-29 00:25:11 +00:00
|
|
|
VOP_UNLOCK(m->object->handle, 0, curproc);
|
1996-11-05 04:19:08 +00:00
|
|
|
goto again1;
|
|
|
|
} else if (m->object->type == OBJT_SWAP ||
|
|
|
|
m->object->type == OBJT_DEFAULT) {
|
|
|
|
vm_pageout_flush(&m, 1, 0);
|
|
|
|
goto again1;
|
|
|
|
}
|
|
|
|
}
|
1998-01-31 11:56:53 +00:00
|
|
|
if ((m->dirty == 0) && (m->busy == 0) && (m->hold_count == 0))
|
1996-11-05 04:19:08 +00:00
|
|
|
vm_page_cache(m);
|
|
|
|
}
|
|
|
|
|
|
|
|
for (m = TAILQ_FIRST(&vm_page_queue_active);
|
|
|
|
m != NULL;
|
|
|
|
m = next) {
|
|
|
|
|
|
|
|
if (m->queue != PQ_ACTIVE) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
next = TAILQ_NEXT(m, pageq);
|
1998-03-01 04:18:54 +00:00
|
|
|
if (vm_page_sleep(m, "vpctw1", &m->busy))
|
1996-11-05 04:19:08 +00:00
|
|
|
goto again1;
|
|
|
|
vm_page_test_dirty(m);
|
|
|
|
if (m->dirty) {
|
|
|
|
if (m->object->type == OBJT_VNODE) {
|
1997-12-29 00:25:11 +00:00
|
|
|
vn_lock(m->object->handle, LK_EXCLUSIVE | LK_RETRY, curproc);
|
This mega-commit is meant to fix numerous interrelated problems. There
has been some bitrot and incorrect assumptions in the vfs_bio code. These
problems have manifest themselves worse on NFS type filesystems, but can
still affect local filesystems under certain circumstances. Most of
the problems have involved mmap consistancy, and as a side-effect broke
the vfs.ioopt code. This code might have been committed seperately, but
almost everything is interrelated.
1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that
are fully valid.
2) Rather than deactivating erroneously read initial (header) pages in
kern_exec, we now free them.
3) Fix the rundown of non-VMIO buffers that are in an inconsistent
(missing vp) state.
4) Fix the disassociation of pages from buffers in brelse. The previous
code had rotted and was faulty in a couple of important circumstances.
5) Remove a gratuitious buffer wakeup in vfs_vmio_release.
6) Remove a crufty and currently unused cluster mechanism for VBLK
files in vfs_bio_awrite. When the code is functional, I'll add back
a cleaner version.
7) The page busy count wakeups assocated with the buffer cache usage were
incorrectly cleaned up in a previous commit by me. Revert to the
original, correct version, but with a cleaner implementation.
8) The cluster read code now tries to keep data associated with buffers
more aggressively (without breaking the heuristics) when it is presumed
that the read data (buffers) will be soon needed.
9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The
delay loop waiting is not useful for filesystem locks, due to the
length of the time intervals.
10) Correct and clean-up spec_getpages.
11) Implement a fully functional nfs_getpages, nfs_putpages.
12) Fix nfs_write so that modifications are coherent with the NFS data on
the server disk (at least as well as NFS seems to allow.)
13) Properly support MS_INVALIDATE on NFS.
14) Properly pass down MS_INVALIDATE to lower levels of the VM code from
vm_map_clean.
15) Better support the notion of pages being busy but valid, so that
fewer in-transit waits occur. (use p->busy more for pageouts instead
of PG_BUSY.) Since the page is fully valid, it is still usable for
reads.
16) It is possible (in error) for cached pages to be busy. Make the
page allocation code handle that case correctly. (It should probably
be a printf or panic, but I want the system to handle coding errors
robustly. I'll probably add a printf.)
17) Correct the design and usage of vm_page_sleep. It didn't handle
consistancy problems very well, so make the design a little less
lofty. After vm_page_sleep, if it ever blocked, it is still important
to relookup the page (if the object generation count changed), and
verify it's status (always.)
18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up.
19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush.
20) Fix vm_pager_put_pages and it's descendents to support an int flag
instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
|
|
|
vm_object_page_clean(m->object, 0, 0, OBJPC_SYNC);
|
1997-12-29 00:25:11 +00:00
|
|
|
VOP_UNLOCK(m->object->handle, 0, curproc);
|
1996-11-05 04:19:08 +00:00
|
|
|
goto again1;
|
|
|
|
} else if (m->object->type == OBJT_SWAP ||
|
|
|
|
m->object->type == OBJT_DEFAULT) {
|
|
|
|
vm_pageout_flush(&m, 1, 0);
|
|
|
|
goto again1;
|
|
|
|
}
|
|
|
|
}
|
1998-01-31 11:56:53 +00:00
|
|
|
if ((m->dirty == 0) && (m->busy == 0) && (m->hold_count == 0))
|
1996-11-05 04:19:08 +00:00
|
|
|
vm_page_cache(m);
|
|
|
|
}
|
|
|
|
|
|
|
|
splx(s);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
start = i;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check successive pages for contiguous and free.
|
|
|
|
*/
|
|
|
|
for (i = start + 1; i < (start + size / PAGE_SIZE); i++) {
|
|
|
|
int pqtype;
|
|
|
|
pqtype = pga[i].queue - pga[i].pc;
|
|
|
|
if ((VM_PAGE_TO_PHYS(&pga[i]) !=
|
|
|
|
(VM_PAGE_TO_PHYS(&pga[i - 1]) + PAGE_SIZE)) ||
|
|
|
|
((pqtype != PQ_ZERO) && (pqtype != PQ_FREE) && (pqtype != PQ_CACHE))) {
|
|
|
|
start++;
|
|
|
|
goto again;
|
|
|
|
}
|
1996-05-18 03:38:05 +00:00
|
|
|
}
|
|
|
|
|
1996-11-05 04:19:08 +00:00
|
|
|
for (i = start; i < (start + size / PAGE_SIZE); i++) {
|
|
|
|
int pqtype;
|
|
|
|
vm_page_t m = &pga[i];
|
1996-05-18 03:38:05 +00:00
|
|
|
|
1996-11-05 04:19:08 +00:00
|
|
|
pqtype = m->queue - m->pc;
|
1998-01-31 11:56:53 +00:00
|
|
|
if (pqtype == PQ_CACHE) {
|
1998-09-04 08:06:57 +00:00
|
|
|
vm_page_busy(m);
|
1996-11-05 04:19:08 +00:00
|
|
|
vm_page_free(m);
|
1998-01-31 11:56:53 +00:00
|
|
|
}
|
1996-05-18 03:38:05 +00:00
|
|
|
|
1996-11-05 04:19:08 +00:00
|
|
|
TAILQ_REMOVE(vm_page_queues[m->queue].pl, m, pageq);
|
1998-01-31 11:56:53 +00:00
|
|
|
(*vm_page_queues[m->queue].lcnt)--;
|
1996-11-05 04:19:08 +00:00
|
|
|
cnt.v_free_count--;
|
|
|
|
m->valid = VM_PAGE_BITS_ALL;
|
|
|
|
m->flags = 0;
|
|
|
|
m->dirty = 0;
|
|
|
|
m->wire_count = 0;
|
|
|
|
m->busy = 0;
|
|
|
|
m->queue = PQ_NONE;
|
|
|
|
m->object = NULL;
|
|
|
|
vm_page_wire(m);
|
|
|
|
}
|
1996-05-18 03:38:05 +00:00
|
|
|
|
1996-11-05 04:19:08 +00:00
|
|
|
/*
|
|
|
|
* We've found a contiguous chunk that meets are requirements.
|
|
|
|
* Allocate kernel VM, unfree and assign the physical pages to it and
|
|
|
|
* return kernel VM pointer.
|
|
|
|
*/
|
1997-02-13 19:37:40 +00:00
|
|
|
tmp_addr = addr = kmem_alloc_pageable(map, size);
|
1996-11-05 04:19:08 +00:00
|
|
|
if (addr == 0) {
|
|
|
|
/*
|
|
|
|
* XXX We almost never run out of kernel virtual
|
|
|
|
* space, so we don't make the allocated memory
|
|
|
|
* above available.
|
|
|
|
*/
|
|
|
|
splx(s);
|
|
|
|
return (NULL);
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = start; i < (start + size / PAGE_SIZE); i++) {
|
|
|
|
vm_page_t m = &pga[i];
|
|
|
|
vm_page_insert(m, kernel_object,
|
|
|
|
OFF_TO_IDX(tmp_addr - VM_MIN_KERNEL_ADDRESS));
|
|
|
|
pmap_kenter(tmp_addr, VM_PAGE_TO_PHYS(m));
|
|
|
|
tmp_addr += PAGE_SIZE;
|
|
|
|
}
|
|
|
|
|
|
|
|
splx(s);
|
|
|
|
return ((void *)addr);
|
|
|
|
}
|
|
|
|
return NULL;
|
1996-05-18 03:38:05 +00:00
|
|
|
}
|
|
|
|
|
1997-02-13 19:37:40 +00:00
|
|
|
void *
|
|
|
|
contigmalloc(size, type, flags, low, high, alignment, boundary)
|
|
|
|
unsigned long size; /* should be size_t here and for malloc() */
|
1997-10-10 18:18:47 +00:00
|
|
|
struct malloc_type *type;
|
1997-02-13 19:37:40 +00:00
|
|
|
int flags;
|
|
|
|
unsigned long low;
|
|
|
|
unsigned long high;
|
|
|
|
unsigned long alignment;
|
|
|
|
unsigned long boundary;
|
|
|
|
{
|
|
|
|
return contigmalloc1(size, type, flags, low, high, alignment, boundary,
|
|
|
|
kernel_map);
|
|
|
|
}
|
|
|
|
|
1996-05-18 03:38:05 +00:00
|
|
|
vm_offset_t
|
|
|
|
vm_page_alloc_contig(size, low, high, alignment)
|
|
|
|
vm_offset_t size;
|
|
|
|
vm_offset_t low;
|
|
|
|
vm_offset_t high;
|
|
|
|
vm_offset_t alignment;
|
|
|
|
{
|
1997-02-13 19:37:40 +00:00
|
|
|
return ((vm_offset_t)contigmalloc1(size, M_DEVBUF, M_NOWAIT, low, high,
|
|
|
|
alignment, 0ul, kernel_map));
|
1996-05-18 03:38:05 +00:00
|
|
|
}
|
1996-09-14 11:54:59 +00:00
|
|
|
|
|
|
|
#include "opt_ddb.h"
|
1995-04-16 09:59:16 +00:00
|
|
|
#ifdef DDB
|
1996-09-14 11:54:59 +00:00
|
|
|
#include <sys/kernel.h>
|
|
|
|
|
|
|
|
#include <ddb/ddb.h>
|
|
|
|
|
|
|
|
DB_SHOW_COMMAND(page, vm_page_print_page_info)
|
These changes embody the support of the fully coherent merged VM buffer cache,
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
|
|
|
{
|
1996-09-14 11:54:59 +00:00
|
|
|
db_printf("cnt.v_free_count: %d\n", cnt.v_free_count);
|
|
|
|
db_printf("cnt.v_cache_count: %d\n", cnt.v_cache_count);
|
|
|
|
db_printf("cnt.v_inactive_count: %d\n", cnt.v_inactive_count);
|
|
|
|
db_printf("cnt.v_active_count: %d\n", cnt.v_active_count);
|
|
|
|
db_printf("cnt.v_wire_count: %d\n", cnt.v_wire_count);
|
|
|
|
db_printf("cnt.v_free_reserved: %d\n", cnt.v_free_reserved);
|
|
|
|
db_printf("cnt.v_free_min: %d\n", cnt.v_free_min);
|
|
|
|
db_printf("cnt.v_free_target: %d\n", cnt.v_free_target);
|
|
|
|
db_printf("cnt.v_cache_min: %d\n", cnt.v_cache_min);
|
|
|
|
db_printf("cnt.v_inactive_target: %d\n", cnt.v_inactive_target);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
1996-09-08 20:44:49 +00:00
|
|
|
|
1996-09-14 11:54:59 +00:00
|
|
|
DB_SHOW_COMMAND(pageq, vm_page_print_pageq_info)
|
1996-09-08 20:44:49 +00:00
|
|
|
{
|
|
|
|
int i;
|
1996-09-14 11:54:59 +00:00
|
|
|
db_printf("PQ_FREE:");
|
1996-09-08 20:44:49 +00:00
|
|
|
for(i=0;i<PQ_L2_SIZE;i++) {
|
1996-09-14 11:54:59 +00:00
|
|
|
db_printf(" %d", *vm_page_queues[PQ_FREE + i].lcnt);
|
1996-09-08 20:44:49 +00:00
|
|
|
}
|
1996-09-14 11:54:59 +00:00
|
|
|
db_printf("\n");
|
1996-09-08 20:44:49 +00:00
|
|
|
|
1996-09-14 11:54:59 +00:00
|
|
|
db_printf("PQ_CACHE:");
|
1996-09-08 20:44:49 +00:00
|
|
|
for(i=0;i<PQ_L2_SIZE;i++) {
|
1996-09-14 11:54:59 +00:00
|
|
|
db_printf(" %d", *vm_page_queues[PQ_CACHE + i].lcnt);
|
1996-09-08 20:44:49 +00:00
|
|
|
}
|
1996-09-14 11:54:59 +00:00
|
|
|
db_printf("\n");
|
1996-09-08 20:44:49 +00:00
|
|
|
|
1996-09-14 11:54:59 +00:00
|
|
|
db_printf("PQ_ZERO:");
|
1996-09-08 20:44:49 +00:00
|
|
|
for(i=0;i<PQ_L2_SIZE;i++) {
|
1996-09-14 11:54:59 +00:00
|
|
|
db_printf(" %d", *vm_page_queues[PQ_ZERO + i].lcnt);
|
1996-09-08 20:44:49 +00:00
|
|
|
}
|
1996-09-14 11:54:59 +00:00
|
|
|
db_printf("\n");
|
1996-09-08 20:44:49 +00:00
|
|
|
|
1996-09-14 11:54:59 +00:00
|
|
|
db_printf("PQ_ACTIVE: %d, PQ_INACTIVE: %d\n",
|
1996-09-08 20:44:49 +00:00
|
|
|
*vm_page_queues[PQ_ACTIVE].lcnt,
|
|
|
|
*vm_page_queues[PQ_INACTIVE].lcnt);
|
|
|
|
}
|
1996-09-14 11:54:59 +00:00
|
|
|
#endif /* DDB */
|