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
|
|
|
* Copyright (c) 1991, 1993
|
|
|
|
* The Regents of the University of California. All rights reserved.
|
|
|
|
*
|
|
|
|
* 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.
|
|
|
|
* 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.
|
|
|
|
*
|
1994-08-02 07:55:43 +00:00
|
|
|
* from: @(#)vm_glue.c 8.6 (Berkeley) 1/5/94
|
1994-05-24 10:09:53 +00:00
|
|
|
*
|
|
|
|
*
|
|
|
|
* Copyright (c) 1987, 1990 Carnegie-Mellon University.
|
|
|
|
* All rights reserved.
|
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.
|
|
|
|
*/
|
|
|
|
|
2003-06-11 23:50:51 +00:00
|
|
|
#include <sys/cdefs.h>
|
|
|
|
__FBSDID("$FreeBSD$");
|
|
|
|
|
1998-09-29 17:33:59 +00:00
|
|
|
#include "opt_vm.h"
|
2003-07-31 01:25:05 +00:00
|
|
|
#include "opt_kstack_pages.h"
|
|
|
|
#include "opt_kstack_max_pages.h"
|
1996-12-22 23:17:09 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <sys/param.h>
|
|
|
|
#include <sys/systm.h>
|
2003-04-29 13:36:06 +00:00
|
|
|
#include <sys/limits.h>
|
2001-05-01 08:13:21 +00:00
|
|
|
#include <sys/lock.h>
|
|
|
|
#include <sys/mutex.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <sys/proc.h>
|
|
|
|
#include <sys/resourcevar.h>
|
1995-03-28 07:58:53 +00:00
|
|
|
#include <sys/shm.h>
|
1995-12-07 12:48:31 +00:00
|
|
|
#include <sys/vmmeter.h>
|
2001-03-28 11:52:56 +00:00
|
|
|
#include <sys/sx.h>
|
1997-12-06 02:23:36 +00:00
|
|
|
#include <sys/sysctl.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1994-05-25 09:21:21 +00:00
|
|
|
#include <sys/kernel.h>
|
2000-09-07 01:33:02 +00:00
|
|
|
#include <sys/ktr.h>
|
1997-04-07 07:16:06 +00:00
|
|
|
#include <sys/unistd.h>
|
1994-05-25 09:21:21 +00:00
|
|
|
|
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/pmap.h>
|
|
|
|
#include <vm/vm_map.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <vm/vm_page.h>
|
1994-05-25 09:21:21 +00:00
|
|
|
#include <vm/vm_pageout.h>
|
2002-07-07 23:05:27 +00:00
|
|
|
#include <vm/vm_object.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <vm/vm_kern.h>
|
1995-12-07 12:48:31 +00:00
|
|
|
#include <vm/vm_extern.h>
|
2002-07-07 23:05:27 +00:00
|
|
|
#include <vm/vm_pager.h>
|
2002-12-15 19:17:57 +00:00
|
|
|
#include <vm/swap_pager.h>
|
1995-12-07 12:48:31 +00:00
|
|
|
|
|
|
|
#include <sys/user.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2001-05-18 00:08:38 +00:00
|
|
|
extern int maxslp;
|
|
|
|
|
1995-08-28 09:19:25 +00:00
|
|
|
/*
|
|
|
|
* System initialization
|
|
|
|
*
|
|
|
|
* Note: proc0 from proc.h
|
|
|
|
*/
|
2002-03-19 22:20:14 +00:00
|
|
|
static void vm_init_limits(void *);
|
1995-09-09 18:10:37 +00:00
|
|
|
SYSINIT(vm_limits, SI_SUB_VM_CONF, SI_ORDER_FIRST, vm_init_limits, &proc0)
|
1995-08-28 09:19:25 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* THIS MUST BE THE LAST INITIALIZATION ITEM!!!
|
|
|
|
*
|
|
|
|
* Note: run scheduling should be divorced from the vm system.
|
|
|
|
*/
|
2002-03-19 22:20:14 +00:00
|
|
|
static void scheduler(void *);
|
2004-01-29 12:35:11 +00:00
|
|
|
SYSINIT(scheduler, SI_SUB_RUN_SCHEDULER, SI_ORDER_ANY, scheduler, NULL)
|
1995-08-28 09:19:25 +00:00
|
|
|
|
2002-01-17 16:46:26 +00:00
|
|
|
#ifndef NO_SWAPPING
|
2002-03-19 22:20:14 +00:00
|
|
|
static void swapout(struct proc *);
|
2002-07-07 23:05:27 +00:00
|
|
|
static void vm_proc_swapin(struct proc *p);
|
|
|
|
static void vm_proc_swapout(struct proc *p);
|
2002-01-17 16:46:26 +00:00
|
|
|
#endif
|
1995-12-14 09:55:16 +00:00
|
|
|
|
2002-06-22 01:26:02 +00:00
|
|
|
/*
|
|
|
|
* MPSAFE
|
2003-01-20 17:46:48 +00:00
|
|
|
*
|
|
|
|
* WARNING! This code calls vm_map_check_protection() which only checks
|
|
|
|
* the associated vm_map_entry range. It does not determine whether the
|
|
|
|
* contents of the memory is actually readable or writable. In most cases
|
|
|
|
* just checking the vm_map_entry is sufficient within the kernel's address
|
|
|
|
* space.
|
2002-06-22 01:26:02 +00:00
|
|
|
*/
|
1994-05-24 10:09:53 +00:00
|
|
|
int
|
|
|
|
kernacc(addr, len, rw)
|
2003-01-21 11:34:57 +00:00
|
|
|
void *addr;
|
1994-05-24 10:09:53 +00:00
|
|
|
int len, rw;
|
|
|
|
{
|
|
|
|
boolean_t rv;
|
|
|
|
vm_offset_t saddr, eaddr;
|
1999-10-30 06:32:05 +00:00
|
|
|
vm_prot_t prot;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2002-01-17 16:46:26 +00:00
|
|
|
KASSERT((rw & ~VM_PROT_ALL) == 0,
|
1999-10-30 06:32:05 +00:00
|
|
|
("illegal ``rw'' argument to kernacc (%x)\n", rw));
|
|
|
|
prot = rw;
|
1998-10-13 08:24:45 +00:00
|
|
|
saddr = trunc_page((vm_offset_t)addr);
|
|
|
|
eaddr = round_page((vm_offset_t)addr + len);
|
2003-11-10 01:37:40 +00:00
|
|
|
vm_map_lock_read(kernel_map);
|
1994-05-24 10:09:53 +00:00
|
|
|
rv = vm_map_check_protection(kernel_map, saddr, eaddr, prot);
|
2003-11-10 01:37:40 +00:00
|
|
|
vm_map_unlock_read(kernel_map);
|
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 (rv == TRUE);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
2002-06-22 01:26:02 +00:00
|
|
|
/*
|
|
|
|
* MPSAFE
|
2003-01-20 17:46:48 +00:00
|
|
|
*
|
|
|
|
* WARNING! This code calls vm_map_check_protection() which only checks
|
|
|
|
* the associated vm_map_entry range. It does not determine whether the
|
|
|
|
* contents of the memory is actually readable or writable. vmapbuf(),
|
|
|
|
* vm_fault_quick(), or copyin()/copout()/su*()/fu*() functions should be
|
|
|
|
* used in conjuction with this call.
|
2002-06-22 01:26:02 +00:00
|
|
|
*/
|
1994-05-24 10:09:53 +00:00
|
|
|
int
|
|
|
|
useracc(addr, len, rw)
|
2003-01-21 11:34:57 +00:00
|
|
|
void *addr;
|
1994-05-24 10:09:53 +00:00
|
|
|
int len, rw;
|
|
|
|
{
|
|
|
|
boolean_t rv;
|
1999-10-30 06:32:05 +00:00
|
|
|
vm_prot_t prot;
|
2002-09-21 22:07:17 +00:00
|
|
|
vm_map_t map;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2002-01-17 16:46:26 +00:00
|
|
|
KASSERT((rw & ~VM_PROT_ALL) == 0,
|
1999-10-30 06:32:05 +00:00
|
|
|
("illegal ``rw'' argument to useracc (%x)\n", rw));
|
|
|
|
prot = rw;
|
2002-09-21 22:07:17 +00:00
|
|
|
map = &curproc->p_vmspace->vm_map;
|
|
|
|
if ((vm_offset_t)addr + len > vm_map_max(map) ||
|
|
|
|
(vm_offset_t)addr + len < (vm_offset_t)addr) {
|
1994-05-25 09:21:21 +00:00
|
|
|
return (FALSE);
|
|
|
|
}
|
2003-11-10 01:37:40 +00:00
|
|
|
vm_map_lock_read(map);
|
2002-09-21 22:07:17 +00:00
|
|
|
rv = vm_map_check_protection(map, trunc_page((vm_offset_t)addr),
|
|
|
|
round_page((vm_offset_t)addr + len), prot);
|
2003-11-10 01:37:40 +00:00
|
|
|
vm_map_unlock_read(map);
|
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 (rv == TRUE);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
2004-03-05 22:03:11 +00:00
|
|
|
int
|
2004-03-15 06:43:51 +00:00
|
|
|
vslock(void *addr, size_t len)
|
2004-03-05 22:03:11 +00:00
|
|
|
{
|
2004-03-15 09:11:23 +00:00
|
|
|
vm_offset_t end, last, start;
|
|
|
|
vm_size_t npages;
|
|
|
|
int error;
|
2004-03-05 22:03:11 +00:00
|
|
|
|
2004-03-15 09:11:23 +00:00
|
|
|
last = (vm_offset_t)addr + len;
|
2004-03-15 06:42:40 +00:00
|
|
|
start = trunc_page((vm_offset_t)addr);
|
2004-03-15 09:11:23 +00:00
|
|
|
end = round_page(last);
|
|
|
|
if (last < (vm_offset_t)addr || end < (vm_offset_t)addr)
|
2004-03-05 22:03:11 +00:00
|
|
|
return (EINVAL);
|
|
|
|
npages = atop(end - start);
|
|
|
|
if (npages > vm_page_max_wired)
|
|
|
|
return (ENOMEM);
|
2004-03-15 06:42:40 +00:00
|
|
|
PROC_LOCK(curproc);
|
2004-03-15 09:11:23 +00:00
|
|
|
if (ptoa(npages +
|
|
|
|
pmap_wired_count(vm_map_pmap(&curproc->p_vmspace->vm_map))) >
|
|
|
|
lim_cur(curproc, RLIMIT_MEMLOCK)) {
|
2004-03-15 06:42:40 +00:00
|
|
|
PROC_UNLOCK(curproc);
|
2004-03-05 22:03:11 +00:00
|
|
|
return (ENOMEM);
|
|
|
|
}
|
2004-03-15 06:42:40 +00:00
|
|
|
PROC_UNLOCK(curproc);
|
2004-03-05 22:03:11 +00:00
|
|
|
#if 0
|
|
|
|
/*
|
|
|
|
* XXX - not yet
|
|
|
|
*
|
|
|
|
* The limit for transient usage of wired pages should be
|
|
|
|
* larger than for "permanent" wired pages (mlock()).
|
|
|
|
*
|
|
|
|
* Also, the sysctl code, which is the only present user
|
|
|
|
* of vslock(), does a hard loop on EAGAIN.
|
|
|
|
*/
|
|
|
|
if (npages + cnt.v_wire_count > vm_page_max_wired)
|
|
|
|
return (EAGAIN);
|
|
|
|
#endif
|
2004-03-15 06:42:40 +00:00
|
|
|
error = vm_map_wire(&curproc->p_vmspace->vm_map, start, end,
|
2004-05-07 11:43:24 +00:00
|
|
|
VM_MAP_WIRE_SYSTEM | VM_MAP_WIRE_NOHOLES);
|
2004-03-15 06:42:40 +00:00
|
|
|
/*
|
|
|
|
* Return EFAULT on error to match copy{in,out}() behaviour
|
|
|
|
* rather than returning ENOMEM like mlock() would.
|
|
|
|
*/
|
|
|
|
return (error == KERN_SUCCESS ? 0 : EFAULT);
|
2004-03-05 22:03:11 +00:00
|
|
|
}
|
|
|
|
|
2004-03-15 06:42:40 +00:00
|
|
|
void
|
2004-03-15 06:43:51 +00:00
|
|
|
vsunlock(void *addr, size_t len)
|
2004-03-05 22:03:11 +00:00
|
|
|
{
|
|
|
|
|
2004-03-15 06:42:40 +00:00
|
|
|
/* Rely on the parameter sanity checks performed by vslock(). */
|
|
|
|
(void)vm_map_unwire(&curproc->p_vmspace->vm_map,
|
|
|
|
trunc_page((vm_offset_t)addr), round_page((vm_offset_t)addr + len),
|
2004-03-15 06:43:51 +00:00
|
|
|
VM_MAP_WIRE_SYSTEM | VM_MAP_WIRE_NOHOLES);
|
2004-03-05 22:03:11 +00:00
|
|
|
}
|
|
|
|
|
2002-07-07 23:05:27 +00:00
|
|
|
/*
|
|
|
|
* Create the U area for a new process.
|
|
|
|
* This routine directly affects the fork perf for a process.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
vm_proc_new(struct proc *p)
|
|
|
|
{
|
|
|
|
vm_page_t ma[UAREA_PAGES];
|
|
|
|
vm_object_t upobj;
|
|
|
|
vm_offset_t up;
|
|
|
|
vm_page_t m;
|
|
|
|
u_int i;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get a kernel virtual address for the U area for this process.
|
|
|
|
*/
|
|
|
|
up = kmem_alloc_nofault(kernel_map, UAREA_PAGES * PAGE_SIZE);
|
|
|
|
if (up == 0)
|
|
|
|
panic("vm_proc_new: upage allocation failed");
|
|
|
|
p->p_uarea = (struct user *)up;
|
|
|
|
|
2003-08-18 01:31:43 +00:00
|
|
|
/*
|
|
|
|
* Allocate object and page(s) for the U area.
|
|
|
|
*/
|
|
|
|
upobj = vm_object_allocate(OBJT_DEFAULT, UAREA_PAGES);
|
|
|
|
p->p_upages_obj = upobj;
|
|
|
|
VM_OBJECT_LOCK(upobj);
|
2002-07-07 23:05:27 +00:00
|
|
|
for (i = 0; i < UAREA_PAGES; i++) {
|
2002-07-29 05:42:44 +00:00
|
|
|
m = vm_page_grab(upobj, i,
|
|
|
|
VM_ALLOC_NORMAL | VM_ALLOC_RETRY | VM_ALLOC_WIRED);
|
2002-07-07 23:05:27 +00:00
|
|
|
ma[i] = m;
|
|
|
|
|
2002-12-24 04:24:58 +00:00
|
|
|
vm_page_lock_queues();
|
2002-07-07 23:05:27 +00:00
|
|
|
vm_page_wakeup(m);
|
|
|
|
m->valid = VM_PAGE_BITS_ALL;
|
2002-12-24 04:24:58 +00:00
|
|
|
vm_page_unlock_queues();
|
2002-07-07 23:05:27 +00:00
|
|
|
}
|
2003-08-18 01:31:43 +00:00
|
|
|
VM_OBJECT_UNLOCK(upobj);
|
2002-07-07 23:05:27 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Enter the pages into the kernel address space.
|
|
|
|
*/
|
|
|
|
pmap_qenter(up, ma, UAREA_PAGES);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Dispose the U area for a process that has exited.
|
|
|
|
* This routine directly impacts the exit perf of a process.
|
|
|
|
* XXX proc_zone is marked UMA_ZONE_NOFREE, so this should never be called.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
vm_proc_dispose(struct proc *p)
|
|
|
|
{
|
|
|
|
vm_object_t upobj;
|
|
|
|
vm_offset_t up;
|
|
|
|
vm_page_t m;
|
|
|
|
|
|
|
|
upobj = p->p_upages_obj;
|
2003-04-25 16:30:02 +00:00
|
|
|
VM_OBJECT_LOCK(upobj);
|
2002-07-08 01:11:10 +00:00
|
|
|
if (upobj->resident_page_count != UAREA_PAGES)
|
|
|
|
panic("vm_proc_dispose: incorrect number of pages in upobj");
|
2002-07-13 19:24:04 +00:00
|
|
|
vm_page_lock_queues();
|
2002-07-08 01:11:10 +00:00
|
|
|
while ((m = TAILQ_FIRST(&upobj->memq)) != NULL) {
|
2002-07-07 23:05:27 +00:00
|
|
|
vm_page_busy(m);
|
|
|
|
vm_page_unwire(m, 0);
|
|
|
|
vm_page_free(m);
|
|
|
|
}
|
2002-07-13 19:24:04 +00:00
|
|
|
vm_page_unlock_queues();
|
2003-04-25 16:30:02 +00:00
|
|
|
VM_OBJECT_UNLOCK(upobj);
|
2002-07-08 01:11:10 +00:00
|
|
|
up = (vm_offset_t)p->p_uarea;
|
2002-07-07 23:05:27 +00:00
|
|
|
pmap_qremove(up, UAREA_PAGES);
|
|
|
|
kmem_free(kernel_map, up, UAREA_PAGES * PAGE_SIZE);
|
|
|
|
vm_object_deallocate(upobj);
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifndef NO_SWAPPING
|
|
|
|
/*
|
|
|
|
* Allow the U area for a process to be prejudicially paged out.
|
|
|
|
*/
|
2002-09-28 17:15:38 +00:00
|
|
|
static void
|
2002-07-07 23:05:27 +00:00
|
|
|
vm_proc_swapout(struct proc *p)
|
|
|
|
{
|
|
|
|
vm_object_t upobj;
|
|
|
|
vm_offset_t up;
|
|
|
|
vm_page_t m;
|
|
|
|
|
|
|
|
upobj = p->p_upages_obj;
|
2003-04-25 16:30:02 +00:00
|
|
|
VM_OBJECT_LOCK(upobj);
|
2002-07-08 01:11:10 +00:00
|
|
|
if (upobj->resident_page_count != UAREA_PAGES)
|
|
|
|
panic("vm_proc_dispose: incorrect number of pages in upobj");
|
2002-07-13 19:24:04 +00:00
|
|
|
vm_page_lock_queues();
|
2002-07-08 01:11:10 +00:00
|
|
|
TAILQ_FOREACH(m, &upobj->memq, listq) {
|
2002-07-07 23:05:27 +00:00
|
|
|
vm_page_dirty(m);
|
|
|
|
vm_page_unwire(m, 0);
|
|
|
|
}
|
2002-07-13 19:24:04 +00:00
|
|
|
vm_page_unlock_queues();
|
2003-04-25 16:30:02 +00:00
|
|
|
VM_OBJECT_UNLOCK(upobj);
|
2002-07-08 01:11:10 +00:00
|
|
|
up = (vm_offset_t)p->p_uarea;
|
2002-07-07 23:05:27 +00:00
|
|
|
pmap_qremove(up, UAREA_PAGES);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Bring the U area for a specified process back in.
|
|
|
|
*/
|
2002-09-28 17:15:38 +00:00
|
|
|
static void
|
2002-07-07 23:05:27 +00:00
|
|
|
vm_proc_swapin(struct proc *p)
|
|
|
|
{
|
|
|
|
vm_page_t ma[UAREA_PAGES];
|
|
|
|
vm_object_t upobj;
|
|
|
|
vm_offset_t up;
|
|
|
|
vm_page_t m;
|
|
|
|
int rv;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
upobj = p->p_upages_obj;
|
2003-06-13 03:02:28 +00:00
|
|
|
VM_OBJECT_LOCK(upobj);
|
2002-07-07 23:05:27 +00:00
|
|
|
for (i = 0; i < UAREA_PAGES; i++) {
|
|
|
|
m = vm_page_grab(upobj, i, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
|
|
|
|
if (m->valid != VM_PAGE_BITS_ALL) {
|
|
|
|
rv = vm_pager_get_pages(upobj, &m, 1, 0);
|
|
|
|
if (rv != VM_PAGER_OK)
|
|
|
|
panic("vm_proc_swapin: cannot get upage");
|
|
|
|
}
|
|
|
|
ma[i] = m;
|
2002-07-12 04:38:51 +00:00
|
|
|
}
|
|
|
|
if (upobj->resident_page_count != UAREA_PAGES)
|
|
|
|
panic("vm_proc_swapin: lost pages from upobj");
|
2002-07-14 19:36:15 +00:00
|
|
|
vm_page_lock_queues();
|
2002-07-12 04:38:51 +00:00
|
|
|
TAILQ_FOREACH(m, &upobj->memq, listq) {
|
|
|
|
m->valid = VM_PAGE_BITS_ALL;
|
2002-07-07 23:05:27 +00:00
|
|
|
vm_page_wire(m);
|
|
|
|
vm_page_wakeup(m);
|
|
|
|
}
|
2002-07-14 19:36:15 +00:00
|
|
|
vm_page_unlock_queues();
|
2003-04-25 16:30:02 +00:00
|
|
|
VM_OBJECT_UNLOCK(upobj);
|
2002-07-08 01:11:10 +00:00
|
|
|
up = (vm_offset_t)p->p_uarea;
|
2002-07-07 23:05:27 +00:00
|
|
|
pmap_qenter(up, ma, UAREA_PAGES);
|
|
|
|
}
|
2002-12-15 19:17:57 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Swap in the UAREAs of all processes swapped out to the given device.
|
|
|
|
* The pages in the UAREA are marked dirty and their swap metadata is freed.
|
|
|
|
*/
|
|
|
|
void
|
2003-08-03 13:35:31 +00:00
|
|
|
vm_proc_swapin_all(struct swdevt *devidx)
|
2002-12-15 19:17:57 +00:00
|
|
|
{
|
|
|
|
struct proc *p;
|
|
|
|
vm_object_t object;
|
|
|
|
vm_page_t m;
|
|
|
|
|
|
|
|
retry:
|
|
|
|
sx_slock(&allproc_lock);
|
|
|
|
FOREACH_PROC_IN_SYSTEM(p) {
|
|
|
|
PROC_LOCK(p);
|
|
|
|
object = p->p_upages_obj;
|
2003-04-28 17:13:53 +00:00
|
|
|
if (object != NULL) {
|
2003-04-25 16:30:02 +00:00
|
|
|
VM_OBJECT_LOCK(object);
|
2003-04-28 17:13:53 +00:00
|
|
|
if (swap_pager_isswapped(object, devidx)) {
|
|
|
|
VM_OBJECT_UNLOCK(object);
|
|
|
|
sx_sunlock(&allproc_lock);
|
|
|
|
faultin(p);
|
|
|
|
PROC_UNLOCK(p);
|
|
|
|
VM_OBJECT_LOCK(object);
|
|
|
|
vm_page_lock_queues();
|
|
|
|
TAILQ_FOREACH(m, &object->memq, listq)
|
|
|
|
vm_page_dirty(m);
|
|
|
|
vm_page_unlock_queues();
|
|
|
|
swap_pager_freespace(object, 0,
|
|
|
|
object->un_pager.swp.swp_bcount);
|
|
|
|
VM_OBJECT_UNLOCK(object);
|
|
|
|
goto retry;
|
|
|
|
}
|
2003-04-25 16:30:02 +00:00
|
|
|
VM_OBJECT_UNLOCK(object);
|
2002-12-15 19:17:57 +00:00
|
|
|
}
|
|
|
|
PROC_UNLOCK(p);
|
|
|
|
}
|
|
|
|
sx_sunlock(&allproc_lock);
|
|
|
|
}
|
2002-07-07 23:05:27 +00:00
|
|
|
#endif
|
|
|
|
|
2003-06-14 23:23:55 +00:00
|
|
|
#ifndef KSTACK_MAX_PAGES
|
|
|
|
#define KSTACK_MAX_PAGES 32
|
|
|
|
#endif
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Create the kernel stack (including pcb for i386) for a new thread.
|
|
|
|
* This routine directly affects the fork perf for a process and
|
|
|
|
* create performance for a thread.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
vm_thread_new(struct thread *td, int pages)
|
|
|
|
{
|
|
|
|
vm_object_t ksobj;
|
|
|
|
vm_offset_t ks;
|
|
|
|
vm_page_t m, ma[KSTACK_MAX_PAGES];
|
|
|
|
int i;
|
|
|
|
|
|
|
|
/* Bounds check */
|
|
|
|
if (pages <= 1)
|
|
|
|
pages = KSTACK_PAGES;
|
|
|
|
else if (pages > KSTACK_MAX_PAGES)
|
|
|
|
pages = KSTACK_MAX_PAGES;
|
|
|
|
/*
|
|
|
|
* Allocate an object for the kstack.
|
|
|
|
*/
|
|
|
|
ksobj = vm_object_allocate(OBJT_DEFAULT, pages);
|
|
|
|
td->td_kstack_obj = ksobj;
|
|
|
|
/*
|
|
|
|
* Get a kernel virtual address for this thread's kstack.
|
|
|
|
*/
|
|
|
|
ks = kmem_alloc_nofault(kernel_map,
|
|
|
|
(pages + KSTACK_GUARD_PAGES) * PAGE_SIZE);
|
|
|
|
if (ks == 0)
|
|
|
|
panic("vm_thread_new: kstack allocation failed");
|
|
|
|
if (KSTACK_GUARD_PAGES != 0) {
|
|
|
|
pmap_qremove(ks, KSTACK_GUARD_PAGES);
|
|
|
|
ks += KSTACK_GUARD_PAGES * PAGE_SIZE;
|
|
|
|
}
|
|
|
|
td->td_kstack = ks;
|
|
|
|
/*
|
|
|
|
* Knowing the number of pages allocated is useful when you
|
|
|
|
* want to deallocate them.
|
|
|
|
*/
|
|
|
|
td->td_kstack_pages = pages;
|
|
|
|
/*
|
|
|
|
* For the length of the stack, link in a real page of ram for each
|
|
|
|
* page of stack.
|
|
|
|
*/
|
|
|
|
VM_OBJECT_LOCK(ksobj);
|
|
|
|
for (i = 0; i < pages; i++) {
|
|
|
|
/*
|
|
|
|
* Get a kernel stack page.
|
|
|
|
*/
|
|
|
|
m = vm_page_grab(ksobj, i,
|
|
|
|
VM_ALLOC_NORMAL | VM_ALLOC_RETRY | VM_ALLOC_WIRED);
|
|
|
|
ma[i] = m;
|
|
|
|
vm_page_lock_queues();
|
|
|
|
vm_page_wakeup(m);
|
|
|
|
m->valid = VM_PAGE_BITS_ALL;
|
|
|
|
vm_page_unlock_queues();
|
|
|
|
}
|
|
|
|
VM_OBJECT_UNLOCK(ksobj);
|
|
|
|
pmap_qenter(ks, ma, pages);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Dispose of a thread's kernel stack.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
vm_thread_dispose(struct thread *td)
|
|
|
|
{
|
|
|
|
vm_object_t ksobj;
|
|
|
|
vm_offset_t ks;
|
|
|
|
vm_page_t m;
|
|
|
|
int i, pages;
|
|
|
|
|
|
|
|
pages = td->td_kstack_pages;
|
|
|
|
ksobj = td->td_kstack_obj;
|
|
|
|
ks = td->td_kstack;
|
|
|
|
pmap_qremove(ks, pages);
|
|
|
|
VM_OBJECT_LOCK(ksobj);
|
|
|
|
for (i = 0; i < pages; i++) {
|
|
|
|
m = vm_page_lookup(ksobj, i);
|
|
|
|
if (m == NULL)
|
|
|
|
panic("vm_thread_dispose: kstack already missing?");
|
|
|
|
vm_page_lock_queues();
|
|
|
|
vm_page_busy(m);
|
|
|
|
vm_page_unwire(m, 0);
|
|
|
|
vm_page_free(m);
|
|
|
|
vm_page_unlock_queues();
|
|
|
|
}
|
|
|
|
VM_OBJECT_UNLOCK(ksobj);
|
|
|
|
vm_object_deallocate(ksobj);
|
|
|
|
kmem_free(kernel_map, ks - (KSTACK_GUARD_PAGES * PAGE_SIZE),
|
|
|
|
(pages + KSTACK_GUARD_PAGES) * PAGE_SIZE);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Allow a thread's kernel stack to be paged out.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
vm_thread_swapout(struct thread *td)
|
|
|
|
{
|
|
|
|
vm_object_t ksobj;
|
|
|
|
vm_page_t m;
|
|
|
|
int i, pages;
|
|
|
|
|
2003-08-16 23:15:15 +00:00
|
|
|
cpu_thread_swapout(td);
|
2003-06-14 23:23:55 +00:00
|
|
|
pages = td->td_kstack_pages;
|
|
|
|
ksobj = td->td_kstack_obj;
|
|
|
|
pmap_qremove(td->td_kstack, pages);
|
|
|
|
VM_OBJECT_LOCK(ksobj);
|
|
|
|
for (i = 0; i < pages; i++) {
|
|
|
|
m = vm_page_lookup(ksobj, i);
|
|
|
|
if (m == NULL)
|
|
|
|
panic("vm_thread_swapout: kstack already missing?");
|
|
|
|
vm_page_lock_queues();
|
|
|
|
vm_page_dirty(m);
|
|
|
|
vm_page_unwire(m, 0);
|
|
|
|
vm_page_unlock_queues();
|
|
|
|
}
|
|
|
|
VM_OBJECT_UNLOCK(ksobj);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Bring the kernel stack for a specified thread back in.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
vm_thread_swapin(struct thread *td)
|
|
|
|
{
|
|
|
|
vm_object_t ksobj;
|
|
|
|
vm_page_t m, ma[KSTACK_MAX_PAGES];
|
|
|
|
int i, pages, rv;
|
|
|
|
|
|
|
|
pages = td->td_kstack_pages;
|
|
|
|
ksobj = td->td_kstack_obj;
|
|
|
|
VM_OBJECT_LOCK(ksobj);
|
|
|
|
for (i = 0; i < pages; i++) {
|
|
|
|
m = vm_page_grab(ksobj, i, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
|
|
|
|
if (m->valid != VM_PAGE_BITS_ALL) {
|
|
|
|
rv = vm_pager_get_pages(ksobj, &m, 1, 0);
|
|
|
|
if (rv != VM_PAGER_OK)
|
|
|
|
panic("vm_thread_swapin: cannot get kstack for proc: %d", td->td_proc->p_pid);
|
|
|
|
m = vm_page_lookup(ksobj, i);
|
|
|
|
m->valid = VM_PAGE_BITS_ALL;
|
|
|
|
}
|
|
|
|
ma[i] = m;
|
|
|
|
vm_page_lock_queues();
|
|
|
|
vm_page_wire(m);
|
|
|
|
vm_page_wakeup(m);
|
|
|
|
vm_page_unlock_queues();
|
|
|
|
}
|
|
|
|
VM_OBJECT_UNLOCK(ksobj);
|
|
|
|
pmap_qenter(td->td_kstack, ma, pages);
|
2003-08-16 23:15:15 +00:00
|
|
|
cpu_thread_swapin(td);
|
2003-06-14 23:23:55 +00:00
|
|
|
}
|
|
|
|
|
2003-06-14 06:20:25 +00:00
|
|
|
/*
|
|
|
|
* Set up a variable-sized alternate kstack.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
vm_thread_new_altkstack(struct thread *td, int pages)
|
|
|
|
{
|
|
|
|
|
|
|
|
td->td_altkstack = td->td_kstack;
|
|
|
|
td->td_altkstack_obj = td->td_kstack_obj;
|
|
|
|
td->td_altkstack_pages = td->td_kstack_pages;
|
|
|
|
|
2003-06-14 23:23:55 +00:00
|
|
|
vm_thread_new(td, pages);
|
2003-06-14 06:20:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Restore the original kstack.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
vm_thread_dispose_altkstack(struct thread *td)
|
|
|
|
{
|
|
|
|
|
2003-06-14 23:23:55 +00:00
|
|
|
vm_thread_dispose(td);
|
2003-06-14 06:20:25 +00:00
|
|
|
|
|
|
|
td->td_kstack = td->td_altkstack;
|
|
|
|
td->td_kstack_obj = td->td_altkstack_obj;
|
|
|
|
td->td_kstack_pages = td->td_altkstack_pages;
|
|
|
|
td->td_altkstack = 0;
|
|
|
|
td->td_altkstack_obj = NULL;
|
|
|
|
td->td_altkstack_pages = 0;
|
|
|
|
}
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Implement fork's actions on an address space.
|
|
|
|
* Here we arrange for the address space to be copied or referenced,
|
|
|
|
* allocate a user struct (pcb and kernel stack), then call the
|
|
|
|
* machine-dependent layer to fill those in and make the new process
|
1997-04-07 07:16:06 +00:00
|
|
|
* ready to run. The new process is set up so that it returns directly
|
|
|
|
* to user mode to avoid stack copying and relocation problems.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
1997-04-07 07:16:06 +00:00
|
|
|
void
|
2002-02-07 20:58:47 +00:00
|
|
|
vm_forkproc(td, p2, td2, flags)
|
2001-09-12 08:38:13 +00:00
|
|
|
struct thread *td;
|
|
|
|
struct proc *p2;
|
2002-02-07 20:58:47 +00:00
|
|
|
struct thread *td2;
|
1997-04-07 07:16:06 +00:00
|
|
|
int flags;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2001-09-12 08:38:13 +00:00
|
|
|
struct proc *p1 = td->td_proc;
|
1994-05-25 09:21:21 +00:00
|
|
|
|
2001-07-04 16:20:28 +00:00
|
|
|
GIANT_REQUIRED;
|
|
|
|
|
1999-12-06 04:53:08 +00:00
|
|
|
if ((flags & RFPROC) == 0) {
|
|
|
|
/*
|
|
|
|
* Divorce the memory, if it is shared, essentially
|
|
|
|
* this changes shared memory amongst threads, into
|
|
|
|
* COW locally.
|
|
|
|
*/
|
|
|
|
if ((flags & RFMEM) == 0) {
|
|
|
|
if (p1->p_vmspace->vm_refcnt > 1) {
|
|
|
|
vmspace_unshare(p1);
|
|
|
|
}
|
|
|
|
}
|
2002-02-07 20:58:47 +00:00
|
|
|
cpu_fork(td, p2, td2, flags);
|
1999-12-06 04:53:08 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
1997-04-13 01:48:35 +00:00
|
|
|
if (flags & RFMEM) {
|
|
|
|
p2->p_vmspace = p1->p_vmspace;
|
2004-07-27 03:53:41 +00:00
|
|
|
atomic_add_int(&p1->p_vmspace->vm_refcnt, 1);
|
1997-04-13 01:48:35 +00:00
|
|
|
}
|
|
|
|
|
1999-09-17 04:56:40 +00:00
|
|
|
while (vm_page_count_severe()) {
|
1994-05-25 09:21:21 +00:00
|
|
|
VM_WAIT;
|
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
|
|
|
|
1997-04-13 01:48:35 +00:00
|
|
|
if ((flags & RFMEM) == 0) {
|
1997-04-07 07:16:06 +00:00
|
|
|
p2->p_vmspace = vmspace_fork(p1->p_vmspace);
|
|
|
|
if (p1->p_vmspace->vm_shm)
|
|
|
|
shmfork(p1, p2);
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1998-12-19 08:23:31 +00:00
|
|
|
/*
|
2004-07-02 03:45:07 +00:00
|
|
|
* p_stats currently points at fields in the user struct.
|
|
|
|
* Copy parts of p_stats; zero the rest of p_stats (statistics).
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
2004-07-02 03:45:07 +00:00
|
|
|
#define RANGEOF(type, start, end) (offsetof(type, end) - offsetof(type, start))
|
|
|
|
|
|
|
|
p2->p_stats = &p2->p_uarea->u_stats;
|
|
|
|
bzero(&p2->p_stats->pstat_startzero,
|
|
|
|
(unsigned) RANGEOF(struct pstats, pstat_startzero, pstat_endzero));
|
|
|
|
bcopy(&p1->p_stats->pstat_startcopy, &p2->p_stats->pstat_startcopy,
|
|
|
|
(unsigned) RANGEOF(struct pstats, pstat_startcopy, pstat_endcopy));
|
|
|
|
#undef RANGEOF
|
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
|
|
|
/*
|
1997-04-07 07:16:06 +00:00
|
|
|
* cpu_fork will copy and update the pcb, set up the kernel stack,
|
|
|
|
* and make the child ready to run.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
2002-02-07 20:58:47 +00:00
|
|
|
cpu_fork(td, p2, td2, flags);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2001-09-10 04:28:58 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Called after process has been wait(2)'ed apon and is being reaped.
|
|
|
|
* The idea is to reclaim resources that we could not reclaim while
|
|
|
|
* the process was still executing.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
vm_waitproc(p)
|
|
|
|
struct proc *p;
|
|
|
|
{
|
|
|
|
|
2002-02-05 21:23:05 +00:00
|
|
|
vmspace_exitfree(p); /* and clean-out the vmspace */
|
2001-09-10 04:28:58 +00:00
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Set default limits for VM system.
|
|
|
|
* Called for proc 0, and then inherited by all others.
|
1995-08-28 09:19:25 +00:00
|
|
|
*
|
|
|
|
* XXX should probably act directly on proc0.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
1995-08-28 09:19:25 +00:00
|
|
|
static void
|
1995-09-09 18:10:37 +00:00
|
|
|
vm_init_limits(udata)
|
|
|
|
void *udata;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2001-07-04 19:00:13 +00:00
|
|
|
struct proc *p = udata;
|
2004-02-04 21:52:57 +00:00
|
|
|
struct plimit *limp;
|
1994-08-04 03:06:48 +00:00
|
|
|
int rss_limit;
|
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
|
|
|
* Set up the initial limits on process VM. Set the maximum resident
|
|
|
|
* set size to be half of (reasonably) available memory. Since this
|
|
|
|
* is a soft limit, it comes into effect only when the system is out
|
|
|
|
* of memory - half of main memory helps to favor smaller processes,
|
1994-08-04 03:06:48 +00:00
|
|
|
* and reduces thrashing of the object cache.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
2004-02-04 21:52:57 +00:00
|
|
|
limp = p->p_limit;
|
|
|
|
limp->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
|
|
|
|
limp->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
|
|
|
|
limp->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
|
|
|
|
limp->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
|
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
|
|
|
/* limit the limit to no less than 2MB */
|
1995-07-10 08:48:58 +00:00
|
|
|
rss_limit = max(cnt.v_free_count, 512);
|
2004-02-04 21:52:57 +00:00
|
|
|
limp->pl_rlimit[RLIMIT_RSS].rlim_cur = ptoa(rss_limit);
|
|
|
|
limp->pl_rlimit[RLIMIT_RSS].rlim_max = RLIM_INFINITY;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
1994-05-25 09:21:21 +00:00
|
|
|
void
|
|
|
|
faultin(p)
|
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
|
|
|
struct proc *p;
|
1994-05-25 09:21:21 +00:00
|
|
|
{
|
2003-04-23 18:21:41 +00:00
|
|
|
#ifdef NO_SWAPPING
|
1994-05-25 09:21:21 +00:00
|
|
|
|
2001-05-15 22:20:44 +00:00
|
|
|
PROC_LOCK_ASSERT(p, MA_OWNED);
|
2002-07-07 23:05:27 +00:00
|
|
|
if ((p->p_sflag & PS_INMEM) == 0)
|
|
|
|
panic("faultin: proc swapped out with NO_SWAPPING!");
|
2003-04-23 18:21:41 +00:00
|
|
|
#else /* !NO_SWAPPING */
|
|
|
|
struct thread *td;
|
|
|
|
|
|
|
|
GIANT_REQUIRED;
|
|
|
|
PROC_LOCK_ASSERT(p, MA_OWNED);
|
2003-04-22 20:00:26 +00:00
|
|
|
/*
|
|
|
|
* If another process is swapping in this process,
|
|
|
|
* just wait until it finishes.
|
|
|
|
*/
|
|
|
|
if (p->p_sflag & PS_SWAPPINGIN)
|
|
|
|
msleep(&p->p_sflag, &p->p_mtx, PVM, "faultin", 0);
|
|
|
|
else if ((p->p_sflag & PS_INMEM) == 0) {
|
2002-07-29 18:33:32 +00:00
|
|
|
/*
|
2003-04-22 20:00:26 +00:00
|
|
|
* Don't let another thread swap process p out while we are
|
|
|
|
* busy swapping it in.
|
2002-07-29 18:33:32 +00:00
|
|
|
*/
|
2003-04-22 20:00:26 +00:00
|
|
|
++p->p_lock;
|
|
|
|
mtx_lock_spin(&sched_lock);
|
2002-07-29 18:33:32 +00:00
|
|
|
p->p_sflag |= PS_SWAPPINGIN;
|
Change and clean the mutex lock interface.
mtx_enter(lock, type) becomes:
mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks)
mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized)
similarily, for releasing a lock, we now have:
mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN.
We change the caller interface for the two different types of locks
because the semantics are entirely different for each case, and this
makes it explicitly clear and, at the same time, it rids us of the
extra `type' argument.
The enter->lock and exit->unlock change has been made with the idea
that we're "locking data" and not "entering locked code" in mind.
Further, remove all additional "flags" previously passed to the
lock acquire/release routines with the exception of two:
MTX_QUIET and MTX_NOSWITCH
The functionality of these flags is preserved and they can be passed
to the lock/unlock routines by calling the corresponding wrappers:
mtx_{lock, unlock}_flags(lock, flag(s)) and
mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN
locks, respectively.
Re-inline some lock acq/rel code; in the sleep lock case, we only
inline the _obtain_lock()s in order to ensure that the inlined code
fits into a cache line. In the spin lock case, we inline recursion and
actually only perform a function call if we need to spin. This change
has been made with the idea that we generally tend to avoid spin locks
and that also the spin locks that we do have and are heavily used
(i.e. sched_lock) do recurse, and therefore in an effort to reduce
function call overhead for some architectures (such as alpha), we
inline recursion for this case.
Create a new malloc type for the witness code and retire from using
the M_DEV type. The new type is called M_WITNESS and is only declared
if WITNESS is enabled.
Begin cleaning up some machdep/mutex.h code - specifically updated the
"optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN
and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently
need those.
Finally, caught up to the interface changes in all sys code.
Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
|
|
|
mtx_unlock_spin(&sched_lock);
|
2001-01-25 01:38:09 +00:00
|
|
|
PROC_UNLOCK(p);
|
1994-05-25 09:21:21 +00:00
|
|
|
|
2002-07-07 23:05:27 +00:00
|
|
|
vm_proc_swapin(p);
|
2003-04-22 20:00:26 +00:00
|
|
|
FOREACH_THREAD_IN_PROC(p, td)
|
2003-06-14 23:23:55 +00:00
|
|
|
vm_thread_swapin(td);
|
1996-10-15 03:16:45 +00:00
|
|
|
|
2001-01-25 01:38:09 +00:00
|
|
|
PROC_LOCK(p);
|
Change and clean the mutex lock interface.
mtx_enter(lock, type) becomes:
mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks)
mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized)
similarily, for releasing a lock, we now have:
mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN.
We change the caller interface for the two different types of locks
because the semantics are entirely different for each case, and this
makes it explicitly clear and, at the same time, it rids us of the
extra `type' argument.
The enter->lock and exit->unlock change has been made with the idea
that we're "locking data" and not "entering locked code" in mind.
Further, remove all additional "flags" previously passed to the
lock acquire/release routines with the exception of two:
MTX_QUIET and MTX_NOSWITCH
The functionality of these flags is preserved and they can be passed
to the lock/unlock routines by calling the corresponding wrappers:
mtx_{lock, unlock}_flags(lock, flag(s)) and
mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN
locks, respectively.
Re-inline some lock acq/rel code; in the sleep lock case, we only
inline the _obtain_lock()s in order to ensure that the inlined code
fits into a cache line. In the spin lock case, we inline recursion and
actually only perform a function call if we need to spin. This change
has been made with the idea that we generally tend to avoid spin locks
and that also the spin locks that we do have and are heavily used
(i.e. sched_lock) do recurse, and therefore in an effort to reduce
function call overhead for some architectures (such as alpha), we
inline recursion for this case.
Create a new malloc type for the witness code and retire from using
the M_DEV type. The new type is called M_WITNESS and is only declared
if WITNESS is enabled.
Begin cleaning up some machdep/mutex.h code - specifically updated the
"optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN
and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently
need those.
Finally, caught up to the interface changes in all sys code.
Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
|
|
|
mtx_lock_spin(&sched_lock);
|
2002-07-30 06:54:05 +00:00
|
|
|
p->p_sflag &= ~PS_SWAPPINGIN;
|
|
|
|
p->p_sflag |= PS_INMEM;
|
2003-04-22 20:00:26 +00:00
|
|
|
FOREACH_THREAD_IN_PROC(p, td) {
|
|
|
|
TD_CLR_SWAPPED(td);
|
2002-09-11 08:13:56 +00:00
|
|
|
if (TD_CAN_RUN(td))
|
|
|
|
setrunnable(td);
|
2003-04-22 20:00:26 +00:00
|
|
|
}
|
|
|
|
mtx_unlock_spin(&sched_lock);
|
1994-05-25 09:21:21 +00:00
|
|
|
|
2002-07-29 18:33:32 +00:00
|
|
|
wakeup(&p->p_sflag);
|
1994-05-25 09:21:21 +00:00
|
|
|
|
2003-04-22 20:00:26 +00:00
|
|
|
/* Allow other threads to swap p out now. */
|
1994-05-25 09:21:21 +00:00
|
|
|
--p->p_lock;
|
|
|
|
}
|
2003-04-23 18:21:41 +00:00
|
|
|
#endif /* NO_SWAPPING */
|
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
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
1994-05-25 09:21:21 +00:00
|
|
|
* This swapin algorithm attempts to swap-in processes only if there
|
|
|
|
* is enough space for them. Of course, if a process waits for a long
|
|
|
|
* time, it will be swapped in anyway.
|
2000-09-07 01:33:02 +00:00
|
|
|
*
|
Part 1 of KSE-III
The ability to schedule multiple threads per process
(one one cpu) by making ALL system calls optionally asynchronous.
to come: ia64 and power-pc patches, patches for gdb, test program (in tools)
Reviewed by: Almost everyone who counts
(at various times, peter, jhb, matt, alfred, mini, bernd,
and a cast of thousands)
NOTE: this is still Beta code, and contains lots of debugging stuff.
expect slight instability in signals..
2002-06-29 17:26:22 +00:00
|
|
|
* XXXKSE - process with the thread with highest priority counts..
|
2001-09-12 08:38:13 +00:00
|
|
|
*
|
2000-09-07 01:33:02 +00:00
|
|
|
* Giant is still held at this point, to be released in tsleep.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
1995-08-28 09:19:25 +00:00
|
|
|
/* ARGSUSED*/
|
|
|
|
static void
|
1995-12-02 17:11:20 +00:00
|
|
|
scheduler(dummy)
|
|
|
|
void *dummy;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2001-07-04 19:00:13 +00:00
|
|
|
struct proc *p;
|
Part 1 of KSE-III
The ability to schedule multiple threads per process
(one one cpu) by making ALL system calls optionally asynchronous.
to come: ia64 and power-pc patches, patches for gdb, test program (in tools)
Reviewed by: Almost everyone who counts
(at various times, peter, jhb, matt, alfred, mini, bernd,
and a cast of thousands)
NOTE: this is still Beta code, and contains lots of debugging stuff.
expect slight instability in signals..
2002-06-29 17:26:22 +00:00
|
|
|
struct thread *td;
|
2001-07-04 19:00:13 +00:00
|
|
|
int pri;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct proc *pp;
|
|
|
|
int ppri;
|
|
|
|
|
2001-05-15 22:20:44 +00:00
|
|
|
mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
|
2001-07-04 16:20:28 +00:00
|
|
|
/* GIANT_REQUIRED */
|
2000-09-07 01:33:02 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
loop:
|
1999-09-17 04:56:40 +00:00
|
|
|
if (vm_page_count_min()) {
|
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_WAIT;
|
1999-09-17 04:56:40 +00:00
|
|
|
goto loop;
|
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
|
|
|
pp = NULL;
|
|
|
|
ppri = INT_MIN;
|
2001-03-28 11:52:56 +00:00
|
|
|
sx_slock(&allproc_lock);
|
2001-09-12 08:38:13 +00:00
|
|
|
FOREACH_PROC_IN_SYSTEM(p) {
|
|
|
|
struct ksegrp *kg;
|
2003-04-22 20:00:26 +00:00
|
|
|
if (p->p_sflag & (PS_INMEM | PS_SWAPPINGOUT | PS_SWAPPINGIN)) {
|
Part 1 of KSE-III
The ability to schedule multiple threads per process
(one one cpu) by making ALL system calls optionally asynchronous.
to come: ia64 and power-pc patches, patches for gdb, test program (in tools)
Reviewed by: Almost everyone who counts
(at various times, peter, jhb, matt, alfred, mini, bernd,
and a cast of thousands)
NOTE: this is still Beta code, and contains lots of debugging stuff.
expect slight instability in signals..
2002-06-29 17:26:22 +00:00
|
|
|
continue;
|
|
|
|
}
|
Change and clean the mutex lock interface.
mtx_enter(lock, type) becomes:
mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks)
mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized)
similarily, for releasing a lock, we now have:
mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN.
We change the caller interface for the two different types of locks
because the semantics are entirely different for each case, and this
makes it explicitly clear and, at the same time, it rids us of the
extra `type' argument.
The enter->lock and exit->unlock change has been made with the idea
that we're "locking data" and not "entering locked code" in mind.
Further, remove all additional "flags" previously passed to the
lock acquire/release routines with the exception of two:
MTX_QUIET and MTX_NOSWITCH
The functionality of these flags is preserved and they can be passed
to the lock/unlock routines by calling the corresponding wrappers:
mtx_{lock, unlock}_flags(lock, flag(s)) and
mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN
locks, respectively.
Re-inline some lock acq/rel code; in the sleep lock case, we only
inline the _obtain_lock()s in order to ensure that the inlined code
fits into a cache line. In the spin lock case, we inline recursion and
actually only perform a function call if we need to spin. This change
has been made with the idea that we generally tend to avoid spin locks
and that also the spin locks that we do have and are heavily used
(i.e. sched_lock) do recurse, and therefore in an effort to reduce
function call overhead for some architectures (such as alpha), we
inline recursion for this case.
Create a new malloc type for the witness code and retire from using
the M_DEV type. The new type is called M_WITNESS and is only declared
if WITNESS is enabled.
Begin cleaning up some machdep/mutex.h code - specifically updated the
"optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN
and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently
need those.
Finally, caught up to the interface changes in all sys code.
Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
|
|
|
mtx_lock_spin(&sched_lock);
|
Part 1 of KSE-III
The ability to schedule multiple threads per process
(one one cpu) by making ALL system calls optionally asynchronous.
to come: ia64 and power-pc patches, patches for gdb, test program (in tools)
Reviewed by: Almost everyone who counts
(at various times, peter, jhb, matt, alfred, mini, bernd,
and a cast of thousands)
NOTE: this is still Beta code, and contains lots of debugging stuff.
expect slight instability in signals..
2002-06-29 17:26:22 +00:00
|
|
|
FOREACH_THREAD_IN_PROC(p, td) {
|
2002-07-29 18:33:32 +00:00
|
|
|
/*
|
2002-09-11 08:13:56 +00:00
|
|
|
* An otherwise runnable thread of a process
|
|
|
|
* swapped out has only the TDI_SWAPPED bit set.
|
|
|
|
*
|
2002-07-29 18:33:32 +00:00
|
|
|
*/
|
2002-09-11 08:13:56 +00:00
|
|
|
if (td->td_inhibitors == TDI_SWAPPED) {
|
Part 1 of KSE-III
The ability to schedule multiple threads per process
(one one cpu) by making ALL system calls optionally asynchronous.
to come: ia64 and power-pc patches, patches for gdb, test program (in tools)
Reviewed by: Almost everyone who counts
(at various times, peter, jhb, matt, alfred, mini, bernd,
and a cast of thousands)
NOTE: this is still Beta code, and contains lots of debugging stuff.
expect slight instability in signals..
2002-06-29 17:26:22 +00:00
|
|
|
kg = td->td_ksegrp;
|
2001-09-12 08:38:13 +00:00
|
|
|
pri = p->p_swtime + kg->kg_slptime;
|
|
|
|
if ((p->p_sflag & PS_SWAPINREQ) == 0) {
|
2004-06-16 00:26:31 +00:00
|
|
|
pri -= p->p_nice * 8;
|
2001-09-12 08:38:13 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* if this ksegrp is higher priority
|
|
|
|
* and there is enough space, then select
|
|
|
|
* this process instead of the previous
|
|
|
|
* selection.
|
|
|
|
*/
|
|
|
|
if (pri > ppri) {
|
|
|
|
pp = p;
|
|
|
|
ppri = pri;
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
}
|
Change and clean the mutex lock interface.
mtx_enter(lock, type) becomes:
mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks)
mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized)
similarily, for releasing a lock, we now have:
mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN.
We change the caller interface for the two different types of locks
because the semantics are entirely different for each case, and this
makes it explicitly clear and, at the same time, it rids us of the
extra `type' argument.
The enter->lock and exit->unlock change has been made with the idea
that we're "locking data" and not "entering locked code" in mind.
Further, remove all additional "flags" previously passed to the
lock acquire/release routines with the exception of two:
MTX_QUIET and MTX_NOSWITCH
The functionality of these flags is preserved and they can be passed
to the lock/unlock routines by calling the corresponding wrappers:
mtx_{lock, unlock}_flags(lock, flag(s)) and
mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN
locks, respectively.
Re-inline some lock acq/rel code; in the sleep lock case, we only
inline the _obtain_lock()s in order to ensure that the inlined code
fits into a cache line. In the spin lock case, we inline recursion and
actually only perform a function call if we need to spin. This change
has been made with the idea that we generally tend to avoid spin locks
and that also the spin locks that we do have and are heavily used
(i.e. sched_lock) do recurse, and therefore in an effort to reduce
function call overhead for some architectures (such as alpha), we
inline recursion for this case.
Create a new malloc type for the witness code and retire from using
the M_DEV type. The new type is called M_WITNESS and is only declared
if WITNESS is enabled.
Begin cleaning up some machdep/mutex.h code - specifically updated the
"optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN
and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently
need those.
Finally, caught up to the interface changes in all sys code.
Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
|
|
|
mtx_unlock_spin(&sched_lock);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2001-03-28 11:52:56 +00:00
|
|
|
sx_sunlock(&allproc_lock);
|
1994-05-25 09:21:21 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
1996-10-17 02:58:20 +00:00
|
|
|
* Nothing to do, back to sleep.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
if ((p = pp) == NULL) {
|
2001-05-18 00:08:38 +00:00
|
|
|
tsleep(&proc0, PVM, "sched", maxslp * hz / 2);
|
1994-05-24 10:09:53 +00:00
|
|
|
goto loop;
|
|
|
|
}
|
2002-07-29 18:33:32 +00:00
|
|
|
PROC_LOCK(p);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Another process may be bringing or may have already
|
|
|
|
* brought this process in while we traverse all threads.
|
|
|
|
* Or, this process may even be being swapped out again.
|
|
|
|
*/
|
2003-04-22 20:00:26 +00:00
|
|
|
if (p->p_sflag & (PS_INMEM | PS_SWAPPINGOUT | PS_SWAPPINGIN)) {
|
2002-07-29 18:33:32 +00:00
|
|
|
PROC_UNLOCK(p);
|
|
|
|
goto loop;
|
|
|
|
}
|
|
|
|
|
2003-04-22 20:00:26 +00:00
|
|
|
mtx_lock_spin(&sched_lock);
|
2001-01-24 11:25:56 +00:00
|
|
|
p->p_sflag &= ~PS_SWAPINREQ;
|
2003-04-22 20:00:26 +00:00
|
|
|
mtx_unlock_spin(&sched_lock);
|
1996-10-17 02:58:20 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
1994-05-25 09:21:21 +00:00
|
|
|
* We would like to bring someone in. (only if there is space).
|
Part 1 of KSE-III
The ability to schedule multiple threads per process
(one one cpu) by making ALL system calls optionally asynchronous.
to come: ia64 and power-pc patches, patches for gdb, test program (in tools)
Reviewed by: Almost everyone who counts
(at various times, peter, jhb, matt, alfred, mini, bernd,
and a cast of thousands)
NOTE: this is still Beta code, and contains lots of debugging stuff.
expect slight instability in signals..
2002-06-29 17:26:22 +00:00
|
|
|
* [What checks the space? ]
|
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
|
|
|
faultin(p);
|
2001-01-25 01:38:09 +00:00
|
|
|
PROC_UNLOCK(p);
|
2003-04-22 20:00:26 +00:00
|
|
|
mtx_lock_spin(&sched_lock);
|
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
|
|
|
p->p_swtime = 0;
|
Change and clean the mutex lock interface.
mtx_enter(lock, type) becomes:
mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks)
mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized)
similarily, for releasing a lock, we now have:
mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN.
We change the caller interface for the two different types of locks
because the semantics are entirely different for each case, and this
makes it explicitly clear and, at the same time, it rids us of the
extra `type' argument.
The enter->lock and exit->unlock change has been made with the idea
that we're "locking data" and not "entering locked code" in mind.
Further, remove all additional "flags" previously passed to the
lock acquire/release routines with the exception of two:
MTX_QUIET and MTX_NOSWITCH
The functionality of these flags is preserved and they can be passed
to the lock/unlock routines by calling the corresponding wrappers:
mtx_{lock, unlock}_flags(lock, flag(s)) and
mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN
locks, respectively.
Re-inline some lock acq/rel code; in the sleep lock case, we only
inline the _obtain_lock()s in order to ensure that the inlined code
fits into a cache line. In the spin lock case, we inline recursion and
actually only perform a function call if we need to spin. This change
has been made with the idea that we generally tend to avoid spin locks
and that also the spin locks that we do have and are heavily used
(i.e. sched_lock) do recurse, and therefore in an effort to reduce
function call overhead for some architectures (such as alpha), we
inline recursion for this case.
Create a new malloc type for the witness code and retire from using
the M_DEV type. The new type is called M_WITNESS and is only declared
if WITNESS is enabled.
Begin cleaning up some machdep/mutex.h code - specifically updated the
"optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN
and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently
need those.
Finally, caught up to the interface changes in all sys code.
Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
|
|
|
mtx_unlock_spin(&sched_lock);
|
1994-05-24 10:09:53 +00:00
|
|
|
goto loop;
|
|
|
|
}
|
|
|
|
|
1996-02-22 10:57:37 +00:00
|
|
|
#ifndef NO_SWAPPING
|
|
|
|
|
1997-12-06 02:23:36 +00:00
|
|
|
/*
|
|
|
|
* Swap_idle_threshold1 is the guaranteed swapped in time for a process
|
|
|
|
*/
|
1998-02-09 06:11:36 +00:00
|
|
|
static int swap_idle_threshold1 = 2;
|
2003-04-17 15:39:12 +00:00
|
|
|
SYSCTL_INT(_vm, OID_AUTO, swap_idle_threshold1, CTLFLAG_RW,
|
2003-04-17 15:44:22 +00:00
|
|
|
&swap_idle_threshold1, 0, "Guaranteed swapped in time for a process");
|
1997-12-06 02:23:36 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Swap_idle_threshold2 is the time that a process can be idle before
|
|
|
|
* it will be swapped out, if idle swapping is enabled.
|
|
|
|
*/
|
1998-02-09 06:11:36 +00:00
|
|
|
static int swap_idle_threshold2 = 10;
|
2003-04-17 15:39:12 +00:00
|
|
|
SYSCTL_INT(_vm, OID_AUTO, swap_idle_threshold2, CTLFLAG_RW,
|
2003-04-17 15:44:22 +00:00
|
|
|
&swap_idle_threshold2, 0, "Time before a process will be swapped out");
|
1997-12-06 02:23:36 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Swapout is driven by the pageout daemon. Very simple, we find eligible
|
|
|
|
* procs and unwire their u-areas. We try to always "swap" at least one
|
|
|
|
* process in case we need the room for a swapin.
|
|
|
|
* If any procs have been sleeping/stopped for at least maxslp seconds,
|
|
|
|
* they are swapped. Else, we swap the longest-sleeping or stopped process,
|
|
|
|
* if any, otherwise the longest-resident process.
|
|
|
|
*/
|
|
|
|
void
|
1997-12-11 02:10:55 +00:00
|
|
|
swapout_procs(action)
|
|
|
|
int action;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2001-07-04 19:00:13 +00:00
|
|
|
struct proc *p;
|
Part 1 of KSE-III
The ability to schedule multiple threads per process
(one one cpu) by making ALL system calls optionally asynchronous.
to come: ia64 and power-pc patches, patches for gdb, test program (in tools)
Reviewed by: Almost everyone who counts
(at various times, peter, jhb, matt, alfred, mini, bernd,
and a cast of thousands)
NOTE: this is still Beta code, and contains lots of debugging stuff.
expect slight instability in signals..
2002-06-29 17:26:22 +00:00
|
|
|
struct thread *td;
|
2001-09-12 08:38:13 +00:00
|
|
|
struct ksegrp *kg;
|
1994-05-24 10:09:53 +00:00
|
|
|
int didswap = 0;
|
|
|
|
|
2001-07-04 16:20:28 +00:00
|
|
|
GIANT_REQUIRED;
|
|
|
|
|
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
|
|
|
retry:
|
2001-05-23 22:35:45 +00:00
|
|
|
sx_slock(&allproc_lock);
|
Part 1 of KSE-III
The ability to schedule multiple threads per process
(one one cpu) by making ALL system calls optionally asynchronous.
to come: ia64 and power-pc patches, patches for gdb, test program (in tools)
Reviewed by: Almost everyone who counts
(at various times, peter, jhb, matt, alfred, mini, bernd,
and a cast of thousands)
NOTE: this is still Beta code, and contains lots of debugging stuff.
expect slight instability in signals..
2002-06-29 17:26:22 +00:00
|
|
|
FOREACH_PROC_IN_SYSTEM(p) {
|
1996-05-18 03:38:05 +00:00
|
|
|
struct vmspace *vm;
|
2001-09-12 08:38:13 +00:00
|
|
|
int minslptime = 100000;
|
2001-01-24 11:25:56 +00:00
|
|
|
|
2002-07-30 06:54:05 +00:00
|
|
|
/*
|
2002-09-09 09:05:06 +00:00
|
|
|
* Watch out for a process in
|
|
|
|
* creation. It may have no
|
2002-10-22 14:31:32 +00:00
|
|
|
* address space or lock yet.
|
|
|
|
*/
|
|
|
|
mtx_lock_spin(&sched_lock);
|
|
|
|
if (p->p_state == PRS_NEW) {
|
|
|
|
mtx_unlock_spin(&sched_lock);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
mtx_unlock_spin(&sched_lock);
|
|
|
|
|
|
|
|
/*
|
2002-09-09 09:05:06 +00:00
|
|
|
* An aio daemon switches its
|
|
|
|
* address space while running.
|
|
|
|
* Perform a quick check whether
|
|
|
|
* a process has P_SYSTEM.
|
2002-07-30 06:54:05 +00:00
|
|
|
*/
|
2003-04-25 20:06:30 +00:00
|
|
|
if ((p->p_flag & P_SYSTEM) != 0)
|
2002-09-09 09:05:06 +00:00
|
|
|
continue;
|
2002-10-22 14:31:32 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Do not swapout a process that
|
|
|
|
* is waiting for VM data
|
|
|
|
* structures as there is a possible
|
|
|
|
* deadlock. Test this first as
|
|
|
|
* this may block.
|
|
|
|
*
|
|
|
|
* Lock the map until swapout
|
|
|
|
* finishes, or a thread of this
|
|
|
|
* process may attempt to alter
|
|
|
|
* the map.
|
|
|
|
*/
|
2003-04-25 20:06:30 +00:00
|
|
|
PROC_LOCK(p);
|
2002-07-30 06:54:05 +00:00
|
|
|
vm = p->p_vmspace;
|
2002-09-09 09:05:06 +00:00
|
|
|
KASSERT(vm != NULL,
|
|
|
|
("swapout_procs: a process has no address space"));
|
2004-07-27 03:53:41 +00:00
|
|
|
atomic_add_int(&vm->vm_refcnt, 1);
|
2002-09-09 09:05:06 +00:00
|
|
|
PROC_UNLOCK(p);
|
2002-07-30 06:54:05 +00:00
|
|
|
if (!vm_map_trylock(&vm->vm_map))
|
|
|
|
goto nextproc1;
|
|
|
|
|
2001-01-24 11:25:56 +00:00
|
|
|
PROC_LOCK(p);
|
2001-01-24 12:23:17 +00:00
|
|
|
if (p->p_lock != 0 ||
|
2002-09-05 07:30:18 +00:00
|
|
|
(p->p_flag & (P_STOPPED_SINGLE|P_TRACED|P_SYSTEM|P_WEXIT)
|
|
|
|
) != 0) {
|
2002-07-30 06:54:05 +00:00
|
|
|
goto nextproc2;
|
2001-01-24 11:25:56 +00:00
|
|
|
}
|
2001-05-19 01:28:09 +00:00
|
|
|
/*
|
|
|
|
* only aiod changes vmspace, however it will be
|
|
|
|
* skipped because of the if statement above checking
|
|
|
|
* for P_SYSTEM
|
|
|
|
*/
|
2003-04-22 20:00:26 +00:00
|
|
|
if ((p->p_sflag & (PS_INMEM|PS_SWAPPINGOUT|PS_SWAPPINGIN)) != PS_INMEM)
|
|
|
|
goto nextproc2;
|
2001-01-24 12:23:17 +00:00
|
|
|
|
Part 1 of KSE-III
The ability to schedule multiple threads per process
(one one cpu) by making ALL system calls optionally asynchronous.
to come: ia64 and power-pc patches, patches for gdb, test program (in tools)
Reviewed by: Almost everyone who counts
(at various times, peter, jhb, matt, alfred, mini, bernd,
and a cast of thousands)
NOTE: this is still Beta code, and contains lots of debugging stuff.
expect slight instability in signals..
2002-06-29 17:26:22 +00:00
|
|
|
switch (p->p_state) {
|
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
|
|
|
default:
|
Part 1 of KSE-III
The ability to schedule multiple threads per process
(one one cpu) by making ALL system calls optionally asynchronous.
to come: ia64 and power-pc patches, patches for gdb, test program (in tools)
Reviewed by: Almost everyone who counts
(at various times, peter, jhb, matt, alfred, mini, bernd,
and a cast of thousands)
NOTE: this is still Beta code, and contains lots of debugging stuff.
expect slight instability in signals..
2002-06-29 17:26:22 +00:00
|
|
|
/* Don't swap out processes in any sort
|
|
|
|
* of 'special' state. */
|
2003-04-25 20:06:30 +00:00
|
|
|
break;
|
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
|
|
|
|
Part 1 of KSE-III
The ability to schedule multiple threads per process
(one one cpu) by making ALL system calls optionally asynchronous.
to come: ia64 and power-pc patches, patches for gdb, test program (in tools)
Reviewed by: Almost everyone who counts
(at various times, peter, jhb, matt, alfred, mini, bernd,
and a cast of thousands)
NOTE: this is still Beta code, and contains lots of debugging stuff.
expect slight instability in signals..
2002-06-29 17:26:22 +00:00
|
|
|
case PRS_NORMAL:
|
2003-04-25 20:06:30 +00:00
|
|
|
mtx_lock_spin(&sched_lock);
|
1994-11-13 12:47:07 +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
|
|
|
* do not swapout a realtime process
|
2001-09-12 08:38:13 +00:00
|
|
|
* Check all the thread groups..
|
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
|
|
|
*/
|
2001-09-12 08:38:13 +00:00
|
|
|
FOREACH_KSEGRP_IN_PROC(p, kg) {
|
2002-07-30 06:54:05 +00:00
|
|
|
if (PRI_IS_REALTIME(kg->kg_pri_class))
|
2001-09-12 08:38:13 +00:00
|
|
|
goto nextproc;
|
|
|
|
|
|
|
|
/*
|
2002-07-30 06:54:05 +00:00
|
|
|
* Guarantee swap_idle_threshold1
|
2001-09-12 08:38:13 +00:00
|
|
|
* time in memory.
|
|
|
|
*/
|
2002-07-30 06:54:05 +00:00
|
|
|
if (kg->kg_slptime < swap_idle_threshold1)
|
2001-09-12 08:38:13 +00:00
|
|
|
goto nextproc;
|
2002-07-30 06:54:05 +00:00
|
|
|
|
2002-07-29 18:33:32 +00:00
|
|
|
/*
|
2002-07-30 06:54:05 +00:00
|
|
|
* Do not swapout a process if it is
|
|
|
|
* waiting on a critical event of some
|
|
|
|
* kind or there is a thread whose
|
|
|
|
* pageable memory may be accessed.
|
2002-07-29 18:33:32 +00:00
|
|
|
*
|
|
|
|
* This could be refined to support
|
|
|
|
* swapping out a thread.
|
|
|
|
*/
|
2002-07-30 06:54:05 +00:00
|
|
|
FOREACH_THREAD_IN_GROUP(kg, td) {
|
2002-07-29 18:33:32 +00:00
|
|
|
if ((td->td_priority) < PSOCK ||
|
2002-07-30 06:54:05 +00:00
|
|
|
!thread_safetoswapout(td))
|
Part 1 of KSE-III
The ability to schedule multiple threads per process
(one one cpu) by making ALL system calls optionally asynchronous.
to come: ia64 and power-pc patches, patches for gdb, test program (in tools)
Reviewed by: Almost everyone who counts
(at various times, peter, jhb, matt, alfred, mini, bernd,
and a cast of thousands)
NOTE: this is still Beta code, and contains lots of debugging stuff.
expect slight instability in signals..
2002-06-29 17:26:22 +00:00
|
|
|
goto nextproc;
|
|
|
|
}
|
2001-09-12 08:38:13 +00:00
|
|
|
/*
|
|
|
|
* If the system is under memory stress,
|
|
|
|
* or if we are swapping
|
|
|
|
* idle processes >= swap_idle_threshold2,
|
|
|
|
* then swap the process out.
|
|
|
|
*/
|
|
|
|
if (((action & VM_SWAP_NORMAL) == 0) &&
|
|
|
|
(((action & VM_SWAP_IDLE) == 0) ||
|
2002-07-30 06:54:05 +00:00
|
|
|
(kg->kg_slptime < swap_idle_threshold2)))
|
2001-09-12 08:38:13 +00:00
|
|
|
goto nextproc;
|
2002-07-30 06:54:05 +00:00
|
|
|
|
2001-09-12 08:38:13 +00:00
|
|
|
if (minslptime > kg->kg_slptime)
|
|
|
|
minslptime = kg->kg_slptime;
|
2000-12-02 03:29:33 +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
|
|
|
|
1994-09-12 15:06:14 +00:00
|
|
|
/*
|
2004-06-27 01:58:12 +00:00
|
|
|
* If the pageout daemon didn't free enough pages,
|
|
|
|
* or if this process is idle and the system is
|
|
|
|
* configured to swap proactively, swap it out.
|
1994-09-12 15:06:14 +00:00
|
|
|
*/
|
1997-12-06 02:23:36 +00:00
|
|
|
if ((action & VM_SWAP_NORMAL) ||
|
|
|
|
((action & VM_SWAP_IDLE) &&
|
2001-09-12 08:38:13 +00:00
|
|
|
(minslptime > swap_idle_threshold2))) {
|
1997-12-06 02:23:36 +00:00
|
|
|
swapout(p);
|
|
|
|
didswap++;
|
2002-07-30 06:54:05 +00:00
|
|
|
mtx_unlock_spin(&sched_lock);
|
2003-04-22 20:00:26 +00:00
|
|
|
PROC_UNLOCK(p);
|
2002-07-30 06:54:05 +00:00
|
|
|
vm_map_unlock(&vm->vm_map);
|
|
|
|
vmspace_free(vm);
|
|
|
|
sx_sunlock(&allproc_lock);
|
1997-12-06 02:23:36 +00:00
|
|
|
goto retry;
|
2001-05-15 22:20:44 +00:00
|
|
|
}
|
2003-04-25 20:06:30 +00:00
|
|
|
nextproc:
|
|
|
|
mtx_unlock_spin(&sched_lock);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2002-07-30 06:54:05 +00:00
|
|
|
nextproc2:
|
|
|
|
PROC_UNLOCK(p);
|
|
|
|
vm_map_unlock(&vm->vm_map);
|
|
|
|
nextproc1:
|
|
|
|
vmspace_free(vm);
|
2002-03-19 11:02:06 +00:00
|
|
|
continue;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2001-03-28 11:52:56 +00:00
|
|
|
sx_sunlock(&allproc_lock);
|
1994-05-25 09:21:21 +00:00
|
|
|
/*
|
|
|
|
* If we swapped something out, and another process needed memory,
|
|
|
|
* then wakeup the sched process.
|
|
|
|
*/
|
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 (didswap)
|
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
|
|
|
wakeup(&proc0);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
1995-12-14 09:55:16 +00:00
|
|
|
static void
|
1994-05-24 10:09:53 +00:00
|
|
|
swapout(p)
|
2001-07-04 19:00:13 +00:00
|
|
|
struct proc *p;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2001-09-12 08:38:13 +00:00
|
|
|
struct thread *td;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2001-05-18 00:08:38 +00:00
|
|
|
PROC_LOCK_ASSERT(p, MA_OWNED);
|
2002-07-30 06:54:05 +00:00
|
|
|
mtx_assert(&sched_lock, MA_OWNED | MA_NOTRECURSED);
|
1996-04-09 04:36:58 +00:00
|
|
|
#if defined(SWAP_DEBUG)
|
|
|
|
printf("swapping out %d\n", p->p_pid);
|
|
|
|
#endif
|
2002-07-29 18:33:32 +00:00
|
|
|
|
2002-07-30 06:54:05 +00:00
|
|
|
/*
|
|
|
|
* The states of this process and its threads may have changed
|
|
|
|
* by now. Assuming that there is only one pageout daemon thread,
|
|
|
|
* this process should still be in memory.
|
|
|
|
*/
|
2003-04-22 20:00:26 +00:00
|
|
|
KASSERT((p->p_sflag & (PS_INMEM|PS_SWAPPINGOUT|PS_SWAPPINGIN)) == PS_INMEM,
|
2002-07-30 06:54:05 +00:00
|
|
|
("swapout: lost a swapout race?"));
|
|
|
|
|
|
|
|
#if defined(INVARIANTS)
|
2002-07-29 18:33:32 +00:00
|
|
|
/*
|
|
|
|
* Make sure that all threads are safe to be swapped out.
|
|
|
|
*
|
|
|
|
* Alternatively, we could swap out only safe threads.
|
|
|
|
*/
|
|
|
|
FOREACH_THREAD_IN_PROC(p, td) {
|
2002-07-30 06:54:05 +00:00
|
|
|
KASSERT(thread_safetoswapout(td),
|
|
|
|
("swapout: there is a thread not safe for swapout"));
|
2002-07-29 18:33:32 +00:00
|
|
|
}
|
2002-07-30 06:54:05 +00:00
|
|
|
#endif /* INVARIANTS */
|
2002-07-29 18:33:32 +00:00
|
|
|
|
1994-05-25 09:21:21 +00:00
|
|
|
++p->p_stats->p_ru.ru_nswap;
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
1994-05-25 09:21:21 +00:00
|
|
|
* remember the process resident count
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
1999-02-19 14:25:37 +00:00
|
|
|
p->p_vmspace->vm_swrss = vmspace_resident_count(p->p_vmspace);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2002-07-30 06:54:05 +00:00
|
|
|
p->p_sflag &= ~PS_INMEM;
|
2003-04-22 20:00:26 +00:00
|
|
|
p->p_sflag |= PS_SWAPPINGOUT;
|
|
|
|
PROC_UNLOCK(p);
|
|
|
|
FOREACH_THREAD_IN_PROC(p, td)
|
|
|
|
TD_SET_SWAPPED(td);
|
Change and clean the mutex lock interface.
mtx_enter(lock, type) becomes:
mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks)
mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized)
similarily, for releasing a lock, we now have:
mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN.
We change the caller interface for the two different types of locks
because the semantics are entirely different for each case, and this
makes it explicitly clear and, at the same time, it rids us of the
extra `type' argument.
The enter->lock and exit->unlock change has been made with the idea
that we're "locking data" and not "entering locked code" in mind.
Further, remove all additional "flags" previously passed to the
lock acquire/release routines with the exception of two:
MTX_QUIET and MTX_NOSWITCH
The functionality of these flags is preserved and they can be passed
to the lock/unlock routines by calling the corresponding wrappers:
mtx_{lock, unlock}_flags(lock, flag(s)) and
mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN
locks, respectively.
Re-inline some lock acq/rel code; in the sleep lock case, we only
inline the _obtain_lock()s in order to ensure that the inlined code
fits into a cache line. In the spin lock case, we inline recursion and
actually only perform a function call if we need to spin. This change
has been made with the idea that we generally tend to avoid spin locks
and that also the spin locks that we do have and are heavily used
(i.e. sched_lock) do recurse, and therefore in an effort to reduce
function call overhead for some architectures (such as alpha), we
inline recursion for this case.
Create a new malloc type for the witness code and retire from using
the M_DEV type. The new type is called M_WITNESS and is only declared
if WITNESS is enabled.
Begin cleaning up some machdep/mutex.h code - specifically updated the
"optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN
and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently
need those.
Finally, caught up to the interface changes in all sys code.
Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
|
|
|
mtx_unlock_spin(&sched_lock);
|
1994-05-25 09:21:21 +00:00
|
|
|
|
2002-07-07 23:05:27 +00:00
|
|
|
vm_proc_swapout(p);
|
2003-04-22 20:00:26 +00:00
|
|
|
FOREACH_THREAD_IN_PROC(p, td)
|
2003-06-14 23:23:55 +00:00
|
|
|
vm_thread_swapout(td);
|
2003-04-22 20:00:26 +00:00
|
|
|
|
|
|
|
PROC_LOCK(p);
|
Change and clean the mutex lock interface.
mtx_enter(lock, type) becomes:
mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks)
mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized)
similarily, for releasing a lock, we now have:
mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN.
We change the caller interface for the two different types of locks
because the semantics are entirely different for each case, and this
makes it explicitly clear and, at the same time, it rids us of the
extra `type' argument.
The enter->lock and exit->unlock change has been made with the idea
that we're "locking data" and not "entering locked code" in mind.
Further, remove all additional "flags" previously passed to the
lock acquire/release routines with the exception of two:
MTX_QUIET and MTX_NOSWITCH
The functionality of these flags is preserved and they can be passed
to the lock/unlock routines by calling the corresponding wrappers:
mtx_{lock, unlock}_flags(lock, flag(s)) and
mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN
locks, respectively.
Re-inline some lock acq/rel code; in the sleep lock case, we only
inline the _obtain_lock()s in order to ensure that the inlined code
fits into a cache line. In the spin lock case, we inline recursion and
actually only perform a function call if we need to spin. This change
has been made with the idea that we generally tend to avoid spin locks
and that also the spin locks that we do have and are heavily used
(i.e. sched_lock) do recurse, and therefore in an effort to reduce
function call overhead for some architectures (such as alpha), we
inline recursion for this case.
Create a new malloc type for the witness code and retire from using
the M_DEV type. The new type is called M_WITNESS and is only declared
if WITNESS is enabled.
Begin cleaning up some machdep/mutex.h code - specifically updated the
"optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN
and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently
need those.
Finally, caught up to the interface changes in all sys code.
Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
|
|
|
mtx_lock_spin(&sched_lock);
|
2003-04-22 20:00:26 +00:00
|
|
|
p->p_sflag &= ~PS_SWAPPINGOUT;
|
1994-05-24 10:09:53 +00:00
|
|
|
p->p_swtime = 0;
|
|
|
|
}
|
1996-02-22 10:57:37 +00:00
|
|
|
#endif /* !NO_SWAPPING */
|