1
0
mirror of https://git.FreeBSD.org/src.git synced 2024-12-21 11:13:30 +00:00
freebsd/sys/vm/vm_map.c

3166 lines
82 KiB
C
Raw Normal View History

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_map.c 8.3 (Berkeley) 1/12/94
1994-05-24 10:09:53 +00:00
*
*
* Copyright (c) 1987, 1990 Carnegie-Mellon University.
* All rights reserved.
*
* Authors: Avadis Tevanian, Jr., Michael Wayne Young
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
*
1994-05-24 10:09:53 +00:00
* Permission to use, copy, modify and distribute this software and
* its documentation is hereby granted, provided that both the copyright
* notice and this permission notice appear in all copies of the
* software, derivative works or modified versions, and any portions
* thereof, and that both notices appear in supporting documentation.
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
*
* CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
* CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
1994-05-24 10:09:53 +00:00
* FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
*
1994-05-24 10:09:53 +00:00
* Carnegie Mellon requests users of this software to return to
*
* Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU
* School of Computer Science
* Carnegie Mellon University
* Pittsburgh PA 15213-3890
*
* any improvements or extensions that they make and grant Carnegie the
* rights to redistribute these changes.
*/
/*
* Virtual memory mapping module.
*/
2003-06-11 23:50:51 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
2001-10-11 17:53:43 +00:00
#include <sys/ktr.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/vmmeter.h>
#include <sys/mman.h>
#include <sys/vnode.h>
#include <sys/resourcevar.h>
#include <sys/file.h>
#include <sys/sysent.h>
#include <sys/shm.h>
1994-05-24 10:09:53 +00:00
#include <vm/vm.h>
#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>
#include <vm/vm_object.h>
#include <vm/vm_pager.h>
#include <vm/vm_kern.h>
#include <vm/vm_extern.h>
#include <vm/swap_pager.h>
#include <vm/uma.h>
1994-05-24 10:09:53 +00:00
/*
* Virtual memory maps provide for the mapping, protection,
* and sharing of virtual memory objects. In addition,
* this module provides for an efficient virtual copy of
* memory from one map to another.
*
* Synchronization is required prior to most operations.
*
* Maps consist of an ordered doubly-linked list of simple
* entries; a single hint is used to speed up lookups.
*
2000-03-26 15:20:23 +00:00
* Since portions of maps are specified by start/end addresses,
1994-05-24 10:09:53 +00:00
* which may not align with existing map entries, all
* routines merely "clip" entries to these start/end values.
* [That is, an entry is split into two, bordering at a
* start or end value.] Note that these clippings may not
* always be necessary (as the two resulting entries are then
* not changed); however, the clipping is done for convenience.
*
* As mentioned above, virtual copy operations are performed
* by copying VM object references from one map to
1994-05-24 10:09:53 +00:00
* another, and then marking both regions as copy-on-write.
*/
/*
* vm_map_startup:
*
* Initialize the vm_map module. Must be called before
* any other vm_map routines.
*
* Map and entry structures are allocated from the general
* purpose memory pool with some exceptions:
*
* - The kernel map and kmem submap are allocated statically.
* - Kernel map entries are allocated out of a static pool.
*
* These restrictions are necessary since malloc() uses the
* maps and requires map entries.
*/
static struct mtx map_sleep_mtx;
static uma_zone_t mapentzone;
static uma_zone_t kmapentzone;
static uma_zone_t mapzone;
static uma_zone_t vmspace_zone;
static struct vm_object kmapentobj;
static int vmspace_zinit(void *mem, int size, int flags);
static void vmspace_zfini(void *mem, int size);
static int vm_map_zinit(void *mem, int ize, int flags);
static void vm_map_zfini(void *mem, int size);
static void _vm_map_init(vm_map_t map, vm_offset_t min, vm_offset_t max);
#ifdef INVARIANTS
static void vm_map_zdtor(void *mem, int size, void *arg);
static void vmspace_zdtor(void *mem, int size, void *arg);
#endif
This set of commits to the VM system does the following, and contain contributions or ideas from Stephen McKay <syssgm@devetir.qld.gov.au>, Alan Cox <alc@cs.rice.edu>, David Greenman <davidg@freebsd.org> and me: More usage of the TAILQ macros. Additional minor fix to queue.h. Performance enhancements to the pageout daemon. Addition of a wait in the case that the pageout daemon has to run immediately. Slightly modify the pageout algorithm. Significant revamp of the pmap/fork code: 1) PTE's and UPAGES's are NO LONGER in the process's map. 2) PTE's and UPAGES's reside in their own objects. 3) TOTAL elimination of recursive page table pagefaults. 4) The page directory now resides in the PTE object. 5) Implemented pmap_copy, thereby speeding up fork time. 6) Changed the pv entries so that the head is a pointer and not an entire entry. 7) Significant cleanup of pmap_protect, and pmap_remove. 8) Removed significant amounts of machine dependent fork code from vm_glue. Pushed much of that code into the machine dependent pmap module. 9) Support more completely the reuse of already zeroed pages (Page table pages and page directories) as being already zeroed. Performance and code cleanups in vm_map: 1) Improved and simplified allocation of map entries. 2) Improved vm_map_copy code. 3) Corrected some minor problems in the simplify code. Implemented splvm (combo of splbio and splimp.) The VM code now seldom uses splhigh. Improved the speed of and simplified kmem_malloc. Minor mod to vm_fault to avoid using pre-zeroed pages in the case of objects with backing objects along with the already existant condition of having a vnode. (If there is a backing object, there will likely be a COW... With a COW, it isn't necessary to start with a pre-zeroed page.) Minor reorg of source to perhaps improve locality of ref.
1996-05-18 03:38:05 +00:00
1995-05-30 08:16:23 +00:00
void
vm_map_startup(void)
1994-05-24 10:09:53 +00:00
{
mtx_init(&map_sleep_mtx, "vm map sleep mutex", NULL, MTX_DEF);
mapzone = uma_zcreate("MAP", sizeof(struct vm_map), NULL,
#ifdef INVARIANTS
vm_map_zdtor,
#else
NULL,
#endif
vm_map_zinit, vm_map_zfini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
uma_prealloc(mapzone, MAX_KMAP);
2003-11-03 16:14:45 +00:00
kmapentzone = uma_zcreate("KMAP ENTRY", sizeof(struct vm_map_entry),
NULL, NULL, NULL, NULL, UMA_ALIGN_PTR,
UMA_ZONE_MTXCLASS | UMA_ZONE_VM);
uma_prealloc(kmapentzone, MAX_KMAPENT);
2003-11-03 16:14:45 +00:00
mapentzone = uma_zcreate("MAP ENTRY", sizeof(struct vm_map_entry),
NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
uma_prealloc(mapentzone, MAX_MAPENT);
}
static void
vmspace_zfini(void *mem, int size)
{
struct vmspace *vm;
vm = (struct vmspace *)mem;
pmap_release(vmspace_pmap(vm));
vm_map_zfini(&vm->vm_map, sizeof(vm->vm_map));
}
static int
vmspace_zinit(void *mem, int size, int flags)
{
struct vmspace *vm;
vm = (struct vmspace *)mem;
(void)vm_map_zinit(&vm->vm_map, sizeof(vm->vm_map), flags);
pmap_pinit(vmspace_pmap(vm));
return (0);
}
static void
vm_map_zfini(void *mem, int size)
{
vm_map_t map;
map = (vm_map_t)mem;
mtx_destroy(&map->system_mtx);
sx_destroy(&map->lock);
}
static int
vm_map_zinit(void *mem, int size, int flags)
{
vm_map_t map;
map = (vm_map_t)mem;
map->nentries = 0;
map->size = 0;
map->infork = 0;
mtx_init(&map->system_mtx, "system map", NULL, MTX_DEF | MTX_DUPOK);
sx_init(&map->lock, "user map");
return (0);
}
#ifdef INVARIANTS
static void
vmspace_zdtor(void *mem, int size, void *arg)
{
struct vmspace *vm;
vm = (struct vmspace *)mem;
vm_map_zdtor(&vm->vm_map, sizeof(vm->vm_map), arg);
}
static void
vm_map_zdtor(void *mem, int size, void *arg)
{
vm_map_t map;
map = (vm_map_t)mem;
KASSERT(map->nentries == 0,
2003-11-03 16:14:45 +00:00
("map %p nentries == %d on free.",
map, map->nentries));
KASSERT(map->size == 0,
("map %p size == %lu on free.",
map, (unsigned long)map->size));
KASSERT(map->infork == 0,
("map %p infork == %d on free.",
map, map->infork));
1994-05-24 10:09:53 +00:00
}
#endif /* INVARIANTS */
1994-05-24 10:09:53 +00:00
/*
* Allocate a vmspace structure, including a vm_map and pmap,
* and initialize those structures. The refcnt is set to 1.
*/
struct vmspace *
VM level code cleanups. 1) Start using TSM. Struct procs continue to point to upages structure, after being freed. Struct vmspace continues to point to pte object and kva space for kstack. u_map is now superfluous. 2) vm_map's don't need to be reference counted. They always exist either in the kernel or in a vmspace. The vmspaces are managed by reference counts. 3) Remove the "wired" vm_map nonsense. 4) No need to keep a cache of kernel stack kva's. 5) Get rid of strange looking ++var, and change to var++. 6) Change more data structures to use our "zone" allocator. Added struct proc, struct vmspace and struct vnode. This saves a significant amount of kva space and physical memory. Additionally, this enables TSM for the zone managed memory. 7) Keep ioopt disabled for now. 8) Remove the now bogus "single use" map concept. 9) Use generation counts or id's for data structures residing in TSM, where it allows us to avoid unneeded restart overhead during traversals, where blocking might occur. 10) Account better for memory deficits, so the pageout daemon will be able to make enough memory available (experimental.) 11) Fix some vnode locking problems. (From Tor, I think.) 12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp. (experimental.) 13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c code. Use generation counts, get rid of unneded collpase operations, and clean up the cluster code. 14) Make vm_zone more suitable for TSM. This commit is partially as a result of discussions and contributions from other people, including DG, Tor Egge, PHK, and probably others that I have forgotten to attribute (so let me know, if I forgot.) This is not the infamous, final cleanup of the vnode stuff, but a necessary step. Vnode mgmt should be correct, but things might still change, and there is still some missing stuff (like ioopt, and physical backing of non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
vmspace_alloc(min, max)
1994-05-24 10:09:53 +00:00
vm_offset_t min, max;
{
struct vmspace *vm;
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 = uma_zalloc(vmspace_zone, M_WAITOK);
CTR1(KTR_VM, "vmspace_alloc: %p", vm);
_vm_map_init(&vm->vm_map, min, max);
vm->vm_map.pmap = vmspace_pmap(vm); /* XXX */
1994-05-24 10:09:53 +00:00
vm->vm_refcnt = 1;
VM level code cleanups. 1) Start using TSM. Struct procs continue to point to upages structure, after being freed. Struct vmspace continues to point to pte object and kva space for kstack. u_map is now superfluous. 2) vm_map's don't need to be reference counted. They always exist either in the kernel or in a vmspace. The vmspaces are managed by reference counts. 3) Remove the "wired" vm_map nonsense. 4) No need to keep a cache of kernel stack kva's. 5) Get rid of strange looking ++var, and change to var++. 6) Change more data structures to use our "zone" allocator. Added struct proc, struct vmspace and struct vnode. This saves a significant amount of kva space and physical memory. Additionally, this enables TSM for the zone managed memory. 7) Keep ioopt disabled for now. 8) Remove the now bogus "single use" map concept. 9) Use generation counts or id's for data structures residing in TSM, where it allows us to avoid unneeded restart overhead during traversals, where blocking might occur. 10) Account better for memory deficits, so the pageout daemon will be able to make enough memory available (experimental.) 11) Fix some vnode locking problems. (From Tor, I think.) 12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp. (experimental.) 13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c code. Use generation counts, get rid of unneded collpase operations, and clean up the cluster code. 14) Make vm_zone more suitable for TSM. This commit is partially as a result of discussions and contributions from other people, including DG, Tor Egge, PHK, and probably others that I have forgotten to attribute (so let me know, if I forgot.) This is not the infamous, final cleanup of the vnode stuff, but a necessary step. Vnode mgmt should be correct, but things might still change, and there is still some missing stuff (like ioopt, and physical backing of non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
vm->vm_shm = NULL;
vm->vm_swrss = 0;
vm->vm_tsize = 0;
vm->vm_dsize = 0;
vm->vm_ssize = 0;
vm->vm_taddr = 0;
vm->vm_daddr = 0;
vm->vm_maxsaddr = 0;
vm->vm_exitingcnt = 0;
1994-05-24 10:09:53 +00:00
return (vm);
}
void
2003-11-03 16:14:45 +00:00
vm_init2(void)
{
uma_zone_set_obj(kmapentzone, &kmapentobj, lmin(cnt.v_page_count,
(VM_MAX_KERNEL_ADDRESS - KERNBASE) / PAGE_SIZE) / 8 +
maxproc * 2 + maxfiles);
vmspace_zone = uma_zcreate("VMSPACE", sizeof(struct vmspace), NULL,
#ifdef INVARIANTS
vmspace_zdtor,
#else
NULL,
#endif
vmspace_zinit, vmspace_zfini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
pmap_init2();
}
static __inline void
vmspace_dofree(struct vmspace *vm)
{
CTR1(KTR_VM, "vmspace_free: %p", vm);
/*
* Make sure any SysV shm is freed, it might not have been in
* exit1().
*/
shmexit(vm);
/*
* Lock the map, to wait out all other references to it.
* Delete all of the mappings and pages they hold, then call
* the pmap module to reclaim anything left.
*/
vm_map_lock(&vm->vm_map);
(void) vm_map_delete(&vm->vm_map, vm->vm_map.min_offset,
vm->vm_map.max_offset);
vm_map_unlock(&vm->vm_map);
uma_zfree(vmspace_zone, vm);
}
1994-05-24 10:09:53 +00:00
void
vmspace_free(struct vmspace *vm)
1994-05-24 10:09:53 +00:00
{
int refcnt;
1994-05-24 10:09:53 +00:00
swap_pager.c: Fixed long standing bug in freeing swap space during object collapses. Fixed 'out of space' messages from printing out too often. Modified to use new kmem_malloc() calling convention. Implemented an additional stat in the swap pager struct to count the amount of space allocated to that pager. This may be removed at some point in the future. Minimized unnecessary wakeups. vm_fault.c: Don't try to collect fault stats on 'swapped' processes - there aren't any upages to store the stats in. Changed read-ahead policy (again!). vm_glue.c: Be sure to gain a reference to the process's map before swapping. Be sure to lose it when done. kern_malloc.c: Added the ability to specify if allocations are at interrupt time or are 'safe'; this affects what types of pages can be allocated. vm_map.c: Fixed a variety of map lock problems; there's still a lurking bug that will eventually bite. vm_object.c: Explicitly initialize the object fields rather than bzeroing the struct. Eliminated the 'rcollapse' code and folded it's functionality into the "real" collapse routine. Moved an object_unlock() so that the backing_object is protected in the qcollapse routine. Make sure nobody fools with the backing_object when we're destroying it. Added some diagnostic code which can be called from the debugger that looks through all the internal objects and makes certain that they all belong to someone. vm_page.c: Fixed a rather serious logic bug that would result in random system crashes. Changed pagedaemon wakeup policy (again!). vm_pageout.c: Removed unnecessary page rotations on the inactive queue. Changed the number of pages to explicitly free to just free_reserved level. Submitted by: John Dyson
1995-02-02 09:09:15 +00:00
if (vm->vm_refcnt == 0)
panic("vmspace_free: attempt to free already freed vmspace");
do
refcnt = vm->vm_refcnt;
while (!atomic_cmpset_int(&vm->vm_refcnt, refcnt, refcnt - 1));
if (refcnt == 1 && vm->vm_exitingcnt == 0)
vmspace_dofree(vm);
}
void
vmspace_exitfree(struct proc *p)
{
struct vmspace *vm;
int exitingcnt;
vm = p->p_vmspace;
p->p_vmspace = NULL;
/*
* cleanup by parent process wait()ing on exiting child. vm_refcnt
* may not be 0 (e.g. fork() and child exits without exec()ing).
* exitingcnt may increment above 0 and drop back down to zero
* several times while vm_refcnt is held non-zero. vm_refcnt
2003-11-03 16:14:45 +00:00
* may also increment above 0 and drop back down to zero several
* times while vm_exitingcnt is held non-zero.
2003-11-03 16:14:45 +00:00
*
* The last wait on the exiting child's vmspace will clean up
* the remainder of the vmspace.
*/
do
exitingcnt = vm->vm_exitingcnt;
while (!atomic_cmpset_int(&vm->vm_exitingcnt, exitingcnt,
exitingcnt - 1));
if (vm->vm_refcnt == 0 && exitingcnt == 1)
vmspace_dofree(vm);
1994-05-24 10:09:53 +00:00
}
void
_vm_map_lock(vm_map_t map, const char *file, int line)
{
if (map->system_map)
_mtx_lock_flags(&map->system_mtx, 0, file, line);
else
_sx_xlock(&map->lock, file, line);
map->timestamp++;
}
void
_vm_map_unlock(vm_map_t map, const char *file, int line)
{
if (map->system_map)
_mtx_unlock_flags(&map->system_mtx, 0, file, line);
else
_sx_xunlock(&map->lock, file, line);
}
void
_vm_map_lock_read(vm_map_t map, const char *file, int line)
{
if (map->system_map)
_mtx_lock_flags(&map->system_mtx, 0, file, line);
else
_sx_xlock(&map->lock, file, line);
}
void
_vm_map_unlock_read(vm_map_t map, const char *file, int line)
{
if (map->system_map)
_mtx_unlock_flags(&map->system_mtx, 0, file, line);
else
_sx_xunlock(&map->lock, file, line);
}
int
_vm_map_trylock(vm_map_t map, const char *file, int line)
{
int error;
error = map->system_map ?
!_mtx_trylock(&map->system_mtx, 0, file, line) :
!_sx_try_xlock(&map->lock, file, line);
if (error == 0)
map->timestamp++;
return (error == 0);
}
int
_vm_map_trylock_read(vm_map_t map, const char *file, int line)
{
int error;
error = map->system_map ?
!_mtx_trylock(&map->system_mtx, 0, file, line) :
!_sx_try_xlock(&map->lock, file, line);
return (error == 0);
}
int
_vm_map_lock_upgrade(vm_map_t map, const char *file, int line)
{
#ifdef INVARIANTS
if (map->system_map) {
_mtx_assert(&map->system_mtx, MA_OWNED, file, line);
} else
_sx_assert(&map->lock, SX_XLOCKED, file, line);
#endif
map->timestamp++;
return (0);
}
void
_vm_map_lock_downgrade(vm_map_t map, const char *file, int line)
{
#ifdef INVARIANTS
if (map->system_map) {
_mtx_assert(&map->system_mtx, MA_OWNED, file, line);
} else
_sx_assert(&map->lock, SX_XLOCKED, file, line);
#endif
}
/*
* vm_map_unlock_and_wait:
*/
int
vm_map_unlock_and_wait(vm_map_t map, boolean_t user_wait)
{
mtx_lock(&map_sleep_mtx);
vm_map_unlock(map);
return (msleep(&map->root, &map_sleep_mtx, PDROP | PVM, "vmmaps", 0));
}
/*
* vm_map_wakeup:
*/
void
vm_map_wakeup(vm_map_t map)
{
/*
* Acquire and release map_sleep_mtx to prevent a wakeup()
* from being performed (and lost) between the vm_map_unlock()
* and the msleep() in vm_map_unlock_and_wait().
*/
mtx_lock(&map_sleep_mtx);
mtx_unlock(&map_sleep_mtx);
wakeup(&map->root);
}
long
vmspace_resident_count(struct vmspace *vmspace)
{
return pmap_resident_count(vmspace_pmap(vmspace));
}
long
vmspace_wired_count(struct vmspace *vmspace)
{
return pmap_wired_count(vmspace_pmap(vmspace));
}
1994-05-24 10:09:53 +00:00
/*
* vm_map_create:
*
* Creates and returns a new empty VM map with
* the given physical map structure, and having
* the given lower and upper address bounds.
*/
1995-05-30 08:16:23 +00:00
vm_map_t
vm_map_create(pmap_t pmap, vm_offset_t min, vm_offset_t max)
1994-05-24 10:09:53 +00:00
{
vm_map_t result;
1994-05-24 10:09:53 +00:00
result = uma_zalloc(mapzone, M_WAITOK);
CTR1(KTR_VM, "vm_map_create: %p", result);
_vm_map_init(result, min, max);
1994-05-24 10:09:53 +00:00
result->pmap = pmap;
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 (result);
1994-05-24 10:09:53 +00:00
}
/*
* Initialize an existing vm_map structure
* such as that in the vmspace structure.
* The pmap is set elsewhere.
*/
static void
_vm_map_init(vm_map_t map, vm_offset_t min, vm_offset_t max)
1994-05-24 10:09:53 +00:00
{
1994-05-24 10:09:53 +00:00
map->header.next = map->header.prev = &map->header;
map->needs_wakeup = FALSE;
map->system_map = 0;
1994-05-24 10:09:53 +00:00
map->min_offset = min;
map->max_offset = max;
map->first_free = &map->header;
map->flags = 0;
map->root = NULL;
1994-05-24 10:09:53 +00:00
map->timestamp = 0;
}
void
vm_map_init(vm_map_t map, vm_offset_t min, vm_offset_t max)
{
_vm_map_init(map, min, max);
mtx_init(&map->system_mtx, "system map", NULL, MTX_DEF | MTX_DUPOK);
sx_init(&map->lock, "user map");
}
This set of commits to the VM system does the following, and contain contributions or ideas from Stephen McKay <syssgm@devetir.qld.gov.au>, Alan Cox <alc@cs.rice.edu>, David Greenman <davidg@freebsd.org> and me: More usage of the TAILQ macros. Additional minor fix to queue.h. Performance enhancements to the pageout daemon. Addition of a wait in the case that the pageout daemon has to run immediately. Slightly modify the pageout algorithm. Significant revamp of the pmap/fork code: 1) PTE's and UPAGES's are NO LONGER in the process's map. 2) PTE's and UPAGES's reside in their own objects. 3) TOTAL elimination of recursive page table pagefaults. 4) The page directory now resides in the PTE object. 5) Implemented pmap_copy, thereby speeding up fork time. 6) Changed the pv entries so that the head is a pointer and not an entire entry. 7) Significant cleanup of pmap_protect, and pmap_remove. 8) Removed significant amounts of machine dependent fork code from vm_glue. Pushed much of that code into the machine dependent pmap module. 9) Support more completely the reuse of already zeroed pages (Page table pages and page directories) as being already zeroed. Performance and code cleanups in vm_map: 1) Improved and simplified allocation of map entries. 2) Improved vm_map_copy code. 3) Corrected some minor problems in the simplify code. Implemented splvm (combo of splbio and splimp.) The VM code now seldom uses splhigh. Improved the speed of and simplified kmem_malloc. Minor mod to vm_fault to avoid using pre-zeroed pages in the case of objects with backing objects along with the already existant condition of having a vnode. (If there is a backing object, there will likely be a COW... With a COW, it isn't necessary to start with a pre-zeroed page.) Minor reorg of source to perhaps improve locality of ref.
1996-05-18 03:38:05 +00:00
/*
* vm_map_entry_dispose: [ internal use only ]
*
* Inverse of vm_map_entry_create.
*/
static void
vm_map_entry_dispose(vm_map_t map, vm_map_entry_t entry)
This set of commits to the VM system does the following, and contain contributions or ideas from Stephen McKay <syssgm@devetir.qld.gov.au>, Alan Cox <alc@cs.rice.edu>, David Greenman <davidg@freebsd.org> and me: More usage of the TAILQ macros. Additional minor fix to queue.h. Performance enhancements to the pageout daemon. Addition of a wait in the case that the pageout daemon has to run immediately. Slightly modify the pageout algorithm. Significant revamp of the pmap/fork code: 1) PTE's and UPAGES's are NO LONGER in the process's map. 2) PTE's and UPAGES's reside in their own objects. 3) TOTAL elimination of recursive page table pagefaults. 4) The page directory now resides in the PTE object. 5) Implemented pmap_copy, thereby speeding up fork time. 6) Changed the pv entries so that the head is a pointer and not an entire entry. 7) Significant cleanup of pmap_protect, and pmap_remove. 8) Removed significant amounts of machine dependent fork code from vm_glue. Pushed much of that code into the machine dependent pmap module. 9) Support more completely the reuse of already zeroed pages (Page table pages and page directories) as being already zeroed. Performance and code cleanups in vm_map: 1) Improved and simplified allocation of map entries. 2) Improved vm_map_copy code. 3) Corrected some minor problems in the simplify code. Implemented splvm (combo of splbio and splimp.) The VM code now seldom uses splhigh. Improved the speed of and simplified kmem_malloc. Minor mod to vm_fault to avoid using pre-zeroed pages in the case of objects with backing objects along with the already existant condition of having a vnode. (If there is a backing object, there will likely be a COW... With a COW, it isn't necessary to start with a pre-zeroed page.) Minor reorg of source to perhaps improve locality of ref.
1996-05-18 03:38:05 +00:00
{
uma_zfree(map->system_map ? kmapentzone : mapentzone, entry);
This set of commits to the VM system does the following, and contain contributions or ideas from Stephen McKay <syssgm@devetir.qld.gov.au>, Alan Cox <alc@cs.rice.edu>, David Greenman <davidg@freebsd.org> and me: More usage of the TAILQ macros. Additional minor fix to queue.h. Performance enhancements to the pageout daemon. Addition of a wait in the case that the pageout daemon has to run immediately. Slightly modify the pageout algorithm. Significant revamp of the pmap/fork code: 1) PTE's and UPAGES's are NO LONGER in the process's map. 2) PTE's and UPAGES's reside in their own objects. 3) TOTAL elimination of recursive page table pagefaults. 4) The page directory now resides in the PTE object. 5) Implemented pmap_copy, thereby speeding up fork time. 6) Changed the pv entries so that the head is a pointer and not an entire entry. 7) Significant cleanup of pmap_protect, and pmap_remove. 8) Removed significant amounts of machine dependent fork code from vm_glue. Pushed much of that code into the machine dependent pmap module. 9) Support more completely the reuse of already zeroed pages (Page table pages and page directories) as being already zeroed. Performance and code cleanups in vm_map: 1) Improved and simplified allocation of map entries. 2) Improved vm_map_copy code. 3) Corrected some minor problems in the simplify code. Implemented splvm (combo of splbio and splimp.) The VM code now seldom uses splhigh. Improved the speed of and simplified kmem_malloc. Minor mod to vm_fault to avoid using pre-zeroed pages in the case of objects with backing objects along with the already existant condition of having a vnode. (If there is a backing object, there will likely be a COW... With a COW, it isn't necessary to start with a pre-zeroed page.) Minor reorg of source to perhaps improve locality of ref.
1996-05-18 03:38:05 +00:00
}
1994-05-24 10:09:53 +00:00
/*
* vm_map_entry_create: [ internal use only ]
*
* Allocates a VM map entry for insertion.
2001-04-12 21:50:03 +00:00
* No entry fields are filled in.
1994-05-24 10:09:53 +00:00
*/
static vm_map_entry_t
vm_map_entry_create(vm_map_t map)
1994-05-24 10:09:53 +00:00
{
vm_map_entry_t new_entry;
if (map->system_map)
new_entry = uma_zalloc(kmapentzone, M_NOWAIT);
else
new_entry = uma_zalloc(mapentzone, M_WAITOK);
if (new_entry == NULL)
panic("vm_map_entry_create: kernel resources exhausted");
return (new_entry);
1994-05-24 10:09:53 +00:00
}
/*
* vm_map_entry_set_behavior:
*
* Set the expected access behavior, either normal, random, or
* sequential.
*/
static __inline void
vm_map_entry_set_behavior(vm_map_entry_t entry, u_char behavior)
{
entry->eflags = (entry->eflags & ~MAP_ENTRY_BEHAV_MASK) |
(behavior & MAP_ENTRY_BEHAV_MASK);
}
/*
* vm_map_entry_splay:
*
* Implements Sleator and Tarjan's top-down splay algorithm. Returns
* the vm_map_entry containing the given address. If, however, that
* address is not found in the vm_map, returns a vm_map_entry that is
* adjacent to the address, coming before or after it.
*/
static vm_map_entry_t
vm_map_entry_splay(vm_offset_t address, vm_map_entry_t root)
{
struct vm_map_entry dummy;
vm_map_entry_t lefttreemax, righttreemin, y;
if (root == NULL)
return (root);
lefttreemax = righttreemin = &dummy;
for (;; root = y) {
if (address < root->start) {
if ((y = root->left) == NULL)
break;
if (address < y->start) {
/* Rotate right. */
root->left = y->right;
y->right = root;
root = y;
if ((y = root->left) == NULL)
break;
}
/* Link into the new root's right tree. */
righttreemin->left = root;
righttreemin = root;
} else if (address >= root->end) {
if ((y = root->right) == NULL)
break;
if (address >= y->end) {
/* Rotate left. */
root->right = y->left;
y->left = root;
root = y;
if ((y = root->right) == NULL)
break;
}
/* Link into the new root's left tree. */
lefttreemax->right = root;
lefttreemax = root;
} else
break;
}
/* Assemble the new root. */
lefttreemax->right = root->left;
righttreemin->left = root->right;
root->left = dummy.right;
root->right = dummy.left;
return (root);
}
1994-05-24 10:09:53 +00:00
/*
* vm_map_entry_{un,}link:
*
* Insert/remove entries from maps.
*/
static void
vm_map_entry_link(vm_map_t map,
vm_map_entry_t after_where,
vm_map_entry_t entry)
{
CTR4(KTR_VM,
"vm_map_entry_link: map %p, nentries %d, entry %p, after %p", map,
map->nentries, entry, after_where);
map->nentries++;
entry->prev = after_where;
entry->next = after_where->next;
entry->next->prev = entry;
after_where->next = entry;
if (after_where != &map->header) {
if (after_where != map->root)
vm_map_entry_splay(after_where->start, map->root);
entry->right = after_where->right;
entry->left = after_where;
after_where->right = NULL;
} else {
entry->right = map->root;
entry->left = NULL;
}
map->root = entry;
}
static void
vm_map_entry_unlink(vm_map_t map,
vm_map_entry_t entry)
{
vm_map_entry_t next, prev, root;
if (entry != map->root)
vm_map_entry_splay(entry->start, map->root);
if (entry->left == NULL)
root = entry->right;
else {
root = vm_map_entry_splay(entry->start, entry->left);
root->right = entry->right;
}
map->root = root;
prev = entry->prev;
next = entry->next;
next->prev = prev;
prev->next = next;
map->nentries--;
CTR3(KTR_VM, "vm_map_entry_unlink: map %p, nentries %d, entry %p", map,
map->nentries, entry);
}
1994-05-24 10:09:53 +00:00
/*
* vm_map_lookup_entry: [ internal use only ]
*
* Finds the map entry containing (or
* immediately preceding) the specified address
* in the given map; the entry is returned
* in the "entry" parameter. The boolean
* result indicates whether the address is
* actually contained in the map.
*/
boolean_t
vm_map_lookup_entry(
vm_map_t map,
vm_offset_t address,
vm_map_entry_t *entry) /* OUT */
{
vm_map_entry_t cur;
cur = vm_map_entry_splay(address, map->root);
if (cur == NULL)
*entry = &map->header;
else {
map->root = cur;
if (address >= cur->start) {
*entry = cur;
if (cur->end > address)
return (TRUE);
} else
*entry = cur->prev;
}
return (FALSE);
}
1994-05-24 10:09:53 +00:00
/*
* vm_map_insert:
*
* Inserts the given whole VM object into the target
* map at the specified address range. The object's
* size should match that of the address range.
*
* Requires that the map be locked, and leaves it so.
*
* If object is non-NULL, ref count must be bumped by caller
* prior to making call to account for the new entry.
1994-05-24 10:09:53 +00:00
*/
int
vm_map_insert(vm_map_t map, vm_object_t object, vm_ooffset_t offset,
vm_offset_t start, vm_offset_t end, vm_prot_t prot, vm_prot_t max,
int cow)
1994-05-24 10:09:53 +00:00
{
vm_map_entry_t new_entry;
vm_map_entry_t prev_entry;
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_map_entry_t temp_entry;
vm_eflags_t protoeflags;
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
* Check that the start and end points are not bogus.
1994-05-24 10:09:53 +00:00
*/
if ((start < map->min_offset) || (end > map->max_offset) ||
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
(start >= end))
return (KERN_INVALID_ADDRESS);
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
* Find the entry prior to the proposed starting address; if it's part
* of an existing entry, this range is bogus.
1994-05-24 10:09:53 +00:00
*/
if (vm_map_lookup_entry(map, start, &temp_entry))
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 (KERN_NO_SPACE);
1994-05-24 10:09:53 +00:00
prev_entry = temp_entry;
/*
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
* Assert that the next entry doesn't overlap the end point.
1994-05-24 10:09:53 +00:00
*/
if ((prev_entry->next != &map->header) &&
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
(prev_entry->next->start < end))
return (KERN_NO_SPACE);
1994-05-24 10:09:53 +00:00
protoeflags = 0;
if (cow & MAP_COPY_ON_WRITE)
protoeflags |= MAP_ENTRY_COW|MAP_ENTRY_NEEDS_COPY;
if (cow & MAP_NOFAULT) {
protoeflags |= MAP_ENTRY_NOFAULT;
KASSERT(object == NULL,
("vm_map_insert: paradoxical MAP_NOFAULT request"));
}
if (cow & MAP_DISABLE_SYNCER)
protoeflags |= MAP_ENTRY_NOSYNC;
if (cow & MAP_DISABLE_COREDUMP)
protoeflags |= MAP_ENTRY_NOCOREDUMP;
if (object != NULL) {
/*
* OBJ_ONEMAPPING must be cleared unless this mapping
* is trivially proven to be the only mapping for any
* of the object's pages. (Object granularity
* reference counting is insufficient to recognize
2003-11-03 16:14:45 +00:00
* aliases with precision.)
*/
VM_OBJECT_LOCK(object);
if (object->ref_count > 1 || object->shadow_count != 0)
vm_object_clear_flag(object, OBJ_ONEMAPPING);
VM_OBJECT_UNLOCK(object);
}
else if ((prev_entry != &map->header) &&
(prev_entry->eflags == protoeflags) &&
(prev_entry->end == start) &&
(prev_entry->wired_count == 0) &&
((prev_entry->object.vm_object == NULL) ||
vm_object_coalesce(prev_entry->object.vm_object,
prev_entry->offset,
(vm_size_t)(prev_entry->end - prev_entry->start),
(vm_size_t)(end - prev_entry->end)))) {
/*
* We were able to extend the object. Determine if we
2003-11-03 16:14:45 +00:00
* can extend the previous map entry to include the
* new range as well.
*/
if ((prev_entry->inheritance == VM_INHERIT_DEFAULT) &&
(prev_entry->protection == prot) &&
(prev_entry->max_protection == max)) {
map->size += (end - prev_entry->end);
prev_entry->end = end;
vm_map_simplify_entry(map, prev_entry);
return (KERN_SUCCESS);
This set of commits to the VM system does the following, and contain contributions or ideas from Stephen McKay <syssgm@devetir.qld.gov.au>, Alan Cox <alc@cs.rice.edu>, David Greenman <davidg@freebsd.org> and me: More usage of the TAILQ macros. Additional minor fix to queue.h. Performance enhancements to the pageout daemon. Addition of a wait in the case that the pageout daemon has to run immediately. Slightly modify the pageout algorithm. Significant revamp of the pmap/fork code: 1) PTE's and UPAGES's are NO LONGER in the process's map. 2) PTE's and UPAGES's reside in their own objects. 3) TOTAL elimination of recursive page table pagefaults. 4) The page directory now resides in the PTE object. 5) Implemented pmap_copy, thereby speeding up fork time. 6) Changed the pv entries so that the head is a pointer and not an entire entry. 7) Significant cleanup of pmap_protect, and pmap_remove. 8) Removed significant amounts of machine dependent fork code from vm_glue. Pushed much of that code into the machine dependent pmap module. 9) Support more completely the reuse of already zeroed pages (Page table pages and page directories) as being already zeroed. Performance and code cleanups in vm_map: 1) Improved and simplified allocation of map entries. 2) Improved vm_map_copy code. 3) Corrected some minor problems in the simplify code. Implemented splvm (combo of splbio and splimp.) The VM code now seldom uses splhigh. Improved the speed of and simplified kmem_malloc. Minor mod to vm_fault to avoid using pre-zeroed pages in the case of objects with backing objects along with the already existant condition of having a vnode. (If there is a backing object, there will likely be a COW... With a COW, it isn't necessary to start with a pre-zeroed page.) Minor reorg of source to perhaps improve locality of ref.
1996-05-18 03:38:05 +00:00
}
/*
* If we can extend the object but cannot extend the
* map entry, we have to create a new map entry. We
* must bump the ref count on the extended object to
* account for it. object may be NULL.
*/
object = prev_entry->object.vm_object;
offset = prev_entry->offset +
(prev_entry->end - prev_entry->start);
vm_object_reference(object);
1994-05-24 10:09:53 +00:00
}
/*
* NOTE: if conditionals fail, object can be NULL here. This occurs
* in things like the buffer map where we manage kva but do not manage
* backing objects.
*/
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
* Create a new entry
1994-05-24 10:09:53 +00:00
*/
new_entry = vm_map_entry_create(map);
new_entry->start = start;
new_entry->end = end;
new_entry->eflags = protoeflags;
1994-05-24 10:09:53 +00:00
new_entry->object.vm_object = object;
new_entry->offset = offset;
new_entry->avail_ssize = 0;
new_entry->inheritance = VM_INHERIT_DEFAULT;
new_entry->protection = prot;
new_entry->max_protection = max;
new_entry->wired_count = 0;
1994-05-24 10:09:53 +00:00
/*
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
* Insert the new entry into the list
1994-05-24 10:09:53 +00:00
*/
vm_map_entry_link(map, prev_entry, new_entry);
map->size += new_entry->end - new_entry->start;
/*
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
* Update the free space hint
1994-05-24 10:09:53 +00:00
*/
if ((map->first_free == prev_entry) &&
(prev_entry->end >= new_entry->start)) {
map->first_free = new_entry;
}
1994-05-24 10:09:53 +00:00
#if 0
/*
* Temporarily removed to avoid MAP_STACK panic, due to
* MAP_STACK being a huge hack. Will be added back in
* when MAP_STACK (and the user stack mapping) is fixed.
*/
/*
* It may be possible to simplify the entry
*/
vm_map_simplify_entry(map, new_entry);
#endif
if (cow & (MAP_PREFAULT|MAP_PREFAULT_PARTIAL)) {
vm_map_pmap_enter(map, start, prot,
object, OFF_TO_IDX(offset), end - start,
cow & MAP_PREFAULT_PARTIAL);
}
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 (KERN_SUCCESS);
1994-05-24 10:09:53 +00:00
}
/*
* Find sufficient space for `length' bytes in the given map, starting at
* `start'. The map must be locked. Returns 0 on success, 1 on no space.
*/
int
vm_map_findspace(
vm_map_t map,
vm_offset_t start,
vm_size_t length,
vm_offset_t *addr)
1994-05-24 10:09:53 +00:00
{
vm_map_entry_t entry, next;
vm_offset_t end;
1994-05-24 10:09:53 +00:00
if (start < map->min_offset)
start = map->min_offset;
if (start > map->max_offset)
return (1);
/*
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
* Look for the first possible address; if there's already something
* at this address, we have to start after it.
1994-05-24 10:09:53 +00:00
*/
if (start == map->min_offset) {
if ((entry = map->first_free) != &map->header)
1994-05-24 10:09:53 +00:00
start = entry->end;
} else {
vm_map_entry_t tmp;
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
if (vm_map_lookup_entry(map, start, &tmp))
start = tmp->end;
entry = tmp;
}
/*
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
* Look through the rest of the map, trying to fit a new region in the
* gap between existing regions, or after the very last region.
1994-05-24 10:09:53 +00:00
*/
for (;; start = (entry = next)->end) {
/*
* Find the end of the proposed new region. Be sure we didn't
* go beyond the end of the map, or wrap around the address;
* if so, we lose. Otherwise, if this is the last entry, or
* if the proposed new region fits before the next entry, we
* win.
*/
end = start + length;
if (end > map->max_offset || end < start)
return (1);
next = entry->next;
if (next == &map->header || next->start >= end)
break;
}
*addr = start;
if (map == kernel_map) {
vm_offset_t ksize;
if ((ksize = round_page(start + length)) > kernel_vm_end) {
pmap_growkernel(ksize);
}
}
1994-05-24 10:09:53 +00:00
return (0);
}
/*
* vm_map_find finds an unallocated region in the target address
* map with the given length. The search is defined to be
* first-fit from the specified address; the region found is
* returned in the same parameter.
*
* If object is non-NULL, ref count must be bumped by caller
* prior to making call to account for the new entry.
1994-05-24 10:09:53 +00:00
*/
int
vm_map_find(vm_map_t map, vm_object_t object, vm_ooffset_t offset,
vm_offset_t *addr, /* IN/OUT */
vm_size_t length, boolean_t find_space, vm_prot_t prot,
vm_prot_t max, int cow)
1994-05-24 10:09:53 +00:00
{
vm_offset_t start;
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
int result, s = 0;
1994-05-24 10:09:53 +00:00
start = *addr;
Introduce numerous SMP friendly changes to the mbuf allocator. Namely, introduce a modified allocation mechanism for mbufs and mbuf clusters; one which can scale under SMP and which offers the possibility of resource reclamation to be implemented in the future. Notable advantages: o Reduce contention for SMP by offering per-CPU pools and locks. o Better use of data cache due to per-CPU pools. o Much less code cache pollution due to excessively large allocation macros. o Framework for `grouping' objects from same page together so as to be able to possibly free wired-down pages back to the system if they are no longer needed by the network stacks. Additional things changed with this addition: - Moved some mbuf specific declarations and initializations from sys/conf/param.c into mbuf-specific code where they belong. - m_getclr() has been renamed to m_get_clrd() because the old name is really confusing. m_getclr() HAS been preserved though and is defined to the new name. No tree sweep has been done "to change the interface," as the old name will continue to be supported and is not depracated. The change was merely done because m_getclr() sounds too much like "m_get a cluster." - TEMPORARILY disabled mbtypes statistics displaying in netstat(1) and systat(1) (see TODO below). - Fixed systat(1) to display number of "free mbufs" based on new per-CPU stat structures. - Fixed netstat(1) to display new per-CPU stats based on sysctl-exported per-CPU stat structures. All infos are fetched via sysctl. TODO (in order of priority): - Re-enable mbtypes statistics in both netstat(1) and systat(1) after introducing an SMP friendly way to collect the mbtypes stats under the already introduced per-CPU locks (i.e. hopefully don't use atomic() - it seems too costly for a mere stat update, especially when other locks are already present). - Optionally have systat(1) display not only "total free mbufs" but also "total free mbufs per CPU pool." - Fix minor length-fetching issues in netstat(1) related to recently re-enabled option to read mbuf stats from a core file. - Move reference counters at least for mbuf clusters into an unused portion of the cluster itself, to save space and need to allocate a counter. - Look into introducing resource freeing possibly from a kproc. Reviewed by (in parts): jlemon, jake, silby, terry Tested by: jlemon (Intel & Alpha), mjacob (Intel & Alpha) Preliminary performance measurements: jlemon (and me, obviously) URL: http://people.freebsd.org/~bmilekic/mb_alloc/
2001-06-22 06:35:32 +00:00
if (map == kmem_map)
This set of commits to the VM system does the following, and contain contributions or ideas from Stephen McKay <syssgm@devetir.qld.gov.au>, Alan Cox <alc@cs.rice.edu>, David Greenman <davidg@freebsd.org> and me: More usage of the TAILQ macros. Additional minor fix to queue.h. Performance enhancements to the pageout daemon. Addition of a wait in the case that the pageout daemon has to run immediately. Slightly modify the pageout algorithm. Significant revamp of the pmap/fork code: 1) PTE's and UPAGES's are NO LONGER in the process's map. 2) PTE's and UPAGES's reside in their own objects. 3) TOTAL elimination of recursive page table pagefaults. 4) The page directory now resides in the PTE object. 5) Implemented pmap_copy, thereby speeding up fork time. 6) Changed the pv entries so that the head is a pointer and not an entire entry. 7) Significant cleanup of pmap_protect, and pmap_remove. 8) Removed significant amounts of machine dependent fork code from vm_glue. Pushed much of that code into the machine dependent pmap module. 9) Support more completely the reuse of already zeroed pages (Page table pages and page directories) as being already zeroed. Performance and code cleanups in vm_map: 1) Improved and simplified allocation of map entries. 2) Improved vm_map_copy code. 3) Corrected some minor problems in the simplify code. Implemented splvm (combo of splbio and splimp.) The VM code now seldom uses splhigh. Improved the speed of and simplified kmem_malloc. Minor mod to vm_fault to avoid using pre-zeroed pages in the case of objects with backing objects along with the already existant condition of having a vnode. (If there is a backing object, there will likely be a COW... With a COW, it isn't necessary to start with a pre-zeroed page.) Minor reorg of source to perhaps improve locality of ref.
1996-05-18 03:38:05 +00:00
s = splvm();
vm_map_lock(map);
1994-05-24 10:09:53 +00:00
if (find_space) {
if (vm_map_findspace(map, start, length, addr)) {
vm_map_unlock(map);
Introduce numerous SMP friendly changes to the mbuf allocator. Namely, introduce a modified allocation mechanism for mbufs and mbuf clusters; one which can scale under SMP and which offers the possibility of resource reclamation to be implemented in the future. Notable advantages: o Reduce contention for SMP by offering per-CPU pools and locks. o Better use of data cache due to per-CPU pools. o Much less code cache pollution due to excessively large allocation macros. o Framework for `grouping' objects from same page together so as to be able to possibly free wired-down pages back to the system if they are no longer needed by the network stacks. Additional things changed with this addition: - Moved some mbuf specific declarations and initializations from sys/conf/param.c into mbuf-specific code where they belong. - m_getclr() has been renamed to m_get_clrd() because the old name is really confusing. m_getclr() HAS been preserved though and is defined to the new name. No tree sweep has been done "to change the interface," as the old name will continue to be supported and is not depracated. The change was merely done because m_getclr() sounds too much like "m_get a cluster." - TEMPORARILY disabled mbtypes statistics displaying in netstat(1) and systat(1) (see TODO below). - Fixed systat(1) to display number of "free mbufs" based on new per-CPU stat structures. - Fixed netstat(1) to display new per-CPU stats based on sysctl-exported per-CPU stat structures. All infos are fetched via sysctl. TODO (in order of priority): - Re-enable mbtypes statistics in both netstat(1) and systat(1) after introducing an SMP friendly way to collect the mbtypes stats under the already introduced per-CPU locks (i.e. hopefully don't use atomic() - it seems too costly for a mere stat update, especially when other locks are already present). - Optionally have systat(1) display not only "total free mbufs" but also "total free mbufs per CPU pool." - Fix minor length-fetching issues in netstat(1) related to recently re-enabled option to read mbuf stats from a core file. - Move reference counters at least for mbuf clusters into an unused portion of the cluster itself, to save space and need to allocate a counter. - Look into introducing resource freeing possibly from a kproc. Reviewed by (in parts): jlemon, jake, silby, terry Tested by: jlemon (Intel & Alpha), mjacob (Intel & Alpha) Preliminary performance measurements: jlemon (and me, obviously) URL: http://people.freebsd.org/~bmilekic/mb_alloc/
2001-06-22 06:35:32 +00:00
if (map == kmem_map)
splx(s);
1994-05-24 10:09:53 +00:00
return (KERN_NO_SPACE);
}
start = *addr;
}
result = vm_map_insert(map, object, offset,
start, start + length, prot, max, cow);
1994-05-24 10:09:53 +00:00
vm_map_unlock(map);
Introduce numerous SMP friendly changes to the mbuf allocator. Namely, introduce a modified allocation mechanism for mbufs and mbuf clusters; one which can scale under SMP and which offers the possibility of resource reclamation to be implemented in the future. Notable advantages: o Reduce contention for SMP by offering per-CPU pools and locks. o Better use of data cache due to per-CPU pools. o Much less code cache pollution due to excessively large allocation macros. o Framework for `grouping' objects from same page together so as to be able to possibly free wired-down pages back to the system if they are no longer needed by the network stacks. Additional things changed with this addition: - Moved some mbuf specific declarations and initializations from sys/conf/param.c into mbuf-specific code where they belong. - m_getclr() has been renamed to m_get_clrd() because the old name is really confusing. m_getclr() HAS been preserved though and is defined to the new name. No tree sweep has been done "to change the interface," as the old name will continue to be supported and is not depracated. The change was merely done because m_getclr() sounds too much like "m_get a cluster." - TEMPORARILY disabled mbtypes statistics displaying in netstat(1) and systat(1) (see TODO below). - Fixed systat(1) to display number of "free mbufs" based on new per-CPU stat structures. - Fixed netstat(1) to display new per-CPU stats based on sysctl-exported per-CPU stat structures. All infos are fetched via sysctl. TODO (in order of priority): - Re-enable mbtypes statistics in both netstat(1) and systat(1) after introducing an SMP friendly way to collect the mbtypes stats under the already introduced per-CPU locks (i.e. hopefully don't use atomic() - it seems too costly for a mere stat update, especially when other locks are already present). - Optionally have systat(1) display not only "total free mbufs" but also "total free mbufs per CPU pool." - Fix minor length-fetching issues in netstat(1) related to recently re-enabled option to read mbuf stats from a core file. - Move reference counters at least for mbuf clusters into an unused portion of the cluster itself, to save space and need to allocate a counter. - Look into introducing resource freeing possibly from a kproc. Reviewed by (in parts): jlemon, jake, silby, terry Tested by: jlemon (Intel & Alpha), mjacob (Intel & Alpha) Preliminary performance measurements: jlemon (and me, obviously) URL: http://people.freebsd.org/~bmilekic/mb_alloc/
2001-06-22 06:35:32 +00:00
if (map == kmem_map)
splx(s);
1994-05-24 10:09:53 +00:00
return (result);
}
/*
* vm_map_simplify_entry:
*
* Simplify the given map entry by merging with either neighbor. This
* routine also has the ability to merge with both neighbors.
*
* The map must be locked.
*
* This routine guarentees that the passed entry remains valid (though
* possibly extended). When merging, this routine may delete one or
* both neighbors.
1994-05-24 10:09:53 +00:00
*/
void
vm_map_simplify_entry(vm_map_t map, vm_map_entry_t entry)
1994-05-24 10:09:53 +00:00
{
vm_map_entry_t next, prev;
vm_size_t prevsize, esize;
if (entry->eflags & (MAP_ENTRY_IN_TRANSITION | MAP_ENTRY_IS_SUB_MAP))
1994-05-24 10:09:53 +00:00
return;
prev = entry->prev;
if (prev != &map->header) {
prevsize = prev->end - prev->start;
if ( (prev->end == entry->start) &&
(prev->object.vm_object == entry->object.vm_object) &&
(!prev->object.vm_object ||
(prev->offset + prevsize == entry->offset)) &&
(prev->eflags == entry->eflags) &&
(prev->protection == entry->protection) &&
(prev->max_protection == entry->max_protection) &&
(prev->inheritance == entry->inheritance) &&
(prev->wired_count == entry->wired_count)) {
if (map->first_free == prev)
map->first_free = entry;
vm_map_entry_unlink(map, prev);
entry->start = prev->start;
entry->offset = prev->offset;
This set of commits to the VM system does the following, and contain contributions or ideas from Stephen McKay <syssgm@devetir.qld.gov.au>, Alan Cox <alc@cs.rice.edu>, David Greenman <davidg@freebsd.org> and me: More usage of the TAILQ macros. Additional minor fix to queue.h. Performance enhancements to the pageout daemon. Addition of a wait in the case that the pageout daemon has to run immediately. Slightly modify the pageout algorithm. Significant revamp of the pmap/fork code: 1) PTE's and UPAGES's are NO LONGER in the process's map. 2) PTE's and UPAGES's reside in their own objects. 3) TOTAL elimination of recursive page table pagefaults. 4) The page directory now resides in the PTE object. 5) Implemented pmap_copy, thereby speeding up fork time. 6) Changed the pv entries so that the head is a pointer and not an entire entry. 7) Significant cleanup of pmap_protect, and pmap_remove. 8) Removed significant amounts of machine dependent fork code from vm_glue. Pushed much of that code into the machine dependent pmap module. 9) Support more completely the reuse of already zeroed pages (Page table pages and page directories) as being already zeroed. Performance and code cleanups in vm_map: 1) Improved and simplified allocation of map entries. 2) Improved vm_map_copy code. 3) Corrected some minor problems in the simplify code. Implemented splvm (combo of splbio and splimp.) The VM code now seldom uses splhigh. Improved the speed of and simplified kmem_malloc. Minor mod to vm_fault to avoid using pre-zeroed pages in the case of objects with backing objects along with the already existant condition of having a vnode. (If there is a backing object, there will likely be a COW... With a COW, it isn't necessary to start with a pre-zeroed page.) Minor reorg of source to perhaps improve locality of ref.
1996-05-18 03:38:05 +00:00
if (prev->object.vm_object)
vm_object_deallocate(prev->object.vm_object);
vm_map_entry_dispose(map, prev);
}
}
next = entry->next;
if (next != &map->header) {
esize = entry->end - entry->start;
if ((entry->end == next->start) &&
(next->object.vm_object == entry->object.vm_object) &&
(!entry->object.vm_object ||
(entry->offset + esize == next->offset)) &&
(next->eflags == entry->eflags) &&
(next->protection == entry->protection) &&
(next->max_protection == entry->max_protection) &&
(next->inheritance == entry->inheritance) &&
(next->wired_count == entry->wired_count)) {
if (map->first_free == next)
map->first_free = entry;
vm_map_entry_unlink(map, next);
entry->end = next->end;
This set of commits to the VM system does the following, and contain contributions or ideas from Stephen McKay <syssgm@devetir.qld.gov.au>, Alan Cox <alc@cs.rice.edu>, David Greenman <davidg@freebsd.org> and me: More usage of the TAILQ macros. Additional minor fix to queue.h. Performance enhancements to the pageout daemon. Addition of a wait in the case that the pageout daemon has to run immediately. Slightly modify the pageout algorithm. Significant revamp of the pmap/fork code: 1) PTE's and UPAGES's are NO LONGER in the process's map. 2) PTE's and UPAGES's reside in their own objects. 3) TOTAL elimination of recursive page table pagefaults. 4) The page directory now resides in the PTE object. 5) Implemented pmap_copy, thereby speeding up fork time. 6) Changed the pv entries so that the head is a pointer and not an entire entry. 7) Significant cleanup of pmap_protect, and pmap_remove. 8) Removed significant amounts of machine dependent fork code from vm_glue. Pushed much of that code into the machine dependent pmap module. 9) Support more completely the reuse of already zeroed pages (Page table pages and page directories) as being already zeroed. Performance and code cleanups in vm_map: 1) Improved and simplified allocation of map entries. 2) Improved vm_map_copy code. 3) Corrected some minor problems in the simplify code. Implemented splvm (combo of splbio and splimp.) The VM code now seldom uses splhigh. Improved the speed of and simplified kmem_malloc. Minor mod to vm_fault to avoid using pre-zeroed pages in the case of objects with backing objects along with the already existant condition of having a vnode. (If there is a backing object, there will likely be a COW... With a COW, it isn't necessary to start with a pre-zeroed page.) Minor reorg of source to perhaps improve locality of ref.
1996-05-18 03:38:05 +00:00
if (next->object.vm_object)
vm_object_deallocate(next->object.vm_object);
vm_map_entry_dispose(map, next);
2003-11-03 16:14:45 +00:00
}
1994-05-24 10:09:53 +00:00
}
}
/*
* vm_map_clip_start: [ internal use only ]
*
* Asserts that the given entry begins at or after
* the specified address; if necessary,
* it splits the entry into two.
*/
#define vm_map_clip_start(map, entry, startaddr) \
{ \
if (startaddr > entry->start) \
_vm_map_clip_start(map, entry, startaddr); \
}
/*
* This routine is called only when it is known that
* the entry must be split.
*/
1995-05-30 08:16:23 +00:00
static void
_vm_map_clip_start(vm_map_t map, vm_map_entry_t entry, vm_offset_t start)
1994-05-24 10:09:53 +00:00
{
vm_map_entry_t new_entry;
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
* Split off the front portion -- note that we must insert the new
* entry BEFORE this one, so that this entry has the specified
* starting address.
1994-05-24 10:09:53 +00:00
*/
vm_map_simplify_entry(map, entry);
/*
* If there is no object backing this entry, we might as well create
* one now. If we defer it, an object can get created after the map
* is clipped, and individual objects will be created for the split-up
* map. This is a bit of a hack, but is also about the best place to
* put this improvement.
*/
if (entry->object.vm_object == NULL && !map->system_map) {
vm_object_t object;
object = vm_object_allocate(OBJT_DEFAULT,
atop(entry->end - entry->start));
entry->object.vm_object = object;
entry->offset = 0;
}
1994-05-24 10:09:53 +00:00
new_entry = vm_map_entry_create(map);
*new_entry = *entry;
new_entry->end = start;
entry->offset += (start - entry->start);
entry->start = start;
vm_map_entry_link(map, entry->prev, new_entry);
if ((entry->eflags & MAP_ENTRY_IS_SUB_MAP) == 0) {
1994-05-24 10:09:53 +00:00
vm_object_reference(new_entry->object.vm_object);
}
1994-05-24 10:09:53 +00:00
}
/*
* vm_map_clip_end: [ internal use only ]
*
* Asserts that the given entry ends at or before
* the specified address; if necessary,
* it splits the entry into two.
*/
#define vm_map_clip_end(map, entry, endaddr) \
{ \
if ((endaddr) < (entry->end)) \
_vm_map_clip_end((map), (entry), (endaddr)); \
1994-05-24 10:09:53 +00:00
}
/*
* This routine is called only when it is known that
* the entry must be split.
*/
1995-05-30 08:16:23 +00:00
static void
_vm_map_clip_end(vm_map_t map, vm_map_entry_t entry, vm_offset_t end)
1994-05-24 10:09:53 +00:00
{
vm_map_entry_t new_entry;
1994-05-24 10:09:53 +00:00
/*
* If there is no object backing this entry, we might as well create
* one now. If we defer it, an object can get created after the map
* is clipped, and individual objects will be created for the split-up
* map. This is a bit of a hack, but is also about the best place to
* put this improvement.
*/
if (entry->object.vm_object == NULL && !map->system_map) {
vm_object_t object;
object = vm_object_allocate(OBJT_DEFAULT,
atop(entry->end - entry->start));
entry->object.vm_object = object;
entry->offset = 0;
}
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
* Create a new entry and insert it AFTER the specified entry
1994-05-24 10:09:53 +00:00
*/
new_entry = vm_map_entry_create(map);
*new_entry = *entry;
new_entry->start = entry->end = end;
new_entry->offset += (end - entry->start);
vm_map_entry_link(map, entry, new_entry);
if ((entry->eflags & MAP_ENTRY_IS_SUB_MAP) == 0) {
1994-05-24 10:09:53 +00:00
vm_object_reference(new_entry->object.vm_object);
}
1994-05-24 10:09:53 +00:00
}
/*
* VM_MAP_RANGE_CHECK: [ internal use only ]
*
* Asserts that the starting and ending region
* addresses fall within the valid range of the map.
*/
#define VM_MAP_RANGE_CHECK(map, start, end) \
{ \
if (start < vm_map_min(map)) \
start = vm_map_min(map); \
if (end > vm_map_max(map)) \
end = vm_map_max(map); \
if (start > end) \
start = end; \
}
/*
* vm_map_submap: [ kernel use only ]
*
* Mark the given range as handled by a subordinate map.
*
* This range must have been created with vm_map_find,
* and no other operations may have been performed on this
* range prior to calling vm_map_submap.
*
* Only a limited number of operations can be performed
* within this rage after calling vm_map_submap:
* vm_fault
* [Don't try vm_map_copy!]
*
* To remove a submapping, one must first remove the
* range from the superior map, and then destroy the
* submap (if desired). [Better yet, don't try it.]
*/
int
vm_map_submap(
vm_map_t map,
vm_offset_t start,
vm_offset_t end,
vm_map_t submap)
1994-05-24 10:09:53 +00:00
{
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
vm_map_entry_t entry;
int result = KERN_INVALID_ARGUMENT;
1994-05-24 10:09:53 +00:00
vm_map_lock(map);
VM_MAP_RANGE_CHECK(map, start, end);
if (vm_map_lookup_entry(map, start, &entry)) {
vm_map_clip_start(map, entry, start);
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
} else
1994-05-24 10:09:53 +00:00
entry = entry->next;
vm_map_clip_end(map, entry, end);
if ((entry->start == start) && (entry->end == end) &&
((entry->eflags & MAP_ENTRY_COW) == 0) &&
(entry->object.vm_object == NULL)) {
VM level code cleanups. 1) Start using TSM. Struct procs continue to point to upages structure, after being freed. Struct vmspace continues to point to pte object and kva space for kstack. u_map is now superfluous. 2) vm_map's don't need to be reference counted. They always exist either in the kernel or in a vmspace. The vmspaces are managed by reference counts. 3) Remove the "wired" vm_map nonsense. 4) No need to keep a cache of kernel stack kva's. 5) Get rid of strange looking ++var, and change to var++. 6) Change more data structures to use our "zone" allocator. Added struct proc, struct vmspace and struct vnode. This saves a significant amount of kva space and physical memory. Additionally, this enables TSM for the zone managed memory. 7) Keep ioopt disabled for now. 8) Remove the now bogus "single use" map concept. 9) Use generation counts or id's for data structures residing in TSM, where it allows us to avoid unneeded restart overhead during traversals, where blocking might occur. 10) Account better for memory deficits, so the pageout daemon will be able to make enough memory available (experimental.) 11) Fix some vnode locking problems. (From Tor, I think.) 12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp. (experimental.) 13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c code. Use generation counts, get rid of unneded collpase operations, and clean up the cluster code. 14) Make vm_zone more suitable for TSM. This commit is partially as a result of discussions and contributions from other people, including DG, Tor Egge, PHK, and probably others that I have forgotten to attribute (so let me know, if I forgot.) This is not the infamous, final cleanup of the vnode stuff, but a necessary step. Vnode mgmt should be correct, but things might still change, and there is still some missing stuff (like ioopt, and physical backing of non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
entry->object.sub_map = submap;
entry->eflags |= MAP_ENTRY_IS_SUB_MAP;
1994-05-24 10:09:53 +00:00
result = KERN_SUCCESS;
}
vm_map_unlock(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 (result);
1994-05-24 10:09:53 +00:00
}
/*
* The maximum number of pages to map
*/
#define MAX_INIT_PT 96
/*
* vm_map_pmap_enter:
*
* Preload read-only mappings for the given object into the specified
* map. This eliminates the soft faults on process startup and
* immediately after an mmap(2).
*/
void
vm_map_pmap_enter(vm_map_t map, vm_offset_t addr, vm_prot_t prot,
vm_object_t object, vm_pindex_t pindex, vm_size_t size, int flags)
{
vm_offset_t tmpidx;
int psize;
vm_page_t p, mpte;
if ((prot & VM_PROT_READ) == 0 || object == NULL)
return;
VM_OBJECT_LOCK(object);
if (object->type == OBJT_DEVICE) {
pmap_object_init_pt(map->pmap, addr, object, pindex, size);
goto unlock_return;
}
psize = atop(size);
if (object->type != OBJT_VNODE ||
((flags & MAP_PREFAULT_PARTIAL) && (psize > MAX_INIT_PT) &&
(object->resident_page_count > MAX_INIT_PT))) {
goto unlock_return;
}
if (psize + pindex > object->size) {
if (object->size < pindex)
goto unlock_return;
psize = object->size - pindex;
}
mpte = NULL;
if ((p = TAILQ_FIRST(&object->memq)) != NULL) {
if (p->pindex < pindex) {
p = vm_page_splay(pindex, object->root);
if ((object->root = p)->pindex < pindex)
p = TAILQ_NEXT(p, listq);
}
}
/*
* Assert: the variable p is either (1) the page with the
* least pindex greater than or equal to the parameter pindex
* or (2) NULL.
*/
for (;
p != NULL && (tmpidx = p->pindex - pindex) < psize;
p = TAILQ_NEXT(p, listq)) {
/*
* don't allow an madvise to blow away our really
* free pages allocating pv entries.
*/
if ((flags & MAP_PREFAULT_MADVISE) &&
cnt.v_free_count < cnt.v_free_reserved) {
break;
}
vm_page_lock_queues();
if ((p->valid & VM_PAGE_BITS_ALL) == VM_PAGE_BITS_ALL &&
(p->busy == 0) &&
(p->flags & (PG_BUSY | PG_FICTITIOUS)) == 0) {
if ((p->queue - p->pc) == PQ_CACHE)
vm_page_deactivate(p);
vm_page_busy(p);
vm_page_unlock_queues();
VM_OBJECT_UNLOCK(object);
2003-11-03 16:14:45 +00:00
mpte = pmap_enter_quick(map->pmap,
addr + ptoa(tmpidx), p, mpte);
VM_OBJECT_LOCK(object);
vm_page_lock_queues();
vm_page_wakeup(p);
}
vm_page_unlock_queues();
}
unlock_return:
VM_OBJECT_UNLOCK(object);
}
1994-05-24 10:09:53 +00:00
/*
* vm_map_protect:
*
* Sets the protection of the specified address
* region in the target map. If "set_max" is
* specified, the maximum protection is to be set;
* otherwise, only the current protection is affected.
*/
int
vm_map_protect(vm_map_t map, vm_offset_t start, vm_offset_t end,
vm_prot_t new_prot, boolean_t set_max)
1994-05-24 10:09:53 +00:00
{
vm_map_entry_t current;
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_map_entry_t entry;
1994-05-24 10:09:53 +00:00
vm_map_lock(map);
VM_MAP_RANGE_CHECK(map, start, end);
if (vm_map_lookup_entry(map, start, &entry)) {
vm_map_clip_start(map, entry, start);
} else {
1994-05-24 10:09:53 +00:00
entry = entry->next;
}
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
* Make a first pass to check for protection violations.
1994-05-24 10:09:53 +00:00
*/
current = entry;
while ((current != &map->header) && (current->start < end)) {
if (current->eflags & MAP_ENTRY_IS_SUB_MAP) {
swap_pager.c: Fixed long standing bug in freeing swap space during object collapses. Fixed 'out of space' messages from printing out too often. Modified to use new kmem_malloc() calling convention. Implemented an additional stat in the swap pager struct to count the amount of space allocated to that pager. This may be removed at some point in the future. Minimized unnecessary wakeups. vm_fault.c: Don't try to collect fault stats on 'swapped' processes - there aren't any upages to store the stats in. Changed read-ahead policy (again!). vm_glue.c: Be sure to gain a reference to the process's map before swapping. Be sure to lose it when done. kern_malloc.c: Added the ability to specify if allocations are at interrupt time or are 'safe'; this affects what types of pages can be allocated. vm_map.c: Fixed a variety of map lock problems; there's still a lurking bug that will eventually bite. vm_object.c: Explicitly initialize the object fields rather than bzeroing the struct. Eliminated the 'rcollapse' code and folded it's functionality into the "real" collapse routine. Moved an object_unlock() so that the backing_object is protected in the qcollapse routine. Make sure nobody fools with the backing_object when we're destroying it. Added some diagnostic code which can be called from the debugger that looks through all the internal objects and makes certain that they all belong to someone. vm_page.c: Fixed a rather serious logic bug that would result in random system crashes. Changed pagedaemon wakeup policy (again!). vm_pageout.c: Removed unnecessary page rotations on the inactive queue. Changed the number of pages to explicitly free to just free_reserved level. Submitted by: John Dyson
1995-02-02 09:09:15 +00:00
vm_map_unlock(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 (KERN_INVALID_ARGUMENT);
swap_pager.c: Fixed long standing bug in freeing swap space during object collapses. Fixed 'out of space' messages from printing out too often. Modified to use new kmem_malloc() calling convention. Implemented an additional stat in the swap pager struct to count the amount of space allocated to that pager. This may be removed at some point in the future. Minimized unnecessary wakeups. vm_fault.c: Don't try to collect fault stats on 'swapped' processes - there aren't any upages to store the stats in. Changed read-ahead policy (again!). vm_glue.c: Be sure to gain a reference to the process's map before swapping. Be sure to lose it when done. kern_malloc.c: Added the ability to specify if allocations are at interrupt time or are 'safe'; this affects what types of pages can be allocated. vm_map.c: Fixed a variety of map lock problems; there's still a lurking bug that will eventually bite. vm_object.c: Explicitly initialize the object fields rather than bzeroing the struct. Eliminated the 'rcollapse' code and folded it's functionality into the "real" collapse routine. Moved an object_unlock() so that the backing_object is protected in the qcollapse routine. Make sure nobody fools with the backing_object when we're destroying it. Added some diagnostic code which can be called from the debugger that looks through all the internal objects and makes certain that they all belong to someone. vm_page.c: Fixed a rather serious logic bug that would result in random system crashes. Changed pagedaemon wakeup policy (again!). vm_pageout.c: Removed unnecessary page rotations on the inactive queue. Changed the number of pages to explicitly free to just free_reserved level. Submitted by: John Dyson
1995-02-02 09:09:15 +00:00
}
1994-05-24 10:09:53 +00:00
if ((new_prot & current->max_protection) != new_prot) {
vm_map_unlock(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 (KERN_PROTECTION_FAILURE);
1994-05-24 10:09:53 +00:00
}
current = current->next;
}
/*
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
* Go back and fix up protections. [Note that clipping is not
* necessary the second time.]
1994-05-24 10:09:53 +00:00
*/
current = entry;
while ((current != &map->header) && (current->start < end)) {
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
vm_prot_t old_prot;
1994-05-24 10:09:53 +00:00
vm_map_clip_end(map, current, end);
old_prot = current->protection;
if (set_max)
current->protection =
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
(current->max_protection = new_prot) &
old_prot;
1994-05-24 10:09:53 +00:00
else
current->protection = new_prot;
/*
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
* Update physical map if necessary. Worry about copy-on-write
* here -- CHECK THIS XXX
1994-05-24 10:09:53 +00:00
*/
if (current->protection != old_prot) {
#define MASK(entry) (((entry)->eflags & MAP_ENTRY_COW) ? ~VM_PROT_WRITE : \
1994-05-24 10:09:53 +00:00
VM_PROT_ALL)
pmap_protect(map->pmap, current->start,
current->end,
current->protection & MASK(current));
1994-05-24 10:09:53 +00:00
#undef MASK
}
vm_map_simplify_entry(map, current);
1994-05-24 10:09:53 +00:00
current = current->next;
}
vm_map_unlock(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 (KERN_SUCCESS);
1994-05-24 10:09:53 +00:00
}
/*
* vm_map_madvise:
*
2003-11-03 16:14:45 +00:00
* This routine traverses a processes map handling the madvise
* system call. Advisories are classified as either those effecting
2003-11-03 16:14:45 +00:00
* the vm_map_entry structure, or those effecting the underlying
* objects.
*/
int
vm_map_madvise(
vm_map_t map,
2003-11-03 16:14:45 +00:00
vm_offset_t start,
vm_offset_t end,
int behav)
{
vm_map_entry_t current, entry;
int modify_map = 0;
/*
* Some madvise calls directly modify the vm_map_entry, in which case
2003-11-03 16:14:45 +00:00
* we need to use an exclusive lock on the map and we need to perform
* various clipping operations. Otherwise we only need a read-lock
* on the map.
*/
switch(behav) {
case MADV_NORMAL:
case MADV_SEQUENTIAL:
case MADV_RANDOM:
case MADV_NOSYNC:
case MADV_AUTOSYNC:
case MADV_NOCORE:
case MADV_CORE:
modify_map = 1;
vm_map_lock(map);
break;
case MADV_WILLNEED:
case MADV_DONTNEED:
case MADV_FREE:
vm_map_lock_read(map);
break;
default:
return (KERN_INVALID_ARGUMENT);
}
/*
* Locate starting entry and clip if necessary.
*/
VM_MAP_RANGE_CHECK(map, start, end);
if (vm_map_lookup_entry(map, start, &entry)) {
if (modify_map)
vm_map_clip_start(map, entry, start);
} else {
entry = entry->next;
}
if (modify_map) {
/*
* madvise behaviors that are implemented in the vm_map_entry.
*
* We clip the vm_map_entry so that behavioral changes are
* limited to the specified address range.
*/
for (current = entry;
(current != &map->header) && (current->start < end);
current = current->next
) {
if (current->eflags & MAP_ENTRY_IS_SUB_MAP)
continue;
vm_map_clip_end(map, current, end);
switch (behav) {
case MADV_NORMAL:
vm_map_entry_set_behavior(current, MAP_ENTRY_BEHAV_NORMAL);
break;
case MADV_SEQUENTIAL:
vm_map_entry_set_behavior(current, MAP_ENTRY_BEHAV_SEQUENTIAL);
break;
case MADV_RANDOM:
vm_map_entry_set_behavior(current, MAP_ENTRY_BEHAV_RANDOM);
break;
case MADV_NOSYNC:
current->eflags |= MAP_ENTRY_NOSYNC;
break;
case MADV_AUTOSYNC:
current->eflags &= ~MAP_ENTRY_NOSYNC;
break;
case MADV_NOCORE:
current->eflags |= MAP_ENTRY_NOCOREDUMP;
break;
case MADV_CORE:
current->eflags &= ~MAP_ENTRY_NOCOREDUMP;
break;
default:
break;
}
vm_map_simplify_entry(map, current);
}
vm_map_unlock(map);
} else {
vm_pindex_t pindex;
int count;
/*
* madvise behaviors that are implemented in the underlying
* vm_object.
*
* Since we don't clip the vm_map_entry, we have to clip
* the vm_object pindex and count.
*/
for (current = entry;
(current != &map->header) && (current->start < end);
current = current->next
) {
vm_offset_t useStart;
if (current->eflags & MAP_ENTRY_IS_SUB_MAP)
continue;
pindex = OFF_TO_IDX(current->offset);
count = atop(current->end - current->start);
useStart = current->start;
if (current->start < start) {
pindex += atop(start - current->start);
count -= atop(start - current->start);
useStart = start;
}
if (current->end > end)
count -= atop(current->end - end);
if (count <= 0)
continue;
vm_object_madvise(current->object.vm_object,
pindex, count, behav);
if (behav == MADV_WILLNEED) {
2003-11-03 16:14:45 +00:00
vm_map_pmap_enter(map,
useStart,
current->protection,
current->object.vm_object,
2003-11-03 16:14:45 +00:00
pindex,
(count << PAGE_SHIFT),
MAP_PREFAULT_MADVISE
);
}
}
vm_map_unlock_read(map);
}
return (0);
2003-11-03 16:14:45 +00:00
}
1994-05-24 10:09:53 +00:00
/*
* vm_map_inherit:
*
* Sets the inheritance of the specified address
* range in the target map. Inheritance
* affects how the map will be shared with
* child maps at the time of vm_map_fork.
*/
int
vm_map_inherit(vm_map_t map, vm_offset_t start, vm_offset_t end,
vm_inherit_t new_inheritance)
1994-05-24 10:09:53 +00:00
{
vm_map_entry_t entry;
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_map_entry_t temp_entry;
1994-05-24 10:09:53 +00:00
switch (new_inheritance) {
case VM_INHERIT_NONE:
case VM_INHERIT_COPY:
case VM_INHERIT_SHARE:
break;
default:
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 (KERN_INVALID_ARGUMENT);
1994-05-24 10:09:53 +00:00
}
vm_map_lock(map);
VM_MAP_RANGE_CHECK(map, start, end);
if (vm_map_lookup_entry(map, start, &temp_entry)) {
entry = temp_entry;
vm_map_clip_start(map, entry, start);
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
} else
1994-05-24 10:09:53 +00:00
entry = temp_entry->next;
while ((entry != &map->header) && (entry->start < end)) {
vm_map_clip_end(map, entry, end);
entry->inheritance = new_inheritance;
vm_map_simplify_entry(map, entry);
1994-05-24 10:09:53 +00:00
entry = entry->next;
}
vm_map_unlock(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 (KERN_SUCCESS);
1994-05-24 10:09:53 +00:00
}
/*
* vm_map_unwire:
*
* Implements both kernel and user unwiring.
*
* Ignores MAP_NOFAULT (wired_count == 0 is okay) for kernel unwiring.
*/
int
vm_map_unwire(vm_map_t map, vm_offset_t start, vm_offset_t end,
int flags)
{
vm_map_entry_t entry, first_entry, tmp_entry;
vm_offset_t saved_start;
unsigned int last_timestamp;
int rv;
boolean_t need_wakeup, result, user_unwire;
user_unwire = (flags & VM_MAP_WIRE_USER) ? TRUE : FALSE;
vm_map_lock(map);
VM_MAP_RANGE_CHECK(map, start, end);
if (!vm_map_lookup_entry(map, start, &first_entry)) {
if (flags & VM_MAP_WIRE_HOLESOK)
first_entry = first_entry->next;
else {
vm_map_unlock(map);
return (KERN_INVALID_ADDRESS);
}
}
last_timestamp = map->timestamp;
entry = first_entry;
while (entry != &map->header && entry->start < end) {
if (entry->eflags & MAP_ENTRY_IN_TRANSITION) {
/*
* We have not yet clipped the entry.
*/
saved_start = (start >= entry->start) ? start :
entry->start;
entry->eflags |= MAP_ENTRY_NEEDS_WAKEUP;
if (vm_map_unlock_and_wait(map, user_unwire)) {
/*
* Allow interruption of user unwiring?
*/
}
vm_map_lock(map);
if (last_timestamp+1 != map->timestamp) {
/*
* Look again for the entry because the map was
* modified while it was unlocked.
* Specifically, the entry may have been
* clipped, merged, or deleted.
*/
if (!vm_map_lookup_entry(map, saved_start,
&tmp_entry)) {
if (flags & VM_MAP_WIRE_HOLESOK)
tmp_entry = tmp_entry->next;
else {
if (saved_start == start) {
/*
* First_entry has been deleted.
*/
vm_map_unlock(map);
return (KERN_INVALID_ADDRESS);
}
end = saved_start;
rv = KERN_INVALID_ADDRESS;
goto done;
}
}
if (entry == first_entry)
first_entry = tmp_entry;
else
first_entry = NULL;
entry = tmp_entry;
}
last_timestamp = map->timestamp;
continue;
}
vm_map_clip_start(map, entry, start);
vm_map_clip_end(map, entry, end);
/*
* Mark the entry in case the map lock is released. (See
* above.)
*/
entry->eflags |= MAP_ENTRY_IN_TRANSITION;
/*
* Check the map for holes in the specified region.
* If VM_MAP_WIRE_HOLESOK was specified, skip this check.
*/
if (((flags & VM_MAP_WIRE_HOLESOK) == 0) &&
(entry->end < end && (entry->next == &map->header ||
entry->next->start > entry->end))) {
end = entry->end;
rv = KERN_INVALID_ADDRESS;
goto done;
}
/*
* If system unwiring, require that the entry is system wired.
*/
if (user_unwire ? vm_map_entry_user_wired_count(entry) == 0 :
(vm_map_entry_system_wired_count(entry) == 0 &&
(entry->eflags & MAP_ENTRY_NOFAULT) == 0)) {
end = entry->end;
rv = KERN_INVALID_ARGUMENT;
goto done;
}
entry = entry->next;
}
rv = KERN_SUCCESS;
done:
need_wakeup = FALSE;
if (first_entry == NULL) {
result = vm_map_lookup_entry(map, start, &first_entry);
if (!result && (flags & VM_MAP_WIRE_HOLESOK))
first_entry = first_entry->next;
else
KASSERT(result, ("vm_map_unwire: lookup failed"));
}
entry = first_entry;
while (entry != &map->header && entry->start < end) {
if (rv == KERN_SUCCESS && (!user_unwire ||
(entry->eflags & MAP_ENTRY_USER_WIRED))) {
if (user_unwire)
entry->eflags &= ~MAP_ENTRY_USER_WIRED;
if (user_unwire ||
(entry->eflags & MAP_ENTRY_NOFAULT) == 0)
entry->wired_count--;
if (entry->wired_count == 0 &&
(entry->eflags & MAP_ENTRY_NOFAULT) == 0) {
/*
* Retain the map lock.
*/
vm_fault_unwire(map, entry->start, entry->end,
entry->object.vm_object != NULL &&
entry->object.vm_object->type == OBJT_DEVICE);
}
}
KASSERT(entry->eflags & MAP_ENTRY_IN_TRANSITION,
("vm_map_unwire: in-transition flag missing"));
entry->eflags &= ~MAP_ENTRY_IN_TRANSITION;
if (entry->eflags & MAP_ENTRY_NEEDS_WAKEUP) {
entry->eflags &= ~MAP_ENTRY_NEEDS_WAKEUP;
need_wakeup = TRUE;
}
vm_map_simplify_entry(map, entry);
entry = entry->next;
}
vm_map_unlock(map);
if (need_wakeup)
vm_map_wakeup(map);
return (rv);
}
/*
* vm_map_wire:
*
* Implements both kernel and user wiring.
*/
int
vm_map_wire(vm_map_t map, vm_offset_t start, vm_offset_t end,
int flags)
{
vm_map_entry_t entry, first_entry, tmp_entry;
vm_offset_t saved_end, saved_start;
unsigned int last_timestamp;
int rv;
boolean_t fictitious, need_wakeup, result, user_wire;
user_wire = (flags & VM_MAP_WIRE_USER) ? TRUE : FALSE;
vm_map_lock(map);
VM_MAP_RANGE_CHECK(map, start, end);
if (!vm_map_lookup_entry(map, start, &first_entry)) {
if (flags & VM_MAP_WIRE_HOLESOK)
first_entry = first_entry->next;
else {
vm_map_unlock(map);
return (KERN_INVALID_ADDRESS);
}
}
last_timestamp = map->timestamp;
entry = first_entry;
while (entry != &map->header && entry->start < end) {
if (entry->eflags & MAP_ENTRY_IN_TRANSITION) {
/*
* We have not yet clipped the entry.
*/
saved_start = (start >= entry->start) ? start :
entry->start;
entry->eflags |= MAP_ENTRY_NEEDS_WAKEUP;
if (vm_map_unlock_and_wait(map, user_wire)) {
/*
* Allow interruption of user wiring?
*/
}
vm_map_lock(map);
if (last_timestamp + 1 != map->timestamp) {
/*
* Look again for the entry because the map was
* modified while it was unlocked.
* Specifically, the entry may have been
* clipped, merged, or deleted.
*/
if (!vm_map_lookup_entry(map, saved_start,
&tmp_entry)) {
if (flags & VM_MAP_WIRE_HOLESOK)
tmp_entry = tmp_entry->next;
else {
if (saved_start == start) {
/*
* first_entry has been deleted.
*/
vm_map_unlock(map);
return (KERN_INVALID_ADDRESS);
}
end = saved_start;
rv = KERN_INVALID_ADDRESS;
goto done;
}
}
if (entry == first_entry)
first_entry = tmp_entry;
else
first_entry = NULL;
entry = tmp_entry;
}
last_timestamp = map->timestamp;
continue;
}
vm_map_clip_start(map, entry, start);
vm_map_clip_end(map, entry, end);
/*
* Mark the entry in case the map lock is released. (See
* above.)
*/
entry->eflags |= MAP_ENTRY_IN_TRANSITION;
/*
* The wired_count keeps track of zero or one user
* wirings. It also keeps track of system wirings if
* MAP_ENTRY_NOFAULT is not set.
*/
if ((user_wire ? vm_map_entry_user_wired_count(entry) == 0 :
(entry->eflags & MAP_ENTRY_NOFAULT) == 0) &&
entry->wired_count++ == 0 &&
(entry->eflags & MAP_ENTRY_NOFAULT) == 0) {
saved_start = entry->start;
saved_end = entry->end;
fictitious = entry->object.vm_object != NULL &&
entry->object.vm_object->type == OBJT_DEVICE;
/*
* Release the map lock, relying on the in-transition
* mark.
*/
vm_map_unlock(map);
rv = vm_fault_wire(map, saved_start, saved_end,
user_wire, fictitious);
vm_map_lock(map);
if (last_timestamp + 1 != map->timestamp) {
/*
* Look again for the entry because the map was
* modified while it was unlocked. The entry
* may have been clipped, but NOT merged or
* deleted.
*/
result = vm_map_lookup_entry(map, saved_start,
&tmp_entry);
KASSERT(result, ("vm_map_wire: lookup failed"));
if (entry == first_entry)
first_entry = tmp_entry;
else
first_entry = NULL;
entry = tmp_entry;
while (entry->end < saved_end) {
if (rv != KERN_SUCCESS) {
KASSERT(entry->wired_count == 1,
("vm_map_wire: bad count"));
entry->wired_count = -1;
}
entry = entry->next;
}
}
last_timestamp = map->timestamp;
if (rv != KERN_SUCCESS) {
KASSERT(entry->wired_count == 1,
("vm_map_wire: bad count"));
/*
* Assign an out-of-range value to represent
* the failure to wire this entry.
*/
entry->wired_count = -1;
end = entry->end;
goto done;
}
}
/*
* Check the map for holes in the specified region.
* If VM_MAP_WIRE_HOLESOK was specified, skip this check.
*/
if (((flags & VM_MAP_WIRE_HOLESOK) == 0) &&
(entry->end < end && (entry->next == &map->header ||
entry->next->start > entry->end))) {
end = entry->end;
rv = KERN_INVALID_ADDRESS;
goto done;
}
entry = entry->next;
}
rv = KERN_SUCCESS;
done:
need_wakeup = FALSE;
if (first_entry == NULL) {
result = vm_map_lookup_entry(map, start, &first_entry);
if (!result && (flags & VM_MAP_WIRE_HOLESOK))
first_entry = first_entry->next;
else
KASSERT(result, ("vm_map_wire: lookup failed"));
}
entry = first_entry;
while (entry != &map->header && entry->start < end) {
if (rv == KERN_SUCCESS) {
/*
* The MAP_ENTRY_USER_WIRED may already have been
* set. If so, this statement has no effect
* (and wired_count will not have changed).
*/
if (user_wire)
entry->eflags |= MAP_ENTRY_USER_WIRED;
} else if (entry->wired_count == -1) {
/*
* Wiring failed on this entry. Thus, unwiring is
* unnecessary.
*/
entry->wired_count = 0;
} else {
if (user_wire ||
(entry->eflags & MAP_ENTRY_NOFAULT) == 0)
entry->wired_count--;
if (entry->wired_count == 0 &&
(entry->eflags & MAP_ENTRY_NOFAULT) == 0) {
/*
* Retain the map lock.
*/
vm_fault_unwire(map, entry->start, entry->end,
entry->object.vm_object != NULL &&
entry->object.vm_object->type == OBJT_DEVICE);
}
}
KASSERT(entry->eflags & MAP_ENTRY_IN_TRANSITION,
("vm_map_wire: in-transition flag missing"));
entry->eflags &= ~MAP_ENTRY_IN_TRANSITION;
if (entry->eflags & MAP_ENTRY_NEEDS_WAKEUP) {
entry->eflags &= ~MAP_ENTRY_NEEDS_WAKEUP;
need_wakeup = TRUE;
}
vm_map_simplify_entry(map, entry);
entry = entry->next;
}
vm_map_unlock(map);
if (need_wakeup)
vm_map_wakeup(map);
return (rv);
}
1994-05-24 10:09:53 +00:00
/*
* vm_map_sync
1994-05-24 10:09:53 +00:00
*
* Push any dirty cached pages in the address range to their pager.
* If syncio is TRUE, dirty pages are written synchronously.
* If invalidate is TRUE, any cached pages are freed as well.
*
* If the size of the region from start to end is zero, we are
* supposed to flush all modified pages within the region containing
* start. Unfortunately, a region can be split or coalesced with
* neighboring regions, making it difficult to determine what the
* original region was. Therefore, we approximate this requirement by
* flushing the current region containing start.
*
1994-05-24 10:09:53 +00:00
* Returns an error if any part of the specified range is not mapped.
*/
int
vm_map_sync(
vm_map_t map,
vm_offset_t start,
vm_offset_t end,
boolean_t syncio,
boolean_t invalidate)
1994-05-24 10:09:53 +00:00
{
vm_map_entry_t current;
1994-05-24 10:09:53 +00:00
vm_map_entry_t entry;
vm_size_t size;
vm_object_t object;
vm_ooffset_t offset;
1994-05-24 10:09:53 +00:00
vm_map_lock_read(map);
VM_MAP_RANGE_CHECK(map, start, end);
if (!vm_map_lookup_entry(map, start, &entry)) {
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 (KERN_INVALID_ADDRESS);
} else if (start == end) {
start = entry->start;
end = entry->end;
1994-05-24 10:09:53 +00:00
}
/*
* Make a first pass to check for user-wired memory and holes.
1994-05-24 10:09:53 +00:00
*/
for (current = entry; current->start < end; current = current->next) {
if (invalidate && (current->eflags & MAP_ENTRY_USER_WIRED)) {
1994-05-24 10:09:53 +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 (KERN_INVALID_ARGUMENT);
1994-05-24 10:09:53 +00:00
}
if (end > current->end &&
(current->next == &map->header ||
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
current->end != current->next->start)) {
1994-05-24 10:09:53 +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 (KERN_INVALID_ADDRESS);
1994-05-24 10:09:53 +00:00
}
}
if (invalidate) {
mtx_lock(&Giant);
pmap_remove(map->pmap, start, end);
mtx_unlock(&Giant);
}
1994-05-24 10:09:53 +00:00
/*
* Make a second pass, cleaning/uncaching pages from the indicated
* objects as we go.
*/
for (current = entry; current->start < end; current = current->next) {
offset = current->offset + (start - current->start);
size = (end <= current->end ? end : current->end) - start;
if (current->eflags & MAP_ENTRY_IS_SUB_MAP) {
vm_map_t smap;
1994-05-24 10:09:53 +00:00
vm_map_entry_t tentry;
vm_size_t tsize;
smap = current->object.sub_map;
1994-05-24 10:09:53 +00:00
vm_map_lock_read(smap);
(void) vm_map_lookup_entry(smap, offset, &tentry);
tsize = tentry->end - offset;
if (tsize < size)
size = tsize;
object = tentry->object.vm_object;
offset = tentry->offset + (offset - tentry->start);
vm_map_unlock_read(smap);
} else {
object = current->object.vm_object;
}
vm_object_sync(object, offset, size, syncio, invalidate);
1994-05-24 10:09:53 +00:00
start += size;
}
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 (KERN_SUCCESS);
1994-05-24 10:09:53 +00:00
}
/*
* vm_map_entry_unwire: [ internal use only ]
*
* Make the region specified by this entry pageable.
*
* The map in question should be locked.
* [This is the reason for this routine's existence.]
*/
2003-11-03 16:14:45 +00:00
static void
vm_map_entry_unwire(vm_map_t map, vm_map_entry_t entry)
1994-05-24 10:09:53 +00:00
{
KASSERT(vm_map_entry_user_wired_count(entry) == 1 &&
vm_map_entry_system_wired_count(entry) == 0,
("system/user wiring mistake"));
vm_fault_unwire(map, entry->start, entry->end,
entry->object.vm_object != NULL &&
entry->object.vm_object->type == OBJT_DEVICE);
1994-05-24 10:09:53 +00:00
entry->wired_count = 0;
}
/*
* vm_map_entry_delete: [ internal use only ]
*
* Deallocate the given entry from the target 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
*/
1996-12-07 07:44:05 +00:00
static void
vm_map_entry_delete(vm_map_t map, vm_map_entry_t entry)
1994-05-24 10:09:53 +00:00
{
vm_object_t object;
vm_pindex_t offidxstart, offidxend, count;
1994-05-24 10:09:53 +00:00
vm_map_entry_unlink(map, entry);
map->size -= entry->end - entry->start;
if ((entry->eflags & MAP_ENTRY_IS_SUB_MAP) == 0 &&
(object = entry->object.vm_object) != NULL) {
count = OFF_TO_IDX(entry->end - entry->start);
offidxstart = OFF_TO_IDX(entry->offset);
offidxend = offidxstart + count;
VM_OBJECT_LOCK(object);
if (object->ref_count != 1 &&
((object->flags & (OBJ_NOSPLIT|OBJ_ONEMAPPING)) == OBJ_ONEMAPPING ||
object == kernel_object || object == kmem_object) &&
(object->type == OBJT_DEFAULT || object->type == OBJT_SWAP)) {
vm_object_collapse(object);
vm_object_page_remove(object, offidxstart, offidxend, FALSE);
if (object->type == OBJT_SWAP)
swap_pager_freespace(object, offidxstart, count);
if (offidxend >= object->size &&
offidxstart < object->size)
object->size = offidxstart;
}
VM_OBJECT_UNLOCK(object);
vm_object_deallocate(object);
}
1994-05-24 10:09:53 +00:00
vm_map_entry_dispose(map, entry);
}
/*
* vm_map_delete: [ internal use only ]
*
* Deallocates the given address range from the target
* map.
*/
int
vm_map_delete(vm_map_t map, vm_offset_t start, vm_offset_t end)
1994-05-24 10:09:53 +00:00
{
vm_map_entry_t entry;
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_map_entry_t first_entry;
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
* Find the start of the region, and clip it
1994-05-24 10:09:53 +00:00
*/
if (!vm_map_lookup_entry(map, start, &first_entry))
1994-05-24 10:09:53 +00:00
entry = first_entry->next;
else {
1994-05-24 10:09:53 +00:00
entry = first_entry;
vm_map_clip_start(map, entry, start);
}
/*
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
* Save the free space hint
1994-05-24 10:09:53 +00:00
*/
This set of commits to the VM system does the following, and contain contributions or ideas from Stephen McKay <syssgm@devetir.qld.gov.au>, Alan Cox <alc@cs.rice.edu>, David Greenman <davidg@freebsd.org> and me: More usage of the TAILQ macros. Additional minor fix to queue.h. Performance enhancements to the pageout daemon. Addition of a wait in the case that the pageout daemon has to run immediately. Slightly modify the pageout algorithm. Significant revamp of the pmap/fork code: 1) PTE's and UPAGES's are NO LONGER in the process's map. 2) PTE's and UPAGES's reside in their own objects. 3) TOTAL elimination of recursive page table pagefaults. 4) The page directory now resides in the PTE object. 5) Implemented pmap_copy, thereby speeding up fork time. 6) Changed the pv entries so that the head is a pointer and not an entire entry. 7) Significant cleanup of pmap_protect, and pmap_remove. 8) Removed significant amounts of machine dependent fork code from vm_glue. Pushed much of that code into the machine dependent pmap module. 9) Support more completely the reuse of already zeroed pages (Page table pages and page directories) as being already zeroed. Performance and code cleanups in vm_map: 1) Improved and simplified allocation of map entries. 2) Improved vm_map_copy code. 3) Corrected some minor problems in the simplify code. Implemented splvm (combo of splbio and splimp.) The VM code now seldom uses splhigh. Improved the speed of and simplified kmem_malloc. Minor mod to vm_fault to avoid using pre-zeroed pages in the case of objects with backing objects along with the already existant condition of having a vnode. (If there is a backing object, there will likely be a COW... With a COW, it isn't necessary to start with a pre-zeroed page.) Minor reorg of source to perhaps improve locality of ref.
1996-05-18 03:38:05 +00:00
if (entry == &map->header) {
map->first_free = &map->header;
} else if (map->first_free->start >= start) {
1994-05-24 10:09:53 +00:00
map->first_free = entry->prev;
}
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
* Step through all entries in this region
1994-05-24 10:09:53 +00:00
*/
while ((entry != &map->header) && (entry->start < end)) {
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_map_entry_t next;
1994-05-24 10:09:53 +00:00
/*
* Wait for wiring or unwiring of an entry to complete.
* Also wait for any system wirings to disappear.
*/
if ((entry->eflags & MAP_ENTRY_IN_TRANSITION) != 0 ||
vm_map_entry_system_wired_count(entry) != 0) {
unsigned int last_timestamp;
vm_offset_t saved_start;
vm_map_entry_t tmp_entry;
saved_start = entry->start;
entry->eflags |= MAP_ENTRY_NEEDS_WAKEUP;
last_timestamp = map->timestamp;
(void) vm_map_unlock_and_wait(map, FALSE);
vm_map_lock(map);
if (last_timestamp + 1 != map->timestamp) {
/*
* Look again for the entry because the map was
* modified while it was unlocked.
* Specifically, the entry may have been
* clipped, merged, or deleted.
*/
if (!vm_map_lookup_entry(map, saved_start,
&tmp_entry))
entry = tmp_entry->next;
else {
entry = tmp_entry;
vm_map_clip_start(map, entry,
saved_start);
}
}
continue;
}
1994-05-24 10:09:53 +00:00
vm_map_clip_end(map, entry, end);
next = entry->next;
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
* Unwire before removing addresses from the pmap; otherwise,
* unwiring will put the entries back in the pmap.
1994-05-24 10:09:53 +00:00
*/
if (entry->wired_count != 0) {
1994-05-24 10:09:53 +00:00
vm_map_entry_unwire(map, entry);
}
1994-05-24 10:09:53 +00:00
if (!map->system_map)
mtx_lock(&Giant);
pmap_remove(map->pmap, entry->start, entry->end);
if (!map->system_map)
mtx_unlock(&Giant);
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
* Delete the entry (which may delete the object) only after
* removing all pmap entries pointing to its pages.
* (Otherwise, its page frames may be reallocated, and any
* modify bits will be set in the wrong object!)
1994-05-24 10:09:53 +00:00
*/
vm_map_entry_delete(map, entry);
entry = next;
}
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 (KERN_SUCCESS);
1994-05-24 10:09:53 +00:00
}
/*
* vm_map_remove:
*
* Remove the given address range from the target map.
* This is the exported form of vm_map_delete.
*/
int
vm_map_remove(vm_map_t map, vm_offset_t start, vm_offset_t end)
1994-05-24 10:09:53 +00:00
{
int result, s = 0;
Introduce numerous SMP friendly changes to the mbuf allocator. Namely, introduce a modified allocation mechanism for mbufs and mbuf clusters; one which can scale under SMP and which offers the possibility of resource reclamation to be implemented in the future. Notable advantages: o Reduce contention for SMP by offering per-CPU pools and locks. o Better use of data cache due to per-CPU pools. o Much less code cache pollution due to excessively large allocation macros. o Framework for `grouping' objects from same page together so as to be able to possibly free wired-down pages back to the system if they are no longer needed by the network stacks. Additional things changed with this addition: - Moved some mbuf specific declarations and initializations from sys/conf/param.c into mbuf-specific code where they belong. - m_getclr() has been renamed to m_get_clrd() because the old name is really confusing. m_getclr() HAS been preserved though and is defined to the new name. No tree sweep has been done "to change the interface," as the old name will continue to be supported and is not depracated. The change was merely done because m_getclr() sounds too much like "m_get a cluster." - TEMPORARILY disabled mbtypes statistics displaying in netstat(1) and systat(1) (see TODO below). - Fixed systat(1) to display number of "free mbufs" based on new per-CPU stat structures. - Fixed netstat(1) to display new per-CPU stats based on sysctl-exported per-CPU stat structures. All infos are fetched via sysctl. TODO (in order of priority): - Re-enable mbtypes statistics in both netstat(1) and systat(1) after introducing an SMP friendly way to collect the mbtypes stats under the already introduced per-CPU locks (i.e. hopefully don't use atomic() - it seems too costly for a mere stat update, especially when other locks are already present). - Optionally have systat(1) display not only "total free mbufs" but also "total free mbufs per CPU pool." - Fix minor length-fetching issues in netstat(1) related to recently re-enabled option to read mbuf stats from a core file. - Move reference counters at least for mbuf clusters into an unused portion of the cluster itself, to save space and need to allocate a counter. - Look into introducing resource freeing possibly from a kproc. Reviewed by (in parts): jlemon, jake, silby, terry Tested by: jlemon (Intel & Alpha), mjacob (Intel & Alpha) Preliminary performance measurements: jlemon (and me, obviously) URL: http://people.freebsd.org/~bmilekic/mb_alloc/
2001-06-22 06:35:32 +00:00
if (map == kmem_map)
This set of commits to the VM system does the following, and contain contributions or ideas from Stephen McKay <syssgm@devetir.qld.gov.au>, Alan Cox <alc@cs.rice.edu>, David Greenman <davidg@freebsd.org> and me: More usage of the TAILQ macros. Additional minor fix to queue.h. Performance enhancements to the pageout daemon. Addition of a wait in the case that the pageout daemon has to run immediately. Slightly modify the pageout algorithm. Significant revamp of the pmap/fork code: 1) PTE's and UPAGES's are NO LONGER in the process's map. 2) PTE's and UPAGES's reside in their own objects. 3) TOTAL elimination of recursive page table pagefaults. 4) The page directory now resides in the PTE object. 5) Implemented pmap_copy, thereby speeding up fork time. 6) Changed the pv entries so that the head is a pointer and not an entire entry. 7) Significant cleanup of pmap_protect, and pmap_remove. 8) Removed significant amounts of machine dependent fork code from vm_glue. Pushed much of that code into the machine dependent pmap module. 9) Support more completely the reuse of already zeroed pages (Page table pages and page directories) as being already zeroed. Performance and code cleanups in vm_map: 1) Improved and simplified allocation of map entries. 2) Improved vm_map_copy code. 3) Corrected some minor problems in the simplify code. Implemented splvm (combo of splbio and splimp.) The VM code now seldom uses splhigh. Improved the speed of and simplified kmem_malloc. Minor mod to vm_fault to avoid using pre-zeroed pages in the case of objects with backing objects along with the already existant condition of having a vnode. (If there is a backing object, there will likely be a COW... With a COW, it isn't necessary to start with a pre-zeroed page.) Minor reorg of source to perhaps improve locality of ref.
1996-05-18 03:38:05 +00:00
s = splvm();
1994-05-24 10:09:53 +00:00
vm_map_lock(map);
VM_MAP_RANGE_CHECK(map, start, end);
result = vm_map_delete(map, start, end);
vm_map_unlock(map);
Introduce numerous SMP friendly changes to the mbuf allocator. Namely, introduce a modified allocation mechanism for mbufs and mbuf clusters; one which can scale under SMP and which offers the possibility of resource reclamation to be implemented in the future. Notable advantages: o Reduce contention for SMP by offering per-CPU pools and locks. o Better use of data cache due to per-CPU pools. o Much less code cache pollution due to excessively large allocation macros. o Framework for `grouping' objects from same page together so as to be able to possibly free wired-down pages back to the system if they are no longer needed by the network stacks. Additional things changed with this addition: - Moved some mbuf specific declarations and initializations from sys/conf/param.c into mbuf-specific code where they belong. - m_getclr() has been renamed to m_get_clrd() because the old name is really confusing. m_getclr() HAS been preserved though and is defined to the new name. No tree sweep has been done "to change the interface," as the old name will continue to be supported and is not depracated. The change was merely done because m_getclr() sounds too much like "m_get a cluster." - TEMPORARILY disabled mbtypes statistics displaying in netstat(1) and systat(1) (see TODO below). - Fixed systat(1) to display number of "free mbufs" based on new per-CPU stat structures. - Fixed netstat(1) to display new per-CPU stats based on sysctl-exported per-CPU stat structures. All infos are fetched via sysctl. TODO (in order of priority): - Re-enable mbtypes statistics in both netstat(1) and systat(1) after introducing an SMP friendly way to collect the mbtypes stats under the already introduced per-CPU locks (i.e. hopefully don't use atomic() - it seems too costly for a mere stat update, especially when other locks are already present). - Optionally have systat(1) display not only "total free mbufs" but also "total free mbufs per CPU pool." - Fix minor length-fetching issues in netstat(1) related to recently re-enabled option to read mbuf stats from a core file. - Move reference counters at least for mbuf clusters into an unused portion of the cluster itself, to save space and need to allocate a counter. - Look into introducing resource freeing possibly from a kproc. Reviewed by (in parts): jlemon, jake, silby, terry Tested by: jlemon (Intel & Alpha), mjacob (Intel & Alpha) Preliminary performance measurements: jlemon (and me, obviously) URL: http://people.freebsd.org/~bmilekic/mb_alloc/
2001-06-22 06:35:32 +00:00
if (map == kmem_map)
splx(s);
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
return (result);
1994-05-24 10:09:53 +00:00
}
/*
* vm_map_check_protection:
*
* Assert that the target map allows the specified privilege on the
* entire address region given. The entire region must be allocated.
*
* WARNING! This code does not and should not check whether the
* contents of the region is accessible. For example a smaller file
* might be mapped into a larger address space.
*
* NOTE! This code is also called by munmap().
*
* The map must be locked. A read lock is sufficient.
1994-05-24 10:09:53 +00:00
*/
1995-05-30 08:16:23 +00:00
boolean_t
vm_map_check_protection(vm_map_t map, vm_offset_t start, vm_offset_t end,
vm_prot_t protection)
1994-05-24 10:09:53 +00:00
{
vm_map_entry_t entry;
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_map_entry_t tmp_entry;
1994-05-24 10:09:53 +00:00
if (!vm_map_lookup_entry(map, start, &tmp_entry))
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 (FALSE);
1994-05-24 10:09:53 +00:00
entry = tmp_entry;
while (start < end) {
if (entry == &map->header)
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 (FALSE);
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
* No holes allowed!
1994-05-24 10:09:53 +00:00
*/
if (start < entry->start)
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
return (FALSE);
1994-05-24 10:09:53 +00:00
/*
* Check protection associated with entry.
*/
if ((entry->protection & protection) != protection)
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 (FALSE);
1994-05-24 10:09:53 +00:00
/* go to next entry */
start = entry->end;
entry = entry->next;
}
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 (TRUE);
1994-05-24 10:09:53 +00:00
}
/*
* vm_map_copy_entry:
*
* Copies the contents of the source entry to the destination
* entry. The entries *must* be aligned properly.
*/
static void
vm_map_copy_entry(
vm_map_t src_map,
vm_map_t dst_map,
2003-11-03 16:14:45 +00:00
vm_map_entry_t src_entry,
vm_map_entry_t dst_entry)
1994-05-24 10:09:53 +00:00
{
vm_object_t src_object;
if ((dst_entry->eflags|src_entry->eflags) & MAP_ENTRY_IS_SUB_MAP)
1994-05-24 10:09:53 +00:00
return;
if (src_entry->wired_count == 0) {
/*
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
* If the source entry is marked needs_copy, it is already
* write-protected.
1994-05-24 10:09:53 +00:00
*/
if ((src_entry->eflags & MAP_ENTRY_NEEDS_COPY) == 0) {
VM level code cleanups. 1) Start using TSM. Struct procs continue to point to upages structure, after being freed. Struct vmspace continues to point to pte object and kva space for kstack. u_map is now superfluous. 2) vm_map's don't need to be reference counted. They always exist either in the kernel or in a vmspace. The vmspaces are managed by reference counts. 3) Remove the "wired" vm_map nonsense. 4) No need to keep a cache of kernel stack kva's. 5) Get rid of strange looking ++var, and change to var++. 6) Change more data structures to use our "zone" allocator. Added struct proc, struct vmspace and struct vnode. This saves a significant amount of kva space and physical memory. Additionally, this enables TSM for the zone managed memory. 7) Keep ioopt disabled for now. 8) Remove the now bogus "single use" map concept. 9) Use generation counts or id's for data structures residing in TSM, where it allows us to avoid unneeded restart overhead during traversals, where blocking might occur. 10) Account better for memory deficits, so the pageout daemon will be able to make enough memory available (experimental.) 11) Fix some vnode locking problems. (From Tor, I think.) 12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp. (experimental.) 13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c code. Use generation counts, get rid of unneded collpase operations, and clean up the cluster code. 14) Make vm_zone more suitable for TSM. This commit is partially as a result of discussions and contributions from other people, including DG, Tor Egge, PHK, and probably others that I have forgotten to attribute (so let me know, if I forgot.) This is not the infamous, final cleanup of the vnode stuff, but a necessary step. Vnode mgmt should be correct, but things might still change, and there is still some missing stuff (like ioopt, and physical backing of non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
pmap_protect(src_map->pmap,
src_entry->start,
src_entry->end,
src_entry->protection & ~VM_PROT_WRITE);
1994-05-24 10:09:53 +00:00
}
This set of commits to the VM system does the following, and contain contributions or ideas from Stephen McKay <syssgm@devetir.qld.gov.au>, Alan Cox <alc@cs.rice.edu>, David Greenman <davidg@freebsd.org> and me: More usage of the TAILQ macros. Additional minor fix to queue.h. Performance enhancements to the pageout daemon. Addition of a wait in the case that the pageout daemon has to run immediately. Slightly modify the pageout algorithm. Significant revamp of the pmap/fork code: 1) PTE's and UPAGES's are NO LONGER in the process's map. 2) PTE's and UPAGES's reside in their own objects. 3) TOTAL elimination of recursive page table pagefaults. 4) The page directory now resides in the PTE object. 5) Implemented pmap_copy, thereby speeding up fork time. 6) Changed the pv entries so that the head is a pointer and not an entire entry. 7) Significant cleanup of pmap_protect, and pmap_remove. 8) Removed significant amounts of machine dependent fork code from vm_glue. Pushed much of that code into the machine dependent pmap module. 9) Support more completely the reuse of already zeroed pages (Page table pages and page directories) as being already zeroed. Performance and code cleanups in vm_map: 1) Improved and simplified allocation of map entries. 2) Improved vm_map_copy code. 3) Corrected some minor problems in the simplify code. Implemented splvm (combo of splbio and splimp.) The VM code now seldom uses splhigh. Improved the speed of and simplified kmem_malloc. Minor mod to vm_fault to avoid using pre-zeroed pages in the case of objects with backing objects along with the already existant condition of having a vnode. (If there is a backing object, there will likely be a COW... With a COW, it isn't necessary to start with a pre-zeroed page.) Minor reorg of source to perhaps improve locality of ref.
1996-05-18 03:38:05 +00:00
1994-05-24 10:09:53 +00:00
/*
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
* Make a copy of the object.
1994-05-24 10:09:53 +00:00
*/
if ((src_object = src_entry->object.vm_object) != NULL) {
VM_OBJECT_LOCK(src_object);
if ((src_object->handle == NULL) &&
(src_object->type == OBJT_DEFAULT ||
src_object->type == OBJT_SWAP)) {
vm_object_collapse(src_object);
if ((src_object->flags & (OBJ_NOSPLIT|OBJ_ONEMAPPING)) == OBJ_ONEMAPPING) {
vm_object_split(src_entry);
src_object = src_entry->object.vm_object;
}
}
vm_object_reference_locked(src_object);
vm_object_clear_flag(src_object, OBJ_ONEMAPPING);
VM_OBJECT_UNLOCK(src_object);
dst_entry->object.vm_object = src_object;
src_entry->eflags |= (MAP_ENTRY_COW|MAP_ENTRY_NEEDS_COPY);
dst_entry->eflags |= (MAP_ENTRY_COW|MAP_ENTRY_NEEDS_COPY);
This set of commits to the VM system does the following, and contain contributions or ideas from Stephen McKay <syssgm@devetir.qld.gov.au>, Alan Cox <alc@cs.rice.edu>, David Greenman <davidg@freebsd.org> and me: More usage of the TAILQ macros. Additional minor fix to queue.h. Performance enhancements to the pageout daemon. Addition of a wait in the case that the pageout daemon has to run immediately. Slightly modify the pageout algorithm. Significant revamp of the pmap/fork code: 1) PTE's and UPAGES's are NO LONGER in the process's map. 2) PTE's and UPAGES's reside in their own objects. 3) TOTAL elimination of recursive page table pagefaults. 4) The page directory now resides in the PTE object. 5) Implemented pmap_copy, thereby speeding up fork time. 6) Changed the pv entries so that the head is a pointer and not an entire entry. 7) Significant cleanup of pmap_protect, and pmap_remove. 8) Removed significant amounts of machine dependent fork code from vm_glue. Pushed much of that code into the machine dependent pmap module. 9) Support more completely the reuse of already zeroed pages (Page table pages and page directories) as being already zeroed. Performance and code cleanups in vm_map: 1) Improved and simplified allocation of map entries. 2) Improved vm_map_copy code. 3) Corrected some minor problems in the simplify code. Implemented splvm (combo of splbio and splimp.) The VM code now seldom uses splhigh. Improved the speed of and simplified kmem_malloc. Minor mod to vm_fault to avoid using pre-zeroed pages in the case of objects with backing objects along with the already existant condition of having a vnode. (If there is a backing object, there will likely be a COW... With a COW, it isn't necessary to start with a pre-zeroed page.) Minor reorg of source to perhaps improve locality of ref.
1996-05-18 03:38:05 +00:00
dst_entry->offset = src_entry->offset;
} else {
dst_entry->object.vm_object = NULL;
dst_entry->offset = 0;
}
1994-05-24 10:09:53 +00:00
pmap_copy(dst_map->pmap, src_map->pmap, dst_entry->start,
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
dst_entry->end - dst_entry->start, src_entry->start);
} else {
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
* Of course, wired down pages can't be set copy-on-write.
* Cause wired pages to be copied into the new map by
* simulating faults (the new pages are pageable)
1994-05-24 10:09:53 +00:00
*/
vm_fault_copy_entry(dst_map, src_map, dst_entry, src_entry);
}
}
/*
* vmspace_map_entry_forked:
* Update the newly-forked vmspace each time a map entry is inherited
* or copied. The values for vm_dsize and vm_tsize are approximate
* (and mostly-obsolete ideas in the face of mmap(2) et al.)
*/
static void
vmspace_map_entry_forked(const struct vmspace *vm1, struct vmspace *vm2,
vm_map_entry_t entry)
{
vm_size_t entrysize;
vm_offset_t newend;
entrysize = entry->end - entry->start;
vm2->vm_map.size += entrysize;
if (entry->eflags & (MAP_ENTRY_GROWS_DOWN | MAP_ENTRY_GROWS_UP)) {
vm2->vm_ssize += btoc(entrysize);
} else if (entry->start >= (vm_offset_t)vm1->vm_daddr &&
entry->start < (vm_offset_t)vm1->vm_daddr + ctob(vm1->vm_dsize)) {
newend = MIN(entry->end,
(vm_offset_t)vm1->vm_daddr + ctob(vm1->vm_dsize));
vm2->vm_dsize += btoc(newend - entry->start);
} else if (entry->start >= (vm_offset_t)vm1->vm_taddr &&
entry->start < (vm_offset_t)vm1->vm_taddr + ctob(vm1->vm_tsize)) {
newend = MIN(entry->end,
(vm_offset_t)vm1->vm_taddr + ctob(vm1->vm_tsize));
vm2->vm_tsize += btoc(newend - entry->start);
}
}
1994-05-24 10:09:53 +00:00
/*
* vmspace_fork:
* Create a new process vmspace structure and vm_map
* based on those of an existing process. The new map
* is based on the old map, according to the inheritance
* values on the regions in that map.
*
* XXX It might be worth coalescing the entries added to the new vmspace.
*
1994-05-24 10:09:53 +00:00
* The source map must not be locked.
*/
struct vmspace *
vmspace_fork(struct vmspace *vm1)
1994-05-24 10:09:53 +00:00
{
struct vmspace *vm2;
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_map_t old_map = &vm1->vm_map;
vm_map_t new_map;
vm_map_entry_t old_entry;
vm_map_entry_t new_entry;
vm_object_t object;
1994-05-24 10:09:53 +00:00
GIANT_REQUIRED;
1994-05-24 10:09:53 +00:00
vm_map_lock(old_map);
old_map->infork = 1;
1994-05-24 10:09:53 +00:00
VM level code cleanups. 1) Start using TSM. Struct procs continue to point to upages structure, after being freed. Struct vmspace continues to point to pte object and kva space for kstack. u_map is now superfluous. 2) vm_map's don't need to be reference counted. They always exist either in the kernel or in a vmspace. The vmspaces are managed by reference counts. 3) Remove the "wired" vm_map nonsense. 4) No need to keep a cache of kernel stack kva's. 5) Get rid of strange looking ++var, and change to var++. 6) Change more data structures to use our "zone" allocator. Added struct proc, struct vmspace and struct vnode. This saves a significant amount of kva space and physical memory. Additionally, this enables TSM for the zone managed memory. 7) Keep ioopt disabled for now. 8) Remove the now bogus "single use" map concept. 9) Use generation counts or id's for data structures residing in TSM, where it allows us to avoid unneeded restart overhead during traversals, where blocking might occur. 10) Account better for memory deficits, so the pageout daemon will be able to make enough memory available (experimental.) 11) Fix some vnode locking problems. (From Tor, I think.) 12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp. (experimental.) 13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c code. Use generation counts, get rid of unneded collpase operations, and clean up the cluster code. 14) Make vm_zone more suitable for TSM. This commit is partially as a result of discussions and contributions from other people, including DG, Tor Egge, PHK, and probably others that I have forgotten to attribute (so let me know, if I forgot.) This is not the infamous, final cleanup of the vnode stuff, but a necessary step. Vnode mgmt should be correct, but things might still change, and there is still some missing stuff (like ioopt, and physical backing of non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
vm2 = vmspace_alloc(old_map->min_offset, old_map->max_offset);
vm2->vm_taddr = vm1->vm_taddr;
vm2->vm_daddr = vm1->vm_daddr;
vm2->vm_maxsaddr = vm1->vm_maxsaddr;
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
new_map = &vm2->vm_map; /* XXX */
new_map->timestamp = 1;
1994-05-24 10:09:53 +00:00
/* Do not inherit the MAP_WIREFUTURE property. */
if ((new_map->flags & MAP_WIREFUTURE) == MAP_WIREFUTURE)
new_map->flags &= ~MAP_WIREFUTURE;
1994-05-24 10:09:53 +00:00
old_entry = old_map->header.next;
while (old_entry != &old_map->header) {
if (old_entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1994-05-24 10:09:53 +00:00
panic("vm_map_fork: encountered a submap");
switch (old_entry->inheritance) {
case VM_INHERIT_NONE:
break;
case VM_INHERIT_SHARE:
/*
* Clone the entry, creating the shared object if necessary.
*/
object = old_entry->object.vm_object;
if (object == NULL) {
object = vm_object_allocate(OBJT_DEFAULT,
atop(old_entry->end - old_entry->start));
old_entry->object.vm_object = object;
old_entry->offset = (vm_offset_t) 0;
}
/*
* Add the reference before calling vm_object_shadow
* to insure that a shadow object is created.
*/
vm_object_reference(object);
if (old_entry->eflags & MAP_ENTRY_NEEDS_COPY) {
vm_object_shadow(&old_entry->object.vm_object,
&old_entry->offset,
atop(old_entry->end - old_entry->start));
old_entry->eflags &= ~MAP_ENTRY_NEEDS_COPY;
/* Transfer the second reference too. */
vm_object_reference(
old_entry->object.vm_object);
vm_object_deallocate(object);
object = old_entry->object.vm_object;
}
VM_OBJECT_LOCK(object);
vm_object_clear_flag(object, OBJ_ONEMAPPING);
VM_OBJECT_UNLOCK(object);
1994-05-24 10:09:53 +00:00
/*
* Clone the entry, referencing the shared object.
1994-05-24 10:09:53 +00:00
*/
new_entry = vm_map_entry_create(new_map);
*new_entry = *old_entry;
new_entry->eflags &= ~MAP_ENTRY_USER_WIRED;
1994-05-24 10:09:53 +00:00
new_entry->wired_count = 0;
/*
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
* Insert the entry into the new map -- we know we're
* inserting at the end of the new map.
1994-05-24 10:09:53 +00:00
*/
vm_map_entry_link(new_map, new_map->header.prev,
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
new_entry);
vmspace_map_entry_forked(vm1, vm2, new_entry);
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
* Update the physical map
1994-05-24 10:09:53 +00:00
*/
pmap_copy(new_map->pmap, old_map->pmap,
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
new_entry->start,
(old_entry->end - old_entry->start),
old_entry->start);
1994-05-24 10:09:53 +00:00
break;
case VM_INHERIT_COPY:
/*
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
* Clone the entry and link into the map.
1994-05-24 10:09:53 +00:00
*/
new_entry = vm_map_entry_create(new_map);
*new_entry = *old_entry;
new_entry->eflags &= ~MAP_ENTRY_USER_WIRED;
1994-05-24 10:09:53 +00:00
new_entry->wired_count = 0;
new_entry->object.vm_object = NULL;
vm_map_entry_link(new_map, new_map->header.prev,
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
new_entry);
vmspace_map_entry_forked(vm1, vm2, new_entry);
vm_map_copy_entry(old_map, new_map, old_entry,
new_entry);
1994-05-24 10:09:53 +00:00
break;
}
old_entry = old_entry->next;
}
old_map->infork = 0;
1994-05-24 10:09:53 +00:00
vm_map_unlock(old_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 (vm2);
1994-05-24 10:09:53 +00:00
}
int
vm_map_stack(vm_map_t map, vm_offset_t addrbos, vm_size_t max_ssize,
vm_prot_t prot, vm_prot_t max, int cow)
{
vm_map_entry_t new_entry, prev_entry;
vm_offset_t bot, top;
vm_size_t init_ssize;
int orient, rv;
Locking for the per-process resource limits structure. - struct plimit includes a mutex to protect a reference count. The plimit structure is treated similarly to struct ucred in that is is always copy on write, so having a reference to a structure is sufficient to read from it without needing a further lock. - The proc lock protects the p_limit pointer and must be held while reading limits from a process to keep the limit structure from changing out from under you while reading from it. - Various global limits that are ints are not protected by a lock since int writes are atomic on all the archs we support and thus a lock wouldn't buy us anything. - All accesses to individual resource limits from a process are abstracted behind a simple lim_rlimit(), lim_max(), and lim_cur() API that return either an rlimit, or the current or max individual limit of the specified resource from a process. - dosetrlimit() was renamed to kern_setrlimit() to match existing style of other similar syscall helper functions. - The alpha OSF/1 compat layer no longer calls getrlimit() and setrlimit() (it didn't used the stackgap when it should have) but uses lim_rlimit() and kern_setrlimit() instead. - The svr4 compat no longer uses the stackgap for resource limits calls, but uses lim_rlimit() and kern_setrlimit() instead. - The ibcs2 compat no longer uses the stackgap for resource limits. It also no longer uses the stackgap for accessing sysctl's for the ibcs2_sysconf() syscall but uses kernel_sysctl() instead. As a result, ibcs2_sysconf() no longer needs Giant. - The p_rlimit macro no longer exists. Submitted by: mtm (mostly, I only did a few cleanups and catchups) Tested on: i386 Compiled on: alpha, amd64
2004-02-04 21:52:57 +00:00
rlim_t vmemlim;
/*
* The stack orientation is piggybacked with the cow argument.
* Extract it into orient and mask the cow argument so that we
* don't pass it around further.
* NOTE: We explicitly allow bi-directional stacks.
*/
orient = cow & (MAP_STACK_GROWS_DOWN|MAP_STACK_GROWS_UP);
cow &= ~orient;
KASSERT(orient != 0, ("No stack grow direction"));
if (addrbos < vm_map_min(map) || addrbos > map->max_offset)
return (KERN_NO_SPACE);
init_ssize = (max_ssize < sgrowsiz) ? max_ssize : sgrowsiz;
Locking for the per-process resource limits structure. - struct plimit includes a mutex to protect a reference count. The plimit structure is treated similarly to struct ucred in that is is always copy on write, so having a reference to a structure is sufficient to read from it without needing a further lock. - The proc lock protects the p_limit pointer and must be held while reading limits from a process to keep the limit structure from changing out from under you while reading from it. - Various global limits that are ints are not protected by a lock since int writes are atomic on all the archs we support and thus a lock wouldn't buy us anything. - All accesses to individual resource limits from a process are abstracted behind a simple lim_rlimit(), lim_max(), and lim_cur() API that return either an rlimit, or the current or max individual limit of the specified resource from a process. - dosetrlimit() was renamed to kern_setrlimit() to match existing style of other similar syscall helper functions. - The alpha OSF/1 compat layer no longer calls getrlimit() and setrlimit() (it didn't used the stackgap when it should have) but uses lim_rlimit() and kern_setrlimit() instead. - The svr4 compat no longer uses the stackgap for resource limits calls, but uses lim_rlimit() and kern_setrlimit() instead. - The ibcs2 compat no longer uses the stackgap for resource limits. It also no longer uses the stackgap for accessing sysctl's for the ibcs2_sysconf() syscall but uses kernel_sysctl() instead. As a result, ibcs2_sysconf() no longer needs Giant. - The p_rlimit macro no longer exists. Submitted by: mtm (mostly, I only did a few cleanups and catchups) Tested on: i386 Compiled on: alpha, amd64
2004-02-04 21:52:57 +00:00
PROC_LOCK(curthread->td_proc);
vmemlim = lim_cur(curthread->td_proc, RLIMIT_VMEM);
PROC_UNLOCK(curthread->td_proc);
vm_map_lock(map);
/* If addr is already mapped, no go */
if (vm_map_lookup_entry(map, addrbos, &prev_entry)) {
vm_map_unlock(map);
return (KERN_NO_SPACE);
}
/* If we would blow our VMEM resource limit, no go */
Locking for the per-process resource limits structure. - struct plimit includes a mutex to protect a reference count. The plimit structure is treated similarly to struct ucred in that is is always copy on write, so having a reference to a structure is sufficient to read from it without needing a further lock. - The proc lock protects the p_limit pointer and must be held while reading limits from a process to keep the limit structure from changing out from under you while reading from it. - Various global limits that are ints are not protected by a lock since int writes are atomic on all the archs we support and thus a lock wouldn't buy us anything. - All accesses to individual resource limits from a process are abstracted behind a simple lim_rlimit(), lim_max(), and lim_cur() API that return either an rlimit, or the current or max individual limit of the specified resource from a process. - dosetrlimit() was renamed to kern_setrlimit() to match existing style of other similar syscall helper functions. - The alpha OSF/1 compat layer no longer calls getrlimit() and setrlimit() (it didn't used the stackgap when it should have) but uses lim_rlimit() and kern_setrlimit() instead. - The svr4 compat no longer uses the stackgap for resource limits calls, but uses lim_rlimit() and kern_setrlimit() instead. - The ibcs2 compat no longer uses the stackgap for resource limits. It also no longer uses the stackgap for accessing sysctl's for the ibcs2_sysconf() syscall but uses kernel_sysctl() instead. As a result, ibcs2_sysconf() no longer needs Giant. - The p_rlimit macro no longer exists. Submitted by: mtm (mostly, I only did a few cleanups and catchups) Tested on: i386 Compiled on: alpha, amd64
2004-02-04 21:52:57 +00:00
if (map->size + init_ssize > vmemlim) {
vm_map_unlock(map);
return (KERN_NO_SPACE);
}
/*
* If we can't accomodate max_ssize in the current mapping, no go.
* However, we need to be aware that subsequent user mappings might
* map into the space we have reserved for stack, and currently this
2003-11-03 16:14:45 +00:00
* space is not protected.
*
* Hopefully we will at least detect this condition when we try to
* grow the stack.
*/
if ((prev_entry->next != &map->header) &&
(prev_entry->next->start < addrbos + max_ssize)) {
vm_map_unlock(map);
return (KERN_NO_SPACE);
}
/*
* We initially map a stack of only init_ssize. We will grow as
* needed later. Depending on the orientation of the stack (i.e.
* the grow direction) we either map at the top of the range, the
* bottom of the range or in the middle.
*
* Note: we would normally expect prot and max to be VM_PROT_ALL,
* and cow to be 0. Possibly we should eliminate these as input
* parameters, and just pass these values here in the insert call.
*/
if (orient == MAP_STACK_GROWS_DOWN)
bot = addrbos + max_ssize - init_ssize;
else if (orient == MAP_STACK_GROWS_UP)
bot = addrbos;
else
bot = round_page(addrbos + max_ssize/2 - init_ssize/2);
top = bot + init_ssize;
rv = vm_map_insert(map, NULL, 0, bot, top, prot, max, cow);
/* Now set the avail_ssize amount. */
if (rv == KERN_SUCCESS) {
if (prev_entry != &map->header)
vm_map_clip_end(map, prev_entry, bot);
new_entry = prev_entry->next;
if (new_entry->end != top || new_entry->start != bot)
panic("Bad entry start/end for new stack entry");
new_entry->avail_ssize = max_ssize - init_ssize;
if (orient & MAP_STACK_GROWS_DOWN)
new_entry->eflags |= MAP_ENTRY_GROWS_DOWN;
if (orient & MAP_STACK_GROWS_UP)
new_entry->eflags |= MAP_ENTRY_GROWS_UP;
}
vm_map_unlock(map);
return (rv);
}
/* Attempts to grow a vm stack entry. Returns KERN_SUCCESS if the
* desired address is already mapped, or if we successfully grow
* the stack. Also returns KERN_SUCCESS if addr is outside the
* stack range (this is strange, but preserves compatibility with
* the grow function in vm_machdep.c).
*/
int
vm_map_growstack(struct proc *p, vm_offset_t addr)
{
vm_map_entry_t next_entry, prev_entry;
vm_map_entry_t new_entry, stack_entry;
struct vmspace *vm = p->p_vmspace;
vm_map_t map = &vm->vm_map;
vm_offset_t end;
size_t grow_amount, max_grow;
Locking for the per-process resource limits structure. - struct plimit includes a mutex to protect a reference count. The plimit structure is treated similarly to struct ucred in that is is always copy on write, so having a reference to a structure is sufficient to read from it without needing a further lock. - The proc lock protects the p_limit pointer and must be held while reading limits from a process to keep the limit structure from changing out from under you while reading from it. - Various global limits that are ints are not protected by a lock since int writes are atomic on all the archs we support and thus a lock wouldn't buy us anything. - All accesses to individual resource limits from a process are abstracted behind a simple lim_rlimit(), lim_max(), and lim_cur() API that return either an rlimit, or the current or max individual limit of the specified resource from a process. - dosetrlimit() was renamed to kern_setrlimit() to match existing style of other similar syscall helper functions. - The alpha OSF/1 compat layer no longer calls getrlimit() and setrlimit() (it didn't used the stackgap when it should have) but uses lim_rlimit() and kern_setrlimit() instead. - The svr4 compat no longer uses the stackgap for resource limits calls, but uses lim_rlimit() and kern_setrlimit() instead. - The ibcs2 compat no longer uses the stackgap for resource limits. It also no longer uses the stackgap for accessing sysctl's for the ibcs2_sysconf() syscall but uses kernel_sysctl() instead. As a result, ibcs2_sysconf() no longer needs Giant. - The p_rlimit macro no longer exists. Submitted by: mtm (mostly, I only did a few cleanups and catchups) Tested on: i386 Compiled on: alpha, amd64
2004-02-04 21:52:57 +00:00
rlim_t stacklim, vmemlim;
int is_procstack, rv;
Retry:
Locking for the per-process resource limits structure. - struct plimit includes a mutex to protect a reference count. The plimit structure is treated similarly to struct ucred in that is is always copy on write, so having a reference to a structure is sufficient to read from it without needing a further lock. - The proc lock protects the p_limit pointer and must be held while reading limits from a process to keep the limit structure from changing out from under you while reading from it. - Various global limits that are ints are not protected by a lock since int writes are atomic on all the archs we support and thus a lock wouldn't buy us anything. - All accesses to individual resource limits from a process are abstracted behind a simple lim_rlimit(), lim_max(), and lim_cur() API that return either an rlimit, or the current or max individual limit of the specified resource from a process. - dosetrlimit() was renamed to kern_setrlimit() to match existing style of other similar syscall helper functions. - The alpha OSF/1 compat layer no longer calls getrlimit() and setrlimit() (it didn't used the stackgap when it should have) but uses lim_rlimit() and kern_setrlimit() instead. - The svr4 compat no longer uses the stackgap for resource limits calls, but uses lim_rlimit() and kern_setrlimit() instead. - The ibcs2 compat no longer uses the stackgap for resource limits. It also no longer uses the stackgap for accessing sysctl's for the ibcs2_sysconf() syscall but uses kernel_sysctl() instead. As a result, ibcs2_sysconf() no longer needs Giant. - The p_rlimit macro no longer exists. Submitted by: mtm (mostly, I only did a few cleanups and catchups) Tested on: i386 Compiled on: alpha, amd64
2004-02-04 21:52:57 +00:00
PROC_LOCK(p);
stacklim = lim_cur(p, RLIMIT_STACK);
vmemlim = lim_cur(p, RLIMIT_VMEM);
Locking for the per-process resource limits structure. - struct plimit includes a mutex to protect a reference count. The plimit structure is treated similarly to struct ucred in that is is always copy on write, so having a reference to a structure is sufficient to read from it without needing a further lock. - The proc lock protects the p_limit pointer and must be held while reading limits from a process to keep the limit structure from changing out from under you while reading from it. - Various global limits that are ints are not protected by a lock since int writes are atomic on all the archs we support and thus a lock wouldn't buy us anything. - All accesses to individual resource limits from a process are abstracted behind a simple lim_rlimit(), lim_max(), and lim_cur() API that return either an rlimit, or the current or max individual limit of the specified resource from a process. - dosetrlimit() was renamed to kern_setrlimit() to match existing style of other similar syscall helper functions. - The alpha OSF/1 compat layer no longer calls getrlimit() and setrlimit() (it didn't used the stackgap when it should have) but uses lim_rlimit() and kern_setrlimit() instead. - The svr4 compat no longer uses the stackgap for resource limits calls, but uses lim_rlimit() and kern_setrlimit() instead. - The ibcs2 compat no longer uses the stackgap for resource limits. It also no longer uses the stackgap for accessing sysctl's for the ibcs2_sysconf() syscall but uses kernel_sysctl() instead. As a result, ibcs2_sysconf() no longer needs Giant. - The p_rlimit macro no longer exists. Submitted by: mtm (mostly, I only did a few cleanups and catchups) Tested on: i386 Compiled on: alpha, amd64
2004-02-04 21:52:57 +00:00
PROC_UNLOCK(p);
vm_map_lock_read(map);
/* If addr is already in the entry range, no need to grow.*/
if (vm_map_lookup_entry(map, addr, &prev_entry)) {
vm_map_unlock_read(map);
return (KERN_SUCCESS);
}
next_entry = prev_entry->next;
if (!(prev_entry->eflags & MAP_ENTRY_GROWS_UP)) {
/*
* This entry does not grow upwards. Since the address lies
* beyond this entry, the next entry (if one exists) has to
* be a downward growable entry. The entry list header is
* never a growable entry, so it suffices to check the flags.
*/
if (!(next_entry->eflags & MAP_ENTRY_GROWS_DOWN)) {
vm_map_unlock_read(map);
return (KERN_SUCCESS);
}
stack_entry = next_entry;
} else {
/*
* This entry grows upward. If the next entry does not at
* least grow downwards, this is the entry we need to grow.
* otherwise we have two possible choices and we have to
* select one.
*/
if (next_entry->eflags & MAP_ENTRY_GROWS_DOWN) {
/*
* We have two choices; grow the entry closest to
* the address to minimize the amount of growth.
*/
if (addr - prev_entry->end <= next_entry->start - addr)
stack_entry = prev_entry;
else
stack_entry = next_entry;
} else
stack_entry = prev_entry;
}
if (stack_entry == next_entry) {
KASSERT(stack_entry->eflags & MAP_ENTRY_GROWS_DOWN, ("foo"));
KASSERT(addr < stack_entry->start, ("foo"));
end = (prev_entry != &map->header) ? prev_entry->end :
stack_entry->start - stack_entry->avail_ssize;
grow_amount = roundup(stack_entry->start - addr, PAGE_SIZE);
max_grow = stack_entry->start - end;
} else {
KASSERT(stack_entry->eflags & MAP_ENTRY_GROWS_UP, ("foo"));
KASSERT(addr >= stack_entry->end, ("foo"));
end = (next_entry != &map->header) ? next_entry->start :
stack_entry->end + stack_entry->avail_ssize;
grow_amount = roundup(addr + 1 - stack_entry->end, PAGE_SIZE);
max_grow = end - stack_entry->end;
}
if (grow_amount > stack_entry->avail_ssize) {
vm_map_unlock_read(map);
return (KERN_NO_SPACE);
}
/*
* If there is no longer enough space between the entries nogo, and
* adjust the available space. Note: this should only happen if the
* user has mapped into the stack area after the stack was created,
* and is probably an error.
*
* This also effectively destroys any guard page the user might have
* intended by limiting the stack size.
*/
if (grow_amount > max_grow) {
if (vm_map_lock_upgrade(map))
goto Retry;
stack_entry->avail_ssize = max_grow;
vm_map_unlock(map);
return (KERN_NO_SPACE);
}
is_procstack = (addr >= (vm_offset_t)vm->vm_maxsaddr) ? 1 : 0;
/*
* If this is the main process stack, see if we're over the stack
* limit.
*/
Locking for the per-process resource limits structure. - struct plimit includes a mutex to protect a reference count. The plimit structure is treated similarly to struct ucred in that is is always copy on write, so having a reference to a structure is sufficient to read from it without needing a further lock. - The proc lock protects the p_limit pointer and must be held while reading limits from a process to keep the limit structure from changing out from under you while reading from it. - Various global limits that are ints are not protected by a lock since int writes are atomic on all the archs we support and thus a lock wouldn't buy us anything. - All accesses to individual resource limits from a process are abstracted behind a simple lim_rlimit(), lim_max(), and lim_cur() API that return either an rlimit, or the current or max individual limit of the specified resource from a process. - dosetrlimit() was renamed to kern_setrlimit() to match existing style of other similar syscall helper functions. - The alpha OSF/1 compat layer no longer calls getrlimit() and setrlimit() (it didn't used the stackgap when it should have) but uses lim_rlimit() and kern_setrlimit() instead. - The svr4 compat no longer uses the stackgap for resource limits calls, but uses lim_rlimit() and kern_setrlimit() instead. - The ibcs2 compat no longer uses the stackgap for resource limits. It also no longer uses the stackgap for accessing sysctl's for the ibcs2_sysconf() syscall but uses kernel_sysctl() instead. As a result, ibcs2_sysconf() no longer needs Giant. - The p_rlimit macro no longer exists. Submitted by: mtm (mostly, I only did a few cleanups and catchups) Tested on: i386 Compiled on: alpha, amd64
2004-02-04 21:52:57 +00:00
if (is_procstack && (ctob(vm->vm_ssize) + grow_amount > stacklim)) {
vm_map_unlock_read(map);
return (KERN_NO_SPACE);
}
/* Round up the grow amount modulo SGROWSIZ */
grow_amount = roundup (grow_amount, sgrowsiz);
if (grow_amount > stack_entry->avail_ssize)
grow_amount = stack_entry->avail_ssize;
Locking for the per-process resource limits structure. - struct plimit includes a mutex to protect a reference count. The plimit structure is treated similarly to struct ucred in that is is always copy on write, so having a reference to a structure is sufficient to read from it without needing a further lock. - The proc lock protects the p_limit pointer and must be held while reading limits from a process to keep the limit structure from changing out from under you while reading from it. - Various global limits that are ints are not protected by a lock since int writes are atomic on all the archs we support and thus a lock wouldn't buy us anything. - All accesses to individual resource limits from a process are abstracted behind a simple lim_rlimit(), lim_max(), and lim_cur() API that return either an rlimit, or the current or max individual limit of the specified resource from a process. - dosetrlimit() was renamed to kern_setrlimit() to match existing style of other similar syscall helper functions. - The alpha OSF/1 compat layer no longer calls getrlimit() and setrlimit() (it didn't used the stackgap when it should have) but uses lim_rlimit() and kern_setrlimit() instead. - The svr4 compat no longer uses the stackgap for resource limits calls, but uses lim_rlimit() and kern_setrlimit() instead. - The ibcs2 compat no longer uses the stackgap for resource limits. It also no longer uses the stackgap for accessing sysctl's for the ibcs2_sysconf() syscall but uses kernel_sysctl() instead. As a result, ibcs2_sysconf() no longer needs Giant. - The p_rlimit macro no longer exists. Submitted by: mtm (mostly, I only did a few cleanups and catchups) Tested on: i386 Compiled on: alpha, amd64
2004-02-04 21:52:57 +00:00
if (is_procstack && (ctob(vm->vm_ssize) + grow_amount > stacklim)) {
grow_amount = stacklim - ctob(vm->vm_ssize);
}
/* If we would blow our VMEM resource limit, no go */
Locking for the per-process resource limits structure. - struct plimit includes a mutex to protect a reference count. The plimit structure is treated similarly to struct ucred in that is is always copy on write, so having a reference to a structure is sufficient to read from it without needing a further lock. - The proc lock protects the p_limit pointer and must be held while reading limits from a process to keep the limit structure from changing out from under you while reading from it. - Various global limits that are ints are not protected by a lock since int writes are atomic on all the archs we support and thus a lock wouldn't buy us anything. - All accesses to individual resource limits from a process are abstracted behind a simple lim_rlimit(), lim_max(), and lim_cur() API that return either an rlimit, or the current or max individual limit of the specified resource from a process. - dosetrlimit() was renamed to kern_setrlimit() to match existing style of other similar syscall helper functions. - The alpha OSF/1 compat layer no longer calls getrlimit() and setrlimit() (it didn't used the stackgap when it should have) but uses lim_rlimit() and kern_setrlimit() instead. - The svr4 compat no longer uses the stackgap for resource limits calls, but uses lim_rlimit() and kern_setrlimit() instead. - The ibcs2 compat no longer uses the stackgap for resource limits. It also no longer uses the stackgap for accessing sysctl's for the ibcs2_sysconf() syscall but uses kernel_sysctl() instead. As a result, ibcs2_sysconf() no longer needs Giant. - The p_rlimit macro no longer exists. Submitted by: mtm (mostly, I only did a few cleanups and catchups) Tested on: i386 Compiled on: alpha, amd64
2004-02-04 21:52:57 +00:00
if (map->size + grow_amount > vmemlim) {
vm_map_unlock_read(map);
return (KERN_NO_SPACE);
}
if (vm_map_lock_upgrade(map))
goto Retry;
if (stack_entry == next_entry) {
/*
* Growing downward.
*/
/* Get the preliminary new entry start value */
addr = stack_entry->start - grow_amount;
/*
* If this puts us into the previous entry, cut back our
* growth to the available space. Also, see the note above.
*/
if (addr < end) {
stack_entry->avail_ssize = max_grow;
addr = end;
}
rv = vm_map_insert(map, NULL, 0, addr, stack_entry->start,
p->p_sysent->sv_stackprot, VM_PROT_ALL, 0);
/* Adjust the available stack space by the amount we grew. */
if (rv == KERN_SUCCESS) {
if (prev_entry != &map->header)
vm_map_clip_end(map, prev_entry, addr);
new_entry = prev_entry->next;
KASSERT(new_entry == stack_entry->prev, ("foo"));
KASSERT(new_entry->end == stack_entry->start, ("foo"));
KASSERT(new_entry->start == addr, ("foo"));
grow_amount = new_entry->end - new_entry->start;
new_entry->avail_ssize = stack_entry->avail_ssize -
grow_amount;
stack_entry->eflags &= ~MAP_ENTRY_GROWS_DOWN;
new_entry->eflags |= MAP_ENTRY_GROWS_DOWN;
}
} else {
/*
* Growing upward.
*/
addr = stack_entry->end + grow_amount;
/*
* If this puts us into the next entry, cut back our growth
* to the available space. Also, see the note above.
*/
if (addr > end) {
stack_entry->avail_ssize = end - stack_entry->end;
addr = end;
}
grow_amount = addr - stack_entry->end;
/* Grow the underlying object if applicable. */
if (stack_entry->object.vm_object == NULL ||
vm_object_coalesce(stack_entry->object.vm_object,
stack_entry->offset,
(vm_size_t)(stack_entry->end - stack_entry->start),
(vm_size_t)grow_amount)) {
map->size += (addr - stack_entry->end);
/* Update the current entry. */
stack_entry->end = addr;
stack_entry->avail_ssize -= grow_amount;
rv = KERN_SUCCESS;
if (next_entry != &map->header)
vm_map_clip_start(map, next_entry, addr);
} else
rv = KERN_FAILURE;
}
if (rv == KERN_SUCCESS && is_procstack)
vm->vm_ssize += btoc(grow_amount);
vm_map_unlock(map);
/*
* Heed the MAP_WIREFUTURE flag if it was set for this process.
*/
if (rv == KERN_SUCCESS && (map->flags & MAP_WIREFUTURE)) {
vm_map_wire(map,
(stack_entry == next_entry) ? addr : addr - grow_amount,
(stack_entry == next_entry) ? stack_entry->start : addr,
(p->p_flag & P_SYSTEM)
? VM_MAP_WIRE_SYSTEM|VM_MAP_WIRE_NOHOLES
: VM_MAP_WIRE_USER|VM_MAP_WIRE_NOHOLES);
}
return (rv);
}
/*
* Unshare the specified VM space for exec. If other processes are
* mapped to it, then create a new one. The new vmspace is null.
*/
void
vmspace_exec(struct proc *p, vm_offset_t minuser, vm_offset_t maxuser)
{
struct vmspace *oldvmspace = p->p_vmspace;
struct vmspace *newvmspace;
GIANT_REQUIRED;
newvmspace = vmspace_alloc(minuser, maxuser);
newvmspace->vm_swrss = oldvmspace->vm_swrss;
/*
* This code is written like this for prototype purposes. The
* goal is to avoid running down the vmspace here, but let the
* other process's that are still using the vmspace to finally
* run it down. Even though there is little or no chance of blocking
* here, it is a good idea to keep this form for future mods.
*/
p->p_vmspace = newvmspace;
if (p == curthread->td_proc) /* XXXKSE ? */
pmap_activate(curthread);
vmspace_free(oldvmspace);
}
/*
* Unshare the specified VM space for forcing COW. This
* is called by rfork, for the (RFMEM|RFPROC) == 0 case.
*/
void
vmspace_unshare(struct proc *p)
{
struct vmspace *oldvmspace = p->p_vmspace;
struct vmspace *newvmspace;
GIANT_REQUIRED;
if (oldvmspace->vm_refcnt == 1)
return;
newvmspace = vmspace_fork(oldvmspace);
p->p_vmspace = newvmspace;
if (p == curthread->td_proc) /* XXXKSE ? */
pmap_activate(curthread);
vmspace_free(oldvmspace);
}
1994-05-24 10:09:53 +00:00
/*
* vm_map_lookup:
*
* Finds the VM object, offset, and
* protection for a given virtual address in the
* specified map, assuming a page fault of the
* type specified.
*
* Leaves the map in question locked for read; return
* values are guaranteed until a vm_map_lookup_done
* call is performed. Note that the map argument
* is in/out; the returned map must be used in
* the call to vm_map_lookup_done.
*
* A handle (out_entry) is returned for use in
* vm_map_lookup_done, to make that fast.
*
* If a lookup is requested with "write protection"
* specified, the map may be changed to perform virtual
* copying operations, although the data referenced will
* remain the same.
*/
int
vm_map_lookup(vm_map_t *var_map, /* IN/OUT */
vm_offset_t vaddr,
vm_prot_t fault_typea,
vm_map_entry_t *out_entry, /* OUT */
vm_object_t *object, /* OUT */
vm_pindex_t *pindex, /* OUT */
vm_prot_t *out_prot, /* OUT */
VM level code cleanups. 1) Start using TSM. Struct procs continue to point to upages structure, after being freed. Struct vmspace continues to point to pte object and kva space for kstack. u_map is now superfluous. 2) vm_map's don't need to be reference counted. They always exist either in the kernel or in a vmspace. The vmspaces are managed by reference counts. 3) Remove the "wired" vm_map nonsense. 4) No need to keep a cache of kernel stack kva's. 5) Get rid of strange looking ++var, and change to var++. 6) Change more data structures to use our "zone" allocator. Added struct proc, struct vmspace and struct vnode. This saves a significant amount of kva space and physical memory. Additionally, this enables TSM for the zone managed memory. 7) Keep ioopt disabled for now. 8) Remove the now bogus "single use" map concept. 9) Use generation counts or id's for data structures residing in TSM, where it allows us to avoid unneeded restart overhead during traversals, where blocking might occur. 10) Account better for memory deficits, so the pageout daemon will be able to make enough memory available (experimental.) 11) Fix some vnode locking problems. (From Tor, I think.) 12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp. (experimental.) 13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c code. Use generation counts, get rid of unneded collpase operations, and clean up the cluster code. 14) Make vm_zone more suitable for TSM. This commit is partially as a result of discussions and contributions from other people, including DG, Tor Egge, PHK, and probably others that I have forgotten to attribute (so let me know, if I forgot.) This is not the infamous, final cleanup of the vnode stuff, but a necessary step. Vnode mgmt should be correct, but things might still change, and there is still some missing stuff (like ioopt, and physical backing of non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
boolean_t *wired) /* OUT */
1994-05-24 10:09:53 +00:00
{
vm_map_entry_t entry;
vm_map_t map = *var_map;
vm_prot_t prot;
vm_prot_t fault_type = fault_typea;
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
RetryLookup:;
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
* Lookup the faulting address.
1994-05-24 10:09:53 +00:00
*/
vm_map_lock_read(map);
#define RETURN(why) \
{ \
vm_map_unlock_read(map); \
return (why); \
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
* If the map has an interesting hint, try it before calling full
* blown lookup routine.
1994-05-24 10:09:53 +00:00
*/
entry = map->root;
1994-05-24 10:09:53 +00:00
*out_entry = entry;
if (entry == NULL ||
1994-05-24 10:09:53 +00:00
(vaddr < entry->start) || (vaddr >= entry->end)) {
/*
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
* Entry was either not a valid hint, or the vaddr was not
* contained in the entry, so do a full lookup.
1994-05-24 10:09:53 +00:00
*/
if (!vm_map_lookup_entry(map, vaddr, out_entry))
1994-05-24 10:09:53 +00:00
RETURN(KERN_INVALID_ADDRESS);
entry = *out_entry;
1994-05-24 10:09:53 +00:00
}
2003-11-03 16:14:45 +00:00
1994-05-24 10:09:53 +00:00
/*
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
* Handle submaps.
1994-05-24 10:09:53 +00:00
*/
if (entry->eflags & MAP_ENTRY_IS_SUB_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
vm_map_t old_map = map;
1994-05-24 10:09:53 +00:00
*var_map = map = entry->object.sub_map;
vm_map_unlock_read(old_map);
goto RetryLookup;
}
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
* Check whether this task is allowed to have this page.
* Note the special case for MAP_ENTRY_COW
* pages with an override. This is to implement a forced
* COW for debuggers.
1994-05-24 10:09:53 +00:00
*/
1998-01-21 12:18:00 +00:00
if (fault_type & VM_PROT_OVERRIDE_WRITE)
prot = entry->max_protection;
else
prot = entry->protection;
fault_type &= (VM_PROT_READ|VM_PROT_WRITE|VM_PROT_EXECUTE);
if ((fault_type & prot) != fault_type) {
RETURN(KERN_PROTECTION_FAILURE);
}
if ((entry->eflags & MAP_ENTRY_USER_WIRED) &&
(entry->eflags & MAP_ENTRY_COW) &&
(fault_type & VM_PROT_WRITE) &&
(fault_typea & VM_PROT_OVERRIDE_WRITE) == 0) {
RETURN(KERN_PROTECTION_FAILURE);
}
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
* If this page is not pageable, we have to get it for all possible
* accesses.
1994-05-24 10:09:53 +00:00
*/
*wired = (entry->wired_count != 0);
if (*wired)
1994-05-24 10:09:53 +00:00
prot = fault_type = entry->protection;
/*
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 the entry was copy-on-write, we either ...
1994-05-24 10:09:53 +00:00
*/
if (entry->eflags & MAP_ENTRY_NEEDS_COPY) {
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 we want to write the page, we may as well handle that
* now since we've got the map locked.
1995-05-30 08:16:23 +00:00
*
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
* If we don't need to write the page, we just demote the
* permissions allowed.
1994-05-24 10:09:53 +00:00
*/
if (fault_type & VM_PROT_WRITE) {
/*
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
* Make a new object, and place it in the object
* chain. Note that no new references have appeared
* -- one just moved from the map to the new
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
* object.
1994-05-24 10:09:53 +00:00
*/
if (vm_map_lock_upgrade(map))
1994-05-24 10:09:53 +00:00
goto RetryLookup;
1994-05-24 10:09:53 +00:00
vm_object_shadow(
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
&entry->object.vm_object,
&entry->offset,
atop(entry->end - entry->start));
entry->eflags &= ~MAP_ENTRY_NEEDS_COPY;
vm_map_lock_downgrade(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
} else {
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
* We're attempting to read a copy-on-write page --
* don't allow writes.
1994-05-24 10:09:53 +00:00
*/
VM level code cleanups. 1) Start using TSM. Struct procs continue to point to upages structure, after being freed. Struct vmspace continues to point to pte object and kva space for kstack. u_map is now superfluous. 2) vm_map's don't need to be reference counted. They always exist either in the kernel or in a vmspace. The vmspaces are managed by reference counts. 3) Remove the "wired" vm_map nonsense. 4) No need to keep a cache of kernel stack kva's. 5) Get rid of strange looking ++var, and change to var++. 6) Change more data structures to use our "zone" allocator. Added struct proc, struct vmspace and struct vnode. This saves a significant amount of kva space and physical memory. Additionally, this enables TSM for the zone managed memory. 7) Keep ioopt disabled for now. 8) Remove the now bogus "single use" map concept. 9) Use generation counts or id's for data structures residing in TSM, where it allows us to avoid unneeded restart overhead during traversals, where blocking might occur. 10) Account better for memory deficits, so the pageout daemon will be able to make enough memory available (experimental.) 11) Fix some vnode locking problems. (From Tor, I think.) 12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp. (experimental.) 13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c code. Use generation counts, get rid of unneded collpase operations, and clean up the cluster code. 14) Make vm_zone more suitable for TSM. This commit is partially as a result of discussions and contributions from other people, including DG, Tor Egge, PHK, and probably others that I have forgotten to attribute (so let me know, if I forgot.) This is not the infamous, final cleanup of the vnode stuff, but a necessary step. Vnode mgmt should be correct, but things might still change, and there is still some missing stuff (like ioopt, and physical backing of non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
prot &= ~VM_PROT_WRITE;
1994-05-24 10:09:53 +00:00
}
}
VM level code cleanups. 1) Start using TSM. Struct procs continue to point to upages structure, after being freed. Struct vmspace continues to point to pte object and kva space for kstack. u_map is now superfluous. 2) vm_map's don't need to be reference counted. They always exist either in the kernel or in a vmspace. The vmspaces are managed by reference counts. 3) Remove the "wired" vm_map nonsense. 4) No need to keep a cache of kernel stack kva's. 5) Get rid of strange looking ++var, and change to var++. 6) Change more data structures to use our "zone" allocator. Added struct proc, struct vmspace and struct vnode. This saves a significant amount of kva space and physical memory. Additionally, this enables TSM for the zone managed memory. 7) Keep ioopt disabled for now. 8) Remove the now bogus "single use" map concept. 9) Use generation counts or id's for data structures residing in TSM, where it allows us to avoid unneeded restart overhead during traversals, where blocking might occur. 10) Account better for memory deficits, so the pageout daemon will be able to make enough memory available (experimental.) 11) Fix some vnode locking problems. (From Tor, I think.) 12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp. (experimental.) 13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c code. Use generation counts, get rid of unneded collpase operations, and clean up the cluster code. 14) Make vm_zone more suitable for TSM. This commit is partially as a result of discussions and contributions from other people, including DG, Tor Egge, PHK, and probably others that I have forgotten to attribute (so let me know, if I forgot.) This is not the infamous, final cleanup of the vnode stuff, but a necessary step. Vnode mgmt should be correct, but things might still change, and there is still some missing stuff (like ioopt, and physical backing of non-merged cache files, debugging of layering concepts.)
1998-01-22 17:30:44 +00:00
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
* Create an object if necessary.
1994-05-24 10:09:53 +00:00
*/
if (entry->object.vm_object == NULL &&
!map->system_map) {
2003-11-03 16:14:45 +00:00
if (vm_map_lock_upgrade(map))
1994-05-24 10:09:53 +00:00
goto RetryLookup;
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
entry->object.vm_object = vm_object_allocate(OBJT_DEFAULT,
atop(entry->end - entry->start));
1994-05-24 10:09:53 +00:00
entry->offset = 0;
vm_map_lock_downgrade(map);
1994-05-24 10:09:53 +00:00
}
1994-05-24 10:09:53 +00:00
/*
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
* Return the object/offset from this entry. If the entry was
* copy-on-write or empty, it has been fixed up.
1994-05-24 10:09:53 +00:00
*/
*pindex = OFF_TO_IDX((vaddr - entry->start) + entry->offset);
1994-05-24 10:09:53 +00:00
*object = entry->object.vm_object;
*out_prot = prot;
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 (KERN_SUCCESS);
1994-05-24 10:09:53 +00:00
#undef RETURN
}
/*
* vm_map_lookup_done:
*
* Releases locks acquired by a vm_map_lookup
* (according to the handle returned by that lookup).
*/
1995-05-30 08:16:23 +00:00
void
vm_map_lookup_done(vm_map_t map, vm_map_entry_t entry)
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
* Unlock the main-level map
1994-05-24 10:09:53 +00:00
*/
vm_map_unlock_read(map);
}
#include "opt_ddb.h"
#ifdef DDB
#include <sys/kernel.h>
#include <ddb/ddb.h>
1994-05-24 10:09:53 +00:00
/*
* vm_map_print: [ debug ]
*/
DB_SHOW_COMMAND(map, vm_map_print)
1994-05-24 10:09:53 +00:00
{
static int nlines;
/* XXX convert args. */
vm_map_t map = (vm_map_t)addr;
boolean_t full = have_addr;
vm_map_entry_t entry;
1994-05-24 10:09:53 +00:00
db_iprintf("Task map %p: pmap=%p, nentries=%d, version=%u\n",
(void *)map,
(void *)map->pmap, map->nentries, map->timestamp);
nlines++;
1994-05-24 10:09:53 +00:00
if (!full && db_indent)
1994-05-24 10:09:53 +00:00
return;
db_indent += 2;
1994-05-24 10:09:53 +00:00
for (entry = map->header.next; entry != &map->header;
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
entry = entry->next) {
1998-07-11 11:30:46 +00:00
db_iprintf("map entry %p: start=%p, end=%p\n",
(void *)entry, (void *)entry->start, (void *)entry->end);
nlines++;
{
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
static char *inheritance_name[4] =
{"share", "copy", "none", "donate_copy"};
db_iprintf(" prot=%x/%x/%s",
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
entry->protection,
entry->max_protection,
inheritance_name[(int)(unsigned char)entry->inheritance]);
1994-05-24 10:09:53 +00:00
if (entry->wired_count != 0)
db_printf(", wired");
1994-05-24 10:09:53 +00:00
}
if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) {
2002-11-07 22:49:07 +00:00
db_printf(", share=%p, offset=0x%jx\n",
(void *)entry->object.sub_map,
2002-11-07 22:49:07 +00:00
(uintmax_t)entry->offset);
nlines++;
1994-05-24 10:09:53 +00:00
if ((entry->prev == &map->header) ||
(entry->prev->object.sub_map !=
entry->object.sub_map)) {
db_indent += 2;
vm_map_print((db_expr_t)(intptr_t)
entry->object.sub_map,
full, 0, (char *)0);
db_indent -= 2;
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
} else {
2002-11-07 22:49:07 +00:00
db_printf(", object=%p, offset=0x%jx",
(void *)entry->object.vm_object,
2002-11-07 22:49:07 +00:00
(uintmax_t)entry->offset);
if (entry->eflags & MAP_ENTRY_COW)
db_printf(", copy (%s)",
(entry->eflags & MAP_ENTRY_NEEDS_COPY) ? "needed" : "done");
db_printf("\n");
nlines++;
1994-05-24 10:09:53 +00:00
if ((entry->prev == &map->header) ||
(entry->prev->object.vm_object !=
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
entry->object.vm_object)) {
db_indent += 2;
vm_object_print((db_expr_t)(intptr_t)
entry->object.vm_object,
full, 0, (char *)0);
nlines += 4;
db_indent -= 2;
1994-05-24 10:09:53 +00:00
}
}
}
db_indent -= 2;
if (db_indent == 0)
nlines = 0;
1994-05-24 10:09:53 +00:00
}
DB_SHOW_COMMAND(procvm, procvm)
{
struct proc *p;
if (have_addr) {
p = (struct proc *) addr;
} else {
p = curproc;
}
1998-07-11 07:46:16 +00:00
db_printf("p = %p, vmspace = %p, map = %p, pmap = %p\n",
(void *)p, (void *)p->p_vmspace, (void *)&p->p_vmspace->vm_map,
(void *)vmspace_pmap(p->p_vmspace));
vm_map_print((db_expr_t)(intptr_t)&p->p_vmspace->vm_map, 1, 0, NULL);
}
#endif /* DDB */