freebsd-skq/sys/vm/vm_object.c

2396 lines
63 KiB
C
Raw Normal View History

/*-
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_object.c 8.5 (Berkeley) 3/22/94
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 object module.
*/
2003-06-11 23:50:51 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_vm.h"
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/lock.h>
#include <sys/mman.h>
#include <sys/mount.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/mutex.h>
#include <sys/proc.h> /* for curproc, pageproc */
#include <sys/socket.h>
#include <sys/resourcevar.h>
#include <sys/vnode.h>
#include <sys/vmmeter.h>
#include <sys/sx.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>
#include <vm/vm_object.h>
1994-05-24 10:09:53 +00:00
#include <vm/vm_page.h>
#include <vm/vm_pageout.h>
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
#include <vm/vm_pager.h>
#include <vm/swap_pager.h>
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
#include <vm/vm_kern.h>
#include <vm/vm_extern.h>
#include <vm/vm_reserv.h>
#include <vm/uma.h>
static int old_msync;
SYSCTL_INT(_vm, OID_AUTO, old_msync, CTLFLAG_RW, &old_msync, 0,
"Use old (insecure) msync behavior");
static int vm_object_page_collect_flush(vm_object_t object, vm_page_t p,
int pagerflags, int flags, boolean_t *clearobjflags,
boolean_t *eio);
static boolean_t vm_object_page_remove_write(vm_page_t p, int flags,
boolean_t *clearobjflags);
static void vm_object_qcollapse(vm_object_t object);
static void vm_object_vndeallocate(vm_object_t object);
1994-05-24 10:09:53 +00:00
/*
* Virtual memory objects maintain the actual data
* associated with allocated virtual memory. A given
* page of memory exists within exactly one object.
*
* An object is only deallocated when all "references"
* are given up. Only one "reference" to a given
* region of an object should be writeable.
*
* Associated with each object is a list of all resident
* memory pages belonging to that object; this list is
* maintained by the "vm_page" module, and locked by the object's
* lock.
*
* Each object also records a "pager" routine which is
* used to retrieve (and store) pages to the proper backing
* storage. In addition, objects may be backed by other
* objects from which they were virtual-copied.
*
* The only items within the object structure which are
* modified after time of creation are:
* reference count locked by object's lock
* pager routine locked by object's lock
*
*/
struct object_q vm_object_list;
struct mtx vm_object_list_mtx; /* lock for object list and count */
struct vm_object kernel_object_store;
struct vm_object kmem_object_store;
1994-05-24 10:09:53 +00:00
static SYSCTL_NODE(_vm_stats, OID_AUTO, object, CTLFLAG_RD, 0,
"VM object stats");
static long object_collapses;
SYSCTL_LONG(_vm_stats_object, OID_AUTO, collapses, CTLFLAG_RD,
&object_collapses, 0, "VM object collapses");
static long object_bypasses;
SYSCTL_LONG(_vm_stats_object, OID_AUTO, bypasses, CTLFLAG_RD,
&object_bypasses, 0, "VM object bypasses");
static uma_zone_t obj_zone;
static int vm_object_zinit(void *mem, int size, int flags);
#ifdef INVARIANTS
static void vm_object_zdtor(void *mem, int size, void *arg);
static void
vm_object_zdtor(void *mem, int size, void *arg)
{
vm_object_t object;
object = (vm_object_t)mem;
KASSERT(TAILQ_EMPTY(&object->memq),
("object %p has resident pages",
object));
#if VM_NRESERVLEVEL > 0
KASSERT(LIST_EMPTY(&object->rvq),
("object %p has reservations",
object));
#endif
Change the management of cached pages (PQ_CACHE) in two fundamental ways: (1) Cached pages are no longer kept in the object's resident page splay tree and memq. Instead, they are kept in a separate per-object splay tree of cached pages. However, access to this new per-object splay tree is synchronized by the _free_ page queues lock, not to be confused with the heavily contended page queues lock. Consequently, a cached page can be reclaimed by vm_page_alloc(9) without acquiring the object's lock or the page queues lock. This solves a problem independently reported by tegge@ and Isilon. Specifically, they observed the page daemon consuming a great deal of CPU time because of pages bouncing back and forth between the cache queue (PQ_CACHE) and the inactive queue (PQ_INACTIVE). The source of this problem turned out to be a deadlock avoidance strategy employed when selecting a cached page to reclaim in vm_page_select_cache(). However, the root cause was really that reclaiming a cached page required the acquisition of an object lock while the page queues lock was already held. Thus, this change addresses the problem at its root, by eliminating the need to acquire the object's lock. Moreover, keeping cached pages in the object's primary splay tree and memq was, in effect, optimizing for the uncommon case. Cached pages are reclaimed far, far more often than they are reactivated. Instead, this change makes reclamation cheaper, especially in terms of synchronization overhead, and reactivation more expensive, because reactivated pages will have to be reentered into the object's primary splay tree and memq. (2) Cached pages are now stored alongside free pages in the physical memory allocator's buddy queues, increasing the likelihood that large allocations of contiguous physical memory (i.e., superpages) will succeed. Finally, as a result of this change long-standing restrictions on when and where a cached page can be reclaimed and returned by vm_page_alloc(9) are eliminated. Specifically, calls to vm_page_alloc(9) specifying VM_ALLOC_INTERRUPT can now reclaim and return a formerly cached page. Consequently, a call to malloc(9) specifying M_NOWAIT is less likely to fail. Discussed with: many over the course of the summer, including jeff@, Justin Husted @ Isilon, peter@, tegge@ Tested by: an earlier version by kris@ Approved by: re (kensmith)
2007-09-25 06:25:06 +00:00
KASSERT(object->cache == NULL,
("object %p has cached pages",
object));
KASSERT(object->paging_in_progress == 0,
("object %p paging_in_progress = %d",
object, object->paging_in_progress));
KASSERT(object->resident_page_count == 0,
("object %p resident_page_count = %d",
object, object->resident_page_count));
KASSERT(object->shadow_count == 0,
("object %p shadow_count = %d",
object, object->shadow_count));
}
#endif
static int
vm_object_zinit(void *mem, int size, int flags)
{
vm_object_t object;
object = (vm_object_t)mem;
bzero(&object->mtx, sizeof(object->mtx));
mtx_init(&object->mtx, "vm object", NULL, MTX_DEF | MTX_DUPOK);
/* These are true for any object that has been freed */
object->paging_in_progress = 0;
object->resident_page_count = 0;
object->shadow_count = 0;
return (0);
}
1994-05-24 10:09:53 +00:00
static void
_vm_object_allocate(objtype_t type, vm_pindex_t size, vm_object_t object)
{
TAILQ_INIT(&object->memq);
LIST_INIT(&object->shadow_head);
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
object->root = NULL;
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
object->type = type;
switch (type) {
case OBJT_DEAD:
panic("_vm_object_allocate: can't create OBJT_DEAD");
case OBJT_DEFAULT:
case OBJT_SWAP:
object->flags = OBJ_ONEMAPPING;
break;
case OBJT_DEVICE:
case OBJT_SG:
object->flags = OBJ_FICTITIOUS | OBJ_UNMANAGED;
break;
case OBJT_MGTDEVICE:
object->flags = OBJ_FICTITIOUS;
break;
case OBJT_PHYS:
object->flags = OBJ_UNMANAGED;
break;
case OBJT_VNODE:
object->flags = 0;
break;
default:
panic("_vm_object_allocate: type %d is undefined", type);
}
object->size = size;
object->generation = 1;
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
object->ref_count = 1;
Add support to the virtual memory system for configuring machine- dependent memory attributes: Rename vm_cache_mode_t to vm_memattr_t. The new name reflects the fact that there are machine-dependent memory attributes that have nothing to do with controlling the cache's behavior. Introduce vm_object_set_memattr() for setting the default memory attributes that will be given to an object's pages. Introduce and use pmap_page_{get,set}_memattr() for getting and setting a page's machine-dependent memory attributes. Add full support for these functions on amd64 and i386 and stubs for them on the other architectures. The function pmap_page_set_memattr() is also responsible for any other machine-dependent aspects of changing a page's memory attributes, such as flushing the cache or updating the direct map. The uses include kmem_alloc_contig(), vm_page_alloc(), and the device pager: kmem_alloc_contig() can now be used to allocate kernel memory with non-default memory attributes on amd64 and i386. vm_page_alloc() and the device pager will set the memory attributes for the real or fictitious page according to the object's default memory attributes. Update the various pmap functions on amd64 and i386 that map pages to incorporate each page's memory attributes in the mapping. Notes: (1) Inherent to this design are safety features that prevent the specification of inconsistent memory attributes by different mappings on amd64 and i386. In addition, the device pager provides a warning when a device driver creates a fictitious page with memory attributes that are inconsistent with the real page that the fictitious page is an alias for. (2) Storing the machine-dependent memory attributes for amd64 and i386 as a dedicated "int" in "struct md_page" represents a compromise between space efficiency and the ease of MFCing these changes to RELENG_7. In collaboration with: jhb Approved by: re (kib)
2009-07-12 23:31:20 +00:00
object->memattr = VM_MEMATTR_DEFAULT;
object->cred = NULL;
object->charge = 0;
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
object->handle = NULL;
object->backing_object = NULL;
object->backing_object_offset = (vm_ooffset_t) 0;
#if VM_NRESERVLEVEL > 0
LIST_INIT(&object->rvq);
#endif
Change the management of cached pages (PQ_CACHE) in two fundamental ways: (1) Cached pages are no longer kept in the object's resident page splay tree and memq. Instead, they are kept in a separate per-object splay tree of cached pages. However, access to this new per-object splay tree is synchronized by the _free_ page queues lock, not to be confused with the heavily contended page queues lock. Consequently, a cached page can be reclaimed by vm_page_alloc(9) without acquiring the object's lock or the page queues lock. This solves a problem independently reported by tegge@ and Isilon. Specifically, they observed the page daemon consuming a great deal of CPU time because of pages bouncing back and forth between the cache queue (PQ_CACHE) and the inactive queue (PQ_INACTIVE). The source of this problem turned out to be a deadlock avoidance strategy employed when selecting a cached page to reclaim in vm_page_select_cache(). However, the root cause was really that reclaiming a cached page required the acquisition of an object lock while the page queues lock was already held. Thus, this change addresses the problem at its root, by eliminating the need to acquire the object's lock. Moreover, keeping cached pages in the object's primary splay tree and memq was, in effect, optimizing for the uncommon case. Cached pages are reclaimed far, far more often than they are reactivated. Instead, this change makes reclamation cheaper, especially in terms of synchronization overhead, and reactivation more expensive, because reactivated pages will have to be reentered into the object's primary splay tree and memq. (2) Cached pages are now stored alongside free pages in the physical memory allocator's buddy queues, increasing the likelihood that large allocations of contiguous physical memory (i.e., superpages) will succeed. Finally, as a result of this change long-standing restrictions on when and where a cached page can be reclaimed and returned by vm_page_alloc(9) are eliminated. Specifically, calls to vm_page_alloc(9) specifying VM_ALLOC_INTERRUPT can now reclaim and return a formerly cached page. Consequently, a call to malloc(9) specifying M_NOWAIT is less likely to fail. Discussed with: many over the course of the summer, including jeff@, Justin Husted @ Isilon, peter@, tegge@ Tested by: an earlier version by kris@ Approved by: re (kensmith)
2007-09-25 06:25:06 +00:00
object->cache = NULL;
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
mtx_lock(&vm_object_list_mtx);
TAILQ_INSERT_TAIL(&vm_object_list, object, object_list);
mtx_unlock(&vm_object_list_mtx);
}
1994-05-24 10:09:53 +00:00
/*
* vm_object_init:
*
* Initialize the VM objects module.
*/
void
vm_object_init(void)
1994-05-24 10:09:53 +00:00
{
TAILQ_INIT(&vm_object_list);
mtx_init(&vm_object_list_mtx, "vm object_list", NULL, MTX_DEF);
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
mtx_init(&kernel_object->mtx, "vm object", "kernel object", MTX_DEF);
_vm_object_allocate(OBJT_PHYS, OFF_TO_IDX(VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS),
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
kernel_object);
#if VM_NRESERVLEVEL > 0
kernel_object->flags |= OBJ_COLORED;
kernel_object->pg_color = (u_short)atop(VM_MIN_KERNEL_ADDRESS);
#endif
1994-05-24 10:09:53 +00:00
mtx_init(&kmem_object->mtx, "vm object", "kmem object", MTX_DEF);
_vm_object_allocate(OBJT_PHYS, OFF_TO_IDX(VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS),
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
kmem_object);
#if VM_NRESERVLEVEL > 0
kmem_object->flags |= OBJ_COLORED;
kmem_object->pg_color = (u_short)atop(VM_MIN_KERNEL_ADDRESS);
#endif
/*
* The lock portion of struct vm_object must be type stable due
* to vm_pageout_fallback_object_lock locking a vm object
* without holding any references to it.
*/
obj_zone = uma_zcreate("VM OBJECT", sizeof (struct vm_object), NULL,
#ifdef INVARIANTS
vm_object_zdtor,
#else
NULL,
#endif
vm_object_zinit, NULL, UMA_ALIGN_PTR, UMA_ZONE_VM|UMA_ZONE_NOFREE);
}
void
vm_object_clear_flag(vm_object_t object, u_short bits)
{
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
object->flags &= ~bits;
}
Add support to the virtual memory system for configuring machine- dependent memory attributes: Rename vm_cache_mode_t to vm_memattr_t. The new name reflects the fact that there are machine-dependent memory attributes that have nothing to do with controlling the cache's behavior. Introduce vm_object_set_memattr() for setting the default memory attributes that will be given to an object's pages. Introduce and use pmap_page_{get,set}_memattr() for getting and setting a page's machine-dependent memory attributes. Add full support for these functions on amd64 and i386 and stubs for them on the other architectures. The function pmap_page_set_memattr() is also responsible for any other machine-dependent aspects of changing a page's memory attributes, such as flushing the cache or updating the direct map. The uses include kmem_alloc_contig(), vm_page_alloc(), and the device pager: kmem_alloc_contig() can now be used to allocate kernel memory with non-default memory attributes on amd64 and i386. vm_page_alloc() and the device pager will set the memory attributes for the real or fictitious page according to the object's default memory attributes. Update the various pmap functions on amd64 and i386 that map pages to incorporate each page's memory attributes in the mapping. Notes: (1) Inherent to this design are safety features that prevent the specification of inconsistent memory attributes by different mappings on amd64 and i386. In addition, the device pager provides a warning when a device driver creates a fictitious page with memory attributes that are inconsistent with the real page that the fictitious page is an alias for. (2) Storing the machine-dependent memory attributes for amd64 and i386 as a dedicated "int" in "struct md_page" represents a compromise between space efficiency and the ease of MFCing these changes to RELENG_7. In collaboration with: jhb Approved by: re (kib)
2009-07-12 23:31:20 +00:00
/*
* Sets the default memory attribute for the specified object. Pages
* that are allocated to this object are by default assigned this memory
* attribute.
*
* Presently, this function must be called before any pages are allocated
* to the object. In the future, this requirement may be relaxed for
* "default" and "swap" objects.
*/
int
vm_object_set_memattr(vm_object_t object, vm_memattr_t memattr)
{
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
switch (object->type) {
case OBJT_DEFAULT:
case OBJT_DEVICE:
case OBJT_MGTDEVICE:
Add support to the virtual memory system for configuring machine- dependent memory attributes: Rename vm_cache_mode_t to vm_memattr_t. The new name reflects the fact that there are machine-dependent memory attributes that have nothing to do with controlling the cache's behavior. Introduce vm_object_set_memattr() for setting the default memory attributes that will be given to an object's pages. Introduce and use pmap_page_{get,set}_memattr() for getting and setting a page's machine-dependent memory attributes. Add full support for these functions on amd64 and i386 and stubs for them on the other architectures. The function pmap_page_set_memattr() is also responsible for any other machine-dependent aspects of changing a page's memory attributes, such as flushing the cache or updating the direct map. The uses include kmem_alloc_contig(), vm_page_alloc(), and the device pager: kmem_alloc_contig() can now be used to allocate kernel memory with non-default memory attributes on amd64 and i386. vm_page_alloc() and the device pager will set the memory attributes for the real or fictitious page according to the object's default memory attributes. Update the various pmap functions on amd64 and i386 that map pages to incorporate each page's memory attributes in the mapping. Notes: (1) Inherent to this design are safety features that prevent the specification of inconsistent memory attributes by different mappings on amd64 and i386. In addition, the device pager provides a warning when a device driver creates a fictitious page with memory attributes that are inconsistent with the real page that the fictitious page is an alias for. (2) Storing the machine-dependent memory attributes for amd64 and i386 as a dedicated "int" in "struct md_page" represents a compromise between space efficiency and the ease of MFCing these changes to RELENG_7. In collaboration with: jhb Approved by: re (kib)
2009-07-12 23:31:20 +00:00
case OBJT_PHYS:
case OBJT_SG:
Add support to the virtual memory system for configuring machine- dependent memory attributes: Rename vm_cache_mode_t to vm_memattr_t. The new name reflects the fact that there are machine-dependent memory attributes that have nothing to do with controlling the cache's behavior. Introduce vm_object_set_memattr() for setting the default memory attributes that will be given to an object's pages. Introduce and use pmap_page_{get,set}_memattr() for getting and setting a page's machine-dependent memory attributes. Add full support for these functions on amd64 and i386 and stubs for them on the other architectures. The function pmap_page_set_memattr() is also responsible for any other machine-dependent aspects of changing a page's memory attributes, such as flushing the cache or updating the direct map. The uses include kmem_alloc_contig(), vm_page_alloc(), and the device pager: kmem_alloc_contig() can now be used to allocate kernel memory with non-default memory attributes on amd64 and i386. vm_page_alloc() and the device pager will set the memory attributes for the real or fictitious page according to the object's default memory attributes. Update the various pmap functions on amd64 and i386 that map pages to incorporate each page's memory attributes in the mapping. Notes: (1) Inherent to this design are safety features that prevent the specification of inconsistent memory attributes by different mappings on amd64 and i386. In addition, the device pager provides a warning when a device driver creates a fictitious page with memory attributes that are inconsistent with the real page that the fictitious page is an alias for. (2) Storing the machine-dependent memory attributes for amd64 and i386 as a dedicated "int" in "struct md_page" represents a compromise between space efficiency and the ease of MFCing these changes to RELENG_7. In collaboration with: jhb Approved by: re (kib)
2009-07-12 23:31:20 +00:00
case OBJT_SWAP:
case OBJT_VNODE:
if (!TAILQ_EMPTY(&object->memq))
return (KERN_FAILURE);
break;
case OBJT_DEAD:
return (KERN_INVALID_ARGUMENT);
default:
panic("vm_object_set_memattr: object %p is of undefined type",
object);
Add support to the virtual memory system for configuring machine- dependent memory attributes: Rename vm_cache_mode_t to vm_memattr_t. The new name reflects the fact that there are machine-dependent memory attributes that have nothing to do with controlling the cache's behavior. Introduce vm_object_set_memattr() for setting the default memory attributes that will be given to an object's pages. Introduce and use pmap_page_{get,set}_memattr() for getting and setting a page's machine-dependent memory attributes. Add full support for these functions on amd64 and i386 and stubs for them on the other architectures. The function pmap_page_set_memattr() is also responsible for any other machine-dependent aspects of changing a page's memory attributes, such as flushing the cache or updating the direct map. The uses include kmem_alloc_contig(), vm_page_alloc(), and the device pager: kmem_alloc_contig() can now be used to allocate kernel memory with non-default memory attributes on amd64 and i386. vm_page_alloc() and the device pager will set the memory attributes for the real or fictitious page according to the object's default memory attributes. Update the various pmap functions on amd64 and i386 that map pages to incorporate each page's memory attributes in the mapping. Notes: (1) Inherent to this design are safety features that prevent the specification of inconsistent memory attributes by different mappings on amd64 and i386. In addition, the device pager provides a warning when a device driver creates a fictitious page with memory attributes that are inconsistent with the real page that the fictitious page is an alias for. (2) Storing the machine-dependent memory attributes for amd64 and i386 as a dedicated "int" in "struct md_page" represents a compromise between space efficiency and the ease of MFCing these changes to RELENG_7. In collaboration with: jhb Approved by: re (kib)
2009-07-12 23:31:20 +00:00
}
object->memattr = memattr;
return (KERN_SUCCESS);
}
void
vm_object_pip_add(vm_object_t object, short i)
{
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
object->paging_in_progress += i;
}
void
vm_object_pip_subtract(vm_object_t object, short i)
{
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
object->paging_in_progress -= i;
}
void
vm_object_pip_wakeup(vm_object_t object)
{
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
object->paging_in_progress--;
if ((object->flags & OBJ_PIPWNT) && object->paging_in_progress == 0) {
vm_object_clear_flag(object, OBJ_PIPWNT);
wakeup(object);
}
}
void
vm_object_pip_wakeupn(vm_object_t object, short i)
{
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
if (i)
object->paging_in_progress -= i;
if ((object->flags & OBJ_PIPWNT) && object->paging_in_progress == 0) {
vm_object_clear_flag(object, OBJ_PIPWNT);
wakeup(object);
}
}
void
vm_object_pip_wait(vm_object_t object, char *waitid)
{
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
while (object->paging_in_progress) {
object->flags |= OBJ_PIPWNT;
VM_OBJECT_SLEEP(object, object, PVM, waitid, 0);
}
}
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
/*
* vm_object_allocate:
*
* Returns a new object with the given size.
*/
vm_object_t
vm_object_allocate(objtype_t type, vm_pindex_t size)
{
vm_object_t object;
object = (vm_object_t)uma_zalloc(obj_zone, M_WAITOK);
_vm_object_allocate(type, size, object);
return (object);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
1994-05-24 10:09:53 +00:00
/*
* vm_object_reference:
*
* Gets another reference to the given object. Note: OBJ_DEAD
* objects can be referenced during final cleaning.
1994-05-24 10:09:53 +00:00
*/
void
vm_object_reference(vm_object_t object)
1994-05-24 10:09:53 +00:00
{
if (object == NULL)
return;
VM_OBJECT_LOCK(object);
vm_object_reference_locked(object);
VM_OBJECT_UNLOCK(object);
}
/*
* vm_object_reference_locked:
*
* Gets another reference to the given object.
*
* The object must be locked.
*/
void
vm_object_reference_locked(vm_object_t object)
{
struct vnode *vp;
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
object->ref_count++;
if (object->type == OBJT_VNODE) {
vp = object->handle;
vref(vp);
}
}
/*
* Handle deallocating an object of type OBJT_VNODE.
*/
static void
vm_object_vndeallocate(vm_object_t object)
{
struct vnode *vp = (struct vnode *) object->handle;
1999-01-10 01:58:29 +00:00
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
KASSERT(object->type == OBJT_VNODE,
1999-01-10 01:58:29 +00:00
("vm_object_vndeallocate: not a vnode object"));
KASSERT(vp != NULL, ("vm_object_vndeallocate: missing vp"));
#ifdef INVARIANTS
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
if (object->ref_count == 0) {
vprint("vm_object_vndeallocate", vp);
panic("vm_object_vndeallocate: bad object reference count");
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
}
#endif
if (object->ref_count > 1) {
object->ref_count--;
VM_OBJECT_UNLOCK(object);
/* vrele may need the vnode lock. */
vrele(vp);
} else {
vhold(vp);
VM_OBJECT_UNLOCK(object);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
vdrop(vp);
VM_OBJECT_LOCK(object);
object->ref_count--;
if (object->type == OBJT_DEAD) {
VM_OBJECT_UNLOCK(object);
VOP_UNLOCK(vp, 0);
} else {
if (object->ref_count == 0)
VOP_UNSET_TEXT(vp);
VM_OBJECT_UNLOCK(object);
vput(vp);
}
}
1994-05-24 10:09:53 +00:00
}
/*
* vm_object_deallocate:
*
* Release a reference to the specified object,
* gained either through a vm_object_allocate
* or a vm_object_reference call. When all references
* are gone, storage associated with this object
* may be relinquished.
*
* No object may be locked.
*/
void
vm_object_deallocate(vm_object_t object)
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_object_t temp;
1994-05-24 10:09:53 +00:00
while (object != NULL) {
VM_OBJECT_LOCK(object);
if (object->type == OBJT_VNODE) {
vm_object_vndeallocate(object);
return;
}
KASSERT(object->ref_count != 0,
("vm_object_deallocate: object deallocated too many times: %d", object->type));
/*
* If the reference count goes to 0 we start calling
* vm_object_terminate() on the object chain.
* A ref count of 1 may be a special case depending on the
* shadow count being 0 or 1.
*/
object->ref_count--;
if (object->ref_count > 1) {
VM_OBJECT_UNLOCK(object);
return;
} else if (object->ref_count == 1) {
if (object->shadow_count == 0 &&
object->handle == NULL &&
(object->type == OBJT_DEFAULT ||
object->type == OBJT_SWAP)) {
vm_object_set_flag(object, OBJ_ONEMAPPING);
} else if ((object->shadow_count == 1) &&
(object->handle == NULL) &&
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
(object->type == OBJT_DEFAULT ||
object->type == OBJT_SWAP)) {
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_object_t robject;
robject = LIST_FIRST(&object->shadow_head);
KASSERT(robject != NULL,
1999-01-10 01:58:29 +00:00
("vm_object_deallocate: ref_count: %d, shadow_count: %d",
object->ref_count,
object->shadow_count));
if (!VM_OBJECT_TRYLOCK(robject)) {
/*
* Avoid a potential deadlock.
*/
object->ref_count++;
VM_OBJECT_UNLOCK(object);
/*
* More likely than not the thread
* holding robject's lock has lower
* priority than the current thread.
* Let the lower priority thread run.
*/
pause("vmo_de", 1);
continue;
}
/*
* Collapse object into its shadow unless its
* shadow is dead. In that case, object will
* be deallocated by the thread that is
* deallocating its shadow.
*/
if ((robject->flags & OBJ_DEAD) == 0 &&
(robject->handle == NULL) &&
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
(robject->type == OBJT_DEFAULT ||
robject->type == OBJT_SWAP)) {
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
robject->ref_count++;
retry:
if (robject->paging_in_progress) {
VM_OBJECT_UNLOCK(object);
vm_object_pip_wait(robject,
"objde1");
temp = robject->backing_object;
if (object == temp) {
VM_OBJECT_LOCK(object);
goto retry;
}
} else if (object->paging_in_progress) {
VM_OBJECT_UNLOCK(robject);
object->flags |= OBJ_PIPWNT;
VM_OBJECT_SLEEP(object, object,
PDROP | PVM, "objde2", 0);
VM_OBJECT_LOCK(robject);
temp = robject->backing_object;
if (object == temp) {
VM_OBJECT_LOCK(object);
goto retry;
}
} else
VM_OBJECT_UNLOCK(object);
if (robject->ref_count == 1) {
robject->ref_count--;
object = robject;
goto doterm;
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 = robject;
vm_object_collapse(object);
VM_OBJECT_UNLOCK(object);
continue;
}
VM_OBJECT_UNLOCK(robject);
}
VM_OBJECT_UNLOCK(object);
return;
}
doterm:
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
temp = object->backing_object;
if (temp != NULL) {
VM_OBJECT_LOCK(temp);
LIST_REMOVE(object, shadow_list);
temp->shadow_count--;
VM_OBJECT_UNLOCK(temp);
object->backing_object = NULL;
}
/*
* Don't double-terminate, we could be in a termination
* recursion due to the terminate having to sync data
* to disk.
*/
if ((object->flags & OBJ_DEAD) == 0)
vm_object_terminate(object);
else
VM_OBJECT_UNLOCK(object);
1994-05-24 10:09:53 +00:00
object = temp;
}
}
/*
* vm_object_destroy removes the object from the global object list
* and frees the space for the object.
*/
void
vm_object_destroy(vm_object_t object)
{
/*
* Remove the object from the global object list.
*/
mtx_lock(&vm_object_list_mtx);
TAILQ_REMOVE(&vm_object_list, object, object_list);
mtx_unlock(&vm_object_list_mtx);
/*
* Release the allocation charge.
*/
if (object->cred != NULL) {
KASSERT(object->type == OBJT_DEFAULT ||
object->type == OBJT_SWAP,
("vm_object_terminate: non-swap obj %p has cred",
object));
swap_release_by_cred(object->charge, object->cred);
object->charge = 0;
crfree(object->cred);
object->cred = NULL;
}
/*
* Free the space for the object.
*/
uma_zfree(obj_zone, object);
}
1994-05-24 10:09:53 +00:00
/*
* vm_object_terminate actually destroys the specified object, freeing
* up all previously used resources.
*
* The object must be locked.
* This routine may block.
1994-05-24 10:09:53 +00:00
*/
void
vm_object_terminate(vm_object_t object)
1994-05-24 10:09:53 +00:00
{
vm_page_t p, p_next;
1994-05-24 10:09:53 +00:00
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
/*
* Make sure no one uses us.
*/
vm_object_set_flag(object, OBJ_DEAD);
1994-05-24 10:09:53 +00:00
/*
* wait for the pageout daemon to be done with the object
1994-05-24 10:09:53 +00:00
*/
vm_object_pip_wait(object, "objtrm");
1994-05-24 10:09:53 +00:00
KASSERT(!object->paging_in_progress,
("vm_object_terminate: pageout in progress"));
/*
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
* Clean and free the pages, as appropriate. All references to the
* object are gone, so we don't need to lock it.
1994-05-24 10:09:53 +00:00
*/
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
if (object->type == OBJT_VNODE) {
struct vnode *vp = (struct vnode *)object->handle;
/*
* Clean pages and flush buffers.
*/
This mega-commit is meant to fix numerous interrelated problems. There has been some bitrot and incorrect assumptions in the vfs_bio code. These problems have manifest themselves worse on NFS type filesystems, but can still affect local filesystems under certain circumstances. Most of the problems have involved mmap consistancy, and as a side-effect broke the vfs.ioopt code. This code might have been committed seperately, but almost everything is interrelated. 1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that are fully valid. 2) Rather than deactivating erroneously read initial (header) pages in kern_exec, we now free them. 3) Fix the rundown of non-VMIO buffers that are in an inconsistent (missing vp) state. 4) Fix the disassociation of pages from buffers in brelse. The previous code had rotted and was faulty in a couple of important circumstances. 5) Remove a gratuitious buffer wakeup in vfs_vmio_release. 6) Remove a crufty and currently unused cluster mechanism for VBLK files in vfs_bio_awrite. When the code is functional, I'll add back a cleaner version. 7) The page busy count wakeups assocated with the buffer cache usage were incorrectly cleaned up in a previous commit by me. Revert to the original, correct version, but with a cleaner implementation. 8) The cluster read code now tries to keep data associated with buffers more aggressively (without breaking the heuristics) when it is presumed that the read data (buffers) will be soon needed. 9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The delay loop waiting is not useful for filesystem locks, due to the length of the time intervals. 10) Correct and clean-up spec_getpages. 11) Implement a fully functional nfs_getpages, nfs_putpages. 12) Fix nfs_write so that modifications are coherent with the NFS data on the server disk (at least as well as NFS seems to allow.) 13) Properly support MS_INVALIDATE on NFS. 14) Properly pass down MS_INVALIDATE to lower levels of the VM code from vm_map_clean. 15) Better support the notion of pages being busy but valid, so that fewer in-transit waits occur. (use p->busy more for pageouts instead of PG_BUSY.) Since the page is fully valid, it is still usable for reads. 16) It is possible (in error) for cached pages to be busy. Make the page allocation code handle that case correctly. (It should probably be a printf or panic, but I want the system to handle coding errors robustly. I'll probably add a printf.) 17) Correct the design and usage of vm_page_sleep. It didn't handle consistancy problems very well, so make the design a little less lofty. After vm_page_sleep, if it ever blocked, it is still important to relookup the page (if the object generation count changed), and verify it's status (always.) 18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up. 19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush. 20) Fix vm_pager_put_pages and it's descendents to support an int flag instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
vm_object_page_clean(object, 0, 0, OBJPC_SYNC);
VM_OBJECT_UNLOCK(object);
vinvalbuf(vp, V_SAVE, 0, 0);
VM_OBJECT_LOCK(object);
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
}
2001-04-13 11:15:40 +00:00
KASSERT(object->ref_count == 0,
("vm_object_terminate: object with references, ref_count=%d",
object->ref_count));
/*
* Free any remaining pageable pages. This also removes them from the
* paging queues. However, don't free wired pages, just remove them
* from the object. Rather than incrementally removing each page from
* the object, the page and object are reset to any empty state.
*/
TAILQ_FOREACH_SAFE(p, &object->memq, listq, p_next) {
KASSERT(!p->busy && (p->oflags & VPO_BUSY) == 0,
("vm_object_terminate: freeing busy page %p", p));
vm_page_lock(p);
/*
* Optimize the page's removal from the object by resetting
* its "object" field. Specifically, if the page is not
* wired, then the effect of this assignment is that
* vm_page_free()'s call to vm_page_remove() will return
* immediately without modifying the page or the object.
*/
p->object = NULL;
if (p->wire_count == 0) {
vm_page_free(p);
PCPU_INC(cnt.v_pfree);
}
vm_page_unlock(p);
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
}
/*
* If the object contained any pages, then reset it to an empty state.
* None of the object's fields, including "resident_page_count", were
* modified by the preceding loop.
*/
if (object->resident_page_count != 0) {
object->root = NULL;
TAILQ_INIT(&object->memq);
object->resident_page_count = 0;
if (object->type == OBJT_VNODE)
vdrop(object->handle);
}
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
#if VM_NRESERVLEVEL > 0
if (__predict_false(!LIST_EMPTY(&object->rvq)))
vm_reserv_break_all(object);
#endif
Change the management of cached pages (PQ_CACHE) in two fundamental ways: (1) Cached pages are no longer kept in the object's resident page splay tree and memq. Instead, they are kept in a separate per-object splay tree of cached pages. However, access to this new per-object splay tree is synchronized by the _free_ page queues lock, not to be confused with the heavily contended page queues lock. Consequently, a cached page can be reclaimed by vm_page_alloc(9) without acquiring the object's lock or the page queues lock. This solves a problem independently reported by tegge@ and Isilon. Specifically, they observed the page daemon consuming a great deal of CPU time because of pages bouncing back and forth between the cache queue (PQ_CACHE) and the inactive queue (PQ_INACTIVE). The source of this problem turned out to be a deadlock avoidance strategy employed when selecting a cached page to reclaim in vm_page_select_cache(). However, the root cause was really that reclaiming a cached page required the acquisition of an object lock while the page queues lock was already held. Thus, this change addresses the problem at its root, by eliminating the need to acquire the object's lock. Moreover, keeping cached pages in the object's primary splay tree and memq was, in effect, optimizing for the uncommon case. Cached pages are reclaimed far, far more often than they are reactivated. Instead, this change makes reclamation cheaper, especially in terms of synchronization overhead, and reactivation more expensive, because reactivated pages will have to be reentered into the object's primary splay tree and memq. (2) Cached pages are now stored alongside free pages in the physical memory allocator's buddy queues, increasing the likelihood that large allocations of contiguous physical memory (i.e., superpages) will succeed. Finally, as a result of this change long-standing restrictions on when and where a cached page can be reclaimed and returned by vm_page_alloc(9) are eliminated. Specifically, calls to vm_page_alloc(9) specifying VM_ALLOC_INTERRUPT can now reclaim and return a formerly cached page. Consequently, a call to malloc(9) specifying M_NOWAIT is less likely to fail. Discussed with: many over the course of the summer, including jeff@, Justin Husted @ Isilon, peter@, tegge@ Tested by: an earlier version by kris@ Approved by: re (kensmith)
2007-09-25 06:25:06 +00:00
if (__predict_false(object->cache != NULL))
vm_page_cache_free(object, 0, 0);
Change the management of cached pages (PQ_CACHE) in two fundamental ways: (1) Cached pages are no longer kept in the object's resident page splay tree and memq. Instead, they are kept in a separate per-object splay tree of cached pages. However, access to this new per-object splay tree is synchronized by the _free_ page queues lock, not to be confused with the heavily contended page queues lock. Consequently, a cached page can be reclaimed by vm_page_alloc(9) without acquiring the object's lock or the page queues lock. This solves a problem independently reported by tegge@ and Isilon. Specifically, they observed the page daemon consuming a great deal of CPU time because of pages bouncing back and forth between the cache queue (PQ_CACHE) and the inactive queue (PQ_INACTIVE). The source of this problem turned out to be a deadlock avoidance strategy employed when selecting a cached page to reclaim in vm_page_select_cache(). However, the root cause was really that reclaiming a cached page required the acquisition of an object lock while the page queues lock was already held. Thus, this change addresses the problem at its root, by eliminating the need to acquire the object's lock. Moreover, keeping cached pages in the object's primary splay tree and memq was, in effect, optimizing for the uncommon case. Cached pages are reclaimed far, far more often than they are reactivated. Instead, this change makes reclamation cheaper, especially in terms of synchronization overhead, and reactivation more expensive, because reactivated pages will have to be reentered into the object's primary splay tree and memq. (2) Cached pages are now stored alongside free pages in the physical memory allocator's buddy queues, increasing the likelihood that large allocations of contiguous physical memory (i.e., superpages) will succeed. Finally, as a result of this change long-standing restrictions on when and where a cached page can be reclaimed and returned by vm_page_alloc(9) are eliminated. Specifically, calls to vm_page_alloc(9) specifying VM_ALLOC_INTERRUPT can now reclaim and return a formerly cached page. Consequently, a call to malloc(9) specifying M_NOWAIT is less likely to fail. Discussed with: many over the course of the summer, including jeff@, Justin Husted @ Isilon, peter@, tegge@ Tested by: an earlier version by kris@ Approved by: re (kensmith)
2007-09-25 06:25:06 +00:00
/*
* Let the pager know object is dead.
*/
vm_pager_deallocate(object);
VM_OBJECT_UNLOCK(object);
vm_object_destroy(object);
1994-05-24 10:09:53 +00:00
}
/*
* Make the page read-only so that we can clear the object flags. However, if
* this is a nosync mmap then the object is likely to stay dirty so do not
* mess with the page and do not clear the object flags. Returns TRUE if the
* page should be flushed, and FALSE otherwise.
*/
static boolean_t
vm_object_page_remove_write(vm_page_t p, int flags, boolean_t *clearobjflags)
{
/*
* If we have been asked to skip nosync pages and this is a
* nosync page, skip it. Note that the object flags were not
* cleared in this case so we do not have to set them.
*/
if ((flags & OBJPC_NOSYNC) != 0 && (p->oflags & VPO_NOSYNC) != 0) {
*clearobjflags = FALSE;
return (FALSE);
} else {
pmap_remove_write(p);
return (p->dirty != 0);
}
}
/*
* vm_object_page_clean
*
* Clean all dirty pages in the specified range of object. Leaves page
* on whatever queue it is currently on. If NOSYNC is set then do not
* write out pages with VPO_NOSYNC set (originally comes from MAP_NOSYNC),
* leaving the object dirty.
*
* When stuffing pages asynchronously, allow clustering. XXX we need a
* synchronous clustering mode implementation.
*
* Odd semantics: if start == end, we clean everything.
*
* The object must be locked.
*
* Returns FALSE if some page from the range was not written, as
* reported by the pager, and TRUE otherwise.
*/
boolean_t
vm_object_page_clean(vm_object_t object, vm_ooffset_t start, vm_ooffset_t end,
int flags)
{
vm_page_t np, p;
vm_pindex_t pi, tend, tstart;
int curgeneration, n, pagerflags;
boolean_t clearobjflags, eio, res;
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
KASSERT(object->type == OBJT_VNODE, ("Not a vnode object"));
if ((object->flags & OBJ_MIGHTBEDIRTY) == 0 ||
object->resident_page_count == 0)
return (TRUE);
pagerflags = (flags & (OBJPC_SYNC | OBJPC_INVAL)) != 0 ?
VM_PAGER_PUT_SYNC : VM_PAGER_CLUSTER_OK;
pagerflags |= (flags & OBJPC_INVAL) != 0 ? VM_PAGER_PUT_INVAL : 0;
tstart = OFF_TO_IDX(start);
tend = (end == 0) ? object->size : OFF_TO_IDX(end + PAGE_MASK);
clearobjflags = tstart == 0 && tend >= object->size;
res = TRUE;
rescan:
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
curgeneration = object->generation;
for (p = vm_page_find_least(object, tstart); p != NULL; p = np) {
pi = p->pindex;
if (pi >= tend)
break;
np = TAILQ_NEXT(p, listq);
if (p->valid == 0)
continue;
if (vm_page_sleep_if_busy(p, TRUE, "vpcwai")) {
if (object->generation != curgeneration) {
if ((flags & OBJPC_SYNC) != 0)
goto rescan;
else
clearobjflags = FALSE;
}
np = vm_page_find_least(object, pi);
continue;
}
if (!vm_object_page_remove_write(p, flags, &clearobjflags))
continue;
n = vm_object_page_collect_flush(object, p, pagerflags,
flags, &clearobjflags, &eio);
if (eio) {
res = FALSE;
clearobjflags = FALSE;
}
if (object->generation != curgeneration) {
if ((flags & OBJPC_SYNC) != 0)
goto rescan;
else
clearobjflags = FALSE;
}
/*
* If the VOP_PUTPAGES() did a truncated write, so
* that even the first page of the run is not fully
* written, vm_pageout_flush() returns 0 as the run
* length. Since the condition that caused truncated
* write may be permanent, e.g. exhausted free space,
* accepting n == 0 would cause an infinite loop.
*
* Forwarding the iterator leaves the unwritten page
* behind, but there is not much we can do there if
* filesystem refuses to write it.
*/
if (n == 0) {
n = 1;
clearobjflags = FALSE;
}
np = vm_page_find_least(object, pi + n);
}
#if 0
VOP_FSYNC(vp, (pagerflags & VM_PAGER_PUT_SYNC) ? MNT_WAIT : 0);
#endif
if (clearobjflags)
vm_object_clear_flag(object, OBJ_MIGHTBEDIRTY);
return (res);
}
static int
vm_object_page_collect_flush(vm_object_t object, vm_page_t p, int pagerflags,
int flags, boolean_t *clearobjflags, boolean_t *eio)
{
vm_page_t ma[vm_pageout_page_count], p_first, tp;
int count, i, mreq, runlen;
vm_page_lock_assert(p, MA_NOTOWNED);
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
count = 1;
mreq = 0;
for (tp = p; count < vm_pageout_page_count; count++) {
tp = vm_page_next(tp);
if (tp == NULL || tp->busy != 0 || (tp->oflags & VPO_BUSY) != 0)
break;
if (!vm_object_page_remove_write(tp, flags, clearobjflags))
break;
}
for (p_first = p; count < vm_pageout_page_count; count++) {
tp = vm_page_prev(p_first);
if (tp == NULL || tp->busy != 0 || (tp->oflags & VPO_BUSY) != 0)
break;
if (!vm_object_page_remove_write(tp, flags, clearobjflags))
break;
p_first = tp;
mreq++;
}
for (tp = p_first, i = 0; i < count; tp = TAILQ_NEXT(tp, listq), i++)
ma[i] = tp;
vm_pageout_flush(ma, count, pagerflags, mreq, &runlen, eio);
return (runlen);
}
/*
* Note that there is absolutely no sense in writing out
* anonymous objects, so we track down the vnode object
* to write out.
* We invalidate (remove) all pages from the address space
* for semantic correctness.
*
* If the backing object is a device object with unmanaged pages, then any
* mappings to the specified range of pages must be removed before this
* function is called.
*
* Note: certain anonymous maps, such as MAP_NOSYNC maps,
* may start out with a NULL object.
*/
boolean_t
vm_object_sync(vm_object_t object, vm_ooffset_t offset, vm_size_t size,
boolean_t syncio, boolean_t invalidate)
{
vm_object_t backing_object;
struct vnode *vp;
struct mount *mp;
int error, flags, fsync_after;
boolean_t res;
if (object == NULL)
return (TRUE);
res = TRUE;
error = 0;
VM_OBJECT_LOCK(object);
while ((backing_object = object->backing_object) != NULL) {
VM_OBJECT_LOCK(backing_object);
offset += object->backing_object_offset;
VM_OBJECT_UNLOCK(object);
object = backing_object;
if (object->size < OFF_TO_IDX(offset + size))
size = IDX_TO_OFF(object->size) - offset;
}
/*
* Flush pages if writing is allowed, invalidate them
* if invalidation requested. Pages undergoing I/O
* will be ignored by vm_object_page_remove().
*
* We cannot lock the vnode and then wait for paging
* to complete without deadlocking against vm_fault.
* Instead we simply call vm_object_page_remove() and
* allow it to block internally on a page-by-page
* basis when it encounters pages undergoing async
* I/O.
*/
if (object->type == OBJT_VNODE &&
(object->flags & OBJ_MIGHTBEDIRTY) != 0) {
vp = object->handle;
VM_OBJECT_UNLOCK(object);
(void) vn_start_write(vp, &mp, V_WAIT);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
if (syncio && !invalidate && offset == 0 &&
OFF_TO_IDX(size) == object->size) {
/*
* If syncing the whole mapping of the file,
* it is faster to schedule all the writes in
* async mode, also allowing the clustering,
* and then wait for i/o to complete.
*/
flags = 0;
fsync_after = TRUE;
} else {
flags = (syncio || invalidate) ? OBJPC_SYNC : 0;
flags |= invalidate ? (OBJPC_SYNC | OBJPC_INVAL) : 0;
fsync_after = FALSE;
}
VM_OBJECT_LOCK(object);
res = vm_object_page_clean(object, offset, offset + size,
flags);
VM_OBJECT_UNLOCK(object);
if (fsync_after)
error = VOP_FSYNC(vp, MNT_WAIT, curthread);
VOP_UNLOCK(vp, 0);
vn_finished_write(mp);
if (error != 0)
res = FALSE;
VM_OBJECT_LOCK(object);
}
if ((object->type == OBJT_VNODE ||
object->type == OBJT_DEVICE) && invalidate) {
if (object->type == OBJT_DEVICE)
/*
* The option OBJPR_NOTMAPPED must be passed here
* because vm_object_page_remove() cannot remove
* unmanaged mappings.
*/
flags = OBJPR_NOTMAPPED;
else if (old_msync)
flags = 0;
else
flags = OBJPR_CLEANONLY;
vm_object_page_remove(object, OFF_TO_IDX(offset),
OFF_TO_IDX(offset + size + PAGE_MASK), flags);
}
VM_OBJECT_UNLOCK(object);
return (res);
}
/*
* vm_object_madvise:
*
* Implements the madvise function at the object/page level.
*
* MADV_WILLNEED (any object)
*
* Activate the specified pages if they are resident.
*
* MADV_DONTNEED (any object)
*
* Deactivate the specified pages if they are resident.
*
* MADV_FREE (OBJT_DEFAULT/OBJT_SWAP objects,
* OBJ_ONEMAPPING only)
*
* Deactivate and clean the specified pages if they are
* resident. This permits the process to reuse the pages
* without faulting or the kernel to reclaim the pages
* without I/O.
*/
void
vm_object_madvise(vm_object_t object, vm_pindex_t pindex, vm_pindex_t end,
int advise)
{
vm_pindex_t tpindex;
vm_object_t backing_object, tobject;
vm_page_t m;
if (object == NULL)
return;
VM_OBJECT_LOCK(object);
/*
* Locate and adjust resident pages
*/
for (; pindex < end; pindex += 1) {
relookup:
tobject = object;
tpindex = pindex;
shadowlookup:
/*
* MADV_FREE only operates on OBJT_DEFAULT or OBJT_SWAP pages
* and those pages must be OBJ_ONEMAPPING.
*/
if (advise == MADV_FREE) {
if ((tobject->type != OBJT_DEFAULT &&
tobject->type != OBJT_SWAP) ||
(tobject->flags & OBJ_ONEMAPPING) == 0) {
goto unlock_tobject;
}
} else if ((tobject->flags & OBJ_UNMANAGED) != 0)
goto unlock_tobject;
m = vm_page_lookup(tobject, tpindex);
Change the management of cached pages (PQ_CACHE) in two fundamental ways: (1) Cached pages are no longer kept in the object's resident page splay tree and memq. Instead, they are kept in a separate per-object splay tree of cached pages. However, access to this new per-object splay tree is synchronized by the _free_ page queues lock, not to be confused with the heavily contended page queues lock. Consequently, a cached page can be reclaimed by vm_page_alloc(9) without acquiring the object's lock or the page queues lock. This solves a problem independently reported by tegge@ and Isilon. Specifically, they observed the page daemon consuming a great deal of CPU time because of pages bouncing back and forth between the cache queue (PQ_CACHE) and the inactive queue (PQ_INACTIVE). The source of this problem turned out to be a deadlock avoidance strategy employed when selecting a cached page to reclaim in vm_page_select_cache(). However, the root cause was really that reclaiming a cached page required the acquisition of an object lock while the page queues lock was already held. Thus, this change addresses the problem at its root, by eliminating the need to acquire the object's lock. Moreover, keeping cached pages in the object's primary splay tree and memq was, in effect, optimizing for the uncommon case. Cached pages are reclaimed far, far more often than they are reactivated. Instead, this change makes reclamation cheaper, especially in terms of synchronization overhead, and reactivation more expensive, because reactivated pages will have to be reentered into the object's primary splay tree and memq. (2) Cached pages are now stored alongside free pages in the physical memory allocator's buddy queues, increasing the likelihood that large allocations of contiguous physical memory (i.e., superpages) will succeed. Finally, as a result of this change long-standing restrictions on when and where a cached page can be reclaimed and returned by vm_page_alloc(9) are eliminated. Specifically, calls to vm_page_alloc(9) specifying VM_ALLOC_INTERRUPT can now reclaim and return a formerly cached page. Consequently, a call to malloc(9) specifying M_NOWAIT is less likely to fail. Discussed with: many over the course of the summer, including jeff@, Justin Husted @ Isilon, peter@, tegge@ Tested by: an earlier version by kris@ Approved by: re (kensmith)
2007-09-25 06:25:06 +00:00
if (m == NULL && advise == MADV_WILLNEED) {
/*
* If the page is cached, reactivate it.
*/
m = vm_page_alloc(tobject, tpindex, VM_ALLOC_IFCACHED |
VM_ALLOC_NOBUSY);
Change the management of cached pages (PQ_CACHE) in two fundamental ways: (1) Cached pages are no longer kept in the object's resident page splay tree and memq. Instead, they are kept in a separate per-object splay tree of cached pages. However, access to this new per-object splay tree is synchronized by the _free_ page queues lock, not to be confused with the heavily contended page queues lock. Consequently, a cached page can be reclaimed by vm_page_alloc(9) without acquiring the object's lock or the page queues lock. This solves a problem independently reported by tegge@ and Isilon. Specifically, they observed the page daemon consuming a great deal of CPU time because of pages bouncing back and forth between the cache queue (PQ_CACHE) and the inactive queue (PQ_INACTIVE). The source of this problem turned out to be a deadlock avoidance strategy employed when selecting a cached page to reclaim in vm_page_select_cache(). However, the root cause was really that reclaiming a cached page required the acquisition of an object lock while the page queues lock was already held. Thus, this change addresses the problem at its root, by eliminating the need to acquire the object's lock. Moreover, keeping cached pages in the object's primary splay tree and memq was, in effect, optimizing for the uncommon case. Cached pages are reclaimed far, far more often than they are reactivated. Instead, this change makes reclamation cheaper, especially in terms of synchronization overhead, and reactivation more expensive, because reactivated pages will have to be reentered into the object's primary splay tree and memq. (2) Cached pages are now stored alongside free pages in the physical memory allocator's buddy queues, increasing the likelihood that large allocations of contiguous physical memory (i.e., superpages) will succeed. Finally, as a result of this change long-standing restrictions on when and where a cached page can be reclaimed and returned by vm_page_alloc(9) are eliminated. Specifically, calls to vm_page_alloc(9) specifying VM_ALLOC_INTERRUPT can now reclaim and return a formerly cached page. Consequently, a call to malloc(9) specifying M_NOWAIT is less likely to fail. Discussed with: many over the course of the summer, including jeff@, Justin Husted @ Isilon, peter@, tegge@ Tested by: an earlier version by kris@ Approved by: re (kensmith)
2007-09-25 06:25:06 +00:00
}
if (m == NULL) {
/*
* There may be swap even if there is no backing page
*/
if (advise == MADV_FREE && tobject->type == OBJT_SWAP)
swap_pager_freespace(tobject, tpindex, 1);
/*
* next object
*/
backing_object = tobject->backing_object;
if (backing_object == NULL)
goto unlock_tobject;
VM_OBJECT_LOCK(backing_object);
tpindex += OFF_TO_IDX(tobject->backing_object_offset);
if (tobject != object)
VM_OBJECT_UNLOCK(tobject);
tobject = backing_object;
goto shadowlookup;
} else if (m->valid != VM_PAGE_BITS_ALL)
goto unlock_tobject;
/*
* If the page is not in a normal state, skip it.
*/
vm_page_lock(m);
if (m->hold_count != 0 || m->wire_count != 0) {
vm_page_unlock(m);
goto unlock_tobject;
}
KASSERT((m->flags & PG_FICTITIOUS) == 0,
("vm_object_madvise: page %p is fictitious", m));
KASSERT((m->oflags & VPO_UNMANAGED) == 0,
("vm_object_madvise: page %p is not managed", m));
if ((m->oflags & VPO_BUSY) || m->busy) {
if (advise == MADV_WILLNEED) {
/*
* Reference the page before unlocking and
* sleeping so that the page daemon is less
* likely to reclaim it.
*/
vm_page_aflag_set(m, PGA_REFERENCED);
}
vm_page_unlock(m);
if (object != tobject)
VM_OBJECT_UNLOCK(object);
m->oflags |= VPO_WANTED;
VM_OBJECT_SLEEP(tobject, m, PDROP | PVM, "madvpo", 0);
VM_OBJECT_LOCK(object);
goto relookup;
}
if (advise == MADV_WILLNEED) {
vm_page_activate(m);
} else if (advise == MADV_DONTNEED) {
vm_page_dontneed(m);
} else if (advise == MADV_FREE) {
/*
* Mark the page clean. This will allow the page
* to be freed up by the system. However, such pages
* are often reused quickly by malloc()/free()
* so we do not do anything that would cause
* a page fault if we can help it.
*
* Specifically, we do not try to actually free
* the page now nor do we try to put it in the
* cache (which would cause a page fault on reuse).
*
* But we do make the page is freeable as we
* can without actually taking the step of unmapping
* it.
*/
pmap_clear_modify(m);
m->dirty = 0;
m->act_count = 0;
vm_page_dontneed(m);
}
vm_page_unlock(m);
if (advise == MADV_FREE && tobject->type == OBJT_SWAP)
swap_pager_freespace(tobject, tpindex, 1);
unlock_tobject:
if (tobject != object)
VM_OBJECT_UNLOCK(tobject);
}
VM_OBJECT_UNLOCK(object);
}
1994-05-24 10:09:53 +00:00
/*
* vm_object_shadow:
*
* Create a new object which is backed by the
* specified existing object range. The source
* object reference is deallocated.
*
* The new object and offset into that object
* are returned in the source parameters.
*/
void
vm_object_shadow(
vm_object_t *object, /* IN/OUT */
vm_ooffset_t *offset, /* IN/OUT */
vm_size_t length)
1994-05-24 10:09:53 +00:00
{
vm_object_t source;
vm_object_t result;
1994-05-24 10:09:53 +00:00
source = *object;
/*
* Don't create the new object if the old object isn't shared.
*/
if (source != NULL) {
VM_OBJECT_LOCK(source);
if (source->ref_count == 1 &&
source->handle == NULL &&
(source->type == OBJT_DEFAULT ||
source->type == OBJT_SWAP)) {
VM_OBJECT_UNLOCK(source);
return;
}
VM_OBJECT_UNLOCK(source);
}
1994-05-24 10:09:53 +00:00
/*
* Allocate a new object with the given length.
1994-05-24 10:09:53 +00:00
*/
result = vm_object_allocate(OBJT_DEFAULT, atop(length));
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
* The new object shadows the source object, adding a reference to it.
* Our caller changes his reference to point to the new object,
* removing a reference to the source object. Net result: no change
* of reference count.
*
* Try to optimize the result object's page color when shadowing
2000-03-26 15:20:23 +00:00
* in order to maintain page coloring consistency in the combined
* shadowed object.
1994-05-24 10:09:53 +00:00
*/
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
result->backing_object = source;
/*
* Store the offset into the source object, and fix up the offset into
* the new object.
*/
result->backing_object_offset = *offset;
if (source != NULL) {
VM_OBJECT_LOCK(source);
LIST_INSERT_HEAD(&source->shadow_head, result, shadow_list);
source->shadow_count++;
#if VM_NRESERVLEVEL > 0
result->flags |= source->flags & OBJ_COLORED;
result->pg_color = (source->pg_color + OFF_TO_IDX(*offset)) &
((1 << (VM_NFREEORDER - 1)) - 1);
#endif
VM_OBJECT_UNLOCK(source);
}
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
/*
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 new things
1994-05-24 10:09:53 +00:00
*/
*offset = 0;
*object = result;
}
/*
* vm_object_split:
*
* Split the pages in a map entry into a new object. This affords
* easier removal of unused pages, and keeps object inheritance from
* being a negative impact on memory usage.
*/
void
vm_object_split(vm_map_entry_t entry)
{
vm_page_t m, m_next;
vm_object_t orig_object, new_object, source;
vm_pindex_t idx, offidxstart;
vm_size_t size;
orig_object = entry->object.vm_object;
if (orig_object->type != OBJT_DEFAULT && orig_object->type != OBJT_SWAP)
return;
if (orig_object->ref_count <= 1)
return;
VM_OBJECT_UNLOCK(orig_object);
offidxstart = OFF_TO_IDX(entry->offset);
size = atop(entry->end - entry->start);
/*
* If swap_pager_copy() is later called, it will convert new_object
* into a swap object.
*/
new_object = vm_object_allocate(OBJT_DEFAULT, size);
/*
* At this point, the new object is still private, so the order in
* which the original and new objects are locked does not matter.
*/
VM_OBJECT_LOCK(new_object);
VM_OBJECT_LOCK(orig_object);
source = orig_object->backing_object;
if (source != NULL) {
VM_OBJECT_LOCK(source);
if ((source->flags & OBJ_DEAD) != 0) {
VM_OBJECT_UNLOCK(source);
VM_OBJECT_UNLOCK(orig_object);
VM_OBJECT_UNLOCK(new_object);
vm_object_deallocate(new_object);
VM_OBJECT_LOCK(orig_object);
return;
}
LIST_INSERT_HEAD(&source->shadow_head,
new_object, shadow_list);
source->shadow_count++;
vm_object_reference_locked(source); /* for new_object */
vm_object_clear_flag(source, OBJ_ONEMAPPING);
VM_OBJECT_UNLOCK(source);
new_object->backing_object_offset =
orig_object->backing_object_offset + entry->offset;
new_object->backing_object = source;
}
if (orig_object->cred != NULL) {
new_object->cred = orig_object->cred;
crhold(orig_object->cred);
new_object->charge = ptoa(size);
KASSERT(orig_object->charge >= ptoa(size),
("orig_object->charge < 0"));
orig_object->charge -= ptoa(size);
}
retry:
m = vm_page_find_least(orig_object, offidxstart);
for (; m != NULL && (idx = m->pindex - offidxstart) < size;
m = m_next) {
m_next = TAILQ_NEXT(m, listq);
/*
* We must wait for pending I/O to complete before we can
* rename the page.
*
* We do not have to VM_PROT_NONE the page as mappings should
* not be changed by this operation.
*/
if ((m->oflags & VPO_BUSY) || m->busy) {
VM_OBJECT_UNLOCK(new_object);
m->oflags |= VPO_WANTED;
VM_OBJECT_SLEEP(orig_object, m, PVM, "spltwt", 0);
VM_OBJECT_LOCK(new_object);
goto retry;
}
#if VM_NRESERVLEVEL > 0
/*
* If some of the reservation's allocated pages remain with
* the original object, then transferring the reservation to
* the new object is neither particularly beneficial nor
* particularly harmful as compared to leaving the reservation
* with the original object. If, however, all of the
* reservation's allocated pages are transferred to the new
* object, then transferring the reservation is typically
* beneficial. Determining which of these two cases applies
* would be more costly than unconditionally renaming the
* reservation.
*/
vm_reserv_rename(m, new_object, orig_object, offidxstart);
#endif
vm_page_lock(m);
vm_page_rename(m, new_object, idx);
vm_page_unlock(m);
/* page automatically made dirty by rename and cache handled */
vm_page_busy(m);
}
if (orig_object->type == OBJT_SWAP) {
/*
* swap_pager_copy() can sleep, in which case the orig_object's
* and new_object's locks are released and reacquired.
*/
swap_pager_copy(orig_object, new_object, offidxstart, 0);
Change the management of cached pages (PQ_CACHE) in two fundamental ways: (1) Cached pages are no longer kept in the object's resident page splay tree and memq. Instead, they are kept in a separate per-object splay tree of cached pages. However, access to this new per-object splay tree is synchronized by the _free_ page queues lock, not to be confused with the heavily contended page queues lock. Consequently, a cached page can be reclaimed by vm_page_alloc(9) without acquiring the object's lock or the page queues lock. This solves a problem independently reported by tegge@ and Isilon. Specifically, they observed the page daemon consuming a great deal of CPU time because of pages bouncing back and forth between the cache queue (PQ_CACHE) and the inactive queue (PQ_INACTIVE). The source of this problem turned out to be a deadlock avoidance strategy employed when selecting a cached page to reclaim in vm_page_select_cache(). However, the root cause was really that reclaiming a cached page required the acquisition of an object lock while the page queues lock was already held. Thus, this change addresses the problem at its root, by eliminating the need to acquire the object's lock. Moreover, keeping cached pages in the object's primary splay tree and memq was, in effect, optimizing for the uncommon case. Cached pages are reclaimed far, far more often than they are reactivated. Instead, this change makes reclamation cheaper, especially in terms of synchronization overhead, and reactivation more expensive, because reactivated pages will have to be reentered into the object's primary splay tree and memq. (2) Cached pages are now stored alongside free pages in the physical memory allocator's buddy queues, increasing the likelihood that large allocations of contiguous physical memory (i.e., superpages) will succeed. Finally, as a result of this change long-standing restrictions on when and where a cached page can be reclaimed and returned by vm_page_alloc(9) are eliminated. Specifically, calls to vm_page_alloc(9) specifying VM_ALLOC_INTERRUPT can now reclaim and return a formerly cached page. Consequently, a call to malloc(9) specifying M_NOWAIT is less likely to fail. Discussed with: many over the course of the summer, including jeff@, Justin Husted @ Isilon, peter@, tegge@ Tested by: an earlier version by kris@ Approved by: re (kensmith)
2007-09-25 06:25:06 +00:00
/*
* Transfer any cached pages from orig_object to new_object.
* If swap_pager_copy() found swapped out pages within the
* specified range of orig_object, then it changed
* new_object's type to OBJT_SWAP when it transferred those
* pages to new_object. Otherwise, new_object's type
* should still be OBJT_DEFAULT and orig_object should not
* contain any cached pages within the specified range.
Change the management of cached pages (PQ_CACHE) in two fundamental ways: (1) Cached pages are no longer kept in the object's resident page splay tree and memq. Instead, they are kept in a separate per-object splay tree of cached pages. However, access to this new per-object splay tree is synchronized by the _free_ page queues lock, not to be confused with the heavily contended page queues lock. Consequently, a cached page can be reclaimed by vm_page_alloc(9) without acquiring the object's lock or the page queues lock. This solves a problem independently reported by tegge@ and Isilon. Specifically, they observed the page daemon consuming a great deal of CPU time because of pages bouncing back and forth between the cache queue (PQ_CACHE) and the inactive queue (PQ_INACTIVE). The source of this problem turned out to be a deadlock avoidance strategy employed when selecting a cached page to reclaim in vm_page_select_cache(). However, the root cause was really that reclaiming a cached page required the acquisition of an object lock while the page queues lock was already held. Thus, this change addresses the problem at its root, by eliminating the need to acquire the object's lock. Moreover, keeping cached pages in the object's primary splay tree and memq was, in effect, optimizing for the uncommon case. Cached pages are reclaimed far, far more often than they are reactivated. Instead, this change makes reclamation cheaper, especially in terms of synchronization overhead, and reactivation more expensive, because reactivated pages will have to be reentered into the object's primary splay tree and memq. (2) Cached pages are now stored alongside free pages in the physical memory allocator's buddy queues, increasing the likelihood that large allocations of contiguous physical memory (i.e., superpages) will succeed. Finally, as a result of this change long-standing restrictions on when and where a cached page can be reclaimed and returned by vm_page_alloc(9) are eliminated. Specifically, calls to vm_page_alloc(9) specifying VM_ALLOC_INTERRUPT can now reclaim and return a formerly cached page. Consequently, a call to malloc(9) specifying M_NOWAIT is less likely to fail. Discussed with: many over the course of the summer, including jeff@, Justin Husted @ Isilon, peter@, tegge@ Tested by: an earlier version by kris@ Approved by: re (kensmith)
2007-09-25 06:25:06 +00:00
*/
if (__predict_false(orig_object->cache != NULL))
vm_page_cache_transfer(orig_object, offidxstart,
new_object);
}
VM_OBJECT_UNLOCK(orig_object);
TAILQ_FOREACH(m, &new_object->memq, listq)
vm_page_wakeup(m);
VM_OBJECT_UNLOCK(new_object);
entry->object.vm_object = new_object;
entry->offset = 0LL;
vm_object_deallocate(orig_object);
VM_OBJECT_LOCK(new_object);
}
#define OBSC_TEST_ALL_SHADOWED 0x0001
#define OBSC_COLLAPSE_NOWAIT 0x0002
#define OBSC_COLLAPSE_WAIT 0x0004
1994-05-24 10:09:53 +00:00
static int
vm_object_backing_scan(vm_object_t object, int op)
{
int r = 1;
vm_page_t p;
vm_object_t backing_object;
vm_pindex_t backing_offset_index;
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
VM_OBJECT_LOCK_ASSERT(object->backing_object, MA_OWNED);
backing_object = object->backing_object;
backing_offset_index = OFF_TO_IDX(object->backing_object_offset);
/*
* Initial conditions
*/
if (op & OBSC_TEST_ALL_SHADOWED) {
/*
Change the management of cached pages (PQ_CACHE) in two fundamental ways: (1) Cached pages are no longer kept in the object's resident page splay tree and memq. Instead, they are kept in a separate per-object splay tree of cached pages. However, access to this new per-object splay tree is synchronized by the _free_ page queues lock, not to be confused with the heavily contended page queues lock. Consequently, a cached page can be reclaimed by vm_page_alloc(9) without acquiring the object's lock or the page queues lock. This solves a problem independently reported by tegge@ and Isilon. Specifically, they observed the page daemon consuming a great deal of CPU time because of pages bouncing back and forth between the cache queue (PQ_CACHE) and the inactive queue (PQ_INACTIVE). The source of this problem turned out to be a deadlock avoidance strategy employed when selecting a cached page to reclaim in vm_page_select_cache(). However, the root cause was really that reclaiming a cached page required the acquisition of an object lock while the page queues lock was already held. Thus, this change addresses the problem at its root, by eliminating the need to acquire the object's lock. Moreover, keeping cached pages in the object's primary splay tree and memq was, in effect, optimizing for the uncommon case. Cached pages are reclaimed far, far more often than they are reactivated. Instead, this change makes reclamation cheaper, especially in terms of synchronization overhead, and reactivation more expensive, because reactivated pages will have to be reentered into the object's primary splay tree and memq. (2) Cached pages are now stored alongside free pages in the physical memory allocator's buddy queues, increasing the likelihood that large allocations of contiguous physical memory (i.e., superpages) will succeed. Finally, as a result of this change long-standing restrictions on when and where a cached page can be reclaimed and returned by vm_page_alloc(9) are eliminated. Specifically, calls to vm_page_alloc(9) specifying VM_ALLOC_INTERRUPT can now reclaim and return a formerly cached page. Consequently, a call to malloc(9) specifying M_NOWAIT is less likely to fail. Discussed with: many over the course of the summer, including jeff@, Justin Husted @ Isilon, peter@, tegge@ Tested by: an earlier version by kris@ Approved by: re (kensmith)
2007-09-25 06:25:06 +00:00
* We do not want to have to test for the existence of cache
* or swap pages in the backing object. XXX but with the
* new swapper this would be pretty easy to do.
*
* XXX what about anonymous MAP_SHARED memory that hasn't
* been ZFOD faulted yet? If we do not test for this, the
* shadow test may succeed! XXX
*/
if (backing_object->type != OBJT_DEFAULT) {
return (0);
}
}
if (op & OBSC_COLLAPSE_WAIT) {
vm_object_set_flag(backing_object, OBJ_DEAD);
}
/*
* Our scan
*/
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
p = TAILQ_FIRST(&backing_object->memq);
while (p) {
vm_page_t next = TAILQ_NEXT(p, listq);
vm_pindex_t new_pindex = p->pindex - backing_offset_index;
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 (op & OBSC_TEST_ALL_SHADOWED) {
vm_page_t pp;
/*
* Ignore pages outside the parent object's range
* and outside the parent object's mapping of the
* backing object.
*
* note that we do not busy the backing object's
* page.
*/
if (
p->pindex < backing_offset_index ||
new_pindex >= object->size
) {
p = next;
continue;
}
/*
* See if the parent has the page or if the parent's
* object pager has the page. If the parent has the
* page but the page is not valid, the parent's
* object pager must have the page.
*
* If this fails, the parent does not completely shadow
* the object and we might as well give up now.
*/
pp = vm_page_lookup(object, new_pindex);
if (
(pp == NULL || pp->valid == 0) &&
!vm_pager_has_page(object, new_pindex, NULL, NULL)
) {
r = 0;
break;
}
}
/*
* Check for busy page
*/
if (op & (OBSC_COLLAPSE_WAIT | OBSC_COLLAPSE_NOWAIT)) {
vm_page_t pp;
if (op & OBSC_COLLAPSE_NOWAIT) {
if ((p->oflags & VPO_BUSY) ||
!p->valid ||
p->busy) {
p = next;
continue;
}
} else if (op & OBSC_COLLAPSE_WAIT) {
if ((p->oflags & VPO_BUSY) || p->busy) {
VM_OBJECT_UNLOCK(object);
p->oflags |= VPO_WANTED;
VM_OBJECT_SLEEP(backing_object, p,
PDROP | PVM, "vmocol", 0);
VM_OBJECT_LOCK(object);
VM_OBJECT_LOCK(backing_object);
/*
* If we slept, anything could have
* happened. Since the object is
* marked dead, the backing offset
* should not have changed so we
* just restart our scan.
*/
p = TAILQ_FIRST(&backing_object->memq);
continue;
}
}
KASSERT(
p->object == backing_object,
("vm_object_backing_scan: object mismatch")
);
/*
* Destroy any associated swap
*/
if (backing_object->type == OBJT_SWAP) {
swap_pager_freespace(
backing_object,
p->pindex,
1
);
}
if (
p->pindex < backing_offset_index ||
new_pindex >= object->size
) {
/*
* Page is out of the parent object's range, we
* can simply destroy it.
*/
vm_page_lock(p);
KASSERT(!pmap_page_is_mapped(p),
("freeing mapped page %p", p));
if (p->wire_count == 0)
vm_page_free(p);
else
vm_page_remove(p);
vm_page_unlock(p);
p = next;
continue;
}
pp = vm_page_lookup(object, new_pindex);
if (
(op & OBSC_COLLAPSE_NOWAIT) != 0 &&
(pp != NULL && pp->valid == 0)
) {
/*
* The page in the parent is not (yet) valid.
* We don't know anything about the state of
* the original page. It might be mapped,
* so we must avoid the next if here.
*
* This is due to a race in vm_fault() where
* we must unbusy the original (backing_obj)
* page before we can (re)lock the parent.
* Hence we can get here.
*/
p = next;
continue;
}
if (
pp != NULL ||
vm_pager_has_page(object, new_pindex, NULL, NULL)
) {
/*
* page already exists in parent OR swap exists
* for this location in the parent. Destroy
* the original page from the backing object.
*
* Leave the parent's page alone
*/
vm_page_lock(p);
KASSERT(!pmap_page_is_mapped(p),
("freeing mapped page %p", p));
if (p->wire_count == 0)
vm_page_free(p);
else
vm_page_remove(p);
vm_page_unlock(p);
p = next;
continue;
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 VM_NRESERVLEVEL > 0
/*
* Rename the reservation.
*/
vm_reserv_rename(p, object, backing_object,
backing_offset_index);
#endif
/*
* Page does not exist in parent, rename the
* page from the backing object to the main object.
*
* If the page was mapped to a process, it can remain
* mapped through the rename.
*/
vm_page_lock(p);
vm_page_rename(p, object, new_pindex);
vm_page_unlock(p);
/* page automatically made dirty by rename */
}
p = next;
}
return (r);
}
/*
* this version of collapse allows the operation to occur earlier and
* when paging_in_progress is true for an object... This is not a complete
* operation, but should plug 99.9% of the rest of the leaks.
*/
static void
vm_object_qcollapse(vm_object_t object)
{
vm_object_t backing_object = object->backing_object;
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
VM_OBJECT_LOCK_ASSERT(backing_object, MA_OWNED);
if (backing_object->ref_count != 1)
return;
vm_object_backing_scan(object, OBSC_COLLAPSE_NOWAIT);
}
1994-05-24 10:09:53 +00:00
/*
* vm_object_collapse:
*
* Collapse an object with the object backing it.
* Pages in the backing object are moved into the
* parent, and the backing object is deallocated.
*/
void
vm_object_collapse(vm_object_t object)
1994-05-24 10:09:53 +00:00
{
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
1994-05-24 10:09:53 +00:00
while (TRUE) {
vm_object_t backing_object;
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
* Verify that the conditions are right for collapse:
1995-05-30 08:16:23 +00:00
*
* The object exists and the backing object exists.
1994-05-24 10:09:53 +00:00
*/
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
if ((backing_object = object->backing_object) == NULL)
break;
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
/*
* we check the backing object first, because it is most likely
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
* not collapsable.
*/
VM_OBJECT_LOCK(backing_object);
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
if (backing_object->handle != NULL ||
(backing_object->type != OBJT_DEFAULT &&
backing_object->type != OBJT_SWAP) ||
(backing_object->flags & OBJ_DEAD) ||
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
object->handle != NULL ||
(object->type != OBJT_DEFAULT &&
object->type != OBJT_SWAP) ||
(object->flags & OBJ_DEAD)) {
VM_OBJECT_UNLOCK(backing_object);
break;
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
}
if (
object->paging_in_progress != 0 ||
backing_object->paging_in_progress != 0
) {
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
vm_object_qcollapse(object);
VM_OBJECT_UNLOCK(backing_object);
break;
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 know that we can either collapse the backing object (if
* the parent is the only reference to it) or (perhaps) have
* the parent bypass the object if the parent happens to shadow
* all the resident pages in the entire backing object.
*
* This is ignoring pager-backed pages such as swap pages.
* vm_object_backing_scan fails the shadowing test in this
* case.
1994-05-24 10:09:53 +00:00
*/
if (backing_object->ref_count == 1) {
/*
* If there is exactly one reference to the backing
* object, we can collapse it into the parent.
1994-05-24 10:09:53 +00:00
*/
vm_object_backing_scan(object, OBSC_COLLAPSE_WAIT);
1994-05-24 10:09:53 +00:00
#if VM_NRESERVLEVEL > 0
/*
* Break any reservations from backing_object.
*/
if (__predict_false(!LIST_EMPTY(&backing_object->rvq)))
vm_reserv_break_all(backing_object);
#endif
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
* Move the pager from backing_object to object.
1994-05-24 10:09:53 +00:00
*/
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
if (backing_object->type == OBJT_SWAP) {
/*
* swap_pager_copy() can sleep, in which case
* the backing_object's and object's locks are
* released and reacquired.
* Since swap_pager_copy() is being asked to
* destroy the source, it will change the
* backing_object's type to OBJT_DEFAULT.
*/
swap_pager_copy(
backing_object,
object,
OFF_TO_IDX(object->backing_object_offset), TRUE);
Change the management of cached pages (PQ_CACHE) in two fundamental ways: (1) Cached pages are no longer kept in the object's resident page splay tree and memq. Instead, they are kept in a separate per-object splay tree of cached pages. However, access to this new per-object splay tree is synchronized by the _free_ page queues lock, not to be confused with the heavily contended page queues lock. Consequently, a cached page can be reclaimed by vm_page_alloc(9) without acquiring the object's lock or the page queues lock. This solves a problem independently reported by tegge@ and Isilon. Specifically, they observed the page daemon consuming a great deal of CPU time because of pages bouncing back and forth between the cache queue (PQ_CACHE) and the inactive queue (PQ_INACTIVE). The source of this problem turned out to be a deadlock avoidance strategy employed when selecting a cached page to reclaim in vm_page_select_cache(). However, the root cause was really that reclaiming a cached page required the acquisition of an object lock while the page queues lock was already held. Thus, this change addresses the problem at its root, by eliminating the need to acquire the object's lock. Moreover, keeping cached pages in the object's primary splay tree and memq was, in effect, optimizing for the uncommon case. Cached pages are reclaimed far, far more often than they are reactivated. Instead, this change makes reclamation cheaper, especially in terms of synchronization overhead, and reactivation more expensive, because reactivated pages will have to be reentered into the object's primary splay tree and memq. (2) Cached pages are now stored alongside free pages in the physical memory allocator's buddy queues, increasing the likelihood that large allocations of contiguous physical memory (i.e., superpages) will succeed. Finally, as a result of this change long-standing restrictions on when and where a cached page can be reclaimed and returned by vm_page_alloc(9) are eliminated. Specifically, calls to vm_page_alloc(9) specifying VM_ALLOC_INTERRUPT can now reclaim and return a formerly cached page. Consequently, a call to malloc(9) specifying M_NOWAIT is less likely to fail. Discussed with: many over the course of the summer, including jeff@, Justin Husted @ Isilon, peter@, tegge@ Tested by: an earlier version by kris@ Approved by: re (kensmith)
2007-09-25 06:25:06 +00:00
/*
* Free any cached pages from backing_object.
*/
if (__predict_false(backing_object->cache != NULL))
vm_page_cache_free(backing_object, 0, 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
* Object now shadows whatever backing_object did.
* Note that the reference to
* backing_object->backing_object moves from within
* backing_object to within object.
1994-05-24 10:09:53 +00:00
*/
LIST_REMOVE(object, shadow_list);
backing_object->shadow_count--;
if (backing_object->backing_object) {
VM_OBJECT_LOCK(backing_object->backing_object);
LIST_REMOVE(backing_object, shadow_list);
LIST_INSERT_HEAD(
&backing_object->backing_object->shadow_head,
object, shadow_list);
/*
* The shadow_count has not changed.
*/
VM_OBJECT_UNLOCK(backing_object->backing_object);
}
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
object->backing_object = backing_object->backing_object;
object->backing_object_offset +=
backing_object->backing_object_offset;
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
* Discard backing_object.
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
* Since the backing object has no pages, no pager left,
* and no object references within it, all that is
* necessary is to dispose of it.
1994-05-24 10:09:53 +00:00
*/
KASSERT(backing_object->ref_count == 1, (
"backing_object %p was somehow re-referenced during collapse!",
backing_object));
VM_OBJECT_UNLOCK(backing_object);
vm_object_destroy(backing_object);
1994-05-24 10:09:53 +00:00
object_collapses++;
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 {
vm_object_t new_backing_object;
1994-05-24 10:09:53 +00:00
/*
* If we do not entirely shadow the backing object,
* there is nothing we can do so we give up.
1994-05-24 10:09:53 +00:00
*/
if (object->resident_page_count != object->size &&
vm_object_backing_scan(object,
OBSC_TEST_ALL_SHADOWED) == 0) {
VM_OBJECT_UNLOCK(backing_object);
break;
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 the parent shadow the next object in the
* chain. Deallocating backing_object will not remove
* it, since its reference count is at least 2.
1994-05-24 10:09:53 +00:00
*/
LIST_REMOVE(object, shadow_list);
backing_object->shadow_count--;
new_backing_object = backing_object->backing_object;
if ((object->backing_object = new_backing_object) != NULL) {
VM_OBJECT_LOCK(new_backing_object);
LIST_INSERT_HEAD(
&new_backing_object->shadow_head,
object,
shadow_list
);
new_backing_object->shadow_count++;
vm_object_reference_locked(new_backing_object);
VM_OBJECT_UNLOCK(new_backing_object);
object->backing_object_offset +=
backing_object->backing_object_offset;
}
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
/*
* Drop the reference count on backing_object. Since
* its ref_count was at least 2, it will not vanish.
1994-05-24 10:09:53 +00:00
*/
backing_object->ref_count--;
VM_OBJECT_UNLOCK(backing_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
object_bypasses++;
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
* Try again with this object's new backing object.
1994-05-24 10:09:53 +00:00
*/
}
}
/*
* vm_object_page_remove:
1994-05-24 10:09:53 +00:00
*
* For the given object, either frees or invalidates each of the
* specified pages. In general, a page is freed. However, if a page is
* wired for any reason other than the existence of a managed, wired
* mapping, then it may be invalidated but not removed from the object.
* Pages are specified by the given range ["start", "end") and the option
* OBJPR_CLEANONLY. As a special case, if "end" is zero, then the range
* extends from "start" to the end of the object. If the option
* OBJPR_CLEANONLY is specified, then only the non-dirty pages within the
* specified range are affected. If the option OBJPR_NOTMAPPED is
* specified, then the pages within the specified range must have no
* mappings. Otherwise, if this option is not specified, any mappings to
* the specified pages are removed before the pages are freed or
* invalidated.
*
* In general, this operation should only be performed on objects that
* contain managed pages. There are, however, two exceptions. First, it
* is performed on the kernel and kmem objects by vm_map_entry_delete().
* Second, it is used by msync(..., MS_INVALIDATE) to invalidate device-
* backed pages. In both of these cases, the option OBJPR_CLEANONLY must
* not be specified and the option OBJPR_NOTMAPPED must be specified.
1994-05-24 10:09:53 +00:00
*
* The object must be locked.
*/
void
vm_object_page_remove(vm_object_t object, vm_pindex_t start, vm_pindex_t end,
int options)
1994-05-24 10:09:53 +00:00
{
vm_page_t p, next;
int wirings;
1994-05-24 10:09:53 +00:00
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
KASSERT((object->flags & OBJ_UNMANAGED) == 0 ||
(options & (OBJPR_CLEANONLY | OBJPR_NOTMAPPED)) == OBJPR_NOTMAPPED,
("vm_object_page_remove: illegal options for object %p", object));
if (object->resident_page_count == 0)
goto skipmemq;
vm_object_pip_add(object, 1);
again:
p = vm_page_find_least(object, start);
/*
* Here, the variable "p" is either (1) the page with the least pindex
* greater than or equal to the parameter "start" or (2) NULL.
*/
for (; p != NULL && (p->pindex < end || end == 0); p = next) {
next = TAILQ_NEXT(p, listq);
/*
* If the page is wired for any reason besides the existence
* of managed, wired mappings, then it cannot be freed. For
* example, fictitious pages, which represent device memory,
* are inherently wired and cannot be freed. They can,
* however, be invalidated if the option OBJPR_CLEANONLY is
* not specified.
*/
vm_page_lock(p);
if ((wirings = p->wire_count) != 0 &&
(wirings = pmap_page_wired_mappings(p)) != p->wire_count) {
if ((options & OBJPR_NOTMAPPED) == 0) {
pmap_remove_all(p);
/* Account for removal of wired mappings. */
if (wirings != 0)
p->wire_count -= wirings;
}
if ((options & OBJPR_CLEANONLY) == 0) {
p->valid = 0;
vm_page_undirty(p);
}
vm_page_unlock(p);
continue;
}
if (vm_page_sleep_if_busy(p, TRUE, "vmopar"))
goto again;
KASSERT((p->flags & PG_FICTITIOUS) == 0,
("vm_object_page_remove: page %p is fictitious", p));
if ((options & OBJPR_CLEANONLY) != 0 && p->valid != 0) {
if ((options & OBJPR_NOTMAPPED) == 0)
pmap_remove_write(p);
if (p->dirty) {
vm_page_unlock(p);
continue;
}
1994-05-24 10:09:53 +00:00
}
if ((options & OBJPR_NOTMAPPED) == 0) {
pmap_remove_all(p);
/* Account for removal of wired mappings. */
if (wirings != 0) {
KASSERT(p->wire_count == wirings,
("inconsistent wire count %d %d %p",
p->wire_count, wirings, p));
p->wire_count = 0;
atomic_subtract_int(&cnt.v_wire_count, 1);
}
}
vm_page_free(p);
vm_page_unlock(p);
1994-05-24 10:09:53 +00:00
}
vm_object_pip_wakeup(object);
skipmemq:
if (__predict_false(object->cache != NULL))
vm_page_cache_free(object, start, end);
1994-05-24 10:09:53 +00:00
}
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
/*
* vm_object_page_cache:
*
* For the given object, attempt to move the specified clean
* pages to the cache queue. If a page is wired for any reason,
* then it will not be changed. Pages are specified by the given
* range ["start", "end"). As a special case, if "end" is zero,
* then the range extends from "start" to the end of the object.
* Any mappings to the specified pages are removed before the
* pages are moved to the cache queue.
*
* This operation should only be performed on objects that
* contain non-fictitious, managed pages.
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
*
* The object must be locked.
*/
void
vm_object_page_cache(vm_object_t object, vm_pindex_t start, vm_pindex_t end)
{
struct mtx *mtx, *new_mtx;
vm_page_t p, next;
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
KASSERT((object->flags & (OBJ_FICTITIOUS | OBJ_UNMANAGED)) == 0,
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
("vm_object_page_cache: illegal object %p", object));
if (object->resident_page_count == 0)
return;
p = vm_page_find_least(object, start);
/*
* Here, the variable "p" is either (1) the page with the least pindex
* greater than or equal to the parameter "start" or (2) NULL.
*/
mtx = NULL;
for (; p != NULL && (p->pindex < end || end == 0); p = next) {
next = TAILQ_NEXT(p, listq);
/*
* Avoid releasing and reacquiring the same page lock.
*/
new_mtx = vm_page_lockptr(p);
if (mtx != new_mtx) {
if (mtx != NULL)
mtx_unlock(mtx);
mtx = new_mtx;
mtx_lock(mtx);
}
vm_page_try_to_cache(p);
}
if (mtx != NULL)
mtx_unlock(mtx);
}
Long, long ago in r27464 special case code for mapping device-backed memory with 4MB pages was added to pmap_object_init_pt(). This code assumes that the pages of a OBJT_DEVICE object are always physically contiguous. Unfortunately, this is not always the case. For example, jhb@ informs me that the recently introduced /dev/ksyms driver creates a OBJT_DEVICE object that violates this assumption. Thus, this revision modifies pmap_object_init_pt() to abort the mapping if the OBJT_DEVICE object's pages are not physically contiguous. This revision also changes some inconsistent if not buggy behavior. For example, the i386 version aborts if the first 4MB virtual page that would be mapped is already valid. However, it incorrectly replaces any subsequent 4MB virtual page mappings that it encounters, potentially leaking a page table page. The amd64 version has a bug of my own creation. It potentially busies the wrong page and always an insufficent number of pages if it blocks allocating a page table page. To my knowledge, there have been no reports of these bugs, hence, their persistance. I suspect that the existing restrictions that pmap_object_init_pt() placed on the OBJT_DEVICE objects that it would choose to map, for example, that the first page must be aligned on a 2 or 4MB physical boundary and that the size of the mapping must be a multiple of the large page size, were enough to avoid triggering the bug for drivers like ksyms. However, one side effect of testing the OBJT_DEVICE object's pages for physical contiguity is that a dubious difference between pmap_object_init_pt() and the standard path for mapping devices pages, i.e., vm_fault(), has been eliminated. Previously, pmap_object_init_pt() would only instantiate the first PG_FICTITOUS page being mapped because it never examined the rest. Now, however, pmap_object_init_pt() uses the new function vm_object_populate() to instantiate them all (in order to support testing their physical contiguity). These pages need to be instantiated for the mechanism that I have prototyped for automatically maintaining the consistency of the PAT settings across multiple mappings, particularly, amd64's direct mapping, to work. (Translation: This change is also being made to support jhb@'s work on the Nvidia feature requests.) Discussed with: jhb@
2009-06-14 19:51:43 +00:00
/*
* Populate the specified range of the object with valid pages. Returns
* TRUE if the range is successfully populated and FALSE otherwise.
*
* Note: This function should be optimized to pass a larger array of
* pages to vm_pager_get_pages() before it is applied to a non-
* OBJT_DEVICE object.
*
* The object must be locked.
*/
boolean_t
vm_object_populate(vm_object_t object, vm_pindex_t start, vm_pindex_t end)
{
vm_page_t m, ma[1];
vm_pindex_t pindex;
int rv;
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
for (pindex = start; pindex < end; pindex++) {
m = vm_page_grab(object, pindex, VM_ALLOC_NORMAL |
VM_ALLOC_RETRY);
if (m->valid != VM_PAGE_BITS_ALL) {
ma[0] = m;
rv = vm_pager_get_pages(object, ma, 1, 0);
m = vm_page_lookup(object, pindex);
if (m == NULL)
break;
if (rv != VM_PAGER_OK) {
vm_page_lock(m);
Long, long ago in r27464 special case code for mapping device-backed memory with 4MB pages was added to pmap_object_init_pt(). This code assumes that the pages of a OBJT_DEVICE object are always physically contiguous. Unfortunately, this is not always the case. For example, jhb@ informs me that the recently introduced /dev/ksyms driver creates a OBJT_DEVICE object that violates this assumption. Thus, this revision modifies pmap_object_init_pt() to abort the mapping if the OBJT_DEVICE object's pages are not physically contiguous. This revision also changes some inconsistent if not buggy behavior. For example, the i386 version aborts if the first 4MB virtual page that would be mapped is already valid. However, it incorrectly replaces any subsequent 4MB virtual page mappings that it encounters, potentially leaking a page table page. The amd64 version has a bug of my own creation. It potentially busies the wrong page and always an insufficent number of pages if it blocks allocating a page table page. To my knowledge, there have been no reports of these bugs, hence, their persistance. I suspect that the existing restrictions that pmap_object_init_pt() placed on the OBJT_DEVICE objects that it would choose to map, for example, that the first page must be aligned on a 2 or 4MB physical boundary and that the size of the mapping must be a multiple of the large page size, were enough to avoid triggering the bug for drivers like ksyms. However, one side effect of testing the OBJT_DEVICE object's pages for physical contiguity is that a dubious difference between pmap_object_init_pt() and the standard path for mapping devices pages, i.e., vm_fault(), has been eliminated. Previously, pmap_object_init_pt() would only instantiate the first PG_FICTITOUS page being mapped because it never examined the rest. Now, however, pmap_object_init_pt() uses the new function vm_object_populate() to instantiate them all (in order to support testing their physical contiguity). These pages need to be instantiated for the mechanism that I have prototyped for automatically maintaining the consistency of the PAT settings across multiple mappings, particularly, amd64's direct mapping, to work. (Translation: This change is also being made to support jhb@'s work on the Nvidia feature requests.) Discussed with: jhb@
2009-06-14 19:51:43 +00:00
vm_page_free(m);
vm_page_unlock(m);
Long, long ago in r27464 special case code for mapping device-backed memory with 4MB pages was added to pmap_object_init_pt(). This code assumes that the pages of a OBJT_DEVICE object are always physically contiguous. Unfortunately, this is not always the case. For example, jhb@ informs me that the recently introduced /dev/ksyms driver creates a OBJT_DEVICE object that violates this assumption. Thus, this revision modifies pmap_object_init_pt() to abort the mapping if the OBJT_DEVICE object's pages are not physically contiguous. This revision also changes some inconsistent if not buggy behavior. For example, the i386 version aborts if the first 4MB virtual page that would be mapped is already valid. However, it incorrectly replaces any subsequent 4MB virtual page mappings that it encounters, potentially leaking a page table page. The amd64 version has a bug of my own creation. It potentially busies the wrong page and always an insufficent number of pages if it blocks allocating a page table page. To my knowledge, there have been no reports of these bugs, hence, their persistance. I suspect that the existing restrictions that pmap_object_init_pt() placed on the OBJT_DEVICE objects that it would choose to map, for example, that the first page must be aligned on a 2 or 4MB physical boundary and that the size of the mapping must be a multiple of the large page size, were enough to avoid triggering the bug for drivers like ksyms. However, one side effect of testing the OBJT_DEVICE object's pages for physical contiguity is that a dubious difference between pmap_object_init_pt() and the standard path for mapping devices pages, i.e., vm_fault(), has been eliminated. Previously, pmap_object_init_pt() would only instantiate the first PG_FICTITOUS page being mapped because it never examined the rest. Now, however, pmap_object_init_pt() uses the new function vm_object_populate() to instantiate them all (in order to support testing their physical contiguity). These pages need to be instantiated for the mechanism that I have prototyped for automatically maintaining the consistency of the PAT settings across multiple mappings, particularly, amd64's direct mapping, to work. (Translation: This change is also being made to support jhb@'s work on the Nvidia feature requests.) Discussed with: jhb@
2009-06-14 19:51:43 +00:00
break;
}
}
/*
* Keep "m" busy because a subsequent iteration may unlock
* the object.
*/
}
if (pindex > start) {
m = vm_page_lookup(object, start);
while (m != NULL && m->pindex < pindex) {
vm_page_wakeup(m);
m = TAILQ_NEXT(m, listq);
}
}
return (pindex == end);
}
1994-05-24 10:09:53 +00:00
/*
* Routine: vm_object_coalesce
* Function: Coalesces two objects backing up adjoining
* regions of memory into a single object.
*
* returns TRUE if objects were combined.
*
* NOTE: Only works at the moment if the second object is NULL -
* if it's not, which object do we lock first?
*
* Parameters:
* prev_object First object to coalesce
* prev_offset Offset into prev_object
* prev_size Size of reference to prev_object
* next_size Size of reference to the second object
* reserved Indicator that extension region has
* swap accounted for
1994-05-24 10:09:53 +00:00
*
* Conditions:
* The object must *not* be locked.
*/
1995-05-30 08:16:23 +00:00
boolean_t
vm_object_coalesce(vm_object_t prev_object, vm_ooffset_t prev_offset,
vm_size_t prev_size, vm_size_t next_size, boolean_t reserved)
1994-05-24 10:09:53 +00:00
{
vm_pindex_t next_pindex;
1994-05-24 10:09:53 +00:00
if (prev_object == NULL)
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);
VM_OBJECT_LOCK(prev_object);
if (prev_object->type != OBJT_DEFAULT &&
prev_object->type != OBJT_SWAP) {
VM_OBJECT_UNLOCK(prev_object);
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
* Try to collapse the object first
1994-05-24 10:09:53 +00:00
*/
vm_object_collapse(prev_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
* Can't coalesce if: . more than one reference . paged out . shadows
* another object . has a copy elsewhere (any of which mean that the
* pages not mapped to prev_entry may be in use anyway)
1994-05-24 10:09:53 +00:00
*/
if (prev_object->backing_object != NULL) {
VM_OBJECT_UNLOCK(prev_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
return (FALSE);
1994-05-24 10:09:53 +00:00
}
prev_size >>= PAGE_SHIFT;
next_size >>= PAGE_SHIFT;
next_pindex = OFF_TO_IDX(prev_offset) + prev_size;
if ((prev_object->ref_count > 1) &&
(prev_object->size != next_pindex)) {
VM_OBJECT_UNLOCK(prev_object);
return (FALSE);
}
/*
* Account for the charge.
*/
if (prev_object->cred != NULL) {
/*
* If prev_object was charged, then this mapping,
* althought not charged now, may become writable
* later. Non-NULL cred in the object would prevent
* swap reservation during enabling of the write
* access, so reserve swap now. Failed reservation
* cause allocation of the separate object for the map
* entry, and swap reservation for this entry is
* managed in appropriate time.
*/
if (!reserved && !swap_reserve_by_cred(ptoa(next_size),
prev_object->cred)) {
return (FALSE);
}
prev_object->charge += ptoa(next_size);
}
1994-05-24 10:09:53 +00:00
/*
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
* Remove any pages that may still be in the object from a previous
* deallocation.
1994-05-24 10:09:53 +00:00
*/
if (next_pindex < prev_object->size) {
vm_object_page_remove(prev_object, next_pindex, next_pindex +
next_size, 0);
if (prev_object->type == OBJT_SWAP)
swap_pager_freespace(prev_object,
next_pindex, next_size);
#if 0
if (prev_object->cred != NULL) {
KASSERT(prev_object->charge >=
ptoa(prev_object->size - next_pindex),
("object %p overcharged 1 %jx %jx", prev_object,
(uintmax_t)next_pindex, (uintmax_t)next_size));
prev_object->charge -= ptoa(prev_object->size -
next_pindex);
}
#endif
}
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
* Extend the object if necessary.
1994-05-24 10:09:53 +00:00
*/
if (next_pindex + next_size > prev_object->size)
prev_object->size = next_pindex + next_size;
1994-05-24 10:09:53 +00:00
VM_OBJECT_UNLOCK(prev_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
return (TRUE);
1994-05-24 10:09:53 +00:00
}
void
vm_object_set_writeable_dirty(vm_object_t object)
{
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
if (object->type != OBJT_VNODE)
return;
object->generation++;
if ((object->flags & OBJ_MIGHTBEDIRTY) != 0)
return;
vm_object_set_flag(object, OBJ_MIGHTBEDIRTY);
}
#include "opt_ddb.h"
#ifdef DDB
#include <sys/kernel.h>
#include <sys/cons.h>
#include <ddb/ddb.h>
static int
_vm_object_in_map(vm_map_t map, vm_object_t object, vm_map_entry_t entry)
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_t tmpm;
vm_map_entry_t tmpe;
vm_object_t obj;
int entcount;
if (map == 0)
return 0;
if (entry == 0) {
tmpe = map->header.next;
entcount = map->nentries;
while (entcount-- && (tmpe != &map->header)) {
if (_vm_object_in_map(map, object, tmpe)) {
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
return 1;
}
tmpe = tmpe->next;
}
} else if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) {
tmpm = entry->object.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
tmpe = tmpm->header.next;
entcount = tmpm->nentries;
while (entcount-- && tmpe != &tmpm->header) {
if (_vm_object_in_map(tmpm, object, tmpe)) {
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
return 1;
}
tmpe = tmpe->next;
}
} else if ((obj = entry->object.vm_object) != NULL) {
2001-07-04 19:00:13 +00:00
for (; obj; obj = obj->backing_object)
if (obj == object) {
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
return 1;
}
}
return 0;
}
static int
vm_object_in_map(vm_object_t object)
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
{
struct proc *p;
/* sx_slock(&allproc_lock); */
FOREACH_PROC_IN_SYSTEM(p) {
if (!p->p_vmspace /* || (p->p_flag & (P_SYSTEM|P_WEXIT)) */)
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
continue;
if (_vm_object_in_map(&p->p_vmspace->vm_map, object, 0)) {
/* sx_sunlock(&allproc_lock); */
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
return 1;
}
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
}
/* sx_sunlock(&allproc_lock); */
if (_vm_object_in_map(kernel_map, object, 0))
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
return 1;
if (_vm_object_in_map(kmem_map, object, 0))
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
return 1;
if (_vm_object_in_map(pager_map, object, 0))
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
return 1;
if (_vm_object_in_map(buffer_map, object, 0))
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
return 1;
return 0;
}
DB_SHOW_COMMAND(vmochk, vm_object_check)
{
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_object_t object;
/*
* make sure that internal objs are in a map somewhere
* and none have zero ref counts.
*/
TAILQ_FOREACH(object, &vm_object_list, object_list) {
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
if (object->handle == NULL &&
(object->type == OBJT_DEFAULT || object->type == OBJT_SWAP)) {
if (object->ref_count == 0) {
db_printf("vmochk: internal obj has zero ref count: %ld\n",
(long)object->size);
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_object_in_map(object)) {
1998-07-11 11:30:46 +00:00
db_printf(
"vmochk: internal obj is not in a map: "
"ref: %d, size: %lu: 0x%lx, backing_object: %p\n",
object->ref_count, (u_long)object->size,
(u_long)object->size,
(void *)object->backing_object);
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
/*
* vm_object_print: [ debug ]
*/
DB_SHOW_COMMAND(object, vm_object_print_static)
1994-05-24 10:09:53 +00:00
{
/* XXX convert args. */
vm_object_t object = (vm_object_t)addr;
boolean_t full = have_addr;
vm_page_t p;
1994-05-24 10:09:53 +00:00
/* XXX count is an (unused) arg. Avoid shadowing it. */
#define count was_count
int count;
1994-05-24 10:09:53 +00:00
if (object == NULL)
return;
db_iprintf(
"Object %p: type=%d, size=0x%jx, res=%d, ref=%d, flags=0x%x ruid %d charge %jx\n",
2002-11-07 23:03:04 +00:00
object, (int)object->type, (uintmax_t)object->size,
object->resident_page_count, object->ref_count, object->flags,
object->cred ? object->cred->cr_ruid : -1, (uintmax_t)object->charge);
2002-11-07 23:03:04 +00:00
db_iprintf(" sref=%d, backing_object(%d)=(%p)+0x%jx\n",
object->shadow_count,
object->backing_object ? object->backing_object->ref_count : 0,
2002-11-07 23:03:04 +00:00
object->backing_object, (uintmax_t)object->backing_object_offset);
1994-05-24 10:09:53 +00:00
if (!full)
return;
db_indent += 2;
1994-05-24 10:09:53 +00:00
count = 0;
TAILQ_FOREACH(p, &object->memq, listq) {
1994-05-24 10:09:53 +00:00
if (count == 0)
db_iprintf("memory:=");
1994-05-24 10:09:53 +00:00
else if (count == 6) {
db_printf("\n");
db_iprintf(" ...");
1994-05-24 10:09:53 +00:00
count = 0;
} else
db_printf(",");
1994-05-24 10:09:53 +00:00
count++;
2002-11-07 23:03:04 +00:00
db_printf("(off=0x%jx,page=0x%jx)",
(uintmax_t)p->pindex, (uintmax_t)VM_PAGE_TO_PHYS(p));
1994-05-24 10:09:53 +00:00
}
if (count != 0)
db_printf("\n");
db_indent -= 2;
1994-05-24 10:09:53 +00:00
}
/* XXX. */
#undef count
/* XXX need this non-static entry for calling from vm_map_print. */
void
vm_object_print(
/* db_expr_t */ long addr,
boolean_t have_addr,
/* db_expr_t */ long count,
char *modif)
{
vm_object_print_static(addr, have_addr, count, modif);
}
DB_SHOW_COMMAND(vmopag, vm_object_print_pages)
{
vm_object_t object;
vm_pindex_t fidx;
vm_paddr_t pa;
vm_page_t m, prev_m;
int rcount, nl, c;
nl = 0;
TAILQ_FOREACH(object, &vm_object_list, object_list) {
1998-07-11 11:30:46 +00:00
db_printf("new object: %p\n", (void *)object);
if (nl > 18) {
c = cngetc();
if (c != ' ')
return;
nl = 0;
}
nl++;
rcount = 0;
fidx = 0;
pa = -1;
TAILQ_FOREACH(m, &object->memq, listq) {
if (m->pindex > 128)
break;
if ((prev_m = TAILQ_PREV(m, pglist, listq)) != NULL &&
prev_m->pindex + 1 != m->pindex) {
if (rcount) {
db_printf(" index(%ld)run(%d)pa(0x%lx)\n",
(long)fidx, rcount, (long)pa);
if (nl > 18) {
c = cngetc();
if (c != ' ')
return;
nl = 0;
}
nl++;
rcount = 0;
}
}
if (rcount &&
(VM_PAGE_TO_PHYS(m) == pa + rcount * PAGE_SIZE)) {
++rcount;
continue;
}
if (rcount) {
db_printf(" index(%ld)run(%d)pa(0x%lx)\n",
(long)fidx, rcount, (long)pa);
if (nl > 18) {
c = cngetc();
if (c != ' ')
return;
nl = 0;
}
nl++;
}
fidx = m->pindex;
pa = VM_PAGE_TO_PHYS(m);
rcount = 1;
}
if (rcount) {
db_printf(" index(%ld)run(%d)pa(0x%lx)\n",
(long)fidx, rcount, (long)pa);
if (nl > 18) {
c = cngetc();
if (c != ' ')
return;
nl = 0;
}
nl++;
}
}
}
#endif /* DDB */