freebsd-dev/sys/vm/vm_glue.c

1094 lines
26 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_glue.c 8.6 (Berkeley) 1/5/94
1994-05-24 10:09:53 +00:00
*
*
* Copyright (c) 1987, 1990 Carnegie-Mellon University.
* All rights reserved.
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
*
1994-05-24 10:09:53 +00:00
* Permission to use, copy, modify and distribute this software and
* its documentation is hereby granted, provided that both the copyright
* notice and this permission notice appear in all copies of the
* software, derivative works or modified versions, and any portions
* thereof, and that both notices appear in supporting documentation.
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
*
* CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
* CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
1994-05-24 10:09:53 +00:00
* FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
*
1994-05-24 10:09:53 +00:00
* Carnegie Mellon requests users of this software to return to
*
* Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU
* School of Computer Science
* Carnegie Mellon University
* Pittsburgh PA 15213-3890
*
* any improvements or extensions that they make and grant Carnegie the
* rights to redistribute these changes.
*/
2003-06-11 23:50:51 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_vm.h"
#include "opt_kstack_pages.h"
#include "opt_kstack_max_pages.h"
#include "opt_kstack_usage_prof.h"
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/limits.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mutex.h>
1994-05-24 10:09:53 +00:00
#include <sys/proc.h>
#include <sys/racct.h>
1994-05-24 10:09:53 +00:00
#include <sys/resourcevar.h>
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
#include <sys/rwlock.h>
#include <sys/sched.h>
#include <sys/sf_buf.h>
#include <sys/shm.h>
#include <sys/vmmeter.h>
#include <sys/vmem.h>
#include <sys/sx.h>
#include <sys/sysctl.h>
#include <sys/_kstack_cache.h>
#include <sys/eventhandler.h>
#include <sys/kernel.h>
#include <sys/ktr.h>
The biggie: Get rid of the UPAGES from the top of the per-process address space. (!) Have each process use the kernel stack and pcb in the kvm space. Since the stacks are at a different address, we cannot copy the stack at fork() and allow the child to return up through the function call tree to return to user mode - create a new execution context and have the new process begin executing from cpu_switch() and go to user mode directly. In theory this should speed up fork a bit. Context switch the tss_esp0 pointer in the common tss. This is a lot simpler since than swithching the gdt[GPROC0_SEL].sd.sd_base pointer to each process's tss since the esp0 pointer is a 32 bit pointer, and the sd_base setting is split into three different bit sections at non-aligned boundaries and requires a lot of twiddling to reset. The 8K of memory at the top of the process space is now empty, and unmapped (and unmappable, it's higher than VM_MAXUSER_ADDRESS). Simplity the pmap code to manage process contexts, we no longer have to double map the UPAGES, this simplifies and should measuably speed up fork(). The following parts came from John Dyson: Set PG_G on the UPAGES that are now in kernel context, and invalidate them when swapping them out. Move the upages object (upobj) from the vmspace to the proc structure. Now that the UPAGES (pcb and kernel stack) are out of user space, make rfork(..RFMEM..) do what was intended by sharing the vmspace entirely via reference counting rather than simply inheriting the mappings.
1997-04-07 07:16:06 +00:00
#include <sys/unistd.h>
1994-05-24 10:09:53 +00:00
#include <vm/vm.h>
#include <vm/vm_param.h>
#include <vm/pmap.h>
#include <vm/vm_map.h>
1994-05-24 10:09:53 +00:00
#include <vm/vm_page.h>
#include <vm/vm_pageout.h>
#include <vm/vm_object.h>
1994-05-24 10:09:53 +00:00
#include <vm/vm_kern.h>
#include <vm/vm_extern.h>
#include <vm/vm_pager.h>
#include <vm/swap_pager.h>
#include <machine/cpu.h>
#ifndef NO_SWAPPING
static int swapout(struct proc *);
static void swapclear(struct proc *);
static void vm_thread_swapin(struct thread *td);
static void vm_thread_swapout(struct thread *td);
#endif
/*
* MPSAFE
*
* WARNING! This code calls vm_map_check_protection() which only checks
* the associated vm_map_entry range. It does not determine whether the
* contents of the memory is actually readable or writable. In most cases
* just checking the vm_map_entry is sufficient within the kernel's address
* space.
*/
1994-05-24 10:09:53 +00:00
int
kernacc(addr, len, rw)
void *addr;
1994-05-24 10:09:53 +00:00
int len, rw;
{
boolean_t rv;
vm_offset_t saddr, eaddr;
vm_prot_t prot;
1994-05-24 10:09:53 +00:00
KASSERT((rw & ~VM_PROT_ALL) == 0,
("illegal ``rw'' argument to kernacc (%x)\n", rw));
if ((vm_offset_t)addr + len > kernel_map->max_offset ||
(vm_offset_t)addr + len < (vm_offset_t)addr)
return (FALSE);
prot = rw;
saddr = trunc_page((vm_offset_t)addr);
eaddr = round_page((vm_offset_t)addr + len);
vm_map_lock_read(kernel_map);
1994-05-24 10:09:53 +00:00
rv = vm_map_check_protection(kernel_map, saddr, eaddr, prot);
vm_map_unlock_read(kernel_map);
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
return (rv == TRUE);
1994-05-24 10:09:53 +00:00
}
/*
* MPSAFE
*
* WARNING! This code calls vm_map_check_protection() which only checks
* the associated vm_map_entry range. It does not determine whether the
* contents of the memory is actually readable or writable. vmapbuf(),
* vm_fault_quick(), or copyin()/copout()/su*()/fu*() functions should be
* used in conjuction with this call.
*/
1994-05-24 10:09:53 +00:00
int
useracc(addr, len, rw)
void *addr;
1994-05-24 10:09:53 +00:00
int len, rw;
{
boolean_t rv;
vm_prot_t prot;
vm_map_t map;
1994-05-24 10:09:53 +00:00
KASSERT((rw & ~VM_PROT_ALL) == 0,
("illegal ``rw'' argument to useracc (%x)\n", rw));
prot = rw;
map = &curproc->p_vmspace->vm_map;
if ((vm_offset_t)addr + len > vm_map_max(map) ||
(vm_offset_t)addr + len < (vm_offset_t)addr) {
return (FALSE);
}
vm_map_lock_read(map);
rv = vm_map_check_protection(map, trunc_page((vm_offset_t)addr),
round_page((vm_offset_t)addr + len), prot);
vm_map_unlock_read(map);
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
return (rv == TRUE);
1994-05-24 10:09:53 +00:00
}
int
2004-03-15 06:43:51 +00:00
vslock(void *addr, size_t len)
{
vm_offset_t end, last, start;
vm_size_t npages;
int error;
last = (vm_offset_t)addr + len;
start = trunc_page((vm_offset_t)addr);
end = round_page(last);
if (last < (vm_offset_t)addr || end < (vm_offset_t)addr)
return (EINVAL);
npages = atop(end - start);
if (npages > vm_page_max_wired)
return (ENOMEM);
#if 0
/*
* XXX - not yet
*
* The limit for transient usage of wired pages should be
* larger than for "permanent" wired pages (mlock()).
*
* Also, the sysctl code, which is the only present user
* of vslock(), does a hard loop on EAGAIN.
*/
if (npages + vm_cnt.v_wire_count > vm_page_max_wired)
return (EAGAIN);
#endif
error = vm_map_wire(&curproc->p_vmspace->vm_map, start, end,
VM_MAP_WIRE_SYSTEM | VM_MAP_WIRE_NOHOLES);
/*
* Return EFAULT on error to match copy{in,out}() behaviour
* rather than returning ENOMEM like mlock() would.
*/
return (error == KERN_SUCCESS ? 0 : EFAULT);
}
void
2004-03-15 06:43:51 +00:00
vsunlock(void *addr, size_t len)
{
/* Rely on the parameter sanity checks performed by vslock(). */
(void)vm_map_unwire(&curproc->p_vmspace->vm_map,
trunc_page((vm_offset_t)addr), round_page((vm_offset_t)addr + len),
2004-03-15 06:43:51 +00:00
VM_MAP_WIRE_SYSTEM | VM_MAP_WIRE_NOHOLES);
}
/*
* Pin the page contained within the given object at the given offset. If the
* page is not resident, allocate and load it using the given object's pager.
* Return the pinned page if successful; otherwise, return NULL.
*/
static vm_page_t
vm_imgact_hold_page(vm_object_t object, vm_ooffset_t offset)
{
vm_page_t m;
vm_pindex_t pindex;
int rv;
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WLOCK(object);
pindex = OFF_TO_IDX(offset);
m = vm_page_grab(object, pindex, VM_ALLOC_NORMAL);
if (m->valid != VM_PAGE_BITS_ALL) {
rv = vm_pager_get_pages(object, &m, 1, 0);
if (rv != VM_PAGER_OK) {
vm_page_lock(m);
vm_page_free(m);
vm_page_unlock(m);
m = NULL;
goto out;
}
}
vm_page_xunbusy(m);
vm_page_lock(m);
vm_page_hold(m);
vm_page_activate(m);
vm_page_unlock(m);
out:
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WUNLOCK(object);
return (m);
}
/*
* Return a CPU private mapping to the page at the given offset within the
* given object. The page is pinned before it is mapped.
*/
struct sf_buf *
vm_imgact_map_page(vm_object_t object, vm_ooffset_t offset)
{
vm_page_t m;
m = vm_imgact_hold_page(object, offset);
if (m == NULL)
return (NULL);
sched_pin();
return (sf_buf_alloc(m, SFB_CPUPRIVATE));
}
/*
* Destroy the given CPU private mapping and unpin the page that it mapped.
*/
void
vm_imgact_unmap_page(struct sf_buf *sf)
{
vm_page_t m;
m = sf_buf_page(sf);
sf_buf_free(sf);
sched_unpin();
vm_page_lock(m);
vm_page_unhold(m);
vm_page_unlock(m);
}
void
vm_sync_icache(vm_map_t map, vm_offset_t va, vm_offset_t sz)
{
pmap_sync_icache(map->pmap, va, sz);
}
struct kstack_cache_entry *kstack_cache;
static int kstack_cache_size = 128;
static int kstacks;
static struct mtx kstack_cache_mtx;
MTX_SYSINIT(kstack_cache, &kstack_cache_mtx, "kstkch", MTX_DEF);
SYSCTL_INT(_vm, OID_AUTO, kstack_cache_size, CTLFLAG_RW, &kstack_cache_size, 0,
"");
SYSCTL_INT(_vm, OID_AUTO, kstacks, CTLFLAG_RD, &kstacks, 0,
"");
#ifndef KSTACK_MAX_PAGES
#define KSTACK_MAX_PAGES 32
#endif
/*
* Create the kernel stack (including pcb for i386) for a new thread.
* This routine directly affects the fork perf for a process and
* create performance for a thread.
*/
int
vm_thread_new(struct thread *td, int pages)
{
vm_object_t ksobj;
vm_offset_t ks;
vm_page_t m, ma[KSTACK_MAX_PAGES];
struct kstack_cache_entry *ks_ce;
int i;
/* Bounds check */
if (pages <= 1)
pages = KSTACK_PAGES;
else if (pages > KSTACK_MAX_PAGES)
pages = KSTACK_MAX_PAGES;
if (pages == KSTACK_PAGES) {
mtx_lock(&kstack_cache_mtx);
if (kstack_cache != NULL) {
ks_ce = kstack_cache;
kstack_cache = ks_ce->next_ks_entry;
mtx_unlock(&kstack_cache_mtx);
td->td_kstack_obj = ks_ce->ksobj;
td->td_kstack = (vm_offset_t)ks_ce;
td->td_kstack_pages = KSTACK_PAGES;
return (1);
}
mtx_unlock(&kstack_cache_mtx);
}
/*
* Allocate an object for the kstack.
*/
ksobj = vm_object_allocate(OBJT_DEFAULT, pages);
/*
* Get a kernel virtual address for this thread's kstack.
*/
#if defined(__mips__)
/*
* We need to align the kstack's mapped address to fit within
* a single TLB entry.
*/
if (vmem_xalloc(kernel_arena, (pages + KSTACK_GUARD_PAGES) * PAGE_SIZE,
PAGE_SIZE * 2, 0, 0, VMEM_ADDR_MIN, VMEM_ADDR_MAX,
M_BESTFIT | M_NOWAIT, &ks)) {
ks = 0;
}
#else
ks = kva_alloc((pages + KSTACK_GUARD_PAGES) * PAGE_SIZE);
#endif
if (ks == 0) {
printf("vm_thread_new: kstack allocation failed\n");
vm_object_deallocate(ksobj);
return (0);
}
atomic_add_int(&kstacks, 1);
if (KSTACK_GUARD_PAGES != 0) {
pmap_qremove(ks, KSTACK_GUARD_PAGES);
ks += KSTACK_GUARD_PAGES * PAGE_SIZE;
}
td->td_kstack_obj = ksobj;
td->td_kstack = ks;
/*
* Knowing the number of pages allocated is useful when you
* want to deallocate them.
*/
td->td_kstack_pages = pages;
/*
* For the length of the stack, link in a real page of ram for each
* page of stack.
*/
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WLOCK(ksobj);
for (i = 0; i < pages; i++) {
/*
* Get a kernel stack page.
*/
m = vm_page_grab(ksobj, i, VM_ALLOC_NOBUSY |
VM_ALLOC_NORMAL | VM_ALLOC_WIRED);
ma[i] = m;
m->valid = VM_PAGE_BITS_ALL;
}
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WUNLOCK(ksobj);
pmap_qenter(ks, ma, pages);
return (1);
}
static void
vm_thread_stack_dispose(vm_object_t ksobj, vm_offset_t ks, int pages)
{
vm_page_t m;
int i;
atomic_add_int(&kstacks, -1);
pmap_qremove(ks, pages);
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WLOCK(ksobj);
for (i = 0; i < pages; i++) {
m = vm_page_lookup(ksobj, i);
if (m == NULL)
panic("vm_thread_dispose: kstack already missing?");
vm_page_lock(m);
vm_page_unwire(m, PQ_INACTIVE);
vm_page_free(m);
vm_page_unlock(m);
}
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WUNLOCK(ksobj);
vm_object_deallocate(ksobj);
kva_free(ks - (KSTACK_GUARD_PAGES * PAGE_SIZE),
(pages + KSTACK_GUARD_PAGES) * PAGE_SIZE);
}
/*
* Dispose of a thread's kernel stack.
*/
void
vm_thread_dispose(struct thread *td)
{
vm_object_t ksobj;
vm_offset_t ks;
struct kstack_cache_entry *ks_ce;
int pages;
pages = td->td_kstack_pages;
ksobj = td->td_kstack_obj;
ks = td->td_kstack;
td->td_kstack = 0;
td->td_kstack_pages = 0;
if (pages == KSTACK_PAGES && kstacks <= kstack_cache_size) {
ks_ce = (struct kstack_cache_entry *)ks;
ks_ce->ksobj = ksobj;
mtx_lock(&kstack_cache_mtx);
ks_ce->next_ks_entry = kstack_cache;
kstack_cache = ks_ce;
mtx_unlock(&kstack_cache_mtx);
return;
}
vm_thread_stack_dispose(ksobj, ks, pages);
}
static void
vm_thread_stack_lowmem(void *nulll)
{
struct kstack_cache_entry *ks_ce, *ks_ce1;
mtx_lock(&kstack_cache_mtx);
ks_ce = kstack_cache;
kstack_cache = NULL;
mtx_unlock(&kstack_cache_mtx);
while (ks_ce != NULL) {
ks_ce1 = ks_ce;
ks_ce = ks_ce->next_ks_entry;
vm_thread_stack_dispose(ks_ce1->ksobj, (vm_offset_t)ks_ce1,
KSTACK_PAGES);
}
}
static void
kstack_cache_init(void *nulll)
{
EVENTHANDLER_REGISTER(vm_lowmem, vm_thread_stack_lowmem, NULL,
EVENTHANDLER_PRI_ANY);
}
SYSINIT(vm_kstacks, SI_SUB_KTHREAD_INIT, SI_ORDER_ANY, kstack_cache_init, NULL);
#ifdef KSTACK_USAGE_PROF
/*
* Track maximum stack used by a thread in kernel.
*/
static int max_kstack_used;
SYSCTL_INT(_debug, OID_AUTO, max_kstack_used, CTLFLAG_RD,
&max_kstack_used, 0,
"Maxiumum stack depth used by a thread in kernel");
void
intr_prof_stack_use(struct thread *td, struct trapframe *frame)
{
vm_offset_t stack_top;
vm_offset_t current;
int used, prev_used;
/*
* Testing for interrupted kernel mode isn't strictly
* needed. It optimizes the execution, since interrupts from
* usermode will have only the trap frame on the stack.
*/
if (TRAPF_USERMODE(frame))
return;
stack_top = td->td_kstack + td->td_kstack_pages * PAGE_SIZE;
current = (vm_offset_t)(uintptr_t)&stack_top;
/*
* Try to detect if interrupt is using kernel thread stack.
* Hardware could use a dedicated stack for interrupt handling.
*/
if (stack_top <= current || current < td->td_kstack)
return;
used = stack_top - current;
for (;;) {
prev_used = max_kstack_used;
if (prev_used >= used)
break;
if (atomic_cmpset_int(&max_kstack_used, prev_used, used))
break;
}
}
#endif /* KSTACK_USAGE_PROF */
#ifndef NO_SWAPPING
/*
* Allow a thread's kernel stack to be paged out.
*/
static void
vm_thread_swapout(struct thread *td)
{
vm_object_t ksobj;
vm_page_t m;
int i, pages;
cpu_thread_swapout(td);
pages = td->td_kstack_pages;
ksobj = td->td_kstack_obj;
pmap_qremove(td->td_kstack, pages);
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WLOCK(ksobj);
for (i = 0; i < pages; i++) {
m = vm_page_lookup(ksobj, i);
if (m == NULL)
panic("vm_thread_swapout: kstack already missing?");
vm_page_dirty(m);
vm_page_lock(m);
vm_page_unwire(m, PQ_INACTIVE);
vm_page_unlock(m);
}
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WUNLOCK(ksobj);
}
/*
* Bring the kernel stack for a specified thread back in.
*/
static void
vm_thread_swapin(struct thread *td)
{
vm_object_t ksobj;
vm_page_t ma[KSTACK_MAX_PAGES];
int i, j, pages, rv;
pages = td->td_kstack_pages;
ksobj = td->td_kstack_obj;
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WLOCK(ksobj);
for (i = 0; i < pages; i++)
ma[i] = vm_page_grab(ksobj, i, VM_ALLOC_NORMAL |
2010-04-13 06:48:37 +00:00
VM_ALLOC_WIRED);
for (i = 0; i < pages; i++) {
if (ma[i]->valid != VM_PAGE_BITS_ALL) {
vm_page_assert_xbusied(ma[i]);
vm_object_pip_add(ksobj, 1);
for (j = i + 1; j < pages; j++) {
if (ma[j]->valid != VM_PAGE_BITS_ALL)
vm_page_assert_xbusied(ma[j]);
if (ma[j]->valid == VM_PAGE_BITS_ALL)
break;
}
rv = vm_pager_get_pages(ksobj, ma + i, j - i, 0);
if (rv != VM_PAGER_OK)
panic("vm_thread_swapin: cannot get kstack for proc: %d",
td->td_proc->p_pid);
/*
* All pages in the array are in place, due to the
* pager is always the swap pager, which doesn't
* free or remove wired non-req pages from object.
*/
vm_object_pip_wakeup(ksobj);
vm_page_xunbusy(ma[i]);
} else if (vm_page_xbusied(ma[i]))
vm_page_xunbusy(ma[i]);
}
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WUNLOCK(ksobj);
pmap_qenter(td->td_kstack, ma, pages);
cpu_thread_swapin(td);
}
#endif /* !NO_SWAPPING */
1994-05-24 10:09:53 +00:00
/*
* Implement fork's actions on an address space.
* Here we arrange for the address space to be copied or referenced,
* allocate a user struct (pcb and kernel stack), then call the
* machine-dependent layer to fill those in and make the new process
The biggie: Get rid of the UPAGES from the top of the per-process address space. (!) Have each process use the kernel stack and pcb in the kvm space. Since the stacks are at a different address, we cannot copy the stack at fork() and allow the child to return up through the function call tree to return to user mode - create a new execution context and have the new process begin executing from cpu_switch() and go to user mode directly. In theory this should speed up fork a bit. Context switch the tss_esp0 pointer in the common tss. This is a lot simpler since than swithching the gdt[GPROC0_SEL].sd.sd_base pointer to each process's tss since the esp0 pointer is a 32 bit pointer, and the sd_base setting is split into three different bit sections at non-aligned boundaries and requires a lot of twiddling to reset. The 8K of memory at the top of the process space is now empty, and unmapped (and unmappable, it's higher than VM_MAXUSER_ADDRESS). Simplity the pmap code to manage process contexts, we no longer have to double map the UPAGES, this simplifies and should measuably speed up fork(). The following parts came from John Dyson: Set PG_G on the UPAGES that are now in kernel context, and invalidate them when swapping them out. Move the upages object (upobj) from the vmspace to the proc structure. Now that the UPAGES (pcb and kernel stack) are out of user space, make rfork(..RFMEM..) do what was intended by sharing the vmspace entirely via reference counting rather than simply inheriting the mappings.
1997-04-07 07:16:06 +00:00
* ready to run. The new process is set up so that it returns directly
* to user mode to avoid stack copying and relocation problems.
1994-05-24 10:09:53 +00:00
*/
int
vm_forkproc(td, p2, td2, vm2, flags)
struct thread *td;
struct proc *p2;
struct thread *td2;
struct vmspace *vm2;
The biggie: Get rid of the UPAGES from the top of the per-process address space. (!) Have each process use the kernel stack and pcb in the kvm space. Since the stacks are at a different address, we cannot copy the stack at fork() and allow the child to return up through the function call tree to return to user mode - create a new execution context and have the new process begin executing from cpu_switch() and go to user mode directly. In theory this should speed up fork a bit. Context switch the tss_esp0 pointer in the common tss. This is a lot simpler since than swithching the gdt[GPROC0_SEL].sd.sd_base pointer to each process's tss since the esp0 pointer is a 32 bit pointer, and the sd_base setting is split into three different bit sections at non-aligned boundaries and requires a lot of twiddling to reset. The 8K of memory at the top of the process space is now empty, and unmapped (and unmappable, it's higher than VM_MAXUSER_ADDRESS). Simplity the pmap code to manage process contexts, we no longer have to double map the UPAGES, this simplifies and should measuably speed up fork(). The following parts came from John Dyson: Set PG_G on the UPAGES that are now in kernel context, and invalidate them when swapping them out. Move the upages object (upobj) from the vmspace to the proc structure. Now that the UPAGES (pcb and kernel stack) are out of user space, make rfork(..RFMEM..) do what was intended by sharing the vmspace entirely via reference counting rather than simply inheriting the mappings.
1997-04-07 07:16:06 +00:00
int flags;
1994-05-24 10:09:53 +00:00
{
struct proc *p1 = td->td_proc;
int error;
1999-12-06 04:53:08 +00:00
if ((flags & RFPROC) == 0) {
/*
* Divorce the memory, if it is shared, essentially
* this changes shared memory amongst threads, into
* COW locally.
*/
if ((flags & RFMEM) == 0) {
if (p1->p_vmspace->vm_refcnt > 1) {
error = vmspace_unshare(p1);
if (error)
return (error);
1999-12-06 04:53:08 +00:00
}
}
cpu_fork(td, p2, td2, flags);
return (0);
1999-12-06 04:53:08 +00:00
}
if (flags & RFMEM) {
p2->p_vmspace = p1->p_vmspace;
atomic_add_int(&p1->p_vmspace->vm_refcnt, 1);
}
while (vm_page_count_severe()) {
VM_WAIT;
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
}
1994-05-24 10:09:53 +00:00
if ((flags & RFMEM) == 0) {
p2->p_vmspace = vm2;
The biggie: Get rid of the UPAGES from the top of the per-process address space. (!) Have each process use the kernel stack and pcb in the kvm space. Since the stacks are at a different address, we cannot copy the stack at fork() and allow the child to return up through the function call tree to return to user mode - create a new execution context and have the new process begin executing from cpu_switch() and go to user mode directly. In theory this should speed up fork a bit. Context switch the tss_esp0 pointer in the common tss. This is a lot simpler since than swithching the gdt[GPROC0_SEL].sd.sd_base pointer to each process's tss since the esp0 pointer is a 32 bit pointer, and the sd_base setting is split into three different bit sections at non-aligned boundaries and requires a lot of twiddling to reset. The 8K of memory at the top of the process space is now empty, and unmapped (and unmappable, it's higher than VM_MAXUSER_ADDRESS). Simplity the pmap code to manage process contexts, we no longer have to double map the UPAGES, this simplifies and should measuably speed up fork(). The following parts came from John Dyson: Set PG_G on the UPAGES that are now in kernel context, and invalidate them when swapping them out. Move the upages object (upobj) from the vmspace to the proc structure. Now that the UPAGES (pcb and kernel stack) are out of user space, make rfork(..RFMEM..) do what was intended by sharing the vmspace entirely via reference counting rather than simply inheriting the mappings.
1997-04-07 07:16:06 +00:00
if (p1->p_vmspace->vm_shm)
shmfork(p1, p2);
}
1994-05-24 10:09:53 +00:00
/*
The biggie: Get rid of the UPAGES from the top of the per-process address space. (!) Have each process use the kernel stack and pcb in the kvm space. Since the stacks are at a different address, we cannot copy the stack at fork() and allow the child to return up through the function call tree to return to user mode - create a new execution context and have the new process begin executing from cpu_switch() and go to user mode directly. In theory this should speed up fork a bit. Context switch the tss_esp0 pointer in the common tss. This is a lot simpler since than swithching the gdt[GPROC0_SEL].sd.sd_base pointer to each process's tss since the esp0 pointer is a 32 bit pointer, and the sd_base setting is split into three different bit sections at non-aligned boundaries and requires a lot of twiddling to reset. The 8K of memory at the top of the process space is now empty, and unmapped (and unmappable, it's higher than VM_MAXUSER_ADDRESS). Simplity the pmap code to manage process contexts, we no longer have to double map the UPAGES, this simplifies and should measuably speed up fork(). The following parts came from John Dyson: Set PG_G on the UPAGES that are now in kernel context, and invalidate them when swapping them out. Move the upages object (upobj) from the vmspace to the proc structure. Now that the UPAGES (pcb and kernel stack) are out of user space, make rfork(..RFMEM..) do what was intended by sharing the vmspace entirely via reference counting rather than simply inheriting the mappings.
1997-04-07 07:16:06 +00:00
* cpu_fork will copy and update the pcb, set up the kernel stack,
* and make the child ready to run.
1994-05-24 10:09:53 +00:00
*/
cpu_fork(td, p2, td2, flags);
return (0);
1994-05-24 10:09:53 +00:00
}
/*
* Called after process has been wait(2)'ed apon and is being reaped.
* The idea is to reclaim resources that we could not reclaim while
* the process was still executing.
*/
void
vm_waitproc(p)
struct proc *p;
{
vmspace_exitfree(p); /* and clean-out the vmspace */
}
1994-05-24 10:09:53 +00:00
void
faultin(p)
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
struct proc *p;
{
#ifdef NO_SWAPPING
PROC_LOCK_ASSERT(p, MA_OWNED);
if ((p->p_flag & P_INMEM) == 0)
panic("faultin: proc swapped out with NO_SWAPPING!");
#else /* !NO_SWAPPING */
struct thread *td;
PROC_LOCK_ASSERT(p, MA_OWNED);
/*
* If another process is swapping in this process,
* just wait until it finishes.
*/
if (p->p_flag & P_SWAPPINGIN) {
while (p->p_flag & P_SWAPPINGIN)
msleep(&p->p_flag, &p->p_mtx, PVM, "faultin", 0);
return;
}
if ((p->p_flag & P_INMEM) == 0) {
/*
* Don't let another thread swap process p out while we are
* busy swapping it in.
*/
++p->p_lock;
p->p_flag |= P_SWAPPINGIN;
PROC_UNLOCK(p);
/*
* We hold no lock here because the list of threads
* can not change while all threads in the process are
* swapped out.
*/
FOREACH_THREAD_IN_PROC(p, td)
vm_thread_swapin(td);
PROC_LOCK(p);
swapclear(p);
p->p_swtick = ticks;
wakeup(&p->p_flag);
/* Allow other threads to swap p out now. */
--p->p_lock;
}
#endif /* NO_SWAPPING */
}
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
/*
* This swapin algorithm attempts to swap-in processes only if there
* is enough space for them. Of course, if a process waits for a long
* time, it will be swapped in anyway.
*
* Giant is held on entry.
1994-05-24 10:09:53 +00:00
*/
void
swapper(void)
1994-05-24 10:09:53 +00:00
{
2001-07-04 19:00:13 +00:00
struct proc *p;
struct thread *td;
1994-05-24 10:09:53 +00:00
struct proc *pp;
int slptime;
int swtime;
1994-05-24 10:09:53 +00:00
int ppri;
int pri;
1994-05-24 10:09:53 +00:00
loop:
if (vm_page_count_min()) {
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
VM_WAIT;
goto loop;
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
}
1994-05-24 10:09:53 +00:00
pp = NULL;
ppri = INT_MIN;
sx_slock(&allproc_lock);
FOREACH_PROC_IN_SYSTEM(p) {
PROC_LOCK(p);
if (p->p_state == PRS_NEW ||
p->p_flag & (P_SWAPPINGOUT | P_SWAPPINGIN | P_INMEM)) {
PROC_UNLOCK(p);
continue;
}
swtime = (ticks - p->p_swtick) / hz;
FOREACH_THREAD_IN_PROC(p, td) {
/*
* An otherwise runnable thread of a process
* swapped out has only the TDI_SWAPPED bit set.
*
*/
thread_lock(td);
if (td->td_inhibitors == TDI_SWAPPED) {
slptime = (ticks - td->td_slptick) / hz;
pri = swtime + slptime;
if ((td->td_flags & TDF_SWAPINREQ) == 0)
pri -= p->p_nice * 8;
/*
* if this thread is higher priority
* and there is enough space, then select
* this process instead of the previous
* selection.
*/
if (pri > ppri) {
pp = p;
ppri = pri;
}
1994-05-24 10:09:53 +00:00
}
thread_unlock(td);
1994-05-24 10:09:53 +00:00
}
PROC_UNLOCK(p);
1994-05-24 10:09:53 +00:00
}
sx_sunlock(&allproc_lock);
1994-05-24 10:09:53 +00:00
/*
* Nothing to do, back to sleep.
1994-05-24 10:09:53 +00:00
*/
if ((p = pp) == NULL) {
tsleep(&proc0, PVM, "swapin", MAXSLP * hz / 2);
1994-05-24 10:09:53 +00:00
goto loop;
}
PROC_LOCK(p);
/*
* Another process may be bringing or may have already
* brought this process in while we traverse all threads.
* Or, this process may even be being swapped out again.
*/
if (p->p_flag & (P_INMEM | P_SWAPPINGOUT | P_SWAPPINGIN)) {
PROC_UNLOCK(p);
goto loop;
}
1994-05-24 10:09:53 +00:00
/*
* We would like to bring someone in. (only if there is space).
* [What checks the space? ]
1994-05-24 10:09:53 +00:00
*/
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
faultin(p);
PROC_UNLOCK(p);
1994-05-24 10:09:53 +00:00
goto loop;
}
If a thread that is swapped out is made runnable, then the setrunnable() routine wakes up proc0 so that proc0 can swap the thread back in. Historically, this has been done by waking up proc0 directly from setrunnable() itself via a wakeup(). When waking up a sleeping thread that was swapped out (the usual case when waking proc0 since only sleeping threads are eligible to be swapped out), this resulted in a bit of recursion (e.g. wakeup() -> setrunnable() -> wakeup()). With sleep queues having separate locks in 6.x and later, this caused a spin lock LOR (sleepq lock -> sched_lock/thread lock -> sleepq lock). An attempt was made to fix this in 7.0 by making the proc0 wakeup use the ithread mechanism for doing the wakeup. However, this required grabbing proc0's thread lock to perform the wakeup. If proc0 was asleep elsewhere in the kernel (e.g. waiting for disk I/O), then this degenerated into the same LOR since the thread lock would be some other sleepq lock. Fix this by deferring the wakeup of the swapper until after the sleepq lock held by the upper layer has been locked. The setrunnable() routine now returns a boolean value to indicate whether or not proc0 needs to be woken up. The end result is that consumers of the sleepq API such as *sleep/wakeup, condition variables, sx locks, and lockmgr, have to wakeup proc0 if they get a non-zero return value from sleepq_abort(), sleepq_broadcast(), or sleepq_signal(). Discussed with: jeff Glanced at by: sam Tested by: Jurgen Weber jurgen - ish com au MFC after: 2 weeks
2008-08-05 20:02:31 +00:00
void
kick_proc0(void)
{
If a thread that is swapped out is made runnable, then the setrunnable() routine wakes up proc0 so that proc0 can swap the thread back in. Historically, this has been done by waking up proc0 directly from setrunnable() itself via a wakeup(). When waking up a sleeping thread that was swapped out (the usual case when waking proc0 since only sleeping threads are eligible to be swapped out), this resulted in a bit of recursion (e.g. wakeup() -> setrunnable() -> wakeup()). With sleep queues having separate locks in 6.x and later, this caused a spin lock LOR (sleepq lock -> sched_lock/thread lock -> sleepq lock). An attempt was made to fix this in 7.0 by making the proc0 wakeup use the ithread mechanism for doing the wakeup. However, this required grabbing proc0's thread lock to perform the wakeup. If proc0 was asleep elsewhere in the kernel (e.g. waiting for disk I/O), then this degenerated into the same LOR since the thread lock would be some other sleepq lock. Fix this by deferring the wakeup of the swapper until after the sleepq lock held by the upper layer has been locked. The setrunnable() routine now returns a boolean value to indicate whether or not proc0 needs to be woken up. The end result is that consumers of the sleepq API such as *sleep/wakeup, condition variables, sx locks, and lockmgr, have to wakeup proc0 if they get a non-zero return value from sleepq_abort(), sleepq_broadcast(), or sleepq_signal(). Discussed with: jeff Glanced at by: sam Tested by: Jurgen Weber jurgen - ish com au MFC after: 2 weeks
2008-08-05 20:02:31 +00:00
wakeup(&proc0);
}
#ifndef NO_SWAPPING
/*
* Swap_idle_threshold1 is the guaranteed swapped in time for a process
*/
1998-02-09 06:11:36 +00:00
static int swap_idle_threshold1 = 2;
SYSCTL_INT(_vm, OID_AUTO, swap_idle_threshold1, CTLFLAG_RW,
&swap_idle_threshold1, 0, "Guaranteed swapped in time for a process");
/*
* Swap_idle_threshold2 is the time that a process can be idle before
* it will be swapped out, if idle swapping is enabled.
*/
1998-02-09 06:11:36 +00:00
static int swap_idle_threshold2 = 10;
SYSCTL_INT(_vm, OID_AUTO, swap_idle_threshold2, CTLFLAG_RW,
&swap_idle_threshold2, 0, "Time before a process will be swapped out");
1994-05-24 10:09:53 +00:00
/*
* First, if any processes have been sleeping or stopped for at least
* "swap_idle_threshold1" seconds, they are swapped out. If, however,
* no such processes exist, then the longest-sleeping or stopped
* process is swapped out. Finally, and only as a last resort, if
* there are no sleeping or stopped processes, the longest-resident
* process is swapped out.
1994-05-24 10:09:53 +00:00
*/
void
swapout_procs(action)
int action;
1994-05-24 10:09:53 +00:00
{
2001-07-04 19:00:13 +00:00
struct proc *p;
struct thread *td;
1994-05-24 10:09:53 +00:00
int didswap = 0;
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
retry:
sx_slock(&allproc_lock);
FOREACH_PROC_IN_SYSTEM(p) {
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
struct vmspace *vm;
int minslptime = 100000;
int slptime;
/*
* Watch out for a process in
* creation. It may have no
* address space or lock yet.
*/
if (p->p_state == PRS_NEW)
continue;
/*
* An aio daemon switches its
* address space while running.
* Perform a quick check whether
* a process has P_SYSTEM.
*/
if ((p->p_flag & P_SYSTEM) != 0)
continue;
/*
* Do not swapout a process that
* is waiting for VM data
* structures as there is a possible
* deadlock. Test this first as
* this may block.
*
* Lock the map until swapout
* finishes, or a thread of this
* process may attempt to alter
* the map.
*/
vm = vmspace_acquire_ref(p);
if (vm == NULL)
continue;
if (!vm_map_trylock(&vm->vm_map))
goto nextproc1;
PROC_LOCK(p);
if (p->p_lock != 0 ||
(p->p_flag & (P_STOPPED_SINGLE|P_TRACED|P_SYSTEM|P_WEXIT)
) != 0) {
goto nextproc;
}
/*
* only aiod changes vmspace, however it will be
* skipped because of the if statement above checking
* for P_SYSTEM
*/
if ((p->p_flag & (P_INMEM|P_SWAPPINGOUT|P_SWAPPINGIN)) != P_INMEM)
goto nextproc;
switch (p->p_state) {
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
default:
/* Don't swap out processes in any sort
* of 'special' state. */
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
case PRS_NORMAL:
/*
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
* do not swapout a realtime process
* Check all the thread groups..
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
*/
FOREACH_THREAD_IN_PROC(p, td) {
thread_lock(td);
if (PRI_IS_REALTIME(td->td_pri_class)) {
thread_unlock(td);
goto nextproc;
}
slptime = (ticks - td->td_slptick) / hz;
/*
* Guarantee swap_idle_threshold1
* time in memory.
*/
if (slptime < swap_idle_threshold1) {
thread_unlock(td);
goto nextproc;
}
/*
* Do not swapout a process if it is
* waiting on a critical event of some
* kind or there is a thread whose
* pageable memory may be accessed.
*
* This could be refined to support
* swapping out a thread.
*/
if (!thread_safetoswapout(td)) {
thread_unlock(td);
goto nextproc;
}
/*
* If the system is under memory stress,
* or if we are swapping
* idle processes >= swap_idle_threshold2,
* then swap the process out.
*/
if (((action & VM_SWAP_NORMAL) == 0) &&
(((action & VM_SWAP_IDLE) == 0) ||
(slptime < swap_idle_threshold2))) {
thread_unlock(td);
goto nextproc;
}
if (minslptime > slptime)
minslptime = slptime;
thread_unlock(td);
2000-12-02 03:29:33 +00:00
}
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
/*
* If the pageout daemon didn't free enough pages,
* or if this process is idle and the system is
* configured to swap proactively, swap it out.
*/
if ((action & VM_SWAP_NORMAL) ||
((action & VM_SWAP_IDLE) &&
(minslptime > swap_idle_threshold2))) {
if (swapout(p) == 0)
didswap++;
PROC_UNLOCK(p);
vm_map_unlock(&vm->vm_map);
vmspace_free(vm);
sx_sunlock(&allproc_lock);
goto retry;
}
1994-05-24 10:09:53 +00:00
}
nextproc:
PROC_UNLOCK(p);
vm_map_unlock(&vm->vm_map);
nextproc1:
vmspace_free(vm);
continue;
1994-05-24 10:09:53 +00:00
}
sx_sunlock(&allproc_lock);
/*
* If we swapped something out, and another process needed memory,
* then wakeup the sched process.
*/
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
if (didswap)
NOTE: libkvm, w, ps, 'top', and any other utility which depends on struct proc or any VM system structure will have to be rebuilt!!! Much needed overhaul of the VM system. Included in this first round of changes: 1) Improved pager interfaces: init, alloc, dealloc, getpages, putpages, haspage, and sync operations are supported. The haspage interface now provides information about clusterability. All pager routines now take struct vm_object's instead of "pagers". 2) Improved data structures. In the previous paradigm, there is constant confusion caused by pagers being both a data structure ("allocate a pager") and a collection of routines. The idea of a pager structure has escentially been eliminated. Objects now have types, and this type is used to index the appropriate pager. In most cases, items in the pager structure were duplicated in the object data structure and thus were unnecessary. In the few cases that remained, a un_pager structure union was created in the object to contain these items. 3) Because of the cleanup of #1 & #2, a lot of unnecessary layering can now be removed. For instance, vm_object_enter(), vm_object_lookup(), vm_object_remove(), and the associated object hash list were some of the things that were removed. 4) simple_lock's removed. Discussion with several people reveals that the SMP locking primitives used in the VM system aren't likely the mechanism that we'll be adopting. Even if it were, the locking that was in the code was very inadequate and would have to be mostly re-done anyway. The locking in a uni-processor kernel was a no-op but went a long way toward making the code difficult to read and debug. 5) Places that attempted to kludge-up the fact that we don't have kernel thread support have been fixed to reflect the reality that we are really dealing with processes, not threads. The VM system didn't have complete thread support, so the comments and mis-named routines were just wrong. We now use tsleep and wakeup directly in the lock routines, for instance. 6) Where appropriate, the pagers have been improved, especially in the pager_alloc routines. Most of the pager_allocs have been rewritten and are now faster and easier to maintain. 7) The pagedaemon pageout clustering algorithm has been rewritten and now tries harder to output an even number of pages before and after the requested page. This is sort of the reverse of the ideal pagein algorithm and should provide better overall performance. 8) Unnecessary (incorrect) casts to caddr_t in calls to tsleep & wakeup have been removed. Some other unnecessary casts have also been removed. 9) Some almost useless debugging code removed. 10) Terminology of shadow objects vs. backing objects straightened out. The fact that the vm_object data structure escentially had this backwards really confused things. The use of "shadow" and "backing object" throughout the code is now internally consistent and correct in the Mach terminology. 11) Several minor bug fixes, including one in the vm daemon that caused 0 RSS objects to not get purged as intended. 12) A "default pager" has now been created which cleans up the transition of objects to the "swap" type. The previous checks throughout the code for swp->pg_data != NULL were really ugly. This change also provides the rudiments for future backing of "anonymous" memory by something other than the swap pager (via the vnode pager, for example), and it allows the decision about which of these pagers to use to be made dynamically (although will need some additional decision code to do this, of course). 13) (dyson) MAP_COPY has been deprecated and the corresponding "copy object" code has been removed. MAP_COPY was undocumented and non- standard. It was furthermore broken in several ways which caused its behavior to degrade to MAP_PRIVATE. Binaries that use MAP_COPY will continue to work correctly, but via the slightly different semantics of MAP_PRIVATE. 14) (dyson) Sharing maps have been removed. It's marginal usefulness in a threads design can be worked around in other ways. Both #12 and #13 were done to simplify the code and improve readability and maintain- ability. (As were most all of these changes) TODO: 1) Rewrite most of the vnode pager to use VOP_GETPAGES/PUTPAGES. Doing this will reduce the vnode pager to a mere fraction of its current size. 2) Rewrite vm_fault and the swap/vnode pagers to use the clustering information provided by the new haspage pager interface. This will substantially reduce the overhead by eliminating a large number of VOP_BMAP() calls. The VOP_BMAP() filesystem interface should be improved to provide both a "behind" and "ahead" indication of contiguousness. 3) Implement the extended features of pager_haspage in swap_pager_haspage(). It currently just says 0 pages ahead/behind. 4) Re-implement the swap device (swstrategy) in a more elegant way, perhaps via a much more general mechanism that could also be used for disk striping of regular filesystems. 5) Do something to improve the architecture of vm_object_collapse(). The fact that it makes calls into the swap pager and knows too much about how the swap pager operates really bothers me. It also doesn't allow for collapsing of non-swap pager objects ("unnamed" objects backed by other pagers).
1995-07-13 08:48:48 +00:00
wakeup(&proc0);
1994-05-24 10:09:53 +00:00
}
static void
swapclear(p)
struct proc *p;
{
struct thread *td;
PROC_LOCK_ASSERT(p, MA_OWNED);
FOREACH_THREAD_IN_PROC(p, td) {
thread_lock(td);
td->td_flags |= TDF_INMEM;
td->td_flags &= ~TDF_SWAPINREQ;
TD_CLR_SWAPPED(td);
if (TD_CAN_RUN(td))
If a thread that is swapped out is made runnable, then the setrunnable() routine wakes up proc0 so that proc0 can swap the thread back in. Historically, this has been done by waking up proc0 directly from setrunnable() itself via a wakeup(). When waking up a sleeping thread that was swapped out (the usual case when waking proc0 since only sleeping threads are eligible to be swapped out), this resulted in a bit of recursion (e.g. wakeup() -> setrunnable() -> wakeup()). With sleep queues having separate locks in 6.x and later, this caused a spin lock LOR (sleepq lock -> sched_lock/thread lock -> sleepq lock). An attempt was made to fix this in 7.0 by making the proc0 wakeup use the ithread mechanism for doing the wakeup. However, this required grabbing proc0's thread lock to perform the wakeup. If proc0 was asleep elsewhere in the kernel (e.g. waiting for disk I/O), then this degenerated into the same LOR since the thread lock would be some other sleepq lock. Fix this by deferring the wakeup of the swapper until after the sleepq lock held by the upper layer has been locked. The setrunnable() routine now returns a boolean value to indicate whether or not proc0 needs to be woken up. The end result is that consumers of the sleepq API such as *sleep/wakeup, condition variables, sx locks, and lockmgr, have to wakeup proc0 if they get a non-zero return value from sleepq_abort(), sleepq_broadcast(), or sleepq_signal(). Discussed with: jeff Glanced at by: sam Tested by: Jurgen Weber jurgen - ish com au MFC after: 2 weeks
2008-08-05 20:02:31 +00:00
if (setrunnable(td)) {
#ifdef INVARIANTS
/*
* XXX: We just cleared TDI_SWAPPED
* above and set TDF_INMEM, so this
* should never happen.
*/
panic("not waking up swapper");
#endif
}
thread_unlock(td);
}
p->p_flag &= ~(P_SWAPPINGIN|P_SWAPPINGOUT);
p->p_flag |= P_INMEM;
}
static int
1994-05-24 10:09:53 +00:00
swapout(p)
2001-07-04 19:00:13 +00:00
struct proc *p;
1994-05-24 10:09:53 +00:00
{
struct thread *td;
1994-05-24 10:09:53 +00:00
PROC_LOCK_ASSERT(p, MA_OWNED);
#if defined(SWAP_DEBUG)
printf("swapping out %d\n", p->p_pid);
#endif
/*
* The states of this process and its threads may have changed
* by now. Assuming that there is only one pageout daemon thread,
* this process should still be in memory.
*/
KASSERT((p->p_flag & (P_INMEM|P_SWAPPINGOUT|P_SWAPPINGIN)) == P_INMEM,
("swapout: lost a swapout race?"));
1994-05-24 10:09:53 +00:00
/*
* remember the process resident count
1994-05-24 10:09:53 +00:00
*/
p->p_vmspace->vm_swrss = vmspace_resident_count(p->p_vmspace);
/*
* Check and mark all threads before we proceed.
*/
p->p_flag &= ~P_INMEM;
p->p_flag |= P_SWAPPINGOUT;
FOREACH_THREAD_IN_PROC(p, td) {
thread_lock(td);
if (!thread_safetoswapout(td)) {
thread_unlock(td);
swapclear(p);
return (EBUSY);
}
td->td_flags &= ~TDF_INMEM;
TD_SET_SWAPPED(td);
thread_unlock(td);
}
td = FIRST_THREAD_IN_PROC(p);
++td->td_ru.ru_nswap;
PROC_UNLOCK(p);
/*
* This list is stable because all threads are now prevented from
* running. The list is only modified in the context of a running
* thread in this process.
*/
FOREACH_THREAD_IN_PROC(p, td)
vm_thread_swapout(td);
PROC_LOCK(p);
p->p_flag &= ~P_SWAPPINGOUT;
p->p_swtick = ticks;
return (0);
1994-05-24 10:09:53 +00:00
}
#endif /* !NO_SWAPPING */