* 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
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* @(#)vm_glue.c 8.1 (Berkeley) 7/15/93
* Copyright (c) 1987, 1990 Carnegie-Mellon University.
* 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.
* CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
* CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
* FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
* 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.
#include <sys/resourcevar.h>
int avefree
= 0; /* XXX */
unsigned maxdmap
= MAXDSIZ
; /* XXX */
int readbuffers
= 0; /* XXX allow kgdb to read kernel buffer pool */
vm_offset_t saddr
, eaddr
;
vm_prot_t prot
= rw
== B_READ
? VM_PROT_READ
: VM_PROT_WRITE
;
saddr
= trunc_page(addr
);
eaddr
= round_page(addr
+len
);
rv
= vm_map_check_protection(kernel_map
, saddr
, eaddr
, prot
);
* XXX there are still some things (e.g. the buffer cache) that
* are managed behind the VM system's back so even though an
* address is accessible in the mind of the VM system, there may
* not be physical pages where the VM thinks there is. This can
* lead to bogus allocation of pages in the kernel address space
* or worse, inconsistencies at the pmap level. We only worry
* about the buffer cache for now.
if (!readbuffers
&& rv
&& (eaddr
> (vm_offset_t
)buffers
&&
saddr
< (vm_offset_t
)buffers
+ MAXBSIZE
* nbuf
))
vm_prot_t prot
= rw
== B_READ
? VM_PROT_READ
: VM_PROT_WRITE
;
rv
= vm_map_check_protection(&curproc
->p_vmspace
->vm_map
,
trunc_page(addr
), round_page(addr
+len
), prot
);
* Change protections on kernel pages from addr to addr+len
* (presumably so debugger can plant a breakpoint).
* We force the protection change at the pmap level. If we were
* to use vm_map_protect a change to allow writing would be lazily-
* applied meaning we would still take a protection fault, something
* we really don't want to do. It would also fragment the kernel
* map unnecessarily. We cannot use pmap_protect since it also won't
* enforce a write-enable request. Using pmap_enter is the only way
* we can ensure the change takes place properly.
vm_offset_t pa
, sva
, eva
;
prot
= rw
== B_READ
? VM_PROT_READ
: VM_PROT_READ
|VM_PROT_WRITE
;
eva
= round_page(addr
+ len
);
for (sva
= trunc_page(addr
); sva
< eva
; sva
+= PAGE_SIZE
) {
* Extract physical address for the page.
* We use a cheezy hack to differentiate physical
* page 0 from an invalid mapping, not that it
pa
= pmap_extract(kernel_pmap
, sva
|1);
panic("chgkprot: invalid page");
pmap_enter(kernel_pmap
, sva
, pa
&~1, prot
, TRUE
);
vm_map_pageable(&curproc
->p_vmspace
->vm_map
, trunc_page(addr
),
round_page(addr
+len
), FALSE
);
vsunlock(addr
, len
, dirtied
)
vm_map_pageable(&curproc
->p_vmspace
->vm_map
, trunc_page(addr
),
round_page(addr
+len
), TRUE
);
* 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
* NOTE: the kernel stack may be at a different location in the child
* process, and thus addresses of automatic variables may be invalid
* after cpu_fork returns in the child process. We do nothing here
* after cpu_fork returns.
register struct proc
*p1
, *p2
;
register struct user
*up
;
* avoid copying any of the parent's pagetables or other per-process
* objects that reside in the map by marking all of them non-inheritable
(void)vm_map_inherit(&p1
->p_vmspace
->vm_map
,
UPT_MIN_ADDRESS
-UPAGES
*NBPG
, VM_MAX_ADDRESS
, VM_INHERIT_NONE
);
p2
->p_vmspace
= vmspace_fork(p1
->p_vmspace
);
if (p1
->p_vmspace
->vm_shm
)
shmfork(p1
, p2
, isvfork
);
* Allocate a wired-down (for now) pcb and kernel stack for the process
addr
= kmem_alloc_pageable(kernel_map
, ctob(UPAGES
));
panic("vm_fork: no more kernel virtual memory");
vm_map_pageable(kernel_map
, addr
, addr
+ ctob(UPAGES
), FALSE
);
/* XXX somehow, on 386, ocassionally pageout removes active, wired down kstack,
and pagetables, WITHOUT going thru vm_page_unwire! Why this appears to work is
not yet clear, yet it does... */
addr
= kmem_alloc(kernel_map
, ctob(UPAGES
));
panic("vm_fork: no more kernel virtual memory");
up
= (struct user
*)addr
;
* p_stats and p_sigacts currently point at fields
* in the user struct but not at &u, instead at p_addr.
* Copy p_sigacts and parts of p_stats; zero the rest
* of p_stats (statistics).
p2
->p_stats
= &up
->u_stats
;
p2
->p_sigacts
= &up
->u_sigacts
;
up
->u_sigacts
= *p1
->p_sigacts
;
bzero(&up
->u_stats
.pstat_startzero
,
(unsigned) ((caddr_t
)&up
->u_stats
.pstat_endzero
-
(caddr_t
)&up
->u_stats
.pstat_startzero
));
bcopy(&p1
->p_stats
->pstat_startcopy
, &up
->u_stats
.pstat_startcopy
,
((caddr_t
)&up
->u_stats
.pstat_endcopy
-
(caddr_t
)&up
->u_stats
.pstat_startcopy
));
{ u_int addr
= UPT_MIN_ADDRESS
- UPAGES
*NBPG
; struct vm_map
*vp
;
vp
= &p2
->p_vmspace
->vm_map
;
(void)vm_deallocate(vp
, addr
, UPT_MAX_ADDRESS
- addr
);
(void)vm_allocate(vp
, &addr
, UPT_MAX_ADDRESS
- addr
, FALSE
);
(void)vm_map_inherit(vp
, addr
, UPT_MAX_ADDRESS
, VM_INHERIT_NONE
);
* cpu_fork will copy and update the kernel stack and pcb,
* and make the child ready to run. It marks the child
* so that it can return differently than the parent.
* It returns twice, once in the parent process and
return (cpu_fork(p1
, p2
));
* Set default limits for VM system.
* Called for proc 0, and then inherited by all others.
* Set up the initial limits on process VM.
* Set the maximum resident set size to be all
* of (reasonably) available memory. This causes
* any single, large process to start random page
* replacement once it fills memory.
p
->p_rlimit
[RLIMIT_STACK
].rlim_cur
= DFLSSIZ
;
p
->p_rlimit
[RLIMIT_STACK
].rlim_max
= MAXSSIZ
;
p
->p_rlimit
[RLIMIT_DATA
].rlim_cur
= DFLDSIZ
;
p
->p_rlimit
[RLIMIT_DATA
].rlim_max
= MAXDSIZ
;
p
->p_rlimit
[RLIMIT_RSS
].rlim_cur
= ptoa(cnt
.v_free_count
);
#include <vm/vm_pageout.h>
* 1. Attempt to swapin every swaped-out, runnable process in
* 2. If not enough memory, wake the pageout daemon and let it
sleep((caddr_t
)&proc0
, PVM
);
for (p
= (struct proc
*)allproc
; p
!= NULL
; p
= p
->p_nxt
) {
if (p
->p_stat
== SRUN
&& (p
->p_flag
& SLOAD
) == 0) {
pri
= p
->p_time
+ p
->p_slptime
- p
->p_nice
* 8;
if (swapdebug
& SDB_FOLLOW
)
printf("sched: running, procp %x pri %d\n", pp
, ppri
);
* Nothing to do, back to sleep
sleep((caddr_t
)&proc0
, PVM
);
* We would like to bring someone in.
* This part is really bogus cuz we could deadlock on memory
* despite our feeble check.
size
= round_page(ctob(UPAGES
));
addr
= (vm_offset_t
) p
->p_addr
;
if (cnt
.v_free_count
> atop(size
)) {
if (swapdebug
& SDB_SWAPIN
)
printf("swapin: pid %d(%s)@%x, pri %d free %d\n",
p
->p_pid
, p
->p_comm
, p
->p_addr
,
vm_map_pageable(kernel_map
, addr
, addr
+size
, FALSE
);
* Not enough memory, jab the pageout daemon and wait til the
if (swapdebug
& SDB_FOLLOW
)
printf("sched: no room for pid %d(%s), free %d\n",
p
->p_pid
, p
->p_comm
, cnt
.v_free_count
);
if (swapdebug
& SDB_FOLLOW
)
printf("sched: room again, free %d\n", cnt
.v_free_count
);
(((p)->p_flag & (SSYS|SLOAD|SKEEP|SWEXIT|SPHYSIO)) == SLOAD)
* Swapout is driven by the pageout daemon. Very simple, we find eligible
* procs and unwire their u-areas. We try to always "swap" at least one
* process in case we need the room for a swapin.
* If any procs have been sleeping/stopped for at least maxslp seconds,
* they are swapped. Else, we swap the longest-sleeping or stopped process,
* if any, otherwise the longest-resident process.
struct proc
*outp
, *outp2
;
for (p
= (struct proc
*)allproc
; p
!= NULL
; p
= p
->p_nxt
) {
if (p
->p_time
> outpri2
) {
if (p
->p_slptime
>= maxslp
) {
} else if (p
->p_slptime
> outpri
) {
* If we didn't get rid of any real duds, toss out the next most
* likely sleeping/stopped or running candidate. We only do this
* if we are real low on memory since we don't gain much by doing
cnt
.v_free_count
<= atop(round_page(ctob(UPAGES
)))) {
if (swapdebug
& SDB_SWAPOUT
)
printf("swapout_threads: no duds, try procp %x\n", p
);
if (swapdebug
& SDB_SWAPOUT
)
printf("swapout: pid %d(%s)@%x, stat %x pri %d free %d\n",
p
->p_pid
, p
->p_comm
, p
->p_addr
, p
->p_stat
,
p
->p_slptime
, cnt
.v_free_count
);
size
= round_page(ctob(UPAGES
));
addr
= (vm_offset_t
) p
->p_addr
;
#if defined(hp300) || defined(luna68k)
* Ugh! u-area is double mapped to a fixed address behind the
* back of the VM system and accesses are usually through that
* address rather than the per-process address. Hence reference
* and modify information are recorded at the fixed address and
* lost at context switch time. We assume the u-struct and
* kernel stack are always accessed/modified and force it to be so.
for (i
= 0; i
< UPAGES
; i
++) {
tmp
= *(long *)addr
; *(long *)addr
= tmp
;
addr
= (vm_offset_t
) p
->p_addr
;
* Be sure to save the floating point coprocessor state before
* paging out the u-struct.
extern struct proc
*machFPCurProcPtr
;
if (p
== machFPCurProcPtr
) {
machFPCurProcPtr
= (struct proc
*)0;
#ifndef i386 /* temporary measure till we find spontaineous unwire of kstack */
vm_map_pageable(kernel_map
, addr
, addr
+size
, TRUE
);
pmap_collect(vm_map_pmap(&p
->p_vmspace
->vm_map
));
* The rest of these routines fake thread handling
assert_wait(event
, ruptible
)
curproc
->p_thread
= event
;
sleep((caddr_t
)curproc
->p_thread
, PVM
);
thread_sleep(event
, lock
, ruptible
)
curproc
->p_thread
= event
;
sleep((caddr_t
)event
, PVM
);
#include <machine/stdarg.h> /* see subr_prf.c */
iprintf(const char *fmt
, ...)
iprintf(fmt
/* , va_alist */)
for (i
= indent
; i
>= 8; i
-= 8)