NetBSD/sys/uvm/uvm_mmap.c

1167 lines
25 KiB
C
Raw Normal View History

/* $NetBSD: uvm_mmap.c,v 1.60 2001/11/10 07:37:00 lukem Exp $ */
/*
* Copyright (c) 1997 Charles D. Cranor and Washington University.
2001-05-25 08:06:11 +04:00
* Copyright (c) 1991, 1993 The Regents of the University of California.
* Copyright (c) 1988 University of Utah.
2001-05-25 08:06:11 +04:00
*
* All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* the Systems Programming Group of the University of Utah Computer
* Science Department.
*
* 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.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the Charles D. Cranor,
2001-05-25 08:06:11 +04:00
* Washington University, 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
* SUCH DAMAGE.
*
* from: Utah $Hdr: vm_mmap.c 1.6 91/10/21$
* @(#)vm_mmap.c 8.5 (Berkeley) 5/19/94
1998-02-07 14:07:38 +03:00
* from: Id: uvm_mmap.c,v 1.1.2.14 1998/01/05 21:04:26 chuck Exp
*/
/*
* uvm_mmap.c: system call interface into VM system, plus kernel vm_mmap
* function.
*/
#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: uvm_mmap.c,v 1.60 2001/11/10 07:37:00 lukem Exp $");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/file.h>
#include <sys/filedesc.h>
#include <sys/resourcevar.h>
#include <sys/mman.h>
#include <sys/mount.h>
#include <sys/proc.h>
#include <sys/malloc.h>
#include <sys/vnode.h>
#include <sys/conf.h>
#include <sys/stat.h>
#include <miscfs/specfs/specdev.h>
#include <sys/syscallargs.h>
#include <uvm/uvm.h>
#include <uvm/uvm_device.h>
/*
* unimplemented VM system calls:
*/
/*
* sys_sbrk: sbrk system call.
*/
/* ARGSUSED */
1998-03-09 03:58:55 +03:00
int
sys_sbrk(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
#if 0
1998-03-09 03:58:55 +03:00
struct sys_sbrk_args /* {
syscallarg(intptr_t) incr;
1999-05-03 13:08:28 +04:00
} */ *uap = v;
#endif
1998-03-09 03:58:55 +03:00
return (ENOSYS);
}
/*
* sys_sstk: sstk system call.
*/
/* ARGSUSED */
1998-03-09 03:58:55 +03:00
int
sys_sstk(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
#if 0
1998-03-09 03:58:55 +03:00
struct sys_sstk_args /* {
1999-05-03 13:08:28 +04:00
syscallarg(int) incr;
} */ *uap = v;
#endif
1998-03-09 03:58:55 +03:00
return (ENOSYS);
}
/*
* sys_mincore: determine if pages are in core or not.
*/
/* ARGSUSED */
1998-03-09 03:58:55 +03:00
int
sys_mincore(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
1998-03-09 03:58:55 +03:00
struct sys_mincore_args /* {
syscallarg(void *) addr;
1999-05-03 13:08:28 +04:00
syscallarg(size_t) len;
syscallarg(char *) vec;
} */ *uap = v;
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
struct vm_page *pg;
char *vec, pgi;
struct uvm_object *uobj;
struct vm_amap *amap;
struct vm_anon *anon;
struct vm_map_entry *entry;
vaddr_t start, end, lim;
struct vm_map *map;
vsize_t len;
int error = 0, npgs;
map = &p->p_vmspace->vm_map;
start = (vaddr_t)SCARG(uap, addr);
len = SCARG(uap, len);
vec = SCARG(uap, vec);
if (start & PAGE_MASK)
return (EINVAL);
len = round_page(len);
end = start + len;
if (end <= start)
return (EINVAL);
npgs = len >> PAGE_SHIFT;
if (uvm_useracc(vec, npgs, B_WRITE) == FALSE)
return (EFAULT);
/*
* Lock down vec, so our returned status isn't outdated by
* storing the status byte for a page.
*/
uvm_vslock(p, vec, npgs, VM_PROT_WRITE);
vm_map_lock_read(map);
if (uvm_map_lookup_entry(map, start, &entry) == FALSE) {
error = ENOMEM;
goto out;
}
for (/* nothing */;
entry != &map->header && entry->start < end;
entry = entry->next) {
KASSERT(!UVM_ET_ISSUBMAP(entry));
KASSERT(start >= entry->start);
/* Make sure there are no holes. */
if (entry->end < end &&
(entry->next == &map->header ||
entry->next->start > entry->end)) {
error = ENOMEM;
goto out;
}
1998-03-09 03:58:55 +03:00
lim = end < entry->end ? end : entry->end;
/*
* Special case for objects with no "real" pages. Those
* are always considered resident (mapped devices).
*/
if (UVM_ET_ISOBJ(entry)) {
KASSERT(!UVM_OBJ_IS_KERN_OBJECT(entry->object.uvm_obj));
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
if (!UVM_OBJ_IS_VNODE(entry->object.uvm_obj)) {
for (/* nothing */; start < lim;
start += PAGE_SIZE, vec++)
subyte(vec, 1);
continue;
}
}
1999-07-11 00:40:23 +04:00
amap = entry->aref.ar_amap; /* top layer */
uobj = entry->object.uvm_obj; /* bottom layer */
if (amap != NULL)
amap_lock(amap);
if (uobj != NULL)
simple_lock(&uobj->vmobjlock);
for (/* nothing */; start < lim; start += PAGE_SIZE, vec++) {
pgi = 0;
if (amap != NULL) {
/* Check the top layer first. */
anon = amap_lookup(&entry->aref,
start - entry->start);
/* Don't need to lock anon here. */
if (anon != NULL && anon->u.an_page != NULL) {
/*
* Anon has the page for this entry
* offset.
*/
pgi = 1;
}
}
if (uobj != NULL && pgi == 0) {
/* Check the bottom layer. */
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
pg = uvm_pagelookup(uobj,
entry->offset + (start - entry->start));
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
if (pg != NULL) {
/*
* Object has the page for this entry
* offset.
*/
pgi = 1;
}
}
(void) subyte(vec, pgi);
}
if (uobj != NULL)
1999-07-01 22:40:39 +04:00
simple_unlock(&uobj->vmobjlock);
if (amap != NULL)
amap_unlock(amap);
}
out:
vm_map_unlock_read(map);
uvm_vsunlock(p, SCARG(uap, vec), npgs);
return (error);
}
/*
* sys_mmap: mmap system call.
*
* => file offest and address may not be page aligned
* - if MAP_FIXED, offset and address must have remainder mod PAGE_SIZE
* - if address isn't page aligned the mapping starts at trunc_page(addr)
* and the return value is adjusted up by the page offset.
*/
1998-03-09 03:58:55 +03:00
int
sys_mmap(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
2000-03-30 16:31:50 +04:00
struct sys_mmap_args /* {
1998-03-09 03:58:55 +03:00
syscallarg(caddr_t) addr;
syscallarg(size_t) len;
syscallarg(int) prot;
syscallarg(int) flags;
syscallarg(int) fd;
syscallarg(long) pad;
syscallarg(off_t) pos;
} */ *uap = v;
vaddr_t addr;
struct vattr va;
1998-03-09 03:58:55 +03:00
off_t pos;
vsize_t size, pageoff;
1998-03-09 03:58:55 +03:00
vm_prot_t prot, maxprot;
int flags, fd;
vaddr_t vm_min_address = VM_MIN_ADDRESS;
2000-03-30 16:31:50 +04:00
struct filedesc *fdp = p->p_fd;
struct file *fp;
1998-03-09 03:58:55 +03:00
struct vnode *vp;
void *handle;
1998-03-09 03:58:55 +03:00
int error;
/*
* first, extract syscall args from the uap.
*/
addr = (vaddr_t)SCARG(uap, addr);
size = (vsize_t)SCARG(uap, len);
1998-03-09 03:58:55 +03:00
prot = SCARG(uap, prot) & VM_PROT_ALL;
flags = SCARG(uap, flags);
fd = SCARG(uap, fd);
pos = SCARG(uap, pos);
/*
* Fixup the old deprecated MAP_COPY into MAP_PRIVATE, and
* validate the flags.
*/
if (flags & MAP_COPY)
flags = (flags & ~MAP_COPY) | MAP_PRIVATE;
if ((flags & (MAP_SHARED|MAP_PRIVATE)) == (MAP_SHARED|MAP_PRIVATE))
return (EINVAL);
1998-03-09 03:58:55 +03:00
/*
* align file position and save offset. adjust size.
*/
pageoff = (pos & PAGE_MASK);
pos -= pageoff;
size += pageoff; /* add offset */
size = (vsize_t)round_page(size); /* round up */
1998-03-09 03:58:55 +03:00
if ((ssize_t) size < 0)
return (EINVAL); /* don't allow wrap */
/*
2001-05-25 08:06:11 +04:00
* now check (MAP_FIXED) or get (!MAP_FIXED) the "addr"
1998-03-09 03:58:55 +03:00
*/
if (flags & MAP_FIXED) {
/* ensure address and file offset are aligned properly */
addr -= pageoff;
if (addr & PAGE_MASK)
return (EINVAL);
if (VM_MAXUSER_ADDRESS > 0 &&
(addr + size) > VM_MAXUSER_ADDRESS)
return (EINVAL);
if (vm_min_address > 0 && addr < vm_min_address)
return (EINVAL);
if (addr > addr + size)
return (EINVAL); /* no wrapping! */
} else {
/*
* not fixed: make sure we skip over the largest possible heap.
* we will refine our guess later (e.g. to account for VAC, etc)
*/
addr = MAX(addr, round_page((vaddr_t)p->p_vmspace->vm_daddr +
MAXDSIZ));
1998-03-09 03:58:55 +03:00
}
/*
* check for file mappings (i.e. not anonymous) and verify file.
*/
if ((flags & MAP_ANON) == 0) {
if ((fp = fd_getfile(fdp, fd)) == NULL)
return (EBADF);
1998-03-09 03:58:55 +03:00
if (fp->f_type != DTYPE_VNODE)
return (ENODEV); /* only mmap vnodes! */
1998-03-09 03:58:55 +03:00
vp = (struct vnode *)fp->f_data; /* convert to vnode */
if (vp->v_type != VREG && vp->v_type != VCHR &&
vp->v_type != VBLK)
return (ENODEV); /* only REG/CHR/BLK support mmap */
1998-03-09 03:58:55 +03:00
if (vp->v_type == VREG && (pos + size) < pos)
return (EOVERFLOW); /* no offset wrapping */
1998-03-09 03:58:55 +03:00
/* special case: catch SunOS style /dev/zero */
if (vp->v_type == VCHR && iszerodev(vp->v_rdev)) {
flags |= MAP_ANON;
goto is_anon;
}
/*
* Old programs may not select a specific sharing type, so
* default to an appropriate one.
*
* XXX: how does MAP_ANON fit in the picture?
*/
if ((flags & (MAP_SHARED|MAP_PRIVATE)) == 0) {
#if defined(DEBUG)
1998-03-09 03:58:55 +03:00
printf("WARNING: defaulted mmap() share type to "
"%s (pid %d comm %s)\n", vp->v_type == VCHR ?
"MAP_SHARED" : "MAP_PRIVATE", p->p_pid,
p->p_comm);
#endif
1998-03-09 03:58:55 +03:00
if (vp->v_type == VCHR)
flags |= MAP_SHARED; /* for a device */
else
flags |= MAP_PRIVATE; /* for a file */
}
2001-05-25 08:06:11 +04:00
/*
1998-03-09 03:58:55 +03:00
* MAP_PRIVATE device mappings don't make sense (and aren't
* supported anyway). However, some programs rely on this,
* so just change it to MAP_SHARED.
*/
if (vp->v_type == VCHR && (flags & MAP_PRIVATE) != 0) {
flags = (flags & ~MAP_PRIVATE) | MAP_SHARED;
}
1998-03-09 03:58:55 +03:00
/*
* now check protection
*/
maxprot = VM_PROT_EXECUTE;
1998-03-09 03:58:55 +03:00
/* check read access */
if (fp->f_flag & FREAD)
maxprot |= VM_PROT_READ;
else if (prot & PROT_READ)
return (EACCES);
/* check write access, shared case first */
1998-03-09 03:58:55 +03:00
if (flags & MAP_SHARED) {
/*
* if the file is writable, only add PROT_WRITE to
* maxprot if the file is not immutable, append-only.
* otherwise, if we have asked for PROT_WRITE, return
* EPERM.
*/
if (fp->f_flag & FWRITE) {
if ((error =
VOP_GETATTR(vp, &va, p->p_ucred, p)))
return (error);
if ((va.va_flags & (IMMUTABLE|APPEND)) == 0)
maxprot |= VM_PROT_WRITE;
else if (prot & PROT_WRITE)
return (EPERM);
}
1998-03-09 03:58:55 +03:00
else if (prot & PROT_WRITE)
return (EACCES);
} else {
/* MAP_PRIVATE mappings can always write to */
maxprot |= VM_PROT_WRITE;
}
handle = vp;
1998-03-09 03:58:55 +03:00
} else { /* MAP_ANON case */
/*
* XXX What do we do about (MAP_SHARED|MAP_PRIVATE) == 0?
*/
1998-03-09 03:58:55 +03:00
if (fd != -1)
return (EINVAL);
is_anon: /* label for SunOS style /dev/zero */
1998-03-09 03:58:55 +03:00
handle = NULL;
maxprot = VM_PROT_ALL;
pos = 0;
}
/*
* XXX (in)sanity check. We don't do proper datasize checking
* XXX for anonymous (or private writable) mmap(). However,
* XXX know that if we're trying to allocate more than the amount
* XXX remaining under our current data size limit, _that_ should
* XXX be disallowed.
*/
if ((flags & MAP_ANON) != 0 ||
((flags & MAP_PRIVATE) != 0 && (prot & PROT_WRITE) != 0)) {
if (size >
(p->p_rlimit[RLIMIT_DATA].rlim_cur -
ctob(p->p_vmspace->vm_dsize))) {
return (ENOMEM);
}
}
1998-03-09 03:58:55 +03:00
/*
* now let kernel internal function uvm_mmap do the work.
*/
1998-03-09 03:58:55 +03:00
error = uvm_mmap(&p->p_vmspace->vm_map, &addr, size, prot, maxprot,
flags, handle, pos, p->p_rlimit[RLIMIT_MEMLOCK].rlim_cur);
1998-03-09 03:58:55 +03:00
if (error == 0)
/* remember to add offset */
*retval = (register_t)(addr + pageoff);
1998-03-09 03:58:55 +03:00
return (error);
}
/*
* sys___msync13: the msync system call (a front-end for flush)
*/
1998-03-09 03:58:55 +03:00
int
sys___msync13(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
1998-03-09 03:58:55 +03:00
struct sys___msync13_args /* {
syscallarg(caddr_t) addr;
syscallarg(size_t) len;
syscallarg(int) flags;
} */ *uap = v;
vaddr_t addr;
vsize_t size, pageoff;
struct vm_map *map;
int error, rv, flags, uvmflags;
1998-03-09 03:58:55 +03:00
/*
* extract syscall args from the uap
*/
addr = (vaddr_t)SCARG(uap, addr);
size = (vsize_t)SCARG(uap, len);
1998-03-09 03:58:55 +03:00
flags = SCARG(uap, flags);
/* sanity check flags */
if ((flags & ~(MS_ASYNC | MS_SYNC | MS_INVALIDATE)) != 0 ||
(flags & (MS_ASYNC | MS_SYNC | MS_INVALIDATE)) == 0 ||
(flags & (MS_ASYNC | MS_SYNC)) == (MS_ASYNC | MS_SYNC))
return (EINVAL);
1998-03-09 03:58:55 +03:00
if ((flags & (MS_ASYNC | MS_SYNC)) == 0)
flags |= MS_SYNC;
1998-03-09 03:58:55 +03:00
/*
* align the address to a page boundary and adjust the size accordingly.
1998-03-09 03:58:55 +03:00
*/
pageoff = (addr & PAGE_MASK);
addr -= pageoff;
size += pageoff;
size = (vsize_t)round_page(size);
1998-03-09 03:58:55 +03:00
/* disallow wrap-around. */
if (addr + size < addr)
return (EINVAL);
/*
* get map
*/
map = &p->p_vmspace->vm_map;
/*
* XXXCDC: do we really need this semantic?
*
* XXX Gak! If size is zero we are supposed to sync "all modified
* pages with the region containing addr". Unfortunately, we
* don't really keep track of individual mmaps so we approximate
* by flushing the range of the map entry containing addr.
* This can be incorrect if the region splits or is coalesced
* with a neighbor.
*/
1998-03-09 03:58:55 +03:00
if (size == 0) {
struct vm_map_entry *entry;
2001-05-25 08:06:11 +04:00
1998-03-09 03:58:55 +03:00
vm_map_lock_read(map);
rv = uvm_map_lookup_entry(map, addr, &entry);
if (rv == TRUE) {
addr = entry->start;
size = entry->end - entry->start;
}
vm_map_unlock_read(map);
if (rv == FALSE)
return (EINVAL);
}
/*
* translate MS_ flags into PGO_ flags
*/
uvmflags = PGO_CLEANIT;
if (flags & MS_INVALIDATE)
uvmflags |= PGO_FREE;
1998-03-09 03:58:55 +03:00
if (flags & MS_SYNC)
uvmflags |= PGO_SYNCIO;
else
uvmflags |= PGO_SYNCIO; /* XXXCDC: force sync for now! */
error = uvm_map_clean(map, addr, addr+size, uvmflags);
return error;
}
/*
* sys_munmap: unmap a users memory
*/
1998-03-09 03:58:55 +03:00
int
sys_munmap(p, v, retval)
2000-03-30 16:31:50 +04:00
struct proc *p;
1998-03-09 03:58:55 +03:00
void *v;
register_t *retval;
{
2000-03-30 16:31:50 +04:00
struct sys_munmap_args /* {
1998-03-09 03:58:55 +03:00
syscallarg(caddr_t) addr;
syscallarg(size_t) len;
} */ *uap = v;
vaddr_t addr;
vsize_t size, pageoff;
struct vm_map *map;
vaddr_t vm_min_address = VM_MIN_ADDRESS;
1998-03-09 03:58:55 +03:00
struct vm_map_entry *dead_entries;
/*
* get syscall args.
1998-03-09 03:58:55 +03:00
*/
addr = (vaddr_t)SCARG(uap, addr);
size = (vsize_t)SCARG(uap, len);
2001-05-25 08:06:11 +04:00
1998-03-09 03:58:55 +03:00
/*
* align the address to a page boundary and adjust the size accordingly.
1998-03-09 03:58:55 +03:00
*/
pageoff = (addr & PAGE_MASK);
addr -= pageoff;
size += pageoff;
size = (vsize_t)round_page(size);
1998-03-09 03:58:55 +03:00
if ((int)size < 0)
return (EINVAL);
if (size == 0)
return (0);
/*
* Check for illegal addresses. Watch out for address wrap...
* Note that VM_*_ADDRESS are not constants due to casts (argh).
*/
if (VM_MAXUSER_ADDRESS > 0 && addr + size > VM_MAXUSER_ADDRESS)
return (EINVAL);
if (vm_min_address > 0 && addr < vm_min_address)
return (EINVAL);
if (addr > addr + size)
return (EINVAL);
map = &p->p_vmspace->vm_map;
/*
2001-05-25 08:06:11 +04:00
* interesting system call semantic: make sure entire range is
1998-03-09 03:58:55 +03:00
* allocated before allowing an unmap.
*/
vm_map_lock(map);
1998-03-09 03:58:55 +03:00
if (!uvm_map_checkprot(map, addr, addr + size, VM_PROT_NONE)) {
vm_map_unlock(map);
return (EINVAL);
}
uvm_unmap_remove(map, addr, addr + size, &dead_entries);
vm_map_unlock(map);
1998-03-09 03:58:55 +03:00
if (dead_entries != NULL)
uvm_unmap_detach(dead_entries, 0);
return (0);
}
/*
* sys_mprotect: the mprotect system call
*/
1998-03-09 03:58:55 +03:00
int
sys_mprotect(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
1998-03-09 03:58:55 +03:00
struct sys_mprotect_args /* {
syscallarg(caddr_t) addr;
syscallarg(int) len;
syscallarg(int) prot;
} */ *uap = v;
vaddr_t addr;
vsize_t size, pageoff;
1998-03-09 03:58:55 +03:00
vm_prot_t prot;
int error;
1998-03-09 03:58:55 +03:00
/*
* extract syscall args from uap
*/
addr = (vaddr_t)SCARG(uap, addr);
size = (vsize_t)SCARG(uap, len);
1998-03-09 03:58:55 +03:00
prot = SCARG(uap, prot) & VM_PROT_ALL;
/*
* align the address to a page boundary and adjust the size accordingly.
1998-03-09 03:58:55 +03:00
*/
1998-03-09 03:58:55 +03:00
pageoff = (addr & PAGE_MASK);
addr -= pageoff;
size += pageoff;
size = (vsize_t)round_page(size);
1998-03-09 03:58:55 +03:00
if ((int)size < 0)
return (EINVAL);
error = uvm_map_protect(&p->p_vmspace->vm_map, addr, addr + size, prot,
FALSE);
return error;
}
/*
* sys_minherit: the minherit system call
*/
1998-03-09 03:58:55 +03:00
int
sys_minherit(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
1998-03-09 03:58:55 +03:00
struct sys_minherit_args /* {
syscallarg(caddr_t) addr;
syscallarg(int) len;
syscallarg(int) inherit;
} */ *uap = v;
vaddr_t addr;
vsize_t size, pageoff;
2000-03-30 16:31:50 +04:00
vm_inherit_t inherit;
int error;
2001-05-25 08:06:11 +04:00
addr = (vaddr_t)SCARG(uap, addr);
size = (vsize_t)SCARG(uap, len);
1998-03-09 03:58:55 +03:00
inherit = SCARG(uap, inherit);
1998-03-09 03:58:55 +03:00
/*
* align the address to a page boundary and adjust the size accordingly.
1998-03-09 03:58:55 +03:00
*/
pageoff = (addr & PAGE_MASK);
addr -= pageoff;
size += pageoff;
size = (vsize_t)round_page(size);
1998-03-09 03:58:55 +03:00
if ((int)size < 0)
return (EINVAL);
error = uvm_map_inherit(&p->p_vmspace->vm_map, addr, addr + size,
inherit);
return error;
}
/*
* sys_madvise: give advice about memory usage.
*/
/* ARGSUSED */
int
sys_madvise(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
struct sys_madvise_args /* {
syscallarg(caddr_t) addr;
syscallarg(size_t) len;
syscallarg(int) behav;
} */ *uap = v;
vaddr_t addr;
vsize_t size, pageoff;
int advice, error;
2001-05-25 08:06:11 +04:00
addr = (vaddr_t)SCARG(uap, addr);
size = (vsize_t)SCARG(uap, len);
advice = SCARG(uap, behav);
/*
* align the address to a page boundary, and adjust the size accordingly
*/
pageoff = (addr & PAGE_MASK);
addr -= pageoff;
size += pageoff;
size = (vsize_t)round_page(size);
if ((ssize_t)size <= 0)
return (EINVAL);
switch (advice) {
case MADV_NORMAL:
case MADV_RANDOM:
case MADV_SEQUENTIAL:
error = uvm_map_advice(&p->p_vmspace->vm_map, addr, addr + size,
advice);
break;
case MADV_WILLNEED:
/*
* Activate all these pages, pre-faulting them in if
* necessary.
*/
/*
* XXX IMPLEMENT ME.
* Should invent a "weak" mode for uvm_fault()
* which would only do the PGO_LOCKED pgo_get().
*/
return (0);
case MADV_DONTNEED:
/*
* Deactivate all these pages. We don't need them
* any more. We don't, however, toss the data in
* the pages.
*/
error = uvm_map_clean(&p->p_vmspace->vm_map, addr, addr + size,
PGO_DEACTIVATE);
break;
case MADV_FREE:
/*
* These pages contain no valid data, and may be
2000-11-25 02:30:01 +03:00
* garbage-collected. Toss all resources, including
1999-07-08 04:52:45 +04:00
* any swap space in use.
*/
error = uvm_map_clean(&p->p_vmspace->vm_map, addr, addr + size,
PGO_FREE);
break;
case MADV_SPACEAVAIL:
/*
* XXXMRG What is this? I think it's:
*
* Ensure that we have allocated backing-store
* for these pages.
*
* This is going to require changes to the page daemon,
* as it will free swap space allocated to pages in core.
* There's also what to do for device/file/anonymous memory.
*/
return (EINVAL);
default:
return (EINVAL);
}
return error;
}
/*
* sys_mlock: memory lock
*/
1998-03-09 03:58:55 +03:00
int
sys_mlock(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
1998-03-09 03:58:55 +03:00
struct sys_mlock_args /* {
syscallarg(const void *) addr;
1998-03-09 03:58:55 +03:00
syscallarg(size_t) len;
} */ *uap = v;
vaddr_t addr;
vsize_t size, pageoff;
1998-03-09 03:58:55 +03:00
int error;
/*
* extract syscall args from uap
*/
addr = (vaddr_t)SCARG(uap, addr);
size = (vsize_t)SCARG(uap, len);
1998-03-09 03:58:55 +03:00
/*
* align the address to a page boundary and adjust the size accordingly
*/
1998-03-09 03:58:55 +03:00
pageoff = (addr & PAGE_MASK);
addr -= pageoff;
size += pageoff;
size = (vsize_t)round_page(size);
2001-05-25 08:06:11 +04:00
1998-03-09 03:58:55 +03:00
/* disallow wrap-around. */
if (addr + size < addr)
1998-03-09 03:58:55 +03:00
return (EINVAL);
if (atop(size) + uvmexp.wired > uvmexp.wiredmax)
return (EAGAIN);
#ifdef pmap_wired_count
1998-03-09 03:58:55 +03:00
if (size + ptoa(pmap_wired_count(vm_map_pmap(&p->p_vmspace->vm_map))) >
p->p_rlimit[RLIMIT_MEMLOCK].rlim_cur)
return (EAGAIN);
#else
1998-03-09 03:58:55 +03:00
if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
return (error);
#endif
error = uvm_map_pageable(&p->p_vmspace->vm_map, addr, addr+size, FALSE,
0);
return error;
}
/*
* sys_munlock: unlock wired pages
*/
1998-03-09 03:58:55 +03:00
int
sys_munlock(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
1998-03-09 03:58:55 +03:00
struct sys_munlock_args /* {
syscallarg(const void *) addr;
1998-03-09 03:58:55 +03:00
syscallarg(size_t) len;
} */ *uap = v;
vaddr_t addr;
vsize_t size, pageoff;
1998-03-09 03:58:55 +03:00
int error;
/*
* extract syscall args from uap
*/
addr = (vaddr_t)SCARG(uap, addr);
size = (vsize_t)SCARG(uap, len);
1998-03-09 03:58:55 +03:00
/*
* align the address to a page boundary, and adjust the size accordingly
*/
1998-03-09 03:58:55 +03:00
pageoff = (addr & PAGE_MASK);
addr -= pageoff;
size += pageoff;
size = (vsize_t)round_page(size);
1998-03-09 03:58:55 +03:00
/* disallow wrap-around. */
if (addr + size < addr)
1998-03-09 03:58:55 +03:00
return (EINVAL);
#ifndef pmap_wired_count
1998-03-09 03:58:55 +03:00
if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
return (error);
#endif
error = uvm_map_pageable(&p->p_vmspace->vm_map, addr, addr+size, TRUE,
0);
return error;
}
/*
* sys_mlockall: lock all pages mapped into an address space.
*/
int
sys_mlockall(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
struct sys_mlockall_args /* {
syscallarg(int) flags;
} */ *uap = v;
int error, flags;
flags = SCARG(uap, flags);
if (flags == 0 ||
(flags & ~(MCL_CURRENT|MCL_FUTURE)) != 0)
return (EINVAL);
#ifndef pmap_wired_count
if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
return (error);
#endif
error = uvm_map_pageable_all(&p->p_vmspace->vm_map, flags,
p->p_rlimit[RLIMIT_MEMLOCK].rlim_cur);
return (error);
}
/*
* sys_munlockall: unlock all pages mapped into an address space.
*/
int
sys_munlockall(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
(void) uvm_map_pageable_all(&p->p_vmspace->vm_map, 0, 0);
return (0);
}
/*
* uvm_mmap: internal version of mmap
*
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
* - used by sys_mmap and various framebuffers
* - handle is a vnode pointer or NULL for MAP_ANON
* - caller must page-align the file offset
*/
1998-03-09 03:58:55 +03:00
int
uvm_mmap(map, addr, size, prot, maxprot, flags, handle, foff, locklimit)
struct vm_map *map;
vaddr_t *addr;
vsize_t size;
1998-03-09 03:58:55 +03:00
vm_prot_t prot, maxprot;
int flags;
void *handle;
voff_t foff;
vsize_t locklimit;
{
1998-03-09 03:58:55 +03:00
struct uvm_object *uobj;
struct vnode *vp;
int error;
1998-03-09 03:58:55 +03:00
int advice = UVM_ADV_NORMAL;
uvm_flag_t uvmflag = 0;
/*
* check params
*/
if (size == 0)
return(0);
if (foff & PAGE_MASK)
return(EINVAL);
if ((prot & maxprot) != prot)
return(EINVAL);
/*
* for non-fixed mappings, round off the suggested address.
* for fixed mappings, check alignment and zap old mappings.
*/
if ((flags & MAP_FIXED) == 0) {
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
*addr = round_page(*addr);
1998-03-09 03:58:55 +03:00
} else {
if (*addr & PAGE_MASK)
return(EINVAL);
uvmflag |= UVM_FLAG_FIXED;
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
(void) uvm_unmap(map, *addr, *addr + size);
1998-03-09 03:58:55 +03:00
}
/*
* handle anon vs. non-anon mappings. for non-anon mappings attach
* to underlying vm object.
*/
if (flags & MAP_ANON) {
foff = UVM_UNKNOWN_OFFSET;
1998-03-09 03:58:55 +03:00
uobj = NULL;
if ((flags & MAP_SHARED) == 0)
/* XXX: defer amap create */
uvmflag |= UVM_FLAG_COPYONW;
else
/* shared: create amap now */
uvmflag |= UVM_FLAG_OVERLAY;
} else {
vp = (struct vnode *)handle;
/*
* Don't allow mmap for EXEC if the file system
* is mounted NOEXEC.
*/
if ((prot & PROT_EXEC) != 0 &&
(vp->v_mount->mnt_flag & MNT_NOEXEC) != 0)
return (EACCES);
1998-03-09 03:58:55 +03:00
if (vp->v_type != VCHR) {
error = VOP_MMAP(vp, 0, curproc->p_ucred, curproc);
if (error) {
return error;
}
uobj = uvn_attach((void *)vp, (flags & MAP_SHARED) ?
1998-03-09 03:58:55 +03:00
maxprot : (maxprot & ~VM_PROT_WRITE));
/* XXX for now, attach doesn't gain a ref */
VREF(vp);
/*
* If the vnode is being mapped with PROT_EXEC,
* then mark it as text.
*/
if (prot & PROT_EXEC)
vn_markexec(vp);
1998-03-09 03:58:55 +03:00
} else {
uobj = udv_attach((void *) &vp->v_rdev,
(flags & MAP_SHARED) ? maxprot :
(maxprot & ~VM_PROT_WRITE), foff, size);
/*
* XXX Some devices don't like to be mapped with
* XXX PROT_EXEC, but we don't really have a
* XXX better way of handling this, right now
*/
if (uobj == NULL && (prot & PROT_EXEC) == 0) {
maxprot &= ~VM_PROT_EXECUTE;
uobj = udv_attach((void *)&vp->v_rdev,
(flags & MAP_SHARED) ? maxprot :
(maxprot & ~VM_PROT_WRITE), foff, size);
}
1998-03-09 03:58:55 +03:00
advice = UVM_ADV_RANDOM;
}
if (uobj == NULL)
return((vp->v_type == VREG) ? ENOMEM : EINVAL);
1998-03-09 03:58:55 +03:00
if ((flags & MAP_SHARED) == 0)
uvmflag |= UVM_FLAG_COPYONW;
}
2001-05-25 08:06:11 +04:00
uvmflag = UVM_MAPFLAG(prot, maxprot,
(flags & MAP_SHARED) ? UVM_INH_SHARE : UVM_INH_COPY,
advice, uvmflag);
error = uvm_map(map, addr, size, uobj, foff, 0, uvmflag);
if (error) {
if (uobj)
uobj->pgops->pgo_detach(uobj);
return error;
}
1998-03-09 03:58:55 +03:00
/*
* POSIX 1003.1b -- if our address space was configured
* to lock all future mappings, wire the one we just made.
1998-03-09 03:58:55 +03:00
*/
if (prot == VM_PROT_NONE) {
1998-03-09 03:58:55 +03:00
/*
* No more work to do in this case.
*/
return (0);
}
vm_map_lock(map);
if (map->flags & VM_MAP_WIREFUTURE) {
if ((atop(size) + uvmexp.wired) > uvmexp.wiredmax
#ifdef pmap_wired_count
|| (locklimit != 0 && (size +
ptoa(pmap_wired_count(vm_map_pmap(map)))) >
locklimit)
#endif
) {
vm_map_unlock(map);
uvm_unmap(map, *addr, *addr + size);
return ENOMEM;
}
/*
* uvm_map_pageable() always returns the map unlocked.
*/
error = uvm_map_pageable(map, *addr, *addr + size,
FALSE, UVM_LK_ENTER);
if (error) {
uvm_unmap(map, *addr, *addr + size);
return error;
}
return (0);
}
vm_map_unlock(map);
return 0;
}