NetBSD/sys/uvm/uvm_mremap.c
rmind e225b7bd09 Welcome to 5.99.53! Merge rmind-uvmplock branch:
- Reorganize locking in UVM and provide extra serialisation for pmap(9).
  New lock order: [vmpage-owner-lock] -> pmap-lock.

- Simplify locking in some pmap(9) modules by removing P->V locking.

- Use lock object on vmobjlock (and thus vnode_t::v_interlock) to share
  the locks amongst UVM objects where necessary (tmpfs, layerfs, unionfs).

- Rewrite and optimise x86 TLB shootdown code, make it simpler and cleaner.
  Add TLBSTATS option for x86 to collect statistics about TLB shootdowns.

- Unify /dev/mem et al in MI code and provide required locking (removes
  kernel-lock on some ports).  Also, avoid cache-aliasing issues.

Thanks to Andrew Doran and Joerg Sonnenberger, as their initial patches
formed the core changes of this branch.
2011-06-12 03:35:36 +00:00

296 lines
7.5 KiB
C

/* $NetBSD: uvm_mremap.c,v 1.17 2011/06/12 03:36:03 rmind Exp $ */
/*-
* Copyright (c)2006,2007,2009 YAMAMOTO Takashi,
* All rights reserved.
*
* 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.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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.
*/
#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: uvm_mremap.c,v 1.17 2011/06/12 03:36:03 rmind Exp $");
#include <sys/param.h>
#include <sys/mman.h>
#include <sys/sched.h>
#include <sys/syscallargs.h>
#include <sys/proc.h>
#include <sys/atomic.h>
#include <uvm/uvm.h>
static int
uvm_mapent_extend(struct vm_map *map, vaddr_t endva, vsize_t size)
{
struct vm_map_entry *entry;
struct vm_map_entry *reserved_entry;
struct uvm_object *uobj;
int error = 0;
vm_map_lock(map);
if (!uvm_map_lookup_entry(map, endva, &reserved_entry)) {
error = ENOENT;
goto done;
}
if (reserved_entry->start != endva ||
reserved_entry->end != endva + size ||
reserved_entry->object.uvm_obj != NULL ||
reserved_entry->aref.ar_amap != NULL ||
reserved_entry->protection != VM_PROT_NONE) {
error = EINVAL;
goto done;
}
entry = reserved_entry->prev;
if (&map->header == entry || entry->end != endva) {
error = EINVAL;
goto done;
}
/*
* now, make reserved_entry compatible with entry, and then
* try to merge.
*/
uobj = entry->object.uvm_obj;
if (uobj) {
voff_t offset = entry->offset;
voff_t newoffset;
newoffset = offset + entry->end - entry->start;
if (newoffset <= offset) {
error = E2BIG; /* XXX */
goto done;
}
mutex_enter(uobj->vmobjlock);
KASSERT(uobj->uo_refs > 0);
atomic_inc_uint(&uobj->uo_refs);
mutex_exit(uobj->vmobjlock);
reserved_entry->object.uvm_obj = uobj;
reserved_entry->offset = newoffset;
}
reserved_entry->etype = entry->etype;
if (UVM_ET_ISCOPYONWRITE(entry)) {
reserved_entry->etype |= UVM_ET_NEEDSCOPY;
}
reserved_entry->flags &= ~UVM_MAP_NOMERGE;
reserved_entry->protection = entry->protection;
reserved_entry->max_protection = entry->max_protection;
reserved_entry->inheritance = entry->inheritance;
reserved_entry->advice = entry->advice;
reserved_entry->wired_count = 0; /* XXX should inherit? */
uvm_mapent_trymerge(map, reserved_entry, 0);
done:
vm_map_unlock(map);
return error;
}
/*
* uvm_mremap: move and/or resize existing mappings.
*/
int
uvm_mremap(struct vm_map *oldmap, vaddr_t oldva, vsize_t oldsize,
struct vm_map *newmap, vaddr_t *newvap, vsize_t newsize,
struct proc *newproc, int flags)
{
vaddr_t dstva;
vsize_t movesize;
vaddr_t newva;
int alignshift;
vaddr_t align = 0;
int error = 0;
const bool fixed = (flags & MAP_FIXED) != 0;
if (fixed) {
newva = *newvap;
} else {
newva = 0;
}
if ((oldva & PAGE_MASK) != 0 ||
(newva & PAGE_MASK) != 0 ||
(oldsize & PAGE_MASK) != 0 ||
(newsize & PAGE_MASK) != 0) {
return EINVAL;
}
/* XXX zero-size should be allowed? */
if (oldva + oldsize <= oldva || newva + newsize <= newva) {
return EINVAL;
}
/*
* Try to see if any requested alignment can even be attempted.
* Make sure we can express the alignment (asking for a >= 4GB
* alignment on an ILP32 architecure make no sense) and the
* alignment is at least for a page sized quanitiy. If the
* request was for a fixed mapping, make sure supplied address
* adheres to the request alignment.
*/
alignshift = (flags & MAP_ALIGNMENT_MASK) >> MAP_ALIGNMENT_SHIFT;
if (alignshift != 0) {
if (alignshift >= sizeof(vaddr_t) * NBBY)
return EINVAL;
align = 1L << alignshift;
if (align < PAGE_SIZE)
return EINVAL;
if (align >= vm_map_max(oldmap))
return ENOMEM;
if ((flags & MAP_FIXED) != 0) {
if ((*newvap & (align - 1)) != 0)
return EINVAL;
align = 0;
}
}
/*
* check the easy cases first.
*/
if ((!fixed || newva == oldva) && newmap == oldmap &&
(align == 0 || (oldva & (align - 1)) == 0)) {
vaddr_t va;
if (newsize == oldsize) {
newva = oldva;
goto done;
}
if (newsize < oldsize) {
uvm_unmap(oldmap, oldva + newsize, oldva + oldsize);
newva = oldva;
goto done;
}
va = oldva + oldsize;
if (uvm_map_reserve(oldmap, newsize - oldsize, 0, 0, &va,
UVM_FLAG_FIXED)) {
newva = oldva;
goto extend;
}
if (fixed) {
return ENOMEM;
}
}
/*
* we need to move mappings.
*/
if (!fixed) {
KASSERT(&newproc->p_vmspace->vm_map == newmap);
newva = newproc->p_emul->e_vm_default_addr(newproc,
(vaddr_t)newproc->p_vmspace->vm_daddr, newsize);
}
dstva = newva;
if (!uvm_map_reserve(newmap, newsize, oldva, align, &dstva,
fixed ? UVM_FLAG_FIXED : 0)) {
return ENOMEM;
}
KASSERT(!fixed || dstva == newva);
newva = dstva;
movesize = MIN(oldsize, newsize);
error = uvm_map_extract(oldmap, oldva, movesize, newmap, &dstva,
UVM_EXTRACT_RESERVED);
KASSERT(dstva == newva);
if (error != 0) {
/*
* undo uvm_map_reserve.
*/
uvm_unmap(newmap, newva, newva + newsize);
return error;
}
if (newsize > oldsize) {
extend:
error = uvm_mapent_extend(newmap, newva + oldsize,
newsize - oldsize);
if (error != 0) {
/*
* undo uvm_map_reserve and uvm_map_extract.
*/
if (newva == oldva && newmap == oldmap) {
uvm_unmap(newmap, newva + oldsize,
newva + newsize);
} else {
uvm_unmap(newmap, newva, newva + newsize);
}
return error;
}
}
/*
* now we won't fail.
* remove original entries unless we did in-place extend.
*/
if (oldva != newva || oldmap != newmap) {
uvm_unmap(oldmap, oldva, oldva + oldsize);
}
done:
*newvap = newva;
return 0;
}
/*
* sys_mremap: mremap system call.
*/
int
sys_mremap(struct lwp *l, const struct sys_mremap_args *uap, register_t *retval)
{
/* {
syscallarg(void *) old_address;
syscallarg(size_t) old_size;
syscallarg(void *) new_address;
syscallarg(size_t) new_size;
syscallarg(int) flags;
} */
struct proc *p;
struct vm_map *map;
vaddr_t oldva;
vaddr_t newva;
size_t oldsize;
size_t newsize;
int flags;
int error;
flags = SCARG(uap, flags);
oldva = (vaddr_t)SCARG(uap, old_address);
oldsize = (vsize_t)(SCARG(uap, old_size));
newva = (vaddr_t)SCARG(uap, new_address);
newsize = (vsize_t)(SCARG(uap, new_size));
if ((flags & ~(MAP_FIXED | MAP_ALIGNMENT_MASK)) != 0) {
error = EINVAL;
goto done;
}
oldsize = round_page(oldsize);
newsize = round_page(newsize);
p = l->l_proc;
map = &p->p_vmspace->vm_map;
error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p, flags);
done:
*retval = (error != 0) ? 0 : (register_t)newva;
return error;
}