int
vsunlock(
user_addr_t addr,
user_size_t len,
__unused int dirtied)
{
#if FIXME /* [ */
pmap_t pmap;
vm_page_t pg;
vm_map_offset_t vaddr;
ppnum_t paddr;
#endif /* FIXME ] */
kern_return_t kret;
vm_map_t map;
map = current_map();
#if FIXME /* [ */
if (dirtied) {
pmap = get_task_pmap(current_task());
for (vaddr = vm_map_trunc_page(addr, PAGE_MASK);
vaddr < vm_map_round_page(addr+len, PAGE_MASK);
vaddr += PAGE_SIZE) {
paddr = pmap_extract(pmap, vaddr);
pg = PHYS_TO_VM_PAGE(paddr);
vm_page_set_modified(pg);
}
}
#endif /* FIXME ] */
#ifdef lint
dirtied++;
#endif /* lint */
kret = vm_map_unwire(map,
vm_map_trunc_page(addr,
vm_map_page_mask(map)),
vm_map_round_page(addr+len,
vm_map_page_mask(map)),
FALSE);
switch (kret) {
case KERN_SUCCESS:
return (0);
case KERN_INVALID_ADDRESS:
case KERN_NO_SPACE:
return (ENOMEM);
case KERN_PROTECTION_FAILURE:
return (EACCES);
default:
return (EINVAL);
}
}
int
vslock(
user_addr_t addr,
user_size_t len)
{
kern_return_t kret;
vm_map_t map;
map = current_map();
kret = vm_map_wire(map,
vm_map_trunc_page(addr,
vm_map_page_mask(map)),
vm_map_round_page(addr+len,
vm_map_page_mask(map)),
VM_PROT_READ | VM_PROT_WRITE,
FALSE);
switch (kret) {
case KERN_SUCCESS:
return (0);
case KERN_INVALID_ADDRESS:
case KERN_NO_SPACE:
return (ENOMEM);
case KERN_PROTECTION_FAILURE:
return (EACCES);
default:
return (EINVAL);
}
}
int
useracc(
user_addr_t addr,
user_size_t len,
int prot)
{
return (vm_map_check_protection(
current_map(),
vm_map_trunc_page(addr), vm_map_round_page(addr+len),
prot == B_READ ? VM_PROT_READ : VM_PROT_WRITE));
}
/*
* On x86_64 systems, kernel extension text must remain within 2GB of the
* kernel's text segment. To ensure this happens, we snag 2GB of kernel VM
* as early as possible for kext allocations.
*/
void
kext_alloc_init(void)
{
#if __x86_64__
kern_return_t rval = 0;
kernel_segment_command_t *text = NULL;
mach_vm_offset_t text_end, text_start;
mach_vm_size_t text_size;
mach_vm_size_t kext_alloc_size;
/* Determine the start of the kernel's __TEXT segment and determine the
* lower bound of the allocated submap for kext allocations.
*/
text = getsegbyname(SEG_TEXT);
text_start = vm_map_trunc_page(text->vmaddr);
text_start &= ~((512ULL * 1024 * 1024 * 1024) - 1);
text_end = vm_map_round_page(text->vmaddr + text->vmsize);
text_size = text_end - text_start;
kext_alloc_base = KEXT_ALLOC_BASE(text_end);
kext_alloc_size = KEXT_ALLOC_SIZE(text_size);
kext_alloc_max = kext_alloc_base + kext_alloc_size;
/* Allocate the subblock of the kernel map */
rval = kmem_suballoc(kernel_map, (vm_offset_t *) &kext_alloc_base,
kext_alloc_size, /* pageable */ TRUE,
VM_FLAGS_FIXED|VM_FLAGS_OVERWRITE,
&g_kext_map);
if (rval != KERN_SUCCESS) {
panic("kext_alloc_init: kmem_suballoc failed 0x%x\n", rval);
}
if ((kext_alloc_base + kext_alloc_size) > kext_alloc_max) {
panic("kext_alloc_init: failed to get first 2GB\n");
}
if (kernel_map->min_offset > kext_alloc_base) {
kernel_map->min_offset = kext_alloc_base;
}
printf("kext submap [0x%llx - 0x%llx], kernel text [0x%llx - 0x%llx]\n",
kext_alloc_base, kext_alloc_max, text->vmaddr,
text->vmaddr + text->vmsize);
#else
g_kext_map = kernel_map;
kext_alloc_base = VM_MIN_KERNEL_ADDRESS;
kext_alloc_max = VM_MAX_KERNEL_ADDRESS;
#endif /* __x86_64__ */
}
int
useracc(
user_addr_t addr,
user_size_t len,
int prot)
{
vm_map_t map;
map = current_map();
return (vm_map_check_protection(
map,
vm_map_trunc_page(addr,
vm_map_page_mask(map)),
vm_map_round_page(addr+len,
vm_map_page_mask(map)),
prot == B_READ ? VM_PROT_READ : VM_PROT_WRITE));
}
int
mlock(__unused proc_t p, struct mlock_args *uap, __unused int32_t *retvalval)
{
vm_map_t user_map;
vm_map_offset_t addr;
vm_map_size_t size, pageoff;
kern_return_t result;
AUDIT_ARG(addr, uap->addr);
AUDIT_ARG(len, uap->len);
addr = (vm_map_offset_t) uap->addr;
size = (vm_map_size_t)uap->len;
/* disable wrap around */
if (addr + size < addr)
return (EINVAL);
if (size == 0)
return (0);
user_map = current_map();
pageoff = (addr & vm_map_page_mask(user_map));
addr -= pageoff;
size = vm_map_round_page(size+pageoff, vm_map_page_mask(user_map));
/* have to call vm_map_wire directly to pass "I don't know" protections */
result = vm_map_wire(user_map, addr, addr+size, VM_PROT_NONE | VM_PROT_MEMORY_TAG_MAKE(VM_KERN_MEMORY_MLOCK), TRUE);
if (result == KERN_RESOURCE_SHORTAGE)
return EAGAIN;
else if (result == KERN_PROTECTION_FAILURE)
return EACCES;
else if (result != KERN_SUCCESS)
return ENOMEM;
return 0; /* KERN_SUCCESS */
}
static
load_return_t
load_dylinker(
struct dylinker_command *lcp,
integer_t archbits,
vm_map_t map,
thread_t thread,
int depth,
load_result_t *result,
boolean_t is_64bit
)
{
char *name;
char *p;
struct vnode *vp = NULLVP; /* set by get_macho_vnode() */
struct mach_header header;
off_t file_offset = 0; /* set by get_macho_vnode() */
off_t macho_size = 0; /* set by get_macho_vnode() */
vm_map_t copy_map;
load_result_t myresult;
kern_return_t ret;
vm_map_copy_t tmp;
mach_vm_offset_t dyl_start, map_addr;
mach_vm_size_t dyl_length;
name = (char *)lcp + lcp->name.offset;
/*
* Check for a proper null terminated string.
*/
p = name;
do {
if (p >= (char *)lcp + lcp->cmdsize)
return(LOAD_BADMACHO);
} while (*p++);
ret = get_macho_vnode(name, archbits, &header, &file_offset, &macho_size, &vp);
if (ret)
return (ret);
myresult = load_result_null;
/*
* First try to map dyld in directly. This should work most of
* the time since there shouldn't normally be something already
* mapped to its address.
*/
ret = parse_machfile(vp, map, thread, &header, file_offset, macho_size,
depth, &myresult);
/*
* If it turned out something was in the way, then we'll take
* take this longer path to map dyld into a temporary map and
* copy it into destination map at a different address.
*/
if (ret == LOAD_NOSPACE) {
/*
* Load the Mach-O.
* Use a temporary map to do the work.
*/
copy_map = vm_map_create(pmap_create(vm_map_round_page(macho_size),
is_64bit),
get_map_min(map), get_map_max(map), TRUE);
if (VM_MAP_NULL == copy_map) {
ret = LOAD_RESOURCE;
goto out;
}
myresult = load_result_null;
ret = parse_machfile(vp, copy_map, thread, &header,
file_offset, macho_size,
depth, &myresult);
if (ret) {
vm_map_deallocate(copy_map);
goto out;
}
if (get_map_nentries(copy_map) > 0) {
dyl_start = mach_get_vm_start(copy_map);
dyl_length = mach_get_vm_end(copy_map) - dyl_start;
map_addr = dyl_start;
ret = mach_vm_allocate(map, &map_addr, dyl_length, VM_FLAGS_ANYWHERE);
if (ret != KERN_SUCCESS) {
vm_map_deallocate(copy_map);
ret = LOAD_NOSPACE;
goto out;
}
ret = vm_map_copyin(copy_map,
(vm_map_address_t)dyl_start,
(vm_map_size_t)dyl_length,
TRUE, &tmp);
if (ret != KERN_SUCCESS) {
(void) vm_map_remove(map,
vm_map_trunc_page(map_addr),
vm_map_round_page(map_addr + dyl_length),
VM_MAP_NO_FLAGS);
vm_map_deallocate(copy_map);
goto out;
}
ret = vm_map_copy_overwrite(map,
(vm_map_address_t)map_addr,
tmp, FALSE);
if (ret != KERN_SUCCESS) {
vm_map_copy_discard(tmp);
(void) vm_map_remove(map,
vm_map_trunc_page(map_addr),
vm_map_round_page(map_addr + dyl_length),
VM_MAP_NO_FLAGS);
vm_map_deallocate(copy_map);
goto out;
}
if (map_addr != dyl_start)
myresult.entry_point += (map_addr - dyl_start);
} else {
ret = LOAD_FAILURE;
}
vm_map_deallocate(copy_map);
}
if (ret == LOAD_SUCCESS) {
result->dynlinker = TRUE;
result->entry_point = myresult.entry_point;
(void)ubc_map(vp, PROT_READ | PROT_EXEC);
}
out:
vnode_put(vp);
return (ret);
}
/*
* XXX Internally, we use VM_PROT_* somewhat interchangeably, but the correct
* XXX usage is PROT_* from an interface perspective. Thus the values of
* XXX VM_PROT_* and PROT_* need to correspond.
*/
int
mmap(proc_t p, struct mmap_args *uap, user_addr_t *retval)
{
/*
* Map in special device (must be SHARED) or file
*/
struct fileproc *fp;
struct vnode *vp;
int flags;
int prot;
int err=0;
vm_map_t user_map;
kern_return_t result;
vm_map_offset_t user_addr;
vm_map_size_t user_size;
vm_object_offset_t pageoff;
vm_object_offset_t file_pos;
int alloc_flags = 0;
vm_tag_t tag = VM_KERN_MEMORY_NONE;
vm_map_kernel_flags_t vmk_flags = VM_MAP_KERNEL_FLAGS_NONE;
boolean_t docow;
vm_prot_t maxprot;
void *handle;
memory_object_t pager = MEMORY_OBJECT_NULL;
memory_object_control_t control;
int mapanon=0;
int fpref=0;
int error =0;
int fd = uap->fd;
int num_retries = 0;
/*
* Note that for UNIX03 conformance, there is additional parameter checking for
* mmap() system call in libsyscall prior to entering the kernel. The sanity
* checks and argument validation done in this function are not the only places
* one can get returned errnos.
*/
user_map = current_map();
user_addr = (vm_map_offset_t)uap->addr;
user_size = (vm_map_size_t) uap->len;
AUDIT_ARG(addr, user_addr);
AUDIT_ARG(len, user_size);
AUDIT_ARG(fd, uap->fd);
prot = (uap->prot & VM_PROT_ALL);
#if 3777787
/*
* Since the hardware currently does not support writing without
* read-before-write, or execution-without-read, if the request is
* for write or execute access, we must imply read access as well;
* otherwise programs expecting this to work will fail to operate.
*/
if (prot & (VM_PROT_EXECUTE | VM_PROT_WRITE))
prot |= VM_PROT_READ;
#endif /* radar 3777787 */
flags = uap->flags;
vp = NULLVP;
/*
* The vm code does not have prototypes & compiler doesn't do the'
* the right thing when you cast 64bit value and pass it in function
* call. So here it is.
*/
file_pos = (vm_object_offset_t)uap->pos;
/* make sure mapping fits into numeric range etc */
if (file_pos + user_size > (vm_object_offset_t)-PAGE_SIZE_64)
return (EINVAL);
/*
* Align the file position to a page boundary,
* and save its page offset component.
*/
pageoff = (file_pos & vm_map_page_mask(user_map));
file_pos -= (vm_object_offset_t)pageoff;
/* Adjust size for rounding (on both ends). */
user_size += pageoff; /* low end... */
user_size = vm_map_round_page(user_size,
vm_map_page_mask(user_map)); /* hi end */
if (flags & MAP_JIT) {
if ((flags & MAP_FIXED) ||
(flags & MAP_SHARED) ||
!(flags & MAP_ANON) ||
(flags & MAP_RESILIENT_CODESIGN) ||
(flags & MAP_RESILIENT_MEDIA)) {
return EINVAL;
}
}
if ((flags & MAP_RESILIENT_CODESIGN) ||
(flags & MAP_RESILIENT_MEDIA)) {
if ((flags & MAP_ANON) ||
(flags & MAP_JIT)) {
return EINVAL;
}
if (prot & (VM_PROT_WRITE | VM_PROT_EXECUTE)) {
return EPERM;
}
}
/*
* Check for illegal addresses. Watch out for address wrap... Note
* that VM_*_ADDRESS are not constants due to casts (argh).
*/
if (flags & MAP_FIXED) {
/*
* The specified address must have the same remainder
* as the file offset taken modulo PAGE_SIZE, so it
* should be aligned after adjustment by pageoff.
*/
user_addr -= pageoff;
if (user_addr & vm_map_page_mask(user_map))
return (EINVAL);
}
#ifdef notyet
/* DO not have apis to get this info, need to wait till then*/
/*
* XXX for non-fixed mappings where no hint is provided or
* the hint would fall in the potential heap space,
* place it after the end of the largest possible heap.
*
* There should really be a pmap call to determine a reasonable
* location.
*/
else if (addr < vm_map_round_page(p->p_vmspace->vm_daddr + MAXDSIZ,
vm_map_page_mask(user_map)))
addr = vm_map_round_page(p->p_vmspace->vm_daddr + MAXDSIZ,
vm_map_page_mask(user_map));
#endif
alloc_flags = 0;
if (flags & MAP_ANON) {
maxprot = VM_PROT_ALL;
#if CONFIG_MACF
/*
* Entitlement check.
*/
error = mac_proc_check_map_anon(p, user_addr, user_size, prot, flags, &maxprot);
if (error) {
return EINVAL;
}
#endif /* MAC */
/*
* Mapping blank space is trivial. Use positive fds as the alias
* value for memory tracking.
*/
if (fd != -1) {
/*
* Use "fd" to pass (some) Mach VM allocation flags,
* (see the VM_FLAGS_* definitions).
*/
alloc_flags = fd & (VM_FLAGS_ALIAS_MASK |
VM_FLAGS_SUPERPAGE_MASK |
VM_FLAGS_PURGABLE |
VM_FLAGS_4GB_CHUNK);
if (alloc_flags != fd) {
/* reject if there are any extra flags */
return EINVAL;
}
VM_GET_FLAGS_ALIAS(alloc_flags, tag);
alloc_flags &= ~VM_FLAGS_ALIAS_MASK;
}
handle = NULL;
file_pos = 0;
mapanon = 1;
} else {
struct vnode_attr va;
vfs_context_t ctx = vfs_context_current();
if (flags & MAP_JIT)
return EINVAL;
/*
* Mapping file, get fp for validation. Obtain vnode and make
* sure it is of appropriate type.
*/
err = fp_lookup(p, fd, &fp, 0);
if (err)
return(err);
fpref = 1;
switch (FILEGLOB_DTYPE(fp->f_fglob)) {
case DTYPE_PSXSHM:
uap->addr = (user_addr_t)user_addr;
uap->len = (user_size_t)user_size;
uap->prot = prot;
uap->flags = flags;
uap->pos = file_pos;
error = pshm_mmap(p, uap, retval, fp, (off_t)pageoff);
goto bad;
case DTYPE_VNODE:
break;
default:
error = EINVAL;
goto bad;
}
vp = (struct vnode *)fp->f_fglob->fg_data;
error = vnode_getwithref(vp);
if(error != 0)
goto bad;
if (vp->v_type != VREG && vp->v_type != VCHR) {
(void)vnode_put(vp);
error = EINVAL;
goto bad;
}
AUDIT_ARG(vnpath, vp, ARG_VNODE1);
/*
* POSIX: mmap needs to update access time for mapped files
*/
if ((vnode_vfsvisflags(vp) & MNT_NOATIME) == 0) {
VATTR_INIT(&va);
nanotime(&va.va_access_time);
VATTR_SET_ACTIVE(&va, va_access_time);
vnode_setattr(vp, &va, ctx);
}
/*
* XXX hack to handle use of /dev/zero to map anon memory (ala
* SunOS).
*/
if (vp->v_type == VCHR || vp->v_type == VSTR) {
(void)vnode_put(vp);
error = ENODEV;
goto bad;
} else {
/*
* Ensure that file and memory protections are
* compatible. Note that we only worry about
* writability if mapping is shared; in this case,
* current and max prot are dictated by the open file.
* XXX use the vnode instead? Problem is: what
* credentials do we use for determination? What if
* proc does a setuid?
*/
maxprot = VM_PROT_EXECUTE; /* ??? */
if (fp->f_fglob->fg_flag & FREAD)
maxprot |= VM_PROT_READ;
else if (prot & PROT_READ) {
(void)vnode_put(vp);
error = EACCES;
goto bad;
}
/*
* If we are sharing potential changes (either via
* MAP_SHARED or via the implicit sharing of character
* device mappings), and we are trying to get write
* permission although we opened it without asking
* for it, bail out.
*/
if ((flags & MAP_SHARED) != 0) {
if ((fp->f_fglob->fg_flag & FWRITE) != 0 &&
/*
* Do not allow writable mappings of
* swap files (see vm_swapfile_pager.c).
*/
!vnode_isswap(vp)) {
/*
* check for write access
*
* Note that we already made this check when granting FWRITE
* against the file, so it seems redundant here.
*/
error = vnode_authorize(vp, NULL, KAUTH_VNODE_CHECKIMMUTABLE, ctx);
/* if not granted for any reason, but we wanted it, bad */
if ((prot & PROT_WRITE) && (error != 0)) {
vnode_put(vp);
goto bad;
}
/* if writable, remember */
if (error == 0)
maxprot |= VM_PROT_WRITE;
} else if ((prot & PROT_WRITE) != 0) {
(void)vnode_put(vp);
error = EACCES;
goto bad;
}
} else
maxprot |= VM_PROT_WRITE;
handle = (void *)vp;
#if CONFIG_MACF
error = mac_file_check_mmap(vfs_context_ucred(ctx),
fp->f_fglob, prot, flags, file_pos, &maxprot);
if (error) {
(void)vnode_put(vp);
goto bad;
}
#endif /* MAC */
}
}
if (user_size == 0) {
if (!mapanon)
(void)vnode_put(vp);
error = 0;
goto bad;
}
/*
* We bend a little - round the start and end addresses
* to the nearest page boundary.
*/
user_size = vm_map_round_page(user_size,
vm_map_page_mask(user_map));
if (file_pos & vm_map_page_mask(user_map)) {
if (!mapanon)
(void)vnode_put(vp);
error = EINVAL;
goto bad;
}
if ((flags & MAP_FIXED) == 0) {
alloc_flags |= VM_FLAGS_ANYWHERE;
user_addr = vm_map_round_page(user_addr,
vm_map_page_mask(user_map));
} else {
if (user_addr != vm_map_trunc_page(user_addr,
vm_map_page_mask(user_map))) {
if (!mapanon)
(void)vnode_put(vp);
error = EINVAL;
goto bad;
}
/*
* mmap(MAP_FIXED) will replace any existing mappings in the
* specified range, if the new mapping is successful.
* If we just deallocate the specified address range here,
* another thread might jump in and allocate memory in that
* range before we get a chance to establish the new mapping,
* and we won't have a chance to restore the old mappings.
* So we use VM_FLAGS_OVERWRITE to let Mach VM know that it
* has to deallocate the existing mappings and establish the
* new ones atomically.
*/
alloc_flags |= VM_FLAGS_FIXED | VM_FLAGS_OVERWRITE;
}
if (flags & MAP_NOCACHE)
alloc_flags |= VM_FLAGS_NO_CACHE;
if (flags & MAP_JIT) {
vmk_flags.vmkf_map_jit = TRUE;
}
if (flags & MAP_RESILIENT_CODESIGN) {
alloc_flags |= VM_FLAGS_RESILIENT_CODESIGN;
}
/*
* Lookup/allocate object.
*/
if (handle == NULL) {
control = NULL;
#ifdef notyet
/* Hmm .. */
#if defined(VM_PROT_READ_IS_EXEC)
if (prot & VM_PROT_READ)
prot |= VM_PROT_EXECUTE;
if (maxprot & VM_PROT_READ)
maxprot |= VM_PROT_EXECUTE;
#endif
#endif
#if 3777787
if (prot & (VM_PROT_EXECUTE | VM_PROT_WRITE))
prot |= VM_PROT_READ;
if (maxprot & (VM_PROT_EXECUTE | VM_PROT_WRITE))
maxprot |= VM_PROT_READ;
#endif /* radar 3777787 */
map_anon_retry:
result = vm_map_enter_mem_object(user_map,
&user_addr, user_size,
0, alloc_flags, vmk_flags,
tag,
IPC_PORT_NULL, 0, FALSE,
prot, maxprot,
(flags & MAP_SHARED) ?
VM_INHERIT_SHARE :
VM_INHERIT_DEFAULT);
/* If a non-binding address was specified for this anonymous
* mapping, retry the mapping with a zero base
* in the event the mapping operation failed due to
* lack of space between the address and the map's maximum.
*/
if ((result == KERN_NO_SPACE) && ((flags & MAP_FIXED) == 0) && user_addr && (num_retries++ == 0)) {
user_addr = vm_map_page_size(user_map);
goto map_anon_retry;
}
} else {
if (vnode_isswap(vp)) {
/*
* Map swap files with a special pager
* that returns obfuscated contents.
*/
control = NULL;
pager = swapfile_pager_setup(vp);
if (pager != MEMORY_OBJECT_NULL) {
control = swapfile_pager_control(pager);
}
} else {
control = ubc_getobject(vp, UBC_FLAGS_NONE);
}
if (control == NULL) {
(void)vnode_put(vp);
error = ENOMEM;
goto bad;
}
/*
* Set credentials:
* FIXME: if we're writing the file we need a way to
* ensure that someone doesn't replace our R/W creds
* with ones that only work for read.
*/
ubc_setthreadcred(vp, p, current_thread());
docow = FALSE;
if ((flags & (MAP_ANON|MAP_SHARED)) == 0) {
docow = TRUE;
}
#ifdef notyet
/* Hmm .. */
#if defined(VM_PROT_READ_IS_EXEC)
if (prot & VM_PROT_READ)
prot |= VM_PROT_EXECUTE;
if (maxprot & VM_PROT_READ)
maxprot |= VM_PROT_EXECUTE;
#endif
#endif /* notyet */
#if 3777787
if (prot & (VM_PROT_EXECUTE | VM_PROT_WRITE))
prot |= VM_PROT_READ;
if (maxprot & (VM_PROT_EXECUTE | VM_PROT_WRITE))
maxprot |= VM_PROT_READ;
#endif /* radar 3777787 */
map_file_retry:
if ((flags & MAP_RESILIENT_CODESIGN) ||
(flags & MAP_RESILIENT_MEDIA)) {
if (prot & (VM_PROT_WRITE | VM_PROT_EXECUTE)) {
assert(!mapanon);
vnode_put(vp);
error = EPERM;
goto bad;
}
/* strictly limit access to "prot" */
maxprot &= prot;
}
vm_object_offset_t end_pos = 0;
if (os_add_overflow(user_size, file_pos, &end_pos)) {
vnode_put(vp);
error = EINVAL;
goto bad;
}
result = vm_map_enter_mem_object_control(user_map,
&user_addr, user_size,
0, alloc_flags, vmk_flags,
tag,
control, file_pos,
docow, prot, maxprot,
(flags & MAP_SHARED) ?
VM_INHERIT_SHARE :
VM_INHERIT_DEFAULT);
/* If a non-binding address was specified for this file backed
* mapping, retry the mapping with a zero base
* in the event the mapping operation failed due to
* lack of space between the address and the map's maximum.
*/
if ((result == KERN_NO_SPACE) && ((flags & MAP_FIXED) == 0) && user_addr && (num_retries++ == 0)) {
user_addr = vm_map_page_size(user_map);
goto map_file_retry;
}
}
if (!mapanon) {
(void)vnode_put(vp);
}
switch (result) {
case KERN_SUCCESS:
*retval = user_addr + pageoff;
error = 0;
break;
case KERN_INVALID_ADDRESS:
case KERN_NO_SPACE:
error = ENOMEM;
break;
case KERN_PROTECTION_FAILURE:
error = EACCES;
break;
default:
error = EINVAL;
break;
}
bad:
if (pager != MEMORY_OBJECT_NULL) {
/*
* Release the reference on the pager.
* If the mapping was successful, it now holds
* an extra reference.
*/
memory_object_deallocate(pager);
}
if (fpref)
fp_drop(p, fd, fp, 0);
KERNEL_DEBUG_CONSTANT((BSDDBG_CODE(DBG_BSD_SC_EXTENDED_INFO, SYS_mmap) | DBG_FUNC_NONE), fd, (uint32_t)(*retval), (uint32_t)user_size, error, 0);
#ifndef CONFIG_EMBEDDED
KERNEL_DEBUG_CONSTANT((BSDDBG_CODE(DBG_BSD_SC_EXTENDED_INFO2, SYS_mmap) | DBG_FUNC_NONE), (uint32_t)(*retval >> 32), (uint32_t)(user_size >> 32),
(uint32_t)(file_pos >> 32), (uint32_t)file_pos, 0);
#endif
return(error);
}
kern_return_t
kmem_alloc_contig(
vm_map_t map,
vm_offset_t *addrp,
vm_size_t size,
vm_offset_t mask,
ppnum_t max_pnum,
ppnum_t pnum_mask,
int flags)
{
vm_object_t object;
vm_object_offset_t offset;
vm_map_offset_t map_addr;
vm_map_offset_t map_mask;
vm_map_size_t map_size, i;
vm_map_entry_t entry;
vm_page_t m, pages;
kern_return_t kr;
if (map == VM_MAP_NULL || (flags & ~(KMA_KOBJECT | KMA_LOMEM | KMA_NOPAGEWAIT)))
return KERN_INVALID_ARGUMENT;
if (size == 0) {
*addrp = 0;
return KERN_INVALID_ARGUMENT;
}
map_size = vm_map_round_page(size);
map_mask = (vm_map_offset_t)mask;
/*
* Allocate a new object (if necessary) and the reference we
* will be donating to the map entry. We must do this before
* locking the map, or risk deadlock with the default pager.
*/
if ((flags & KMA_KOBJECT) != 0) {
object = kernel_object;
vm_object_reference(object);
} else {
object = vm_object_allocate(map_size);
}
kr = vm_map_find_space(map, &map_addr, map_size, map_mask, 0, &entry);
if (KERN_SUCCESS != kr) {
vm_object_deallocate(object);
return kr;
}
entry->object.vm_object = object;
entry->offset = offset = (object == kernel_object) ?
map_addr : 0;
/* Take an extra object ref in case the map entry gets deleted */
vm_object_reference(object);
vm_map_unlock(map);
kr = cpm_allocate(CAST_DOWN(vm_size_t, map_size), &pages, max_pnum, pnum_mask, FALSE, flags);
if (kr != KERN_SUCCESS) {
vm_map_remove(map, vm_map_trunc_page(map_addr),
vm_map_round_page(map_addr + map_size), 0);
vm_object_deallocate(object);
*addrp = 0;
return kr;
}
vm_object_lock(object);
for (i = 0; i < map_size; i += PAGE_SIZE) {
m = pages;
pages = NEXT_PAGE(m);
*(NEXT_PAGE_PTR(m)) = VM_PAGE_NULL;
m->busy = FALSE;
vm_page_insert(m, object, offset + i);
}
vm_object_unlock(object);
if ((kr = vm_map_wire(map, vm_map_trunc_page(map_addr),
vm_map_round_page(map_addr + map_size), VM_PROT_DEFAULT, FALSE))
!= KERN_SUCCESS) {
if (object == kernel_object) {
vm_object_lock(object);
vm_object_page_remove(object, offset, offset + map_size);
vm_object_unlock(object);
}
vm_map_remove(map, vm_map_trunc_page(map_addr),
vm_map_round_page(map_addr + map_size), 0);
vm_object_deallocate(object);
return kr;
}
vm_object_deallocate(object);
if (object == kernel_object)
vm_map_simplify(map, map_addr);
*addrp = (vm_offset_t) map_addr;
assert((vm_map_offset_t) *addrp == map_addr);
return KERN_SUCCESS;
}
kern_return_t
map_fd_funneled(
int fd,
vm_object_offset_t offset,
vm_offset_t *va,
boolean_t findspace,
vm_size_t size)
{
kern_return_t result;
struct fileproc *fp;
struct vnode *vp;
void * pager;
vm_offset_t map_addr=0;
vm_size_t map_size;
int err=0;
vm_map_t my_map;
proc_t p = current_proc();
struct vnode_attr vattr;
/*
* Find the inode; verify that it's a regular file.
*/
err = fp_lookup(p, fd, &fp, 0);
if (err)
return(err);
if (fp->f_fglob->fg_type != DTYPE_VNODE){
err = KERN_INVALID_ARGUMENT;
goto bad;
}
if (!(fp->f_fglob->fg_flag & FREAD)) {
err = KERN_PROTECTION_FAILURE;
goto bad;
}
vp = (struct vnode *)fp->f_fglob->fg_data;
err = vnode_getwithref(vp);
if(err != 0)
goto bad;
if (vp->v_type != VREG) {
(void)vnode_put(vp);
err = KERN_INVALID_ARGUMENT;
goto bad;
}
AUDIT_ARG(vnpath, vp, ARG_VNODE1);
/*
* POSIX: mmap needs to update access time for mapped files
*/
if ((vnode_vfsvisflags(vp) & MNT_NOATIME) == 0) {
VATTR_INIT(&vattr);
nanotime(&vattr.va_access_time);
VATTR_SET_ACTIVE(&vattr, va_access_time);
vnode_setattr(vp, &vattr, vfs_context_current());
}
if (offset & PAGE_MASK_64) {
printf("map_fd: file offset not page aligned(%d : %s)\n",p->p_pid, p->p_comm);
(void)vnode_put(vp);
err = KERN_INVALID_ARGUMENT;
goto bad;
}
map_size = round_page(size);
/*
* Allow user to map in a zero length file.
*/
if (size == 0) {
(void)vnode_put(vp);
err = KERN_SUCCESS;
goto bad;
}
/*
* Map in the file.
*/
pager = (void *)ubc_getpager(vp);
if (pager == NULL) {
(void)vnode_put(vp);
err = KERN_FAILURE;
goto bad;
}
my_map = current_map();
result = vm_map_64(
my_map,
&map_addr, map_size, (vm_offset_t)0,
VM_FLAGS_ANYWHERE, pager, offset, TRUE,
VM_PROT_DEFAULT, VM_PROT_ALL,
VM_INHERIT_DEFAULT);
if (result != KERN_SUCCESS) {
(void)vnode_put(vp);
err = result;
goto bad;
}
if (!findspace) {
vm_offset_t dst_addr;
vm_map_copy_t tmp;
if (copyin(CAST_USER_ADDR_T(va), &dst_addr, sizeof (dst_addr)) ||
trunc_page_32(dst_addr) != dst_addr) {
(void) vm_map_remove(
my_map,
map_addr, map_addr + map_size,
VM_MAP_NO_FLAGS);
(void)vnode_put(vp);
err = KERN_INVALID_ADDRESS;
goto bad;
}
result = vm_map_copyin(my_map, (vm_map_address_t)map_addr,
(vm_map_size_t)map_size, TRUE, &tmp);
if (result != KERN_SUCCESS) {
(void) vm_map_remove(my_map, vm_map_trunc_page(map_addr),
vm_map_round_page(map_addr + map_size),
VM_MAP_NO_FLAGS);
(void)vnode_put(vp);
err = result;
goto bad;
}
result = vm_map_copy_overwrite(my_map,
(vm_map_address_t)dst_addr, tmp, FALSE);
if (result != KERN_SUCCESS) {
vm_map_copy_discard(tmp);
(void)vnode_put(vp);
err = result;
goto bad;
}
} else {
if (copyout(&map_addr, CAST_USER_ADDR_T(va), sizeof (map_addr))) {
(void) vm_map_remove(my_map, vm_map_trunc_page(map_addr),
vm_map_round_page(map_addr + map_size),
VM_MAP_NO_FLAGS);
(void)vnode_put(vp);
err = KERN_INVALID_ADDRESS;
goto bad;
}
}
ubc_setthreadcred(vp, current_proc(), current_thread());
(void)ubc_map(vp, (PROT_READ | PROT_EXEC));
(void)vnode_put(vp);
err = 0;
bad:
fp_drop(p, fd, fp, 0);
return (err);
}
kern_return_t
mach_port_space_info(
ipc_space_t space,
ipc_info_space_t *infop,
ipc_info_name_array_t *tablep,
mach_msg_type_number_t *tableCntp,
__unused ipc_info_tree_name_array_t *treep,
__unused mach_msg_type_number_t *treeCntp)
{
ipc_info_name_t *table_info;
vm_offset_t table_addr;
vm_size_t table_size, table_size_needed;
ipc_entry_t table;
ipc_entry_num_t tsize;
mach_port_index_t index;
kern_return_t kr;
vm_map_copy_t copy;
if (space == IS_NULL)
return KERN_INVALID_TASK;
#if !(DEVELOPMENT | DEBUG)
const boolean_t dbg_ok = (mac_task_check_expose_task(kernel_task) == 0);
#else
const boolean_t dbg_ok = TRUE;
#endif
/* start with in-line memory */
table_size = 0;
for (;;) {
is_read_lock(space);
if (!is_active(space)) {
is_read_unlock(space);
if (table_size != 0)
kmem_free(ipc_kernel_map,
table_addr, table_size);
return KERN_INVALID_TASK;
}
table_size_needed =
vm_map_round_page((space->is_table_size
* sizeof(ipc_info_name_t)),
VM_MAP_PAGE_MASK(ipc_kernel_map));
if (table_size_needed == table_size)
break;
is_read_unlock(space);
if (table_size != table_size_needed) {
if (table_size != 0)
kmem_free(ipc_kernel_map, table_addr, table_size);
kr = kmem_alloc(ipc_kernel_map, &table_addr, table_size_needed, VM_KERN_MEMORY_IPC);
if (kr != KERN_SUCCESS) {
return KERN_RESOURCE_SHORTAGE;
}
table_size = table_size_needed;
}
}
/* space is read-locked and active; we have enough wired memory */
/* get the overall space info */
infop->iis_genno_mask = MACH_PORT_NGEN(MACH_PORT_DEAD);
infop->iis_table_size = space->is_table_size;
infop->iis_table_next = space->is_table_next->its_size;
/* walk the table for this space */
table = space->is_table;
tsize = space->is_table_size;
table_info = (ipc_info_name_array_t)table_addr;
for (index = 0; index < tsize; index++) {
ipc_info_name_t *iin = &table_info[index];
ipc_entry_t entry = &table[index];
ipc_entry_bits_t bits;
bits = entry->ie_bits;
iin->iin_name = MACH_PORT_MAKE(index, IE_BITS_GEN(bits));
iin->iin_collision = 0;
iin->iin_type = IE_BITS_TYPE(bits);
if ((entry->ie_bits & MACH_PORT_TYPE_PORT_RIGHTS) != MACH_PORT_TYPE_NONE &&
entry->ie_request != IE_REQ_NONE) {
__IGNORE_WCASTALIGN(ipc_port_t port = (ipc_port_t) entry->ie_object);
assert(IP_VALID(port));
ip_lock(port);
iin->iin_type |= ipc_port_request_type(port, iin->iin_name, entry->ie_request);
ip_unlock(port);
}
iin->iin_urefs = IE_BITS_UREFS(bits);
iin->iin_object = (dbg_ok) ? (natural_t)VM_KERNEL_ADDRPERM((uintptr_t)entry->ie_object) : 0;
iin->iin_next = entry->ie_next;
iin->iin_hash = entry->ie_index;
}
is_read_unlock(space);
/* prepare the table out-of-line data for return */
if (table_size > 0) {
vm_size_t used_table_size;
used_table_size = infop->iis_table_size * sizeof(ipc_info_name_t);
if (table_size > used_table_size)
bzero((char *)&table_info[infop->iis_table_size],
table_size - used_table_size);
kr = vm_map_unwire(
ipc_kernel_map,
vm_map_trunc_page(table_addr,
VM_MAP_PAGE_MASK(ipc_kernel_map)),
vm_map_round_page(table_addr + table_size,
VM_MAP_PAGE_MASK(ipc_kernel_map)),
FALSE);
assert(kr == KERN_SUCCESS);
kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)table_addr,
(vm_map_size_t)used_table_size, TRUE, ©);
assert(kr == KERN_SUCCESS);
*tablep = (ipc_info_name_t *)copy;
*tableCntp = infop->iis_table_size;
} else {
*tablep = (ipc_info_name_t *)0;
*tableCntp = 0;
}
/* splay tree is obsolete, no work to do... */
*treep = (ipc_info_tree_name_t *)0;
*treeCntp = 0;
return KERN_SUCCESS;
}
kern_return_t
kernel_memory_allocate(
register vm_map_t map,
register vm_offset_t *addrp,
register vm_size_t size,
register vm_offset_t mask,
int flags)
{
vm_object_t object;
vm_object_offset_t offset;
vm_object_offset_t pg_offset;
vm_map_entry_t entry;
vm_map_offset_t map_addr, fill_start;
vm_map_offset_t map_mask;
vm_map_size_t map_size, fill_size;
kern_return_t kr;
vm_page_t mem;
vm_page_t guard_page_list = NULL;
vm_page_t wired_page_list = NULL;
int guard_page_count = 0;
int wired_page_count = 0;
int i;
int vm_alloc_flags;
if (! vm_kernel_ready) {
panic("kernel_memory_allocate: VM is not ready");
}
if (size == 0) {
*addrp = 0;
return KERN_INVALID_ARGUMENT;
}
map_size = vm_map_round_page(size);
map_mask = (vm_map_offset_t) mask;
vm_alloc_flags = 0;
/*
* limit the size of a single extent of wired memory
* to try and limit the damage to the system if
* too many pages get wired down
*/
if (map_size > (1 << 30)) {
return KERN_RESOURCE_SHORTAGE;
}
/*
* Guard pages:
*
* Guard pages are implemented as ficticious pages. By placing guard pages
* on either end of a stack, they can help detect cases where a thread walks
* off either end of its stack. They are allocated and set up here and attempts
* to access those pages are trapped in vm_fault_page().
*
* The map_size we were passed may include extra space for
* guard pages. If those were requested, then back it out of fill_size
* since vm_map_find_space() takes just the actual size not including
* guard pages. Similarly, fill_start indicates where the actual pages
* will begin in the range.
*/
fill_start = 0;
fill_size = map_size;
if (flags & KMA_GUARD_FIRST) {
vm_alloc_flags |= VM_FLAGS_GUARD_BEFORE;
fill_start += PAGE_SIZE_64;
fill_size -= PAGE_SIZE_64;
if (map_size < fill_start + fill_size) {
/* no space for a guard page */
*addrp = 0;
return KERN_INVALID_ARGUMENT;
}
guard_page_count++;
}
if (flags & KMA_GUARD_LAST) {
vm_alloc_flags |= VM_FLAGS_GUARD_AFTER;
fill_size -= PAGE_SIZE_64;
if (map_size <= fill_start + fill_size) {
/* no space for a guard page */
*addrp = 0;
return KERN_INVALID_ARGUMENT;
}
guard_page_count++;
}
wired_page_count = (int) (fill_size / PAGE_SIZE_64);
assert(wired_page_count * PAGE_SIZE_64 == fill_size);
for (i = 0; i < guard_page_count; i++) {
for (;;) {
mem = vm_page_grab_guard();
if (mem != VM_PAGE_NULL)
break;
if (flags & KMA_NOPAGEWAIT) {
kr = KERN_RESOURCE_SHORTAGE;
goto out;
}
vm_page_more_fictitious();
}
mem->pageq.next = (queue_entry_t)guard_page_list;
guard_page_list = mem;
}
for (i = 0; i < wired_page_count; i++) {
uint64_t unavailable;
for (;;) {
if (flags & KMA_LOMEM)
mem = vm_page_grablo();
else
mem = vm_page_grab();
if (mem != VM_PAGE_NULL)
break;
if (flags & KMA_NOPAGEWAIT) {
kr = KERN_RESOURCE_SHORTAGE;
goto out;
}
if ((flags & KMA_LOMEM) && (vm_lopage_needed == TRUE)) {
kr = KERN_RESOURCE_SHORTAGE;
goto out;
}
unavailable = (vm_page_wire_count + vm_page_free_target) * PAGE_SIZE;
if (unavailable > max_mem || map_size > (max_mem - unavailable)) {
kr = KERN_RESOURCE_SHORTAGE;
goto out;
}
VM_PAGE_WAIT();
}
mem->pageq.next = (queue_entry_t)wired_page_list;
wired_page_list = mem;
}
/*
* Allocate a new object (if necessary). We must do this before
* locking the map, or risk deadlock with the default pager.
*/
if ((flags & KMA_KOBJECT) != 0) {
object = kernel_object;
vm_object_reference(object);
} else {
object = vm_object_allocate(map_size);
}
kr = vm_map_find_space(map, &map_addr,
fill_size, map_mask,
vm_alloc_flags, &entry);
if (KERN_SUCCESS != kr) {
vm_object_deallocate(object);
goto out;
}
entry->object.vm_object = object;
entry->offset = offset = (object == kernel_object) ?
map_addr : 0;
entry->wired_count++;
if (flags & KMA_PERMANENT)
entry->permanent = TRUE;
if (object != kernel_object)
vm_object_reference(object);
vm_object_lock(object);
vm_map_unlock(map);
pg_offset = 0;
if (fill_start) {
if (guard_page_list == NULL)
panic("kernel_memory_allocate: guard_page_list == NULL");
mem = guard_page_list;
guard_page_list = (vm_page_t)mem->pageq.next;
mem->pageq.next = NULL;
vm_page_insert(mem, object, offset + pg_offset);
mem->busy = FALSE;
pg_offset += PAGE_SIZE_64;
}
for (pg_offset = fill_start; pg_offset < fill_start + fill_size; pg_offset += PAGE_SIZE_64) {
if (wired_page_list == NULL)
panic("kernel_memory_allocate: wired_page_list == NULL");
mem = wired_page_list;
wired_page_list = (vm_page_t)mem->pageq.next;
mem->pageq.next = NULL;
mem->wire_count++;
vm_page_insert(mem, object, offset + pg_offset);
mem->busy = FALSE;
mem->pmapped = TRUE;
mem->wpmapped = TRUE;
PMAP_ENTER(kernel_pmap, map_addr + pg_offset, mem,
VM_PROT_READ | VM_PROT_WRITE, object->wimg_bits & VM_WIMG_MASK, TRUE);
if (flags & KMA_NOENCRYPT) {
bzero(CAST_DOWN(void *, (map_addr + pg_offset)), PAGE_SIZE);
pmap_set_noencrypt(mem->phys_page);
}
}
int
mincore(__unused proc_t p, struct mincore_args *uap, __unused int32_t *retval)
{
mach_vm_offset_t addr, first_addr, end;
vm_map_t map;
user_addr_t vec;
int error;
int vecindex, lastvecindex;
int mincoreinfo=0;
int pqueryinfo;
kern_return_t ret;
int numref;
char c;
map = current_map();
/*
* Make sure that the addresses presented are valid for user
* mode.
*/
first_addr = addr = vm_map_trunc_page(uap->addr,
vm_map_page_mask(map));
end = addr + vm_map_round_page(uap->len,
vm_map_page_mask(map));
if (end < addr)
return (EINVAL);
/*
* Address of byte vector
*/
vec = uap->vec;
map = current_map();
/*
* Do this on a map entry basis so that if the pages are not
* in the current processes address space, we can easily look
* up the pages elsewhere.
*/
lastvecindex = -1;
for( ; addr < end; addr += PAGE_SIZE ) {
pqueryinfo = 0;
ret = mach_vm_page_query(map, addr, &pqueryinfo, &numref);
if (ret != KERN_SUCCESS)
pqueryinfo = 0;
mincoreinfo = 0;
if (pqueryinfo & VM_PAGE_QUERY_PAGE_PRESENT)
mincoreinfo |= MINCORE_INCORE;
if (pqueryinfo & VM_PAGE_QUERY_PAGE_REF)
mincoreinfo |= MINCORE_REFERENCED;
if (pqueryinfo & VM_PAGE_QUERY_PAGE_DIRTY)
mincoreinfo |= MINCORE_MODIFIED;
/*
* calculate index into user supplied byte vector
*/
vecindex = (addr - first_addr)>> PAGE_SHIFT;
/*
* If we have skipped map entries, we need to make sure that
* the byte vector is zeroed for those skipped entries.
*/
while((lastvecindex + 1) < vecindex) {
c = 0;
error = copyout(&c, vec + lastvecindex, 1);
if (error) {
return (EFAULT);
}
++lastvecindex;
}
/*
* Pass the page information to the user
*/
c = (char)mincoreinfo;
error = copyout(&c, vec + vecindex, 1);
if (error) {
return (EFAULT);
}
lastvecindex = vecindex;
}
/*
* Zero the last entries in the byte vector.
*/
vecindex = (end - first_addr) >> PAGE_SHIFT;
while((lastvecindex + 1) < vecindex) {
c = 0;
error = copyout(&c, vec + lastvecindex, 1);
if (error) {
return (EFAULT);
}
++lastvecindex;
}
return (0);
}
/*
* Return an array of virtual pages that are mapped to a task.
*/
kern_return_t
vm32_mapped_pages_info(
__DEBUG_ONLY vm_map_t map,
__DEBUG_ONLY page_address_array_t *pages,
__DEBUG_ONLY mach_msg_type_number_t *pages_count)
{
#if !MACH_VM_DEBUG
return KERN_FAILURE;
#else
pmap_t pmap;
vm_size_t size, size_used;
unsigned int actual, space;
page_address_array_t list;
vm_offset_t addr = 0;
if (map == VM_MAP_NULL)
return (KERN_INVALID_ARGUMENT);
pmap = map->pmap;
size = pmap_resident_count(pmap) * sizeof(vm_offset_t);
size = vm_map_round_page(size,
VM_MAP_PAGE_MASK(ipc_kernel_map));
for (;;) {
(void) vm_allocate(ipc_kernel_map, &addr, size, VM_FLAGS_ANYWHERE | VM_MAKE_TAG(VM_KERN_MEMORY_IPC));
(void) vm_map_unwire(
ipc_kernel_map,
vm_map_trunc_page(addr,
VM_MAP_PAGE_MASK(ipc_kernel_map)),
vm_map_round_page(addr + size,
VM_MAP_PAGE_MASK(ipc_kernel_map)),
FALSE);
list = (page_address_array_t) addr;
space = (unsigned int) (size / sizeof(vm_offset_t));
actual = pmap_list_resident_pages(pmap,
list,
space);
if (actual <= space)
break;
/*
* Free memory if not enough
*/
(void) kmem_free(ipc_kernel_map, addr, size);
/*
* Try again, doubling the size
*/
size = vm_map_round_page(actual * sizeof(vm_offset_t),
VM_MAP_PAGE_MASK(ipc_kernel_map));
}
if (actual == 0) {
*pages = 0;
*pages_count = 0;
(void) kmem_free(ipc_kernel_map, addr, size);
}
else {
vm_size_t vmsize_used;
*pages_count = actual;
size_used = (actual * sizeof(vm_offset_t));
vmsize_used = vm_map_round_page(size_used,
VM_MAP_PAGE_MASK(ipc_kernel_map));
(void) vm_map_wire(
ipc_kernel_map,
vm_map_trunc_page(addr,
VM_MAP_PAGE_MASK(ipc_kernel_map)),
vm_map_round_page(addr + size,
VM_MAP_PAGE_MASK(ipc_kernel_map)),
VM_PROT_READ|VM_PROT_WRITE,
FALSE);
(void) vm_map_copyin(ipc_kernel_map,
(vm_map_address_t)addr,
(vm_map_size_t)size_used,
TRUE,
(vm_map_copy_t *)pages);
if (vmsize_used != size) {
(void) kmem_free(ipc_kernel_map,
addr + vmsize_used,
size - vmsize_used);
}
}
return (KERN_SUCCESS);
#endif /* MACH_VM_DEBUG */
}
/*
* On x86_64 systems, kernel extension text must remain within 2GB of the
* kernel's text segment. To ensure this happens, we snag 2GB of kernel VM
* as early as possible for kext allocations.
*/
void
kext_alloc_init(void)
{
#if CONFIG_KEXT_BASEMENT
kern_return_t rval = 0;
kernel_segment_command_t *text = NULL;
kernel_segment_command_t *prelinkTextSegment = NULL;
mach_vm_offset_t text_end, text_start;
mach_vm_size_t text_size;
mach_vm_size_t kext_alloc_size;
/* Determine the start of the kernel's __TEXT segment and determine the
* lower bound of the allocated submap for kext allocations.
*/
text = getsegbyname(SEG_TEXT);
text_start = vm_map_trunc_page(text->vmaddr,
VM_MAP_PAGE_MASK(kernel_map));
text_start &= ~((512ULL * 1024 * 1024 * 1024) - 1);
text_end = vm_map_round_page(text->vmaddr + text->vmsize,
VM_MAP_PAGE_MASK(kernel_map));
text_size = text_end - text_start;
kext_alloc_base = KEXT_ALLOC_BASE(text_end);
kext_alloc_size = KEXT_ALLOC_SIZE(text_size);
kext_alloc_max = kext_alloc_base + kext_alloc_size;
/* Post boot kext allocation will start after the prelinked kexts */
prelinkTextSegment = getsegbyname("__PRELINK_TEXT");
if (prelinkTextSegment) {
/* use kext_post_boot_base to start allocations past all the prelinked
* kexts
*/
kext_post_boot_base =
vm_map_round_page(kext_alloc_base + prelinkTextSegment->vmsize,
VM_MAP_PAGE_MASK(kernel_map));
}
else {
kext_post_boot_base = kext_alloc_base;
}
/* Allocate the sub block of the kernel map */
rval = kmem_suballoc(kernel_map, (vm_offset_t *) &kext_alloc_base,
kext_alloc_size, /* pageable */ TRUE,
VM_FLAGS_FIXED|VM_FLAGS_OVERWRITE,
&g_kext_map);
if (rval != KERN_SUCCESS) {
panic("kext_alloc_init: kmem_suballoc failed 0x%x\n", rval);
}
if ((kext_alloc_base + kext_alloc_size) > kext_alloc_max) {
panic("kext_alloc_init: failed to get first 2GB\n");
}
if (kernel_map->min_offset > kext_alloc_base) {
kernel_map->min_offset = kext_alloc_base;
}
printf("kext submap [0x%lx - 0x%lx], kernel text [0x%lx - 0x%lx]\n",
VM_KERNEL_UNSLIDE(kext_alloc_base),
VM_KERNEL_UNSLIDE(kext_alloc_max),
VM_KERNEL_UNSLIDE(text->vmaddr),
VM_KERNEL_UNSLIDE(text->vmaddr + text->vmsize));
#else
g_kext_map = kernel_map;
kext_alloc_base = VM_MIN_KERNEL_ADDRESS;
kext_alloc_max = VM_MAX_KERNEL_ADDRESS;
#endif /* CONFIG_KEXT_BASEMENT */
}
static
load_return_t
load_segment(
struct load_command *lcp,
uint32_t filetype,
void * control,
off_t pager_offset,
off_t macho_size,
struct vnode *vp,
vm_map_t map,
int64_t slide,
load_result_t *result
)
{
struct segment_command_64 segment_command, *scp;
kern_return_t ret;
vm_map_offset_t map_addr, map_offset;
vm_map_size_t map_size, seg_size, delta_size;
vm_prot_t initprot;
vm_prot_t maxprot;
size_t segment_command_size, total_section_size,
single_section_size;
boolean_t prohibit_pagezero_mapping = FALSE;
if (LC_SEGMENT_64 == lcp->cmd) {
segment_command_size = sizeof(struct segment_command_64);
single_section_size = sizeof(struct section_64);
} else {
segment_command_size = sizeof(struct segment_command);
single_section_size = sizeof(struct section);
}
if (lcp->cmdsize < segment_command_size)
return (LOAD_BADMACHO);
total_section_size = lcp->cmdsize - segment_command_size;
if (LC_SEGMENT_64 == lcp->cmd)
scp = (struct segment_command_64 *)lcp;
else {
scp = &segment_command;
widen_segment_command((struct segment_command *)lcp, scp);
}
/*
* Make sure what we get from the file is really ours (as specified
* by macho_size).
*/
if (scp->fileoff + scp->filesize < scp->fileoff ||
scp->fileoff + scp->filesize > (uint64_t)macho_size)
return (LOAD_BADMACHO);
/*
* Ensure that the number of sections specified would fit
* within the load command size.
*/
if (total_section_size / single_section_size < scp->nsects)
return (LOAD_BADMACHO);
/*
* Make sure the segment is page-aligned in the file.
*/
if ((scp->fileoff & PAGE_MASK_64) != 0)
return (LOAD_BADMACHO);
/*
* Round sizes to page size.
*/
seg_size = round_page_64(scp->vmsize);
map_size = round_page_64(scp->filesize);
map_addr = trunc_page_64(scp->vmaddr); /* JVXXX note that in XNU TOT this is round instead of trunc for 64 bits */
seg_size = vm_map_round_page(seg_size, vm_map_page_mask(map));
map_size = vm_map_round_page(map_size, vm_map_page_mask(map));
if (seg_size == 0)
return (KERN_SUCCESS);
if (map_addr == 0 &&
map_size == 0 &&
seg_size != 0 &&
(scp->initprot & VM_PROT_ALL) == VM_PROT_NONE &&
(scp->maxprot & VM_PROT_ALL) == VM_PROT_NONE) {
/*
* For PIE, extend page zero rather than moving it. Extending
* page zero keeps early allocations from falling predictably
* between the end of page zero and the beginning of the first
* slid segment.
*/
seg_size += slide;
slide = 0;
/* XXX (4596982) this interferes with Rosetta, so limit to 64-bit tasks */
if (scp->cmd == LC_SEGMENT_64) {
prohibit_pagezero_mapping = TRUE;
}
if (prohibit_pagezero_mapping) {
/*
* This is a "page zero" segment: it starts at address 0,
* is not mapped from the binary file and is not accessible.
* User-space should never be able to access that memory, so
* make it completely off limits by raising the VM map's
* minimum offset.
*/
ret = vm_map_raise_min_offset(map, seg_size);
if (ret != KERN_SUCCESS) {
return (LOAD_FAILURE);
}
return (LOAD_SUCCESS);
}
}
/* If a non-zero slide was specified by the caller, apply now */
map_addr += slide;
if (map_addr < result->min_vm_addr)
result->min_vm_addr = map_addr;
if (map_addr+seg_size > result->max_vm_addr)
result->max_vm_addr = map_addr+seg_size;
if (map == VM_MAP_NULL)
return (LOAD_SUCCESS);
map_offset = pager_offset + scp->fileoff; /* limited to 32 bits */
if (map_size > 0) {
initprot = (scp->initprot) & VM_PROT_ALL;
maxprot = (scp->maxprot) & VM_PROT_ALL;
/*
* Map a copy of the file into the address space.
*/
ret = vm_map_enter_mem_object_control(map,
&map_addr, map_size, (mach_vm_offset_t)0,
VM_FLAGS_FIXED, control, map_offset, TRUE,
initprot, maxprot,
VM_INHERIT_DEFAULT);
if (ret != KERN_SUCCESS) {
return (LOAD_NOSPACE);
}
/*
* If the file didn't end on a page boundary,
* we need to zero the leftover.
*/
delta_size = map_size - scp->filesize;
#if FIXME
if (delta_size > 0) {
mach_vm_offset_t tmp;
ret = mach_vm_allocate(kernel_map, &tmp, delta_size, VM_FLAGS_ANYWHERE);
if (ret != KERN_SUCCESS)
return(LOAD_RESOURCE);
if (copyout(tmp, map_addr + scp->filesize,
delta_size)) {
(void) mach_vm_deallocate(
kernel_map, tmp, delta_size);
return (LOAD_FAILURE);
}
(void) mach_vm_deallocate(kernel_map, tmp, delta_size);
}
#endif /* FIXME */
}
/*
* If the virtual size of the segment is greater
* than the size from the file, we need to allocate
* zero fill memory for the rest.
*/
delta_size = seg_size - map_size;
if (delta_size > 0) {
mach_vm_offset_t tmp = map_addr + map_size;
ret = mach_vm_map(map, &tmp, delta_size, 0, VM_FLAGS_FIXED,
NULL, 0, FALSE,
scp->initprot, scp->maxprot,
VM_INHERIT_DEFAULT);
if (ret != KERN_SUCCESS)
return(LOAD_NOSPACE);
}
if ( (scp->fileoff == 0) && (scp->filesize != 0) )
result->mach_header = map_addr;
if (scp->flags & SG_PROTECTED_VERSION_1) {
ret = unprotect_segment(scp->fileoff,
scp->filesize,
vp,
pager_offset,
map,
map_addr,
map_size);
} else {
ret = LOAD_SUCCESS;
}
if (LOAD_SUCCESS == ret && filetype == MH_DYLINKER &&
result->all_image_info_addr == MACH_VM_MIN_ADDRESS)
note_all_image_info_section(scp,
LC_SEGMENT_64 == lcp->cmd, single_section_size,
(const char *)lcp + segment_command_size, slide, result);
if ((result->entry_point >= map_addr) && (result->entry_point < (map_addr + map_size)))
result->validentry = 1;
return ret;
}
kern_return_t
mach_port_space_info(
ipc_space_t space,
ipc_info_space_t *infop,
ipc_info_name_array_t *tablep,
mach_msg_type_number_t *tableCntp,
ipc_info_tree_name_array_t *treep,
mach_msg_type_number_t *treeCntp)
{
ipc_info_name_t *table_info;
vm_offset_t table_addr;
vm_size_t table_size, table_size_needed;
ipc_info_tree_name_t *tree_info;
vm_offset_t tree_addr;
vm_size_t tree_size, tree_size_needed;
ipc_tree_entry_t tentry;
ipc_entry_t table;
ipc_entry_num_t tsize;
mach_port_index_t index;
kern_return_t kr;
vm_map_copy_t copy;
if (space == IS_NULL)
return KERN_INVALID_TASK;
/* start with in-line memory */
table_size = 0;
tree_size = 0;
for (;;) {
is_read_lock(space);
if (!space->is_active) {
is_read_unlock(space);
if (table_size != 0)
kmem_free(ipc_kernel_map,
table_addr, table_size);
if (tree_size != 0)
kmem_free(ipc_kernel_map,
tree_addr, tree_size);
return KERN_INVALID_TASK;
}
table_size_needed = round_page(space->is_table_size
* sizeof(ipc_info_name_t));
tree_size_needed = round_page(space->is_tree_total
* sizeof(ipc_info_tree_name_t));
if ((table_size_needed == table_size) &&
(tree_size_needed == tree_size))
break;
is_read_unlock(space);
if (table_size != table_size_needed) {
if (table_size != 0)
kmem_free(ipc_kernel_map, table_addr, table_size);
kr = kmem_alloc(ipc_kernel_map, &table_addr, table_size_needed);
if (kr != KERN_SUCCESS) {
if (tree_size != 0)
kmem_free(ipc_kernel_map, tree_addr, tree_size);
return KERN_RESOURCE_SHORTAGE;
}
table_size = table_size_needed;
}
if (tree_size != tree_size_needed) {
if (tree_size != 0)
kmem_free(ipc_kernel_map, tree_addr, tree_size);
kr = kmem_alloc(ipc_kernel_map, &tree_addr, tree_size_needed);
if (kr != KERN_SUCCESS) {
if (table_size != 0)
kmem_free(ipc_kernel_map, table_addr, table_size);
return KERN_RESOURCE_SHORTAGE;
}
tree_size = tree_size_needed;
}
}
/* space is read-locked and active; we have enough wired memory */
/* get the overall space info */
infop->iis_genno_mask = MACH_PORT_NGEN(MACH_PORT_DEAD);
infop->iis_table_size = space->is_table_size;
infop->iis_table_next = space->is_table_next->its_size;
infop->iis_tree_size = space->is_tree_total;
infop->iis_tree_small = space->is_tree_small;
infop->iis_tree_hash = space->is_tree_hash;
/* walk the table for this space */
table = space->is_table;
tsize = space->is_table_size;
table_info = (ipc_info_name_array_t)table_addr;
for (index = 0; index < tsize; index++) {
ipc_info_name_t *iin = &table_info[index];
ipc_entry_t entry = &table[index];
ipc_entry_bits_t bits;
bits = entry->ie_bits;
iin->iin_name = MACH_PORT_MAKE(index, IE_BITS_GEN(bits));
iin->iin_collision = (bits & IE_BITS_COLLISION) ? TRUE : FALSE;
iin->iin_type = IE_BITS_TYPE(bits);
if (entry->ie_request)
iin->iin_type |= MACH_PORT_TYPE_DNREQUEST;
iin->iin_urefs = IE_BITS_UREFS(bits);
iin->iin_object = (vm_offset_t) entry->ie_object;
iin->iin_next = entry->ie_next;
iin->iin_hash = entry->ie_index;
}
/* walk the splay tree for this space */
tree_info = (ipc_info_tree_name_array_t)tree_addr;
for (tentry = ipc_splay_traverse_start(&space->is_tree), index = 0;
tentry != ITE_NULL;
tentry = ipc_splay_traverse_next(&space->is_tree, FALSE)) {
ipc_info_tree_name_t *iitn = &tree_info[index++];
ipc_info_name_t *iin = &iitn->iitn_name;
ipc_entry_t entry = &tentry->ite_entry;
ipc_entry_bits_t bits = entry->ie_bits;
assert(IE_BITS_TYPE(bits) != MACH_PORT_TYPE_NONE);
iin->iin_name = tentry->ite_name;
iin->iin_collision = (bits & IE_BITS_COLLISION) ? TRUE : FALSE;
iin->iin_type = IE_BITS_TYPE(bits);
if (entry->ie_request)
iin->iin_type |= MACH_PORT_TYPE_DNREQUEST;
iin->iin_urefs = IE_BITS_UREFS(bits);
iin->iin_object = (vm_offset_t) entry->ie_object;
iin->iin_next = entry->ie_next;
iin->iin_hash = entry->ie_index;
if (tentry->ite_lchild == ITE_NULL)
iitn->iitn_lchild = MACH_PORT_NULL;
else
iitn->iitn_lchild = tentry->ite_lchild->ite_name;
if (tentry->ite_rchild == ITE_NULL)
iitn->iitn_rchild = MACH_PORT_NULL;
else
iitn->iitn_rchild = tentry->ite_rchild->ite_name;
}
ipc_splay_traverse_finish(&space->is_tree);
is_read_unlock(space);
/* prepare the table out-of-line data for return */
if (table_size > 0) {
if (table_size > infop->iis_table_size * sizeof(ipc_info_name_t))
bzero((char *)&table_info[infop->iis_table_size],
table_size - infop->iis_table_size * sizeof(ipc_info_name_t));
kr = vm_map_unwire(ipc_kernel_map, vm_map_trunc_page(table_addr),
vm_map_round_page(table_addr + table_size), FALSE);
assert(kr == KERN_SUCCESS);
kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)table_addr,
(vm_map_size_t)table_size, TRUE, ©);
assert(kr == KERN_SUCCESS);
*tablep = (ipc_info_name_t *)copy;
*tableCntp = infop->iis_table_size;
} else {
*tablep = (ipc_info_name_t *)0;
*tableCntp = 0;
}
/* prepare the tree out-of-line data for return */
if (tree_size > 0) {
if (tree_size > infop->iis_tree_size * sizeof(ipc_info_tree_name_t))
bzero((char *)&tree_info[infop->iis_tree_size],
tree_size - infop->iis_tree_size * sizeof(ipc_info_tree_name_t));
kr = vm_map_unwire(ipc_kernel_map, vm_map_trunc_page(tree_addr),
vm_map_round_page(tree_addr + tree_size), FALSE);
assert(kr == KERN_SUCCESS);
kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)tree_addr,
(vm_map_size_t)tree_size, TRUE, ©);
assert(kr == KERN_SUCCESS);
*treep = (ipc_info_tree_name_t *)copy;
*treeCntp = infop->iis_tree_size;
} else {
*treep = (ipc_info_tree_name_t *)0;
*treeCntp = 0;
}
return KERN_SUCCESS;
}
kern_return_t
vm32_region_info_64(
__DEBUG_ONLY vm_map_t map,
__DEBUG_ONLY vm32_offset_t address,
__DEBUG_ONLY vm_info_region_64_t *regionp,
__DEBUG_ONLY vm_info_object_array_t *objectsp,
__DEBUG_ONLY mach_msg_type_number_t *objectsCntp)
{
#if !MACH_VM_DEBUG
return KERN_FAILURE;
#else
vm_map_copy_t copy;
vm_offset_t addr = 0; /* memory for OOL data */
vm_size_t size; /* size of the memory */
unsigned int room; /* room for this many objects */
unsigned int used; /* actually this many objects */
vm_info_region_64_t region;
kern_return_t kr;
if (map == VM_MAP_NULL)
return KERN_INVALID_TASK;
size = 0; /* no memory allocated yet */
for (;;) {
vm_map_t cmap; /* current map in traversal */
vm_map_t nmap; /* next map to look at */
vm_map_entry_t entry;
vm_object_t object, cobject, nobject;
/* nothing is locked */
vm_map_lock_read(map);
for (cmap = map;; cmap = nmap) {
/* cmap is read-locked */
if (!vm_map_lookup_entry(cmap, address, &entry)) {
entry = entry->vme_next;
if (entry == vm_map_to_entry(cmap)) {
vm_map_unlock_read(cmap);
if (size != 0)
kmem_free(ipc_kernel_map,
addr, size);
return KERN_NO_SPACE;
}
}
if (entry->is_sub_map)
nmap = VME_SUBMAP(entry);
else
break;
/* move down to the lower map */
vm_map_lock_read(nmap);
vm_map_unlock_read(cmap);
}
/* cmap is read-locked; we have a real entry */
object = VME_OBJECT(entry);
region.vir_start = (natural_t) entry->vme_start;
region.vir_end = (natural_t) entry->vme_end;
region.vir_object = (natural_t)(uintptr_t) object;
region.vir_offset = VME_OFFSET(entry);
region.vir_needs_copy = entry->needs_copy;
region.vir_protection = entry->protection;
region.vir_max_protection = entry->max_protection;
region.vir_inheritance = entry->inheritance;
region.vir_wired_count = entry->wired_count;
region.vir_user_wired_count = entry->user_wired_count;
used = 0;
room = (unsigned int) (size / sizeof(vm_info_object_t));
if (object == VM_OBJECT_NULL) {
vm_map_unlock_read(cmap);
/* no memory needed */
break;
}
vm_object_lock(object);
vm_map_unlock_read(cmap);
for (cobject = object;; cobject = nobject) {
/* cobject is locked */
if (used < room) {
vm_info_object_t *vio =
&((vm_info_object_t *) addr)[used];
vio->vio_object =
(natural_t)(uintptr_t) cobject;
vio->vio_size =
(natural_t) cobject->vo_size;
vio->vio_ref_count =
cobject->ref_count;
vio->vio_resident_page_count =
cobject->resident_page_count;
vio->vio_copy =
(natural_t)(uintptr_t) cobject->copy;
vio->vio_shadow =
(natural_t)(uintptr_t) cobject->shadow;
vio->vio_shadow_offset =
(natural_t) cobject->vo_shadow_offset;
vio->vio_paging_offset =
(natural_t) cobject->paging_offset;
vio->vio_copy_strategy =
cobject->copy_strategy;
vio->vio_last_alloc =
(vm_offset_t) cobject->last_alloc;
vio->vio_paging_in_progress =
cobject->paging_in_progress +
cobject->activity_in_progress;
vio->vio_pager_created =
cobject->pager_created;
vio->vio_pager_initialized =
cobject->pager_initialized;
vio->vio_pager_ready =
cobject->pager_ready;
vio->vio_can_persist =
cobject->can_persist;
vio->vio_internal =
cobject->internal;
vio->vio_temporary =
cobject->temporary;
vio->vio_alive =
cobject->alive;
vio->vio_purgable =
(cobject->purgable != VM_PURGABLE_DENY);
vio->vio_purgable_volatile =
(cobject->purgable == VM_PURGABLE_VOLATILE ||
cobject->purgable == VM_PURGABLE_EMPTY);
}
used++;
nobject = cobject->shadow;
if (nobject == VM_OBJECT_NULL) {
vm_object_unlock(cobject);
break;
}
vm_object_lock(nobject);
vm_object_unlock(cobject);
}
/* nothing locked */
if (used <= room)
break;
/* must allocate more memory */
if (size != 0)
kmem_free(ipc_kernel_map, addr, size);
size = vm_map_round_page(2 * used * sizeof(vm_info_object_t),
VM_MAP_PAGE_MASK(ipc_kernel_map));
kr = vm_allocate(ipc_kernel_map, &addr, size, VM_FLAGS_ANYWHERE | VM_MAKE_TAG(VM_KERN_MEMORY_IPC));
if (kr != KERN_SUCCESS)
return KERN_RESOURCE_SHORTAGE;
kr = vm_map_wire(
ipc_kernel_map,
vm_map_trunc_page(addr,
VM_MAP_PAGE_MASK(ipc_kernel_map)),
vm_map_round_page(addr + size,
VM_MAP_PAGE_MASK(ipc_kernel_map)),
VM_PROT_READ|VM_PROT_WRITE,
FALSE);
assert(kr == KERN_SUCCESS);
}
/* free excess memory; make remaining memory pageable */
if (used == 0) {
copy = VM_MAP_COPY_NULL;
if (size != 0)
kmem_free(ipc_kernel_map, addr, size);
} else {
vm_size_t size_used = (used * sizeof(vm_info_object_t));
vm_size_t vmsize_used = vm_map_round_page(size_used,
VM_MAP_PAGE_MASK(ipc_kernel_map));
kr = vm_map_unwire(
ipc_kernel_map,
vm_map_trunc_page(addr,
VM_MAP_PAGE_MASK(ipc_kernel_map)),
vm_map_round_page(addr + size_used,
VM_MAP_PAGE_MASK(ipc_kernel_map)),
FALSE);
assert(kr == KERN_SUCCESS);
kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)addr,
(vm_map_size_t)size_used, TRUE, ©);
assert(kr == KERN_SUCCESS);
if (size != vmsize_used)
kmem_free(ipc_kernel_map,
addr + vmsize_used, size - vmsize_used);
}
*regionp = region;
*objectsp = (vm_info_object_array_t) copy;
*objectsCntp = used;
return KERN_SUCCESS;
#endif /* MACH_VM_DEBUG */
}
int
mprotect(__unused proc_t p, struct mprotect_args *uap, __unused int32_t *retval)
{
vm_prot_t prot;
mach_vm_offset_t user_addr;
mach_vm_size_t user_size;
kern_return_t result;
vm_map_t user_map;
#if CONFIG_MACF
int error;
#endif
AUDIT_ARG(addr, uap->addr);
AUDIT_ARG(len, uap->len);
AUDIT_ARG(value32, uap->prot);
user_map = current_map();
user_addr = (mach_vm_offset_t) uap->addr;
user_size = (mach_vm_size_t) uap->len;
prot = (vm_prot_t)(uap->prot & (VM_PROT_ALL | VM_PROT_TRUSTED | VM_PROT_STRIP_READ));
if (user_addr & vm_map_page_mask(user_map)) {
/* UNIX SPEC: user address is not page-aligned, return EINVAL */
return EINVAL;
}
#ifdef notyet
/* Hmm .. */
#if defined(VM_PROT_READ_IS_EXEC)
if (prot & VM_PROT_READ)
prot |= VM_PROT_EXECUTE;
#endif
#endif /* notyet */
#if 3936456
if (prot & (VM_PROT_EXECUTE | VM_PROT_WRITE))
prot |= VM_PROT_READ;
#endif /* 3936456 */
#if defined(__arm64__)
if (prot & VM_PROT_STRIP_READ)
prot &= ~(VM_PROT_READ | VM_PROT_STRIP_READ);
#endif
#if CONFIG_MACF
/*
* The MAC check for mprotect is of limited use for 2 reasons:
* Without mmap revocation, the caller could have asked for the max
* protections initially instead of a reduced set, so a mprotect
* check would offer no new security.
* It is not possible to extract the vnode from the pager object(s)
* of the target memory range.
* However, the MAC check may be used to prevent a process from,
* e.g., making the stack executable.
*/
error = mac_proc_check_mprotect(p, user_addr,
user_size, prot);
if (error)
return (error);
#endif
if(prot & VM_PROT_TRUSTED) {
#if CONFIG_DYNAMIC_CODE_SIGNING
/* CODE SIGNING ENFORCEMENT - JIT support */
/* The special protection value VM_PROT_TRUSTED requests that we treat
* this page as if it had a valid code signature.
* If this is enabled, there MUST be a MAC policy implementing the
* mac_proc_check_mprotect() hook above. Otherwise, Codesigning will be
* compromised because the check would always succeed and thusly any
* process could sign dynamically. */
result = vm_map_sign(
user_map,
vm_map_trunc_page(user_addr,
vm_map_page_mask(user_map)),
vm_map_round_page(user_addr+user_size,
vm_map_page_mask(user_map)));
switch (result) {
case KERN_SUCCESS:
break;
case KERN_INVALID_ADDRESS:
/* UNIX SPEC: for an invalid address range, return ENOMEM */
return ENOMEM;
default:
return EINVAL;
}
#else
return ENOTSUP;
#endif
}
prot &= ~VM_PROT_TRUSTED;
result = mach_vm_protect(user_map, user_addr, user_size,
FALSE, prot);
switch (result) {
case KERN_SUCCESS:
return (0);
case KERN_PROTECTION_FAILURE:
return (EACCES);
case KERN_INVALID_ADDRESS:
/* UNIX SPEC: for an invalid address range, return ENOMEM */
return ENOMEM;
}
return (EINVAL);
}
kern_return_t
host_virtual_physical_table_info(
__DEBUG_ONLY host_t host,
__DEBUG_ONLY hash_info_bucket_array_t *infop,
__DEBUG_ONLY mach_msg_type_number_t *countp)
{
#if !MACH_VM_DEBUG
return KERN_FAILURE;
#else
vm_offset_t addr = 0;
vm_size_t size = 0;
hash_info_bucket_t *info;
unsigned int potential, actual;
kern_return_t kr;
if (host == HOST_NULL)
return KERN_INVALID_HOST;
/* start with in-line data */
info = *infop;
potential = *countp;
for (;;) {
actual = vm_page_info(info, potential);
if (actual <= potential)
break;
/* allocate more memory */
if (info != *infop)
kmem_free(ipc_kernel_map, addr, size);
size = vm_map_round_page(actual * sizeof *info,
VM_MAP_PAGE_MASK(ipc_kernel_map));
kr = vm_allocate(ipc_kernel_map, &addr, size,
VM_FLAGS_ANYWHERE | VM_MAKE_TAG(VM_KERN_MEMORY_IPC));
if (kr != KERN_SUCCESS)
return KERN_RESOURCE_SHORTAGE;
info = (hash_info_bucket_t *) addr;
potential = (unsigned int) (size/sizeof (*info));
}
if (info == *infop) {
/* data fit in-line; nothing to deallocate */
*countp = actual;
} else if (actual == 0) {
kmem_free(ipc_kernel_map, addr, size);
*countp = 0;
} else {
vm_map_copy_t copy;
vm_size_t used, vmused;
used = (actual * sizeof(*info));
vmused = vm_map_round_page(used, VM_MAP_PAGE_MASK(ipc_kernel_map));
if (vmused != size)
kmem_free(ipc_kernel_map, addr + vmused, size - vmused);
kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)addr,
(vm_map_size_t)used, TRUE, ©);
assert(kr == KERN_SUCCESS);
*infop = (hash_info_bucket_t *) copy;
*countp = actual;
}
return KERN_SUCCESS;
#endif /* MACH_VM_DEBUG */
}
/*
* host_processor_info
*
* Return info about the processors on this host. It will return
* the number of processors, and the specific type of info requested
* in an OOL array.
*/
kern_return_t
host_processor_info(
host_t host,
processor_flavor_t flavor,
natural_t *out_pcount,
processor_info_array_t *out_array,
mach_msg_type_number_t *out_array_count)
{
kern_return_t result;
processor_t processor;
host_t thost;
processor_info_t info;
unsigned int icount, tcount;
unsigned int pcount, i;
vm_offset_t addr;
vm_size_t size, needed;
vm_map_copy_t copy;
if (host == HOST_NULL)
return (KERN_INVALID_ARGUMENT);
result = processor_info_count(flavor, &icount);
if (result != KERN_SUCCESS)
return (result);
pcount = processor_count;
assert(pcount != 0);
needed = pcount * icount * sizeof(natural_t);
size = round_page(needed);
result = kmem_alloc(ipc_kernel_map, &addr, size);
if (result != KERN_SUCCESS)
return (KERN_RESOURCE_SHORTAGE);
info = (processor_info_t) addr;
processor = processor_list;
tcount = icount;
result = processor_info(processor, flavor, &thost, info, &tcount);
if (result != KERN_SUCCESS) {
kmem_free(ipc_kernel_map, addr, size);
return (result);
}
if (pcount > 1) {
for (i = 1; i < pcount; i++) {
simple_lock(&processor_list_lock);
processor = processor->processor_list;
simple_unlock(&processor_list_lock);
info += icount;
tcount = icount;
result = processor_info(processor, flavor, &thost, info, &tcount);
if (result != KERN_SUCCESS) {
kmem_free(ipc_kernel_map, addr, size);
return (result);
}
}
}
if (size != needed)
bzero((char *) addr + needed, size - needed);
result = vm_map_unwire(ipc_kernel_map, vm_map_trunc_page(addr),
vm_map_round_page(addr + size), FALSE);
assert(result == KERN_SUCCESS);
result = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)addr,
(vm_map_size_t)size, TRUE, ©);
assert(result == KERN_SUCCESS);
*out_pcount = pcount;
*out_array = (processor_info_array_t) copy;
*out_array_count = pcount * icount;
return (KERN_SUCCESS);
}
