/*
* 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;
register struct vnode *vp;
int flags;
int prot, file_prot;
int err=0;
vm_map_t user_map;
kern_return_t result;
mach_vm_offset_t user_addr;
mach_vm_size_t user_size;
vm_object_offset_t pageoff;
vm_object_offset_t file_pos;
int alloc_flags=0;
boolean_t docow;
vm_prot_t maxprot;
void *handle;
vm_pager_t pager;
int mapanon=0;
int fpref=0;
int error =0;
int fd = uap->fd;
user_addr = (mach_vm_offset_t)uap->addr;
user_size = (mach_vm_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 & PAGE_MASK);
file_pos -= (vm_object_offset_t)pageoff;
/* Adjust size for rounding (on both ends). */
user_size += pageoff; /* low end... */
user_size = mach_vm_round_page(user_size); /* hi end */
/*
* 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 & PAGE_MASK)
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 < mach_vm_round_page(p->p_vmspace->vm_daddr + MAXDSIZ))
addr = mach_vm_round_page(p->p_vmspace->vm_daddr + MAXDSIZ);
#endif
alloc_flags = 0;
if (flags & MAP_ANON) {
/*
* 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_PURGABLE);
if (alloc_flags != fd) {
/* reject if there are any extra flags */
return EINVAL;
}
}
handle = NULL;
maxprot = VM_PROT_ALL;
file_pos = 0;
mapanon = 1;
} else {
struct vnode_attr va;
vfs_context_t ctx = vfs_context_current();
/*
* 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;
if(fp->f_fglob->fg_type == 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;
}
if (fp->f_fglob->fg_type != DTYPE_VNODE) {
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) {
/*
* 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, &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 = mach_vm_round_page(user_size);
if (file_pos & PAGE_MASK_64) {
if (!mapanon)
(void)vnode_put(vp);
error = EINVAL;
goto bad;
}
user_map = current_map();
if ((flags & MAP_FIXED) == 0) {
alloc_flags |= VM_FLAGS_ANYWHERE;
user_addr = mach_vm_round_page(user_addr);
} else {
if (user_addr != mach_vm_trunc_page(user_addr)) {
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;
/*
* Lookup/allocate object.
*/
if (handle == NULL) {
pager = 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 */
result = vm_map_enter_mem_object(user_map,
&user_addr, user_size,
0, alloc_flags,
IPC_PORT_NULL, 0, FALSE,
prot, maxprot,
(flags & MAP_SHARED) ?
VM_INHERIT_SHARE :
VM_INHERIT_DEFAULT);
if (result != KERN_SUCCESS)
goto out;
} else {
pager = (vm_pager_t)ubc_getpager(vp);
if (pager == 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 */
result = vm_map_enter_mem_object(user_map,
&user_addr, user_size,
0, alloc_flags,
(ipc_port_t)pager, file_pos,
docow, prot, maxprot,
(flags & MAP_SHARED) ?
VM_INHERIT_SHARE :
VM_INHERIT_DEFAULT);
if (result != KERN_SUCCESS) {
(void)vnode_put(vp);
goto out;
}
file_prot = prot & (PROT_READ | PROT_WRITE | PROT_EXEC);
if (docow) {
/* private mapping: won't write to the file */
file_prot &= ~PROT_WRITE;
}
(void) ubc_map(vp, file_prot);
}
if (!mapanon)
(void)vnode_put(vp);
out:
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 (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);
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);
return(error);
}
/*
* Routine: macx_swapon
* Function:
* Syscall interface to add a file to backing store
*/
int
macx_swapon(
struct macx_swapon_args *args)
{
int size = args->size;
vnode_t vp = (vnode_t)NULL;
struct nameidata nd, *ndp;
register int error;
kern_return_t kr;
mach_port_t backing_store;
memory_object_default_t default_pager;
int i;
boolean_t funnel_state;
off_t file_size;
vfs_context_t ctx = vfs_context_current();
struct proc *p = current_proc();
int dp_cluster_size;
AUDIT_MACH_SYSCALL_ENTER(AUE_SWAPON);
AUDIT_ARG(value32, args->priority);
funnel_state = thread_funnel_set(kernel_flock, TRUE);
ndp = &nd;
if ((error = suser(kauth_cred_get(), 0)))
goto swapon_bailout;
/*
* Get a vnode for the paging area.
*/
NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1,
((IS_64BIT_PROCESS(p)) ? UIO_USERSPACE64 : UIO_USERSPACE32),
(user_addr_t) args->filename, ctx);
if ((error = namei(ndp)))
goto swapon_bailout;
nameidone(ndp);
vp = ndp->ni_vp;
if (vp->v_type != VREG) {
error = EINVAL;
goto swapon_bailout;
}
/* get file size */
if ((error = vnode_size(vp, &file_size, ctx)) != 0)
goto swapon_bailout;
#if CONFIG_MACF
vnode_lock(vp);
error = mac_system_check_swapon(vfs_context_ucred(ctx), vp);
vnode_unlock(vp);
if (error)
goto swapon_bailout;
#endif
/* resize to desired size if it's too small */
if ((file_size < (off_t)size) && ((error = vnode_setsize(vp, (off_t)size, 0, ctx)) != 0))
goto swapon_bailout;
if (default_pager_init_flag == 0) {
start_def_pager(NULL);
default_pager_init_flag = 1;
}
/* add new backing store to list */
i = 0;
while(bs_port_table[i].vp != 0) {
if(i == MAX_BACKING_STORE)
break;
i++;
}
if(i == MAX_BACKING_STORE) {
error = ENOMEM;
goto swapon_bailout;
}
/* remember the vnode. This vnode has namei() reference */
bs_port_table[i].vp = vp;
/*
* Look to see if we are already paging to this file.
*/
/* make certain the copy send of kernel call will work */
default_pager = MEMORY_OBJECT_DEFAULT_NULL;
kr = host_default_memory_manager(host_priv_self(), &default_pager, 0);
if(kr != KERN_SUCCESS) {
error = EAGAIN;
bs_port_table[i].vp = 0;
goto swapon_bailout;
}
if (vp->v_mount->mnt_kern_flag & MNTK_SSD) {
/*
* keep the cluster size small since the
* seek cost is effectively 0 which means
* we don't care much about fragmentation
*/
dp_isssd = TRUE;
dp_cluster_size = 2 * PAGE_SIZE;
} else {
/*
* use the default cluster size
*/
dp_isssd = FALSE;
dp_cluster_size = 0;
}
kr = default_pager_backing_store_create(default_pager,
-1, /* default priority */
dp_cluster_size,
&backing_store);
memory_object_default_deallocate(default_pager);
if(kr != KERN_SUCCESS) {
error = ENOMEM;
bs_port_table[i].vp = 0;
goto swapon_bailout;
}
/* Mark this vnode as being used for swapfile */
vnode_lock_spin(vp);
SET(vp->v_flag, VSWAP);
vnode_unlock(vp);
/*
* NOTE: we are able to supply PAGE_SIZE here instead of
* an actual record size or block number because:
* a: we do not support offsets from the beginning of the
* file (allowing for non page size/record modulo offsets.
* b: because allow paging will be done modulo page size
*/
kr = default_pager_add_file(backing_store, (vnode_ptr_t) vp,
PAGE_SIZE, (int)(file_size/PAGE_SIZE));
if(kr != KERN_SUCCESS) {
bs_port_table[i].vp = 0;
if(kr == KERN_INVALID_ARGUMENT)
error = EINVAL;
else
error = ENOMEM;
/* This vnode is not to be used for swapfile */
vnode_lock_spin(vp);
CLR(vp->v_flag, VSWAP);
vnode_unlock(vp);
goto swapon_bailout;
}
bs_port_table[i].bs = (void *)backing_store;
error = 0;
ubc_setthreadcred(vp, p, current_thread());
/*
* take a long term reference on the vnode to keep
* vnreclaim() away from this vnode.
*/
vnode_ref(vp);
swapon_bailout:
if (vp) {
vnode_put(vp);
}
(void) thread_funnel_set(kernel_flock, FALSE);
AUDIT_MACH_SYSCALL_EXIT(error);
return(error);
}
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);
}
