#include <sys/ubc.h>
#include <sys/ubc_internal.h>
#include <sys/sysproto.h>
-#if CONFIG_PROTECT
-#include <sys/cprotect.h>
-#endif
#include <sys/syscall.h>
#include <sys/kdebug.h>
#include <mach/mach_vm.h>
#include <mach/vm_map.h>
#include <mach/host_priv.h>
+#include <mach/sdt.h>
#include <machine/machine_routines.h>
#include <vm/vm_pager.h>
#include <vm/vm_protos.h>
+#if CONFIG_MACF
+#include <security/mac_framework.h>
+#endif
+
/*
* XXX Internally, we use VM_PROT_* somewhat interchangeably, but the correct
* XXX usage is PROT_* from an interface perspective. Thus the values of
* Map in special device (must be SHARED) or file
*/
struct fileproc *fp;
- register struct vnode *vp;
+ struct vnode *vp;
int flags;
int prot;
int err=0;
vm_map_size_t user_size;
vm_object_offset_t pageoff;
vm_object_offset_t file_pos;
- int alloc_flags=0;
+ 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;
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;
user_size = vm_map_round_page(user_size,
vm_map_page_mask(user_map)); /* hi end */
- if ((flags & MAP_JIT) && ((flags & MAP_FIXED) || (flags & MAP_SHARED) || !(flags & MAP_ANON))){
- return EINVAL;
+ 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).
* 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 |
+ alloc_flags = fd & (VM_FLAGS_ALIAS_MASK |
+ VM_FLAGS_SUPERPAGE_MASK |
VM_FLAGS_PURGABLE);
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;
handle = (void *)vp;
#if CONFIG_MACF
error = mac_file_check_mmap(vfs_context_ucred(ctx),
- fp->f_fglob, prot, flags, &maxprot);
+ fp->f_fglob, prot, flags, file_pos, &maxprot);
if (error) {
(void)vnode_put(vp);
goto bad;
}
#endif /* MAC */
-
-#if CONFIG_PROTECT
- {
- error = cp_handle_vnop(vp, CP_READ_ACCESS | CP_WRITE_ACCESS, 0);
- if (error) {
- (void) vnode_put(vp);
- goto bad;
- }
- }
-#endif /* CONFIG_PROTECT */
-
-
}
}
if (flags & MAP_NOCACHE)
alloc_flags |= VM_FLAGS_NO_CACHE;
- if (flags & MAP_JIT){
- alloc_flags |= VM_FLAGS_MAP_JIT;
+ if (flags & MAP_JIT) {
+ vmk_flags.vmkf_map_jit = TRUE;
}
+
+ if (flags & MAP_RESILIENT_CODESIGN) {
+ alloc_flags |= VM_FLAGS_RESILIENT_CODESIGN;
+ }
+
/*
* Lookup/allocate object.
*/
map_anon_retry:
result = vm_map_enter_mem_object(user_map,
&user_addr, user_size,
- 0, alloc_flags,
+ 0, alloc_flags, vmk_flags,
+ tag,
IPC_PORT_NULL, 0, FALSE,
prot, maxprot,
(flags & MAP_SHARED) ?
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;
+ }
result = vm_map_enter_mem_object_control(user_map,
&user_addr, user_size,
- 0, alloc_flags,
+ 0, alloc_flags, vmk_flags,
+ tag,
control, file_pos,
docow, prot, maxprot,
(flags & MAP_SHARED) ?
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);
}
user_map = current_map();
addr = (mach_vm_offset_t) uap->addr;
size = (mach_vm_size_t)uap->len;
+#ifndef CONFIG_EMBEDDED
KERNEL_DEBUG_CONSTANT((BSDDBG_CODE(DBG_BSD_SC_EXTENDED_INFO, SYS_msync) | DBG_FUNC_NONE), (uint32_t)(addr >> 32), (uint32_t)(size >> 32), 0, 0, 0);
+#endif
if (addr & vm_map_page_mask(user_map)) {
/* UNIX SPEC: user address is not page-aligned, return EINVAL */
return EINVAL;
int
mprotect(__unused proc_t p, struct mprotect_args *uap, __unused int32_t *retval)
{
- register vm_prot_t prot;
+ vm_prot_t prot;
mach_vm_offset_t user_addr;
mach_vm_size_t user_size;
kern_return_t result;
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));
+ 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 */
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:
{
mach_vm_offset_t addr;
mach_vm_size_t size;
- register vm_inherit_t inherit;
+ vm_inherit_t inherit;
vm_map_t user_map;
kern_return_t result;
case MADV_CAN_REUSE:
new_behavior = VM_BEHAVIOR_CAN_REUSE;
break;
+ case MADV_PAGEOUT:
+#if MACH_ASSERT
+ new_behavior = VM_BEHAVIOR_PAGEOUT;
+ break;
+#else /* MACH_ASSERT */
+ return ENOTSUP;
+#endif /* MACH_ASSERT */
default:
return(EINVAL);
}
start = (mach_vm_offset_t) uap->addr;
size = (mach_vm_size_t) uap->len;
+#if __arm64__
+ if (start == 0 &&
+ size != 0 &&
+ (uap->behav == MADV_FREE ||
+ uap->behav == MADV_FREE_REUSABLE)) {
+ printf("** FOURK_COMPAT: %d[%s] "
+ "failing madvise(0x%llx,0x%llx,%s)\n",
+ p->p_pid, p->p_comm, start, size,
+ ((uap->behav == MADV_FREE_REUSABLE)
+ ? "MADV_FREE_REUSABLE"
+ : "MADV_FREE"));
+ DTRACE_VM3(fourk_compat_madvise,
+ uint64_t, start,
+ uint64_t, size,
+ int, uap->behav);
+ return EINVAL;
+ }
+#endif /* __arm64__ */
+
user_map = current_map();
result = mach_vm_behavior_set(user_map, start, size, new_behavior);
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;
+ mach_vm_offset_t addr = 0, first_addr = 0, end = 0, cur_end = 0;
+ vm_map_t map = VM_MAP_NULL;
+ user_addr_t vec = 0;
+ int error = 0;
+ int vecindex = 0, lastvecindex = 0;
int mincoreinfo=0;
- int pqueryinfo;
- kern_return_t ret;
- int numref;
-
- char c;
+ int pqueryinfo = 0;
+ unsigned int pqueryinfo_vec_size = 0;
+ vm_page_info_basic_t info = NULL;
+ mach_msg_type_number_t count = 0;
+ char *kernel_vec = NULL;
+ int req_vec_size_pages = 0, cur_vec_size_pages = 0;
+ kern_return_t kr = KERN_SUCCESS;
map = current_map();
*/
first_addr = addr = vm_map_trunc_page(uap->addr,
vm_map_page_mask(map));
- end = addr + vm_map_round_page(uap->len,
+ end = vm_map_round_page(uap->addr + uap->len,
vm_map_page_mask(map));
if (end < addr)
return (EINVAL);
+ if (end == addr)
+ return (0);
+
/*
- * Address of byte vector
+ * We are going to loop through the whole 'req_vec_size' pages
+ * range in chunks of 'cur_vec_size'.
*/
- vec = uap->vec;
- map = current_map();
+ req_vec_size_pages = (end - addr) >> PAGE_SHIFT;
+ cur_vec_size_pages = MIN(req_vec_size_pages, (int)(MAX_PAGE_RANGE_QUERY >> PAGE_SHIFT));
+
+ kernel_vec = (void*) _MALLOC(cur_vec_size_pages * sizeof(char), M_TEMP, M_WAITOK);
+
+ if (kernel_vec == NULL) {
+ return (ENOMEM);
+ }
/*
- * 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.
+ * Address of byte vector
*/
- 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;
+ vec = uap->vec;
+
+ pqueryinfo_vec_size = cur_vec_size_pages * sizeof(struct vm_page_info_basic);
+ info = (void*) _MALLOC(pqueryinfo_vec_size, M_TEMP, M_WAITOK);
+
+ if (info == NULL) {
+ FREE(kernel_vec, M_TEMP);
+ return (ENOMEM);
+ }
+
+ while (addr < end) {
+
+ cur_end = addr + (cur_vec_size_pages * PAGE_SIZE_64);
+
+ count = VM_PAGE_INFO_BASIC_COUNT;
+ kr = vm_map_page_range_info_internal(map,
+ addr,
+ cur_end,
+ VM_PAGE_INFO_BASIC,
+ (vm_page_info_t) info,
+ &count);
+
+ assert(kr == KERN_SUCCESS);
/*
- * If we have skipped map entries, we need to make sure that
- * the byte vector is zeroed for those skipped entries.
+ * 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.
*/
- while((lastvecindex + 1) < vecindex) {
- c = 0;
- error = copyout(&c, vec + lastvecindex, 1);
- if (error) {
- return (EFAULT);
- }
- ++lastvecindex;
+ lastvecindex = -1;
+ for( ; addr < cur_end; addr += PAGE_SIZE ) {
+
+ pqueryinfo = info[lastvecindex + 1].disposition;
+
+ 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;
+ if (pqueryinfo & VM_PAGE_QUERY_PAGE_PAGED_OUT)
+ mincoreinfo |= MINCORE_PAGED_OUT;
+ if (pqueryinfo & VM_PAGE_QUERY_PAGE_COPIED)
+ mincoreinfo |= MINCORE_COPIED;
+ if ((pqueryinfo & VM_PAGE_QUERY_PAGE_EXTERNAL) == 0)
+ mincoreinfo |= MINCORE_ANONYMOUS;
+ /*
+ * calculate index into user supplied byte vector
+ */
+ vecindex = (addr - first_addr)>> PAGE_SHIFT;
+ kernel_vec[vecindex] = (char)mincoreinfo;
+ lastvecindex = vecindex;
}
- /*
- * Pass the page information to the user
- */
- c = (char)mincoreinfo;
- error = copyout(&c, vec + vecindex, 1);
+
+ assert(vecindex == (cur_vec_size_pages - 1));
+
+ error = copyout(kernel_vec, vec, cur_vec_size_pages * sizeof(char) /* a char per page */);
+
if (error) {
- return (EFAULT);
+ break;
}
- lastvecindex = vecindex;
+
+ /*
+ * For the next chunk, we'll need:
+ * - bump the location in the user buffer for our next disposition.
+ * - new length
+ * - starting address
+ */
+ vec += cur_vec_size_pages * sizeof(char);
+ req_vec_size_pages = (end - addr) >> PAGE_SHIFT;
+ cur_vec_size_pages = MIN(req_vec_size_pages, (int)(MAX_PAGE_RANGE_QUERY >> PAGE_SHIFT));
+
+ first_addr = addr;
}
+ FREE(kernel_vec, M_TEMP);
+ FREE(info, M_TEMP);
- /*
- * 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;
+ if (error) {
+ return (EFAULT);
}
-
+
return (0);
}
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, TRUE);
+ result = vm_map_wire_kernel(user_map, addr, addr+size, VM_PROT_NONE, 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;
user_map = current_map();
/* JMM - need to remove all wirings by spec - this just removes one */
- result = mach_vm_wire(host_priv_self(), user_map, addr, size, VM_PROT_NONE);
+ result = mach_vm_wire_kernel(host_priv_self(), user_map, addr, size, VM_PROT_NONE, VM_KERN_MEMORY_MLOCK);
return (result == KERN_SUCCESS ? 0 : ENOMEM);
}
uint32_t cryptid;
cpu_type_t cputype;
cpu_subtype_t cpusubtype;
- pager_crypt_info_t crypt_info;
+ pager_crypt_info_t crypt_info;
const char * cryptname = 0;
char *vpath;
int len, ret;
kprintf("%s vpath %s cryptid 0x%08x cputype 0x%08x cpusubtype 0x%08x range 0x%016llx size 0x%016llx\n",
__FUNCTION__, vpath, cryptid, cputype, cpusubtype, (uint64_t)user_addr, (uint64_t)user_size);
#endif
-
+
/* set up decrypter first */
crypt_file_data_t crypt_data = {
.filename = vpath,
.cputype = cputype,
.cpusubtype = cpusubtype };
result = text_crypter_create(&crypt_info, cryptname, (void*)&crypt_data);
+#if VM_MAP_DEBUG_APPLE_PROTECT
+ if (vm_map_debug_apple_protect) {
+ printf("APPLE_PROTECT: %d[%s] map %p [0x%llx:0x%llx] %s(%s) -> 0x%x\n",
+ p->p_pid, p->p_comm,
+ user_map,
+ (uint64_t) user_addr,
+ (uint64_t) (user_addr + user_size),
+ __FUNCTION__, vpath, result);
+ }
+#endif /* VM_MAP_DEBUG_APPLE_PROTECT */
FREE_ZONE(vpath, MAXPATHLEN, M_NAMEI);
if(result) {
}
/* now remap using the decrypter */
- result = vm_map_apple_protected(user_map, user_addr, user_addr+user_size, &crypt_info);
+ vm_object_offset_t crypto_backing_offset;
+ crypto_backing_offset = -1; /* i.e. use map entry's offset */
+ result = vm_map_apple_protected(user_map,
+ user_addr,
+ user_addr+user_size,
+ crypto_backing_offset,
+ &crypt_info);
if (result) {
printf("%s: mapping failed with %d\n", __FUNCTION__, result);
- crypt_info.crypt_end(crypt_info.crypt_ops);
+ }
+
+ if (result) {
return (EPERM);
}
-
return 0;
}
#endif /* CONFIG_CODE_DECRYPTION */
projects / apple / xnu.git / blobdiff