]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/kern/mach_loader.c
xnu-3248.60.10.tar.gz
[apple/xnu.git] / bsd / kern / mach_loader.c
index 47253a898b268ca575643d8055bdfa6442133395..d4f084e31169a178eb71f03042774388db6e6ea1 100644 (file)
@@ -49,6 +49,7 @@
 #include <sys/fcntl.h>
 #include <sys/ubc_internal.h>
 #include <sys/imgact.h>
+#include <sys/codesign.h>
 
 #include <mach/mach_types.h>
 #include <mach/vm_map.h>       /* vm_allocate() */
 #include <vm/vm_pager.h>
 #include <vm/vnode_pager.h>
 #include <vm/vm_protos.h> 
+#include <IOKit/IOReturn.h>    /* for kIOReturnNotPrivileged */
 
-/*
- * XXX vm/pmap.h should not treat these prototypes as MACH_KERNEL_PRIVATE
- * when KERNEL is defined.
- */
-extern pmap_t  pmap_create(vm_map_size_t size, boolean_t is_64bit);
-extern void    pmap_switch(pmap_t);
+#include <os/overflow.h>
 
 /*
- * XXX kern/thread.h should not treat these prototypes as MACH_KERNEL_PRIVATE
+ * XXX vm/pmap.h should not treat these prototypes as MACH_KERNEL_PRIVATE
  * when KERNEL is defined.
  */
-extern kern_return_t   thread_setstatus(thread_t thread, int flavor,
-                               thread_state_t tstate,
-                               mach_msg_type_number_t count);
-
-extern kern_return_t    thread_state_initialize(thread_t thread);
-
+extern pmap_t  pmap_create(ledger_t ledger, vm_map_size_t size,
+                               boolean_t is_64bit);
 
 /* XXX should have prototypes in a shared header file */
 extern int     get_map_nentries(vm_map_t);
@@ -109,17 +102,25 @@ static load_result_t load_result_null = {
        .mach_header = MACH_VM_MIN_ADDRESS,
        .entry_point = MACH_VM_MIN_ADDRESS,
        .user_stack = MACH_VM_MIN_ADDRESS,
+       .user_stack_size = 0,
        .all_image_info_addr = MACH_VM_MIN_ADDRESS,
        .all_image_info_size = 0,
        .thread_count = 0,
        .unixproc = 0,
        .dynlinker = 0,
-       .customstack = 0,
+       .needs_dynlinker = 0,
+       .prog_allocated_stack = 0,
+       .prog_stack_size = 0,
        .validentry = 0,
+       .using_lcmain = 0,
        .csflags = 0,
+       .has_pagezero = 0,
        .uuid = { 0 },
        .min_vm_addr = MACH_VM_MAX_ADDRESS,
-       .max_vm_addr = MACH_VM_MIN_ADDRESS
+       .max_vm_addr = MACH_VM_MIN_ADDRESS,
+       .cs_end_offset = 0,
+       .threadstate = NULL,
+       .threadstate_sz = 0
 };
 
 /*
@@ -135,6 +136,7 @@ parse_machfile(
        off_t                   macho_size,
        int                     depth,
        int64_t                 slide,
+       int64_t                 dyld_slide,     
        load_result_t           *result
 );
 
@@ -151,6 +153,13 @@ load_segment(
        load_result_t                   *result
 );
 
+static load_return_t
+load_uuid(
+       struct uuid_command             *uulp,
+       char                            *command_end,
+       load_result_t                   *result
+);
+
 static load_return_t
 load_code_signature(
        struct linkedit_data_command    *lcp,
@@ -166,9 +175,22 @@ set_code_unprotect(
        struct encryption_info_command  *lcp,
        caddr_t                         addr,
        vm_map_t                        map,
-       struct vnode                    *vp);
+       int64_t                         slide,
+       struct vnode                    *vp,
+       off_t                           macho_offset,
+       cpu_type_t                      cputype,
+       cpu_subtype_t                   cpusubtype);
 #endif
 
+static
+load_return_t
+load_main(
+       struct entry_point_command      *epc,
+       thread_t                thread,
+       int64_t                         slide,
+       load_result_t           *result
+);
+
 static load_return_t
 load_unixthread(
        struct thread_command   *tcp,
@@ -181,7 +203,8 @@ static load_return_t
 load_threadstate(
        thread_t                thread,
        uint32_t        *ts,
-       uint32_t        total_size
+       uint32_t        total_size,
+       load_result_t *
 );
 
 static load_return_t
@@ -270,35 +293,40 @@ note_all_image_info_section(const struct segment_command_64 *scp,
        }
 }
 
+
 load_return_t
 load_machfile(
        struct image_params     *imgp,
        struct mach_header      *header,
        thread_t                thread,
-       vm_map_t                new_map,
+       vm_map_t                *mapp,
        load_result_t           *result
 )
 {
        struct vnode            *vp = imgp->ip_vp;
        off_t                   file_offset = imgp->ip_arch_offset;
        off_t                   macho_size = imgp->ip_arch_size;
-       
+       off_t                   file_size = imgp->ip_vattr->va_data_size;
+       vm_map_t                new_map = *mapp;
        pmap_t                  pmap = 0;       /* protected by create_map */
        vm_map_t                map;
-       vm_map_t                old_map;
-       task_t                  old_task = TASK_NULL; /* protected by create_map */
        load_result_t           myresult;
        load_return_t           lret;
        boolean_t create_map = FALSE;
+       boolean_t enforce_hard_pagezero = TRUE;
        int spawn = (imgp->ip_flags & IMGPF_SPAWN);
        task_t task = current_task();
        proc_t p = current_proc();
        mach_vm_offset_t        aslr_offset = 0;
+       mach_vm_offset_t        dyld_aslr_offset = 0;
        kern_return_t           kret;
 
+       if (macho_size > file_size) {
+               return(LOAD_BADMACHO);
+       }
+
        if (new_map == VM_MAP_NULL) {
                create_map = TRUE;
-               old_task = current_task();
        }
 
        /*
@@ -309,24 +337,38 @@ load_machfile(
         */
        if (spawn) {
                create_map = TRUE;
-               old_task = get_threadtask(thread);
        }
 
        if (create_map) {
-               pmap = pmap_create((vm_map_size_t) 0, (imgp->ip_flags & IMGPF_IS_64BIT));
+               task_t ledger_task;
+               if (imgp->ip_new_thread) {
+                       ledger_task = get_threadtask(imgp->ip_new_thread);
+               } else {
+                       ledger_task = task;
+               }
+               pmap = pmap_create(get_task_ledger(ledger_task),
+                                  (vm_map_size_t) 0,
+                                  ((imgp->ip_flags & IMGPF_IS_64BIT) != 0));
                pal_switch_pmap(thread, pmap, imgp->ip_flags & IMGPF_IS_64BIT);
                map = vm_map_create(pmap,
                                0,
-                               vm_compute_max_offset((imgp->ip_flags & IMGPF_IS_64BIT)),
+                               vm_compute_max_offset(((imgp->ip_flags & IMGPF_IS_64BIT) == IMGPF_IS_64BIT)),
                                TRUE);
-
        } else
                map = new_map;
 
+#if   (__ARM_ARCH_7K__ >= 2) && defined(PLATFORM_WatchOS)
+       /* enforce 16KB alignment for watch targets with new ABI */
+       vm_map_set_page_shift(map, SIXTEENK_PAGE_SHIFT);
+#endif /* __arm64__ */
+
 #ifndef        CONFIG_ENFORCE_SIGNED_CODE
-       /* This turns off faulting for executable pages, which allows to 
-        * circumvent Code Signing Enforcement */
-       if ( (header->flags & MH_ALLOW_STACK_EXECUTION) )
+       /* This turns off faulting for executable pages, which allows
+        * to circumvent Code Signing Enforcement. The per process
+        * flag (CS_ENFORCEMENT) is not set yet, but we can use the
+        * global flag.
+        */
+       if ( !cs_enforcement(NULL) && (header->flags & MH_ALLOW_STACK_EXECUTION) )
                vm_map_disable_NX(map);
 #endif
 
@@ -336,12 +378,20 @@ load_machfile(
                vm_map_disallow_data_exec(map);
        
        /*
-        * Compute a random offset for ASLR.
+        * Compute a random offset for ASLR, and an independent random offset for dyld.
         */
        if (!(imgp->ip_flags & IMGPF_DISABLE_ASLR)) {
+               uint64_t max_slide_pages;
+
+               max_slide_pages = vm_map_get_max_aslr_slide_pages(map);
+
                aslr_offset = random();
-               aslr_offset %= 1 << ((imgp->ip_flags & IMGPF_IS_64BIT) ? 16 : 8);
-               aslr_offset <<= PAGE_SHIFT;
+               aslr_offset %= max_slide_pages;
+               aslr_offset <<= vm_map_page_shift(map);
+
+               dyld_aslr_offset = random();
+               dyld_aslr_offset %= max_slide_pages;
+               dyld_aslr_offset <<= vm_map_page_shift(map);
        }
        
        if (!result)
@@ -350,7 +400,7 @@ load_machfile(
        *result = load_result_null;
 
        lret = parse_machfile(vp, map, thread, header, file_offset, macho_size,
-                             0, (int64_t)aslr_offset, result);
+                             0, (int64_t)aslr_offset, (int64_t)dyld_aslr_offset, result);
 
        if (lret != LOAD_SUCCESS) {
                if (create_map) {
@@ -359,32 +409,34 @@ load_machfile(
                return(lret);
        }
 
+#if __x86_64__
        /*
-        * For 64-bit users, check for presence of a 4GB page zero
-        * which will enable the kernel to share the user's address space
-        * and hence avoid TLB flushes on kernel entry/exit
-        */ 
-
-       if ((imgp->ip_flags & IMGPF_IS_64BIT) &&
-            vm_map_has_4GB_pagezero(map))
-               vm_map_set_4GB_pagezero(map);
-
-       /*
-        *      Commit to new map.
-        *
-        *      Swap the new map for the old, which  consumes our new map
-        *      reference but each leaves us responsible for the old_map reference.
-        *      That lets us get off the pmap associated with it, and
-        *      then we can release it.
+        * On x86, for compatibility, don't enforce the hard page-zero restriction for 32-bit binaries.
         */
+       if ((imgp->ip_flags & IMGPF_IS_64BIT) == 0) {
+               enforce_hard_pagezero = FALSE;
+       }
+#endif
+       /*
+        * Check to see if the page zero is enforced by the map->min_offset.
+        */ 
+       if (enforce_hard_pagezero &&
+           (vm_map_has_hard_pagezero(map, 0x1000) == FALSE)) {
+               {
+                       if (create_map) {
+                               vm_map_deallocate(map); /* will lose pmap reference too */
+                       }
+                       return (LOAD_BADMACHO);
+               }
+       }
 
-        if (create_map) {
+       if (create_map) {
                /*
                 * If this is an exec, then we are going to destroy the old
                 * task, and it's correct to halt it; if it's spawn, the
                 * task is not yet running, and it makes no sense.
                 */
-               if (!spawn) {
+               if (!spawn) {
                        /*
                         * Mark the task as halting and start the other
                         * threads towards terminating themselves.  Then
@@ -396,30 +448,39 @@ load_machfile(
                         *
                         * NOTE: task_start_halt() makes sure that no new
                         * threads are created in the task during the transition.
-                        * We need to mark the workqueue as exiting before we
-                        * wait for threads to terminate (at the end of which
-                        * we no longer have a prohibition on thread creation).
-                        * 
-                        * Finally, clean up any lingering workqueue data structures
-                        * that may have been left behind by the workqueue threads
-                        * as they exited (and then clean up the work queue itself).
-                        */
-                       kret = task_start_halt(task);
-                       if (kret != KERN_SUCCESS) {
-                               return(kret);           
-                       }
-                       proc_transcommit(p, 0);
-                       workqueue_mark_exiting(p);
-                       task_complete_halt(task);
-                       workqueue_exit(p);
-               }
-               old_map = swap_task_map(old_task, thread, map, !spawn);
-               vm_map_clear_4GB_pagezero(old_map);
-               vm_map_deallocate(old_map);
+                        * We need to mark the workqueue as exiting before we
+                        * wait for threads to terminate (at the end of which
+                        * we no longer have a prohibition on thread creation).
+                        * 
+                        * Finally, clean up any lingering workqueue data structures
+                        * that may have been left behind by the workqueue threads
+                        * as they exited (and then clean up the work queue itself).
+                        */
+                       kret = task_start_halt(task);
+                       if (kret != KERN_SUCCESS) {
+                               vm_map_deallocate(map); /* will lose pmap reference too */
+                               return (LOAD_FAILURE);
+                       }
+                       proc_transcommit(p, 0);
+                       workqueue_mark_exiting(p);
+                       task_complete_halt(task);
+                       workqueue_exit(p);
+                       kqueue_dealloc(p->p_wqkqueue);
+                       p->p_wqkqueue = NULL;
+               }
+               *mapp = map;
        }
        return(LOAD_SUCCESS);
 }
 
+int macho_printf = 0;
+#define MACHO_PRINTF(args)                             \
+       do {                                            \
+               if (macho_printf) {                     \
+                       printf args;                    \
+               }                                       \
+       } while (0)
+
 /*
  * The file size of a mach-o file is limited to 32 bits; this is because
  * this is the limit on the kalloc() of enough bytes for a mach_header and
@@ -442,13 +503,13 @@ parse_machfile(
        off_t                   macho_size,
        int                     depth,
        int64_t                 aslr_offset,
+       int64_t                 dyld_aslr_offset,
        load_result_t           *result
 )
 {
        uint32_t                ncmds;
        struct load_command     *lcp;
        struct dylinker_command *dlp = 0;
-       struct uuid_command     *uulp = 0;
        integer_t               dlarchbits = 0;
        void *                  control;
        load_return_t           ret = LOAD_SUCCESS;
@@ -460,7 +521,7 @@ parse_machfile(
        int                     pass;
        proc_t                  p = current_proc();             /* XXXX */
        int                     error;
-       int resid=0;
+       int                     resid = 0;
        size_t                  mach_header_sz = sizeof(struct mach_header);
        boolean_t               abi64;
        boolean_t               got_code_signatures = FALSE;
@@ -474,7 +535,7 @@ parse_machfile(
        /*
         *      Break infinite recursion
         */
-       if (depth > 6) {
+       if (depth > 1) {
                return(LOAD_FAILURE);
        }
 
@@ -483,7 +544,7 @@ parse_machfile(
        /*
         *      Check to see if right machine type.
         */
-       if (((cpu_type_t)(header->cputype & ~CPU_ARCH_MASK) != cpu_type()) ||
+       if (((cpu_type_t)(header->cputype & ~CPU_ARCH_MASK) != (cpu_type() & ~CPU_ARCH_MASK)) ||
            !grade_binary(header->cputype, 
                header->cpusubtype & ~CPU_SUBTYPE_MASK))
                return(LOAD_BADARCH);
@@ -492,21 +553,12 @@ parse_machfile(
                
        switch (header->filetype) {
        
-       case MH_OBJECT:
        case MH_EXECUTE:
-       case MH_PRELOAD:
                if (depth != 1) {
                        return (LOAD_FAILURE);
                }
-               break;
-               
-       case MH_FVMLIB:
-       case MH_DYLIB:
-               if (depth == 1) {
-                       return (LOAD_FAILURE);
-               }
-               break;
 
+               break;
        case MH_DYLINKER:
                if (depth != 2) {
                        return (LOAD_FAILURE);
@@ -549,11 +601,18 @@ parse_machfile(
        error = vn_rdwr(UIO_READ, vp, addr, size, file_offset,
            UIO_SYSSPACE, 0, kauth_cred_get(), &resid, p);
        if (error) {
-               if (kl_addr )
+               if (kl_addr)
                        kfree(kl_addr, kl_size);
                return(LOAD_IOERROR);
        }
 
+       if (resid) {
+               /* We must be able to read in as much as the mach_header indicated */
+               if (kl_addr)
+                       kfree(kl_addr, kl_size);
+               return(LOAD_BADMACHO);
+       }
+
        /*
         *      For PIE and dyld, slide everything by the ASLR offset.
         */
@@ -561,21 +620,32 @@ parse_machfile(
                slide = aslr_offset;
        }
 
-       /*
-        *      Scan through the commands, processing each one as necessary.
+        /*
+        *  Scan through the commands, processing each one as necessary.
+        *  We parse in three passes through the headers:
+        *  0: determine if TEXT and DATA boundary can be page-aligned
+        *  1: thread state, uuid, code signature
+        *  2: segments
+        *  3: dyld, encryption, check entry point
         */
-       for (pass = 1; pass <= 3; pass++) {
+       
+       for (pass = 0; pass <= 3; pass++) {
+
+               if (pass == 0) {
+                       /* see if we need to adjust the slide to re-align... */
+                       /* no re-alignment needed on X86_64 or ARM32 kernel */
+                       continue;
+               } else if (pass == 1) {
+               }
 
-#if CONFIG_EMBEDDED
                /*
                 * Check that the entry point is contained in an executable segments
                 */ 
-               if ((pass == 3) && (result->validentry == 0)) {
+               if ((pass == 3) && (!result->using_lcmain && result->validentry == 0)) {
                        thread_state_initialize(thread);
                        ret = LOAD_FAILURE;
                        break;
                }
-#endif
 
                /*
                 * Loop through each of the load_commands indicated by the
@@ -615,18 +685,54 @@ parse_machfile(
                         */
                        switch(lcp->cmd) {
                        case LC_SEGMENT:
+                               if (pass == 0) {
+                                       break;
+                               }
+
+                               if (pass != 2)
+                                       break;
+
+                               if (abi64) {
+                                       /*
+                                        * Having an LC_SEGMENT command for the
+                                        * wrong ABI is invalid <rdar://problem/11021230>
+                                        */
+                                       ret = LOAD_BADMACHO;
+                                       break;
+                               }
+
+                               ret = load_segment(lcp,
+                                                  header->filetype,
+                                                  control,
+                                                  file_offset,
+                                                  macho_size,
+                                                  vp,
+                                                  map,
+                                                  slide,
+                                                  result);
+                               break;
                        case LC_SEGMENT_64:
                                if (pass != 2)
                                        break;
+
+                               if (!abi64) {
+                                       /*
+                                        * Having an LC_SEGMENT_64 command for the
+                                        * wrong ABI is invalid <rdar://problem/11021230>
+                                        */
+                                       ret = LOAD_BADMACHO;
+                                       break;
+                               }
+
                                ret = load_segment(lcp,
-                                                  header->filetype,
-                                                  control,
-                                                  file_offset,
-                                                  macho_size,
-                                                  vp,
-                                                  map,
-                                                  slide,
-                                                  result);
+                                                  header->filetype,
+                                                  control,
+                                                  file_offset,
+                                                  macho_size,
+                                                  vp,
+                                                  map,
+                                                  slide,
+                                                  result);
                                break;
                        case LC_UNIXTHREAD:
                                if (pass != 1)
@@ -637,6 +743,17 @@ parse_machfile(
                                                 slide,
                                                 result);
                                break;
+                       case LC_MAIN:
+                               if (pass != 1)
+                                       break;
+                               if (depth != 1)
+                                       break;
+                               ret = load_main(
+                                                (struct entry_point_command *) lcp,
+                                                thread,
+                                                slide,
+                                                result);
+                               break;
                        case LC_LOAD_DYLINKER:
                                if (pass != 3)
                                        break;
@@ -649,8 +766,9 @@ parse_machfile(
                                break;
                        case LC_UUID:
                                if (pass == 1 && depth == 1) {
-                                       uulp = (struct uuid_command *)lcp;
-                                       memcpy(&result->uuid[0], &uulp->uuid[0], sizeof(result->uuid));
+                                       ret = load_uuid((struct uuid_command *) lcp,
+                                                       (char *)addr + mach_header_sz + header->sizeofcmds,
+                                                       result);
                                }
                                break;
                        case LC_CODE_SIGNATURE:
@@ -667,30 +785,80 @@ parse_machfile(
                                        file_offset,
                                        macho_size,
                                        header->cputype,
-                                       (depth == 1) ? result : NULL);
+                                       result);
                                if (ret != LOAD_SUCCESS) {
                                        printf("proc %d: load code signature error %d "
                                               "for file \"%s\"\n",
                                               p->p_pid, ret, vp->v_name);
-                                       ret = LOAD_SUCCESS; /* ignore error */
+                                       /*
+                                        * Allow injections to be ignored on devices w/o enforcement enabled
+                                        */
+                                       if (!cs_enforcement(NULL))
+                                           ret = LOAD_SUCCESS; /* ignore error */
+
                                } else {
                                        got_code_signatures = TRUE;
                                }
+
+                               if (got_code_signatures) {
+                                       unsigned tainted = CS_VALIDATE_TAINTED;
+                                       boolean_t valid = FALSE;
+                                       struct cs_blob *blobs;
+                                       vm_size_t off = 0;
+
+
+                                       if (cs_debug > 10)
+                                               printf("validating initial pages of %s\n", vp->v_name);
+                                       blobs = ubc_get_cs_blobs(vp);
+                                       
+                                       while (off < size && ret == LOAD_SUCCESS) {
+                                            tainted = CS_VALIDATE_TAINTED;
+
+                                            valid = cs_validate_page(blobs,
+                                                                     NULL,
+                                                                     file_offset + off,
+                                                                     addr + off,
+                                                                     &tainted);
+                                            if (!valid || (tainted & CS_VALIDATE_TAINTED)) {
+                                                    if (cs_debug)
+                                                            printf("CODE SIGNING: %s[%d]: invalid initial page at offset %lld validated:%d tainted:%d csflags:0x%x\n", 
+                                                                   vp->v_name, p->p_pid, (long long)(file_offset + off), valid, tainted, result->csflags);
+                                                    if (cs_enforcement(NULL) ||
+                                                        (result->csflags & (CS_HARD|CS_KILL|CS_ENFORCEMENT))) {
+                                                            ret = LOAD_FAILURE;
+                                                    }
+                                                    result->csflags &= ~CS_VALID;
+                                            }
+                                            off += PAGE_SIZE;
+                                       }
+                               }
+
                                break;
 #if CONFIG_CODE_DECRYPTION
                        case LC_ENCRYPTION_INFO:
+                       case LC_ENCRYPTION_INFO_64:
                                if (pass != 3)
                                        break;
                                ret = set_code_unprotect(
                                        (struct encryption_info_command *) lcp,
-                                       addr, map, vp);
+                                       addr, map, slide, vp, file_offset,
+                                       header->cputype, header->cpusubtype);
                                if (ret != LOAD_SUCCESS) {
                                        printf("proc %d: set_code_unprotect() error %d "
                                               "for file \"%s\"\n",
                                               p->p_pid, ret, vp->v_name);
-                                       /* Don't let the app run if it's 
+                                       /* 
+                                        * Don't let the app run if it's 
                                         * encrypted but we failed to set up the
-                                        * decrypter */
+                                        * decrypter. If the keys are missing it will
+                                        * return LOAD_DECRYPTFAIL.
+                                        */
+                                        if (ret == LOAD_DECRYPTFAIL) {
+                                               /* failed to load due to missing FP keys */
+                                               proc_lock(p);
+                                               p->p_lflag |= P_LTERM_DECRYPTFAIL;
+                                               proc_unlock(p);
+                                        }
                                         psignal(p, SIGKILL);
                                }
                                break;
@@ -706,27 +874,53 @@ parse_machfile(
                if (ret != LOAD_SUCCESS)
                        break;
        }
-       if (ret == LOAD_SUCCESS) { 
-           if (! got_code_signatures) {
-                   struct cs_blob *blob;
-                   /* no embedded signatures: look for detached ones */
-                   blob = ubc_cs_blob_get(vp, -1, file_offset);
-                   if (blob != NULL) {
-                           /* get flags to be applied to the process */
-                           result->csflags |= blob->csb_flags;
-                   }
-           }
 
-           if (dlp != 0) {
-                   /* load the dylinker, and always slide it by the ASLR
-                    * offset regardless of PIE */
-                   ret = load_dylinker(dlp, dlarchbits, map, thread, depth, aslr_offset, result);
-           }
+       if (ret == LOAD_SUCCESS) { 
+               if (! got_code_signatures) {
+                       if (cs_enforcement(NULL)) {
+                               ret = LOAD_FAILURE;
+                       } else {
+                               /*
+                                * No embedded signatures: look for detached by taskgated,
+                                * this is only done on OSX, on embedded platforms we expect everything
+                                * to be have embedded signatures.
+                                */
+                               struct cs_blob *blob;
+
+                               blob = ubc_cs_blob_get(vp, -1, file_offset);
+                               if (blob != NULL) {
+                                       unsigned int cs_flag_data = blob->csb_flags;
+                                       if(0 != ubc_cs_generation_check(vp)) {
+                                               if (0 != ubc_cs_blob_revalidate(vp, blob, 0)) {
+                                                       /* clear out the flag data if revalidation fails */
+                                                       cs_flag_data = 0;
+                                                       result->csflags &= ~CS_VALID;
+                                               }
+                                       }
+                                       /* get flags to be applied to the process */
+                                       result->csflags |= cs_flag_data;
+                               }
+                       }
+               }
 
-           if(depth == 1) {
-               if (result->thread_count == 0) {
+               /* Make sure if we need dyld, we got it */
+               if (result->needs_dynlinker && !dlp) {
                        ret = LOAD_FAILURE;
                }
+
+               if ((ret == LOAD_SUCCESS) && (dlp != 0)) {
+                       /*
+                        * load the dylinker, and slide it by the independent DYLD ASLR
+                        * offset regardless of the PIE-ness of the main binary.
+                        */
+                       ret = load_dylinker(dlp, dlarchbits, map, thread, depth,
+                                           dyld_aslr_offset, result);
+               }
+               
+               if((ret == LOAD_SUCCESS) && (depth == 1)) {
+                       if (result->thread_count == 0) {
+                               ret = LOAD_FAILURE;
+                       }
            }
        }
 
@@ -738,10 +932,10 @@ parse_machfile(
 
 #if CONFIG_CODE_DECRYPTION
 
-#define        APPLE_UNPROTECTED_HEADER_SIZE   (3 * PAGE_SIZE_64)
+#define        APPLE_UNPROTECTED_HEADER_SIZE   (3 * 4096)
 
 static load_return_t
-unprotect_segment(
+unprotect_dsmos_segment(
        uint64_t        file_off,
        uint64_t        file_size,
        struct vnode    *vp,
@@ -781,9 +975,28 @@ unprotect_segment(
                crypt_info.crypt_end = NULL;
 #pragma unused(vp, macho_offset)
                crypt_info.crypt_ops = (void *)0x2e69cf40;
+               vm_map_offset_t crypto_backing_offset;
+               crypto_backing_offset = -1; /* i.e. use map entry's offset */
+#if DEVELOPMENT || DEBUG
+               struct proc *p;
+               p = current_proc();
+               printf("APPLE_PROTECT: %d[%s] map %p [0x%llx:0x%llx] %s(%s)\n",
+                      p->p_pid, p->p_comm, map,
+                      (uint64_t) map_addr, (uint64_t) (map_addr + map_size),
+                      __FUNCTION__, vp->v_name);
+#endif /* DEVELOPMENT || DEBUG */
+
+               /* The DSMOS pager can only be used by apple signed code */
+               struct cs_blob * blob = csvnode_get_blob(vp, file_off);
+               if( blob == NULL || !blob->csb_platform_binary || blob->csb_platform_path)
+               {
+                       return LOAD_FAILURE;
+               }
+
                kr = vm_map_apple_protected(map,
                                            map_addr,
                                            map_addr + map_size,
+                                           crypto_backing_offset,
                                            &crypt_info);
        }
 
@@ -794,7 +1007,7 @@ unprotect_segment(
 }
 #else  /* CONFIG_CODE_DECRYPTION */
 static load_return_t
-unprotect_segment(
+unprotect_dsmos_segment(
        __unused        uint64_t        file_off,
        __unused        uint64_t        file_size,
        __unused        struct vnode    *vp,
@@ -807,29 +1020,166 @@ unprotect_segment(
 }
 #endif /* CONFIG_CODE_DECRYPTION */
 
+
+/*
+ * map_segment:
+ *     Maps a Mach-O segment, taking care of mis-alignment (wrt the system
+ *     page size) issues.
+ * 
+ *     The mapping might result in 1, 2 or 3 map entries:
+ *     1. for the first page, which could be overlap with the previous
+ *        mapping,
+ *     2. for the center (if applicable),
+ *     3. for the last page, which could overlap with the next mapping.
+ *
+ *     For each of those map entries, we might have to interpose a
+ *     "fourk_pager" to deal with mis-alignment wrt the system page size,
+ *     either in the mapping address and/or size or the file offset and/or
+ *     size.
+ *     The "fourk_pager" itself would be mapped with proper alignment
+ *     wrt the system page size and would then be populated with the
+ *     information about the intended mapping, with a "4KB" granularity.
+ */
+static kern_return_t
+map_segment(
+       vm_map_t                map,
+       vm_map_offset_t         vm_start,
+       vm_map_offset_t         vm_end,
+       memory_object_control_t control,
+       vm_map_offset_t         file_start,
+       vm_map_offset_t         file_end,
+       vm_prot_t               initprot,
+       vm_prot_t               maxprot)
+{
+       int             extra_vm_flags, cur_extra_vm_flags;
+       vm_map_offset_t cur_offset, cur_start, cur_end;
+       kern_return_t   ret;
+       vm_map_offset_t effective_page_mask;
+       
+       if (vm_end < vm_start ||
+           file_end < file_start) {
+               return LOAD_BADMACHO;
+       }
+       if (vm_end == vm_start ||
+           file_end == file_start) {
+               /* nothing to map... */
+               return LOAD_SUCCESS;
+       }
+
+       effective_page_mask = MAX(PAGE_MASK, vm_map_page_mask(map));
+
+       extra_vm_flags = 0;
+       if (vm_map_page_aligned(vm_start, effective_page_mask) &&
+           vm_map_page_aligned(vm_end, effective_page_mask) &&
+           vm_map_page_aligned(file_start, effective_page_mask) &&
+           vm_map_page_aligned(file_end, effective_page_mask)) {
+               /* all page-aligned and map-aligned: proceed */
+       } else {
+               panic("map_segment: unexpected mis-alignment "
+                     "vm[0x%llx:0x%llx] file[0x%llx:0x%llx]\n",
+                     (uint64_t) vm_start,
+                     (uint64_t) vm_end,
+                     (uint64_t) file_start,
+                     (uint64_t) file_end);
+       }
+
+       cur_offset = 0;
+       cur_start = vm_start;
+       cur_end = vm_start;
+       if (cur_end >= vm_start + (file_end - file_start)) {
+               /* all mapped: done */
+               goto done;
+       }
+       if (vm_map_round_page(cur_end, effective_page_mask) >=
+           vm_map_trunc_page(vm_start + (file_end - file_start),
+                             effective_page_mask)) {
+               /* no middle */
+       } else {
+               cur_start = cur_end;
+               if ((vm_start & effective_page_mask) !=
+                   (file_start & effective_page_mask)) {
+                       /* one 4K pager for the middle */
+                       cur_extra_vm_flags = extra_vm_flags;
+               } else {
+                       /* regular mapping for the middle */
+                       cur_extra_vm_flags = 0;
+               }
+               cur_end = vm_map_trunc_page(vm_start + (file_end -
+                                                       file_start),
+                                           effective_page_mask);
+               if (control != MEMORY_OBJECT_CONTROL_NULL) {
+                       ret = vm_map_enter_mem_object_control(
+                               map,
+                               &cur_start,
+                               cur_end - cur_start,
+                               (mach_vm_offset_t)0,
+                               VM_FLAGS_FIXED | cur_extra_vm_flags,
+                               control,
+                               file_start + cur_offset,
+                               TRUE, /* copy */
+                               initprot, maxprot,
+                               VM_INHERIT_DEFAULT);
+               } else {
+                       ret = vm_map_enter_mem_object(
+                               map,
+                               &cur_start,
+                               cur_end - cur_start,
+                               (mach_vm_offset_t)0,
+                               VM_FLAGS_FIXED | cur_extra_vm_flags,
+                               IPC_PORT_NULL,
+                               0, /* offset */
+                               TRUE, /* copy */
+                               initprot, maxprot,
+                               VM_INHERIT_DEFAULT);
+               }
+               if (ret != KERN_SUCCESS) {
+                       return (LOAD_NOSPACE);
+               }
+               cur_offset += cur_end - cur_start;
+       }
+       if (cur_end >= vm_start + (file_end - file_start)) {
+               /* all mapped: done */
+               goto done;
+       }
+       cur_start = cur_end;
+done:
+       assert(cur_end >= vm_start + (file_end - file_start));
+       return LOAD_SUCCESS;
+}
+
 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 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;
-       mach_vm_offset_t        map_addr, map_offset;
-       mach_vm_size_t          map_size, seg_size, delta_size;
+       vm_map_size_t           delta_size;
        vm_prot_t               initprot;
        vm_prot_t               maxprot;
        size_t                  segment_command_size, total_section_size,
                                single_section_size;
-       
+       vm_map_offset_t         file_offset, file_size;
+       vm_map_offset_t         vm_offset, vm_size;
+       vm_map_offset_t         vm_start, vm_end, vm_end_aligned;
+       vm_map_offset_t         file_start, file_end;
+       kern_return_t           kr;
+       boolean_t               verbose;
+       vm_map_size_t           effective_page_size;
+       vm_map_offset_t         effective_page_mask;
+
+       effective_page_size = MAX(PAGE_SIZE, vm_map_page_size(map));
+       effective_page_mask = MAX(PAGE_MASK, vm_map_page_mask(map));
+
+       verbose = FALSE;
        if (LC_SEGMENT_64 == lcp->cmd) {
                segment_command_size = sizeof(struct segment_command_64);
                single_section_size  = sizeof(struct section_64);
@@ -841,43 +1191,80 @@ load_segment(
                return (LOAD_BADMACHO);
        total_section_size = lcp->cmdsize - segment_command_size;
 
-       if (LC_SEGMENT_64 == lcp->cmd)
+       if (LC_SEGMENT_64 == lcp->cmd) {
                scp = (struct segment_command_64 *)lcp;
-       else {
+       else {
                scp = &segment_command;
                widen_segment_command((struct segment_command *)lcp, scp);
        }
 
+       if (verbose) {
+               MACHO_PRINTF(("+++ load_segment %s "
+                             "vm[0x%llx:0x%llx] file[0x%llx:0x%llx] "
+                             "prot %d/%d flags 0x%x\n",
+                             scp->segname,
+                             (uint64_t)(slide + scp->vmaddr),
+                             (uint64_t)(slide + scp->vmaddr + scp->vmsize),
+                             pager_offset + scp->fileoff,
+                             pager_offset + scp->fileoff + scp->filesize,
+                             scp->initprot,
+                             scp->maxprot,
+                             scp->flags));
+       }
+
        /*
         * 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)
+           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)
+       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)
+       file_offset = pager_offset + scp->fileoff;      /* limited to 32 bits */
+       file_size = scp->filesize;
+       if ((file_offset & PAGE_MASK_64) != 0 ||
+               /* we can't mmap() it if it's not page-aligned in the file */
+           (file_offset & vm_map_page_mask(map)) != 0) {
+               /*
+                * The 1st test would have failed if the system's page size
+                * was what this process believe is the page size, so let's
+                * fail here too for the sake of consistency.
+                */
                return (LOAD_BADMACHO);
+       }
 
        /*
-        *      Round sizes to page size.
+        * If we have a code signature attached for this slice
+        * require that the segments are within the signed part
+        * of the file.
         */
-       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 */
-       if (seg_size == 0)
-               return (KERN_SUCCESS);
-       if (map_addr == 0 &&
-           map_size == 0 &&
-           seg_size != 0 &&
+       if (result->cs_end_offset &&
+           result->cs_end_offset < (off_t)scp->fileoff &&
+           result->cs_end_offset - scp->fileoff < scp->filesize)
+        {
+               if (cs_debug)
+                       printf("section outside code signature\n");
+               return LOAD_BADMACHO;
+       }
+
+       vm_offset = scp->vmaddr + slide;
+       vm_size = scp->vmsize;
+
+       if (vm_size == 0)
+               return (LOAD_SUCCESS);
+       if (scp->vmaddr == 0 &&
+           file_size == 0 &&
+           vm_size != 0 &&
            (scp->initprot & VM_PROT_ALL) == VM_PROT_NONE &&
            (scp->maxprot & VM_PROT_ALL) == VM_PROT_NONE) {
                /*
@@ -886,65 +1273,96 @@ load_segment(
                 * 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) {
-                       /*
-                        * 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);
+               /*
+                * 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.
+                */
+               vm_end = vm_offset + vm_size;
+               if (vm_end < vm_offset) {
+                       return (LOAD_BADMACHO);
+               }
+               if (verbose) {
+                       MACHO_PRINTF(("++++++ load_segment: "
+                                     "page_zero up to 0x%llx\n",
+                                     (uint64_t) vm_end));
+               }
+               {
+                       vm_end = vm_map_round_page(vm_end,
+                                                  PAGE_MASK_64);
+                       vm_end_aligned = vm_end;
                }
+               ret = vm_map_raise_min_offset(map,
+                                             vm_end_aligned);
+                       
+               if (ret != KERN_SUCCESS) {
+                       return (LOAD_FAILURE);
+               }
+               return (LOAD_SUCCESS);
+       } else {
        }
 
-       /* If a non-zero slide was specified by the caller, apply now */
-       map_addr += slide;
+       {
+               file_start = vm_map_trunc_page(file_offset,
+                                              effective_page_mask);
+               file_end = vm_map_round_page(file_offset + file_size,
+                                            effective_page_mask);
+               vm_start = vm_map_trunc_page(vm_offset,
+                                            effective_page_mask);
+               vm_end = vm_map_round_page(vm_offset + vm_size,
+                                          effective_page_mask);
+       }
 
-       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 (vm_start < result->min_vm_addr)
+               result->min_vm_addr = vm_start;
+       if (vm_end > result->max_vm_addr)
+               result->max_vm_addr = vm_end;
 
        if (map == VM_MAP_NULL)
                return (LOAD_SUCCESS);
 
-       map_offset = pager_offset + scp->fileoff;       /* limited to 32 bits */
-
-       if (map_size > 0) {
+       if (vm_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 (verbose) {
+                       MACHO_PRINTF(("++++++ load_segment: "
+                                     "mapping at vm [0x%llx:0x%llx] of "
+                                     "file [0x%llx:0x%llx]\n",
+                                     (uint64_t) vm_start,
+                                     (uint64_t) vm_end,
+                                     (uint64_t) file_start,
+                                     (uint64_t) file_end));
+               }
+               ret = map_segment(map,
+                                 vm_start,
+                                 vm_end,
+                                 control,
+                                 file_start,
+                                 file_end,
+                                 initprot,
+                                 maxprot);
+               if (ret) {
+                       return LOAD_NOSPACE;
+               }
+
+#if FIXME
                /*
                 *      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)
+                       ret = mach_vm_allocate(kernel_map, &tmp, delta_size, VM_FLAGS_ANYWHERE| VM_MAKE_TAG(VM_KERN_MEMORY_BSD));
+                       if (ret != KERN_SUCCESS) {
                                return(LOAD_RESOURCE);
+                       }
        
                        if (copyout(tmp, map_addr + scp->filesize,
                                                                delta_size)) {
@@ -963,44 +1381,156 @@ load_segment(
         *      than the size from the file, we need to allocate
         *      zero fill memory for the rest.
         */
-       delta_size = seg_size - map_size;
+       if ((vm_end - vm_start) > (file_end - file_start)) {
+               delta_size = (vm_end - vm_start) - (file_end - file_start);
+       } else {
+               delta_size = 0;
+       }
        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)
+               mach_vm_offset_t tmp;
+
+               tmp = vm_start + (file_end - file_start);
+               if (verbose) {
+                       MACHO_PRINTF(("++++++ load_segment: "
+                                     "delta mapping vm [0x%llx:0x%llx]\n",
+                                     (uint64_t) tmp,
+                                     (uint64_t) (tmp + delta_size)));
+               }
+               kr = map_segment(map,
+                                tmp,
+                                tmp + delta_size,
+                                MEMORY_OBJECT_CONTROL_NULL,
+                                0,
+                                delta_size,
+                                scp->initprot,
+                                scp->maxprot);
+               if (kr != KERN_SUCCESS) {
                        return(LOAD_NOSPACE);
+               }
        }
 
        if ( (scp->fileoff == 0) && (scp->filesize != 0) )
-               result->mach_header = map_addr;
+               result->mach_header = vm_offset;
 
        if (scp->flags & SG_PROTECTED_VERSION_1) {
-               ret = unprotect_segment(scp->fileoff,
-                                       scp->filesize,
-                                       vp,
-                                       pager_offset,
-                                       map,
-                                       map_addr,
-                                       map_size);
+               ret = unprotect_dsmos_segment(file_start,
+                                             file_end - file_start,
+                                             vp,
+                                             pager_offset,
+                                             map,
+                                             vm_start,
+                                             vm_end - vm_start);
+               if (ret != LOAD_SUCCESS) {
+                       return ret;
+               }
        } else {
                ret = LOAD_SUCCESS;
        }
-       if (LOAD_SUCCESS == ret && filetype == MH_DYLINKER &&
-           result->all_image_info_addr == MACH_VM_MIN_ADDRESS)
+
+       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);
+                                           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;
+       if (result->entry_point != MACH_VM_MIN_ADDRESS) {
+               if ((result->entry_point >= vm_offset) && (result->entry_point < (vm_offset + vm_size))) {
+                       if ((scp->initprot & (VM_PROT_READ|VM_PROT_EXECUTE)) == (VM_PROT_READ|VM_PROT_EXECUTE)) {
+                               result->validentry = 1;
+                       } else {
+                               /* right range but wrong protections, unset if previously validated */
+                               result->validentry = 0;
+                       }
+               }
+       }
 
        return ret;
 }
 
+static
+load_return_t
+load_uuid(
+       struct uuid_command     *uulp,
+       char                    *command_end,
+       load_result_t           *result
+)
+{
+               /*
+                * We need to check the following for this command:
+                * - The command size should be atleast the size of struct uuid_command
+                * - The UUID part of the command should be completely within the mach-o header
+                */
+
+               if ((uulp->cmdsize < sizeof(struct uuid_command)) ||
+                   (((char *)uulp + sizeof(struct uuid_command)) > command_end)) {
+                       return (LOAD_BADMACHO);
+               }
+               
+               memcpy(&result->uuid[0], &uulp->uuid[0], sizeof(result->uuid));
+               return (LOAD_SUCCESS);
+}
+
+static
+load_return_t
+load_main(
+       struct entry_point_command      *epc,
+       thread_t                thread,
+       int64_t                         slide,
+       load_result_t           *result
+)
+{
+       mach_vm_offset_t addr;
+       kern_return_t   ret;
+       
+       if (epc->cmdsize < sizeof(*epc))
+               return (LOAD_BADMACHO);
+       if (result->thread_count != 0) {
+               return (LOAD_FAILURE);
+       }
+
+       if (thread == THREAD_NULL)
+               return (LOAD_SUCCESS);
+       
+       /* LC_MAIN specifies stack size but not location */
+       if (epc->stacksize) {
+               result->prog_stack_size = 1;
+               result->user_stack_size = epc->stacksize;
+       } else {
+               result->prog_stack_size = 0;
+               result->user_stack_size = MAXSSIZ;
+       }
+       result->prog_allocated_stack = 0;
+
+       /* use default location for stack */
+       ret = thread_userstackdefault(thread, &addr);
+       if (ret != KERN_SUCCESS)
+               return(LOAD_FAILURE);
+
+       /* The stack slides down from the default location */
+       result->user_stack = addr;
+       result->user_stack -= slide;
+
+       if (result->using_lcmain || result->entry_point != MACH_VM_MIN_ADDRESS) {
+               /* Already processed LC_MAIN or LC_UNIXTHREAD */
+               return (LOAD_FAILURE);
+       }
+
+       /* kernel does *not* use entryoff from LC_MAIN.  Dyld uses it. */
+       result->needs_dynlinker = TRUE;
+       result->using_lcmain = TRUE;
+       result->unixproc = TRUE;
+       result->thread_count++;
+
+       return(LOAD_SUCCESS);
+}
+
+
 static
 load_return_t
 load_unixthread(
@@ -1012,11 +1542,11 @@ load_unixthread(
 {
        load_return_t   ret;
        int customstack =0;
+       mach_vm_offset_t addr;
        
        if (tcp->cmdsize < sizeof(*tcp))
                return (LOAD_BADMACHO);
        if (result->thread_count != 0) {
-               printf("load_unixthread: already have a thread!");
                return (LOAD_FAILURE);
        }
 
@@ -1027,32 +1557,46 @@ load_unixthread(
                       (uint32_t *)(((vm_offset_t)tcp) + 
                                sizeof(struct thread_command)),
                       tcp->cmdsize - sizeof(struct thread_command),
-                      &result->user_stack,
+                      &addr,
                           &customstack);
        if (ret != LOAD_SUCCESS)
                return(ret);
 
-       if (customstack)
-               result->customstack = 1;
-       else
-               result->customstack = 0;
+       /* LC_UNIXTHREAD optionally specifies stack size and location */
+    
+       if (customstack) {
+               result->prog_stack_size = 0;    /* unknown */
+               result->prog_allocated_stack = 1;
+       } else {
+               result->prog_allocated_stack = 0;
+               result->prog_stack_size = 0;
+               result->user_stack_size = MAXSSIZ;
+       }
 
-       result->user_stack += slide;
+       /* The stack slides down from the default location */
+       result->user_stack = addr;
+       result->user_stack -= slide;
 
        ret = load_threadentry(thread,
                       (uint32_t *)(((vm_offset_t)tcp) + 
                                sizeof(struct thread_command)),
                       tcp->cmdsize - sizeof(struct thread_command),
-                      &result->entry_point);
+                      &addr);
        if (ret != LOAD_SUCCESS)
                return(ret);
 
+       if (result->using_lcmain || result->entry_point != MACH_VM_MIN_ADDRESS) {
+               /* Already processed LC_MAIN or LC_UNIXTHREAD */
+               return (LOAD_FAILURE);
+       }
+
+       result->entry_point = addr;
        result->entry_point += slide;
 
        ret = load_threadstate(thread,
-                      (uint32_t *)(((vm_offset_t)tcp) + 
-                               sizeof(struct thread_command)),
-                      tcp->cmdsize - sizeof(struct thread_command));
+                      (uint32_t *)(((vm_offset_t)tcp) + sizeof(struct thread_command)),
+                      tcp->cmdsize - sizeof(struct thread_command),
+                      result);
        if (ret != LOAD_SUCCESS)
                return (ret);
 
@@ -1067,45 +1611,58 @@ load_return_t
 load_threadstate(
        thread_t        thread,
        uint32_t        *ts,
-       uint32_t        total_size
+       uint32_t        total_size,
+       load_result_t   *result
 )
 {
-       kern_return_t   ret;
        uint32_t        size;
        int             flavor;
        uint32_t        thread_size;
+       uint32_t        *local_ts = NULL;
+       uint32_t        local_ts_size = 0;
+       int             ret;
+
+       (void)thread;
+
+       if (total_size > 0) {
+               local_ts_size = total_size;
+               local_ts = kalloc(local_ts_size);
+               if (local_ts == NULL) {
+                       return LOAD_FAILURE;
+               }
+               memcpy(local_ts, ts, local_ts_size);
+               ts = local_ts;
+       }
 
-    ret = thread_state_initialize( thread );
-    if (ret != KERN_SUCCESS) {
-        return(LOAD_FAILURE);
-    }
-    
        /*
-        *      Set the new thread state; iterate through the state flavors in
-     *  the mach-o file.
+        * Validate the new thread state; iterate through the state flavors in
+        * the Mach-O file.
+        * XXX: we should validate the machine state here, to avoid failing at
+        * activation time where we can't bail out cleanly.
         */
        while (total_size > 0) {
                flavor = *ts++;
                size = *ts++;
-               if (UINT32_MAX-2 < size ||
-                   UINT32_MAX/sizeof(uint32_t) < size+2)
-                       return (LOAD_BADMACHO);
-               thread_size = (size+2)*sizeof(uint32_t);
-               if (thread_size > total_size)
-                       return(LOAD_BADMACHO);
-               total_size -= thread_size;
-               /*
-                * Third argument is a kernel space pointer; it gets cast
-                * to the appropriate type in machine_thread_set_state()
-                * based on the value of flavor.
-                */
-               ret = thread_setstatus(thread, flavor, (thread_state_t)ts, size);
-               if (ret != KERN_SUCCESS) {
-                       return(LOAD_FAILURE);
+
+               if (os_add_overflow(size, UINT32_C(2), &thread_size) ||
+                   os_mul_overflow(thread_size, (uint32_t)sizeof(uint32_t), &thread_size) ||
+                   os_sub_overflow(total_size, thread_size, &total_size)) {
+                       ret = LOAD_BADMACHO;
+                       goto bad;
                }
+
                ts += size;     /* ts is a (uint32_t *) */
        }
-       return(LOAD_SUCCESS);
+
+       result->threadstate = local_ts;
+       result->threadstate_sz = local_ts_size;
+       return LOAD_SUCCESS;
+
+bad:
+       if (local_ts) {
+               kfree(local_ts, local_ts_size);
+       }
+       return ret;
 }
 
 static
@@ -1199,6 +1756,8 @@ struct macho_data {
        } __header;
 };
 
+#define DEFAULT_DYLD_PATH "/usr/lib/dyld"
+
 static load_return_t
 load_dylinker(
        struct dylinker_command *lcp,
@@ -1238,6 +1797,12 @@ load_dylinker(
                        return(LOAD_BADMACHO);
        } while (*p++);
 
+#if !(DEVELOPMENT || DEBUG)
+       if (0 != strcmp(name, DEFAULT_DYLD_PATH)) {
+               return (LOAD_BADMACHO);
+       }
+#endif
+
        /* Allocate wad-of-data from heap to reduce excessively deep stacks */
 
        MALLOC(dyld_data, void *, sizeof (*dyld_data), M_TEMP, M_WAITOK);
@@ -1259,7 +1824,7 @@ load_dylinker(
         */
 
        ret = parse_machfile(vp, map, thread, header, file_offset,
-           macho_size, depth, slide, myresult);
+                            macho_size, depth, slide, 0, myresult);
 
        /*
         *      If it turned out something was in the way, then we'll take
@@ -1280,7 +1845,8 @@ load_dylinker(
                 * subsequent map attempt (with a slide) in "myresult"
                 */
                ret = parse_machfile(vp, VM_MAP_NULL, THREAD_NULL, header,
-                   file_offset, macho_size, depth, 0 /* slide */, myresult);
+                                    file_offset, macho_size, depth,
+                                    0 /* slide */, 0, myresult);
 
                if (ret != LOAD_SUCCESS) {
                        goto out;
@@ -1315,18 +1881,30 @@ load_dylinker(
                *myresult = load_result_null;
 
                ret = parse_machfile(vp, map, thread, header,
-                   file_offset, macho_size, depth, slide_amount, myresult);
+                                    file_offset, macho_size, depth,
+                                    slide_amount, 0, myresult);
 
                if (ret) {
                        goto out;
                }
        }
 
-       if (ret == LOAD_SUCCESS) {              
+       if (ret == LOAD_SUCCESS) {
+               if (result->threadstate) {
+                       /* don't use the app's threadstate if we have a dyld */
+                       kfree(result->threadstate, result->threadstate_sz);
+               }
+               result->threadstate = myresult->threadstate;
+               result->threadstate_sz = myresult->threadstate_sz;
+
                result->dynlinker = TRUE;
                result->entry_point = myresult->entry_point;
+               result->validentry = myresult->validentry;
                result->all_image_info_addr = myresult->all_image_info_addr;
                result->all_image_info_size = myresult->all_image_info_size;
+               if (myresult->platform_binary) {
+                       result->csflags |= CS_DYLD_PLATFORM;
+               }
        }
 out:
        vnode_put(vp);
@@ -1362,13 +1940,19 @@ load_code_signature(
                goto out;
        }
 
-       blob = ubc_cs_blob_get(vp, cputype, -1);
+       blob = ubc_cs_blob_get(vp, cputype, macho_offset);
        if (blob != NULL) {
                /* we already have a blob for this vnode and cputype */
                if (blob->csb_cpu_type == cputype &&
                    blob->csb_base_offset == macho_offset &&
                    blob->csb_mem_size == lcp->datasize) {
-                       /* it matches the blob we want here: we're done */
+                       /* it matches the blob we want here, lets verify the version */
+                       if(0 != ubc_cs_generation_check(vp)) {
+                               if (0 != ubc_cs_blob_revalidate(vp, blob, 0)) {
+                                       ret = LOAD_FAILURE; /* set error same as from ubc_cs_blob_add */
+                                       goto out;
+                               }
+                       }
                        ret = LOAD_SUCCESS;
                } else {
                        /* the blob has changed for this vnode: fail ! */
@@ -1404,7 +1988,9 @@ load_code_signature(
                            cputype,
                            macho_offset,
                            addr,
-                           lcp->datasize)) {
+                           lcp->datasize,
+                           0,
+                           &blob)) {
                ret = LOAD_FAILURE;
                goto out;
        } else {
@@ -1416,12 +2002,15 @@ load_code_signature(
        ubc_cs_validation_bitmap_allocate( vp );
 #endif
                
-       blob = ubc_cs_blob_get(vp, cputype, -1);
-
        ret = LOAD_SUCCESS;
 out:
-       if (result && ret == LOAD_SUCCESS) {
+       if (ret == LOAD_SUCCESS) {
+               if (blob == NULL)
+                       panic("sucess, but no blob!");
+
                result->csflags |= blob->csb_flags;
+               result->platform_binary = blob->csb_platform_binary;
+               result->cs_end_offset = blob->csb_end_offset;
        }
        if (addr != 0) {
                ubc_cs_blob_deallocate(addr, blob_size);
@@ -1436,12 +2025,16 @@ out:
 
 static load_return_t
 set_code_unprotect(
-                  struct encryption_info_command *eip,
-                  caddr_t addr,        
-                  vm_map_t map,
-                  struct vnode *vp)
+       struct encryption_info_command *eip,
+       caddr_t addr,   
+       vm_map_t map,
+       int64_t slide,
+       struct vnode *vp,
+       off_t macho_offset,
+       cpu_type_t cputype,
+       cpu_subtype_t cpusubtype)
 {
-       int result, len;
+       int error, len;
        pager_crypt_info_t crypt_info;
        const char * cryptname = 0;
        char *vpath;
@@ -1450,6 +2043,7 @@ set_code_unprotect(
        struct segment_command_64 *seg64;
        struct segment_command *seg32;
        vm_map_offset_t map_offset, map_size;
+       vm_object_offset_t crypto_backing_offset;
        kern_return_t kr;
 
        if (eip->cmdsize < sizeof(*eip)) return LOAD_BADMACHO;
@@ -1477,20 +2071,34 @@ set_code_unprotect(
        if(vpath == NULL) return LOAD_FAILURE;
        
        len = MAXPATHLEN;
-       result = vn_getpath(vp, vpath, &len);
-       if(result) {
+       error = vn_getpath(vp, vpath, &len);
+       if (error) {
                FREE_ZONE(vpath, MAXPATHLEN, M_NAMEI);
                return LOAD_FAILURE;
        }
        
        /* set up decrypter first */
-       kr=text_crypter_create(&crypt_info, cryptname, (void*)vpath);
+       crypt_file_data_t crypt_data = {
+               .filename = vpath,
+               .cputype = cputype,
+               .cpusubtype = cpusubtype};
+       kr=text_crypter_create(&crypt_info, cryptname, (void*)&crypt_data);
+#if DEVELOPMENT || DEBUG
+       struct proc *p;
+       p  = current_proc();
+       printf("APPLE_PROTECT: %d[%s] map %p %s(%s) -> 0x%x\n",
+              p->p_pid, p->p_comm, map, __FUNCTION__, vpath, kr);
+#endif /* DEVELOPMENT || DEBUG */
        FREE_ZONE(vpath, MAXPATHLEN, M_NAMEI);
        
        if(kr) {
                printf("set_code_unprotect: unable to create decrypter %s, kr=%d\n",
                       cryptname, kr);
-               return LOAD_RESOURCE;
+               if (kr == kIOReturnNotPrivileged) {
+                       /* text encryption returned decryption failure */
+                       return(LOAD_DECRYPTFAIL);
+                }else
+                       return LOAD_RESOURCE;
        }
        
        /* this is terrible, but we have to rescan the load commands to find the
@@ -1517,8 +2125,9 @@ set_code_unprotect(
                                if ((seg64->fileoff <= eip->cryptoff) &&
                                    (seg64->fileoff+seg64->filesize >= 
                                     eip->cryptoff+eip->cryptsize)) {
-                                       map_offset = seg64->vmaddr + eip->cryptoff - seg64->fileoff;
+                                       map_offset = seg64->vmaddr + eip->cryptoff - seg64->fileoff + slide;
                                        map_size = eip->cryptsize;
+                                       crypto_backing_offset = macho_offset + eip->cryptoff;
                                        goto remap_now;
                                }
                        case LC_SEGMENT:
@@ -1526,8 +2135,9 @@ set_code_unprotect(
                                if ((seg32->fileoff <= eip->cryptoff) &&
                                    (seg32->fileoff+seg32->filesize >= 
                                     eip->cryptoff+eip->cryptsize)) {
-                                       map_offset = seg32->vmaddr + eip->cryptoff - seg32->fileoff;
+                                       map_offset = seg32->vmaddr + eip->cryptoff - seg32->fileoff + slide;
                                        map_size = eip->cryptsize;
+                                       crypto_backing_offset = macho_offset + eip->cryptoff;
                                        goto remap_now;
                                }
                }
@@ -1538,10 +2148,16 @@ set_code_unprotect(
        
 remap_now:
        /* now remap using the decrypter */
-       kr = vm_map_apple_protected(map, map_offset, map_offset+map_size, &crypt_info);
-       if(kr) {
+       MACHO_PRINTF(("+++ set_code_unprotect: vm[0x%llx:0x%llx]\n",
+                     (uint64_t) map_offset,
+                     (uint64_t) (map_offset+map_size)));
+       kr = vm_map_apple_protected(map,
+                                   map_offset,
+                                   map_offset+map_size,
+                                   crypto_backing_offset,
+                                   &crypt_info);
+       if (kr) {
                printf("set_code_unprotect(): mapping failed with %x\n", kr);
-               crypt_info.crypt_end(crypt_info.crypt_ops);
                return LOAD_PROTECT;
        }
        
@@ -1623,7 +2239,7 @@ get_macho_vnode(
        }
 
        /* check access */
-       if ((error = vnode_authorize(vp, NULL, KAUTH_VNODE_EXECUTE, ctx)) != 0) {
+       if ((error = vnode_authorize(vp, NULL, KAUTH_VNODE_EXECUTE | KAUTH_VNODE_READ_DATA, ctx)) != 0) {
                error = LOAD_PROTECT;
                goto bad1;
        }
@@ -1640,21 +2256,32 @@ get_macho_vnode(
                goto bad2;
        }
 
+       if (resid) {
+               error = LOAD_BADMACHO;
+               goto bad2;
+       }
+
        if (header->mach_header.magic == MH_MAGIC ||
            header->mach_header.magic == MH_MAGIC_64) {
                is_fat = FALSE;
-       } else if (header->fat_header.magic == FAT_MAGIC ||
-           header->fat_header.magic == FAT_CIGAM) {
-               is_fat = TRUE;
+       } else if (OSSwapBigToHostInt32(header->fat_header.magic) == FAT_MAGIC) {
+           is_fat = TRUE;
        } else {
                error = LOAD_BADMACHO;
                goto bad2;
        }
 
        if (is_fat) {
+
+               error = fatfile_validate_fatarches((vm_offset_t)(&header->fat_header),
+                                               sizeof(*header));
+               if (error != LOAD_SUCCESS) {
+                       goto bad2;
+               }
+
                /* Look up our architecture in the fat file. */
-               error = fatfile_getarch_with_bits(vp, archbits,
-                   (vm_offset_t)(&header->fat_header), &fat_arch);
+               error = fatfile_getarch_with_bits(archbits,
+                                               (vm_offset_t)(&header->fat_header), sizeof(*header), &fat_arch);
                if (error != LOAD_SUCCESS)
                        goto bad2;
 
@@ -1667,6 +2294,11 @@ get_macho_vnode(
                        goto bad2;
                }
 
+               if (resid) {
+                       error = LOAD_BADMACHO;
+                       goto bad2;
+               }
+
                /* Is this really a Mach-O? */
                if (header->mach_header.magic != MH_MAGIC &&
                    header->mach_header.magic != MH_MAGIC_64) {