Loading...
--- dyld/dyld-44.4/src/dyldInitialization.cpp
+++ dyld/dyld-96.2/src/dyldInitialization.cpp
@@ -22,8 +22,11 @@
* @APPLE_LICENSE_HEADER_END@
*/
+#define __STDC_LIMIT_MACROS
+#include <stdint.h>
#include <stddef.h>
#include <string.h>
+#include <stdlib.h>
#include <mach/mach.h>
#include <mach-o/loader.h>
#include <mach-o/ldsyms.h>
@@ -31,7 +34,15 @@
#if __ppc__ || __ppc64__
#include <mach-o/ppc/reloc.h>
#endif
+#if __x86_64__
+ #include <mach-o/x86_64/reloc.h>
+#endif
#include "dyld.h"
+
+#ifndef MH_PIE
+ #define MH_PIE 0x200000
+#endif
+
#if __LP64__
#define macho_header mach_header_64
@@ -46,6 +57,15 @@
#define macho_section section
#define RELOC_SIZE 2
#endif
+
+#if __x86_64__
+ #define POINTER_RELOC X86_64_RELOC_UNSIGNED
+#else
+ #define POINTER_RELOC GENERIC_RELOC_VANILLA
+#endif
+
+// from dyld.cpp
+namespace dyld { extern bool isRosetta(); };
//
@@ -94,20 +114,22 @@
}
}
-
//
// If the kernel does not load dyld at its preferred address, we need to apply
// fixups to various initialized parts of the __DATA segment
//
static void rebaseDyld(const struct macho_header* mh, intptr_t slide)
{
- // get interesting pointers into dyld
+ // rebase non-lazy pointers (which all point internal to dyld, since dyld uses no shared libraries)
+ // and get interesting pointers into dyld
const uint32_t cmd_count = mh->ncmds;
const struct load_command* const cmds = (struct load_command*)(((char*)mh)+sizeof(macho_header));
const struct load_command* cmd = cmds;
const struct macho_segment_command* linkEditSeg = NULL;
+#if __x86_64__
+ const struct macho_segment_command* firstWritableSeg = NULL;
+#endif
const struct dysymtab_command* dynamicSymbolTable = NULL;
- const struct macho_section* nonLazySection = NULL;
for (uint32_t i = 0; i < cmd_count; ++i) {
switch (cmd->cmd) {
case LC_SEGMENT_COMMAND:
@@ -119,9 +141,19 @@
const struct macho_section* const sectionsEnd = §ionsStart[seg->nsects];
for (const struct macho_section* sect=sectionsStart; sect < sectionsEnd; ++sect) {
const uint8_t type = sect->flags & SECTION_TYPE;
- if ( type == S_NON_LAZY_SYMBOL_POINTERS )
- nonLazySection = sect;
+ if ( type == S_NON_LAZY_SYMBOL_POINTERS ) {
+ // rebase non-lazy pointers (which all point internal to dyld, since dyld uses no shared libraries)
+ const uint32_t pointerCount = sect->size / sizeof(uintptr_t);
+ uintptr_t* const symbolPointers = (uintptr_t*)(sect->addr + slide);
+ for (uint32_t j=0; j < pointerCount; ++j) {
+ symbolPointers[j] += slide;
+ }
+ }
}
+#if __x86_64__
+ if ( (firstWritableSeg == NULL) && (seg->initprot & VM_PROT_WRITE) )
+ firstWritableSeg = seg;
+#endif
}
break;
case LC_DYSYMTAB:
@@ -132,50 +164,29 @@
}
// use reloc's to rebase all random data pointers
+#if __x86_64__
+ const uintptr_t relocBase = firstWritableSeg->vmaddr + slide;
+#else
const uintptr_t relocBase = (uintptr_t)mh;
+#endif
const relocation_info* const relocsStart = (struct relocation_info*)(linkEditSeg->vmaddr + slide + dynamicSymbolTable->locreloff - linkEditSeg->fileoff);
const relocation_info* const relocsEnd = &relocsStart[dynamicSymbolTable->nlocrel];
for (const relocation_info* reloc=relocsStart; reloc < relocsEnd; ++reloc) {
- if ( (reloc->r_address & R_SCATTERED) == 0 ) {
- if (reloc->r_length == RELOC_SIZE) {
- switch(reloc->r_type) {
- case GENERIC_RELOC_VANILLA:
- *((uintptr_t*)(reloc->r_address + relocBase)) += slide;
- break;
- }
- }
- }
- else {
- const struct scattered_relocation_info* sreloc = (struct scattered_relocation_info*)reloc;
- if (sreloc->r_length == RELOC_SIZE) {
- uintptr_t* locationToFix = (uintptr_t*)(sreloc->r_address + relocBase);
- switch(sreloc->r_type) {
- case GENERIC_RELOC_VANILLA:
- #if __ppc__ || __ppc64__
- case PPC_RELOC_PB_LA_PTR:
- #elif __i386__
- case GENERIC_RELOC_PB_LA_PTR:
- #endif
- // Note the use of PB_LA_PTR is unique here. Seems like ld should strip out all lazy pointers
- // but it does not. But, since all lazy-pointers point within dyld, they can be slid too
- *locationToFix += slide;
- break;
- }
- }
- }
- }
-
- // rebase non-lazy pointers (which all point internal to dyld, since dyld uses no shared libraries)
- if ( nonLazySection != NULL ) {
- const uint32_t pointerCount = nonLazySection->size / sizeof(uintptr_t);
- uintptr_t* const symbolPointers = (uintptr_t*)(nonLazySection->addr + slide);
- for (uint32_t j=0; j < pointerCount; ++j) {
- symbolPointers[j] += slide;
- }
- }
-
-
-}
+ #if __ppc__ || __ppc64__ || __i36__
+ if ( (reloc->r_address & R_SCATTERED) != 0 )
+ throw "scattered relocation in dyld";
+ #endif
+ if ( reloc->r_length != RELOC_SIZE )
+ throw "relocation in dyld has wrong size";
+
+ if ( reloc->r_type != POINTER_RELOC )
+ throw "relocation in dyld has wrong type";
+
+ // update pointer by amount dyld slid
+ *((uintptr_t*)(reloc->r_address + relocBase)) += slide;
+ }
+}
+
//
// For some reason the kernel loads dyld with __TEXT and __LINKEDIT writable
@@ -195,7 +206,7 @@
vm_size_t size = seg->vmsize;
const bool setCurrentPermissions = false;
vm_protect(mach_task_self(), addr, size, setCurrentPermissions, seg->initprot);
- //fprintf(stderr, "dyld: segment %s, 0x%08X -> 0x%08X, set to %d\n", seg->segname, addr, addr+size-1, seg->initprot);
+ //dyld::log("dyld: segment %s, 0x%08X -> 0x%08X, set to %d\n", seg->segname, addr, addr+size-1, seg->initprot);
}
break;
}
@@ -204,8 +215,113 @@
}
+
+//
+// re-map the main executable to a new random address
+//
+static const struct mach_header* randomizeExecutableLoadAddress(const struct mach_header* orgMH, uintptr_t* appsSlide)
+{
+#if __ppc__
+ // don't slide PIE programs running under rosetta
+ if ( dyld::isRosetta() )
+ return orgMH;
+#endif
+ // count segments
+ uint32_t segCount = 0;
+ const uint32_t cmd_count = orgMH->ncmds;
+ const struct load_command* const cmds = (struct load_command*)(((char*)orgMH)+sizeof(macho_header));
+ const struct load_command* cmd = cmds;
+ for (uint32_t i = 0; i < cmd_count; ++i) {
+ if ( cmd->cmd == LC_SEGMENT_COMMAND ) {
+ const struct macho_segment_command* segCmd = (struct macho_segment_command*)cmd;
+ // page-zero and custom stacks don't move
+ if ( (strcmp(segCmd->segname, "__PAGEZERO") != 0) && (strcmp(segCmd->segname, "__UNIXSTACK") != 0) )
+ ++segCount;
+ }
+ cmd = (const struct load_command*)(((char*)cmd)+cmd->cmdsize);
+ }
+
+ // make copy of segment info
+ macho_segment_command segs[segCount];
+ uint32_t index = 0;
+ uintptr_t highestAddressUsed = 0;
+ uintptr_t lowestAddressUsed = UINTPTR_MAX;
+ cmd = cmds;
+ for (uint32_t i = 0; i < cmd_count; ++i) {
+ if ( cmd->cmd == LC_SEGMENT_COMMAND ) {
+ const struct macho_segment_command* segCmd = (struct macho_segment_command*)cmd;
+ if ( (strcmp(segCmd->segname, "__PAGEZERO") != 0) && (strcmp(segCmd->segname, "__UNIXSTACK") != 0) ) {
+ segs[index++] = *segCmd;
+ if ( (segCmd->vmaddr + segCmd->vmsize) > highestAddressUsed )
+ highestAddressUsed = ((segCmd->vmaddr + segCmd->vmsize) + 4095) & -4096;
+ if ( segCmd->vmaddr < lowestAddressUsed )
+ lowestAddressUsed = segCmd->vmaddr;
+ // do nothing if kernel has already randomized load address
+ if ( (strcmp(segCmd->segname, "__TEXT") == 0) && (segCmd->vmaddr != (uintptr_t)orgMH) )
+ return orgMH;
+ }
+ }
+ cmd = (const struct load_command*)(((char*)cmd)+cmd->cmdsize);
+ }
+
+ // choose a random new base address
+#if __LP64__
+ uintptr_t highestAddressPossible = highestAddressUsed + 0x100000000ULL;
+#else
+ uintptr_t highestAddressPossible = 0x80000000;
+#endif
+ uintptr_t sizeNeeded = highestAddressUsed-lowestAddressUsed;
+ if ( (highestAddressPossible-sizeNeeded) < highestAddressUsed ) {
+ // new and old segments will overlap
+ // need better algorithm for remapping
+ // punt and don't re-map
+ return orgMH;
+ }
+ uintptr_t possibleRange = (highestAddressPossible-sizeNeeded) - highestAddressUsed;
+ uintptr_t newBaseAddress = highestAddressUsed + ((arc4random() % possibleRange) & -4096);
+
+ vm_address_t addr = newBaseAddress;
+ // reserve new address range
+ if ( vm_allocate(mach_task_self(), &addr, sizeNeeded, VM_FLAGS_FIXED) == KERN_SUCCESS ) {
+ // copy each segment to new address
+ for (uint32_t i = 0; i < segCount; ++i) {
+ uintptr_t newSegAddress = segs[i].vmaddr - lowestAddressUsed + newBaseAddress;
+ if ( (vm_copy(mach_task_self(), segs[i].vmaddr, segs[i].vmsize, newSegAddress) != KERN_SUCCESS)
+ || (vm_protect(mach_task_self(), newSegAddress, segs[i].vmsize, true, segs[i].maxprot) != KERN_SUCCESS)
+ || (vm_protect(mach_task_self(), newSegAddress, segs[i].vmsize, false, segs[i].initprot) != KERN_SUCCESS) ) {
+ // can't copy so dealloc new region and run with original base address
+ vm_deallocate(mach_task_self(), newBaseAddress, sizeNeeded);
+ dyld::warn("could not relocate position independent exectable\n");
+ return orgMH;
+ }
+ }
+ // unmap original segments
+ vm_deallocate(mach_task_self(), lowestAddressUsed, highestAddressUsed-lowestAddressUsed);
+
+ // run with newly mapped executable
+ *appsSlide = newBaseAddress - lowestAddressUsed;
+ return (const struct mach_header*)newBaseAddress;
+ }
+
+ // can't get new range, so don't slide to random address
+ return orgMH;
+}
+
+
extern "C" void dyld_exceptions_init(const struct macho_header*, uintptr_t slide); // in dyldExceptions.cpp
extern "C" void mach_init();
+
+//
+// _pthread_keys is partitioned in a lower part that dyld will use; libSystem
+// will use the upper part. We set __pthread_tsd_first to 1 as the start of
+// the lower part. Libc will take #1 and c++ exceptions will take #2. There
+// is one free key=3 left.
+//
+extern "C" {
+ extern int __pthread_tsd_first;
+ extern void _pthread_keys_init();
+}
+
//
// This is code to bootstrap dyld. This work in normally done for a program by dyld and crt.
@@ -222,6 +338,12 @@
rebaseDyld(dyldsMachHeader, slide);
}
+ uintptr_t appsSlide = 0;
+
+ // set pthread keys to dyld range
+ __pthread_tsd_first = 1;
+ _pthread_keys_init();
+
// enable C++ exceptions to work inside dyld
dyld_exceptions_init(dyldsMachHeader, slide);
@@ -242,8 +364,12 @@
// run all C++ initializers inside dyld
runDyldInitializers(dyldsMachHeader, slide, argc, argv, envp, apple);
+ // if main executable was linked -pie, then randomize its load address
+ if ( appsMachHeader->flags & MH_PIE )
+ appsMachHeader = randomizeExecutableLoadAddress(appsMachHeader, &appsSlide);
+
// now that we are done bootstrapping dyld, call dyld's main
- return dyld::_main(appsMachHeader, argc, argv, envp, apple);
+ return dyld::_main(appsMachHeader, appsSlide, argc, argv, envp, apple);
}