0e179328f9
Change-Id: I62aefc53c3f24254ca193d29c80050ed20a671bf
1427 lines
52 KiB
C++
1427 lines
52 KiB
C++
/*
|
|
* Copyright (C) 2012 The Android Open Source Project
|
|
* All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
* * Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* * Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in
|
|
* the documentation and/or other materials provided with the
|
|
* distribution.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
|
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
|
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
|
|
* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
|
|
* COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
|
|
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
|
|
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
|
|
* OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
|
|
* AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
|
|
* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
|
|
* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
|
* SUCH DAMAGE.
|
|
*/
|
|
|
|
#include "linker_phdr.h"
|
|
|
|
#include <errno.h>
|
|
#include <string.h>
|
|
#include <sys/mman.h>
|
|
#include <sys/prctl.h>
|
|
#include <sys/types.h>
|
|
#include <sys/stat.h>
|
|
#include <unistd.h>
|
|
|
|
#include "linker.h"
|
|
#include "linker_dlwarning.h"
|
|
#include "linker_globals.h"
|
|
#include "linker_debug.h"
|
|
#include "linker_utils.h"
|
|
|
|
#include "private/bionic_asm_note.h"
|
|
#include "private/CFIShadow.h" // For kLibraryAlignment
|
|
#include "private/elf_note.h"
|
|
|
|
static int GetTargetElfMachine() {
|
|
#if defined(__arm__)
|
|
return EM_ARM;
|
|
#elif defined(__aarch64__)
|
|
return EM_AARCH64;
|
|
#elif defined(__i386__)
|
|
return EM_386;
|
|
#elif defined(__riscv)
|
|
return EM_RISCV;
|
|
#elif defined(__x86_64__)
|
|
return EM_X86_64;
|
|
#endif
|
|
}
|
|
|
|
/**
|
|
TECHNICAL NOTE ON ELF LOADING.
|
|
|
|
An ELF file's program header table contains one or more PT_LOAD
|
|
segments, which corresponds to portions of the file that need to
|
|
be mapped into the process' address space.
|
|
|
|
Each loadable segment has the following important properties:
|
|
|
|
p_offset -> segment file offset
|
|
p_filesz -> segment file size
|
|
p_memsz -> segment memory size (always >= p_filesz)
|
|
p_vaddr -> segment's virtual address
|
|
p_flags -> segment flags (e.g. readable, writable, executable)
|
|
p_align -> segment's in-memory and in-file alignment
|
|
|
|
We will ignore the p_paddr field of ElfW(Phdr) for now.
|
|
|
|
The loadable segments can be seen as a list of [p_vaddr ... p_vaddr+p_memsz)
|
|
ranges of virtual addresses. A few rules apply:
|
|
|
|
- the virtual address ranges should not overlap.
|
|
|
|
- if a segment's p_filesz is smaller than its p_memsz, the extra bytes
|
|
between them should always be initialized to 0.
|
|
|
|
- ranges do not necessarily start or end at page boundaries. Two distinct
|
|
segments can have their start and end on the same page. In this case, the
|
|
page inherits the mapping flags of the latter segment.
|
|
|
|
Finally, the real load addrs of each segment is not p_vaddr. Instead the
|
|
loader decides where to load the first segment, then will load all others
|
|
relative to the first one to respect the initial range layout.
|
|
|
|
For example, consider the following list:
|
|
|
|
[ offset:0, filesz:0x4000, memsz:0x4000, vaddr:0x30000 ],
|
|
[ offset:0x4000, filesz:0x2000, memsz:0x8000, vaddr:0x40000 ],
|
|
|
|
This corresponds to two segments that cover these virtual address ranges:
|
|
|
|
0x30000...0x34000
|
|
0x40000...0x48000
|
|
|
|
If the loader decides to load the first segment at address 0xa0000000
|
|
then the segments' load address ranges will be:
|
|
|
|
0xa0030000...0xa0034000
|
|
0xa0040000...0xa0048000
|
|
|
|
In other words, all segments must be loaded at an address that has the same
|
|
constant offset from their p_vaddr value. This offset is computed as the
|
|
difference between the first segment's load address, and its p_vaddr value.
|
|
|
|
However, in practice, segments do _not_ start at page boundaries. Since we
|
|
can only memory-map at page boundaries, this means that the bias is
|
|
computed as:
|
|
|
|
load_bias = phdr0_load_address - page_start(phdr0->p_vaddr)
|
|
|
|
(NOTE: The value must be used as a 32-bit unsigned integer, to deal with
|
|
possible wrap around UINT32_MAX for possible large p_vaddr values).
|
|
|
|
And that the phdr0_load_address must start at a page boundary, with
|
|
the segment's real content starting at:
|
|
|
|
phdr0_load_address + page_offset(phdr0->p_vaddr)
|
|
|
|
Note that ELF requires the following condition to make the mmap()-ing work:
|
|
|
|
page_offset(phdr0->p_vaddr) == page_offset(phdr0->p_offset)
|
|
|
|
The load_bias must be added to any p_vaddr value read from the ELF file to
|
|
determine the corresponding memory address.
|
|
|
|
**/
|
|
|
|
#define MAYBE_MAP_FLAG(x, from, to) (((x) & (from)) ? (to) : 0)
|
|
#define PFLAGS_TO_PROT(x) (MAYBE_MAP_FLAG((x), PF_X, PROT_EXEC) | \
|
|
MAYBE_MAP_FLAG((x), PF_R, PROT_READ) | \
|
|
MAYBE_MAP_FLAG((x), PF_W, PROT_WRITE))
|
|
|
|
static const size_t kPageSize = page_size();
|
|
|
|
/*
|
|
* Generic PMD size calculation:
|
|
* - Each page table (PT) is of size 1 page.
|
|
* - Each page table entry (PTE) is of size 64 bits.
|
|
* - Each PTE locates one physical page frame (PFN) of size 1 page.
|
|
* - A PMD entry locates 1 page table (PT)
|
|
*
|
|
* PMD size = Num entries in a PT * page_size
|
|
*/
|
|
static const size_t kPmdSize = (kPageSize / sizeof(uint64_t)) * kPageSize;
|
|
|
|
ElfReader::ElfReader()
|
|
: did_read_(false), did_load_(false), fd_(-1), file_offset_(0), file_size_(0), phdr_num_(0),
|
|
phdr_table_(nullptr), shdr_table_(nullptr), shdr_num_(0), dynamic_(nullptr), strtab_(nullptr),
|
|
strtab_size_(0), load_start_(nullptr), load_size_(0), load_bias_(0), loaded_phdr_(nullptr),
|
|
mapped_by_caller_(false) {
|
|
}
|
|
|
|
bool ElfReader::Read(const char* name, int fd, off64_t file_offset, off64_t file_size) {
|
|
if (did_read_) {
|
|
return true;
|
|
}
|
|
name_ = name;
|
|
fd_ = fd;
|
|
file_offset_ = file_offset;
|
|
file_size_ = file_size;
|
|
|
|
if (ReadElfHeader() &&
|
|
VerifyElfHeader() &&
|
|
ReadProgramHeaders() &&
|
|
ReadSectionHeaders() &&
|
|
ReadDynamicSection() &&
|
|
ReadPadSegmentNote()) {
|
|
did_read_ = true;
|
|
}
|
|
|
|
return did_read_;
|
|
}
|
|
|
|
bool ElfReader::Load(address_space_params* address_space) {
|
|
CHECK(did_read_);
|
|
if (did_load_) {
|
|
return true;
|
|
}
|
|
bool reserveSuccess = ReserveAddressSpace(address_space);
|
|
if (reserveSuccess && LoadSegments() && FindPhdr() &&
|
|
FindGnuPropertySection()) {
|
|
did_load_ = true;
|
|
#if defined(__aarch64__)
|
|
// For Armv8.5-A loaded executable segments may require PROT_BTI.
|
|
if (note_gnu_property_.IsBTICompatible()) {
|
|
did_load_ = (phdr_table_protect_segments(phdr_table_, phdr_num_, load_bias_,
|
|
should_pad_segments_, ¬e_gnu_property_) == 0);
|
|
}
|
|
#endif
|
|
}
|
|
if (reserveSuccess && !did_load_) {
|
|
if (load_start_ != nullptr && load_size_ != 0) {
|
|
if (!mapped_by_caller_) {
|
|
munmap(load_start_, load_size_);
|
|
}
|
|
}
|
|
}
|
|
|
|
return did_load_;
|
|
}
|
|
|
|
const char* ElfReader::get_string(ElfW(Word) index) const {
|
|
CHECK(strtab_ != nullptr);
|
|
CHECK(index < strtab_size_);
|
|
|
|
return strtab_ + index;
|
|
}
|
|
|
|
bool ElfReader::ReadElfHeader() {
|
|
ssize_t rc = TEMP_FAILURE_RETRY(pread64(fd_, &header_, sizeof(header_), file_offset_));
|
|
if (rc < 0) {
|
|
DL_ERR("can't read file \"%s\": %s", name_.c_str(), strerror(errno));
|
|
return false;
|
|
}
|
|
|
|
if (rc != sizeof(header_)) {
|
|
DL_ERR("\"%s\" is too small to be an ELF executable: only found %zd bytes", name_.c_str(),
|
|
static_cast<size_t>(rc));
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
static const char* EM_to_string(int em) {
|
|
if (em == EM_386) return "EM_386";
|
|
if (em == EM_AARCH64) return "EM_AARCH64";
|
|
if (em == EM_ARM) return "EM_ARM";
|
|
if (em == EM_RISCV) return "EM_RISCV";
|
|
if (em == EM_X86_64) return "EM_X86_64";
|
|
return "EM_???";
|
|
}
|
|
|
|
bool ElfReader::VerifyElfHeader() {
|
|
if (memcmp(header_.e_ident, ELFMAG, SELFMAG) != 0) {
|
|
DL_ERR("\"%s\" has bad ELF magic: %02x%02x%02x%02x", name_.c_str(),
|
|
header_.e_ident[0], header_.e_ident[1], header_.e_ident[2], header_.e_ident[3]);
|
|
return false;
|
|
}
|
|
|
|
// Try to give a clear diagnostic for ELF class mismatches, since they're
|
|
// an easy mistake to make during the 32-bit/64-bit transition period.
|
|
int elf_class = header_.e_ident[EI_CLASS];
|
|
#if defined(__LP64__)
|
|
if (elf_class != ELFCLASS64) {
|
|
if (elf_class == ELFCLASS32) {
|
|
DL_ERR("\"%s\" is 32-bit instead of 64-bit", name_.c_str());
|
|
} else {
|
|
DL_ERR("\"%s\" has unknown ELF class: %d", name_.c_str(), elf_class);
|
|
}
|
|
return false;
|
|
}
|
|
#else
|
|
if (elf_class != ELFCLASS32) {
|
|
if (elf_class == ELFCLASS64) {
|
|
DL_ERR("\"%s\" is 64-bit instead of 32-bit", name_.c_str());
|
|
} else {
|
|
DL_ERR("\"%s\" has unknown ELF class: %d", name_.c_str(), elf_class);
|
|
}
|
|
return false;
|
|
}
|
|
#endif
|
|
|
|
if (header_.e_ident[EI_DATA] != ELFDATA2LSB) {
|
|
DL_ERR("\"%s\" not little-endian: %d", name_.c_str(), header_.e_ident[EI_DATA]);
|
|
return false;
|
|
}
|
|
|
|
if (header_.e_type != ET_DYN) {
|
|
DL_ERR("\"%s\" has unexpected e_type: %d", name_.c_str(), header_.e_type);
|
|
return false;
|
|
}
|
|
|
|
if (header_.e_version != EV_CURRENT) {
|
|
DL_ERR("\"%s\" has unexpected e_version: %d", name_.c_str(), header_.e_version);
|
|
return false;
|
|
}
|
|
|
|
if (header_.e_machine != GetTargetElfMachine()) {
|
|
DL_ERR("\"%s\" is for %s (%d) instead of %s (%d)",
|
|
name_.c_str(),
|
|
EM_to_string(header_.e_machine), header_.e_machine,
|
|
EM_to_string(GetTargetElfMachine()), GetTargetElfMachine());
|
|
return false;
|
|
}
|
|
|
|
if (header_.e_shentsize != sizeof(ElfW(Shdr))) {
|
|
// Fail if app is targeting Android O or above
|
|
if (get_application_target_sdk_version() >= 26) {
|
|
DL_ERR_AND_LOG("\"%s\" has unsupported e_shentsize: 0x%x (expected 0x%zx)",
|
|
name_.c_str(), header_.e_shentsize, sizeof(ElfW(Shdr)));
|
|
return false;
|
|
}
|
|
DL_WARN_documented_change(26,
|
|
"invalid-elf-header_section-headers-enforced-for-api-level-26",
|
|
"\"%s\" has unsupported e_shentsize 0x%x (expected 0x%zx)",
|
|
name_.c_str(), header_.e_shentsize, sizeof(ElfW(Shdr)));
|
|
add_dlwarning(name_.c_str(), "has invalid ELF header");
|
|
}
|
|
|
|
if (header_.e_shstrndx == 0) {
|
|
// Fail if app is targeting Android O or above
|
|
if (get_application_target_sdk_version() >= 26) {
|
|
DL_ERR_AND_LOG("\"%s\" has invalid e_shstrndx", name_.c_str());
|
|
return false;
|
|
}
|
|
|
|
DL_WARN_documented_change(26,
|
|
"invalid-elf-header_section-headers-enforced-for-api-level-26",
|
|
"\"%s\" has invalid e_shstrndx", name_.c_str());
|
|
add_dlwarning(name_.c_str(), "has invalid ELF header");
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
bool ElfReader::CheckFileRange(ElfW(Addr) offset, size_t size, size_t alignment) {
|
|
off64_t range_start;
|
|
off64_t range_end;
|
|
|
|
// Only header can be located at the 0 offset... This function called to
|
|
// check DYNSYM and DYNAMIC sections and phdr/shdr - none of them can be
|
|
// at offset 0.
|
|
|
|
return offset > 0 &&
|
|
safe_add(&range_start, file_offset_, offset) &&
|
|
safe_add(&range_end, range_start, size) &&
|
|
(range_start < file_size_) &&
|
|
(range_end <= file_size_) &&
|
|
((offset % alignment) == 0);
|
|
}
|
|
|
|
// Loads the program header table from an ELF file into a read-only private
|
|
// anonymous mmap-ed block.
|
|
bool ElfReader::ReadProgramHeaders() {
|
|
phdr_num_ = header_.e_phnum;
|
|
|
|
// Like the kernel, we only accept program header tables that
|
|
// are smaller than 64KiB.
|
|
if (phdr_num_ < 1 || phdr_num_ > 65536/sizeof(ElfW(Phdr))) {
|
|
DL_ERR("\"%s\" has invalid e_phnum: %zd", name_.c_str(), phdr_num_);
|
|
return false;
|
|
}
|
|
|
|
// Boundary checks
|
|
size_t size = phdr_num_ * sizeof(ElfW(Phdr));
|
|
if (!CheckFileRange(header_.e_phoff, size, alignof(ElfW(Phdr)))) {
|
|
DL_ERR_AND_LOG("\"%s\" has invalid phdr offset/size: %zu/%zu",
|
|
name_.c_str(),
|
|
static_cast<size_t>(header_.e_phoff),
|
|
size);
|
|
return false;
|
|
}
|
|
|
|
if (!phdr_fragment_.Map(fd_, file_offset_, header_.e_phoff, size)) {
|
|
DL_ERR("\"%s\" phdr mmap failed: %s", name_.c_str(), strerror(errno));
|
|
return false;
|
|
}
|
|
|
|
phdr_table_ = static_cast<ElfW(Phdr)*>(phdr_fragment_.data());
|
|
return true;
|
|
}
|
|
|
|
bool ElfReader::ReadSectionHeaders() {
|
|
shdr_num_ = header_.e_shnum;
|
|
|
|
if (shdr_num_ == 0) {
|
|
DL_ERR_AND_LOG("\"%s\" has no section headers", name_.c_str());
|
|
return false;
|
|
}
|
|
|
|
size_t size = shdr_num_ * sizeof(ElfW(Shdr));
|
|
if (!CheckFileRange(header_.e_shoff, size, alignof(const ElfW(Shdr)))) {
|
|
DL_ERR_AND_LOG("\"%s\" has invalid shdr offset/size: %zu/%zu",
|
|
name_.c_str(),
|
|
static_cast<size_t>(header_.e_shoff),
|
|
size);
|
|
return false;
|
|
}
|
|
|
|
if (!shdr_fragment_.Map(fd_, file_offset_, header_.e_shoff, size)) {
|
|
DL_ERR("\"%s\" shdr mmap failed: %s", name_.c_str(), strerror(errno));
|
|
return false;
|
|
}
|
|
|
|
shdr_table_ = static_cast<const ElfW(Shdr)*>(shdr_fragment_.data());
|
|
return true;
|
|
}
|
|
|
|
bool ElfReader::ReadDynamicSection() {
|
|
// 1. Find .dynamic section (in section headers)
|
|
const ElfW(Shdr)* dynamic_shdr = nullptr;
|
|
for (size_t i = 0; i < shdr_num_; ++i) {
|
|
if (shdr_table_[i].sh_type == SHT_DYNAMIC) {
|
|
dynamic_shdr = &shdr_table_ [i];
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (dynamic_shdr == nullptr) {
|
|
DL_ERR_AND_LOG("\"%s\" .dynamic section header was not found", name_.c_str());
|
|
return false;
|
|
}
|
|
|
|
// Make sure dynamic_shdr offset and size matches PT_DYNAMIC phdr
|
|
size_t pt_dynamic_offset = 0;
|
|
size_t pt_dynamic_filesz = 0;
|
|
for (size_t i = 0; i < phdr_num_; ++i) {
|
|
const ElfW(Phdr)* phdr = &phdr_table_[i];
|
|
if (phdr->p_type == PT_DYNAMIC) {
|
|
pt_dynamic_offset = phdr->p_offset;
|
|
pt_dynamic_filesz = phdr->p_filesz;
|
|
}
|
|
}
|
|
|
|
if (pt_dynamic_offset != dynamic_shdr->sh_offset) {
|
|
if (get_application_target_sdk_version() >= 26) {
|
|
DL_ERR_AND_LOG("\"%s\" .dynamic section has invalid offset: 0x%zx, "
|
|
"expected to match PT_DYNAMIC offset: 0x%zx",
|
|
name_.c_str(),
|
|
static_cast<size_t>(dynamic_shdr->sh_offset),
|
|
pt_dynamic_offset);
|
|
return false;
|
|
}
|
|
DL_WARN_documented_change(26,
|
|
"invalid-elf-header_section-headers-enforced-for-api-level-26",
|
|
"\"%s\" .dynamic section has invalid offset: 0x%zx "
|
|
"(expected to match PT_DYNAMIC offset 0x%zx)",
|
|
name_.c_str(),
|
|
static_cast<size_t>(dynamic_shdr->sh_offset),
|
|
pt_dynamic_offset);
|
|
add_dlwarning(name_.c_str(), "invalid .dynamic section");
|
|
}
|
|
|
|
if (pt_dynamic_filesz != dynamic_shdr->sh_size) {
|
|
if (get_application_target_sdk_version() >= 26) {
|
|
DL_ERR_AND_LOG("\"%s\" .dynamic section has invalid size: 0x%zx, "
|
|
"expected to match PT_DYNAMIC filesz: 0x%zx",
|
|
name_.c_str(),
|
|
static_cast<size_t>(dynamic_shdr->sh_size),
|
|
pt_dynamic_filesz);
|
|
return false;
|
|
}
|
|
DL_WARN_documented_change(26,
|
|
"invalid-elf-header_section-headers-enforced-for-api-level-26",
|
|
"\"%s\" .dynamic section has invalid size: 0x%zx "
|
|
"(expected to match PT_DYNAMIC filesz 0x%zx)",
|
|
name_.c_str(),
|
|
static_cast<size_t>(dynamic_shdr->sh_size),
|
|
pt_dynamic_filesz);
|
|
add_dlwarning(name_.c_str(), "invalid .dynamic section");
|
|
}
|
|
|
|
if (dynamic_shdr->sh_link >= shdr_num_) {
|
|
DL_ERR_AND_LOG("\"%s\" .dynamic section has invalid sh_link: %d",
|
|
name_.c_str(),
|
|
dynamic_shdr->sh_link);
|
|
return false;
|
|
}
|
|
|
|
const ElfW(Shdr)* strtab_shdr = &shdr_table_[dynamic_shdr->sh_link];
|
|
|
|
if (strtab_shdr->sh_type != SHT_STRTAB) {
|
|
DL_ERR_AND_LOG("\"%s\" .dynamic section has invalid link(%d) sh_type: %d (expected SHT_STRTAB)",
|
|
name_.c_str(), dynamic_shdr->sh_link, strtab_shdr->sh_type);
|
|
return false;
|
|
}
|
|
|
|
if (!CheckFileRange(dynamic_shdr->sh_offset, dynamic_shdr->sh_size, alignof(const ElfW(Dyn)))) {
|
|
DL_ERR_AND_LOG("\"%s\" has invalid offset/size of .dynamic section", name_.c_str());
|
|
return false;
|
|
}
|
|
|
|
if (!dynamic_fragment_.Map(fd_, file_offset_, dynamic_shdr->sh_offset, dynamic_shdr->sh_size)) {
|
|
DL_ERR("\"%s\" dynamic section mmap failed: %s", name_.c_str(), strerror(errno));
|
|
return false;
|
|
}
|
|
|
|
dynamic_ = static_cast<const ElfW(Dyn)*>(dynamic_fragment_.data());
|
|
|
|
if (!CheckFileRange(strtab_shdr->sh_offset, strtab_shdr->sh_size, alignof(const char))) {
|
|
DL_ERR_AND_LOG("\"%s\" has invalid offset/size of the .strtab section linked from .dynamic section",
|
|
name_.c_str());
|
|
return false;
|
|
}
|
|
|
|
if (!strtab_fragment_.Map(fd_, file_offset_, strtab_shdr->sh_offset, strtab_shdr->sh_size)) {
|
|
DL_ERR("\"%s\" strtab section mmap failed: %s", name_.c_str(), strerror(errno));
|
|
return false;
|
|
}
|
|
|
|
strtab_ = static_cast<const char*>(strtab_fragment_.data());
|
|
strtab_size_ = strtab_fragment_.size();
|
|
return true;
|
|
}
|
|
|
|
/* Returns the size of the extent of all the possibly non-contiguous
|
|
* loadable segments in an ELF program header table. This corresponds
|
|
* to the page-aligned size in bytes that needs to be reserved in the
|
|
* process' address space. If there are no loadable segments, 0 is
|
|
* returned.
|
|
*
|
|
* If out_min_vaddr or out_max_vaddr are not null, they will be
|
|
* set to the minimum and maximum addresses of pages to be reserved,
|
|
* or 0 if there is nothing to load.
|
|
*/
|
|
size_t phdr_table_get_load_size(const ElfW(Phdr)* phdr_table, size_t phdr_count,
|
|
ElfW(Addr)* out_min_vaddr,
|
|
ElfW(Addr)* out_max_vaddr) {
|
|
ElfW(Addr) min_vaddr = UINTPTR_MAX;
|
|
ElfW(Addr) max_vaddr = 0;
|
|
|
|
bool found_pt_load = false;
|
|
for (size_t i = 0; i < phdr_count; ++i) {
|
|
const ElfW(Phdr)* phdr = &phdr_table[i];
|
|
|
|
if (phdr->p_type != PT_LOAD) {
|
|
continue;
|
|
}
|
|
found_pt_load = true;
|
|
|
|
if (phdr->p_vaddr < min_vaddr) {
|
|
min_vaddr = phdr->p_vaddr;
|
|
}
|
|
|
|
if (phdr->p_vaddr + phdr->p_memsz > max_vaddr) {
|
|
max_vaddr = phdr->p_vaddr + phdr->p_memsz;
|
|
}
|
|
}
|
|
if (!found_pt_load) {
|
|
min_vaddr = 0;
|
|
}
|
|
|
|
min_vaddr = page_start(min_vaddr);
|
|
max_vaddr = page_end(max_vaddr);
|
|
|
|
if (out_min_vaddr != nullptr) {
|
|
*out_min_vaddr = min_vaddr;
|
|
}
|
|
if (out_max_vaddr != nullptr) {
|
|
*out_max_vaddr = max_vaddr;
|
|
}
|
|
return max_vaddr - min_vaddr;
|
|
}
|
|
|
|
// Returns the maximum p_align associated with a loadable segment in the ELF
|
|
// program header table. Used to determine whether the file should be loaded at
|
|
// a specific virtual address alignment for use with huge pages.
|
|
size_t phdr_table_get_maximum_alignment(const ElfW(Phdr)* phdr_table, size_t phdr_count) {
|
|
size_t maximum_alignment = page_size();
|
|
|
|
for (size_t i = 0; i < phdr_count; ++i) {
|
|
const ElfW(Phdr)* phdr = &phdr_table[i];
|
|
|
|
// p_align must be 0, 1, or a positive, integral power of two.
|
|
if (phdr->p_type != PT_LOAD || ((phdr->p_align & (phdr->p_align - 1)) != 0)) {
|
|
continue;
|
|
}
|
|
|
|
if (phdr->p_align > maximum_alignment) {
|
|
maximum_alignment = phdr->p_align;
|
|
}
|
|
}
|
|
|
|
#if defined(__LP64__)
|
|
return maximum_alignment;
|
|
#else
|
|
return page_size();
|
|
#endif
|
|
}
|
|
|
|
// Reserve a virtual address range such that if it's limits were extended to the next 2**align
|
|
// boundary, it would not overlap with any existing mappings.
|
|
static void* ReserveWithAlignmentPadding(size_t size, size_t mapping_align, size_t start_align,
|
|
void** out_gap_start, size_t* out_gap_size) {
|
|
int mmap_flags = MAP_PRIVATE | MAP_ANONYMOUS;
|
|
// Reserve enough space to properly align the library's start address.
|
|
mapping_align = std::max(mapping_align, start_align);
|
|
if (mapping_align == page_size()) {
|
|
void* mmap_ptr = mmap(nullptr, size, PROT_NONE, mmap_flags, -1, 0);
|
|
if (mmap_ptr == MAP_FAILED) {
|
|
return nullptr;
|
|
}
|
|
return mmap_ptr;
|
|
}
|
|
|
|
// Minimum alignment of shared library gap. For efficiency, this should match the second level
|
|
// page size of the platform.
|
|
#if defined(__LP64__)
|
|
constexpr size_t kGapAlignment = 1ul << 21; // 2MB
|
|
#else
|
|
constexpr size_t kGapAlignment = 0;
|
|
#endif
|
|
// Maximum gap size, in the units of kGapAlignment.
|
|
constexpr size_t kMaxGapUnits = 32;
|
|
// Allocate enough space so that the end of the desired region aligned up is still inside the
|
|
// mapping.
|
|
size_t mmap_size = align_up(size, mapping_align) + mapping_align - page_size();
|
|
uint8_t* mmap_ptr =
|
|
reinterpret_cast<uint8_t*>(mmap(nullptr, mmap_size, PROT_NONE, mmap_flags, -1, 0));
|
|
if (mmap_ptr == MAP_FAILED) {
|
|
return nullptr;
|
|
}
|
|
size_t gap_size = 0;
|
|
size_t first_byte = reinterpret_cast<size_t>(align_up(mmap_ptr, mapping_align));
|
|
size_t last_byte = reinterpret_cast<size_t>(align_down(mmap_ptr + mmap_size, mapping_align) - 1);
|
|
if (kGapAlignment && first_byte / kGapAlignment != last_byte / kGapAlignment) {
|
|
// This library crosses a 2MB boundary and will fragment a new huge page.
|
|
// Lets take advantage of that and insert a random number of inaccessible huge pages before that
|
|
// to improve address randomization and make it harder to locate this library code by probing.
|
|
munmap(mmap_ptr, mmap_size);
|
|
mapping_align = std::max(mapping_align, kGapAlignment);
|
|
gap_size =
|
|
kGapAlignment * (is_first_stage_init() ? 1 : arc4random_uniform(kMaxGapUnits - 1) + 1);
|
|
mmap_size = align_up(size + gap_size, mapping_align) + mapping_align - page_size();
|
|
mmap_ptr = reinterpret_cast<uint8_t*>(mmap(nullptr, mmap_size, PROT_NONE, mmap_flags, -1, 0));
|
|
if (mmap_ptr == MAP_FAILED) {
|
|
return nullptr;
|
|
}
|
|
}
|
|
|
|
uint8_t *gap_end, *gap_start;
|
|
if (gap_size) {
|
|
gap_end = align_down(mmap_ptr + mmap_size, kGapAlignment);
|
|
gap_start = gap_end - gap_size;
|
|
} else {
|
|
gap_start = gap_end = mmap_ptr + mmap_size;
|
|
}
|
|
|
|
uint8_t* first = align_up(mmap_ptr, mapping_align);
|
|
uint8_t* last = align_down(gap_start, mapping_align) - size;
|
|
|
|
// arc4random* is not available in first stage init because /dev/urandom hasn't yet been
|
|
// created. Don't randomize then.
|
|
size_t n = is_first_stage_init() ? 0 : arc4random_uniform((last - first) / start_align + 1);
|
|
uint8_t* start = first + n * start_align;
|
|
// Unmap the extra space around the allocation.
|
|
// Keep it mapped PROT_NONE on 64-bit targets where address space is plentiful to make it harder
|
|
// to defeat ASLR by probing for readable memory mappings.
|
|
munmap(mmap_ptr, start - mmap_ptr);
|
|
munmap(start + size, gap_start - (start + size));
|
|
if (gap_end != mmap_ptr + mmap_size) {
|
|
munmap(gap_end, mmap_ptr + mmap_size - gap_end);
|
|
}
|
|
*out_gap_start = gap_start;
|
|
*out_gap_size = gap_size;
|
|
return start;
|
|
}
|
|
|
|
// Reserve a virtual address range big enough to hold all loadable
|
|
// segments of a program header table. This is done by creating a
|
|
// private anonymous mmap() with PROT_NONE.
|
|
bool ElfReader::ReserveAddressSpace(address_space_params* address_space) {
|
|
ElfW(Addr) min_vaddr;
|
|
load_size_ = phdr_table_get_load_size(phdr_table_, phdr_num_, &min_vaddr);
|
|
if (load_size_ == 0) {
|
|
DL_ERR("\"%s\" has no loadable segments", name_.c_str());
|
|
return false;
|
|
}
|
|
|
|
uint8_t* addr = reinterpret_cast<uint8_t*>(min_vaddr);
|
|
void* start;
|
|
|
|
if (load_size_ > address_space->reserved_size) {
|
|
if (address_space->must_use_address) {
|
|
DL_ERR("reserved address space %zd smaller than %zd bytes needed for \"%s\"",
|
|
load_size_ - address_space->reserved_size, load_size_, name_.c_str());
|
|
return false;
|
|
}
|
|
size_t start_alignment = page_size();
|
|
if (get_transparent_hugepages_supported() && get_application_target_sdk_version() >= 31) {
|
|
size_t maximum_alignment = phdr_table_get_maximum_alignment(phdr_table_, phdr_num_);
|
|
// Limit alignment to PMD size as other alignments reduce the number of
|
|
// bits available for ASLR for no benefit.
|
|
start_alignment = maximum_alignment == kPmdSize ? kPmdSize : page_size();
|
|
}
|
|
start = ReserveWithAlignmentPadding(load_size_, kLibraryAlignment, start_alignment, &gap_start_,
|
|
&gap_size_);
|
|
if (start == nullptr) {
|
|
DL_ERR("couldn't reserve %zd bytes of address space for \"%s\"", load_size_, name_.c_str());
|
|
return false;
|
|
}
|
|
} else {
|
|
start = address_space->start_addr;
|
|
gap_start_ = nullptr;
|
|
gap_size_ = 0;
|
|
mapped_by_caller_ = true;
|
|
|
|
// Update the reserved address space to subtract the space used by this library.
|
|
address_space->start_addr = reinterpret_cast<uint8_t*>(address_space->start_addr) + load_size_;
|
|
address_space->reserved_size -= load_size_;
|
|
}
|
|
|
|
load_start_ = start;
|
|
load_bias_ = reinterpret_cast<uint8_t*>(start) - addr;
|
|
return true;
|
|
}
|
|
|
|
// Find the ELF note of type NT_ANDROID_TYPE_PAD_SEGMENT and check that the desc value is 1.
|
|
bool ElfReader::ReadPadSegmentNote() {
|
|
// The ELF can have multiple PT_NOTE's, check them all
|
|
for (size_t i = 0; i < phdr_num_; ++i) {
|
|
const ElfW(Phdr)* phdr = &phdr_table_[i];
|
|
|
|
if (phdr->p_type != PT_NOTE) {
|
|
continue;
|
|
}
|
|
|
|
// Some obfuscated ELFs may contain "empty" PT_NOTE program headers that don't
|
|
// point to any part of the ELF (p_memsz == 0). Skip these since there is
|
|
// nothing to decode. See: b/324468126
|
|
if (phdr->p_memsz == 0) {
|
|
continue;
|
|
}
|
|
|
|
// If the PT_NOTE extends beyond the file. The ELF is doing something
|
|
// strange -- obfuscation, embedding hidden loaders, ...
|
|
//
|
|
// It doesn't contain the pad_segment note. Skip it to avoid SIGBUS
|
|
// by accesses beyond the file.
|
|
off64_t note_end_off = file_offset_ + phdr->p_offset + phdr->p_filesz;
|
|
if (note_end_off > file_size_) {
|
|
continue;
|
|
}
|
|
|
|
// note_fragment is scoped to within the loop so that there is
|
|
// at most 1 PT_NOTE mapped at anytime during this search.
|
|
MappedFileFragment note_fragment;
|
|
if (!note_fragment.Map(fd_, file_offset_, phdr->p_offset, phdr->p_memsz)) {
|
|
DL_ERR("\"%s\": PT_NOTE mmap(nullptr, %p, PROT_READ, MAP_PRIVATE, %d, %p) failed: %m",
|
|
name_.c_str(), reinterpret_cast<void*>(phdr->p_memsz), fd_,
|
|
reinterpret_cast<void*>(page_start(file_offset_ + phdr->p_offset)));
|
|
return false;
|
|
}
|
|
|
|
const ElfW(Nhdr)* note_hdr = nullptr;
|
|
const char* note_desc = nullptr;
|
|
if (!__get_elf_note(NT_ANDROID_TYPE_PAD_SEGMENT, "Android",
|
|
reinterpret_cast<ElfW(Addr)>(note_fragment.data()),
|
|
phdr, ¬e_hdr, ¬e_desc)) {
|
|
continue;
|
|
}
|
|
|
|
if (note_hdr->n_descsz != sizeof(ElfW(Word))) {
|
|
DL_ERR("\"%s\" NT_ANDROID_TYPE_PAD_SEGMENT note has unexpected n_descsz: %u",
|
|
name_.c_str(), reinterpret_cast<unsigned int>(note_hdr->n_descsz));
|
|
return false;
|
|
}
|
|
|
|
// 1 == enabled, 0 == disabled
|
|
should_pad_segments_ = *reinterpret_cast<const ElfW(Word)*>(note_desc) == 1;
|
|
return true;
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
static inline void _extend_load_segment_vma(const ElfW(Phdr)* phdr_table, size_t phdr_count,
|
|
size_t phdr_idx, ElfW(Addr)* p_memsz,
|
|
ElfW(Addr)* p_filesz, bool should_pad_segments) {
|
|
const ElfW(Phdr)* phdr = &phdr_table[phdr_idx];
|
|
const ElfW(Phdr)* next = nullptr;
|
|
size_t next_idx = phdr_idx + 1;
|
|
|
|
if (phdr->p_align == kPageSize || !should_pad_segments) {
|
|
return;
|
|
}
|
|
|
|
if (next_idx < phdr_count && phdr_table[next_idx].p_type == PT_LOAD) {
|
|
next = &phdr_table[next_idx];
|
|
}
|
|
|
|
// If this is the last LOAD segment, no extension is needed
|
|
if (!next || *p_memsz != *p_filesz) {
|
|
return;
|
|
}
|
|
|
|
ElfW(Addr) next_start = page_start(next->p_vaddr);
|
|
ElfW(Addr) curr_end = page_end(phdr->p_vaddr + *p_memsz);
|
|
|
|
// If adjacent segment mappings overlap, no extension is needed.
|
|
if (curr_end >= next_start) {
|
|
return;
|
|
}
|
|
|
|
// Extend the LOAD segment mapping to be contiguous with that of
|
|
// the next LOAD segment.
|
|
ElfW(Addr) extend = next_start - curr_end;
|
|
*p_memsz += extend;
|
|
*p_filesz += extend;
|
|
}
|
|
|
|
bool ElfReader::LoadSegments() {
|
|
for (size_t i = 0; i < phdr_num_; ++i) {
|
|
const ElfW(Phdr)* phdr = &phdr_table_[i];
|
|
|
|
if (phdr->p_type != PT_LOAD) {
|
|
continue;
|
|
}
|
|
|
|
ElfW(Addr) p_memsz = phdr->p_memsz;
|
|
ElfW(Addr) p_filesz = phdr->p_filesz;
|
|
_extend_load_segment_vma(phdr_table_, phdr_num_, i, &p_memsz, &p_filesz, should_pad_segments_);
|
|
|
|
// Segment addresses in memory.
|
|
ElfW(Addr) seg_start = phdr->p_vaddr + load_bias_;
|
|
ElfW(Addr) seg_end = seg_start + p_memsz;
|
|
|
|
ElfW(Addr) seg_page_start = page_start(seg_start);
|
|
ElfW(Addr) seg_page_end = page_end(seg_end);
|
|
|
|
ElfW(Addr) seg_file_end = seg_start + p_filesz;
|
|
|
|
// File offsets.
|
|
ElfW(Addr) file_start = phdr->p_offset;
|
|
ElfW(Addr) file_end = file_start + p_filesz;
|
|
|
|
ElfW(Addr) file_page_start = page_start(file_start);
|
|
ElfW(Addr) file_length = file_end - file_page_start;
|
|
|
|
if (file_size_ <= 0) {
|
|
DL_ERR("\"%s\" invalid file size: %" PRId64, name_.c_str(), file_size_);
|
|
return false;
|
|
}
|
|
|
|
if (file_start + phdr->p_filesz > static_cast<size_t>(file_size_)) {
|
|
DL_ERR("invalid ELF file \"%s\" load segment[%zd]:"
|
|
" p_offset (%p) + p_filesz (%p) ( = %p) past end of file (0x%" PRIx64 ")",
|
|
name_.c_str(), i, reinterpret_cast<void*>(phdr->p_offset),
|
|
reinterpret_cast<void*>(phdr->p_filesz),
|
|
reinterpret_cast<void*>(file_start + phdr->p_filesz), file_size_);
|
|
return false;
|
|
}
|
|
|
|
if (file_length != 0) {
|
|
int prot = PFLAGS_TO_PROT(phdr->p_flags);
|
|
if ((prot & (PROT_EXEC | PROT_WRITE)) == (PROT_EXEC | PROT_WRITE)) {
|
|
// W + E PT_LOAD segments are not allowed in O.
|
|
if (get_application_target_sdk_version() >= 26) {
|
|
DL_ERR_AND_LOG("\"%s\": W+E load segments are not allowed", name_.c_str());
|
|
return false;
|
|
}
|
|
DL_WARN_documented_change(26,
|
|
"writable-and-executable-segments-enforced-for-api-level-26",
|
|
"\"%s\" has load segments that are both writable and executable",
|
|
name_.c_str());
|
|
add_dlwarning(name_.c_str(), "W+E load segments");
|
|
}
|
|
|
|
void* seg_addr = mmap64(reinterpret_cast<void*>(seg_page_start),
|
|
file_length,
|
|
prot,
|
|
MAP_FIXED|MAP_PRIVATE,
|
|
fd_,
|
|
file_offset_ + file_page_start);
|
|
if (seg_addr == MAP_FAILED) {
|
|
DL_ERR("couldn't map \"%s\" segment %zd: %s", name_.c_str(), i, strerror(errno));
|
|
return false;
|
|
}
|
|
|
|
// Mark segments as huge page eligible if they meet the requirements
|
|
// (executable and PMD aligned).
|
|
if ((phdr->p_flags & PF_X) && phdr->p_align == kPmdSize &&
|
|
get_transparent_hugepages_supported()) {
|
|
madvise(seg_addr, file_length, MADV_HUGEPAGE);
|
|
}
|
|
}
|
|
|
|
// if the segment is writable, and does not end on a page boundary,
|
|
// zero-fill it until the page limit.
|
|
//
|
|
// Do not attempt to zero the extended region past the first partial page,
|
|
// since doing so may:
|
|
// 1) Result in a SIGBUS, as the region is not backed by the underlying
|
|
// file.
|
|
// 2) Break the COW backing, faulting in new anon pages for a region
|
|
// that will not be used.
|
|
|
|
// _seg_file_end = unextended seg_file_end
|
|
uint64_t _seg_file_end = seg_start + phdr->p_filesz;
|
|
if ((phdr->p_flags & PF_W) != 0 && page_offset(_seg_file_end) > 0) {
|
|
memset(reinterpret_cast<void*>(_seg_file_end), 0, kPageSize - page_offset(_seg_file_end));
|
|
}
|
|
|
|
seg_file_end = page_end(seg_file_end);
|
|
|
|
// seg_file_end is now the first page address after the file
|
|
// content. If seg_end is larger, we need to zero anything
|
|
// between them. This is done by using a private anonymous
|
|
// map for all extra pages.
|
|
if (seg_page_end > seg_file_end) {
|
|
size_t zeromap_size = seg_page_end - seg_file_end;
|
|
void* zeromap = mmap(reinterpret_cast<void*>(seg_file_end),
|
|
zeromap_size,
|
|
PFLAGS_TO_PROT(phdr->p_flags),
|
|
MAP_FIXED|MAP_ANONYMOUS|MAP_PRIVATE,
|
|
-1,
|
|
0);
|
|
if (zeromap == MAP_FAILED) {
|
|
DL_ERR("couldn't zero fill \"%s\" gap: %s", name_.c_str(), strerror(errno));
|
|
return false;
|
|
}
|
|
|
|
prctl(PR_SET_VMA, PR_SET_VMA_ANON_NAME, zeromap, zeromap_size, ".bss");
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
/* Used internally. Used to set the protection bits of all loaded segments
|
|
* with optional extra flags (i.e. really PROT_WRITE). Used by
|
|
* phdr_table_protect_segments and phdr_table_unprotect_segments.
|
|
*/
|
|
static int _phdr_table_set_load_prot(const ElfW(Phdr)* phdr_table, size_t phdr_count,
|
|
ElfW(Addr) load_bias, int extra_prot_flags,
|
|
bool should_pad_segments) {
|
|
for (size_t i = 0; i < phdr_count; ++i) {
|
|
const ElfW(Phdr)* phdr = &phdr_table[i];
|
|
|
|
if (phdr->p_type != PT_LOAD || (phdr->p_flags & PF_W) != 0) {
|
|
continue;
|
|
}
|
|
|
|
ElfW(Addr) p_memsz = phdr->p_memsz;
|
|
ElfW(Addr) p_filesz = phdr->p_filesz;
|
|
_extend_load_segment_vma(phdr_table, phdr_count, i, &p_memsz, &p_filesz, should_pad_segments);
|
|
|
|
ElfW(Addr) seg_page_start = page_start(phdr->p_vaddr + load_bias);
|
|
ElfW(Addr) seg_page_end = page_end(phdr->p_vaddr + p_memsz + load_bias);
|
|
|
|
int prot = PFLAGS_TO_PROT(phdr->p_flags) | extra_prot_flags;
|
|
if ((prot & PROT_WRITE) != 0) {
|
|
// make sure we're never simultaneously writable / executable
|
|
prot &= ~PROT_EXEC;
|
|
}
|
|
#if defined(__aarch64__)
|
|
if ((prot & PROT_EXEC) == 0) {
|
|
// Though it is not specified don't add PROT_BTI if segment is not
|
|
// executable.
|
|
prot &= ~PROT_BTI;
|
|
}
|
|
#endif
|
|
|
|
int ret =
|
|
mprotect(reinterpret_cast<void*>(seg_page_start), seg_page_end - seg_page_start, prot);
|
|
if (ret < 0) {
|
|
return -1;
|
|
}
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
/* Restore the original protection modes for all loadable segments.
|
|
* You should only call this after phdr_table_unprotect_segments and
|
|
* applying all relocations.
|
|
*
|
|
* AArch64: also called from linker_main and ElfReader::Load to apply
|
|
* PROT_BTI for loaded main so and other so-s.
|
|
*
|
|
* Input:
|
|
* phdr_table -> program header table
|
|
* phdr_count -> number of entries in tables
|
|
* load_bias -> load bias
|
|
* should_pad_segments -> Are segments extended to avoid gaps in the memory map
|
|
* prop -> GnuPropertySection or nullptr
|
|
* Return:
|
|
* 0 on success, -1 on failure (error code in errno).
|
|
*/
|
|
int phdr_table_protect_segments(const ElfW(Phdr)* phdr_table, size_t phdr_count,
|
|
ElfW(Addr) load_bias, bool should_pad_segments,
|
|
const GnuPropertySection* prop __unused) {
|
|
int prot = 0;
|
|
#if defined(__aarch64__)
|
|
if ((prop != nullptr) && prop->IsBTICompatible()) {
|
|
prot |= PROT_BTI;
|
|
}
|
|
#endif
|
|
return _phdr_table_set_load_prot(phdr_table, phdr_count, load_bias, prot, should_pad_segments);
|
|
}
|
|
|
|
/* Change the protection of all loaded segments in memory to writable.
|
|
* This is useful before performing relocations. Once completed, you
|
|
* will have to call phdr_table_protect_segments to restore the original
|
|
* protection flags on all segments.
|
|
*
|
|
* Note that some writable segments can also have their content turned
|
|
* to read-only by calling phdr_table_protect_gnu_relro. This is no
|
|
* performed here.
|
|
*
|
|
* Input:
|
|
* phdr_table -> program header table
|
|
* phdr_count -> number of entries in tables
|
|
* load_bias -> load bias
|
|
* should_pad_segments -> Are segments extended to avoid gaps in the memory map
|
|
* Return:
|
|
* 0 on success, -1 on failure (error code in errno).
|
|
*/
|
|
int phdr_table_unprotect_segments(const ElfW(Phdr)* phdr_table,
|
|
size_t phdr_count, ElfW(Addr) load_bias,
|
|
bool should_pad_segments) {
|
|
return _phdr_table_set_load_prot(phdr_table, phdr_count, load_bias, PROT_WRITE,
|
|
should_pad_segments);
|
|
}
|
|
|
|
static inline void _extend_gnu_relro_prot_end(const ElfW(Phdr)* relro_phdr,
|
|
const ElfW(Phdr)* phdr_table, size_t phdr_count,
|
|
ElfW(Addr) load_bias, ElfW(Addr)* seg_page_end,
|
|
bool should_pad_segments) {
|
|
// Find the index and phdr of the LOAD containing the GNU_RELRO segment
|
|
for (size_t index = 0; index < phdr_count; ++index) {
|
|
const ElfW(Phdr)* phdr = &phdr_table[index];
|
|
|
|
if (phdr->p_type == PT_LOAD && phdr->p_vaddr == relro_phdr->p_vaddr) {
|
|
// If the PT_GNU_RELRO mem size is not at least as large as the corresponding
|
|
// LOAD segment mem size, we need to protect only a partial region of the
|
|
// LOAD segment and therefore cannot avoid a VMA split.
|
|
//
|
|
// Note: Don't check the page-aligned mem sizes since the extended protection
|
|
// may incorrectly write protect non-relocation data.
|
|
//
|
|
// Example:
|
|
//
|
|
// |---- 3K ----|-- 1K --|---- 3K ---- |-- 1K --|
|
|
// ----------------------------------------------------------------
|
|
// | | | | |
|
|
// SEG X | RO | RO | RW | | SEG Y
|
|
// | | | | |
|
|
// ----------------------------------------------------------------
|
|
// | | |
|
|
// | | |
|
|
// | | |
|
|
// relro_vaddr relro_vaddr relro_vaddr
|
|
// (load_vaddr) + +
|
|
// relro_memsz load_memsz
|
|
//
|
|
// ----------------------------------------------------------------
|
|
// | PAGE | PAGE |
|
|
// ----------------------------------------------------------------
|
|
// | Potential |
|
|
// |----- Extended RO ----|
|
|
// | Protection |
|
|
//
|
|
// If the check below uses page aligned mem sizes it will cause incorrect write
|
|
// protection of the 3K RW part of the LOAD segment containing the GNU_RELRO.
|
|
if (relro_phdr->p_memsz < phdr->p_memsz) {
|
|
return;
|
|
}
|
|
|
|
ElfW(Addr) p_memsz = phdr->p_memsz;
|
|
ElfW(Addr) p_filesz = phdr->p_filesz;
|
|
|
|
// Attempt extending the VMA (mprotect range). Without extending the range,
|
|
// mprotect will only RO protect a part of the extended RW LOAD segment, which
|
|
// will leave an extra split RW VMA (the gap).
|
|
_extend_load_segment_vma(phdr_table, phdr_count, index, &p_memsz, &p_filesz,
|
|
should_pad_segments);
|
|
|
|
*seg_page_end = page_end(phdr->p_vaddr + p_memsz + load_bias);
|
|
return;
|
|
}
|
|
}
|
|
}
|
|
|
|
/* Used internally by phdr_table_protect_gnu_relro and
|
|
* phdr_table_unprotect_gnu_relro.
|
|
*/
|
|
static int _phdr_table_set_gnu_relro_prot(const ElfW(Phdr)* phdr_table, size_t phdr_count,
|
|
ElfW(Addr) load_bias, int prot_flags,
|
|
bool should_pad_segments) {
|
|
const ElfW(Phdr)* phdr = phdr_table;
|
|
const ElfW(Phdr)* phdr_limit = phdr + phdr_count;
|
|
|
|
for (phdr = phdr_table; phdr < phdr_limit; phdr++) {
|
|
if (phdr->p_type != PT_GNU_RELRO) {
|
|
continue;
|
|
}
|
|
|
|
// Tricky: what happens when the relro segment does not start
|
|
// or end at page boundaries? We're going to be over-protective
|
|
// here and put every page touched by the segment as read-only.
|
|
|
|
// This seems to match Ian Lance Taylor's description of the
|
|
// feature at http://www.airs.com/blog/archives/189.
|
|
|
|
// Extract:
|
|
// Note that the current dynamic linker code will only work
|
|
// correctly if the PT_GNU_RELRO segment starts on a page
|
|
// boundary. This is because the dynamic linker rounds the
|
|
// p_vaddr field down to the previous page boundary. If
|
|
// there is anything on the page which should not be read-only,
|
|
// the program is likely to fail at runtime. So in effect the
|
|
// linker must only emit a PT_GNU_RELRO segment if it ensures
|
|
// that it starts on a page boundary.
|
|
ElfW(Addr) seg_page_start = page_start(phdr->p_vaddr) + load_bias;
|
|
ElfW(Addr) seg_page_end = page_end(phdr->p_vaddr + phdr->p_memsz) + load_bias;
|
|
_extend_gnu_relro_prot_end(phdr, phdr_table, phdr_count, load_bias, &seg_page_end,
|
|
should_pad_segments);
|
|
|
|
int ret = mprotect(reinterpret_cast<void*>(seg_page_start),
|
|
seg_page_end - seg_page_start,
|
|
prot_flags);
|
|
if (ret < 0) {
|
|
return -1;
|
|
}
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
/* Apply GNU relro protection if specified by the program header. This will
|
|
* turn some of the pages of a writable PT_LOAD segment to read-only, as
|
|
* specified by one or more PT_GNU_RELRO segments. This must be always
|
|
* performed after relocations.
|
|
*
|
|
* The areas typically covered are .got and .data.rel.ro, these are
|
|
* read-only from the program's POV, but contain absolute addresses
|
|
* that need to be relocated before use.
|
|
*
|
|
* Input:
|
|
* phdr_table -> program header table
|
|
* phdr_count -> number of entries in tables
|
|
* load_bias -> load bias
|
|
* should_pad_segments -> Were segments extended to avoid gaps in the memory map
|
|
* Return:
|
|
* 0 on success, -1 on failure (error code in errno).
|
|
*/
|
|
int phdr_table_protect_gnu_relro(const ElfW(Phdr)* phdr_table, size_t phdr_count,
|
|
ElfW(Addr) load_bias, bool should_pad_segments) {
|
|
return _phdr_table_set_gnu_relro_prot(phdr_table, phdr_count, load_bias, PROT_READ,
|
|
should_pad_segments);
|
|
}
|
|
|
|
/* Serialize the GNU relro segments to the given file descriptor. This can be
|
|
* performed after relocations to allow another process to later share the
|
|
* relocated segment, if it was loaded at the same address.
|
|
*
|
|
* Input:
|
|
* phdr_table -> program header table
|
|
* phdr_count -> number of entries in tables
|
|
* load_bias -> load bias
|
|
* fd -> writable file descriptor to use
|
|
* file_offset -> pointer to offset into file descriptor to use/update
|
|
* Return:
|
|
* 0 on success, -1 on failure (error code in errno).
|
|
*/
|
|
int phdr_table_serialize_gnu_relro(const ElfW(Phdr)* phdr_table,
|
|
size_t phdr_count,
|
|
ElfW(Addr) load_bias,
|
|
int fd,
|
|
size_t* file_offset) {
|
|
const ElfW(Phdr)* phdr = phdr_table;
|
|
const ElfW(Phdr)* phdr_limit = phdr + phdr_count;
|
|
|
|
for (phdr = phdr_table; phdr < phdr_limit; phdr++) {
|
|
if (phdr->p_type != PT_GNU_RELRO) {
|
|
continue;
|
|
}
|
|
|
|
ElfW(Addr) seg_page_start = page_start(phdr->p_vaddr) + load_bias;
|
|
ElfW(Addr) seg_page_end = page_end(phdr->p_vaddr + phdr->p_memsz) + load_bias;
|
|
ssize_t size = seg_page_end - seg_page_start;
|
|
|
|
ssize_t written = TEMP_FAILURE_RETRY(write(fd, reinterpret_cast<void*>(seg_page_start), size));
|
|
if (written != size) {
|
|
return -1;
|
|
}
|
|
void* map = mmap(reinterpret_cast<void*>(seg_page_start), size, PROT_READ,
|
|
MAP_PRIVATE|MAP_FIXED, fd, *file_offset);
|
|
if (map == MAP_FAILED) {
|
|
return -1;
|
|
}
|
|
*file_offset += size;
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
/* Where possible, replace the GNU relro segments with mappings of the given
|
|
* file descriptor. This can be performed after relocations to allow a file
|
|
* previously created by phdr_table_serialize_gnu_relro in another process to
|
|
* replace the dirty relocated pages, saving memory, if it was loaded at the
|
|
* same address. We have to compare the data before we map over it, since some
|
|
* parts of the relro segment may not be identical due to other libraries in
|
|
* the process being loaded at different addresses.
|
|
*
|
|
* Input:
|
|
* phdr_table -> program header table
|
|
* phdr_count -> number of entries in tables
|
|
* load_bias -> load bias
|
|
* fd -> readable file descriptor to use
|
|
* file_offset -> pointer to offset into file descriptor to use/update
|
|
* Return:
|
|
* 0 on success, -1 on failure (error code in errno).
|
|
*/
|
|
int phdr_table_map_gnu_relro(const ElfW(Phdr)* phdr_table,
|
|
size_t phdr_count,
|
|
ElfW(Addr) load_bias,
|
|
int fd,
|
|
size_t* file_offset) {
|
|
// Map the file at a temporary location so we can compare its contents.
|
|
struct stat file_stat;
|
|
if (TEMP_FAILURE_RETRY(fstat(fd, &file_stat)) != 0) {
|
|
return -1;
|
|
}
|
|
off_t file_size = file_stat.st_size;
|
|
void* temp_mapping = nullptr;
|
|
if (file_size > 0) {
|
|
temp_mapping = mmap(nullptr, file_size, PROT_READ, MAP_PRIVATE, fd, 0);
|
|
if (temp_mapping == MAP_FAILED) {
|
|
return -1;
|
|
}
|
|
}
|
|
|
|
// Iterate over the relro segments and compare/remap the pages.
|
|
const ElfW(Phdr)* phdr = phdr_table;
|
|
const ElfW(Phdr)* phdr_limit = phdr + phdr_count;
|
|
|
|
for (phdr = phdr_table; phdr < phdr_limit; phdr++) {
|
|
if (phdr->p_type != PT_GNU_RELRO) {
|
|
continue;
|
|
}
|
|
|
|
ElfW(Addr) seg_page_start = page_start(phdr->p_vaddr) + load_bias;
|
|
ElfW(Addr) seg_page_end = page_end(phdr->p_vaddr + phdr->p_memsz) + load_bias;
|
|
|
|
char* file_base = static_cast<char*>(temp_mapping) + *file_offset;
|
|
char* mem_base = reinterpret_cast<char*>(seg_page_start);
|
|
size_t match_offset = 0;
|
|
size_t size = seg_page_end - seg_page_start;
|
|
|
|
if (file_size - *file_offset < size) {
|
|
// File is too short to compare to this segment. The contents are likely
|
|
// different as well (it's probably for a different library version) so
|
|
// just don't bother checking.
|
|
break;
|
|
}
|
|
|
|
while (match_offset < size) {
|
|
// Skip over dissimilar pages.
|
|
while (match_offset < size &&
|
|
memcmp(mem_base + match_offset, file_base + match_offset, page_size()) != 0) {
|
|
match_offset += page_size();
|
|
}
|
|
|
|
// Count similar pages.
|
|
size_t mismatch_offset = match_offset;
|
|
while (mismatch_offset < size &&
|
|
memcmp(mem_base + mismatch_offset, file_base + mismatch_offset, page_size()) == 0) {
|
|
mismatch_offset += page_size();
|
|
}
|
|
|
|
// Map over similar pages.
|
|
if (mismatch_offset > match_offset) {
|
|
void* map = mmap(mem_base + match_offset, mismatch_offset - match_offset,
|
|
PROT_READ, MAP_PRIVATE|MAP_FIXED, fd, *file_offset + match_offset);
|
|
if (map == MAP_FAILED) {
|
|
munmap(temp_mapping, file_size);
|
|
return -1;
|
|
}
|
|
}
|
|
|
|
match_offset = mismatch_offset;
|
|
}
|
|
|
|
// Add to the base file offset in case there are multiple relro segments.
|
|
*file_offset += size;
|
|
}
|
|
munmap(temp_mapping, file_size);
|
|
return 0;
|
|
}
|
|
|
|
|
|
#if defined(__arm__)
|
|
/* Return the address and size of the .ARM.exidx section in memory,
|
|
* if present.
|
|
*
|
|
* Input:
|
|
* phdr_table -> program header table
|
|
* phdr_count -> number of entries in tables
|
|
* load_bias -> load bias
|
|
* Output:
|
|
* arm_exidx -> address of table in memory (null on failure).
|
|
* arm_exidx_count -> number of items in table (0 on failure).
|
|
* Return:
|
|
* 0 on success, -1 on failure (_no_ error code in errno)
|
|
*/
|
|
int phdr_table_get_arm_exidx(const ElfW(Phdr)* phdr_table, size_t phdr_count,
|
|
ElfW(Addr) load_bias,
|
|
ElfW(Addr)** arm_exidx, size_t* arm_exidx_count) {
|
|
const ElfW(Phdr)* phdr = phdr_table;
|
|
const ElfW(Phdr)* phdr_limit = phdr + phdr_count;
|
|
|
|
for (phdr = phdr_table; phdr < phdr_limit; phdr++) {
|
|
if (phdr->p_type != PT_ARM_EXIDX) {
|
|
continue;
|
|
}
|
|
|
|
*arm_exidx = reinterpret_cast<ElfW(Addr)*>(load_bias + phdr->p_vaddr);
|
|
*arm_exidx_count = phdr->p_memsz / 8;
|
|
return 0;
|
|
}
|
|
*arm_exidx = nullptr;
|
|
*arm_exidx_count = 0;
|
|
return -1;
|
|
}
|
|
#endif
|
|
|
|
/* Return the address and size of the ELF file's .dynamic section in memory,
|
|
* or null if missing.
|
|
*
|
|
* Input:
|
|
* phdr_table -> program header table
|
|
* phdr_count -> number of entries in tables
|
|
* load_bias -> load bias
|
|
* Output:
|
|
* dynamic -> address of table in memory (null on failure).
|
|
* dynamic_flags -> protection flags for section (unset on failure)
|
|
* Return:
|
|
* void
|
|
*/
|
|
void phdr_table_get_dynamic_section(const ElfW(Phdr)* phdr_table, size_t phdr_count,
|
|
ElfW(Addr) load_bias, ElfW(Dyn)** dynamic,
|
|
ElfW(Word)* dynamic_flags) {
|
|
*dynamic = nullptr;
|
|
for (size_t i = 0; i<phdr_count; ++i) {
|
|
const ElfW(Phdr)& phdr = phdr_table[i];
|
|
if (phdr.p_type == PT_DYNAMIC) {
|
|
*dynamic = reinterpret_cast<ElfW(Dyn)*>(load_bias + phdr.p_vaddr);
|
|
if (dynamic_flags) {
|
|
*dynamic_flags = phdr.p_flags;
|
|
}
|
|
return;
|
|
}
|
|
}
|
|
}
|
|
|
|
/* Return the program interpreter string, or nullptr if missing.
|
|
*
|
|
* Input:
|
|
* phdr_table -> program header table
|
|
* phdr_count -> number of entries in tables
|
|
* load_bias -> load bias
|
|
* Return:
|
|
* pointer to the program interpreter string.
|
|
*/
|
|
const char* phdr_table_get_interpreter_name(const ElfW(Phdr)* phdr_table, size_t phdr_count,
|
|
ElfW(Addr) load_bias) {
|
|
for (size_t i = 0; i<phdr_count; ++i) {
|
|
const ElfW(Phdr)& phdr = phdr_table[i];
|
|
if (phdr.p_type == PT_INTERP) {
|
|
return reinterpret_cast<const char*>(load_bias + phdr.p_vaddr);
|
|
}
|
|
}
|
|
return nullptr;
|
|
}
|
|
|
|
// Sets loaded_phdr_ to the address of the program header table as it appears
|
|
// in the loaded segments in memory. This is in contrast with phdr_table_,
|
|
// which is temporary and will be released before the library is relocated.
|
|
bool ElfReader::FindPhdr() {
|
|
const ElfW(Phdr)* phdr_limit = phdr_table_ + phdr_num_;
|
|
|
|
// If there is a PT_PHDR, use it directly.
|
|
for (const ElfW(Phdr)* phdr = phdr_table_; phdr < phdr_limit; ++phdr) {
|
|
if (phdr->p_type == PT_PHDR) {
|
|
return CheckPhdr(load_bias_ + phdr->p_vaddr);
|
|
}
|
|
}
|
|
|
|
// Otherwise, check the first loadable segment. If its file offset
|
|
// is 0, it starts with the ELF header, and we can trivially find the
|
|
// loaded program header from it.
|
|
for (const ElfW(Phdr)* phdr = phdr_table_; phdr < phdr_limit; ++phdr) {
|
|
if (phdr->p_type == PT_LOAD) {
|
|
if (phdr->p_offset == 0) {
|
|
ElfW(Addr) elf_addr = load_bias_ + phdr->p_vaddr;
|
|
const ElfW(Ehdr)* ehdr = reinterpret_cast<const ElfW(Ehdr)*>(elf_addr);
|
|
ElfW(Addr) offset = ehdr->e_phoff;
|
|
return CheckPhdr(reinterpret_cast<ElfW(Addr)>(ehdr) + offset);
|
|
}
|
|
break;
|
|
}
|
|
}
|
|
|
|
DL_ERR("can't find loaded phdr for \"%s\"", name_.c_str());
|
|
return false;
|
|
}
|
|
|
|
// Tries to find .note.gnu.property section.
|
|
// It is not considered an error if such section is missing.
|
|
bool ElfReader::FindGnuPropertySection() {
|
|
#if defined(__aarch64__)
|
|
note_gnu_property_ = GnuPropertySection(phdr_table_, phdr_num_, load_start(), name_.c_str());
|
|
#endif
|
|
return true;
|
|
}
|
|
|
|
// Ensures that our program header is actually within a loadable
|
|
// segment. This should help catch badly-formed ELF files that
|
|
// would cause the linker to crash later when trying to access it.
|
|
bool ElfReader::CheckPhdr(ElfW(Addr) loaded) {
|
|
const ElfW(Phdr)* phdr_limit = phdr_table_ + phdr_num_;
|
|
ElfW(Addr) loaded_end = loaded + (phdr_num_ * sizeof(ElfW(Phdr)));
|
|
for (const ElfW(Phdr)* phdr = phdr_table_; phdr < phdr_limit; ++phdr) {
|
|
if (phdr->p_type != PT_LOAD) {
|
|
continue;
|
|
}
|
|
ElfW(Addr) seg_start = phdr->p_vaddr + load_bias_;
|
|
ElfW(Addr) seg_end = phdr->p_filesz + seg_start;
|
|
if (seg_start <= loaded && loaded_end <= seg_end) {
|
|
loaded_phdr_ = reinterpret_cast<const ElfW(Phdr)*>(loaded);
|
|
return true;
|
|
}
|
|
}
|
|
DL_ERR("\"%s\" loaded phdr %p not in loadable segment",
|
|
name_.c_str(), reinterpret_cast<void*>(loaded));
|
|
return false;
|
|
}
|