diff options
Diffstat (limited to 'toolkit/crashreporter/process_reader/src/platform')
3 files changed, 748 insertions, 0 deletions
diff --git a/toolkit/crashreporter/process_reader/src/platform/linux.rs b/toolkit/crashreporter/process_reader/src/platform/linux.rs new file mode 100644 index 0000000000..4a87e5480b --- /dev/null +++ b/toolkit/crashreporter/process_reader/src/platform/linux.rs @@ -0,0 +1,322 @@ +/* This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at https://mozilla.org/MPL/2.0/. */ + +use std::{ + cmp::min, + fs::File, + io::{BufRead, BufReader, Error}, + mem::{size_of, MaybeUninit}, + ptr::null_mut, + slice, +}; + +use crate::{ + error::{ProcessReaderError, PtraceError, ReadError}, + ProcessHandle, ProcessReader, +}; + +use goblin::elf::{ + self, + program_header::{PF_R, PT_NOTE}, + Elf, ProgramHeader, +}; +use libc::{ + c_int, c_long, c_void, pid_t, ptrace, waitpid, EINTR, PTRACE_ATTACH, PTRACE_DETACH, + PTRACE_PEEKDATA, __WALL, +}; + +impl ProcessReader { + pub fn new(process: ProcessHandle) -> Result<ProcessReader, ProcessReaderError> { + let pid: pid_t = process; + + ptrace_attach(pid)?; + + let mut status: i32 = 0; + + loop { + let res = unsafe { waitpid(pid, &mut status as *mut _, __WALL) }; + if res < 0 { + match get_errno() { + EINTR => continue, + _ => { + ptrace_detach(pid)?; + return Err(ProcessReaderError::WaitPidError); + } + } + } else { + break; + } + } + + Ok(ProcessReader { process: pid }) + } + + pub fn find_module(&self, module_name: &str) -> Result<usize, ProcessReaderError> { + let maps_file = File::open(format!("/proc/{}/maps", self.process))?; + + BufReader::new(maps_file) + .lines() + .flatten() + .map(|line| parse_proc_maps_line(&line)) + .filter_map(Result::ok) + .find_map(|(name, address)| { + if name.is_some_and(|name| name.eq(module_name)) { + Some(address) + } else { + None + } + }) + .ok_or(ProcessReaderError::ModuleNotFound) + } + + pub fn find_program_note( + &self, + module_address: usize, + note_type: u32, + note_size: usize, + ) -> Result<usize, ProcessReaderError> { + let header_bytes = self.copy_array(module_address, size_of::<elf::Header>())?; + let elf_header = Elf::parse_header(&header_bytes)?; + + let program_header_bytes = self.copy_array( + module_address + (elf_header.e_phoff as usize), + (elf_header.e_phnum as usize) * (elf_header.e_phentsize as usize), + )?; + + let mut elf = Elf::lazy_parse(elf_header)?; + let context = goblin::container::Ctx { + container: elf.header.container()?, + le: elf.header.endianness()?, + }; + + elf.program_headers = ProgramHeader::parse( + &program_header_bytes, + 0, + elf_header.e_phnum as usize, + context, + )?; + + self.find_note_in_headers(&elf, module_address, note_type, note_size) + } + + fn find_note_in_headers( + &self, + elf: &Elf, + address: usize, + note_type: u32, + note_size: usize, + ) -> Result<usize, ProcessReaderError> { + for program_header in elf.program_headers.iter() { + // We're looking for a note in the program headers, it needs to be + // readable and it needs to be at least as large as the + // requested size. + if (program_header.p_type == PT_NOTE) + && ((program_header.p_flags & PF_R) != 0 + && (program_header.p_memsz as usize >= note_size)) + { + // Iterate over the notes + let notes_address = address + program_header.p_offset as usize; + let mut notes_offset = 0; + let notes_size = program_header.p_memsz as usize; + while notes_offset < notes_size { + let note_address = notes_address + notes_offset; + if let Ok(note) = self.copy_object::<goblin::elf::note::Nhdr32>(note_address) { + if note.n_type == note_type { + return Ok(note_address); + } + + notes_offset += size_of::<goblin::elf::note::Nhdr32>() + + (note.n_descsz as usize) + + (note.n_namesz as usize); + } else { + break; + } + } + } + } + + Err(ProcessReaderError::NoteNotFound) + } + + pub fn copy_object_shallow<T>(&self, src: usize) -> Result<MaybeUninit<T>, ReadError> { + let data = self.copy_array(src, size_of::<T>())?; + let mut object = MaybeUninit::<T>::uninit(); + let uninitialized_object = uninit_as_bytes_mut(&mut object); + + for (index, &value) in data.iter().enumerate() { + uninitialized_object[index].write(value); + } + + Ok(object) + } + + pub fn copy_object<T>(&self, src: usize) -> Result<T, ReadError> { + self.copy_object_shallow(src) + .map(|object| unsafe { object.assume_init() }) + } + + pub fn copy_array<T>(&self, src: usize, num: usize) -> Result<Vec<T>, ReadError> { + let mut array = Vec::<MaybeUninit<T>>::with_capacity(num); + let num_bytes = num * size_of::<T>(); + let mut array_buffer = array.as_mut_ptr() as *mut u8; + let mut index = 0; + + while index < num_bytes { + let word = ptrace_read(self.process, src + index)?; + let len = min(size_of::<c_long>(), num_bytes - index); + let word_as_bytes = word.to_ne_bytes(); + for &byte in word_as_bytes.iter().take(len) { + unsafe { + array_buffer.write(byte); + array_buffer = array_buffer.add(1); + } + } + + index += size_of::<c_long>(); + } + + unsafe { + array.set_len(num); + Ok(std::mem::transmute(array)) + } + } +} + +impl Drop for ProcessReader { + fn drop(&mut self) { + let _ignored = ptrace_detach(self.process); + } +} + +fn parse_proc_maps_line(line: &str) -> Result<(Option<String>, usize), ProcessReaderError> { + let mut splits = line.trim().splitn(6, ' '); + let address_str = splits + .next() + .ok_or(ProcessReaderError::ProcMapsParseError)?; + let _perms_str = splits + .next() + .ok_or(ProcessReaderError::ProcMapsParseError)?; + let _offset_str = splits + .next() + .ok_or(ProcessReaderError::ProcMapsParseError)?; + let _dev_str = splits + .next() + .ok_or(ProcessReaderError::ProcMapsParseError)?; + let _inode_str = splits + .next() + .ok_or(ProcessReaderError::ProcMapsParseError)?; + let path_str = splits + .next() + .ok_or(ProcessReaderError::ProcMapsParseError)?; + + let address = get_proc_maps_address(address_str)?; + + // Note that we don't care if the mapped file has been deleted because + // we're reading everything from memory. + let name = path_str + .trim_end_matches(" (deleted)") + .rsplit('/') + .next() + .map(String::from); + + Ok((name, address)) +} + +fn get_proc_maps_address(addresses: &str) -> Result<usize, ProcessReaderError> { + let begin = addresses + .split('-') + .next() + .ok_or(ProcessReaderError::ProcMapsParseError)?; + usize::from_str_radix(begin, 16).map_err(ProcessReaderError::from) +} + +fn uninit_as_bytes_mut<T>(elem: &mut MaybeUninit<T>) -> &mut [MaybeUninit<u8>] { + // SAFETY: MaybeUninit<u8> is always valid, even for padding bytes + unsafe { slice::from_raw_parts_mut(elem.as_mut_ptr() as *mut MaybeUninit<u8>, size_of::<T>()) } +} + +/*********************************************************************** + ***** libc helpers ***** + ***********************************************************************/ + +fn get_errno() -> c_int { + #[cfg(target_os = "linux")] + unsafe { + *libc::__errno_location() + } + #[cfg(target_os = "android")] + unsafe { + *libc::__errno() + } +} + +fn clear_errno() { + #[cfg(target_os = "linux")] + unsafe { + *libc::__errno_location() = 0; + } + #[cfg(target_os = "android")] + unsafe { + *libc::__errno() = 0; + } +} + +#[derive(Clone, Copy)] +enum PTraceOperation { + Attach, + Detach, + PeekData, +} + +#[cfg(all(target_os = "linux", target_env = "gnu"))] +type PTraceOperationNative = libc::c_uint; +#[cfg(all(target_os = "linux", target_env = "musl"))] +type PTraceOperationNative = libc::c_int; +#[cfg(target_os = "android")] +type PTraceOperationNative = c_int; + +impl From<PTraceOperation> for PTraceOperationNative { + fn from(val: PTraceOperation) -> Self { + match val { + PTraceOperation::Attach => PTRACE_ATTACH, + PTraceOperation::Detach => PTRACE_DETACH, + PTraceOperation::PeekData => PTRACE_PEEKDATA, + } + } +} + +fn ptrace_attach(pid: pid_t) -> Result<(), PtraceError> { + ptrace_helper(pid, PTraceOperation::Attach, 0).map(|_r| ()) +} + +fn ptrace_detach(pid: pid_t) -> Result<(), PtraceError> { + ptrace_helper(pid, PTraceOperation::Detach, 0).map(|_r| ()) +} + +fn ptrace_read(pid: libc::pid_t, addr: usize) -> Result<c_long, PtraceError> { + ptrace_helper(pid, PTraceOperation::PeekData, addr) +} + +fn ptrace_helper(pid: pid_t, op: PTraceOperation, addr: usize) -> Result<c_long, PtraceError> { + clear_errno(); + let result = unsafe { ptrace(op.into(), pid, addr, null_mut::<c_void>()) }; + + if result == -1 { + let errno = get_errno(); + if errno != 0 { + let error = match op { + PTraceOperation::Attach => PtraceError::TraceError(Error::from_raw_os_error(errno)), + PTraceOperation::Detach => PtraceError::TraceError(Error::from_raw_os_error(errno)), + PTraceOperation::PeekData => { + PtraceError::ReadError(Error::from_raw_os_error(errno)) + } + }; + Err(error) + } else { + Ok(result) + } + } else { + Ok(result) + } +} diff --git a/toolkit/crashreporter/process_reader/src/platform/macos.rs b/toolkit/crashreporter/process_reader/src/platform/macos.rs new file mode 100644 index 0000000000..a0dfa2b8fd --- /dev/null +++ b/toolkit/crashreporter/process_reader/src/platform/macos.rs @@ -0,0 +1,225 @@ +/* This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at https://mozilla.org/MPL/2.0/. */ + +use goblin::mach::{ + header::{Header64, MH_DYLIB, MH_EXECUTE, MH_MAGIC_64}, + load_command::{LoadCommandHeader, Section64, SegmentCommand64, LC_SEGMENT_64}, +}; +use mach2::{ + kern_return::KERN_SUCCESS, + task::task_info, + task_info::{task_dyld_info, TASK_DYLD_ALL_IMAGE_INFO_64, TASK_DYLD_INFO}, + vm::mach_vm_read_overwrite, +}; +use std::mem::{size_of, MaybeUninit}; + +use crate::{ + error::{ProcessReaderError, ReadError}, + ProcessHandle, ProcessReader, +}; + +#[repr(C)] +#[derive(Copy, Clone, Debug)] +struct AllImagesInfo { + // VERSION 1 + pub version: u32, + /// The number of [`ImageInfo`] structs at that following address + info_array_count: u32, + /// The address in the process where the array of [`ImageInfo`] structs is + info_array_addr: u64, + /// A function pointer, unused + _notification: u64, + /// Unused + _process_detached_from_shared_region: bool, + // VERSION 2 + lib_system_initialized: bool, + // Note that crashpad adds a 32-bit int here to get proper alignment when + // building on 32-bit targets...but we explicitly don't care about 32-bit + // targets since Apple doesn't + pub dyld_image_load_address: u64, +} + +/// `dyld_image_info` from <usr/include/mach-o/dyld_images.h> +#[repr(C)] +#[derive(Debug, Clone, Copy)] +struct ImageInfo { + /// The address in the process where the image is loaded + pub load_address: u64, + /// The address in the process where the image's file path can be read + pub file_path: u64, + /// Timestamp for when the image's file was last modified + pub file_mod_date: u64, +} + +const DATA_SEGMENT: &[u8; 16] = b"__DATA\0\0\0\0\0\0\0\0\0\0"; + +impl ProcessReader { + pub fn new(process: ProcessHandle) -> Result<ProcessReader, ProcessReaderError> { + Ok(ProcessReader { process }) + } + + pub fn find_module(&self, module_name: &str) -> Result<usize, ProcessReaderError> { + let dyld_info = self.task_info()?; + if (dyld_info.all_image_info_format as u32) != TASK_DYLD_ALL_IMAGE_INFO_64 { + return Err(ProcessReaderError::ImageFormatError); + } + + let all_image_info_size = dyld_info.all_image_info_size; + let all_image_info_addr = dyld_info.all_image_info_addr; + if (all_image_info_size as usize) < size_of::<AllImagesInfo>() { + return Err(ProcessReaderError::ImageFormatError); + } + + let all_images_info = self.copy_object::<AllImagesInfo>(all_image_info_addr as _)?; + + // Load the images + let images = self.copy_array::<ImageInfo>( + all_images_info.info_array_addr as _, + all_images_info.info_array_count as _, + )?; + + images + .iter() + .find(|&image| { + let image_path = self.copy_null_terminated_string(image.file_path as usize); + + if let Ok(image_path) = image_path { + if let Some(image_name) = image_path.into_bytes().rsplit(|&b| b == b'/').next() + { + image_name.eq(module_name.as_bytes()) + } else { + false + } + } else { + false + } + }) + .map(|image| image.load_address as usize) + .ok_or(ProcessReaderError::ModuleNotFound) + } + + pub fn find_section( + &self, + module_address: usize, + section_name: &[u8; 16], + ) -> Result<usize, ProcessReaderError> { + let header = self.copy_object::<Header64>(module_address)?; + let mut address = module_address + size_of::<Header64>(); + + if header.magic == MH_MAGIC_64 + && (header.filetype == MH_EXECUTE || header.filetype == MH_DYLIB) + { + let end_of_commands = address + (header.sizeofcmds as usize); + + while address < end_of_commands { + let command = self.copy_object::<LoadCommandHeader>(address)?; + + if command.cmd == LC_SEGMENT_64 { + if let Ok(offset) = self.find_section_in_segment(address, section_name) { + return module_address + .checked_add(offset) + .ok_or(ProcessReaderError::InvalidAddress); + } + } + + address += command.cmdsize as usize; + } + } + + Err(ProcessReaderError::SectionNotFound) + } + + fn find_section_in_segment( + &self, + segment_address: usize, + section_name: &[u8; 16], + ) -> Result<usize, ProcessReaderError> { + let segment = self.copy_object::<SegmentCommand64>(segment_address)?; + + if segment.segname.eq(DATA_SEGMENT) { + let sections_addr = segment_address + size_of::<SegmentCommand64>(); + let sections = self.copy_array::<Section64>(sections_addr, segment.nsects as usize)?; + for section in §ions { + if section.sectname.eq(section_name) { + return Ok(section.offset as usize); + } + } + } + + Err(ProcessReaderError::SectionNotFound) + } + + fn task_info(&self) -> Result<task_dyld_info, ProcessReaderError> { + let mut info = std::mem::MaybeUninit::<task_dyld_info>::uninit(); + let mut count = (std::mem::size_of::<task_dyld_info>() / std::mem::size_of::<u32>()) as u32; + + let res = unsafe { + task_info( + self.process, + TASK_DYLD_INFO, + info.as_mut_ptr().cast(), + &mut count, + ) + }; + + if res == KERN_SUCCESS { + // SAFETY: this will be initialized if the call succeeded + unsafe { Ok(info.assume_init()) } + } else { + Err(ProcessReaderError::TaskInfoError) + } + } + + pub fn copy_object_shallow<T>(&self, src: usize) -> Result<MaybeUninit<T>, ReadError> { + let mut object = MaybeUninit::<T>::uninit(); + let mut size: u64 = 0; + let res = unsafe { + mach_vm_read_overwrite( + self.process, + src as u64, + size_of::<T>() as u64, + object.as_mut_ptr() as _, + &mut size as _, + ) + }; + + if res == KERN_SUCCESS { + Ok(object) + } else { + Err(ReadError::MachError) + } + } + + pub fn copy_object<T>(&self, src: usize) -> Result<T, ReadError> { + let object = self.copy_object_shallow(src)?; + Ok(unsafe { object.assume_init() }) + } + + pub fn copy_array<T>(&self, src: usize, num: usize) -> Result<Vec<T>, ReadError> { + let mut array: Vec<MaybeUninit<T>> = Vec::with_capacity(num); + let mut size: u64 = 0; + let res = unsafe { + mach_vm_read_overwrite( + self.process, + src as u64, + (num * size_of::<T>()) as u64, + array.as_mut_ptr() as _, + &mut size as _, + ) + }; + + if res == KERN_SUCCESS { + unsafe { + array.set_len(num); + Ok(std::mem::transmute(array)) + } + } else { + Err(ReadError::MachError) + } + } +} + +impl Drop for ProcessReader { + fn drop(&mut self) {} +} diff --git a/toolkit/crashreporter/process_reader/src/platform/windows.rs b/toolkit/crashreporter/process_reader/src/platform/windows.rs new file mode 100644 index 0000000000..2d8c63444e --- /dev/null +++ b/toolkit/crashreporter/process_reader/src/platform/windows.rs @@ -0,0 +1,201 @@ +/* This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at https://mozilla.org/MPL/2.0/. */ + +use std::{ + convert::TryInto, + ffi::OsString, + mem::{size_of, MaybeUninit}, + os::windows::ffi::OsStringExt, + ptr::null_mut, +}; + +use windows_sys::Win32::{ + Foundation::{FALSE, HMODULE, MAX_PATH}, + System::{ + Diagnostics::Debug::ReadProcessMemory, + ProcessStatus::{ + K32EnumProcessModules, K32GetModuleBaseNameW, K32GetModuleInformation, MODULEINFO, + }, + }, +}; + +use crate::{ + error::{ProcessReaderError, ReadError}, + ProcessHandle, ProcessReader, +}; + +impl ProcessReader { + pub fn new(process: ProcessHandle) -> Result<ProcessReader, ProcessReaderError> { + Ok(ProcessReader { process }) + } + + pub fn find_module(&self, module_name: &str) -> Result<usize, ProcessReaderError> { + let modules = self.get_module_list()?; + + let module = modules.iter().find_map(|&module| { + let name = self.get_module_name(module); + // Crude way of mimicking Windows lower-case comparisons but + // sufficient for our use-cases. + if name.is_some_and(|name| name.eq_ignore_ascii_case(module_name)) { + self.get_module_info(module) + .map(|module| module.lpBaseOfDll as usize) + } else { + None + } + }); + + module.ok_or(ProcessReaderError::ModuleNotFound) + } + + pub fn find_section( + &self, + module_address: usize, + section_name: &[u8; 8], + ) -> Result<usize, ProcessReaderError> { + // We read only the first page from the module, this should be more than + // enough to read the header and section list. In the future we might do + // this incrementally but for now goblin requires an array to parse + // so we can't do it just yet. + const PAGE_SIZE: usize = 4096; + let bytes = self.copy_array(module_address as _, PAGE_SIZE)?; + let header = goblin::pe::header::Header::parse(&bytes)?; + + // Skip the PE header so we can parse the sections + let optional_header_offset = header.dos_header.pe_pointer as usize + + goblin::pe::header::SIZEOF_PE_MAGIC + + goblin::pe::header::SIZEOF_COFF_HEADER; + let offset = + &mut (optional_header_offset + header.coff_header.size_of_optional_header as usize); + + let sections = header.coff_header.sections(&bytes, offset)?; + + for section in sections { + if section.name.eq(section_name) { + let address = module_address.checked_add(section.virtual_address as usize); + return address.ok_or(ProcessReaderError::InvalidAddress); + } + } + + Err(ProcessReaderError::SectionNotFound) + } + + fn get_module_list(&self) -> Result<Vec<HMODULE>, ProcessReaderError> { + let mut module_num: usize = 100; + let mut required_buffer_size: u32 = 0; + let mut module_array = Vec::<HMODULE>::with_capacity(module_num); + + loop { + let buffer_size: u32 = (module_num * size_of::<HMODULE>()).try_into()?; + let res = unsafe { + K32EnumProcessModules( + self.process, + module_array.as_mut_ptr() as *mut _, + buffer_size, + &mut required_buffer_size as *mut _, + ) + }; + + module_num = required_buffer_size as usize / size_of::<HMODULE>(); + + if res == 0 { + if required_buffer_size > buffer_size { + module_array = Vec::<HMODULE>::with_capacity(module_num); + } else { + return Err(ProcessReaderError::EnumProcessModulesError); + } + } else { + break; + } + } + + // SAFETY: module_array has been filled by K32EnumProcessModules() + unsafe { + module_array.set_len(module_num); + }; + + Ok(module_array) + } + + fn get_module_name(&self, module: HMODULE) -> Option<String> { + let mut path: [u16; MAX_PATH as usize] = [0; MAX_PATH as usize]; + let res = unsafe { + K32GetModuleBaseNameW(self.process, module, (&mut path).as_mut_ptr(), MAX_PATH) + }; + + if res == 0 { + None + } else { + let name = OsString::from_wide(&path[0..res as usize]); + let name = name.to_str()?; + Some(name.to_string()) + } + } + + fn get_module_info(&self, module: HMODULE) -> Option<MODULEINFO> { + let mut info: MaybeUninit<MODULEINFO> = MaybeUninit::uninit(); + let res = unsafe { + K32GetModuleInformation( + self.process, + module, + info.as_mut_ptr(), + size_of::<MODULEINFO>() as u32, + ) + }; + + if res == 0 { + None + } else { + let info = unsafe { info.assume_init() }; + Some(info) + } + } + + pub fn copy_object_shallow<T>(&self, src: usize) -> Result<MaybeUninit<T>, ReadError> { + let mut object = MaybeUninit::<T>::uninit(); + let res = unsafe { + ReadProcessMemory( + self.process, + src as _, + object.as_mut_ptr() as _, + size_of::<T>(), + null_mut(), + ) + }; + + if res != FALSE { + Ok(object) + } else { + Err(ReadError::ReadProcessMemoryError) + } + } + + pub fn copy_object<T>(&self, src: usize) -> Result<T, ReadError> { + let object = self.copy_object_shallow(src)?; + Ok(unsafe { object.assume_init() }) + } + + pub fn copy_array<T>(&self, src: usize, num: usize) -> Result<Vec<T>, ReadError> { + let num_of_bytes = num * size_of::<T>(); + let mut array: Vec<T> = Vec::with_capacity(num); + let res = unsafe { + ReadProcessMemory( + self.process, + src as _, + array.as_mut_ptr() as _, + num_of_bytes, + null_mut(), + ) + }; + + if res != FALSE { + unsafe { + array.set_len(num); + } + + Ok(array) + } else { + Err(ReadError::ReadProcessMemoryError) + } + } +} |