diff options
Diffstat (limited to 'library/std/src/personality')
| -rw-r--r-- | library/std/src/personality/dwarf/eh.rs | 192 | ||||
| -rw-r--r-- | library/std/src/personality/dwarf/mod.rs | 73 | ||||
| -rw-r--r-- | library/std/src/personality/dwarf/tests.rs | 19 | ||||
| -rw-r--r-- | library/std/src/personality/emcc.rs | 20 | ||||
| -rw-r--r-- | library/std/src/personality/gcc.rs | 279 |
5 files changed, 583 insertions, 0 deletions
diff --git a/library/std/src/personality/dwarf/eh.rs b/library/std/src/personality/dwarf/eh.rs new file mode 100644 index 000000000..8799137b7 --- /dev/null +++ b/library/std/src/personality/dwarf/eh.rs @@ -0,0 +1,192 @@ +//! Parsing of GCC-style Language-Specific Data Area (LSDA) +//! For details see: +//! * <https://refspecs.linuxfoundation.org/LSB_3.0.0/LSB-PDA/LSB-PDA/ehframechpt.html> +//! * <https://itanium-cxx-abi.github.io/cxx-abi/exceptions.pdf> +//! * <https://www.airs.com/blog/archives/460> +//! * <https://www.airs.com/blog/archives/464> +//! +//! A reference implementation may be found in the GCC source tree +//! (`<root>/libgcc/unwind-c.c` as of this writing). + +#![allow(non_upper_case_globals)] +#![allow(unused)] + +use super::DwarfReader; +use core::mem; + +pub const DW_EH_PE_omit: u8 = 0xFF; +pub const DW_EH_PE_absptr: u8 = 0x00; + +pub const DW_EH_PE_uleb128: u8 = 0x01; +pub const DW_EH_PE_udata2: u8 = 0x02; +pub const DW_EH_PE_udata4: u8 = 0x03; +pub const DW_EH_PE_udata8: u8 = 0x04; +pub const DW_EH_PE_sleb128: u8 = 0x09; +pub const DW_EH_PE_sdata2: u8 = 0x0A; +pub const DW_EH_PE_sdata4: u8 = 0x0B; +pub const DW_EH_PE_sdata8: u8 = 0x0C; + +pub const DW_EH_PE_pcrel: u8 = 0x10; +pub const DW_EH_PE_textrel: u8 = 0x20; +pub const DW_EH_PE_datarel: u8 = 0x30; +pub const DW_EH_PE_funcrel: u8 = 0x40; +pub const DW_EH_PE_aligned: u8 = 0x50; + +pub const DW_EH_PE_indirect: u8 = 0x80; + +#[derive(Copy, Clone)] +pub struct EHContext<'a> { + pub ip: usize, // Current instruction pointer + pub func_start: usize, // Address of the current function + pub get_text_start: &'a dyn Fn() -> usize, // Get address of the code section + pub get_data_start: &'a dyn Fn() -> usize, // Get address of the data section +} + +pub enum EHAction { + None, + Cleanup(usize), + Catch(usize), + Terminate, +} + +pub const USING_SJLJ_EXCEPTIONS: bool = cfg!(all(target_os = "ios", target_arch = "arm")); + +pub unsafe fn find_eh_action(lsda: *const u8, context: &EHContext<'_>) -> Result<EHAction, ()> { + if lsda.is_null() { + return Ok(EHAction::None); + } + + let func_start = context.func_start; + let mut reader = DwarfReader::new(lsda); + + let start_encoding = reader.read::<u8>(); + // base address for landing pad offsets + let lpad_base = if start_encoding != DW_EH_PE_omit { + read_encoded_pointer(&mut reader, context, start_encoding)? + } else { + func_start + }; + + let ttype_encoding = reader.read::<u8>(); + if ttype_encoding != DW_EH_PE_omit { + // Rust doesn't analyze exception types, so we don't care about the type table + reader.read_uleb128(); + } + + let call_site_encoding = reader.read::<u8>(); + let call_site_table_length = reader.read_uleb128(); + let action_table = reader.ptr.add(call_site_table_length as usize); + let ip = context.ip; + + if !USING_SJLJ_EXCEPTIONS { + while reader.ptr < action_table { + let cs_start = read_encoded_pointer(&mut reader, context, call_site_encoding)?; + let cs_len = read_encoded_pointer(&mut reader, context, call_site_encoding)?; + let cs_lpad = read_encoded_pointer(&mut reader, context, call_site_encoding)?; + let cs_action = reader.read_uleb128(); + // Callsite table is sorted by cs_start, so if we've passed the ip, we + // may stop searching. + if ip < func_start + cs_start { + break; + } + if ip < func_start + cs_start + cs_len { + if cs_lpad == 0 { + return Ok(EHAction::None); + } else { + let lpad = lpad_base + cs_lpad; + return Ok(interpret_cs_action(cs_action, lpad)); + } + } + } + // Ip is not present in the table. This should not happen... but it does: issue #35011. + // So rather than returning EHAction::Terminate, we do this. + Ok(EHAction::None) + } else { + // SjLj version: + // The "IP" is an index into the call-site table, with two exceptions: + // -1 means 'no-action', and 0 means 'terminate'. + match ip as isize { + -1 => return Ok(EHAction::None), + 0 => return Ok(EHAction::Terminate), + _ => (), + } + let mut idx = ip; + loop { + let cs_lpad = reader.read_uleb128(); + let cs_action = reader.read_uleb128(); + idx -= 1; + if idx == 0 { + // Can never have null landing pad for sjlj -- that would have + // been indicated by a -1 call site index. + let lpad = (cs_lpad + 1) as usize; + return Ok(interpret_cs_action(cs_action, lpad)); + } + } + } +} + +fn interpret_cs_action(cs_action: u64, lpad: usize) -> EHAction { + if cs_action == 0 { + // If cs_action is 0 then this is a cleanup (Drop::drop). We run these + // for both Rust panics and foreign exceptions. + EHAction::Cleanup(lpad) + } else { + // Stop unwinding Rust panics at catch_unwind. + EHAction::Catch(lpad) + } +} + +#[inline] +fn round_up(unrounded: usize, align: usize) -> Result<usize, ()> { + if align.is_power_of_two() { Ok((unrounded + align - 1) & !(align - 1)) } else { Err(()) } +} + +unsafe fn read_encoded_pointer( + reader: &mut DwarfReader, + context: &EHContext<'_>, + encoding: u8, +) -> Result<usize, ()> { + if encoding == DW_EH_PE_omit { + return Err(()); + } + + // DW_EH_PE_aligned implies it's an absolute pointer value + if encoding == DW_EH_PE_aligned { + reader.ptr = round_up(reader.ptr as usize, mem::size_of::<usize>())? as *const u8; + return Ok(reader.read::<usize>()); + } + + let mut result = match encoding & 0x0F { + DW_EH_PE_absptr => reader.read::<usize>(), + DW_EH_PE_uleb128 => reader.read_uleb128() as usize, + DW_EH_PE_udata2 => reader.read::<u16>() as usize, + DW_EH_PE_udata4 => reader.read::<u32>() as usize, + DW_EH_PE_udata8 => reader.read::<u64>() as usize, + DW_EH_PE_sleb128 => reader.read_sleb128() as usize, + DW_EH_PE_sdata2 => reader.read::<i16>() as usize, + DW_EH_PE_sdata4 => reader.read::<i32>() as usize, + DW_EH_PE_sdata8 => reader.read::<i64>() as usize, + _ => return Err(()), + }; + + result += match encoding & 0x70 { + DW_EH_PE_absptr => 0, + // relative to address of the encoded value, despite the name + DW_EH_PE_pcrel => reader.ptr as usize, + DW_EH_PE_funcrel => { + if context.func_start == 0 { + return Err(()); + } + context.func_start + } + DW_EH_PE_textrel => (*context.get_text_start)(), + DW_EH_PE_datarel => (*context.get_data_start)(), + _ => return Err(()), + }; + + if encoding & DW_EH_PE_indirect != 0 { + result = *(result as *const usize); + } + + Ok(result) +} diff --git a/library/std/src/personality/dwarf/mod.rs b/library/std/src/personality/dwarf/mod.rs new file mode 100644 index 000000000..652fbe95a --- /dev/null +++ b/library/std/src/personality/dwarf/mod.rs @@ -0,0 +1,73 @@ +//! Utilities for parsing DWARF-encoded data streams. +//! See <http://www.dwarfstd.org>, +//! DWARF-4 standard, Section 7 - "Data Representation" + +// This module is used only by x86_64-pc-windows-gnu for now, but we +// are compiling it everywhere to avoid regressions. +#![allow(unused)] + +#[cfg(test)] +mod tests; + +pub mod eh; + +use core::mem; + +pub struct DwarfReader { + pub ptr: *const u8, +} + +#[repr(C, packed)] +struct Unaligned<T>(T); + +impl DwarfReader { + pub fn new(ptr: *const u8) -> DwarfReader { + DwarfReader { ptr } + } + + // DWARF streams are packed, so e.g., a u32 would not necessarily be aligned + // on a 4-byte boundary. This may cause problems on platforms with strict + // alignment requirements. By wrapping data in a "packed" struct, we are + // telling the backend to generate "misalignment-safe" code. + pub unsafe fn read<T: Copy>(&mut self) -> T { + let Unaligned(result) = *(self.ptr as *const Unaligned<T>); + self.ptr = self.ptr.add(mem::size_of::<T>()); + result + } + + // ULEB128 and SLEB128 encodings are defined in Section 7.6 - "Variable + // Length Data". + pub unsafe fn read_uleb128(&mut self) -> u64 { + let mut shift: usize = 0; + let mut result: u64 = 0; + let mut byte: u8; + loop { + byte = self.read::<u8>(); + result |= ((byte & 0x7F) as u64) << shift; + shift += 7; + if byte & 0x80 == 0 { + break; + } + } + result + } + + pub unsafe fn read_sleb128(&mut self) -> i64 { + let mut shift: u32 = 0; + let mut result: u64 = 0; + let mut byte: u8; + loop { + byte = self.read::<u8>(); + result |= ((byte & 0x7F) as u64) << shift; + shift += 7; + if byte & 0x80 == 0 { + break; + } + } + // sign-extend + if shift < u64::BITS && (byte & 0x40) != 0 { + result |= (!0 as u64) << shift; + } + result as i64 + } +} diff --git a/library/std/src/personality/dwarf/tests.rs b/library/std/src/personality/dwarf/tests.rs new file mode 100644 index 000000000..1644f3708 --- /dev/null +++ b/library/std/src/personality/dwarf/tests.rs @@ -0,0 +1,19 @@ +use super::*; + +#[test] +fn dwarf_reader() { + let encoded: &[u8] = &[1, 2, 3, 4, 5, 6, 7, 0xE5, 0x8E, 0x26, 0x9B, 0xF1, 0x59, 0xFF, 0xFF]; + + let mut reader = DwarfReader::new(encoded.as_ptr()); + + unsafe { + assert!(reader.read::<u8>() == u8::to_be(1u8)); + assert!(reader.read::<u16>() == u16::to_be(0x0203)); + assert!(reader.read::<u32>() == u32::to_be(0x04050607)); + + assert!(reader.read_uleb128() == 624485); + assert!(reader.read_sleb128() == -624485); + + assert!(reader.read::<i8>() == i8::to_be(-1)); + } +} diff --git a/library/std/src/personality/emcc.rs b/library/std/src/personality/emcc.rs new file mode 100644 index 000000000..f942bdf18 --- /dev/null +++ b/library/std/src/personality/emcc.rs @@ -0,0 +1,20 @@ +//! On Emscripten Rust panics are wrapped in C++ exceptions, so we just forward +//! to `__gxx_personality_v0` which is provided by Emscripten. + +use libc::c_int; +use unwind as uw; + +// This is required by the compiler to exist (e.g., it's a lang item), but it's +// never actually called by the compiler. Emscripten EH doesn't use a +// personality function at all, it instead uses __cxa_find_matching_catch. +// Wasm error handling would use __gxx_personality_wasm0. +#[lang = "eh_personality"] +unsafe extern "C" fn rust_eh_personality( + _version: c_int, + _actions: uw::_Unwind_Action, + _exception_class: uw::_Unwind_Exception_Class, + _exception_object: *mut uw::_Unwind_Exception, + _context: *mut uw::_Unwind_Context, +) -> uw::_Unwind_Reason_Code { + core::intrinsics::abort() +} diff --git a/library/std/src/personality/gcc.rs b/library/std/src/personality/gcc.rs new file mode 100644 index 000000000..7f0b0439c --- /dev/null +++ b/library/std/src/personality/gcc.rs @@ -0,0 +1,279 @@ +//! Implementation of panics backed by libgcc/libunwind (in some form). +//! +//! For background on exception handling and stack unwinding please see +//! "Exception Handling in LLVM" (llvm.org/docs/ExceptionHandling.html) and +//! documents linked from it. +//! These are also good reads: +//! * <https://itanium-cxx-abi.github.io/cxx-abi/abi-eh.html> +//! * <https://monoinfinito.wordpress.com/series/exception-handling-in-c/> +//! * <https://www.airs.com/blog/index.php?s=exception+frames> +//! +//! ## A brief summary +//! +//! Exception handling happens in two phases: a search phase and a cleanup +//! phase. +//! +//! In both phases the unwinder walks stack frames from top to bottom using +//! information from the stack frame unwind sections of the current process's +//! modules ("module" here refers to an OS module, i.e., an executable or a +//! dynamic library). +//! +//! For each stack frame, it invokes the associated "personality routine", whose +//! address is also stored in the unwind info section. +//! +//! In the search phase, the job of a personality routine is to examine +//! exception object being thrown, and to decide whether it should be caught at +//! that stack frame. Once the handler frame has been identified, cleanup phase +//! begins. +//! +//! In the cleanup phase, the unwinder invokes each personality routine again. +//! This time it decides which (if any) cleanup code needs to be run for +//! the current stack frame. If so, the control is transferred to a special +//! branch in the function body, the "landing pad", which invokes destructors, +//! frees memory, etc. At the end of the landing pad, control is transferred +//! back to the unwinder and unwinding resumes. +//! +//! Once stack has been unwound down to the handler frame level, unwinding stops +//! and the last personality routine transfers control to the catch block. + +use super::dwarf::eh::{self, EHAction, EHContext}; +use libc::{c_int, uintptr_t}; +use unwind as uw; + +// Register ids were lifted from LLVM's TargetLowering::getExceptionPointerRegister() +// and TargetLowering::getExceptionSelectorRegister() for each architecture, +// then mapped to DWARF register numbers via register definition tables +// (typically <arch>RegisterInfo.td, search for "DwarfRegNum"). +// See also https://llvm.org/docs/WritingAnLLVMBackend.html#defining-a-register. + +#[cfg(target_arch = "x86")] +const UNWIND_DATA_REG: (i32, i32) = (0, 2); // EAX, EDX + +#[cfg(target_arch = "x86_64")] +const UNWIND_DATA_REG: (i32, i32) = (0, 1); // RAX, RDX + +#[cfg(any(target_arch = "arm", target_arch = "aarch64"))] +const UNWIND_DATA_REG: (i32, i32) = (0, 1); // R0, R1 / X0, X1 + +#[cfg(target_arch = "m68k")] +const UNWIND_DATA_REG: (i32, i32) = (0, 1); // D0, D1 + +#[cfg(any(target_arch = "mips", target_arch = "mips64"))] +const UNWIND_DATA_REG: (i32, i32) = (4, 5); // A0, A1 + +#[cfg(any(target_arch = "powerpc", target_arch = "powerpc64"))] +const UNWIND_DATA_REG: (i32, i32) = (3, 4); // R3, R4 / X3, X4 + +#[cfg(target_arch = "s390x")] +const UNWIND_DATA_REG: (i32, i32) = (6, 7); // R6, R7 + +#[cfg(any(target_arch = "sparc", target_arch = "sparc64"))] +const UNWIND_DATA_REG: (i32, i32) = (24, 25); // I0, I1 + +#[cfg(target_arch = "hexagon")] +const UNWIND_DATA_REG: (i32, i32) = (0, 1); // R0, R1 + +#[cfg(any(target_arch = "riscv64", target_arch = "riscv32"))] +const UNWIND_DATA_REG: (i32, i32) = (10, 11); // x10, x11 + +// The following code is based on GCC's C and C++ personality routines. For reference, see: +// https://github.com/gcc-mirror/gcc/blob/master/libstdc++-v3/libsupc++/eh_personality.cc +// https://github.com/gcc-mirror/gcc/blob/trunk/libgcc/unwind-c.c + +cfg_if::cfg_if! { + if #[cfg(all(target_arch = "arm", not(target_os = "ios"), not(target_os = "watchos"), not(target_os = "netbsd")))] { + // ARM EHABI personality routine. + // https://infocenter.arm.com/help/topic/com.arm.doc.ihi0038b/IHI0038B_ehabi.pdf + // + // iOS uses the default routine instead since it uses SjLj unwinding. + #[lang = "eh_personality"] + unsafe extern "C" fn rust_eh_personality( + state: uw::_Unwind_State, + exception_object: *mut uw::_Unwind_Exception, + context: *mut uw::_Unwind_Context, + ) -> uw::_Unwind_Reason_Code { + let state = state as c_int; + let action = state & uw::_US_ACTION_MASK as c_int; + let search_phase = if action == uw::_US_VIRTUAL_UNWIND_FRAME as c_int { + // Backtraces on ARM will call the personality routine with + // state == _US_VIRTUAL_UNWIND_FRAME | _US_FORCE_UNWIND. In those cases + // we want to continue unwinding the stack, otherwise all our backtraces + // would end at __rust_try + if state & uw::_US_FORCE_UNWIND as c_int != 0 { + return continue_unwind(exception_object, context); + } + true + } else if action == uw::_US_UNWIND_FRAME_STARTING as c_int { + false + } else if action == uw::_US_UNWIND_FRAME_RESUME as c_int { + return continue_unwind(exception_object, context); + } else { + return uw::_URC_FAILURE; + }; + + // The DWARF unwinder assumes that _Unwind_Context holds things like the function + // and LSDA pointers, however ARM EHABI places them into the exception object. + // To preserve signatures of functions like _Unwind_GetLanguageSpecificData(), which + // take only the context pointer, GCC personality routines stash a pointer to + // exception_object in the context, using location reserved for ARM's + // "scratch register" (r12). + uw::_Unwind_SetGR(context, uw::UNWIND_POINTER_REG, exception_object as uw::_Unwind_Ptr); + // ...A more principled approach would be to provide the full definition of ARM's + // _Unwind_Context in our libunwind bindings and fetch the required data from there + // directly, bypassing DWARF compatibility functions. + + let eh_action = match find_eh_action(context) { + Ok(action) => action, + Err(_) => return uw::_URC_FAILURE, + }; + if search_phase { + match eh_action { + EHAction::None | EHAction::Cleanup(_) => { + return continue_unwind(exception_object, context); + } + EHAction::Catch(_) => { + // EHABI requires the personality routine to update the + // SP value in the barrier cache of the exception object. + (*exception_object).private[5] = + uw::_Unwind_GetGR(context, uw::UNWIND_SP_REG); + return uw::_URC_HANDLER_FOUND; + } + EHAction::Terminate => return uw::_URC_FAILURE, + } + } else { + match eh_action { + EHAction::None => return continue_unwind(exception_object, context), + EHAction::Cleanup(lpad) | EHAction::Catch(lpad) => { + uw::_Unwind_SetGR( + context, + UNWIND_DATA_REG.0, + exception_object as uintptr_t, + ); + uw::_Unwind_SetGR(context, UNWIND_DATA_REG.1, 0); + uw::_Unwind_SetIP(context, lpad); + return uw::_URC_INSTALL_CONTEXT; + } + EHAction::Terminate => return uw::_URC_FAILURE, + } + } + + // On ARM EHABI the personality routine is responsible for actually + // unwinding a single stack frame before returning (ARM EHABI Sec. 6.1). + unsafe fn continue_unwind( + exception_object: *mut uw::_Unwind_Exception, + context: *mut uw::_Unwind_Context, + ) -> uw::_Unwind_Reason_Code { + if __gnu_unwind_frame(exception_object, context) == uw::_URC_NO_REASON { + uw::_URC_CONTINUE_UNWIND + } else { + uw::_URC_FAILURE + } + } + // defined in libgcc + extern "C" { + fn __gnu_unwind_frame( + exception_object: *mut uw::_Unwind_Exception, + context: *mut uw::_Unwind_Context, + ) -> uw::_Unwind_Reason_Code; + } + } + } else { + // Default personality routine, which is used directly on most targets + // and indirectly on Windows x86_64 via SEH. + unsafe extern "C" fn rust_eh_personality_impl( + version: c_int, + actions: uw::_Unwind_Action, + _exception_class: uw::_Unwind_Exception_Class, + exception_object: *mut uw::_Unwind_Exception, + context: *mut uw::_Unwind_Context, + ) -> uw::_Unwind_Reason_Code { + if version != 1 { + return uw::_URC_FATAL_PHASE1_ERROR; + } + let eh_action = match find_eh_action(context) { + Ok(action) => action, + Err(_) => return uw::_URC_FATAL_PHASE1_ERROR, + }; + if actions as i32 & uw::_UA_SEARCH_PHASE as i32 != 0 { + match eh_action { + EHAction::None | EHAction::Cleanup(_) => uw::_URC_CONTINUE_UNWIND, + EHAction::Catch(_) => uw::_URC_HANDLER_FOUND, + EHAction::Terminate => uw::_URC_FATAL_PHASE1_ERROR, + } + } else { + match eh_action { + EHAction::None => uw::_URC_CONTINUE_UNWIND, + EHAction::Cleanup(lpad) | EHAction::Catch(lpad) => { + uw::_Unwind_SetGR( + context, + UNWIND_DATA_REG.0, + exception_object as uintptr_t, + ); + uw::_Unwind_SetGR(context, UNWIND_DATA_REG.1, 0); + uw::_Unwind_SetIP(context, lpad); + uw::_URC_INSTALL_CONTEXT + } + EHAction::Terminate => uw::_URC_FATAL_PHASE2_ERROR, + } + } + } + + cfg_if::cfg_if! { + if #[cfg(all(windows, target_arch = "x86_64", target_env = "gnu"))] { + // On x86_64 MinGW targets, the unwinding mechanism is SEH however the unwind + // handler data (aka LSDA) uses GCC-compatible encoding. + #[lang = "eh_personality"] + #[allow(nonstandard_style)] + unsafe extern "C" fn rust_eh_personality( + exceptionRecord: *mut uw::EXCEPTION_RECORD, + establisherFrame: uw::LPVOID, + contextRecord: *mut uw::CONTEXT, + dispatcherContext: *mut uw::DISPATCHER_CONTEXT, + ) -> uw::EXCEPTION_DISPOSITION { + uw::_GCC_specific_handler( + exceptionRecord, + establisherFrame, + contextRecord, + dispatcherContext, + rust_eh_personality_impl, + ) + } + } else { + // The personality routine for most of our targets. + #[lang = "eh_personality"] + unsafe extern "C" fn rust_eh_personality( + version: c_int, + actions: uw::_Unwind_Action, + exception_class: uw::_Unwind_Exception_Class, + exception_object: *mut uw::_Unwind_Exception, + context: *mut uw::_Unwind_Context, + ) -> uw::_Unwind_Reason_Code { + rust_eh_personality_impl( + version, + actions, + exception_class, + exception_object, + context, + ) + } + } + } + } +} + +unsafe fn find_eh_action(context: *mut uw::_Unwind_Context) -> Result<EHAction, ()> { + let lsda = uw::_Unwind_GetLanguageSpecificData(context) as *const u8; + let mut ip_before_instr: c_int = 0; + let ip = uw::_Unwind_GetIPInfo(context, &mut ip_before_instr); + let eh_context = EHContext { + // The return address points 1 byte past the call instruction, + // which could be in the next IP range in LSDA range table. + // + // `ip = -1` has special meaning, so use wrapping sub to allow for that + ip: if ip_before_instr != 0 { ip } else { ip.wrapping_sub(1) }, + func_start: uw::_Unwind_GetRegionStart(context), + get_text_start: &|| uw::_Unwind_GetTextRelBase(context), + get_data_start: &|| uw::_Unwind_GetDataRelBase(context), + }; + eh::find_eh_action(lsda, &eh_context) +} |