rust/src/libpanic_unwind/seh.rs

327 lines
12 KiB
Rust
Raw Normal View History

// Copyright 2015 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
trans: Reimplement unwinding on MSVC This commit transitions the compiler to using the new exception handling instructions in LLVM for implementing unwinding for MSVC. This affects both 32 and 64-bit MSVC as they're both now using SEH-based strategies. In terms of standard library support, lots more details about how SEH unwinding is implemented can be found in the commits. In terms of trans, this change necessitated a few modifications: * Branches were added to detect when the old landingpad instruction is used or the new cleanuppad instruction is used to `trans::cleanup`. * The return value from `cleanuppad` is not stored in an `alloca` (because it cannot be). * Each block in trans now has an `Option<LandingPad>` instead of `is_lpad: bool` for indicating whether it's in a landing pad or not. The new exception handling intrinsics require that on MSVC each `call` inside of a landing pad is annotated with which landing pad that it's in. This change to the basic block means that whenever a `call` or `invoke` instruction is generated we know whether to annotate it as part of a cleanuppad or not. * Lots of modifications were made to the instruction builders to construct the new instructions as well as pass the tagging information for the call/invoke instructions. * The translation of the `try` intrinsics for MSVC has been overhauled to use the new `catchpad` instruction. The filter function is now also a rustc-generated function instead of a purely libstd-defined function. The libstd definition still exists, it just has a stable ABI across architectures and leaves some of the really weird implementation details to the compiler (e.g. the `localescape` and `localrecover` intrinsics).
2015-10-23 20:18:44 -05:00
//! Windows SEH
msvc: Implement runtime support for unwinding Now that LLVM has been updated, the only remaining roadblock to implementing unwinding for MSVC is to fill out the runtime support in `std::rt::unwind::seh`. This commit does precisely that, fixing up some other bits and pieces along the way: * The `seh` unwinding module now uses `RaiseException` to initiate a panic. * The `rust_try.ll` file was rewritten for MSVC (as it's quite different) and is located at `rust_try_msvc_64.ll`, only included on MSVC builds for now. * The personality function for all landing pads generated by LLVM is hard-wired to `__C_specific_handler` instead of the standard `rust_eh_personality` lang item. This is required to get LLVM to emit SEH unwinding information instead of DWARF unwinding information. This also means that on MSVC the `rust_eh_personality` function is entirely unused (but is defined as it's a lang item). More details about how panicking works on SEH can be found in the `rust_try_msvc_64.ll` or `seh.rs` files, but I'm always open to adding more comments! A key aspect of this PR is missing, however, which is that **unwinding is still turned off by default for MSVC**. There is a [bug in llvm][llvm-bug] which causes optimizations to inline enough landing pads that LLVM chokes. If the compiler is optimized at `-O1` (where inlining isn't enabled) then it can bootstrap with unwinding enabled, but when optimized at `-O2` (inlining is enabled) then it hits a fatal LLVM error. [llvm-bug]: https://llvm.org/bugs/show_bug.cgi?id=23884
2015-06-19 16:57:06 -05:00
//!
//! On Windows (currently only on MSVC), the default exception handling
//! mechanism is Structured Exception Handling (SEH). This is quite different
//! than Dwarf-based exception handling (e.g. what other unix platforms use) in
//! terms of compiler internals, so LLVM is required to have a good deal of
trans: Reimplement unwinding on MSVC This commit transitions the compiler to using the new exception handling instructions in LLVM for implementing unwinding for MSVC. This affects both 32 and 64-bit MSVC as they're both now using SEH-based strategies. In terms of standard library support, lots more details about how SEH unwinding is implemented can be found in the commits. In terms of trans, this change necessitated a few modifications: * Branches were added to detect when the old landingpad instruction is used or the new cleanuppad instruction is used to `trans::cleanup`. * The return value from `cleanuppad` is not stored in an `alloca` (because it cannot be). * Each block in trans now has an `Option<LandingPad>` instead of `is_lpad: bool` for indicating whether it's in a landing pad or not. The new exception handling intrinsics require that on MSVC each `call` inside of a landing pad is annotated with which landing pad that it's in. This change to the basic block means that whenever a `call` or `invoke` instruction is generated we know whether to annotate it as part of a cleanuppad or not. * Lots of modifications were made to the instruction builders to construct the new instructions as well as pass the tagging information for the call/invoke instructions. * The translation of the `try` intrinsics for MSVC has been overhauled to use the new `catchpad` instruction. The filter function is now also a rustc-generated function instead of a purely libstd-defined function. The libstd definition still exists, it just has a stable ABI across architectures and leaves some of the really weird implementation details to the compiler (e.g. the `localescape` and `localrecover` intrinsics).
2015-10-23 20:18:44 -05:00
//! extra support for SEH.
msvc: Implement runtime support for unwinding Now that LLVM has been updated, the only remaining roadblock to implementing unwinding for MSVC is to fill out the runtime support in `std::rt::unwind::seh`. This commit does precisely that, fixing up some other bits and pieces along the way: * The `seh` unwinding module now uses `RaiseException` to initiate a panic. * The `rust_try.ll` file was rewritten for MSVC (as it's quite different) and is located at `rust_try_msvc_64.ll`, only included on MSVC builds for now. * The personality function for all landing pads generated by LLVM is hard-wired to `__C_specific_handler` instead of the standard `rust_eh_personality` lang item. This is required to get LLVM to emit SEH unwinding information instead of DWARF unwinding information. This also means that on MSVC the `rust_eh_personality` function is entirely unused (but is defined as it's a lang item). More details about how panicking works on SEH can be found in the `rust_try_msvc_64.ll` or `seh.rs` files, but I'm always open to adding more comments! A key aspect of this PR is missing, however, which is that **unwinding is still turned off by default for MSVC**. There is a [bug in llvm][llvm-bug] which causes optimizations to inline enough landing pads that LLVM chokes. If the compiler is optimized at `-O1` (where inlining isn't enabled) then it can bootstrap with unwinding enabled, but when optimized at `-O2` (inlining is enabled) then it hits a fatal LLVM error. [llvm-bug]: https://llvm.org/bugs/show_bug.cgi?id=23884
2015-06-19 16:57:06 -05:00
//!
//! In a nutshell, what happens here is:
//!
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
//! 1. The `panic` function calls the standard Windows function
//! `_CxxThrowException` to throw a C++-like exception, triggering the
//! unwinding process.
trans: Reimplement unwinding on MSVC This commit transitions the compiler to using the new exception handling instructions in LLVM for implementing unwinding for MSVC. This affects both 32 and 64-bit MSVC as they're both now using SEH-based strategies. In terms of standard library support, lots more details about how SEH unwinding is implemented can be found in the commits. In terms of trans, this change necessitated a few modifications: * Branches were added to detect when the old landingpad instruction is used or the new cleanuppad instruction is used to `trans::cleanup`. * The return value from `cleanuppad` is not stored in an `alloca` (because it cannot be). * Each block in trans now has an `Option<LandingPad>` instead of `is_lpad: bool` for indicating whether it's in a landing pad or not. The new exception handling intrinsics require that on MSVC each `call` inside of a landing pad is annotated with which landing pad that it's in. This change to the basic block means that whenever a `call` or `invoke` instruction is generated we know whether to annotate it as part of a cleanuppad or not. * Lots of modifications were made to the instruction builders to construct the new instructions as well as pass the tagging information for the call/invoke instructions. * The translation of the `try` intrinsics for MSVC has been overhauled to use the new `catchpad` instruction. The filter function is now also a rustc-generated function instead of a purely libstd-defined function. The libstd definition still exists, it just has a stable ABI across architectures and leaves some of the really weird implementation details to the compiler (e.g. the `localescape` and `localrecover` intrinsics).
2015-10-23 20:18:44 -05:00
//! 2. All landing pads generated by the compiler use the personality function
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
//! `__CxxFrameHandler3`, a function in the CRT, and the unwinding code in
//! Windows will use this personality function to execute all cleanup code on
//! the stack.
trans: Reimplement unwinding on MSVC This commit transitions the compiler to using the new exception handling instructions in LLVM for implementing unwinding for MSVC. This affects both 32 and 64-bit MSVC as they're both now using SEH-based strategies. In terms of standard library support, lots more details about how SEH unwinding is implemented can be found in the commits. In terms of trans, this change necessitated a few modifications: * Branches were added to detect when the old landingpad instruction is used or the new cleanuppad instruction is used to `trans::cleanup`. * The return value from `cleanuppad` is not stored in an `alloca` (because it cannot be). * Each block in trans now has an `Option<LandingPad>` instead of `is_lpad: bool` for indicating whether it's in a landing pad or not. The new exception handling intrinsics require that on MSVC each `call` inside of a landing pad is annotated with which landing pad that it's in. This change to the basic block means that whenever a `call` or `invoke` instruction is generated we know whether to annotate it as part of a cleanuppad or not. * Lots of modifications were made to the instruction builders to construct the new instructions as well as pass the tagging information for the call/invoke instructions. * The translation of the `try` intrinsics for MSVC has been overhauled to use the new `catchpad` instruction. The filter function is now also a rustc-generated function instead of a purely libstd-defined function. The libstd definition still exists, it just has a stable ABI across architectures and leaves some of the really weird implementation details to the compiler (e.g. the `localescape` and `localrecover` intrinsics).
2015-10-23 20:18:44 -05:00
//! 3. All compiler-generated calls to `invoke` have a landing pad set as a
//! `cleanuppad` LLVM instruction, which indicates the start of the cleanup
//! routine. The personality (in step 2, defined in the CRT) is responsible
//! for running the cleanup routines.
//! 4. Eventually the "catch" code in the `try` intrinsic (generated by the
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
//! compiler) is executed and indicates that control should come back to
trans: Reimplement unwinding on MSVC This commit transitions the compiler to using the new exception handling instructions in LLVM for implementing unwinding for MSVC. This affects both 32 and 64-bit MSVC as they're both now using SEH-based strategies. In terms of standard library support, lots more details about how SEH unwinding is implemented can be found in the commits. In terms of trans, this change necessitated a few modifications: * Branches were added to detect when the old landingpad instruction is used or the new cleanuppad instruction is used to `trans::cleanup`. * The return value from `cleanuppad` is not stored in an `alloca` (because it cannot be). * Each block in trans now has an `Option<LandingPad>` instead of `is_lpad: bool` for indicating whether it's in a landing pad or not. The new exception handling intrinsics require that on MSVC each `call` inside of a landing pad is annotated with which landing pad that it's in. This change to the basic block means that whenever a `call` or `invoke` instruction is generated we know whether to annotate it as part of a cleanuppad or not. * Lots of modifications were made to the instruction builders to construct the new instructions as well as pass the tagging information for the call/invoke instructions. * The translation of the `try` intrinsics for MSVC has been overhauled to use the new `catchpad` instruction. The filter function is now also a rustc-generated function instead of a purely libstd-defined function. The libstd definition still exists, it just has a stable ABI across architectures and leaves some of the really weird implementation details to the compiler (e.g. the `localescape` and `localrecover` intrinsics).
2015-10-23 20:18:44 -05:00
//! Rust. This is done via a `catchswitch` plus a `catchpad` instruction in
//! LLVM IR terms, finally returning normal control to the program with a
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
//! `catchret` instruction.
msvc: Implement runtime support for unwinding Now that LLVM has been updated, the only remaining roadblock to implementing unwinding for MSVC is to fill out the runtime support in `std::rt::unwind::seh`. This commit does precisely that, fixing up some other bits and pieces along the way: * The `seh` unwinding module now uses `RaiseException` to initiate a panic. * The `rust_try.ll` file was rewritten for MSVC (as it's quite different) and is located at `rust_try_msvc_64.ll`, only included on MSVC builds for now. * The personality function for all landing pads generated by LLVM is hard-wired to `__C_specific_handler` instead of the standard `rust_eh_personality` lang item. This is required to get LLVM to emit SEH unwinding information instead of DWARF unwinding information. This also means that on MSVC the `rust_eh_personality` function is entirely unused (but is defined as it's a lang item). More details about how panicking works on SEH can be found in the `rust_try_msvc_64.ll` or `seh.rs` files, but I'm always open to adding more comments! A key aspect of this PR is missing, however, which is that **unwinding is still turned off by default for MSVC**. There is a [bug in llvm][llvm-bug] which causes optimizations to inline enough landing pads that LLVM chokes. If the compiler is optimized at `-O1` (where inlining isn't enabled) then it can bootstrap with unwinding enabled, but when optimized at `-O2` (inlining is enabled) then it hits a fatal LLVM error. [llvm-bug]: https://llvm.org/bugs/show_bug.cgi?id=23884
2015-06-19 16:57:06 -05:00
//!
//! Some specific differences from the gcc-based exception handling are:
//!
//! * Rust has no custom personality function, it is instead *always*
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
//! `__CxxFrameHandler3`. Additionally, no extra filtering is performed, so we
//! end up catching any C++ exceptions that happen to look like the kind we're
//! throwing. Note that throwing an exception into Rust is undefined behavior
//! anyway, so this should be fine.
msvc: Implement runtime support for unwinding Now that LLVM has been updated, the only remaining roadblock to implementing unwinding for MSVC is to fill out the runtime support in `std::rt::unwind::seh`. This commit does precisely that, fixing up some other bits and pieces along the way: * The `seh` unwinding module now uses `RaiseException` to initiate a panic. * The `rust_try.ll` file was rewritten for MSVC (as it's quite different) and is located at `rust_try_msvc_64.ll`, only included on MSVC builds for now. * The personality function for all landing pads generated by LLVM is hard-wired to `__C_specific_handler` instead of the standard `rust_eh_personality` lang item. This is required to get LLVM to emit SEH unwinding information instead of DWARF unwinding information. This also means that on MSVC the `rust_eh_personality` function is entirely unused (but is defined as it's a lang item). More details about how panicking works on SEH can be found in the `rust_try_msvc_64.ll` or `seh.rs` files, but I'm always open to adding more comments! A key aspect of this PR is missing, however, which is that **unwinding is still turned off by default for MSVC**. There is a [bug in llvm][llvm-bug] which causes optimizations to inline enough landing pads that LLVM chokes. If the compiler is optimized at `-O1` (where inlining isn't enabled) then it can bootstrap with unwinding enabled, but when optimized at `-O2` (inlining is enabled) then it hits a fatal LLVM error. [llvm-bug]: https://llvm.org/bugs/show_bug.cgi?id=23884
2015-06-19 16:57:06 -05:00
//! * We've got some data to transmit across the unwinding boundary,
rustc: Implement custom panic runtimes This commit is an implementation of [RFC 1513] which allows applications to alter the behavior of panics at compile time. A new compiler flag, `-C panic`, is added and accepts the values `unwind` or `panic`, with the default being `unwind`. This model affects how code is generated for the local crate, skipping generation of landing pads with `-C panic=abort`. [RFC 1513]: https://github.com/rust-lang/rfcs/blob/master/text/1513-less-unwinding.md Panic implementations are then provided by crates tagged with `#![panic_runtime]` and lazily required by crates with `#![needs_panic_runtime]`. The panic strategy (`-C panic` value) of the panic runtime must match the final product, and if the panic strategy is not `abort` then the entire DAG must have the same panic strategy. With the `-C panic=abort` strategy, users can expect a stable method to disable generation of landing pads, improving optimization in niche scenarios, decreasing compile time, and decreasing output binary size. With the `-C panic=unwind` strategy users can expect the existing ability to isolate failure in Rust code from the outside world. Organizationally, this commit dismantles the `sys_common::unwind` module in favor of some bits moving part of it to `libpanic_unwind` and the rest into the `panicking` module in libstd. The custom panic runtime support is pretty similar to the custom allocator support with the only major difference being how the panic runtime is injected (takes the `-C panic` flag into account).
2016-04-08 18:18:40 -05:00
//! specifically a `Box<Any + Send>`. Like with Dwarf exceptions
trans: Reimplement unwinding on MSVC This commit transitions the compiler to using the new exception handling instructions in LLVM for implementing unwinding for MSVC. This affects both 32 and 64-bit MSVC as they're both now using SEH-based strategies. In terms of standard library support, lots more details about how SEH unwinding is implemented can be found in the commits. In terms of trans, this change necessitated a few modifications: * Branches were added to detect when the old landingpad instruction is used or the new cleanuppad instruction is used to `trans::cleanup`. * The return value from `cleanuppad` is not stored in an `alloca` (because it cannot be). * Each block in trans now has an `Option<LandingPad>` instead of `is_lpad: bool` for indicating whether it's in a landing pad or not. The new exception handling intrinsics require that on MSVC each `call` inside of a landing pad is annotated with which landing pad that it's in. This change to the basic block means that whenever a `call` or `invoke` instruction is generated we know whether to annotate it as part of a cleanuppad or not. * Lots of modifications were made to the instruction builders to construct the new instructions as well as pass the tagging information for the call/invoke instructions. * The translation of the `try` intrinsics for MSVC has been overhauled to use the new `catchpad` instruction. The filter function is now also a rustc-generated function instead of a purely libstd-defined function. The libstd definition still exists, it just has a stable ABI across architectures and leaves some of the really weird implementation details to the compiler (e.g. the `localescape` and `localrecover` intrinsics).
2015-10-23 20:18:44 -05:00
//! these two pointers are stored as a payload in the exception itself. On
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
//! MSVC, however, there's no need for an extra heap allocation because the
//! call stack is preserved while filter functions are being executed. This
//! means that the pointers are passed directly to `_CxxThrowException` which
//! are then recovered in the filter function to be written to the stack frame
//! of the `try` intrinsic.
msvc: Implement runtime support for unwinding Now that LLVM has been updated, the only remaining roadblock to implementing unwinding for MSVC is to fill out the runtime support in `std::rt::unwind::seh`. This commit does precisely that, fixing up some other bits and pieces along the way: * The `seh` unwinding module now uses `RaiseException` to initiate a panic. * The `rust_try.ll` file was rewritten for MSVC (as it's quite different) and is located at `rust_try_msvc_64.ll`, only included on MSVC builds for now. * The personality function for all landing pads generated by LLVM is hard-wired to `__C_specific_handler` instead of the standard `rust_eh_personality` lang item. This is required to get LLVM to emit SEH unwinding information instead of DWARF unwinding information. This also means that on MSVC the `rust_eh_personality` function is entirely unused (but is defined as it's a lang item). More details about how panicking works on SEH can be found in the `rust_try_msvc_64.ll` or `seh.rs` files, but I'm always open to adding more comments! A key aspect of this PR is missing, however, which is that **unwinding is still turned off by default for MSVC**. There is a [bug in llvm][llvm-bug] which causes optimizations to inline enough landing pads that LLVM chokes. If the compiler is optimized at `-O1` (where inlining isn't enabled) then it can bootstrap with unwinding enabled, but when optimized at `-O2` (inlining is enabled) then it hits a fatal LLVM error. [llvm-bug]: https://llvm.org/bugs/show_bug.cgi?id=23884
2015-06-19 16:57:06 -05:00
//!
trans: Reimplement unwinding on MSVC This commit transitions the compiler to using the new exception handling instructions in LLVM for implementing unwinding for MSVC. This affects both 32 and 64-bit MSVC as they're both now using SEH-based strategies. In terms of standard library support, lots more details about how SEH unwinding is implemented can be found in the commits. In terms of trans, this change necessitated a few modifications: * Branches were added to detect when the old landingpad instruction is used or the new cleanuppad instruction is used to `trans::cleanup`. * The return value from `cleanuppad` is not stored in an `alloca` (because it cannot be). * Each block in trans now has an `Option<LandingPad>` instead of `is_lpad: bool` for indicating whether it's in a landing pad or not. The new exception handling intrinsics require that on MSVC each `call` inside of a landing pad is annotated with which landing pad that it's in. This change to the basic block means that whenever a `call` or `invoke` instruction is generated we know whether to annotate it as part of a cleanuppad or not. * Lots of modifications were made to the instruction builders to construct the new instructions as well as pass the tagging information for the call/invoke instructions. * The translation of the `try` intrinsics for MSVC has been overhauled to use the new `catchpad` instruction. The filter function is now also a rustc-generated function instead of a purely libstd-defined function. The libstd definition still exists, it just has a stable ABI across architectures and leaves some of the really weird implementation details to the compiler (e.g. the `localescape` and `localrecover` intrinsics).
2015-10-23 20:18:44 -05:00
//! [win64]: http://msdn.microsoft.com/en-us/library/1eyas8tf.aspx
//! [llvm]: http://llvm.org/docs/ExceptionHandling.html#background-on-windows-exceptions
msvc: Implement runtime support for unwinding Now that LLVM has been updated, the only remaining roadblock to implementing unwinding for MSVC is to fill out the runtime support in `std::rt::unwind::seh`. This commit does precisely that, fixing up some other bits and pieces along the way: * The `seh` unwinding module now uses `RaiseException` to initiate a panic. * The `rust_try.ll` file was rewritten for MSVC (as it's quite different) and is located at `rust_try_msvc_64.ll`, only included on MSVC builds for now. * The personality function for all landing pads generated by LLVM is hard-wired to `__C_specific_handler` instead of the standard `rust_eh_personality` lang item. This is required to get LLVM to emit SEH unwinding information instead of DWARF unwinding information. This also means that on MSVC the `rust_eh_personality` function is entirely unused (but is defined as it's a lang item). More details about how panicking works on SEH can be found in the `rust_try_msvc_64.ll` or `seh.rs` files, but I'm always open to adding more comments! A key aspect of this PR is missing, however, which is that **unwinding is still turned off by default for MSVC**. There is a [bug in llvm][llvm-bug] which causes optimizations to inline enough landing pads that LLVM chokes. If the compiler is optimized at `-O1` (where inlining isn't enabled) then it can bootstrap with unwinding enabled, but when optimized at `-O2` (inlining is enabled) then it hits a fatal LLVM error. [llvm-bug]: https://llvm.org/bugs/show_bug.cgi?id=23884
2015-06-19 16:57:06 -05:00
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
#![allow(bad_style)]
#![allow(private_no_mangle_fns)]
rustc: Implement custom panic runtimes This commit is an implementation of [RFC 1513] which allows applications to alter the behavior of panics at compile time. A new compiler flag, `-C panic`, is added and accepts the values `unwind` or `panic`, with the default being `unwind`. This model affects how code is generated for the local crate, skipping generation of landing pads with `-C panic=abort`. [RFC 1513]: https://github.com/rust-lang/rfcs/blob/master/text/1513-less-unwinding.md Panic implementations are then provided by crates tagged with `#![panic_runtime]` and lazily required by crates with `#![needs_panic_runtime]`. The panic strategy (`-C panic` value) of the panic runtime must match the final product, and if the panic strategy is not `abort` then the entire DAG must have the same panic strategy. With the `-C panic=abort` strategy, users can expect a stable method to disable generation of landing pads, improving optimization in niche scenarios, decreasing compile time, and decreasing output binary size. With the `-C panic=unwind` strategy users can expect the existing ability to isolate failure in Rust code from the outside world. Organizationally, this commit dismantles the `sys_common::unwind` module in favor of some bits moving part of it to `libpanic_unwind` and the rest into the `panicking` module in libstd. The custom panic runtime support is pretty similar to the custom allocator support with the only major difference being how the panic runtime is injected (takes the `-C panic` flag into account).
2016-04-08 18:18:40 -05:00
use alloc::boxed::Box;
use core::any::Any;
use core::mem;
use core::raw;
use windows as c;
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
use libc::{c_int, c_uint};
// First up, a whole bunch of type definitions. There's a few platform-specific
// oddities here, and a lot that's just blatantly copied from LLVM. The purpose
// of all this is to implement the `panic` function below through a call to
// `_CxxThrowException`.
//
// This function takes two arguments. The first is a pointer to the data we're
// passing in, which in this case is our trait object. Pretty easy to find! The
// next, however, is more complicated. This is a pointer to a `_ThrowInfo`
// structure, and it generally is just intended to just describe the exception
// being thrown.
//
// Currently the definition of this type [1] is a little hairy, and the main
// oddity (and difference from the online article) is that on 32-bit the
// pointers are pointers but on 64-bit the pointers are expressed as 32-bit
// offsets from the `__ImageBase` symbol. The `ptr_t` and `ptr!` macro in the
// modules below are used to express this.
//
// The maze of type definitions also closely follows what LLVM emits for this
// sort of operation. For example, if you compile this C++ code on MSVC and emit
// the LLVM IR:
//
// #include <stdin.h>
//
// void foo() {
// uint64_t a[2] = {0, 1};
// throw a;
// }
//
// That's essentially what we're trying to emulate. Most of the constant values
// below were just copied from LLVM, I'm at least not 100% sure what's going on
// everywhere. For example the `.PA_K\0` and `.PEA_K\0` strings below (stuck in
// the names of a few of these) I'm not actually sure what they do, but it seems
// to mirror what LLVM does!
//
// In any case, these structures are all constructed in a similar manner, and
// it's just somewhat verbose for us.
//
// [1]: http://www.geoffchappell.com/studies/msvc/language/predefined/
#[cfg(target_arch = "x86")]
#[macro_use]
mod imp {
pub type ptr_t = *mut u8;
pub const OFFSET: i32 = 4;
pub const NAME1: [u8; 7] = [b'.', b'P', b'A', b'_', b'K', 0, 0];
pub const NAME2: [u8; 7] = [b'.', b'P', b'A', b'X', 0, 0, 0];
macro_rules! ptr {
(0) => (0 as *mut u8);
($e:expr) => ($e as *mut u8);
}
}
#[cfg(target_arch = "x86_64")]
#[macro_use]
mod imp {
pub type ptr_t = u32;
pub const OFFSET: i32 = 8;
pub const NAME1: [u8; 7] = [b'.', b'P', b'E', b'A', b'_', b'K', 0];
pub const NAME2: [u8; 7] = [b'.', b'P', b'E', b'A', b'X', 0, 0];
extern {
pub static __ImageBase: u8;
}
macro_rules! ptr {
(0) => (0);
($e:expr) => {
(($e as usize) - (&imp::__ImageBase as *const _ as usize)) as u32
}
}
}
#[repr(C)]
pub struct _ThrowInfo {
pub attribues: c_uint,
pub pnfnUnwind: imp::ptr_t,
pub pForwardCompat: imp::ptr_t,
pub pCatchableTypeArray: imp::ptr_t,
}
#[repr(C)]
pub struct _CatchableTypeArray {
pub nCatchableTypes: c_int,
pub arrayOfCatchableTypes: [imp::ptr_t; 2],
}
msvc: Implement runtime support for unwinding Now that LLVM has been updated, the only remaining roadblock to implementing unwinding for MSVC is to fill out the runtime support in `std::rt::unwind::seh`. This commit does precisely that, fixing up some other bits and pieces along the way: * The `seh` unwinding module now uses `RaiseException` to initiate a panic. * The `rust_try.ll` file was rewritten for MSVC (as it's quite different) and is located at `rust_try_msvc_64.ll`, only included on MSVC builds for now. * The personality function for all landing pads generated by LLVM is hard-wired to `__C_specific_handler` instead of the standard `rust_eh_personality` lang item. This is required to get LLVM to emit SEH unwinding information instead of DWARF unwinding information. This also means that on MSVC the `rust_eh_personality` function is entirely unused (but is defined as it's a lang item). More details about how panicking works on SEH can be found in the `rust_try_msvc_64.ll` or `seh.rs` files, but I'm always open to adding more comments! A key aspect of this PR is missing, however, which is that **unwinding is still turned off by default for MSVC**. There is a [bug in llvm][llvm-bug] which causes optimizations to inline enough landing pads that LLVM chokes. If the compiler is optimized at `-O1` (where inlining isn't enabled) then it can bootstrap with unwinding enabled, but when optimized at `-O2` (inlining is enabled) then it hits a fatal LLVM error. [llvm-bug]: https://llvm.org/bugs/show_bug.cgi?id=23884
2015-06-19 16:57:06 -05:00
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
#[repr(C)]
pub struct _CatchableType {
pub properties: c_uint,
pub pType: imp::ptr_t,
pub thisDisplacement: _PMD,
pub sizeOrOffset: c_int,
pub copy_function: imp::ptr_t,
}
#[repr(C)]
pub struct _PMD {
pub mdisp: c_int,
pub pdisp: c_int,
pub vdisp: c_int,
}
#[repr(C)]
pub struct _TypeDescriptor {
pub pVFTable: *const u8,
pub spare: *mut u8,
pub name: [u8; 7],
}
static mut THROW_INFO: _ThrowInfo = _ThrowInfo {
attribues: 0,
pnfnUnwind: ptr!(0),
pForwardCompat: ptr!(0),
pCatchableTypeArray: ptr!(0),
};
static mut CATCHABLE_TYPE_ARRAY: _CatchableTypeArray = _CatchableTypeArray {
nCatchableTypes: 2,
arrayOfCatchableTypes: [
ptr!(0),
ptr!(0),
],
};
static mut CATCHABLE_TYPE1: _CatchableType = _CatchableType {
properties: 1,
pType: ptr!(0),
thisDisplacement: _PMD {
mdisp: 0,
pdisp: -1,
vdisp: 0,
},
sizeOrOffset: imp::OFFSET,
copy_function: ptr!(0),
};
static mut CATCHABLE_TYPE2: _CatchableType = _CatchableType {
properties: 1,
pType: ptr!(0),
thisDisplacement: _PMD {
mdisp: 0,
pdisp: -1,
vdisp: 0,
},
sizeOrOffset: imp::OFFSET,
copy_function: ptr!(0),
};
extern {
// The leading `\x01` byte here is actually a magical signal to LLVM to
// *not* apply any other mangling like prefixing with a `_` character.
//
// This symbol is the vtable used by C++'s `std::type_info`. Objects of type
// `std::type_info`, type descriptors, have a pointer to this table. Type
// descriptors are referenced by the C++ EH structures defined above and
// that we construct below.
#[link_name = "\x01??_7type_info@@6B@"]
static TYPE_INFO_VTABLE: *const u8;
}
// We use #[lang = "msvc_try_filter"] here as this is the type descriptor which
// we'll use in LLVM's `catchpad` instruction which ends up also being passed as
// an argument to the C++ personality function.
//
// Again, I'm not entirely sure what this is describing, it just seems to work.
#[cfg_attr(all(not(test), not(stage0)),
lang = "msvc_try_filter")]
static mut TYPE_DESCRIPTOR1: _TypeDescriptor = _TypeDescriptor {
pVFTable: &TYPE_INFO_VTABLE as *const _ as *const _,
spare: 0 as *mut _,
name: imp::NAME1,
};
static mut TYPE_DESCRIPTOR2: _TypeDescriptor = _TypeDescriptor {
pVFTable: &TYPE_INFO_VTABLE as *const _ as *const _,
spare: 0 as *mut _,
name: imp::NAME2,
};
rustc: Implement custom panic runtimes This commit is an implementation of [RFC 1513] which allows applications to alter the behavior of panics at compile time. A new compiler flag, `-C panic`, is added and accepts the values `unwind` or `panic`, with the default being `unwind`. This model affects how code is generated for the local crate, skipping generation of landing pads with `-C panic=abort`. [RFC 1513]: https://github.com/rust-lang/rfcs/blob/master/text/1513-less-unwinding.md Panic implementations are then provided by crates tagged with `#![panic_runtime]` and lazily required by crates with `#![needs_panic_runtime]`. The panic strategy (`-C panic` value) of the panic runtime must match the final product, and if the panic strategy is not `abort` then the entire DAG must have the same panic strategy. With the `-C panic=abort` strategy, users can expect a stable method to disable generation of landing pads, improving optimization in niche scenarios, decreasing compile time, and decreasing output binary size. With the `-C panic=unwind` strategy users can expect the existing ability to isolate failure in Rust code from the outside world. Organizationally, this commit dismantles the `sys_common::unwind` module in favor of some bits moving part of it to `libpanic_unwind` and the rest into the `panicking` module in libstd. The custom panic runtime support is pretty similar to the custom allocator support with the only major difference being how the panic runtime is injected (takes the `-C panic` flag into account).
2016-04-08 18:18:40 -05:00
pub unsafe fn panic(data: Box<Any + Send>) -> u32 {
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
use core::intrinsics::atomic_store;
// _CxxThrowException executes entirely on this stack frame, so there's no
// need to otherwise transfer `data` to the heap. We just pass a stack
// pointer to this function.
rustc: Implement custom panic runtimes This commit is an implementation of [RFC 1513] which allows applications to alter the behavior of panics at compile time. A new compiler flag, `-C panic`, is added and accepts the values `unwind` or `panic`, with the default being `unwind`. This model affects how code is generated for the local crate, skipping generation of landing pads with `-C panic=abort`. [RFC 1513]: https://github.com/rust-lang/rfcs/blob/master/text/1513-less-unwinding.md Panic implementations are then provided by crates tagged with `#![panic_runtime]` and lazily required by crates with `#![needs_panic_runtime]`. The panic strategy (`-C panic` value) of the panic runtime must match the final product, and if the panic strategy is not `abort` then the entire DAG must have the same panic strategy. With the `-C panic=abort` strategy, users can expect a stable method to disable generation of landing pads, improving optimization in niche scenarios, decreasing compile time, and decreasing output binary size. With the `-C panic=unwind` strategy users can expect the existing ability to isolate failure in Rust code from the outside world. Organizationally, this commit dismantles the `sys_common::unwind` module in favor of some bits moving part of it to `libpanic_unwind` and the rest into the `panicking` module in libstd. The custom panic runtime support is pretty similar to the custom allocator support with the only major difference being how the panic runtime is injected (takes the `-C panic` flag into account).
2016-04-08 18:18:40 -05:00
//
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
// The first argument is the payload being thrown (our two pointers), and
// the second argument is the type information object describing the
// exception (constructed above).
rustc: Implement custom panic runtimes This commit is an implementation of [RFC 1513] which allows applications to alter the behavior of panics at compile time. A new compiler flag, `-C panic`, is added and accepts the values `unwind` or `panic`, with the default being `unwind`. This model affects how code is generated for the local crate, skipping generation of landing pads with `-C panic=abort`. [RFC 1513]: https://github.com/rust-lang/rfcs/blob/master/text/1513-less-unwinding.md Panic implementations are then provided by crates tagged with `#![panic_runtime]` and lazily required by crates with `#![needs_panic_runtime]`. The panic strategy (`-C panic` value) of the panic runtime must match the final product, and if the panic strategy is not `abort` then the entire DAG must have the same panic strategy. With the `-C panic=abort` strategy, users can expect a stable method to disable generation of landing pads, improving optimization in niche scenarios, decreasing compile time, and decreasing output binary size. With the `-C panic=unwind` strategy users can expect the existing ability to isolate failure in Rust code from the outside world. Organizationally, this commit dismantles the `sys_common::unwind` module in favor of some bits moving part of it to `libpanic_unwind` and the rest into the `panicking` module in libstd. The custom panic runtime support is pretty similar to the custom allocator support with the only major difference being how the panic runtime is injected (takes the `-C panic` flag into account).
2016-04-08 18:18:40 -05:00
let ptrs = mem::transmute::<_, raw::TraitObject>(data);
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
let mut ptrs = [ptrs.data as u64, ptrs.vtable as u64];
let mut ptrs_ptr = ptrs.as_mut_ptr();
// This... may seems surprising, and justifiably so. On 32-bit MSVC the
// pointers between these structure are just that, pointers. On 64-bit MSVC,
// however, the pointers between structures are rather expressed as 32-bit
// offsets from `__ImageBase`.
//
// Consequently, on 32-bit MSVC we can declare all these pointers in the
// `static`s above. On 64-bit MSVC, we would have to express subtraction of
// pointers in statics, which Rust does not currently allow, so we can't
// actually do that.
//
// The next best thing, then is to fill in these structures at runtime
// (panicking is already the "slow path" anyway). So here we reinterpret all
// of these pointer fields as 32-bit integers and then store the
// relevant value into it (atomically, as concurrent panics may be
// happening). Technically the runtime will probably do a nonatomic read of
// these fields, but in theory they never read the *wrong* value so it
// shouldn't be too bad...
//
// In any case, we basically need to do something like this until we can
// express more operations in statics (and we may never be able to).
atomic_store(&mut THROW_INFO.pCatchableTypeArray as *mut _ as *mut u32,
ptr!(&CATCHABLE_TYPE_ARRAY as *const _) as u32);
atomic_store(&mut CATCHABLE_TYPE_ARRAY.arrayOfCatchableTypes[0] as *mut _ as *mut u32,
ptr!(&CATCHABLE_TYPE1 as *const _) as u32);
atomic_store(&mut CATCHABLE_TYPE_ARRAY.arrayOfCatchableTypes[1] as *mut _ as *mut u32,
ptr!(&CATCHABLE_TYPE2 as *const _) as u32);
atomic_store(&mut CATCHABLE_TYPE1.pType as *mut _ as *mut u32,
ptr!(&TYPE_DESCRIPTOR1 as *const _) as u32);
atomic_store(&mut CATCHABLE_TYPE2.pType as *mut _ as *mut u32,
ptr!(&TYPE_DESCRIPTOR2 as *const _) as u32);
c::_CxxThrowException(&mut ptrs_ptr as *mut _ as *mut _,
&mut THROW_INFO as *mut _ as *mut _);
rustc: Implement custom panic runtimes This commit is an implementation of [RFC 1513] which allows applications to alter the behavior of panics at compile time. A new compiler flag, `-C panic`, is added and accepts the values `unwind` or `panic`, with the default being `unwind`. This model affects how code is generated for the local crate, skipping generation of landing pads with `-C panic=abort`. [RFC 1513]: https://github.com/rust-lang/rfcs/blob/master/text/1513-less-unwinding.md Panic implementations are then provided by crates tagged with `#![panic_runtime]` and lazily required by crates with `#![needs_panic_runtime]`. The panic strategy (`-C panic` value) of the panic runtime must match the final product, and if the panic strategy is not `abort` then the entire DAG must have the same panic strategy. With the `-C panic=abort` strategy, users can expect a stable method to disable generation of landing pads, improving optimization in niche scenarios, decreasing compile time, and decreasing output binary size. With the `-C panic=unwind` strategy users can expect the existing ability to isolate failure in Rust code from the outside world. Organizationally, this commit dismantles the `sys_common::unwind` module in favor of some bits moving part of it to `libpanic_unwind` and the rest into the `panicking` module in libstd. The custom panic runtime support is pretty similar to the custom allocator support with the only major difference being how the panic runtime is injected (takes the `-C panic` flag into account).
2016-04-08 18:18:40 -05:00
u32::max_value()
}
trans: Reimplement unwinding on MSVC This commit transitions the compiler to using the new exception handling instructions in LLVM for implementing unwinding for MSVC. This affects both 32 and 64-bit MSVC as they're both now using SEH-based strategies. In terms of standard library support, lots more details about how SEH unwinding is implemented can be found in the commits. In terms of trans, this change necessitated a few modifications: * Branches were added to detect when the old landingpad instruction is used or the new cleanuppad instruction is used to `trans::cleanup`. * The return value from `cleanuppad` is not stored in an `alloca` (because it cannot be). * Each block in trans now has an `Option<LandingPad>` instead of `is_lpad: bool` for indicating whether it's in a landing pad or not. The new exception handling intrinsics require that on MSVC each `call` inside of a landing pad is annotated with which landing pad that it's in. This change to the basic block means that whenever a `call` or `invoke` instruction is generated we know whether to annotate it as part of a cleanuppad or not. * Lots of modifications were made to the instruction builders to construct the new instructions as well as pass the tagging information for the call/invoke instructions. * The translation of the `try` intrinsics for MSVC has been overhauled to use the new `catchpad` instruction. The filter function is now also a rustc-generated function instead of a purely libstd-defined function. The libstd definition still exists, it just has a stable ABI across architectures and leaves some of the really weird implementation details to the compiler (e.g. the `localescape` and `localrecover` intrinsics).
2015-10-23 20:18:44 -05:00
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
pub fn payload() -> [u64; 2] {
rustc: Implement custom panic runtimes This commit is an implementation of [RFC 1513] which allows applications to alter the behavior of panics at compile time. A new compiler flag, `-C panic`, is added and accepts the values `unwind` or `panic`, with the default being `unwind`. This model affects how code is generated for the local crate, skipping generation of landing pads with `-C panic=abort`. [RFC 1513]: https://github.com/rust-lang/rfcs/blob/master/text/1513-less-unwinding.md Panic implementations are then provided by crates tagged with `#![panic_runtime]` and lazily required by crates with `#![needs_panic_runtime]`. The panic strategy (`-C panic` value) of the panic runtime must match the final product, and if the panic strategy is not `abort` then the entire DAG must have the same panic strategy. With the `-C panic=abort` strategy, users can expect a stable method to disable generation of landing pads, improving optimization in niche scenarios, decreasing compile time, and decreasing output binary size. With the `-C panic=unwind` strategy users can expect the existing ability to isolate failure in Rust code from the outside world. Organizationally, this commit dismantles the `sys_common::unwind` module in favor of some bits moving part of it to `libpanic_unwind` and the rest into the `panicking` module in libstd. The custom panic runtime support is pretty similar to the custom allocator support with the only major difference being how the panic runtime is injected (takes the `-C panic` flag into account).
2016-04-08 18:18:40 -05:00
[0; 2]
}
msvc: Implement runtime support for unwinding Now that LLVM has been updated, the only remaining roadblock to implementing unwinding for MSVC is to fill out the runtime support in `std::rt::unwind::seh`. This commit does precisely that, fixing up some other bits and pieces along the way: * The `seh` unwinding module now uses `RaiseException` to initiate a panic. * The `rust_try.ll` file was rewritten for MSVC (as it's quite different) and is located at `rust_try_msvc_64.ll`, only included on MSVC builds for now. * The personality function for all landing pads generated by LLVM is hard-wired to `__C_specific_handler` instead of the standard `rust_eh_personality` lang item. This is required to get LLVM to emit SEH unwinding information instead of DWARF unwinding information. This also means that on MSVC the `rust_eh_personality` function is entirely unused (but is defined as it's a lang item). More details about how panicking works on SEH can be found in the `rust_try_msvc_64.ll` or `seh.rs` files, but I'm always open to adding more comments! A key aspect of this PR is missing, however, which is that **unwinding is still turned off by default for MSVC**. There is a [bug in llvm][llvm-bug] which causes optimizations to inline enough landing pads that LLVM chokes. If the compiler is optimized at `-O1` (where inlining isn't enabled) then it can bootstrap with unwinding enabled, but when optimized at `-O2` (inlining is enabled) then it hits a fatal LLVM error. [llvm-bug]: https://llvm.org/bugs/show_bug.cgi?id=23884
2015-06-19 16:57:06 -05:00
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
pub unsafe fn cleanup(payload: [u64; 2]) -> Box<Any + Send> {
rustc: Implement custom panic runtimes This commit is an implementation of [RFC 1513] which allows applications to alter the behavior of panics at compile time. A new compiler flag, `-C panic`, is added and accepts the values `unwind` or `panic`, with the default being `unwind`. This model affects how code is generated for the local crate, skipping generation of landing pads with `-C panic=abort`. [RFC 1513]: https://github.com/rust-lang/rfcs/blob/master/text/1513-less-unwinding.md Panic implementations are then provided by crates tagged with `#![panic_runtime]` and lazily required by crates with `#![needs_panic_runtime]`. The panic strategy (`-C panic` value) of the panic runtime must match the final product, and if the panic strategy is not `abort` then the entire DAG must have the same panic strategy. With the `-C panic=abort` strategy, users can expect a stable method to disable generation of landing pads, improving optimization in niche scenarios, decreasing compile time, and decreasing output binary size. With the `-C panic=unwind` strategy users can expect the existing ability to isolate failure in Rust code from the outside world. Organizationally, this commit dismantles the `sys_common::unwind` module in favor of some bits moving part of it to `libpanic_unwind` and the rest into the `panicking` module in libstd. The custom panic runtime support is pretty similar to the custom allocator support with the only major difference being how the panic runtime is injected (takes the `-C panic` flag into account).
2016-04-08 18:18:40 -05:00
mem::transmute(raw::TraitObject {
data: payload[0] as *mut _,
vtable: payload[1] as *mut _,
})
}
trans: Reimplement unwinding on MSVC This commit transitions the compiler to using the new exception handling instructions in LLVM for implementing unwinding for MSVC. This affects both 32 and 64-bit MSVC as they're both now using SEH-based strategies. In terms of standard library support, lots more details about how SEH unwinding is implemented can be found in the commits. In terms of trans, this change necessitated a few modifications: * Branches were added to detect when the old landingpad instruction is used or the new cleanuppad instruction is used to `trans::cleanup`. * The return value from `cleanuppad` is not stored in an `alloca` (because it cannot be). * Each block in trans now has an `Option<LandingPad>` instead of `is_lpad: bool` for indicating whether it's in a landing pad or not. The new exception handling intrinsics require that on MSVC each `call` inside of a landing pad is annotated with which landing pad that it's in. This change to the basic block means that whenever a `call` or `invoke` instruction is generated we know whether to annotate it as part of a cleanuppad or not. * Lots of modifications were made to the instruction builders to construct the new instructions as well as pass the tagging information for the call/invoke instructions. * The translation of the `try` intrinsics for MSVC has been overhauled to use the new `catchpad` instruction. The filter function is now also a rustc-generated function instead of a purely libstd-defined function. The libstd definition still exists, it just has a stable ABI across architectures and leaves some of the really weird implementation details to the compiler (e.g. the `localescape` and `localrecover` intrinsics).
2015-10-23 20:18:44 -05:00
rustc: Implement custom panic runtimes This commit is an implementation of [RFC 1513] which allows applications to alter the behavior of panics at compile time. A new compiler flag, `-C panic`, is added and accepts the values `unwind` or `panic`, with the default being `unwind`. This model affects how code is generated for the local crate, skipping generation of landing pads with `-C panic=abort`. [RFC 1513]: https://github.com/rust-lang/rfcs/blob/master/text/1513-less-unwinding.md Panic implementations are then provided by crates tagged with `#![panic_runtime]` and lazily required by crates with `#![needs_panic_runtime]`. The panic strategy (`-C panic` value) of the panic runtime must match the final product, and if the panic strategy is not `abort` then the entire DAG must have the same panic strategy. With the `-C panic=abort` strategy, users can expect a stable method to disable generation of landing pads, improving optimization in niche scenarios, decreasing compile time, and decreasing output binary size. With the `-C panic=unwind` strategy users can expect the existing ability to isolate failure in Rust code from the outside world. Organizationally, this commit dismantles the `sys_common::unwind` module in favor of some bits moving part of it to `libpanic_unwind` and the rest into the `panicking` module in libstd. The custom panic runtime support is pretty similar to the custom allocator support with the only major difference being how the panic runtime is injected (takes the `-C panic` flag into account).
2016-04-08 18:18:40 -05:00
#[lang = "msvc_try_filter"]
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
#[cfg(stage0)]
unsafe extern fn __rust_try_filter(_eh_ptrs: *mut u8,
_payload: *mut u8) -> i32 {
return 0
}
trans: Reimplement unwinding on MSVC This commit transitions the compiler to using the new exception handling instructions in LLVM for implementing unwinding for MSVC. This affects both 32 and 64-bit MSVC as they're both now using SEH-based strategies. In terms of standard library support, lots more details about how SEH unwinding is implemented can be found in the commits. In terms of trans, this change necessitated a few modifications: * Branches were added to detect when the old landingpad instruction is used or the new cleanuppad instruction is used to `trans::cleanup`. * The return value from `cleanuppad` is not stored in an `alloca` (because it cannot be). * Each block in trans now has an `Option<LandingPad>` instead of `is_lpad: bool` for indicating whether it's in a landing pad or not. The new exception handling intrinsics require that on MSVC each `call` inside of a landing pad is annotated with which landing pad that it's in. This change to the basic block means that whenever a `call` or `invoke` instruction is generated we know whether to annotate it as part of a cleanuppad or not. * Lots of modifications were made to the instruction builders to construct the new instructions as well as pass the tagging information for the call/invoke instructions. * The translation of the `try` intrinsics for MSVC has been overhauled to use the new `catchpad` instruction. The filter function is now also a rustc-generated function instead of a purely libstd-defined function. The libstd definition still exists, it just has a stable ABI across architectures and leaves some of the really weird implementation details to the compiler (e.g. the `localescape` and `localrecover` intrinsics).
2015-10-23 20:18:44 -05:00
// This is required by the compiler to exist (e.g. it's a lang item), but
// it's never actually called by the compiler because __C_specific_handler
// or _except_handler3 is the personality function that is always used.
// Hence this is just an aborting stub.
#[lang = "eh_personality"]
trans: Reimplement unwinding on MSVC This commit transitions the compiler to using the new exception handling instructions in LLVM for implementing unwinding for MSVC. This affects both 32 and 64-bit MSVC as they're both now using SEH-based strategies. In terms of standard library support, lots more details about how SEH unwinding is implemented can be found in the commits. In terms of trans, this change necessitated a few modifications: * Branches were added to detect when the old landingpad instruction is used or the new cleanuppad instruction is used to `trans::cleanup`. * The return value from `cleanuppad` is not stored in an `alloca` (because it cannot be). * Each block in trans now has an `Option<LandingPad>` instead of `is_lpad: bool` for indicating whether it's in a landing pad or not. The new exception handling intrinsics require that on MSVC each `call` inside of a landing pad is annotated with which landing pad that it's in. This change to the basic block means that whenever a `call` or `invoke` instruction is generated we know whether to annotate it as part of a cleanuppad or not. * Lots of modifications were made to the instruction builders to construct the new instructions as well as pass the tagging information for the call/invoke instructions. * The translation of the `try` intrinsics for MSVC has been overhauled to use the new `catchpad` instruction. The filter function is now also a rustc-generated function instead of a purely libstd-defined function. The libstd definition still exists, it just has a stable ABI across architectures and leaves some of the really weird implementation details to the compiler (e.g. the `localescape` and `localrecover` intrinsics).
2015-10-23 20:18:44 -05:00
#[cfg(not(test))]
msvc: Implement runtime support for unwinding Now that LLVM has been updated, the only remaining roadblock to implementing unwinding for MSVC is to fill out the runtime support in `std::rt::unwind::seh`. This commit does precisely that, fixing up some other bits and pieces along the way: * The `seh` unwinding module now uses `RaiseException` to initiate a panic. * The `rust_try.ll` file was rewritten for MSVC (as it's quite different) and is located at `rust_try_msvc_64.ll`, only included on MSVC builds for now. * The personality function for all landing pads generated by LLVM is hard-wired to `__C_specific_handler` instead of the standard `rust_eh_personality` lang item. This is required to get LLVM to emit SEH unwinding information instead of DWARF unwinding information. This also means that on MSVC the `rust_eh_personality` function is entirely unused (but is defined as it's a lang item). More details about how panicking works on SEH can be found in the `rust_try_msvc_64.ll` or `seh.rs` files, but I'm always open to adding more comments! A key aspect of this PR is missing, however, which is that **unwinding is still turned off by default for MSVC**. There is a [bug in llvm][llvm-bug] which causes optimizations to inline enough landing pads that LLVM chokes. If the compiler is optimized at `-O1` (where inlining isn't enabled) then it can bootstrap with unwinding enabled, but when optimized at `-O2` (inlining is enabled) then it hits a fatal LLVM error. [llvm-bug]: https://llvm.org/bugs/show_bug.cgi?id=23884
2015-06-19 16:57:06 -05:00
fn rust_eh_personality() {
rustc: Use C++ personalities on MSVC Currently the compiler has two relatively critical bugs in the implementation of MSVC unwinding: * #33112 - faults like segfaults and illegal instructions will run destructors in Rust, meaning we keep running code after a super-fatal exception has happened. * #33116 - When compiling with LTO plus `-Z no-landing-pads` (or `-C panic=abort` with the previous commit) LLVM won't remove all `invoke` instructions, meaning that some landing pads stick around and cleanups may be run due to the previous bug. These both stem from the flavor of "personality function" that Rust uses for unwinding on MSVC. On 32-bit this is `_except_handler3` and on 64-bit this is `__C_specific_handler`, but they both essentially are the "most generic" personality functions for catching exceptions and running cleanups. That is, thse two personalities will run cleanups for all exceptions unconditionally, so when we use them we run cleanups for **all SEH exceptions** (include things like segfaults). Note that this also explains why LLVM won't optimize away `invoke` instructions. These functions can legitimately still unwind (the `nounwind` attribute only seems to apply to "C++ exception-like unwining"). Also note that the standard library only *catches* Rust exceptions, not others like segfaults and illegal instructions. LLVM has support for another personality, `__CxxFrameHandler3`, which does not run cleanups for general exceptions, only C++ exceptions thrown by `_CxxThrowException`. This essentially ideally matches our use case, so this commit moves us over to using this well-known personality function as well as exception-throwing function. This doesn't *seem* to pull in any extra runtime dependencies just yet, but if it does we can perhaps try to work out how to implement more of it in Rust rather than relying on MSVCRT runtime bits. More details about how this is actually implemented can be found in the changes itself, but this... Closes #33112 Closes #33116
2016-04-26 16:30:01 -05:00
unsafe { ::core::intrinsics::abort() }
msvc: Implement runtime support for unwinding Now that LLVM has been updated, the only remaining roadblock to implementing unwinding for MSVC is to fill out the runtime support in `std::rt::unwind::seh`. This commit does precisely that, fixing up some other bits and pieces along the way: * The `seh` unwinding module now uses `RaiseException` to initiate a panic. * The `rust_try.ll` file was rewritten for MSVC (as it's quite different) and is located at `rust_try_msvc_64.ll`, only included on MSVC builds for now. * The personality function for all landing pads generated by LLVM is hard-wired to `__C_specific_handler` instead of the standard `rust_eh_personality` lang item. This is required to get LLVM to emit SEH unwinding information instead of DWARF unwinding information. This also means that on MSVC the `rust_eh_personality` function is entirely unused (but is defined as it's a lang item). More details about how panicking works on SEH can be found in the `rust_try_msvc_64.ll` or `seh.rs` files, but I'm always open to adding more comments! A key aspect of this PR is missing, however, which is that **unwinding is still turned off by default for MSVC**. There is a [bug in llvm][llvm-bug] which causes optimizations to inline enough landing pads that LLVM chokes. If the compiler is optimized at `-O1` (where inlining isn't enabled) then it can bootstrap with unwinding enabled, but when optimized at `-O2` (inlining is enabled) then it hits a fatal LLVM error. [llvm-bug]: https://llvm.org/bugs/show_bug.cgi?id=23884
2015-06-19 16:57:06 -05:00
}