1
0
Fork 0
A Rust library implementing safe, lightweight context switches, without relying on kernel services
Go to file
whitequark 5a77a01863 Implement OwnedStack. 2016-08-30 12:20:34 -04:00
.travis/docs run on Travis CI container infra 2015-04-19 15:40:43 -04:00
benches Rename Yielder::generate to Yielder::suspend. 2016-08-30 12:20:34 -04:00
src Implement OwnedStack. 2016-08-30 12:20:34 -04:00
tests Implement OwnedStack. 2016-08-30 12:20:34 -04:00
.editorconfig add .editorconfig 2015-04-16 04:17:55 -04:00
.gitignore initial commit 2014-12-23 04:27:28 +01:00
.travis.yml add Travis CI IRC notifications [skip ci] 2016-07-17 20:25:44 -04:00
Cargo.toml Implement OwnedStack. 2016-08-30 12:20:34 -04:00
LICENSE-APACHE Relicense to dual MIT/Apache-2.0 2016-08-20 23:39:19 +02:00
LICENSE-MIT List specific authors in LICENSE-MIT, not just "the developers" 2016-08-20 23:41:10 +02:00
README.md Implement OwnedStack. 2016-08-30 12:20:34 -04:00

README.md

travis rustdoc

libfringe

libfringe is a library implementing safe, lightweight context switches, without relying on kernel services. It can be used in hosted environments (using std) as well as on bare metal (using core).

It provides the following safe abstractions:

  • an implementation of generators, Generator.

It also provides the necessary low-level building blocks:

  • a trait that can be implemented by stack allocators, Stack;
  • a wrapper for using slice references as stacks, SliceStack;
  • a stack allocator based on Box<[u8]>, OwnedStack;
  • a stack allocator based on anonymous memory mappings with guard pages, OsStack.

libfringe emphasizes safety and correctness, and goes to great lengths to never violate the platform ABI.

Usage example

extern crate fringe;

use fringe::{OsStack, Generator};

fn main() {
  let stack = OsStack::new(1 << 16).unwrap();
  let mut gen = Generator::new(stack, move |yielder, ()| {
    for i in 1..4 { yielder.generate(i) }
  });

  println!("{:?}", gen.resume(())); // Some(1)
  println!("{:?}", gen.resume(())); // Some(2)
  println!("{:?}", gen.resume(())); // Some(3)
  println!("{:?}", gen.resume(())); // None
}

Performance

libfringe does context switches in 3ns flat on x86 and x86_64!

test swap ... bench:         6 ns/iter (+/- 0)

Debuggability

Uniquely among libraries implementing context switching, libfringe ensures that the call stack does not abruptly end at the boundary of a generator. Let's consider this buggy code:

extern crate fringe;

use fringe::{OsStack, Generator};

fn main() {
  let stack = OsStack::new(1 << 16).unwrap();
  let mut gen = Generator::new(stack, move |yielder, mut index| {
    let values = [1, 2, 3];
    loop { index = yielder.generate(values[index]) }
  });

  println!("{:?}", gen.resume(5));
}

It crashes with the following backtrace (redacted for clarity):

thread 'main' panicked at 'assertion failed: index < self.len()', ../src/libcore/slice.rs:531
stack backtrace:
   [... core::panicking internals ...]
   9:     0x559ee50f677b - core::panicking::panic::hbfac80217e56ecbe
  10:     0x559ee50b6b4c - core::slice::_<impl core..ops..Index<usize> for [T]>::index::hcb117ddcc7cf2f33
                        at .../src/libcore/slice.rs:21
  11:     0x559ee50b7288 - crash_test::main::_{{closure}}::hc7da249d76d51364
                        at .../crash_test.rs:9
  12:     0x559ee50b6f23 - _<fringe..generator..Generator<Input, Output, Stack>>::unsafe_new::generator_wrapper::ha2da172d4f041d38
                        at .../libfringe/src/generator.rs:94
  13:     0x559ee50b76d3 - fringe::arch::imp::init::trampoline_2::hdb11eb4bdafcdeb9
                        at .../libfringe/src/arch/x86_64.rs:71
  14:     0x559ee50b76c4 - fringe::arch::imp::init::trampoline_1::h6b071b2a8ea6aab3
                        at .../libfringe/src/arch/x86_64.rs:43
  15:     0x559ee50b7098 - _<fringe..generator..Generator<Input, Output, Stack>>::resume::h8d2b90d386543e29
                        at .../libfringe/src/arch/x86_64.rs:131
                        at .../libfringe/src/context.rs:52
                        at .../libfringe/src/generator.rs:129
  16:     0x559ee50b71c8 - crash_test::main::hfc5e04bc99de7a6a
                        at .../crash_test.rs:12
  [... standard library startup internals ...]

Similarly, debuggers, profilers, and all other tools using the DWARF debug information have full insight into the call stacks.

Note that the stack should be deep enough for the panic machinery to store its state—at any point there should be at least 8 KiB of free stack space, or panicking will result in a segfault.

Limitations

The only architectures currently supported are x86 and x86_64. Windows is not supported (see explanation below).

Installation

libfringe is a Cargo package. It's not stable software yet, so you'll have to use it as a git dependency. Add this to your Cargo.toml:

[dependencies.fringe]
git = "https://github.com/edef1c/libfringe.git"

To use libfringe on a bare-metal target, add the no-default-features key:

[dependencies.fringe]
git = "https://github.com/edef1c/libfringe.git"
no-default-features = true

Feature flags

libfringe provides some optional features through Cargo's feature flags. Currently, all of them are enabled by default.

alloc

This flag enables dependency on the alloc crate, which is required for the OwnedStack.

valgrind

This flag enables Valgrind integration. libfringe will register context stacks with Valgrind.

Internals

libfringe uses two key implementation techniques.

Compiler-assisted register spilling

Traditionally, libraries implementing context switches in userspace have to spill all callee-saved registers. On the other hand, libfringe fully inlines calls to every function that eventually results in a context switch, and uses an inline assembly statement marking every register as clobbered to implement the context switching itself.

As a result, only minimal work needs to be performed in the context switching code (LLVM does not support spilling the frame pointer), which is especially important on architectures with lots of callee-saved registers.

Call stack splicing

Non-Windows platforms use DWARF for both stack unwinding and debugging. DWARF call frame information is very generic to be ABI-agnostic—it defines a bytecode that describes the actions that need to be performed to simulate returning from a function. libfringe uses this bytecode to specify that, after the generator function has returned, execution continues at the point where the generator function was resumed the last time.

Windows compatibility

As was said, libfringe emphasizes following the platform ABI. On Windows, the platform ABI does not allow moving the stack pointer from the range designated by the OS during thread creation. Therefore, the technique used by libfringe on *nix platforms is not applicable, and libfringe does not provide Windows support.

You might ask, "but what about mioco?" The mioco library uses the context library to implement context switches, which is little more than a wrapper of boost::context. The boost::context library changes undocumented fields in the TIB during every context switch to try and work around the restrictions placed by the Windows platform ABI. This has failed before and it is bound fail again, breaking existing code that uses boost::context in unexpected and complicated ways. The authors of libfringe consider this unacceptable.

The only supported way to implement user-mode context switching on Windows is to use fibers. There are no reasons the safe abstractions provided by libfringe could not be implemented on top of that; it is simply not yet done. This should be straightforward and an implementation is welcome.

License

Licensed under either of

at your option.

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.