dbg_breakpoint

Crates.iodbg_breakpoint
lib.rsdbg_breakpoint
version0.1.1
sourcesrc
created_at2024-09-25 22:58:08.726301
updated_at2024-12-01 08:11:42.895258
descriptionSet breakpoints with the `breakpoint!()` macro on many target architectures and popular OSes like FreeBSD, macOS, iOS, Linux distro's, Windows without using the nightly toolchain. Break into the debugger with an easy `breakpoint_if_debugging()` call, too!
homepagehttps://github.com/kromych/dbg_breakpoint
repositoryhttps://github.com/kromych/dbg_breakpoint
max_upload_size
id1386748
size24,194
(kromych)

documentation

https://github.com/kromych/dbg_breakpoint

README

Breakpoints when the debugger is attached

Set breakpoints with the breakpoint!() macro on many target architectures and popular OSes like FreeBSD, macOS, iOS, Linux distro's, Windows without using the nightly toolchain. Break into the debugger with an easy breakpoint_if_debugging() call, too!

Well, sure, but why?

  • It might be more convinient to add the call to breakpoint_if_debugging from inside the comfort of your editor than to remember the incantion in the debugger,
  • Some callsites like lambdas and async routines/coroutines can be tricky to set a breakpoint to in the debugger due to name mangling or because the toolchain doesn't give them a name that is easily-discovered/human-friendly,
  • Can add this to your #[panic_handler] to break into the debugger on a panic.

This model might be reminiscent of "semihosting" where the execution environment includes a host or a debugger who's services might be requested by the program.

Here is the example of how one can make use of this: runme.rs. Do exercise extreme caution when using any of this in the production environment, i.e. out of the inner development loop. Heisenbugs and crashes might be sighted.

Platform- and target-specific notes follow.

Windows

The library provides breakpoint_if_debugging() and breakpoint_if_debugging_seh() The latter might be useful to detect the debugger if it is trying to hide its presence via some cheap tricks.

Linux, macOS and FreeBSD

The debugger detection logic will detect any tracer like strace as the debugger, and if the tracer isn't able to skip over the breakpoint CPU instruction, the program will crash. That can be fixed by handling SIGTRAP inside your program.

arm64

brk #imm16 is used for breakpoint on arm64.

Just FYI, the #imm16 value can be inside the Linux kernel 6.1 at the time of writing:

  • 0x004: for installing kprobes
  • 0x005: for installing uprobes
  • 0x006: for kprobe software single-step
  • 0x400 - 0x7ff: kgdb
  • 0x100: for triggering a fault on purpose (reserved)
  • 0x400: for dynamic BRK instruction
  • 0x401: for compile time BRK instruction
  • 0x800: kernel-mode BUG() and WARN() traps
  • 0x9xx: tag-based KASAN trap (allowed values 0x900 - 0x9ff)
  • 0x8xxx: Control-Flow Integrity traps

Here, we're talking the user mode yet the above illustrates the point that the value supplied after brk influences what to expect.

For __builtin_trap(), gcc produces brk #0x3e8, clang generates brk #1. This library uses 0xf000 as the debuggers on Windows and macOS skip over the debug trap automatically in this case by advancing the instruction pointer behind the curtain.

See also

License

dbg_breakpoint is free and open source. All code in this repository is dual-licensed under either:

at your option.

Commit count: 15

cargo fmt