This adds trapframe unwinding support through dwarf cfi. It has been tested under lldb.
MFC after: 2 weeks
MFC to: stable/14, stable/15
Sponsored by: FreeBSD Foundation
Details
Details
Diff Detail
Diff Detail
- Repository
- rG FreeBSD src repository
- Lint
Lint Skipped - Unit
No Test Coverage - Build Status
Buildable 76864 Build 73747: arc lint + arc unit
Event Timeline
Comment Actions
Is this necessary? Do you not want LLDB to print trapframes as special, like KGDB does, and like GDB does for signal frames?
| sys/riscv/include/asm.h | ||
|---|---|---|
| 48–50 | This seems wrong | |
Comment Actions
An lldb dev working on Darwin's kernel frame unwinding was against porting kgdb's sniffer method to lldb. Instead, he recommended using CFI.
CFI can mark itself as signal frame using .cfi_signal_frame as I wrote in trap_cfi_entry in trap_cfi.h
| sys/riscv/include/asm.h | ||
|---|---|---|
| 48–50 | I didn't want to include .cfi_startproc and .cfi_endproc in ENTRY() and END() but in trap_cfi macros. The only cfi usage I found in riscv was rtld_start, so it is possible to remove .cfi_startproc and endproc from ENTRY() and END() without much refactoring. arm64 is not the case because it has a special security instruction that must be come after cfi_startproc and refactoring them will require too much change. | |
| sys/riscv/include/asm.h | ||
|---|---|---|
| 48–50 | Let me rephrase: please don't do this. ENTRY and END should have all the boilerplate for a normal function, which they now do not. | |