[AIX] Port to 64-bit AIX

[bzEq]

We are porting Rust to AIX, see rust-lang/compiler-team#553. This PR adds corresponding assembly in psm for AIX.

[nagisa]

Please also adjust the support table in the README.

[nagisa]

This appears to be using tabs for whitespace. Is it actually required by the assembler? If not, please replace with spaces to keep this consistent with the rest of the project.

(also applies across this file)

[bzEq]

Testing the PR and filling the README.

[bzEq]

@nagisa For the traceback table part, can add @xingxue-ibm who implemented the libunwind of LLVM on AIX, as reviewer.

[bzEq]

running 5 tests
test catch_panic_leaf ... ok
test deep ... ok
test catch_panic ... ok
test catch_panic_inside ... ok
test panic ... ok

test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.08s

@nagisa Is it sufficient to set Callstack as Yes?

[nagisa]

For the traceback table part, can add xingxue-ibm who implemented the libunwind of LLVM on AIX, as reviewer.

I think it is fine as long as it is confirmed these are largely working. Given that unwinding appears to work okay (this is tested by unit tests), I’m happy to not block this on additional reviews.

---

Is it sufficient to set Callstack as Yes?

Unfortunately no, we don’t have unit tests for that currently.

The kind of testing I did for this was stepping through each instruction of the rust_psm_on_stack functions in one of the examples running under a debugger and seeing if a backtrace (bt) produces meaningful output. Another test is seeing if backtrace can be obtained from the function running on the new stack. The meaningful output in this sense should include frames from the old stack as well. So the expected output is someething along the lines of:

$ gdb target/debug/examples/on_stack_fibo
(gdb) br *0x000055555555e600   -- the first instruction of rust_psm_on_stack
(gdb) run
(gdb) bt
#0  0x000055555555e600 in rust_psm_on_stack ()
#1  0x000055555555c374 in psm::rust_psm_on_stack (...) at src/lib.rs:120
#2  psm::on_stack<core::option::Option<u64>, on_stack_fibo::main::{closure_env#0}> (...) at src/lib.rs:201
#3  0x000055555555e2d1 in on_stack_fibo::main (...) at examples/on_stack_fibo.rs:49
(gdb) ni
(gdb) bt
...
(gdb) ni
(gdb) bt

repeating until the exit of the function.

[bzEq]

Here's my gdb script and result

b rust_psm_on_stack
r
while 1
  ni
  bt
end

Run with gdb target/debug/examples/on_stack_fibo < dbg_cmd >debug.log 2>&1
debug.log outputs(only head and tail pasted)

Thread 2 hit Breakpoint 1, 0x000000010014cd60 in rust_psm_on_stack ()
(gdb)  > > >0x000000010014cd64 in rust_psm_on_stack ()
#0  0x000000010014cd64 in rust_psm_on_stack ()
#1  0x000000010014cf20 in psm::on_stack<core::option::Option<u64>, on_stack_fibo::main::{closure_env#0}> (base=0x11007a000, size=131072, callback=
...)
    at src/lib.rs:193
#2  0x0000000100000d2c in on_stack_fibo::main ()
    at examples/on_stack_fibo.rs:49
0x000000010014cd68 in rust_psm_on_stack ()
...
#0  0x000000010014cd80 in rust_psm_on_stack ()
#1  0x000000010014d064 in psm::rust_psm_on_stack (data=1152921504606842056,
    return_ptr=1152921504606842072,
    callback=@0x11001ad20: 0x10014ce00 <psm::on_stack::with_on_stack<core::option::Option<u64>, on_stack_fibo::main::{closure_env#0}>>, sp=0x11009a000)
    at src/lib.rs:118
#2  psm::on_stack<core::option::Option<u64>, on_stack_fibo::main::{closure_env#0}> (base=0x11007a000, size=131072, callback=...) at src/lib.rs:199
#3  0x0000000100000d2c in on_stack_fibo::main ()
    at examples/on_stack_fibo.rs:49
0x000000010014d064 in psm::rust_psm_on_stack (data=1152921504606842056,
    return_ptr=1152921504606842072,
    callback=@0x11001ad20: 0x10014ce00 <psm::on_stack::with_on_stack<core::option::Option<u64>, on_stack_fibo::main::{closure_env#0}>>, sp=0x11009a000)
    at src/lib.rs:118
...
#0  0x090000000005ffa8 in exit () from /usr/lib/libc.a(shr_64.o)
#1  0x0000000100000554 in __start ()
[Inferior 1 (process 11666350) exited normally]
No stack.
(gdb) quit

[nagisa]

Yeah, seems fine to mark that as supported in that case. In an effort to save some time for yourself and myself, I’ll merge this now and I’ll adjust the table as a follow-up. Thanks for contributing!