[llvm] [AArch64] Stack probing for function prologues (PR #66524)

[Momchil Velikov via llvm-commits]

momchil-velikov wrote:

> Apologies for still not being able to create a reproducible example I can share but what I am seeing is the stack probe write overwriting the value at the tip of the stack when I step debug execution:

Can you spot a place where the probe instruction is *not* immediately after a decrement of the stack (disregarding some random register-to-register arithmetic that may appear)?

If you can't (and you should not find such a case), perhaps some instruction writes below the `sp`? That would be against the AArch64.

Is redzone enabled (`--aarch64-redzone=true|false`) "? If it is on, can you check if it makes a difference if you disable it? 

> > All the stack probing should have already finished before the call to `malloc`.

> Only for the containing function, the functions which have their stack probes inlined will be in the middle of the function which then results in this null-deref.

The stack probes (except `alloca` ones)  are emitted by the PrologEpilogInsertion pass, which is very late in the pipeline.
I didn't think inlining could happen after that pass. Can you share yous compilation pipeline (`-- debug-pass=Structure`)?


https://github.com/llvm/llvm-project/pull/66524