Rationale behind "while" in arm_inline_hash_lookup(...) about mambo

[lgeek]

There are a few infinite loops like that in MAMBO. They are basically asserts which stop execution and allow you to attach a debugger and inspect the context instead of exiting. I'll admit that doing that silently is less than ideal.

In this case, the issue is that the values saved at scanner_arm.c#L411 and which would be restored by the dispatcher are stale (because those values would be overwritten by the POP). I think the best solution is to POP (reglist & 3) before inserting the lookup routine, but I'll need to take a closer look. By the way, where have you encountered this instruction? It doesn't look compiler generated, which is why I've never encountered it before.

[alirazeen]

What you say makes sense. So in that case, technically speaking, shouldn't the assert statement also check if r2 is being POPed? I'm aware that arm_inline_hash_lookup doesn't use r2 but then the values saved in scanner_arm.c#L411 saves r2 as well.

[lgeek]

The updated value of R2 (i.e. after the POP) is saved at scanner_arm.c#L428, so it doesn't have to be included in the assert. Now, I've spent a bit of time looking at the file history to try to understand why R2 is also saved by the STM, but as far as I can tell it's always been redundant.