wasm2c: parametrize memory bounds checks on a per-memory basis

[keithw]

(Sequenced behind #2506)

This PR allows "software-bounds-checked" memories and "guard-page-checked" memories to coexist in the same module.

It creates two versions of every memory operation: an unrestricted version (that works with any memory) and a _default32 version (for memories with default page size and i32 indexing).

The unrestricted version calls MEMCHECK_GENERAL, which does a 64-bit software RANGE_CHECK to check that the operation reads/writes within the bounds of the memory.

The _default32 version calls MEMCHECK_DEFAULT32, which is the same as the old MEMCHECK: if the runtime declares WASM_RT_MEMCHECK_GUARD_PAGES, it will do nothing. Otherwise it will do a 32-bit software RANGE_CHECK (which seems to be one less instruction than the 64-bit RANGE_CHECK).

This is a stepping stone to supporting custom-page-sizes (which will need to be software bounds-checked) (#2508).

[shravanrn]

Not a strong opinion, but would it be simpler to have something like?

#define DEFINE_STORE(name, t1, t2, MEMCHK, suffix)                                  \
  static inline void name##_##suffix(wasm_rt_memory_t* mem, u64 addr,  t2 value) {  \
    MEMCHK(mem, addr, t1);                                                          \
    t1 wrapped = (t1)value;                                                         \
    wasm_rt_memcpy(MEM_ADDR_MEMOP(mem, addr, sizeof(t1)), &wrapped, sizeof(t1));    \
  }

DEFINE_STORE(i32_store, u32, u32, MEMCHECK_DEFAULT32, _default32);
DEFINE_STORE(i32_store, u32, u32, MEMCHECK_GENERAL, _general);

[keithw]

I guess I was trying to avoid the churn of duplicating every occurrence of DEFINE_LOAD/DEFINE_STORE/DEFINE_SIMD_LOAD_FUNC/DEFINE_SIMD_LOAD_LANE/DEFINE_SHARED_LOAD/etc. ... I think on balance the status quo is probably the less-bad but I don't feel that strongly either.

[shravanrn]

Nit: Might be cleaner to have

// Check if we should reallocate using mmap
#if WASM_RT_USE_MMAP
if(!is64) {
  ... // original code
  return;
}
#endif

// Fallback to malloc
memory->data = calloc(byte_length, 1);

[keithw]

Yeah... I agree this one is clearer but it seemed nice to have consistency with grow_memory_impl where the early-return is harder to do. :-/