Better support for large map_size

[acdxvfsvd]

By default ld_preload_fuzz reads host_config.bitmap_size from the configuration and maps a memory area with the corresponding size, then it overwrites agent_config.coverage_bitmap_size when processing an LTO-compiled binary but doesn't reallocate the memory.

trace_buffer = mmap((void*)NULL, host_config.bitmap_size, PROT_READ | PROT_WRITE, MAP_SHARED | MAP_ANONYMOUS, -1, 0);
memset(trace_buffer, 0xff, host_config.bitmap_size);
// ...
if (get_harness_state()->afl_mode && __afl_final_loc_ptr){
    unsigned int map_size = __afl_final_loc == 0 ? 65536 : __afl_final_loc;
    agent_config.coverage_bitmap_size = map_size;
}

There is no such argument about bitmap_size when generating configuration using nyx_config_gen.py, so the value will come from the included default_config_kernel.ron, which is 0x10000.
This works well in most cases but may crash when the map_size of the binary is larger than 0x10000. It would be great to reallocate the memory after detecting the map_size, or just add an argument to the config generator for customizing the map_size.

[schumilo]

I was able to successfully reproduce the bug and it seems that my patch has now fixed the issue.