Buffer overflow on too many break-points

[pruzko]

If you set a break-point on the same address numerous times the python process crashes. This happens e.g. when you set a bp from a callback function:

addr_1 = 0x...
addr_2 = 0x...

def cb_1(): print('1')

def cb_2():
    with vm.break_on_physical(addr_1, cb_1)
        ...

with vm.break_on_physical(addr_2, cb_2)
    ...

# guest is calling addr_2 and addr_1 fiercely

The solution is rather obvious - one must keep track of addresses that already have a bp set and prevent duplicate bps. However, I thought SIGSEGV is not intended behavior so I did a little investigation.

Long story short, the buffer in struct Worker gets overflown and the co_thread pointer gets malformed. A following call to co_switch is going to kill the python process.

[bamiaux]

I will look into it, meanwhile, can you test with a bigger stacksize ?

icebox/src/icebox/icebox/core/state.cpp

Line 104 in 8a97bf4

constexpr auto g_stack_size = 0x400000; // 4mb stack size

[bamiaux]

The coroutine stuff is not really required anymore since we do not use page faults anymore. I should probably remove it