How to register custom exception in Intel SGX? - c++

Question

I'm having difficulty to get my custom exception handler to work.
This is the Enclave code:
#include "Enclave_DivideZero_t.h"
#include "sgx_trts_exception.h"
#include "sgx_trts.h"
#include <string>
static char buf[200] = "Handler not called";
int divide_by_zero_handler(sgx_exception_info_t* info) {
buf[0] = '1';
return EXCEPTION_CONTINUE_EXECUTION;
}
void Enclave_DivideByZero() {
Ocall_printf(buf);
if (sgx_register_exception_handler(1, divide_by_zero_handler) == NULL) {
Ocall_printf("register failed");
} else {
Ocall_printf("register success");
}
int a(1);
int b(3/(a-a));
(void) a;
(void) b;
Ocall_printf(buf);
}
We used buf as an indication of whether the handler has been actually executed. However, the output is this:
Enclave created!
[Ocall printf] - Handler not called
[Ocall printf] - register success
[Ocall printf] - Handler not called <-- should be: "1andler not called" ("1andler" instead of "Handler")
Also, here is the App code (i.e. the untrusted code)
#include "stdafx.h"
#include "sgx_urts.h"
#include "Enclave_DivideZero_u.h"
#define ENCLAVE_FILE _T("Enclave_DivideZero.signed.dll")
sgx_status_t createEnclave(sgx_enclave_id_t *eid) {
sgx_status_t ret = SGX_SUCCESS;
sgx_launch_token_t token = {0};
int updated = 0;
ret = sgx_create_enclave(ENCLAVE_FILE, SGX_DEBUG_FLAG, &token, &updated, eid, NULL);
return ret;
}
void Ocall_printf( char* str) {
printf("[Ocall printf] - %s\n", str);
}
int _tmain(int argc, _TCHAR* argv[]) {
sgx_enclave_id_t eid;
sgx_status_t res = createEnclave(&eid);
if (res != SGX_SUCCESS) {
printf("App: error-, failed to create enclave.\n");
return -1;
} else {
printf("Enclave created!\n");
}
Enclave_DivideByZero(eid);
return 0;
}
The Problem: As you can see from the output indicates that the handler is registered successfully, but is not executed.
1- Why the registration doesn't work?
2- I tried to put the registeration with the App code, and for this I had to add the handler in the edl, the problem for this is passing the sgx_exception_info_t *info param, which is not clear which flags it needs (i.e. [in, .. <'flags'>]). So, is it correct to define it inside the Enclave?
p.s. I ran the code with Prerelease Mode.
[EDIT] I documented the project + code that I conducted on SGX here

Answer 1

So the question is a bit old now, but I got stuck at the same place, so maybe someone can still use my insight.
In trts_veh.cpp, where it is checked whether the exception was handled, it says in a comment that the "instruction triggering the exception will be executed again". This leads to the exception handling loop, as you noticed yourself.
However, you can increase the rip yourself, as it is stored in the pointer to the sgx_exception_info_t, by using "info->cpu_context.rip += 2;". Adding this inside your exception handler should do the trick.