Logsload723

There is no support of the 128 bit integers (128uint/128int) in Windows and Linux SGX SDK Great notice about basics of SGX threading Intel Management Engine (Intel ME) for the SGX remote attestation. The Intel SGX SDK’s solution to this problem is to copy the contents of data buffers into and out of enclaves, and have the ECALLs and OCALLs operate on these copies of the original memory buffer. When you pass a pointer into an enclave, you specify in the EDL whether the buffer referenced by the pointer is being pass into the call, out of. Compared to other SGX software stacks, the official Intel SGX SDK seems to be the most secure open-source solution, designed for security and with defenses against Spectre attacks. Azure also exclusivelysupportstheIntelSGXSDKinitscloudenvironment.

1. ECREATE

• [Intel SGX Explained p63] Section 5.3.1. Creation
• [Programming References p21] Section 5.3. ECREATE

An enclave is born when the system software issues the ECREATE instruction, which turns a free EPC page into the SECS for the new enclave.

ECREATE copies an SECS structure outside the EPC into an SECS page inside the EPC. The internal structure of SECS is not accessible to software.
Software sets the following fields in the source structure: SECS:BASEADDR, SECS:SIZE, and ATTRIBUTES.

ECREATE validates the information used to initialize the SECS, and results in a page fault (#PF) or general protection fault (#PF) if the information is not valid.
ECREATE will also fault if SECS target page is in use; already valid; outside the EPC; adresses are not aligned; unused PAGEINFO fields are not zero.

2. EADD

• [Intel SGX Explained p64] Section 5.3.2. Loading
• [Programming References p11] Section 5.3. EADD

The system software can use EADD instructions to load the initial code and data into the enclave. EADD is used to create both TCS pages and regular pages.
This function copies a source page from non-enclave memory into the EPC, associates the EPC page with an SECS page residing in the EPC, and stores the linear address and security attributes in EPCM.

EADD reads its input data from a Page Information (PAGEINFO) structure.

The PAGEINFO structure contains

• The virtual address of the EPC page (LINADDR)
• The virtual address of the non-EPC page whose contents will be copied into the newly allocated EPC page (SRCPGE)
• A virtual address that resolves to the SECS of the enclave that will own the page (SECS).
• The virtual address, pointing to a Security Information (SECINFO) structure, which contains the newly allocated EPC page’s access permissions (R, W, X) and its EPCM page type (RT_REG or PT_TCS).

{: .center-image}

EADD validates its inputs, and modifies the newly allocated EPC page and its EPCM entry.

EADD ensures

• The EPC page is not allocated to another enclave.
• The page’s virtual address falls within the enclave’s ELRANGE.
• All the reserved fields in SECINFO are set to zero.

lin_addr` to get a simulated enclave, `ce`. According to the paper *'Intel SGX Explained'*, it is the virtual address of the EPC page. Meanwhile, according to the *'Intel SGX Programming Reference'*, it seems that virtual address of the EPC page is passed as a parameter of `EADD` function, `epc_lin_addr`.Are `pi->lin_addr` and `epc_lin_addr` different? They are almost same, but different, according to the driver API function, `add_enclave_page()`.```c++linux-sgx/sdk/simulation/driver_api/driver_api.cppint add_enclave_page(sgx_enclave_id_t enclave_id, void* source, size_t offset, const sec_info_t &secinfo, uint32_t attr){ sec_info_t sinfo; page_info_t pinfo; CEnclaveMngr* mngr; CEnclaveSim* ce; ... memset(&pinfo, 0, sizeof(pinfo)); pinfo.secs = ce->get_secs(); pinfo.lin_addr = (char*)ce->get_secs()->base + offset; pinfo.src_page = source; pinfo.sec_info = &sinfo; // Passing NULL here when there is no EPC mgmt. return (int)DoEADD_SW(&pinfo, GET_PTR(void, ce->get_secs()->base, offset));}````_EADD` instruction is called during `DoEADD_SW` function is handled.`pinfo.lin_addr` is initialized as `(char*)ce->get_secs()->base + offset`, and `epc_lin_addr` is set as `GET_PTR(void, ce->get_secs()->base, offset)`.`GET_PTR` is defined as `#define GET_PTR(t, p, offset) reinterpret_case( reinterpret_case(p) + static_cast(offset) )` in `linux-sgx/common/inc/internal/util.h`.While `epc_lin_addr` saves the address `ce->get_secs()->base + sizeof(size_t) * offset`, `pinfo.lin_addr` saves the address `ce->get_secs()->base + offset`.Of course, there will be no problem on calling `get_enclave()`, as it is defined as follows.```c++linux-sgx/sdk/simulation/uinst/enclave_mngr.cppCEnclaveSim* CEnclaveMngr::get_enclave(const void* base_addr){ CEnclaveSim* ce = NULL; ... std::list::iterator it = m_enclave_list.begin(); for (; it != m_enclave_list.end(); ++it) { secs_t* secs = (* it)->get_secs(); if (base_addr >= secs->base && PTR_DIFF(base_addr, secs->base) < secs->size) { ce = * it; break; } } ...}```It just checks which enclave `base_addr` is in its ELRANGE (`BASEADDR` ~ `BASEADDR+SIZE`).-->

2-1. How a free EPC page is selected in SGX simulation mode?

Simulation implementation might be different from hardware implementation. In simulation mode, ECREATE allocates all EPC pages via mmap().

As SGX simulation simulates SGX behavior by software, it copies EPC page data with virtual address. Hence _EADD() does not have detailed information about picking a physical EPC page.

2.2. How system software selects a EPC page in SGX hardware mode?

SGX simulation code can’t tell EPC page allocation in detail, as it is also a software, so it cannot use the physical address.

To understand actual implementation, I first tried to understand Intel SGX Programming Reference deeply. There is a table explaining inputs for the instruction.

Table. Instruction Operand Encoding{: .center}

Op/En	EAX	RBX	RCX
IR	EADD (in)	Address of PAGEINFO (in)	Address of the destination EPC page (in)

As you see the above table, addresses of PAGEINFO and target EPC page should be saved in the register RBX and RCX, respectively. The target EPC page is already determined, which means system software is responsible for selecting one. This is also explained in ISCA ‘15 tutorial slide. [link]

{: .center-image width=“600px”}

Then this means I need to search a code that calls EADD instruction.

Simulation function which calls EADD instruction is add_enclave_page() in sdk/simulation/driver_api/driver_api.cpp. This is a simulation code because it is in simulation directory.
Then there should be a function with the same name for hardware?

And yes. There is.

Note that PSW (Platform SoftWare) is for actual hardware. It calls ioctl(), which calls sgx_ioctl_enclave_add_page() in linux-sgx-driver. linux-sgx-driver is separately provided in [here].

In linux-sgx-driver/isgx_ioctl.c,

ioctl() from enclave_creator_hw.cpp

• calls isgx_ioctl_enclave_add_page() at linux-sgx-driver/isgx_ioctl.c:548
• calls __enclave_add_page() at linux-sgx-driver/isgx_ioctl.c:440
• calls construct_enclave_page at linux-sgx-driver/isgx_ioctl.c:122
• calls isgx_alloc_epc_page() at linux-sgx-driver/isgx_page_cache.c:429
• calls isgx_alloc_epc_page_fast() at linux-sgx-driver/isgx_page_cache.c:411
  which picks the first entry from driver’s EPC page list.

isgx Linux SGX driver manages EPC page instances by using the linked list(static LIST_HEAD(isgx_free_list) at isgx_page_cache.c:25) and the number of free EPC pages(unsigned int isgx_nr_free_epc_pages at isgx_page_cache.c:31).

The type of EPC pages is struct isgx_epc_page, defined as follows.

Here, pa is the physical address for the EPC page. How pa is determined?

Each EPC page that is put into free list is allocated in the function isgx_page_cache_init() at linux-sgx-driver/isgx_page_cache.c:360.

Each EPC page has the physical address as start + i. start is the first parameter of the function.

The function isgx_page_cache_init() is called in isgx_init() at linux-sgx-driver/isgx_main.c:190.

Is There An Sgx Sdk Download

As shown above, start is the value of isgx_epc_base variable.
isgx_epc_base is initialized by the function isgx_init_platform() at linux-sgx-driver/isgx_main.c:133.

You can see what isgx_epc_base value is in your machine, as SGX driver prints it in kernel message buffer by default.

{: .center-image}

My machine tells that 32MiB of EPC is allocated with the physical base memory address 0x80000000.

All EPC page instance struct isgx_epc_page has a pa variable, which contains the physical address of the EPC page, and is initialized as EPC base address + offset.

And, managing page mapping table is also a responsibility of system software. Linux SGX driver insert PTE via calling vm_insert_pfn() at linux-sgx-driver/isgx_util.c:67 by using epc_page->pa and expected virtual address enclave_page->addr as follows.

To be concluded, when a SGX platform is initialized, several EPC page instances (struct isgx_epc_page) are allocated to represent all EPC pages. System software manages them as a linked list, called isgx_free_list. When EADD is called, system software picks a free EPC page instance from the list, and create a page table entry, pointing the physical address that is saved in epc_page->pa, with the expected virtual address enclave_page->addr, a part of user enclave’s ELRANGE.

3. EEXTEND

• [Intel SGX Explained p64] Section 5.3.2. Loading
• [Programming References p31] Section 5.3. EEXTEND

While loading an enclave, the system software will also use the EEXTEND instruction, which updates the enclave’s measurement used in the software attestation process.
It updates the MRENCLAVE measurement register of an SECS with the measurement of an EXTEND string compromising of “EEXTEND” || ENCLAVEOFFSET || PADDING || 256 bytes of the enclave page.

Is There An Sgx Sdk

RCX register contains the effective address of the 256 byte region of an EPC page to be measured.

From Intel 64 and IA-32 Architectures Software Developer’s Manual, Volume 3D, Part 4:Section 39.1.2. EADD and EEXTEND Interaction

Software can measure a 256 byte region as determined by the by the developer by invoking EEXTEND. Thus to measure an entire 4KB page, system software must execute EEXTEND 16 times. Each invocation of EEXTEND adds to the cryptographic log information about which region is being measured and the measurement of the section.

Is There An Sgx Sdk Free

4. EINIT

• [Intel SGX Explained p64] Section 5.3.3. Initialization
• [Programming References p34] Section 5.3. EINIT

This function is the final instruction executed in the enclave build process. After EINIT, the MRENCLAVE measurement is cimplete, and the enclave is ready to start user code execution using EENTER instruction.

When EINIT completes successfully, it sets the enclave’s INIT attribute to true. This opens the way for ring 3 application software to execute the enclave’s code, using the SGX instructions.

On the other hand, once INIT is set to true, EADDcannot be invoked on that enclave anymore, so the system software must load all the pages that make up the enclave’s initial state before executing the EINIT instruction.

References

• Intel Software Guard Extensions Programming Reference. [link]
• Intel SGX Explained. [link]
• Intel SGX Tutorial Slide presented in ISCA 2015.[link]
• Intel SGX SDK Github Repository. [link]
• Intel SGX Linux Driver Github Repository.[link]

License

Is There An Sgx Sdk App

All source codes are from Intel SGX SDK Github repository and Intel SGX Linux driver Github repository, released under BSD License 2.0 and GNU General Public License 2.0, respectively.

Intel SGX SDK

Copyright (C) 2011-2017 Intel Corporation. All rights reserved.
Redistribution and use in source and binary forms, with or withoutmodification, are permitted provided that the following conditions are met:

• Redistributions of source code must retain the above copyright
  notice, this list of conditions and the following disclaimer.
• Redistributions in binary form must reproduce the above copyright
  notice, this list of conditions and the following disclaimer in the
  documentation and/or other materials provided with the distribution.
• Neither the name of the nor the
  names of its contributors may be used to endorse or promote products
  derived from this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” ANDANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIEDWARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE AREDISCLAIMED. IN NO EVENT SHALL

This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; version 2 of the License.