226 lines · plain
1===============2ShadowCallStack3===============4 5.. contents::6 :local:7 8Introduction9============10 11ShadowCallStack is an instrumentation pass, currently only implemented for12aarch64 and RISC-V, that protects programs against return address overwrites13(e.g. stack buffer overflows.) It works by saving a function's return address14to a separately allocated 'shadow call stack' in the function prolog in15non-leaf functions and loading the return address from the shadow call stack16in the function epilog. The return address is also stored on the regular stack17for compatibility with unwinders, but is otherwise unused.18 19The aarch64 implementation is considered production ready, and20an `implementation of the runtime`_ has been added to Android's libc21(bionic). An x86_64 implementation was evaluated using Chromium and was found22to have critical performance and security deficiencies--it was removed in23LLVM 9.0. Details on the x86_64 implementation can be found in the24`Clang 7.0.1 documentation`_.25 26.. _`implementation of the runtime`: https://android.googlesource.com/platform/bionic/+/808d176e7e0dd727c7f929622ec017f6e065c582/libc/bionic/pthread_create.cpp#12827.. _`Clang 7.0.1 documentation`: https://releases.llvm.org/7.0.1/tools/clang/docs/ShadowCallStack.html28 29Comparison30----------31 32To optimize for memory consumption and cache locality, the shadow call33stack stores only an array of return addresses. This is in contrast to other34schemes, like :doc:`SafeStack`, that mirror the entire stack and trade-off35consuming more memory for shorter function prologs and epilogs with fewer36memory accesses.37 38`Return Flow Guard`_ is a pure software implementation of shadow call stacks39on x86_64. Like the previous implementation of ShadowCallStack on x86_64, it is40inherently racy due to the architecture's use of the stack for calls and41returns.42 43Intel `Control-flow Enforcement Technology`_ (CET) is a proposed hardware44extension that would add native support to use a shadow stack to store/check45return addresses at call/return time. Being a hardware implementation, it46would not suffer from race conditions and would not incur the overhead of47function instrumentation, but it does require operating system support.48 49.. _`Return Flow Guard`: https://xlab.tencent.com/en/2016/11/02/return-flow-guard/50.. _`Control-flow Enforcement Technology`: https://software.intel.com/sites/default/files/managed/4d/2a/control-flow-enforcement-technology-preview.pdf51 52Compatibility53-------------54 55A runtime is not provided in compiler-rt so one must be provided by the56compiled application or the operating system. Integrating the runtime into57the operating system should be preferred since otherwise all thread creation58and destruction would need to be intercepted by the application.59 60The instrumentation makes use of the platform register ``x18`` on AArch64,61``x3`` (``gp``) on RISC-V with software shadow stack and ``ssp`` on RISC-V with62hardware shadow stack, which needs `Zicfiss`_ and ``-fcf-protection=return``.63Users can choose between the software and hardware based shadow stack64implementation on RISC-V backend by passing ``-fsanitize=shadow-call-stack``65or ``Zicfiss`` with ``-fcf-protection=return``.66For simplicity we will refer to this as the ``SCSReg``. On some platforms,67``SCSReg`` is reserved, and on others, it is designated as a scratch register.68This generally means that any code that may run on the same thread as code69compiled with ShadowCallStack must either target one of the platforms whose ABI70reserves ``SCSReg`` (currently Android, Darwin, Fuchsia and Windows) or be71compiled with a flag to reserve that register (e.g., ``-ffixed-x18``). If72absolutely necessary, code compiled without reserving the register may be run on73the same thread as code that uses ShadowCallStack by saving the register value74temporarily on the stack (`example in Android`_) but this should be done with75care since it risks leaking the shadow call stack address.76 77.. _`Zicfiss`: https://github.com/riscv/riscv-cfi/blob/main/cfi_backward.adoc78.. _`example in Android`: https://android-review.googlesource.com/c/platform/frameworks/base/+/80371779 80Because it requires a dedicated register, the ShadowCallStack feature is81incompatible with any other feature that may use ``SCSReg``. However, there is82no inherent reason why ShadowCallStack needs to use a specific register; in83principle, a platform could choose to reserve and use another register for84ShadowCallStack, but this would be incompatible with the ABI standards85published in AAPCS64 and the RISC-V psABI.86 87Special unwind information is required on functions that are compiled88with ShadowCallStack and that may be unwound, i.e. functions compiled with89``-fexceptions`` (which is the default in C++). Some unwinders (such as the90libgcc 4.9 unwinder) do not understand this unwind info and will segfault91when encountering it. LLVM libunwind processes this unwind info correctly,92however. This means that if exceptions are used together with ShadowCallStack,93the program must use a compatible unwinder.94 95Security96========97 98ShadowCallStack is intended to be a stronger alternative to99``-fstack-protector``. It protects from non-linear overflows and arbitrary100memory writes to the return address slot.101 102The instrumentation makes use of the ``SCSReg`` register to reference the shadow103call stack, meaning that references to the shadow call stack do not have104to be stored in memory. This makes it possible to implement a runtime that105avoids exposing the address of the shadow call stack to attackers that can106read arbitrary memory. However, attackers could still try to exploit side107channels exposed by the operating system `[1]`_ `[2]`_ or processor `[3]`_108to discover the address of the shadow call stack.109 110.. _`[1]`: https://eyalitkin.wordpress.com/2017/09/01/cartography-lighting-up-the-shadows/111.. _`[2]`: https://www.blackhat.com/docs/eu-16/materials/eu-16-Goktas-Bypassing-Clangs-SafeStack.pdf112.. _`[3]`: https://www.vusec.net/projects/anc/113 114Unless care is taken when allocating the shadow call stack, it may be115possible for an attacker to guess its address using the addresses of116other allocations. Therefore, the address should be chosen to make this117difficult. One way to do this is to allocate a large guard region without118read/write permissions, randomly select a small region within it to be119used as the address of the shadow call stack and mark only that region as120read/write. This also mitigates somewhat against processor side channels.121The intent is that the Android runtime `will do this`_, but the platform will122first need to be `changed`_ to avoid using ``setrlimit(RLIMIT_AS)`` to limit123memory allocations in certain processes, as this also limits the number of124guard regions that can be allocated.125 126.. _`will do this`: https://android-review.googlesource.com/c/platform/bionic/+/891622127.. _`changed`: https://android-review.googlesource.com/c/platform/frameworks/av/+/837745128 129The runtime will need the address of the shadow call stack in order to130deallocate it when destroying the thread. If the entire program is compiled131with ``SCSReg`` reserved, this is trivial: the address can be derived from the132value stored in ``SCSReg`` (e.g. by masking out the lower bits). If a guard133region is used, the address of the start of the guard region could then be134stored at the start of the shadow call stack itself. But if it is possible135for code compiled without reserving ``SCSReg`` to run on a thread managed by the136runtime, which is the case on Android for example, the address must be stored137somewhere else instead. On Android we store the address of the start of the138guard region in TLS and deallocate the entire guard region including the139shadow call stack at thread exit. This is considered acceptable given that140the address of the start of the guard region is already somewhat guessable.141 142One way in which the address of the shadow call stack could leak is in the143``jmp_buf`` data structure used by ``setjmp`` and ``longjmp``. The Android144runtime `avoids this`_ by only storing the low bits of ``SCSReg`` in the145``jmp_buf``, which requires the address of the shadow call stack to be146aligned to its size.147 148.. _`avoids this`: https://android.googlesource.com/platform/bionic/+/808d176e7e0dd727c7f929622ec017f6e065c582/libc/arch-arm64/bionic/setjmp.S#49149 150The architecture's call and return instructions (``bl`` and ``ret``) operate on151a register rather than the stack, which means that leaf functions are generally152protected from return address overwrites even without ShadowCallStack.153 154Usage155=====156 157To enable ShadowCallStack, just pass the ``-fsanitize=shadow-call-stack`` flag158to both compile and link command lines. On aarch64, you also need to pass159``-ffixed-x18`` unless your target already reserves ``x18``. No additional flags160need to be passed on RISC-V because the software based shadow stack uses161``x3`` (``gp``), which is always reserved, and the hardware based shadow call162stack uses a dedicated register, ``ssp``.163However, it is important to disable GP relaxation in the linker when using the164software based shadow call stack on RISC-V. This can be done with the165``--no-relax-gp`` flag in GNU ld, and is off by default in LLD.166 167Low-level API168-------------169 170``__has_feature(shadow_call_stack)``171~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~172 173In some cases one may need to execute different code depending on whether174ShadowCallStack is enabled. The macro ``__has_feature(shadow_call_stack)`` can175be used for this purpose.176 177.. code-block:: c178 179 #if defined(__has_feature)180 # if __has_feature(shadow_call_stack)181 // code that builds only under ShadowCallStack182 # endif183 #endif184 185``__attribute__((no_sanitize("shadow-call-stack")))``186~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~187 188Use ``__attribute__((no_sanitize("shadow-call-stack")))`` on a function189declaration to specify that the shadow call stack instrumentation should not be190applied to that function, even if enabled globally.191 192Example193=======194 195The following example code:196 197.. code-block:: c++198 199 int foo() {200 return bar() + 1;201 }202 203Generates the following aarch64 assembly when compiled with ``-O2``:204 205.. code-block:: none206 207 stp x29, x30, [sp, #-16]!208 mov x29, sp209 bl bar210 add w0, w0, #1211 ldp x29, x30, [sp], #16212 ret213 214Adding ``-fsanitize=shadow-call-stack`` would output the following assembly:215 216.. code-block:: none217 218 str x30, [x18], #8219 stp x29, x30, [sp, #-16]!220 mov x29, sp221 bl bar222 add w0, w0, #1223 ldp x29, x30, [sp], #16224 ldr x30, [x18, #-8]!225 ret226