180 lines · plain
1.. SPDX-License-Identifier: GPL-2.02 3======================================================4Control-flow Enforcement Technology (CET) Shadow Stack5======================================================6 7CET Background8==============9 10Control-flow Enforcement Technology (CET) covers several related x86 processor11features that provide protection against control flow hijacking attacks. CET12can protect both applications and the kernel.13 14CET introduces shadow stack and indirect branch tracking (IBT). A shadow stack15is a secondary stack allocated from memory which cannot be directly modified by16applications. When executing a CALL instruction, the processor pushes the17return address to both the normal stack and the shadow stack. Upon18function return, the processor pops the shadow stack copy and compares it19to the normal stack copy. If the two differ, the processor raises a20control-protection fault. IBT verifies indirect CALL/JMP targets are intended21as marked by the compiler with 'ENDBR' opcodes. Not all CPU's have both Shadow22Stack and Indirect Branch Tracking. Today in the 64-bit kernel, only userspace23shadow stack and kernel IBT are supported.24 25Requirements to use Shadow Stack26================================27 28To use userspace shadow stack you need HW that supports it, a kernel29configured with it and userspace libraries compiled with it.30 31The kernel Kconfig option is X86_USER_SHADOW_STACK. When compiled in, shadow32stacks can be disabled at runtime with the kernel parameter: nousershstk.33 34To build a user shadow stack enabled kernel, Binutils v2.29 or LLVM v6 or later35are required.36 37At run time, /proc/cpuinfo shows CET features if the processor supports38CET. "user_shstk" means that userspace shadow stack is supported on the current39kernel and HW.40 41Application Enabling42====================43 44An application's CET capability is marked in its ELF note and can be verified45from readelf/llvm-readelf output::46 47 readelf -n <application> | grep -a SHSTK48 properties: x86 feature: SHSTK49 50The kernel does not process these applications markers directly. Applications51or loaders must enable CET features using the interface described in section 4.52Typically this would be done in dynamic loader or static runtime objects, as is53the case in GLIBC.54 55Enabling arch_prctl()'s56=======================57 58Elf features should be enabled by the loader using the below arch_prctl's. They59are only supported in 64 bit user applications. These operate on the features60on a per-thread basis. The enablement status is inherited on clone, so if the61feature is enabled on the first thread, it will propagate to all the thread's62in an app.63 64arch_prctl(ARCH_SHSTK_ENABLE, unsigned long feature)65 Enable a single feature specified in 'feature'. Can only operate on66 one feature at a time.67 68arch_prctl(ARCH_SHSTK_DISABLE, unsigned long feature)69 Disable a single feature specified in 'feature'. Can only operate on70 one feature at a time.71 72arch_prctl(ARCH_SHSTK_LOCK, unsigned long features)73 Lock in features at their current enabled or disabled status. 'features'74 is a mask of all features to lock. All bits set are processed, unset bits75 are ignored. The mask is ORed with the existing value. So any feature bits76 set here cannot be enabled or disabled afterwards.77 78arch_prctl(ARCH_SHSTK_UNLOCK, unsigned long features)79 Unlock features. 'features' is a mask of all features to unlock. All80 bits set are processed, unset bits are ignored. Only works via ptrace.81 82arch_prctl(ARCH_SHSTK_STATUS, unsigned long addr)83 Copy the currently enabled features to the address passed in addr. The84 features are described using the bits passed into the others in85 'features'.86 87The return values are as follows. On success, return 0. On error, errno can88be::89 90 -EPERM if any of the passed feature are locked.91 -ENOTSUPP if the feature is not supported by the hardware or92 kernel.93 -EINVAL arguments (non existing feature, etc)94 -EFAULT if could not copy information back to userspace95 96The feature's bits supported are::97 98 ARCH_SHSTK_SHSTK - Shadow stack99 ARCH_SHSTK_WRSS - WRSS100 101Currently shadow stack and WRSS are supported via this interface. WRSS102can only be enabled with shadow stack, and is automatically disabled103if shadow stack is disabled.104 105Proc Status106===========107To check if an application is actually running with shadow stack, the108user can read the /proc/$PID/status. It will report "wrss" or "shstk"109depending on what is enabled. The lines look like this::110 111 x86_Thread_features: shstk wrss112 x86_Thread_features_locked: shstk wrss113 114Implementation of the Shadow Stack115==================================116 117Shadow Stack Size118-----------------119 120A task's shadow stack is allocated from memory to a fixed size of121MIN(RLIMIT_STACK, 4 GB). In other words, the shadow stack is allocated to122the maximum size of the normal stack, but capped to 4 GB. In the case123of the clone3 syscall, there is a stack size passed in and shadow stack124uses this instead of the rlimit.125 126Signal127------128 129The main program and its signal handlers use the same shadow stack. Because130the shadow stack stores only return addresses, a large shadow stack covers131the condition that both the program stack and the signal alternate stack run132out.133 134When a signal happens, the old pre-signal state is pushed on the stack. When135shadow stack is enabled, the shadow stack specific state is pushed onto the136shadow stack. Today this is only the old SSP (shadow stack pointer), pushed137in a special format with bit 63 set. On sigreturn this old SSP token is138verified and restored by the kernel. The kernel will also push the normal139restorer address to the shadow stack to help userspace avoid a shadow stack140violation on the sigreturn path that goes through the restorer.141 142So the shadow stack signal frame format is as follows::143 144 |1...old SSP| - Pointer to old pre-signal ssp in sigframe token format145 (bit 63 set to 1)146 | ...| - Other state may be added in the future147 148 14932 bit ABI signals are not supported in shadow stack processes. Linux prevents15032 bit execution while shadow stack is enabled by the allocating shadow stacks151outside of the 32 bit address space. When execution enters 32 bit mode, either152via far call or returning to userspace, a #GP is generated by the hardware153which, will be delivered to the process as a segfault. When transitioning to154userspace the register's state will be as if the userspace ip being returned to155caused the segfault.156 157Fork158----159 160The shadow stack's vma has VM_SHADOW_STACK flag set; its PTEs are required161to be read-only and dirty. When a shadow stack PTE is not RO and dirty, a162shadow access triggers a page fault with the shadow stack access bit set163in the page fault error code.164 165When a task forks a child, its shadow stack PTEs are copied and both the166parent's and the child's shadow stack PTEs are cleared of the dirty bit.167Upon the next shadow stack access, the resulting shadow stack page fault168is handled by page copy/re-use.169 170When a pthread child is created, the kernel allocates a new shadow stack171for the new thread. New shadow stack creation behaves like mmap() with respect172to ASLR behavior. Similarly, on thread exit the thread's shadow stack is173disabled.174 175Exec176----177 178On exec, shadow stack features are disabled by the kernel. At which point,179userspace can choose to re-enable, or lock them.180