FreeBSD Manual Pages
MITIGATIONS(7) Miscellaneous Information Manual MITIGATIONS(7) NAME mitigations -- FreeBSD Security Vulnerability Mitigations SYNOPSIS In FreeBSD, various security mitigations are employed to limit the impact of vulnerabilities and protect the system from malicious attacks. Some of these mitigations have run-time controls to enable them on a global or per- process basis, some are optionally enabled or disabled at compile time, and some are inherent to the implementation and have no controls. The following vulnerability mitigations are covered in this document: * Address Space Layout Randomization (ASLR) * Position Independent Executable (PIE) * Write XOR Execute page protection policy * PROT_MAX * Relocation Read-Only (RELRO) * Bind Now * Stack Overflow Protection * Supervisor Mode Memory Protection * Capsicum * Firmware and Microcode * Architectural Vulnerability Mitigations Please note that the effectiveness and availability of these mitigations may vary depending on the FreeBSD version and system configuration. DESCRIPTION Security vulnerability mitigations are techniques employed in FreeBSD to limit the potential impact of security vulnerabilities in software and hardware. It is essential to understand that mitigations do not directly address the underlying security issues. They are not a substitute for se- cure coding practices. Mitigations serve as an additional layer of de- fense, helping to reduce the likelihood of a successful exploitation of vulnerabilities by making it more difficult for attackers to achieve their objectives. This manual page describes the security mitigations implemented in FreeBSD to enhance the overall security of the operating system. Each mitigation is designed to protect against specific types of attacks and vulnerabili- ties. SOFTWARE VULNERABILITY MITIGATIONS Address Space Layout Randomization (ASLR) Address Space Layout Randomization (ASLR) is a security mitigation tech- nique that works by randomizing the memory addresses where system and ap- plication code, data, and libraries are loaded, making it more challenging for attackers to predict the memory layout and exploit vulnerabilities. ASLR introduces randomness into the memory layout during process execution, reducing the predictability of memory addresses. ASLR is intended to make exploitation more difficult in the event that an attacker discovers a soft- ware vulnerability, such as a buffer overflow. ASLR can be enabled on both a global and per-process basis. Global control is provided by a separate set of sysctl(8) knobs for 32- and 64-bit processes. It can be or disabled on a per-process basis via proccon- trol(1). Note that an ASLR mode change takes effect upon address space change, i.e., upon execve(2). Global controls for 32-bit processes: kern.elf32.aslr.enable Enable ASLR for 32-bit ELF binaries, other than Position Independent Executable (PIE) binaries. kern.elf32.aslr.pie_enable Enable ASLR for 32-bit Position Independent Ex- ecutable (PIE) ELF binaries. kern.elf32.aslr.honor_sbrk Reserve the legacy sbrk(2) region for compati- bility with older binaries. kern.elf32.aslr.stack Randomize the stack location for 32-bit ELF bi- naries. Global controls for 64-bit processes: kern.elf64.aslr.enable Enable ASLR for 64-bit ELF binaries, other than Position Independent Executable (PIE) binaries. kern.elf64.aslr.pie_enable Enable ASLR for 64-bit Position Independent Ex- ecutable (PIE) ELF binaries. kern.elf64.aslr.honor_sbrk Reserve the legacy sbrk(2) region for compati- bility with older binaries. kern.elf64.aslr.stack Randomize the stack location for 64-bit ELF bi- naries. To execute a command with ASLR enabled or disabled: proccontrol -m aslr [-s enable | disable] command Position Independent Executable (PIE) PIE binaries are executable files that do not have a fixed load address. They can be loaded at an arbitrary memory address by the rtld(1) run-time linker. With ASLR they are loaded at a random address on each execution. Write XOR Execute page protection policy Write XOR Execute (W^X) is a vulnerability mitigation strategy that strengthens the security of the system by controlling memory access permis- sions. Under the W^X mitigation, memory pages may be writable (W) or executable (E), but not both at the same time. This means that code execution is pre- vented in areas of memory that are designated as writable, and writing or modification of memory is restricted in areas marked for execution. Appli- cations that perform Just In Time (JIT) compilation need to be adapted to be compatible with W^X. There are separate sysctl(8) knobs to control W^X policy enforcement for 32- and 64-bit processes. The W^X policy is enabled by setting the appro- priate allow_wx sysctl to 0. kern.elf32.allow_wx Allow 32-bit processes to map pages simultaneously writable and executable. kern.elf64.allow_wx Allow 64-bit processes to map pages simultaneously writable and executable. PROT_MAX PROT_MAX is a FreeBSD-specific extension to mmap(2). PROT_MAX provides the ability to set the maximum protection of a region allocated by mmap(2) and later altered by mprotect(2). For example, memory allocated originally with an mmap prot argument of PROT_MAX(PROT_READ | PROT_WRITE) | PROT_READ may be made writable by a future mprotect(2) call, but may not be made exe- cutable. Relocation Read-Only (RELRO) Relocation Read-Only (RELRO) is a mitigation tool that makes certain por- tions of a program's address space that contain ELF metadata read-only, af- ter relocation processing by rtld(1). When enabled in isolation the RELRO option provides partial RELRO support. In this case the Procedure Linkage Table (PLT)-related part of the Global Offset Table (GOT) (in the section typically named .got.plt) remains writable. RELRO is enabled by default. The src.conf(5) build-time option WITHOUT_RELRO may be used to disable it. BIND_NOW The WITH_BIND_NOW src.conf(5) build-time option causes binaries to be built with the DF_BIND_NOW flag set. The run-time loader rtld(1) will then per- form all relocation processing when the process starts, instead of on de- mand (on the first access to each symbol). When enabled in combination with RELRO (which is enabled by default) this provides full RELRO. The entire GOT (.got and .got.plt) are made read-only at program startup, preventing attacks on the relocation table. Note that this results in a nonstandard Application Binary Interface (ABI), and it is possible that some applications may not function correctly. Stack Overflow Protection FreeBSD supports stack overflow protection using the Stack Smashing Protec- tor (SSP) compiler feature. Stack clash protection is also enabled, if supported by the compiler for the given architecture. In userland, SSP adds a per-process randomized canary at the end of every stack frame which is checked for corruption upon return from the function, and stack probing in PAGE_SIZE chunks. In the kernel, a single randomized canary is used globally except on aarch64, which has a PERTHREAD_SSP config(8) option to enable per-thread randomized canaries. If stack corruption is detected, then the process aborts to avoid potentially malicious execution as a re- sult of the corruption. SSP may be enabled or disabled when building FreeBSD base with the src.conf(5) SSP knob. When WITH_SSP is enabled, which is the default, world is built with the -fstack-protector-strong and -fstack-clash-protection compiler options. The kernel is built with the -fstack-protector option. In addition to SSP, a "FORTIFY_SOURCE" implementation is supported up to level 2 by defining _FORTIFY_SOURCE to 1 or 2 before including any FreeBSD headers. FreeBSD world builds can set FORTIFY_SOURCE in the environment or /etc/src-env.conf to provide a default value for _FORTIFY_SOURCE. When en- abled, "FORTIFY_SOURCE" enables extra bounds checking in various functions that accept buffers to be written into. These functions currently have ex- tra bounds checking support: bcopy() bzero() fgets() getcwd() gets() memcpy() memmove() memset() read() readlink() snprintf() sprintf() stpcpy() stpncpy() strcat() strcpy() strncat() strncpy() vsnprintf() vsprintf() "FORTIFY_SOURCE" requires compiler support from clang(1) or gcc(1), which provide the __builtin_object_size(3) function that is used to determine the bounds of an object. This feature works best at optimization levels -O1 and above, as some object sizes may be less obvious without some data that the compiler would collect in an optimization pass. Similar to SSP, violating the bounds of an object will cause the program to abort in an effort to avoid malicious execution. This effectively provides finer-grained protection than SSP for some class of function and system calls, along with some protection for buffers allocated as part of the pro- gram data. Supervisor mode memory protection Certain processors include features that prevent unintended access to mem- ory pages accessible to userspace (non-privileged) code, while in a privi- leged mode. One feature prevents execution, intended to mitigate exploita- tion of kernel vulnerabilities from userland. Another feature prevents un- intended reads from or writes to user space memory from the kernel. This also provides effective protection against NULL pointer dereferences from kernel. Architecture Feature Access Type Prevented amd64 SMAP Read / Write amd64 SMEP Execute arm64 PAN Read / Write arm64 PXN Execute riscv SUM Read / Write riscv - Execute These features are automatically used by the kernel. There is no user-fac- ing configuration. Capsicum Capsicum is a lightweight OS capability and sandbox framework. See cap- sicum(4) for more information. HARDWARE VULNERABILITY MITIGATIONS Firmware and Microcode Recent years have seen an unending stream of new hardware vulnerabilities, notably CPU ones generally caused by detectable microarchitectural side-ef- fects of speculative execution which leak private data from some other thread or process or sometimes even internal CPU state that is normally in- accessible. Hardware vendors usually address these vulnerabilities as they are discovered by releasing microcode updates, which may then be bundled into platform firmware updates (historically called BIOS updates for PCs) or packages to be updated by the operating system at boot time. Platform firmware updates, if available from the manufacturer, are the best defense as they provide coverage during early boot. Install them with sysutils/flashrom from the FreeBSD Ports Collection. If platform firmware updates are no longer available, packaged microcode is available for installation at sysutils/cpu-microcode and can be loaded at runtime using loader.conf(5), see the package message for more details. The best defense overall against hardware vulnerabilities is to timely ap- ply these updates when available, as early as possible in the boot process, and to disable the affected hardware's problematic functionalities when possible (e.g., CPU Simultaneous Multi-Threading). Software mitigations are only partial substitutes for these, but they can be helpful on out-of- support hardware or as complements for just-discovered vulnerabilities not yet addressed by vendors. Some software mitigations depend on hardware ca- pabilities provided by a microcode update. Architectural Vulnerability Mitigations FreeBSD's usual policy is to apply by default all OS-level mitigations that do not require recompilation, except those the particular hardware it is running on is known not to be vulnerable to (which sometimes requires firmware updates), or those that are extremely detrimental to performance in proportion to the protection they actually provide. OS-level mitiga- tions generally can have noticeable performance impacts on specific work- loads. If your threat model allows it, you may want to try disabling some of them in order to possibly get better performance. Conversely, minimiz- ing the risks may require you to explicitly enable the most expensive ones. The description of each vulnerability/mitigation indicates whether it is enabled or disabled by default and under which conditions. It also lists the knobs to tweak to force a particular status. Zenbleed The "Zenbleed" vulnerability exclusively affects AMD processors based on the Zen2 microarchitecture. In contrast with, e.g., Meltdown and the dif- ferent variants of Spectre, which leak data by leaving microarchitectural traces, Zenbleed is a genuine hardware bug affecting the CPU's architec- tural state. With particular sequences of instructions whose last ones are mispredicted by speculative execution, it is possible to make appear in an XMM register data previously put in some XMM register by some preceding or concurrent task executing on the same physical core (disabling Simultaneous Muti-Threading (SMT) is thus not a sufficient protection). According to the vulnerability's discoverer, all Zen2-based processors are affected (see https://lock.cmpxchg8b.com/zenbleed.html). As of August 2023, AMD has not publicly listed any corresponding errata but has issued a security bulletin (AMD-SB-7008) entitled "Cross-Process Information Leak" indicating that platform firmware fixing the vulnerability will be distrib- uted to manufacturers no sooner than the end of 2023, except for Rome processors for which it is already available. No standalone CPU microcodes have been announced so far. The only readily-applicable fix mentioned by the discoverer is to set a bit of an undocumented MSR, which reportedly completely stops XMM register leaks. FreeBSD currently sets this bit by default on all Zen2 processors. In the future, it might set it by default only on those Zen2 processors whose mi- crocode has not been updated to revisions fixing the vulnerability, once such microcode updates have been actually released and community-tested. To this mitigation are associated the following knobs: machdep.mitigations.zenbleed.enable A read-write integer tunable and sysctl indicating whether the mit- igation should be forcibly disabled (0), enabled (1) or if it is left to FreeBSD to selectively apply it (2). Any other integer value is silently converted to and treated as value 2. Note that this setting is silently ignored when running on non-Zen2 proces- sors to ease applying a common configuration to heterogeneous ma- chines. machdep.mitigations.zenbleed.state A read-only string indicating the current mitigation state. It can be either "Not applicable", if the processor is not Zen2-based, "Mitigation enabled" or "Mitigation disabled". This state is auto- matically updated each time the sysctl machdep.mitigations.zenbleed.enable is written to. Note that it can become inaccurate if the chicken bit is set or cleared directly via cpuctl(4) (which includes the cpucontrol(8) utility). The performance impact and threat models related to these mitigations should be considered when configuring and deploying them in a FreeBSD sys- tem. Additional mitigation knobs are listed in the "KNOBS AND TWEAKS" section of security(7). SEE ALSO elfctl(1), proccontrol(1), rtld(1), mmap(2), src.conf(5), sysctl.conf(5), security(7), cpucontrol(8), sysctl(8) FreeBSD ports 15.quarterly January 29, 2025 MITIGATIONS(7)
NAME | SYNOPSIS | DESCRIPTION | SOFTWARE VULNERABILITY MITIGATIONS | HARDWARE VULNERABILITY MITIGATIONS | SEE ALSO
Want to link to this manual page? Use this URL:
<https://man.freebsd.org/cgi/man.cgi?query=mitigations&sektion=7&manpath=FreeBSD+15.1-RELEASE+and+Ports.quarterly>