freebsd-src/share/man/man7/mitigations.7

.\" Copyright © 2023 The FreeBSD Foundation
.\"
.\" This documentation was written by Ed Maste <emaste@freebsd.org>, and
.\" Olivier Certner <olce.freebsd@certner.fr> at Kumacom SAS, under
.\" sponsorship of the FreeBSD Foundation.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\"    notice, this list of conditions and the following disclaimer.
.\" 2. 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.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\" ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.Dd June 1, 2024
.Dt MITIGATIONS 7
.Os
.Sh NAME
.Nm mitigations
.Nd FreeBSD Security Vulnerability Mitigations
.Sh SYNOPSIS
In
.Fx ,
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.
.Pp
The following vulnerability mitigations are covered in this document:
.Bl -bullet -compact
.It
Address Space Layout Randomization (ASLR)
.It
Position Independent Executable (PIE)
.It
Write XOR Execute page protection policy
.It
.Dv PROT_MAX
.It
Relocation Read-Only (RELRO)
.It
Bind Now
.It
Stack Overflow Protection
.It
Supervisor Mode Memory Protection
.It
Hardware Vulnerability Mitigation Controls
.It
Capsicum
.El
.Pp
Please note that the effectiveness and availability of these mitigations may
vary depending on the
.Fx
version and system configuration.
.Sh DESCRIPTION
Security vulnerability mitigations are techniques employed in
.Fx
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 secure coding practices.
Mitigations serve as an additional layer of defense, helping to reduce the
likelihood of a successful exploitation of vulnerabilities by making it
more difficult for attackers to achieve their objectives.
.Pp
This manual page describes the security mitigations implemented in
.Fx
to enhance the overall security of the operating system.
Each mitigation is designed to protect against specific types of attacks
and vulnerabilities.
.\"
.Sh SOFTWARE VULNERABILITY MITIGATIONS
.Ss Address Space Layout Randomization (ASLR)
Address Space Layout Randomization (ASLR) is a security mitigation technique
that works by randomizing the memory addresses where system and application
code, data, and libraries are loaded, making it more challenging for attackers
to predict the memory layout and exploit vulnerabilities.
.Pp
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 software vulnerability, such as a buffer overflow.
.Pp
ASLR can be enabled on both a global and per-process basis.
Global control is provided by a separate set of
.Xr sysctl 8
knobs for 32- and 64-bit processes.
It can be or disabled on a per-process basis via
.Xr proccontrol 1 .
Note that an ASLR mode change takes effect upon address space change,
i.e., upon
.Xr execve 2 .
.Pp
Global controls for 32-bit processes:
.Bl -tag -width kern.elf32.aslr.pie_enable
.It Va kern.elf32.aslr.enable
Enable ASLR for 32-bit ELF binaries, other than Position Independent
Executable (PIE) binaries.
.It Va kern.elf32.aslr.pie_enable
Enable ASLR for 32-bit Position Independent Executable (PIE) ELF binaries.
.It Va kern.elf32.aslr.honor_sbrk
Reserve the legacy
.Xr sbrk 2
region for compatibility with older binaries.
.It Va kern.elf32.aslr.stack
Randomize the stack location for 32-bit ELF binaries.
.El
.Pp
Global controls for 64-bit processes:
.Bl -tag -width kern.elf64.aslr.pie_enable
.It Va kern.elf64.aslr.enable
Enable ASLR for 64-bit ELF binaries, other than Position Independent
Executable (PIE) binaries.
.It Va kern.elf64.aslr.pie_enable
Enable ASLR for 64-bit Position Independent Executable (PIE) ELF binaries.
.It Va kern.elf64.aslr.honor_sbrk
Reserve the legacy
.Xr sbrk 2
region for compatibility with older binaries.
.It Va kern.elf64.aslr.stack
Randomize the stack location for 64-bit ELF binaries.
.El
.Pp
To execute a command with ASLR enabled or disabled:
.Pp
proccontrol
.Fl m Ar aslr
.Op Fl s Ar enable | disable
.Ar command
.\"
.Ss 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
.Xr rtld
run-time linker.
With ASLR they are loaded at a random address on each execution.
.\"
.Ss 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 permissions.
.Pp
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 prevented in areas of memory that are
designated as writable, and writing or modification of memory is restricted in
areas marked for execution.
Applications that perform Just In Time (JIT) compilation need to be adapted
to be compatible with W^X.
.Pp
There are separate
.Xr sysctl 8
knobs to control W^X policy enforcement for 32- and 64-bit processes.
The W^X policy is enabled by setting the appropriate
.Dv allow_wx
sysctl to 0.
.Bl -tag -width kern.elf64.allow_wx
.It Va kern.elf32.allow_wx
Allow 32-bit processes to map pages simultaneously writable and executable.
.It Va kern.elf64.allow_wx
Allow 64-bit processes to map pages simultaneously writable and executable.
.El
.\"
.Ss PROT_MAX
.Dv PROT_MAX
is a FreeBSD-specific extension to
.Xr mmap 2 .
.Dv PROT_MAX
provides the ability to set the maximum protection of a region allocated by
.Xr mmap
and later altered by
.Xr mprotect .
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
.Xr mprotect
call, but may not be made executable.
.\"
.Ss Relocation Read-Only (RELRO)
Relocation Read-Only (RELRO) is a mitigation tool that makes certain portions
of a program's address space that contain ELF metadata read-only, after
relocation processing by
.Xr rtld 1 .
.Pp
When enabled in isolation the RELRO option provides
.Em 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.
.Pp
RELRO is enabled by default.
The
.Xr src.conf 5
build-time option
.Va WITHOUT_RELRO
may be used to disable it.
.Ss BIND_NOW
The
.Va WITH_BIND_NOW
.Xr src.conf 5
build-time option causes binaries to be built with the
.Dv DF_BIND_NOW
flag set.
The run-time loader
.Xr rtld 1
will then perform all relocation processing when the process starts, instead of
on demand (on the first access to each symbol).
.Pp
When enabled in combination with
.Dv RELRO
(which is enabled by default) this provides
.Em 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.
.\"
.Ss Stack Overflow Protection
.Fx
supports stack overflow protection using the Stack Smashing Protector
.Pq SSP
compiler feature.
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.
In the kernel, a single randomized canary is used globally except on aarch64,
which has a
.Dv PERTHREAD_SSP
.Xr 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 result of the corruption.
SSP may be enabled or disabled when building
.Fx
base with the
.Xr src.conf 5
SSP knob.
.Pp
When
.Va WITH_SSP
is enabled, which is the default, world is built with the
.Fl fstack-protector-strong
compiler option.
The kernel is built with the
.Fl fstack-protector
option.
.Pp
In addition to SSP, a
.Dq FORTIFY_SOURCE
implementation is supported up to level 2 by defining
.Va _FORTIFY_SOURCE
to
.Dv 1
or
.Dv 2
before including any
.Fx
headers.
.Fx
world builds can set
.Va FORTIFY_SOURCE
to provide a default value for
.Va _FORTIFY_SOURCE .
When enabled,
.Dq FORTIFY_SOURCE
enables extra bounds checking in various functions that accept buffers to be
written into.
These functions currently have extra bounds checking support:
.Bl -column -offset indent "snprintf" "memmove" "strncpy" "vsnprintf" "readlink"
.It bcopy    Ta bzero    Ta fgets     Ta getcwd    Ta gets
.It memcpy   Ta memmove  Ta memset    Ta read      Ta readlink
.It snprintf Ta sprintf  Ta stpcpy    Ta stpncpy   Ta strcat
.It strcpy   Ta strncat  Ta strncpy   Ta vsnprintf Ta vsprintf
.El
.Pp
.Dq FORTIFY_SOURCE
requires compiler support from
.Xr clang 1
or
.Xr gcc 1 ,
which provide the
.Xr __builtin_object_size 3
function that is used to determine the bounds of an object.
This feature works best at optimization levels
.Fl O1
and above, as some object sizes may be less obvious without some data that the
compiler would collect in an optimization pass.
.Pp
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 program data.
.\"
.Ss Supervisor mode memory protection
Certain processors include features that prevent unintended access to memory
pages accessible to userspace (non-privileged) code, while in a privileged
mode.
One feature prevents execution, intended to mitigate exploitation of kernel
vulnerabilities from userland.
Another feature prevents unintended reads from or writes to user space memory
from the kernel.
This also provides effective protection against NULL pointer dereferences from
kernel.
.Bl -column -offset indent "Architecture" "Feature" "Access Type Prevented"
.It Sy Architecture Ta Sy Feature Ta Sy Access Type Prevented
.It amd64       Ta SMAP  Ta Read / Write
.It amd64       Ta SMEP  Ta Execute
.It arm64       Ta PAN   Ta Read / Write
.It arm64       Ta PXN   Ta Execute
.It riscv       Ta SUM   Ta Read / Write
.It riscv       Ta -     Ta Execute
.El
.Pp
These features are automatically used by the kernel.
There is no user-facing configuration.
.Ss Hardware vulnerability controls
See
.Xr security 7
for more information.
.\"
.Ss Capsicum
Capsicum is a lightweight OS capability and sandbox framework.
See
.Xr capsicum 4
for more information.
.Pp
.Sh HARDWARE VULNERABILITY MITIGATIONS
Recent years have seen an unending stream of new hardware vulnerabilities,
notably CPU ones generally caused by detectable microarchitectural side-effects
of speculative execution which leak private data from some other thread or
process or sometimes even internal CPU state that is normally inaccessible.
Hardware vendors usually address these vulnerabilities as they are discovered by
releasing microcode updates, which may then be bundled into platform firmware
updates
.Pq historically called BIOS updates for PCs .
.Pp
The best defense overall against hardware vulnerabilities is to timely apply
these updates when available 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 capabilities provided by a
microcode update.
.Pp
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
.Pq which sometimes requires firmware updates ,
or those that are extremely detrimental to performance in proportion to the
protection they actually provide.
OS-level mitigations generally can have noticeable performance impacts on
specific workloads.
If your threat model allows it, you may want to try disabling some of them in
order to possibly get better performance.
Conversely, minimizing 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.
.Ss Zenbleed
The
.Dq Zenbleed
vulnerability exclusively affects AMD processors based on the Zen2
microarchitecture.
In contrast with, e.g., Meltdown and the different variants of Spectre, which
leak data by leaving microarchitectural traces, Zenbleed is a genuine hardware
bug affecting the CPU's architectural 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
.Po disabling Simultaneous Muti-Threading
.Pq SMT
is thus not a sufficient protection
.Pc .
.Pp
According to the vulnerability's discoverer, all Zen2-based processors are
affected
.Po see
.Lk https://lock.cmpxchg8b.com/zenbleed.html
.Pc .
As of August 2023, AMD has not publicly listed any corresponding errata but has
issued a security bulletin
.Pq AMD-SB-7008
entitled
.Dq Cross-Process Information Leak
indicating that platform firmware fixing the vulnerability will be distributed
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.
.Pp
.Fx
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
microcode 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:
.Bl -tag -width indent
.It Va machdep.mitigations.zenbleed.enable
A read-write integer tunable and sysctl indicating whether the mitigation should
be forcibly disabled (0), enabled (1) or if it is left to
.Fx
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 processors
to ease applying a common configuration to heterogeneous machines.
.It Va machdep.mitigations.zenbleed.state
A read-only string indicating the current mitigation state.
It can be either
.Dq Not applicable ,
if the processor is not Zen2-based,
.Dq Mitigation enabled
or
.Dq Mitigation disabled .
This state is automatically updated each time the sysctl
.Va machdep.mitigations.zenbleed.enable
is written to.
Note that it can become inaccurate if the chicken bit is set or cleared
directly via
.Xr cpuctl 4
.Po which includes the
.Xr cpucontrol 8
utility
.Pc .
.El
.Pp
The performance impact and threat models related to these mitigations
should be considered when configuring and deploying them in a
.Fx
system.
.Pp
.Sh SEE ALSO
.Xr elfctl 1 ,
.Xr proccontrol 1 ,
.Xr rtld 1 ,
.Xr mmap 2 ,
.Xr src.conf 5 ,
.Xr sysctl.conf 5 ,
.Xr security 7 ,
.Xr cpucontrol 8 ,
.Xr sysctl 8