xref: /freebsd/share/man/man7/mitigations.7 (revision 59f5f100b774de8824fb2fc1a8a11a93bbc2dafd)
1.\"-
2.\" SPDX-License-Identifer: BSD-2-Clause
3.\"
4.\" Copyright © 2023 The FreeBSD Foundation
5.\"
6.\" This documentation was written by Ed Maste <emaste@freebsd.org>, and
7.\" Olivier Certner <olce.freebsd@certner.fr> at Kumacom SAS, under
8.\" sponsorship of the FreeBSD Foundation.
9.\"
10.\" Redistribution and use in source and binary forms, with or without
11.\" modification, are permitted provided that the following conditions
12.\" are met:
13.\" 1. Redistributions of source code must retain the above copyright
14.\"    notice, this list of conditions and the following disclaimer.
15.\" 2. Redistributions in binary form must reproduce the above copyright
16.\"    notice, this list of conditions and the following disclaimer in the
17.\"    documentation and/or other materials provided with the distribution.
18.\"
19.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
20.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22.\" ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
23.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29.\" SUCH DAMAGE.
30.\"
31.Dd July 23, 2024
32.Dt MITIGATIONS 7
33.Os
34.Sh NAME
35.Nm mitigations
36.Nd FreeBSD Security Vulnerability Mitigations
37.Sh SYNOPSIS
38In
39.Fx ,
40various security mitigations are employed to limit the impact of
41vulnerabilities and protect the system from malicious attacks.
42Some of these mitigations have run-time controls to enable them on a global
43or per-process basis, some are optionally enabled or disabled at compile time,
44and some are inherent to the implementation and have no controls.
45.Pp
46The following vulnerability mitigations are covered in this document:
47.Pp
48.Bl -bullet -compact
49.It
50Address Space Layout Randomization (ASLR)
51.It
52Position Independent Executable (PIE)
53.It
54Write XOR Execute page protection policy
55.It
56.Dv PROT_MAX
57.It
58Relocation Read-Only (RELRO)
59.It
60Bind Now
61.It
62Stack Overflow Protection
63.It
64Supervisor Mode Memory Protection
65.It
66Capsicum
67.It
68Firmware and Microcode
69.It
70Architectural Vulnerability Mitigations
71.El
72.Pp
73Please note that the effectiveness and availability of these mitigations may
74vary depending on the
75.Fx
76version and system configuration.
77.Sh DESCRIPTION
78Security vulnerability mitigations are techniques employed in
79.Fx
80to limit the potential impact of security vulnerabilities in software and
81hardware.
82It is essential to understand that mitigations do not directly address the
83underlying security issues.
84They are not a substitute for secure coding practices.
85Mitigations serve as an additional layer of defense, helping to reduce the
86likelihood of a successful exploitation of vulnerabilities by making it
87more difficult for attackers to achieve their objectives.
88.Pp
89This manual page describes the security mitigations implemented in
90.Fx
91to enhance the overall security of the operating system.
92Each mitigation is designed to protect against specific types of attacks
93and vulnerabilities.
94.\"
95.Sh SOFTWARE VULNERABILITY MITIGATIONS
96.Ss Address Space Layout Randomization (ASLR)
97Address Space Layout Randomization (ASLR) is a security mitigation technique
98that works by randomizing the memory addresses where system and application
99code, data, and libraries are loaded, making it more challenging for attackers
100to predict the memory layout and exploit vulnerabilities.
101.Pp
102ASLR introduces randomness into the memory layout during process execution,
103reducing the predictability of memory addresses.
104ASLR is intended to make exploitation more difficult in the event that an
105attacker discovers a software vulnerability, such as a buffer overflow.
106.Pp
107ASLR can be enabled on both a global and per-process basis.
108Global control is provided by a separate set of
109.Xr sysctl 8
110knobs for 32- and 64-bit processes.
111It can be or disabled on a per-process basis via
112.Xr proccontrol 1 .
113Note that an ASLR mode change takes effect upon address space change,
114i.e., upon
115.Xr execve 2 .
116.Pp
117Global controls for 32-bit processes:
118.Bl -tag -width kern.elf32.aslr.pie_enable
119.It Va kern.elf32.aslr.enable
120Enable ASLR for 32-bit ELF binaries, other than Position Independent
121Executable (PIE) binaries.
122.It Va kern.elf32.aslr.pie_enable
123Enable ASLR for 32-bit Position Independent Executable (PIE) ELF binaries.
124.It Va kern.elf32.aslr.honor_sbrk
125Reserve the legacy
126.Xr sbrk 2
127region for compatibility with older binaries.
128.It Va kern.elf32.aslr.stack
129Randomize the stack location for 32-bit ELF binaries.
130.El
131.Pp
132Global controls for 64-bit processes:
133.Bl -tag -width kern.elf64.aslr.pie_enable
134.It Va kern.elf64.aslr.enable
135Enable ASLR for 64-bit ELF binaries, other than Position Independent
136Executable (PIE) binaries.
137.It Va kern.elf64.aslr.pie_enable
138Enable ASLR for 64-bit Position Independent Executable (PIE) ELF binaries.
139.It Va kern.elf64.aslr.honor_sbrk
140Reserve the legacy
141.Xr sbrk 2
142region for compatibility with older binaries.
143.It Va kern.elf64.aslr.stack
144Randomize the stack location for 64-bit ELF binaries.
145.El
146.Pp
147To execute a command with ASLR enabled or disabled:
148.Pp
149proccontrol
150.Fl m Ar aslr
151.Op Fl s Ar enable | disable
152.Ar command
153.\"
154.Ss Position Independent Executable (PIE)
155PIE binaries are executable files that do not have a fixed load address.
156They can be loaded at an arbitrary memory address by the
157.Xr rtld
158run-time linker.
159With ASLR they are loaded at a random address on each execution.
160.\"
161.Ss Write XOR Execute page protection policy
162Write XOR Execute (W^X) is a vulnerability mitigation strategy that strengthens
163the security of the system by controlling memory access permissions.
164.Pp
165Under the W^X mitigation, memory pages may be writable (W) or executable (E),
166but not both at the same time.
167This means that code execution is prevented in areas of memory that are
168designated as writable, and writing or modification of memory is restricted in
169areas marked for execution.
170Applications that perform Just In Time (JIT) compilation need to be adapted
171to be compatible with W^X.
172.Pp
173There are separate
174.Xr sysctl 8
175knobs to control W^X policy enforcement for 32- and 64-bit processes.
176The W^X policy is enabled by setting the appropriate
177.Dv allow_wx
178sysctl to 0.
179.Bl -tag -width kern.elf64.allow_wx
180.It Va kern.elf32.allow_wx
181Allow 32-bit processes to map pages simultaneously writable and executable.
182.It Va kern.elf64.allow_wx
183Allow 64-bit processes to map pages simultaneously writable and executable.
184.El
185.\"
186.Ss PROT_MAX
187.Dv PROT_MAX
188is a FreeBSD-specific extension to
189.Xr mmap 2 .
190.Dv PROT_MAX
191provides the ability to set the maximum protection of a region allocated by
192.Xr mmap
193and later altered by
194.Xr mprotect .
195For example, memory allocated originally with an mmap prot argument of
196PROT_MAX(PROT_READ | PROT_WRITE) | PROT_READ
197may be made writable by a future
198.Xr mprotect
199call, but may not be made executable.
200.\"
201.Ss Relocation Read-Only (RELRO)
202Relocation Read-Only (RELRO) is a mitigation tool that makes certain portions
203of a program's address space that contain ELF metadata read-only, after
204relocation processing by
205.Xr rtld 1 .
206.Pp
207When enabled in isolation the RELRO option provides
208.Em partial RELRO
209support.
210In this case the Procedure Linkage Table (PLT)-related part of the
211Global Offset Table (GOT) (in the section typically named .got.plt) remains
212writable.
213.Pp
214RELRO is enabled by default.
215The
216.Xr src.conf 5
217build-time option
218.Va WITHOUT_RELRO
219may be used to disable it.
220.Ss BIND_NOW
221The
222.Va WITH_BIND_NOW
223.Xr src.conf 5
224build-time option causes binaries to be built with the
225.Dv DF_BIND_NOW
226flag set.
227The run-time loader
228.Xr rtld 1
229will then perform all relocation processing when the process starts, instead of
230on demand (on the first access to each symbol).
231.Pp
232When enabled in combination with
233.Dv RELRO
234(which is enabled by default) this provides
235.Em full RELRO .
236The entire GOT (.got and .got.plt) are made read-only at program startup,
237preventing attacks on the relocation table.
238Note that this results in a nonstandard Application Binary Interface (ABI),
239and it is possible that some applications may not function correctly.
240.\"
241.Ss Stack Overflow Protection
242.Fx
243supports stack overflow protection using the Stack Smashing Protector
244.Pq SSP
245compiler feature.
246In userland, SSP adds a per-process randomized canary at the end of every stack
247frame which is checked for corruption upon return from the function.
248In the kernel, a single randomized canary is used globally except on aarch64,
249which has a
250.Dv PERTHREAD_SSP
251.Xr config 8
252option to enable per-thread randomized canaries.
253If stack corruption is detected, then the process aborts to avoid potentially
254malicious execution as a result of the corruption.
255SSP may be enabled or disabled when building
256.Fx
257base with the
258.Xr src.conf 5
259SSP knob.
260.Pp
261When
262.Va WITH_SSP
263is enabled, which is the default, world is built with the
264.Fl fstack-protector-strong
265compiler option.
266The kernel is built with the
267.Fl fstack-protector
268option.
269.Pp
270In addition to SSP, a
271.Dq FORTIFY_SOURCE
272implementation is supported up to level 2 by defining
273.Va _FORTIFY_SOURCE
274to
275.Dv 1
276or
277.Dv 2
278before including any
279.Fx
280headers.
281.Fx
282world builds can set
283.Va FORTIFY_SOURCE
284in the environment or
285.Pa /etc/src-env.conf
286to provide a default value for
287.Va _FORTIFY_SOURCE .
288When enabled,
289.Dq FORTIFY_SOURCE
290enables extra bounds checking in various functions that accept buffers to be
291written into.
292These functions currently have extra bounds checking support:
293.Bl -column -offset indent "snprintf" "memmove" "strncpy" "vsnprintf" "readlink"
294.It bcopy    Ta bzero    Ta fgets     Ta getcwd    Ta gets
295.It memcpy   Ta memmove  Ta memset    Ta read      Ta readlink
296.It snprintf Ta sprintf  Ta stpcpy    Ta stpncpy   Ta strcat
297.It strcpy   Ta strncat  Ta strncpy   Ta vsnprintf Ta vsprintf
298.El
299.Pp
300.Dq FORTIFY_SOURCE
301requires compiler support from
302.Xr clang 1
303or
304.Xr gcc 1 ,
305which provide the
306.Xr __builtin_object_size 3
307function that is used to determine the bounds of an object.
308This feature works best at optimization levels
309.Fl O1
310and above, as some object sizes may be less obvious without some data that the
311compiler would collect in an optimization pass.
312.Pp
313Similar to SSP, violating the bounds of an object will cause the program to
314abort in an effort to avoid malicious execution.
315This effectively provides finer-grained protection than SSP for some class of
316function and system calls, along with some protection for buffers allocated as
317part of the program data.
318.\"
319.Ss Supervisor mode memory protection
320Certain processors include features that prevent unintended access to memory
321pages accessible to userspace (non-privileged) code, while in a privileged
322mode.
323One feature prevents execution, intended to mitigate exploitation of kernel
324vulnerabilities from userland.
325Another feature prevents unintended reads from or writes to user space memory
326from the kernel.
327This also provides effective protection against NULL pointer dereferences from
328kernel.
329.Bl -column -offset indent "Architecture" "Feature" "Access Type Prevented"
330.It Sy Architecture Ta Sy Feature Ta Sy Access Type Prevented
331.It amd64       Ta SMAP  Ta Read / Write
332.It amd64       Ta SMEP  Ta Execute
333.It arm64       Ta PAN   Ta Read / Write
334.It arm64       Ta PXN   Ta Execute
335.It riscv       Ta SUM   Ta Read / Write
336.It riscv       Ta -     Ta Execute
337.El
338.Pp
339These features are automatically used by the kernel.
340There is no user-facing configuration.
341.\"
342.Ss Capsicum
343Capsicum is a lightweight OS capability and sandbox framework.
344See
345.Xr capsicum 4
346for more information.
347.Sh HARDWARE VULNERABILITY MITIGATIONS
348.Ss Firmware and Microcode
349Recent years have seen an unending stream of new hardware vulnerabilities,
350notably CPU ones generally caused by detectable microarchitectural side-effects
351of speculative execution which leak private data from some other thread or
352process or sometimes even internal CPU state that is normally inaccessible.
353Hardware vendors usually address these vulnerabilities as they are discovered by
354releasing microcode updates, which may then be bundled into platform firmware
355updates
356.Pq historically called BIOS updates for PCs
357or packages to be updated by the operating system at boot time.
358.Pp
359Platform firmware updates, if available from the manufacturer,
360are the best defense as they provide coverage during early boot.
361Install them with
362.Pa sysutils/flashrom
363from the
364.Fx
365Ports Collection.
366.Pp
367If platform firmware updates are no longer available,
368packaged microcode is available for installation at
369.Pa sysutils/cpu-microcode
370and can be loaded at runtime using
371.Xr loader.conf 5 ,
372see the package message for more details.
373.Pp
374The best defense overall against hardware vulnerabilities is to timely apply
375these updates when available, as early as possible in the boot process,
376and to disable the affected hardware's problematic functionalities when possible
377(e.g., CPU Simultaneous Multi-Threading).
378Software mitigations are only partial substitutes for these, but they can be
379helpful on out-of-support hardware or as complements for just-discovered
380vulnerabilities not yet addressed by vendors.
381Some software mitigations depend on hardware capabilities provided by a
382microcode update.
383.Ss Architectural Vulnerability Mitigations
384.Fx Ap s
385usual policy is to apply by default all OS-level mitigations that do
386not require recompilation, except those the particular hardware it is running on
387is known not to be vulnerable to
388.Pq which sometimes requires firmware updates ,
389or those that are extremely detrimental to performance in proportion to the
390protection they actually provide.
391OS-level mitigations generally can have noticeable performance impacts on
392specific workloads.
393If your threat model allows it, you may want to try disabling some of them in
394order to possibly get better performance.
395Conversely, minimizing the risks may require you to explicitly enable the most
396expensive ones.
397The description of each vulnerability/mitigation indicates whether it is enabled
398or disabled by default and under which conditions.
399It also lists the knobs to tweak to force a particular status.
400.Ss Zenbleed
401The
402.Dq Zenbleed
403vulnerability exclusively affects AMD processors based on the Zen2
404microarchitecture.
405In contrast with, e.g., Meltdown and the different variants of Spectre, which
406leak data by leaving microarchitectural traces, Zenbleed is a genuine hardware
407bug affecting the CPU's architectural state.
408With particular sequences of instructions whose last ones are mispredicted by
409speculative execution, it is possible to make appear in an XMM register data
410previously put in some XMM register by some preceding or concurrent task
411executing on the same physical core
412.Po disabling Simultaneous Muti-Threading
413.Pq SMT
414is thus not a sufficient protection
415.Pc .
416.Pp
417According to the vulnerability's discoverer, all Zen2-based processors are
418affected
419.Po see
420.Lk https://lock.cmpxchg8b.com/zenbleed.html
421.Pc .
422As of August 2023, AMD has not publicly listed any corresponding errata but has
423issued a security bulletin
424.Pq AMD-SB-7008
425entitled
426.Dq Cross-Process Information Leak
427indicating that platform firmware fixing the vulnerability will be distributed
428to manufacturers no sooner than the end of 2023, except for Rome processors for
429which it is already available.
430No standalone CPU microcodes have been announced so far.
431The only readily-applicable fix mentioned by the discoverer is to set a bit of
432an undocumented MSR, which reportedly completely stops XMM register leaks.
433.Pp
434.Fx
435currently sets this bit by default on all Zen2 processors.
436In the future, it might set it by default only on those Zen2 processors whose
437microcode has not been updated to revisions fixing the vulnerability, once such
438microcode updates have been actually released and community-tested.
439To this mitigation are associated the following knobs:
440.Bl -tag -width indent
441.It Va machdep.mitigations.zenbleed.enable
442A read-write integer tunable and sysctl indicating whether the mitigation should
443be forcibly disabled (0), enabled (1) or if it is left to
444.Fx
445to selectively apply it (2).
446Any other integer value is silently converted to and treated as value 2.
447Note that this setting is silently ignored when running on non-Zen2 processors
448to ease applying a common configuration to heterogeneous machines.
449.It Va machdep.mitigations.zenbleed.state
450A read-only string indicating the current mitigation state.
451It can be either
452.Dq Not applicable ,
453if the processor is not Zen2-based,
454.Dq Mitigation enabled
455or
456.Dq Mitigation disabled .
457This state is automatically updated each time the sysctl
458.Va machdep.mitigations.zenbleed.enable
459is written to.
460Note that it can become inaccurate if the chicken bit is set or cleared
461directly via
462.Xr cpuctl 4
463.Po which includes the
464.Xr cpucontrol 8
465utility
466.Pc .
467.El
468.Pp
469The performance impact and threat models related to these mitigations
470should be considered when configuring and deploying them in a
471.Fx
472system.
473.Pp
474Additional mitigation knobs are listed in the
475.Sx KNOBS AND TWEAKS
476section of
477.Xr security 7 .
478.Sh SEE ALSO
479.Xr elfctl 1 ,
480.Xr proccontrol 1 ,
481.Xr rtld 1 ,
482.Xr mmap 2 ,
483.Xr src.conf 5 ,
484.Xr sysctl.conf 5 ,
485.Xr security 7 ,
486.Xr cpucontrol 8 ,
487.Xr sysctl 8
488