summaryrefslogtreecommitdiff
path: root/Documentation/security
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2016-05-19 18:07:25 -0700
committerLinus Torvalds <torvalds@linux-foundation.org>2016-05-19 18:07:25 -0700
commite9ad9b9bd3a3b95c89a29b2a197476e662db4233 (patch)
tree0b5e6c09953edc1a0bfab0b2aa964b80bceb5efd /Documentation/security
parent78975f23cba0cd195db01cdbd6eb48138a655890 (diff)
parent9f8036643dd9609b329aa1b89c9a95981e9ba62f (diff)
Merge tag 'docs-for-linus' of git://git.lwn.net/linux
Pull Documentation updates from Jon Corbet: "A bit busier this time around. The most interesting thing (IMO) this time around is some beginning infrastructural work to allow documents to be written using restructured text. Maybe someday, in a galaxy far far away, we'll be able to eliminate the DocBook dependency and have a much better integrated set of kernel docs. Someday. Beyond that, there's a new document on security hardening from Kees, the movement of some sample code over to samples/, a number of improvements to the serial docs from Geert, and the usual collection of corrections, typo fixes, etc" * tag 'docs-for-linus' of git://git.lwn.net/linux: (55 commits) doc: self-protection: provide initial details serial: doc: Use port->state instead of info serial: doc: Always refer to tty_port->mutex Documentation: vm: Spelling s/paltform/platform/g Documentation/memcg: update kmem limit doc as codes behavior docproc: print a comment about autogeneration for rst output docproc: add support for reStructuredText format via --rst option docproc: abstract terminating lines at first space docproc: abstract docproc directive detection docproc: reduce unnecessary indentation docproc: add variables for subcommand and filename kernel-doc: use rst C domain directives and references for types kernel-doc: produce RestructuredText output kernel-doc: rewrite usage description, remove duplicated comments Doc: correct the location of sysrq.c Documentation: fix common spelling mistakes samples: v4l: from Documentation to samples directory samples: connector: from Documentation to samples directory Documentation: xillybus: fix spelling mistake Documentation: x86: fix spelling mistakes ...
Diffstat (limited to 'Documentation/security')
-rw-r--r--Documentation/security/self-protection.txt261
1 files changed, 261 insertions, 0 deletions
diff --git a/Documentation/security/self-protection.txt b/Documentation/security/self-protection.txt
new file mode 100644
index 000000000000..babd6378ec05
--- /dev/null
+++ b/Documentation/security/self-protection.txt
@@ -0,0 +1,261 @@
+# Kernel Self-Protection
+
+Kernel self-protection is the design and implementation of systems and
+structures within the Linux kernel to protect against security flaws in
+the kernel itself. This covers a wide range of issues, including removing
+entire classes of bugs, blocking security flaw exploitation methods,
+and actively detecting attack attempts. Not all topics are explored in
+this document, but it should serve as a reasonable starting point and
+answer any frequently asked questions. (Patches welcome, of course!)
+
+In the worst-case scenario, we assume an unprivileged local attacker
+has arbitrary read and write access to the kernel's memory. In many
+cases, bugs being exploited will not provide this level of access,
+but with systems in place that defend against the worst case we'll
+cover the more limited cases as well. A higher bar, and one that should
+still be kept in mind, is protecting the kernel against a _privileged_
+local attacker, since the root user has access to a vastly increased
+attack surface. (Especially when they have the ability to load arbitrary
+kernel modules.)
+
+The goals for successful self-protection systems would be that they
+are effective, on by default, require no opt-in by developers, have no
+performance impact, do not impede kernel debugging, and have tests. It
+is uncommon that all these goals can be met, but it is worth explicitly
+mentioning them, since these aspects need to be explored, dealt with,
+and/or accepted.
+
+
+## Attack Surface Reduction
+
+The most fundamental defense against security exploits is to reduce the
+areas of the kernel that can be used to redirect execution. This ranges
+from limiting the exposed APIs available to userspace, making in-kernel
+APIs hard to use incorrectly, minimizing the areas of writable kernel
+memory, etc.
+
+### Strict kernel memory permissions
+
+When all of kernel memory is writable, it becomes trivial for attacks
+to redirect execution flow. To reduce the availability of these targets
+the kernel needs to protect its memory with a tight set of permissions.
+
+#### Executable code and read-only data must not be writable
+
+Any areas of the kernel with executable memory must not be writable.
+While this obviously includes the kernel text itself, we must consider
+all additional places too: kernel modules, JIT memory, etc. (There are
+temporary exceptions to this rule to support things like instruction
+alternatives, breakpoints, kprobes, etc. If these must exist in a
+kernel, they are implemented in a way where the memory is temporarily
+made writable during the update, and then returned to the original
+permissions.)
+
+In support of this are (the poorly named) CONFIG_DEBUG_RODATA and
+CONFIG_DEBUG_SET_MODULE_RONX, which seek to make sure that code is not
+writable, data is not executable, and read-only data is neither writable
+nor executable.
+
+#### Function pointers and sensitive variables must not be writable
+
+Vast areas of kernel memory contain function pointers that are looked
+up by the kernel and used to continue execution (e.g. descriptor/vector
+tables, file/network/etc operation structures, etc). The number of these
+variables must be reduced to an absolute minimum.
+
+Many such variables can be made read-only by setting them "const"
+so that they live in the .rodata section instead of the .data section
+of the kernel, gaining the protection of the kernel's strict memory
+permissions as described above.
+
+For variables that are initialized once at __init time, these can
+be marked with the (new and under development) __ro_after_init
+attribute.
+
+What remains are variables that are updated rarely (e.g. GDT). These
+will need another infrastructure (similar to the temporary exceptions
+made to kernel code mentioned above) that allow them to spend the rest
+of their lifetime read-only. (For example, when being updated, only the
+CPU thread performing the update would be given uninterruptible write
+access to the memory.)
+
+#### Segregation of kernel memory from userspace memory
+
+The kernel must never execute userspace memory. The kernel must also never
+access userspace memory without explicit expectation to do so. These
+rules can be enforced either by support of hardware-based restrictions
+(x86's SMEP/SMAP, ARM's PXN/PAN) or via emulation (ARM's Memory Domains).
+By blocking userspace memory in this way, execution and data parsing
+cannot be passed to trivially-controlled userspace memory, forcing
+attacks to operate entirely in kernel memory.
+
+### Reduced access to syscalls
+
+One trivial way to eliminate many syscalls for 64-bit systems is building
+without CONFIG_COMPAT. However, this is rarely a feasible scenario.
+
+The "seccomp" system provides an opt-in feature made available to
+userspace, which provides a way to reduce the number of kernel entry
+points available to a running process. This limits the breadth of kernel
+code that can be reached, possibly reducing the availability of a given
+bug to an attack.
+
+An area of improvement would be creating viable ways to keep access to
+things like compat, user namespaces, BPF creation, and perf limited only
+to trusted processes. This would keep the scope of kernel entry points
+restricted to the more regular set of normally available to unprivileged
+userspace.
+
+### Restricting access to kernel modules
+
+The kernel should never allow an unprivileged user the ability to
+load specific kernel modules, since that would provide a facility to
+unexpectedly extend the available attack surface. (The on-demand loading
+of modules via their predefined subsystems, e.g. MODULE_ALIAS_*, is
+considered "expected" here, though additional consideration should be
+given even to these.) For example, loading a filesystem module via an
+unprivileged socket API is nonsense: only the root or physically local
+user should trigger filesystem module loading. (And even this can be up
+for debate in some scenarios.)
+
+To protect against even privileged users, systems may need to either
+disable module loading entirely (e.g. monolithic kernel builds or
+modules_disabled sysctl), or provide signed modules (e.g.
+CONFIG_MODULE_SIG_FORCE, or dm-crypt with LoadPin), to keep from having
+root load arbitrary kernel code via the module loader interface.
+
+
+## Memory integrity
+
+There are many memory structures in the kernel that are regularly abused
+to gain execution control during an attack, By far the most commonly
+understood is that of the stack buffer overflow in which the return
+address stored on the stack is overwritten. Many other examples of this
+kind of attack exist, and protections exist to defend against them.
+
+### Stack buffer overflow
+
+The classic stack buffer overflow involves writing past the expected end
+of a variable stored on the stack, ultimately writing a controlled value
+to the stack frame's stored return address. The most widely used defense
+is the presence of a stack canary between the stack variables and the
+return address (CONFIG_CC_STACKPROTECTOR), which is verified just before
+the function returns. Other defenses include things like shadow stacks.
+
+### Stack depth overflow
+
+A less well understood attack is using a bug that triggers the
+kernel to consume stack memory with deep function calls or large stack
+allocations. With this attack it is possible to write beyond the end of
+the kernel's preallocated stack space and into sensitive structures. Two
+important changes need to be made for better protections: moving the
+sensitive thread_info structure elsewhere, and adding a faulting memory
+hole at the bottom of the stack to catch these overflows.
+
+### Heap memory integrity
+
+The structures used to track heap free lists can be sanity-checked during
+allocation and freeing to make sure they aren't being used to manipulate
+other memory areas.
+
+### Counter integrity
+
+Many places in the kernel use atomic counters to track object references
+or perform similar lifetime management. When these counters can be made
+to wrap (over or under) this traditionally exposes a use-after-free
+flaw. By trapping atomic wrapping, this class of bug vanishes.
+
+### Size calculation overflow detection
+
+Similar to counter overflow, integer overflows (usually size calculations)
+need to be detected at runtime to kill this class of bug, which
+traditionally leads to being able to write past the end of kernel buffers.
+
+
+## Statistical defenses
+
+While many protections can be considered deterministic (e.g. read-only
+memory cannot be written to), some protections provide only statistical
+defense, in that an attack must gather enough information about a
+running system to overcome the defense. While not perfect, these do
+provide meaningful defenses.
+
+### Canaries, blinding, and other secrets
+
+It should be noted that things like the stack canary discussed earlier
+are technically statistical defenses, since they rely on a (leakable)
+secret value.
+
+Blinding literal values for things like JITs, where the executable
+contents may be partially under the control of userspace, need a similar
+secret value.
+
+It is critical that the secret values used must be separate (e.g.
+different canary per stack) and high entropy (e.g. is the RNG actually
+working?) in order to maximize their success.
+
+### Kernel Address Space Layout Randomization (KASLR)
+
+Since the location of kernel memory is almost always instrumental in
+mounting a successful attack, making the location non-deterministic
+raises the difficulty of an exploit. (Note that this in turn makes
+the value of leaks higher, since they may be used to discover desired
+memory locations.)
+
+#### Text and module base
+
+By relocating the physical and virtual base address of the kernel at
+boot-time (CONFIG_RANDOMIZE_BASE), attacks needing kernel code will be
+frustrated. Additionally, offsetting the module loading base address
+means that even systems that load the same set of modules in the same
+order every boot will not share a common base address with the rest of
+the kernel text.
+
+#### Stack base
+
+If the base address of the kernel stack is not the same between processes,
+or even not the same between syscalls, targets on or beyond the stack
+become more difficult to locate.
+
+#### Dynamic memory base
+
+Much of the kernel's dynamic memory (e.g. kmalloc, vmalloc, etc) ends up
+being relatively deterministic in layout due to the order of early-boot
+initializations. If the base address of these areas is not the same
+between boots, targeting them is frustrated, requiring a leak specific
+to the region.
+
+
+## Preventing Leaks
+
+Since the locations of sensitive structures are the primary target for
+attacks, it is important to defend against leaks of both kernel memory
+addresses and kernel memory contents (since they may contain kernel
+addresses or other sensitive things like canary values).
+
+### Unique identifiers
+
+Kernel memory addresses must never be used as identifiers exposed to
+userspace. Instead, use an atomic counter, an idr, or similar unique
+identifier.
+
+### Memory initialization
+
+Memory copied to userspace must always be fully initialized. If not
+explicitly memset(), this will require changes to the compiler to make
+sure structure holes are cleared.
+
+### Memory poisoning
+
+When releasing memory, it is best to poison the contents (clear stack on
+syscall return, wipe heap memory on a free), to avoid reuse attacks that
+rely on the old contents of memory. This frustrates many uninitialized
+variable attacks, stack info leaks, heap info leaks, and use-after-free
+attacks.
+
+### Destination tracking
+
+To help kill classes of bugs that result in kernel addresses being
+written to userspace, the destination of writes needs to be tracked. If
+the buffer is destined for userspace (e.g. seq_file backed /proc files),
+it should automatically censor sensitive values.