1 The Kernel Address Sanitizer (KASAN)
2 ====================================
7 Kernel Address Sanitizer (KASAN) is a dynamic memory safety error detector
8 designed to find out-of-bounds and use-after-free bugs.
10 KASAN has three modes:
13 2. Software Tag-Based KASAN
14 3. Hardware Tag-Based KASAN
16 Generic KASAN, enabled with CONFIG_KASAN_GENERIC, is the mode intended for
17 debugging, similar to userspace ASan. This mode is supported on many CPU
18 architectures, but it has significant performance and memory overheads.
20 Software Tag-Based KASAN or SW_TAGS KASAN, enabled with CONFIG_KASAN_SW_TAGS,
21 can be used for both debugging and dogfood testing, similar to userspace HWASan.
22 This mode is only supported for arm64, but its moderate memory overhead allows
23 using it for testing on memory-restricted devices with real workloads.
25 Hardware Tag-Based KASAN or HW_TAGS KASAN, enabled with CONFIG_KASAN_HW_TAGS,
26 is the mode intended to be used as an in-field memory bug detector or as a
27 security mitigation. This mode only works on arm64 CPUs that support MTE
28 (Memory Tagging Extension), but it has low memory and performance overheads and
29 thus can be used in production.
31 For details about the memory and performance impact of each KASAN mode, see the
32 descriptions of the corresponding Kconfig options.
34 The Generic and the Software Tag-Based modes are commonly referred to as the
35 software modes. The Software Tag-Based and the Hardware Tag-Based modes are
36 referred to as the tag-based modes.
44 Generic KASAN is supported on x86_64, arm, arm64, powerpc, riscv, s390, and
45 xtensa, and the tag-based KASAN modes are supported only on arm64.
50 Software KASAN modes use compile-time instrumentation to insert validity checks
51 before every memory access and thus require a compiler version that provides
52 support for that. The Hardware Tag-Based mode relies on hardware to perform
53 these checks but still requires a compiler version that supports the memory
56 Generic KASAN requires GCC version 8.3.0 or later
57 or any Clang version supported by the kernel.
59 Software Tag-Based KASAN requires GCC 11+
60 or any Clang version supported by the kernel.
62 Hardware Tag-Based KASAN requires GCC 10+ or Clang 12+.
67 Generic KASAN supports finding bugs in all of slab, page_alloc, vmap, vmalloc,
68 stack, and global memory.
70 Software Tag-Based KASAN supports slab, page_alloc, vmalloc, and stack memory.
72 Hardware Tag-Based KASAN supports slab, page_alloc, and non-executable vmalloc
75 For slab, both software KASAN modes support SLUB and SLAB allocators, while
76 Hardware Tag-Based KASAN only supports SLUB.
81 To enable KASAN, configure the kernel with::
85 and choose between ``CONFIG_KASAN_GENERIC`` (to enable Generic KASAN),
86 ``CONFIG_KASAN_SW_TAGS`` (to enable Software Tag-Based KASAN), and
87 ``CONFIG_KASAN_HW_TAGS`` (to enable Hardware Tag-Based KASAN).
89 For the software modes, also choose between ``CONFIG_KASAN_OUTLINE`` and
90 ``CONFIG_KASAN_INLINE``. Outline and inline are compiler instrumentation types.
91 The former produces a smaller binary while the latter is up to 2 times faster.
93 To include alloc and free stack traces of affected slab objects into reports,
94 enable ``CONFIG_STACKTRACE``. To include alloc and free stack traces of affected
95 physical pages, enable ``CONFIG_PAGE_OWNER`` and boot with ``page_owner=on``.
100 KASAN is affected by the generic ``panic_on_warn`` command line parameter.
101 When it is enabled, KASAN panics the kernel after printing a bug report.
103 By default, KASAN prints a bug report only for the first invalid memory access.
104 With ``kasan_multi_shot``, KASAN prints a report on every invalid access. This
105 effectively disables ``panic_on_warn`` for KASAN reports.
107 Alternatively, independent of ``panic_on_warn``, the ``kasan.fault=`` boot
108 parameter can be used to control panic and reporting behaviour:
110 - ``kasan.fault=report`` or ``=panic`` controls whether to only print a KASAN
111 report or also panic the kernel (default: ``report``). The panic happens even
112 if ``kasan_multi_shot`` is enabled.
114 Software and Hardware Tag-Based KASAN modes (see the section about various
115 modes below) support altering stack trace collection behavior:
117 - ``kasan.stacktrace=off`` or ``=on`` disables or enables alloc and free stack
118 traces collection (default: ``on``).
119 - ``kasan.stack_ring_size=<number of entries>`` specifies the number of entries
120 in the stack ring (default: ``32768``).
122 Hardware Tag-Based KASAN mode is intended for use in production as a security
123 mitigation. Therefore, it supports additional boot parameters that allow
124 disabling KASAN altogether or controlling its features:
126 - ``kasan=off`` or ``=on`` controls whether KASAN is enabled (default: ``on``).
128 - ``kasan.mode=sync``, ``=async`` or ``=asymm`` controls whether KASAN
129 is configured in synchronous, asynchronous or asymmetric mode of
130 execution (default: ``sync``).
131 Synchronous mode: a bad access is detected immediately when a tag
133 Asynchronous mode: a bad access detection is delayed. When a tag check
134 fault occurs, the information is stored in hardware (in the TFSR_EL1
135 register for arm64). The kernel periodically checks the hardware and
136 only reports tag faults during these checks.
137 Asymmetric mode: a bad access is detected synchronously on reads and
138 asynchronously on writes.
140 - ``kasan.vmalloc=off`` or ``=on`` disables or enables tagging of vmalloc
141 allocations (default: ``on``).
143 - ``kasan.page_alloc.sample=<sampling interval>`` makes KASAN tag only every
144 Nth page_alloc allocation with the order equal or greater than
145 ``kasan.page_alloc.sample.order``, where N is the value of the ``sample``
146 parameter (default: ``1``, or tag every such allocation).
147 This parameter is intended to mitigate the performance overhead introduced
149 Note that enabling this parameter makes Hardware Tag-Based KASAN skip checks
150 of allocations chosen by sampling and thus miss bad accesses to these
151 allocations. Use the default value for accurate bug detection.
153 - ``kasan.page_alloc.sample.order=<minimum page order>`` specifies the minimum
154 order of allocations that are affected by sampling (default: ``3``).
155 Only applies when ``kasan.page_alloc.sample`` is set to a value greater
157 This parameter is intended to allow sampling only large page_alloc
158 allocations, which is the biggest source of the performance overhead.
163 A typical KASAN report looks like this::
165 ==================================================================
166 BUG: KASAN: slab-out-of-bounds in kmalloc_oob_right+0xa8/0xbc [test_kasan]
167 Write of size 1 at addr ffff8801f44ec37b by task insmod/2760
169 CPU: 1 PID: 2760 Comm: insmod Not tainted 4.19.0-rc3+ #698
170 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014
173 print_address_description+0x73/0x280
174 kasan_report+0x144/0x187
175 __asan_report_store1_noabort+0x17/0x20
176 kmalloc_oob_right+0xa8/0xbc [test_kasan]
177 kmalloc_tests_init+0x16/0x700 [test_kasan]
178 do_one_initcall+0xa5/0x3ae
179 do_init_module+0x1b6/0x547
180 load_module+0x75df/0x8070
181 __do_sys_init_module+0x1c6/0x200
182 __x64_sys_init_module+0x6e/0xb0
183 do_syscall_64+0x9f/0x2c0
184 entry_SYSCALL_64_after_hwframe+0x44/0xa9
185 RIP: 0033:0x7f96443109da
186 RSP: 002b:00007ffcf0b51b08 EFLAGS: 00000202 ORIG_RAX: 00000000000000af
187 RAX: ffffffffffffffda RBX: 000055dc3ee521a0 RCX: 00007f96443109da
188 RDX: 00007f96445cff88 RSI: 0000000000057a50 RDI: 00007f9644992000
189 RBP: 000055dc3ee510b0 R08: 0000000000000003 R09: 0000000000000000
190 R10: 00007f964430cd0a R11: 0000000000000202 R12: 00007f96445cff88
191 R13: 000055dc3ee51090 R14: 0000000000000000 R15: 0000000000000000
193 Allocated by task 2760:
195 kasan_kmalloc+0xa7/0xd0
196 kmem_cache_alloc_trace+0xe1/0x1b0
197 kmalloc_oob_right+0x56/0xbc [test_kasan]
198 kmalloc_tests_init+0x16/0x700 [test_kasan]
199 do_one_initcall+0xa5/0x3ae
200 do_init_module+0x1b6/0x547
201 load_module+0x75df/0x8070
202 __do_sys_init_module+0x1c6/0x200
203 __x64_sys_init_module+0x6e/0xb0
204 do_syscall_64+0x9f/0x2c0
205 entry_SYSCALL_64_after_hwframe+0x44/0xa9
209 __kasan_slab_free+0x135/0x190
210 kasan_slab_free+0xe/0x10
212 umh_complete+0x6a/0xa0
213 call_usermodehelper_exec_async+0x4c3/0x640
214 ret_from_fork+0x35/0x40
216 The buggy address belongs to the object at ffff8801f44ec300
217 which belongs to the cache kmalloc-128 of size 128
218 The buggy address is located 123 bytes inside of
219 128-byte region [ffff8801f44ec300, ffff8801f44ec380)
220 The buggy address belongs to the page:
221 page:ffffea0007d13b00 count:1 mapcount:0 mapping:ffff8801f7001640 index:0x0
222 flags: 0x200000000000100(slab)
223 raw: 0200000000000100 ffffea0007d11dc0 0000001a0000001a ffff8801f7001640
224 raw: 0000000000000000 0000000080150015 00000001ffffffff 0000000000000000
225 page dumped because: kasan: bad access detected
227 Memory state around the buggy address:
228 ffff8801f44ec200: fc fc fc fc fc fc fc fc fb fb fb fb fb fb fb fb
229 ffff8801f44ec280: fb fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
230 >ffff8801f44ec300: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 03
232 ffff8801f44ec380: fc fc fc fc fc fc fc fc fb fb fb fb fb fb fb fb
233 ffff8801f44ec400: fb fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
234 ==================================================================
236 The report header summarizes what kind of bug happened and what kind of access
237 caused it. It is followed by a stack trace of the bad access, a stack trace of
238 where the accessed memory was allocated (in case a slab object was accessed),
239 and a stack trace of where the object was freed (in case of a use-after-free
240 bug report). Next comes a description of the accessed slab object and the
241 information about the accessed memory page.
243 In the end, the report shows the memory state around the accessed address.
244 Internally, KASAN tracks memory state separately for each memory granule, which
245 is either 8 or 16 aligned bytes depending on KASAN mode. Each number in the
246 memory state section of the report shows the state of one of the memory
247 granules that surround the accessed address.
249 For Generic KASAN, the size of each memory granule is 8. The state of each
250 granule is encoded in one shadow byte. Those 8 bytes can be accessible,
251 partially accessible, freed, or be a part of a redzone. KASAN uses the following
252 encoding for each shadow byte: 00 means that all 8 bytes of the corresponding
253 memory region are accessible; number N (1 <= N <= 7) means that the first N
254 bytes are accessible, and other (8 - N) bytes are not; any negative value
255 indicates that the entire 8-byte word is inaccessible. KASAN uses different
256 negative values to distinguish between different kinds of inaccessible memory
257 like redzones or freed memory (see mm/kasan/kasan.h).
259 In the report above, the arrow points to the shadow byte ``03``, which means
260 that the accessed address is partially accessible.
262 For tag-based KASAN modes, this last report section shows the memory tags around
263 the accessed address (see the `Implementation details`_ section).
265 Note that KASAN bug titles (like ``slab-out-of-bounds`` or ``use-after-free``)
266 are best-effort: KASAN prints the most probable bug type based on the limited
267 information it has. The actual type of the bug might be different.
269 Generic KASAN also reports up to two auxiliary call stack traces. These stack
270 traces point to places in code that interacted with the object but that are not
271 directly present in the bad access stack trace. Currently, this includes
272 call_rcu() and workqueue queuing.
274 Implementation details
275 ----------------------
280 Software KASAN modes use shadow memory to record whether each byte of memory is
281 safe to access and use compile-time instrumentation to insert shadow memory
282 checks before each memory access.
284 Generic KASAN dedicates 1/8th of kernel memory to its shadow memory (16TB
285 to cover 128TB on x86_64) and uses direct mapping with a scale and offset to
286 translate a memory address to its corresponding shadow address.
288 Here is the function which translates an address to its corresponding shadow
291 static inline void *kasan_mem_to_shadow(const void *addr)
293 return (void *)((unsigned long)addr >> KASAN_SHADOW_SCALE_SHIFT)
294 + KASAN_SHADOW_OFFSET;
297 where ``KASAN_SHADOW_SCALE_SHIFT = 3``.
299 Compile-time instrumentation is used to insert memory access checks. Compiler
300 inserts function calls (``__asan_load*(addr)``, ``__asan_store*(addr)``) before
301 each memory access of size 1, 2, 4, 8, or 16. These functions check whether
302 memory accesses are valid or not by checking corresponding shadow memory.
304 With inline instrumentation, instead of making function calls, the compiler
305 directly inserts the code to check shadow memory. This option significantly
306 enlarges the kernel, but it gives an x1.1-x2 performance boost over the
307 outline-instrumented kernel.
309 Generic KASAN is the only mode that delays the reuse of freed objects via
310 quarantine (see mm/kasan/quarantine.c for implementation).
312 Software Tag-Based KASAN
313 ~~~~~~~~~~~~~~~~~~~~~~~~
315 Software Tag-Based KASAN uses a software memory tagging approach to checking
316 access validity. It is currently only implemented for the arm64 architecture.
318 Software Tag-Based KASAN uses the Top Byte Ignore (TBI) feature of arm64 CPUs
319 to store a pointer tag in the top byte of kernel pointers. It uses shadow memory
320 to store memory tags associated with each 16-byte memory cell (therefore, it
321 dedicates 1/16th of the kernel memory for shadow memory).
323 On each memory allocation, Software Tag-Based KASAN generates a random tag, tags
324 the allocated memory with this tag, and embeds the same tag into the returned
327 Software Tag-Based KASAN uses compile-time instrumentation to insert checks
328 before each memory access. These checks make sure that the tag of the memory
329 that is being accessed is equal to the tag of the pointer that is used to access
330 this memory. In case of a tag mismatch, Software Tag-Based KASAN prints a bug
333 Software Tag-Based KASAN also has two instrumentation modes (outline, which
334 emits callbacks to check memory accesses; and inline, which performs the shadow
335 memory checks inline). With outline instrumentation mode, a bug report is
336 printed from the function that performs the access check. With inline
337 instrumentation, a ``brk`` instruction is emitted by the compiler, and a
338 dedicated ``brk`` handler is used to print bug reports.
340 Software Tag-Based KASAN uses 0xFF as a match-all pointer tag (accesses through
341 pointers with the 0xFF pointer tag are not checked). The value 0xFE is currently
342 reserved to tag freed memory regions.
344 Hardware Tag-Based KASAN
345 ~~~~~~~~~~~~~~~~~~~~~~~~
347 Hardware Tag-Based KASAN is similar to the software mode in concept but uses
348 hardware memory tagging support instead of compiler instrumentation and
351 Hardware Tag-Based KASAN is currently only implemented for arm64 architecture
352 and based on both arm64 Memory Tagging Extension (MTE) introduced in ARMv8.5
353 Instruction Set Architecture and Top Byte Ignore (TBI).
355 Special arm64 instructions are used to assign memory tags for each allocation.
356 Same tags are assigned to pointers to those allocations. On every memory
357 access, hardware makes sure that the tag of the memory that is being accessed is
358 equal to the tag of the pointer that is used to access this memory. In case of a
359 tag mismatch, a fault is generated, and a report is printed.
361 Hardware Tag-Based KASAN uses 0xFF as a match-all pointer tag (accesses through
362 pointers with the 0xFF pointer tag are not checked). The value 0xFE is currently
363 reserved to tag freed memory regions.
365 If the hardware does not support MTE (pre ARMv8.5), Hardware Tag-Based KASAN
366 will not be enabled. In this case, all KASAN boot parameters are ignored.
368 Note that enabling CONFIG_KASAN_HW_TAGS always results in in-kernel TBI being
369 enabled. Even when ``kasan.mode=off`` is provided or when the hardware does not
370 support MTE (but supports TBI).
372 Hardware Tag-Based KASAN only reports the first found bug. After that, MTE tag
373 checking gets disabled.
378 The contents of this section are only applicable to software KASAN modes.
380 The kernel maps memory in several different parts of the address space.
381 The range of kernel virtual addresses is large: there is not enough real
382 memory to support a real shadow region for every address that could be
383 accessed by the kernel. Therefore, KASAN only maps real shadow for certain
384 parts of the address space.
389 By default, architectures only map real memory over the shadow region
390 for the linear mapping (and potentially other small areas). For all
391 other areas - such as vmalloc and vmemmap space - a single read-only
392 page is mapped over the shadow area. This read-only shadow page
393 declares all memory accesses as permitted.
395 This presents a problem for modules: they do not live in the linear
396 mapping but in a dedicated module space. By hooking into the module
397 allocator, KASAN temporarily maps real shadow memory to cover them.
398 This allows detection of invalid accesses to module globals, for example.
400 This also creates an incompatibility with ``VMAP_STACK``: if the stack
401 lives in vmalloc space, it will be shadowed by the read-only page, and
402 the kernel will fault when trying to set up the shadow data for stack
408 With ``CONFIG_KASAN_VMALLOC``, KASAN can cover vmalloc space at the
409 cost of greater memory usage. Currently, this is supported on x86,
410 arm64, riscv, s390, and powerpc.
412 This works by hooking into vmalloc and vmap and dynamically
413 allocating real shadow memory to back the mappings.
415 Most mappings in vmalloc space are small, requiring less than a full
416 page of shadow space. Allocating a full shadow page per mapping would
417 therefore be wasteful. Furthermore, to ensure that different mappings
418 use different shadow pages, mappings would have to be aligned to
419 ``KASAN_GRANULE_SIZE * PAGE_SIZE``.
421 Instead, KASAN shares backing space across multiple mappings. It allocates
422 a backing page when a mapping in vmalloc space uses a particular page
423 of the shadow region. This page can be shared by other vmalloc
426 KASAN hooks into the vmap infrastructure to lazily clean up unused shadow
429 To avoid the difficulties around swapping mappings around, KASAN expects
430 that the part of the shadow region that covers the vmalloc space will
431 not be covered by the early shadow page but will be left unmapped.
432 This will require changes in arch-specific code.
434 This allows ``VMAP_STACK`` support on x86 and can simplify support of
435 architectures that do not have a fixed module region.
443 Software KASAN modes use compiler instrumentation to insert validity checks.
444 Such instrumentation might be incompatible with some parts of the kernel, and
445 therefore needs to be disabled.
447 Other parts of the kernel might access metadata for allocated objects.
448 Normally, KASAN detects and reports such accesses, but in some cases (e.g.,
449 in memory allocators), these accesses are valid.
451 For software KASAN modes, to disable instrumentation for a specific file or
452 directory, add a ``KASAN_SANITIZE`` annotation to the respective kernel
455 - For a single file (e.g., main.o)::
457 KASAN_SANITIZE_main.o := n
459 - For all files in one directory::
463 For software KASAN modes, to disable instrumentation on a per-function basis,
464 use the KASAN-specific ``__no_sanitize_address`` function attribute or the
465 generic ``noinstr`` one.
467 Note that disabling compiler instrumentation (either on a per-file or a
468 per-function basis) makes KASAN ignore the accesses that happen directly in
469 that code for software KASAN modes. It does not help when the accesses happen
470 indirectly (through calls to instrumented functions) or with Hardware
471 Tag-Based KASAN, which does not use compiler instrumentation.
473 For software KASAN modes, to disable KASAN reports in a part of the kernel code
474 for the current task, annotate this part of the code with a
475 ``kasan_disable_current()``/``kasan_enable_current()`` section. This also
476 disables the reports for indirect accesses that happen through function calls.
478 For tag-based KASAN modes, to disable access checking, use
479 ``kasan_reset_tag()`` or ``page_kasan_tag_reset()``. Note that temporarily
480 disabling access checking via ``page_kasan_tag_reset()`` requires saving and
481 restoring the per-page KASAN tag via ``page_kasan_tag``/``page_kasan_tag_set``.
486 There are KASAN tests that allow verifying that KASAN works and can detect
487 certain types of memory corruptions. The tests consist of two parts:
489 1. Tests that are integrated with the KUnit Test Framework. Enabled with
490 ``CONFIG_KASAN_KUNIT_TEST``. These tests can be run and partially verified
491 automatically in a few different ways; see the instructions below.
493 2. Tests that are currently incompatible with KUnit. Enabled with
494 ``CONFIG_KASAN_MODULE_TEST`` and can only be run as a module. These tests can
495 only be verified manually by loading the kernel module and inspecting the
496 kernel log for KASAN reports.
498 Each KUnit-compatible KASAN test prints one of multiple KASAN reports if an
499 error is detected. Then the test prints its number and status.
503 ok 28 - kmalloc_double_kzfree
505 When a test fails due to a failed ``kmalloc``::
507 # kmalloc_large_oob_right: ASSERTION FAILED at lib/test_kasan.c:163
508 Expected ptr is not null, but is
509 not ok 4 - kmalloc_large_oob_right
511 When a test fails due to a missing KASAN report::
513 # kmalloc_double_kzfree: EXPECTATION FAILED at lib/test_kasan.c:974
514 KASAN failure expected in "kfree_sensitive(ptr)", but none occurred
515 not ok 44 - kmalloc_double_kzfree
518 At the end the cumulative status of all KASAN tests is printed. On success::
522 Or, if one of the tests failed::
526 There are a few ways to run KUnit-compatible KASAN tests.
530 With ``CONFIG_KUNIT`` enabled, KASAN-KUnit tests can be built as a loadable
531 module and run by loading ``test_kasan.ko`` with ``insmod`` or ``modprobe``.
535 With ``CONFIG_KUNIT`` built-in, KASAN-KUnit tests can be built-in as well.
536 In this case, the tests will run at boot as a late-init call.
540 With ``CONFIG_KUNIT`` and ``CONFIG_KASAN_KUNIT_TEST`` built-in, it is also
541 possible to use ``kunit_tool`` to see the results of KUnit tests in a more
542 readable way. This will not print the KASAN reports of the tests that passed.
543 See `KUnit documentation <https://www.kernel.org/doc/html/latest/dev-tools/kunit/index.html>`_
544 for more up-to-date information on ``kunit_tool``.
546 .. _KUnit: https://www.kernel.org/doc/html/latest/dev-tools/kunit/index.html