1 The Kernel Address Sanitizer (KASAN)
2 ====================================
7 KernelAddressSANitizer (KASAN) is a dynamic memory safety error detector
8 designed to find out-of-bound and use-after-free bugs. KASAN has three modes:
10 1. generic KASAN (similar to userspace ASan),
11 2. software tag-based KASAN (similar to userspace HWASan),
12 3. hardware tag-based KASAN (based on hardware memory tagging).
14 Generic KASAN is mainly used for debugging due to a large memory overhead.
15 Software tag-based KASAN can be used for dogfood testing as it has a lower
16 memory overhead that allows using it with real workloads. Hardware tag-based
17 KASAN comes with low memory and performance overheads and, therefore, can be
18 used in production. Either as an in-field memory bug detector or as a security
21 Software KASAN modes (#1 and #2) use compile-time instrumentation to insert
22 validity checks before every memory access and, therefore, require a compiler
23 version that supports that.
25 Generic KASAN is supported in GCC and Clang. With GCC, it requires version
26 8.3.0 or later. Any supported Clang version is compatible, but detection of
27 out-of-bounds accesses for global variables is only supported since Clang 11.
29 Software tag-based KASAN mode is only supported in Clang.
31 The hardware KASAN mode (#3) relies on hardware to perform the checks but
32 still requires a compiler version that supports memory tagging instructions.
33 This mode is supported in GCC 10+ and Clang 11+.
35 Both software KASAN modes work with SLUB and SLAB memory allocators,
36 while the hardware tag-based KASAN currently only supports SLUB.
38 Currently, generic KASAN is supported for the x86_64, arm, arm64, xtensa, s390,
39 and riscv architectures, and tag-based KASAN modes are supported only for arm64.
44 To enable KASAN, configure the kernel with::
48 and choose between ``CONFIG_KASAN_GENERIC`` (to enable generic KASAN),
49 ``CONFIG_KASAN_SW_TAGS`` (to enable software tag-based KASAN), and
50 ``CONFIG_KASAN_HW_TAGS`` (to enable hardware tag-based KASAN).
52 For software modes, also choose between ``CONFIG_KASAN_OUTLINE`` and
53 ``CONFIG_KASAN_INLINE``. Outline and inline are compiler instrumentation types.
54 The former produces a smaller binary while the latter is 1.1-2 times faster.
56 To include alloc and free stack traces of affected slab objects into reports,
57 enable ``CONFIG_STACKTRACE``. To include alloc and free stack traces of affected
58 physical pages, enable ``CONFIG_PAGE_OWNER`` and boot with ``page_owner=on``.
63 A typical KASAN report looks like this::
65 ==================================================================
66 BUG: KASAN: slab-out-of-bounds in kmalloc_oob_right+0xa8/0xbc [test_kasan]
67 Write of size 1 at addr ffff8801f44ec37b by task insmod/2760
69 CPU: 1 PID: 2760 Comm: insmod Not tainted 4.19.0-rc3+ #698
70 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014
73 print_address_description+0x73/0x280
74 kasan_report+0x144/0x187
75 __asan_report_store1_noabort+0x17/0x20
76 kmalloc_oob_right+0xa8/0xbc [test_kasan]
77 kmalloc_tests_init+0x16/0x700 [test_kasan]
78 do_one_initcall+0xa5/0x3ae
79 do_init_module+0x1b6/0x547
80 load_module+0x75df/0x8070
81 __do_sys_init_module+0x1c6/0x200
82 __x64_sys_init_module+0x6e/0xb0
83 do_syscall_64+0x9f/0x2c0
84 entry_SYSCALL_64_after_hwframe+0x44/0xa9
85 RIP: 0033:0x7f96443109da
86 RSP: 002b:00007ffcf0b51b08 EFLAGS: 00000202 ORIG_RAX: 00000000000000af
87 RAX: ffffffffffffffda RBX: 000055dc3ee521a0 RCX: 00007f96443109da
88 RDX: 00007f96445cff88 RSI: 0000000000057a50 RDI: 00007f9644992000
89 RBP: 000055dc3ee510b0 R08: 0000000000000003 R09: 0000000000000000
90 R10: 00007f964430cd0a R11: 0000000000000202 R12: 00007f96445cff88
91 R13: 000055dc3ee51090 R14: 0000000000000000 R15: 0000000000000000
93 Allocated by task 2760:
95 kasan_kmalloc+0xa7/0xd0
96 kmem_cache_alloc_trace+0xe1/0x1b0
97 kmalloc_oob_right+0x56/0xbc [test_kasan]
98 kmalloc_tests_init+0x16/0x700 [test_kasan]
99 do_one_initcall+0xa5/0x3ae
100 do_init_module+0x1b6/0x547
101 load_module+0x75df/0x8070
102 __do_sys_init_module+0x1c6/0x200
103 __x64_sys_init_module+0x6e/0xb0
104 do_syscall_64+0x9f/0x2c0
105 entry_SYSCALL_64_after_hwframe+0x44/0xa9
109 __kasan_slab_free+0x135/0x190
110 kasan_slab_free+0xe/0x10
112 umh_complete+0x6a/0xa0
113 call_usermodehelper_exec_async+0x4c3/0x640
114 ret_from_fork+0x35/0x40
116 The buggy address belongs to the object at ffff8801f44ec300
117 which belongs to the cache kmalloc-128 of size 128
118 The buggy address is located 123 bytes inside of
119 128-byte region [ffff8801f44ec300, ffff8801f44ec380)
120 The buggy address belongs to the page:
121 page:ffffea0007d13b00 count:1 mapcount:0 mapping:ffff8801f7001640 index:0x0
122 flags: 0x200000000000100(slab)
123 raw: 0200000000000100 ffffea0007d11dc0 0000001a0000001a ffff8801f7001640
124 raw: 0000000000000000 0000000080150015 00000001ffffffff 0000000000000000
125 page dumped because: kasan: bad access detected
127 Memory state around the buggy address:
128 ffff8801f44ec200: fc fc fc fc fc fc fc fc fb fb fb fb fb fb fb fb
129 ffff8801f44ec280: fb fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
130 >ffff8801f44ec300: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 03
132 ffff8801f44ec380: fc fc fc fc fc fc fc fc fb fb fb fb fb fb fb fb
133 ffff8801f44ec400: fb fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
134 ==================================================================
136 The report header summarizes what kind of bug happened and what kind of access
137 caused it. It is followed by a stack trace of the bad access, a stack trace of
138 where the accessed memory was allocated (in case a slab object was accessed),
139 and a stack trace of where the object was freed (in case of a use-after-free
140 bug report). Next comes a description of the accessed slab object and the
141 information about the accessed memory page.
143 In the end, the report shows the memory state around the accessed address.
144 Internally, KASAN tracks memory state separately for each memory granule, which
145 is either 8 or 16 aligned bytes depending on KASAN mode. Each number in the
146 memory state section of the report shows the state of one of the memory
147 granules that surround the accessed address.
149 For generic KASAN, the size of each memory granule is 8. The state of each
150 granule is encoded in one shadow byte. Those 8 bytes can be accessible,
151 partially accessible, freed, or be a part of a redzone. KASAN uses the following
152 encoding for each shadow byte: 00 means that all 8 bytes of the corresponding
153 memory region are accessible; number N (1 <= N <= 7) means that the first N
154 bytes are accessible, and other (8 - N) bytes are not; any negative value
155 indicates that the entire 8-byte word is inaccessible. KASAN uses different
156 negative values to distinguish between different kinds of inaccessible memory
157 like redzones or freed memory (see mm/kasan/kasan.h).
159 In the report above, the arrow points to the shadow byte ``03``, which means
160 that the accessed address is partially accessible.
162 For tag-based KASAN modes, this last report section shows the memory tags around
163 the accessed address (see the `Implementation details`_ section).
165 Note that KASAN bug titles (like ``slab-out-of-bounds`` or ``use-after-free``)
166 are best-effort: KASAN prints the most probable bug type based on the limited
167 information it has. The actual type of the bug might be different.
169 Generic KASAN also reports up to two auxiliary call stack traces. These stack
170 traces point to places in code that interacted with the object but that are not
171 directly present in the bad access stack trace. Currently, this includes
172 call_rcu() and workqueue queuing.
177 KASAN is affected by the generic ``panic_on_warn`` command line parameter.
178 When it is enabled, KASAN panics the kernel after printing a bug report.
180 By default, KASAN prints a bug report only for the first invalid memory access.
181 With ``kasan_multi_shot``, KASAN prints a report on every invalid access. This
182 effectively disables ``panic_on_warn`` for KASAN reports.
184 Alternatively, independent of ``panic_on_warn`` the ``kasan.fault=`` boot
185 parameter can be used to control panic and reporting behaviour:
187 - ``kasan.fault=report`` or ``=panic`` controls whether to only print a KASAN
188 report or also panic the kernel (default: ``report``). The panic happens even
189 if ``kasan_multi_shot`` is enabled.
191 Hardware tag-based KASAN mode (see the section about various modes below) is
192 intended for use in production as a security mitigation. Therefore, it supports
193 additional boot parameters that allow disabling KASAN or controlling features:
195 - ``kasan=off`` or ``=on`` controls whether KASAN is enabled (default: ``on``).
197 - ``kasan.mode=sync`` or ``=async`` controls whether KASAN is configured in
198 synchronous or asynchronous mode of execution (default: ``sync``).
199 Synchronous mode: a bad access is detected immediately when a tag
201 Asynchronous mode: a bad access detection is delayed. When a tag check
202 fault occurs, the information is stored in hardware (in the TFSR_EL1
203 register for arm64). The kernel periodically checks the hardware and
204 only reports tag faults during these checks.
206 - ``kasan.stacktrace=off`` or ``=on`` disables or enables alloc and free stack
207 traces collection (default: ``on``).
209 Implementation details
210 ----------------------
215 Software KASAN modes use shadow memory to record whether each byte of memory is
216 safe to access and use compile-time instrumentation to insert shadow memory
217 checks before each memory access.
219 Generic KASAN dedicates 1/8th of kernel memory to its shadow memory (16TB
220 to cover 128TB on x86_64) and uses direct mapping with a scale and offset to
221 translate a memory address to its corresponding shadow address.
223 Here is the function which translates an address to its corresponding shadow
226 static inline void *kasan_mem_to_shadow(const void *addr)
228 return (void *)((unsigned long)addr >> KASAN_SHADOW_SCALE_SHIFT)
229 + KASAN_SHADOW_OFFSET;
232 where ``KASAN_SHADOW_SCALE_SHIFT = 3``.
234 Compile-time instrumentation is used to insert memory access checks. Compiler
235 inserts function calls (``__asan_load*(addr)``, ``__asan_store*(addr)``) before
236 each memory access of size 1, 2, 4, 8, or 16. These functions check whether
237 memory accesses are valid or not by checking corresponding shadow memory.
239 With inline instrumentation, instead of making function calls, the compiler
240 directly inserts the code to check shadow memory. This option significantly
241 enlarges the kernel, but it gives an x1.1-x2 performance boost over the
242 outline-instrumented kernel.
244 Generic KASAN is the only mode that delays the reuse of freed objects via
245 quarantine (see mm/kasan/quarantine.c for implementation).
247 Software tag-based KASAN
248 ~~~~~~~~~~~~~~~~~~~~~~~~
250 Software tag-based KASAN uses a software memory tagging approach to checking
251 access validity. It is currently only implemented for the arm64 architecture.
253 Software tag-based KASAN uses the Top Byte Ignore (TBI) feature of arm64 CPUs
254 to store a pointer tag in the top byte of kernel pointers. It uses shadow memory
255 to store memory tags associated with each 16-byte memory cell (therefore, it
256 dedicates 1/16th of the kernel memory for shadow memory).
258 On each memory allocation, software tag-based KASAN generates a random tag, tags
259 the allocated memory with this tag, and embeds the same tag into the returned
262 Software tag-based KASAN uses compile-time instrumentation to insert checks
263 before each memory access. These checks make sure that the tag of the memory
264 that is being accessed is equal to the tag of the pointer that is used to access
265 this memory. In case of a tag mismatch, software tag-based KASAN prints a bug
268 Software tag-based KASAN also has two instrumentation modes (outline, which
269 emits callbacks to check memory accesses; and inline, which performs the shadow
270 memory checks inline). With outline instrumentation mode, a bug report is
271 printed from the function that performs the access check. With inline
272 instrumentation, a ``brk`` instruction is emitted by the compiler, and a
273 dedicated ``brk`` handler is used to print bug reports.
275 Software tag-based KASAN uses 0xFF as a match-all pointer tag (accesses through
276 pointers with the 0xFF pointer tag are not checked). The value 0xFE is currently
277 reserved to tag freed memory regions.
279 Software tag-based KASAN currently only supports tagging of slab and page_alloc
282 Hardware tag-based KASAN
283 ~~~~~~~~~~~~~~~~~~~~~~~~
285 Hardware tag-based KASAN is similar to the software mode in concept but uses
286 hardware memory tagging support instead of compiler instrumentation and
289 Hardware tag-based KASAN is currently only implemented for arm64 architecture
290 and based on both arm64 Memory Tagging Extension (MTE) introduced in ARMv8.5
291 Instruction Set Architecture and Top Byte Ignore (TBI).
293 Special arm64 instructions are used to assign memory tags for each allocation.
294 Same tags are assigned to pointers to those allocations. On every memory
295 access, hardware makes sure that the tag of the memory that is being accessed is
296 equal to the tag of the pointer that is used to access this memory. In case of a
297 tag mismatch, a fault is generated, and a report is printed.
299 Hardware tag-based KASAN uses 0xFF as a match-all pointer tag (accesses through
300 pointers with the 0xFF pointer tag are not checked). The value 0xFE is currently
301 reserved to tag freed memory regions.
303 Hardware tag-based KASAN currently only supports tagging of slab and page_alloc
306 If the hardware does not support MTE (pre ARMv8.5), hardware tag-based KASAN
307 will not be enabled. In this case, all KASAN boot parameters are ignored.
309 Note that enabling CONFIG_KASAN_HW_TAGS always results in in-kernel TBI being
310 enabled. Even when ``kasan.mode=off`` is provided or when the hardware does not
311 support MTE (but supports TBI).
313 Hardware tag-based KASAN only reports the first found bug. After that, MTE tag
314 checking gets disabled.
319 The kernel maps memory in several different parts of the address space.
320 The range of kernel virtual addresses is large: there is not enough real
321 memory to support a real shadow region for every address that could be
322 accessed by the kernel. Therefore, KASAN only maps real shadow for certain
323 parts of the address space.
328 By default, architectures only map real memory over the shadow region
329 for the linear mapping (and potentially other small areas). For all
330 other areas - such as vmalloc and vmemmap space - a single read-only
331 page is mapped over the shadow area. This read-only shadow page
332 declares all memory accesses as permitted.
334 This presents a problem for modules: they do not live in the linear
335 mapping but in a dedicated module space. By hooking into the module
336 allocator, KASAN temporarily maps real shadow memory to cover them.
337 This allows detection of invalid accesses to module globals, for example.
339 This also creates an incompatibility with ``VMAP_STACK``: if the stack
340 lives in vmalloc space, it will be shadowed by the read-only page, and
341 the kernel will fault when trying to set up the shadow data for stack
347 With ``CONFIG_KASAN_VMALLOC``, KASAN can cover vmalloc space at the
348 cost of greater memory usage. Currently, this is supported on x86,
349 riscv, s390, and powerpc.
351 This works by hooking into vmalloc and vmap and dynamically
352 allocating real shadow memory to back the mappings.
354 Most mappings in vmalloc space are small, requiring less than a full
355 page of shadow space. Allocating a full shadow page per mapping would
356 therefore be wasteful. Furthermore, to ensure that different mappings
357 use different shadow pages, mappings would have to be aligned to
358 ``KASAN_GRANULE_SIZE * PAGE_SIZE``.
360 Instead, KASAN shares backing space across multiple mappings. It allocates
361 a backing page when a mapping in vmalloc space uses a particular page
362 of the shadow region. This page can be shared by other vmalloc
365 KASAN hooks into the vmap infrastructure to lazily clean up unused shadow
368 To avoid the difficulties around swapping mappings around, KASAN expects
369 that the part of the shadow region that covers the vmalloc space will
370 not be covered by the early shadow page but will be left unmapped.
371 This will require changes in arch-specific code.
373 This allows ``VMAP_STACK`` support on x86 and can simplify support of
374 architectures that do not have a fixed module region.
382 Software KASAN modes use compiler instrumentation to insert validity checks.
383 Such instrumentation might be incompatible with some parts of the kernel, and
384 therefore needs to be disabled.
386 Other parts of the kernel might access metadata for allocated objects.
387 Normally, KASAN detects and reports such accesses, but in some cases (e.g.,
388 in memory allocators), these accesses are valid.
390 For software KASAN modes, to disable instrumentation for a specific file or
391 directory, add a ``KASAN_SANITIZE`` annotation to the respective kernel
394 - For a single file (e.g., main.o)::
396 KASAN_SANITIZE_main.o := n
398 - For all files in one directory::
402 For software KASAN modes, to disable instrumentation on a per-function basis,
403 use the KASAN-specific ``__no_sanitize_address`` function attribute or the
404 generic ``noinstr`` one.
406 Note that disabling compiler instrumentation (either on a per-file or a
407 per-function basis) makes KASAN ignore the accesses that happen directly in
408 that code for software KASAN modes. It does not help when the accesses happen
409 indirectly (through calls to instrumented functions) or with the hardware
410 tag-based mode that does not use compiler instrumentation.
412 For software KASAN modes, to disable KASAN reports in a part of the kernel code
413 for the current task, annotate this part of the code with a
414 ``kasan_disable_current()``/``kasan_enable_current()`` section. This also
415 disables the reports for indirect accesses that happen through function calls.
417 For tag-based KASAN modes (include the hardware one), to disable access
418 checking, use ``kasan_reset_tag()`` or ``page_kasan_tag_reset()``. Note that
419 temporarily disabling access checking via ``page_kasan_tag_reset()`` requires
420 saving and restoring the per-page KASAN tag via
421 ``page_kasan_tag``/``page_kasan_tag_set``.
426 There are KASAN tests that allow verifying that KASAN works and can detect
427 certain types of memory corruptions. The tests consist of two parts:
429 1. Tests that are integrated with the KUnit Test Framework. Enabled with
430 ``CONFIG_KASAN_KUNIT_TEST``. These tests can be run and partially verified
431 automatically in a few different ways; see the instructions below.
433 2. Tests that are currently incompatible with KUnit. Enabled with
434 ``CONFIG_KASAN_MODULE_TEST`` and can only be run as a module. These tests can
435 only be verified manually by loading the kernel module and inspecting the
436 kernel log for KASAN reports.
438 Each KUnit-compatible KASAN test prints one of multiple KASAN reports if an
439 error is detected. Then the test prints its number and status.
443 ok 28 - kmalloc_double_kzfree
445 When a test fails due to a failed ``kmalloc``::
447 # kmalloc_large_oob_right: ASSERTION FAILED at lib/test_kasan.c:163
448 Expected ptr is not null, but is
449 not ok 4 - kmalloc_large_oob_right
451 When a test fails due to a missing KASAN report::
453 # kmalloc_double_kzfree: EXPECTATION FAILED at lib/test_kasan.c:974
454 KASAN failure expected in "kfree_sensitive(ptr)", but none occurred
455 not ok 44 - kmalloc_double_kzfree
458 At the end the cumulative status of all KASAN tests is printed. On success::
462 Or, if one of the tests failed::
466 There are a few ways to run KUnit-compatible KASAN tests.
470 With ``CONFIG_KUNIT`` enabled, KASAN-KUnit tests can be built as a loadable
471 module and run by loading ``test_kasan.ko`` with ``insmod`` or ``modprobe``.
475 With ``CONFIG_KUNIT`` built-in, KASAN-KUnit tests can be built-in as well.
476 In this case, the tests will run at boot as a late-init call.
480 With ``CONFIG_KUNIT`` and ``CONFIG_KASAN_KUNIT_TEST`` built-in, it is also
481 possible to use ``kunit_tool`` to see the results of KUnit tests in a more
482 readable way. This will not print the KASAN reports of the tests that passed.
483 See `KUnit documentation <https://www.kernel.org/doc/html/latest/dev-tools/kunit/index.html>`_
484 for more up-to-date information on ``kunit_tool``.
486 .. _KUnit: https://www.kernel.org/doc/html/latest/dev-tools/kunit/index.html