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 Software KASAN modes (1 and 2) use compile-time instrumentation to insert
15 validity checks before every memory access, and therefore require a compiler
16 version that supports that.
18 Generic KASAN is supported in both GCC and Clang. With GCC it requires version
19 8.3.0 or later. Any supported Clang version is compatible, but detection of
20 out-of-bounds accesses for global variables is only supported since Clang 11.
22 Tag-based KASAN is only supported in Clang.
24 Currently generic KASAN is supported for the x86_64, arm, arm64, xtensa, s390
25 and riscv architectures, and tag-based KASAN modes are supported only for arm64.
30 To enable KASAN configure kernel with::
34 and choose between CONFIG_KASAN_GENERIC (to enable generic KASAN),
35 CONFIG_KASAN_SW_TAGS (to enable software tag-based KASAN), and
36 CONFIG_KASAN_HW_TAGS (to enable hardware tag-based KASAN).
38 For software modes, you also need to choose between CONFIG_KASAN_OUTLINE and
39 CONFIG_KASAN_INLINE. Outline and inline are compiler instrumentation types.
40 The former produces smaller binary while the latter is 1.1 - 2 times faster.
42 Both software KASAN modes work with both SLUB and SLAB memory allocators,
43 while the hardware tag-based KASAN currently only support SLUB.
45 For better error reports that include stack traces, enable CONFIG_STACKTRACE.
47 To augment reports with last allocation and freeing stack of the physical page,
48 it is recommended to enable also CONFIG_PAGE_OWNER and boot with page_owner=on.
53 A typical out-of-bounds access generic KASAN report looks like this::
55 ==================================================================
56 BUG: KASAN: slab-out-of-bounds in kmalloc_oob_right+0xa8/0xbc [test_kasan]
57 Write of size 1 at addr ffff8801f44ec37b by task insmod/2760
59 CPU: 1 PID: 2760 Comm: insmod Not tainted 4.19.0-rc3+ #698
60 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014
63 print_address_description+0x73/0x280
64 kasan_report+0x144/0x187
65 __asan_report_store1_noabort+0x17/0x20
66 kmalloc_oob_right+0xa8/0xbc [test_kasan]
67 kmalloc_tests_init+0x16/0x700 [test_kasan]
68 do_one_initcall+0xa5/0x3ae
69 do_init_module+0x1b6/0x547
70 load_module+0x75df/0x8070
71 __do_sys_init_module+0x1c6/0x200
72 __x64_sys_init_module+0x6e/0xb0
73 do_syscall_64+0x9f/0x2c0
74 entry_SYSCALL_64_after_hwframe+0x44/0xa9
75 RIP: 0033:0x7f96443109da
76 RSP: 002b:00007ffcf0b51b08 EFLAGS: 00000202 ORIG_RAX: 00000000000000af
77 RAX: ffffffffffffffda RBX: 000055dc3ee521a0 RCX: 00007f96443109da
78 RDX: 00007f96445cff88 RSI: 0000000000057a50 RDI: 00007f9644992000
79 RBP: 000055dc3ee510b0 R08: 0000000000000003 R09: 0000000000000000
80 R10: 00007f964430cd0a R11: 0000000000000202 R12: 00007f96445cff88
81 R13: 000055dc3ee51090 R14: 0000000000000000 R15: 0000000000000000
83 Allocated by task 2760:
85 kasan_kmalloc+0xa7/0xd0
86 kmem_cache_alloc_trace+0xe1/0x1b0
87 kmalloc_oob_right+0x56/0xbc [test_kasan]
88 kmalloc_tests_init+0x16/0x700 [test_kasan]
89 do_one_initcall+0xa5/0x3ae
90 do_init_module+0x1b6/0x547
91 load_module+0x75df/0x8070
92 __do_sys_init_module+0x1c6/0x200
93 __x64_sys_init_module+0x6e/0xb0
94 do_syscall_64+0x9f/0x2c0
95 entry_SYSCALL_64_after_hwframe+0x44/0xa9
99 __kasan_slab_free+0x135/0x190
100 kasan_slab_free+0xe/0x10
102 umh_complete+0x6a/0xa0
103 call_usermodehelper_exec_async+0x4c3/0x640
104 ret_from_fork+0x35/0x40
106 The buggy address belongs to the object at ffff8801f44ec300
107 which belongs to the cache kmalloc-128 of size 128
108 The buggy address is located 123 bytes inside of
109 128-byte region [ffff8801f44ec300, ffff8801f44ec380)
110 The buggy address belongs to the page:
111 page:ffffea0007d13b00 count:1 mapcount:0 mapping:ffff8801f7001640 index:0x0
112 flags: 0x200000000000100(slab)
113 raw: 0200000000000100 ffffea0007d11dc0 0000001a0000001a ffff8801f7001640
114 raw: 0000000000000000 0000000080150015 00000001ffffffff 0000000000000000
115 page dumped because: kasan: bad access detected
117 Memory state around the buggy address:
118 ffff8801f44ec200: fc fc fc fc fc fc fc fc fb fb fb fb fb fb fb fb
119 ffff8801f44ec280: fb fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
120 >ffff8801f44ec300: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 03
122 ffff8801f44ec380: fc fc fc fc fc fc fc fc fb fb fb fb fb fb fb fb
123 ffff8801f44ec400: fb fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
124 ==================================================================
126 The header of the report provides a short summary of what kind of bug happened
127 and what kind of access caused it. It's followed by a stack trace of the bad
128 access, a stack trace of where the accessed memory was allocated (in case bad
129 access happens on a slab object), and a stack trace of where the object was
130 freed (in case of a use-after-free bug report). Next comes a description of
131 the accessed slab object and information about the accessed memory page.
133 In the last section the report shows memory state around the accessed address.
134 Internally KASAN tracks memory state separately for each memory granule, which
135 is either 8 or 16 aligned bytes depending on KASAN mode. Each number in the
136 memory state section of the report shows the state of one of the memory
137 granules that surround the accessed address.
139 For generic KASAN the size of each memory granule is 8. The state of each
140 granule is encoded in one shadow byte. Those 8 bytes can be accessible,
141 partially accessible, freed or be a part of a redzone. KASAN uses the following
142 encoding for each shadow byte: 0 means that all 8 bytes of the corresponding
143 memory region are accessible; number N (1 <= N <= 7) means that the first N
144 bytes are accessible, and other (8 - N) bytes are not; any negative value
145 indicates that the entire 8-byte word is inaccessible. KASAN uses different
146 negative values to distinguish between different kinds of inaccessible memory
147 like redzones or freed memory (see mm/kasan/kasan.h).
149 In the report above the arrows point to the shadow byte 03, which means that
150 the accessed address is partially accessible. For tag-based KASAN modes this
151 last report section shows the memory tags around the accessed address
152 (see the `Implementation details`_ section).
157 Hardware tag-based KASAN mode (see the section about various modes below) is
158 intended for use in production as a security mitigation. Therefore, it supports
159 boot parameters that allow to disable KASAN competely or otherwise control
160 particular KASAN features.
162 - ``kasan=off`` or ``=on`` controls whether KASAN is enabled (default: ``on``).
164 - ``kasan.stacktrace=off`` or ``=on`` disables or enables alloc and free stack
165 traces collection (default: ``on``).
167 - ``kasan.fault=report`` or ``=panic`` controls whether to only print a KASAN
168 report or also panic the kernel (default: ``report``). Note, that tag
169 checking gets disabled after the first reported bug.
174 Software KASAN modes use compiler instrumentation to insert validity checks.
175 Such instrumentation might be incompatible with some part of the kernel, and
176 therefore needs to be disabled. To disable instrumentation for specific files
177 or directories, add a line similar to the following to the respective kernel
180 - For a single file (e.g. main.o)::
182 KASAN_SANITIZE_main.o := n
184 - For all files in one directory::
189 Implementation details
190 ----------------------
195 From a high level perspective, KASAN's approach to memory error detection is
196 similar to that of kmemcheck: use shadow memory to record whether each byte of
197 memory is safe to access, and use compile-time instrumentation to insert checks
198 of shadow memory on each memory access.
200 Generic KASAN dedicates 1/8th of kernel memory to its shadow memory (e.g. 16TB
201 to cover 128TB on x86_64) and uses direct mapping with a scale and offset to
202 translate a memory address to its corresponding shadow address.
204 Here is the function which translates an address to its corresponding shadow
207 static inline void *kasan_mem_to_shadow(const void *addr)
209 return ((unsigned long)addr >> KASAN_SHADOW_SCALE_SHIFT)
210 + KASAN_SHADOW_OFFSET;
213 where ``KASAN_SHADOW_SCALE_SHIFT = 3``.
215 Compile-time instrumentation is used to insert memory access checks. Compiler
216 inserts function calls (__asan_load*(addr), __asan_store*(addr)) before each
217 memory access of size 1, 2, 4, 8 or 16. These functions check whether memory
218 access is valid or not by checking corresponding shadow memory.
220 GCC 5.0 has possibility to perform inline instrumentation. Instead of making
221 function calls GCC directly inserts the code to check the shadow memory.
222 This option significantly enlarges kernel but it gives x1.1-x2 performance
223 boost over outline instrumented kernel.
225 Generic KASAN also reports the last 2 call stacks to creation of work that
226 potentially has access to an object. Call stacks for the following are shown:
227 call_rcu() and workqueue queuing.
229 Generic KASAN is the only mode that delays the reuse of freed object via
230 quarantine (see mm/kasan/quarantine.c for implementation).
232 Software tag-based KASAN
233 ~~~~~~~~~~~~~~~~~~~~~~~~
235 Software tag-based KASAN requires software memory tagging support in the form
236 of HWASan-like compiler instrumentation (see HWASan documentation for details).
238 Software tag-based KASAN is currently only implemented for arm64 architecture.
240 Software tag-based KASAN uses the Top Byte Ignore (TBI) feature of arm64 CPUs
241 to store a pointer tag in the top byte of kernel pointers. Like generic KASAN
242 it uses shadow memory to store memory tags associated with each 16-byte memory
243 cell (therefore it dedicates 1/16th of the kernel memory for shadow memory).
245 On each memory allocation software tag-based KASAN generates a random tag, tags
246 the allocated memory with this tag, and embeds this tag into the returned
249 Software tag-based KASAN uses compile-time instrumentation to insert checks
250 before each memory access. These checks make sure that tag of the memory that
251 is being accessed is equal to tag of the pointer that is used to access this
252 memory. In case of a tag mismatch software tag-based KASAN prints a bug report.
254 Software tag-based KASAN also has two instrumentation modes (outline, that
255 emits callbacks to check memory accesses; and inline, that performs the shadow
256 memory checks inline). With outline instrumentation mode, a bug report is
257 simply printed from the function that performs the access check. With inline
258 instrumentation a brk instruction is emitted by the compiler, and a dedicated
259 brk handler is used to print bug reports.
261 Software tag-based KASAN uses 0xFF as a match-all pointer tag (accesses through
262 pointers with 0xFF pointer tag aren't checked). The value 0xFE is currently
263 reserved to tag freed memory regions.
265 Software tag-based KASAN currently only supports tagging of
266 kmem_cache_alloc/kmalloc and page_alloc memory.
268 Hardware tag-based KASAN
269 ~~~~~~~~~~~~~~~~~~~~~~~~
271 Hardware tag-based KASAN is similar to the software mode in concept, but uses
272 hardware memory tagging support instead of compiler instrumentation and
275 Hardware tag-based KASAN is currently only implemented for arm64 architecture
276 and based on both arm64 Memory Tagging Extension (MTE) introduced in ARMv8.5
277 Instruction Set Architecture, and Top Byte Ignore (TBI).
279 Special arm64 instructions are used to assign memory tags for each allocation.
280 Same tags are assigned to pointers to those allocations. On every memory
281 access, hardware makes sure that tag of the memory that is being accessed is
282 equal to tag of the pointer that is used to access this memory. In case of a
283 tag mismatch a fault is generated and a report is printed.
285 Hardware tag-based KASAN uses 0xFF as a match-all pointer tag (accesses through
286 pointers with 0xFF pointer tag aren't checked). The value 0xFE is currently
287 reserved to tag freed memory regions.
289 Hardware tag-based KASAN currently only supports tagging of
290 kmem_cache_alloc/kmalloc and page_alloc memory.
292 If the hardware doesn't support MTE (pre ARMv8.5), hardware tag-based KASAN
293 won't be enabled. In this case all boot parameters are ignored.
295 Note, that enabling CONFIG_KASAN_HW_TAGS always results in in-kernel TBI being
296 enabled. Even when kasan.mode=off is provided, or when the hardware doesn't
297 support MTE (but supports TBI).
299 Hardware tag-based KASAN only reports the first found bug. After that MTE tag
300 checking gets disabled.
302 What memory accesses are sanitised by KASAN?
303 --------------------------------------------
305 The kernel maps memory in a number of different parts of the address
306 space. This poses something of a problem for KASAN, which requires
307 that all addresses accessed by instrumented code have a valid shadow
310 The range of kernel virtual addresses is large: there is not enough
311 real memory to support a real shadow region for every address that
312 could be accessed by the kernel.
317 By default, architectures only map real memory over the shadow region
318 for the linear mapping (and potentially other small areas). For all
319 other areas - such as vmalloc and vmemmap space - a single read-only
320 page is mapped over the shadow area. This read-only shadow page
321 declares all memory accesses as permitted.
323 This presents a problem for modules: they do not live in the linear
324 mapping, but in a dedicated module space. By hooking in to the module
325 allocator, KASAN can temporarily map real shadow memory to cover
326 them. This allows detection of invalid accesses to module globals, for
329 This also creates an incompatibility with ``VMAP_STACK``: if the stack
330 lives in vmalloc space, it will be shadowed by the read-only page, and
331 the kernel will fault when trying to set up the shadow data for stack
337 With ``CONFIG_KASAN_VMALLOC``, KASAN can cover vmalloc space at the
338 cost of greater memory usage. Currently this is only supported on x86.
340 This works by hooking into vmalloc and vmap, and dynamically
341 allocating real shadow memory to back the mappings.
343 Most mappings in vmalloc space are small, requiring less than a full
344 page of shadow space. Allocating a full shadow page per mapping would
345 therefore be wasteful. Furthermore, to ensure that different mappings
346 use different shadow pages, mappings would have to be aligned to
347 ``KASAN_GRANULE_SIZE * PAGE_SIZE``.
349 Instead, KASAN shares backing space across multiple mappings. It allocates
350 a backing page when a mapping in vmalloc space uses a particular page
351 of the shadow region. This page can be shared by other vmalloc
354 KASAN hooks into the vmap infrastructure to lazily clean up unused shadow
357 To avoid the difficulties around swapping mappings around, KASAN expects
358 that the part of the shadow region that covers the vmalloc space will
359 not be covered by the early shadow page, but will be left
360 unmapped. This will require changes in arch-specific code.
362 This allows ``VMAP_STACK`` support on x86, and can simplify support of
363 architectures that do not have a fixed module region.
365 CONFIG_KASAN_KUNIT_TEST and CONFIG_KASAN_MODULE_TEST
366 ----------------------------------------------------
368 KASAN tests consist of two parts:
370 1. Tests that are integrated with the KUnit Test Framework. Enabled with
371 ``CONFIG_KASAN_KUNIT_TEST``. These tests can be run and partially verified
372 automatically in a few different ways, see the instructions below.
374 2. Tests that are currently incompatible with KUnit. Enabled with
375 ``CONFIG_KASAN_MODULE_TEST`` and can only be run as a module. These tests can
376 only be verified manually, by loading the kernel module and inspecting the
377 kernel log for KASAN reports.
379 Each KUnit-compatible KASAN test prints a KASAN report if an error is detected.
380 Then the test prints its number and status.
384 ok 28 - kmalloc_double_kzfree
386 When a test fails due to a failed ``kmalloc``::
388 # kmalloc_large_oob_right: ASSERTION FAILED at lib/test_kasan.c:163
389 Expected ptr is not null, but is
390 not ok 4 - kmalloc_large_oob_right
392 When a test fails due to a missing KASAN report::
394 # kmalloc_double_kzfree: EXPECTATION FAILED at lib/test_kasan.c:629
395 Expected kasan_data->report_expected == kasan_data->report_found, but
396 kasan_data->report_expected == 1
397 kasan_data->report_found == 0
398 not ok 28 - kmalloc_double_kzfree
400 At the end the cumulative status of all KASAN tests is printed. On success::
404 Or, if one of the tests failed::
409 There are a few ways to run KUnit-compatible KASAN tests.
414 With ``CONFIG_KUNIT`` enabled, ``CONFIG_KASAN_KUNIT_TEST`` can be built as
415 a loadable module and run on any architecture that supports KASAN by loading
416 the module with insmod or modprobe. The module is called ``test_kasan``.
421 With ``CONFIG_KUNIT`` built-in, ``CONFIG_KASAN_KUNIT_TEST`` can be built-in
422 on any architecure that supports KASAN. These and any other KUnit tests enabled
423 will run and print the results at boot as a late-init call.
428 With ``CONFIG_KUNIT`` and ``CONFIG_KASAN_KUNIT_TEST`` built-in, it's also
429 possible use ``kunit_tool`` to see the results of these and other KUnit tests
430 in a more readable way. This will not print the KASAN reports of the tests that
431 passed. Use `KUnit documentation <https://www.kernel.org/doc/html/latest/dev-tools/kunit/index.html>`_
432 for more up-to-date information on ``kunit_tool``.
434 .. _KUnit: https://www.kernel.org/doc/html/latest/dev-tools/kunit/index.html