1The Kernel Address Sanitizer (KASAN)
2====================================
3
4Overview
5--------
6
7Kernel Address Sanitizer (KASAN) is a dynamic memory safety error detector
8designed to find out-of-bounds and use-after-free bugs.
9
10KASAN has three modes:
11
121. Generic KASAN
132. Software Tag-Based KASAN
143. Hardware Tag-Based KASAN
15
16Generic KASAN, enabled with CONFIG_KASAN_GENERIC, is the mode intended for
17debugging, similar to userspace ASan. This mode is supported on many CPU
18architectures, but it has significant performance and memory overheads.
19
20Software Tag-Based KASAN or SW_TAGS KASAN, enabled with CONFIG_KASAN_SW_TAGS,
21can be used for both debugging and dogfood testing, similar to userspace HWASan.
22This mode is only supported for arm64, but its moderate memory overhead allows
23using it for testing on memory-restricted devices with real workloads.
24
25Hardware Tag-Based KASAN or HW_TAGS KASAN, enabled with CONFIG_KASAN_HW_TAGS,
26is the mode intended to be used as an in-field memory bug detector or as a
27security mitigation. This mode only works on arm64 CPUs that support MTE
28(Memory Tagging Extension), but it has low memory and performance overheads and
29thus can be used in production.
30
31For details about the memory and performance impact of each KASAN mode, see the
32descriptions of the corresponding Kconfig options.
33
34The Generic and the Software Tag-Based modes are commonly referred to as the
35software modes. The Software Tag-Based and the Hardware Tag-Based modes are
36referred to as the tag-based modes.
37
38Support
39-------
40
41Architectures
42~~~~~~~~~~~~~
43
44Generic KASAN is supported on x86_64, arm, arm64, powerpc, riscv, s390, and
45xtensa, and the tag-based KASAN modes are supported only on arm64.
46
47Compilers
48~~~~~~~~~
49
50Software KASAN modes use compile-time instrumentation to insert validity checks
51before every memory access and thus require a compiler version that provides
52support for that. The Hardware Tag-Based mode relies on hardware to perform
53these checks but still requires a compiler version that supports the memory
54tagging instructions.
55
56Generic KASAN requires GCC version 8.3.0 or later
57or any Clang version supported by the kernel.
58
59Software Tag-Based KASAN requires GCC 11+
60or any Clang version supported by the kernel.
61
62Hardware Tag-Based KASAN requires GCC 10+ or Clang 12+.
63
64Memory types
65~~~~~~~~~~~~
66
67Generic KASAN supports finding bugs in all of slab, page_alloc, vmap, vmalloc,
68stack, and global memory.
69
70Software Tag-Based KASAN supports slab, page_alloc, vmalloc, and stack memory.
71
72Hardware Tag-Based KASAN supports slab, page_alloc, and non-executable vmalloc
73memory.
74
75For slab, both software KASAN modes support SLUB and SLAB allocators, while
76Hardware Tag-Based KASAN only supports SLUB.
77
78Usage
79-----
80
81To enable KASAN, configure the kernel with::
82
83	  CONFIG_KASAN=y
84
85and 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).
88
89For the software modes, also choose between ``CONFIG_KASAN_OUTLINE`` and
90``CONFIG_KASAN_INLINE``. Outline and inline are compiler instrumentation types.
91The former produces a smaller binary while the latter is up to 2 times faster.
92
93To include alloc and free stack traces of affected slab objects into reports,
94enable ``CONFIG_STACKTRACE``. To include alloc and free stack traces of affected
95physical pages, enable ``CONFIG_PAGE_OWNER`` and boot with ``page_owner=on``.
96
97Boot parameters
98~~~~~~~~~~~~~~~
99
100KASAN is affected by the generic ``panic_on_warn`` command line parameter.
101When it is enabled, KASAN panics the kernel after printing a bug report.
102
103By default, KASAN prints a bug report only for the first invalid memory access.
104With ``kasan_multi_shot``, KASAN prints a report on every invalid access. This
105effectively disables ``panic_on_warn`` for KASAN reports.
106
107Alternatively, independent of ``panic_on_warn``, the ``kasan.fault=`` boot
108parameter can be used to control panic and reporting behaviour:
109
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.
113
114Software and Hardware Tag-Based KASAN modes (see the section about various
115modes below) support altering stack trace collection behavior:
116
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``).
121
122Hardware Tag-Based KASAN mode is intended for use in production as a security
123mitigation. Therefore, it supports additional boot parameters that allow
124disabling KASAN altogether or controlling its features:
125
126- ``kasan=off`` or ``=on`` controls whether KASAN is enabled (default: ``on``).
127
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
132  check fault occurs.
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.
139
140- ``kasan.vmalloc=off`` or ``=on`` disables or enables tagging of vmalloc
141  allocations (default: ``on``).
142
143Error reports
144~~~~~~~~~~~~~
145
146A typical KASAN report looks like this::
147
148    ==================================================================
149    BUG: KASAN: slab-out-of-bounds in kmalloc_oob_right+0xa8/0xbc [test_kasan]
150    Write of size 1 at addr ffff8801f44ec37b by task insmod/2760
151
152    CPU: 1 PID: 2760 Comm: insmod Not tainted 4.19.0-rc3+ #698
153    Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014
154    Call Trace:
155     dump_stack+0x94/0xd8
156     print_address_description+0x73/0x280
157     kasan_report+0x144/0x187
158     __asan_report_store1_noabort+0x17/0x20
159     kmalloc_oob_right+0xa8/0xbc [test_kasan]
160     kmalloc_tests_init+0x16/0x700 [test_kasan]
161     do_one_initcall+0xa5/0x3ae
162     do_init_module+0x1b6/0x547
163     load_module+0x75df/0x8070
164     __do_sys_init_module+0x1c6/0x200
165     __x64_sys_init_module+0x6e/0xb0
166     do_syscall_64+0x9f/0x2c0
167     entry_SYSCALL_64_after_hwframe+0x44/0xa9
168    RIP: 0033:0x7f96443109da
169    RSP: 002b:00007ffcf0b51b08 EFLAGS: 00000202 ORIG_RAX: 00000000000000af
170    RAX: ffffffffffffffda RBX: 000055dc3ee521a0 RCX: 00007f96443109da
171    RDX: 00007f96445cff88 RSI: 0000000000057a50 RDI: 00007f9644992000
172    RBP: 000055dc3ee510b0 R08: 0000000000000003 R09: 0000000000000000
173    R10: 00007f964430cd0a R11: 0000000000000202 R12: 00007f96445cff88
174    R13: 000055dc3ee51090 R14: 0000000000000000 R15: 0000000000000000
175
176    Allocated by task 2760:
177     save_stack+0x43/0xd0
178     kasan_kmalloc+0xa7/0xd0
179     kmem_cache_alloc_trace+0xe1/0x1b0
180     kmalloc_oob_right+0x56/0xbc [test_kasan]
181     kmalloc_tests_init+0x16/0x700 [test_kasan]
182     do_one_initcall+0xa5/0x3ae
183     do_init_module+0x1b6/0x547
184     load_module+0x75df/0x8070
185     __do_sys_init_module+0x1c6/0x200
186     __x64_sys_init_module+0x6e/0xb0
187     do_syscall_64+0x9f/0x2c0
188     entry_SYSCALL_64_after_hwframe+0x44/0xa9
189
190    Freed by task 815:
191     save_stack+0x43/0xd0
192     __kasan_slab_free+0x135/0x190
193     kasan_slab_free+0xe/0x10
194     kfree+0x93/0x1a0
195     umh_complete+0x6a/0xa0
196     call_usermodehelper_exec_async+0x4c3/0x640
197     ret_from_fork+0x35/0x40
198
199    The buggy address belongs to the object at ffff8801f44ec300
200     which belongs to the cache kmalloc-128 of size 128
201    The buggy address is located 123 bytes inside of
202     128-byte region [ffff8801f44ec300, ffff8801f44ec380)
203    The buggy address belongs to the page:
204    page:ffffea0007d13b00 count:1 mapcount:0 mapping:ffff8801f7001640 index:0x0
205    flags: 0x200000000000100(slab)
206    raw: 0200000000000100 ffffea0007d11dc0 0000001a0000001a ffff8801f7001640
207    raw: 0000000000000000 0000000080150015 00000001ffffffff 0000000000000000
208    page dumped because: kasan: bad access detected
209
210    Memory state around the buggy address:
211     ffff8801f44ec200: fc fc fc fc fc fc fc fc fb fb fb fb fb fb fb fb
212     ffff8801f44ec280: fb fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
213    >ffff8801f44ec300: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 03
214                                                                    ^
215     ffff8801f44ec380: fc fc fc fc fc fc fc fc fb fb fb fb fb fb fb fb
216     ffff8801f44ec400: fb fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
217    ==================================================================
218
219The report header summarizes what kind of bug happened and what kind of access
220caused it. It is followed by a stack trace of the bad access, a stack trace of
221where the accessed memory was allocated (in case a slab object was accessed),
222and a stack trace of where the object was freed (in case of a use-after-free
223bug report). Next comes a description of the accessed slab object and the
224information about the accessed memory page.
225
226In the end, the report shows the memory state around the accessed address.
227Internally, KASAN tracks memory state separately for each memory granule, which
228is either 8 or 16 aligned bytes depending on KASAN mode. Each number in the
229memory state section of the report shows the state of one of the memory
230granules that surround the accessed address.
231
232For Generic KASAN, the size of each memory granule is 8. The state of each
233granule is encoded in one shadow byte. Those 8 bytes can be accessible,
234partially accessible, freed, or be a part of a redzone. KASAN uses the following
235encoding for each shadow byte: 00 means that all 8 bytes of the corresponding
236memory region are accessible; number N (1 <= N <= 7) means that the first N
237bytes are accessible, and other (8 - N) bytes are not; any negative value
238indicates that the entire 8-byte word is inaccessible. KASAN uses different
239negative values to distinguish between different kinds of inaccessible memory
240like redzones or freed memory (see mm/kasan/kasan.h).
241
242In the report above, the arrow points to the shadow byte ``03``, which means
243that the accessed address is partially accessible.
244
245For tag-based KASAN modes, this last report section shows the memory tags around
246the accessed address (see the `Implementation details`_ section).
247
248Note that KASAN bug titles (like ``slab-out-of-bounds`` or ``use-after-free``)
249are best-effort: KASAN prints the most probable bug type based on the limited
250information it has. The actual type of the bug might be different.
251
252Generic KASAN also reports up to two auxiliary call stack traces. These stack
253traces point to places in code that interacted with the object but that are not
254directly present in the bad access stack trace. Currently, this includes
255call_rcu() and workqueue queuing.
256
257Implementation details
258----------------------
259
260Generic KASAN
261~~~~~~~~~~~~~
262
263Software KASAN modes use shadow memory to record whether each byte of memory is
264safe to access and use compile-time instrumentation to insert shadow memory
265checks before each memory access.
266
267Generic KASAN dedicates 1/8th of kernel memory to its shadow memory (16TB
268to cover 128TB on x86_64) and uses direct mapping with a scale and offset to
269translate a memory address to its corresponding shadow address.
270
271Here is the function which translates an address to its corresponding shadow
272address::
273
274    static inline void *kasan_mem_to_shadow(const void *addr)
275    {
276	return (void *)((unsigned long)addr >> KASAN_SHADOW_SCALE_SHIFT)
277		+ KASAN_SHADOW_OFFSET;
278    }
279
280where ``KASAN_SHADOW_SCALE_SHIFT = 3``.
281
282Compile-time instrumentation is used to insert memory access checks. Compiler
283inserts function calls (``__asan_load*(addr)``, ``__asan_store*(addr)``) before
284each memory access of size 1, 2, 4, 8, or 16. These functions check whether
285memory accesses are valid or not by checking corresponding shadow memory.
286
287With inline instrumentation, instead of making function calls, the compiler
288directly inserts the code to check shadow memory. This option significantly
289enlarges the kernel, but it gives an x1.1-x2 performance boost over the
290outline-instrumented kernel.
291
292Generic KASAN is the only mode that delays the reuse of freed objects via
293quarantine (see mm/kasan/quarantine.c for implementation).
294
295Software Tag-Based KASAN
296~~~~~~~~~~~~~~~~~~~~~~~~
297
298Software Tag-Based KASAN uses a software memory tagging approach to checking
299access validity. It is currently only implemented for the arm64 architecture.
300
301Software Tag-Based KASAN uses the Top Byte Ignore (TBI) feature of arm64 CPUs
302to store a pointer tag in the top byte of kernel pointers. It uses shadow memory
303to store memory tags associated with each 16-byte memory cell (therefore, it
304dedicates 1/16th of the kernel memory for shadow memory).
305
306On each memory allocation, Software Tag-Based KASAN generates a random tag, tags
307the allocated memory with this tag, and embeds the same tag into the returned
308pointer.
309
310Software Tag-Based KASAN uses compile-time instrumentation to insert checks
311before each memory access. These checks make sure that the tag of the memory
312that is being accessed is equal to the tag of the pointer that is used to access
313this memory. In case of a tag mismatch, Software Tag-Based KASAN prints a bug
314report.
315
316Software Tag-Based KASAN also has two instrumentation modes (outline, which
317emits callbacks to check memory accesses; and inline, which performs the shadow
318memory checks inline). With outline instrumentation mode, a bug report is
319printed from the function that performs the access check. With inline
320instrumentation, a ``brk`` instruction is emitted by the compiler, and a
321dedicated ``brk`` handler is used to print bug reports.
322
323Software Tag-Based KASAN uses 0xFF as a match-all pointer tag (accesses through
324pointers with the 0xFF pointer tag are not checked). The value 0xFE is currently
325reserved to tag freed memory regions.
326
327Hardware Tag-Based KASAN
328~~~~~~~~~~~~~~~~~~~~~~~~
329
330Hardware Tag-Based KASAN is similar to the software mode in concept but uses
331hardware memory tagging support instead of compiler instrumentation and
332shadow memory.
333
334Hardware Tag-Based KASAN is currently only implemented for arm64 architecture
335and based on both arm64 Memory Tagging Extension (MTE) introduced in ARMv8.5
336Instruction Set Architecture and Top Byte Ignore (TBI).
337
338Special arm64 instructions are used to assign memory tags for each allocation.
339Same tags are assigned to pointers to those allocations. On every memory
340access, hardware makes sure that the tag of the memory that is being accessed is
341equal to the tag of the pointer that is used to access this memory. In case of a
342tag mismatch, a fault is generated, and a report is printed.
343
344Hardware Tag-Based KASAN uses 0xFF as a match-all pointer tag (accesses through
345pointers with the 0xFF pointer tag are not checked). The value 0xFE is currently
346reserved to tag freed memory regions.
347
348If the hardware does not support MTE (pre ARMv8.5), Hardware Tag-Based KASAN
349will not be enabled. In this case, all KASAN boot parameters are ignored.
350
351Note that enabling CONFIG_KASAN_HW_TAGS always results in in-kernel TBI being
352enabled. Even when ``kasan.mode=off`` is provided or when the hardware does not
353support MTE (but supports TBI).
354
355Hardware Tag-Based KASAN only reports the first found bug. After that, MTE tag
356checking gets disabled.
357
358Shadow memory
359-------------
360
361The contents of this section are only applicable to software KASAN modes.
362
363The kernel maps memory in several different parts of the address space.
364The range of kernel virtual addresses is large: there is not enough real
365memory to support a real shadow region for every address that could be
366accessed by the kernel. Therefore, KASAN only maps real shadow for certain
367parts of the address space.
368
369Default behaviour
370~~~~~~~~~~~~~~~~~
371
372By default, architectures only map real memory over the shadow region
373for the linear mapping (and potentially other small areas). For all
374other areas - such as vmalloc and vmemmap space - a single read-only
375page is mapped over the shadow area. This read-only shadow page
376declares all memory accesses as permitted.
377
378This presents a problem for modules: they do not live in the linear
379mapping but in a dedicated module space. By hooking into the module
380allocator, KASAN temporarily maps real shadow memory to cover them.
381This allows detection of invalid accesses to module globals, for example.
382
383This also creates an incompatibility with ``VMAP_STACK``: if the stack
384lives in vmalloc space, it will be shadowed by the read-only page, and
385the kernel will fault when trying to set up the shadow data for stack
386variables.
387
388CONFIG_KASAN_VMALLOC
389~~~~~~~~~~~~~~~~~~~~
390
391With ``CONFIG_KASAN_VMALLOC``, KASAN can cover vmalloc space at the
392cost of greater memory usage. Currently, this is supported on x86,
393arm64, riscv, s390, and powerpc.
394
395This works by hooking into vmalloc and vmap and dynamically
396allocating real shadow memory to back the mappings.
397
398Most mappings in vmalloc space are small, requiring less than a full
399page of shadow space. Allocating a full shadow page per mapping would
400therefore be wasteful. Furthermore, to ensure that different mappings
401use different shadow pages, mappings would have to be aligned to
402``KASAN_GRANULE_SIZE * PAGE_SIZE``.
403
404Instead, KASAN shares backing space across multiple mappings. It allocates
405a backing page when a mapping in vmalloc space uses a particular page
406of the shadow region. This page can be shared by other vmalloc
407mappings later on.
408
409KASAN hooks into the vmap infrastructure to lazily clean up unused shadow
410memory.
411
412To avoid the difficulties around swapping mappings around, KASAN expects
413that the part of the shadow region that covers the vmalloc space will
414not be covered by the early shadow page but will be left unmapped.
415This will require changes in arch-specific code.
416
417This allows ``VMAP_STACK`` support on x86 and can simplify support of
418architectures that do not have a fixed module region.
419
420For developers
421--------------
422
423Ignoring accesses
424~~~~~~~~~~~~~~~~~
425
426Software KASAN modes use compiler instrumentation to insert validity checks.
427Such instrumentation might be incompatible with some parts of the kernel, and
428therefore needs to be disabled.
429
430Other parts of the kernel might access metadata for allocated objects.
431Normally, KASAN detects and reports such accesses, but in some cases (e.g.,
432in memory allocators), these accesses are valid.
433
434For software KASAN modes, to disable instrumentation for a specific file or
435directory, add a ``KASAN_SANITIZE`` annotation to the respective kernel
436Makefile:
437
438- For a single file (e.g., main.o)::
439
440    KASAN_SANITIZE_main.o := n
441
442- For all files in one directory::
443
444    KASAN_SANITIZE := n
445
446For software KASAN modes, to disable instrumentation on a per-function basis,
447use the KASAN-specific ``__no_sanitize_address`` function attribute or the
448generic ``noinstr`` one.
449
450Note that disabling compiler instrumentation (either on a per-file or a
451per-function basis) makes KASAN ignore the accesses that happen directly in
452that code for software KASAN modes. It does not help when the accesses happen
453indirectly (through calls to instrumented functions) or with Hardware
454Tag-Based KASAN, which does not use compiler instrumentation.
455
456For software KASAN modes, to disable KASAN reports in a part of the kernel code
457for the current task, annotate this part of the code with a
458``kasan_disable_current()``/``kasan_enable_current()`` section. This also
459disables the reports for indirect accesses that happen through function calls.
460
461For tag-based KASAN modes, to disable access checking, use
462``kasan_reset_tag()`` or ``page_kasan_tag_reset()``. Note that temporarily
463disabling access checking via ``page_kasan_tag_reset()`` requires saving and
464restoring the per-page KASAN tag via ``page_kasan_tag``/``page_kasan_tag_set``.
465
466Tests
467~~~~~
468
469There are KASAN tests that allow verifying that KASAN works and can detect
470certain types of memory corruptions. The tests consist of two parts:
471
4721. Tests that are integrated with the KUnit Test Framework. Enabled with
473``CONFIG_KASAN_KUNIT_TEST``. These tests can be run and partially verified
474automatically in a few different ways; see the instructions below.
475
4762. Tests that are currently incompatible with KUnit. Enabled with
477``CONFIG_KASAN_MODULE_TEST`` and can only be run as a module. These tests can
478only be verified manually by loading the kernel module and inspecting the
479kernel log for KASAN reports.
480
481Each KUnit-compatible KASAN test prints one of multiple KASAN reports if an
482error is detected. Then the test prints its number and status.
483
484When a test passes::
485
486        ok 28 - kmalloc_double_kzfree
487
488When a test fails due to a failed ``kmalloc``::
489
490        # kmalloc_large_oob_right: ASSERTION FAILED at lib/test_kasan.c:163
491        Expected ptr is not null, but is
492        not ok 4 - kmalloc_large_oob_right
493
494When a test fails due to a missing KASAN report::
495
496        # kmalloc_double_kzfree: EXPECTATION FAILED at lib/test_kasan.c:974
497        KASAN failure expected in "kfree_sensitive(ptr)", but none occurred
498        not ok 44 - kmalloc_double_kzfree
499
500
501At the end the cumulative status of all KASAN tests is printed. On success::
502
503        ok 1 - kasan
504
505Or, if one of the tests failed::
506
507        not ok 1 - kasan
508
509There are a few ways to run KUnit-compatible KASAN tests.
510
5111. Loadable module
512
513   With ``CONFIG_KUNIT`` enabled, KASAN-KUnit tests can be built as a loadable
514   module and run by loading ``test_kasan.ko`` with ``insmod`` or ``modprobe``.
515
5162. Built-In
517
518   With ``CONFIG_KUNIT`` built-in, KASAN-KUnit tests can be built-in as well.
519   In this case, the tests will run at boot as a late-init call.
520
5213. Using kunit_tool
522
523   With ``CONFIG_KUNIT`` and ``CONFIG_KASAN_KUNIT_TEST`` built-in, it is also
524   possible to use ``kunit_tool`` to see the results of KUnit tests in a more
525   readable way. This will not print the KASAN reports of the tests that passed.
526   See `KUnit documentation <https://www.kernel.org/doc/html/latest/dev-tools/kunit/index.html>`_
527   for more up-to-date information on ``kunit_tool``.
528
529.. _KUnit: https://www.kernel.org/doc/html/latest/dev-tools/kunit/index.html
530