Andrey Ryabinin | 0b24bec | 2015-02-13 14:39:17 -0800 | [diff] [blame] | 1 | Kernel address sanitizer |
| 2 | ================ |
| 3 | |
| 4 | 0. Overview |
| 5 | =========== |
| 6 | |
| 7 | Kernel Address sanitizer (KASan) is a dynamic memory error detector. It provides |
| 8 | a fast and comprehensive solution for finding use-after-free and out-of-bounds |
| 9 | bugs. |
| 10 | |
| 11 | KASan uses compile-time instrumentation for checking every memory access, |
Andrey Ryabinin | bebf56a | 2015-02-13 14:40:17 -0800 | [diff] [blame] | 12 | therefore you will need a certain version of GCC > 4.9.2 |
Andrey Ryabinin | 0b24bec | 2015-02-13 14:39:17 -0800 | [diff] [blame] | 13 | |
| 14 | Currently KASan is supported only for x86_64 architecture and requires that the |
| 15 | kernel be built with the SLUB allocator. |
| 16 | |
| 17 | 1. Usage |
| 18 | ========= |
| 19 | |
| 20 | To enable KASAN configure kernel with: |
| 21 | |
| 22 | CONFIG_KASAN = y |
| 23 | |
| 24 | and choose between CONFIG_KASAN_OUTLINE and CONFIG_KASAN_INLINE. Outline/inline |
| 25 | is compiler instrumentation types. The former produces smaller binary the |
| 26 | latter is 1.1 - 2 times faster. Inline instrumentation requires GCC 5.0 or |
| 27 | latter. |
| 28 | |
| 29 | Currently KASAN works only with the SLUB memory allocator. |
| 30 | For better bug detection and nicer report, enable CONFIG_STACKTRACE and put |
| 31 | at least 'slub_debug=U' in the boot cmdline. |
| 32 | |
| 33 | To disable instrumentation for specific files or directories, add a line |
| 34 | similar to the following to the respective kernel Makefile: |
| 35 | |
| 36 | For a single file (e.g. main.o): |
| 37 | KASAN_SANITIZE_main.o := n |
| 38 | |
| 39 | For all files in one directory: |
| 40 | KASAN_SANITIZE := n |
| 41 | |
| 42 | 1.1 Error reports |
| 43 | ========== |
| 44 | |
| 45 | A typical out of bounds access report looks like this: |
| 46 | |
| 47 | ================================================================== |
| 48 | BUG: AddressSanitizer: out of bounds access in kmalloc_oob_right+0x65/0x75 [test_kasan] at addr ffff8800693bc5d3 |
| 49 | Write of size 1 by task modprobe/1689 |
| 50 | ============================================================================= |
| 51 | BUG kmalloc-128 (Not tainted): kasan error |
| 52 | ----------------------------------------------------------------------------- |
| 53 | |
| 54 | Disabling lock debugging due to kernel taint |
| 55 | INFO: Allocated in kmalloc_oob_right+0x3d/0x75 [test_kasan] age=0 cpu=0 pid=1689 |
| 56 | __slab_alloc+0x4b4/0x4f0 |
| 57 | kmem_cache_alloc_trace+0x10b/0x190 |
| 58 | kmalloc_oob_right+0x3d/0x75 [test_kasan] |
| 59 | init_module+0x9/0x47 [test_kasan] |
| 60 | do_one_initcall+0x99/0x200 |
| 61 | load_module+0x2cb3/0x3b20 |
| 62 | SyS_finit_module+0x76/0x80 |
| 63 | system_call_fastpath+0x12/0x17 |
| 64 | INFO: Slab 0xffffea0001a4ef00 objects=17 used=7 fp=0xffff8800693bd728 flags=0x100000000004080 |
| 65 | INFO: Object 0xffff8800693bc558 @offset=1368 fp=0xffff8800693bc720 |
| 66 | |
| 67 | Bytes b4 ffff8800693bc548: 00 00 00 00 00 00 00 00 5a 5a 5a 5a 5a 5a 5a 5a ........ZZZZZZZZ |
| 68 | Object ffff8800693bc558: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk |
| 69 | Object ffff8800693bc568: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk |
| 70 | Object ffff8800693bc578: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk |
| 71 | Object ffff8800693bc588: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk |
| 72 | Object ffff8800693bc598: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk |
| 73 | Object ffff8800693bc5a8: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk |
| 74 | Object ffff8800693bc5b8: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk |
| 75 | Object ffff8800693bc5c8: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b a5 kkkkkkkkkkkkkkk. |
| 76 | Redzone ffff8800693bc5d8: cc cc cc cc cc cc cc cc ........ |
| 77 | Padding ffff8800693bc718: 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZ |
| 78 | CPU: 0 PID: 1689 Comm: modprobe Tainted: G B 3.18.0-rc1-mm1+ #98 |
| 79 | Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.7.5-0-ge51488c-20140602_164612-nilsson.home.kraxel.org 04/01/2014 |
| 80 | ffff8800693bc000 0000000000000000 ffff8800693bc558 ffff88006923bb78 |
| 81 | ffffffff81cc68ae 00000000000000f3 ffff88006d407600 ffff88006923bba8 |
| 82 | ffffffff811fd848 ffff88006d407600 ffffea0001a4ef00 ffff8800693bc558 |
| 83 | Call Trace: |
| 84 | [<ffffffff81cc68ae>] dump_stack+0x46/0x58 |
| 85 | [<ffffffff811fd848>] print_trailer+0xf8/0x160 |
| 86 | [<ffffffffa00026a7>] ? kmem_cache_oob+0xc3/0xc3 [test_kasan] |
| 87 | [<ffffffff811ff0f5>] object_err+0x35/0x40 |
| 88 | [<ffffffffa0002065>] ? kmalloc_oob_right+0x65/0x75 [test_kasan] |
| 89 | [<ffffffff8120b9fa>] kasan_report_error+0x38a/0x3f0 |
| 90 | [<ffffffff8120a79f>] ? kasan_poison_shadow+0x2f/0x40 |
| 91 | [<ffffffff8120b344>] ? kasan_unpoison_shadow+0x14/0x40 |
| 92 | [<ffffffff8120a79f>] ? kasan_poison_shadow+0x2f/0x40 |
| 93 | [<ffffffffa00026a7>] ? kmem_cache_oob+0xc3/0xc3 [test_kasan] |
| 94 | [<ffffffff8120a995>] __asan_store1+0x75/0xb0 |
| 95 | [<ffffffffa0002601>] ? kmem_cache_oob+0x1d/0xc3 [test_kasan] |
| 96 | [<ffffffffa0002065>] ? kmalloc_oob_right+0x65/0x75 [test_kasan] |
| 97 | [<ffffffffa0002065>] kmalloc_oob_right+0x65/0x75 [test_kasan] |
| 98 | [<ffffffffa00026b0>] init_module+0x9/0x47 [test_kasan] |
| 99 | [<ffffffff810002d9>] do_one_initcall+0x99/0x200 |
| 100 | [<ffffffff811e4e5c>] ? __vunmap+0xec/0x160 |
| 101 | [<ffffffff81114f63>] load_module+0x2cb3/0x3b20 |
| 102 | [<ffffffff8110fd70>] ? m_show+0x240/0x240 |
| 103 | [<ffffffff81115f06>] SyS_finit_module+0x76/0x80 |
| 104 | [<ffffffff81cd3129>] system_call_fastpath+0x12/0x17 |
| 105 | Memory state around the buggy address: |
| 106 | ffff8800693bc300: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc |
| 107 | ffff8800693bc380: fc fc 00 00 00 00 00 00 00 00 00 00 00 00 00 fc |
| 108 | ffff8800693bc400: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc |
| 109 | ffff8800693bc480: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc |
| 110 | ffff8800693bc500: fc fc fc fc fc fc fc fc fc fc fc 00 00 00 00 00 |
| 111 | >ffff8800693bc580: 00 00 00 00 00 00 00 00 00 00 03 fc fc fc fc fc |
| 112 | ^ |
| 113 | ffff8800693bc600: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc |
| 114 | ffff8800693bc680: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc |
| 115 | ffff8800693bc700: fc fc fc fc fb fb fb fb fb fb fb fb fb fb fb fb |
| 116 | ffff8800693bc780: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb |
| 117 | ffff8800693bc800: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb |
| 118 | ================================================================== |
| 119 | |
| 120 | First sections describe slub object where bad access happened. |
| 121 | See 'SLUB Debug output' section in Documentation/vm/slub.txt for details. |
| 122 | |
| 123 | In the last section the report shows memory state around the accessed address. |
| 124 | Reading this part requires some more understanding of how KASAN works. |
| 125 | |
| 126 | Each 8 bytes of memory are encoded in one shadow byte as accessible, |
| 127 | partially accessible, freed or they can be part of a redzone. |
| 128 | We use the following encoding for each shadow byte: 0 means that all 8 bytes |
| 129 | of the corresponding memory region are accessible; number N (1 <= N <= 7) means |
| 130 | that the first N bytes are accessible, and other (8 - N) bytes are not; |
| 131 | any negative value indicates that the entire 8-byte word is inaccessible. |
| 132 | We use different negative values to distinguish between different kinds of |
| 133 | inaccessible memory like redzones or freed memory (see mm/kasan/kasan.h). |
| 134 | |
| 135 | In the report above the arrows point to the shadow byte 03, which means that |
| 136 | the accessed address is partially accessible. |
| 137 | |
| 138 | |
| 139 | 2. Implementation details |
| 140 | ======================== |
| 141 | |
| 142 | From a high level, our approach to memory error detection is similar to that |
| 143 | of kmemcheck: use shadow memory to record whether each byte of memory is safe |
| 144 | to access, and use compile-time instrumentation to check shadow memory on each |
| 145 | memory access. |
| 146 | |
| 147 | AddressSanitizer dedicates 1/8 of kernel memory to its shadow memory |
| 148 | (e.g. 16TB to cover 128TB on x86_64) and uses direct mapping with a scale and |
| 149 | offset to translate a memory address to its corresponding shadow address. |
| 150 | |
| 151 | Here is the function witch translate an address to its corresponding shadow |
| 152 | address: |
| 153 | |
| 154 | static inline void *kasan_mem_to_shadow(const void *addr) |
| 155 | { |
| 156 | return ((unsigned long)addr >> KASAN_SHADOW_SCALE_SHIFT) |
| 157 | + KASAN_SHADOW_OFFSET; |
| 158 | } |
| 159 | |
| 160 | where KASAN_SHADOW_SCALE_SHIFT = 3. |
| 161 | |
| 162 | Compile-time instrumentation used for checking memory accesses. Compiler inserts |
| 163 | function calls (__asan_load*(addr), __asan_store*(addr)) before each memory |
| 164 | access of size 1, 2, 4, 8 or 16. These functions check whether memory access is |
| 165 | valid or not by checking corresponding shadow memory. |
| 166 | |
| 167 | GCC 5.0 has possibility to perform inline instrumentation. Instead of making |
| 168 | function calls GCC directly inserts the code to check the shadow memory. |
| 169 | This option significantly enlarges kernel but it gives x1.1-x2 performance |
| 170 | boost over outline instrumented kernel. |