Loading...
Note: File does not exist in v3.1.
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * This is for all the tests related to logic bugs (e.g. bad dereferences,
4 * bad alignment, bad loops, bad locking, bad scheduling, deep stacks, and
5 * lockups) along with other things that don't fit well into existing LKDTM
6 * test source files.
7 */
8#include "lkdtm.h"
9#include <linux/list.h>
10#include <linux/sched.h>
11#include <linux/sched/signal.h>
12#include <linux/sched/task_stack.h>
13#include <linux/uaccess.h>
14#include <linux/slab.h>
15
16#if IS_ENABLED(CONFIG_X86_32) && !IS_ENABLED(CONFIG_UML)
17#include <asm/desc.h>
18#endif
19
20struct lkdtm_list {
21 struct list_head node;
22};
23
24/*
25 * Make sure our attempts to over run the kernel stack doesn't trigger
26 * a compiler warning when CONFIG_FRAME_WARN is set. Then make sure we
27 * recurse past the end of THREAD_SIZE by default.
28 */
29#if defined(CONFIG_FRAME_WARN) && (CONFIG_FRAME_WARN > 0)
30#define REC_STACK_SIZE (_AC(CONFIG_FRAME_WARN, UL) / 2)
31#else
32#define REC_STACK_SIZE (THREAD_SIZE / 8)
33#endif
34#define REC_NUM_DEFAULT ((THREAD_SIZE / REC_STACK_SIZE) * 2)
35
36static int recur_count = REC_NUM_DEFAULT;
37
38static DEFINE_SPINLOCK(lock_me_up);
39
40/*
41 * Make sure compiler does not optimize this function or stack frame away:
42 * - function marked noinline
43 * - stack variables are marked volatile
44 * - stack variables are written (memset()) and read (pr_info())
45 * - function has external effects (pr_info())
46 * */
47static int noinline recursive_loop(int remaining)
48{
49 volatile char buf[REC_STACK_SIZE];
50
51 memset((void *)buf, remaining & 0xFF, sizeof(buf));
52 pr_info("loop %d/%d ...\n", (int)buf[remaining % sizeof(buf)],
53 recur_count);
54 if (!remaining)
55 return 0;
56 else
57 return recursive_loop(remaining - 1);
58}
59
60/* If the depth is negative, use the default, otherwise keep parameter. */
61void __init lkdtm_bugs_init(int *recur_param)
62{
63 if (*recur_param < 0)
64 *recur_param = recur_count;
65 else
66 recur_count = *recur_param;
67}
68
69void lkdtm_PANIC(void)
70{
71 panic("dumptest");
72}
73
74void lkdtm_BUG(void)
75{
76 BUG();
77}
78
79static int warn_counter;
80
81void lkdtm_WARNING(void)
82{
83 WARN_ON(++warn_counter);
84}
85
86void lkdtm_WARNING_MESSAGE(void)
87{
88 WARN(1, "Warning message trigger count: %d\n", ++warn_counter);
89}
90
91void lkdtm_EXCEPTION(void)
92{
93 *((volatile int *) 0) = 0;
94}
95
96void lkdtm_LOOP(void)
97{
98 for (;;)
99 ;
100}
101
102void lkdtm_EXHAUST_STACK(void)
103{
104 pr_info("Calling function with %lu frame size to depth %d ...\n",
105 REC_STACK_SIZE, recur_count);
106 recursive_loop(recur_count);
107 pr_info("FAIL: survived without exhausting stack?!\n");
108}
109
110static noinline void __lkdtm_CORRUPT_STACK(void *stack)
111{
112 memset(stack, '\xff', 64);
113}
114
115/* This should trip the stack canary, not corrupt the return address. */
116noinline void lkdtm_CORRUPT_STACK(void)
117{
118 /* Use default char array length that triggers stack protection. */
119 char data[8] __aligned(sizeof(void *));
120
121 pr_info("Corrupting stack containing char array ...\n");
122 __lkdtm_CORRUPT_STACK((void *)&data);
123}
124
125/* Same as above but will only get a canary with -fstack-protector-strong */
126noinline void lkdtm_CORRUPT_STACK_STRONG(void)
127{
128 union {
129 unsigned short shorts[4];
130 unsigned long *ptr;
131 } data __aligned(sizeof(void *));
132
133 pr_info("Corrupting stack containing union ...\n");
134 __lkdtm_CORRUPT_STACK((void *)&data);
135}
136
137void lkdtm_UNALIGNED_LOAD_STORE_WRITE(void)
138{
139 static u8 data[5] __attribute__((aligned(4))) = {1, 2, 3, 4, 5};
140 u32 *p;
141 u32 val = 0x12345678;
142
143 p = (u32 *)(data + 1);
144 if (*p == 0)
145 val = 0x87654321;
146 *p = val;
147}
148
149void lkdtm_SOFTLOCKUP(void)
150{
151 preempt_disable();
152 for (;;)
153 cpu_relax();
154}
155
156void lkdtm_HARDLOCKUP(void)
157{
158 local_irq_disable();
159 for (;;)
160 cpu_relax();
161}
162
163void lkdtm_SPINLOCKUP(void)
164{
165 /* Must be called twice to trigger. */
166 spin_lock(&lock_me_up);
167 /* Let sparse know we intended to exit holding the lock. */
168 __release(&lock_me_up);
169}
170
171void lkdtm_HUNG_TASK(void)
172{
173 set_current_state(TASK_UNINTERRUPTIBLE);
174 schedule();
175}
176
177volatile unsigned int huge = INT_MAX - 2;
178volatile unsigned int ignored;
179
180void lkdtm_OVERFLOW_SIGNED(void)
181{
182 int value;
183
184 value = huge;
185 pr_info("Normal signed addition ...\n");
186 value += 1;
187 ignored = value;
188
189 pr_info("Overflowing signed addition ...\n");
190 value += 4;
191 ignored = value;
192}
193
194
195void lkdtm_OVERFLOW_UNSIGNED(void)
196{
197 unsigned int value;
198
199 value = huge;
200 pr_info("Normal unsigned addition ...\n");
201 value += 1;
202 ignored = value;
203
204 pr_info("Overflowing unsigned addition ...\n");
205 value += 4;
206 ignored = value;
207}
208
209/* Intentionally using old-style flex array definition of 1 byte. */
210struct array_bounds_flex_array {
211 int one;
212 int two;
213 char data[1];
214};
215
216struct array_bounds {
217 int one;
218 int two;
219 char data[8];
220 int three;
221};
222
223void lkdtm_ARRAY_BOUNDS(void)
224{
225 struct array_bounds_flex_array *not_checked;
226 struct array_bounds *checked;
227 volatile int i;
228
229 not_checked = kmalloc(sizeof(*not_checked) * 2, GFP_KERNEL);
230 checked = kmalloc(sizeof(*checked) * 2, GFP_KERNEL);
231
232 pr_info("Array access within bounds ...\n");
233 /* For both, touch all bytes in the actual member size. */
234 for (i = 0; i < sizeof(checked->data); i++)
235 checked->data[i] = 'A';
236 /*
237 * For the uninstrumented flex array member, also touch 1 byte
238 * beyond to verify it is correctly uninstrumented.
239 */
240 for (i = 0; i < sizeof(not_checked->data) + 1; i++)
241 not_checked->data[i] = 'A';
242
243 pr_info("Array access beyond bounds ...\n");
244 for (i = 0; i < sizeof(checked->data) + 1; i++)
245 checked->data[i] = 'B';
246
247 kfree(not_checked);
248 kfree(checked);
249 pr_err("FAIL: survived array bounds overflow!\n");
250}
251
252void lkdtm_CORRUPT_LIST_ADD(void)
253{
254 /*
255 * Initially, an empty list via LIST_HEAD:
256 * test_head.next = &test_head
257 * test_head.prev = &test_head
258 */
259 LIST_HEAD(test_head);
260 struct lkdtm_list good, bad;
261 void *target[2] = { };
262 void *redirection = ⌖
263
264 pr_info("attempting good list addition\n");
265
266 /*
267 * Adding to the list performs these actions:
268 * test_head.next->prev = &good.node
269 * good.node.next = test_head.next
270 * good.node.prev = test_head
271 * test_head.next = good.node
272 */
273 list_add(&good.node, &test_head);
274
275 pr_info("attempting corrupted list addition\n");
276 /*
277 * In simulating this "write what where" primitive, the "what" is
278 * the address of &bad.node, and the "where" is the address held
279 * by "redirection".
280 */
281 test_head.next = redirection;
282 list_add(&bad.node, &test_head);
283
284 if (target[0] == NULL && target[1] == NULL)
285 pr_err("Overwrite did not happen, but no BUG?!\n");
286 else
287 pr_err("list_add() corruption not detected!\n");
288}
289
290void lkdtm_CORRUPT_LIST_DEL(void)
291{
292 LIST_HEAD(test_head);
293 struct lkdtm_list item;
294 void *target[2] = { };
295 void *redirection = ⌖
296
297 list_add(&item.node, &test_head);
298
299 pr_info("attempting good list removal\n");
300 list_del(&item.node);
301
302 pr_info("attempting corrupted list removal\n");
303 list_add(&item.node, &test_head);
304
305 /* As with the list_add() test above, this corrupts "next". */
306 item.node.next = redirection;
307 list_del(&item.node);
308
309 if (target[0] == NULL && target[1] == NULL)
310 pr_err("Overwrite did not happen, but no BUG?!\n");
311 else
312 pr_err("list_del() corruption not detected!\n");
313}
314
315/* Test if unbalanced set_fs(KERNEL_DS)/set_fs(USER_DS) check exists. */
316void lkdtm_CORRUPT_USER_DS(void)
317{
318 pr_info("setting bad task size limit\n");
319 set_fs(KERNEL_DS);
320
321 /* Make sure we do not keep running with a KERNEL_DS! */
322 force_sig(SIGKILL);
323}
324
325/* Test that VMAP_STACK is actually allocating with a leading guard page */
326void lkdtm_STACK_GUARD_PAGE_LEADING(void)
327{
328 const unsigned char *stack = task_stack_page(current);
329 const unsigned char *ptr = stack - 1;
330 volatile unsigned char byte;
331
332 pr_info("attempting bad read from page below current stack\n");
333
334 byte = *ptr;
335
336 pr_err("FAIL: accessed page before stack! (byte: %x)\n", byte);
337}
338
339/* Test that VMAP_STACK is actually allocating with a trailing guard page */
340void lkdtm_STACK_GUARD_PAGE_TRAILING(void)
341{
342 const unsigned char *stack = task_stack_page(current);
343 const unsigned char *ptr = stack + THREAD_SIZE;
344 volatile unsigned char byte;
345
346 pr_info("attempting bad read from page above current stack\n");
347
348 byte = *ptr;
349
350 pr_err("FAIL: accessed page after stack! (byte: %x)\n", byte);
351}
352
353void lkdtm_UNSET_SMEP(void)
354{
355#if IS_ENABLED(CONFIG_X86_64) && !IS_ENABLED(CONFIG_UML)
356#define MOV_CR4_DEPTH 64
357 void (*direct_write_cr4)(unsigned long val);
358 unsigned char *insn;
359 unsigned long cr4;
360 int i;
361
362 cr4 = native_read_cr4();
363
364 if ((cr4 & X86_CR4_SMEP) != X86_CR4_SMEP) {
365 pr_err("FAIL: SMEP not in use\n");
366 return;
367 }
368 cr4 &= ~(X86_CR4_SMEP);
369
370 pr_info("trying to clear SMEP normally\n");
371 native_write_cr4(cr4);
372 if (cr4 == native_read_cr4()) {
373 pr_err("FAIL: pinning SMEP failed!\n");
374 cr4 |= X86_CR4_SMEP;
375 pr_info("restoring SMEP\n");
376 native_write_cr4(cr4);
377 return;
378 }
379 pr_info("ok: SMEP did not get cleared\n");
380
381 /*
382 * To test the post-write pinning verification we need to call
383 * directly into the middle of native_write_cr4() where the
384 * cr4 write happens, skipping any pinning. This searches for
385 * the cr4 writing instruction.
386 */
387 insn = (unsigned char *)native_write_cr4;
388 for (i = 0; i < MOV_CR4_DEPTH; i++) {
389 /* mov %rdi, %cr4 */
390 if (insn[i] == 0x0f && insn[i+1] == 0x22 && insn[i+2] == 0xe7)
391 break;
392 /* mov %rdi,%rax; mov %rax, %cr4 */
393 if (insn[i] == 0x48 && insn[i+1] == 0x89 &&
394 insn[i+2] == 0xf8 && insn[i+3] == 0x0f &&
395 insn[i+4] == 0x22 && insn[i+5] == 0xe0)
396 break;
397 }
398 if (i >= MOV_CR4_DEPTH) {
399 pr_info("ok: cannot locate cr4 writing call gadget\n");
400 return;
401 }
402 direct_write_cr4 = (void *)(insn + i);
403
404 pr_info("trying to clear SMEP with call gadget\n");
405 direct_write_cr4(cr4);
406 if (native_read_cr4() & X86_CR4_SMEP) {
407 pr_info("ok: SMEP removal was reverted\n");
408 } else {
409 pr_err("FAIL: cleared SMEP not detected!\n");
410 cr4 |= X86_CR4_SMEP;
411 pr_info("restoring SMEP\n");
412 native_write_cr4(cr4);
413 }
414#else
415 pr_err("XFAIL: this test is x86_64-only\n");
416#endif
417}
418
419void lkdtm_DOUBLE_FAULT(void)
420{
421#if IS_ENABLED(CONFIG_X86_32) && !IS_ENABLED(CONFIG_UML)
422 /*
423 * Trigger #DF by setting the stack limit to zero. This clobbers
424 * a GDT TLS slot, which is okay because the current task will die
425 * anyway due to the double fault.
426 */
427 struct desc_struct d = {
428 .type = 3, /* expand-up, writable, accessed data */
429 .p = 1, /* present */
430 .d = 1, /* 32-bit */
431 .g = 0, /* limit in bytes */
432 .s = 1, /* not system */
433 };
434
435 local_irq_disable();
436 write_gdt_entry(get_cpu_gdt_rw(smp_processor_id()),
437 GDT_ENTRY_TLS_MIN, &d, DESCTYPE_S);
438
439 /*
440 * Put our zero-limit segment in SS and then trigger a fault. The
441 * 4-byte access to (%esp) will fault with #SS, and the attempt to
442 * deliver the fault will recursively cause #SS and result in #DF.
443 * This whole process happens while NMIs and MCEs are blocked by the
444 * MOV SS window. This is nice because an NMI with an invalid SS
445 * would also double-fault, resulting in the NMI or MCE being lost.
446 */
447 asm volatile ("movw %0, %%ss; addl $0, (%%esp)" ::
448 "r" ((unsigned short)(GDT_ENTRY_TLS_MIN << 3)));
449
450 pr_err("FAIL: tried to double fault but didn't die\n");
451#else
452 pr_err("XFAIL: this test is ia32-only\n");
453#endif
454}
455
456#ifdef CONFIG_ARM64
457static noinline void change_pac_parameters(void)
458{
459 if (IS_ENABLED(CONFIG_ARM64_PTR_AUTH)) {
460 /* Reset the keys of current task */
461 ptrauth_thread_init_kernel(current);
462 ptrauth_thread_switch_kernel(current);
463 }
464}
465#endif
466
467noinline void lkdtm_CORRUPT_PAC(void)
468{
469#ifdef CONFIG_ARM64
470#define CORRUPT_PAC_ITERATE 10
471 int i;
472
473 if (!IS_ENABLED(CONFIG_ARM64_PTR_AUTH))
474 pr_err("FAIL: kernel not built with CONFIG_ARM64_PTR_AUTH\n");
475
476 if (!system_supports_address_auth()) {
477 pr_err("FAIL: CPU lacks pointer authentication feature\n");
478 return;
479 }
480
481 pr_info("changing PAC parameters to force function return failure...\n");
482 /*
483 * PAC is a hash value computed from input keys, return address and
484 * stack pointer. As pac has fewer bits so there is a chance of
485 * collision, so iterate few times to reduce the collision probability.
486 */
487 for (i = 0; i < CORRUPT_PAC_ITERATE; i++)
488 change_pac_parameters();
489
490 pr_err("FAIL: survived PAC changes! Kernel may be unstable from here\n");
491#else
492 pr_err("XFAIL: this test is arm64-only\n");
493#endif
494}