1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * Kernel Probes (KProbes)
4 *
5 * Copyright (C) IBM Corporation, 2002, 2004
6 *
7 * 2002-Oct Created by Vamsi Krishna S <vamsi_krishna@in.ibm.com> Kernel
8 * Probes initial implementation ( includes contributions from
9 * Rusty Russell).
10 * 2004-July Suparna Bhattacharya <suparna@in.ibm.com> added jumper probes
11 * interface to access function arguments.
12 * 2004-Oct Jim Keniston <jkenisto@us.ibm.com> and Prasanna S Panchamukhi
13 * <prasanna@in.ibm.com> adapted for x86_64 from i386.
14 * 2005-Mar Roland McGrath <roland@redhat.com>
15 * Fixed to handle %rip-relative addressing mode correctly.
16 * 2005-May Hien Nguyen <hien@us.ibm.com>, Jim Keniston
17 * <jkenisto@us.ibm.com> and Prasanna S Panchamukhi
18 * <prasanna@in.ibm.com> added function-return probes.
19 * 2005-May Rusty Lynch <rusty.lynch@intel.com>
20 * Added function return probes functionality
21 * 2006-Feb Masami Hiramatsu <hiramatu@sdl.hitachi.co.jp> added
22 * kprobe-booster and kretprobe-booster for i386.
23 * 2007-Dec Masami Hiramatsu <mhiramat@redhat.com> added kprobe-booster
24 * and kretprobe-booster for x86-64
25 * 2007-Dec Masami Hiramatsu <mhiramat@redhat.com>, Arjan van de Ven
26 * <arjan@infradead.org> and Jim Keniston <jkenisto@us.ibm.com>
27 * unified x86 kprobes code.
28 */
29 #include <linux/kprobes.h>
30 #include <linux/ptrace.h>
31 #include <linux/string.h>
32 #include <linux/slab.h>
33 #include <linux/hardirq.h>
34 #include <linux/preempt.h>
35 #include <linux/sched/debug.h>
36 #include <linux/perf_event.h>
37 #include <linux/extable.h>
38 #include <linux/kdebug.h>
39 #include <linux/kallsyms.h>
40 #include <linux/kgdb.h>
41 #include <linux/ftrace.h>
42 #include <linux/kasan.h>
43 #include <linux/objtool.h>
44 #include <linux/vmalloc.h>
45 #include <linux/pgtable.h>
46 #include <linux/set_memory.h>
47 #include <linux/cfi.h>
48 #include <linux/execmem.h>
49
50 #include <asm/text-patching.h>
51 #include <asm/cacheflush.h>
52 #include <asm/desc.h>
53 #include <linux/uaccess.h>
54 #include <asm/alternative.h>
55 #include <asm/insn.h>
56 #include <asm/debugreg.h>
57 #include <asm/ibt.h>
58
59 #include "common.h"
60
61 DEFINE_PER_CPU(struct kprobe *, current_kprobe) = NULL;
62 DEFINE_PER_CPU(struct kprobe_ctlblk, kprobe_ctlblk);
63
64 #define W(row, b0, b1, b2, b3, b4, b5, b6, b7, b8, b9, ba, bb, bc, bd, be, bf)\
65 (((b0##UL << 0x0)|(b1##UL << 0x1)|(b2##UL << 0x2)|(b3##UL << 0x3) | \
66 (b4##UL << 0x4)|(b5##UL << 0x5)|(b6##UL << 0x6)|(b7##UL << 0x7) | \
67 (b8##UL << 0x8)|(b9##UL << 0x9)|(ba##UL << 0xa)|(bb##UL << 0xb) | \
68 (bc##UL << 0xc)|(bd##UL << 0xd)|(be##UL << 0xe)|(bf##UL << 0xf)) \
69 << (row % 32))
70 /*
71 * Undefined/reserved opcodes, conditional jump, Opcode Extension
72 * Groups, and some special opcodes can not boost.
73 * This is non-const and volatile to keep gcc from statically
74 * optimizing it out, as variable_test_bit makes gcc think only
75 * *(unsigned long*) is used.
76 */
77 static volatile u32 twobyte_is_boostable[256 / 32] = {
78 /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */
79 /* ---------------------------------------------- */
80 W(0x00, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0) | /* 00 */
81 W(0x10, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1) , /* 10 */
82 W(0x20, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) | /* 20 */
83 W(0x30, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) , /* 30 */
84 W(0x40, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* 40 */
85 W(0x50, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) , /* 50 */
86 W(0x60, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1) | /* 60 */
87 W(0x70, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1) , /* 70 */
88 W(0x80, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) | /* 80 */
89 W(0x90, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 90 */
90 W(0xa0, 1, 1, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1, 1, 1, 0, 1) | /* a0 */
91 W(0xb0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1) , /* b0 */
92 W(0xc0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1) | /* c0 */
93 W(0xd0, 0, 1, 1, 1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 0, 1) , /* d0 */
94 W(0xe0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 0, 1) | /* e0 */
95 W(0xf0, 0, 1, 1, 1, 0, 1, 0, 0, 1, 1, 1, 0, 1, 1, 1, 0) /* f0 */
96 /* ----------------------------------------------- */
97 /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */
98 };
99 #undef W
100
101 struct kretprobe_blackpoint kretprobe_blacklist[] = {
102 {"__switch_to", }, /* This function switches only current task, but
103 doesn't switch kernel stack.*/
104 {NULL, NULL} /* Terminator */
105 };
106
107 const int kretprobe_blacklist_size = ARRAY_SIZE(kretprobe_blacklist);
108
109 static nokprobe_inline void
110 __synthesize_relative_insn(void *dest, void *from, void *to, u8 op)
111 {
112 struct __arch_relative_insn {
113 u8 op;
114 s32 raddr;
115 } __packed *insn;
116
117 insn = (struct __arch_relative_insn *)dest;
118 insn->raddr = (s32)((long)(to) - ((long)(from) + 5));
119 insn->op = op;
120 }
121
122 /* Insert a jump instruction at address 'from', which jumps to address 'to'.*/
123 void synthesize_reljump(void *dest, void *from, void *to)
124 {
125 __synthesize_relative_insn(dest, from, to, JMP32_INSN_OPCODE);
126 }
127 NOKPROBE_SYMBOL(synthesize_reljump);
128
129 /* Insert a call instruction at address 'from', which calls address 'to'.*/
130 void synthesize_relcall(void *dest, void *from, void *to)
131 {
132 __synthesize_relative_insn(dest, from, to, CALL_INSN_OPCODE);
133 }
134 NOKPROBE_SYMBOL(synthesize_relcall);
135
136 /*
137 * Returns non-zero if INSN is boostable.
138 * RIP relative instructions are adjusted at copying time in 64 bits mode
139 */
140 bool can_boost(struct insn *insn, void *addr)
141 {
142 kprobe_opcode_t opcode;
143 insn_byte_t prefix;
144 int i;
145
146 if (search_exception_tables((unsigned long)addr))
147 return false; /* Page fault may occur on this address. */
148
149 /* 2nd-byte opcode */
150 if (insn->opcode.nbytes == 2)
151 return test_bit(insn->opcode.bytes[1],
152 (unsigned long *)twobyte_is_boostable);
153
154 if (insn->opcode.nbytes != 1)
155 return false;
156
157 for_each_insn_prefix(insn, i, prefix) {
158 insn_attr_t attr;
159
160 attr = inat_get_opcode_attribute(prefix);
161 /* Can't boost Address-size override prefix and CS override prefix */
162 if (prefix == 0x2e || inat_is_address_size_prefix(attr))
163 return false;
164 }
165
166 opcode = insn->opcode.bytes[0];
167
168 switch (opcode) {
169 case 0x62: /* bound */
170 case 0x70 ... 0x7f: /* Conditional jumps */
171 case 0x9a: /* Call far */
172 case 0xcc ... 0xce: /* software exceptions */
173 case 0xd6: /* (UD) */
174 case 0xd8 ... 0xdf: /* ESC */
175 case 0xe0 ... 0xe3: /* LOOP*, JCXZ */
176 case 0xe8 ... 0xe9: /* near Call, JMP */
177 case 0xeb: /* Short JMP */
178 case 0xf0 ... 0xf4: /* LOCK/REP, HLT */
179 /* ... are not boostable */
180 return false;
181 case 0xc0 ... 0xc1: /* Grp2 */
182 case 0xd0 ... 0xd3: /* Grp2 */
183 /*
184 * AMD uses nnn == 110 as SHL/SAL, but Intel makes it reserved.
185 */
186 return X86_MODRM_REG(insn->modrm.bytes[0]) != 0b110;
187 case 0xf6 ... 0xf7: /* Grp3 */
188 /* AMD uses nnn == 001 as TEST, but Intel makes it reserved. */
189 return X86_MODRM_REG(insn->modrm.bytes[0]) != 0b001;
190 case 0xfe: /* Grp4 */
191 /* Only INC and DEC are boostable */
192 return X86_MODRM_REG(insn->modrm.bytes[0]) == 0b000 ||
193 X86_MODRM_REG(insn->modrm.bytes[0]) == 0b001;
194 case 0xff: /* Grp5 */
195 /* Only INC, DEC, and indirect JMP are boostable */
196 return X86_MODRM_REG(insn->modrm.bytes[0]) == 0b000 ||
197 X86_MODRM_REG(insn->modrm.bytes[0]) == 0b001 ||
198 X86_MODRM_REG(insn->modrm.bytes[0]) == 0b100;
199 default:
200 return true;
201 }
202 }
203
204 static unsigned long
205 __recover_probed_insn(kprobe_opcode_t *buf, unsigned long addr)
206 {
207 struct kprobe *kp;
208 bool faddr;
209
210 kp = get_kprobe((void *)addr);
211 faddr = ftrace_location(addr) == addr;
212 /*
213 * Use the current code if it is not modified by Kprobe
214 * and it cannot be modified by ftrace.
215 */
216 if (!kp && !faddr)
217 return addr;
218
219 /*
220 * Basically, kp->ainsn.insn has an original instruction.
221 * However, RIP-relative instruction can not do single-stepping
222 * at different place, __copy_instruction() tweaks the displacement of
223 * that instruction. In that case, we can't recover the instruction
224 * from the kp->ainsn.insn.
225 *
226 * On the other hand, in case on normal Kprobe, kp->opcode has a copy
227 * of the first byte of the probed instruction, which is overwritten
228 * by int3. And the instruction at kp->addr is not modified by kprobes
229 * except for the first byte, we can recover the original instruction
230 * from it and kp->opcode.
231 *
232 * In case of Kprobes using ftrace, we do not have a copy of
233 * the original instruction. In fact, the ftrace location might
234 * be modified at anytime and even could be in an inconsistent state.
235 * Fortunately, we know that the original code is the ideal 5-byte
236 * long NOP.
237 */
238 if (copy_from_kernel_nofault(buf, (void *)addr,
239 MAX_INSN_SIZE * sizeof(kprobe_opcode_t)))
240 return 0UL;
241
242 if (faddr)
243 memcpy(buf, x86_nops[5], 5);
244 else
245 buf[0] = kp->opcode;
246 return (unsigned long)buf;
247 }
248
249 /*
250 * Recover the probed instruction at addr for further analysis.
251 * Caller must lock kprobes by kprobe_mutex, or disable preemption
252 * for preventing to release referencing kprobes.
253 * Returns zero if the instruction can not get recovered (or access failed).
254 */
255 unsigned long recover_probed_instruction(kprobe_opcode_t *buf, unsigned long addr)
256 {
257 unsigned long __addr;
258
259 __addr = __recover_optprobed_insn(buf, addr);
260 if (__addr != addr)
261 return __addr;
262
263 return __recover_probed_insn(buf, addr);
264 }
265
266 /* Check if insn is INT or UD */
267 static inline bool is_exception_insn(struct insn *insn)
268 {
269 /* UD uses 0f escape */
270 if (insn->opcode.bytes[0] == 0x0f) {
271 /* UD0 / UD1 / UD2 */
272 return insn->opcode.bytes[1] == 0xff ||
273 insn->opcode.bytes[1] == 0xb9 ||
274 insn->opcode.bytes[1] == 0x0b;
275 }
276
277 /* INT3 / INT n / INTO / INT1 */
278 return insn->opcode.bytes[0] == 0xcc ||
279 insn->opcode.bytes[0] == 0xcd ||
280 insn->opcode.bytes[0] == 0xce ||
281 insn->opcode.bytes[0] == 0xf1;
282 }
283
284 /*
285 * Check if paddr is at an instruction boundary and that instruction can
286 * be probed
287 */
288 static bool can_probe(unsigned long paddr)
289 {
290 unsigned long addr, __addr, offset = 0;
291 struct insn insn;
292 kprobe_opcode_t buf[MAX_INSN_SIZE];
293
294 if (!kallsyms_lookup_size_offset(paddr, NULL, &offset))
295 return false;
296
297 /* Decode instructions */
298 addr = paddr - offset;
299 while (addr < paddr) {
300 /*
301 * Check if the instruction has been modified by another
302 * kprobe, in which case we replace the breakpoint by the
303 * original instruction in our buffer.
304 * Also, jump optimization will change the breakpoint to
305 * relative-jump. Since the relative-jump itself is
306 * normally used, we just go through if there is no kprobe.
307 */
308 __addr = recover_probed_instruction(buf, addr);
309 if (!__addr)
310 return false;
311
312 if (insn_decode_kernel(&insn, (void *)__addr) < 0)
313 return false;
314
315 #ifdef CONFIG_KGDB
316 /*
317 * If there is a dynamically installed kgdb sw breakpoint,
318 * this function should not be probed.
319 */
320 if (insn.opcode.bytes[0] == INT3_INSN_OPCODE &&
321 kgdb_has_hit_break(addr))
322 return false;
323 #endif
324 addr += insn.length;
325 }
326
327 /* Check if paddr is at an instruction boundary */
328 if (addr != paddr)
329 return false;
330
331 __addr = recover_probed_instruction(buf, addr);
332 if (!__addr)
333 return false;
334
335 if (insn_decode_kernel(&insn, (void *)__addr) < 0)
336 return false;
337
338 /* INT and UD are special and should not be kprobed */
339 if (is_exception_insn(&insn))
340 return false;
341
342 if (IS_ENABLED(CONFIG_CFI_CLANG)) {
343 /*
344 * The compiler generates the following instruction sequence
345 * for indirect call checks and cfi.c decodes this;
346 *
347 * movl -<id>, %r10d ; 6 bytes
348 * addl -4(%reg), %r10d ; 4 bytes
349 * je .Ltmp1 ; 2 bytes
350 * ud2 ; <- regs->ip
351 * .Ltmp1:
352 *
353 * Also, these movl and addl are used for showing expected
354 * type. So those must not be touched.
355 */
356 if (insn.opcode.value == 0xBA)
357 offset = 12;
358 else if (insn.opcode.value == 0x3)
359 offset = 6;
360 else
361 goto out;
362
363 /* This movl/addl is used for decoding CFI. */
364 if (is_cfi_trap(addr + offset))
365 return false;
366 }
367
368 out:
369 return true;
370 }
371
372 /* If x86 supports IBT (ENDBR) it must be skipped. */
373 kprobe_opcode_t *arch_adjust_kprobe_addr(unsigned long addr, unsigned long offset,
374 bool *on_func_entry)
375 {
376 u32 insn;
377
378 /*
379 * Since 'addr' is not guaranteed to be safe to access, use
380 * copy_from_kernel_nofault() to read the instruction:
381 */
382 if (copy_from_kernel_nofault(&insn, (void *)addr, sizeof(u32)))
383 return NULL;
384
385 if (is_endbr(insn)) {
386 *on_func_entry = !offset || offset == 4;
387 if (*on_func_entry)
388 offset = 4;
389
390 } else {
391 *on_func_entry = !offset;
392 }
393
394 return (kprobe_opcode_t *)(addr + offset);
395 }
396
397 /*
398 * Copy an instruction with recovering modified instruction by kprobes
399 * and adjust the displacement if the instruction uses the %rip-relative
400 * addressing mode. Note that since @real will be the final place of copied
401 * instruction, displacement must be adjust by @real, not @dest.
402 * This returns the length of copied instruction, or 0 if it has an error.
403 */
404 int __copy_instruction(u8 *dest, u8 *src, u8 *real, struct insn *insn)
405 {
406 kprobe_opcode_t buf[MAX_INSN_SIZE];
407 unsigned long recovered_insn = recover_probed_instruction(buf, (unsigned long)src);
408 int ret;
409
410 if (!recovered_insn || !insn)
411 return 0;
412
413 /* This can access kernel text if given address is not recovered */
414 if (copy_from_kernel_nofault(dest, (void *)recovered_insn,
415 MAX_INSN_SIZE))
416 return 0;
417
418 ret = insn_decode_kernel(insn, dest);
419 if (ret < 0)
420 return 0;
421
422 /* We can not probe force emulate prefixed instruction */
423 if (insn_has_emulate_prefix(insn))
424 return 0;
425
426 /* Another subsystem puts a breakpoint, failed to recover */
427 if (insn->opcode.bytes[0] == INT3_INSN_OPCODE)
428 return 0;
429
430 /* We should not singlestep on the exception masking instructions */
431 if (insn_masking_exception(insn))
432 return 0;
433
434 #ifdef CONFIG_X86_64
435 /* Only x86_64 has RIP relative instructions */
436 if (insn_rip_relative(insn)) {
437 s64 newdisp;
438 u8 *disp;
439 /*
440 * The copied instruction uses the %rip-relative addressing
441 * mode. Adjust the displacement for the difference between
442 * the original location of this instruction and the location
443 * of the copy that will actually be run. The tricky bit here
444 * is making sure that the sign extension happens correctly in
445 * this calculation, since we need a signed 32-bit result to
446 * be sign-extended to 64 bits when it's added to the %rip
447 * value and yield the same 64-bit result that the sign-
448 * extension of the original signed 32-bit displacement would
449 * have given.
450 */
451 newdisp = (u8 *) src + (s64) insn->displacement.value
452 - (u8 *) real;
453 if ((s64) (s32) newdisp != newdisp) {
454 pr_err("Kprobes error: new displacement does not fit into s32 (%llx)\n", newdisp);
455 return 0;
456 }
457 disp = (u8 *) dest + insn_offset_displacement(insn);
458 *(s32 *) disp = (s32) newdisp;
459 }
460 #endif
461 return insn->length;
462 }
463
464 /* Prepare reljump or int3 right after instruction */
465 static int prepare_singlestep(kprobe_opcode_t *buf, struct kprobe *p,
466 struct insn *insn)
467 {
468 int len = insn->length;
469
470 if (!IS_ENABLED(CONFIG_PREEMPTION) &&
471 !p->post_handler && can_boost(insn, p->addr) &&
472 MAX_INSN_SIZE - len >= JMP32_INSN_SIZE) {
473 /*
474 * These instructions can be executed directly if it
475 * jumps back to correct address.
476 */
477 synthesize_reljump(buf + len, p->ainsn.insn + len,
478 p->addr + insn->length);
479 len += JMP32_INSN_SIZE;
480 p->ainsn.boostable = 1;
481 } else {
482 /* Otherwise, put an int3 for trapping singlestep */
483 if (MAX_INSN_SIZE - len < INT3_INSN_SIZE)
484 return -ENOSPC;
485
486 buf[len] = INT3_INSN_OPCODE;
487 len += INT3_INSN_SIZE;
488 }
489
490 return len;
491 }
492
493 /* Make page to RO mode when allocate it */
494 void *alloc_insn_page(void)
495 {
496 void *page;
497
498 page = execmem_alloc(EXECMEM_KPROBES, PAGE_SIZE);
499 if (!page)
500 return NULL;
501
502 /*
503 * TODO: Once additional kernel code protection mechanisms are set, ensure
504 * that the page was not maliciously altered and it is still zeroed.
505 */
506 set_memory_rox((unsigned long)page, 1);
507
508 return page;
509 }
510
511 /* Kprobe x86 instruction emulation - only regs->ip or IF flag modifiers */
512
513 static void kprobe_emulate_ifmodifiers(struct kprobe *p, struct pt_regs *regs)
514 {
515 switch (p->ainsn.opcode) {
516 case 0xfa: /* cli */
517 regs->flags &= ~(X86_EFLAGS_IF);
518 break;
519 case 0xfb: /* sti */
520 regs->flags |= X86_EFLAGS_IF;
521 break;
522 case 0x9c: /* pushf */
523 int3_emulate_push(regs, regs->flags);
524 break;
525 case 0x9d: /* popf */
526 regs->flags = int3_emulate_pop(regs);
527 break;
528 }
529 regs->ip = regs->ip - INT3_INSN_SIZE + p->ainsn.size;
530 }
531 NOKPROBE_SYMBOL(kprobe_emulate_ifmodifiers);
532
533 static void kprobe_emulate_ret(struct kprobe *p, struct pt_regs *regs)
534 {
535 int3_emulate_ret(regs);
536 }
537 NOKPROBE_SYMBOL(kprobe_emulate_ret);
538
539 static void kprobe_emulate_call(struct kprobe *p, struct pt_regs *regs)
540 {
541 unsigned long func = regs->ip - INT3_INSN_SIZE + p->ainsn.size;
542
543 func += p->ainsn.rel32;
544 int3_emulate_call(regs, func);
545 }
546 NOKPROBE_SYMBOL(kprobe_emulate_call);
547
548 static void kprobe_emulate_jmp(struct kprobe *p, struct pt_regs *regs)
549 {
550 unsigned long ip = regs->ip - INT3_INSN_SIZE + p->ainsn.size;
551
552 ip += p->ainsn.rel32;
553 int3_emulate_jmp(regs, ip);
554 }
555 NOKPROBE_SYMBOL(kprobe_emulate_jmp);
556
557 static void kprobe_emulate_jcc(struct kprobe *p, struct pt_regs *regs)
558 {
559 unsigned long ip = regs->ip - INT3_INSN_SIZE + p->ainsn.size;
560
561 int3_emulate_jcc(regs, p->ainsn.jcc.type, ip, p->ainsn.rel32);
562 }
563 NOKPROBE_SYMBOL(kprobe_emulate_jcc);
564
565 static void kprobe_emulate_loop(struct kprobe *p, struct pt_regs *regs)
566 {
567 unsigned long ip = regs->ip - INT3_INSN_SIZE + p->ainsn.size;
568 bool match;
569
570 if (p->ainsn.loop.type != 3) { /* LOOP* */
571 if (p->ainsn.loop.asize == 32)
572 match = ((*(u32 *)®s->cx)--) != 0;
573 #ifdef CONFIG_X86_64
574 else if (p->ainsn.loop.asize == 64)
575 match = ((*(u64 *)®s->cx)--) != 0;
576 #endif
577 else
578 match = ((*(u16 *)®s->cx)--) != 0;
579 } else { /* JCXZ */
580 if (p->ainsn.loop.asize == 32)
581 match = *(u32 *)(®s->cx) == 0;
582 #ifdef CONFIG_X86_64
583 else if (p->ainsn.loop.asize == 64)
584 match = *(u64 *)(®s->cx) == 0;
585 #endif
586 else
587 match = *(u16 *)(®s->cx) == 0;
588 }
589
590 if (p->ainsn.loop.type == 0) /* LOOPNE */
591 match = match && !(regs->flags & X86_EFLAGS_ZF);
592 else if (p->ainsn.loop.type == 1) /* LOOPE */
593 match = match && (regs->flags & X86_EFLAGS_ZF);
594
595 if (match)
596 ip += p->ainsn.rel32;
597 int3_emulate_jmp(regs, ip);
598 }
599 NOKPROBE_SYMBOL(kprobe_emulate_loop);
600
601 static const int addrmode_regoffs[] = {
602 offsetof(struct pt_regs, ax),
603 offsetof(struct pt_regs, cx),
604 offsetof(struct pt_regs, dx),
605 offsetof(struct pt_regs, bx),
606 offsetof(struct pt_regs, sp),
607 offsetof(struct pt_regs, bp),
608 offsetof(struct pt_regs, si),
609 offsetof(struct pt_regs, di),
610 #ifdef CONFIG_X86_64
611 offsetof(struct pt_regs, r8),
612 offsetof(struct pt_regs, r9),
613 offsetof(struct pt_regs, r10),
614 offsetof(struct pt_regs, r11),
615 offsetof(struct pt_regs, r12),
616 offsetof(struct pt_regs, r13),
617 offsetof(struct pt_regs, r14),
618 offsetof(struct pt_regs, r15),
619 #endif
620 };
621
622 static void kprobe_emulate_call_indirect(struct kprobe *p, struct pt_regs *regs)
623 {
624 unsigned long offs = addrmode_regoffs[p->ainsn.indirect.reg];
625
626 int3_emulate_push(regs, regs->ip - INT3_INSN_SIZE + p->ainsn.size);
627 int3_emulate_jmp(regs, regs_get_register(regs, offs));
628 }
629 NOKPROBE_SYMBOL(kprobe_emulate_call_indirect);
630
631 static void kprobe_emulate_jmp_indirect(struct kprobe *p, struct pt_regs *regs)
632 {
633 unsigned long offs = addrmode_regoffs[p->ainsn.indirect.reg];
634
635 int3_emulate_jmp(regs, regs_get_register(regs, offs));
636 }
637 NOKPROBE_SYMBOL(kprobe_emulate_jmp_indirect);
638
639 static int prepare_emulation(struct kprobe *p, struct insn *insn)
640 {
641 insn_byte_t opcode = insn->opcode.bytes[0];
642
643 switch (opcode) {
644 case 0xfa: /* cli */
645 case 0xfb: /* sti */
646 case 0x9c: /* pushfl */
647 case 0x9d: /* popf/popfd */
648 /*
649 * IF modifiers must be emulated since it will enable interrupt while
650 * int3 single stepping.
651 */
652 p->ainsn.emulate_op = kprobe_emulate_ifmodifiers;
653 p->ainsn.opcode = opcode;
654 break;
655 case 0xc2: /* ret/lret */
656 case 0xc3:
657 case 0xca:
658 case 0xcb:
659 p->ainsn.emulate_op = kprobe_emulate_ret;
660 break;
661 case 0x9a: /* far call absolute -- segment is not supported */
662 case 0xea: /* far jmp absolute -- segment is not supported */
663 case 0xcc: /* int3 */
664 case 0xcf: /* iret -- in-kernel IRET is not supported */
665 return -EOPNOTSUPP;
666 break;
667 case 0xe8: /* near call relative */
668 p->ainsn.emulate_op = kprobe_emulate_call;
669 if (insn->immediate.nbytes == 2)
670 p->ainsn.rel32 = *(s16 *)&insn->immediate.value;
671 else
672 p->ainsn.rel32 = *(s32 *)&insn->immediate.value;
673 break;
674 case 0xeb: /* short jump relative */
675 case 0xe9: /* near jump relative */
676 p->ainsn.emulate_op = kprobe_emulate_jmp;
677 if (insn->immediate.nbytes == 1)
678 p->ainsn.rel32 = *(s8 *)&insn->immediate.value;
679 else if (insn->immediate.nbytes == 2)
680 p->ainsn.rel32 = *(s16 *)&insn->immediate.value;
681 else
682 p->ainsn.rel32 = *(s32 *)&insn->immediate.value;
683 break;
684 case 0x70 ... 0x7f:
685 /* 1 byte conditional jump */
686 p->ainsn.emulate_op = kprobe_emulate_jcc;
687 p->ainsn.jcc.type = opcode & 0xf;
688 p->ainsn.rel32 = insn->immediate.value;
689 break;
690 case 0x0f:
691 opcode = insn->opcode.bytes[1];
692 if ((opcode & 0xf0) == 0x80) {
693 /* 2 bytes Conditional Jump */
694 p->ainsn.emulate_op = kprobe_emulate_jcc;
695 p->ainsn.jcc.type = opcode & 0xf;
696 if (insn->immediate.nbytes == 2)
697 p->ainsn.rel32 = *(s16 *)&insn->immediate.value;
698 else
699 p->ainsn.rel32 = *(s32 *)&insn->immediate.value;
700 } else if (opcode == 0x01 &&
701 X86_MODRM_REG(insn->modrm.bytes[0]) == 0 &&
702 X86_MODRM_MOD(insn->modrm.bytes[0]) == 3) {
703 /* VM extensions - not supported */
704 return -EOPNOTSUPP;
705 }
706 break;
707 case 0xe0: /* Loop NZ */
708 case 0xe1: /* Loop */
709 case 0xe2: /* Loop */
710 case 0xe3: /* J*CXZ */
711 p->ainsn.emulate_op = kprobe_emulate_loop;
712 p->ainsn.loop.type = opcode & 0x3;
713 p->ainsn.loop.asize = insn->addr_bytes * 8;
714 p->ainsn.rel32 = *(s8 *)&insn->immediate.value;
715 break;
716 case 0xff:
717 /*
718 * Since the 0xff is an extended group opcode, the instruction
719 * is determined by the MOD/RM byte.
720 */
721 opcode = insn->modrm.bytes[0];
722 switch (X86_MODRM_REG(opcode)) {
723 case 0b010: /* FF /2, call near, absolute indirect */
724 p->ainsn.emulate_op = kprobe_emulate_call_indirect;
725 break;
726 case 0b100: /* FF /4, jmp near, absolute indirect */
727 p->ainsn.emulate_op = kprobe_emulate_jmp_indirect;
728 break;
729 case 0b011: /* FF /3, call far, absolute indirect */
730 case 0b101: /* FF /5, jmp far, absolute indirect */
731 return -EOPNOTSUPP;
732 }
733
734 if (!p->ainsn.emulate_op)
735 break;
736
737 if (insn->addr_bytes != sizeof(unsigned long))
738 return -EOPNOTSUPP; /* Don't support different size */
739 if (X86_MODRM_MOD(opcode) != 3)
740 return -EOPNOTSUPP; /* TODO: support memory addressing */
741
742 p->ainsn.indirect.reg = X86_MODRM_RM(opcode);
743 #ifdef CONFIG_X86_64
744 if (X86_REX_B(insn->rex_prefix.value))
745 p->ainsn.indirect.reg += 8;
746 #endif
747 break;
748 default:
749 break;
750 }
751 p->ainsn.size = insn->length;
752
753 return 0;
754 }
755
756 static int arch_copy_kprobe(struct kprobe *p)
757 {
758 struct insn insn;
759 kprobe_opcode_t buf[MAX_INSN_SIZE];
760 int ret, len;
761
762 /* Copy an instruction with recovering if other optprobe modifies it.*/
763 len = __copy_instruction(buf, p->addr, p->ainsn.insn, &insn);
764 if (!len)
765 return -EINVAL;
766
767 /* Analyze the opcode and setup emulate functions */
768 ret = prepare_emulation(p, &insn);
769 if (ret < 0)
770 return ret;
771
772 /* Add int3 for single-step or booster jmp */
773 len = prepare_singlestep(buf, p, &insn);
774 if (len < 0)
775 return len;
776
777 /* Also, displacement change doesn't affect the first byte */
778 p->opcode = buf[0];
779
780 p->ainsn.tp_len = len;
781 perf_event_text_poke(p->ainsn.insn, NULL, 0, buf, len);
782
783 /* OK, write back the instruction(s) into ROX insn buffer */
784 text_poke(p->ainsn.insn, buf, len);
785
786 return 0;
787 }
788
789 int arch_prepare_kprobe(struct kprobe *p)
790 {
791 int ret;
792
793 if (alternatives_text_reserved(p->addr, p->addr))
794 return -EINVAL;
795
796 if (!can_probe((unsigned long)p->addr))
797 return -EILSEQ;
798
799 memset(&p->ainsn, 0, sizeof(p->ainsn));
800
801 /* insn: must be on special executable page on x86. */
802 p->ainsn.insn = get_insn_slot();
803 if (!p->ainsn.insn)
804 return -ENOMEM;
805
806 ret = arch_copy_kprobe(p);
807 if (ret) {
808 free_insn_slot(p->ainsn.insn, 0);
809 p->ainsn.insn = NULL;
810 }
811
812 return ret;
813 }
814
815 void arch_arm_kprobe(struct kprobe *p)
816 {
817 u8 int3 = INT3_INSN_OPCODE;
818
819 text_poke(p->addr, &int3, 1);
820 text_poke_sync();
821 perf_event_text_poke(p->addr, &p->opcode, 1, &int3, 1);
822 }
823
824 void arch_disarm_kprobe(struct kprobe *p)
825 {
826 u8 int3 = INT3_INSN_OPCODE;
827
828 perf_event_text_poke(p->addr, &int3, 1, &p->opcode, 1);
829 text_poke(p->addr, &p->opcode, 1);
830 text_poke_sync();
831 }
832
833 void arch_remove_kprobe(struct kprobe *p)
834 {
835 if (p->ainsn.insn) {
836 /* Record the perf event before freeing the slot */
837 perf_event_text_poke(p->ainsn.insn, p->ainsn.insn,
838 p->ainsn.tp_len, NULL, 0);
839 free_insn_slot(p->ainsn.insn, p->ainsn.boostable);
840 p->ainsn.insn = NULL;
841 }
842 }
843
844 static nokprobe_inline void
845 save_previous_kprobe(struct kprobe_ctlblk *kcb)
846 {
847 kcb->prev_kprobe.kp = kprobe_running();
848 kcb->prev_kprobe.status = kcb->kprobe_status;
849 kcb->prev_kprobe.old_flags = kcb->kprobe_old_flags;
850 kcb->prev_kprobe.saved_flags = kcb->kprobe_saved_flags;
851 }
852
853 static nokprobe_inline void
854 restore_previous_kprobe(struct kprobe_ctlblk *kcb)
855 {
856 __this_cpu_write(current_kprobe, kcb->prev_kprobe.kp);
857 kcb->kprobe_status = kcb->prev_kprobe.status;
858 kcb->kprobe_old_flags = kcb->prev_kprobe.old_flags;
859 kcb->kprobe_saved_flags = kcb->prev_kprobe.saved_flags;
860 }
861
862 static nokprobe_inline void
863 set_current_kprobe(struct kprobe *p, struct pt_regs *regs,
864 struct kprobe_ctlblk *kcb)
865 {
866 __this_cpu_write(current_kprobe, p);
867 kcb->kprobe_saved_flags = kcb->kprobe_old_flags
868 = (regs->flags & X86_EFLAGS_IF);
869 }
870
871 static void kprobe_post_process(struct kprobe *cur, struct pt_regs *regs,
872 struct kprobe_ctlblk *kcb)
873 {
874 /* Restore back the original saved kprobes variables and continue. */
875 if (kcb->kprobe_status == KPROBE_REENTER) {
876 /* This will restore both kcb and current_kprobe */
877 restore_previous_kprobe(kcb);
878 } else {
879 /*
880 * Always update the kcb status because
881 * reset_curent_kprobe() doesn't update kcb.
882 */
883 kcb->kprobe_status = KPROBE_HIT_SSDONE;
884 if (cur->post_handler)
885 cur->post_handler(cur, regs, 0);
886 reset_current_kprobe();
887 }
888 }
889 NOKPROBE_SYMBOL(kprobe_post_process);
890
891 static void setup_singlestep(struct kprobe *p, struct pt_regs *regs,
892 struct kprobe_ctlblk *kcb, int reenter)
893 {
894 if (setup_detour_execution(p, regs, reenter))
895 return;
896
897 #if !defined(CONFIG_PREEMPTION)
898 if (p->ainsn.boostable) {
899 /* Boost up -- we can execute copied instructions directly */
900 if (!reenter)
901 reset_current_kprobe();
902 /*
903 * Reentering boosted probe doesn't reset current_kprobe,
904 * nor set current_kprobe, because it doesn't use single
905 * stepping.
906 */
907 regs->ip = (unsigned long)p->ainsn.insn;
908 return;
909 }
910 #endif
911 if (reenter) {
912 save_previous_kprobe(kcb);
913 set_current_kprobe(p, regs, kcb);
914 kcb->kprobe_status = KPROBE_REENTER;
915 } else
916 kcb->kprobe_status = KPROBE_HIT_SS;
917
918 if (p->ainsn.emulate_op) {
919 p->ainsn.emulate_op(p, regs);
920 kprobe_post_process(p, regs, kcb);
921 return;
922 }
923
924 /* Disable interrupt, and set ip register on trampoline */
925 regs->flags &= ~X86_EFLAGS_IF;
926 regs->ip = (unsigned long)p->ainsn.insn;
927 }
928 NOKPROBE_SYMBOL(setup_singlestep);
929
930 /*
931 * Called after single-stepping. p->addr is the address of the
932 * instruction whose first byte has been replaced by the "int3"
933 * instruction. To avoid the SMP problems that can occur when we
934 * temporarily put back the original opcode to single-step, we
935 * single-stepped a copy of the instruction. The address of this
936 * copy is p->ainsn.insn. We also doesn't use trap, but "int3" again
937 * right after the copied instruction.
938 * Different from the trap single-step, "int3" single-step can not
939 * handle the instruction which changes the ip register, e.g. jmp,
940 * call, conditional jmp, and the instructions which changes the IF
941 * flags because interrupt must be disabled around the single-stepping.
942 * Such instructions are software emulated, but others are single-stepped
943 * using "int3".
944 *
945 * When the 2nd "int3" handled, the regs->ip and regs->flags needs to
946 * be adjusted, so that we can resume execution on correct code.
947 */
948 static void resume_singlestep(struct kprobe *p, struct pt_regs *regs,
949 struct kprobe_ctlblk *kcb)
950 {
951 unsigned long copy_ip = (unsigned long)p->ainsn.insn;
952 unsigned long orig_ip = (unsigned long)p->addr;
953
954 /* Restore saved interrupt flag and ip register */
955 regs->flags |= kcb->kprobe_saved_flags;
956 /* Note that regs->ip is executed int3 so must be a step back */
957 regs->ip += (orig_ip - copy_ip) - INT3_INSN_SIZE;
958 }
959 NOKPROBE_SYMBOL(resume_singlestep);
960
961 /*
962 * We have reentered the kprobe_handler(), since another probe was hit while
963 * within the handler. We save the original kprobes variables and just single
964 * step on the instruction of the new probe without calling any user handlers.
965 */
966 static int reenter_kprobe(struct kprobe *p, struct pt_regs *regs,
967 struct kprobe_ctlblk *kcb)
968 {
969 switch (kcb->kprobe_status) {
970 case KPROBE_HIT_SSDONE:
971 case KPROBE_HIT_ACTIVE:
972 case KPROBE_HIT_SS:
973 kprobes_inc_nmissed_count(p);
974 setup_singlestep(p, regs, kcb, 1);
975 break;
976 case KPROBE_REENTER:
977 /* A probe has been hit in the codepath leading up to, or just
978 * after, single-stepping of a probed instruction. This entire
979 * codepath should strictly reside in .kprobes.text section.
980 * Raise a BUG or we'll continue in an endless reentering loop
981 * and eventually a stack overflow.
982 */
983 pr_err("Unrecoverable kprobe detected.\n");
984 dump_kprobe(p);
985 BUG();
986 default:
987 /* impossible cases */
988 WARN_ON(1);
989 return 0;
990 }
991
992 return 1;
993 }
994 NOKPROBE_SYMBOL(reenter_kprobe);
995
996 static nokprobe_inline int kprobe_is_ss(struct kprobe_ctlblk *kcb)
997 {
998 return (kcb->kprobe_status == KPROBE_HIT_SS ||
999 kcb->kprobe_status == KPROBE_REENTER);
1000 }
1001
1002 /*
1003 * Interrupts are disabled on entry as trap3 is an interrupt gate and they
1004 * remain disabled throughout this function.
1005 */
1006 int kprobe_int3_handler(struct pt_regs *regs)
1007 {
1008 kprobe_opcode_t *addr;
1009 struct kprobe *p;
1010 struct kprobe_ctlblk *kcb;
1011
1012 if (user_mode(regs))
1013 return 0;
1014
1015 addr = (kprobe_opcode_t *)(regs->ip - sizeof(kprobe_opcode_t));
1016 /*
1017 * We don't want to be preempted for the entire duration of kprobe
1018 * processing. Since int3 and debug trap disables irqs and we clear
1019 * IF while singlestepping, it must be no preemptible.
1020 */
1021
1022 kcb = get_kprobe_ctlblk();
1023 p = get_kprobe(addr);
1024
1025 if (p) {
1026 if (kprobe_running()) {
1027 if (reenter_kprobe(p, regs, kcb))
1028 return 1;
1029 } else {
1030 set_current_kprobe(p, regs, kcb);
1031 kcb->kprobe_status = KPROBE_HIT_ACTIVE;
1032
1033 /*
1034 * If we have no pre-handler or it returned 0, we
1035 * continue with normal processing. If we have a
1036 * pre-handler and it returned non-zero, that means
1037 * user handler setup registers to exit to another
1038 * instruction, we must skip the single stepping.
1039 */
1040 if (!p->pre_handler || !p->pre_handler(p, regs))
1041 setup_singlestep(p, regs, kcb, 0);
1042 else
1043 reset_current_kprobe();
1044 return 1;
1045 }
1046 } else if (kprobe_is_ss(kcb)) {
1047 p = kprobe_running();
1048 if ((unsigned long)p->ainsn.insn < regs->ip &&
1049 (unsigned long)p->ainsn.insn + MAX_INSN_SIZE > regs->ip) {
1050 /* Most provably this is the second int3 for singlestep */
1051 resume_singlestep(p, regs, kcb);
1052 kprobe_post_process(p, regs, kcb);
1053 return 1;
1054 }
1055 } /* else: not a kprobe fault; let the kernel handle it */
1056
1057 return 0;
1058 }
1059 NOKPROBE_SYMBOL(kprobe_int3_handler);
1060
1061 int kprobe_fault_handler(struct pt_regs *regs, int trapnr)
1062 {
1063 struct kprobe *cur = kprobe_running();
1064 struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
1065
1066 if (unlikely(regs->ip == (unsigned long)cur->ainsn.insn)) {
1067 /* This must happen on single-stepping */
1068 WARN_ON(kcb->kprobe_status != KPROBE_HIT_SS &&
1069 kcb->kprobe_status != KPROBE_REENTER);
1070 /*
1071 * We are here because the instruction being single
1072 * stepped caused a page fault. We reset the current
1073 * kprobe and the ip points back to the probe address
1074 * and allow the page fault handler to continue as a
1075 * normal page fault.
1076 */
1077 regs->ip = (unsigned long)cur->addr;
1078
1079 /*
1080 * If the IF flag was set before the kprobe hit,
1081 * don't touch it:
1082 */
1083 regs->flags |= kcb->kprobe_old_flags;
1084
1085 if (kcb->kprobe_status == KPROBE_REENTER)
1086 restore_previous_kprobe(kcb);
1087 else
1088 reset_current_kprobe();
1089 }
1090
1091 return 0;
1092 }
1093 NOKPROBE_SYMBOL(kprobe_fault_handler);
1094
1095 int __init arch_populate_kprobe_blacklist(void)
1096 {
1097 return kprobe_add_area_blacklist((unsigned long)__entry_text_start,
1098 (unsigned long)__entry_text_end);
1099 }
1100
1101 int __init arch_init_kprobes(void)
1102 {
1103 return 0;
1104 }
1105
1106 int arch_trampoline_kprobe(struct kprobe *p)
1107 {
1108 return 0;
1109 }
1110
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.