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