xref: /openbmc/linux/arch/x86/kernel/kprobes/core.c (revision 35267cea)
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/ftrace.h>
41 #include <linux/kasan.h>
42 #include <linux/moduleloader.h>
43 #include <linux/objtool.h>
44 #include <linux/vmalloc.h>
45 #include <linux/pgtable.h>
46 
47 #include <asm/text-patching.h>
48 #include <asm/cacheflush.h>
49 #include <asm/desc.h>
50 #include <linux/uaccess.h>
51 #include <asm/alternative.h>
52 #include <asm/insn.h>
53 #include <asm/debugreg.h>
54 #include <asm/set_memory.h>
55 
56 #include "common.h"
57 
58 DEFINE_PER_CPU(struct kprobe *, current_kprobe) = NULL;
59 DEFINE_PER_CPU(struct kprobe_ctlblk, kprobe_ctlblk);
60 
61 #define stack_addr(regs) ((unsigned long *)regs->sp)
62 
63 #define W(row, b0, b1, b2, b3, b4, b5, b6, b7, b8, b9, ba, bb, bc, bd, be, bf)\
64 	(((b0##UL << 0x0)|(b1##UL << 0x1)|(b2##UL << 0x2)|(b3##UL << 0x3) |   \
65 	  (b4##UL << 0x4)|(b5##UL << 0x5)|(b6##UL << 0x6)|(b7##UL << 0x7) |   \
66 	  (b8##UL << 0x8)|(b9##UL << 0x9)|(ba##UL << 0xa)|(bb##UL << 0xb) |   \
67 	  (bc##UL << 0xc)|(bd##UL << 0xd)|(be##UL << 0xe)|(bf##UL << 0xf))    \
68 	 << (row % 32))
69 	/*
70 	 * Undefined/reserved opcodes, conditional jump, Opcode Extension
71 	 * Groups, and some special opcodes can not boost.
72 	 * This is non-const and volatile to keep gcc from statically
73 	 * optimizing it out, as variable_test_bit makes gcc think only
74 	 * *(unsigned long*) is used.
75 	 */
76 static volatile u32 twobyte_is_boostable[256 / 32] = {
77 	/*      0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f          */
78 	/*      ----------------------------------------------          */
79 	W(0x00, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0) | /* 00 */
80 	W(0x10, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1) , /* 10 */
81 	W(0x20, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) | /* 20 */
82 	W(0x30, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) , /* 30 */
83 	W(0x40, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* 40 */
84 	W(0x50, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) , /* 50 */
85 	W(0x60, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1) | /* 60 */
86 	W(0x70, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1) , /* 70 */
87 	W(0x80, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) | /* 80 */
88 	W(0x90, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 90 */
89 	W(0xa0, 1, 1, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1, 1, 1, 0, 1) | /* a0 */
90 	W(0xb0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1) , /* b0 */
91 	W(0xc0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1) | /* c0 */
92 	W(0xd0, 0, 1, 1, 1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 0, 1) , /* d0 */
93 	W(0xe0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 0, 1) | /* e0 */
94 	W(0xf0, 0, 1, 1, 1, 0, 1, 0, 0, 1, 1, 1, 0, 1, 1, 1, 0)   /* f0 */
95 	/*      -----------------------------------------------         */
96 	/*      0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f          */
97 };
98 #undef W
99 
100 struct kretprobe_blackpoint kretprobe_blacklist[] = {
101 	{"__switch_to", }, /* This function switches only current task, but
102 			      doesn't switch kernel stack.*/
103 	{NULL, NULL}	/* Terminator */
104 };
105 
106 const int kretprobe_blacklist_size = ARRAY_SIZE(kretprobe_blacklist);
107 
108 static nokprobe_inline void
109 __synthesize_relative_insn(void *dest, void *from, void *to, u8 op)
110 {
111 	struct __arch_relative_insn {
112 		u8 op;
113 		s32 raddr;
114 	} __packed *insn;
115 
116 	insn = (struct __arch_relative_insn *)dest;
117 	insn->raddr = (s32)((long)(to) - ((long)(from) + 5));
118 	insn->op = op;
119 }
120 
121 /* Insert a jump instruction at address 'from', which jumps to address 'to'.*/
122 void synthesize_reljump(void *dest, void *from, void *to)
123 {
124 	__synthesize_relative_insn(dest, from, to, JMP32_INSN_OPCODE);
125 }
126 NOKPROBE_SYMBOL(synthesize_reljump);
127 
128 /* Insert a call instruction at address 'from', which calls address 'to'.*/
129 void synthesize_relcall(void *dest, void *from, void *to)
130 {
131 	__synthesize_relative_insn(dest, from, to, CALL_INSN_OPCODE);
132 }
133 NOKPROBE_SYMBOL(synthesize_relcall);
134 
135 /*
136  * Returns non-zero if INSN is boostable.
137  * RIP relative instructions are adjusted at copying time in 64 bits mode
138  */
139 int can_boost(struct insn *insn, void *addr)
140 {
141 	kprobe_opcode_t opcode;
142 	insn_byte_t prefix;
143 	int i;
144 
145 	if (search_exception_tables((unsigned long)addr))
146 		return 0;	/* 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 0;
155 
156 	for_each_insn_prefix(insn, i, 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 0;
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 0xc0 ... 0xc1:	/* Grp2 */
172 	case 0xcc ... 0xce:	/* software exceptions */
173 	case 0xd0 ... 0xd3:	/* Grp2 */
174 	case 0xd6:		/* (UD) */
175 	case 0xd8 ... 0xdf:	/* ESC */
176 	case 0xe0 ... 0xe3:	/* LOOP*, JCXZ */
177 	case 0xe8 ... 0xe9:	/* near Call, JMP */
178 	case 0xeb:		/* Short JMP */
179 	case 0xf0 ... 0xf4:	/* LOCK/REP, HLT */
180 	case 0xf6 ... 0xf7:	/* Grp3 */
181 	case 0xfe:		/* Grp4 */
182 		/* ... are not boostable */
183 		return 0;
184 	case 0xff:		/* Grp5 */
185 		/* Only indirect jmp is boostable */
186 		return X86_MODRM_REG(insn->modrm.bytes[0]) == 4;
187 	default:
188 		return 1;
189 	}
190 }
191 
192 static unsigned long
193 __recover_probed_insn(kprobe_opcode_t *buf, unsigned long addr)
194 {
195 	struct kprobe *kp;
196 	unsigned long faddr;
197 
198 	kp = get_kprobe((void *)addr);
199 	faddr = ftrace_location(addr);
200 	/*
201 	 * Addresses inside the ftrace location are refused by
202 	 * arch_check_ftrace_location(). Something went terribly wrong
203 	 * if such an address is checked here.
204 	 */
205 	if (WARN_ON(faddr && faddr != addr))
206 		return 0UL;
207 	/*
208 	 * Use the current code if it is not modified by Kprobe
209 	 * and it cannot be modified by ftrace.
210 	 */
211 	if (!kp && !faddr)
212 		return addr;
213 
214 	/*
215 	 * Basically, kp->ainsn.insn has an original instruction.
216 	 * However, RIP-relative instruction can not do single-stepping
217 	 * at different place, __copy_instruction() tweaks the displacement of
218 	 * that instruction. In that case, we can't recover the instruction
219 	 * from the kp->ainsn.insn.
220 	 *
221 	 * On the other hand, in case on normal Kprobe, kp->opcode has a copy
222 	 * of the first byte of the probed instruction, which is overwritten
223 	 * by int3. And the instruction at kp->addr is not modified by kprobes
224 	 * except for the first byte, we can recover the original instruction
225 	 * from it and kp->opcode.
226 	 *
227 	 * In case of Kprobes using ftrace, we do not have a copy of
228 	 * the original instruction. In fact, the ftrace location might
229 	 * be modified at anytime and even could be in an inconsistent state.
230 	 * Fortunately, we know that the original code is the ideal 5-byte
231 	 * long NOP.
232 	 */
233 	if (copy_from_kernel_nofault(buf, (void *)addr,
234 		MAX_INSN_SIZE * sizeof(kprobe_opcode_t)))
235 		return 0UL;
236 
237 	if (faddr)
238 		memcpy(buf, x86_nops[5], 5);
239 	else
240 		buf[0] = kp->opcode;
241 	return (unsigned long)buf;
242 }
243 
244 /*
245  * Recover the probed instruction at addr for further analysis.
246  * Caller must lock kprobes by kprobe_mutex, or disable preemption
247  * for preventing to release referencing kprobes.
248  * Returns zero if the instruction can not get recovered (or access failed).
249  */
250 unsigned long recover_probed_instruction(kprobe_opcode_t *buf, unsigned long addr)
251 {
252 	unsigned long __addr;
253 
254 	__addr = __recover_optprobed_insn(buf, addr);
255 	if (__addr != addr)
256 		return __addr;
257 
258 	return __recover_probed_insn(buf, addr);
259 }
260 
261 /* Check if paddr is at an instruction boundary */
262 static int can_probe(unsigned long paddr)
263 {
264 	unsigned long addr, __addr, offset = 0;
265 	struct insn insn;
266 	kprobe_opcode_t buf[MAX_INSN_SIZE];
267 
268 	if (!kallsyms_lookup_size_offset(paddr, NULL, &offset))
269 		return 0;
270 
271 	/* Decode instructions */
272 	addr = paddr - offset;
273 	while (addr < paddr) {
274 		int ret;
275 
276 		/*
277 		 * Check if the instruction has been modified by another
278 		 * kprobe, in which case we replace the breakpoint by the
279 		 * original instruction in our buffer.
280 		 * Also, jump optimization will change the breakpoint to
281 		 * relative-jump. Since the relative-jump itself is
282 		 * normally used, we just go through if there is no kprobe.
283 		 */
284 		__addr = recover_probed_instruction(buf, addr);
285 		if (!__addr)
286 			return 0;
287 
288 		ret = insn_decode_kernel(&insn, (void *)__addr);
289 		if (ret < 0)
290 			return 0;
291 
292 		/*
293 		 * Another debugging subsystem might insert this breakpoint.
294 		 * In that case, we can't recover it.
295 		 */
296 		if (insn.opcode.bytes[0] == INT3_INSN_OPCODE)
297 			return 0;
298 		addr += insn.length;
299 	}
300 
301 	return (addr == paddr);
302 }
303 
304 /*
305  * Copy an instruction with recovering modified instruction by kprobes
306  * and adjust the displacement if the instruction uses the %rip-relative
307  * addressing mode. Note that since @real will be the final place of copied
308  * instruction, displacement must be adjust by @real, not @dest.
309  * This returns the length of copied instruction, or 0 if it has an error.
310  */
311 int __copy_instruction(u8 *dest, u8 *src, u8 *real, struct insn *insn)
312 {
313 	kprobe_opcode_t buf[MAX_INSN_SIZE];
314 	unsigned long recovered_insn = recover_probed_instruction(buf, (unsigned long)src);
315 	int ret;
316 
317 	if (!recovered_insn || !insn)
318 		return 0;
319 
320 	/* This can access kernel text if given address is not recovered */
321 	if (copy_from_kernel_nofault(dest, (void *)recovered_insn,
322 			MAX_INSN_SIZE))
323 		return 0;
324 
325 	ret = insn_decode_kernel(insn, dest);
326 	if (ret < 0)
327 		return 0;
328 
329 	/* We can not probe force emulate prefixed instruction */
330 	if (insn_has_emulate_prefix(insn))
331 		return 0;
332 
333 	/* Another subsystem puts a breakpoint, failed to recover */
334 	if (insn->opcode.bytes[0] == INT3_INSN_OPCODE)
335 		return 0;
336 
337 	/* We should not singlestep on the exception masking instructions */
338 	if (insn_masking_exception(insn))
339 		return 0;
340 
341 #ifdef CONFIG_X86_64
342 	/* Only x86_64 has RIP relative instructions */
343 	if (insn_rip_relative(insn)) {
344 		s64 newdisp;
345 		u8 *disp;
346 		/*
347 		 * The copied instruction uses the %rip-relative addressing
348 		 * mode.  Adjust the displacement for the difference between
349 		 * the original location of this instruction and the location
350 		 * of the copy that will actually be run.  The tricky bit here
351 		 * is making sure that the sign extension happens correctly in
352 		 * this calculation, since we need a signed 32-bit result to
353 		 * be sign-extended to 64 bits when it's added to the %rip
354 		 * value and yield the same 64-bit result that the sign-
355 		 * extension of the original signed 32-bit displacement would
356 		 * have given.
357 		 */
358 		newdisp = (u8 *) src + (s64) insn->displacement.value
359 			  - (u8 *) real;
360 		if ((s64) (s32) newdisp != newdisp) {
361 			pr_err("Kprobes error: new displacement does not fit into s32 (%llx)\n", newdisp);
362 			return 0;
363 		}
364 		disp = (u8 *) dest + insn_offset_displacement(insn);
365 		*(s32 *) disp = (s32) newdisp;
366 	}
367 #endif
368 	return insn->length;
369 }
370 
371 /* Prepare reljump or int3 right after instruction */
372 static int prepare_singlestep(kprobe_opcode_t *buf, struct kprobe *p,
373 			      struct insn *insn)
374 {
375 	int len = insn->length;
376 
377 	if (!IS_ENABLED(CONFIG_PREEMPTION) &&
378 	    !p->post_handler && can_boost(insn, p->addr) &&
379 	    MAX_INSN_SIZE - len >= JMP32_INSN_SIZE) {
380 		/*
381 		 * These instructions can be executed directly if it
382 		 * jumps back to correct address.
383 		 */
384 		synthesize_reljump(buf + len, p->ainsn.insn + len,
385 				   p->addr + insn->length);
386 		len += JMP32_INSN_SIZE;
387 		p->ainsn.boostable = 1;
388 	} else {
389 		/* Otherwise, put an int3 for trapping singlestep */
390 		if (MAX_INSN_SIZE - len < INT3_INSN_SIZE)
391 			return -ENOSPC;
392 
393 		buf[len] = INT3_INSN_OPCODE;
394 		len += INT3_INSN_SIZE;
395 	}
396 
397 	return len;
398 }
399 
400 /* Make page to RO mode when allocate it */
401 void *alloc_insn_page(void)
402 {
403 	void *page;
404 
405 	page = module_alloc(PAGE_SIZE);
406 	if (!page)
407 		return NULL;
408 
409 	set_vm_flush_reset_perms(page);
410 	/*
411 	 * First make the page read-only, and only then make it executable to
412 	 * prevent it from being W+X in between.
413 	 */
414 	set_memory_ro((unsigned long)page, 1);
415 
416 	/*
417 	 * TODO: Once additional kernel code protection mechanisms are set, ensure
418 	 * that the page was not maliciously altered and it is still zeroed.
419 	 */
420 	set_memory_x((unsigned long)page, 1);
421 
422 	return page;
423 }
424 
425 /* Kprobe x86 instruction emulation - only regs->ip or IF flag modifiers */
426 
427 static void kprobe_emulate_ifmodifiers(struct kprobe *p, struct pt_regs *regs)
428 {
429 	switch (p->ainsn.opcode) {
430 	case 0xfa:	/* cli */
431 		regs->flags &= ~(X86_EFLAGS_IF);
432 		break;
433 	case 0xfb:	/* sti */
434 		regs->flags |= X86_EFLAGS_IF;
435 		break;
436 	case 0x9c:	/* pushf */
437 		int3_emulate_push(regs, regs->flags);
438 		break;
439 	case 0x9d:	/* popf */
440 		regs->flags = int3_emulate_pop(regs);
441 		break;
442 	}
443 	regs->ip = regs->ip - INT3_INSN_SIZE + p->ainsn.size;
444 }
445 NOKPROBE_SYMBOL(kprobe_emulate_ifmodifiers);
446 
447 static void kprobe_emulate_ret(struct kprobe *p, struct pt_regs *regs)
448 {
449 	int3_emulate_ret(regs);
450 }
451 NOKPROBE_SYMBOL(kprobe_emulate_ret);
452 
453 static void kprobe_emulate_call(struct kprobe *p, struct pt_regs *regs)
454 {
455 	unsigned long func = regs->ip - INT3_INSN_SIZE + p->ainsn.size;
456 
457 	func += p->ainsn.rel32;
458 	int3_emulate_call(regs, func);
459 }
460 NOKPROBE_SYMBOL(kprobe_emulate_call);
461 
462 static nokprobe_inline
463 void __kprobe_emulate_jmp(struct kprobe *p, struct pt_regs *regs, bool cond)
464 {
465 	unsigned long ip = regs->ip - INT3_INSN_SIZE + p->ainsn.size;
466 
467 	if (cond)
468 		ip += p->ainsn.rel32;
469 	int3_emulate_jmp(regs, ip);
470 }
471 
472 static void kprobe_emulate_jmp(struct kprobe *p, struct pt_regs *regs)
473 {
474 	__kprobe_emulate_jmp(p, regs, true);
475 }
476 NOKPROBE_SYMBOL(kprobe_emulate_jmp);
477 
478 static const unsigned long jcc_mask[6] = {
479 	[0] = X86_EFLAGS_OF,
480 	[1] = X86_EFLAGS_CF,
481 	[2] = X86_EFLAGS_ZF,
482 	[3] = X86_EFLAGS_CF | X86_EFLAGS_ZF,
483 	[4] = X86_EFLAGS_SF,
484 	[5] = X86_EFLAGS_PF,
485 };
486 
487 static void kprobe_emulate_jcc(struct kprobe *p, struct pt_regs *regs)
488 {
489 	bool invert = p->ainsn.jcc.type & 1;
490 	bool match;
491 
492 	if (p->ainsn.jcc.type < 0xc) {
493 		match = regs->flags & jcc_mask[p->ainsn.jcc.type >> 1];
494 	} else {
495 		match = ((regs->flags & X86_EFLAGS_SF) >> X86_EFLAGS_SF_BIT) ^
496 			((regs->flags & X86_EFLAGS_OF) >> X86_EFLAGS_OF_BIT);
497 		if (p->ainsn.jcc.type >= 0xe)
498 			match = match && (regs->flags & X86_EFLAGS_ZF);
499 	}
500 	__kprobe_emulate_jmp(p, regs, (match && !invert) || (!match && invert));
501 }
502 NOKPROBE_SYMBOL(kprobe_emulate_jcc);
503 
504 static void kprobe_emulate_loop(struct kprobe *p, struct pt_regs *regs)
505 {
506 	bool match;
507 
508 	if (p->ainsn.loop.type != 3) {	/* LOOP* */
509 		if (p->ainsn.loop.asize == 32)
510 			match = ((*(u32 *)&regs->cx)--) != 0;
511 #ifdef CONFIG_X86_64
512 		else if (p->ainsn.loop.asize == 64)
513 			match = ((*(u64 *)&regs->cx)--) != 0;
514 #endif
515 		else
516 			match = ((*(u16 *)&regs->cx)--) != 0;
517 	} else {			/* JCXZ */
518 		if (p->ainsn.loop.asize == 32)
519 			match = *(u32 *)(&regs->cx) == 0;
520 #ifdef CONFIG_X86_64
521 		else if (p->ainsn.loop.asize == 64)
522 			match = *(u64 *)(&regs->cx) == 0;
523 #endif
524 		else
525 			match = *(u16 *)(&regs->cx) == 0;
526 	}
527 
528 	if (p->ainsn.loop.type == 0)	/* LOOPNE */
529 		match = match && !(regs->flags & X86_EFLAGS_ZF);
530 	else if (p->ainsn.loop.type == 1)	/* LOOPE */
531 		match = match && (regs->flags & X86_EFLAGS_ZF);
532 
533 	__kprobe_emulate_jmp(p, regs, match);
534 }
535 NOKPROBE_SYMBOL(kprobe_emulate_loop);
536 
537 static const int addrmode_regoffs[] = {
538 	offsetof(struct pt_regs, ax),
539 	offsetof(struct pt_regs, cx),
540 	offsetof(struct pt_regs, dx),
541 	offsetof(struct pt_regs, bx),
542 	offsetof(struct pt_regs, sp),
543 	offsetof(struct pt_regs, bp),
544 	offsetof(struct pt_regs, si),
545 	offsetof(struct pt_regs, di),
546 #ifdef CONFIG_X86_64
547 	offsetof(struct pt_regs, r8),
548 	offsetof(struct pt_regs, r9),
549 	offsetof(struct pt_regs, r10),
550 	offsetof(struct pt_regs, r11),
551 	offsetof(struct pt_regs, r12),
552 	offsetof(struct pt_regs, r13),
553 	offsetof(struct pt_regs, r14),
554 	offsetof(struct pt_regs, r15),
555 #endif
556 };
557 
558 static void kprobe_emulate_call_indirect(struct kprobe *p, struct pt_regs *regs)
559 {
560 	unsigned long offs = addrmode_regoffs[p->ainsn.indirect.reg];
561 
562 	int3_emulate_call(regs, regs_get_register(regs, offs));
563 }
564 NOKPROBE_SYMBOL(kprobe_emulate_call_indirect);
565 
566 static void kprobe_emulate_jmp_indirect(struct kprobe *p, struct pt_regs *regs)
567 {
568 	unsigned long offs = addrmode_regoffs[p->ainsn.indirect.reg];
569 
570 	int3_emulate_jmp(regs, regs_get_register(regs, offs));
571 }
572 NOKPROBE_SYMBOL(kprobe_emulate_jmp_indirect);
573 
574 static int prepare_emulation(struct kprobe *p, struct insn *insn)
575 {
576 	insn_byte_t opcode = insn->opcode.bytes[0];
577 
578 	switch (opcode) {
579 	case 0xfa:		/* cli */
580 	case 0xfb:		/* sti */
581 	case 0x9c:		/* pushfl */
582 	case 0x9d:		/* popf/popfd */
583 		/*
584 		 * IF modifiers must be emulated since it will enable interrupt while
585 		 * int3 single stepping.
586 		 */
587 		p->ainsn.emulate_op = kprobe_emulate_ifmodifiers;
588 		p->ainsn.opcode = opcode;
589 		break;
590 	case 0xc2:	/* ret/lret */
591 	case 0xc3:
592 	case 0xca:
593 	case 0xcb:
594 		p->ainsn.emulate_op = kprobe_emulate_ret;
595 		break;
596 	case 0x9a:	/* far call absolute -- segment is not supported */
597 	case 0xea:	/* far jmp absolute -- segment is not supported */
598 	case 0xcc:	/* int3 */
599 	case 0xcf:	/* iret -- in-kernel IRET is not supported */
600 		return -EOPNOTSUPP;
601 		break;
602 	case 0xe8:	/* near call relative */
603 		p->ainsn.emulate_op = kprobe_emulate_call;
604 		if (insn->immediate.nbytes == 2)
605 			p->ainsn.rel32 = *(s16 *)&insn->immediate.value;
606 		else
607 			p->ainsn.rel32 = *(s32 *)&insn->immediate.value;
608 		break;
609 	case 0xeb:	/* short jump relative */
610 	case 0xe9:	/* near jump relative */
611 		p->ainsn.emulate_op = kprobe_emulate_jmp;
612 		if (insn->immediate.nbytes == 1)
613 			p->ainsn.rel32 = *(s8 *)&insn->immediate.value;
614 		else if (insn->immediate.nbytes == 2)
615 			p->ainsn.rel32 = *(s16 *)&insn->immediate.value;
616 		else
617 			p->ainsn.rel32 = *(s32 *)&insn->immediate.value;
618 		break;
619 	case 0x70 ... 0x7f:
620 		/* 1 byte conditional jump */
621 		p->ainsn.emulate_op = kprobe_emulate_jcc;
622 		p->ainsn.jcc.type = opcode & 0xf;
623 		p->ainsn.rel32 = *(char *)insn->immediate.bytes;
624 		break;
625 	case 0x0f:
626 		opcode = insn->opcode.bytes[1];
627 		if ((opcode & 0xf0) == 0x80) {
628 			/* 2 bytes Conditional Jump */
629 			p->ainsn.emulate_op = kprobe_emulate_jcc;
630 			p->ainsn.jcc.type = opcode & 0xf;
631 			if (insn->immediate.nbytes == 2)
632 				p->ainsn.rel32 = *(s16 *)&insn->immediate.value;
633 			else
634 				p->ainsn.rel32 = *(s32 *)&insn->immediate.value;
635 		} else if (opcode == 0x01 &&
636 			   X86_MODRM_REG(insn->modrm.bytes[0]) == 0 &&
637 			   X86_MODRM_MOD(insn->modrm.bytes[0]) == 3) {
638 			/* VM extensions - not supported */
639 			return -EOPNOTSUPP;
640 		}
641 		break;
642 	case 0xe0:	/* Loop NZ */
643 	case 0xe1:	/* Loop */
644 	case 0xe2:	/* Loop */
645 	case 0xe3:	/* J*CXZ */
646 		p->ainsn.emulate_op = kprobe_emulate_loop;
647 		p->ainsn.loop.type = opcode & 0x3;
648 		p->ainsn.loop.asize = insn->addr_bytes * 8;
649 		p->ainsn.rel32 = *(s8 *)&insn->immediate.value;
650 		break;
651 	case 0xff:
652 		/*
653 		 * Since the 0xff is an extended group opcode, the instruction
654 		 * is determined by the MOD/RM byte.
655 		 */
656 		opcode = insn->modrm.bytes[0];
657 		if ((opcode & 0x30) == 0x10) {
658 			if ((opcode & 0x8) == 0x8)
659 				return -EOPNOTSUPP;	/* far call */
660 			/* call absolute, indirect */
661 			p->ainsn.emulate_op = kprobe_emulate_call_indirect;
662 		} else if ((opcode & 0x30) == 0x20) {
663 			if ((opcode & 0x8) == 0x8)
664 				return -EOPNOTSUPP;	/* far jmp */
665 			/* jmp near absolute indirect */
666 			p->ainsn.emulate_op = kprobe_emulate_jmp_indirect;
667 		} else
668 			break;
669 
670 		if (insn->addr_bytes != sizeof(unsigned long))
671 			return -EOPNOTSUPP;	/* Don't support different size */
672 		if (X86_MODRM_MOD(opcode) != 3)
673 			return -EOPNOTSUPP;	/* TODO: support memory addressing */
674 
675 		p->ainsn.indirect.reg = X86_MODRM_RM(opcode);
676 #ifdef CONFIG_X86_64
677 		if (X86_REX_B(insn->rex_prefix.value))
678 			p->ainsn.indirect.reg += 8;
679 #endif
680 		break;
681 	default:
682 		break;
683 	}
684 	p->ainsn.size = insn->length;
685 
686 	return 0;
687 }
688 
689 static int arch_copy_kprobe(struct kprobe *p)
690 {
691 	struct insn insn;
692 	kprobe_opcode_t buf[MAX_INSN_SIZE];
693 	int ret, len;
694 
695 	/* Copy an instruction with recovering if other optprobe modifies it.*/
696 	len = __copy_instruction(buf, p->addr, p->ainsn.insn, &insn);
697 	if (!len)
698 		return -EINVAL;
699 
700 	/* Analyze the opcode and setup emulate functions */
701 	ret = prepare_emulation(p, &insn);
702 	if (ret < 0)
703 		return ret;
704 
705 	/* Add int3 for single-step or booster jmp */
706 	len = prepare_singlestep(buf, p, &insn);
707 	if (len < 0)
708 		return len;
709 
710 	/* Also, displacement change doesn't affect the first byte */
711 	p->opcode = buf[0];
712 
713 	p->ainsn.tp_len = len;
714 	perf_event_text_poke(p->ainsn.insn, NULL, 0, buf, len);
715 
716 	/* OK, write back the instruction(s) into ROX insn buffer */
717 	text_poke(p->ainsn.insn, buf, len);
718 
719 	return 0;
720 }
721 
722 int arch_prepare_kprobe(struct kprobe *p)
723 {
724 	int ret;
725 
726 	if (alternatives_text_reserved(p->addr, p->addr))
727 		return -EINVAL;
728 
729 	if (!can_probe((unsigned long)p->addr))
730 		return -EILSEQ;
731 
732 	memset(&p->ainsn, 0, sizeof(p->ainsn));
733 
734 	/* insn: must be on special executable page on x86. */
735 	p->ainsn.insn = get_insn_slot();
736 	if (!p->ainsn.insn)
737 		return -ENOMEM;
738 
739 	ret = arch_copy_kprobe(p);
740 	if (ret) {
741 		free_insn_slot(p->ainsn.insn, 0);
742 		p->ainsn.insn = NULL;
743 	}
744 
745 	return ret;
746 }
747 
748 void arch_arm_kprobe(struct kprobe *p)
749 {
750 	u8 int3 = INT3_INSN_OPCODE;
751 
752 	text_poke(p->addr, &int3, 1);
753 	text_poke_sync();
754 	perf_event_text_poke(p->addr, &p->opcode, 1, &int3, 1);
755 }
756 
757 void arch_disarm_kprobe(struct kprobe *p)
758 {
759 	u8 int3 = INT3_INSN_OPCODE;
760 
761 	perf_event_text_poke(p->addr, &int3, 1, &p->opcode, 1);
762 	text_poke(p->addr, &p->opcode, 1);
763 	text_poke_sync();
764 }
765 
766 void arch_remove_kprobe(struct kprobe *p)
767 {
768 	if (p->ainsn.insn) {
769 		/* Record the perf event before freeing the slot */
770 		perf_event_text_poke(p->ainsn.insn, p->ainsn.insn,
771 				     p->ainsn.tp_len, NULL, 0);
772 		free_insn_slot(p->ainsn.insn, p->ainsn.boostable);
773 		p->ainsn.insn = NULL;
774 	}
775 }
776 
777 static nokprobe_inline void
778 save_previous_kprobe(struct kprobe_ctlblk *kcb)
779 {
780 	kcb->prev_kprobe.kp = kprobe_running();
781 	kcb->prev_kprobe.status = kcb->kprobe_status;
782 	kcb->prev_kprobe.old_flags = kcb->kprobe_old_flags;
783 	kcb->prev_kprobe.saved_flags = kcb->kprobe_saved_flags;
784 }
785 
786 static nokprobe_inline void
787 restore_previous_kprobe(struct kprobe_ctlblk *kcb)
788 {
789 	__this_cpu_write(current_kprobe, kcb->prev_kprobe.kp);
790 	kcb->kprobe_status = kcb->prev_kprobe.status;
791 	kcb->kprobe_old_flags = kcb->prev_kprobe.old_flags;
792 	kcb->kprobe_saved_flags = kcb->prev_kprobe.saved_flags;
793 }
794 
795 static nokprobe_inline void
796 set_current_kprobe(struct kprobe *p, struct pt_regs *regs,
797 		   struct kprobe_ctlblk *kcb)
798 {
799 	__this_cpu_write(current_kprobe, p);
800 	kcb->kprobe_saved_flags = kcb->kprobe_old_flags
801 		= (regs->flags & X86_EFLAGS_IF);
802 }
803 
804 void arch_prepare_kretprobe(struct kretprobe_instance *ri, struct pt_regs *regs)
805 {
806 	unsigned long *sara = stack_addr(regs);
807 
808 	ri->ret_addr = (kprobe_opcode_t *) *sara;
809 	ri->fp = sara;
810 
811 	/* Replace the return addr with trampoline addr */
812 	*sara = (unsigned long) &kretprobe_trampoline;
813 }
814 NOKPROBE_SYMBOL(arch_prepare_kretprobe);
815 
816 static void kprobe_post_process(struct kprobe *cur, struct pt_regs *regs,
817 			       struct kprobe_ctlblk *kcb)
818 {
819 	if ((kcb->kprobe_status != KPROBE_REENTER) && cur->post_handler) {
820 		kcb->kprobe_status = KPROBE_HIT_SSDONE;
821 		cur->post_handler(cur, regs, 0);
822 	}
823 
824 	/* Restore back the original saved kprobes variables and continue. */
825 	if (kcb->kprobe_status == KPROBE_REENTER)
826 		restore_previous_kprobe(kcb);
827 	else
828 		reset_current_kprobe();
829 }
830 NOKPROBE_SYMBOL(kprobe_post_process);
831 
832 static void setup_singlestep(struct kprobe *p, struct pt_regs *regs,
833 			     struct kprobe_ctlblk *kcb, int reenter)
834 {
835 	if (setup_detour_execution(p, regs, reenter))
836 		return;
837 
838 #if !defined(CONFIG_PREEMPTION)
839 	if (p->ainsn.boostable) {
840 		/* Boost up -- we can execute copied instructions directly */
841 		if (!reenter)
842 			reset_current_kprobe();
843 		/*
844 		 * Reentering boosted probe doesn't reset current_kprobe,
845 		 * nor set current_kprobe, because it doesn't use single
846 		 * stepping.
847 		 */
848 		regs->ip = (unsigned long)p->ainsn.insn;
849 		return;
850 	}
851 #endif
852 	if (reenter) {
853 		save_previous_kprobe(kcb);
854 		set_current_kprobe(p, regs, kcb);
855 		kcb->kprobe_status = KPROBE_REENTER;
856 	} else
857 		kcb->kprobe_status = KPROBE_HIT_SS;
858 
859 	if (p->ainsn.emulate_op) {
860 		p->ainsn.emulate_op(p, regs);
861 		kprobe_post_process(p, regs, kcb);
862 		return;
863 	}
864 
865 	/* Disable interrupt, and set ip register on trampoline */
866 	regs->flags &= ~X86_EFLAGS_IF;
867 	regs->ip = (unsigned long)p->ainsn.insn;
868 }
869 NOKPROBE_SYMBOL(setup_singlestep);
870 
871 /*
872  * Called after single-stepping.  p->addr is the address of the
873  * instruction whose first byte has been replaced by the "int3"
874  * instruction.  To avoid the SMP problems that can occur when we
875  * temporarily put back the original opcode to single-step, we
876  * single-stepped a copy of the instruction.  The address of this
877  * copy is p->ainsn.insn. We also doesn't use trap, but "int3" again
878  * right after the copied instruction.
879  * Different from the trap single-step, "int3" single-step can not
880  * handle the instruction which changes the ip register, e.g. jmp,
881  * call, conditional jmp, and the instructions which changes the IF
882  * flags because interrupt must be disabled around the single-stepping.
883  * Such instructions are software emulated, but others are single-stepped
884  * using "int3".
885  *
886  * When the 2nd "int3" handled, the regs->ip and regs->flags needs to
887  * be adjusted, so that we can resume execution on correct code.
888  */
889 static void resume_singlestep(struct kprobe *p, struct pt_regs *regs,
890 			      struct kprobe_ctlblk *kcb)
891 {
892 	unsigned long copy_ip = (unsigned long)p->ainsn.insn;
893 	unsigned long orig_ip = (unsigned long)p->addr;
894 
895 	/* Restore saved interrupt flag and ip register */
896 	regs->flags |= kcb->kprobe_saved_flags;
897 	/* Note that regs->ip is executed int3 so must be a step back */
898 	regs->ip += (orig_ip - copy_ip) - INT3_INSN_SIZE;
899 }
900 NOKPROBE_SYMBOL(resume_singlestep);
901 
902 /*
903  * We have reentered the kprobe_handler(), since another probe was hit while
904  * within the handler. We save the original kprobes variables and just single
905  * step on the instruction of the new probe without calling any user handlers.
906  */
907 static int reenter_kprobe(struct kprobe *p, struct pt_regs *regs,
908 			  struct kprobe_ctlblk *kcb)
909 {
910 	switch (kcb->kprobe_status) {
911 	case KPROBE_HIT_SSDONE:
912 	case KPROBE_HIT_ACTIVE:
913 	case KPROBE_HIT_SS:
914 		kprobes_inc_nmissed_count(p);
915 		setup_singlestep(p, regs, kcb, 1);
916 		break;
917 	case KPROBE_REENTER:
918 		/* A probe has been hit in the codepath leading up to, or just
919 		 * after, single-stepping of a probed instruction. This entire
920 		 * codepath should strictly reside in .kprobes.text section.
921 		 * Raise a BUG or we'll continue in an endless reentering loop
922 		 * and eventually a stack overflow.
923 		 */
924 		pr_err("Unrecoverable kprobe detected.\n");
925 		dump_kprobe(p);
926 		BUG();
927 	default:
928 		/* impossible cases */
929 		WARN_ON(1);
930 		return 0;
931 	}
932 
933 	return 1;
934 }
935 NOKPROBE_SYMBOL(reenter_kprobe);
936 
937 static nokprobe_inline int kprobe_is_ss(struct kprobe_ctlblk *kcb)
938 {
939 	return (kcb->kprobe_status == KPROBE_HIT_SS ||
940 		kcb->kprobe_status == KPROBE_REENTER);
941 }
942 
943 /*
944  * Interrupts are disabled on entry as trap3 is an interrupt gate and they
945  * remain disabled throughout this function.
946  */
947 int kprobe_int3_handler(struct pt_regs *regs)
948 {
949 	kprobe_opcode_t *addr;
950 	struct kprobe *p;
951 	struct kprobe_ctlblk *kcb;
952 
953 	if (user_mode(regs))
954 		return 0;
955 
956 	addr = (kprobe_opcode_t *)(regs->ip - sizeof(kprobe_opcode_t));
957 	/*
958 	 * We don't want to be preempted for the entire duration of kprobe
959 	 * processing. Since int3 and debug trap disables irqs and we clear
960 	 * IF while singlestepping, it must be no preemptible.
961 	 */
962 
963 	kcb = get_kprobe_ctlblk();
964 	p = get_kprobe(addr);
965 
966 	if (p) {
967 		if (kprobe_running()) {
968 			if (reenter_kprobe(p, regs, kcb))
969 				return 1;
970 		} else {
971 			set_current_kprobe(p, regs, kcb);
972 			kcb->kprobe_status = KPROBE_HIT_ACTIVE;
973 
974 			/*
975 			 * If we have no pre-handler or it returned 0, we
976 			 * continue with normal processing.  If we have a
977 			 * pre-handler and it returned non-zero, that means
978 			 * user handler setup registers to exit to another
979 			 * instruction, we must skip the single stepping.
980 			 */
981 			if (!p->pre_handler || !p->pre_handler(p, regs))
982 				setup_singlestep(p, regs, kcb, 0);
983 			else
984 				reset_current_kprobe();
985 			return 1;
986 		}
987 	} else if (kprobe_is_ss(kcb)) {
988 		p = kprobe_running();
989 		if ((unsigned long)p->ainsn.insn < regs->ip &&
990 		    (unsigned long)p->ainsn.insn + MAX_INSN_SIZE > regs->ip) {
991 			/* Most provably this is the second int3 for singlestep */
992 			resume_singlestep(p, regs, kcb);
993 			kprobe_post_process(p, regs, kcb);
994 			return 1;
995 		}
996 	}
997 
998 	if (*addr != INT3_INSN_OPCODE) {
999 		/*
1000 		 * The breakpoint instruction was removed right
1001 		 * after we hit it.  Another cpu has removed
1002 		 * either a probepoint or a debugger breakpoint
1003 		 * at this address.  In either case, no further
1004 		 * handling of this interrupt is appropriate.
1005 		 * Back up over the (now missing) int3 and run
1006 		 * the original instruction.
1007 		 */
1008 		regs->ip = (unsigned long)addr;
1009 		return 1;
1010 	} /* else: not a kprobe fault; let the kernel handle it */
1011 
1012 	return 0;
1013 }
1014 NOKPROBE_SYMBOL(kprobe_int3_handler);
1015 
1016 /*
1017  * When a retprobed function returns, this code saves registers and
1018  * calls trampoline_handler() runs, which calls the kretprobe's handler.
1019  */
1020 asm(
1021 	".text\n"
1022 	".global kretprobe_trampoline\n"
1023 	".type kretprobe_trampoline, @function\n"
1024 	"kretprobe_trampoline:\n"
1025 	/* We don't bother saving the ss register */
1026 #ifdef CONFIG_X86_64
1027 	"	pushq %rsp\n"
1028 	"	pushfq\n"
1029 	SAVE_REGS_STRING
1030 	"	movq %rsp, %rdi\n"
1031 	"	call trampoline_handler\n"
1032 	/* Replace saved sp with true return address. */
1033 	"	movq %rax, 19*8(%rsp)\n"
1034 	RESTORE_REGS_STRING
1035 	"	popfq\n"
1036 #else
1037 	"	pushl %esp\n"
1038 	"	pushfl\n"
1039 	SAVE_REGS_STRING
1040 	"	movl %esp, %eax\n"
1041 	"	call trampoline_handler\n"
1042 	/* Replace saved sp with true return address. */
1043 	"	movl %eax, 15*4(%esp)\n"
1044 	RESTORE_REGS_STRING
1045 	"	popfl\n"
1046 #endif
1047 	"	ret\n"
1048 	".size kretprobe_trampoline, .-kretprobe_trampoline\n"
1049 );
1050 NOKPROBE_SYMBOL(kretprobe_trampoline);
1051 STACK_FRAME_NON_STANDARD(kretprobe_trampoline);
1052 
1053 
1054 /*
1055  * Called from kretprobe_trampoline
1056  */
1057 __used __visible void *trampoline_handler(struct pt_regs *regs)
1058 {
1059 	/* fixup registers */
1060 	regs->cs = __KERNEL_CS;
1061 #ifdef CONFIG_X86_32
1062 	regs->gs = 0;
1063 #endif
1064 	regs->ip = (unsigned long)&kretprobe_trampoline;
1065 	regs->orig_ax = ~0UL;
1066 
1067 	return (void *)kretprobe_trampoline_handler(regs, &kretprobe_trampoline, &regs->sp);
1068 }
1069 NOKPROBE_SYMBOL(trampoline_handler);
1070 
1071 int kprobe_fault_handler(struct pt_regs *regs, int trapnr)
1072 {
1073 	struct kprobe *cur = kprobe_running();
1074 	struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
1075 
1076 	if (unlikely(regs->ip == (unsigned long)cur->ainsn.insn)) {
1077 		/* This must happen on single-stepping */
1078 		WARN_ON(kcb->kprobe_status != KPROBE_HIT_SS &&
1079 			kcb->kprobe_status != KPROBE_REENTER);
1080 		/*
1081 		 * We are here because the instruction being single
1082 		 * stepped caused a page fault. We reset the current
1083 		 * kprobe and the ip points back to the probe address
1084 		 * and allow the page fault handler to continue as a
1085 		 * normal page fault.
1086 		 */
1087 		regs->ip = (unsigned long)cur->addr;
1088 
1089 		/*
1090 		 * If the IF flag was set before the kprobe hit,
1091 		 * don't touch it:
1092 		 */
1093 		regs->flags |= kcb->kprobe_old_flags;
1094 
1095 		if (kcb->kprobe_status == KPROBE_REENTER)
1096 			restore_previous_kprobe(kcb);
1097 		else
1098 			reset_current_kprobe();
1099 	}
1100 
1101 	return 0;
1102 }
1103 NOKPROBE_SYMBOL(kprobe_fault_handler);
1104 
1105 int __init arch_populate_kprobe_blacklist(void)
1106 {
1107 	return kprobe_add_area_blacklist((unsigned long)__entry_text_start,
1108 					 (unsigned long)__entry_text_end);
1109 }
1110 
1111 int __init arch_init_kprobes(void)
1112 {
1113 	return 0;
1114 }
1115 
1116 int arch_trampoline_kprobe(struct kprobe *p)
1117 {
1118 	return 0;
1119 }
1120