kprobes.c 27.3 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
/*
 *  Kernel Probes (KProbes)
 *  arch/ia64/kernel/kprobes.c
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
 *
 * Copyright (C) IBM Corporation, 2002, 2004
 * Copyright (C) Intel Corporation, 2005
 *
 * 2005-Apr     Rusty Lynch <rusty.lynch@intel.com> and Anil S Keshavamurthy
 *              <anil.s.keshavamurthy@intel.com> adapted from i386
 */

#include <linux/kprobes.h>
#include <linux/ptrace.h>
#include <linux/string.h>
#include <linux/slab.h>
#include <linux/preempt.h>
#include <linux/moduleloader.h>
32
#include <linux/kdebug.h>
33 34

#include <asm/pgtable.h>
35
#include <asm/sections.h>
36
#include <asm/uaccess.h>
37

38 39
extern void jprobe_inst_return(void);

40 41
DEFINE_PER_CPU(struct kprobe *, current_kprobe) = NULL;
DEFINE_PER_CPU(struct kprobe_ctlblk, kprobe_ctlblk);
42

43 44
struct kretprobe_blackpoint kretprobe_blacklist[] = {{NULL, NULL}};

45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80
enum instruction_type {A, I, M, F, B, L, X, u};
static enum instruction_type bundle_encoding[32][3] = {
  { M, I, I },				/* 00 */
  { M, I, I },				/* 01 */
  { M, I, I },				/* 02 */
  { M, I, I },				/* 03 */
  { M, L, X },				/* 04 */
  { M, L, X },				/* 05 */
  { u, u, u },  			/* 06 */
  { u, u, u },  			/* 07 */
  { M, M, I },				/* 08 */
  { M, M, I },				/* 09 */
  { M, M, I },				/* 0A */
  { M, M, I },				/* 0B */
  { M, F, I },				/* 0C */
  { M, F, I },				/* 0D */
  { M, M, F },				/* 0E */
  { M, M, F },				/* 0F */
  { M, I, B },				/* 10 */
  { M, I, B },				/* 11 */
  { M, B, B },				/* 12 */
  { M, B, B },				/* 13 */
  { u, u, u },  			/* 14 */
  { u, u, u },  			/* 15 */
  { B, B, B },				/* 16 */
  { B, B, B },				/* 17 */
  { M, M, B },				/* 18 */
  { M, M, B },				/* 19 */
  { u, u, u },  			/* 1A */
  { u, u, u },  			/* 1B */
  { M, F, B },				/* 1C */
  { M, F, B },				/* 1D */
  { u, u, u },  			/* 1E */
  { u, u, u },  			/* 1F */
};

81 82 83 84 85
/*
 * In this function we check to see if the instruction
 * is IP relative instruction and update the kprobe
 * inst flag accordingly
 */
86 87 88 89
static void __kprobes update_kprobe_inst_flag(uint template, uint  slot,
					      uint major_opcode,
					      unsigned long kprobe_inst,
					      struct kprobe *p)
90
{
R
Rusty Lynch 已提交
91 92
	p->ainsn.inst_flag = 0;
	p->ainsn.target_br_reg = 0;
T
Tony Luck 已提交
93
	p->ainsn.slot = slot;
94

95
	/* Check for Break instruction
B
bibo,mao 已提交
96
	 * Bits 37:40 Major opcode to be zero
97 98 99 100 101 102 103 104 105
	 * Bits 27:32 X6 to be zero
	 * Bits 32:35 X3 to be zero
	 */
	if ((!major_opcode) && (!((kprobe_inst >> 27) & 0x1FF)) ) {
		/* is a break instruction */
	 	p->ainsn.inst_flag |= INST_FLAG_BREAK_INST;
		return;
	}

106 107 108 109
	if (bundle_encoding[template][slot] == B) {
		switch (major_opcode) {
		  case INDIRECT_CALL_OPCODE:
	 		p->ainsn.inst_flag |= INST_FLAG_FIX_BRANCH_REG;
B
bibo,mao 已提交
110 111
			p->ainsn.target_br_reg = ((kprobe_inst >> 6) & 0x7);
			break;
112 113 114
		  case IP_RELATIVE_PREDICT_OPCODE:
		  case IP_RELATIVE_BRANCH_OPCODE:
			p->ainsn.inst_flag |= INST_FLAG_FIX_RELATIVE_IP_ADDR;
B
bibo,mao 已提交
115
			break;
116
		  case IP_RELATIVE_CALL_OPCODE:
B
bibo,mao 已提交
117 118 119 120
			p->ainsn.inst_flag |= INST_FLAG_FIX_RELATIVE_IP_ADDR;
			p->ainsn.inst_flag |= INST_FLAG_FIX_BRANCH_REG;
			p->ainsn.target_br_reg = ((kprobe_inst >> 6) & 0x7);
			break;
121
		}
B
bibo,mao 已提交
122
	} else if (bundle_encoding[template][slot] == X) {
123 124 125 126 127 128 129 130 131
		switch (major_opcode) {
		  case LONG_CALL_OPCODE:
			p->ainsn.inst_flag |= INST_FLAG_FIX_BRANCH_REG;
			p->ainsn.target_br_reg = ((kprobe_inst >> 6) & 0x7);
		  break;
		}
	}
	return;
}
132

133 134 135 136 137 138
/*
 * In this function we check to see if the instruction
 * (qp) cmpx.crel.ctype p1,p2=r2,r3
 * on which we are inserting kprobe is cmp instruction
 * with ctype as unc.
 */
139 140 141
static uint __kprobes is_cmp_ctype_unc_inst(uint template, uint slot,
					    uint major_opcode,
					    unsigned long kprobe_inst)
142 143 144 145 146 147 148 149 150 151 152 153 154 155
{
	cmp_inst_t cmp_inst;
	uint ctype_unc = 0;

	if (!((bundle_encoding[template][slot] == I) ||
		(bundle_encoding[template][slot] == M)))
		goto out;

	if (!((major_opcode == 0xC) || (major_opcode == 0xD) ||
		(major_opcode == 0xE)))
		goto out;

	cmp_inst.l = kprobe_inst;
	if ((cmp_inst.f.x2 == 0) || (cmp_inst.f.x2 == 1)) {
S
Simon Arlott 已提交
156
		/* Integer compare - Register Register (A6 type)*/
157 158 159 160
		if ((cmp_inst.f.tb == 0) && (cmp_inst.f.ta == 0)
				&&(cmp_inst.f.c == 1))
			ctype_unc = 1;
	} else if ((cmp_inst.f.x2 == 2)||(cmp_inst.f.x2 == 3)) {
S
Simon Arlott 已提交
161
		/* Integer compare - Immediate Register (A8 type)*/
162 163 164 165 166 167 168
		if ((cmp_inst.f.ta == 0) &&(cmp_inst.f.c == 1))
			ctype_unc = 1;
	}
out:
	return ctype_unc;
}

169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184
/*
 * In this function we check to see if the instruction
 * on which we are inserting kprobe is supported.
 * Returns qp value if supported
 * Returns -EINVAL if unsupported
 */
static int __kprobes unsupported_inst(uint template, uint  slot,
				      uint major_opcode,
				      unsigned long kprobe_inst,
				      unsigned long addr)
{
	int qp;

	qp = kprobe_inst & 0x3f;
	if (is_cmp_ctype_unc_inst(template, slot, major_opcode, kprobe_inst)) {
		if (slot == 1 && qp)  {
185 186
			printk(KERN_WARNING "Kprobes on cmp unc "
					"instruction on slot 1 at <0x%lx> "
187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
					"is not supported\n", addr);
			return -EINVAL;

		}
		qp = 0;
	}
	else if (bundle_encoding[template][slot] == I) {
		if (major_opcode == 0) {
			/*
			 * Check for Integer speculation instruction
			 * - Bit 33-35 to be equal to 0x1
			 */
			if (((kprobe_inst >> 33) & 0x7) == 1) {
				printk(KERN_WARNING
					"Kprobes on speculation inst at <0x%lx> not supported\n",
						addr);
				return -EINVAL;
			}
			/*
			 * IP relative mov instruction
			 *  - Bit 27-35 to be equal to 0x30
			 */
			if (((kprobe_inst >> 27) & 0x1FF) == 0x30) {
				printk(KERN_WARNING
					"Kprobes on \"mov r1=ip\" at <0x%lx> not supported\n",
						addr);
				return -EINVAL;

			}
		}
		else if ((major_opcode == 5) &&	!(kprobe_inst & (0xFUl << 33)) &&
				(kprobe_inst & (0x1UL << 12))) {
			/* test bit instructions, tbit,tnat,tf
			 * bit 33-36 to be equal to 0
			 * bit 12 to be equal to 1
			 */
			if (slot == 1 && qp) {
224 225
				printk(KERN_WARNING "Kprobes on test bit "
						"instruction on slot at <0x%lx> "
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244
						"is not supported\n", addr);
				return -EINVAL;
			}
			qp = 0;
		}
	}
	else if (bundle_encoding[template][slot] == B) {
		if (major_opcode == 7) {
			/* IP-Relative Predict major code is 7 */
			printk(KERN_WARNING "Kprobes on IP-Relative"
					"Predict is not supported\n");
			return -EINVAL;
		}
		else if (major_opcode == 2) {
			/* Indirect Predict, major code is 2
			 * bit 27-32 to be equal to 10 or 11
			 */
			int x6=(kprobe_inst >> 27) & 0x3F;
			if ((x6 == 0x10) || (x6 == 0x11)) {
245
				printk(KERN_WARNING "Kprobes on "
246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281
					"Indirect Predict is not supported\n");
				return -EINVAL;
			}
		}
	}
	/* kernel does not use float instruction, here for safety kprobe
	 * will judge whether it is fcmp/flass/float approximation instruction
	 */
	else if (unlikely(bundle_encoding[template][slot] == F)) {
		if ((major_opcode == 4 || major_opcode == 5) &&
				(kprobe_inst  & (0x1 << 12))) {
			/* fcmp/fclass unc instruction */
			if (slot == 1 && qp) {
				printk(KERN_WARNING "Kprobes on fcmp/fclass "
					"instruction on slot at <0x%lx> "
					"is not supported\n", addr);
				return -EINVAL;

			}
			qp = 0;
		}
		if ((major_opcode == 0 || major_opcode == 1) &&
			(kprobe_inst & (0x1UL << 33))) {
			/* float Approximation instruction */
			if (slot == 1 && qp) {
				printk(KERN_WARNING "Kprobes on float Approx "
					"instr at <0x%lx> is not supported\n",
						addr);
				return -EINVAL;
			}
			qp = 0;
		}
	}
	return qp;
}

282 283 284 285
/*
 * In this function we override the bundle with
 * the break instruction at the given slot.
 */
286 287 288
static void __kprobes prepare_break_inst(uint template, uint  slot,
					 uint major_opcode,
					 unsigned long kprobe_inst,
289 290
					 struct kprobe *p,
					 int qp)
291 292
{
	unsigned long break_inst = BREAK_INST;
293
	bundle_t *bundle = &p->opcode.bundle;
294 295 296

	/*
	 * Copy the original kprobe_inst qualifying predicate(qp)
297
	 * to the break instruction
298
	 */
299
	break_inst |= qp;
300 301 302 303 304 305 306 307 308 309 310 311

	switch (slot) {
	  case 0:
		bundle->quad0.slot0 = break_inst;
		break;
	  case 1:
		bundle->quad0.slot1_p0 = break_inst;
		bundle->quad1.slot1_p1 = break_inst >> (64-46);
		break;
	  case 2:
		bundle->quad1.slot2 = break_inst;
		break;
R
Rusty Lynch 已提交
312
	}
313

314 315 316 317 318 319 320 321
	/*
	 * Update the instruction flag, so that we can
	 * emulate the instruction properly after we
	 * single step on original instruction
	 */
	update_kprobe_inst_flag(template, slot, major_opcode, kprobe_inst, p);
}

322
static void __kprobes get_kprobe_inst(bundle_t *bundle, uint slot,
323 324 325 326 327 328
	       	unsigned long *kprobe_inst, uint *major_opcode)
{
	unsigned long kprobe_inst_p0, kprobe_inst_p1;
	unsigned int template;

	template = bundle->quad0.template;
329 330

	switch (slot) {
331
	  case 0:
B
bibo,mao 已提交
332 333 334
		*major_opcode = (bundle->quad0.slot0 >> SLOT0_OPCODE_SHIFT);
		*kprobe_inst = bundle->quad0.slot0;
		  break;
335
	  case 1:
B
bibo,mao 已提交
336 337 338 339
		*major_opcode = (bundle->quad1.slot1_p1 >> SLOT1_p1_OPCODE_SHIFT);
		kprobe_inst_p0 = bundle->quad0.slot1_p0;
		kprobe_inst_p1 = bundle->quad1.slot1_p1;
		*kprobe_inst = kprobe_inst_p0 | (kprobe_inst_p1 << (64-46));
340
		break;
341
	  case 2:
B
bibo,mao 已提交
342 343
		*major_opcode = (bundle->quad1.slot2 >> SLOT2_OPCODE_SHIFT);
		*kprobe_inst = bundle->quad1.slot2;
344 345
		break;
	}
346
}
347

348
/* Returns non-zero if the addr is in the Interrupt Vector Table */
349
static int __kprobes in_ivt_functions(unsigned long addr)
350 351 352 353 354
{
	return (addr >= (unsigned long)__start_ivt_text
		&& addr < (unsigned long)__end_ivt_text);
}

355 356
static int __kprobes valid_kprobe_addr(int template, int slot,
				       unsigned long addr)
357 358
{
	if ((slot > 2) || ((bundle_encoding[template][1] == L) && slot > 1)) {
359 360
		printk(KERN_WARNING "Attempting to insert unaligned kprobe "
				"at 0x%lx\n", addr);
361
		return -EINVAL;
R
Rusty Lynch 已提交
362
	}
363

B
bibo,mao 已提交
364 365
	if (in_ivt_functions(addr)) {
		printk(KERN_WARNING "Kprobes can't be inserted inside "
366
				"IVT functions at 0x%lx\n", addr);
B
bibo,mao 已提交
367 368
		return -EINVAL;
	}
369

370 371 372
	return 0;
}

373
static void __kprobes save_previous_kprobe(struct kprobe_ctlblk *kcb)
374
{
375 376 377 378
	unsigned int i;
	i = atomic_add_return(1, &kcb->prev_kprobe_index);
	kcb->prev_kprobe[i-1].kp = kprobe_running();
	kcb->prev_kprobe[i-1].status = kcb->kprobe_status;
379 380
}

381
static void __kprobes restore_previous_kprobe(struct kprobe_ctlblk *kcb)
382
{
383
	unsigned int i;
384 385 386 387
	i = atomic_read(&kcb->prev_kprobe_index);
	__get_cpu_var(current_kprobe) = kcb->prev_kprobe[i-1].kp;
	kcb->kprobe_status = kcb->prev_kprobe[i-1].status;
	atomic_sub(1, &kcb->prev_kprobe_index);
388 389
}

390
static void __kprobes set_current_kprobe(struct kprobe *p,
391
			struct kprobe_ctlblk *kcb)
392
{
393
	__get_cpu_var(current_kprobe) = p;
394 395
}

396 397 398 399 400 401 402 403 404 405 406 407
static void kretprobe_trampoline(void)
{
}

/*
 * At this point the target function has been tricked into
 * returning into our trampoline.  Lookup the associated instance
 * and then:
 *    - call the handler function
 *    - cleanup by marking the instance as unused
 *    - long jump back to the original return address
 */
408
int __kprobes trampoline_probe_handler(struct kprobe *p, struct pt_regs *regs)
409 410
{
	struct kretprobe_instance *ri = NULL;
411
	struct hlist_head *head, empty_rp;
412
	struct hlist_node *node, *tmp;
413
	unsigned long flags, orig_ret_address = 0;
414 415 416
	unsigned long trampoline_address =
		((struct fnptr *)kretprobe_trampoline)->ip;

417
	INIT_HLIST_HEAD(&empty_rp);
418
	spin_lock_irqsave(&kretprobe_lock, flags);
419
	head = kretprobe_inst_table_head(current);
420 421 422 423 424 425 426 427 428 429 430 431 432 433

	/*
	 * It is possible to have multiple instances associated with a given
	 * task either because an multiple functions in the call path
	 * have a return probe installed on them, and/or more then one return
	 * return probe was registered for a target function.
	 *
	 * We can handle this because:
	 *     - instances are always inserted at the head of the list
	 *     - when multiple return probes are registered for the same
	 *       function, the first instance's ret_addr will point to the
	 *       real return address, and all the rest will point to
	 *       kretprobe_trampoline
	 */
434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450
	hlist_for_each_entry_safe(ri, node, tmp, head, hlist) {
		if (ri->task != current)
			/* another task is sharing our hash bucket */
			continue;

		orig_ret_address = (unsigned long)ri->ret_addr;
		if (orig_ret_address != trampoline_address)
			/*
			 * This is the real return address. Any other
			 * instances associated with this task are for
			 * other calls deeper on the call stack
			 */
			break;
	}

	regs->cr_iip = orig_ret_address;

451
	hlist_for_each_entry_safe(ri, node, tmp, head, hlist) {
452
		if (ri->task != current)
453
			/* another task is sharing our hash bucket */
454
			continue;
455 456 457 458 459

		if (ri->rp && ri->rp->handler)
			ri->rp->handler(ri, regs);

		orig_ret_address = (unsigned long)ri->ret_addr;
460
		recycle_rp_inst(ri, &empty_rp);
461 462 463 464 465 466 467 468 469 470

		if (orig_ret_address != trampoline_address)
			/*
			 * This is the real return address. Any other
			 * instances associated with this task are for
			 * other calls deeper on the call stack
			 */
			break;
	}

471 472
	kretprobe_assert(ri, orig_ret_address, trampoline_address);

473
	reset_current_kprobe();
474
	spin_unlock_irqrestore(&kretprobe_lock, flags);
475 476
	preempt_enable_no_resched();

477 478 479 480
	hlist_for_each_entry_safe(ri, node, tmp, &empty_rp, hlist) {
		hlist_del(&ri->hlist);
		kfree(ri);
	}
481 482 483 484 485
	/*
	 * By returning a non-zero value, we are telling
	 * kprobe_handler() that we don't want the post_handler
	 * to run (and have re-enabled preemption)
	 */
486
	return 1;
487 488
}

489
/* Called with kretprobe_lock held */
490
void __kprobes arch_prepare_kretprobe(struct kretprobe_instance *ri,
491
				      struct pt_regs *regs)
492
{
493
	ri->ret_addr = (kprobe_opcode_t *)regs->b0;
494

495 496
	/* Replace the return addr with trampoline addr */
	regs->b0 = ((struct fnptr *)kretprobe_trampoline)->ip;
497 498
}

499
int __kprobes arch_prepare_kprobe(struct kprobe *p)
500 501 502 503 504
{
	unsigned long addr = (unsigned long) p->addr;
	unsigned long *kprobe_addr = (unsigned long *)(addr & ~0xFULL);
	unsigned long kprobe_inst=0;
	unsigned int slot = addr & 0xf, template, major_opcode = 0;
505
	bundle_t *bundle;
506
	int qp;
507

508
	bundle = &((kprobe_opcode_t *)kprobe_addr)->bundle;
B
bibo,mao 已提交
509
	template = bundle->quad0.template;
510 511 512 513 514

	if(valid_kprobe_addr(template, slot, addr))
		return -EINVAL;

	/* Move to slot 2, if bundle is MLX type and kprobe slot is 1 */
B
bibo,mao 已提交
515 516
	if (slot == 1 && bundle_encoding[template][1] == L)
		slot++;
517 518 519 520

	/* Get kprobe_inst and major_opcode from the bundle */
	get_kprobe_inst(bundle, slot, &kprobe_inst, &major_opcode);

521 522 523
	qp = unsupported_inst(template, slot, major_opcode, kprobe_inst, addr);
	if (qp < 0)
		return -EINVAL;
R
Rusty Lynch 已提交
524

525 526 527 528 529
	p->ainsn.insn = get_insn_slot();
	if (!p->ainsn.insn)
		return -ENOMEM;
	memcpy(&p->opcode, kprobe_addr, sizeof(kprobe_opcode_t));
	memcpy(p->ainsn.insn, kprobe_addr, sizeof(kprobe_opcode_t));
R
Rusty Lynch 已提交
530

531
	prepare_break_inst(template, slot, major_opcode, kprobe_inst, p, qp);
532

533
	return 0;
534 535
}

536
void __kprobes arch_arm_kprobe(struct kprobe *p)
R
Rusty Lynch 已提交
537
{
T
Tony Luck 已提交
538 539 540 541 542 543
	unsigned long arm_addr;
	bundle_t *src, *dest;

	arm_addr = ((unsigned long)p->addr) & ~0xFUL;
	dest = &((kprobe_opcode_t *)arm_addr)->bundle;
	src = &p->opcode.bundle;
R
Rusty Lynch 已提交
544

545 546
	flush_icache_range((unsigned long)p->ainsn.insn,
			(unsigned long)p->ainsn.insn + sizeof(kprobe_opcode_t));
T
Tony Luck 已提交
547 548 549 550 551 552 553 554 555 556 557
	switch (p->ainsn.slot) {
		case 0:
			dest->quad0.slot0 = src->quad0.slot0;
			break;
		case 1:
			dest->quad1.slot1_p1 = src->quad1.slot1_p1;
			break;
		case 2:
			dest->quad1.slot2 = src->quad1.slot2;
			break;
	}
558
	flush_icache_range(arm_addr, arm_addr + sizeof(kprobe_opcode_t));
559 560
}

561
void __kprobes arch_disarm_kprobe(struct kprobe *p)
562
{
T
Tony Luck 已提交
563 564
	unsigned long arm_addr;
	bundle_t *src, *dest;
565

T
Tony Luck 已提交
566 567
	arm_addr = ((unsigned long)p->addr) & ~0xFUL;
	dest = &((kprobe_opcode_t *)arm_addr)->bundle;
568
	/* p->ainsn.insn contains the original unaltered kprobe_opcode_t */
T
Tony Luck 已提交
569 570 571 572 573 574 575 576 577 578 579 580
	src = &p->ainsn.insn->bundle;
	switch (p->ainsn.slot) {
		case 0:
			dest->quad0.slot0 = src->quad0.slot0;
			break;
		case 1:
			dest->quad1.slot1_p1 = src->quad1.slot1_p1;
			break;
		case 2:
			dest->quad1.slot2 = src->quad1.slot2;
			break;
	}
581
	flush_icache_range(arm_addr, arm_addr + sizeof(kprobe_opcode_t));
582 583
}

584 585 586
void __kprobes arch_remove_kprobe(struct kprobe *p)
{
	mutex_lock(&kprobe_mutex);
587
	free_insn_slot(p->ainsn.insn, 0);
588 589
	mutex_unlock(&kprobe_mutex);
}
590 591 592 593
/*
 * We are resuming execution after a single step fault, so the pt_regs
 * structure reflects the register state after we executed the instruction
 * located in the kprobe (p->ainsn.insn.bundle).  We still need to adjust
594 595 596
 * the ip to point back to the original stack address. To set the IP address
 * to original stack address, handle the case where we need to fixup the
 * relative IP address and/or fixup branch register.
597
 */
598
static void __kprobes resume_execution(struct kprobe *p, struct pt_regs *regs)
599
{
B
bibo,mao 已提交
600 601 602 603
	unsigned long bundle_addr = (unsigned long) (&p->ainsn.insn->bundle);
	unsigned long resume_addr = (unsigned long)p->addr & ~0xFULL;
	unsigned long template;
	int slot = ((unsigned long)p->addr & 0xf);
604

605
	template = p->ainsn.insn->bundle.quad0.template;
606

B
bibo,mao 已提交
607 608
	if (slot == 1 && bundle_encoding[template][1] == L)
		slot = 2;
609 610 611 612 613

	if (p->ainsn.inst_flag) {

		if (p->ainsn.inst_flag & INST_FLAG_FIX_RELATIVE_IP_ADDR) {
			/* Fix relative IP address */
B
bibo,mao 已提交
614 615
			regs->cr_iip = (regs->cr_iip - bundle_addr) +
					resume_addr;
616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649
		}

		if (p->ainsn.inst_flag & INST_FLAG_FIX_BRANCH_REG) {
		/*
		 * Fix target branch register, software convention is
		 * to use either b0 or b6 or b7, so just checking
		 * only those registers
		 */
			switch (p->ainsn.target_br_reg) {
			case 0:
				if ((regs->b0 == bundle_addr) ||
					(regs->b0 == bundle_addr + 0x10)) {
					regs->b0 = (regs->b0 - bundle_addr) +
						resume_addr;
				}
				break;
			case 6:
				if ((regs->b6 == bundle_addr) ||
					(regs->b6 == bundle_addr + 0x10)) {
					regs->b6 = (regs->b6 - bundle_addr) +
						resume_addr;
				}
				break;
			case 7:
				if ((regs->b7 == bundle_addr) ||
					(regs->b7 == bundle_addr + 0x10)) {
					regs->b7 = (regs->b7 - bundle_addr) +
						resume_addr;
				}
				break;
			} /* end switch */
		}
		goto turn_ss_off;
	}
650

651
	if (slot == 2) {
B
bibo,mao 已提交
652 653 654 655 656 657 658
		if (regs->cr_iip == bundle_addr + 0x10) {
			regs->cr_iip = resume_addr + 0x10;
		}
	} else {
		if (regs->cr_iip == bundle_addr) {
			regs->cr_iip = resume_addr;
		}
659
	}
660

661
turn_ss_off:
B
bibo,mao 已提交
662 663
	/* Turn off Single Step bit */
	ia64_psr(regs)->ss = 0;
664 665
}

666
static void __kprobes prepare_ss(struct kprobe *p, struct pt_regs *regs)
667
{
668
	unsigned long bundle_addr = (unsigned long) &p->ainsn.insn->bundle;
669 670
	unsigned long slot = (unsigned long)p->addr & 0xf;

671 672 673 674 675
	/* single step inline if break instruction */
	if (p->ainsn.inst_flag == INST_FLAG_BREAK_INST)
		regs->cr_iip = (unsigned long)p->addr & ~0xFULL;
	else
		regs->cr_iip = bundle_addr & ~0xFULL;
676 677 678 679 680 681 682 683 684 685

	if (slot > 2)
		slot = 0;

	ia64_psr(regs)->ri = slot;

	/* turn on single stepping */
	ia64_psr(regs)->ss = 1;
}

686 687 688 689 690 691 692 693 694 695 696 697 698
static int __kprobes is_ia64_break_inst(struct pt_regs *regs)
{
	unsigned int slot = ia64_psr(regs)->ri;
	unsigned int template, major_opcode;
	unsigned long kprobe_inst;
	unsigned long *kprobe_addr = (unsigned long *)regs->cr_iip;
	bundle_t bundle;

	memcpy(&bundle, kprobe_addr, sizeof(bundle_t));
	template = bundle.quad0.template;

	/* Move to slot 2, if bundle is MLX type and kprobe slot is 1 */
	if (slot == 1 && bundle_encoding[template][1] == L)
B
bibo,mao 已提交
699
		slot++;
700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717

	/* Get Kprobe probe instruction at given slot*/
	get_kprobe_inst(&bundle, slot, &kprobe_inst, &major_opcode);

	/* For break instruction,
	 * Bits 37:40 Major opcode to be zero
	 * Bits 27:32 X6 to be zero
	 * Bits 32:35 X3 to be zero
	 */
	if (major_opcode || ((kprobe_inst >> 27) & 0x1FF) ) {
		/* Not a break instruction */
		return 0;
	}

	/* Is a break instruction */
	return 1;
}

718
static int __kprobes pre_kprobes_handler(struct die_args *args)
719 720 721
{
	struct kprobe *p;
	int ret = 0;
722
	struct pt_regs *regs = args->regs;
723
	kprobe_opcode_t *addr = (kprobe_opcode_t *)instruction_pointer(regs);
724 725 726 727 728 729 730 731
	struct kprobe_ctlblk *kcb;

	/*
	 * We don't want to be preempted for the entire
	 * duration of kprobe processing
	 */
	preempt_disable();
	kcb = get_kprobe_ctlblk();
732 733 734 735 736

	/* Handle recursion cases */
	if (kprobe_running()) {
		p = get_kprobe(addr);
		if (p) {
737
			if ((kcb->kprobe_status == KPROBE_HIT_SS) &&
738
	 		     (p->ainsn.inst_flag == INST_FLAG_BREAK_INST)) {
B
bibo,mao 已提交
739
				ia64_psr(regs)->ss = 0;
740 741
				goto no_kprobe;
			}
742 743 744 745 746 747
			/* We have reentered the pre_kprobe_handler(), since
			 * another probe was hit while within the handler.
			 * We here save the original kprobes variables and
			 * just single step on the instruction of the new probe
			 * without calling any user handlers.
			 */
748 749
			save_previous_kprobe(kcb);
			set_current_kprobe(p, kcb);
750
			kprobes_inc_nmissed_count(p);
751
			prepare_ss(p, regs);
752
			kcb->kprobe_status = KPROBE_REENTER;
753
			return 1;
754
		} else if (args->err == __IA64_BREAK_JPROBE) {
755 756 757
			/*
			 * jprobe instrumented function just completed
			 */
758
			p = __get_cpu_var(current_kprobe);
759 760 761
			if (p->break_handler && p->break_handler(p, regs)) {
				goto ss_probe;
			}
762 763 764 765 766 767 768
		} else if (!is_ia64_break_inst(regs)) {
			/* The breakpoint instruction was removed by
			 * another cpu right after we hit, no further
			 * handling of this interrupt is appropriate
			 */
			ret = 1;
			goto no_kprobe;
769 770 771
		} else {
			/* Not our break */
			goto no_kprobe;
772 773 774 775 776
		}
	}

	p = get_kprobe(addr);
	if (!p) {
777 778 779 780 781 782 783 784 785 786 787 788 789
		if (!is_ia64_break_inst(regs)) {
			/*
			 * The breakpoint instruction was removed right
			 * after we hit it.  Another cpu has removed
			 * either a probepoint or a debugger breakpoint
			 * at this address.  In either case, no further
			 * handling of this interrupt is appropriate.
			 */
			ret = 1;

		}

		/* Not one of our break, let kernel handle it */
790 791 792
		goto no_kprobe;
	}

793 794
	set_current_kprobe(p, kcb);
	kcb->kprobe_status = KPROBE_HIT_ACTIVE;
795 796 797 798

	if (p->pre_handler && p->pre_handler(p, regs))
		/*
		 * Our pre-handler is specifically requesting that we just
799 800
		 * do a return.  This is used for both the jprobe pre-handler
		 * and the kretprobe trampoline
801 802 803 804 805
		 */
		return 1;

ss_probe:
	prepare_ss(p, regs);
806
	kcb->kprobe_status = KPROBE_HIT_SS;
807 808 809
	return 1;

no_kprobe:
810
	preempt_enable_no_resched();
811 812 813
	return ret;
}

814
static int __kprobes post_kprobes_handler(struct pt_regs *regs)
815
{
816 817 818 819
	struct kprobe *cur = kprobe_running();
	struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();

	if (!cur)
820 821
		return 0;

822 823 824
	if ((kcb->kprobe_status != KPROBE_REENTER) && cur->post_handler) {
		kcb->kprobe_status = KPROBE_HIT_SSDONE;
		cur->post_handler(cur, regs, 0);
825
	}
826

827
	resume_execution(cur, regs);
828

829
	/*Restore back the original saved kprobes variables and continue. */
830 831
	if (kcb->kprobe_status == KPROBE_REENTER) {
		restore_previous_kprobe(kcb);
832 833
		goto out;
	}
834
	reset_current_kprobe();
835 836

out:
837 838 839 840
	preempt_enable_no_resched();
	return 1;
}

841
int __kprobes kprobes_fault_handler(struct pt_regs *regs, int trapnr)
842
{
843 844 845
	struct kprobe *cur = kprobe_running();
	struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();

846

847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862
	switch(kcb->kprobe_status) {
	case KPROBE_HIT_SS:
	case KPROBE_REENTER:
		/*
		 * We are here because the instruction being single
		 * stepped caused a page fault. We reset the current
		 * kprobe and the instruction pointer points back to
		 * the probe address and allow the page fault handler
		 * to continue as a normal page fault.
		 */
		regs->cr_iip = ((unsigned long)cur->addr) & ~0xFULL;
		ia64_psr(regs)->ri = ((unsigned long)cur->addr) & 0xf;
		if (kcb->kprobe_status == KPROBE_REENTER)
			restore_previous_kprobe(kcb);
		else
			reset_current_kprobe();
863
		preempt_enable_no_resched();
864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882
		break;
	case KPROBE_HIT_ACTIVE:
	case KPROBE_HIT_SSDONE:
		/*
		 * We increment the nmissed count for accounting,
		 * we can also use npre/npostfault count for accouting
		 * these specific fault cases.
		 */
		kprobes_inc_nmissed_count(cur);

		/*
		 * We come here because instructions in the pre/post
		 * handler caused the page_fault, this could happen
		 * if handler tries to access user space by
		 * copy_from_user(), get_user() etc. Let the
		 * user-specified handler try to fix it first.
		 */
		if (cur->fault_handler && cur->fault_handler(cur, regs, trapnr))
			return 1;
883 884 885 886 887 888
		/*
		 * In case the user-specified fault handler returned
		 * zero, try to fix up.
		 */
		if (ia64_done_with_exception(regs))
			return 1;
889 890 891 892 893 894 895

		/*
		 * Let ia64_do_page_fault() fix it.
		 */
		break;
	default:
		break;
896 897 898 899 900
	}

	return 0;
}

901 902
int __kprobes kprobe_exceptions_notify(struct notifier_block *self,
				       unsigned long val, void *data)
903 904
{
	struct die_args *args = (struct die_args *)data;
905 906
	int ret = NOTIFY_DONE;

907 908 909
	if (args->regs && user_mode(args->regs))
		return ret;

910 911
	switch(val) {
	case DIE_BREAK:
912
		/* err is break number from ia64_bad_break() */
T
Tony Luck 已提交
913 914 915
		if ((args->err >> 12) == (__IA64_BREAK_KPROBE >> 12)
			|| args->err == __IA64_BREAK_JPROBE
			|| args->err == 0)
916 917
			if (pre_kprobes_handler(args))
				ret = NOTIFY_STOP;
918
		break;
919 920 921 922 923
	case DIE_FAULT:
		/* err is vector number from ia64_fault() */
		if (args->err == 36)
			if (post_kprobes_handler(args->regs))
				ret = NOTIFY_STOP;
924 925 926 927
		break;
	default:
		break;
	}
928
	return ret;
929 930
}

931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951
struct param_bsp_cfm {
	unsigned long ip;
	unsigned long *bsp;
	unsigned long cfm;
};

static void ia64_get_bsp_cfm(struct unw_frame_info *info, void *arg)
{
	unsigned long ip;
	struct param_bsp_cfm *lp = arg;

	do {
		unw_get_ip(info, &ip);
		if (ip == 0)
			break;
		if (ip == lp->ip) {
			unw_get_bsp(info, (unsigned long*)&lp->bsp);
			unw_get_cfm(info, (unsigned long*)&lp->cfm);
			return;
		}
	} while (unw_unwind(info) >= 0);
M
Matthew Wilcox 已提交
952
	lp->bsp = NULL;
953 954 955 956
	lp->cfm = 0;
	return;
}

957 958 959 960 961
unsigned long arch_deref_entry_point(void *entry)
{
	return ((struct fnptr *)entry)->ip;
}

962
int __kprobes setjmp_pre_handler(struct kprobe *p, struct pt_regs *regs)
963
{
964
	struct jprobe *jp = container_of(p, struct jprobe, kp);
965
	unsigned long addr = arch_deref_entry_point(jp->entry);
966
	struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
967 968 969 970 971 972
	struct param_bsp_cfm pa;
	int bytes;

	/*
	 * Callee owns the argument space and could overwrite it, eg
	 * tail call optimization. So to be absolutely safe
S
Simon Arlott 已提交
973
	 * we save the argument space before transferring the control
974 975 976 977 978 979 980 981 982 983 984 985
	 * to instrumented jprobe function which runs in
	 * the process context
	 */
	pa.ip = regs->cr_iip;
	unw_init_running(ia64_get_bsp_cfm, &pa);
	bytes = (char *)ia64_rse_skip_regs(pa.bsp, pa.cfm & 0x3f)
				- (char *)pa.bsp;
	memcpy( kcb->jprobes_saved_stacked_regs,
		pa.bsp,
		bytes );
	kcb->bsp = pa.bsp;
	kcb->cfm = pa.cfm;
986

987
	/* save architectural state */
988
	kcb->jprobe_saved_regs = *regs;
989 990 991 992 993 994 995 996 997 998

	/* after rfi, execute the jprobe instrumented function */
	regs->cr_iip = addr & ~0xFULL;
	ia64_psr(regs)->ri = addr & 0xf;
	regs->r1 = ((struct fnptr *)(jp->entry))->gp;

	/*
	 * fix the return address to our jprobe_inst_return() function
	 * in the jprobes.S file
	 */
B
bibo,mao 已提交
999
	regs->b0 = ((struct fnptr *)(jprobe_inst_return))->ip;
1000 1001

	return 1;
1002 1003
}

1004
int __kprobes longjmp_break_handler(struct kprobe *p, struct pt_regs *regs)
1005
{
1006
	struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
1007
	int bytes;
1008

1009
	/* restoring architectural state */
1010
	*regs = kcb->jprobe_saved_regs;
1011 1012 1013 1014 1015 1016 1017 1018 1019 1020

	/* restoring the original argument space */
	flush_register_stack();
	bytes = (char *)ia64_rse_skip_regs(kcb->bsp, kcb->cfm & 0x3f)
				- (char *)kcb->bsp;
	memcpy( kcb->bsp,
		kcb->jprobes_saved_stacked_regs,
		bytes );
	invalidate_stacked_regs();

1021
	preempt_enable_no_resched();
1022
	return 1;
1023
}
1024 1025 1026 1027 1028

static struct kprobe trampoline_p = {
	.pre_handler = trampoline_probe_handler
};

1029
int __init arch_init_kprobes(void)
1030 1031 1032 1033 1034
{
	trampoline_p.addr =
		(kprobe_opcode_t *)((struct fnptr *)kretprobe_trampoline)->ip;
	return register_kprobe(&trampoline_p);
}
1035 1036 1037 1038 1039 1040 1041 1042 1043

int __kprobes arch_trampoline_kprobe(struct kprobe *p)
{
	if (p->addr ==
		(kprobe_opcode_t *)((struct fnptr *)kretprobe_trampoline)->ip)
		return 1;

	return 0;
}