arch/x86_64/kernel/kprobes.c

   1 /*
   2  *  Kernel Probes (KProbes)
   3  *  arch/x86_64/kernel/kprobes.c
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License as published by
   7  * the Free Software Foundation; either version 2 of the License, or
   8  * (at your option) any later version.
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License
  16  * along with this program; if not, write to the Free Software
  17  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
  18  *
  19  * Copyright (C) IBM Corporation, 2002, 2004
  20  *
  21  * 2002-Oct     Created by Vamsi Krishna S <vamsi_krishna@in.ibm.com> Kernel
  22  *              Probes initial implementation ( includes contributions from
  23  *              Rusty Russell).
  24  * 2004-July    Suparna Bhattacharya <suparna@in.ibm.com> added jumper probes
  25  *              interface to access function arguments.
  26  * 2004-Oct     Jim Keniston <kenistoj@us.ibm.com> and Prasanna S Panchamukhi
  27  *              <prasanna@in.ibm.com> adapted for x86_64
  28  * 2005-Mar     Roland McGrath <roland@redhat.com>
  29  *              Fixed to handle %rip-relative addressing mode correctly.
  30  * 2005-May     Rusty Lynch <rusty.lynch@intel.com>
  31  *              Added function return probes functionality
  32  */
  33
  34 #include <linux/config.h>
  35 #include <linux/kprobes.h>
  36 #include <linux/ptrace.h>
  37 #include <linux/string.h>
  38 #include <linux/slab.h>
  39 #include <linux/preempt.h>
  40
  41 #include <asm/cacheflush.h>
  42 #include <asm/pgtable.h>
  43 #include <asm/kdebug.h>
  44
  45 void jprobe_return_end(void);
  46 void __kprobes arch_copy_kprobe(struct kprobe *p);
  47
  48 DEFINE_PER_CPU(struct kprobe *, current_kprobe) = NULL;
  49 DEFINE_PER_CPU(struct kprobe_ctlblk, kprobe_ctlblk);
  50
  51 /*
  52  * returns non-zero if opcode modifies the interrupt flag.
  53  */
  54 static inline int is_IF_modifier(kprobe_opcode_t *insn)
  55 {
  56         switch (*insn) {
  57         case 0xfa:              /* cli */
  58         case 0xfb:              /* sti */
  59         case 0xcf:              /* iret/iretd */
  60         case 0x9d:              /* popf/popfd */
  61                 return 1;
  62         }
  63
  64         if (*insn  >= 0x40 && *insn <= 0x4f && *++insn == 0xcf)
  65                 return 1;
  66         return 0;
  67 }
  68
  69 int __kprobes arch_prepare_kprobe(struct kprobe *p)
  70 {
  71         /* insn: must be on special executable page on x86_64. */
  72         p->ainsn.insn = get_insn_slot();
  73         if (!p->ainsn.insn) {
  74                 return -ENOMEM;
  75         }
  76         arch_copy_kprobe(p);
  77         return 0;
  78 }
  79
  80 /*
  81  * Determine if the instruction uses the %rip-relative addressing mode.
  82  * If it does, return the address of the 32-bit displacement word.
  83  * If not, return null.
  84  */
  85 static inline s32 *is_riprel(u8 *insn)
  86 {
  87 #define W(row,b0,b1,b2,b3,b4,b5,b6,b7,b8,b9,ba,bb,bc,bd,be,bf)                \
  88         (((b0##UL << 0x0)|(b1##UL << 0x1)|(b2##UL << 0x2)|(b3##UL << 0x3) |   \
  89           (b4##UL << 0x4)|(b5##UL << 0x5)|(b6##UL << 0x6)|(b7##UL << 0x7) |   \
  90           (b8##UL << 0x8)|(b9##UL << 0x9)|(ba##UL << 0xa)|(bb##UL << 0xb) |   \
  91           (bc##UL << 0xc)|(bd##UL << 0xd)|(be##UL << 0xe)|(bf##UL << 0xf))    \
  92          << (row % 64))
  93         static const u64 onebyte_has_modrm[256 / 64] = {
  94                 /*      0 1 2 3 4 5 6 7 8 9 a b c d e f         */
  95                 /*      -------------------------------         */
  96                 W(0x00, 1,1,1,1,0,0,0,0,1,1,1,1,0,0,0,0)| /* 00 */
  97                 W(0x10, 1,1,1,1,0,0,0,0,1,1,1,1,0,0,0,0)| /* 10 */
  98                 W(0x20, 1,1,1,1,0,0,0,0,1,1,1,1,0,0,0,0)| /* 20 */
  99                 W(0x30, 1,1,1,1,0,0,0,0,1,1,1,1,0,0,0,0), /* 30 */
 100                 W(0x40, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0)| /* 40 */
 101                 W(0x50, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0)| /* 50 */
 102                 W(0x60, 0,0,1,1,0,0,0,0,0,1,0,1,0,0,0,0)| /* 60 */
 103                 W(0x70, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0), /* 70 */
 104                 W(0x80, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)| /* 80 */
 105                 W(0x90, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0)| /* 90 */
 106                 W(0xa0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0)| /* a0 */
 107                 W(0xb0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0), /* b0 */
 108                 W(0xc0, 1,1,0,0,1,1,1,1,0,0,0,0,0,0,0,0)| /* c0 */
 109                 W(0xd0, 1,1,1,1,0,0,0,0,1,1,1,1,1,1,1,1)| /* d0 */
 110                 W(0xe0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0)| /* e0 */
 111                 W(0xf0, 0,0,0,0,0,0,1,1,0,0,0,0,0,0,1,1)  /* f0 */
 112                 /*      -------------------------------         */
 113                 /*      0 1 2 3 4 5 6 7 8 9 a b c d e f         */
 114         };
 115         static const u64 twobyte_has_modrm[256 / 64] = {
 116                 /*      0 1 2 3 4 5 6 7 8 9 a b c d e f         */
 117                 /*      -------------------------------         */
 118                 W(0x00, 1,1,1,1,0,0,0,0,0,0,0,0,0,1,0,1)| /* 0f */
 119                 W(0x10, 1,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0)| /* 1f */
 120                 W(0x20, 1,1,1,1,1,0,1,0,1,1,1,1,1,1,1,1)| /* 2f */
 121                 W(0x30, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0), /* 3f */
 122                 W(0x40, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)| /* 4f */
 123                 W(0x50, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)| /* 5f */
 124                 W(0x60, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)| /* 6f */
 125                 W(0x70, 1,1,1,1,1,1,1,0,0,0,0,0,1,1,1,1), /* 7f */
 126                 W(0x80, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0)| /* 8f */
 127                 W(0x90, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)| /* 9f */
 128                 W(0xa0, 0,0,0,1,1,1,1,1,0,0,0,1,1,1,1,1)| /* af */
 129                 W(0xb0, 1,1,1,1,1,1,1,1,0,0,1,1,1,1,1,1), /* bf */
 130                 W(0xc0, 1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0)| /* cf */
 131                 W(0xd0, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)| /* df */
 132                 W(0xe0, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)| /* ef */
 133                 W(0xf0, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,0)  /* ff */
 134                 /*      -------------------------------         */
 135                 /*      0 1 2 3 4 5 6 7 8 9 a b c d e f         */
 136         };
 137 #undef  W
 138         int need_modrm;
 139
 140         /* Skip legacy instruction prefixes.  */
 141         while (1) {
 142                 switch (*insn) {
 143                 case 0x66:
 144                 case 0x67:
 145                 case 0x2e:
 146                 case 0x3e:
 147                 case 0x26:
 148                 case 0x64:
 149                 case 0x65:
 150                 case 0x36:
 151                 case 0xf0:
 152                 case 0xf3:
 153                 case 0xf2:
 154                         ++insn;
 155                         continue;
 156                 }
 157                 break;
 158         }
 159
 160         /* Skip REX instruction prefix.  */
 161         if ((*insn & 0xf0) == 0x40)
 162                 ++insn;
 163
 164         if (*insn == 0x0f) {    /* Two-byte opcode.  */
 165                 ++insn;
 166                 need_modrm = test_bit(*insn, twobyte_has_modrm);
 167         } else {                /* One-byte opcode.  */
 168                 need_modrm = test_bit(*insn, onebyte_has_modrm);
 169         }
 170
 171         if (need_modrm) {
 172                 u8 modrm = *++insn;
 173                 if ((modrm & 0xc7) == 0x05) { /* %rip+disp32 addressing mode */
 174                         /* Displacement follows ModRM byte.  */
 175                         return (s32 *) ++insn;
 176                 }
 177         }
 178
 179         /* No %rip-relative addressing mode here.  */
 180         return NULL;
 181 }
 182
 183 void __kprobes arch_copy_kprobe(struct kprobe *p)
 184 {
 185         s32 *ripdisp;
 186         memcpy(p->ainsn.insn, p->addr, MAX_INSN_SIZE);
 187         ripdisp = is_riprel(p->ainsn.insn);
 188         if (ripdisp) {
 189                 /*
 190                  * The copied instruction uses the %rip-relative
 191                  * addressing mode.  Adjust the displacement for the
 192                  * difference between the original location of this
 193                  * instruction and the location of the copy that will
 194                  * actually be run.  The tricky bit here is making sure
 195                  * that the sign extension happens correctly in this
 196                  * calculation, since we need a signed 32-bit result to
 197                  * be sign-extended to 64 bits when it's added to the
 198                  * %rip value and yield the same 64-bit result that the
 199                  * sign-extension of the original signed 32-bit
 200                  * displacement would have given.
 201                  */
 202                 s64 disp = (u8 *) p->addr + *ripdisp - (u8 *) p->ainsn.insn;
 203                 BUG_ON((s64) (s32) disp != disp); /* Sanity check.  */
 204                 *ripdisp = disp;
 205         }
 206         p->opcode = *p->addr;
 207 }
 208
 209 void __kprobes arch_arm_kprobe(struct kprobe *p)
 210 {
 211         *p->addr = BREAKPOINT_INSTRUCTION;
 212         flush_icache_range((unsigned long) p->addr,
 213                            (unsigned long) p->addr + sizeof(kprobe_opcode_t));
 214 }
 215
 216 void __kprobes arch_disarm_kprobe(struct kprobe *p)
 217 {
 218         *p->addr = p->opcode;
 219         flush_icache_range((unsigned long) p->addr,
 220                            (unsigned long) p->addr + sizeof(kprobe_opcode_t));
 221 }
 222
 223 void __kprobes arch_remove_kprobe(struct kprobe *p)
 224 {
 225         free_insn_slot(p->ainsn.insn);
 226 }
 227
 228 static inline void save_previous_kprobe(struct kprobe_ctlblk *kcb)
 229 {
 230         kcb->prev_kprobe.kp = kprobe_running();
 231         kcb->prev_kprobe.status = kcb->kprobe_status;
 232         kcb->prev_kprobe.old_rflags = kcb->kprobe_old_rflags;
 233         kcb->prev_kprobe.saved_rflags = kcb->kprobe_saved_rflags;
 234 }
 235
 236 static inline void restore_previous_kprobe(struct kprobe_ctlblk *kcb)
 237 {
 238         __get_cpu_var(current_kprobe) = kcb->prev_kprobe.kp;
 239         kcb->kprobe_status = kcb->prev_kprobe.status;
 240         kcb->kprobe_old_rflags = kcb->prev_kprobe.old_rflags;
 241         kcb->kprobe_saved_rflags = kcb->prev_kprobe.saved_rflags;
 242 }
 243
 244 static inline void set_current_kprobe(struct kprobe *p, struct pt_regs *regs,
 245                                 struct kprobe_ctlblk *kcb)
 246 {
 247         __get_cpu_var(current_kprobe) = p;
 248         kcb->kprobe_saved_rflags = kcb->kprobe_old_rflags
 249                 = (regs->eflags & (TF_MASK | IF_MASK));
 250         if (is_IF_modifier(p->ainsn.insn))
 251                 kcb->kprobe_saved_rflags &= ~IF_MASK;
 252 }
 253
 254 static void __kprobes prepare_singlestep(struct kprobe *p, struct pt_regs *regs)
 255 {
 256         regs->eflags |= TF_MASK;
 257         regs->eflags &= ~IF_MASK;
 258         /*single step inline if the instruction is an int3*/
 259         if (p->opcode == BREAKPOINT_INSTRUCTION)
 260                 regs->rip = (unsigned long)p->addr;
 261         else
 262                 regs->rip = (unsigned long)p->ainsn.insn;
 263 }
 264
 265 /* Called with kretprobe_lock held */
 266 void __kprobes arch_prepare_kretprobe(struct kretprobe *rp,
 267                                       struct pt_regs *regs)
 268 {
 269         unsigned long *sara = (unsigned long *)regs->rsp;
 270         struct kretprobe_instance *ri;
 271
 272         if ((ri = get_free_rp_inst(rp)) != NULL) {
 273                 ri->rp = rp;
 274                 ri->task = current;
 275                 ri->ret_addr = (kprobe_opcode_t *) *sara;
 276
 277                 /* Replace the return addr with trampoline addr */
 278                 *sara = (unsigned long) &kretprobe_trampoline;
 279
 280                 add_rp_inst(ri);
 281         } else {
 282                 rp->nmissed++;
 283         }
 284 }
 285
 286 int __kprobes kprobe_handler(struct pt_regs *regs)
 287 {
 288         struct kprobe *p;
 289         int ret = 0;
 290         kprobe_opcode_t *addr = (kprobe_opcode_t *)(regs->rip - sizeof(kprobe_opcode_t));
 291         struct kprobe_ctlblk *kcb;
 292
 293         /*
 294          * We don't want to be preempted for the entire
 295          * duration of kprobe processing
 296          */
 297         preempt_disable();
 298         kcb = get_kprobe_ctlblk();
 299
 300         /* Check we're not actually recursing */
 301         if (kprobe_running()) {
 302                 p = get_kprobe(addr);
 303                 if (p) {
 304                         if (kcb->kprobe_status == KPROBE_HIT_SS &&
 305                                 *p->ainsn.insn == BREAKPOINT_INSTRUCTION) {
 306                                 regs->eflags &= ~TF_MASK;
 307                                 regs->eflags |= kcb->kprobe_saved_rflags;
 308                                 goto no_kprobe;
 309                         } else if (kcb->kprobe_status == KPROBE_HIT_SSDONE) {
 310                                 /* TODO: Provide re-entrancy from
 311                                  * post_kprobes_handler() and avoid exception
 312                                  * stack corruption while single-stepping on
 313                                  * the instruction of the new probe.
 314                                  */
 315                                 arch_disarm_kprobe(p);
 316                                 regs->rip = (unsigned long)p->addr;
 317                                 reset_current_kprobe();
 318                                 ret = 1;
 319                         } else {
 320                                 /* We have reentered the kprobe_handler(), since
 321                                  * another probe was hit while within the
 322                                  * handler. We here save the original kprobe
 323                                  * variables and just single step on instruction
 324                                  * of the new probe without calling any user
 325                                  * handlers.
 326                                  */
 327                                 save_previous_kprobe(kcb);
 328                                 set_current_kprobe(p, regs, kcb);
 329                                 kprobes_inc_nmissed_count(p);
 330                                 prepare_singlestep(p, regs);
 331                                 kcb->kprobe_status = KPROBE_REENTER;
 332                                 return 1;
 333                         }
 334                 } else {
 335                         p = __get_cpu_var(current_kprobe);
 336                         if (p->break_handler && p->break_handler(p, regs)) {
 337                                 goto ss_probe;
 338                         }
 339                 }
 340                 goto no_kprobe;
 341         }
 342
 343         p = get_kprobe(addr);
 344         if (!p) {
 345                 if (*addr != BREAKPOINT_INSTRUCTION) {
 346                         /*
 347                          * The breakpoint instruction was removed right
 348                          * after we hit it.  Another cpu has removed
 349                          * either a probepoint or a debugger breakpoint
 350                          * at this address.  In either case, no further
 351                          * handling of this interrupt is appropriate.
 352                          * Back up over the (now missing) int3 and run
 353                          * the original instruction.
 354                          */
 355                         regs->rip = (unsigned long)addr;
 356                         ret = 1;
 357                 }
 358                 /* Not one of ours: let kernel handle it */
 359                 goto no_kprobe;
 360         }
 361
 362         set_current_kprobe(p, regs, kcb);
 363         kcb->kprobe_status = KPROBE_HIT_ACTIVE;
 364
 365         if (p->pre_handler && p->pre_handler(p, regs))
 366                 /* handler has already set things up, so skip ss setup */
 367                 return 1;
 368
 369 ss_probe:
 370         prepare_singlestep(p, regs);
 371         kcb->kprobe_status = KPROBE_HIT_SS;
 372         return 1;
 373
 374 no_kprobe:
 375         preempt_enable_no_resched();
 376         return ret;
 377 }
 378
 379 /*
 380  * For function-return probes, init_kprobes() establishes a probepoint
 381  * here. When a retprobed function returns, this probe is hit and
 382  * trampoline_probe_handler() runs, calling the kretprobe's handler.
 383  */
 384  void kretprobe_trampoline_holder(void)
 385  {
 386         asm volatile (  ".global kretprobe_trampoline\n"
 387                         "kretprobe_trampoline: \n"
 388                         "nop\n");
 389  }
 390
 391 /*
 392  * Called when we hit the probe point at kretprobe_trampoline
 393  */
 394 int __kprobes trampoline_probe_handler(struct kprobe *p, struct pt_regs *regs)
 395 {
 396         struct kretprobe_instance *ri = NULL;
 397         struct hlist_head *head;
 398         struct hlist_node *node, *tmp;
 399         unsigned long flags, orig_ret_address = 0;
 400         unsigned long trampoline_address =(unsigned long)&kretprobe_trampoline;
 401
 402         spin_lock_irqsave(&kretprobe_lock, flags);
 403         head = kretprobe_inst_table_head(current);
 404
 405         /*
 406          * It is possible to have multiple instances associated with a given
 407          * task either because an multiple functions in the call path
 408          * have a return probe installed on them, and/or more then one return
 409          * return probe was registered for a target function.
 410          *
 411          * We can handle this because:
 412          *     - instances are always inserted at the head of the list
 413          *     - when multiple return probes are registered for the same
 414          *       function, the first instance's ret_addr will point to the
 415          *       real return address, and all the rest will point to
 416          *       kretprobe_trampoline
 417          */
 418         hlist_for_each_entry_safe(ri, node, tmp, head, hlist) {
 419                 if (ri->task != current)
 420                         /* another task is sharing our hash bucket */
 421                         continue;
 422
 423                 if (ri->rp && ri->rp->handler)
 424                         ri->rp->handler(ri, regs);
 425
 426                 orig_ret_address = (unsigned long)ri->ret_addr;
 427                 recycle_rp_inst(ri);
 428
 429                 if (orig_ret_address != trampoline_address)
 430                         /*
 431                          * This is the real return address. Any other
 432                          * instances associated with this task are for
 433                          * other calls deeper on the call stack
 434                          */
 435                         break;
 436         }
 437
 438         BUG_ON(!orig_ret_address || (orig_ret_address == trampoline_address));
 439         regs->rip = orig_ret_address;
 440
 441         reset_current_kprobe();
 442         spin_unlock_irqrestore(&kretprobe_lock, flags);
 443         preempt_enable_no_resched();
 444
 445         /*
 446          * By returning a non-zero value, we are telling
 447          * kprobe_handler() that we don't want the post_handler
 448          * to run (and have re-enabled preemption)
 449          */
 450         return 1;
 451 }
 452
 453 /*
 454  * Called after single-stepping.  p->addr is the address of the
 455  * instruction whose first byte has been replaced by the "int 3"
 456  * instruction.  To avoid the SMP problems that can occur when we
 457  * temporarily put back the original opcode to single-step, we
 458  * single-stepped a copy of the instruction.  The address of this
 459  * copy is p->ainsn.insn.
 460  *
 461  * This function prepares to return from the post-single-step
 462  * interrupt.  We have to fix up the stack as follows:
 463  *
 464  * 0) Except in the case of absolute or indirect jump or call instructions,
 465  * the new rip is relative to the copied instruction.  We need to make
 466  * it relative to the original instruction.
 467  *
 468  * 1) If the single-stepped instruction was pushfl, then the TF and IF
 469  * flags are set in the just-pushed eflags, and may need to be cleared.
 470  *
 471  * 2) If the single-stepped instruction was a call, the return address
 472  * that is atop the stack is the address following the copied instruction.
 473  * We need to make it the address following the original instruction.
 474  */
 475 static void __kprobes resume_execution(struct kprobe *p,
 476                 struct pt_regs *regs, struct kprobe_ctlblk *kcb)
 477 {
 478         unsigned long *tos = (unsigned long *)regs->rsp;
 479         unsigned long next_rip = 0;
 480         unsigned long copy_rip = (unsigned long)p->ainsn.insn;
 481         unsigned long orig_rip = (unsigned long)p->addr;
 482         kprobe_opcode_t *insn = p->ainsn.insn;
 483
 484         /*skip the REX prefix*/
 485         if (*insn >= 0x40 && *insn <= 0x4f)
 486                 insn++;
 487
 488         switch (*insn) {
 489         case 0x9c:              /* pushfl */
 490                 *tos &= ~(TF_MASK | IF_MASK);
 491                 *tos |= kcb->kprobe_old_rflags;
 492                 break;
 493         case 0xc3:              /* ret/lret */
 494         case 0xcb:
 495         case 0xc2:
 496         case 0xca:
 497                 regs->eflags &= ~TF_MASK;
 498                 /* rip is already adjusted, no more changes required*/
 499                 return;
 500         case 0xe8:              /* call relative - Fix return addr */
 501                 *tos = orig_rip + (*tos - copy_rip);
 502                 break;
 503         case 0xff:
 504                 if ((*insn & 0x30) == 0x10) {
 505                         /* call absolute, indirect */
 506                         /* Fix return addr; rip is correct. */
 507                         next_rip = regs->rip;
 508                         *tos = orig_rip + (*tos - copy_rip);
 509                 } else if (((*insn & 0x31) == 0x20) ||  /* jmp near, absolute indirect */
 510                            ((*insn & 0x31) == 0x21)) {  /* jmp far, absolute indirect */
 511                         /* rip is correct. */
 512                         next_rip = regs->rip;
 513                 }
 514                 break;
 515         case 0xea:              /* jmp absolute -- rip is correct */
 516                 next_rip = regs->rip;
 517                 break;
 518         default:
 519                 break;
 520         }
 521
 522         regs->eflags &= ~TF_MASK;
 523         if (next_rip) {
 524                 regs->rip = next_rip;
 525         } else {
 526                 regs->rip = orig_rip + (regs->rip - copy_rip);
 527         }
 528 }
 529
 530 int __kprobes post_kprobe_handler(struct pt_regs *regs)
 531 {
 532         struct kprobe *cur = kprobe_running();
 533         struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
 534
 535         if (!cur)
 536                 return 0;
 537
 538         if ((kcb->kprobe_status != KPROBE_REENTER) && cur->post_handler) {
 539                 kcb->kprobe_status = KPROBE_HIT_SSDONE;
 540                 cur->post_handler(cur, regs, 0);
 541         }
 542
 543         resume_execution(cur, regs, kcb);
 544         regs->eflags |= kcb->kprobe_saved_rflags;
 545
 546         /* Restore the original saved kprobes variables and continue. */
 547         if (kcb->kprobe_status == KPROBE_REENTER) {
 548                 restore_previous_kprobe(kcb);
 549                 goto out;
 550         }
 551         reset_current_kprobe();
 552 out:
 553         preempt_enable_no_resched();
 554
 555         /*
 556          * if somebody else is singlestepping across a probe point, eflags
 557          * will have TF set, in which case, continue the remaining processing
 558          * of do_debug, as if this is not a probe hit.
 559          */
 560         if (regs->eflags & TF_MASK)
 561                 return 0;
 562
 563         return 1;
 564 }
 565
 566 int __kprobes kprobe_fault_handler(struct pt_regs *regs, int trapnr)
 567 {
 568         struct kprobe *cur = kprobe_running();
 569         struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
 570
 571         if (cur->fault_handler && cur->fault_handler(cur, regs, trapnr))
 572                 return 1;
 573
 574         if (kcb->kprobe_status & KPROBE_HIT_SS) {
 575                 resume_execution(cur, regs, kcb);
 576                 regs->eflags |= kcb->kprobe_old_rflags;
 577
 578                 reset_current_kprobe();
 579                 preempt_enable_no_resched();
 580         }
 581         return 0;
 582 }
 583
 584 /*
 585  * Wrapper routine for handling exceptions.
 586  */
 587 int __kprobes kprobe_exceptions_notify(struct notifier_block *self,
 588                                        unsigned long val, void *data)
 589 {
 590         struct die_args *args = (struct die_args *)data;
 591         int ret = NOTIFY_DONE;
 592
 593         switch (val) {
 594         case DIE_INT3:
 595                 if (kprobe_handler(args->regs))
 596                         ret = NOTIFY_STOP;
 597                 break;
 598         case DIE_DEBUG:
 599                 if (post_kprobe_handler(args->regs))
 600                         ret = NOTIFY_STOP;
 601                 break;
 602         case DIE_GPF:
 603         case DIE_PAGE_FAULT:
 604                 /* kprobe_running() needs smp_processor_id() */
 605                 preempt_disable();
 606                 if (kprobe_running() &&
 607                     kprobe_fault_handler(args->regs, args->trapnr))
 608                         ret = NOTIFY_STOP;
 609                 preempt_enable();
 610                 break;
 611         default:
 612                 break;
 613         }
 614         return ret;
 615 }
 616
 617 int __kprobes setjmp_pre_handler(struct kprobe *p, struct pt_regs *regs)
 618 {
 619         struct jprobe *jp = container_of(p, struct jprobe, kp);
 620         unsigned long addr;
 621         struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
 622
 623         kcb->jprobe_saved_regs = *regs;
 624         kcb->jprobe_saved_rsp = (long *) regs->rsp;
 625         addr = (unsigned long)(kcb->jprobe_saved_rsp);
 626         /*
 627          * As Linus pointed out, gcc assumes that the callee
 628          * owns the argument space and could overwrite it, e.g.
 629          * tailcall optimization. So, to be absolutely safe
 630          * we also save and restore enough stack bytes to cover
 631          * the argument area.
 632          */
 633         memcpy(kcb->jprobes_stack, (kprobe_opcode_t *)addr,
 634                         MIN_STACK_SIZE(addr));
 635         regs->eflags &= ~IF_MASK;
 636         regs->rip = (unsigned long)(jp->entry);
 637         return 1;
 638 }
 639
 640 void __kprobes jprobe_return(void)
 641 {
 642         struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
 643
 644         asm volatile ("       xchg   %%rbx,%%rsp     \n"
 645                       "       int3                      \n"
 646                       "       .globl jprobe_return_end  \n"
 647                       "       jprobe_return_end:        \n"
 648                       "       nop                       \n"::"b"
 649                       (kcb->jprobe_saved_rsp):"memory");
 650 }
 651
 652 int __kprobes longjmp_break_handler(struct kprobe *p, struct pt_regs *regs)
 653 {
 654         struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
 655         u8 *addr = (u8 *) (regs->rip - 1);
 656         unsigned long stack_addr = (unsigned long)(kcb->jprobe_saved_rsp);
 657         struct jprobe *jp = container_of(p, struct jprobe, kp);
 658
 659         if ((addr > (u8 *) jprobe_return) && (addr < (u8 *) jprobe_return_end)) {
 660                 if ((long *)regs->rsp != kcb->jprobe_saved_rsp) {
 661                         struct pt_regs *saved_regs =
 662                             container_of(kcb->jprobe_saved_rsp,
 663                                             struct pt_regs, rsp);
 664                         printk("current rsp %p does not match saved rsp %p\n",
 665                                (long *)regs->rsp, kcb->jprobe_saved_rsp);
 666                         printk("Saved registers for jprobe %p\n", jp);
 667                         show_registers(saved_regs);
 668                         printk("Current registers\n");
 669                         show_registers(regs);
 670                         BUG();
 671                 }
 672                 *regs = kcb->jprobe_saved_regs;
 673                 memcpy((kprobe_opcode_t *) stack_addr, kcb->jprobes_stack,
 674                        MIN_STACK_SIZE(stack_addr));
 675                 preempt_enable_no_resched();
 676                 return 1;
 677         }
 678         return 0;
 679 }
 680
 681 static struct kprobe trampoline_p = {
 682         .addr = (kprobe_opcode_t *) &kretprobe_trampoline,
 683         .pre_handler = trampoline_probe_handler
 684 };
 685
 686 int __init arch_init_kprobes(void)
 687 {
 688         return register_kprobe(&trampoline_p);
 689 }