[PATCH] kprobes: fix namespace problem and sparc64 build
[linux-2.6.git] / 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/spinlock.h>
38 #include <linux/string.h>
39 #include <linux/slab.h>
40 #include <linux/preempt.h>
41
42 #include <asm/cacheflush.h>
43 #include <asm/pgtable.h>
44 #include <asm/kdebug.h>
45
46 static DECLARE_MUTEX(kprobe_mutex);
47
48 static struct kprobe *current_kprobe;
49 static unsigned long kprobe_status, kprobe_old_rflags, kprobe_saved_rflags;
50 static struct kprobe *kprobe_prev;
51 static unsigned long kprobe_status_prev, kprobe_old_rflags_prev, kprobe_saved_rflags_prev;
52 static struct pt_regs jprobe_saved_regs;
53 static long *jprobe_saved_rsp;
54 void jprobe_return_end(void);
55
56 /* copy of the kernel stack at the probe fire time */
57 static kprobe_opcode_t jprobes_stack[MAX_STACK_SIZE];
58
59 /*
60  * returns non-zero if opcode modifies the interrupt flag.
61  */
62 static inline int is_IF_modifier(kprobe_opcode_t *insn)
63 {
64         switch (*insn) {
65         case 0xfa:              /* cli */
66         case 0xfb:              /* sti */
67         case 0xcf:              /* iret/iretd */
68         case 0x9d:              /* popf/popfd */
69                 return 1;
70         }
71
72         if (*insn  >= 0x40 && *insn <= 0x4f && *++insn == 0xcf)
73                 return 1;
74         return 0;
75 }
76
77 int arch_prepare_kprobe(struct kprobe *p)
78 {
79         /* insn: must be on special executable page on x86_64. */
80         up(&kprobe_mutex);
81         p->ainsn.insn = get_insn_slot();
82         down(&kprobe_mutex);
83         if (!p->ainsn.insn) {
84                 return -ENOMEM;
85         }
86         return 0;
87 }
88
89 /*
90  * Determine if the instruction uses the %rip-relative addressing mode.
91  * If it does, return the address of the 32-bit displacement word.
92  * If not, return null.
93  */
94 static inline s32 *is_riprel(u8 *insn)
95 {
96 #define W(row,b0,b1,b2,b3,b4,b5,b6,b7,b8,b9,ba,bb,bc,bd,be,bf)                \
97         (((b0##UL << 0x0)|(b1##UL << 0x1)|(b2##UL << 0x2)|(b3##UL << 0x3) |   \
98           (b4##UL << 0x4)|(b5##UL << 0x5)|(b6##UL << 0x6)|(b7##UL << 0x7) |   \
99           (b8##UL << 0x8)|(b9##UL << 0x9)|(ba##UL << 0xa)|(bb##UL << 0xb) |   \
100           (bc##UL << 0xc)|(bd##UL << 0xd)|(be##UL << 0xe)|(bf##UL << 0xf))    \
101          << (row % 64))
102         static const u64 onebyte_has_modrm[256 / 64] = {
103                 /*      0 1 2 3 4 5 6 7 8 9 a b c d e f         */
104                 /*      -------------------------------         */
105                 W(0x00, 1,1,1,1,0,0,0,0,1,1,1,1,0,0,0,0)| /* 00 */
106                 W(0x10, 1,1,1,1,0,0,0,0,1,1,1,1,0,0,0,0)| /* 10 */
107                 W(0x20, 1,1,1,1,0,0,0,0,1,1,1,1,0,0,0,0)| /* 20 */
108                 W(0x30, 1,1,1,1,0,0,0,0,1,1,1,1,0,0,0,0), /* 30 */
109                 W(0x40, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0)| /* 40 */
110                 W(0x50, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0)| /* 50 */
111                 W(0x60, 0,0,1,1,0,0,0,0,0,1,0,1,0,0,0,0)| /* 60 */
112                 W(0x70, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0), /* 70 */
113                 W(0x80, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)| /* 80 */
114                 W(0x90, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0)| /* 90 */
115                 W(0xa0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0)| /* a0 */
116                 W(0xb0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0), /* b0 */
117                 W(0xc0, 1,1,0,0,1,1,1,1,0,0,0,0,0,0,0,0)| /* c0 */
118                 W(0xd0, 1,1,1,1,0,0,0,0,1,1,1,1,1,1,1,1)| /* d0 */
119                 W(0xe0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0)| /* e0 */
120                 W(0xf0, 0,0,0,0,0,0,1,1,0,0,0,0,0,0,1,1)  /* f0 */
121                 /*      -------------------------------         */
122                 /*      0 1 2 3 4 5 6 7 8 9 a b c d e f         */
123         };
124         static const u64 twobyte_has_modrm[256 / 64] = {
125                 /*      0 1 2 3 4 5 6 7 8 9 a b c d e f         */
126                 /*      -------------------------------         */
127                 W(0x00, 1,1,1,1,0,0,0,0,0,0,0,0,0,1,0,1)| /* 0f */
128                 W(0x10, 1,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0)| /* 1f */
129                 W(0x20, 1,1,1,1,1,0,1,0,1,1,1,1,1,1,1,1)| /* 2f */
130                 W(0x30, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0), /* 3f */
131                 W(0x40, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)| /* 4f */
132                 W(0x50, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)| /* 5f */
133                 W(0x60, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)| /* 6f */
134                 W(0x70, 1,1,1,1,1,1,1,0,0,0,0,0,1,1,1,1), /* 7f */
135                 W(0x80, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0)| /* 8f */
136                 W(0x90, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)| /* 9f */
137                 W(0xa0, 0,0,0,1,1,1,1,1,0,0,0,1,1,1,1,1)| /* af */
138                 W(0xb0, 1,1,1,1,1,1,1,1,0,0,1,1,1,1,1,1), /* bf */
139                 W(0xc0, 1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0)| /* cf */
140                 W(0xd0, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)| /* df */
141                 W(0xe0, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)| /* ef */
142                 W(0xf0, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,0)  /* ff */
143                 /*      -------------------------------         */
144                 /*      0 1 2 3 4 5 6 7 8 9 a b c d e f         */
145         };
146 #undef  W
147         int need_modrm;
148
149         /* Skip legacy instruction prefixes.  */
150         while (1) {
151                 switch (*insn) {
152                 case 0x66:
153                 case 0x67:
154                 case 0x2e:
155                 case 0x3e:
156                 case 0x26:
157                 case 0x64:
158                 case 0x65:
159                 case 0x36:
160                 case 0xf0:
161                 case 0xf3:
162                 case 0xf2:
163                         ++insn;
164                         continue;
165                 }
166                 break;
167         }
168
169         /* Skip REX instruction prefix.  */
170         if ((*insn & 0xf0) == 0x40)
171                 ++insn;
172
173         if (*insn == 0x0f) {    /* Two-byte opcode.  */
174                 ++insn;
175                 need_modrm = test_bit(*insn, twobyte_has_modrm);
176         } else {                /* One-byte opcode.  */
177                 need_modrm = test_bit(*insn, onebyte_has_modrm);
178         }
179
180         if (need_modrm) {
181                 u8 modrm = *++insn;
182                 if ((modrm & 0xc7) == 0x05) { /* %rip+disp32 addressing mode */
183                         /* Displacement follows ModRM byte.  */
184                         return (s32 *) ++insn;
185                 }
186         }
187
188         /* No %rip-relative addressing mode here.  */
189         return NULL;
190 }
191
192 void arch_copy_kprobe(struct kprobe *p)
193 {
194         s32 *ripdisp;
195         memcpy(p->ainsn.insn, p->addr, MAX_INSN_SIZE);
196         ripdisp = is_riprel(p->ainsn.insn);
197         if (ripdisp) {
198                 /*
199                  * The copied instruction uses the %rip-relative
200                  * addressing mode.  Adjust the displacement for the
201                  * difference between the original location of this
202                  * instruction and the location of the copy that will
203                  * actually be run.  The tricky bit here is making sure
204                  * that the sign extension happens correctly in this
205                  * calculation, since we need a signed 32-bit result to
206                  * be sign-extended to 64 bits when it's added to the
207                  * %rip value and yield the same 64-bit result that the
208                  * sign-extension of the original signed 32-bit
209                  * displacement would have given.
210                  */
211                 s64 disp = (u8 *) p->addr + *ripdisp - (u8 *) p->ainsn.insn;
212                 BUG_ON((s64) (s32) disp != disp); /* Sanity check.  */
213                 *ripdisp = disp;
214         }
215         p->opcode = *p->addr;
216 }
217
218 void arch_arm_kprobe(struct kprobe *p)
219 {
220         *p->addr = BREAKPOINT_INSTRUCTION;
221         flush_icache_range((unsigned long) p->addr,
222                            (unsigned long) p->addr + sizeof(kprobe_opcode_t));
223 }
224
225 void arch_disarm_kprobe(struct kprobe *p)
226 {
227         *p->addr = p->opcode;
228         flush_icache_range((unsigned long) p->addr,
229                            (unsigned long) p->addr + sizeof(kprobe_opcode_t));
230 }
231
232 void arch_remove_kprobe(struct kprobe *p)
233 {
234         up(&kprobe_mutex);
235         free_insn_slot(p->ainsn.insn);
236         down(&kprobe_mutex);
237 }
238
239 static inline void save_previous_kprobe(void)
240 {
241         kprobe_prev = current_kprobe;
242         kprobe_status_prev = kprobe_status;
243         kprobe_old_rflags_prev = kprobe_old_rflags;
244         kprobe_saved_rflags_prev = kprobe_saved_rflags;
245 }
246
247 static inline void restore_previous_kprobe(void)
248 {
249         current_kprobe = kprobe_prev;
250         kprobe_status = kprobe_status_prev;
251         kprobe_old_rflags = kprobe_old_rflags_prev;
252         kprobe_saved_rflags = kprobe_saved_rflags_prev;
253 }
254
255 static inline void set_current_kprobe(struct kprobe *p, struct pt_regs *regs)
256 {
257         current_kprobe = p;
258         kprobe_saved_rflags = kprobe_old_rflags
259                 = (regs->eflags & (TF_MASK | IF_MASK));
260         if (is_IF_modifier(p->ainsn.insn))
261                 kprobe_saved_rflags &= ~IF_MASK;
262 }
263
264 static void prepare_singlestep(struct kprobe *p, struct pt_regs *regs)
265 {
266         regs->eflags |= TF_MASK;
267         regs->eflags &= ~IF_MASK;
268         /*single step inline if the instruction is an int3*/
269         if (p->opcode == BREAKPOINT_INSTRUCTION)
270                 regs->rip = (unsigned long)p->addr;
271         else
272                 regs->rip = (unsigned long)p->ainsn.insn;
273 }
274
275 void arch_prepare_kretprobe(struct kretprobe *rp, struct pt_regs *regs)
276 {
277         unsigned long *sara = (unsigned long *)regs->rsp;
278         struct kretprobe_instance *ri;
279
280         if ((ri = get_free_rp_inst(rp)) != NULL) {
281                 ri->rp = rp;
282                 ri->task = current;
283                 ri->ret_addr = (kprobe_opcode_t *) *sara;
284
285                 /* Replace the return addr with trampoline addr */
286                 *sara = (unsigned long) &kretprobe_trampoline;
287
288                 add_rp_inst(ri);
289         } else {
290                 rp->nmissed++;
291         }
292 }
293
294 /*
295  * Interrupts are disabled on entry as trap3 is an interrupt gate and they
296  * remain disabled thorough out this function.
297  */
298 int kprobe_handler(struct pt_regs *regs)
299 {
300         struct kprobe *p;
301         int ret = 0;
302         kprobe_opcode_t *addr = (kprobe_opcode_t *)(regs->rip - sizeof(kprobe_opcode_t));
303
304         /* We're in an interrupt, but this is clear and BUG()-safe. */
305         preempt_disable();
306
307         /* Check we're not actually recursing */
308         if (kprobe_running()) {
309                 /* We *are* holding lock here, so this is safe.
310                    Disarm the probe we just hit, and ignore it. */
311                 p = get_kprobe(addr);
312                 if (p) {
313                         if (kprobe_status == KPROBE_HIT_SS) {
314                                 regs->eflags &= ~TF_MASK;
315                                 regs->eflags |= kprobe_saved_rflags;
316                                 unlock_kprobes();
317                                 goto no_kprobe;
318                         } else if (kprobe_status == KPROBE_HIT_SSDONE) {
319                                 /* TODO: Provide re-entrancy from
320                                  * post_kprobes_handler() and avoid exception
321                                  * stack corruption while single-stepping on
322                                  * the instruction of the new probe.
323                                  */
324                                 arch_disarm_kprobe(p);
325                                 regs->rip = (unsigned long)p->addr;
326                                 ret = 1;
327                         } else {
328                                 /* We have reentered the kprobe_handler(), since
329                                  * another probe was hit while within the
330                                  * handler. We here save the original kprobe
331                                  * variables and just single step on instruction
332                                  * of the new probe without calling any user
333                                  * handlers.
334                                  */
335                                 save_previous_kprobe();
336                                 set_current_kprobe(p, regs);
337                                 p->nmissed++;
338                                 prepare_singlestep(p, regs);
339                                 kprobe_status = KPROBE_REENTER;
340                                 return 1;
341                         }
342                 } else {
343                         p = current_kprobe;
344                         if (p->break_handler && p->break_handler(p, regs)) {
345                                 goto ss_probe;
346                         }
347                 }
348                 /* If it's not ours, can't be delete race, (we hold lock). */
349                 goto no_kprobe;
350         }
351
352         lock_kprobes();
353         p = get_kprobe(addr);
354         if (!p) {
355                 unlock_kprobes();
356                 if (*addr != BREAKPOINT_INSTRUCTION) {
357                         /*
358                          * The breakpoint instruction was removed right
359                          * after we hit it.  Another cpu has removed
360                          * either a probepoint or a debugger breakpoint
361                          * at this address.  In either case, no further
362                          * handling of this interrupt is appropriate.
363                          */
364                         ret = 1;
365                 }
366                 /* Not one of ours: let kernel handle it */
367                 goto no_kprobe;
368         }
369
370         kprobe_status = KPROBE_HIT_ACTIVE;
371         set_current_kprobe(p, regs);
372
373         if (p->pre_handler && p->pre_handler(p, regs))
374                 /* handler has already set things up, so skip ss setup */
375                 return 1;
376
377 ss_probe:
378         prepare_singlestep(p, regs);
379         kprobe_status = KPROBE_HIT_SS;
380         return 1;
381
382 no_kprobe:
383         preempt_enable_no_resched();
384         return ret;
385 }
386
387 /*
388  * For function-return probes, init_kprobes() establishes a probepoint
389  * here. When a retprobed function returns, this probe is hit and
390  * trampoline_probe_handler() runs, calling the kretprobe's handler.
391  */
392  void kretprobe_trampoline_holder(void)
393  {
394         asm volatile (  ".global kretprobe_trampoline\n"
395                         "kretprobe_trampoline: \n"
396                         "nop\n");
397  }
398
399 /*
400  * Called when we hit the probe point at kretprobe_trampoline
401  */
402 int trampoline_probe_handler(struct kprobe *p, struct pt_regs *regs)
403 {
404         struct kretprobe_instance *ri = NULL;
405         struct hlist_head *head;
406         struct hlist_node *node, *tmp;
407         unsigned long orig_ret_address = 0;
408         unsigned long trampoline_address =(unsigned long)&kretprobe_trampoline;
409
410         head = kretprobe_inst_table_head(current);
411
412         /*
413          * It is possible to have multiple instances associated with a given
414          * task either because an multiple functions in the call path
415          * have a return probe installed on them, and/or more then one return
416          * return probe was registered for a target function.
417          *
418          * We can handle this because:
419          *     - instances are always inserted at the head of the list
420          *     - when multiple return probes are registered for the same
421          *       function, the first instance's ret_addr will point to the
422          *       real return address, and all the rest will point to
423          *       kretprobe_trampoline
424          */
425         hlist_for_each_entry_safe(ri, node, tmp, head, hlist) {
426                 if (ri->task != current)
427                         /* another task is sharing our hash bucket */
428                         continue;
429
430                 if (ri->rp && ri->rp->handler)
431                         ri->rp->handler(ri, regs);
432
433                 orig_ret_address = (unsigned long)ri->ret_addr;
434                 recycle_rp_inst(ri);
435
436                 if (orig_ret_address != trampoline_address)
437                         /*
438                          * This is the real return address. Any other
439                          * instances associated with this task are for
440                          * other calls deeper on the call stack
441                          */
442                         break;
443         }
444
445         BUG_ON(!orig_ret_address || (orig_ret_address == trampoline_address));
446         regs->rip = orig_ret_address;
447
448         unlock_kprobes();
449         preempt_enable_no_resched();
450
451         /*
452          * By returning a non-zero value, we are telling
453          * kprobe_handler() that we have handled unlocking
454          * and re-enabling preemption.
455          */
456         return 1;
457 }
458
459 /*
460  * Called after single-stepping.  p->addr is the address of the
461  * instruction whose first byte has been replaced by the "int 3"
462  * instruction.  To avoid the SMP problems that can occur when we
463  * temporarily put back the original opcode to single-step, we
464  * single-stepped a copy of the instruction.  The address of this
465  * copy is p->ainsn.insn.
466  *
467  * This function prepares to return from the post-single-step
468  * interrupt.  We have to fix up the stack as follows:
469  *
470  * 0) Except in the case of absolute or indirect jump or call instructions,
471  * the new rip is relative to the copied instruction.  We need to make
472  * it relative to the original instruction.
473  *
474  * 1) If the single-stepped instruction was pushfl, then the TF and IF
475  * flags are set in the just-pushed eflags, and may need to be cleared.
476  *
477  * 2) If the single-stepped instruction was a call, the return address
478  * that is atop the stack is the address following the copied instruction.
479  * We need to make it the address following the original instruction.
480  */
481 static void resume_execution(struct kprobe *p, struct pt_regs *regs)
482 {
483         unsigned long *tos = (unsigned long *)regs->rsp;
484         unsigned long next_rip = 0;
485         unsigned long copy_rip = (unsigned long)p->ainsn.insn;
486         unsigned long orig_rip = (unsigned long)p->addr;
487         kprobe_opcode_t *insn = p->ainsn.insn;
488
489         /*skip the REX prefix*/
490         if (*insn >= 0x40 && *insn <= 0x4f)
491                 insn++;
492
493         switch (*insn) {
494         case 0x9c:              /* pushfl */
495                 *tos &= ~(TF_MASK | IF_MASK);
496                 *tos |= kprobe_old_rflags;
497                 break;
498         case 0xc3:              /* ret/lret */
499         case 0xcb:
500         case 0xc2:
501         case 0xca:
502                 regs->eflags &= ~TF_MASK;
503                 /* rip is already adjusted, no more changes required*/
504                 return;
505         case 0xe8:              /* call relative - Fix return addr */
506                 *tos = orig_rip + (*tos - copy_rip);
507                 break;
508         case 0xff:
509                 if ((*insn & 0x30) == 0x10) {
510                         /* call absolute, indirect */
511                         /* Fix return addr; rip is correct. */
512                         next_rip = regs->rip;
513                         *tos = orig_rip + (*tos - copy_rip);
514                 } else if (((*insn & 0x31) == 0x20) ||  /* jmp near, absolute indirect */
515                            ((*insn & 0x31) == 0x21)) {  /* jmp far, absolute indirect */
516                         /* rip is correct. */
517                         next_rip = regs->rip;
518                 }
519                 break;
520         case 0xea:              /* jmp absolute -- rip is correct */
521                 next_rip = regs->rip;
522                 break;
523         default:
524                 break;
525         }
526
527         regs->eflags &= ~TF_MASK;
528         if (next_rip) {
529                 regs->rip = next_rip;
530         } else {
531                 regs->rip = orig_rip + (regs->rip - copy_rip);
532         }
533 }
534
535 /*
536  * Interrupts are disabled on entry as trap1 is an interrupt gate and they
537  * remain disabled thoroughout this function.  And we hold kprobe lock.
538  */
539 int post_kprobe_handler(struct pt_regs *regs)
540 {
541         if (!kprobe_running())
542                 return 0;
543
544         if ((kprobe_status != KPROBE_REENTER) && current_kprobe->post_handler) {
545                 kprobe_status = KPROBE_HIT_SSDONE;
546                 current_kprobe->post_handler(current_kprobe, regs, 0);
547         }
548
549         resume_execution(current_kprobe, regs);
550         regs->eflags |= kprobe_saved_rflags;
551
552         /* Restore the original saved kprobes variables and continue. */
553         if (kprobe_status == KPROBE_REENTER) {
554                 restore_previous_kprobe();
555                 goto out;
556         } else {
557                 unlock_kprobes();
558         }
559 out:
560         preempt_enable_no_resched();
561
562         /*
563          * if somebody else is singlestepping across a probe point, eflags
564          * will have TF set, in which case, continue the remaining processing
565          * of do_debug, as if this is not a probe hit.
566          */
567         if (regs->eflags & TF_MASK)
568                 return 0;
569
570         return 1;
571 }
572
573 /* Interrupts disabled, kprobe_lock held. */
574 int kprobe_fault_handler(struct pt_regs *regs, int trapnr)
575 {
576         if (current_kprobe->fault_handler
577             && current_kprobe->fault_handler(current_kprobe, regs, trapnr))
578                 return 1;
579
580         if (kprobe_status & KPROBE_HIT_SS) {
581                 resume_execution(current_kprobe, regs);
582                 regs->eflags |= kprobe_old_rflags;
583
584                 unlock_kprobes();
585                 preempt_enable_no_resched();
586         }
587         return 0;
588 }
589
590 /*
591  * Wrapper routine for handling exceptions.
592  */
593 int kprobe_exceptions_notify(struct notifier_block *self, unsigned long val,
594                              void *data)
595 {
596         struct die_args *args = (struct die_args *)data;
597         switch (val) {
598         case DIE_INT3:
599                 if (kprobe_handler(args->regs))
600                         return NOTIFY_STOP;
601                 break;
602         case DIE_DEBUG:
603                 if (post_kprobe_handler(args->regs))
604                         return NOTIFY_STOP;
605                 break;
606         case DIE_GPF:
607                 if (kprobe_running() &&
608                     kprobe_fault_handler(args->regs, args->trapnr))
609                         return NOTIFY_STOP;
610                 break;
611         case DIE_PAGE_FAULT:
612                 if (kprobe_running() &&
613                     kprobe_fault_handler(args->regs, args->trapnr))
614                         return NOTIFY_STOP;
615                 break;
616         default:
617                 break;
618         }
619         return NOTIFY_DONE;
620 }
621
622 int setjmp_pre_handler(struct kprobe *p, struct pt_regs *regs)
623 {
624         struct jprobe *jp = container_of(p, struct jprobe, kp);
625         unsigned long addr;
626
627         jprobe_saved_regs = *regs;
628         jprobe_saved_rsp = (long *) regs->rsp;
629         addr = (unsigned long)jprobe_saved_rsp;
630         /*
631          * As Linus pointed out, gcc assumes that the callee
632          * owns the argument space and could overwrite it, e.g.
633          * tailcall optimization. So, to be absolutely safe
634          * we also save and restore enough stack bytes to cover
635          * the argument area.
636          */
637         memcpy(jprobes_stack, (kprobe_opcode_t *) addr, MIN_STACK_SIZE(addr));
638         regs->eflags &= ~IF_MASK;
639         regs->rip = (unsigned long)(jp->entry);
640         return 1;
641 }
642
643 void jprobe_return(void)
644 {
645         preempt_enable_no_resched();
646         asm volatile ("       xchg   %%rbx,%%rsp     \n"
647                       "       int3                      \n"
648                       "       .globl jprobe_return_end  \n"
649                       "       jprobe_return_end:        \n"
650                       "       nop                       \n"::"b"
651                       (jprobe_saved_rsp):"memory");
652 }
653
654 int longjmp_break_handler(struct kprobe *p, struct pt_regs *regs)
655 {
656         u8 *addr = (u8 *) (regs->rip - 1);
657         unsigned long stack_addr = (unsigned long)jprobe_saved_rsp;
658         struct jprobe *jp = container_of(p, struct jprobe, kp);
659
660         if ((addr > (u8 *) jprobe_return) && (addr < (u8 *) jprobe_return_end)) {
661                 if ((long *)regs->rsp != jprobe_saved_rsp) {
662                         struct pt_regs *saved_regs =
663                             container_of(jprobe_saved_rsp, struct pt_regs, rsp);
664                         printk("current rsp %p does not match saved rsp %p\n",
665                                (long *)regs->rsp, 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 = jprobe_saved_regs;
673                 memcpy((kprobe_opcode_t *) stack_addr, jprobes_stack,
674                        MIN_STACK_SIZE(stack_addr));
675                 return 1;
676         }
677         return 0;
678 }
679
680 static struct kprobe trampoline_p = {
681         .addr = (kprobe_opcode_t *) &kretprobe_trampoline,
682         .pre_handler = trampoline_probe_handler
683 };
684
685 int __init arch_init_kprobes(void)
686 {
687         return register_kprobe(&trampoline_p);
688 }