Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* By Ross Biro 1/23/92 */ |
2 | /* | |
3 | * Pentium III FXSR, SSE support | |
4 | * Gareth Hughes <gareth@valinux.com>, May 2000 | |
5 | */ | |
6 | ||
7 | #include <linux/kernel.h> | |
8 | #include <linux/sched.h> | |
9 | #include <linux/mm.h> | |
10 | #include <linux/smp.h> | |
1da177e4 LT |
11 | #include <linux/errno.h> |
12 | #include <linux/ptrace.h> | |
13 | #include <linux/user.h> | |
14 | #include <linux/security.h> | |
15 | #include <linux/audit.h> | |
16 | #include <linux/seccomp.h> | |
7ed20e1a | 17 | #include <linux/signal.h> |
1da177e4 LT |
18 | |
19 | #include <asm/uaccess.h> | |
20 | #include <asm/pgtable.h> | |
21 | #include <asm/system.h> | |
22 | #include <asm/processor.h> | |
23 | #include <asm/i387.h> | |
24 | #include <asm/debugreg.h> | |
25 | #include <asm/ldt.h> | |
26 | #include <asm/desc.h> | |
27 | ||
28 | /* | |
29 | * does not yet catch signals sent when the child dies. | |
30 | * in exit.c or in signal.c. | |
31 | */ | |
32 | ||
9f155b98 CE |
33 | /* |
34 | * Determines which flags the user has access to [1 = access, 0 = no access]. | |
3c36c6aa | 35 | * Prohibits changing ID(21), VIP(20), VIF(19), VM(17), NT(14), IOPL(12-13), IF(9). |
9f155b98 CE |
36 | * Also masks reserved bits (31-22, 15, 5, 3, 1). |
37 | */ | |
3c36c6aa | 38 | #define FLAG_MASK 0x00050dd5 |
1da177e4 LT |
39 | |
40 | /* set's the trap flag. */ | |
41 | #define TRAP_FLAG 0x100 | |
42 | ||
43 | /* | |
44 | * Offset of eflags on child stack.. | |
45 | */ | |
8701ea95 | 46 | #define EFL_OFFSET offsetof(struct pt_regs, eflags) |
1da177e4 LT |
47 | |
48 | static inline struct pt_regs *get_child_regs(struct task_struct *task) | |
49 | { | |
50 | void *stack_top = (void *)task->thread.esp0; | |
51 | return stack_top - sizeof(struct pt_regs); | |
52 | } | |
53 | ||
54 | /* | |
8701ea95 JF |
55 | * This routine will get a word off of the processes privileged stack. |
56 | * the offset is bytes into the pt_regs structure on the stack. | |
57 | * This routine assumes that all the privileged stacks are in our | |
1da177e4 LT |
58 | * data space. |
59 | */ | |
60 | static inline int get_stack_long(struct task_struct *task, int offset) | |
61 | { | |
62 | unsigned char *stack; | |
63 | ||
8701ea95 | 64 | stack = (unsigned char *)task->thread.esp0 - sizeof(struct pt_regs); |
1da177e4 LT |
65 | stack += offset; |
66 | return (*((int *)stack)); | |
67 | } | |
68 | ||
69 | /* | |
8701ea95 JF |
70 | * This routine will put a word on the processes privileged stack. |
71 | * the offset is bytes into the pt_regs structure on the stack. | |
72 | * This routine assumes that all the privileged stacks are in our | |
1da177e4 LT |
73 | * data space. |
74 | */ | |
75 | static inline int put_stack_long(struct task_struct *task, int offset, | |
76 | unsigned long data) | |
77 | { | |
78 | unsigned char * stack; | |
79 | ||
8701ea95 | 80 | stack = (unsigned char *)task->thread.esp0 - sizeof(struct pt_regs); |
1da177e4 LT |
81 | stack += offset; |
82 | *(unsigned long *) stack = data; | |
83 | return 0; | |
84 | } | |
85 | ||
86 | static int putreg(struct task_struct *child, | |
87 | unsigned long regno, unsigned long value) | |
88 | { | |
89 | switch (regno >> 2) { | |
464d1a78 | 90 | case GS: |
1da177e4 LT |
91 | if (value && (value & 3) != 3) |
92 | return -EIO; | |
464d1a78 | 93 | child->thread.gs = value; |
1da177e4 | 94 | return 0; |
1da177e4 LT |
95 | case DS: |
96 | case ES: | |
464d1a78 | 97 | case FS: |
1da177e4 LT |
98 | if (value && (value & 3) != 3) |
99 | return -EIO; | |
100 | value &= 0xffff; | |
101 | break; | |
102 | case SS: | |
103 | case CS: | |
104 | if ((value & 3) != 3) | |
105 | return -EIO; | |
106 | value &= 0xffff; | |
107 | break; | |
108 | case EFL: | |
109 | value &= FLAG_MASK; | |
110 | value |= get_stack_long(child, EFL_OFFSET) & ~FLAG_MASK; | |
111 | break; | |
112 | } | |
464d1a78 | 113 | if (regno > FS*4) |
66e10a44 | 114 | regno -= 1*4; |
8701ea95 | 115 | put_stack_long(child, regno, value); |
1da177e4 LT |
116 | return 0; |
117 | } | |
118 | ||
119 | static unsigned long getreg(struct task_struct *child, | |
120 | unsigned long regno) | |
121 | { | |
122 | unsigned long retval = ~0UL; | |
123 | ||
124 | switch (regno >> 2) { | |
464d1a78 JF |
125 | case GS: |
126 | retval = child->thread.gs; | |
1da177e4 | 127 | break; |
1da177e4 LT |
128 | case DS: |
129 | case ES: | |
464d1a78 | 130 | case FS: |
1da177e4 LT |
131 | case SS: |
132 | case CS: | |
133 | retval = 0xffff; | |
134 | /* fall through */ | |
135 | default: | |
464d1a78 | 136 | if (regno > FS*4) |
66e10a44 | 137 | regno -= 1*4; |
1da177e4 LT |
138 | retval &= get_stack_long(child, regno); |
139 | } | |
140 | return retval; | |
141 | } | |
142 | ||
143 | #define LDT_SEGMENT 4 | |
144 | ||
145 | static unsigned long convert_eip_to_linear(struct task_struct *child, struct pt_regs *regs) | |
146 | { | |
147 | unsigned long addr, seg; | |
148 | ||
149 | addr = regs->eip; | |
150 | seg = regs->xcs & 0xffff; | |
151 | if (regs->eflags & VM_MASK) { | |
152 | addr = (addr & 0xffff) + (seg << 4); | |
153 | return addr; | |
154 | } | |
155 | ||
156 | /* | |
157 | * We'll assume that the code segments in the GDT | |
158 | * are all zero-based. That is largely true: the | |
159 | * TLS segments are used for data, and the PNPBIOS | |
160 | * and APM bios ones we just ignore here. | |
161 | */ | |
162 | if (seg & LDT_SEGMENT) { | |
163 | u32 *desc; | |
164 | unsigned long base; | |
165 | ||
29eb5110 | 166 | seg &= ~7UL; |
1da177e4 | 167 | |
de8aacbe | 168 | mutex_lock(&child->mm->context.lock); |
29eb5110 RM |
169 | if (unlikely((seg >> 3) >= child->mm->context.size)) |
170 | addr = -1L; /* bogus selector, access would fault */ | |
171 | else { | |
172 | desc = child->mm->context.ldt + seg; | |
173 | base = ((desc[0] >> 16) | | |
174 | ((desc[1] & 0xff) << 16) | | |
175 | (desc[1] & 0xff000000)); | |
176 | ||
177 | /* 16-bit code segment? */ | |
178 | if (!((desc[1] >> 22) & 1)) | |
179 | addr &= 0xffff; | |
180 | addr += base; | |
181 | } | |
de8aacbe | 182 | mutex_unlock(&child->mm->context.lock); |
1da177e4 LT |
183 | } |
184 | return addr; | |
185 | } | |
186 | ||
2ade2920 | 187 | static inline int is_setting_trap_flag(struct task_struct *child, struct pt_regs *regs) |
1da177e4 LT |
188 | { |
189 | int i, copied; | |
2ade2920 | 190 | unsigned char opcode[15]; |
1da177e4 LT |
191 | unsigned long addr = convert_eip_to_linear(child, regs); |
192 | ||
193 | copied = access_process_vm(child, addr, opcode, sizeof(opcode), 0); | |
194 | for (i = 0; i < copied; i++) { | |
195 | switch (opcode[i]) { | |
2ade2920 CE |
196 | /* popf and iret */ |
197 | case 0x9d: case 0xcf: | |
1da177e4 LT |
198 | return 1; |
199 | /* opcode and address size prefixes */ | |
200 | case 0x66: case 0x67: | |
201 | continue; | |
202 | /* irrelevant prefixes (segment overrides and repeats) */ | |
203 | case 0x26: case 0x2e: | |
204 | case 0x36: case 0x3e: | |
205 | case 0x64: case 0x65: | |
206 | case 0xf0: case 0xf2: case 0xf3: | |
207 | continue; | |
208 | ||
209 | /* | |
210 | * pushf: NOTE! We should probably not let | |
211 | * the user see the TF bit being set. But | |
212 | * it's more pain than it's worth to avoid | |
213 | * it, and a debugger could emulate this | |
214 | * all in user space if it _really_ cares. | |
215 | */ | |
216 | case 0x9c: | |
217 | default: | |
218 | return 0; | |
219 | } | |
220 | } | |
221 | return 0; | |
222 | } | |
223 | ||
224 | static void set_singlestep(struct task_struct *child) | |
225 | { | |
226 | struct pt_regs *regs = get_child_regs(child); | |
227 | ||
228 | /* | |
229 | * Always set TIF_SINGLESTEP - this guarantees that | |
230 | * we single-step system calls etc.. This will also | |
231 | * cause us to set TF when returning to user mode. | |
232 | */ | |
233 | set_tsk_thread_flag(child, TIF_SINGLESTEP); | |
234 | ||
235 | /* | |
236 | * If TF was already set, don't do anything else | |
237 | */ | |
238 | if (regs->eflags & TRAP_FLAG) | |
239 | return; | |
240 | ||
241 | /* Set TF on the kernel stack.. */ | |
242 | regs->eflags |= TRAP_FLAG; | |
243 | ||
244 | /* | |
245 | * ..but if TF is changed by the instruction we will trace, | |
246 | * don't mark it as being "us" that set it, so that we | |
247 | * won't clear it by hand later. | |
248 | */ | |
2ade2920 | 249 | if (is_setting_trap_flag(child, regs)) |
1da177e4 LT |
250 | return; |
251 | ||
252 | child->ptrace |= PT_DTRACE; | |
253 | } | |
254 | ||
255 | static void clear_singlestep(struct task_struct *child) | |
256 | { | |
257 | /* Always clear TIF_SINGLESTEP... */ | |
258 | clear_tsk_thread_flag(child, TIF_SINGLESTEP); | |
259 | ||
260 | /* But touch TF only if it was set by us.. */ | |
261 | if (child->ptrace & PT_DTRACE) { | |
262 | struct pt_regs *regs = get_child_regs(child); | |
263 | regs->eflags &= ~TRAP_FLAG; | |
264 | child->ptrace &= ~PT_DTRACE; | |
265 | } | |
266 | } | |
267 | ||
268 | /* | |
269 | * Called by kernel/ptrace.c when detaching.. | |
270 | * | |
271 | * Make sure the single step bit is not set. | |
272 | */ | |
273 | void ptrace_disable(struct task_struct *child) | |
274 | { | |
275 | clear_singlestep(child); | |
ab1c23c2 | 276 | clear_tsk_thread_flag(child, TIF_SYSCALL_EMU); |
1da177e4 LT |
277 | } |
278 | ||
279 | /* | |
280 | * Perform get_thread_area on behalf of the traced child. | |
281 | */ | |
282 | static int | |
283 | ptrace_get_thread_area(struct task_struct *child, | |
284 | int idx, struct user_desc __user *user_desc) | |
285 | { | |
286 | struct user_desc info; | |
287 | struct desc_struct *desc; | |
288 | ||
289 | /* | |
290 | * Get the current Thread-Local Storage area: | |
291 | */ | |
292 | ||
293 | #define GET_BASE(desc) ( \ | |
294 | (((desc)->a >> 16) & 0x0000ffff) | \ | |
295 | (((desc)->b << 16) & 0x00ff0000) | \ | |
296 | ( (desc)->b & 0xff000000) ) | |
297 | ||
298 | #define GET_LIMIT(desc) ( \ | |
299 | ((desc)->a & 0x0ffff) | \ | |
300 | ((desc)->b & 0xf0000) ) | |
301 | ||
302 | #define GET_32BIT(desc) (((desc)->b >> 22) & 1) | |
303 | #define GET_CONTENTS(desc) (((desc)->b >> 10) & 3) | |
304 | #define GET_WRITABLE(desc) (((desc)->b >> 9) & 1) | |
305 | #define GET_LIMIT_PAGES(desc) (((desc)->b >> 23) & 1) | |
306 | #define GET_PRESENT(desc) (((desc)->b >> 15) & 1) | |
307 | #define GET_USEABLE(desc) (((desc)->b >> 20) & 1) | |
308 | ||
309 | if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX) | |
310 | return -EINVAL; | |
311 | ||
312 | desc = child->thread.tls_array + idx - GDT_ENTRY_TLS_MIN; | |
313 | ||
314 | info.entry_number = idx; | |
315 | info.base_addr = GET_BASE(desc); | |
316 | info.limit = GET_LIMIT(desc); | |
317 | info.seg_32bit = GET_32BIT(desc); | |
318 | info.contents = GET_CONTENTS(desc); | |
319 | info.read_exec_only = !GET_WRITABLE(desc); | |
320 | info.limit_in_pages = GET_LIMIT_PAGES(desc); | |
321 | info.seg_not_present = !GET_PRESENT(desc); | |
322 | info.useable = GET_USEABLE(desc); | |
323 | ||
324 | if (copy_to_user(user_desc, &info, sizeof(info))) | |
325 | return -EFAULT; | |
326 | ||
327 | return 0; | |
328 | } | |
329 | ||
330 | /* | |
331 | * Perform set_thread_area on behalf of the traced child. | |
332 | */ | |
333 | static int | |
334 | ptrace_set_thread_area(struct task_struct *child, | |
335 | int idx, struct user_desc __user *user_desc) | |
336 | { | |
337 | struct user_desc info; | |
338 | struct desc_struct *desc; | |
339 | ||
340 | if (copy_from_user(&info, user_desc, sizeof(info))) | |
341 | return -EFAULT; | |
342 | ||
343 | if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX) | |
344 | return -EINVAL; | |
345 | ||
346 | desc = child->thread.tls_array + idx - GDT_ENTRY_TLS_MIN; | |
347 | if (LDT_empty(&info)) { | |
348 | desc->a = 0; | |
349 | desc->b = 0; | |
350 | } else { | |
351 | desc->a = LDT_entry_a(&info); | |
352 | desc->b = LDT_entry_b(&info); | |
353 | } | |
354 | ||
355 | return 0; | |
356 | } | |
357 | ||
481bed45 | 358 | long arch_ptrace(struct task_struct *child, long request, long addr, long data) |
1da177e4 | 359 | { |
1da177e4 LT |
360 | struct user * dummy = NULL; |
361 | int i, ret; | |
362 | unsigned long __user *datap = (unsigned long __user *)data; | |
363 | ||
1da177e4 LT |
364 | switch (request) { |
365 | /* when I and D space are separate, these will need to be fixed. */ | |
366 | case PTRACE_PEEKTEXT: /* read word at location addr. */ | |
76647323 AD |
367 | case PTRACE_PEEKDATA: |
368 | ret = generic_ptrace_peekdata(child, addr, data); | |
1da177e4 | 369 | break; |
1da177e4 LT |
370 | |
371 | /* read the word at location addr in the USER area. */ | |
372 | case PTRACE_PEEKUSR: { | |
373 | unsigned long tmp; | |
374 | ||
375 | ret = -EIO; | |
376 | if ((addr & 3) || addr < 0 || | |
377 | addr > sizeof(struct user) - 3) | |
378 | break; | |
379 | ||
380 | tmp = 0; /* Default return condition */ | |
381 | if(addr < FRAME_SIZE*sizeof(long)) | |
382 | tmp = getreg(child, addr); | |
383 | if(addr >= (long) &dummy->u_debugreg[0] && | |
384 | addr <= (long) &dummy->u_debugreg[7]){ | |
385 | addr -= (long) &dummy->u_debugreg[0]; | |
386 | addr = addr >> 2; | |
387 | tmp = child->thread.debugreg[addr]; | |
388 | } | |
389 | ret = put_user(tmp, datap); | |
390 | break; | |
391 | } | |
392 | ||
393 | /* when I and D space are separate, this will have to be fixed. */ | |
394 | case PTRACE_POKETEXT: /* write the word at location addr. */ | |
395 | case PTRACE_POKEDATA: | |
f284ce72 | 396 | ret = generic_ptrace_pokedata(child, addr, data); |
1da177e4 LT |
397 | break; |
398 | ||
399 | case PTRACE_POKEUSR: /* write the word at location addr in the USER area */ | |
400 | ret = -EIO; | |
401 | if ((addr & 3) || addr < 0 || | |
402 | addr > sizeof(struct user) - 3) | |
403 | break; | |
404 | ||
405 | if (addr < FRAME_SIZE*sizeof(long)) { | |
406 | ret = putreg(child, addr, data); | |
407 | break; | |
408 | } | |
409 | /* We need to be very careful here. We implicitly | |
410 | want to modify a portion of the task_struct, and we | |
411 | have to be selective about what portions we allow someone | |
412 | to modify. */ | |
413 | ||
414 | ret = -EIO; | |
415 | if(addr >= (long) &dummy->u_debugreg[0] && | |
416 | addr <= (long) &dummy->u_debugreg[7]){ | |
417 | ||
418 | if(addr == (long) &dummy->u_debugreg[4]) break; | |
419 | if(addr == (long) &dummy->u_debugreg[5]) break; | |
420 | if(addr < (long) &dummy->u_debugreg[4] && | |
421 | ((unsigned long) data) >= TASK_SIZE-3) break; | |
422 | ||
423 | /* Sanity-check data. Take one half-byte at once with | |
424 | * check = (val >> (16 + 4*i)) & 0xf. It contains the | |
425 | * R/Wi and LENi bits; bits 0 and 1 are R/Wi, and bits | |
426 | * 2 and 3 are LENi. Given a list of invalid values, | |
427 | * we do mask |= 1 << invalid_value, so that | |
428 | * (mask >> check) & 1 is a correct test for invalid | |
429 | * values. | |
430 | * | |
431 | * R/Wi contains the type of the breakpoint / | |
432 | * watchpoint, LENi contains the length of the watched | |
433 | * data in the watchpoint case. | |
434 | * | |
435 | * The invalid values are: | |
436 | * - LENi == 0x10 (undefined), so mask |= 0x0f00. | |
437 | * - R/Wi == 0x10 (break on I/O reads or writes), so | |
438 | * mask |= 0x4444. | |
439 | * - R/Wi == 0x00 && LENi != 0x00, so we have mask |= | |
440 | * 0x1110. | |
441 | * | |
442 | * Finally, mask = 0x0f00 | 0x4444 | 0x1110 == 0x5f54. | |
443 | * | |
444 | * See the Intel Manual "System Programming Guide", | |
445 | * 15.2.4 | |
446 | * | |
447 | * Note that LENi == 0x10 is defined on x86_64 in long | |
448 | * mode (i.e. even for 32-bit userspace software, but | |
449 | * 64-bit kernel), so the x86_64 mask value is 0x5454. | |
450 | * See the AMD manual no. 24593 (AMD64 System | |
451 | * Programming)*/ | |
452 | ||
453 | if(addr == (long) &dummy->u_debugreg[7]) { | |
454 | data &= ~DR_CONTROL_RESERVED; | |
455 | for(i=0; i<4; i++) | |
456 | if ((0x5f54 >> ((data >> (16 + 4*i)) & 0xf)) & 1) | |
457 | goto out_tsk; | |
b3cf2576 SE |
458 | if (data) |
459 | set_tsk_thread_flag(child, TIF_DEBUG); | |
460 | else | |
461 | clear_tsk_thread_flag(child, TIF_DEBUG); | |
1da177e4 | 462 | } |
1da177e4 LT |
463 | addr -= (long) &dummy->u_debugreg; |
464 | addr = addr >> 2; | |
465 | child->thread.debugreg[addr] = data; | |
466 | ret = 0; | |
467 | } | |
468 | break; | |
469 | ||
ed75e8d5 | 470 | case PTRACE_SYSEMU: /* continue and stop at next syscall, which will not be executed */ |
1da177e4 LT |
471 | case PTRACE_SYSCALL: /* continue and stop at next (return from) syscall */ |
472 | case PTRACE_CONT: /* restart after signal. */ | |
473 | ret = -EIO; | |
7ed20e1a | 474 | if (!valid_signal(data)) |
1da177e4 | 475 | break; |
ed75e8d5 LV |
476 | if (request == PTRACE_SYSEMU) { |
477 | set_tsk_thread_flag(child, TIF_SYSCALL_EMU); | |
c8c86cec BS |
478 | clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE); |
479 | } else if (request == PTRACE_SYSCALL) { | |
1da177e4 | 480 | set_tsk_thread_flag(child, TIF_SYSCALL_TRACE); |
c8c86cec | 481 | clear_tsk_thread_flag(child, TIF_SYSCALL_EMU); |
ed75e8d5 | 482 | } else { |
c8c86cec | 483 | clear_tsk_thread_flag(child, TIF_SYSCALL_EMU); |
1da177e4 LT |
484 | clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE); |
485 | } | |
486 | child->exit_code = data; | |
487 | /* make sure the single step bit is not set. */ | |
488 | clear_singlestep(child); | |
489 | wake_up_process(child); | |
490 | ret = 0; | |
491 | break; | |
492 | ||
493 | /* | |
494 | * make the child exit. Best I can do is send it a sigkill. | |
495 | * perhaps it should be put in the status that it wants to | |
496 | * exit. | |
497 | */ | |
498 | case PTRACE_KILL: | |
499 | ret = 0; | |
500 | if (child->exit_state == EXIT_ZOMBIE) /* already dead */ | |
501 | break; | |
502 | child->exit_code = SIGKILL; | |
503 | /* make sure the single step bit is not set. */ | |
504 | clear_singlestep(child); | |
505 | wake_up_process(child); | |
506 | break; | |
507 | ||
1b38f006 | 508 | case PTRACE_SYSEMU_SINGLESTEP: /* Same as SYSEMU, but singlestep if not syscall */ |
1da177e4 LT |
509 | case PTRACE_SINGLESTEP: /* set the trap flag. */ |
510 | ret = -EIO; | |
7ed20e1a | 511 | if (!valid_signal(data)) |
1da177e4 | 512 | break; |
1b38f006 BS |
513 | |
514 | if (request == PTRACE_SYSEMU_SINGLESTEP) | |
515 | set_tsk_thread_flag(child, TIF_SYSCALL_EMU); | |
516 | else | |
517 | clear_tsk_thread_flag(child, TIF_SYSCALL_EMU); | |
518 | ||
1da177e4 LT |
519 | clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE); |
520 | set_singlestep(child); | |
521 | child->exit_code = data; | |
522 | /* give it a chance to run. */ | |
523 | wake_up_process(child); | |
524 | ret = 0; | |
525 | break; | |
1da177e4 LT |
526 | |
527 | case PTRACE_GETREGS: { /* Get all gp regs from the child. */ | |
528 | if (!access_ok(VERIFY_WRITE, datap, FRAME_SIZE*sizeof(long))) { | |
529 | ret = -EIO; | |
530 | break; | |
531 | } | |
532 | for ( i = 0; i < FRAME_SIZE*sizeof(long); i += sizeof(long) ) { | |
533 | __put_user(getreg(child, i), datap); | |
534 | datap++; | |
535 | } | |
536 | ret = 0; | |
537 | break; | |
538 | } | |
539 | ||
540 | case PTRACE_SETREGS: { /* Set all gp regs in the child. */ | |
541 | unsigned long tmp; | |
542 | if (!access_ok(VERIFY_READ, datap, FRAME_SIZE*sizeof(long))) { | |
543 | ret = -EIO; | |
544 | break; | |
545 | } | |
546 | for ( i = 0; i < FRAME_SIZE*sizeof(long); i += sizeof(long) ) { | |
547 | __get_user(tmp, datap); | |
548 | putreg(child, i, tmp); | |
549 | datap++; | |
550 | } | |
551 | ret = 0; | |
552 | break; | |
553 | } | |
554 | ||
555 | case PTRACE_GETFPREGS: { /* Get the child FPU state. */ | |
556 | if (!access_ok(VERIFY_WRITE, datap, | |
557 | sizeof(struct user_i387_struct))) { | |
558 | ret = -EIO; | |
559 | break; | |
560 | } | |
561 | ret = 0; | |
562 | if (!tsk_used_math(child)) | |
563 | init_fpu(child); | |
564 | get_fpregs((struct user_i387_struct __user *)data, child); | |
565 | break; | |
566 | } | |
567 | ||
568 | case PTRACE_SETFPREGS: { /* Set the child FPU state. */ | |
569 | if (!access_ok(VERIFY_READ, datap, | |
570 | sizeof(struct user_i387_struct))) { | |
571 | ret = -EIO; | |
572 | break; | |
573 | } | |
574 | set_stopped_child_used_math(child); | |
575 | set_fpregs(child, (struct user_i387_struct __user *)data); | |
576 | ret = 0; | |
577 | break; | |
578 | } | |
579 | ||
580 | case PTRACE_GETFPXREGS: { /* Get the child extended FPU state. */ | |
581 | if (!access_ok(VERIFY_WRITE, datap, | |
582 | sizeof(struct user_fxsr_struct))) { | |
583 | ret = -EIO; | |
584 | break; | |
585 | } | |
586 | if (!tsk_used_math(child)) | |
587 | init_fpu(child); | |
588 | ret = get_fpxregs((struct user_fxsr_struct __user *)data, child); | |
589 | break; | |
590 | } | |
591 | ||
592 | case PTRACE_SETFPXREGS: { /* Set the child extended FPU state. */ | |
593 | if (!access_ok(VERIFY_READ, datap, | |
594 | sizeof(struct user_fxsr_struct))) { | |
595 | ret = -EIO; | |
596 | break; | |
597 | } | |
598 | set_stopped_child_used_math(child); | |
599 | ret = set_fpxregs(child, (struct user_fxsr_struct __user *)data); | |
600 | break; | |
601 | } | |
602 | ||
603 | case PTRACE_GET_THREAD_AREA: | |
604 | ret = ptrace_get_thread_area(child, addr, | |
605 | (struct user_desc __user *) data); | |
606 | break; | |
607 | ||
608 | case PTRACE_SET_THREAD_AREA: | |
609 | ret = ptrace_set_thread_area(child, addr, | |
610 | (struct user_desc __user *) data); | |
611 | break; | |
612 | ||
613 | default: | |
614 | ret = ptrace_request(child, request, addr, data); | |
615 | break; | |
616 | } | |
481bed45 | 617 | out_tsk: |
1da177e4 LT |
618 | return ret; |
619 | } | |
620 | ||
621 | void send_sigtrap(struct task_struct *tsk, struct pt_regs *regs, int error_code) | |
622 | { | |
623 | struct siginfo info; | |
624 | ||
625 | tsk->thread.trap_no = 1; | |
626 | tsk->thread.error_code = error_code; | |
627 | ||
628 | memset(&info, 0, sizeof(info)); | |
629 | info.si_signo = SIGTRAP; | |
630 | info.si_code = TRAP_BRKPT; | |
631 | ||
632 | /* User-mode eip? */ | |
fa1e1bdf | 633 | info.si_addr = user_mode_vm(regs) ? (void __user *) regs->eip : NULL; |
1da177e4 LT |
634 | |
635 | /* Send us the fakey SIGTRAP */ | |
636 | force_sig_info(SIGTRAP, &info, tsk); | |
637 | } | |
638 | ||
639 | /* notification of system call entry/exit | |
640 | * - triggered by current->work.syscall_trace | |
641 | */ | |
642 | __attribute__((regparm(3))) | |
ed75e8d5 | 643 | int do_syscall_trace(struct pt_regs *regs, int entryexit) |
1da177e4 | 644 | { |
4c7fc722 AA |
645 | int is_sysemu = test_thread_flag(TIF_SYSCALL_EMU); |
646 | /* | |
647 | * With TIF_SYSCALL_EMU set we want to ignore TIF_SINGLESTEP for syscall | |
648 | * interception | |
649 | */ | |
1b38f006 | 650 | int is_singlestep = !is_sysemu && test_thread_flag(TIF_SINGLESTEP); |
4c7fc722 | 651 | int ret = 0; |
1b38f006 | 652 | |
1da177e4 | 653 | /* do the secure computing check first */ |
4c7fc722 AA |
654 | if (!entryexit) |
655 | secure_computing(regs->orig_eax); | |
1da177e4 | 656 | |
ab1c23c2 BS |
657 | if (unlikely(current->audit_context)) { |
658 | if (entryexit) | |
5411be59 | 659 | audit_syscall_exit(AUDITSC_RESULT(regs->eax), |
4c7fc722 | 660 | regs->eax); |
ab1c23c2 BS |
661 | /* Debug traps, when using PTRACE_SINGLESTEP, must be sent only |
662 | * on the syscall exit path. Normally, when TIF_SYSCALL_AUDIT is | |
663 | * not used, entry.S will call us only on syscall exit, not | |
664 | * entry; so when TIF_SYSCALL_AUDIT is used we must avoid | |
665 | * calling send_sigtrap() on syscall entry. | |
666 | * | |
667 | * Note that when PTRACE_SYSEMU_SINGLESTEP is used, | |
668 | * is_singlestep is false, despite his name, so we will still do | |
669 | * the correct thing. | |
670 | */ | |
671 | else if (is_singlestep) | |
672 | goto out; | |
673 | } | |
1da177e4 LT |
674 | |
675 | if (!(current->ptrace & PT_PTRACED)) | |
2fd6f58b | 676 | goto out; |
1da177e4 | 677 | |
1b38f006 BS |
678 | /* If a process stops on the 1st tracepoint with SYSCALL_TRACE |
679 | * and then is resumed with SYSEMU_SINGLESTEP, it will come in | |
680 | * here. We have to check this and return */ | |
681 | if (is_sysemu && entryexit) | |
682 | return 0; | |
ed75e8d5 | 683 | |
1da177e4 | 684 | /* Fake a debug trap */ |
c8c86cec | 685 | if (is_singlestep) |
1da177e4 LT |
686 | send_sigtrap(current, regs, 0); |
687 | ||
c8c86cec | 688 | if (!test_thread_flag(TIF_SYSCALL_TRACE) && !is_sysemu) |
2fd6f58b | 689 | goto out; |
1da177e4 LT |
690 | |
691 | /* the 0x80 provides a way for the tracing parent to distinguish | |
692 | between a syscall stop and SIGTRAP delivery */ | |
ed75e8d5 | 693 | /* Note that the debugger could change the result of test_thread_flag!*/ |
4c7fc722 | 694 | ptrace_notify(SIGTRAP | ((current->ptrace & PT_TRACESYSGOOD) ? 0x80:0)); |
1da177e4 LT |
695 | |
696 | /* | |
697 | * this isn't the same as continuing with a signal, but it will do | |
698 | * for normal use. strace only continues with a signal if the | |
699 | * stopping signal is not SIGTRAP. -brl | |
700 | */ | |
701 | if (current->exit_code) { | |
702 | send_sig(current->exit_code, current, 1); | |
703 | current->exit_code = 0; | |
704 | } | |
ed75e8d5 | 705 | ret = is_sysemu; |
4c7fc722 | 706 | out: |
2fd6f58b | 707 | if (unlikely(current->audit_context) && !entryexit) |
5411be59 | 708 | audit_syscall_entry(AUDIT_ARCH_I386, regs->orig_eax, |
2fd6f58b | 709 | regs->ebx, regs->ecx, regs->edx, regs->esi); |
c8c86cec BS |
710 | if (ret == 0) |
711 | return 0; | |
712 | ||
1b38f006 | 713 | regs->orig_eax = -1; /* force skip of syscall restarting */ |
c8c86cec | 714 | if (unlikely(current->audit_context)) |
5411be59 | 715 | audit_syscall_exit(AUDITSC_RESULT(regs->eax), regs->eax); |
c8c86cec | 716 | return 1; |
1da177e4 | 717 | } |