Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * linux/kernel/ptrace.c | |
4 | * | |
5 | * (C) Copyright 1999 Linus Torvalds | |
6 | * | |
7 | * Common interfaces for "ptrace()" which we do not want | |
8 | * to continually duplicate across every architecture. | |
9 | */ | |
10 | ||
c59ede7b | 11 | #include <linux/capability.h> |
9984de1a | 12 | #include <linux/export.h> |
1da177e4 | 13 | #include <linux/sched.h> |
6e84f315 | 14 | #include <linux/sched/mm.h> |
f7ccbae4 | 15 | #include <linux/sched/coredump.h> |
29930025 | 16 | #include <linux/sched/task.h> |
1da177e4 LT |
17 | #include <linux/errno.h> |
18 | #include <linux/mm.h> | |
19 | #include <linux/highmem.h> | |
20 | #include <linux/pagemap.h> | |
1da177e4 LT |
21 | #include <linux/ptrace.h> |
22 | #include <linux/security.h> | |
7ed20e1a | 23 | #include <linux/signal.h> |
a27bb332 | 24 | #include <linux/uio.h> |
a5cb013d | 25 | #include <linux/audit.h> |
b488893a | 26 | #include <linux/pid_namespace.h> |
f17d30a8 | 27 | #include <linux/syscalls.h> |
3a709703 | 28 | #include <linux/uaccess.h> |
2225a122 | 29 | #include <linux/regset.h> |
bf26c018 | 30 | #include <linux/hw_breakpoint.h> |
f701e5b7 | 31 | #include <linux/cn_proc.h> |
84c751bd | 32 | #include <linux/compat.h> |
fcfc2aa0 | 33 | #include <linux/sched/signal.h> |
90f093fa | 34 | #include <linux/minmax.h> |
3f67987c | 35 | #include <linux/syscall_user_dispatch.h> |
1da177e4 | 36 | |
201766a2 EK |
37 | #include <asm/syscall.h> /* for syscall_get_* */ |
38 | ||
84d77d3f EB |
39 | /* |
40 | * Access another process' address space via ptrace. | |
41 | * Source/target buffer must be kernel space, | |
42 | * Do not walk the page table directly, use get_user_pages | |
43 | */ | |
44 | int ptrace_access_vm(struct task_struct *tsk, unsigned long addr, | |
45 | void *buf, int len, unsigned int gup_flags) | |
46 | { | |
47 | struct mm_struct *mm; | |
48 | int ret; | |
49 | ||
50 | mm = get_task_mm(tsk); | |
51 | if (!mm) | |
52 | return 0; | |
53 | ||
54 | if (!tsk->ptrace || | |
55 | (current != tsk->parent) || | |
56 | ((get_dumpable(mm) != SUID_DUMP_USER) && | |
57 | !ptracer_capable(tsk, mm->user_ns))) { | |
58 | mmput(mm); | |
59 | return 0; | |
60 | } | |
61 | ||
c43cfa42 | 62 | ret = access_remote_vm(mm, addr, buf, len, gup_flags); |
84d77d3f EB |
63 | mmput(mm); |
64 | ||
65 | return ret; | |
66 | } | |
67 | ||
bf53de90 | 68 | |
c70d9d80 EB |
69 | void __ptrace_link(struct task_struct *child, struct task_struct *new_parent, |
70 | const struct cred *ptracer_cred) | |
71 | { | |
72 | BUG_ON(!list_empty(&child->ptrace_entry)); | |
73 | list_add(&child->ptrace_entry, &new_parent->ptraced); | |
74 | child->parent = new_parent; | |
75 | child->ptracer_cred = get_cred(ptracer_cred); | |
76 | } | |
77 | ||
1da177e4 LT |
78 | /* |
79 | * ptrace a task: make the debugger its new parent and | |
80 | * move it to the ptrace list. | |
81 | * | |
82 | * Must be called with the tasklist lock write-held. | |
83 | */ | |
c70d9d80 | 84 | static void ptrace_link(struct task_struct *child, struct task_struct *new_parent) |
1da177e4 | 85 | { |
6994eefb | 86 | __ptrace_link(child, new_parent, current_cred()); |
1da177e4 | 87 | } |
3a709703 | 88 | |
e3bd058f TH |
89 | /** |
90 | * __ptrace_unlink - unlink ptracee and restore its execution state | |
91 | * @child: ptracee to be unlinked | |
1da177e4 | 92 | * |
0e9f0a4a TH |
93 | * Remove @child from the ptrace list, move it back to the original parent, |
94 | * and restore the execution state so that it conforms to the group stop | |
95 | * state. | |
96 | * | |
97 | * Unlinking can happen via two paths - explicit PTRACE_DETACH or ptracer | |
98 | * exiting. For PTRACE_DETACH, unless the ptracee has been killed between | |
99 | * ptrace_check_attach() and here, it's guaranteed to be in TASK_TRACED. | |
100 | * If the ptracer is exiting, the ptracee can be in any state. | |
101 | * | |
102 | * After detach, the ptracee should be in a state which conforms to the | |
103 | * group stop. If the group is stopped or in the process of stopping, the | |
104 | * ptracee should be put into TASK_STOPPED; otherwise, it should be woken | |
105 | * up from TASK_TRACED. | |
106 | * | |
107 | * If the ptracee is in TASK_TRACED and needs to be moved to TASK_STOPPED, | |
108 | * it goes through TRACED -> RUNNING -> STOPPED transition which is similar | |
109 | * to but in the opposite direction of what happens while attaching to a | |
110 | * stopped task. However, in this direction, the intermediate RUNNING | |
111 | * state is not hidden even from the current ptracer and if it immediately | |
112 | * re-attaches and performs a WNOHANG wait(2), it may fail. | |
e3bd058f TH |
113 | * |
114 | * CONTEXT: | |
115 | * write_lock_irq(tasklist_lock) | |
1da177e4 | 116 | */ |
36c8b586 | 117 | void __ptrace_unlink(struct task_struct *child) |
1da177e4 | 118 | { |
64b875f7 | 119 | const struct cred *old_cred; |
5ecfbae0 ON |
120 | BUG_ON(!child->ptrace); |
121 | ||
64c19ba2 | 122 | clear_task_syscall_work(child, SYSCALL_TRACE); |
64eb35f7 GKB |
123 | #if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU) |
124 | clear_task_syscall_work(child, SYSCALL_EMU); | |
15532fd6 | 125 | #endif |
0a5bf409 | 126 | |
f470021a RM |
127 | child->parent = child->real_parent; |
128 | list_del_init(&child->ptrace_entry); | |
64b875f7 EB |
129 | old_cred = child->ptracer_cred; |
130 | child->ptracer_cred = NULL; | |
131 | put_cred(old_cred); | |
1da177e4 | 132 | |
1da177e4 | 133 | spin_lock(&child->sighand->siglock); |
1333ab03 | 134 | child->ptrace = 0; |
73ddff2b TH |
135 | /* |
136 | * Clear all pending traps and TRAPPING. TRAPPING should be | |
137 | * cleared regardless of JOBCTL_STOP_PENDING. Do it explicitly. | |
138 | */ | |
139 | task_clear_jobctl_pending(child, JOBCTL_TRAP_MASK); | |
140 | task_clear_jobctl_trapping(child); | |
141 | ||
0e9f0a4a | 142 | /* |
a8f072c1 | 143 | * Reinstate JOBCTL_STOP_PENDING if group stop is in effect and |
0e9f0a4a TH |
144 | * @child isn't dead. |
145 | */ | |
146 | if (!(child->flags & PF_EXITING) && | |
147 | (child->signal->flags & SIGNAL_STOP_STOPPED || | |
27bbb2a0 | 148 | child->signal->group_stop_count)) |
a8f072c1 | 149 | child->jobctl |= JOBCTL_STOP_PENDING; |
0e9f0a4a TH |
150 | |
151 | /* | |
152 | * If transition to TASK_STOPPED is pending or in TASK_TRACED, kick | |
153 | * @child in the butt. Note that @resume should be used iff @child | |
154 | * is in TASK_TRACED; otherwise, we might unduly disrupt | |
155 | * TASK_KILLABLE sleeps. | |
156 | */ | |
a8f072c1 | 157 | if (child->jobctl & JOBCTL_STOP_PENDING || task_is_traced(child)) |
910ffdb1 | 158 | ptrace_signal_wake_up(child, true); |
0e9f0a4a | 159 | |
1da177e4 | 160 | spin_unlock(&child->sighand->siglock); |
1da177e4 LT |
161 | } |
162 | ||
dbb5afad ON |
163 | static bool looks_like_a_spurious_pid(struct task_struct *task) |
164 | { | |
165 | if (task->exit_code != ((PTRACE_EVENT_EXEC << 8) | SIGTRAP)) | |
166 | return false; | |
167 | ||
168 | if (task_pid_vnr(task) == task->ptrace_message) | |
169 | return false; | |
170 | /* | |
171 | * The tracee changed its pid but the PTRACE_EVENT_EXEC event | |
172 | * was not wait()'ed, most probably debugger targets the old | |
173 | * leader which was destroyed in de_thread(). | |
174 | */ | |
175 | return true; | |
176 | } | |
177 | ||
31cae1ea PZ |
178 | /* |
179 | * Ensure that nothing can wake it up, even SIGKILL | |
180 | * | |
181 | * A task is switched to this state while a ptrace operation is in progress; | |
182 | * such that the ptrace operation is uninterruptible. | |
183 | */ | |
9899d11f ON |
184 | static bool ptrace_freeze_traced(struct task_struct *task) |
185 | { | |
186 | bool ret = false; | |
187 | ||
188 | /* Lockless, nobody but us can set this flag */ | |
189 | if (task->jobctl & JOBCTL_LISTENING) | |
190 | return ret; | |
191 | ||
192 | spin_lock_irq(&task->sighand->siglock); | |
dbb5afad ON |
193 | if (task_is_traced(task) && !looks_like_a_spurious_pid(task) && |
194 | !__fatal_signal_pending(task)) { | |
2500ad1c | 195 | task->jobctl |= JOBCTL_PTRACE_FROZEN; |
9899d11f ON |
196 | ret = true; |
197 | } | |
198 | spin_unlock_irq(&task->sighand->siglock); | |
199 | ||
200 | return ret; | |
201 | } | |
202 | ||
203 | static void ptrace_unfreeze_traced(struct task_struct *task) | |
204 | { | |
2500ad1c | 205 | unsigned long flags; |
9899d11f | 206 | |
5402e97a | 207 | /* |
2500ad1c EB |
208 | * The child may be awake and may have cleared |
209 | * JOBCTL_PTRACE_FROZEN (see ptrace_resume). The child will | |
210 | * not set JOBCTL_PTRACE_FROZEN or enter __TASK_TRACED anew. | |
5402e97a | 211 | */ |
2500ad1c EB |
212 | if (lock_task_sighand(task, &flags)) { |
213 | task->jobctl &= ~JOBCTL_PTRACE_FROZEN; | |
31cae1ea | 214 | if (__fatal_signal_pending(task)) { |
de2a3477 | 215 | task->jobctl &= ~JOBCTL_TRACED; |
5402e97a | 216 | wake_up_state(task, __TASK_TRACED); |
31cae1ea | 217 | } |
2500ad1c | 218 | unlock_task_sighand(task, &flags); |
5402e97a | 219 | } |
9899d11f ON |
220 | } |
221 | ||
755e276b TH |
222 | /** |
223 | * ptrace_check_attach - check whether ptracee is ready for ptrace operation | |
224 | * @child: ptracee to check for | |
225 | * @ignore_state: don't check whether @child is currently %TASK_TRACED | |
226 | * | |
227 | * Check whether @child is being ptraced by %current and ready for further | |
228 | * ptrace operations. If @ignore_state is %false, @child also should be in | |
229 | * %TASK_TRACED state and on return the child is guaranteed to be traced | |
230 | * and not executing. If @ignore_state is %true, @child can be in any | |
231 | * state. | |
232 | * | |
233 | * CONTEXT: | |
234 | * Grabs and releases tasklist_lock and @child->sighand->siglock. | |
235 | * | |
236 | * RETURNS: | |
237 | * 0 on success, -ESRCH if %child is not ready. | |
1da177e4 | 238 | */ |
edea0d03 | 239 | static int ptrace_check_attach(struct task_struct *child, bool ignore_state) |
1da177e4 LT |
240 | { |
241 | int ret = -ESRCH; | |
242 | ||
243 | /* | |
244 | * We take the read lock around doing both checks to close a | |
245 | * possible race where someone else was tracing our child and | |
246 | * detached between these two checks. After this locked check, | |
247 | * we are sure that this is our traced child and that can only | |
248 | * be changed by us so it's not changing right after this. | |
249 | */ | |
250 | read_lock(&tasklist_lock); | |
9899d11f | 251 | if (child->ptrace && child->parent == current) { |
c0c0b649 ON |
252 | /* |
253 | * child->sighand can't be NULL, release_task() | |
254 | * does ptrace_unlink() before __exit_signal(). | |
255 | */ | |
9899d11f | 256 | if (ignore_state || ptrace_freeze_traced(child)) |
321fb561 | 257 | ret = 0; |
1da177e4 LT |
258 | } |
259 | read_unlock(&tasklist_lock); | |
260 | ||
7b0fe136 | 261 | if (!ret && !ignore_state && |
f5d39b02 | 262 | WARN_ON_ONCE(!wait_task_inactive(child, __TASK_TRACED|TASK_FROZEN))) |
7b0fe136 | 263 | ret = -ESRCH; |
1da177e4 | 264 | |
1da177e4 LT |
265 | return ret; |
266 | } | |
267 | ||
cf237052 | 268 | static bool ptrace_has_cap(struct user_namespace *ns, unsigned int mode) |
69f594a3 EP |
269 | { |
270 | if (mode & PTRACE_MODE_NOAUDIT) | |
cf237052 MS |
271 | return ns_capable_noaudit(ns, CAP_SYS_PTRACE); |
272 | return ns_capable(ns, CAP_SYS_PTRACE); | |
69f594a3 EP |
273 | } |
274 | ||
9f99798f TH |
275 | /* Returns 0 on success, -errno on denial. */ |
276 | static int __ptrace_may_access(struct task_struct *task, unsigned int mode) | |
ab8d11be | 277 | { |
c69e8d9c | 278 | const struct cred *cred = current_cred(), *tcred; |
bfedb589 | 279 | struct mm_struct *mm; |
caaee623 JH |
280 | kuid_t caller_uid; |
281 | kgid_t caller_gid; | |
282 | ||
283 | if (!(mode & PTRACE_MODE_FSCREDS) == !(mode & PTRACE_MODE_REALCREDS)) { | |
284 | WARN(1, "denying ptrace access check without PTRACE_MODE_*CREDS\n"); | |
285 | return -EPERM; | |
286 | } | |
b6dff3ec | 287 | |
df26c40e EB |
288 | /* May we inspect the given task? |
289 | * This check is used both for attaching with ptrace | |
290 | * and for allowing access to sensitive information in /proc. | |
291 | * | |
292 | * ptrace_attach denies several cases that /proc allows | |
293 | * because setting up the necessary parent/child relationship | |
294 | * or halting the specified task is impossible. | |
295 | */ | |
caaee623 | 296 | |
df26c40e | 297 | /* Don't let security modules deny introspection */ |
73af963f | 298 | if (same_thread_group(task, current)) |
df26c40e | 299 | return 0; |
c69e8d9c | 300 | rcu_read_lock(); |
caaee623 JH |
301 | if (mode & PTRACE_MODE_FSCREDS) { |
302 | caller_uid = cred->fsuid; | |
303 | caller_gid = cred->fsgid; | |
304 | } else { | |
305 | /* | |
306 | * Using the euid would make more sense here, but something | |
307 | * in userland might rely on the old behavior, and this | |
308 | * shouldn't be a security problem since | |
309 | * PTRACE_MODE_REALCREDS implies that the caller explicitly | |
310 | * used a syscall that requests access to another process | |
311 | * (and not a filesystem syscall to procfs). | |
312 | */ | |
313 | caller_uid = cred->uid; | |
314 | caller_gid = cred->gid; | |
315 | } | |
c69e8d9c | 316 | tcred = __task_cred(task); |
caaee623 JH |
317 | if (uid_eq(caller_uid, tcred->euid) && |
318 | uid_eq(caller_uid, tcred->suid) && | |
319 | uid_eq(caller_uid, tcred->uid) && | |
320 | gid_eq(caller_gid, tcred->egid) && | |
321 | gid_eq(caller_gid, tcred->sgid) && | |
322 | gid_eq(caller_gid, tcred->gid)) | |
8409cca7 | 323 | goto ok; |
cf237052 | 324 | if (ptrace_has_cap(tcred->user_ns, mode)) |
8409cca7 SH |
325 | goto ok; |
326 | rcu_read_unlock(); | |
327 | return -EPERM; | |
328 | ok: | |
c69e8d9c | 329 | rcu_read_unlock(); |
f6581f5b JH |
330 | /* |
331 | * If a task drops privileges and becomes nondumpable (through a syscall | |
332 | * like setresuid()) while we are trying to access it, we must ensure | |
333 | * that the dumpability is read after the credentials; otherwise, | |
334 | * we may be able to attach to a task that we shouldn't be able to | |
335 | * attach to (as if the task had dropped privileges without becoming | |
336 | * nondumpable). | |
337 | * Pairs with a write barrier in commit_creds(). | |
338 | */ | |
339 | smp_rmb(); | |
bfedb589 EB |
340 | mm = task->mm; |
341 | if (mm && | |
342 | ((get_dumpable(mm) != SUID_DUMP_USER) && | |
cf237052 | 343 | !ptrace_has_cap(mm->user_ns, mode))) |
bfedb589 | 344 | return -EPERM; |
ab8d11be | 345 | |
9e48858f | 346 | return security_ptrace_access_check(task, mode); |
ab8d11be MS |
347 | } |
348 | ||
006ebb40 | 349 | bool ptrace_may_access(struct task_struct *task, unsigned int mode) |
ab8d11be MS |
350 | { |
351 | int err; | |
352 | task_lock(task); | |
006ebb40 | 353 | err = __ptrace_may_access(task, mode); |
ab8d11be | 354 | task_unlock(task); |
3a709703 | 355 | return !err; |
ab8d11be MS |
356 | } |
357 | ||
ee1fee90 JH |
358 | static int check_ptrace_options(unsigned long data) |
359 | { | |
360 | if (data & ~(unsigned long)PTRACE_O_MASK) | |
361 | return -EINVAL; | |
362 | ||
363 | if (unlikely(data & PTRACE_O_SUSPEND_SECCOMP)) { | |
364 | if (!IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) || | |
365 | !IS_ENABLED(CONFIG_SECCOMP)) | |
366 | return -EINVAL; | |
367 | ||
368 | if (!capable(CAP_SYS_ADMIN)) | |
369 | return -EPERM; | |
370 | ||
371 | if (seccomp_mode(¤t->seccomp) != SECCOMP_MODE_DISABLED || | |
372 | current->ptrace & PT_SUSPEND_SECCOMP) | |
373 | return -EPERM; | |
374 | } | |
375 | return 0; | |
376 | } | |
377 | ||
5431fdd2 PZ |
378 | static inline void ptrace_set_stopped(struct task_struct *task) |
379 | { | |
380 | guard(spinlock)(&task->sighand->siglock); | |
381 | ||
382 | /* | |
383 | * If the task is already STOPPED, set JOBCTL_TRAP_STOP and | |
384 | * TRAPPING, and kick it so that it transits to TRACED. TRAPPING | |
385 | * will be cleared if the child completes the transition or any | |
386 | * event which clears the group stop states happens. We'll wait | |
387 | * for the transition to complete before returning from this | |
388 | * function. | |
389 | * | |
390 | * This hides STOPPED -> RUNNING -> TRACED transition from the | |
391 | * attaching thread but a different thread in the same group can | |
392 | * still observe the transient RUNNING state. IOW, if another | |
393 | * thread's WNOHANG wait(2) on the stopped tracee races against | |
394 | * ATTACH, the wait(2) may fail due to the transient RUNNING. | |
395 | * | |
396 | * The following task_is_stopped() test is safe as both transitions | |
397 | * in and out of STOPPED are protected by siglock. | |
398 | */ | |
399 | if (task_is_stopped(task) && | |
400 | task_set_jobctl_pending(task, JOBCTL_TRAP_STOP | JOBCTL_TRAPPING)) { | |
401 | task->jobctl &= ~JOBCTL_STOPPED; | |
402 | signal_wake_up_state(task, __TASK_STOPPED); | |
403 | } | |
404 | } | |
405 | ||
3544d72a | 406 | static int ptrace_attach(struct task_struct *task, long request, |
aa9147c9 | 407 | unsigned long addr, |
3544d72a | 408 | unsigned long flags) |
1da177e4 | 409 | { |
3544d72a | 410 | bool seize = (request == PTRACE_SEIZE); |
1da177e4 | 411 | int retval; |
f5b40e36 | 412 | |
aa9147c9 DV |
413 | if (seize) { |
414 | if (addr != 0) | |
5431fdd2 | 415 | return -EIO; |
ee1fee90 JH |
416 | /* |
417 | * This duplicates the check in check_ptrace_options() because | |
418 | * ptrace_attach() and ptrace_setoptions() have historically | |
419 | * used different error codes for unknown ptrace options. | |
420 | */ | |
aa9147c9 | 421 | if (flags & ~(unsigned long)PTRACE_O_MASK) |
5431fdd2 PZ |
422 | return -EIO; |
423 | ||
ee1fee90 JH |
424 | retval = check_ptrace_options(flags); |
425 | if (retval) | |
426 | return retval; | |
aa9147c9 DV |
427 | flags = PT_PTRACED | PT_SEIZED | (flags << PT_OPT_FLAG_SHIFT); |
428 | } else { | |
429 | flags = PT_PTRACED; | |
430 | } | |
3544d72a | 431 | |
a5cb013d AV |
432 | audit_ptrace(task); |
433 | ||
e8b33b8c | 434 | if (unlikely(task->flags & PF_KTHREAD)) |
5431fdd2 | 435 | return -EPERM; |
bac0abd6 | 436 | if (same_thread_group(task, current)) |
5431fdd2 | 437 | return -EPERM; |
f5b40e36 | 438 | |
f2f0b00a ON |
439 | /* |
440 | * Protect exec's credential calculations against our interference; | |
86b6c1f3 | 441 | * SUID, SGID and LSM creds get determined differently |
5e751e99 | 442 | * under ptrace. |
d84f4f99 | 443 | */ |
5431fdd2 PZ |
444 | scoped_cond_guard (mutex_intr, return -ERESTARTNOINTR, |
445 | &task->signal->cred_guard_mutex) { | |
f5b40e36 | 446 | |
5431fdd2 PZ |
447 | scoped_guard (task_lock, task) { |
448 | retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH_REALCREDS); | |
449 | if (retval) | |
450 | return retval; | |
451 | } | |
1da177e4 | 452 | |
5431fdd2 PZ |
453 | scoped_guard (write_lock_irq, &tasklist_lock) { |
454 | if (unlikely(task->exit_state)) | |
455 | return -EPERM; | |
456 | if (task->ptrace) | |
457 | return -EPERM; | |
b79b7ba9 | 458 | |
5431fdd2 | 459 | task->ptrace = flags; |
1da177e4 | 460 | |
5431fdd2 | 461 | ptrace_link(task, current); |
3544d72a | 462 | |
5431fdd2 PZ |
463 | /* SEIZE doesn't trap tracee on attach */ |
464 | if (!seize) | |
465 | send_sig_info(SIGSTOP, SEND_SIG_PRIV, task); | |
b79b7ba9 | 466 | |
5431fdd2 PZ |
467 | ptrace_set_stopped(task); |
468 | } | |
469 | } | |
d79fdd6d TH |
470 | |
471 | /* | |
5431fdd2 PZ |
472 | * We do not bother to change retval or clear JOBCTL_TRAPPING |
473 | * if wait_on_bit() was interrupted by SIGKILL. The tracer will | |
474 | * not return to user-mode, it will exit and clear this bit in | |
475 | * __ptrace_unlink() if it wasn't already cleared by the tracee; | |
476 | * and until then nobody can ptrace this task. | |
d79fdd6d | 477 | */ |
5431fdd2 PZ |
478 | wait_on_bit(&task->jobctl, JOBCTL_TRAPPING_BIT, TASK_KILLABLE); |
479 | proc_ptrace_connector(task, PTRACE_ATTACH); | |
d79fdd6d | 480 | |
5431fdd2 | 481 | return 0; |
1da177e4 LT |
482 | } |
483 | ||
f2f0b00a ON |
484 | /** |
485 | * ptrace_traceme -- helper for PTRACE_TRACEME | |
486 | * | |
487 | * Performs checks and sets PT_PTRACED. | |
488 | * Should be used by all ptrace implementations for PTRACE_TRACEME. | |
489 | */ | |
e3e89cc5 | 490 | static int ptrace_traceme(void) |
f2f0b00a ON |
491 | { |
492 | int ret = -EPERM; | |
493 | ||
4b105cbb ON |
494 | write_lock_irq(&tasklist_lock); |
495 | /* Are we already being traced? */ | |
f2f0b00a | 496 | if (!current->ptrace) { |
f2f0b00a | 497 | ret = security_ptrace_traceme(current->parent); |
f2f0b00a ON |
498 | /* |
499 | * Check PF_EXITING to ensure ->real_parent has not passed | |
500 | * exit_ptrace(). Otherwise we don't report the error but | |
501 | * pretend ->real_parent untraces us right after return. | |
502 | */ | |
503 | if (!ret && !(current->real_parent->flags & PF_EXITING)) { | |
504 | current->ptrace = PT_PTRACED; | |
c70d9d80 | 505 | ptrace_link(current, current->real_parent); |
f2f0b00a | 506 | } |
f2f0b00a | 507 | } |
4b105cbb ON |
508 | write_unlock_irq(&tasklist_lock); |
509 | ||
f2f0b00a ON |
510 | return ret; |
511 | } | |
512 | ||
39c626ae ON |
513 | /* |
514 | * Called with irqs disabled, returns true if childs should reap themselves. | |
515 | */ | |
516 | static int ignoring_children(struct sighand_struct *sigh) | |
517 | { | |
518 | int ret; | |
519 | spin_lock(&sigh->siglock); | |
520 | ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) || | |
521 | (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT); | |
522 | spin_unlock(&sigh->siglock); | |
523 | return ret; | |
524 | } | |
525 | ||
526 | /* | |
527 | * Called with tasklist_lock held for writing. | |
528 | * Unlink a traced task, and clean it up if it was a traced zombie. | |
529 | * Return true if it needs to be reaped with release_task(). | |
530 | * (We can't call release_task() here because we already hold tasklist_lock.) | |
531 | * | |
532 | * If it's a zombie, our attachedness prevented normal parent notification | |
533 | * or self-reaping. Do notification now if it would have happened earlier. | |
534 | * If it should reap itself, return true. | |
535 | * | |
a7f0765e ON |
536 | * If it's our own child, there is no notification to do. But if our normal |
537 | * children self-reap, then this child was prevented by ptrace and we must | |
538 | * reap it now, in that case we must also wake up sub-threads sleeping in | |
539 | * do_wait(). | |
39c626ae ON |
540 | */ |
541 | static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p) | |
542 | { | |
9843a1e9 ON |
543 | bool dead; |
544 | ||
39c626ae ON |
545 | __ptrace_unlink(p); |
546 | ||
9843a1e9 ON |
547 | if (p->exit_state != EXIT_ZOMBIE) |
548 | return false; | |
549 | ||
550 | dead = !thread_group_leader(p); | |
551 | ||
552 | if (!dead && thread_group_empty(p)) { | |
553 | if (!same_thread_group(p->real_parent, tracer)) | |
554 | dead = do_notify_parent(p, p->exit_signal); | |
555 | else if (ignoring_children(tracer->sighand)) { | |
556 | __wake_up_parent(p, tracer); | |
9843a1e9 | 557 | dead = true; |
39c626ae ON |
558 | } |
559 | } | |
9843a1e9 ON |
560 | /* Mark it as in the process of being reaped. */ |
561 | if (dead) | |
562 | p->exit_state = EXIT_DEAD; | |
563 | return dead; | |
39c626ae ON |
564 | } |
565 | ||
e3e89cc5 | 566 | static int ptrace_detach(struct task_struct *child, unsigned int data) |
1da177e4 | 567 | { |
7ed20e1a | 568 | if (!valid_signal(data)) |
5ecfbae0 | 569 | return -EIO; |
1da177e4 LT |
570 | |
571 | /* Architecture-specific hardware disable .. */ | |
572 | ptrace_disable(child); | |
573 | ||
95c3eb76 | 574 | write_lock_irq(&tasklist_lock); |
39c626ae | 575 | /* |
64a4096c ON |
576 | * We rely on ptrace_freeze_traced(). It can't be killed and |
577 | * untraced by another thread, it can't be a zombie. | |
39c626ae | 578 | */ |
64a4096c ON |
579 | WARN_ON(!child->ptrace || child->exit_state); |
580 | /* | |
581 | * tasklist_lock avoids the race with wait_task_stopped(), see | |
582 | * the comment in ptrace_resume(). | |
583 | */ | |
584 | child->exit_code = data; | |
585 | __ptrace_detach(current, child); | |
1da177e4 LT |
586 | write_unlock_irq(&tasklist_lock); |
587 | ||
f701e5b7 | 588 | proc_ptrace_connector(child, PTRACE_DETACH); |
4576145c | 589 | |
1da177e4 LT |
590 | return 0; |
591 | } | |
592 | ||
39c626ae | 593 | /* |
c7e49c14 | 594 | * Detach all tasks we were using ptrace on. Called with tasklist held |
7c8bd232 | 595 | * for writing. |
39c626ae | 596 | */ |
7c8bd232 | 597 | void exit_ptrace(struct task_struct *tracer, struct list_head *dead) |
39c626ae ON |
598 | { |
599 | struct task_struct *p, *n; | |
c7e49c14 | 600 | |
39c626ae | 601 | list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) { |
992fb6e1 | 602 | if (unlikely(p->ptrace & PT_EXITKILL)) |
079b22dc | 603 | send_sig_info(SIGKILL, SEND_SIG_PRIV, p); |
992fb6e1 | 604 | |
39c626ae | 605 | if (__ptrace_detach(tracer, p)) |
7c8bd232 | 606 | list_add(&p->ptrace_entry, dead); |
39c626ae ON |
607 | } |
608 | } | |
609 | ||
1da177e4 LT |
610 | int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len) |
611 | { | |
612 | int copied = 0; | |
613 | ||
614 | while (len > 0) { | |
615 | char buf[128]; | |
616 | int this_len, retval; | |
617 | ||
618 | this_len = (len > sizeof(buf)) ? sizeof(buf) : len; | |
84d77d3f EB |
619 | retval = ptrace_access_vm(tsk, src, buf, this_len, FOLL_FORCE); |
620 | ||
1da177e4 LT |
621 | if (!retval) { |
622 | if (copied) | |
623 | break; | |
624 | return -EIO; | |
625 | } | |
626 | if (copy_to_user(dst, buf, retval)) | |
627 | return -EFAULT; | |
628 | copied += retval; | |
629 | src += retval; | |
630 | dst += retval; | |
3a709703 | 631 | len -= retval; |
1da177e4 LT |
632 | } |
633 | return copied; | |
634 | } | |
635 | ||
636 | int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len) | |
637 | { | |
638 | int copied = 0; | |
639 | ||
640 | while (len > 0) { | |
641 | char buf[128]; | |
642 | int this_len, retval; | |
643 | ||
644 | this_len = (len > sizeof(buf)) ? sizeof(buf) : len; | |
645 | if (copy_from_user(buf, src, this_len)) | |
646 | return -EFAULT; | |
84d77d3f | 647 | retval = ptrace_access_vm(tsk, dst, buf, this_len, |
f307ab6d | 648 | FOLL_FORCE | FOLL_WRITE); |
1da177e4 LT |
649 | if (!retval) { |
650 | if (copied) | |
651 | break; | |
652 | return -EIO; | |
653 | } | |
654 | copied += retval; | |
655 | src += retval; | |
656 | dst += retval; | |
3a709703 | 657 | len -= retval; |
1da177e4 LT |
658 | } |
659 | return copied; | |
660 | } | |
661 | ||
4abf9869 | 662 | static int ptrace_setoptions(struct task_struct *child, unsigned long data) |
1da177e4 | 663 | { |
86b6c1f3 | 664 | unsigned flags; |
ee1fee90 | 665 | int ret; |
86b6c1f3 | 666 | |
ee1fee90 JH |
667 | ret = check_ptrace_options(data); |
668 | if (ret) | |
669 | return ret; | |
13c4a901 | 670 | |
86b6c1f3 DV |
671 | /* Avoid intermediate state when all opts are cleared */ |
672 | flags = child->ptrace; | |
673 | flags &= ~(PTRACE_O_MASK << PT_OPT_FLAG_SHIFT); | |
674 | flags |= (data << PT_OPT_FLAG_SHIFT); | |
675 | child->ptrace = flags; | |
1da177e4 | 676 | |
8c5cf9e5 | 677 | return 0; |
1da177e4 LT |
678 | } |
679 | ||
ae7795bc | 680 | static int ptrace_getsiginfo(struct task_struct *child, kernel_siginfo_t *info) |
1da177e4 | 681 | { |
e4961254 | 682 | unsigned long flags; |
1da177e4 LT |
683 | int error = -ESRCH; |
684 | ||
e4961254 | 685 | if (lock_task_sighand(child, &flags)) { |
1da177e4 | 686 | error = -EINVAL; |
1da177e4 | 687 | if (likely(child->last_siginfo != NULL)) { |
0752d7bf | 688 | copy_siginfo(info, child->last_siginfo); |
1da177e4 LT |
689 | error = 0; |
690 | } | |
e4961254 | 691 | unlock_task_sighand(child, &flags); |
1da177e4 | 692 | } |
1da177e4 LT |
693 | return error; |
694 | } | |
695 | ||
ae7795bc | 696 | static int ptrace_setsiginfo(struct task_struct *child, const kernel_siginfo_t *info) |
1da177e4 | 697 | { |
e4961254 | 698 | unsigned long flags; |
1da177e4 LT |
699 | int error = -ESRCH; |
700 | ||
e4961254 | 701 | if (lock_task_sighand(child, &flags)) { |
1da177e4 | 702 | error = -EINVAL; |
1da177e4 | 703 | if (likely(child->last_siginfo != NULL)) { |
0752d7bf | 704 | copy_siginfo(child->last_siginfo, info); |
1da177e4 LT |
705 | error = 0; |
706 | } | |
e4961254 | 707 | unlock_task_sighand(child, &flags); |
1da177e4 | 708 | } |
1da177e4 LT |
709 | return error; |
710 | } | |
711 | ||
84c751bd AV |
712 | static int ptrace_peek_siginfo(struct task_struct *child, |
713 | unsigned long addr, | |
714 | unsigned long data) | |
715 | { | |
716 | struct ptrace_peeksiginfo_args arg; | |
717 | struct sigpending *pending; | |
718 | struct sigqueue *q; | |
719 | int ret, i; | |
720 | ||
721 | ret = copy_from_user(&arg, (void __user *) addr, | |
722 | sizeof(struct ptrace_peeksiginfo_args)); | |
723 | if (ret) | |
724 | return -EFAULT; | |
725 | ||
726 | if (arg.flags & ~PTRACE_PEEKSIGINFO_SHARED) | |
727 | return -EINVAL; /* unknown flags */ | |
728 | ||
729 | if (arg.nr < 0) | |
730 | return -EINVAL; | |
731 | ||
f6e2aa91 EB |
732 | /* Ensure arg.off fits in an unsigned long */ |
733 | if (arg.off > ULONG_MAX) | |
734 | return 0; | |
735 | ||
84c751bd AV |
736 | if (arg.flags & PTRACE_PEEKSIGINFO_SHARED) |
737 | pending = &child->signal->shared_pending; | |
738 | else | |
739 | pending = &child->pending; | |
740 | ||
741 | for (i = 0; i < arg.nr; ) { | |
ae7795bc | 742 | kernel_siginfo_t info; |
f6e2aa91 EB |
743 | unsigned long off = arg.off + i; |
744 | bool found = false; | |
84c751bd AV |
745 | |
746 | spin_lock_irq(&child->sighand->siglock); | |
747 | list_for_each_entry(q, &pending->list, list) { | |
748 | if (!off--) { | |
f6e2aa91 | 749 | found = true; |
84c751bd AV |
750 | copy_siginfo(&info, &q->info); |
751 | break; | |
752 | } | |
753 | } | |
754 | spin_unlock_irq(&child->sighand->siglock); | |
755 | ||
f6e2aa91 | 756 | if (!found) /* beyond the end of the list */ |
84c751bd AV |
757 | break; |
758 | ||
759 | #ifdef CONFIG_COMPAT | |
5c465217 | 760 | if (unlikely(in_compat_syscall())) { |
84c751bd AV |
761 | compat_siginfo_t __user *uinfo = compat_ptr(data); |
762 | ||
cc731525 | 763 | if (copy_siginfo_to_user32(uinfo, &info)) { |
706b23bd MD |
764 | ret = -EFAULT; |
765 | break; | |
766 | } | |
767 | ||
84c751bd AV |
768 | } else |
769 | #endif | |
770 | { | |
771 | siginfo_t __user *uinfo = (siginfo_t __user *) data; | |
772 | ||
cc731525 | 773 | if (copy_siginfo_to_user(uinfo, &info)) { |
706b23bd MD |
774 | ret = -EFAULT; |
775 | break; | |
776 | } | |
84c751bd AV |
777 | } |
778 | ||
779 | data += sizeof(siginfo_t); | |
780 | i++; | |
781 | ||
782 | if (signal_pending(current)) | |
783 | break; | |
784 | ||
785 | cond_resched(); | |
786 | } | |
787 | ||
788 | if (i > 0) | |
789 | return i; | |
790 | ||
791 | return ret; | |
792 | } | |
36df29d7 | 793 | |
90f093fa PF |
794 | #ifdef CONFIG_RSEQ |
795 | static long ptrace_get_rseq_configuration(struct task_struct *task, | |
796 | unsigned long size, void __user *data) | |
797 | { | |
798 | struct ptrace_rseq_configuration conf = { | |
799 | .rseq_abi_pointer = (u64)(uintptr_t)task->rseq, | |
ee3e3ac0 | 800 | .rseq_abi_size = task->rseq_len, |
90f093fa PF |
801 | .signature = task->rseq_sig, |
802 | .flags = 0, | |
803 | }; | |
804 | ||
805 | size = min_t(unsigned long, size, sizeof(conf)); | |
806 | if (copy_to_user(data, &conf, size)) | |
807 | return -EFAULT; | |
808 | return sizeof(conf); | |
809 | } | |
810 | #endif | |
811 | ||
36df29d7 | 812 | #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP) |
36df29d7 | 813 | |
5b88abbf RM |
814 | #ifdef PTRACE_SINGLEBLOCK |
815 | #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK) | |
816 | #else | |
817 | #define is_singleblock(request) 0 | |
818 | #endif | |
819 | ||
36df29d7 RM |
820 | #ifdef PTRACE_SYSEMU |
821 | #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP) | |
822 | #else | |
823 | #define is_sysemu_singlestep(request) 0 | |
824 | #endif | |
825 | ||
4abf9869 NK |
826 | static int ptrace_resume(struct task_struct *child, long request, |
827 | unsigned long data) | |
36df29d7 RM |
828 | { |
829 | if (!valid_signal(data)) | |
830 | return -EIO; | |
831 | ||
832 | if (request == PTRACE_SYSCALL) | |
64c19ba2 | 833 | set_task_syscall_work(child, SYSCALL_TRACE); |
36df29d7 | 834 | else |
64c19ba2 | 835 | clear_task_syscall_work(child, SYSCALL_TRACE); |
36df29d7 | 836 | |
64eb35f7 | 837 | #if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU) |
36df29d7 | 838 | if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP) |
64eb35f7 | 839 | set_task_syscall_work(child, SYSCALL_EMU); |
36df29d7 | 840 | else |
64eb35f7 | 841 | clear_task_syscall_work(child, SYSCALL_EMU); |
36df29d7 RM |
842 | #endif |
843 | ||
5b88abbf RM |
844 | if (is_singleblock(request)) { |
845 | if (unlikely(!arch_has_block_step())) | |
846 | return -EIO; | |
847 | user_enable_block_step(child); | |
848 | } else if (is_singlestep(request) || is_sysemu_singlestep(request)) { | |
36df29d7 RM |
849 | if (unlikely(!arch_has_single_step())) |
850 | return -EIO; | |
851 | user_enable_single_step(child); | |
3a709703 | 852 | } else { |
36df29d7 | 853 | user_disable_single_step(child); |
3a709703 | 854 | } |
36df29d7 | 855 | |
b72c1869 ON |
856 | /* |
857 | * Change ->exit_code and ->state under siglock to avoid the race | |
858 | * with wait_task_stopped() in between; a non-zero ->exit_code will | |
859 | * wrongly look like another report from tracee. | |
860 | * | |
861 | * Note that we need siglock even if ->exit_code == data and/or this | |
862 | * status was not reported yet, the new status must not be cleared by | |
863 | * wait_task_stopped() after resume. | |
b72c1869 | 864 | */ |
5b4197cb | 865 | spin_lock_irq(&child->sighand->siglock); |
36df29d7 | 866 | child->exit_code = data; |
31cae1ea | 867 | child->jobctl &= ~JOBCTL_TRACED; |
0666fb51 | 868 | wake_up_state(child, __TASK_TRACED); |
5b4197cb | 869 | spin_unlock_irq(&child->sighand->siglock); |
36df29d7 RM |
870 | |
871 | return 0; | |
872 | } | |
873 | ||
2225a122 SS |
874 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
875 | ||
876 | static const struct user_regset * | |
877 | find_regset(const struct user_regset_view *view, unsigned int type) | |
878 | { | |
879 | const struct user_regset *regset; | |
880 | int n; | |
881 | ||
882 | for (n = 0; n < view->n; ++n) { | |
883 | regset = view->regsets + n; | |
884 | if (regset->core_note_type == type) | |
885 | return regset; | |
886 | } | |
887 | ||
888 | return NULL; | |
889 | } | |
890 | ||
891 | static int ptrace_regset(struct task_struct *task, int req, unsigned int type, | |
892 | struct iovec *kiov) | |
893 | { | |
894 | const struct user_regset_view *view = task_user_regset_view(task); | |
895 | const struct user_regset *regset = find_regset(view, type); | |
896 | int regset_no; | |
897 | ||
898 | if (!regset || (kiov->iov_len % regset->size) != 0) | |
c6a0dd7e | 899 | return -EINVAL; |
2225a122 SS |
900 | |
901 | regset_no = regset - view->regsets; | |
902 | kiov->iov_len = min(kiov->iov_len, | |
903 | (__kernel_size_t) (regset->n * regset->size)); | |
904 | ||
905 | if (req == PTRACE_GETREGSET) | |
906 | return copy_regset_to_user(task, view, regset_no, 0, | |
907 | kiov->iov_len, kiov->iov_base); | |
908 | else | |
909 | return copy_regset_from_user(task, view, regset_no, 0, | |
910 | kiov->iov_len, kiov->iov_base); | |
911 | } | |
912 | ||
e8440c14 JS |
913 | /* |
914 | * This is declared in linux/regset.h and defined in machine-dependent | |
915 | * code. We put the export here, near the primary machine-neutral use, | |
916 | * to ensure no machine forgets it. | |
917 | */ | |
918 | EXPORT_SYMBOL_GPL(task_user_regset_view); | |
201766a2 EK |
919 | |
920 | static unsigned long | |
921 | ptrace_get_syscall_info_entry(struct task_struct *child, struct pt_regs *regs, | |
922 | struct ptrace_syscall_info *info) | |
923 | { | |
924 | unsigned long args[ARRAY_SIZE(info->entry.args)]; | |
925 | int i; | |
926 | ||
927 | info->op = PTRACE_SYSCALL_INFO_ENTRY; | |
928 | info->entry.nr = syscall_get_nr(child, regs); | |
929 | syscall_get_arguments(child, regs, args); | |
930 | for (i = 0; i < ARRAY_SIZE(args); i++) | |
931 | info->entry.args[i] = args[i]; | |
932 | ||
933 | /* args is the last field in struct ptrace_syscall_info.entry */ | |
934 | return offsetofend(struct ptrace_syscall_info, entry.args); | |
935 | } | |
936 | ||
937 | static unsigned long | |
938 | ptrace_get_syscall_info_seccomp(struct task_struct *child, struct pt_regs *regs, | |
939 | struct ptrace_syscall_info *info) | |
940 | { | |
941 | /* | |
942 | * As struct ptrace_syscall_info.entry is currently a subset | |
943 | * of struct ptrace_syscall_info.seccomp, it makes sense to | |
944 | * initialize that subset using ptrace_get_syscall_info_entry(). | |
945 | * This can be reconsidered in the future if these structures | |
946 | * diverge significantly enough. | |
947 | */ | |
948 | ptrace_get_syscall_info_entry(child, regs, info); | |
949 | info->op = PTRACE_SYSCALL_INFO_SECCOMP; | |
950 | info->seccomp.ret_data = child->ptrace_message; | |
951 | ||
952 | /* ret_data is the last field in struct ptrace_syscall_info.seccomp */ | |
953 | return offsetofend(struct ptrace_syscall_info, seccomp.ret_data); | |
954 | } | |
955 | ||
956 | static unsigned long | |
957 | ptrace_get_syscall_info_exit(struct task_struct *child, struct pt_regs *regs, | |
958 | struct ptrace_syscall_info *info) | |
959 | { | |
960 | info->op = PTRACE_SYSCALL_INFO_EXIT; | |
961 | info->exit.rval = syscall_get_error(child, regs); | |
962 | info->exit.is_error = !!info->exit.rval; | |
963 | if (!info->exit.is_error) | |
964 | info->exit.rval = syscall_get_return_value(child, regs); | |
965 | ||
966 | /* is_error is the last field in struct ptrace_syscall_info.exit */ | |
967 | return offsetofend(struct ptrace_syscall_info, exit.is_error); | |
968 | } | |
969 | ||
970 | static int | |
971 | ptrace_get_syscall_info(struct task_struct *child, unsigned long user_size, | |
972 | void __user *datavp) | |
973 | { | |
974 | struct pt_regs *regs = task_pt_regs(child); | |
975 | struct ptrace_syscall_info info = { | |
976 | .op = PTRACE_SYSCALL_INFO_NONE, | |
977 | .arch = syscall_get_arch(child), | |
978 | .instruction_pointer = instruction_pointer(regs), | |
979 | .stack_pointer = user_stack_pointer(regs), | |
980 | }; | |
981 | unsigned long actual_size = offsetof(struct ptrace_syscall_info, entry); | |
982 | unsigned long write_size; | |
983 | ||
984 | /* | |
985 | * This does not need lock_task_sighand() to access | |
986 | * child->last_siginfo because ptrace_freeze_traced() | |
987 | * called earlier by ptrace_check_attach() ensures that | |
988 | * the tracee cannot go away and clear its last_siginfo. | |
989 | */ | |
990 | switch (child->last_siginfo ? child->last_siginfo->si_code : 0) { | |
991 | case SIGTRAP | 0x80: | |
992 | switch (child->ptrace_message) { | |
993 | case PTRACE_EVENTMSG_SYSCALL_ENTRY: | |
994 | actual_size = ptrace_get_syscall_info_entry(child, regs, | |
995 | &info); | |
996 | break; | |
997 | case PTRACE_EVENTMSG_SYSCALL_EXIT: | |
998 | actual_size = ptrace_get_syscall_info_exit(child, regs, | |
999 | &info); | |
1000 | break; | |
1001 | } | |
1002 | break; | |
1003 | case SIGTRAP | (PTRACE_EVENT_SECCOMP << 8): | |
1004 | actual_size = ptrace_get_syscall_info_seccomp(child, regs, | |
1005 | &info); | |
1006 | break; | |
1007 | } | |
1008 | ||
1009 | write_size = min(actual_size, user_size); | |
1010 | return copy_to_user(datavp, &info, write_size) ? -EFAULT : actual_size; | |
1011 | } | |
1012 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
2225a122 | 1013 | |
1da177e4 | 1014 | int ptrace_request(struct task_struct *child, long request, |
4abf9869 | 1015 | unsigned long addr, unsigned long data) |
1da177e4 | 1016 | { |
fca26f26 | 1017 | bool seized = child->ptrace & PT_SEIZED; |
1da177e4 | 1018 | int ret = -EIO; |
ae7795bc | 1019 | kernel_siginfo_t siginfo, *si; |
9fed81dc NK |
1020 | void __user *datavp = (void __user *) data; |
1021 | unsigned long __user *datalp = datavp; | |
fca26f26 | 1022 | unsigned long flags; |
1da177e4 LT |
1023 | |
1024 | switch (request) { | |
16c3e389 RM |
1025 | case PTRACE_PEEKTEXT: |
1026 | case PTRACE_PEEKDATA: | |
1027 | return generic_ptrace_peekdata(child, addr, data); | |
1028 | case PTRACE_POKETEXT: | |
1029 | case PTRACE_POKEDATA: | |
1030 | return generic_ptrace_pokedata(child, addr, data); | |
1031 | ||
1da177e4 LT |
1032 | #ifdef PTRACE_OLDSETOPTIONS |
1033 | case PTRACE_OLDSETOPTIONS: | |
1034 | #endif | |
1035 | case PTRACE_SETOPTIONS: | |
1036 | ret = ptrace_setoptions(child, data); | |
1037 | break; | |
1038 | case PTRACE_GETEVENTMSG: | |
9fed81dc | 1039 | ret = put_user(child->ptrace_message, datalp); |
1da177e4 | 1040 | break; |
e16b2781 | 1041 | |
84c751bd AV |
1042 | case PTRACE_PEEKSIGINFO: |
1043 | ret = ptrace_peek_siginfo(child, addr, data); | |
1044 | break; | |
1045 | ||
1da177e4 | 1046 | case PTRACE_GETSIGINFO: |
e16b2781 RM |
1047 | ret = ptrace_getsiginfo(child, &siginfo); |
1048 | if (!ret) | |
9fed81dc | 1049 | ret = copy_siginfo_to_user(datavp, &siginfo); |
1da177e4 | 1050 | break; |
e16b2781 | 1051 | |
1da177e4 | 1052 | case PTRACE_SETSIGINFO: |
4cd2e0e7 EB |
1053 | ret = copy_siginfo_from_user(&siginfo, datavp); |
1054 | if (!ret) | |
e16b2781 | 1055 | ret = ptrace_setsiginfo(child, &siginfo); |
1da177e4 | 1056 | break; |
e16b2781 | 1057 | |
fcfc2aa0 AV |
1058 | case PTRACE_GETSIGMASK: { |
1059 | sigset_t *mask; | |
1060 | ||
29000cae AV |
1061 | if (addr != sizeof(sigset_t)) { |
1062 | ret = -EINVAL; | |
1063 | break; | |
1064 | } | |
1065 | ||
fcfc2aa0 AV |
1066 | if (test_tsk_restore_sigmask(child)) |
1067 | mask = &child->saved_sigmask; | |
1068 | else | |
1069 | mask = &child->blocked; | |
1070 | ||
1071 | if (copy_to_user(datavp, mask, sizeof(sigset_t))) | |
29000cae AV |
1072 | ret = -EFAULT; |
1073 | else | |
1074 | ret = 0; | |
1075 | ||
1076 | break; | |
fcfc2aa0 | 1077 | } |
29000cae AV |
1078 | |
1079 | case PTRACE_SETSIGMASK: { | |
1080 | sigset_t new_set; | |
1081 | ||
1082 | if (addr != sizeof(sigset_t)) { | |
1083 | ret = -EINVAL; | |
1084 | break; | |
1085 | } | |
1086 | ||
1087 | if (copy_from_user(&new_set, datavp, sizeof(sigset_t))) { | |
1088 | ret = -EFAULT; | |
1089 | break; | |
1090 | } | |
1091 | ||
1092 | sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1093 | ||
1094 | /* | |
1095 | * Every thread does recalc_sigpending() after resume, so | |
1096 | * retarget_shared_pending() and recalc_sigpending() are not | |
1097 | * called here. | |
1098 | */ | |
1099 | spin_lock_irq(&child->sighand->siglock); | |
1100 | child->blocked = new_set; | |
1101 | spin_unlock_irq(&child->sighand->siglock); | |
1102 | ||
fcfc2aa0 AV |
1103 | clear_tsk_restore_sigmask(child); |
1104 | ||
29000cae AV |
1105 | ret = 0; |
1106 | break; | |
1107 | } | |
1108 | ||
fca26f26 TH |
1109 | case PTRACE_INTERRUPT: |
1110 | /* | |
1111 | * Stop tracee without any side-effect on signal or job | |
1112 | * control. At least one trap is guaranteed to happen | |
1113 | * after this request. If @child is already trapped, the | |
1114 | * current trap is not disturbed and another trap will | |
1115 | * happen after the current trap is ended with PTRACE_CONT. | |
1116 | * | |
1117 | * The actual trap might not be PTRACE_EVENT_STOP trap but | |
1118 | * the pending condition is cleared regardless. | |
1119 | */ | |
1120 | if (unlikely(!seized || !lock_task_sighand(child, &flags))) | |
1121 | break; | |
1122 | ||
544b2c91 TH |
1123 | /* |
1124 | * INTERRUPT doesn't disturb existing trap sans one | |
1125 | * exception. If ptracer issued LISTEN for the current | |
1126 | * STOP, this INTERRUPT should clear LISTEN and re-trap | |
1127 | * tracee into STOP. | |
1128 | */ | |
fca26f26 | 1129 | if (likely(task_set_jobctl_pending(child, JOBCTL_TRAP_STOP))) |
910ffdb1 | 1130 | ptrace_signal_wake_up(child, child->jobctl & JOBCTL_LISTENING); |
544b2c91 TH |
1131 | |
1132 | unlock_task_sighand(child, &flags); | |
1133 | ret = 0; | |
1134 | break; | |
1135 | ||
1136 | case PTRACE_LISTEN: | |
1137 | /* | |
1138 | * Listen for events. Tracee must be in STOP. It's not | |
1139 | * resumed per-se but is not considered to be in TRACED by | |
1140 | * wait(2) or ptrace(2). If an async event (e.g. group | |
1141 | * stop state change) happens, tracee will enter STOP trap | |
1142 | * again. Alternatively, ptracer can issue INTERRUPT to | |
1143 | * finish listening and re-trap tracee into STOP. | |
1144 | */ | |
1145 | if (unlikely(!seized || !lock_task_sighand(child, &flags))) | |
1146 | break; | |
1147 | ||
1148 | si = child->last_siginfo; | |
f9d81f61 ON |
1149 | if (likely(si && (si->si_code >> 8) == PTRACE_EVENT_STOP)) { |
1150 | child->jobctl |= JOBCTL_LISTENING; | |
1151 | /* | |
1152 | * If NOTIFY is set, it means event happened between | |
1153 | * start of this trap and now. Trigger re-trap. | |
1154 | */ | |
1155 | if (child->jobctl & JOBCTL_TRAP_NOTIFY) | |
910ffdb1 | 1156 | ptrace_signal_wake_up(child, true); |
f9d81f61 ON |
1157 | ret = 0; |
1158 | } | |
fca26f26 | 1159 | unlock_task_sighand(child, &flags); |
fca26f26 TH |
1160 | break; |
1161 | ||
1bcf5482 AD |
1162 | case PTRACE_DETACH: /* detach a process that was attached. */ |
1163 | ret = ptrace_detach(child, data); | |
1164 | break; | |
36df29d7 | 1165 | |
9c1a1259 MF |
1166 | #ifdef CONFIG_BINFMT_ELF_FDPIC |
1167 | case PTRACE_GETFDPIC: { | |
e0129ef9 | 1168 | struct mm_struct *mm = get_task_mm(child); |
9c1a1259 MF |
1169 | unsigned long tmp = 0; |
1170 | ||
e0129ef9 ON |
1171 | ret = -ESRCH; |
1172 | if (!mm) | |
1173 | break; | |
1174 | ||
9c1a1259 MF |
1175 | switch (addr) { |
1176 | case PTRACE_GETFDPIC_EXEC: | |
e0129ef9 | 1177 | tmp = mm->context.exec_fdpic_loadmap; |
9c1a1259 MF |
1178 | break; |
1179 | case PTRACE_GETFDPIC_INTERP: | |
e0129ef9 | 1180 | tmp = mm->context.interp_fdpic_loadmap; |
9c1a1259 MF |
1181 | break; |
1182 | default: | |
1183 | break; | |
1184 | } | |
e0129ef9 | 1185 | mmput(mm); |
9c1a1259 | 1186 | |
9fed81dc | 1187 | ret = put_user(tmp, datalp); |
9c1a1259 MF |
1188 | break; |
1189 | } | |
1190 | #endif | |
1191 | ||
36df29d7 | 1192 | case PTRACE_SINGLESTEP: |
5b88abbf RM |
1193 | #ifdef PTRACE_SINGLEBLOCK |
1194 | case PTRACE_SINGLEBLOCK: | |
1195 | #endif | |
36df29d7 RM |
1196 | #ifdef PTRACE_SYSEMU |
1197 | case PTRACE_SYSEMU: | |
1198 | case PTRACE_SYSEMU_SINGLESTEP: | |
1199 | #endif | |
1200 | case PTRACE_SYSCALL: | |
1201 | case PTRACE_CONT: | |
1202 | return ptrace_resume(child, request, data); | |
1203 | ||
1204 | case PTRACE_KILL: | |
6a2d90ba EB |
1205 | send_sig_info(SIGKILL, SEND_SIG_NOINFO, child); |
1206 | return 0; | |
36df29d7 | 1207 | |
2225a122 SS |
1208 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
1209 | case PTRACE_GETREGSET: | |
29000cae | 1210 | case PTRACE_SETREGSET: { |
2225a122 | 1211 | struct iovec kiov; |
9fed81dc | 1212 | struct iovec __user *uiov = datavp; |
2225a122 | 1213 | |
96d4f267 | 1214 | if (!access_ok(uiov, sizeof(*uiov))) |
2225a122 SS |
1215 | return -EFAULT; |
1216 | ||
1217 | if (__get_user(kiov.iov_base, &uiov->iov_base) || | |
1218 | __get_user(kiov.iov_len, &uiov->iov_len)) | |
1219 | return -EFAULT; | |
1220 | ||
1221 | ret = ptrace_regset(child, request, addr, &kiov); | |
1222 | if (!ret) | |
1223 | ret = __put_user(kiov.iov_len, &uiov->iov_len); | |
1224 | break; | |
1225 | } | |
201766a2 EK |
1226 | |
1227 | case PTRACE_GET_SYSCALL_INFO: | |
1228 | ret = ptrace_get_syscall_info(child, addr, datavp); | |
1229 | break; | |
2225a122 | 1230 | #endif |
f8e529ed TA |
1231 | |
1232 | case PTRACE_SECCOMP_GET_FILTER: | |
1233 | ret = seccomp_get_filter(child, addr, datavp); | |
1234 | break; | |
1235 | ||
26500475 TA |
1236 | case PTRACE_SECCOMP_GET_METADATA: |
1237 | ret = seccomp_get_metadata(child, addr, datavp); | |
1238 | break; | |
1239 | ||
90f093fa PF |
1240 | #ifdef CONFIG_RSEQ |
1241 | case PTRACE_GET_RSEQ_CONFIGURATION: | |
1242 | ret = ptrace_get_rseq_configuration(child, addr, datavp); | |
1243 | break; | |
1244 | #endif | |
1245 | ||
3f67987c GP |
1246 | case PTRACE_SET_SYSCALL_USER_DISPATCH_CONFIG: |
1247 | ret = syscall_user_dispatch_set_config(child, addr, datavp); | |
1248 | break; | |
1249 | ||
1250 | case PTRACE_GET_SYSCALL_USER_DISPATCH_CONFIG: | |
1251 | ret = syscall_user_dispatch_get_config(child, addr, datavp); | |
1252 | break; | |
1253 | ||
1da177e4 LT |
1254 | default: |
1255 | break; | |
1256 | } | |
1257 | ||
1258 | return ret; | |
1259 | } | |
481bed45 | 1260 | |
4abf9869 NK |
1261 | SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr, |
1262 | unsigned long, data) | |
481bed45 CH |
1263 | { |
1264 | struct task_struct *child; | |
1265 | long ret; | |
1266 | ||
6b9c7ed8 CH |
1267 | if (request == PTRACE_TRACEME) { |
1268 | ret = ptrace_traceme(); | |
481bed45 | 1269 | goto out; |
6b9c7ed8 CH |
1270 | } |
1271 | ||
2ee08260 MR |
1272 | child = find_get_task_by_vpid(pid); |
1273 | if (!child) { | |
1274 | ret = -ESRCH; | |
6b9c7ed8 CH |
1275 | goto out; |
1276 | } | |
481bed45 | 1277 | |
3544d72a | 1278 | if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) { |
aa9147c9 | 1279 | ret = ptrace_attach(child, request, addr, data); |
005f18df | 1280 | goto out_put_task_struct; |
481bed45 CH |
1281 | } |
1282 | ||
fca26f26 TH |
1283 | ret = ptrace_check_attach(child, request == PTRACE_KILL || |
1284 | request == PTRACE_INTERRUPT); | |
481bed45 CH |
1285 | if (ret < 0) |
1286 | goto out_put_task_struct; | |
1287 | ||
1288 | ret = arch_ptrace(child, request, addr, data); | |
9899d11f ON |
1289 | if (ret || request != PTRACE_DETACH) |
1290 | ptrace_unfreeze_traced(child); | |
481bed45 CH |
1291 | |
1292 | out_put_task_struct: | |
1293 | put_task_struct(child); | |
1294 | out: | |
481bed45 CH |
1295 | return ret; |
1296 | } | |
76647323 | 1297 | |
4abf9869 NK |
1298 | int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr, |
1299 | unsigned long data) | |
76647323 AD |
1300 | { |
1301 | unsigned long tmp; | |
1302 | int copied; | |
1303 | ||
84d77d3f | 1304 | copied = ptrace_access_vm(tsk, addr, &tmp, sizeof(tmp), FOLL_FORCE); |
76647323 AD |
1305 | if (copied != sizeof(tmp)) |
1306 | return -EIO; | |
1307 | return put_user(tmp, (unsigned long __user *)data); | |
1308 | } | |
f284ce72 | 1309 | |
4abf9869 NK |
1310 | int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr, |
1311 | unsigned long data) | |
f284ce72 AD |
1312 | { |
1313 | int copied; | |
1314 | ||
84d77d3f | 1315 | copied = ptrace_access_vm(tsk, addr, &data, sizeof(data), |
f307ab6d | 1316 | FOLL_FORCE | FOLL_WRITE); |
f284ce72 AD |
1317 | return (copied == sizeof(data)) ? 0 : -EIO; |
1318 | } | |
032d82d9 | 1319 | |
96b8936a | 1320 | #if defined CONFIG_COMPAT |
032d82d9 RM |
1321 | |
1322 | int compat_ptrace_request(struct task_struct *child, compat_long_t request, | |
1323 | compat_ulong_t addr, compat_ulong_t data) | |
1324 | { | |
1325 | compat_ulong_t __user *datap = compat_ptr(data); | |
1326 | compat_ulong_t word; | |
ae7795bc | 1327 | kernel_siginfo_t siginfo; |
032d82d9 RM |
1328 | int ret; |
1329 | ||
1330 | switch (request) { | |
1331 | case PTRACE_PEEKTEXT: | |
1332 | case PTRACE_PEEKDATA: | |
84d77d3f | 1333 | ret = ptrace_access_vm(child, addr, &word, sizeof(word), |
f307ab6d | 1334 | FOLL_FORCE); |
032d82d9 RM |
1335 | if (ret != sizeof(word)) |
1336 | ret = -EIO; | |
1337 | else | |
1338 | ret = put_user(word, datap); | |
1339 | break; | |
1340 | ||
1341 | case PTRACE_POKETEXT: | |
1342 | case PTRACE_POKEDATA: | |
84d77d3f | 1343 | ret = ptrace_access_vm(child, addr, &data, sizeof(data), |
f307ab6d | 1344 | FOLL_FORCE | FOLL_WRITE); |
032d82d9 RM |
1345 | ret = (ret != sizeof(data) ? -EIO : 0); |
1346 | break; | |
1347 | ||
1348 | case PTRACE_GETEVENTMSG: | |
1349 | ret = put_user((compat_ulong_t) child->ptrace_message, datap); | |
1350 | break; | |
1351 | ||
e16b2781 RM |
1352 | case PTRACE_GETSIGINFO: |
1353 | ret = ptrace_getsiginfo(child, &siginfo); | |
1354 | if (!ret) | |
1355 | ret = copy_siginfo_to_user32( | |
1356 | (struct compat_siginfo __user *) datap, | |
1357 | &siginfo); | |
1358 | break; | |
1359 | ||
1360 | case PTRACE_SETSIGINFO: | |
4cd2e0e7 EB |
1361 | ret = copy_siginfo_from_user32( |
1362 | &siginfo, (struct compat_siginfo __user *) datap); | |
1363 | if (!ret) | |
e16b2781 RM |
1364 | ret = ptrace_setsiginfo(child, &siginfo); |
1365 | break; | |
2225a122 SS |
1366 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
1367 | case PTRACE_GETREGSET: | |
1368 | case PTRACE_SETREGSET: | |
1369 | { | |
1370 | struct iovec kiov; | |
1371 | struct compat_iovec __user *uiov = | |
1372 | (struct compat_iovec __user *) datap; | |
1373 | compat_uptr_t ptr; | |
1374 | compat_size_t len; | |
1375 | ||
96d4f267 | 1376 | if (!access_ok(uiov, sizeof(*uiov))) |
2225a122 SS |
1377 | return -EFAULT; |
1378 | ||
1379 | if (__get_user(ptr, &uiov->iov_base) || | |
1380 | __get_user(len, &uiov->iov_len)) | |
1381 | return -EFAULT; | |
1382 | ||
1383 | kiov.iov_base = compat_ptr(ptr); | |
1384 | kiov.iov_len = len; | |
1385 | ||
1386 | ret = ptrace_regset(child, request, addr, &kiov); | |
1387 | if (!ret) | |
1388 | ret = __put_user(kiov.iov_len, &uiov->iov_len); | |
1389 | break; | |
1390 | } | |
1391 | #endif | |
e16b2781 | 1392 | |
032d82d9 RM |
1393 | default: |
1394 | ret = ptrace_request(child, request, addr, data); | |
1395 | } | |
1396 | ||
1397 | return ret; | |
1398 | } | |
c269f196 | 1399 | |
62a6fa97 HC |
1400 | COMPAT_SYSCALL_DEFINE4(ptrace, compat_long_t, request, compat_long_t, pid, |
1401 | compat_long_t, addr, compat_long_t, data) | |
c269f196 RM |
1402 | { |
1403 | struct task_struct *child; | |
1404 | long ret; | |
1405 | ||
c269f196 RM |
1406 | if (request == PTRACE_TRACEME) { |
1407 | ret = ptrace_traceme(); | |
1408 | goto out; | |
1409 | } | |
1410 | ||
2ee08260 MR |
1411 | child = find_get_task_by_vpid(pid); |
1412 | if (!child) { | |
1413 | ret = -ESRCH; | |
c269f196 RM |
1414 | goto out; |
1415 | } | |
1416 | ||
3544d72a | 1417 | if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) { |
aa9147c9 | 1418 | ret = ptrace_attach(child, request, addr, data); |
c269f196 RM |
1419 | goto out_put_task_struct; |
1420 | } | |
1421 | ||
fca26f26 TH |
1422 | ret = ptrace_check_attach(child, request == PTRACE_KILL || |
1423 | request == PTRACE_INTERRUPT); | |
9899d11f | 1424 | if (!ret) { |
c269f196 | 1425 | ret = compat_arch_ptrace(child, request, addr, data); |
9899d11f ON |
1426 | if (ret || request != PTRACE_DETACH) |
1427 | ptrace_unfreeze_traced(child); | |
1428 | } | |
c269f196 RM |
1429 | |
1430 | out_put_task_struct: | |
1431 | put_task_struct(child); | |
1432 | out: | |
c269f196 RM |
1433 | return ret; |
1434 | } | |
96b8936a | 1435 | #endif /* CONFIG_COMPAT */ |