| 1 | // SPDX-License-Identifier: GPL-2.0-only |
| 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 | |
| 11 | #include <linux/capability.h> |
| 12 | #include <linux/export.h> |
| 13 | #include <linux/sched.h> |
| 14 | #include <linux/sched/mm.h> |
| 15 | #include <linux/sched/coredump.h> |
| 16 | #include <linux/sched/task.h> |
| 17 | #include <linux/errno.h> |
| 18 | #include <linux/mm.h> |
| 19 | #include <linux/highmem.h> |
| 20 | #include <linux/pagemap.h> |
| 21 | #include <linux/ptrace.h> |
| 22 | #include <linux/security.h> |
| 23 | #include <linux/signal.h> |
| 24 | #include <linux/uio.h> |
| 25 | #include <linux/audit.h> |
| 26 | #include <linux/pid_namespace.h> |
| 27 | #include <linux/syscalls.h> |
| 28 | #include <linux/uaccess.h> |
| 29 | #include <linux/regset.h> |
| 30 | #include <linux/hw_breakpoint.h> |
| 31 | #include <linux/cn_proc.h> |
| 32 | #include <linux/compat.h> |
| 33 | #include <linux/sched/signal.h> |
| 34 | #include <linux/minmax.h> |
| 35 | #include <linux/syscall_user_dispatch.h> |
| 36 | |
| 37 | #include <asm/syscall.h> /* for syscall_get_* */ |
| 38 | |
| 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(task: 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, ns: mm->user_ns))) { |
| 58 | mmput(mm); |
| 59 | return 0; |
| 60 | } |
| 61 | |
| 62 | ret = access_remote_vm(mm, addr, buf, len, gup_flags); |
| 63 | mmput(mm); |
| 64 | |
| 65 | return ret; |
| 66 | } |
| 67 | |
| 68 | |
| 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(new: &child->ptrace_entry, head: &new_parent->ptraced); |
| 74 | child->parent = new_parent; |
| 75 | child->ptracer_cred = get_cred(cred: ptracer_cred); |
| 76 | } |
| 77 | |
| 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 | */ |
| 84 | static void ptrace_link(struct task_struct *child, struct task_struct *new_parent) |
| 85 | { |
| 86 | __ptrace_link(child, new_parent, current_cred()); |
| 87 | } |
| 88 | |
| 89 | /** |
| 90 | * __ptrace_unlink - unlink ptracee and restore its execution state |
| 91 | * @child: ptracee to be unlinked |
| 92 | * |
| 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. |
| 113 | * |
| 114 | * CONTEXT: |
| 115 | * write_lock_irq(tasklist_lock) |
| 116 | */ |
| 117 | void __ptrace_unlink(struct task_struct *child) |
| 118 | { |
| 119 | const struct cred *old_cred; |
| 120 | BUG_ON(!child->ptrace); |
| 121 | |
| 122 | clear_task_syscall_work(child, SYSCALL_TRACE); |
| 123 | #if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU) |
| 124 | clear_task_syscall_work(child, SYSCALL_EMU); |
| 125 | #endif |
| 126 | |
| 127 | child->parent = child->real_parent; |
| 128 | list_del_init(entry: &child->ptrace_entry); |
| 129 | old_cred = child->ptracer_cred; |
| 130 | child->ptracer_cred = NULL; |
| 131 | put_cred(cred: old_cred); |
| 132 | |
| 133 | spin_lock(lock: &child->sighand->siglock); |
| 134 | child->ptrace = 0; |
| 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(task: child, JOBCTL_TRAP_MASK); |
| 140 | task_clear_jobctl_trapping(task: child); |
| 141 | |
| 142 | /* |
| 143 | * Reinstate JOBCTL_STOP_PENDING if group stop is in effect and |
| 144 | * @child isn't dead. |
| 145 | */ |
| 146 | if (!(child->flags & PF_EXITING) && |
| 147 | (child->signal->flags & SIGNAL_STOP_STOPPED || |
| 148 | child->signal->group_stop_count)) |
| 149 | child->jobctl |= JOBCTL_STOP_PENDING; |
| 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 | */ |
| 157 | if (child->jobctl & JOBCTL_STOP_PENDING || task_is_traced(child)) |
| 158 | ptrace_signal_wake_up(t: child, resume: true); |
| 159 | |
| 160 | spin_unlock(lock: &child->sighand->siglock); |
| 161 | } |
| 162 | |
| 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(tsk: 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 | |
| 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 | */ |
| 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(lock: &task->sighand->siglock); |
| 193 | if (task_is_traced(task) && !looks_like_a_spurious_pid(task) && |
| 194 | !__fatal_signal_pending(p: task)) { |
| 195 | task->jobctl |= JOBCTL_PTRACE_FROZEN; |
| 196 | ret = true; |
| 197 | } |
| 198 | spin_unlock_irq(lock: &task->sighand->siglock); |
| 199 | |
| 200 | return ret; |
| 201 | } |
| 202 | |
| 203 | static void ptrace_unfreeze_traced(struct task_struct *task) |
| 204 | { |
| 205 | unsigned long flags; |
| 206 | |
| 207 | /* |
| 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. |
| 211 | */ |
| 212 | if (lock_task_sighand(task, flags: &flags)) { |
| 213 | task->jobctl &= ~JOBCTL_PTRACE_FROZEN; |
| 214 | if (__fatal_signal_pending(p: task)) { |
| 215 | task->jobctl &= ~JOBCTL_TRACED; |
| 216 | wake_up_state(tsk: task, __TASK_TRACED); |
| 217 | } |
| 218 | unlock_task_sighand(task, flags: &flags); |
| 219 | } |
| 220 | } |
| 221 | |
| 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. |
| 238 | */ |
| 239 | static int ptrace_check_attach(struct task_struct *child, bool ignore_state) |
| 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); |
| 251 | if (child->ptrace && child->parent == current) { |
| 252 | /* |
| 253 | * child->sighand can't be NULL, release_task() |
| 254 | * does ptrace_unlink() before __exit_signal(). |
| 255 | */ |
| 256 | if (ignore_state || ptrace_freeze_traced(task: child)) |
| 257 | ret = 0; |
| 258 | } |
| 259 | read_unlock(&tasklist_lock); |
| 260 | |
| 261 | if (!ret && !ignore_state && |
| 262 | WARN_ON_ONCE(!wait_task_inactive(child, __TASK_TRACED|TASK_FROZEN))) |
| 263 | ret = -ESRCH; |
| 264 | |
| 265 | return ret; |
| 266 | } |
| 267 | |
| 268 | static bool ptrace_has_cap(struct user_namespace *ns, unsigned int mode) |
| 269 | { |
| 270 | if (mode & PTRACE_MODE_NOAUDIT) |
| 271 | return ns_capable_noaudit(ns, CAP_SYS_PTRACE); |
| 272 | return ns_capable(ns, CAP_SYS_PTRACE); |
| 273 | } |
| 274 | |
| 275 | /* Returns 0 on success, -errno on denial. */ |
| 276 | static int __ptrace_may_access(struct task_struct *task, unsigned int mode) |
| 277 | { |
| 278 | const struct cred *cred = current_cred(), *tcred; |
| 279 | struct mm_struct *mm; |
| 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 | } |
| 287 | |
| 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 | */ |
| 296 | |
| 297 | /* Don't let security modules deny introspection */ |
| 298 | if (same_thread_group(p1: task, current)) |
| 299 | return 0; |
| 300 | rcu_read_lock(); |
| 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 | } |
| 316 | tcred = __task_cred(task); |
| 317 | if (uid_eq(left: caller_uid, right: tcred->euid) && |
| 318 | uid_eq(left: caller_uid, right: tcred->suid) && |
| 319 | uid_eq(left: caller_uid, right: tcred->uid) && |
| 320 | gid_eq(left: caller_gid, right: tcred->egid) && |
| 321 | gid_eq(left: caller_gid, right: tcred->sgid) && |
| 322 | gid_eq(left: caller_gid, right: tcred->gid)) |
| 323 | goto ok; |
| 324 | if (ptrace_has_cap(ns: tcred->user_ns, mode)) |
| 325 | goto ok; |
| 326 | rcu_read_unlock(); |
| 327 | return -EPERM; |
| 328 | ok: |
| 329 | rcu_read_unlock(); |
| 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(); |
| 340 | mm = task->mm; |
| 341 | if (mm && |
| 342 | ((get_dumpable(mm) != SUID_DUMP_USER) && |
| 343 | !ptrace_has_cap(ns: mm->user_ns, mode))) |
| 344 | return -EPERM; |
| 345 | |
| 346 | return security_ptrace_access_check(child: task, mode); |
| 347 | } |
| 348 | |
| 349 | bool ptrace_may_access(struct task_struct *task, unsigned int mode) |
| 350 | { |
| 351 | int err; |
| 352 | task_lock(p: task); |
| 353 | err = __ptrace_may_access(task, mode); |
| 354 | task_unlock(p: task); |
| 355 | return !err; |
| 356 | } |
| 357 | |
| 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(s: ¤t->seccomp) != SECCOMP_MODE_DISABLED || |
| 372 | current->ptrace & PT_SUSPEND_SECCOMP) |
| 373 | return -EPERM; |
| 374 | } |
| 375 | return 0; |
| 376 | } |
| 377 | |
| 378 | static inline void ptrace_set_stopped(struct task_struct *task, bool seize) |
| 379 | { |
| 380 | guard(spinlock)(l: &task->sighand->siglock); |
| 381 | |
| 382 | /* SEIZE doesn't trap tracee on attach */ |
| 383 | if (!seize) |
| 384 | send_signal_locked(SIGSTOP, SEND_SIG_PRIV, p: task, type: PIDTYPE_PID); |
| 385 | /* |
| 386 | * If the task is already STOPPED, set JOBCTL_TRAP_STOP and |
| 387 | * TRAPPING, and kick it so that it transits to TRACED. TRAPPING |
| 388 | * will be cleared if the child completes the transition or any |
| 389 | * event which clears the group stop states happens. We'll wait |
| 390 | * for the transition to complete before returning from this |
| 391 | * function. |
| 392 | * |
| 393 | * This hides STOPPED -> RUNNING -> TRACED transition from the |
| 394 | * attaching thread but a different thread in the same group can |
| 395 | * still observe the transient RUNNING state. IOW, if another |
| 396 | * thread's WNOHANG wait(2) on the stopped tracee races against |
| 397 | * ATTACH, the wait(2) may fail due to the transient RUNNING. |
| 398 | * |
| 399 | * The following task_is_stopped() test is safe as both transitions |
| 400 | * in and out of STOPPED are protected by siglock. |
| 401 | */ |
| 402 | if (task_is_stopped(task) && |
| 403 | task_set_jobctl_pending(task, JOBCTL_TRAP_STOP | JOBCTL_TRAPPING)) { |
| 404 | task->jobctl &= ~JOBCTL_STOPPED; |
| 405 | signal_wake_up_state(t: task, __TASK_STOPPED); |
| 406 | } |
| 407 | } |
| 408 | |
| 409 | static int ptrace_attach(struct task_struct *task, long request, |
| 410 | unsigned long addr, |
| 411 | unsigned long flags) |
| 412 | { |
| 413 | bool seize = (request == PTRACE_SEIZE); |
| 414 | int retval; |
| 415 | |
| 416 | if (seize) { |
| 417 | if (addr != 0) |
| 418 | return -EIO; |
| 419 | /* |
| 420 | * This duplicates the check in check_ptrace_options() because |
| 421 | * ptrace_attach() and ptrace_setoptions() have historically |
| 422 | * used different error codes for unknown ptrace options. |
| 423 | */ |
| 424 | if (flags & ~(unsigned long)PTRACE_O_MASK) |
| 425 | return -EIO; |
| 426 | |
| 427 | retval = check_ptrace_options(data: flags); |
| 428 | if (retval) |
| 429 | return retval; |
| 430 | flags = PT_PTRACED | PT_SEIZED | (flags << PT_OPT_FLAG_SHIFT); |
| 431 | } else { |
| 432 | flags = PT_PTRACED; |
| 433 | } |
| 434 | |
| 435 | audit_ptrace(t: task); |
| 436 | |
| 437 | if (unlikely(task->flags & PF_KTHREAD)) |
| 438 | return -EPERM; |
| 439 | if (same_thread_group(p1: task, current)) |
| 440 | return -EPERM; |
| 441 | |
| 442 | /* |
| 443 | * Protect exec's credential calculations against our interference; |
| 444 | * SUID, SGID and LSM creds get determined differently |
| 445 | * under ptrace. |
| 446 | */ |
| 447 | scoped_cond_guard (mutex_intr, return -ERESTARTNOINTR, |
| 448 | &task->signal->cred_guard_mutex) { |
| 449 | |
| 450 | scoped_guard (task_lock, task) { |
| 451 | retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH_REALCREDS); |
| 452 | if (retval) |
| 453 | return retval; |
| 454 | } |
| 455 | |
| 456 | scoped_guard (write_lock_irq, &tasklist_lock) { |
| 457 | if (unlikely(task->exit_state)) |
| 458 | return -EPERM; |
| 459 | if (task->ptrace) |
| 460 | return -EPERM; |
| 461 | |
| 462 | task->ptrace = flags; |
| 463 | ptrace_link(child: task, current); |
| 464 | ptrace_set_stopped(task, seize); |
| 465 | } |
| 466 | } |
| 467 | |
| 468 | /* |
| 469 | * We do not bother to change retval or clear JOBCTL_TRAPPING |
| 470 | * if wait_on_bit() was interrupted by SIGKILL. The tracer will |
| 471 | * not return to user-mode, it will exit and clear this bit in |
| 472 | * __ptrace_unlink() if it wasn't already cleared by the tracee; |
| 473 | * and until then nobody can ptrace this task. |
| 474 | */ |
| 475 | wait_on_bit(word: &task->jobctl, JOBCTL_TRAPPING_BIT, TASK_KILLABLE); |
| 476 | proc_ptrace_connector(task, PTRACE_ATTACH); |
| 477 | |
| 478 | return 0; |
| 479 | } |
| 480 | |
| 481 | /** |
| 482 | * ptrace_traceme -- helper for PTRACE_TRACEME |
| 483 | * |
| 484 | * Performs checks and sets PT_PTRACED. |
| 485 | * Should be used by all ptrace implementations for PTRACE_TRACEME. |
| 486 | */ |
| 487 | static int ptrace_traceme(void) |
| 488 | { |
| 489 | int ret = -EPERM; |
| 490 | |
| 491 | write_lock_irq(&tasklist_lock); |
| 492 | /* Are we already being traced? */ |
| 493 | if (!current->ptrace) { |
| 494 | ret = security_ptrace_traceme(current->parent); |
| 495 | /* |
| 496 | * Check PF_EXITING to ensure ->real_parent has not passed |
| 497 | * exit_ptrace(). Otherwise we don't report the error but |
| 498 | * pretend ->real_parent untraces us right after return. |
| 499 | */ |
| 500 | if (!ret && !(current->real_parent->flags & PF_EXITING)) { |
| 501 | current->ptrace = PT_PTRACED; |
| 502 | ptrace_link(current, current->real_parent); |
| 503 | } |
| 504 | } |
| 505 | write_unlock_irq(&tasklist_lock); |
| 506 | |
| 507 | return ret; |
| 508 | } |
| 509 | |
| 510 | /* |
| 511 | * Called with irqs disabled, returns true if childs should reap themselves. |
| 512 | */ |
| 513 | static int ignoring_children(struct sighand_struct *sigh) |
| 514 | { |
| 515 | int ret; |
| 516 | spin_lock(lock: &sigh->siglock); |
| 517 | ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) || |
| 518 | (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT); |
| 519 | spin_unlock(lock: &sigh->siglock); |
| 520 | return ret; |
| 521 | } |
| 522 | |
| 523 | /* |
| 524 | * Called with tasklist_lock held for writing. |
| 525 | * Unlink a traced task, and clean it up if it was a traced zombie. |
| 526 | * Return true if it needs to be reaped with release_task(). |
| 527 | * (We can't call release_task() here because we already hold tasklist_lock.) |
| 528 | * |
| 529 | * If it's a zombie, our attachedness prevented normal parent notification |
| 530 | * or self-reaping. Do notification now if it would have happened earlier. |
| 531 | * If it should reap itself, return true. |
| 532 | * |
| 533 | * If it's our own child, there is no notification to do. But if our normal |
| 534 | * children self-reap, then this child was prevented by ptrace and we must |
| 535 | * reap it now, in that case we must also wake up sub-threads sleeping in |
| 536 | * do_wait(). |
| 537 | */ |
| 538 | static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p) |
| 539 | { |
| 540 | bool dead; |
| 541 | |
| 542 | __ptrace_unlink(child: p); |
| 543 | |
| 544 | if (p->exit_state != EXIT_ZOMBIE) |
| 545 | return false; |
| 546 | |
| 547 | dead = !thread_group_leader(p); |
| 548 | |
| 549 | if (!dead && thread_group_empty(p)) { |
| 550 | if (!same_thread_group(p1: p->real_parent, p2: tracer)) |
| 551 | dead = do_notify_parent(p, p->exit_signal); |
| 552 | else if (ignoring_children(sigh: tracer->sighand)) { |
| 553 | __wake_up_parent(p, parent: tracer); |
| 554 | dead = true; |
| 555 | } |
| 556 | } |
| 557 | /* Mark it as in the process of being reaped. */ |
| 558 | if (dead) |
| 559 | p->exit_state = EXIT_DEAD; |
| 560 | return dead; |
| 561 | } |
| 562 | |
| 563 | static int ptrace_detach(struct task_struct *child, unsigned int data) |
| 564 | { |
| 565 | if (!valid_signal(sig: data)) |
| 566 | return -EIO; |
| 567 | |
| 568 | /* Architecture-specific hardware disable .. */ |
| 569 | ptrace_disable(child); |
| 570 | |
| 571 | write_lock_irq(&tasklist_lock); |
| 572 | /* |
| 573 | * We rely on ptrace_freeze_traced(). It can't be killed and |
| 574 | * untraced by another thread, it can't be a zombie. |
| 575 | */ |
| 576 | WARN_ON(!child->ptrace || child->exit_state); |
| 577 | /* |
| 578 | * tasklist_lock avoids the race with wait_task_stopped(), see |
| 579 | * the comment in ptrace_resume(). |
| 580 | */ |
| 581 | child->exit_code = data; |
| 582 | __ptrace_detach(current, p: child); |
| 583 | write_unlock_irq(&tasklist_lock); |
| 584 | |
| 585 | proc_ptrace_connector(task: child, PTRACE_DETACH); |
| 586 | |
| 587 | return 0; |
| 588 | } |
| 589 | |
| 590 | /* |
| 591 | * Detach all tasks we were using ptrace on. Called with tasklist held |
| 592 | * for writing. |
| 593 | */ |
| 594 | void exit_ptrace(struct task_struct *tracer, struct list_head *dead) |
| 595 | { |
| 596 | struct task_struct *p, *n; |
| 597 | |
| 598 | list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) { |
| 599 | if (unlikely(p->ptrace & PT_EXITKILL)) |
| 600 | send_sig_info(SIGKILL, SEND_SIG_PRIV, p); |
| 601 | |
| 602 | if (__ptrace_detach(tracer, p)) |
| 603 | list_add(new: &p->ptrace_entry, head: dead); |
| 604 | } |
| 605 | } |
| 606 | |
| 607 | int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len) |
| 608 | { |
| 609 | int copied = 0; |
| 610 | |
| 611 | while (len > 0) { |
| 612 | char buf[128]; |
| 613 | int this_len, retval; |
| 614 | |
| 615 | this_len = (len > sizeof(buf)) ? sizeof(buf) : len; |
| 616 | retval = ptrace_access_vm(tsk, addr: src, buf, len: this_len, gup_flags: FOLL_FORCE); |
| 617 | |
| 618 | if (!retval) { |
| 619 | if (copied) |
| 620 | break; |
| 621 | return -EIO; |
| 622 | } |
| 623 | if (copy_to_user(to: dst, from: buf, n: retval)) |
| 624 | return -EFAULT; |
| 625 | copied += retval; |
| 626 | src += retval; |
| 627 | dst += retval; |
| 628 | len -= retval; |
| 629 | } |
| 630 | return copied; |
| 631 | } |
| 632 | |
| 633 | int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len) |
| 634 | { |
| 635 | int copied = 0; |
| 636 | |
| 637 | while (len > 0) { |
| 638 | char buf[128]; |
| 639 | int this_len, retval; |
| 640 | |
| 641 | this_len = (len > sizeof(buf)) ? sizeof(buf) : len; |
| 642 | if (copy_from_user(to: buf, from: src, n: this_len)) |
| 643 | return -EFAULT; |
| 644 | retval = ptrace_access_vm(tsk, addr: dst, buf, len: this_len, |
| 645 | gup_flags: FOLL_FORCE | FOLL_WRITE); |
| 646 | if (!retval) { |
| 647 | if (copied) |
| 648 | break; |
| 649 | return -EIO; |
| 650 | } |
| 651 | copied += retval; |
| 652 | src += retval; |
| 653 | dst += retval; |
| 654 | len -= retval; |
| 655 | } |
| 656 | return copied; |
| 657 | } |
| 658 | |
| 659 | static int ptrace_setoptions(struct task_struct *child, unsigned long data) |
| 660 | { |
| 661 | unsigned flags; |
| 662 | int ret; |
| 663 | |
| 664 | ret = check_ptrace_options(data); |
| 665 | if (ret) |
| 666 | return ret; |
| 667 | |
| 668 | /* Avoid intermediate state when all opts are cleared */ |
| 669 | flags = child->ptrace; |
| 670 | flags &= ~(PTRACE_O_MASK << PT_OPT_FLAG_SHIFT); |
| 671 | flags |= (data << PT_OPT_FLAG_SHIFT); |
| 672 | child->ptrace = flags; |
| 673 | |
| 674 | return 0; |
| 675 | } |
| 676 | |
| 677 | static int ptrace_getsiginfo(struct task_struct *child, kernel_siginfo_t *info) |
| 678 | { |
| 679 | unsigned long flags; |
| 680 | int error = -ESRCH; |
| 681 | |
| 682 | if (lock_task_sighand(task: child, flags: &flags)) { |
| 683 | error = -EINVAL; |
| 684 | if (likely(child->last_siginfo != NULL)) { |
| 685 | copy_siginfo(to: info, from: child->last_siginfo); |
| 686 | error = 0; |
| 687 | } |
| 688 | unlock_task_sighand(task: child, flags: &flags); |
| 689 | } |
| 690 | return error; |
| 691 | } |
| 692 | |
| 693 | static int ptrace_setsiginfo(struct task_struct *child, const kernel_siginfo_t *info) |
| 694 | { |
| 695 | unsigned long flags; |
| 696 | int error = -ESRCH; |
| 697 | |
| 698 | if (lock_task_sighand(task: child, flags: &flags)) { |
| 699 | error = -EINVAL; |
| 700 | if (likely(child->last_siginfo != NULL)) { |
| 701 | copy_siginfo(to: child->last_siginfo, from: info); |
| 702 | error = 0; |
| 703 | } |
| 704 | unlock_task_sighand(task: child, flags: &flags); |
| 705 | } |
| 706 | return error; |
| 707 | } |
| 708 | |
| 709 | static int ptrace_peek_siginfo(struct task_struct *child, |
| 710 | unsigned long addr, |
| 711 | unsigned long data) |
| 712 | { |
| 713 | struct ptrace_peeksiginfo_args arg; |
| 714 | struct sigpending *pending; |
| 715 | struct sigqueue *q; |
| 716 | int ret, i; |
| 717 | |
| 718 | ret = copy_from_user(to: &arg, from: (void __user *) addr, |
| 719 | n: sizeof(struct ptrace_peeksiginfo_args)); |
| 720 | if (ret) |
| 721 | return -EFAULT; |
| 722 | |
| 723 | if (arg.flags & ~PTRACE_PEEKSIGINFO_SHARED) |
| 724 | return -EINVAL; /* unknown flags */ |
| 725 | |
| 726 | if (arg.nr < 0) |
| 727 | return -EINVAL; |
| 728 | |
| 729 | /* Ensure arg.off fits in an unsigned long */ |
| 730 | if (arg.off > ULONG_MAX) |
| 731 | return 0; |
| 732 | |
| 733 | if (arg.flags & PTRACE_PEEKSIGINFO_SHARED) |
| 734 | pending = &child->signal->shared_pending; |
| 735 | else |
| 736 | pending = &child->pending; |
| 737 | |
| 738 | for (i = 0; i < arg.nr; ) { |
| 739 | kernel_siginfo_t info; |
| 740 | unsigned long off = arg.off + i; |
| 741 | bool found = false; |
| 742 | |
| 743 | spin_lock_irq(lock: &child->sighand->siglock); |
| 744 | list_for_each_entry(q, &pending->list, list) { |
| 745 | if (!off--) { |
| 746 | found = true; |
| 747 | copy_siginfo(to: &info, from: &q->info); |
| 748 | break; |
| 749 | } |
| 750 | } |
| 751 | spin_unlock_irq(lock: &child->sighand->siglock); |
| 752 | |
| 753 | if (!found) /* beyond the end of the list */ |
| 754 | break; |
| 755 | |
| 756 | #ifdef CONFIG_COMPAT |
| 757 | if (unlikely(in_compat_syscall())) { |
| 758 | compat_siginfo_t __user *uinfo = compat_ptr(uptr: data); |
| 759 | |
| 760 | if (copy_siginfo_to_user32(to: uinfo, from: &info)) { |
| 761 | ret = -EFAULT; |
| 762 | break; |
| 763 | } |
| 764 | |
| 765 | } else |
| 766 | #endif |
| 767 | { |
| 768 | siginfo_t __user *uinfo = (siginfo_t __user *) data; |
| 769 | |
| 770 | if (copy_siginfo_to_user(to: uinfo, from: &info)) { |
| 771 | ret = -EFAULT; |
| 772 | break; |
| 773 | } |
| 774 | } |
| 775 | |
| 776 | data += sizeof(siginfo_t); |
| 777 | i++; |
| 778 | |
| 779 | if (signal_pending(current)) |
| 780 | break; |
| 781 | |
| 782 | cond_resched(); |
| 783 | } |
| 784 | |
| 785 | if (i > 0) |
| 786 | return i; |
| 787 | |
| 788 | return ret; |
| 789 | } |
| 790 | |
| 791 | #ifdef CONFIG_RSEQ |
| 792 | static long ptrace_get_rseq_configuration(struct task_struct *task, |
| 793 | unsigned long size, void __user *data) |
| 794 | { |
| 795 | struct ptrace_rseq_configuration conf = { |
| 796 | .rseq_abi_pointer = (u64)(uintptr_t)task->rseq, |
| 797 | .rseq_abi_size = task->rseq_len, |
| 798 | .signature = task->rseq_sig, |
| 799 | .flags = 0, |
| 800 | }; |
| 801 | |
| 802 | size = min_t(unsigned long, size, sizeof(conf)); |
| 803 | if (copy_to_user(to: data, from: &conf, n: size)) |
| 804 | return -EFAULT; |
| 805 | return sizeof(conf); |
| 806 | } |
| 807 | #endif |
| 808 | |
| 809 | #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP) |
| 810 | |
| 811 | #ifdef PTRACE_SINGLEBLOCK |
| 812 | #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK) |
| 813 | #else |
| 814 | #define is_singleblock(request) 0 |
| 815 | #endif |
| 816 | |
| 817 | #ifdef PTRACE_SYSEMU |
| 818 | #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP) |
| 819 | #else |
| 820 | #define is_sysemu_singlestep(request) 0 |
| 821 | #endif |
| 822 | |
| 823 | static int ptrace_resume(struct task_struct *child, long request, |
| 824 | unsigned long data) |
| 825 | { |
| 826 | if (!valid_signal(sig: data)) |
| 827 | return -EIO; |
| 828 | |
| 829 | if (request == PTRACE_SYSCALL) |
| 830 | set_task_syscall_work(child, SYSCALL_TRACE); |
| 831 | else |
| 832 | clear_task_syscall_work(child, SYSCALL_TRACE); |
| 833 | |
| 834 | #if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU) |
| 835 | if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP) |
| 836 | set_task_syscall_work(child, SYSCALL_EMU); |
| 837 | else |
| 838 | clear_task_syscall_work(child, SYSCALL_EMU); |
| 839 | #endif |
| 840 | |
| 841 | if (is_singleblock(request)) { |
| 842 | if (unlikely(!arch_has_block_step())) |
| 843 | return -EIO; |
| 844 | user_enable_block_step(child); |
| 845 | } else if (is_singlestep(request) || is_sysemu_singlestep(request)) { |
| 846 | if (unlikely(!arch_has_single_step())) |
| 847 | return -EIO; |
| 848 | user_enable_single_step(child); |
| 849 | } else { |
| 850 | user_disable_single_step(child); |
| 851 | } |
| 852 | |
| 853 | /* |
| 854 | * Change ->exit_code and ->state under siglock to avoid the race |
| 855 | * with wait_task_stopped() in between; a non-zero ->exit_code will |
| 856 | * wrongly look like another report from tracee. |
| 857 | * |
| 858 | * Note that we need siglock even if ->exit_code == data and/or this |
| 859 | * status was not reported yet, the new status must not be cleared by |
| 860 | * wait_task_stopped() after resume. |
| 861 | */ |
| 862 | spin_lock_irq(lock: &child->sighand->siglock); |
| 863 | child->exit_code = data; |
| 864 | child->jobctl &= ~JOBCTL_TRACED; |
| 865 | wake_up_state(tsk: child, __TASK_TRACED); |
| 866 | spin_unlock_irq(lock: &child->sighand->siglock); |
| 867 | |
| 868 | return 0; |
| 869 | } |
| 870 | |
| 871 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
| 872 | |
| 873 | static const struct user_regset * |
| 874 | find_regset(const struct user_regset_view *view, unsigned int type) |
| 875 | { |
| 876 | const struct user_regset *regset; |
| 877 | int n; |
| 878 | |
| 879 | for (n = 0; n < view->n; ++n) { |
| 880 | regset = view->regsets + n; |
| 881 | if (regset->core_note_type == type) |
| 882 | return regset; |
| 883 | } |
| 884 | |
| 885 | return NULL; |
| 886 | } |
| 887 | |
| 888 | static int ptrace_regset(struct task_struct *task, int req, unsigned int type, |
| 889 | struct iovec *kiov) |
| 890 | { |
| 891 | const struct user_regset_view *view = task_user_regset_view(tsk: task); |
| 892 | const struct user_regset *regset = find_regset(view, type); |
| 893 | int regset_no; |
| 894 | |
| 895 | if (!regset || (kiov->iov_len % regset->size) != 0) |
| 896 | return -EINVAL; |
| 897 | |
| 898 | regset_no = regset - view->regsets; |
| 899 | kiov->iov_len = min(kiov->iov_len, |
| 900 | (__kernel_size_t) (regset->n * regset->size)); |
| 901 | |
| 902 | if (req == PTRACE_GETREGSET) |
| 903 | return copy_regset_to_user(target: task, view, setno: regset_no, offset: 0, |
| 904 | size: kiov->iov_len, data: kiov->iov_base); |
| 905 | else |
| 906 | return copy_regset_from_user(target: task, view, setno: regset_no, offset: 0, |
| 907 | size: kiov->iov_len, data: kiov->iov_base); |
| 908 | } |
| 909 | |
| 910 | /* |
| 911 | * This is declared in linux/regset.h and defined in machine-dependent |
| 912 | * code. We put the export here, near the primary machine-neutral use, |
| 913 | * to ensure no machine forgets it. |
| 914 | */ |
| 915 | EXPORT_SYMBOL_GPL(task_user_regset_view); |
| 916 | |
| 917 | static unsigned long |
| 918 | ptrace_get_syscall_info_entry(struct task_struct *child, struct pt_regs *regs, |
| 919 | struct ptrace_syscall_info *info) |
| 920 | { |
| 921 | unsigned long args[ARRAY_SIZE(info->entry.args)]; |
| 922 | int i; |
| 923 | |
| 924 | info->entry.nr = syscall_get_nr(task: child, regs); |
| 925 | syscall_get_arguments(task: child, regs, args); |
| 926 | for (i = 0; i < ARRAY_SIZE(args); i++) |
| 927 | info->entry.args[i] = args[i]; |
| 928 | |
| 929 | /* args is the last field in struct ptrace_syscall_info.entry */ |
| 930 | return offsetofend(struct ptrace_syscall_info, entry.args); |
| 931 | } |
| 932 | |
| 933 | static unsigned long |
| 934 | ptrace_get_syscall_info_seccomp(struct task_struct *child, struct pt_regs *regs, |
| 935 | struct ptrace_syscall_info *info) |
| 936 | { |
| 937 | /* |
| 938 | * As struct ptrace_syscall_info.entry is currently a subset |
| 939 | * of struct ptrace_syscall_info.seccomp, it makes sense to |
| 940 | * initialize that subset using ptrace_get_syscall_info_entry(). |
| 941 | * This can be reconsidered in the future if these structures |
| 942 | * diverge significantly enough. |
| 943 | */ |
| 944 | ptrace_get_syscall_info_entry(child, regs, info); |
| 945 | info->seccomp.ret_data = child->ptrace_message; |
| 946 | |
| 947 | /* |
| 948 | * ret_data is the last non-reserved field |
| 949 | * in struct ptrace_syscall_info.seccomp |
| 950 | */ |
| 951 | return offsetofend(struct ptrace_syscall_info, seccomp.ret_data); |
| 952 | } |
| 953 | |
| 954 | static unsigned long |
| 955 | ptrace_get_syscall_info_exit(struct task_struct *child, struct pt_regs *regs, |
| 956 | struct ptrace_syscall_info *info) |
| 957 | { |
| 958 | info->exit.rval = syscall_get_error(task: child, regs); |
| 959 | info->exit.is_error = !!info->exit.rval; |
| 960 | if (!info->exit.is_error) |
| 961 | info->exit.rval = syscall_get_return_value(task: child, regs); |
| 962 | |
| 963 | /* is_error is the last field in struct ptrace_syscall_info.exit */ |
| 964 | return offsetofend(struct ptrace_syscall_info, exit.is_error); |
| 965 | } |
| 966 | |
| 967 | static int |
| 968 | ptrace_get_syscall_info_op(struct task_struct *child) |
| 969 | { |
| 970 | /* |
| 971 | * This does not need lock_task_sighand() to access |
| 972 | * child->last_siginfo because ptrace_freeze_traced() |
| 973 | * called earlier by ptrace_check_attach() ensures that |
| 974 | * the tracee cannot go away and clear its last_siginfo. |
| 975 | */ |
| 976 | switch (child->last_siginfo ? child->last_siginfo->si_code : 0) { |
| 977 | case SIGTRAP | 0x80: |
| 978 | switch (child->ptrace_message) { |
| 979 | case PTRACE_EVENTMSG_SYSCALL_ENTRY: |
| 980 | return PTRACE_SYSCALL_INFO_ENTRY; |
| 981 | case PTRACE_EVENTMSG_SYSCALL_EXIT: |
| 982 | return PTRACE_SYSCALL_INFO_EXIT; |
| 983 | default: |
| 984 | return PTRACE_SYSCALL_INFO_NONE; |
| 985 | } |
| 986 | case SIGTRAP | (PTRACE_EVENT_SECCOMP << 8): |
| 987 | return PTRACE_SYSCALL_INFO_SECCOMP; |
| 988 | default: |
| 989 | return PTRACE_SYSCALL_INFO_NONE; |
| 990 | } |
| 991 | } |
| 992 | |
| 993 | static int |
| 994 | ptrace_get_syscall_info(struct task_struct *child, unsigned long user_size, |
| 995 | void __user *datavp) |
| 996 | { |
| 997 | struct pt_regs *regs = task_pt_regs(child); |
| 998 | struct ptrace_syscall_info info = { |
| 999 | .op = ptrace_get_syscall_info_op(child), |
| 1000 | .arch = syscall_get_arch(task: child), |
| 1001 | .instruction_pointer = instruction_pointer(regs), |
| 1002 | .stack_pointer = user_stack_pointer(regs), |
| 1003 | }; |
| 1004 | unsigned long actual_size = offsetof(struct ptrace_syscall_info, entry); |
| 1005 | unsigned long write_size; |
| 1006 | |
| 1007 | switch (info.op) { |
| 1008 | case PTRACE_SYSCALL_INFO_ENTRY: |
| 1009 | actual_size = ptrace_get_syscall_info_entry(child, regs, info: &info); |
| 1010 | break; |
| 1011 | case PTRACE_SYSCALL_INFO_EXIT: |
| 1012 | actual_size = ptrace_get_syscall_info_exit(child, regs, info: &info); |
| 1013 | break; |
| 1014 | case PTRACE_SYSCALL_INFO_SECCOMP: |
| 1015 | actual_size = ptrace_get_syscall_info_seccomp(child, regs, info: &info); |
| 1016 | break; |
| 1017 | } |
| 1018 | |
| 1019 | write_size = min(actual_size, user_size); |
| 1020 | return copy_to_user(to: datavp, from: &info, n: write_size) ? -EFAULT : actual_size; |
| 1021 | } |
| 1022 | |
| 1023 | static int |
| 1024 | ptrace_set_syscall_info_entry(struct task_struct *child, struct pt_regs *regs, |
| 1025 | struct ptrace_syscall_info *info) |
| 1026 | { |
| 1027 | unsigned long args[ARRAY_SIZE(info->entry.args)]; |
| 1028 | int nr = info->entry.nr; |
| 1029 | int i; |
| 1030 | |
| 1031 | /* |
| 1032 | * Check that the syscall number specified in info->entry.nr |
| 1033 | * is either a value of type "int" or a sign-extended value |
| 1034 | * of type "int". |
| 1035 | */ |
| 1036 | if (nr != info->entry.nr) |
| 1037 | return -ERANGE; |
| 1038 | |
| 1039 | for (i = 0; i < ARRAY_SIZE(args); i++) { |
| 1040 | args[i] = info->entry.args[i]; |
| 1041 | /* |
| 1042 | * Check that the syscall argument specified in |
| 1043 | * info->entry.args[i] is either a value of type |
| 1044 | * "unsigned long" or a sign-extended value of type "long". |
| 1045 | */ |
| 1046 | if (args[i] != info->entry.args[i]) |
| 1047 | return -ERANGE; |
| 1048 | } |
| 1049 | |
| 1050 | syscall_set_nr(task: child, regs, nr); |
| 1051 | /* |
| 1052 | * If the syscall number is set to -1, setting syscall arguments is not |
| 1053 | * just pointless, it would also clobber the syscall return value on |
| 1054 | * those architectures that share the same register both for the first |
| 1055 | * argument of syscall and its return value. |
| 1056 | */ |
| 1057 | if (nr != -1) |
| 1058 | syscall_set_arguments(task: child, regs, args); |
| 1059 | |
| 1060 | return 0; |
| 1061 | } |
| 1062 | |
| 1063 | static int |
| 1064 | ptrace_set_syscall_info_seccomp(struct task_struct *child, struct pt_regs *regs, |
| 1065 | struct ptrace_syscall_info *info) |
| 1066 | { |
| 1067 | /* |
| 1068 | * info->entry is currently a subset of info->seccomp, |
| 1069 | * info->seccomp.ret_data is currently ignored. |
| 1070 | */ |
| 1071 | return ptrace_set_syscall_info_entry(child, regs, info); |
| 1072 | } |
| 1073 | |
| 1074 | static int |
| 1075 | ptrace_set_syscall_info_exit(struct task_struct *child, struct pt_regs *regs, |
| 1076 | struct ptrace_syscall_info *info) |
| 1077 | { |
| 1078 | long rval = info->exit.rval; |
| 1079 | |
| 1080 | /* |
| 1081 | * Check that the return value specified in info->exit.rval |
| 1082 | * is either a value of type "long" or a sign-extended value |
| 1083 | * of type "long". |
| 1084 | */ |
| 1085 | if (rval != info->exit.rval) |
| 1086 | return -ERANGE; |
| 1087 | |
| 1088 | if (info->exit.is_error) |
| 1089 | syscall_set_return_value(task: child, regs, error: rval, val: 0); |
| 1090 | else |
| 1091 | syscall_set_return_value(task: child, regs, error: 0, val: rval); |
| 1092 | |
| 1093 | return 0; |
| 1094 | } |
| 1095 | |
| 1096 | static int |
| 1097 | ptrace_set_syscall_info(struct task_struct *child, unsigned long user_size, |
| 1098 | const void __user *datavp) |
| 1099 | { |
| 1100 | struct pt_regs *regs = task_pt_regs(child); |
| 1101 | struct ptrace_syscall_info info; |
| 1102 | |
| 1103 | if (user_size < sizeof(info)) |
| 1104 | return -EINVAL; |
| 1105 | |
| 1106 | /* |
| 1107 | * The compatibility is tracked by info.op and info.flags: if user-space |
| 1108 | * does not instruct us to use unknown extra bits from future versions |
| 1109 | * of ptrace_syscall_info, we are not going to read them either. |
| 1110 | */ |
| 1111 | if (copy_from_user(to: &info, from: datavp, n: sizeof(info))) |
| 1112 | return -EFAULT; |
| 1113 | |
| 1114 | /* Reserved for future use. */ |
| 1115 | if (info.flags || info.reserved) |
| 1116 | return -EINVAL; |
| 1117 | |
| 1118 | /* Changing the type of the system call stop is not supported yet. */ |
| 1119 | if (ptrace_get_syscall_info_op(child) != info.op) |
| 1120 | return -EINVAL; |
| 1121 | |
| 1122 | switch (info.op) { |
| 1123 | case PTRACE_SYSCALL_INFO_ENTRY: |
| 1124 | return ptrace_set_syscall_info_entry(child, regs, info: &info); |
| 1125 | case PTRACE_SYSCALL_INFO_EXIT: |
| 1126 | return ptrace_set_syscall_info_exit(child, regs, info: &info); |
| 1127 | case PTRACE_SYSCALL_INFO_SECCOMP: |
| 1128 | return ptrace_set_syscall_info_seccomp(child, regs, info: &info); |
| 1129 | default: |
| 1130 | /* Other types of system call stops are not supported yet. */ |
| 1131 | return -EINVAL; |
| 1132 | } |
| 1133 | } |
| 1134 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ |
| 1135 | |
| 1136 | int ptrace_request(struct task_struct *child, long request, |
| 1137 | unsigned long addr, unsigned long data) |
| 1138 | { |
| 1139 | bool seized = child->ptrace & PT_SEIZED; |
| 1140 | int ret = -EIO; |
| 1141 | kernel_siginfo_t siginfo, *si; |
| 1142 | void __user *datavp = (void __user *) data; |
| 1143 | unsigned long __user *datalp = datavp; |
| 1144 | unsigned long flags; |
| 1145 | |
| 1146 | switch (request) { |
| 1147 | case PTRACE_PEEKTEXT: |
| 1148 | case PTRACE_PEEKDATA: |
| 1149 | return generic_ptrace_peekdata(tsk: child, addr, data); |
| 1150 | case PTRACE_POKETEXT: |
| 1151 | case PTRACE_POKEDATA: |
| 1152 | return generic_ptrace_pokedata(tsk: child, addr, data); |
| 1153 | |
| 1154 | #ifdef PTRACE_OLDSETOPTIONS |
| 1155 | case PTRACE_OLDSETOPTIONS: |
| 1156 | #endif |
| 1157 | case PTRACE_SETOPTIONS: |
| 1158 | ret = ptrace_setoptions(child, data); |
| 1159 | break; |
| 1160 | case PTRACE_GETEVENTMSG: |
| 1161 | ret = put_user(child->ptrace_message, datalp); |
| 1162 | break; |
| 1163 | |
| 1164 | case PTRACE_PEEKSIGINFO: |
| 1165 | ret = ptrace_peek_siginfo(child, addr, data); |
| 1166 | break; |
| 1167 | |
| 1168 | case PTRACE_GETSIGINFO: |
| 1169 | ret = ptrace_getsiginfo(child, info: &siginfo); |
| 1170 | if (!ret) |
| 1171 | ret = copy_siginfo_to_user(to: datavp, from: &siginfo); |
| 1172 | break; |
| 1173 | |
| 1174 | case PTRACE_SETSIGINFO: |
| 1175 | ret = copy_siginfo_from_user(to: &siginfo, from: datavp); |
| 1176 | if (!ret) |
| 1177 | ret = ptrace_setsiginfo(child, info: &siginfo); |
| 1178 | break; |
| 1179 | |
| 1180 | case PTRACE_GETSIGMASK: { |
| 1181 | sigset_t *mask; |
| 1182 | |
| 1183 | if (addr != sizeof(sigset_t)) { |
| 1184 | ret = -EINVAL; |
| 1185 | break; |
| 1186 | } |
| 1187 | |
| 1188 | if (test_tsk_restore_sigmask(task: child)) |
| 1189 | mask = &child->saved_sigmask; |
| 1190 | else |
| 1191 | mask = &child->blocked; |
| 1192 | |
| 1193 | if (copy_to_user(to: datavp, from: mask, n: sizeof(sigset_t))) |
| 1194 | ret = -EFAULT; |
| 1195 | else |
| 1196 | ret = 0; |
| 1197 | |
| 1198 | break; |
| 1199 | } |
| 1200 | |
| 1201 | case PTRACE_SETSIGMASK: { |
| 1202 | sigset_t new_set; |
| 1203 | |
| 1204 | if (addr != sizeof(sigset_t)) { |
| 1205 | ret = -EINVAL; |
| 1206 | break; |
| 1207 | } |
| 1208 | |
| 1209 | if (copy_from_user(to: &new_set, from: datavp, n: sizeof(sigset_t))) { |
| 1210 | ret = -EFAULT; |
| 1211 | break; |
| 1212 | } |
| 1213 | |
| 1214 | sigdelsetmask(set: &new_set, sigmask(SIGKILL)|sigmask(SIGSTOP)); |
| 1215 | |
| 1216 | /* |
| 1217 | * Every thread does recalc_sigpending() after resume, so |
| 1218 | * retarget_shared_pending() and recalc_sigpending() are not |
| 1219 | * called here. |
| 1220 | */ |
| 1221 | spin_lock_irq(lock: &child->sighand->siglock); |
| 1222 | child->blocked = new_set; |
| 1223 | spin_unlock_irq(lock: &child->sighand->siglock); |
| 1224 | |
| 1225 | clear_tsk_restore_sigmask(task: child); |
| 1226 | |
| 1227 | ret = 0; |
| 1228 | break; |
| 1229 | } |
| 1230 | |
| 1231 | case PTRACE_INTERRUPT: |
| 1232 | /* |
| 1233 | * Stop tracee without any side-effect on signal or job |
| 1234 | * control. At least one trap is guaranteed to happen |
| 1235 | * after this request. If @child is already trapped, the |
| 1236 | * current trap is not disturbed and another trap will |
| 1237 | * happen after the current trap is ended with PTRACE_CONT. |
| 1238 | * |
| 1239 | * The actual trap might not be PTRACE_EVENT_STOP trap but |
| 1240 | * the pending condition is cleared regardless. |
| 1241 | */ |
| 1242 | if (unlikely(!seized || !lock_task_sighand(child, &flags))) |
| 1243 | break; |
| 1244 | |
| 1245 | /* |
| 1246 | * INTERRUPT doesn't disturb existing trap sans one |
| 1247 | * exception. If ptracer issued LISTEN for the current |
| 1248 | * STOP, this INTERRUPT should clear LISTEN and re-trap |
| 1249 | * tracee into STOP. |
| 1250 | */ |
| 1251 | if (likely(task_set_jobctl_pending(child, JOBCTL_TRAP_STOP))) |
| 1252 | ptrace_signal_wake_up(t: child, resume: child->jobctl & JOBCTL_LISTENING); |
| 1253 | |
| 1254 | unlock_task_sighand(task: child, flags: &flags); |
| 1255 | ret = 0; |
| 1256 | break; |
| 1257 | |
| 1258 | case PTRACE_LISTEN: |
| 1259 | /* |
| 1260 | * Listen for events. Tracee must be in STOP. It's not |
| 1261 | * resumed per-se but is not considered to be in TRACED by |
| 1262 | * wait(2) or ptrace(2). If an async event (e.g. group |
| 1263 | * stop state change) happens, tracee will enter STOP trap |
| 1264 | * again. Alternatively, ptracer can issue INTERRUPT to |
| 1265 | * finish listening and re-trap tracee into STOP. |
| 1266 | */ |
| 1267 | if (unlikely(!seized || !lock_task_sighand(child, &flags))) |
| 1268 | break; |
| 1269 | |
| 1270 | si = child->last_siginfo; |
| 1271 | if (likely(si && (si->si_code >> 8) == PTRACE_EVENT_STOP)) { |
| 1272 | child->jobctl |= JOBCTL_LISTENING; |
| 1273 | /* |
| 1274 | * If NOTIFY is set, it means event happened between |
| 1275 | * start of this trap and now. Trigger re-trap. |
| 1276 | */ |
| 1277 | if (child->jobctl & JOBCTL_TRAP_NOTIFY) |
| 1278 | ptrace_signal_wake_up(t: child, resume: true); |
| 1279 | ret = 0; |
| 1280 | } |
| 1281 | unlock_task_sighand(task: child, flags: &flags); |
| 1282 | break; |
| 1283 | |
| 1284 | case PTRACE_DETACH: /* detach a process that was attached. */ |
| 1285 | ret = ptrace_detach(child, data); |
| 1286 | break; |
| 1287 | |
| 1288 | #ifdef CONFIG_BINFMT_ELF_FDPIC |
| 1289 | case PTRACE_GETFDPIC: { |
| 1290 | struct mm_struct *mm = get_task_mm(child); |
| 1291 | unsigned long tmp = 0; |
| 1292 | |
| 1293 | ret = -ESRCH; |
| 1294 | if (!mm) |
| 1295 | break; |
| 1296 | |
| 1297 | switch (addr) { |
| 1298 | case PTRACE_GETFDPIC_EXEC: |
| 1299 | tmp = mm->context.exec_fdpic_loadmap; |
| 1300 | break; |
| 1301 | case PTRACE_GETFDPIC_INTERP: |
| 1302 | tmp = mm->context.interp_fdpic_loadmap; |
| 1303 | break; |
| 1304 | default: |
| 1305 | break; |
| 1306 | } |
| 1307 | mmput(mm); |
| 1308 | |
| 1309 | ret = put_user(tmp, datalp); |
| 1310 | break; |
| 1311 | } |
| 1312 | #endif |
| 1313 | |
| 1314 | case PTRACE_SINGLESTEP: |
| 1315 | #ifdef PTRACE_SINGLEBLOCK |
| 1316 | case PTRACE_SINGLEBLOCK: |
| 1317 | #endif |
| 1318 | #ifdef PTRACE_SYSEMU |
| 1319 | case PTRACE_SYSEMU: |
| 1320 | case PTRACE_SYSEMU_SINGLESTEP: |
| 1321 | #endif |
| 1322 | case PTRACE_SYSCALL: |
| 1323 | case PTRACE_CONT: |
| 1324 | return ptrace_resume(child, request, data); |
| 1325 | |
| 1326 | case PTRACE_KILL: |
| 1327 | send_sig_info(SIGKILL, SEND_SIG_NOINFO, child); |
| 1328 | return 0; |
| 1329 | |
| 1330 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
| 1331 | case PTRACE_GETREGSET: |
| 1332 | case PTRACE_SETREGSET: { |
| 1333 | struct iovec kiov; |
| 1334 | struct iovec __user *uiov = datavp; |
| 1335 | |
| 1336 | if (!access_ok(uiov, sizeof(*uiov))) |
| 1337 | return -EFAULT; |
| 1338 | |
| 1339 | if (__get_user(kiov.iov_base, &uiov->iov_base) || |
| 1340 | __get_user(kiov.iov_len, &uiov->iov_len)) |
| 1341 | return -EFAULT; |
| 1342 | |
| 1343 | ret = ptrace_regset(task: child, req: request, type: addr, kiov: &kiov); |
| 1344 | if (!ret) |
| 1345 | ret = __put_user(kiov.iov_len, &uiov->iov_len); |
| 1346 | break; |
| 1347 | } |
| 1348 | |
| 1349 | case PTRACE_GET_SYSCALL_INFO: |
| 1350 | ret = ptrace_get_syscall_info(child, user_size: addr, datavp); |
| 1351 | break; |
| 1352 | |
| 1353 | case PTRACE_SET_SYSCALL_INFO: |
| 1354 | ret = ptrace_set_syscall_info(child, user_size: addr, datavp); |
| 1355 | break; |
| 1356 | #endif |
| 1357 | |
| 1358 | case PTRACE_SECCOMP_GET_FILTER: |
| 1359 | ret = seccomp_get_filter(task: child, n: addr, data: datavp); |
| 1360 | break; |
| 1361 | |
| 1362 | case PTRACE_SECCOMP_GET_METADATA: |
| 1363 | ret = seccomp_get_metadata(task: child, filter_off: addr, data: datavp); |
| 1364 | break; |
| 1365 | |
| 1366 | #ifdef CONFIG_RSEQ |
| 1367 | case PTRACE_GET_RSEQ_CONFIGURATION: |
| 1368 | ret = ptrace_get_rseq_configuration(task: child, size: addr, data: datavp); |
| 1369 | break; |
| 1370 | #endif |
| 1371 | |
| 1372 | case PTRACE_SET_SYSCALL_USER_DISPATCH_CONFIG: |
| 1373 | ret = syscall_user_dispatch_set_config(task: child, size: addr, data: datavp); |
| 1374 | break; |
| 1375 | |
| 1376 | case PTRACE_GET_SYSCALL_USER_DISPATCH_CONFIG: |
| 1377 | ret = syscall_user_dispatch_get_config(task: child, size: addr, data: datavp); |
| 1378 | break; |
| 1379 | |
| 1380 | default: |
| 1381 | break; |
| 1382 | } |
| 1383 | |
| 1384 | return ret; |
| 1385 | } |
| 1386 | |
| 1387 | SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr, |
| 1388 | unsigned long, data) |
| 1389 | { |
| 1390 | struct task_struct *child; |
| 1391 | long ret; |
| 1392 | |
| 1393 | if (request == PTRACE_TRACEME) { |
| 1394 | ret = ptrace_traceme(); |
| 1395 | goto out; |
| 1396 | } |
| 1397 | |
| 1398 | child = find_get_task_by_vpid(nr: pid); |
| 1399 | if (!child) { |
| 1400 | ret = -ESRCH; |
| 1401 | goto out; |
| 1402 | } |
| 1403 | |
| 1404 | if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) { |
| 1405 | ret = ptrace_attach(task: child, request, addr, flags: data); |
| 1406 | goto out_put_task_struct; |
| 1407 | } |
| 1408 | |
| 1409 | ret = ptrace_check_attach(child, ignore_state: request == PTRACE_KILL || |
| 1410 | request == PTRACE_INTERRUPT); |
| 1411 | if (ret < 0) |
| 1412 | goto out_put_task_struct; |
| 1413 | |
| 1414 | ret = arch_ptrace(child, request, addr, data); |
| 1415 | if (ret || request != PTRACE_DETACH) |
| 1416 | ptrace_unfreeze_traced(task: child); |
| 1417 | |
| 1418 | out_put_task_struct: |
| 1419 | put_task_struct(t: child); |
| 1420 | out: |
| 1421 | return ret; |
| 1422 | } |
| 1423 | |
| 1424 | int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr, |
| 1425 | unsigned long data) |
| 1426 | { |
| 1427 | unsigned long tmp; |
| 1428 | int copied; |
| 1429 | |
| 1430 | copied = ptrace_access_vm(tsk, addr, buf: &tmp, len: sizeof(tmp), gup_flags: FOLL_FORCE); |
| 1431 | if (copied != sizeof(tmp)) |
| 1432 | return -EIO; |
| 1433 | return put_user(tmp, (unsigned long __user *)data); |
| 1434 | } |
| 1435 | |
| 1436 | int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr, |
| 1437 | unsigned long data) |
| 1438 | { |
| 1439 | int copied; |
| 1440 | |
| 1441 | copied = ptrace_access_vm(tsk, addr, buf: &data, len: sizeof(data), |
| 1442 | gup_flags: FOLL_FORCE | FOLL_WRITE); |
| 1443 | return (copied == sizeof(data)) ? 0 : -EIO; |
| 1444 | } |
| 1445 | |
| 1446 | #if defined CONFIG_COMPAT |
| 1447 | |
| 1448 | int compat_ptrace_request(struct task_struct *child, compat_long_t request, |
| 1449 | compat_ulong_t addr, compat_ulong_t data) |
| 1450 | { |
| 1451 | compat_ulong_t __user *datap = compat_ptr(uptr: data); |
| 1452 | compat_ulong_t word; |
| 1453 | kernel_siginfo_t siginfo; |
| 1454 | int ret; |
| 1455 | |
| 1456 | switch (request) { |
| 1457 | case PTRACE_PEEKTEXT: |
| 1458 | case PTRACE_PEEKDATA: |
| 1459 | ret = ptrace_access_vm(tsk: child, addr, buf: &word, len: sizeof(word), |
| 1460 | gup_flags: FOLL_FORCE); |
| 1461 | if (ret != sizeof(word)) |
| 1462 | ret = -EIO; |
| 1463 | else |
| 1464 | ret = put_user(word, datap); |
| 1465 | break; |
| 1466 | |
| 1467 | case PTRACE_POKETEXT: |
| 1468 | case PTRACE_POKEDATA: |
| 1469 | ret = ptrace_access_vm(tsk: child, addr, buf: &data, len: sizeof(data), |
| 1470 | gup_flags: FOLL_FORCE | FOLL_WRITE); |
| 1471 | ret = (ret != sizeof(data) ? -EIO : 0); |
| 1472 | break; |
| 1473 | |
| 1474 | case PTRACE_GETEVENTMSG: |
| 1475 | ret = put_user((compat_ulong_t) child->ptrace_message, datap); |
| 1476 | break; |
| 1477 | |
| 1478 | case PTRACE_GETSIGINFO: |
| 1479 | ret = ptrace_getsiginfo(child, info: &siginfo); |
| 1480 | if (!ret) |
| 1481 | ret = copy_siginfo_to_user32( |
| 1482 | to: (struct compat_siginfo __user *) datap, |
| 1483 | from: &siginfo); |
| 1484 | break; |
| 1485 | |
| 1486 | case PTRACE_SETSIGINFO: |
| 1487 | ret = copy_siginfo_from_user32( |
| 1488 | to: &siginfo, from: (struct compat_siginfo __user *) datap); |
| 1489 | if (!ret) |
| 1490 | ret = ptrace_setsiginfo(child, info: &siginfo); |
| 1491 | break; |
| 1492 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
| 1493 | case PTRACE_GETREGSET: |
| 1494 | case PTRACE_SETREGSET: |
| 1495 | { |
| 1496 | struct iovec kiov; |
| 1497 | struct compat_iovec __user *uiov = |
| 1498 | (struct compat_iovec __user *) datap; |
| 1499 | compat_uptr_t ptr; |
| 1500 | compat_size_t len; |
| 1501 | |
| 1502 | if (!access_ok(uiov, sizeof(*uiov))) |
| 1503 | return -EFAULT; |
| 1504 | |
| 1505 | if (__get_user(ptr, &uiov->iov_base) || |
| 1506 | __get_user(len, &uiov->iov_len)) |
| 1507 | return -EFAULT; |
| 1508 | |
| 1509 | kiov.iov_base = compat_ptr(uptr: ptr); |
| 1510 | kiov.iov_len = len; |
| 1511 | |
| 1512 | ret = ptrace_regset(task: child, req: request, type: addr, kiov: &kiov); |
| 1513 | if (!ret) |
| 1514 | ret = __put_user(kiov.iov_len, &uiov->iov_len); |
| 1515 | break; |
| 1516 | } |
| 1517 | #endif |
| 1518 | |
| 1519 | default: |
| 1520 | ret = ptrace_request(child, request, addr, data); |
| 1521 | } |
| 1522 | |
| 1523 | return ret; |
| 1524 | } |
| 1525 | |
| 1526 | COMPAT_SYSCALL_DEFINE4(ptrace, compat_long_t, request, compat_long_t, pid, |
| 1527 | compat_long_t, addr, compat_long_t, data) |
| 1528 | { |
| 1529 | struct task_struct *child; |
| 1530 | long ret; |
| 1531 | |
| 1532 | if (request == PTRACE_TRACEME) { |
| 1533 | ret = ptrace_traceme(); |
| 1534 | goto out; |
| 1535 | } |
| 1536 | |
| 1537 | child = find_get_task_by_vpid(nr: pid); |
| 1538 | if (!child) { |
| 1539 | ret = -ESRCH; |
| 1540 | goto out; |
| 1541 | } |
| 1542 | |
| 1543 | if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) { |
| 1544 | ret = ptrace_attach(task: child, request, addr, flags: data); |
| 1545 | goto out_put_task_struct; |
| 1546 | } |
| 1547 | |
| 1548 | ret = ptrace_check_attach(child, ignore_state: request == PTRACE_KILL || |
| 1549 | request == PTRACE_INTERRUPT); |
| 1550 | if (!ret) { |
| 1551 | ret = compat_arch_ptrace(child, request, addr, data); |
| 1552 | if (ret || request != PTRACE_DETACH) |
| 1553 | ptrace_unfreeze_traced(task: child); |
| 1554 | } |
| 1555 | |
| 1556 | out_put_task_struct: |
| 1557 | put_task_struct(t: child); |
| 1558 | out: |
| 1559 | return ret; |
| 1560 | } |
| 1561 | #endif /* CONFIG_COMPAT */ |
| 1562 | |