include/linux/ptrace.h

   1 /* SPDX-License-Identifier: GPL-2.0 */
   2 #ifndef _LINUX_PTRACE_H
   3 #define _LINUX_PTRACE_H
   4
   5 #include <linux/compiler.h>             /* For unlikely.  */
   6 #include <linux/sched.h>                /* For struct task_struct.  */
   7 #include <linux/sched/signal.h>         /* For send_sig(), same_thread_group(), etc. */
   8 #include <linux/err.h>                  /* for IS_ERR_VALUE */
   9 #include <linux/bug.h>                  /* For BUG_ON.  */
  10 #include <linux/pid_namespace.h>        /* For task_active_pid_ns.  */
  11 #include <uapi/linux/ptrace.h>
  12
  13 extern int ptrace_access_vm(struct task_struct *tsk, unsigned long addr,
  14                             void *buf, int len, unsigned int gup_flags);
  15
  16 /*
  17  * Ptrace flags
  18  *
  19  * The owner ship rules for task->ptrace which holds the ptrace
  20  * flags is simple.  When a task is running it owns it's task->ptrace
  21  * flags.  When the a task is stopped the ptracer owns task->ptrace.
  22  */
  23
  24 #define PT_SEIZED       0x00010000      /* SEIZE used, enable new behavior */
  25 #define PT_PTRACED      0x00000001
  26 #define PT_DTRACE       0x00000002      /* delayed trace (used on m68k, i386) */
  27
  28 #define PT_OPT_FLAG_SHIFT       3
  29 /* PT_TRACE_* event enable flags */
  30 #define PT_EVENT_FLAG(event)    (1 << (PT_OPT_FLAG_SHIFT + (event)))
  31 #define PT_TRACESYSGOOD         PT_EVENT_FLAG(0)
  32 #define PT_TRACE_FORK           PT_EVENT_FLAG(PTRACE_EVENT_FORK)
  33 #define PT_TRACE_VFORK          PT_EVENT_FLAG(PTRACE_EVENT_VFORK)
  34 #define PT_TRACE_CLONE          PT_EVENT_FLAG(PTRACE_EVENT_CLONE)
  35 #define PT_TRACE_EXEC           PT_EVENT_FLAG(PTRACE_EVENT_EXEC)
  36 #define PT_TRACE_VFORK_DONE     PT_EVENT_FLAG(PTRACE_EVENT_VFORK_DONE)
  37 #define PT_TRACE_EXIT           PT_EVENT_FLAG(PTRACE_EVENT_EXIT)
  38 #define PT_TRACE_SECCOMP        PT_EVENT_FLAG(PTRACE_EVENT_SECCOMP)
  39
  40 #define PT_EXITKILL             (PTRACE_O_EXITKILL << PT_OPT_FLAG_SHIFT)
  41 #define PT_SUSPEND_SECCOMP      (PTRACE_O_SUSPEND_SECCOMP << PT_OPT_FLAG_SHIFT)
  42
  43 /* single stepping state bits (used on ARM and PA-RISC) */
  44 #define PT_SINGLESTEP_BIT       31
  45 #define PT_SINGLESTEP           (1<<PT_SINGLESTEP_BIT)
  46 #define PT_BLOCKSTEP_BIT        30
  47 #define PT_BLOCKSTEP            (1<<PT_BLOCKSTEP_BIT)
  48
  49 extern long arch_ptrace(struct task_struct *child, long request,
  50                         unsigned long addr, unsigned long data);
  51 extern int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len);
  52 extern int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len);
  53 extern void ptrace_disable(struct task_struct *);
  54 extern int ptrace_request(struct task_struct *child, long request,
  55                           unsigned long addr, unsigned long data);
  56 extern void ptrace_notify(int exit_code);
  57 extern void __ptrace_link(struct task_struct *child,
  58                           struct task_struct *new_parent,
  59                           const struct cred *ptracer_cred);
  60 extern void __ptrace_unlink(struct task_struct *child);
  61 extern void exit_ptrace(struct task_struct *tracer, struct list_head *dead);
  62 #define PTRACE_MODE_READ        0x01
  63 #define PTRACE_MODE_ATTACH      0x02
  64 #define PTRACE_MODE_NOAUDIT     0x04
  65 #define PTRACE_MODE_FSCREDS     0x08
  66 #define PTRACE_MODE_REALCREDS   0x10
  67 #define PTRACE_MODE_SCHED       0x20
  68 #define PTRACE_MODE_IBPB        0x40
  69
  70 /* shorthands for READ/ATTACH and FSCREDS/REALCREDS combinations */
  71 #define PTRACE_MODE_READ_FSCREDS (PTRACE_MODE_READ | PTRACE_MODE_FSCREDS)
  72 #define PTRACE_MODE_READ_REALCREDS (PTRACE_MODE_READ | PTRACE_MODE_REALCREDS)
  73 #define PTRACE_MODE_ATTACH_FSCREDS (PTRACE_MODE_ATTACH | PTRACE_MODE_FSCREDS)
  74 #define PTRACE_MODE_ATTACH_REALCREDS (PTRACE_MODE_ATTACH | PTRACE_MODE_REALCREDS)
  75 #define PTRACE_MODE_SPEC_IBPB (PTRACE_MODE_ATTACH_REALCREDS | PTRACE_MODE_IBPB)
  76
  77 /**
  78  * ptrace_may_access - check whether the caller is permitted to access
  79  * a target task.
  80  * @task: target task
  81  * @mode: selects type of access and caller credentials
  82  *
  83  * Returns true on success, false on denial.
  84  *
  85  * One of the flags PTRACE_MODE_FSCREDS and PTRACE_MODE_REALCREDS must
  86  * be set in @mode to specify whether the access was requested through
  87  * a filesystem syscall (should use effective capabilities and fsuid
  88  * of the caller) or through an explicit syscall such as
  89  * process_vm_writev or ptrace (and should use the real credentials).
  90  */
  91 extern bool ptrace_may_access(struct task_struct *task, unsigned int mode);
  92
  93 /**
  94  * ptrace_may_access - check whether the caller is permitted to access
  95  * a target task.
  96  * @task: target task
  97  * @mode: selects type of access and caller credentials
  98  *
  99  * Returns true on success, false on denial.
 100  *
 101  * Similar to ptrace_may_access(). Only to be called from context switch
 102  * code. Does not call into audit and the regular LSM hooks due to locking
 103  * constraints.
 104  */
 105 extern bool ptrace_may_access_sched(struct task_struct *task, unsigned int mode);
 106
 107 static inline int ptrace_reparented(struct task_struct *child)
 108 {
 109         return !same_thread_group(child->real_parent, child->parent);
 110 }
 111
 112 static inline void ptrace_unlink(struct task_struct *child)
 113 {
 114         if (unlikely(child->ptrace))
 115                 __ptrace_unlink(child);
 116 }
 117
 118 int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr,
 119                             unsigned long data);
 120 int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr,
 121                             unsigned long data);
 122
 123 /**
 124  * ptrace_parent - return the task that is tracing the given task
 125  * @task: task to consider
 126  *
 127  * Returns %NULL if no one is tracing @task, or the &struct task_struct
 128  * pointer to its tracer.
 129  *
 130  * Must called under rcu_read_lock().  The pointer returned might be kept
 131  * live only by RCU.  During exec, this may be called with task_lock() held
 132  * on @task, still held from when check_unsafe_exec() was called.
 133  */
 134 static inline struct task_struct *ptrace_parent(struct task_struct *task)
 135 {
 136         if (unlikely(task->ptrace))
 137                 return rcu_dereference(task->parent);
 138         return NULL;
 139 }
 140
 141 /**
 142  * ptrace_event_enabled - test whether a ptrace event is enabled
 143  * @task: ptracee of interest
 144  * @event: %PTRACE_EVENT_* to test
 145  *
 146  * Test whether @event is enabled for ptracee @task.
 147  *
 148  * Returns %true if @event is enabled, %false otherwise.
 149  */
 150 static inline bool ptrace_event_enabled(struct task_struct *task, int event)
 151 {
 152         return task->ptrace & PT_EVENT_FLAG(event);
 153 }
 154
 155 /**
 156  * ptrace_event - possibly stop for a ptrace event notification
 157  * @event:      %PTRACE_EVENT_* value to report
 158  * @message:    value for %PTRACE_GETEVENTMSG to return
 159  *
 160  * Check whether @event is enabled and, if so, report @event and @message
 161  * to the ptrace parent.
 162  *
 163  * Called without locks.
 164  */
 165 static inline void ptrace_event(int event, unsigned long message)
 166 {
 167         if (unlikely(ptrace_event_enabled(current, event))) {
 168                 current->ptrace_message = message;
 169                 ptrace_notify((event << 8) | SIGTRAP);
 170         } else if (event == PTRACE_EVENT_EXEC) {
 171                 /* legacy EXEC report via SIGTRAP */
 172                 if ((current->ptrace & (PT_PTRACED|PT_SEIZED)) == PT_PTRACED)
 173                         send_sig(SIGTRAP, current, 0);
 174         }
 175 }
 176
 177 /**
 178  * ptrace_event_pid - possibly stop for a ptrace event notification
 179  * @event:      %PTRACE_EVENT_* value to report
 180  * @pid:        process identifier for %PTRACE_GETEVENTMSG to return
 181  *
 182  * Check whether @event is enabled and, if so, report @event and @pid
 183  * to the ptrace parent.  @pid is reported as the pid_t seen from the
 184  * the ptrace parent's pid namespace.
 185  *
 186  * Called without locks.
 187  */
 188 static inline void ptrace_event_pid(int event, struct pid *pid)
 189 {
 190         /*
 191          * FIXME: There's a potential race if a ptracer in a different pid
 192          * namespace than parent attaches between computing message below and
 193          * when we acquire tasklist_lock in ptrace_stop().  If this happens,
 194          * the ptracer will get a bogus pid from PTRACE_GETEVENTMSG.
 195          */
 196         unsigned long message = 0;
 197         struct pid_namespace *ns;
 198
 199         rcu_read_lock();
 200         ns = task_active_pid_ns(rcu_dereference(current->parent));
 201         if (ns)
 202                 message = pid_nr_ns(pid, ns);
 203         rcu_read_unlock();
 204
 205         ptrace_event(event, message);
 206 }
 207
 208 /**
 209  * ptrace_init_task - initialize ptrace state for a new child
 210  * @child:              new child task
 211  * @ptrace:             true if child should be ptrace'd by parent's tracer
 212  *
 213  * This is called immediately after adding @child to its parent's children
 214  * list.  @ptrace is false in the normal case, and true to ptrace @child.
 215  *
 216  * Called with current's siglock and write_lock_irq(&tasklist_lock) held.
 217  */
 218 static inline void ptrace_init_task(struct task_struct *child, bool ptrace)
 219 {
 220         INIT_LIST_HEAD(&child->ptrace_entry);
 221         INIT_LIST_HEAD(&child->ptraced);
 222         child->jobctl = 0;
 223         child->ptrace = 0;
 224         child->parent = child->real_parent;
 225
 226         if (unlikely(ptrace) && current->ptrace) {
 227                 child->ptrace = current->ptrace;
 228                 __ptrace_link(child, current->parent, current->ptracer_cred);
 229
 230                 if (child->ptrace & PT_SEIZED)
 231                         task_set_jobctl_pending(child, JOBCTL_TRAP_STOP);
 232                 else
 233                         sigaddset(&child->pending.signal, SIGSTOP);
 234         }
 235         else
 236                 child->ptracer_cred = NULL;
 237 }
 238
 239 /**
 240  * ptrace_release_task - final ptrace-related cleanup of a zombie being reaped
 241  * @task:       task in %EXIT_DEAD state
 242  *
 243  * Called with write_lock(&tasklist_lock) held.
 244  */
 245 static inline void ptrace_release_task(struct task_struct *task)
 246 {
 247         BUG_ON(!list_empty(&task->ptraced));
 248         ptrace_unlink(task);
 249         BUG_ON(!list_empty(&task->ptrace_entry));
 250 }
 251
 252 #ifndef force_successful_syscall_return
 253 /*
 254  * System call handlers that, upon successful completion, need to return a
 255  * negative value should call force_successful_syscall_return() right before
 256  * returning.  On architectures where the syscall convention provides for a
 257  * separate error flag (e.g., alpha, ia64, ppc{,64}, sparc{,64}, possibly
 258  * others), this macro can be used to ensure that the error flag will not get
 259  * set.  On architectures which do not support a separate error flag, the macro
 260  * is a no-op and the spurious error condition needs to be filtered out by some
 261  * other means (e.g., in user-level, by passing an extra argument to the
 262  * syscall handler, or something along those lines).
 263  */
 264 #define force_successful_syscall_return() do { } while (0)
 265 #endif
 266
 267 #ifndef is_syscall_success
 268 /*
 269  * On most systems we can tell if a syscall is a success based on if the retval
 270  * is an error value.  On some systems like ia64 and powerpc they have different
 271  * indicators of success/failure and must define their own.
 272  */
 273 #define is_syscall_success(regs) (!IS_ERR_VALUE((unsigned long)(regs_return_value(regs))))
 274 #endif
 275
 276 /*
 277  * <asm/ptrace.h> should define the following things inside #ifdef __KERNEL__.
 278  *
 279  * These do-nothing inlines are used when the arch does not
 280  * implement single-step.  The kerneldoc comments are here
 281  * to document the interface for all arch definitions.
 282  */
 283
 284 #ifndef arch_has_single_step
 285 /**
 286  * arch_has_single_step - does this CPU support user-mode single-step?
 287  *
 288  * If this is defined, then there must be function declarations or
 289  * inlines for user_enable_single_step() and user_disable_single_step().
 290  * arch_has_single_step() should evaluate to nonzero iff the machine
 291  * supports instruction single-step for user mode.
 292  * It can be a constant or it can test a CPU feature bit.
 293  */
 294 #define arch_has_single_step()          (0)
 295
 296 /**
 297  * user_enable_single_step - single-step in user-mode task
 298  * @task: either current or a task stopped in %TASK_TRACED
 299  *
 300  * This can only be called when arch_has_single_step() has returned nonzero.
 301  * Set @task so that when it returns to user mode, it will trap after the
 302  * next single instruction executes.  If arch_has_block_step() is defined,
 303  * this must clear the effects of user_enable_block_step() too.
 304  */
 305 static inline void user_enable_single_step(struct task_struct *task)
 306 {
 307         BUG();                  /* This can never be called.  */
 308 }
 309
 310 /**
 311  * user_disable_single_step - cancel user-mode single-step
 312  * @task: either current or a task stopped in %TASK_TRACED
 313  *
 314  * Clear @task of the effects of user_enable_single_step() and
 315  * user_enable_block_step().  This can be called whether or not either
 316  * of those was ever called on @task, and even if arch_has_single_step()
 317  * returned zero.
 318  */
 319 static inline void user_disable_single_step(struct task_struct *task)
 320 {
 321 }
 322 #else
 323 extern void user_enable_single_step(struct task_struct *);
 324 extern void user_disable_single_step(struct task_struct *);
 325 #endif  /* arch_has_single_step */
 326
 327 #ifndef arch_has_block_step
 328 /**
 329  * arch_has_block_step - does this CPU support user-mode block-step?
 330  *
 331  * If this is defined, then there must be a function declaration or inline
 332  * for user_enable_block_step(), and arch_has_single_step() must be defined
 333  * too.  arch_has_block_step() should evaluate to nonzero iff the machine
 334  * supports step-until-branch for user mode.  It can be a constant or it
 335  * can test a CPU feature bit.
 336  */
 337 #define arch_has_block_step()           (0)
 338
 339 /**
 340  * user_enable_block_step - step until branch in user-mode task
 341  * @task: either current or a task stopped in %TASK_TRACED
 342  *
 343  * This can only be called when arch_has_block_step() has returned nonzero,
 344  * and will never be called when single-instruction stepping is being used.
 345  * Set @task so that when it returns to user mode, it will trap after the
 346  * next branch or trap taken.
 347  */
 348 static inline void user_enable_block_step(struct task_struct *task)
 349 {
 350         BUG();                  /* This can never be called.  */
 351 }
 352 #else
 353 extern void user_enable_block_step(struct task_struct *);
 354 #endif  /* arch_has_block_step */
 355
 356 #ifdef ARCH_HAS_USER_SINGLE_STEP_REPORT
 357 extern void user_single_step_report(struct pt_regs *regs);
 358 #else
 359 static inline void user_single_step_report(struct pt_regs *regs)
 360 {
 361         kernel_siginfo_t info;
 362         clear_siginfo(&info);
 363         info.si_signo = SIGTRAP;
 364         info.si_errno = 0;
 365         info.si_code = SI_USER;
 366         info.si_pid = 0;
 367         info.si_uid = 0;
 368         force_sig_info(info.si_signo, &info, current);
 369 }
 370 #endif
 371
 372 #ifndef arch_ptrace_stop_needed
 373 /**
 374  * arch_ptrace_stop_needed - Decide whether arch_ptrace_stop() should be called
 375  * @code:       current->exit_code value ptrace will stop with
 376  * @info:       siginfo_t pointer (or %NULL) for signal ptrace will stop with
 377  *
 378  * This is called with the siglock held, to decide whether or not it's
 379  * necessary to release the siglock and call arch_ptrace_stop() with the
 380  * same @code and @info arguments.  It can be defined to a constant if
 381  * arch_ptrace_stop() is never required, or always is.  On machines where
 382  * this makes sense, it should be defined to a quick test to optimize out
 383  * calling arch_ptrace_stop() when it would be superfluous.  For example,
 384  * if the thread has not been back to user mode since the last stop, the
 385  * thread state might indicate that nothing needs to be done.
 386  *
 387  * This is guaranteed to be invoked once before a task stops for ptrace and
 388  * may include arch-specific operations necessary prior to a ptrace stop.
 389  */
 390 #define arch_ptrace_stop_needed(code, info)     (0)
 391 #endif
 392
 393 #ifndef arch_ptrace_stop
 394 /**
 395  * arch_ptrace_stop - Do machine-specific work before stopping for ptrace
 396  * @code:       current->exit_code value ptrace will stop with
 397  * @info:       siginfo_t pointer (or %NULL) for signal ptrace will stop with
 398  *
 399  * This is called with no locks held when arch_ptrace_stop_needed() has
 400  * just returned nonzero.  It is allowed to block, e.g. for user memory
 401  * access.  The arch can have machine-specific work to be done before
 402  * ptrace stops.  On ia64, register backing store gets written back to user
 403  * memory here.  Since this can be costly (requires dropping the siglock),
 404  * we only do it when the arch requires it for this particular stop, as
 405  * indicated by arch_ptrace_stop_needed().
 406  */
 407 #define arch_ptrace_stop(code, info)            do { } while (0)
 408 #endif
 409
 410 #ifndef current_pt_regs
 411 #define current_pt_regs() task_pt_regs(current)
 412 #endif
 413
 414 /*
 415  * unlike current_pt_regs(), this one is equal to task_pt_regs(current)
 416  * on *all* architectures; the only reason to have a per-arch definition
 417  * is optimisation.
 418  */
 419 #ifndef signal_pt_regs
 420 #define signal_pt_regs() task_pt_regs(current)
 421 #endif
 422
 423 #ifndef current_user_stack_pointer
 424 #define current_user_stack_pointer() user_stack_pointer(current_pt_regs())
 425 #endif
 426
 427 extern int task_current_syscall(struct task_struct *target, long *callno,
 428                                 unsigned long args[6], unsigned int maxargs,
 429                                 unsigned long *sp, unsigned long *pc);
 430
 431 #endif