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