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 #include <linux/seccomp.h>
13
14 /* Add sp to seccomp_data, as seccomp is user API, we don't want to modify it */
15 struct syscall_info {
16 __u64 sp;
17 struct seccomp_data data;
18 };
19
20 extern int ptrace_access_vm(struct task_struct *tsk, unsigned long addr,
21 void *buf, int len, unsigned int gup_flags);
22
23 /*
24 * Ptrace flags
25 *
26 * The owner ship rules for task->ptrace which holds the ptrace
27 * flags is simple. When a task is running it owns it's task->ptrace
28 * flags. When the a task is stopped the ptracer owns task->ptrace.
29 */
30
31 #define PT_SEIZED 0x00010000 /* SEIZE used, enable new behavior */
32 #define PT_PTRACED 0x00000001
33 #define PT_DTRACE 0x00000002 /* delayed trace (used on m68k, i386) */
34
35 #define PT_OPT_FLAG_SHIFT 3
36 /* PT_TRACE_* event enable flags */
37 #define PT_EVENT_FLAG(event) (1 << (PT_OPT_FLAG_SHIFT + (event)))
38 #define PT_TRACESYSGOOD PT_EVENT_FLAG(0)
39 #define PT_TRACE_FORK PT_EVENT_FLAG(PTRACE_EVENT_FORK)
40 #define PT_TRACE_VFORK PT_EVENT_FLAG(PTRACE_EVENT_VFORK)
41 #define PT_TRACE_CLONE PT_EVENT_FLAG(PTRACE_EVENT_CLONE)
42 #define PT_TRACE_EXEC PT_EVENT_FLAG(PTRACE_EVENT_EXEC)
43 #define PT_TRACE_VFORK_DONE PT_EVENT_FLAG(PTRACE_EVENT_VFORK_DONE)
44 #define PT_TRACE_EXIT PT_EVENT_FLAG(PTRACE_EVENT_EXIT)
45 #define PT_TRACE_SECCOMP PT_EVENT_FLAG(PTRACE_EVENT_SECCOMP)
46
47 #define PT_EXITKILL (PTRACE_O_EXITKILL << PT_OPT_FLAG_SHIFT)
48 #define PT_SUSPEND_SECCOMP (PTRACE_O_SUSPEND_SECCOMP << PT_OPT_FLAG_SHIFT)
49
50 /* single stepping state bits (used on ARM and PA-RISC) */
51 #define PT_SINGLESTEP_BIT 31
52 #define PT_SINGLESTEP (1<<PT_SINGLESTEP_BIT)
53 #define PT_BLOCKSTEP_BIT 30
54 #define PT_BLOCKSTEP (1<<PT_BLOCKSTEP_BIT)
55
56 extern long arch_ptrace(struct task_struct *child, long request,
57 unsigned long addr, unsigned long data);
58 extern int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len);
59 extern int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len);
60 extern void ptrace_disable(struct task_struct *);
61 extern int ptrace_request(struct task_struct *child, long request,
62 unsigned long addr, unsigned long data);
63 extern void ptrace_notify(int exit_code);
64 extern void __ptrace_link(struct task_struct *child,
65 struct task_struct *new_parent,
66 const struct cred *ptracer_cred);
67 extern void __ptrace_unlink(struct task_struct *child);
68 extern void exit_ptrace(struct task_struct *tracer, struct list_head *dead);
69 #define PTRACE_MODE_READ 0x01
70 #define PTRACE_MODE_ATTACH 0x02
71 #define PTRACE_MODE_NOAUDIT 0x04
72 #define PTRACE_MODE_FSCREDS 0x08
73 #define PTRACE_MODE_REALCREDS 0x10
74
75 /* shorthands for READ/ATTACH and FSCREDS/REALCREDS combinations */
76 #define PTRACE_MODE_READ_FSCREDS (PTRACE_MODE_READ | PTRACE_MODE_FSCREDS)
77 #define PTRACE_MODE_READ_REALCREDS (PTRACE_MODE_READ | PTRACE_MODE_REALCREDS)
78 #define PTRACE_MODE_ATTACH_FSCREDS (PTRACE_MODE_ATTACH | PTRACE_MODE_FSCREDS)
79 #define PTRACE_MODE_ATTACH_REALCREDS (PTRACE_MODE_ATTACH | PTRACE_MODE_REALCREDS)
80
81 /**
82 * ptrace_may_access - check whether the caller is permitted to access
83 * a target task.
84 * @task: target task
85 * @mode: selects type of access and caller credentials
86 *
87 * Returns true on success, false on denial.
88 *
89 * One of the flags PTRACE_MODE_FSCREDS and PTRACE_MODE_REALCREDS must
90 * be set in @mode to specify whether the access was requested through
91 * a filesystem syscall (should use effective capabilities and fsuid
92 * of the caller) or through an explicit syscall such as
93 * process_vm_writev or ptrace (and should use the real credentials).
94 */
95 extern bool ptrace_may_access(struct task_struct *task, unsigned int mode);
96
ptrace_reparented(struct task_struct * child)97 static inline int ptrace_reparented(struct task_struct *child)
98 {
99 return !same_thread_group(child->real_parent, child->parent);
100 }
101
ptrace_unlink(struct task_struct * child)102 static inline void ptrace_unlink(struct task_struct *child)
103 {
104 if (unlikely(child->ptrace))
105 __ptrace_unlink(child);
106 }
107
108 int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr,
109 unsigned long data);
110 int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr,
111 unsigned long data);
112
113 /**
114 * ptrace_parent - return the task that is tracing the given task
115 * @task: task to consider
116 *
117 * Returns %NULL if no one is tracing @task, or the &struct task_struct
118 * pointer to its tracer.
119 *
120 * Must called under rcu_read_lock(). The pointer returned might be kept
121 * live only by RCU. During exec, this may be called with task_lock() held
122 * on @task, still held from when check_unsafe_exec() was called.
123 */
ptrace_parent(struct task_struct * task)124 static inline struct task_struct *ptrace_parent(struct task_struct *task)
125 {
126 if (unlikely(task->ptrace))
127 return rcu_dereference(task->parent);
128 return NULL;
129 }
130
131 /**
132 * ptrace_event_enabled - test whether a ptrace event is enabled
133 * @task: ptracee of interest
134 * @event: %PTRACE_EVENT_* to test
135 *
136 * Test whether @event is enabled for ptracee @task.
137 *
138 * Returns %true if @event is enabled, %false otherwise.
139 */
ptrace_event_enabled(struct task_struct * task,int event)140 static inline bool ptrace_event_enabled(struct task_struct *task, int event)
141 {
142 return task->ptrace & PT_EVENT_FLAG(event);
143 }
144
145 /**
146 * ptrace_event - possibly stop for a ptrace event notification
147 * @event: %PTRACE_EVENT_* value to report
148 * @message: value for %PTRACE_GETEVENTMSG to return
149 *
150 * Check whether @event is enabled and, if so, report @event and @message
151 * to the ptrace parent.
152 *
153 * Called without locks.
154 */
ptrace_event(int event,unsigned long message)155 static inline void ptrace_event(int event, unsigned long message)
156 {
157 if (unlikely(ptrace_event_enabled(current, event))) {
158 current->ptrace_message = message;
159 ptrace_notify((event << 8) | SIGTRAP);
160 } else if (event == PTRACE_EVENT_EXEC) {
161 /* legacy EXEC report via SIGTRAP */
162 if ((current->ptrace & (PT_PTRACED|PT_SEIZED)) == PT_PTRACED)
163 send_sig(SIGTRAP, current, 0);
164 }
165 }
166
167 /**
168 * ptrace_event_pid - possibly stop for a ptrace event notification
169 * @event: %PTRACE_EVENT_* value to report
170 * @pid: process identifier for %PTRACE_GETEVENTMSG to return
171 *
172 * Check whether @event is enabled and, if so, report @event and @pid
173 * to the ptrace parent. @pid is reported as the pid_t seen from the
174 * the ptrace parent's pid namespace.
175 *
176 * Called without locks.
177 */
ptrace_event_pid(int event,struct pid * pid)178 static inline void ptrace_event_pid(int event, struct pid *pid)
179 {
180 /*
181 * FIXME: There's a potential race if a ptracer in a different pid
182 * namespace than parent attaches between computing message below and
183 * when we acquire tasklist_lock in ptrace_stop(). If this happens,
184 * the ptracer will get a bogus pid from PTRACE_GETEVENTMSG.
185 */
186 unsigned long message = 0;
187 struct pid_namespace *ns;
188
189 rcu_read_lock();
190 ns = task_active_pid_ns(rcu_dereference(current->parent));
191 if (ns)
192 message = pid_nr_ns(pid, ns);
193 rcu_read_unlock();
194
195 ptrace_event(event, message);
196 }
197
198 /**
199 * ptrace_init_task - initialize ptrace state for a new child
200 * @child: new child task
201 * @ptrace: true if child should be ptrace'd by parent's tracer
202 *
203 * This is called immediately after adding @child to its parent's children
204 * list. @ptrace is false in the normal case, and true to ptrace @child.
205 *
206 * Called with current's siglock and write_lock_irq(&tasklist_lock) held.
207 */
ptrace_init_task(struct task_struct * child,bool ptrace)208 static inline void ptrace_init_task(struct task_struct *child, bool ptrace)
209 {
210 INIT_LIST_HEAD(&child->ptrace_entry);
211 INIT_LIST_HEAD(&child->ptraced);
212 child->jobctl = 0;
213 child->ptrace = 0;
214 child->parent = child->real_parent;
215
216 if (unlikely(ptrace) && current->ptrace) {
217 child->ptrace = current->ptrace;
218 __ptrace_link(child, current->parent, current->ptracer_cred);
219
220 if (child->ptrace & PT_SEIZED)
221 task_set_jobctl_pending(child, JOBCTL_TRAP_STOP);
222 else
223 sigaddset(&child->pending.signal, SIGSTOP);
224 }
225 else
226 child->ptracer_cred = NULL;
227 }
228
229 /**
230 * ptrace_release_task - final ptrace-related cleanup of a zombie being reaped
231 * @task: task in %EXIT_DEAD state
232 *
233 * Called with write_lock(&tasklist_lock) held.
234 */
ptrace_release_task(struct task_struct * task)235 static inline void ptrace_release_task(struct task_struct *task)
236 {
237 BUG_ON(!list_empty(&task->ptraced));
238 ptrace_unlink(task);
239 BUG_ON(!list_empty(&task->ptrace_entry));
240 }
241
242 #ifndef force_successful_syscall_return
243 /*
244 * System call handlers that, upon successful completion, need to return a
245 * negative value should call force_successful_syscall_return() right before
246 * returning. On architectures where the syscall convention provides for a
247 * separate error flag (e.g., alpha, ia64, ppc{,64}, sparc{,64}, possibly
248 * others), this macro can be used to ensure that the error flag will not get
249 * set. On architectures which do not support a separate error flag, the macro
250 * is a no-op and the spurious error condition needs to be filtered out by some
251 * other means (e.g., in user-level, by passing an extra argument to the
252 * syscall handler, or something along those lines).
253 */
254 #define force_successful_syscall_return() do { } while (0)
255 #endif
256
257 #ifndef is_syscall_success
258 /*
259 * On most systems we can tell if a syscall is a success based on if the retval
260 * is an error value. On some systems like ia64 and powerpc they have different
261 * indicators of success/failure and must define their own.
262 */
263 #define is_syscall_success(regs) (!IS_ERR_VALUE((unsigned long)(regs_return_value(regs))))
264 #endif
265
266 /*
267 * <asm/ptrace.h> should define the following things inside #ifdef __KERNEL__.
268 *
269 * These do-nothing inlines are used when the arch does not
270 * implement single-step. The kerneldoc comments are here
271 * to document the interface for all arch definitions.
272 */
273
274 #ifndef arch_has_single_step
275 /**
276 * arch_has_single_step - does this CPU support user-mode single-step?
277 *
278 * If this is defined, then there must be function declarations or
279 * inlines for user_enable_single_step() and user_disable_single_step().
280 * arch_has_single_step() should evaluate to nonzero iff the machine
281 * supports instruction single-step for user mode.
282 * It can be a constant or it can test a CPU feature bit.
283 */
284 #define arch_has_single_step() (0)
285
286 /**
287 * user_enable_single_step - single-step in user-mode task
288 * @task: either current or a task stopped in %TASK_TRACED
289 *
290 * This can only be called when arch_has_single_step() has returned nonzero.
291 * Set @task so that when it returns to user mode, it will trap after the
292 * next single instruction executes. If arch_has_block_step() is defined,
293 * this must clear the effects of user_enable_block_step() too.
294 */
user_enable_single_step(struct task_struct * task)295 static inline void user_enable_single_step(struct task_struct *task)
296 {
297 BUG(); /* This can never be called. */
298 }
299
300 /**
301 * user_disable_single_step - cancel user-mode single-step
302 * @task: either current or a task stopped in %TASK_TRACED
303 *
304 * Clear @task of the effects of user_enable_single_step() and
305 * user_enable_block_step(). This can be called whether or not either
306 * of those was ever called on @task, and even if arch_has_single_step()
307 * returned zero.
308 */
user_disable_single_step(struct task_struct * task)309 static inline void user_disable_single_step(struct task_struct *task)
310 {
311 }
312 #else
313 extern void user_enable_single_step(struct task_struct *);
314 extern void user_disable_single_step(struct task_struct *);
315 #endif /* arch_has_single_step */
316
317 #ifndef arch_has_block_step
318 /**
319 * arch_has_block_step - does this CPU support user-mode block-step?
320 *
321 * If this is defined, then there must be a function declaration or inline
322 * for user_enable_block_step(), and arch_has_single_step() must be defined
323 * too. arch_has_block_step() should evaluate to nonzero iff the machine
324 * supports step-until-branch for user mode. It can be a constant or it
325 * can test a CPU feature bit.
326 */
327 #define arch_has_block_step() (0)
328
329 /**
330 * user_enable_block_step - step until branch in user-mode task
331 * @task: either current or a task stopped in %TASK_TRACED
332 *
333 * This can only be called when arch_has_block_step() has returned nonzero,
334 * and will never be called when single-instruction stepping is being used.
335 * Set @task so that when it returns to user mode, it will trap after the
336 * next branch or trap taken.
337 */
user_enable_block_step(struct task_struct * task)338 static inline void user_enable_block_step(struct task_struct *task)
339 {
340 BUG(); /* This can never be called. */
341 }
342 #else
343 extern void user_enable_block_step(struct task_struct *);
344 #endif /* arch_has_block_step */
345
346 #ifdef ARCH_HAS_USER_SINGLE_STEP_REPORT
347 extern void user_single_step_report(struct pt_regs *regs);
348 #else
user_single_step_report(struct pt_regs * regs)349 static inline void user_single_step_report(struct pt_regs *regs)
350 {
351 kernel_siginfo_t info;
352 clear_siginfo(&info);
353 info.si_signo = SIGTRAP;
354 info.si_errno = 0;
355 info.si_code = SI_USER;
356 info.si_pid = 0;
357 info.si_uid = 0;
358 force_sig_info(&info);
359 }
360 #endif
361
362 #ifndef arch_ptrace_stop_needed
363 /**
364 * arch_ptrace_stop_needed - Decide whether arch_ptrace_stop() should be called
365 * @code: current->exit_code value ptrace will stop with
366 * @info: siginfo_t pointer (or %NULL) for signal ptrace will stop with
367 *
368 * This is called with the siglock held, to decide whether or not it's
369 * necessary to release the siglock and call arch_ptrace_stop() with the
370 * same @code and @info arguments. It can be defined to a constant if
371 * arch_ptrace_stop() is never required, or always is. On machines where
372 * this makes sense, it should be defined to a quick test to optimize out
373 * calling arch_ptrace_stop() when it would be superfluous. For example,
374 * if the thread has not been back to user mode since the last stop, the
375 * thread state might indicate that nothing needs to be done.
376 *
377 * This is guaranteed to be invoked once before a task stops for ptrace and
378 * may include arch-specific operations necessary prior to a ptrace stop.
379 */
380 #define arch_ptrace_stop_needed(code, info) (0)
381 #endif
382
383 #ifndef arch_ptrace_stop
384 /**
385 * arch_ptrace_stop - Do machine-specific work before stopping for ptrace
386 * @code: current->exit_code value ptrace will stop with
387 * @info: siginfo_t pointer (or %NULL) for signal ptrace will stop with
388 *
389 * This is called with no locks held when arch_ptrace_stop_needed() has
390 * just returned nonzero. It is allowed to block, e.g. for user memory
391 * access. The arch can have machine-specific work to be done before
392 * ptrace stops. On ia64, register backing store gets written back to user
393 * memory here. Since this can be costly (requires dropping the siglock),
394 * we only do it when the arch requires it for this particular stop, as
395 * indicated by arch_ptrace_stop_needed().
396 */
397 #define arch_ptrace_stop(code, info) do { } while (0)
398 #endif
399
400 #ifndef current_pt_regs
401 #define current_pt_regs() task_pt_regs(current)
402 #endif
403
404 /*
405 * unlike current_pt_regs(), this one is equal to task_pt_regs(current)
406 * on *all* architectures; the only reason to have a per-arch definition
407 * is optimisation.
408 */
409 #ifndef signal_pt_regs
410 #define signal_pt_regs() task_pt_regs(current)
411 #endif
412
413 #ifndef current_user_stack_pointer
414 #define current_user_stack_pointer() user_stack_pointer(current_pt_regs())
415 #endif
416
417 extern int task_current_syscall(struct task_struct *target, struct syscall_info *info);
418
419 extern void sigaction_compat_abi(struct k_sigaction *act, struct k_sigaction *oact);
420 #endif
421