1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3  * Tracing hooks
4  *
5  * Copyright (C) 2008-2009 Red Hat, Inc.  All rights reserved.
6  *
7  * This file defines hook entry points called by core code where
8  * user tracing/debugging support might need to do something.  These
9  * entry points are called tracehook_*().  Each hook declared below
10  * has a detailed kerneldoc comment giving the context (locking et
11  * al) from which it is called, and the meaning of its return value.
12  *
13  * Each function here typically has only one call site, so it is ok
14  * to have some nontrivial tracehook_*() inlines.  In all cases, the
15  * fast path when no tracing is enabled should be very short.
16  *
17  * The purpose of this file and the tracehook_* layer is to consolidate
18  * the interface that the kernel core and arch code uses to enable any
19  * user debugging or tracing facility (such as ptrace).  The interfaces
20  * here are carefully documented so that maintainers of core and arch
21  * code do not need to think about the implementation details of the
22  * tracing facilities.  Likewise, maintainers of the tracing code do not
23  * need to understand all the calling core or arch code in detail, just
24  * documented circumstances of each call, such as locking conditions.
25  *
26  * If the calling core code changes so that locking is different, then
27  * it is ok to change the interface documented here.  The maintainer of
28  * core code changing should notify the maintainers of the tracing code
29  * that they need to work out the change.
30  *
31  * Some tracehook_*() inlines take arguments that the current tracing
32  * implementations might not necessarily use.  These function signatures
33  * are chosen to pass in all the information that is on hand in the
34  * caller and might conceivably be relevant to a tracer, so that the
35  * core code won't have to be updated when tracing adds more features.
36  * If a call site changes so that some of those parameters are no longer
37  * already on hand without extra work, then the tracehook_* interface
38  * can change so there is no make-work burden on the core code.  The
39  * maintainer of core code changing should notify the maintainers of the
40  * tracing code that they need to work out the change.
41  */
42 
43 #ifndef _LINUX_TRACEHOOK_H
44 #define _LINUX_TRACEHOOK_H	1
45 
46 #include <linux/sched.h>
47 #include <linux/ptrace.h>
48 #include <linux/security.h>
49 #include <linux/task_work.h>
50 #include <linux/memcontrol.h>
51 #include <linux/blk-cgroup.h>
52 struct linux_binprm;
53 
54 /*
55  * ptrace report for syscall entry and exit looks identical.
56  */
ptrace_report_syscall(struct pt_regs * regs,unsigned long message)57 static inline int ptrace_report_syscall(struct pt_regs *regs,
58 					unsigned long message)
59 {
60 	int ptrace = current->ptrace;
61 
62 	if (!(ptrace & PT_PTRACED))
63 		return 0;
64 
65 	current->ptrace_message = message;
66 	ptrace_notify(SIGTRAP | ((ptrace & PT_TRACESYSGOOD) ? 0x80 : 0));
67 
68 	/*
69 	 * this isn't the same as continuing with a signal, but it will do
70 	 * for normal use.  strace only continues with a signal if the
71 	 * stopping signal is not SIGTRAP.  -brl
72 	 */
73 	if (current->exit_code) {
74 		send_sig(current->exit_code, current, 1);
75 		current->exit_code = 0;
76 	}
77 
78 	current->ptrace_message = 0;
79 	return fatal_signal_pending(current);
80 }
81 
82 /**
83  * tracehook_report_syscall_entry - task is about to attempt a system call
84  * @regs:		user register state of current task
85  *
86  * This will be called if %SYSCALL_WORK_SYSCALL_TRACE or
87  * %SYSCALL_WORK_SYSCALL_EMU have been set, when the current task has just
88  * entered the kernel for a system call.  Full user register state is
89  * available here.  Changing the values in @regs can affect the system
90  * call number and arguments to be tried.  It is safe to block here,
91  * preventing the system call from beginning.
92  *
93  * Returns zero normally, or nonzero if the calling arch code should abort
94  * the system call.  That must prevent normal entry so no system call is
95  * made.  If @task ever returns to user mode after this, its register state
96  * is unspecified, but should be something harmless like an %ENOSYS error
97  * return.  It should preserve enough information so that syscall_rollback()
98  * can work (see asm-generic/syscall.h).
99  *
100  * Called without locks, just after entering kernel mode.
101  */
tracehook_report_syscall_entry(struct pt_regs * regs)102 static inline __must_check int tracehook_report_syscall_entry(
103 	struct pt_regs *regs)
104 {
105 	return ptrace_report_syscall(regs, PTRACE_EVENTMSG_SYSCALL_ENTRY);
106 }
107 
108 /**
109  * tracehook_report_syscall_exit - task has just finished a system call
110  * @regs:		user register state of current task
111  * @step:		nonzero if simulating single-step or block-step
112  *
113  * This will be called if %SYSCALL_WORK_SYSCALL_TRACE has been set, when
114  * the current task has just finished an attempted system call.  Full
115  * user register state is available here.  It is safe to block here,
116  * preventing signals from being processed.
117  *
118  * If @step is nonzero, this report is also in lieu of the normal
119  * trap that would follow the system call instruction because
120  * user_enable_block_step() or user_enable_single_step() was used.
121  * In this case, %SYSCALL_WORK_SYSCALL_TRACE might not be set.
122  *
123  * Called without locks, just before checking for pending signals.
124  */
tracehook_report_syscall_exit(struct pt_regs * regs,int step)125 static inline void tracehook_report_syscall_exit(struct pt_regs *regs, int step)
126 {
127 	if (step)
128 		user_single_step_report(regs);
129 	else
130 		ptrace_report_syscall(regs, PTRACE_EVENTMSG_SYSCALL_EXIT);
131 }
132 
133 /**
134  * tracehook_signal_handler - signal handler setup is complete
135  * @stepping:		nonzero if debugger single-step or block-step in use
136  *
137  * Called by the arch code after a signal handler has been set up.
138  * Register and stack state reflects the user handler about to run.
139  * Signal mask changes have already been made.
140  *
141  * Called without locks, shortly before returning to user mode
142  * (or handling more signals).
143  */
tracehook_signal_handler(int stepping)144 static inline void tracehook_signal_handler(int stepping)
145 {
146 	if (stepping)
147 		ptrace_notify(SIGTRAP);
148 }
149 
150 /**
151  * set_notify_resume - cause tracehook_notify_resume() to be called
152  * @task:		task that will call tracehook_notify_resume()
153  *
154  * Calling this arranges that @task will call tracehook_notify_resume()
155  * before returning to user mode.  If it's already running in user mode,
156  * it will enter the kernel and call tracehook_notify_resume() soon.
157  * If it's blocked, it will not be woken.
158  */
set_notify_resume(struct task_struct * task)159 static inline void set_notify_resume(struct task_struct *task)
160 {
161 #ifdef TIF_NOTIFY_RESUME
162 	if (!test_and_set_tsk_thread_flag(task, TIF_NOTIFY_RESUME))
163 		kick_process(task);
164 #endif
165 }
166 
167 /**
168  * tracehook_notify_resume - report when about to return to user mode
169  * @regs:		user-mode registers of @current task
170  *
171  * This is called when %TIF_NOTIFY_RESUME has been set.  Now we are
172  * about to return to user mode, and the user state in @regs can be
173  * inspected or adjusted.  The caller in arch code has cleared
174  * %TIF_NOTIFY_RESUME before the call.  If the flag gets set again
175  * asynchronously, this will be called again before we return to
176  * user mode.
177  *
178  * Called without locks.
179  */
tracehook_notify_resume(struct pt_regs * regs)180 static inline void tracehook_notify_resume(struct pt_regs *regs)
181 {
182 	clear_thread_flag(TIF_NOTIFY_RESUME);
183 	/*
184 	 * This barrier pairs with task_work_add()->set_notify_resume() after
185 	 * hlist_add_head(task->task_works);
186 	 */
187 	smp_mb__after_atomic();
188 	if (unlikely(current->task_works))
189 		task_work_run();
190 
191 #ifdef CONFIG_KEYS_REQUEST_CACHE
192 	if (unlikely(current->cached_requested_key)) {
193 		key_put(current->cached_requested_key);
194 		current->cached_requested_key = NULL;
195 	}
196 #endif
197 
198 	mem_cgroup_handle_over_high();
199 	blkcg_maybe_throttle_current();
200 
201 	rseq_handle_notify_resume(NULL, regs);
202 }
203 
204 /*
205  * called by exit_to_user_mode_loop() if ti_work & _TIF_NOTIFY_SIGNAL. This
206  * is currently used by TWA_SIGNAL based task_work, which requires breaking
207  * wait loops to ensure that task_work is noticed and run.
208  */
tracehook_notify_signal(void)209 static inline void tracehook_notify_signal(void)
210 {
211 	clear_thread_flag(TIF_NOTIFY_SIGNAL);
212 	smp_mb__after_atomic();
213 	if (current->task_works)
214 		task_work_run();
215 }
216 
217 /*
218  * Called when we have work to process from exit_to_user_mode_loop()
219  */
set_notify_signal(struct task_struct * task)220 static inline void set_notify_signal(struct task_struct *task)
221 {
222 	if (!test_and_set_tsk_thread_flag(task, TIF_NOTIFY_SIGNAL) &&
223 	    !wake_up_state(task, TASK_INTERRUPTIBLE))
224 		kick_process(task);
225 }
226 
227 #endif	/* <linux/tracehook.h> */
228