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