1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef __LINUX_ENTRYCOMMON_H
3 #define __LINUX_ENTRYCOMMON_H
5 #include <linux/tracehook.h>
6 #include <linux/syscalls.h>
7 #include <linux/seccomp.h>
8 #include <linux/sched.h>
10 #include <asm/entry-common.h>
13 * Define dummy _TIF work flags if not defined by the architecture or for
14 * disabled functionality.
16 #ifndef _TIF_SYSCALL_AUDIT
17 # define _TIF_SYSCALL_AUDIT (0)
20 #ifndef _TIF_PATCH_PENDING
21 # define _TIF_PATCH_PENDING (0)
25 # define _TIF_UPROBE (0)
28 #ifndef _TIF_NOTIFY_SIGNAL
29 # define _TIF_NOTIFY_SIGNAL (0)
33 * TIF flags handled in syscall_enter_from_user_mode()
35 #ifndef ARCH_SYSCALL_ENTER_WORK
36 # define ARCH_SYSCALL_ENTER_WORK (0)
39 #define SYSCALL_ENTER_WORK \
40 (_TIF_SYSCALL_AUDIT | \
41 ARCH_SYSCALL_ENTER_WORK)
44 * TIF flags handled in syscall_exit_to_user_mode()
46 #ifndef ARCH_SYSCALL_EXIT_WORK
47 # define ARCH_SYSCALL_EXIT_WORK (0)
50 #define SYSCALL_EXIT_WORK \
51 (_TIF_SYSCALL_AUDIT | \
52 ARCH_SYSCALL_EXIT_WORK)
54 #define SYSCALL_WORK_ENTER (SYSCALL_WORK_SECCOMP | \
55 SYSCALL_WORK_SYSCALL_TRACEPOINT | \
56 SYSCALL_WORK_SYSCALL_TRACE | \
57 SYSCALL_WORK_SYSCALL_EMU)
58 #define SYSCALL_WORK_EXIT (SYSCALL_WORK_SYSCALL_TRACEPOINT | \
59 SYSCALL_WORK_SYSCALL_TRACE)
62 * TIF flags handled in exit_to_user_mode_loop()
64 #ifndef ARCH_EXIT_TO_USER_MODE_WORK
65 # define ARCH_EXIT_TO_USER_MODE_WORK (0)
68 #define EXIT_TO_USER_MODE_WORK \
69 (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_UPROBE | \
70 _TIF_NEED_RESCHED | _TIF_PATCH_PENDING | _TIF_NOTIFY_SIGNAL | \
71 ARCH_EXIT_TO_USER_MODE_WORK)
74 * arch_check_user_regs - Architecture specific sanity check for user mode regs
75 * @regs: Pointer to currents pt_regs
77 * Defaults to an empty implementation. Can be replaced by architecture
80 * Invoked from syscall_enter_from_user_mode() in the non-instrumentable
81 * section. Use __always_inline so the compiler cannot push it out of line
82 * and make it instrumentable.
84 static __always_inline void arch_check_user_regs(struct pt_regs *regs);
86 #ifndef arch_check_user_regs
87 static __always_inline void arch_check_user_regs(struct pt_regs *regs) {}
91 * arch_syscall_enter_tracehook - Wrapper around tracehook_report_syscall_entry()
92 * @regs: Pointer to currents pt_regs
94 * Returns: 0 on success or an error code to skip the syscall.
96 * Defaults to tracehook_report_syscall_entry(). Can be replaced by
97 * architecture specific code.
99 * Invoked from syscall_enter_from_user_mode()
101 static inline __must_check int arch_syscall_enter_tracehook(struct pt_regs *regs);
103 #ifndef arch_syscall_enter_tracehook
104 static inline __must_check int arch_syscall_enter_tracehook(struct pt_regs *regs)
106 return tracehook_report_syscall_entry(regs);
111 * syscall_enter_from_user_mode_prepare - Establish state and enable interrupts
112 * @regs: Pointer to currents pt_regs
114 * Invoked from architecture specific syscall entry code with interrupts
115 * disabled. The calling code has to be non-instrumentable. When the
116 * function returns all state is correct, interrupts are enabled and the
117 * subsequent functions can be instrumented.
119 * This handles lockdep, RCU (context tracking) and tracing state.
121 * This is invoked when there is extra architecture specific functionality
122 * to be done between establishing state and handling user mode entry work.
124 void syscall_enter_from_user_mode_prepare(struct pt_regs *regs);
127 * syscall_enter_from_user_mode_work - Check and handle work before invoking
129 * @regs: Pointer to currents pt_regs
130 * @syscall: The syscall number
132 * Invoked from architecture specific syscall entry code with interrupts
133 * enabled after invoking syscall_enter_from_user_mode_prepare() and extra
134 * architecture specific work.
136 * Returns: The original or a modified syscall number
138 * If the returned syscall number is -1 then the syscall should be
139 * skipped. In this case the caller may invoke syscall_set_error() or
140 * syscall_set_return_value() first. If neither of those are called and -1
141 * is returned, then the syscall will fail with ENOSYS.
143 * It handles the following work items:
145 * 1) TIF flag dependent invocations of arch_syscall_enter_tracehook(),
146 * __secure_computing(), trace_sys_enter()
147 * 2) Invocation of audit_syscall_entry()
149 long syscall_enter_from_user_mode_work(struct pt_regs *regs, long syscall);
152 * syscall_enter_from_user_mode - Establish state and check and handle work
153 * before invoking a syscall
154 * @regs: Pointer to currents pt_regs
155 * @syscall: The syscall number
157 * Invoked from architecture specific syscall entry code with interrupts
158 * disabled. The calling code has to be non-instrumentable. When the
159 * function returns all state is correct, interrupts are enabled and the
160 * subsequent functions can be instrumented.
162 * This is combination of syscall_enter_from_user_mode_prepare() and
163 * syscall_enter_from_user_mode_work().
165 * Returns: The original or a modified syscall number. See
166 * syscall_enter_from_user_mode_work() for further explanation.
168 long syscall_enter_from_user_mode(struct pt_regs *regs, long syscall);
171 * local_irq_enable_exit_to_user - Exit to user variant of local_irq_enable()
172 * @ti_work: Cached TIF flags gathered with interrupts disabled
174 * Defaults to local_irq_enable(). Can be supplied by architecture specific
177 static inline void local_irq_enable_exit_to_user(unsigned long ti_work);
179 #ifndef local_irq_enable_exit_to_user
180 static inline void local_irq_enable_exit_to_user(unsigned long ti_work)
187 * local_irq_disable_exit_to_user - Exit to user variant of local_irq_disable()
189 * Defaults to local_irq_disable(). Can be supplied by architecture specific
192 static inline void local_irq_disable_exit_to_user(void);
194 #ifndef local_irq_disable_exit_to_user
195 static inline void local_irq_disable_exit_to_user(void)
202 * arch_exit_to_user_mode_work - Architecture specific TIF work for exit
204 * @regs: Pointer to currents pt_regs
205 * @ti_work: Cached TIF flags gathered with interrupts disabled
207 * Invoked from exit_to_user_mode_loop() with interrupt enabled
209 * Defaults to NOOP. Can be supplied by architecture specific code.
211 static inline void arch_exit_to_user_mode_work(struct pt_regs *regs,
212 unsigned long ti_work);
214 #ifndef arch_exit_to_user_mode_work
215 static inline void arch_exit_to_user_mode_work(struct pt_regs *regs,
216 unsigned long ti_work)
222 * arch_exit_to_user_mode_prepare - Architecture specific preparation for
224 * @regs: Pointer to currents pt_regs
225 * @ti_work: Cached TIF flags gathered with interrupts disabled
227 * Invoked from exit_to_user_mode_prepare() with interrupt disabled as the last
228 * function before return. Defaults to NOOP.
230 static inline void arch_exit_to_user_mode_prepare(struct pt_regs *regs,
231 unsigned long ti_work);
233 #ifndef arch_exit_to_user_mode_prepare
234 static inline void arch_exit_to_user_mode_prepare(struct pt_regs *regs,
235 unsigned long ti_work)
241 * arch_exit_to_user_mode - Architecture specific final work before
244 * Invoked from exit_to_user_mode() with interrupt disabled as the last
245 * function before return. Defaults to NOOP.
247 * This needs to be __always_inline because it is non-instrumentable code
248 * invoked after context tracking switched to user mode.
250 * An architecture implementation must not do anything complex, no locking
251 * etc. The main purpose is for speculation mitigations.
253 static __always_inline void arch_exit_to_user_mode(void);
255 #ifndef arch_exit_to_user_mode
256 static __always_inline void arch_exit_to_user_mode(void) { }
260 * arch_do_signal_or_restart - Architecture specific signal delivery function
261 * @regs: Pointer to currents pt_regs
262 * @has_signal: actual signal to handle
264 * Invoked from exit_to_user_mode_loop().
266 void arch_do_signal_or_restart(struct pt_regs *regs, bool has_signal);
269 * arch_syscall_exit_tracehook - Wrapper around tracehook_report_syscall_exit()
270 * @regs: Pointer to currents pt_regs
271 * @step: Indicator for single step
273 * Defaults to tracehook_report_syscall_exit(). Can be replaced by
274 * architecture specific code.
276 * Invoked from syscall_exit_to_user_mode()
278 static inline void arch_syscall_exit_tracehook(struct pt_regs *regs, bool step);
280 #ifndef arch_syscall_exit_tracehook
281 static inline void arch_syscall_exit_tracehook(struct pt_regs *regs, bool step)
283 tracehook_report_syscall_exit(regs, step);
288 * syscall_exit_to_user_mode - Handle work before returning to user mode
289 * @regs: Pointer to currents pt_regs
291 * Invoked with interrupts enabled and fully valid regs. Returns with all
292 * work handled, interrupts disabled such that the caller can immediately
293 * switch to user mode. Called from architecture specific syscall and ret
297 * 1) One-time syscall exit work:
298 * - rseq syscall exit
301 * - tracehook (single stepping)
303 * 2) Preparatory work
304 * - Exit to user mode loop (common TIF handling). Invokes
305 * arch_exit_to_user_mode_work() for architecture specific TIF work
306 * - Architecture specific one time work arch_exit_to_user_mode_prepare()
307 * - Address limit and lockdep checks
309 * 3) Final transition (lockdep, tracing, context tracking, RCU). Invokes
310 * arch_exit_to_user_mode() to handle e.g. speculation mitigations
312 void syscall_exit_to_user_mode(struct pt_regs *regs);
315 * irqentry_enter_from_user_mode - Establish state before invoking the irq handler
316 * @regs: Pointer to currents pt_regs
318 * Invoked from architecture specific entry code with interrupts disabled.
319 * Can only be called when the interrupt entry came from user mode. The
320 * calling code must be non-instrumentable. When the function returns all
321 * state is correct and the subsequent functions can be instrumented.
323 * The function establishes state (lockdep, RCU (context tracking), tracing)
325 void irqentry_enter_from_user_mode(struct pt_regs *regs);
328 * irqentry_exit_to_user_mode - Interrupt exit work
329 * @regs: Pointer to current's pt_regs
331 * Invoked with interrupts disbled and fully valid regs. Returns with all
332 * work handled, interrupts disabled such that the caller can immediately
333 * switch to user mode. Called from architecture specific interrupt
336 * The call order is #2 and #3 as described in syscall_exit_to_user_mode().
337 * Interrupt exit is not invoking #1 which is the syscall specific one time
340 void irqentry_exit_to_user_mode(struct pt_regs *regs);
342 #ifndef irqentry_state
344 * struct irqentry_state - Opaque object for exception state storage
345 * @exit_rcu: Used exclusively in the irqentry_*() calls; signals whether the
346 * exit path has to invoke rcu_irq_exit().
347 * @lockdep: Used exclusively in the irqentry_nmi_*() calls; ensures that
348 * lockdep state is restored correctly on exit from nmi.
350 * This opaque object is filled in by the irqentry_*_enter() functions and
351 * must be passed back into the corresponding irqentry_*_exit() functions
352 * when the exception is complete.
354 * Callers of irqentry_*_[enter|exit]() must consider this structure opaque
355 * and all members private. Descriptions of the members are provided to aid in
356 * the maintenance of the irqentry_*() functions.
358 typedef struct irqentry_state {
367 * irqentry_enter - Handle state tracking on ordinary interrupt entries
368 * @regs: Pointer to pt_regs of interrupted context
371 * - lockdep irqflag state tracking as low level ASM entry disabled
374 * - Context tracking if the exception hit user mode.
376 * - The hardirq tracer to keep the state consistent as low level ASM
377 * entry disabled interrupts.
379 * As a precondition, this requires that the entry came from user mode,
380 * idle, or a kernel context in which RCU is watching.
382 * For kernel mode entries RCU handling is done conditional. If RCU is
383 * watching then the only RCU requirement is to check whether the tick has
384 * to be restarted. If RCU is not watching then rcu_irq_enter() has to be
385 * invoked on entry and rcu_irq_exit() on exit.
387 * Avoiding the rcu_irq_enter/exit() calls is an optimization but also
388 * solves the problem of kernel mode pagefaults which can schedule, which
389 * is not possible after invoking rcu_irq_enter() without undoing it.
391 * For user mode entries irqentry_enter_from_user_mode() is invoked to
392 * establish the proper context for NOHZ_FULL. Otherwise scheduling on exit
393 * would not be possible.
395 * Returns: An opaque object that must be passed to idtentry_exit()
397 irqentry_state_t noinstr irqentry_enter(struct pt_regs *regs);
400 * irqentry_exit_cond_resched - Conditionally reschedule on return from interrupt
402 * Conditional reschedule with additional sanity checks.
404 void irqentry_exit_cond_resched(void);
407 * irqentry_exit - Handle return from exception that used irqentry_enter()
408 * @regs: Pointer to pt_regs (exception entry regs)
409 * @state: Return value from matching call to irqentry_enter()
411 * Depending on the return target (kernel/user) this runs the necessary
412 * preemption and work checks if possible and required and returns to
413 * the caller with interrupts disabled and no further work pending.
415 * This is the last action before returning to the low level ASM code which
416 * just needs to return to the appropriate context.
418 * Counterpart to irqentry_enter().
420 void noinstr irqentry_exit(struct pt_regs *regs, irqentry_state_t state);
423 * irqentry_nmi_enter - Handle NMI entry
424 * @regs: Pointer to currents pt_regs
426 * Similar to irqentry_enter() but taking care of the NMI constraints.
428 irqentry_state_t noinstr irqentry_nmi_enter(struct pt_regs *regs);
431 * irqentry_nmi_exit - Handle return from NMI handling
432 * @regs: Pointer to pt_regs (NMI entry regs)
433 * @irq_state: Return value from matching call to irqentry_nmi_enter()
435 * Last action before returning to the low level assembly code.
437 * Counterpart to irqentry_nmi_enter().
439 void noinstr irqentry_nmi_exit(struct pt_regs *regs, irqentry_state_t irq_state);