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_PATCH_PENDING
17 # define _TIF_PATCH_PENDING (0)
21 # define _TIF_UPROBE (0)
24 #ifndef _TIF_NOTIFY_SIGNAL
25 # define _TIF_NOTIFY_SIGNAL (0)
29 * SYSCALL_WORK flags handled in syscall_enter_from_user_mode()
31 #ifndef ARCH_SYSCALL_WORK_ENTER
32 # define ARCH_SYSCALL_WORK_ENTER (0)
36 * SYSCALL_WORK flags handled in syscall_exit_to_user_mode()
38 #ifndef ARCH_SYSCALL_WORK_EXIT
39 # define ARCH_SYSCALL_WORK_EXIT (0)
42 #define SYSCALL_WORK_ENTER (SYSCALL_WORK_SECCOMP | \
43 SYSCALL_WORK_SYSCALL_TRACEPOINT | \
44 SYSCALL_WORK_SYSCALL_TRACE | \
45 SYSCALL_WORK_SYSCALL_EMU | \
46 SYSCALL_WORK_SYSCALL_AUDIT | \
47 SYSCALL_WORK_SYSCALL_USER_DISPATCH | \
48 ARCH_SYSCALL_WORK_ENTER)
49 #define SYSCALL_WORK_EXIT (SYSCALL_WORK_SYSCALL_TRACEPOINT | \
50 SYSCALL_WORK_SYSCALL_TRACE | \
51 SYSCALL_WORK_SYSCALL_AUDIT | \
52 SYSCALL_WORK_SYSCALL_USER_DISPATCH | \
53 ARCH_SYSCALL_WORK_EXIT)
56 * TIF flags handled in exit_to_user_mode_loop()
58 #ifndef ARCH_EXIT_TO_USER_MODE_WORK
59 # define ARCH_EXIT_TO_USER_MODE_WORK (0)
62 #define EXIT_TO_USER_MODE_WORK \
63 (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_UPROBE | \
64 _TIF_NEED_RESCHED | _TIF_PATCH_PENDING | _TIF_NOTIFY_SIGNAL | \
65 ARCH_EXIT_TO_USER_MODE_WORK)
68 * arch_check_user_regs - Architecture specific sanity check for user mode regs
69 * @regs: Pointer to currents pt_regs
71 * Defaults to an empty implementation. Can be replaced by architecture
74 * Invoked from syscall_enter_from_user_mode() in the non-instrumentable
75 * section. Use __always_inline so the compiler cannot push it out of line
76 * and make it instrumentable.
78 static __always_inline void arch_check_user_regs(struct pt_regs *regs);
80 #ifndef arch_check_user_regs
81 static __always_inline void arch_check_user_regs(struct pt_regs *regs) {}
85 * arch_syscall_enter_tracehook - Wrapper around tracehook_report_syscall_entry()
86 * @regs: Pointer to currents pt_regs
88 * Returns: 0 on success or an error code to skip the syscall.
90 * Defaults to tracehook_report_syscall_entry(). Can be replaced by
91 * architecture specific code.
93 * Invoked from syscall_enter_from_user_mode()
95 static inline __must_check int arch_syscall_enter_tracehook(struct pt_regs *regs);
97 #ifndef arch_syscall_enter_tracehook
98 static inline __must_check int arch_syscall_enter_tracehook(struct pt_regs *regs)
100 return tracehook_report_syscall_entry(regs);
105 * enter_from_user_mode - Establish state when coming from user mode
107 * Syscall/interrupt entry disables interrupts, but user mode is traced as
108 * interrupts enabled. Also with NO_HZ_FULL RCU might be idle.
110 * 1) Tell lockdep that interrupts are disabled
111 * 2) Invoke context tracking if enabled to reactivate RCU
112 * 3) Trace interrupts off state
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 and interrupts are still
117 * disabled. The subsequent functions can be instrumented.
119 * This is invoked when there is architecture specific functionality to be
120 * done between establishing state and enabling interrupts. The caller must
121 * enable interrupts before invoking syscall_enter_from_user_mode_work().
123 void enter_from_user_mode(struct pt_regs *regs);
126 * syscall_enter_from_user_mode_prepare - Establish state and enable interrupts
127 * @regs: Pointer to currents pt_regs
129 * Invoked from architecture specific syscall entry code with interrupts
130 * disabled. The calling code has to be non-instrumentable. When the
131 * function returns all state is correct, interrupts are enabled and the
132 * subsequent functions can be instrumented.
134 * This handles lockdep, RCU (context tracking) and tracing state, i.e.
135 * the functionality provided by enter_from_user_mode().
137 * This is invoked when there is extra architecture specific functionality
138 * to be done between establishing state and handling user mode entry work.
140 void syscall_enter_from_user_mode_prepare(struct pt_regs *regs);
143 * syscall_enter_from_user_mode_work - Check and handle work before invoking
145 * @regs: Pointer to currents pt_regs
146 * @syscall: The syscall number
148 * Invoked from architecture specific syscall entry code with interrupts
149 * enabled after invoking syscall_enter_from_user_mode_prepare() and extra
150 * architecture specific work.
152 * Returns: The original or a modified syscall number
154 * If the returned syscall number is -1 then the syscall should be
155 * skipped. In this case the caller may invoke syscall_set_error() or
156 * syscall_set_return_value() first. If neither of those are called and -1
157 * is returned, then the syscall will fail with ENOSYS.
159 * It handles the following work items:
161 * 1) syscall_work flag dependent invocations of
162 * arch_syscall_enter_tracehook(), __secure_computing(), trace_sys_enter()
163 * 2) Invocation of audit_syscall_entry()
165 long syscall_enter_from_user_mode_work(struct pt_regs *regs, long syscall);
168 * syscall_enter_from_user_mode - Establish state and check and handle work
169 * before invoking a syscall
170 * @regs: Pointer to currents pt_regs
171 * @syscall: The syscall number
173 * Invoked from architecture specific syscall entry code with interrupts
174 * disabled. The calling code has to be non-instrumentable. When the
175 * function returns all state is correct, interrupts are enabled and the
176 * subsequent functions can be instrumented.
178 * This is combination of syscall_enter_from_user_mode_prepare() and
179 * syscall_enter_from_user_mode_work().
181 * Returns: The original or a modified syscall number. See
182 * syscall_enter_from_user_mode_work() for further explanation.
184 long syscall_enter_from_user_mode(struct pt_regs *regs, long syscall);
187 * local_irq_enable_exit_to_user - Exit to user variant of local_irq_enable()
188 * @ti_work: Cached TIF flags gathered with interrupts disabled
190 * Defaults to local_irq_enable(). Can be supplied by architecture specific
193 static inline void local_irq_enable_exit_to_user(unsigned long ti_work);
195 #ifndef local_irq_enable_exit_to_user
196 static inline void local_irq_enable_exit_to_user(unsigned long ti_work)
203 * local_irq_disable_exit_to_user - Exit to user variant of local_irq_disable()
205 * Defaults to local_irq_disable(). Can be supplied by architecture specific
208 static inline void local_irq_disable_exit_to_user(void);
210 #ifndef local_irq_disable_exit_to_user
211 static inline void local_irq_disable_exit_to_user(void)
218 * arch_exit_to_user_mode_work - Architecture specific TIF work for exit
220 * @regs: Pointer to currents pt_regs
221 * @ti_work: Cached TIF flags gathered with interrupts disabled
223 * Invoked from exit_to_user_mode_loop() with interrupt enabled
225 * Defaults to NOOP. Can be supplied by architecture specific code.
227 static inline void arch_exit_to_user_mode_work(struct pt_regs *regs,
228 unsigned long ti_work);
230 #ifndef arch_exit_to_user_mode_work
231 static inline void arch_exit_to_user_mode_work(struct pt_regs *regs,
232 unsigned long ti_work)
238 * arch_exit_to_user_mode_prepare - Architecture specific preparation for
240 * @regs: Pointer to currents pt_regs
241 * @ti_work: Cached TIF flags gathered with interrupts disabled
243 * Invoked from exit_to_user_mode_prepare() with interrupt disabled as the last
244 * function before return. Defaults to NOOP.
246 static inline void arch_exit_to_user_mode_prepare(struct pt_regs *regs,
247 unsigned long ti_work);
249 #ifndef arch_exit_to_user_mode_prepare
250 static inline void arch_exit_to_user_mode_prepare(struct pt_regs *regs,
251 unsigned long ti_work)
257 * arch_exit_to_user_mode - Architecture specific final work before
260 * Invoked from exit_to_user_mode() with interrupt disabled as the last
261 * function before return. Defaults to NOOP.
263 * This needs to be __always_inline because it is non-instrumentable code
264 * invoked after context tracking switched to user mode.
266 * An architecture implementation must not do anything complex, no locking
267 * etc. The main purpose is for speculation mitigations.
269 static __always_inline void arch_exit_to_user_mode(void);
271 #ifndef arch_exit_to_user_mode
272 static __always_inline void arch_exit_to_user_mode(void) { }
276 * arch_do_signal_or_restart - Architecture specific signal delivery function
277 * @regs: Pointer to currents pt_regs
278 * @has_signal: actual signal to handle
280 * Invoked from exit_to_user_mode_loop().
282 void arch_do_signal_or_restart(struct pt_regs *regs, bool has_signal);
285 * arch_syscall_exit_tracehook - Wrapper around tracehook_report_syscall_exit()
286 * @regs: Pointer to currents pt_regs
287 * @step: Indicator for single step
289 * Defaults to tracehook_report_syscall_exit(). Can be replaced by
290 * architecture specific code.
292 * Invoked from syscall_exit_to_user_mode()
294 static inline void arch_syscall_exit_tracehook(struct pt_regs *regs, bool step);
296 #ifndef arch_syscall_exit_tracehook
297 static inline void arch_syscall_exit_tracehook(struct pt_regs *regs, bool step)
299 tracehook_report_syscall_exit(regs, step);
304 * exit_to_user_mode - Fixup state when exiting to user mode
306 * Syscall/interrupt exit enables interrupts, but the kernel state is
307 * interrupts disabled when this is invoked. Also tell RCU about it.
309 * 1) Trace interrupts on state
310 * 2) Invoke context tracking if enabled to adjust RCU state
311 * 3) Invoke architecture specific last minute exit code, e.g. speculation
312 * mitigations, etc.: arch_exit_to_user_mode()
313 * 4) Tell lockdep that interrupts are enabled
315 * Invoked from architecture specific code when syscall_exit_to_user_mode()
316 * is not suitable as the last step before returning to userspace. Must be
317 * invoked with interrupts disabled and the caller must be
318 * non-instrumentable.
319 * The caller has to invoke syscall_exit_to_user_mode_work() before this.
321 void exit_to_user_mode(void);
324 * syscall_exit_to_user_mode_work - Handle work before returning to user mode
325 * @regs: Pointer to currents pt_regs
327 * Same as step 1 and 2 of syscall_exit_to_user_mode() but without calling
328 * exit_to_user_mode() to perform the final transition to user mode.
330 * Calling convention is the same as for syscall_exit_to_user_mode() and it
331 * returns with all work handled and interrupts disabled. The caller must
332 * invoke exit_to_user_mode() before actually switching to user mode to
333 * make the final state transitions. Interrupts must stay disabled between
334 * return from this function and the invocation of exit_to_user_mode().
336 void syscall_exit_to_user_mode_work(struct pt_regs *regs);
339 * syscall_exit_to_user_mode - Handle work before returning to user mode
340 * @regs: Pointer to currents pt_regs
342 * Invoked with interrupts enabled and fully valid regs. Returns with all
343 * work handled, interrupts disabled such that the caller can immediately
344 * switch to user mode. Called from architecture specific syscall and ret
348 * 1) One-time syscall exit work:
349 * - rseq syscall exit
352 * - tracehook (single stepping)
354 * 2) Preparatory work
355 * - Exit to user mode loop (common TIF handling). Invokes
356 * arch_exit_to_user_mode_work() for architecture specific TIF work
357 * - Architecture specific one time work arch_exit_to_user_mode_prepare()
358 * - Address limit and lockdep checks
360 * 3) Final transition (lockdep, tracing, context tracking, RCU), i.e. the
361 * functionality in exit_to_user_mode().
363 * This is a combination of syscall_exit_to_user_mode_work() (1,2) and
364 * exit_to_user_mode(). This function is preferred unless there is a
365 * compelling architectural reason to use the seperate functions.
367 void syscall_exit_to_user_mode(struct pt_regs *regs);
370 * irqentry_enter_from_user_mode - Establish state before invoking the irq handler
371 * @regs: Pointer to currents pt_regs
373 * Invoked from architecture specific entry code with interrupts disabled.
374 * Can only be called when the interrupt entry came from user mode. The
375 * calling code must be non-instrumentable. When the function returns all
376 * state is correct and the subsequent functions can be instrumented.
378 * The function establishes state (lockdep, RCU (context tracking), tracing)
380 void irqentry_enter_from_user_mode(struct pt_regs *regs);
383 * irqentry_exit_to_user_mode - Interrupt exit work
384 * @regs: Pointer to current's pt_regs
386 * Invoked with interrupts disbled and fully valid regs. Returns with all
387 * work handled, interrupts disabled such that the caller can immediately
388 * switch to user mode. Called from architecture specific interrupt
391 * The call order is #2 and #3 as described in syscall_exit_to_user_mode().
392 * Interrupt exit is not invoking #1 which is the syscall specific one time
395 void irqentry_exit_to_user_mode(struct pt_regs *regs);
397 #ifndef irqentry_state
399 * struct irqentry_state - Opaque object for exception state storage
400 * @exit_rcu: Used exclusively in the irqentry_*() calls; signals whether the
401 * exit path has to invoke rcu_irq_exit().
402 * @lockdep: Used exclusively in the irqentry_nmi_*() calls; ensures that
403 * lockdep state is restored correctly on exit from nmi.
405 * This opaque object is filled in by the irqentry_*_enter() functions and
406 * must be passed back into the corresponding irqentry_*_exit() functions
407 * when the exception is complete.
409 * Callers of irqentry_*_[enter|exit]() must consider this structure opaque
410 * and all members private. Descriptions of the members are provided to aid in
411 * the maintenance of the irqentry_*() functions.
413 typedef struct irqentry_state {
422 * irqentry_enter - Handle state tracking on ordinary interrupt entries
423 * @regs: Pointer to pt_regs of interrupted context
426 * - lockdep irqflag state tracking as low level ASM entry disabled
429 * - Context tracking if the exception hit user mode.
431 * - The hardirq tracer to keep the state consistent as low level ASM
432 * entry disabled interrupts.
434 * As a precondition, this requires that the entry came from user mode,
435 * idle, or a kernel context in which RCU is watching.
437 * For kernel mode entries RCU handling is done conditional. If RCU is
438 * watching then the only RCU requirement is to check whether the tick has
439 * to be restarted. If RCU is not watching then rcu_irq_enter() has to be
440 * invoked on entry and rcu_irq_exit() on exit.
442 * Avoiding the rcu_irq_enter/exit() calls is an optimization but also
443 * solves the problem of kernel mode pagefaults which can schedule, which
444 * is not possible after invoking rcu_irq_enter() without undoing it.
446 * For user mode entries irqentry_enter_from_user_mode() is invoked to
447 * establish the proper context for NOHZ_FULL. Otherwise scheduling on exit
448 * would not be possible.
450 * Returns: An opaque object that must be passed to idtentry_exit()
452 irqentry_state_t noinstr irqentry_enter(struct pt_regs *regs);
455 * irqentry_exit_cond_resched - Conditionally reschedule on return from interrupt
457 * Conditional reschedule with additional sanity checks.
459 void irqentry_exit_cond_resched(void);
462 * irqentry_exit - Handle return from exception that used irqentry_enter()
463 * @regs: Pointer to pt_regs (exception entry regs)
464 * @state: Return value from matching call to irqentry_enter()
466 * Depending on the return target (kernel/user) this runs the necessary
467 * preemption and work checks if possible and required and returns to
468 * the caller with interrupts disabled and no further work pending.
470 * This is the last action before returning to the low level ASM code which
471 * just needs to return to the appropriate context.
473 * Counterpart to irqentry_enter().
475 void noinstr irqentry_exit(struct pt_regs *regs, irqentry_state_t state);
478 * irqentry_nmi_enter - Handle NMI entry
479 * @regs: Pointer to currents pt_regs
481 * Similar to irqentry_enter() but taking care of the NMI constraints.
483 irqentry_state_t noinstr irqentry_nmi_enter(struct pt_regs *regs);
486 * irqentry_nmi_exit - Handle return from NMI handling
487 * @regs: Pointer to pt_regs (NMI entry regs)
488 * @irq_state: Return value from matching call to irqentry_nmi_enter()
490 * Last action before returning to the low level assembly code.
492 * Counterpart to irqentry_nmi_enter().
494 void noinstr irqentry_nmi_exit(struct pt_regs *regs, irqentry_state_t irq_state);