1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef __LINUX_ENTRYCOMMON_H
3 #define __LINUX_ENTRYCOMMON_H
5 #include <linux/static_call_types.h>
6 #include <linux/tracehook.h>
7 #include <linux/syscalls.h>
8 #include <linux/seccomp.h>
9 #include <linux/sched.h>
11 #include <asm/entry-common.h>
14 * Define dummy _TIF work flags if not defined by the architecture or for
15 * disabled functionality.
17 #ifndef _TIF_PATCH_PENDING
18 # define _TIF_PATCH_PENDING (0)
22 # define _TIF_UPROBE (0)
26 * SYSCALL_WORK flags handled in syscall_enter_from_user_mode()
28 #ifndef ARCH_SYSCALL_WORK_ENTER
29 # define ARCH_SYSCALL_WORK_ENTER (0)
33 * SYSCALL_WORK flags handled in syscall_exit_to_user_mode()
35 #ifndef ARCH_SYSCALL_WORK_EXIT
36 # define ARCH_SYSCALL_WORK_EXIT (0)
39 #define SYSCALL_WORK_ENTER (SYSCALL_WORK_SECCOMP | \
40 SYSCALL_WORK_SYSCALL_TRACEPOINT | \
41 SYSCALL_WORK_SYSCALL_TRACE | \
42 SYSCALL_WORK_SYSCALL_EMU | \
43 SYSCALL_WORK_SYSCALL_AUDIT | \
44 SYSCALL_WORK_SYSCALL_USER_DISPATCH | \
45 ARCH_SYSCALL_WORK_ENTER)
46 #define SYSCALL_WORK_EXIT (SYSCALL_WORK_SYSCALL_TRACEPOINT | \
47 SYSCALL_WORK_SYSCALL_TRACE | \
48 SYSCALL_WORK_SYSCALL_AUDIT | \
49 SYSCALL_WORK_SYSCALL_USER_DISPATCH | \
50 SYSCALL_WORK_SYSCALL_EXIT_TRAP | \
51 ARCH_SYSCALL_WORK_EXIT)
54 * TIF flags handled in exit_to_user_mode_loop()
56 #ifndef ARCH_EXIT_TO_USER_MODE_WORK
57 # define ARCH_EXIT_TO_USER_MODE_WORK (0)
60 #ifdef CONFIG_PREEMPT_LAZY
61 # define _TIF_NEED_RESCHED_MASK (_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY)
63 # define _TIF_NEED_RESCHED_MASK (_TIF_NEED_RESCHED)
66 #define EXIT_TO_USER_MODE_WORK \
67 (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_UPROBE | \
68 _TIF_NEED_RESCHED_MASK | _TIF_PATCH_PENDING | _TIF_NOTIFY_SIGNAL | \
69 ARCH_EXIT_TO_USER_MODE_WORK)
72 * arch_check_user_regs - Architecture specific sanity check for user mode regs
73 * @regs: Pointer to currents pt_regs
75 * Defaults to an empty implementation. Can be replaced by architecture
78 * Invoked from syscall_enter_from_user_mode() in the non-instrumentable
79 * section. Use __always_inline so the compiler cannot push it out of line
80 * and make it instrumentable.
82 static __always_inline void arch_check_user_regs(struct pt_regs *regs);
84 #ifndef arch_check_user_regs
85 static __always_inline void arch_check_user_regs(struct pt_regs *regs) {}
89 * arch_syscall_enter_tracehook - Wrapper around tracehook_report_syscall_entry()
90 * @regs: Pointer to currents pt_regs
92 * Returns: 0 on success or an error code to skip the syscall.
94 * Defaults to tracehook_report_syscall_entry(). Can be replaced by
95 * architecture specific code.
97 * Invoked from syscall_enter_from_user_mode()
99 static inline __must_check int arch_syscall_enter_tracehook(struct pt_regs *regs);
101 #ifndef arch_syscall_enter_tracehook
102 static inline __must_check int arch_syscall_enter_tracehook(struct pt_regs *regs)
104 return tracehook_report_syscall_entry(regs);
109 * enter_from_user_mode - Establish state when coming from user mode
111 * Syscall/interrupt entry disables interrupts, but user mode is traced as
112 * interrupts enabled. Also with NO_HZ_FULL RCU might be idle.
114 * 1) Tell lockdep that interrupts are disabled
115 * 2) Invoke context tracking if enabled to reactivate RCU
116 * 3) Trace interrupts off state
118 * Invoked from architecture specific syscall entry code with interrupts
119 * disabled. The calling code has to be non-instrumentable. When the
120 * function returns all state is correct and interrupts are still
121 * disabled. The subsequent functions can be instrumented.
123 * This is invoked when there is architecture specific functionality to be
124 * done between establishing state and enabling interrupts. The caller must
125 * enable interrupts before invoking syscall_enter_from_user_mode_work().
127 void enter_from_user_mode(struct pt_regs *regs);
130 * syscall_enter_from_user_mode_prepare - Establish state and enable interrupts
131 * @regs: Pointer to currents pt_regs
133 * Invoked from architecture specific syscall entry code with interrupts
134 * disabled. The calling code has to be non-instrumentable. When the
135 * function returns all state is correct, interrupts are enabled and the
136 * subsequent functions can be instrumented.
138 * This handles lockdep, RCU (context tracking) and tracing state, i.e.
139 * the functionality provided by enter_from_user_mode().
141 * This is invoked when there is extra architecture specific functionality
142 * to be done between establishing state and handling user mode entry work.
144 void syscall_enter_from_user_mode_prepare(struct pt_regs *regs);
147 * syscall_enter_from_user_mode_work - Check and handle work before invoking
149 * @regs: Pointer to currents pt_regs
150 * @syscall: The syscall number
152 * Invoked from architecture specific syscall entry code with interrupts
153 * enabled after invoking syscall_enter_from_user_mode_prepare() and extra
154 * architecture specific work.
156 * Returns: The original or a modified syscall number
158 * If the returned syscall number is -1 then the syscall should be
159 * skipped. In this case the caller may invoke syscall_set_error() or
160 * syscall_set_return_value() first. If neither of those are called and -1
161 * is returned, then the syscall will fail with ENOSYS.
163 * It handles the following work items:
165 * 1) syscall_work flag dependent invocations of
166 * arch_syscall_enter_tracehook(), __secure_computing(), trace_sys_enter()
167 * 2) Invocation of audit_syscall_entry()
169 long syscall_enter_from_user_mode_work(struct pt_regs *regs, long syscall);
172 * syscall_enter_from_user_mode - Establish state and check and handle work
173 * before invoking a syscall
174 * @regs: Pointer to currents pt_regs
175 * @syscall: The syscall number
177 * Invoked from architecture specific syscall entry code with interrupts
178 * disabled. The calling code has to be non-instrumentable. When the
179 * function returns all state is correct, interrupts are enabled and the
180 * subsequent functions can be instrumented.
182 * This is combination of syscall_enter_from_user_mode_prepare() and
183 * syscall_enter_from_user_mode_work().
185 * Returns: The original or a modified syscall number. See
186 * syscall_enter_from_user_mode_work() for further explanation.
188 long syscall_enter_from_user_mode(struct pt_regs *regs, long syscall);
191 * local_irq_enable_exit_to_user - Exit to user variant of local_irq_enable()
192 * @ti_work: Cached TIF flags gathered with interrupts disabled
194 * Defaults to local_irq_enable(). Can be supplied by architecture specific
197 static inline void local_irq_enable_exit_to_user(unsigned long ti_work);
199 #ifndef local_irq_enable_exit_to_user
200 static inline void local_irq_enable_exit_to_user(unsigned long ti_work)
207 * local_irq_disable_exit_to_user - Exit to user variant of local_irq_disable()
209 * Defaults to local_irq_disable(). Can be supplied by architecture specific
212 static inline void local_irq_disable_exit_to_user(void);
214 #ifndef local_irq_disable_exit_to_user
215 static inline void local_irq_disable_exit_to_user(void)
222 * arch_exit_to_user_mode_work - Architecture specific TIF work for exit
224 * @regs: Pointer to currents pt_regs
225 * @ti_work: Cached TIF flags gathered with interrupts disabled
227 * Invoked from exit_to_user_mode_loop() with interrupt enabled
229 * Defaults to NOOP. Can be supplied by architecture specific code.
231 static inline void arch_exit_to_user_mode_work(struct pt_regs *regs,
232 unsigned long ti_work);
234 #ifndef arch_exit_to_user_mode_work
235 static inline void arch_exit_to_user_mode_work(struct pt_regs *regs,
236 unsigned long ti_work)
242 * arch_exit_to_user_mode_prepare - Architecture specific preparation for
244 * @regs: Pointer to currents pt_regs
245 * @ti_work: Cached TIF flags gathered with interrupts disabled
247 * Invoked from exit_to_user_mode_prepare() with interrupt disabled as the last
248 * function before return. Defaults to NOOP.
250 static inline void arch_exit_to_user_mode_prepare(struct pt_regs *regs,
251 unsigned long ti_work);
253 #ifndef arch_exit_to_user_mode_prepare
254 static inline void arch_exit_to_user_mode_prepare(struct pt_regs *regs,
255 unsigned long ti_work)
261 * arch_exit_to_user_mode - Architecture specific final work before
264 * Invoked from exit_to_user_mode() with interrupt disabled as the last
265 * function before return. Defaults to NOOP.
267 * This needs to be __always_inline because it is non-instrumentable code
268 * invoked after context tracking switched to user mode.
270 * An architecture implementation must not do anything complex, no locking
271 * etc. The main purpose is for speculation mitigations.
273 static __always_inline void arch_exit_to_user_mode(void);
275 #ifndef arch_exit_to_user_mode
276 static __always_inline void arch_exit_to_user_mode(void) { }
280 * arch_do_signal_or_restart - Architecture specific signal delivery function
281 * @regs: Pointer to currents pt_regs
282 * @has_signal: actual signal to handle
284 * Invoked from exit_to_user_mode_loop().
286 void arch_do_signal_or_restart(struct pt_regs *regs, bool has_signal);
289 * arch_syscall_exit_tracehook - Wrapper around tracehook_report_syscall_exit()
290 * @regs: Pointer to currents pt_regs
291 * @step: Indicator for single step
293 * Defaults to tracehook_report_syscall_exit(). Can be replaced by
294 * architecture specific code.
296 * Invoked from syscall_exit_to_user_mode()
298 static inline void arch_syscall_exit_tracehook(struct pt_regs *regs, bool step);
300 #ifndef arch_syscall_exit_tracehook
301 static inline void arch_syscall_exit_tracehook(struct pt_regs *regs, bool step)
303 tracehook_report_syscall_exit(regs, step);
308 * exit_to_user_mode - Fixup state when exiting to user mode
310 * Syscall/interrupt exit enables interrupts, but the kernel state is
311 * interrupts disabled when this is invoked. Also tell RCU about it.
313 * 1) Trace interrupts on state
314 * 2) Invoke context tracking if enabled to adjust RCU state
315 * 3) Invoke architecture specific last minute exit code, e.g. speculation
316 * mitigations, etc.: arch_exit_to_user_mode()
317 * 4) Tell lockdep that interrupts are enabled
319 * Invoked from architecture specific code when syscall_exit_to_user_mode()
320 * is not suitable as the last step before returning to userspace. Must be
321 * invoked with interrupts disabled and the caller must be
322 * non-instrumentable.
323 * The caller has to invoke syscall_exit_to_user_mode_work() before this.
325 void exit_to_user_mode(void);
328 * syscall_exit_to_user_mode_work - Handle work before returning to user mode
329 * @regs: Pointer to currents pt_regs
331 * Same as step 1 and 2 of syscall_exit_to_user_mode() but without calling
332 * exit_to_user_mode() to perform the final transition to user mode.
334 * Calling convention is the same as for syscall_exit_to_user_mode() and it
335 * returns with all work handled and interrupts disabled. The caller must
336 * invoke exit_to_user_mode() before actually switching to user mode to
337 * make the final state transitions. Interrupts must stay disabled between
338 * return from this function and the invocation of exit_to_user_mode().
340 void syscall_exit_to_user_mode_work(struct pt_regs *regs);
343 * syscall_exit_to_user_mode - Handle work before returning to user mode
344 * @regs: Pointer to currents pt_regs
346 * Invoked with interrupts enabled and fully valid regs. Returns with all
347 * work handled, interrupts disabled such that the caller can immediately
348 * switch to user mode. Called from architecture specific syscall and ret
352 * 1) One-time syscall exit work:
353 * - rseq syscall exit
356 * - tracehook (single stepping)
358 * 2) Preparatory work
359 * - Exit to user mode loop (common TIF handling). Invokes
360 * arch_exit_to_user_mode_work() for architecture specific TIF work
361 * - Architecture specific one time work arch_exit_to_user_mode_prepare()
362 * - Address limit and lockdep checks
364 * 3) Final transition (lockdep, tracing, context tracking, RCU), i.e. the
365 * functionality in exit_to_user_mode().
367 * This is a combination of syscall_exit_to_user_mode_work() (1,2) and
368 * exit_to_user_mode(). This function is preferred unless there is a
369 * compelling architectural reason to use the separate functions.
371 void syscall_exit_to_user_mode(struct pt_regs *regs);
374 * irqentry_enter_from_user_mode - Establish state before invoking the irq handler
375 * @regs: Pointer to currents pt_regs
377 * Invoked from architecture specific entry code with interrupts disabled.
378 * Can only be called when the interrupt entry came from user mode. The
379 * calling code must be non-instrumentable. When the function returns all
380 * state is correct and the subsequent functions can be instrumented.
382 * The function establishes state (lockdep, RCU (context tracking), tracing)
384 void irqentry_enter_from_user_mode(struct pt_regs *regs);
387 * irqentry_exit_to_user_mode - Interrupt exit work
388 * @regs: Pointer to current's pt_regs
390 * Invoked with interrupts disabled and fully valid regs. Returns with all
391 * work handled, interrupts disabled such that the caller can immediately
392 * switch to user mode. Called from architecture specific interrupt
395 * The call order is #2 and #3 as described in syscall_exit_to_user_mode().
396 * Interrupt exit is not invoking #1 which is the syscall specific one time
399 void irqentry_exit_to_user_mode(struct pt_regs *regs);
401 #ifndef irqentry_state
403 * struct irqentry_state - Opaque object for exception state storage
404 * @exit_rcu: Used exclusively in the irqentry_*() calls; signals whether the
405 * exit path has to invoke rcu_irq_exit().
406 * @lockdep: Used exclusively in the irqentry_nmi_*() calls; ensures that
407 * lockdep state is restored correctly on exit from nmi.
409 * This opaque object is filled in by the irqentry_*_enter() functions and
410 * must be passed back into the corresponding irqentry_*_exit() functions
411 * when the exception is complete.
413 * Callers of irqentry_*_[enter|exit]() must consider this structure opaque
414 * and all members private. Descriptions of the members are provided to aid in
415 * the maintenance of the irqentry_*() functions.
417 typedef struct irqentry_state {
426 * irqentry_enter - Handle state tracking on ordinary interrupt entries
427 * @regs: Pointer to pt_regs of interrupted context
430 * - lockdep irqflag state tracking as low level ASM entry disabled
433 * - Context tracking if the exception hit user mode.
435 * - The hardirq tracer to keep the state consistent as low level ASM
436 * entry disabled interrupts.
438 * As a precondition, this requires that the entry came from user mode,
439 * idle, or a kernel context in which RCU is watching.
441 * For kernel mode entries RCU handling is done conditional. If RCU is
442 * watching then the only RCU requirement is to check whether the tick has
443 * to be restarted. If RCU is not watching then rcu_irq_enter() has to be
444 * invoked on entry and rcu_irq_exit() on exit.
446 * Avoiding the rcu_irq_enter/exit() calls is an optimization but also
447 * solves the problem of kernel mode pagefaults which can schedule, which
448 * is not possible after invoking rcu_irq_enter() without undoing it.
450 * For user mode entries irqentry_enter_from_user_mode() is invoked to
451 * establish the proper context for NOHZ_FULL. Otherwise scheduling on exit
452 * would not be possible.
454 * Returns: An opaque object that must be passed to idtentry_exit()
456 irqentry_state_t noinstr irqentry_enter(struct pt_regs *regs);
459 * irqentry_exit_cond_resched - Conditionally reschedule on return from interrupt
461 * Conditional reschedule with additional sanity checks.
463 void irqentry_exit_cond_resched(void);
464 #ifdef CONFIG_PREEMPT_DYNAMIC
465 DECLARE_STATIC_CALL(irqentry_exit_cond_resched, irqentry_exit_cond_resched);
469 * irqentry_exit - Handle return from exception that used irqentry_enter()
470 * @regs: Pointer to pt_regs (exception entry regs)
471 * @state: Return value from matching call to irqentry_enter()
473 * Depending on the return target (kernel/user) this runs the necessary
474 * preemption and work checks if possible and required and returns to
475 * the caller with interrupts disabled and no further work pending.
477 * This is the last action before returning to the low level ASM code which
478 * just needs to return to the appropriate context.
480 * Counterpart to irqentry_enter().
482 void noinstr irqentry_exit(struct pt_regs *regs, irqentry_state_t state);
485 * irqentry_nmi_enter - Handle NMI entry
486 * @regs: Pointer to currents pt_regs
488 * Similar to irqentry_enter() but taking care of the NMI constraints.
490 irqentry_state_t noinstr irqentry_nmi_enter(struct pt_regs *regs);
493 * irqentry_nmi_exit - Handle return from NMI handling
494 * @regs: Pointer to pt_regs (NMI entry regs)
495 * @irq_state: Return value from matching call to irqentry_nmi_enter()
497 * Last action before returning to the low level assembly code.
499 * Counterpart to irqentry_nmi_enter().
501 void noinstr irqentry_nmi_exit(struct pt_regs *regs, irqentry_state_t irq_state);