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_EMU
17 # define _TIF_SYSCALL_EMU (0)
20 #ifndef _TIF_SYSCALL_TRACEPOINT
21 # define _TIF_SYSCALL_TRACEPOINT (0)
25 # define _TIF_SECCOMP (0)
28 #ifndef _TIF_SYSCALL_AUDIT
29 # define _TIF_SYSCALL_AUDIT (0)
32 #ifndef _TIF_PATCH_PENDING
33 # define _TIF_PATCH_PENDING (0)
37 # define _TIF_UPROBE (0)
41 * TIF flags handled in syscall_enter_from_user_mode()
43 #ifndef ARCH_SYSCALL_ENTER_WORK
44 # define ARCH_SYSCALL_ENTER_WORK (0)
47 #define SYSCALL_ENTER_WORK \
48 (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_AUDIT | _TIF_SECCOMP | \
49 _TIF_SYSCALL_TRACEPOINT | _TIF_SYSCALL_EMU | \
50 ARCH_SYSCALL_ENTER_WORK)
53 * TIF flags handled in syscall_exit_to_user_mode()
55 #ifndef ARCH_SYSCALL_EXIT_WORK
56 # define ARCH_SYSCALL_EXIT_WORK (0)
59 #define SYSCALL_EXIT_WORK \
60 (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_AUDIT | \
61 _TIF_SYSCALL_TRACEPOINT | ARCH_SYSCALL_EXIT_WORK)
64 * TIF flags handled in exit_to_user_mode_loop()
66 #ifndef ARCH_EXIT_TO_USER_MODE_WORK
67 # define ARCH_EXIT_TO_USER_MODE_WORK (0)
70 #define EXIT_TO_USER_MODE_WORK \
71 (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_UPROBE | \
72 _TIF_NEED_RESCHED | _TIF_PATCH_PENDING | _TIF_NOTIFY_SIGNAL | \
73 ARCH_EXIT_TO_USER_MODE_WORK)
76 * arch_check_user_regs - Architecture specific sanity check for user mode regs
77 * @regs: Pointer to currents pt_regs
79 * Defaults to an empty implementation. Can be replaced by architecture
82 * Invoked from syscall_enter_from_user_mode() in the non-instrumentable
83 * section. Use __always_inline so the compiler cannot push it out of line
84 * and make it instrumentable.
86 static __always_inline void arch_check_user_regs(struct pt_regs *regs);
88 #ifndef arch_check_user_regs
89 static __always_inline void arch_check_user_regs(struct pt_regs *regs) {}
93 * arch_syscall_enter_tracehook - Wrapper around tracehook_report_syscall_entry()
94 * @regs: Pointer to currents pt_regs
96 * Returns: 0 on success or an error code to skip the syscall.
98 * Defaults to tracehook_report_syscall_entry(). Can be replaced by
99 * architecture specific code.
101 * Invoked from syscall_enter_from_user_mode()
103 static inline __must_check int arch_syscall_enter_tracehook(struct pt_regs *regs);
105 #ifndef arch_syscall_enter_tracehook
106 static inline __must_check int arch_syscall_enter_tracehook(struct pt_regs *regs)
108 return tracehook_report_syscall_entry(regs);
113 * syscall_enter_from_user_mode_prepare - Establish state and enable interrupts
114 * @regs: Pointer to currents pt_regs
116 * Invoked from architecture specific syscall entry code with interrupts
117 * disabled. The calling code has to be non-instrumentable. When the
118 * function returns all state is correct, interrupts are enabled and the
119 * subsequent functions can be instrumented.
121 * This handles lockdep, RCU (context tracking) and tracing state.
123 * This is invoked when there is extra architecture specific functionality
124 * to be done between establishing state and handling user mode entry work.
126 void syscall_enter_from_user_mode_prepare(struct pt_regs *regs);
129 * syscall_enter_from_user_mode_work - Check and handle work before invoking
131 * @regs: Pointer to currents pt_regs
132 * @syscall: The syscall number
134 * Invoked from architecture specific syscall entry code with interrupts
135 * enabled after invoking syscall_enter_from_user_mode_prepare() and extra
136 * architecture specific work.
138 * Returns: The original or a modified syscall number
140 * If the returned syscall number is -1 then the syscall should be
141 * skipped. In this case the caller may invoke syscall_set_error() or
142 * syscall_set_return_value() first. If neither of those are called and -1
143 * is returned, then the syscall will fail with ENOSYS.
145 * It handles the following work items:
147 * 1) TIF flag dependent invocations of arch_syscall_enter_tracehook(),
148 * __secure_computing(), trace_sys_enter()
149 * 2) Invocation of audit_syscall_entry()
151 long syscall_enter_from_user_mode_work(struct pt_regs *regs, long syscall);
154 * syscall_enter_from_user_mode - Establish state and check and handle work
155 * before invoking a syscall
156 * @regs: Pointer to currents pt_regs
157 * @syscall: The syscall number
159 * Invoked from architecture specific syscall entry code with interrupts
160 * disabled. The calling code has to be non-instrumentable. When the
161 * function returns all state is correct, interrupts are enabled and the
162 * subsequent functions can be instrumented.
164 * This is combination of syscall_enter_from_user_mode_prepare() and
165 * syscall_enter_from_user_mode_work().
167 * Returns: The original or a modified syscall number. See
168 * syscall_enter_from_user_mode_work() for further explanation.
170 long syscall_enter_from_user_mode(struct pt_regs *regs, long syscall);
173 * local_irq_enable_exit_to_user - Exit to user variant of local_irq_enable()
174 * @ti_work: Cached TIF flags gathered with interrupts disabled
176 * Defaults to local_irq_enable(). Can be supplied by architecture specific
179 static inline void local_irq_enable_exit_to_user(unsigned long ti_work);
181 #ifndef local_irq_enable_exit_to_user
182 static inline void local_irq_enable_exit_to_user(unsigned long ti_work)
189 * local_irq_disable_exit_to_user - Exit to user variant of local_irq_disable()
191 * Defaults to local_irq_disable(). Can be supplied by architecture specific
194 static inline void local_irq_disable_exit_to_user(void);
196 #ifndef local_irq_disable_exit_to_user
197 static inline void local_irq_disable_exit_to_user(void)
204 * arch_exit_to_user_mode_work - Architecture specific TIF work for exit
206 * @regs: Pointer to currents pt_regs
207 * @ti_work: Cached TIF flags gathered with interrupts disabled
209 * Invoked from exit_to_user_mode_loop() with interrupt enabled
211 * Defaults to NOOP. Can be supplied by architecture specific code.
213 static inline void arch_exit_to_user_mode_work(struct pt_regs *regs,
214 unsigned long ti_work);
216 #ifndef arch_exit_to_user_mode_work
217 static inline void arch_exit_to_user_mode_work(struct pt_regs *regs,
218 unsigned long ti_work)
224 * arch_exit_to_user_mode_prepare - Architecture specific preparation for
226 * @regs: Pointer to currents pt_regs
227 * @ti_work: Cached TIF flags gathered with interrupts disabled
229 * Invoked from exit_to_user_mode_prepare() with interrupt disabled as the last
230 * function before return. Defaults to NOOP.
232 static inline void arch_exit_to_user_mode_prepare(struct pt_regs *regs,
233 unsigned long ti_work);
235 #ifndef arch_exit_to_user_mode_prepare
236 static inline void arch_exit_to_user_mode_prepare(struct pt_regs *regs,
237 unsigned long ti_work)
243 * arch_exit_to_user_mode - Architecture specific final work before
246 * Invoked from exit_to_user_mode() with interrupt disabled as the last
247 * function before return. Defaults to NOOP.
249 * This needs to be __always_inline because it is non-instrumentable code
250 * invoked after context tracking switched to user mode.
252 * An architecture implementation must not do anything complex, no locking
253 * etc. The main purpose is for speculation mitigations.
255 static __always_inline void arch_exit_to_user_mode(void);
257 #ifndef arch_exit_to_user_mode
258 static __always_inline void arch_exit_to_user_mode(void) { }
262 * arch_do_signal_or_restart - Architecture specific signal delivery function
263 * @regs: Pointer to currents pt_regs
264 * @has_signal: actual signal to handle
266 * Invoked from exit_to_user_mode_loop().
268 void arch_do_signal_or_restart(struct pt_regs *regs, bool has_signal);
271 * arch_syscall_exit_tracehook - Wrapper around tracehook_report_syscall_exit()
272 * @regs: Pointer to currents pt_regs
273 * @step: Indicator for single step
275 * Defaults to tracehook_report_syscall_exit(). Can be replaced by
276 * architecture specific code.
278 * Invoked from syscall_exit_to_user_mode()
280 static inline void arch_syscall_exit_tracehook(struct pt_regs *regs, bool step);
282 #ifndef arch_syscall_exit_tracehook
283 static inline void arch_syscall_exit_tracehook(struct pt_regs *regs, bool step)
285 tracehook_report_syscall_exit(regs, step);
290 * syscall_exit_to_user_mode - Handle work before returning to user mode
291 * @regs: Pointer to currents pt_regs
293 * Invoked with interrupts enabled and fully valid regs. Returns with all
294 * work handled, interrupts disabled such that the caller can immediately
295 * switch to user mode. Called from architecture specific syscall and ret
299 * 1) One-time syscall exit work:
300 * - rseq syscall exit
303 * - tracehook (single stepping)
305 * 2) Preparatory work
306 * - Exit to user mode loop (common TIF handling). Invokes
307 * arch_exit_to_user_mode_work() for architecture specific TIF work
308 * - Architecture specific one time work arch_exit_to_user_mode_prepare()
309 * - Address limit and lockdep checks
311 * 3) Final transition (lockdep, tracing, context tracking, RCU). Invokes
312 * arch_exit_to_user_mode() to handle e.g. speculation mitigations
314 void syscall_exit_to_user_mode(struct pt_regs *regs);
317 * irqentry_enter_from_user_mode - Establish state before invoking the irq handler
318 * @regs: Pointer to currents pt_regs
320 * Invoked from architecture specific entry code with interrupts disabled.
321 * Can only be called when the interrupt entry came from user mode. The
322 * calling code must be non-instrumentable. When the function returns all
323 * state is correct and the subsequent functions can be instrumented.
325 * The function establishes state (lockdep, RCU (context tracking), tracing)
327 void irqentry_enter_from_user_mode(struct pt_regs *regs);
330 * irqentry_exit_to_user_mode - Interrupt exit work
331 * @regs: Pointer to current's pt_regs
333 * Invoked with interrupts disbled and fully valid regs. Returns with all
334 * work handled, interrupts disabled such that the caller can immediately
335 * switch to user mode. Called from architecture specific interrupt
338 * The call order is #2 and #3 as described in syscall_exit_to_user_mode().
339 * Interrupt exit is not invoking #1 which is the syscall specific one time
342 void irqentry_exit_to_user_mode(struct pt_regs *regs);
344 #ifndef irqentry_state
346 * struct irqentry_state - Opaque object for exception state storage
347 * @exit_rcu: Used exclusively in the irqentry_*() calls; signals whether the
348 * exit path has to invoke rcu_irq_exit().
349 * @lockdep: Used exclusively in the irqentry_nmi_*() calls; ensures that
350 * lockdep state is restored correctly on exit from nmi.
352 * This opaque object is filled in by the irqentry_*_enter() functions and
353 * must be passed back into the corresponding irqentry_*_exit() functions
354 * when the exception is complete.
356 * Callers of irqentry_*_[enter|exit]() must consider this structure opaque
357 * and all members private. Descriptions of the members are provided to aid in
358 * the maintenance of the irqentry_*() functions.
360 typedef struct irqentry_state {
369 * irqentry_enter - Handle state tracking on ordinary interrupt entries
370 * @regs: Pointer to pt_regs of interrupted context
373 * - lockdep irqflag state tracking as low level ASM entry disabled
376 * - Context tracking if the exception hit user mode.
378 * - The hardirq tracer to keep the state consistent as low level ASM
379 * entry disabled interrupts.
381 * As a precondition, this requires that the entry came from user mode,
382 * idle, or a kernel context in which RCU is watching.
384 * For kernel mode entries RCU handling is done conditional. If RCU is
385 * watching then the only RCU requirement is to check whether the tick has
386 * to be restarted. If RCU is not watching then rcu_irq_enter() has to be
387 * invoked on entry and rcu_irq_exit() on exit.
389 * Avoiding the rcu_irq_enter/exit() calls is an optimization but also
390 * solves the problem of kernel mode pagefaults which can schedule, which
391 * is not possible after invoking rcu_irq_enter() without undoing it.
393 * For user mode entries irqentry_enter_from_user_mode() is invoked to
394 * establish the proper context for NOHZ_FULL. Otherwise scheduling on exit
395 * would not be possible.
397 * Returns: An opaque object that must be passed to idtentry_exit()
399 irqentry_state_t noinstr irqentry_enter(struct pt_regs *regs);
402 * irqentry_exit_cond_resched - Conditionally reschedule on return from interrupt
404 * Conditional reschedule with additional sanity checks.
406 void irqentry_exit_cond_resched(void);
409 * irqentry_exit - Handle return from exception that used irqentry_enter()
410 * @regs: Pointer to pt_regs (exception entry regs)
411 * @state: Return value from matching call to irqentry_enter()
413 * Depending on the return target (kernel/user) this runs the necessary
414 * preemption and work checks if possible and reguired and returns to
415 * the caller with interrupts disabled and no further work pending.
417 * This is the last action before returning to the low level ASM code which
418 * just needs to return to the appropriate context.
420 * Counterpart to irqentry_enter().
422 void noinstr irqentry_exit(struct pt_regs *regs, irqentry_state_t state);
425 * irqentry_nmi_enter - Handle NMI entry
426 * @regs: Pointer to currents pt_regs
428 * Similar to irqentry_enter() but taking care of the NMI constraints.
430 irqentry_state_t noinstr irqentry_nmi_enter(struct pt_regs *regs);
433 * irqentry_nmi_exit - Handle return from NMI handling
434 * @regs: Pointer to pt_regs (NMI entry regs)
435 * @irq_state: Return value from matching call to irqentry_nmi_enter()
437 * Last action before returning to the low level assmenbly code.
439 * Counterpart to irqentry_nmi_enter().
441 void noinstr irqentry_nmi_exit(struct pt_regs *regs, irqentry_state_t irq_state);