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_usermode()
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 | \
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 - Check and handle work before invoking
115 * @regs: Pointer to currents pt_regs
116 * @syscall: The syscall number
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 the subsequent functions can be
123 * Returns: The original or a modified syscall number
125 * If the returned syscall number is -1 then the syscall should be
126 * skipped. In this case the caller may invoke syscall_set_error() or
127 * syscall_set_return_value() first. If neither of those are called and -1
128 * is returned, then the syscall will fail with ENOSYS.
130 * The following functionality is handled here:
132 * 1) Establish state (lockdep, RCU (context tracking), tracing)
133 * 2) TIF flag dependent invocations of arch_syscall_enter_tracehook(),
134 * __secure_computing(), trace_sys_enter()
135 * 3) Invocation of audit_syscall_entry()
137 long syscall_enter_from_user_mode(struct pt_regs *regs, long syscall);
140 * local_irq_enable_exit_to_user - Exit to user variant of local_irq_enable()
141 * @ti_work: Cached TIF flags gathered with interrupts disabled
143 * Defaults to local_irq_enable(). Can be supplied by architecture specific
146 static inline void local_irq_enable_exit_to_user(unsigned long ti_work);
148 #ifndef local_irq_enable_exit_to_user
149 static inline void local_irq_enable_exit_to_user(unsigned long ti_work)
156 * local_irq_disable_exit_to_user - Exit to user variant of local_irq_disable()
158 * Defaults to local_irq_disable(). Can be supplied by architecture specific
161 static inline void local_irq_disable_exit_to_user(void);
163 #ifndef local_irq_disable_exit_to_user
164 static inline void local_irq_disable_exit_to_user(void)
171 * arch_exit_to_user_mode_work - Architecture specific TIF work for exit
173 * @regs: Pointer to currents pt_regs
174 * @ti_work: Cached TIF flags gathered with interrupts disabled
176 * Invoked from exit_to_user_mode_loop() with interrupt enabled
178 * Defaults to NOOP. Can be supplied by architecture specific code.
180 static inline void arch_exit_to_user_mode_work(struct pt_regs *regs,
181 unsigned long ti_work);
183 #ifndef arch_exit_to_user_mode_work
184 static inline void arch_exit_to_user_mode_work(struct pt_regs *regs,
185 unsigned long ti_work)
191 * arch_exit_to_user_mode_prepare - Architecture specific preparation for
193 * @regs: Pointer to currents pt_regs
194 * @ti_work: Cached TIF flags gathered with interrupts disabled
196 * Invoked from exit_to_user_mode_prepare() with interrupt disabled as the last
197 * function before return. Defaults to NOOP.
199 static inline void arch_exit_to_user_mode_prepare(struct pt_regs *regs,
200 unsigned long ti_work);
202 #ifndef arch_exit_to_user_mode_prepare
203 static inline void arch_exit_to_user_mode_prepare(struct pt_regs *regs,
204 unsigned long ti_work)
210 * arch_exit_to_user_mode - Architecture specific final work before
213 * Invoked from exit_to_user_mode() with interrupt disabled as the last
214 * function before return. Defaults to NOOP.
216 * This needs to be __always_inline because it is non-instrumentable code
217 * invoked after context tracking switched to user mode.
219 * An architecture implementation must not do anything complex, no locking
220 * etc. The main purpose is for speculation mitigations.
222 static __always_inline void arch_exit_to_user_mode(void);
224 #ifndef arch_exit_to_user_mode
225 static __always_inline void arch_exit_to_user_mode(void) { }
229 * arch_do_signal - Architecture specific signal delivery function
230 * @regs: Pointer to currents pt_regs
232 * Invoked from exit_to_user_mode_loop().
234 void arch_do_signal(struct pt_regs *regs);
237 * arch_syscall_exit_tracehook - Wrapper around tracehook_report_syscall_exit()
238 * @regs: Pointer to currents pt_regs
239 * @step: Indicator for single step
241 * Defaults to tracehook_report_syscall_exit(). Can be replaced by
242 * architecture specific code.
244 * Invoked from syscall_exit_to_user_mode()
246 static inline void arch_syscall_exit_tracehook(struct pt_regs *regs, bool step);
248 #ifndef arch_syscall_exit_tracehook
249 static inline void arch_syscall_exit_tracehook(struct pt_regs *regs, bool step)
251 tracehook_report_syscall_exit(regs, step);
256 * syscall_exit_to_user_mode - Handle work before returning to user mode
257 * @regs: Pointer to currents pt_regs
259 * Invoked with interrupts enabled and fully valid regs. Returns with all
260 * work handled, interrupts disabled such that the caller can immediately
261 * switch to user mode. Called from architecture specific syscall and ret
265 * 1) One-time syscall exit work:
266 * - rseq syscall exit
269 * - tracehook (single stepping)
271 * 2) Preparatory work
272 * - Exit to user mode loop (common TIF handling). Invokes
273 * arch_exit_to_user_mode_work() for architecture specific TIF work
274 * - Architecture specific one time work arch_exit_to_user_mode_prepare()
275 * - Address limit and lockdep checks
277 * 3) Final transition (lockdep, tracing, context tracking, RCU). Invokes
278 * arch_exit_to_user_mode() to handle e.g. speculation mitigations
280 void syscall_exit_to_user_mode(struct pt_regs *regs);
283 * irqentry_enter_from_user_mode - Establish state before invoking the irq handler
284 * @regs: Pointer to currents pt_regs
286 * Invoked from architecture specific entry code with interrupts disabled.
287 * Can only be called when the interrupt entry came from user mode. The
288 * calling code must be non-instrumentable. When the function returns all
289 * state is correct and the subsequent functions can be instrumented.
291 * The function establishes state (lockdep, RCU (context tracking), tracing)
293 void irqentry_enter_from_user_mode(struct pt_regs *regs);
296 * irqentry_exit_to_user_mode - Interrupt exit work
297 * @regs: Pointer to current's pt_regs
299 * Invoked with interrupts disbled and fully valid regs. Returns with all
300 * work handled, interrupts disabled such that the caller can immediately
301 * switch to user mode. Called from architecture specific interrupt
304 * The call order is #2 and #3 as described in syscall_exit_to_user_mode().
305 * Interrupt exit is not invoking #1 which is the syscall specific one time
308 void irqentry_exit_to_user_mode(struct pt_regs *regs);
310 #ifndef irqentry_state
311 typedef struct irqentry_state {
317 * irqentry_enter - Handle state tracking on ordinary interrupt entries
318 * @regs: Pointer to pt_regs of interrupted context
321 * - lockdep irqflag state tracking as low level ASM entry disabled
324 * - Context tracking if the exception hit user mode.
326 * - The hardirq tracer to keep the state consistent as low level ASM
327 * entry disabled interrupts.
329 * As a precondition, this requires that the entry came from user mode,
330 * idle, or a kernel context in which RCU is watching.
332 * For kernel mode entries RCU handling is done conditional. If RCU is
333 * watching then the only RCU requirement is to check whether the tick has
334 * to be restarted. If RCU is not watching then rcu_irq_enter() has to be
335 * invoked on entry and rcu_irq_exit() on exit.
337 * Avoiding the rcu_irq_enter/exit() calls is an optimization but also
338 * solves the problem of kernel mode pagefaults which can schedule, which
339 * is not possible after invoking rcu_irq_enter() without undoing it.
341 * For user mode entries irqentry_enter_from_user_mode() is invoked to
342 * establish the proper context for NOHZ_FULL. Otherwise scheduling on exit
343 * would not be possible.
345 * Returns: An opaque object that must be passed to idtentry_exit()
347 irqentry_state_t noinstr irqentry_enter(struct pt_regs *regs);
350 * irqentry_exit_cond_resched - Conditionally reschedule on return from interrupt
352 * Conditional reschedule with additional sanity checks.
354 void irqentry_exit_cond_resched(void);
357 * irqentry_exit - Handle return from exception that used irqentry_enter()
358 * @regs: Pointer to pt_regs (exception entry regs)
359 * @state: Return value from matching call to irqentry_enter()
361 * Depending on the return target (kernel/user) this runs the necessary
362 * preemption and work checks if possible and reguired and returns to
363 * the caller with interrupts disabled and no further work pending.
365 * This is the last action before returning to the low level ASM code which
366 * just needs to return to the appropriate context.
368 * Counterpart to irqentry_enter().
370 void noinstr irqentry_exit(struct pt_regs *regs, irqentry_state_t state);