2 * SPDX-License-Identifier: BSD-4-Clause
4 * Copyright (C) 1994, David Greenman
5 * Copyright (c) 1990, 1993
6 * The Regents of the University of California. All rights reserved.
7 * Copyright (c) 2007 The FreeBSD Foundation
9 * This code is derived from software contributed to Berkeley by
10 * the University of Utah, and William Jolitz.
12 * Portions of this software were developed by A. Joseph Koshy under
13 * sponsorship from the FreeBSD Foundation and Google, Inc.
15 * Redistribution and use in source and binary forms, with or without
16 * modification, are permitted provided that the following conditions
18 * 1. Redistributions of source code must retain the above copyright
19 * notice, this list of conditions and the following disclaimer.
20 * 2. Redistributions in binary form must reproduce the above copyright
21 * notice, this list of conditions and the following disclaimer in the
22 * documentation and/or other materials provided with the distribution.
23 * 3. All advertising materials mentioning features or use of this software
24 * must display the following acknowledgement:
25 * This product includes software developed by the University of
26 * California, Berkeley and its contributors.
27 * 4. Neither the name of the University nor the names of its contributors
28 * may be used to endorse or promote products derived from this software
29 * without specific prior written permission.
31 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
32 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
33 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
34 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
35 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
36 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
37 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
38 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
39 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
40 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
43 * from: @(#)trap.c 7.4 (Berkeley) 5/13/91
46 #include <sys/cdefs.h>
47 __FBSDID("$FreeBSD$");
49 #include "opt_hwpmc_hooks.h"
50 #include "opt_ktrace.h"
51 #include "opt_sched.h"
53 #include <sys/param.h>
55 #include <sys/capsicum.h>
56 #include <sys/event.h>
57 #include <sys/kernel.h>
60 #include <sys/mutex.h>
61 #include <sys/pmckern.h>
64 #include <sys/ptrace.h>
65 #include <sys/racct.h>
66 #include <sys/resourcevar.h>
67 #include <sys/sched.h>
68 #include <sys/signalvar.h>
69 #include <sys/syscall.h>
70 #include <sys/syscallsubr.h>
71 #include <sys/sysent.h>
72 #include <sys/systm.h>
73 #include <sys/vmmeter.h>
76 #include <sys/ktrace.h>
78 #include <security/audit/audit.h>
80 #include <machine/cpu.h>
87 #include <sys/pmckern.h>
90 #include <security/mac/mac_framework.h>
93 * Define the code needed before returning to user mode, for trap and
97 userret(struct thread *td, struct trapframe *frame)
99 struct proc *p = td->td_proc;
101 CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid,
103 KASSERT((p->p_flag & P_WEXIT) == 0,
104 ("Exiting process returns to usermode"));
107 * Check that we called signotify() enough. For
108 * multi-threaded processes, where signal distribution might
109 * change due to other threads changing sigmask, the check is
110 * racy and cannot be performed reliably.
111 * If current process is vfork child, indicated by P_PPWAIT, then
112 * issignal() ignores stops, so we block the check to avoid
113 * classifying pending signals.
115 if (p->p_numthreads == 1) {
118 if ((p->p_flag & P_PPWAIT) == 0 &&
119 (td->td_pflags & TDP_SIGFASTBLOCK) == 0 &&
120 SIGPENDING(td) && !td_ast_pending(td, TDA_AST) &&
121 !td_ast_pending(td, TDA_SIG)) {
124 "failed to set signal flags for ast p %p "
125 "td %p td_ast %#x fl %#x",
126 p, td, td->td_ast, td->td_flags);
134 * Charge system time if profiling.
136 if (__predict_false(p->p_flag & P_PROFIL))
137 addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio);
140 if (PMC_THREAD_HAS_SAMPLES(td))
141 PMC_CALL_HOOK(td, PMC_FN_THR_USERRET, NULL);
145 * @gallatin is adament that this needs to go here, I
146 * am not so sure. Running hpts is a lot like
147 * a lro_flush() that happens while a user process
148 * is running. But he may know best so I will go
149 * with his view of accounting. :-)
154 * Let the scheduler adjust our priority etc.
159 * Check for misbehavior.
161 * In case there is a callchain tracing ongoing because of
162 * hwpmc(4), skip the scheduler pinning check.
163 * hwpmc(4) subsystem, infact, will collect callchain informations
164 * at ast() checkpoint, which is past userret().
166 WITNESS_WARN(WARN_PANIC, NULL, "userret: returning");
167 KASSERT(td->td_critnest == 0,
168 ("userret: Returning in a critical section"));
169 KASSERT(td->td_locks == 0,
170 ("userret: Returning with %d locks held", td->td_locks));
171 KASSERT(td->td_rw_rlocks == 0,
172 ("userret: Returning with %d rwlocks held in read mode",
174 KASSERT(td->td_sx_slocks == 0,
175 ("userret: Returning with %d sx locks held in shared mode",
177 KASSERT(td->td_lk_slocks == 0,
178 ("userret: Returning with %d lockmanager locks held in shared mode",
180 KASSERT((td->td_pflags & TDP_NOFAULTING) == 0,
181 ("userret: Returning with pagefaults disabled"));
182 if (__predict_false(!THREAD_CAN_SLEEP())) {
184 epoch_trace_list(curthread);
186 KASSERT(0, ("userret: Returning with sleep disabled"));
188 KASSERT(td->td_pinned == 0 || (td->td_pflags & TDP_CALLCHAIN) != 0,
189 ("userret: Returning with with pinned thread"));
190 KASSERT(td->td_vp_reserved == NULL,
191 ("userret: Returning with preallocated vnode"));
192 KASSERT((td->td_flags & (TDF_SBDRY | TDF_SEINTR | TDF_SERESTART)) == 0,
193 ("userret: Returning with stop signals deferred"));
194 KASSERT(td->td_vslock_sz == 0,
195 ("userret: Returning with vslock-wired space"));
197 /* Unfortunately td_vnet_lpush needs VNET_DEBUG. */
198 VNET_ASSERT(curvnet == NULL,
199 ("%s: Returning on td %p (pid %d, %s) with vnet %p set in %s",
200 __func__, td, p->p_pid, td->td_name, curvnet,
201 (td->td_vnet_lpush != NULL) ? td->td_vnet_lpush : "N/A"));
206 ast_prep(struct thread *td, int tda __unused)
210 if (td->td_cowgen != atomic_load_int(&td->td_proc->p_cowgen))
211 thread_cow_update(td);
218 void (*ae_f)(struct thread *td, int ast);
221 _Static_assert(TDAI(TDA_MAX) <= UINT_MAX, "Too many ASTs");
223 static struct ast_entry ast_entries[TDA_MAX] __read_mostly = {
224 [TDA_AST] = { .ae_f = ast_prep, .ae_flags = ASTR_UNCOND},
228 ast_register(int ast, int flags, int tdp,
229 void (*f)(struct thread *, int asts))
231 struct ast_entry *ae;
233 MPASS(ast < TDA_MAX);
234 MPASS((flags & ASTR_TDP) == 0 || ((flags & ASTR_ASTF_REQUIRED) != 0
235 && __bitcount(tdp) == 1));
236 ae = &ast_entries[ast];
237 MPASS(ae->ae_f == NULL);
238 ae->ae_flags = flags;
240 atomic_interrupt_fence();
245 * XXXKIB Note that the deregistration of an AST handler does not
246 * drain threads possibly executing it, which affects unloadable
247 * modules. The issue is either handled by the subsystem using
248 * handlers, or simply ignored. Fixing the problem is considered not
249 * worth the overhead.
252 ast_deregister(int ast)
254 struct ast_entry *ae;
256 MPASS(ast < TDA_MAX);
257 ae = &ast_entries[ast];
258 MPASS(ae->ae_f != NULL);
260 atomic_interrupt_fence();
266 ast_sched_locked(struct thread *td, int tda)
268 THREAD_LOCK_ASSERT(td, MA_OWNED);
269 MPASS(tda < TDA_MAX);
271 td->td_ast |= TDAI(tda);
275 ast_unsched_locked(struct thread *td, int tda)
277 THREAD_LOCK_ASSERT(td, MA_OWNED);
278 MPASS(tda < TDA_MAX);
280 td->td_ast &= ~TDAI(tda);
284 ast_sched(struct thread *td, int tda)
287 ast_sched_locked(td, tda);
292 ast_sched_mask(struct thread *td, int ast)
300 ast_handler_calc_tdp_run(struct thread *td, const struct ast_entry *ae)
302 return ((ae->ae_flags & ASTR_TDP) == 0 ||
303 (td->td_pflags & ae->ae_tdp) != 0);
307 * Process an asynchronous software trap.
310 ast_handler(struct thread *td, struct trapframe *framep, bool dtor)
312 struct ast_entry *ae;
313 void (*f)(struct thread *td, int asts);
317 if (framep != NULL) {
318 kmsan_mark(framep, sizeof(*framep), KMSAN_STATE_INITED);
319 td->td_frame = framep;
322 if (__predict_true(!dtor)) {
323 WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode");
324 mtx_assert(&Giant, MA_NOTOWNED);
325 THREAD_LOCK_ASSERT(td, MA_NOTOWNED);
328 * This updates the td_ast for the checks below in one
329 * atomic operation with turning off all scheduled AST's.
330 * If another AST is triggered while we are handling the
331 * AST's saved in td_ast, the td_ast is again non-zero and
332 * ast() will be called again.
340 * The td thread's td_lock is not guaranteed to exist,
341 * the thread might be not initialized enough when it's
342 * destructor is called. It is safe to read and
343 * update td_ast without locking since the thread is
344 * not runnable or visible to other threads.
350 CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, td->td_proc->p_pid,
351 td->td_proc->p_comm);
352 KASSERT(framep == NULL || TRAPF_USERMODE(framep),
353 ("ast in kernel mode"));
355 for (a = 0; a < nitems(ast_entries); a++) {
356 ae = &ast_entries[a];
360 atomic_interrupt_fence();
363 if (__predict_false(framep == NULL)) {
364 if ((ae->ae_flags & ASTR_KCLEAR) != 0)
365 run = ast_handler_calc_tdp_run(td, ae);
367 if ((ae->ae_flags & ASTR_UNCOND) != 0)
369 else if ((ae->ae_flags & ASTR_ASTF_REQUIRED) != 0 &&
370 (td_ast & TDAI(a)) != 0)
371 run = ast_handler_calc_tdp_run(td, ae);
379 ast(struct trapframe *framep)
384 ast_handler(td, framep, false);
389 ast_kclear(struct thread *td)
391 ast_handler(td, NULL, td != curthread);
395 syscallname(struct proc *p, u_int code)
397 static const char unknown[] = "unknown";
398 struct sysentvec *sv;
401 if (sv->sv_syscallnames == NULL || code >= sv->sv_size)
403 return (sv->sv_syscallnames[code]);