1 /*-
2 * SPDX-License-Identifier: BSD-4-Clause
3 *
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
8 *
9 * This code is derived from software contributed to Berkeley by
10 * the University of Utah, and William Jolitz.
11 *
12 * Portions of this software were developed by A. Joseph Koshy under
13 * sponsorship from the FreeBSD Foundation and Google, Inc.
14 *
15 * Redistribution and use in source and binary forms, with or without
16 * modification, are permitted provided that the following conditions
17 * are met:
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.
30 *
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
41 * SUCH DAMAGE.
42 *
43 * from: @(#)trap.c 7.4 (Berkeley) 5/13/91
44 */
45
46 #include <sys/cdefs.h>
47 #include "opt_hwpmc_hooks.h"
48 #include "opt_ktrace.h"
49 #include "opt_sched.h"
50
51 #include <sys/param.h>
52 #include <sys/bus.h>
53 #include <sys/capsicum.h>
54 #include <sys/event.h>
55 #include <sys/kernel.h>
56 #include <sys/lock.h>
57 #include <sys/mutex.h>
58 #include <sys/pmckern.h>
59 #include <sys/proc.h>
60 #include <sys/ktr.h>
61 #include <sys/ptrace.h>
62 #include <sys/racct.h>
63 #include <sys/resourcevar.h>
64 #include <sys/sched.h>
65 #include <sys/signalvar.h>
66 #include <sys/syscall.h>
67 #include <sys/syscallsubr.h>
68 #include <sys/sysent.h>
69 #include <sys/systm.h>
70 #include <sys/vmmeter.h>
71 #ifdef KTRACE
72 #include <sys/uio.h>
73 #include <sys/ktrace.h>
74 #endif
75 #include <security/audit/audit.h>
76
77 #include <machine/cpu.h>
78
79 #ifdef VIMAGE
80 #include <net/vnet.h>
81 #endif
82
83 #ifdef HWPMC_HOOKS
84 #include <sys/pmckern.h>
85 #endif
86
87 #include <security/mac/mac_framework.h>
88
89 void (*softdep_ast_cleanup)(struct thread *);
90
91 /*
92 * Define the code needed before returning to user mode, for trap and
93 * syscall.
94 */
95 void
userret(struct thread * td,struct trapframe * frame)96 userret(struct thread *td, struct trapframe *frame)
97 {
98 struct proc *p = td->td_proc;
99
100 CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid,
101 td->td_name);
102 KASSERT((p->p_flag & P_WEXIT) == 0,
103 ("Exiting process returns to usermode"));
104 #ifdef DIAGNOSTIC
105 /*
106 * Check that we called signotify() enough. For
107 * multi-threaded processes, where signal distribution might
108 * change due to other threads changing sigmask, the check is
109 * racy and cannot be performed reliably.
110 * If current process is vfork child, indicated by P_PPWAIT, then
111 * issignal() ignores stops, so we block the check to avoid
112 * classifying pending signals.
113 */
114 if (p->p_numthreads == 1) {
115 PROC_LOCK(p);
116 thread_lock(td);
117 if ((p->p_flag & P_PPWAIT) == 0 &&
118 (td->td_pflags & TDP_SIGFASTBLOCK) == 0) {
119 if (SIGPENDING(td) && (td->td_flags &
120 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) !=
121 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) {
122 thread_unlock(td);
123 panic(
124 "failed to set signal flags for ast p %p td %p fl %x",
125 p, td, td->td_flags);
126 }
127 }
128 thread_unlock(td);
129 PROC_UNLOCK(p);
130 }
131 #endif
132
133 /*
134 * Charge system time if profiling.
135 */
136 if (__predict_false(p->p_flag & P_PROFIL))
137 addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio);
138
139 #ifdef HWPMC_HOOKS
140 if (PMC_THREAD_HAS_SAMPLES(td))
141 PMC_CALL_HOOK(td, PMC_FN_THR_USERRET, NULL);
142 #endif
143 #ifdef TCPHPTS
144 /*
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. :-)
150 */
151 tcp_run_hpts();
152 #endif
153 /*
154 * Let the scheduler adjust our priority etc.
155 */
156 sched_userret(td);
157
158 /*
159 * Check for misbehavior.
160 *
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().
165 */
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",
173 td->td_rw_rlocks));
174 KASSERT(td->td_sx_slocks == 0,
175 ("userret: Returning with %d sx locks held in shared mode",
176 td->td_sx_slocks));
177 KASSERT(td->td_lk_slocks == 0,
178 ("userret: Returning with %d lockmanager locks held in shared mode",
179 td->td_lk_slocks));
180 KASSERT((td->td_pflags & TDP_NOFAULTING) == 0,
181 ("userret: Returning with pagefaults disabled"));
182 if (__predict_false(!THREAD_CAN_SLEEP())) {
183 #ifdef EPOCH_TRACE
184 epoch_trace_list(curthread);
185 #endif
186 KASSERT(0, ("userret: Returning with sleep disabled"));
187 }
188 KASSERT(td->td_pinned == 0 || (td->td_pflags & TDP_CALLCHAIN) != 0,
189 ("userret: Returning 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"));
196 #ifdef VIMAGE
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"));
202 #endif
203 }
204
205 /*
206 * Process an asynchronous software trap.
207 * This is relatively easy.
208 * This function will return with preemption disabled.
209 */
210 void
ast(struct trapframe * framep)211 ast(struct trapframe *framep)
212 {
213 struct thread *td;
214 struct proc *p;
215 int flags, old_boundary, sig;
216 bool resched_sigs;
217
218 td = curthread;
219 p = td->td_proc;
220
221 CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid,
222 p->p_comm);
223 KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode"));
224 WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode");
225 mtx_assert(&Giant, MA_NOTOWNED);
226 THREAD_LOCK_ASSERT(td, MA_NOTOWNED);
227 td->td_frame = framep;
228 td->td_pticks = 0;
229
230 /*
231 * This updates the td_flag's for the checks below in one
232 * "atomic" operation with turning off the astpending flag.
233 * If another AST is triggered while we are handling the
234 * AST's saved in flags, the astpending flag will be set and
235 * ast() will be called again.
236 */
237 thread_lock(td);
238 flags = td->td_flags;
239 td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK |
240 TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND |
241 TDF_KQTICKLED);
242 thread_unlock(td);
243 VM_CNT_INC(v_trap);
244
245 if (td->td_cowgen != p->p_cowgen)
246 thread_cow_update(td);
247 if (td->td_pflags & TDP_OWEUPC && p->p_flag & P_PROFIL) {
248 addupc_task(td, td->td_profil_addr, td->td_profil_ticks);
249 td->td_profil_ticks = 0;
250 td->td_pflags &= ~TDP_OWEUPC;
251 }
252 #ifdef HWPMC_HOOKS
253 /* Handle Software PMC callchain capture. */
254 if (PMC_IS_PENDING_CALLCHAIN(td))
255 PMC_CALL_HOOK_UNLOCKED(td, PMC_FN_USER_CALLCHAIN_SOFT, (void *) framep);
256 #endif
257 if (flags & TDF_ALRMPEND) {
258 PROC_LOCK(p);
259 kern_psignal(p, SIGVTALRM);
260 PROC_UNLOCK(p);
261 }
262 if (flags & TDF_PROFPEND) {
263 PROC_LOCK(p);
264 kern_psignal(p, SIGPROF);
265 PROC_UNLOCK(p);
266 }
267 #ifdef MAC
268 if (flags & TDF_MACPEND)
269 mac_thread_userret(td);
270 #endif
271 if (flags & TDF_NEEDRESCHED) {
272 #ifdef KTRACE
273 if (KTRPOINT(td, KTR_CSW))
274 ktrcsw(1, 1, __func__);
275 #endif
276 thread_lock(td);
277 sched_prio(td, td->td_user_pri);
278 mi_switch(SW_INVOL | SWT_NEEDRESCHED);
279 #ifdef KTRACE
280 if (KTRPOINT(td, KTR_CSW))
281 ktrcsw(0, 1, __func__);
282 #endif
283 }
284
285 td_softdep_cleanup(td);
286 MPASS(td->td_su == NULL);
287
288 /*
289 * If this thread tickled GEOM, we need to wait for the giggling to
290 * stop before we return to userland
291 */
292 if (__predict_false(td->td_pflags & TDP_GEOM))
293 g_waitidle();
294
295 #ifdef DIAGNOSTIC
296 if (p->p_numthreads == 1 && (flags & TDF_NEEDSIGCHK) == 0) {
297 PROC_LOCK(p);
298 thread_lock(td);
299 /*
300 * Note that TDF_NEEDSIGCHK should be re-read from
301 * td_flags, since signal might have been delivered
302 * after we cleared td_flags above. This is one of
303 * the reason for looping check for AST condition.
304 * See comment in userret() about P_PPWAIT.
305 */
306 if ((p->p_flag & P_PPWAIT) == 0 &&
307 (td->td_pflags & TDP_SIGFASTBLOCK) == 0) {
308 if (SIGPENDING(td) && (td->td_flags &
309 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) !=
310 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) {
311 thread_unlock(td); /* fix dumps */
312 panic(
313 "failed2 to set signal flags for ast p %p td %p fl %x %x",
314 p, td, flags, td->td_flags);
315 }
316 }
317 thread_unlock(td);
318 PROC_UNLOCK(p);
319 }
320 #endif
321
322 /*
323 * Check for signals. Unlocked reads of p_pendingcnt or
324 * p_siglist might cause process-directed signal to be handled
325 * later.
326 */
327 if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 ||
328 !SIGISEMPTY(p->p_siglist)) {
329 sigfastblock_fetch(td);
330 PROC_LOCK(p);
331 old_boundary = ~TDB_BOUNDARY | (td->td_dbgflags & TDB_BOUNDARY);
332 td->td_dbgflags |= TDB_BOUNDARY;
333 mtx_lock(&p->p_sigacts->ps_mtx);
334 while ((sig = cursig(td)) != 0) {
335 KASSERT(sig >= 0, ("sig %d", sig));
336 postsig(sig);
337 }
338 mtx_unlock(&p->p_sigacts->ps_mtx);
339 td->td_dbgflags &= old_boundary;
340 PROC_UNLOCK(p);
341 resched_sigs = true;
342 } else {
343 resched_sigs = false;
344 }
345
346 if ((flags & TDF_KQTICKLED) != 0)
347 kqueue_drain_schedtask();
348
349 /*
350 * Handle deferred update of the fast sigblock value, after
351 * the postsig() loop was performed.
352 */
353 sigfastblock_setpend(td, resched_sigs);
354
355 #ifdef KTRACE
356 KTRUSERRET(td);
357 #endif
358
359 /*
360 * We need to check to see if we have to exit or wait due to a
361 * single threading requirement or some other STOP condition.
362 */
363 if (flags & TDF_NEEDSUSPCHK) {
364 PROC_LOCK(p);
365 thread_suspend_check(0);
366 PROC_UNLOCK(p);
367 }
368
369 if (td->td_pflags & TDP_OLDMASK) {
370 td->td_pflags &= ~TDP_OLDMASK;
371 kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0);
372 }
373
374 #ifdef RACCT
375 if (__predict_false(racct_enable && p->p_throttled != 0))
376 racct_proc_throttled(p);
377 #endif
378
379 userret(td, framep);
380 }
381
382 const char *
syscallname(struct proc * p,u_int code)383 syscallname(struct proc *p, u_int code)
384 {
385 static const char unknown[] = "unknown";
386 struct sysentvec *sv;
387
388 sv = p->p_sysent;
389 if (sv->sv_syscallnames == NULL || code >= sv->sv_size)
390 return (unknown);
391 return (sv->sv_syscallnames[code]);
392 }
393