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 __FBSDID("$FreeBSD$"); 48 49 #include "opt_hwpmc_hooks.h" 50 #include "opt_ktrace.h" 51 #include "opt_sched.h" 52 53 #include <sys/param.h> 54 #include <sys/bus.h> 55 #include <sys/capsicum.h> 56 #include <sys/event.h> 57 #include <sys/kernel.h> 58 #include <sys/lock.h> 59 #include <sys/msan.h> 60 #include <sys/mutex.h> 61 #include <sys/pmckern.h> 62 #include <sys/proc.h> 63 #include <sys/ktr.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> 74 #ifdef KTRACE 75 #include <sys/uio.h> 76 #include <sys/ktrace.h> 77 #endif 78 #include <security/audit/audit.h> 79 80 #include <machine/cpu.h> 81 82 #ifdef VIMAGE 83 #include <net/vnet.h> 84 #endif 85 86 #ifdef HWPMC_HOOKS 87 #include <sys/pmckern.h> 88 #endif 89 90 #include <security/mac/mac_framework.h> 91 92 void (*softdep_ast_cleanup)(struct thread *); 93 94 /* 95 * Define the code needed before returning to user mode, for trap and 96 * syscall. 97 */ 98 void 99 userret(struct thread *td, struct trapframe *frame) 100 { 101 struct proc *p = td->td_proc; 102 103 CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid, 104 td->td_name); 105 KASSERT((p->p_flag & P_WEXIT) == 0, 106 ("Exiting process returns to usermode")); 107 #ifdef DIAGNOSTIC 108 /* 109 * Check that we called signotify() enough. For 110 * multi-threaded processes, where signal distribution might 111 * change due to other threads changing sigmask, the check is 112 * racy and cannot be performed reliably. 113 * If current process is vfork child, indicated by P_PPWAIT, then 114 * issignal() ignores stops, so we block the check to avoid 115 * classifying pending signals. 116 */ 117 if (p->p_numthreads == 1) { 118 PROC_LOCK(p); 119 thread_lock(td); 120 if ((p->p_flag & P_PPWAIT) == 0 && 121 (td->td_pflags & TDP_SIGFASTBLOCK) == 0) { 122 if (SIGPENDING(td) && (td->td_flags & 123 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) != 124 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) { 125 thread_unlock(td); 126 panic( 127 "failed to set signal flags for ast p %p td %p fl %x", 128 p, td, td->td_flags); 129 } 130 } 131 thread_unlock(td); 132 PROC_UNLOCK(p); 133 } 134 #endif 135 136 /* 137 * Charge system time if profiling. 138 */ 139 if (__predict_false(p->p_flag & P_PROFIL)) 140 addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio); 141 142 #ifdef HWPMC_HOOKS 143 if (PMC_THREAD_HAS_SAMPLES(td)) 144 PMC_CALL_HOOK(td, PMC_FN_THR_USERRET, NULL); 145 #endif 146 #ifdef TCPHPTS 147 /* 148 * @gallatin is adament that this needs to go here, I 149 * am not so sure. Running hpts is a lot like 150 * a lro_flush() that happens while a user process 151 * is running. But he may know best so I will go 152 * with his view of accounting. :-) 153 */ 154 tcp_run_hpts(); 155 #endif 156 /* 157 * Let the scheduler adjust our priority etc. 158 */ 159 sched_userret(td); 160 161 /* 162 * Check for misbehavior. 163 * 164 * In case there is a callchain tracing ongoing because of 165 * hwpmc(4), skip the scheduler pinning check. 166 * hwpmc(4) subsystem, infact, will collect callchain informations 167 * at ast() checkpoint, which is past userret(). 168 */ 169 WITNESS_WARN(WARN_PANIC, NULL, "userret: returning"); 170 KASSERT(td->td_critnest == 0, 171 ("userret: Returning in a critical section")); 172 KASSERT(td->td_locks == 0, 173 ("userret: Returning with %d locks held", td->td_locks)); 174 KASSERT(td->td_rw_rlocks == 0, 175 ("userret: Returning with %d rwlocks held in read mode", 176 td->td_rw_rlocks)); 177 KASSERT(td->td_sx_slocks == 0, 178 ("userret: Returning with %d sx locks held in shared mode", 179 td->td_sx_slocks)); 180 KASSERT(td->td_lk_slocks == 0, 181 ("userret: Returning with %d lockmanager locks held in shared mode", 182 td->td_lk_slocks)); 183 KASSERT((td->td_pflags & TDP_NOFAULTING) == 0, 184 ("userret: Returning with pagefaults disabled")); 185 if (__predict_false(!THREAD_CAN_SLEEP())) { 186 #ifdef EPOCH_TRACE 187 epoch_trace_list(curthread); 188 #endif 189 KASSERT(0, ("userret: Returning with sleep disabled")); 190 } 191 KASSERT(td->td_pinned == 0 || (td->td_pflags & TDP_CALLCHAIN) != 0, 192 ("userret: Returning with with pinned thread")); 193 KASSERT(td->td_vp_reserved == NULL, 194 ("userret: Returning with preallocated vnode")); 195 KASSERT((td->td_flags & (TDF_SBDRY | TDF_SEINTR | TDF_SERESTART)) == 0, 196 ("userret: Returning with stop signals deferred")); 197 KASSERT(td->td_vslock_sz == 0, 198 ("userret: Returning with vslock-wired space")); 199 #ifdef VIMAGE 200 /* Unfortunately td_vnet_lpush needs VNET_DEBUG. */ 201 VNET_ASSERT(curvnet == NULL, 202 ("%s: Returning on td %p (pid %d, %s) with vnet %p set in %s", 203 __func__, td, p->p_pid, td->td_name, curvnet, 204 (td->td_vnet_lpush != NULL) ? td->td_vnet_lpush : "N/A")); 205 #endif 206 } 207 208 /* 209 * Process an asynchronous software trap. 210 * This is relatively easy. 211 * This function will return with preemption disabled. 212 */ 213 void 214 ast(struct trapframe *framep) 215 { 216 struct thread *td; 217 struct proc *p; 218 int flags, sig; 219 bool resched_sigs; 220 221 kmsan_mark(framep, sizeof(*framep), KMSAN_STATE_INITED); 222 223 td = curthread; 224 p = td->td_proc; 225 226 CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid, 227 p->p_comm); 228 KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode")); 229 WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode"); 230 mtx_assert(&Giant, MA_NOTOWNED); 231 THREAD_LOCK_ASSERT(td, MA_NOTOWNED); 232 td->td_frame = framep; 233 td->td_pticks = 0; 234 235 /* 236 * This updates the td_flag's for the checks below in one 237 * "atomic" operation with turning off the astpending flag. 238 * If another AST is triggered while we are handling the 239 * AST's saved in flags, the astpending flag will be set and 240 * ast() will be called again. 241 */ 242 thread_lock(td); 243 flags = td->td_flags; 244 td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK | 245 TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND | 246 TDF_KQTICKLED); 247 thread_unlock(td); 248 VM_CNT_INC(v_trap); 249 250 if (td->td_cowgen != p->p_cowgen) 251 thread_cow_update(td); 252 if (td->td_pflags & TDP_OWEUPC && p->p_flag & P_PROFIL) { 253 addupc_task(td, td->td_profil_addr, td->td_profil_ticks); 254 td->td_profil_ticks = 0; 255 td->td_pflags &= ~TDP_OWEUPC; 256 } 257 #ifdef HWPMC_HOOKS 258 /* Handle Software PMC callchain capture. */ 259 if (PMC_IS_PENDING_CALLCHAIN(td)) 260 PMC_CALL_HOOK_UNLOCKED(td, PMC_FN_USER_CALLCHAIN_SOFT, (void *) framep); 261 #endif 262 if (flags & TDF_ALRMPEND) { 263 PROC_LOCK(p); 264 kern_psignal(p, SIGVTALRM); 265 PROC_UNLOCK(p); 266 } 267 if (flags & TDF_PROFPEND) { 268 PROC_LOCK(p); 269 kern_psignal(p, SIGPROF); 270 PROC_UNLOCK(p); 271 } 272 #ifdef MAC 273 if (flags & TDF_MACPEND) 274 mac_thread_userret(td); 275 #endif 276 if (flags & TDF_NEEDRESCHED) { 277 #ifdef KTRACE 278 if (KTRPOINT(td, KTR_CSW)) 279 ktrcsw(1, 1, __func__); 280 #endif 281 thread_lock(td); 282 sched_prio(td, td->td_user_pri); 283 mi_switch(SW_INVOL | SWT_NEEDRESCHED); 284 #ifdef KTRACE 285 if (KTRPOINT(td, KTR_CSW)) 286 ktrcsw(0, 1, __func__); 287 #endif 288 } 289 290 td_softdep_cleanup(td); 291 MPASS(td->td_su == NULL); 292 293 /* 294 * If this thread tickled GEOM, we need to wait for the giggling to 295 * stop before we return to userland 296 */ 297 if (__predict_false(td->td_pflags & TDP_GEOM)) 298 g_waitidle(); 299 300 #ifdef DIAGNOSTIC 301 if (p->p_numthreads == 1 && (flags & TDF_NEEDSIGCHK) == 0) { 302 PROC_LOCK(p); 303 thread_lock(td); 304 /* 305 * Note that TDF_NEEDSIGCHK should be re-read from 306 * td_flags, since signal might have been delivered 307 * after we cleared td_flags above. This is one of 308 * the reason for looping check for AST condition. 309 * See comment in userret() about P_PPWAIT. 310 */ 311 if ((p->p_flag & P_PPWAIT) == 0 && 312 (td->td_pflags & TDP_SIGFASTBLOCK) == 0) { 313 if (SIGPENDING(td) && (td->td_flags & 314 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) != 315 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) { 316 thread_unlock(td); /* fix dumps */ 317 panic( 318 "failed2 to set signal flags for ast p %p td %p fl %x %x", 319 p, td, flags, td->td_flags); 320 } 321 } 322 thread_unlock(td); 323 PROC_UNLOCK(p); 324 } 325 #endif 326 327 /* 328 * Check for signals. Unlocked reads of p_pendingcnt or 329 * p_siglist might cause process-directed signal to be handled 330 * later. 331 */ 332 if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 || 333 !SIGISEMPTY(p->p_siglist)) { 334 sigfastblock_fetch(td); 335 PROC_LOCK(p); 336 mtx_lock(&p->p_sigacts->ps_mtx); 337 while ((sig = cursig(td)) != 0) { 338 KASSERT(sig >= 0, ("sig %d", sig)); 339 postsig(sig); 340 } 341 mtx_unlock(&p->p_sigacts->ps_mtx); 342 PROC_UNLOCK(p); 343 resched_sigs = true; 344 } else { 345 resched_sigs = false; 346 } 347 348 if ((flags & TDF_KQTICKLED) != 0) 349 kqueue_drain_schedtask(); 350 351 /* 352 * Handle deferred update of the fast sigblock value, after 353 * the postsig() loop was performed. 354 */ 355 sigfastblock_setpend(td, resched_sigs); 356 357 #ifdef KTRACE 358 KTRUSERRET(td); 359 #endif 360 361 /* 362 * We need to check to see if we have to exit or wait due to a 363 * single threading requirement or some other STOP condition. 364 */ 365 if (flags & TDF_NEEDSUSPCHK) { 366 PROC_LOCK(p); 367 thread_suspend_check(0); 368 PROC_UNLOCK(p); 369 } 370 371 if (td->td_pflags & TDP_OLDMASK) { 372 td->td_pflags &= ~TDP_OLDMASK; 373 kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0); 374 } 375 376 #ifdef RACCT 377 if (__predict_false(racct_enable && p->p_throttled != 0)) 378 racct_proc_throttled(p); 379 #endif 380 381 userret(td, framep); 382 } 383 384 const char * 385 syscallname(struct proc *p, u_int code) 386 { 387 static const char unknown[] = "unknown"; 388 struct sysentvec *sv; 389 390 sv = p->p_sysent; 391 if (sv->sv_syscallnames == NULL || code >= sv->sv_size) 392 return (unknown); 393 return (sv->sv_syscallnames[code]); 394 } 395