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/kernel.h> 57 #include <sys/lock.h> 58 #include <sys/msan.h> 59 #include <sys/mutex.h> 60 #include <sys/pmckern.h> 61 #include <sys/proc.h> 62 #include <sys/ktr.h> 63 #include <sys/ptrace.h> 64 #include <sys/racct.h> 65 #include <sys/resourcevar.h> 66 #include <sys/sched.h> 67 #include <sys/signalvar.h> 68 #include <sys/syscall.h> 69 #include <sys/syscallsubr.h> 70 #include <sys/sysent.h> 71 #include <sys/systm.h> 72 #include <sys/vmmeter.h> 73 #ifdef KTRACE 74 #include <sys/uio.h> 75 #include <sys/ktrace.h> 76 #endif 77 #include <security/audit/audit.h> 78 79 #include <machine/cpu.h> 80 81 #ifdef VIMAGE 82 #include <net/vnet.h> 83 #endif 84 85 #ifdef HWPMC_HOOKS 86 #include <sys/pmckern.h> 87 #endif 88 89 #include <security/mac/mac_framework.h> 90 91 void (*softdep_ast_cleanup)(struct thread *); 92 93 /* 94 * Define the code needed before returning to user mode, for trap and 95 * syscall. 96 */ 97 void 98 userret(struct thread *td, struct trapframe *frame) 99 { 100 struct proc *p = td->td_proc; 101 102 CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid, 103 td->td_name); 104 KASSERT((p->p_flag & P_WEXIT) == 0, 105 ("Exiting process returns to usermode")); 106 #ifdef DIAGNOSTIC 107 /* 108 * Check that we called signotify() enough. For 109 * multi-threaded processes, where signal distribution might 110 * change due to other threads changing sigmask, the check is 111 * racy and cannot be performed reliably. 112 * If current process is vfork child, indicated by P_PPWAIT, then 113 * issignal() ignores stops, so we block the check to avoid 114 * classifying pending signals. 115 */ 116 if (p->p_numthreads == 1) { 117 PROC_LOCK(p); 118 thread_lock(td); 119 if ((p->p_flag & P_PPWAIT) == 0 && 120 (td->td_pflags & TDP_SIGFASTBLOCK) == 0) { 121 if (SIGPENDING(td) && (td->td_flags & 122 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) != 123 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) { 124 thread_unlock(td); 125 panic( 126 "failed to set signal flags for ast p %p td %p fl %x", 127 p, td, td->td_flags); 128 } 129 } 130 thread_unlock(td); 131 PROC_UNLOCK(p); 132 } 133 #endif 134 135 /* 136 * Charge system time if profiling. 137 */ 138 if (__predict_false(p->p_flag & P_PROFIL)) 139 addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio); 140 141 #ifdef HWPMC_HOOKS 142 if (PMC_THREAD_HAS_SAMPLES(td)) 143 PMC_CALL_HOOK(td, PMC_FN_THR_USERRET, NULL); 144 #endif 145 #ifdef TCPHPTS 146 /* 147 * @gallatin is adament that this needs to go here, I 148 * am not so sure. Running hpts is a lot like 149 * a lro_flush() that happens while a user process 150 * is running. But he may know best so I will go 151 * with his view of accounting. :-) 152 */ 153 tcp_run_hpts(); 154 #endif 155 /* 156 * Let the scheduler adjust our priority etc. 157 */ 158 sched_userret(td); 159 160 /* 161 * Check for misbehavior. 162 * 163 * In case there is a callchain tracing ongoing because of 164 * hwpmc(4), skip the scheduler pinning check. 165 * hwpmc(4) subsystem, infact, will collect callchain informations 166 * at ast() checkpoint, which is past userret(). 167 */ 168 WITNESS_WARN(WARN_PANIC, NULL, "userret: returning"); 169 KASSERT(td->td_critnest == 0, 170 ("userret: Returning in a critical section")); 171 KASSERT(td->td_locks == 0, 172 ("userret: Returning with %d locks held", td->td_locks)); 173 KASSERT(td->td_rw_rlocks == 0, 174 ("userret: Returning with %d rwlocks held in read mode", 175 td->td_rw_rlocks)); 176 KASSERT(td->td_sx_slocks == 0, 177 ("userret: Returning with %d sx locks held in shared mode", 178 td->td_sx_slocks)); 179 KASSERT(td->td_lk_slocks == 0, 180 ("userret: Returning with %d lockmanager locks held in shared mode", 181 td->td_lk_slocks)); 182 KASSERT((td->td_pflags & TDP_NOFAULTING) == 0, 183 ("userret: Returning with pagefaults disabled")); 184 if (__predict_false(!THREAD_CAN_SLEEP())) { 185 #ifdef EPOCH_TRACE 186 epoch_trace_list(curthread); 187 #endif 188 KASSERT(0, ("userret: Returning with sleep disabled")); 189 } 190 KASSERT(td->td_pinned == 0 || (td->td_pflags & TDP_CALLCHAIN) != 0, 191 ("userret: Returning with with pinned thread")); 192 KASSERT(td->td_vp_reserved == NULL, 193 ("userret: Returning with preallocated vnode")); 194 KASSERT((td->td_flags & (TDF_SBDRY | TDF_SEINTR | TDF_SERESTART)) == 0, 195 ("userret: Returning with stop signals deferred")); 196 KASSERT(td->td_vslock_sz == 0, 197 ("userret: Returning with vslock-wired space")); 198 #ifdef VIMAGE 199 /* Unfortunately td_vnet_lpush needs VNET_DEBUG. */ 200 VNET_ASSERT(curvnet == NULL, 201 ("%s: Returning on td %p (pid %d, %s) with vnet %p set in %s", 202 __func__, td, p->p_pid, td->td_name, curvnet, 203 (td->td_vnet_lpush != NULL) ? td->td_vnet_lpush : "N/A")); 204 #endif 205 } 206 207 /* 208 * Process an asynchronous software trap. 209 * This is relatively easy. 210 * This function will return with preemption disabled. 211 */ 212 void 213 ast(struct trapframe *framep) 214 { 215 struct thread *td; 216 struct proc *p; 217 int flags, sig; 218 bool resched_sigs; 219 220 kmsan_mark(framep, sizeof(*framep), KMSAN_STATE_INITED); 221 222 td = curthread; 223 p = td->td_proc; 224 225 CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid, 226 p->p_comm); 227 KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode")); 228 WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode"); 229 mtx_assert(&Giant, MA_NOTOWNED); 230 THREAD_LOCK_ASSERT(td, MA_NOTOWNED); 231 td->td_frame = framep; 232 td->td_pticks = 0; 233 234 /* 235 * This updates the td_flag's for the checks below in one 236 * "atomic" operation with turning off the astpending flag. 237 * If another AST is triggered while we are handling the 238 * AST's saved in flags, the astpending flag will be set and 239 * ast() will be called again. 240 */ 241 thread_lock(td); 242 flags = td->td_flags; 243 td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK | 244 TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND); 245 thread_unlock(td); 246 VM_CNT_INC(v_trap); 247 248 if (td->td_cowgen != p->p_cowgen) 249 thread_cow_update(td); 250 if (td->td_pflags & TDP_OWEUPC && p->p_flag & P_PROFIL) { 251 addupc_task(td, td->td_profil_addr, td->td_profil_ticks); 252 td->td_profil_ticks = 0; 253 td->td_pflags &= ~TDP_OWEUPC; 254 } 255 #ifdef HWPMC_HOOKS 256 /* Handle Software PMC callchain capture. */ 257 if (PMC_IS_PENDING_CALLCHAIN(td)) 258 PMC_CALL_HOOK_UNLOCKED(td, PMC_FN_USER_CALLCHAIN_SOFT, (void *) framep); 259 #endif 260 if (flags & TDF_ALRMPEND) { 261 PROC_LOCK(p); 262 kern_psignal(p, SIGVTALRM); 263 PROC_UNLOCK(p); 264 } 265 if (flags & TDF_PROFPEND) { 266 PROC_LOCK(p); 267 kern_psignal(p, SIGPROF); 268 PROC_UNLOCK(p); 269 } 270 #ifdef MAC 271 if (flags & TDF_MACPEND) 272 mac_thread_userret(td); 273 #endif 274 if (flags & TDF_NEEDRESCHED) { 275 #ifdef KTRACE 276 if (KTRPOINT(td, KTR_CSW)) 277 ktrcsw(1, 1, __func__); 278 #endif 279 thread_lock(td); 280 sched_prio(td, td->td_user_pri); 281 mi_switch(SW_INVOL | SWT_NEEDRESCHED); 282 #ifdef KTRACE 283 if (KTRPOINT(td, KTR_CSW)) 284 ktrcsw(0, 1, __func__); 285 #endif 286 } 287 288 td_softdep_cleanup(td); 289 MPASS(td->td_su == NULL); 290 291 /* 292 * If this thread tickled GEOM, we need to wait for the giggling to 293 * stop before we return to userland 294 */ 295 if (__predict_false(td->td_pflags & TDP_GEOM)) 296 g_waitidle(); 297 298 #ifdef DIAGNOSTIC 299 if (p->p_numthreads == 1 && (flags & TDF_NEEDSIGCHK) == 0) { 300 PROC_LOCK(p); 301 thread_lock(td); 302 /* 303 * Note that TDF_NEEDSIGCHK should be re-read from 304 * td_flags, since signal might have been delivered 305 * after we cleared td_flags above. This is one of 306 * the reason for looping check for AST condition. 307 * See comment in userret() about P_PPWAIT. 308 */ 309 if ((p->p_flag & P_PPWAIT) == 0 && 310 (td->td_pflags & TDP_SIGFASTBLOCK) == 0) { 311 if (SIGPENDING(td) && (td->td_flags & 312 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) != 313 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) { 314 thread_unlock(td); /* fix dumps */ 315 panic( 316 "failed2 to set signal flags for ast p %p td %p fl %x %x", 317 p, td, flags, td->td_flags); 318 } 319 } 320 thread_unlock(td); 321 PROC_UNLOCK(p); 322 } 323 #endif 324 325 /* 326 * Check for signals. Unlocked reads of p_pendingcnt or 327 * p_siglist might cause process-directed signal to be handled 328 * later. 329 */ 330 if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 || 331 !SIGISEMPTY(p->p_siglist)) { 332 sigfastblock_fetch(td); 333 PROC_LOCK(p); 334 mtx_lock(&p->p_sigacts->ps_mtx); 335 while ((sig = cursig(td)) != 0) { 336 KASSERT(sig >= 0, ("sig %d", sig)); 337 postsig(sig); 338 } 339 mtx_unlock(&p->p_sigacts->ps_mtx); 340 PROC_UNLOCK(p); 341 resched_sigs = true; 342 } else { 343 resched_sigs = false; 344 } 345 346 /* 347 * Handle deferred update of the fast sigblock value, after 348 * the postsig() loop was performed. 349 */ 350 sigfastblock_setpend(td, resched_sigs); 351 352 #ifdef KTRACE 353 KTRUSERRET(td); 354 #endif 355 356 /* 357 * We need to check to see if we have to exit or wait due to a 358 * single threading requirement or some other STOP condition. 359 */ 360 if (flags & TDF_NEEDSUSPCHK) { 361 PROC_LOCK(p); 362 thread_suspend_check(0); 363 PROC_UNLOCK(p); 364 } 365 366 if (td->td_pflags & TDP_OLDMASK) { 367 td->td_pflags &= ~TDP_OLDMASK; 368 kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0); 369 } 370 371 #ifdef RACCT 372 if (__predict_false(racct_enable && p->p_throttled != 0)) 373 racct_proc_throttled(p); 374 #endif 375 376 userret(td, framep); 377 } 378 379 const char * 380 syscallname(struct proc *p, u_int code) 381 { 382 static const char unknown[] = "unknown"; 383 struct sysentvec *sv; 384 385 sv = p->p_sysent; 386 if (sv->sv_syscallnames == NULL || code >= sv->sv_size) 387 return (unknown); 388 return (sv->sv_syscallnames[code]); 389 } 390