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/mutex.h> 59 #include <sys/pmckern.h> 60 #include <sys/proc.h> 61 #include <sys/ktr.h> 62 #include <sys/pioctl.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 KASSERT(!SIGPENDING(td) || (td->td_flags & 121 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) == 122 (TDF_NEEDSIGCHK | TDF_ASTPENDING), 123 ("failed to set signal flags for ast p %p " 124 "td %p fl %x", p, td, td->td_flags)); 125 } 126 thread_unlock(td); 127 PROC_UNLOCK(p); 128 } 129 #endif 130 #ifdef KTRACE 131 KTRUSERRET(td); 132 #endif 133 td_softdep_cleanup(td); 134 MPASS(td->td_su == NULL); 135 136 /* 137 * If this thread tickled GEOM, we need to wait for the giggling to 138 * stop before we return to userland 139 */ 140 if (td->td_pflags & TDP_GEOM) 141 g_waitidle(); 142 143 /* 144 * Charge system time if profiling. 145 */ 146 if (p->p_flag & P_PROFIL) 147 addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio); 148 149 #ifdef HWPMC_HOOKS 150 if (PMC_THREAD_HAS_SAMPLES(td)) 151 PMC_CALL_HOOK(td, PMC_FN_THR_USERRET, NULL); 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(1, ("userret: Returning with sleep disabled")); 187 } 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_su == NULL, 195 ("userret: Returning with SU cleanup request not handled")); 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 #ifdef RACCT 206 if (__predict_false(racct_enable && p->p_throttled != 0)) 207 racct_proc_throttled(p); 208 #endif 209 } 210 211 /* 212 * Process an asynchronous software trap. 213 * This is relatively easy. 214 * This function will return with preemption disabled. 215 */ 216 void 217 ast(struct trapframe *framep) 218 { 219 struct thread *td; 220 struct proc *p; 221 int flags; 222 int sig; 223 224 td = curthread; 225 p = td->td_proc; 226 227 CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid, 228 p->p_comm); 229 KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode")); 230 WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode"); 231 mtx_assert(&Giant, MA_NOTOWNED); 232 THREAD_LOCK_ASSERT(td, MA_NOTOWNED); 233 td->td_frame = framep; 234 td->td_pticks = 0; 235 236 /* 237 * This updates the td_flag's for the checks below in one 238 * "atomic" operation with turning off the astpending flag. 239 * If another AST is triggered while we are handling the 240 * AST's saved in flags, the astpending flag will be set and 241 * ast() will be called again. 242 */ 243 thread_lock(td); 244 flags = td->td_flags; 245 td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK | 246 TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND); 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 #ifdef DIAGNOSTIC 291 if (p->p_numthreads == 1 && (flags & TDF_NEEDSIGCHK) == 0) { 292 PROC_LOCK(p); 293 thread_lock(td); 294 /* 295 * Note that TDF_NEEDSIGCHK should be re-read from 296 * td_flags, since signal might have been delivered 297 * after we cleared td_flags above. This is one of 298 * the reason for looping check for AST condition. 299 * See comment in userret() about P_PPWAIT. 300 */ 301 if ((p->p_flag & P_PPWAIT) == 0) { 302 KASSERT(!SIGPENDING(td) || (td->td_flags & 303 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) == 304 (TDF_NEEDSIGCHK | TDF_ASTPENDING), 305 ("failed2 to set signal flags for ast p %p td %p " 306 "fl %x %x", p, td, flags, td->td_flags)); 307 } 308 thread_unlock(td); 309 PROC_UNLOCK(p); 310 } 311 #endif 312 313 /* 314 * Check for signals. Unlocked reads of p_pendingcnt or 315 * p_siglist might cause process-directed signal to be handled 316 * later. 317 */ 318 if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 || 319 !SIGISEMPTY(p->p_siglist)) { 320 PROC_LOCK(p); 321 mtx_lock(&p->p_sigacts->ps_mtx); 322 while ((sig = cursig(td)) != 0) { 323 KASSERT(sig >= 0, ("sig %d", sig)); 324 postsig(sig); 325 } 326 mtx_unlock(&p->p_sigacts->ps_mtx); 327 PROC_UNLOCK(p); 328 } 329 /* 330 * We need to check to see if we have to exit or wait due to a 331 * single threading requirement or some other STOP condition. 332 */ 333 if (flags & TDF_NEEDSUSPCHK) { 334 PROC_LOCK(p); 335 thread_suspend_check(0); 336 PROC_UNLOCK(p); 337 } 338 339 if (td->td_pflags & TDP_OLDMASK) { 340 td->td_pflags &= ~TDP_OLDMASK; 341 kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0); 342 } 343 344 userret(td, framep); 345 } 346 347 const char * 348 syscallname(struct proc *p, u_int code) 349 { 350 static const char unknown[] = "unknown"; 351 struct sysentvec *sv; 352 353 sv = p->p_sysent; 354 if (sv->sv_syscallnames == NULL || code >= sv->sv_size) 355 return (unknown); 356 return (sv->sv_syscallnames[code]); 357 } 358