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_epochnest == 0, 170 ("userret: Returning in an epoch 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_pflags & TDP_NOFAULTING) == 0, 180 ("userret: Returning with pagefaults disabled")); 181 KASSERT(td->td_no_sleeping == 0, 182 ("userret: Returning with sleep disabled")); 183 KASSERT(td->td_pinned == 0 || (td->td_pflags & TDP_CALLCHAIN) != 0, 184 ("userret: Returning with with pinned thread")); 185 KASSERT(td->td_vp_reserv == 0, 186 ("userret: Returning while holding vnode reservation")); 187 KASSERT((td->td_flags & (TDF_SBDRY | TDF_SEINTR | TDF_SERESTART)) == 0, 188 ("userret: Returning with stop signals deferred")); 189 KASSERT(td->td_su == NULL, 190 ("userret: Returning with SU cleanup request not handled")); 191 KASSERT(td->td_vslock_sz == 0, 192 ("userret: Returning with vslock-wired space")); 193 #ifdef VIMAGE 194 /* Unfortunately td_vnet_lpush needs VNET_DEBUG. */ 195 VNET_ASSERT(curvnet == NULL, 196 ("%s: Returning on td %p (pid %d, %s) with vnet %p set in %s", 197 __func__, td, p->p_pid, td->td_name, curvnet, 198 (td->td_vnet_lpush != NULL) ? td->td_vnet_lpush : "N/A")); 199 #endif 200 #ifdef RACCT 201 if (__predict_false(racct_enable && p->p_throttled != 0)) 202 racct_proc_throttled(p); 203 #endif 204 } 205 206 /* 207 * Process an asynchronous software trap. 208 * This is relatively easy. 209 * This function will return with preemption disabled. 210 */ 211 void 212 ast(struct trapframe *framep) 213 { 214 struct thread *td; 215 struct proc *p; 216 int flags; 217 int sig; 218 219 td = curthread; 220 p = td->td_proc; 221 222 CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid, 223 p->p_comm); 224 KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode")); 225 WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode"); 226 mtx_assert(&Giant, MA_NOTOWNED); 227 THREAD_LOCK_ASSERT(td, MA_NOTOWNED); 228 td->td_frame = framep; 229 td->td_pticks = 0; 230 231 /* 232 * This updates the td_flag's for the checks below in one 233 * "atomic" operation with turning off the astpending flag. 234 * If another AST is triggered while we are handling the 235 * AST's saved in flags, the astpending flag will be set and 236 * ast() will be called again. 237 */ 238 thread_lock(td); 239 flags = td->td_flags; 240 td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK | 241 TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND); 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, NULL); 279 thread_unlock(td); 280 #ifdef KTRACE 281 if (KTRPOINT(td, KTR_CSW)) 282 ktrcsw(0, 1, __func__); 283 #endif 284 } 285 286 #ifdef DIAGNOSTIC 287 if (p->p_numthreads == 1 && (flags & TDF_NEEDSIGCHK) == 0) { 288 PROC_LOCK(p); 289 thread_lock(td); 290 /* 291 * Note that TDF_NEEDSIGCHK should be re-read from 292 * td_flags, since signal might have been delivered 293 * after we cleared td_flags above. This is one of 294 * the reason for looping check for AST condition. 295 * See comment in userret() about P_PPWAIT. 296 */ 297 if ((p->p_flag & P_PPWAIT) == 0) { 298 KASSERT(!SIGPENDING(td) || (td->td_flags & 299 (TDF_NEEDSIGCHK | TDF_ASTPENDING)) == 300 (TDF_NEEDSIGCHK | TDF_ASTPENDING), 301 ("failed2 to set signal flags for ast p %p td %p " 302 "fl %x %x", p, td, flags, td->td_flags)); 303 } 304 thread_unlock(td); 305 PROC_UNLOCK(p); 306 } 307 #endif 308 309 /* 310 * Check for signals. Unlocked reads of p_pendingcnt or 311 * p_siglist might cause process-directed signal to be handled 312 * later. 313 */ 314 if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 || 315 !SIGISEMPTY(p->p_siglist)) { 316 PROC_LOCK(p); 317 mtx_lock(&p->p_sigacts->ps_mtx); 318 while ((sig = cursig(td)) != 0) { 319 KASSERT(sig >= 0, ("sig %d", sig)); 320 postsig(sig); 321 } 322 mtx_unlock(&p->p_sigacts->ps_mtx); 323 PROC_UNLOCK(p); 324 } 325 /* 326 * We need to check to see if we have to exit or wait due to a 327 * single threading requirement or some other STOP condition. 328 */ 329 if (flags & TDF_NEEDSUSPCHK) { 330 PROC_LOCK(p); 331 thread_suspend_check(0); 332 PROC_UNLOCK(p); 333 } 334 335 if (td->td_pflags & TDP_OLDMASK) { 336 td->td_pflags &= ~TDP_OLDMASK; 337 kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0); 338 } 339 340 userret(td, framep); 341 } 342 343 const char * 344 syscallname(struct proc *p, u_int code) 345 { 346 static const char unknown[] = "unknown"; 347 struct sysentvec *sv; 348 349 sv = p->p_sysent; 350 if (sv->sv_syscallnames == NULL || code >= sv->sv_size) 351 return (unknown); 352 return (sv->sv_syscallnames[code]); 353 } 354