1 /*- 2 * Copyright (C) 1994, David Greenman 3 * Copyright (c) 1990, 1993 4 * The Regents of the University of California. All rights reserved. 5 * Copyright (c) 2007 The FreeBSD Foundation 6 * 7 * This code is derived from software contributed to Berkeley by 8 * the University of Utah, and William Jolitz. 9 * 10 * Portions of this software were developed by A. Joseph Koshy under 11 * sponsorship from the FreeBSD Foundation and Google, Inc. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 3. All advertising materials mentioning features or use of this software 22 * must display the following acknowledgement: 23 * This product includes software developed by the University of 24 * California, Berkeley and its contributors. 25 * 4. Neither the name of the University nor the names of its contributors 26 * may be used to endorse or promote products derived from this software 27 * without specific prior written permission. 28 * 29 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 30 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 31 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 32 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 33 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 34 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 35 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 36 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 38 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 39 * SUCH DAMAGE. 40 * 41 * from: @(#)trap.c 7.4 (Berkeley) 5/13/91 42 */ 43 44 #include <sys/cdefs.h> 45 __FBSDID("$FreeBSD$"); 46 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/kernel.h> 55 #include <sys/lock.h> 56 #include <sys/mutex.h> 57 #include <sys/pmckern.h> 58 #include <sys/proc.h> 59 #include <sys/ktr.h> 60 #include <sys/pioctl.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 /* 90 * Define the code needed before returning to user mode, for trap and 91 * syscall. 92 */ 93 void 94 userret(struct thread *td, struct trapframe *frame) 95 { 96 struct proc *p = td->td_proc; 97 98 CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid, 99 td->td_name); 100 KASSERT((p->p_flag & P_WEXIT) == 0, 101 ("Exiting process returns to usermode")); 102 #if 0 103 #ifdef DIAGNOSTIC 104 /* Check that we called signotify() enough. */ 105 PROC_LOCK(p); 106 thread_lock(td); 107 if (SIGPENDING(td) && ((td->td_flags & TDF_NEEDSIGCHK) == 0 || 108 (td->td_flags & TDF_ASTPENDING) == 0)) 109 printf("failed to set signal flags properly for ast()\n"); 110 thread_unlock(td); 111 PROC_UNLOCK(p); 112 #endif 113 #endif 114 #ifdef KTRACE 115 KTRUSERRET(td); 116 #endif 117 /* 118 * If this thread tickled GEOM, we need to wait for the giggling to 119 * stop before we return to userland 120 */ 121 if (td->td_pflags & TDP_GEOM) 122 g_waitidle(); 123 124 /* 125 * Charge system time if profiling. 126 */ 127 if (p->p_flag & P_PROFIL) 128 addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio); 129 /* 130 * Let the scheduler adjust our priority etc. 131 */ 132 sched_userret(td); 133 134 /* 135 * Check for misbehavior. 136 * 137 * In case there is a callchain tracing ongoing because of 138 * hwpmc(4), skip the scheduler pinning check. 139 * hwpmc(4) subsystem, infact, will collect callchain informations 140 * at ast() checkpoint, which is past userret(). 141 */ 142 WITNESS_WARN(WARN_PANIC, NULL, "userret: returning"); 143 KASSERT(td->td_critnest == 0, 144 ("userret: Returning in a critical section")); 145 KASSERT(td->td_locks == 0, 146 ("userret: Returning with %d locks held", td->td_locks)); 147 KASSERT(td->td_rw_rlocks == 0, 148 ("userret: Returning with %d rwlocks held in read mode", 149 td->td_rw_rlocks)); 150 KASSERT((td->td_pflags & TDP_NOFAULTING) == 0, 151 ("userret: Returning with pagefaults disabled")); 152 KASSERT(td->td_no_sleeping == 0, 153 ("userret: Returning with sleep disabled")); 154 KASSERT(td->td_pinned == 0 || (td->td_pflags & TDP_CALLCHAIN) != 0, 155 ("userret: Returning with with pinned thread")); 156 KASSERT(td->td_vp_reserv == 0, 157 ("userret: Returning while holding vnode reservation")); 158 KASSERT((td->td_flags & TDF_SBDRY) == 0, 159 ("userret: Returning with stop signals deferred")); 160 #ifdef VIMAGE 161 /* Unfortunately td_vnet_lpush needs VNET_DEBUG. */ 162 VNET_ASSERT(curvnet == NULL, 163 ("%s: Returning on td %p (pid %d, %s) with vnet %p set in %s", 164 __func__, td, p->p_pid, td->td_name, curvnet, 165 (td->td_vnet_lpush != NULL) ? td->td_vnet_lpush : "N/A")); 166 #endif 167 #ifdef RACCT 168 if (racct_enable) { 169 PROC_LOCK(p); 170 while (p->p_throttled == 1) 171 msleep(p->p_racct, &p->p_mtx, 0, "racct", 0); 172 PROC_UNLOCK(p); 173 } 174 #endif 175 } 176 177 /* 178 * Process an asynchronous software trap. 179 * This is relatively easy. 180 * This function will return with preemption disabled. 181 */ 182 void 183 ast(struct trapframe *framep) 184 { 185 struct thread *td; 186 struct proc *p; 187 int flags; 188 int sig; 189 190 td = curthread; 191 p = td->td_proc; 192 193 CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid, 194 p->p_comm); 195 KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode")); 196 WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode"); 197 mtx_assert(&Giant, MA_NOTOWNED); 198 THREAD_LOCK_ASSERT(td, MA_NOTOWNED); 199 td->td_frame = framep; 200 td->td_pticks = 0; 201 202 /* 203 * This updates the td_flag's for the checks below in one 204 * "atomic" operation with turning off the astpending flag. 205 * If another AST is triggered while we are handling the 206 * AST's saved in flags, the astpending flag will be set and 207 * ast() will be called again. 208 */ 209 thread_lock(td); 210 flags = td->td_flags; 211 td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK | 212 TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND); 213 thread_unlock(td); 214 PCPU_INC(cnt.v_trap); 215 216 if (td->td_ucred != p->p_ucred) 217 cred_update_thread(td); 218 if (td->td_pflags & TDP_OWEUPC && p->p_flag & P_PROFIL) { 219 addupc_task(td, td->td_profil_addr, td->td_profil_ticks); 220 td->td_profil_ticks = 0; 221 td->td_pflags &= ~TDP_OWEUPC; 222 } 223 #ifdef HWPMC_HOOKS 224 /* Handle Software PMC callchain capture. */ 225 if (PMC_IS_PENDING_CALLCHAIN(td)) 226 PMC_CALL_HOOK_UNLOCKED(td, PMC_FN_USER_CALLCHAIN_SOFT, (void *) framep); 227 #endif 228 if (flags & TDF_ALRMPEND) { 229 PROC_LOCK(p); 230 kern_psignal(p, SIGVTALRM); 231 PROC_UNLOCK(p); 232 } 233 if (flags & TDF_PROFPEND) { 234 PROC_LOCK(p); 235 kern_psignal(p, SIGPROF); 236 PROC_UNLOCK(p); 237 } 238 #ifdef MAC 239 if (flags & TDF_MACPEND) 240 mac_thread_userret(td); 241 #endif 242 if (flags & TDF_NEEDRESCHED) { 243 #ifdef KTRACE 244 if (KTRPOINT(td, KTR_CSW)) 245 ktrcsw(1, 1, __func__); 246 #endif 247 thread_lock(td); 248 sched_prio(td, td->td_user_pri); 249 mi_switch(SW_INVOL | SWT_NEEDRESCHED, NULL); 250 thread_unlock(td); 251 #ifdef KTRACE 252 if (KTRPOINT(td, KTR_CSW)) 253 ktrcsw(0, 1, __func__); 254 #endif 255 } 256 257 /* 258 * Check for signals. Unlocked reads of p_pendingcnt or 259 * p_siglist might cause process-directed signal to be handled 260 * later. 261 */ 262 if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 || 263 !SIGISEMPTY(p->p_siglist)) { 264 PROC_LOCK(p); 265 mtx_lock(&p->p_sigacts->ps_mtx); 266 while ((sig = cursig(td)) != 0) 267 postsig(sig); 268 mtx_unlock(&p->p_sigacts->ps_mtx); 269 PROC_UNLOCK(p); 270 } 271 /* 272 * We need to check to see if we have to exit or wait due to a 273 * single threading requirement or some other STOP condition. 274 */ 275 if (flags & TDF_NEEDSUSPCHK) { 276 PROC_LOCK(p); 277 thread_suspend_check(0); 278 PROC_UNLOCK(p); 279 } 280 281 if (td->td_pflags & TDP_OLDMASK) { 282 td->td_pflags &= ~TDP_OLDMASK; 283 kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0); 284 } 285 286 userret(td, framep); 287 } 288 289 const char * 290 syscallname(struct proc *p, u_int code) 291 { 292 static const char unknown[] = "unknown"; 293 struct sysentvec *sv; 294 295 sv = p->p_sysent; 296 if (sv->sv_syscallnames == NULL || code >= sv->sv_size) 297 return (unknown); 298 return (sv->sv_syscallnames[code]); 299 } 300