subr_syscall.c revision 209258
1208963Srdivacky/*- 2208963Srdivacky * Copyright (C) 1994, David Greenman 3208963Srdivacky * Copyright (c) 1990, 1993 4208963Srdivacky * The Regents of the University of California. All rights reserved. 5208963Srdivacky * Copyright (c) 2007 The FreeBSD Foundation 6208963Srdivacky * 7208963Srdivacky * This code is derived from software contributed to Berkeley by 8208963Srdivacky * the University of Utah, and William Jolitz. 9208963Srdivacky * 10208963Srdivacky * Portions of this software were developed by A. Joseph Koshy under 11208963Srdivacky * sponsorship from the FreeBSD Foundation and Google, Inc. 12208963Srdivacky * 13208963Srdivacky * Redistribution and use in source and binary forms, with or without 14208963Srdivacky * modification, are permitted provided that the following conditions 15208963Srdivacky * are met: 16208963Srdivacky * 1. Redistributions of source code must retain the above copyright 17208963Srdivacky * 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: head/sys/kern/subr_trap.c 209258 2010-06-17 17:34:45Z rpaulo $"); 46 47#include "opt_ktrace.h" 48#include "opt_kdtrace.h" 49#ifdef __i386__ 50#include "opt_npx.h" 51#endif 52#include "opt_sched.h" 53 54#include <sys/param.h> 55#include <sys/bus.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/resourcevar.h> 65#include <sys/sched.h> 66#include <sys/signalvar.h> 67#include <sys/syscall.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#include <machine/pcb.h> 79 80#ifdef XEN 81#include <vm/vm.h> 82#include <vm/vm_param.h> 83#include <vm/pmap.h> 84#endif 85 86#include <security/mac/mac_framework.h> 87 88/* 89 * Define the code needed before returning to user mode, for trap and 90 * syscall. 91 */ 92void 93userret(struct thread *td, struct trapframe *frame) 94{ 95 struct proc *p = td->td_proc; 96 97 CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid, 98 td->td_name); 99#if 0 100#ifdef DIAGNOSTIC 101 /* Check that we called signotify() enough. */ 102 PROC_LOCK(p); 103 thread_lock(td); 104 if (SIGPENDING(td) && ((td->td_flags & TDF_NEEDSIGCHK) == 0 || 105 (td->td_flags & TDF_ASTPENDING) == 0)) 106 printf("failed to set signal flags properly for ast()\n"); 107 thread_unlock(td); 108 PROC_UNLOCK(p); 109#endif 110#endif 111#ifdef KTRACE 112 KTRUSERRET(td); 113#endif 114 /* 115 * If this thread tickled GEOM, we need to wait for the giggling to 116 * stop before we return to userland 117 */ 118 if (td->td_pflags & TDP_GEOM) 119 g_waitidle(); 120 121 /* 122 * Charge system time if profiling. 123 */ 124 if (p->p_flag & P_PROFIL) { 125 addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio); 126 } 127 /* 128 * Let the scheduler adjust our priority etc. 129 */ 130 sched_userret(td); 131 KASSERT(td->td_locks == 0, 132 ("userret: Returning with %d locks held.", td->td_locks)); 133#ifdef XEN 134 PT_UPDATES_FLUSH(); 135#endif 136} 137 138/* 139 * Process an asynchronous software trap. 140 * This is relatively easy. 141 * This function will return with preemption disabled. 142 */ 143void 144ast(struct trapframe *framep) 145{ 146 struct thread *td; 147 struct proc *p; 148 int flags; 149 int sig; 150#if defined(DEV_NPX) && !defined(SMP) 151 int ucode; 152 ksiginfo_t ksi; 153#endif 154 155 td = curthread; 156 p = td->td_proc; 157 158 CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid, 159 p->p_comm); 160 KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode")); 161 WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode"); 162 mtx_assert(&Giant, MA_NOTOWNED); 163 THREAD_LOCK_ASSERT(td, MA_NOTOWNED); 164 td->td_frame = framep; 165 td->td_pticks = 0; 166 167 /* 168 * This updates the td_flag's for the checks below in one 169 * "atomic" operation with turning off the astpending flag. 170 * If another AST is triggered while we are handling the 171 * AST's saved in flags, the astpending flag will be set and 172 * ast() will be called again. 173 */ 174 thread_lock(td); 175 flags = td->td_flags; 176 td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK | 177 TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND); 178 thread_unlock(td); 179 PCPU_INC(cnt.v_trap); 180 181 if (td->td_ucred != p->p_ucred) 182 cred_update_thread(td); 183 if (td->td_pflags & TDP_OWEUPC && p->p_flag & P_PROFIL) { 184 addupc_task(td, td->td_profil_addr, td->td_profil_ticks); 185 td->td_profil_ticks = 0; 186 td->td_pflags &= ~TDP_OWEUPC; 187 } 188 if (flags & TDF_ALRMPEND) { 189 PROC_LOCK(p); 190 psignal(p, SIGVTALRM); 191 PROC_UNLOCK(p); 192 } 193#if defined(DEV_NPX) && !defined(SMP) 194 if (PCPU_GET(curpcb)->pcb_flags & PCB_NPXTRAP) { 195 atomic_clear_int(&PCPU_GET(curpcb)->pcb_flags, 196 PCB_NPXTRAP); 197 ucode = npxtrap(); 198 if (ucode != -1) { 199 ksiginfo_init_trap(&ksi); 200 ksi.ksi_signo = SIGFPE; 201 ksi.ksi_code = ucode; 202 trapsignal(td, &ksi); 203 } 204 } 205#endif 206 if (flags & TDF_PROFPEND) { 207 PROC_LOCK(p); 208 psignal(p, SIGPROF); 209 PROC_UNLOCK(p); 210 } 211#ifdef MAC 212 if (flags & TDF_MACPEND) 213 mac_thread_userret(td); 214#endif 215 if (flags & TDF_NEEDRESCHED) { 216#ifdef KTRACE 217 if (KTRPOINT(td, KTR_CSW)) 218 ktrcsw(1, 1); 219#endif 220 thread_lock(td); 221 sched_prio(td, td->td_user_pri); 222 mi_switch(SW_INVOL | SWT_NEEDRESCHED, NULL); 223 thread_unlock(td); 224#ifdef KTRACE 225 if (KTRPOINT(td, KTR_CSW)) 226 ktrcsw(0, 1); 227#endif 228 } 229 230 /* 231 * Check for signals. Unlocked reads of p_pendingcnt or 232 * p_siglist might cause process-directed signal to be handled 233 * later. 234 */ 235 if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 || 236 !SIGISEMPTY(p->p_siglist)) { 237 PROC_LOCK(p); 238 mtx_lock(&p->p_sigacts->ps_mtx); 239 while ((sig = cursig(td, SIG_STOP_ALLOWED)) != 0) 240 postsig(sig); 241 mtx_unlock(&p->p_sigacts->ps_mtx); 242 PROC_UNLOCK(p); 243 } 244 /* 245 * We need to check to see if we have to exit or wait due to a 246 * single threading requirement or some other STOP condition. 247 */ 248 if (flags & TDF_NEEDSUSPCHK) { 249 PROC_LOCK(p); 250 thread_suspend_check(0); 251 PROC_UNLOCK(p); 252 } 253 254 if (td->td_pflags & TDP_OLDMASK) { 255 td->td_pflags &= ~TDP_OLDMASK; 256 kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0); 257 } 258 259 userret(td, framep); 260 mtx_assert(&Giant, MA_NOTOWNED); 261} 262 263#ifdef HAVE_SYSCALL_ARGS_DEF 264const char * 265syscallname(struct proc *p, u_int code) 266{ 267 static const char unknown[] = "unknown"; 268 269 if (p->p_sysent->sv_syscallnames == NULL) 270 return (unknown); 271 return (p->p_sysent->sv_syscallnames[code]); 272} 273 274int 275syscallenter(struct thread *td, struct syscall_args *sa) 276{ 277 struct proc *p; 278 int error, traced; 279 280 PCPU_INC(cnt.v_syscall); 281 p = td->td_proc; 282 td->td_syscalls++; 283 284 td->td_pticks = 0; 285 if (td->td_ucred != p->p_ucred) 286 cred_update_thread(td); 287 if (p->p_flag & P_TRACED) { 288 traced = 1; 289 PROC_LOCK(p); 290 td->td_dbgflags &= ~TDB_USERWR; 291 td->td_dbgflags |= TDB_SCE; 292 PROC_UNLOCK(p); 293 } else 294 traced = 0; 295 error = (p->p_sysent->sv_fetch_syscall_args)(td, sa); 296#ifdef KTRACE 297 if (KTRPOINT(td, KTR_SYSCALL)) 298 ktrsyscall(sa->code, sa->narg, sa->args); 299#endif 300 301 CTR6(KTR_SYSC, 302"syscall: td=%p pid %d %s (%#lx, %#lx, %#lx)", 303 td, td->td_proc->p_pid, syscallname(p, sa->code), 304 sa->args[0], sa->args[1], sa->args[2]); 305 306 if (error == 0) { 307 STOPEVENT(p, S_SCE, sa->narg); 308 PTRACESTOP_SC(p, td, S_PT_SCE); 309 if (td->td_dbgflags & TDB_USERWR) { 310 /* 311 * Reread syscall number and arguments if 312 * debugger modified registers or memory. 313 */ 314 error = (p->p_sysent->sv_fetch_syscall_args)(td, sa); 315#ifdef KTRACE 316 if (KTRPOINT(td, KTR_SYSCALL)) 317 ktrsyscall(sa->code, sa->narg, sa->args); 318#endif 319 if (error != 0) 320 goto retval; 321 } 322 323#ifdef KDTRACE_HOOKS 324 /* 325 * If the systrace module has registered it's probe 326 * callback and if there is a probe active for the 327 * syscall 'entry', process the probe. 328 */ 329 if (systrace_probe_func != NULL && sa->callp->sy_entry != 0) 330 (*systrace_probe_func)(sa->callp->sy_entry, sa->code, 331 sa->callp, sa->args); 332#endif 333 334 AUDIT_SYSCALL_ENTER(sa->code, td); 335 error = (sa->callp->sy_call)(td, sa->args); 336 AUDIT_SYSCALL_EXIT(error, td); 337 338 /* Save the latest error return value. */ 339 td->td_errno = error; 340 341#ifdef KDTRACE_HOOKS 342 /* 343 * If the systrace module has registered it's probe 344 * callback and if there is a probe active for the 345 * syscall 'return', process the probe. 346 */ 347 if (systrace_probe_func != NULL && sa->callp->sy_return != 0) 348 (*systrace_probe_func)(sa->callp->sy_return, sa->code, 349 sa->callp, sa->args); 350#endif 351 CTR4(KTR_SYSC, "syscall: p=%p error=%d return %#lx %#lx", 352 p, error, td->td_retval[0], td->td_retval[1]); 353 } 354 retval: 355 if (traced) { 356 PROC_LOCK(p); 357 td->td_dbgflags &= ~TDB_SCE; 358 PROC_UNLOCK(p); 359 } 360 (p->p_sysent->sv_set_syscall_retval)(td, error); 361 return (error); 362} 363 364void 365syscallret(struct thread *td, int error, struct syscall_args *sa __unused) 366{ 367 struct proc *p; 368 int traced; 369 370 p = td->td_proc; 371 372 /* 373 * Check for misbehavior. 374 */ 375 WITNESS_WARN(WARN_PANIC, NULL, "System call %s returning", 376 syscallname(p, sa->code)); 377 KASSERT(td->td_critnest == 0, 378 ("System call %s returning in a critical section", 379 syscallname(p, sa->code))); 380 KASSERT(td->td_locks == 0, 381 ("System call %s returning with %d locks held", 382 syscallname(p, sa->code), td->td_locks)); 383 384 /* 385 * Handle reschedule and other end-of-syscall issues 386 */ 387 userret(td, td->td_frame); 388 389 CTR4(KTR_SYSC, "syscall %s exit thread %p pid %d proc %s", 390 syscallname(p, sa->code), td, td->td_proc->p_pid, td->td_name); 391 392#ifdef KTRACE 393 if (KTRPOINT(td, KTR_SYSRET)) 394 ktrsysret(sa->code, error, td->td_retval[0]); 395#endif 396 397 if (p->p_flag & P_TRACED) { 398 traced = 1; 399 PROC_LOCK(p); 400 td->td_dbgflags |= TDB_SCX; 401 PROC_UNLOCK(p); 402 } else 403 traced = 0; 404 /* 405 * This works because errno is findable through the 406 * register set. If we ever support an emulation where this 407 * is not the case, this code will need to be revisited. 408 */ 409 STOPEVENT(p, S_SCX, sa->code); 410 PTRACESTOP_SC(p, td, S_PT_SCX); 411 if (traced || (td->td_dbgflags & TDB_EXEC) != 0) { 412 PROC_LOCK(p); 413 td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC); 414 PROC_UNLOCK(p); 415 } 416} 417#endif /* HAVE_SYSCALL_ARGS_DEF */ 418