subr_syscall.c revision 208453
1118824Sharti/*- 2118824Sharti * Copyright (C) 1994, David Greenman 3118824Sharti * Copyright (c) 1990, 1993 4119075Sobrien * The Regents of the University of California. All rights reserved. 5118824Sharti * Copyright (c) 2007 The FreeBSD Foundation 6118824Sharti * 7119075Sobrien * This code is derived from software contributed to Berkeley by 8118824Sharti * the University of Utah, and William Jolitz. 9133565Sharti * 10133565Sharti * Portions of this software were developed by A. Joseph Koshy under 11133565Sharti * sponsorship from the FreeBSD Foundation and Google, Inc. 12133565Sharti * 13133565Sharti * Redistribution and use in source and binary forms, with or without 14133565Sharti * modification, are permitted provided that the following conditions 15133565Sharti * are met: 16118824Sharti * 1. Redistributions of source code must retain the above copyright 17132492Sharti * notice, this list of conditions and the following disclaimer. 18129215Scognet * 2. Redistributions in binary form must reproduce the above copyright 19133565Sharti * notice, this list of conditions and the following disclaimer in the 20133565Sharti * documentation and/or other materials provided with the distribution. 21133565Sharti * 3. All advertising materials mentioning features or use of this software 22133565Sharti * must display the following acknowledgement: 23133565Sharti * This product includes software developed by the University of 24133565Sharti * California, Berkeley and its contributors. 25133565Sharti * 4. Neither the name of the University nor the names of its contributors 26133565Sharti * may be used to endorse or promote products derived from this software 27133565Sharti * without specific prior written permission. 28133565Sharti * 29133565Sharti * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 30129215Scognet * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 31129215Scognet * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 32129215Scognet * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 33138809Sobrien * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 34129215Scognet * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 35118824Sharti * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 36133565Sharti * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 37119075Sobrien * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 38118824Sharti * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 39133565Sharti * SUCH DAMAGE. 40133565Sharti * 41133565Sharti * from: @(#)trap.c 7.4 (Berkeley) 5/13/91 42133565Sharti */ 43133565Sharti 44133565Sharti#include <sys/cdefs.h> 45133565Sharti__FBSDID("$FreeBSD: head/sys/kern/subr_trap.c 208453 2010-05-23 18:32:02Z kib $"); 46118824Sharti 47#include "opt_ktrace.h" 48#ifdef __i386__ 49#include "opt_npx.h" 50#endif 51#include "opt_sched.h" 52 53#include <sys/param.h> 54#include <sys/bus.h> 55#include <sys/kernel.h> 56#include <sys/lock.h> 57#include <sys/mutex.h> 58#include <sys/pmckern.h> 59#include <sys/proc.h> 60#include <sys/ktr.h> 61#include <sys/pioctl.h> 62#include <sys/ptrace.h> 63#include <sys/resourcevar.h> 64#include <sys/sched.h> 65#include <sys/signalvar.h> 66#include <sys/syscall.h> 67#include <sys/sysent.h> 68#include <sys/systm.h> 69#include <sys/vmmeter.h> 70#ifdef KTRACE 71#include <sys/uio.h> 72#include <sys/ktrace.h> 73#endif 74#include <security/audit/audit.h> 75 76#include <machine/cpu.h> 77#include <machine/pcb.h> 78 79#ifdef XEN 80#include <vm/vm.h> 81#include <vm/vm_param.h> 82#include <vm/pmap.h> 83#endif 84 85#include <security/mac/mac_framework.h> 86 87/* 88 * Define the code needed before returning to user mode, for trap and 89 * syscall. 90 */ 91void 92userret(struct thread *td, struct trapframe *frame) 93{ 94 struct proc *p = td->td_proc; 95 96 CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid, 97 td->td_name); 98#if 0 99#ifdef DIAGNOSTIC 100 /* Check that we called signotify() enough. */ 101 PROC_LOCK(p); 102 thread_lock(td); 103 if (SIGPENDING(td) && ((td->td_flags & TDF_NEEDSIGCHK) == 0 || 104 (td->td_flags & TDF_ASTPENDING) == 0)) 105 printf("failed to set signal flags properly for ast()\n"); 106 thread_unlock(td); 107 PROC_UNLOCK(p); 108#endif 109#endif 110#ifdef KTRACE 111 KTRUSERRET(td); 112#endif 113 /* 114 * If this thread tickled GEOM, we need to wait for the giggling to 115 * stop before we return to userland 116 */ 117 if (td->td_pflags & TDP_GEOM) 118 g_waitidle(); 119 120 /* 121 * Charge system time if profiling. 122 */ 123 if (p->p_flag & P_PROFIL) { 124 addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio); 125 } 126 /* 127 * Let the scheduler adjust our priority etc. 128 */ 129 sched_userret(td); 130 KASSERT(td->td_locks == 0, 131 ("userret: Returning with %d locks held.", td->td_locks)); 132#ifdef XEN 133 PT_UPDATES_FLUSH(); 134#endif 135} 136 137/* 138 * Process an asynchronous software trap. 139 * This is relatively easy. 140 * This function will return with preemption disabled. 141 */ 142void 143ast(struct trapframe *framep) 144{ 145 struct thread *td; 146 struct proc *p; 147 int flags; 148 int sig; 149#if defined(DEV_NPX) && !defined(SMP) 150 int ucode; 151 ksiginfo_t ksi; 152#endif 153 154 td = curthread; 155 p = td->td_proc; 156 157 CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid, 158 p->p_comm); 159 KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode")); 160 WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode"); 161 mtx_assert(&Giant, MA_NOTOWNED); 162 THREAD_LOCK_ASSERT(td, MA_NOTOWNED); 163 td->td_frame = framep; 164 td->td_pticks = 0; 165 166 /* 167 * This updates the td_flag's for the checks below in one 168 * "atomic" operation with turning off the astpending flag. 169 * If another AST is triggered while we are handling the 170 * AST's saved in flags, the astpending flag will be set and 171 * ast() will be called again. 172 */ 173 thread_lock(td); 174 flags = td->td_flags; 175 td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK | 176 TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND); 177 thread_unlock(td); 178 PCPU_INC(cnt.v_trap); 179 180 if (td->td_ucred != p->p_ucred) 181 cred_update_thread(td); 182 if (td->td_pflags & TDP_OWEUPC && p->p_flag & P_PROFIL) { 183 addupc_task(td, td->td_profil_addr, td->td_profil_ticks); 184 td->td_profil_ticks = 0; 185 td->td_pflags &= ~TDP_OWEUPC; 186 } 187 if (flags & TDF_ALRMPEND) { 188 PROC_LOCK(p); 189 psignal(p, SIGVTALRM); 190 PROC_UNLOCK(p); 191 } 192#if defined(DEV_NPX) && !defined(SMP) 193 if (PCPU_GET(curpcb)->pcb_flags & PCB_NPXTRAP) { 194 atomic_clear_int(&PCPU_GET(curpcb)->pcb_flags, 195 PCB_NPXTRAP); 196 ucode = npxtrap(); 197 if (ucode != -1) { 198 ksiginfo_init_trap(&ksi); 199 ksi.ksi_signo = SIGFPE; 200 ksi.ksi_code = ucode; 201 trapsignal(td, &ksi); 202 } 203 } 204#endif 205 if (flags & TDF_PROFPEND) { 206 PROC_LOCK(p); 207 psignal(p, SIGPROF); 208 PROC_UNLOCK(p); 209 } 210#ifdef MAC 211 if (flags & TDF_MACPEND) 212 mac_thread_userret(td); 213#endif 214 if (flags & TDF_NEEDRESCHED) { 215#ifdef KTRACE 216 if (KTRPOINT(td, KTR_CSW)) 217 ktrcsw(1, 1); 218#endif 219 thread_lock(td); 220 sched_prio(td, td->td_user_pri); 221 mi_switch(SW_INVOL | SWT_NEEDRESCHED, NULL); 222 thread_unlock(td); 223#ifdef KTRACE 224 if (KTRPOINT(td, KTR_CSW)) 225 ktrcsw(0, 1); 226#endif 227 } 228 229 /* 230 * Check for signals. Unlocked reads of p_pendingcnt or 231 * p_siglist might cause process-directed signal to be handled 232 * later. 233 */ 234 if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 || 235 !SIGISEMPTY(p->p_siglist)) { 236 PROC_LOCK(p); 237 mtx_lock(&p->p_sigacts->ps_mtx); 238 while ((sig = cursig(td, SIG_STOP_ALLOWED)) != 0) 239 postsig(sig); 240 mtx_unlock(&p->p_sigacts->ps_mtx); 241 PROC_UNLOCK(p); 242 } 243 /* 244 * We need to check to see if we have to exit or wait due to a 245 * single threading requirement or some other STOP condition. 246 */ 247 if (flags & TDF_NEEDSUSPCHK) { 248 PROC_LOCK(p); 249 thread_suspend_check(0); 250 PROC_UNLOCK(p); 251 } 252 253 if (td->td_pflags & TDP_OLDMASK) { 254 td->td_pflags &= ~TDP_OLDMASK; 255 kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0); 256 } 257 258 userret(td, framep); 259 mtx_assert(&Giant, MA_NOTOWNED); 260} 261 262#ifdef HAVE_SYSCALL_ARGS_DEF 263static const char *syscallname(struct proc *p, u_int code) __unused; 264static const 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