subr_syscall.c revision 219133
14Srgrimes/*- 21690Sdg * Copyright (C) 1994, David Greenman 31690Sdg * Copyright (c) 1990, 1993 41690Sdg * The Regents of the University of California. All rights reserved. 5174395Sjkoshy * Copyright (c) 2007 The FreeBSD Foundation 64Srgrimes * 74Srgrimes * This code is derived from software contributed to Berkeley by 84Srgrimes * the University of Utah, and William Jolitz. 94Srgrimes * 10174395Sjkoshy * Portions of this software were developed by A. Joseph Koshy under 11174395Sjkoshy * sponsorship from the FreeBSD Foundation and Google, Inc. 12174395Sjkoshy * 134Srgrimes * Redistribution and use in source and binary forms, with or without 144Srgrimes * modification, are permitted provided that the following conditions 154Srgrimes * are met: 164Srgrimes * 1. Redistributions of source code must retain the above copyright 174Srgrimes * notice, this list of conditions and the following disclaimer. 184Srgrimes * 2. Redistributions in binary form must reproduce the above copyright 194Srgrimes * notice, this list of conditions and the following disclaimer in the 204Srgrimes * documentation and/or other materials provided with the distribution. 214Srgrimes * 3. All advertising materials mentioning features or use of this software 224Srgrimes * must display the following acknowledgement: 234Srgrimes * This product includes software developed by the University of 244Srgrimes * California, Berkeley and its contributors. 254Srgrimes * 4. Neither the name of the University nor the names of its contributors 264Srgrimes * may be used to endorse or promote products derived from this software 274Srgrimes * without specific prior written permission. 284Srgrimes * 294Srgrimes * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 304Srgrimes * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 314Srgrimes * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 324Srgrimes * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 334Srgrimes * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 344Srgrimes * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 354Srgrimes * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 364Srgrimes * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 374Srgrimes * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 384Srgrimes * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 394Srgrimes * SUCH DAMAGE. 404Srgrimes * 41608Srgrimes * from: @(#)trap.c 7.4 (Berkeley) 5/13/91 424Srgrimes */ 434Srgrimes 44116182Sobrien#include <sys/cdefs.h> 45116182Sobrien__FBSDID("$FreeBSD: head/sys/kern/subr_trap.c 219133 2011-03-01 13:32:07Z rwatson $"); 46116182Sobrien 47219133Srwatson#include "opt_capabilities.h" 48118240Speter#include "opt_ktrace.h" 49209258Srpaulo#include "opt_kdtrace.h" 50170640Sjeff#include "opt_sched.h" 5113203Swollman 521549Srgrimes#include <sys/param.h> 5365557Sjasone#include <sys/bus.h> 54219133Srwatson#include <sys/capability.h> 551549Srgrimes#include <sys/kernel.h> 5678983Sjhb#include <sys/lock.h> 5767365Sjhb#include <sys/mutex.h> 58174395Sjkoshy#include <sys/pmckern.h> 5978983Sjhb#include <sys/proc.h> 6099072Sjulian#include <sys/ktr.h> 61208453Skib#include <sys/pioctl.h> 62208453Skib#include <sys/ptrace.h> 6331389Sbde#include <sys/resourcevar.h> 64104964Sjeff#include <sys/sched.h> 6531389Sbde#include <sys/signalvar.h> 66208453Skib#include <sys/syscall.h> 67209613Sjhb#include <sys/syscallsubr.h> 68208453Skib#include <sys/sysent.h> 6978983Sjhb#include <sys/systm.h> 7012662Sdg#include <sys/vmmeter.h> 71118240Speter#ifdef KTRACE 72118240Speter#include <sys/uio.h> 73118240Speter#include <sys/ktrace.h> 74118240Speter#endif 75208453Skib#include <security/audit/audit.h> 76118240Speter 771549Srgrimes#include <machine/cpu.h> 781549Srgrimes 79218688Sbz#ifdef VIMAGE 80218688Sbz#include <net/vnet.h> 81218688Sbz#endif 82218688Sbz 83184042Skmacy#ifdef XEN 84184042Skmacy#include <vm/vm.h> 85184042Skmacy#include <vm/vm_param.h> 86184042Skmacy#include <vm/pmap.h> 87184042Skmacy#endif 88184042Skmacy 89163606Srwatson#include <security/mac/mac_framework.h> 90163606Srwatson 9178983Sjhb/* 92167211Srwatson * Define the code needed before returning to user mode, for trap and 93167211Srwatson * syscall. 9478983Sjhb */ 9571527Sjhbvoid 96155455Sphkuserret(struct thread *td, struct trapframe *frame) 971690Sdg{ 9883366Sjulian struct proc *p = td->td_proc; 99757Sdg 10099072Sjulian CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid, 101173601Sjulian td->td_name); 102197963Skib#if 0 103126661Srwatson#ifdef DIAGNOSTIC 104110190Sjulian /* Check that we called signotify() enough. */ 10578636Sjhb PROC_LOCK(p); 106170307Sjeff thread_lock(td); 107112888Sjeff if (SIGPENDING(td) && ((td->td_flags & TDF_NEEDSIGCHK) == 0 || 108111032Sjulian (td->td_flags & TDF_ASTPENDING) == 0)) 109102266Srwatson printf("failed to set signal flags properly for ast()\n"); 110170307Sjeff thread_unlock(td); 11182585Sdillon PROC_UNLOCK(p); 11293793Sbde#endif 113197963Skib#endif 114152376Srwatson#ifdef KTRACE 115152376Srwatson KTRUSERRET(td); 116152376Srwatson#endif 11793793Sbde /* 118136837Sphk * If this thread tickled GEOM, we need to wait for the giggling to 119136837Sphk * stop before we return to userland 120136837Sphk */ 121136837Sphk if (td->td_pflags & TDP_GEOM) 122136837Sphk g_waitidle(); 123136837Sphk 124136837Sphk /* 125110190Sjulian * Charge system time if profiling. 126110190Sjulian */ 127213236Semaste if (p->p_flag & P_PROFIL) 128155455Sphk addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio); 129139324Sjeff /* 130139324Sjeff * Let the scheduler adjust our priority etc. 131139324Sjeff */ 132139324Sjeff sched_userret(td); 133144061Sjeff KASSERT(td->td_locks == 0, 134144061Sjeff ("userret: Returning with %d locks held.", td->td_locks)); 135218688Sbz#ifdef VIMAGE 136218688Sbz /* Unfortunately td_vnet_lpush needs VNET_DEBUG. */ 137218688Sbz VNET_ASSERT(curvnet == NULL, 138218688Sbz ("%s: Returning on td %p (pid %d, %s) with vnet %p set in %s", 139218688Sbz __func__, td, p->p_pid, td->td_name, curvnet, 140218688Sbz (td->td_vnet_lpush != NULL) ? td->td_vnet_lpush : "N/A")); 141218688Sbz#endif 142184042Skmacy#ifdef XEN 143184042Skmacy PT_UPDATES_FLUSH(); 144184042Skmacy#endif 1451690Sdg} 1461690Sdg 1474Srgrimes/* 14878983Sjhb * Process an asynchronous software trap. 14978983Sjhb * This is relatively easy. 15081493Sjhb * This function will return with preemption disabled. 1514Srgrimes */ 152798Swollmanvoid 15399072Sjulianast(struct trapframe *framep) 15465557Sjasone{ 155104297Sjhb struct thread *td; 156104297Sjhb struct proc *p; 15783366Sjulian int flags; 15893793Sbde int sig; 15965557Sjasone 160104297Sjhb td = curthread; 161104297Sjhb p = td->td_proc; 162104378Sjmallett 16399072Sjulian CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid, 16499072Sjulian p->p_comm); 16572911Sjhb KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode")); 166111883Sjhb WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode"); 16781493Sjhb mtx_assert(&Giant, MA_NOTOWNED); 168170307Sjeff THREAD_LOCK_ASSERT(td, MA_NOTOWNED); 16993390Sjake td->td_frame = framep; 170155455Sphk td->td_pticks = 0; 171104297Sjhb 17293390Sjake /* 173172207Sjeff * This updates the td_flag's for the checks below in one 17493390Sjake * "atomic" operation with turning off the astpending flag. 17593390Sjake * If another AST is triggered while we are handling the 176172207Sjeff * AST's saved in flags, the astpending flag will be set and 17793390Sjake * ast() will be called again. 17893390Sjake */ 179170307Sjeff thread_lock(td); 180170307Sjeff flags = td->td_flags; 181177471Sjeff td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK | 182177471Sjeff TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND); 183170307Sjeff thread_unlock(td); 184170292Sattilio PCPU_INC(cnt.v_trap); 185135573Sjhb 18693390Sjake if (td->td_ucred != p->p_ucred) 18793390Sjake cred_update_thread(td); 188132266Sjhb if (td->td_pflags & TDP_OWEUPC && p->p_flag & P_PROFIL) { 189132266Sjhb addupc_task(td, td->td_profil_addr, td->td_profil_ticks); 190132266Sjhb td->td_profil_ticks = 0; 191132266Sjhb td->td_pflags &= ~TDP_OWEUPC; 192131437Sjhb } 193172207Sjeff if (flags & TDF_ALRMPEND) { 19493390Sjake PROC_LOCK(p); 19593390Sjake psignal(p, SIGVTALRM); 19693390Sjake PROC_UNLOCK(p); 19793390Sjake } 198172207Sjeff if (flags & TDF_PROFPEND) { 19993390Sjake PROC_LOCK(p); 20093390Sjake psignal(p, SIGPROF); 20193390Sjake PROC_UNLOCK(p); 20293390Sjake } 203106655Srwatson#ifdef MAC 204172207Sjeff if (flags & TDF_MACPEND) 205106655Srwatson mac_thread_userret(td); 206106655Srwatson#endif 207111032Sjulian if (flags & TDF_NEEDRESCHED) { 208118240Speter#ifdef KTRACE 209118240Speter if (KTRPOINT(td, KTR_CSW)) 210119781Speter ktrcsw(1, 1); 211118240Speter#endif 212170307Sjeff thread_lock(td); 213163709Sjb sched_prio(td, td->td_user_pri); 214178272Sjeff mi_switch(SW_INVOL | SWT_NEEDRESCHED, NULL); 215170307Sjeff thread_unlock(td); 216118240Speter#ifdef KTRACE 217118240Speter if (KTRPOINT(td, KTR_CSW)) 218119781Speter ktrcsw(0, 1); 219118240Speter#endif 22093793Sbde } 221197963Skib 222197963Skib /* 223197963Skib * Check for signals. Unlocked reads of p_pendingcnt or 224197963Skib * p_siglist might cause process-directed signal to be handled 225197963Skib * later. 226197963Skib */ 227197963Skib if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 || 228197963Skib !SIGISEMPTY(p->p_siglist)) { 22993793Sbde PROC_LOCK(p); 230114983Sjhb mtx_lock(&p->p_sigacts->ps_mtx); 231195702Skib while ((sig = cursig(td, SIG_STOP_ALLOWED)) != 0) 23293793Sbde postsig(sig); 233114983Sjhb mtx_unlock(&p->p_sigacts->ps_mtx); 23493793Sbde PROC_UNLOCK(p); 23593793Sbde } 236177471Sjeff /* 237177471Sjeff * We need to check to see if we have to exit or wait due to a 238177471Sjeff * single threading requirement or some other STOP condition. 239177471Sjeff */ 240177471Sjeff if (flags & TDF_NEEDSUSPCHK) { 241177471Sjeff PROC_LOCK(p); 242177471Sjeff thread_suspend_check(0); 243177471Sjeff PROC_UNLOCK(p); 244177471Sjeff } 24565557Sjasone 246198508Skib if (td->td_pflags & TDP_OLDMASK) { 247198508Skib td->td_pflags &= ~TDP_OLDMASK; 248198508Skib kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0); 249198508Skib } 250198508Skib 251155455Sphk userret(td, framep); 25281493Sjhb mtx_assert(&Giant, MA_NOTOWNED); 25324691Speter} 254208453Skib 255208453Skib#ifdef HAVE_SYSCALL_ARGS_DEF 256208566Skibconst char * 257208453Skibsyscallname(struct proc *p, u_int code) 258208453Skib{ 259208453Skib static const char unknown[] = "unknown"; 260209697Skib struct sysentvec *sv; 261208453Skib 262209697Skib sv = p->p_sysent; 263209697Skib if (sv->sv_syscallnames == NULL || code >= sv->sv_size) 264208453Skib return (unknown); 265209697Skib return (sv->sv_syscallnames[code]); 266208453Skib} 267208453Skib 268208453Skibint 269208453Skibsyscallenter(struct thread *td, struct syscall_args *sa) 270208453Skib{ 271208453Skib struct proc *p; 272208453Skib int error, traced; 273208453Skib 274208453Skib PCPU_INC(cnt.v_syscall); 275208453Skib p = td->td_proc; 276208453Skib 277208453Skib td->td_pticks = 0; 278208453Skib if (td->td_ucred != p->p_ucred) 279208453Skib cred_update_thread(td); 280208453Skib if (p->p_flag & P_TRACED) { 281208453Skib traced = 1; 282208453Skib PROC_LOCK(p); 283208453Skib td->td_dbgflags &= ~TDB_USERWR; 284208453Skib td->td_dbgflags |= TDB_SCE; 285208453Skib PROC_UNLOCK(p); 286208453Skib } else 287208453Skib traced = 0; 288208453Skib error = (p->p_sysent->sv_fetch_syscall_args)(td, sa); 289208453Skib#ifdef KTRACE 290208453Skib if (KTRPOINT(td, KTR_SYSCALL)) 291208453Skib ktrsyscall(sa->code, sa->narg, sa->args); 292208453Skib#endif 293208453Skib 294208453Skib CTR6(KTR_SYSC, 295208453Skib"syscall: td=%p pid %d %s (%#lx, %#lx, %#lx)", 296208453Skib td, td->td_proc->p_pid, syscallname(p, sa->code), 297208453Skib sa->args[0], sa->args[1], sa->args[2]); 298208453Skib 299208453Skib if (error == 0) { 300208453Skib STOPEVENT(p, S_SCE, sa->narg); 301208453Skib PTRACESTOP_SC(p, td, S_PT_SCE); 302208453Skib if (td->td_dbgflags & TDB_USERWR) { 303208453Skib /* 304208453Skib * Reread syscall number and arguments if 305208453Skib * debugger modified registers or memory. 306208453Skib */ 307208453Skib error = (p->p_sysent->sv_fetch_syscall_args)(td, sa); 308208453Skib#ifdef KTRACE 309208453Skib if (KTRPOINT(td, KTR_SYSCALL)) 310208453Skib ktrsyscall(sa->code, sa->narg, sa->args); 311208453Skib#endif 312208453Skib if (error != 0) 313208453Skib goto retval; 314208453Skib } 315219133Srwatson 316219133Srwatson#ifdef CAPABILITIES 317219133Srwatson /* 318219133Srwatson * In capability mode, we only allow access to system calls 319219133Srwatson * flagged with SYF_CAPENABLED. 320219133Srwatson */ 321219133Srwatson if (IN_CAPABILITY_MODE(td) && 322219133Srwatson !(sa->callp->sy_flags & SYF_CAPENABLED)) { 323219133Srwatson error = ECAPMODE; 324219133Srwatson goto retval; 325219133Srwatson } 326219133Srwatson#endif 327219133Srwatson 328209579Skib error = syscall_thread_enter(td, sa->callp); 329209579Skib if (error != 0) 330209579Skib goto retval; 331208453Skib 332208453Skib#ifdef KDTRACE_HOOKS 333208453Skib /* 334208453Skib * If the systrace module has registered it's probe 335208453Skib * callback and if there is a probe active for the 336208453Skib * syscall 'entry', process the probe. 337208453Skib */ 338208453Skib if (systrace_probe_func != NULL && sa->callp->sy_entry != 0) 339208453Skib (*systrace_probe_func)(sa->callp->sy_entry, sa->code, 340211617Srpaulo sa->callp, sa->args, 0); 341208453Skib#endif 342208453Skib 343208453Skib AUDIT_SYSCALL_ENTER(sa->code, td); 344208453Skib error = (sa->callp->sy_call)(td, sa->args); 345208453Skib AUDIT_SYSCALL_EXIT(error, td); 346208453Skib 347208453Skib /* Save the latest error return value. */ 348208453Skib td->td_errno = error; 349208453Skib 350208453Skib#ifdef KDTRACE_HOOKS 351208453Skib /* 352208453Skib * If the systrace module has registered it's probe 353208453Skib * callback and if there is a probe active for the 354208453Skib * syscall 'return', process the probe. 355208453Skib */ 356208453Skib if (systrace_probe_func != NULL && sa->callp->sy_return != 0) 357208453Skib (*systrace_probe_func)(sa->callp->sy_return, sa->code, 358211617Srpaulo sa->callp, NULL, (error) ? -1 : td->td_retval[0]); 359208453Skib#endif 360209579Skib syscall_thread_exit(td, sa->callp); 361208453Skib CTR4(KTR_SYSC, "syscall: p=%p error=%d return %#lx %#lx", 362208453Skib p, error, td->td_retval[0], td->td_retval[1]); 363208453Skib } 364208453Skib retval: 365208453Skib if (traced) { 366208453Skib PROC_LOCK(p); 367208453Skib td->td_dbgflags &= ~TDB_SCE; 368208453Skib PROC_UNLOCK(p); 369208453Skib } 370208453Skib (p->p_sysent->sv_set_syscall_retval)(td, error); 371208453Skib return (error); 372208453Skib} 373208453Skib 374208453Skibvoid 375208453Skibsyscallret(struct thread *td, int error, struct syscall_args *sa __unused) 376208453Skib{ 377208453Skib struct proc *p; 378208453Skib int traced; 379208453Skib 380208453Skib p = td->td_proc; 381208453Skib 382208453Skib /* 383208453Skib * Check for misbehavior. 384208453Skib */ 385208453Skib WITNESS_WARN(WARN_PANIC, NULL, "System call %s returning", 386208453Skib syscallname(p, sa->code)); 387208453Skib KASSERT(td->td_critnest == 0, 388208453Skib ("System call %s returning in a critical section", 389208453Skib syscallname(p, sa->code))); 390208453Skib KASSERT(td->td_locks == 0, 391208453Skib ("System call %s returning with %d locks held", 392208453Skib syscallname(p, sa->code), td->td_locks)); 393208453Skib 394208453Skib /* 395208453Skib * Handle reschedule and other end-of-syscall issues 396208453Skib */ 397208453Skib userret(td, td->td_frame); 398208453Skib 399208453Skib CTR4(KTR_SYSC, "syscall %s exit thread %p pid %d proc %s", 400208453Skib syscallname(p, sa->code), td, td->td_proc->p_pid, td->td_name); 401208453Skib 402208453Skib#ifdef KTRACE 403208453Skib if (KTRPOINT(td, KTR_SYSRET)) 404208453Skib ktrsysret(sa->code, error, td->td_retval[0]); 405208453Skib#endif 406208453Skib 407208453Skib if (p->p_flag & P_TRACED) { 408208453Skib traced = 1; 409208453Skib PROC_LOCK(p); 410208453Skib td->td_dbgflags |= TDB_SCX; 411208453Skib PROC_UNLOCK(p); 412208453Skib } else 413208453Skib traced = 0; 414208453Skib /* 415208453Skib * This works because errno is findable through the 416208453Skib * register set. If we ever support an emulation where this 417208453Skib * is not the case, this code will need to be revisited. 418208453Skib */ 419208453Skib STOPEVENT(p, S_SCX, sa->code); 420208453Skib PTRACESTOP_SC(p, td, S_PT_SCX); 421217819Skib if (traced || (td->td_dbgflags & (TDB_EXEC | TDB_FORK)) != 0) { 422208453Skib PROC_LOCK(p); 423217819Skib td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC | TDB_FORK); 424208453Skib PROC_UNLOCK(p); 425208453Skib } 426208453Skib} 427208453Skib#endif /* HAVE_SYSCALL_ARGS_DEF */ 428