subr_syscall.c revision 111032
1228753Smm/*- 2228753Smm * Copyright (C) 1994, David Greenman 3228753Smm * Copyright (c) 1990, 1993 4228753Smm * The Regents of the University of California. All rights reserved. 5228753Smm * 6228753Smm * This code is derived from software contributed to Berkeley by 7228753Smm * the University of Utah, and William Jolitz. 8228753Smm * 9228753Smm * Redistribution and use in source and binary forms, with or without 10228753Smm * modification, are permitted provided that the following conditions 11228753Smm * are met: 12228753Smm * 1. Redistributions of source code must retain the above copyright 13228753Smm * notice, this list of conditions and the following disclaimer. 14228753Smm * 2. Redistributions in binary form must reproduce the above copyright 15228753Smm * notice, this list of conditions and the following disclaimer in the 16228753Smm * documentation and/or other materials provided with the distribution. 17228753Smm * 3. All advertising materials mentioning features or use of this software 18228753Smm * must display the following acknowledgement: 19228753Smm * This product includes software developed by the University of 20228753Smm * California, Berkeley and its contributors. 21228753Smm * 4. Neither the name of the University nor the names of its contributors 22228753Smm * may be used to endorse or promote products derived from this software 23228753Smm * without specific prior written permission. 24228753Smm * 25228753Smm * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 26228753Smm * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27228753Smm * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28231200Smm * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 29228753Smm * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 30228753Smm * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 31228753Smm * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 32228753Smm * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 33228753Smm * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 34228753Smm * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 35228753Smm * SUCH DAMAGE. 36228753Smm * 37228753Smm * from: @(#)trap.c 7.4 (Berkeley) 5/13/91 38228753Smm * $FreeBSD: head/sys/kern/subr_trap.c 111032 2003-02-17 09:55:10Z julian $ 39228753Smm */ 40228753Smm 41228753Smm#include "opt_mac.h" 42228753Smm#ifdef __i386__ 43228753Smm#include "opt_npx.h" 44228753Smm#endif 45228753Smm 46228753Smm#include <sys/param.h> 47228753Smm#include <sys/bus.h> 48228753Smm#include <sys/kernel.h> 49228753Smm#include <sys/lock.h> 50228753Smm#include <sys/mac.h> 51228753Smm#include <sys/mutex.h> 52228753Smm#include <sys/proc.h> 53228753Smm#include <sys/kse.h> 54228753Smm#include <sys/ktr.h> 55228753Smm#include <sys/resourcevar.h> 56228753Smm#include <sys/sched.h> 57228753Smm#include <sys/signalvar.h> 58228753Smm#include <sys/systm.h> 59228753Smm#include <sys/vmmeter.h> 60228753Smm#include <machine/cpu.h> 61228753Smm#include <machine/pcb.h> 62228753Smm 63228753Smm/* 64228753Smm * Define the code needed before returning to user mode, for 65228753Smm * trap and syscall. 66228753Smm * 67228753Smm * MPSAFE 68228753Smm */ 69228753Smmvoid 70228753Smmuserret(td, frame, oticks) 71228753Smm struct thread *td; 72228753Smm struct trapframe *frame; 73228753Smm u_int oticks; 74228753Smm{ 75231200Smm struct proc *p = td->td_proc; 76231200Smm 77228753Smm CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid, 78231200Smm p->p_comm); 79228753Smm#ifdef INVARIANTS 80231200Smm /* Check that we called signotify() enough. */ 81231200Smm mtx_lock(&Giant); 82231200Smm PROC_LOCK(p); 83231200Smm mtx_lock_spin(&sched_lock); 84231200Smm if (SIGPENDING(p) && ((p->p_sflag & PS_NEEDSIGCHK) == 0 || 85231200Smm (td->td_flags & TDF_ASTPENDING) == 0)) 86231200Smm printf("failed to set signal flags properly for ast()\n"); 87228753Smm mtx_unlock_spin(&sched_lock); 88231200Smm PROC_UNLOCK(p); 89228753Smm mtx_unlock(&Giant); 90231200Smm#endif 91231200Smm 92231200Smm /* 93231200Smm * Let the scheduler adjust our priority etc. 94228753Smm */ 95228753Smm sched_userret(td); 96228753Smm 97248616Smm /* 98228753Smm * We need to check to see if we have to exit or wait due to a 99228753Smm * single threading requirement or some other STOP condition. 100228753Smm * Don't bother doing all the work if the stop bits are not set 101228753Smm * at this time.. If we miss it, we miss it.. no big deal. 102228753Smm */ 103228753Smm if (P_SHOULDSTOP(p)) { 104228753Smm PROC_LOCK(p); 105228753Smm thread_suspend_check(0); /* Can suspend or kill */ 106248616Smm PROC_UNLOCK(p); 107228753Smm } 108228753Smm 109228753Smm /* 110228753Smm * Do special thread processing, e.g. upcall tweaking and such. 111228753Smm */ 112228753Smm if (p->p_flag & P_KSES) { 113228753Smm thread_userret(td, frame); 114228753Smm } 115228753Smm 116228753Smm /* 117228753Smm * Charge system time if profiling. 118228753Smm * 119228753Smm * XXX should move PS_PROFIL to a place that can obviously be 120228753Smm * accessed safely without sched_lock. 121228753Smm */ 122228753Smm if (p->p_sflag & PS_PROFIL) { 123228753Smm quad_t ticks; 124228753Smm 125228753Smm mtx_lock_spin(&sched_lock); 126228753Smm ticks = td->td_sticks - oticks; 127228753Smm mtx_unlock_spin(&sched_lock); 128228753Smm addupc_task(td, TRAPF_PC(frame), (u_int)ticks * psratio); 129228753Smm } 130228753Smm} 131228753Smm 132228753Smm/* 133228753Smm * Process an asynchronous software trap. 134228753Smm * This is relatively easy. 135228753Smm * This function will return with preemption disabled. 136228753Smm */ 137228753Smmvoid 138228753Smmast(struct trapframe *framep) 139228753Smm{ 140231200Smm struct thread *td; 141228753Smm struct proc *p; 142228753Smm struct kse *ke; 143228753Smm struct ksegrp *kg; 144228753Smm struct rlimit *rlim; 145228753Smm u_int prticks, sticks; 146228753Smm int sflag; 147228753Smm int flags; 148228753Smm int sig; 149228753Smm#if defined(DEV_NPX) && !defined(SMP) 150228753Smm int ucode; 151228753Smm#endif 152228753Smm 153228753Smm td = curthread; 154228753Smm p = td->td_proc; 155228753Smm kg = td->td_ksegrp; 156228753Smm 157228753Smm CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid, 158228753Smm p->p_comm); 159228753Smm KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode")); 160228753Smm#ifdef WITNESS 161228753Smm if (witness_list(td)) 162228753Smm panic("Returning to user mode with mutex(s) held"); 163228753Smm#endif 164228753Smm mtx_assert(&Giant, MA_NOTOWNED); 165228753Smm mtx_assert(&sched_lock, MA_NOTOWNED); 166228753Smm td->td_frame = framep; 167228753Smm 168228753Smm /* 169228753Smm * This updates the p_sflag's for the checks below in one 170228753Smm * "atomic" operation with turning off the astpending flag. 171228753Smm * If another AST is triggered while we are handling the 172228753Smm * AST's saved in sflag, the astpending flag will be set and 173228753Smm * ast() will be called again. 174248616Smm */ 175228753Smm mtx_lock_spin(&sched_lock); 176228753Smm ke = td->td_kse; 177228753Smm sticks = td->td_sticks; 178248616Smm flags = td->td_flags; 179228753Smm sflag = p->p_sflag; 180228753Smm p->p_sflag &= ~(PS_ALRMPEND | PS_NEEDSIGCHK | PS_PROFPEND | PS_XCPU); 181228753Smm#ifdef MAC 182228753Smm p->p_sflag &= ~PS_MACPEND; 183228753Smm#endif 184228753Smm td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDRESCHED | TDF_OWEUPC); 185228753Smm cnt.v_soft++; 186228753Smm prticks = 0; 187228753Smm if (flags & TDF_OWEUPC && sflag & PS_PROFIL) { 188228753Smm prticks = p->p_stats->p_prof.pr_ticks; 189228753Smm p->p_stats->p_prof.pr_ticks = 0; 190228753Smm } 191228753Smm mtx_unlock_spin(&sched_lock); 192228753Smm /* 193228753Smm * XXXKSE While the fact that we owe a user profiling 194228753Smm * tick is stored per KSE in this code, the statistics 195228753Smm * themselves are still stored per process. 196248616Smm * This should probably change, by which I mean that 197228753Smm * possibly the location of both might change. 198228753Smm */ 199228753Smm 200228753Smm if (td->td_ucred != p->p_ucred) 201231200Smm cred_update_thread(td); 202228753Smm if (flags & TDF_OWEUPC && sflag & PS_PROFIL) 203228753Smm addupc_task(td, p->p_stats->p_prof.pr_addr, prticks); 204228753Smm if (sflag & PS_ALRMPEND) { 205228753Smm PROC_LOCK(p); 206228753Smm psignal(p, SIGVTALRM); 207228753Smm PROC_UNLOCK(p); 208228753Smm } 209228753Smm#if defined(DEV_NPX) && !defined(SMP) 210228753Smm if (PCPU_GET(curpcb)->pcb_flags & PCB_NPXTRAP) { 211228753Smm atomic_clear_int(&PCPU_GET(curpcb)->pcb_flags, 212228753Smm PCB_NPXTRAP); 213228753Smm ucode = npxtrap(); 214228753Smm if (ucode != -1) { 215228753Smm trapsignal(p, SIGFPE, ucode); 216228753Smm } 217228753Smm } 218228753Smm#endif 219228753Smm if (sflag & PS_PROFPEND) { 220228753Smm PROC_LOCK(p); 221228753Smm psignal(p, SIGPROF); 222228753Smm PROC_UNLOCK(p); 223228753Smm } 224228753Smm if (sflag & PS_XCPU) { 225228753Smm PROC_LOCK(p); 226228753Smm rlim = &p->p_rlimit[RLIMIT_CPU]; 227228753Smm if (p->p_runtime.sec >= rlim->rlim_max) 228228753Smm killproc(p, "exceeded maximum CPU limit"); 229228753Smm else { 230228753Smm psignal(p, SIGXCPU); 231228753Smm mtx_lock_spin(&sched_lock); 232228753Smm if (p->p_cpulimit < rlim->rlim_max) 233228753Smm p->p_cpulimit += 5; 234228753Smm mtx_unlock_spin(&sched_lock); 235228753Smm } 236228753Smm PROC_UNLOCK(p); 237228753Smm } 238228753Smm#ifdef MAC 239228753Smm if (sflag & PS_MACPEND) 240228753Smm mac_thread_userret(td); 241228753Smm#endif 242228753Smm if (flags & TDF_NEEDRESCHED) { 243228753Smm mtx_lock_spin(&sched_lock); 244228753Smm sched_prio(td, kg->kg_user_pri); 245228753Smm p->p_stats->p_ru.ru_nivcsw++; 246228753Smm mi_switch(); 247228753Smm mtx_unlock_spin(&sched_lock); 248228753Smm } 249228753Smm if (sflag & PS_NEEDSIGCHK) { 250228753Smm PROC_LOCK(p); 251228753Smm while ((sig = cursig(td)) != 0) 252228753Smm postsig(sig); 253228753Smm PROC_UNLOCK(p); 254228753Smm } 255228753Smm 256228753Smm userret(td, framep, sticks); 257228753Smm#ifdef DIAGNOSTIC 258228753Smm cred_free_thread(td); 259228753Smm#endif 260228753Smm mtx_assert(&Giant, MA_NOTOWNED); 261228753Smm} 262228753Smm