lwproc.c revision 1.7
1/* $NetBSD: lwproc.c,v 1.7 2011/01/02 12:52:25 pooka Exp $ */ 2 3/* 4 * Copyright (c) 2010, 2011 Antti Kantee. All Rights Reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS 16 * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 17 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 18 * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 21 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 25 * SUCH DAMAGE. 26 */ 27 28#include <sys/cdefs.h> 29__KERNEL_RCSID(0, "$NetBSD: lwproc.c,v 1.7 2011/01/02 12:52:25 pooka Exp $"); 30 31#include <sys/param.h> 32#include <sys/atomic.h> 33#include <sys/filedesc.h> 34#include <sys/kauth.h> 35#include <sys/kmem.h> 36#include <sys/lwp.h> 37#include <sys/pool.h> 38#include <sys/proc.h> 39#include <sys/queue.h> 40#include <sys/resourcevar.h> 41#include <sys/uidinfo.h> 42 43#include <rump/rumpuser.h> 44 45#include "rump_private.h" 46 47static void 48lwproc_proc_free(struct proc *p) 49{ 50 kauth_cred_t cred; 51 52 mutex_enter(proc_lock); 53 54 KASSERT(p->p_nlwps == 0); 55 KASSERT(LIST_EMPTY(&p->p_lwps)); 56 KASSERT(p->p_stat == SIDL || p->p_stat == SDEAD); 57 58 LIST_REMOVE(p, p_list); 59 LIST_REMOVE(p, p_sibling); 60 proc_free_pid(p->p_pid); /* decrements nprocs */ 61 proc_leavepgrp(p); /* releases proc_lock */ 62 63 cred = p->p_cred; 64 chgproccnt(kauth_cred_getuid(cred), -1); 65 if (rump_proc_vfs_release) 66 rump_proc_vfs_release(p); 67 68 limfree(p->p_limit); 69 pstatsfree(p->p_stats); 70 kauth_cred_free(p->p_cred); 71 proc_finispecific(p); 72 73 mutex_obj_free(p->p_lock); 74 mutex_destroy(&p->p_stmutex); 75 mutex_destroy(&p->p_auxlock); 76 rw_destroy(&p->p_reflock); 77 cv_destroy(&p->p_waitcv); 78 cv_destroy(&p->p_lwpcv); 79 80 /* non-kernel vmspaces are not shared */ 81 if (p->p_vmspace != vmspace_kernel()) { 82 KASSERT(p->p_vmspace->vm_refcnt == 1); 83 kmem_free(p->p_vmspace, sizeof(*p->p_vmspace)); 84 } 85 86 proc_free_mem(p); 87} 88 89/* 90 * Allocate a new process. Mostly mimic fork by 91 * copying the properties of the parent. However, there are some 92 * differences. For example, we never share the fd table. 93 * 94 * Switch to the new lwp and return a pointer to it. 95 */ 96static struct proc * 97lwproc_newproc(struct proc *parent, int flags) 98{ 99 uid_t uid = kauth_cred_getuid(parent->p_cred); 100 struct proc *p; 101 102 /* maxproc not enforced */ 103 atomic_inc_uint(&nprocs); 104 105 /* allocate process */ 106 p = proc_alloc(); 107 memset(&p->p_startzero, 0, 108 offsetof(struct proc, p_endzero) 109 - offsetof(struct proc, p_startzero)); 110 memcpy(&p->p_startcopy, &parent->p_startcopy, 111 offsetof(struct proc, p_endcopy) 112 - offsetof(struct proc, p_startcopy)); 113 114 p->p_stats = pstatscopy(parent->p_stats); 115 116 p->p_vmspace = vmspace_kernel(); 117 p->p_emul = &emul_netbsd; 118 119 if ((flags & RUMP_RFCFDG) == 0) 120 KASSERT(parent == curproc); 121 if (flags & RUMP_RFFDG) 122 p->p_fd = fd_copy(); 123 else if (flags & RUMP_RFCFDG) 124 p->p_fd = fd_init(NULL); 125 else 126 fd_share(p); 127 128 lim_addref(parent->p_limit); 129 p->p_limit = parent->p_limit; 130 131 LIST_INIT(&p->p_lwps); 132 LIST_INIT(&p->p_children); 133 134 p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 135 mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_NONE); 136 mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE); 137 rw_init(&p->p_reflock); 138 cv_init(&p->p_waitcv, "pwait"); 139 cv_init(&p->p_lwpcv, "plwp"); 140 141 p->p_pptr = parent; 142 p->p_ppid = parent->p_pid; 143 144 kauth_proc_fork(parent, p); 145 146 /* initialize cwd in rump kernels with vfs */ 147 if (rump_proc_vfs_init) 148 rump_proc_vfs_init(p); 149 150 chgproccnt(uid, 1); /* not enforced */ 151 152 /* publish proc various proc lists */ 153 mutex_enter(proc_lock); 154 LIST_INSERT_HEAD(&allproc, p, p_list); 155 LIST_INSERT_HEAD(&parent->p_children, p, p_sibling); 156 LIST_INSERT_AFTER(parent, p, p_pglist); 157 mutex_exit(proc_lock); 158 159 return p; 160} 161 162static void 163lwproc_freelwp(struct lwp *l) 164{ 165 struct proc *p; 166 bool freeproc; 167 168 p = l->l_proc; 169 mutex_enter(p->p_lock); 170 171 /* XXX: l_refcnt */ 172 KASSERT(l->l_flag & LW_WEXIT); 173 KASSERT(l->l_refcnt == 0); 174 175 /* ok, zero references, continue with nuke */ 176 LIST_REMOVE(l, l_sibling); 177 KASSERT(p->p_nlwps >= 1); 178 if (--p->p_nlwps == 0) { 179 KASSERT(p != &proc0); 180 p->p_stat = SDEAD; 181 } 182 freeproc = p->p_nlwps == 0; 183 cv_broadcast(&p->p_lwpcv); /* nobody sleeps on this in rump? */ 184 kauth_cred_free(l->l_cred); 185 mutex_exit(p->p_lock); 186 187 mutex_enter(proc_lock); 188 LIST_REMOVE(l, l_list); 189 mutex_exit(proc_lock); 190 191 if (l->l_name) 192 kmem_free(l->l_name, MAXCOMLEN); 193 lwp_finispecific(l); 194 195 kmem_free(l, sizeof(*l)); 196 197 if (p->p_stat == SDEAD) 198 lwproc_proc_free(p); 199} 200 201/* 202 * called with p_lock held, releases lock before return 203 */ 204static void 205lwproc_makelwp(struct proc *p, struct lwp *l, bool doswitch, bool procmake) 206{ 207 208 p->p_nlwps++; 209 l->l_refcnt = 1; 210 l->l_proc = p; 211 212 l->l_lid = p->p_nlwpid++; 213 LIST_INSERT_HEAD(&p->p_lwps, l, l_sibling); 214 mutex_exit(p->p_lock); 215 216 lwp_update_creds(l); 217 218 l->l_fd = p->p_fd; 219 l->l_cpu = NULL; 220 l->l_target_cpu = rump_cpu; /* Initial target CPU always the same */ 221 222 lwp_initspecific(l); 223 224 if (doswitch) { 225 rump_lwproc_switch(l); 226 } 227 228 /* filedesc already has refcount 1 when process is created */ 229 if (!procmake) { 230 fd_hold(l); 231 } 232 233 mutex_enter(proc_lock); 234 LIST_INSERT_HEAD(&alllwp, l, l_list); 235 mutex_exit(proc_lock); 236} 237 238struct lwp * 239rump__lwproc_alloclwp(struct proc *p) 240{ 241 struct lwp *l; 242 bool newproc = false; 243 244 if (p == NULL) { 245 p = lwproc_newproc(&proc0, 0); 246 newproc = true; 247 } 248 249 l = kmem_zalloc(sizeof(*l), KM_SLEEP); 250 251 mutex_enter(p->p_lock); 252 lwproc_makelwp(p, l, false, newproc); 253 254 return l; 255} 256 257int 258rump_lwproc_newlwp(pid_t pid) 259{ 260 struct proc *p; 261 struct lwp *l; 262 263 l = kmem_zalloc(sizeof(*l), KM_SLEEP); 264 mutex_enter(proc_lock); 265 p = proc_find_raw(pid); 266 if (p == NULL) { 267 mutex_exit(proc_lock); 268 kmem_free(l, sizeof(*l)); 269 return ESRCH; 270 } 271 mutex_enter(p->p_lock); 272 mutex_exit(proc_lock); 273 lwproc_makelwp(p, l, true, false); 274 275 return 0; 276} 277 278int 279rump_lwproc_rfork(int flags) 280{ 281 struct proc *p; 282 struct lwp *l; 283 284 if (flags & ~(RUMP_RFFDG|RUMP_RFCFDG) || 285 (~flags & (RUMP_RFFDG|RUMP_RFCFDG)) == 0) 286 return EINVAL; 287 288 p = lwproc_newproc(curproc, flags); 289 l = kmem_zalloc(sizeof(*l), KM_SLEEP); 290 mutex_enter(p->p_lock); 291 lwproc_makelwp(p, l, true, true); 292 293 return 0; 294} 295 296/* 297 * Switch to a new process/thread. Release previous one if 298 * deemed to be exiting. This is considered a slow path for 299 * rump kernel entry. 300 */ 301void 302rump_lwproc_switch(struct lwp *newlwp) 303{ 304 struct lwp *l = curlwp; 305 306 KASSERT(!(l->l_flag & LW_WEXIT) || newlwp); 307 308 if (__predict_false(newlwp && (newlwp->l_pflag & LP_RUNNING))) 309 panic("lwp %p (%d:%d) already running", 310 newlwp, newlwp->l_proc->p_pid, newlwp->l_lid); 311 312 if (newlwp == NULL) { 313 l->l_pflag &= ~LP_RUNNING; 314 l->l_flag |= LW_RUMP_CLEAR; 315 return; 316 } 317 318 /* fd_free() must be called from curlwp context. talk about ugh */ 319 if (l->l_flag & LW_WEXIT) { 320 fd_free(); 321 } 322 323 rumpuser_set_curlwp(NULL); 324 325 newlwp->l_cpu = newlwp->l_target_cpu = l->l_cpu; 326 newlwp->l_mutex = l->l_mutex; 327 newlwp->l_pflag |= LP_RUNNING; 328 329 rumpuser_set_curlwp(newlwp); 330 331 /* 332 * Check if the thread should get a signal. This is 333 * mostly to satisfy the "record" rump sigmodel. 334 */ 335 mutex_enter(newlwp->l_proc->p_lock); 336 if (sigispending(newlwp, 0)) { 337 newlwp->l_flag |= LW_PENDSIG; 338 } 339 mutex_exit(newlwp->l_proc->p_lock); 340 341 l->l_mutex = NULL; 342 l->l_cpu = NULL; 343 l->l_pflag &= ~LP_RUNNING; 344 l->l_flag &= ~LW_PENDSIG; 345 346 if (l->l_flag & LW_WEXIT) { 347 lwproc_freelwp(l); 348 } 349} 350 351void 352rump_lwproc_releaselwp(void) 353{ 354 struct proc *p; 355 struct lwp *l = curlwp; 356 357 if (l->l_refcnt == 0 && l->l_flag & LW_WEXIT) 358 panic("releasing non-pertinent lwp"); 359 360 p = l->l_proc; 361 mutex_enter(p->p_lock); 362 KASSERT(l->l_refcnt != 0); 363 l->l_refcnt--; 364 mutex_exit(p->p_lock); 365 l->l_flag |= LW_WEXIT; /* will be released when unscheduled */ 366} 367 368struct lwp * 369rump_lwproc_curlwp(void) 370{ 371 struct lwp *l = curlwp; 372 373 if (l->l_flag & LW_WEXIT) 374 return NULL; 375 return l; 376} 377