nfs_iod.c revision 1.3.2.4
1 /* $NetBSD: nfs_iod.c,v 1.3.2.4 2010/03/11 15:04:31 yamt Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * Rick Macklem at The University of Guelph. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)nfs_syscalls.c 8.5 (Berkeley) 3/30/95 35 */ 36 37 #include <sys/cdefs.h> 38 __KERNEL_RCSID(0, "$NetBSD: nfs_iod.c,v 1.3.2.4 2010/03/11 15:04:31 yamt Exp $"); 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/kernel.h> 43 #include <sys/file.h> 44 #include <sys/stat.h> 45 #include <sys/vnode.h> 46 #include <sys/mount.h> 47 #include <sys/proc.h> 48 #include <sys/uio.h> 49 #include <sys/malloc.h> 50 #include <sys/kmem.h> 51 #include <sys/buf.h> 52 #include <sys/mbuf.h> 53 #include <sys/socket.h> 54 #include <sys/socketvar.h> 55 #include <sys/signalvar.h> 56 #include <sys/domain.h> 57 #include <sys/protosw.h> 58 #include <sys/namei.h> 59 #include <sys/syslog.h> 60 #include <sys/filedesc.h> 61 #include <sys/kthread.h> 62 #include <sys/kauth.h> 63 #include <sys/syscallargs.h> 64 65 #include <netinet/in.h> 66 #include <netinet/tcp.h> 67 #include <nfs/xdr_subs.h> 68 #include <nfs/rpcv2.h> 69 #include <nfs/nfsproto.h> 70 #include <nfs/nfs.h> 71 #include <nfs/nfsm_subs.h> 72 #include <nfs/nfsrvcache.h> 73 #include <nfs/nfsmount.h> 74 #include <nfs/nfsnode.h> 75 #include <nfs/nfsrtt.h> 76 #include <nfs/nfs_var.h> 77 78 extern int nuidhash_max; 79 80 /* 81 * locking order: 82 * nfs_iodlist_lock -> nid_lock -> nm_lock 83 */ 84 kmutex_t nfs_iodlist_lock; 85 struct nfs_iodlist nfs_iodlist_idle; 86 struct nfs_iodlist nfs_iodlist_all; 87 int nfs_niothreads = -1; /* == "0, and has never been set" */ 88 int nfs_defect = 0; 89 90 /* 91 * Asynchronous I/O threads for client nfs. 92 * They do read-ahead and write-behind operations on the block I/O cache. 93 * Never returns unless it fails or gets killed. 94 */ 95 96 static void 97 nfssvc_iod(void *arg) 98 { 99 struct buf *bp; 100 struct nfs_iod *myiod; 101 struct nfsmount *nmp; 102 103 myiod = kmem_alloc(sizeof(*myiod), KM_SLEEP); 104 mutex_init(&myiod->nid_lock, MUTEX_DEFAULT, IPL_NONE); 105 cv_init(&myiod->nid_cv, "nfsiod"); 106 myiod->nid_exiting = false; 107 myiod->nid_mount = NULL; 108 mutex_enter(&nfs_iodlist_lock); 109 LIST_INSERT_HEAD(&nfs_iodlist_all, myiod, nid_all); 110 mutex_exit(&nfs_iodlist_lock); 111 112 for (;;) { 113 mutex_enter(&nfs_iodlist_lock); 114 LIST_INSERT_HEAD(&nfs_iodlist_idle, myiod, nid_idle); 115 mutex_exit(&nfs_iodlist_lock); 116 117 mutex_enter(&myiod->nid_lock); 118 while (/*CONSTCOND*/ true) { 119 nmp = myiod->nid_mount; 120 if (nmp) { 121 myiod->nid_mount = NULL; 122 break; 123 } 124 if (__predict_false(myiod->nid_exiting)) { 125 /* 126 * drop nid_lock to preserve locking order. 127 */ 128 mutex_exit(&myiod->nid_lock); 129 mutex_enter(&nfs_iodlist_lock); 130 mutex_enter(&myiod->nid_lock); 131 /* 132 * recheck nid_mount because nfs_asyncio can 133 * pick us in the meantime as we are still on 134 * nfs_iodlist_lock. 135 */ 136 if (myiod->nid_mount != NULL) { 137 mutex_exit(&nfs_iodlist_lock); 138 continue; 139 } 140 LIST_REMOVE(myiod, nid_idle); 141 mutex_exit(&nfs_iodlist_lock); 142 goto quit; 143 } 144 cv_wait(&myiod->nid_cv, &myiod->nid_lock); 145 } 146 mutex_exit(&myiod->nid_lock); 147 148 mutex_enter(&nmp->nm_lock); 149 while ((bp = TAILQ_FIRST(&nmp->nm_bufq)) != NULL) { 150 /* Take one off the front of the list */ 151 TAILQ_REMOVE(&nmp->nm_bufq, bp, b_freelist); 152 nmp->nm_bufqlen--; 153 if (nmp->nm_bufqlen < 2 * nmp->nm_bufqiods) { 154 cv_broadcast(&nmp->nm_aiocv); 155 } 156 mutex_exit(&nmp->nm_lock); 157 (void)nfs_doio(bp); 158 mutex_enter(&nmp->nm_lock); 159 /* 160 * If there are more than one iod on this mount, 161 * then defect so that the iods can be shared out 162 * fairly between the mounts 163 */ 164 if (nfs_defect && nmp->nm_bufqiods > 1) { 165 break; 166 } 167 } 168 KASSERT(nmp->nm_bufqiods > 0); 169 nmp->nm_bufqiods--; 170 mutex_exit(&nmp->nm_lock); 171 } 172 quit: 173 KASSERT(myiod->nid_mount == NULL); 174 mutex_exit(&myiod->nid_lock); 175 176 cv_destroy(&myiod->nid_cv); 177 mutex_destroy(&myiod->nid_lock); 178 kmem_free(myiod, sizeof(*myiod)); 179 180 kthread_exit(0); 181 } 182 183 void 184 nfs_iodinit(void) 185 { 186 187 mutex_init(&nfs_iodlist_lock, MUTEX_DEFAULT, IPL_NONE); 188 LIST_INIT(&nfs_iodlist_all); 189 LIST_INIT(&nfs_iodlist_idle); 190 } 191 192 void 193 nfs_iodfini(void) 194 { 195 int error; 196 197 error = nfs_set_niothreads(0); 198 KASSERT(error == 0); 199 mutex_destroy(&nfs_iodlist_lock); 200 } 201 202 int 203 nfs_set_niothreads(int newval) 204 { 205 struct nfs_iod *nid; 206 int error = 0; 207 int hold_count; 208 209 KERNEL_UNLOCK_ALL(curlwp, &hold_count); 210 211 mutex_enter(&nfs_iodlist_lock); 212 /* clamp to sane range */ 213 nfs_niothreads = max(0, min(newval, NFS_MAXASYNCDAEMON)); 214 215 while (nfs_numasync != nfs_niothreads && error == 0) { 216 while (nfs_numasync < nfs_niothreads) { 217 218 /* 219 * kthread_create can wait for pagedaemon and 220 * pagedaemon can wait for nfsiod which needs to acquire 221 * nfs_iodlist_lock. 222 */ 223 224 mutex_exit(&nfs_iodlist_lock); 225 error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL, 226 nfssvc_iod, NULL, NULL, "nfsio"); 227 mutex_enter(&nfs_iodlist_lock); 228 if (error) { 229 /* give up */ 230 nfs_niothreads = nfs_numasync; 231 break; 232 } 233 nfs_numasync++; 234 } 235 while (nfs_numasync > nfs_niothreads) { 236 nid = LIST_FIRST(&nfs_iodlist_all); 237 if (nid == NULL) { 238 /* iod has not started yet. */ 239 kpause("nfsiorm", false, hz, &nfs_iodlist_lock); 240 continue; 241 } 242 LIST_REMOVE(nid, nid_all); 243 mutex_enter(&nid->nid_lock); 244 KASSERT(!nid->nid_exiting); 245 nid->nid_exiting = true; 246 cv_signal(&nid->nid_cv); 247 mutex_exit(&nid->nid_lock); 248 nfs_numasync--; 249 } 250 } 251 mutex_exit(&nfs_iodlist_lock); 252 253 KERNEL_LOCK(hold_count, curlwp); 254 return error; 255 } 256 257 /* 258 * Get an authorization string for the uid by having the mount_nfs sitting 259 * on this mount point porpous out of the kernel and do it. 260 */ 261 int 262 nfs_getauth(struct nfsmount *nmp, struct nfsreq *rep, kauth_cred_t cred, char **auth_str, int *auth_len, char *verf_str, int *verf_len, NFSKERBKEY_T key) 263 /* key: return session key */ 264 { 265 int error = 0; 266 267 while ((nmp->nm_iflag & NFSMNT_WAITAUTH) == 0) { 268 nmp->nm_iflag |= NFSMNT_WANTAUTH; 269 (void) tsleep((void *)&nmp->nm_authtype, PSOCK, 270 "nfsauth1", 2 * hz); 271 error = nfs_sigintr(nmp, rep, rep->r_lwp); 272 if (error) { 273 nmp->nm_iflag &= ~NFSMNT_WANTAUTH; 274 return (error); 275 } 276 } 277 nmp->nm_iflag &= ~(NFSMNT_WAITAUTH | NFSMNT_WANTAUTH); 278 nmp->nm_authstr = *auth_str = (char *)malloc(RPCAUTH_MAXSIZ, M_TEMP, M_WAITOK); 279 nmp->nm_authlen = RPCAUTH_MAXSIZ; 280 nmp->nm_verfstr = verf_str; 281 nmp->nm_verflen = *verf_len; 282 nmp->nm_authuid = kauth_cred_geteuid(cred); 283 wakeup((void *)&nmp->nm_authstr); 284 285 /* 286 * And wait for mount_nfs to do its stuff. 287 */ 288 while ((nmp->nm_iflag & NFSMNT_HASAUTH) == 0 && error == 0) { 289 (void) tsleep((void *)&nmp->nm_authlen, PSOCK, 290 "nfsauth2", 2 * hz); 291 error = nfs_sigintr(nmp, rep, rep->r_lwp); 292 } 293 if (nmp->nm_iflag & NFSMNT_AUTHERR) { 294 nmp->nm_iflag &= ~NFSMNT_AUTHERR; 295 error = EAUTH; 296 } 297 if (error) 298 free((void *)*auth_str, M_TEMP); 299 else { 300 *auth_len = nmp->nm_authlen; 301 *verf_len = nmp->nm_verflen; 302 memcpy(key, nmp->nm_key, sizeof (NFSKERBKEY_T)); 303 } 304 nmp->nm_iflag &= ~NFSMNT_HASAUTH; 305 nmp->nm_iflag |= NFSMNT_WAITAUTH; 306 if (nmp->nm_iflag & NFSMNT_WANTAUTH) { 307 nmp->nm_iflag &= ~NFSMNT_WANTAUTH; 308 wakeup((void *)&nmp->nm_authtype); 309 } 310 return (error); 311 } 312 313 /* 314 * Get a nickname authenticator and verifier. 315 */ 316 int 317 nfs_getnickauth(struct nfsmount *nmp, kauth_cred_t cred, char **auth_str, 318 int *auth_len, char *verf_str, int verf_len) 319 { 320 struct timeval ktvin, ktvout, tv; 321 struct nfsuid *nuidp; 322 u_int32_t *nickp, *verfp; 323 324 memset(&ktvout, 0, sizeof ktvout); /* XXX gcc */ 325 326 #ifdef DIAGNOSTIC 327 if (verf_len < (4 * NFSX_UNSIGNED)) 328 panic("nfs_getnickauth verf too small"); 329 #endif 330 LIST_FOREACH(nuidp, NMUIDHASH(nmp, kauth_cred_geteuid(cred)), nu_hash) { 331 if (kauth_cred_geteuid(nuidp->nu_cr) == kauth_cred_geteuid(cred)) 332 break; 333 } 334 if (!nuidp || nuidp->nu_expire < time_second) 335 return (EACCES); 336 337 /* 338 * Move to the end of the lru list (end of lru == most recently used). 339 */ 340 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru); 341 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, nu_lru); 342 343 nickp = (u_int32_t *)malloc(2 * NFSX_UNSIGNED, M_TEMP, M_WAITOK); 344 *nickp++ = txdr_unsigned(RPCAKN_NICKNAME); 345 *nickp = txdr_unsigned(nuidp->nu_nickname); 346 *auth_str = (char *)nickp; 347 *auth_len = 2 * NFSX_UNSIGNED; 348 349 /* 350 * Now we must encrypt the verifier and package it up. 351 */ 352 verfp = (u_int32_t *)verf_str; 353 *verfp++ = txdr_unsigned(RPCAKN_NICKNAME); 354 getmicrotime(&tv); 355 if (tv.tv_sec > nuidp->nu_timestamp.tv_sec || 356 (tv.tv_sec == nuidp->nu_timestamp.tv_sec && 357 tv.tv_usec > nuidp->nu_timestamp.tv_usec)) 358 nuidp->nu_timestamp = tv; 359 else 360 nuidp->nu_timestamp.tv_usec++; 361 ktvin.tv_sec = txdr_unsigned(nuidp->nu_timestamp.tv_sec); 362 ktvin.tv_usec = txdr_unsigned(nuidp->nu_timestamp.tv_usec); 363 364 /* 365 * Now encrypt the timestamp verifier in ecb mode using the session 366 * key. 367 */ 368 #ifdef NFSKERB 369 XXX 370 #endif 371 372 *verfp++ = ktvout.tv_sec; 373 *verfp++ = ktvout.tv_usec; 374 *verfp = 0; 375 return (0); 376 } 377 378 /* 379 * Save the current nickname in a hash list entry on the mount point. 380 */ 381 int 382 nfs_savenickauth(struct nfsmount *nmp, kauth_cred_t cred, int len, NFSKERBKEY_T key, struct mbuf **mdp, char **dposp, struct mbuf *mrep) 383 { 384 struct nfsuid *nuidp; 385 u_int32_t *tl; 386 int32_t t1; 387 struct mbuf *md = *mdp; 388 struct timeval ktvin, ktvout; 389 u_int32_t nick; 390 char *dpos = *dposp, *cp2; 391 int deltasec, error = 0; 392 393 memset(&ktvout, 0, sizeof ktvout); /* XXX gcc */ 394 395 if (len == (3 * NFSX_UNSIGNED)) { 396 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 397 ktvin.tv_sec = *tl++; 398 ktvin.tv_usec = *tl++; 399 nick = fxdr_unsigned(u_int32_t, *tl); 400 401 /* 402 * Decrypt the timestamp in ecb mode. 403 */ 404 #ifdef NFSKERB 405 XXX 406 #endif 407 ktvout.tv_sec = fxdr_unsigned(long, ktvout.tv_sec); 408 ktvout.tv_usec = fxdr_unsigned(long, ktvout.tv_usec); 409 deltasec = time_second - ktvout.tv_sec; 410 if (deltasec < 0) 411 deltasec = -deltasec; 412 /* 413 * If ok, add it to the hash list for the mount point. 414 */ 415 if (deltasec <= NFS_KERBCLOCKSKEW) { 416 if (nmp->nm_numuids < nuidhash_max) { 417 nmp->nm_numuids++; 418 nuidp = kmem_alloc(sizeof(*nuidp), KM_SLEEP); 419 } else { 420 nuidp = TAILQ_FIRST(&nmp->nm_uidlruhead); 421 LIST_REMOVE(nuidp, nu_hash); 422 TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, 423 nu_lru); 424 } 425 nuidp->nu_flag = 0; 426 kauth_cred_seteuid(nuidp->nu_cr, kauth_cred_geteuid(cred)); 427 nuidp->nu_expire = time_second + NFS_KERBTTL; 428 nuidp->nu_timestamp = ktvout; 429 nuidp->nu_nickname = nick; 430 memcpy(nuidp->nu_key, key, sizeof (NFSKERBKEY_T)); 431 TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, 432 nu_lru); 433 LIST_INSERT_HEAD(NMUIDHASH(nmp, kauth_cred_geteuid(cred)), 434 nuidp, nu_hash); 435 } 436 } else 437 nfsm_adv(nfsm_rndup(len)); 438 nfsmout: 439 *mdp = md; 440 *dposp = dpos; 441 return (error); 442 } 443
Indexes created Tue Sep 30 10:09:55 GMT 2025