1 /*- 2 * Copyright (c) 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software contributed to Berkeley by 6 * Rick Macklem at The University of Guelph. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 4. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * from nfs_subs.c 8.8 (Berkeley) 5/22/95 33 */ 34 35 #include <sys/cdefs.h> 36 __FBSDID("$FreeBSD$"); 37 38 /* 39 * These functions support the macros and help fiddle mbuf chains for 40 * the nfs op functions. They do things like create the rpc header and 41 * copy data between mbuf chains and uio lists. 42 */ 43 44 #include <sys/param.h> 45 #include <sys/systm.h> 46 #include <sys/kernel.h> 47 #include <sys/bio.h> 48 #include <sys/buf.h> 49 #include <sys/proc.h> 50 #include <sys/mount.h> 51 #include <sys/vnode.h> 52 #include <sys/namei.h> 53 #include <sys/mbuf.h> 54 #include <sys/socket.h> 55 #include <sys/stat.h> 56 #include <sys/malloc.h> 57 #include <sys/sysent.h> 58 #include <sys/syscall.h> 59 #include <sys/sysproto.h> 60 61 #include <vm/vm.h> 62 #include <vm/vm_object.h> 63 #include <vm/vm_extern.h> 64 #include <vm/uma.h> 65 66 #include <fs/nfs/nfsport.h> 67 #include <fs/nfsclient/nfsnode.h> 68 #include <fs/nfsclient/nfsmount.h> 69 #include <fs/nfsclient/nfs.h> 70 #include <fs/nfsclient/nfs_lock.h> 71 72 #include <netinet/in.h> 73 74 /* 75 * Note that stdarg.h and the ANSI style va_start macro is used for both 76 * ANSI and traditional C compilers. 77 */ 78 #include <machine/stdarg.h> 79 80 extern struct mtx ncl_iod_mutex; 81 extern enum nfsiod_state ncl_iodwant[NFS_MAXRAHEAD]; 82 extern struct nfsmount *ncl_iodmount[NFS_MAXRAHEAD]; 83 extern int ncl_numasync; 84 extern unsigned int ncl_iodmax; 85 extern struct nfsstats newnfsstats; 86 87 int 88 ncl_uninit(struct vfsconf *vfsp) 89 { 90 /* 91 * XXX: Unloading of nfscl module is unsupported. 92 */ 93 #if 0 94 int i; 95 96 /* 97 * Tell all nfsiod processes to exit. Clear ncl_iodmax, and wakeup 98 * any sleeping nfsiods so they check ncl_iodmax and exit. 99 */ 100 mtx_lock(&ncl_iod_mutex); 101 ncl_iodmax = 0; 102 for (i = 0; i < ncl_numasync; i++) 103 if (ncl_iodwant[i] == NFSIOD_AVAILABLE) 104 wakeup(&ncl_iodwant[i]); 105 /* The last nfsiod to exit will wake us up when ncl_numasync hits 0 */ 106 while (ncl_numasync) 107 msleep(&ncl_numasync, &ncl_iod_mutex, PWAIT, "ioddie", 0); 108 mtx_unlock(&ncl_iod_mutex); 109 ncl_nhuninit(); 110 return (0); 111 #else 112 return (EOPNOTSUPP); 113 #endif 114 } 115 116 void 117 ncl_dircookie_lock(struct nfsnode *np) 118 { 119 mtx_lock(&np->n_mtx); 120 while (np->n_flag & NDIRCOOKIELK) 121 (void) msleep(&np->n_flag, &np->n_mtx, PZERO, "nfsdirlk", 0); 122 np->n_flag |= NDIRCOOKIELK; 123 mtx_unlock(&np->n_mtx); 124 } 125 126 void 127 ncl_dircookie_unlock(struct nfsnode *np) 128 { 129 mtx_lock(&np->n_mtx); 130 np->n_flag &= ~NDIRCOOKIELK; 131 wakeup(&np->n_flag); 132 mtx_unlock(&np->n_mtx); 133 } 134 135 int 136 ncl_upgrade_vnlock(struct vnode *vp) 137 { 138 int old_lock; 139 140 ASSERT_VOP_LOCKED(vp, "ncl_upgrade_vnlock"); 141 old_lock = VOP_ISLOCKED(vp); 142 if (old_lock != LK_EXCLUSIVE) { 143 KASSERT(old_lock == LK_SHARED, 144 ("ncl_upgrade_vnlock: wrong old_lock %d", old_lock)); 145 /* Upgrade to exclusive lock, this might block */ 146 vn_lock(vp, LK_UPGRADE | LK_RETRY); 147 } 148 return (old_lock); 149 } 150 151 void 152 ncl_downgrade_vnlock(struct vnode *vp, int old_lock) 153 { 154 if (old_lock != LK_EXCLUSIVE) { 155 KASSERT(old_lock == LK_SHARED, ("wrong old_lock %d", old_lock)); 156 /* Downgrade from exclusive lock. */ 157 vn_lock(vp, LK_DOWNGRADE | LK_RETRY); 158 } 159 } 160 161 void 162 ncl_printf(const char *fmt, ...) 163 { 164 va_list ap; 165 166 mtx_lock(&Giant); 167 va_start(ap, fmt); 168 printf(fmt, ap); 169 va_end(ap); 170 mtx_unlock(&Giant); 171 } 172 173 #ifdef NFS_ACDEBUG 174 #include <sys/sysctl.h> 175 SYSCTL_DECL(_vfs_newnfs); 176 static int nfs_acdebug; 177 SYSCTL_INT(_vfs_newnfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, ""); 178 #endif 179 180 /* 181 * Check the time stamp 182 * If the cache is valid, copy contents to *vap and return 0 183 * otherwise return an error 184 */ 185 int 186 ncl_getattrcache(struct vnode *vp, struct vattr *vaper) 187 { 188 struct nfsnode *np; 189 struct vattr *vap; 190 struct nfsmount *nmp; 191 int timeo, mustflush; 192 193 np = VTONFS(vp); 194 vap = &np->n_vattr.na_vattr; 195 nmp = VFSTONFS(vp->v_mount); 196 mustflush = nfscl_mustflush(vp); /* must be before mtx_lock() */ 197 #ifdef NFS_ACDEBUG 198 mtx_lock(&Giant); /* ncl_printf() */ 199 #endif 200 mtx_lock(&np->n_mtx); 201 /* XXX n_mtime doesn't seem to be updated on a miss-and-reload */ 202 timeo = (time_second - np->n_mtime.tv_sec) / 10; 203 204 #ifdef NFS_ACDEBUG 205 if (nfs_acdebug>1) 206 ncl_printf("nfs_getattrcache: initial timeo = %d\n", timeo); 207 #endif 208 209 if (vap->va_type == VDIR) { 210 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin) 211 timeo = nmp->nm_acdirmin; 212 else if (timeo > nmp->nm_acdirmax) 213 timeo = nmp->nm_acdirmax; 214 } else { 215 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin) 216 timeo = nmp->nm_acregmin; 217 else if (timeo > nmp->nm_acregmax) 218 timeo = nmp->nm_acregmax; 219 } 220 221 #ifdef NFS_ACDEBUG 222 if (nfs_acdebug > 2) 223 ncl_printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n", 224 nmp->nm_acregmin, nmp->nm_acregmax, 225 nmp->nm_acdirmin, nmp->nm_acdirmax); 226 227 if (nfs_acdebug) 228 ncl_printf("nfs_getattrcache: age = %d; final timeo = %d\n", 229 (time_second - np->n_attrstamp), timeo); 230 #endif 231 232 if ((time_second - np->n_attrstamp) >= timeo && 233 mustflush != 0) { 234 newnfsstats.attrcache_misses++; 235 mtx_unlock(&np->n_mtx); 236 #ifdef NFS_ACDEBUG 237 mtx_unlock(&Giant); /* ncl_printf() */ 238 #endif 239 return( ENOENT); 240 } 241 newnfsstats.attrcache_hits++; 242 if (vap->va_size != np->n_size) { 243 if (vap->va_type == VREG) { 244 if (np->n_flag & NMODIFIED) { 245 if (vap->va_size < np->n_size) 246 vap->va_size = np->n_size; 247 else 248 np->n_size = vap->va_size; 249 } else { 250 np->n_size = vap->va_size; 251 } 252 vnode_pager_setsize(vp, np->n_size); 253 } else { 254 np->n_size = vap->va_size; 255 } 256 } 257 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr)); 258 if (np->n_flag & NCHG) { 259 if (np->n_flag & NACC) 260 vaper->va_atime = np->n_atim; 261 if (np->n_flag & NUPD) 262 vaper->va_mtime = np->n_mtim; 263 } 264 mtx_unlock(&np->n_mtx); 265 #ifdef NFS_ACDEBUG 266 mtx_unlock(&Giant); /* ncl_printf() */ 267 #endif 268 return (0); 269 } 270 271 static nfsuint64 nfs_nullcookie = { { 0, 0 } }; 272 /* 273 * This function finds the directory cookie that corresponds to the 274 * logical byte offset given. 275 */ 276 nfsuint64 * 277 ncl_getcookie(struct nfsnode *np, off_t off, int add) 278 { 279 struct nfsdmap *dp, *dp2; 280 int pos; 281 nfsuint64 *retval = NULL; 282 283 pos = (uoff_t)off / NFS_DIRBLKSIZ; 284 if (pos == 0 || off < 0) { 285 #ifdef DIAGNOSTIC 286 if (add) 287 panic("nfs getcookie add at <= 0"); 288 #endif 289 return (&nfs_nullcookie); 290 } 291 pos--; 292 dp = LIST_FIRST(&np->n_cookies); 293 if (!dp) { 294 if (add) { 295 MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap), 296 M_NFSDIROFF, M_WAITOK); 297 dp->ndm_eocookie = 0; 298 LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list); 299 } else 300 goto out; 301 } 302 while (pos >= NFSNUMCOOKIES) { 303 pos -= NFSNUMCOOKIES; 304 if (LIST_NEXT(dp, ndm_list)) { 305 if (!add && dp->ndm_eocookie < NFSNUMCOOKIES && 306 pos >= dp->ndm_eocookie) 307 goto out; 308 dp = LIST_NEXT(dp, ndm_list); 309 } else if (add) { 310 MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap), 311 M_NFSDIROFF, M_WAITOK); 312 dp2->ndm_eocookie = 0; 313 LIST_INSERT_AFTER(dp, dp2, ndm_list); 314 dp = dp2; 315 } else 316 goto out; 317 } 318 if (pos >= dp->ndm_eocookie) { 319 if (add) 320 dp->ndm_eocookie = pos + 1; 321 else 322 goto out; 323 } 324 retval = &dp->ndm_cookies[pos]; 325 out: 326 return (retval); 327 } 328 329 /* 330 * Invalidate cached directory information, except for the actual directory 331 * blocks (which are invalidated separately). 332 * Done mainly to avoid the use of stale offset cookies. 333 */ 334 void 335 ncl_invaldir(struct vnode *vp) 336 { 337 struct nfsnode *np = VTONFS(vp); 338 339 #ifdef DIAGNOSTIC 340 if (vp->v_type != VDIR) 341 panic("nfs: invaldir not dir"); 342 #endif 343 ncl_dircookie_lock(np); 344 np->n_direofoffset = 0; 345 np->n_cookieverf.nfsuquad[0] = 0; 346 np->n_cookieverf.nfsuquad[1] = 0; 347 if (LIST_FIRST(&np->n_cookies)) 348 LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0; 349 ncl_dircookie_unlock(np); 350 } 351 352 /* 353 * The write verifier has changed (probably due to a server reboot), so all 354 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the 355 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT 356 * and B_CLUSTEROK flags. Once done the new write verifier can be set for the 357 * mount point. 358 * 359 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data 360 * writes are not clusterable. 361 */ 362 void 363 ncl_clearcommit(struct mount *mp) 364 { 365 struct vnode *vp, *nvp; 366 struct buf *bp, *nbp; 367 struct bufobj *bo; 368 369 MNT_ILOCK(mp); 370 MNT_VNODE_FOREACH(vp, mp, nvp) { 371 bo = &vp->v_bufobj; 372 VI_LOCK(vp); 373 if (vp->v_iflag & VI_DOOMED) { 374 VI_UNLOCK(vp); 375 continue; 376 } 377 vholdl(vp); 378 VI_UNLOCK(vp); 379 MNT_IUNLOCK(mp); 380 BO_LOCK(bo); 381 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) { 382 if (!BUF_ISLOCKED(bp) && 383 (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT)) 384 == (B_DELWRI | B_NEEDCOMMIT)) 385 bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK); 386 } 387 BO_UNLOCK(bo); 388 vdrop(vp); 389 MNT_ILOCK(mp); 390 } 391 MNT_IUNLOCK(mp); 392 } 393 394 /* 395 * Called once to initialize data structures... 396 */ 397 int 398 ncl_init(struct vfsconf *vfsp) 399 { 400 int i; 401 402 /* Ensure async daemons disabled */ 403 for (i = 0; i < NFS_MAXRAHEAD; i++) { 404 ncl_iodwant[i] = NFSIOD_NOT_AVAILABLE; 405 ncl_iodmount[i] = NULL; 406 } 407 ncl_nhinit(); /* Init the nfsnode table */ 408 409 return (0); 410 } 411 412