nfs_clsubs.c revision 210201
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: head/sys/fs/nfsclient/nfs_clsubs.c 210201 2010-07-18 00:24:01Z rmacklem $"); 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 80extern struct mtx ncl_iod_mutex; 81extern enum nfsiod_state ncl_iodwant[NFS_MAXRAHEAD]; 82extern struct nfsmount *ncl_iodmount[NFS_MAXRAHEAD]; 83extern int ncl_numasync; 84extern unsigned int ncl_iodmax; 85extern struct nfsstats newnfsstats; 86 87int 88ncl_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 116void 117ncl_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 126void 127ncl_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 135int 136ncl_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 151void 152ncl_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 161void 162ncl_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> 175SYSCTL_DECL(_vfs_newnfs); 176static int nfs_acdebug; 177SYSCTL_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 */ 185int 186ncl_getattrcache(struct vnode *vp, struct vattr *vaper) 187{ 188 struct nfsnode *np; 189 struct vattr *vap; 190 struct nfsmount *nmp; 191 int timeo; 192 boolean_t mustflush; 193 194 np = VTONFS(vp); 195 vap = &np->n_vattr.na_vattr; 196 nmp = VFSTONFS(vp->v_mount); 197 mustflush = nfscl_mustflush(vp); /* must be before mtx_lock() */ 198#ifdef NFS_ACDEBUG 199 mtx_lock(&Giant); /* ncl_printf() */ 200#endif 201 mtx_lock(&np->n_mtx); 202 /* XXX n_mtime doesn't seem to be updated on a miss-and-reload */ 203 timeo = (time_second - np->n_mtime.tv_sec) / 10; 204 205#ifdef NFS_ACDEBUG 206 if (nfs_acdebug>1) 207 ncl_printf("nfs_getattrcache: initial timeo = %d\n", timeo); 208#endif 209 210 if (vap->va_type == VDIR) { 211 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin) 212 timeo = nmp->nm_acdirmin; 213 else if (timeo > nmp->nm_acdirmax) 214 timeo = nmp->nm_acdirmax; 215 } else { 216 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin) 217 timeo = nmp->nm_acregmin; 218 else if (timeo > nmp->nm_acregmax) 219 timeo = nmp->nm_acregmax; 220 } 221 222#ifdef NFS_ACDEBUG 223 if (nfs_acdebug > 2) 224 ncl_printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n", 225 nmp->nm_acregmin, nmp->nm_acregmax, 226 nmp->nm_acdirmin, nmp->nm_acdirmax); 227 228 if (nfs_acdebug) 229 ncl_printf("nfs_getattrcache: age = %d; final timeo = %d\n", 230 (time_second - np->n_attrstamp), timeo); 231#endif 232 233 if ((time_second - np->n_attrstamp) >= timeo && 234 (mustflush || np->n_attrstamp == 0)) { 235 newnfsstats.attrcache_misses++; 236 mtx_unlock(&np->n_mtx); 237#ifdef NFS_ACDEBUG 238 mtx_unlock(&Giant); /* ncl_printf() */ 239#endif 240 return( ENOENT); 241 } 242 newnfsstats.attrcache_hits++; 243 if (vap->va_size != np->n_size) { 244 if (vap->va_type == VREG) { 245 if (np->n_flag & NMODIFIED) { 246 if (vap->va_size < np->n_size) 247 vap->va_size = np->n_size; 248 else 249 np->n_size = vap->va_size; 250 } else { 251 np->n_size = vap->va_size; 252 } 253 vnode_pager_setsize(vp, np->n_size); 254 } else { 255 np->n_size = vap->va_size; 256 } 257 } 258 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr)); 259 if (np->n_flag & NCHG) { 260 if (np->n_flag & NACC) 261 vaper->va_atime = np->n_atim; 262 if (np->n_flag & NUPD) 263 vaper->va_mtime = np->n_mtim; 264 } 265 mtx_unlock(&np->n_mtx); 266#ifdef NFS_ACDEBUG 267 mtx_unlock(&Giant); /* ncl_printf() */ 268#endif 269 return (0); 270} 271 272static nfsuint64 nfs_nullcookie = { { 0, 0 } }; 273/* 274 * This function finds the directory cookie that corresponds to the 275 * logical byte offset given. 276 */ 277nfsuint64 * 278ncl_getcookie(struct nfsnode *np, off_t off, int add) 279{ 280 struct nfsdmap *dp, *dp2; 281 int pos; 282 nfsuint64 *retval = NULL; 283 284 pos = (uoff_t)off / NFS_DIRBLKSIZ; 285 if (pos == 0 || off < 0) { 286 KASSERT(!add, ("nfs getcookie add at <= 0")); 287 return (&nfs_nullcookie); 288 } 289 pos--; 290 dp = LIST_FIRST(&np->n_cookies); 291 if (!dp) { 292 if (add) { 293 MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap), 294 M_NFSDIROFF, M_WAITOK); 295 dp->ndm_eocookie = 0; 296 LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list); 297 } else 298 goto out; 299 } 300 while (pos >= NFSNUMCOOKIES) { 301 pos -= NFSNUMCOOKIES; 302 if (LIST_NEXT(dp, ndm_list)) { 303 if (!add && dp->ndm_eocookie < NFSNUMCOOKIES && 304 pos >= dp->ndm_eocookie) 305 goto out; 306 dp = LIST_NEXT(dp, ndm_list); 307 } else if (add) { 308 MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap), 309 M_NFSDIROFF, M_WAITOK); 310 dp2->ndm_eocookie = 0; 311 LIST_INSERT_AFTER(dp, dp2, ndm_list); 312 dp = dp2; 313 } else 314 goto out; 315 } 316 if (pos >= dp->ndm_eocookie) { 317 if (add) 318 dp->ndm_eocookie = pos + 1; 319 else 320 goto out; 321 } 322 retval = &dp->ndm_cookies[pos]; 323out: 324 return (retval); 325} 326 327/* 328 * Invalidate cached directory information, except for the actual directory 329 * blocks (which are invalidated separately). 330 * Done mainly to avoid the use of stale offset cookies. 331 */ 332void 333ncl_invaldir(struct vnode *vp) 334{ 335 struct nfsnode *np = VTONFS(vp); 336 337 KASSERT(vp->v_type == VDIR, ("nfs: invaldir not dir")); 338 ncl_dircookie_lock(np); 339 np->n_direofoffset = 0; 340 np->n_cookieverf.nfsuquad[0] = 0; 341 np->n_cookieverf.nfsuquad[1] = 0; 342 if (LIST_FIRST(&np->n_cookies)) 343 LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0; 344 ncl_dircookie_unlock(np); 345} 346 347/* 348 * The write verifier has changed (probably due to a server reboot), so all 349 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the 350 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT 351 * and B_CLUSTEROK flags. Once done the new write verifier can be set for the 352 * mount point. 353 * 354 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data 355 * writes are not clusterable. 356 */ 357void 358ncl_clearcommit(struct mount *mp) 359{ 360 struct vnode *vp, *nvp; 361 struct buf *bp, *nbp; 362 struct bufobj *bo; 363 364 MNT_ILOCK(mp); 365 MNT_VNODE_FOREACH(vp, mp, nvp) { 366 bo = &vp->v_bufobj; 367 VI_LOCK(vp); 368 if (vp->v_iflag & VI_DOOMED) { 369 VI_UNLOCK(vp); 370 continue; 371 } 372 vholdl(vp); 373 VI_UNLOCK(vp); 374 MNT_IUNLOCK(mp); 375 BO_LOCK(bo); 376 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) { 377 if (!BUF_ISLOCKED(bp) && 378 (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT)) 379 == (B_DELWRI | B_NEEDCOMMIT)) 380 bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK); 381 } 382 BO_UNLOCK(bo); 383 vdrop(vp); 384 MNT_ILOCK(mp); 385 } 386 MNT_IUNLOCK(mp); 387} 388 389/* 390 * Called once to initialize data structures... 391 */ 392int 393ncl_init(struct vfsconf *vfsp) 394{ 395 int i; 396 397 /* Ensure async daemons disabled */ 398 for (i = 0; i < NFS_MAXRAHEAD; i++) { 399 ncl_iodwant[i] = NFSIOD_NOT_AVAILABLE; 400 ncl_iodmount[i] = NULL; 401 } 402 ncl_nhinit(); /* Init the nfsnode table */ 403 404 return (0); 405} 406 407