nfs_clsubs.c revision 306663
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: stable/10/sys/fs/nfsclient/nfs_clsubs.c 306663 2016-10-03 23:17:57Z rmacklem $"); 37 38#include "opt_kdtrace.h" 39 40/* 41 * These functions support the macros and help fiddle mbuf chains for 42 * the nfs op functions. They do things like create the rpc header and 43 * copy data between mbuf chains and uio lists. 44 */ 45 46#include <sys/param.h> 47#include <sys/systm.h> 48#include <sys/kernel.h> 49#include <sys/bio.h> 50#include <sys/buf.h> 51#include <sys/proc.h> 52#include <sys/mount.h> 53#include <sys/vnode.h> 54#include <sys/namei.h> 55#include <sys/mbuf.h> 56#include <sys/socket.h> 57#include <sys/stat.h> 58#include <sys/malloc.h> 59#include <sys/sysent.h> 60#include <sys/syscall.h> 61#include <sys/sysproto.h> 62#include <sys/taskqueue.h> 63 64#include <vm/vm.h> 65#include <vm/vm_object.h> 66#include <vm/vm_extern.h> 67#include <vm/uma.h> 68 69#include <fs/nfs/nfsport.h> 70#include <fs/nfsclient/nfsnode.h> 71#include <fs/nfsclient/nfsmount.h> 72#include <fs/nfsclient/nfs.h> 73#include <fs/nfsclient/nfs_kdtrace.h> 74 75#include <netinet/in.h> 76 77/* 78 * Note that stdarg.h and the ANSI style va_start macro is used for both 79 * ANSI and traditional C compilers. 80 */ 81#include <machine/stdarg.h> 82 83extern struct mtx ncl_iod_mutex; 84extern enum nfsiod_state ncl_iodwant[NFS_MAXASYNCDAEMON]; 85extern struct nfsmount *ncl_iodmount[NFS_MAXASYNCDAEMON]; 86extern int ncl_numasync; 87extern unsigned int ncl_iodmax; 88extern struct nfsstats newnfsstats; 89 90struct task ncl_nfsiodnew_task; 91 92int 93ncl_uninit(struct vfsconf *vfsp) 94{ 95 /* 96 * XXX: Unloading of nfscl module is unsupported. 97 */ 98#if 0 99 int i; 100 101 /* 102 * Tell all nfsiod processes to exit. Clear ncl_iodmax, and wakeup 103 * any sleeping nfsiods so they check ncl_iodmax and exit. 104 */ 105 mtx_lock(&ncl_iod_mutex); 106 ncl_iodmax = 0; 107 for (i = 0; i < ncl_numasync; i++) 108 if (ncl_iodwant[i] == NFSIOD_AVAILABLE) 109 wakeup(&ncl_iodwant[i]); 110 /* The last nfsiod to exit will wake us up when ncl_numasync hits 0 */ 111 while (ncl_numasync) 112 msleep(&ncl_numasync, &ncl_iod_mutex, PWAIT, "ioddie", 0); 113 mtx_unlock(&ncl_iod_mutex); 114 ncl_nhuninit(); 115 return (0); 116#else 117 return (EOPNOTSUPP); 118#endif 119} 120 121void 122ncl_dircookie_lock(struct nfsnode *np) 123{ 124 mtx_lock(&np->n_mtx); 125 while (np->n_flag & NDIRCOOKIELK) 126 (void) msleep(&np->n_flag, &np->n_mtx, PZERO, "nfsdirlk", 0); 127 np->n_flag |= NDIRCOOKIELK; 128 mtx_unlock(&np->n_mtx); 129} 130 131void 132ncl_dircookie_unlock(struct nfsnode *np) 133{ 134 mtx_lock(&np->n_mtx); 135 np->n_flag &= ~NDIRCOOKIELK; 136 wakeup(&np->n_flag); 137 mtx_unlock(&np->n_mtx); 138} 139 140int 141ncl_upgrade_vnlock(struct vnode *vp) 142{ 143 int old_lock; 144 145 ASSERT_VOP_LOCKED(vp, "ncl_upgrade_vnlock"); 146 old_lock = NFSVOPISLOCKED(vp); 147 if (old_lock != LK_EXCLUSIVE) { 148 KASSERT(old_lock == LK_SHARED, 149 ("ncl_upgrade_vnlock: wrong old_lock %d", old_lock)); 150 /* Upgrade to exclusive lock, this might block */ 151 NFSVOPLOCK(vp, LK_UPGRADE | LK_RETRY); 152 } 153 return (old_lock); 154} 155 156void 157ncl_downgrade_vnlock(struct vnode *vp, int old_lock) 158{ 159 if (old_lock != LK_EXCLUSIVE) { 160 KASSERT(old_lock == LK_SHARED, ("wrong old_lock %d", old_lock)); 161 /* Downgrade from exclusive lock. */ 162 NFSVOPLOCK(vp, LK_DOWNGRADE | LK_RETRY); 163 } 164} 165 166#ifdef NFS_ACDEBUG 167#include <sys/sysctl.h> 168SYSCTL_DECL(_vfs_nfs); 169static int nfs_acdebug; 170SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, ""); 171#endif 172 173/* 174 * Check the time stamp 175 * If the cache is valid, copy contents to *vap and return 0 176 * otherwise return an error 177 */ 178int 179ncl_getattrcache(struct vnode *vp, struct vattr *vaper) 180{ 181 struct nfsnode *np; 182 struct vattr *vap; 183 struct nfsmount *nmp; 184 int timeo, mustflush; 185 186 np = VTONFS(vp); 187 vap = &np->n_vattr.na_vattr; 188 nmp = VFSTONFS(vp->v_mount); 189 mustflush = nfscl_mustflush(vp); /* must be before mtx_lock() */ 190 mtx_lock(&np->n_mtx); 191 /* XXX n_mtime doesn't seem to be updated on a miss-and-reload */ 192 timeo = (time_second - np->n_mtime.tv_sec) / 10; 193 194#ifdef NFS_ACDEBUG 195 if (nfs_acdebug>1) 196 printf("ncl_getattrcache: initial timeo = %d\n", timeo); 197#endif 198 199 if (vap->va_type == VDIR) { 200 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin) 201 timeo = nmp->nm_acdirmin; 202 else if (timeo > nmp->nm_acdirmax) 203 timeo = nmp->nm_acdirmax; 204 } else { 205 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin) 206 timeo = nmp->nm_acregmin; 207 else if (timeo > nmp->nm_acregmax) 208 timeo = nmp->nm_acregmax; 209 } 210 211#ifdef NFS_ACDEBUG 212 if (nfs_acdebug > 2) 213 printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n", 214 nmp->nm_acregmin, nmp->nm_acregmax, 215 nmp->nm_acdirmin, nmp->nm_acdirmax); 216 217 if (nfs_acdebug) 218 printf("ncl_getattrcache: age = %d; final timeo = %d\n", 219 (time_second - np->n_attrstamp), timeo); 220#endif 221 222 if ((time_second - np->n_attrstamp) >= timeo && 223 (mustflush != 0 || np->n_attrstamp == 0)) { 224 newnfsstats.attrcache_misses++; 225 mtx_unlock(&np->n_mtx); 226 KDTRACE_NFS_ATTRCACHE_GET_MISS(vp); 227 return( ENOENT); 228 } 229 newnfsstats.attrcache_hits++; 230 if (vap->va_size != np->n_size) { 231 if (vap->va_type == VREG) { 232 if (np->n_flag & NMODIFIED) { 233 if (vap->va_size < np->n_size) 234 vap->va_size = np->n_size; 235 else 236 np->n_size = vap->va_size; 237 } else { 238 np->n_size = vap->va_size; 239 } 240 vnode_pager_setsize(vp, np->n_size); 241 } else { 242 np->n_size = vap->va_size; 243 } 244 } 245 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr)); 246 if (np->n_flag & NCHG) { 247 if (np->n_flag & NACC) 248 vaper->va_atime = np->n_atim; 249 if (np->n_flag & NUPD) 250 vaper->va_mtime = np->n_mtim; 251 } 252 mtx_unlock(&np->n_mtx); 253 KDTRACE_NFS_ATTRCACHE_GET_HIT(vp, vap); 254 return (0); 255} 256 257static nfsuint64 nfs_nullcookie = { { 0, 0 } }; 258/* 259 * This function finds the directory cookie that corresponds to the 260 * logical byte offset given. 261 */ 262nfsuint64 * 263ncl_getcookie(struct nfsnode *np, off_t off, int add) 264{ 265 struct nfsdmap *dp, *dp2; 266 int pos; 267 nfsuint64 *retval = NULL; 268 269 pos = (uoff_t)off / NFS_DIRBLKSIZ; 270 if (pos == 0 || off < 0) { 271 KASSERT(!add, ("nfs getcookie add at <= 0")); 272 return (&nfs_nullcookie); 273 } 274 pos--; 275 dp = LIST_FIRST(&np->n_cookies); 276 if (!dp) { 277 if (add) { 278 MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap), 279 M_NFSDIROFF, M_WAITOK); 280 dp->ndm_eocookie = 0; 281 LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list); 282 } else 283 goto out; 284 } 285 while (pos >= NFSNUMCOOKIES) { 286 pos -= NFSNUMCOOKIES; 287 if (LIST_NEXT(dp, ndm_list)) { 288 if (!add && dp->ndm_eocookie < NFSNUMCOOKIES && 289 pos >= dp->ndm_eocookie) 290 goto out; 291 dp = LIST_NEXT(dp, ndm_list); 292 } else if (add) { 293 MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap), 294 M_NFSDIROFF, M_WAITOK); 295 dp2->ndm_eocookie = 0; 296 LIST_INSERT_AFTER(dp, dp2, ndm_list); 297 dp = dp2; 298 } else 299 goto out; 300 } 301 if (pos >= dp->ndm_eocookie) { 302 if (add) 303 dp->ndm_eocookie = pos + 1; 304 else 305 goto out; 306 } 307 retval = &dp->ndm_cookies[pos]; 308out: 309 return (retval); 310} 311 312/* 313 * Invalidate cached directory information, except for the actual directory 314 * blocks (which are invalidated separately). 315 * Done mainly to avoid the use of stale offset cookies. 316 */ 317void 318ncl_invaldir(struct vnode *vp) 319{ 320 struct nfsnode *np = VTONFS(vp); 321 322 KASSERT(vp->v_type == VDIR, ("nfs: invaldir not dir")); 323 ncl_dircookie_lock(np); 324 np->n_direofoffset = 0; 325 np->n_cookieverf.nfsuquad[0] = 0; 326 np->n_cookieverf.nfsuquad[1] = 0; 327 if (LIST_FIRST(&np->n_cookies)) 328 LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0; 329 ncl_dircookie_unlock(np); 330} 331 332/* 333 * The write verifier has changed (probably due to a server reboot), so all 334 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the 335 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT 336 * and B_CLUSTEROK flags. Once done the new write verifier can be set for the 337 * mount point. 338 * 339 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data 340 * writes are not clusterable. 341 */ 342void 343ncl_clearcommit(struct mount *mp) 344{ 345 struct vnode *vp, *nvp; 346 struct buf *bp, *nbp; 347 struct bufobj *bo; 348 349 MNT_VNODE_FOREACH_ALL(vp, mp, nvp) { 350 bo = &vp->v_bufobj; 351 vholdl(vp); 352 VI_UNLOCK(vp); 353 BO_LOCK(bo); 354 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) { 355 if (!BUF_ISLOCKED(bp) && 356 (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT)) 357 == (B_DELWRI | B_NEEDCOMMIT)) 358 bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK); 359 } 360 BO_UNLOCK(bo); 361 vdrop(vp); 362 } 363} 364 365/* 366 * Called once to initialize data structures... 367 */ 368int 369ncl_init(struct vfsconf *vfsp) 370{ 371 int i; 372 373 /* Ensure async daemons disabled */ 374 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) { 375 ncl_iodwant[i] = NFSIOD_NOT_AVAILABLE; 376 ncl_iodmount[i] = NULL; 377 } 378 TASK_INIT(&ncl_nfsiodnew_task, 0, ncl_nfsiodnew_tq, NULL); 379 ncl_nhinit(); /* Init the nfsnode table */ 380 381 return (0); 382} 383 384