1191783Srmacklem/*- 2191783Srmacklem * Copyright (c) 1989, 1993 3191783Srmacklem * The Regents of the University of California. All rights reserved. 4191783Srmacklem * 5191783Srmacklem * This code is derived from software contributed to Berkeley by 6191783Srmacklem * Rick Macklem at The University of Guelph. 7191783Srmacklem * 8191783Srmacklem * Redistribution and use in source and binary forms, with or without 9191783Srmacklem * modification, are permitted provided that the following conditions 10191783Srmacklem * are met: 11191783Srmacklem * 1. Redistributions of source code must retain the above copyright 12191783Srmacklem * notice, this list of conditions and the following disclaimer. 13191783Srmacklem * 2. Redistributions in binary form must reproduce the above copyright 14191783Srmacklem * notice, this list of conditions and the following disclaimer in the 15191783Srmacklem * documentation and/or other materials provided with the distribution. 16191783Srmacklem * 4. Neither the name of the University nor the names of its contributors 17191783Srmacklem * may be used to endorse or promote products derived from this software 18191783Srmacklem * without specific prior written permission. 19191783Srmacklem * 20191783Srmacklem * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21191783Srmacklem * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22191783Srmacklem * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23191783Srmacklem * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24191783Srmacklem * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25191783Srmacklem * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26191783Srmacklem * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27191783Srmacklem * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28191783Srmacklem * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29191783Srmacklem * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30191783Srmacklem * SUCH DAMAGE. 31191783Srmacklem * 32191783Srmacklem * from nfs_subs.c 8.8 (Berkeley) 5/22/95 33191783Srmacklem */ 34191783Srmacklem 35191783Srmacklem#include <sys/cdefs.h> 36191783Srmacklem__FBSDID("$FreeBSD: releng/10.2/sys/fs/nfsclient/nfs_clsubs.c 234386 2012-04-17 16:28:22Z mckusick $"); 37191783Srmacklem 38223280Srmacklem#include "opt_kdtrace.h" 39223280Srmacklem 40191783Srmacklem/* 41191783Srmacklem * These functions support the macros and help fiddle mbuf chains for 42191783Srmacklem * the nfs op functions. They do things like create the rpc header and 43191783Srmacklem * copy data between mbuf chains and uio lists. 44191783Srmacklem */ 45191783Srmacklem 46191783Srmacklem#include <sys/param.h> 47191783Srmacklem#include <sys/systm.h> 48191783Srmacklem#include <sys/kernel.h> 49191783Srmacklem#include <sys/bio.h> 50191783Srmacklem#include <sys/buf.h> 51191783Srmacklem#include <sys/proc.h> 52191783Srmacklem#include <sys/mount.h> 53191783Srmacklem#include <sys/vnode.h> 54191783Srmacklem#include <sys/namei.h> 55191783Srmacklem#include <sys/mbuf.h> 56191783Srmacklem#include <sys/socket.h> 57191783Srmacklem#include <sys/stat.h> 58191783Srmacklem#include <sys/malloc.h> 59191783Srmacklem#include <sys/sysent.h> 60191783Srmacklem#include <sys/syscall.h> 61191783Srmacklem#include <sys/sysproto.h> 62220683Srmacklem#include <sys/taskqueue.h> 63191783Srmacklem 64191783Srmacklem#include <vm/vm.h> 65191783Srmacklem#include <vm/vm_object.h> 66191783Srmacklem#include <vm/vm_extern.h> 67191783Srmacklem#include <vm/uma.h> 68191783Srmacklem 69191783Srmacklem#include <fs/nfs/nfsport.h> 70191783Srmacklem#include <fs/nfsclient/nfsnode.h> 71191783Srmacklem#include <fs/nfsclient/nfsmount.h> 72191783Srmacklem#include <fs/nfsclient/nfs.h> 73223280Srmacklem#include <fs/nfsclient/nfs_kdtrace.h> 74191783Srmacklem 75191783Srmacklem#include <netinet/in.h> 76191783Srmacklem 77191783Srmacklem/* 78191783Srmacklem * Note that stdarg.h and the ANSI style va_start macro is used for both 79191783Srmacklem * ANSI and traditional C compilers. 80191783Srmacklem */ 81191783Srmacklem#include <machine/stdarg.h> 82191783Srmacklem 83191783Srmacklemextern struct mtx ncl_iod_mutex; 84220683Srmacklemextern enum nfsiod_state ncl_iodwant[NFS_MAXASYNCDAEMON]; 85220683Srmacklemextern struct nfsmount *ncl_iodmount[NFS_MAXASYNCDAEMON]; 86191783Srmacklemextern int ncl_numasync; 87191783Srmacklemextern unsigned int ncl_iodmax; 88191783Srmacklemextern struct nfsstats newnfsstats; 89191783Srmacklem 90220683Srmacklemstruct task ncl_nfsiodnew_task; 91220683Srmacklem 92191783Srmacklemint 93191783Srmacklemncl_uninit(struct vfsconf *vfsp) 94191783Srmacklem{ 95198291Sjh /* 96198291Sjh * XXX: Unloading of nfscl module is unsupported. 97198291Sjh */ 98198291Sjh#if 0 99191783Srmacklem int i; 100191783Srmacklem 101191783Srmacklem /* 102191783Srmacklem * Tell all nfsiod processes to exit. Clear ncl_iodmax, and wakeup 103191783Srmacklem * any sleeping nfsiods so they check ncl_iodmax and exit. 104191783Srmacklem */ 105191783Srmacklem mtx_lock(&ncl_iod_mutex); 106191783Srmacklem ncl_iodmax = 0; 107191783Srmacklem for (i = 0; i < ncl_numasync; i++) 108203119Srmacklem if (ncl_iodwant[i] == NFSIOD_AVAILABLE) 109191783Srmacklem wakeup(&ncl_iodwant[i]); 110191783Srmacklem /* The last nfsiod to exit will wake us up when ncl_numasync hits 0 */ 111191783Srmacklem while (ncl_numasync) 112191783Srmacklem msleep(&ncl_numasync, &ncl_iod_mutex, PWAIT, "ioddie", 0); 113191783Srmacklem mtx_unlock(&ncl_iod_mutex); 114191783Srmacklem ncl_nhuninit(); 115191783Srmacklem return (0); 116198291Sjh#else 117198291Sjh return (EOPNOTSUPP); 118198291Sjh#endif 119191783Srmacklem} 120191783Srmacklem 121191783Srmacklemvoid 122191783Srmacklemncl_dircookie_lock(struct nfsnode *np) 123191783Srmacklem{ 124191783Srmacklem mtx_lock(&np->n_mtx); 125191783Srmacklem while (np->n_flag & NDIRCOOKIELK) 126191783Srmacklem (void) msleep(&np->n_flag, &np->n_mtx, PZERO, "nfsdirlk", 0); 127191783Srmacklem np->n_flag |= NDIRCOOKIELK; 128191783Srmacklem mtx_unlock(&np->n_mtx); 129191783Srmacklem} 130191783Srmacklem 131191783Srmacklemvoid 132191783Srmacklemncl_dircookie_unlock(struct nfsnode *np) 133191783Srmacklem{ 134191783Srmacklem mtx_lock(&np->n_mtx); 135191783Srmacklem np->n_flag &= ~NDIRCOOKIELK; 136191783Srmacklem wakeup(&np->n_flag); 137191783Srmacklem mtx_unlock(&np->n_mtx); 138191783Srmacklem} 139191783Srmacklem 140191783Srmacklemint 141191783Srmacklemncl_upgrade_vnlock(struct vnode *vp) 142191783Srmacklem{ 143191783Srmacklem int old_lock; 144196332Srmacklem 145196332Srmacklem ASSERT_VOP_LOCKED(vp, "ncl_upgrade_vnlock"); 146224083Szack old_lock = NFSVOPISLOCKED(vp); 147196332Srmacklem if (old_lock != LK_EXCLUSIVE) { 148196332Srmacklem KASSERT(old_lock == LK_SHARED, 149196332Srmacklem ("ncl_upgrade_vnlock: wrong old_lock %d", old_lock)); 150196332Srmacklem /* Upgrade to exclusive lock, this might block */ 151224081Szack NFSVOPLOCK(vp, LK_UPGRADE | LK_RETRY); 152191783Srmacklem } 153196332Srmacklem return (old_lock); 154191783Srmacklem} 155191783Srmacklem 156191783Srmacklemvoid 157191783Srmacklemncl_downgrade_vnlock(struct vnode *vp, int old_lock) 158191783Srmacklem{ 159191783Srmacklem if (old_lock != LK_EXCLUSIVE) { 160196332Srmacklem KASSERT(old_lock == LK_SHARED, ("wrong old_lock %d", old_lock)); 161196332Srmacklem /* Downgrade from exclusive lock. */ 162224081Szack NFSVOPLOCK(vp, LK_DOWNGRADE | LK_RETRY); 163191783Srmacklem } 164191783Srmacklem} 165191783Srmacklem 166191783Srmacklemvoid 167191783Srmacklemncl_printf(const char *fmt, ...) 168191783Srmacklem{ 169191783Srmacklem va_list ap; 170191783Srmacklem 171191783Srmacklem mtx_lock(&Giant); 172191783Srmacklem va_start(ap, fmt); 173221429Sru vprintf(fmt, ap); 174191783Srmacklem va_end(ap); 175191783Srmacklem mtx_unlock(&Giant); 176191783Srmacklem} 177191783Srmacklem 178191783Srmacklem#ifdef NFS_ACDEBUG 179191783Srmacklem#include <sys/sysctl.h> 180221973SrmacklemSYSCTL_DECL(_vfs_nfs); 181191783Srmacklemstatic int nfs_acdebug; 182221973SrmacklemSYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, ""); 183191783Srmacklem#endif 184191783Srmacklem 185191783Srmacklem/* 186191783Srmacklem * Check the time stamp 187191783Srmacklem * If the cache is valid, copy contents to *vap and return 0 188191783Srmacklem * otherwise return an error 189191783Srmacklem */ 190191783Srmacklemint 191191783Srmacklemncl_getattrcache(struct vnode *vp, struct vattr *vaper) 192191783Srmacklem{ 193191783Srmacklem struct nfsnode *np; 194191783Srmacklem struct vattr *vap; 195191783Srmacklem struct nfsmount *nmp; 196210786Srmacklem int timeo, mustflush; 197191783Srmacklem 198191783Srmacklem np = VTONFS(vp); 199191783Srmacklem vap = &np->n_vattr.na_vattr; 200191783Srmacklem nmp = VFSTONFS(vp->v_mount); 201208254Srmacklem mustflush = nfscl_mustflush(vp); /* must be before mtx_lock() */ 202191783Srmacklem#ifdef NFS_ACDEBUG 203191783Srmacklem mtx_lock(&Giant); /* ncl_printf() */ 204191783Srmacklem#endif 205191783Srmacklem mtx_lock(&np->n_mtx); 206191783Srmacklem /* XXX n_mtime doesn't seem to be updated on a miss-and-reload */ 207191783Srmacklem timeo = (time_second - np->n_mtime.tv_sec) / 10; 208191783Srmacklem 209191783Srmacklem#ifdef NFS_ACDEBUG 210191783Srmacklem if (nfs_acdebug>1) 211191783Srmacklem ncl_printf("nfs_getattrcache: initial timeo = %d\n", timeo); 212191783Srmacklem#endif 213191783Srmacklem 214191783Srmacklem if (vap->va_type == VDIR) { 215191783Srmacklem if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin) 216191783Srmacklem timeo = nmp->nm_acdirmin; 217191783Srmacklem else if (timeo > nmp->nm_acdirmax) 218191783Srmacklem timeo = nmp->nm_acdirmax; 219191783Srmacklem } else { 220191783Srmacklem if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin) 221191783Srmacklem timeo = nmp->nm_acregmin; 222191783Srmacklem else if (timeo > nmp->nm_acregmax) 223191783Srmacklem timeo = nmp->nm_acregmax; 224191783Srmacklem } 225191783Srmacklem 226191783Srmacklem#ifdef NFS_ACDEBUG 227191783Srmacklem if (nfs_acdebug > 2) 228191783Srmacklem ncl_printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n", 229191783Srmacklem nmp->nm_acregmin, nmp->nm_acregmax, 230191783Srmacklem nmp->nm_acdirmin, nmp->nm_acdirmax); 231191783Srmacklem 232191783Srmacklem if (nfs_acdebug) 233191783Srmacklem ncl_printf("nfs_getattrcache: age = %d; final timeo = %d\n", 234191783Srmacklem (time_second - np->n_attrstamp), timeo); 235191783Srmacklem#endif 236191783Srmacklem 237208254Srmacklem if ((time_second - np->n_attrstamp) >= timeo && 238210786Srmacklem (mustflush != 0 || np->n_attrstamp == 0)) { 239191783Srmacklem newnfsstats.attrcache_misses++; 240191783Srmacklem mtx_unlock(&np->n_mtx); 241208254Srmacklem#ifdef NFS_ACDEBUG 242208254Srmacklem mtx_unlock(&Giant); /* ncl_printf() */ 243208254Srmacklem#endif 244223280Srmacklem KDTRACE_NFS_ATTRCACHE_GET_MISS(vp); 245191783Srmacklem return( ENOENT); 246191783Srmacklem } 247191783Srmacklem newnfsstats.attrcache_hits++; 248191783Srmacklem if (vap->va_size != np->n_size) { 249191783Srmacklem if (vap->va_type == VREG) { 250191783Srmacklem if (np->n_flag & NMODIFIED) { 251191783Srmacklem if (vap->va_size < np->n_size) 252191783Srmacklem vap->va_size = np->n_size; 253191783Srmacklem else 254191783Srmacklem np->n_size = vap->va_size; 255191783Srmacklem } else { 256191783Srmacklem np->n_size = vap->va_size; 257191783Srmacklem } 258191783Srmacklem vnode_pager_setsize(vp, np->n_size); 259191783Srmacklem } else { 260191783Srmacklem np->n_size = vap->va_size; 261191783Srmacklem } 262191783Srmacklem } 263191783Srmacklem bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr)); 264191783Srmacklem if (np->n_flag & NCHG) { 265191783Srmacklem if (np->n_flag & NACC) 266191783Srmacklem vaper->va_atime = np->n_atim; 267191783Srmacklem if (np->n_flag & NUPD) 268191783Srmacklem vaper->va_mtime = np->n_mtim; 269191783Srmacklem } 270191783Srmacklem mtx_unlock(&np->n_mtx); 271191783Srmacklem#ifdef NFS_ACDEBUG 272191783Srmacklem mtx_unlock(&Giant); /* ncl_printf() */ 273191783Srmacklem#endif 274223280Srmacklem KDTRACE_NFS_ATTRCACHE_GET_HIT(vp, vap); 275191783Srmacklem return (0); 276191783Srmacklem} 277191783Srmacklem 278191783Srmacklemstatic nfsuint64 nfs_nullcookie = { { 0, 0 } }; 279191783Srmacklem/* 280191783Srmacklem * This function finds the directory cookie that corresponds to the 281191783Srmacklem * logical byte offset given. 282191783Srmacklem */ 283191783Srmacklemnfsuint64 * 284191783Srmacklemncl_getcookie(struct nfsnode *np, off_t off, int add) 285191783Srmacklem{ 286191783Srmacklem struct nfsdmap *dp, *dp2; 287191783Srmacklem int pos; 288191783Srmacklem nfsuint64 *retval = NULL; 289191783Srmacklem 290191783Srmacklem pos = (uoff_t)off / NFS_DIRBLKSIZ; 291191783Srmacklem if (pos == 0 || off < 0) { 292209120Skib KASSERT(!add, ("nfs getcookie add at <= 0")); 293191783Srmacklem return (&nfs_nullcookie); 294191783Srmacklem } 295191783Srmacklem pos--; 296191783Srmacklem dp = LIST_FIRST(&np->n_cookies); 297191783Srmacklem if (!dp) { 298191783Srmacklem if (add) { 299191783Srmacklem MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap), 300191783Srmacklem M_NFSDIROFF, M_WAITOK); 301191783Srmacklem dp->ndm_eocookie = 0; 302191783Srmacklem LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list); 303191783Srmacklem } else 304191783Srmacklem goto out; 305191783Srmacklem } 306191783Srmacklem while (pos >= NFSNUMCOOKIES) { 307191783Srmacklem pos -= NFSNUMCOOKIES; 308191783Srmacklem if (LIST_NEXT(dp, ndm_list)) { 309191783Srmacklem if (!add && dp->ndm_eocookie < NFSNUMCOOKIES && 310191783Srmacklem pos >= dp->ndm_eocookie) 311191783Srmacklem goto out; 312191783Srmacklem dp = LIST_NEXT(dp, ndm_list); 313191783Srmacklem } else if (add) { 314191783Srmacklem MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap), 315191783Srmacklem M_NFSDIROFF, M_WAITOK); 316191783Srmacklem dp2->ndm_eocookie = 0; 317191783Srmacklem LIST_INSERT_AFTER(dp, dp2, ndm_list); 318191783Srmacklem dp = dp2; 319191783Srmacklem } else 320191783Srmacklem goto out; 321191783Srmacklem } 322191783Srmacklem if (pos >= dp->ndm_eocookie) { 323191783Srmacklem if (add) 324191783Srmacklem dp->ndm_eocookie = pos + 1; 325191783Srmacklem else 326191783Srmacklem goto out; 327191783Srmacklem } 328191783Srmacklem retval = &dp->ndm_cookies[pos]; 329191783Srmacklemout: 330191783Srmacklem return (retval); 331191783Srmacklem} 332191783Srmacklem 333191783Srmacklem/* 334191783Srmacklem * Invalidate cached directory information, except for the actual directory 335191783Srmacklem * blocks (which are invalidated separately). 336191783Srmacklem * Done mainly to avoid the use of stale offset cookies. 337191783Srmacklem */ 338191783Srmacklemvoid 339191783Srmacklemncl_invaldir(struct vnode *vp) 340191783Srmacklem{ 341191783Srmacklem struct nfsnode *np = VTONFS(vp); 342191783Srmacklem 343209120Skib KASSERT(vp->v_type == VDIR, ("nfs: invaldir not dir")); 344191783Srmacklem ncl_dircookie_lock(np); 345191783Srmacklem np->n_direofoffset = 0; 346191783Srmacklem np->n_cookieverf.nfsuquad[0] = 0; 347191783Srmacklem np->n_cookieverf.nfsuquad[1] = 0; 348191783Srmacklem if (LIST_FIRST(&np->n_cookies)) 349191783Srmacklem LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0; 350191783Srmacklem ncl_dircookie_unlock(np); 351191783Srmacklem} 352191783Srmacklem 353191783Srmacklem/* 354191783Srmacklem * The write verifier has changed (probably due to a server reboot), so all 355191783Srmacklem * B_NEEDCOMMIT blocks will have to be written again. Since they are on the 356191783Srmacklem * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT 357191783Srmacklem * and B_CLUSTEROK flags. Once done the new write verifier can be set for the 358191783Srmacklem * mount point. 359191783Srmacklem * 360191783Srmacklem * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data 361191783Srmacklem * writes are not clusterable. 362191783Srmacklem */ 363191783Srmacklemvoid 364191783Srmacklemncl_clearcommit(struct mount *mp) 365191783Srmacklem{ 366191783Srmacklem struct vnode *vp, *nvp; 367191783Srmacklem struct buf *bp, *nbp; 368191783Srmacklem struct bufobj *bo; 369191783Srmacklem 370234386Smckusick MNT_VNODE_FOREACH_ALL(vp, mp, nvp) { 371191783Srmacklem bo = &vp->v_bufobj; 372191783Srmacklem vholdl(vp); 373191783Srmacklem VI_UNLOCK(vp); 374191783Srmacklem BO_LOCK(bo); 375191783Srmacklem TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) { 376191783Srmacklem if (!BUF_ISLOCKED(bp) && 377191783Srmacklem (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT)) 378191783Srmacklem == (B_DELWRI | B_NEEDCOMMIT)) 379191783Srmacklem bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK); 380191783Srmacklem } 381191783Srmacklem BO_UNLOCK(bo); 382191783Srmacklem vdrop(vp); 383191783Srmacklem } 384191783Srmacklem} 385191783Srmacklem 386191783Srmacklem/* 387191783Srmacklem * Called once to initialize data structures... 388191783Srmacklem */ 389191783Srmacklemint 390191783Srmacklemncl_init(struct vfsconf *vfsp) 391191783Srmacklem{ 392191783Srmacklem int i; 393191783Srmacklem 394191783Srmacklem /* Ensure async daemons disabled */ 395220683Srmacklem for (i = 0; i < NFS_MAXASYNCDAEMON; i++) { 396203119Srmacklem ncl_iodwant[i] = NFSIOD_NOT_AVAILABLE; 397191783Srmacklem ncl_iodmount[i] = NULL; 398191783Srmacklem } 399220683Srmacklem TASK_INIT(&ncl_nfsiodnew_task, 0, ncl_nfsiodnew_tq, NULL); 400191783Srmacklem ncl_nhinit(); /* Init the nfsnode table */ 401191783Srmacklem 402191783Srmacklem return (0); 403191783Srmacklem} 404191783Srmacklem 405