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 * 3. 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_node.c 8.6 (Berkeley) 5/22/95 33 */ 34 35 #include <sys/cdefs.h> 36 __FBSDID("$FreeBSD$"); 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/fcntl.h> 41 #include <sys/lock.h> 42 #include <sys/malloc.h> 43 #include <sys/mount.h> 44 #include <sys/namei.h> 45 #include <sys/proc.h> 46 #include <sys/socket.h> 47 #include <sys/sysctl.h> 48 #include <sys/taskqueue.h> 49 #include <sys/vnode.h> 50 51 #include <vm/uma.h> 52 53 #include <fs/nfs/nfsport.h> 54 #include <fs/nfsclient/nfsnode.h> 55 #include <fs/nfsclient/nfsmount.h> 56 #include <fs/nfsclient/nfs.h> 57 #include <fs/nfsclient/nfs_kdtrace.h> 58 59 #include <nfs/nfs_lock.h> 60 61 extern struct vop_vector newnfs_vnodeops; 62 extern struct buf_ops buf_ops_newnfs; 63 MALLOC_DECLARE(M_NEWNFSREQ); 64 65 uma_zone_t newnfsnode_zone; 66 67 const char nfs_vnode_tag[] = "nfs"; 68 69 static void nfs_freesillyrename(void *arg, __unused int pending); 70 71 void 72 ncl_nhinit(void) 73 { 74 75 newnfsnode_zone = uma_zcreate("NCLNODE", sizeof(struct nfsnode), NULL, 76 NULL, NULL, NULL, UMA_ALIGN_PTR, 0); 77 } 78 79 void 80 ncl_nhuninit(void) 81 { 82 uma_zdestroy(newnfsnode_zone); 83 } 84 85 /* 86 * ONLY USED FOR THE ROOT DIRECTORY. nfscl_nget() does the rest. If this 87 * function is going to be used to get Regular Files, code must be added 88 * to fill in the "struct nfsv4node". 89 * Look up a vnode/nfsnode by file handle. 90 * Callers must check for mount points!! 91 * In all cases, a pointer to a 92 * nfsnode structure is returned. 93 */ 94 int 95 ncl_nget(struct mount *mntp, u_int8_t *fhp, int fhsize, struct nfsnode **npp, 96 int lkflags) 97 { 98 struct thread *td = curthread; /* XXX */ 99 struct nfsnode *np; 100 struct vnode *vp; 101 struct vnode *nvp; 102 int error; 103 u_int hash; 104 struct nfsmount *nmp; 105 struct nfsfh *nfhp; 106 107 nmp = VFSTONFS(mntp); 108 *npp = NULL; 109 110 hash = fnv_32_buf(fhp, fhsize, FNV1_32_INIT); 111 112 MALLOC(nfhp, struct nfsfh *, sizeof (struct nfsfh) + fhsize, 113 M_NFSFH, M_WAITOK); 114 bcopy(fhp, &nfhp->nfh_fh[0], fhsize); 115 nfhp->nfh_len = fhsize; 116 error = vfs_hash_get(mntp, hash, lkflags, 117 td, &nvp, newnfs_vncmpf, nfhp); 118 FREE(nfhp, M_NFSFH); 119 if (error) 120 return (error); 121 if (nvp != NULL) { 122 *npp = VTONFS(nvp); 123 return (0); 124 } 125 np = uma_zalloc(newnfsnode_zone, M_WAITOK | M_ZERO); 126 127 error = getnewvnode(nfs_vnode_tag, mntp, &newnfs_vnodeops, &nvp); 128 if (error) { 129 uma_zfree(newnfsnode_zone, np); 130 return (error); 131 } 132 vp = nvp; 133 KASSERT(vp->v_bufobj.bo_bsize != 0, ("ncl_nget: bo_bsize == 0")); 134 vp->v_bufobj.bo_ops = &buf_ops_newnfs; 135 vp->v_data = np; 136 np->n_vnode = vp; 137 /* 138 * Initialize the mutex even if the vnode is going to be a loser. 139 * This simplifies the logic in reclaim, which can then unconditionally 140 * destroy the mutex (in the case of the loser, or if hash_insert 141 * happened to return an error no special casing is needed). 142 */ 143 mtx_init(&np->n_mtx, "NEWNFSnode lock", NULL, MTX_DEF | MTX_DUPOK); 144 lockinit(&np->n_excl, PVFS, "nfsupg", VLKTIMEOUT, LK_NOSHARE | 145 LK_CANRECURSE); 146 147 /* 148 * NFS supports recursive and shared locking. 149 */ 150 lockmgr(vp->v_vnlock, LK_EXCLUSIVE | LK_NOWITNESS, NULL); 151 VN_LOCK_AREC(vp); 152 VN_LOCK_ASHARE(vp); 153 /* 154 * Are we getting the root? If so, make sure the vnode flags 155 * are correct 156 */ 157 if ((fhsize == nmp->nm_fhsize) && 158 !bcmp(fhp, nmp->nm_fh, fhsize)) { 159 if (vp->v_type == VNON) 160 vp->v_type = VDIR; 161 vp->v_vflag |= VV_ROOT; 162 } 163 164 MALLOC(np->n_fhp, struct nfsfh *, sizeof (struct nfsfh) + fhsize, 165 M_NFSFH, M_WAITOK); 166 bcopy(fhp, np->n_fhp->nfh_fh, fhsize); 167 np->n_fhp->nfh_len = fhsize; 168 error = insmntque(vp, mntp); 169 if (error != 0) { 170 *npp = NULL; 171 FREE((caddr_t)np->n_fhp, M_NFSFH); 172 mtx_destroy(&np->n_mtx); 173 lockdestroy(&np->n_excl); 174 uma_zfree(newnfsnode_zone, np); 175 return (error); 176 } 177 error = vfs_hash_insert(vp, hash, lkflags, 178 td, &nvp, newnfs_vncmpf, np->n_fhp); 179 if (error) 180 return (error); 181 if (nvp != NULL) { 182 *npp = VTONFS(nvp); 183 /* vfs_hash_insert() vput()'s the losing vnode */ 184 return (0); 185 } 186 *npp = np; 187 188 return (0); 189 } 190 191 /* 192 * Do the vrele(sp->s_dvp) as a separate task in order to avoid a 193 * deadlock because of a LOR when vrele() locks the directory vnode. 194 */ 195 static void 196 nfs_freesillyrename(void *arg, __unused int pending) 197 { 198 struct sillyrename *sp; 199 200 sp = arg; 201 vrele(sp->s_dvp); 202 free(sp, M_NEWNFSREQ); 203 } 204 205 static void 206 ncl_releasesillyrename(struct vnode *vp, struct thread *td) 207 { 208 struct nfsnode *np; 209 struct sillyrename *sp; 210 211 ASSERT_VOP_ELOCKED(vp, "releasesillyrename"); 212 np = VTONFS(vp); 213 mtx_assert(&np->n_mtx, MA_OWNED); 214 if (vp->v_type != VDIR) { 215 sp = np->n_sillyrename; 216 np->n_sillyrename = NULL; 217 } else 218 sp = NULL; 219 if (sp != NULL) { 220 mtx_unlock(&np->n_mtx); 221 (void) ncl_vinvalbuf(vp, 0, td, 1); 222 /* 223 * Remove the silly file that was rename'd earlier 224 */ 225 ncl_removeit(sp, vp); 226 crfree(sp->s_cred); 227 TASK_INIT(&sp->s_task, 0, nfs_freesillyrename, sp); 228 taskqueue_enqueue(taskqueue_thread, &sp->s_task); 229 mtx_lock(&np->n_mtx); 230 } 231 } 232 233 int 234 ncl_inactive(struct vop_inactive_args *ap) 235 { 236 struct vnode *vp = ap->a_vp; 237 struct nfsnode *np; 238 boolean_t retv; 239 240 if (NFS_ISV4(vp) && vp->v_type == VREG) { 241 /* 242 * Since mmap()'d files do I/O after VOP_CLOSE(), the NFSv4 243 * Close operations are delayed until now. Any dirty 244 * buffers/pages must be flushed before the close, so that the 245 * stateid is available for the writes. 246 */ 247 if (vp->v_object != NULL) { 248 VM_OBJECT_WLOCK(vp->v_object); 249 retv = vm_object_page_clean(vp->v_object, 0, 0, 250 OBJPC_SYNC); 251 VM_OBJECT_WUNLOCK(vp->v_object); 252 } else 253 retv = TRUE; 254 if (retv == TRUE) { 255 (void)ncl_flush(vp, MNT_WAIT, ap->a_td, 1, 0); 256 (void)nfsrpc_close(vp, 1, ap->a_td); 257 } 258 } 259 260 np = VTONFS(vp); 261 mtx_lock(&np->n_mtx); 262 ncl_releasesillyrename(vp, ap->a_td); 263 264 /* 265 * NMODIFIED means that there might be dirty/stale buffers 266 * associated with the NFS vnode. 267 * NDSCOMMIT means that the file is on a pNFS server and commits 268 * should be done to the DS. 269 * None of the other flags are meaningful after the vnode is unused. 270 */ 271 np->n_flag &= (NMODIFIED | NDSCOMMIT); 272 mtx_unlock(&np->n_mtx); 273 return (0); 274 } 275 276 /* 277 * Reclaim an nfsnode so that it can be used for other purposes. 278 */ 279 int 280 ncl_reclaim(struct vop_reclaim_args *ap) 281 { 282 struct vnode *vp = ap->a_vp; 283 struct nfsnode *np = VTONFS(vp); 284 struct nfsdmap *dp, *dp2; 285 286 /* 287 * If the NLM is running, give it a chance to abort pending 288 * locks. 289 */ 290 if (nfs_reclaim_p != NULL) 291 nfs_reclaim_p(ap); 292 293 mtx_lock(&np->n_mtx); 294 ncl_releasesillyrename(vp, ap->a_td); 295 mtx_unlock(&np->n_mtx); 296 297 /* 298 * Destroy the vm object and flush associated pages. 299 */ 300 vnode_destroy_vobject(vp); 301 302 if (NFS_ISV4(vp) && vp->v_type == VREG) 303 /* 304 * We can now safely close any remaining NFSv4 Opens for 305 * this file. Most opens will have already been closed by 306 * ncl_inactive(), but there are cases where it is not 307 * called, so we need to do it again here. 308 */ 309 (void) nfsrpc_close(vp, 1, ap->a_td); 310 311 vfs_hash_remove(vp); 312 313 /* 314 * Call nfscl_reclaimnode() to save attributes in the delegation, 315 * as required. 316 */ 317 if (vp->v_type == VREG) 318 nfscl_reclaimnode(vp); 319 320 /* 321 * Free up any directory cookie structures and 322 * large file handle structures that might be associated with 323 * this nfs node. 324 */ 325 if (vp->v_type == VDIR) { 326 dp = LIST_FIRST(&np->n_cookies); 327 while (dp) { 328 dp2 = dp; 329 dp = LIST_NEXT(dp, ndm_list); 330 FREE((caddr_t)dp2, M_NFSDIROFF); 331 } 332 } 333 if (np->n_writecred != NULL) 334 crfree(np->n_writecred); 335 FREE((caddr_t)np->n_fhp, M_NFSFH); 336 if (np->n_v4 != NULL) 337 FREE((caddr_t)np->n_v4, M_NFSV4NODE); 338 mtx_destroy(&np->n_mtx); 339 lockdestroy(&np->n_excl); 340 uma_zfree(newnfsnode_zone, vp->v_data); 341 vp->v_data = NULL; 342 return (0); 343 } 344 345 /* 346 * Invalidate both the access and attribute caches for this vnode. 347 */ 348 void 349 ncl_invalcaches(struct vnode *vp) 350 { 351 struct nfsnode *np = VTONFS(vp); 352 int i; 353 354 mtx_lock(&np->n_mtx); 355 for (i = 0; i < NFS_ACCESSCACHESIZE; i++) 356 np->n_accesscache[i].stamp = 0; 357 KDTRACE_NFS_ACCESSCACHE_FLUSH_DONE(vp); 358 np->n_attrstamp = 0; 359 KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp); 360 mtx_unlock(&np->n_mtx); 361 } 362