1 /* $NetBSD: nfs_node.c,v 1.107 2008/11/19 18:36:09 ad Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * Rick Macklem at The University of Guelph. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)nfs_node.c 8.6 (Berkeley) 5/22/95 35 */ 36 37 #include <sys/cdefs.h> 38 __KERNEL_RCSID(0, "$NetBSD: nfs_node.c,v 1.107 2008/11/19 18:36:09 ad Exp $"); 39 40 #ifdef _KERNEL_OPT 41 #include "opt_nfs.h" 42 #endif 43 44 #include <sys/param.h> 45 #include <sys/systm.h> 46 #include <sys/proc.h> 47 #include <sys/mount.h> 48 #include <sys/namei.h> 49 #include <sys/vnode.h> 50 #include <sys/kernel.h> 51 #include <sys/pool.h> 52 #include <sys/lock.h> 53 #include <sys/hash.h> 54 #include <sys/kauth.h> 55 56 #include <nfs/rpcv2.h> 57 #include <nfs/nfsproto.h> 58 #include <nfs/nfs.h> 59 #include <nfs/nfsnode.h> 60 #include <nfs/nfsmount.h> 61 #include <nfs/nfs_var.h> 62 63 struct pool nfs_node_pool; 64 struct pool nfs_vattr_pool; 65 66 MALLOC_JUSTDEFINE(M_NFSNODE, "NFS node", "NFS vnode private part"); 67 68 extern int prtactive; 69 70 void nfs_gop_size(struct vnode *, off_t, off_t *, int); 71 int nfs_gop_alloc(struct vnode *, off_t, off_t, int, kauth_cred_t); 72 int nfs_gop_write(struct vnode *, struct vm_page **, int, int); 73 74 static const struct genfs_ops nfs_genfsops = { 75 .gop_size = nfs_gop_size, 76 .gop_alloc = nfs_gop_alloc, 77 .gop_write = nfs_gop_write, 78 }; 79 80 /* 81 * Reinitialize inode hash table. 82 */ 83 void 84 nfs_node_init() 85 { 86 malloc_type_attach(M_NFSNODE); 87 pool_init(&nfs_node_pool, sizeof(struct nfsnode), 0, 0, 0, "nfsnodepl", 88 &pool_allocator_nointr, IPL_NONE); 89 pool_init(&nfs_vattr_pool, sizeof(struct vattr), 0, 0, 0, "nfsvapl", 90 &pool_allocator_nointr, IPL_NONE); 91 } 92 93 /* 94 * Free resources previously allocated in nfs_node_reinit(). 95 */ 96 void 97 nfs_node_done() 98 { 99 pool_destroy(&nfs_node_pool); 100 pool_destroy(&nfs_vattr_pool); 101 malloc_type_detach(M_NFSNODE); 102 } 103 104 #define RBTONFSNODE(node) \ 105 (void *)((uintptr_t)(node) - offsetof(struct nfsnode, n_rbnode)) 106 107 struct fh_match { 108 nfsfh_t *fhm_fhp; 109 size_t fhm_fhsize; 110 size_t fhm_fhoffset; 111 }; 112 113 static int 114 nfs_compare_nodes(const struct rb_node *parent, const struct rb_node *node) 115 { 116 const struct nfsnode * const pnp = RBTONFSNODE(parent); 117 const struct nfsnode * const np = RBTONFSNODE(node); 118 119 if (pnp->n_fhsize != np->n_fhsize) 120 return np->n_fhsize - pnp->n_fhsize; 121 122 return memcmp(np->n_fhp, pnp->n_fhp, np->n_fhsize); 123 } 124 125 static int 126 nfs_compare_node_fh(const struct rb_node *b, const void *key) 127 { 128 const struct nfsnode * const pnp = RBTONFSNODE(b); 129 const struct fh_match * const fhm = key; 130 131 if (pnp->n_fhsize != fhm->fhm_fhsize) 132 return fhm->fhm_fhsize - pnp->n_fhsize; 133 134 return memcmp(fhm->fhm_fhp, pnp->n_fhp, pnp->n_fhsize); 135 } 136 137 static const struct rb_tree_ops nfs_node_rbtree_ops = { 138 .rbto_compare_nodes = nfs_compare_nodes, 139 .rbto_compare_key = nfs_compare_node_fh, 140 }; 141 142 void 143 nfs_rbtinit(struct nfsmount *nmp) 144 { 145 rb_tree_init(&nmp->nm_rbtree, &nfs_node_rbtree_ops); 146 } 147 148 149 /* 150 * Look up a vnode/nfsnode by file handle. 151 * Callers must check for mount points!! 152 * In all cases, a pointer to a 153 * nfsnode structure is returned. 154 */ 155 int 156 nfs_nget1(mntp, fhp, fhsize, npp, lkflags) 157 struct mount *mntp; 158 nfsfh_t *fhp; 159 int fhsize; 160 struct nfsnode **npp; 161 int lkflags; 162 { 163 struct nfsnode *np; 164 struct vnode *vp; 165 struct nfsmount *nmp = VFSTONFS(mntp); 166 int error; 167 struct fh_match fhm; 168 struct rb_node *node; 169 170 fhm.fhm_fhp = fhp; 171 fhm.fhm_fhsize = fhsize; 172 173 loop: 174 rw_enter(&nmp->nm_rbtlock, RW_READER); 175 node = rb_tree_find_node(&nmp->nm_rbtree, &fhm); 176 if (node != NULL) { 177 np = RBTONFSNODE(node); 178 vp = NFSTOV(np); 179 mutex_enter(&vp->v_interlock); 180 rw_exit(&nmp->nm_rbtlock); 181 error = vget(vp, LK_EXCLUSIVE | LK_INTERLOCK | lkflags); 182 if (error == EBUSY) 183 return error; 184 if (error) 185 goto loop; 186 *npp = np; 187 return(0); 188 } 189 rw_exit(&nmp->nm_rbtlock); 190 191 error = getnewvnode(VT_NFS, mntp, nfsv2_vnodeop_p, &vp); 192 if (error) { 193 *npp = 0; 194 return (error); 195 } 196 np = pool_get(&nfs_node_pool, PR_WAITOK); 197 memset(np, 0, sizeof *np); 198 np->n_vnode = vp; 199 200 /* 201 * Insert the nfsnode in the hash queue for its new file handle 202 */ 203 204 if (fhsize > NFS_SMALLFH) { 205 np->n_fhp = kmem_alloc(fhsize, KM_SLEEP); 206 } else 207 np->n_fhp = &np->n_fh; 208 memcpy(np->n_fhp, fhp, fhsize); 209 np->n_fhsize = fhsize; 210 np->n_accstamp = -1; 211 np->n_vattr = pool_get(&nfs_vattr_pool, PR_WAITOK); 212 213 rw_enter(&nmp->nm_rbtlock, RW_WRITER); 214 if (NULL != rb_tree_find_node(&nmp->nm_rbtree, &fhm)) { 215 rw_exit(&nmp->nm_rbtlock); 216 if (fhsize > NFS_SMALLFH) { 217 kmem_free(np->n_fhp, fhsize); 218 } 219 pool_put(&nfs_vattr_pool, np->n_vattr); 220 pool_put(&nfs_node_pool, np); 221 ungetnewvnode(vp); 222 goto loop; 223 } 224 vp->v_data = np; 225 genfs_node_init(vp, &nfs_genfsops); 226 /* 227 * Initalize read/write creds to useful values. VOP_OPEN will 228 * overwrite these. 229 */ 230 np->n_rcred = curlwp->l_cred; 231 kauth_cred_hold(np->n_rcred); 232 np->n_wcred = curlwp->l_cred; 233 kauth_cred_hold(np->n_wcred); 234 vlockmgr(&vp->v_lock, LK_EXCLUSIVE); 235 NFS_INVALIDATE_ATTRCACHE(np); 236 uvm_vnp_setsize(vp, 0); 237 rb_tree_insert_node(&nmp->nm_rbtree, &np->n_rbnode); 238 rw_exit(&nmp->nm_rbtlock); 239 240 *npp = np; 241 return (0); 242 } 243 244 int 245 nfs_inactive(v) 246 void *v; 247 { 248 struct vop_inactive_args /* { 249 struct vnode *a_vp; 250 bool *a_recycle; 251 } */ *ap = v; 252 struct nfsnode *np; 253 struct sillyrename *sp; 254 struct vnode *vp = ap->a_vp; 255 256 np = VTONFS(vp); 257 if (vp->v_type != VDIR) { 258 sp = np->n_sillyrename; 259 np->n_sillyrename = (struct sillyrename *)0; 260 } else 261 sp = NULL; 262 if (sp != NULL) 263 nfs_vinvalbuf(vp, 0, sp->s_cred, curlwp, 1); 264 *ap->a_recycle = (np->n_flag & NREMOVED) != 0; 265 np->n_flag &= 266 (NMODIFIED | NFLUSHINPROG | NFLUSHWANT | NEOFVALID | NTRUNCDELAYED); 267 268 if (vp->v_type == VDIR && np->n_dircache) 269 nfs_invaldircache(vp, 270 NFS_INVALDIRCACHE_FORCE | NFS_INVALDIRCACHE_KEEPEOF); 271 272 VOP_UNLOCK(vp, 0); 273 274 if (sp != NULL) { 275 int error; 276 277 /* 278 * Remove the silly file that was rename'd earlier 279 * 280 * Just in case our thread also has the parent node locked, 281 * we use LK_CANRECURSE. 282 */ 283 284 error = vn_lock(sp->s_dvp, LK_EXCLUSIVE | LK_CANRECURSE); 285 if (error || sp->s_dvp->v_data == NULL) { 286 /* XXX should recover */ 287 printf("%s: vp=%p error=%d\n", 288 __func__, sp->s_dvp, error); 289 } else { 290 nfs_removeit(sp); 291 } 292 kauth_cred_free(sp->s_cred); 293 vput(sp->s_dvp); 294 kmem_free(sp, sizeof(*sp)); 295 } 296 297 return (0); 298 } 299 300 /* 301 * Reclaim an nfsnode so that it can be used for other purposes. 302 */ 303 int 304 nfs_reclaim(v) 305 void *v; 306 { 307 struct vop_reclaim_args /* { 308 struct vnode *a_vp; 309 } */ *ap = v; 310 struct vnode *vp = ap->a_vp; 311 struct nfsnode *np = VTONFS(vp); 312 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 313 314 if (prtactive && vp->v_usecount > 1) 315 vprint("nfs_reclaim: pushing active", vp); 316 317 rw_enter(&nmp->nm_rbtlock, RW_WRITER); 318 rb_tree_remove_node(&nmp->nm_rbtree, &np->n_rbnode); 319 rw_exit(&nmp->nm_rbtlock); 320 321 /* 322 * Free up any directory cookie structures and 323 * large file handle structures that might be associated with 324 * this nfs node. 325 */ 326 if (vp->v_type == VDIR && np->n_dircache != NULL) { 327 nfs_invaldircache(vp, NFS_INVALDIRCACHE_FORCE); 328 hashdone(np->n_dircache, HASH_LIST, nfsdirhashmask); 329 } 330 KASSERT(np->n_dirgens == NULL); 331 332 if (np->n_fhsize > NFS_SMALLFH) 333 kmem_free(np->n_fhp, np->n_fhsize); 334 335 pool_put(&nfs_vattr_pool, np->n_vattr); 336 if (np->n_rcred) 337 kauth_cred_free(np->n_rcred); 338 339 if (np->n_wcred) 340 kauth_cred_free(np->n_wcred); 341 342 cache_purge(vp); 343 if (vp->v_type == VREG) { 344 mutex_destroy(&np->n_commitlock); 345 } 346 genfs_node_destroy(vp); 347 pool_put(&nfs_node_pool, np); 348 vp->v_data = NULL; 349 return (0); 350 } 351 352 void 353 nfs_gop_size(struct vnode *vp, off_t size, off_t *eobp, int flags) 354 { 355 356 *eobp = MAX(size, vp->v_size); 357 } 358 359 int 360 nfs_gop_alloc(struct vnode *vp, off_t off, off_t len, int flags, 361 kauth_cred_t cred) 362 { 363 364 return 0; 365 } 366 367 int 368 nfs_gop_write(struct vnode *vp, struct vm_page **pgs, int npages, int flags) 369 { 370 int i; 371 372 for (i = 0; i < npages; i++) { 373 pmap_page_protect(pgs[i], VM_PROT_READ); 374 } 375 return genfs_gop_write(vp, pgs, npages, flags); 376 } 377