1 /*- 2 * Copyright (c) 2001 Dag-Erling Co�dan Sm�rgrav 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer 10 * in this position and unchanged. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. The name of the author may not be used to endorse or promote products 15 * derived from this software without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 #include "opt_pseudofs.h" 33 34 #include <sys/param.h> 35 #include <sys/kernel.h> 36 #include <sys/systm.h> 37 #include <sys/eventhandler.h> 38 #include <sys/lock.h> 39 #include <sys/malloc.h> 40 #include <sys/mutex.h> 41 #include <sys/proc.h> 42 #include <sys/sysctl.h> 43 #include <sys/vnode.h> 44 45 #include <fs/pseudofs/pseudofs.h> 46 #include <fs/pseudofs/pseudofs_internal.h> 47 48 static MALLOC_DEFINE(M_PFSVNCACHE, "pfs_vncache", "pseudofs vnode cache"); 49 50 static struct mtx pfs_vncache_mutex; 51 static struct pfs_vdata *pfs_vncache; 52 static eventhandler_tag pfs_exit_tag; 53 static void pfs_exit(void *arg, struct proc *p); 54 55 SYSCTL_NODE(_vfs_pfs, OID_AUTO, vncache, CTLFLAG_RW, 0, 56 "pseudofs vnode cache"); 57 58 static int pfs_vncache_entries; 59 SYSCTL_INT(_vfs_pfs_vncache, OID_AUTO, entries, CTLFLAG_RD, 60 &pfs_vncache_entries, 0, 61 "number of entries in the vnode cache"); 62 63 static int pfs_vncache_maxentries; 64 SYSCTL_INT(_vfs_pfs_vncache, OID_AUTO, maxentries, CTLFLAG_RD, 65 &pfs_vncache_maxentries, 0, 66 "highest number of entries in the vnode cache"); 67 68 static int pfs_vncache_hits; 69 SYSCTL_INT(_vfs_pfs_vncache, OID_AUTO, hits, CTLFLAG_RD, 70 &pfs_vncache_hits, 0, 71 "number of cache hits since initialization"); 72 73 static int pfs_vncache_misses; 74 SYSCTL_INT(_vfs_pfs_vncache, OID_AUTO, misses, CTLFLAG_RD, 75 &pfs_vncache_misses, 0, 76 "number of cache misses since initialization"); 77 78 extern struct vop_vector pfs_vnodeops; /* XXX -> .h file */ 79 80 /* 81 * Initialize vnode cache 82 */ 83 void 84 pfs_vncache_load(void) 85 { 86 mtx_init(&pfs_vncache_mutex, "pseudofs_vncache", NULL, MTX_DEF); 87 pfs_exit_tag = EVENTHANDLER_REGISTER(process_exit, pfs_exit, NULL, 88 EVENTHANDLER_PRI_ANY); 89 } 90 91 /* 92 * Tear down vnode cache 93 */ 94 void 95 pfs_vncache_unload(void) 96 { 97 EVENTHANDLER_DEREGISTER(process_exit, pfs_exit_tag); 98 if (pfs_vncache_entries != 0) 99 printf("pfs_vncache_unload(): %d entries remaining\n", 100 pfs_vncache_entries); 101 mtx_destroy(&pfs_vncache_mutex); 102 } 103 104 /* 105 * Allocate a vnode 106 */ 107 int 108 pfs_vncache_alloc(struct mount *mp, struct vnode **vpp, 109 struct pfs_node *pn, pid_t pid) 110 { 111 struct pfs_vdata *pvd; 112 int error; 113 114 /* 115 * See if the vnode is in the cache. 116 * XXX linear search is not very efficient. 117 */ 118 mtx_lock(&pfs_vncache_mutex); 119 for (pvd = pfs_vncache; pvd; pvd = pvd->pvd_next) { 120 if (pvd->pvd_pn == pn && pvd->pvd_pid == pid && 121 pvd->pvd_vnode->v_mount == mp) { 122 if (vget(pvd->pvd_vnode, 0, curthread) == 0) { 123 ++pfs_vncache_hits; 124 *vpp = pvd->pvd_vnode; 125 mtx_unlock(&pfs_vncache_mutex); 126 /* XXX see comment at top of pfs_lookup() */ 127 cache_purge(*vpp); 128 vn_lock(*vpp, LK_RETRY | LK_EXCLUSIVE, 129 curthread); 130 return (0); 131 } 132 /* XXX if this can happen, we're in trouble */ 133 break; 134 } 135 } 136 mtx_unlock(&pfs_vncache_mutex); 137 ++pfs_vncache_misses; 138 139 /* nope, get a new one */ 140 MALLOC(pvd, struct pfs_vdata *, sizeof *pvd, M_PFSVNCACHE, M_WAITOK); 141 if (++pfs_vncache_entries > pfs_vncache_maxentries) 142 pfs_vncache_maxentries = pfs_vncache_entries; 143 error = getnewvnode("pseudofs", mp, &pfs_vnodeops, vpp); 144 if (error) { 145 FREE(pvd, M_PFSVNCACHE); 146 return (error); 147 } 148 pvd->pvd_pn = pn; 149 pvd->pvd_pid = pid; 150 (*vpp)->v_data = pvd; 151 switch (pn->pn_type) { 152 case pfstype_root: 153 (*vpp)->v_vflag = VV_ROOT; 154 #if 0 155 printf("root vnode allocated\n"); 156 #endif 157 /* fall through */ 158 case pfstype_dir: 159 case pfstype_this: 160 case pfstype_parent: 161 case pfstype_procdir: 162 (*vpp)->v_type = VDIR; 163 break; 164 case pfstype_file: 165 (*vpp)->v_type = VREG; 166 break; 167 case pfstype_symlink: 168 (*vpp)->v_type = VLNK; 169 break; 170 case pfstype_none: 171 KASSERT(0, ("pfs_vncache_alloc called for null node\n")); 172 default: 173 panic("%s has unexpected type: %d", pn->pn_name, pn->pn_type); 174 } 175 /* 176 * Propagate flag through to vnode so users know it can change 177 * if the process changes (i.e. execve) 178 */ 179 if ((pn->pn_flags & PFS_PROCDEP) != 0) 180 (*vpp)->v_vflag |= VV_PROCDEP; 181 pvd->pvd_vnode = *vpp; 182 mtx_lock(&pfs_vncache_mutex); 183 pvd->pvd_prev = NULL; 184 pvd->pvd_next = pfs_vncache; 185 if (pvd->pvd_next) 186 pvd->pvd_next->pvd_prev = pvd; 187 pfs_vncache = pvd; 188 mtx_unlock(&pfs_vncache_mutex); 189 (*vpp)->v_vnlock->lk_flags |= LK_CANRECURSE; 190 vn_lock(*vpp, LK_RETRY | LK_EXCLUSIVE, curthread); 191 return (0); 192 } 193 194 /* 195 * Free a vnode 196 */ 197 int 198 pfs_vncache_free(struct vnode *vp) 199 { 200 struct pfs_vdata *pvd; 201 202 mtx_lock(&pfs_vncache_mutex); 203 pvd = (struct pfs_vdata *)vp->v_data; 204 KASSERT(pvd != NULL, ("pfs_vncache_free(): no vnode data\n")); 205 if (pvd->pvd_next) 206 pvd->pvd_next->pvd_prev = pvd->pvd_prev; 207 if (pvd->pvd_prev) 208 pvd->pvd_prev->pvd_next = pvd->pvd_next; 209 else 210 pfs_vncache = pvd->pvd_next; 211 mtx_unlock(&pfs_vncache_mutex); 212 213 --pfs_vncache_entries; 214 FREE(pvd, M_PFSVNCACHE); 215 vp->v_data = NULL; 216 return (0); 217 } 218 219 /* 220 * Free all vnodes associated with a defunct process 221 * 222 * XXXRW: It is unfortunate that pfs_exit() always acquires and releases two 223 * mutexes (one of which is Giant) for every process exit, even if procfs 224 * isn't mounted. 225 */ 226 static void 227 pfs_exit(void *arg, struct proc *p) 228 { 229 struct pfs_vdata *pvd; 230 struct vnode *vnp; 231 232 if (pfs_vncache == NULL) 233 return; 234 mtx_lock(&Giant); 235 /* 236 * This is extremely inefficient due to the fact that vgone() not 237 * only indirectly modifies the vnode cache, but may also sleep. 238 * We can neither hold pfs_vncache_mutex across a vgone() call, 239 * nor make any assumptions about the state of the cache after 240 * vgone() returns. In consequence, we must start over after 241 * every vgone() call, and keep trying until we manage to traverse 242 * the entire cache. 243 * 244 * The only way to improve this situation is to change the data 245 * structure used to implement the cache. An obvious choice in 246 * this particular case would be a BST sorted by PID. 247 */ 248 mtx_lock(&pfs_vncache_mutex); 249 pvd = pfs_vncache; 250 while (pvd != NULL) { 251 if (pvd->pvd_pid == p->p_pid) { 252 vnp = pvd->pvd_vnode; 253 vhold(vnp); 254 mtx_unlock(&pfs_vncache_mutex); 255 VOP_LOCK(vnp, LK_EXCLUSIVE, curthread); 256 vgone(vnp); 257 VOP_UNLOCK(vnp, 0, curthread); 258 vdrop(vnp); 259 mtx_lock(&pfs_vncache_mutex); 260 pvd = pfs_vncache; 261 } else { 262 pvd = pvd->pvd_next; 263 } 264 } 265 mtx_unlock(&pfs_vncache_mutex); 266 mtx_unlock(&Giant); 267 } 268 269 /* 270 * Disable a pseudofs node, and free all vnodes associated with it 271 */ 272 int 273 pfs_disable(struct pfs_node *pn) 274 { 275 struct pfs_vdata *pvd; 276 struct vnode *vnp; 277 278 if (pn->pn_flags & PFS_DISABLED) 279 return (0); 280 pn->pn_flags |= PFS_DISABLED; 281 /* XXX see comment above nearly identical code in pfs_exit() */ 282 mtx_lock(&pfs_vncache_mutex); 283 pvd = pfs_vncache; 284 while (pvd != NULL) { 285 if (pvd->pvd_pn == pn) { 286 vnp = pvd->pvd_vnode; 287 vhold(vnp); 288 mtx_unlock(&pfs_vncache_mutex); 289 VOP_LOCK(vnp, LK_EXCLUSIVE, curthread); 290 vgone(vnp); 291 VOP_UNLOCK(vnp, 0, curthread); 292 vdrop(vnp); 293 mtx_lock(&pfs_vncache_mutex); 294 pvd = pfs_vncache; 295 } else { 296 pvd = pvd->pvd_next; 297 } 298 } 299 mtx_unlock(&pfs_vncache_mutex); 300 return (0); 301 } 302 303 /* 304 * Re-enable a disabled pseudofs node 305 */ 306 int 307 pfs_enable(struct pfs_node *pn) 308 { 309 pn->pn_flags &= ~PFS_DISABLED; 310 return (0); 311 } 312