nfs_commonport.c revision 221306
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 */ 33 34#include <sys/cdefs.h> 35__FBSDID("$FreeBSD: head/sys/fs/nfs/nfs_commonport.c 221306 2011-05-01 22:19:52Z rmacklem $"); 36 37/* 38 * Functions that need to be different for different versions of BSD 39 * kernel should be kept here, along with any global storage specific 40 * to this BSD variant. 41 */ 42#include <fs/nfs/nfsport.h> 43#include <sys/sysctl.h> 44#include <vm/vm.h> 45#include <vm/vm_object.h> 46#include <vm/vm_page.h> 47#include <vm/vm_param.h> 48#include <vm/vm_map.h> 49#include <vm/vm_kern.h> 50#include <vm/vm_extern.h> 51#include <vm/uma.h> 52#include <vm/uma_int.h> 53 54extern int nfscl_ticks; 55extern int nfsrv_nfsuserd; 56extern struct nfssockreq nfsrv_nfsuserdsock; 57extern void (*nfsd_call_recall)(struct vnode *, int, struct ucred *, 58 struct thread *); 59extern int nfsrv_useacl; 60struct mount nfsv4root_mnt; 61int newnfs_numnfsd = 0; 62struct nfsstats newnfsstats; 63int nfs_numnfscbd = 0; 64char nfsv4_callbackaddr[INET6_ADDRSTRLEN]; 65struct callout newnfsd_callout; 66void (*nfsd_call_servertimer)(void) = NULL; 67void (*ncl_call_invalcaches)(struct vnode *) = NULL; 68 69static int nfs_realign_test; 70static int nfs_realign_count; 71 72SYSCTL_NODE(_vfs, OID_AUTO, newnfs, CTLFLAG_RW, 0, "New NFS filesystem"); 73SYSCTL_INT(_vfs_newnfs, OID_AUTO, realign_test, CTLFLAG_RW, &nfs_realign_test, 74 0, "Number of realign tests done"); 75SYSCTL_INT(_vfs_newnfs, OID_AUTO, realign_count, CTLFLAG_RW, &nfs_realign_count, 76 0, "Number of mbuf realignments done"); 77SYSCTL_INT(_vfs_newnfs, OID_AUTO, nfs4acl_enable, CTLFLAG_RW, &nfsrv_useacl, 78 0, "Enable NFSv4 ACLs"); 79SYSCTL_STRING(_vfs_newnfs, OID_AUTO, callback_addr, CTLFLAG_RW, 80 nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr), 81 "NFSv4 callback addr for server to use"); 82 83/* 84 * Defines for malloc 85 * (Here for FreeBSD, since they allocate storage.) 86 */ 87MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache"); 88MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id"); 89MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state", 90 "NFSD V4 State (Openowner, Open, Lockowner, Delegation"); 91MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock"); 92MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file"); 93MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string"); 94MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map"); 95MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header"); 96MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle"); 97MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner"); 98MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open"); 99MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation"); 100MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client"); 101MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner"); 102MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock"); 103MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "New nfs vnode"); 104MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "New nfs Direct IO buffer"); 105MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroffdiroff", 106 "New NFS directory offset data"); 107MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback", 108 "New NFS local lock rollback"); 109 110/* 111 * Definition of mutex locks. 112 * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list 113 * and assorted other nfsd structures. 114 * Giant is used to protect the nfsd list and count, which is just 115 * updated when nfsd's start/stop and is grabbed for nfsrvd_dorpc() 116 * for the VFS ops. 117 */ 118struct mtx newnfsd_mtx; 119struct mtx nfs_sockl_mutex; 120struct mtx nfs_state_mutex; 121struct mtx nfs_nameid_mutex; 122struct mtx nfs_req_mutex; 123struct mtx nfs_slock_mutex; 124 125/* local functions */ 126static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *); 127 128#ifdef __NO_STRICT_ALIGNMENT 129/* 130 * These architectures don't need re-alignment, so just return. 131 */ 132void 133newnfs_realign(struct mbuf **pm) 134{ 135 136 return; 137} 138#else /* !__NO_STRICT_ALIGNMENT */ 139/* 140 * newnfs_realign: 141 * 142 * Check for badly aligned mbuf data and realign by copying the unaligned 143 * portion of the data into a new mbuf chain and freeing the portions 144 * of the old chain that were replaced. 145 * 146 * We cannot simply realign the data within the existing mbuf chain 147 * because the underlying buffers may contain other rpc commands and 148 * we cannot afford to overwrite them. 149 * 150 * We would prefer to avoid this situation entirely. The situation does 151 * not occur with NFS/UDP and is supposed to only occassionally occur 152 * with TCP. Use vfs.nfs.realign_count and realign_test to check this. 153 * 154 */ 155void 156newnfs_realign(struct mbuf **pm) 157{ 158 struct mbuf *m, *n; 159 int off, space; 160 161 ++nfs_realign_test; 162 while ((m = *pm) != NULL) { 163 if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) { 164 /* 165 * NB: we can't depend on m_pkthdr.len to help us 166 * decide what to do here. May not be worth doing 167 * the m_length calculation as m_copyback will 168 * expand the mbuf chain below as needed. 169 */ 170 space = m_length(m, NULL); 171 if (space >= MINCLSIZE) { 172 /* NB: m_copyback handles space > MCLBYTES */ 173 n = m_getcl(M_WAITOK, MT_DATA, 0); 174 } else 175 n = m_get(M_WAITOK, MT_DATA); 176 if (n == NULL) 177 return; 178 /* 179 * Align the remainder of the mbuf chain. 180 */ 181 n->m_len = 0; 182 off = 0; 183 while (m != NULL) { 184 m_copyback(n, off, m->m_len, mtod(m, caddr_t)); 185 off += m->m_len; 186 m = m->m_next; 187 } 188 m_freem(*pm); 189 *pm = n; 190 ++nfs_realign_count; 191 break; 192 } 193 pm = &m->m_next; 194 } 195} 196#endif /* __NO_STRICT_ALIGNMENT */ 197 198#ifdef notdef 199static void 200nfsrv_object_create(struct vnode *vp, struct thread *td) 201{ 202 203 if (vp == NULL || vp->v_type != VREG) 204 return; 205 (void) vfs_object_create(vp, td, td->td_ucred); 206} 207#endif 208 209/* 210 * Look up a file name. Basically just initialize stuff and call namei(). 211 */ 212int 213nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p) 214{ 215 int error; 216 217 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE, UIO_USERSPACE, fname, 218 p); 219 error = namei(ndp); 220 if (!error) { 221 NDFREE(ndp, NDF_ONLY_PNBUF); 222 } 223 return (error); 224} 225 226/* 227 * Copy NFS uid, gids to the cred structure. 228 */ 229void 230newnfs_copycred(struct nfscred *nfscr, struct ucred *cr) 231{ 232 233 KASSERT(nfscr->nfsc_ngroups >= 0, 234 ("newnfs_copycred: negative nfsc_ngroups")); 235 cr->cr_uid = nfscr->nfsc_uid; 236 crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups); 237} 238 239/* 240 * Map args from nfsmsleep() to msleep(). 241 */ 242int 243nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg, 244 struct timespec *ts) 245{ 246 u_int64_t nsecval; 247 int error, timeo; 248 249 if (ts) { 250 timeo = hz * ts->tv_sec; 251 nsecval = (u_int64_t)ts->tv_nsec; 252 nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) / 253 1000000000; 254 timeo += (int)nsecval; 255 } else { 256 timeo = 0; 257 } 258 error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo); 259 return (error); 260} 261 262/* 263 * Get the file system info for the server. For now, just assume FFS. 264 */ 265void 266nfsvno_getfs(struct nfsfsinfo *sip, int isdgram) 267{ 268 int pref; 269 270 /* 271 * XXX 272 * There should be file system VFS OP(s) to get this information. 273 * For now, assume ufs. 274 */ 275 if (isdgram) 276 pref = NFS_MAXDGRAMDATA; 277 else 278 pref = NFS_MAXDATA; 279 sip->fs_rtmax = NFS_MAXDATA; 280 sip->fs_rtpref = pref; 281 sip->fs_rtmult = NFS_FABLKSIZE; 282 sip->fs_wtmax = NFS_MAXDATA; 283 sip->fs_wtpref = pref; 284 sip->fs_wtmult = NFS_FABLKSIZE; 285 sip->fs_dtpref = pref; 286 sip->fs_maxfilesize = 0xffffffffffffffffull; 287 sip->fs_timedelta.tv_sec = 0; 288 sip->fs_timedelta.tv_nsec = 1; 289 sip->fs_properties = (NFSV3FSINFO_LINK | 290 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 291 NFSV3FSINFO_CANSETTIME); 292} 293 294/* Fake nfsrv_atroot. Just return 0 */ 295int 296nfsrv_atroot(struct vnode *vp, long *retp) 297{ 298 299 return (0); 300} 301 302/* 303 * Set the credentials to refer to root. 304 * If only the various BSDen could agree on whether cr_gid is a separate 305 * field or cr_groups[0]... 306 */ 307void 308newnfs_setroot(struct ucred *cred) 309{ 310 311 cred->cr_uid = 0; 312 cred->cr_groups[0] = 0; 313 cred->cr_ngroups = 1; 314} 315 316/* 317 * Get the client credential. Used for Renew and recovery. 318 */ 319struct ucred * 320newnfs_getcred(void) 321{ 322 struct ucred *cred; 323 struct thread *td = curthread; 324 325 cred = crdup(td->td_ucred); 326 newnfs_setroot(cred); 327 return (cred); 328} 329 330/* 331 * Nfs timer routine 332 * Call the nfsd's timer function once/sec. 333 */ 334void 335newnfs_timer(void *arg) 336{ 337 static time_t lasttime = 0; 338 /* 339 * Call the server timer, if set up. 340 * The argument indicates if it is the next second and therefore 341 * leases should be checked. 342 */ 343 if (lasttime != NFSD_MONOSEC) { 344 lasttime = NFSD_MONOSEC; 345 if (nfsd_call_servertimer != NULL) 346 (*nfsd_call_servertimer)(); 347 } 348 callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL); 349} 350 351 352/* 353 * Sleep for a short period of time unless errval == NFSERR_GRACE, where 354 * the sleep should be for 5 seconds. 355 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on 356 * an event that never gets a wakeup. Only return EINTR or 0. 357 */ 358int 359nfs_catnap(int prio, int errval, const char *wmesg) 360{ 361 static int non_event; 362 int ret; 363 364 if (errval == NFSERR_GRACE) 365 ret = tsleep(&non_event, prio, wmesg, 5 * hz); 366 else 367 ret = tsleep(&non_event, prio, wmesg, 1); 368 if (ret != EINTR) 369 ret = 0; 370 return (ret); 371} 372 373/* 374 * Get referral. For now, just fail. 375 */ 376struct nfsreferral * 377nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno) 378{ 379 380 return (NULL); 381} 382 383static int 384nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap) 385{ 386 int error; 387 388 error = nfssvc_call(td, uap, td->td_ucred); 389 return (error); 390} 391 392static int 393nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred) 394{ 395 int error = EINVAL; 396 struct nfsd_idargs nid; 397 398 if (uap->flag & NFSSVC_IDNAME) { 399 error = copyin(uap->argp, (caddr_t)&nid, sizeof (nid)); 400 if (error) 401 return (error); 402 error = nfssvc_idname(&nid); 403 return (error); 404 } else if (uap->flag & NFSSVC_GETSTATS) { 405 error = copyout(&newnfsstats, 406 CAST_USER_ADDR_T(uap->argp), sizeof (newnfsstats)); 407 if ((uap->flag & NFSSVC_ZEROSTATS) != 0 && error == 0) 408 bzero(&newnfsstats, sizeof(newnfsstats)); 409 return (error); 410 } else if (uap->flag & NFSSVC_NFSUSERDPORT) { 411 u_short sockport; 412 413 error = copyin(uap->argp, (caddr_t)&sockport, 414 sizeof (u_short)); 415 if (!error) 416 error = nfsrv_nfsuserdport(sockport, p); 417 } else if (uap->flag & NFSSVC_NFSUSERDDELPORT) { 418 nfsrv_nfsuserddelport(); 419 error = 0; 420 } 421 return (error); 422} 423 424/* 425 * called by all three modevent routines, so that it gets things 426 * initialized soon enough. 427 */ 428void 429newnfs_portinit(void) 430{ 431 static int inited = 0; 432 433 if (inited) 434 return; 435 inited = 1; 436 /* Initialize SMP locks used by both client and server. */ 437 mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF); 438 mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF); 439} 440 441/* 442 * Determine if the file system supports NFSv4 ACLs. 443 * Return 1 if it does, 0 otherwise. 444 */ 445int 446nfs_supportsnfsv4acls(struct vnode *vp) 447{ 448 int error; 449 register_t retval; 450 451 ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls"); 452 453 if (nfsrv_useacl == 0) 454 return (0); 455 error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval); 456 if (error == 0 && retval != 0) 457 return (1); 458 return (0); 459} 460 461extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *); 462 463/* 464 * Called once to initialize data structures... 465 */ 466static int 467nfscommon_modevent(module_t mod, int type, void *data) 468{ 469 int error = 0; 470 static int loaded = 0; 471 472 switch (type) { 473 case MOD_LOAD: 474 if (loaded) 475 return (0); 476 newnfs_portinit(); 477 mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF); 478 mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF); 479 mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF); 480 mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF); 481 mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL, 482 MTX_DEF); 483 callout_init(&newnfsd_callout, CALLOUT_MPSAFE); 484 newnfs_init(); 485 nfsd_call_nfscommon = nfssvc_nfscommon; 486 loaded = 1; 487 break; 488 489 case MOD_UNLOAD: 490 if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != 0 || 491 nfs_numnfscbd != 0) { 492 error = EBUSY; 493 break; 494 } 495 496 nfsd_call_nfscommon = NULL; 497 callout_drain(&newnfsd_callout); 498 /* and get rid of the mutexes */ 499 mtx_destroy(&nfs_nameid_mutex); 500 mtx_destroy(&newnfsd_mtx); 501 mtx_destroy(&nfs_state_mutex); 502 mtx_destroy(&nfs_sockl_mutex); 503 mtx_destroy(&nfs_slock_mutex); 504 mtx_destroy(&nfs_req_mutex); 505 mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx); 506 loaded = 0; 507 break; 508 default: 509 error = EOPNOTSUPP; 510 break; 511 } 512 return error; 513} 514static moduledata_t nfscommon_mod = { 515 "nfscommon", 516 nfscommon_modevent, 517 NULL, 518}; 519DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY); 520 521/* So that loader and kldload(2) can find us, wherever we are.. */ 522MODULE_VERSION(nfscommon, 1); 523MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1); 524MODULE_DEPEND(nfscommon, krpc, 1, 1, 1); 525 526