nfs_commonport.c revision 192695
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 192695 2009-05-24 19:46:12Z 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, newnfs_realign_test, CTLFLAG_RW, &nfs_realign_test, 0, ""); 74SYSCTL_INT(_vfs_newnfs, OID_AUTO, newnfs_realign_count, CTLFLAG_RW, &nfs_realign_count, 0, ""); 75SYSCTL_INT(_vfs_newnfs, OID_AUTO, nfs4acl_enable, CTLFLAG_RW, &nfsrv_useacl, 0, ""); 76SYSCTL_STRING(_vfs_newnfs, OID_AUTO, callback_addr, CTLFLAG_RW, 77 nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr), ""); 78 79/* 80 * Defines for malloc 81 * (Here for FreeBSD, since they allocate storage.) 82 */ 83MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache"); 84MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id"); 85MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state", "NFSD V4 State (Openowner, Open, Lockowner, Delegation"); 86MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock"); 87MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file"); 88MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string"); 89MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map"); 90MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header"); 91MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle"); 92MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner"); 93MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open"); 94MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation"); 95MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client"); 96MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner"); 97MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock"); 98MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "New nfs vnode"); 99MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "New nfs Direct IO buffer"); 100MALLOC_DEFINE(M_NEWNFSDIROFF, "Newnfscl_diroff", "New NFS directory offset data"); 101 102/* 103 * Definition of mutex locks. 104 * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list 105 * and assorted other nfsd structures. 106 * Giant is used to protect the nfsd list and count, which is just 107 * updated when nfsd's start/stop and is grabbed for nfsrvd_dorpc() 108 * for the VFS ops. 109 */ 110struct mtx newnfsd_mtx; 111struct mtx nfs_sockl_mutex; 112struct mtx nfs_state_mutex; 113struct mtx nfs_nameid_mutex; 114struct mtx nfs_req_mutex; 115struct mtx nfs_slock_mutex; 116 117/* local functions */ 118static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *); 119 120#if defined(__i386__) 121/* 122 * These architectures don't need re-alignment, so just return. 123 */ 124void 125newnfs_realign(struct mbuf **pm) 126{ 127 128 return; 129} 130#else 131/* 132 * newnfs_realign: 133 * 134 * Check for badly aligned mbuf data and realign by copying the unaligned 135 * portion of the data into a new mbuf chain and freeing the portions 136 * of the old chain that were replaced. 137 * 138 * We cannot simply realign the data within the existing mbuf chain 139 * because the underlying buffers may contain other rpc commands and 140 * we cannot afford to overwrite them. 141 * 142 * We would prefer to avoid this situation entirely. The situation does 143 * not occur with NFS/UDP and is supposed to only occassionally occur 144 * with TCP. Use vfs.nfs.realign_count and realign_test to check this. 145 * 146 */ 147void 148newnfs_realign(struct mbuf **pm) 149{ 150 struct mbuf *m, *n; 151 int off, space; 152 153 ++nfs_realign_test; 154 while ((m = *pm) != NULL) { 155 if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) { 156 /* 157 * NB: we can't depend on m_pkthdr.len to help us 158 * decide what to do here. May not be worth doing 159 * the m_length calculation as m_copyback will 160 * expand the mbuf chain below as needed. 161 */ 162 space = m_length(m, NULL); 163 if (space >= MINCLSIZE) { 164 /* NB: m_copyback handles space > MCLBYTES */ 165 n = m_getcl(M_WAITOK, MT_DATA, 0); 166 } else 167 n = m_get(M_WAITOK, MT_DATA); 168 if (n == NULL) 169 return; 170 /* 171 * Align the remainder of the mbuf chain. 172 */ 173 n->m_len = 0; 174 off = 0; 175 while (m != NULL) { 176 m_copyback(n, off, m->m_len, mtod(m, caddr_t)); 177 off += m->m_len; 178 m = m->m_next; 179 } 180 m_freem(*pm); 181 *pm = n; 182 ++nfs_realign_count; 183 break; 184 } 185 pm = &m->m_next; 186 } 187} 188#endif /* !__i386__ */ 189 190#ifdef notdef 191static void 192nfsrv_object_create(struct vnode *vp, struct thread *td) 193{ 194 195 if (vp == NULL || vp->v_type != VREG) 196 return; 197 (void) vfs_object_create(vp, td, td->td_ucred); 198} 199#endif 200 201/* 202 * Look up a file name. Basically just initialize stuff and call namei(). 203 */ 204int 205nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p) 206{ 207 int error; 208 209 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname, p); 210 error = namei(ndp); 211 if (!error) { 212 NDFREE(ndp, NDF_ONLY_PNBUF); 213 } 214 return (error); 215} 216 217/* 218 * Copy NFS uid, gids to the cred structure. 219 */ 220void 221newnfs_copycred(struct nfscred *nfscr, struct ucred *cr) 222{ 223 int ngroups, i; 224 225 cr->cr_uid = nfscr->nfsc_uid; 226 ngroups = (nfscr->nfsc_ngroups < NGROUPS) ? 227 nfscr->nfsc_ngroups : NGROUPS; 228 for (i = 0; i < ngroups; i++) 229 cr->cr_groups[i] = nfscr->nfsc_groups[i]; 230 cr->cr_ngroups = ngroups; 231} 232 233/* 234 * Map args from nfsmsleep() to msleep(). 235 */ 236int 237nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg, 238 struct timespec *ts) 239{ 240 u_int64_t nsecval; 241 int error, timeo; 242 243 if (ts) { 244 timeo = hz * ts->tv_sec; 245 nsecval = (u_int64_t)ts->tv_nsec; 246 nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) / 247 1000000000; 248 timeo += (int)nsecval; 249 } else { 250 timeo = 0; 251 } 252 error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo); 253 return (error); 254} 255 256/* 257 * Get the file system info for the server. For now, just assume FFS. 258 */ 259void 260nfsvno_getfs(struct nfsfsinfo *sip, int isdgram) 261{ 262 int pref; 263 264 /* 265 * XXX 266 * There should be file system VFS OP(s) to get this information. 267 * For now, assume ufs. 268 */ 269 if (isdgram) 270 pref = NFS_MAXDGRAMDATA; 271 else 272 pref = NFS_MAXDATA; 273 sip->fs_rtmax = NFS_MAXDATA; 274 sip->fs_rtpref = pref; 275 sip->fs_rtmult = NFS_FABLKSIZE; 276 sip->fs_wtmax = NFS_MAXDATA; 277 sip->fs_wtpref = pref; 278 sip->fs_wtmult = NFS_FABLKSIZE; 279 sip->fs_dtpref = pref; 280 sip->fs_maxfilesize = 0xffffffffffffffffull; 281 sip->fs_timedelta.tv_sec = 0; 282 sip->fs_timedelta.tv_nsec = 1; 283 sip->fs_properties = (NFSV3FSINFO_LINK | 284 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 285 NFSV3FSINFO_CANSETTIME); 286} 287 288/* Fake nfsrv_atroot. Just return 0 */ 289int 290nfsrv_atroot(struct vnode *vp, long *retp) 291{ 292 293 return (0); 294} 295 296/* 297 * Set the credentials to refer to root. 298 * If only the various BSDen could agree on whether cr_gid is a separate 299 * field or cr_groups[0]... 300 */ 301void 302newnfs_setroot(struct ucred *cred) 303{ 304 305 cred->cr_uid = 0; 306 cred->cr_groups[0] = 0; 307 cred->cr_ngroups = 1; 308} 309 310/* 311 * Get the client credential. Used for Renew and recovery. 312 */ 313struct ucred * 314newnfs_getcred(void) 315{ 316 struct ucred *cred; 317 struct thread *td = curthread; 318 319 cred = crdup(td->td_ucred); 320 newnfs_setroot(cred); 321 return (cred); 322} 323 324/* 325 * Nfs timer routine 326 * Call the nfsd's timer function once/sec. 327 */ 328void 329newnfs_timer(void *arg) 330{ 331 static time_t lasttime = 0; 332 /* 333 * Call the server timer, if set up. 334 * The argument indicates if it is the next second and therefore 335 * leases should be checked. 336 */ 337 if (lasttime != NFSD_MONOSEC) { 338 lasttime = NFSD_MONOSEC; 339 if (nfsd_call_servertimer != NULL) 340 (*nfsd_call_servertimer)(); 341 } 342 callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL); 343} 344 345 346/* 347 * sleep for a short period of time. 348 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on 349 * an event that never gets a wakeup. Only return EINTR or 0. 350 */ 351int 352nfs_catnap(int prio, const char *wmesg) 353{ 354 static int non_event; 355 int ret; 356 357 ret = tsleep(&non_event, prio, wmesg, 1); 358 if (ret != EINTR) 359 ret = 0; 360 return (ret); 361} 362 363/* 364 * Get referral. For now, just fail. 365 */ 366struct nfsreferral * 367nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno) 368{ 369 370 return (NULL); 371} 372 373static int 374nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap) 375{ 376 int error; 377 378 error = nfssvc_call(td, uap, td->td_ucred); 379 return (error); 380} 381 382static int 383nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred) 384{ 385 int error = EINVAL; 386 struct nfsd_idargs nid; 387 388 if (uap->flag & NFSSVC_IDNAME) { 389 error = copyin(uap->argp, (caddr_t)&nid, sizeof (nid)); 390 if (error) 391 return (error); 392 error = nfssvc_idname(&nid); 393 return (error); 394 } else if (uap->flag & NFSSVC_GETSTATS) { 395 error = copyout(&newnfsstats, 396 CAST_USER_ADDR_T(uap->argp), sizeof (newnfsstats)); 397 return (error); 398 } else if (uap->flag & NFSSVC_NFSUSERDPORT) { 399 u_short sockport; 400 401 error = copyin(uap->argp, (caddr_t)&sockport, 402 sizeof (u_short)); 403 if (!error) 404 error = nfsrv_nfsuserdport(sockport, p); 405 } else if (uap->flag & NFSSVC_NFSUSERDDELPORT) { 406 nfsrv_nfsuserddelport(); 407 error = 0; 408 } 409 return (error); 410} 411 412/* 413 * called by all three modevent routines, so that it gets things 414 * initialized soon enough. 415 */ 416void 417newnfs_portinit(void) 418{ 419 static int inited = 0; 420 421 if (inited) 422 return; 423 inited = 1; 424 /* Initialize SMP locks used by both client and server. */ 425 mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF); 426 mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF); 427} 428 429extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *); 430 431/* 432 * Called once to initialize data structures... 433 */ 434static int 435nfscommon_modevent(module_t mod, int type, void *data) 436{ 437 int error = 0; 438 static int loaded = 0; 439 440 switch (type) { 441 case MOD_LOAD: 442 if (loaded) 443 return (0); 444 newnfs_portinit(); 445 mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF); 446 mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF); 447 mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF); 448 mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF); 449 mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL, 450 MTX_DEF); 451 callout_init(&newnfsd_callout, CALLOUT_MPSAFE); 452 newnfs_init(); 453 nfsd_call_nfscommon = nfssvc_nfscommon; 454 loaded = 1; 455 break; 456 457 case MOD_UNLOAD: 458 if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != 0 || 459 nfs_numnfscbd != 0) { 460 error = EBUSY; 461 break; 462 } 463 464 nfsd_call_nfscommon = NULL; 465 callout_drain(&newnfsd_callout); 466 /* and get rid of the mutexes */ 467 mtx_destroy(&nfs_nameid_mutex); 468 mtx_destroy(&newnfsd_mtx); 469 mtx_destroy(&nfs_state_mutex); 470 mtx_destroy(&nfs_sockl_mutex); 471 mtx_destroy(&nfs_slock_mutex); 472 mtx_destroy(&nfs_req_mutex); 473 mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx); 474 loaded = 0; 475 break; 476 default: 477 error = EOPNOTSUPP; 478 break; 479 } 480 return error; 481} 482static moduledata_t nfscommon_mod = { 483 "nfscommon", 484 nfscommon_modevent, 485 NULL, 486}; 487DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY); 488 489/* So that loader and kldload(2) can find us, wherever we are.. */ 490MODULE_VERSION(nfscommon, 1); 491MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1); 492MODULE_DEPEND(nfscommon, krpc, 1, 1, 1); 493 494