proc->thread stage 4: rework the VFS and DEVICE subsystems to take thread
[dragonfly.git] / sys / vfs / nfs / nfs_vfsops.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1989, 1993, 1995
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. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the University of
19 * California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * @(#)nfs_vfsops.c 8.12 (Berkeley) 5/20/95
37 * $FreeBSD: src/sys/nfs/nfs_vfsops.c,v 1.91.2.7 2003/01/27 20:04:08 dillon Exp $
dadab5e9 38 * $DragonFly: src/sys/vfs/nfs/nfs_vfsops.c,v 1.3 2003/06/25 03:56:07 dillon Exp $
984263bc
MD
39 */
40
41#include "opt_bootp.h"
42
43#include <sys/param.h>
44#include <sys/sockio.h>
45#include <sys/proc.h>
46#include <sys/vnode.h>
47#include <sys/kernel.h>
48#include <sys/sysctl.h>
49#include <sys/malloc.h>
50#include <sys/mount.h>
51#include <sys/mbuf.h>
52#include <sys/socket.h>
53#include <sys/socketvar.h>
54#include <sys/systm.h>
55
56#include <vm/vm.h>
57#include <vm/vm_extern.h>
58#include <vm/vm_zone.h>
59
60#include <net/if.h>
61#include <net/route.h>
62#include <netinet/in.h>
63
64#include <nfs/rpcv2.h>
65#include <nfs/nfsproto.h>
66#include <nfs/nfs.h>
67#include <nfs/nfsnode.h>
68#include <nfs/nfsmount.h>
69#include <nfs/xdr_subs.h>
70#include <nfs/nfsm_subs.h>
71#include <nfs/nfsdiskless.h>
72#include <nfs/nqnfs.h>
73
74extern int nfs_mountroot __P((struct mount *mp));
75
76extern int nfs_ticks;
77
78MALLOC_DEFINE(M_NFSREQ, "NFS req", "NFS request header");
79MALLOC_DEFINE(M_NFSBIGFH, "NFSV3 bigfh", "NFS version 3 file handle");
80MALLOC_DEFINE(M_NFSD, "NFS daemon", "Nfs server daemon structure");
81MALLOC_DEFINE(M_NFSDIROFF, "NFSV3 diroff", "NFS directory offset data");
82MALLOC_DEFINE(M_NFSRVDESC, "NFSV3 srvdesc", "NFS server socket descriptor");
83MALLOC_DEFINE(M_NFSUID, "NFS uid", "Nfs uid mapping structure");
84MALLOC_DEFINE(M_NQLEASE, "NQNFS Lease", "Nqnfs lease");
85MALLOC_DEFINE(M_NFSHASH, "NFS hash", "NFS hash tables");
86
87vm_zone_t nfsmount_zone;
88
89struct nfsstats nfsstats;
90SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW, 0, "NFS filesystem");
91SYSCTL_STRUCT(_vfs_nfs, NFS_NFSSTATS, nfsstats, CTLFLAG_RD,
92 &nfsstats, nfsstats, "");
93static int nfs_ip_paranoia = 1;
94SYSCTL_INT(_vfs_nfs, OID_AUTO, nfs_ip_paranoia, CTLFLAG_RW,
95 &nfs_ip_paranoia, 0, "");
96#ifdef NFS_DEBUG
97int nfs_debug;
98SYSCTL_INT(_vfs_nfs, OID_AUTO, debug, CTLFLAG_RW, &nfs_debug, 0, "");
99#endif
100
101static int nfs_iosize __P((struct nfsmount *nmp));
102static void nfs_decode_args __P((struct nfsmount *nmp,
103 struct nfs_args *argp));
104static int mountnfs __P((struct nfs_args *,struct mount *,
105 struct sockaddr *,char *,char *,struct vnode **));
106static int nfs_mount __P(( struct mount *mp, char *path, caddr_t data,
dadab5e9 107 struct nameidata *ndp, struct thread *td));
984263bc 108static int nfs_unmount __P(( struct mount *mp, int mntflags,
dadab5e9 109 struct thread *td));
984263bc
MD
110static int nfs_root __P(( struct mount *mp, struct vnode **vpp));
111static int nfs_statfs __P(( struct mount *mp, struct statfs *sbp,
dadab5e9 112 struct thread *td));
984263bc 113static int nfs_sync __P(( struct mount *mp, int waitfor,
dadab5e9 114 struct ucred *cred, struct thread *td));
984263bc
MD
115
116/*
117 * nfs vfs operations.
118 */
119static struct vfsops nfs_vfsops = {
120 nfs_mount,
121 vfs_stdstart,
122 nfs_unmount,
123 nfs_root,
124 vfs_stdquotactl,
125 nfs_statfs,
126 nfs_sync,
127 vfs_stdvget,
128 vfs_stdfhtovp, /* shouldn't happen */
129 vfs_stdcheckexp,
130 vfs_stdvptofh, /* shouldn't happen */
131 nfs_init,
132 nfs_uninit,
133 vfs_stdextattrctl,
134};
135VFS_SET(nfs_vfsops, nfs, VFCF_NETWORK);
136
137/*
138 * This structure must be filled in by a primary bootstrap or bootstrap
139 * server for a diskless/dataless machine. It is initialized below just
140 * to ensure that it is allocated to initialized data (.data not .bss).
141 */
142struct nfs_diskless nfs_diskless = { { { 0 } } };
143struct nfsv3_diskless nfsv3_diskless = { { { 0 } } };
144int nfs_diskless_valid = 0;
145
146SYSCTL_INT(_vfs_nfs, OID_AUTO, diskless_valid, CTLFLAG_RD,
147 &nfs_diskless_valid, 0, "");
148
149SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_rootpath, CTLFLAG_RD,
150 nfsv3_diskless.root_hostnam, 0, "");
151
152SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_rootaddr, CTLFLAG_RD,
153 &nfsv3_diskless.root_saddr, sizeof nfsv3_diskless.root_saddr,
154 "%Ssockaddr_in", "");
155
156SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_swappath, CTLFLAG_RD,
157 nfsv3_diskless.swap_hostnam, 0, "");
158
159SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_swapaddr, CTLFLAG_RD,
160 &nfsv3_diskless.swap_saddr, sizeof nfsv3_diskless.swap_saddr,
161 "%Ssockaddr_in","");
162
163
164void nfsargs_ntoh __P((struct nfs_args *));
165static int nfs_mountdiskless __P((char *, char *, int,
166 struct sockaddr_in *, struct nfs_args *,
dadab5e9 167 struct thread *, struct vnode **,
984263bc
MD
168 struct mount **));
169static void nfs_convert_diskless __P((void));
170static void nfs_convert_oargs __P((struct nfs_args *args,
171 struct onfs_args *oargs));
172
173static int
174nfs_iosize(nmp)
175 struct nfsmount* nmp;
176{
177 int iosize;
178
179 /*
180 * Calculate the size used for io buffers. Use the larger
181 * of the two sizes to minimise nfs requests but make sure
182 * that it is at least one VM page to avoid wasting buffer
183 * space.
184 */
185 iosize = max(nmp->nm_rsize, nmp->nm_wsize);
186 if (iosize < PAGE_SIZE) iosize = PAGE_SIZE;
187 return iosize;
188}
189
190static void
191nfs_convert_oargs(args, oargs)
192 struct nfs_args *args;
193 struct onfs_args *oargs;
194{
195 args->version = NFS_ARGSVERSION;
196 args->addr = oargs->addr;
197 args->addrlen = oargs->addrlen;
198 args->sotype = oargs->sotype;
199 args->proto = oargs->proto;
200 args->fh = oargs->fh;
201 args->fhsize = oargs->fhsize;
202 args->flags = oargs->flags;
203 args->wsize = oargs->wsize;
204 args->rsize = oargs->rsize;
205 args->readdirsize = oargs->readdirsize;
206 args->timeo = oargs->timeo;
207 args->retrans = oargs->retrans;
208 args->maxgrouplist = oargs->maxgrouplist;
209 args->readahead = oargs->readahead;
210 args->leaseterm = oargs->leaseterm;
211 args->deadthresh = oargs->deadthresh;
212 args->hostname = oargs->hostname;
213}
214
215static void
216nfs_convert_diskless()
217{
218 bcopy(&nfs_diskless.myif, &nfsv3_diskless.myif,
219 sizeof(struct ifaliasreq));
220 bcopy(&nfs_diskless.mygateway, &nfsv3_diskless.mygateway,
221 sizeof(struct sockaddr_in));
222 nfs_convert_oargs(&nfsv3_diskless.swap_args,&nfs_diskless.swap_args);
223 nfsv3_diskless.swap_fhsize = NFSX_V2FH;
224 bcopy(nfs_diskless.swap_fh,nfsv3_diskless.swap_fh,NFSX_V2FH);
225 bcopy(&nfs_diskless.swap_saddr,&nfsv3_diskless.swap_saddr,
226 sizeof(struct sockaddr_in));
227 bcopy(nfs_diskless.swap_hostnam,nfsv3_diskless.swap_hostnam, MNAMELEN);
228 nfsv3_diskless.swap_nblks = nfs_diskless.swap_nblks;
229 bcopy(&nfs_diskless.swap_ucred, &nfsv3_diskless.swap_ucred,
230 sizeof(struct ucred));
231 nfs_convert_oargs(&nfsv3_diskless.root_args,&nfs_diskless.root_args);
232 nfsv3_diskless.root_fhsize = NFSX_V2FH;
233 bcopy(nfs_diskless.root_fh,nfsv3_diskless.root_fh,NFSX_V2FH);
234 bcopy(&nfs_diskless.root_saddr,&nfsv3_diskless.root_saddr,
235 sizeof(struct sockaddr_in));
236 bcopy(nfs_diskless.root_hostnam,nfsv3_diskless.root_hostnam, MNAMELEN);
237 nfsv3_diskless.root_time = nfs_diskless.root_time;
238 bcopy(nfs_diskless.my_hostnam,nfsv3_diskless.my_hostnam,
239 MAXHOSTNAMELEN);
240 nfs_diskless_valid = 3;
241}
242
243/*
244 * nfs statfs call
245 */
246int
dadab5e9 247nfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
984263bc
MD
248{
249 register struct vnode *vp;
250 register struct nfs_statfs *sfp;
251 register caddr_t cp;
252 register u_int32_t *tl;
253 register int32_t t1, t2;
254 caddr_t bpos, dpos, cp2;
255 struct nfsmount *nmp = VFSTONFS(mp);
256 int error = 0, v3 = (nmp->nm_flag & NFSMNT_NFSV3), retattr;
257 struct mbuf *mreq, *mrep, *md, *mb, *mb2;
258 struct ucred *cred;
259 struct nfsnode *np;
260 u_quad_t tquad;
261
262#ifndef nolint
263 sfp = (struct nfs_statfs *)0;
264#endif
265 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
266 if (error)
267 return (error);
268 vp = NFSTOV(np);
269 cred = crget();
270 cred->cr_ngroups = 1;
271 if (v3 && (nmp->nm_state & NFSSTA_GOTFSINFO) == 0)
dadab5e9 272 (void)nfs_fsinfo(nmp, vp, cred, td);
984263bc
MD
273 nfsstats.rpccnt[NFSPROC_FSSTAT]++;
274 nfsm_reqhead(vp, NFSPROC_FSSTAT, NFSX_FH(v3));
275 nfsm_fhtom(vp, v3);
dadab5e9 276 nfsm_request(vp, NFSPROC_FSSTAT, td, cred);
984263bc
MD
277 if (v3)
278 nfsm_postop_attr(vp, retattr);
279 if (error) {
280 if (mrep != NULL)
281 m_freem(mrep);
282 goto nfsmout;
283 }
284 nfsm_dissect(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
285 sbp->f_flags = nmp->nm_flag;
286 sbp->f_iosize = nfs_iosize(nmp);
287 if (v3) {
288 sbp->f_bsize = NFS_FABLKSIZE;
289 tquad = fxdr_hyper(&sfp->sf_tbytes);
290 sbp->f_blocks = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
291 tquad = fxdr_hyper(&sfp->sf_fbytes);
292 sbp->f_bfree = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
293 tquad = fxdr_hyper(&sfp->sf_abytes);
294 sbp->f_bavail = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
295 sbp->f_files = (fxdr_unsigned(int32_t,
296 sfp->sf_tfiles.nfsuquad[1]) & 0x7fffffff);
297 sbp->f_ffree = (fxdr_unsigned(int32_t,
298 sfp->sf_ffiles.nfsuquad[1]) & 0x7fffffff);
299 } else {
300 sbp->f_bsize = fxdr_unsigned(int32_t, sfp->sf_bsize);
301 sbp->f_blocks = fxdr_unsigned(int32_t, sfp->sf_blocks);
302 sbp->f_bfree = fxdr_unsigned(int32_t, sfp->sf_bfree);
303 sbp->f_bavail = fxdr_unsigned(int32_t, sfp->sf_bavail);
304 sbp->f_files = 0;
305 sbp->f_ffree = 0;
306 }
307 if (sbp != &mp->mnt_stat) {
308 sbp->f_type = mp->mnt_vfc->vfc_typenum;
309 bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
310 bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
311 }
312 nfsm_reqdone;
313 vput(vp);
314 crfree(cred);
315 return (error);
316}
317
318/*
319 * nfs version 3 fsinfo rpc call
320 */
321int
dadab5e9
MD
322nfs_fsinfo(struct nfsmount *nmp, struct vnode *vp,
323 struct ucred *cred, struct thread *td)
984263bc
MD
324{
325 register struct nfsv3_fsinfo *fsp;
326 register caddr_t cp;
327 register int32_t t1, t2;
328 register u_int32_t *tl, pref, max;
329 caddr_t bpos, dpos, cp2;
330 int error = 0, retattr;
331 struct mbuf *mreq, *mrep, *md, *mb, *mb2;
332 u_int64_t maxfsize;
333
334 nfsstats.rpccnt[NFSPROC_FSINFO]++;
335 nfsm_reqhead(vp, NFSPROC_FSINFO, NFSX_FH(1));
336 nfsm_fhtom(vp, 1);
dadab5e9 337 nfsm_request(vp, NFSPROC_FSINFO, td, cred);
984263bc
MD
338 nfsm_postop_attr(vp, retattr);
339 if (!error) {
340 nfsm_dissect(fsp, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
341 pref = fxdr_unsigned(u_int32_t, fsp->fs_wtpref);
342 if (pref < nmp->nm_wsize && pref >= NFS_FABLKSIZE)
343 nmp->nm_wsize = (pref + NFS_FABLKSIZE - 1) &
344 ~(NFS_FABLKSIZE - 1);
345 max = fxdr_unsigned(u_int32_t, fsp->fs_wtmax);
346 if (max < nmp->nm_wsize && max > 0) {
347 nmp->nm_wsize = max & ~(NFS_FABLKSIZE - 1);
348 if (nmp->nm_wsize == 0)
349 nmp->nm_wsize = max;
350 }
351 pref = fxdr_unsigned(u_int32_t, fsp->fs_rtpref);
352 if (pref < nmp->nm_rsize && pref >= NFS_FABLKSIZE)
353 nmp->nm_rsize = (pref + NFS_FABLKSIZE - 1) &
354 ~(NFS_FABLKSIZE - 1);
355 max = fxdr_unsigned(u_int32_t, fsp->fs_rtmax);
356 if (max < nmp->nm_rsize && max > 0) {
357 nmp->nm_rsize = max & ~(NFS_FABLKSIZE - 1);
358 if (nmp->nm_rsize == 0)
359 nmp->nm_rsize = max;
360 }
361 pref = fxdr_unsigned(u_int32_t, fsp->fs_dtpref);
362 if (pref < nmp->nm_readdirsize && pref >= NFS_DIRBLKSIZ)
363 nmp->nm_readdirsize = (pref + NFS_DIRBLKSIZ - 1) &
364 ~(NFS_DIRBLKSIZ - 1);
365 if (max < nmp->nm_readdirsize && max > 0) {
366 nmp->nm_readdirsize = max & ~(NFS_DIRBLKSIZ - 1);
367 if (nmp->nm_readdirsize == 0)
368 nmp->nm_readdirsize = max;
369 }
370 maxfsize = fxdr_hyper(&fsp->fs_maxfilesize);
371 if (maxfsize > 0 && maxfsize < nmp->nm_maxfilesize)
372 nmp->nm_maxfilesize = maxfsize;
373 nmp->nm_state |= NFSSTA_GOTFSINFO;
374 }
375 nfsm_reqdone;
376 return (error);
377}
378
379/*
380 * Mount a remote root fs via. nfs. This depends on the info in the
381 * nfs_diskless structure that has been filled in properly by some primary
382 * bootstrap.
383 * It goes something like this:
384 * - do enough of "ifconfig" by calling ifioctl() so that the system
385 * can talk to the server
386 * - If nfs_diskless.mygateway is filled in, use that address as
387 * a default gateway.
388 * - build the rootfs mount point and call mountnfs() to do the rest.
389 */
390int
391nfs_mountroot(mp)
392 struct mount *mp;
393{
394 struct mount *swap_mp;
395 struct nfsv3_diskless *nd = &nfsv3_diskless;
396 struct socket *so;
397 struct vnode *vp;
dadab5e9 398 struct thread *td = curthread; /* XXX */
984263bc
MD
399 int error, i;
400 u_long l;
401 char buf[128];
402
403#if defined(BOOTP_NFSROOT) && defined(BOOTP)
404 bootpc_init(); /* use bootp to get nfs_diskless filled in */
405#endif
406
407 /*
408 * XXX time must be non-zero when we init the interface or else
409 * the arp code will wedge...
410 */
411 while (time_second == 0)
412 tsleep(&time_second, PZERO+8, "arpkludge", 10);
413
414 if (nfs_diskless_valid==1)
415 nfs_convert_diskless();
416
417 /*
418 * XXX splnet, so networks will receive...
419 */
420 splnet();
421
422#ifdef notyet
423 /* Set up swap credentials. */
424 proc0.p_ucred->cr_uid = ntohl(nd->swap_ucred.cr_uid);
425 proc0.p_ucred->cr_gid = ntohl(nd->swap_ucred.cr_gid);
426 if ((proc0.p_ucred->cr_ngroups = ntohs(nd->swap_ucred.cr_ngroups)) >
427 NGROUPS)
428 proc0.p_ucred->cr_ngroups = NGROUPS;
429 for (i = 0; i < proc0.p_ucred->cr_ngroups; i++)
430 proc0.p_ucred->cr_groups[i] = ntohl(nd->swap_ucred.cr_groups[i]);
431#endif
432
433 /*
434 * Do enough of ifconfig(8) so that the critical net interface can
435 * talk to the server.
436 */
dadab5e9 437 error = socreate(nd->myif.ifra_addr.sa_family, &so, SOCK_DGRAM, 0, td);
984263bc
MD
438 if (error)
439 panic("nfs_mountroot: socreate(%04x): %d",
440 nd->myif.ifra_addr.sa_family, error);
441
442#if 0 /* XXX Bad idea */
443 /*
444 * We might not have been told the right interface, so we pass
445 * over the first ten interfaces of the same kind, until we get
446 * one of them configured.
447 */
448
449 for (i = strlen(nd->myif.ifra_name) - 1;
450 nd->myif.ifra_name[i] >= '0' &&
451 nd->myif.ifra_name[i] <= '9';
452 nd->myif.ifra_name[i] ++) {
dadab5e9 453 error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, td);
984263bc
MD
454 if(!error)
455 break;
456 }
457#endif
dadab5e9 458 error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, td);
984263bc
MD
459 if (error)
460 panic("nfs_mountroot: SIOCAIFADDR: %d", error);
461 soclose(so);
462
463 /*
464 * If the gateway field is filled in, set it as the default route.
465 */
466 if (nd->mygateway.sin_len != 0) {
467 struct sockaddr_in mask, sin;
468
469 bzero((caddr_t)&mask, sizeof(mask));
470 sin = mask;
471 sin.sin_family = AF_INET;
472 sin.sin_len = sizeof(sin);
473 error = rtrequest(RTM_ADD, (struct sockaddr *)&sin,
474 (struct sockaddr *)&nd->mygateway,
475 (struct sockaddr *)&mask,
476 RTF_UP | RTF_GATEWAY, (struct rtentry **)0);
477 if (error)
478 panic("nfs_mountroot: RTM_ADD: %d", error);
479 }
480
481 /*
482 * Create the rootfs mount point.
483 */
484 nd->root_args.fh = nd->root_fh;
485 nd->root_args.fhsize = nd->root_fhsize;
486 l = ntohl(nd->root_saddr.sin_addr.s_addr);
487 snprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s",
488 (l >> 24) & 0xff, (l >> 16) & 0xff,
489 (l >> 8) & 0xff, (l >> 0) & 0xff,nd->root_hostnam);
490 printf("NFS ROOT: %s\n",buf);
491 if ((error = nfs_mountdiskless(buf, "/", MNT_RDONLY,
dadab5e9 492 &nd->root_saddr, &nd->root_args, td, &vp, &mp)) != 0) {
984263bc
MD
493 if (swap_mp) {
494 mp->mnt_vfc->vfc_refcount--;
495 free(swap_mp, M_MOUNT);
496 }
497 return (error);
498 }
499
500 swap_mp = NULL;
501 if (nd->swap_nblks) {
502
503 /* Convert to DEV_BSIZE instead of Kilobyte */
504 nd->swap_nblks *= 2;
505
506 /*
507 * Create a fake mount point just for the swap vnode so that the
508 * swap file can be on a different server from the rootfs.
509 */
510 nd->swap_args.fh = nd->swap_fh;
511 nd->swap_args.fhsize = nd->swap_fhsize;
512 l = ntohl(nd->swap_saddr.sin_addr.s_addr);
513 snprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s",
514 (l >> 24) & 0xff, (l >> 16) & 0xff,
515 (l >> 8) & 0xff, (l >> 0) & 0xff,nd->swap_hostnam);
516 printf("NFS SWAP: %s\n",buf);
517 if ((error = nfs_mountdiskless(buf, "/swap", 0,
dadab5e9 518 &nd->swap_saddr, &nd->swap_args, td, &vp, &swap_mp)) != 0)
984263bc 519 return (error);
dadab5e9 520 vfs_unbusy(swap_mp, td);
984263bc
MD
521
522 VTONFS(vp)->n_size = VTONFS(vp)->n_vattr.va_size =
523 nd->swap_nblks * DEV_BSIZE ;
524
525 /*
526 * Since the swap file is not the root dir of a file system,
527 * hack it to a regular file.
528 */
529 vp->v_type = VREG;
530 vp->v_flag = 0;
531 VREF(vp);
dadab5e9 532 swaponvp(td, vp, NODEV, nd->swap_nblks);
984263bc
MD
533 }
534
535 mp->mnt_flag |= MNT_ROOTFS;
536 mp->mnt_vnodecovered = NULLVP;
537 rootvp = vp;
dadab5e9 538 vfs_unbusy(mp, td);
984263bc
MD
539
540 /*
541 * This is not really an nfs issue, but it is much easier to
542 * set hostname here and then let the "/etc/rc.xxx" files
543 * mount the right /var based upon its preset value.
544 */
545 bcopy(nd->my_hostnam, hostname, MAXHOSTNAMELEN);
546 hostname[MAXHOSTNAMELEN - 1] = '\0';
547 for (i = 0; i < MAXHOSTNAMELEN; i++)
548 if (hostname[i] == '\0')
549 break;
550 inittodr(ntohl(nd->root_time));
551 return (0);
552}
553
554/*
555 * Internal version of mount system call for diskless setup.
556 */
557static int
dadab5e9
MD
558nfs_mountdiskless(char *path, char *which, int mountflag,
559 struct sockaddr_in *sin, struct nfs_args *args, struct thread *td,
560 struct vnode **vpp, struct mount **mpp)
984263bc
MD
561{
562 struct mount *mp;
563 struct sockaddr *nam;
564 int error;
565 int didalloc = 0;
566
567 mp = *mpp;
568
569 if (mp == NULL) {
570 if ((error = vfs_rootmountalloc("nfs", path, &mp)) != 0) {
571 printf("nfs_mountroot: NFS not configured");
572 return (error);
573 }
574 didalloc = 1;
575 }
576
577 mp->mnt_kern_flag = 0;
578 mp->mnt_flag = mountflag;
579 nam = dup_sockaddr((struct sockaddr *)sin, 1);
580 if ((error = mountnfs(args, mp, nam, which, path, vpp)) != 0) {
581 printf("nfs_mountroot: mount %s on %s: %d", path, which, error);
582 mp->mnt_vfc->vfc_refcount--;
dadab5e9 583 vfs_unbusy(mp, td);
984263bc
MD
584 if (didalloc)
585 free(mp, M_MOUNT);
586 FREE(nam, M_SONAME);
587 return (error);
588 }
589 (void) copystr(which, mp->mnt_stat.f_mntonname, MNAMELEN - 1, 0);
590 *mpp = mp;
591 return (0);
592}
593
594static void
595nfs_decode_args(nmp, argp)
596 struct nfsmount *nmp;
597 struct nfs_args *argp;
598{
599 int s;
600 int adjsock;
601 int maxio;
602
603 s = splnet();
604 /*
605 * Silently clear NFSMNT_NOCONN if it's a TCP mount, it makes
606 * no sense in that context.
607 */
608 if (argp->sotype == SOCK_STREAM)
609 nmp->nm_flag &= ~NFSMNT_NOCONN;
610
611 /* Also clear RDIRPLUS if not NFSv3, it crashes some servers */
612 if ((argp->flags & NFSMNT_NFSV3) == 0)
613 nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
614
615 /* Re-bind if rsrvd port requested and wasn't on one */
616 adjsock = !(nmp->nm_flag & NFSMNT_RESVPORT)
617 && (argp->flags & NFSMNT_RESVPORT);
618 /* Also re-bind if we're switching to/from a connected UDP socket */
619 adjsock |= ((nmp->nm_flag & NFSMNT_NOCONN) !=
620 (argp->flags & NFSMNT_NOCONN));
621
622 /* Update flags atomically. Don't change the lock bits. */
623 nmp->nm_flag = argp->flags | nmp->nm_flag;
624 splx(s);
625
626 if ((argp->flags & NFSMNT_TIMEO) && argp->timeo > 0) {
627 nmp->nm_timeo = (argp->timeo * NFS_HZ + 5) / 10;
628 if (nmp->nm_timeo < NFS_MINTIMEO)
629 nmp->nm_timeo = NFS_MINTIMEO;
630 else if (nmp->nm_timeo > NFS_MAXTIMEO)
631 nmp->nm_timeo = NFS_MAXTIMEO;
632 }
633
634 if ((argp->flags & NFSMNT_RETRANS) && argp->retrans > 1) {
635 nmp->nm_retry = argp->retrans;
636 if (nmp->nm_retry > NFS_MAXREXMIT)
637 nmp->nm_retry = NFS_MAXREXMIT;
638 }
639
640 if (argp->flags & NFSMNT_NFSV3) {
641 if (argp->sotype == SOCK_DGRAM)
642 maxio = NFS_MAXDGRAMDATA;
643 else
644 maxio = NFS_MAXDATA;
645 } else
646 maxio = NFS_V2MAXDATA;
647
648 if ((argp->flags & NFSMNT_WSIZE) && argp->wsize > 0) {
649 nmp->nm_wsize = argp->wsize;
650 /* Round down to multiple of blocksize */
651 nmp->nm_wsize &= ~(NFS_FABLKSIZE - 1);
652 if (nmp->nm_wsize <= 0)
653 nmp->nm_wsize = NFS_FABLKSIZE;
654 }
655 if (nmp->nm_wsize > maxio)
656 nmp->nm_wsize = maxio;
657 if (nmp->nm_wsize > MAXBSIZE)
658 nmp->nm_wsize = MAXBSIZE;
659
660 if ((argp->flags & NFSMNT_RSIZE) && argp->rsize > 0) {
661 nmp->nm_rsize = argp->rsize;
662 /* Round down to multiple of blocksize */
663 nmp->nm_rsize &= ~(NFS_FABLKSIZE - 1);
664 if (nmp->nm_rsize <= 0)
665 nmp->nm_rsize = NFS_FABLKSIZE;
666 }
667 if (nmp->nm_rsize > maxio)
668 nmp->nm_rsize = maxio;
669 if (nmp->nm_rsize > MAXBSIZE)
670 nmp->nm_rsize = MAXBSIZE;
671
672 if ((argp->flags & NFSMNT_READDIRSIZE) && argp->readdirsize > 0) {
673 nmp->nm_readdirsize = argp->readdirsize;
674 }
675 if (nmp->nm_readdirsize > maxio)
676 nmp->nm_readdirsize = maxio;
677 if (nmp->nm_readdirsize > nmp->nm_rsize)
678 nmp->nm_readdirsize = nmp->nm_rsize;
679
680 if ((argp->flags & NFSMNT_ACREGMIN) && argp->acregmin >= 0)
681 nmp->nm_acregmin = argp->acregmin;
682 else
683 nmp->nm_acregmin = NFS_MINATTRTIMO;
684 if ((argp->flags & NFSMNT_ACREGMAX) && argp->acregmax >= 0)
685 nmp->nm_acregmax = argp->acregmax;
686 else
687 nmp->nm_acregmax = NFS_MAXATTRTIMO;
688 if ((argp->flags & NFSMNT_ACDIRMIN) && argp->acdirmin >= 0)
689 nmp->nm_acdirmin = argp->acdirmin;
690 else
691 nmp->nm_acdirmin = NFS_MINDIRATTRTIMO;
692 if ((argp->flags & NFSMNT_ACDIRMAX) && argp->acdirmax >= 0)
693 nmp->nm_acdirmax = argp->acdirmax;
694 else
695 nmp->nm_acdirmax = NFS_MAXDIRATTRTIMO;
696 if (nmp->nm_acdirmin > nmp->nm_acdirmax)
697 nmp->nm_acdirmin = nmp->nm_acdirmax;
698 if (nmp->nm_acregmin > nmp->nm_acregmax)
699 nmp->nm_acregmin = nmp->nm_acregmax;
700
701 if ((argp->flags & NFSMNT_MAXGRPS) && argp->maxgrouplist >= 0) {
702 if (argp->maxgrouplist <= NFS_MAXGRPS)
703 nmp->nm_numgrps = argp->maxgrouplist;
704 else
705 nmp->nm_numgrps = NFS_MAXGRPS;
706 }
707 if ((argp->flags & NFSMNT_READAHEAD) && argp->readahead >= 0) {
708 if (argp->readahead <= NFS_MAXRAHEAD)
709 nmp->nm_readahead = argp->readahead;
710 else
711 nmp->nm_readahead = NFS_MAXRAHEAD;
712 }
713 if ((argp->flags & NFSMNT_LEASETERM) && argp->leaseterm >= 2) {
714 if (argp->leaseterm <= NQ_MAXLEASE)
715 nmp->nm_leaseterm = argp->leaseterm;
716 else
717 nmp->nm_leaseterm = NQ_MAXLEASE;
718 }
719 if ((argp->flags & NFSMNT_DEADTHRESH) && argp->deadthresh >= 1) {
720 if (argp->deadthresh <= NQ_NEVERDEAD)
721 nmp->nm_deadthresh = argp->deadthresh;
722 else
723 nmp->nm_deadthresh = NQ_NEVERDEAD;
724 }
725
726 adjsock |= ((nmp->nm_sotype != argp->sotype) ||
727 (nmp->nm_soproto != argp->proto));
728 nmp->nm_sotype = argp->sotype;
729 nmp->nm_soproto = argp->proto;
730
731 if (nmp->nm_so && adjsock) {
732 nfs_safedisconnect(nmp);
733 if (nmp->nm_sotype == SOCK_DGRAM)
734 while (nfs_connect(nmp, (struct nfsreq *)0)) {
735 printf("nfs_args: retrying connect\n");
736 (void) tsleep((caddr_t)&lbolt,
737 PSOCK, "nfscon", 0);
738 }
739 }
740}
741
742/*
743 * VFS Operations.
744 *
745 * mount system call
746 * It seems a bit dumb to copyinstr() the host and path here and then
747 * bcopy() them in mountnfs(), but I wanted to detect errors before
748 * doing the sockargs() call because sockargs() allocates an mbuf and
749 * an error after that means that I have to release the mbuf.
750 */
751/* ARGSUSED */
752static int
dadab5e9
MD
753nfs_mount(struct mount *mp, char *path, caddr_t data,
754 struct nameidata *ndp, struct thread *td)
984263bc
MD
755{
756 int error;
757 struct nfs_args args;
758 struct sockaddr *nam;
759 struct vnode *vp;
760 char pth[MNAMELEN], hst[MNAMELEN];
761 size_t len;
762 u_char nfh[NFSX_V3FHMAX];
763
764 if (path == NULL) {
765 nfs_mountroot(mp);
766 return (0);
767 }
768 error = copyin(data, (caddr_t)&args, sizeof (struct nfs_args));
769 if (error)
770 return (error);
771 if (args.version != NFS_ARGSVERSION) {
772#ifdef COMPAT_PRELITE2
773 /*
774 * If the argument version is unknown, then assume the
775 * caller is a pre-lite2 4.4BSD client and convert its
776 * arguments.
777 */
778 struct onfs_args oargs;
779 error = copyin(data, (caddr_t)&oargs, sizeof (struct onfs_args));
780 if (error)
781 return (error);
782 nfs_convert_oargs(&args,&oargs);
783#else /* !COMPAT_PRELITE2 */
784 return (EPROGMISMATCH);
785#endif /* COMPAT_PRELITE2 */
786 }
787 if (mp->mnt_flag & MNT_UPDATE) {
788 register struct nfsmount *nmp = VFSTONFS(mp);
789
790 if (nmp == NULL)
791 return (EIO);
792 /*
793 * When doing an update, we can't change from or to
794 * v3 and/or nqnfs, or change cookie translation
795 */
796 args.flags = (args.flags &
797 ~(NFSMNT_NFSV3|NFSMNT_NQNFS /*|NFSMNT_XLATECOOKIE*/)) |
798 (nmp->nm_flag &
799 (NFSMNT_NFSV3|NFSMNT_NQNFS /*|NFSMNT_XLATECOOKIE*/));
800 nfs_decode_args(nmp, &args);
801 return (0);
802 }
803
804 /*
805 * Make the nfs_ip_paranoia sysctl serve as the default connection
806 * or no-connection mode for those protocols that support
807 * no-connection mode (the flag will be cleared later for protocols
808 * that do not support no-connection mode). This will allow a client
809 * to receive replies from a different IP then the request was
810 * sent to. Note: default value for nfs_ip_paranoia is 1 (paranoid),
811 * not 0.
812 */
813 if (nfs_ip_paranoia == 0)
814 args.flags |= NFSMNT_NOCONN;
815 if (args.fhsize < 0 || args.fhsize > NFSX_V3FHMAX)
816 return (EINVAL);
817 error = copyin((caddr_t)args.fh, (caddr_t)nfh, args.fhsize);
818 if (error)
819 return (error);
820 error = copyinstr(path, pth, MNAMELEN-1, &len);
821 if (error)
822 return (error);
823 bzero(&pth[len], MNAMELEN - len);
824 error = copyinstr(args.hostname, hst, MNAMELEN-1, &len);
825 if (error)
826 return (error);
827 bzero(&hst[len], MNAMELEN - len);
828 /* sockargs() call must be after above copyin() calls */
829 error = getsockaddr(&nam, (caddr_t)args.addr, args.addrlen);
830 if (error)
831 return (error);
832 args.fh = nfh;
833 error = mountnfs(&args, mp, nam, pth, hst, &vp);
834 return (error);
835}
836
837/*
838 * Common code for mount and mountroot
839 */
840static int
dadab5e9
MD
841mountnfs(struct nfs_args *argp, struct mount *mp, struct sockaddr *nam,
842 char *pth, char *hst, struct vnode **vpp)
984263bc
MD
843{
844 register struct nfsmount *nmp;
845 struct nfsnode *np;
846 int error;
847 struct vattr attrs;
848
849 if (mp->mnt_flag & MNT_UPDATE) {
850 nmp = VFSTONFS(mp);
851 /* update paths, file handles, etc, here XXX */
852 FREE(nam, M_SONAME);
853 return (0);
854 } else {
855 nmp = zalloc(nfsmount_zone);
856 bzero((caddr_t)nmp, sizeof (struct nfsmount));
857 TAILQ_INIT(&nmp->nm_uidlruhead);
858 TAILQ_INIT(&nmp->nm_bufq);
859 mp->mnt_data = (qaddr_t)nmp;
860 }
861 vfs_getnewfsid(mp);
862 nmp->nm_mountp = mp;
863 if (argp->flags & NFSMNT_NQNFS)
864 /*
865 * We have to set mnt_maxsymlink to a non-zero value so
866 * that COMPAT_43 routines will know that we are setting
867 * the d_type field in directories (and can zero it for
868 * unsuspecting binaries).
869 */
870 mp->mnt_maxsymlinklen = 1;
871
872 /*
873 * V2 can only handle 32 bit filesizes. A 4GB-1 limit may be too
874 * high, depending on whether we end up with negative offsets in
875 * the client or server somewhere. 2GB-1 may be safer.
876 *
877 * For V3, nfs_fsinfo will adjust this as necessary. Assume maximum
878 * that we can handle until we find out otherwise.
879 * XXX Our "safe" limit on the client is what we can store in our
880 * buffer cache using signed(!) block numbers.
881 */
882 if ((argp->flags & NFSMNT_NFSV3) == 0)
883 nmp->nm_maxfilesize = 0xffffffffLL;
884 else
885 nmp->nm_maxfilesize = (u_int64_t)0x80000000 * DEV_BSIZE - 1;
886
887 nmp->nm_timeo = NFS_TIMEO;
888 nmp->nm_retry = NFS_RETRANS;
889 nmp->nm_wsize = NFS_WSIZE;
890 nmp->nm_rsize = NFS_RSIZE;
891 nmp->nm_readdirsize = NFS_READDIRSIZE;
892 nmp->nm_numgrps = NFS_MAXGRPS;
893 nmp->nm_readahead = NFS_DEFRAHEAD;
894 nmp->nm_leaseterm = NQ_DEFLEASE;
895 nmp->nm_deadthresh = NQ_DEADTHRESH;
896 CIRCLEQ_INIT(&nmp->nm_timerhead);
897 nmp->nm_inprog = NULLVP;
898 nmp->nm_fhsize = argp->fhsize;
899 bcopy((caddr_t)argp->fh, (caddr_t)nmp->nm_fh, argp->fhsize);
900 bcopy(hst, mp->mnt_stat.f_mntfromname, MNAMELEN);
901 bcopy(pth, mp->mnt_stat.f_mntonname, MNAMELEN);
902 nmp->nm_nam = nam;
903 /* Set up the sockets and per-host congestion */
904 nmp->nm_sotype = argp->sotype;
905 nmp->nm_soproto = argp->proto;
906
907 nfs_decode_args(nmp, argp);
908
909 /*
910 * For Connection based sockets (TCP,...) defer the connect until
911 * the first request, in case the server is not responding.
912 */
913 if (nmp->nm_sotype == SOCK_DGRAM &&
914 (error = nfs_connect(nmp, (struct nfsreq *)0)))
915 goto bad;
916
917 /*
918 * This is silly, but it has to be set so that vinifod() works.
919 * We do not want to do an nfs_statfs() here since we can get
920 * stuck on a dead server and we are holding a lock on the mount
921 * point.
922 */
923 mp->mnt_stat.f_iosize = nfs_iosize(nmp);
924 /*
925 * A reference count is needed on the nfsnode representing the
926 * remote root. If this object is not persistent, then backward
927 * traversals of the mount point (i.e. "..") will not work if
928 * the nfsnode gets flushed out of the cache. Ufs does not have
929 * this problem, because one can identify root inodes by their
930 * number == ROOTINO (2).
931 */
932 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
933 if (error)
934 goto bad;
935 *vpp = NFSTOV(np);
936
937 /*
938 * Get file attributes for the mountpoint. This has the side
939 * effect of filling in (*vpp)->v_type with the correct value.
940 */
dadab5e9 941 VOP_GETATTR(*vpp, &attrs, curproc->p_ucred, curthread);
984263bc
MD
942
943 /*
944 * Lose the lock but keep the ref.
945 */
dadab5e9 946 VOP_UNLOCK(*vpp, 0, curthread);
984263bc
MD
947
948 return (0);
949bad:
950 nfs_disconnect(nmp);
951 zfree(nfsmount_zone, nmp);
952 FREE(nam, M_SONAME);
953 return (error);
954}
955
956/*
957 * unmount system call
958 */
959static int
dadab5e9 960nfs_unmount(struct mount *mp, int mntflags, struct thread *td)
984263bc
MD
961{
962 register struct nfsmount *nmp;
963 int error, flags = 0;
964
965 if (mntflags & MNT_FORCE)
966 flags |= FORCECLOSE;
967 nmp = VFSTONFS(mp);
968 /*
969 * Goes something like this..
970 * - Call vflush() to clear out vnodes for this file system
971 * - Close the socket
972 * - Free up the data structures
973 */
974 /* In the forced case, cancel any outstanding requests. */
975 if (flags & FORCECLOSE) {
976 error = nfs_nmcancelreqs(nmp);
977 if (error)
978 return (error);
979 }
980 /*
981 * Must handshake with nqnfs_clientd() if it is active.
982 */
983 nmp->nm_state |= NFSSTA_DISMINPROG;
984 while (nmp->nm_inprog != NULLVP)
985 (void) tsleep((caddr_t)&lbolt, PSOCK, "nfsdism", 0);
986
987 /* We hold 1 extra ref on the root vnode; see comment in mountnfs(). */
988 error = vflush(mp, 1, flags);
989 if (error) {
990 nmp->nm_state &= ~NFSSTA_DISMINPROG;
991 return (error);
992 }
993
994 /*
995 * We are now committed to the unmount.
996 * For NQNFS, let the server daemon free the nfsmount structure.
997 */
998 if (nmp->nm_flag & (NFSMNT_NQNFS | NFSMNT_KERB))
999 nmp->nm_state |= NFSSTA_DISMNT;
1000
1001 nfs_disconnect(nmp);
1002 FREE(nmp->nm_nam, M_SONAME);
1003
1004 if ((nmp->nm_flag & (NFSMNT_NQNFS | NFSMNT_KERB)) == 0)
1005 zfree(nfsmount_zone, nmp);
1006 return (0);
1007}
1008
1009/*
1010 * Return root of a filesystem
1011 */
1012static int
1013nfs_root(mp, vpp)
1014 struct mount *mp;
1015 struct vnode **vpp;
1016{
1017 register struct vnode *vp;
1018 struct nfsmount *nmp;
1019 struct nfsnode *np;
1020 int error;
1021
1022 nmp = VFSTONFS(mp);
1023 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
1024 if (error)
1025 return (error);
1026 vp = NFSTOV(np);
1027 if (vp->v_type == VNON)
1028 vp->v_type = VDIR;
1029 vp->v_flag = VROOT;
1030 *vpp = vp;
1031 return (0);
1032}
1033
1034extern int syncprt;
1035
1036/*
1037 * Flush out the buffer cache
1038 */
1039/* ARGSUSED */
1040static int
dadab5e9 1041nfs_sync(struct mount *mp, int waitfor, struct ucred *cred, struct thread *td)
984263bc 1042{
dadab5e9 1043 struct vnode *vp;
984263bc
MD
1044 int error, allerror = 0;
1045
1046 /*
1047 * Force stale buffer cache information to be flushed.
1048 */
1049loop:
1050 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
1051 vp != NULL;
1052 vp = TAILQ_NEXT(vp, v_nmntvnodes)) {
1053 /*
1054 * If the vnode that we are about to sync is no longer
1055 * associated with this mount point, start over.
1056 */
1057 if (vp->v_mount != mp)
1058 goto loop;
1059 if (VOP_ISLOCKED(vp, NULL) || TAILQ_EMPTY(&vp->v_dirtyblkhd) ||
1060 waitfor == MNT_LAZY)
1061 continue;
dadab5e9 1062 if (vget(vp, LK_EXCLUSIVE, td))
984263bc 1063 goto loop;
dadab5e9 1064 error = VOP_FSYNC(vp, cred, waitfor, td);
984263bc
MD
1065 if (error)
1066 allerror = error;
1067 vput(vp);
1068 }
1069 return (allerror);
1070}
1071