kernel: Make SMP support default (and non-optional).
[dragonfly.git] / sys / vfs / nfs / nfs_vfsops.c
CommitLineData
c9669adf 1/*
984263bc
MD
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 $
38 */
39
40#include "opt_bootp.h"
0e8ff41a 41#include "opt_nfsroot.h"
984263bc
MD
42
43#include <sys/param.h>
44#include <sys/sockio.h>
45#include <sys/proc.h>
46#include <sys/vnode.h>
9ba76b73 47#include <sys/fcntl.h>
984263bc
MD
48#include <sys/kernel.h>
49#include <sys/sysctl.h>
50#include <sys/malloc.h>
51#include <sys/mount.h>
52#include <sys/mbuf.h>
53#include <sys/socket.h>
54#include <sys/socketvar.h>
55#include <sys/systm.h>
0704a975 56#include <sys/objcache.h>
984263bc
MD
57
58#include <vm/vm.h>
59#include <vm/vm_extern.h>
984263bc
MD
60
61#include <net/if.h>
62#include <net/route.h>
63#include <netinet/in.h>
64
165dba55 65#include <sys/thread2.h>
8684e6f9 66#include <sys/mutex2.h>
165dba55 67
1f2de5d4
MD
68#include "rpcv2.h"
69#include "nfsproto.h"
70#include "nfs.h"
1f2de5d4 71#include "nfsmount.h"
c1cf1e59 72#include "nfsnode.h"
1f2de5d4
MD
73#include "xdr_subs.h"
74#include "nfsm_subs.h"
75#include "nfsdiskless.h"
0e8ff41a 76#include "nfsmountrpc.h"
984263bc 77
c8a30e1e
JS
78extern int nfs_mountroot(struct mount *mp);
79extern void bootpc_init(void);
984263bc 80
66a1ddf5
MD
81extern struct vop_ops nfsv2_vnode_vops;
82extern struct vop_ops nfsv2_fifo_vops;
83extern struct vop_ops nfsv2_spec_vops;
984263bc
MD
84
85MALLOC_DEFINE(M_NFSREQ, "NFS req", "NFS request header");
86MALLOC_DEFINE(M_NFSBIGFH, "NFSV3 bigfh", "NFS version 3 file handle");
87MALLOC_DEFINE(M_NFSD, "NFS daemon", "Nfs server daemon structure");
88MALLOC_DEFINE(M_NFSDIROFF, "NFSV3 diroff", "NFS directory offset data");
89MALLOC_DEFINE(M_NFSRVDESC, "NFSV3 srvdesc", "NFS server socket descriptor");
90MALLOC_DEFINE(M_NFSUID, "NFS uid", "Nfs uid mapping structure");
984263bc
MD
91MALLOC_DEFINE(M_NFSHASH, "NFS hash", "NFS hash tables");
92
0704a975 93struct objcache *nfsmount_objcache;
984263bc
MD
94
95struct nfsstats nfsstats;
96SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW, 0, "NFS filesystem");
093e85dc
SG
97SYSCTL_STRUCT(_vfs_nfs, NFS_NFSSTATS, nfsstats, CTLFLAG_RD, &nfsstats, nfsstats,
98 "Nfs stats structure");
984263bc 99static int nfs_ip_paranoia = 1;
093e85dc
SG
100SYSCTL_INT(_vfs_nfs, OID_AUTO, nfs_ip_paranoia, CTLFLAG_RW, &nfs_ip_paranoia, 0,
101 "Enable no-connection mode for protocols that support no-connection mode");
984263bc
MD
102#ifdef NFS_DEBUG
103int nfs_debug;
104SYSCTL_INT(_vfs_nfs, OID_AUTO, debug, CTLFLAG_RW, &nfs_debug, 0, "");
105#endif
106
eefb74c1
MD
107/*
108 * Tunable to determine the Read/Write unit size. Maximum value
109 * is NFS_MAXDATA. We also default to NFS_MAXDATA.
110 */
111static int nfs_io_size = NFS_MAXDATA;
112SYSCTL_INT(_vfs_nfs, OID_AUTO, nfs_io_size, CTLFLAG_RW,
113 &nfs_io_size, 0, "NFS optimal I/O unit size");
114
a6ee311a
RG
115static void nfs_decode_args (struct nfsmount *nmp,
116 struct nfs_args *argp);
117static int mountnfs (struct nfs_args *,struct mount *,
118 struct sockaddr *,char *,char *,struct vnode **);
119static int nfs_mount ( struct mount *mp, char *path, caddr_t data,
acde96db
MD
120 struct ucred *cred);
121static int nfs_unmount ( struct mount *mp, int mntflags);
a6ee311a
RG
122static int nfs_root ( struct mount *mp, struct vnode **vpp);
123static int nfs_statfs ( struct mount *mp, struct statfs *sbp,
acde96db 124 struct ucred *cred);
2ef35c61
AH
125static int nfs_statvfs(struct mount *mp, struct statvfs *sbp,
126 struct ucred *cred);
87de5057 127static int nfs_sync ( struct mount *mp, int waitfor);
984263bc
MD
128
129/*
130 * nfs vfs operations.
131 */
132static struct vfsops nfs_vfsops = {
43c45e8f
HP
133 .vfs_mount = nfs_mount,
134 .vfs_unmount = nfs_unmount,
135 .vfs_root = nfs_root,
136 .vfs_statfs = nfs_statfs,
2ef35c61 137 .vfs_statvfs = nfs_statvfs,
43c45e8f
HP
138 .vfs_sync = nfs_sync,
139 .vfs_init = nfs_init,
140 .vfs_uninit = nfs_uninit
984263bc
MD
141};
142VFS_SET(nfs_vfsops, nfs, VFCF_NETWORK);
e5e63c20 143MODULE_VERSION(nfs, 1);
984263bc
MD
144
145/*
146 * This structure must be filled in by a primary bootstrap or bootstrap
147 * server for a diskless/dataless machine. It is initialized below just
148 * to ensure that it is allocated to initialized data (.data not .bss).
149 */
150struct nfs_diskless nfs_diskless = { { { 0 } } };
151struct nfsv3_diskless nfsv3_diskless = { { { 0 } } };
152int nfs_diskless_valid = 0;
153
2ef35c61 154SYSCTL_INT(_vfs_nfs, OID_AUTO, diskless_valid, CTLFLAG_RD,
093e85dc
SG
155 &nfs_diskless_valid, 0,
156 "NFS diskless params were obtained");
984263bc
MD
157
158SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_rootpath, CTLFLAG_RD,
093e85dc
SG
159 nfsv3_diskless.root_hostnam, 0,
160 "Host name for mount point");
984263bc
MD
161
162SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_rootaddr, CTLFLAG_RD,
163 &nfsv3_diskless.root_saddr, sizeof nfsv3_diskless.root_saddr,
093e85dc 164 "%Ssockaddr_in", "Address of root server");
984263bc
MD
165
166SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_swappath, CTLFLAG_RD,
093e85dc
SG
167 nfsv3_diskless.swap_hostnam, 0,
168 "Host name for mount ppoint");
984263bc
MD
169
170SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_swapaddr, CTLFLAG_RD,
2ef35c61 171 &nfsv3_diskless.swap_saddr, sizeof nfsv3_diskless.swap_saddr,
093e85dc 172 "%Ssockaddr_in", "Address of swap server");
984263bc
MD
173
174
a6ee311a
RG
175void nfsargs_ntoh (struct nfs_args *);
176static int nfs_mountdiskless (char *, char *, int,
984263bc 177 struct sockaddr_in *, struct nfs_args *,
dadab5e9 178 struct thread *, struct vnode **,
a6ee311a
RG
179 struct mount **);
180static void nfs_convert_diskless (void);
181static void nfs_convert_oargs (struct nfs_args *args,
182 struct onfs_args *oargs);
984263bc 183
eefb74c1
MD
184/*
185 * Calculate the buffer I/O block size to use. The maximum V2 block size
186 * is typically 8K, the maximum datagram size is typically 16K, and the
187 * maximum V3 block size is typically 32K. The buffer cache tends to work
188 * best with 16K blocks but we allow 32K for TCP connections.
189 *
190 * We force the block size to be at least a page for buffer cache efficiency.
191 */
3fcb1ab8 192static int
eefb74c1 193nfs_iosize(int v3, int sotype)
984263bc
MD
194{
195 int iosize;
eefb74c1 196 int iomax;
984263bc 197
eefb74c1
MD
198 if (v3) {
199 if (sotype == SOCK_STREAM)
200 iomax = NFS_MAXDATA;
201 else
202 iomax = NFS_MAXDGRAMDATA;
203 } else {
204 iomax = NFS_V2MAXDATA;
205 }
206 if ((iosize = nfs_io_size) > iomax)
207 iosize = iomax;
208 if (iosize < PAGE_SIZE)
209 iosize = PAGE_SIZE;
8d429613
MD
210
211 /*
212 * This is an aweful hack but until the buffer cache is rewritten
213 * we need it. The problem is that when you combine write() with
214 * mmap() the vm_page->valid bits can become weird looking
215 * (e.g. 0xfc). This occurs because NFS uses piecemeal buffers
216 * at the file EOF. To solve the problem the BIO system needs to
217 * be guarenteed that the NFS iosize for regular files will be a
218 * multiple of PAGE_SIZE so it can invalidate the whole page
219 * rather then just the piece of it owned by the buffer when
220 * NFS does vinvalbuf() calls.
221 */
222 if (iosize & PAGE_MASK)
223 iosize = (iosize & ~PAGE_MASK) + PAGE_SIZE;
984263bc
MD
224 return iosize;
225}
226
227static void
3fcb1ab8 228nfs_convert_oargs(struct nfs_args *args, struct onfs_args *oargs)
984263bc
MD
229{
230 args->version = NFS_ARGSVERSION;
231 args->addr = oargs->addr;
232 args->addrlen = oargs->addrlen;
233 args->sotype = oargs->sotype;
234 args->proto = oargs->proto;
235 args->fh = oargs->fh;
236 args->fhsize = oargs->fhsize;
237 args->flags = oargs->flags;
238 args->wsize = oargs->wsize;
239 args->rsize = oargs->rsize;
240 args->readdirsize = oargs->readdirsize;
241 args->timeo = oargs->timeo;
242 args->retrans = oargs->retrans;
243 args->maxgrouplist = oargs->maxgrouplist;
244 args->readahead = oargs->readahead;
984263bc
MD
245 args->deadthresh = oargs->deadthresh;
246 args->hostname = oargs->hostname;
247}
248
249static void
3fcb1ab8 250nfs_convert_diskless(void)
984263bc 251{
0e8ff41a
MD
252 int i;
253
984263bc
MD
254 bcopy(&nfs_diskless.myif, &nfsv3_diskless.myif,
255 sizeof(struct ifaliasreq));
256 bcopy(&nfs_diskless.mygateway, &nfsv3_diskless.mygateway,
257 sizeof(struct sockaddr_in));
226341c1 258 nfs_convert_oargs(&nfsv3_diskless.swap_args, &nfs_diskless.swap_args);
0e8ff41a 259
226341c1
MD
260 /*
261 * Copy the NFS handle passed from the diskless code.
262 *
263 * XXX CURRENTLY DISABLED - bootp passes us a NFSv2 handle which
264 * will fail utterly with HAMMER due to limitations with NFSv2
265 * directory cookies.
266 */
267 bcopy(nfs_diskless.swap_fh, nfsv3_diskless.swap_fh, NFSX_V2FH);
0e8ff41a
MD
268 nfsv3_diskless.swap_fhsize = NFSX_V2FH;
269 for (i = NFSX_V2FH - 1; i >= 0; --i) {
270 if (nfs_diskless.swap_fh[i])
271 break;
272 }
273 if (i < 0)
274 nfsv3_diskless.swap_fhsize = 0;
226341c1 275 nfsv3_diskless.swap_fhsize = 0; /* FORCE DISABLE */
0e8ff41a 276
984263bc
MD
277 bcopy(&nfs_diskless.swap_saddr,&nfsv3_diskless.swap_saddr,
278 sizeof(struct sockaddr_in));
279 bcopy(nfs_diskless.swap_hostnam,nfsv3_diskless.swap_hostnam, MNAMELEN);
280 nfsv3_diskless.swap_nblks = nfs_diskless.swap_nblks;
281 bcopy(&nfs_diskless.swap_ucred, &nfsv3_diskless.swap_ucred,
282 sizeof(struct ucred));
226341c1 283 nfs_convert_oargs(&nfsv3_diskless.root_args, &nfs_diskless.root_args);
0e8ff41a 284
226341c1
MD
285 /*
286 * Copy the NFS handle passed from the diskless code.
287 *
288 * XXX CURRENTLY DISABLED - bootp passes us a NFSv2 handle which
289 * will fail utterly with HAMMER due to limitations with NFSv2
290 * directory cookies.
291 */
292 bcopy(nfs_diskless.root_fh, nfsv3_diskless.root_fh, NFSX_V2FH);
0e8ff41a
MD
293 nfsv3_diskless.root_fhsize = NFSX_V2FH;
294 for (i = NFSX_V2FH - 1; i >= 0; --i) {
295 if (nfs_diskless.root_fh[i])
296 break;
297 }
298 if (i < 0)
299 nfsv3_diskless.root_fhsize = 0;
226341c1 300 nfsv3_diskless.root_fhsize = 0; /* FORCE DISABLE */
0e8ff41a 301
984263bc
MD
302 bcopy(&nfs_diskless.root_saddr,&nfsv3_diskless.root_saddr,
303 sizeof(struct sockaddr_in));
304 bcopy(nfs_diskless.root_hostnam,nfsv3_diskless.root_hostnam, MNAMELEN);
305 nfsv3_diskless.root_time = nfs_diskless.root_time;
306 bcopy(nfs_diskless.my_hostnam,nfsv3_diskless.my_hostnam,
307 MAXHOSTNAMELEN);
308 nfs_diskless_valid = 3;
309}
310
311/*
312 * nfs statfs call
313 */
314int
acde96db 315nfs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
984263bc 316{
40393ded
RG
317 struct vnode *vp;
318 struct nfs_statfs *sfp;
984263bc 319 struct nfsmount *nmp = VFSTONFS(mp);
acde96db 320 thread_t td = curthread;
42edf14f 321 int error = 0, retattr;
984263bc
MD
322 struct nfsnode *np;
323 u_quad_t tquad;
42edf14f
MD
324 struct nfsm_info info;
325
326 info.mrep = NULL;
327 info.v3 = (nmp->nm_flag & NFSMNT_NFSV3);
984263bc 328
c6b43e93
MD
329 lwkt_gettoken(&nmp->nm_token);
330
984263bc 331#ifndef nolint
60233e58 332 sfp = NULL;
984263bc
MD
333#endif
334 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
c6b43e93
MD
335 if (error) {
336 lwkt_reltoken(&nmp->nm_token);
984263bc 337 return (error);
c6b43e93 338 }
984263bc 339 vp = NFSTOV(np);
acde96db 340 /* ignore the passed cred */
984263bc
MD
341 cred = crget();
342 cred->cr_ngroups = 1;
42edf14f 343 if (info.v3 && (nmp->nm_state & NFSSTA_GOTFSINFO) == 0)
3b568787 344 (void)nfs_fsinfo(nmp, vp, td);
984263bc 345 nfsstats.rpccnt[NFSPROC_FSSTAT]++;
42edf14f
MD
346 nfsm_reqhead(&info, vp, NFSPROC_FSSTAT, NFSX_FH(info.v3));
347 ERROROUT(nfsm_fhtom(&info, vp));
348 NEGKEEPOUT(nfsm_request(&info, vp, NFSPROC_FSSTAT, td, cred, &error));
349 if (info.v3) {
350 ERROROUT(nfsm_postop_attr(&info, vp, &retattr,
351 NFS_LATTR_NOSHRINK));
352 }
984263bc 353 if (error) {
42edf14f
MD
354 if (info.mrep != NULL)
355 m_freem(info.mrep);
984263bc
MD
356 goto nfsmout;
357 }
42edf14f 358 NULLOUT(sfp = nfsm_dissect(&info, NFSX_STATFS(info.v3)));
984263bc 359 sbp->f_flags = nmp->nm_flag;
eefb74c1 360
42edf14f 361 if (info.v3) {
984263bc
MD
362 sbp->f_bsize = NFS_FABLKSIZE;
363 tquad = fxdr_hyper(&sfp->sf_tbytes);
364 sbp->f_blocks = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
365 tquad = fxdr_hyper(&sfp->sf_fbytes);
366 sbp->f_bfree = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
367 tquad = fxdr_hyper(&sfp->sf_abytes);
368 sbp->f_bavail = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
369 sbp->f_files = (fxdr_unsigned(int32_t,
370 sfp->sf_tfiles.nfsuquad[1]) & 0x7fffffff);
371 sbp->f_ffree = (fxdr_unsigned(int32_t,
372 sfp->sf_ffiles.nfsuquad[1]) & 0x7fffffff);
373 } else {
374 sbp->f_bsize = fxdr_unsigned(int32_t, sfp->sf_bsize);
375 sbp->f_blocks = fxdr_unsigned(int32_t, sfp->sf_blocks);
376 sbp->f_bfree = fxdr_unsigned(int32_t, sfp->sf_bfree);
377 sbp->f_bavail = fxdr_unsigned(int32_t, sfp->sf_bavail);
378 sbp->f_files = 0;
379 sbp->f_ffree = 0;
380 }
b672fc16
MD
381
382 /*
383 * Some values are pre-set in mnt_stat. Note in particular f_iosize
384 * cannot be changed once the filesystem is mounted as it is used
385 * as the basis for BIOs.
386 */
984263bc
MD
387 if (sbp != &mp->mnt_stat) {
388 sbp->f_type = mp->mnt_vfc->vfc_typenum;
984263bc 389 bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
b672fc16 390 sbp->f_iosize = mp->mnt_stat.f_iosize;
984263bc 391 }
42edf14f
MD
392 m_freem(info.mrep);
393 info.mrep = NULL;
6b08710e 394nfsmout:
984263bc
MD
395 vput(vp);
396 crfree(cred);
c6b43e93 397 lwkt_reltoken(&nmp->nm_token);
984263bc
MD
398 return (error);
399}
400
2ef35c61
AH
401static int
402nfs_statvfs(struct mount *mp, struct statvfs *sbp, struct ucred *cred)
403{
404 struct vnode *vp;
405 struct nfs_statfs *sfp;
406 struct nfsmount *nmp = VFSTONFS(mp);
407 thread_t td = curthread;
408 int error = 0, retattr;
409 struct nfsnode *np;
410 struct nfsm_info info;
411
412 info.mrep = NULL;
413 info.v3 = (nmp->nm_flag & NFSMNT_NFSV3);
c6b43e93 414 lwkt_gettoken(&nmp->nm_token);
2ef35c61
AH
415
416#ifndef nolint
417 sfp = NULL;
418#endif
419 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
c6b43e93
MD
420 if (error) {
421 lwkt_reltoken(&nmp->nm_token);
2ef35c61 422 return (error);
c6b43e93 423 }
2ef35c61
AH
424 vp = NFSTOV(np);
425 /* ignore the passed cred */
426 cred = crget();
427 cred->cr_ngroups = 1;
428 if (info.v3 && (nmp->nm_state & NFSSTA_GOTFSINFO) == 0)
429 (void)nfs_fsinfo(nmp, vp, td);
430 nfsstats.rpccnt[NFSPROC_FSSTAT]++;
431 nfsm_reqhead(&info, vp, NFSPROC_FSSTAT, NFSX_FH(info.v3));
432 ERROROUT(nfsm_fhtom(&info, vp));
433 NEGKEEPOUT(nfsm_request(&info, vp, NFSPROC_FSSTAT, td, cred, &error));
434 if (info.v3) {
435 ERROROUT(nfsm_postop_attr(&info, vp, &retattr,
436 NFS_LATTR_NOSHRINK));
437 }
438 if (error) {
439 if (info.mrep != NULL)
440 m_freem(info.mrep);
441 goto nfsmout;
442 }
443 NULLOUT(sfp = nfsm_dissect(&info, NFSX_STATFS(info.v3)));
444 sbp->f_flag = nmp->nm_flag;
445 sbp->f_owner = nmp->nm_cred->cr_ruid;
446
447 if (info.v3) {
448 sbp->f_bsize = NFS_FABLKSIZE;
449 sbp->f_frsize = NFS_FABLKSIZE;
450 sbp->f_blocks = (fxdr_hyper(&sfp->sf_tbytes) /
451 ((u_quad_t)NFS_FABLKSIZE));
452 sbp->f_bfree = (fxdr_hyper(&sfp->sf_fbytes) /
453 ((u_quad_t)NFS_FABLKSIZE));
454 sbp->f_bavail = (fxdr_hyper(&sfp->sf_abytes) /
455 ((u_quad_t)NFS_FABLKSIZE));
456 sbp->f_files = fxdr_hyper(&sfp->sf_tfiles);
457 sbp->f_ffree = fxdr_hyper(&sfp->sf_ffiles);
458 sbp->f_favail = fxdr_hyper(&sfp->sf_afiles);
459 } else {
460 sbp->f_bsize = fxdr_unsigned(int32_t, sfp->sf_bsize);
461 sbp->f_blocks = fxdr_unsigned(int32_t, sfp->sf_blocks);
462 sbp->f_bfree = fxdr_unsigned(int32_t, sfp->sf_bfree);
463 sbp->f_bavail = fxdr_unsigned(int32_t, sfp->sf_bavail);
464 sbp->f_files = 0;
465 sbp->f_ffree = 0;
466 sbp->f_favail = 0;
467 }
468 sbp->f_syncreads = 0;
469 sbp->f_syncwrites = 0;
470 sbp->f_asyncreads = 0;
471 sbp->f_asyncwrites = 0;
472 sbp->f_type = mp->mnt_vfc->vfc_typenum;
473
474 m_freem(info.mrep);
475 info.mrep = NULL;
476nfsmout:
477 vput(vp);
478 crfree(cred);
c6b43e93 479 lwkt_reltoken(&nmp->nm_token);
2ef35c61
AH
480 return (error);
481}
482
984263bc
MD
483/*
484 * nfs version 3 fsinfo rpc call
485 */
486int
3b568787 487nfs_fsinfo(struct nfsmount *nmp, struct vnode *vp, struct thread *td)
984263bc 488{
40393ded 489 struct nfsv3_fsinfo *fsp;
42edf14f 490 u_int32_t pref, max;
984263bc 491 int error = 0, retattr;
984263bc 492 u_int64_t maxfsize;
42edf14f 493 struct nfsm_info info;
984263bc 494
42edf14f 495 info.v3 = 1;
984263bc 496 nfsstats.rpccnt[NFSPROC_FSINFO]++;
42edf14f
MD
497 nfsm_reqhead(&info, vp, NFSPROC_FSINFO, NFSX_FH(1));
498 ERROROUT(nfsm_fhtom(&info, vp));
499 NEGKEEPOUT(nfsm_request(&info, vp, NFSPROC_FSINFO, td,
500 nfs_vpcred(vp, ND_READ), &error));
501 ERROROUT(nfsm_postop_attr(&info, vp, &retattr, NFS_LATTR_NOSHRINK));
502 if (error == 0) {
503 NULLOUT(fsp = nfsm_dissect(&info, NFSX_V3FSINFO));
984263bc
MD
504 pref = fxdr_unsigned(u_int32_t, fsp->fs_wtpref);
505 if (pref < nmp->nm_wsize && pref >= NFS_FABLKSIZE)
506 nmp->nm_wsize = (pref + NFS_FABLKSIZE - 1) &
507 ~(NFS_FABLKSIZE - 1);
508 max = fxdr_unsigned(u_int32_t, fsp->fs_wtmax);
509 if (max < nmp->nm_wsize && max > 0) {
510 nmp->nm_wsize = max & ~(NFS_FABLKSIZE - 1);
511 if (nmp->nm_wsize == 0)
512 nmp->nm_wsize = max;
513 }
514 pref = fxdr_unsigned(u_int32_t, fsp->fs_rtpref);
515 if (pref < nmp->nm_rsize && pref >= NFS_FABLKSIZE)
516 nmp->nm_rsize = (pref + NFS_FABLKSIZE - 1) &
517 ~(NFS_FABLKSIZE - 1);
518 max = fxdr_unsigned(u_int32_t, fsp->fs_rtmax);
519 if (max < nmp->nm_rsize && max > 0) {
520 nmp->nm_rsize = max & ~(NFS_FABLKSIZE - 1);
521 if (nmp->nm_rsize == 0)
522 nmp->nm_rsize = max;
523 }
524 pref = fxdr_unsigned(u_int32_t, fsp->fs_dtpref);
525 if (pref < nmp->nm_readdirsize && pref >= NFS_DIRBLKSIZ)
526 nmp->nm_readdirsize = (pref + NFS_DIRBLKSIZ - 1) &
527 ~(NFS_DIRBLKSIZ - 1);
528 if (max < nmp->nm_readdirsize && max > 0) {
529 nmp->nm_readdirsize = max & ~(NFS_DIRBLKSIZ - 1);
530 if (nmp->nm_readdirsize == 0)
531 nmp->nm_readdirsize = max;
532 }
533 maxfsize = fxdr_hyper(&fsp->fs_maxfilesize);
534 if (maxfsize > 0 && maxfsize < nmp->nm_maxfilesize)
535 nmp->nm_maxfilesize = maxfsize;
536 nmp->nm_state |= NFSSTA_GOTFSINFO;
b672fc16
MD
537
538 /*
539 * Use the smaller of rsize/wsize for the biosize.
540 */
541 if (nmp->nm_rsize < nmp->nm_wsize)
542 nmp->nm_mountp->mnt_stat.f_iosize = nmp->nm_rsize;
543 else
544 nmp->nm_mountp->mnt_stat.f_iosize = nmp->nm_wsize;
984263bc 545 }
42edf14f
MD
546 m_freem(info.mrep);
547 info.mrep = NULL;
6b08710e 548nfsmout:
984263bc
MD
549 return (error);
550}
551
552/*
553 * Mount a remote root fs via. nfs. This depends on the info in the
554 * nfs_diskless structure that has been filled in properly by some primary
555 * bootstrap.
556 * It goes something like this:
557 * - do enough of "ifconfig" by calling ifioctl() so that the system
558 * can talk to the server
559 * - If nfs_diskless.mygateway is filled in, use that address as
560 * a default gateway.
561 * - build the rootfs mount point and call mountnfs() to do the rest.
562 */
563int
acde96db 564nfs_mountroot(struct mount *mp)
984263bc
MD
565{
566 struct mount *swap_mp;
567 struct nfsv3_diskless *nd = &nfsv3_diskless;
568 struct socket *so;
569 struct vnode *vp;
dadab5e9 570 struct thread *td = curthread; /* XXX */
984263bc
MD
571 int error, i;
572 u_long l;
573 char buf[128];
574
575#if defined(BOOTP_NFSROOT) && defined(BOOTP)
576 bootpc_init(); /* use bootp to get nfs_diskless filled in */
577#endif
578
579 /*
580 * XXX time must be non-zero when we init the interface or else
581 * the arp code will wedge...
582 */
3a6f9faf
MD
583 while (mycpu->gd_time_seconds == 0)
584 tsleep(mycpu, 0, "arpkludge", 10);
984263bc 585
a2103e90
MD
586 /*
587 * The boot code may have passed us a diskless structure.
588 */
226341c1 589 kprintf("DISKLESS %d\n", nfs_diskless_valid);
2ef35c61 590 if (nfs_diskless_valid == 1)
0b9984b8
MD
591 nfs_convert_diskless();
592
226341c1
MD
593 /*
594 * NFSv3 is required.
595 */
596 nd->root_args.flags |= NFSMNT_NFSV3 | NFSMNT_RDIRPLUS;
597 nd->swap_args.flags |= NFSMNT_NFSV3;
598
a2103e90 599#define SINP(sockaddr) ((struct sockaddr_in *)(sockaddr))
086c1d7e 600 kprintf("nfs_mountroot: interface %s ip %s",
2ef35c61 601 nd->myif.ifra_name,
a2103e90 602 inet_ntoa(SINP(&nd->myif.ifra_addr)->sin_addr));
2ef35c61 603 kprintf(" bcast %s",
a2103e90 604 inet_ntoa(SINP(&nd->myif.ifra_broadaddr)->sin_addr));
2ef35c61 605 kprintf(" mask %s\n",
a2103e90
MD
606 inet_ntoa(SINP(&nd->myif.ifra_mask)->sin_addr));
607#undef SINP
984263bc
MD
608
609 /*
610 * XXX splnet, so networks will receive...
611 */
165dba55 612 crit_enter();
984263bc 613
984263bc 614 /*
a2103e90 615 * BOOTP does not necessarily have to be compiled into the kernel
2ef35c61 616 * for an NFS root to work. If we inherited the network
a2103e90
MD
617 * configuration for PXEBOOT then pxe_setup_nfsdiskless() has figured
618 * out our interface for us and all we need to do is ifconfig the
619 * interface. We only do this if the interface has not already been
620 * ifconfig'd by e.g. BOOTP.
984263bc 621 */
dadab5e9 622 error = socreate(nd->myif.ifra_addr.sa_family, &so, SOCK_DGRAM, 0, td);
a2103e90 623 if (error) {
984263bc
MD
624 panic("nfs_mountroot: socreate(%04x): %d",
625 nd->myif.ifra_addr.sa_family, error);
984263bc 626 }
a2103e90 627
87de5057 628 error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, proc0.p_ucred);
984263bc
MD
629 if (error)
630 panic("nfs_mountroot: SIOCAIFADDR: %d", error);
a2103e90 631
9ba76b73 632 soclose(so, FNONBLOCK);
984263bc
MD
633
634 /*
635 * If the gateway field is filled in, set it as the default route.
636 */
637 if (nd->mygateway.sin_len != 0) {
638 struct sockaddr_in mask, sin;
639
640 bzero((caddr_t)&mask, sizeof(mask));
641 sin = mask;
642 sin.sin_family = AF_INET;
643 sin.sin_len = sizeof(sin);
086c1d7e 644 kprintf("nfs_mountroot: gateway %s\n",
0b9984b8 645 inet_ntoa(nd->mygateway.sin_addr));
ecdefdda
MD
646 error = rtrequest_global(RTM_ADD, (struct sockaddr *)&sin,
647 (struct sockaddr *)&nd->mygateway,
648 (struct sockaddr *)&mask,
649 RTF_UP | RTF_GATEWAY);
984263bc 650 if (error)
086c1d7e 651 kprintf("nfs_mountroot: unable to set gateway, error %d, continuing anyway\n", error);
984263bc
MD
652 }
653
654 /*
655 * Create the rootfs mount point.
656 */
657 nd->root_args.fh = nd->root_fh;
658 nd->root_args.fhsize = nd->root_fhsize;
659 l = ntohl(nd->root_saddr.sin_addr.s_addr);
f8c7a42d 660 ksnprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s",
984263bc
MD
661 (l >> 24) & 0xff, (l >> 16) & 0xff,
662 (l >> 8) & 0xff, (l >> 0) & 0xff,nd->root_hostnam);
5dacbb2e 663 kprintf("NFS_ROOT: %s\n",buf);
226341c1
MD
664 error = nfs_mountdiskless(buf, "/", MNT_RDONLY, &nd->root_saddr,
665 &nd->root_args, td, &vp, &mp);
666 if (error) {
5dacbb2e 667 mp->mnt_vfc->vfc_refcount--;
0b9984b8 668 crit_exit();
984263bc
MD
669 return (error);
670 }
671
672 swap_mp = NULL;
673 if (nd->swap_nblks) {
674
675 /* Convert to DEV_BSIZE instead of Kilobyte */
676 nd->swap_nblks *= 2;
677
678 /*
679 * Create a fake mount point just for the swap vnode so that the
680 * swap file can be on a different server from the rootfs.
681 */
682 nd->swap_args.fh = nd->swap_fh;
683 nd->swap_args.fhsize = nd->swap_fhsize;
684 l = ntohl(nd->swap_saddr.sin_addr.s_addr);
f8c7a42d 685 ksnprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s",
984263bc
MD
686 (l >> 24) & 0xff, (l >> 16) & 0xff,
687 (l >> 8) & 0xff, (l >> 0) & 0xff,nd->swap_hostnam);
086c1d7e 688 kprintf("NFS SWAP: %s\n",buf);
226341c1
MD
689 error = nfs_mountdiskless(buf, "/swap", 0, &nd->swap_saddr,
690 &nd->swap_args, td, &vp, &swap_mp);
691 if (error) {
0b9984b8 692 crit_exit();
984263bc 693 return (error);
0b9984b8 694 }
f9642f56 695 vfs_unbusy(swap_mp);
984263bc 696
2ef35c61 697 VTONFS(vp)->n_size = VTONFS(vp)->n_vattr.va_size =
984263bc 698 nd->swap_nblks * DEV_BSIZE ;
2ef35c61 699
984263bc
MD
700 /*
701 * Since the swap file is not the root dir of a file system,
702 * hack it to a regular file.
703 */
2247fe02 704 vclrflags(vp, VROOT);
597aea93 705 vref(vp);
1c843a13 706 nfs_setvtype(vp, VREG);
e4c9c0c8 707 swaponvp(td, vp, nd->swap_nblks);
984263bc
MD
708 }
709
710 mp->mnt_flag |= MNT_ROOTFS;
984263bc
MD
711
712 /*
713 * This is not really an nfs issue, but it is much easier to
714 * set hostname here and then let the "/etc/rc.xxx" files
715 * mount the right /var based upon its preset value.
716 */
717 bcopy(nd->my_hostnam, hostname, MAXHOSTNAMELEN);
718 hostname[MAXHOSTNAMELEN - 1] = '\0';
719 for (i = 0; i < MAXHOSTNAMELEN; i++)
720 if (hostname[i] == '\0')
721 break;
722 inittodr(ntohl(nd->root_time));
0b9984b8 723 crit_exit();
984263bc
MD
724 return (0);
725}
726
727/*
728 * Internal version of mount system call for diskless setup.
729 */
730static int
dadab5e9
MD
731nfs_mountdiskless(char *path, char *which, int mountflag,
732 struct sockaddr_in *sin, struct nfs_args *args, struct thread *td,
733 struct vnode **vpp, struct mount **mpp)
984263bc
MD
734{
735 struct mount *mp;
736 struct sockaddr *nam;
984263bc 737 int didalloc = 0;
0e8ff41a 738 int error;
984263bc
MD
739
740 mp = *mpp;
741
742 if (mp == NULL) {
743 if ((error = vfs_rootmountalloc("nfs", path, &mp)) != 0) {
086c1d7e 744 kprintf("nfs_mountroot: NFS not configured");
984263bc
MD
745 return (error);
746 }
747 didalloc = 1;
748 }
984263bc
MD
749 mp->mnt_kern_flag = 0;
750 mp->mnt_flag = mountflag;
cfa2ba21 751 nam = dup_sockaddr((struct sockaddr *)sin);
0e8ff41a
MD
752
753#if defined(BOOTP) || defined(NFS_ROOT)
754 if (args->fhsize == 0) {
df0a038e
SW
755 char *xpath = path;
756
226341c1
MD
757 kprintf("NFS_ROOT: No FH passed from loader, attempting "
758 "mount rpc...");
5dacbb2e
MD
759 while (*xpath && *xpath != ':')
760 ++xpath;
761 if (*xpath)
762 ++xpath;
0e8ff41a 763 args->fhsize = 0;
5dacbb2e 764 error = md_mount(sin, xpath, args->fh, &args->fhsize, args, td);
0e8ff41a 765 if (error) {
5dacbb2e 766 kprintf("failed error %d.\n", error);
0e8ff41a
MD
767 goto haderror;
768 }
086c1d7e 769 kprintf("success!\n");
0e8ff41a
MD
770 }
771#endif
772
984263bc 773 if ((error = mountnfs(args, mp, nam, which, path, vpp)) != 0) {
04ef15fb 774#if defined(BOOTP) || defined(NFS_ROOT)
0e8ff41a 775haderror:
04ef15fb 776#endif
086c1d7e 777 kprintf("nfs_mountroot: mount %s on %s: %d", path, which, error);
984263bc 778 mp->mnt_vfc->vfc_refcount--;
984263bc 779 if (didalloc)
efda3bd0 780 kfree(mp, M_MOUNT);
884717e1 781 kfree(nam, M_SONAME);
984263bc
MD
782 return (error);
783 }
984263bc
MD
784 *mpp = mp;
785 return (0);
786}
787
788static void
3fcb1ab8 789nfs_decode_args(struct nfsmount *nmp, struct nfs_args *argp)
984263bc 790{
984263bc
MD
791 int adjsock;
792 int maxio;
793
165dba55 794 crit_enter();
984263bc
MD
795 /*
796 * Silently clear NFSMNT_NOCONN if it's a TCP mount, it makes
797 * no sense in that context.
798 */
5e6f1ca5 799 if (nmp->nm_sotype == SOCK_STREAM) {
984263bc 800 nmp->nm_flag &= ~NFSMNT_NOCONN;
5e6f1ca5
MD
801 argp->flags &= ~NFSMNT_NOCONN;
802 }
984263bc 803
5e6f1ca5 804 /*
226341c1 805 * readdirplus is NFSv3 only.
5e6f1ca5 806 */
5e6f1ca5 807 if ((argp->flags & NFSMNT_NFSV3) == 0) {
984263bc 808 nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
5e6f1ca5
MD
809 argp->flags &= ~NFSMNT_RDIRPLUS;
810 }
984263bc 811
e20dbf61
MD
812 /*
813 * Re-bind if rsrvd port flag has changed
814 */
815 adjsock = (nmp->nm_flag & NFSMNT_RESVPORT) !=
816 (argp->flags & NFSMNT_RESVPORT);
984263bc
MD
817
818 /* Update flags atomically. Don't change the lock bits. */
819 nmp->nm_flag = argp->flags | nmp->nm_flag;
165dba55 820 crit_exit();
984263bc
MD
821
822 if ((argp->flags & NFSMNT_TIMEO) && argp->timeo > 0) {
823 nmp->nm_timeo = (argp->timeo * NFS_HZ + 5) / 10;
824 if (nmp->nm_timeo < NFS_MINTIMEO)
825 nmp->nm_timeo = NFS_MINTIMEO;
826 else if (nmp->nm_timeo > NFS_MAXTIMEO)
827 nmp->nm_timeo = NFS_MAXTIMEO;
828 }
829
830 if ((argp->flags & NFSMNT_RETRANS) && argp->retrans > 1) {
831 nmp->nm_retry = argp->retrans;
832 if (nmp->nm_retry > NFS_MAXREXMIT)
833 nmp->nm_retry = NFS_MAXREXMIT;
834 }
835
e20dbf61
MD
836 /*
837 * These parameters effect the buffer cache and cannot be changed
838 * once we've successfully mounted.
839 */
840 if ((nmp->nm_state & NFSSTA_GOTFSINFO) == 0) {
841 maxio = nfs_iosize(argp->flags & NFSMNT_NFSV3, nmp->nm_sotype);
842
843 if ((argp->flags & NFSMNT_WSIZE) && argp->wsize > 0) {
844 nmp->nm_wsize = argp->wsize;
845 /* Round down to multiple of blocksize */
846 nmp->nm_wsize &= ~(NFS_FABLKSIZE - 1);
847 if (nmp->nm_wsize <= 0)
848 nmp->nm_wsize = NFS_FABLKSIZE;
849 }
850 if (nmp->nm_wsize > maxio)
851 nmp->nm_wsize = maxio;
852 if (nmp->nm_wsize > MAXBSIZE)
853 nmp->nm_wsize = MAXBSIZE;
854
855 if ((argp->flags & NFSMNT_RSIZE) && argp->rsize > 0) {
856 nmp->nm_rsize = argp->rsize;
857 /* Round down to multiple of blocksize */
858 nmp->nm_rsize &= ~(NFS_FABLKSIZE - 1);
859 if (nmp->nm_rsize <= 0)
860 nmp->nm_rsize = NFS_FABLKSIZE;
861 }
862 if (nmp->nm_rsize > maxio)
863 nmp->nm_rsize = maxio;
864 if (nmp->nm_rsize > MAXBSIZE)
865 nmp->nm_rsize = MAXBSIZE;
866
867 if ((argp->flags & NFSMNT_READDIRSIZE) &&
868 argp->readdirsize > 0) {
869 nmp->nm_readdirsize = argp->readdirsize;
870 }
871 if (nmp->nm_readdirsize > maxio)
872 nmp->nm_readdirsize = maxio;
873 if (nmp->nm_readdirsize > nmp->nm_rsize)
874 nmp->nm_readdirsize = nmp->nm_rsize;
984263bc 875 }
984263bc
MD
876
877 if ((argp->flags & NFSMNT_ACREGMIN) && argp->acregmin >= 0)
878 nmp->nm_acregmin = argp->acregmin;
879 else
880 nmp->nm_acregmin = NFS_MINATTRTIMO;
881 if ((argp->flags & NFSMNT_ACREGMAX) && argp->acregmax >= 0)
882 nmp->nm_acregmax = argp->acregmax;
883 else
884 nmp->nm_acregmax = NFS_MAXATTRTIMO;
885 if ((argp->flags & NFSMNT_ACDIRMIN) && argp->acdirmin >= 0)
886 nmp->nm_acdirmin = argp->acdirmin;
887 else
888 nmp->nm_acdirmin = NFS_MINDIRATTRTIMO;
889 if ((argp->flags & NFSMNT_ACDIRMAX) && argp->acdirmax >= 0)
890 nmp->nm_acdirmax = argp->acdirmax;
891 else
892 nmp->nm_acdirmax = NFS_MAXDIRATTRTIMO;
893 if (nmp->nm_acdirmin > nmp->nm_acdirmax)
894 nmp->nm_acdirmin = nmp->nm_acdirmax;
895 if (nmp->nm_acregmin > nmp->nm_acregmax)
896 nmp->nm_acregmin = nmp->nm_acregmax;
897
898 if ((argp->flags & NFSMNT_MAXGRPS) && argp->maxgrouplist >= 0) {
899 if (argp->maxgrouplist <= NFS_MAXGRPS)
900 nmp->nm_numgrps = argp->maxgrouplist;
901 else
902 nmp->nm_numgrps = NFS_MAXGRPS;
903 }
904 if ((argp->flags & NFSMNT_READAHEAD) && argp->readahead >= 0) {
905 if (argp->readahead <= NFS_MAXRAHEAD)
906 nmp->nm_readahead = argp->readahead;
907 else
908 nmp->nm_readahead = NFS_MAXRAHEAD;
909 }
984263bc 910 if ((argp->flags & NFSMNT_DEADTHRESH) && argp->deadthresh >= 1) {
e07fef60 911 if (argp->deadthresh <= NFS_NEVERDEAD)
984263bc
MD
912 nmp->nm_deadthresh = argp->deadthresh;
913 else
e07fef60 914 nmp->nm_deadthresh = NFS_NEVERDEAD;
984263bc
MD
915 }
916
984263bc
MD
917 if (nmp->nm_so && adjsock) {
918 nfs_safedisconnect(nmp);
919 if (nmp->nm_sotype == SOCK_DGRAM)
60233e58 920 while (nfs_connect(nmp, NULL)) {
086c1d7e 921 kprintf("nfs_args: retrying connect\n");
377d4740 922 (void) tsleep((caddr_t)&lbolt, 0, "nfscon", 0);
984263bc
MD
923 }
924 }
925}
926
927/*
928 * VFS Operations.
929 *
930 * mount system call
931 * It seems a bit dumb to copyinstr() the host and path here and then
932 * bcopy() them in mountnfs(), but I wanted to detect errors before
933 * doing the sockargs() call because sockargs() allocates an mbuf and
934 * an error after that means that I have to release the mbuf.
935 */
936/* ARGSUSED */
937static int
acde96db 938nfs_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
984263bc
MD
939{
940 int error;
941 struct nfs_args args;
942 struct sockaddr *nam;
943 struct vnode *vp;
944 char pth[MNAMELEN], hst[MNAMELEN];
945 size_t len;
946 u_char nfh[NFSX_V3FHMAX];
947
948 if (path == NULL) {
949 nfs_mountroot(mp);
950 return (0);
951 }
952 error = copyin(data, (caddr_t)&args, sizeof (struct nfs_args));
953 if (error)
954 return (error);
955 if (args.version != NFS_ARGSVERSION) {
956#ifdef COMPAT_PRELITE2
957 /*
958 * If the argument version is unknown, then assume the
959 * caller is a pre-lite2 4.4BSD client and convert its
960 * arguments.
961 */
962 struct onfs_args oargs;
963 error = copyin(data, (caddr_t)&oargs, sizeof (struct onfs_args));
964 if (error)
965 return (error);
966 nfs_convert_oargs(&args,&oargs);
967#else /* !COMPAT_PRELITE2 */
968 return (EPROGMISMATCH);
969#endif /* COMPAT_PRELITE2 */
970 }
971 if (mp->mnt_flag & MNT_UPDATE) {
40393ded 972 struct nfsmount *nmp = VFSTONFS(mp);
984263bc
MD
973
974 if (nmp == NULL)
975 return (EIO);
976 /*
977 * When doing an update, we can't change from or to
e20dbf61 978 * v3, or change cookie translation, or rsize or wsize.
984263bc 979 */
e20dbf61
MD
980 args.flags &= ~(NFSMNT_NFSV3 | NFSMNT_RSIZE | NFSMNT_WSIZE);
981 args.flags |= nmp->nm_flag & (NFSMNT_NFSV3);
984263bc
MD
982 nfs_decode_args(nmp, &args);
983 return (0);
984 }
985
986 /*
987 * Make the nfs_ip_paranoia sysctl serve as the default connection
2ef35c61 988 * or no-connection mode for those protocols that support
984263bc
MD
989 * no-connection mode (the flag will be cleared later for protocols
990 * that do not support no-connection mode). This will allow a client
991 * to receive replies from a different IP then the request was
992 * sent to. Note: default value for nfs_ip_paranoia is 1 (paranoid),
993 * not 0.
994 */
995 if (nfs_ip_paranoia == 0)
996 args.flags |= NFSMNT_NOCONN;
997 if (args.fhsize < 0 || args.fhsize > NFSX_V3FHMAX)
998 return (EINVAL);
999 error = copyin((caddr_t)args.fh, (caddr_t)nfh, args.fhsize);
1000 if (error)
1001 return (error);
1002 error = copyinstr(path, pth, MNAMELEN-1, &len);
1003 if (error)
1004 return (error);
1005 bzero(&pth[len], MNAMELEN - len);
1006 error = copyinstr(args.hostname, hst, MNAMELEN-1, &len);
1007 if (error)
1008 return (error);
1009 bzero(&hst[len], MNAMELEN - len);
1010 /* sockargs() call must be after above copyin() calls */
1011 error = getsockaddr(&nam, (caddr_t)args.addr, args.addrlen);
1012 if (error)
1013 return (error);
1014 args.fh = nfh;
1015 error = mountnfs(&args, mp, nam, pth, hst, &vp);
1016 return (error);
1017}
1018
1019/*
1020 * Common code for mount and mountroot
1021 */
1022static int
dadab5e9
MD
1023mountnfs(struct nfs_args *argp, struct mount *mp, struct sockaddr *nam,
1024 char *pth, char *hst, struct vnode **vpp)
984263bc 1025{
40393ded 1026 struct nfsmount *nmp;
984263bc
MD
1027 struct nfsnode *np;
1028 int error;
13ddc895
MD
1029 int rxcpu;
1030 int txcpu;
984263bc
MD
1031
1032 if (mp->mnt_flag & MNT_UPDATE) {
1033 nmp = VFSTONFS(mp);
1034 /* update paths, file handles, etc, here XXX */
884717e1 1035 kfree(nam, M_SONAME);
984263bc
MD
1036 return (0);
1037 } else {
0704a975 1038 nmp = objcache_get(nfsmount_objcache, M_WAITOK);
984263bc 1039 bzero((caddr_t)nmp, sizeof (struct nfsmount));
8684e6f9
MD
1040 mtx_init(&nmp->nm_rxlock);
1041 mtx_init(&nmp->nm_txlock);
984263bc 1042 TAILQ_INIT(&nmp->nm_uidlruhead);
81b5c339 1043 TAILQ_INIT(&nmp->nm_bioq);
e21aec5b 1044 TAILQ_INIT(&nmp->nm_reqq);
edb90c22
MD
1045 TAILQ_INIT(&nmp->nm_reqtxq);
1046 TAILQ_INIT(&nmp->nm_reqrxq);
984263bc 1047 mp->mnt_data = (qaddr_t)nmp;
a3c18566 1048 lwkt_token_init(&nmp->nm_token, "nfs_token");
984263bc
MD
1049 }
1050 vfs_getnewfsid(mp);
1051 nmp->nm_mountp = mp;
c6b43e93 1052 mp->mnt_kern_flag |= MNTK_ALL_MPSAFE;
54c0e7fb 1053 mp->mnt_kern_flag |= MNTK_THR_SYNC;
c6b43e93
MD
1054
1055 lwkt_gettoken(&nmp->nm_token);
984263bc
MD
1056
1057 /*
1058 * V2 can only handle 32 bit filesizes. A 4GB-1 limit may be too
1059 * high, depending on whether we end up with negative offsets in
1060 * the client or server somewhere. 2GB-1 may be safer.
1061 *
1062 * For V3, nfs_fsinfo will adjust this as necessary. Assume maximum
5f73a1d1
MD
1063 * that we can handle until we find out otherwise. Note that seek
1064 * offsets are signed.
984263bc
MD
1065 */
1066 if ((argp->flags & NFSMNT_NFSV3) == 0)
1067 nmp->nm_maxfilesize = 0xffffffffLL;
1068 else
5f73a1d1 1069 nmp->nm_maxfilesize = 0x7fffffffffffffffLL;
984263bc
MD
1070
1071 nmp->nm_timeo = NFS_TIMEO;
1072 nmp->nm_retry = NFS_RETRANS;
eefb74c1
MD
1073 nmp->nm_wsize = nfs_iosize(argp->flags & NFSMNT_NFSV3, argp->sotype);
1074 nmp->nm_rsize = nmp->nm_wsize;
984263bc
MD
1075 nmp->nm_readdirsize = NFS_READDIRSIZE;
1076 nmp->nm_numgrps = NFS_MAXGRPS;
1077 nmp->nm_readahead = NFS_DEFRAHEAD;
0639df29 1078 nmp->nm_deadthresh = NFS_DEADTHRESH;
984263bc
MD
1079 nmp->nm_fhsize = argp->fhsize;
1080 bcopy((caddr_t)argp->fh, (caddr_t)nmp->nm_fh, argp->fhsize);
1081 bcopy(hst, mp->mnt_stat.f_mntfromname, MNAMELEN);
984263bc
MD
1082 nmp->nm_nam = nam;
1083 /* Set up the sockets and per-host congestion */
1084 nmp->nm_sotype = argp->sotype;
1085 nmp->nm_soproto = argp->proto;
3b568787 1086 nmp->nm_cred = crhold(proc0.p_ucred);
984263bc
MD
1087
1088 nfs_decode_args(nmp, argp);
1089
1090 /*
1091 * For Connection based sockets (TCP,...) defer the connect until
1092 * the first request, in case the server is not responding.
1093 */
1094 if (nmp->nm_sotype == SOCK_DGRAM &&
60233e58 1095 (error = nfs_connect(nmp, NULL)))
984263bc
MD
1096 goto bad;
1097
1098 /*
1099 * This is silly, but it has to be set so that vinifod() works.
1100 * We do not want to do an nfs_statfs() here since we can get
1101 * stuck on a dead server and we are holding a lock on the mount
1102 * point.
1103 */
2ef35c61 1104 mp->mnt_stat.f_iosize =
eefb74c1
MD
1105 nfs_iosize(nmp->nm_flag & NFSMNT_NFSV3, nmp->nm_sotype);
1106
0961aa92
MD
1107 /*
1108 * Install vop_ops for our vnops
1109 */
66a1ddf5
MD
1110 vfs_add_vnodeops(mp, &nfsv2_vnode_vops, &mp->mnt_vn_norm_ops);
1111 vfs_add_vnodeops(mp, &nfsv2_spec_vops, &mp->mnt_vn_spec_ops);
1112 vfs_add_vnodeops(mp, &nfsv2_fifo_vops, &mp->mnt_vn_fifo_ops);
0961aa92 1113
984263bc
MD
1114 /*
1115 * A reference count is needed on the nfsnode representing the
1116 * remote root. If this object is not persistent, then backward
1117 * traversals of the mount point (i.e. "..") will not work if
1118 * the nfsnode gets flushed out of the cache. Ufs does not have
1119 * this problem, because one can identify root inodes by their
1120 * number == ROOTINO (2).
1121 */
1122 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
1123 if (error)
1124 goto bad;
1125 *vpp = NFSTOV(np);
1126
1127 /*
eefb74c1
MD
1128 * Retrieval of mountpoint attributes is delayed until nfs_rot
1129 * or nfs_statfs are first called. This will happen either when
1130 * we first traverse the mount point or if somebody does a df(1).
1131 *
1132 * NFSSTA_GOTFSINFO is used to flag if we have successfully
1133 * retrieved mountpoint attributes. In the case of NFSv3 we
1134 * also flag static fsinfo.
984263bc 1135 */
eefb74c1
MD
1136 if (*vpp != NULL)
1137 (*vpp)->v_type = VNON;
984263bc
MD
1138
1139 /*
1140 * Lose the lock but keep the ref.
1141 */
a11aaa81 1142 vn_unlock(*vpp);
c6b43e93 1143 lwkt_gettoken(&nfs_token);
9a702b27 1144 TAILQ_INSERT_TAIL(&nfs_mountq, nmp, nm_entry);
c6b43e93 1145 lwkt_reltoken(&nfs_token);
984263bc 1146
13ddc895
MD
1147 switch(ncpus) {
1148 case 0:
f8565b0f 1149 case 1:
13ddc895
MD
1150 rxcpu = 0;
1151 txcpu = 0;
1152 break;
f8565b0f 1153 case 2:
13ddc895
MD
1154 rxcpu = 0;
1155 txcpu = 1;
1156 break;
1157 default:
d2d8515b
MD
1158 rxcpu = -1;
1159 txcpu = -1;
13ddc895
MD
1160 break;
1161 }
13ddc895 1162
52e1cf57
MD
1163 /*
1164 * Start the reader and writer threads.
1165 */
1166 lwkt_create(nfssvc_iod_reader, nmp, &nmp->nm_rxthread,
fdce8919 1167 NULL, 0, rxcpu, "nfsiod_rx");
52e1cf57 1168 lwkt_create(nfssvc_iod_writer, nmp, &nmp->nm_txthread,
fdce8919 1169 NULL, 0, txcpu, "nfsiod_tx");
c6b43e93 1170 lwkt_reltoken(&nmp->nm_token);
984263bc
MD
1171 return (0);
1172bad:
1173 nfs_disconnect(nmp);
c6b43e93 1174 lwkt_reltoken(&nmp->nm_token);
3b568787 1175 nfs_free_mount(nmp);
984263bc
MD
1176 return (error);
1177}
1178
1179/*
1180 * unmount system call
1181 */
1182static int
acde96db 1183nfs_unmount(struct mount *mp, int mntflags)
984263bc 1184{
40393ded 1185 struct nfsmount *nmp;
984263bc
MD
1186 int error, flags = 0;
1187
984263bc 1188 nmp = VFSTONFS(mp);
c6b43e93 1189 lwkt_gettoken(&nmp->nm_token);
d9adbeaf
MD
1190 if (mntflags & MNT_FORCE) {
1191 flags |= FORCECLOSE;
1192 nmp->nm_flag |= NFSMNT_FORCE;
1193 }
1194
984263bc
MD
1195 /*
1196 * Goes something like this..
1197 * - Call vflush() to clear out vnodes for this file system
1198 * - Close the socket
1199 * - Free up the data structures
1200 */
1201 /* In the forced case, cancel any outstanding requests. */
1202 if (flags & FORCECLOSE) {
1203 error = nfs_nmcancelreqs(nmp);
d9adbeaf
MD
1204 if (error) {
1205 kprintf("NFS: %s: Unable to cancel all requests\n",
1206 mp->mnt_stat.f_mntfromname);
1207 /* continue anyway */
1208 }
984263bc 1209 }
d9adbeaf 1210
984263bc 1211 /*
e07fef60 1212 * Must handshake with nfs_clientd() if it is active. XXX
984263bc
MD
1213 */
1214 nmp->nm_state |= NFSSTA_DISMINPROG;
984263bc 1215
d9adbeaf
MD
1216 /*
1217 * We hold 1 extra ref on the root vnode; see comment in mountnfs().
1218 *
1219 * If this doesn't work and we are doing a forced unmount we continue
1220 * anyway.
1221 */
984263bc
MD
1222 error = vflush(mp, 1, flags);
1223 if (error) {
1224 nmp->nm_state &= ~NFSSTA_DISMINPROG;
c6b43e93
MD
1225 if ((flags & FORCECLOSE) == 0) {
1226 lwkt_reltoken(&nmp->nm_token);
d9adbeaf 1227 return (error);
c6b43e93 1228 }
984263bc
MD
1229 }
1230
1231 /*
1232 * We are now committed to the unmount.
1233 * For NQNFS, let the server daemon free the nfsmount structure.
1234 */
e07fef60 1235 if (nmp->nm_flag & NFSMNT_KERB)
984263bc 1236 nmp->nm_state |= NFSSTA_DISMNT;
13ddc895 1237 nfssvc_iod_stop1(nmp);
984263bc 1238 nfs_disconnect(nmp);
13ddc895 1239 nfssvc_iod_stop2(nmp);
c6b43e93
MD
1240
1241 lwkt_gettoken(&nfs_token);
9a702b27 1242 TAILQ_REMOVE(&nfs_mountq, nmp, nm_entry);
c6b43e93
MD
1243 lwkt_reltoken(&nfs_token);
1244
1245 lwkt_reltoken(&nmp->nm_token);
984263bc 1246
52e1cf57 1247 if ((nmp->nm_flag & NFSMNT_KERB) == 0) {
3b568787 1248 nfs_free_mount(nmp);
52e1cf57 1249 }
984263bc
MD
1250 return (0);
1251}
1252
3b568787
MD
1253void
1254nfs_free_mount(struct nfsmount *nmp)
1255{
1256 if (nmp->nm_cred) {
1257 crfree(nmp->nm_cred);
1258 nmp->nm_cred = NULL;
1259 }
13ddc895 1260 if (nmp->nm_nam) {
884717e1 1261 kfree(nmp->nm_nam, M_SONAME);
13ddc895
MD
1262 nmp->nm_nam = NULL;
1263 }
0704a975 1264 objcache_put(nfsmount_objcache, nmp);
3b568787
MD
1265}
1266
984263bc
MD
1267/*
1268 * Return root of a filesystem
1269 */
1270static int
3fcb1ab8 1271nfs_root(struct mount *mp, struct vnode **vpp)
984263bc 1272{
40393ded 1273 struct vnode *vp;
984263bc 1274 struct nfsmount *nmp;
eefb74c1 1275 struct vattr attrs;
984263bc
MD
1276 struct nfsnode *np;
1277 int error;
1278
1279 nmp = VFSTONFS(mp);
c6b43e93 1280 lwkt_gettoken(&nmp->nm_token);
984263bc 1281 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
c6b43e93
MD
1282 if (error) {
1283 lwkt_reltoken(&nmp->nm_token);
984263bc 1284 return (error);
c6b43e93 1285 }
984263bc 1286 vp = NFSTOV(np);
eefb74c1
MD
1287
1288 /*
1289 * Get transfer parameters and root vnode attributes
b672fc16
MD
1290 *
1291 * NOTE: nfs_fsinfo() is expected to override the default
1292 * f_iosize we set.
eefb74c1
MD
1293 */
1294 if ((nmp->nm_state & NFSSTA_GOTFSINFO) == 0) {
1295 if (nmp->nm_flag & NFSMNT_NFSV3) {
eefb74c1 1296 mp->mnt_stat.f_iosize = nfs_iosize(1, nmp->nm_sotype);
b672fc16 1297 error = nfs_fsinfo(nmp, vp, curthread);
eefb74c1 1298 } else {
87de5057 1299 if ((error = VOP_GETATTR(vp, &attrs)) == 0)
eefb74c1 1300 nmp->nm_state |= NFSSTA_GOTFSINFO;
2ef35c61 1301
eefb74c1 1302 }
14aa2dd7
MD
1303 } else {
1304 /*
1305 * The root vnode is usually cached by the namecache so do not
1306 * try to avoid going over the wire even if we have previous
1307 * information cached. A stale NFS mount can loop
1308 * forever resolving the root vnode if we return no-error when
1309 * there is in fact an error.
1310 */
1311 np->n_attrstamp = 0;
1312 error = VOP_GETATTR(vp, &attrs);
eefb74c1 1313 }
984263bc 1314 if (vp->v_type == VNON)
1c843a13 1315 nfs_setvtype(vp, VDIR);
2247fe02 1316 vsetflags(vp, VROOT);
14aa2dd7
MD
1317 if (error)
1318 vput(vp);
1319 else
1320 *vpp = vp;
c6b43e93 1321 lwkt_reltoken(&nmp->nm_token);
14aa2dd7 1322 return (error);
984263bc
MD
1323}
1324
b2eb81cd
MD
1325struct scaninfo {
1326 int rescan;
b2eb81cd
MD
1327 int waitfor;
1328 int allerror;
1329};
1330
1331static int nfs_sync_scan1(struct mount *mp, struct vnode *vp, void *data);
5fd012e0 1332static int nfs_sync_scan2(struct mount *mp, struct vnode *vp, void *data);
b2eb81cd 1333
984263bc
MD
1334/*
1335 * Flush out the buffer cache
1336 */
1337/* ARGSUSED */
1338static int
87de5057 1339nfs_sync(struct mount *mp, int waitfor)
984263bc 1340{
c6b43e93 1341 struct nfsmount *nmp = VFSTONFS(mp);
b2eb81cd
MD
1342 struct scaninfo scaninfo;
1343 int error;
1344
a482a28a 1345 scaninfo.rescan = 1;
b2eb81cd
MD
1346 scaninfo.waitfor = waitfor;
1347 scaninfo.allerror = 0;
984263bc
MD
1348
1349 /*
1350 * Force stale buffer cache information to be flushed.
1351 */
c6b43e93 1352 lwkt_gettoken(&nmp->nm_token);
b2eb81cd
MD
1353 error = 0;
1354 while (error == 0 && scaninfo.rescan) {
1355 scaninfo.rescan = 0;
5fd012e0 1356 error = vmntvnodescan(mp, VMSC_GETVP, nfs_sync_scan1,
b2eb81cd 1357 nfs_sync_scan2, &scaninfo);
984263bc 1358 }
c6b43e93 1359 lwkt_reltoken(&nmp->nm_token);
b2eb81cd
MD
1360 return(error);
1361}
1362
3fcb1ab8 1363static int
b2eb81cd
MD
1364nfs_sync_scan1(struct mount *mp, struct vnode *vp, void *data)
1365{
1366 struct scaninfo *info = data;
1367
a11aaa81 1368 if (vn_islocked(vp) || RB_EMPTY(&vp->v_rbdirty_tree))
b2eb81cd 1369 return(-1);
28271622 1370 if (info->waitfor & MNT_LAZY)
b2eb81cd
MD
1371 return(-1);
1372 return(0);
1373}
1374
3fcb1ab8 1375static int
5fd012e0 1376nfs_sync_scan2(struct mount *mp, struct vnode *vp, void *data)
b2eb81cd
MD
1377{
1378 struct scaninfo *info = data;
1379 int error;
1380
52174f71 1381 error = VOP_FSYNC(vp, info->waitfor, 0);
b2eb81cd
MD
1382 if (error)
1383 info->allerror = error;
b2eb81cd 1384 return(0);
984263bc
MD
1385}
1386