kernel - Add many sysctl definitions, sysv, vfs, nfs, etc.
[dragonfly.git] / sys / vfs / nfs / nfs_vfsops.c
CommitLineData
093e85dc 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 $
df0a038e 38 * $DragonFly: src/sys/vfs/nfs/nfs_vfsops.c,v 1.54 2008/07/31 20:23:40 swildner Exp $
984263bc
MD
39 */
40
41#include "opt_bootp.h"
0e8ff41a 42#include "opt_nfsroot.h"
984263bc
MD
43
44#include <sys/param.h>
45#include <sys/sockio.h>
46#include <sys/proc.h>
47#include <sys/vnode.h>
9ba76b73 48#include <sys/fcntl.h>
984263bc
MD
49#include <sys/kernel.h>
50#include <sys/sysctl.h>
51#include <sys/malloc.h>
52#include <sys/mount.h>
53#include <sys/mbuf.h>
54#include <sys/socket.h>
55#include <sys/socketvar.h>
56#include <sys/systm.h>
57
58#include <vm/vm.h>
59#include <vm/vm_extern.h>
60#include <vm/vm_zone.h>
61
62#include <net/if.h>
63#include <net/route.h>
64#include <netinet/in.h>
65
165dba55 66#include <sys/thread2.h>
8684e6f9 67#include <sys/mutex2.h>
165dba55 68
1f2de5d4
MD
69#include "rpcv2.h"
70#include "nfsproto.h"
71#include "nfs.h"
1f2de5d4 72#include "nfsmount.h"
c1cf1e59 73#include "nfsnode.h"
1f2de5d4
MD
74#include "xdr_subs.h"
75#include "nfsm_subs.h"
76#include "nfsdiskless.h"
0e8ff41a 77#include "nfsmountrpc.h"
984263bc 78
c8a30e1e
JS
79extern int nfs_mountroot(struct mount *mp);
80extern void bootpc_init(void);
984263bc 81
66a1ddf5
MD
82extern struct vop_ops nfsv2_vnode_vops;
83extern struct vop_ops nfsv2_fifo_vops;
84extern struct vop_ops nfsv2_spec_vops;
984263bc
MD
85
86MALLOC_DEFINE(M_NFSREQ, "NFS req", "NFS request header");
87MALLOC_DEFINE(M_NFSBIGFH, "NFSV3 bigfh", "NFS version 3 file handle");
88MALLOC_DEFINE(M_NFSD, "NFS daemon", "Nfs server daemon structure");
89MALLOC_DEFINE(M_NFSDIROFF, "NFSV3 diroff", "NFS directory offset data");
90MALLOC_DEFINE(M_NFSRVDESC, "NFSV3 srvdesc", "NFS server socket descriptor");
91MALLOC_DEFINE(M_NFSUID, "NFS uid", "Nfs uid mapping structure");
984263bc
MD
92MALLOC_DEFINE(M_NFSHASH, "NFS hash", "NFS hash tables");
93
94vm_zone_t nfsmount_zone;
95
96struct nfsstats nfsstats;
97SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW, 0, "NFS filesystem");
093e85dc
SG
98SYSCTL_STRUCT(_vfs_nfs, NFS_NFSSTATS, nfsstats, CTLFLAG_RD, &nfsstats, nfsstats,
99 "Nfs stats structure");
984263bc 100static int nfs_ip_paranoia = 1;
093e85dc
SG
101SYSCTL_INT(_vfs_nfs, OID_AUTO, nfs_ip_paranoia, CTLFLAG_RW, &nfs_ip_paranoia, 0,
102 "Enable no-connection mode for protocols that support no-connection mode");
984263bc
MD
103#ifdef NFS_DEBUG
104int nfs_debug;
105SYSCTL_INT(_vfs_nfs, OID_AUTO, debug, CTLFLAG_RW, &nfs_debug, 0, "");
106#endif
107
eefb74c1
MD
108/*
109 * Tunable to determine the Read/Write unit size. Maximum value
110 * is NFS_MAXDATA. We also default to NFS_MAXDATA.
111 */
112static int nfs_io_size = NFS_MAXDATA;
113SYSCTL_INT(_vfs_nfs, OID_AUTO, nfs_io_size, CTLFLAG_RW,
114 &nfs_io_size, 0, "NFS optimal I/O unit size");
115
a6ee311a
RG
116static void nfs_decode_args (struct nfsmount *nmp,
117 struct nfs_args *argp);
118static int mountnfs (struct nfs_args *,struct mount *,
119 struct sockaddr *,char *,char *,struct vnode **);
120static int nfs_mount ( struct mount *mp, char *path, caddr_t data,
acde96db
MD
121 struct ucred *cred);
122static int nfs_unmount ( struct mount *mp, int mntflags);
a6ee311a
RG
123static int nfs_root ( struct mount *mp, struct vnode **vpp);
124static int nfs_statfs ( struct mount *mp, struct statfs *sbp,
acde96db 125 struct ucred *cred);
2ef35c61
AH
126static int nfs_statvfs(struct mount *mp, struct statvfs *sbp,
127 struct ucred *cred);
87de5057 128static int nfs_sync ( struct mount *mp, int waitfor);
984263bc
MD
129
130/*
131 * nfs vfs operations.
132 */
133static struct vfsops nfs_vfsops = {
43c45e8f
HP
134 .vfs_mount = nfs_mount,
135 .vfs_unmount = nfs_unmount,
136 .vfs_root = nfs_root,
137 .vfs_statfs = nfs_statfs,
2ef35c61 138 .vfs_statvfs = nfs_statvfs,
43c45e8f
HP
139 .vfs_sync = nfs_sync,
140 .vfs_init = nfs_init,
141 .vfs_uninit = nfs_uninit
984263bc
MD
142};
143VFS_SET(nfs_vfsops, nfs, VFCF_NETWORK);
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;
f9642f56 711 vfs_unbusy(mp);
984263bc
MD
712
713 /*
714 * This is not really an nfs issue, but it is much easier to
715 * set hostname here and then let the "/etc/rc.xxx" files
716 * mount the right /var based upon its preset value.
717 */
718 bcopy(nd->my_hostnam, hostname, MAXHOSTNAMELEN);
719 hostname[MAXHOSTNAMELEN - 1] = '\0';
720 for (i = 0; i < MAXHOSTNAMELEN; i++)
721 if (hostname[i] == '\0')
722 break;
723 inittodr(ntohl(nd->root_time));
0b9984b8 724 crit_exit();
984263bc
MD
725 return (0);
726}
727
728/*
729 * Internal version of mount system call for diskless setup.
730 */
731static int
dadab5e9
MD
732nfs_mountdiskless(char *path, char *which, int mountflag,
733 struct sockaddr_in *sin, struct nfs_args *args, struct thread *td,
734 struct vnode **vpp, struct mount **mpp)
984263bc
MD
735{
736 struct mount *mp;
737 struct sockaddr *nam;
984263bc 738 int didalloc = 0;
0e8ff41a 739 int error;
984263bc
MD
740
741 mp = *mpp;
742
743 if (mp == NULL) {
744 if ((error = vfs_rootmountalloc("nfs", path, &mp)) != 0) {
086c1d7e 745 kprintf("nfs_mountroot: NFS not configured");
984263bc
MD
746 return (error);
747 }
748 didalloc = 1;
749 }
984263bc
MD
750 mp->mnt_kern_flag = 0;
751 mp->mnt_flag = mountflag;
cfa2ba21 752 nam = dup_sockaddr((struct sockaddr *)sin);
0e8ff41a
MD
753
754#if defined(BOOTP) || defined(NFS_ROOT)
755 if (args->fhsize == 0) {
df0a038e
SW
756 char *xpath = path;
757
226341c1
MD
758 kprintf("NFS_ROOT: No FH passed from loader, attempting "
759 "mount rpc...");
5dacbb2e
MD
760 while (*xpath && *xpath != ':')
761 ++xpath;
762 if (*xpath)
763 ++xpath;
0e8ff41a 764 args->fhsize = 0;
5dacbb2e 765 error = md_mount(sin, xpath, args->fh, &args->fhsize, args, td);
0e8ff41a 766 if (error) {
5dacbb2e 767 kprintf("failed error %d.\n", error);
0e8ff41a
MD
768 goto haderror;
769 }
086c1d7e 770 kprintf("success!\n");
0e8ff41a
MD
771 }
772#endif
773
984263bc 774 if ((error = mountnfs(args, mp, nam, which, path, vpp)) != 0) {
04ef15fb 775#if defined(BOOTP) || defined(NFS_ROOT)
0e8ff41a 776haderror:
04ef15fb 777#endif
086c1d7e 778 kprintf("nfs_mountroot: mount %s on %s: %d", path, which, error);
984263bc 779 mp->mnt_vfc->vfc_refcount--;
f9642f56 780 vfs_unbusy(mp);
984263bc 781 if (didalloc)
efda3bd0 782 kfree(mp, M_MOUNT);
984263bc
MD
783 FREE(nam, M_SONAME);
784 return (error);
785 }
984263bc
MD
786 *mpp = mp;
787 return (0);
788}
789
790static void
3fcb1ab8 791nfs_decode_args(struct nfsmount *nmp, struct nfs_args *argp)
984263bc 792{
984263bc
MD
793 int adjsock;
794 int maxio;
795
165dba55 796 crit_enter();
984263bc
MD
797 /*
798 * Silently clear NFSMNT_NOCONN if it's a TCP mount, it makes
799 * no sense in that context.
800 */
5e6f1ca5 801 if (nmp->nm_sotype == SOCK_STREAM) {
984263bc 802 nmp->nm_flag &= ~NFSMNT_NOCONN;
5e6f1ca5
MD
803 argp->flags &= ~NFSMNT_NOCONN;
804 }
984263bc 805
5e6f1ca5 806 /*
226341c1 807 * readdirplus is NFSv3 only.
5e6f1ca5 808 */
5e6f1ca5 809 if ((argp->flags & NFSMNT_NFSV3) == 0) {
984263bc 810 nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
5e6f1ca5
MD
811 argp->flags &= ~NFSMNT_RDIRPLUS;
812 }
984263bc 813
e20dbf61
MD
814 /*
815 * Re-bind if rsrvd port flag has changed
816 */
817 adjsock = (nmp->nm_flag & NFSMNT_RESVPORT) !=
818 (argp->flags & NFSMNT_RESVPORT);
984263bc
MD
819
820 /* Update flags atomically. Don't change the lock bits. */
821 nmp->nm_flag = argp->flags | nmp->nm_flag;
165dba55 822 crit_exit();
984263bc
MD
823
824 if ((argp->flags & NFSMNT_TIMEO) && argp->timeo > 0) {
825 nmp->nm_timeo = (argp->timeo * NFS_HZ + 5) / 10;
826 if (nmp->nm_timeo < NFS_MINTIMEO)
827 nmp->nm_timeo = NFS_MINTIMEO;
828 else if (nmp->nm_timeo > NFS_MAXTIMEO)
829 nmp->nm_timeo = NFS_MAXTIMEO;
830 }
831
832 if ((argp->flags & NFSMNT_RETRANS) && argp->retrans > 1) {
833 nmp->nm_retry = argp->retrans;
834 if (nmp->nm_retry > NFS_MAXREXMIT)
835 nmp->nm_retry = NFS_MAXREXMIT;
836 }
837
e20dbf61
MD
838 /*
839 * These parameters effect the buffer cache and cannot be changed
840 * once we've successfully mounted.
841 */
842 if ((nmp->nm_state & NFSSTA_GOTFSINFO) == 0) {
843 maxio = nfs_iosize(argp->flags & NFSMNT_NFSV3, nmp->nm_sotype);
844
845 if ((argp->flags & NFSMNT_WSIZE) && argp->wsize > 0) {
846 nmp->nm_wsize = argp->wsize;
847 /* Round down to multiple of blocksize */
848 nmp->nm_wsize &= ~(NFS_FABLKSIZE - 1);
849 if (nmp->nm_wsize <= 0)
850 nmp->nm_wsize = NFS_FABLKSIZE;
851 }
852 if (nmp->nm_wsize > maxio)
853 nmp->nm_wsize = maxio;
854 if (nmp->nm_wsize > MAXBSIZE)
855 nmp->nm_wsize = MAXBSIZE;
856
857 if ((argp->flags & NFSMNT_RSIZE) && argp->rsize > 0) {
858 nmp->nm_rsize = argp->rsize;
859 /* Round down to multiple of blocksize */
860 nmp->nm_rsize &= ~(NFS_FABLKSIZE - 1);
861 if (nmp->nm_rsize <= 0)
862 nmp->nm_rsize = NFS_FABLKSIZE;
863 }
864 if (nmp->nm_rsize > maxio)
865 nmp->nm_rsize = maxio;
866 if (nmp->nm_rsize > MAXBSIZE)
867 nmp->nm_rsize = MAXBSIZE;
868
869 if ((argp->flags & NFSMNT_READDIRSIZE) &&
870 argp->readdirsize > 0) {
871 nmp->nm_readdirsize = argp->readdirsize;
872 }
873 if (nmp->nm_readdirsize > maxio)
874 nmp->nm_readdirsize = maxio;
875 if (nmp->nm_readdirsize > nmp->nm_rsize)
876 nmp->nm_readdirsize = nmp->nm_rsize;
984263bc 877 }
984263bc
MD
878
879 if ((argp->flags & NFSMNT_ACREGMIN) && argp->acregmin >= 0)
880 nmp->nm_acregmin = argp->acregmin;
881 else
882 nmp->nm_acregmin = NFS_MINATTRTIMO;
883 if ((argp->flags & NFSMNT_ACREGMAX) && argp->acregmax >= 0)
884 nmp->nm_acregmax = argp->acregmax;
885 else
886 nmp->nm_acregmax = NFS_MAXATTRTIMO;
887 if ((argp->flags & NFSMNT_ACDIRMIN) && argp->acdirmin >= 0)
888 nmp->nm_acdirmin = argp->acdirmin;
889 else
890 nmp->nm_acdirmin = NFS_MINDIRATTRTIMO;
891 if ((argp->flags & NFSMNT_ACDIRMAX) && argp->acdirmax >= 0)
892 nmp->nm_acdirmax = argp->acdirmax;
893 else
894 nmp->nm_acdirmax = NFS_MAXDIRATTRTIMO;
895 if (nmp->nm_acdirmin > nmp->nm_acdirmax)
896 nmp->nm_acdirmin = nmp->nm_acdirmax;
897 if (nmp->nm_acregmin > nmp->nm_acregmax)
898 nmp->nm_acregmin = nmp->nm_acregmax;
899
900 if ((argp->flags & NFSMNT_MAXGRPS) && argp->maxgrouplist >= 0) {
901 if (argp->maxgrouplist <= NFS_MAXGRPS)
902 nmp->nm_numgrps = argp->maxgrouplist;
903 else
904 nmp->nm_numgrps = NFS_MAXGRPS;
905 }
906 if ((argp->flags & NFSMNT_READAHEAD) && argp->readahead >= 0) {
907 if (argp->readahead <= NFS_MAXRAHEAD)
908 nmp->nm_readahead = argp->readahead;
909 else
910 nmp->nm_readahead = NFS_MAXRAHEAD;
911 }
984263bc 912 if ((argp->flags & NFSMNT_DEADTHRESH) && argp->deadthresh >= 1) {
e07fef60 913 if (argp->deadthresh <= NFS_NEVERDEAD)
984263bc
MD
914 nmp->nm_deadthresh = argp->deadthresh;
915 else
e07fef60 916 nmp->nm_deadthresh = NFS_NEVERDEAD;
984263bc
MD
917 }
918
984263bc
MD
919 if (nmp->nm_so && adjsock) {
920 nfs_safedisconnect(nmp);
921 if (nmp->nm_sotype == SOCK_DGRAM)
60233e58 922 while (nfs_connect(nmp, NULL)) {
086c1d7e 923 kprintf("nfs_args: retrying connect\n");
377d4740 924 (void) tsleep((caddr_t)&lbolt, 0, "nfscon", 0);
984263bc
MD
925 }
926 }
927}
928
929/*
930 * VFS Operations.
931 *
932 * mount system call
933 * It seems a bit dumb to copyinstr() the host and path here and then
934 * bcopy() them in mountnfs(), but I wanted to detect errors before
935 * doing the sockargs() call because sockargs() allocates an mbuf and
936 * an error after that means that I have to release the mbuf.
937 */
938/* ARGSUSED */
939static int
acde96db 940nfs_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
984263bc
MD
941{
942 int error;
943 struct nfs_args args;
944 struct sockaddr *nam;
945 struct vnode *vp;
946 char pth[MNAMELEN], hst[MNAMELEN];
947 size_t len;
948 u_char nfh[NFSX_V3FHMAX];
949
950 if (path == NULL) {
951 nfs_mountroot(mp);
952 return (0);
953 }
954 error = copyin(data, (caddr_t)&args, sizeof (struct nfs_args));
955 if (error)
956 return (error);
957 if (args.version != NFS_ARGSVERSION) {
958#ifdef COMPAT_PRELITE2
959 /*
960 * If the argument version is unknown, then assume the
961 * caller is a pre-lite2 4.4BSD client and convert its
962 * arguments.
963 */
964 struct onfs_args oargs;
965 error = copyin(data, (caddr_t)&oargs, sizeof (struct onfs_args));
966 if (error)
967 return (error);
968 nfs_convert_oargs(&args,&oargs);
969#else /* !COMPAT_PRELITE2 */
970 return (EPROGMISMATCH);
971#endif /* COMPAT_PRELITE2 */
972 }
973 if (mp->mnt_flag & MNT_UPDATE) {
40393ded 974 struct nfsmount *nmp = VFSTONFS(mp);
984263bc
MD
975
976 if (nmp == NULL)
977 return (EIO);
978 /*
979 * When doing an update, we can't change from or to
e20dbf61 980 * v3, or change cookie translation, or rsize or wsize.
984263bc 981 */
e20dbf61
MD
982 args.flags &= ~(NFSMNT_NFSV3 | NFSMNT_RSIZE | NFSMNT_WSIZE);
983 args.flags |= nmp->nm_flag & (NFSMNT_NFSV3);
984263bc
MD
984 nfs_decode_args(nmp, &args);
985 return (0);
986 }
987
988 /*
989 * Make the nfs_ip_paranoia sysctl serve as the default connection
2ef35c61 990 * or no-connection mode for those protocols that support
984263bc
MD
991 * no-connection mode (the flag will be cleared later for protocols
992 * that do not support no-connection mode). This will allow a client
993 * to receive replies from a different IP then the request was
994 * sent to. Note: default value for nfs_ip_paranoia is 1 (paranoid),
995 * not 0.
996 */
997 if (nfs_ip_paranoia == 0)
998 args.flags |= NFSMNT_NOCONN;
999 if (args.fhsize < 0 || args.fhsize > NFSX_V3FHMAX)
1000 return (EINVAL);
1001 error = copyin((caddr_t)args.fh, (caddr_t)nfh, args.fhsize);
1002 if (error)
1003 return (error);
1004 error = copyinstr(path, pth, MNAMELEN-1, &len);
1005 if (error)
1006 return (error);
1007 bzero(&pth[len], MNAMELEN - len);
1008 error = copyinstr(args.hostname, hst, MNAMELEN-1, &len);
1009 if (error)
1010 return (error);
1011 bzero(&hst[len], MNAMELEN - len);
1012 /* sockargs() call must be after above copyin() calls */
1013 error = getsockaddr(&nam, (caddr_t)args.addr, args.addrlen);
1014 if (error)
1015 return (error);
1016 args.fh = nfh;
1017 error = mountnfs(&args, mp, nam, pth, hst, &vp);
1018 return (error);
1019}
1020
1021/*
1022 * Common code for mount and mountroot
1023 */
1024static int
dadab5e9
MD
1025mountnfs(struct nfs_args *argp, struct mount *mp, struct sockaddr *nam,
1026 char *pth, char *hst, struct vnode **vpp)
984263bc 1027{
40393ded 1028 struct nfsmount *nmp;
984263bc
MD
1029 struct nfsnode *np;
1030 int error;
13ddc895
MD
1031 int rxcpu;
1032 int txcpu;
984263bc
MD
1033
1034 if (mp->mnt_flag & MNT_UPDATE) {
1035 nmp = VFSTONFS(mp);
1036 /* update paths, file handles, etc, here XXX */
1037 FREE(nam, M_SONAME);
1038 return (0);
1039 } else {
1040 nmp = zalloc(nfsmount_zone);
1041 bzero((caddr_t)nmp, sizeof (struct nfsmount));
8684e6f9
MD
1042 mtx_init(&nmp->nm_rxlock);
1043 mtx_init(&nmp->nm_txlock);
984263bc 1044 TAILQ_INIT(&nmp->nm_uidlruhead);
81b5c339 1045 TAILQ_INIT(&nmp->nm_bioq);
e21aec5b 1046 TAILQ_INIT(&nmp->nm_reqq);
edb90c22
MD
1047 TAILQ_INIT(&nmp->nm_reqtxq);
1048 TAILQ_INIT(&nmp->nm_reqrxq);
984263bc 1049 mp->mnt_data = (qaddr_t)nmp;
c6b43e93 1050 lwkt_token_init(&nmp->nm_token, 1, "nfs_token");
984263bc
MD
1051 }
1052 vfs_getnewfsid(mp);
1053 nmp->nm_mountp = mp;
c6b43e93
MD
1054 mp->mnt_kern_flag |= MNTK_ALL_MPSAFE;
1055
1056 lwkt_gettoken(&nmp->nm_token);
984263bc
MD
1057
1058 /*
1059 * V2 can only handle 32 bit filesizes. A 4GB-1 limit may be too
1060 * high, depending on whether we end up with negative offsets in
1061 * the client or server somewhere. 2GB-1 may be safer.
1062 *
1063 * For V3, nfs_fsinfo will adjust this as necessary. Assume maximum
1064 * that we can handle until we find out otherwise.
1065 * XXX Our "safe" limit on the client is what we can store in our
1066 * buffer cache using signed(!) block numbers.
1067 */
1068 if ((argp->flags & NFSMNT_NFSV3) == 0)
1069 nmp->nm_maxfilesize = 0xffffffffLL;
1070 else
1071 nmp->nm_maxfilesize = (u_int64_t)0x80000000 * DEV_BSIZE - 1;
1072
1073 nmp->nm_timeo = NFS_TIMEO;
1074 nmp->nm_retry = NFS_RETRANS;
eefb74c1
MD
1075 nmp->nm_wsize = nfs_iosize(argp->flags & NFSMNT_NFSV3, argp->sotype);
1076 nmp->nm_rsize = nmp->nm_wsize;
984263bc
MD
1077 nmp->nm_readdirsize = NFS_READDIRSIZE;
1078 nmp->nm_numgrps = NFS_MAXGRPS;
1079 nmp->nm_readahead = NFS_DEFRAHEAD;
0639df29 1080 nmp->nm_deadthresh = NFS_DEADTHRESH;
984263bc
MD
1081 nmp->nm_fhsize = argp->fhsize;
1082 bcopy((caddr_t)argp->fh, (caddr_t)nmp->nm_fh, argp->fhsize);
1083 bcopy(hst, mp->mnt_stat.f_mntfromname, MNAMELEN);
984263bc
MD
1084 nmp->nm_nam = nam;
1085 /* Set up the sockets and per-host congestion */
1086 nmp->nm_sotype = argp->sotype;
1087 nmp->nm_soproto = argp->proto;
3b568787 1088 nmp->nm_cred = crhold(proc0.p_ucred);
984263bc
MD
1089
1090 nfs_decode_args(nmp, argp);
1091
1092 /*
1093 * For Connection based sockets (TCP,...) defer the connect until
1094 * the first request, in case the server is not responding.
1095 */
1096 if (nmp->nm_sotype == SOCK_DGRAM &&
60233e58 1097 (error = nfs_connect(nmp, NULL)))
984263bc
MD
1098 goto bad;
1099
1100 /*
1101 * This is silly, but it has to be set so that vinifod() works.
1102 * We do not want to do an nfs_statfs() here since we can get
1103 * stuck on a dead server and we are holding a lock on the mount
1104 * point.
1105 */
2ef35c61 1106 mp->mnt_stat.f_iosize =
eefb74c1
MD
1107 nfs_iosize(nmp->nm_flag & NFSMNT_NFSV3, nmp->nm_sotype);
1108
984263bc 1109 /*
0961aa92
MD
1110 * Install vop_ops for our vnops
1111 */
66a1ddf5
MD
1112 vfs_add_vnodeops(mp, &nfsv2_vnode_vops, &mp->mnt_vn_norm_ops);
1113 vfs_add_vnodeops(mp, &nfsv2_spec_vops, &mp->mnt_vn_spec_ops);
1114 vfs_add_vnodeops(mp, &nfsv2_fifo_vops, &mp->mnt_vn_fifo_ops);
0961aa92
MD
1115
1116 /*
984263bc
MD
1117 * A reference count is needed on the nfsnode representing the
1118 * remote root. If this object is not persistent, then backward
1119 * traversals of the mount point (i.e. "..") will not work if
1120 * the nfsnode gets flushed out of the cache. Ufs does not have
1121 * this problem, because one can identify root inodes by their
1122 * number == ROOTINO (2).
1123 */
1124 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
1125 if (error)
1126 goto bad;
1127 *vpp = NFSTOV(np);
1128
1129 /*
eefb74c1
MD
1130 * Retrieval of mountpoint attributes is delayed until nfs_rot
1131 * or nfs_statfs are first called. This will happen either when
1132 * we first traverse the mount point or if somebody does a df(1).
1133 *
1134 * NFSSTA_GOTFSINFO is used to flag if we have successfully
1135 * retrieved mountpoint attributes. In the case of NFSv3 we
1136 * also flag static fsinfo.
984263bc 1137 */
eefb74c1
MD
1138 if (*vpp != NULL)
1139 (*vpp)->v_type = VNON;
984263bc
MD
1140
1141 /*
1142 * Lose the lock but keep the ref.
1143 */
a11aaa81 1144 vn_unlock(*vpp);
c6b43e93 1145 lwkt_gettoken(&nfs_token);
9a702b27 1146 TAILQ_INSERT_TAIL(&nfs_mountq, nmp, nm_entry);
c6b43e93 1147 lwkt_reltoken(&nfs_token);
984263bc 1148
13ddc895
MD
1149#ifdef SMP
1150 switch(ncpus) {
1151 case 0:
f8565b0f 1152 case 1:
13ddc895
MD
1153 rxcpu = 0;
1154 txcpu = 0;
1155 break;
f8565b0f 1156 case 2:
13ddc895
MD
1157 rxcpu = 0;
1158 txcpu = 1;
1159 break;
1160 default:
1161 rxcpu = 1;
1162 txcpu = 2;
1163 break;
1164 }
1165#else
1166 rxcpu = 0;
1167 txcpu = 0;
1168#endif
1169
52e1cf57
MD
1170 /*
1171 * Start the reader and writer threads.
1172 */
1173 lwkt_create(nfssvc_iod_reader, nmp, &nmp->nm_rxthread,
fdce8919 1174 NULL, 0, rxcpu, "nfsiod_rx");
52e1cf57 1175 lwkt_create(nfssvc_iod_writer, nmp, &nmp->nm_txthread,
fdce8919 1176 NULL, 0, txcpu, "nfsiod_tx");
c6b43e93 1177 lwkt_reltoken(&nmp->nm_token);
984263bc
MD
1178 return (0);
1179bad:
1180 nfs_disconnect(nmp);
c6b43e93 1181 lwkt_reltoken(&nmp->nm_token);
3b568787 1182 nfs_free_mount(nmp);
984263bc
MD
1183 return (error);
1184}
1185
1186/*
1187 * unmount system call
1188 */
1189static int
acde96db 1190nfs_unmount(struct mount *mp, int mntflags)
984263bc 1191{
40393ded 1192 struct nfsmount *nmp;
984263bc
MD
1193 int error, flags = 0;
1194
984263bc 1195 nmp = VFSTONFS(mp);
c6b43e93 1196 lwkt_gettoken(&nmp->nm_token);
d9adbeaf
MD
1197 if (mntflags & MNT_FORCE) {
1198 flags |= FORCECLOSE;
1199 nmp->nm_flag |= NFSMNT_FORCE;
1200 }
1201
984263bc
MD
1202 /*
1203 * Goes something like this..
1204 * - Call vflush() to clear out vnodes for this file system
1205 * - Close the socket
1206 * - Free up the data structures
1207 */
1208 /* In the forced case, cancel any outstanding requests. */
1209 if (flags & FORCECLOSE) {
1210 error = nfs_nmcancelreqs(nmp);
d9adbeaf
MD
1211 if (error) {
1212 kprintf("NFS: %s: Unable to cancel all requests\n",
1213 mp->mnt_stat.f_mntfromname);
1214 /* continue anyway */
1215 }
984263bc 1216 }
d9adbeaf 1217
984263bc 1218 /*
e07fef60 1219 * Must handshake with nfs_clientd() if it is active. XXX
984263bc
MD
1220 */
1221 nmp->nm_state |= NFSSTA_DISMINPROG;
984263bc 1222
d9adbeaf
MD
1223 /*
1224 * We hold 1 extra ref on the root vnode; see comment in mountnfs().
1225 *
1226 * If this doesn't work and we are doing a forced unmount we continue
1227 * anyway.
1228 */
984263bc
MD
1229 error = vflush(mp, 1, flags);
1230 if (error) {
1231 nmp->nm_state &= ~NFSSTA_DISMINPROG;
c6b43e93
MD
1232 if ((flags & FORCECLOSE) == 0) {
1233 lwkt_reltoken(&nmp->nm_token);
d9adbeaf 1234 return (error);
c6b43e93 1235 }
984263bc
MD
1236 }
1237
1238 /*
1239 * We are now committed to the unmount.
1240 * For NQNFS, let the server daemon free the nfsmount structure.
1241 */
e07fef60 1242 if (nmp->nm_flag & NFSMNT_KERB)
984263bc 1243 nmp->nm_state |= NFSSTA_DISMNT;
13ddc895 1244 nfssvc_iod_stop1(nmp);
984263bc 1245 nfs_disconnect(nmp);
13ddc895 1246 nfssvc_iod_stop2(nmp);
c6b43e93
MD
1247
1248 lwkt_gettoken(&nfs_token);
9a702b27 1249 TAILQ_REMOVE(&nfs_mountq, nmp, nm_entry);
c6b43e93
MD
1250 lwkt_reltoken(&nfs_token);
1251
1252 lwkt_reltoken(&nmp->nm_token);
984263bc 1253
52e1cf57 1254 if ((nmp->nm_flag & NFSMNT_KERB) == 0) {
3b568787 1255 nfs_free_mount(nmp);
52e1cf57 1256 }
984263bc
MD
1257 return (0);
1258}
1259
3b568787
MD
1260void
1261nfs_free_mount(struct nfsmount *nmp)
1262{
1263 if (nmp->nm_cred) {
1264 crfree(nmp->nm_cred);
1265 nmp->nm_cred = NULL;
1266 }
13ddc895
MD
1267 if (nmp->nm_nam) {
1268 FREE(nmp->nm_nam, M_SONAME);
1269 nmp->nm_nam = NULL;
1270 }
3b568787
MD
1271 zfree(nfsmount_zone, nmp);
1272}
1273
984263bc
MD
1274/*
1275 * Return root of a filesystem
1276 */
1277static int
3fcb1ab8 1278nfs_root(struct mount *mp, struct vnode **vpp)
984263bc 1279{
40393ded 1280 struct vnode *vp;
984263bc 1281 struct nfsmount *nmp;
eefb74c1 1282 struct vattr attrs;
984263bc
MD
1283 struct nfsnode *np;
1284 int error;
1285
1286 nmp = VFSTONFS(mp);
c6b43e93 1287 lwkt_gettoken(&nmp->nm_token);
984263bc 1288 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
c6b43e93
MD
1289 if (error) {
1290 lwkt_reltoken(&nmp->nm_token);
984263bc 1291 return (error);
c6b43e93 1292 }
984263bc 1293 vp = NFSTOV(np);
eefb74c1
MD
1294
1295 /*
1296 * Get transfer parameters and root vnode attributes
b672fc16
MD
1297 *
1298 * NOTE: nfs_fsinfo() is expected to override the default
1299 * f_iosize we set.
eefb74c1
MD
1300 */
1301 if ((nmp->nm_state & NFSSTA_GOTFSINFO) == 0) {
1302 if (nmp->nm_flag & NFSMNT_NFSV3) {
eefb74c1 1303 mp->mnt_stat.f_iosize = nfs_iosize(1, nmp->nm_sotype);
b672fc16 1304 error = nfs_fsinfo(nmp, vp, curthread);
eefb74c1 1305 } else {
87de5057 1306 if ((error = VOP_GETATTR(vp, &attrs)) == 0)
eefb74c1 1307 nmp->nm_state |= NFSSTA_GOTFSINFO;
2ef35c61 1308
eefb74c1 1309 }
14aa2dd7
MD
1310 } else {
1311 /*
1312 * The root vnode is usually cached by the namecache so do not
1313 * try to avoid going over the wire even if we have previous
1314 * information cached. A stale NFS mount can loop
1315 * forever resolving the root vnode if we return no-error when
1316 * there is in fact an error.
1317 */
1318 np->n_attrstamp = 0;
1319 error = VOP_GETATTR(vp, &attrs);
eefb74c1 1320 }
984263bc 1321 if (vp->v_type == VNON)
1c843a13 1322 nfs_setvtype(vp, VDIR);
2247fe02 1323 vsetflags(vp, VROOT);
14aa2dd7
MD
1324 if (error)
1325 vput(vp);
1326 else
1327 *vpp = vp;
c6b43e93 1328 lwkt_reltoken(&nmp->nm_token);
14aa2dd7 1329 return (error);
984263bc
MD
1330}
1331
b2eb81cd
MD
1332struct scaninfo {
1333 int rescan;
b2eb81cd
MD
1334 int waitfor;
1335 int allerror;
1336};
1337
1338static int nfs_sync_scan1(struct mount *mp, struct vnode *vp, void *data);
5fd012e0 1339static int nfs_sync_scan2(struct mount *mp, struct vnode *vp, void *data);
b2eb81cd 1340
984263bc
MD
1341/*
1342 * Flush out the buffer cache
1343 */
1344/* ARGSUSED */
1345static int
87de5057 1346nfs_sync(struct mount *mp, int waitfor)
984263bc 1347{
c6b43e93 1348 struct nfsmount *nmp = VFSTONFS(mp);
b2eb81cd
MD
1349 struct scaninfo scaninfo;
1350 int error;
1351
a482a28a 1352 scaninfo.rescan = 1;
b2eb81cd
MD
1353 scaninfo.waitfor = waitfor;
1354 scaninfo.allerror = 0;
984263bc
MD
1355
1356 /*
1357 * Force stale buffer cache information to be flushed.
1358 */
c6b43e93 1359 lwkt_gettoken(&nmp->nm_token);
b2eb81cd
MD
1360 error = 0;
1361 while (error == 0 && scaninfo.rescan) {
1362 scaninfo.rescan = 0;
5fd012e0 1363 error = vmntvnodescan(mp, VMSC_GETVP, nfs_sync_scan1,
b2eb81cd 1364 nfs_sync_scan2, &scaninfo);
984263bc 1365 }
c6b43e93 1366 lwkt_reltoken(&nmp->nm_token);
b2eb81cd
MD
1367 return(error);
1368}
1369
3fcb1ab8 1370static int
b2eb81cd
MD
1371nfs_sync_scan1(struct mount *mp, struct vnode *vp, void *data)
1372{
1373 struct scaninfo *info = data;
1374
a11aaa81 1375 if (vn_islocked(vp) || RB_EMPTY(&vp->v_rbdirty_tree))
b2eb81cd
MD
1376 return(-1);
1377 if (info->waitfor == MNT_LAZY)
1378 return(-1);
1379 return(0);
1380}
1381
3fcb1ab8 1382static int
5fd012e0 1383nfs_sync_scan2(struct mount *mp, struct vnode *vp, void *data)
b2eb81cd
MD
1384{
1385 struct scaninfo *info = data;
1386 int error;
1387
52174f71 1388 error = VOP_FSYNC(vp, info->waitfor, 0);
b2eb81cd
MD
1389 if (error)
1390 info->allerror = error;
b2eb81cd 1391 return(0);
984263bc
MD
1392}
1393