1 /*-
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Rick Macklem at The University of Guelph.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 4. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)nfs_subs.c  8.8 (Berkeley) 5/22/95
33  */
34 
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD: stable/9/sys/nfs/nfs_common.c 247502 2013-02-28 21:57:38Z jhb $");
37 
38 /*
39  * These functions support the macros and help fiddle mbuf chains for
40  * the nfs op functions.  They do things like create the rpc header and
41  * copy data between mbuf chains and uio lists.
42  */
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/proc.h>
50 #include <sys/mount.h>
51 #include <sys/vnode.h>
52 #include <sys/namei.h>
53 #include <sys/mbuf.h>
54 #include <sys/socket.h>
55 #include <sys/stat.h>
56 #include <sys/malloc.h>
57 #include <sys/module.h>
58 #include <sys/sysent.h>
59 #include <sys/syscall.h>
60 #include <sys/sysctl.h>
61 
62 #include <vm/vm.h>
63 #include <vm/vm_object.h>
64 #include <vm/vm_extern.h>
65 
66 #include <nfs/nfsproto.h>
67 #include <nfsserver/nfs.h>
68 #include <nfs/xdr_subs.h>
69 #include <nfs/nfs_common.h>
70 
71 #include <netinet/in.h>
72 
73 enum vtype nv3tov_type[8]= {
74 	VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO
75 };
76 nfstype nfsv3_type[9] = {
77 	NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK, NFFIFO, NFNON
78 };
79 
80 static void *nfsm_dissect_xx_sub(int s, struct mbuf **md, caddr_t *dpos,
81     int how);
82 
83 SYSCTL_NODE(_vfs, OID_AUTO, nfs_common, CTLFLAG_RD, 0, "NFS common support");
84 
85 static int nfs_realign_test;
86 SYSCTL_INT(_vfs_nfs_common, OID_AUTO, realign_test, CTLFLAG_RD,
87     &nfs_realign_test, 0, "Number of realign tests done");
88 
89 static int nfs_realign_count;
90 SYSCTL_INT(_vfs_nfs_common, OID_AUTO, realign_count, CTLFLAG_RD,
91     &nfs_realign_count, 0, "Number of mbuf realignments done");
92 
93 /*
94  * copies mbuf chain to the uio scatter/gather list
95  */
96 int
nfsm_mbuftouio(struct mbuf ** mrep,struct uio * uiop,int siz,caddr_t * dpos)97 nfsm_mbuftouio(struct mbuf **mrep, struct uio *uiop, int siz, caddr_t *dpos)
98 {
99 	char *mbufcp, *uiocp;
100 	int xfer, left, len;
101 	struct mbuf *mp;
102 	long uiosiz, rem;
103 	int error = 0;
104 
105 	mp = *mrep;
106 	mbufcp = *dpos;
107 	len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
108 	rem = nfsm_rndup(siz)-siz;
109 	while (siz > 0) {
110 		if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
111 			return (EFBIG);
112 		left = uiop->uio_iov->iov_len;
113 		uiocp = uiop->uio_iov->iov_base;
114 		if (left > siz)
115 			left = siz;
116 		uiosiz = left;
117 		while (left > 0) {
118 			while (len == 0) {
119 				mp = mp->m_next;
120 				if (mp == NULL)
121 					return (EBADRPC);
122 				mbufcp = mtod(mp, caddr_t);
123 				len = mp->m_len;
124 			}
125 			xfer = (left > len) ? len : left;
126 #ifdef notdef
127 			/* Not Yet.. */
128 			if (uiop->uio_iov->iov_op != NULL)
129 				(*(uiop->uio_iov->iov_op))
130 				(mbufcp, uiocp, xfer);
131 			else
132 #endif
133 			if (uiop->uio_segflg == UIO_SYSSPACE)
134 				bcopy(mbufcp, uiocp, xfer);
135 			else
136 				copyout(mbufcp, uiocp, xfer);
137 			left -= xfer;
138 			len -= xfer;
139 			mbufcp += xfer;
140 			uiocp += xfer;
141 			uiop->uio_offset += xfer;
142 			uiop->uio_resid -= xfer;
143 		}
144 		if (uiop->uio_iov->iov_len <= siz) {
145 			uiop->uio_iovcnt--;
146 			uiop->uio_iov++;
147 		} else {
148 			uiop->uio_iov->iov_base =
149 			    (char *)uiop->uio_iov->iov_base + uiosiz;
150 			uiop->uio_iov->iov_len -= uiosiz;
151 		}
152 		siz -= uiosiz;
153 	}
154 	*dpos = mbufcp;
155 	*mrep = mp;
156 	if (rem > 0) {
157 		if (len < rem)
158 			error = nfs_adv(mrep, dpos, rem, len);
159 		else
160 			*dpos += rem;
161 	}
162 	return (error);
163 }
164 
165 /*
166  * Help break down an mbuf chain by setting the first siz bytes contiguous
167  * pointed to by returned val.
168  * This is used by the macros nfsm_dissect for tough
169  * cases. (The macros use the vars. dpos and dpos2)
170  */
171 void *
nfsm_disct(struct mbuf ** mdp,caddr_t * dposp,int siz,int left,int how)172 nfsm_disct(struct mbuf **mdp, caddr_t *dposp, int siz, int left, int how)
173 {
174 	struct mbuf *mp, *mp2;
175 	int siz2, xfer;
176 	caddr_t ptr, npos = NULL;
177 	void *ret;
178 
179 	mp = *mdp;
180 	while (left == 0) {
181 		*mdp = mp = mp->m_next;
182 		if (mp == NULL)
183 			return (NULL);
184 		left = mp->m_len;
185 		*dposp = mtod(mp, caddr_t);
186 	}
187 	if (left >= siz) {
188 		ret = *dposp;
189 		*dposp += siz;
190 	} else if (mp->m_next == NULL) {
191 		return (NULL);
192 	} else if (siz > MHLEN) {
193 		panic("nfs S too big");
194 	} else {
195 		MGET(mp2, how, MT_DATA);
196 		if (mp2 == NULL)
197 			return (NULL);
198 		mp2->m_len = siz;
199 		mp2->m_next = mp->m_next;
200 		mp->m_next = mp2;
201 		mp->m_len -= left;
202 		mp = mp2;
203 		ptr = mtod(mp, caddr_t);
204 		ret = ptr;
205 		bcopy(*dposp, ptr, left);		/* Copy what was left */
206 		siz2 = siz-left;
207 		ptr += left;
208 		mp2 = mp->m_next;
209 		npos = mtod(mp2, caddr_t);
210 		/* Loop around copying up the siz2 bytes */
211 		while (siz2 > 0) {
212 			if (mp2 == NULL)
213 				return (NULL);
214 			xfer = (siz2 > mp2->m_len) ? mp2->m_len : siz2;
215 			if (xfer > 0) {
216 				bcopy(mtod(mp2, caddr_t), ptr, xfer);
217 				mp2->m_data += xfer;
218 				mp2->m_len -= xfer;
219 				ptr += xfer;
220 				siz2 -= xfer;
221 			}
222 			if (siz2 > 0) {
223 				mp2 = mp2->m_next;
224 				if (mp2 != NULL)
225 					npos = mtod(mp2, caddr_t);
226 			}
227 		}
228 		*mdp = mp2;
229 		*dposp = mtod(mp2, caddr_t);
230 		if (!nfsm_aligned(*dposp, u_int32_t)) {
231 			bcopy(*dposp, npos, mp2->m_len);
232 			mp2->m_data = npos;
233 			*dposp = npos;
234 		}
235 	}
236 	return (ret);
237 }
238 
239 /*
240  * Advance the position in the mbuf chain.
241  */
242 int
nfs_adv(struct mbuf ** mdp,caddr_t * dposp,int offs,int left)243 nfs_adv(struct mbuf **mdp, caddr_t *dposp, int offs, int left)
244 {
245 	struct mbuf *m;
246 	int s;
247 
248 	m = *mdp;
249 	s = left;
250 	while (s < offs) {
251 		offs -= s;
252 		m = m->m_next;
253 		if (m == NULL)
254 			return (EBADRPC);
255 		s = m->m_len;
256 	}
257 	*mdp = m;
258 	*dposp = mtod(m, caddr_t)+offs;
259 	return (0);
260 }
261 
262 void *
nfsm_build_xx(int s,struct mbuf ** mb,caddr_t * bpos)263 nfsm_build_xx(int s, struct mbuf **mb, caddr_t *bpos)
264 {
265 	struct mbuf *mb2;
266 	void *ret;
267 
268 	if (s > M_TRAILINGSPACE(*mb)) {
269 		MGET(mb2, M_WAIT, MT_DATA);
270 		if (s > MLEN)
271 			panic("build > MLEN");
272 		(*mb)->m_next = mb2;
273 		*mb = mb2;
274 		(*mb)->m_len = 0;
275 		*bpos = mtod(*mb, caddr_t);
276 	}
277 	ret = *bpos;
278 	(*mb)->m_len += s;
279 	*bpos += s;
280 	return (ret);
281 }
282 
283 void *
nfsm_dissect_xx(int s,struct mbuf ** md,caddr_t * dpos)284 nfsm_dissect_xx(int s, struct mbuf **md, caddr_t *dpos)
285 {
286 
287 	return (nfsm_dissect_xx_sub(s, md, dpos, M_WAIT));
288 }
289 
290 void *
nfsm_dissect_xx_nonblock(int s,struct mbuf ** md,caddr_t * dpos)291 nfsm_dissect_xx_nonblock(int s, struct mbuf **md, caddr_t *dpos)
292 {
293 
294 	return (nfsm_dissect_xx_sub(s, md, dpos, M_DONTWAIT));
295 }
296 
297 static void *
nfsm_dissect_xx_sub(int s,struct mbuf ** md,caddr_t * dpos,int how)298 nfsm_dissect_xx_sub(int s, struct mbuf **md, caddr_t *dpos, int how)
299 {
300 	int t1;
301 	char *cp2;
302 	void *ret;
303 
304 	t1 = mtod(*md, caddr_t) + (*md)->m_len - *dpos;
305 	if (t1 >= s) {
306 		ret = *dpos;
307 		*dpos += s;
308 		return (ret);
309 	}
310 	cp2 = nfsm_disct(md, dpos, s, t1, how);
311 	return (cp2);
312 }
313 
314 int
nfsm_strsiz_xx(int * s,int m,struct mbuf ** mb,caddr_t * bpos)315 nfsm_strsiz_xx(int *s, int m, struct mbuf **mb, caddr_t *bpos)
316 {
317 	u_int32_t *tl;
318 
319 	tl = nfsm_dissect_xx(NFSX_UNSIGNED, mb, bpos);
320 	if (tl == NULL)
321 		return (EBADRPC);
322 	*s = fxdr_unsigned(int32_t, *tl);
323 	if (*s > m)
324 		return (EBADRPC);
325 	return (0);
326 }
327 
328 int
nfsm_adv_xx(int s,struct mbuf ** md,caddr_t * dpos)329 nfsm_adv_xx(int s, struct mbuf **md, caddr_t *dpos)
330 {
331 	int t1;
332 
333 	t1 = mtod(*md, caddr_t) + (*md)->m_len - *dpos;
334 	if (t1 >= s) {
335 		*dpos += s;
336 		return (0);
337 	}
338 	t1 = nfs_adv(md, dpos, s, t1);
339 	if (t1)
340 		return (t1);
341 	return (0);
342 }
343 
344 /*
345  * Check for badly aligned mbuf data and realign by copying the unaligned
346  * portion of the data into a new mbuf chain and freeing the portions of the
347  * old chain that were replaced.
348  *
349  * We cannot simply realign the data within the existing mbuf chain because
350  * the underlying buffers may contain other rpc commands and we cannot afford
351  * to overwrite them.
352  *
353  * We would prefer to avoid this situation entirely.  The situation does not
354  * occur with NFS/UDP and is supposed to only occassionally occur with TCP.
355  * Use vfs.nfs_common.realign_count and realign_test to check this.
356  */
357 int
nfs_realign(struct mbuf ** pm,int how)358 nfs_realign(struct mbuf **pm, int how)
359 {
360 	struct mbuf *m, *n;
361 	int off;
362 
363 	++nfs_realign_test;
364 	while ((m = *pm) != NULL) {
365 		if (!nfsm_aligned(m->m_len, u_int32_t) ||
366 		    !nfsm_aligned(mtod(m, intptr_t), u_int32_t)) {
367 			/*
368 			 * NB: we can't depend on m_pkthdr.len to help us
369 			 * decide what to do here.  May not be worth doing
370 			 * the m_length calculation as m_copyback will
371 			 * expand the mbuf chain below as needed.
372 			 */
373 			if (m_length(m, NULL) >= MINCLSIZE) {
374 				/* NB: m_copyback handles space > MCLBYTES */
375 				n = m_getcl(how, MT_DATA, 0);
376 			} else
377 				n = m_get(how, MT_DATA);
378 			if (n == NULL)
379 				return (ENOMEM);
380 			/*
381 			 * Align the remainder of the mbuf chain.
382 			 */
383 			n->m_len = 0;
384 			off = 0;
385 			while (m != NULL) {
386 				m_copyback(n, off, m->m_len, mtod(m, caddr_t));
387 				off += m->m_len;
388 				m = m->m_next;
389 			}
390 			m_freem(*pm);
391 			*pm = n;
392 			++nfs_realign_count;
393 			break;
394 		}
395 		pm = &m->m_next;
396 	}
397 	return (0);
398 }
399 
400 static moduledata_t nfs_common_mod = {
401 	"nfs_common",
402 	NULL,
403 	NULL
404 };
405 
406 DECLARE_MODULE(nfs_common, nfs_common_mod, SI_SUB_VFS, SI_ORDER_ANY);
407 MODULE_VERSION(nfs_common, 1);
408