1/*-
2 * Copyright (c) 1997 Berkeley Software Design, Inc. All rights reserved.
3 *
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
6 * are met:
7 * 1. Redistributions of source code must retain the above copyright
8 *    notice, this list of conditions and the following disclaimer.
9 * 2. Redistributions in binary form must reproduce the above copyright
10 *    notice, this list of conditions and the following disclaimer in the
11 *    documentation and/or other materials provided with the distribution.
12 * 3. Berkeley Software Design Inc's name may not be used to endorse or
13 *    promote products derived from this software without specific prior
14 *    written permission.
15 *
16 * THIS SOFTWARE IS PROVIDED BY BERKELEY SOFTWARE DESIGN INC ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED.  IN NO EVENT SHALL BERKELEY SOFTWARE DESIGN INC BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 *
28 *      from BSDI nfs_lock.c,v 2.4 1998/12/14 23:49:56 jch Exp
29 */
30
31#include <sys/cdefs.h>
32__FBSDID("$FreeBSD$");
33
34#include <sys/param.h>
35#include <sys/systm.h>
36#include <sys/conf.h>
37#include <sys/fcntl.h>
38#include <sys/kernel.h>		/* for hz */
39#include <sys/limits.h>
40#include <sys/lock.h>
41#include <sys/malloc.h>
42#include <sys/lockf.h>		/* for hz */ /* Must come after sys/malloc.h */
43#include <sys/mbuf.h>
44#include <sys/mount.h>
45#include <sys/namei.h>
46#include <sys/priv.h>
47#include <sys/proc.h>
48#include <sys/resourcevar.h>
49#include <sys/socket.h>
50#include <sys/socket.h>
51#include <sys/unistd.h>
52#include <sys/vnode.h>
53
54#include <net/if.h>
55
56#include <nfs/nfsproto.h>
57#include <nfs/nfs_lock.h>
58#include <nfsclient/nfs.h>
59#include <nfsclient/nfsmount.h>
60#include <nfsclient/nfsnode.h>
61#include <nfsclient/nlminfo.h>
62
63extern void (*nlminfo_release_p)(struct proc *p);
64
65vop_advlock_t	*nfs_advlock_p = nfs_dolock;
66vop_reclaim_t	*nfs_reclaim_p = NULL;
67
68static MALLOC_DEFINE(M_NFSLOCK, "nfsclient_lock", "NFS lock request");
69static MALLOC_DEFINE(M_NLMINFO, "nfsclient_nlminfo",
70    "NFS lock process structure");
71
72static int nfslockdans(struct thread *td, struct lockd_ans *ansp);
73static void nlminfo_release(struct proc *p);
74/*
75 * --------------------------------------------------------------------
76 * A miniature device driver which the userland uses to talk to us.
77 *
78 */
79
80static struct cdev *nfslock_dev;
81static struct mtx nfslock_mtx;
82static int nfslock_isopen;
83static TAILQ_HEAD(,__lock_msg)	nfslock_list;
84
85static int
86nfslock_open(struct cdev *dev, int oflags, int devtype, struct thread *td)
87{
88	int error;
89
90	error = priv_check(td, PRIV_NFS_LOCKD);
91	if (error)
92		return (error);
93
94	mtx_lock(&nfslock_mtx);
95	if (!nfslock_isopen) {
96		error = 0;
97		nfslock_isopen = 1;
98	} else {
99		error = EOPNOTSUPP;
100	}
101	mtx_unlock(&nfslock_mtx);
102
103	return (error);
104}
105
106static int
107nfslock_close(struct cdev *dev, int fflag, int devtype, struct thread *td)
108{
109	struct __lock_msg *lm;
110
111	mtx_lock(&nfslock_mtx);
112	nfslock_isopen = 0;
113	while (!TAILQ_EMPTY(&nfslock_list)) {
114		lm = TAILQ_FIRST(&nfslock_list);
115		/* XXX: answer request */
116		TAILQ_REMOVE(&nfslock_list, lm, lm_link);
117		free(lm, M_NFSLOCK);
118	}
119	mtx_unlock(&nfslock_mtx);
120	return (0);
121}
122
123static int
124nfslock_read(struct cdev *dev, struct uio *uio, int ioflag)
125{
126	int error;
127	struct __lock_msg *lm;
128
129	if (uio->uio_resid != sizeof *lm)
130		return (EOPNOTSUPP);
131	lm = NULL;
132	error = 0;
133	mtx_lock(&nfslock_mtx);
134	while (TAILQ_EMPTY(&nfslock_list)) {
135		error = msleep(&nfslock_list, &nfslock_mtx, PSOCK | PCATCH,
136		    "nfslockd", 0);
137		if (error)
138			break;
139	}
140	if (!error) {
141		lm = TAILQ_FIRST(&nfslock_list);
142		TAILQ_REMOVE(&nfslock_list, lm, lm_link);
143	}
144	mtx_unlock(&nfslock_mtx);
145	if (!error) {
146		error = uiomove(lm, sizeof *lm, uio);
147		free(lm, M_NFSLOCK);
148	}
149	return (error);
150}
151
152static int
153nfslock_write(struct cdev *dev, struct uio *uio, int ioflag)
154{
155	struct lockd_ans la;
156	int error;
157
158	if (uio->uio_resid != sizeof la)
159		return (EOPNOTSUPP);
160	error = uiomove(&la, sizeof la, uio);
161	if (!error)
162		error = nfslockdans(curthread, &la);
163	return (error);
164}
165
166static int
167nfslock_send(struct __lock_msg *lm)
168{
169	struct __lock_msg *lm2;
170	int error;
171
172	error = 0;
173	lm2 = malloc(sizeof *lm2, M_NFSLOCK, M_WAITOK);
174	mtx_lock(&nfslock_mtx);
175	if (nfslock_isopen) {
176		memcpy(lm2, lm, sizeof *lm2);
177		TAILQ_INSERT_TAIL(&nfslock_list, lm2, lm_link);
178		wakeup(&nfslock_list);
179	} else {
180		error = EOPNOTSUPP;
181	}
182	mtx_unlock(&nfslock_mtx);
183	if (error)
184		free(lm2, M_NFSLOCK);
185	return (error);
186}
187
188static struct cdevsw nfslock_cdevsw = {
189	.d_version =	D_VERSION,
190	.d_open =	nfslock_open,
191	.d_close =	nfslock_close,
192	.d_read =	nfslock_read,
193	.d_write =	nfslock_write,
194	.d_name =	"nfslock"
195};
196
197static int
198nfslock_modevent(module_t mod __unused, int type, void *data __unused)
199{
200
201	switch (type) {
202	case MOD_LOAD:
203		if (bootverbose)
204			printf("nfslock: pseudo-device\n");
205		mtx_init(&nfslock_mtx, "nfslock", NULL, MTX_DEF);
206		TAILQ_INIT(&nfslock_list);
207		nlminfo_release_p = nlminfo_release;
208		nfslock_dev = make_dev(&nfslock_cdevsw, 0,
209		    UID_ROOT, GID_KMEM, 0600, _PATH_NFSLCKDEV);
210		return (0);
211	default:
212		return (EOPNOTSUPP);
213	}
214}
215
216DEV_MODULE(nfslock, nfslock_modevent, NULL);
217MODULE_VERSION(nfslock, 1);
218
219
220/*
221 * XXX
222 * We have to let the process know if the call succeeded.  I'm using an extra
223 * field in the p_nlminfo field in the proc structure, as it is already for
224 * lockd stuff.
225 */
226
227/*
228 * nfs_advlock --
229 *      NFS advisory byte-level locks.
230 *
231 * The vnode shall be (shared) locked on the entry, it is
232 * unconditionally unlocked after.
233 */
234int
235nfs_dolock(struct vop_advlock_args *ap)
236{
237	LOCKD_MSG msg;
238	struct thread *td;
239	struct vnode *vp;
240	int error;
241	struct flock *fl;
242	struct proc *p;
243	struct nfsmount *nmp;
244
245	td = curthread;
246	p = td->td_proc;
247
248	vp = ap->a_vp;
249	fl = ap->a_fl;
250	nmp = VFSTONFS(vp->v_mount);
251
252	ASSERT_VOP_LOCKED(vp, "nfs_dolock");
253
254	nmp->nm_getinfo(vp, msg.lm_fh, &msg.lm_fh_len, &msg.lm_addr,
255	    &msg.lm_nfsv3, NULL, NULL);
256	VOP_UNLOCK(vp, 0);
257
258	/*
259	 * the NLM protocol doesn't allow the server to return an error
260	 * on ranges, so we do it.
261	 */
262	if (fl->l_whence != SEEK_END) {
263		if ((fl->l_whence != SEEK_CUR && fl->l_whence != SEEK_SET) ||
264		    fl->l_start < 0 ||
265		    (fl->l_len < 0 &&
266		     (fl->l_start == 0 || fl->l_start + fl->l_len < 0)))
267			return (EINVAL);
268		if (fl->l_len > 0 &&
269			 (fl->l_len - 1 > OFF_MAX - fl->l_start))
270			return (EOVERFLOW);
271	}
272
273	/*
274	 * Fill in the information structure.
275	 */
276	msg.lm_version = LOCKD_MSG_VERSION;
277	msg.lm_msg_ident.pid = p->p_pid;
278
279	mtx_lock(&Giant);
280	/*
281	 * if there is no nfsowner table yet, allocate one.
282	 */
283	if (p->p_nlminfo == NULL) {
284		p->p_nlminfo = malloc(sizeof(struct nlminfo),
285		    M_NLMINFO, M_WAITOK | M_ZERO);
286		p->p_nlminfo->pid_start = p->p_stats->p_start;
287		timevaladd(&p->p_nlminfo->pid_start, &boottime);
288	}
289	msg.lm_msg_ident.pid_start = p->p_nlminfo->pid_start;
290	msg.lm_msg_ident.msg_seq = ++(p->p_nlminfo->msg_seq);
291
292	msg.lm_fl = *fl;
293	msg.lm_wait = ap->a_flags & F_WAIT;
294	msg.lm_getlk = ap->a_op == F_GETLK;
295	cru2x(td->td_ucred, &msg.lm_cred);
296
297	for (;;) {
298		error = nfslock_send(&msg);
299		if (error)
300			goto out;
301
302		/* Unlocks succeed immediately.  */
303		if (fl->l_type == F_UNLCK)
304			goto out;
305
306		/*
307		 * Retry after 20 seconds if we haven't gotten a response yet.
308		 * This number was picked out of thin air... but is longer
309		 * then even a reasonably loaded system should take (at least
310		 * on a local network).  XXX Probably should use a back-off
311		 * scheme.
312		 *
313		 * XXX: No PCATCH here since we currently have no useful
314		 * way to signal to the userland rpc.lockd that the request
315		 * has been aborted.  Once the rpc.lockd implementation
316		 * can handle aborts, and we report them properly,
317		 * PCATCH can be put back.  In the mean time, if we did
318		 * permit aborting, the lock attempt would "get lost"
319		 * and the lock would get stuck in the locked state.
320		 */
321		error = tsleep(p->p_nlminfo, PUSER, "lockd", 20*hz);
322		if (error != 0) {
323			if (error == EWOULDBLOCK) {
324				/*
325				 * We timed out, so we rewrite the request
326				 * to the fifo.
327				 */
328				continue;
329			}
330
331			break;
332		}
333
334		if (msg.lm_getlk && p->p_nlminfo->retcode == 0) {
335			if (p->p_nlminfo->set_getlk_pid) {
336				fl->l_sysid = 0; /* XXX */
337				fl->l_pid = p->p_nlminfo->getlk_pid;
338			} else {
339				fl->l_type = F_UNLCK;
340			}
341		}
342		error = p->p_nlminfo->retcode;
343		break;
344	}
345 out:
346	mtx_unlock(&Giant);
347	return (error);
348}
349
350/*
351 * nfslockdans --
352 *      NFS advisory byte-level locks answer from the lock daemon.
353 */
354static int
355nfslockdans(struct thread *td, struct lockd_ans *ansp)
356{
357	struct proc *targetp;
358
359	/* the version should match, or we're out of sync */
360	if (ansp->la_vers != LOCKD_ANS_VERSION)
361		return (EINVAL);
362
363	/* Find the process, set its return errno and wake it up. */
364	if ((targetp = pfind(ansp->la_msg_ident.pid)) == NULL)
365		return (ESRCH);
366
367	/* verify the pid hasn't been reused (if we can), and it isn't waiting
368	 * for an answer from a more recent request.  We return an EPIPE if
369	 * the match fails, because we've already used ESRCH above, and this
370	 * is sort of like writing on a pipe after the reader has closed it.
371	 */
372	if (targetp->p_nlminfo == NULL ||
373	    ((ansp->la_msg_ident.msg_seq != -1) &&
374	      (timevalcmp(&targetp->p_nlminfo->pid_start,
375			&ansp->la_msg_ident.pid_start, !=) ||
376	       targetp->p_nlminfo->msg_seq != ansp->la_msg_ident.msg_seq))) {
377		PROC_UNLOCK(targetp);
378		return (EPIPE);
379	}
380
381	targetp->p_nlminfo->retcode = ansp->la_errno;
382	targetp->p_nlminfo->set_getlk_pid = ansp->la_set_getlk_pid;
383	targetp->p_nlminfo->getlk_pid = ansp->la_getlk_pid;
384
385	wakeup(targetp->p_nlminfo);
386
387	PROC_UNLOCK(targetp);
388	return (0);
389}
390
391/*
392 * Free nlminfo attached to process.
393 */
394void
395nlminfo_release(struct proc *p)
396{
397	free(p->p_nlminfo, M_NLMINFO);
398	p->p_nlminfo = NULL;
399}
400