2 * Copyright (c) 1997 Berkeley Software Design, Inc. All rights reserved.
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
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
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
28 * from BSDI nfs_lock.c,v 2.4 1998/12/14 23:49:56 jch Exp
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
34 #include <sys/param.h>
35 #include <sys/systm.h>
37 #include <sys/fcntl.h>
38 #include <sys/kernel.h> /* for hz */
39 #include <sys/limits.h>
41 #include <sys/malloc.h>
42 #include <sys/lockf.h> /* for hz */ /* Must come after sys/malloc.h */
44 #include <sys/mount.h>
45 #include <sys/namei.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>
56 #include <nfs/nfsproto.h>
57 #include <nfsclient/nfs.h>
58 #include <nfsclient/nfsmount.h>
59 #include <nfsclient/nfsnode.h>
60 #include <nfsclient/nfs_lock.h>
61 #include <nfsclient/nlminfo.h>
63 extern void (*nlminfo_release_p)(struct proc *p);
65 MALLOC_DEFINE(M_NFSLOCK, "nfsclient_lock", "NFS lock request");
66 MALLOC_DEFINE(M_NLMINFO, "nfsclient_nlminfo", "NFS lock process structure");
68 static int nfslockdans(struct thread *td, struct lockd_ans *ansp);
69 static void nlminfo_release(struct proc *p);
71 * --------------------------------------------------------------------
72 * A miniature device driver which the userland uses to talk to us.
76 static struct cdev *nfslock_dev;
77 static struct mtx nfslock_mtx;
78 static int nfslock_isopen;
79 static TAILQ_HEAD(,__lock_msg) nfslock_list;
82 nfslock_open(struct cdev *dev, int oflags, int devtype, struct thread *td)
86 error = priv_check(td, PRIV_NFS_LOCKD);
90 mtx_lock(&nfslock_mtx);
91 if (!nfslock_isopen) {
97 mtx_unlock(&nfslock_mtx);
103 nfslock_close(struct cdev *dev, int fflag, int devtype, struct thread *td)
105 struct __lock_msg *lm;
107 mtx_lock(&nfslock_mtx);
109 while (!TAILQ_EMPTY(&nfslock_list)) {
110 lm = TAILQ_FIRST(&nfslock_list);
111 /* XXX: answer request */
112 TAILQ_REMOVE(&nfslock_list, lm, lm_link);
115 mtx_unlock(&nfslock_mtx);
120 nfslock_read(struct cdev *dev, struct uio *uio, int ioflag)
123 struct __lock_msg *lm;
125 if (uio->uio_resid != sizeof *lm)
129 mtx_lock(&nfslock_mtx);
130 while (TAILQ_EMPTY(&nfslock_list)) {
131 error = msleep(&nfslock_list, &nfslock_mtx, PSOCK | PCATCH,
137 lm = TAILQ_FIRST(&nfslock_list);
138 TAILQ_REMOVE(&nfslock_list, lm, lm_link);
140 mtx_unlock(&nfslock_mtx);
142 error = uiomove(lm, sizeof *lm, uio);
149 nfslock_write(struct cdev *dev, struct uio *uio, int ioflag)
154 if (uio->uio_resid != sizeof la)
156 error = uiomove(&la, sizeof la, uio);
158 error = nfslockdans(curthread, &la);
163 nfslock_send(struct __lock_msg *lm)
165 struct __lock_msg *lm2;
169 lm2 = malloc(sizeof *lm2, M_NFSLOCK, M_WAITOK);
170 mtx_lock(&nfslock_mtx);
171 if (nfslock_isopen) {
172 memcpy(lm2, lm, sizeof *lm2);
173 TAILQ_INSERT_TAIL(&nfslock_list, lm2, lm_link);
174 wakeup(&nfslock_list);
178 mtx_unlock(&nfslock_mtx);
180 free(lm2, M_NFSLOCK);
184 static struct cdevsw nfslock_cdevsw = {
185 .d_version = D_VERSION,
186 .d_open = nfslock_open,
187 .d_close = nfslock_close,
188 .d_read = nfslock_read,
189 .d_write = nfslock_write,
194 nfslock_modevent(module_t mod __unused, int type, void *data __unused)
200 printf("nfslock: pseudo-device\n");
201 mtx_init(&nfslock_mtx, "nfslock", NULL, MTX_DEF);
202 TAILQ_INIT(&nfslock_list);
203 nlminfo_release_p = nlminfo_release;
204 nfslock_dev = make_dev(&nfslock_cdevsw, 0,
205 UID_ROOT, GID_KMEM, 0600, _PATH_NFSLCKDEV);
212 DEV_MODULE(nfslock, nfslock_modevent, NULL);
213 MODULE_VERSION(nfslock, 1);
218 * We have to let the process know if the call succeeded. I'm using an extra
219 * field in the p_nlminfo field in the proc structure, as it is already for
225 * NFS advisory byte-level locks.
227 * The vnode shall be (shared) locked on the entry, it is
228 * unconditionally unlocked after.
231 nfs_dolock(struct vop_advlock_args *ap)
246 ASSERT_VOP_LOCKED(vp, "nfs_dolock");
248 bcopy(VFSTONFS(vp->v_mount)->nm_nam, &msg.lm_addr,
249 min(sizeof msg.lm_addr, VFSTONFS(vp->v_mount)->nm_nam->sa_len));
250 msg.lm_fh_len = NFS_ISV3(vp) ? VTONFS(vp)->n_fhsize : NFSX_V2FH;
251 bcopy(VTONFS(vp)->n_fhp, msg.lm_fh, msg.lm_fh_len);
252 msg.lm_nfsv3 = NFS_ISV3(vp);
256 * the NLM protocol doesn't allow the server to return an error
257 * on ranges, so we do it.
259 if (fl->l_whence != SEEK_END) {
260 if ((fl->l_whence != SEEK_CUR && fl->l_whence != SEEK_SET) ||
263 (fl->l_start == 0 || fl->l_start + fl->l_len < 0)))
266 (fl->l_len - 1 > OFF_MAX - fl->l_start))
271 * Fill in the information structure.
273 msg.lm_version = LOCKD_MSG_VERSION;
274 msg.lm_msg_ident.pid = p->p_pid;
278 * if there is no nfsowner table yet, allocate one.
280 if (p->p_nlminfo == NULL) {
281 p->p_nlminfo = malloc(sizeof(struct nlminfo),
282 M_NLMINFO, M_WAITOK | M_ZERO);
283 p->p_nlminfo->pid_start = p->p_stats->p_start;
284 timevaladd(&p->p_nlminfo->pid_start, &boottime);
286 msg.lm_msg_ident.pid_start = p->p_nlminfo->pid_start;
287 msg.lm_msg_ident.msg_seq = ++(p->p_nlminfo->msg_seq);
290 msg.lm_wait = ap->a_flags & F_WAIT;
291 msg.lm_getlk = ap->a_op == F_GETLK;
292 cru2x(td->td_ucred, &msg.lm_cred);
295 error = nfslock_send(&msg);
299 /* Unlocks succeed immediately. */
300 if (fl->l_type == F_UNLCK)
304 * Retry after 20 seconds if we haven't gotten a response yet.
305 * This number was picked out of thin air... but is longer
306 * then even a reasonably loaded system should take (at least
307 * on a local network). XXX Probably should use a back-off
310 * XXX: No PCATCH here since we currently have no useful
311 * way to signal to the userland rpc.lockd that the request
312 * has been aborted. Once the rpc.lockd implementation
313 * can handle aborts, and we report them properly,
314 * PCATCH can be put back. In the mean time, if we did
315 * permit aborting, the lock attempt would "get lost"
316 * and the lock would get stuck in the locked state.
318 error = tsleep(p->p_nlminfo, PUSER, "lockd", 20*hz);
320 if (error == EWOULDBLOCK) {
322 * We timed out, so we rewrite the request
331 if (msg.lm_getlk && p->p_nlminfo->retcode == 0) {
332 if (p->p_nlminfo->set_getlk_pid) {
333 fl->l_sysid = 0; /* XXX */
334 fl->l_pid = p->p_nlminfo->getlk_pid;
336 fl->l_type = F_UNLCK;
339 error = p->p_nlminfo->retcode;
349 * NFS advisory byte-level locks answer from the lock daemon.
352 nfslockdans(struct thread *td, struct lockd_ans *ansp)
354 struct proc *targetp;
356 /* the version should match, or we're out of sync */
357 if (ansp->la_vers != LOCKD_ANS_VERSION)
360 /* Find the process, set its return errno and wake it up. */
361 if ((targetp = pfind(ansp->la_msg_ident.pid)) == NULL)
364 /* verify the pid hasn't been reused (if we can), and it isn't waiting
365 * for an answer from a more recent request. We return an EPIPE if
366 * the match fails, because we've already used ESRCH above, and this
367 * is sort of like writing on a pipe after the reader has closed it.
369 if (targetp->p_nlminfo == NULL ||
370 ((ansp->la_msg_ident.msg_seq != -1) &&
371 (timevalcmp(&targetp->p_nlminfo->pid_start,
372 &ansp->la_msg_ident.pid_start, !=) ||
373 targetp->p_nlminfo->msg_seq != ansp->la_msg_ident.msg_seq))) {
374 PROC_UNLOCK(targetp);
378 targetp->p_nlminfo->retcode = ansp->la_errno;
379 targetp->p_nlminfo->set_getlk_pid = ansp->la_set_getlk_pid;
380 targetp->p_nlminfo->getlk_pid = ansp->la_getlk_pid;
382 wakeup(targetp->p_nlminfo);
384 PROC_UNLOCK(targetp);
389 * Free nlminfo attached to process.
392 nlminfo_release(struct proc *p)
394 free(p->p_nlminfo, M_NLMINFO);