2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
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.
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
32 * @(#)nfs_syscalls.c 8.5 (Berkeley) 3/30/95
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/sysproto.h>
41 #include <sys/kernel.h>
42 #include <sys/sysctl.h>
44 #include <sys/filedesc.h>
45 #include <sys/vnode.h>
46 #include <sys/malloc.h>
47 #include <sys/mount.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/domain.h>
55 #include <sys/protosw.h>
56 #include <sys/namei.h>
57 #include <sys/unistd.h>
58 #include <sys/kthread.h>
59 #include <sys/fcntl.h>
60 #include <sys/lockf.h>
61 #include <sys/mutex.h>
63 #include <netinet/in.h>
64 #include <netinet/tcp.h>
66 #include <nfs/xdr_subs.h>
67 #include <nfs/nfsproto.h>
68 #include <nfsclient/nfs.h>
69 #include <nfsclient/nfsm_subs.h>
70 #include <nfsclient/nfsmount.h>
71 #include <nfsclient/nfsnode.h>
72 #include <nfsclient/nfs_lock.h>
74 static MALLOC_DEFINE(M_NFSSVC, "nfsclient_srvsock", "Nfs server structure");
76 static void nfssvc_iod(void *);
78 static int nfs_asyncdaemon[NFS_MAXASYNCDAEMON];
80 SYSCTL_DECL(_vfs_nfs);
82 /* Maximum number of seconds a nfsiod kthread will sleep before exiting */
83 static unsigned int nfs_iodmaxidle = 120;
84 SYSCTL_UINT(_vfs_nfs, OID_AUTO, iodmaxidle, CTLFLAG_RW, &nfs_iodmaxidle, 0,
85 "Max number of seconds an nfsiod kthread will sleep before exiting");
87 /* Maximum number of nfsiod kthreads */
88 unsigned int nfs_iodmax = 20;
90 /* Minimum number of nfsiod kthreads to keep as spares */
91 static unsigned int nfs_iodmin = 0;
94 sysctl_iodmin(SYSCTL_HANDLER_ARGS)
100 error = sysctl_handle_int(oidp, &newmin, 0, req);
101 if (error || (req->newptr == NULL))
103 mtx_lock(&nfs_iod_mtx);
104 if (newmin > nfs_iodmax) {
109 if (nfs_numasync >= nfs_iodmin)
112 * If the current number of nfsiod is lower
113 * than the new minimum, create some more.
115 for (i = nfs_iodmin - nfs_numasync; i > 0; i--)
118 mtx_unlock(&nfs_iod_mtx);
121 SYSCTL_PROC(_vfs_nfs, OID_AUTO, iodmin, CTLTYPE_UINT | CTLFLAG_RW, 0,
122 sizeof (nfs_iodmin), sysctl_iodmin, "IU",
123 "Min number of nfsiod kthreads to keep as spares");
127 sysctl_iodmax(SYSCTL_HANDLER_ARGS)
133 error = sysctl_handle_int(oidp, &newmax, 0, req);
134 if (error || (req->newptr == NULL))
136 if (newmax > NFS_MAXASYNCDAEMON)
138 mtx_lock(&nfs_iod_mtx);
140 if (nfs_numasync <= nfs_iodmax)
143 * If there are some asleep nfsiods that should
144 * exit, wakeup() them so that they check nfs_iodmax
145 * and exit. Those who are active will exit as
146 * soon as they finish I/O.
148 iod = nfs_numasync - 1;
149 for (i = 0; i < nfs_numasync - nfs_iodmax; i++) {
150 if (nfs_iodwant[iod] == NFSIOD_AVAILABLE)
151 wakeup(&nfs_iodwant[iod]);
155 mtx_unlock(&nfs_iod_mtx);
158 SYSCTL_PROC(_vfs_nfs, OID_AUTO, iodmax, CTLTYPE_UINT | CTLFLAG_RW, 0,
159 sizeof (nfs_iodmax), sysctl_iodmax, "IU",
160 "Max number of nfsiod kthreads");
163 nfs_nfsiodnew(int set_iodwant)
168 if (nfs_numasync >= nfs_iodmax)
171 for (i = 0; i < nfs_iodmax; i++)
172 if (nfs_asyncdaemon[i] == 0) {
173 nfs_asyncdaemon[i]++;
180 nfs_iodwant[i] = NFSIOD_CREATED_FOR_NFS_ASYNCIO;
181 mtx_unlock(&nfs_iod_mtx);
182 error = kproc_create(nfssvc_iod, nfs_asyncdaemon + i, NULL, RFHIGHPID,
183 0, "nfsiod %d", newiod);
184 mtx_lock(&nfs_iod_mtx);
187 nfs_iodwant[i] = NFSIOD_NOT_AVAILABLE;
195 nfsiod_setup(void *dummy)
200 TUNABLE_INT_FETCH("vfs.nfs.iodmin", &nfs_iodmin);
201 mtx_lock(&nfs_iod_mtx);
202 /* Silently limit the start number of nfsiod's */
203 if (nfs_iodmin > NFS_MAXASYNCDAEMON)
204 nfs_iodmin = NFS_MAXASYNCDAEMON;
206 for (i = 0; i < nfs_iodmin; i++) {
207 error = nfs_nfsiodnew(0);
209 panic("nfsiod_setup: nfs_nfsiodnew failed");
211 mtx_unlock(&nfs_iod_mtx);
213 SYSINIT(nfsiod, SI_SUB_KTHREAD_IDLE, SI_ORDER_ANY, nfsiod_setup, NULL);
215 static int nfs_defect = 0;
216 SYSCTL_INT(_vfs_nfs, OID_AUTO, defect, CTLFLAG_RW, &nfs_defect, 0,
217 "Allow nfsiods to migrate serving different mounts");
220 * Asynchronous I/O daemons for client nfs.
221 * They do read-ahead and write-behind operations on the block I/O cache.
222 * Returns if we hit the timeout defined by the iodmaxidle sysctl.
225 nfssvc_iod(void *instance)
228 struct nfsmount *nmp;
232 mtx_lock(&nfs_iod_mtx);
233 myiod = (int *)instance - nfs_asyncdaemon;
238 while (((nmp = nfs_iodmount[myiod]) == NULL)
239 || !TAILQ_FIRST(&nmp->nm_bufq)) {
240 if (myiod >= nfs_iodmax)
244 if (nfs_iodwant[myiod] == NFSIOD_NOT_AVAILABLE)
245 nfs_iodwant[myiod] = NFSIOD_AVAILABLE;
246 nfs_iodmount[myiod] = NULL;
248 * Always keep at least nfs_iodmin kthreads.
250 timo = (myiod < nfs_iodmin) ? 0 : nfs_iodmaxidle * hz;
251 error = msleep(&nfs_iodwant[myiod], &nfs_iod_mtx, PWAIT | PCATCH,
254 nmp = nfs_iodmount[myiod];
256 * Rechecking the nm_bufq closes a rare race where the
257 * nfsiod is woken up at the exact time the idle timeout
260 if (nmp && TAILQ_FIRST(&nmp->nm_bufq))
267 while ((bp = TAILQ_FIRST(&nmp->nm_bufq)) != NULL) {
268 int giant_locked = 0;
270 /* Take one off the front of the list */
271 TAILQ_REMOVE(&nmp->nm_bufq, bp, b_freelist);
273 if (nmp->nm_bufqwant && nmp->nm_bufqlen <= nfs_numasync) {
274 nmp->nm_bufqwant = 0;
275 wakeup(&nmp->nm_bufq);
277 mtx_unlock(&nfs_iod_mtx);
278 if (NFS_ISV4(bp->b_vp)) {
282 if (bp->b_flags & B_DIRECT) {
283 KASSERT((bp->b_iocmd == BIO_WRITE), ("nfscvs_iod: BIO_WRITE not set"));
284 (void)nfs_doio_directwrite(bp);
286 if (bp->b_iocmd == BIO_READ)
287 (void) nfs_doio(bp->b_vp, bp, bp->b_rcred, NULL);
289 (void) nfs_doio(bp->b_vp, bp, bp->b_wcred, NULL);
293 mtx_lock(&nfs_iod_mtx);
295 * If there are more than one iod on this mount, then defect
296 * so that the iods can be shared out fairly between the mounts
298 if (nfs_defect && nmp->nm_bufqiods > 1) {
300 ("nfssvc_iod: iod %d defecting from mount %p\n",
302 nfs_iodmount[myiod] = NULL;
309 nfs_asyncdaemon[myiod] = 0;
312 nfs_iodwant[myiod] = NFSIOD_NOT_AVAILABLE;
313 nfs_iodmount[myiod] = NULL;
314 /* Someone may be waiting for the last nfsiod to terminate. */
315 if (--nfs_numasync == 0)
316 wakeup(&nfs_numasync);
317 mtx_unlock(&nfs_iod_mtx);
318 if ((error == 0) || (error == EWOULDBLOCK))
320 /* Abnormal termination */