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 * from 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 <fs/nfs/nfsport.h>
67 #include <fs/nfsclient/nfsmount.h>
68 #include <fs/nfsclient/nfs.h>
69 #include <fs/nfsclient/nfsnode.h>
71 extern struct mtx ncl_iod_mutex;
74 enum nfsiod_state ncl_iodwant[NFS_MAXRAHEAD];
75 struct nfsmount *ncl_iodmount[NFS_MAXRAHEAD];
77 static void nfssvc_iod(void *);
79 static int nfs_asyncdaemon[NFS_MAXRAHEAD];
81 SYSCTL_DECL(_vfs_newnfs);
83 /* Maximum number of seconds a nfsiod kthread will sleep before exiting */
84 static unsigned int ncl_iodmaxidle = 120;
85 SYSCTL_UINT(_vfs_newnfs, OID_AUTO, iodmaxidle, CTLFLAG_RW, &ncl_iodmaxidle, 0, "");
87 /* Maximum number of nfsiod kthreads */
88 unsigned int ncl_iodmax = NFS_MAXRAHEAD;
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(&ncl_iod_mutex);
104 if (newmin > ncl_iodmax) {
109 if (ncl_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 - ncl_numasync; i > 0; i--)
118 mtx_unlock(&ncl_iod_mutex);
121 SYSCTL_PROC(_vfs_newnfs, OID_AUTO, iodmin, CTLTYPE_UINT | CTLFLAG_RW, 0,
122 sizeof (nfs_iodmin), sysctl_iodmin, "IU", "");
126 sysctl_iodmax(SYSCTL_HANDLER_ARGS)
132 error = sysctl_handle_int(oidp, &newmax, 0, req);
133 if (error || (req->newptr == NULL))
135 if (newmax > NFS_MAXRAHEAD)
137 mtx_lock(&ncl_iod_mutex);
139 if (ncl_numasync <= ncl_iodmax)
142 * If there are some asleep nfsiods that should
143 * exit, wakeup() them so that they check ncl_iodmax
144 * and exit. Those who are active will exit as
145 * soon as they finish I/O.
147 iod = ncl_numasync - 1;
148 for (i = 0; i < ncl_numasync - ncl_iodmax; i++) {
149 if (ncl_iodwant[iod] == NFSIOD_AVAILABLE)
150 wakeup(&ncl_iodwant[iod]);
154 mtx_unlock(&ncl_iod_mutex);
157 SYSCTL_PROC(_vfs_newnfs, OID_AUTO, iodmax, CTLTYPE_UINT | CTLFLAG_RW, 0,
158 sizeof (ncl_iodmax), sysctl_iodmax, "IU", "");
161 ncl_nfsiodnew(int set_iodwant)
166 if (ncl_numasync >= ncl_iodmax)
169 for (i = 0; i < ncl_iodmax; i++)
170 if (nfs_asyncdaemon[i] == 0) {
171 nfs_asyncdaemon[i]++;
178 ncl_iodwant[i] = NFSIOD_CREATED_FOR_NFS_ASYNCIO;
179 mtx_unlock(&ncl_iod_mutex);
180 error = kproc_create(nfssvc_iod, nfs_asyncdaemon + i, NULL, RFHIGHPID,
181 0, "nfsiod %d", newiod);
182 mtx_lock(&ncl_iod_mutex);
185 ncl_iodwant[i] = NFSIOD_NOT_AVAILABLE;
193 nfsiod_setup(void *dummy)
198 TUNABLE_INT_FETCH("vfs.newnfs.iodmin", &nfs_iodmin);
200 mtx_lock(&ncl_iod_mutex);
201 /* Silently limit the start number of nfsiod's */
202 if (nfs_iodmin > NFS_MAXRAHEAD)
203 nfs_iodmin = NFS_MAXRAHEAD;
205 for (i = 0; i < nfs_iodmin; i++) {
206 error = ncl_nfsiodnew(0);
208 panic("newnfsiod_setup: ncl_nfsiodnew failed");
210 mtx_unlock(&ncl_iod_mutex);
212 SYSINIT(newnfsiod, SI_SUB_KTHREAD_IDLE, SI_ORDER_ANY, nfsiod_setup, NULL);
214 static int nfs_defect = 0;
215 SYSCTL_INT(_vfs_newnfs, OID_AUTO, defect, CTLFLAG_RW, &nfs_defect, 0, "");
218 * Asynchronous I/O daemons for client nfs.
219 * They do read-ahead and write-behind operations on the block I/O cache.
220 * Returns if we hit the timeout defined by the iodmaxidle sysctl.
223 nfssvc_iod(void *instance)
226 struct nfsmount *nmp;
230 mtx_lock(&ncl_iod_mutex);
231 myiod = (int *)instance - nfs_asyncdaemon;
236 while (((nmp = ncl_iodmount[myiod]) == NULL)
237 || !TAILQ_FIRST(&nmp->nm_bufq)) {
238 if (myiod >= ncl_iodmax)
242 if (ncl_iodwant[myiod] == NFSIOD_NOT_AVAILABLE)
243 ncl_iodwant[myiod] = NFSIOD_AVAILABLE;
244 ncl_iodmount[myiod] = NULL;
246 * Always keep at least nfs_iodmin kthreads.
248 timo = (myiod < nfs_iodmin) ? 0 : ncl_iodmaxidle * hz;
249 error = msleep(&ncl_iodwant[myiod], &ncl_iod_mutex, PWAIT | PCATCH,
252 nmp = ncl_iodmount[myiod];
254 * Rechecking the nm_bufq closes a rare race where the
255 * nfsiod is woken up at the exact time the idle timeout
258 if (nmp && TAILQ_FIRST(&nmp->nm_bufq))
265 while ((bp = TAILQ_FIRST(&nmp->nm_bufq)) != NULL) {
267 /* Take one off the front of the list */
268 TAILQ_REMOVE(&nmp->nm_bufq, bp, b_freelist);
270 if (nmp->nm_bufqwant && nmp->nm_bufqlen <= ncl_numasync) {
271 nmp->nm_bufqwant = 0;
272 wakeup(&nmp->nm_bufq);
274 mtx_unlock(&ncl_iod_mutex);
275 if (bp->b_flags & B_DIRECT) {
276 KASSERT((bp->b_iocmd == BIO_WRITE), ("nfscvs_iod: BIO_WRITE not set"));
277 (void)ncl_doio_directwrite(bp);
279 if (bp->b_iocmd == BIO_READ)
280 (void) ncl_doio(bp->b_vp, bp, bp->b_rcred,
283 (void) ncl_doio(bp->b_vp, bp, bp->b_wcred,
286 mtx_lock(&ncl_iod_mutex);
288 * If there are more than one iod on this mount, then defect
289 * so that the iods can be shared out fairly between the mounts
291 if (nfs_defect && nmp->nm_bufqiods > 1) {
293 ("nfssvc_iod: iod %d defecting from mount %p\n",
295 ncl_iodmount[myiod] = NULL;
302 nfs_asyncdaemon[myiod] = 0;
305 ncl_iodwant[myiod] = NFSIOD_NOT_AVAILABLE;
306 ncl_iodmount[myiod] = NULL;
307 /* Someone may be waiting for the last nfsiod to terminate. */
308 if (--ncl_numasync == 0)
309 wakeup(&ncl_numasync);
310 mtx_unlock(&ncl_iod_mutex);
311 if ((error == 0) || (error == EWOULDBLOCK))
313 /* Abnormal termination */