nfs_lock.c revision 75631
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 * $FreeBSD: head/sys/nfsclient/nfs_lock.c 75631 2001-04-17 20:45:23Z alfred $ 30 */ 31 32#include <sys/param.h> 33#include <sys/systm.h> 34#include <sys/fcntl.h> 35#include <sys/malloc.h> 36#include <sys/mbuf.h> 37#include <sys/mount.h> 38#include <sys/namei.h> 39#include <sys/proc.h> 40#include <sys/socket.h> 41#include <sys/unistd.h> 42#include <sys/vnode.h> 43#include <sys/resourcevar.h> 44#include <sys/kernel.h> /* for hz */ 45#include <sys/lockf.h> /* for hz */ 46 47#include <net/if.h> 48 49#include <nfs/rpcv2.h> 50#include <nfs/nfsproto.h> 51#include <nfs/nfs.h> 52#include <nfs/nfsmount.h> 53#include <nfs/nfsnode.h> 54#include <nfs/nfs_lock.h> 55#include <nfs/nlminfo.h> 56 57#define NFSOWNER_1ST_LEVEL_START 1 /* initial entries */ 58#define NFSOWNER_2ND_LEVEL 256 /* some power of 2 */ 59 60#define NFSOWNER(tbl, i) \ 61 (tbl)[(i) / NFSOWNER_2ND_LEVEL][(i) % NFSOWNER_2ND_LEVEL] 62 63/* 64 * XXX 65 * We have to let the process know if the call succeeded. I'm using an extra 66 * field in the p_nlminfo field in the proc structure, as it is already for 67 * lockd stuff. 68 */ 69 70/* 71 * nfs_advlock -- 72 * NFS advisory byte-level locks. 73 */ 74int 75nfs_dolock(ap) 76 struct vop_advlock_args /* { 77 struct vnode *a_vp; 78 caddr_t a_id; 79 int a_op; 80 struct flock *a_fl; 81 int a_flags; 82 } */ *ap; 83{ 84 LOCKD_MSG msg; 85 struct nameidata nd; 86 struct proc *p; 87 uid_t saved_uid; 88 struct vnode *vp, *wvp; 89 int error, error1; 90 struct flock *fl; 91 int fmode, ioflg; 92 93 p = curproc; 94 vp = ap->a_vp; 95 fl = ap->a_fl; 96 97 /* 98 * the NLM protocol doesn't allow the server to return an error 99 * on ranges, so we do it. Note that we should be returning 100 * EOVERFLOW in some cases, but we don't have it. 101 */ 102 if (fl->l_start < 0 || fl->l_len < 0 || 103 ((fl->l_len != 0 && 104 (fl->l_start + fl->l_len - 1) < 0))) 105 return (EINVAL); 106 107 /* 108 * Fill in the information structure. 109 */ 110 msg.lm_version = LOCKD_MSG_VERSION; 111 msg.lm_msg_ident.pid = p->p_pid; 112 /* 113 * if there is no nfsowner table yet, allocate one. 114 */ 115 if (p->p_nlminfo == NULL) { 116 MALLOC(p->p_nlminfo, struct nlminfo *, 117 sizeof(struct nlminfo), M_LOCKF, M_WAITOK | M_ZERO); 118 p->p_nlminfo->pid_start = p->p_stats->p_start; 119 } 120 msg.lm_msg_ident.pid_start = p->p_nlminfo->pid_start; 121 msg.lm_msg_ident.msg_seq = ++(p->p_nlminfo->msg_seq); 122 123 msg.lm_fl = *fl; 124 msg.lm_wait = ap->a_flags & F_WAIT; 125 msg.lm_getlk = ap->a_op == F_GETLK; 126 /* 127 * XXX -- I think this is wrong for anything other AF_INET. 128 */ 129 msg.lm_addr = *(VFSTONFS(vp->v_mount)->nm_nam); 130 msg.lm_fh_len = NFS_ISV3(vp) ? VTONFS(vp)->n_fhsize : NFSX_V2FH; 131 bcopy(VTONFS(vp)->n_fhp, msg.lm_fh, msg.lm_fh_len); 132 msg.lm_nfsv3 = NFS_ISV3(vp); 133 msg.lm_cred = *(p->p_ucred); 134 135 /* 136 * Open the lock fifo. If for any reason we don't find the fifo, it 137 * means that the lock daemon isn't running. Translate any missing 138 * file error message for the user, otherwise the application will 139 * complain that the user's file is missing, which isn't the case. 140 * Note that we use proc0's cred, so the fifo is opened as root. 141 */ 142 NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, _PATH_LCKFIFO, p); 143 144 /* 145 * XXX Hack to temporarily allow this process (regardless of it's creds) 146 * to open the fifo we need to write to. vn_open() really should 147 * take a ucred (and once it does, this code should be fixed to use 148 * proc0's ucred. 149 */ 150 saved_uid = p->p_ucred->cr_uid; 151 p->p_ucred->cr_uid = 0; /* temporarly run the vn_open as root */ 152 153 fmode = FFLAGS(O_WRONLY); 154 error = vn_open(&nd, &fmode, 0); 155 p->p_ucred->cr_uid = saved_uid; 156 if (error != 0) { 157 return (error == ENOENT ? EOPNOTSUPP : error); 158 } 159 wvp = nd.ni_vp; 160 VOP_UNLOCK(wvp, 0, p); /* vn_open leaves it locked */ 161 162 163 ioflg = IO_UNIT; 164 for (;;) { 165 VOP_LEASE(wvp, p, proc0.p_ucred, LEASE_WRITE); 166 167 error = vn_rdwr(UIO_WRITE, wvp, (caddr_t)&msg, sizeof(msg), 0, 168 UIO_SYSSPACE, ioflg, proc0.p_ucred, NULL, p); 169 170 if (error && (((ioflg & IO_NDELAY) == 0) || error != EAGAIN)) { 171 break; 172 } 173 /* 174 * If we're locking a file, wait for an answer. Unlocks succeed 175 * immediately. 176 */ 177 if (fl->l_type == F_UNLCK) 178 /* 179 * XXX this isn't exactly correct. The client side 180 * needs to continue sending it's unlock until 181 * it gets a responce back. 182 */ 183 break; 184 185 /* 186 * retry after 20 seconds if we haven't gotten a responce yet. 187 * This number was picked out of thin air... but is longer 188 * then even a reasonably loaded system should take (at least 189 * on a local network). XXX Probably should use a back-off 190 * scheme. 191 */ 192 if ((error = tsleep((void *)p->p_nlminfo, 193 PCATCH | PUSER, "lockd", 20*hz)) != 0) { 194 if (error == EWOULDBLOCK) { 195 /* 196 * We timed out, so we rewrite the request 197 * to the fifo, but only if it isn't already 198 * full. 199 */ 200 ioflg |= IO_NDELAY; 201 continue; 202 } 203 204 break; 205 } 206 207 if (msg.lm_getlk && p->p_nlminfo->retcode == 0) { 208 if (p->p_nlminfo->set_getlk_pid) { 209 fl->l_pid = p->p_nlminfo->getlk_pid; 210 } else { 211 fl->l_type = F_UNLCK; 212 } 213 } 214 error = p->p_nlminfo->retcode; 215 break; 216 } 217 218 if ((error1 = vn_close(wvp, FWRITE, proc0.p_ucred, p)) && error == 0) 219 return (error1); 220 221 return (error); 222} 223 224/* 225 * nfslockdans -- 226 * NFS advisory byte-level locks answer from the lock daemon. 227 */ 228int 229nfslockdans(p, ansp) 230 struct proc *p; 231 struct lockd_ans *ansp; 232{ 233 int error; 234 235 /* Let root, or someone who once was root (lockd generally 236 * switches to the daemon uid once it is done setting up) make 237 * this call 238 */ 239 if ((error = suser(p)) != 0 && p->p_cred->p_svuid != 0) 240 return (error); 241 242 /* the version should match, or we're out of sync */ 243 if (ansp->la_vers != LOCKD_ANS_VERSION) 244 return (EINVAL); 245 246 /* Find the process, set its return errno and wake it up. */ 247 if ((p = pfind(ansp->la_msg_ident.pid)) == NULL) 248 return (ESRCH); 249 250 /* verify the pid hasn't been reused (if we can), and it isn't waiting 251 * for an answer from a more recent request. We return an EPIPE if 252 * the match fails, because we've already used ESRCH above, and this 253 * is sort of like writing on a pipe after the reader has closed it. 254 */ 255 if (p->p_nlminfo == NULL || 256 ((ansp->la_msg_ident.msg_seq != -1) && 257 (timevalcmp(&p->p_nlminfo->pid_start, 258 &ansp->la_msg_ident.pid_start, !=) || 259 p->p_nlminfo->msg_seq != ansp->la_msg_ident.msg_seq))) 260 return (EPIPE); 261 262 p->p_nlminfo->retcode = ansp->la_errno; 263 p->p_nlminfo->set_getlk_pid = ansp->la_set_getlk_pid; 264 p->p_nlminfo->getlk_pid = ansp->la_getlk_pid; 265 266 (void)wakeup((void *)p->p_nlminfo); 267 268 return (0); 269} 270