sys/vfs/nfs/nfs_iod.c

   1 /*
   2  * Copyright (c) 2009 The DragonFly Project.  All rights reserved.
   3  *
   4  * This code is derived from software contributed to The DragonFly Project
   5  * by Matthew Dillon <dillon@backplane.com>
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  *
  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
  15  *    the documentation and/or other materials provided with the
  16  *    distribution.
  17  * 3. Neither the name of The DragonFly Project nor the names of its
  18  *    contributors may be used to endorse or promote products derived
  19  *    from this software without specific, prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
  25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
  27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
  29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  32  * SUCH DAMAGE.
  33  */
  34 /*
  35  * NFSIOD operations - now built into the kernel.
  36  */
  37 #include <sys/param.h>
  38 #include <sys/systm.h>
  39 #include <sys/proc.h>
  40 #include <sys/malloc.h>
  41 #include <sys/mount.h>
  42 #include <sys/kernel.h>
  43 #include <sys/mbuf.h>
  44 #include <sys/vnode.h>
  45 #include <sys/fcntl.h>
  46 #include <sys/protosw.h>
  47 #include <sys/resourcevar.h>
  48 #include <sys/socket.h>
  49 #include <sys/socketvar.h>
  50 #include <sys/socketops.h>
  51 #include <sys/syslog.h>
  52 #include <sys/thread.h>
  53 #include <sys/tprintf.h>
  54 #include <sys/sysctl.h>
  55 #include <sys/signalvar.h>
  56 #include <sys/mutex.h>
  57
  58 #include <sys/signal2.h>
  59 #include <sys/mutex2.h>
  60
  61 #include <netinet/in.h>
  62 #include <netinet/tcp.h>
  63 #include <sys/thread2.h>
  64
  65 #include "rpcv2.h"
  66 #include "nfsproto.h"
  67 #include "nfs.h"
  68 #include "xdr_subs.h"
  69 #include "nfsm_subs.h"
  70 #include "nfsmount.h"
  71 #include "nfsnode.h"
  72 #include "nfsrtt.h"
  73
  74 void
  75 nfssvc_iod_reader(void *arg)
  76 {
  77         struct nfsmount *nmp = arg;
  78         struct nfsm_info *info;
  79         struct nfsreq *req;
  80         int error;
  81
  82         if (nmp->nm_rxstate == NFSSVC_INIT)
  83                 nmp->nm_rxstate = NFSSVC_PENDING;
  84         crit_enter();
  85         for (;;) {
  86                 if (nmp->nm_rxstate == NFSSVC_WAITING) {
  87                         if (TAILQ_FIRST(&nmp->nm_reqq) == NULL &&
  88                             TAILQ_FIRST(&nmp->nm_reqrxq) == NULL) {
  89                                 tsleep(&nmp->nm_rxstate, 0, "nfsidl", 0);
  90                         } else {
  91                                 /*
  92                                  * This can happen during shutdown, we don't
  93                                  * want to hardloop.
  94                                  */
  95                                 error = nfs_reply(nmp, NULL);
  96                                 if (error && error != EWOULDBLOCK) {
  97                                         tsleep(&nmp->nm_rxstate, 0,
  98                                                 "nfsxxx", hz / 10);
  99                                 }
 100                         }
 101                         continue;
 102                 }
 103                 if (nmp->nm_rxstate != NFSSVC_PENDING)
 104                         break;
 105                 nmp->nm_rxstate = NFSSVC_WAITING;
 106
 107                 /*
 108                  * Process requests which have received replies.  Only
 109                  * process the post-reply states.  If we get EINPROGRESS
 110                  * it means the request went back to an auth or retransmit
 111                  * state and we let the iod_writer thread deal with it.
 112                  *
 113                  * Any lock on the request is strictly temporary due to
 114                  * MP races (XXX).
 115                  *
 116                  * If the request completes we run the info->done call
 117                  * to finish up the I/O.
 118                  */
 119                 while ((req = TAILQ_FIRST(&nmp->nm_reqrxq)) != NULL) {
 120                         if (req->r_flags & R_LOCKED) {
 121                                 while (req->r_flags & R_LOCKED) {
 122                                         req->r_flags |= R_WANTED;
 123                                         tsleep(req, 0, "nfstrac", 0);
 124                                 }
 125                                 continue;
 126                         }
 127                         TAILQ_REMOVE(&nmp->nm_reqrxq, req, r_chain);
 128                         crit_exit();
 129                         info = req->r_info;
 130                         KKASSERT(info);
 131                         info->error = nfs_request(info,
 132                                                   NFSM_STATE_PROCESSREPLY,
 133                                                   NFSM_STATE_DONE);
 134                         if (info->error == EINPROGRESS) {
 135                                 kprintf("rxq: move info %p back to txq\n", info);
 136                                 TAILQ_INSERT_TAIL(&nmp->nm_reqtxq, req, r_chain);
 137                                 nfssvc_iod_writer_wakeup(nmp);
 138                         } else {
 139                                 atomic_subtract_int(&nmp->nm_bioqlen, 1);
 140                                 info->done(info);
 141                         }
 142                         crit_enter();
 143                 }
 144         }
 145         crit_exit();
 146         nmp->nm_rxthread = NULL;
 147         nmp->nm_rxstate = NFSSVC_DONE;
 148         wakeup(&nmp->nm_rxthread);
 149 }
 150
 151 /*
 152  * The writer sits on the send side of the client's socket and
 153  * does both the initial processing of BIOs and also transmission
 154  * and retransmission of nfsreq's.
 155  *
 156  * The writer processes both new BIOs from nm_bioq and retransmit
 157  * or state machine jumpbacks from nm_reqtxq
 158  */
 159 void
 160 nfssvc_iod_writer(void *arg)
 161 {
 162         struct nfsmount *nmp = arg;
 163         struct bio *bio;
 164         struct nfsreq *req;
 165         struct vnode *vp;
 166         nfsm_info_t info;
 167
 168         if (nmp->nm_txstate == NFSSVC_INIT)
 169                 nmp->nm_txstate = NFSSVC_PENDING;
 170         crit_enter();
 171         for (;;) {
 172                 if (nmp->nm_txstate == NFSSVC_WAITING) {
 173                         tsleep(&nmp->nm_txstate, 0, "nfsidl", 0);
 174                         continue;
 175                 }
 176                 if (nmp->nm_txstate != NFSSVC_PENDING)
 177                         break;
 178                 nmp->nm_txstate = NFSSVC_WAITING;
 179
 180                 /*
 181                  * Eep, we could blow out the mbuf allocator if we just
 182                  * did everything the kernel wanted us to do.
 183                  */
 184                 while ((bio = TAILQ_FIRST(&nmp->nm_bioq)) != NULL) {
 185                         if (nmp->nm_reqqlen >= NFS_MAXASYNCBIO)
 186                                 break;
 187                         TAILQ_REMOVE(&nmp->nm_bioq, bio, bio_act);
 188                         vp = bio->bio_driver_info;
 189                         crit_exit();
 190                         nfs_startio(vp, bio, NULL);
 191                         crit_enter();
 192                 }
 193
 194                 /*
 195                  * Process reauths & retransmits.  If we get an EINPROGRESS
 196                  * it means the state transitioned to WAITREPLY or later.
 197                  * Otherwise the request completed (probably with an error
 198                  * since we didn't get to a replied state).
 199                  */
 200                 while ((req = TAILQ_FIRST(&nmp->nm_reqtxq)) != NULL) {
 201                         TAILQ_REMOVE(&nmp->nm_reqtxq, req, r_chain);
 202                         info = req->r_info;
 203                         KKASSERT(info);
 204                         crit_exit();
 205                         info->error = nfs_request(info,
 206                                                   NFSM_STATE_AUTH,
 207                                                   NFSM_STATE_WAITREPLY);
 208                         crit_enter();
 209                         if (info->error == EINPROGRESS) {
 210                                 ;
 211                         } else {
 212                                 atomic_subtract_int(&nmp->nm_bioqlen, 1);
 213                                 info->done(info);
 214                         }
 215                 }
 216         }
 217         crit_exit();
 218         nmp->nm_txthread = NULL;
 219         nmp->nm_txstate = NFSSVC_DONE;
 220         wakeup(&nmp->nm_txthread);
 221 }
 222
 223 void
 224 nfssvc_iod_stop1(struct nfsmount *nmp)
 225 {
 226         crit_enter();
 227         nmp->nm_txstate = NFSSVC_STOPPING;
 228         nmp->nm_rxstate = NFSSVC_STOPPING;
 229         crit_exit();
 230 }
 231
 232 void
 233 nfssvc_iod_stop2(struct nfsmount *nmp)
 234 {
 235         wakeup(&nmp->nm_txstate);
 236         while (nmp->nm_txthread)
 237                 tsleep(&nmp->nm_txthread, 0, "nfssttx", hz*2);
 238         wakeup(&nmp->nm_rxstate);
 239         while (nmp->nm_rxthread)
 240                 tsleep(&nmp->nm_rxthread, 0, "nfsstrx", hz*2);
 241 }
 242
 243 void
 244 nfssvc_iod_writer_wakeup(struct nfsmount *nmp)
 245 {
 246         if (nmp->nm_txstate == NFSSVC_WAITING) {
 247                 nmp->nm_txstate = NFSSVC_PENDING;
 248                 wakeup(&nmp->nm_txstate);
 249         }
 250 }
 251
 252 void
 253 nfssvc_iod_reader_wakeup(struct nfsmount *nmp)
 254 {
 255         if (nmp->nm_rxstate == NFSSVC_WAITING) {
 256                 nmp->nm_rxstate = NFSSVC_PENDING;
 257                 wakeup(&nmp->nm_rxstate);
 258         }
 259 }