The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/nfsclient/nfs_krpc.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 1989, 1991, 1993, 1995
    3  *      The Regents of the University of California.  All rights reserved.
    4  *
    5  * This code is derived from software contributed to Berkeley by
    6  * Rick Macklem at The University of Guelph.
    7  *
    8  * Redistribution and use in source and binary forms, with or without
    9  * modification, are permitted provided that the following conditions
   10  * are met:
   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.
   19  *
   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
   30  * SUCH DAMAGE.
   31  *
   32  *      @(#)nfs_socket.c        8.5 (Berkeley) 3/30/95
   33  */
   34 
   35 #include <sys/cdefs.h>
   36 __FBSDID("$FreeBSD: releng/9.0/sys/nfsclient/nfs_krpc.c 223309 2011-06-19 22:08:55Z rmacklem $");
   37 
   38 /*
   39  * Socket operations for use by nfs
   40  */
   41 
   42 #include "opt_inet6.h"
   43 #include "opt_kdtrace.h"
   44 #include "opt_kgssapi.h"
   45 
   46 #include <sys/param.h>
   47 #include <sys/systm.h>
   48 #include <sys/kernel.h>
   49 #include <sys/limits.h>
   50 #include <sys/lock.h>
   51 #include <sys/malloc.h>
   52 #include <sys/mbuf.h>
   53 #include <sys/mount.h>
   54 #include <sys/mutex.h>
   55 #include <sys/proc.h>
   56 #include <sys/signalvar.h>
   57 #include <sys/syscallsubr.h>
   58 #include <sys/sysctl.h>
   59 #include <sys/syslog.h>
   60 #include <sys/vnode.h>
   61 
   62 #include <rpc/rpc.h>
   63 
   64 #include <nfs/nfsproto.h>
   65 #include <nfsclient/nfs.h>
   66 #include <nfs/xdr_subs.h>
   67 #include <nfsclient/nfsm_subs.h>
   68 #include <nfsclient/nfsmount.h>
   69 #include <nfsclient/nfsnode.h>
   70 
   71 #ifdef KDTRACE_HOOKS
   72 #include <sys/dtrace_bsd.h>
   73 
   74 dtrace_nfsclient_nfs23_start_probe_func_t
   75     dtrace_nfsclient_nfs23_start_probe;
   76 
   77 dtrace_nfsclient_nfs23_done_probe_func_t
   78     dtrace_nfsclient_nfs23_done_probe;
   79 
   80 /*
   81  * Registered probes by RPC type.
   82  */
   83 uint32_t        nfsclient_nfs2_start_probes[NFS_NPROCS];
   84 uint32_t        nfsclient_nfs2_done_probes[NFS_NPROCS];
   85 
   86 uint32_t        nfsclient_nfs3_start_probes[NFS_NPROCS];
   87 uint32_t        nfsclient_nfs3_done_probes[NFS_NPROCS];
   88 #endif
   89 
   90 static int      nfs_bufpackets = 4;
   91 static int      nfs_reconnects;
   92 static int      nfs3_jukebox_delay = 10;
   93 static int      nfs_skip_wcc_data_onerr = 1;
   94 static int      fake_wchan;
   95 
   96 SYSCTL_DECL(_vfs_oldnfs);
   97 
   98 SYSCTL_INT(_vfs_oldnfs, OID_AUTO, bufpackets, CTLFLAG_RW, &nfs_bufpackets, 0,
   99     "Buffer reservation size 2 < x < 64");
  100 SYSCTL_INT(_vfs_oldnfs, OID_AUTO, reconnects, CTLFLAG_RD, &nfs_reconnects, 0,
  101     "Number of times the nfs client has had to reconnect");
  102 SYSCTL_INT(_vfs_oldnfs, OID_AUTO, nfs3_jukebox_delay, CTLFLAG_RW,
  103     &nfs3_jukebox_delay, 0,
  104     "Number of seconds to delay a retry after receiving EJUKEBOX");
  105 SYSCTL_INT(_vfs_oldnfs, OID_AUTO, skip_wcc_data_onerr, CTLFLAG_RW,
  106     &nfs_skip_wcc_data_onerr, 0,
  107     "Disable weak cache consistency checking when server returns an error");
  108 
  109 static void     nfs_down(struct nfsmount *, struct thread *, const char *,
  110     int, int);
  111 static void     nfs_up(struct nfsmount *, struct thread *, const char *,
  112     int, int);
  113 static int      nfs_msg(struct thread *, const char *, const char *, int);
  114 
  115 extern int nfsv2_procid[];
  116 
  117 struct nfs_cached_auth {
  118         int             ca_refs; /* refcount, including 1 from the cache */
  119         uid_t           ca_uid;  /* uid that corresponds to this auth */
  120         AUTH            *ca_auth; /* RPC auth handle */
  121 };
  122 
  123 /*
  124  * RTT estimator
  125  */
  126 
  127 static enum nfs_rto_timer_t nfs_proct[NFS_NPROCS] = {
  128         NFS_DEFAULT_TIMER,      /* NULL */
  129         NFS_GETATTR_TIMER,      /* GETATTR */
  130         NFS_DEFAULT_TIMER,      /* SETATTR */
  131         NFS_LOOKUP_TIMER,       /* LOOKUP */
  132         NFS_GETATTR_TIMER,      /* ACCESS */
  133         NFS_READ_TIMER,         /* READLINK */
  134         NFS_READ_TIMER,         /* READ */
  135         NFS_WRITE_TIMER,        /* WRITE */
  136         NFS_DEFAULT_TIMER,      /* CREATE */
  137         NFS_DEFAULT_TIMER,      /* MKDIR */
  138         NFS_DEFAULT_TIMER,      /* SYMLINK */
  139         NFS_DEFAULT_TIMER,      /* MKNOD */
  140         NFS_DEFAULT_TIMER,      /* REMOVE */
  141         NFS_DEFAULT_TIMER,      /* RMDIR */
  142         NFS_DEFAULT_TIMER,      /* RENAME */
  143         NFS_DEFAULT_TIMER,      /* LINK */
  144         NFS_READ_TIMER,         /* READDIR */
  145         NFS_READ_TIMER,         /* READDIRPLUS */
  146         NFS_DEFAULT_TIMER,      /* FSSTAT */
  147         NFS_DEFAULT_TIMER,      /* FSINFO */
  148         NFS_DEFAULT_TIMER,      /* PATHCONF */
  149         NFS_DEFAULT_TIMER,      /* COMMIT */
  150         NFS_DEFAULT_TIMER,      /* NOOP */
  151 };
  152 
  153 /*
  154  * Choose the correct RTT timer for this NFS procedure.
  155  */
  156 static inline enum nfs_rto_timer_t
  157 nfs_rto_timer(u_int32_t procnum)
  158 {
  159 
  160         return (nfs_proct[procnum]);
  161 }
  162 
  163 /*
  164  * Initialize the RTT estimator state for a new mount point.
  165  */
  166 static void
  167 nfs_init_rtt(struct nfsmount *nmp)
  168 {
  169         int i;
  170 
  171         for (i = 0; i < NFS_MAX_TIMER; i++) {
  172                 nmp->nm_timers[i].rt_srtt = hz;
  173                 nmp->nm_timers[i].rt_deviate = 0;
  174                 nmp->nm_timers[i].rt_rtxcur = hz;
  175         }
  176 }
  177 
  178 /*
  179  * Initialize sockets and congestion for a new NFS connection.
  180  * We do not free the sockaddr if error.
  181  */
  182 int
  183 nfs_connect(struct nfsmount *nmp)
  184 {
  185         int rcvreserve, sndreserve;
  186         int pktscale;
  187         struct sockaddr *saddr;
  188         struct ucred *origcred;
  189         struct thread *td = curthread;
  190         CLIENT *client;
  191         struct netconfig *nconf;
  192         rpcvers_t vers;
  193         int one = 1, retries;
  194 
  195         /*
  196          * We need to establish the socket using the credentials of
  197          * the mountpoint.  Some parts of this process (such as
  198          * sobind() and soconnect()) will use the curent thread's
  199          * credential instead of the socket credential.  To work
  200          * around this, temporarily change the current thread's
  201          * credential to that of the mountpoint.
  202          *
  203          * XXX: It would be better to explicitly pass the correct
  204          * credential to sobind() and soconnect().
  205          */
  206         origcred = td->td_ucred;
  207         td->td_ucred = nmp->nm_mountp->mnt_cred;
  208         saddr = nmp->nm_nam;
  209 
  210         vers = NFS_VER2;
  211         if (nmp->nm_flag & NFSMNT_NFSV3)
  212                 vers = NFS_VER3;
  213         else if (nmp->nm_flag & NFSMNT_NFSV4)
  214                 vers = NFS_VER4;
  215         if (saddr->sa_family == AF_INET)
  216                 if (nmp->nm_sotype == SOCK_DGRAM)
  217                         nconf = getnetconfigent("udp");
  218                 else
  219                         nconf = getnetconfigent("tcp");
  220         else
  221                 if (nmp->nm_sotype == SOCK_DGRAM)
  222                         nconf = getnetconfigent("udp6");
  223                 else
  224                         nconf = getnetconfigent("tcp6");
  225 
  226         /*
  227          * Get buffer reservation size from sysctl, but impose reasonable
  228          * limits.
  229          */
  230         pktscale = nfs_bufpackets;
  231         if (pktscale < 2)
  232                 pktscale = 2;
  233         if (pktscale > 64)
  234                 pktscale = 64;
  235         mtx_lock(&nmp->nm_mtx);
  236         if (nmp->nm_sotype == SOCK_DGRAM) {
  237                 sndreserve = (nmp->nm_wsize + NFS_MAXPKTHDR) * pktscale;
  238                 rcvreserve = (max(nmp->nm_rsize, nmp->nm_readdirsize) +
  239                     NFS_MAXPKTHDR) * pktscale;
  240         } else if (nmp->nm_sotype == SOCK_SEQPACKET) {
  241                 sndreserve = (nmp->nm_wsize + NFS_MAXPKTHDR) * pktscale;
  242                 rcvreserve = (max(nmp->nm_rsize, nmp->nm_readdirsize) +
  243                     NFS_MAXPKTHDR) * pktscale;
  244         } else {
  245                 if (nmp->nm_sotype != SOCK_STREAM)
  246                         panic("nfscon sotype");
  247                 sndreserve = (nmp->nm_wsize + NFS_MAXPKTHDR +
  248                     sizeof (u_int32_t)) * pktscale;
  249                 rcvreserve = (nmp->nm_rsize + NFS_MAXPKTHDR +
  250                     sizeof (u_int32_t)) * pktscale;
  251         }
  252         mtx_unlock(&nmp->nm_mtx);
  253 
  254         client = clnt_reconnect_create(nconf, saddr, NFS_PROG, vers,
  255             sndreserve, rcvreserve);
  256         CLNT_CONTROL(client, CLSET_WAITCHAN, "nfsreq");
  257         if (nmp->nm_flag & NFSMNT_INT)
  258                 CLNT_CONTROL(client, CLSET_INTERRUPTIBLE, &one);
  259         if (nmp->nm_flag & NFSMNT_RESVPORT)
  260                 CLNT_CONTROL(client, CLSET_PRIVPORT, &one);
  261         if (nmp->nm_flag & NFSMNT_SOFT)
  262                 retries = nmp->nm_retry;
  263         else
  264                 retries = INT_MAX;
  265         CLNT_CONTROL(client, CLSET_RETRIES, &retries);
  266 
  267         mtx_lock(&nmp->nm_mtx);
  268         if (nmp->nm_client) {
  269                 /*
  270                  * Someone else already connected.
  271                  */
  272                 CLNT_RELEASE(client);
  273         } else
  274                 nmp->nm_client = client;
  275 
  276         /*
  277          * Protocols that do not require connections may be optionally left
  278          * unconnected for servers that reply from a port other than NFS_PORT.
  279          */
  280         if (!(nmp->nm_flag & NFSMNT_NOCONN)) {
  281                 mtx_unlock(&nmp->nm_mtx);
  282                 CLNT_CONTROL(client, CLSET_CONNECT, &one);
  283         } else
  284                 mtx_unlock(&nmp->nm_mtx);
  285 
  286         /* Restore current thread's credentials. */
  287         td->td_ucred = origcred;
  288 
  289         mtx_lock(&nmp->nm_mtx);
  290         /* Initialize other non-zero congestion variables. */
  291         nfs_init_rtt(nmp);
  292         mtx_unlock(&nmp->nm_mtx);
  293         return (0);
  294 }
  295 
  296 /*
  297  * NFS disconnect.  Clean up and unlink.
  298  */
  299 void
  300 nfs_disconnect(struct nfsmount *nmp)
  301 {
  302         CLIENT *client;
  303 
  304         mtx_lock(&nmp->nm_mtx);
  305         if (nmp->nm_client) {
  306                 client = nmp->nm_client;
  307                 nmp->nm_client = NULL;
  308                 mtx_unlock(&nmp->nm_mtx);
  309                 rpc_gss_secpurge_call(client);
  310                 CLNT_CLOSE(client);
  311                 CLNT_RELEASE(client);
  312         } else
  313                 mtx_unlock(&nmp->nm_mtx);
  314 }
  315 
  316 void
  317 nfs_safedisconnect(struct nfsmount *nmp)
  318 {
  319 
  320         nfs_disconnect(nmp);
  321 }
  322 
  323 static AUTH *
  324 nfs_getauth(struct nfsmount *nmp, struct ucred *cred)
  325 {
  326         rpc_gss_service_t svc;
  327         AUTH *auth;
  328 
  329         switch (nmp->nm_secflavor) {
  330         case RPCSEC_GSS_KRB5:
  331         case RPCSEC_GSS_KRB5I:
  332         case RPCSEC_GSS_KRB5P:
  333                 if (!nmp->nm_mech_oid)
  334                         if (!rpc_gss_mech_to_oid_call("kerberosv5",
  335                             &nmp->nm_mech_oid))
  336                                 return (NULL);
  337                 if (nmp->nm_secflavor == RPCSEC_GSS_KRB5)
  338                         svc = rpc_gss_svc_none;
  339                 else if (nmp->nm_secflavor == RPCSEC_GSS_KRB5I)
  340                         svc = rpc_gss_svc_integrity;
  341                 else
  342                         svc = rpc_gss_svc_privacy;
  343                 auth = rpc_gss_secfind_call(nmp->nm_client, cred,
  344                     nmp->nm_principal, nmp->nm_mech_oid, svc);
  345                 if (auth)
  346                         return (auth);
  347                 /* fallthrough */
  348         case AUTH_SYS:
  349         default:
  350                 return (authunix_create(cred));
  351 
  352         }
  353 }
  354 
  355 /*
  356  * Callback from the RPC code to generate up/down notifications.
  357  */
  358 
  359 struct nfs_feedback_arg {
  360         struct nfsmount *nf_mount;
  361         int             nf_lastmsg;     /* last tprintf */
  362         int             nf_tprintfmsg;
  363         struct thread   *nf_td;
  364 };
  365 
  366 static void
  367 nfs_feedback(int type, int proc, void *arg)
  368 {
  369         struct nfs_feedback_arg *nf = (struct nfs_feedback_arg *) arg;
  370         struct nfsmount *nmp = nf->nf_mount;
  371         struct timeval now;
  372 
  373         getmicrouptime(&now);
  374 
  375         switch (type) {
  376         case FEEDBACK_REXMIT2:
  377         case FEEDBACK_RECONNECT:
  378                 if (nf->nf_lastmsg + nmp->nm_tprintf_delay < now.tv_sec) {
  379                         nfs_down(nmp, nf->nf_td,
  380                             "not responding", 0, NFSSTA_TIMEO);
  381                         nf->nf_tprintfmsg = TRUE;
  382                         nf->nf_lastmsg = now.tv_sec;
  383                 }
  384                 break;
  385 
  386         case FEEDBACK_OK:
  387                 nfs_up(nf->nf_mount, nf->nf_td,
  388                     "is alive again", NFSSTA_TIMEO, nf->nf_tprintfmsg);
  389                 break;
  390         }
  391 }
  392 
  393 /*
  394  * nfs_request - goes something like this
  395  *      - fill in request struct
  396  *      - links it into list
  397  *      - calls nfs_send() for first transmit
  398  *      - calls nfs_receive() to get reply
  399  *      - break down rpc header and return with nfs reply pointed to
  400  *        by mrep or error
  401  * nb: always frees up mreq mbuf list
  402  */
  403 int
  404 nfs_request(struct vnode *vp, struct mbuf *mreq, int procnum,
  405     struct thread *td, struct ucred *cred, struct mbuf **mrp,
  406     struct mbuf **mdp, caddr_t *dposp)
  407 {
  408         struct mbuf *mrep;
  409         u_int32_t *tl;
  410         struct nfsmount *nmp;
  411         struct mbuf *md;
  412         time_t waituntil;
  413         caddr_t dpos;
  414         int error = 0;
  415         struct timeval now;
  416         AUTH *auth = NULL;
  417         enum nfs_rto_timer_t timer;
  418         struct nfs_feedback_arg nf;
  419         struct rpc_callextra ext;
  420         enum clnt_stat stat;
  421         struct timeval timo;
  422 
  423         /* Reject requests while attempting a forced unmount. */
  424         if (vp->v_mount->mnt_kern_flag & MNTK_UNMOUNTF) {
  425                 m_freem(mreq);
  426                 return (ESTALE);
  427         }
  428         nmp = VFSTONFS(vp->v_mount);
  429         bzero(&nf, sizeof(struct nfs_feedback_arg));
  430         nf.nf_mount = nmp;
  431         nf.nf_td = td;
  432         getmicrouptime(&now);
  433         nf.nf_lastmsg = now.tv_sec -
  434             ((nmp->nm_tprintf_delay) - (nmp->nm_tprintf_initial_delay));
  435 
  436         /*
  437          * XXX if not already connected call nfs_connect now.  Longer
  438          * term, change nfs_mount to call nfs_connect unconditionally
  439          * and let clnt_reconnect_create handle reconnects.
  440          */
  441         if (!nmp->nm_client)
  442                 nfs_connect(nmp);
  443 
  444         auth = nfs_getauth(nmp, cred);
  445         if (!auth) {
  446                 m_freem(mreq);
  447                 return (EACCES);
  448         }
  449         bzero(&ext, sizeof(ext));
  450         ext.rc_auth = auth;
  451 
  452         ext.rc_feedback = nfs_feedback;
  453         ext.rc_feedback_arg = &nf;
  454 
  455         /*
  456          * Use a conservative timeout for RPCs other than getattr,
  457          * lookup, read or write.  The justification for doing "other"
  458          * this way is that these RPCs happen so infrequently that
  459          * timer est. would probably be stale.  Also, since many of
  460          * these RPCs are non-idempotent, a conservative timeout is
  461          * desired.
  462          */
  463         timer = nfs_rto_timer(procnum);
  464         if (timer != NFS_DEFAULT_TIMER)
  465                 ext.rc_timers = &nmp->nm_timers[timer - 1];
  466         else
  467                 ext.rc_timers = NULL;
  468 
  469 #ifdef KDTRACE_HOOKS
  470         if (dtrace_nfsclient_nfs23_start_probe != NULL) {
  471                 uint32_t probe_id;
  472                 int probe_procnum;
  473 
  474                 if (nmp->nm_flag & NFSMNT_NFSV3) {
  475                         probe_id = nfsclient_nfs3_start_probes[procnum];
  476                         probe_procnum = procnum;
  477                 } else {
  478                         probe_id = nfsclient_nfs2_start_probes[procnum];
  479                         probe_procnum = nfsv2_procid[procnum];
  480                 }
  481                 if (probe_id != 0)
  482                         (dtrace_nfsclient_nfs23_start_probe)(probe_id, vp,
  483                             mreq, cred, probe_procnum);
  484         }
  485 #endif
  486 
  487         nfsstats.rpcrequests++;
  488 tryagain:
  489         timo.tv_sec = nmp->nm_timeo / NFS_HZ;
  490         timo.tv_usec = (nmp->nm_timeo * 1000000) / NFS_HZ;
  491         mrep = NULL;
  492         stat = CLNT_CALL_MBUF(nmp->nm_client, &ext,
  493             (nmp->nm_flag & NFSMNT_NFSV3) ? procnum : nfsv2_procid[procnum],
  494             mreq, &mrep, timo);
  495 
  496         /*
  497          * If there was a successful reply and a tprintf msg.
  498          * tprintf a response.
  499          */
  500         if (stat == RPC_SUCCESS)
  501                 error = 0;
  502         else if (stat == RPC_TIMEDOUT)
  503                 error = ETIMEDOUT;
  504         else if (stat == RPC_VERSMISMATCH)
  505                 error = EOPNOTSUPP;
  506         else if (stat == RPC_PROGVERSMISMATCH)
  507                 error = EPROTONOSUPPORT;
  508         else
  509                 error = EACCES;
  510         if (error)
  511                 goto nfsmout;
  512 
  513         KASSERT(mrep != NULL, ("mrep shouldn't be NULL if no error\n"));
  514 
  515         /*
  516          * Search for any mbufs that are not a multiple of 4 bytes long
  517          * or with m_data not longword aligned.
  518          * These could cause pointer alignment problems, so copy them to
  519          * well aligned mbufs.
  520          */
  521         error = nfs_realign(&mrep, M_DONTWAIT);
  522         if (error == ENOMEM) {
  523                 m_freem(mrep);
  524                 AUTH_DESTROY(auth);
  525                 return (error);
  526         }
  527 
  528         md = mrep;
  529         dpos = mtod(mrep, caddr_t);
  530         tl = nfsm_dissect(u_int32_t *, NFSX_UNSIGNED);
  531         if (*tl != 0) {
  532                 error = fxdr_unsigned(int, *tl);
  533                 if ((nmp->nm_flag & NFSMNT_NFSV3) &&
  534                     error == NFSERR_TRYLATER) {
  535                         m_freem(mrep);
  536                         error = 0;
  537                         waituntil = time_second + nfs3_jukebox_delay;
  538                         while (time_second < waituntil)
  539                                 (void)tsleep(&fake_wchan, PSOCK, "nqnfstry",
  540                                     hz);
  541                         goto tryagain;
  542                 }
  543 
  544                 /*
  545                  * If the File Handle was stale, invalidate the lookup
  546                  * cache, just in case.
  547                  */
  548                 if (error == ESTALE)
  549                         nfs_purgecache(vp);
  550                 /*
  551                  * Skip wcc data on NFS errors for now.  NetApp filers
  552                  * return corrupt postop attrs in the wcc data for NFS
  553                  * err EROFS.  Not sure if they could return corrupt
  554                  * postop attrs for others errors.
  555                  */
  556                 if ((nmp->nm_flag & NFSMNT_NFSV3) &&
  557                     !nfs_skip_wcc_data_onerr) {
  558                         *mrp = mrep;
  559                         *mdp = md;
  560                         *dposp = dpos;
  561                         error |= NFSERR_RETERR;
  562                 } else
  563                         m_freem(mrep);
  564                 goto nfsmout;
  565         }
  566 
  567 #ifdef KDTRACE_HOOKS
  568         if (dtrace_nfsclient_nfs23_done_probe != NULL) {
  569                 uint32_t probe_id;
  570                 int probe_procnum;
  571 
  572                 if (nmp->nm_flag & NFSMNT_NFSV3) {
  573                         probe_id = nfsclient_nfs3_done_probes[procnum];
  574                         probe_procnum = procnum;
  575                 } else {
  576                         probe_id = nfsclient_nfs2_done_probes[procnum];
  577                         probe_procnum = (nmp->nm_flag & NFSMNT_NFSV3) ?
  578                             procnum : nfsv2_procid[procnum];
  579                 }
  580                 if (probe_id != 0)
  581                         (dtrace_nfsclient_nfs23_done_probe)(probe_id, vp,
  582                             mreq, cred, probe_procnum, 0);
  583         }
  584 #endif
  585         m_freem(mreq);
  586         *mrp = mrep;
  587         *mdp = md;
  588         *dposp = dpos;
  589         AUTH_DESTROY(auth);
  590         return (0);
  591 
  592 nfsmout:
  593 #ifdef KDTRACE_HOOKS
  594         if (dtrace_nfsclient_nfs23_done_probe != NULL) {
  595                 uint32_t probe_id;
  596                 int probe_procnum;
  597 
  598                 if (nmp->nm_flag & NFSMNT_NFSV3) {
  599                         probe_id = nfsclient_nfs3_done_probes[procnum];
  600                         probe_procnum = procnum;
  601                 } else {
  602                         probe_id = nfsclient_nfs2_done_probes[procnum];
  603                         probe_procnum = (nmp->nm_flag & NFSMNT_NFSV3) ?
  604                             procnum : nfsv2_procid[procnum];
  605                 }
  606                 if (probe_id != 0)
  607                         (dtrace_nfsclient_nfs23_done_probe)(probe_id, vp,
  608                             mreq, cred, probe_procnum, error);
  609         }
  610 #endif
  611         m_freem(mreq);
  612         if (auth)
  613                 AUTH_DESTROY(auth);
  614         return (error);
  615 }
  616 
  617 /*
  618  * Mark all of an nfs mount's outstanding requests with R_SOFTTERM and
  619  * wait for all requests to complete.  This is used by forced unmounts
  620  * to terminate any outstanding RPCs.
  621  */
  622 int
  623 nfs_nmcancelreqs(struct nfsmount *nmp)
  624 {
  625 
  626         if (nmp->nm_client)
  627                 CLNT_CLOSE(nmp->nm_client);
  628         return (0);
  629 }
  630 
  631 /*
  632  * Any signal that can interrupt an NFS operation in an intr mount
  633  * should be added to this set.  SIGSTOP and SIGKILL cannot be masked.
  634  */
  635 int nfs_sig_set[] = {
  636         SIGINT,
  637         SIGTERM,
  638         SIGHUP,
  639         SIGKILL,
  640         SIGSTOP,
  641         SIGQUIT
  642 };
  643 
  644 /*
  645  * Check to see if one of the signals in our subset is pending on
  646  * the process (in an intr mount).
  647  */
  648 static int
  649 nfs_sig_pending(sigset_t set)
  650 {
  651         int i;
  652 
  653         for (i = 0 ; i < sizeof(nfs_sig_set)/sizeof(int) ; i++)
  654                 if (SIGISMEMBER(set, nfs_sig_set[i]))
  655                         return (1);
  656         return (0);
  657 }
  658 
  659 /*
  660  * The set/restore sigmask functions are used to (temporarily) overwrite
  661  * the process p_sigmask during an RPC call (for example).  These are also
  662  * used in other places in the NFS client that might tsleep().
  663  */
  664 void
  665 nfs_set_sigmask(struct thread *td, sigset_t *oldset)
  666 {
  667         sigset_t newset;
  668         int i;
  669         struct proc *p;
  670 
  671         SIGFILLSET(newset);
  672         if (td == NULL)
  673                 td = curthread; /* XXX */
  674         p = td->td_proc;
  675         /* Remove the NFS set of signals from newset. */
  676         PROC_LOCK(p);
  677         mtx_lock(&p->p_sigacts->ps_mtx);
  678         for (i = 0 ; i < sizeof(nfs_sig_set)/sizeof(int) ; i++) {
  679                 /*
  680                  * But make sure we leave the ones already masked
  681                  * by the process, i.e. remove the signal from the
  682                  * temporary signalmask only if it wasn't already
  683                  * in p_sigmask.
  684                  */
  685                 if (!SIGISMEMBER(td->td_sigmask, nfs_sig_set[i]) &&
  686                     !SIGISMEMBER(p->p_sigacts->ps_sigignore, nfs_sig_set[i]))
  687                         SIGDELSET(newset, nfs_sig_set[i]);
  688         }
  689         mtx_unlock(&p->p_sigacts->ps_mtx);
  690         PROC_UNLOCK(p);
  691         kern_sigprocmask(td, SIG_SETMASK, &newset, oldset, 0);
  692 }
  693 
  694 void
  695 nfs_restore_sigmask(struct thread *td, sigset_t *set)
  696 {
  697         if (td == NULL)
  698                 td = curthread; /* XXX */
  699         kern_sigprocmask(td, SIG_SETMASK, set, NULL, 0);
  700 }
  701 
  702 /*
  703  * NFS wrapper to msleep(), that shoves a new p_sigmask and restores the
  704  * old one after msleep() returns.
  705  */
  706 int
  707 nfs_msleep(struct thread *td, void *ident, struct mtx *mtx, int priority,
  708     char *wmesg, int timo)
  709 {
  710         sigset_t oldset;
  711         int error;
  712         struct proc *p;
  713 
  714         if ((priority & PCATCH) == 0)
  715                 return msleep(ident, mtx, priority, wmesg, timo);
  716         if (td == NULL)
  717                 td = curthread; /* XXX */
  718         nfs_set_sigmask(td, &oldset);
  719         error = msleep(ident, mtx, priority, wmesg, timo);
  720         nfs_restore_sigmask(td, &oldset);
  721         p = td->td_proc;
  722         return (error);
  723 }
  724 
  725 /*
  726  * Test for a termination condition pending on the process.
  727  * This is used for NFSMNT_INT mounts.
  728  */
  729 int
  730 nfs_sigintr(struct nfsmount *nmp, struct thread *td)
  731 {
  732         struct proc *p;
  733         sigset_t tmpset;
  734 
  735         /* Terminate all requests while attempting a forced unmount. */
  736         if (nmp->nm_mountp->mnt_kern_flag & MNTK_UNMOUNTF)
  737                 return (EIO);
  738         if (!(nmp->nm_flag & NFSMNT_INT))
  739                 return (0);
  740         if (td == NULL)
  741                 return (0);
  742         p = td->td_proc;
  743         PROC_LOCK(p);
  744         tmpset = p->p_siglist;
  745         SIGSETOR(tmpset, td->td_siglist);
  746         SIGSETNAND(tmpset, td->td_sigmask);
  747         mtx_lock(&p->p_sigacts->ps_mtx);
  748         SIGSETNAND(tmpset, p->p_sigacts->ps_sigignore);
  749         mtx_unlock(&p->p_sigacts->ps_mtx);
  750         if ((SIGNOTEMPTY(p->p_siglist) || SIGNOTEMPTY(td->td_siglist))
  751             && nfs_sig_pending(tmpset)) {
  752                 PROC_UNLOCK(p);
  753                 return (EINTR);
  754         }
  755         PROC_UNLOCK(p);
  756         return (0);
  757 }
  758 
  759 static int
  760 nfs_msg(struct thread *td, const char *server, const char *msg, int error)
  761 {
  762         struct proc *p;
  763 
  764         p = td ? td->td_proc : NULL;
  765         if (error)
  766                 tprintf(p, LOG_INFO, "nfs server %s: %s, error %d\n", server,
  767                     msg, error);
  768         else
  769                 tprintf(p, LOG_INFO, "nfs server %s: %s\n", server, msg);
  770         return (0);
  771 }
  772 
  773 static void
  774 nfs_down(struct nfsmount *nmp, struct thread *td, const char *msg,
  775     int error, int flags)
  776 {
  777         if (nmp == NULL)
  778                 return;
  779         mtx_lock(&nmp->nm_mtx);
  780         if ((flags & NFSSTA_TIMEO) && !(nmp->nm_state & NFSSTA_TIMEO)) {
  781                 nmp->nm_state |= NFSSTA_TIMEO;
  782                 mtx_unlock(&nmp->nm_mtx);
  783                 vfs_event_signal(&nmp->nm_mountp->mnt_stat.f_fsid,
  784                     VQ_NOTRESP, 0);
  785         } else
  786                 mtx_unlock(&nmp->nm_mtx);
  787         mtx_lock(&nmp->nm_mtx);
  788         if ((flags & NFSSTA_LOCKTIMEO) &&
  789             !(nmp->nm_state & NFSSTA_LOCKTIMEO)) {
  790                 nmp->nm_state |= NFSSTA_LOCKTIMEO;
  791                 mtx_unlock(&nmp->nm_mtx);
  792                 vfs_event_signal(&nmp->nm_mountp->mnt_stat.f_fsid,
  793                     VQ_NOTRESPLOCK, 0);
  794         } else
  795                 mtx_unlock(&nmp->nm_mtx);
  796         nfs_msg(td, nmp->nm_mountp->mnt_stat.f_mntfromname, msg, error);
  797 }
  798 
  799 static void
  800 nfs_up(struct nfsmount *nmp, struct thread *td, const char *msg,
  801     int flags, int tprintfmsg)
  802 {
  803         if (nmp == NULL)
  804                 return;
  805         if (tprintfmsg)
  806                 nfs_msg(td, nmp->nm_mountp->mnt_stat.f_mntfromname, msg, 0);
  807 
  808         mtx_lock(&nmp->nm_mtx);
  809         if ((flags & NFSSTA_TIMEO) && (nmp->nm_state & NFSSTA_TIMEO)) {
  810                 nmp->nm_state &= ~NFSSTA_TIMEO;
  811                 mtx_unlock(&nmp->nm_mtx);
  812                 vfs_event_signal(&nmp->nm_mountp->mnt_stat.f_fsid,
  813                     VQ_NOTRESP, 1);
  814         } else
  815                 mtx_unlock(&nmp->nm_mtx);
  816 
  817         mtx_lock(&nmp->nm_mtx);
  818         if ((flags & NFSSTA_LOCKTIMEO) &&
  819             (nmp->nm_state & NFSSTA_LOCKTIMEO)) {
  820                 nmp->nm_state &= ~NFSSTA_LOCKTIMEO;
  821                 mtx_unlock(&nmp->nm_mtx);
  822                 vfs_event_signal(&nmp->nm_mountp->mnt_stat.f_fsid,
  823                     VQ_NOTRESPLOCK, 1);
  824         } else
  825                 mtx_unlock(&nmp->nm_mtx);
  826 }

Cache object: bb6c7b327ae3a079b302154aee64da3b


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.