Print this page
    
Fix NFS design problems re. multiple zone keys
Make NFS server zone-specific data all have the same lifetime
Fix rfs4_clean_state_exi
Fix exi_cache_reclaim
Fix mistakes in zone keys work
More fixes re. exi_zoneid and exi_tree
(danmcd -> Keep some ASSERT()s around for readability.)
    
      
        | Split | 
	Close | 
      
      | Expand all | 
      | Collapse all | 
    
    
          --- old/usr/src/uts/common/fs/nfs/nfs_server.c
          +++ new/usr/src/uts/common/fs/nfs/nfs_server.c
   1    1  /*
   2    2   * CDDL HEADER START
   3    3   *
   4    4   * The contents of this file are subject to the terms of the
   5    5   * Common Development and Distribution License (the "License").
   6    6   * You may not use this file except in compliance with the License.
   7    7   *
   8    8   * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9    9   * or http://www.opensolaris.org/os/licensing.
  10   10   * See the License for the specific language governing permissions
  11   11   * and limitations under the License.
  12   12   *
  13   13   * When distributing Covered Code, include this CDDL HEADER in each
  14   14   * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15   15   * If applicable, add the following below this CDDL HEADER, with the
  16   16   * fields enclosed by brackets "[]" replaced with your own identifying
  17   17   * information: Portions Copyright [yyyy] [name of copyright owner]
  18   18   *
  19   19   * CDDL HEADER END
  20   20   */
  21   21  /*
  22   22   * Copyright (c) 1990, 2010, Oracle and/or its affiliates. All rights reserved.
  23   23   * Copyright (c) 2011 Bayard G. Bell. All rights reserved.
  24   24   * Copyright (c) 2013 by Delphix. All rights reserved.
  25   25   * Copyright (c) 2017 Joyent Inc
  26   26   * Copyright 2019 Nexenta by DDN, Inc.
  27   27   */
  28   28  
  29   29  /*
  30   30   *      Copyright (c) 1983,1984,1985,1986,1987,1988,1989  AT&T.
  31   31   *      All rights reserved.
  32   32   *      Use is subject to license terms.
  33   33   */
  34   34  
  35   35  /*
  36   36   * Copyright (c) 2011 Bayard G. Bell. All rights reserved.
  37   37   * Copyright (c) 2013 by Delphix. All rights reserved.
  38   38   * Copyright 2018 Nexenta Systems, Inc.
  39   39   * Copyright (c) 2017 Joyent Inc
  40   40   */
  41   41  
  42   42  #include <sys/param.h>
  43   43  #include <sys/types.h>
  44   44  #include <sys/systm.h>
  45   45  #include <sys/cred.h>
  46   46  #include <sys/proc.h>
  47   47  #include <sys/user.h>
  48   48  #include <sys/buf.h>
  49   49  #include <sys/vfs.h>
  50   50  #include <sys/vnode.h>
  51   51  #include <sys/pathname.h>
  52   52  #include <sys/uio.h>
  53   53  #include <sys/file.h>
  54   54  #include <sys/stat.h>
  55   55  #include <sys/errno.h>
  56   56  #include <sys/socket.h>
  57   57  #include <sys/sysmacros.h>
  58   58  #include <sys/siginfo.h>
  59   59  #include <sys/tiuser.h>
  60   60  #include <sys/statvfs.h>
  61   61  #include <sys/stream.h>
  62   62  #include <sys/strsun.h>
  63   63  #include <sys/strsubr.h>
  64   64  #include <sys/stropts.h>
  65   65  #include <sys/timod.h>
  66   66  #include <sys/t_kuser.h>
  67   67  #include <sys/kmem.h>
  68   68  #include <sys/kstat.h>
  69   69  #include <sys/dirent.h>
  70   70  #include <sys/cmn_err.h>
  71   71  #include <sys/debug.h>
  72   72  #include <sys/unistd.h>
  73   73  #include <sys/vtrace.h>
  74   74  #include <sys/mode.h>
  75   75  #include <sys/acl.h>
  76   76  #include <sys/sdt.h>
  77   77  #include <sys/debug.h>
  78   78  
  79   79  #include <rpc/types.h>
  80   80  #include <rpc/auth.h>
  81   81  #include <rpc/auth_unix.h>
  82   82  #include <rpc/auth_des.h>
  83   83  #include <rpc/svc.h>
  84   84  #include <rpc/xdr.h>
  85   85  #include <rpc/rpc_rdma.h>
  86   86  
  87   87  #include <nfs/nfs.h>
  88   88  #include <nfs/export.h>
  89   89  #include <nfs/nfssys.h>
  90   90  #include <nfs/nfs_clnt.h>
  91   91  #include <nfs/nfs_acl.h>
  92   92  #include <nfs/nfs_log.h>
  93   93  #include <nfs/lm.h>
  94   94  #include <nfs/nfs_dispatch.h>
  95   95  #include <nfs/nfs4_drc.h>
  96   96  
  97   97  #include <sys/modctl.h>
  98   98  #include <sys/cladm.h>
  99   99  #include <sys/clconf.h>
 100  100  
 101  101  #include <sys/tsol/label.h>
 102  102  
 103  103  #define MAXHOST 32
 104  104  const char *kinet_ntop6(uchar_t *, char *, size_t);
 105  105  
 106  106  /*
 107  107   * Module linkage information.
  
    | 
      ↓ open down ↓ | 
    107 lines elided | 
    
      ↑ open up ↑ | 
  
 108  108   */
 109  109  
 110  110  static struct modlmisc modlmisc = {
 111  111          &mod_miscops, "NFS server module"
 112  112  };
 113  113  
 114  114  static struct modlinkage modlinkage = {
 115  115          MODREV_1, (void *)&modlmisc, NULL
 116  116  };
 117  117  
 118      -zone_key_t nfssrv_zone_key;
      118 +zone_key_t      nfssrv_zone_key;
      119 +list_t          nfssrv_globals_list;
      120 +krwlock_t       nfssrv_globals_rwl;
      121 +
 119  122  kmem_cache_t *nfs_xuio_cache;
 120  123  int nfs_loaned_buffers = 0;
 121  124  
 122  125  int
 123  126  _init(void)
 124  127  {
 125  128          int status;
 126  129  
 127  130          nfs_srvinit();
 128  131  
 129  132          status = mod_install((struct modlinkage *)&modlinkage);
 130  133          if (status != 0) {
 131  134                  /*
 132  135                   * Could not load module, cleanup previous
 133  136                   * initialization work.
 134  137                   */
 135  138                  nfs_srvfini();
 136  139  
 137  140                  return (status);
 138  141          }
 139  142  
 140  143          /*
 141  144           * Initialise some placeholders for nfssys() calls. These have
 142  145           * to be declared by the nfs module, since that handles nfssys()
 143  146           * calls - also used by NFS clients - but are provided by this
 144  147           * nfssrv module. These also then serve as confirmation to the
 145  148           * relevant code in nfs that nfssrv has been loaded, as they're
 146  149           * initially NULL.
 147  150           */
 148  151          nfs_srv_quiesce_func = nfs_srv_quiesce_all;
 149  152          nfs_srv_dss_func = rfs4_dss_setpaths;
 150  153  
 151  154          /* setup DSS paths here; must be done before initial server startup */
 152  155          rfs4_dss_paths = rfs4_dss_oldpaths = NULL;
 153  156  
 154  157          /* initialize the copy reduction caches */
 155  158  
 156  159          nfs_xuio_cache = kmem_cache_create("nfs_xuio_cache",
 157  160              sizeof (nfs_xuio_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
 158  161  
 159  162          return (status);
 160  163  }
 161  164  
 162  165  int
 163  166  _fini()
 164  167  {
 165  168          return (EBUSY);
 166  169  }
 167  170  
 168  171  int
 169  172  _info(struct modinfo *modinfop)
 170  173  {
 171  174          return (mod_info(&modlinkage, modinfop));
 172  175  }
 173  176  
 174  177  /*
 175  178   * PUBLICFH_CHECK() checks if the dispatch routine supports
 176  179   * RPC_PUBLICFH_OK, if the filesystem is exported public, and if the
 177  180   * incoming request is using the public filehandle. The check duplicates
 178  181   * the exportmatch() call done in checkexport(), and we should consider
 179  182   * modifying those routines to avoid the duplication. For now, we optimize
 180  183   * by calling exportmatch() only after checking that the dispatch routine
 181  184   * supports RPC_PUBLICFH_OK, and if the filesystem is explicitly exported
 182  185   * public (i.e., not the placeholder).
 183  186   */
 184  187  #define PUBLICFH_CHECK(ne, disp, exi, fsid, xfid) \
 185  188                  ((disp->dis_flags & RPC_PUBLICFH_OK) && \
 186  189                  ((exi->exi_export.ex_flags & EX_PUBLIC) || \
 187  190                  (exi == ne->exi_public && exportmatch(ne->exi_root, \
 188  191                  fsid, xfid))))
 189  192  
 190  193  static void     nfs_srv_shutdown_all(int);
 191  194  static void     rfs4_server_start(nfs_globals_t *, int);
 192  195  static void     nullfree(void);
 193  196  static void     rfs_dispatch(struct svc_req *, SVCXPRT *);
  
    | 
      ↓ open down ↓ | 
    65 lines elided | 
    
      ↑ open up ↑ | 
  
 194  197  static void     acl_dispatch(struct svc_req *, SVCXPRT *);
 195  198  static void     common_dispatch(struct svc_req *, SVCXPRT *,
 196  199                  rpcvers_t, rpcvers_t, char *,
 197  200                  struct rpc_disptable *);
 198  201  static  int     checkauth(struct exportinfo *, struct svc_req *, cred_t *, int,
 199  202                  bool_t, bool_t *);
 200  203  static char     *client_name(struct svc_req *req);
 201  204  static char     *client_addr(struct svc_req *req, char *buf);
 202  205  extern  int     sec_svc_getcred(struct svc_req *, cred_t *cr, char **, int *);
 203  206  extern  bool_t  sec_svc_inrootlist(int, caddr_t, int, caddr_t *);
 204      -static void     *nfs_srv_zone_init(zoneid_t);
 205      -static void     nfs_srv_zone_fini(zoneid_t, void *);
      207 +static void     *nfs_server_zone_init(zoneid_t);
      208 +static void     nfs_server_zone_fini(zoneid_t, void *);
      209 +static void     nfs_server_zone_shutdown(zoneid_t, void *);
 206  210  
 207  211  #define NFSLOG_COPY_NETBUF(exi, xprt, nb)       {               \
 208  212          (nb)->maxlen = (xprt)->xp_rtaddr.maxlen;                \
 209  213          (nb)->len = (xprt)->xp_rtaddr.len;                      \
 210  214          (nb)->buf = kmem_alloc((nb)->len, KM_SLEEP);            \
 211  215          bcopy((xprt)->xp_rtaddr.buf, (nb)->buf, (nb)->len);     \
 212  216          }
 213  217  
 214  218  /*
 215  219   * Public Filehandle common nfs routines
 216  220   */
 217  221  static int      MCLpath(char **);
 218  222  static void     URLparse(char *);
 219  223  
 220  224  /*
 221  225   * NFS callout table.
 222  226   * This table is used by svc_getreq() to dispatch a request with
 223  227   * a given prog/vers pair to an appropriate service provider
 224  228   * dispatch routine.
 225  229   *
 226  230   * NOTE: ordering is relied upon below when resetting the version min/max
 227  231   * for NFS_PROGRAM.  Careful, if this is ever changed.
 228  232   */
 229  233  static SVC_CALLOUT __nfs_sc_clts[] = {
 230  234          { NFS_PROGRAM,     NFS_VERSMIN,     NFS_VERSMAX,        rfs_dispatch },
 231  235          { NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX,    acl_dispatch }
 232  236  };
 233  237  
 234  238  static SVC_CALLOUT_TABLE nfs_sct_clts = {
 235  239          sizeof (__nfs_sc_clts) / sizeof (__nfs_sc_clts[0]), FALSE,
 236  240          __nfs_sc_clts
 237  241  };
 238  242  
 239  243  static SVC_CALLOUT __nfs_sc_cots[] = {
 240  244          { NFS_PROGRAM,     NFS_VERSMIN,     NFS_VERSMAX,        rfs_dispatch },
 241  245          { NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX,    acl_dispatch }
 242  246  };
 243  247  
 244  248  static SVC_CALLOUT_TABLE nfs_sct_cots = {
 245  249          sizeof (__nfs_sc_cots) / sizeof (__nfs_sc_cots[0]), FALSE, __nfs_sc_cots
 246  250  };
 247  251  
 248  252  static SVC_CALLOUT __nfs_sc_rdma[] = {
 249  253          { NFS_PROGRAM,     NFS_VERSMIN,     NFS_VERSMAX,        rfs_dispatch },
 250  254          { NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX,    acl_dispatch }
 251  255  };
 252  256  
 253  257  static SVC_CALLOUT_TABLE nfs_sct_rdma = {
 254  258          sizeof (__nfs_sc_rdma) / sizeof (__nfs_sc_rdma[0]), FALSE, __nfs_sc_rdma
 255  259  };
  
    | 
      ↓ open down ↓ | 
    40 lines elided | 
    
      ↑ open up ↑ | 
  
 256  260  
 257  261  /*
 258  262   * DSS: distributed stable storage
 259  263   * lists of all DSS paths: current, and before last warmstart
 260  264   */
 261  265  nvlist_t *rfs4_dss_paths, *rfs4_dss_oldpaths;
 262  266  
 263  267  int rfs4_dispatch(struct rpcdisp *, struct svc_req *, SVCXPRT *, char *);
 264  268  bool_t rfs4_minorvers_mismatch(struct svc_req *, SVCXPRT *, void *);
 265  269  
      270 +nfs_globals_t *
      271 +nfs_srv_getzg(void)
      272 +{
      273 +        nfs_globals_t *ng;
      274 +
      275 +        ng = zone_getspecific(nfssrv_zone_key, curzone);
      276 +        return (ng);
      277 +}
      278 +
 266  279  /*
 267  280   * Will be called at the point the server pool is being unregistered
 268  281   * from the pool list. From that point onwards, the pool is waiting
 269  282   * to be drained and as such the server state is stale and pertains
 270  283   * to the old instantiation of the NFS server pool.
 271  284   */
 272  285  void
 273  286  nfs_srv_offline(void)
 274  287  {
 275  288          nfs_globals_t *ng;
 276  289  
 277      -        ng = zone_getspecific(nfssrv_zone_key, curzone);
      290 +        ng = nfs_srv_getzg();
 278  291  
 279  292          mutex_enter(&ng->nfs_server_upordown_lock);
 280  293          if (ng->nfs_server_upordown == NFS_SERVER_RUNNING) {
 281  294                  ng->nfs_server_upordown = NFS_SERVER_OFFLINE;
 282  295          }
 283  296          mutex_exit(&ng->nfs_server_upordown_lock);
 284  297  }
 285  298  
 286  299  /*
 287  300   * Will be called at the point the server pool is being destroyed so
 288  301   * all transports have been closed and no service threads are in
 289  302   * existence.
 290  303   *
 291  304   * If we quiesce the server, we're shutting it down without destroying the
 292  305   * server state. This allows it to warm start subsequently.
 293  306   */
 294  307  void
 295  308  nfs_srv_stop_all(void)
 296  309  {
 297  310          int quiesce = 0;
 298  311          nfs_srv_shutdown_all(quiesce);
 299  312  }
 300  313  
 301  314  /*
 302  315   * This alternative shutdown routine can be requested via nfssys()
 303  316   */
  
    | 
      ↓ open down ↓ | 
    16 lines elided | 
    
      ↑ open up ↑ | 
  
 304  317  void
 305  318  nfs_srv_quiesce_all(void)
 306  319  {
 307  320          int quiesce = 1;
 308  321          nfs_srv_shutdown_all(quiesce);
 309  322  }
 310  323  
 311  324  static void
 312  325  nfs_srv_shutdown_all(int quiesce)
 313  326  {
 314      -        nfs_globals_t *ng = zone_getspecific(nfssrv_zone_key, curzone);
      327 +        nfs_globals_t *ng = nfs_srv_getzg();
 315  328  
 316  329          mutex_enter(&ng->nfs_server_upordown_lock);
 317  330          if (quiesce) {
 318  331                  if (ng->nfs_server_upordown == NFS_SERVER_RUNNING ||
 319  332                      ng->nfs_server_upordown == NFS_SERVER_OFFLINE) {
 320  333                          ng->nfs_server_upordown = NFS_SERVER_QUIESCED;
 321  334                          cv_signal(&ng->nfs_server_upordown_cv);
 322  335  
 323  336                          /* reset DSS state */
 324  337                          rfs4_dss_numnewpaths = 0;
 325  338                          rfs4_dss_newpaths = NULL;
 326  339  
 327  340                          cmn_err(CE_NOTE, "nfs_server: server is now quiesced; "
 328  341                              "NFSv4 state has been preserved");
 329  342                  }
 330  343          } else {
 331  344                  if (ng->nfs_server_upordown == NFS_SERVER_OFFLINE) {
 332  345                          ng->nfs_server_upordown = NFS_SERVER_STOPPING;
 333  346                          mutex_exit(&ng->nfs_server_upordown_lock);
 334  347                          rfs4_state_zone_fini();
 335  348                          rfs4_fini_drc();
 336  349                          mutex_enter(&ng->nfs_server_upordown_lock);
 337  350                          ng->nfs_server_upordown = NFS_SERVER_STOPPED;
 338  351  
 339  352                          /* reset DSS state */
 340  353                          rfs4_dss_numnewpaths = 0;
 341  354                          rfs4_dss_newpaths = NULL;
 342  355  
 343  356                          cv_signal(&ng->nfs_server_upordown_cv);
 344  357                  }
 345  358          }
 346  359          mutex_exit(&ng->nfs_server_upordown_lock);
 347  360  }
 348  361  
 349  362  static int
 350  363  nfs_srv_set_sc_versions(struct file *fp, SVC_CALLOUT_TABLE **sctpp,
 351  364      rpcvers_t versmin, rpcvers_t versmax)
 352  365  {
 353  366          struct strioctl strioc;
 354  367          struct T_info_ack tinfo;
 355  368          int             error, retval;
 356  369  
 357  370          /*
 358  371           * Find out what type of transport this is.
 359  372           */
 360  373          strioc.ic_cmd = TI_GETINFO;
 361  374          strioc.ic_timout = -1;
 362  375          strioc.ic_len = sizeof (tinfo);
 363  376          strioc.ic_dp = (char *)&tinfo;
 364  377          tinfo.PRIM_type = T_INFO_REQ;
 365  378  
 366  379          error = strioctl(fp->f_vnode, I_STR, (intptr_t)&strioc, 0, K_TO_K,
 367  380              CRED(), &retval);
 368  381          if (error || retval)
 369  382                  return (error);
 370  383  
 371  384          /*
 372  385           * Based on our query of the transport type...
 373  386           *
 374  387           * Reset the min/max versions based on the caller's request
 375  388           * NOTE: This assumes that NFS_PROGRAM is first in the array!!
 376  389           * And the second entry is the NFS_ACL_PROGRAM.
 377  390           */
 378  391          switch (tinfo.SERV_type) {
 379  392          case T_CLTS:
 380  393                  if (versmax == NFS_V4)
 381  394                          return (EINVAL);
 382  395                  __nfs_sc_clts[0].sc_versmin = versmin;
 383  396                  __nfs_sc_clts[0].sc_versmax = versmax;
 384  397                  __nfs_sc_clts[1].sc_versmin = versmin;
 385  398                  __nfs_sc_clts[1].sc_versmax = versmax;
 386  399                  *sctpp = &nfs_sct_clts;
 387  400                  break;
 388  401          case T_COTS:
 389  402          case T_COTS_ORD:
 390  403                  __nfs_sc_cots[0].sc_versmin = versmin;
 391  404                  __nfs_sc_cots[0].sc_versmax = versmax;
 392  405                  /* For the NFS_ACL program, check the max version */
 393  406                  if (versmax > NFS_ACL_VERSMAX)
 394  407                          versmax = NFS_ACL_VERSMAX;
 395  408                  __nfs_sc_cots[1].sc_versmin = versmin;
 396  409                  __nfs_sc_cots[1].sc_versmax = versmax;
 397  410                  *sctpp = &nfs_sct_cots;
 398  411                  break;
 399  412          default:
 400  413                  error = EINVAL;
 401  414          }
 402  415  
 403  416          return (error);
 404  417  }
 405  418  
 406  419  /*
 407  420   * NFS Server system call.
 408  421   * Does all of the work of running a NFS server.
 409  422   * uap->fd is the fd of an open transport provider
 410  423   */
 411  424  int
 412  425  nfs_svc(struct nfs_svc_args *arg, model_t model)
 413  426  {
 414  427          nfs_globals_t *ng;
 415  428          file_t *fp;
 416  429          SVCMASTERXPRT *xprt;
 417  430          int error;
 418  431          int readsize;
  
    | 
      ↓ open down ↓ | 
    94 lines elided | 
    
      ↑ open up ↑ | 
  
 419  432          char buf[KNC_STRSIZE];
 420  433          size_t len;
 421  434          STRUCT_HANDLE(nfs_svc_args, uap);
 422  435          struct netbuf addrmask;
 423  436          SVC_CALLOUT_TABLE *sctp = NULL;
 424  437  
 425  438  #ifdef lint
 426  439          model = model;          /* STRUCT macros don't always refer to it */
 427  440  #endif
 428  441  
 429      -        ng = zone_getspecific(nfssrv_zone_key, curzone);
      442 +        ng = nfs_srv_getzg();
 430  443          STRUCT_SET_HANDLE(uap, model, arg);
 431  444  
 432  445          /* Check privileges in nfssys() */
 433  446  
 434  447          if ((fp = getf(STRUCT_FGET(uap, fd))) == NULL)
 435  448                  return (EBADF);
 436  449  
      450 +        /* Setup global file handle in nfs_export */
      451 +        if ((error = nfs_export_get_rootfh(ng)) != 0)
      452 +                return (error);
      453 +
 437  454          /*
 438  455           * Set read buffer size to rsize
 439  456           * and add room for RPC headers.
 440  457           */
 441  458          readsize = nfs3tsize() + (RPC_MAXDATASIZE - NFS_MAXDATA);
 442  459          if (readsize < RPC_MAXDATASIZE)
 443  460                  readsize = RPC_MAXDATASIZE;
 444  461  
 445  462          error = copyinstr((const char *)STRUCT_FGETP(uap, netid), buf,
 446  463              KNC_STRSIZE, &len);
 447  464          if (error) {
 448  465                  releasef(STRUCT_FGET(uap, fd));
 449  466                  return (error);
 450  467          }
 451  468  
 452  469          addrmask.len = STRUCT_FGET(uap, addrmask.len);
 453  470          addrmask.maxlen = STRUCT_FGET(uap, addrmask.maxlen);
 454  471          addrmask.buf = kmem_alloc(addrmask.maxlen, KM_SLEEP);
 455  472          error = copyin(STRUCT_FGETP(uap, addrmask.buf), addrmask.buf,
 456  473              addrmask.len);
 457  474          if (error) {
 458  475                  releasef(STRUCT_FGET(uap, fd));
 459  476                  kmem_free(addrmask.buf, addrmask.maxlen);
 460  477                  return (error);
 461  478          }
 462  479  
 463  480          ng->nfs_versmin = STRUCT_FGET(uap, versmin);
 464  481          ng->nfs_versmax = STRUCT_FGET(uap, versmax);
 465  482  
 466  483          /* Double check the vers min/max ranges */
 467  484          if ((ng->nfs_versmin > ng->nfs_versmax) ||
 468  485              (ng->nfs_versmin < NFS_VERSMIN) ||
 469  486              (ng->nfs_versmax > NFS_VERSMAX)) {
 470  487                  ng->nfs_versmin = NFS_VERSMIN_DEFAULT;
 471  488                  ng->nfs_versmax = NFS_VERSMAX_DEFAULT;
 472  489          }
 473  490  
 474  491          if (error = nfs_srv_set_sc_versions(fp, &sctp, ng->nfs_versmin,
 475  492              ng->nfs_versmax)) {
 476  493                  releasef(STRUCT_FGET(uap, fd));
 477  494                  kmem_free(addrmask.buf, addrmask.maxlen);
 478  495                  return (error);
 479  496          }
 480  497  
 481  498          /* Initialize nfsv4 server */
 482  499          if (ng->nfs_versmax == (rpcvers_t)NFS_V4)
 483  500                  rfs4_server_start(ng, STRUCT_FGET(uap, delegation));
 484  501  
 485  502          /* Create a transport handle. */
 486  503          error = svc_tli_kcreate(fp, readsize, buf, &addrmask, &xprt,
 487  504              sctp, NULL, NFS_SVCPOOL_ID, TRUE);
 488  505  
 489  506          if (error)
 490  507                  kmem_free(addrmask.buf, addrmask.maxlen);
 491  508  
 492  509          releasef(STRUCT_FGET(uap, fd));
 493  510  
 494  511          /* HA-NFSv4: save the cluster nodeid */
 495  512          if (cluster_bootflags & CLUSTER_BOOTED)
 496  513                  lm_global_nlmid = clconf_get_nodeid();
 497  514  
 498  515          return (error);
 499  516  }
 500  517  
 501  518  static void
 502  519  rfs4_server_start(nfs_globals_t *ng, int nfs4_srv_delegation)
 503  520  {
 504  521          /*
 505  522           * Determine if the server has previously been "started" and
 506  523           * if not, do the per instance initialization
 507  524           */
 508  525          mutex_enter(&ng->nfs_server_upordown_lock);
 509  526  
 510  527          if (ng->nfs_server_upordown != NFS_SERVER_RUNNING) {
 511  528                  /* Do we need to stop and wait on the previous server? */
 512  529                  while (ng->nfs_server_upordown == NFS_SERVER_STOPPING ||
 513  530                      ng->nfs_server_upordown == NFS_SERVER_OFFLINE)
 514  531                          cv_wait(&ng->nfs_server_upordown_cv,
 515  532                              &ng->nfs_server_upordown_lock);
 516  533  
 517  534                  if (ng->nfs_server_upordown != NFS_SERVER_RUNNING) {
 518  535                          (void) svc_pool_control(NFS_SVCPOOL_ID,
 519  536                              SVCPSET_UNREGISTER_PROC, (void *)&nfs_srv_offline);
 520  537                          (void) svc_pool_control(NFS_SVCPOOL_ID,
 521  538                              SVCPSET_SHUTDOWN_PROC, (void *)&nfs_srv_stop_all);
 522  539  
 523  540                          rfs4_do_server_start(ng->nfs_server_upordown,
 524  541                              nfs4_srv_delegation,
 525  542                              cluster_bootflags & CLUSTER_BOOTED);
 526  543  
 527  544                          ng->nfs_server_upordown = NFS_SERVER_RUNNING;
 528  545                  }
 529  546                  cv_signal(&ng->nfs_server_upordown_cv);
 530  547          }
 531  548          mutex_exit(&ng->nfs_server_upordown_lock);
 532  549  }
 533  550  
 534  551  /*
 535  552   * If RDMA device available,
 536  553   * start RDMA listener.
 537  554   */
 538  555  int
 539  556  rdma_start(struct rdma_svc_args *rsa)
 540  557  {
 541  558          nfs_globals_t *ng;
 542  559          int error;
 543  560          rdma_xprt_group_t started_rdma_xprts;
 544  561          rdma_stat stat;
  
    | 
      ↓ open down ↓ | 
    98 lines elided | 
    
      ↑ open up ↑ | 
  
 545  562          int svc_state = 0;
 546  563  
 547  564          /* Double check the vers min/max ranges */
 548  565          if ((rsa->nfs_versmin > rsa->nfs_versmax) ||
 549  566              (rsa->nfs_versmin < NFS_VERSMIN) ||
 550  567              (rsa->nfs_versmax > NFS_VERSMAX)) {
 551  568                  rsa->nfs_versmin = NFS_VERSMIN_DEFAULT;
 552  569                  rsa->nfs_versmax = NFS_VERSMAX_DEFAULT;
 553  570          }
 554  571  
 555      -        ng = zone_getspecific(nfssrv_zone_key, curzone);
      572 +        ng = nfs_srv_getzg();
 556  573          ng->nfs_versmin = rsa->nfs_versmin;
 557  574          ng->nfs_versmax = rsa->nfs_versmax;
 558  575  
 559  576          /* Set the versions in the callout table */
 560  577          __nfs_sc_rdma[0].sc_versmin = rsa->nfs_versmin;
 561  578          __nfs_sc_rdma[0].sc_versmax = rsa->nfs_versmax;
 562  579          /* For the NFS_ACL program, check the max version */
 563  580          __nfs_sc_rdma[1].sc_versmin = rsa->nfs_versmin;
 564  581          if (rsa->nfs_versmax > NFS_ACL_VERSMAX)
 565  582                  __nfs_sc_rdma[1].sc_versmax = NFS_ACL_VERSMAX;
 566  583          else
 567  584                  __nfs_sc_rdma[1].sc_versmax = rsa->nfs_versmax;
 568  585  
 569  586          /* Initialize nfsv4 server */
 570  587          if (rsa->nfs_versmax == (rpcvers_t)NFS_V4)
 571  588                  rfs4_server_start(ng, rsa->delegation);
 572  589  
 573  590          started_rdma_xprts.rtg_count = 0;
 574  591          started_rdma_xprts.rtg_listhead = NULL;
 575  592          started_rdma_xprts.rtg_poolid = rsa->poolid;
 576  593  
 577  594  restart:
 578  595          error = svc_rdma_kcreate(rsa->netid, &nfs_sct_rdma, rsa->poolid,
 579  596              &started_rdma_xprts);
 580  597  
 581  598          svc_state = !error;
 582  599  
 583  600          while (!error) {
 584  601  
 585  602                  /*
 586  603                   * wait till either interrupted by a signal on
 587  604                   * nfs service stop/restart or signalled by a
 588  605                   * rdma attach/detatch.
 589  606                   */
 590  607  
 591  608                  stat = rdma_kwait();
 592  609  
 593  610                  /*
 594  611                   * stop services if running -- either on a HCA detach event
 595  612                   * or if the nfs service is stopped/restarted.
 596  613                   */
 597  614  
 598  615                  if ((stat == RDMA_HCA_DETACH || stat == RDMA_INTR) &&
 599  616                      svc_state) {
 600  617                          rdma_stop(&started_rdma_xprts);
 601  618                          svc_state = 0;
 602  619                  }
 603  620  
 604  621                  /*
 605  622                   * nfs service stop/restart, break out of the
 606  623                   * wait loop and return;
 607  624                   */
 608  625                  if (stat == RDMA_INTR)
 609  626                          return (0);
 610  627  
 611  628                  /*
 612  629                   * restart stopped services on a HCA attach event
 613  630                   * (if not already running)
 614  631                   */
 615  632  
 616  633                  if ((stat == RDMA_HCA_ATTACH) && (svc_state == 0))
 617  634                          goto restart;
 618  635  
 619  636                  /*
 620  637                   * loop until a nfs service stop/restart
 621  638                   */
 622  639          }
 623  640  
 624  641          return (error);
 625  642  }
 626  643  
 627  644  /* ARGSUSED */
 628  645  void
 629  646  rpc_null(caddr_t *argp, caddr_t *resp, struct exportinfo *exi,
 630  647      struct svc_req *req, cred_t *cr, bool_t ro)
 631  648  {
 632  649  }
 633  650  
 634  651  /* ARGSUSED */
 635  652  void
 636  653  rpc_null_v3(caddr_t *argp, caddr_t *resp, struct exportinfo *exi,
 637  654      struct svc_req *req, cred_t *cr, bool_t ro)
 638  655  {
 639  656          DTRACE_NFSV3_4(op__null__start, struct svc_req *, req,
 640  657              cred_t *, cr, vnode_t *, NULL, struct exportinfo *, exi);
 641  658          DTRACE_NFSV3_4(op__null__done, struct svc_req *, req,
 642  659              cred_t *, cr, vnode_t *, NULL, struct exportinfo *, exi);
 643  660  }
 644  661  
 645  662  /* ARGSUSED */
 646  663  static void
 647  664  rfs_error(caddr_t *argp, caddr_t *resp, struct exportinfo *exi,
 648  665      struct svc_req *req, cred_t *cr, bool_t ro)
 649  666  {
 650  667          /* return (EOPNOTSUPP); */
 651  668  }
 652  669  
 653  670  static void
 654  671  nullfree(void)
 655  672  {
 656  673  }
 657  674  
 658  675  static char *rfscallnames_v2[] = {
 659  676          "RFS2_NULL",
 660  677          "RFS2_GETATTR",
 661  678          "RFS2_SETATTR",
 662  679          "RFS2_ROOT",
 663  680          "RFS2_LOOKUP",
 664  681          "RFS2_READLINK",
 665  682          "RFS2_READ",
 666  683          "RFS2_WRITECACHE",
 667  684          "RFS2_WRITE",
 668  685          "RFS2_CREATE",
 669  686          "RFS2_REMOVE",
 670  687          "RFS2_RENAME",
 671  688          "RFS2_LINK",
 672  689          "RFS2_SYMLINK",
 673  690          "RFS2_MKDIR",
 674  691          "RFS2_RMDIR",
 675  692          "RFS2_READDIR",
 676  693          "RFS2_STATFS"
 677  694  };
 678  695  
 679  696  static struct rpcdisp rfsdisptab_v2[] = {
 680  697          /*
 681  698           * NFS VERSION 2
 682  699           */
 683  700  
 684  701          /* RFS_NULL = 0 */
 685  702          {rpc_null,
 686  703              xdr_void, NULL_xdrproc_t, 0,
 687  704              xdr_void, NULL_xdrproc_t, 0,
 688  705              nullfree, RPC_IDEMPOTENT,
 689  706              0},
 690  707  
 691  708          /* RFS_GETATTR = 1 */
 692  709          {rfs_getattr,
 693  710              xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t),
 694  711              xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat),
 695  712              nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP,
 696  713              rfs_getattr_getfh},
 697  714  
 698  715          /* RFS_SETATTR = 2 */
 699  716          {rfs_setattr,
 700  717              xdr_saargs, NULL_xdrproc_t, sizeof (struct nfssaargs),
 701  718              xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat),
 702  719              nullfree, RPC_MAPRESP,
 703  720              rfs_setattr_getfh},
 704  721  
 705  722          /* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */
 706  723          {rfs_error,
 707  724              xdr_void, NULL_xdrproc_t, 0,
 708  725              xdr_void, NULL_xdrproc_t, 0,
 709  726              nullfree, RPC_IDEMPOTENT,
 710  727              0},
 711  728  
 712  729          /* RFS_LOOKUP = 4 */
 713  730          {rfs_lookup,
 714  731              xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs),
 715  732              xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres),
 716  733              nullfree, RPC_IDEMPOTENT|RPC_MAPRESP|RPC_PUBLICFH_OK,
 717  734              rfs_lookup_getfh},
 718  735  
 719  736          /* RFS_READLINK = 5 */
 720  737          {rfs_readlink,
 721  738              xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t),
 722  739              xdr_rdlnres, NULL_xdrproc_t, sizeof (struct nfsrdlnres),
 723  740              rfs_rlfree, RPC_IDEMPOTENT,
 724  741              rfs_readlink_getfh},
 725  742  
 726  743          /* RFS_READ = 6 */
 727  744          {rfs_read,
 728  745              xdr_readargs, NULL_xdrproc_t, sizeof (struct nfsreadargs),
 729  746              xdr_rdresult, NULL_xdrproc_t, sizeof (struct nfsrdresult),
 730  747              rfs_rdfree, RPC_IDEMPOTENT,
 731  748              rfs_read_getfh},
 732  749  
 733  750          /* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */
 734  751          {rfs_error,
 735  752              xdr_void, NULL_xdrproc_t, 0,
 736  753              xdr_void, NULL_xdrproc_t, 0,
 737  754              nullfree, RPC_IDEMPOTENT,
 738  755              0},
 739  756  
 740  757          /* RFS_WRITE = 8 */
 741  758          {rfs_write,
 742  759              xdr_writeargs, NULL_xdrproc_t, sizeof (struct nfswriteargs),
 743  760              xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat),
 744  761              nullfree, RPC_MAPRESP,
 745  762              rfs_write_getfh},
 746  763  
 747  764          /* RFS_CREATE = 9 */
 748  765          {rfs_create,
 749  766              xdr_creatargs, NULL_xdrproc_t, sizeof (struct nfscreatargs),
 750  767              xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres),
 751  768              nullfree, RPC_MAPRESP,
 752  769              rfs_create_getfh},
 753  770  
 754  771          /* RFS_REMOVE = 10 */
 755  772          {rfs_remove,
 756  773              xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs),
 757  774  #ifdef _LITTLE_ENDIAN
 758  775              xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
 759  776  #else
 760  777              xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
 761  778  #endif
 762  779              nullfree, RPC_MAPRESP,
 763  780              rfs_remove_getfh},
 764  781  
 765  782          /* RFS_RENAME = 11 */
 766  783          {rfs_rename,
 767  784              xdr_rnmargs, NULL_xdrproc_t, sizeof (struct nfsrnmargs),
 768  785  #ifdef _LITTLE_ENDIAN
 769  786              xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
 770  787  #else
 771  788              xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
 772  789  #endif
 773  790              nullfree, RPC_MAPRESP,
 774  791              rfs_rename_getfh},
 775  792  
 776  793          /* RFS_LINK = 12 */
 777  794          {rfs_link,
 778  795              xdr_linkargs, NULL_xdrproc_t, sizeof (struct nfslinkargs),
 779  796  #ifdef _LITTLE_ENDIAN
 780  797              xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
 781  798  #else
 782  799              xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
 783  800  #endif
 784  801              nullfree, RPC_MAPRESP,
 785  802              rfs_link_getfh},
 786  803  
 787  804          /* RFS_SYMLINK = 13 */
 788  805          {rfs_symlink,
 789  806              xdr_slargs, NULL_xdrproc_t, sizeof (struct nfsslargs),
 790  807  #ifdef _LITTLE_ENDIAN
 791  808              xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
 792  809  #else
 793  810              xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
 794  811  #endif
 795  812              nullfree, RPC_MAPRESP,
 796  813              rfs_symlink_getfh},
 797  814  
 798  815          /* RFS_MKDIR = 14 */
 799  816          {rfs_mkdir,
 800  817              xdr_creatargs, NULL_xdrproc_t, sizeof (struct nfscreatargs),
 801  818              xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres),
 802  819              nullfree, RPC_MAPRESP,
 803  820              rfs_mkdir_getfh},
 804  821  
 805  822          /* RFS_RMDIR = 15 */
 806  823          {rfs_rmdir,
 807  824              xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs),
 808  825  #ifdef _LITTLE_ENDIAN
 809  826              xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
 810  827  #else
 811  828              xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
 812  829  #endif
 813  830              nullfree, RPC_MAPRESP,
 814  831              rfs_rmdir_getfh},
 815  832  
 816  833          /* RFS_READDIR = 16 */
 817  834          {rfs_readdir,
 818  835              xdr_rddirargs, NULL_xdrproc_t, sizeof (struct nfsrddirargs),
 819  836              xdr_putrddirres, NULL_xdrproc_t, sizeof (struct nfsrddirres),
 820  837              rfs_rddirfree, RPC_IDEMPOTENT,
 821  838              rfs_readdir_getfh},
 822  839  
 823  840          /* RFS_STATFS = 17 */
 824  841          {rfs_statfs,
 825  842              xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t),
 826  843              xdr_statfs, xdr_faststatfs, sizeof (struct nfsstatfs),
 827  844              nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP,
 828  845              rfs_statfs_getfh},
 829  846  };
 830  847  
 831  848  static char *rfscallnames_v3[] = {
 832  849          "RFS3_NULL",
 833  850          "RFS3_GETATTR",
 834  851          "RFS3_SETATTR",
 835  852          "RFS3_LOOKUP",
 836  853          "RFS3_ACCESS",
 837  854          "RFS3_READLINK",
 838  855          "RFS3_READ",
 839  856          "RFS3_WRITE",
 840  857          "RFS3_CREATE",
 841  858          "RFS3_MKDIR",
 842  859          "RFS3_SYMLINK",
 843  860          "RFS3_MKNOD",
 844  861          "RFS3_REMOVE",
 845  862          "RFS3_RMDIR",
 846  863          "RFS3_RENAME",
 847  864          "RFS3_LINK",
 848  865          "RFS3_READDIR",
 849  866          "RFS3_READDIRPLUS",
 850  867          "RFS3_FSSTAT",
 851  868          "RFS3_FSINFO",
 852  869          "RFS3_PATHCONF",
 853  870          "RFS3_COMMIT"
 854  871  };
 855  872  
 856  873  static struct rpcdisp rfsdisptab_v3[] = {
 857  874          /*
 858  875           * NFS VERSION 3
 859  876           */
 860  877  
 861  878          /* RFS_NULL = 0 */
 862  879          {rpc_null_v3,
 863  880              xdr_void, NULL_xdrproc_t, 0,
 864  881              xdr_void, NULL_xdrproc_t, 0,
 865  882              nullfree, RPC_IDEMPOTENT,
 866  883              0},
 867  884  
 868  885          /* RFS3_GETATTR = 1 */
 869  886          {rfs3_getattr,
 870  887              xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (GETATTR3args),
 871  888              xdr_GETATTR3res, NULL_xdrproc_t, sizeof (GETATTR3res),
 872  889              nullfree, (RPC_IDEMPOTENT | RPC_ALLOWANON),
 873  890              rfs3_getattr_getfh},
 874  891  
 875  892          /* RFS3_SETATTR = 2 */
 876  893          {rfs3_setattr,
 877  894              xdr_SETATTR3args, NULL_xdrproc_t, sizeof (SETATTR3args),
 878  895              xdr_SETATTR3res, NULL_xdrproc_t, sizeof (SETATTR3res),
 879  896              nullfree, 0,
 880  897              rfs3_setattr_getfh},
 881  898  
 882  899          /* RFS3_LOOKUP = 3 */
 883  900          {rfs3_lookup,
 884  901              xdr_diropargs3, NULL_xdrproc_t, sizeof (LOOKUP3args),
 885  902              xdr_LOOKUP3res, NULL_xdrproc_t, sizeof (LOOKUP3res),
 886  903              nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK),
 887  904              rfs3_lookup_getfh},
 888  905  
 889  906          /* RFS3_ACCESS = 4 */
 890  907          {rfs3_access,
 891  908              xdr_ACCESS3args, NULL_xdrproc_t, sizeof (ACCESS3args),
 892  909              xdr_ACCESS3res, NULL_xdrproc_t, sizeof (ACCESS3res),
 893  910              nullfree, RPC_IDEMPOTENT,
 894  911              rfs3_access_getfh},
 895  912  
 896  913          /* RFS3_READLINK = 5 */
 897  914          {rfs3_readlink,
 898  915              xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (READLINK3args),
 899  916              xdr_READLINK3res, NULL_xdrproc_t, sizeof (READLINK3res),
 900  917              rfs3_readlink_free, RPC_IDEMPOTENT,
 901  918              rfs3_readlink_getfh},
 902  919  
 903  920          /* RFS3_READ = 6 */
 904  921          {rfs3_read,
 905  922              xdr_READ3args, NULL_xdrproc_t, sizeof (READ3args),
 906  923              xdr_READ3res, NULL_xdrproc_t, sizeof (READ3res),
 907  924              rfs3_read_free, RPC_IDEMPOTENT,
 908  925              rfs3_read_getfh},
 909  926  
 910  927          /* RFS3_WRITE = 7 */
 911  928          {rfs3_write,
 912  929              xdr_WRITE3args, NULL_xdrproc_t, sizeof (WRITE3args),
 913  930              xdr_WRITE3res, NULL_xdrproc_t, sizeof (WRITE3res),
 914  931              nullfree, 0,
 915  932              rfs3_write_getfh},
 916  933  
 917  934          /* RFS3_CREATE = 8 */
 918  935          {rfs3_create,
 919  936              xdr_CREATE3args, NULL_xdrproc_t, sizeof (CREATE3args),
 920  937              xdr_CREATE3res, NULL_xdrproc_t, sizeof (CREATE3res),
 921  938              nullfree, 0,
 922  939              rfs3_create_getfh},
 923  940  
 924  941          /* RFS3_MKDIR = 9 */
 925  942          {rfs3_mkdir,
 926  943              xdr_MKDIR3args, NULL_xdrproc_t, sizeof (MKDIR3args),
 927  944              xdr_MKDIR3res, NULL_xdrproc_t, sizeof (MKDIR3res),
 928  945              nullfree, 0,
 929  946              rfs3_mkdir_getfh},
 930  947  
 931  948          /* RFS3_SYMLINK = 10 */
 932  949          {rfs3_symlink,
 933  950              xdr_SYMLINK3args, NULL_xdrproc_t, sizeof (SYMLINK3args),
 934  951              xdr_SYMLINK3res, NULL_xdrproc_t, sizeof (SYMLINK3res),
 935  952              nullfree, 0,
 936  953              rfs3_symlink_getfh},
 937  954  
 938  955          /* RFS3_MKNOD = 11 */
 939  956          {rfs3_mknod,
 940  957              xdr_MKNOD3args, NULL_xdrproc_t, sizeof (MKNOD3args),
 941  958              xdr_MKNOD3res, NULL_xdrproc_t, sizeof (MKNOD3res),
 942  959              nullfree, 0,
 943  960              rfs3_mknod_getfh},
 944  961  
 945  962          /* RFS3_REMOVE = 12 */
 946  963          {rfs3_remove,
 947  964              xdr_diropargs3, NULL_xdrproc_t, sizeof (REMOVE3args),
 948  965              xdr_REMOVE3res, NULL_xdrproc_t, sizeof (REMOVE3res),
 949  966              nullfree, 0,
 950  967              rfs3_remove_getfh},
 951  968  
 952  969          /* RFS3_RMDIR = 13 */
 953  970          {rfs3_rmdir,
 954  971              xdr_diropargs3, NULL_xdrproc_t, sizeof (RMDIR3args),
 955  972              xdr_RMDIR3res, NULL_xdrproc_t, sizeof (RMDIR3res),
 956  973              nullfree, 0,
 957  974              rfs3_rmdir_getfh},
 958  975  
 959  976          /* RFS3_RENAME = 14 */
 960  977          {rfs3_rename,
 961  978              xdr_RENAME3args, NULL_xdrproc_t, sizeof (RENAME3args),
 962  979              xdr_RENAME3res, NULL_xdrproc_t, sizeof (RENAME3res),
 963  980              nullfree, 0,
 964  981              rfs3_rename_getfh},
 965  982  
 966  983          /* RFS3_LINK = 15 */
 967  984          {rfs3_link,
 968  985              xdr_LINK3args, NULL_xdrproc_t, sizeof (LINK3args),
 969  986              xdr_LINK3res, NULL_xdrproc_t, sizeof (LINK3res),
 970  987              nullfree, 0,
 971  988              rfs3_link_getfh},
 972  989  
 973  990          /* RFS3_READDIR = 16 */
 974  991          {rfs3_readdir,
 975  992              xdr_READDIR3args, NULL_xdrproc_t, sizeof (READDIR3args),
 976  993              xdr_READDIR3res, NULL_xdrproc_t, sizeof (READDIR3res),
 977  994              rfs3_readdir_free, RPC_IDEMPOTENT,
 978  995              rfs3_readdir_getfh},
 979  996  
 980  997          /* RFS3_READDIRPLUS = 17 */
 981  998          {rfs3_readdirplus,
 982  999              xdr_READDIRPLUS3args, NULL_xdrproc_t, sizeof (READDIRPLUS3args),
 983 1000              xdr_READDIRPLUS3res, NULL_xdrproc_t, sizeof (READDIRPLUS3res),
 984 1001              rfs3_readdirplus_free, RPC_AVOIDWORK,
 985 1002              rfs3_readdirplus_getfh},
 986 1003  
 987 1004          /* RFS3_FSSTAT = 18 */
 988 1005          {rfs3_fsstat,
 989 1006              xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (FSSTAT3args),
 990 1007              xdr_FSSTAT3res, NULL_xdrproc_t, sizeof (FSSTAT3res),
 991 1008              nullfree, RPC_IDEMPOTENT,
 992 1009              rfs3_fsstat_getfh},
 993 1010  
 994 1011          /* RFS3_FSINFO = 19 */
 995 1012          {rfs3_fsinfo,
 996 1013              xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (FSINFO3args),
 997 1014              xdr_FSINFO3res, NULL_xdrproc_t, sizeof (FSINFO3res),
 998 1015              nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON,
 999 1016              rfs3_fsinfo_getfh},
1000 1017  
1001 1018          /* RFS3_PATHCONF = 20 */
1002 1019          {rfs3_pathconf,
1003 1020              xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (PATHCONF3args),
1004 1021              xdr_PATHCONF3res, NULL_xdrproc_t, sizeof (PATHCONF3res),
1005 1022              nullfree, RPC_IDEMPOTENT,
1006 1023              rfs3_pathconf_getfh},
1007 1024  
1008 1025          /* RFS3_COMMIT = 21 */
1009 1026          {rfs3_commit,
1010 1027              xdr_COMMIT3args, NULL_xdrproc_t, sizeof (COMMIT3args),
1011 1028              xdr_COMMIT3res, NULL_xdrproc_t, sizeof (COMMIT3res),
1012 1029              nullfree, RPC_IDEMPOTENT,
1013 1030              rfs3_commit_getfh},
1014 1031  };
1015 1032  
1016 1033  static char *rfscallnames_v4[] = {
1017 1034          "RFS4_NULL",
1018 1035          "RFS4_COMPOUND",
1019 1036          "RFS4_NULL",
1020 1037          "RFS4_NULL",
1021 1038          "RFS4_NULL",
1022 1039          "RFS4_NULL",
1023 1040          "RFS4_NULL",
1024 1041          "RFS4_NULL",
1025 1042          "RFS4_CREATE"
1026 1043  };
1027 1044  
1028 1045  static struct rpcdisp rfsdisptab_v4[] = {
1029 1046          /*
1030 1047           * NFS VERSION 4
1031 1048           */
1032 1049  
1033 1050          /* RFS_NULL = 0 */
1034 1051          {rpc_null,
1035 1052              xdr_void, NULL_xdrproc_t, 0,
1036 1053              xdr_void, NULL_xdrproc_t, 0,
1037 1054              nullfree, RPC_IDEMPOTENT, 0},
1038 1055  
1039 1056          /* RFS4_compound = 1 */
1040 1057          {rfs4_compound,
1041 1058              xdr_COMPOUND4args_srv, NULL_xdrproc_t, sizeof (COMPOUND4args),
1042 1059              xdr_COMPOUND4res_srv, NULL_xdrproc_t, sizeof (COMPOUND4res),
1043 1060              rfs4_compound_free, 0, 0},
1044 1061  };
1045 1062  
1046 1063  union rfs_args {
1047 1064          /*
1048 1065           * NFS VERSION 2
1049 1066           */
1050 1067  
1051 1068          /* RFS_NULL = 0 */
1052 1069  
1053 1070          /* RFS_GETATTR = 1 */
1054 1071          fhandle_t nfs2_getattr_args;
1055 1072  
1056 1073          /* RFS_SETATTR = 2 */
1057 1074          struct nfssaargs nfs2_setattr_args;
1058 1075  
1059 1076          /* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */
1060 1077  
1061 1078          /* RFS_LOOKUP = 4 */
1062 1079          struct nfsdiropargs nfs2_lookup_args;
1063 1080  
1064 1081          /* RFS_READLINK = 5 */
1065 1082          fhandle_t nfs2_readlink_args;
1066 1083  
1067 1084          /* RFS_READ = 6 */
1068 1085          struct nfsreadargs nfs2_read_args;
1069 1086  
1070 1087          /* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */
1071 1088  
1072 1089          /* RFS_WRITE = 8 */
1073 1090          struct nfswriteargs nfs2_write_args;
1074 1091  
1075 1092          /* RFS_CREATE = 9 */
1076 1093          struct nfscreatargs nfs2_create_args;
1077 1094  
1078 1095          /* RFS_REMOVE = 10 */
1079 1096          struct nfsdiropargs nfs2_remove_args;
1080 1097  
1081 1098          /* RFS_RENAME = 11 */
1082 1099          struct nfsrnmargs nfs2_rename_args;
1083 1100  
1084 1101          /* RFS_LINK = 12 */
1085 1102          struct nfslinkargs nfs2_link_args;
1086 1103  
1087 1104          /* RFS_SYMLINK = 13 */
1088 1105          struct nfsslargs nfs2_symlink_args;
1089 1106  
1090 1107          /* RFS_MKDIR = 14 */
1091 1108          struct nfscreatargs nfs2_mkdir_args;
1092 1109  
1093 1110          /* RFS_RMDIR = 15 */
1094 1111          struct nfsdiropargs nfs2_rmdir_args;
1095 1112  
1096 1113          /* RFS_READDIR = 16 */
1097 1114          struct nfsrddirargs nfs2_readdir_args;
1098 1115  
1099 1116          /* RFS_STATFS = 17 */
1100 1117          fhandle_t nfs2_statfs_args;
1101 1118  
1102 1119          /*
1103 1120           * NFS VERSION 3
1104 1121           */
1105 1122  
1106 1123          /* RFS_NULL = 0 */
1107 1124  
1108 1125          /* RFS3_GETATTR = 1 */
1109 1126          GETATTR3args nfs3_getattr_args;
1110 1127  
1111 1128          /* RFS3_SETATTR = 2 */
1112 1129          SETATTR3args nfs3_setattr_args;
1113 1130  
1114 1131          /* RFS3_LOOKUP = 3 */
1115 1132          LOOKUP3args nfs3_lookup_args;
1116 1133  
1117 1134          /* RFS3_ACCESS = 4 */
1118 1135          ACCESS3args nfs3_access_args;
1119 1136  
1120 1137          /* RFS3_READLINK = 5 */
1121 1138          READLINK3args nfs3_readlink_args;
1122 1139  
1123 1140          /* RFS3_READ = 6 */
1124 1141          READ3args nfs3_read_args;
1125 1142  
1126 1143          /* RFS3_WRITE = 7 */
1127 1144          WRITE3args nfs3_write_args;
1128 1145  
1129 1146          /* RFS3_CREATE = 8 */
1130 1147          CREATE3args nfs3_create_args;
1131 1148  
1132 1149          /* RFS3_MKDIR = 9 */
1133 1150          MKDIR3args nfs3_mkdir_args;
1134 1151  
1135 1152          /* RFS3_SYMLINK = 10 */
1136 1153          SYMLINK3args nfs3_symlink_args;
1137 1154  
1138 1155          /* RFS3_MKNOD = 11 */
1139 1156          MKNOD3args nfs3_mknod_args;
1140 1157  
1141 1158          /* RFS3_REMOVE = 12 */
1142 1159          REMOVE3args nfs3_remove_args;
1143 1160  
1144 1161          /* RFS3_RMDIR = 13 */
1145 1162          RMDIR3args nfs3_rmdir_args;
1146 1163  
1147 1164          /* RFS3_RENAME = 14 */
1148 1165          RENAME3args nfs3_rename_args;
1149 1166  
1150 1167          /* RFS3_LINK = 15 */
1151 1168          LINK3args nfs3_link_args;
1152 1169  
1153 1170          /* RFS3_READDIR = 16 */
1154 1171          READDIR3args nfs3_readdir_args;
1155 1172  
1156 1173          /* RFS3_READDIRPLUS = 17 */
1157 1174          READDIRPLUS3args nfs3_readdirplus_args;
1158 1175  
1159 1176          /* RFS3_FSSTAT = 18 */
1160 1177          FSSTAT3args nfs3_fsstat_args;
1161 1178  
1162 1179          /* RFS3_FSINFO = 19 */
1163 1180          FSINFO3args nfs3_fsinfo_args;
1164 1181  
1165 1182          /* RFS3_PATHCONF = 20 */
1166 1183          PATHCONF3args nfs3_pathconf_args;
1167 1184  
1168 1185          /* RFS3_COMMIT = 21 */
1169 1186          COMMIT3args nfs3_commit_args;
1170 1187  
1171 1188          /*
1172 1189           * NFS VERSION 4
1173 1190           */
1174 1191  
1175 1192          /* RFS_NULL = 0 */
1176 1193  
1177 1194          /* COMPUND = 1 */
1178 1195          COMPOUND4args nfs4_compound_args;
1179 1196  };
1180 1197  
1181 1198  union rfs_res {
1182 1199          /*
1183 1200           * NFS VERSION 2
1184 1201           */
1185 1202  
1186 1203          /* RFS_NULL = 0 */
1187 1204  
1188 1205          /* RFS_GETATTR = 1 */
1189 1206          struct nfsattrstat nfs2_getattr_res;
1190 1207  
1191 1208          /* RFS_SETATTR = 2 */
1192 1209          struct nfsattrstat nfs2_setattr_res;
1193 1210  
1194 1211          /* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */
1195 1212  
1196 1213          /* RFS_LOOKUP = 4 */
1197 1214          struct nfsdiropres nfs2_lookup_res;
1198 1215  
1199 1216          /* RFS_READLINK = 5 */
1200 1217          struct nfsrdlnres nfs2_readlink_res;
1201 1218  
1202 1219          /* RFS_READ = 6 */
1203 1220          struct nfsrdresult nfs2_read_res;
1204 1221  
1205 1222          /* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */
1206 1223  
1207 1224          /* RFS_WRITE = 8 */
1208 1225          struct nfsattrstat nfs2_write_res;
1209 1226  
1210 1227          /* RFS_CREATE = 9 */
1211 1228          struct nfsdiropres nfs2_create_res;
1212 1229  
1213 1230          /* RFS_REMOVE = 10 */
1214 1231          enum nfsstat nfs2_remove_res;
1215 1232  
1216 1233          /* RFS_RENAME = 11 */
1217 1234          enum nfsstat nfs2_rename_res;
1218 1235  
1219 1236          /* RFS_LINK = 12 */
1220 1237          enum nfsstat nfs2_link_res;
1221 1238  
1222 1239          /* RFS_SYMLINK = 13 */
1223 1240          enum nfsstat nfs2_symlink_res;
1224 1241  
1225 1242          /* RFS_MKDIR = 14 */
1226 1243          struct nfsdiropres nfs2_mkdir_res;
1227 1244  
1228 1245          /* RFS_RMDIR = 15 */
1229 1246          enum nfsstat nfs2_rmdir_res;
1230 1247  
1231 1248          /* RFS_READDIR = 16 */
1232 1249          struct nfsrddirres nfs2_readdir_res;
1233 1250  
1234 1251          /* RFS_STATFS = 17 */
1235 1252          struct nfsstatfs nfs2_statfs_res;
1236 1253  
1237 1254          /*
1238 1255           * NFS VERSION 3
1239 1256           */
1240 1257  
1241 1258          /* RFS_NULL = 0 */
1242 1259  
1243 1260          /* RFS3_GETATTR = 1 */
1244 1261          GETATTR3res nfs3_getattr_res;
1245 1262  
1246 1263          /* RFS3_SETATTR = 2 */
1247 1264          SETATTR3res nfs3_setattr_res;
1248 1265  
1249 1266          /* RFS3_LOOKUP = 3 */
1250 1267          LOOKUP3res nfs3_lookup_res;
1251 1268  
1252 1269          /* RFS3_ACCESS = 4 */
1253 1270          ACCESS3res nfs3_access_res;
1254 1271  
1255 1272          /* RFS3_READLINK = 5 */
1256 1273          READLINK3res nfs3_readlink_res;
1257 1274  
1258 1275          /* RFS3_READ = 6 */
1259 1276          READ3res nfs3_read_res;
1260 1277  
1261 1278          /* RFS3_WRITE = 7 */
1262 1279          WRITE3res nfs3_write_res;
1263 1280  
1264 1281          /* RFS3_CREATE = 8 */
1265 1282          CREATE3res nfs3_create_res;
1266 1283  
1267 1284          /* RFS3_MKDIR = 9 */
1268 1285          MKDIR3res nfs3_mkdir_res;
1269 1286  
1270 1287          /* RFS3_SYMLINK = 10 */
1271 1288          SYMLINK3res nfs3_symlink_res;
1272 1289  
1273 1290          /* RFS3_MKNOD = 11 */
1274 1291          MKNOD3res nfs3_mknod_res;
1275 1292  
1276 1293          /* RFS3_REMOVE = 12 */
1277 1294          REMOVE3res nfs3_remove_res;
1278 1295  
1279 1296          /* RFS3_RMDIR = 13 */
1280 1297          RMDIR3res nfs3_rmdir_res;
1281 1298  
1282 1299          /* RFS3_RENAME = 14 */
1283 1300          RENAME3res nfs3_rename_res;
1284 1301  
1285 1302          /* RFS3_LINK = 15 */
1286 1303          LINK3res nfs3_link_res;
1287 1304  
1288 1305          /* RFS3_READDIR = 16 */
1289 1306          READDIR3res nfs3_readdir_res;
1290 1307  
1291 1308          /* RFS3_READDIRPLUS = 17 */
1292 1309          READDIRPLUS3res nfs3_readdirplus_res;
1293 1310  
1294 1311          /* RFS3_FSSTAT = 18 */
1295 1312          FSSTAT3res nfs3_fsstat_res;
1296 1313  
1297 1314          /* RFS3_FSINFO = 19 */
1298 1315          FSINFO3res nfs3_fsinfo_res;
1299 1316  
1300 1317          /* RFS3_PATHCONF = 20 */
1301 1318          PATHCONF3res nfs3_pathconf_res;
1302 1319  
1303 1320          /* RFS3_COMMIT = 21 */
1304 1321          COMMIT3res nfs3_commit_res;
1305 1322  
1306 1323          /*
1307 1324           * NFS VERSION 4
1308 1325           */
1309 1326  
1310 1327          /* RFS_NULL = 0 */
1311 1328  
1312 1329          /* RFS4_COMPOUND = 1 */
1313 1330          COMPOUND4res nfs4_compound_res;
1314 1331  
1315 1332  };
1316 1333  
1317 1334  static struct rpc_disptable rfs_disptable[] = {
1318 1335          {sizeof (rfsdisptab_v2) / sizeof (rfsdisptab_v2[0]),
1319 1336              rfscallnames_v2,
1320 1337              &rfsproccnt_v2_ptr, rfsdisptab_v2},
1321 1338          {sizeof (rfsdisptab_v3) / sizeof (rfsdisptab_v3[0]),
1322 1339              rfscallnames_v3,
1323 1340              &rfsproccnt_v3_ptr, rfsdisptab_v3},
1324 1341          {sizeof (rfsdisptab_v4) / sizeof (rfsdisptab_v4[0]),
1325 1342              rfscallnames_v4,
1326 1343              &rfsproccnt_v4_ptr, rfsdisptab_v4},
1327 1344  };
1328 1345  
1329 1346  /*
1330 1347   * If nfs_portmon is set, then clients are required to use privileged
1331 1348   * ports (ports < IPPORT_RESERVED) in order to get NFS services.
1332 1349   *
1333 1350   * N.B.: this attempt to carry forward the already ill-conceived notion
1334 1351   * of privileged ports for TCP/UDP is really quite ineffectual.  Not only
1335 1352   * is it transport-dependent, it's laughably easy to spoof.  If you're
1336 1353   * really interested in security, you must start with secure RPC instead.
1337 1354   */
1338 1355  static int nfs_portmon = 0;
1339 1356  
1340 1357  #ifdef DEBUG
1341 1358  static int cred_hits = 0;
1342 1359  static int cred_misses = 0;
1343 1360  #endif
1344 1361  
1345 1362  #ifdef DEBUG
1346 1363  /*
1347 1364   * Debug code to allow disabling of rfs_dispatch() use of
1348 1365   * fastxdrargs() and fastxdrres() calls for testing purposes.
1349 1366   */
1350 1367  static int rfs_no_fast_xdrargs = 0;
1351 1368  static int rfs_no_fast_xdrres = 0;
1352 1369  #endif
1353 1370  
1354 1371  union acl_args {
1355 1372          /*
1356 1373           * ACL VERSION 2
1357 1374           */
1358 1375  
1359 1376          /* ACL2_NULL = 0 */
1360 1377  
1361 1378          /* ACL2_GETACL = 1 */
1362 1379          GETACL2args acl2_getacl_args;
1363 1380  
1364 1381          /* ACL2_SETACL = 2 */
1365 1382          SETACL2args acl2_setacl_args;
1366 1383  
1367 1384          /* ACL2_GETATTR = 3 */
1368 1385          GETATTR2args acl2_getattr_args;
1369 1386  
1370 1387          /* ACL2_ACCESS = 4 */
1371 1388          ACCESS2args acl2_access_args;
1372 1389  
1373 1390          /* ACL2_GETXATTRDIR = 5 */
1374 1391          GETXATTRDIR2args acl2_getxattrdir_args;
1375 1392  
1376 1393          /*
1377 1394           * ACL VERSION 3
1378 1395           */
1379 1396  
1380 1397          /* ACL3_NULL = 0 */
1381 1398  
1382 1399          /* ACL3_GETACL = 1 */
1383 1400          GETACL3args acl3_getacl_args;
1384 1401  
1385 1402          /* ACL3_SETACL = 2 */
1386 1403          SETACL3args acl3_setacl;
1387 1404  
1388 1405          /* ACL3_GETXATTRDIR = 3 */
1389 1406          GETXATTRDIR3args acl3_getxattrdir_args;
1390 1407  
1391 1408  };
1392 1409  
1393 1410  union acl_res {
1394 1411          /*
1395 1412           * ACL VERSION 2
1396 1413           */
1397 1414  
1398 1415          /* ACL2_NULL = 0 */
1399 1416  
1400 1417          /* ACL2_GETACL = 1 */
1401 1418          GETACL2res acl2_getacl_res;
1402 1419  
1403 1420          /* ACL2_SETACL = 2 */
1404 1421          SETACL2res acl2_setacl_res;
1405 1422  
1406 1423          /* ACL2_GETATTR = 3 */
1407 1424          GETATTR2res acl2_getattr_res;
1408 1425  
1409 1426          /* ACL2_ACCESS = 4 */
1410 1427          ACCESS2res acl2_access_res;
1411 1428  
1412 1429          /* ACL2_GETXATTRDIR = 5 */
1413 1430          GETXATTRDIR2args acl2_getxattrdir_res;
1414 1431  
1415 1432          /*
1416 1433           * ACL VERSION 3
1417 1434           */
1418 1435  
1419 1436          /* ACL3_NULL = 0 */
1420 1437  
1421 1438          /* ACL3_GETACL = 1 */
1422 1439          GETACL3res acl3_getacl_res;
1423 1440  
1424 1441          /* ACL3_SETACL = 2 */
1425 1442          SETACL3res acl3_setacl_res;
1426 1443  
1427 1444          /* ACL3_GETXATTRDIR = 3 */
1428 1445          GETXATTRDIR3res acl3_getxattrdir_res;
1429 1446  
1430 1447  };
1431 1448  
1432 1449  static bool_t
1433 1450  auth_tooweak(struct svc_req *req, char *res)
1434 1451  {
1435 1452  
1436 1453          if (req->rq_vers == NFS_VERSION && req->rq_proc == RFS_LOOKUP) {
1437 1454                  struct nfsdiropres *dr = (struct nfsdiropres *)res;
1438 1455                  if ((enum wnfsstat)dr->dr_status == WNFSERR_CLNT_FLAVOR)
1439 1456                          return (TRUE);
1440 1457          } else if (req->rq_vers == NFS_V3 && req->rq_proc == NFSPROC3_LOOKUP) {
1441 1458                  LOOKUP3res *resp = (LOOKUP3res *)res;
1442 1459                  if ((enum wnfsstat)resp->status == WNFSERR_CLNT_FLAVOR)
1443 1460                          return (TRUE);
1444 1461          }
1445 1462          return (FALSE);
1446 1463  }
1447 1464  
1448 1465  
1449 1466  static void
1450 1467  common_dispatch(struct svc_req *req, SVCXPRT *xprt, rpcvers_t min_vers,
1451 1468      rpcvers_t max_vers, char *pgmname, struct rpc_disptable *disptable)
1452 1469  {
1453 1470          int which;
1454 1471          rpcvers_t vers;
1455 1472          char *args;
1456 1473          union {
1457 1474                          union rfs_args ra;
1458 1475                          union acl_args aa;
1459 1476                  } args_buf;
1460 1477          char *res;
1461 1478          union {
1462 1479                          union rfs_res rr;
1463 1480                          union acl_res ar;
1464 1481                  } res_buf;
1465 1482          struct rpcdisp *disp = NULL;
1466 1483          int dis_flags = 0;
1467 1484          cred_t *cr;
1468 1485          int error = 0;
1469 1486          int anon_ok;
1470 1487          struct exportinfo *exi = NULL;
1471 1488          unsigned int nfslog_rec_id;
1472 1489          int dupstat;
1473 1490          struct dupreq *dr;
1474 1491          int authres;
1475 1492          bool_t publicfh_ok = FALSE;
1476 1493          enum_t auth_flavor;
1477 1494          bool_t dupcached = FALSE;
1478 1495          struct netbuf   nb;
1479 1496          bool_t logging_enabled = FALSE;
1480 1497          struct exportinfo *nfslog_exi = NULL;
1481 1498          char **procnames;
1482 1499          char cbuf[INET6_ADDRSTRLEN];    /* to hold both IPv4 and IPv6 addr */
1483 1500          bool_t ro = FALSE;
1484 1501          nfs_export_t *ne = nfs_get_export();
1485 1502  
1486 1503          vers = req->rq_vers;
1487 1504  
1488 1505          if (vers < min_vers || vers > max_vers) {
1489 1506                  svcerr_progvers(req->rq_xprt, min_vers, max_vers);
1490 1507                  error++;
1491 1508                  cmn_err(CE_NOTE, "%s: bad version number %u", pgmname, vers);
1492 1509                  goto done;
1493 1510          }
1494 1511          vers -= min_vers;
1495 1512  
1496 1513          which = req->rq_proc;
1497 1514          if (which < 0 || which >= disptable[(int)vers].dis_nprocs) {
1498 1515                  svcerr_noproc(req->rq_xprt);
1499 1516                  error++;
1500 1517                  goto done;
1501 1518          }
1502 1519  
1503 1520          (*(disptable[(int)vers].dis_proccntp))[which].value.ui64++;
1504 1521  
1505 1522          disp = &disptable[(int)vers].dis_table[which];
1506 1523          procnames = disptable[(int)vers].dis_procnames;
1507 1524  
1508 1525          auth_flavor = req->rq_cred.oa_flavor;
1509 1526  
1510 1527          /*
1511 1528           * Deserialize into the args struct.
1512 1529           */
1513 1530          args = (char *)&args_buf;
1514 1531  
1515 1532  #ifdef DEBUG
1516 1533          if (rfs_no_fast_xdrargs || (auth_flavor == RPCSEC_GSS) ||
1517 1534              disp->dis_fastxdrargs == NULL_xdrproc_t ||
1518 1535              !SVC_GETARGS(xprt, disp->dis_fastxdrargs, (char *)&args))
1519 1536  #else
1520 1537          if ((auth_flavor == RPCSEC_GSS) ||
1521 1538              disp->dis_fastxdrargs == NULL_xdrproc_t ||
1522 1539              !SVC_GETARGS(xprt, disp->dis_fastxdrargs, (char *)&args))
1523 1540  #endif
1524 1541          {
1525 1542                  bzero(args, disp->dis_argsz);
1526 1543                  if (!SVC_GETARGS(xprt, disp->dis_xdrargs, args)) {
1527 1544                          error++;
1528 1545                          /*
1529 1546                           * Check if we are outside our capabilities.
1530 1547                           */
1531 1548                          if (rfs4_minorvers_mismatch(req, xprt, (void *)args))
1532 1549                                  goto done;
1533 1550  
1534 1551                          svcerr_decode(xprt);
1535 1552                          cmn_err(CE_NOTE,
1536 1553                              "Failed to decode arguments for %s version %u "
1537 1554                              "procedure %s client %s%s",
1538 1555                              pgmname, vers + min_vers, procnames[which],
1539 1556                              client_name(req), client_addr(req, cbuf));
1540 1557                          goto done;
1541 1558                  }
1542 1559          }
1543 1560  
1544 1561          /*
1545 1562           * If Version 4 use that specific dispatch function.
1546 1563           */
1547 1564          if (req->rq_vers == 4) {
1548 1565                  error += rfs4_dispatch(disp, req, xprt, args);
1549 1566                  goto done;
1550 1567          }
1551 1568  
1552 1569          dis_flags = disp->dis_flags;
1553 1570  
1554 1571          /*
1555 1572           * Find export information and check authentication,
1556 1573           * setting the credential if everything is ok.
1557 1574           */
1558 1575          if (disp->dis_getfh != NULL) {
1559 1576                  void *fh;
1560 1577                  fsid_t *fsid;
1561 1578                  fid_t *fid, *xfid;
1562 1579                  fhandle_t *fh2;
1563 1580                  nfs_fh3 *fh3;
1564 1581  
1565 1582                  fh = (*disp->dis_getfh)(args);
1566 1583                  switch (req->rq_vers) {
1567 1584                  case NFS_VERSION:
1568 1585                          fh2 = (fhandle_t *)fh;
1569 1586                          fsid = &fh2->fh_fsid;
1570 1587                          fid = (fid_t *)&fh2->fh_len;
1571 1588                          xfid = (fid_t *)&fh2->fh_xlen;
1572 1589                          break;
1573 1590                  case NFS_V3:
1574 1591                          fh3 = (nfs_fh3 *)fh;
1575 1592                          fsid = &fh3->fh3_fsid;
1576 1593                          fid = FH3TOFIDP(fh3);
1577 1594                          xfid = FH3TOXFIDP(fh3);
1578 1595                          break;
1579 1596                  }
1580 1597  
1581 1598                  /*
1582 1599                   * Fix for bug 1038302 - corbin
1583 1600                   * There is a problem here if anonymous access is
1584 1601                   * disallowed.  If the current request is part of the
1585 1602                   * client's mount process for the requested filesystem,
1586 1603                   * then it will carry root (uid 0) credentials on it, and
1587 1604                   * will be denied by checkauth if that client does not
1588 1605                   * have explicit root=0 permission.  This will cause the
1589 1606                   * client's mount operation to fail.  As a work-around,
1590 1607                   * we check here to see if the request is a getattr or
1591 1608                   * statfs operation on the exported vnode itself, and
1592 1609                   * pass a flag to checkauth with the result of this test.
1593 1610                   *
1594 1611                   * The filehandle refers to the mountpoint itself if
1595 1612                   * the fh_data and fh_xdata portions of the filehandle
1596 1613                   * are equal.
1597 1614                   *
1598 1615                   * Added anon_ok argument to checkauth().
1599 1616                   */
1600 1617  
1601 1618                  if ((dis_flags & RPC_ALLOWANON) && EQFID(fid, xfid))
1602 1619                          anon_ok = 1;
1603 1620                  else
1604 1621                          anon_ok = 0;
1605 1622  
1606 1623                  cr = xprt->xp_cred;
1607 1624                  ASSERT(cr != NULL);
1608 1625  #ifdef DEBUG
1609 1626                  {
1610 1627                          if (crgetref(cr) != 1) {
1611 1628                                  crfree(cr);
1612 1629                                  cr = crget();
1613 1630                                  xprt->xp_cred = cr;
1614 1631                                  cred_misses++;
1615 1632                          } else
1616 1633                                  cred_hits++;
1617 1634                  }
1618 1635  #else
1619 1636                  if (crgetref(cr) != 1) {
1620 1637                          crfree(cr);
1621 1638                          cr = crget();
1622 1639                          xprt->xp_cred = cr;
1623 1640                  }
1624 1641  #endif
1625 1642  
1626 1643                  exi = checkexport(fsid, xfid);
1627 1644  
1628 1645                  if (exi != NULL) {
1629 1646                          publicfh_ok = PUBLICFH_CHECK(ne, disp, exi, fsid, xfid);
1630 1647  
1631 1648                          /*
1632 1649                           * Don't allow non-V4 clients access
1633 1650                           * to pseudo exports
1634 1651                           */
1635 1652                          if (PSEUDO(exi)) {
1636 1653                                  svcerr_weakauth(xprt);
1637 1654                                  error++;
1638 1655                                  goto done;
1639 1656                          }
1640 1657  
1641 1658                          authres = checkauth(exi, req, cr, anon_ok, publicfh_ok,
1642 1659                              &ro);
1643 1660                          /*
1644 1661                           * authres >  0: authentication OK - proceed
1645 1662                           * authres == 0: authentication weak - return error
1646 1663                           * authres <  0: authentication timeout - drop
1647 1664                           */
1648 1665                          if (authres <= 0) {
1649 1666                                  if (authres == 0) {
1650 1667                                          svcerr_weakauth(xprt);
1651 1668                                          error++;
1652 1669                                  }
1653 1670                                  goto done;
1654 1671                          }
1655 1672                  }
1656 1673          } else
1657 1674                  cr = NULL;
1658 1675  
1659 1676          if ((dis_flags & RPC_MAPRESP) && (auth_flavor != RPCSEC_GSS)) {
1660 1677                  res = (char *)SVC_GETRES(xprt, disp->dis_ressz);
1661 1678                  if (res == NULL)
1662 1679                          res = (char *)&res_buf;
1663 1680          } else
1664 1681                  res = (char *)&res_buf;
1665 1682  
1666 1683          if (!(dis_flags & RPC_IDEMPOTENT)) {
1667 1684                  dupstat = SVC_DUP_EXT(xprt, req, res, disp->dis_ressz, &dr,
1668 1685                      &dupcached);
1669 1686  
1670 1687                  switch (dupstat) {
1671 1688                  case DUP_ERROR:
1672 1689                          svcerr_systemerr(xprt);
1673 1690                          error++;
1674 1691                          goto done;
1675 1692                          /* NOTREACHED */
1676 1693                  case DUP_INPROGRESS:
1677 1694                          if (res != (char *)&res_buf)
1678 1695                                  SVC_FREERES(xprt);
1679 1696                          error++;
1680 1697                          goto done;
1681 1698                          /* NOTREACHED */
1682 1699                  case DUP_NEW:
1683 1700                  case DUP_DROP:
1684 1701                          curthread->t_flag |= T_DONTPEND;
1685 1702  
1686 1703                          (*disp->dis_proc)(args, res, exi, req, cr, ro);
1687 1704  
1688 1705                          curthread->t_flag &= ~T_DONTPEND;
1689 1706                          if (curthread->t_flag & T_WOULDBLOCK) {
1690 1707                                  curthread->t_flag &= ~T_WOULDBLOCK;
1691 1708                                  SVC_DUPDONE_EXT(xprt, dr, res, NULL,
1692 1709                                      disp->dis_ressz, DUP_DROP);
1693 1710                                  if (res != (char *)&res_buf)
1694 1711                                          SVC_FREERES(xprt);
1695 1712                                  error++;
1696 1713                                  goto done;
1697 1714                          }
1698 1715                          if (dis_flags & RPC_AVOIDWORK) {
1699 1716                                  SVC_DUPDONE_EXT(xprt, dr, res, NULL,
1700 1717                                      disp->dis_ressz, DUP_DROP);
1701 1718                          } else {
1702 1719                                  SVC_DUPDONE_EXT(xprt, dr, res,
1703 1720                                      disp->dis_resfree == nullfree ? NULL :
1704 1721                                      disp->dis_resfree,
1705 1722                                      disp->dis_ressz, DUP_DONE);
1706 1723                                  dupcached = TRUE;
1707 1724                          }
1708 1725                          break;
1709 1726                  case DUP_DONE:
1710 1727                          break;
1711 1728                  }
1712 1729  
1713 1730          } else {
1714 1731                  curthread->t_flag |= T_DONTPEND;
1715 1732  
1716 1733                  (*disp->dis_proc)(args, res, exi, req, cr, ro);
1717 1734  
1718 1735                  curthread->t_flag &= ~T_DONTPEND;
1719 1736                  if (curthread->t_flag & T_WOULDBLOCK) {
1720 1737                          curthread->t_flag &= ~T_WOULDBLOCK;
1721 1738                          if (res != (char *)&res_buf)
1722 1739                                  SVC_FREERES(xprt);
1723 1740                          error++;
1724 1741                          goto done;
1725 1742                  }
1726 1743          }
1727 1744  
1728 1745          if (auth_tooweak(req, res)) {
1729 1746                  svcerr_weakauth(xprt);
1730 1747                  error++;
1731 1748                  goto done;
1732 1749          }
1733 1750  
1734 1751          /*
1735 1752           * Check to see if logging has been enabled on the server.
1736 1753           * If so, then obtain the export info struct to be used for
1737 1754           * the later writing of the log record.  This is done for
1738 1755           * the case that a lookup is done across a non-logged public
1739 1756           * file system.
1740 1757           */
1741 1758          if (nfslog_buffer_list != NULL) {
1742 1759                  nfslog_exi = nfslog_get_exi(ne, exi, req, res, &nfslog_rec_id);
1743 1760                  /*
1744 1761                   * Is logging enabled?
1745 1762                   */
1746 1763                  logging_enabled = (nfslog_exi != NULL);
1747 1764  
1748 1765                  /*
1749 1766                   * Copy the netbuf for logging purposes, before it is
1750 1767                   * freed by svc_sendreply().
1751 1768                   */
1752 1769                  if (logging_enabled) {
1753 1770                          NFSLOG_COPY_NETBUF(nfslog_exi, xprt, &nb);
1754 1771                          /*
1755 1772                           * If RPC_MAPRESP flag set (i.e. in V2 ops) the
1756 1773                           * res gets copied directly into the mbuf and
1757 1774                           * may be freed soon after the sendreply. So we
1758 1775                           * must copy it here to a safe place...
1759 1776                           */
1760 1777                          if (res != (char *)&res_buf) {
1761 1778                                  bcopy(res, (char *)&res_buf, disp->dis_ressz);
1762 1779                          }
1763 1780                  }
1764 1781          }
1765 1782  
1766 1783          /*
1767 1784           * Serialize and send results struct
1768 1785           */
1769 1786  #ifdef DEBUG
1770 1787          if (rfs_no_fast_xdrres == 0 && res != (char *)&res_buf)
1771 1788  #else
1772 1789          if (res != (char *)&res_buf)
1773 1790  #endif
1774 1791          {
1775 1792                  if (!svc_sendreply(xprt, disp->dis_fastxdrres, res)) {
1776 1793                          cmn_err(CE_NOTE, "%s: bad sendreply", pgmname);
1777 1794                          svcerr_systemerr(xprt);
1778 1795                          error++;
1779 1796                  }
1780 1797          } else {
1781 1798                  if (!svc_sendreply(xprt, disp->dis_xdrres, res)) {
1782 1799                          cmn_err(CE_NOTE, "%s: bad sendreply", pgmname);
1783 1800                          svcerr_systemerr(xprt);
1784 1801                          error++;
1785 1802                  }
1786 1803          }
1787 1804  
1788 1805          /*
1789 1806           * Log if needed
1790 1807           */
1791 1808          if (logging_enabled) {
1792 1809                  nfslog_write_record(nfslog_exi, req, args, (char *)&res_buf,
1793 1810                      cr, &nb, nfslog_rec_id, NFSLOG_ONE_BUFFER);
1794 1811                  exi_rele(nfslog_exi);
1795 1812                  kmem_free((&nb)->buf, (&nb)->len);
1796 1813          }
1797 1814  
1798 1815          /*
1799 1816           * Free results struct. With the addition of NFS V4 we can
1800 1817           * have non-idempotent procedures with functions.
1801 1818           */
1802 1819          if (disp->dis_resfree != nullfree && dupcached == FALSE) {
1803 1820                  (*disp->dis_resfree)(res);
1804 1821          }
1805 1822  
1806 1823  done:
1807 1824          /*
1808 1825           * Free arguments struct
1809 1826           */
1810 1827          if (disp) {
1811 1828                  if (!SVC_FREEARGS(xprt, disp->dis_xdrargs, args)) {
1812 1829                          cmn_err(CE_NOTE, "%s: bad freeargs", pgmname);
1813 1830                          error++;
1814 1831                  }
1815 1832          } else {
1816 1833                  if (!SVC_FREEARGS(xprt, (xdrproc_t)0, (caddr_t)0)) {
1817 1834                          cmn_err(CE_NOTE, "%s: bad freeargs", pgmname);
1818 1835                          error++;
1819 1836                  }
1820 1837          }
1821 1838  
1822 1839          if (exi != NULL)
1823 1840                  exi_rele(exi);
1824 1841  
1825 1842          global_svstat_ptr[req->rq_vers][NFS_BADCALLS].value.ui64 += error;
1826 1843  
1827 1844          global_svstat_ptr[req->rq_vers][NFS_CALLS].value.ui64++;
1828 1845  }
1829 1846  
1830 1847  static void
1831 1848  rfs_dispatch(struct svc_req *req, SVCXPRT *xprt)
1832 1849  {
1833 1850          common_dispatch(req, xprt, NFS_VERSMIN, NFS_VERSMAX,
1834 1851              "NFS", rfs_disptable);
1835 1852  }
1836 1853  
1837 1854  static char *aclcallnames_v2[] = {
1838 1855          "ACL2_NULL",
1839 1856          "ACL2_GETACL",
1840 1857          "ACL2_SETACL",
1841 1858          "ACL2_GETATTR",
1842 1859          "ACL2_ACCESS",
1843 1860          "ACL2_GETXATTRDIR"
1844 1861  };
1845 1862  
1846 1863  static struct rpcdisp acldisptab_v2[] = {
1847 1864          /*
1848 1865           * ACL VERSION 2
1849 1866           */
1850 1867  
1851 1868          /* ACL2_NULL = 0 */
1852 1869          {rpc_null,
1853 1870              xdr_void, NULL_xdrproc_t, 0,
1854 1871              xdr_void, NULL_xdrproc_t, 0,
1855 1872              nullfree, RPC_IDEMPOTENT,
1856 1873              0},
1857 1874  
1858 1875          /* ACL2_GETACL = 1 */
1859 1876          {acl2_getacl,
1860 1877              xdr_GETACL2args, xdr_fastGETACL2args, sizeof (GETACL2args),
1861 1878              xdr_GETACL2res, NULL_xdrproc_t, sizeof (GETACL2res),
1862 1879              acl2_getacl_free, RPC_IDEMPOTENT,
1863 1880              acl2_getacl_getfh},
1864 1881  
1865 1882          /* ACL2_SETACL = 2 */
1866 1883          {acl2_setacl,
1867 1884              xdr_SETACL2args, NULL_xdrproc_t, sizeof (SETACL2args),
1868 1885  #ifdef _LITTLE_ENDIAN
1869 1886              xdr_SETACL2res, xdr_fastSETACL2res, sizeof (SETACL2res),
1870 1887  #else
1871 1888              xdr_SETACL2res, NULL_xdrproc_t, sizeof (SETACL2res),
1872 1889  #endif
1873 1890              nullfree, RPC_MAPRESP,
1874 1891              acl2_setacl_getfh},
1875 1892  
1876 1893          /* ACL2_GETATTR = 3 */
1877 1894          {acl2_getattr,
1878 1895              xdr_GETATTR2args, xdr_fastGETATTR2args, sizeof (GETATTR2args),
1879 1896  #ifdef _LITTLE_ENDIAN
1880 1897              xdr_GETATTR2res, xdr_fastGETATTR2res, sizeof (GETATTR2res),
1881 1898  #else
1882 1899              xdr_GETATTR2res, NULL_xdrproc_t, sizeof (GETATTR2res),
1883 1900  #endif
1884 1901              nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP,
1885 1902              acl2_getattr_getfh},
1886 1903  
1887 1904          /* ACL2_ACCESS = 4 */
1888 1905          {acl2_access,
1889 1906              xdr_ACCESS2args, xdr_fastACCESS2args, sizeof (ACCESS2args),
1890 1907  #ifdef _LITTLE_ENDIAN
1891 1908              xdr_ACCESS2res, xdr_fastACCESS2res, sizeof (ACCESS2res),
1892 1909  #else
1893 1910              xdr_ACCESS2res, NULL_xdrproc_t, sizeof (ACCESS2res),
1894 1911  #endif
1895 1912              nullfree, RPC_IDEMPOTENT|RPC_MAPRESP,
1896 1913              acl2_access_getfh},
1897 1914  
1898 1915          /* ACL2_GETXATTRDIR = 5 */
1899 1916          {acl2_getxattrdir,
1900 1917              xdr_GETXATTRDIR2args, NULL_xdrproc_t, sizeof (GETXATTRDIR2args),
1901 1918              xdr_GETXATTRDIR2res, NULL_xdrproc_t, sizeof (GETXATTRDIR2res),
1902 1919              nullfree, RPC_IDEMPOTENT,
1903 1920              acl2_getxattrdir_getfh},
1904 1921  };
1905 1922  
1906 1923  static char *aclcallnames_v3[] = {
1907 1924          "ACL3_NULL",
1908 1925          "ACL3_GETACL",
1909 1926          "ACL3_SETACL",
1910 1927          "ACL3_GETXATTRDIR"
1911 1928  };
1912 1929  
1913 1930  static struct rpcdisp acldisptab_v3[] = {
1914 1931          /*
1915 1932           * ACL VERSION 3
1916 1933           */
1917 1934  
1918 1935          /* ACL3_NULL = 0 */
1919 1936          {rpc_null,
1920 1937              xdr_void, NULL_xdrproc_t, 0,
1921 1938              xdr_void, NULL_xdrproc_t, 0,
1922 1939              nullfree, RPC_IDEMPOTENT,
1923 1940              0},
1924 1941  
1925 1942          /* ACL3_GETACL = 1 */
1926 1943          {acl3_getacl,
1927 1944              xdr_GETACL3args, NULL_xdrproc_t, sizeof (GETACL3args),
1928 1945              xdr_GETACL3res, NULL_xdrproc_t, sizeof (GETACL3res),
1929 1946              acl3_getacl_free, RPC_IDEMPOTENT,
1930 1947              acl3_getacl_getfh},
1931 1948  
1932 1949          /* ACL3_SETACL = 2 */
1933 1950          {acl3_setacl,
1934 1951              xdr_SETACL3args, NULL_xdrproc_t, sizeof (SETACL3args),
1935 1952              xdr_SETACL3res, NULL_xdrproc_t, sizeof (SETACL3res),
1936 1953              nullfree, 0,
1937 1954              acl3_setacl_getfh},
1938 1955  
1939 1956          /* ACL3_GETXATTRDIR = 3 */
1940 1957          {acl3_getxattrdir,
1941 1958              xdr_GETXATTRDIR3args, NULL_xdrproc_t, sizeof (GETXATTRDIR3args),
1942 1959              xdr_GETXATTRDIR3res, NULL_xdrproc_t, sizeof (GETXATTRDIR3res),
1943 1960              nullfree, RPC_IDEMPOTENT,
1944 1961              acl3_getxattrdir_getfh},
1945 1962  };
1946 1963  
1947 1964  static struct rpc_disptable acl_disptable[] = {
1948 1965          {sizeof (acldisptab_v2) / sizeof (acldisptab_v2[0]),
1949 1966                  aclcallnames_v2,
1950 1967                  &aclproccnt_v2_ptr, acldisptab_v2},
1951 1968          {sizeof (acldisptab_v3) / sizeof (acldisptab_v3[0]),
1952 1969                  aclcallnames_v3,
1953 1970                  &aclproccnt_v3_ptr, acldisptab_v3},
1954 1971  };
1955 1972  
1956 1973  static void
1957 1974  acl_dispatch(struct svc_req *req, SVCXPRT *xprt)
1958 1975  {
1959 1976          common_dispatch(req, xprt, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX,
1960 1977              "ACL", acl_disptable);
1961 1978  }
1962 1979  
1963 1980  int
1964 1981  checkwin(int flavor, int window, struct svc_req *req)
1965 1982  {
1966 1983          struct authdes_cred *adc;
1967 1984  
1968 1985          switch (flavor) {
1969 1986          case AUTH_DES:
1970 1987                  adc = (struct authdes_cred *)req->rq_clntcred;
1971 1988                  CTASSERT(sizeof (struct authdes_cred) <= RQCRED_SIZE);
1972 1989                  if (adc->adc_fullname.window > window)
1973 1990                          return (0);
1974 1991                  break;
1975 1992  
1976 1993          default:
1977 1994                  break;
1978 1995          }
1979 1996          return (1);
1980 1997  }
1981 1998  
1982 1999  
1983 2000  /*
1984 2001   * checkauth() will check the access permission against the export
1985 2002   * information.  Then map root uid/gid to appropriate uid/gid.
1986 2003   *
1987 2004   * This routine is used by NFS V3 and V2 code.
1988 2005   */
1989 2006  static int
1990 2007  checkauth(struct exportinfo *exi, struct svc_req *req, cred_t *cr, int anon_ok,
1991 2008      bool_t publicfh_ok, bool_t *ro)
1992 2009  {
1993 2010          int i, nfsflavor, rpcflavor, stat, access;
1994 2011          struct secinfo *secp;
1995 2012          caddr_t principal;
1996 2013          char buf[INET6_ADDRSTRLEN]; /* to hold both IPv4 and IPv6 addr */
1997 2014          int anon_res = 0;
1998 2015  
1999 2016          uid_t uid;
2000 2017          gid_t gid;
2001 2018          uint_t ngids;
2002 2019          gid_t *gids;
2003 2020  
2004 2021          /*
2005 2022           * Check for privileged port number
2006 2023           * N.B.:  this assumes that we know the format of a netbuf.
2007 2024           */
2008 2025          if (nfs_portmon) {
2009 2026                  struct sockaddr *ca;
2010 2027                  ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
2011 2028  
2012 2029                  if (ca == NULL)
2013 2030                          return (0);
2014 2031  
2015 2032                  if ((ca->sa_family == AF_INET &&
2016 2033                      ntohs(((struct sockaddr_in *)ca)->sin_port) >=
2017 2034                      IPPORT_RESERVED) ||
2018 2035                      (ca->sa_family == AF_INET6 &&
2019 2036                      ntohs(((struct sockaddr_in6 *)ca)->sin6_port) >=
2020 2037                      IPPORT_RESERVED)) {
2021 2038                          cmn_err(CE_NOTE,
2022 2039                              "nfs_server: client %s%ssent NFS request from "
2023 2040                              "unprivileged port",
2024 2041                              client_name(req), client_addr(req, buf));
2025 2042                          return (0);
2026 2043                  }
2027 2044          }
2028 2045  
2029 2046          /*
2030 2047           *  return 1 on success or 0 on failure
2031 2048           */
2032 2049          stat = sec_svc_getcred(req, cr, &principal, &nfsflavor);
2033 2050  
2034 2051          /*
2035 2052           * A failed AUTH_UNIX sec_svc_getcred() implies we couldn't set
2036 2053           * the credentials; below we map that to anonymous.
2037 2054           */
2038 2055          if (!stat && nfsflavor != AUTH_UNIX) {
2039 2056                  cmn_err(CE_NOTE,
2040 2057                      "nfs_server: couldn't get unix cred for %s",
2041 2058                      client_name(req));
2042 2059                  return (0);
2043 2060          }
2044 2061  
2045 2062          /*
2046 2063           * Short circuit checkauth() on operations that support the
2047 2064           * public filehandle, and if the request for that operation
2048 2065           * is using the public filehandle. Note that we must call
2049 2066           * sec_svc_getcred() first so that xp_cookie is set to the
2050 2067           * right value. Normally xp_cookie is just the RPC flavor
2051 2068           * of the the request, but in the case of RPCSEC_GSS it
2052 2069           * could be a pseudo flavor.
2053 2070           */
2054 2071          if (publicfh_ok)
2055 2072                  return (1);
2056 2073  
2057 2074          rpcflavor = req->rq_cred.oa_flavor;
2058 2075          /*
2059 2076           * Check if the auth flavor is valid for this export
2060 2077           */
2061 2078          access = nfsauth_access(exi, req, cr, &uid, &gid, &ngids, &gids);
2062 2079          if (access & NFSAUTH_DROP)
2063 2080                  return (-1);    /* drop the request */
2064 2081  
2065 2082          if (access & NFSAUTH_RO)
2066 2083                  *ro = TRUE;
2067 2084  
2068 2085          if (access & NFSAUTH_DENIED) {
2069 2086                  /*
2070 2087                   * If anon_ok == 1 and we got NFSAUTH_DENIED, it was
2071 2088                   * probably due to the flavor not matching during
2072 2089                   * the mount attempt. So map the flavor to AUTH_NONE
2073 2090                   * so that the credentials get mapped to the anonymous
2074 2091                   * user.
2075 2092                   */
2076 2093                  if (anon_ok == 1)
2077 2094                          rpcflavor = AUTH_NONE;
2078 2095                  else
2079 2096                          return (0);     /* deny access */
2080 2097  
2081 2098          } else if (access & NFSAUTH_MAPNONE) {
2082 2099                  /*
2083 2100                   * Access was granted even though the flavor mismatched
2084 2101                   * because AUTH_NONE was one of the exported flavors.
2085 2102                   */
2086 2103                  rpcflavor = AUTH_NONE;
2087 2104  
2088 2105          } else if (access & NFSAUTH_WRONGSEC) {
2089 2106                  /*
2090 2107                   * NFSAUTH_WRONGSEC is used for NFSv4. If we get here,
2091 2108                   * it means a client ignored the list of allowed flavors
2092 2109                   * returned via the MOUNT protocol. So we just disallow it!
2093 2110                   */
2094 2111                  return (0);
2095 2112          }
2096 2113  
2097 2114          if (rpcflavor != AUTH_SYS)
2098 2115                  kmem_free(gids, ngids * sizeof (gid_t));
2099 2116  
2100 2117          switch (rpcflavor) {
2101 2118          case AUTH_NONE:
2102 2119                  anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2103 2120                      exi->exi_export.ex_anon);
2104 2121                  (void) crsetgroups(cr, 0, NULL);
2105 2122                  break;
2106 2123  
2107 2124          case AUTH_UNIX:
2108 2125                  if (!stat || crgetuid(cr) == 0 && !(access & NFSAUTH_UIDMAP)) {
2109 2126                          anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2110 2127                              exi->exi_export.ex_anon);
2111 2128                          (void) crsetgroups(cr, 0, NULL);
2112 2129                  } else if (crgetuid(cr) == 0 && access & NFSAUTH_ROOT) {
2113 2130                          /*
2114 2131                           * It is root, so apply rootid to get real UID
2115 2132                           * Find the secinfo structure.  We should be able
2116 2133                           * to find it by the time we reach here.
2117 2134                           * nfsauth_access() has done the checking.
2118 2135                           */
2119 2136                          secp = NULL;
2120 2137                          for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
2121 2138                                  struct secinfo *sptr;
2122 2139                                  sptr = &exi->exi_export.ex_secinfo[i];
2123 2140                                  if (sptr->s_secinfo.sc_nfsnum == nfsflavor) {
2124 2141                                          secp = sptr;
2125 2142                                          break;
2126 2143                                  }
2127 2144                          }
2128 2145                          if (secp != NULL) {
2129 2146                                  (void) crsetugid(cr, secp->s_rootid,
2130 2147                                      secp->s_rootid);
2131 2148                                  (void) crsetgroups(cr, 0, NULL);
2132 2149                          }
2133 2150                  } else if (crgetuid(cr) != uid || crgetgid(cr) != gid) {
2134 2151                          if (crsetugid(cr, uid, gid) != 0)
2135 2152                                  anon_res = crsetugid(cr,
2136 2153                                      exi->exi_export.ex_anon,
2137 2154                                      exi->exi_export.ex_anon);
2138 2155                          (void) crsetgroups(cr, 0, NULL);
2139 2156                  } else if (access & NFSAUTH_GROUPS) {
2140 2157                          (void) crsetgroups(cr, ngids, gids);
2141 2158                  }
2142 2159  
2143 2160                  kmem_free(gids, ngids * sizeof (gid_t));
2144 2161  
2145 2162                  break;
2146 2163  
2147 2164          case AUTH_DES:
2148 2165          case RPCSEC_GSS:
2149 2166                  /*
2150 2167                   *  Find the secinfo structure.  We should be able
2151 2168                   *  to find it by the time we reach here.
2152 2169                   *  nfsauth_access() has done the checking.
2153 2170                   */
2154 2171                  secp = NULL;
2155 2172                  for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
2156 2173                          if (exi->exi_export.ex_secinfo[i].s_secinfo.sc_nfsnum ==
2157 2174                              nfsflavor) {
2158 2175                                  secp = &exi->exi_export.ex_secinfo[i];
2159 2176                                  break;
2160 2177                          }
2161 2178                  }
2162 2179  
2163 2180                  if (!secp) {
2164 2181                          cmn_err(CE_NOTE, "nfs_server: client %s%shad "
2165 2182                              "no secinfo data for flavor %d",
2166 2183                              client_name(req), client_addr(req, buf),
2167 2184                              nfsflavor);
2168 2185                          return (0);
2169 2186                  }
2170 2187  
2171 2188                  if (!checkwin(rpcflavor, secp->s_window, req)) {
2172 2189                          cmn_err(CE_NOTE,
2173 2190                              "nfs_server: client %s%sused invalid "
2174 2191                              "auth window value",
2175 2192                              client_name(req), client_addr(req, buf));
2176 2193                          return (0);
2177 2194                  }
2178 2195  
2179 2196                  /*
2180 2197                   * Map root principals listed in the share's root= list to root,
2181 2198                   * and map any others principals that were mapped to root by RPC
2182 2199                   * to anon.
2183 2200                   */
2184 2201                  if (principal && sec_svc_inrootlist(rpcflavor, principal,
2185 2202                      secp->s_rootcnt, secp->s_rootnames)) {
2186 2203                          if (crgetuid(cr) == 0 && secp->s_rootid == 0)
2187 2204                                  return (1);
2188 2205  
2189 2206  
2190 2207                          (void) crsetugid(cr, secp->s_rootid, secp->s_rootid);
2191 2208  
2192 2209                          /*
2193 2210                           * NOTE: If and when kernel-land privilege tracing is
2194 2211                           * added this may have to be replaced with code that
2195 2212                           * retrieves root's supplementary groups (e.g., using
2196 2213                           * kgss_get_group_info().  In the meantime principals
2197 2214                           * mapped to uid 0 get all privileges, so setting cr's
2198 2215                           * supplementary groups for them does nothing.
2199 2216                           */
2200 2217                          (void) crsetgroups(cr, 0, NULL);
2201 2218  
2202 2219                          return (1);
2203 2220                  }
2204 2221  
2205 2222                  /*
2206 2223                   * Not a root princ, or not in root list, map UID 0/nobody to
2207 2224                   * the anon ID for the share.  (RPC sets cr's UIDs and GIDs to
2208 2225                   * UID_NOBODY and GID_NOBODY, respectively.)
2209 2226                   */
2210 2227                  if (crgetuid(cr) != 0 &&
2211 2228                      (crgetuid(cr) != UID_NOBODY || crgetgid(cr) != GID_NOBODY))
2212 2229                          return (1);
2213 2230  
2214 2231                  anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2215 2232                      exi->exi_export.ex_anon);
2216 2233                  (void) crsetgroups(cr, 0, NULL);
2217 2234                  break;
2218 2235          default:
2219 2236                  return (0);
2220 2237          } /* switch on rpcflavor */
2221 2238  
2222 2239          /*
2223 2240           * Even if anon access is disallowed via ex_anon == -1, we allow
2224 2241           * this access if anon_ok is set.  So set creds to the default
2225 2242           * "nobody" id.
2226 2243           */
2227 2244          if (anon_res != 0) {
2228 2245                  if (anon_ok == 0) {
2229 2246                          cmn_err(CE_NOTE,
2230 2247                              "nfs_server: client %s%ssent wrong "
2231 2248                              "authentication for %s",
2232 2249                              client_name(req), client_addr(req, buf),
2233 2250                              exi->exi_export.ex_path ?
2234 2251                              exi->exi_export.ex_path : "?");
2235 2252                          return (0);
2236 2253                  }
2237 2254  
2238 2255                  if (crsetugid(cr, UID_NOBODY, GID_NOBODY) != 0)
2239 2256                          return (0);
2240 2257          }
2241 2258  
2242 2259          return (1);
2243 2260  }
2244 2261  
2245 2262  /*
2246 2263   * returns 0 on failure, -1 on a drop, -2 on wrong security flavor,
2247 2264   * and 1 on success
2248 2265   */
2249 2266  int
2250 2267  checkauth4(struct compound_state *cs, struct svc_req *req)
2251 2268  {
2252 2269          int i, rpcflavor, access;
2253 2270          struct secinfo *secp;
2254 2271          char buf[MAXHOST + 1];
2255 2272          int anon_res = 0, nfsflavor;
2256 2273          struct exportinfo *exi;
2257 2274          cred_t  *cr;
2258 2275          caddr_t principal;
2259 2276  
2260 2277          uid_t uid;
2261 2278          gid_t gid;
2262 2279          uint_t ngids;
2263 2280          gid_t *gids;
2264 2281  
2265 2282          exi = cs->exi;
2266 2283          cr = cs->cr;
2267 2284          principal = cs->principal;
2268 2285          nfsflavor = cs->nfsflavor;
2269 2286  
2270 2287          ASSERT(cr != NULL);
2271 2288  
2272 2289          rpcflavor = req->rq_cred.oa_flavor;
2273 2290          cs->access &= ~CS_ACCESS_LIMITED;
2274 2291  
2275 2292          /*
2276 2293           * Check for privileged port number
2277 2294           * N.B.:  this assumes that we know the format of a netbuf.
2278 2295           */
2279 2296          if (nfs_portmon) {
2280 2297                  struct sockaddr *ca;
2281 2298                  ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
2282 2299  
2283 2300                  if (ca == NULL)
2284 2301                          return (0);
2285 2302  
2286 2303                  if ((ca->sa_family == AF_INET &&
2287 2304                      ntohs(((struct sockaddr_in *)ca)->sin_port) >=
2288 2305                      IPPORT_RESERVED) ||
2289 2306                      (ca->sa_family == AF_INET6 &&
2290 2307                      ntohs(((struct sockaddr_in6 *)ca)->sin6_port) >=
2291 2308                      IPPORT_RESERVED)) {
2292 2309                          cmn_err(CE_NOTE,
2293 2310                              "nfs_server: client %s%ssent NFSv4 request from "
2294 2311                              "unprivileged port",
2295 2312                              client_name(req), client_addr(req, buf));
2296 2313                          return (0);
2297 2314                  }
2298 2315          }
2299 2316  
2300 2317          /*
2301 2318           * Check the access right per auth flavor on the vnode of
2302 2319           * this export for the given request.
2303 2320           */
2304 2321          access = nfsauth4_access(cs->exi, cs->vp, req, cr, &uid, &gid, &ngids,
2305 2322              &gids);
2306 2323  
2307 2324          if (access & NFSAUTH_WRONGSEC)
2308 2325                  return (-2);    /* no access for this security flavor */
2309 2326  
2310 2327          if (access & NFSAUTH_DROP)
2311 2328                  return (-1);    /* drop the request */
2312 2329  
2313 2330          if (access & NFSAUTH_DENIED) {
2314 2331  
2315 2332                  if (exi->exi_export.ex_seccnt > 0)
2316 2333                          return (0);     /* deny access */
2317 2334  
2318 2335          } else if (access & NFSAUTH_LIMITED) {
2319 2336  
2320 2337                  cs->access |= CS_ACCESS_LIMITED;
2321 2338  
2322 2339          } else if (access & NFSAUTH_MAPNONE) {
2323 2340                  /*
2324 2341                   * Access was granted even though the flavor mismatched
2325 2342                   * because AUTH_NONE was one of the exported flavors.
2326 2343                   */
2327 2344                  rpcflavor = AUTH_NONE;
2328 2345          }
2329 2346  
2330 2347          /*
2331 2348           * XXX probably need to redo some of it for nfsv4?
2332 2349           * return 1 on success or 0 on failure
2333 2350           */
2334 2351  
2335 2352          if (rpcflavor != AUTH_SYS)
2336 2353                  kmem_free(gids, ngids * sizeof (gid_t));
2337 2354  
2338 2355          switch (rpcflavor) {
2339 2356          case AUTH_NONE:
2340 2357                  anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2341 2358                      exi->exi_export.ex_anon);
2342 2359                  (void) crsetgroups(cr, 0, NULL);
2343 2360                  break;
2344 2361  
2345 2362          case AUTH_UNIX:
2346 2363                  if (crgetuid(cr) == 0 && !(access & NFSAUTH_UIDMAP)) {
2347 2364                          anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2348 2365                              exi->exi_export.ex_anon);
2349 2366                          (void) crsetgroups(cr, 0, NULL);
2350 2367                  } else if (crgetuid(cr) == 0 && access & NFSAUTH_ROOT) {
2351 2368                          /*
2352 2369                           * It is root, so apply rootid to get real UID
2353 2370                           * Find the secinfo structure.  We should be able
2354 2371                           * to find it by the time we reach here.
2355 2372                           * nfsauth_access() has done the checking.
2356 2373                           */
2357 2374                          secp = NULL;
2358 2375                          for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
2359 2376                                  struct secinfo *sptr;
2360 2377                                  sptr = &exi->exi_export.ex_secinfo[i];
2361 2378                                  if (sptr->s_secinfo.sc_nfsnum == nfsflavor) {
2362 2379                                          secp = &exi->exi_export.ex_secinfo[i];
2363 2380                                          break;
2364 2381                                  }
2365 2382                          }
2366 2383                          if (secp != NULL) {
2367 2384                                  (void) crsetugid(cr, secp->s_rootid,
2368 2385                                      secp->s_rootid);
2369 2386                                  (void) crsetgroups(cr, 0, NULL);
2370 2387                          }
2371 2388                  } else if (crgetuid(cr) != uid || crgetgid(cr) != gid) {
2372 2389                          if (crsetugid(cr, uid, gid) != 0)
2373 2390                                  anon_res = crsetugid(cr,
2374 2391                                      exi->exi_export.ex_anon,
2375 2392                                      exi->exi_export.ex_anon);
2376 2393                          (void) crsetgroups(cr, 0, NULL);
2377 2394                  } if (access & NFSAUTH_GROUPS) {
2378 2395                          (void) crsetgroups(cr, ngids, gids);
2379 2396                  }
2380 2397  
2381 2398                  kmem_free(gids, ngids * sizeof (gid_t));
2382 2399  
2383 2400                  break;
2384 2401  
2385 2402          default:
2386 2403                  /*
2387 2404                   *  Find the secinfo structure.  We should be able
2388 2405                   *  to find it by the time we reach here.
2389 2406                   *  nfsauth_access() has done the checking.
2390 2407                   */
2391 2408                  secp = NULL;
2392 2409                  for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
2393 2410                          if (exi->exi_export.ex_secinfo[i].s_secinfo.sc_nfsnum ==
2394 2411                              nfsflavor) {
2395 2412                                  secp = &exi->exi_export.ex_secinfo[i];
2396 2413                                  break;
2397 2414                          }
2398 2415                  }
2399 2416  
2400 2417                  if (!secp) {
2401 2418                          cmn_err(CE_NOTE, "nfs_server: client %s%shad "
2402 2419                              "no secinfo data for flavor %d",
2403 2420                              client_name(req), client_addr(req, buf),
2404 2421                              nfsflavor);
2405 2422                          return (0);
2406 2423                  }
2407 2424  
2408 2425                  if (!checkwin(rpcflavor, secp->s_window, req)) {
2409 2426                          cmn_err(CE_NOTE,
2410 2427                              "nfs_server: client %s%sused invalid "
2411 2428                              "auth window value",
2412 2429                              client_name(req), client_addr(req, buf));
2413 2430                          return (0);
2414 2431                  }
2415 2432  
2416 2433                  /*
2417 2434                   * Map root principals listed in the share's root= list to root,
2418 2435                   * and map any others principals that were mapped to root by RPC
2419 2436                   * to anon. If not going to anon, set to rootid (root_mapping).
2420 2437                   */
2421 2438                  if (principal && sec_svc_inrootlist(rpcflavor, principal,
2422 2439                      secp->s_rootcnt, secp->s_rootnames)) {
2423 2440                          if (crgetuid(cr) == 0 && secp->s_rootid == 0)
2424 2441                                  return (1);
2425 2442  
2426 2443                          (void) crsetugid(cr, secp->s_rootid, secp->s_rootid);
2427 2444  
2428 2445                          /*
2429 2446                           * NOTE: If and when kernel-land privilege tracing is
2430 2447                           * added this may have to be replaced with code that
2431 2448                           * retrieves root's supplementary groups (e.g., using
2432 2449                           * kgss_get_group_info().  In the meantime principals
2433 2450                           * mapped to uid 0 get all privileges, so setting cr's
2434 2451                           * supplementary groups for them does nothing.
2435 2452                           */
2436 2453                          (void) crsetgroups(cr, 0, NULL);
2437 2454  
2438 2455                          return (1);
2439 2456                  }
2440 2457  
2441 2458                  /*
2442 2459                   * Not a root princ, or not in root list, map UID 0/nobody to
2443 2460                   * the anon ID for the share.  (RPC sets cr's UIDs and GIDs to
2444 2461                   * UID_NOBODY and GID_NOBODY, respectively.)
2445 2462                   */
2446 2463                  if (crgetuid(cr) != 0 &&
2447 2464                      (crgetuid(cr) != UID_NOBODY || crgetgid(cr) != GID_NOBODY))
2448 2465                          return (1);
2449 2466  
2450 2467                  anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2451 2468                      exi->exi_export.ex_anon);
2452 2469                  (void) crsetgroups(cr, 0, NULL);
2453 2470                  break;
2454 2471          } /* switch on rpcflavor */
2455 2472  
2456 2473          /*
2457 2474           * Even if anon access is disallowed via ex_anon == -1, we allow
2458 2475           * this access if anon_ok is set.  So set creds to the default
2459 2476           * "nobody" id.
2460 2477           */
2461 2478  
2462 2479          if (anon_res != 0) {
2463 2480                  cmn_err(CE_NOTE,
2464 2481                      "nfs_server: client %s%ssent wrong "
2465 2482                      "authentication for %s",
2466 2483                      client_name(req), client_addr(req, buf),
2467 2484                      exi->exi_export.ex_path ?
2468 2485                      exi->exi_export.ex_path : "?");
2469 2486                  return (0);
2470 2487          }
2471 2488  
2472 2489          return (1);
2473 2490  }
2474 2491  
2475 2492  
2476 2493  static char *
2477 2494  client_name(struct svc_req *req)
2478 2495  {
2479 2496          char *hostname = NULL;
2480 2497  
2481 2498          /*
2482 2499           * If it's a Unix cred then use the
2483 2500           * hostname from the credential.
2484 2501           */
2485 2502          if (req->rq_cred.oa_flavor == AUTH_UNIX) {
2486 2503                  hostname = ((struct authunix_parms *)
2487 2504                      req->rq_clntcred)->aup_machname;
2488 2505          }
2489 2506          if (hostname == NULL)
2490 2507                  hostname = "";
2491 2508  
2492 2509          return (hostname);
2493 2510  }
2494 2511  
2495 2512  static char *
2496 2513  client_addr(struct svc_req *req, char *buf)
2497 2514  {
2498 2515          struct sockaddr *ca;
2499 2516          uchar_t *b;
2500 2517          char *frontspace = "";
2501 2518  
2502 2519          /*
2503 2520           * We assume we are called in tandem with client_name and the
2504 2521           * format string looks like "...client %s%sblah blah..."
2505 2522           *
2506 2523           * If it's a Unix cred then client_name returned
2507 2524           * a host name, so we need insert a space between host name
2508 2525           * and IP address.
2509 2526           */
2510 2527          if (req->rq_cred.oa_flavor == AUTH_UNIX)
2511 2528                  frontspace = " ";
2512 2529  
2513 2530          /*
2514 2531           * Convert the caller's IP address to a dotted string
2515 2532           */
2516 2533          ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
2517 2534  
2518 2535          if (ca->sa_family == AF_INET) {
2519 2536                  b = (uchar_t *)&((struct sockaddr_in *)ca)->sin_addr;
2520 2537                  (void) sprintf(buf, "%s(%d.%d.%d.%d) ", frontspace,
2521 2538                      b[0] & 0xFF, b[1] & 0xFF, b[2] & 0xFF, b[3] & 0xFF);
2522 2539          } else if (ca->sa_family == AF_INET6) {
2523 2540                  struct sockaddr_in6 *sin6;
2524 2541                  sin6 = (struct sockaddr_in6 *)ca;
2525 2542                  (void) kinet_ntop6((uchar_t *)&sin6->sin6_addr,
2526 2543                      buf, INET6_ADDRSTRLEN);
2527 2544  
2528 2545          } else {
2529 2546  
2530 2547                  /*
2531 2548                   * No IP address to print. If there was a host name
2532 2549                   * printed, then we print a space.
2533 2550                   */
2534 2551                  (void) sprintf(buf, frontspace);
2535 2552          }
2536 2553  
2537 2554          return (buf);
2538 2555  }
2539 2556  
  
    | 
      ↓ open down ↓ | 
    1974 lines elided | 
    
      ↑ open up ↑ | 
  
2540 2557  /*
2541 2558   * NFS Server initialization routine.  This routine should only be called
2542 2559   * once.  It performs the following tasks:
2543 2560   *      - Call sub-initialization routines (localize access to variables)
2544 2561   *      - Initialize all locks
2545 2562   *      - initialize the version 3 write verifier
2546 2563   */
2547 2564  void
2548 2565  nfs_srvinit(void)
2549 2566  {
2550      -        /* NFS server zone-specific global variables */
2551      -        zone_key_create(&nfssrv_zone_key, nfs_srv_zone_init,
2552      -            NULL, nfs_srv_zone_fini);
2553 2567  
     2568 +        /* Truly global stuff in this module (not per zone) */
     2569 +        rw_init(&nfssrv_globals_rwl, NULL, RW_DEFAULT, NULL);
     2570 +        list_create(&nfssrv_globals_list, sizeof (nfs_globals_t),
     2571 +            offsetof (nfs_globals_t, nfs_g_link));
     2572 +
     2573 +        /* The order here is important */
2554 2574          nfs_exportinit();
2555 2575          rfs_srvrinit();
2556 2576          rfs3_srvrinit();
2557 2577          rfs4_srvrinit();
2558 2578          nfsauth_init();
     2579 +
     2580 +        /*
     2581 +         * NFS server zone-specific global variables
     2582 +         * Note the zone_init is called for the GZ here.
     2583 +         */
     2584 +        zone_key_create(&nfssrv_zone_key, nfs_server_zone_init,
     2585 +            nfs_server_zone_shutdown, nfs_server_zone_fini);
2559 2586  }
2560 2587  
2561 2588  /*
2562 2589   * NFS Server finalization routine. This routine is called to cleanup the
2563 2590   * initialization work previously performed if the NFS server module could
2564 2591   * not be loaded correctly.
2565 2592   */
2566 2593  void
2567 2594  nfs_srvfini(void)
2568 2595  {
     2596 +
     2597 +        /*
     2598 +         * NFS server zone-specific global variables
     2599 +         * Note the zone_fini is called for the GZ here.
     2600 +         */
     2601 +        (void) zone_key_delete(nfssrv_zone_key);
     2602 +
     2603 +        /* The order here is important (reverse of init) */
2569 2604          nfsauth_fini();
2570 2605          rfs4_srvrfini();
2571 2606          rfs3_srvrfini();
2572 2607          rfs_srvrfini();
2573 2608          nfs_exportfini();
2574 2609  
2575      -        (void) zone_key_delete(nfssrv_zone_key);
     2610 +        /* Truly global stuff in this module (not per zone) */
     2611 +        list_destroy(&nfssrv_globals_list);
     2612 +        rw_destroy(&nfssrv_globals_rwl);
2576 2613  }
2577 2614  
2578      -/* ARGSUSED */
     2615 +/*
     2616 + * Zone init, shutdown, fini functions for the NFS server
     2617 + *
     2618 + * This design is careful to create the entire hierarhcy of
     2619 + * NFS server "globals" (including those created by various
     2620 + * per-module *_zone_init functions, etc.) so that all these
     2621 + * objects have exactly the same lifetime.
     2622 + *
     2623 + * These objects are also kept on a list for two reasons:
     2624 + * 1: It makes finding these in  mdb _much_ easier.
     2625 + * 2: It allows operating across all zone globals for
     2626 + *    functions like nfs_auth.c:exi_cache_reclaim
     2627 + */
2579 2628  static void *
2580      -nfs_srv_zone_init(zoneid_t zoneid)
     2629 +nfs_server_zone_init(zoneid_t zoneid)
2581 2630  {
2582 2631          nfs_globals_t *ng;
2583 2632  
2584 2633          ng = kmem_zalloc(sizeof (*ng), KM_SLEEP);
2585 2634  
2586 2635          ng->nfs_versmin = NFS_VERSMIN_DEFAULT;
2587 2636          ng->nfs_versmax = NFS_VERSMAX_DEFAULT;
2588 2637  
2589 2638          /* Init the stuff to control start/stop */
2590 2639          ng->nfs_server_upordown = NFS_SERVER_STOPPED;
2591 2640          mutex_init(&ng->nfs_server_upordown_lock, NULL, MUTEX_DEFAULT, NULL);
2592 2641          cv_init(&ng->nfs_server_upordown_cv, NULL, CV_DEFAULT, NULL);
2593 2642          mutex_init(&ng->rdma_wait_mutex, NULL, MUTEX_DEFAULT, NULL);
2594 2643          cv_init(&ng->rdma_wait_cv, NULL, CV_DEFAULT, NULL);
2595 2644  
     2645 +        ng->nfs_zoneid = zoneid;
     2646 +
     2647 +        /*
     2648 +         * Order here is important.
     2649 +         * export init must precede srv init calls.
     2650 +         */
     2651 +        nfs_export_zone_init(ng);
     2652 +        rfs_srv_zone_init(ng);
     2653 +        rfs3_srv_zone_init(ng);
     2654 +        rfs4_srv_zone_init(ng);
     2655 +        nfsauth_zone_init(ng);
     2656 +
     2657 +        rw_enter(&nfssrv_globals_rwl, RW_WRITER);
     2658 +        list_insert_tail(&nfssrv_globals_list, ng);
     2659 +        rw_exit(&nfssrv_globals_rwl);
     2660 +
2596 2661          return (ng);
2597 2662  }
2598 2663  
2599 2664  /* ARGSUSED */
2600 2665  static void
2601      -nfs_srv_zone_fini(zoneid_t zoneid, void *data)
     2666 +nfs_server_zone_shutdown(zoneid_t zoneid, void *data)
2602 2667  {
2603 2668          nfs_globals_t *ng;
2604 2669  
2605 2670          ng = (nfs_globals_t *)data;
     2671 +
     2672 +        /*
     2673 +         * Order is like _fini, but only
     2674 +         * some modules need this hook.
     2675 +         */
     2676 +        nfsauth_zone_shutdown(ng);
     2677 +        nfs_export_zone_shutdown(ng);
     2678 +}
     2679 +
     2680 +/* ARGSUSED */
     2681 +static void
     2682 +nfs_server_zone_fini(zoneid_t zoneid, void *data)
     2683 +{
     2684 +        nfs_globals_t *ng;
     2685 +
     2686 +        ng = (nfs_globals_t *)data;
     2687 +
     2688 +        rw_enter(&nfssrv_globals_rwl, RW_WRITER);
     2689 +        list_remove(&nfssrv_globals_list, ng);
     2690 +        rw_exit(&nfssrv_globals_rwl);
     2691 +
     2692 +        /*
     2693 +         * Order here is important.
     2694 +         * reverse order from init
     2695 +         */
     2696 +        nfsauth_zone_fini(ng);
     2697 +        rfs4_srv_zone_fini(ng);
     2698 +        rfs3_srv_zone_fini(ng);
     2699 +        rfs_srv_zone_fini(ng);
     2700 +        nfs_export_zone_fini(ng);
     2701 +
2606 2702          mutex_destroy(&ng->nfs_server_upordown_lock);
2607 2703          cv_destroy(&ng->nfs_server_upordown_cv);
2608 2704          mutex_destroy(&ng->rdma_wait_mutex);
2609 2705          cv_destroy(&ng->rdma_wait_cv);
2610 2706  
2611 2707          kmem_free(ng, sizeof (*ng));
2612 2708  }
2613 2709  
2614 2710  /*
2615 2711   * Set up an iovec array of up to cnt pointers.
2616 2712   */
2617 2713  void
2618 2714  mblk_to_iov(mblk_t *m, int cnt, struct iovec *iovp)
2619 2715  {
2620 2716          while (m != NULL && cnt-- > 0) {
2621 2717                  iovp->iov_base = (caddr_t)m->b_rptr;
2622 2718                  iovp->iov_len = (m->b_wptr - m->b_rptr);
2623 2719                  iovp++;
2624 2720                  m = m->b_cont;
2625 2721          }
2626 2722  }
2627 2723  
2628 2724  /*
2629 2725   * Common code between NFS Version 2 and NFS Version 3 for the public
2630 2726   * filehandle multicomponent lookups.
2631 2727   */
2632 2728  
2633 2729  /*
2634 2730   * Public filehandle evaluation of a multi-component lookup, following
2635 2731   * symbolic links, if necessary. This may result in a vnode in another
2636 2732   * filesystem, which is OK as long as the other filesystem is exported.
2637 2733   *
2638 2734   * Note that the exi will be set either to NULL or a new reference to the
2639 2735   * exportinfo struct that corresponds to the vnode of the multi-component path.
2640 2736   * It is the callers responsibility to release this reference.
2641 2737   */
2642 2738  int
2643 2739  rfs_publicfh_mclookup(char *p, vnode_t *dvp, cred_t *cr, vnode_t **vpp,
2644 2740      struct exportinfo **exi, struct sec_ol *sec)
2645 2741  {
2646 2742          int pathflag;
2647 2743          vnode_t *mc_dvp = NULL;
2648 2744          vnode_t *realvp;
2649 2745          int error;
2650 2746  
2651 2747          *exi = NULL;
2652 2748  
2653 2749          /*
2654 2750           * check if the given path is a url or native path. Since p is
2655 2751           * modified by MCLpath(), it may be empty after returning from
2656 2752           * there, and should be checked.
2657 2753           */
2658 2754          if ((pathflag = MCLpath(&p)) == -1)
2659 2755                  return (EIO);
2660 2756  
2661 2757          /*
2662 2758           * If pathflag is SECURITY_QUERY, turn the SEC_QUERY bit
2663 2759           * on in sec->sec_flags. This bit will later serve as an
2664 2760           * indication in makefh_ol() or makefh3_ol() to overload the
2665 2761           * filehandle to contain the sec modes used by the server for
2666 2762           * the path.
2667 2763           */
2668 2764          if (pathflag == SECURITY_QUERY) {
2669 2765                  if ((sec->sec_index = (uint_t)(*p)) > 0) {
2670 2766                          sec->sec_flags |= SEC_QUERY;
2671 2767                          p++;
2672 2768                          if ((pathflag = MCLpath(&p)) == -1)
2673 2769                                  return (EIO);
2674 2770                  } else {
2675 2771                          cmn_err(CE_NOTE,
2676 2772                              "nfs_server: invalid security index %d, "
2677 2773                              "violating WebNFS SNEGO protocol.", sec->sec_index);
2678 2774                          return (EIO);
2679 2775                  }
2680 2776          }
2681 2777  
2682 2778          if (p[0] == '\0') {
2683 2779                  error = ENOENT;
2684 2780                  goto publicfh_done;
2685 2781          }
2686 2782  
2687 2783          error = rfs_pathname(p, &mc_dvp, vpp, dvp, cr, pathflag);
2688 2784  
2689 2785          /*
2690 2786           * If name resolves to "/" we get EINVAL since we asked for
2691 2787           * the vnode of the directory that the file is in. Try again
2692 2788           * with NULL directory vnode.
2693 2789           */
2694 2790          if (error == EINVAL) {
2695 2791                  error = rfs_pathname(p, NULL, vpp, dvp, cr, pathflag);
2696 2792                  if (!error) {
2697 2793                          ASSERT(*vpp != NULL);
2698 2794                          if ((*vpp)->v_type == VDIR) {
2699 2795                                  VN_HOLD(*vpp);
2700 2796                                  mc_dvp = *vpp;
2701 2797                          } else {
2702 2798                                  /*
2703 2799                                   * This should not happen, the filesystem is
2704 2800                                   * in an inconsistent state. Fail the lookup
2705 2801                                   * at this point.
2706 2802                                   */
2707 2803                                  VN_RELE(*vpp);
2708 2804                                  error = EINVAL;
2709 2805                          }
2710 2806                  }
2711 2807          }
2712 2808  
2713 2809          if (error)
2714 2810                  goto publicfh_done;
2715 2811  
2716 2812          if (*vpp == NULL) {
2717 2813                  error = ENOENT;
2718 2814                  goto publicfh_done;
2719 2815          }
2720 2816  
2721 2817          ASSERT(mc_dvp != NULL);
2722 2818          ASSERT(*vpp != NULL);
2723 2819  
2724 2820          if ((*vpp)->v_type == VDIR) {
2725 2821                  do {
2726 2822                          /*
2727 2823                           * *vpp may be an AutoFS node, so we perform
2728 2824                           * a VOP_ACCESS() to trigger the mount of the intended
2729 2825                           * filesystem, so we can perform the lookup in the
2730 2826                           * intended filesystem.
2731 2827                           */
2732 2828                          (void) VOP_ACCESS(*vpp, 0, 0, cr, NULL);
2733 2829  
2734 2830                          /*
2735 2831                           * If vnode is covered, get the
2736 2832                           * the topmost vnode.
2737 2833                           */
2738 2834                          if (vn_mountedvfs(*vpp) != NULL) {
2739 2835                                  error = traverse(vpp);
2740 2836                                  if (error) {
2741 2837                                          VN_RELE(*vpp);
2742 2838                                          goto publicfh_done;
2743 2839                                  }
2744 2840                          }
2745 2841  
2746 2842                          if (VOP_REALVP(*vpp, &realvp, NULL) == 0 &&
2747 2843                              realvp != *vpp) {
2748 2844                                  /*
2749 2845                                   * If realvp is different from *vpp
2750 2846                                   * then release our reference on *vpp, so that
2751 2847                                   * the export access check be performed on the
2752 2848                                   * real filesystem instead.
2753 2849                                   */
2754 2850                                  VN_HOLD(realvp);
2755 2851                                  VN_RELE(*vpp);
2756 2852                                  *vpp = realvp;
2757 2853                          } else {
2758 2854                                  break;
2759 2855                          }
2760 2856                  /* LINTED */
2761 2857                  } while (TRUE);
2762 2858  
2763 2859                  /*
2764 2860                   * Let nfs_vptexi() figure what the real parent is.
2765 2861                   */
2766 2862                  VN_RELE(mc_dvp);
2767 2863                  mc_dvp = NULL;
2768 2864  
2769 2865          } else {
2770 2866                  /*
2771 2867                   * If vnode is covered, get the
2772 2868                   * the topmost vnode.
2773 2869                   */
2774 2870                  if (vn_mountedvfs(mc_dvp) != NULL) {
2775 2871                          error = traverse(&mc_dvp);
2776 2872                          if (error) {
2777 2873                                  VN_RELE(*vpp);
2778 2874                                  goto publicfh_done;
2779 2875                          }
2780 2876                  }
2781 2877  
2782 2878                  if (VOP_REALVP(mc_dvp, &realvp, NULL) == 0 &&
2783 2879                      realvp != mc_dvp) {
2784 2880                          /*
2785 2881                           * *vpp is a file, obtain realvp of the parent
2786 2882                           * directory vnode.
2787 2883                           */
2788 2884                          VN_HOLD(realvp);
2789 2885                          VN_RELE(mc_dvp);
2790 2886                          mc_dvp = realvp;
2791 2887                  }
2792 2888          }
2793 2889  
2794 2890          /*
2795 2891           * The pathname may take us from the public filesystem to another.
2796 2892           * If that's the case then just set the exportinfo to the new export
2797 2893           * and build filehandle for it. Thanks to per-access checking there's
2798 2894           * no security issues with doing this. If the client is not allowed
2799 2895           * access to this new export then it will get an access error when it
2800 2896           * tries to use the filehandle
2801 2897           */
2802 2898          if (error = nfs_check_vpexi(mc_dvp, *vpp, kcred, exi)) {
2803 2899                  VN_RELE(*vpp);
2804 2900                  goto publicfh_done;
2805 2901          }
2806 2902  
2807 2903          /*
2808 2904           * Not allowed access to pseudo exports.
2809 2905           */
2810 2906          if (PSEUDO(*exi)) {
2811 2907                  error = ENOENT;
2812 2908                  VN_RELE(*vpp);
2813 2909                  goto publicfh_done;
2814 2910          }
2815 2911  
2816 2912          /*
2817 2913           * Do a lookup for the index file. We know the index option doesn't
2818 2914           * allow paths through handling in the share command, so mc_dvp will
2819 2915           * be the parent for the index file vnode, if its present. Use
2820 2916           * temporary pointers to preserve and reuse the vnode pointers of the
2821 2917           * original directory in case there's no index file. Note that the
2822 2918           * index file is a native path, and should not be interpreted by
2823 2919           * the URL parser in rfs_pathname()
2824 2920           */
2825 2921          if (((*exi)->exi_export.ex_flags & EX_INDEX) &&
2826 2922              ((*vpp)->v_type == VDIR) && (pathflag == URLPATH)) {
2827 2923                  vnode_t *tvp, *tmc_dvp; /* temporary vnode pointers */
2828 2924  
2829 2925                  tmc_dvp = mc_dvp;
2830 2926                  mc_dvp = tvp = *vpp;
2831 2927  
2832 2928                  error = rfs_pathname((*exi)->exi_export.ex_index, NULL, vpp,
2833 2929                      mc_dvp, cr, NATIVEPATH);
2834 2930  
2835 2931                  if (error == ENOENT) {
2836 2932                          *vpp = tvp;
2837 2933                          mc_dvp = tmc_dvp;
2838 2934                          error = 0;
2839 2935                  } else {        /* ok or error other than ENOENT */
2840 2936                          if (tmc_dvp)
2841 2937                                  VN_RELE(tmc_dvp);
2842 2938                          if (error)
2843 2939                                  goto publicfh_done;
2844 2940  
2845 2941                          /*
2846 2942                           * Found a valid vp for index "filename". Sanity check
2847 2943                           * for odd case where a directory is provided as index
2848 2944                           * option argument and leads us to another filesystem
2849 2945                           */
2850 2946  
2851 2947                          /* Release the reference on the old exi value */
2852 2948                          ASSERT(*exi != NULL);
2853 2949                          exi_rele(*exi);
2854 2950  
2855 2951                          if (error = nfs_check_vpexi(mc_dvp, *vpp, kcred, exi)) {
2856 2952                                  VN_RELE(*vpp);
2857 2953                                  goto publicfh_done;
2858 2954                          }
2859 2955                  }
2860 2956          }
2861 2957  
2862 2958  publicfh_done:
2863 2959          if (mc_dvp)
2864 2960                  VN_RELE(mc_dvp);
2865 2961  
2866 2962          return (error);
2867 2963  }
2868 2964  
2869 2965  /*
2870 2966   * Evaluate a multi-component path
2871 2967   */
2872 2968  int
2873 2969  rfs_pathname(
2874 2970          char *path,                     /* pathname to evaluate */
2875 2971          vnode_t **dirvpp,               /* ret for ptr to parent dir vnode */
2876 2972          vnode_t **compvpp,              /* ret for ptr to component vnode */
2877 2973          vnode_t *startdvp,              /* starting vnode */
2878 2974          cred_t *cr,                     /* user's credential */
2879 2975          int pathflag)                   /* flag to identify path, e.g. URL */
2880 2976  {
2881 2977          char namebuf[TYPICALMAXPATHLEN];
2882 2978          struct pathname pn;
2883 2979          int error;
2884 2980  
2885 2981          /*
2886 2982           * If pathname starts with '/', then set startdvp to root.
2887 2983           */
2888 2984          if (*path == '/') {
2889 2985                  while (*path == '/')
2890 2986                          path++;
2891 2987  
2892 2988                  startdvp = ZONE_ROOTVP();
2893 2989          }
2894 2990  
2895 2991          error = pn_get_buf(path, UIO_SYSSPACE, &pn, namebuf, sizeof (namebuf));
2896 2992          if (error == 0) {
2897 2993                  /*
2898 2994                   * Call the URL parser for URL paths to modify the original
2899 2995                   * string to handle any '%' encoded characters that exist.
2900 2996                   * Done here to avoid an extra bcopy in the lookup.
2901 2997                   * We need to be careful about pathlen's. We know that
2902 2998                   * rfs_pathname() is called with a non-empty path. However,
2903 2999                   * it could be emptied due to the path simply being all /'s,
2904 3000                   * which is valid to proceed with the lookup, or due to the
2905 3001                   * URL parser finding an encoded null character at the
2906 3002                   * beginning of path which should not proceed with the lookup.
2907 3003                   */
2908 3004                  if (pn.pn_pathlen != 0 && pathflag == URLPATH) {
2909 3005                          URLparse(pn.pn_path);
2910 3006                          if ((pn.pn_pathlen = strlen(pn.pn_path)) == 0)
2911 3007                                  return (ENOENT);
2912 3008                  }
2913 3009                  VN_HOLD(startdvp);
2914 3010                  error = lookuppnvp(&pn, NULL, NO_FOLLOW, dirvpp, compvpp,
2915 3011                      ZONE_ROOTVP(), startdvp, cr);
2916 3012          }
2917 3013          if (error == ENAMETOOLONG) {
2918 3014                  /*
2919 3015                   * This thread used a pathname > TYPICALMAXPATHLEN bytes long.
2920 3016                   */
2921 3017                  if (error = pn_get(path, UIO_SYSSPACE, &pn))
2922 3018                          return (error);
2923 3019                  if (pn.pn_pathlen != 0 && pathflag == URLPATH) {
2924 3020                          URLparse(pn.pn_path);
2925 3021                          if ((pn.pn_pathlen = strlen(pn.pn_path)) == 0) {
2926 3022                                  pn_free(&pn);
2927 3023                                  return (ENOENT);
2928 3024                          }
2929 3025                  }
2930 3026                  VN_HOLD(startdvp);
2931 3027                  error = lookuppnvp(&pn, NULL, NO_FOLLOW, dirvpp, compvpp,
2932 3028                      ZONE_ROOTVP(), startdvp, cr);
2933 3029                  pn_free(&pn);
2934 3030          }
2935 3031  
2936 3032          return (error);
2937 3033  }
2938 3034  
2939 3035  /*
2940 3036   * Adapt the multicomponent lookup path depending on the pathtype
2941 3037   */
2942 3038  static int
2943 3039  MCLpath(char **path)
2944 3040  {
2945 3041          unsigned char c = (unsigned char)**path;
2946 3042  
2947 3043          /*
2948 3044           * If the MCL path is between 0x20 and 0x7E (graphic printable
2949 3045           * character of the US-ASCII coded character set), its a URL path,
2950 3046           * per RFC 1738.
2951 3047           */
2952 3048          if (c >= 0x20 && c <= 0x7E)
2953 3049                  return (URLPATH);
2954 3050  
2955 3051          /*
2956 3052           * If the first octet of the MCL path is not an ASCII character
2957 3053           * then it must be interpreted as a tag value that describes the
2958 3054           * format of the remaining octets of the MCL path.
2959 3055           *
2960 3056           * If the first octet of the MCL path is 0x81 it is a query
2961 3057           * for the security info.
2962 3058           */
2963 3059          switch (c) {
2964 3060          case 0x80:      /* native path, i.e. MCL via mount protocol */
2965 3061                  (*path)++;
2966 3062                  return (NATIVEPATH);
2967 3063          case 0x81:      /* security query */
2968 3064                  (*path)++;
2969 3065                  return (SECURITY_QUERY);
2970 3066          default:
2971 3067                  return (-1);
2972 3068          }
2973 3069  }
2974 3070  
2975 3071  #define fromhex(c)  ((c >= '0' && c <= '9') ? (c - '0') : \
2976 3072                          ((c >= 'A' && c <= 'F') ? (c - 'A' + 10) :\
2977 3073                          ((c >= 'a' && c <= 'f') ? (c - 'a' + 10) : 0)))
2978 3074  
2979 3075  /*
2980 3076   * The implementation of URLparse guarantees that the final string will
2981 3077   * fit in the original one. Replaces '%' occurrences followed by 2 characters
2982 3078   * with its corresponding hexadecimal character.
2983 3079   */
2984 3080  static void
2985 3081  URLparse(char *str)
2986 3082  {
2987 3083          char *p, *q;
2988 3084  
2989 3085          p = q = str;
2990 3086          while (*p) {
2991 3087                  *q = *p;
2992 3088                  if (*p++ == '%') {
2993 3089                          if (*p) {
2994 3090                                  *q = fromhex(*p) * 16;
2995 3091                                  p++;
2996 3092                                  if (*p) {
2997 3093                                          *q += fromhex(*p);
2998 3094                                          p++;
2999 3095                                  }
3000 3096                          }
3001 3097                  }
3002 3098                  q++;
3003 3099          }
3004 3100          *q = '\0';
3005 3101  }
3006 3102  
3007 3103  
3008 3104  /*
3009 3105   * Get the export information for the lookup vnode, and verify its
3010 3106   * useable.
3011 3107   */
3012 3108  int
3013 3109  nfs_check_vpexi(vnode_t *mc_dvp, vnode_t *vp, cred_t *cr,
3014 3110      struct exportinfo **exi)
3015 3111  {
3016 3112          int walk;
3017 3113          int error = 0;
3018 3114  
3019 3115          *exi = nfs_vptoexi(mc_dvp, vp, cr, &walk, NULL, FALSE);
3020 3116          if (*exi == NULL)
3021 3117                  error = EACCES;
3022 3118          else {
3023 3119                  /*
3024 3120                   * If nosub is set for this export then
3025 3121                   * a lookup relative to the public fh
3026 3122                   * must not terminate below the
3027 3123                   * exported directory.
3028 3124                   */
3029 3125                  if ((*exi)->exi_export.ex_flags & EX_NOSUB && walk > 0)
3030 3126                          error = EACCES;
3031 3127          }
3032 3128  
3033 3129          return (error);
3034 3130  }
3035 3131  
3036 3132  /*
3037 3133   * Used by NFSv3 and NFSv4 server to query label of
3038 3134   * a pathname component during lookup/access ops.
3039 3135   */
3040 3136  ts_label_t *
3041 3137  nfs_getflabel(vnode_t *vp, struct exportinfo *exi)
3042 3138  {
3043 3139          zone_t *zone;
3044 3140          ts_label_t *zone_label;
3045 3141          char *path;
3046 3142  
3047 3143          mutex_enter(&vp->v_lock);
3048 3144          if (vp->v_path != vn_vpath_empty) {
3049 3145                  zone = zone_find_by_any_path(vp->v_path, B_FALSE);
3050 3146                  mutex_exit(&vp->v_lock);
3051 3147          } else {
3052 3148                  /*
3053 3149                   * v_path not cached. Fall back on pathname of exported
3054 3150                   * file system as we rely on pathname from which we can
3055 3151                   * derive a label. The exported file system portion of
3056 3152                   * path is sufficient to obtain a label.
3057 3153                   */
3058 3154                  path = exi->exi_export.ex_path;
3059 3155                  if (path == NULL) {
3060 3156                          mutex_exit(&vp->v_lock);
3061 3157                          return (NULL);
3062 3158                  }
3063 3159                  zone = zone_find_by_any_path(path, B_FALSE);
3064 3160                  mutex_exit(&vp->v_lock);
3065 3161          }
3066 3162          /*
3067 3163           * Caller has verified that the file is either
3068 3164           * exported or visible. So if the path falls in
3069 3165           * global zone, admin_low is returned; otherwise
3070 3166           * the zone's label is returned.
3071 3167           */
3072 3168          zone_label = zone->zone_slabel;
3073 3169          label_hold(zone_label);
3074 3170          zone_rele(zone);
3075 3171          return (zone_label);
3076 3172  }
3077 3173  
3078 3174  /*
3079 3175   * TX NFS routine used by NFSv3 and NFSv4 to do label check
3080 3176   * on client label and server's file object lable.
3081 3177   */
3082 3178  boolean_t
3083 3179  do_rfs_label_check(bslabel_t *clabel, vnode_t *vp, int flag,
3084 3180      struct exportinfo *exi)
3085 3181  {
3086 3182          bslabel_t *slabel;
3087 3183          ts_label_t *tslabel;
3088 3184          boolean_t result;
3089 3185  
3090 3186          if ((tslabel = nfs_getflabel(vp, exi)) == NULL) {
3091 3187                  return (B_FALSE);
3092 3188          }
3093 3189          slabel = label2bslabel(tslabel);
3094 3190          DTRACE_PROBE4(tx__rfs__log__info__labelcheck, char *,
3095 3191              "comparing server's file label(1) with client label(2) (vp(3))",
3096 3192              bslabel_t *, slabel, bslabel_t *, clabel, vnode_t *, vp);
3097 3193  
3098 3194          if (flag == EQUALITY_CHECK)
3099 3195                  result = blequal(clabel, slabel);
3100 3196          else
3101 3197                  result = bldominates(clabel, slabel);
3102 3198          label_rele(tslabel);
3103 3199          return (result);
3104 3200  }
3105 3201  
3106 3202  /*
3107 3203   * Callback function to return the loaned buffers.
3108 3204   * Calls VOP_RETZCBUF() only after all uio_iov[]
3109 3205   * buffers are returned. nu_ref maintains the count.
3110 3206   */
3111 3207  void
3112 3208  rfs_free_xuio(void *free_arg)
3113 3209  {
3114 3210          uint_t ref;
3115 3211          nfs_xuio_t *nfsuiop = (nfs_xuio_t *)free_arg;
3116 3212  
3117 3213          ref = atomic_dec_uint_nv(&nfsuiop->nu_ref);
3118 3214  
3119 3215          /*
3120 3216           * Call VOP_RETZCBUF() only when all the iov buffers
3121 3217           * are sent OTW.
3122 3218           */
3123 3219          if (ref != 0)
3124 3220                  return;
3125 3221  
3126 3222          if (((uio_t *)nfsuiop)->uio_extflg & UIO_XUIO) {
3127 3223                  (void) VOP_RETZCBUF(nfsuiop->nu_vp, (xuio_t *)free_arg, NULL,
3128 3224                      NULL);
3129 3225                  VN_RELE(nfsuiop->nu_vp);
3130 3226          }
3131 3227  
3132 3228          kmem_cache_free(nfs_xuio_cache, free_arg);
3133 3229  }
3134 3230  
3135 3231  xuio_t *
3136 3232  rfs_setup_xuio(vnode_t *vp)
3137 3233  {
3138 3234          nfs_xuio_t *nfsuiop;
3139 3235  
3140 3236          nfsuiop = kmem_cache_alloc(nfs_xuio_cache, KM_SLEEP);
3141 3237  
3142 3238          bzero(nfsuiop, sizeof (nfs_xuio_t));
3143 3239          nfsuiop->nu_vp = vp;
3144 3240  
3145 3241          /*
3146 3242           * ref count set to 1. more may be added
3147 3243           * if multiple mblks refer to multiple iov's.
3148 3244           * This is done in uio_to_mblk().
3149 3245           */
3150 3246  
3151 3247          nfsuiop->nu_ref = 1;
3152 3248  
3153 3249          nfsuiop->nu_frtn.free_func = rfs_free_xuio;
3154 3250          nfsuiop->nu_frtn.free_arg = (char *)nfsuiop;
3155 3251  
3156 3252          nfsuiop->nu_uio.xu_type = UIOTYPE_ZEROCOPY;
3157 3253  
3158 3254          return (&nfsuiop->nu_uio);
3159 3255  }
3160 3256  
3161 3257  mblk_t *
3162 3258  uio_to_mblk(uio_t *uiop)
3163 3259  {
3164 3260          struct iovec *iovp;
3165 3261          int i;
3166 3262          mblk_t *mp, *mp1;
3167 3263          nfs_xuio_t *nfsuiop = (nfs_xuio_t *)uiop;
3168 3264  
3169 3265          if (uiop->uio_iovcnt == 0)
3170 3266                  return (NULL);
3171 3267  
3172 3268          iovp = uiop->uio_iov;
3173 3269          mp = mp1 = esballoca((uchar_t *)iovp->iov_base, iovp->iov_len,
3174 3270              BPRI_MED, &nfsuiop->nu_frtn);
3175 3271          ASSERT(mp != NULL);
3176 3272  
3177 3273          mp->b_wptr += iovp->iov_len;
3178 3274          mp->b_datap->db_type = M_DATA;
3179 3275  
3180 3276          for (i = 1; i < uiop->uio_iovcnt; i++) {
3181 3277                  iovp = (uiop->uio_iov + i);
3182 3278  
3183 3279                  mp1->b_cont = esballoca(
3184 3280                      (uchar_t *)iovp->iov_base, iovp->iov_len, BPRI_MED,
3185 3281                      &nfsuiop->nu_frtn);
3186 3282  
3187 3283                  mp1 = mp1->b_cont;
3188 3284                  ASSERT(mp1 != NULL);
3189 3285                  mp1->b_wptr += iovp->iov_len;
3190 3286                  mp1->b_datap->db_type = M_DATA;
3191 3287          }
3192 3288  
3193 3289          nfsuiop->nu_ref = uiop->uio_iovcnt;
3194 3290  
3195 3291          return (mp);
3196 3292  }
3197 3293  
3198 3294  /*
3199 3295   * Allocate memory to hold data for a read request of len bytes.
3200 3296   *
3201 3297   * We don't allocate buffers greater than kmem_max_cached in size to avoid
3202 3298   * allocating memory from the kmem_oversized arena.  If we allocate oversized
3203 3299   * buffers, we incur heavy cross-call activity when freeing these large buffers
3204 3300   * in the TCP receive path. Note that we can't set b_wptr here since the
3205 3301   * length of the data returned may differ from the length requested when
3206 3302   * reading the end of a file; we set b_wptr in rfs_rndup_mblks() once the
3207 3303   * length of the read is known.
3208 3304   */
3209 3305  mblk_t *
3210 3306  rfs_read_alloc(uint_t len, struct iovec **iov, int *iovcnt)
3211 3307  {
3212 3308          struct iovec *iovarr;
3213 3309          mblk_t *mp, **mpp = ∓
3214 3310          size_t mpsize;
3215 3311          uint_t remain = len;
3216 3312          int i, err = 0;
3217 3313  
3218 3314          *iovcnt = howmany(len, kmem_max_cached);
3219 3315  
3220 3316          iovarr = kmem_alloc(*iovcnt * sizeof (struct iovec), KM_SLEEP);
3221 3317          *iov = iovarr;
3222 3318  
3223 3319          for (i = 0; i < *iovcnt; remain -= mpsize, i++) {
3224 3320                  ASSERT(remain <= len);
3225 3321                  /*
3226 3322                   * We roundup the size we allocate to a multiple of
3227 3323                   * BYTES_PER_XDR_UNIT (4 bytes) so that the call to
3228 3324                   * xdrmblk_putmblk() never fails.
3229 3325                   */
3230 3326                  ASSERT(kmem_max_cached % BYTES_PER_XDR_UNIT == 0);
3231 3327                  mpsize = MIN(kmem_max_cached, remain);
3232 3328                  *mpp = allocb_wait(RNDUP(mpsize), BPRI_MED, STR_NOSIG, &err);
3233 3329                  ASSERT(*mpp != NULL);
3234 3330                  ASSERT(err == 0);
3235 3331  
3236 3332                  iovarr[i].iov_base = (caddr_t)(*mpp)->b_rptr;
3237 3333                  iovarr[i].iov_len = mpsize;
3238 3334                  mpp = &(*mpp)->b_cont;
3239 3335          }
3240 3336          return (mp);
3241 3337  }
3242 3338  
3243 3339  void
3244 3340  rfs_rndup_mblks(mblk_t *mp, uint_t len, int buf_loaned)
3245 3341  {
3246 3342          int i;
3247 3343          int alloc_err = 0;
3248 3344          mblk_t *rmp;
3249 3345          uint_t mpsize, remainder;
3250 3346  
3251 3347          remainder = P2NPHASE(len, BYTES_PER_XDR_UNIT);
3252 3348  
3253 3349          /*
3254 3350           * Non copy-reduction case.  This function assumes that blocks were
3255 3351           * allocated in multiples of BYTES_PER_XDR_UNIT bytes, which makes this
3256 3352           * padding safe without bounds checking.
3257 3353           */
3258 3354          if (!buf_loaned) {
3259 3355                  /*
3260 3356                   * Set the size of each mblk in the chain until we've consumed
3261 3357                   * the specified length for all but the last one.
3262 3358                   */
3263 3359                  while ((mpsize = MBLKSIZE(mp)) < len) {
3264 3360                          ASSERT(mpsize % BYTES_PER_XDR_UNIT == 0);
3265 3361                          mp->b_wptr += mpsize;
3266 3362                          len -= mpsize;
3267 3363                          mp = mp->b_cont;
3268 3364                          ASSERT(mp != NULL);
3269 3365                  }
3270 3366  
3271 3367                  ASSERT(len + remainder <= mpsize);
3272 3368                  mp->b_wptr += len;
3273 3369                  for (i = 0; i < remainder; i++)
3274 3370                          *mp->b_wptr++ = '\0';
3275 3371                  return;
3276 3372          }
3277 3373  
3278 3374          /*
3279 3375           * No remainder mblk required.
3280 3376           */
3281 3377          if (remainder == 0)
3282 3378                  return;
3283 3379  
3284 3380          /*
3285 3381           * Get to the last mblk in the chain.
3286 3382           */
3287 3383          while (mp->b_cont != NULL)
3288 3384                  mp = mp->b_cont;
3289 3385  
3290 3386          /*
3291 3387           * In case of copy-reduction mblks, the size of the mblks are fixed
3292 3388           * and are of the size of the loaned buffers.  Allocate a remainder
3293 3389           * mblk and chain it to the data buffers. This is sub-optimal, but not
3294 3390           * expected to happen commonly.
3295 3391           */
3296 3392          rmp = allocb_wait(remainder, BPRI_MED, STR_NOSIG, &alloc_err);
3297 3393          ASSERT(rmp != NULL);
3298 3394          ASSERT(alloc_err == 0);
3299 3395  
3300 3396          for (i = 0; i < remainder; i++)
3301 3397                  *rmp->b_wptr++ = '\0';
3302 3398  
3303 3399          rmp->b_datap->db_type = M_DATA;
3304 3400          mp->b_cont = rmp;
3305 3401  }
  
    | 
      ↓ open down ↓ | 
    690 lines elided | 
    
      ↑ open up ↑ | 
  
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX