17c478bd9Sstevel@tonic-gate /* 27c478bd9Sstevel@tonic-gate * CDDL HEADER START 37c478bd9Sstevel@tonic-gate * 47c478bd9Sstevel@tonic-gate * The contents of this file are subject to the terms of the 5f86c6ccaSdm * Common Development and Distribution License (the "License"). 6f86c6ccaSdm * You may not use this file except in compliance with the License. 77c478bd9Sstevel@tonic-gate * 87c478bd9Sstevel@tonic-gate * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 97c478bd9Sstevel@tonic-gate * or http://www.opensolaris.org/os/licensing. 107c478bd9Sstevel@tonic-gate * See the License for the specific language governing permissions 117c478bd9Sstevel@tonic-gate * and limitations under the License. 127c478bd9Sstevel@tonic-gate * 137c478bd9Sstevel@tonic-gate * When distributing Covered Code, include this CDDL HEADER in each 147c478bd9Sstevel@tonic-gate * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 157c478bd9Sstevel@tonic-gate * If applicable, add the following below this CDDL HEADER, with the 167c478bd9Sstevel@tonic-gate * fields enclosed by brackets "[]" replaced with your own identifying 177c478bd9Sstevel@tonic-gate * information: Portions Copyright [yyyy] [name of copyright owner] 187c478bd9Sstevel@tonic-gate * 197c478bd9Sstevel@tonic-gate * CDDL HEADER END 207c478bd9Sstevel@tonic-gate */ 217c478bd9Sstevel@tonic-gate /* 22*d7de0ceaSRobert Harris * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved. 237c478bd9Sstevel@tonic-gate */ 247c478bd9Sstevel@tonic-gate 257c478bd9Sstevel@tonic-gate /* 267c478bd9Sstevel@tonic-gate * Copyright (c) 1983,1984,1985,1986,1987,1988,1989 AT&T. 277c478bd9Sstevel@tonic-gate * All Rights Reserved 287c478bd9Sstevel@tonic-gate */ 297c478bd9Sstevel@tonic-gate 307c478bd9Sstevel@tonic-gate #include <sys/param.h> 317c478bd9Sstevel@tonic-gate #include <sys/types.h> 327c478bd9Sstevel@tonic-gate #include <sys/systm.h> 337c478bd9Sstevel@tonic-gate #include <sys/cred.h> 347c478bd9Sstevel@tonic-gate #include <sys/vfs.h> 35aa59c4cbSrsb #include <sys/vfs_opreg.h> 367c478bd9Sstevel@tonic-gate #include <sys/vnode.h> 377c478bd9Sstevel@tonic-gate #include <sys/pathname.h> 387c478bd9Sstevel@tonic-gate #include <sys/sysmacros.h> 397c478bd9Sstevel@tonic-gate #include <sys/kmem.h> 407c478bd9Sstevel@tonic-gate #include <sys/mkdev.h> 417c478bd9Sstevel@tonic-gate #include <sys/mount.h> 427c478bd9Sstevel@tonic-gate #include <sys/statvfs.h> 437c478bd9Sstevel@tonic-gate #include <sys/errno.h> 447c478bd9Sstevel@tonic-gate #include <sys/debug.h> 457c478bd9Sstevel@tonic-gate #include <sys/cmn_err.h> 467c478bd9Sstevel@tonic-gate #include <sys/utsname.h> 477c478bd9Sstevel@tonic-gate #include <sys/bootconf.h> 487c478bd9Sstevel@tonic-gate #include <sys/modctl.h> 497c478bd9Sstevel@tonic-gate #include <sys/acl.h> 507c478bd9Sstevel@tonic-gate #include <sys/flock.h> 517c478bd9Sstevel@tonic-gate #include <sys/time.h> 527c478bd9Sstevel@tonic-gate #include <sys/disp.h> 537c478bd9Sstevel@tonic-gate #include <sys/policy.h> 547c478bd9Sstevel@tonic-gate #include <sys/socket.h> 557c478bd9Sstevel@tonic-gate #include <sys/netconfig.h> 567c478bd9Sstevel@tonic-gate #include <sys/dnlc.h> 577c478bd9Sstevel@tonic-gate #include <sys/list.h> 5845916cd2Sjpk #include <sys/mntent.h> 5945916cd2Sjpk #include <sys/tsol/label.h> 607c478bd9Sstevel@tonic-gate 617c478bd9Sstevel@tonic-gate #include <rpc/types.h> 627c478bd9Sstevel@tonic-gate #include <rpc/auth.h> 637c478bd9Sstevel@tonic-gate #include <rpc/rpcsec_gss.h> 647c478bd9Sstevel@tonic-gate #include <rpc/clnt.h> 657c478bd9Sstevel@tonic-gate 667c478bd9Sstevel@tonic-gate #include <nfs/nfs.h> 677c478bd9Sstevel@tonic-gate #include <nfs/nfs_clnt.h> 687c478bd9Sstevel@tonic-gate #include <nfs/mount.h> 697c478bd9Sstevel@tonic-gate #include <nfs/nfs_acl.h> 707c478bd9Sstevel@tonic-gate 717c478bd9Sstevel@tonic-gate #include <fs/fs_subr.h> 727c478bd9Sstevel@tonic-gate 737c478bd9Sstevel@tonic-gate #include <nfs/nfs4.h> 747c478bd9Sstevel@tonic-gate #include <nfs/rnode4.h> 757c478bd9Sstevel@tonic-gate #include <nfs/nfs4_clnt.h> 7639d3e169Sevanl #include <sys/fs/autofs.h> 7739d3e169Sevanl 782f172c55SRobert Thurlow #include <sys/sdt.h> 792f172c55SRobert Thurlow 807c478bd9Sstevel@tonic-gate 817c478bd9Sstevel@tonic-gate /* 827c478bd9Sstevel@tonic-gate * Arguments passed to thread to free data structures from forced unmount. 837c478bd9Sstevel@tonic-gate */ 847c478bd9Sstevel@tonic-gate 857c478bd9Sstevel@tonic-gate typedef struct { 86b9238976Sth vfs_t *fm_vfsp; 87b9238976Sth int fm_flag; 88b9238976Sth cred_t *fm_cr; 897c478bd9Sstevel@tonic-gate } freemountargs_t; 907c478bd9Sstevel@tonic-gate 91b9238976Sth static void async_free_mount(vfs_t *, int, cred_t *); 92b9238976Sth static void nfs4_free_mount(vfs_t *, int, cred_t *); 937c478bd9Sstevel@tonic-gate static void nfs4_free_mount_thread(freemountargs_t *); 947c478bd9Sstevel@tonic-gate static int nfs4_chkdup_servinfo4(servinfo4_t *, servinfo4_t *); 957c478bd9Sstevel@tonic-gate 967c478bd9Sstevel@tonic-gate /* 977c478bd9Sstevel@tonic-gate * From rpcsec module (common/rpcsec). 987c478bd9Sstevel@tonic-gate */ 997c478bd9Sstevel@tonic-gate extern int sec_clnt_loadinfo(struct sec_data *, struct sec_data **, model_t); 1007c478bd9Sstevel@tonic-gate extern void sec_clnt_freeinfo(struct sec_data *); 1017c478bd9Sstevel@tonic-gate 1027c478bd9Sstevel@tonic-gate /* 1037c478bd9Sstevel@tonic-gate * The order and contents of this structure must be kept in sync with that of 1047c478bd9Sstevel@tonic-gate * rfsreqcnt_v4_tmpl in nfs_stats.c 1057c478bd9Sstevel@tonic-gate */ 1067c478bd9Sstevel@tonic-gate static char *rfsnames_v4[] = { 1077c478bd9Sstevel@tonic-gate "null", "compound", "reserved", "access", "close", "commit", "create", 1087c478bd9Sstevel@tonic-gate "delegpurge", "delegreturn", "getattr", "getfh", "link", "lock", 1097c478bd9Sstevel@tonic-gate "lockt", "locku", "lookup", "lookupp", "nverify", "open", "openattr", 1107c478bd9Sstevel@tonic-gate "open_confirm", "open_downgrade", "putfh", "putpubfh", "putrootfh", 1117c478bd9Sstevel@tonic-gate "read", "readdir", "readlink", "remove", "rename", "renew", 1127c478bd9Sstevel@tonic-gate "restorefh", "savefh", "secinfo", "setattr", "setclientid", 1137c478bd9Sstevel@tonic-gate "setclientid_confirm", "verify", "write" 1147c478bd9Sstevel@tonic-gate }; 1157c478bd9Sstevel@tonic-gate 1167c478bd9Sstevel@tonic-gate /* 1177c478bd9Sstevel@tonic-gate * nfs4_max_mount_retry is the number of times the client will redrive 1187c478bd9Sstevel@tonic-gate * a mount compound before giving up and returning failure. The intent 1197c478bd9Sstevel@tonic-gate * is to redrive mount compounds which fail NFS4ERR_STALE so that 1207c478bd9Sstevel@tonic-gate * if a component of the server path being mounted goes stale, it can 1217c478bd9Sstevel@tonic-gate * "recover" by redriving the mount compund (LOOKUP ops). This recovery 1227c478bd9Sstevel@tonic-gate * code is needed outside of the recovery framework because mount is a 1237c478bd9Sstevel@tonic-gate * special case. The client doesn't create vnodes/rnodes for components 1247c478bd9Sstevel@tonic-gate * of the server path being mounted. The recovery code recovers real 1257c478bd9Sstevel@tonic-gate * client objects, not STALE FHs which map to components of the server 1267c478bd9Sstevel@tonic-gate * path being mounted. 1277c478bd9Sstevel@tonic-gate * 1287c478bd9Sstevel@tonic-gate * We could just fail the mount on the first time, but that would 1297c478bd9Sstevel@tonic-gate * instantly trigger failover (from nfs4_mount), and the client should 1307c478bd9Sstevel@tonic-gate * try to re-lookup the STALE FH before doing failover. The easiest 1317c478bd9Sstevel@tonic-gate * way to "re-lookup" is to simply redrive the mount compound. 1327c478bd9Sstevel@tonic-gate */ 1337c478bd9Sstevel@tonic-gate static int nfs4_max_mount_retry = 2; 1347c478bd9Sstevel@tonic-gate 1357c478bd9Sstevel@tonic-gate /* 1367c478bd9Sstevel@tonic-gate * nfs4 vfs operations. 1377c478bd9Sstevel@tonic-gate */ 138b9238976Sth int nfs4_mount(vfs_t *, vnode_t *, struct mounta *, cred_t *); 1397c478bd9Sstevel@tonic-gate static int nfs4_unmount(vfs_t *, int, cred_t *); 1407c478bd9Sstevel@tonic-gate static int nfs4_root(vfs_t *, vnode_t **); 1417c478bd9Sstevel@tonic-gate static int nfs4_statvfs(vfs_t *, struct statvfs64 *); 1427c478bd9Sstevel@tonic-gate static int nfs4_sync(vfs_t *, short, cred_t *); 1437c478bd9Sstevel@tonic-gate static int nfs4_vget(vfs_t *, vnode_t **, fid_t *); 1447c478bd9Sstevel@tonic-gate static int nfs4_mountroot(vfs_t *, whymountroot_t); 1457c478bd9Sstevel@tonic-gate static void nfs4_freevfs(vfs_t *); 1467c478bd9Sstevel@tonic-gate 1477c478bd9Sstevel@tonic-gate static int nfs4rootvp(vnode_t **, vfs_t *, struct servinfo4 *, 1487c478bd9Sstevel@tonic-gate int, cred_t *, zone_t *); 1497c478bd9Sstevel@tonic-gate 1507c478bd9Sstevel@tonic-gate vfsops_t *nfs4_vfsops; 1517c478bd9Sstevel@tonic-gate 1527c478bd9Sstevel@tonic-gate int nfs4_vfsinit(void); 1537c478bd9Sstevel@tonic-gate void nfs4_vfsfini(void); 1547c478bd9Sstevel@tonic-gate static void nfs4setclientid_init(void); 1557c478bd9Sstevel@tonic-gate static void nfs4setclientid_fini(void); 1567c478bd9Sstevel@tonic-gate static void nfs4setclientid_otw(mntinfo4_t *, servinfo4_t *, cred_t *, 1577c478bd9Sstevel@tonic-gate struct nfs4_server *, nfs4_error_t *, int *); 1587c478bd9Sstevel@tonic-gate static void destroy_nfs4_server(nfs4_server_t *); 1597c478bd9Sstevel@tonic-gate static void remove_mi(nfs4_server_t *, mntinfo4_t *); 1607c478bd9Sstevel@tonic-gate 161b9238976Sth extern void nfs4_ephemeral_init(void); 162b9238976Sth extern void nfs4_ephemeral_fini(void); 163b9238976Sth 1642f172c55SRobert Thurlow /* referral related routines */ 1652f172c55SRobert Thurlow static servinfo4_t *copy_svp(servinfo4_t *); 1662f172c55SRobert Thurlow static void free_knconf_contents(struct knetconfig *k); 1672f172c55SRobert Thurlow static char *extract_referral_point(const char *, int); 1682f172c55SRobert Thurlow static void setup_newsvpath(servinfo4_t *, int); 1692f172c55SRobert Thurlow static void update_servinfo4(servinfo4_t *, fs_location4 *, 1702f172c55SRobert Thurlow struct nfs_fsl_info *, char *, int); 1712f172c55SRobert Thurlow 1727c478bd9Sstevel@tonic-gate /* 1737c478bd9Sstevel@tonic-gate * Initialize the vfs structure 1747c478bd9Sstevel@tonic-gate */ 1757c478bd9Sstevel@tonic-gate 1767c478bd9Sstevel@tonic-gate static int nfs4fstyp; 1777c478bd9Sstevel@tonic-gate 1787c478bd9Sstevel@tonic-gate 1797c478bd9Sstevel@tonic-gate /* 1807c478bd9Sstevel@tonic-gate * Debug variable to check for rdma based 1817c478bd9Sstevel@tonic-gate * transport startup and cleanup. Controlled 1827c478bd9Sstevel@tonic-gate * through /etc/system. Off by default. 1837c478bd9Sstevel@tonic-gate */ 1847c478bd9Sstevel@tonic-gate extern int rdma_debug; 1857c478bd9Sstevel@tonic-gate 1867c478bd9Sstevel@tonic-gate int 1877c478bd9Sstevel@tonic-gate nfs4init(int fstyp, char *name) 1887c478bd9Sstevel@tonic-gate { 1897c478bd9Sstevel@tonic-gate static const fs_operation_def_t nfs4_vfsops_template[] = { 190aa59c4cbSrsb VFSNAME_MOUNT, { .vfs_mount = nfs4_mount }, 191aa59c4cbSrsb VFSNAME_UNMOUNT, { .vfs_unmount = nfs4_unmount }, 192aa59c4cbSrsb VFSNAME_ROOT, { .vfs_root = nfs4_root }, 193aa59c4cbSrsb VFSNAME_STATVFS, { .vfs_statvfs = nfs4_statvfs }, 194aa59c4cbSrsb VFSNAME_SYNC, { .vfs_sync = nfs4_sync }, 195aa59c4cbSrsb VFSNAME_VGET, { .vfs_vget = nfs4_vget }, 196aa59c4cbSrsb VFSNAME_MOUNTROOT, { .vfs_mountroot = nfs4_mountroot }, 197aa59c4cbSrsb VFSNAME_FREEVFS, { .vfs_freevfs = nfs4_freevfs }, 198aa59c4cbSrsb NULL, NULL 1997c478bd9Sstevel@tonic-gate }; 2007c478bd9Sstevel@tonic-gate int error; 2017c478bd9Sstevel@tonic-gate 202b9238976Sth nfs4_vfsops = NULL; 203b9238976Sth nfs4_vnodeops = NULL; 204b9238976Sth nfs4_trigger_vnodeops = NULL; 205b9238976Sth 2067c478bd9Sstevel@tonic-gate error = vfs_setfsops(fstyp, nfs4_vfsops_template, &nfs4_vfsops); 2077c478bd9Sstevel@tonic-gate if (error != 0) { 2087c478bd9Sstevel@tonic-gate zcmn_err(GLOBAL_ZONEID, CE_WARN, 2097c478bd9Sstevel@tonic-gate "nfs4init: bad vfs ops template"); 210b9238976Sth goto out; 2117c478bd9Sstevel@tonic-gate } 2127c478bd9Sstevel@tonic-gate 2137c478bd9Sstevel@tonic-gate error = vn_make_ops(name, nfs4_vnodeops_template, &nfs4_vnodeops); 2147c478bd9Sstevel@tonic-gate if (error != 0) { 2157c478bd9Sstevel@tonic-gate zcmn_err(GLOBAL_ZONEID, CE_WARN, 2167c478bd9Sstevel@tonic-gate "nfs4init: bad vnode ops template"); 217b9238976Sth goto out; 2187c478bd9Sstevel@tonic-gate } 2197c478bd9Sstevel@tonic-gate 220b9238976Sth error = vn_make_ops("nfs4_trigger", nfs4_trigger_vnodeops_template, 221b9238976Sth &nfs4_trigger_vnodeops); 222b9238976Sth if (error != 0) { 223b9238976Sth zcmn_err(GLOBAL_ZONEID, CE_WARN, 224b9238976Sth "nfs4init: bad trigger vnode ops template"); 225b9238976Sth goto out; 226b9238976Sth } 2277c478bd9Sstevel@tonic-gate 228b9238976Sth nfs4fstyp = fstyp; 2297c478bd9Sstevel@tonic-gate (void) nfs4_vfsinit(); 2307c478bd9Sstevel@tonic-gate (void) nfs4_init_dot_entries(); 2317c478bd9Sstevel@tonic-gate 232b9238976Sth out: 233b9238976Sth if (error) { 234b9238976Sth if (nfs4_trigger_vnodeops != NULL) 235b9238976Sth vn_freevnodeops(nfs4_trigger_vnodeops); 236b9238976Sth 237b9238976Sth if (nfs4_vnodeops != NULL) 238b9238976Sth vn_freevnodeops(nfs4_vnodeops); 239b9238976Sth 240b9238976Sth (void) vfs_freevfsops_by_type(fstyp); 241b9238976Sth } 242b9238976Sth 243b9238976Sth return (error); 2447c478bd9Sstevel@tonic-gate } 2457c478bd9Sstevel@tonic-gate 2467c478bd9Sstevel@tonic-gate void 2477c478bd9Sstevel@tonic-gate nfs4fini(void) 2487c478bd9Sstevel@tonic-gate { 2497c478bd9Sstevel@tonic-gate (void) nfs4_destroy_dot_entries(); 2507c478bd9Sstevel@tonic-gate nfs4_vfsfini(); 2517c478bd9Sstevel@tonic-gate } 2527c478bd9Sstevel@tonic-gate 2537c478bd9Sstevel@tonic-gate /* 2547c478bd9Sstevel@tonic-gate * Create a new sec_data structure to store AUTH_DH related data: 2557c478bd9Sstevel@tonic-gate * netname, syncaddr, knetconfig. There is no AUTH_F_RPCTIMESYNC 2567c478bd9Sstevel@tonic-gate * flag set for NFS V4 since we are avoiding to contact the rpcbind 2577c478bd9Sstevel@tonic-gate * daemon and is using the IP time service (IPPORT_TIMESERVER). 2587c478bd9Sstevel@tonic-gate * 2597c478bd9Sstevel@tonic-gate * sec_data can be freed by sec_clnt_freeinfo(). 2607c478bd9Sstevel@tonic-gate */ 261b9238976Sth static struct sec_data * 2627c478bd9Sstevel@tonic-gate create_authdh_data(char *netname, int nlen, struct netbuf *syncaddr, 2637c478bd9Sstevel@tonic-gate struct knetconfig *knconf) { 2647c478bd9Sstevel@tonic-gate struct sec_data *secdata; 2657c478bd9Sstevel@tonic-gate dh_k4_clntdata_t *data; 2667c478bd9Sstevel@tonic-gate char *pf, *p; 2677c478bd9Sstevel@tonic-gate 2687c478bd9Sstevel@tonic-gate if (syncaddr == NULL || syncaddr->buf == NULL || nlen == 0) 2697c478bd9Sstevel@tonic-gate return (NULL); 2707c478bd9Sstevel@tonic-gate 2717c478bd9Sstevel@tonic-gate secdata = kmem_alloc(sizeof (*secdata), KM_SLEEP); 2727c478bd9Sstevel@tonic-gate secdata->flags = 0; 2737c478bd9Sstevel@tonic-gate 2747c478bd9Sstevel@tonic-gate data = kmem_alloc(sizeof (*data), KM_SLEEP); 2757c478bd9Sstevel@tonic-gate 2767c478bd9Sstevel@tonic-gate data->syncaddr.maxlen = syncaddr->maxlen; 2777c478bd9Sstevel@tonic-gate data->syncaddr.len = syncaddr->len; 2787c478bd9Sstevel@tonic-gate data->syncaddr.buf = (char *)kmem_alloc(syncaddr->len, KM_SLEEP); 2797c478bd9Sstevel@tonic-gate bcopy(syncaddr->buf, data->syncaddr.buf, syncaddr->len); 2807c478bd9Sstevel@tonic-gate 2817c478bd9Sstevel@tonic-gate /* 2827c478bd9Sstevel@tonic-gate * duplicate the knconf information for the 2837c478bd9Sstevel@tonic-gate * new opaque data. 2847c478bd9Sstevel@tonic-gate */ 2857c478bd9Sstevel@tonic-gate data->knconf = kmem_alloc(sizeof (*knconf), KM_SLEEP); 2867c478bd9Sstevel@tonic-gate *data->knconf = *knconf; 2877c478bd9Sstevel@tonic-gate pf = kmem_alloc(KNC_STRSIZE, KM_SLEEP); 2887c478bd9Sstevel@tonic-gate p = kmem_alloc(KNC_STRSIZE, KM_SLEEP); 2897c478bd9Sstevel@tonic-gate bcopy(knconf->knc_protofmly, pf, KNC_STRSIZE); 2907c478bd9Sstevel@tonic-gate bcopy(knconf->knc_proto, p, KNC_STRSIZE); 2917c478bd9Sstevel@tonic-gate data->knconf->knc_protofmly = pf; 2927c478bd9Sstevel@tonic-gate data->knconf->knc_proto = p; 2937c478bd9Sstevel@tonic-gate 2947c478bd9Sstevel@tonic-gate /* move server netname to the sec_data structure */ 2957c478bd9Sstevel@tonic-gate data->netname = kmem_alloc(nlen, KM_SLEEP); 2967c478bd9Sstevel@tonic-gate bcopy(netname, data->netname, nlen); 2977c478bd9Sstevel@tonic-gate data->netnamelen = (int)nlen; 2987c478bd9Sstevel@tonic-gate 2997c478bd9Sstevel@tonic-gate secdata->secmod = AUTH_DH; 3007c478bd9Sstevel@tonic-gate secdata->rpcflavor = AUTH_DH; 3017c478bd9Sstevel@tonic-gate secdata->data = (caddr_t)data; 3027c478bd9Sstevel@tonic-gate 3037c478bd9Sstevel@tonic-gate return (secdata); 3047c478bd9Sstevel@tonic-gate } 3057c478bd9Sstevel@tonic-gate 306b9238976Sth /* 307b9238976Sth * Returns (deep) copy of sec_data_t. Allocates all memory required; caller 308b9238976Sth * is responsible for freeing. 309b9238976Sth */ 310b9238976Sth sec_data_t * 311b9238976Sth copy_sec_data(sec_data_t *fsecdata) { 312b9238976Sth sec_data_t *tsecdata; 313b9238976Sth 314b9238976Sth if (fsecdata == NULL) 315b9238976Sth return (NULL); 316b9238976Sth 317b9238976Sth if (fsecdata->rpcflavor == AUTH_DH) { 318b9238976Sth dh_k4_clntdata_t *fdata = (dh_k4_clntdata_t *)fsecdata->data; 319b9238976Sth 320b9238976Sth if (fdata == NULL) 321b9238976Sth return (NULL); 322b9238976Sth 323b9238976Sth tsecdata = (sec_data_t *)create_authdh_data(fdata->netname, 324b9238976Sth fdata->netnamelen, &fdata->syncaddr, fdata->knconf); 325b9238976Sth 326b9238976Sth return (tsecdata); 327b9238976Sth } 328b9238976Sth 329b9238976Sth tsecdata = kmem_zalloc(sizeof (sec_data_t), KM_SLEEP); 330b9238976Sth 331b9238976Sth tsecdata->secmod = fsecdata->secmod; 332b9238976Sth tsecdata->rpcflavor = fsecdata->rpcflavor; 333b9238976Sth tsecdata->flags = fsecdata->flags; 334b9238976Sth tsecdata->uid = fsecdata->uid; 335b9238976Sth 336b9238976Sth if (fsecdata->rpcflavor == RPCSEC_GSS) { 337b9238976Sth gss_clntdata_t *gcd = (gss_clntdata_t *)fsecdata->data; 338b9238976Sth 339b9238976Sth tsecdata->data = (caddr_t)copy_sec_data_gss(gcd); 340b9238976Sth } else { 341b9238976Sth tsecdata->data = NULL; 342b9238976Sth } 343b9238976Sth 344b9238976Sth return (tsecdata); 345b9238976Sth } 346b9238976Sth 347b9238976Sth gss_clntdata_t * 348b9238976Sth copy_sec_data_gss(gss_clntdata_t *fdata) 349b9238976Sth { 350b9238976Sth gss_clntdata_t *tdata; 351b9238976Sth 352b9238976Sth if (fdata == NULL) 353b9238976Sth return (NULL); 354b9238976Sth 355b9238976Sth tdata = kmem_zalloc(sizeof (gss_clntdata_t), KM_SLEEP); 356b9238976Sth 357b9238976Sth tdata->mechanism.length = fdata->mechanism.length; 358b9238976Sth tdata->mechanism.elements = kmem_zalloc(fdata->mechanism.length, 359b9238976Sth KM_SLEEP); 360b9238976Sth bcopy(fdata->mechanism.elements, tdata->mechanism.elements, 361b9238976Sth fdata->mechanism.length); 362b9238976Sth 363b9238976Sth tdata->service = fdata->service; 364b9238976Sth 365b9238976Sth (void) strcpy(tdata->uname, fdata->uname); 366b9238976Sth (void) strcpy(tdata->inst, fdata->inst); 367b9238976Sth (void) strcpy(tdata->realm, fdata->realm); 368b9238976Sth 369b9238976Sth tdata->qop = fdata->qop; 370b9238976Sth 371b9238976Sth return (tdata); 372b9238976Sth } 373b9238976Sth 3747c478bd9Sstevel@tonic-gate static int 3757c478bd9Sstevel@tonic-gate nfs4_chkdup_servinfo4(servinfo4_t *svp_head, servinfo4_t *svp) 3767c478bd9Sstevel@tonic-gate { 3777c478bd9Sstevel@tonic-gate servinfo4_t *si; 3787c478bd9Sstevel@tonic-gate 3797c478bd9Sstevel@tonic-gate /* 3807c478bd9Sstevel@tonic-gate * Iterate over the servinfo4 list to make sure 3817c478bd9Sstevel@tonic-gate * we do not have a duplicate. Skip any servinfo4 3827c478bd9Sstevel@tonic-gate * that has been marked "NOT IN USE" 3837c478bd9Sstevel@tonic-gate */ 3847c478bd9Sstevel@tonic-gate for (si = svp_head; si; si = si->sv_next) { 3857c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&si->sv_lock, RW_READER, 0); 3867c478bd9Sstevel@tonic-gate if (si->sv_flags & SV4_NOTINUSE) { 3877c478bd9Sstevel@tonic-gate nfs_rw_exit(&si->sv_lock); 3887c478bd9Sstevel@tonic-gate continue; 3897c478bd9Sstevel@tonic-gate } 3907c478bd9Sstevel@tonic-gate nfs_rw_exit(&si->sv_lock); 3917c478bd9Sstevel@tonic-gate if (si == svp) 3927c478bd9Sstevel@tonic-gate continue; 3937c478bd9Sstevel@tonic-gate if (si->sv_addr.len == svp->sv_addr.len && 3947c478bd9Sstevel@tonic-gate strcmp(si->sv_knconf->knc_protofmly, 395b9238976Sth svp->sv_knconf->knc_protofmly) == 0 && 3967c478bd9Sstevel@tonic-gate bcmp(si->sv_addr.buf, svp->sv_addr.buf, 397b9238976Sth si->sv_addr.len) == 0) { 3987c478bd9Sstevel@tonic-gate /* it's a duplicate */ 3997c478bd9Sstevel@tonic-gate return (1); 4007c478bd9Sstevel@tonic-gate } 4017c478bd9Sstevel@tonic-gate } 4027c478bd9Sstevel@tonic-gate /* it's not a duplicate */ 4037c478bd9Sstevel@tonic-gate return (0); 4047c478bd9Sstevel@tonic-gate } 4057c478bd9Sstevel@tonic-gate 40639d3e169Sevanl void 40739d3e169Sevanl nfs4_free_args(struct nfs_args *nargs) 40839d3e169Sevanl { 40939d3e169Sevanl if (nargs->knconf) { 41039d3e169Sevanl if (nargs->knconf->knc_protofmly) 41139d3e169Sevanl kmem_free(nargs->knconf->knc_protofmly, 412b9238976Sth KNC_STRSIZE); 41339d3e169Sevanl if (nargs->knconf->knc_proto) 41439d3e169Sevanl kmem_free(nargs->knconf->knc_proto, KNC_STRSIZE); 41539d3e169Sevanl kmem_free(nargs->knconf, sizeof (*nargs->knconf)); 41639d3e169Sevanl nargs->knconf = NULL; 41739d3e169Sevanl } 41839d3e169Sevanl 41939d3e169Sevanl if (nargs->fh) { 42039d3e169Sevanl kmem_free(nargs->fh, strlen(nargs->fh) + 1); 42139d3e169Sevanl nargs->fh = NULL; 42239d3e169Sevanl } 42339d3e169Sevanl 42439d3e169Sevanl if (nargs->hostname) { 42539d3e169Sevanl kmem_free(nargs->hostname, strlen(nargs->hostname) + 1); 42639d3e169Sevanl nargs->hostname = NULL; 42739d3e169Sevanl } 42839d3e169Sevanl 42939d3e169Sevanl if (nargs->addr) { 43039d3e169Sevanl if (nargs->addr->buf) { 43139d3e169Sevanl ASSERT(nargs->addr->len); 43239d3e169Sevanl kmem_free(nargs->addr->buf, nargs->addr->len); 43339d3e169Sevanl } 43439d3e169Sevanl kmem_free(nargs->addr, sizeof (struct netbuf)); 43539d3e169Sevanl nargs->addr = NULL; 43639d3e169Sevanl } 43739d3e169Sevanl 43839d3e169Sevanl if (nargs->syncaddr) { 43939d3e169Sevanl ASSERT(nargs->syncaddr->len); 44039d3e169Sevanl if (nargs->syncaddr->buf) { 44139d3e169Sevanl ASSERT(nargs->syncaddr->len); 44239d3e169Sevanl kmem_free(nargs->syncaddr->buf, nargs->syncaddr->len); 44339d3e169Sevanl } 44439d3e169Sevanl kmem_free(nargs->syncaddr, sizeof (struct netbuf)); 44539d3e169Sevanl nargs->syncaddr = NULL; 44639d3e169Sevanl } 44739d3e169Sevanl 44839d3e169Sevanl if (nargs->netname) { 44939d3e169Sevanl kmem_free(nargs->netname, strlen(nargs->netname) + 1); 45039d3e169Sevanl nargs->netname = NULL; 45139d3e169Sevanl } 45239d3e169Sevanl 45339d3e169Sevanl if (nargs->nfs_ext_u.nfs_extA.secdata) { 45439d3e169Sevanl sec_clnt_freeinfo( 455b9238976Sth nargs->nfs_ext_u.nfs_extA.secdata); 45639d3e169Sevanl nargs->nfs_ext_u.nfs_extA.secdata = NULL; 45739d3e169Sevanl } 45839d3e169Sevanl } 45939d3e169Sevanl 46039d3e169Sevanl 46139d3e169Sevanl int 46239d3e169Sevanl nfs4_copyin(char *data, int datalen, struct nfs_args *nargs) 46339d3e169Sevanl { 46439d3e169Sevanl 46539d3e169Sevanl int error; 46639d3e169Sevanl size_t hlen; /* length of hostname */ 46739d3e169Sevanl size_t nlen; /* length of netname */ 46839d3e169Sevanl char netname[MAXNETNAMELEN+1]; /* server's netname */ 46939d3e169Sevanl struct netbuf addr; /* server's address */ 47039d3e169Sevanl struct netbuf syncaddr; /* AUTH_DES time sync addr */ 47139d3e169Sevanl struct knetconfig *knconf; /* transport structure */ 47239d3e169Sevanl struct sec_data *secdata = NULL; /* security data */ 47339d3e169Sevanl STRUCT_DECL(nfs_args, args); /* nfs mount arguments */ 47439d3e169Sevanl STRUCT_DECL(knetconfig, knconf_tmp); 47539d3e169Sevanl STRUCT_DECL(netbuf, addr_tmp); 47639d3e169Sevanl int flags; 47739d3e169Sevanl char *p, *pf; 47839d3e169Sevanl struct pathname pn; 47939d3e169Sevanl char *userbufptr; 48039d3e169Sevanl 48139d3e169Sevanl 48239d3e169Sevanl bzero(nargs, sizeof (*nargs)); 48339d3e169Sevanl 48439d3e169Sevanl STRUCT_INIT(args, get_udatamodel()); 48539d3e169Sevanl bzero(STRUCT_BUF(args), SIZEOF_STRUCT(nfs_args, DATAMODEL_NATIVE)); 48639d3e169Sevanl if (copyin(data, STRUCT_BUF(args), MIN(datalen, 48739d3e169Sevanl STRUCT_SIZE(args)))) 48839d3e169Sevanl return (EFAULT); 48939d3e169Sevanl 49039d3e169Sevanl nargs->wsize = STRUCT_FGET(args, wsize); 49139d3e169Sevanl nargs->rsize = STRUCT_FGET(args, rsize); 49239d3e169Sevanl nargs->timeo = STRUCT_FGET(args, timeo); 49339d3e169Sevanl nargs->retrans = STRUCT_FGET(args, retrans); 49439d3e169Sevanl nargs->acregmin = STRUCT_FGET(args, acregmin); 49539d3e169Sevanl nargs->acregmax = STRUCT_FGET(args, acregmax); 49639d3e169Sevanl nargs->acdirmin = STRUCT_FGET(args, acdirmin); 49739d3e169Sevanl nargs->acdirmax = STRUCT_FGET(args, acdirmax); 49839d3e169Sevanl 49939d3e169Sevanl flags = STRUCT_FGET(args, flags); 50039d3e169Sevanl nargs->flags = flags; 50139d3e169Sevanl 50239d3e169Sevanl addr.buf = NULL; 50339d3e169Sevanl syncaddr.buf = NULL; 50439d3e169Sevanl 50539d3e169Sevanl 50639d3e169Sevanl /* 50739d3e169Sevanl * Allocate space for a knetconfig structure and 50839d3e169Sevanl * its strings and copy in from user-land. 50939d3e169Sevanl */ 51039d3e169Sevanl knconf = kmem_zalloc(sizeof (*knconf), KM_SLEEP); 51139d3e169Sevanl STRUCT_INIT(knconf_tmp, get_udatamodel()); 51239d3e169Sevanl if (copyin(STRUCT_FGETP(args, knconf), STRUCT_BUF(knconf_tmp), 51339d3e169Sevanl STRUCT_SIZE(knconf_tmp))) { 51439d3e169Sevanl kmem_free(knconf, sizeof (*knconf)); 51539d3e169Sevanl return (EFAULT); 51639d3e169Sevanl } 51739d3e169Sevanl 51839d3e169Sevanl knconf->knc_semantics = STRUCT_FGET(knconf_tmp, knc_semantics); 51939d3e169Sevanl knconf->knc_protofmly = STRUCT_FGETP(knconf_tmp, knc_protofmly); 52039d3e169Sevanl knconf->knc_proto = STRUCT_FGETP(knconf_tmp, knc_proto); 52139d3e169Sevanl if (get_udatamodel() != DATAMODEL_LP64) { 52239d3e169Sevanl knconf->knc_rdev = expldev(STRUCT_FGET(knconf_tmp, knc_rdev)); 52339d3e169Sevanl } else { 52439d3e169Sevanl knconf->knc_rdev = STRUCT_FGET(knconf_tmp, knc_rdev); 52539d3e169Sevanl } 52639d3e169Sevanl 52739d3e169Sevanl pf = kmem_alloc(KNC_STRSIZE, KM_SLEEP); 52839d3e169Sevanl p = kmem_alloc(KNC_STRSIZE, KM_SLEEP); 52939d3e169Sevanl error = copyinstr(knconf->knc_protofmly, pf, KNC_STRSIZE, NULL); 53039d3e169Sevanl if (error) { 53139d3e169Sevanl kmem_free(pf, KNC_STRSIZE); 53239d3e169Sevanl kmem_free(p, KNC_STRSIZE); 53339d3e169Sevanl kmem_free(knconf, sizeof (*knconf)); 53439d3e169Sevanl return (error); 53539d3e169Sevanl } 53639d3e169Sevanl 53739d3e169Sevanl error = copyinstr(knconf->knc_proto, p, KNC_STRSIZE, NULL); 53839d3e169Sevanl if (error) { 53939d3e169Sevanl kmem_free(pf, KNC_STRSIZE); 54039d3e169Sevanl kmem_free(p, KNC_STRSIZE); 54139d3e169Sevanl kmem_free(knconf, sizeof (*knconf)); 54239d3e169Sevanl return (error); 54339d3e169Sevanl } 54439d3e169Sevanl 54539d3e169Sevanl 54639d3e169Sevanl knconf->knc_protofmly = pf; 54739d3e169Sevanl knconf->knc_proto = p; 54839d3e169Sevanl 54939d3e169Sevanl nargs->knconf = knconf; 55039d3e169Sevanl 55139d3e169Sevanl /* 55239d3e169Sevanl * Get server address 55339d3e169Sevanl */ 55439d3e169Sevanl STRUCT_INIT(addr_tmp, get_udatamodel()); 55539d3e169Sevanl if (copyin(STRUCT_FGETP(args, addr), STRUCT_BUF(addr_tmp), 55639d3e169Sevanl STRUCT_SIZE(addr_tmp))) { 55739d3e169Sevanl error = EFAULT; 55839d3e169Sevanl goto errout; 55939d3e169Sevanl } 56039d3e169Sevanl 5617e450cbcSJames Wahlig nargs->addr = kmem_zalloc(sizeof (struct netbuf), KM_SLEEP); 56239d3e169Sevanl userbufptr = STRUCT_FGETP(addr_tmp, buf); 56339d3e169Sevanl addr.len = STRUCT_FGET(addr_tmp, len); 56439d3e169Sevanl addr.buf = kmem_alloc(addr.len, KM_SLEEP); 56539d3e169Sevanl addr.maxlen = addr.len; 56639d3e169Sevanl if (copyin(userbufptr, addr.buf, addr.len)) { 56739d3e169Sevanl kmem_free(addr.buf, addr.len); 56839d3e169Sevanl error = EFAULT; 56939d3e169Sevanl goto errout; 57039d3e169Sevanl } 57139d3e169Sevanl bcopy(&addr, nargs->addr, sizeof (struct netbuf)); 57239d3e169Sevanl 57339d3e169Sevanl /* 57439d3e169Sevanl * Get the root fhandle 57539d3e169Sevanl */ 57639d3e169Sevanl error = pn_get(STRUCT_FGETP(args, fh), UIO_USERSPACE, &pn); 57739d3e169Sevanl if (error) 57839d3e169Sevanl goto errout; 57939d3e169Sevanl 58039d3e169Sevanl /* Volatile fh: keep server paths, so use actual-size strings */ 58139d3e169Sevanl nargs->fh = kmem_alloc(pn.pn_pathlen + 1, KM_SLEEP); 58239d3e169Sevanl bcopy(pn.pn_path, nargs->fh, pn.pn_pathlen); 58339d3e169Sevanl nargs->fh[pn.pn_pathlen] = '\0'; 58439d3e169Sevanl pn_free(&pn); 58539d3e169Sevanl 58639d3e169Sevanl 58739d3e169Sevanl /* 58839d3e169Sevanl * Get server's hostname 58939d3e169Sevanl */ 59039d3e169Sevanl if (flags & NFSMNT_HOSTNAME) { 59139d3e169Sevanl error = copyinstr(STRUCT_FGETP(args, hostname), 592b9238976Sth netname, sizeof (netname), &hlen); 59339d3e169Sevanl if (error) 59439d3e169Sevanl goto errout; 59539d3e169Sevanl nargs->hostname = kmem_zalloc(hlen, KM_SLEEP); 59639d3e169Sevanl (void) strcpy(nargs->hostname, netname); 59739d3e169Sevanl 59839d3e169Sevanl } else { 59939d3e169Sevanl nargs->hostname = NULL; 60039d3e169Sevanl } 60139d3e169Sevanl 60239d3e169Sevanl 60339d3e169Sevanl /* 60439d3e169Sevanl * If there are syncaddr and netname data, load them in. This is 60539d3e169Sevanl * to support data needed for NFSV4 when AUTH_DH is the negotiated 60639d3e169Sevanl * flavor via SECINFO. (instead of using MOUNT protocol in V3). 60739d3e169Sevanl */ 60839d3e169Sevanl netname[0] = '\0'; 60939d3e169Sevanl if (flags & NFSMNT_SECURE) { 61039d3e169Sevanl 61139d3e169Sevanl /* get syncaddr */ 61239d3e169Sevanl STRUCT_INIT(addr_tmp, get_udatamodel()); 61339d3e169Sevanl if (copyin(STRUCT_FGETP(args, syncaddr), STRUCT_BUF(addr_tmp), 614b9238976Sth STRUCT_SIZE(addr_tmp))) { 61539d3e169Sevanl error = EINVAL; 61639d3e169Sevanl goto errout; 61739d3e169Sevanl } 61839d3e169Sevanl userbufptr = STRUCT_FGETP(addr_tmp, buf); 61939d3e169Sevanl syncaddr.len = STRUCT_FGET(addr_tmp, len); 62039d3e169Sevanl syncaddr.buf = kmem_alloc(syncaddr.len, KM_SLEEP); 62139d3e169Sevanl syncaddr.maxlen = syncaddr.len; 62239d3e169Sevanl if (copyin(userbufptr, syncaddr.buf, syncaddr.len)) { 62339d3e169Sevanl kmem_free(syncaddr.buf, syncaddr.len); 62439d3e169Sevanl error = EFAULT; 62539d3e169Sevanl goto errout; 62639d3e169Sevanl } 62739d3e169Sevanl 62839d3e169Sevanl nargs->syncaddr = kmem_alloc(sizeof (struct netbuf), KM_SLEEP); 62939d3e169Sevanl bcopy(&syncaddr, nargs->syncaddr, sizeof (struct netbuf)); 63039d3e169Sevanl 63122d5e933Skr /* get server's netname */ 63222d5e933Skr if (copyinstr(STRUCT_FGETP(args, netname), netname, 633b9238976Sth sizeof (netname), &nlen)) { 63422d5e933Skr error = EFAULT; 63522d5e933Skr goto errout; 63622d5e933Skr } 63722d5e933Skr 63822d5e933Skr netname[nlen] = '\0'; 63922d5e933Skr nargs->netname = kmem_zalloc(nlen, KM_SLEEP); 64022d5e933Skr (void) strcpy(nargs->netname, netname); 64122d5e933Skr } 64239d3e169Sevanl 64339d3e169Sevanl /* 64439d3e169Sevanl * Get the extention data which has the security data structure. 64539d3e169Sevanl * This includes data for AUTH_SYS as well. 64639d3e169Sevanl */ 64739d3e169Sevanl if (flags & NFSMNT_NEWARGS) { 64839d3e169Sevanl nargs->nfs_args_ext = STRUCT_FGET(args, nfs_args_ext); 64939d3e169Sevanl if (nargs->nfs_args_ext == NFS_ARGS_EXTA || 650b9238976Sth nargs->nfs_args_ext == NFS_ARGS_EXTB) { 65139d3e169Sevanl /* 65239d3e169Sevanl * Indicating the application is using the new 65339d3e169Sevanl * sec_data structure to pass in the security 65439d3e169Sevanl * data. 65539d3e169Sevanl */ 65639d3e169Sevanl if (STRUCT_FGETP(args, 65739d3e169Sevanl nfs_ext_u.nfs_extA.secdata) != NULL) { 65839d3e169Sevanl error = sec_clnt_loadinfo( 65939d3e169Sevanl (struct sec_data *)STRUCT_FGETP(args, 660b9238976Sth nfs_ext_u.nfs_extA.secdata), 66139d3e169Sevanl &secdata, get_udatamodel()); 66239d3e169Sevanl } 66339d3e169Sevanl nargs->nfs_ext_u.nfs_extA.secdata = secdata; 66439d3e169Sevanl } 66539d3e169Sevanl } 66639d3e169Sevanl 66739d3e169Sevanl if (error) 66839d3e169Sevanl goto errout; 66939d3e169Sevanl 67039d3e169Sevanl /* 67139d3e169Sevanl * Failover support: 67239d3e169Sevanl * 67339d3e169Sevanl * We may have a linked list of nfs_args structures, 67439d3e169Sevanl * which means the user is looking for failover. If 67539d3e169Sevanl * the mount is either not "read-only" or "soft", 67639d3e169Sevanl * we want to bail out with EINVAL. 67739d3e169Sevanl */ 67839d3e169Sevanl if (nargs->nfs_args_ext == NFS_ARGS_EXTB) 67939d3e169Sevanl nargs->nfs_ext_u.nfs_extB.next = 680b9238976Sth STRUCT_FGETP(args, nfs_ext_u.nfs_extB.next); 68139d3e169Sevanl 68239d3e169Sevanl errout: 68339d3e169Sevanl if (error) 68439d3e169Sevanl nfs4_free_args(nargs); 68539d3e169Sevanl 68639d3e169Sevanl return (error); 68739d3e169Sevanl } 68839d3e169Sevanl 68939d3e169Sevanl 6907c478bd9Sstevel@tonic-gate /* 6917c478bd9Sstevel@tonic-gate * nfs mount vfsop 6927c478bd9Sstevel@tonic-gate * Set up mount info record and attach it to vfs struct. 6937c478bd9Sstevel@tonic-gate */ 694b9238976Sth int 6957c478bd9Sstevel@tonic-gate nfs4_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr) 6967c478bd9Sstevel@tonic-gate { 6977c478bd9Sstevel@tonic-gate char *data = uap->dataptr; 6987c478bd9Sstevel@tonic-gate int error; 6997c478bd9Sstevel@tonic-gate vnode_t *rtvp; /* the server's root */ 7007c478bd9Sstevel@tonic-gate mntinfo4_t *mi; /* mount info, pointed at by vfs */ 7017c478bd9Sstevel@tonic-gate struct knetconfig *rdma_knconf; /* rdma transport structure */ 7027c478bd9Sstevel@tonic-gate rnode4_t *rp; 7037c478bd9Sstevel@tonic-gate struct servinfo4 *svp; /* nfs server info */ 7047c478bd9Sstevel@tonic-gate struct servinfo4 *svp_tail = NULL; /* previous nfs server info */ 7057c478bd9Sstevel@tonic-gate struct servinfo4 *svp_head; /* first nfs server info */ 7067c478bd9Sstevel@tonic-gate struct servinfo4 *svp_2ndlast; /* 2nd last in server info list */ 7077c478bd9Sstevel@tonic-gate struct sec_data *secdata; /* security data */ 70839d3e169Sevanl struct nfs_args *args = NULL; 70950a83466Sjwahlig int flags, addr_type, removed; 710108322fbScarlsonj zone_t *zone = nfs_zone(); 7117c478bd9Sstevel@tonic-gate nfs4_error_t n4e; 71245916cd2Sjpk zone_t *mntzone = NULL; 7137c478bd9Sstevel@tonic-gate 7147c478bd9Sstevel@tonic-gate if (secpolicy_fs_mount(cr, mvp, vfsp) != 0) 7157c478bd9Sstevel@tonic-gate return (EPERM); 7167c478bd9Sstevel@tonic-gate if (mvp->v_type != VDIR) 7177c478bd9Sstevel@tonic-gate return (ENOTDIR); 718b9238976Sth 7197c478bd9Sstevel@tonic-gate /* 7207c478bd9Sstevel@tonic-gate * get arguments 7217c478bd9Sstevel@tonic-gate * 7227c478bd9Sstevel@tonic-gate * nfs_args is now versioned and is extensible, so 7237c478bd9Sstevel@tonic-gate * uap->datalen might be different from sizeof (args) 7247c478bd9Sstevel@tonic-gate * in a compatible situation. 7257c478bd9Sstevel@tonic-gate */ 7267c478bd9Sstevel@tonic-gate more: 72739d3e169Sevanl if (!(uap->flags & MS_SYSSPACE)) { 72839d3e169Sevanl if (args == NULL) 72939d3e169Sevanl args = kmem_zalloc(sizeof (struct nfs_args), KM_SLEEP); 73039d3e169Sevanl else 73139d3e169Sevanl nfs4_free_args(args); 73239d3e169Sevanl error = nfs4_copyin(data, uap->datalen, args); 73339d3e169Sevanl if (error) { 73439d3e169Sevanl if (args) { 73539d3e169Sevanl kmem_free(args, sizeof (*args)); 73639d3e169Sevanl } 73739d3e169Sevanl return (error); 73839d3e169Sevanl } 73939d3e169Sevanl } else { 74039d3e169Sevanl args = (struct nfs_args *)data; 74139d3e169Sevanl } 7427c478bd9Sstevel@tonic-gate 74339d3e169Sevanl flags = args->flags; 7447c478bd9Sstevel@tonic-gate 7457c478bd9Sstevel@tonic-gate /* 7467c478bd9Sstevel@tonic-gate * If the request changes the locking type, disallow the remount, 7477c478bd9Sstevel@tonic-gate * because it's questionable whether we can transfer the 7487c478bd9Sstevel@tonic-gate * locking state correctly. 7497c478bd9Sstevel@tonic-gate */ 7507c478bd9Sstevel@tonic-gate if (uap->flags & MS_REMOUNT) { 75139d3e169Sevanl if (!(uap->flags & MS_SYSSPACE)) { 75239d3e169Sevanl nfs4_free_args(args); 75339d3e169Sevanl kmem_free(args, sizeof (*args)); 75439d3e169Sevanl } 7557c478bd9Sstevel@tonic-gate if ((mi = VFTOMI4(vfsp)) != NULL) { 7567c478bd9Sstevel@tonic-gate uint_t new_mi_llock; 7577c478bd9Sstevel@tonic-gate uint_t old_mi_llock; 7587c478bd9Sstevel@tonic-gate new_mi_llock = (flags & NFSMNT_LLOCK) ? 1 : 0; 7597c478bd9Sstevel@tonic-gate old_mi_llock = (mi->mi_flags & MI4_LLOCK) ? 1 : 0; 7607c478bd9Sstevel@tonic-gate if (old_mi_llock != new_mi_llock) 7617c478bd9Sstevel@tonic-gate return (EBUSY); 7627c478bd9Sstevel@tonic-gate } 7637c478bd9Sstevel@tonic-gate return (0); 7647c478bd9Sstevel@tonic-gate } 7657c478bd9Sstevel@tonic-gate 766b9238976Sth /* 767b9238976Sth * For ephemeral mount trigger stub vnodes, we have two problems 768b9238976Sth * to solve: racing threads will likely fail the v_count check, and 769b9238976Sth * we want only one to proceed with the mount. 770b9238976Sth * 771b9238976Sth * For stubs, if the mount has already occurred (via a racing thread), 772b9238976Sth * just return success. If not, skip the v_count check and proceed. 773b9238976Sth * Note that we are already serialised at this point. 774b9238976Sth */ 7757c478bd9Sstevel@tonic-gate mutex_enter(&mvp->v_lock); 776b9238976Sth if (vn_matchops(mvp, nfs4_trigger_vnodeops)) { 777b9238976Sth /* mntpt is a v4 stub vnode */ 778b9238976Sth ASSERT(RP_ISSTUB(VTOR4(mvp))); 779b9238976Sth ASSERT(!(uap->flags & MS_OVERLAY)); 780b9238976Sth ASSERT(!(mvp->v_flag & VROOT)); 781b9238976Sth if (vn_mountedvfs(mvp) != NULL) { 782b9238976Sth /* ephemeral mount has already occurred */ 783b9238976Sth ASSERT(uap->flags & MS_SYSSPACE); 784b9238976Sth mutex_exit(&mvp->v_lock); 785b9238976Sth return (0); 786b9238976Sth } 787b9238976Sth } else { 788b9238976Sth /* mntpt is a non-v4 or v4 non-stub vnode */ 789b9238976Sth if (!(uap->flags & MS_OVERLAY) && 790b9238976Sth (mvp->v_count != 1 || (mvp->v_flag & VROOT))) { 791b9238976Sth mutex_exit(&mvp->v_lock); 792b9238976Sth if (!(uap->flags & MS_SYSSPACE)) { 793b9238976Sth nfs4_free_args(args); 794b9238976Sth kmem_free(args, sizeof (*args)); 795b9238976Sth } 796b9238976Sth return (EBUSY); 79739d3e169Sevanl } 7987c478bd9Sstevel@tonic-gate } 7997c478bd9Sstevel@tonic-gate mutex_exit(&mvp->v_lock); 8007c478bd9Sstevel@tonic-gate 8017c478bd9Sstevel@tonic-gate /* make sure things are zeroed for errout: */ 8027c478bd9Sstevel@tonic-gate rtvp = NULL; 8037c478bd9Sstevel@tonic-gate mi = NULL; 8047c478bd9Sstevel@tonic-gate secdata = NULL; 8057c478bd9Sstevel@tonic-gate 8067c478bd9Sstevel@tonic-gate /* 8077c478bd9Sstevel@tonic-gate * A valid knetconfig structure is required. 8087c478bd9Sstevel@tonic-gate */ 80939d3e169Sevanl if (!(flags & NFSMNT_KNCONF) || 810b9238976Sth args->knconf == NULL || args->knconf->knc_protofmly == NULL || 811b9238976Sth args->knconf->knc_proto == NULL || 812b9238976Sth (strcmp(args->knconf->knc_proto, NC_UDP) == 0)) { 81339d3e169Sevanl if (!(uap->flags & MS_SYSSPACE)) { 81439d3e169Sevanl nfs4_free_args(args); 81539d3e169Sevanl kmem_free(args, sizeof (*args)); 81639d3e169Sevanl } 8177c478bd9Sstevel@tonic-gate return (EINVAL); 81839d3e169Sevanl } 81939d3e169Sevanl 82039d3e169Sevanl if ((strlen(args->knconf->knc_protofmly) >= KNC_STRSIZE) || 821b9238976Sth (strlen(args->knconf->knc_proto) >= KNC_STRSIZE)) { 82239d3e169Sevanl if (!(uap->flags & MS_SYSSPACE)) { 82339d3e169Sevanl nfs4_free_args(args); 82439d3e169Sevanl kmem_free(args, sizeof (*args)); 82539d3e169Sevanl } 82639d3e169Sevanl return (EINVAL); 82739d3e169Sevanl } 82839d3e169Sevanl 8297c478bd9Sstevel@tonic-gate /* 8307c478bd9Sstevel@tonic-gate * Allocate a servinfo4 struct. 8317c478bd9Sstevel@tonic-gate */ 8327c478bd9Sstevel@tonic-gate svp = kmem_zalloc(sizeof (*svp), KM_SLEEP); 8337c478bd9Sstevel@tonic-gate nfs_rw_init(&svp->sv_lock, NULL, RW_DEFAULT, NULL); 8347c478bd9Sstevel@tonic-gate if (svp_tail) { 8357c478bd9Sstevel@tonic-gate svp_2ndlast = svp_tail; 8367c478bd9Sstevel@tonic-gate svp_tail->sv_next = svp; 8377c478bd9Sstevel@tonic-gate } else { 8387c478bd9Sstevel@tonic-gate svp_head = svp; 8397c478bd9Sstevel@tonic-gate svp_2ndlast = svp; 8407c478bd9Sstevel@tonic-gate } 8417c478bd9Sstevel@tonic-gate 8427c478bd9Sstevel@tonic-gate svp_tail = svp; 84339d3e169Sevanl svp->sv_knconf = args->knconf; 84439d3e169Sevanl args->knconf = NULL; 8457c478bd9Sstevel@tonic-gate 8467c478bd9Sstevel@tonic-gate /* 8477c478bd9Sstevel@tonic-gate * Get server address 8487c478bd9Sstevel@tonic-gate */ 84939d3e169Sevanl if (args->addr == NULL || args->addr->buf == NULL) { 85039d3e169Sevanl error = EINVAL; 8517c478bd9Sstevel@tonic-gate goto errout; 8527c478bd9Sstevel@tonic-gate } 8537c478bd9Sstevel@tonic-gate 85439d3e169Sevanl svp->sv_addr.maxlen = args->addr->maxlen; 85539d3e169Sevanl svp->sv_addr.len = args->addr->len; 85639d3e169Sevanl svp->sv_addr.buf = args->addr->buf; 85739d3e169Sevanl args->addr->buf = NULL; 85839d3e169Sevanl 8597c478bd9Sstevel@tonic-gate /* 8607c478bd9Sstevel@tonic-gate * Get the root fhandle 8617c478bd9Sstevel@tonic-gate */ 86239d3e169Sevanl if (args->fh == NULL || (strlen(args->fh) >= MAXPATHLEN)) { 86339d3e169Sevanl error = EINVAL; 8647c478bd9Sstevel@tonic-gate goto errout; 86539d3e169Sevanl } 8667c478bd9Sstevel@tonic-gate 86739d3e169Sevanl svp->sv_path = args->fh; 86839d3e169Sevanl svp->sv_pathlen = strlen(args->fh) + 1; 86939d3e169Sevanl args->fh = NULL; 8707c478bd9Sstevel@tonic-gate 8717c478bd9Sstevel@tonic-gate /* 8727c478bd9Sstevel@tonic-gate * Get server's hostname 8737c478bd9Sstevel@tonic-gate */ 8747c478bd9Sstevel@tonic-gate if (flags & NFSMNT_HOSTNAME) { 87539d3e169Sevanl if (args->hostname == NULL || (strlen(args->hostname) > 876b9238976Sth MAXNETNAMELEN)) { 87739d3e169Sevanl error = EINVAL; 8787c478bd9Sstevel@tonic-gate goto errout; 87939d3e169Sevanl } 88039d3e169Sevanl svp->sv_hostnamelen = strlen(args->hostname) + 1; 88139d3e169Sevanl svp->sv_hostname = args->hostname; 88239d3e169Sevanl args->hostname = NULL; 8837c478bd9Sstevel@tonic-gate } else { 8847c478bd9Sstevel@tonic-gate char *p = "unknown-host"; 88539d3e169Sevanl svp->sv_hostnamelen = strlen(p) + 1; 88639d3e169Sevanl svp->sv_hostname = kmem_zalloc(svp->sv_hostnamelen, KM_SLEEP); 88739d3e169Sevanl (void) strcpy(svp->sv_hostname, p); 8887c478bd9Sstevel@tonic-gate } 8897c478bd9Sstevel@tonic-gate 8907c478bd9Sstevel@tonic-gate /* 8917c478bd9Sstevel@tonic-gate * RDMA MOUNT SUPPORT FOR NFS v4. 8927c478bd9Sstevel@tonic-gate * Establish, is it possible to use RDMA, if so overload the 8937c478bd9Sstevel@tonic-gate * knconf with rdma specific knconf and free the orignal knconf. 8947c478bd9Sstevel@tonic-gate */ 8957c478bd9Sstevel@tonic-gate if ((flags & NFSMNT_TRYRDMA) || (flags & NFSMNT_DORDMA)) { 8967c478bd9Sstevel@tonic-gate /* 8977c478bd9Sstevel@tonic-gate * Determine the addr type for RDMA, IPv4 or v6. 8987c478bd9Sstevel@tonic-gate */ 8997c478bd9Sstevel@tonic-gate if (strcmp(svp->sv_knconf->knc_protofmly, NC_INET) == 0) 9007c478bd9Sstevel@tonic-gate addr_type = AF_INET; 9017c478bd9Sstevel@tonic-gate else if (strcmp(svp->sv_knconf->knc_protofmly, NC_INET6) == 0) 9027c478bd9Sstevel@tonic-gate addr_type = AF_INET6; 9037c478bd9Sstevel@tonic-gate 9047c478bd9Sstevel@tonic-gate if (rdma_reachable(addr_type, &svp->sv_addr, 905b9238976Sth &rdma_knconf) == 0) { 9067c478bd9Sstevel@tonic-gate /* 9077c478bd9Sstevel@tonic-gate * If successful, hijack the orignal knconf and 9087c478bd9Sstevel@tonic-gate * replace with the new one, depending on the flags. 9097c478bd9Sstevel@tonic-gate */ 9107c478bd9Sstevel@tonic-gate svp->sv_origknconf = svp->sv_knconf; 9117c478bd9Sstevel@tonic-gate svp->sv_knconf = rdma_knconf; 9127c478bd9Sstevel@tonic-gate } else { 9137c478bd9Sstevel@tonic-gate if (flags & NFSMNT_TRYRDMA) { 9147c478bd9Sstevel@tonic-gate #ifdef DEBUG 9157c478bd9Sstevel@tonic-gate if (rdma_debug) 9167c478bd9Sstevel@tonic-gate zcmn_err(getzoneid(), CE_WARN, 9177c478bd9Sstevel@tonic-gate "no RDMA onboard, revert\n"); 9187c478bd9Sstevel@tonic-gate #endif 9197c478bd9Sstevel@tonic-gate } 9207c478bd9Sstevel@tonic-gate 9217c478bd9Sstevel@tonic-gate if (flags & NFSMNT_DORDMA) { 9227c478bd9Sstevel@tonic-gate /* 9237c478bd9Sstevel@tonic-gate * If proto=rdma is specified and no RDMA 9247c478bd9Sstevel@tonic-gate * path to this server is avialable then 9257c478bd9Sstevel@tonic-gate * ditch this server. 9267c478bd9Sstevel@tonic-gate * This is not included in the mountable 9277c478bd9Sstevel@tonic-gate * server list or the replica list. 9287c478bd9Sstevel@tonic-gate * Check if more servers are specified; 9297c478bd9Sstevel@tonic-gate * Failover case, otherwise bail out of mount. 9307c478bd9Sstevel@tonic-gate */ 931b9238976Sth if (args->nfs_args_ext == NFS_ARGS_EXTB && 932b9238976Sth args->nfs_ext_u.nfs_extB.next != NULL) { 93339d3e169Sevanl data = (char *) 934b9238976Sth args->nfs_ext_u.nfs_extB.next; 9357c478bd9Sstevel@tonic-gate if (uap->flags & MS_RDONLY && 9367c478bd9Sstevel@tonic-gate !(flags & NFSMNT_SOFT)) { 9377c478bd9Sstevel@tonic-gate if (svp_head->sv_next == NULL) { 9387c478bd9Sstevel@tonic-gate svp_tail = NULL; 9397c478bd9Sstevel@tonic-gate svp_2ndlast = NULL; 9407c478bd9Sstevel@tonic-gate sv4_free(svp_head); 9417c478bd9Sstevel@tonic-gate goto more; 9427c478bd9Sstevel@tonic-gate } else { 9437c478bd9Sstevel@tonic-gate svp_tail = svp_2ndlast; 9447c478bd9Sstevel@tonic-gate svp_2ndlast->sv_next = 9457c478bd9Sstevel@tonic-gate NULL; 9467c478bd9Sstevel@tonic-gate sv4_free(svp); 9477c478bd9Sstevel@tonic-gate goto more; 9487c478bd9Sstevel@tonic-gate } 9497c478bd9Sstevel@tonic-gate } 9507c478bd9Sstevel@tonic-gate } else { 9517c478bd9Sstevel@tonic-gate /* 9527c478bd9Sstevel@tonic-gate * This is the last server specified 9537c478bd9Sstevel@tonic-gate * in the nfs_args list passed down 9547c478bd9Sstevel@tonic-gate * and its not rdma capable. 9557c478bd9Sstevel@tonic-gate */ 9567c478bd9Sstevel@tonic-gate if (svp_head->sv_next == NULL) { 9577c478bd9Sstevel@tonic-gate /* 9587c478bd9Sstevel@tonic-gate * Is this the only one 9597c478bd9Sstevel@tonic-gate */ 9607c478bd9Sstevel@tonic-gate error = EINVAL; 9617c478bd9Sstevel@tonic-gate #ifdef DEBUG 9627c478bd9Sstevel@tonic-gate if (rdma_debug) 9637c478bd9Sstevel@tonic-gate zcmn_err(getzoneid(), 9647c478bd9Sstevel@tonic-gate CE_WARN, 9657c478bd9Sstevel@tonic-gate "No RDMA srv"); 9667c478bd9Sstevel@tonic-gate #endif 9677c478bd9Sstevel@tonic-gate goto errout; 9687c478bd9Sstevel@tonic-gate } else { 9697c478bd9Sstevel@tonic-gate /* 9707c478bd9Sstevel@tonic-gate * There is list, since some 9717c478bd9Sstevel@tonic-gate * servers specified before 9727c478bd9Sstevel@tonic-gate * this passed all requirements 9737c478bd9Sstevel@tonic-gate */ 9747c478bd9Sstevel@tonic-gate svp_tail = svp_2ndlast; 9757c478bd9Sstevel@tonic-gate svp_2ndlast->sv_next = NULL; 9767c478bd9Sstevel@tonic-gate sv4_free(svp); 9777c478bd9Sstevel@tonic-gate goto proceed; 9787c478bd9Sstevel@tonic-gate } 9797c478bd9Sstevel@tonic-gate } 9807c478bd9Sstevel@tonic-gate } 9817c478bd9Sstevel@tonic-gate } 9827c478bd9Sstevel@tonic-gate } 9837c478bd9Sstevel@tonic-gate 9847c478bd9Sstevel@tonic-gate /* 9857c478bd9Sstevel@tonic-gate * If there are syncaddr and netname data, load them in. This is 9867c478bd9Sstevel@tonic-gate * to support data needed for NFSV4 when AUTH_DH is the negotiated 9877c478bd9Sstevel@tonic-gate * flavor via SECINFO. (instead of using MOUNT protocol in V3). 9887c478bd9Sstevel@tonic-gate */ 98939d3e169Sevanl if (args->flags & NFSMNT_SECURE) { 99039d3e169Sevanl svp->sv_dhsec = create_authdh_data(args->netname, 991b9238976Sth strlen(args->netname), 992b9238976Sth args->syncaddr, svp->sv_knconf); 9937c478bd9Sstevel@tonic-gate } 9947c478bd9Sstevel@tonic-gate 9957c478bd9Sstevel@tonic-gate /* 9967c478bd9Sstevel@tonic-gate * Get the extention data which has the security data structure. 9977c478bd9Sstevel@tonic-gate * This includes data for AUTH_SYS as well. 9987c478bd9Sstevel@tonic-gate */ 9997c478bd9Sstevel@tonic-gate if (flags & NFSMNT_NEWARGS) { 100039d3e169Sevanl switch (args->nfs_args_ext) { 10017c478bd9Sstevel@tonic-gate case NFS_ARGS_EXTA: 10027c478bd9Sstevel@tonic-gate case NFS_ARGS_EXTB: 10037c478bd9Sstevel@tonic-gate /* 10047c478bd9Sstevel@tonic-gate * Indicating the application is using the new 10057c478bd9Sstevel@tonic-gate * sec_data structure to pass in the security 10067c478bd9Sstevel@tonic-gate * data. 10077c478bd9Sstevel@tonic-gate */ 100839d3e169Sevanl secdata = args->nfs_ext_u.nfs_extA.secdata; 100939d3e169Sevanl if (secdata == NULL) { 10107c478bd9Sstevel@tonic-gate error = EINVAL; 101139d3e169Sevanl } else if (uap->flags & MS_SYSSPACE) { 101239d3e169Sevanl /* 101339d3e169Sevanl * Need to validate the flavor here if 101439d3e169Sevanl * sysspace, userspace was already 101539d3e169Sevanl * validate from the nfs_copyin function. 101639d3e169Sevanl */ 101739d3e169Sevanl switch (secdata->rpcflavor) { 101839d3e169Sevanl case AUTH_NONE: 101939d3e169Sevanl case AUTH_UNIX: 102039d3e169Sevanl case AUTH_LOOPBACK: 102139d3e169Sevanl case AUTH_DES: 102239d3e169Sevanl case RPCSEC_GSS: 102339d3e169Sevanl break; 102439d3e169Sevanl default: 102539d3e169Sevanl error = EINVAL; 102639d3e169Sevanl goto errout; 102739d3e169Sevanl } 10287c478bd9Sstevel@tonic-gate } 102939d3e169Sevanl args->nfs_ext_u.nfs_extA.secdata = NULL; 10307c478bd9Sstevel@tonic-gate break; 10317c478bd9Sstevel@tonic-gate 10327c478bd9Sstevel@tonic-gate default: 10337c478bd9Sstevel@tonic-gate error = EINVAL; 10347c478bd9Sstevel@tonic-gate break; 10357c478bd9Sstevel@tonic-gate } 10367c478bd9Sstevel@tonic-gate 10377c478bd9Sstevel@tonic-gate } else if (flags & NFSMNT_SECURE) { 10387c478bd9Sstevel@tonic-gate /* 10397c478bd9Sstevel@tonic-gate * NFSMNT_SECURE is deprecated but we keep it 1040b9238976Sth * to support the rogue user-generated application 10417c478bd9Sstevel@tonic-gate * that may use this undocumented interface to do 1042b9238976Sth * AUTH_DH security, e.g. our own rexd. 1043b9238976Sth * 1044b9238976Sth * Also note that NFSMNT_SECURE is used for passing 1045b9238976Sth * AUTH_DH info to be used in negotiation. 10467c478bd9Sstevel@tonic-gate */ 104739d3e169Sevanl secdata = create_authdh_data(args->netname, 1048b9238976Sth strlen(args->netname), args->syncaddr, svp->sv_knconf); 10497c478bd9Sstevel@tonic-gate 10507c478bd9Sstevel@tonic-gate } else { 10517c478bd9Sstevel@tonic-gate secdata = kmem_alloc(sizeof (*secdata), KM_SLEEP); 10527c478bd9Sstevel@tonic-gate secdata->secmod = secdata->rpcflavor = AUTH_SYS; 10537c478bd9Sstevel@tonic-gate secdata->data = NULL; 10547c478bd9Sstevel@tonic-gate } 10557c478bd9Sstevel@tonic-gate 10567c478bd9Sstevel@tonic-gate svp->sv_secdata = secdata; 10577c478bd9Sstevel@tonic-gate 10587c478bd9Sstevel@tonic-gate /* 10597c478bd9Sstevel@tonic-gate * User does not explictly specify a flavor, and a user 10607c478bd9Sstevel@tonic-gate * defined default flavor is passed down. 10617c478bd9Sstevel@tonic-gate */ 10627c478bd9Sstevel@tonic-gate if (flags & NFSMNT_SECDEFAULT) { 10637c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_WRITER, 0); 10647c478bd9Sstevel@tonic-gate svp->sv_flags |= SV4_TRYSECDEFAULT; 10657c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 10667c478bd9Sstevel@tonic-gate } 10677c478bd9Sstevel@tonic-gate 10687c478bd9Sstevel@tonic-gate /* 10697c478bd9Sstevel@tonic-gate * Failover support: 10707c478bd9Sstevel@tonic-gate * 10717c478bd9Sstevel@tonic-gate * We may have a linked list of nfs_args structures, 10727c478bd9Sstevel@tonic-gate * which means the user is looking for failover. If 10737c478bd9Sstevel@tonic-gate * the mount is either not "read-only" or "soft", 10747c478bd9Sstevel@tonic-gate * we want to bail out with EINVAL. 10757c478bd9Sstevel@tonic-gate */ 107639d3e169Sevanl if (args->nfs_args_ext == NFS_ARGS_EXTB && 107739d3e169Sevanl args->nfs_ext_u.nfs_extB.next != NULL) { 10787c478bd9Sstevel@tonic-gate if (uap->flags & MS_RDONLY && !(flags & NFSMNT_SOFT)) { 107939d3e169Sevanl data = (char *)args->nfs_ext_u.nfs_extB.next; 10807c478bd9Sstevel@tonic-gate goto more; 10817c478bd9Sstevel@tonic-gate } 10827c478bd9Sstevel@tonic-gate error = EINVAL; 10837c478bd9Sstevel@tonic-gate goto errout; 10847c478bd9Sstevel@tonic-gate } 10857c478bd9Sstevel@tonic-gate 10867c478bd9Sstevel@tonic-gate /* 10877c478bd9Sstevel@tonic-gate * Determine the zone we're being mounted into. 10887c478bd9Sstevel@tonic-gate */ 108945916cd2Sjpk zone_hold(mntzone = zone); /* start with this assumption */ 10907c478bd9Sstevel@tonic-gate if (getzoneid() == GLOBAL_ZONEID) { 109145916cd2Sjpk zone_rele(mntzone); 10927c478bd9Sstevel@tonic-gate mntzone = zone_find_by_path(refstr_value(vfsp->vfs_mntpt)); 10937c478bd9Sstevel@tonic-gate ASSERT(mntzone != NULL); 10947c478bd9Sstevel@tonic-gate if (mntzone != zone) { 10957c478bd9Sstevel@tonic-gate error = EBUSY; 10967c478bd9Sstevel@tonic-gate goto errout; 10977c478bd9Sstevel@tonic-gate } 10987c478bd9Sstevel@tonic-gate } 10997c478bd9Sstevel@tonic-gate 110045916cd2Sjpk if (is_system_labeled()) { 110145916cd2Sjpk error = nfs_mount_label_policy(vfsp, &svp->sv_addr, 110245916cd2Sjpk svp->sv_knconf, cr); 110345916cd2Sjpk 110445916cd2Sjpk if (error > 0) 110545916cd2Sjpk goto errout; 110645916cd2Sjpk 110745916cd2Sjpk if (error == -1) { 110845916cd2Sjpk /* change mount to read-only to prevent write-down */ 110945916cd2Sjpk vfs_setmntopt(vfsp, MNTOPT_RO, NULL, 0); 111045916cd2Sjpk } 111145916cd2Sjpk } 111245916cd2Sjpk 11137c478bd9Sstevel@tonic-gate /* 11147c478bd9Sstevel@tonic-gate * Stop the mount from going any further if the zone is going away. 11157c478bd9Sstevel@tonic-gate */ 111645916cd2Sjpk if (zone_status_get(mntzone) >= ZONE_IS_SHUTTING_DOWN) { 11177c478bd9Sstevel@tonic-gate error = EBUSY; 11187c478bd9Sstevel@tonic-gate goto errout; 11197c478bd9Sstevel@tonic-gate } 11207c478bd9Sstevel@tonic-gate 11217c478bd9Sstevel@tonic-gate /* 11227c478bd9Sstevel@tonic-gate * Get root vnode. 11237c478bd9Sstevel@tonic-gate */ 11247c478bd9Sstevel@tonic-gate proceed: 112545916cd2Sjpk error = nfs4rootvp(&rtvp, vfsp, svp_head, flags, cr, mntzone); 112650a83466Sjwahlig if (error) { 112750a83466Sjwahlig /* if nfs4rootvp failed, it will free svp_head */ 112850a83466Sjwahlig svp_head = NULL; 11297c478bd9Sstevel@tonic-gate goto errout; 113050a83466Sjwahlig } 11317c478bd9Sstevel@tonic-gate 11327c478bd9Sstevel@tonic-gate mi = VTOMI4(rtvp); 11337c478bd9Sstevel@tonic-gate 11347c478bd9Sstevel@tonic-gate /* 11357c478bd9Sstevel@tonic-gate * Send client id to the server, if necessary 11367c478bd9Sstevel@tonic-gate */ 11377c478bd9Sstevel@tonic-gate nfs4_error_zinit(&n4e); 11387c478bd9Sstevel@tonic-gate nfs4setclientid(mi, cr, FALSE, &n4e); 1139b9238976Sth 11407c478bd9Sstevel@tonic-gate error = n4e.error; 11417c478bd9Sstevel@tonic-gate 11427c478bd9Sstevel@tonic-gate if (error) 11437c478bd9Sstevel@tonic-gate goto errout; 11447c478bd9Sstevel@tonic-gate 11457c478bd9Sstevel@tonic-gate /* 11467c478bd9Sstevel@tonic-gate * Set option fields in the mount info record 11477c478bd9Sstevel@tonic-gate */ 11487c478bd9Sstevel@tonic-gate 11497c478bd9Sstevel@tonic-gate if (svp_head->sv_next) { 11507c478bd9Sstevel@tonic-gate mutex_enter(&mi->mi_lock); 11517c478bd9Sstevel@tonic-gate mi->mi_flags |= MI4_LLOCK; 11527c478bd9Sstevel@tonic-gate mutex_exit(&mi->mi_lock); 11537c478bd9Sstevel@tonic-gate } 115439d3e169Sevanl error = nfs4_setopts(rtvp, DATAMODEL_NATIVE, args); 1155b9238976Sth if (error) 1156b9238976Sth goto errout; 1157b9238976Sth 1158b9238976Sth /* 1159b9238976Sth * Time to tie in the mirror mount info at last! 1160b9238976Sth */ 1161b9238976Sth if (flags & NFSMNT_EPHEMERAL) 1162d3a14591SThomas Haynes error = nfs4_record_ephemeral_mount(mi, mvp); 11637c478bd9Sstevel@tonic-gate 11647c478bd9Sstevel@tonic-gate errout: 11657c478bd9Sstevel@tonic-gate if (error) { 11667c478bd9Sstevel@tonic-gate if (rtvp != NULL) { 11677c478bd9Sstevel@tonic-gate rp = VTOR4(rtvp); 11687c478bd9Sstevel@tonic-gate if (rp->r_flags & R4HASHED) 11697c478bd9Sstevel@tonic-gate rp4_rmhash(rp); 11707c478bd9Sstevel@tonic-gate } 11717c478bd9Sstevel@tonic-gate if (mi != NULL) { 11727c478bd9Sstevel@tonic-gate nfs4_async_stop(vfsp); 11737c478bd9Sstevel@tonic-gate nfs4_async_manager_stop(vfsp); 11747c478bd9Sstevel@tonic-gate nfs4_remove_mi_from_server(mi, NULL); 117550a83466Sjwahlig if (rtvp != NULL) 11767c478bd9Sstevel@tonic-gate VN_RELE(rtvp); 117745916cd2Sjpk if (mntzone != NULL) 117845916cd2Sjpk zone_rele(mntzone); 117950a83466Sjwahlig /* need to remove it from the zone */ 118050a83466Sjwahlig removed = nfs4_mi_zonelist_remove(mi); 118150a83466Sjwahlig if (removed) 118250a83466Sjwahlig zone_rele(mi->mi_zone); 118350a83466Sjwahlig MI4_RELE(mi); 118439d3e169Sevanl if (!(uap->flags & MS_SYSSPACE) && args) { 118539d3e169Sevanl nfs4_free_args(args); 118639d3e169Sevanl kmem_free(args, sizeof (*args)); 118739d3e169Sevanl } 11887c478bd9Sstevel@tonic-gate return (error); 11897c478bd9Sstevel@tonic-gate } 119050a83466Sjwahlig if (svp_head) 119150a83466Sjwahlig sv4_free(svp_head); 11927c478bd9Sstevel@tonic-gate } 11937c478bd9Sstevel@tonic-gate 119439d3e169Sevanl if (!(uap->flags & MS_SYSSPACE) && args) { 119539d3e169Sevanl nfs4_free_args(args); 119639d3e169Sevanl kmem_free(args, sizeof (*args)); 119739d3e169Sevanl } 11987c478bd9Sstevel@tonic-gate if (rtvp != NULL) 11997c478bd9Sstevel@tonic-gate VN_RELE(rtvp); 12007c478bd9Sstevel@tonic-gate 120145916cd2Sjpk if (mntzone != NULL) 120245916cd2Sjpk zone_rele(mntzone); 120345916cd2Sjpk 12047c478bd9Sstevel@tonic-gate return (error); 12057c478bd9Sstevel@tonic-gate } 12067c478bd9Sstevel@tonic-gate 120739d3e169Sevanl #ifdef DEBUG 12087c478bd9Sstevel@tonic-gate #define VERS_MSG "NFS4 server " 12097c478bd9Sstevel@tonic-gate #else 12107c478bd9Sstevel@tonic-gate #define VERS_MSG "NFS server " 12117c478bd9Sstevel@tonic-gate #endif 12127c478bd9Sstevel@tonic-gate 121339d3e169Sevanl #define READ_MSG \ 12147c478bd9Sstevel@tonic-gate VERS_MSG "%s returned 0 for read transfer size" 121539d3e169Sevanl #define WRITE_MSG \ 12167c478bd9Sstevel@tonic-gate VERS_MSG "%s returned 0 for write transfer size" 121739d3e169Sevanl #define SIZE_MSG \ 12187c478bd9Sstevel@tonic-gate VERS_MSG "%s returned 0 for maximum file size" 12197c478bd9Sstevel@tonic-gate 12207c478bd9Sstevel@tonic-gate /* 12217c478bd9Sstevel@tonic-gate * Get the symbolic link text from the server for a given filehandle 12227c478bd9Sstevel@tonic-gate * of that symlink. 12237c478bd9Sstevel@tonic-gate * 122439d3e169Sevanl * (get symlink text) PUTFH READLINK 12257c478bd9Sstevel@tonic-gate */ 12267c478bd9Sstevel@tonic-gate static int 12277c478bd9Sstevel@tonic-gate getlinktext_otw(mntinfo4_t *mi, nfs_fh4 *fh, char **linktextp, cred_t *cr, 1228b9238976Sth int flags) 12297c478bd9Sstevel@tonic-gate { 12307c478bd9Sstevel@tonic-gate COMPOUND4args_clnt args; 12317c478bd9Sstevel@tonic-gate COMPOUND4res_clnt res; 12327c478bd9Sstevel@tonic-gate int doqueue; 12337c478bd9Sstevel@tonic-gate nfs_argop4 argop[2]; 12347c478bd9Sstevel@tonic-gate nfs_resop4 *resop; 12357c478bd9Sstevel@tonic-gate READLINK4res *lr_res; 12367c478bd9Sstevel@tonic-gate uint_t len; 12377c478bd9Sstevel@tonic-gate bool_t needrecov = FALSE; 12387c478bd9Sstevel@tonic-gate nfs4_recov_state_t recov_state; 12397c478bd9Sstevel@tonic-gate nfs4_sharedfh_t *sfh; 12407c478bd9Sstevel@tonic-gate nfs4_error_t e; 12417c478bd9Sstevel@tonic-gate int num_retry = nfs4_max_mount_retry; 12427c478bd9Sstevel@tonic-gate int recovery = !(flags & NFS4_GETFH_NEEDSOP); 12437c478bd9Sstevel@tonic-gate 12447c478bd9Sstevel@tonic-gate sfh = sfh4_get(fh, mi); 12457c478bd9Sstevel@tonic-gate recov_state.rs_flags = 0; 12467c478bd9Sstevel@tonic-gate recov_state.rs_num_retry_despite_err = 0; 12477c478bd9Sstevel@tonic-gate 12487c478bd9Sstevel@tonic-gate recov_retry: 12497c478bd9Sstevel@tonic-gate nfs4_error_zinit(&e); 12507c478bd9Sstevel@tonic-gate 12517c478bd9Sstevel@tonic-gate args.array_len = 2; 12527c478bd9Sstevel@tonic-gate args.array = argop; 12537c478bd9Sstevel@tonic-gate args.ctag = TAG_GET_SYMLINK; 12547c478bd9Sstevel@tonic-gate 12557c478bd9Sstevel@tonic-gate if (! recovery) { 12567c478bd9Sstevel@tonic-gate e.error = nfs4_start_op(mi, NULL, NULL, &recov_state); 12577c478bd9Sstevel@tonic-gate if (e.error) { 12587c478bd9Sstevel@tonic-gate sfh4_rele(&sfh); 12597c478bd9Sstevel@tonic-gate return (e.error); 12607c478bd9Sstevel@tonic-gate } 12617c478bd9Sstevel@tonic-gate } 12627c478bd9Sstevel@tonic-gate 12637c478bd9Sstevel@tonic-gate /* 0. putfh symlink fh */ 12647c478bd9Sstevel@tonic-gate argop[0].argop = OP_CPUTFH; 12657c478bd9Sstevel@tonic-gate argop[0].nfs_argop4_u.opcputfh.sfh = sfh; 12667c478bd9Sstevel@tonic-gate 12677c478bd9Sstevel@tonic-gate /* 1. readlink */ 12687c478bd9Sstevel@tonic-gate argop[1].argop = OP_READLINK; 12697c478bd9Sstevel@tonic-gate 12707c478bd9Sstevel@tonic-gate doqueue = 1; 12717c478bd9Sstevel@tonic-gate 12727c478bd9Sstevel@tonic-gate rfs4call(mi, &args, &res, cr, &doqueue, 0, &e); 12737c478bd9Sstevel@tonic-gate 12747c478bd9Sstevel@tonic-gate needrecov = nfs4_needs_recovery(&e, FALSE, mi->mi_vfsp); 12757c478bd9Sstevel@tonic-gate 12767c478bd9Sstevel@tonic-gate if (needrecov && !recovery && num_retry-- > 0) { 12777c478bd9Sstevel@tonic-gate 12787c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_recov_debug, (CE_NOTE, 1279b9238976Sth "getlinktext_otw: initiating recovery\n")); 12807c478bd9Sstevel@tonic-gate 12817c478bd9Sstevel@tonic-gate if (nfs4_start_recovery(&e, mi, NULL, NULL, NULL, NULL, 12822f172c55SRobert Thurlow OP_READLINK, NULL, NULL, NULL) == FALSE) { 1283b9238976Sth nfs4_end_op(mi, NULL, NULL, &recov_state, needrecov); 1284b9238976Sth if (!e.error) 1285b9238976Sth (void) xdr_free(xdr_COMPOUND4res_clnt, 1286b9238976Sth (caddr_t)&res); 12877c478bd9Sstevel@tonic-gate goto recov_retry; 12887c478bd9Sstevel@tonic-gate } 12897c478bd9Sstevel@tonic-gate } 12907c478bd9Sstevel@tonic-gate 12917c478bd9Sstevel@tonic-gate /* 12927c478bd9Sstevel@tonic-gate * If non-NFS4 pcol error and/or we weren't able to recover. 12937c478bd9Sstevel@tonic-gate */ 12947c478bd9Sstevel@tonic-gate if (e.error != 0) { 12957c478bd9Sstevel@tonic-gate if (! recovery) 12967c478bd9Sstevel@tonic-gate nfs4_end_op(mi, NULL, NULL, &recov_state, needrecov); 12977c478bd9Sstevel@tonic-gate sfh4_rele(&sfh); 12987c478bd9Sstevel@tonic-gate return (e.error); 12997c478bd9Sstevel@tonic-gate } 13007c478bd9Sstevel@tonic-gate 13017c478bd9Sstevel@tonic-gate if (res.status) { 13027c478bd9Sstevel@tonic-gate e.error = geterrno4(res.status); 13037c478bd9Sstevel@tonic-gate (void) xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&res); 13047c478bd9Sstevel@tonic-gate if (! recovery) 13057c478bd9Sstevel@tonic-gate nfs4_end_op(mi, NULL, NULL, &recov_state, needrecov); 13067c478bd9Sstevel@tonic-gate sfh4_rele(&sfh); 13077c478bd9Sstevel@tonic-gate return (e.error); 13087c478bd9Sstevel@tonic-gate } 13097c478bd9Sstevel@tonic-gate 13107c478bd9Sstevel@tonic-gate /* res.status == NFS4_OK */ 13117c478bd9Sstevel@tonic-gate ASSERT(res.status == NFS4_OK); 13127c478bd9Sstevel@tonic-gate 131339d3e169Sevanl resop = &res.array[1]; /* readlink res */ 13147c478bd9Sstevel@tonic-gate lr_res = &resop->nfs_resop4_u.opreadlink; 13157c478bd9Sstevel@tonic-gate 13167c478bd9Sstevel@tonic-gate /* treat symlink name as data */ 13177c478bd9Sstevel@tonic-gate *linktextp = utf8_to_str(&lr_res->link, &len, NULL); 13187c478bd9Sstevel@tonic-gate 13197c478bd9Sstevel@tonic-gate if (! recovery) 13207c478bd9Sstevel@tonic-gate nfs4_end_op(mi, NULL, NULL, &recov_state, needrecov); 13217c478bd9Sstevel@tonic-gate sfh4_rele(&sfh); 13227c478bd9Sstevel@tonic-gate (void) xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&res); 13237c478bd9Sstevel@tonic-gate return (0); 13247c478bd9Sstevel@tonic-gate } 13257c478bd9Sstevel@tonic-gate 13267c478bd9Sstevel@tonic-gate /* 13277c478bd9Sstevel@tonic-gate * Skip over consecutive slashes and "/./" in a pathname. 13287c478bd9Sstevel@tonic-gate */ 13297c478bd9Sstevel@tonic-gate void 13307c478bd9Sstevel@tonic-gate pathname_skipslashdot(struct pathname *pnp) 13317c478bd9Sstevel@tonic-gate { 13327c478bd9Sstevel@tonic-gate char *c1, *c2; 13337c478bd9Sstevel@tonic-gate 13347c478bd9Sstevel@tonic-gate while (pnp->pn_pathlen > 0 && *pnp->pn_path == '/') { 13357c478bd9Sstevel@tonic-gate 13367c478bd9Sstevel@tonic-gate c1 = pnp->pn_path + 1; 13377c478bd9Sstevel@tonic-gate c2 = pnp->pn_path + 2; 13387c478bd9Sstevel@tonic-gate 13397c478bd9Sstevel@tonic-gate if (*c1 == '.' && (*c2 == '/' || *c2 == '\0')) { 13407c478bd9Sstevel@tonic-gate pnp->pn_path = pnp->pn_path + 2; /* skip "/." */ 13417c478bd9Sstevel@tonic-gate pnp->pn_pathlen = pnp->pn_pathlen - 2; 13427c478bd9Sstevel@tonic-gate } else { 13437c478bd9Sstevel@tonic-gate pnp->pn_path++; 13447c478bd9Sstevel@tonic-gate pnp->pn_pathlen--; 13457c478bd9Sstevel@tonic-gate } 13467c478bd9Sstevel@tonic-gate } 13477c478bd9Sstevel@tonic-gate } 13487c478bd9Sstevel@tonic-gate 13497c478bd9Sstevel@tonic-gate /* 13507c478bd9Sstevel@tonic-gate * Resolve a symbolic link path. The symlink is in the nth component of 13517c478bd9Sstevel@tonic-gate * svp->sv_path and has an nfs4 file handle "fh". 13527c478bd9Sstevel@tonic-gate * Upon return, the sv_path will point to the new path that has the nth 13537c478bd9Sstevel@tonic-gate * component resolved to its symlink text. 13547c478bd9Sstevel@tonic-gate */ 13557c478bd9Sstevel@tonic-gate int 13567c478bd9Sstevel@tonic-gate resolve_sympath(mntinfo4_t *mi, servinfo4_t *svp, int nth, nfs_fh4 *fh, 1357b9238976Sth cred_t *cr, int flags) 13587c478bd9Sstevel@tonic-gate { 13597c478bd9Sstevel@tonic-gate char *oldpath; 13607c478bd9Sstevel@tonic-gate char *symlink, *newpath; 13617c478bd9Sstevel@tonic-gate struct pathname oldpn, newpn; 13627c478bd9Sstevel@tonic-gate char component[MAXNAMELEN]; 13637c478bd9Sstevel@tonic-gate int i, addlen, error = 0; 13647c478bd9Sstevel@tonic-gate int oldpathlen; 13657c478bd9Sstevel@tonic-gate 13667c478bd9Sstevel@tonic-gate /* Get the symbolic link text over the wire. */ 13677c478bd9Sstevel@tonic-gate error = getlinktext_otw(mi, fh, &symlink, cr, flags); 13687c478bd9Sstevel@tonic-gate 13697c478bd9Sstevel@tonic-gate if (error || symlink == NULL || strlen(symlink) == 0) 13707c478bd9Sstevel@tonic-gate return (error); 13717c478bd9Sstevel@tonic-gate 13727c478bd9Sstevel@tonic-gate /* 13737c478bd9Sstevel@tonic-gate * Compose the new pathname. 13747c478bd9Sstevel@tonic-gate * Note: 13757c478bd9Sstevel@tonic-gate * - only the nth component is resolved for the pathname. 13767c478bd9Sstevel@tonic-gate * - pathname.pn_pathlen does not count the ending null byte. 13777c478bd9Sstevel@tonic-gate */ 13787c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_READER, 0); 13797c478bd9Sstevel@tonic-gate oldpath = svp->sv_path; 13807c478bd9Sstevel@tonic-gate oldpathlen = svp->sv_pathlen; 13817c478bd9Sstevel@tonic-gate if (error = pn_get(oldpath, UIO_SYSSPACE, &oldpn)) { 13827c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 13837c478bd9Sstevel@tonic-gate kmem_free(symlink, strlen(symlink) + 1); 13847c478bd9Sstevel@tonic-gate return (error); 13857c478bd9Sstevel@tonic-gate } 13867c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 13877c478bd9Sstevel@tonic-gate pn_alloc(&newpn); 13887c478bd9Sstevel@tonic-gate 13897c478bd9Sstevel@tonic-gate /* 13907c478bd9Sstevel@tonic-gate * Skip over previous components from the oldpath so that the 13917c478bd9Sstevel@tonic-gate * oldpn.pn_path will point to the symlink component. Skip 13927c478bd9Sstevel@tonic-gate * leading slashes and "/./" (no OP_LOOKUP on ".") so that 13937c478bd9Sstevel@tonic-gate * pn_getcompnent can get the component. 13947c478bd9Sstevel@tonic-gate */ 13957c478bd9Sstevel@tonic-gate for (i = 1; i < nth; i++) { 13967c478bd9Sstevel@tonic-gate pathname_skipslashdot(&oldpn); 13977c478bd9Sstevel@tonic-gate error = pn_getcomponent(&oldpn, component); 13987c478bd9Sstevel@tonic-gate if (error) 13997c478bd9Sstevel@tonic-gate goto out; 14007c478bd9Sstevel@tonic-gate } 14017c478bd9Sstevel@tonic-gate 14027c478bd9Sstevel@tonic-gate /* 14037c478bd9Sstevel@tonic-gate * Copy the old path upto the component right before the symlink 14047c478bd9Sstevel@tonic-gate * if the symlink is not an absolute path. 14057c478bd9Sstevel@tonic-gate */ 14067c478bd9Sstevel@tonic-gate if (symlink[0] != '/') { 14077c478bd9Sstevel@tonic-gate addlen = oldpn.pn_path - oldpn.pn_buf; 14087c478bd9Sstevel@tonic-gate bcopy(oldpn.pn_buf, newpn.pn_path, addlen); 14097c478bd9Sstevel@tonic-gate newpn.pn_pathlen += addlen; 14107c478bd9Sstevel@tonic-gate newpn.pn_path += addlen; 14117c478bd9Sstevel@tonic-gate newpn.pn_buf[newpn.pn_pathlen] = '/'; 14127c478bd9Sstevel@tonic-gate newpn.pn_pathlen++; 14137c478bd9Sstevel@tonic-gate newpn.pn_path++; 14147c478bd9Sstevel@tonic-gate } 14157c478bd9Sstevel@tonic-gate 14167c478bd9Sstevel@tonic-gate /* copy the resolved symbolic link text */ 14177c478bd9Sstevel@tonic-gate addlen = strlen(symlink); 14187c478bd9Sstevel@tonic-gate if (newpn.pn_pathlen + addlen >= newpn.pn_bufsize) { 14197c478bd9Sstevel@tonic-gate error = ENAMETOOLONG; 14207c478bd9Sstevel@tonic-gate goto out; 14217c478bd9Sstevel@tonic-gate } 14227c478bd9Sstevel@tonic-gate bcopy(symlink, newpn.pn_path, addlen); 14237c478bd9Sstevel@tonic-gate newpn.pn_pathlen += addlen; 14247c478bd9Sstevel@tonic-gate newpn.pn_path += addlen; 14257c478bd9Sstevel@tonic-gate 14267c478bd9Sstevel@tonic-gate /* 14277c478bd9Sstevel@tonic-gate * Check if there is any remaining path after the symlink component. 14287c478bd9Sstevel@tonic-gate * First, skip the symlink component. 14297c478bd9Sstevel@tonic-gate */ 14307c478bd9Sstevel@tonic-gate pathname_skipslashdot(&oldpn); 14317c478bd9Sstevel@tonic-gate if (error = pn_getcomponent(&oldpn, component)) 14327c478bd9Sstevel@tonic-gate goto out; 14337c478bd9Sstevel@tonic-gate 14347c478bd9Sstevel@tonic-gate addlen = pn_pathleft(&oldpn); /* includes counting the slash */ 14357c478bd9Sstevel@tonic-gate 14367c478bd9Sstevel@tonic-gate /* 14377c478bd9Sstevel@tonic-gate * Copy the remaining path to the new pathname if there is any. 14387c478bd9Sstevel@tonic-gate */ 14397c478bd9Sstevel@tonic-gate if (addlen > 0) { 14407c478bd9Sstevel@tonic-gate if (newpn.pn_pathlen + addlen >= newpn.pn_bufsize) { 14417c478bd9Sstevel@tonic-gate error = ENAMETOOLONG; 14427c478bd9Sstevel@tonic-gate goto out; 14437c478bd9Sstevel@tonic-gate } 14447c478bd9Sstevel@tonic-gate bcopy(oldpn.pn_path, newpn.pn_path, addlen); 14457c478bd9Sstevel@tonic-gate newpn.pn_pathlen += addlen; 14467c478bd9Sstevel@tonic-gate } 14477c478bd9Sstevel@tonic-gate newpn.pn_buf[newpn.pn_pathlen] = '\0'; 14487c478bd9Sstevel@tonic-gate 14497c478bd9Sstevel@tonic-gate /* get the newpath and store it in the servinfo4_t */ 14507c478bd9Sstevel@tonic-gate newpath = kmem_alloc(newpn.pn_pathlen + 1, KM_SLEEP); 14517c478bd9Sstevel@tonic-gate bcopy(newpn.pn_buf, newpath, newpn.pn_pathlen); 14527c478bd9Sstevel@tonic-gate newpath[newpn.pn_pathlen] = '\0'; 14537c478bd9Sstevel@tonic-gate 14547c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_WRITER, 0); 14557c478bd9Sstevel@tonic-gate svp->sv_path = newpath; 14567c478bd9Sstevel@tonic-gate svp->sv_pathlen = strlen(newpath) + 1; 14577c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 14587c478bd9Sstevel@tonic-gate 14597c478bd9Sstevel@tonic-gate kmem_free(oldpath, oldpathlen); 14607c478bd9Sstevel@tonic-gate out: 14617c478bd9Sstevel@tonic-gate kmem_free(symlink, strlen(symlink) + 1); 14627c478bd9Sstevel@tonic-gate pn_free(&newpn); 14637c478bd9Sstevel@tonic-gate pn_free(&oldpn); 14647c478bd9Sstevel@tonic-gate 14657c478bd9Sstevel@tonic-gate return (error); 14667c478bd9Sstevel@tonic-gate } 14677c478bd9Sstevel@tonic-gate 14682f172c55SRobert Thurlow /* 14692f172c55SRobert Thurlow * This routine updates servinfo4 structure with the new referred server 14702f172c55SRobert Thurlow * info. 14712f172c55SRobert Thurlow * nfsfsloc has the location related information 14722f172c55SRobert Thurlow * fsp has the hostname and pathname info. 14732f172c55SRobert Thurlow * new path = pathname from referral + part of orig pathname(based on nth). 14742f172c55SRobert Thurlow */ 14752f172c55SRobert Thurlow static void 14762f172c55SRobert Thurlow update_servinfo4(servinfo4_t *svp, fs_location4 *fsp, 14772f172c55SRobert Thurlow struct nfs_fsl_info *nfsfsloc, char *orig_path, int nth) 14782f172c55SRobert Thurlow { 14792f172c55SRobert Thurlow struct knetconfig *knconf, *svknconf; 14802f172c55SRobert Thurlow struct netbuf *saddr; 14812f172c55SRobert Thurlow sec_data_t *secdata; 14822f172c55SRobert Thurlow utf8string *host; 14832f172c55SRobert Thurlow int i = 0, num_slashes = 0; 14842f172c55SRobert Thurlow char *p, *spath, *op, *new_path; 14852f172c55SRobert Thurlow 14862f172c55SRobert Thurlow /* Update knconf */ 14872f172c55SRobert Thurlow knconf = svp->sv_knconf; 14882f172c55SRobert Thurlow free_knconf_contents(knconf); 14892f172c55SRobert Thurlow bzero(knconf, sizeof (struct knetconfig)); 14902f172c55SRobert Thurlow svknconf = nfsfsloc->knconf; 14912f172c55SRobert Thurlow knconf->knc_semantics = svknconf->knc_semantics; 14922f172c55SRobert Thurlow knconf->knc_protofmly = kmem_zalloc(KNC_STRSIZE, KM_SLEEP); 14932f172c55SRobert Thurlow knconf->knc_proto = kmem_zalloc(KNC_STRSIZE, KM_SLEEP); 14942f172c55SRobert Thurlow knconf->knc_rdev = svknconf->knc_rdev; 14952f172c55SRobert Thurlow bcopy(svknconf->knc_protofmly, knconf->knc_protofmly, KNC_STRSIZE); 14962f172c55SRobert Thurlow bcopy(svknconf->knc_proto, knconf->knc_proto, KNC_STRSIZE); 14972f172c55SRobert Thurlow 14982f172c55SRobert Thurlow /* Update server address */ 14992f172c55SRobert Thurlow saddr = &svp->sv_addr; 15002f172c55SRobert Thurlow if (saddr->buf != NULL) 15012f172c55SRobert Thurlow kmem_free(saddr->buf, saddr->maxlen); 15022f172c55SRobert Thurlow saddr->buf = kmem_alloc(nfsfsloc->addr->maxlen, KM_SLEEP); 15032f172c55SRobert Thurlow saddr->len = nfsfsloc->addr->len; 15042f172c55SRobert Thurlow saddr->maxlen = nfsfsloc->addr->maxlen; 15052f172c55SRobert Thurlow bcopy(nfsfsloc->addr->buf, saddr->buf, nfsfsloc->addr->len); 15062f172c55SRobert Thurlow 15072f172c55SRobert Thurlow /* Update server name */ 15082f172c55SRobert Thurlow host = fsp->server_val; 15092f172c55SRobert Thurlow kmem_free(svp->sv_hostname, svp->sv_hostnamelen); 15102f172c55SRobert Thurlow svp->sv_hostname = kmem_zalloc(host->utf8string_len + 1, KM_SLEEP); 15112f172c55SRobert Thurlow bcopy(host->utf8string_val, svp->sv_hostname, host->utf8string_len); 15122f172c55SRobert Thurlow svp->sv_hostname[host->utf8string_len] = '\0'; 15132f172c55SRobert Thurlow svp->sv_hostnamelen = host->utf8string_len + 1; 15142f172c55SRobert Thurlow 15152f172c55SRobert Thurlow /* 15162f172c55SRobert Thurlow * Update server path. 15172f172c55SRobert Thurlow * We need to setup proper path here. 15182f172c55SRobert Thurlow * For ex., If we got a path name serv1:/rp/aaa/bbb 15192f172c55SRobert Thurlow * where aaa is a referral and points to serv2:/rpool/aa 15202f172c55SRobert Thurlow * we need to set the path to serv2:/rpool/aa/bbb 15212f172c55SRobert Thurlow * The first part of this below code generates /rpool/aa 15222f172c55SRobert Thurlow * and the second part appends /bbb to the server path. 15232f172c55SRobert Thurlow */ 15242f172c55SRobert Thurlow spath = p = kmem_zalloc(MAXPATHLEN, KM_SLEEP); 15252f172c55SRobert Thurlow *p++ = '/'; 15262f172c55SRobert Thurlow for (i = 0; i < fsp->rootpath.pathname4_len; i++) { 15272f172c55SRobert Thurlow component4 *comp; 15282f172c55SRobert Thurlow 15292f172c55SRobert Thurlow comp = &fsp->rootpath.pathname4_val[i]; 15302f172c55SRobert Thurlow /* If no space, null the string and bail */ 15312f172c55SRobert Thurlow if ((p - spath) + comp->utf8string_len + 1 > MAXPATHLEN) { 15322f172c55SRobert Thurlow p = spath + MAXPATHLEN - 1; 15332f172c55SRobert Thurlow spath[0] = '\0'; 15342f172c55SRobert Thurlow break; 15352f172c55SRobert Thurlow } 15362f172c55SRobert Thurlow bcopy(comp->utf8string_val, p, comp->utf8string_len); 15372f172c55SRobert Thurlow p += comp->utf8string_len; 15382f172c55SRobert Thurlow *p++ = '/'; 15392f172c55SRobert Thurlow } 15402f172c55SRobert Thurlow if (fsp->rootpath.pathname4_len != 0) 15412f172c55SRobert Thurlow *(p - 1) = '\0'; 15422f172c55SRobert Thurlow else 15432f172c55SRobert Thurlow *p = '\0'; 15442f172c55SRobert Thurlow p = spath; 15452f172c55SRobert Thurlow 15462f172c55SRobert Thurlow new_path = kmem_zalloc(MAXPATHLEN, KM_SLEEP); 15472f172c55SRobert Thurlow (void) strlcpy(new_path, p, MAXPATHLEN); 15482f172c55SRobert Thurlow kmem_free(p, MAXPATHLEN); 15492f172c55SRobert Thurlow i = strlen(new_path); 15502f172c55SRobert Thurlow 15512f172c55SRobert Thurlow for (op = orig_path; *op; op++) { 15522f172c55SRobert Thurlow if (*op == '/') 15532f172c55SRobert Thurlow num_slashes++; 15542f172c55SRobert Thurlow if (num_slashes == nth + 2) { 15552f172c55SRobert Thurlow while (*op != '\0') { 15562f172c55SRobert Thurlow new_path[i] = *op; 15572f172c55SRobert Thurlow i++; 15582f172c55SRobert Thurlow op++; 15592f172c55SRobert Thurlow } 15602f172c55SRobert Thurlow break; 15612f172c55SRobert Thurlow } 15622f172c55SRobert Thurlow } 15632f172c55SRobert Thurlow new_path[i] = '\0'; 15642f172c55SRobert Thurlow 15652f172c55SRobert Thurlow kmem_free(svp->sv_path, svp->sv_pathlen); 15662f172c55SRobert Thurlow svp->sv_pathlen = strlen(new_path) + 1; 15672f172c55SRobert Thurlow svp->sv_path = kmem_alloc(svp->sv_pathlen, KM_SLEEP); 15682f172c55SRobert Thurlow bcopy(new_path, svp->sv_path, svp->sv_pathlen); 15692f172c55SRobert Thurlow kmem_free(new_path, MAXPATHLEN); 15702f172c55SRobert Thurlow 15712f172c55SRobert Thurlow /* 15722f172c55SRobert Thurlow * All the security data is specific to old server. 15732f172c55SRobert Thurlow * Clean it up except secdata which deals with mount options. 15742f172c55SRobert Thurlow * We need to inherit that data. Copy secdata into our new servinfo4. 15752f172c55SRobert Thurlow */ 15762f172c55SRobert Thurlow if (svp->sv_dhsec) { 15772f172c55SRobert Thurlow sec_clnt_freeinfo(svp->sv_dhsec); 15782f172c55SRobert Thurlow svp->sv_dhsec = NULL; 15792f172c55SRobert Thurlow } 15802f172c55SRobert Thurlow if (svp->sv_save_secinfo && 15812f172c55SRobert Thurlow svp->sv_save_secinfo != svp->sv_secinfo) { 15822f172c55SRobert Thurlow secinfo_free(svp->sv_save_secinfo); 15832f172c55SRobert Thurlow svp->sv_save_secinfo = NULL; 15842f172c55SRobert Thurlow } 15852f172c55SRobert Thurlow if (svp->sv_secinfo) { 15862f172c55SRobert Thurlow secinfo_free(svp->sv_secinfo); 15872f172c55SRobert Thurlow svp->sv_secinfo = NULL; 15882f172c55SRobert Thurlow } 15892f172c55SRobert Thurlow svp->sv_currsec = NULL; 15902f172c55SRobert Thurlow 15912f172c55SRobert Thurlow secdata = kmem_alloc(sizeof (*secdata), KM_SLEEP); 15922f172c55SRobert Thurlow *secdata = *svp->sv_secdata; 15932f172c55SRobert Thurlow secdata->data = NULL; 15942f172c55SRobert Thurlow if (svp->sv_secdata) { 15952f172c55SRobert Thurlow sec_clnt_freeinfo(svp->sv_secdata); 15962f172c55SRobert Thurlow svp->sv_secdata = NULL; 15972f172c55SRobert Thurlow } 15982f172c55SRobert Thurlow svp->sv_secdata = secdata; 15992f172c55SRobert Thurlow } 16002f172c55SRobert Thurlow 16012f172c55SRobert Thurlow /* 16022f172c55SRobert Thurlow * Resolve a referral. The referral is in the n+1th component of 16032f172c55SRobert Thurlow * svp->sv_path and has a parent nfs4 file handle "fh". 16042f172c55SRobert Thurlow * Upon return, the sv_path will point to the new path that has referral 16052f172c55SRobert Thurlow * component resolved to its referred path and part of original path. 16062f172c55SRobert Thurlow * Hostname and other address information is also updated. 16072f172c55SRobert Thurlow */ 16082f172c55SRobert Thurlow int 16092f172c55SRobert Thurlow resolve_referral(mntinfo4_t *mi, servinfo4_t *svp, cred_t *cr, int nth, 16102f172c55SRobert Thurlow nfs_fh4 *fh) 16112f172c55SRobert Thurlow { 16122f172c55SRobert Thurlow nfs4_sharedfh_t *sfh; 16132f172c55SRobert Thurlow struct nfs_fsl_info nfsfsloc; 16142f172c55SRobert Thurlow nfs4_ga_res_t garp; 16152f172c55SRobert Thurlow COMPOUND4res_clnt callres; 16162f172c55SRobert Thurlow fs_location4 *fsp; 16172f172c55SRobert Thurlow char *nm, *orig_path; 16182f172c55SRobert Thurlow int orig_pathlen = 0, ret = -1, index; 16192f172c55SRobert Thurlow 16202f172c55SRobert Thurlow if (svp->sv_pathlen <= 0) 16212f172c55SRobert Thurlow return (ret); 16222f172c55SRobert Thurlow 16232f172c55SRobert Thurlow (void) nfs_rw_enter_sig(&svp->sv_lock, RW_WRITER, 0); 16242f172c55SRobert Thurlow orig_pathlen = svp->sv_pathlen; 16252f172c55SRobert Thurlow orig_path = kmem_alloc(orig_pathlen, KM_SLEEP); 16262f172c55SRobert Thurlow bcopy(svp->sv_path, orig_path, orig_pathlen); 16272f172c55SRobert Thurlow nm = extract_referral_point(svp->sv_path, nth); 16282f172c55SRobert Thurlow setup_newsvpath(svp, nth); 16292f172c55SRobert Thurlow nfs_rw_exit(&svp->sv_lock); 16302f172c55SRobert Thurlow 16312f172c55SRobert Thurlow sfh = sfh4_get(fh, mi); 16322f172c55SRobert Thurlow index = nfs4_process_referral(mi, sfh, nm, cr, 16332f172c55SRobert Thurlow &garp, &callres, &nfsfsloc); 16342f172c55SRobert Thurlow sfh4_rele(&sfh); 16352f172c55SRobert Thurlow kmem_free(nm, MAXPATHLEN); 16362f172c55SRobert Thurlow if (index < 0) { 16372f172c55SRobert Thurlow kmem_free(orig_path, orig_pathlen); 16382f172c55SRobert Thurlow return (index); 16392f172c55SRobert Thurlow } 16402f172c55SRobert Thurlow 16412f172c55SRobert Thurlow fsp = &garp.n4g_ext_res->n4g_fslocations.locations_val[index]; 16422f172c55SRobert Thurlow (void) nfs_rw_enter_sig(&svp->sv_lock, RW_WRITER, 0); 16432f172c55SRobert Thurlow update_servinfo4(svp, fsp, &nfsfsloc, orig_path, nth); 16442f172c55SRobert Thurlow nfs_rw_exit(&svp->sv_lock); 16452f172c55SRobert Thurlow 16462f172c55SRobert Thurlow mutex_enter(&mi->mi_lock); 16472f172c55SRobert Thurlow mi->mi_vfs_referral_loop_cnt++; 16482f172c55SRobert Thurlow mutex_exit(&mi->mi_lock); 16492f172c55SRobert Thurlow 16502f172c55SRobert Thurlow ret = 0; 16512f172c55SRobert Thurlow bad: 16522f172c55SRobert Thurlow /* Free up XDR memory allocated in nfs4_process_referral() */ 16532f172c55SRobert Thurlow xdr_free(xdr_nfs_fsl_info, (char *)&nfsfsloc); 16542f172c55SRobert Thurlow xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&callres); 16552f172c55SRobert Thurlow kmem_free(orig_path, orig_pathlen); 16562f172c55SRobert Thurlow 16572f172c55SRobert Thurlow return (ret); 16582f172c55SRobert Thurlow } 16592f172c55SRobert Thurlow 16607c478bd9Sstevel@tonic-gate /* 16617c478bd9Sstevel@tonic-gate * Get the root filehandle for the given filesystem and server, and update 16627c478bd9Sstevel@tonic-gate * svp. 16637c478bd9Sstevel@tonic-gate * 16647c478bd9Sstevel@tonic-gate * If NFS4_GETFH_NEEDSOP is set, then use nfs4_start_fop and nfs4_end_fop 16657c478bd9Sstevel@tonic-gate * to coordinate with recovery. Otherwise, the caller is assumed to be 16667c478bd9Sstevel@tonic-gate * the recovery thread or have already done a start_fop. 16677c478bd9Sstevel@tonic-gate * 16687c478bd9Sstevel@tonic-gate * Errors are returned by the nfs4_error_t parameter. 16697c478bd9Sstevel@tonic-gate */ 16707c478bd9Sstevel@tonic-gate static void 16717c478bd9Sstevel@tonic-gate nfs4getfh_otw(struct mntinfo4 *mi, servinfo4_t *svp, vtype_t *vtp, 1672b9238976Sth int flags, cred_t *cr, nfs4_error_t *ep) 16737c478bd9Sstevel@tonic-gate { 16747c478bd9Sstevel@tonic-gate COMPOUND4args_clnt args; 16757c478bd9Sstevel@tonic-gate COMPOUND4res_clnt res; 16767c478bd9Sstevel@tonic-gate int doqueue = 1; 16777c478bd9Sstevel@tonic-gate nfs_argop4 *argop; 16787c478bd9Sstevel@tonic-gate nfs_resop4 *resop; 16797c478bd9Sstevel@tonic-gate nfs4_ga_res_t *garp; 16807c478bd9Sstevel@tonic-gate int num_argops; 16817c478bd9Sstevel@tonic-gate lookup4_param_t lookuparg; 16827c478bd9Sstevel@tonic-gate nfs_fh4 *tmpfhp; 16837c478bd9Sstevel@tonic-gate nfs_fh4 *resfhp; 16847c478bd9Sstevel@tonic-gate bool_t needrecov = FALSE; 16857c478bd9Sstevel@tonic-gate nfs4_recov_state_t recov_state; 16867c478bd9Sstevel@tonic-gate int llndx; 16877c478bd9Sstevel@tonic-gate int nthcomp; 16887c478bd9Sstevel@tonic-gate int recovery = !(flags & NFS4_GETFH_NEEDSOP); 16897c478bd9Sstevel@tonic-gate 16907c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_READER, 0); 16917c478bd9Sstevel@tonic-gate ASSERT(svp->sv_path != NULL); 16927c478bd9Sstevel@tonic-gate if (svp->sv_path[0] == '\0') { 16937c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 16947c478bd9Sstevel@tonic-gate nfs4_error_init(ep, EINVAL); 16957c478bd9Sstevel@tonic-gate return; 16967c478bd9Sstevel@tonic-gate } 16977c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 16987c478bd9Sstevel@tonic-gate 16997c478bd9Sstevel@tonic-gate recov_state.rs_flags = 0; 17007c478bd9Sstevel@tonic-gate recov_state.rs_num_retry_despite_err = 0; 17012f172c55SRobert Thurlow 17027c478bd9Sstevel@tonic-gate recov_retry: 17032f172c55SRobert Thurlow if (mi->mi_vfs_referral_loop_cnt >= NFS4_REFERRAL_LOOP_MAX) { 17042f172c55SRobert Thurlow DTRACE_PROBE3(nfs4clnt__debug__referral__loop, mntinfo4 *, 17052f172c55SRobert Thurlow mi, servinfo4_t *, svp, char *, "nfs4getfh_otw"); 17062f172c55SRobert Thurlow nfs4_error_init(ep, EINVAL); 17072f172c55SRobert Thurlow return; 17082f172c55SRobert Thurlow } 17097c478bd9Sstevel@tonic-gate nfs4_error_zinit(ep); 17107c478bd9Sstevel@tonic-gate 17117c478bd9Sstevel@tonic-gate if (!recovery) { 17127c478bd9Sstevel@tonic-gate ep->error = nfs4_start_fop(mi, NULL, NULL, OH_MOUNT, 1713b9238976Sth &recov_state, NULL); 17147c478bd9Sstevel@tonic-gate 17157c478bd9Sstevel@tonic-gate /* 17167c478bd9Sstevel@tonic-gate * If recovery has been started and this request as 17177c478bd9Sstevel@tonic-gate * initiated by a mount, then we must wait for recovery 17187c478bd9Sstevel@tonic-gate * to finish before proceeding, otherwise, the error 17197c478bd9Sstevel@tonic-gate * cleanup would remove data structures needed by the 17207c478bd9Sstevel@tonic-gate * recovery thread. 17217c478bd9Sstevel@tonic-gate */ 17227c478bd9Sstevel@tonic-gate if (ep->error) { 17237c478bd9Sstevel@tonic-gate mutex_enter(&mi->mi_lock); 17247c478bd9Sstevel@tonic-gate if (mi->mi_flags & MI4_MOUNTING) { 17257c478bd9Sstevel@tonic-gate mi->mi_flags |= MI4_RECOV_FAIL; 17267c478bd9Sstevel@tonic-gate mi->mi_error = EIO; 17277c478bd9Sstevel@tonic-gate 17287c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_recov_debug, (CE_NOTE, 17297c478bd9Sstevel@tonic-gate "nfs4getfh_otw: waiting 4 recovery\n")); 17307c478bd9Sstevel@tonic-gate 17317c478bd9Sstevel@tonic-gate while (mi->mi_flags & MI4_RECOV_ACTIV) 17327c478bd9Sstevel@tonic-gate cv_wait(&mi->mi_failover_cv, 17337c478bd9Sstevel@tonic-gate &mi->mi_lock); 17347c478bd9Sstevel@tonic-gate } 17357c478bd9Sstevel@tonic-gate mutex_exit(&mi->mi_lock); 17367c478bd9Sstevel@tonic-gate return; 17377c478bd9Sstevel@tonic-gate } 17387c478bd9Sstevel@tonic-gate 17397c478bd9Sstevel@tonic-gate /* 17407c478bd9Sstevel@tonic-gate * If the client does not specify a specific flavor to use 17417c478bd9Sstevel@tonic-gate * and has not gotten a secinfo list from the server yet, 17427c478bd9Sstevel@tonic-gate * retrieve the secinfo list from the server and use a 17437c478bd9Sstevel@tonic-gate * flavor from the list to mount. 17447c478bd9Sstevel@tonic-gate * 17457c478bd9Sstevel@tonic-gate * If fail to get the secinfo list from the server, then 17467c478bd9Sstevel@tonic-gate * try the default flavor. 17477c478bd9Sstevel@tonic-gate */ 17487c478bd9Sstevel@tonic-gate if ((svp->sv_flags & SV4_TRYSECDEFAULT) && 17497c478bd9Sstevel@tonic-gate svp->sv_secinfo == NULL) { 17507c478bd9Sstevel@tonic-gate (void) nfs4_secinfo_path(mi, cr, FALSE); 17517c478bd9Sstevel@tonic-gate } 17527c478bd9Sstevel@tonic-gate } 17537c478bd9Sstevel@tonic-gate 17547c478bd9Sstevel@tonic-gate if (recovery) 17557c478bd9Sstevel@tonic-gate args.ctag = TAG_REMAP_MOUNT; 17567c478bd9Sstevel@tonic-gate else 17577c478bd9Sstevel@tonic-gate args.ctag = TAG_MOUNT; 17587c478bd9Sstevel@tonic-gate 17597c478bd9Sstevel@tonic-gate lookuparg.l4_getattrs = LKP4_ALL_ATTRIBUTES; 17607c478bd9Sstevel@tonic-gate lookuparg.argsp = &args; 17617c478bd9Sstevel@tonic-gate lookuparg.resp = &res; 17627c478bd9Sstevel@tonic-gate lookuparg.header_len = 2; /* Putrootfh, getfh */ 17637c478bd9Sstevel@tonic-gate lookuparg.trailer_len = 0; 17647c478bd9Sstevel@tonic-gate lookuparg.ga_bits = FATTR4_FSINFO_MASK; 17657c478bd9Sstevel@tonic-gate lookuparg.mi = mi; 17667c478bd9Sstevel@tonic-gate 17677c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_READER, 0); 17687c478bd9Sstevel@tonic-gate ASSERT(svp->sv_path != NULL); 17697c478bd9Sstevel@tonic-gate llndx = nfs4lookup_setup(svp->sv_path, &lookuparg, 0); 17707c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 17717c478bd9Sstevel@tonic-gate 17727c478bd9Sstevel@tonic-gate argop = args.array; 17737c478bd9Sstevel@tonic-gate num_argops = args.array_len; 17747c478bd9Sstevel@tonic-gate 17757c478bd9Sstevel@tonic-gate /* choose public or root filehandle */ 17767c478bd9Sstevel@tonic-gate if (flags & NFS4_GETFH_PUBLIC) 17777c478bd9Sstevel@tonic-gate argop[0].argop = OP_PUTPUBFH; 17787c478bd9Sstevel@tonic-gate else 17797c478bd9Sstevel@tonic-gate argop[0].argop = OP_PUTROOTFH; 17807c478bd9Sstevel@tonic-gate 17817c478bd9Sstevel@tonic-gate /* get fh */ 17827c478bd9Sstevel@tonic-gate argop[1].argop = OP_GETFH; 17837c478bd9Sstevel@tonic-gate 17847c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_call_debug, (CE_NOTE, 17857c478bd9Sstevel@tonic-gate "nfs4getfh_otw: %s call, mi 0x%p", 17867c478bd9Sstevel@tonic-gate needrecov ? "recov" : "first", (void *)mi)); 17877c478bd9Sstevel@tonic-gate 17887c478bd9Sstevel@tonic-gate rfs4call(mi, &args, &res, cr, &doqueue, RFSCALL_SOFT, ep); 17897c478bd9Sstevel@tonic-gate 17907c478bd9Sstevel@tonic-gate needrecov = nfs4_needs_recovery(ep, FALSE, mi->mi_vfsp); 17917c478bd9Sstevel@tonic-gate 17927c478bd9Sstevel@tonic-gate if (needrecov) { 17937c478bd9Sstevel@tonic-gate bool_t abort; 17947c478bd9Sstevel@tonic-gate 17957c478bd9Sstevel@tonic-gate if (recovery) { 17967c478bd9Sstevel@tonic-gate nfs4args_lookup_free(argop, num_argops); 17977c478bd9Sstevel@tonic-gate kmem_free(argop, 1798b9238976Sth lookuparg.arglen * sizeof (nfs_argop4)); 17997c478bd9Sstevel@tonic-gate if (!ep->error) 18007c478bd9Sstevel@tonic-gate (void) xdr_free(xdr_COMPOUND4res_clnt, 1801b9238976Sth (caddr_t)&res); 18027c478bd9Sstevel@tonic-gate return; 18037c478bd9Sstevel@tonic-gate } 18047c478bd9Sstevel@tonic-gate 18057c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_recov_debug, 18067c478bd9Sstevel@tonic-gate (CE_NOTE, "nfs4getfh_otw: initiating recovery\n")); 18077c478bd9Sstevel@tonic-gate 18087c478bd9Sstevel@tonic-gate abort = nfs4_start_recovery(ep, mi, NULL, 18092f172c55SRobert Thurlow NULL, NULL, NULL, OP_GETFH, NULL, NULL, NULL); 18107c478bd9Sstevel@tonic-gate if (!ep->error) { 18117c478bd9Sstevel@tonic-gate ep->error = geterrno4(res.status); 18127c478bd9Sstevel@tonic-gate (void) xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&res); 18137c478bd9Sstevel@tonic-gate } 18147c478bd9Sstevel@tonic-gate nfs4args_lookup_free(argop, num_argops); 18157c478bd9Sstevel@tonic-gate kmem_free(argop, lookuparg.arglen * sizeof (nfs_argop4)); 18167c478bd9Sstevel@tonic-gate nfs4_end_fop(mi, NULL, NULL, OH_MOUNT, &recov_state, needrecov); 18177c478bd9Sstevel@tonic-gate /* have another go? */ 18187c478bd9Sstevel@tonic-gate if (abort == FALSE) 18197c478bd9Sstevel@tonic-gate goto recov_retry; 18207c478bd9Sstevel@tonic-gate return; 18217c478bd9Sstevel@tonic-gate } 18227c478bd9Sstevel@tonic-gate 18237c478bd9Sstevel@tonic-gate /* 18247c478bd9Sstevel@tonic-gate * No recovery, but check if error is set. 18257c478bd9Sstevel@tonic-gate */ 18267c478bd9Sstevel@tonic-gate if (ep->error) { 18277c478bd9Sstevel@tonic-gate nfs4args_lookup_free(argop, num_argops); 18287c478bd9Sstevel@tonic-gate kmem_free(argop, lookuparg.arglen * sizeof (nfs_argop4)); 18297c478bd9Sstevel@tonic-gate if (!recovery) 18307c478bd9Sstevel@tonic-gate nfs4_end_fop(mi, NULL, NULL, OH_MOUNT, &recov_state, 1831b9238976Sth needrecov); 18327c478bd9Sstevel@tonic-gate return; 18337c478bd9Sstevel@tonic-gate } 18347c478bd9Sstevel@tonic-gate 18357c478bd9Sstevel@tonic-gate is_link_err: 18367c478bd9Sstevel@tonic-gate 18377c478bd9Sstevel@tonic-gate /* for non-recovery errors */ 18382f172c55SRobert Thurlow if (res.status && res.status != NFS4ERR_SYMLINK && 18392f172c55SRobert Thurlow res.status != NFS4ERR_MOVED) { 18407c478bd9Sstevel@tonic-gate if (!recovery) { 18417c478bd9Sstevel@tonic-gate nfs4_end_fop(mi, NULL, NULL, OH_MOUNT, &recov_state, 1842b9238976Sth needrecov); 18437c478bd9Sstevel@tonic-gate } 18447c478bd9Sstevel@tonic-gate nfs4args_lookup_free(argop, num_argops); 18457c478bd9Sstevel@tonic-gate kmem_free(argop, lookuparg.arglen * sizeof (nfs_argop4)); 18467c478bd9Sstevel@tonic-gate (void) xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&res); 18477c478bd9Sstevel@tonic-gate return; 18487c478bd9Sstevel@tonic-gate } 18497c478bd9Sstevel@tonic-gate 18507c478bd9Sstevel@tonic-gate /* 18517c478bd9Sstevel@tonic-gate * If any intermediate component in the path is a symbolic link, 18527c478bd9Sstevel@tonic-gate * resolve the symlink, then try mount again using the new path. 18537c478bd9Sstevel@tonic-gate */ 18542f172c55SRobert Thurlow if (res.status == NFS4ERR_SYMLINK || res.status == NFS4ERR_MOVED) { 18557c478bd9Sstevel@tonic-gate int where; 18567c478bd9Sstevel@tonic-gate 18572f172c55SRobert Thurlow /* 18582f172c55SRobert Thurlow * Need to call nfs4_end_op before resolve_sympath to avoid 18592f172c55SRobert Thurlow * potential nfs4_start_op deadlock. 18602f172c55SRobert Thurlow */ 18612f172c55SRobert Thurlow if (!recovery) 18622f172c55SRobert Thurlow nfs4_end_fop(mi, NULL, NULL, OH_MOUNT, &recov_state, 18632f172c55SRobert Thurlow needrecov); 18642f172c55SRobert Thurlow 18657c478bd9Sstevel@tonic-gate /* 18667c478bd9Sstevel@tonic-gate * This must be from OP_LOOKUP failure. The (cfh) for this 18677c478bd9Sstevel@tonic-gate * OP_LOOKUP is a symlink node. Found out where the 18687c478bd9Sstevel@tonic-gate * OP_GETFH is for the (cfh) that is a symlink node. 18697c478bd9Sstevel@tonic-gate * 18707c478bd9Sstevel@tonic-gate * Example: 18717c478bd9Sstevel@tonic-gate * (mount) PUTROOTFH, GETFH, LOOKUP comp1, GETFH, GETATTR, 18727c478bd9Sstevel@tonic-gate * LOOKUP comp2, GETFH, GETATTR, LOOKUP comp3, GETFH, GETATTR 18737c478bd9Sstevel@tonic-gate * 18747c478bd9Sstevel@tonic-gate * LOOKUP comp3 fails with SYMLINK because comp2 is a symlink. 18757c478bd9Sstevel@tonic-gate * In this case, where = 7, nthcomp = 2. 18767c478bd9Sstevel@tonic-gate */ 18777c478bd9Sstevel@tonic-gate where = res.array_len - 2; 18787c478bd9Sstevel@tonic-gate ASSERT(where > 0); 18797c478bd9Sstevel@tonic-gate 18802f172c55SRobert Thurlow if (res.status == NFS4ERR_SYMLINK) { 18817c478bd9Sstevel@tonic-gate 18822f172c55SRobert Thurlow resop = &res.array[where - 1]; 18832f172c55SRobert Thurlow ASSERT(resop->resop == OP_GETFH); 18842f172c55SRobert Thurlow tmpfhp = &resop->nfs_resop4_u.opgetfh.object; 18852f172c55SRobert Thurlow nthcomp = res.array_len/3 - 1; 18862f172c55SRobert Thurlow ep->error = resolve_sympath(mi, svp, nthcomp, 18872f172c55SRobert Thurlow tmpfhp, cr, flags); 18887c478bd9Sstevel@tonic-gate 18892f172c55SRobert Thurlow } else if (res.status == NFS4ERR_MOVED) { 18902f172c55SRobert Thurlow 18912f172c55SRobert Thurlow resop = &res.array[where - 2]; 18922f172c55SRobert Thurlow ASSERT(resop->resop == OP_GETFH); 18932f172c55SRobert Thurlow tmpfhp = &resop->nfs_resop4_u.opgetfh.object; 18942f172c55SRobert Thurlow nthcomp = res.array_len/3 - 1; 18952f172c55SRobert Thurlow ep->error = resolve_referral(mi, svp, cr, nthcomp, 18962f172c55SRobert Thurlow tmpfhp); 18972f172c55SRobert Thurlow } 18987c478bd9Sstevel@tonic-gate 18997c478bd9Sstevel@tonic-gate nfs4args_lookup_free(argop, num_argops); 19007c478bd9Sstevel@tonic-gate kmem_free(argop, lookuparg.arglen * sizeof (nfs_argop4)); 19017c478bd9Sstevel@tonic-gate (void) xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&res); 19027c478bd9Sstevel@tonic-gate 19037c478bd9Sstevel@tonic-gate if (ep->error) 19047c478bd9Sstevel@tonic-gate return; 19057c478bd9Sstevel@tonic-gate 19067c478bd9Sstevel@tonic-gate goto recov_retry; 19077c478bd9Sstevel@tonic-gate } 19087c478bd9Sstevel@tonic-gate 19097c478bd9Sstevel@tonic-gate /* getfh */ 19107c478bd9Sstevel@tonic-gate resop = &res.array[res.array_len - 2]; 19117c478bd9Sstevel@tonic-gate ASSERT(resop->resop == OP_GETFH); 19127c478bd9Sstevel@tonic-gate resfhp = &resop->nfs_resop4_u.opgetfh.object; 19137c478bd9Sstevel@tonic-gate 19147c478bd9Sstevel@tonic-gate /* getattr fsinfo res */ 19157c478bd9Sstevel@tonic-gate resop++; 19167c478bd9Sstevel@tonic-gate garp = &resop->nfs_resop4_u.opgetattr.ga_res; 19177c478bd9Sstevel@tonic-gate 19187c478bd9Sstevel@tonic-gate *vtp = garp->n4g_va.va_type; 19197c478bd9Sstevel@tonic-gate 19207c478bd9Sstevel@tonic-gate mi->mi_fh_expire_type = garp->n4g_ext_res->n4g_fet; 19217c478bd9Sstevel@tonic-gate 19227c478bd9Sstevel@tonic-gate mutex_enter(&mi->mi_lock); 19237c478bd9Sstevel@tonic-gate if (garp->n4g_ext_res->n4g_pc4.pc4_link_support) 19247c478bd9Sstevel@tonic-gate mi->mi_flags |= MI4_LINK; 19257c478bd9Sstevel@tonic-gate if (garp->n4g_ext_res->n4g_pc4.pc4_symlink_support) 19267c478bd9Sstevel@tonic-gate mi->mi_flags |= MI4_SYMLINK; 19277c478bd9Sstevel@tonic-gate if (garp->n4g_ext_res->n4g_suppattrs & FATTR4_ACL_MASK) 19287c478bd9Sstevel@tonic-gate mi->mi_flags |= MI4_ACL; 19297c478bd9Sstevel@tonic-gate mutex_exit(&mi->mi_lock); 19307c478bd9Sstevel@tonic-gate 19317c478bd9Sstevel@tonic-gate if (garp->n4g_ext_res->n4g_maxread == 0) 19327c478bd9Sstevel@tonic-gate mi->mi_tsize = 1933b9238976Sth MIN(MAXBSIZE, mi->mi_tsize); 19347c478bd9Sstevel@tonic-gate else 19357c478bd9Sstevel@tonic-gate mi->mi_tsize = 1936b9238976Sth MIN(garp->n4g_ext_res->n4g_maxread, 1937b9238976Sth mi->mi_tsize); 19387c478bd9Sstevel@tonic-gate 19397c478bd9Sstevel@tonic-gate if (garp->n4g_ext_res->n4g_maxwrite == 0) 19407c478bd9Sstevel@tonic-gate mi->mi_stsize = 1941b9238976Sth MIN(MAXBSIZE, mi->mi_stsize); 19427c478bd9Sstevel@tonic-gate else 19437c478bd9Sstevel@tonic-gate mi->mi_stsize = 1944b9238976Sth MIN(garp->n4g_ext_res->n4g_maxwrite, 1945b9238976Sth mi->mi_stsize); 19467c478bd9Sstevel@tonic-gate 19477c478bd9Sstevel@tonic-gate if (garp->n4g_ext_res->n4g_maxfilesize != 0) 19487c478bd9Sstevel@tonic-gate mi->mi_maxfilesize = 1949b9238976Sth MIN(garp->n4g_ext_res->n4g_maxfilesize, 1950b9238976Sth mi->mi_maxfilesize); 19517c478bd9Sstevel@tonic-gate 19527c478bd9Sstevel@tonic-gate /* 19537c478bd9Sstevel@tonic-gate * If the final component is a a symbolic link, resolve the symlink, 19547c478bd9Sstevel@tonic-gate * then try mount again using the new path. 19557c478bd9Sstevel@tonic-gate * 19567c478bd9Sstevel@tonic-gate * Assume no symbolic link for root filesysm "/". 19577c478bd9Sstevel@tonic-gate */ 19587c478bd9Sstevel@tonic-gate if (*vtp == VLNK) { 19597c478bd9Sstevel@tonic-gate /* 19607c478bd9Sstevel@tonic-gate * nthcomp is the total result length minus 19617c478bd9Sstevel@tonic-gate * the 1st 2 OPs (PUTROOTFH, GETFH), 19627c478bd9Sstevel@tonic-gate * then divided by 3 (LOOKUP,GETFH,GETATTR) 19637c478bd9Sstevel@tonic-gate * 19647c478bd9Sstevel@tonic-gate * e.g. PUTROOTFH GETFH LOOKUP 1st-comp GETFH GETATTR 19657c478bd9Sstevel@tonic-gate * LOOKUP 2nd-comp GETFH GETATTR 19667c478bd9Sstevel@tonic-gate * 19677c478bd9Sstevel@tonic-gate * (8 - 2)/3 = 2 19687c478bd9Sstevel@tonic-gate */ 19697c478bd9Sstevel@tonic-gate nthcomp = (res.array_len - 2)/3; 19707c478bd9Sstevel@tonic-gate 19717c478bd9Sstevel@tonic-gate /* 19727c478bd9Sstevel@tonic-gate * Need to call nfs4_end_op before resolve_sympath to avoid 19737c478bd9Sstevel@tonic-gate * potential nfs4_start_op deadlock. See RFE 4777612. 19747c478bd9Sstevel@tonic-gate */ 19757c478bd9Sstevel@tonic-gate if (!recovery) 19767c478bd9Sstevel@tonic-gate nfs4_end_fop(mi, NULL, NULL, OH_MOUNT, &recov_state, 1977b9238976Sth needrecov); 19787c478bd9Sstevel@tonic-gate 19797c478bd9Sstevel@tonic-gate ep->error = resolve_sympath(mi, svp, nthcomp, resfhp, cr, 1980b9238976Sth flags); 19817c478bd9Sstevel@tonic-gate 19827c478bd9Sstevel@tonic-gate nfs4args_lookup_free(argop, num_argops); 19837c478bd9Sstevel@tonic-gate kmem_free(argop, lookuparg.arglen * sizeof (nfs_argop4)); 19847c478bd9Sstevel@tonic-gate (void) xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&res); 19857c478bd9Sstevel@tonic-gate 19867c478bd9Sstevel@tonic-gate if (ep->error) 19877c478bd9Sstevel@tonic-gate return; 19887c478bd9Sstevel@tonic-gate 19897c478bd9Sstevel@tonic-gate goto recov_retry; 19907c478bd9Sstevel@tonic-gate } 19917c478bd9Sstevel@tonic-gate 19927c478bd9Sstevel@tonic-gate /* 19937c478bd9Sstevel@tonic-gate * We need to figure out where in the compound the getfh 19947c478bd9Sstevel@tonic-gate * for the parent directory is. If the object to be mounted is 19957c478bd9Sstevel@tonic-gate * the root, then there is no lookup at all: 19967c478bd9Sstevel@tonic-gate * PUTROOTFH, GETFH. 19977c478bd9Sstevel@tonic-gate * If the object to be mounted is in the root, then the compound is: 19987c478bd9Sstevel@tonic-gate * PUTROOTFH, GETFH, LOOKUP, GETFH, GETATTR. 19997c478bd9Sstevel@tonic-gate * In either of these cases, the index of the GETFH is 1. 20007c478bd9Sstevel@tonic-gate * If it is not at the root, then it's something like: 20017c478bd9Sstevel@tonic-gate * PUTROOTFH, GETFH, LOOKUP, GETFH, GETATTR, 20027c478bd9Sstevel@tonic-gate * LOOKUP, GETFH, GETATTR 20037c478bd9Sstevel@tonic-gate * In this case, the index is llndx (last lookup index) - 2. 20047c478bd9Sstevel@tonic-gate */ 20057c478bd9Sstevel@tonic-gate if (llndx == -1 || llndx == 2) 20067c478bd9Sstevel@tonic-gate resop = &res.array[1]; 20077c478bd9Sstevel@tonic-gate else { 20087c478bd9Sstevel@tonic-gate ASSERT(llndx > 2); 20097c478bd9Sstevel@tonic-gate resop = &res.array[llndx-2]; 20107c478bd9Sstevel@tonic-gate } 20117c478bd9Sstevel@tonic-gate 20127c478bd9Sstevel@tonic-gate ASSERT(resop->resop == OP_GETFH); 20137c478bd9Sstevel@tonic-gate tmpfhp = &resop->nfs_resop4_u.opgetfh.object; 20147c478bd9Sstevel@tonic-gate 20157c478bd9Sstevel@tonic-gate /* save the filehandles for the replica */ 20167c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_WRITER, 0); 20177c478bd9Sstevel@tonic-gate ASSERT(tmpfhp->nfs_fh4_len <= NFS4_FHSIZE); 20187c478bd9Sstevel@tonic-gate svp->sv_pfhandle.fh_len = tmpfhp->nfs_fh4_len; 20197c478bd9Sstevel@tonic-gate bcopy(tmpfhp->nfs_fh4_val, svp->sv_pfhandle.fh_buf, 20207c478bd9Sstevel@tonic-gate tmpfhp->nfs_fh4_len); 20217c478bd9Sstevel@tonic-gate ASSERT(resfhp->nfs_fh4_len <= NFS4_FHSIZE); 20227c478bd9Sstevel@tonic-gate svp->sv_fhandle.fh_len = resfhp->nfs_fh4_len; 20237c478bd9Sstevel@tonic-gate bcopy(resfhp->nfs_fh4_val, svp->sv_fhandle.fh_buf, resfhp->nfs_fh4_len); 20247c478bd9Sstevel@tonic-gate 20257c478bd9Sstevel@tonic-gate /* initialize fsid and supp_attrs for server fs */ 20267c478bd9Sstevel@tonic-gate svp->sv_fsid = garp->n4g_fsid; 20277c478bd9Sstevel@tonic-gate svp->sv_supp_attrs = 2028b9238976Sth garp->n4g_ext_res->n4g_suppattrs | FATTR4_MANDATTR_MASK; 20297c478bd9Sstevel@tonic-gate 20307c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 20317c478bd9Sstevel@tonic-gate nfs4args_lookup_free(argop, num_argops); 20327c478bd9Sstevel@tonic-gate kmem_free(argop, lookuparg.arglen * sizeof (nfs_argop4)); 20337c478bd9Sstevel@tonic-gate (void) xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&res); 20347c478bd9Sstevel@tonic-gate if (!recovery) 20357c478bd9Sstevel@tonic-gate nfs4_end_fop(mi, NULL, NULL, OH_MOUNT, &recov_state, needrecov); 20367c478bd9Sstevel@tonic-gate } 20377c478bd9Sstevel@tonic-gate 20382f172c55SRobert Thurlow /* 20392f172c55SRobert Thurlow * Save a copy of Servinfo4_t structure. 20402f172c55SRobert Thurlow * We might need when there is a failure in getting file handle 20412f172c55SRobert Thurlow * in case of a referral to replace servinfo4 struct and try again. 20422f172c55SRobert Thurlow */ 20432f172c55SRobert Thurlow static struct servinfo4 * 20442f172c55SRobert Thurlow copy_svp(servinfo4_t *nsvp) 20452f172c55SRobert Thurlow { 20462f172c55SRobert Thurlow servinfo4_t *svp = NULL; 20472f172c55SRobert Thurlow struct knetconfig *sknconf, *tknconf; 20482f172c55SRobert Thurlow struct netbuf *saddr, *taddr; 20492f172c55SRobert Thurlow 20502f172c55SRobert Thurlow svp = kmem_zalloc(sizeof (*svp), KM_SLEEP); 20512f172c55SRobert Thurlow nfs_rw_init(&svp->sv_lock, NULL, RW_DEFAULT, NULL); 20522f172c55SRobert Thurlow svp->sv_flags = nsvp->sv_flags; 20532f172c55SRobert Thurlow svp->sv_fsid = nsvp->sv_fsid; 20542f172c55SRobert Thurlow svp->sv_hostnamelen = nsvp->sv_hostnamelen; 20552f172c55SRobert Thurlow svp->sv_pathlen = nsvp->sv_pathlen; 20562f172c55SRobert Thurlow svp->sv_supp_attrs = nsvp->sv_supp_attrs; 20572f172c55SRobert Thurlow 20582f172c55SRobert Thurlow svp->sv_path = kmem_alloc(svp->sv_pathlen, KM_SLEEP); 20592f172c55SRobert Thurlow svp->sv_hostname = kmem_alloc(svp->sv_hostnamelen, KM_SLEEP); 20602f172c55SRobert Thurlow bcopy(nsvp->sv_hostname, svp->sv_hostname, svp->sv_hostnamelen); 20612f172c55SRobert Thurlow bcopy(nsvp->sv_path, svp->sv_path, svp->sv_pathlen); 20622f172c55SRobert Thurlow 20632f172c55SRobert Thurlow saddr = &nsvp->sv_addr; 20642f172c55SRobert Thurlow taddr = &svp->sv_addr; 20652f172c55SRobert Thurlow taddr->maxlen = saddr->maxlen; 20662f172c55SRobert Thurlow taddr->len = saddr->len; 20672f172c55SRobert Thurlow if (saddr->len > 0) { 20682f172c55SRobert Thurlow taddr->buf = kmem_zalloc(saddr->maxlen, KM_SLEEP); 20692f172c55SRobert Thurlow bcopy(saddr->buf, taddr->buf, saddr->len); 20702f172c55SRobert Thurlow } 20712f172c55SRobert Thurlow 20722f172c55SRobert Thurlow svp->sv_knconf = kmem_zalloc(sizeof (struct knetconfig), KM_SLEEP); 20732f172c55SRobert Thurlow sknconf = nsvp->sv_knconf; 20742f172c55SRobert Thurlow tknconf = svp->sv_knconf; 20752f172c55SRobert Thurlow tknconf->knc_semantics = sknconf->knc_semantics; 20762f172c55SRobert Thurlow tknconf->knc_rdev = sknconf->knc_rdev; 20772f172c55SRobert Thurlow if (sknconf->knc_proto != NULL) { 20782f172c55SRobert Thurlow tknconf->knc_proto = kmem_zalloc(KNC_STRSIZE, KM_SLEEP); 20792f172c55SRobert Thurlow bcopy(sknconf->knc_proto, (char *)tknconf->knc_proto, 20802f172c55SRobert Thurlow KNC_STRSIZE); 20812f172c55SRobert Thurlow } 20822f172c55SRobert Thurlow if (sknconf->knc_protofmly != NULL) { 20832f172c55SRobert Thurlow tknconf->knc_protofmly = kmem_zalloc(KNC_STRSIZE, KM_SLEEP); 20842f172c55SRobert Thurlow bcopy(sknconf->knc_protofmly, (char *)tknconf->knc_protofmly, 20852f172c55SRobert Thurlow KNC_STRSIZE); 20862f172c55SRobert Thurlow } 20872f172c55SRobert Thurlow 20882f172c55SRobert Thurlow if (nsvp->sv_origknconf != NULL) { 20892f172c55SRobert Thurlow svp->sv_origknconf = kmem_zalloc(sizeof (struct knetconfig), 20902f172c55SRobert Thurlow KM_SLEEP); 20912f172c55SRobert Thurlow sknconf = nsvp->sv_origknconf; 20922f172c55SRobert Thurlow tknconf = svp->sv_origknconf; 20932f172c55SRobert Thurlow tknconf->knc_semantics = sknconf->knc_semantics; 20942f172c55SRobert Thurlow tknconf->knc_rdev = sknconf->knc_rdev; 20952f172c55SRobert Thurlow if (sknconf->knc_proto != NULL) { 20962f172c55SRobert Thurlow tknconf->knc_proto = kmem_zalloc(KNC_STRSIZE, KM_SLEEP); 20972f172c55SRobert Thurlow bcopy(sknconf->knc_proto, (char *)tknconf->knc_proto, 20982f172c55SRobert Thurlow KNC_STRSIZE); 20992f172c55SRobert Thurlow } 21002f172c55SRobert Thurlow if (sknconf->knc_protofmly != NULL) { 21012f172c55SRobert Thurlow tknconf->knc_protofmly = kmem_zalloc(KNC_STRSIZE, 21022f172c55SRobert Thurlow KM_SLEEP); 21032f172c55SRobert Thurlow bcopy(sknconf->knc_protofmly, 21042f172c55SRobert Thurlow (char *)tknconf->knc_protofmly, KNC_STRSIZE); 21052f172c55SRobert Thurlow } 21062f172c55SRobert Thurlow } 21072f172c55SRobert Thurlow 21082f172c55SRobert Thurlow svp->sv_secdata = copy_sec_data(nsvp->sv_secdata); 21092f172c55SRobert Thurlow svp->sv_dhsec = copy_sec_data(svp->sv_dhsec); 21102f172c55SRobert Thurlow /* 21112f172c55SRobert Thurlow * Rest of the security information is not copied as they are built 21122f172c55SRobert Thurlow * with the information available from secdata and dhsec. 21132f172c55SRobert Thurlow */ 21142f172c55SRobert Thurlow svp->sv_next = NULL; 21152f172c55SRobert Thurlow 21162f172c55SRobert Thurlow return (svp); 21172f172c55SRobert Thurlow } 21182f172c55SRobert Thurlow 21192f172c55SRobert Thurlow servinfo4_t * 21202f172c55SRobert Thurlow restore_svp(mntinfo4_t *mi, servinfo4_t *svp, servinfo4_t *origsvp) 21212f172c55SRobert Thurlow { 21222f172c55SRobert Thurlow servinfo4_t *srvnext, *tmpsrv; 21232f172c55SRobert Thurlow 21242f172c55SRobert Thurlow if (strcmp(svp->sv_hostname, origsvp->sv_hostname) != 0) { 21252f172c55SRobert Thurlow /* 21262f172c55SRobert Thurlow * Since the hostname changed, we must be dealing 21272f172c55SRobert Thurlow * with a referral, and the lookup failed. We will 21282f172c55SRobert Thurlow * restore the whole servinfo4_t to what it was before. 21292f172c55SRobert Thurlow */ 21302f172c55SRobert Thurlow srvnext = svp->sv_next; 21312f172c55SRobert Thurlow svp->sv_next = NULL; 21322f172c55SRobert Thurlow tmpsrv = copy_svp(origsvp); 21332f172c55SRobert Thurlow sv4_free(svp); 21342f172c55SRobert Thurlow svp = tmpsrv; 21352f172c55SRobert Thurlow svp->sv_next = srvnext; 21362f172c55SRobert Thurlow mutex_enter(&mi->mi_lock); 21372f172c55SRobert Thurlow mi->mi_servers = svp; 21382f172c55SRobert Thurlow mi->mi_curr_serv = svp; 21392f172c55SRobert Thurlow mutex_exit(&mi->mi_lock); 21402f172c55SRobert Thurlow 21412f172c55SRobert Thurlow } else if (origsvp->sv_pathlen != svp->sv_pathlen) { 21422f172c55SRobert Thurlow 21432f172c55SRobert Thurlow /* 21442f172c55SRobert Thurlow * For symlink case: restore original path because 21452f172c55SRobert Thurlow * it might have contained symlinks that were 21462f172c55SRobert Thurlow * expanded by nfsgetfh_otw before the failure occurred. 21472f172c55SRobert Thurlow */ 21485301ec54SRobert Thurlow (void) nfs_rw_enter_sig(&svp->sv_lock, RW_READER, 0); 21492f172c55SRobert Thurlow kmem_free(svp->sv_path, svp->sv_pathlen); 21502f172c55SRobert Thurlow svp->sv_path = 21512f172c55SRobert Thurlow kmem_alloc(origsvp->sv_pathlen, KM_SLEEP); 21522f172c55SRobert Thurlow svp->sv_pathlen = origsvp->sv_pathlen; 21532f172c55SRobert Thurlow bcopy(origsvp->sv_path, svp->sv_path, 21542f172c55SRobert Thurlow origsvp->sv_pathlen); 21552f172c55SRobert Thurlow nfs_rw_exit(&svp->sv_lock); 21562f172c55SRobert Thurlow } 21572f172c55SRobert Thurlow return (svp); 21582f172c55SRobert Thurlow } 21592f172c55SRobert Thurlow 21607c478bd9Sstevel@tonic-gate static ushort_t nfs4_max_threads = 8; /* max number of active async threads */ 2161c242f9a0Schunli zhang - Sun Microsystems - Irvine United States uint_t nfs4_bsize = 32 * 1024; /* client `block' size */ 21627c478bd9Sstevel@tonic-gate static uint_t nfs4_async_clusters = 1; /* # of reqs from each async queue */ 21637c478bd9Sstevel@tonic-gate static uint_t nfs4_cots_timeo = NFS_COTS_TIMEO; 21647c478bd9Sstevel@tonic-gate 21657c478bd9Sstevel@tonic-gate /* 21667c478bd9Sstevel@tonic-gate * Remap the root filehandle for the given filesystem. 21677c478bd9Sstevel@tonic-gate * 21687c478bd9Sstevel@tonic-gate * results returned via the nfs4_error_t parameter. 21697c478bd9Sstevel@tonic-gate */ 21707c478bd9Sstevel@tonic-gate void 21717c478bd9Sstevel@tonic-gate nfs4_remap_root(mntinfo4_t *mi, nfs4_error_t *ep, int flags) 21727c478bd9Sstevel@tonic-gate { 21732f172c55SRobert Thurlow struct servinfo4 *svp, *origsvp; 21747c478bd9Sstevel@tonic-gate vtype_t vtype; 21757c478bd9Sstevel@tonic-gate nfs_fh4 rootfh; 21767c478bd9Sstevel@tonic-gate int getfh_flags; 21772f172c55SRobert Thurlow int num_retry; 21787c478bd9Sstevel@tonic-gate 21797c478bd9Sstevel@tonic-gate mutex_enter(&mi->mi_lock); 21808c9e5ad2Saalok 21818c9e5ad2Saalok remap_retry: 21827c478bd9Sstevel@tonic-gate svp = mi->mi_curr_serv; 21837c478bd9Sstevel@tonic-gate getfh_flags = 2184b9238976Sth (flags & NFS4_REMAP_NEEDSOP) ? NFS4_GETFH_NEEDSOP : 0; 21857c478bd9Sstevel@tonic-gate getfh_flags |= 2186b9238976Sth (mi->mi_flags & MI4_PUBLIC) ? NFS4_GETFH_PUBLIC : 0; 21877c478bd9Sstevel@tonic-gate mutex_exit(&mi->mi_lock); 21887c478bd9Sstevel@tonic-gate 21897c478bd9Sstevel@tonic-gate /* 21907c478bd9Sstevel@tonic-gate * Just in case server path being mounted contains 21917c478bd9Sstevel@tonic-gate * symlinks and fails w/STALE, save the initial sv_path 21927c478bd9Sstevel@tonic-gate * so we can redrive the initial mount compound with the 21937c478bd9Sstevel@tonic-gate * initial sv_path -- not a symlink-expanded version. 21947c478bd9Sstevel@tonic-gate * 21957c478bd9Sstevel@tonic-gate * This could only happen if a symlink was expanded 21967c478bd9Sstevel@tonic-gate * and the expanded mount compound failed stale. Because 21977c478bd9Sstevel@tonic-gate * it could be the case that the symlink was removed at 21987c478bd9Sstevel@tonic-gate * the server (and replaced with another symlink/dir, 21997c478bd9Sstevel@tonic-gate * we need to use the initial sv_path when attempting 22007c478bd9Sstevel@tonic-gate * to re-lookup everything and recover. 22017c478bd9Sstevel@tonic-gate */ 22027c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_READER, 0); 22032f172c55SRobert Thurlow origsvp = copy_svp(svp); 22047c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 22057c478bd9Sstevel@tonic-gate 22067c478bd9Sstevel@tonic-gate num_retry = nfs4_max_mount_retry; 22077c478bd9Sstevel@tonic-gate 22087c478bd9Sstevel@tonic-gate do { 22097c478bd9Sstevel@tonic-gate /* 22107c478bd9Sstevel@tonic-gate * Get the root fh from the server. Retry nfs4_max_mount_retry 22117c478bd9Sstevel@tonic-gate * (2) times if it fails with STALE since the recovery 22127c478bd9Sstevel@tonic-gate * infrastructure doesn't do STALE recovery for components 22137c478bd9Sstevel@tonic-gate * of the server path to the object being mounted. 22147c478bd9Sstevel@tonic-gate */ 22157c478bd9Sstevel@tonic-gate nfs4getfh_otw(mi, svp, &vtype, getfh_flags, CRED(), ep); 22167c478bd9Sstevel@tonic-gate 22177c478bd9Sstevel@tonic-gate if (ep->error == 0 && ep->stat == NFS4_OK) 22187c478bd9Sstevel@tonic-gate break; 22197c478bd9Sstevel@tonic-gate 22207c478bd9Sstevel@tonic-gate /* 22217c478bd9Sstevel@tonic-gate * For some reason, the mount compound failed. Before 22222f172c55SRobert Thurlow * retrying, we need to restore original conditions. 22237c478bd9Sstevel@tonic-gate */ 22242f172c55SRobert Thurlow svp = restore_svp(mi, svp, origsvp); 22257c478bd9Sstevel@tonic-gate 22267c478bd9Sstevel@tonic-gate } while (num_retry-- > 0); 22277c478bd9Sstevel@tonic-gate 22282f172c55SRobert Thurlow sv4_free(origsvp); 22297c478bd9Sstevel@tonic-gate 22307c478bd9Sstevel@tonic-gate if (ep->error != 0 || ep->stat != 0) { 22317c478bd9Sstevel@tonic-gate return; 22327c478bd9Sstevel@tonic-gate } 22337c478bd9Sstevel@tonic-gate 22347c478bd9Sstevel@tonic-gate if (vtype != VNON && vtype != mi->mi_type) { 22357c478bd9Sstevel@tonic-gate /* shouldn't happen */ 22367c478bd9Sstevel@tonic-gate zcmn_err(mi->mi_zone->zone_id, CE_WARN, 2237b9238976Sth "nfs4_remap_root: server root vnode type (%d) doesn't " 2238b9238976Sth "match mount info (%d)", vtype, mi->mi_type); 22397c478bd9Sstevel@tonic-gate } 22407c478bd9Sstevel@tonic-gate 22417c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_READER, 0); 22427c478bd9Sstevel@tonic-gate rootfh.nfs_fh4_val = svp->sv_fhandle.fh_buf; 22437c478bd9Sstevel@tonic-gate rootfh.nfs_fh4_len = svp->sv_fhandle.fh_len; 22447c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 22457c478bd9Sstevel@tonic-gate sfh4_update(mi->mi_rootfh, &rootfh); 22467c478bd9Sstevel@tonic-gate 22477c478bd9Sstevel@tonic-gate /* 22488c9e5ad2Saalok * It's possible that recovery took place on the filesystem 22498c9e5ad2Saalok * and the server has been updated between the time we did 22508c9e5ad2Saalok * the nfs4getfh_otw and now. Re-drive the otw operation 22518c9e5ad2Saalok * to make sure we have a good fh. 22527c478bd9Sstevel@tonic-gate */ 22537c478bd9Sstevel@tonic-gate mutex_enter(&mi->mi_lock); 22548c9e5ad2Saalok if (mi->mi_curr_serv != svp) 22558c9e5ad2Saalok goto remap_retry; 22568c9e5ad2Saalok 22577c478bd9Sstevel@tonic-gate mutex_exit(&mi->mi_lock); 22587c478bd9Sstevel@tonic-gate } 22597c478bd9Sstevel@tonic-gate 22607c478bd9Sstevel@tonic-gate static int 22617c478bd9Sstevel@tonic-gate nfs4rootvp(vnode_t **rtvpp, vfs_t *vfsp, struct servinfo4 *svp_head, 2262b9238976Sth int flags, cred_t *cr, zone_t *zone) 22637c478bd9Sstevel@tonic-gate { 22647c478bd9Sstevel@tonic-gate vnode_t *rtvp = NULL; 22657c478bd9Sstevel@tonic-gate mntinfo4_t *mi; 22667c478bd9Sstevel@tonic-gate dev_t nfs_dev; 22677c478bd9Sstevel@tonic-gate int error = 0; 22687c478bd9Sstevel@tonic-gate rnode4_t *rp; 22692f172c55SRobert Thurlow int i, len; 22707c478bd9Sstevel@tonic-gate struct vattr va; 22717c478bd9Sstevel@tonic-gate vtype_t vtype = VNON; 22727c478bd9Sstevel@tonic-gate vtype_t tmp_vtype = VNON; 22737c478bd9Sstevel@tonic-gate struct servinfo4 *firstsvp = NULL, *svp = svp_head; 22747c478bd9Sstevel@tonic-gate nfs4_oo_hash_bucket_t *bucketp; 22757c478bd9Sstevel@tonic-gate nfs_fh4 fh; 22767c478bd9Sstevel@tonic-gate char *droptext = ""; 22777c478bd9Sstevel@tonic-gate struct nfs_stats *nfsstatsp; 22787c478bd9Sstevel@tonic-gate nfs4_fname_t *mfname; 22797c478bd9Sstevel@tonic-gate nfs4_error_t e; 22802f172c55SRobert Thurlow int num_retry, removed; 22817c478bd9Sstevel@tonic-gate cred_t *lcr = NULL, *tcr = cr; 22822f172c55SRobert Thurlow struct servinfo4 *origsvp; 22832f172c55SRobert Thurlow char *resource; 22847c478bd9Sstevel@tonic-gate 2285108322fbScarlsonj nfsstatsp = zone_getspecific(nfsstat_zone_key, nfs_zone()); 22867c478bd9Sstevel@tonic-gate ASSERT(nfsstatsp != NULL); 22877c478bd9Sstevel@tonic-gate 2288108322fbScarlsonj ASSERT(nfs_zone() == zone); 22897c478bd9Sstevel@tonic-gate ASSERT(crgetref(cr)); 22907c478bd9Sstevel@tonic-gate 22917c478bd9Sstevel@tonic-gate /* 22927c478bd9Sstevel@tonic-gate * Create a mount record and link it to the vfs struct. 22937c478bd9Sstevel@tonic-gate */ 22947c478bd9Sstevel@tonic-gate mi = kmem_zalloc(sizeof (*mi), KM_SLEEP); 22957c478bd9Sstevel@tonic-gate mutex_init(&mi->mi_lock, NULL, MUTEX_DEFAULT, NULL); 22967c478bd9Sstevel@tonic-gate nfs_rw_init(&mi->mi_recovlock, NULL, RW_DEFAULT, NULL); 22977c478bd9Sstevel@tonic-gate nfs_rw_init(&mi->mi_rename_lock, NULL, RW_DEFAULT, NULL); 22987c478bd9Sstevel@tonic-gate nfs_rw_init(&mi->mi_fh_lock, NULL, RW_DEFAULT, NULL); 22997c478bd9Sstevel@tonic-gate 23007c478bd9Sstevel@tonic-gate if (!(flags & NFSMNT_SOFT)) 23017c478bd9Sstevel@tonic-gate mi->mi_flags |= MI4_HARD; 23027c478bd9Sstevel@tonic-gate if ((flags & NFSMNT_NOPRINT)) 23037c478bd9Sstevel@tonic-gate mi->mi_flags |= MI4_NOPRINT; 23047c478bd9Sstevel@tonic-gate if (flags & NFSMNT_INT) 23057c478bd9Sstevel@tonic-gate mi->mi_flags |= MI4_INT; 23067c478bd9Sstevel@tonic-gate if (flags & NFSMNT_PUBLIC) 23077c478bd9Sstevel@tonic-gate mi->mi_flags |= MI4_PUBLIC; 2308b9238976Sth if (flags & NFSMNT_MIRRORMOUNT) 2309b9238976Sth mi->mi_flags |= MI4_MIRRORMOUNT; 23102f172c55SRobert Thurlow if (flags & NFSMNT_REFERRAL) 23112f172c55SRobert Thurlow mi->mi_flags |= MI4_REFERRAL; 23127c478bd9Sstevel@tonic-gate mi->mi_retrans = NFS_RETRIES; 23137c478bd9Sstevel@tonic-gate if (svp->sv_knconf->knc_semantics == NC_TPI_COTS_ORD || 23147c478bd9Sstevel@tonic-gate svp->sv_knconf->knc_semantics == NC_TPI_COTS) 23157c478bd9Sstevel@tonic-gate mi->mi_timeo = nfs4_cots_timeo; 23167c478bd9Sstevel@tonic-gate else 23177c478bd9Sstevel@tonic-gate mi->mi_timeo = NFS_TIMEO; 23187c478bd9Sstevel@tonic-gate mi->mi_prog = NFS_PROGRAM; 23197c478bd9Sstevel@tonic-gate mi->mi_vers = NFS_V4; 23207c478bd9Sstevel@tonic-gate mi->mi_rfsnames = rfsnames_v4; 23217c478bd9Sstevel@tonic-gate mi->mi_reqs = nfsstatsp->nfs_stats_v4.rfsreqcnt_ptr; 23227c478bd9Sstevel@tonic-gate cv_init(&mi->mi_failover_cv, NULL, CV_DEFAULT, NULL); 23237c478bd9Sstevel@tonic-gate mi->mi_servers = svp; 23247c478bd9Sstevel@tonic-gate mi->mi_curr_serv = svp; 23257c478bd9Sstevel@tonic-gate mi->mi_acregmin = SEC2HR(ACREGMIN); 23267c478bd9Sstevel@tonic-gate mi->mi_acregmax = SEC2HR(ACREGMAX); 23277c478bd9Sstevel@tonic-gate mi->mi_acdirmin = SEC2HR(ACDIRMIN); 23287c478bd9Sstevel@tonic-gate mi->mi_acdirmax = SEC2HR(ACDIRMAX); 23297c478bd9Sstevel@tonic-gate mi->mi_fh_expire_type = FH4_PERSISTENT; 23307c478bd9Sstevel@tonic-gate mi->mi_clientid_next = NULL; 23317c478bd9Sstevel@tonic-gate mi->mi_clientid_prev = NULL; 23323b895386SPavel Filipensky mi->mi_srv = NULL; 23337c478bd9Sstevel@tonic-gate mi->mi_grace_wait = 0; 23347c478bd9Sstevel@tonic-gate mi->mi_error = 0; 23357c478bd9Sstevel@tonic-gate mi->mi_srvsettime = 0; 23363b895386SPavel Filipensky mi->mi_srvset_cnt = 0; 23377c478bd9Sstevel@tonic-gate 233850a83466Sjwahlig mi->mi_count = 1; 233950a83466Sjwahlig 23407c478bd9Sstevel@tonic-gate mi->mi_tsize = nfs4_tsize(svp->sv_knconf); 23417c478bd9Sstevel@tonic-gate mi->mi_stsize = mi->mi_tsize; 23427c478bd9Sstevel@tonic-gate 23437c478bd9Sstevel@tonic-gate if (flags & NFSMNT_DIRECTIO) 23447c478bd9Sstevel@tonic-gate mi->mi_flags |= MI4_DIRECTIO; 23457c478bd9Sstevel@tonic-gate 23467c478bd9Sstevel@tonic-gate mi->mi_flags |= MI4_MOUNTING; 23477c478bd9Sstevel@tonic-gate 23487c478bd9Sstevel@tonic-gate /* 23497c478bd9Sstevel@tonic-gate * Make a vfs struct for nfs. We do this here instead of below 23507c478bd9Sstevel@tonic-gate * because rtvp needs a vfs before we can do a getattr on it. 23517c478bd9Sstevel@tonic-gate * 23527c478bd9Sstevel@tonic-gate * Assign a unique device id to the mount 23537c478bd9Sstevel@tonic-gate */ 23547c478bd9Sstevel@tonic-gate mutex_enter(&nfs_minor_lock); 23557c478bd9Sstevel@tonic-gate do { 23567c478bd9Sstevel@tonic-gate nfs_minor = (nfs_minor + 1) & MAXMIN32; 23577c478bd9Sstevel@tonic-gate nfs_dev = makedevice(nfs_major, nfs_minor); 23587c478bd9Sstevel@tonic-gate } while (vfs_devismounted(nfs_dev)); 23597c478bd9Sstevel@tonic-gate mutex_exit(&nfs_minor_lock); 23607c478bd9Sstevel@tonic-gate 23617c478bd9Sstevel@tonic-gate vfsp->vfs_dev = nfs_dev; 23627c478bd9Sstevel@tonic-gate vfs_make_fsid(&vfsp->vfs_fsid, nfs_dev, nfs4fstyp); 23637c478bd9Sstevel@tonic-gate vfsp->vfs_data = (caddr_t)mi; 23647c478bd9Sstevel@tonic-gate vfsp->vfs_fstype = nfsfstyp; 23657c478bd9Sstevel@tonic-gate vfsp->vfs_bsize = nfs4_bsize; 23667c478bd9Sstevel@tonic-gate 23677c478bd9Sstevel@tonic-gate /* 23687c478bd9Sstevel@tonic-gate * Initialize fields used to support async putpage operations. 23697c478bd9Sstevel@tonic-gate */ 23707c478bd9Sstevel@tonic-gate for (i = 0; i < NFS4_ASYNC_TYPES; i++) 23717c478bd9Sstevel@tonic-gate mi->mi_async_clusters[i] = nfs4_async_clusters; 23727c478bd9Sstevel@tonic-gate mi->mi_async_init_clusters = nfs4_async_clusters; 23730776f5e6SVallish Vaidyeshwara mi->mi_async_curr[NFS4_ASYNC_QUEUE] = 23740776f5e6SVallish Vaidyeshwara mi->mi_async_curr[NFS4_ASYNC_PGOPS_QUEUE] = &mi->mi_async_reqs[0]; 23757c478bd9Sstevel@tonic-gate mi->mi_max_threads = nfs4_max_threads; 23767c478bd9Sstevel@tonic-gate mutex_init(&mi->mi_async_lock, NULL, MUTEX_DEFAULT, NULL); 23777c478bd9Sstevel@tonic-gate cv_init(&mi->mi_async_reqs_cv, NULL, CV_DEFAULT, NULL); 23780776f5e6SVallish Vaidyeshwara cv_init(&mi->mi_async_work_cv[NFS4_ASYNC_QUEUE], NULL, CV_DEFAULT, 23790776f5e6SVallish Vaidyeshwara NULL); 23800776f5e6SVallish Vaidyeshwara cv_init(&mi->mi_async_work_cv[NFS4_ASYNC_PGOPS_QUEUE], NULL, 23810776f5e6SVallish Vaidyeshwara CV_DEFAULT, NULL); 23827c478bd9Sstevel@tonic-gate cv_init(&mi->mi_async_cv, NULL, CV_DEFAULT, NULL); 23837c478bd9Sstevel@tonic-gate cv_init(&mi->mi_inact_req_cv, NULL, CV_DEFAULT, NULL); 23847c478bd9Sstevel@tonic-gate 23857c478bd9Sstevel@tonic-gate mi->mi_vfsp = vfsp; 23867c478bd9Sstevel@tonic-gate zone_hold(mi->mi_zone = zone); 23877c478bd9Sstevel@tonic-gate nfs4_mi_zonelist_add(mi); 23887c478bd9Sstevel@tonic-gate 23897c478bd9Sstevel@tonic-gate /* 23907c478bd9Sstevel@tonic-gate * Initialize the <open owner/cred> hash table. 23917c478bd9Sstevel@tonic-gate */ 23927c478bd9Sstevel@tonic-gate for (i = 0; i < NFS4_NUM_OO_BUCKETS; i++) { 23937c478bd9Sstevel@tonic-gate bucketp = &(mi->mi_oo_list[i]); 23947c478bd9Sstevel@tonic-gate mutex_init(&bucketp->b_lock, NULL, MUTEX_DEFAULT, NULL); 23957c478bd9Sstevel@tonic-gate list_create(&bucketp->b_oo_hash_list, 23967c478bd9Sstevel@tonic-gate sizeof (nfs4_open_owner_t), 23977c478bd9Sstevel@tonic-gate offsetof(nfs4_open_owner_t, oo_hash_node)); 23987c478bd9Sstevel@tonic-gate } 23997c478bd9Sstevel@tonic-gate 24007c478bd9Sstevel@tonic-gate /* 24017c478bd9Sstevel@tonic-gate * Initialize the freed open owner list. 24027c478bd9Sstevel@tonic-gate */ 24037c478bd9Sstevel@tonic-gate mi->mi_foo_num = 0; 24047c478bd9Sstevel@tonic-gate mi->mi_foo_max = NFS4_NUM_FREED_OPEN_OWNERS; 24057c478bd9Sstevel@tonic-gate list_create(&mi->mi_foo_list, sizeof (nfs4_open_owner_t), 24067c478bd9Sstevel@tonic-gate offsetof(nfs4_open_owner_t, oo_foo_node)); 24077c478bd9Sstevel@tonic-gate 24087c478bd9Sstevel@tonic-gate list_create(&mi->mi_lost_state, sizeof (nfs4_lost_rqst_t), 24097c478bd9Sstevel@tonic-gate offsetof(nfs4_lost_rqst_t, lr_node)); 24107c478bd9Sstevel@tonic-gate 24117c478bd9Sstevel@tonic-gate list_create(&mi->mi_bseqid_list, sizeof (nfs4_bseqid_entry_t), 24127c478bd9Sstevel@tonic-gate offsetof(nfs4_bseqid_entry_t, bs_node)); 24137c478bd9Sstevel@tonic-gate 24147c478bd9Sstevel@tonic-gate /* 24157c478bd9Sstevel@tonic-gate * Initialize the msg buffer. 24167c478bd9Sstevel@tonic-gate */ 24177c478bd9Sstevel@tonic-gate list_create(&mi->mi_msg_list, sizeof (nfs4_debug_msg_t), 24187c478bd9Sstevel@tonic-gate offsetof(nfs4_debug_msg_t, msg_node)); 24197c478bd9Sstevel@tonic-gate mi->mi_msg_count = 0; 24207c478bd9Sstevel@tonic-gate mutex_init(&mi->mi_msg_list_lock, NULL, MUTEX_DEFAULT, NULL); 24217c478bd9Sstevel@tonic-gate 24227c478bd9Sstevel@tonic-gate /* 24237c478bd9Sstevel@tonic-gate * Initialize kstats 24247c478bd9Sstevel@tonic-gate */ 24257c478bd9Sstevel@tonic-gate nfs4_mnt_kstat_init(vfsp); 24267c478bd9Sstevel@tonic-gate 24277c478bd9Sstevel@tonic-gate /* 2428bbf2a467SNagakiran Rajashekar * Initialize the shared filehandle pool. 24297c478bd9Sstevel@tonic-gate */ 24307c478bd9Sstevel@tonic-gate sfh4_createtab(&mi->mi_filehandles); 24317c478bd9Sstevel@tonic-gate 24327c478bd9Sstevel@tonic-gate /* 24337c478bd9Sstevel@tonic-gate * Save server path we're attempting to mount. 24347c478bd9Sstevel@tonic-gate */ 24357c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_WRITER, 0); 24362f172c55SRobert Thurlow origsvp = copy_svp(svp); 24377c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 24387c478bd9Sstevel@tonic-gate 24397c478bd9Sstevel@tonic-gate /* 24407c478bd9Sstevel@tonic-gate * Make the GETFH call to get root fh for each replica. 24417c478bd9Sstevel@tonic-gate */ 24427c478bd9Sstevel@tonic-gate if (svp_head->sv_next) 24437c478bd9Sstevel@tonic-gate droptext = ", dropping replica"; 24447c478bd9Sstevel@tonic-gate 24457c478bd9Sstevel@tonic-gate /* 24467c478bd9Sstevel@tonic-gate * If the uid is set then set the creds for secure mounts 24477c478bd9Sstevel@tonic-gate * by proxy processes such as automountd. 24487c478bd9Sstevel@tonic-gate */ 24497c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_READER, 0); 2450f722863dSSiyamaladevi Santhana Krishnan if (svp->sv_secdata->uid != 0 && 2451f722863dSSiyamaladevi Santhana Krishnan svp->sv_secdata->rpcflavor == RPCSEC_GSS) { 24527c478bd9Sstevel@tonic-gate lcr = crdup(cr); 24537c478bd9Sstevel@tonic-gate (void) crsetugid(lcr, svp->sv_secdata->uid, crgetgid(cr)); 24547c478bd9Sstevel@tonic-gate tcr = lcr; 24557c478bd9Sstevel@tonic-gate } 24567c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 24577c478bd9Sstevel@tonic-gate for (svp = svp_head; svp; svp = svp->sv_next) { 24587c478bd9Sstevel@tonic-gate if (nfs4_chkdup_servinfo4(svp_head, svp)) { 24597c478bd9Sstevel@tonic-gate nfs_cmn_err(error, CE_WARN, 2460b9238976Sth VERS_MSG "Host %s is a duplicate%s", 2461b9238976Sth svp->sv_hostname, droptext); 24627c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_WRITER, 0); 24637c478bd9Sstevel@tonic-gate svp->sv_flags |= SV4_NOTINUSE; 24647c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 24657c478bd9Sstevel@tonic-gate continue; 24667c478bd9Sstevel@tonic-gate } 24677c478bd9Sstevel@tonic-gate mi->mi_curr_serv = svp; 24687c478bd9Sstevel@tonic-gate 24697c478bd9Sstevel@tonic-gate /* 24707c478bd9Sstevel@tonic-gate * Just in case server path being mounted contains 24717c478bd9Sstevel@tonic-gate * symlinks and fails w/STALE, save the initial sv_path 24727c478bd9Sstevel@tonic-gate * so we can redrive the initial mount compound with the 24737c478bd9Sstevel@tonic-gate * initial sv_path -- not a symlink-expanded version. 24747c478bd9Sstevel@tonic-gate * 24757c478bd9Sstevel@tonic-gate * This could only happen if a symlink was expanded 24767c478bd9Sstevel@tonic-gate * and the expanded mount compound failed stale. Because 24777c478bd9Sstevel@tonic-gate * it could be the case that the symlink was removed at 24787c478bd9Sstevel@tonic-gate * the server (and replaced with another symlink/dir, 24797c478bd9Sstevel@tonic-gate * we need to use the initial sv_path when attempting 24807c478bd9Sstevel@tonic-gate * to re-lookup everything and recover. 24817c478bd9Sstevel@tonic-gate * 24827c478bd9Sstevel@tonic-gate * Other mount errors should evenutally be handled here also 24837c478bd9Sstevel@tonic-gate * (NFS4ERR_DELAY, NFS4ERR_RESOURCE). For now, all mount 24847c478bd9Sstevel@tonic-gate * failures will result in mount being redriven a few times. 24857c478bd9Sstevel@tonic-gate */ 24867c478bd9Sstevel@tonic-gate num_retry = nfs4_max_mount_retry; 24877c478bd9Sstevel@tonic-gate do { 24887c478bd9Sstevel@tonic-gate nfs4getfh_otw(mi, svp, &tmp_vtype, 24897c478bd9Sstevel@tonic-gate ((flags & NFSMNT_PUBLIC) ? NFS4_GETFH_PUBLIC : 0) | 24907c478bd9Sstevel@tonic-gate NFS4_GETFH_NEEDSOP, tcr, &e); 24917c478bd9Sstevel@tonic-gate 24927c478bd9Sstevel@tonic-gate if (e.error == 0 && e.stat == NFS4_OK) 24937c478bd9Sstevel@tonic-gate break; 24947c478bd9Sstevel@tonic-gate 24957c478bd9Sstevel@tonic-gate /* 24962f172c55SRobert Thurlow * For some reason, the mount compound failed. Before 24972f172c55SRobert Thurlow * retrying, we need to restore original conditions. 24987c478bd9Sstevel@tonic-gate */ 24992f172c55SRobert Thurlow svp = restore_svp(mi, svp, origsvp); 25002f172c55SRobert Thurlow svp_head = svp; 25017c478bd9Sstevel@tonic-gate 25027c478bd9Sstevel@tonic-gate } while (num_retry-- > 0); 25037c478bd9Sstevel@tonic-gate error = e.error ? e.error : geterrno4(e.stat); 25047c478bd9Sstevel@tonic-gate if (error) { 25057c478bd9Sstevel@tonic-gate nfs_cmn_err(error, CE_WARN, 2506b9238976Sth VERS_MSG "initial call to %s failed%s: %m", 2507b9238976Sth svp->sv_hostname, droptext); 25087c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_WRITER, 0); 25097c478bd9Sstevel@tonic-gate svp->sv_flags |= SV4_NOTINUSE; 25107c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 25117c478bd9Sstevel@tonic-gate mi->mi_flags &= ~MI4_RECOV_FAIL; 25127c478bd9Sstevel@tonic-gate mi->mi_error = 0; 25137c478bd9Sstevel@tonic-gate continue; 25147c478bd9Sstevel@tonic-gate } 25157c478bd9Sstevel@tonic-gate 25167c478bd9Sstevel@tonic-gate if (tmp_vtype == VBAD) { 25177c478bd9Sstevel@tonic-gate zcmn_err(mi->mi_zone->zone_id, CE_WARN, 2518b9238976Sth VERS_MSG "%s returned a bad file type for " 2519b9238976Sth "root%s", svp->sv_hostname, droptext); 25207c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_WRITER, 0); 25217c478bd9Sstevel@tonic-gate svp->sv_flags |= SV4_NOTINUSE; 25227c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 25237c478bd9Sstevel@tonic-gate continue; 25247c478bd9Sstevel@tonic-gate } 25257c478bd9Sstevel@tonic-gate 25267c478bd9Sstevel@tonic-gate if (vtype == VNON) { 25277c478bd9Sstevel@tonic-gate vtype = tmp_vtype; 25287c478bd9Sstevel@tonic-gate } else if (vtype != tmp_vtype) { 25297c478bd9Sstevel@tonic-gate zcmn_err(mi->mi_zone->zone_id, CE_WARN, 2530b9238976Sth VERS_MSG "%s returned a different file type " 2531b9238976Sth "for root%s", svp->sv_hostname, droptext); 25327c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_WRITER, 0); 25337c478bd9Sstevel@tonic-gate svp->sv_flags |= SV4_NOTINUSE; 25347c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 25357c478bd9Sstevel@tonic-gate continue; 25367c478bd9Sstevel@tonic-gate } 25377c478bd9Sstevel@tonic-gate if (firstsvp == NULL) 25387c478bd9Sstevel@tonic-gate firstsvp = svp; 25397c478bd9Sstevel@tonic-gate } 25407c478bd9Sstevel@tonic-gate 25417c478bd9Sstevel@tonic-gate if (firstsvp == NULL) { 25427c478bd9Sstevel@tonic-gate if (error == 0) 25437c478bd9Sstevel@tonic-gate error = ENOENT; 25447c478bd9Sstevel@tonic-gate goto bad; 25457c478bd9Sstevel@tonic-gate } 25467c478bd9Sstevel@tonic-gate 25477c478bd9Sstevel@tonic-gate mi->mi_curr_serv = svp = firstsvp; 25487c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_READER, 0); 25497c478bd9Sstevel@tonic-gate ASSERT((mi->mi_curr_serv->sv_flags & SV4_NOTINUSE) == 0); 25507c478bd9Sstevel@tonic-gate fh.nfs_fh4_len = svp->sv_fhandle.fh_len; 25517c478bd9Sstevel@tonic-gate fh.nfs_fh4_val = svp->sv_fhandle.fh_buf; 25527c478bd9Sstevel@tonic-gate mi->mi_rootfh = sfh4_get(&fh, mi); 25537c478bd9Sstevel@tonic-gate fh.nfs_fh4_len = svp->sv_pfhandle.fh_len; 25547c478bd9Sstevel@tonic-gate fh.nfs_fh4_val = svp->sv_pfhandle.fh_buf; 25557c478bd9Sstevel@tonic-gate mi->mi_srvparentfh = sfh4_get(&fh, mi); 25567c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 25577c478bd9Sstevel@tonic-gate 25587c478bd9Sstevel@tonic-gate /* 2559bbf2a467SNagakiran Rajashekar * Get the fname for filesystem root. 25607c478bd9Sstevel@tonic-gate */ 2561bbf2a467SNagakiran Rajashekar mi->mi_fname = fn_get(NULL, ".", mi->mi_rootfh); 25627c478bd9Sstevel@tonic-gate mfname = mi->mi_fname; 25637c478bd9Sstevel@tonic-gate fn_hold(mfname); 2564bbf2a467SNagakiran Rajashekar 2565bbf2a467SNagakiran Rajashekar /* 2566bbf2a467SNagakiran Rajashekar * Make the root vnode without attributes. 2567bbf2a467SNagakiran Rajashekar */ 25687c478bd9Sstevel@tonic-gate rtvp = makenfs4node_by_fh(mi->mi_rootfh, NULL, 25697c478bd9Sstevel@tonic-gate &mfname, NULL, mi, cr, gethrtime()); 25707c478bd9Sstevel@tonic-gate rtvp->v_type = vtype; 25717c478bd9Sstevel@tonic-gate 25727c478bd9Sstevel@tonic-gate mi->mi_curread = mi->mi_tsize; 25737c478bd9Sstevel@tonic-gate mi->mi_curwrite = mi->mi_stsize; 25747c478bd9Sstevel@tonic-gate 25757c478bd9Sstevel@tonic-gate /* 25767c478bd9Sstevel@tonic-gate * Start the manager thread responsible for handling async worker 25777c478bd9Sstevel@tonic-gate * threads. 25787c478bd9Sstevel@tonic-gate */ 257950a83466Sjwahlig MI4_HOLD(mi); 25807c478bd9Sstevel@tonic-gate VFS_HOLD(vfsp); /* add reference for thread */ 25817c478bd9Sstevel@tonic-gate mi->mi_manager_thread = zthread_create(NULL, 0, nfs4_async_manager, 2582b9238976Sth vfsp, 0, minclsyspri); 25837c478bd9Sstevel@tonic-gate ASSERT(mi->mi_manager_thread != NULL); 258450a83466Sjwahlig 25857c478bd9Sstevel@tonic-gate /* 25867c478bd9Sstevel@tonic-gate * Create the thread that handles over-the-wire calls for 25877c478bd9Sstevel@tonic-gate * VOP_INACTIVE. 25887c478bd9Sstevel@tonic-gate * This needs to happen after the manager thread is created. 25897c478bd9Sstevel@tonic-gate */ 259050a83466Sjwahlig MI4_HOLD(mi); 25917c478bd9Sstevel@tonic-gate mi->mi_inactive_thread = zthread_create(NULL, 0, nfs4_inactive_thread, 2592b9238976Sth mi, 0, minclsyspri); 25937c478bd9Sstevel@tonic-gate ASSERT(mi->mi_inactive_thread != NULL); 25947c478bd9Sstevel@tonic-gate 25957c478bd9Sstevel@tonic-gate /* If we didn't get a type, get one now */ 25967c478bd9Sstevel@tonic-gate if (rtvp->v_type == VNON) { 25977c478bd9Sstevel@tonic-gate va.va_mask = AT_TYPE; 25987c478bd9Sstevel@tonic-gate error = nfs4getattr(rtvp, &va, tcr); 25997c478bd9Sstevel@tonic-gate if (error) 26007c478bd9Sstevel@tonic-gate goto bad; 26017c478bd9Sstevel@tonic-gate rtvp->v_type = va.va_type; 26027c478bd9Sstevel@tonic-gate } 26037c478bd9Sstevel@tonic-gate 26047c478bd9Sstevel@tonic-gate mi->mi_type = rtvp->v_type; 26057c478bd9Sstevel@tonic-gate 26067c478bd9Sstevel@tonic-gate mutex_enter(&mi->mi_lock); 26077c478bd9Sstevel@tonic-gate mi->mi_flags &= ~MI4_MOUNTING; 26087c478bd9Sstevel@tonic-gate mutex_exit(&mi->mi_lock); 26097c478bd9Sstevel@tonic-gate 26102f172c55SRobert Thurlow /* Update VFS with new server and path info */ 26112f172c55SRobert Thurlow if ((strcmp(svp->sv_hostname, origsvp->sv_hostname) != 0) || 26122f172c55SRobert Thurlow (strcmp(svp->sv_path, origsvp->sv_path) != 0)) { 26132f172c55SRobert Thurlow len = svp->sv_hostnamelen + svp->sv_pathlen; 26142f172c55SRobert Thurlow resource = kmem_zalloc(len, KM_SLEEP); 26152f172c55SRobert Thurlow (void) strcat(resource, svp->sv_hostname); 26162f172c55SRobert Thurlow (void) strcat(resource, ":"); 26172f172c55SRobert Thurlow (void) strcat(resource, svp->sv_path); 2618*d7de0ceaSRobert Harris vfs_setresource(vfsp, resource, 0); 26192f172c55SRobert Thurlow kmem_free(resource, len); 26202f172c55SRobert Thurlow } 26212f172c55SRobert Thurlow 26222f172c55SRobert Thurlow sv4_free(origsvp); 26237c478bd9Sstevel@tonic-gate *rtvpp = rtvp; 26247c478bd9Sstevel@tonic-gate if (lcr != NULL) 26257c478bd9Sstevel@tonic-gate crfree(lcr); 26267c478bd9Sstevel@tonic-gate 26277c478bd9Sstevel@tonic-gate return (0); 26287c478bd9Sstevel@tonic-gate bad: 26297c478bd9Sstevel@tonic-gate /* 26307c478bd9Sstevel@tonic-gate * An error occurred somewhere, need to clean up... 26317c478bd9Sstevel@tonic-gate */ 26327c478bd9Sstevel@tonic-gate if (lcr != NULL) 26337c478bd9Sstevel@tonic-gate crfree(lcr); 2634b9238976Sth 26357c478bd9Sstevel@tonic-gate if (rtvp != NULL) { 26367c478bd9Sstevel@tonic-gate /* 26377c478bd9Sstevel@tonic-gate * We need to release our reference to the root vnode and 26387c478bd9Sstevel@tonic-gate * destroy the mntinfo4 struct that we just created. 26397c478bd9Sstevel@tonic-gate */ 26407c478bd9Sstevel@tonic-gate rp = VTOR4(rtvp); 26417c478bd9Sstevel@tonic-gate if (rp->r_flags & R4HASHED) 26427c478bd9Sstevel@tonic-gate rp4_rmhash(rp); 26437c478bd9Sstevel@tonic-gate VN_RELE(rtvp); 26447c478bd9Sstevel@tonic-gate } 26457c478bd9Sstevel@tonic-gate nfs4_async_stop(vfsp); 26467c478bd9Sstevel@tonic-gate nfs4_async_manager_stop(vfsp); 264750a83466Sjwahlig removed = nfs4_mi_zonelist_remove(mi); 264850a83466Sjwahlig if (removed) 264950a83466Sjwahlig zone_rele(mi->mi_zone); 265050a83466Sjwahlig 265150a83466Sjwahlig /* 265250a83466Sjwahlig * This releases the initial "hold" of the mi since it will never 265350a83466Sjwahlig * be referenced by the vfsp. Also, when mount returns to vfs.c 265450a83466Sjwahlig * with an error, the vfsp will be destroyed, not rele'd. 265550a83466Sjwahlig */ 265650a83466Sjwahlig MI4_RELE(mi); 265750a83466Sjwahlig 26582f172c55SRobert Thurlow if (origsvp != NULL) 26592f172c55SRobert Thurlow sv4_free(origsvp); 26602f172c55SRobert Thurlow 26617c478bd9Sstevel@tonic-gate *rtvpp = NULL; 26627c478bd9Sstevel@tonic-gate return (error); 26637c478bd9Sstevel@tonic-gate } 26647c478bd9Sstevel@tonic-gate 26657c478bd9Sstevel@tonic-gate /* 26667c478bd9Sstevel@tonic-gate * vfs operations 26677c478bd9Sstevel@tonic-gate */ 26687c478bd9Sstevel@tonic-gate static int 26697c478bd9Sstevel@tonic-gate nfs4_unmount(vfs_t *vfsp, int flag, cred_t *cr) 26707c478bd9Sstevel@tonic-gate { 2671b9238976Sth mntinfo4_t *mi; 2672b9238976Sth ushort_t omax; 2673b9238976Sth int removed; 2674b9238976Sth 2675d3a14591SThomas Haynes bool_t must_unlock; 2676b9238976Sth 2677b9238976Sth nfs4_ephemeral_tree_t *eph_tree; 26787c478bd9Sstevel@tonic-gate 26797c478bd9Sstevel@tonic-gate if (secpolicy_fs_unmount(cr, vfsp) != 0) 26807c478bd9Sstevel@tonic-gate return (EPERM); 26817c478bd9Sstevel@tonic-gate 26827c478bd9Sstevel@tonic-gate mi = VFTOMI4(vfsp); 26837c478bd9Sstevel@tonic-gate 26847c478bd9Sstevel@tonic-gate if (flag & MS_FORCE) { 26857c478bd9Sstevel@tonic-gate vfsp->vfs_flag |= VFS_UNMOUNTED; 2686108322fbScarlsonj if (nfs_zone() != mi->mi_zone) { 26877c478bd9Sstevel@tonic-gate /* 26887c478bd9Sstevel@tonic-gate * If the request is coming from the wrong zone, 26897c478bd9Sstevel@tonic-gate * we don't want to create any new threads, and 26907c478bd9Sstevel@tonic-gate * performance is not a concern. Do everything 26917c478bd9Sstevel@tonic-gate * inline. 26927c478bd9Sstevel@tonic-gate */ 26937c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_zone_debug, (CE_NOTE, 26947c478bd9Sstevel@tonic-gate "nfs4_unmount x-zone forced unmount of vfs %p\n", 26957c478bd9Sstevel@tonic-gate (void *)vfsp)); 2696b9238976Sth nfs4_free_mount(vfsp, flag, cr); 26977c478bd9Sstevel@tonic-gate } else { 26987c478bd9Sstevel@tonic-gate /* 26997c478bd9Sstevel@tonic-gate * Free data structures asynchronously, to avoid 27007c478bd9Sstevel@tonic-gate * blocking the current thread (for performance 27017c478bd9Sstevel@tonic-gate * reasons only). 27027c478bd9Sstevel@tonic-gate */ 2703b9238976Sth async_free_mount(vfsp, flag, cr); 27047c478bd9Sstevel@tonic-gate } 2705b9238976Sth 27067c478bd9Sstevel@tonic-gate return (0); 27077c478bd9Sstevel@tonic-gate } 2708b9238976Sth 27097c478bd9Sstevel@tonic-gate /* 27107c478bd9Sstevel@tonic-gate * Wait until all asynchronous putpage operations on 27117c478bd9Sstevel@tonic-gate * this file system are complete before flushing rnodes 27127c478bd9Sstevel@tonic-gate * from the cache. 27137c478bd9Sstevel@tonic-gate */ 27147c478bd9Sstevel@tonic-gate omax = mi->mi_max_threads; 2715b9238976Sth if (nfs4_async_stop_sig(vfsp)) 27167c478bd9Sstevel@tonic-gate return (EINTR); 2717b9238976Sth 27187c478bd9Sstevel@tonic-gate r4flush(vfsp, cr); 2719b9238976Sth 2720eabd0450Sth /* 2721eabd0450Sth * About the only reason that this would fail would be 2722eabd0450Sth * that the harvester is already busy tearing down this 2723eabd0450Sth * node. So we fail back to the caller and let them try 2724eabd0450Sth * again when needed. 2725eabd0450Sth */ 2726eabd0450Sth if (nfs4_ephemeral_umount(mi, flag, cr, 27272f172c55SRobert Thurlow &must_unlock, &eph_tree)) { 2728d3a14591SThomas Haynes ASSERT(must_unlock == FALSE); 2729eabd0450Sth mutex_enter(&mi->mi_async_lock); 2730eabd0450Sth mi->mi_max_threads = omax; 2731eabd0450Sth mutex_exit(&mi->mi_async_lock); 2732eabd0450Sth 2733eabd0450Sth return (EBUSY); 2734eabd0450Sth } 2735b9238976Sth 27367c478bd9Sstevel@tonic-gate /* 27377c478bd9Sstevel@tonic-gate * If there are any active vnodes on this file system, 2738b9238976Sth * then the file system is busy and can't be unmounted. 27397c478bd9Sstevel@tonic-gate */ 27407c478bd9Sstevel@tonic-gate if (check_rtable4(vfsp)) { 27412f172c55SRobert Thurlow nfs4_ephemeral_umount_unlock(&must_unlock, &eph_tree); 2742b9238976Sth 27437c478bd9Sstevel@tonic-gate mutex_enter(&mi->mi_async_lock); 27447c478bd9Sstevel@tonic-gate mi->mi_max_threads = omax; 27457c478bd9Sstevel@tonic-gate mutex_exit(&mi->mi_async_lock); 2746b9238976Sth 27477c478bd9Sstevel@tonic-gate return (EBUSY); 27487c478bd9Sstevel@tonic-gate } 2749b9238976Sth 2750b9238976Sth /* 2751b9238976Sth * The unmount can't fail from now on, so record any 2752b9238976Sth * ephemeral changes. 2753b9238976Sth */ 27542f172c55SRobert Thurlow nfs4_ephemeral_umount_activate(mi, &must_unlock, &eph_tree); 2755b9238976Sth 27567c478bd9Sstevel@tonic-gate /* 2757b9238976Sth * There are no active files that could require over-the-wire 2758b9238976Sth * calls to the server, so stop the async manager and the 2759b9238976Sth * inactive thread. 27607c478bd9Sstevel@tonic-gate */ 27617c478bd9Sstevel@tonic-gate nfs4_async_manager_stop(vfsp); 2762b9238976Sth 27637c478bd9Sstevel@tonic-gate /* 27647c478bd9Sstevel@tonic-gate * Destroy all rnodes belonging to this file system from the 27657c478bd9Sstevel@tonic-gate * rnode hash queues and purge any resources allocated to 27667c478bd9Sstevel@tonic-gate * them. 27677c478bd9Sstevel@tonic-gate */ 27687c478bd9Sstevel@tonic-gate destroy_rtable4(vfsp, cr); 27697c478bd9Sstevel@tonic-gate vfsp->vfs_flag |= VFS_UNMOUNTED; 277050a83466Sjwahlig 27717c478bd9Sstevel@tonic-gate nfs4_remove_mi_from_server(mi, NULL); 277250a83466Sjwahlig removed = nfs4_mi_zonelist_remove(mi); 277350a83466Sjwahlig if (removed) 277450a83466Sjwahlig zone_rele(mi->mi_zone); 277550a83466Sjwahlig 27767c478bd9Sstevel@tonic-gate return (0); 27777c478bd9Sstevel@tonic-gate } 27787c478bd9Sstevel@tonic-gate 27797c478bd9Sstevel@tonic-gate /* 27807c478bd9Sstevel@tonic-gate * find root of nfs 27817c478bd9Sstevel@tonic-gate */ 27827c478bd9Sstevel@tonic-gate static int 27837c478bd9Sstevel@tonic-gate nfs4_root(vfs_t *vfsp, vnode_t **vpp) 27847c478bd9Sstevel@tonic-gate { 27857c478bd9Sstevel@tonic-gate mntinfo4_t *mi; 27867c478bd9Sstevel@tonic-gate vnode_t *vp; 27877c478bd9Sstevel@tonic-gate nfs4_fname_t *mfname; 27887c478bd9Sstevel@tonic-gate servinfo4_t *svp; 27897c478bd9Sstevel@tonic-gate 27907c478bd9Sstevel@tonic-gate mi = VFTOMI4(vfsp); 27917c478bd9Sstevel@tonic-gate 2792108322fbScarlsonj if (nfs_zone() != mi->mi_zone) 27937c478bd9Sstevel@tonic-gate return (EPERM); 27947c478bd9Sstevel@tonic-gate 27957c478bd9Sstevel@tonic-gate svp = mi->mi_curr_serv; 27967c478bd9Sstevel@tonic-gate if (svp) { 27977c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_READER, 0); 27987c478bd9Sstevel@tonic-gate if (svp->sv_flags & SV4_ROOT_STALE) { 27997c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 28007c478bd9Sstevel@tonic-gate 28017c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_WRITER, 0); 28027c478bd9Sstevel@tonic-gate if (svp->sv_flags & SV4_ROOT_STALE) { 28037c478bd9Sstevel@tonic-gate svp->sv_flags &= ~SV4_ROOT_STALE; 28047c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 28057c478bd9Sstevel@tonic-gate return (ENOENT); 28067c478bd9Sstevel@tonic-gate } 28077c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 28087c478bd9Sstevel@tonic-gate } else 28097c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 28107c478bd9Sstevel@tonic-gate } 28117c478bd9Sstevel@tonic-gate 28127c478bd9Sstevel@tonic-gate mfname = mi->mi_fname; 28137c478bd9Sstevel@tonic-gate fn_hold(mfname); 28147c478bd9Sstevel@tonic-gate vp = makenfs4node_by_fh(mi->mi_rootfh, NULL, &mfname, NULL, 28157c478bd9Sstevel@tonic-gate VFTOMI4(vfsp), CRED(), gethrtime()); 28167c478bd9Sstevel@tonic-gate 28177c478bd9Sstevel@tonic-gate if (VTOR4(vp)->r_flags & R4STALE) { 28187c478bd9Sstevel@tonic-gate VN_RELE(vp); 28197c478bd9Sstevel@tonic-gate return (ENOENT); 28207c478bd9Sstevel@tonic-gate } 28217c478bd9Sstevel@tonic-gate 28227c478bd9Sstevel@tonic-gate ASSERT(vp->v_type == VNON || vp->v_type == mi->mi_type); 28237c478bd9Sstevel@tonic-gate 28247c478bd9Sstevel@tonic-gate vp->v_type = mi->mi_type; 28257c478bd9Sstevel@tonic-gate 28267c478bd9Sstevel@tonic-gate *vpp = vp; 28277c478bd9Sstevel@tonic-gate 28287c478bd9Sstevel@tonic-gate return (0); 28297c478bd9Sstevel@tonic-gate } 28307c478bd9Sstevel@tonic-gate 28317c478bd9Sstevel@tonic-gate static int 28327c478bd9Sstevel@tonic-gate nfs4_statfs_otw(vnode_t *vp, struct statvfs64 *sbp, cred_t *cr) 28337c478bd9Sstevel@tonic-gate { 28347c478bd9Sstevel@tonic-gate int error; 28357c478bd9Sstevel@tonic-gate nfs4_ga_res_t gar; 28367c478bd9Sstevel@tonic-gate nfs4_ga_ext_res_t ger; 28377c478bd9Sstevel@tonic-gate 28387c478bd9Sstevel@tonic-gate gar.n4g_ext_res = &ger; 28397c478bd9Sstevel@tonic-gate 28407c478bd9Sstevel@tonic-gate if (error = nfs4_attr_otw(vp, TAG_FSINFO, &gar, 28417c478bd9Sstevel@tonic-gate NFS4_STATFS_ATTR_MASK, cr)) 28427c478bd9Sstevel@tonic-gate return (error); 28437c478bd9Sstevel@tonic-gate 28447c478bd9Sstevel@tonic-gate *sbp = gar.n4g_ext_res->n4g_sb; 28457c478bd9Sstevel@tonic-gate 28467c478bd9Sstevel@tonic-gate return (0); 28477c478bd9Sstevel@tonic-gate } 28487c478bd9Sstevel@tonic-gate 28497c478bd9Sstevel@tonic-gate /* 28507c478bd9Sstevel@tonic-gate * Get file system statistics. 28517c478bd9Sstevel@tonic-gate */ 28527c478bd9Sstevel@tonic-gate static int 28537c478bd9Sstevel@tonic-gate nfs4_statvfs(vfs_t *vfsp, struct statvfs64 *sbp) 28547c478bd9Sstevel@tonic-gate { 28557c478bd9Sstevel@tonic-gate int error; 28567c478bd9Sstevel@tonic-gate vnode_t *vp; 28577c478bd9Sstevel@tonic-gate cred_t *cr; 28587c478bd9Sstevel@tonic-gate 28597c478bd9Sstevel@tonic-gate error = nfs4_root(vfsp, &vp); 28607c478bd9Sstevel@tonic-gate if (error) 28617c478bd9Sstevel@tonic-gate return (error); 28627c478bd9Sstevel@tonic-gate 28637c478bd9Sstevel@tonic-gate cr = CRED(); 28647c478bd9Sstevel@tonic-gate 28657c478bd9Sstevel@tonic-gate error = nfs4_statfs_otw(vp, sbp, cr); 28667c478bd9Sstevel@tonic-gate if (!error) { 28677c478bd9Sstevel@tonic-gate (void) strncpy(sbp->f_basetype, 2868b9238976Sth vfssw[vfsp->vfs_fstype].vsw_name, FSTYPSZ); 28697c478bd9Sstevel@tonic-gate sbp->f_flag = vf_to_stf(vfsp->vfs_flag); 28707c478bd9Sstevel@tonic-gate } else { 28717c478bd9Sstevel@tonic-gate nfs4_purge_stale_fh(error, vp, cr); 28727c478bd9Sstevel@tonic-gate } 28737c478bd9Sstevel@tonic-gate 28747c478bd9Sstevel@tonic-gate VN_RELE(vp); 28757c478bd9Sstevel@tonic-gate 28767c478bd9Sstevel@tonic-gate return (error); 28777c478bd9Sstevel@tonic-gate } 28787c478bd9Sstevel@tonic-gate 28797c478bd9Sstevel@tonic-gate static kmutex_t nfs4_syncbusy; 28807c478bd9Sstevel@tonic-gate 28817c478bd9Sstevel@tonic-gate /* 28827c478bd9Sstevel@tonic-gate * Flush dirty nfs files for file system vfsp. 28837c478bd9Sstevel@tonic-gate * If vfsp == NULL, all nfs files are flushed. 28847c478bd9Sstevel@tonic-gate * 28857c478bd9Sstevel@tonic-gate * SYNC_CLOSE in flag is passed to us to 28867c478bd9Sstevel@tonic-gate * indicate that we are shutting down and or 28877c478bd9Sstevel@tonic-gate * rebooting. 28887c478bd9Sstevel@tonic-gate */ 28897c478bd9Sstevel@tonic-gate static int 28907c478bd9Sstevel@tonic-gate nfs4_sync(vfs_t *vfsp, short flag, cred_t *cr) 28917c478bd9Sstevel@tonic-gate { 28927c478bd9Sstevel@tonic-gate /* 28937c478bd9Sstevel@tonic-gate * Cross-zone calls are OK here, since this translates to a 28947c478bd9Sstevel@tonic-gate * VOP_PUTPAGE(B_ASYNC), which gets picked up by the right zone. 28957c478bd9Sstevel@tonic-gate */ 28967c478bd9Sstevel@tonic-gate if (!(flag & SYNC_ATTR) && mutex_tryenter(&nfs4_syncbusy) != 0) { 28977c478bd9Sstevel@tonic-gate r4flush(vfsp, cr); 28987c478bd9Sstevel@tonic-gate mutex_exit(&nfs4_syncbusy); 28997c478bd9Sstevel@tonic-gate } 29007c478bd9Sstevel@tonic-gate 29017c478bd9Sstevel@tonic-gate /* 29027c478bd9Sstevel@tonic-gate * if SYNC_CLOSE is set then we know that 29037c478bd9Sstevel@tonic-gate * the system is rebooting, mark the mntinfo 29047c478bd9Sstevel@tonic-gate * for later examination. 29057c478bd9Sstevel@tonic-gate */ 29067c478bd9Sstevel@tonic-gate if (vfsp && (flag & SYNC_CLOSE)) { 29077c478bd9Sstevel@tonic-gate mntinfo4_t *mi; 29087c478bd9Sstevel@tonic-gate 29097c478bd9Sstevel@tonic-gate mi = VFTOMI4(vfsp); 29107c478bd9Sstevel@tonic-gate if (!(mi->mi_flags & MI4_SHUTDOWN)) { 29117c478bd9Sstevel@tonic-gate mutex_enter(&mi->mi_lock); 29127c478bd9Sstevel@tonic-gate mi->mi_flags |= MI4_SHUTDOWN; 29137c478bd9Sstevel@tonic-gate mutex_exit(&mi->mi_lock); 29147c478bd9Sstevel@tonic-gate } 29157c478bd9Sstevel@tonic-gate } 29167c478bd9Sstevel@tonic-gate return (0); 29177c478bd9Sstevel@tonic-gate } 29187c478bd9Sstevel@tonic-gate 29197c478bd9Sstevel@tonic-gate /* 29207c478bd9Sstevel@tonic-gate * vget is difficult, if not impossible, to support in v4 because we don't 29217c478bd9Sstevel@tonic-gate * know the parent directory or name, which makes it impossible to create a 29227c478bd9Sstevel@tonic-gate * useful shadow vnode. And we need the shadow vnode for things like 29237c478bd9Sstevel@tonic-gate * OPEN. 29247c478bd9Sstevel@tonic-gate */ 29257c478bd9Sstevel@tonic-gate 29267c478bd9Sstevel@tonic-gate /* ARGSUSED */ 29277c478bd9Sstevel@tonic-gate /* 29287c478bd9Sstevel@tonic-gate * XXX Check nfs4_vget_pseudo() for dependency. 29297c478bd9Sstevel@tonic-gate */ 29307c478bd9Sstevel@tonic-gate static int 29317c478bd9Sstevel@tonic-gate nfs4_vget(vfs_t *vfsp, vnode_t **vpp, fid_t *fidp) 29327c478bd9Sstevel@tonic-gate { 29337c478bd9Sstevel@tonic-gate return (EREMOTE); 29347c478bd9Sstevel@tonic-gate } 29357c478bd9Sstevel@tonic-gate 29367c478bd9Sstevel@tonic-gate /* 29377c478bd9Sstevel@tonic-gate * nfs4_mountroot get called in the case where we are diskless booting. All 29387c478bd9Sstevel@tonic-gate * we need from here is the ability to get the server info and from there we 29397c478bd9Sstevel@tonic-gate * can simply call nfs4_rootvp. 29407c478bd9Sstevel@tonic-gate */ 29417c478bd9Sstevel@tonic-gate /* ARGSUSED */ 29427c478bd9Sstevel@tonic-gate static int 29437c478bd9Sstevel@tonic-gate nfs4_mountroot(vfs_t *vfsp, whymountroot_t why) 29447c478bd9Sstevel@tonic-gate { 29457c478bd9Sstevel@tonic-gate vnode_t *rtvp; 29467c478bd9Sstevel@tonic-gate char root_hostname[SYS_NMLN+1]; 29477c478bd9Sstevel@tonic-gate struct servinfo4 *svp; 29487c478bd9Sstevel@tonic-gate int error; 29497c478bd9Sstevel@tonic-gate int vfsflags; 29507c478bd9Sstevel@tonic-gate size_t size; 29517c478bd9Sstevel@tonic-gate char *root_path; 29527c478bd9Sstevel@tonic-gate struct pathname pn; 29537c478bd9Sstevel@tonic-gate char *name; 29547c478bd9Sstevel@tonic-gate cred_t *cr; 29557c478bd9Sstevel@tonic-gate mntinfo4_t *mi; 29567c478bd9Sstevel@tonic-gate struct nfs_args args; /* nfs mount arguments */ 29577c478bd9Sstevel@tonic-gate static char token[10]; 29587c478bd9Sstevel@tonic-gate nfs4_error_t n4e; 29597c478bd9Sstevel@tonic-gate 29607c478bd9Sstevel@tonic-gate bzero(&args, sizeof (args)); 29617c478bd9Sstevel@tonic-gate 29627c478bd9Sstevel@tonic-gate /* do this BEFORE getfile which causes xid stamps to be initialized */ 29637c478bd9Sstevel@tonic-gate clkset(-1L); /* hack for now - until we get time svc? */ 29647c478bd9Sstevel@tonic-gate 29657c478bd9Sstevel@tonic-gate if (why == ROOT_REMOUNT) { 29667c478bd9Sstevel@tonic-gate /* 29677c478bd9Sstevel@tonic-gate * Shouldn't happen. 29687c478bd9Sstevel@tonic-gate */ 29697c478bd9Sstevel@tonic-gate panic("nfs4_mountroot: why == ROOT_REMOUNT"); 29707c478bd9Sstevel@tonic-gate } 29717c478bd9Sstevel@tonic-gate 29727c478bd9Sstevel@tonic-gate if (why == ROOT_UNMOUNT) { 29737c478bd9Sstevel@tonic-gate /* 29747c478bd9Sstevel@tonic-gate * Nothing to do for NFS. 29757c478bd9Sstevel@tonic-gate */ 29767c478bd9Sstevel@tonic-gate return (0); 29777c478bd9Sstevel@tonic-gate } 29787c478bd9Sstevel@tonic-gate 29797c478bd9Sstevel@tonic-gate /* 29807c478bd9Sstevel@tonic-gate * why == ROOT_INIT 29817c478bd9Sstevel@tonic-gate */ 29827c478bd9Sstevel@tonic-gate 29837c478bd9Sstevel@tonic-gate name = token; 29847c478bd9Sstevel@tonic-gate *name = 0; 29857c478bd9Sstevel@tonic-gate (void) getfsname("root", name, sizeof (token)); 29867c478bd9Sstevel@tonic-gate 29877c478bd9Sstevel@tonic-gate pn_alloc(&pn); 29887c478bd9Sstevel@tonic-gate root_path = pn.pn_path; 29897c478bd9Sstevel@tonic-gate 29907c478bd9Sstevel@tonic-gate svp = kmem_zalloc(sizeof (*svp), KM_SLEEP); 29917c478bd9Sstevel@tonic-gate nfs_rw_init(&svp->sv_lock, NULL, RW_DEFAULT, NULL); 29927c478bd9Sstevel@tonic-gate svp->sv_knconf = kmem_zalloc(sizeof (*svp->sv_knconf), KM_SLEEP); 29937c478bd9Sstevel@tonic-gate svp->sv_knconf->knc_protofmly = kmem_alloc(KNC_STRSIZE, KM_SLEEP); 29947c478bd9Sstevel@tonic-gate svp->sv_knconf->knc_proto = kmem_alloc(KNC_STRSIZE, KM_SLEEP); 29957c478bd9Sstevel@tonic-gate 29967c478bd9Sstevel@tonic-gate /* 29977c478bd9Sstevel@tonic-gate * Get server address 29987c478bd9Sstevel@tonic-gate * Get the root path 29997c478bd9Sstevel@tonic-gate * Get server's transport 30007c478bd9Sstevel@tonic-gate * Get server's hostname 30017c478bd9Sstevel@tonic-gate * Get options 30027c478bd9Sstevel@tonic-gate */ 30037c478bd9Sstevel@tonic-gate args.addr = &svp->sv_addr; 30047c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_READER, 0); 30057c478bd9Sstevel@tonic-gate args.fh = (char *)&svp->sv_fhandle; 30067c478bd9Sstevel@tonic-gate args.knconf = svp->sv_knconf; 30077c478bd9Sstevel@tonic-gate args.hostname = root_hostname; 30087c478bd9Sstevel@tonic-gate vfsflags = 0; 30097c478bd9Sstevel@tonic-gate if (error = mount_root(*name ? name : "root", root_path, NFS_V4, 30107c478bd9Sstevel@tonic-gate &args, &vfsflags)) { 30117c478bd9Sstevel@tonic-gate if (error == EPROTONOSUPPORT) 30127c478bd9Sstevel@tonic-gate nfs_cmn_err(error, CE_WARN, "nfs4_mountroot: " 30137c478bd9Sstevel@tonic-gate "mount_root failed: server doesn't support NFS V4"); 30147c478bd9Sstevel@tonic-gate else 30157c478bd9Sstevel@tonic-gate nfs_cmn_err(error, CE_WARN, 30167c478bd9Sstevel@tonic-gate "nfs4_mountroot: mount_root failed: %m"); 30177c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 30187c478bd9Sstevel@tonic-gate sv4_free(svp); 30197c478bd9Sstevel@tonic-gate pn_free(&pn); 30207c478bd9Sstevel@tonic-gate return (error); 30217c478bd9Sstevel@tonic-gate } 30227c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 30237c478bd9Sstevel@tonic-gate svp->sv_hostnamelen = (int)(strlen(root_hostname) + 1); 30247c478bd9Sstevel@tonic-gate svp->sv_hostname = kmem_alloc(svp->sv_hostnamelen, KM_SLEEP); 30257c478bd9Sstevel@tonic-gate (void) strcpy(svp->sv_hostname, root_hostname); 30267c478bd9Sstevel@tonic-gate 30277c478bd9Sstevel@tonic-gate svp->sv_pathlen = (int)(strlen(root_path) + 1); 30287c478bd9Sstevel@tonic-gate svp->sv_path = kmem_alloc(svp->sv_pathlen, KM_SLEEP); 30297c478bd9Sstevel@tonic-gate (void) strcpy(svp->sv_path, root_path); 30307c478bd9Sstevel@tonic-gate 30317c478bd9Sstevel@tonic-gate /* 30327c478bd9Sstevel@tonic-gate * Force root partition to always be mounted with AUTH_UNIX for now 30337c478bd9Sstevel@tonic-gate */ 30347c478bd9Sstevel@tonic-gate svp->sv_secdata = kmem_alloc(sizeof (*svp->sv_secdata), KM_SLEEP); 30357c478bd9Sstevel@tonic-gate svp->sv_secdata->secmod = AUTH_UNIX; 30367c478bd9Sstevel@tonic-gate svp->sv_secdata->rpcflavor = AUTH_UNIX; 30377c478bd9Sstevel@tonic-gate svp->sv_secdata->data = NULL; 30387c478bd9Sstevel@tonic-gate 30397c478bd9Sstevel@tonic-gate cr = crgetcred(); 30407c478bd9Sstevel@tonic-gate rtvp = NULL; 30417c478bd9Sstevel@tonic-gate 30427c478bd9Sstevel@tonic-gate error = nfs4rootvp(&rtvp, vfsp, svp, args.flags, cr, global_zone); 30437c478bd9Sstevel@tonic-gate 30447c478bd9Sstevel@tonic-gate if (error) { 30457c478bd9Sstevel@tonic-gate crfree(cr); 30467c478bd9Sstevel@tonic-gate pn_free(&pn); 3047ab7762b6Smaheshvs sv4_free(svp); 3048ab7762b6Smaheshvs return (error); 30497c478bd9Sstevel@tonic-gate } 30507c478bd9Sstevel@tonic-gate 30517c478bd9Sstevel@tonic-gate mi = VTOMI4(rtvp); 30527c478bd9Sstevel@tonic-gate 30537c478bd9Sstevel@tonic-gate /* 30547c478bd9Sstevel@tonic-gate * Send client id to the server, if necessary 30557c478bd9Sstevel@tonic-gate */ 30567c478bd9Sstevel@tonic-gate nfs4_error_zinit(&n4e); 30577c478bd9Sstevel@tonic-gate nfs4setclientid(mi, cr, FALSE, &n4e); 30587c478bd9Sstevel@tonic-gate error = n4e.error; 30597c478bd9Sstevel@tonic-gate 30607c478bd9Sstevel@tonic-gate crfree(cr); 30617c478bd9Sstevel@tonic-gate 30627c478bd9Sstevel@tonic-gate if (error) { 30637c478bd9Sstevel@tonic-gate pn_free(&pn); 30647c478bd9Sstevel@tonic-gate goto errout; 30657c478bd9Sstevel@tonic-gate } 30667c478bd9Sstevel@tonic-gate 30677c478bd9Sstevel@tonic-gate error = nfs4_setopts(rtvp, DATAMODEL_NATIVE, &args); 30687c478bd9Sstevel@tonic-gate if (error) { 30697c478bd9Sstevel@tonic-gate nfs_cmn_err(error, CE_WARN, 30707c478bd9Sstevel@tonic-gate "nfs4_mountroot: invalid root mount options"); 30717c478bd9Sstevel@tonic-gate pn_free(&pn); 30727c478bd9Sstevel@tonic-gate goto errout; 30737c478bd9Sstevel@tonic-gate } 30747c478bd9Sstevel@tonic-gate 30757c478bd9Sstevel@tonic-gate (void) vfs_lock_wait(vfsp); 30767c478bd9Sstevel@tonic-gate vfs_add(NULL, vfsp, vfsflags); 30777c478bd9Sstevel@tonic-gate vfs_unlock(vfsp); 30787c478bd9Sstevel@tonic-gate 30797c478bd9Sstevel@tonic-gate size = strlen(svp->sv_hostname); 30807c478bd9Sstevel@tonic-gate (void) strcpy(rootfs.bo_name, svp->sv_hostname); 30817c478bd9Sstevel@tonic-gate rootfs.bo_name[size] = ':'; 30827c478bd9Sstevel@tonic-gate (void) strcpy(&rootfs.bo_name[size + 1], root_path); 30837c478bd9Sstevel@tonic-gate 30847c478bd9Sstevel@tonic-gate pn_free(&pn); 30857c478bd9Sstevel@tonic-gate 30867c478bd9Sstevel@tonic-gate errout: 30877c478bd9Sstevel@tonic-gate if (error) { 30887c478bd9Sstevel@tonic-gate sv4_free(svp); 30897c478bd9Sstevel@tonic-gate nfs4_async_stop(vfsp); 30907c478bd9Sstevel@tonic-gate nfs4_async_manager_stop(vfsp); 30917c478bd9Sstevel@tonic-gate } 30927c478bd9Sstevel@tonic-gate 30937c478bd9Sstevel@tonic-gate if (rtvp != NULL) 30947c478bd9Sstevel@tonic-gate VN_RELE(rtvp); 30957c478bd9Sstevel@tonic-gate 30967c478bd9Sstevel@tonic-gate return (error); 30977c478bd9Sstevel@tonic-gate } 30987c478bd9Sstevel@tonic-gate 30997c478bd9Sstevel@tonic-gate /* 31007c478bd9Sstevel@tonic-gate * Initialization routine for VFS routines. Should only be called once 31017c478bd9Sstevel@tonic-gate */ 31027c478bd9Sstevel@tonic-gate int 31037c478bd9Sstevel@tonic-gate nfs4_vfsinit(void) 31047c478bd9Sstevel@tonic-gate { 31057c478bd9Sstevel@tonic-gate mutex_init(&nfs4_syncbusy, NULL, MUTEX_DEFAULT, NULL); 31067c478bd9Sstevel@tonic-gate nfs4setclientid_init(); 3107b9238976Sth nfs4_ephemeral_init(); 31087c478bd9Sstevel@tonic-gate return (0); 31097c478bd9Sstevel@tonic-gate } 31107c478bd9Sstevel@tonic-gate 31117c478bd9Sstevel@tonic-gate void 31127c478bd9Sstevel@tonic-gate nfs4_vfsfini(void) 31137c478bd9Sstevel@tonic-gate { 3114b9238976Sth nfs4_ephemeral_fini(); 31157c478bd9Sstevel@tonic-gate nfs4setclientid_fini(); 31167c478bd9Sstevel@tonic-gate mutex_destroy(&nfs4_syncbusy); 31177c478bd9Sstevel@tonic-gate } 31187c478bd9Sstevel@tonic-gate 31197c478bd9Sstevel@tonic-gate void 31207c478bd9Sstevel@tonic-gate nfs4_freevfs(vfs_t *vfsp) 31217c478bd9Sstevel@tonic-gate { 31227c478bd9Sstevel@tonic-gate mntinfo4_t *mi; 31237c478bd9Sstevel@tonic-gate 312450a83466Sjwahlig /* need to release the initial hold */ 31257c478bd9Sstevel@tonic-gate mi = VFTOMI4(vfsp); 3126b87f76edSThomas Haynes 3127b87f76edSThomas Haynes /* 3128b87f76edSThomas Haynes * At this point, we can no longer reference the vfs 3129b87f76edSThomas Haynes * and need to inform other holders of the reference 3130b87f76edSThomas Haynes * to the mntinfo4_t. 3131b87f76edSThomas Haynes */ 3132b87f76edSThomas Haynes mi->mi_vfsp = NULL; 3133b87f76edSThomas Haynes 313450a83466Sjwahlig MI4_RELE(mi); 31357c478bd9Sstevel@tonic-gate } 31367c478bd9Sstevel@tonic-gate 31377c478bd9Sstevel@tonic-gate /* 31387c478bd9Sstevel@tonic-gate * Client side SETCLIENTID and SETCLIENTID_CONFIRM 31397c478bd9Sstevel@tonic-gate */ 31407c478bd9Sstevel@tonic-gate struct nfs4_server nfs4_server_lst = 31417c478bd9Sstevel@tonic-gate { &nfs4_server_lst, &nfs4_server_lst }; 31427c478bd9Sstevel@tonic-gate 31437c478bd9Sstevel@tonic-gate kmutex_t nfs4_server_lst_lock; 31447c478bd9Sstevel@tonic-gate 31457c478bd9Sstevel@tonic-gate static void 31467c478bd9Sstevel@tonic-gate nfs4setclientid_init(void) 31477c478bd9Sstevel@tonic-gate { 31487c478bd9Sstevel@tonic-gate mutex_init(&nfs4_server_lst_lock, NULL, MUTEX_DEFAULT, NULL); 31497c478bd9Sstevel@tonic-gate } 31507c478bd9Sstevel@tonic-gate 31517c478bd9Sstevel@tonic-gate static void 31527c478bd9Sstevel@tonic-gate nfs4setclientid_fini(void) 31537c478bd9Sstevel@tonic-gate { 31547c478bd9Sstevel@tonic-gate mutex_destroy(&nfs4_server_lst_lock); 31557c478bd9Sstevel@tonic-gate } 31567c478bd9Sstevel@tonic-gate 31577c478bd9Sstevel@tonic-gate int nfs4_retry_sclid_delay = NFS4_RETRY_SCLID_DELAY; 31587c478bd9Sstevel@tonic-gate int nfs4_num_sclid_retries = NFS4_NUM_SCLID_RETRIES; 31597c478bd9Sstevel@tonic-gate 31607c478bd9Sstevel@tonic-gate /* 31617c478bd9Sstevel@tonic-gate * Set the clientid for the server for "mi". No-op if the clientid is 31627c478bd9Sstevel@tonic-gate * already set. 31637c478bd9Sstevel@tonic-gate * 31647c478bd9Sstevel@tonic-gate * The recovery boolean should be set to TRUE if this function was called 3165a092743bSek * by the recovery code, and FALSE otherwise. This is used to determine 3166a092743bSek * if we need to call nfs4_start/end_op as well as grab the mi_recovlock 3167a092743bSek * for adding a mntinfo4_t to a nfs4_server_t. 31687c478bd9Sstevel@tonic-gate * 31697c478bd9Sstevel@tonic-gate * Error is returned via 'n4ep'. If there was a 'n4ep->stat' error, then 31707c478bd9Sstevel@tonic-gate * 'n4ep->error' is set to geterrno4(n4ep->stat). 31717c478bd9Sstevel@tonic-gate */ 31727c478bd9Sstevel@tonic-gate void 31737c478bd9Sstevel@tonic-gate nfs4setclientid(mntinfo4_t *mi, cred_t *cr, bool_t recovery, nfs4_error_t *n4ep) 31747c478bd9Sstevel@tonic-gate { 31757c478bd9Sstevel@tonic-gate struct nfs4_server *np; 31767c478bd9Sstevel@tonic-gate struct servinfo4 *svp = mi->mi_curr_serv; 31777c478bd9Sstevel@tonic-gate nfs4_recov_state_t recov_state; 31787c478bd9Sstevel@tonic-gate int num_retries = 0; 3179f64c4ae1Sdm bool_t retry; 31807c478bd9Sstevel@tonic-gate cred_t *lcr = NULL; 31817c478bd9Sstevel@tonic-gate int retry_inuse = 1; /* only retry once on NFS4ERR_CLID_INUSE */ 31827c478bd9Sstevel@tonic-gate time_t lease_time = 0; 31837c478bd9Sstevel@tonic-gate 31847c478bd9Sstevel@tonic-gate recov_state.rs_flags = 0; 31857c478bd9Sstevel@tonic-gate recov_state.rs_num_retry_despite_err = 0; 31867c478bd9Sstevel@tonic-gate ASSERT(n4ep != NULL); 31877c478bd9Sstevel@tonic-gate 31887c478bd9Sstevel@tonic-gate recov_retry: 3189f64c4ae1Sdm retry = FALSE; 31907c478bd9Sstevel@tonic-gate nfs4_error_zinit(n4ep); 3191a092743bSek if (!recovery) 3192a092743bSek (void) nfs_rw_enter_sig(&mi->mi_recovlock, RW_READER, 0); 3193a092743bSek 3194f86c6ccaSdm mutex_enter(&nfs4_server_lst_lock); 3195f86c6ccaSdm np = servinfo4_to_nfs4_server(svp); /* This locks np if it is found */ 3196f86c6ccaSdm mutex_exit(&nfs4_server_lst_lock); 3197f86c6ccaSdm if (!np) { 3198f86c6ccaSdm struct nfs4_server *tnp; 3199f86c6ccaSdm np = new_nfs4_server(svp, cr); 320016237317Sdm mutex_enter(&np->s_lock); 32017c478bd9Sstevel@tonic-gate 3202f86c6ccaSdm mutex_enter(&nfs4_server_lst_lock); 3203f86c6ccaSdm tnp = servinfo4_to_nfs4_server(svp); 3204f86c6ccaSdm if (tnp) { 3205f86c6ccaSdm /* 3206f86c6ccaSdm * another thread snuck in and put server on list. 3207f86c6ccaSdm * since we aren't adding it to the nfs4_server_list 3208f86c6ccaSdm * we need to set the ref count to 0 and destroy it. 3209f86c6ccaSdm */ 3210f86c6ccaSdm np->s_refcnt = 0; 3211f86c6ccaSdm destroy_nfs4_server(np); 3212f86c6ccaSdm np = tnp; 3213f86c6ccaSdm } else { 3214f86c6ccaSdm /* 3215f86c6ccaSdm * do not give list a reference until everything 3216f86c6ccaSdm * succeeds 3217f86c6ccaSdm */ 3218f86c6ccaSdm insque(np, &nfs4_server_lst); 3219f86c6ccaSdm } 3220f86c6ccaSdm mutex_exit(&nfs4_server_lst_lock); 3221f86c6ccaSdm } 3222f86c6ccaSdm ASSERT(MUTEX_HELD(&np->s_lock)); 32237c478bd9Sstevel@tonic-gate /* 3224f86c6ccaSdm * If we find the server already has N4S_CLIENTID_SET, then 3225f86c6ccaSdm * just return, we've already done SETCLIENTID to that server 32267c478bd9Sstevel@tonic-gate */ 3227f86c6ccaSdm if (np->s_flags & N4S_CLIENTID_SET) { 32287c478bd9Sstevel@tonic-gate /* add mi to np's mntinfo4_list */ 32297c478bd9Sstevel@tonic-gate nfs4_add_mi_to_server(np, mi); 3230a092743bSek if (!recovery) 3231a092743bSek nfs_rw_exit(&mi->mi_recovlock); 32327c478bd9Sstevel@tonic-gate mutex_exit(&np->s_lock); 32337c478bd9Sstevel@tonic-gate nfs4_server_rele(np); 32347c478bd9Sstevel@tonic-gate return; 32357c478bd9Sstevel@tonic-gate } 3236f86c6ccaSdm mutex_exit(&np->s_lock); 3237f86c6ccaSdm 32387c478bd9Sstevel@tonic-gate 3239a092743bSek /* 3240a092743bSek * Drop the mi_recovlock since nfs4_start_op will 3241a092743bSek * acquire it again for us. 3242a092743bSek */ 3243f86c6ccaSdm if (!recovery) { 3244a092743bSek nfs_rw_exit(&mi->mi_recovlock); 3245a092743bSek 32467c478bd9Sstevel@tonic-gate n4ep->error = nfs4_start_op(mi, NULL, NULL, &recov_state); 32477c478bd9Sstevel@tonic-gate if (n4ep->error) { 32487c478bd9Sstevel@tonic-gate nfs4_server_rele(np); 32497c478bd9Sstevel@tonic-gate return; 32507c478bd9Sstevel@tonic-gate } 32517c478bd9Sstevel@tonic-gate } 32527c478bd9Sstevel@tonic-gate 32537c478bd9Sstevel@tonic-gate mutex_enter(&np->s_lock); 3254f86c6ccaSdm while (np->s_flags & N4S_CLIENTID_PEND) { 3255f86c6ccaSdm if (!cv_wait_sig(&np->s_clientid_pend, &np->s_lock)) { 3256f86c6ccaSdm mutex_exit(&np->s_lock); 3257f86c6ccaSdm nfs4_server_rele(np); 3258f86c6ccaSdm if (!recovery) 3259f86c6ccaSdm nfs4_end_op(mi, NULL, NULL, &recov_state, 3260f86c6ccaSdm recovery); 3261f86c6ccaSdm n4ep->error = EINTR; 3262f86c6ccaSdm return; 3263f86c6ccaSdm } 3264f86c6ccaSdm } 32657c478bd9Sstevel@tonic-gate 32667c478bd9Sstevel@tonic-gate if (np->s_flags & N4S_CLIENTID_SET) { 32677c478bd9Sstevel@tonic-gate /* XXX copied/pasted from above */ 32687c478bd9Sstevel@tonic-gate /* add mi to np's mntinfo4_list */ 32697c478bd9Sstevel@tonic-gate nfs4_add_mi_to_server(np, mi); 32707c478bd9Sstevel@tonic-gate mutex_exit(&np->s_lock); 32717c478bd9Sstevel@tonic-gate nfs4_server_rele(np); 32727c478bd9Sstevel@tonic-gate if (!recovery) 32737c478bd9Sstevel@tonic-gate nfs4_end_op(mi, NULL, NULL, &recov_state, recovery); 32747c478bd9Sstevel@tonic-gate return; 32757c478bd9Sstevel@tonic-gate } 32767c478bd9Sstevel@tonic-gate 3277f86c6ccaSdm /* 3278f86c6ccaSdm * Reset the N4S_CB_PINGED flag. This is used to 3279f86c6ccaSdm * indicate if we have received a CB_NULL from the 3280f86c6ccaSdm * server. Also we reset the waiter flag. 3281f86c6ccaSdm */ 3282f86c6ccaSdm np->s_flags &= ~(N4S_CB_PINGED | N4S_CB_WAITER); 3283f86c6ccaSdm /* any failure must now clear this flag */ 3284f86c6ccaSdm np->s_flags |= N4S_CLIENTID_PEND; 3285f86c6ccaSdm mutex_exit(&np->s_lock); 32867c478bd9Sstevel@tonic-gate nfs4setclientid_otw(mi, svp, cr, np, n4ep, &retry_inuse); 32877c478bd9Sstevel@tonic-gate 32887c478bd9Sstevel@tonic-gate if (n4ep->error == EACCES) { 32897c478bd9Sstevel@tonic-gate /* 32907c478bd9Sstevel@tonic-gate * If the uid is set then set the creds for secure mounts 32917c478bd9Sstevel@tonic-gate * by proxy processes such as automountd. 32927c478bd9Sstevel@tonic-gate */ 32937c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&svp->sv_lock, RW_READER, 0); 32947c478bd9Sstevel@tonic-gate if (svp->sv_secdata->uid != 0) { 32957c478bd9Sstevel@tonic-gate lcr = crdup(cr); 32967c478bd9Sstevel@tonic-gate (void) crsetugid(lcr, svp->sv_secdata->uid, 32977c478bd9Sstevel@tonic-gate crgetgid(cr)); 32987c478bd9Sstevel@tonic-gate } 32997c478bd9Sstevel@tonic-gate nfs_rw_exit(&svp->sv_lock); 33007c478bd9Sstevel@tonic-gate 3301f86c6ccaSdm if (lcr != NULL) { 3302f86c6ccaSdm mutex_enter(&np->s_lock); 3303f86c6ccaSdm crfree(np->s_cred); 3304f86c6ccaSdm np->s_cred = lcr; 3305f86c6ccaSdm mutex_exit(&np->s_lock); 33067c478bd9Sstevel@tonic-gate nfs4setclientid_otw(mi, svp, lcr, np, n4ep, 3307b9238976Sth &retry_inuse); 3308f86c6ccaSdm } 33097c478bd9Sstevel@tonic-gate } 3310f86c6ccaSdm mutex_enter(&np->s_lock); 33117c478bd9Sstevel@tonic-gate lease_time = np->s_lease_time; 3312f86c6ccaSdm np->s_flags &= ~N4S_CLIENTID_PEND; 33137c478bd9Sstevel@tonic-gate mutex_exit(&np->s_lock); 33147c478bd9Sstevel@tonic-gate 33157c478bd9Sstevel@tonic-gate if (n4ep->error != 0 || n4ep->stat != NFS4_OK) { 33167c478bd9Sstevel@tonic-gate /* 33177c478bd9Sstevel@tonic-gate * Start recovery if failover is a possibility. If 33187c478bd9Sstevel@tonic-gate * invoked by the recovery thread itself, then just 33197c478bd9Sstevel@tonic-gate * return and let it handle the failover first. NB: 33207c478bd9Sstevel@tonic-gate * recovery is not allowed if the mount is in progress 33217c478bd9Sstevel@tonic-gate * since the infrastructure is not sufficiently setup 33227c478bd9Sstevel@tonic-gate * to allow it. Just return the error (after suitable 33237c478bd9Sstevel@tonic-gate * retries). 33247c478bd9Sstevel@tonic-gate */ 33257c478bd9Sstevel@tonic-gate if (FAILOVER_MOUNT4(mi) && nfs4_try_failover(n4ep)) { 33267c478bd9Sstevel@tonic-gate (void) nfs4_start_recovery(n4ep, mi, NULL, 33272f172c55SRobert Thurlow NULL, NULL, NULL, OP_SETCLIENTID, NULL, NULL, NULL); 33287c478bd9Sstevel@tonic-gate /* 33297c478bd9Sstevel@tonic-gate * Don't retry here, just return and let 33307c478bd9Sstevel@tonic-gate * recovery take over. 33317c478bd9Sstevel@tonic-gate */ 33327c478bd9Sstevel@tonic-gate if (recovery) 33337c478bd9Sstevel@tonic-gate retry = FALSE; 33347c478bd9Sstevel@tonic-gate } else if (nfs4_rpc_retry_error(n4ep->error) || 3335b9238976Sth n4ep->stat == NFS4ERR_RESOURCE || 3336b9238976Sth n4ep->stat == NFS4ERR_STALE_CLIENTID) { 33377c478bd9Sstevel@tonic-gate 3338b9238976Sth retry = TRUE; 3339b9238976Sth /* 3340b9238976Sth * Always retry if in recovery or once had 3341b9238976Sth * contact with the server (but now it's 3342b9238976Sth * overloaded). 3343b9238976Sth */ 3344b9238976Sth if (recovery == TRUE || 3345b9238976Sth n4ep->error == ETIMEDOUT || 3346b9238976Sth n4ep->error == ECONNRESET) 33477c478bd9Sstevel@tonic-gate num_retries = 0; 3348b9238976Sth } else if (retry_inuse && n4ep->error == 0 && 3349b9238976Sth n4ep->stat == NFS4ERR_CLID_INUSE) { 3350b9238976Sth retry = TRUE; 3351b9238976Sth num_retries = 0; 33527c478bd9Sstevel@tonic-gate } 3353f86c6ccaSdm } else { 3354f64c4ae1Sdm /* 3355f64c4ae1Sdm * Since everything succeeded give the list a reference count if 3356f64c4ae1Sdm * it hasn't been given one by add_new_nfs4_server() or if this 3357f64c4ae1Sdm * is not a recovery situation in which case it is already on 3358f64c4ae1Sdm * the list. 3359f64c4ae1Sdm */ 3360f86c6ccaSdm mutex_enter(&np->s_lock); 3361f64c4ae1Sdm if ((np->s_flags & N4S_INSERTED) == 0) { 3362f64c4ae1Sdm np->s_refcnt++; 3363f64c4ae1Sdm np->s_flags |= N4S_INSERTED; 3364f64c4ae1Sdm } 3365f86c6ccaSdm mutex_exit(&np->s_lock); 33667c478bd9Sstevel@tonic-gate } 33677c478bd9Sstevel@tonic-gate 33687c478bd9Sstevel@tonic-gate if (!recovery) 33697c478bd9Sstevel@tonic-gate nfs4_end_op(mi, NULL, NULL, &recov_state, recovery); 3370f86c6ccaSdm 33717c478bd9Sstevel@tonic-gate 33727c478bd9Sstevel@tonic-gate if (retry && num_retries++ < nfs4_num_sclid_retries) { 33737c478bd9Sstevel@tonic-gate if (retry_inuse) { 33747c478bd9Sstevel@tonic-gate delay(SEC_TO_TICK(lease_time + nfs4_retry_sclid_delay)); 33757c478bd9Sstevel@tonic-gate retry_inuse = 0; 33767c478bd9Sstevel@tonic-gate } else 33777c478bd9Sstevel@tonic-gate delay(SEC_TO_TICK(nfs4_retry_sclid_delay)); 3378f86c6ccaSdm 3379f86c6ccaSdm nfs4_server_rele(np); 33807c478bd9Sstevel@tonic-gate goto recov_retry; 33817c478bd9Sstevel@tonic-gate } 33827c478bd9Sstevel@tonic-gate 3383f86c6ccaSdm 33847c478bd9Sstevel@tonic-gate if (n4ep->error == 0) 33857c478bd9Sstevel@tonic-gate n4ep->error = geterrno4(n4ep->stat); 3386f86c6ccaSdm 3387f86c6ccaSdm /* broadcast before release in case no other threads are waiting */ 3388f86c6ccaSdm cv_broadcast(&np->s_clientid_pend); 3389f86c6ccaSdm nfs4_server_rele(np); 33907c478bd9Sstevel@tonic-gate } 33917c478bd9Sstevel@tonic-gate 33927c478bd9Sstevel@tonic-gate int nfs4setclientid_otw_debug = 0; 33937c478bd9Sstevel@tonic-gate 33947c478bd9Sstevel@tonic-gate /* 33957c478bd9Sstevel@tonic-gate * This function handles the recovery of STALE_CLIENTID for SETCLIENTID_CONFRIM, 33967c478bd9Sstevel@tonic-gate * but nothing else; the calling function must be designed to handle those 33977c478bd9Sstevel@tonic-gate * other errors. 33987c478bd9Sstevel@tonic-gate */ 33997c478bd9Sstevel@tonic-gate static void 34007c478bd9Sstevel@tonic-gate nfs4setclientid_otw(mntinfo4_t *mi, struct servinfo4 *svp, cred_t *cr, 3401b9238976Sth struct nfs4_server *np, nfs4_error_t *ep, int *retry_inusep) 34027c478bd9Sstevel@tonic-gate { 34037c478bd9Sstevel@tonic-gate COMPOUND4args_clnt args; 34047c478bd9Sstevel@tonic-gate COMPOUND4res_clnt res; 34057c478bd9Sstevel@tonic-gate nfs_argop4 argop[3]; 34067c478bd9Sstevel@tonic-gate SETCLIENTID4args *s_args; 34077c478bd9Sstevel@tonic-gate SETCLIENTID4resok *s_resok; 34087c478bd9Sstevel@tonic-gate int doqueue = 1; 34097c478bd9Sstevel@tonic-gate nfs4_ga_res_t *garp = NULL; 34107c478bd9Sstevel@tonic-gate timespec_t prop_time, after_time; 34117c478bd9Sstevel@tonic-gate verifier4 verf; 34127c478bd9Sstevel@tonic-gate clientid4 tmp_clientid; 34137c478bd9Sstevel@tonic-gate 3414f86c6ccaSdm ASSERT(!MUTEX_HELD(&np->s_lock)); 34157c478bd9Sstevel@tonic-gate 34167c478bd9Sstevel@tonic-gate args.ctag = TAG_SETCLIENTID; 34177c478bd9Sstevel@tonic-gate 34187c478bd9Sstevel@tonic-gate args.array = argop; 34197c478bd9Sstevel@tonic-gate args.array_len = 3; 34207c478bd9Sstevel@tonic-gate 34217c478bd9Sstevel@tonic-gate /* PUTROOTFH */ 34227c478bd9Sstevel@tonic-gate argop[0].argop = OP_PUTROOTFH; 34237c478bd9Sstevel@tonic-gate 34247c478bd9Sstevel@tonic-gate /* GETATTR */ 34257c478bd9Sstevel@tonic-gate argop[1].argop = OP_GETATTR; 34267c478bd9Sstevel@tonic-gate argop[1].nfs_argop4_u.opgetattr.attr_request = FATTR4_LEASE_TIME_MASK; 34277c478bd9Sstevel@tonic-gate argop[1].nfs_argop4_u.opgetattr.mi = mi; 34287c478bd9Sstevel@tonic-gate 34297c478bd9Sstevel@tonic-gate /* SETCLIENTID */ 34307c478bd9Sstevel@tonic-gate argop[2].argop = OP_SETCLIENTID; 34317c478bd9Sstevel@tonic-gate 34327c478bd9Sstevel@tonic-gate s_args = &argop[2].nfs_argop4_u.opsetclientid; 34337c478bd9Sstevel@tonic-gate 3434f86c6ccaSdm mutex_enter(&np->s_lock); 3435f86c6ccaSdm 34367c478bd9Sstevel@tonic-gate s_args->client.verifier = np->clidtosend.verifier; 34377c478bd9Sstevel@tonic-gate s_args->client.id_len = np->clidtosend.id_len; 34387c478bd9Sstevel@tonic-gate ASSERT(s_args->client.id_len <= NFS4_OPAQUE_LIMIT); 34397c478bd9Sstevel@tonic-gate s_args->client.id_val = np->clidtosend.id_val; 34407c478bd9Sstevel@tonic-gate 34417c478bd9Sstevel@tonic-gate /* 34427c478bd9Sstevel@tonic-gate * Callback needs to happen on non-RDMA transport 34437c478bd9Sstevel@tonic-gate * Check if we have saved the original knetconfig 34447c478bd9Sstevel@tonic-gate * if so, use that instead. 34457c478bd9Sstevel@tonic-gate */ 34467c478bd9Sstevel@tonic-gate if (svp->sv_origknconf != NULL) 34477c478bd9Sstevel@tonic-gate nfs4_cb_args(np, svp->sv_origknconf, s_args); 34487c478bd9Sstevel@tonic-gate else 34497c478bd9Sstevel@tonic-gate nfs4_cb_args(np, svp->sv_knconf, s_args); 34507c478bd9Sstevel@tonic-gate 3451f86c6ccaSdm mutex_exit(&np->s_lock); 3452f86c6ccaSdm 3453f86c6ccaSdm rfs4call(mi, &args, &res, cr, &doqueue, 0, ep); 34547c478bd9Sstevel@tonic-gate 34557c478bd9Sstevel@tonic-gate if (ep->error) 34567c478bd9Sstevel@tonic-gate return; 34577c478bd9Sstevel@tonic-gate 34587c478bd9Sstevel@tonic-gate /* getattr lease_time res */ 3459e557fb2cSDai Ngo if ((res.array_len >= 2) && 3460e557fb2cSDai Ngo (res.array[1].nfs_resop4_u.opgetattr.status == NFS4_OK)) { 34617c478bd9Sstevel@tonic-gate garp = &res.array[1].nfs_resop4_u.opgetattr.ga_res; 34627c478bd9Sstevel@tonic-gate 34637c478bd9Sstevel@tonic-gate #ifndef _LP64 34647c478bd9Sstevel@tonic-gate /* 34657c478bd9Sstevel@tonic-gate * The 32 bit client cannot handle a lease time greater than 34667c478bd9Sstevel@tonic-gate * (INT32_MAX/1000000). This is due to the use of the 34677c478bd9Sstevel@tonic-gate * lease_time in calls to drv_usectohz() in 34687c478bd9Sstevel@tonic-gate * nfs4_renew_lease_thread(). The problem is that 34697c478bd9Sstevel@tonic-gate * drv_usectohz() takes a time_t (which is just a long = 4 34707c478bd9Sstevel@tonic-gate * bytes) as its parameter. The lease_time is multiplied by 34717c478bd9Sstevel@tonic-gate * 1000000 to convert seconds to usecs for the parameter. If 34727c478bd9Sstevel@tonic-gate * a number bigger than (INT32_MAX/1000000) is used then we 34737c478bd9Sstevel@tonic-gate * overflow on the 32bit client. 34747c478bd9Sstevel@tonic-gate */ 34757c478bd9Sstevel@tonic-gate if (garp->n4g_ext_res->n4g_leasetime > (INT32_MAX/1000000)) { 34767c478bd9Sstevel@tonic-gate garp->n4g_ext_res->n4g_leasetime = INT32_MAX/1000000; 34777c478bd9Sstevel@tonic-gate } 34787c478bd9Sstevel@tonic-gate #endif 34797c478bd9Sstevel@tonic-gate 3480f86c6ccaSdm mutex_enter(&np->s_lock); 34817c478bd9Sstevel@tonic-gate np->s_lease_time = garp->n4g_ext_res->n4g_leasetime; 34827c478bd9Sstevel@tonic-gate 34837c478bd9Sstevel@tonic-gate /* 34847c478bd9Sstevel@tonic-gate * Keep track of the lease period for the mi's 34857c478bd9Sstevel@tonic-gate * mi_msg_list. We need an appropiate time 34867c478bd9Sstevel@tonic-gate * bound to associate past facts with a current 34877c478bd9Sstevel@tonic-gate * event. The lease period is perfect for this. 34887c478bd9Sstevel@tonic-gate */ 34897c478bd9Sstevel@tonic-gate mutex_enter(&mi->mi_msg_list_lock); 34907c478bd9Sstevel@tonic-gate mi->mi_lease_period = np->s_lease_time; 34917c478bd9Sstevel@tonic-gate mutex_exit(&mi->mi_msg_list_lock); 3492f86c6ccaSdm mutex_exit(&np->s_lock); 34937c478bd9Sstevel@tonic-gate } 34947c478bd9Sstevel@tonic-gate 34957c478bd9Sstevel@tonic-gate 34967c478bd9Sstevel@tonic-gate if (res.status == NFS4ERR_CLID_INUSE) { 34977c478bd9Sstevel@tonic-gate clientaddr4 *clid_inuse; 34987c478bd9Sstevel@tonic-gate 34997c478bd9Sstevel@tonic-gate if (!(*retry_inusep)) { 35007c478bd9Sstevel@tonic-gate clid_inuse = &res.array->nfs_resop4_u. 3501b9238976Sth opsetclientid.SETCLIENTID4res_u.client_using; 35027c478bd9Sstevel@tonic-gate 35037c478bd9Sstevel@tonic-gate zcmn_err(mi->mi_zone->zone_id, CE_NOTE, 35047c478bd9Sstevel@tonic-gate "NFS4 mount (SETCLIENTID failed)." 35057c478bd9Sstevel@tonic-gate " nfs4_client_id.id is in" 35067c478bd9Sstevel@tonic-gate "use already by: r_netid<%s> r_addr<%s>", 35077c478bd9Sstevel@tonic-gate clid_inuse->r_netid, clid_inuse->r_addr); 35087c478bd9Sstevel@tonic-gate } 35097c478bd9Sstevel@tonic-gate 35107c478bd9Sstevel@tonic-gate /* 35117c478bd9Sstevel@tonic-gate * XXX - The client should be more robust in its 35127c478bd9Sstevel@tonic-gate * handling of clientid in use errors (regen another 35137c478bd9Sstevel@tonic-gate * clientid and try again?) 35147c478bd9Sstevel@tonic-gate */ 35157c478bd9Sstevel@tonic-gate (void) xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&res); 35167c478bd9Sstevel@tonic-gate return; 35177c478bd9Sstevel@tonic-gate } 35187c478bd9Sstevel@tonic-gate 35197c478bd9Sstevel@tonic-gate if (res.status) { 35207c478bd9Sstevel@tonic-gate (void) xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&res); 35217c478bd9Sstevel@tonic-gate return; 35227c478bd9Sstevel@tonic-gate } 35237c478bd9Sstevel@tonic-gate 35247c478bd9Sstevel@tonic-gate s_resok = &res.array[2].nfs_resop4_u. 3525b9238976Sth opsetclientid.SETCLIENTID4res_u.resok4; 35267c478bd9Sstevel@tonic-gate 35277c478bd9Sstevel@tonic-gate tmp_clientid = s_resok->clientid; 35287c478bd9Sstevel@tonic-gate 35297c478bd9Sstevel@tonic-gate verf = s_resok->setclientid_confirm; 35307c478bd9Sstevel@tonic-gate 35317c478bd9Sstevel@tonic-gate #ifdef DEBUG 35327c478bd9Sstevel@tonic-gate if (nfs4setclientid_otw_debug) { 35337c478bd9Sstevel@tonic-gate union { 35347c478bd9Sstevel@tonic-gate clientid4 clientid; 35357c478bd9Sstevel@tonic-gate int foo[2]; 35367c478bd9Sstevel@tonic-gate } cid; 35377c478bd9Sstevel@tonic-gate 35387c478bd9Sstevel@tonic-gate cid.clientid = s_resok->clientid; 35397c478bd9Sstevel@tonic-gate 35407c478bd9Sstevel@tonic-gate zcmn_err(mi->mi_zone->zone_id, CE_NOTE, 35417c478bd9Sstevel@tonic-gate "nfs4setclientid_otw: OK, clientid = %x,%x, " 35427c478bd9Sstevel@tonic-gate "verifier = %" PRIx64 "\n", cid.foo[0], cid.foo[1], verf); 35437c478bd9Sstevel@tonic-gate } 35447c478bd9Sstevel@tonic-gate #endif 35457c478bd9Sstevel@tonic-gate 35467c478bd9Sstevel@tonic-gate (void) xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&res); 35477c478bd9Sstevel@tonic-gate 35487c478bd9Sstevel@tonic-gate /* Confirm the client id and get the lease_time attribute */ 35497c478bd9Sstevel@tonic-gate 35507c478bd9Sstevel@tonic-gate args.ctag = TAG_SETCLIENTID_CF; 35517c478bd9Sstevel@tonic-gate 35527c478bd9Sstevel@tonic-gate args.array = argop; 35537c478bd9Sstevel@tonic-gate args.array_len = 1; 35547c478bd9Sstevel@tonic-gate 35557c478bd9Sstevel@tonic-gate argop[0].argop = OP_SETCLIENTID_CONFIRM; 35567c478bd9Sstevel@tonic-gate 35577c478bd9Sstevel@tonic-gate argop[0].nfs_argop4_u.opsetclientid_confirm.clientid = tmp_clientid; 35587c478bd9Sstevel@tonic-gate argop[0].nfs_argop4_u.opsetclientid_confirm.setclientid_confirm = verf; 35597c478bd9Sstevel@tonic-gate 35607c478bd9Sstevel@tonic-gate /* used to figure out RTT for np */ 35617c478bd9Sstevel@tonic-gate gethrestime(&prop_time); 35627c478bd9Sstevel@tonic-gate 35637c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_lease_debug, (CE_NOTE, "nfs4setlientid_otw: " 3564b9238976Sth "start time: %ld sec %ld nsec", prop_time.tv_sec, 3565b9238976Sth prop_time.tv_nsec)); 35667c478bd9Sstevel@tonic-gate 35677c478bd9Sstevel@tonic-gate rfs4call(mi, &args, &res, cr, &doqueue, 0, ep); 35687c478bd9Sstevel@tonic-gate 35697c478bd9Sstevel@tonic-gate gethrestime(&after_time); 3570f86c6ccaSdm mutex_enter(&np->s_lock); 35717c478bd9Sstevel@tonic-gate np->propagation_delay.tv_sec = 3572b9238976Sth MAX(1, after_time.tv_sec - prop_time.tv_sec); 3573f86c6ccaSdm mutex_exit(&np->s_lock); 35747c478bd9Sstevel@tonic-gate 35757c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_lease_debug, (CE_NOTE, "nfs4setlcientid_otw: " 3576b9238976Sth "finish time: %ld sec ", after_time.tv_sec)); 35777c478bd9Sstevel@tonic-gate 35787c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_lease_debug, (CE_NOTE, "nfs4setclientid_otw: " 3579b9238976Sth "propagation delay set to %ld sec", 3580b9238976Sth np->propagation_delay.tv_sec)); 35817c478bd9Sstevel@tonic-gate 35827c478bd9Sstevel@tonic-gate if (ep->error) 35837c478bd9Sstevel@tonic-gate return; 35847c478bd9Sstevel@tonic-gate 35857c478bd9Sstevel@tonic-gate if (res.status == NFS4ERR_CLID_INUSE) { 35867c478bd9Sstevel@tonic-gate clientaddr4 *clid_inuse; 35877c478bd9Sstevel@tonic-gate 35887c478bd9Sstevel@tonic-gate if (!(*retry_inusep)) { 35897c478bd9Sstevel@tonic-gate clid_inuse = &res.array->nfs_resop4_u. 3590b9238976Sth opsetclientid.SETCLIENTID4res_u.client_using; 35917c478bd9Sstevel@tonic-gate 35927c478bd9Sstevel@tonic-gate zcmn_err(mi->mi_zone->zone_id, CE_NOTE, 35937c478bd9Sstevel@tonic-gate "SETCLIENTID_CONFIRM failed. " 35947c478bd9Sstevel@tonic-gate "nfs4_client_id.id is in use already by: " 35957c478bd9Sstevel@tonic-gate "r_netid<%s> r_addr<%s>", 35967c478bd9Sstevel@tonic-gate clid_inuse->r_netid, clid_inuse->r_addr); 35977c478bd9Sstevel@tonic-gate } 35987c478bd9Sstevel@tonic-gate 35997c478bd9Sstevel@tonic-gate (void) xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&res); 36007c478bd9Sstevel@tonic-gate return; 36017c478bd9Sstevel@tonic-gate } 36027c478bd9Sstevel@tonic-gate 36037c478bd9Sstevel@tonic-gate if (res.status) { 36047c478bd9Sstevel@tonic-gate (void) xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&res); 36057c478bd9Sstevel@tonic-gate return; 36067c478bd9Sstevel@tonic-gate } 36077c478bd9Sstevel@tonic-gate 3608f86c6ccaSdm mutex_enter(&np->s_lock); 36097c478bd9Sstevel@tonic-gate np->clientid = tmp_clientid; 36107c478bd9Sstevel@tonic-gate np->s_flags |= N4S_CLIENTID_SET; 36117c478bd9Sstevel@tonic-gate 36127c478bd9Sstevel@tonic-gate /* Add mi to np's mntinfo4 list */ 36137c478bd9Sstevel@tonic-gate nfs4_add_mi_to_server(np, mi); 36147c478bd9Sstevel@tonic-gate 36157c478bd9Sstevel@tonic-gate if (np->lease_valid == NFS4_LEASE_NOT_STARTED) { 36167c478bd9Sstevel@tonic-gate /* 36177c478bd9Sstevel@tonic-gate * Start lease management thread. 36187c478bd9Sstevel@tonic-gate * Keep trying until we succeed. 36197c478bd9Sstevel@tonic-gate */ 36207c478bd9Sstevel@tonic-gate 36217c478bd9Sstevel@tonic-gate np->s_refcnt++; /* pass reference to thread */ 36227c478bd9Sstevel@tonic-gate (void) zthread_create(NULL, 0, nfs4_renew_lease_thread, np, 0, 3623b9238976Sth minclsyspri); 36247c478bd9Sstevel@tonic-gate } 3625f86c6ccaSdm mutex_exit(&np->s_lock); 36267c478bd9Sstevel@tonic-gate 36277c478bd9Sstevel@tonic-gate (void) xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&res); 36287c478bd9Sstevel@tonic-gate } 36297c478bd9Sstevel@tonic-gate 36307c478bd9Sstevel@tonic-gate /* 36317c478bd9Sstevel@tonic-gate * Add mi to sp's mntinfo4_list if it isn't already in the list. Makes 36327c478bd9Sstevel@tonic-gate * mi's clientid the same as sp's. 36337c478bd9Sstevel@tonic-gate * Assumes sp is locked down. 36347c478bd9Sstevel@tonic-gate */ 36357c478bd9Sstevel@tonic-gate void 36367c478bd9Sstevel@tonic-gate nfs4_add_mi_to_server(nfs4_server_t *sp, mntinfo4_t *mi) 36377c478bd9Sstevel@tonic-gate { 36387c478bd9Sstevel@tonic-gate mntinfo4_t *tmi; 36397c478bd9Sstevel@tonic-gate int in_list = 0; 36407c478bd9Sstevel@tonic-gate 3641a092743bSek ASSERT(nfs_rw_lock_held(&mi->mi_recovlock, RW_READER) || 3642a092743bSek nfs_rw_lock_held(&mi->mi_recovlock, RW_WRITER)); 36437c478bd9Sstevel@tonic-gate ASSERT(sp != &nfs4_server_lst); 36447c478bd9Sstevel@tonic-gate ASSERT(MUTEX_HELD(&sp->s_lock)); 36457c478bd9Sstevel@tonic-gate 36467c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_lease_debug, (CE_NOTE, 3647b9238976Sth "nfs4_add_mi_to_server: add mi %p to sp %p", 3648b9238976Sth (void*)mi, (void*)sp)); 36497c478bd9Sstevel@tonic-gate 36507c478bd9Sstevel@tonic-gate for (tmi = sp->mntinfo4_list; 36517c478bd9Sstevel@tonic-gate tmi != NULL; 36527c478bd9Sstevel@tonic-gate tmi = tmi->mi_clientid_next) { 36537c478bd9Sstevel@tonic-gate if (tmi == mi) { 36547c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_lease_debug, 3655b9238976Sth (CE_NOTE, 3656b9238976Sth "nfs4_add_mi_to_server: mi in list")); 36577c478bd9Sstevel@tonic-gate in_list = 1; 36587c478bd9Sstevel@tonic-gate } 36597c478bd9Sstevel@tonic-gate } 36607c478bd9Sstevel@tonic-gate 36617c478bd9Sstevel@tonic-gate /* 36627c478bd9Sstevel@tonic-gate * First put a hold on the mntinfo4's vfsp so that references via 36637c478bd9Sstevel@tonic-gate * mntinfo4_list will be valid. 36647c478bd9Sstevel@tonic-gate */ 36657c478bd9Sstevel@tonic-gate if (!in_list) 36667c478bd9Sstevel@tonic-gate VFS_HOLD(mi->mi_vfsp); 36677c478bd9Sstevel@tonic-gate 36687c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_lease_debug, (CE_NOTE, "nfs4_add_mi_to_server: " 3669b9238976Sth "hold vfs %p for mi: %p", (void*)mi->mi_vfsp, (void*)mi)); 36707c478bd9Sstevel@tonic-gate 36717c478bd9Sstevel@tonic-gate if (!in_list) { 36727c478bd9Sstevel@tonic-gate if (sp->mntinfo4_list) 36737c478bd9Sstevel@tonic-gate sp->mntinfo4_list->mi_clientid_prev = mi; 36747c478bd9Sstevel@tonic-gate mi->mi_clientid_next = sp->mntinfo4_list; 36753b895386SPavel Filipensky mi->mi_srv = sp; 36767c478bd9Sstevel@tonic-gate sp->mntinfo4_list = mi; 36777c478bd9Sstevel@tonic-gate mi->mi_srvsettime = gethrestime_sec(); 36783b895386SPavel Filipensky mi->mi_srvset_cnt++; 36797c478bd9Sstevel@tonic-gate } 36807c478bd9Sstevel@tonic-gate 36817c478bd9Sstevel@tonic-gate /* set mi's clientid to that of sp's for later matching */ 36827c478bd9Sstevel@tonic-gate mi->mi_clientid = sp->clientid; 36837c478bd9Sstevel@tonic-gate 36847c478bd9Sstevel@tonic-gate /* 36857c478bd9Sstevel@tonic-gate * Update the clientid for any other mi's belonging to sp. This 36867c478bd9Sstevel@tonic-gate * must be done here while we hold sp->s_lock, so that 36877c478bd9Sstevel@tonic-gate * find_nfs4_server() continues to work. 36887c478bd9Sstevel@tonic-gate */ 36897c478bd9Sstevel@tonic-gate 36907c478bd9Sstevel@tonic-gate for (tmi = sp->mntinfo4_list; 36917c478bd9Sstevel@tonic-gate tmi != NULL; 36927c478bd9Sstevel@tonic-gate tmi = tmi->mi_clientid_next) { 36937c478bd9Sstevel@tonic-gate if (tmi != mi) { 36947c478bd9Sstevel@tonic-gate tmi->mi_clientid = sp->clientid; 36957c478bd9Sstevel@tonic-gate } 36967c478bd9Sstevel@tonic-gate } 36977c478bd9Sstevel@tonic-gate } 36987c478bd9Sstevel@tonic-gate 36997c478bd9Sstevel@tonic-gate /* 37007c478bd9Sstevel@tonic-gate * Remove the mi from sp's mntinfo4_list and release its reference. 37017c478bd9Sstevel@tonic-gate * Exception: if mi still has open files, flag it for later removal (when 37027c478bd9Sstevel@tonic-gate * all the files are closed). 37037c478bd9Sstevel@tonic-gate * 37047c478bd9Sstevel@tonic-gate * If this is the last mntinfo4 in sp's list then tell the lease renewal 37057c478bd9Sstevel@tonic-gate * thread to exit. 37067c478bd9Sstevel@tonic-gate */ 37077c478bd9Sstevel@tonic-gate static void 37087c478bd9Sstevel@tonic-gate nfs4_remove_mi_from_server_nolock(mntinfo4_t *mi, nfs4_server_t *sp) 37097c478bd9Sstevel@tonic-gate { 37107c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_lease_debug, (CE_NOTE, 3711b9238976Sth "nfs4_remove_mi_from_server_nolock: remove mi %p from sp %p", 3712b9238976Sth (void*)mi, (void*)sp)); 37137c478bd9Sstevel@tonic-gate 37147c478bd9Sstevel@tonic-gate ASSERT(sp != NULL); 37157c478bd9Sstevel@tonic-gate ASSERT(MUTEX_HELD(&sp->s_lock)); 37167c478bd9Sstevel@tonic-gate ASSERT(mi->mi_open_files >= 0); 37177c478bd9Sstevel@tonic-gate 37187c478bd9Sstevel@tonic-gate /* 37197c478bd9Sstevel@tonic-gate * First make sure this mntinfo4 can be taken off of the list, 37207c478bd9Sstevel@tonic-gate * ie: it doesn't have any open files remaining. 37217c478bd9Sstevel@tonic-gate */ 37227c478bd9Sstevel@tonic-gate if (mi->mi_open_files > 0) { 37237c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_lease_debug, (CE_NOTE, 3724b9238976Sth "nfs4_remove_mi_from_server_nolock: don't " 3725b9238976Sth "remove mi since it still has files open")); 37267c478bd9Sstevel@tonic-gate 37277c478bd9Sstevel@tonic-gate mutex_enter(&mi->mi_lock); 37287c478bd9Sstevel@tonic-gate mi->mi_flags |= MI4_REMOVE_ON_LAST_CLOSE; 37297c478bd9Sstevel@tonic-gate mutex_exit(&mi->mi_lock); 37307c478bd9Sstevel@tonic-gate return; 37317c478bd9Sstevel@tonic-gate } 37327c478bd9Sstevel@tonic-gate 373350a83466Sjwahlig VFS_HOLD(mi->mi_vfsp); 37347c478bd9Sstevel@tonic-gate remove_mi(sp, mi); 373550a83466Sjwahlig VFS_RELE(mi->mi_vfsp); 37367c478bd9Sstevel@tonic-gate 37377c478bd9Sstevel@tonic-gate if (sp->mntinfo4_list == NULL) { 37387c478bd9Sstevel@tonic-gate /* last fs unmounted, kill the thread */ 37397c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_lease_debug, (CE_NOTE, 3740b9238976Sth "remove_mi_from_nfs4_server_nolock: kill the thread")); 37417c478bd9Sstevel@tonic-gate nfs4_mark_srv_dead(sp); 37427c478bd9Sstevel@tonic-gate } 37437c478bd9Sstevel@tonic-gate } 37447c478bd9Sstevel@tonic-gate 37457c478bd9Sstevel@tonic-gate /* 37467c478bd9Sstevel@tonic-gate * Remove mi from sp's mntinfo4_list and release the vfs reference. 37477c478bd9Sstevel@tonic-gate */ 37487c478bd9Sstevel@tonic-gate static void 37497c478bd9Sstevel@tonic-gate remove_mi(nfs4_server_t *sp, mntinfo4_t *mi) 37507c478bd9Sstevel@tonic-gate { 37517c478bd9Sstevel@tonic-gate ASSERT(MUTEX_HELD(&sp->s_lock)); 37527c478bd9Sstevel@tonic-gate 37537c478bd9Sstevel@tonic-gate /* 37547c478bd9Sstevel@tonic-gate * We release a reference, and the caller must still have a 37557c478bd9Sstevel@tonic-gate * reference. 37567c478bd9Sstevel@tonic-gate */ 37577c478bd9Sstevel@tonic-gate ASSERT(mi->mi_vfsp->vfs_count >= 2); 37587c478bd9Sstevel@tonic-gate 37597c478bd9Sstevel@tonic-gate if (mi->mi_clientid_prev) { 37607c478bd9Sstevel@tonic-gate mi->mi_clientid_prev->mi_clientid_next = mi->mi_clientid_next; 37617c478bd9Sstevel@tonic-gate } else { 37627c478bd9Sstevel@tonic-gate /* This is the first mi in sp's mntinfo4_list */ 37637c478bd9Sstevel@tonic-gate /* 37647c478bd9Sstevel@tonic-gate * Make sure the first mntinfo4 in the list is the actual 37657c478bd9Sstevel@tonic-gate * mntinfo4 passed in. 37667c478bd9Sstevel@tonic-gate */ 37677c478bd9Sstevel@tonic-gate ASSERT(sp->mntinfo4_list == mi); 37687c478bd9Sstevel@tonic-gate 37697c478bd9Sstevel@tonic-gate sp->mntinfo4_list = mi->mi_clientid_next; 37707c478bd9Sstevel@tonic-gate } 37717c478bd9Sstevel@tonic-gate if (mi->mi_clientid_next) 37727c478bd9Sstevel@tonic-gate mi->mi_clientid_next->mi_clientid_prev = mi->mi_clientid_prev; 37737c478bd9Sstevel@tonic-gate 37747c478bd9Sstevel@tonic-gate /* Now mark the mntinfo4's links as being removed */ 37757c478bd9Sstevel@tonic-gate mi->mi_clientid_prev = mi->mi_clientid_next = NULL; 37763b895386SPavel Filipensky mi->mi_srv = NULL; 377722dc8f51SPavel Filipensky mi->mi_srvset_cnt++; 37787c478bd9Sstevel@tonic-gate 37797c478bd9Sstevel@tonic-gate VFS_RELE(mi->mi_vfsp); 37807c478bd9Sstevel@tonic-gate } 37817c478bd9Sstevel@tonic-gate 37827c478bd9Sstevel@tonic-gate /* 37837c478bd9Sstevel@tonic-gate * Free all the entries in sp's mntinfo4_list. 37847c478bd9Sstevel@tonic-gate */ 37857c478bd9Sstevel@tonic-gate static void 37867c478bd9Sstevel@tonic-gate remove_all_mi(nfs4_server_t *sp) 37877c478bd9Sstevel@tonic-gate { 37887c478bd9Sstevel@tonic-gate mntinfo4_t *mi; 37897c478bd9Sstevel@tonic-gate 37907c478bd9Sstevel@tonic-gate ASSERT(MUTEX_HELD(&sp->s_lock)); 37917c478bd9Sstevel@tonic-gate 37927c478bd9Sstevel@tonic-gate while (sp->mntinfo4_list != NULL) { 37937c478bd9Sstevel@tonic-gate mi = sp->mntinfo4_list; 37947c478bd9Sstevel@tonic-gate /* 37957c478bd9Sstevel@tonic-gate * Grab a reference in case there is only one left (which 37967c478bd9Sstevel@tonic-gate * remove_mi() frees). 37977c478bd9Sstevel@tonic-gate */ 37987c478bd9Sstevel@tonic-gate VFS_HOLD(mi->mi_vfsp); 37997c478bd9Sstevel@tonic-gate remove_mi(sp, mi); 38007c478bd9Sstevel@tonic-gate VFS_RELE(mi->mi_vfsp); 38017c478bd9Sstevel@tonic-gate } 38027c478bd9Sstevel@tonic-gate } 38037c478bd9Sstevel@tonic-gate 38047c478bd9Sstevel@tonic-gate /* 38057c478bd9Sstevel@tonic-gate * Remove the mi from sp's mntinfo4_list as above, and rele the vfs. 38067c478bd9Sstevel@tonic-gate * 38077c478bd9Sstevel@tonic-gate * This version can be called with a null nfs4_server_t arg, 38087c478bd9Sstevel@tonic-gate * and will either find the right one and handle locking, or 38097c478bd9Sstevel@tonic-gate * do nothing because the mi wasn't added to an sp's mntinfo4_list. 38107c478bd9Sstevel@tonic-gate */ 38117c478bd9Sstevel@tonic-gate void 38127c478bd9Sstevel@tonic-gate nfs4_remove_mi_from_server(mntinfo4_t *mi, nfs4_server_t *esp) 38137c478bd9Sstevel@tonic-gate { 38147c478bd9Sstevel@tonic-gate nfs4_server_t *sp; 38157c478bd9Sstevel@tonic-gate 38163b895386SPavel Filipensky if (esp) { 38173b895386SPavel Filipensky nfs4_remove_mi_from_server_nolock(mi, esp); 38183b895386SPavel Filipensky return; 38193b895386SPavel Filipensky } 38207c478bd9Sstevel@tonic-gate 38213b895386SPavel Filipensky (void) nfs_rw_enter_sig(&mi->mi_recovlock, RW_READER, 0); 382222dc8f51SPavel Filipensky if (sp = find_nfs4_server_all(mi, 1)) { 38237c478bd9Sstevel@tonic-gate nfs4_remove_mi_from_server_nolock(mi, sp); 38243b895386SPavel Filipensky mutex_exit(&sp->s_lock); 382522dc8f51SPavel Filipensky nfs4_server_rele(sp); 38267c478bd9Sstevel@tonic-gate } 38273b895386SPavel Filipensky nfs_rw_exit(&mi->mi_recovlock); 38287c478bd9Sstevel@tonic-gate } 38297c478bd9Sstevel@tonic-gate 38307c478bd9Sstevel@tonic-gate /* 38317c478bd9Sstevel@tonic-gate * Return TRUE if the given server has any non-unmounted filesystems. 38327c478bd9Sstevel@tonic-gate */ 38337c478bd9Sstevel@tonic-gate 38347c478bd9Sstevel@tonic-gate bool_t 38357c478bd9Sstevel@tonic-gate nfs4_fs_active(nfs4_server_t *sp) 38367c478bd9Sstevel@tonic-gate { 38377c478bd9Sstevel@tonic-gate mntinfo4_t *mi; 38387c478bd9Sstevel@tonic-gate 38397c478bd9Sstevel@tonic-gate ASSERT(MUTEX_HELD(&sp->s_lock)); 38407c478bd9Sstevel@tonic-gate 38417c478bd9Sstevel@tonic-gate for (mi = sp->mntinfo4_list; mi != NULL; mi = mi->mi_clientid_next) { 38427c478bd9Sstevel@tonic-gate if (!(mi->mi_vfsp->vfs_flag & VFS_UNMOUNTED)) 38437c478bd9Sstevel@tonic-gate return (TRUE); 38447c478bd9Sstevel@tonic-gate } 38457c478bd9Sstevel@tonic-gate 38467c478bd9Sstevel@tonic-gate return (FALSE); 38477c478bd9Sstevel@tonic-gate } 38487c478bd9Sstevel@tonic-gate 38497c478bd9Sstevel@tonic-gate /* 38507c478bd9Sstevel@tonic-gate * Mark sp as finished and notify any waiters. 38517c478bd9Sstevel@tonic-gate */ 38527c478bd9Sstevel@tonic-gate 38537c478bd9Sstevel@tonic-gate void 38547c478bd9Sstevel@tonic-gate nfs4_mark_srv_dead(nfs4_server_t *sp) 38557c478bd9Sstevel@tonic-gate { 38567c478bd9Sstevel@tonic-gate ASSERT(MUTEX_HELD(&sp->s_lock)); 38577c478bd9Sstevel@tonic-gate 38587c478bd9Sstevel@tonic-gate sp->s_thread_exit = NFS4_THREAD_EXIT; 38597c478bd9Sstevel@tonic-gate cv_broadcast(&sp->cv_thread_exit); 38607c478bd9Sstevel@tonic-gate } 38617c478bd9Sstevel@tonic-gate 38627c478bd9Sstevel@tonic-gate /* 38637c478bd9Sstevel@tonic-gate * Create a new nfs4_server_t structure. 38647c478bd9Sstevel@tonic-gate * Returns new node unlocked and not in list, but with a reference count of 38657c478bd9Sstevel@tonic-gate * 1. 38667c478bd9Sstevel@tonic-gate */ 38677c478bd9Sstevel@tonic-gate struct nfs4_server * 38687c478bd9Sstevel@tonic-gate new_nfs4_server(struct servinfo4 *svp, cred_t *cr) 38697c478bd9Sstevel@tonic-gate { 38707c478bd9Sstevel@tonic-gate struct nfs4_server *np; 38717c478bd9Sstevel@tonic-gate timespec_t tt; 38727c478bd9Sstevel@tonic-gate union { 38737c478bd9Sstevel@tonic-gate struct { 38747c478bd9Sstevel@tonic-gate uint32_t sec; 38757c478bd9Sstevel@tonic-gate uint32_t subsec; 38767c478bd9Sstevel@tonic-gate } un_curtime; 38777c478bd9Sstevel@tonic-gate verifier4 un_verifier; 38787c478bd9Sstevel@tonic-gate } nfs4clientid_verifier; 38792f172c55SRobert Thurlow /* 38802f172c55SRobert Thurlow * We change this ID string carefully and with the Solaris 38812f172c55SRobert Thurlow * NFS server behaviour in mind. "+referrals" indicates 38822f172c55SRobert Thurlow * a client that can handle an NFSv4 referral. 38832f172c55SRobert Thurlow */ 38842f172c55SRobert Thurlow char id_val[] = "Solaris: %s, NFSv4 kernel client +referrals"; 38857c478bd9Sstevel@tonic-gate int len; 38867c478bd9Sstevel@tonic-gate 38877c478bd9Sstevel@tonic-gate np = kmem_zalloc(sizeof (struct nfs4_server), KM_SLEEP); 38887c478bd9Sstevel@tonic-gate np->saddr.len = svp->sv_addr.len; 38897c478bd9Sstevel@tonic-gate np->saddr.maxlen = svp->sv_addr.maxlen; 38907c478bd9Sstevel@tonic-gate np->saddr.buf = kmem_alloc(svp->sv_addr.maxlen, KM_SLEEP); 38917c478bd9Sstevel@tonic-gate bcopy(svp->sv_addr.buf, np->saddr.buf, svp->sv_addr.len); 38927c478bd9Sstevel@tonic-gate np->s_refcnt = 1; 38937c478bd9Sstevel@tonic-gate 38947c478bd9Sstevel@tonic-gate /* 38957c478bd9Sstevel@tonic-gate * Build the nfs_client_id4 for this server mount. Ensure 38967c478bd9Sstevel@tonic-gate * the verifier is useful and that the identification is 38977c478bd9Sstevel@tonic-gate * somehow based on the server's address for the case of 38987c478bd9Sstevel@tonic-gate * multi-homed servers. 38997c478bd9Sstevel@tonic-gate */ 39007c478bd9Sstevel@tonic-gate nfs4clientid_verifier.un_verifier = 0; 39017c478bd9Sstevel@tonic-gate gethrestime(&tt); 39027c478bd9Sstevel@tonic-gate nfs4clientid_verifier.un_curtime.sec = (uint32_t)tt.tv_sec; 39037c478bd9Sstevel@tonic-gate nfs4clientid_verifier.un_curtime.subsec = (uint32_t)tt.tv_nsec; 39047c478bd9Sstevel@tonic-gate np->clidtosend.verifier = nfs4clientid_verifier.un_verifier; 39057c478bd9Sstevel@tonic-gate 39067c478bd9Sstevel@tonic-gate /* 39077c478bd9Sstevel@tonic-gate * calculate the length of the opaque identifier. Subtract 2 39087c478bd9Sstevel@tonic-gate * for the "%s" and add the traditional +1 for null 39097c478bd9Sstevel@tonic-gate * termination. 39107c478bd9Sstevel@tonic-gate */ 39117c478bd9Sstevel@tonic-gate len = strlen(id_val) - 2 + strlen(uts_nodename()) + 1; 39127c478bd9Sstevel@tonic-gate np->clidtosend.id_len = len + np->saddr.maxlen; 39137c478bd9Sstevel@tonic-gate 39147c478bd9Sstevel@tonic-gate np->clidtosend.id_val = kmem_alloc(np->clidtosend.id_len, KM_SLEEP); 39157c478bd9Sstevel@tonic-gate (void) sprintf(np->clidtosend.id_val, id_val, uts_nodename()); 39167c478bd9Sstevel@tonic-gate bcopy(np->saddr.buf, &np->clidtosend.id_val[len], np->saddr.len); 39177c478bd9Sstevel@tonic-gate 39187c478bd9Sstevel@tonic-gate np->s_flags = 0; 39197c478bd9Sstevel@tonic-gate np->mntinfo4_list = NULL; 39207c478bd9Sstevel@tonic-gate /* save cred for issuing rfs4calls inside the renew thread */ 39217c478bd9Sstevel@tonic-gate crhold(cr); 39227c478bd9Sstevel@tonic-gate np->s_cred = cr; 39237c478bd9Sstevel@tonic-gate cv_init(&np->cv_thread_exit, NULL, CV_DEFAULT, NULL); 39247c478bd9Sstevel@tonic-gate mutex_init(&np->s_lock, NULL, MUTEX_DEFAULT, NULL); 39257c478bd9Sstevel@tonic-gate nfs_rw_init(&np->s_recovlock, NULL, RW_DEFAULT, NULL); 39267c478bd9Sstevel@tonic-gate list_create(&np->s_deleg_list, sizeof (rnode4_t), 39277c478bd9Sstevel@tonic-gate offsetof(rnode4_t, r_deleg_link)); 39287c478bd9Sstevel@tonic-gate np->s_thread_exit = 0; 39297c478bd9Sstevel@tonic-gate np->state_ref_count = 0; 39307c478bd9Sstevel@tonic-gate np->lease_valid = NFS4_LEASE_NOT_STARTED; 39317c478bd9Sstevel@tonic-gate cv_init(&np->s_cv_otw_count, NULL, CV_DEFAULT, NULL); 3932f86c6ccaSdm cv_init(&np->s_clientid_pend, NULL, CV_DEFAULT, NULL); 39337c478bd9Sstevel@tonic-gate np->s_otw_call_count = 0; 39347c478bd9Sstevel@tonic-gate cv_init(&np->wait_cb_null, NULL, CV_DEFAULT, NULL); 39357c478bd9Sstevel@tonic-gate np->zoneid = getzoneid(); 39367c478bd9Sstevel@tonic-gate np->zone_globals = nfs4_get_callback_globals(); 39377c478bd9Sstevel@tonic-gate ASSERT(np->zone_globals != NULL); 39387c478bd9Sstevel@tonic-gate return (np); 39397c478bd9Sstevel@tonic-gate } 39407c478bd9Sstevel@tonic-gate 39417c478bd9Sstevel@tonic-gate /* 39427c478bd9Sstevel@tonic-gate * Create a new nfs4_server_t structure and add it to the list. 39437c478bd9Sstevel@tonic-gate * Returns new node locked; reference must eventually be freed. 39447c478bd9Sstevel@tonic-gate */ 39457c478bd9Sstevel@tonic-gate static struct nfs4_server * 39467c478bd9Sstevel@tonic-gate add_new_nfs4_server(struct servinfo4 *svp, cred_t *cr) 39477c478bd9Sstevel@tonic-gate { 39487c478bd9Sstevel@tonic-gate nfs4_server_t *sp; 39497c478bd9Sstevel@tonic-gate 39507c478bd9Sstevel@tonic-gate ASSERT(MUTEX_HELD(&nfs4_server_lst_lock)); 39517c478bd9Sstevel@tonic-gate sp = new_nfs4_server(svp, cr); 39527c478bd9Sstevel@tonic-gate mutex_enter(&sp->s_lock); 39537c478bd9Sstevel@tonic-gate insque(sp, &nfs4_server_lst); 39547c478bd9Sstevel@tonic-gate sp->s_refcnt++; /* list gets a reference */ 3955f64c4ae1Sdm sp->s_flags |= N4S_INSERTED; 39567c478bd9Sstevel@tonic-gate sp->clientid = 0; 39577c478bd9Sstevel@tonic-gate return (sp); 39587c478bd9Sstevel@tonic-gate } 39597c478bd9Sstevel@tonic-gate 39607c478bd9Sstevel@tonic-gate int nfs4_server_t_debug = 0; 39617c478bd9Sstevel@tonic-gate 39627c478bd9Sstevel@tonic-gate #ifdef lint 39637c478bd9Sstevel@tonic-gate extern void 39647c478bd9Sstevel@tonic-gate dumpnfs4slist(char *, mntinfo4_t *, clientid4, servinfo4_t *); 39657c478bd9Sstevel@tonic-gate #endif 39667c478bd9Sstevel@tonic-gate 39677c478bd9Sstevel@tonic-gate #ifndef lint 39687c478bd9Sstevel@tonic-gate #ifdef DEBUG 39697c478bd9Sstevel@tonic-gate void 39707c478bd9Sstevel@tonic-gate dumpnfs4slist(char *txt, mntinfo4_t *mi, clientid4 clientid, servinfo4_t *srv_p) 39717c478bd9Sstevel@tonic-gate { 39727c478bd9Sstevel@tonic-gate int hash16(void *p, int len); 39737c478bd9Sstevel@tonic-gate nfs4_server_t *np; 39747c478bd9Sstevel@tonic-gate 39757c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_server_t_debug, (CE_NOTE, 39767c478bd9Sstevel@tonic-gate "dumping nfs4_server_t list in %s", txt)); 39777c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_server_t_debug, (CE_CONT, 39787c478bd9Sstevel@tonic-gate "mi 0x%p, want clientid %llx, addr %d/%04X", 39797c478bd9Sstevel@tonic-gate mi, (longlong_t)clientid, srv_p->sv_addr.len, 39807c478bd9Sstevel@tonic-gate hash16((void *)srv_p->sv_addr.buf, srv_p->sv_addr.len))); 39817c478bd9Sstevel@tonic-gate for (np = nfs4_server_lst.forw; np != &nfs4_server_lst; 39827c478bd9Sstevel@tonic-gate np = np->forw) { 39837c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_server_t_debug, (CE_CONT, 39847c478bd9Sstevel@tonic-gate "node 0x%p, clientid %llx, addr %d/%04X, cnt %d", 39857c478bd9Sstevel@tonic-gate np, (longlong_t)np->clientid, np->saddr.len, 39867c478bd9Sstevel@tonic-gate hash16((void *)np->saddr.buf, np->saddr.len), 39877c478bd9Sstevel@tonic-gate np->state_ref_count)); 39887c478bd9Sstevel@tonic-gate if (np->saddr.len == srv_p->sv_addr.len && 39897c478bd9Sstevel@tonic-gate bcmp(np->saddr.buf, srv_p->sv_addr.buf, 39907c478bd9Sstevel@tonic-gate np->saddr.len) == 0) 39917c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_server_t_debug, (CE_CONT, 39927c478bd9Sstevel@tonic-gate " - address matches")); 39937c478bd9Sstevel@tonic-gate if (np->clientid == clientid || np->clientid == 0) 39947c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_server_t_debug, (CE_CONT, 39957c478bd9Sstevel@tonic-gate " - clientid matches")); 39967c478bd9Sstevel@tonic-gate if (np->s_thread_exit != NFS4_THREAD_EXIT) 39977c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_server_t_debug, (CE_CONT, 39987c478bd9Sstevel@tonic-gate " - thread not exiting")); 39997c478bd9Sstevel@tonic-gate } 40007c478bd9Sstevel@tonic-gate delay(hz); 40017c478bd9Sstevel@tonic-gate } 40027c478bd9Sstevel@tonic-gate #endif 40037c478bd9Sstevel@tonic-gate #endif 40047c478bd9Sstevel@tonic-gate 40057c478bd9Sstevel@tonic-gate 40067c478bd9Sstevel@tonic-gate /* 40077c478bd9Sstevel@tonic-gate * Move a mntinfo4_t from one server list to another. 40087c478bd9Sstevel@tonic-gate * Locking of the two nfs4_server_t nodes will be done in list order. 40097c478bd9Sstevel@tonic-gate * 40107c478bd9Sstevel@tonic-gate * Returns NULL if the current nfs4_server_t for the filesystem could not 40117c478bd9Sstevel@tonic-gate * be found (e.g., due to forced unmount). Otherwise returns a reference 40127c478bd9Sstevel@tonic-gate * to the new nfs4_server_t, which must eventually be freed. 40137c478bd9Sstevel@tonic-gate */ 40147c478bd9Sstevel@tonic-gate nfs4_server_t * 40157c478bd9Sstevel@tonic-gate nfs4_move_mi(mntinfo4_t *mi, servinfo4_t *old, servinfo4_t *new) 40167c478bd9Sstevel@tonic-gate { 40177c478bd9Sstevel@tonic-gate nfs4_server_t *p, *op = NULL, *np = NULL; 40187c478bd9Sstevel@tonic-gate int num_open; 4019108322fbScarlsonj zoneid_t zoneid = nfs_zoneid(); 40207c478bd9Sstevel@tonic-gate 4021108322fbScarlsonj ASSERT(nfs_zone() == mi->mi_zone); 40227c478bd9Sstevel@tonic-gate 40237c478bd9Sstevel@tonic-gate mutex_enter(&nfs4_server_lst_lock); 40247c478bd9Sstevel@tonic-gate #ifdef DEBUG 40257c478bd9Sstevel@tonic-gate if (nfs4_server_t_debug) 40267c478bd9Sstevel@tonic-gate dumpnfs4slist("nfs4_move_mi", mi, (clientid4)0, new); 40277c478bd9Sstevel@tonic-gate #endif 40287c478bd9Sstevel@tonic-gate for (p = nfs4_server_lst.forw; p != &nfs4_server_lst; p = p->forw) { 40297c478bd9Sstevel@tonic-gate if (p->zoneid != zoneid) 40307c478bd9Sstevel@tonic-gate continue; 40317c478bd9Sstevel@tonic-gate if (p->saddr.len == old->sv_addr.len && 40327c478bd9Sstevel@tonic-gate bcmp(p->saddr.buf, old->sv_addr.buf, p->saddr.len) == 0 && 40337c478bd9Sstevel@tonic-gate p->s_thread_exit != NFS4_THREAD_EXIT) { 40347c478bd9Sstevel@tonic-gate op = p; 40357c478bd9Sstevel@tonic-gate mutex_enter(&op->s_lock); 40367c478bd9Sstevel@tonic-gate op->s_refcnt++; 40377c478bd9Sstevel@tonic-gate } 40387c478bd9Sstevel@tonic-gate if (p->saddr.len == new->sv_addr.len && 40397c478bd9Sstevel@tonic-gate bcmp(p->saddr.buf, new->sv_addr.buf, p->saddr.len) == 0 && 40407c478bd9Sstevel@tonic-gate p->s_thread_exit != NFS4_THREAD_EXIT) { 40417c478bd9Sstevel@tonic-gate np = p; 40427c478bd9Sstevel@tonic-gate mutex_enter(&np->s_lock); 40437c478bd9Sstevel@tonic-gate } 40447c478bd9Sstevel@tonic-gate if (op != NULL && np != NULL) 40457c478bd9Sstevel@tonic-gate break; 40467c478bd9Sstevel@tonic-gate } 40477c478bd9Sstevel@tonic-gate if (op == NULL) { 40487c478bd9Sstevel@tonic-gate /* 40497c478bd9Sstevel@tonic-gate * Filesystem has been forcibly unmounted. Bail out. 40507c478bd9Sstevel@tonic-gate */ 40517c478bd9Sstevel@tonic-gate if (np != NULL) 40527c478bd9Sstevel@tonic-gate mutex_exit(&np->s_lock); 40537c478bd9Sstevel@tonic-gate mutex_exit(&nfs4_server_lst_lock); 40547c478bd9Sstevel@tonic-gate return (NULL); 40557c478bd9Sstevel@tonic-gate } 40567c478bd9Sstevel@tonic-gate if (np != NULL) { 40577c478bd9Sstevel@tonic-gate np->s_refcnt++; 40587c478bd9Sstevel@tonic-gate } else { 40597c478bd9Sstevel@tonic-gate #ifdef DEBUG 40607c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_failover_debug, (CE_NOTE, 40617c478bd9Sstevel@tonic-gate "nfs4_move_mi: no target nfs4_server, will create.")); 40627c478bd9Sstevel@tonic-gate #endif 40637c478bd9Sstevel@tonic-gate np = add_new_nfs4_server(new, kcred); 40647c478bd9Sstevel@tonic-gate } 40657c478bd9Sstevel@tonic-gate mutex_exit(&nfs4_server_lst_lock); 40667c478bd9Sstevel@tonic-gate 40677c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_failover_debug, (CE_NOTE, 40687c478bd9Sstevel@tonic-gate "nfs4_move_mi: for mi 0x%p, " 40697c478bd9Sstevel@tonic-gate "old servinfo4 0x%p, new servinfo4 0x%p, " 40707c478bd9Sstevel@tonic-gate "old nfs4_server 0x%p, new nfs4_server 0x%p, ", 40717c478bd9Sstevel@tonic-gate (void*)mi, (void*)old, (void*)new, 40727c478bd9Sstevel@tonic-gate (void*)op, (void*)np)); 40737c478bd9Sstevel@tonic-gate ASSERT(op != NULL && np != NULL); 40747c478bd9Sstevel@tonic-gate 40757c478bd9Sstevel@tonic-gate /* discard any delegations */ 40767c478bd9Sstevel@tonic-gate nfs4_deleg_discard(mi, op); 40777c478bd9Sstevel@tonic-gate 40787c478bd9Sstevel@tonic-gate num_open = mi->mi_open_files; 40797c478bd9Sstevel@tonic-gate mi->mi_open_files = 0; 40807c478bd9Sstevel@tonic-gate op->state_ref_count -= num_open; 40817c478bd9Sstevel@tonic-gate ASSERT(op->state_ref_count >= 0); 40827c478bd9Sstevel@tonic-gate np->state_ref_count += num_open; 40837c478bd9Sstevel@tonic-gate nfs4_remove_mi_from_server_nolock(mi, op); 40847c478bd9Sstevel@tonic-gate mi->mi_open_files = num_open; 40857c478bd9Sstevel@tonic-gate NFS4_DEBUG(nfs4_client_failover_debug, (CE_NOTE, 40867c478bd9Sstevel@tonic-gate "nfs4_move_mi: mi_open_files %d, op->cnt %d, np->cnt %d", 40877c478bd9Sstevel@tonic-gate mi->mi_open_files, op->state_ref_count, np->state_ref_count)); 40887c478bd9Sstevel@tonic-gate 40897c478bd9Sstevel@tonic-gate nfs4_add_mi_to_server(np, mi); 40907c478bd9Sstevel@tonic-gate 40917c478bd9Sstevel@tonic-gate mutex_exit(&op->s_lock); 40927c478bd9Sstevel@tonic-gate mutex_exit(&np->s_lock); 409395c7fa91SPavel Filipensky nfs4_server_rele(op); 40947c478bd9Sstevel@tonic-gate 40957c478bd9Sstevel@tonic-gate return (np); 40967c478bd9Sstevel@tonic-gate } 40977c478bd9Sstevel@tonic-gate 40987c478bd9Sstevel@tonic-gate /* 4099f86c6ccaSdm * Need to have the nfs4_server_lst_lock. 41007c478bd9Sstevel@tonic-gate * Search the nfs4_server list to find a match on this servinfo4 41017c478bd9Sstevel@tonic-gate * based on its address. 41027c478bd9Sstevel@tonic-gate * 41037c478bd9Sstevel@tonic-gate * Returns NULL if no match is found. Otherwise returns a reference (which 41047c478bd9Sstevel@tonic-gate * must eventually be freed) to a locked nfs4_server. 41057c478bd9Sstevel@tonic-gate */ 41067c478bd9Sstevel@tonic-gate nfs4_server_t * 41077c478bd9Sstevel@tonic-gate servinfo4_to_nfs4_server(servinfo4_t *srv_p) 41087c478bd9Sstevel@tonic-gate { 41097c478bd9Sstevel@tonic-gate nfs4_server_t *np; 4110108322fbScarlsonj zoneid_t zoneid = nfs_zoneid(); 41117c478bd9Sstevel@tonic-gate 4112f86c6ccaSdm ASSERT(MUTEX_HELD(&nfs4_server_lst_lock)); 41137c478bd9Sstevel@tonic-gate for (np = nfs4_server_lst.forw; np != &nfs4_server_lst; np = np->forw) { 41147c478bd9Sstevel@tonic-gate if (np->zoneid == zoneid && 41157c478bd9Sstevel@tonic-gate np->saddr.len == srv_p->sv_addr.len && 41167c478bd9Sstevel@tonic-gate bcmp(np->saddr.buf, srv_p->sv_addr.buf, 4117b9238976Sth np->saddr.len) == 0 && 41187c478bd9Sstevel@tonic-gate np->s_thread_exit != NFS4_THREAD_EXIT) { 41197c478bd9Sstevel@tonic-gate mutex_enter(&np->s_lock); 41207c478bd9Sstevel@tonic-gate np->s_refcnt++; 41217c478bd9Sstevel@tonic-gate return (np); 41227c478bd9Sstevel@tonic-gate } 41237c478bd9Sstevel@tonic-gate } 41247c478bd9Sstevel@tonic-gate return (NULL); 41257c478bd9Sstevel@tonic-gate } 41267c478bd9Sstevel@tonic-gate 41277c478bd9Sstevel@tonic-gate /* 41287c478bd9Sstevel@tonic-gate * Locks the nfs4_server down if it is found and returns a reference that 41297c478bd9Sstevel@tonic-gate * must eventually be freed. 413022dc8f51SPavel Filipensky */ 413122dc8f51SPavel Filipensky static nfs4_server_t * 413222dc8f51SPavel Filipensky lookup_nfs4_server(nfs4_server_t *sp, int any_state) 413322dc8f51SPavel Filipensky { 413422dc8f51SPavel Filipensky nfs4_server_t *np; 413522dc8f51SPavel Filipensky 413622dc8f51SPavel Filipensky mutex_enter(&nfs4_server_lst_lock); 413722dc8f51SPavel Filipensky for (np = nfs4_server_lst.forw; np != &nfs4_server_lst; np = np->forw) { 413822dc8f51SPavel Filipensky mutex_enter(&np->s_lock); 413922dc8f51SPavel Filipensky if (np == sp && np->s_refcnt > 0 && 414022dc8f51SPavel Filipensky (np->s_thread_exit != NFS4_THREAD_EXIT || any_state)) { 414122dc8f51SPavel Filipensky mutex_exit(&nfs4_server_lst_lock); 414222dc8f51SPavel Filipensky np->s_refcnt++; 414322dc8f51SPavel Filipensky return (np); 414422dc8f51SPavel Filipensky } 414522dc8f51SPavel Filipensky mutex_exit(&np->s_lock); 414622dc8f51SPavel Filipensky } 414722dc8f51SPavel Filipensky mutex_exit(&nfs4_server_lst_lock); 414822dc8f51SPavel Filipensky 414922dc8f51SPavel Filipensky return (NULL); 415022dc8f51SPavel Filipensky } 415122dc8f51SPavel Filipensky 415222dc8f51SPavel Filipensky /* 41537c478bd9Sstevel@tonic-gate * The caller should be holding mi->mi_recovlock, and it should continue to 41547c478bd9Sstevel@tonic-gate * hold the lock until done with the returned nfs4_server_t. Once 41557c478bd9Sstevel@tonic-gate * mi->mi_recovlock is released, there is no guarantee that the returned 41567c478bd9Sstevel@tonic-gate * mi->nfs4_server_t will continue to correspond to mi. 41577c478bd9Sstevel@tonic-gate */ 41587c478bd9Sstevel@tonic-gate nfs4_server_t * 41597c478bd9Sstevel@tonic-gate find_nfs4_server(mntinfo4_t *mi) 41607c478bd9Sstevel@tonic-gate { 416122dc8f51SPavel Filipensky ASSERT(nfs_rw_lock_held(&mi->mi_recovlock, RW_READER) || 416222dc8f51SPavel Filipensky nfs_rw_lock_held(&mi->mi_recovlock, RW_WRITER)); 416322dc8f51SPavel Filipensky 416422dc8f51SPavel Filipensky return (lookup_nfs4_server(mi->mi_srv, 0)); 41657c478bd9Sstevel@tonic-gate } 41667c478bd9Sstevel@tonic-gate 41677c478bd9Sstevel@tonic-gate /* 416822dc8f51SPavel Filipensky * Same as above, but takes an "any_state" parameter which can be 41697c478bd9Sstevel@tonic-gate * set to 1 if the caller wishes to find nfs4_server_t's which 41707c478bd9Sstevel@tonic-gate * have been marked for termination by the exit of the renew 41717c478bd9Sstevel@tonic-gate * thread. This should only be used by operations which are 41727c478bd9Sstevel@tonic-gate * cleaning up and will not cause an OTW op. 41737c478bd9Sstevel@tonic-gate */ 41747c478bd9Sstevel@tonic-gate nfs4_server_t * 417522dc8f51SPavel Filipensky find_nfs4_server_all(mntinfo4_t *mi, int any_state) 41767c478bd9Sstevel@tonic-gate { 41777c478bd9Sstevel@tonic-gate ASSERT(nfs_rw_lock_held(&mi->mi_recovlock, RW_READER) || 41787c478bd9Sstevel@tonic-gate nfs_rw_lock_held(&mi->mi_recovlock, RW_WRITER)); 41797c478bd9Sstevel@tonic-gate 418022dc8f51SPavel Filipensky return (lookup_nfs4_server(mi->mi_srv, any_state)); 418122dc8f51SPavel Filipensky } 418222dc8f51SPavel Filipensky 418322dc8f51SPavel Filipensky /* 418422dc8f51SPavel Filipensky * Lock sp, but only if it's still active (in the list and hasn't been 418522dc8f51SPavel Filipensky * flagged as exiting) or 'any_state' is non-zero. 418622dc8f51SPavel Filipensky * Returns TRUE if sp got locked and adds a reference to sp. 418722dc8f51SPavel Filipensky */ 418822dc8f51SPavel Filipensky bool_t 418922dc8f51SPavel Filipensky nfs4_server_vlock(nfs4_server_t *sp, int any_state) 419022dc8f51SPavel Filipensky { 419122dc8f51SPavel Filipensky return (lookup_nfs4_server(sp, any_state) != NULL); 41927c478bd9Sstevel@tonic-gate } 41937c478bd9Sstevel@tonic-gate 41947c478bd9Sstevel@tonic-gate /* 41957c478bd9Sstevel@tonic-gate * Release the reference to sp and destroy it if that's the last one. 41967c478bd9Sstevel@tonic-gate */ 41977c478bd9Sstevel@tonic-gate 41987c478bd9Sstevel@tonic-gate void 41997c478bd9Sstevel@tonic-gate nfs4_server_rele(nfs4_server_t *sp) 42007c478bd9Sstevel@tonic-gate { 42017c478bd9Sstevel@tonic-gate mutex_enter(&sp->s_lock); 42027c478bd9Sstevel@tonic-gate ASSERT(sp->s_refcnt > 0); 42037c478bd9Sstevel@tonic-gate sp->s_refcnt--; 42047c478bd9Sstevel@tonic-gate if (sp->s_refcnt > 0) { 42057c478bd9Sstevel@tonic-gate mutex_exit(&sp->s_lock); 42067c478bd9Sstevel@tonic-gate return; 42077c478bd9Sstevel@tonic-gate } 42087c478bd9Sstevel@tonic-gate mutex_exit(&sp->s_lock); 4209f86c6ccaSdm 42107c478bd9Sstevel@tonic-gate mutex_enter(&nfs4_server_lst_lock); 42117c478bd9Sstevel@tonic-gate mutex_enter(&sp->s_lock); 42127c478bd9Sstevel@tonic-gate if (sp->s_refcnt > 0) { 42137c478bd9Sstevel@tonic-gate mutex_exit(&sp->s_lock); 42147c478bd9Sstevel@tonic-gate mutex_exit(&nfs4_server_lst_lock); 42157c478bd9Sstevel@tonic-gate return; 42167c478bd9Sstevel@tonic-gate } 4217f86c6ccaSdm remque(sp); 4218f86c6ccaSdm sp->forw = sp->back = NULL; 42197c478bd9Sstevel@tonic-gate mutex_exit(&nfs4_server_lst_lock); 42207c478bd9Sstevel@tonic-gate destroy_nfs4_server(sp); 42217c478bd9Sstevel@tonic-gate } 42227c478bd9Sstevel@tonic-gate 42237c478bd9Sstevel@tonic-gate static void 42247c478bd9Sstevel@tonic-gate destroy_nfs4_server(nfs4_server_t *sp) 42257c478bd9Sstevel@tonic-gate { 42267c478bd9Sstevel@tonic-gate ASSERT(MUTEX_HELD(&sp->s_lock)); 42277c478bd9Sstevel@tonic-gate ASSERT(sp->s_refcnt == 0); 42287c478bd9Sstevel@tonic-gate ASSERT(sp->s_otw_call_count == 0); 42297c478bd9Sstevel@tonic-gate 42307c478bd9Sstevel@tonic-gate remove_all_mi(sp); 42317c478bd9Sstevel@tonic-gate 42327c478bd9Sstevel@tonic-gate crfree(sp->s_cred); 42337c478bd9Sstevel@tonic-gate kmem_free(sp->saddr.buf, sp->saddr.maxlen); 42347c478bd9Sstevel@tonic-gate kmem_free(sp->clidtosend.id_val, sp->clidtosend.id_len); 42357c478bd9Sstevel@tonic-gate mutex_exit(&sp->s_lock); 42367c478bd9Sstevel@tonic-gate 42377c478bd9Sstevel@tonic-gate /* destroy the nfs4_server */ 42387c478bd9Sstevel@tonic-gate nfs4callback_destroy(sp); 42397c478bd9Sstevel@tonic-gate list_destroy(&sp->s_deleg_list); 42407c478bd9Sstevel@tonic-gate mutex_destroy(&sp->s_lock); 42417c478bd9Sstevel@tonic-gate cv_destroy(&sp->cv_thread_exit); 42427c478bd9Sstevel@tonic-gate cv_destroy(&sp->s_cv_otw_count); 4243f86c6ccaSdm cv_destroy(&sp->s_clientid_pend); 42447c478bd9Sstevel@tonic-gate cv_destroy(&sp->wait_cb_null); 42457c478bd9Sstevel@tonic-gate nfs_rw_destroy(&sp->s_recovlock); 42467c478bd9Sstevel@tonic-gate kmem_free(sp, sizeof (*sp)); 42477c478bd9Sstevel@tonic-gate } 42487c478bd9Sstevel@tonic-gate 42497c478bd9Sstevel@tonic-gate /* 42507c478bd9Sstevel@tonic-gate * Fork off a thread to free the data structures for a mount. 42517c478bd9Sstevel@tonic-gate */ 42527c478bd9Sstevel@tonic-gate 42537c478bd9Sstevel@tonic-gate static void 4254b9238976Sth async_free_mount(vfs_t *vfsp, int flag, cred_t *cr) 42557c478bd9Sstevel@tonic-gate { 42567c478bd9Sstevel@tonic-gate freemountargs_t *args; 42577c478bd9Sstevel@tonic-gate args = kmem_alloc(sizeof (freemountargs_t), KM_SLEEP); 42587c478bd9Sstevel@tonic-gate args->fm_vfsp = vfsp; 42597c478bd9Sstevel@tonic-gate VFS_HOLD(vfsp); 426050a83466Sjwahlig MI4_HOLD(VFTOMI4(vfsp)); 4261b9238976Sth args->fm_flag = flag; 42627c478bd9Sstevel@tonic-gate args->fm_cr = cr; 42637c478bd9Sstevel@tonic-gate crhold(cr); 42647c478bd9Sstevel@tonic-gate (void) zthread_create(NULL, 0, nfs4_free_mount_thread, args, 0, 42657c478bd9Sstevel@tonic-gate minclsyspri); 42667c478bd9Sstevel@tonic-gate } 42677c478bd9Sstevel@tonic-gate 42687c478bd9Sstevel@tonic-gate static void 42697c478bd9Sstevel@tonic-gate nfs4_free_mount_thread(freemountargs_t *args) 42707c478bd9Sstevel@tonic-gate { 427150a83466Sjwahlig mntinfo4_t *mi; 4272b9238976Sth nfs4_free_mount(args->fm_vfsp, args->fm_flag, args->fm_cr); 427350a83466Sjwahlig mi = VFTOMI4(args->fm_vfsp); 42747c478bd9Sstevel@tonic-gate crfree(args->fm_cr); 427550a83466Sjwahlig VFS_RELE(args->fm_vfsp); 427650a83466Sjwahlig MI4_RELE(mi); 42777c478bd9Sstevel@tonic-gate kmem_free(args, sizeof (freemountargs_t)); 42787c478bd9Sstevel@tonic-gate zthread_exit(); 42797c478bd9Sstevel@tonic-gate /* NOTREACHED */ 42807c478bd9Sstevel@tonic-gate } 42817c478bd9Sstevel@tonic-gate 42827c478bd9Sstevel@tonic-gate /* 42837c478bd9Sstevel@tonic-gate * Thread to free the data structures for a given filesystem. 42847c478bd9Sstevel@tonic-gate */ 42857c478bd9Sstevel@tonic-gate static void 4286b9238976Sth nfs4_free_mount(vfs_t *vfsp, int flag, cred_t *cr) 42877c478bd9Sstevel@tonic-gate { 4288b9238976Sth mntinfo4_t *mi = VFTOMI4(vfsp); 4289b9238976Sth nfs4_server_t *sp; 4290b9238976Sth callb_cpr_t cpr_info; 4291b9238976Sth kmutex_t cpr_lock; 4292b9238976Sth boolean_t async_thread; 4293b9238976Sth int removed; 4294b9238976Sth 4295d3a14591SThomas Haynes bool_t must_unlock; 4296b9238976Sth nfs4_ephemeral_tree_t *eph_tree; 42977c478bd9Sstevel@tonic-gate 42987c478bd9Sstevel@tonic-gate /* 42997c478bd9Sstevel@tonic-gate * We need to participate in the CPR framework if this is a kernel 43007c478bd9Sstevel@tonic-gate * thread. 43017c478bd9Sstevel@tonic-gate */ 4302108322fbScarlsonj async_thread = (curproc == nfs_zone()->zone_zsched); 43037c478bd9Sstevel@tonic-gate if (async_thread) { 43047c478bd9Sstevel@tonic-gate mutex_init(&cpr_lock, NULL, MUTEX_DEFAULT, NULL); 43057c478bd9Sstevel@tonic-gate CALLB_CPR_INIT(&cpr_info, &cpr_lock, callb_generic_cpr, 43067c478bd9Sstevel@tonic-gate "nfsv4AsyncUnmount"); 43077c478bd9Sstevel@tonic-gate } 43087c478bd9Sstevel@tonic-gate 43097c478bd9Sstevel@tonic-gate /* 43107c478bd9Sstevel@tonic-gate * We need to wait for all outstanding OTW calls 43117c478bd9Sstevel@tonic-gate * and recovery to finish before we remove the mi 43127c478bd9Sstevel@tonic-gate * from the nfs4_server_t, as current pending 43137c478bd9Sstevel@tonic-gate * calls might still need this linkage (in order 43147c478bd9Sstevel@tonic-gate * to find a nfs4_server_t from a mntinfo4_t). 43157c478bd9Sstevel@tonic-gate */ 43167c478bd9Sstevel@tonic-gate (void) nfs_rw_enter_sig(&mi->mi_recovlock, RW_READER, FALSE); 43177c478bd9Sstevel@tonic-gate sp = find_nfs4_server(mi); 43187c478bd9Sstevel@tonic-gate nfs_rw_exit(&mi->mi_recovlock); 43197c478bd9Sstevel@tonic-gate 43207c478bd9Sstevel@tonic-gate if (sp) { 43217c478bd9Sstevel@tonic-gate while (sp->s_otw_call_count != 0) { 43227c478bd9Sstevel@tonic-gate if (async_thread) { 43237c478bd9Sstevel@tonic-gate mutex_enter(&cpr_lock); 43247c478bd9Sstevel@tonic-gate CALLB_CPR_SAFE_BEGIN(&cpr_info); 43257c478bd9Sstevel@tonic-gate mutex_exit(&cpr_lock); 43267c478bd9Sstevel@tonic-gate } 43277c478bd9Sstevel@tonic-gate cv_wait(&sp->s_cv_otw_count, &sp->s_lock); 43287c478bd9Sstevel@tonic-gate if (async_thread) { 43297c478bd9Sstevel@tonic-gate mutex_enter(&cpr_lock); 43307c478bd9Sstevel@tonic-gate CALLB_CPR_SAFE_END(&cpr_info, &cpr_lock); 43317c478bd9Sstevel@tonic-gate mutex_exit(&cpr_lock); 43327c478bd9Sstevel@tonic-gate } 43337c478bd9Sstevel@tonic-gate } 43347c478bd9Sstevel@tonic-gate mutex_exit(&sp->s_lock); 43357c478bd9Sstevel@tonic-gate nfs4_server_rele(sp); 43367c478bd9Sstevel@tonic-gate sp = NULL; 43377c478bd9Sstevel@tonic-gate } 43387c478bd9Sstevel@tonic-gate 43397c478bd9Sstevel@tonic-gate mutex_enter(&mi->mi_lock); 43407c478bd9Sstevel@tonic-gate while (mi->mi_in_recovery != 0) { 43417c478bd9Sstevel@tonic-gate if (async_thread) { 43427c478bd9Sstevel@tonic-gate mutex_enter(&cpr_lock); 43437c478bd9Sstevel@tonic-gate CALLB_CPR_SAFE_BEGIN(&cpr_info); 43447c478bd9Sstevel@tonic-gate mutex_exit(&cpr_lock); 43457c478bd9Sstevel@tonic-gate } 43467c478bd9Sstevel@tonic-gate cv_wait(&mi->mi_cv_in_recov, &mi->mi_lock); 43477c478bd9Sstevel@tonic-gate if (async_thread) { 43487c478bd9Sstevel@tonic-gate mutex_enter(&cpr_lock); 43497c478bd9Sstevel@tonic-gate CALLB_CPR_SAFE_END(&cpr_info, &cpr_lock); 43507c478bd9Sstevel@tonic-gate mutex_exit(&cpr_lock); 43517c478bd9Sstevel@tonic-gate } 43527c478bd9Sstevel@tonic-gate } 43537c478bd9Sstevel@tonic-gate mutex_exit(&mi->mi_lock); 43547c478bd9Sstevel@tonic-gate 4355eabd0450Sth /* 4356eabd0450Sth * If we got an error, then do not nuke the 4357eabd0450Sth * tree. Either the harvester is busy reclaiming 4358eabd0450Sth * this node or we ran into some busy condition. 4359eabd0450Sth * 4360eabd0450Sth * The harvester will eventually come along and cleanup. 4361eabd0450Sth * The only problem would be the root mount point. 4362eabd0450Sth * 4363eabd0450Sth * Since the busy node can occur for a variety 4364eabd0450Sth * of reasons and can result in an entry staying 4365eabd0450Sth * in df output but no longer accessible from the 4366eabd0450Sth * directory tree, we are okay. 4367eabd0450Sth */ 4368eabd0450Sth if (!nfs4_ephemeral_umount(mi, flag, cr, 43692f172c55SRobert Thurlow &must_unlock, &eph_tree)) 4370eabd0450Sth nfs4_ephemeral_umount_activate(mi, &must_unlock, 43712f172c55SRobert Thurlow &eph_tree); 4372b9238976Sth 43737c478bd9Sstevel@tonic-gate /* 43747c478bd9Sstevel@tonic-gate * The original purge of the dnlc via 'dounmount' 43757c478bd9Sstevel@tonic-gate * doesn't guarantee that another dnlc entry was not 43767c478bd9Sstevel@tonic-gate * added while we waitied for all outstanding OTW 43777c478bd9Sstevel@tonic-gate * and recovery calls to finish. So re-purge the 43787c478bd9Sstevel@tonic-gate * dnlc now. 43797c478bd9Sstevel@tonic-gate */ 43807c478bd9Sstevel@tonic-gate (void) dnlc_purge_vfsp(vfsp, 0); 43817c478bd9Sstevel@tonic-gate 43827c478bd9Sstevel@tonic-gate /* 43837c478bd9Sstevel@tonic-gate * We need to explicitly stop the manager thread; the asyc worker 43847c478bd9Sstevel@tonic-gate * threads can timeout and exit on their own. 43857c478bd9Sstevel@tonic-gate */ 438650a83466Sjwahlig mutex_enter(&mi->mi_async_lock); 438750a83466Sjwahlig mi->mi_max_threads = 0; 43880776f5e6SVallish Vaidyeshwara NFS4_WAKEALL_ASYNC_WORKERS(mi->mi_async_work_cv); 438950a83466Sjwahlig mutex_exit(&mi->mi_async_lock); 439050a83466Sjwahlig if (mi->mi_manager_thread) 439150a83466Sjwahlig nfs4_async_manager_stop(vfsp); 43927c478bd9Sstevel@tonic-gate 43937c478bd9Sstevel@tonic-gate destroy_rtable4(vfsp, cr); 43947c478bd9Sstevel@tonic-gate 43957c478bd9Sstevel@tonic-gate nfs4_remove_mi_from_server(mi, NULL); 43967c478bd9Sstevel@tonic-gate 43977c478bd9Sstevel@tonic-gate if (async_thread) { 43987c478bd9Sstevel@tonic-gate mutex_enter(&cpr_lock); 43997c478bd9Sstevel@tonic-gate CALLB_CPR_EXIT(&cpr_info); /* drops cpr_lock */ 44007c478bd9Sstevel@tonic-gate mutex_destroy(&cpr_lock); 44017c478bd9Sstevel@tonic-gate } 440250a83466Sjwahlig 440350a83466Sjwahlig removed = nfs4_mi_zonelist_remove(mi); 440450a83466Sjwahlig if (removed) 440550a83466Sjwahlig zone_rele(mi->mi_zone); 44067c478bd9Sstevel@tonic-gate } 44072f172c55SRobert Thurlow 44082f172c55SRobert Thurlow /* Referral related sub-routines */ 44092f172c55SRobert Thurlow 44102f172c55SRobert Thurlow /* Freeup knetconfig */ 44112f172c55SRobert Thurlow static void 44122f172c55SRobert Thurlow free_knconf_contents(struct knetconfig *k) 44132f172c55SRobert Thurlow { 44142f172c55SRobert Thurlow if (k == NULL) 44152f172c55SRobert Thurlow return; 44162f172c55SRobert Thurlow if (k->knc_protofmly) 44172f172c55SRobert Thurlow kmem_free(k->knc_protofmly, KNC_STRSIZE); 44182f172c55SRobert Thurlow if (k->knc_proto) 44192f172c55SRobert Thurlow kmem_free(k->knc_proto, KNC_STRSIZE); 44202f172c55SRobert Thurlow } 44212f172c55SRobert Thurlow 44222f172c55SRobert Thurlow /* 44232f172c55SRobert Thurlow * This updates newpath variable with exact name component from the 44242f172c55SRobert Thurlow * path which gave us a NFS4ERR_MOVED error. 44252f172c55SRobert Thurlow * If the path is /rp/aaa/bbb and nth value is 1, aaa is returned. 44262f172c55SRobert Thurlow */ 44272f172c55SRobert Thurlow static char * 44282f172c55SRobert Thurlow extract_referral_point(const char *svp, int nth) 44292f172c55SRobert Thurlow { 44302f172c55SRobert Thurlow int num_slashes = 0; 44312f172c55SRobert Thurlow const char *p; 44322f172c55SRobert Thurlow char *newpath = NULL; 44332f172c55SRobert Thurlow int i = 0; 44342f172c55SRobert Thurlow 44352f172c55SRobert Thurlow newpath = kmem_zalloc(MAXPATHLEN, KM_SLEEP); 44362f172c55SRobert Thurlow for (p = svp; *p; p++) { 44372f172c55SRobert Thurlow if (*p == '/') 44382f172c55SRobert Thurlow num_slashes++; 44392f172c55SRobert Thurlow if (num_slashes == nth + 1) { 44402f172c55SRobert Thurlow p++; 44412f172c55SRobert Thurlow while (*p != '/') { 44422f172c55SRobert Thurlow if (*p == '\0') 44432f172c55SRobert Thurlow break; 44442f172c55SRobert Thurlow newpath[i] = *p; 44452f172c55SRobert Thurlow i++; 44462f172c55SRobert Thurlow p++; 44472f172c55SRobert Thurlow } 44482f172c55SRobert Thurlow newpath[i++] = '\0'; 44492f172c55SRobert Thurlow break; 44502f172c55SRobert Thurlow } 44512f172c55SRobert Thurlow } 44522f172c55SRobert Thurlow return (newpath); 44532f172c55SRobert Thurlow } 44542f172c55SRobert Thurlow 44552f172c55SRobert Thurlow /* 44562f172c55SRobert Thurlow * This sets up a new path in sv_path to do a lookup of the referral point. 44572f172c55SRobert Thurlow * If the path is /rp/aaa/bbb and the referral point is aaa, 44582f172c55SRobert Thurlow * this updates /rp/aaa. This path will be used to get referral 44592f172c55SRobert Thurlow * location. 44602f172c55SRobert Thurlow */ 44612f172c55SRobert Thurlow static void 44622f172c55SRobert Thurlow setup_newsvpath(servinfo4_t *svp, int nth) 44632f172c55SRobert Thurlow { 44642f172c55SRobert Thurlow int num_slashes = 0, pathlen, i = 0; 44652f172c55SRobert Thurlow char *newpath, *p; 44662f172c55SRobert Thurlow 44672f172c55SRobert Thurlow newpath = kmem_zalloc(MAXPATHLEN, KM_SLEEP); 44682f172c55SRobert Thurlow for (p = svp->sv_path; *p; p++) { 44692f172c55SRobert Thurlow newpath[i] = *p; 44702f172c55SRobert Thurlow if (*p == '/') 44712f172c55SRobert Thurlow num_slashes++; 44722f172c55SRobert Thurlow if (num_slashes == nth + 1) { 44732f172c55SRobert Thurlow newpath[i] = '\0'; 44742f172c55SRobert Thurlow pathlen = strlen(newpath) + 1; 44752f172c55SRobert Thurlow kmem_free(svp->sv_path, svp->sv_pathlen); 44762f172c55SRobert Thurlow svp->sv_path = kmem_alloc(pathlen, KM_SLEEP); 44772f172c55SRobert Thurlow svp->sv_pathlen = pathlen; 44782f172c55SRobert Thurlow bcopy(newpath, svp->sv_path, pathlen); 44792f172c55SRobert Thurlow break; 44802f172c55SRobert Thurlow } 44812f172c55SRobert Thurlow i++; 44822f172c55SRobert Thurlow } 44832f172c55SRobert Thurlow kmem_free(newpath, MAXPATHLEN); 44842f172c55SRobert Thurlow } 4485