diff --git a/sys/fs/nfs/nfs_commonacl.c b/sys/fs/nfs/nfs_commonacl.c
index 19492675e731..b733dc52803f 100644
--- a/sys/fs/nfs/nfs_commonacl.c
+++ b/sys/fs/nfs/nfs_commonacl.c
@@ -1,487 +1,487 @@
/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (c) 2009 Rick Macklem, University of Guelph
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <fs/nfs/nfsport.h>
extern int nfsrv_useacl;
static int nfsrv_acemasktoperm(u_int32_t acetype, u_int32_t mask, int owner,
enum vtype type, acl_perm_t *permp);
/*
* Handle xdr for an ace.
*/
int
nfsrv_dissectace(struct nfsrv_descript *nd, struct acl_entry *acep,
- int *aceerrp, int *acesizep, NFSPROC_T *p)
+ bool server, int *aceerrp, int *acesizep, NFSPROC_T *p)
{
u_int32_t *tl;
int len, gotid = 0, owner = 0, error = 0, aceerr = 0;
u_char *name, namestr[NFSV4_SMALLSTR + 1];
u_int32_t flag, mask, acetype;
gid_t gid;
uid_t uid;
*aceerrp = 0;
acep->ae_flags = 0;
NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
acetype = fxdr_unsigned(u_int32_t, *tl++);
flag = fxdr_unsigned(u_int32_t, *tl++);
mask = fxdr_unsigned(u_int32_t, *tl++);
len = fxdr_unsigned(int, *tl);
/*
* The RFCs do not specify a limit to the length of the "who", but
* NFSV4_OPAQUELIMIT (1024) should be sufficient.
*/
if (len < 0 || len > NFSV4_OPAQUELIMIT) {
error = NFSERR_BADXDR;
goto nfsmout;
} else if (len == 0) {
/* Netapp filers return a 0 length who for nil users */
acep->ae_tag = ACL_UNDEFINED_TAG;
acep->ae_id = ACL_UNDEFINED_ID;
acep->ae_perm = (acl_perm_t)0;
acep->ae_entry_type = ACL_ENTRY_TYPE_DENY;
if (acesizep)
*acesizep = 4 * NFSX_UNSIGNED;
error = 0;
goto nfsmout;
}
if (len > NFSV4_SMALLSTR)
name = malloc(len + 1, M_NFSSTRING, M_WAITOK);
else
name = namestr;
error = nfsrv_mtostr(nd, name, len);
if (error) {
if (len > NFSV4_SMALLSTR)
free(name, M_NFSSTRING);
goto nfsmout;
}
if (len == 6) {
if (!NFSBCMP(name, "OWNER@", 6)) {
acep->ae_tag = ACL_USER_OBJ;
acep->ae_id = ACL_UNDEFINED_ID;
owner = 1;
gotid = 1;
} else if (!NFSBCMP(name, "GROUP@", 6)) {
acep->ae_tag = ACL_GROUP_OBJ;
acep->ae_id = ACL_UNDEFINED_ID;
gotid = 1;
}
} else if (len == 9 && !NFSBCMP(name, "EVERYONE@", 9)) {
acep->ae_tag = ACL_EVERYONE;
acep->ae_id = ACL_UNDEFINED_ID;
gotid = 1;
}
if (gotid == 0) {
if (flag & NFSV4ACE_IDENTIFIERGROUP) {
acep->ae_tag = ACL_GROUP;
aceerr = nfsv4_strtogid(nd, name, len, &gid);
if (aceerr == 0)
acep->ae_id = (uid_t)gid;
} else {
acep->ae_tag = ACL_USER;
aceerr = nfsv4_strtouid(nd, name, len, &uid);
if (aceerr == 0)
acep->ae_id = uid;
}
}
if (len > NFSV4_SMALLSTR)
free(name, M_NFSSTRING);
if (aceerr == 0) {
/*
* Handle the flags.
*/
flag &= ~NFSV4ACE_IDENTIFIERGROUP;
if (flag & NFSV4ACE_FILEINHERIT) {
flag &= ~NFSV4ACE_FILEINHERIT;
acep->ae_flags |= ACL_ENTRY_FILE_INHERIT;
}
if (flag & NFSV4ACE_DIRECTORYINHERIT) {
flag &= ~NFSV4ACE_DIRECTORYINHERIT;
acep->ae_flags |= ACL_ENTRY_DIRECTORY_INHERIT;
}
if (flag & NFSV4ACE_NOPROPAGATEINHERIT) {
flag &= ~NFSV4ACE_NOPROPAGATEINHERIT;
acep->ae_flags |= ACL_ENTRY_NO_PROPAGATE_INHERIT;
}
if (flag & NFSV4ACE_INHERITONLY) {
flag &= ~NFSV4ACE_INHERITONLY;
acep->ae_flags |= ACL_ENTRY_INHERIT_ONLY;
}
if (flag & NFSV4ACE_SUCCESSFULACCESS) {
flag &= ~NFSV4ACE_SUCCESSFULACCESS;
acep->ae_flags |= ACL_ENTRY_SUCCESSFUL_ACCESS;
}
if (flag & NFSV4ACE_FAILEDACCESS) {
flag &= ~NFSV4ACE_FAILEDACCESS;
acep->ae_flags |= ACL_ENTRY_FAILED_ACCESS;
}
/*
* Set ae_entry_type.
*/
if (acetype == NFSV4ACE_ALLOWEDTYPE)
acep->ae_entry_type = ACL_ENTRY_TYPE_ALLOW;
else if (acetype == NFSV4ACE_DENIEDTYPE)
acep->ae_entry_type = ACL_ENTRY_TYPE_DENY;
- else if (acetype == NFSV4ACE_AUDITTYPE)
+ else if (!server && acetype == NFSV4ACE_AUDITTYPE)
acep->ae_entry_type = ACL_ENTRY_TYPE_AUDIT;
- else if (acetype == NFSV4ACE_ALARMTYPE)
+ else if (!server && acetype == NFSV4ACE_ALARMTYPE)
acep->ae_entry_type = ACL_ENTRY_TYPE_ALARM;
else
aceerr = NFSERR_ATTRNOTSUPP;
}
/*
* Now, check for unsupported flag bits.
*/
if (aceerr == 0 && flag != 0)
aceerr = NFSERR_ATTRNOTSUPP;
/*
* And turn the mask into perm bits.
*/
if (aceerr == 0)
aceerr = nfsrv_acemasktoperm(acetype, mask, owner, VREG,
&acep->ae_perm);
*aceerrp = aceerr;
if (acesizep)
*acesizep = NFSM_RNDUP(len) + (4 * NFSX_UNSIGNED);
error = 0;
nfsmout:
NFSEXITCODE(error);
return (error);
}
/*
* Turn an NFSv4 ace mask into R/W/X flag bits.
*/
static int
nfsrv_acemasktoperm(u_int32_t acetype, u_int32_t mask, int owner,
enum vtype type, acl_perm_t *permp)
{
acl_perm_t perm = 0x0;
int error = 0;
if (mask & NFSV4ACE_READDATA) {
mask &= ~NFSV4ACE_READDATA;
perm |= ACL_READ_DATA;
}
if (mask & NFSV4ACE_LISTDIRECTORY) {
mask &= ~NFSV4ACE_LISTDIRECTORY;
perm |= ACL_LIST_DIRECTORY;
}
if (mask & NFSV4ACE_WRITEDATA) {
mask &= ~NFSV4ACE_WRITEDATA;
perm |= ACL_WRITE_DATA;
}
if (mask & NFSV4ACE_ADDFILE) {
mask &= ~NFSV4ACE_ADDFILE;
perm |= ACL_ADD_FILE;
}
if (mask & NFSV4ACE_APPENDDATA) {
mask &= ~NFSV4ACE_APPENDDATA;
perm |= ACL_APPEND_DATA;
}
if (mask & NFSV4ACE_ADDSUBDIRECTORY) {
mask &= ~NFSV4ACE_ADDSUBDIRECTORY;
perm |= ACL_ADD_SUBDIRECTORY;
}
if (mask & NFSV4ACE_READNAMEDATTR) {
mask &= ~NFSV4ACE_READNAMEDATTR;
perm |= ACL_READ_NAMED_ATTRS;
}
if (mask & NFSV4ACE_WRITENAMEDATTR) {
mask &= ~NFSV4ACE_WRITENAMEDATTR;
perm |= ACL_WRITE_NAMED_ATTRS;
}
if (mask & NFSV4ACE_EXECUTE) {
mask &= ~NFSV4ACE_EXECUTE;
perm |= ACL_EXECUTE;
}
if (mask & NFSV4ACE_SEARCH) {
mask &= ~NFSV4ACE_SEARCH;
perm |= ACL_EXECUTE;
}
if (mask & NFSV4ACE_DELETECHILD) {
mask &= ~NFSV4ACE_DELETECHILD;
perm |= ACL_DELETE_CHILD;
}
if (mask & NFSV4ACE_READATTRIBUTES) {
mask &= ~NFSV4ACE_READATTRIBUTES;
perm |= ACL_READ_ATTRIBUTES;
}
if (mask & NFSV4ACE_WRITEATTRIBUTES) {
mask &= ~NFSV4ACE_WRITEATTRIBUTES;
perm |= ACL_WRITE_ATTRIBUTES;
}
if (mask & NFSV4ACE_DELETE) {
mask &= ~NFSV4ACE_DELETE;
perm |= ACL_DELETE;
}
if (mask & NFSV4ACE_READACL) {
mask &= ~NFSV4ACE_READACL;
perm |= ACL_READ_ACL;
}
if (mask & NFSV4ACE_WRITEACL) {
mask &= ~NFSV4ACE_WRITEACL;
perm |= ACL_WRITE_ACL;
}
if (mask & NFSV4ACE_WRITEOWNER) {
mask &= ~NFSV4ACE_WRITEOWNER;
perm |= ACL_WRITE_OWNER;
}
if (mask & NFSV4ACE_SYNCHRONIZE) {
mask &= ~NFSV4ACE_SYNCHRONIZE;
perm |= ACL_SYNCHRONIZE;
}
if (mask != 0) {
error = NFSERR_ATTRNOTSUPP;
goto out;
}
*permp = perm;
out:
NFSEXITCODE(error);
return (error);
}
/* local functions */
static int nfsrv_buildace(struct nfsrv_descript *, u_char *, int,
enum vtype, int, int, struct acl_entry *);
/*
* This function builds an NFS ace.
*/
static int
nfsrv_buildace(struct nfsrv_descript *nd, u_char *name, int namelen,
enum vtype type, int group, int owner, struct acl_entry *ace)
{
u_int32_t *tl, aceflag = 0x0, acemask = 0x0, acetype;
int full_len;
full_len = NFSM_RNDUP(namelen);
NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED + full_len);
/*
* Fill in the ace type.
*/
if (ace->ae_entry_type & ACL_ENTRY_TYPE_ALLOW)
acetype = NFSV4ACE_ALLOWEDTYPE;
else if (ace->ae_entry_type & ACL_ENTRY_TYPE_DENY)
acetype = NFSV4ACE_DENIEDTYPE;
else if (ace->ae_entry_type & ACL_ENTRY_TYPE_AUDIT)
acetype = NFSV4ACE_AUDITTYPE;
else
acetype = NFSV4ACE_ALARMTYPE;
*tl++ = txdr_unsigned(acetype);
/*
* Set the flag bits from the ACL.
*/
if (ace->ae_flags & ACL_ENTRY_FILE_INHERIT)
aceflag |= NFSV4ACE_FILEINHERIT;
if (ace->ae_flags & ACL_ENTRY_DIRECTORY_INHERIT)
aceflag |= NFSV4ACE_DIRECTORYINHERIT;
if (ace->ae_flags & ACL_ENTRY_NO_PROPAGATE_INHERIT)
aceflag |= NFSV4ACE_NOPROPAGATEINHERIT;
if (ace->ae_flags & ACL_ENTRY_INHERIT_ONLY)
aceflag |= NFSV4ACE_INHERITONLY;
if (ace->ae_flags & ACL_ENTRY_SUCCESSFUL_ACCESS)
aceflag |= NFSV4ACE_SUCCESSFULACCESS;
if (ace->ae_flags & ACL_ENTRY_FAILED_ACCESS)
aceflag |= NFSV4ACE_FAILEDACCESS;
if (group)
aceflag |= NFSV4ACE_IDENTIFIERGROUP;
*tl++ = txdr_unsigned(aceflag);
if (type == VDIR) {
if (ace->ae_perm & ACL_LIST_DIRECTORY)
acemask |= NFSV4ACE_LISTDIRECTORY;
if (ace->ae_perm & ACL_ADD_FILE)
acemask |= NFSV4ACE_ADDFILE;
if (ace->ae_perm & ACL_ADD_SUBDIRECTORY)
acemask |= NFSV4ACE_ADDSUBDIRECTORY;
if (ace->ae_perm & ACL_READ_NAMED_ATTRS)
acemask |= NFSV4ACE_READNAMEDATTR;
if (ace->ae_perm & ACL_WRITE_NAMED_ATTRS)
acemask |= NFSV4ACE_WRITENAMEDATTR;
if (ace->ae_perm & ACL_EXECUTE)
acemask |= NFSV4ACE_SEARCH;
if (ace->ae_perm & ACL_DELETE_CHILD)
acemask |= NFSV4ACE_DELETECHILD;
if (ace->ae_perm & ACL_READ_ATTRIBUTES)
acemask |= NFSV4ACE_READATTRIBUTES;
if (ace->ae_perm & ACL_WRITE_ATTRIBUTES)
acemask |= NFSV4ACE_WRITEATTRIBUTES;
if (ace->ae_perm & ACL_DELETE)
acemask |= NFSV4ACE_DELETE;
if (ace->ae_perm & ACL_READ_ACL)
acemask |= NFSV4ACE_READACL;
if (ace->ae_perm & ACL_WRITE_ACL)
acemask |= NFSV4ACE_WRITEACL;
if (ace->ae_perm & ACL_WRITE_OWNER)
acemask |= NFSV4ACE_WRITEOWNER;
if (ace->ae_perm & ACL_SYNCHRONIZE)
acemask |= NFSV4ACE_SYNCHRONIZE;
} else {
if (ace->ae_perm & ACL_READ_DATA)
acemask |= NFSV4ACE_READDATA;
if (ace->ae_perm & ACL_WRITE_DATA)
acemask |= NFSV4ACE_WRITEDATA;
if (ace->ae_perm & ACL_APPEND_DATA)
acemask |= NFSV4ACE_APPENDDATA;
if (ace->ae_perm & ACL_READ_NAMED_ATTRS)
acemask |= NFSV4ACE_READNAMEDATTR;
if (ace->ae_perm & ACL_WRITE_NAMED_ATTRS)
acemask |= NFSV4ACE_WRITENAMEDATTR;
if (ace->ae_perm & ACL_EXECUTE)
acemask |= NFSV4ACE_EXECUTE;
if (ace->ae_perm & ACL_READ_ATTRIBUTES)
acemask |= NFSV4ACE_READATTRIBUTES;
if (ace->ae_perm & ACL_WRITE_ATTRIBUTES)
acemask |= NFSV4ACE_WRITEATTRIBUTES;
if (ace->ae_perm & ACL_DELETE)
acemask |= NFSV4ACE_DELETE;
if (ace->ae_perm & ACL_READ_ACL)
acemask |= NFSV4ACE_READACL;
if (ace->ae_perm & ACL_WRITE_ACL)
acemask |= NFSV4ACE_WRITEACL;
if (ace->ae_perm & ACL_WRITE_OWNER)
acemask |= NFSV4ACE_WRITEOWNER;
if (ace->ae_perm & ACL_SYNCHRONIZE)
acemask |= NFSV4ACE_SYNCHRONIZE;
}
*tl++ = txdr_unsigned(acemask);
*tl++ = txdr_unsigned(namelen);
if (full_len - namelen)
*(tl + (namelen / NFSX_UNSIGNED)) = 0x0;
NFSBCOPY(name, (caddr_t)tl, namelen);
return (full_len + 4 * NFSX_UNSIGNED);
}
/*
* Build an NFSv4 ACL.
*/
int
nfsrv_buildacl(struct nfsrv_descript *nd, NFSACL_T *aclp, enum vtype type,
NFSPROC_T *p)
{
int i, entrycnt = 0, retlen;
u_int32_t *entrycntp;
int isowner, isgroup, namelen, malloced;
u_char *name, namestr[NFSV4_SMALLSTR];
NFSM_BUILD(entrycntp, u_int32_t *, NFSX_UNSIGNED);
retlen = NFSX_UNSIGNED;
/*
* Loop through the acl entries, building each one.
*/
for (i = 0; i < aclp->acl_cnt; i++) {
isowner = isgroup = malloced = 0;
switch (aclp->acl_entry[i].ae_tag) {
case ACL_USER_OBJ:
isowner = 1;
name = "OWNER@";
namelen = 6;
break;
case ACL_GROUP_OBJ:
isgroup = 1;
name = "GROUP@";
namelen = 6;
break;
case ACL_EVERYONE:
name = "EVERYONE@";
namelen = 9;
break;
case ACL_USER:
name = namestr;
nfsv4_uidtostr(aclp->acl_entry[i].ae_id, &name,
&namelen);
if (name != namestr)
malloced = 1;
break;
case ACL_GROUP:
isgroup = 1;
name = namestr;
nfsv4_gidtostr((gid_t)aclp->acl_entry[i].ae_id, &name,
&namelen);
if (name != namestr)
malloced = 1;
break;
default:
continue;
}
retlen += nfsrv_buildace(nd, name, namelen, type, isgroup,
isowner, &aclp->acl_entry[i]);
entrycnt++;
if (malloced)
free(name, M_NFSSTRING);
}
*entrycntp = txdr_unsigned(entrycnt);
return (retlen);
}
/*
* Compare two NFSv4 acls.
* Return 0 if they are the same, 1 if not the same.
*/
int
nfsrv_compareacl(NFSACL_T *aclp1, NFSACL_T *aclp2)
{
int i;
struct acl_entry *acep1, *acep2;
if (aclp1->acl_cnt != aclp2->acl_cnt)
return (1);
acep1 = aclp1->acl_entry;
acep2 = aclp2->acl_entry;
for (i = 0; i < aclp1->acl_cnt; i++) {
if (acep1->ae_tag != acep2->ae_tag)
return (1);
switch (acep1->ae_tag) {
case ACL_GROUP:
case ACL_USER:
if (acep1->ae_id != acep2->ae_id)
return (1);
/* fall through */
case ACL_USER_OBJ:
case ACL_GROUP_OBJ:
case ACL_OTHER:
if (acep1->ae_perm != acep2->ae_perm)
return (1);
}
acep1++;
acep2++;
}
return (0);
}
diff --git a/sys/fs/nfs/nfs_commonsubs.c b/sys/fs/nfs/nfs_commonsubs.c
index 8bfe031f01ab..4be1c415bf67 100644
--- a/sys/fs/nfs/nfs_commonsubs.c
+++ b/sys/fs/nfs/nfs_commonsubs.c
@@ -1,4986 +1,4986 @@
/*-
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 1989, 1993
* The Regents of the University of California. All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* Rick Macklem at The University of Guelph.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
/*
* These functions support the macros and help fiddle mbuf chains for
* the nfs op functions. They do things like create the rpc header and
* copy data between mbuf chains and uio lists.
*/
#include "opt_inet.h"
#include "opt_inet6.h"
#include <fs/nfs/nfsport.h>
#include <sys/extattr.h>
#include <security/mac/mac_framework.h>
#include <vm/vm_param.h>
/*
* Data items converted to xdr at startup, since they are constant
* This is kinda hokey, but may save a little time doing byte swaps
*/
u_int32_t newnfs_true, newnfs_false, newnfs_xdrneg1;
/* And other global data */
nfstype nfsv34_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK,
NFFIFO, NFNON };
enum vtype newnv2tov_type[8] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON };
enum vtype nv34tov_type[8]={ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO };
struct timeval nfsboottime; /* Copy boottime once, so it never changes */
int nfscl_ticks;
int nfsrv_useacl = 1;
struct nfssockreq nfsrv_nfsuserdsock;
nfsuserd_state nfsrv_nfsuserd = NOTRUNNING;
static int nfsrv_userdupcalls = 0;
struct nfsreqhead nfsd_reqq;
uid_t nfsrv_defaultuid = UID_NOBODY;
gid_t nfsrv_defaultgid = GID_NOGROUP;
int nfsrv_lease = NFSRV_LEASE;
int ncl_mbuf_mlen = MLEN;
int nfsd_enable_stringtouid = 0;
int nfsrv_doflexfile = 0;
static int nfs_enable_uidtostring = 0;
NFSNAMEIDMUTEX;
NFSSOCKMUTEX;
extern int nfsrv_lughashsize;
extern struct mtx nfsrv_dslock_mtx;
extern volatile int nfsrv_devidcnt;
extern int nfscl_debuglevel;
extern struct nfsdevicehead nfsrv_devidhead;
extern struct nfsstatsv1 nfsstatsv1;
extern uint32_t nfs_srvmaxio;
SYSCTL_DECL(_vfs_nfs);
SYSCTL_INT(_vfs_nfs, OID_AUTO, enable_uidtostring, CTLFLAG_RW,
&nfs_enable_uidtostring, 0, "Make nfs always send numeric owner_names");
int nfsrv_maxpnfsmirror = 1;
SYSCTL_INT(_vfs_nfs, OID_AUTO, pnfsmirror, CTLFLAG_RD,
&nfsrv_maxpnfsmirror, 0, "Mirror level for pNFS service");
/*
* This array of structures indicates, for V4:
* retfh - which of 3 types of calling args are used
* 0 - doesn't change cfh or use a sfh
* 1 - replaces cfh with a new one (unless it returns an error status)
* 2 - uses cfh and sfh
* needscfh - if the op wants a cfh and premtime
* 0 - doesn't use a cfh
* 1 - uses a cfh, but doesn't want pre-op attributes
* 2 - uses a cfh and wants pre-op attributes
* savereply - indicates a non-idempotent Op
* 0 - not non-idempotent
* 1 - non-idempotent
* Ops that are ordered via seqid# are handled separately from these
* non-idempotent Ops.
* Define it here, since it is used by both the client and server.
*/
struct nfsv4_opflag nfsv4_opflag[NFSV42_NOPS] = {
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* undef */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* undef */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* undef */
{ 0, 1, 0, 0, LK_SHARED, 1, 1 }, /* Access */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* Close */
{ 0, 2, 0, 1, LK_EXCLUSIVE, 1, 1 }, /* Commit */
{ 1, 2, 1, 1, LK_EXCLUSIVE, 1, 1 }, /* Create */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* Delegpurge */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* Delegreturn */
{ 0, 1, 0, 0, LK_SHARED, 1, 1 }, /* Getattr */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* GetFH */
{ 2, 1, 1, 1, LK_EXCLUSIVE, 1, 1 }, /* Link */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* Lock */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* LockT */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* LockU */
{ 1, 2, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Lookup */
{ 1, 2, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Lookupp */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* NVerify */
{ 1, 1, 0, 1, LK_EXCLUSIVE, 1, 0 }, /* Open */
{ 1, 1, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* OpenAttr */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* OpenConfirm */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* OpenDowngrade */
{ 1, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* PutFH */
{ 1, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* PutPubFH */
{ 1, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* PutRootFH */
{ 0, 1, 0, 0, LK_SHARED, 1, 0 }, /* Read */
{ 0, 1, 0, 0, LK_SHARED, 1, 1 }, /* Readdir */
{ 0, 1, 0, 0, LK_SHARED, 1, 1 }, /* ReadLink */
{ 0, 2, 1, 1, LK_EXCLUSIVE, 1, 1 }, /* Remove */
{ 2, 1, 1, 1, LK_EXCLUSIVE, 1, 1 }, /* Rename */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* Renew */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* RestoreFH */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* SaveFH */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* SecInfo */
{ 0, 2, 1, 1, LK_EXCLUSIVE, 1, 0 }, /* Setattr */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* SetClientID */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* SetClientIDConfirm */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Verify */
{ 0, 2, 1, 1, LK_EXCLUSIVE, 1, 0 }, /* Write */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* ReleaseLockOwner */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Backchannel Ctrl */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 0, 0 }, /* Bind Conn to Sess */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 0, 0 }, /* Exchange ID */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 0, 0 }, /* Create Session */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 0, 0 }, /* Destroy Session */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* Free StateID */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Get Dir Deleg */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Get Device Info */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Get Device List */
{ 0, 1, 0, 1, LK_EXCLUSIVE, 1, 1 }, /* Layout Commit */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Layout Get */
{ 0, 1, 0, 1, LK_EXCLUSIVE, 1, 0 }, /* Layout Return */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Secinfo No name */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* Sequence */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Set SSV */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Test StateID */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Want Delegation */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 0, 0 }, /* Destroy ClientID */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* Reclaim Complete */
{ 0, 1, 1, 1, LK_EXCLUSIVE, 1, 0 }, /* Allocate */
{ 2, 1, 1, 0, LK_SHARED, 1, 0 }, /* Copy */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Copy Notify */
{ 0, 2, 1, 1, LK_EXCLUSIVE, 1, 0 }, /* Deallocate */
{ 0, 1, 0, 0, LK_SHARED, 1, 0 }, /* IO Advise */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* Layout Error */
{ 0, 1, 0, 0, LK_EXCLUSIVE, 1, 0 }, /* Layout Stats */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Offload Cancel */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Offload Status */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Read Plus */
{ 0, 1, 0, 0, LK_SHARED, 1, 0 }, /* Seek */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Write Same */
{ 0, 0, 0, 0, LK_EXCLUSIVE, 1, 1 }, /* Clone */
{ 0, 1, 0, 0, LK_SHARED, 1, 1 }, /* Getxattr */
{ 0, 1, 1, 1, LK_EXCLUSIVE, 1, 1 }, /* Setxattr */
{ 0, 1, 0, 0, LK_SHARED, 1, 1 }, /* Listxattrs */
{ 0, 1, 1, 1, LK_EXCLUSIVE, 1, 1 }, /* Removexattr */
};
static int ncl_mbuf_mhlen = MHLEN;
static int nfsrv_usercnt = 0;
static int nfsrv_dnsnamelen;
static u_char *nfsrv_dnsname = NULL;
static int nfsrv_usermax = 999999999;
struct nfsrv_lughash {
struct mtx mtx;
struct nfsuserhashhead lughead;
};
static struct nfsrv_lughash *nfsuserhash;
static struct nfsrv_lughash *nfsusernamehash;
static struct nfsrv_lughash *nfsgrouphash;
static struct nfsrv_lughash *nfsgroupnamehash;
/*
* This static array indicates whether or not the RPC generates a large
* reply. This is used by nfs_reply() to decide whether or not an mbuf
* cluster should be allocated. (If a cluster is required by an RPC
* marked 0 in this array, the code will still work, just not quite as
* efficiently.)
*/
static int nfs_bigreply[NFSV42_NPROCS] = { 0, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0,
0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0,
1, 0, 0, 1, 0, 0, 0, 0 };
/* local functions */
static int nfsrv_skipace(struct nfsrv_descript *nd, int *acesizep);
static void nfsv4_wanted(struct nfsv4lock *lp);
static uint32_t nfsv4_filesavail(struct statfs *, struct mount *);
static int nfsrv_cmpmixedcase(u_char *cp, u_char *cp2, int len);
static int nfsrv_getuser(int procnum, uid_t uid, gid_t gid, char *name);
static void nfsrv_removeuser(struct nfsusrgrp *usrp, int isuser);
static int nfsrv_getrefstr(struct nfsrv_descript *, u_char **, u_char **,
int *, int *);
static void nfsrv_refstrbigenough(int, u_char **, u_char **, int *);
static struct {
int op;
int opcnt;
const u_char *tag;
int taglen;
} nfsv4_opmap[NFSV42_NPROCS] = {
{ 0, 1, "Null", 4 },
{ NFSV4OP_GETATTR, 1, "Getattr", 7, },
{ NFSV4OP_SETATTR, 2, "Setattr", 7, },
{ NFSV4OP_LOOKUP, 3, "Lookup", 6, },
{ NFSV4OP_ACCESS, 2, "Access", 6, },
{ NFSV4OP_READLINK, 2, "Readlink", 8, },
{ NFSV4OP_READ, 1, "Read", 4, },
{ NFSV4OP_WRITE, 2, "Write", 5, },
{ NFSV4OP_OPEN, 5, "Open", 4, },
{ NFSV4OP_CREATE, 5, "Create", 6, },
{ NFSV4OP_CREATE, 1, "Create", 6, },
{ NFSV4OP_CREATE, 3, "Create", 6, },
{ NFSV4OP_REMOVE, 1, "Remove", 6, },
{ NFSV4OP_REMOVE, 1, "Remove", 6, },
{ NFSV4OP_SAVEFH, 5, "Rename", 6, },
{ NFSV4OP_SAVEFH, 4, "Link", 4, },
{ NFSV4OP_READDIR, 2, "Readdir", 7, },
{ NFSV4OP_READDIR, 2, "Readdir", 7, },
{ NFSV4OP_GETATTR, 1, "Getattr", 7, },
{ NFSV4OP_GETATTR, 1, "Getattr", 7, },
{ NFSV4OP_GETATTR, 1, "Getattr", 7, },
{ NFSV4OP_COMMIT, 2, "Commit", 6, },
{ NFSV4OP_LOOKUPP, 3, "Lookupp", 7, },
{ NFSV4OP_SETCLIENTID, 1, "SetClientID", 11, },
{ NFSV4OP_SETCLIENTIDCFRM, 1, "SetClientIDConfirm", 18, },
{ NFSV4OP_LOCK, 1, "Lock", 4, },
{ NFSV4OP_LOCKU, 1, "LockU", 5, },
{ NFSV4OP_OPEN, 2, "Open", 4, },
{ NFSV4OP_CLOSE, 1, "Close", 5, },
{ NFSV4OP_OPENCONFIRM, 1, "Openconfirm", 11, },
{ NFSV4OP_LOCKT, 1, "LockT", 5, },
{ NFSV4OP_OPENDOWNGRADE, 1, "Opendowngrade", 13, },
{ NFSV4OP_RENEW, 1, "Renew", 5, },
{ NFSV4OP_PUTROOTFH, 1, "Dirpath", 7, },
{ NFSV4OP_RELEASELCKOWN, 1, "Rellckown", 9, },
{ NFSV4OP_DELEGRETURN, 1, "Delegret", 8, },
{ NFSV4OP_DELEGRETURN, 3, "DelegRemove", 11, },
{ NFSV4OP_DELEGRETURN, 7, "DelegRename1", 12, },
{ NFSV4OP_DELEGRETURN, 9, "DelegRename2", 12, },
{ NFSV4OP_GETATTR, 1, "Getacl", 6, },
{ NFSV4OP_SETATTR, 1, "Setacl", 6, },
{ NFSV4OP_EXCHANGEID, 1, "ExchangeID", 10, },
{ NFSV4OP_CREATESESSION, 1, "CreateSession", 13, },
{ NFSV4OP_DESTROYSESSION, 1, "DestroySession", 14, },
{ NFSV4OP_DESTROYCLIENTID, 1, "DestroyClient", 13, },
{ NFSV4OP_FREESTATEID, 1, "FreeStateID", 11, },
{ NFSV4OP_LAYOUTGET, 1, "LayoutGet", 9, },
{ NFSV4OP_GETDEVINFO, 1, "GetDeviceInfo", 13, },
{ NFSV4OP_LAYOUTCOMMIT, 1, "LayoutCommit", 12, },
{ NFSV4OP_LAYOUTRETURN, 1, "LayoutReturn", 12, },
{ NFSV4OP_RECLAIMCOMPL, 1, "ReclaimComplete", 15, },
{ NFSV4OP_WRITE, 1, "WriteDS", 7, },
{ NFSV4OP_READ, 1, "ReadDS", 6, },
{ NFSV4OP_COMMIT, 1, "CommitDS", 8, },
{ NFSV4OP_OPEN, 3, "OpenLayoutGet", 13, },
{ NFSV4OP_OPEN, 8, "CreateLayGet", 12, },
{ NFSV4OP_IOADVISE, 1, "Advise", 6, },
{ NFSV4OP_ALLOCATE, 2, "Allocate", 8, },
{ NFSV4OP_SAVEFH, 5, "Copy", 4, },
{ NFSV4OP_SEEK, 2, "Seek", 4, },
{ NFSV4OP_SEEK, 1, "SeekDS", 6, },
{ NFSV4OP_GETXATTR, 2, "Getxattr", 8, },
{ NFSV4OP_SETXATTR, 2, "Setxattr", 8, },
{ NFSV4OP_REMOVEXATTR, 2, "Rmxattr", 7, },
{ NFSV4OP_LISTXATTRS, 2, "Listxattr", 9, },
{ NFSV4OP_BINDCONNTOSESS, 1, "BindConSess", 11, },
{ NFSV4OP_LOOKUP, 5, "LookupOpen", 10, },
{ NFSV4OP_DEALLOCATE, 2, "Deallocate", 10, },
{ NFSV4OP_LAYOUTERROR, 1, "LayoutError", 11, },
};
/*
* NFS RPCS that have large request message size.
*/
static int nfs_bigrequest[NFSV42_NPROCS] = {
0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0,
0
};
/*
* Start building a request. Mostly just put the first file handle in
* place.
*/
void
nfscl_reqstart(struct nfsrv_descript *nd, int procnum, struct nfsmount *nmp,
u_int8_t *nfhp, int fhlen, u_int32_t **opcntpp, struct nfsclsession *sep,
int vers, int minorvers)
{
struct mbuf *mb;
u_int32_t *tl;
int opcnt;
nfsattrbit_t attrbits;
/*
* First, fill in some of the fields of nd.
*/
nd->nd_slotseq = NULL;
if (vers == NFS_VER4) {
nd->nd_flag = ND_NFSV4 | ND_NFSCL;
if (minorvers == NFSV41_MINORVERSION)
nd->nd_flag |= ND_NFSV41;
else if (minorvers == NFSV42_MINORVERSION)
nd->nd_flag |= (ND_NFSV41 | ND_NFSV42);
} else if (vers == NFS_VER3)
nd->nd_flag = ND_NFSV3 | ND_NFSCL;
else {
if (NFSHASNFSV4(nmp)) {
nd->nd_flag = ND_NFSV4 | ND_NFSCL;
if (nmp->nm_minorvers == 1)
nd->nd_flag |= ND_NFSV41;
else if (nmp->nm_minorvers == 2)
nd->nd_flag |= (ND_NFSV41 | ND_NFSV42);
} else if (NFSHASNFSV3(nmp))
nd->nd_flag = ND_NFSV3 | ND_NFSCL;
else
nd->nd_flag = ND_NFSV2 | ND_NFSCL;
}
nd->nd_procnum = procnum;
nd->nd_repstat = 0;
nd->nd_maxextsiz = 0;
/*
* Get the first mbuf for the request.
*/
if (nfs_bigrequest[procnum])
NFSMCLGET(mb, M_WAITOK);
else
NFSMGET(mb);
mb->m_len = 0;
nd->nd_mreq = nd->nd_mb = mb;
nd->nd_bpos = mtod(mb, char *);
/*
* And fill the first file handle into the request.
*/
if (nd->nd_flag & ND_NFSV4) {
opcnt = nfsv4_opmap[procnum].opcnt +
nfsv4_opflag[nfsv4_opmap[procnum].op].needscfh;
if ((nd->nd_flag & ND_NFSV41) != 0) {
opcnt += nfsv4_opflag[nfsv4_opmap[procnum].op].needsseq;
if (procnum == NFSPROC_RENEW)
/*
* For the special case of Renew, just do a
* Sequence Op.
*/
opcnt = 1;
else if (procnum == NFSPROC_WRITEDS ||
procnum == NFSPROC_COMMITDS)
/*
* For the special case of a Writeor Commit to
* a DS, the opcnt == 3, for Sequence, PutFH,
* Write/Commit.
*/
opcnt = 3;
}
/*
* What should the tag really be?
*/
(void) nfsm_strtom(nd, nfsv4_opmap[procnum].tag,
nfsv4_opmap[procnum].taglen);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
if ((nd->nd_flag & ND_NFSV42) != 0)
*tl++ = txdr_unsigned(NFSV42_MINORVERSION);
else if ((nd->nd_flag & ND_NFSV41) != 0)
*tl++ = txdr_unsigned(NFSV41_MINORVERSION);
else
*tl++ = txdr_unsigned(NFSV4_MINORVERSION);
if (opcntpp != NULL)
*opcntpp = tl;
*tl = txdr_unsigned(opcnt);
if ((nd->nd_flag & ND_NFSV41) != 0 &&
nfsv4_opflag[nfsv4_opmap[procnum].op].needsseq > 0) {
if (nfsv4_opflag[nfsv4_opmap[procnum].op].loopbadsess >
0)
nd->nd_flag |= ND_LOOPBADSESS;
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_SEQUENCE);
if (sep == NULL) {
sep = nfsmnt_mdssession(nmp);
nfsv4_setsequence(nmp, nd, sep,
nfs_bigreply[procnum]);
} else
nfsv4_setsequence(nmp, nd, sep,
nfs_bigreply[procnum]);
}
if (nfsv4_opflag[nfsv4_opmap[procnum].op].needscfh > 0) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_PUTFH);
(void) nfsm_fhtom(nd, nfhp, fhlen, 0);
if (nfsv4_opflag[nfsv4_opmap[procnum].op].needscfh
== 2 && procnum != NFSPROC_WRITEDS &&
procnum != NFSPROC_COMMITDS) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
/*
* For Lookup Ops, we want all the directory
* attributes, so we can load the name cache.
*/
if (procnum == NFSPROC_LOOKUP ||
procnum == NFSPROC_LOOKUPP ||
procnum == NFSPROC_LOOKUPOPEN)
NFSGETATTR_ATTRBIT(&attrbits);
else {
NFSWCCATTR_ATTRBIT(&attrbits);
nd->nd_flag |= ND_V4WCCATTR;
}
(void) nfsrv_putattrbit(nd, &attrbits);
}
}
if (procnum != NFSPROC_RENEW ||
(nd->nd_flag & ND_NFSV41) == 0) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(nfsv4_opmap[procnum].op);
}
} else {
(void) nfsm_fhtom(nd, nfhp, fhlen, 0);
}
if (procnum < NFSV42_NPROCS)
NFSINCRGLOBAL(nfsstatsv1.rpccnt[procnum]);
}
/*
* Put a state Id in the mbuf list.
*/
void
nfsm_stateidtom(struct nfsrv_descript *nd, nfsv4stateid_t *stateidp, int flag)
{
nfsv4stateid_t *st;
NFSM_BUILD(st, nfsv4stateid_t *, NFSX_STATEID);
if (flag == NFSSTATEID_PUTALLZERO) {
st->seqid = 0;
st->other[0] = 0;
st->other[1] = 0;
st->other[2] = 0;
} else if (flag == NFSSTATEID_PUTALLONE) {
st->seqid = 0xffffffff;
st->other[0] = 0xffffffff;
st->other[1] = 0xffffffff;
st->other[2] = 0xffffffff;
} else if (flag == NFSSTATEID_PUTSEQIDZERO) {
st->seqid = 0;
st->other[0] = stateidp->other[0];
st->other[1] = stateidp->other[1];
st->other[2] = stateidp->other[2];
} else {
st->seqid = stateidp->seqid;
st->other[0] = stateidp->other[0];
st->other[1] = stateidp->other[1];
st->other[2] = stateidp->other[2];
}
}
/*
* Fill in the setable attributes. The full argument indicates whether
* to fill in them all or just mode and time.
*/
void
nfscl_fillsattr(struct nfsrv_descript *nd, struct vattr *vap,
struct vnode *vp, int flags, u_int32_t rdev)
{
u_int32_t *tl;
struct nfsv2_sattr *sp;
nfsattrbit_t attrbits;
struct nfsnode *np;
switch (nd->nd_flag & (ND_NFSV2 | ND_NFSV3 | ND_NFSV4)) {
case ND_NFSV2:
NFSM_BUILD(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
if (vap->va_mode == (mode_t)VNOVAL)
sp->sa_mode = newnfs_xdrneg1;
else
sp->sa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
if (vap->va_uid == (uid_t)VNOVAL)
sp->sa_uid = newnfs_xdrneg1;
else
sp->sa_uid = txdr_unsigned(vap->va_uid);
if (vap->va_gid == (gid_t)VNOVAL)
sp->sa_gid = newnfs_xdrneg1;
else
sp->sa_gid = txdr_unsigned(vap->va_gid);
if (flags & NFSSATTR_SIZE0)
sp->sa_size = 0;
else if (flags & NFSSATTR_SIZENEG1)
sp->sa_size = newnfs_xdrneg1;
else if (flags & NFSSATTR_SIZERDEV)
sp->sa_size = txdr_unsigned(rdev);
else
sp->sa_size = txdr_unsigned(vap->va_size);
txdr_nfsv2time(&vap->va_atime, &sp->sa_atime);
txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime);
break;
case ND_NFSV3:
if (vap->va_mode != (mode_t)VNOVAL) {
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = newnfs_true;
*tl = txdr_unsigned(vap->va_mode);
} else {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = newnfs_false;
}
if ((flags & NFSSATTR_FULL) && vap->va_uid != (uid_t)VNOVAL) {
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = newnfs_true;
*tl = txdr_unsigned(vap->va_uid);
} else {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = newnfs_false;
}
if ((flags & NFSSATTR_FULL) && vap->va_gid != (gid_t)VNOVAL) {
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = newnfs_true;
*tl = txdr_unsigned(vap->va_gid);
} else {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = newnfs_false;
}
if ((flags & NFSSATTR_FULL) && vap->va_size != VNOVAL) {
NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
*tl++ = newnfs_true;
txdr_hyper(vap->va_size, tl);
} else {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = newnfs_false;
}
if (vap->va_atime.tv_sec != VNOVAL) {
if ((vap->va_vaflags & VA_UTIMES_NULL) == 0) {
NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT);
txdr_nfsv3time(&vap->va_atime, tl);
} else {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER);
}
} else {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE);
}
if (vap->va_mtime.tv_sec != VNOVAL) {
if ((vap->va_vaflags & VA_UTIMES_NULL) == 0) {
NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT);
txdr_nfsv3time(&vap->va_mtime, tl);
} else {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER);
}
} else {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE);
}
break;
case ND_NFSV4:
NFSZERO_ATTRBIT(&attrbits);
if (vap->va_mode != (mode_t)VNOVAL)
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_MODE);
if ((flags & NFSSATTR_FULL) && vap->va_uid != (uid_t)VNOVAL)
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_OWNER);
if ((flags & NFSSATTR_FULL) && vap->va_gid != (gid_t)VNOVAL)
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_OWNERGROUP);
if ((flags & NFSSATTR_FULL) && vap->va_size != VNOVAL)
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SIZE);
if (vap->va_atime.tv_sec != VNOVAL)
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEACCESSSET);
if (vap->va_mtime.tv_sec != VNOVAL)
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFYSET);
if (vap->va_birthtime.tv_sec != VNOVAL &&
strcmp(vp->v_mount->mnt_vfc->vfc_name, "nfs") == 0) {
/*
* We can only test for support of TimeCreate if
* the "vp" argument is for an NFS vnode.
*/
np = VTONFS(vp);
if (NFSISSET_ATTRBIT(&np->n_vattr.na_suppattr,
NFSATTRBIT_TIMECREATE))
NFSSETBIT_ATTRBIT(&attrbits,
NFSATTRBIT_TIMECREATE);
}
(void) nfsv4_fillattr(nd, vp->v_mount, vp, NULL, vap, NULL, 0,
&attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
break;
}
}
#ifndef APPLE
/*
* copies mbuf chain to the uio scatter/gather list
*/
int
nfsm_mbufuio(struct nfsrv_descript *nd, struct uio *uiop, int siz)
{
char *mbufcp, *uiocp;
int xfer, left, len;
struct mbuf *mp;
long uiosiz, rem;
int error = 0;
mp = nd->nd_md;
mbufcp = nd->nd_dpos;
len = mtod(mp, caddr_t) + mp->m_len - mbufcp;
rem = NFSM_RNDUP(siz) - siz;
while (siz > 0) {
if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL) {
error = EBADRPC;
goto out;
}
left = uiop->uio_iov->iov_len;
uiocp = uiop->uio_iov->iov_base;
if (left > siz)
left = siz;
uiosiz = left;
while (left > 0) {
while (len == 0) {
mp = mp->m_next;
if (mp == NULL) {
error = EBADRPC;
goto out;
}
mbufcp = mtod(mp, caddr_t);
len = mp->m_len;
KASSERT(len >= 0,
("len %d, corrupted mbuf?", len));
}
xfer = (left > len) ? len : left;
#ifdef notdef
/* Not Yet.. */
if (uiop->uio_iov->iov_op != NULL)
(*(uiop->uio_iov->iov_op))
(mbufcp, uiocp, xfer);
else
#endif
if (uiop->uio_segflg == UIO_SYSSPACE)
NFSBCOPY(mbufcp, uiocp, xfer);
else
copyout(mbufcp, uiocp, xfer);
left -= xfer;
len -= xfer;
mbufcp += xfer;
uiocp += xfer;
uiop->uio_offset += xfer;
uiop->uio_resid -= xfer;
}
if (uiop->uio_iov->iov_len <= siz) {
uiop->uio_iovcnt--;
uiop->uio_iov++;
} else {
uiop->uio_iov->iov_base = (void *)
((char *)uiop->uio_iov->iov_base + uiosiz);
uiop->uio_iov->iov_len -= uiosiz;
}
siz -= uiosiz;
}
nd->nd_dpos = mbufcp;
nd->nd_md = mp;
if (rem > 0) {
if (len < rem)
error = nfsm_advance(nd, rem, len);
else
nd->nd_dpos += rem;
}
out:
NFSEXITCODE2(error, nd);
return (error);
}
#endif /* !APPLE */
/*
* Help break down an mbuf chain by setting the first siz bytes contiguous
* pointed to by returned val.
* This is used by the macro NFSM_DISSECT for tough
* cases.
*/
void *
nfsm_dissct(struct nfsrv_descript *nd, int siz, int how)
{
struct mbuf *mp2;
int siz2, xfer;
caddr_t p;
int left;
caddr_t retp;
retp = NULL;
left = mtod(nd->nd_md, caddr_t) + nd->nd_md->m_len - nd->nd_dpos;
while (left == 0) {
nd->nd_md = nd->nd_md->m_next;
if (nd->nd_md == NULL)
return (retp);
left = nd->nd_md->m_len;
nd->nd_dpos = mtod(nd->nd_md, caddr_t);
}
if (left >= siz) {
retp = nd->nd_dpos;
nd->nd_dpos += siz;
} else if (nd->nd_md->m_next == NULL) {
return (retp);
} else if (siz > ncl_mbuf_mhlen) {
panic("nfs S too big");
} else {
MGET(mp2, MT_DATA, how);
if (mp2 == NULL)
return (NULL);
mp2->m_next = nd->nd_md->m_next;
nd->nd_md->m_next = mp2;
nd->nd_md->m_len -= left;
nd->nd_md = mp2;
retp = p = mtod(mp2, caddr_t);
NFSBCOPY(nd->nd_dpos, p, left); /* Copy what was left */
siz2 = siz - left;
p += left;
mp2 = mp2->m_next;
/* Loop around copying up the siz2 bytes */
while (siz2 > 0) {
if (mp2 == NULL)
return (NULL);
xfer = (siz2 > mp2->m_len) ? mp2->m_len : siz2;
if (xfer > 0) {
NFSBCOPY(mtod(mp2, caddr_t), p, xfer);
mp2->m_data += xfer;
mp2->m_len -= xfer;
p += xfer;
siz2 -= xfer;
}
if (siz2 > 0)
mp2 = mp2->m_next;
}
nd->nd_md->m_len = siz;
nd->nd_md = mp2;
nd->nd_dpos = mtod(mp2, caddr_t);
}
return (retp);
}
/*
* Advance the position in the mbuf chain.
* If offs == 0, this is a no-op, but it is simpler to just return from
* here than check for offs > 0 for all calls to nfsm_advance.
* If left == -1, it should be calculated here.
*/
int
nfsm_advance(struct nfsrv_descript *nd, int offs, int left)
{
int error = 0;
if (offs == 0)
goto out;
/*
* A negative offs might indicate a corrupted mbuf chain and,
* as such, a printf is logged.
*/
if (offs < 0) {
printf("nfsrv_advance: negative offs\n");
error = EBADRPC;
goto out;
}
/*
* If left == -1, calculate it here.
*/
if (left == -1)
left = mtod(nd->nd_md, caddr_t) + nd->nd_md->m_len -
nd->nd_dpos;
/*
* Loop around, advancing over the mbuf data.
*/
while (offs > left) {
offs -= left;
nd->nd_md = nd->nd_md->m_next;
if (nd->nd_md == NULL) {
error = EBADRPC;
goto out;
}
left = nd->nd_md->m_len;
nd->nd_dpos = mtod(nd->nd_md, caddr_t);
}
nd->nd_dpos += offs;
out:
NFSEXITCODE(error);
return (error);
}
/*
* Copy a string into mbuf(s).
* Return the number of bytes output, including XDR overheads.
*/
int
nfsm_strtom(struct nfsrv_descript *nd, const char *cp, int siz)
{
struct mbuf *m2;
int xfer, left;
struct mbuf *m1;
int rem, bytesize;
u_int32_t *tl;
char *cp2;
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(siz);
rem = NFSM_RNDUP(siz) - siz;
bytesize = NFSX_UNSIGNED + siz + rem;
m2 = nd->nd_mb;
cp2 = nd->nd_bpos;
if ((nd->nd_flag & ND_EXTPG) != 0)
left = nd->nd_bextpgsiz;
else
left = M_TRAILINGSPACE(m2);
KASSERT(((m2->m_flags & (M_EXT | M_EXTPG)) ==
(M_EXT | M_EXTPG) && (nd->nd_flag & ND_EXTPG) != 0) ||
((m2->m_flags & (M_EXT | M_EXTPG)) !=
(M_EXT | M_EXTPG) && (nd->nd_flag & ND_EXTPG) == 0),
("nfsm_strtom: ext_pgs and non-ext_pgs mbufs mixed"));
/*
* Loop around copying the string to mbuf(s).
*/
while (siz > 0) {
if (left == 0) {
if ((nd->nd_flag & ND_EXTPG) != 0) {
m2 = nfsm_add_ext_pgs(m2,
nd->nd_maxextsiz, &nd->nd_bextpg);
cp2 = (char *)(void *)PHYS_TO_DMAP(
m2->m_epg_pa[nd->nd_bextpg]);
nd->nd_bextpgsiz = left = PAGE_SIZE;
} else {
if (siz > ncl_mbuf_mlen)
NFSMCLGET(m1, M_WAITOK);
else
NFSMGET(m1);
m1->m_len = 0;
cp2 = mtod(m1, char *);
left = M_TRAILINGSPACE(m1);
m2->m_next = m1;
m2 = m1;
}
}
if (left >= siz)
xfer = siz;
else
xfer = left;
NFSBCOPY(cp, cp2, xfer);
cp += xfer;
cp2 += xfer;
m2->m_len += xfer;
siz -= xfer;
left -= xfer;
if ((nd->nd_flag & ND_EXTPG) != 0) {
nd->nd_bextpgsiz -= xfer;
m2->m_epg_last_len += xfer;
}
if (siz == 0 && rem) {
if (left < rem)
panic("nfsm_strtom");
NFSBZERO(cp2, rem);
m2->m_len += rem;
cp2 += rem;
if ((nd->nd_flag & ND_EXTPG) != 0) {
nd->nd_bextpgsiz -= rem;
m2->m_epg_last_len += rem;
}
}
}
nd->nd_mb = m2;
if ((nd->nd_flag & ND_EXTPG) != 0)
nd->nd_bpos = cp2;
else
nd->nd_bpos = mtod(m2, char *) + m2->m_len;
return (bytesize);
}
/*
* Called once to initialize data structures...
*/
void
newnfs_init(void)
{
static int nfs_inited = 0;
if (nfs_inited)
return;
nfs_inited = 1;
newnfs_true = txdr_unsigned(TRUE);
newnfs_false = txdr_unsigned(FALSE);
newnfs_xdrneg1 = txdr_unsigned(-1);
nfscl_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
if (nfscl_ticks < 1)
nfscl_ticks = 1;
NFSSETBOOTTIME(nfsboottime);
/*
* Initialize reply list and start timer
*/
TAILQ_INIT(&nfsd_reqq);
}
/*
* Put a file handle in an mbuf list.
* If the size argument == 0, just use the default size.
* set_true == 1 if there should be an newnfs_true prepended on the file handle.
* Return the number of bytes output, including XDR overhead.
*/
int
nfsm_fhtom(struct nfsrv_descript *nd, u_int8_t *fhp, int size, int set_true)
{
u_int32_t *tl;
u_int8_t *cp;
int fullsiz, bytesize = 0;
if (size == 0)
size = NFSX_MYFH;
switch (nd->nd_flag & (ND_NFSV2 | ND_NFSV3 | ND_NFSV4)) {
case ND_NFSV2:
if (size > NFSX_V2FH)
panic("fh size > NFSX_V2FH for NFSv2");
NFSM_BUILD(cp, u_int8_t *, NFSX_V2FH);
NFSBCOPY(fhp, cp, size);
if (size < NFSX_V2FH)
NFSBZERO(cp + size, NFSX_V2FH - size);
bytesize = NFSX_V2FH;
break;
case ND_NFSV3:
case ND_NFSV4:
fullsiz = NFSM_RNDUP(size);
if (set_true) {
bytesize = 2 * NFSX_UNSIGNED + fullsiz;
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = newnfs_true;
} else {
bytesize = NFSX_UNSIGNED + fullsiz;
}
(void) nfsm_strtom(nd, fhp, size);
break;
}
return (bytesize);
}
/*
* This function compares two net addresses by family and returns TRUE
* if they are the same host.
* If there is any doubt, return FALSE.
* The AF_INET family is handled as a special case so that address mbufs
* don't need to be saved to store "struct in_addr", which is only 4 bytes.
*/
int
nfsaddr_match(int family, union nethostaddr *haddr, NFSSOCKADDR_T nam)
{
#ifdef INET
struct sockaddr_in *inetaddr;
#endif
switch (family) {
#ifdef INET
case AF_INET:
inetaddr = NFSSOCKADDR(nam, struct sockaddr_in *);
if (inetaddr->sin_family == AF_INET &&
inetaddr->sin_addr.s_addr == haddr->had_inet.s_addr)
return (1);
break;
#endif
#ifdef INET6
case AF_INET6:
{
struct sockaddr_in6 *inetaddr6;
inetaddr6 = NFSSOCKADDR(nam, struct sockaddr_in6 *);
/* XXX - should test sin6_scope_id ? */
if (inetaddr6->sin6_family == AF_INET6 &&
IN6_ARE_ADDR_EQUAL(&inetaddr6->sin6_addr,
&haddr->had_inet6))
return (1);
}
break;
#endif
}
return (0);
}
/*
* Similar to the above, but takes to NFSSOCKADDR_T args.
*/
int
nfsaddr2_match(NFSSOCKADDR_T nam1, NFSSOCKADDR_T nam2)
{
struct sockaddr_in *addr1, *addr2;
struct sockaddr *inaddr;
inaddr = NFSSOCKADDR(nam1, struct sockaddr *);
switch (inaddr->sa_family) {
case AF_INET:
addr1 = NFSSOCKADDR(nam1, struct sockaddr_in *);
addr2 = NFSSOCKADDR(nam2, struct sockaddr_in *);
if (addr2->sin_family == AF_INET &&
addr1->sin_addr.s_addr == addr2->sin_addr.s_addr)
return (1);
break;
#ifdef INET6
case AF_INET6:
{
struct sockaddr_in6 *inet6addr1, *inet6addr2;
inet6addr1 = NFSSOCKADDR(nam1, struct sockaddr_in6 *);
inet6addr2 = NFSSOCKADDR(nam2, struct sockaddr_in6 *);
/* XXX - should test sin6_scope_id ? */
if (inet6addr2->sin6_family == AF_INET6 &&
IN6_ARE_ADDR_EQUAL(&inet6addr1->sin6_addr,
&inet6addr2->sin6_addr))
return (1);
}
break;
#endif
}
return (0);
}
/*
* Dissect a file handle on the client.
*/
int
nfsm_getfh(struct nfsrv_descript *nd, struct nfsfh **nfhpp)
{
u_int32_t *tl;
struct nfsfh *nfhp;
int error, len;
*nfhpp = NULL;
if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
len > NFSX_FHMAX) {
error = EBADRPC;
goto nfsmout;
}
} else
len = NFSX_V2FH;
nfhp = malloc(sizeof (struct nfsfh) + len,
M_NFSFH, M_WAITOK);
error = nfsrv_mtostr(nd, nfhp->nfh_fh, len);
if (error) {
free(nfhp, M_NFSFH);
goto nfsmout;
}
nfhp->nfh_len = len;
*nfhpp = nfhp;
nfsmout:
NFSEXITCODE2(error, nd);
return (error);
}
/*
* Break down the nfsv4 acl.
* If the aclp == NULL or won't fit in an acl, just discard the acl info.
*/
int
-nfsrv_dissectacl(struct nfsrv_descript *nd, NFSACL_T *aclp, int *aclerrp,
- int *aclsizep, __unused NFSPROC_T *p)
+nfsrv_dissectacl(struct nfsrv_descript *nd, NFSACL_T *aclp, bool server,
+ int *aclerrp, int *aclsizep, __unused NFSPROC_T *p)
{
u_int32_t *tl;
int i, aclsize;
int acecnt, error = 0, aceerr = 0, acesize;
*aclerrp = 0;
if (aclp)
aclp->acl_cnt = 0;
/*
* Parse out the ace entries and expect them to conform to
* what can be supported by R/W/X bits.
*/
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
aclsize = NFSX_UNSIGNED;
acecnt = fxdr_unsigned(int, *tl);
/*
* The RFCs do not define a fixed limit to the number of ACEs in
* an ACL, but 10240 should be more than sufficient.
*/
if (acecnt < 0 || acecnt > 10240) {
error = NFSERR_BADXDR;
goto nfsmout;
}
if (acecnt > ACL_MAX_ENTRIES)
aceerr = NFSERR_ATTRNOTSUPP;
if (nfsrv_useacl == 0)
aceerr = NFSERR_ATTRNOTSUPP;
for (i = 0; i < acecnt; i++) {
if (aclp && !aceerr)
error = nfsrv_dissectace(nd, &aclp->acl_entry[i],
- &aceerr, &acesize, p);
+ server, &aceerr, &acesize, p);
else
error = nfsrv_skipace(nd, &acesize);
if (error)
goto nfsmout;
aclsize += acesize;
}
if (aclp && !aceerr)
aclp->acl_cnt = acecnt;
if (aceerr)
*aclerrp = aceerr;
if (aclsizep)
*aclsizep = aclsize;
nfsmout:
NFSEXITCODE2(error, nd);
return (error);
}
/*
* Skip over an NFSv4 ace entry. Just dissect the xdr and discard it.
*/
static int
nfsrv_skipace(struct nfsrv_descript *nd, int *acesizep)
{
u_int32_t *tl;
int error, len = 0;
NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
len = fxdr_unsigned(int, *(tl + 3));
error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
nfsmout:
*acesizep = NFSM_RNDUP(len) + (4 * NFSX_UNSIGNED);
NFSEXITCODE2(error, nd);
return (error);
}
/*
* Get attribute bits from an mbuf list.
* Returns EBADRPC for a parsing error, 0 otherwise.
* If the clearinvalid flag is set, clear the bits not supported.
*/
int
nfsrv_getattrbits(struct nfsrv_descript *nd, nfsattrbit_t *attrbitp, int *cntp,
int *retnotsupp)
{
u_int32_t *tl;
int cnt, i, outcnt;
int error = 0;
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
cnt = fxdr_unsigned(int, *tl);
if (cnt < 0) {
error = NFSERR_BADXDR;
goto nfsmout;
}
if (cnt > NFSATTRBIT_MAXWORDS)
outcnt = NFSATTRBIT_MAXWORDS;
else
outcnt = cnt;
NFSZERO_ATTRBIT(attrbitp);
if (outcnt > 0) {
NFSM_DISSECT(tl, u_int32_t *, outcnt * NFSX_UNSIGNED);
for (i = 0; i < outcnt; i++)
attrbitp->bits[i] = fxdr_unsigned(u_int32_t, *tl++);
}
for (i = 0; i < (cnt - outcnt); i++) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (retnotsupp != NULL && *tl != 0)
*retnotsupp = NFSERR_ATTRNOTSUPP;
}
if (cntp)
*cntp = NFSX_UNSIGNED + (cnt * NFSX_UNSIGNED);
nfsmout:
NFSEXITCODE2(error, nd);
return (error);
}
/*
* Get the attributes for V4.
* If the compare flag is true, test for any attribute changes,
* otherwise return the attribute values.
* These attributes cover fields in "struct vattr", "struct statfs",
* "struct nfsfsinfo", the file handle and the lease duration.
* The value of retcmpp is set to 1 if all attributes are the same,
* and 0 otherwise.
* Returns EBADRPC if it can't be parsed, 0 otherwise.
*/
int
nfsv4_loadattr(struct nfsrv_descript *nd, vnode_t vp,
struct nfsvattr *nap, struct nfsfh **nfhpp, fhandle_t *fhp, int fhsize,
struct nfsv3_pathconf *pc, struct statfs *sbp, struct nfsstatfs *sfp,
struct nfsfsinfo *fsp, NFSACL_T *aclp, int compare, int *retcmpp,
u_int32_t *leasep, u_int32_t *rderrp, NFSPROC_T *p, struct ucred *cred)
{
u_int32_t *tl;
int i = 0, j, k, l = 0, m, bitpos, attrsum = 0;
int error, tfhsize, aceerr, attrsize, cnt, retnotsup;
u_char *cp, *cp2, namestr[NFSV4_SMALLSTR + 1];
nfsattrbit_t attrbits, retattrbits, checkattrbits;
struct nfsfh *tnfhp;
struct nfsreferral *refp;
u_quad_t tquad;
nfsquad_t tnfsquad;
struct timespec temptime;
uid_t uid;
gid_t gid;
u_int32_t freenum = 0, tuint;
u_int64_t uquad = 0, thyp, thyp2;
#ifdef QUOTA
struct dqblk dqb;
uid_t savuid;
#endif
CTASSERT(sizeof(ino_t) == sizeof(uint64_t));
if (compare) {
retnotsup = 0;
error = nfsrv_getattrbits(nd, &attrbits, NULL, &retnotsup);
} else {
error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
}
if (error)
goto nfsmout;
if (compare) {
*retcmpp = retnotsup;
} else {
/*
* Just set default values to some of the important ones.
*/
if (nap != NULL) {
nap->na_type = VREG;
nap->na_mode = 0;
nap->na_rdev = (NFSDEV_T)0;
nap->na_mtime.tv_sec = 0;
nap->na_mtime.tv_nsec = 0;
nap->na_btime.tv_sec = -1;
nap->na_btime.tv_nsec = 0;
nap->na_gen = 0;
nap->na_flags = 0;
nap->na_blocksize = NFS_FABLKSIZE;
}
if (sbp != NULL) {
sbp->f_bsize = NFS_FABLKSIZE;
sbp->f_blocks = 0;
sbp->f_bfree = 0;
sbp->f_bavail = 0;
sbp->f_files = 0;
sbp->f_ffree = 0;
}
if (fsp != NULL) {
fsp->fs_rtmax = 8192;
fsp->fs_rtpref = 8192;
fsp->fs_maxname = NFS_MAXNAMLEN;
fsp->fs_wtmax = 8192;
fsp->fs_wtpref = 8192;
fsp->fs_wtmult = NFS_FABLKSIZE;
fsp->fs_dtpref = 8192;
fsp->fs_maxfilesize = 0xffffffffffffffffull;
fsp->fs_timedelta.tv_sec = 0;
fsp->fs_timedelta.tv_nsec = 1;
fsp->fs_properties = (NFSV3_FSFLINK | NFSV3_FSFSYMLINK |
NFSV3_FSFHOMOGENEOUS | NFSV3_FSFCANSETTIME);
}
if (pc != NULL) {
pc->pc_linkmax = NFS_LINK_MAX;
pc->pc_namemax = NAME_MAX;
pc->pc_notrunc = 0;
pc->pc_chownrestricted = 0;
pc->pc_caseinsensitive = 0;
pc->pc_casepreserving = 1;
}
if (sfp != NULL) {
sfp->sf_ffiles = UINT64_MAX;
sfp->sf_tfiles = UINT64_MAX;
sfp->sf_afiles = UINT64_MAX;
sfp->sf_fbytes = UINT64_MAX;
sfp->sf_tbytes = UINT64_MAX;
sfp->sf_abytes = UINT64_MAX;
}
}
/*
* Loop around getting the attributes.
*/
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
attrsize = fxdr_unsigned(int, *tl);
for (bitpos = 0; bitpos < NFSATTRBIT_MAX; bitpos++) {
if (attrsum > attrsize) {
error = NFSERR_BADXDR;
goto nfsmout;
}
if (NFSISSET_ATTRBIT(&attrbits, bitpos))
switch (bitpos) {
case NFSATTRBIT_SUPPORTEDATTRS:
retnotsup = 0;
if (compare || nap == NULL)
error = nfsrv_getattrbits(nd, &retattrbits,
&cnt, &retnotsup);
else
error = nfsrv_getattrbits(nd, &nap->na_suppattr,
&cnt, &retnotsup);
if (error)
goto nfsmout;
if (compare && !(*retcmpp)) {
NFSSETSUPP_ATTRBIT(&checkattrbits, nd);
/* Some filesystem do not support NFSv4ACL */
if (nfsrv_useacl == 0 || nfs_supportsnfsv4acls(vp) == 0) {
NFSCLRBIT_ATTRBIT(&checkattrbits, NFSATTRBIT_ACL);
NFSCLRBIT_ATTRBIT(&checkattrbits, NFSATTRBIT_ACLSUPPORT);
}
if (!NFSEQUAL_ATTRBIT(&retattrbits, &checkattrbits)
|| retnotsup)
*retcmpp = NFSERR_NOTSAME;
}
attrsum += cnt;
break;
case NFSATTRBIT_TYPE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare) {
if (!(*retcmpp)) {
if (nap->na_type != nfsv34tov_type(*tl))
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL) {
nap->na_type = nfsv34tov_type(*tl);
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_FHEXPIRETYPE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare && !(*retcmpp)) {
if (fxdr_unsigned(int, *tl) !=
NFSV4FHTYPE_PERSISTENT)
*retcmpp = NFSERR_NOTSAME;
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_CHANGE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
if (compare) {
if (!(*retcmpp)) {
if (nap->na_filerev != fxdr_hyper(tl))
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL) {
nap->na_filerev = fxdr_hyper(tl);
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_SIZE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
if (compare) {
if (!(*retcmpp)) {
if (nap->na_size != fxdr_hyper(tl))
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL) {
nap->na_size = fxdr_hyper(tl);
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_LINKSUPPORT:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare) {
if (!(*retcmpp)) {
if (fsp->fs_properties & NFSV3_FSFLINK) {
if (*tl == newnfs_false)
*retcmpp = NFSERR_NOTSAME;
} else {
if (*tl == newnfs_true)
*retcmpp = NFSERR_NOTSAME;
}
}
} else if (fsp != NULL) {
if (*tl == newnfs_true)
fsp->fs_properties |= NFSV3_FSFLINK;
else
fsp->fs_properties &= ~NFSV3_FSFLINK;
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_SYMLINKSUPPORT:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare) {
if (!(*retcmpp)) {
if (fsp->fs_properties & NFSV3_FSFSYMLINK) {
if (*tl == newnfs_false)
*retcmpp = NFSERR_NOTSAME;
} else {
if (*tl == newnfs_true)
*retcmpp = NFSERR_NOTSAME;
}
}
} else if (fsp != NULL) {
if (*tl == newnfs_true)
fsp->fs_properties |= NFSV3_FSFSYMLINK;
else
fsp->fs_properties &= ~NFSV3_FSFSYMLINK;
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_NAMEDATTR:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare && !(*retcmpp)) {
if (*tl != newnfs_false)
*retcmpp = NFSERR_NOTSAME;
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_FSID:
NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
thyp = fxdr_hyper(tl);
tl += 2;
thyp2 = fxdr_hyper(tl);
if (compare) {
if (*retcmpp == 0) {
if (thyp != (u_int64_t)
vp->v_mount->mnt_stat.f_fsid.val[0] ||
thyp2 != (u_int64_t)
vp->v_mount->mnt_stat.f_fsid.val[1])
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL) {
nap->na_filesid[0] = thyp;
nap->na_filesid[1] = thyp2;
}
attrsum += (4 * NFSX_UNSIGNED);
break;
case NFSATTRBIT_UNIQUEHANDLES:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare && !(*retcmpp)) {
if (*tl != newnfs_true)
*retcmpp = NFSERR_NOTSAME;
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_LEASETIME:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare) {
if (fxdr_unsigned(int, *tl) != nfsrv_lease &&
!(*retcmpp))
*retcmpp = NFSERR_NOTSAME;
} else if (leasep != NULL) {
*leasep = fxdr_unsigned(u_int32_t, *tl);
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_RDATTRERROR:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare) {
if (!(*retcmpp))
*retcmpp = NFSERR_INVAL;
} else if (rderrp != NULL) {
*rderrp = fxdr_unsigned(u_int32_t, *tl);
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_ACL:
if (compare) {
if (!(*retcmpp)) {
if (nfsrv_useacl && nfs_supportsnfsv4acls(vp)) {
NFSACL_T *naclp;
naclp = acl_alloc(M_WAITOK);
- error = nfsrv_dissectacl(nd, naclp, &aceerr,
- &cnt, p);
+ error = nfsrv_dissectacl(nd, naclp, true,
+ &aceerr, &cnt, p);
if (error) {
acl_free(naclp);
goto nfsmout;
}
if (aceerr || aclp == NULL ||
nfsrv_compareacl(aclp, naclp))
*retcmpp = NFSERR_NOTSAME;
acl_free(naclp);
} else {
- error = nfsrv_dissectacl(nd, NULL, &aceerr,
- &cnt, p);
+ error = nfsrv_dissectacl(nd, NULL, true,
+ &aceerr, &cnt, p);
if (error)
goto nfsmout;
*retcmpp = NFSERR_ATTRNOTSUPP;
}
}
} else {
if (vp != NULL && aclp != NULL)
- error = nfsrv_dissectacl(nd, aclp, &aceerr,
- &cnt, p);
+ error = nfsrv_dissectacl(nd, aclp, false,
+ &aceerr, &cnt, p);
else
- error = nfsrv_dissectacl(nd, NULL, &aceerr,
- &cnt, p);
+ error = nfsrv_dissectacl(nd, NULL, false,
+ &aceerr, &cnt, p);
if (error)
goto nfsmout;
}
attrsum += cnt;
break;
case NFSATTRBIT_ACLSUPPORT:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare && !(*retcmpp)) {
if (nfsrv_useacl && nfs_supportsnfsv4acls(vp)) {
if (fxdr_unsigned(u_int32_t, *tl) !=
NFSV4ACE_SUPTYPES)
*retcmpp = NFSERR_NOTSAME;
} else {
*retcmpp = NFSERR_ATTRNOTSUPP;
}
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_ARCHIVE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare && !(*retcmpp))
*retcmpp = NFSERR_ATTRNOTSUPP;
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_CANSETTIME:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare) {
if (!(*retcmpp)) {
if (fsp->fs_properties & NFSV3_FSFCANSETTIME) {
if (*tl == newnfs_false)
*retcmpp = NFSERR_NOTSAME;
} else {
if (*tl == newnfs_true)
*retcmpp = NFSERR_NOTSAME;
}
}
} else if (fsp != NULL) {
if (*tl == newnfs_true)
fsp->fs_properties |= NFSV3_FSFCANSETTIME;
else
fsp->fs_properties &= ~NFSV3_FSFCANSETTIME;
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_CASEINSENSITIVE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare) {
if (!(*retcmpp)) {
if (*tl != newnfs_false)
*retcmpp = NFSERR_NOTSAME;
}
} else if (pc != NULL) {
pc->pc_caseinsensitive =
fxdr_unsigned(u_int32_t, *tl);
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_CASEPRESERVING:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare) {
if (!(*retcmpp)) {
if (*tl != newnfs_true)
*retcmpp = NFSERR_NOTSAME;
}
} else if (pc != NULL) {
pc->pc_casepreserving =
fxdr_unsigned(u_int32_t, *tl);
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_CHOWNRESTRICTED:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare) {
if (!(*retcmpp)) {
if (*tl != newnfs_true)
*retcmpp = NFSERR_NOTSAME;
}
} else if (pc != NULL) {
pc->pc_chownrestricted =
fxdr_unsigned(u_int32_t, *tl);
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_FILEHANDLE:
error = nfsm_getfh(nd, &tnfhp);
if (error)
goto nfsmout;
tfhsize = tnfhp->nfh_len;
if (compare) {
if (!(*retcmpp) &&
!NFSRV_CMPFH(tnfhp->nfh_fh, tfhsize,
fhp, fhsize))
*retcmpp = NFSERR_NOTSAME;
free(tnfhp, M_NFSFH);
} else if (nfhpp != NULL) {
*nfhpp = tnfhp;
} else {
free(tnfhp, M_NFSFH);
}
attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(tfhsize));
break;
case NFSATTRBIT_FILEID:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
thyp = fxdr_hyper(tl);
if (compare) {
if (!(*retcmpp)) {
if (nap->na_fileid != thyp)
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL)
nap->na_fileid = thyp;
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_FILESAVAIL:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
if (compare) {
uquad = nfsv4_filesavail(sbp, vp->v_mount);
if (!(*retcmpp) && uquad != fxdr_hyper(tl))
*retcmpp = NFSERR_NOTSAME;
} else if (sfp != NULL) {
sfp->sf_afiles = fxdr_hyper(tl);
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_FILESFREE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
if (compare) {
uquad = (uint64_t)sbp->f_ffree;
if (!(*retcmpp) && uquad != fxdr_hyper(tl))
*retcmpp = NFSERR_NOTSAME;
} else if (sfp != NULL) {
sfp->sf_ffiles = fxdr_hyper(tl);
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_FILESTOTAL:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
if (compare) {
uquad = sbp->f_files;
if (!(*retcmpp) && uquad != fxdr_hyper(tl))
*retcmpp = NFSERR_NOTSAME;
} else if (sfp != NULL) {
sfp->sf_tfiles = fxdr_hyper(tl);
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_FSLOCATIONS:
error = nfsrv_getrefstr(nd, &cp, &cp2, &l, &m);
if (error)
goto nfsmout;
attrsum += l;
if (compare && !(*retcmpp)) {
refp = nfsv4root_getreferral(vp, NULL, 0);
if (refp != NULL) {
if (cp == NULL || cp2 == NULL ||
strcmp(cp, "/") ||
strcmp(cp2, refp->nfr_srvlist))
*retcmpp = NFSERR_NOTSAME;
} else if (m == 0) {
*retcmpp = NFSERR_NOTSAME;
}
}
if (cp != NULL)
free(cp, M_NFSSTRING);
if (cp2 != NULL)
free(cp2, M_NFSSTRING);
break;
case NFSATTRBIT_HIDDEN:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare && !(*retcmpp))
*retcmpp = NFSERR_ATTRNOTSUPP;
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_HOMOGENEOUS:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare) {
if (!(*retcmpp)) {
if (fsp->fs_properties &
NFSV3_FSFHOMOGENEOUS) {
if (*tl == newnfs_false)
*retcmpp = NFSERR_NOTSAME;
} else {
if (*tl == newnfs_true)
*retcmpp = NFSERR_NOTSAME;
}
}
} else if (fsp != NULL) {
if (*tl == newnfs_true)
fsp->fs_properties |= NFSV3_FSFHOMOGENEOUS;
else
fsp->fs_properties &= ~NFSV3_FSFHOMOGENEOUS;
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_MAXFILESIZE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
tnfsquad.qval = fxdr_hyper(tl);
if (compare) {
if (!(*retcmpp)) {
tquad = NFSRV_MAXFILESIZE;
if (tquad != tnfsquad.qval)
*retcmpp = NFSERR_NOTSAME;
}
} else if (fsp != NULL) {
fsp->fs_maxfilesize = tnfsquad.qval;
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_MAXLINK:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare) {
if (!(*retcmpp)) {
if (fxdr_unsigned(int, *tl) != NFS_LINK_MAX)
*retcmpp = NFSERR_NOTSAME;
}
} else if (pc != NULL) {
pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl);
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_MAXNAME:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare) {
if (!(*retcmpp)) {
if (fsp->fs_maxname !=
fxdr_unsigned(u_int32_t, *tl))
*retcmpp = NFSERR_NOTSAME;
}
} else {
tuint = fxdr_unsigned(u_int32_t, *tl);
/*
* Some Linux NFSv4 servers report this
* as 0 or 4billion, so I'll set it to
* NFS_MAXNAMLEN. If a server actually creates
* a name longer than NFS_MAXNAMLEN, it will
* get an error back.
*/
if (tuint == 0 || tuint > NFS_MAXNAMLEN)
tuint = NFS_MAXNAMLEN;
if (fsp != NULL)
fsp->fs_maxname = tuint;
if (pc != NULL)
pc->pc_namemax = tuint;
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_MAXREAD:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
if (compare) {
if (!(*retcmpp)) {
if (fsp->fs_rtmax != fxdr_unsigned(u_int32_t,
*(tl + 1)) || *tl != 0)
*retcmpp = NFSERR_NOTSAME;
}
} else if (fsp != NULL) {
fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *++tl);
fsp->fs_rtpref = fsp->fs_rtmax;
fsp->fs_dtpref = fsp->fs_rtpref;
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_MAXWRITE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
if (compare) {
if (!(*retcmpp)) {
if (fsp->fs_wtmax != fxdr_unsigned(u_int32_t,
*(tl + 1)) || *tl != 0)
*retcmpp = NFSERR_NOTSAME;
}
} else if (fsp != NULL) {
fsp->fs_wtmax = fxdr_unsigned(int, *++tl);
fsp->fs_wtpref = fsp->fs_wtmax;
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_MIMETYPE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
i = fxdr_unsigned(int, *tl);
attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(i));
error = nfsm_advance(nd, NFSM_RNDUP(i), -1);
if (error)
goto nfsmout;
if (compare && !(*retcmpp))
*retcmpp = NFSERR_ATTRNOTSUPP;
break;
case NFSATTRBIT_MODE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare) {
if (!(*retcmpp)) {
if (nap->na_mode != nfstov_mode(*tl))
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL) {
nap->na_mode = nfstov_mode(*tl);
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_NOTRUNC:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare) {
if (!(*retcmpp)) {
if (*tl != newnfs_true)
*retcmpp = NFSERR_NOTSAME;
}
} else if (pc != NULL) {
pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl);
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_NUMLINKS:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
tuint = fxdr_unsigned(u_int32_t, *tl);
if (compare) {
if (!(*retcmpp)) {
if ((u_int32_t)nap->na_nlink != tuint)
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL) {
nap->na_nlink = tuint;
}
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_OWNER:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
j = fxdr_unsigned(int, *tl);
if (j < 0) {
error = NFSERR_BADXDR;
goto nfsmout;
}
attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(j));
if (j > NFSV4_SMALLSTR)
cp = malloc(j + 1, M_NFSSTRING, M_WAITOK);
else
cp = namestr;
error = nfsrv_mtostr(nd, cp, j);
if (error) {
if (j > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
goto nfsmout;
}
if (compare) {
if (!(*retcmpp)) {
if (nfsv4_strtouid(nd, cp, j, &uid) ||
nap->na_uid != uid)
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL) {
if (nfsv4_strtouid(nd, cp, j, &uid))
nap->na_uid = nfsrv_defaultuid;
else
nap->na_uid = uid;
}
if (j > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
break;
case NFSATTRBIT_OWNERGROUP:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
j = fxdr_unsigned(int, *tl);
if (j < 0) {
error = NFSERR_BADXDR;
goto nfsmout;
}
attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(j));
if (j > NFSV4_SMALLSTR)
cp = malloc(j + 1, M_NFSSTRING, M_WAITOK);
else
cp = namestr;
error = nfsrv_mtostr(nd, cp, j);
if (error) {
if (j > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
goto nfsmout;
}
if (compare) {
if (!(*retcmpp)) {
if (nfsv4_strtogid(nd, cp, j, &gid) ||
nap->na_gid != gid)
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL) {
if (nfsv4_strtogid(nd, cp, j, &gid))
nap->na_gid = nfsrv_defaultgid;
else
nap->na_gid = gid;
}
if (j > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
break;
case NFSATTRBIT_QUOTAHARD:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
if (sbp != NULL) {
if (priv_check_cred(cred, PRIV_VFS_EXCEEDQUOTA))
freenum = sbp->f_bfree;
else
freenum = sbp->f_bavail;
#ifdef QUOTA
/*
* ufs_quotactl() insists that the uid argument
* equal p_ruid for non-root quota access, so
* we'll just make sure that's the case.
*/
savuid = p->p_cred->p_ruid;
p->p_cred->p_ruid = cred->cr_uid;
if (!VFS_QUOTACTL(vp->v_mount,QCMD(Q_GETQUOTA,
USRQUOTA), cred->cr_uid, &dqb))
freenum = min(dqb.dqb_bhardlimit, freenum);
p->p_cred->p_ruid = savuid;
#endif /* QUOTA */
uquad = (u_int64_t)freenum;
NFSQUOTABLKTOBYTE(uquad, sbp->f_bsize);
}
if (compare && !(*retcmpp)) {
if (uquad != fxdr_hyper(tl))
*retcmpp = NFSERR_NOTSAME;
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_QUOTASOFT:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
if (sbp != NULL) {
if (priv_check_cred(cred, PRIV_VFS_EXCEEDQUOTA))
freenum = sbp->f_bfree;
else
freenum = sbp->f_bavail;
#ifdef QUOTA
/*
* ufs_quotactl() insists that the uid argument
* equal p_ruid for non-root quota access, so
* we'll just make sure that's the case.
*/
savuid = p->p_cred->p_ruid;
p->p_cred->p_ruid = cred->cr_uid;
if (!VFS_QUOTACTL(vp->v_mount,QCMD(Q_GETQUOTA,
USRQUOTA), cred->cr_uid, &dqb))
freenum = min(dqb.dqb_bsoftlimit, freenum);
p->p_cred->p_ruid = savuid;
#endif /* QUOTA */
uquad = (u_int64_t)freenum;
NFSQUOTABLKTOBYTE(uquad, sbp->f_bsize);
}
if (compare && !(*retcmpp)) {
if (uquad != fxdr_hyper(tl))
*retcmpp = NFSERR_NOTSAME;
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_QUOTAUSED:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
if (sbp != NULL) {
freenum = 0;
#ifdef QUOTA
/*
* ufs_quotactl() insists that the uid argument
* equal p_ruid for non-root quota access, so
* we'll just make sure that's the case.
*/
savuid = p->p_cred->p_ruid;
p->p_cred->p_ruid = cred->cr_uid;
if (!VFS_QUOTACTL(vp->v_mount,QCMD(Q_GETQUOTA,
USRQUOTA), cred->cr_uid, &dqb))
freenum = dqb.dqb_curblocks;
p->p_cred->p_ruid = savuid;
#endif /* QUOTA */
uquad = (u_int64_t)freenum;
NFSQUOTABLKTOBYTE(uquad, sbp->f_bsize);
}
if (compare && !(*retcmpp)) {
if (uquad != fxdr_hyper(tl))
*retcmpp = NFSERR_NOTSAME;
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_RAWDEV:
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4SPECDATA);
j = fxdr_unsigned(int, *tl++);
k = fxdr_unsigned(int, *tl);
if (compare) {
if (!(*retcmpp)) {
if (nap->na_rdev != NFSMAKEDEV(j, k))
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL) {
nap->na_rdev = NFSMAKEDEV(j, k);
}
attrsum += NFSX_V4SPECDATA;
break;
case NFSATTRBIT_SPACEAVAIL:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
if (compare) {
if (priv_check_cred(cred,
PRIV_VFS_BLOCKRESERVE))
uquad = sbp->f_bfree;
else
uquad = (uint64_t)sbp->f_bavail;
uquad *= sbp->f_bsize;
if (!(*retcmpp) && uquad != fxdr_hyper(tl))
*retcmpp = NFSERR_NOTSAME;
} else if (sfp != NULL) {
sfp->sf_abytes = fxdr_hyper(tl);
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_SPACEFREE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
if (compare) {
uquad = sbp->f_bfree;
uquad *= sbp->f_bsize;
if (!(*retcmpp) && uquad != fxdr_hyper(tl))
*retcmpp = NFSERR_NOTSAME;
} else if (sfp != NULL) {
sfp->sf_fbytes = fxdr_hyper(tl);
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_SPACETOTAL:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
if (compare) {
uquad = sbp->f_blocks;
uquad *= sbp->f_bsize;
if (!(*retcmpp) && uquad != fxdr_hyper(tl))
*retcmpp = NFSERR_NOTSAME;
} else if (sfp != NULL) {
sfp->sf_tbytes = fxdr_hyper(tl);
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_SPACEUSED:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
thyp = fxdr_hyper(tl);
if (compare) {
if (!(*retcmpp)) {
if ((u_int64_t)nap->na_bytes != thyp)
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL) {
nap->na_bytes = thyp;
}
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_SYSTEM:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (compare && !(*retcmpp))
*retcmpp = NFSERR_ATTRNOTSUPP;
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_TIMEACCESS:
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
fxdr_nfsv4time(tl, &temptime);
if (compare) {
if (!(*retcmpp)) {
if (!NFS_CMPTIME(temptime, nap->na_atime))
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL) {
nap->na_atime = temptime;
}
attrsum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMEACCESSSET:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
attrsum += NFSX_UNSIGNED;
i = fxdr_unsigned(int, *tl);
if (i == NFSV4SATTRTIME_TOCLIENT) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
attrsum += NFSX_V4TIME;
}
if (compare && !(*retcmpp))
*retcmpp = NFSERR_INVAL;
break;
case NFSATTRBIT_TIMEBACKUP:
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
if (compare && !(*retcmpp))
*retcmpp = NFSERR_ATTRNOTSUPP;
attrsum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMECREATE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
fxdr_nfsv4time(tl, &temptime);
if (compare) {
if (!(*retcmpp)) {
if (!NFS_CMPTIME(temptime, nap->na_btime))
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL) {
nap->na_btime = temptime;
}
attrsum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMEDELTA:
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
if (fsp != NULL) {
if (compare) {
if (!(*retcmpp)) {
if ((u_int32_t)fsp->fs_timedelta.tv_sec !=
fxdr_unsigned(u_int32_t, *(tl + 1)) ||
(u_int32_t)fsp->fs_timedelta.tv_nsec !=
(fxdr_unsigned(u_int32_t, *(tl + 2)) %
1000000000) ||
*tl != 0)
*retcmpp = NFSERR_NOTSAME;
}
} else {
fxdr_nfsv4time(tl, &fsp->fs_timedelta);
}
}
attrsum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMEMETADATA:
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
fxdr_nfsv4time(tl, &temptime);
if (compare) {
if (!(*retcmpp)) {
if (!NFS_CMPTIME(temptime, nap->na_ctime))
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL) {
nap->na_ctime = temptime;
}
attrsum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMEMODIFY:
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
fxdr_nfsv4time(tl, &temptime);
if (compare) {
if (!(*retcmpp)) {
if (!NFS_CMPTIME(temptime, nap->na_mtime))
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL) {
nap->na_mtime = temptime;
}
attrsum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMEMODIFYSET:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
attrsum += NFSX_UNSIGNED;
i = fxdr_unsigned(int, *tl);
if (i == NFSV4SATTRTIME_TOCLIENT) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
attrsum += NFSX_V4TIME;
}
if (compare && !(*retcmpp))
*retcmpp = NFSERR_INVAL;
break;
case NFSATTRBIT_MOUNTEDONFILEID:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
thyp = fxdr_hyper(tl);
if (compare) {
if (!(*retcmpp)) {
if (!vp || !nfsrv_atroot(vp, &thyp2))
thyp2 = nap->na_fileid;
if (thyp2 != thyp)
*retcmpp = NFSERR_NOTSAME;
}
} else if (nap != NULL)
nap->na_mntonfileno = thyp;
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_SUPPATTREXCLCREAT:
retnotsup = 0;
error = nfsrv_getattrbits(nd, &retattrbits,
&cnt, &retnotsup);
if (error)
goto nfsmout;
if (compare && !(*retcmpp)) {
NFSSETSUPP_ATTRBIT(&checkattrbits, nd);
NFSCLRNOTSETABLE_ATTRBIT(&checkattrbits, nd);
NFSCLRBIT_ATTRBIT(&checkattrbits,
NFSATTRBIT_TIMEACCESSSET);
if (!NFSEQUAL_ATTRBIT(&retattrbits, &checkattrbits)
|| retnotsup)
*retcmpp = NFSERR_NOTSAME;
}
attrsum += cnt;
break;
case NFSATTRBIT_FSLAYOUTTYPE:
case NFSATTRBIT_LAYOUTTYPE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
attrsum += NFSX_UNSIGNED;
i = fxdr_unsigned(int, *tl);
/*
* The RFCs do not define an upper limit for the
* number of layout types, but 32 should be more
* than enough.
*/
if (i < 0 || i > 32) {
error = NFSERR_BADXDR;
goto nfsmout;
}
if (i > 0) {
NFSM_DISSECT(tl, u_int32_t *, i *
NFSX_UNSIGNED);
attrsum += i * NFSX_UNSIGNED;
j = fxdr_unsigned(int, *tl);
if (i == 1 && compare && !(*retcmpp) &&
(((nfsrv_doflexfile != 0 ||
nfsrv_maxpnfsmirror > 1) &&
j != NFSLAYOUT_FLEXFILE) ||
(nfsrv_doflexfile == 0 &&
j != NFSLAYOUT_NFSV4_1_FILES)))
*retcmpp = NFSERR_NOTSAME;
}
if (nfsrv_devidcnt == 0) {
if (compare && !(*retcmpp) && i > 0)
*retcmpp = NFSERR_NOTSAME;
} else {
if (compare && !(*retcmpp) && i != 1)
*retcmpp = NFSERR_NOTSAME;
}
break;
case NFSATTRBIT_LAYOUTALIGNMENT:
case NFSATTRBIT_LAYOUTBLKSIZE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
attrsum += NFSX_UNSIGNED;
i = fxdr_unsigned(int, *tl);
if (compare && !(*retcmpp) && i != nfs_srvmaxio)
*retcmpp = NFSERR_NOTSAME;
break;
default:
printf("EEK! nfsv4_loadattr unknown attr=%d\n",
bitpos);
if (compare && !(*retcmpp))
*retcmpp = NFSERR_ATTRNOTSUPP;
/*
* and get out of the loop, since we can't parse
* the unknown attrbute data.
*/
bitpos = NFSATTRBIT_MAX;
break;
}
}
/*
* some clients pad the attrlist, so we need to skip over the
* padding.
*/
if (attrsum > attrsize) {
error = NFSERR_BADXDR;
} else {
attrsize = NFSM_RNDUP(attrsize);
if (attrsum < attrsize)
error = nfsm_advance(nd, attrsize - attrsum, -1);
}
nfsmout:
NFSEXITCODE2(error, nd);
return (error);
}
/*
* Implement sleep locks for newnfs. The nfslock_usecnt allows for a
* shared lock and the NFSXXX_LOCK flag permits an exclusive lock.
* The first argument is a pointer to an nfsv4lock structure.
* The second argument is 1 iff a blocking lock is wanted.
* If this argument is 0, the call waits until no thread either wants nor
* holds an exclusive lock.
* It returns 1 if the lock was acquired, 0 otherwise.
* If several processes call this function concurrently wanting the exclusive
* lock, one will get the lock and the rest will return without getting the
* lock. (If the caller must have the lock, it simply calls this function in a
* loop until the function returns 1 to indicate the lock was acquired.)
* Any usecnt must be decremented by calling nfsv4_relref() before
* calling nfsv4_lock(). It was done this way, so nfsv4_lock() could
* be called in a loop.
* The isleptp argument is set to indicate if the call slept, iff not NULL
* and the mp argument indicates to check for a forced dismount, iff not
* NULL.
*/
int
nfsv4_lock(struct nfsv4lock *lp, int iwantlock, int *isleptp,
struct mtx *mutex, struct mount *mp)
{
if (isleptp)
*isleptp = 0;
/*
* If a lock is wanted, loop around until the lock is acquired by
* someone and then released. If I want the lock, try to acquire it.
* For a lock to be issued, no lock must be in force and the usecnt
* must be zero.
*/
if (iwantlock) {
if (!(lp->nfslock_lock & NFSV4LOCK_LOCK) &&
lp->nfslock_usecnt == 0) {
lp->nfslock_lock &= ~NFSV4LOCK_LOCKWANTED;
lp->nfslock_lock |= NFSV4LOCK_LOCK;
return (1);
}
lp->nfslock_lock |= NFSV4LOCK_LOCKWANTED;
}
while (lp->nfslock_lock & (NFSV4LOCK_LOCK | NFSV4LOCK_LOCKWANTED)) {
if (mp != NULL && NFSCL_FORCEDISM(mp)) {
lp->nfslock_lock &= ~NFSV4LOCK_LOCKWANTED;
return (0);
}
lp->nfslock_lock |= NFSV4LOCK_WANTED;
if (isleptp)
*isleptp = 1;
msleep(&lp->nfslock_lock, mutex, PVFS, "nfsv4lck", hz);
if (iwantlock && !(lp->nfslock_lock & NFSV4LOCK_LOCK) &&
lp->nfslock_usecnt == 0) {
lp->nfslock_lock &= ~NFSV4LOCK_LOCKWANTED;
lp->nfslock_lock |= NFSV4LOCK_LOCK;
return (1);
}
}
return (0);
}
/*
* Release the lock acquired by nfsv4_lock().
* The second argument is set to 1 to indicate the nfslock_usecnt should be
* incremented, as well.
*/
void
nfsv4_unlock(struct nfsv4lock *lp, int incref)
{
lp->nfslock_lock &= ~NFSV4LOCK_LOCK;
if (incref)
lp->nfslock_usecnt++;
nfsv4_wanted(lp);
}
/*
* Release a reference cnt.
*/
void
nfsv4_relref(struct nfsv4lock *lp)
{
if (lp->nfslock_usecnt <= 0)
panic("nfsv4root ref cnt");
lp->nfslock_usecnt--;
if (lp->nfslock_usecnt == 0)
nfsv4_wanted(lp);
}
/*
* Get a reference cnt.
* This function will wait for any exclusive lock to be released, but will
* not wait for threads that want the exclusive lock. If priority needs
* to be given to threads that need the exclusive lock, a call to nfsv4_lock()
* with the 2nd argument == 0 should be done before calling nfsv4_getref().
* If the mp argument is not NULL, check for NFSCL_FORCEDISM() being set and
* return without getting a refcnt for that case.
*/
void
nfsv4_getref(struct nfsv4lock *lp, int *isleptp, struct mtx *mutex,
struct mount *mp)
{
if (isleptp)
*isleptp = 0;
/*
* Wait for a lock held.
*/
while (lp->nfslock_lock & NFSV4LOCK_LOCK) {
if (mp != NULL && NFSCL_FORCEDISM(mp))
return;
lp->nfslock_lock |= NFSV4LOCK_WANTED;
if (isleptp)
*isleptp = 1;
msleep(&lp->nfslock_lock, mutex, PVFS, "nfsv4gr", hz);
}
if (mp != NULL && NFSCL_FORCEDISM(mp))
return;
lp->nfslock_usecnt++;
}
/*
* Get a reference as above, but return failure instead of sleeping if
* an exclusive lock is held.
*/
int
nfsv4_getref_nonblock(struct nfsv4lock *lp)
{
if ((lp->nfslock_lock & NFSV4LOCK_LOCK) != 0)
return (0);
lp->nfslock_usecnt++;
return (1);
}
/*
* Test for a lock. Return 1 if locked, 0 otherwise.
*/
int
nfsv4_testlock(struct nfsv4lock *lp)
{
if ((lp->nfslock_lock & NFSV4LOCK_LOCK) == 0 &&
lp->nfslock_usecnt == 0)
return (0);
return (1);
}
/*
* Wake up anyone sleeping, waiting for this lock.
*/
static void
nfsv4_wanted(struct nfsv4lock *lp)
{
if (lp->nfslock_lock & NFSV4LOCK_WANTED) {
lp->nfslock_lock &= ~NFSV4LOCK_WANTED;
wakeup((caddr_t)&lp->nfslock_lock);
}
}
/*
* Copy a string from an mbuf list into a character array.
* Return EBADRPC if there is an mbuf error,
* 0 otherwise.
*/
int
nfsrv_mtostr(struct nfsrv_descript *nd, char *str, int siz)
{
char *cp;
int xfer, len;
struct mbuf *mp;
int rem, error = 0;
mp = nd->nd_md;
cp = nd->nd_dpos;
len = mtod(mp, caddr_t) + mp->m_len - cp;
rem = NFSM_RNDUP(siz) - siz;
while (siz > 0) {
if (len > siz)
xfer = siz;
else
xfer = len;
NFSBCOPY(cp, str, xfer);
str += xfer;
siz -= xfer;
if (siz > 0) {
mp = mp->m_next;
if (mp == NULL) {
error = EBADRPC;
goto out;
}
cp = mtod(mp, caddr_t);
len = mp->m_len;
} else {
cp += xfer;
len -= xfer;
}
}
*str = '\0';
nd->nd_dpos = cp;
nd->nd_md = mp;
if (rem > 0) {
if (len < rem)
error = nfsm_advance(nd, rem, len);
else
nd->nd_dpos += rem;
}
out:
NFSEXITCODE2(error, nd);
return (error);
}
/*
* Fill in the attributes as marked by the bitmap (V4).
*/
int
nfsv4_fillattr(struct nfsrv_descript *nd, struct mount *mp, vnode_t vp,
NFSACL_T *saclp, struct vattr *vap, fhandle_t *fhp, int rderror,
nfsattrbit_t *attrbitp, struct ucred *cred, NFSPROC_T *p, int isdgram,
int reterr, int supports_nfsv4acls, int at_root, uint64_t mounted_on_fileno,
struct statfs *pnfssf)
{
int bitpos, retnum = 0;
u_int32_t *tl;
int siz, prefixnum, error;
u_char *cp, namestr[NFSV4_SMALLSTR];
nfsattrbit_t attrbits, retbits;
nfsattrbit_t *retbitp = &retbits;
u_int32_t freenum, *retnump;
u_int64_t uquad;
struct statfs *fs;
struct nfsfsinfo fsinf;
struct timespec temptime;
NFSACL_T *aclp, *naclp = NULL;
size_t atsiz;
bool xattrsupp;
#ifdef QUOTA
struct dqblk dqb;
uid_t savuid;
#endif
/*
* First, set the bits that can be filled and get fsinfo.
*/
NFSSET_ATTRBIT(retbitp, attrbitp);
/*
* If both p and cred are NULL, it is a client side setattr call.
* If both p and cred are not NULL, it is a server side reply call.
* If p is not NULL and cred is NULL, it is a client side callback
* reply call.
*/
if (p == NULL && cred == NULL) {
NFSCLRNOTSETABLE_ATTRBIT(retbitp, nd);
aclp = saclp;
} else {
NFSCLRNOTFILLABLE_ATTRBIT(retbitp, nd);
naclp = acl_alloc(M_WAITOK);
aclp = naclp;
}
nfsvno_getfs(&fsinf, isdgram);
#ifndef APPLE
/*
* Get the VFS_STATFS(), since some attributes need them.
*/
fs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
if (NFSISSETSTATFS_ATTRBIT(retbitp)) {
error = VFS_STATFS(mp, fs);
if (error != 0) {
if (reterr) {
nd->nd_repstat = NFSERR_ACCES;
free(fs, M_STATFS);
return (0);
}
NFSCLRSTATFS_ATTRBIT(retbitp);
}
/*
* Since NFS handles these values as unsigned on the
* wire, there is no way to represent negative values,
* so set them to 0. Without this, they will appear
* to be very large positive values for clients like
* Solaris10.
*/
if (fs->f_bavail < 0)
fs->f_bavail = 0;
if (fs->f_ffree < 0)
fs->f_ffree = 0;
}
#endif
/*
* And the NFSv4 ACL...
*/
if (NFSISSET_ATTRBIT(retbitp, NFSATTRBIT_ACLSUPPORT) &&
(nfsrv_useacl == 0 || ((cred != NULL || p != NULL) &&
supports_nfsv4acls == 0))) {
NFSCLRBIT_ATTRBIT(retbitp, NFSATTRBIT_ACLSUPPORT);
}
if (NFSISSET_ATTRBIT(retbitp, NFSATTRBIT_ACL)) {
if (nfsrv_useacl == 0 || ((cred != NULL || p != NULL) &&
supports_nfsv4acls == 0)) {
NFSCLRBIT_ATTRBIT(retbitp, NFSATTRBIT_ACL);
} else if (naclp != NULL) {
if (NFSVOPLOCK(vp, LK_SHARED) == 0) {
error = VOP_ACCESSX(vp, VREAD_ACL, cred, p);
if (error == 0)
error = VOP_GETACL(vp, ACL_TYPE_NFS4,
naclp, cred, p);
NFSVOPUNLOCK(vp);
} else
error = NFSERR_PERM;
if (error != 0) {
if (reterr) {
nd->nd_repstat = NFSERR_ACCES;
free(fs, M_STATFS);
return (0);
}
NFSCLRBIT_ATTRBIT(retbitp, NFSATTRBIT_ACL);
}
}
}
/* Check to see if Extended Attributes are supported. */
xattrsupp = false;
if (NFSISSET_ATTRBIT(retbitp, NFSATTRBIT_XATTRSUPPORT)) {
if (NFSVOPLOCK(vp, LK_SHARED) == 0) {
error = VOP_GETEXTATTR(vp, EXTATTR_NAMESPACE_USER,
"xxx", NULL, &atsiz, cred, p);
NFSVOPUNLOCK(vp);
if (error != EOPNOTSUPP)
xattrsupp = true;
}
}
/*
* Put out the attribute bitmap for the ones being filled in
* and get the field for the number of attributes returned.
*/
prefixnum = nfsrv_putattrbit(nd, retbitp);
NFSM_BUILD(retnump, u_int32_t *, NFSX_UNSIGNED);
prefixnum += NFSX_UNSIGNED;
/*
* Now, loop around filling in the attributes for each bit set.
*/
for (bitpos = 0; bitpos < NFSATTRBIT_MAX; bitpos++) {
if (NFSISSET_ATTRBIT(retbitp, bitpos)) {
switch (bitpos) {
case NFSATTRBIT_SUPPORTEDATTRS:
NFSSETSUPP_ATTRBIT(&attrbits, nd);
if (nfsrv_useacl == 0 || ((cred != NULL || p != NULL)
&& supports_nfsv4acls == 0)) {
NFSCLRBIT_ATTRBIT(&attrbits,NFSATTRBIT_ACLSUPPORT);
NFSCLRBIT_ATTRBIT(&attrbits,NFSATTRBIT_ACL);
}
retnum += nfsrv_putattrbit(nd, &attrbits);
break;
case NFSATTRBIT_TYPE:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = vtonfsv34_type(vap->va_type);
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_FHEXPIRETYPE:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4FHTYPE_PERSISTENT);
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_CHANGE:
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
txdr_hyper(vap->va_filerev, tl);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_SIZE:
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
txdr_hyper(vap->va_size, tl);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_LINKSUPPORT:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
if (fsinf.fs_properties & NFSV3FSINFO_LINK)
*tl = newnfs_true;
else
*tl = newnfs_false;
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_SYMLINKSUPPORT:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
if (fsinf.fs_properties & NFSV3FSINFO_SYMLINK)
*tl = newnfs_true;
else
*tl = newnfs_false;
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_NAMEDATTR:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = newnfs_false;
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_FSID:
NFSM_BUILD(tl, u_int32_t *, NFSX_V4FSID);
*tl++ = 0;
*tl++ = txdr_unsigned(mp->mnt_stat.f_fsid.val[0]);
*tl++ = 0;
*tl = txdr_unsigned(mp->mnt_stat.f_fsid.val[1]);
retnum += NFSX_V4FSID;
break;
case NFSATTRBIT_UNIQUEHANDLES:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = newnfs_true;
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_LEASETIME:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(nfsrv_lease);
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_RDATTRERROR:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(rderror);
retnum += NFSX_UNSIGNED;
break;
/*
* Recommended Attributes. (Only the supported ones.)
*/
case NFSATTRBIT_ACL:
retnum += nfsrv_buildacl(nd, aclp, vnode_vtype(vp), p);
break;
case NFSATTRBIT_ACLSUPPORT:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4ACE_SUPTYPES);
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_CANSETTIME:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
if (fsinf.fs_properties & NFSV3FSINFO_CANSETTIME)
*tl = newnfs_true;
else
*tl = newnfs_false;
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_CASEINSENSITIVE:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = newnfs_false;
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_CASEPRESERVING:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = newnfs_true;
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_CHOWNRESTRICTED:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = newnfs_true;
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_FILEHANDLE:
retnum += nfsm_fhtom(nd, (u_int8_t *)fhp, 0, 0);
break;
case NFSATTRBIT_FILEID:
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
uquad = vap->va_fileid;
txdr_hyper(uquad, tl);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_FILESAVAIL:
freenum = nfsv4_filesavail(fs, mp);
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
*tl++ = 0;
*tl = txdr_unsigned(freenum);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_FILESFREE:
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
*tl++ = 0;
*tl = txdr_unsigned(fs->f_ffree);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_FILESTOTAL:
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
*tl++ = 0;
*tl = txdr_unsigned(fs->f_files);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_FSLOCATIONS:
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = 0;
*tl = 0;
retnum += 2 * NFSX_UNSIGNED;
break;
case NFSATTRBIT_HOMOGENEOUS:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
if (fsinf.fs_properties & NFSV3FSINFO_HOMOGENEOUS)
*tl = newnfs_true;
else
*tl = newnfs_false;
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_MAXFILESIZE:
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
uquad = NFSRV_MAXFILESIZE;
txdr_hyper(uquad, tl);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_MAXLINK:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFS_LINK_MAX);
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_MAXNAME:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFS_MAXNAMLEN);
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_MAXREAD:
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
*tl++ = 0;
*tl = txdr_unsigned(fsinf.fs_rtmax);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_MAXWRITE:
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
*tl++ = 0;
*tl = txdr_unsigned(fsinf.fs_wtmax);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_MODE:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = vtonfsv34_mode(vap->va_mode);
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_NOTRUNC:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = newnfs_true;
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_NUMLINKS:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(vap->va_nlink);
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_OWNER:
cp = namestr;
nfsv4_uidtostr(vap->va_uid, &cp, &siz);
retnum += nfsm_strtom(nd, cp, siz);
if (cp != namestr)
free(cp, M_NFSSTRING);
break;
case NFSATTRBIT_OWNERGROUP:
cp = namestr;
nfsv4_gidtostr(vap->va_gid, &cp, &siz);
retnum += nfsm_strtom(nd, cp, siz);
if (cp != namestr)
free(cp, M_NFSSTRING);
break;
case NFSATTRBIT_QUOTAHARD:
if (priv_check_cred(cred, PRIV_VFS_EXCEEDQUOTA))
freenum = fs->f_bfree;
else
freenum = fs->f_bavail;
#ifdef QUOTA
/*
* ufs_quotactl() insists that the uid argument
* equal p_ruid for non-root quota access, so
* we'll just make sure that's the case.
*/
savuid = p->p_cred->p_ruid;
p->p_cred->p_ruid = cred->cr_uid;
if (!VFS_QUOTACTL(mp, QCMD(Q_GETQUOTA,USRQUOTA),
cred->cr_uid, &dqb))
freenum = min(dqb.dqb_bhardlimit, freenum);
p->p_cred->p_ruid = savuid;
#endif /* QUOTA */
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
uquad = (u_int64_t)freenum;
NFSQUOTABLKTOBYTE(uquad, fs->f_bsize);
txdr_hyper(uquad, tl);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_QUOTASOFT:
if (priv_check_cred(cred, PRIV_VFS_EXCEEDQUOTA))
freenum = fs->f_bfree;
else
freenum = fs->f_bavail;
#ifdef QUOTA
/*
* ufs_quotactl() insists that the uid argument
* equal p_ruid for non-root quota access, so
* we'll just make sure that's the case.
*/
savuid = p->p_cred->p_ruid;
p->p_cred->p_ruid = cred->cr_uid;
if (!VFS_QUOTACTL(mp, QCMD(Q_GETQUOTA,USRQUOTA),
cred->cr_uid, &dqb))
freenum = min(dqb.dqb_bsoftlimit, freenum);
p->p_cred->p_ruid = savuid;
#endif /* QUOTA */
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
uquad = (u_int64_t)freenum;
NFSQUOTABLKTOBYTE(uquad, fs->f_bsize);
txdr_hyper(uquad, tl);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_QUOTAUSED:
freenum = 0;
#ifdef QUOTA
/*
* ufs_quotactl() insists that the uid argument
* equal p_ruid for non-root quota access, so
* we'll just make sure that's the case.
*/
savuid = p->p_cred->p_ruid;
p->p_cred->p_ruid = cred->cr_uid;
if (!VFS_QUOTACTL(mp, QCMD(Q_GETQUOTA,USRQUOTA),
cred->cr_uid, &dqb))
freenum = dqb.dqb_curblocks;
p->p_cred->p_ruid = savuid;
#endif /* QUOTA */
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
uquad = (u_int64_t)freenum;
NFSQUOTABLKTOBYTE(uquad, fs->f_bsize);
txdr_hyper(uquad, tl);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_RAWDEV:
NFSM_BUILD(tl, u_int32_t *, NFSX_V4SPECDATA);
*tl++ = txdr_unsigned(NFSMAJOR(vap->va_rdev));
*tl = txdr_unsigned(NFSMINOR(vap->va_rdev));
retnum += NFSX_V4SPECDATA;
break;
case NFSATTRBIT_SPACEAVAIL:
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
if (priv_check_cred(cred, PRIV_VFS_BLOCKRESERVE)) {
if (pnfssf != NULL)
uquad = (u_int64_t)pnfssf->f_bfree;
else
uquad = (u_int64_t)fs->f_bfree;
} else {
if (pnfssf != NULL)
uquad = (u_int64_t)pnfssf->f_bavail;
else
uquad = (u_int64_t)fs->f_bavail;
}
if (pnfssf != NULL)
uquad *= pnfssf->f_bsize;
else
uquad *= fs->f_bsize;
txdr_hyper(uquad, tl);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_SPACEFREE:
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
if (pnfssf != NULL) {
uquad = (u_int64_t)pnfssf->f_bfree;
uquad *= pnfssf->f_bsize;
} else {
uquad = (u_int64_t)fs->f_bfree;
uquad *= fs->f_bsize;
}
txdr_hyper(uquad, tl);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_SPACETOTAL:
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
if (pnfssf != NULL) {
uquad = (u_int64_t)pnfssf->f_blocks;
uquad *= pnfssf->f_bsize;
} else {
uquad = (u_int64_t)fs->f_blocks;
uquad *= fs->f_bsize;
}
txdr_hyper(uquad, tl);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_SPACEUSED:
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
txdr_hyper(vap->va_bytes, tl);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_TIMEACCESS:
NFSM_BUILD(tl, u_int32_t *, NFSX_V4TIME);
txdr_nfsv4time(&vap->va_atime, tl);
retnum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMEACCESSSET:
if ((vap->va_vaflags & VA_UTIMES_NULL) == 0) {
NFSM_BUILD(tl, u_int32_t *, NFSX_V4SETTIME);
*tl++ = txdr_unsigned(NFSV4SATTRTIME_TOCLIENT);
txdr_nfsv4time(&vap->va_atime, tl);
retnum += NFSX_V4SETTIME;
} else {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4SATTRTIME_TOSERVER);
retnum += NFSX_UNSIGNED;
}
break;
case NFSATTRBIT_TIMEDELTA:
NFSM_BUILD(tl, u_int32_t *, NFSX_V4TIME);
temptime.tv_sec = 0;
temptime.tv_nsec = 1000000000 / hz;
txdr_nfsv4time(&temptime, tl);
retnum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMEMETADATA:
NFSM_BUILD(tl, u_int32_t *, NFSX_V4TIME);
txdr_nfsv4time(&vap->va_ctime, tl);
retnum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMEMODIFY:
NFSM_BUILD(tl, u_int32_t *, NFSX_V4TIME);
txdr_nfsv4time(&vap->va_mtime, tl);
retnum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMECREATE:
NFSM_BUILD(tl, u_int32_t *, NFSX_V4TIME);
txdr_nfsv4time(&vap->va_birthtime, tl);
retnum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMEMODIFYSET:
if ((vap->va_vaflags & VA_UTIMES_NULL) == 0) {
NFSM_BUILD(tl, u_int32_t *, NFSX_V4SETTIME);
*tl++ = txdr_unsigned(NFSV4SATTRTIME_TOCLIENT);
txdr_nfsv4time(&vap->va_mtime, tl);
retnum += NFSX_V4SETTIME;
} else {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4SATTRTIME_TOSERVER);
retnum += NFSX_UNSIGNED;
}
break;
case NFSATTRBIT_MOUNTEDONFILEID:
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER);
if (at_root != 0)
uquad = mounted_on_fileno;
else
uquad = vap->va_fileid;
txdr_hyper(uquad, tl);
retnum += NFSX_HYPER;
break;
case NFSATTRBIT_SUPPATTREXCLCREAT:
NFSSETSUPP_ATTRBIT(&attrbits, nd);
NFSCLRNOTSETABLE_ATTRBIT(&attrbits, nd);
NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEACCESSSET);
retnum += nfsrv_putattrbit(nd, &attrbits);
break;
case NFSATTRBIT_FSLAYOUTTYPE:
case NFSATTRBIT_LAYOUTTYPE:
if (nfsrv_devidcnt == 0)
siz = 1;
else
siz = 2;
if (siz == 2) {
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(1); /* One entry. */
if (nfsrv_doflexfile != 0 ||
nfsrv_maxpnfsmirror > 1)
*tl = txdr_unsigned(NFSLAYOUT_FLEXFILE);
else
*tl = txdr_unsigned(
NFSLAYOUT_NFSV4_1_FILES);
} else {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = 0;
}
retnum += siz * NFSX_UNSIGNED;
break;
case NFSATTRBIT_LAYOUTALIGNMENT:
case NFSATTRBIT_LAYOUTBLKSIZE:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(nfs_srvmaxio);
retnum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_XATTRSUPPORT:
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
if (xattrsupp)
*tl = newnfs_true;
else
*tl = newnfs_false;
retnum += NFSX_UNSIGNED;
break;
default:
printf("EEK! Bad V4 attribute bitpos=%d\n", bitpos);
}
}
}
if (naclp != NULL)
acl_free(naclp);
free(fs, M_STATFS);
*retnump = txdr_unsigned(retnum);
return (retnum + prefixnum);
}
/*
* Calculate the files available attribute value.
*/
static uint32_t
nfsv4_filesavail(struct statfs *fs, struct mount *mp)
{
uint32_t freenum;
#ifdef QUOTA
struct dqblk dqb;
uid_t savuid;
NFSPROC_T *p;
#endif
/*
* Check quota and use min(quota, f_ffree).
*/
freenum = fs->f_ffree;
#ifdef QUOTA
/*
* This is old OpenBSD code that does not build
* for FreeBSD. I do not know if doing this is
* useful, so I will just leave the code here.
*/
p = curthread();
/*
* ufs_quotactl() insists that the uid argument
* equal p_ruid for non-root quota access, so
* we'll just make sure that's the case.
*/
savuid = p->p_cred->p_ruid;
p->p_cred->p_ruid = cred->cr_uid;
if (!VFS_QUOTACTL(mp, QCMD(Q_GETQUOTA,USRQUOTA),
cred->cr_uid, &dqb))
freenum = min(dqb.dqb_isoftlimit-dqb.dqb_curinodes,
freenum);
p->p_cred->p_ruid = savuid;
#endif /* QUOTA */
return (freenum);
}
/*
* Put the attribute bits onto an mbuf list.
* Return the number of bytes of output generated.
*/
int
nfsrv_putattrbit(struct nfsrv_descript *nd, nfsattrbit_t *attrbitp)
{
u_int32_t *tl;
int cnt, i, bytesize;
for (cnt = NFSATTRBIT_MAXWORDS; cnt > 0; cnt--)
if (attrbitp->bits[cnt - 1])
break;
bytesize = (cnt + 1) * NFSX_UNSIGNED;
NFSM_BUILD(tl, u_int32_t *, bytesize);
*tl++ = txdr_unsigned(cnt);
for (i = 0; i < cnt; i++)
*tl++ = txdr_unsigned(attrbitp->bits[i]);
return (bytesize);
}
/*
* Convert a uid to a string.
* If the lookup fails, just output the digits.
* uid - the user id
* cpp - points to a buffer of size NFSV4_SMALLSTR
* (malloc a larger one, as required)
* retlenp - pointer to length to be returned
*/
void
nfsv4_uidtostr(uid_t uid, u_char **cpp, int *retlenp)
{
int i;
struct nfsusrgrp *usrp;
u_char *cp = *cpp;
uid_t tmp;
int cnt, hasampersand, len = NFSV4_SMALLSTR, ret;
struct nfsrv_lughash *hp;
cnt = 0;
tryagain:
if (nfsrv_dnsnamelen > 0 && !nfs_enable_uidtostring) {
/*
* Always map nfsrv_defaultuid to "nobody".
*/
if (uid == nfsrv_defaultuid) {
i = nfsrv_dnsnamelen + 7;
if (i > len) {
if (len > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
cp = malloc(i, M_NFSSTRING, M_WAITOK);
*cpp = cp;
len = i;
goto tryagain;
}
*retlenp = i;
NFSBCOPY("nobody@", cp, 7);
cp += 7;
NFSBCOPY(nfsrv_dnsname, cp, nfsrv_dnsnamelen);
return;
}
hasampersand = 0;
hp = NFSUSERHASH(uid);
mtx_lock(&hp->mtx);
TAILQ_FOREACH(usrp, &hp->lughead, lug_numhash) {
if (usrp->lug_uid == uid) {
if (usrp->lug_expiry < NFSD_MONOSEC)
break;
/*
* If the name doesn't already have an '@'
* in it, append @domainname to it.
*/
for (i = 0; i < usrp->lug_namelen; i++) {
if (usrp->lug_name[i] == '@') {
hasampersand = 1;
break;
}
}
if (hasampersand)
i = usrp->lug_namelen;
else
i = usrp->lug_namelen +
nfsrv_dnsnamelen + 1;
if (i > len) {
mtx_unlock(&hp->mtx);
if (len > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
cp = malloc(i, M_NFSSTRING, M_WAITOK);
*cpp = cp;
len = i;
goto tryagain;
}
*retlenp = i;
NFSBCOPY(usrp->lug_name, cp, usrp->lug_namelen);
if (!hasampersand) {
cp += usrp->lug_namelen;
*cp++ = '@';
NFSBCOPY(nfsrv_dnsname, cp, nfsrv_dnsnamelen);
}
TAILQ_REMOVE(&hp->lughead, usrp, lug_numhash);
TAILQ_INSERT_TAIL(&hp->lughead, usrp,
lug_numhash);
mtx_unlock(&hp->mtx);
return;
}
}
mtx_unlock(&hp->mtx);
cnt++;
ret = nfsrv_getuser(RPCNFSUSERD_GETUID, uid, (gid_t)0, NULL);
if (ret == 0 && cnt < 2)
goto tryagain;
}
/*
* No match, just return a string of digits.
*/
tmp = uid;
i = 0;
while (tmp || i == 0) {
tmp /= 10;
i++;
}
len = (i > len) ? len : i;
*retlenp = len;
cp += (len - 1);
tmp = uid;
for (i = 0; i < len; i++) {
*cp-- = '0' + (tmp % 10);
tmp /= 10;
}
return;
}
/*
* Get a credential for the uid with the server's group list.
* If none is found, just return the credential passed in after
* logging a warning message.
*/
struct ucred *
nfsrv_getgrpscred(struct ucred *oldcred)
{
struct nfsusrgrp *usrp;
struct ucred *newcred;
int cnt, ret;
uid_t uid;
struct nfsrv_lughash *hp;
cnt = 0;
uid = oldcred->cr_uid;
tryagain:
if (nfsrv_dnsnamelen > 0) {
hp = NFSUSERHASH(uid);
mtx_lock(&hp->mtx);
TAILQ_FOREACH(usrp, &hp->lughead, lug_numhash) {
if (usrp->lug_uid == uid) {
if (usrp->lug_expiry < NFSD_MONOSEC)
break;
if (usrp->lug_cred != NULL) {
newcred = crhold(usrp->lug_cred);
crfree(oldcred);
} else
newcred = oldcred;
TAILQ_REMOVE(&hp->lughead, usrp, lug_numhash);
TAILQ_INSERT_TAIL(&hp->lughead, usrp,
lug_numhash);
mtx_unlock(&hp->mtx);
return (newcred);
}
}
mtx_unlock(&hp->mtx);
cnt++;
ret = nfsrv_getuser(RPCNFSUSERD_GETUID, uid, (gid_t)0, NULL);
if (ret == 0 && cnt < 2)
goto tryagain;
}
return (oldcred);
}
/*
* Convert a string to a uid.
* If no conversion is possible return NFSERR_BADOWNER, otherwise
* return 0.
* If this is called from a client side mount using AUTH_SYS and the
* string is made up entirely of digits, just convert the string to
* a number.
*/
int
nfsv4_strtouid(struct nfsrv_descript *nd, u_char *str, int len, uid_t *uidp)
{
int i;
char *cp, *endstr, *str0;
struct nfsusrgrp *usrp;
int cnt, ret;
int error = 0;
uid_t tuid;
struct nfsrv_lughash *hp, *hp2;
if (len == 0) {
error = NFSERR_BADOWNER;
goto out;
}
/* If a string of digits and an AUTH_SYS mount, just convert it. */
str0 = str;
tuid = (uid_t)strtoul(str0, &endstr, 10);
if ((endstr - str0) == len) {
/* A numeric string. */
if ((nd->nd_flag & ND_KERBV) == 0 &&
((nd->nd_flag & ND_NFSCL) != 0 ||
nfsd_enable_stringtouid != 0))
*uidp = tuid;
else
error = NFSERR_BADOWNER;
goto out;
}
/*
* Look for an '@'.
*/
cp = strchr(str0, '@');
if (cp != NULL)
i = (int)(cp++ - str0);
else
i = len;
cnt = 0;
tryagain:
if (nfsrv_dnsnamelen > 0) {
/*
* If an '@' is found and the domain name matches, search for
* the name with dns stripped off.
* Mixed case alpahbetics will match for the domain name, but
* all upper case will not.
*/
if (cnt == 0 && i < len && i > 0 &&
(len - 1 - i) == nfsrv_dnsnamelen &&
!nfsrv_cmpmixedcase(cp, nfsrv_dnsname, nfsrv_dnsnamelen)) {
len -= (nfsrv_dnsnamelen + 1);
*(cp - 1) = '\0';
}
/*
* Check for the special case of "nobody".
*/
if (len == 6 && !NFSBCMP(str, "nobody", 6)) {
*uidp = nfsrv_defaultuid;
error = 0;
goto out;
}
hp = NFSUSERNAMEHASH(str, len);
mtx_lock(&hp->mtx);
TAILQ_FOREACH(usrp, &hp->lughead, lug_namehash) {
if (usrp->lug_namelen == len &&
!NFSBCMP(usrp->lug_name, str, len)) {
if (usrp->lug_expiry < NFSD_MONOSEC)
break;
hp2 = NFSUSERHASH(usrp->lug_uid);
mtx_lock(&hp2->mtx);
TAILQ_REMOVE(&hp2->lughead, usrp, lug_numhash);
TAILQ_INSERT_TAIL(&hp2->lughead, usrp,
lug_numhash);
*uidp = usrp->lug_uid;
mtx_unlock(&hp2->mtx);
mtx_unlock(&hp->mtx);
error = 0;
goto out;
}
}
mtx_unlock(&hp->mtx);
cnt++;
ret = nfsrv_getuser(RPCNFSUSERD_GETUSER, (uid_t)0, (gid_t)0,
str);
if (ret == 0 && cnt < 2)
goto tryagain;
}
error = NFSERR_BADOWNER;
out:
NFSEXITCODE(error);
return (error);
}
/*
* Convert a gid to a string.
* gid - the group id
* cpp - points to a buffer of size NFSV4_SMALLSTR
* (malloc a larger one, as required)
* retlenp - pointer to length to be returned
*/
void
nfsv4_gidtostr(gid_t gid, u_char **cpp, int *retlenp)
{
int i;
struct nfsusrgrp *usrp;
u_char *cp = *cpp;
gid_t tmp;
int cnt, hasampersand, len = NFSV4_SMALLSTR, ret;
struct nfsrv_lughash *hp;
cnt = 0;
tryagain:
if (nfsrv_dnsnamelen > 0 && !nfs_enable_uidtostring) {
/*
* Always map nfsrv_defaultgid to "nogroup".
*/
if (gid == nfsrv_defaultgid) {
i = nfsrv_dnsnamelen + 8;
if (i > len) {
if (len > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
cp = malloc(i, M_NFSSTRING, M_WAITOK);
*cpp = cp;
len = i;
goto tryagain;
}
*retlenp = i;
NFSBCOPY("nogroup@", cp, 8);
cp += 8;
NFSBCOPY(nfsrv_dnsname, cp, nfsrv_dnsnamelen);
return;
}
hasampersand = 0;
hp = NFSGROUPHASH(gid);
mtx_lock(&hp->mtx);
TAILQ_FOREACH(usrp, &hp->lughead, lug_numhash) {
if (usrp->lug_gid == gid) {
if (usrp->lug_expiry < NFSD_MONOSEC)
break;
/*
* If the name doesn't already have an '@'
* in it, append @domainname to it.
*/
for (i = 0; i < usrp->lug_namelen; i++) {
if (usrp->lug_name[i] == '@') {
hasampersand = 1;
break;
}
}
if (hasampersand)
i = usrp->lug_namelen;
else
i = usrp->lug_namelen +
nfsrv_dnsnamelen + 1;
if (i > len) {
mtx_unlock(&hp->mtx);
if (len > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
cp = malloc(i, M_NFSSTRING, M_WAITOK);
*cpp = cp;
len = i;
goto tryagain;
}
*retlenp = i;
NFSBCOPY(usrp->lug_name, cp, usrp->lug_namelen);
if (!hasampersand) {
cp += usrp->lug_namelen;
*cp++ = '@';
NFSBCOPY(nfsrv_dnsname, cp, nfsrv_dnsnamelen);
}
TAILQ_REMOVE(&hp->lughead, usrp, lug_numhash);
TAILQ_INSERT_TAIL(&hp->lughead, usrp,
lug_numhash);
mtx_unlock(&hp->mtx);
return;
}
}
mtx_unlock(&hp->mtx);
cnt++;
ret = nfsrv_getuser(RPCNFSUSERD_GETGID, (uid_t)0, gid, NULL);
if (ret == 0 && cnt < 2)
goto tryagain;
}
/*
* No match, just return a string of digits.
*/
tmp = gid;
i = 0;
while (tmp || i == 0) {
tmp /= 10;
i++;
}
len = (i > len) ? len : i;
*retlenp = len;
cp += (len - 1);
tmp = gid;
for (i = 0; i < len; i++) {
*cp-- = '0' + (tmp % 10);
tmp /= 10;
}
return;
}
/*
* Convert a string to a gid.
* If no conversion is possible return NFSERR_BADOWNER, otherwise
* return 0.
* If this is called from a client side mount using AUTH_SYS and the
* string is made up entirely of digits, just convert the string to
* a number.
*/
int
nfsv4_strtogid(struct nfsrv_descript *nd, u_char *str, int len, gid_t *gidp)
{
int i;
char *cp, *endstr, *str0;
struct nfsusrgrp *usrp;
int cnt, ret;
int error = 0;
gid_t tgid;
struct nfsrv_lughash *hp, *hp2;
if (len == 0) {
error = NFSERR_BADOWNER;
goto out;
}
/* If a string of digits and an AUTH_SYS mount, just convert it. */
str0 = str;
tgid = (gid_t)strtoul(str0, &endstr, 10);
if ((endstr - str0) == len) {
/* A numeric string. */
if ((nd->nd_flag & ND_KERBV) == 0 &&
((nd->nd_flag & ND_NFSCL) != 0 ||
nfsd_enable_stringtouid != 0))
*gidp = tgid;
else
error = NFSERR_BADOWNER;
goto out;
}
/*
* Look for an '@'.
*/
cp = strchr(str0, '@');
if (cp != NULL)
i = (int)(cp++ - str0);
else
i = len;
cnt = 0;
tryagain:
if (nfsrv_dnsnamelen > 0) {
/*
* If an '@' is found and the dns name matches, search for the
* name with the dns stripped off.
*/
if (cnt == 0 && i < len && i > 0 &&
(len - 1 - i) == nfsrv_dnsnamelen &&
!nfsrv_cmpmixedcase(cp, nfsrv_dnsname, nfsrv_dnsnamelen)) {
len -= (nfsrv_dnsnamelen + 1);
*(cp - 1) = '\0';
}
/*
* Check for the special case of "nogroup".
*/
if (len == 7 && !NFSBCMP(str, "nogroup", 7)) {
*gidp = nfsrv_defaultgid;
error = 0;
goto out;
}
hp = NFSGROUPNAMEHASH(str, len);
mtx_lock(&hp->mtx);
TAILQ_FOREACH(usrp, &hp->lughead, lug_namehash) {
if (usrp->lug_namelen == len &&
!NFSBCMP(usrp->lug_name, str, len)) {
if (usrp->lug_expiry < NFSD_MONOSEC)
break;
hp2 = NFSGROUPHASH(usrp->lug_gid);
mtx_lock(&hp2->mtx);
TAILQ_REMOVE(&hp2->lughead, usrp, lug_numhash);
TAILQ_INSERT_TAIL(&hp2->lughead, usrp,
lug_numhash);
*gidp = usrp->lug_gid;
mtx_unlock(&hp2->mtx);
mtx_unlock(&hp->mtx);
error = 0;
goto out;
}
}
mtx_unlock(&hp->mtx);
cnt++;
ret = nfsrv_getuser(RPCNFSUSERD_GETGROUP, (uid_t)0, (gid_t)0,
str);
if (ret == 0 && cnt < 2)
goto tryagain;
}
error = NFSERR_BADOWNER;
out:
NFSEXITCODE(error);
return (error);
}
/*
* Cmp len chars, allowing mixed case in the first argument to match lower
* case in the second, but not if the first argument is all upper case.
* Return 0 for a match, 1 otherwise.
*/
static int
nfsrv_cmpmixedcase(u_char *cp, u_char *cp2, int len)
{
int i;
u_char tmp;
int fndlower = 0;
for (i = 0; i < len; i++) {
if (*cp >= 'A' && *cp <= 'Z') {
tmp = *cp++ + ('a' - 'A');
} else {
tmp = *cp++;
if (tmp >= 'a' && tmp <= 'z')
fndlower = 1;
}
if (tmp != *cp2++)
return (1);
}
if (fndlower)
return (0);
else
return (1);
}
/*
* Set the port for the nfsuserd.
*/
int
nfsrv_nfsuserdport(struct nfsuserd_args *nargs, NFSPROC_T *p)
{
struct nfssockreq *rp;
#ifdef INET
struct sockaddr_in *ad;
#endif
#ifdef INET6
struct sockaddr_in6 *ad6;
const struct in6_addr in6loopback = IN6ADDR_LOOPBACK_INIT;
#endif
int error;
NFSLOCKNAMEID();
if (nfsrv_nfsuserd != NOTRUNNING) {
NFSUNLOCKNAMEID();
error = EPERM;
goto out;
}
nfsrv_nfsuserd = STARTSTOP;
/*
* Set up the socket record and connect.
* Set nr_client NULL before unlocking, just to ensure that no other
* process/thread/core will use a bogus old value. This could only
* occur if the use of the nameid lock to protect nfsrv_nfsuserd is
* broken.
*/
rp = &nfsrv_nfsuserdsock;
rp->nr_client = NULL;
NFSUNLOCKNAMEID();
rp->nr_sotype = SOCK_DGRAM;
rp->nr_soproto = IPPROTO_UDP;
rp->nr_lock = (NFSR_RESERVEDPORT | NFSR_LOCALHOST);
rp->nr_cred = NULL;
rp->nr_prog = RPCPROG_NFSUSERD;
error = 0;
switch (nargs->nuserd_family) {
#ifdef INET
case AF_INET:
rp->nr_nam = malloc(sizeof(struct sockaddr_in), M_SONAME,
M_WAITOK | M_ZERO);
ad = (struct sockaddr_in *)rp->nr_nam;
ad->sin_len = sizeof(struct sockaddr_in);
ad->sin_family = AF_INET;
ad->sin_addr.s_addr = htonl(INADDR_LOOPBACK);
ad->sin_port = nargs->nuserd_port;
break;
#endif
#ifdef INET6
case AF_INET6:
rp->nr_nam = malloc(sizeof(struct sockaddr_in6), M_SONAME,
M_WAITOK | M_ZERO);
ad6 = (struct sockaddr_in6 *)rp->nr_nam;
ad6->sin6_len = sizeof(struct sockaddr_in6);
ad6->sin6_family = AF_INET6;
ad6->sin6_addr = in6loopback;
ad6->sin6_port = nargs->nuserd_port;
break;
#endif
default:
error = ENXIO;
}
rp->nr_vers = RPCNFSUSERD_VERS;
if (error == 0)
error = newnfs_connect(NULL, rp, NFSPROCCRED(p), p, 0, false,
&rp->nr_client);
if (error == 0) {
NFSLOCKNAMEID();
nfsrv_nfsuserd = RUNNING;
NFSUNLOCKNAMEID();
} else {
free(rp->nr_nam, M_SONAME);
NFSLOCKNAMEID();
nfsrv_nfsuserd = NOTRUNNING;
NFSUNLOCKNAMEID();
}
out:
NFSEXITCODE(error);
return (error);
}
/*
* Delete the nfsuserd port.
*/
void
nfsrv_nfsuserddelport(void)
{
NFSLOCKNAMEID();
if (nfsrv_nfsuserd != RUNNING) {
NFSUNLOCKNAMEID();
return;
}
nfsrv_nfsuserd = STARTSTOP;
/* Wait for all upcalls to complete. */
while (nfsrv_userdupcalls > 0)
msleep(&nfsrv_userdupcalls, NFSNAMEIDMUTEXPTR, PVFS,
"nfsupcalls", 0);
NFSUNLOCKNAMEID();
newnfs_disconnect(NULL, &nfsrv_nfsuserdsock);
free(nfsrv_nfsuserdsock.nr_nam, M_SONAME);
NFSLOCKNAMEID();
nfsrv_nfsuserd = NOTRUNNING;
NFSUNLOCKNAMEID();
}
/*
* Do upcalls to the nfsuserd, for cache misses of the owner/ownergroup
* name<-->id cache.
* Returns 0 upon success, non-zero otherwise.
*/
static int
nfsrv_getuser(int procnum, uid_t uid, gid_t gid, char *name)
{
u_int32_t *tl;
struct nfsrv_descript *nd;
int len;
struct nfsrv_descript nfsd;
struct ucred *cred;
int error;
NFSLOCKNAMEID();
if (nfsrv_nfsuserd != RUNNING) {
NFSUNLOCKNAMEID();
error = EPERM;
goto out;
}
/*
* Maintain a count of upcalls in progress, so that nfsrv_X()
* can wait until no upcalls are in progress.
*/
nfsrv_userdupcalls++;
NFSUNLOCKNAMEID();
KASSERT(nfsrv_userdupcalls > 0,
("nfsrv_getuser: non-positive upcalls"));
nd = &nfsd;
cred = newnfs_getcred();
nd->nd_flag = ND_GSSINITREPLY;
nfsrvd_rephead(nd);
nd->nd_procnum = procnum;
if (procnum == RPCNFSUSERD_GETUID || procnum == RPCNFSUSERD_GETGID) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
if (procnum == RPCNFSUSERD_GETUID)
*tl = txdr_unsigned(uid);
else
*tl = txdr_unsigned(gid);
} else {
len = strlen(name);
(void) nfsm_strtom(nd, name, len);
}
error = newnfs_request(nd, NULL, NULL, &nfsrv_nfsuserdsock, NULL, NULL,
cred, RPCPROG_NFSUSERD, RPCNFSUSERD_VERS, NULL, 0, NULL, NULL);
NFSLOCKNAMEID();
if (--nfsrv_userdupcalls == 0 && nfsrv_nfsuserd == STARTSTOP)
wakeup(&nfsrv_userdupcalls);
NFSUNLOCKNAMEID();
NFSFREECRED(cred);
if (!error) {
m_freem(nd->nd_mrep);
error = nd->nd_repstat;
}
out:
NFSEXITCODE(error);
return (error);
}
/*
* This function is called from the nfssvc(2) system call, to update the
* kernel user/group name list(s) for the V4 owner and ownergroup attributes.
*/
int
nfssvc_idname(struct nfsd_idargs *nidp)
{
struct nfsusrgrp *nusrp, *usrp, *newusrp;
struct nfsrv_lughash *hp_name, *hp_idnum, *thp;
int i, group_locked, groupname_locked, user_locked, username_locked;
int error = 0;
u_char *cp;
gid_t *grps;
struct ucred *cr;
static int onethread = 0;
static time_t lasttime = 0;
if (nidp->nid_namelen <= 0 || nidp->nid_namelen > MAXHOSTNAMELEN) {
error = EINVAL;
goto out;
}
if (nidp->nid_flag & NFSID_INITIALIZE) {
cp = malloc(nidp->nid_namelen + 1, M_NFSSTRING, M_WAITOK);
error = copyin(nidp->nid_name, cp, nidp->nid_namelen);
if (error != 0) {
free(cp, M_NFSSTRING);
goto out;
}
if (atomic_cmpset_acq_int(&nfsrv_dnsnamelen, 0, 0) == 0) {
/*
* Free up all the old stuff and reinitialize hash
* lists. All mutexes for both lists must be locked,
* with the user/group name ones before the uid/gid
* ones, to avoid a LOR.
*/
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_lock(&nfsusernamehash[i].mtx);
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_lock(&nfsuserhash[i].mtx);
for (i = 0; i < nfsrv_lughashsize; i++)
TAILQ_FOREACH_SAFE(usrp,
&nfsuserhash[i].lughead, lug_numhash, nusrp)
nfsrv_removeuser(usrp, 1);
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_unlock(&nfsuserhash[i].mtx);
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_unlock(&nfsusernamehash[i].mtx);
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_lock(&nfsgroupnamehash[i].mtx);
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_lock(&nfsgrouphash[i].mtx);
for (i = 0; i < nfsrv_lughashsize; i++)
TAILQ_FOREACH_SAFE(usrp,
&nfsgrouphash[i].lughead, lug_numhash,
nusrp)
nfsrv_removeuser(usrp, 0);
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_unlock(&nfsgrouphash[i].mtx);
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_unlock(&nfsgroupnamehash[i].mtx);
free(nfsrv_dnsname, M_NFSSTRING);
nfsrv_dnsname = NULL;
}
if (nfsuserhash == NULL) {
/* Allocate the hash tables. */
nfsuserhash = malloc(sizeof(struct nfsrv_lughash) *
nfsrv_lughashsize, M_NFSUSERGROUP, M_WAITOK |
M_ZERO);
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_init(&nfsuserhash[i].mtx, "nfsuidhash",
NULL, MTX_DEF | MTX_DUPOK);
nfsusernamehash = malloc(sizeof(struct nfsrv_lughash) *
nfsrv_lughashsize, M_NFSUSERGROUP, M_WAITOK |
M_ZERO);
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_init(&nfsusernamehash[i].mtx,
"nfsusrhash", NULL, MTX_DEF |
MTX_DUPOK);
nfsgrouphash = malloc(sizeof(struct nfsrv_lughash) *
nfsrv_lughashsize, M_NFSUSERGROUP, M_WAITOK |
M_ZERO);
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_init(&nfsgrouphash[i].mtx, "nfsgidhash",
NULL, MTX_DEF | MTX_DUPOK);
nfsgroupnamehash = malloc(sizeof(struct nfsrv_lughash) *
nfsrv_lughashsize, M_NFSUSERGROUP, M_WAITOK |
M_ZERO);
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_init(&nfsgroupnamehash[i].mtx,
"nfsgrphash", NULL, MTX_DEF | MTX_DUPOK);
}
/* (Re)initialize the list heads. */
for (i = 0; i < nfsrv_lughashsize; i++)
TAILQ_INIT(&nfsuserhash[i].lughead);
for (i = 0; i < nfsrv_lughashsize; i++)
TAILQ_INIT(&nfsusernamehash[i].lughead);
for (i = 0; i < nfsrv_lughashsize; i++)
TAILQ_INIT(&nfsgrouphash[i].lughead);
for (i = 0; i < nfsrv_lughashsize; i++)
TAILQ_INIT(&nfsgroupnamehash[i].lughead);
/*
* Put name in "DNS" string.
*/
nfsrv_dnsname = cp;
nfsrv_defaultuid = nidp->nid_uid;
nfsrv_defaultgid = nidp->nid_gid;
nfsrv_usercnt = 0;
nfsrv_usermax = nidp->nid_usermax;
atomic_store_rel_int(&nfsrv_dnsnamelen, nidp->nid_namelen);
goto out;
}
/*
* malloc the new one now, so any potential sleep occurs before
* manipulation of the lists.
*/
newusrp = malloc(sizeof(struct nfsusrgrp) + nidp->nid_namelen,
M_NFSUSERGROUP, M_WAITOK | M_ZERO);
error = copyin(nidp->nid_name, newusrp->lug_name,
nidp->nid_namelen);
if (error == 0 && nidp->nid_ngroup > 0 &&
(nidp->nid_flag & NFSID_ADDUID) != 0) {
grps = malloc(sizeof(gid_t) * nidp->nid_ngroup, M_TEMP,
M_WAITOK);
error = copyin(nidp->nid_grps, grps,
sizeof(gid_t) * nidp->nid_ngroup);
if (error == 0) {
/*
* Create a credential just like svc_getcred(),
* but using the group list provided.
*/
cr = crget();
cr->cr_uid = cr->cr_ruid = cr->cr_svuid = nidp->nid_uid;
crsetgroups(cr, nidp->nid_ngroup, grps);
cr->cr_rgid = cr->cr_svgid = cr->cr_groups[0];
cr->cr_prison = &prison0;
prison_hold(cr->cr_prison);
#ifdef MAC
mac_cred_associate_nfsd(cr);
#endif
newusrp->lug_cred = cr;
}
free(grps, M_TEMP);
}
if (error) {
free(newusrp, M_NFSUSERGROUP);
goto out;
}
newusrp->lug_namelen = nidp->nid_namelen;
/*
* The lock order is username[0]->[nfsrv_lughashsize - 1] followed
* by uid[0]->[nfsrv_lughashsize - 1], with the same for group.
* The flags user_locked, username_locked, group_locked and
* groupname_locked are set to indicate all of those hash lists are
* locked. hp_name != NULL and hp_idnum != NULL indicates that
* the respective one mutex is locked.
*/
user_locked = username_locked = group_locked = groupname_locked = 0;
hp_name = hp_idnum = NULL;
/*
* Delete old entries, as required.
*/
if (nidp->nid_flag & (NFSID_DELUID | NFSID_ADDUID)) {
/* Must lock all username hash lists first, to avoid a LOR. */
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_lock(&nfsusernamehash[i].mtx);
username_locked = 1;
hp_idnum = NFSUSERHASH(nidp->nid_uid);
mtx_lock(&hp_idnum->mtx);
TAILQ_FOREACH_SAFE(usrp, &hp_idnum->lughead, lug_numhash,
nusrp) {
if (usrp->lug_uid == nidp->nid_uid)
nfsrv_removeuser(usrp, 1);
}
} else if (nidp->nid_flag & (NFSID_DELUSERNAME | NFSID_ADDUSERNAME)) {
hp_name = NFSUSERNAMEHASH(newusrp->lug_name,
newusrp->lug_namelen);
mtx_lock(&hp_name->mtx);
TAILQ_FOREACH_SAFE(usrp, &hp_name->lughead, lug_namehash,
nusrp) {
if (usrp->lug_namelen == newusrp->lug_namelen &&
!NFSBCMP(usrp->lug_name, newusrp->lug_name,
usrp->lug_namelen)) {
thp = NFSUSERHASH(usrp->lug_uid);
mtx_lock(&thp->mtx);
nfsrv_removeuser(usrp, 1);
mtx_unlock(&thp->mtx);
}
}
hp_idnum = NFSUSERHASH(nidp->nid_uid);
mtx_lock(&hp_idnum->mtx);
} else if (nidp->nid_flag & (NFSID_DELGID | NFSID_ADDGID)) {
/* Must lock all groupname hash lists first, to avoid a LOR. */
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_lock(&nfsgroupnamehash[i].mtx);
groupname_locked = 1;
hp_idnum = NFSGROUPHASH(nidp->nid_gid);
mtx_lock(&hp_idnum->mtx);
TAILQ_FOREACH_SAFE(usrp, &hp_idnum->lughead, lug_numhash,
nusrp) {
if (usrp->lug_gid == nidp->nid_gid)
nfsrv_removeuser(usrp, 0);
}
} else if (nidp->nid_flag & (NFSID_DELGROUPNAME | NFSID_ADDGROUPNAME)) {
hp_name = NFSGROUPNAMEHASH(newusrp->lug_name,
newusrp->lug_namelen);
mtx_lock(&hp_name->mtx);
TAILQ_FOREACH_SAFE(usrp, &hp_name->lughead, lug_namehash,
nusrp) {
if (usrp->lug_namelen == newusrp->lug_namelen &&
!NFSBCMP(usrp->lug_name, newusrp->lug_name,
usrp->lug_namelen)) {
thp = NFSGROUPHASH(usrp->lug_gid);
mtx_lock(&thp->mtx);
nfsrv_removeuser(usrp, 0);
mtx_unlock(&thp->mtx);
}
}
hp_idnum = NFSGROUPHASH(nidp->nid_gid);
mtx_lock(&hp_idnum->mtx);
}
/*
* Now, we can add the new one.
*/
if (nidp->nid_usertimeout)
newusrp->lug_expiry = NFSD_MONOSEC + nidp->nid_usertimeout;
else
newusrp->lug_expiry = NFSD_MONOSEC + 5;
if (nidp->nid_flag & (NFSID_ADDUID | NFSID_ADDUSERNAME)) {
newusrp->lug_uid = nidp->nid_uid;
thp = NFSUSERHASH(newusrp->lug_uid);
mtx_assert(&thp->mtx, MA_OWNED);
TAILQ_INSERT_TAIL(&thp->lughead, newusrp, lug_numhash);
thp = NFSUSERNAMEHASH(newusrp->lug_name, newusrp->lug_namelen);
mtx_assert(&thp->mtx, MA_OWNED);
TAILQ_INSERT_TAIL(&thp->lughead, newusrp, lug_namehash);
atomic_add_int(&nfsrv_usercnt, 1);
} else if (nidp->nid_flag & (NFSID_ADDGID | NFSID_ADDGROUPNAME)) {
newusrp->lug_gid = nidp->nid_gid;
thp = NFSGROUPHASH(newusrp->lug_gid);
mtx_assert(&thp->mtx, MA_OWNED);
TAILQ_INSERT_TAIL(&thp->lughead, newusrp, lug_numhash);
thp = NFSGROUPNAMEHASH(newusrp->lug_name, newusrp->lug_namelen);
mtx_assert(&thp->mtx, MA_OWNED);
TAILQ_INSERT_TAIL(&thp->lughead, newusrp, lug_namehash);
atomic_add_int(&nfsrv_usercnt, 1);
} else {
if (newusrp->lug_cred != NULL)
crfree(newusrp->lug_cred);
free(newusrp, M_NFSUSERGROUP);
}
/*
* Once per second, allow one thread to trim the cache.
*/
if (lasttime < NFSD_MONOSEC &&
atomic_cmpset_acq_int(&onethread, 0, 1) != 0) {
/*
* First, unlock the single mutexes, so that all entries
* can be locked and any LOR is avoided.
*/
if (hp_name != NULL) {
mtx_unlock(&hp_name->mtx);
hp_name = NULL;
}
if (hp_idnum != NULL) {
mtx_unlock(&hp_idnum->mtx);
hp_idnum = NULL;
}
if ((nidp->nid_flag & (NFSID_DELUID | NFSID_ADDUID |
NFSID_DELUSERNAME | NFSID_ADDUSERNAME)) != 0) {
if (username_locked == 0) {
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_lock(&nfsusernamehash[i].mtx);
username_locked = 1;
}
KASSERT(user_locked == 0,
("nfssvc_idname: user_locked"));
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_lock(&nfsuserhash[i].mtx);
user_locked = 1;
for (i = 0; i < nfsrv_lughashsize; i++) {
TAILQ_FOREACH_SAFE(usrp,
&nfsuserhash[i].lughead, lug_numhash,
nusrp)
if (usrp->lug_expiry < NFSD_MONOSEC)
nfsrv_removeuser(usrp, 1);
}
for (i = 0; i < nfsrv_lughashsize; i++) {
/*
* Trim the cache using an approximate LRU
* algorithm. This code deletes the least
* recently used entry on each hash list.
*/
if (nfsrv_usercnt <= nfsrv_usermax)
break;
usrp = TAILQ_FIRST(&nfsuserhash[i].lughead);
if (usrp != NULL)
nfsrv_removeuser(usrp, 1);
}
} else {
if (groupname_locked == 0) {
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_lock(&nfsgroupnamehash[i].mtx);
groupname_locked = 1;
}
KASSERT(group_locked == 0,
("nfssvc_idname: group_locked"));
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_lock(&nfsgrouphash[i].mtx);
group_locked = 1;
for (i = 0; i < nfsrv_lughashsize; i++) {
TAILQ_FOREACH_SAFE(usrp,
&nfsgrouphash[i].lughead, lug_numhash,
nusrp)
if (usrp->lug_expiry < NFSD_MONOSEC)
nfsrv_removeuser(usrp, 0);
}
for (i = 0; i < nfsrv_lughashsize; i++) {
/*
* Trim the cache using an approximate LRU
* algorithm. This code deletes the least
* recently user entry on each hash list.
*/
if (nfsrv_usercnt <= nfsrv_usermax)
break;
usrp = TAILQ_FIRST(&nfsgrouphash[i].lughead);
if (usrp != NULL)
nfsrv_removeuser(usrp, 0);
}
}
lasttime = NFSD_MONOSEC;
atomic_store_rel_int(&onethread, 0);
}
/* Now, unlock all locked mutexes. */
if (hp_idnum != NULL)
mtx_unlock(&hp_idnum->mtx);
if (hp_name != NULL)
mtx_unlock(&hp_name->mtx);
if (user_locked != 0)
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_unlock(&nfsuserhash[i].mtx);
if (username_locked != 0)
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_unlock(&nfsusernamehash[i].mtx);
if (group_locked != 0)
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_unlock(&nfsgrouphash[i].mtx);
if (groupname_locked != 0)
for (i = 0; i < nfsrv_lughashsize; i++)
mtx_unlock(&nfsgroupnamehash[i].mtx);
out:
NFSEXITCODE(error);
return (error);
}
/*
* Remove a user/group name element.
*/
static void
nfsrv_removeuser(struct nfsusrgrp *usrp, int isuser)
{
struct nfsrv_lughash *hp;
if (isuser != 0) {
hp = NFSUSERHASH(usrp->lug_uid);
mtx_assert(&hp->mtx, MA_OWNED);
TAILQ_REMOVE(&hp->lughead, usrp, lug_numhash);
hp = NFSUSERNAMEHASH(usrp->lug_name, usrp->lug_namelen);
mtx_assert(&hp->mtx, MA_OWNED);
TAILQ_REMOVE(&hp->lughead, usrp, lug_namehash);
} else {
hp = NFSGROUPHASH(usrp->lug_gid);
mtx_assert(&hp->mtx, MA_OWNED);
TAILQ_REMOVE(&hp->lughead, usrp, lug_numhash);
hp = NFSGROUPNAMEHASH(usrp->lug_name, usrp->lug_namelen);
mtx_assert(&hp->mtx, MA_OWNED);
TAILQ_REMOVE(&hp->lughead, usrp, lug_namehash);
}
atomic_add_int(&nfsrv_usercnt, -1);
if (usrp->lug_cred != NULL)
crfree(usrp->lug_cred);
free(usrp, M_NFSUSERGROUP);
}
/*
* Free up all the allocations related to the name<-->id cache.
* This function should only be called when the nfsuserd daemon isn't
* running, since it doesn't do any locking.
* This function is meant to be used when the nfscommon module is unloaded.
*/
void
nfsrv_cleanusergroup(void)
{
struct nfsrv_lughash *hp, *hp2;
struct nfsusrgrp *nusrp, *usrp;
int i;
if (nfsuserhash == NULL)
return;
for (i = 0; i < nfsrv_lughashsize; i++) {
hp = &nfsuserhash[i];
TAILQ_FOREACH_SAFE(usrp, &hp->lughead, lug_numhash, nusrp) {
TAILQ_REMOVE(&hp->lughead, usrp, lug_numhash);
hp2 = NFSUSERNAMEHASH(usrp->lug_name,
usrp->lug_namelen);
TAILQ_REMOVE(&hp2->lughead, usrp, lug_namehash);
if (usrp->lug_cred != NULL)
crfree(usrp->lug_cred);
free(usrp, M_NFSUSERGROUP);
}
hp = &nfsgrouphash[i];
TAILQ_FOREACH_SAFE(usrp, &hp->lughead, lug_numhash, nusrp) {
TAILQ_REMOVE(&hp->lughead, usrp, lug_numhash);
hp2 = NFSGROUPNAMEHASH(usrp->lug_name,
usrp->lug_namelen);
TAILQ_REMOVE(&hp2->lughead, usrp, lug_namehash);
if (usrp->lug_cred != NULL)
crfree(usrp->lug_cred);
free(usrp, M_NFSUSERGROUP);
}
mtx_destroy(&nfsuserhash[i].mtx);
mtx_destroy(&nfsusernamehash[i].mtx);
mtx_destroy(&nfsgroupnamehash[i].mtx);
mtx_destroy(&nfsgrouphash[i].mtx);
}
free(nfsuserhash, M_NFSUSERGROUP);
free(nfsusernamehash, M_NFSUSERGROUP);
free(nfsgrouphash, M_NFSUSERGROUP);
free(nfsgroupnamehash, M_NFSUSERGROUP);
free(nfsrv_dnsname, M_NFSSTRING);
}
/*
* This function scans a byte string and checks for UTF-8 compliance.
* It returns 0 if it conforms and NFSERR_INVAL if not.
*/
int
nfsrv_checkutf8(u_int8_t *cp, int len)
{
u_int32_t val = 0x0;
int cnt = 0, gotd = 0, shift = 0;
u_int8_t byte;
static int utf8_shift[5] = { 7, 11, 16, 21, 26 };
int error = 0;
/*
* Here are what the variables are used for:
* val - the calculated value of a multibyte char, used to check
* that it was coded with the correct range
* cnt - the number of 10xxxxxx bytes to follow
* gotd - set for a char of Dxxx, so D800<->DFFF can be checked for
* shift - lower order bits of range (ie. "val >> shift" should
* not be 0, in other words, dividing by the lower bound
* of the range should get a non-zero value)
* byte - used to calculate cnt
*/
while (len > 0) {
if (cnt > 0) {
/* This handles the 10xxxxxx bytes */
if ((*cp & 0xc0) != 0x80 ||
(gotd && (*cp & 0x20))) {
error = NFSERR_INVAL;
goto out;
}
gotd = 0;
val <<= 6;
val |= (*cp & 0x3f);
cnt--;
if (cnt == 0 && (val >> shift) == 0x0) {
error = NFSERR_INVAL;
goto out;
}
} else if (*cp & 0x80) {
/* first byte of multi byte char */
byte = *cp;
while ((byte & 0x40) && cnt < 6) {
cnt++;
byte <<= 1;
}
if (cnt == 0 || cnt == 6) {
error = NFSERR_INVAL;
goto out;
}
val = (*cp & (0x3f >> cnt));
shift = utf8_shift[cnt - 1];
if (cnt == 2 && val == 0xd)
/* Check for the 0xd800-0xdfff case */
gotd = 1;
}
cp++;
len--;
}
if (cnt > 0)
error = NFSERR_INVAL;
out:
NFSEXITCODE(error);
return (error);
}
/*
* Parse the xdr for an NFSv4 FsLocations attribute. Return two malloc'd
* strings, one with the root path in it and the other with the list of
* locations. The list is in the same format as is found in nfr_refs.
* It is a "," separated list of entries, where each of them is of the
* form <server>:<rootpath>. For example
* "nfsv4-test:/sub2,nfsv4-test2:/user/mnt,nfsv4-test2:/user/mnt2"
* The nilp argument is set to 1 for the special case of a null fs_root
* and an empty server list.
* It returns NFSERR_BADXDR, if the xdr can't be parsed and returns the
* number of xdr bytes parsed in sump.
*/
static int
nfsrv_getrefstr(struct nfsrv_descript *nd, u_char **fsrootp, u_char **srvp,
int *sump, int *nilp)
{
u_int32_t *tl;
u_char *cp = NULL, *cp2 = NULL, *cp3, *str;
int i, j, len, stringlen, cnt, slen, siz, xdrsum, error = 0, nsrv;
struct list {
SLIST_ENTRY(list) next;
int len;
u_char host[1];
} *lsp, *nlsp;
SLIST_HEAD(, list) head;
*fsrootp = NULL;
*srvp = NULL;
*nilp = 0;
/*
* Get the fs_root path and check for the special case of null path
* and 0 length server list.
*/
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
len = fxdr_unsigned(int, *tl);
if (len < 0 || len > 10240) {
error = NFSERR_BADXDR;
goto nfsmout;
}
if (len == 0) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (*tl != 0) {
error = NFSERR_BADXDR;
goto nfsmout;
}
*nilp = 1;
*sump = 2 * NFSX_UNSIGNED;
error = 0;
goto nfsmout;
}
cp = malloc(len + 1, M_NFSSTRING, M_WAITOK);
error = nfsrv_mtostr(nd, cp, len);
if (!error) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
cnt = fxdr_unsigned(int, *tl);
if (cnt <= 0)
error = NFSERR_BADXDR;
}
if (error)
goto nfsmout;
/*
* Now, loop through the location list and make up the srvlist.
*/
xdrsum = (2 * NFSX_UNSIGNED) + NFSM_RNDUP(len);
cp2 = cp3 = malloc(1024, M_NFSSTRING, M_WAITOK);
slen = 1024;
siz = 0;
for (i = 0; i < cnt; i++) {
SLIST_INIT(&head);
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
nsrv = fxdr_unsigned(int, *tl);
if (nsrv <= 0) {
error = NFSERR_BADXDR;
goto nfsmout;
}
/*
* Handle the first server by putting it in the srvstr.
*/
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
len = fxdr_unsigned(int, *tl);
if (len <= 0 || len > 1024) {
error = NFSERR_BADXDR;
goto nfsmout;
}
nfsrv_refstrbigenough(siz + len + 3, &cp2, &cp3, &slen);
if (cp3 != cp2) {
*cp3++ = ',';
siz++;
}
error = nfsrv_mtostr(nd, cp3, len);
if (error)
goto nfsmout;
cp3 += len;
*cp3++ = ':';
siz += (len + 1);
xdrsum += (2 * NFSX_UNSIGNED) + NFSM_RNDUP(len);
for (j = 1; j < nsrv; j++) {
/*
* Yuck, put them in an slist and process them later.
*/
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
len = fxdr_unsigned(int, *tl);
if (len <= 0 || len > 1024) {
error = NFSERR_BADXDR;
goto nfsmout;
}
lsp = (struct list *)malloc(sizeof (struct list)
+ len, M_TEMP, M_WAITOK);
error = nfsrv_mtostr(nd, lsp->host, len);
if (error)
goto nfsmout;
xdrsum += NFSX_UNSIGNED + NFSM_RNDUP(len);
lsp->len = len;
SLIST_INSERT_HEAD(&head, lsp, next);
}
/*
* Finally, we can get the path.
*/
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
len = fxdr_unsigned(int, *tl);
if (len <= 0 || len > 1024) {
error = NFSERR_BADXDR;
goto nfsmout;
}
nfsrv_refstrbigenough(siz + len + 1, &cp2, &cp3, &slen);
error = nfsrv_mtostr(nd, cp3, len);
if (error)
goto nfsmout;
xdrsum += NFSX_UNSIGNED + NFSM_RNDUP(len);
str = cp3;
stringlen = len;
cp3 += len;
siz += len;
SLIST_FOREACH_SAFE(lsp, &head, next, nlsp) {
nfsrv_refstrbigenough(siz + lsp->len + stringlen + 3,
&cp2, &cp3, &slen);
*cp3++ = ',';
NFSBCOPY(lsp->host, cp3, lsp->len);
cp3 += lsp->len;
*cp3++ = ':';
NFSBCOPY(str, cp3, stringlen);
cp3 += stringlen;
*cp3 = '\0';
siz += (lsp->len + stringlen + 2);
free(lsp, M_TEMP);
}
}
*fsrootp = cp;
*srvp = cp2;
*sump = xdrsum;
NFSEXITCODE2(0, nd);
return (0);
nfsmout:
if (cp != NULL)
free(cp, M_NFSSTRING);
if (cp2 != NULL)
free(cp2, M_NFSSTRING);
NFSEXITCODE2(error, nd);
return (error);
}
/*
* Make the malloc'd space large enough. This is a pain, but the xdr
* doesn't set an upper bound on the side, so...
*/
static void
nfsrv_refstrbigenough(int siz, u_char **cpp, u_char **cpp2, int *slenp)
{
u_char *cp;
int i;
if (siz <= *slenp)
return;
cp = malloc(siz + 1024, M_NFSSTRING, M_WAITOK);
NFSBCOPY(*cpp, cp, *slenp);
free(*cpp, M_NFSSTRING);
i = *cpp2 - *cpp;
*cpp = cp;
*cpp2 = cp + i;
*slenp = siz + 1024;
}
/*
* Initialize the reply header data structures.
*/
void
nfsrvd_rephead(struct nfsrv_descript *nd)
{
struct mbuf *mreq;
if ((nd->nd_flag & ND_EXTPG) != 0) {
mreq = mb_alloc_ext_plus_pages(PAGE_SIZE, M_WAITOK);
nd->nd_mreq = nd->nd_mb = mreq;
nd->nd_bpos = (char *)(void *)
PHYS_TO_DMAP(mreq->m_epg_pa[0]);
nd->nd_bextpg = 0;
nd->nd_bextpgsiz = PAGE_SIZE;
} else {
/*
* If this is a big reply, use a cluster.
*/
if ((nd->nd_flag & ND_GSSINITREPLY) == 0 &&
nfs_bigreply[nd->nd_procnum]) {
NFSMCLGET(mreq, M_WAITOK);
nd->nd_mreq = mreq;
nd->nd_mb = mreq;
} else {
NFSMGET(mreq);
nd->nd_mreq = mreq;
nd->nd_mb = mreq;
}
nd->nd_bpos = mtod(mreq, char *);
mreq->m_len = 0;
}
if ((nd->nd_flag & ND_GSSINITREPLY) == 0)
NFSM_BUILD(nd->nd_errp, int *, NFSX_UNSIGNED);
}
/*
* Lock a socket against others.
* Currently used to serialize connect/disconnect attempts.
*/
int
newnfs_sndlock(int *flagp)
{
struct timespec ts;
NFSLOCKSOCK();
while (*flagp & NFSR_SNDLOCK) {
*flagp |= NFSR_WANTSND;
ts.tv_sec = 0;
ts.tv_nsec = 0;
(void) nfsmsleep((caddr_t)flagp, NFSSOCKMUTEXPTR,
PZERO - 1, "nfsndlck", &ts);
}
*flagp |= NFSR_SNDLOCK;
NFSUNLOCKSOCK();
return (0);
}
/*
* Unlock the stream socket for others.
*/
void
newnfs_sndunlock(int *flagp)
{
NFSLOCKSOCK();
if ((*flagp & NFSR_SNDLOCK) == 0)
panic("nfs sndunlock");
*flagp &= ~NFSR_SNDLOCK;
if (*flagp & NFSR_WANTSND) {
*flagp &= ~NFSR_WANTSND;
wakeup((caddr_t)flagp);
}
NFSUNLOCKSOCK();
}
int
nfsv4_getipaddr(struct nfsrv_descript *nd, struct sockaddr_in *sin,
struct sockaddr_in6 *sin6, sa_family_t *saf, int *isudp)
{
struct in_addr saddr;
uint32_t portnum, *tl;
int i, j, k;
sa_family_t af = AF_UNSPEC;
char addr[64], protocol[5], *cp;
int cantparse = 0, error = 0;
uint16_t portv;
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
i = fxdr_unsigned(int, *tl);
if (i >= 3 && i <= 4) {
error = nfsrv_mtostr(nd, protocol, i);
if (error)
goto nfsmout;
if (strcmp(protocol, "tcp") == 0) {
af = AF_INET;
*isudp = 0;
} else if (strcmp(protocol, "udp") == 0) {
af = AF_INET;
*isudp = 1;
} else if (strcmp(protocol, "tcp6") == 0) {
af = AF_INET6;
*isudp = 0;
} else if (strcmp(protocol, "udp6") == 0) {
af = AF_INET6;
*isudp = 1;
} else
cantparse = 1;
} else {
cantparse = 1;
if (i > 0) {
error = nfsm_advance(nd, NFSM_RNDUP(i), -1);
if (error)
goto nfsmout;
}
}
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
i = fxdr_unsigned(int, *tl);
if (i < 0) {
error = NFSERR_BADXDR;
goto nfsmout;
} else if (cantparse == 0 && i >= 11 && i < 64) {
/*
* The shortest address is 11chars and the longest is < 64.
*/
error = nfsrv_mtostr(nd, addr, i);
if (error)
goto nfsmout;
/* Find the port# at the end and extract that. */
i = strlen(addr);
k = 0;
cp = &addr[i - 1];
/* Count back two '.'s from end to get port# field. */
for (j = 0; j < i; j++) {
if (*cp == '.') {
k++;
if (k == 2)
break;
}
cp--;
}
if (k == 2) {
/*
* The NFSv4 port# is appended as .N.N, where N is
* a decimal # in the range 0-255, just like an inet4
* address. Cheat and use inet_aton(), which will
* return a Class A address and then shift the high
* order 8bits over to convert it to the port#.
*/
*cp++ = '\0';
if (inet_aton(cp, &saddr) == 1) {
portnum = ntohl(saddr.s_addr);
portv = (uint16_t)((portnum >> 16) |
(portnum & 0xff));
} else
cantparse = 1;
} else
cantparse = 1;
if (cantparse == 0) {
if (af == AF_INET) {
if (inet_pton(af, addr, &sin->sin_addr) == 1) {
sin->sin_len = sizeof(*sin);
sin->sin_family = AF_INET;
sin->sin_port = htons(portv);
*saf = af;
return (0);
}
} else {
if (inet_pton(af, addr, &sin6->sin6_addr)
== 1) {
sin6->sin6_len = sizeof(*sin6);
sin6->sin6_family = AF_INET6;
sin6->sin6_port = htons(portv);
*saf = af;
return (0);
}
}
}
} else {
if (i > 0) {
error = nfsm_advance(nd, NFSM_RNDUP(i), -1);
if (error)
goto nfsmout;
}
}
error = EPERM;
nfsmout:
return (error);
}
/*
* Handle an NFSv4.1 Sequence request for the session.
* If reply != NULL, use it to return the cached reply, as required.
* The client gets a cached reply via this call for callbacks, however the
* server gets a cached reply via the nfsv4_seqsess_cacherep() call.
*/
int
nfsv4_seqsession(uint32_t seqid, uint32_t slotid, uint32_t highslot,
struct nfsslot *slots, struct mbuf **reply, uint16_t maxslot)
{
struct mbuf *m;
int error;
error = 0;
if (reply != NULL)
*reply = NULL;
if (slotid > maxslot)
return (NFSERR_BADSLOT);
if (seqid == slots[slotid].nfssl_seq) {
/* A retry. */
if (slots[slotid].nfssl_inprog != 0)
error = NFSERR_DELAY;
else if (slots[slotid].nfssl_reply != NULL) {
if (reply != NULL) {
m = m_copym(slots[slotid].nfssl_reply, 0,
M_COPYALL, M_NOWAIT);
if (m != NULL)
*reply = m;
else {
*reply = slots[slotid].nfssl_reply;
slots[slotid].nfssl_reply = NULL;
}
}
slots[slotid].nfssl_inprog = 1;
error = NFSERR_REPLYFROMCACHE;
} else
/* No reply cached, so just do it. */
slots[slotid].nfssl_inprog = 1;
} else if ((slots[slotid].nfssl_seq + 1) == seqid) {
if (slots[slotid].nfssl_reply != NULL)
m_freem(slots[slotid].nfssl_reply);
slots[slotid].nfssl_reply = NULL;
slots[slotid].nfssl_inprog = 1;
slots[slotid].nfssl_seq++;
} else
error = NFSERR_SEQMISORDERED;
return (error);
}
/*
* Cache this reply for the slot.
* Use the "rep" argument to return the cached reply if repstat is set to
* NFSERR_REPLYFROMCACHE. The client never sets repstat to this value.
*/
void
nfsv4_seqsess_cacherep(uint32_t slotid, struct nfsslot *slots, int repstat,
struct mbuf **rep)
{
struct mbuf *m;
if (repstat == NFSERR_REPLYFROMCACHE) {
if (slots[slotid].nfssl_reply != NULL) {
/*
* We cannot sleep here, but copy will usually
* succeed.
*/
m = m_copym(slots[slotid].nfssl_reply, 0, M_COPYALL,
M_NOWAIT);
if (m != NULL)
*rep = m;
else {
/*
* Multiple retries would be extremely rare,
* so using the cached reply will likely
* be ok.
*/
*rep = slots[slotid].nfssl_reply;
slots[slotid].nfssl_reply = NULL;
}
} else
*rep = NULL;
} else {
if (slots[slotid].nfssl_reply != NULL)
m_freem(slots[slotid].nfssl_reply);
slots[slotid].nfssl_reply = *rep;
}
slots[slotid].nfssl_inprog = 0;
}
/*
* Generate the xdr for an NFSv4.1 Sequence Operation.
*/
void
nfsv4_setsequence(struct nfsmount *nmp, struct nfsrv_descript *nd,
struct nfsclsession *sep, int dont_replycache)
{
uint32_t *tl, slotseq = 0;
int error, maxslot, slotpos;
uint8_t sessionid[NFSX_V4SESSIONID];
error = nfsv4_sequencelookup(nmp, sep, &slotpos, &maxslot, &slotseq,
sessionid);
nd->nd_maxreq = sep->nfsess_maxreq;
nd->nd_maxresp = sep->nfsess_maxresp;
/* Build the Sequence arguments. */
NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID + 4 * NFSX_UNSIGNED);
nd->nd_sequence = tl;
bcopy(sessionid, tl, NFSX_V4SESSIONID);
tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
nd->nd_slotseq = tl;
if (error == 0) {
nd->nd_flag |= ND_HASSLOTID;
nd->nd_slotid = slotpos;
*tl++ = txdr_unsigned(slotseq);
*tl++ = txdr_unsigned(slotpos);
*tl++ = txdr_unsigned(maxslot);
if (dont_replycache == 0)
*tl = newnfs_true;
else
*tl = newnfs_false;
} else {
/*
* There are two errors and the rest of the session can
* just be zeros.
* NFSERR_BADSESSION: This bad session should just generate
* the same error again when the RPC is retried.
* ESTALE: A forced dismount is in progress and will cause the
* RPC to fail later.
*/
*tl++ = 0;
*tl++ = 0;
*tl++ = 0;
*tl = 0;
}
nd->nd_flag |= ND_HASSEQUENCE;
}
int
nfsv4_sequencelookup(struct nfsmount *nmp, struct nfsclsession *sep,
int *slotposp, int *maxslotp, uint32_t *slotseqp, uint8_t *sessionid)
{
int i, maxslot, slotpos;
uint64_t bitval;
/* Find an unused slot. */
slotpos = -1;
maxslot = -1;
mtx_lock(&sep->nfsess_mtx);
do {
if (nmp != NULL && sep->nfsess_defunct != 0) {
/* Just return the bad session. */
bcopy(sep->nfsess_sessionid, sessionid,
NFSX_V4SESSIONID);
mtx_unlock(&sep->nfsess_mtx);
return (NFSERR_BADSESSION);
}
bitval = 1;
for (i = 0; i < sep->nfsess_foreslots; i++) {
if ((bitval & sep->nfsess_slots) == 0) {
slotpos = i;
sep->nfsess_slots |= bitval;
sep->nfsess_slotseq[i]++;
*slotseqp = sep->nfsess_slotseq[i];
break;
}
bitval <<= 1;
}
if (slotpos == -1) {
/*
* If a forced dismount is in progress, just return.
* This RPC attempt will fail when it calls
* newnfs_request().
*/
if (nmp != NULL && NFSCL_FORCEDISM(nmp->nm_mountp)) {
mtx_unlock(&sep->nfsess_mtx);
return (ESTALE);
}
/* Wake up once/sec, to check for a forced dismount. */
(void)mtx_sleep(&sep->nfsess_slots, &sep->nfsess_mtx,
PZERO, "nfsclseq", hz);
}
} while (slotpos == -1);
/* Now, find the highest slot in use. (nfsc_slots is 64bits) */
bitval = 1;
for (i = 0; i < 64; i++) {
if ((bitval & sep->nfsess_slots) != 0)
maxslot = i;
bitval <<= 1;
}
bcopy(sep->nfsess_sessionid, sessionid, NFSX_V4SESSIONID);
mtx_unlock(&sep->nfsess_mtx);
*slotposp = slotpos;
*maxslotp = maxslot;
return (0);
}
/*
* Free a session slot.
*/
void
nfsv4_freeslot(struct nfsclsession *sep, int slot, bool resetseq)
{
uint64_t bitval;
bitval = 1;
if (slot > 0)
bitval <<= slot;
mtx_lock(&sep->nfsess_mtx);
if (resetseq)
sep->nfsess_slotseq[slot]--;
if ((bitval & sep->nfsess_slots) == 0)
printf("freeing free slot!!\n");
sep->nfsess_slots &= ~bitval;
wakeup(&sep->nfsess_slots);
mtx_unlock(&sep->nfsess_mtx);
}
/*
* Search for a matching pnfsd DS, based on the nmp arg.
* Return one if found, NULL otherwise.
*/
struct nfsdevice *
nfsv4_findmirror(struct nfsmount *nmp)
{
struct nfsdevice *ds;
mtx_assert(NFSDDSMUTEXPTR, MA_OWNED);
/*
* Search the DS server list for a match with nmp.
*/
if (nfsrv_devidcnt == 0)
return (NULL);
TAILQ_FOREACH(ds, &nfsrv_devidhead, nfsdev_list) {
if (ds->nfsdev_nmp == nmp) {
NFSCL_DEBUG(4, "nfsv4_findmirror: fnd main ds\n");
break;
}
}
return (ds);
}
/*
* Fill in the fields of "struct nfsrv_descript".
*/
void
nfsm_set(struct nfsrv_descript *nd, u_int offs)
{
struct mbuf *m;
int rlen;
m = nd->nd_mb;
if ((m->m_flags & M_EXTPG) != 0) {
nd->nd_bextpg = 0;
while (offs > 0) {
if (nd->nd_bextpg == 0)
rlen = m_epg_pagelen(m, 0, m->m_epg_1st_off);
else
rlen = m_epg_pagelen(m, nd->nd_bextpg, 0);
if (offs <= rlen)
break;
offs -= rlen;
nd->nd_bextpg++;
if (nd->nd_bextpg == m->m_epg_npgs) {
printf("nfsm_set: build offs "
"out of range\n");
nd->nd_bextpg--;
break;
}
}
nd->nd_bpos = (char *)(void *)
PHYS_TO_DMAP(m->m_epg_pa[nd->nd_bextpg]);
if (nd->nd_bextpg == 0)
nd->nd_bpos += m->m_epg_1st_off;
if (offs > 0) {
nd->nd_bpos += offs;
nd->nd_bextpgsiz = rlen - offs;
} else if (nd->nd_bextpg == 0)
nd->nd_bextpgsiz = PAGE_SIZE - m->m_epg_1st_off;
else
nd->nd_bextpgsiz = PAGE_SIZE;
} else
nd->nd_bpos = mtod(m, char *) + offs;
}
/*
* Grow a ext_pgs mbuf list. Either allocate another page or add
* an mbuf to the list.
*/
struct mbuf *
nfsm_add_ext_pgs(struct mbuf *m, int maxextsiz, int *bextpg)
{
struct mbuf *mp;
vm_page_t pg;
if ((m->m_epg_npgs + 1) * PAGE_SIZE > maxextsiz) {
mp = mb_alloc_ext_plus_pages(PAGE_SIZE, M_WAITOK);
*bextpg = 0;
m->m_next = mp;
} else {
pg = vm_page_alloc_noobj(VM_ALLOC_WAITOK | VM_ALLOC_NODUMP |
VM_ALLOC_WIRED);
m->m_epg_pa[m->m_epg_npgs] = VM_PAGE_TO_PHYS(pg);
*bextpg = m->m_epg_npgs;
m->m_epg_npgs++;
m->m_epg_last_len = 0;
mp = m;
}
return (mp);
}
diff --git a/sys/fs/nfs/nfs_var.h b/sys/fs/nfs/nfs_var.h
index 860b7b7ccfc3..345851573bd9 100644
--- a/sys/fs/nfs/nfs_var.h
+++ b/sys/fs/nfs/nfs_var.h
@@ -1,796 +1,796 @@
/*-
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 1989, 1993
* The Regents of the University of California. All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* Rick Macklem at The University of Guelph.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* $FreeBSD$
*/
/*
* XXX needs <nfs/rpcv2.h> and <nfs/nfs.h> because of typedefs
*/
struct uio;
struct ucred;
struct nfscred;
NFSPROC_T;
struct buf;
struct sockaddr_in;
struct nfs_dlmount;
struct file;
struct nfsmount;
struct socket;
struct nfsreq;
struct nfssockreq;
struct vattr;
struct nameidata;
struct nfsnode;
struct nfsfh;
struct sillyrename;
struct componentname;
struct nfsd_srvargs;
struct nfsrv_descript;
struct nfs_fattr;
union nethostaddr;
struct nfsstate;
struct nfslock;
struct nfsclient;
struct nfslayout;
struct nfsdsession;
struct nfslockconflict;
struct nfsd_idargs;
struct nfsd_clid;
struct nfsusrgrp;
struct nfsclowner;
struct nfsclopen;
struct nfsclopenhead;
struct nfsclclient;
struct nfsclsession;
struct nfscllockowner;
struct nfscllock;
struct nfscldeleg;
struct nfscllayout;
struct nfscldevinfo;
struct nfsv4lock;
struct nfsvattr;
struct nfs_vattr;
struct NFSSVCARGS;
struct nfsdevice;
struct pnfsdsfile;
struct pnfsdsattr;
#ifdef __FreeBSD__
NFS_ACCESS_ARGS;
NFS_OPEN_ARGS;
NFS_GETATTR_ARGS;
NFS_LOOKUP_ARGS;
NFS_READDIR_ARGS;
#endif
/* nfs_nfsdstate.c */
int nfsrv_setclient(struct nfsrv_descript *, struct nfsclient **,
nfsquad_t *, nfsquad_t *, NFSPROC_T *);
int nfsrv_getclient(nfsquad_t, int, struct nfsclient **, struct nfsdsession *,
nfsquad_t, uint32_t, struct nfsrv_descript *, NFSPROC_T *);
int nfsrv_destroyclient(nfsquad_t, NFSPROC_T *);
int nfsrv_destroysession(struct nfsrv_descript *, uint8_t *);
int nfsrv_bindconnsess(struct nfsrv_descript *, uint8_t *, int *);
int nfsrv_freestateid(struct nfsrv_descript *, nfsv4stateid_t *, NFSPROC_T *);
int nfsrv_teststateid(struct nfsrv_descript *, nfsv4stateid_t *, NFSPROC_T *);
int nfsrv_adminrevoke(struct nfsd_clid *, NFSPROC_T *);
void nfsrv_dumpclients(struct nfsd_dumpclients *, int);
void nfsrv_dumplocks(vnode_t, struct nfsd_dumplocks *, int, NFSPROC_T *);
int nfsrv_lockctrl(vnode_t, struct nfsstate **,
struct nfslock **, struct nfslockconflict *, nfsquad_t, nfsv4stateid_t *,
struct nfsexstuff *, struct nfsrv_descript *, NFSPROC_T *);
int nfsrv_openctrl(struct nfsrv_descript *, vnode_t,
struct nfsstate **, nfsquad_t, nfsv4stateid_t *, nfsv4stateid_t *,
u_int32_t *, struct nfsexstuff *, NFSPROC_T *, u_quad_t);
int nfsrv_opencheck(nfsquad_t, nfsv4stateid_t *, struct nfsstate *,
vnode_t, struct nfsrv_descript *, NFSPROC_T *, int);
int nfsrv_openupdate(vnode_t, struct nfsstate *, nfsquad_t,
nfsv4stateid_t *, struct nfsrv_descript *, NFSPROC_T *, int *);
int nfsrv_delegupdate(struct nfsrv_descript *, nfsquad_t, nfsv4stateid_t *,
vnode_t, int, struct ucred *, NFSPROC_T *, int *);
int nfsrv_releaselckown(struct nfsstate *, nfsquad_t, NFSPROC_T *);
void nfsrv_zapclient(struct nfsclient *, NFSPROC_T *);
int nfssvc_idname(struct nfsd_idargs *);
void nfsrv_servertimer(void);
int nfsrv_getclientipaddr(struct nfsrv_descript *, struct nfsclient *);
void nfsrv_setupstable(NFSPROC_T *);
void nfsrv_updatestable(NFSPROC_T *);
void nfsrv_writestable(u_char *, int, int, NFSPROC_T *);
void nfsrv_throwawayopens(NFSPROC_T *);
int nfsrv_checkremove(vnode_t, int, struct nfsrv_descript *, nfsquad_t,
NFSPROC_T *);
void nfsd_recalldelegation(vnode_t, NFSPROC_T *);
void nfsd_disabledelegation(vnode_t, NFSPROC_T *);
int nfsrv_checksetattr(vnode_t, struct nfsrv_descript *,
nfsv4stateid_t *, struct nfsvattr *, nfsattrbit_t *, struct nfsexstuff *,
NFSPROC_T *);
int nfsrv_checkgetattr(struct nfsrv_descript *, vnode_t,
struct nfsvattr *, nfsattrbit_t *, NFSPROC_T *);
int nfsrv_nfsuserdport(struct nfsuserd_args *, NFSPROC_T *);
void nfsrv_nfsuserddelport(void);
void nfsrv_throwawayallstate(NFSPROC_T *);
int nfsrv_checksequence(struct nfsrv_descript *, uint32_t, uint32_t *,
uint32_t *, int, uint32_t *, NFSPROC_T *);
int nfsrv_checkreclaimcomplete(struct nfsrv_descript *, int);
void nfsrv_cache_session(struct nfsrv_descript *, struct mbuf **);
void nfsrv_freeallbackchannel_xprts(void);
int nfsrv_layoutcommit(struct nfsrv_descript *, vnode_t, int, int, uint64_t,
uint64_t, uint64_t, int, struct timespec *, int, nfsv4stateid_t *,
int, char *, int *, uint64_t *, struct ucred *, NFSPROC_T *);
int nfsrv_layoutget(struct nfsrv_descript *, vnode_t, struct nfsexstuff *,
int, int *, uint64_t *, uint64_t *, uint64_t, nfsv4stateid_t *, int, int *,
int *, char *, struct ucred *, NFSPROC_T *);
void nfsrv_flexmirrordel(char *, NFSPROC_T *);
void nfsrv_recalloldlayout(NFSPROC_T *);
int nfsrv_layoutreturn(struct nfsrv_descript *, vnode_t, int, int, uint64_t,
uint64_t, int, int, nfsv4stateid_t *, int, uint32_t *, int *,
struct ucred *, NFSPROC_T *);
int nfsrv_getdevinfo(char *, int, uint32_t *, uint32_t *, int *, char **);
void nfsrv_freeonedevid(struct nfsdevice *);
void nfsrv_freealllayoutsanddevids(void);
void nfsrv_freefilelayouts(fhandle_t *);
int nfsrv_deldsserver(int, char *, NFSPROC_T *);
struct nfsdevice *nfsrv_deldsnmp(int, struct nfsmount *, NFSPROC_T *);
int nfsrv_delds(char *, NFSPROC_T *);
int nfsrv_createdevids(struct nfsd_nfsd_args *, NFSPROC_T *);
int nfsrv_checkdsattr(vnode_t, NFSPROC_T *);
int nfsrv_copymr(vnode_t, vnode_t, vnode_t, struct nfsdevice *,
struct pnfsdsfile *, struct pnfsdsfile *, int, struct ucred *, NFSPROC_T *);
int nfsrv_mdscopymr(char *, char *, char *, char *, int *, char *, NFSPROC_T *,
struct vnode **, struct vnode **, struct pnfsdsfile **, struct nfsdevice **,
struct nfsdevice **);
void nfsrv_marknospc(char *, bool);
/* nfs_nfsdserv.c */
int nfsrvd_access(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_getattr(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_setattr(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_lookup(struct nfsrv_descript *, int,
vnode_t, vnode_t *, fhandle_t *, struct nfsexstuff *);
int nfsrvd_readlink(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_read(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_write(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_create(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_mknod(struct nfsrv_descript *, int,
vnode_t, vnode_t *, fhandle_t *, struct nfsexstuff *);
int nfsrvd_remove(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_rename(struct nfsrv_descript *, int,
vnode_t, vnode_t, struct nfsexstuff *, struct nfsexstuff *);
int nfsrvd_link(struct nfsrv_descript *, int,
vnode_t, vnode_t, struct nfsexstuff *, struct nfsexstuff *);
int nfsrvd_symlink(struct nfsrv_descript *, int,
vnode_t, vnode_t *, fhandle_t *, struct nfsexstuff *);
int nfsrvd_mkdir(struct nfsrv_descript *, int,
vnode_t, vnode_t *, fhandle_t *, struct nfsexstuff *);
int nfsrvd_readdir(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_readdirplus(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_commit(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_statfs(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_fsinfo(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_close(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_delegpurge(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_delegreturn(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_getfh(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_lock(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_lockt(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_locku(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_openconfirm(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_opendowngrade(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_renew(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_secinfo(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_secinfononame(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_setclientid(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_setclientidcfrm(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_verify(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_open(struct nfsrv_descript *, int,
vnode_t, vnode_t *, fhandle_t *, struct nfsexstuff *);
int nfsrvd_openattr(struct nfsrv_descript *, int,
vnode_t, vnode_t *, fhandle_t *, struct nfsexstuff *);
int nfsrvd_releaselckown(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_pathconf(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_exchangeid(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_createsession(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_sequence(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_reclaimcomplete(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_destroyclientid(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_bindconnsess(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_destroysession(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_freestateid(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_layoutget(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_getdevinfo(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_layoutcommit(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_layoutreturn(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_ioadvise(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_layouterror(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_layoutstats(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_teststateid(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_allocate(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_deallocate(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_copy_file_range(struct nfsrv_descript *, int,
vnode_t, vnode_t, struct nfsexstuff *, struct nfsexstuff *);
int nfsrvd_seek(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_getxattr(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_setxattr(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_rmxattr(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_listxattr(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
int nfsrvd_notsupp(struct nfsrv_descript *, int,
vnode_t, struct nfsexstuff *);
/* nfs_nfsdsocket.c */
void nfsrvd_rephead(struct nfsrv_descript *);
void nfsrvd_dorpc(struct nfsrv_descript *, int, u_char *, int, u_int32_t);
/* nfs_nfsdcache.c */
void nfsrvd_initcache(void);
int nfsrvd_getcache(struct nfsrv_descript *);
struct nfsrvcache *nfsrvd_updatecache(struct nfsrv_descript *);
void nfsrvd_sentcache(struct nfsrvcache *, int, uint32_t);
void nfsrvd_cleancache(void);
void nfsrvd_refcache(struct nfsrvcache *);
void nfsrvd_derefcache(struct nfsrvcache *);
void nfsrvd_delcache(struct nfsrvcache *);
void nfsrc_trimcache(uint64_t, uint32_t, int);
/* nfs_commonsubs.c */
void nfscl_reqstart(struct nfsrv_descript *, int, struct nfsmount *,
u_int8_t *, int, u_int32_t **, struct nfsclsession *, int, int);
void nfsm_stateidtom(struct nfsrv_descript *, nfsv4stateid_t *, int);
void nfscl_fillsattr(struct nfsrv_descript *, struct vattr *,
vnode_t, int, u_int32_t);
void newnfs_init(void);
int nfsaddr_match(int, union nethostaddr *, NFSSOCKADDR_T);
int nfsaddr2_match(NFSSOCKADDR_T, NFSSOCKADDR_T);
int nfsm_strtom(struct nfsrv_descript *, const char *, int);
int nfsm_mbufuio(struct nfsrv_descript *, struct uio *, int);
int nfsm_fhtom(struct nfsrv_descript *, u_int8_t *, int, int);
int nfsm_advance(struct nfsrv_descript *, int, int);
void *nfsm_dissct(struct nfsrv_descript *, int, int);
void newnfs_copycred(struct nfscred *, struct ucred *);
void newnfs_copyincred(struct ucred *, struct nfscred *);
-int nfsrv_dissectacl(struct nfsrv_descript *, NFSACL_T *, int *,
+int nfsrv_dissectacl(struct nfsrv_descript *, NFSACL_T *, bool, int *,
int *, NFSPROC_T *);
int nfsrv_getattrbits(struct nfsrv_descript *, nfsattrbit_t *, int *,
int *);
int nfsv4_loadattr(struct nfsrv_descript *, vnode_t,
struct nfsvattr *, struct nfsfh **, fhandle_t *, int,
struct nfsv3_pathconf *, struct statfs *, struct nfsstatfs *,
struct nfsfsinfo *, NFSACL_T *,
int, int *, u_int32_t *, u_int32_t *, NFSPROC_T *, struct ucred *);
int nfsv4_lock(struct nfsv4lock *, int, int *, struct mtx *, struct mount *);
void nfsv4_unlock(struct nfsv4lock *, int);
void nfsv4_relref(struct nfsv4lock *);
void nfsv4_getref(struct nfsv4lock *, int *, struct mtx *, struct mount *);
int nfsv4_getref_nonblock(struct nfsv4lock *);
int nfsv4_testlock(struct nfsv4lock *);
int nfsrv_mtostr(struct nfsrv_descript *, char *, int);
void nfsrv_cleanusergroup(void);
int nfsrv_checkutf8(u_int8_t *, int);
int newnfs_sndlock(int *);
void newnfs_sndunlock(int *);
int nfsv4_getipaddr(struct nfsrv_descript *, struct sockaddr_in *,
struct sockaddr_in6 *, sa_family_t *, int *);
int nfsv4_seqsession(uint32_t, uint32_t, uint32_t, struct nfsslot *,
struct mbuf **, uint16_t);
void nfsv4_seqsess_cacherep(uint32_t, struct nfsslot *, int, struct mbuf **);
void nfsv4_setsequence(struct nfsmount *, struct nfsrv_descript *,
struct nfsclsession *, int);
int nfsv4_sequencelookup(struct nfsmount *, struct nfsclsession *, int *,
int *, uint32_t *, uint8_t *);
void nfsv4_freeslot(struct nfsclsession *, int, bool);
struct ucred *nfsrv_getgrpscred(struct ucred *);
struct nfsdevice *nfsv4_findmirror(struct nfsmount *);
void nfsm_set(struct nfsrv_descript *, u_int);
struct mbuf *nfsm_add_ext_pgs(struct mbuf *, int, int *);
/* nfs_clcomsubs.c */
void nfsm_uiombuf(struct nfsrv_descript *, struct uio *, int);
struct mbuf *nfsm_uiombuflist(struct uio *, int, u_int);
nfsuint64 *nfscl_getcookie(struct nfsnode *, off_t off, int);
u_int8_t *nfscl_getmyip(struct nfsmount *, struct in6_addr *, int *);
int nfsm_getfh(struct nfsrv_descript *, struct nfsfh **);
int nfscl_mtofh(struct nfsrv_descript *, struct nfsfh **,
struct nfsvattr *, int *);
int nfscl_postop_attr(struct nfsrv_descript *, struct nfsvattr *, int *,
void *);
int nfscl_wcc_data(struct nfsrv_descript *, vnode_t,
struct nfsvattr *, int *, int *, void *);
int nfsm_loadattr(struct nfsrv_descript *, struct nfsvattr *);
int nfscl_request(struct nfsrv_descript *, vnode_t,
NFSPROC_T *, struct ucred *, void *);
/* nfs_nfsdsubs.c */
void nfsd_fhtovp(struct nfsrv_descript *, struct nfsrvfh *, int,
vnode_t *, struct nfsexstuff *, mount_t *, int, int);
int nfsd_excred(struct nfsrv_descript *, struct nfsexstuff *, struct ucred *,
bool);
int nfsrv_mtofh(struct nfsrv_descript *, struct nfsrvfh *);
int nfsrv_putattrbit(struct nfsrv_descript *, nfsattrbit_t *);
void nfsrv_wcc(struct nfsrv_descript *, int, struct nfsvattr *, int,
struct nfsvattr *);
int nfsv4_fillattr(struct nfsrv_descript *, struct mount *, vnode_t, NFSACL_T *,
struct vattr *, fhandle_t *, int, nfsattrbit_t *,
struct ucred *, NFSPROC_T *, int, int, int, int, uint64_t, struct statfs *);
void nfsrv_fillattr(struct nfsrv_descript *, struct nfsvattr *);
struct mbuf *nfsrv_adj(struct mbuf *, int, int);
void nfsrv_postopattr(struct nfsrv_descript *, int, struct nfsvattr *);
int nfsd_errmap(struct nfsrv_descript *);
void nfsv4_uidtostr(uid_t, u_char **, int *);
int nfsv4_strtouid(struct nfsrv_descript *, u_char *, int, uid_t *);
void nfsv4_gidtostr(gid_t, u_char **, int *);
int nfsv4_strtogid(struct nfsrv_descript *, u_char *, int, gid_t *);
int nfsrv_checkuidgid(struct nfsrv_descript *, struct nfsvattr *);
void nfsrv_fixattr(struct nfsrv_descript *, vnode_t,
struct nfsvattr *, NFSACL_T *, NFSPROC_T *, nfsattrbit_t *,
struct nfsexstuff *);
int nfsrv_errmoved(int);
int nfsrv_putreferralattr(struct nfsrv_descript *, nfsattrbit_t *,
struct nfsreferral *, int, int *);
int nfsrv_parsename(struct nfsrv_descript *, char *, u_long *,
NFSPATHLEN_T *);
void nfsd_init(void);
int nfsd_checkrootexp(struct nfsrv_descript *);
void nfsd_getminorvers(struct nfsrv_descript *, u_char *, u_char **, int *,
u_int32_t *);
/* nfs_clvfsops.c */
void nfscl_retopts(struct nfsmount *, char *, size_t);
/* nfs_commonport.c */
int nfsrv_lookupfilename(struct nameidata *, char *, NFSPROC_T *);
void nfsrv_object_create(vnode_t, NFSPROC_T *);
int nfsrv_mallocmget_limit(void);
int nfsvno_v4rootexport(struct nfsrv_descript *);
void newnfs_portinit(void);
struct ucred *newnfs_getcred(void);
void newnfs_setroot(struct ucred *);
int nfs_catnap(int, int, const char *);
struct nfsreferral *nfsv4root_getreferral(vnode_t, vnode_t, u_int32_t);
int nfsvno_pathconf(vnode_t, int, long *, struct ucred *, NFSPROC_T *);
int nfsrv_atroot(vnode_t, uint64_t *);
int nfs_supportsnfsv4acls(vnode_t);
/* nfs_commonacl.c */
int nfsrv_dissectace(struct nfsrv_descript *, struct acl_entry *,
- int *, int *, NFSPROC_T *);
+ bool, int *, int *, NFSPROC_T *);
int nfsrv_buildacl(struct nfsrv_descript *, NFSACL_T *, enum vtype,
NFSPROC_T *);
int nfsrv_compareacl(NFSACL_T *, NFSACL_T *);
/* nfs_clrpcops.c */
int nfsrpc_null(vnode_t, struct ucred *, NFSPROC_T *);
int nfsrpc_access(vnode_t, int, struct ucred *, NFSPROC_T *,
struct nfsvattr *, int *);
int nfsrpc_accessrpc(vnode_t, u_int32_t, struct ucred *,
NFSPROC_T *, struct nfsvattr *, int *, u_int32_t *, void *);
int nfsrpc_open(vnode_t, int, struct ucred *, NFSPROC_T *);
int nfsrpc_openrpc(struct nfsmount *, vnode_t, u_int8_t *, int, u_int8_t *, int,
u_int32_t, struct nfsclopen *, u_int8_t *, int, struct nfscldeleg **, int,
u_int32_t, struct ucred *, NFSPROC_T *, int, int);
int nfsrpc_opendowngrade(vnode_t, u_int32_t, struct nfsclopen *,
struct ucred *, NFSPROC_T *);
int nfsrpc_close(vnode_t, int, NFSPROC_T *);
int nfsrpc_closerpc(struct nfsrv_descript *, struct nfsmount *,
struct nfsclopen *, struct ucred *, NFSPROC_T *, int);
int nfsrpc_openconfirm(vnode_t, u_int8_t *, int, struct nfsclopen *,
struct ucred *, NFSPROC_T *);
int nfsrpc_setclient(struct nfsmount *, struct nfsclclient *, int,
bool *, struct ucred *, NFSPROC_T *);
int nfsrpc_getattr(vnode_t, struct ucred *, NFSPROC_T *,
struct nfsvattr *, void *);
int nfsrpc_getattrnovp(struct nfsmount *, u_int8_t *, int, int,
struct ucred *, NFSPROC_T *, struct nfsvattr *, u_int64_t *, uint32_t *);
int nfsrpc_setattr(vnode_t, struct vattr *, NFSACL_T *, struct ucred *,
NFSPROC_T *, struct nfsvattr *, int *, void *);
int nfsrpc_lookup(vnode_t, char *, int, struct ucred *, NFSPROC_T *,
struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *, int *,
void *, uint32_t);
int nfsrpc_readlink(vnode_t, struct uio *, struct ucred *,
NFSPROC_T *, struct nfsvattr *, int *, void *);
int nfsrpc_read(vnode_t, struct uio *, struct ucred *, NFSPROC_T *,
struct nfsvattr *, int *, void *);
int nfsrpc_write(vnode_t, struct uio *, int *, int *,
struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *, int);
int nfsrpc_mknod(vnode_t, char *, int, struct vattr *, u_int32_t,
enum vtype, struct ucred *, NFSPROC_T *, struct nfsvattr *,
struct nfsvattr *, struct nfsfh **, int *, int *, void *);
int nfsrpc_create(vnode_t, char *, int, struct vattr *, nfsquad_t,
int, struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
struct nfsfh **, int *, int *, void *);
int nfsrpc_remove(vnode_t, char *, int, vnode_t, struct ucred *, NFSPROC_T *,
struct nfsvattr *, int *, void *);
int nfsrpc_rename(vnode_t, vnode_t, char *, int, vnode_t, vnode_t, char *, int,
struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
int *, int *, void *, void *);
int nfsrpc_link(vnode_t, vnode_t, char *, int,
struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
int *, int *, void *);
int nfsrpc_symlink(vnode_t, char *, int, const char *, struct vattr *,
struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
struct nfsfh **, int *, int *, void *);
int nfsrpc_mkdir(vnode_t, char *, int, struct vattr *,
struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
struct nfsfh **, int *, int *, void *);
int nfsrpc_rmdir(vnode_t, char *, int, struct ucred *, NFSPROC_T *,
struct nfsvattr *, int *, void *);
int nfsrpc_readdir(vnode_t, struct uio *, nfsuint64 *, struct ucred *,
NFSPROC_T *, struct nfsvattr *, int *, int *, void *);
int nfsrpc_readdirplus(vnode_t, struct uio *, nfsuint64 *,
struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, int *, void *);
int nfsrpc_commit(vnode_t, u_quad_t, int, struct ucred *,
NFSPROC_T *, struct nfsvattr *, int *, void *);
int nfsrpc_advlock(vnode_t, off_t, int, struct flock *, int,
struct ucred *, NFSPROC_T *, void *, int);
int nfsrpc_lockt(struct nfsrv_descript *, vnode_t,
struct nfsclclient *, u_int64_t, u_int64_t, struct flock *,
struct ucred *, NFSPROC_T *, void *, int);
int nfsrpc_lock(struct nfsrv_descript *, struct nfsmount *, vnode_t,
u_int8_t *, int, struct nfscllockowner *, int, int, u_int64_t,
u_int64_t, short, struct ucred *, NFSPROC_T *, int);
int nfsrpc_statfs(vnode_t, struct nfsstatfs *, struct nfsfsinfo *,
struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *);
int nfsrpc_fsinfo(vnode_t, struct nfsfsinfo *, struct ucred *,
NFSPROC_T *, struct nfsvattr *, int *, void *);
int nfsrpc_pathconf(vnode_t, struct nfsv3_pathconf *,
struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *);
int nfsrpc_renew(struct nfsclclient *, struct nfsclds *, struct ucred *,
NFSPROC_T *);
int nfsrpc_rellockown(struct nfsmount *, struct nfscllockowner *, uint8_t *,
int, struct ucred *, NFSPROC_T *);
int nfsrpc_getdirpath(struct nfsmount *, u_char *, struct ucred *,
NFSPROC_T *);
int nfsrpc_delegreturn(struct nfscldeleg *, struct ucred *,
struct nfsmount *, NFSPROC_T *, int);
int nfsrpc_getacl(vnode_t, struct ucred *, NFSPROC_T *, NFSACL_T *, void *);
int nfsrpc_setacl(vnode_t, struct ucred *, NFSPROC_T *, NFSACL_T *, void *);
int nfsrpc_exchangeid(struct nfsmount *, struct nfsclclient *,
struct nfssockreq *, int, uint32_t, struct nfsclds **, struct ucred *,
NFSPROC_T *);
int nfsrpc_createsession(struct nfsmount *, struct nfsclsession *,
struct nfssockreq *, struct nfsclds *, uint32_t, int, struct ucred *,
NFSPROC_T *);
int nfsrpc_destroysession(struct nfsmount *, struct nfsclclient *,
struct ucred *, NFSPROC_T *);
int nfsrpc_destroyclient(struct nfsmount *, struct nfsclclient *,
struct ucred *, NFSPROC_T *);
int nfsrpc_getdeviceinfo(struct nfsmount *, uint8_t *, int, uint32_t *,
struct nfscldevinfo **, struct ucred *, NFSPROC_T *);
int nfsrpc_layoutcommit(struct nfsmount *, uint8_t *, int, int,
uint64_t, uint64_t, uint64_t, nfsv4stateid_t *, int, struct ucred *,
NFSPROC_T *, void *);
int nfsrpc_layoutreturn(struct nfsmount *, uint8_t *, int, int, int, uint32_t,
int, uint64_t, uint64_t, nfsv4stateid_t *, struct ucred *, NFSPROC_T *,
uint32_t, uint32_t, char *);
int nfsrpc_reclaimcomplete(struct nfsmount *, struct ucred *, NFSPROC_T *);
int nfsrpc_advise(vnode_t, off_t, uint64_t, int, struct ucred *, NFSPROC_T *);
int nfscl_doiods(vnode_t, struct uio *, int *, int *, uint32_t, int,
struct ucred *, NFSPROC_T *);
int nfscl_findlayoutforio(struct nfscllayout *, uint64_t, uint32_t,
struct nfsclflayout **);
void nfscl_freenfsclds(struct nfsclds *);
int nfsrpc_allocate(vnode_t, off_t, off_t, struct nfsvattr *, int *,
struct ucred *, NFSPROC_T *, void *);
int nfsrpc_deallocate(vnode_t, off_t, off_t, struct nfsvattr *, int *,
struct ucred *, NFSPROC_T *, void *);
int nfsrpc_copy_file_range(vnode_t, off_t *, vnode_t, off_t *, size_t *,
unsigned int, int *, struct nfsvattr *, int *, struct nfsvattr *,
struct ucred *, bool, bool *);
int nfsrpc_seek(vnode_t, off_t *, bool *, int, struct ucred *,
struct nfsvattr *, int *);
int nfsrpc_getextattr(vnode_t, const char *, struct uio *, ssize_t *,
struct nfsvattr *, int *, struct ucred *, NFSPROC_T *);
int nfsrpc_setextattr(vnode_t, const char *, struct uio *, struct nfsvattr *,
int *, struct ucred *, NFSPROC_T *);
int nfsrpc_listextattr(vnode_t, uint64_t *, struct uio *, size_t *, bool *,
struct nfsvattr *, int *, struct ucred *, NFSPROC_T *);
int nfsrpc_rmextattr(vnode_t, const char *, struct nfsvattr *, int *,
struct ucred *, NFSPROC_T *);
void nfsrpc_bindconnsess(CLIENT *, void *, struct ucred *);
/* nfs_clstate.c */
int nfscl_open(vnode_t, u_int8_t *, int, u_int32_t, int,
struct ucred *, NFSPROC_T *, struct nfsclowner **, struct nfsclopen **,
int *, int *, int, bool);
int nfscl_getstateid(vnode_t, u_int8_t *, int, u_int32_t, int, struct ucred *,
NFSPROC_T *, nfsv4stateid_t *, void **);
void nfscl_ownerrelease(struct nfsmount *, struct nfsclowner *, int, int, int);
void nfscl_openrelease(struct nfsmount *, struct nfsclopen *, int, int);
int nfscl_getcl(struct mount *, struct ucred *, NFSPROC_T *, bool, bool,
struct nfsclclient **);
struct nfsclclient *nfscl_findcl(struct nfsmount *);
void nfscl_clientrelease(struct nfsclclient *);
void nfscl_freelock(struct nfscllock *, int);
void nfscl_freelockowner(struct nfscllockowner *, int);
int nfscl_getbytelock(vnode_t, u_int64_t, u_int64_t, short,
struct ucred *, NFSPROC_T *, struct nfsclclient *, int, void *, int,
u_int8_t *, u_int8_t *, struct nfscllockowner **, int *, int *);
int nfscl_relbytelock(vnode_t, u_int64_t, u_int64_t,
struct ucred *, NFSPROC_T *, int, struct nfsclclient *,
void *, int, struct nfscllockowner **, int *);
int nfscl_checkwritelocked(vnode_t, struct flock *,
struct ucred *, NFSPROC_T *, void *, int);
void nfscl_lockrelease(struct nfscllockowner *, int, int);
void nfscl_fillclid(u_int64_t, char *, u_int8_t *, u_int16_t);
void nfscl_filllockowner(void *, u_int8_t *, int);
void nfscl_freeopen(struct nfsclopen *, int, bool);
void nfscl_umount(struct nfsmount *, NFSPROC_T *, struct nfscldeleghead *);
void nfscl_renewthread(struct nfsclclient *, NFSPROC_T *);
void nfscl_initiate_recovery(struct nfsclclient *);
int nfscl_hasexpired(struct nfsclclient *, u_int32_t, NFSPROC_T *);
void nfscl_dumpstate(struct nfsmount *, int, int, int, int);
void nfscl_dupopen(vnode_t, int);
int nfscl_getclose(vnode_t, struct nfsclclient **);
int nfscl_doclose(vnode_t, struct nfsclclient **, NFSPROC_T *);
int nfsrpc_doclose(struct nfsmount *, struct nfsclopen *, NFSPROC_T *, bool,
bool);
int nfscl_deleg(mount_t, struct nfsclclient *, u_int8_t *, int,
struct ucred *, NFSPROC_T *, struct nfscldeleg **);
void nfscl_lockinit(struct nfsv4lock *);
void nfscl_lockexcl(struct nfsv4lock *, void *);
void nfscl_lockunlock(struct nfsv4lock *);
void nfscl_lockderef(struct nfsv4lock *);
void nfscl_delegreturnvp(vnode_t, NFSPROC_T *);
void nfscl_docb(struct nfsrv_descript *, NFSPROC_T *);
void nfscl_releasealllocks(struct nfsclclient *, vnode_t, NFSPROC_T *, void *,
int);
int nfscl_lockt(vnode_t, struct nfsclclient *, u_int64_t,
u_int64_t, struct flock *, NFSPROC_T *, void *, int);
int nfscl_mustflush(vnode_t);
int nfscl_nodeleg(vnode_t, int);
int nfscl_removedeleg(vnode_t, NFSPROC_T *, nfsv4stateid_t *);
int nfscl_getref(struct nfsmount *);
void nfscl_relref(struct nfsmount *);
int nfscl_renamedeleg(vnode_t, nfsv4stateid_t *, int *, vnode_t,
nfsv4stateid_t *, int *, NFSPROC_T *);
void nfscl_reclaimnode(vnode_t);
void nfscl_newnode(vnode_t);
void nfscl_delegmodtime(vnode_t);
void nfscl_deleggetmodtime(vnode_t, struct timespec *);
int nfscl_trydelegreturn(struct nfscldeleg *, struct ucred *,
struct nfsmount *, NFSPROC_T *);
int nfscl_tryclose(struct nfsclopen *, struct ucred *,
struct nfsmount *, NFSPROC_T *, bool);
void nfscl_cleanup(NFSPROC_T *);
int nfscl_layout(struct nfsmount *, vnode_t, u_int8_t *, int, nfsv4stateid_t *,
int, int, struct nfsclflayouthead *, struct nfscllayout **, struct ucred *,
NFSPROC_T *);
struct nfscllayout *nfscl_getlayout(struct nfsclclient *, uint8_t *, int,
uint64_t, uint32_t, struct nfsclflayout **, int *);
void nfscl_dserr(uint32_t, uint32_t, struct nfscldevinfo *,
struct nfscllayout *, struct nfsclds *);
void nfscl_cancelreqs(struct nfsclds *);
void nfscl_rellayout(struct nfscllayout *, int);
struct nfscldevinfo *nfscl_getdevinfo(struct nfsclclient *, uint8_t *,
struct nfscldevinfo *);
void nfscl_reldevinfo(struct nfscldevinfo *);
int nfscl_adddevinfo(struct nfsmount *, struct nfscldevinfo *, int,
struct nfsclflayout *);
void nfscl_freelayout(struct nfscllayout *);
void nfscl_freeflayout(struct nfsclflayout *);
void nfscl_freedevinfo(struct nfscldevinfo *);
int nfscl_layoutcommit(vnode_t, NFSPROC_T *);
/* nfs_clport.c */
int nfscl_nget(mount_t, vnode_t, struct nfsfh *,
struct componentname *, NFSPROC_T *, struct nfsnode **, void *, int);
NFSPROC_T *nfscl_getparent(NFSPROC_T *);
void nfscl_start_renewthread(struct nfsclclient *);
void nfscl_loadsbinfo(struct nfsmount *, struct nfsstatfs *, void *);
void nfscl_loadfsinfo (struct nfsmount *, struct nfsfsinfo *);
void nfscl_delegreturn(struct nfscldeleg *, int, struct nfsmount *,
struct ucred *, NFSPROC_T *);
void nfsrvd_cbinit(int);
int nfscl_checksattr(struct vattr *, struct nfsvattr *);
int nfscl_ngetreopen(mount_t, u_int8_t *, int, NFSPROC_T *,
struct nfsnode **);
int nfscl_procdoesntexist(u_int8_t *);
int nfscl_maperr(NFSPROC_T *, int, uid_t, gid_t);
/* nfs_clsubs.c */
void nfscl_init(void);
/* nfs_clbio.c */
int ncl_flush(vnode_t, int, NFSPROC_T *, int, int);
/* nfs_clnode.c */
void ncl_invalcaches(vnode_t);
/* nfs_nfsdport.c */
int nfsvno_getattr(vnode_t, struct nfsvattr *, struct nfsrv_descript *,
NFSPROC_T *, int, nfsattrbit_t *);
int nfsvno_setattr(vnode_t, struct nfsvattr *, struct ucred *,
NFSPROC_T *, struct nfsexstuff *);
int nfsvno_getfh(vnode_t, fhandle_t *, NFSPROC_T *);
int nfsvno_accchk(vnode_t, accmode_t, struct ucred *,
struct nfsexstuff *, NFSPROC_T *, int, int, u_int32_t *);
int nfsvno_namei(struct nfsrv_descript *, struct nameidata *,
vnode_t, int, struct nfsexstuff *, vnode_t *);
void nfsvno_setpathbuf(struct nameidata *, char **, u_long **);
void nfsvno_relpathbuf(struct nameidata *);
int nfsvno_readlink(vnode_t, struct ucred *, int, NFSPROC_T *, struct mbuf **,
struct mbuf **, int *);
int nfsvno_read(vnode_t, off_t, int, struct ucred *, int, NFSPROC_T *,
struct mbuf **, struct mbuf **);
int nfsvno_write(vnode_t, off_t, int, int *, struct mbuf *, char *,
struct ucred *, NFSPROC_T *);
int nfsvno_createsub(struct nfsrv_descript *, struct nameidata *,
vnode_t *, struct nfsvattr *, int *, int32_t *, NFSDEV_T,
struct nfsexstuff *);
int nfsvno_mknod(struct nameidata *, struct nfsvattr *, struct ucred *,
NFSPROC_T *);
int nfsvno_mkdir(struct nameidata *,
struct nfsvattr *, uid_t, struct ucred *, NFSPROC_T *,
struct nfsexstuff *);
int nfsvno_symlink(struct nameidata *, struct nfsvattr *, char *, int, int,
uid_t, struct ucred *, NFSPROC_T *, struct nfsexstuff *);
int nfsvno_getsymlink(struct nfsrv_descript *, struct nfsvattr *,
NFSPROC_T *, char **, int *);
int nfsvno_removesub(struct nameidata *, int, struct ucred *, NFSPROC_T *,
struct nfsexstuff *);
int nfsvno_rmdirsub(struct nameidata *, int, struct ucred *, NFSPROC_T *,
struct nfsexstuff *);
int nfsvno_rename(struct nameidata *, struct nameidata *, u_int32_t,
u_int32_t, struct ucred *, NFSPROC_T *);
int nfsvno_link(struct nameidata *, vnode_t, struct ucred *,
NFSPROC_T *, struct nfsexstuff *);
int nfsvno_fsync(vnode_t, u_int64_t, int, struct ucred *, NFSPROC_T *);
int nfsvno_statfs(vnode_t, struct statfs *);
void nfsvno_getfs(struct nfsfsinfo *, int);
void nfsvno_open(struct nfsrv_descript *, struct nameidata *, nfsquad_t,
nfsv4stateid_t *, struct nfsstate *, int *, struct nfsvattr *, int32_t *,
int, NFSACL_T *, nfsattrbit_t *, struct ucred *,
struct nfsexstuff *, vnode_t *);
int nfsvno_updfilerev(vnode_t, struct nfsvattr *, struct nfsrv_descript *,
NFSPROC_T *);
int nfsvno_fillattr(struct nfsrv_descript *, struct mount *, vnode_t,
struct nfsvattr *, fhandle_t *, int, nfsattrbit_t *,
struct ucred *, NFSPROC_T *, int, int, int, int, uint64_t);
int nfsrv_sattr(struct nfsrv_descript *, vnode_t, struct nfsvattr *, nfsattrbit_t *,
NFSACL_T *, NFSPROC_T *);
int nfsv4_sattr(struct nfsrv_descript *, vnode_t, struct nfsvattr *, nfsattrbit_t *,
NFSACL_T *, NFSPROC_T *);
int nfsvno_checkexp(mount_t, NFSSOCKADDR_T, struct nfsexstuff *,
struct ucred **);
int nfsvno_fhtovp(mount_t, fhandle_t *, NFSSOCKADDR_T, int,
vnode_t *, struct nfsexstuff *, struct ucred **);
vnode_t nfsvno_getvp(fhandle_t *);
int nfsvno_advlock(vnode_t, int, u_int64_t, u_int64_t, NFSPROC_T *);
int nfsrv_v4rootexport(void *, struct ucred *, NFSPROC_T *);
int nfsvno_testexp(struct nfsrv_descript *, struct nfsexstuff *);
uint32_t nfsrv_hashfh(fhandle_t *);
uint32_t nfsrv_hashsessionid(uint8_t *);
void nfsrv_backupstable(void);
int nfsrv_dsgetdevandfh(struct vnode *, NFSPROC_T *, int *, fhandle_t *,
char *);
int nfsrv_dsgetsockmnt(struct vnode *, int, char *, int *, int *,
NFSPROC_T *, struct vnode **, fhandle_t *, char *, char *,
struct vnode **, struct nfsmount **, struct nfsmount *, int *, int *);
int nfsrv_dscreate(struct vnode *, struct vattr *, struct vattr *,
fhandle_t *, struct pnfsdsfile *, struct pnfsdsattr *, char *,
struct ucred *, NFSPROC_T *, struct vnode **);
int nfsrv_updatemdsattr(struct vnode *, struct nfsvattr *, NFSPROC_T *);
void nfsrv_killrpcs(struct nfsmount *);
int nfsrv_setacl(struct vnode *, NFSACL_T *, struct ucred *, NFSPROC_T *);
int nfsvno_seek(struct nfsrv_descript *, struct vnode *, u_long, off_t *, int,
bool *, struct ucred *, NFSPROC_T *);
int nfsvno_allocate(struct vnode *, off_t, off_t, struct ucred *, NFSPROC_T *);
int nfsvno_deallocate(struct vnode *, off_t, off_t, struct ucred *,
NFSPROC_T *);
int nfsvno_getxattr(struct vnode *, char *, uint32_t, struct ucred *,
uint64_t, int, struct thread *, struct mbuf **, struct mbuf **, int *);
int nfsvno_setxattr(struct vnode *, char *, int, struct mbuf *, char *,
struct ucred *, struct thread *);
int nfsvno_rmxattr(struct nfsrv_descript *, struct vnode *, char *,
struct ucred *, struct thread *);
int nfsvno_listxattr(struct vnode *, uint64_t, struct ucred *, struct thread *,
u_char **, uint32_t *, bool *);
void nfsm_trimtrailing(struct nfsrv_descript *, struct mbuf *, char *, int,
int);
bool nfsrv_checkwrongsec(struct nfsrv_descript *, int, enum vtype);
void nfsrv_checknospc(void);
/* nfs_commonkrpc.c */
int newnfs_nmcancelreqs(struct nfsmount *);
void newnfs_set_sigmask(struct thread *, sigset_t *);
void newnfs_restore_sigmask(struct thread *, sigset_t *);
int newnfs_msleep(struct thread *, void *, struct mtx *, int, char *, int);
int newnfs_request(struct nfsrv_descript *, struct nfsmount *,
struct nfsclient *, struct nfssockreq *, vnode_t, NFSPROC_T *,
struct ucred *, u_int32_t, u_int32_t, u_char *, int, u_int64_t *,
struct nfsclsession *);
int newnfs_connect(struct nfsmount *, struct nfssockreq *,
struct ucred *, NFSPROC_T *, int, bool, struct __rpc_client **);
void newnfs_disconnect(struct nfsmount *, struct nfssockreq *);
int newnfs_sigintr(struct nfsmount *, NFSPROC_T *);
/* nfs_nfsdkrpc.c */
int nfsrvd_addsock(struct file *);
int nfsrvd_nfsd(NFSPROC_T *, struct nfsd_nfsd_args *);
void nfsrvd_init(int);
/* nfs_clkrpc.c */
int nfscbd_addsock(struct file *);
int nfscbd_nfsd(NFSPROC_T *, struct nfsd_nfscbd_args *);
diff --git a/sys/fs/nfsclient/nfs_clrpcops.c b/sys/fs/nfsclient/nfs_clrpcops.c
index e7558b85a46f..8886ccc8429e 100644
--- a/sys/fs/nfsclient/nfs_clrpcops.c
+++ b/sys/fs/nfsclient/nfs_clrpcops.c
@@ -1,9131 +1,9131 @@
/*-
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 1989, 1993
* The Regents of the University of California. All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* Rick Macklem at The University of Guelph.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
/*
* Rpc op calls, generally called from the vnode op calls or through the
* buffer cache, for NFS v2, 3 and 4.
* These do not normally make any changes to vnode arguments or use
* structures that might change between the VFS variants. The returned
* arguments are all at the end, after the NFSPROC_T *p one.
*/
#include "opt_inet6.h"
#include <fs/nfs/nfsport.h>
#include <fs/nfsclient/nfs.h>
#include <sys/extattr.h>
#include <sys/sysctl.h>
#include <sys/taskqueue.h>
SYSCTL_DECL(_vfs_nfs);
static int nfsignore_eexist = 0;
SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
&nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");
static int nfscl_dssameconn = 0;
SYSCTL_INT(_vfs_nfs, OID_AUTO, dssameconn, CTLFLAG_RW,
&nfscl_dssameconn, 0, "Use same TCP connection to multiple DSs");
static uint64_t nfs_maxcopyrange = SSIZE_MAX;
SYSCTL_U64(_vfs_nfs, OID_AUTO, maxcopyrange, CTLFLAG_RW,
&nfs_maxcopyrange, 0, "Max size of a Copy so RPC times reasonable");
/*
* Global variables
*/
extern struct nfsstatsv1 nfsstatsv1;
extern int nfs_numnfscbd;
extern struct timeval nfsboottime;
extern u_int32_t newnfs_false, newnfs_true;
extern nfstype nfsv34_type[9];
extern int nfsrv_useacl;
extern char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
extern int nfscl_debuglevel;
extern int nfs_pnfsiothreads;
extern u_long sb_max_adj;
NFSCLSTATEMUTEX;
int nfstest_outofseq = 0;
int nfscl_assumeposixlocks = 1;
int nfscl_enablecallb = 0;
short nfsv4_cbport = NFSV4_CBPORT;
int nfstest_openallsetattr = 0;
#define DIRHDSIZ offsetof(struct dirent, d_name)
/*
* nfscl_getsameserver() can return one of three values:
* NFSDSP_USETHISSESSION - Use this session for the DS.
* NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
* session.
* NFSDSP_NOTFOUND - No matching server was found.
*/
enum nfsclds_state {
NFSDSP_USETHISSESSION = 0,
NFSDSP_SEQTHISSESSION = 1,
NFSDSP_NOTFOUND = 2,
};
/*
* Do a write RPC on a DS data file, using this structure for the arguments,
* so that this function can be executed by a separate kernel process.
*/
struct nfsclwritedsdorpc {
int done;
int inprog;
struct task tsk;
struct vnode *vp;
int iomode;
int must_commit;
nfsv4stateid_t *stateidp;
struct nfsclds *dsp;
uint64_t off;
int len;
#ifdef notyet
int advise;
#endif
struct nfsfh *fhp;
struct mbuf *m;
int vers;
int minorvers;
struct ucred *cred;
NFSPROC_T *p;
int err;
};
static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *);
static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *, void *);
static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
void *);
static int nfsrpc_deallocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
struct nfsvattr *, int *, struct ucred *, NFSPROC_T *, void *);
static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
struct nfsvattr *, struct nfsfh **, int *, int *, void *);
static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
int *, void *, int *);
static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
struct nfscllockowner *, u_int64_t, u_int64_t,
u_int32_t, struct ucred *, NFSPROC_T *, int);
static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
struct acl *, nfsv4stateid_t *, void *);
static int nfsrpc_layouterror(struct nfsmount *, uint8_t *, int, uint64_t,
uint64_t, nfsv4stateid_t *, struct ucred *, NFSPROC_T *, uint32_t,
uint32_t, char *);
static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
uint32_t, uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
struct ucred *, NFSPROC_T *);
static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
struct sockaddr_in6 *, sa_family_t, int, int, struct nfsclds **,
NFSPROC_T *);
static void nfscl_initsessionslots(struct nfsclsession *);
static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
NFSPROC_T *);
static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
struct nfsclds *, uint64_t, int, struct nfsfh *, int, int, int,
struct ucred *, NFSPROC_T *);
static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
struct nfsfh *, int, int, int, int, struct ucred *, NFSPROC_T *);
static int nfsio_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
static int nfsrpc_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
struct ucred *, NFSPROC_T *);
static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
struct nfsclds *, struct nfsclds **, uint32_t *);
static int nfsio_commitds(vnode_t, uint64_t, int, struct nfsclds *,
struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
NFSPROC_T *);
static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
#ifdef notyet
static int nfsio_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
NFSPROC_T *);
static int nfsrpc_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
#endif
static int nfsrpc_allocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
struct nfsvattr *, int *, struct ucred *, NFSPROC_T *, void *);
static void nfsrv_setuplayoutget(struct nfsrv_descript *, int, uint64_t,
uint64_t, uint64_t, nfsv4stateid_t *, int, int, int);
static int nfsrv_parseug(struct nfsrv_descript *, int, uid_t *, gid_t *,
NFSPROC_T *);
static int nfsrv_parselayoutget(struct nfsmount *, struct nfsrv_descript *,
nfsv4stateid_t *, int *, struct nfsclflayouthead *);
static int nfsrpc_getopenlayout(struct nfsmount *, vnode_t, u_int8_t *,
int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
struct nfscldeleg **, struct ucred *, NFSPROC_T *);
static int nfsrpc_getcreatelayout(vnode_t, char *, int, struct vattr *,
nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
struct nfsfh **, int *, int *, void *, int *);
static int nfsrpc_openlayoutrpc(struct nfsmount *, vnode_t, u_int8_t *,
int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
struct nfscldeleg **, nfsv4stateid_t *, int, int, int, int *,
struct nfsclflayouthead *, int *, struct ucred *, NFSPROC_T *);
static int nfsrpc_createlayout(vnode_t, char *, int, struct vattr *,
nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
struct nfsfh **, int *, int *, void *, int *, nfsv4stateid_t *,
int, int, int, int *, struct nfsclflayouthead *, int *);
static int nfsrpc_layoutget(struct nfsmount *, uint8_t *, int, int, uint64_t,
uint64_t, uint64_t, int, int, nfsv4stateid_t *, int *,
struct nfsclflayouthead *, struct ucred *, NFSPROC_T *, void *);
static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *,
int, nfsv4stateid_t *, int, uint32_t *, struct nfscllayout **,
struct nfsclflayouthead *, int, int, int *, struct ucred *, NFSPROC_T *);
static int nfsrpc_copyrpc(vnode_t, off_t, vnode_t, off_t, size_t *,
nfsv4stateid_t *, nfsv4stateid_t *, struct nfsvattr *, int *,
struct nfsvattr *, int *, bool, int *, struct ucred *, NFSPROC_T *);
static int nfsrpc_seekrpc(vnode_t, off_t *, nfsv4stateid_t *, bool *,
int, struct nfsvattr *, int *, struct ucred *);
static struct mbuf *nfsm_split(struct mbuf *, uint64_t);
int nfs_pnfsio(task_fn_t *, void *);
/*
* nfs null call from vfs.
*/
int
nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
{
int error;
struct nfsrv_descript nfsd, *nd = &nfsd;
NFSCL_REQSTART(nd, NFSPROC_NULL, vp);
error = nfscl_request(nd, vp, p, cred, NULL);
if (nd->nd_repstat && !error)
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* nfs access rpc op.
* For nfs version 3 and 4, use the access rpc to check accessibility. If file
* modes are changed on the server, accesses might still fail later.
*/
int
nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
{
int error;
u_int32_t mode, rmode;
if (acmode & VREAD)
mode = NFSACCESS_READ;
else
mode = 0;
if (vnode_vtype(vp) == VDIR) {
if (acmode & VWRITE)
mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
NFSACCESS_DELETE);
if (acmode & VEXEC)
mode |= NFSACCESS_LOOKUP;
} else {
if (acmode & VWRITE)
mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
if (acmode & VEXEC)
mode |= NFSACCESS_EXECUTE;
}
/*
* Now, just call nfsrpc_accessrpc() to do the actual RPC.
*/
error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode,
NULL);
/*
* The NFS V3 spec does not clarify whether or not
* the returned access bits can be a superset of
* the ones requested, so...
*/
if (!error && (rmode & mode) != mode)
error = EACCES;
return (error);
}
/*
* The actual rpc, separated out for Darwin.
*/
int
nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep,
void *stuff)
{
u_int32_t *tl;
u_int32_t supported, rmode;
int error;
struct nfsrv_descript nfsd, *nd = &nfsd;
nfsattrbit_t attrbits;
*attrflagp = 0;
supported = mode;
NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(mode);
if (nd->nd_flag & ND_NFSV4) {
/*
* And do a Getattr op.
*/
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
(void) nfsrv_putattrbit(nd, &attrbits);
}
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
if (nd->nd_flag & ND_NFSV3) {
error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
if (error)
goto nfsmout;
}
if (!nd->nd_repstat) {
if (nd->nd_flag & ND_NFSV4) {
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
supported = fxdr_unsigned(u_int32_t, *tl++);
} else {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
}
rmode = fxdr_unsigned(u_int32_t, *tl);
if (nd->nd_flag & ND_NFSV4)
error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
/*
* It's not obvious what should be done about
* unsupported access modes. For now, be paranoid
* and clear the unsupported ones.
*/
rmode &= supported;
*rmodep = rmode;
} else
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* nfs open rpc
*/
int
nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
{
struct nfsclopen *op;
struct nfscldeleg *dp;
struct nfsfh *nfhp;
struct nfsnode *np = VTONFS(vp);
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
u_int32_t mode, clidrev;
int ret, newone, error, expireret = 0, retrycnt;
/*
* For NFSv4, Open Ops are only done on Regular Files.
*/
if (vnode_vtype(vp) != VREG)
return (0);
mode = 0;
if (amode & FREAD)
mode |= NFSV4OPEN_ACCESSREAD;
if (amode & FWRITE)
mode |= NFSV4OPEN_ACCESSWRITE;
nfhp = np->n_fhp;
retrycnt = 0;
#ifdef notdef
{ char name[100]; int namel;
namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
bcopy(NFS4NODENAME(np->n_v4), name, namel);
name[namel] = '\0';
printf("rpcopen p=0x%x name=%s",p->p_pid,name);
if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
else printf(" fhl=0\n");
}
#endif
do {
dp = NULL;
error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
cred, p, NULL, &op, &newone, &ret, 1, true);
if (error) {
return (error);
}
if (nmp->nm_clp != NULL)
clidrev = nmp->nm_clp->nfsc_clientidrev;
else
clidrev = 0;
if (ret == NFSCLOPEN_DOOPEN) {
if (np->n_v4 != NULL) {
/*
* For the first attempt, try and get a layout, if
* pNFS is enabled for the mount.
*/
if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
nfs_numnfscbd == 0 ||
(np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
error = nfsrpc_openrpc(nmp, vp,
np->n_v4->n4_data,
np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
np->n_fhp->nfh_len, mode, op,
NFS4NODENAME(np->n_v4),
np->n_v4->n4_namelen,
&dp, 0, 0x0, cred, p, 0, 0);
else
error = nfsrpc_getopenlayout(nmp, vp,
np->n_v4->n4_data,
np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
np->n_fhp->nfh_len, mode, op,
NFS4NODENAME(np->n_v4),
np->n_v4->n4_namelen, &dp, cred, p);
if (dp != NULL) {
#ifdef APPLE
OSBitAndAtomic((int32_t)~NDELEGMOD, (UInt32 *)&np->n_flag);
#else
NFSLOCKNODE(np);
np->n_flag &= ~NDELEGMOD;
/*
* Invalidate the attribute cache, so that
* attributes that pre-date the issue of a
* delegation are not cached, since the
* cached attributes will remain valid while
* the delegation is held.
*/
NFSINVALATTRCACHE(np);
NFSUNLOCKNODE(np);
#endif
(void) nfscl_deleg(nmp->nm_mountp,
op->nfso_own->nfsow_clp,
nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
}
} else {
error = EIO;
}
newnfs_copyincred(cred, &op->nfso_cred);
} else if (ret == NFSCLOPEN_SETCRED)
/*
* This is a new local open on a delegation. It needs
* to have credentials so that an open can be done
* against the server during recovery.
*/
newnfs_copyincred(cred, &op->nfso_cred);
/*
* nfso_opencnt is the count of how many VOP_OPEN()s have
* been done on this Open successfully and a VOP_CLOSE()
* is expected for each of these.
* If error is non-zero, don't increment it, since the Open
* hasn't succeeded yet.
*/
if (!error) {
op->nfso_opencnt++;
if (NFSHASNFSV4N(nmp) && NFSHASONEOPENOWN(nmp)) {
NFSLOCKNODE(np);
np->n_openstateid = op;
NFSUNLOCKNODE(np);
}
}
nfscl_openrelease(nmp, op, error, newone);
if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_BADSESSION) {
(void) nfs_catnap(PZERO, error, "nfs_open");
} else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
&& clidrev != 0) {
expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
retrycnt++;
}
} while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_BADSESSION ||
((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
expireret == 0 && clidrev != 0 && retrycnt < 4));
if (error && retrycnt >= 4)
error = EIO;
return (error);
}
/*
* the actual open rpc
*/
int
nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
u_int8_t *name, int namelen, struct nfscldeleg **dpp,
int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
int syscred, int recursed)
{
u_int32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfscldeleg *dp, *ndp = NULL;
struct nfsvattr nfsva;
u_int32_t rflags, deleg;
nfsattrbit_t attrbits;
int error, ret, acesize, limitby;
struct nfsclsession *tsep;
dp = *dpp;
*dpp = NULL;
nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
tsep = nfsmnt_mdssession(nmp);
*tl++ = tsep->nfsess_clientid.lval[0];
*tl = tsep->nfsess_clientid.lval[1];
(void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
if (reclaim) {
*tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(delegtype);
} else {
if (dp != NULL) {
*tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
if (NFSHASNFSV4N(nmp))
*tl++ = 0;
else
*tl++ = dp->nfsdl_stateid.seqid;
*tl++ = dp->nfsdl_stateid.other[0];
*tl++ = dp->nfsdl_stateid.other[1];
*tl = dp->nfsdl_stateid.other[2];
} else {
*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
}
(void) nfsm_strtom(nd, name, namelen);
}
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
(void) nfsrv_putattrbit(nd, &attrbits);
if (syscred)
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error)
return (error);
NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
if (!nd->nd_repstat) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
6 * NFSX_UNSIGNED);
op->nfso_stateid.seqid = *tl++;
op->nfso_stateid.other[0] = *tl++;
op->nfso_stateid.other[1] = *tl++;
op->nfso_stateid.other[2] = *tl;
rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
if (error)
goto nfsmout;
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
deleg = fxdr_unsigned(u_int32_t, *tl);
if (deleg == NFSV4OPEN_DELEGATEREAD ||
deleg == NFSV4OPEN_DELEGATEWRITE) {
if (!(op->nfso_own->nfsow_clp->nfsc_flags &
NFSCLFLAGS_FIRSTDELEG))
op->nfso_own->nfsow_clp->nfsc_flags |=
(NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
ndp = malloc(
sizeof (struct nfscldeleg) + newfhlen,
M_NFSCLDELEG, M_WAITOK);
LIST_INIT(&ndp->nfsdl_owner);
LIST_INIT(&ndp->nfsdl_lock);
ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
ndp->nfsdl_fhlen = newfhlen;
NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
newnfs_copyincred(cred, &ndp->nfsdl_cred);
nfscl_lockinit(&ndp->nfsdl_rwlock);
NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
NFSX_UNSIGNED);
ndp->nfsdl_stateid.seqid = *tl++;
ndp->nfsdl_stateid.other[0] = *tl++;
ndp->nfsdl_stateid.other[1] = *tl++;
ndp->nfsdl_stateid.other[2] = *tl++;
ret = fxdr_unsigned(int, *tl);
if (deleg == NFSV4OPEN_DELEGATEWRITE) {
ndp->nfsdl_flags = NFSCLDL_WRITE;
/*
* Indicates how much the file can grow.
*/
NFSM_DISSECT(tl, u_int32_t *,
3 * NFSX_UNSIGNED);
limitby = fxdr_unsigned(int, *tl++);
switch (limitby) {
case NFSV4OPEN_LIMITSIZE:
ndp->nfsdl_sizelimit = fxdr_hyper(tl);
break;
case NFSV4OPEN_LIMITBLOCKS:
ndp->nfsdl_sizelimit =
fxdr_unsigned(u_int64_t, *tl++);
ndp->nfsdl_sizelimit *=
fxdr_unsigned(u_int64_t, *tl);
break;
default:
error = NFSERR_BADXDR;
goto nfsmout;
}
} else {
ndp->nfsdl_flags = NFSCLDL_READ;
}
if (ret)
ndp->nfsdl_flags |= NFSCLDL_RECALL;
- error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
- &acesize, p);
+ error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, false,
+ &ret, &acesize, p);
if (error)
goto nfsmout;
} else if (deleg != NFSV4OPEN_DELEGATENONE) {
error = NFSERR_BADXDR;
goto nfsmout;
}
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
NULL, NULL, NULL, p, cred);
if (error)
goto nfsmout;
if (ndp != NULL) {
ndp->nfsdl_change = nfsva.na_filerev;
ndp->nfsdl_modtime = nfsva.na_mtime;
ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
}
if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
do {
ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
cred, p);
if (ret == NFSERR_DELAY)
(void) nfs_catnap(PZERO, ret, "nfs_open");
} while (ret == NFSERR_DELAY);
error = ret;
}
if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
nfscl_assumeposixlocks)
op->nfso_posixlock = 1;
else
op->nfso_posixlock = 0;
/*
* If the server is handing out delegations, but we didn't
* get one because an OpenConfirm was required, try the
* Open again, to get a delegation. This is a harmless no-op,
* from a server's point of view.
*/
if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
(op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
&& !error && dp == NULL && ndp == NULL && !recursed) {
do {
ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
cred, p, syscred, 1);
if (ret == NFSERR_DELAY)
(void) nfs_catnap(PZERO, ret, "nfs_open2");
} while (ret == NFSERR_DELAY);
if (ret) {
if (ndp != NULL) {
free(ndp, M_NFSCLDELEG);
ndp = NULL;
}
if (ret == NFSERR_STALECLIENTID ||
ret == NFSERR_STALEDONTRECOVER ||
ret == NFSERR_BADSESSION)
error = ret;
}
}
}
if (nd->nd_repstat != 0 && error == 0)
error = nd->nd_repstat;
if (error == NFSERR_STALECLIENTID)
nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
nfsmout:
if (!error)
*dpp = ndp;
else if (ndp != NULL)
free(ndp, M_NFSCLDELEG);
m_freem(nd->nd_mrep);
return (error);
}
/*
* open downgrade rpc
*/
int
nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
struct ucred *cred, NFSPROC_T *p)
{
u_int32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
int error;
NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
if (NFSHASNFSV4N(VFSTONFS(vp->v_mount)))
*tl++ = 0;
else
*tl++ = op->nfso_stateid.seqid;
*tl++ = op->nfso_stateid.other[0];
*tl++ = op->nfso_stateid.other[1];
*tl++ = op->nfso_stateid.other[2];
*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
*tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
error = nfscl_request(nd, vp, p, cred, NULL);
if (error)
return (error);
NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
if (!nd->nd_repstat) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
op->nfso_stateid.seqid = *tl++;
op->nfso_stateid.other[0] = *tl++;
op->nfso_stateid.other[1] = *tl++;
op->nfso_stateid.other[2] = *tl;
}
if (nd->nd_repstat && error == 0)
error = nd->nd_repstat;
if (error == NFSERR_STALESTATEID)
nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* V4 Close operation.
*/
int
nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
{
struct nfsclclient *clp;
int error;
if (vnode_vtype(vp) != VREG)
return (0);
if (doclose)
error = nfscl_doclose(vp, &clp, p);
else {
error = nfscl_getclose(vp, &clp);
if (error == 0)
nfscl_clientrelease(clp);
}
return (error);
}
/*
* Close the open.
*/
int
nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p,
bool loop_on_delayed, bool freeop)
{
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfscllockowner *lp, *nlp;
struct nfscllock *lop, *nlop;
struct ucred *tcred;
u_int64_t off = 0, len = 0;
u_int32_t type = NFSV4LOCKT_READ;
int error, do_unlock, trycnt;
tcred = newnfs_getcred();
newnfs_copycred(&op->nfso_cred, tcred);
/*
* (Theoretically this could be done in the same
* compound as the close, but having multiple
* sequenced Ops in the same compound might be
* too scary for some servers.)
*/
if (op->nfso_posixlock) {
off = 0;
len = NFS64BITSSET;
type = NFSV4LOCKT_READ;
}
/*
* Since this function is only called from VOP_INACTIVE(), no
* other thread will be manipulating this Open. As such, the
* lock lists are not being changed by other threads, so it should
* be safe to do this without locking.
*/
LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
do_unlock = 1;
LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
if (op->nfso_posixlock == 0) {
off = lop->nfslo_first;
len = lop->nfslo_end - lop->nfslo_first;
if (lop->nfslo_type == F_WRLCK)
type = NFSV4LOCKT_WRITE;
else
type = NFSV4LOCKT_READ;
}
if (do_unlock) {
trycnt = 0;
do {
error = nfsrpc_locku(nd, nmp, lp, off,
len, type, tcred, p, 0);
if ((nd->nd_repstat == NFSERR_GRACE ||
nd->nd_repstat == NFSERR_DELAY) &&
error == 0)
(void) nfs_catnap(PZERO,
(int)nd->nd_repstat,
"nfs_close");
} while ((nd->nd_repstat == NFSERR_GRACE ||
nd->nd_repstat == NFSERR_DELAY) &&
error == 0 && trycnt++ < 5);
if (op->nfso_posixlock)
do_unlock = 0;
}
nfscl_freelock(lop, 0);
}
/*
* Do a ReleaseLockOwner.
* The lock owner name nfsl_owner may be used by other opens for
* other files but the lock_owner4 name that nfsrpc_rellockown()
* puts on the wire has the file handle for this file appended
* to it, so it can be done now.
*/
(void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
lp->nfsl_open->nfso_fhlen, tcred, p);
}
/*
* There could be other Opens for different files on the same
* OpenOwner, so locking is required.
*/
NFSLOCKCLSTATE();
nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
NFSUNLOCKCLSTATE();
do {
error = nfscl_tryclose(op, tcred, nmp, p, loop_on_delayed);
if (error == NFSERR_GRACE)
(void) nfs_catnap(PZERO, error, "nfs_close");
} while (error == NFSERR_GRACE);
NFSLOCKCLSTATE();
nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
nfscl_freelockowner(lp, 0);
if (freeop && error != NFSERR_DELAY)
nfscl_freeopen(op, 0, true);
NFSUNLOCKCLSTATE();
NFSFREECRED(tcred);
return (error);
}
/*
* The actual Close RPC.
*/
int
nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
int syscred)
{
u_int32_t *tl;
int error;
nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
op->nfso_fhlen, NULL, NULL, 0, 0);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
if (NFSHASNFSV4N(nmp)) {
*tl++ = 0;
*tl++ = 0;
} else {
*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
*tl++ = op->nfso_stateid.seqid;
}
*tl++ = op->nfso_stateid.other[0];
*tl++ = op->nfso_stateid.other[1];
*tl = op->nfso_stateid.other[2];
if (syscred)
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error)
return (error);
if (!NFSHASNFSV4N(nmp))
NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
if (nd->nd_repstat == 0)
NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
error = nd->nd_repstat;
if (!NFSHASNFSV4N(nmp) && error == NFSERR_STALESTATEID)
nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* V4 Open Confirm RPC.
*/
int
nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
{
u_int32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfsmount *nmp;
int error;
nmp = VFSTONFS(vp->v_mount);
if (NFSHASNFSV4N(nmp))
return (0); /* No confirmation for NFSv4.1. */
nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
0, 0);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
*tl++ = op->nfso_stateid.seqid;
*tl++ = op->nfso_stateid.other[0];
*tl++ = op->nfso_stateid.other[1];
*tl++ = op->nfso_stateid.other[2];
*tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
error = nfscl_request(nd, vp, p, cred, NULL);
if (error)
return (error);
NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
if (!nd->nd_repstat) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
op->nfso_stateid.seqid = *tl++;
op->nfso_stateid.other[0] = *tl++;
op->nfso_stateid.other[1] = *tl++;
op->nfso_stateid.other[2] = *tl;
}
error = nd->nd_repstat;
if (error == NFSERR_STALESTATEID)
nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
* when a mount has just occurred and when the server replies NFSERR_EXPIRED.
*/
int
nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
bool *retokp, struct ucred *cred, NFSPROC_T *p)
{
u_int32_t *tl;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
u_short port;
int error, isinet6 = 0, callblen;
nfsquad_t confirm;
static u_int32_t rev = 0;
struct nfsclds *dsp, *odsp;
struct in6_addr a6;
struct nfsclsession *tsep;
struct rpc_reconupcall recon;
struct nfscl_reconarg *rcp;
if (nfsboottime.tv_sec == 0)
NFSSETBOOTTIME(nfsboottime);
if (NFSHASNFSV4N(nmp)) {
error = NFSERR_BADSESSION;
odsp = dsp = NULL;
if (retokp != NULL) {
NFSLOCKMNT(nmp);
odsp = TAILQ_FIRST(&nmp->nm_sess);
NFSUNLOCKMNT(nmp);
}
if (odsp != NULL) {
/*
* When a session already exists, first try a
* CreateSession with the extant ClientID.
*/
dsp = malloc(sizeof(struct nfsclds) +
odsp->nfsclds_servownlen + 1, M_NFSCLDS,
M_WAITOK | M_ZERO);
dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
dsp->nfsclds_servownlen = odsp->nfsclds_servownlen;
dsp->nfsclds_sess.nfsess_clientid =
odsp->nfsclds_sess.nfsess_clientid;
dsp->nfsclds_sess.nfsess_sequenceid =
odsp->nfsclds_sess.nfsess_sequenceid;
dsp->nfsclds_flags = odsp->nfsclds_flags;
if (dsp->nfsclds_servownlen > 0)
memcpy(dsp->nfsclds_serverown,
odsp->nfsclds_serverown,
dsp->nfsclds_servownlen + 1);
mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
NULL, MTX_DEF);
nfscl_initsessionslots(&dsp->nfsclds_sess);
error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
&nmp->nm_sockreq, NULL,
dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
NFSCL_DEBUG(1, "create session for extant "
"ClientID=%d\n", error);
if (error != 0) {
nfscl_freenfsclds(dsp);
dsp = NULL;
/*
* If *retokp is true, return any error other
* than NFSERR_STALECLIENTID,
* NFSERR_BADSESSION or NFSERR_STALEDONTRECOVER
* so that nfscl_recover() will not loop.
*/
if (*retokp)
return (NFSERR_IO);
} else
*retokp = true;
} else if (retokp != NULL && *retokp)
return (NFSERR_IO);
if (error != 0) {
/*
* Either there was no previous session or the
* CreateSession attempt failed, so...
* do an ExchangeID followed by the CreateSession.
*/
clp->nfsc_rev = rev++;
error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq, 0,
NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp,
cred, p);
NFSCL_DEBUG(1, "aft exch=%d\n", error);
if (error == 0)
error = nfsrpc_createsession(nmp,
&dsp->nfsclds_sess, &nmp->nm_sockreq, NULL,
dsp->nfsclds_sess.nfsess_sequenceid, 1,
cred, p);
NFSCL_DEBUG(1, "aft createsess=%d\n", error);
}
if (error == 0) {
/*
* If the session supports a backchannel, set up
* the BindConnectionToSession call in the krpc
* so that it is done on a reconnection.
*/
if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0) {
rcp = mem_alloc(sizeof(*rcp));
rcp->minorvers = nmp->nm_minorvers;
memcpy(rcp->sessionid,
dsp->nfsclds_sess.nfsess_sessionid,
NFSX_V4SESSIONID);
recon.call = nfsrpc_bindconnsess;
recon.arg = rcp;
CLNT_CONTROL(nmp->nm_client, CLSET_RECONUPCALL,
&recon);
}
NFSLOCKMNT(nmp);
/*
* The old sessions cannot be safely free'd
* here, since they may still be used by
* in-progress RPCs.
*/
tsep = NULL;
if (TAILQ_FIRST(&nmp->nm_sess) != NULL)
tsep = NFSMNT_MDSSESSION(nmp);
TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
nfsclds_list);
/*
* Wake up RPCs waiting for a slot on the
* old session. These will then fail with
* NFSERR_BADSESSION and be retried with the
* new session by nfsv4_setsequence().
* Also wakeup() processes waiting for the
* new session.
*/
if (tsep != NULL)
wakeup(&tsep->nfsess_slots);
wakeup(&nmp->nm_sess);
NFSUNLOCKMNT(nmp);
} else if (dsp != NULL)
nfscl_freenfsclds(dsp);
if (error == 0 && reclaim == 0) {
error = nfsrpc_reclaimcomplete(nmp, cred, p);
NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
if (error == NFSERR_COMPLETEALREADY ||
error == NFSERR_NOTSUPP)
/* Ignore this error. */
error = 0;
}
return (error);
} else if (retokp != NULL && *retokp)
return (NFSERR_IO);
clp->nfsc_rev = rev++;
/*
* Allocate a single session structure for NFSv4.0, because some of
* the fields are used by NFSv4.0 although it doesn't do a session.
*/
dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
NFSLOCKMNT(nmp);
TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
tsep = NFSMNT_MDSSESSION(nmp);
NFSUNLOCKMNT(nmp);
nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(nfsboottime.tv_sec);
*tl = txdr_unsigned(clp->nfsc_rev);
(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
/*
* set up the callback address
*/
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFS_CALLBCKPROG);
callblen = strlen(nfsv4_callbackaddr);
if (callblen == 0)
cp = nfscl_getmyip(nmp, &a6, &isinet6);
if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
(callblen > 0 || cp != NULL)) {
port = htons(nfsv4_cbport);
cp2 = (u_int8_t *)&port;
#ifdef INET6
if ((callblen > 0 &&
strchr(nfsv4_callbackaddr, ':')) || isinet6) {
char ip6buf[INET6_ADDRSTRLEN], *ip6add;
(void) nfsm_strtom(nd, "tcp6", 4);
if (callblen == 0) {
ip6_sprintf(ip6buf, (struct in6_addr *)cp);
ip6add = ip6buf;
} else {
ip6add = nfsv4_callbackaddr;
}
snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
ip6add, cp2[0], cp2[1]);
} else
#endif
{
(void) nfsm_strtom(nd, "tcp", 3);
if (callblen == 0)
snprintf(addr, INET6_ADDRSTRLEN + 9,
"%d.%d.%d.%d.%d.%d", cp[0], cp[1],
cp[2], cp[3], cp2[0], cp2[1]);
else
snprintf(addr, INET6_ADDRSTRLEN + 9,
"%s.%d.%d", nfsv4_callbackaddr,
cp2[0], cp2[1]);
}
(void) nfsm_strtom(nd, addr, strlen(addr));
} else {
(void) nfsm_strtom(nd, "tcp", 3);
(void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
}
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(clp->nfsc_cbident);
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error)
return (error);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
tsep->nfsess_clientid.lval[0] = *tl++;
tsep->nfsess_clientid.lval[1] = *tl++;
confirm.lval[0] = *tl++;
confirm.lval[1] = *tl;
m_freem(nd->nd_mrep);
nd->nd_mrep = NULL;
/*
* and confirm it.
*/
nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
NULL, 0, 0);
NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
*tl++ = tsep->nfsess_clientid.lval[0];
*tl++ = tsep->nfsess_clientid.lval[1];
*tl++ = confirm.lval[0];
*tl = confirm.lval[1];
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error)
return (error);
m_freem(nd->nd_mrep);
nd->nd_mrep = NULL;
}
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* nfs getattr call.
*/
int
nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
struct nfsvattr *nap, void *stuff)
{
struct nfsrv_descript nfsd, *nd = &nfsd;
int error;
nfsattrbit_t attrbits;
NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
if (nd->nd_flag & ND_NFSV4) {
NFSGETATTR_ATTRBIT(&attrbits);
(void) nfsrv_putattrbit(nd, &attrbits);
}
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
if (!nd->nd_repstat)
error = nfsm_loadattr(nd, nap);
else
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* nfs getattr call with non-vnode arguments.
*/
int
nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
uint32_t *leasep)
{
struct nfsrv_descript nfsd, *nd = &nfsd;
int error, vers = NFS_VER2;
nfsattrbit_t attrbits;
nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0);
if (nd->nd_flag & ND_NFSV4) {
vers = NFS_VER4;
NFSGETATTR_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
(void) nfsrv_putattrbit(nd, &attrbits);
} else if (nd->nd_flag & ND_NFSV3) {
vers = NFS_VER3;
}
if (syscred)
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, vers, NULL, 1, xidp, NULL);
if (error)
return (error);
if (nd->nd_repstat == 0) {
if ((nd->nd_flag & ND_NFSV4) != 0)
error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
NULL, NULL);
else
error = nfsm_loadattr(nd, nap);
} else
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* Do an nfs setattr operation.
*/
int
nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
void *stuff)
{
int error, expireret = 0, openerr, retrycnt;
u_int32_t clidrev = 0, mode;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
struct nfsfh *nfhp;
nfsv4stateid_t stateid;
void *lckp;
if (nmp->nm_clp != NULL)
clidrev = nmp->nm_clp->nfsc_clientidrev;
if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
mode = NFSV4OPEN_ACCESSWRITE;
else
mode = NFSV4OPEN_ACCESSREAD;
retrycnt = 0;
do {
lckp = NULL;
openerr = 1;
if (NFSHASNFSV4(nmp)) {
nfhp = VTONFS(vp)->n_fhp;
error = nfscl_getstateid(vp, nfhp->nfh_fh,
nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
if (error && vnode_vtype(vp) == VREG &&
(mode == NFSV4OPEN_ACCESSWRITE ||
nfstest_openallsetattr)) {
/*
* No Open stateid, so try and open the file
* now.
*/
if (mode == NFSV4OPEN_ACCESSWRITE)
openerr = nfsrpc_open(vp, FWRITE, cred,
p);
else
openerr = nfsrpc_open(vp, FREAD, cred,
p);
if (!openerr)
(void) nfscl_getstateid(vp,
nfhp->nfh_fh, nfhp->nfh_len,
mode, 0, cred, p, &stateid, &lckp);
}
}
if (vap != NULL)
error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
rnap, attrflagp, stuff);
else
error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
stuff);
if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
NFSLOCKMNT(nmp);
nmp->nm_state |= NFSSTA_OPENMODE;
NFSUNLOCKMNT(nmp);
}
if (error == NFSERR_STALESTATEID)
nfscl_initiate_recovery(nmp->nm_clp);
if (lckp != NULL)
nfscl_lockderef(lckp);
if (!openerr)
(void) nfsrpc_close(vp, 0, p);
if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
(void) nfs_catnap(PZERO, error, "nfs_setattr");
} else if ((error == NFSERR_EXPIRED ||
error == NFSERR_BADSTATEID) && clidrev != 0) {
expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
}
retrycnt++;
} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_BADSESSION ||
(error == NFSERR_OLDSTATEID && retrycnt < 20) ||
((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
expireret == 0 && clidrev != 0 && retrycnt < 4) ||
(error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
retrycnt < 4));
if (error && retrycnt >= 4)
error = EIO;
return (error);
}
static int
nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
struct nfsvattr *rnap, int *attrflagp, void *stuff)
{
u_int32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
int error;
nfsattrbit_t attrbits;
*attrflagp = 0;
NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
if (nd->nd_flag & ND_NFSV4)
nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
vap->va_type = vnode_vtype(vp);
nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
if (nd->nd_flag & ND_NFSV3) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = newnfs_false;
} else if (nd->nd_flag & ND_NFSV4) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
(void) nfsrv_putattrbit(nd, &attrbits);
}
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, stuff);
if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
error = nfscl_postop_attr(nd, rnap, attrflagp, stuff);
m_freem(nd->nd_mrep);
if (nd->nd_repstat && !error)
error = nd->nd_repstat;
return (error);
}
/*
* nfs lookup rpc
*/
int
nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff,
uint32_t openmode)
{
uint32_t deleg, rflags, *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfsmount *nmp;
struct nfsnode *np;
struct nfsfh *nfhp;
nfsattrbit_t attrbits;
int error = 0, lookupp = 0, newone, ret, retop;
uint8_t own[NFSV4CL_LOCKNAMELEN];
struct nfsclopen *op;
struct nfscldeleg *ndp;
nfsv4stateid_t stateid;
*attrflagp = 0;
*dattrflagp = 0;
if (vnode_vtype(dvp) != VDIR)
return (ENOTDIR);
nmp = VFSTONFS(dvp->v_mount);
if (len > NFS_MAXNAMLEN)
return (ENAMETOOLONG);
if (NFSHASNFSV4(nmp) && len == 1 &&
name[0] == '.') {
/*
* Just return the current dir's fh.
*/
np = VTONFS(dvp);
nfhp = malloc(sizeof (struct nfsfh) +
np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
nfhp->nfh_len = np->n_fhp->nfh_len;
NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
*nfhpp = nfhp;
return (0);
}
if (NFSHASNFSV4(nmp) && len == 2 &&
name[0] == '.' && name[1] == '.') {
lookupp = 1;
openmode = 0;
NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
} else if (openmode != 0) {
NFSCL_REQSTART(nd, NFSPROC_LOOKUPOPEN, dvp);
nfsm_strtom(nd, name, len);
} else {
NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
(void) nfsm_strtom(nd, name, len);
}
if (nd->nd_flag & ND_NFSV4) {
NFSGETATTR_ATTRBIT(&attrbits);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV4OP_GETFH);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &attrbits);
if (openmode != 0) {
/* Test for a VREG file. */
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_VERIFY);
nfsrv_putattrbit(nd, &attrbits);
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSX_UNSIGNED);
*tl = vtonfsv34_type(VREG);
/* Attempt the Open for VREG. */
nfscl_filllockowner(NULL, own, F_POSIX);
NFSM_BUILD(tl, uint32_t *, 6 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV4OP_OPEN);
*tl++ = 0; /* seqid, ignored. */
*tl++ = txdr_unsigned(openmode);
*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
*tl++ = 0; /* ClientID, ignored. */
*tl = 0;
nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN);
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
*tl = txdr_unsigned(NFSV4OPEN_CLAIMFH);
}
}
error = nfscl_request(nd, dvp, p, cred, stuff);
if (error)
return (error);
ndp = NULL;
if (nd->nd_repstat) {
/*
* When an NFSv4 Lookupp returns ENOENT, it means that
* the lookup is at the root of an fs, so return this dir.
*/
if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
np = VTONFS(dvp);
nfhp = malloc(sizeof (struct nfsfh) +
np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
nfhp->nfh_len = np->n_fhp->nfh_len;
NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
*nfhpp = nfhp;
m_freem(nd->nd_mrep);
return (0);
}
if (nd->nd_flag & ND_NFSV3)
error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
ND_NFSV4) {
/* Load the directory attributes. */
error = nfsm_loadattr(nd, dnap);
if (error == 0)
*dattrflagp = 1;
else
goto nfsmout;
}
/* Check Lookup operation reply status. */
if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (*++tl != 0)
goto nfsmout;
}
/* Look for GetFH reply. */
if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (*++tl != 0)
goto nfsmout;
error = nfsm_getfh(nd, nfhpp);
if (error)
goto nfsmout;
}
/* Look for Getattr reply. */
if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (*++tl != 0)
goto nfsmout;
error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
if (error == 0)
/* Successfully got Lookup done. */
nd->nd_repstat = 0;
}
goto nfsmout;
}
if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
/* Load the directory attributes. */
error = nfsm_loadattr(nd, dnap);
if (error != 0)
goto nfsmout;
*dattrflagp = 1;
/* Skip over the Lookup and GetFH operation status values. */
NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
}
error = nfsm_getfh(nd, nfhpp);
if (error)
goto nfsmout;
error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
if (openmode != 0 && error == 0) {
NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID +
10 * NFSX_UNSIGNED);
tl += 4; /* Skip over Verify+Open status. */
stateid.seqid = *tl++;
stateid.other[0] = *tl++;
stateid.other[1] = *tl++;
stateid.other[2] = *tl;
rflags = fxdr_unsigned(uint32_t, *(tl + 6));
error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
if (error != 0)
goto nfsmout;
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
deleg = fxdr_unsigned(uint32_t, *tl);
if (deleg == NFSV4OPEN_DELEGATEREAD ||
deleg == NFSV4OPEN_DELEGATEWRITE) {
/*
* Just need to fill in the fields used by
* nfscl_trydelegreturn().
* Mark the mount point as acquiring
* delegations, so NFSPROC_LOOKUPOPEN will
* no longer be done.
*/
NFSLOCKMNT(nmp);
nmp->nm_privflag |= NFSMNTP_DELEGISSUED;
NFSUNLOCKMNT(nmp);
ndp = malloc(sizeof(struct nfscldeleg) +
(*nfhpp)->nfh_len, M_NFSCLDELEG, M_WAITOK);
ndp->nfsdl_fhlen = (*nfhpp)->nfh_len;
NFSBCOPY((*nfhpp)->nfh_fh, ndp->nfsdl_fh,
ndp->nfsdl_fhlen);
newnfs_copyincred(cred, &ndp->nfsdl_cred);
NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
ndp->nfsdl_stateid.seqid = *tl++;
ndp->nfsdl_stateid.other[0] = *tl++;
ndp->nfsdl_stateid.other[1] = *tl++;
ndp->nfsdl_stateid.other[2] = *tl++;
} else if (deleg != NFSV4OPEN_DELEGATENONE) {
error = NFSERR_BADXDR;
goto nfsmout;
}
ret = nfscl_open(dvp, (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len,
openmode, 0, cred, p, NULL, &op, &newone, &retop, 1, true);
if (ret != 0)
goto nfsmout;
if (newone != 0) {
op->nfso_stateid.seqid = stateid.seqid;
op->nfso_stateid.other[0] = stateid.other[0];
op->nfso_stateid.other[1] = stateid.other[1];
op->nfso_stateid.other[2] = stateid.other[2];
op->nfso_mode = openmode;
} else {
op->nfso_stateid.seqid = stateid.seqid;
if (retop == NFSCLOPEN_DOOPEN)
op->nfso_mode |= openmode;
}
if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
nfscl_assumeposixlocks)
op->nfso_posixlock = 1;
else
op->nfso_posixlock = 0;
nfscl_openrelease(nmp, op, 0, 0);
if (ndp != NULL) {
/*
* Since we do not have the vnode, we
* cannot invalidate cached attributes.
* Just return the delegation.
*/
nfscl_trydelegreturn(ndp, cred, nmp, p);
}
}
if ((nd->nd_flag & ND_NFSV3) && !error)
error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
nfsmout:
m_freem(nd->nd_mrep);
if (!error && nd->nd_repstat)
error = nd->nd_repstat;
free(ndp, M_NFSCLDELEG);
return (error);
}
/*
* Do a readlink rpc.
*/
int
nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
{
u_int32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfsnode *np = VTONFS(vp);
nfsattrbit_t attrbits;
int error, len, cangetattr = 1;
*attrflagp = 0;
NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
if (nd->nd_flag & ND_NFSV4) {
/*
* And do a Getattr op.
*/
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
(void) nfsrv_putattrbit(nd, &attrbits);
}
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
if (nd->nd_flag & ND_NFSV3)
error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
if (!nd->nd_repstat && !error) {
NFSM_STRSIZ(len, NFS_MAXPATHLEN);
/*
* This seems weird to me, but must have been added to
* FreeBSD for some reason. The only thing I can think of
* is that there was/is some server that replies with
* more link data than it should?
*/
if (len == NFS_MAXPATHLEN) {
NFSLOCKNODE(np);
if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
len = np->n_size;
cangetattr = 0;
}
NFSUNLOCKNODE(np);
}
error = nfsm_mbufuio(nd, uiop, len);
if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
}
if (nd->nd_repstat && !error)
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* Read operation.
*/
int
nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
{
int error, expireret = 0, retrycnt;
u_int32_t clidrev = 0;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
struct nfsnode *np = VTONFS(vp);
struct ucred *newcred;
struct nfsfh *nfhp = NULL;
nfsv4stateid_t stateid;
void *lckp;
if (nmp->nm_clp != NULL)
clidrev = nmp->nm_clp->nfsc_clientidrev;
newcred = cred;
if (NFSHASNFSV4(nmp)) {
nfhp = np->n_fhp;
newcred = NFSNEWCRED(cred);
}
retrycnt = 0;
do {
lckp = NULL;
if (NFSHASNFSV4(nmp))
(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
&lckp);
error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
attrflagp, stuff);
if (error == NFSERR_OPENMODE) {
NFSLOCKMNT(nmp);
nmp->nm_state |= NFSSTA_OPENMODE;
NFSUNLOCKMNT(nmp);
}
if (error == NFSERR_STALESTATEID)
nfscl_initiate_recovery(nmp->nm_clp);
if (lckp != NULL)
nfscl_lockderef(lckp);
if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
(void) nfs_catnap(PZERO, error, "nfs_read");
} else if ((error == NFSERR_EXPIRED ||
error == NFSERR_BADSTATEID) && clidrev != 0) {
expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
}
retrycnt++;
} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_BADSESSION ||
(error == NFSERR_OLDSTATEID && retrycnt < 20) ||
((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
expireret == 0 && clidrev != 0 && retrycnt < 4) ||
(error == NFSERR_OPENMODE && retrycnt < 4));
if (error && retrycnt >= 4)
error = EIO;
if (NFSHASNFSV4(nmp))
NFSFREECRED(newcred);
return (error);
}
/*
* The actual read RPC.
*/
static int
nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
int *attrflagp, void *stuff)
{
u_int32_t *tl;
int error = 0, len, retlen, tsiz, eof = 0;
struct nfsrv_descript nfsd;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
struct nfsrv_descript *nd = &nfsd;
int rsize;
off_t tmp_off;
*attrflagp = 0;
tsiz = uiop->uio_resid;
tmp_off = uiop->uio_offset + tsiz;
NFSLOCKMNT(nmp);
if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
NFSUNLOCKMNT(nmp);
return (EFBIG);
}
rsize = nmp->nm_rsize;
NFSUNLOCKMNT(nmp);
nd->nd_mrep = NULL;
while (tsiz > 0) {
*attrflagp = 0;
len = (tsiz > rsize) ? rsize : tsiz;
NFSCL_REQSTART(nd, NFSPROC_READ, vp);
if (nd->nd_flag & ND_NFSV4)
nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
if (nd->nd_flag & ND_NFSV2) {
*tl++ = txdr_unsigned(uiop->uio_offset);
*tl++ = txdr_unsigned(len);
*tl = 0;
} else {
txdr_hyper(uiop->uio_offset, tl);
*(tl + 2) = txdr_unsigned(len);
}
/*
* Since I can't do a Getattr for NFSv4 for Write, there
* doesn't seem any point in doing one here, either.
* (See the comment in nfsrpc_writerpc() for more info.)
*/
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
if (nd->nd_flag & ND_NFSV3) {
error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
} else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
error = nfsm_loadattr(nd, nap);
if (!error)
*attrflagp = 1;
}
if (nd->nd_repstat || error) {
if (!error)
error = nd->nd_repstat;
goto nfsmout;
}
if (nd->nd_flag & ND_NFSV3) {
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
eof = fxdr_unsigned(int, *(tl + 1));
} else if (nd->nd_flag & ND_NFSV4) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
eof = fxdr_unsigned(int, *tl);
}
NFSM_STRSIZ(retlen, len);
error = nfsm_mbufuio(nd, uiop, retlen);
if (error)
goto nfsmout;
m_freem(nd->nd_mrep);
nd->nd_mrep = NULL;
tsiz -= retlen;
if (!(nd->nd_flag & ND_NFSV2)) {
if (eof || retlen == 0)
tsiz = 0;
} else if (retlen < len)
tsiz = 0;
}
return (0);
nfsmout:
if (nd->nd_mrep != NULL)
m_freem(nd->nd_mrep);
return (error);
}
/*
* nfs write operation
* When called_from_strategy != 0, it should return EIO for an error that
* indicates recovery is in progress, so that the buffer will be left
* dirty and be written back to the server later. If it loops around,
* the recovery thread could get stuck waiting for the buffer and recovery
* will then deadlock.
*/
int
nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
void *stuff, int called_from_strategy)
{
int error, expireret = 0, retrycnt, nostateid;
u_int32_t clidrev = 0;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
struct nfsnode *np = VTONFS(vp);
struct ucred *newcred;
struct nfsfh *nfhp = NULL;
nfsv4stateid_t stateid;
void *lckp;
KASSERT(*must_commit >= 0 && *must_commit <= 2,
("nfsrpc_write: must_commit out of range=%d", *must_commit));
if (nmp->nm_clp != NULL)
clidrev = nmp->nm_clp->nfsc_clientidrev;
newcred = cred;
if (NFSHASNFSV4(nmp)) {
newcred = NFSNEWCRED(cred);
nfhp = np->n_fhp;
}
retrycnt = 0;
do {
lckp = NULL;
nostateid = 0;
if (NFSHASNFSV4(nmp)) {
(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
&lckp);
if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
stateid.other[2] == 0) {
nostateid = 1;
NFSCL_DEBUG(1, "stateid0 in write\n");
}
}
/*
* If there is no stateid for NFSv4, it means this is an
* extraneous write after close. Basically a poorly
* implemented buffer cache. Just don't do the write.
*/
if (nostateid)
error = 0;
else
error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
newcred, &stateid, p, nap, attrflagp, stuff);
if (error == NFSERR_STALESTATEID)
nfscl_initiate_recovery(nmp->nm_clp);
if (lckp != NULL)
nfscl_lockderef(lckp);
if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
(void) nfs_catnap(PZERO, error, "nfs_write");
} else if ((error == NFSERR_EXPIRED ||
error == NFSERR_BADSTATEID) && clidrev != 0) {
expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
}
retrycnt++;
} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
(error == NFSERR_OLDSTATEID && retrycnt < 20) ||
((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
expireret == 0 && clidrev != 0 && retrycnt < 4));
if (error != 0 && (retrycnt >= 4 ||
((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
error = EIO;
if (NFSHASNFSV4(nmp))
NFSFREECRED(newcred);
return (error);
}
/*
* The actual write RPC.
*/
static int
nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
{
u_int32_t *tl;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
struct nfsnode *np = VTONFS(vp);
int error = 0, len, tsiz, rlen, commit, committed = NFSWRITE_FILESYNC;
int wccflag = 0, wsize;
int32_t backup;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
nfsattrbit_t attrbits;
off_t tmp_off;
KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
*attrflagp = 0;
tsiz = uiop->uio_resid;
tmp_off = uiop->uio_offset + tsiz;
NFSLOCKMNT(nmp);
if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
NFSUNLOCKMNT(nmp);
return (EFBIG);
}
wsize = nmp->nm_wsize;
NFSUNLOCKMNT(nmp);
nd->nd_mrep = NULL; /* NFSv2 sometimes does a write with */
nd->nd_repstat = 0; /* uio_resid == 0, so the while is not done */
while (tsiz > 0) {
*attrflagp = 0;
len = (tsiz > wsize) ? wsize : tsiz;
NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
if (nd->nd_flag & ND_NFSV4) {
nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
txdr_hyper(uiop->uio_offset, tl);
tl += 2;
*tl++ = txdr_unsigned(*iomode);
*tl = txdr_unsigned(len);
} else if (nd->nd_flag & ND_NFSV3) {
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
txdr_hyper(uiop->uio_offset, tl);
tl += 2;
*tl++ = txdr_unsigned(len);
*tl++ = txdr_unsigned(*iomode);
*tl = txdr_unsigned(len);
} else {
u_int32_t x;
NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
/*
* Not sure why someone changed this, since the
* RFC clearly states that "beginoffset" and
* "totalcount" are ignored, but it wouldn't
* surprise me if there's a busted server out there.
*/
/* Set both "begin" and "current" to non-garbage. */
x = txdr_unsigned((u_int32_t)uiop->uio_offset);
*tl++ = x; /* "begin offset" */
*tl++ = x; /* "current offset" */
x = txdr_unsigned(len);
*tl++ = x; /* total to this offset */
*tl = x; /* size of this write */
}
nfsm_uiombuf(nd, uiop, len);
/*
* Although it is tempting to do a normal Getattr Op in the
* NFSv4 compound, the result can be a nearly hung client
* system if the Getattr asks for Owner and/or OwnerGroup.
* It occurs when the client can't map either the Owner or
* Owner_group name in the Getattr reply to a uid/gid. When
* there is a cache miss, the kernel does an upcall to the
* nfsuserd. Then, it can try and read the local /etc/passwd
* or /etc/group file. It can then block in getnewbuf(),
* waiting for dirty writes to be pushed to the NFS server.
* The only reason this doesn't result in a complete
* deadlock, is that the upcall times out and allows
* the write to complete. However, progress is so slow
* that it might just as well be deadlocked.
* As such, we get the rest of the attributes, but not
* Owner or Owner_group.
* nb: nfscl_loadattrcache() needs to be told that these
* partial attributes from a write rpc are being
* passed in, via a argument flag.
*/
if (nd->nd_flag & ND_NFSV4) {
NFSWRITEGETATTR_ATTRBIT(&attrbits);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &attrbits);
}
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
if (nd->nd_repstat) {
/*
* In case the rpc gets retried, roll
* the uio fields changed by nfsm_uiombuf()
* back.
*/
uiop->uio_offset -= len;
uiop->uio_resid += len;
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base - len;
uiop->uio_iov->iov_len += len;
}
if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
error = nfscl_wcc_data(nd, vp, nap, attrflagp,
&wccflag, stuff);
if (error)
goto nfsmout;
}
if (!nd->nd_repstat) {
if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
+ NFSX_VERF);
rlen = fxdr_unsigned(int, *tl++);
if (rlen == 0) {
error = NFSERR_IO;
goto nfsmout;
} else if (rlen < len) {
backup = len - rlen;
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base -
backup;
uiop->uio_iov->iov_len += backup;
uiop->uio_offset -= backup;
uiop->uio_resid += backup;
len = rlen;
}
commit = fxdr_unsigned(int, *tl++);
/*
* Return the lowest commitment level
* obtained by any of the RPCs.
*/
if (committed == NFSWRITE_FILESYNC)
committed = commit;
else if (committed == NFSWRITE_DATASYNC &&
commit == NFSWRITE_UNSTABLE)
committed = commit;
NFSLOCKMNT(nmp);
if (!NFSHASWRITEVERF(nmp)) {
NFSBCOPY((caddr_t)tl,
(caddr_t)&nmp->nm_verf[0],
NFSX_VERF);
NFSSETWRITEVERF(nmp);
} else if (NFSBCMP(tl, nmp->nm_verf,
NFSX_VERF) && *must_commit != 2) {
*must_commit = 1;
NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
}
NFSUNLOCKMNT(nmp);
}
if (nd->nd_flag & ND_NFSV4)
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
error = nfsm_loadattr(nd, nap);
if (!error)
*attrflagp = NFS_LATTR_NOSHRINK;
}
} else {
error = nd->nd_repstat;
}
if (error)
goto nfsmout;
NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
m_freem(nd->nd_mrep);
nd->nd_mrep = NULL;
tsiz -= len;
}
nfsmout:
if (nd->nd_mrep != NULL)
m_freem(nd->nd_mrep);
*iomode = committed;
if (nd->nd_repstat && !error)
error = nd->nd_repstat;
return (error);
}
/*
* Do an nfs deallocate operation.
*/
int
nfsrpc_deallocate(vnode_t vp, off_t offs, off_t len, struct nfsvattr *nap,
int *attrflagp, struct ucred *cred, NFSPROC_T *p, void *stuff)
{
int error, expireret = 0, openerr, retrycnt;
uint32_t clidrev = 0;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
struct nfsfh *nfhp;
nfsv4stateid_t stateid;
void *lckp;
if (nmp->nm_clp != NULL)
clidrev = nmp->nm_clp->nfsc_clientidrev;
retrycnt = 0;
do {
lckp = NULL;
openerr = 1;
nfhp = VTONFS(vp)->n_fhp;
error = nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
if (error != 0) {
/*
* No Open stateid, so try and open the file
* now.
*/
openerr = nfsrpc_open(vp, FWRITE, cred, p);
if (openerr == 0)
nfscl_getstateid(vp, nfhp->nfh_fh,
nfhp->nfh_len, NFSV4OPEN_ACCESSWRITE, 0,
cred, p, &stateid, &lckp);
}
error = nfsrpc_deallocaterpc(vp, offs, len, &stateid, nap,
attrflagp, cred, p, stuff);
if (error == NFSERR_STALESTATEID)
nfscl_initiate_recovery(nmp->nm_clp);
if (lckp != NULL)
nfscl_lockderef(lckp);
if (openerr == 0)
nfsrpc_close(vp, 0, p);
if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
(void) nfs_catnap(PZERO, error, "nfs_deallocate");
} else if ((error == NFSERR_EXPIRED ||
error == NFSERR_BADSTATEID) && clidrev != 0) {
expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
}
retrycnt++;
} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_BADSESSION ||
(error == NFSERR_OLDSTATEID && retrycnt < 20) ||
((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
expireret == 0 && clidrev != 0 && retrycnt < 4));
if (error && retrycnt >= 4)
error = EIO;
return (error);
}
/*
* The actual deallocate RPC.
*/
static int
nfsrpc_deallocaterpc(vnode_t vp, off_t offs, off_t len,
nfsv4stateid_t *stateidp, struct nfsvattr *nap, int *attrflagp,
struct ucred *cred, NFSPROC_T *p, void *stuff)
{
uint32_t *tl;
struct nfsnode *np = VTONFS(vp);
int error, wccflag;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
nfsattrbit_t attrbits;
*attrflagp = 0;
NFSCL_REQSTART(nd, NFSPROC_DEALLOCATE, vp);
nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
txdr_hyper(offs, tl);
tl += 2;
txdr_hyper(len, tl);
NFSWRITEGETATTR_ATTRBIT(&attrbits);
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
nfsrv_putattrbit(nd, &attrbits);
error = nfscl_request(nd, vp, p, cred, stuff);
if (error != 0)
return (error);
wccflag = 0;
error = nfscl_wcc_data(nd, vp, nap, attrflagp, &wccflag, stuff);
if (error != 0)
goto nfsmout;
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
error = nfsm_loadattr(nd, nap);
if (error != 0)
goto nfsmout;
*attrflagp = NFS_LATTR_NOSHRINK;
}
NFSWRITERPC_SETTIME(wccflag, np, nap, 1);
nfsmout:
m_freem(nd->nd_mrep);
if (nd->nd_repstat != 0 && error == 0)
error = nd->nd_repstat;
return (error);
}
/*
* nfs mknod rpc
* For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
* mode set to specify the file type and the size field for rdev.
*/
int
nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
int *attrflagp, int *dattrflagp, void *dstuff)
{
u_int32_t *tl;
int error = 0;
struct nfsrv_descript nfsd, *nd = &nfsd;
nfsattrbit_t attrbits;
*nfhpp = NULL;
*attrflagp = 0;
*dattrflagp = 0;
if (namelen > NFS_MAXNAMLEN)
return (ENAMETOOLONG);
NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
if (nd->nd_flag & ND_NFSV4) {
if (vtyp == VBLK || vtyp == VCHR) {
NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
*tl++ = vtonfsv34_type(vtyp);
*tl++ = txdr_unsigned(NFSMAJOR(rdev));
*tl = txdr_unsigned(NFSMINOR(rdev));
} else {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = vtonfsv34_type(vtyp);
}
}
(void) nfsm_strtom(nd, name, namelen);
if (nd->nd_flag & ND_NFSV3) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = vtonfsv34_type(vtyp);
}
if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
nfscl_fillsattr(nd, vap, dvp, 0, 0);
if ((nd->nd_flag & ND_NFSV3) &&
(vtyp == VCHR || vtyp == VBLK)) {
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSMAJOR(rdev));
*tl = txdr_unsigned(NFSMINOR(rdev));
}
if (nd->nd_flag & ND_NFSV4) {
NFSGETATTR_ATTRBIT(&attrbits);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV4OP_GETFH);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &attrbits);
}
if (nd->nd_flag & ND_NFSV2)
nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
error = nfscl_request(nd, dvp, p, cred, dstuff);
if (error)
return (error);
if (nd->nd_flag & ND_NFSV4)
error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
if (!nd->nd_repstat) {
if (nd->nd_flag & ND_NFSV4) {
NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
if (error)
goto nfsmout;
}
error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
if (error)
goto nfsmout;
}
if (nd->nd_flag & ND_NFSV3)
error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
if (!error && nd->nd_repstat)
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* nfs file create call
* Mostly just call the approriate routine. (I separated out v4, so that
* error recovery wouldn't be as difficult.)
*/
int
nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
int *attrflagp, int *dattrflagp, void *dstuff)
{
int error = 0, newone, expireret = 0, retrycnt, unlocked;
struct nfsclowner *owp;
struct nfscldeleg *dp;
struct nfsmount *nmp = VFSTONFS(dvp->v_mount);
u_int32_t clidrev;
if (NFSHASNFSV4(nmp)) {
retrycnt = 0;
do {
dp = NULL;
error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
NULL, 1, true);
if (error)
return (error);
if (nmp->nm_clp != NULL)
clidrev = nmp->nm_clp->nfsc_clientidrev;
else
clidrev = 0;
if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
nfs_numnfscbd == 0 || retrycnt > 0)
error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
attrflagp, dattrflagp, dstuff, &unlocked);
else
error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
attrflagp, dattrflagp, dstuff, &unlocked);
/*
* There is no need to invalidate cached attributes here,
* since new post-delegation issue attributes are always
* returned by nfsrpc_createv4() and these will update the
* attribute cache.
*/
if (dp != NULL)
(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
(*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_BADSESSION) {
(void) nfs_catnap(PZERO, error, "nfs_open");
} else if ((error == NFSERR_EXPIRED ||
error == NFSERR_BADSTATEID) && clidrev != 0) {
expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
retrycnt++;
}
} while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_BADSESSION ||
((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
expireret == 0 && clidrev != 0 && retrycnt < 4));
if (error && retrycnt >= 4)
error = EIO;
} else {
error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
dstuff);
}
return (error);
}
/*
* The create rpc for v2 and 3.
*/
static int
nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
int *attrflagp, int *dattrflagp, void *dstuff)
{
u_int32_t *tl;
int error = 0;
struct nfsrv_descript nfsd, *nd = &nfsd;
*nfhpp = NULL;
*attrflagp = 0;
*dattrflagp = 0;
if (namelen > NFS_MAXNAMLEN)
return (ENAMETOOLONG);
NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
(void) nfsm_strtom(nd, name, namelen);
if (nd->nd_flag & ND_NFSV3) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
if (fmode & O_EXCL) {
*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
*tl++ = cverf.lval[0];
*tl = cverf.lval[1];
} else {
*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
nfscl_fillsattr(nd, vap, dvp, 0, 0);
}
} else {
nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
}
error = nfscl_request(nd, dvp, p, cred, dstuff);
if (error)
return (error);
if (nd->nd_repstat == 0) {
error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
if (error)
goto nfsmout;
}
if (nd->nd_flag & ND_NFSV3)
error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
if (nd->nd_repstat != 0 && error == 0)
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
static int
nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
int *dattrflagp, void *dstuff, int *unlockedp)
{
u_int32_t *tl;
int error = 0, deleg, newone, ret, acesize, limitby;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfsclopen *op;
struct nfscldeleg *dp = NULL;
struct nfsnode *np;
struct nfsfh *nfhp;
nfsattrbit_t attrbits;
nfsv4stateid_t stateid;
u_int32_t rflags;
struct nfsmount *nmp;
struct nfsclsession *tsep;
nmp = VFSTONFS(dvp->v_mount);
np = VTONFS(dvp);
*unlockedp = 0;
*nfhpp = NULL;
*dpp = NULL;
*attrflagp = 0;
*dattrflagp = 0;
if (namelen > NFS_MAXNAMLEN)
return (ENAMETOOLONG);
NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
/*
* For V4, this is actually an Open op.
*/
NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(owp->nfsow_seqid);
*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
NFSV4OPEN_ACCESSREAD);
*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
tsep = nfsmnt_mdssession(nmp);
*tl++ = tsep->nfsess_clientid.lval[0];
*tl = tsep->nfsess_clientid.lval[1];
(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
if (fmode & O_EXCL) {
if (NFSHASNFSV4N(nmp)) {
if (NFSHASSESSPERSIST(nmp)) {
/* Use GUARDED for persistent sessions. */
*tl = txdr_unsigned(NFSCREATE_GUARDED);
nfscl_fillsattr(nd, vap, dvp, 0, 0);
} else {
/* Otherwise, use EXCLUSIVE4_1. */
*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
*tl++ = cverf.lval[0];
*tl = cverf.lval[1];
nfscl_fillsattr(nd, vap, dvp, 0, 0);
}
} else {
/* NFSv4.0 */
*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
*tl++ = cverf.lval[0];
*tl = cverf.lval[1];
}
} else {
*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
nfscl_fillsattr(nd, vap, dvp, 0, 0);
}
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
(void) nfsm_strtom(nd, name, namelen);
/* Get the new file's handle and attributes. */
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV4OP_GETFH);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
(void) nfsrv_putattrbit(nd, &attrbits);
/* Get the directory's post-op attributes. */
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_PUTFH);
(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &attrbits);
error = nfscl_request(nd, dvp, p, cred, dstuff);
if (error)
return (error);
NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
6 * NFSX_UNSIGNED);
stateid.seqid = *tl++;
stateid.other[0] = *tl++;
stateid.other[1] = *tl++;
stateid.other[2] = *tl;
rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
if (error)
goto nfsmout;
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
deleg = fxdr_unsigned(int, *tl);
if (deleg == NFSV4OPEN_DELEGATEREAD ||
deleg == NFSV4OPEN_DELEGATEWRITE) {
if (!(owp->nfsow_clp->nfsc_flags &
NFSCLFLAGS_FIRSTDELEG))
owp->nfsow_clp->nfsc_flags |=
(NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
dp = malloc(
sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
M_NFSCLDELEG, M_WAITOK);
LIST_INIT(&dp->nfsdl_owner);
LIST_INIT(&dp->nfsdl_lock);
dp->nfsdl_clp = owp->nfsow_clp;
newnfs_copyincred(cred, &dp->nfsdl_cred);
nfscl_lockinit(&dp->nfsdl_rwlock);
NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
NFSX_UNSIGNED);
dp->nfsdl_stateid.seqid = *tl++;
dp->nfsdl_stateid.other[0] = *tl++;
dp->nfsdl_stateid.other[1] = *tl++;
dp->nfsdl_stateid.other[2] = *tl++;
ret = fxdr_unsigned(int, *tl);
if (deleg == NFSV4OPEN_DELEGATEWRITE) {
dp->nfsdl_flags = NFSCLDL_WRITE;
/*
* Indicates how much the file can grow.
*/
NFSM_DISSECT(tl, u_int32_t *,
3 * NFSX_UNSIGNED);
limitby = fxdr_unsigned(int, *tl++);
switch (limitby) {
case NFSV4OPEN_LIMITSIZE:
dp->nfsdl_sizelimit = fxdr_hyper(tl);
break;
case NFSV4OPEN_LIMITBLOCKS:
dp->nfsdl_sizelimit =
fxdr_unsigned(u_int64_t, *tl++);
dp->nfsdl_sizelimit *=
fxdr_unsigned(u_int64_t, *tl);
break;
default:
error = NFSERR_BADXDR;
goto nfsmout;
}
} else {
dp->nfsdl_flags = NFSCLDL_READ;
}
if (ret)
dp->nfsdl_flags |= NFSCLDL_RECALL;
- error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
- &acesize, p);
+ error = nfsrv_dissectace(nd, &dp->nfsdl_ace, false,
+ &ret, &acesize, p);
if (error)
goto nfsmout;
} else if (deleg != NFSV4OPEN_DELEGATENONE) {
error = NFSERR_BADXDR;
goto nfsmout;
}
error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
if (error)
goto nfsmout;
/* Get rid of the PutFH and Getattr status values. */
NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
/* Load the directory attributes. */
error = nfsm_loadattr(nd, dnap);
if (error)
goto nfsmout;
*dattrflagp = 1;
if (dp != NULL && *attrflagp) {
dp->nfsdl_change = nnap->na_filerev;
dp->nfsdl_modtime = nnap->na_mtime;
dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
}
/*
* We can now complete the Open state.
*/
nfhp = *nfhpp;
if (dp != NULL) {
dp->nfsdl_fhlen = nfhp->nfh_len;
NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
}
/*
* Get an Open structure that will be
* attached to the OpenOwner, acquired already.
*/
error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
(NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
cred, p, NULL, &op, &newone, NULL, 0, false);
if (error)
goto nfsmout;
op->nfso_stateid = stateid;
newnfs_copyincred(cred, &op->nfso_cred);
if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
do {
ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
nfhp->nfh_len, op, cred, p);
if (ret == NFSERR_DELAY)
(void) nfs_catnap(PZERO, ret, "nfs_create");
} while (ret == NFSERR_DELAY);
error = ret;
}
/*
* If the server is handing out delegations, but we didn't
* get one because an OpenConfirm was required, try the
* Open again, to get a delegation. This is a harmless no-op,
* from a server's point of view.
*/
if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
(owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
!error && dp == NULL) {
do {
ret = nfsrpc_openrpc(VFSTONFS(dvp->v_mount), dvp,
np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
nfhp->nfh_fh, nfhp->nfh_len,
(NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
if (ret == NFSERR_DELAY)
(void) nfs_catnap(PZERO, ret, "nfs_crt2");
} while (ret == NFSERR_DELAY);
if (ret) {
if (dp != NULL) {
free(dp, M_NFSCLDELEG);
dp = NULL;
}
if (ret == NFSERR_STALECLIENTID ||
ret == NFSERR_STALEDONTRECOVER ||
ret == NFSERR_BADSESSION)
error = ret;
}
}
nfscl_openrelease(nmp, op, error, newone);
*unlockedp = 1;
}
if (nd->nd_repstat != 0 && error == 0)
error = nd->nd_repstat;
if (error == NFSERR_STALECLIENTID)
nfscl_initiate_recovery(owp->nfsow_clp);
nfsmout:
if (!error)
*dpp = dp;
else if (dp != NULL)
free(dp, M_NFSCLDELEG);
m_freem(nd->nd_mrep);
return (error);
}
/*
* Nfs remove rpc
*/
int
nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
void *dstuff)
{
u_int32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfsnode *np;
struct nfsmount *nmp;
nfsv4stateid_t dstateid;
int error, ret = 0, i;
*dattrflagp = 0;
if (namelen > NFS_MAXNAMLEN)
return (ENAMETOOLONG);
nmp = VFSTONFS(dvp->v_mount);
tryagain:
if (NFSHASNFSV4(nmp) && ret == 0) {
ret = nfscl_removedeleg(vp, p, &dstateid);
if (ret == 1) {
NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
NFSX_UNSIGNED);
if (NFSHASNFSV4N(nmp))
*tl++ = 0;
else
*tl++ = dstateid.seqid;
*tl++ = dstateid.other[0];
*tl++ = dstateid.other[1];
*tl++ = dstateid.other[2];
*tl = txdr_unsigned(NFSV4OP_PUTFH);
np = VTONFS(dvp);
(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
np->n_fhp->nfh_len, 0);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_REMOVE);
}
} else {
ret = 0;
}
if (ret == 0)
NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
(void) nfsm_strtom(nd, name, namelen);
error = nfscl_request(nd, dvp, p, cred, dstuff);
if (error)
return (error);
if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
/* For NFSv4, parse out any Delereturn replies. */
if (ret > 0 && nd->nd_repstat != 0 &&
(nd->nd_flag & ND_NOMOREDATA)) {
/*
* If the Delegreturn failed, try again without
* it. The server will Recall, as required.
*/
m_freem(nd->nd_mrep);
goto tryagain;
}
for (i = 0; i < (ret * 2); i++) {
if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
ND_NFSV4) {
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
if (*(tl + 1))
nd->nd_flag |= ND_NOMOREDATA;
}
}
error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
}
if (nd->nd_repstat && !error)
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* Do an nfs rename rpc.
*/
int
nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
{
u_int32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfsmount *nmp;
struct nfsnode *np;
nfsattrbit_t attrbits;
nfsv4stateid_t fdstateid, tdstateid;
int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
*fattrflagp = 0;
*tattrflagp = 0;
nmp = VFSTONFS(fdvp->v_mount);
if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
return (ENAMETOOLONG);
tryagain:
if (NFSHASNFSV4(nmp) && ret == 0) {
ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
&tdstateid, &gottd, p);
if (gotfd && gottd) {
NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
} else if (gotfd) {
NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
} else if (gottd) {
NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
}
if (gotfd) {
NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
if (NFSHASNFSV4N(nmp))
*tl++ = 0;
else
*tl++ = fdstateid.seqid;
*tl++ = fdstateid.other[0];
*tl++ = fdstateid.other[1];
*tl = fdstateid.other[2];
if (gottd) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_PUTFH);
np = VTONFS(tvp);
(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
np->n_fhp->nfh_len, 0);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
}
}
if (gottd) {
NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
if (NFSHASNFSV4N(nmp))
*tl++ = 0;
else
*tl++ = tdstateid.seqid;
*tl++ = tdstateid.other[0];
*tl++ = tdstateid.other[1];
*tl = tdstateid.other[2];
}
if (ret > 0) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_PUTFH);
np = VTONFS(fdvp);
(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
np->n_fhp->nfh_len, 0);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_SAVEFH);
}
} else {
ret = 0;
}
if (ret == 0)
NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
if (nd->nd_flag & ND_NFSV4) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSWCCATTR_ATTRBIT(&attrbits);
(void) nfsrv_putattrbit(nd, &attrbits);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_PUTFH);
(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
VTONFS(tdvp)->n_fhp->nfh_len, 0);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &attrbits);
nd->nd_flag |= ND_V4WCCATTR;
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_RENAME);
}
(void) nfsm_strtom(nd, fnameptr, fnamelen);
if (!(nd->nd_flag & ND_NFSV4))
(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
VTONFS(tdvp)->n_fhp->nfh_len, 0);
(void) nfsm_strtom(nd, tnameptr, tnamelen);
error = nfscl_request(nd, fdvp, p, cred, fstuff);
if (error)
return (error);
if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
/* For NFSv4, parse out any Delereturn replies. */
if (ret > 0 && nd->nd_repstat != 0 &&
(nd->nd_flag & ND_NOMOREDATA)) {
/*
* If the Delegreturn failed, try again without
* it. The server will Recall, as required.
*/
m_freem(nd->nd_mrep);
goto tryagain;
}
for (i = 0; i < (ret * 2); i++) {
if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
ND_NFSV4) {
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
if (*(tl + 1)) {
if (i == 0 && ret > 1) {
/*
* If the Delegreturn failed, try again
* without it. The server will Recall, as
* required.
* If ret > 1, the first iteration of this
* loop is the second DelegReturn result.
*/
m_freem(nd->nd_mrep);
goto tryagain;
} else {
nd->nd_flag |= ND_NOMOREDATA;
}
}
}
}
/* Now, the first wcc attribute reply. */
if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
if (*(tl + 1))
nd->nd_flag |= ND_NOMOREDATA;
}
error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
fstuff);
/* and the second wcc attribute reply. */
if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
!error) {
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
if (*(tl + 1))
nd->nd_flag |= ND_NOMOREDATA;
}
if (!error)
error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
NULL, tstuff);
}
if (nd->nd_repstat && !error)
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* nfs hard link create rpc
*/
int
nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
{
u_int32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
nfsattrbit_t attrbits;
int error = 0;
*attrflagp = 0;
*dattrflagp = 0;
if (namelen > NFS_MAXNAMLEN)
return (ENAMETOOLONG);
NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
if (nd->nd_flag & ND_NFSV4) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_PUTFH);
}
(void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
VTONFS(dvp)->n_fhp->nfh_len, 0);
if (nd->nd_flag & ND_NFSV4) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSWCCATTR_ATTRBIT(&attrbits);
(void) nfsrv_putattrbit(nd, &attrbits);
nd->nd_flag |= ND_V4WCCATTR;
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_LINK);
}
(void) nfsm_strtom(nd, name, namelen);
error = nfscl_request(nd, vp, p, cred, dstuff);
if (error)
return (error);
if (nd->nd_flag & ND_NFSV3) {
error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
if (!error)
error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
NULL, dstuff);
} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
/*
* First, parse out the PutFH and Getattr result.
*/
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
if (!(*(tl + 1)))
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
if (*(tl + 1))
nd->nd_flag |= ND_NOMOREDATA;
/*
* Get the pre-op attributes.
*/
error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
}
if (nd->nd_repstat && !error)
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* nfs symbolic link create rpc
*/
int
nfsrpc_symlink(vnode_t dvp, char *name, int namelen, const char *target,
struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
int *dattrflagp, void *dstuff)
{
u_int32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfsmount *nmp;
int slen, error = 0;
*nfhpp = NULL;
*attrflagp = 0;
*dattrflagp = 0;
nmp = VFSTONFS(dvp->v_mount);
slen = strlen(target);
if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
return (ENAMETOOLONG);
NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
if (nd->nd_flag & ND_NFSV4) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFLNK);
(void) nfsm_strtom(nd, target, slen);
}
(void) nfsm_strtom(nd, name, namelen);
if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
nfscl_fillsattr(nd, vap, dvp, 0, 0);
if (!(nd->nd_flag & ND_NFSV4))
(void) nfsm_strtom(nd, target, slen);
if (nd->nd_flag & ND_NFSV2)
nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
error = nfscl_request(nd, dvp, p, cred, dstuff);
if (error)
return (error);
if (nd->nd_flag & ND_NFSV4)
error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
if ((nd->nd_flag & ND_NFSV3) && !error) {
if (!nd->nd_repstat)
error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
if (!error)
error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
NULL, dstuff);
}
if (nd->nd_repstat && !error)
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
/*
* Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
* Only do this if vfs.nfs.ignore_eexist is set.
* Never do this for NFSv4.1 or later minor versions, since sessions
* should guarantee "exactly once" RPC semantics.
*/
if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
nmp->nm_minorvers == 0))
error = 0;
return (error);
}
/*
* nfs make dir rpc
*/
int
nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
int *dattrflagp, void *dstuff)
{
u_int32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
nfsattrbit_t attrbits;
int error = 0;
struct nfsfh *fhp;
struct nfsmount *nmp;
*nfhpp = NULL;
*attrflagp = 0;
*dattrflagp = 0;
nmp = VFSTONFS(dvp->v_mount);
fhp = VTONFS(dvp)->n_fhp;
if (namelen > NFS_MAXNAMLEN)
return (ENAMETOOLONG);
NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
if (nd->nd_flag & ND_NFSV4) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFDIR);
}
(void) nfsm_strtom(nd, name, namelen);
nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
if (nd->nd_flag & ND_NFSV4) {
NFSGETATTR_ATTRBIT(&attrbits);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV4OP_GETFH);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &attrbits);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_PUTFH);
(void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &attrbits);
}
error = nfscl_request(nd, dvp, p, cred, dstuff);
if (error)
return (error);
if (nd->nd_flag & ND_NFSV4)
error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
if (!nd->nd_repstat && !error) {
if (nd->nd_flag & ND_NFSV4) {
NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
}
if (!error)
error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
/* Get rid of the PutFH and Getattr status values. */
NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
/* Load the directory attributes. */
error = nfsm_loadattr(nd, dnap);
if (error == 0)
*dattrflagp = 1;
}
}
if ((nd->nd_flag & ND_NFSV3) && !error)
error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
if (nd->nd_repstat && !error)
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
/*
* Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
* Only do this if vfs.nfs.ignore_eexist is set.
* Never do this for NFSv4.1 or later minor versions, since sessions
* should guarantee "exactly once" RPC semantics.
*/
if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
nmp->nm_minorvers == 0))
error = 0;
return (error);
}
/*
* nfs remove directory call
*/
int
nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
{
struct nfsrv_descript nfsd, *nd = &nfsd;
int error = 0;
*dattrflagp = 0;
if (namelen > NFS_MAXNAMLEN)
return (ENAMETOOLONG);
NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
(void) nfsm_strtom(nd, name, namelen);
error = nfscl_request(nd, dvp, p, cred, dstuff);
if (error)
return (error);
if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
if (nd->nd_repstat && !error)
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
/*
* Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
*/
if (error == ENOENT)
error = 0;
return (error);
}
/*
* Readdir rpc.
* Always returns with either uio_resid unchanged, if you are at the
* end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
* filled in.
* I felt this would allow caching of directory blocks more easily
* than returning a pertially filled block.
* Directory offset cookies:
* Oh my, what to do with them...
* I can think of three ways to deal with them:
* 1 - have the layer above these RPCs maintain a map between logical
* directory byte offsets and the NFS directory offset cookies
* 2 - pass the opaque directory offset cookies up into userland
* and let the libc functions deal with them, via the system call
* 3 - return them to userland in the "struct dirent", so future versions
* of libc can use them and do whatever is necessary to make things work
* above these rpc calls, in the meantime
* For now, I do #3 by "hiding" the directory offset cookies after the
* d_name field in struct dirent. This is space inside d_reclen that
* will be ignored by anything that doesn't know about them.
* The directory offset cookies are filled in as the last 8 bytes of
* each directory entry, after d_name. Someday, the userland libc
* functions may be able to use these. In the meantime, it satisfies
* OpenBSD's requirements for cookies being returned.
* If expects the directory offset cookie for the read to be in uio_offset
* and returns the one for the next entry after this directory block in
* there, as well.
*/
int
nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
int *eofp, void *stuff)
{
int len, left;
struct dirent *dp = NULL;
u_int32_t *tl;
nfsquad_t cookie, ncookie;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
struct nfsnode *dnp = VTONFS(vp);
struct nfsvattr nfsva;
struct nfsrv_descript nfsd, *nd = &nfsd;
int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
char *cp;
nfsattrbit_t attrbits, dattrbits;
u_int32_t rderr, *tl2 = NULL;
size_t tresid;
KASSERT(uiop->uio_iovcnt == 1 &&
(uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
("nfs readdirrpc bad uio"));
ncookie.lval[0] = ncookie.lval[1] = 0;
/*
* There is no point in reading a lot more than uio_resid, however
* adding one additional DIRBLKSIZ makes sense. Since uio_resid
* and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
* will never make readsize > nm_readdirsize.
*/
readsize = nmp->nm_readdirsize;
if (readsize > uiop->uio_resid)
readsize = uiop->uio_resid + DIRBLKSIZ;
*attrflagp = 0;
if (eofp)
*eofp = 0;
tresid = uiop->uio_resid;
cookie.lval[0] = cookiep->nfsuquad[0];
cookie.lval[1] = cookiep->nfsuquad[1];
nd->nd_mrep = NULL;
/*
* For NFSv4, first create the "." and ".." entries.
*/
if (NFSHASNFSV4(nmp)) {
reqsize = 6 * NFSX_UNSIGNED;
NFSGETATTR_ATTRBIT(&dattrbits);
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
NFSATTRBIT_MOUNTEDONFILEID)) {
NFSSETBIT_ATTRBIT(&attrbits,
NFSATTRBIT_MOUNTEDONFILEID);
gotmnton = 1;
} else {
/*
* Must fake it. Use the fileno, except when the
* fsid is != to that of the directory. For that
* case, generate a fake fileno that is not the same.
*/
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
gotmnton = 0;
}
/*
* Joy, oh joy. For V4 we get to hand craft '.' and '..'.
*/
if (uiop->uio_offset == 0) {
NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV4OP_GETFH);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &attrbits);
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
dotfileid = 0; /* Fake out the compiler. */
if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
error = nfsm_loadattr(nd, &nfsva);
if (error != 0)
goto nfsmout;
dotfileid = nfsva.na_fileid;
}
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
len = fxdr_unsigned(int, *(tl + 4));
if (len > 0 && len <= NFSX_V4FHMAX)
error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
else
error = EPERM;
if (!error) {
NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
nfsva.na_mntonfileno = UINT64_MAX;
error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
NULL, NULL, NULL, p, cred);
if (error) {
dotdotfileid = dotfileid;
} else if (gotmnton) {
if (nfsva.na_mntonfileno != UINT64_MAX)
dotdotfileid = nfsva.na_mntonfileno;
else
dotdotfileid = nfsva.na_fileid;
} else if (nfsva.na_filesid[0] ==
dnp->n_vattr.na_filesid[0] &&
nfsva.na_filesid[1] ==
dnp->n_vattr.na_filesid[1]) {
dotdotfileid = nfsva.na_fileid;
} else {
do {
fakefileno--;
} while (fakefileno ==
nfsva.na_fileid);
dotdotfileid = fakefileno;
}
}
} else if (nd->nd_repstat == NFSERR_NOENT) {
/*
* Lookupp returns NFSERR_NOENT when we are
* at the root, so just use the current dir.
*/
nd->nd_repstat = 0;
dotdotfileid = dotfileid;
} else {
error = nd->nd_repstat;
}
m_freem(nd->nd_mrep);
if (error)
return (error);
nd->nd_mrep = NULL;
dp = (struct dirent *)uiop->uio_iov->iov_base;
dp->d_pad0 = dp->d_pad1 = 0;
dp->d_off = 0;
dp->d_type = DT_DIR;
dp->d_fileno = dotfileid;
dp->d_namlen = 1;
*((uint64_t *)dp->d_name) = 0; /* Zero pad it. */
dp->d_name[0] = '.';
dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
/*
* Just make these offset cookie 0.
*/
tl = (u_int32_t *)&dp->d_name[8];
*tl++ = 0;
*tl = 0;
blksiz += dp->d_reclen;
uiop->uio_resid -= dp->d_reclen;
uiop->uio_offset += dp->d_reclen;
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base + dp->d_reclen;
uiop->uio_iov->iov_len -= dp->d_reclen;
dp = (struct dirent *)uiop->uio_iov->iov_base;
dp->d_pad0 = dp->d_pad1 = 0;
dp->d_off = 0;
dp->d_type = DT_DIR;
dp->d_fileno = dotdotfileid;
dp->d_namlen = 2;
*((uint64_t *)dp->d_name) = 0;
dp->d_name[0] = '.';
dp->d_name[1] = '.';
dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
/*
* Just make these offset cookie 0.
*/
tl = (u_int32_t *)&dp->d_name[8];
*tl++ = 0;
*tl = 0;
blksiz += dp->d_reclen;
uiop->uio_resid -= dp->d_reclen;
uiop->uio_offset += dp->d_reclen;
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base + dp->d_reclen;
uiop->uio_iov->iov_len -= dp->d_reclen;
}
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
} else {
reqsize = 5 * NFSX_UNSIGNED;
}
/*
* Loop around doing readdir rpc's of size readsize.
* The stopping criteria is EOF or buffer full.
*/
while (more_dirs && bigenough) {
*attrflagp = 0;
NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
if (nd->nd_flag & ND_NFSV2) {
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = cookie.lval[1];
*tl = txdr_unsigned(readsize);
} else {
NFSM_BUILD(tl, u_int32_t *, reqsize);
*tl++ = cookie.lval[0];
*tl++ = cookie.lval[1];
if (cookie.qval == 0) {
*tl++ = 0;
*tl++ = 0;
} else {
NFSLOCKNODE(dnp);
*tl++ = dnp->n_cookieverf.nfsuquad[0];
*tl++ = dnp->n_cookieverf.nfsuquad[1];
NFSUNLOCKNODE(dnp);
}
if (nd->nd_flag & ND_NFSV4) {
*tl++ = txdr_unsigned(readsize);
*tl = txdr_unsigned(readsize);
(void) nfsrv_putattrbit(nd, &attrbits);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &dattrbits);
} else {
*tl = txdr_unsigned(readsize);
}
}
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
if (!(nd->nd_flag & ND_NFSV2)) {
if (nd->nd_flag & ND_NFSV3)
error = nfscl_postop_attr(nd, nap, attrflagp,
stuff);
if (!nd->nd_repstat && !error) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
NFSLOCKNODE(dnp);
dnp->n_cookieverf.nfsuquad[0] = *tl++;
dnp->n_cookieverf.nfsuquad[1] = *tl;
NFSUNLOCKNODE(dnp);
}
}
if (nd->nd_repstat || error) {
if (!error)
error = nd->nd_repstat;
goto nfsmout;
}
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
more_dirs = fxdr_unsigned(int, *tl);
if (!more_dirs)
tryformoredirs = 0;
/* loop through the dir entries, doctoring them to 4bsd form */
while (more_dirs && bigenough) {
if (nd->nd_flag & ND_NFSV4) {
NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
ncookie.lval[0] = *tl++;
ncookie.lval[1] = *tl++;
len = fxdr_unsigned(int, *tl);
} else if (nd->nd_flag & ND_NFSV3) {
NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
nfsva.na_fileid = fxdr_hyper(tl);
tl += 2;
len = fxdr_unsigned(int, *tl);
} else {
NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
nfsva.na_fileid = fxdr_unsigned(uint64_t,
*tl++);
len = fxdr_unsigned(int, *tl);
}
if (len <= 0 || len > NFS_MAXNAMLEN) {
error = EBADRPC;
goto nfsmout;
}
tlen = roundup2(len, 8);
if (tlen == len)
tlen += 8; /* To ensure null termination. */
left = DIRBLKSIZ - blksiz;
if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
NFSBZERO(uiop->uio_iov->iov_base, left);
dp->d_reclen += left;
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base + left;
uiop->uio_iov->iov_len -= left;
uiop->uio_resid -= left;
uiop->uio_offset += left;
blksiz = 0;
}
if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
uiop->uio_resid)
bigenough = 0;
if (bigenough) {
dp = (struct dirent *)uiop->uio_iov->iov_base;
dp->d_pad0 = dp->d_pad1 = 0;
dp->d_off = 0;
dp->d_namlen = len;
dp->d_reclen = _GENERIC_DIRLEN(len) +
NFSX_HYPER;
dp->d_type = DT_UNKNOWN;
blksiz += dp->d_reclen;
if (blksiz == DIRBLKSIZ)
blksiz = 0;
uiop->uio_resid -= DIRHDSIZ;
uiop->uio_offset += DIRHDSIZ;
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base + DIRHDSIZ;
uiop->uio_iov->iov_len -= DIRHDSIZ;
error = nfsm_mbufuio(nd, uiop, len);
if (error)
goto nfsmout;
cp = uiop->uio_iov->iov_base;
tlen -= len;
NFSBZERO(cp, tlen);
cp += tlen; /* points to cookie storage */
tl2 = (u_int32_t *)cp;
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base + tlen +
NFSX_HYPER;
uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
uiop->uio_resid -= tlen + NFSX_HYPER;
uiop->uio_offset += (tlen + NFSX_HYPER);
} else {
error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
if (error)
goto nfsmout;
}
if (nd->nd_flag & ND_NFSV4) {
rderr = 0;
nfsva.na_mntonfileno = UINT64_MAX;
error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
NULL, NULL, &rderr, p, cred);
if (error)
goto nfsmout;
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
} else if (nd->nd_flag & ND_NFSV3) {
NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
ncookie.lval[0] = *tl++;
ncookie.lval[1] = *tl++;
} else {
NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
ncookie.lval[0] = 0;
ncookie.lval[1] = *tl++;
}
if (bigenough) {
if (nd->nd_flag & ND_NFSV4) {
if (rderr) {
dp->d_fileno = 0;
} else {
if (gotmnton) {
if (nfsva.na_mntonfileno != UINT64_MAX)
dp->d_fileno = nfsva.na_mntonfileno;
else
dp->d_fileno = nfsva.na_fileid;
} else if (nfsva.na_filesid[0] ==
dnp->n_vattr.na_filesid[0] &&
nfsva.na_filesid[1] ==
dnp->n_vattr.na_filesid[1]) {
dp->d_fileno = nfsva.na_fileid;
} else {
do {
fakefileno--;
} while (fakefileno ==
nfsva.na_fileid);
dp->d_fileno = fakefileno;
}
dp->d_type = vtonfs_dtype(nfsva.na_type);
}
} else {
dp->d_fileno = nfsva.na_fileid;
}
*tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
ncookie.lval[0];
*tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
ncookie.lval[1];
}
more_dirs = fxdr_unsigned(int, *tl);
}
/*
* If at end of rpc data, get the eof boolean
*/
if (!more_dirs) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
eof = fxdr_unsigned(int, *tl);
if (tryformoredirs)
more_dirs = !eof;
if (nd->nd_flag & ND_NFSV4) {
error = nfscl_postop_attr(nd, nap, attrflagp,
stuff);
if (error)
goto nfsmout;
}
}
m_freem(nd->nd_mrep);
nd->nd_mrep = NULL;
}
/*
* Fill last record, iff any, out to a multiple of DIRBLKSIZ
* by increasing d_reclen for the last record.
*/
if (blksiz > 0) {
left = DIRBLKSIZ - blksiz;
NFSBZERO(uiop->uio_iov->iov_base, left);
dp->d_reclen += left;
uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
left;
uiop->uio_iov->iov_len -= left;
uiop->uio_resid -= left;
uiop->uio_offset += left;
}
/*
* If returning no data, assume end of file.
* If not bigenough, return not end of file, since you aren't
* returning all the data
* Otherwise, return the eof flag from the server.
*/
if (eofp) {
if (tresid == ((size_t)(uiop->uio_resid)))
*eofp = 1;
else if (!bigenough)
*eofp = 0;
else
*eofp = eof;
}
/*
* Add extra empty records to any remaining DIRBLKSIZ chunks.
*/
while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
dp = (struct dirent *)uiop->uio_iov->iov_base;
NFSBZERO(dp, DIRBLKSIZ);
dp->d_type = DT_UNKNOWN;
tl = (u_int32_t *)&dp->d_name[4];
*tl++ = cookie.lval[0];
*tl = cookie.lval[1];
dp->d_reclen = DIRBLKSIZ;
uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
DIRBLKSIZ;
uiop->uio_iov->iov_len -= DIRBLKSIZ;
uiop->uio_resid -= DIRBLKSIZ;
uiop->uio_offset += DIRBLKSIZ;
}
nfsmout:
if (nd->nd_mrep != NULL)
m_freem(nd->nd_mrep);
return (error);
}
#ifndef APPLE
/*
* NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
* (Also used for NFS V4 when mount flag set.)
* (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
*/
int
nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
int *eofp, void *stuff)
{
int len, left;
struct dirent *dp = NULL;
u_int32_t *tl;
vnode_t newvp = NULLVP;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nameidata nami, *ndp = &nami;
struct componentname *cnp = &ndp->ni_cnd;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
struct nfsnode *dnp = VTONFS(vp), *np;
struct nfsvattr nfsva;
struct nfsfh *nfhp;
nfsquad_t cookie, ncookie;
int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
int isdotdot = 0, unlocknewvp = 0;
u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
u_int64_t fileno = 0;
char *cp;
nfsattrbit_t attrbits, dattrbits;
size_t tresid;
u_int32_t *tl2 = NULL, rderr;
struct timespec dctime, ts;
bool attr_ok;
KASSERT(uiop->uio_iovcnt == 1 &&
(uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
("nfs readdirplusrpc bad uio"));
ncookie.lval[0] = ncookie.lval[1] = 0;
timespecclear(&dctime);
*attrflagp = 0;
if (eofp != NULL)
*eofp = 0;
ndp->ni_dvp = vp;
nd->nd_mrep = NULL;
cookie.lval[0] = cookiep->nfsuquad[0];
cookie.lval[1] = cookiep->nfsuquad[1];
tresid = uiop->uio_resid;
/*
* For NFSv4, first create the "." and ".." entries.
*/
if (NFSHASNFSV4(nmp)) {
NFSGETATTR_ATTRBIT(&dattrbits);
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
NFSATTRBIT_MOUNTEDONFILEID)) {
NFSSETBIT_ATTRBIT(&attrbits,
NFSATTRBIT_MOUNTEDONFILEID);
gotmnton = 1;
} else {
/*
* Must fake it. Use the fileno, except when the
* fsid is != to that of the directory. For that
* case, generate a fake fileno that is not the same.
*/
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
gotmnton = 0;
}
/*
* Joy, oh joy. For V4 we get to hand craft '.' and '..'.
*/
if (uiop->uio_offset == 0) {
NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV4OP_GETFH);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &attrbits);
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
dotfileid = 0; /* Fake out the compiler. */
if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
error = nfsm_loadattr(nd, &nfsva);
if (error != 0)
goto nfsmout;
dctime = nfsva.na_ctime;
dotfileid = nfsva.na_fileid;
}
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
len = fxdr_unsigned(int, *(tl + 4));
if (len > 0 && len <= NFSX_V4FHMAX)
error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
else
error = EPERM;
if (!error) {
NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
nfsva.na_mntonfileno = UINT64_MAX;
error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
NULL, NULL, NULL, p, cred);
if (error) {
dotdotfileid = dotfileid;
} else if (gotmnton) {
if (nfsva.na_mntonfileno != UINT64_MAX)
dotdotfileid = nfsva.na_mntonfileno;
else
dotdotfileid = nfsva.na_fileid;
} else if (nfsva.na_filesid[0] ==
dnp->n_vattr.na_filesid[0] &&
nfsva.na_filesid[1] ==
dnp->n_vattr.na_filesid[1]) {
dotdotfileid = nfsva.na_fileid;
} else {
do {
fakefileno--;
} while (fakefileno ==
nfsva.na_fileid);
dotdotfileid = fakefileno;
}
}
} else if (nd->nd_repstat == NFSERR_NOENT) {
/*
* Lookupp returns NFSERR_NOENT when we are
* at the root, so just use the current dir.
*/
nd->nd_repstat = 0;
dotdotfileid = dotfileid;
} else {
error = nd->nd_repstat;
}
m_freem(nd->nd_mrep);
if (error)
return (error);
nd->nd_mrep = NULL;
dp = (struct dirent *)uiop->uio_iov->iov_base;
dp->d_pad0 = dp->d_pad1 = 0;
dp->d_off = 0;
dp->d_type = DT_DIR;
dp->d_fileno = dotfileid;
dp->d_namlen = 1;
*((uint64_t *)dp->d_name) = 0; /* Zero pad it. */
dp->d_name[0] = '.';
dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
/*
* Just make these offset cookie 0.
*/
tl = (u_int32_t *)&dp->d_name[8];
*tl++ = 0;
*tl = 0;
blksiz += dp->d_reclen;
uiop->uio_resid -= dp->d_reclen;
uiop->uio_offset += dp->d_reclen;
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base + dp->d_reclen;
uiop->uio_iov->iov_len -= dp->d_reclen;
dp = (struct dirent *)uiop->uio_iov->iov_base;
dp->d_pad0 = dp->d_pad1 = 0;
dp->d_off = 0;
dp->d_type = DT_DIR;
dp->d_fileno = dotdotfileid;
dp->d_namlen = 2;
*((uint64_t *)dp->d_name) = 0;
dp->d_name[0] = '.';
dp->d_name[1] = '.';
dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
/*
* Just make these offset cookie 0.
*/
tl = (u_int32_t *)&dp->d_name[8];
*tl++ = 0;
*tl = 0;
blksiz += dp->d_reclen;
uiop->uio_resid -= dp->d_reclen;
uiop->uio_offset += dp->d_reclen;
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base + dp->d_reclen;
uiop->uio_iov->iov_len -= dp->d_reclen;
}
NFSREADDIRPLUS_ATTRBIT(&attrbits);
if (gotmnton)
NFSSETBIT_ATTRBIT(&attrbits,
NFSATTRBIT_MOUNTEDONFILEID);
if (!NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
NFSATTRBIT_TIMECREATE))
NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMECREATE);
}
/*
* Loop around doing readdir rpc's of size nm_readdirsize.
* The stopping criteria is EOF or buffer full.
*/
while (more_dirs && bigenough) {
*attrflagp = 0;
NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
*tl++ = cookie.lval[0];
*tl++ = cookie.lval[1];
if (cookie.qval == 0) {
*tl++ = 0;
*tl++ = 0;
} else {
NFSLOCKNODE(dnp);
*tl++ = dnp->n_cookieverf.nfsuquad[0];
*tl++ = dnp->n_cookieverf.nfsuquad[1];
NFSUNLOCKNODE(dnp);
}
*tl++ = txdr_unsigned(nmp->nm_readdirsize);
*tl = txdr_unsigned(nmp->nm_readdirsize);
if (nd->nd_flag & ND_NFSV4) {
(void) nfsrv_putattrbit(nd, &attrbits);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &dattrbits);
}
nanouptime(&ts);
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
if (nd->nd_flag & ND_NFSV3)
error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
if (nd->nd_repstat || error) {
if (!error)
error = nd->nd_repstat;
goto nfsmout;
}
if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
dctime = nap->na_ctime;
NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
NFSLOCKNODE(dnp);
dnp->n_cookieverf.nfsuquad[0] = *tl++;
dnp->n_cookieverf.nfsuquad[1] = *tl++;
NFSUNLOCKNODE(dnp);
more_dirs = fxdr_unsigned(int, *tl);
if (!more_dirs)
tryformoredirs = 0;
/* loop through the dir entries, doctoring them to 4bsd form */
while (more_dirs && bigenough) {
NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
if (nd->nd_flag & ND_NFSV4) {
ncookie.lval[0] = *tl++;
ncookie.lval[1] = *tl++;
} else {
fileno = fxdr_hyper(tl);
tl += 2;
}
len = fxdr_unsigned(int, *tl);
if (len <= 0 || len > NFS_MAXNAMLEN) {
error = EBADRPC;
goto nfsmout;
}
tlen = roundup2(len, 8);
if (tlen == len)
tlen += 8; /* To ensure null termination. */
left = DIRBLKSIZ - blksiz;
if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
NFSBZERO(uiop->uio_iov->iov_base, left);
dp->d_reclen += left;
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base + left;
uiop->uio_iov->iov_len -= left;
uiop->uio_resid -= left;
uiop->uio_offset += left;
blksiz = 0;
}
if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
uiop->uio_resid)
bigenough = 0;
if (bigenough) {
dp = (struct dirent *)uiop->uio_iov->iov_base;
dp->d_pad0 = dp->d_pad1 = 0;
dp->d_off = 0;
dp->d_namlen = len;
dp->d_reclen = _GENERIC_DIRLEN(len) +
NFSX_HYPER;
dp->d_type = DT_UNKNOWN;
blksiz += dp->d_reclen;
if (blksiz == DIRBLKSIZ)
blksiz = 0;
uiop->uio_resid -= DIRHDSIZ;
uiop->uio_offset += DIRHDSIZ;
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base + DIRHDSIZ;
uiop->uio_iov->iov_len -= DIRHDSIZ;
cnp->cn_nameptr = uiop->uio_iov->iov_base;
cnp->cn_namelen = len;
NFSCNHASHZERO(cnp);
error = nfsm_mbufuio(nd, uiop, len);
if (error)
goto nfsmout;
cp = uiop->uio_iov->iov_base;
tlen -= len;
NFSBZERO(cp, tlen);
cp += tlen; /* points to cookie storage */
tl2 = (u_int32_t *)cp;
if (len == 2 && cnp->cn_nameptr[0] == '.' &&
cnp->cn_nameptr[1] == '.')
isdotdot = 1;
else
isdotdot = 0;
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base + tlen +
NFSX_HYPER;
uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
uiop->uio_resid -= tlen + NFSX_HYPER;
uiop->uio_offset += (tlen + NFSX_HYPER);
} else {
error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
if (error)
goto nfsmout;
}
nfhp = NULL;
if (nd->nd_flag & ND_NFSV3) {
NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
ncookie.lval[0] = *tl++;
ncookie.lval[1] = *tl++;
attrflag = fxdr_unsigned(int, *tl);
if (attrflag) {
error = nfsm_loadattr(nd, &nfsva);
if (error)
goto nfsmout;
}
NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
if (*tl) {
error = nfsm_getfh(nd, &nfhp);
if (error)
goto nfsmout;
}
if (!attrflag && nfhp != NULL) {
free(nfhp, M_NFSFH);
nfhp = NULL;
}
} else {
rderr = 0;
nfsva.na_mntonfileno = 0xffffffff;
error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
NULL, NULL, &rderr, p, cred);
if (error)
goto nfsmout;
}
if (bigenough) {
if (nd->nd_flag & ND_NFSV4) {
if (rderr) {
dp->d_fileno = 0;
} else if (gotmnton) {
if (nfsva.na_mntonfileno != 0xffffffff)
dp->d_fileno = nfsva.na_mntonfileno;
else
dp->d_fileno = nfsva.na_fileid;
} else if (nfsva.na_filesid[0] ==
dnp->n_vattr.na_filesid[0] &&
nfsva.na_filesid[1] ==
dnp->n_vattr.na_filesid[1]) {
dp->d_fileno = nfsva.na_fileid;
} else {
do {
fakefileno--;
} while (fakefileno ==
nfsva.na_fileid);
dp->d_fileno = fakefileno;
}
} else {
dp->d_fileno = fileno;
}
*tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
ncookie.lval[0];
*tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
ncookie.lval[1];
if (nfhp != NULL) {
attr_ok = true;
if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
VREF(vp);
newvp = vp;
unlocknewvp = 0;
free(nfhp, M_NFSFH);
np = dnp;
} else if (isdotdot != 0) {
/*
* Skip doing a nfscl_nget() call for "..".
* There's a race between acquiring the nfs
* node here and lookups that look for the
* directory being read (in the parent).
* It would try to get a lock on ".." here,
* owning the lock on the directory being
* read. Lookup will hold the lock on ".."
* and try to acquire the lock on the
* directory being read.
* If the directory is unlocked/relocked,
* then there is a LOR with the buflock
* vp is relocked.
*/
free(nfhp, M_NFSFH);
} else {
error = nfscl_nget(vp->v_mount, vp,
nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
if (!error) {
newvp = NFSTOV(np);
unlocknewvp = 1;
/*
* If n_localmodtime >= time before RPC,
* then a file modification operation,
* such as VOP_SETATTR() of size, has
* occurred while the Lookup RPC and
* acquisition of the vnode happened. As
* such, the attributes might be stale,
* with possibly an incorrect size.
*/
NFSLOCKNODE(np);
if (timespecisset(
&np->n_localmodtime) &&
timespeccmp(&np->n_localmodtime,
&ts, >=)) {
NFSCL_DEBUG(4, "nfsrpc_readdirplus:"
" localmod stale attributes\n");
attr_ok = false;
}
NFSUNLOCKNODE(np);
}
}
nfhp = NULL;
if (newvp != NULLVP) {
if (attr_ok)
error = nfscl_loadattrcache(&newvp,
&nfsva, NULL, NULL, 0, 0);
if (error) {
if (unlocknewvp)
vput(newvp);
else
vrele(newvp);
goto nfsmout;
}
dp->d_type =
vtonfs_dtype(np->n_vattr.na_type);
ndp->ni_vp = newvp;
NFSCNHASH(cnp, HASHINIT);
if (cnp->cn_namelen <= NCHNAMLEN &&
ndp->ni_dvp != ndp->ni_vp &&
(newvp->v_type != VDIR ||
dctime.tv_sec != 0)) {
cache_enter_time_flags(ndp->ni_dvp,
ndp->ni_vp, cnp,
&nfsva.na_ctime,
newvp->v_type != VDIR ? NULL :
&dctime, VFS_CACHE_DROPOLD);
}
if (unlocknewvp)
vput(newvp);
else
vrele(newvp);
newvp = NULLVP;
}
}
} else if (nfhp != NULL) {
free(nfhp, M_NFSFH);
}
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
more_dirs = fxdr_unsigned(int, *tl);
}
/*
* If at end of rpc data, get the eof boolean
*/
if (!more_dirs) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
eof = fxdr_unsigned(int, *tl);
if (tryformoredirs)
more_dirs = !eof;
if (nd->nd_flag & ND_NFSV4) {
error = nfscl_postop_attr(nd, nap, attrflagp,
stuff);
if (error)
goto nfsmout;
}
}
m_freem(nd->nd_mrep);
nd->nd_mrep = NULL;
}
/*
* Fill last record, iff any, out to a multiple of DIRBLKSIZ
* by increasing d_reclen for the last record.
*/
if (blksiz > 0) {
left = DIRBLKSIZ - blksiz;
NFSBZERO(uiop->uio_iov->iov_base, left);
dp->d_reclen += left;
uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
left;
uiop->uio_iov->iov_len -= left;
uiop->uio_resid -= left;
uiop->uio_offset += left;
}
/*
* If returning no data, assume end of file.
* If not bigenough, return not end of file, since you aren't
* returning all the data
* Otherwise, return the eof flag from the server.
*/
if (eofp != NULL) {
if (tresid == uiop->uio_resid)
*eofp = 1;
else if (!bigenough)
*eofp = 0;
else
*eofp = eof;
}
/*
* Add extra empty records to any remaining DIRBLKSIZ chunks.
*/
while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
dp = (struct dirent *)uiop->uio_iov->iov_base;
NFSBZERO(dp, DIRBLKSIZ);
dp->d_type = DT_UNKNOWN;
tl = (u_int32_t *)&dp->d_name[4];
*tl++ = cookie.lval[0];
*tl = cookie.lval[1];
dp->d_reclen = DIRBLKSIZ;
uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
DIRBLKSIZ;
uiop->uio_iov->iov_len -= DIRBLKSIZ;
uiop->uio_resid -= DIRBLKSIZ;
uiop->uio_offset += DIRBLKSIZ;
}
nfsmout:
if (nd->nd_mrep != NULL)
m_freem(nd->nd_mrep);
return (error);
}
#endif /* !APPLE */
/*
* Nfs commit rpc
*/
int
nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
{
u_int32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
nfsattrbit_t attrbits;
int error;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
*attrflagp = 0;
NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
txdr_hyper(offset, tl);
tl += 2;
*tl = txdr_unsigned(cnt);
if (nd->nd_flag & ND_NFSV4) {
/*
* And do a Getattr op.
*/
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
(void) nfsrv_putattrbit(nd, &attrbits);
}
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, stuff);
if (!error && !nd->nd_repstat) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
NFSLOCKMNT(nmp);
if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
nd->nd_repstat = NFSERR_STALEWRITEVERF;
}
NFSUNLOCKMNT(nmp);
if (nd->nd_flag & ND_NFSV4)
error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
}
nfsmout:
if (!error && nd->nd_repstat)
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* NFS byte range lock rpc.
* (Mostly just calls one of the three lower level RPC routines.)
*/
int
nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
{
struct nfscllockowner *lp;
struct nfsclclient *clp;
struct nfsfh *nfhp;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
u_int64_t off, len;
off_t start, end;
u_int32_t clidrev = 0;
int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
int callcnt, dorpc;
/*
* Convert the flock structure into a start and end and do POSIX
* bounds checking.
*/
switch (fl->l_whence) {
case SEEK_SET:
case SEEK_CUR:
/*
* Caller is responsible for adding any necessary offset
* when SEEK_CUR is used.
*/
start = fl->l_start;
off = fl->l_start;
break;
case SEEK_END:
start = size + fl->l_start;
off = size + fl->l_start;
break;
default:
return (EINVAL);
}
if (start < 0)
return (EINVAL);
if (fl->l_len != 0) {
end = start + fl->l_len - 1;
if (end < start)
return (EINVAL);
}
len = fl->l_len;
if (len == 0)
len = NFS64BITSSET;
retrycnt = 0;
do {
nd->nd_repstat = 0;
if (op == F_GETLK) {
error = nfscl_getcl(vp->v_mount, cred, p, false, true, &clp);
if (error)
return (error);
error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
if (!error) {
clidrev = clp->nfsc_clientidrev;
error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
p, id, flags);
} else if (error == -1) {
error = 0;
}
nfscl_clientrelease(clp);
} else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
/*
* We must loop around for all lockowner cases.
*/
callcnt = 0;
error = nfscl_getcl(vp->v_mount, cred, p, false, true, &clp);
if (error)
return (error);
do {
error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
clp, id, flags, &lp, &dorpc);
/*
* If it returns a NULL lp, we're done.
*/
if (lp == NULL) {
if (callcnt == 0)
nfscl_clientrelease(clp);
else
nfscl_releasealllocks(clp, vp, p, id, flags);
return (error);
}
if (nmp->nm_clp != NULL)
clidrev = nmp->nm_clp->nfsc_clientidrev;
else
clidrev = 0;
/*
* If the server doesn't support Posix lock semantics,
* only allow locks on the entire file, since it won't
* handle overlapping byte ranges.
* There might still be a problem when a lock
* upgrade/downgrade (read<->write) occurs, since the
* server "might" expect an unlock first?
*/
if (dorpc && (lp->nfsl_open->nfso_posixlock ||
(off == 0 && len == NFS64BITSSET))) {
/*
* Since the lock records will go away, we must
* wait for grace and delay here.
*/
do {
error = nfsrpc_locku(nd, nmp, lp, off, len,
NFSV4LOCKT_READ, cred, p, 0);
if ((nd->nd_repstat == NFSERR_GRACE ||
nd->nd_repstat == NFSERR_DELAY) &&
error == 0)
(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
"nfs_advlock");
} while ((nd->nd_repstat == NFSERR_GRACE ||
nd->nd_repstat == NFSERR_DELAY) && error == 0);
}
callcnt++;
} while (error == 0 && nd->nd_repstat == 0);
nfscl_releasealllocks(clp, vp, p, id, flags);
} else if (op == F_SETLK) {
error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
if (error || donelocally) {
return (error);
}
if (nmp->nm_clp != NULL)
clidrev = nmp->nm_clp->nfsc_clientidrev;
else
clidrev = 0;
nfhp = VTONFS(vp)->n_fhp;
if (!lp->nfsl_open->nfso_posixlock &&
(off != 0 || len != NFS64BITSSET)) {
error = EINVAL;
} else {
error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
nfhp->nfh_len, lp, newone, reclaim, off,
len, fl->l_type, cred, p, 0);
}
if (!error)
error = nd->nd_repstat;
nfscl_lockrelease(lp, error, newone);
} else {
error = EINVAL;
}
if (!error)
error = nd->nd_repstat;
if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
error == NFSERR_STALEDONTRECOVER ||
error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
error == NFSERR_BADSESSION) {
(void) nfs_catnap(PZERO, error, "nfs_advlock");
} else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
&& clidrev != 0) {
expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
retrycnt++;
}
} while (error == NFSERR_GRACE ||
error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
error == NFSERR_BADSESSION ||
((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
expireret == 0 && clidrev != 0 && retrycnt < 4));
if (error && retrycnt >= 4)
error = EIO;
return (error);
}
/*
* The lower level routine for the LockT case.
*/
int
nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
struct ucred *cred, NFSPROC_T *p, void *id, int flags)
{
u_int32_t *tl;
int error, type, size;
uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
struct nfsnode *np;
struct nfsmount *nmp;
struct nfsclsession *tsep;
nmp = VFSTONFS(vp->v_mount);
NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
if (fl->l_type == F_RDLCK)
*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
else
*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
txdr_hyper(off, tl);
tl += 2;
txdr_hyper(len, tl);
tl += 2;
tsep = nfsmnt_mdssession(nmp);
*tl++ = tsep->nfsess_clientid.lval[0];
*tl = tsep->nfsess_clientid.lval[1];
nfscl_filllockowner(id, own, flags);
np = VTONFS(vp);
NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
np->n_fhp->nfh_len);
(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
error = nfscl_request(nd, vp, p, cred, NULL);
if (error)
return (error);
if (nd->nd_repstat == 0) {
fl->l_type = F_UNLCK;
} else if (nd->nd_repstat == NFSERR_DENIED) {
nd->nd_repstat = 0;
fl->l_whence = SEEK_SET;
NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
fl->l_start = fxdr_hyper(tl);
tl += 2;
len = fxdr_hyper(tl);
tl += 2;
if (len == NFS64BITSSET)
fl->l_len = 0;
else
fl->l_len = len;
type = fxdr_unsigned(int, *tl++);
if (type == NFSV4LOCKT_WRITE)
fl->l_type = F_WRLCK;
else
fl->l_type = F_RDLCK;
/*
* XXX For now, I have no idea what to do with the
* conflicting lock_owner, so I'll just set the pid == 0
* and skip over the lock_owner.
*/
fl->l_pid = (pid_t)0;
tl += 2;
size = fxdr_unsigned(int, *tl);
if (size < 0 || size > NFSV4_OPAQUELIMIT)
error = EBADRPC;
if (!error)
error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
} else if (nd->nd_repstat == NFSERR_STALECLIENTID)
nfscl_initiate_recovery(clp);
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* Lower level function that performs the LockU RPC.
*/
static int
nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
{
u_int32_t *tl;
int error;
nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0);
NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(type);
*tl = txdr_unsigned(lp->nfsl_seqid);
if (nfstest_outofseq &&
(arc4random() % nfstest_outofseq) == 0)
*tl = txdr_unsigned(lp->nfsl_seqid + 1);
tl++;
if (NFSHASNFSV4N(nmp))
*tl++ = 0;
else
*tl++ = lp->nfsl_stateid.seqid;
*tl++ = lp->nfsl_stateid.other[0];
*tl++ = lp->nfsl_stateid.other[1];
*tl++ = lp->nfsl_stateid.other[2];
txdr_hyper(off, tl);
tl += 2;
txdr_hyper(len, tl);
if (syscred)
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
if (error)
return (error);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
lp->nfsl_stateid.seqid = *tl++;
lp->nfsl_stateid.other[0] = *tl++;
lp->nfsl_stateid.other[1] = *tl++;
lp->nfsl_stateid.other[2] = *tl;
} else if (nd->nd_repstat == NFSERR_STALESTATEID)
nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* The actual Lock RPC.
*/
int
nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
NFSPROC_T *p, int syscred)
{
u_int32_t *tl;
int error, size;
uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
struct nfsclsession *tsep;
nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
if (type == F_RDLCK)
*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
else
*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
*tl++ = txdr_unsigned(reclaim);
txdr_hyper(off, tl);
tl += 2;
txdr_hyper(len, tl);
tl += 2;
if (newone) {
*tl = newnfs_true;
NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2 * NFSX_UNSIGNED + NFSX_HYPER);
*tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
if (NFSHASNFSV4N(nmp))
*tl++ = 0;
else
*tl++ = lp->nfsl_open->nfso_stateid.seqid;
*tl++ = lp->nfsl_open->nfso_stateid.other[0];
*tl++ = lp->nfsl_open->nfso_stateid.other[1];
*tl++ = lp->nfsl_open->nfso_stateid.other[2];
*tl++ = txdr_unsigned(lp->nfsl_seqid);
tsep = nfsmnt_mdssession(nmp);
*tl++ = tsep->nfsess_clientid.lval[0];
*tl = tsep->nfsess_clientid.lval[1];
NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
} else {
*tl = newnfs_false;
NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
if (NFSHASNFSV4N(nmp))
*tl++ = 0;
else
*tl++ = lp->nfsl_stateid.seqid;
*tl++ = lp->nfsl_stateid.other[0];
*tl++ = lp->nfsl_stateid.other[1];
*tl++ = lp->nfsl_stateid.other[2];
*tl = txdr_unsigned(lp->nfsl_seqid);
if (nfstest_outofseq &&
(arc4random() % nfstest_outofseq) == 0)
*tl = txdr_unsigned(lp->nfsl_seqid + 1);
}
if (syscred)
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error)
return (error);
if (newone)
NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
lp->nfsl_stateid.seqid = *tl++;
lp->nfsl_stateid.other[0] = *tl++;
lp->nfsl_stateid.other[1] = *tl++;
lp->nfsl_stateid.other[2] = *tl;
} else if (nd->nd_repstat == NFSERR_DENIED) {
NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
size = fxdr_unsigned(int, *(tl + 7));
if (size < 0 || size > NFSV4_OPAQUELIMIT)
error = EBADRPC;
if (!error)
error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
} else if (nd->nd_repstat == NFSERR_STALESTATEID)
nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* nfs statfs rpc
* (always called with the vp for the mount point)
*/
int
nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
void *stuff)
{
u_int32_t *tl = NULL;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfsmount *nmp;
nfsattrbit_t attrbits;
int error;
*attrflagp = 0;
nmp = VFSTONFS(vp->v_mount);
if (NFSHASNFSV4(nmp)) {
/*
* For V4, you actually do a getattr.
*/
NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
NFSSTATFS_GETATTRBIT(&attrbits);
(void) nfsrv_putattrbit(nd, &attrbits);
nd->nd_flag |= ND_USEGSSNAME;
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
if (nd->nd_repstat == 0) {
error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
cred);
if (!error) {
nmp->nm_fsid[0] = nap->na_filesid[0];
nmp->nm_fsid[1] = nap->na_filesid[1];
NFSSETHASSETFSID(nmp);
*attrflagp = 1;
}
} else {
error = nd->nd_repstat;
}
if (error)
goto nfsmout;
} else {
NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
if (nd->nd_flag & ND_NFSV3) {
error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
if (error)
goto nfsmout;
}
if (nd->nd_repstat) {
error = nd->nd_repstat;
goto nfsmout;
}
NFSM_DISSECT(tl, u_int32_t *,
NFSX_STATFS(nd->nd_flag & ND_NFSV3));
}
if (NFSHASNFSV3(nmp)) {
sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
} else if (NFSHASNFSV4(nmp) == 0) {
sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
}
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* nfs pathconf rpc
*/
int
nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
void *stuff)
{
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfsmount *nmp;
u_int32_t *tl;
nfsattrbit_t attrbits;
int error;
*attrflagp = 0;
nmp = VFSTONFS(vp->v_mount);
if (NFSHASNFSV4(nmp)) {
/*
* For V4, you actually do a getattr.
*/
NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
NFSPATHCONF_GETATTRBIT(&attrbits);
(void) nfsrv_putattrbit(nd, &attrbits);
nd->nd_flag |= ND_USEGSSNAME;
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
if (nd->nd_repstat == 0) {
error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
cred);
if (!error)
*attrflagp = 1;
} else {
error = nd->nd_repstat;
}
} else {
NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
if (nd->nd_repstat && !error)
error = nd->nd_repstat;
if (!error) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
pc->pc_chownrestricted =
fxdr_unsigned(u_int32_t, *tl++);
pc->pc_caseinsensitive =
fxdr_unsigned(u_int32_t, *tl++);
pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
}
}
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* nfs version 3 fsinfo rpc call
*/
int
nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
{
u_int32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
int error;
*attrflagp = 0;
NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
if (nd->nd_repstat && !error)
error = nd->nd_repstat;
if (!error) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
fsp->fs_maxfilesize = fxdr_hyper(tl);
tl += 2;
fxdr_nfsv3time(tl, &fsp->fs_timedelta);
tl += 2;
fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
}
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* This function performs the Renew RPC.
*/
int
nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
NFSPROC_T *p)
{
u_int32_t *tl;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
struct nfsmount *nmp;
int error;
struct nfssockreq *nrp;
struct nfsclsession *tsep;
nmp = clp->nfsc_nmp;
if (nmp == NULL)
return (0);
if (dsp == NULL)
nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
0);
else
nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
&dsp->nfsclds_sess, 0, 0);
if (!NFSHASNFSV4N(nmp)) {
/* NFSv4.1 just uses a Sequence Op and not a Renew. */
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
tsep = nfsmnt_mdssession(nmp);
*tl++ = tsep->nfsess_clientid.lval[0];
*tl = tsep->nfsess_clientid.lval[1];
}
nrp = NULL;
if (dsp != NULL)
nrp = dsp->nfsclds_sockp;
if (nrp == NULL)
/* If NULL, use the MDS socket. */
nrp = &nmp->nm_sockreq;
nd->nd_flag |= ND_USEGSSNAME;
if (dsp == NULL)
error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
else {
error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
if (error == ENXIO)
nfscl_cancelreqs(dsp);
}
if (error)
return (error);
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* This function performs the Releaselockowner RPC.
*/
int
nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
{
struct nfsrv_descript nfsd, *nd = &nfsd;
u_int32_t *tl;
int error;
uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
struct nfsclsession *tsep;
if (NFSHASNFSV4N(nmp)) {
/* For NFSv4.1, do a FreeStateID. */
nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
NULL, 0, 0);
nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
} else {
nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
NULL, 0, 0);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
tsep = nfsmnt_mdssession(nmp);
*tl++ = tsep->nfsess_clientid.lval[0];
*tl = tsep->nfsess_clientid.lval[1];
NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
}
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error)
return (error);
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* This function performs the Compound to get the mount pt FH.
*/
int
nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
NFSPROC_T *p)
{
u_int32_t *tl;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
u_char *cp, *cp2;
int error, cnt, len, setnil;
u_int32_t *opcntp;
nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
0);
cp = dirpath;
cnt = 0;
do {
setnil = 0;
while (*cp == '/')
cp++;
cp2 = cp;
while (*cp2 != '\0' && *cp2 != '/')
cp2++;
if (*cp2 == '/') {
setnil = 1;
*cp2 = '\0';
}
if (cp2 != cp) {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_LOOKUP);
nfsm_strtom(nd, cp, strlen(cp));
cnt++;
}
if (setnil)
*cp2++ = '/';
cp = cp2;
} while (*cp != '\0');
if (NFSHASNFSV4N(nmp))
/* Has a Sequence Op done by nfscl_reqstart(). */
*opcntp = txdr_unsigned(3 + cnt);
else
*opcntp = txdr_unsigned(2 + cnt);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETFH);
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error)
return (error);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
tl += (2 + 2 * cnt);
if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
len > NFSX_FHMAX) {
nd->nd_repstat = NFSERR_BADXDR;
} else {
nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
if (nd->nd_repstat == 0)
nmp->nm_fhsize = len;
}
}
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* This function performs the Delegreturn RPC.
*/
int
nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
struct nfsmount *nmp, NFSPROC_T *p, int syscred)
{
u_int32_t *tl;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
int error;
nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
dp->nfsdl_fhlen, NULL, NULL, 0, 0);
NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
if (NFSHASNFSV4N(nmp))
*tl++ = 0;
else
*tl++ = dp->nfsdl_stateid.seqid;
*tl++ = dp->nfsdl_stateid.other[0];
*tl++ = dp->nfsdl_stateid.other[1];
*tl = dp->nfsdl_stateid.other[2];
if (syscred)
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error)
return (error);
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* nfs getacl call.
*/
int
nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
struct acl *aclp, void *stuff)
{
struct nfsrv_descript nfsd, *nd = &nfsd;
int error;
nfsattrbit_t attrbits;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
return (EOPNOTSUPP);
NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
(void) nfsrv_putattrbit(nd, &attrbits);
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
if (!nd->nd_repstat)
error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
else
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* nfs setacl call.
*/
int
nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
struct acl *aclp, void *stuff)
{
int error;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
return (EOPNOTSUPP);
error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
return (error);
}
/*
* nfs setacl call.
*/
static int
nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
struct acl *aclp, nfsv4stateid_t *stateidp, void *stuff)
{
struct nfsrv_descript nfsd, *nd = &nfsd;
int error;
nfsattrbit_t attrbits;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
if (!NFSHASNFSV4(nmp))
return (EOPNOTSUPP);
NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
(void) nfsv4_fillattr(nd, vp->v_mount, vp, aclp, NULL, NULL, 0,
&attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
error = nfscl_request(nd, vp, p, cred, stuff);
if (error)
return (error);
/* Don't care about the pre/postop attributes */
m_freem(nd->nd_mrep);
return (nd->nd_repstat);
}
/*
* Do the NFSv4.1 Exchange ID.
*/
int
nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
struct nfssockreq *nrp, int minorvers, uint32_t exchflags,
struct nfsclds **dspp, struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl, v41flags;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
struct nfsclds *dsp;
struct timespec verstime;
int error, len;
*dspp = NULL;
if (minorvers == 0)
minorvers = nmp->nm_minorvers;
nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL,
NFS_VER4, minorvers);
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* Client owner */
*tl = txdr_unsigned(clp->nfsc_rev);
(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(exchflags);
*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
/* Set the implementation id4 */
*tl = txdr_unsigned(1);
(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
(void) nfsm_strtom(nd, version, strlen(version));
NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
verstime.tv_sec = 1293840000; /* Jan 1, 2011 */
verstime.tv_nsec = 0;
txdr_nfsv4time(&verstime, tl);
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
(int)nd->nd_repstat);
if (error != 0)
return (error);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
len = fxdr_unsigned(int, *(tl + 7));
if (len < 0 || len > NFSV4_OPAQUELIMIT) {
error = NFSERR_BADXDR;
goto nfsmout;
}
dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
M_WAITOK | M_ZERO);
dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
dsp->nfsclds_servownlen = len;
dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
dsp->nfsclds_sess.nfsess_sequenceid =
fxdr_unsigned(uint32_t, *tl++);
v41flags = fxdr_unsigned(uint32_t, *tl);
if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
NFSHASPNFSOPT(nmp)) {
NFSCL_DEBUG(1, "set PNFS\n");
NFSLOCKMNT(nmp);
nmp->nm_state |= NFSSTA_PNFS;
NFSUNLOCKMNT(nmp);
dsp->nfsclds_flags |= NFSCLDS_MDS;
}
if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
dsp->nfsclds_flags |= NFSCLDS_DS;
if (minorvers == NFSV42_MINORVERSION)
dsp->nfsclds_flags |= NFSCLDS_MINORV2;
if (len > 0)
nd->nd_repstat = nfsrv_mtostr(nd,
dsp->nfsclds_serverown, len);
if (nd->nd_repstat == 0) {
mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
NULL, MTX_DEF);
nfscl_initsessionslots(&dsp->nfsclds_sess);
*dspp = dsp;
} else
free(dsp, M_NFSCLDS);
}
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* Do the NFSv4.1 Create Session.
*/
int
nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
struct nfssockreq *nrp, struct nfsclds *dsp, uint32_t sequenceid, int mds,
struct ucred *cred, NFSPROC_T *p)
{
uint32_t crflags, maxval, *tl;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
int error, irdcnt, minorvers;
/* Make sure nm_rsize, nm_wsize is set. */
if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
nmp->nm_rsize = NFS_MAXBSIZE;
if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
nmp->nm_wsize = NFS_MAXBSIZE;
if (dsp == NULL)
minorvers = nmp->nm_minorvers;
else if ((dsp->nfsclds_flags & NFSCLDS_MINORV2) != 0)
minorvers = NFSV42_MINORVERSION;
else
minorvers = NFSV41_MINORVERSION;
nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL,
NFS_VER4, minorvers);
NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
*tl++ = sep->nfsess_clientid.lval[0];
*tl++ = sep->nfsess_clientid.lval[1];
*tl++ = txdr_unsigned(sequenceid);
crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
crflags |= NFSV4CRSESS_CONNBACKCHAN;
*tl = txdr_unsigned(crflags);
/* Fill in fore channel attributes. */
NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
*tl++ = 0; /* Header pad size */
if ((nd->nd_flag & ND_NFSV42) != 0 && mds != 0 && sb_max_adj >=
nmp->nm_wsize && sb_max_adj >= nmp->nm_rsize) {
/*
* NFSv4.2 Extended Attribute operations may want to do
* requests/replies that are larger than nm_rsize/nm_wsize.
*/
*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
} else {
*tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);
*tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);
}
*tl++ = txdr_unsigned(4096); /* Max response size cached */
*tl++ = txdr_unsigned(20); /* Max operations */
*tl++ = txdr_unsigned(64); /* Max slots */
*tl = 0; /* No rdma ird */
/* Fill in back channel attributes. */
NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
*tl++ = 0; /* Header pad size */
*tl++ = txdr_unsigned(10000); /* Max request size */
*tl++ = txdr_unsigned(10000); /* Max response size */
*tl++ = txdr_unsigned(4096); /* Max response size cached */
*tl++ = txdr_unsigned(4); /* Max operations */
*tl++ = txdr_unsigned(NFSV4_CBSLOTS); /* Max slots */
*tl = 0; /* No rdma ird */
NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFS_CALLBCKPROG); /* Call back prog # */
/* Allow AUTH_SYS callbacks as uid, gid == 0. */
*tl++ = txdr_unsigned(1); /* Auth_sys only */
*tl++ = txdr_unsigned(AUTH_SYS); /* AUTH_SYS type */
*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
*tl++ = 0; /* Null machine name */
*tl++ = 0; /* Uid == 0 */
*tl++ = 0; /* Gid == 0 */
*tl = 0; /* No additional gids */
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0)
return (error);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
2 * NFSX_UNSIGNED);
bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
crflags = fxdr_unsigned(uint32_t, *tl);
if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
NFSLOCKMNT(nmp);
nmp->nm_state |= NFSSTA_SESSPERSIST;
NFSUNLOCKMNT(nmp);
}
/* Get the fore channel slot count. */
NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
tl++; /* Skip the header pad size. */
/* Make sure nm_wsize is small enough. */
maxval = fxdr_unsigned(uint32_t, *tl++);
while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
if (nmp->nm_wsize > 8096)
nmp->nm_wsize /= 2;
else
break;
}
sep->nfsess_maxreq = maxval;
/* Make sure nm_rsize is small enough. */
maxval = fxdr_unsigned(uint32_t, *tl++);
while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
if (nmp->nm_rsize > 8096)
nmp->nm_rsize /= 2;
else
break;
}
sep->nfsess_maxresp = maxval;
sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
tl++;
sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
irdcnt = fxdr_unsigned(int, *tl);
if (irdcnt < 0 || irdcnt > 1) {
error = NFSERR_BADXDR;
goto nfsmout;
}
if (irdcnt > 0)
NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
/* and the back channel slot count. */
NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
tl += 5;
sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
}
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* Do the NFSv4.1 Destroy Session.
*/
int
nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
int error;
struct nfsclsession *tsep;
nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL, 0,
0);
NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
tsep = nfsmnt_mdssession(nmp);
bcopy(tsep->nfsess_sessionid, tl, NFSX_V4SESSIONID);
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0)
return (error);
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* Do the NFSv4.1 Destroy Client.
*/
int
nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
int error;
struct nfsclsession *tsep;
nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
0);
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
tsep = nfsmnt_mdssession(nmp);
*tl++ = tsep->nfsess_clientid.lval[0];
*tl = tsep->nfsess_clientid.lval[1];
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0)
return (error);
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* Do the NFSv4.1 LayoutGet.
*/
static int
nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p,
void *stuff)
{
struct nfsrv_descript nfsd, *nd = &nfsd;
int error;
nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
0);
nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
layouttype, layoutlen, 0);
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
if (error != 0)
return (error);
if (nd->nd_repstat == 0)
error = nfsrv_parselayoutget(nmp, nd, stateidp, retonclosep,
flhp);
if (error == 0 && nd->nd_repstat != 0)
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* Do the NFSv4.1 Get Device Info.
*/
int
nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
NFSPROC_T *p)
{
uint32_t cnt, *tl, vers, minorvers;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
struct sockaddr_in sin, ssin;
struct sockaddr_in6 sin6, ssin6;
struct nfsclds *dsp = NULL, **dspp, **gotdspp;
struct nfscldevinfo *ndi;
int addrcnt = 0, bitcnt, error, gotminor, gotvers, i, isudp, j;
int stripecnt;
uint8_t stripeindex;
sa_family_t af, safilled;
ssin.sin_port = 0; /* To shut up compiler. */
ssin.sin_addr.s_addr = 0; /* ditto */
*ndip = NULL;
ndi = NULL;
gotdspp = NULL;
nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
0);
NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
*tl++ = txdr_unsigned(layouttype);
*tl++ = txdr_unsigned(100000);
if (notifybitsp != NULL && *notifybitsp != 0) {
*tl = txdr_unsigned(1); /* One word of bits. */
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(*notifybitsp);
} else
*tl = txdr_unsigned(0);
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0)
return (error);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (layouttype != fxdr_unsigned(int, *tl))
printf("EEK! devinfo layout type not same!\n");
if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
stripecnt = fxdr_unsigned(int, *tl);
NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
if (stripecnt < 1 || stripecnt > 4096) {
printf("pNFS File layout devinfo stripecnt %d:"
" out of range\n", stripecnt);
error = NFSERR_BADXDR;
goto nfsmout;
}
NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
NFSX_UNSIGNED);
addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
if (addrcnt < 1 || addrcnt > 128) {
printf("NFS devinfo addrcnt %d: out of range\n",
addrcnt);
error = NFSERR_BADXDR;
goto nfsmout;
}
/*
* Now we know how many stripe indices and addresses, so
* we can allocate the structure the correct size.
*/
i = (stripecnt * sizeof(uint8_t)) /
sizeof(struct nfsclds *) + 1;
NFSCL_DEBUG(4, "stripeindices=%d\n", i);
ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
M_ZERO);
NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
NFSX_V4DEVICEID);
ndi->nfsdi_refcnt = 0;
ndi->nfsdi_flags = NFSDI_FILELAYOUT;
ndi->nfsdi_stripecnt = stripecnt;
ndi->nfsdi_addrcnt = addrcnt;
/* Fill in the stripe indices. */
for (i = 0; i < stripecnt; i++) {
stripeindex = fxdr_unsigned(uint8_t, *tl++);
NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
if (stripeindex >= addrcnt) {
printf("pNFS File Layout devinfo"
" stripeindex %d: too big\n",
(int)stripeindex);
error = NFSERR_BADXDR;
goto nfsmout;
}
nfsfldi_setstripeindex(ndi, i, stripeindex);
}
} else if (layouttype == NFSLAYOUT_FLEXFILE) {
/* For Flex File, we only get one address list. */
ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
M_NFSDEVINFO, M_WAITOK | M_ZERO);
NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
NFSX_V4DEVICEID);
ndi->nfsdi_refcnt = 0;
ndi->nfsdi_flags = NFSDI_FLEXFILE;
addrcnt = ndi->nfsdi_addrcnt = 1;
}
/* Now, dissect the server address(es). */
safilled = AF_UNSPEC;
for (i = 0; i < addrcnt; i++) {
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
cnt = fxdr_unsigned(uint32_t, *tl);
if (cnt == 0) {
printf("NFS devinfo 0 len addrlist\n");
error = NFSERR_BADXDR;
goto nfsmout;
}
dspp = nfsfldi_addr(ndi, i);
safilled = AF_UNSPEC;
for (j = 0; j < cnt; j++) {
error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
&isudp);
if (error != 0 && error != EPERM) {
error = NFSERR_BADXDR;
goto nfsmout;
}
if (error == 0 && isudp == 0) {
/*
* The priority is:
* - Same address family.
* Save the address and dspp, so that
* the connection can be done after
* parsing is complete.
*/
if (safilled == AF_UNSPEC ||
(af == nmp->nm_nam->sa_family &&
safilled != nmp->nm_nam->sa_family)
) {
if (af == AF_INET)
ssin = sin;
else
ssin6 = sin6;
safilled = af;
gotdspp = dspp;
}
}
}
}
gotvers = NFS_VER4; /* Default NFSv4.1 for File Layout. */
gotminor = NFSV41_MINORVERSION;
/* For Flex File, we will take one of the versions to use. */
if (layouttype == NFSLAYOUT_FLEXFILE) {
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
j = fxdr_unsigned(int, *tl);
if (j < 1 || j > NFSDEV_MAXVERS) {
printf("pNFS: too many versions\n");
error = NFSERR_BADXDR;
goto nfsmout;
}
gotvers = 0;
gotminor = 0;
for (i = 0; i < j; i++) {
NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
vers = fxdr_unsigned(uint32_t, *tl++);
minorvers = fxdr_unsigned(uint32_t, *tl++);
if (vers == NFS_VER3)
minorvers = 0;
if ((vers == NFS_VER4 && ((minorvers ==
NFSV41_MINORVERSION && gotminor == 0) ||
minorvers == NFSV42_MINORVERSION)) ||
(vers == NFS_VER3 && gotvers == 0)) {
gotvers = vers;
gotminor = minorvers;
/* We'll take this one. */
ndi->nfsdi_versindex = i;
ndi->nfsdi_vers = vers;
ndi->nfsdi_minorvers = minorvers;
ndi->nfsdi_rsize = fxdr_unsigned(
uint32_t, *tl++);
ndi->nfsdi_wsize = fxdr_unsigned(
uint32_t, *tl++);
if (*tl == newnfs_true)
ndi->nfsdi_flags |=
NFSDI_TIGHTCOUPLED;
else
ndi->nfsdi_flags &=
~NFSDI_TIGHTCOUPLED;
}
}
if (gotvers == 0) {
printf("pNFS: no NFSv3, NFSv4.1 or NFSv4.2\n");
error = NFSERR_BADXDR;
goto nfsmout;
}
}
/* And the notify bits. */
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
bitcnt = fxdr_unsigned(int, *tl);
if (bitcnt > 0) {
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
if (notifybitsp != NULL)
*notifybitsp =
fxdr_unsigned(uint32_t, *tl);
}
if (safilled != AF_UNSPEC) {
KASSERT(ndi != NULL, ("ndi is NULL"));
*ndip = ndi;
} else
error = EPERM;
if (error == 0) {
/*
* Now we can do a TCP connection for the correct
* NFS version and IP address.
*/
error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
gotvers, gotminor, &dsp, p);
}
if (error == 0) {
KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
*gotdspp = dsp;
}
}
if (nd->nd_repstat != 0 && error == 0)
error = nd->nd_repstat;
nfsmout:
if (error != 0 && ndi != NULL)
nfscl_freedevinfo(ndi);
m_freem(nd->nd_mrep);
return (error);
}
/*
* Do the NFSv4.1 LayoutCommit.
*/
int
nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
int layouttype, struct ucred *cred, NFSPROC_T *p, void *stuff)
{
uint32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
int error;
nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
0, 0);
NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
NFSX_STATEID);
txdr_hyper(off, tl);
tl += 2;
txdr_hyper(len, tl);
tl += 2;
if (reclaim != 0)
*tl++ = newnfs_true;
else
*tl++ = newnfs_false;
*tl++ = txdr_unsigned(stateidp->seqid);
*tl++ = stateidp->other[0];
*tl++ = stateidp->other[1];
*tl++ = stateidp->other[2];
*tl++ = newnfs_true;
if (lastbyte < off)
lastbyte = off;
else if (lastbyte >= (off + len))
lastbyte = off + len - 1;
txdr_hyper(lastbyte, tl);
tl += 2;
*tl++ = newnfs_false;
*tl++ = txdr_unsigned(layouttype);
/* All supported layouts are 0 length. */
*tl = txdr_unsigned(0);
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0)
return (error);
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* Do the NFSv4.1 LayoutReturn.
*/
int
nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
uint32_t stat, uint32_t op, char *devid)
{
uint32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
uint64_t tu64;
int error;
nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
0, 0);
NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
if (reclaim != 0)
*tl++ = newnfs_true;
else
*tl++ = newnfs_false;
*tl++ = txdr_unsigned(layouttype);
*tl++ = txdr_unsigned(iomode);
*tl = txdr_unsigned(layoutreturn);
if (layoutreturn == NFSLAYOUTRETURN_FILE) {
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
NFSX_UNSIGNED);
txdr_hyper(offset, tl);
tl += 2;
txdr_hyper(len, tl);
tl += 2;
NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
*tl++ = txdr_unsigned(stateidp->seqid);
*tl++ = stateidp->other[0];
*tl++ = stateidp->other[1];
*tl++ = stateidp->other[2];
if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
*tl = txdr_unsigned(0);
else if (layouttype == NFSLAYOUT_FLEXFILE) {
if (stat != 0) {
*tl = txdr_unsigned(2 * NFSX_HYPER +
NFSX_STATEID + NFSX_V4DEVICEID + 5 *
NFSX_UNSIGNED);
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER +
NFSX_STATEID + NFSX_V4DEVICEID + 5 *
NFSX_UNSIGNED);
*tl++ = txdr_unsigned(1); /* One error. */
tu64 = 0; /* Offset. */
txdr_hyper(tu64, tl); tl += 2;
tu64 = UINT64_MAX; /* Length. */
txdr_hyper(tu64, tl); tl += 2;
NFSBCOPY(stateidp, tl, NFSX_STATEID);
tl += (NFSX_STATEID / NFSX_UNSIGNED);
*tl++ = txdr_unsigned(1); /* One error. */
NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
*tl++ = txdr_unsigned(stat);
*tl++ = txdr_unsigned(op);
} else {
*tl = txdr_unsigned(2 * NFSX_UNSIGNED);
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
/* No ioerrs. */
*tl++ = 0;
}
*tl = 0; /* No stats yet. */
}
}
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0)
return (error);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
if (*tl != 0) {
NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
stateidp->other[0] = *tl++;
stateidp->other[1] = *tl++;
stateidp->other[2] = *tl;
}
} else
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* Do the NFSv4.2 LayoutError.
*/
static int
nfsrpc_layouterror(struct nfsmount *nmp, uint8_t *fh, int fhlen, uint64_t offset,
uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
uint32_t stat, uint32_t op, char *devid)
{
uint32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
int error;
nfscl_reqstart(nd, NFSPROC_LAYOUTERROR, nmp, fh, fhlen, NULL, NULL,
0, 0);
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
txdr_hyper(offset, tl); tl += 2;
txdr_hyper(len, tl); tl += 2;
*tl++ = txdr_unsigned(stateidp->seqid);
*tl++ = stateidp->other[0];
*tl++ = stateidp->other[1];
*tl++ = stateidp->other[2];
*tl++ = txdr_unsigned(1);
NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
*tl++ = txdr_unsigned(stat);
*tl = txdr_unsigned(op);
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0)
return (error);
if (nd->nd_repstat != 0)
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* Acquire a layout and devinfo, if possible. The caller must have acquired
* a reference count on the nfsclclient structure before calling this.
* Return the layout in lypp with a reference count on it, if successful.
*/
static int
nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
int iomode, uint32_t rw, uint32_t *notifybitsp, nfsv4stateid_t *stateidp,
uint64_t off, struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
{
struct nfscllayout *lyp;
struct nfsclflayout *flp;
struct nfsclflayouthead flh;
int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
nfsv4stateid_t stateid;
struct nfsclsession *tsep;
*lypp = NULL;
if (NFSHASFLEXFILE(nmp))
layouttype = NFSLAYOUT_FLEXFILE;
else
layouttype = NFSLAYOUT_NFSV4_1_FILES;
/*
* If lyp is returned non-NULL, there will be a refcnt (shared lock)
* on it, iff flp != NULL or a lock (exclusive lock) on it iff
* flp == NULL.
*/
lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
off, rw, &flp, &recalled);
islocked = 0;
if (lyp == NULL || flp == NULL) {
if (recalled != 0)
return (EIO);
LIST_INIT(&flh);
tsep = nfsmnt_mdssession(nmp);
layoutlen = tsep->nfsess_maxcache -
(NFSX_STATEID + 3 * NFSX_UNSIGNED);
if (lyp == NULL) {
stateid.seqid = 0;
stateid.other[0] = stateidp->other[0];
stateid.other[1] = stateidp->other[1];
stateid.other[2] = stateidp->other[2];
error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
(uint64_t)0, layouttype, layoutlen, &stateid,
&retonclose, &flh, cred, p, NULL);
} else {
islocked = 1;
stateid.seqid = lyp->nfsly_stateid.seqid;
stateid.other[0] = lyp->nfsly_stateid.other[0];
stateid.other[1] = lyp->nfsly_stateid.other[1];
stateid.other[2] = lyp->nfsly_stateid.other[2];
error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
nfhp->nfh_len, iomode, off, UINT64_MAX,
(uint64_t)0, layouttype, layoutlen, &stateid,
&retonclose, &flh, cred, p, NULL);
}
error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
&flh, layouttype, error, NULL, cred, p);
if (error == 0)
*lypp = lyp;
else if (islocked != 0)
nfscl_rellayout(lyp, 1);
} else
*lypp = lyp;
return (error);
}
/*
* Do a TCP connection plus exchange id and create session.
* If successful, a "struct nfsclds" is linked into the list for the
* mount point and a pointer to it is returned.
*/
static int
nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
struct sockaddr_in6 *sin6, sa_family_t af, int vers, int minorvers,
struct nfsclds **dspp, NFSPROC_T *p)
{
struct sockaddr_in *msad, *sad;
struct sockaddr_in6 *msad6, *sad6;
struct nfsclclient *clp;
struct nfssockreq *nrp;
struct nfsclds *dsp, *tdsp;
int error, firsttry;
enum nfsclds_state retv;
uint32_t sequenceid = 0;
KASSERT(nmp->nm_sockreq.nr_cred != NULL,
("nfsrpc_fillsa: NULL nr_cred"));
NFSLOCKCLSTATE();
clp = nmp->nm_clp;
NFSUNLOCKCLSTATE();
if (clp == NULL)
return (EPERM);
if (af == AF_INET) {
NFSLOCKMNT(nmp);
/*
* Check to see if we already have a session for this
* address that is usable for a DS.
* Note that the MDS's address is in a different place
* than the sessions already acquired for DS's.
*/
msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
tdsp = TAILQ_FIRST(&nmp->nm_sess);
while (tdsp != NULL) {
if (msad != NULL && msad->sin_family == AF_INET &&
sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
sin->sin_port == msad->sin_port &&
(tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
tdsp->nfsclds_sess.nfsess_defunct == 0) {
*dspp = tdsp;
NFSUNLOCKMNT(nmp);
NFSCL_DEBUG(4, "fnd same addr\n");
return (0);
}
tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
msad = (struct sockaddr_in *)
tdsp->nfsclds_sockp->nr_nam;
else
msad = NULL;
}
NFSUNLOCKMNT(nmp);
/* No IP address match, so look for new/trunked one. */
sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
sad->sin_len = sizeof(*sad);
sad->sin_family = AF_INET;
sad->sin_port = sin->sin_port;
sad->sin_addr.s_addr = sin->sin_addr.s_addr;
nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
nrp->nr_nam = (struct sockaddr *)sad;
} else if (af == AF_INET6) {
NFSLOCKMNT(nmp);
/*
* Check to see if we already have a session for this
* address that is usable for a DS.
* Note that the MDS's address is in a different place
* than the sessions already acquired for DS's.
*/
msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
tdsp = TAILQ_FIRST(&nmp->nm_sess);
while (tdsp != NULL) {
if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
&msad6->sin6_addr) &&
sin6->sin6_port == msad6->sin6_port &&
(tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
tdsp->nfsclds_sess.nfsess_defunct == 0) {
*dspp = tdsp;
NFSUNLOCKMNT(nmp);
return (0);
}
tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
msad6 = (struct sockaddr_in6 *)
tdsp->nfsclds_sockp->nr_nam;
else
msad6 = NULL;
}
NFSUNLOCKMNT(nmp);
/* No IP address match, so look for new/trunked one. */
sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
sad6->sin6_len = sizeof(*sad6);
sad6->sin6_family = AF_INET6;
sad6->sin6_port = sin6->sin6_port;
NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
sizeof(struct in6_addr));
nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
nrp->nr_nam = (struct sockaddr *)sad6;
} else
return (EPERM);
nrp->nr_sotype = SOCK_STREAM;
mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
nrp->nr_prog = NFS_PROG;
nrp->nr_vers = vers;
/*
* Use the credentials that were used for the mount, which are
* in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
* Ref. counting the credentials with crhold() is probably not
* necessary, since nm_sockreq.nr_cred won't be crfree()'d until
* unmount, but I did it anyhow.
*/
nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
error = newnfs_connect(nmp, nrp, NULL, p, 0, false, &nrp->nr_client);
NFSCL_DEBUG(3, "DS connect=%d\n", error);
dsp = NULL;
/* Now, do the exchangeid and create session. */
if (error == 0) {
if (vers == NFS_VER4) {
firsttry = 0;
do {
error = nfsrpc_exchangeid(nmp, clp, nrp,
minorvers, NFSV4EXCH_USEPNFSDS, &dsp,
nrp->nr_cred, p);
NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
if (error == NFSERR_MINORVERMISMATCH)
minorvers = NFSV42_MINORVERSION;
} while (error == NFSERR_MINORVERMISMATCH &&
firsttry++ == 0);
if (error != 0)
newnfs_disconnect(NULL, nrp);
} else {
dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
M_WAITOK | M_ZERO);
dsp->nfsclds_flags |= NFSCLDS_DS;
dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
NULL, MTX_DEF);
}
}
if (error == 0) {
dsp->nfsclds_sockp = nrp;
if (vers == NFS_VER4) {
NFSLOCKMNT(nmp);
retv = nfscl_getsameserver(nmp, dsp, &tdsp,
&sequenceid);
NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
if (retv == NFSDSP_USETHISSESSION &&
nfscl_dssameconn != 0) {
NFSLOCKDS(tdsp);
tdsp->nfsclds_flags |= NFSCLDS_SAMECONN;
NFSUNLOCKDS(tdsp);
NFSUNLOCKMNT(nmp);
/*
* If there is already a session for this
* server, use it.
*/
newnfs_disconnect(NULL, nrp);
nfscl_freenfsclds(dsp);
*dspp = tdsp;
return (0);
}
if (retv == NFSDSP_NOTFOUND)
sequenceid =
dsp->nfsclds_sess.nfsess_sequenceid;
NFSUNLOCKMNT(nmp);
error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
nrp, dsp, sequenceid, 0, nrp->nr_cred, p);
NFSCL_DEBUG(3, "DS createsess=%d\n", error);
}
} else {
NFSFREECRED(nrp->nr_cred);
NFSFREEMUTEX(&nrp->nr_mtx);
free(nrp->nr_nam, M_SONAME);
free(nrp, M_NFSSOCKREQ);
}
if (error == 0) {
NFSCL_DEBUG(3, "add DS session\n");
/*
* Put it at the end of the list. That way the list
* is ordered by when the entry was added. This matters
* since the one done first is the one that should be
* used for sequencid'ing any subsequent create sessions.
*/
NFSLOCKMNT(nmp);
TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
NFSUNLOCKMNT(nmp);
*dspp = dsp;
} else if (dsp != NULL) {
newnfs_disconnect(NULL, nrp);
nfscl_freenfsclds(dsp);
}
return (error);
}
/*
* Do the NFSv4.1 Reclaim Complete.
*/
int
nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
int error;
nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
0);
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
*tl = newnfs_false;
nd->nd_flag |= ND_USEGSSNAME;
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0)
return (error);
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* Initialize the slot tables for a session.
*/
static void
nfscl_initsessionslots(struct nfsclsession *sep)
{
int i;
for (i = 0; i < NFSV4_CBSLOTS; i++) {
if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
m_freem(sep->nfsess_cbslots[i].nfssl_reply);
NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
}
for (i = 0; i < 64; i++)
sep->nfsess_slotseq[i] = 0;
sep->nfsess_slots = 0;
}
/*
* Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
*/
int
nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
{
struct nfsnode *np = VTONFS(vp);
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
struct nfscllayout *layp;
struct nfscldevinfo *dip;
struct nfsclflayout *rflp;
struct mbuf *m, *m2;
struct nfsclwritedsdorpc *drpc, *tdrpc;
nfsv4stateid_t stateid;
struct ucred *newcred;
uint64_t lastbyte, len, off, oresid, xfer;
int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
void *lckp;
uint8_t *dev;
void *iovbase = NULL;
size_t iovlen = 0;
off_t offs = 0;
ssize_t resid = 0;
uint32_t op;
if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
(np->n_flag & NNOLAYOUT) != 0)
return (EIO);
/* Now, get a reference cnt on the clientid for this mount. */
if (nfscl_getref(nmp) == 0)
return (EIO);
/* Find an appropriate stateid. */
newcred = NFSNEWCRED(cred);
error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
rwaccess, 1, newcred, p, &stateid, &lckp);
if (error != 0) {
NFSFREECRED(newcred);
nfscl_relref(nmp);
return (error);
}
/* Search for a layout for this file. */
off = uiop->uio_offset;
layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
np->n_fhp->nfh_len, off, rwaccess, &rflp, &recalled);
if (layp == NULL || rflp == NULL) {
if (recalled != 0) {
NFSFREECRED(newcred);
if (lckp != NULL)
nfscl_lockderef(lckp);
nfscl_relref(nmp);
return (EIO);
}
if (layp != NULL) {
nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
layp = NULL;
}
/* Try and get a Layout, if it is supported. */
if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
(np->n_flag & NWRITEOPENED) != 0)
iolaymode = NFSLAYOUTIOMODE_RW;
else
iolaymode = NFSLAYOUTIOMODE_READ;
error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
rwaccess, NULL, &stateid, off, &layp, newcred, p);
if (error != 0) {
NFSLOCKNODE(np);
np->n_flag |= NNOLAYOUT;
NFSUNLOCKNODE(np);
if (lckp != NULL)
nfscl_lockderef(lckp);
NFSFREECRED(newcred);
if (layp != NULL)
nfscl_rellayout(layp, 0);
nfscl_relref(nmp);
return (error);
}
}
/*
* Loop around finding a layout that works for the first part of
* this I/O operation, and then call the function that actually
* does the RPC.
*/
eof = 0;
len = (uint64_t)uiop->uio_resid;
while (len > 0 && error == 0 && eof == 0) {
off = uiop->uio_offset;
error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
if (error == 0) {
oresid = xfer = (uint64_t)uiop->uio_resid;
if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
xfer = rflp->nfsfl_end - rflp->nfsfl_off;
/*
* For Flex File layout with mirrored DSs, select one
* of them at random for reads. For writes and commits,
* do all mirrors.
*/
m = NULL;
tdrpc = drpc = NULL;
firstmirror = 0;
mirrorcnt = 1;
if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
(mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
if (rwaccess == NFSV4OPEN_ACCESSREAD) {
firstmirror = arc4random() % mirrorcnt;
mirrorcnt = firstmirror + 1;
} else {
if (docommit == 0) {
/*
* Save values, so uiop can be
* rolled back upon a write
* error.
*/
offs = uiop->uio_offset;
resid = uiop->uio_resid;
iovbase =
uiop->uio_iov->iov_base;
iovlen = uiop->uio_iov->iov_len;
m = nfsm_uiombuflist(uiop, len,
0);
}
tdrpc = drpc = malloc(sizeof(*drpc) *
(mirrorcnt - 1), M_TEMP, M_WAITOK |
M_ZERO);
}
}
for (i = firstmirror; i < mirrorcnt && error == 0; i++){
m2 = NULL;
if (m != NULL && i < mirrorcnt - 1)
m2 = m_copym(m, 0, M_COPYALL, M_WAITOK);
else {
m2 = m;
m = NULL;
}
if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0) {
dev = rflp->nfsfl_ffm[i].dev;
dip = nfscl_getdevinfo(nmp->nm_clp, dev,
rflp->nfsfl_ffm[i].devp);
} else {
dev = rflp->nfsfl_dev;
dip = nfscl_getdevinfo(nmp->nm_clp, dev,
rflp->nfsfl_devp);
}
if (dip != NULL) {
if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
!= 0)
error = nfscl_dofflayoutio(vp,
uiop, iomode, must_commit,
&eof, &stateid, rwaccess,
dip, layp, rflp, off, xfer,
i, docommit, m2, tdrpc,
newcred, p);
else
error = nfscl_doflayoutio(vp,
uiop, iomode, must_commit,
&eof, &stateid, rwaccess,
dip, layp, rflp, off, xfer,
docommit, newcred, p);
nfscl_reldevinfo(dip);
} else {
if (m2 != NULL)
m_freem(m2);
error = EIO;
}
tdrpc++;
}
if (m != NULL)
m_freem(m);
tdrpc = drpc;
timo = hz / 50; /* Wait for 20msec. */
if (timo < 1)
timo = 1;
for (i = firstmirror; i < mirrorcnt - 1 &&
tdrpc != NULL; i++, tdrpc++) {
/*
* For the unused drpc entries, both inprog and
* err == 0, so this loop won't break.
*/
while (tdrpc->inprog != 0 && tdrpc->done == 0)
tsleep(&tdrpc->tsk, PVFS, "clrpcio",
timo);
if (error == 0 && tdrpc->err != 0)
error = tdrpc->err;
if (rwaccess != NFSV4OPEN_ACCESSREAD &&
docommit == 0 && *must_commit == 0 &&
tdrpc->must_commit == 1)
*must_commit = 1;
}
free(drpc, M_TEMP);
if (error == 0) {
if (mirrorcnt > 1 && rwaccess ==
NFSV4OPEN_ACCESSWRITE && docommit == 0) {
NFSLOCKCLSTATE();
layp->nfsly_flags |= NFSLY_WRITTEN;
NFSUNLOCKCLSTATE();
}
lastbyte = off + xfer - 1;
NFSLOCKCLSTATE();
if (lastbyte > layp->nfsly_lastbyte)
layp->nfsly_lastbyte = lastbyte;
NFSUNLOCKCLSTATE();
} else if (error == NFSERR_OPENMODE &&
rwaccess == NFSV4OPEN_ACCESSREAD) {
NFSLOCKMNT(nmp);
nmp->nm_state |= NFSSTA_OPENMODE;
NFSUNLOCKMNT(nmp);
} else if ((error == NFSERR_NOSPC ||
error == NFSERR_IO || error == NFSERR_NXIO) &&
nmp->nm_minorvers == NFSV42_MINORVERSION) {
if (docommit != 0)
op = NFSV4OP_COMMIT;
else if (rwaccess == NFSV4OPEN_ACCESSREAD)
op = NFSV4OP_READ;
else
op = NFSV4OP_WRITE;
nfsrpc_layouterror(nmp, np->n_fhp->nfh_fh,
np->n_fhp->nfh_len, off, xfer,
&layp->nfsly_stateid, newcred, p, error, op,
dip->nfsdi_deviceid);
error = EIO;
} else
error = EIO;
if (error == 0)
len -= (oresid - (uint64_t)uiop->uio_resid);
else if (mirrorcnt > 1 && rwaccess ==
NFSV4OPEN_ACCESSWRITE && docommit == 0) {
/*
* In case the rpc gets retried, roll the
* uio fields changed by nfsm_uiombuflist()
* back.
*/
uiop->uio_offset = offs;
uiop->uio_resid = resid;
uiop->uio_iov->iov_base = iovbase;
uiop->uio_iov->iov_len = iovlen;
}
}
}
if (lckp != NULL)
nfscl_lockderef(lckp);
NFSFREECRED(newcred);
nfscl_rellayout(layp, 0);
nfscl_relref(nmp);
return (error);
}
/*
* Find a file layout that will handle the first bytes of the requested
* range and return the information from it needed to the I/O operation.
*/
int
nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
struct nfsclflayout **retflpp)
{
struct nfsclflayout *flp, *nflp, *rflp;
uint32_t rw;
rflp = NULL;
rw = rwaccess;
/* For reading, do the Read list first and then the Write list. */
do {
if (rw == NFSV4OPEN_ACCESSREAD)
flp = LIST_FIRST(&lyp->nfsly_flayread);
else
flp = LIST_FIRST(&lyp->nfsly_flayrw);
while (flp != NULL) {
nflp = LIST_NEXT(flp, nfsfl_list);
if (flp->nfsfl_off > off)
break;
if (flp->nfsfl_end > off &&
(rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
rflp = flp;
flp = nflp;
}
if (rw == NFSV4OPEN_ACCESSREAD)
rw = NFSV4OPEN_ACCESSWRITE;
else
rw = 0;
} while (rw != 0);
if (rflp != NULL) {
/* This one covers the most bytes starting at off. */
*retflpp = rflp;
return (0);
}
return (EIO);
}
/*
* Do I/O using an NFSv4.1 or NFSv4.2 file layout.
*/
static int
nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
{
uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
int commit_thru_mds, error, stripe_index, stripe_pos, minorvers;
struct nfsnode *np;
struct nfsfh *fhp;
struct nfsclds **dspp;
np = VTONFS(vp);
rel_off = off - flp->nfsfl_patoff;
stripe_unit_size = flp->nfsfl_util & NFSFLAYUTIL_STRIPE_MASK;
stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
dp->nfsdi_stripecnt;
transfer = stripe_unit_size - (rel_off % stripe_unit_size);
error = 0;
/* Loop around, doing I/O for each stripe unit. */
while (len > 0 && error == 0) {
stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
dspp = nfsfldi_addr(dp, stripe_index);
if (((*dspp)->nfsclds_flags & NFSCLDS_MINORV2) != 0)
minorvers = NFSV42_MINORVERSION;
else
minorvers = NFSV41_MINORVERSION;
if (len > transfer && docommit == 0)
xfer = transfer;
else
xfer = len;
if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
/* Dense layout. */
if (stripe_pos >= flp->nfsfl_fhcnt)
return (EIO);
fhp = flp->nfsfl_fh[stripe_pos];
io_off = (rel_off / (stripe_unit_size *
dp->nfsdi_stripecnt)) * stripe_unit_size +
rel_off % stripe_unit_size;
} else {
/* Sparse layout. */
if (flp->nfsfl_fhcnt > 1) {
if (stripe_index >= flp->nfsfl_fhcnt)
return (EIO);
fhp = flp->nfsfl_fh[stripe_index];
} else if (flp->nfsfl_fhcnt == 1)
fhp = flp->nfsfl_fh[0];
else
fhp = np->n_fhp;
io_off = off;
}
if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
commit_thru_mds = 1;
if (docommit != 0)
error = EIO;
} else {
commit_thru_mds = 0;
NFSLOCKNODE(np);
np->n_flag |= NDSCOMMIT;
NFSUNLOCKNODE(np);
}
if (docommit != 0) {
if (error == 0)
error = nfsrpc_commitds(vp, io_off, xfer,
*dspp, fhp, NFS_VER4, minorvers, cred, p);
if (error == 0) {
/*
* Set both eof and uio_resid = 0 to end any
* loops.
*/
*eofp = 1;
uiop->uio_resid = 0;
} else {
NFSLOCKNODE(np);
np->n_flag &= ~NDSCOMMIT;
NFSUNLOCKNODE(np);
}
} else if (rwflag == NFSV4OPEN_ACCESSREAD)
error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
io_off, xfer, fhp, 0, NFS_VER4, minorvers, cred, p);
else {
error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
0, NFS_VER4, minorvers, cred, p);
if (error == 0) {
NFSLOCKCLSTATE();
lyp->nfsly_flags |= NFSLY_WRITTEN;
NFSUNLOCKCLSTATE();
}
}
if (error == 0) {
transfer = stripe_unit_size;
stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
len -= xfer;
off += xfer;
}
}
return (error);
}
/*
* Do I/O using an NFSv4.1 flex file layout.
*/
static int
nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
uint64_t len, int mirror, int docommit, struct mbuf *mp,
struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
{
uint64_t xfer;
int error;
struct nfsnode *np;
struct nfsfh *fhp;
struct nfsclds **dspp;
struct ucred *tcred;
struct mbuf *m, *m2;
uint32_t copylen;
np = VTONFS(vp);
error = 0;
NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
(uintmax_t)len);
/* Loop around, doing I/O for each stripe unit. */
while (len > 0 && error == 0) {
dspp = nfsfldi_addr(dp, 0);
fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
stateidp = &flp->nfsfl_ffm[mirror].st;
NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
tcred = NFSNEWCRED(cred);
tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
tcred->cr_ngroups = 1;
} else
tcred = cred;
if (rwflag == NFSV4OPEN_ACCESSREAD)
copylen = dp->nfsdi_rsize;
else {
copylen = dp->nfsdi_wsize;
if (len > copylen && mp != NULL) {
/*
* When a mirrored configuration needs to do
* multiple writes to each mirror, all writes
* except the last one must be a multiple of
* 4 bytes. This is required so that the XDR
* does not need padding.
* If possible, clip the size to an exact
* multiple of the mbuf length, so that the
* split will be on an mbuf boundary.
*/
copylen &= 0xfffffffc;
if (copylen > mp->m_len)
copylen = copylen / mp->m_len *
mp->m_len;
}
}
NFSLOCKNODE(np);
np->n_flag |= NDSCOMMIT;
NFSUNLOCKNODE(np);
if (len > copylen && docommit == 0)
xfer = copylen;
else
xfer = len;
if (docommit != 0) {
if (error == 0) {
/*
* Do last mirrored DS commit with this thread.
*/
if (mirror < flp->nfsfl_mirrorcnt - 1)
error = nfsio_commitds(vp, off, xfer,
*dspp, fhp, dp->nfsdi_vers,
dp->nfsdi_minorvers, drpc, tcred,
p);
else
error = nfsrpc_commitds(vp, off, xfer,
*dspp, fhp, dp->nfsdi_vers,
dp->nfsdi_minorvers, tcred, p);
NFSCL_DEBUG(4, "commitds=%d\n", error);
if (error != 0 && error != EACCES && error !=
ESTALE) {
NFSCL_DEBUG(4,
"DS layreterr for commit\n");
nfscl_dserr(NFSV4OP_COMMIT, error, dp,
lyp, *dspp);
}
}
NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
if (error == 0) {
/*
* Set both eof and uio_resid = 0 to end any
* loops.
*/
*eofp = 1;
uiop->uio_resid = 0;
} else {
NFSLOCKNODE(np);
np->n_flag &= ~NDSCOMMIT;
NFSUNLOCKNODE(np);
}
} else if (rwflag == NFSV4OPEN_ACCESSREAD) {
error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
off, xfer, fhp, 1, dp->nfsdi_vers,
dp->nfsdi_minorvers, tcred, p);
NFSCL_DEBUG(4, "readds=%d\n", error);
if (error != 0 && error != EACCES && error != ESTALE) {
NFSCL_DEBUG(4, "DS layreterr for read\n");
nfscl_dserr(NFSV4OP_READ, error, dp, lyp,
*dspp);
}
} else {
if (flp->nfsfl_mirrorcnt == 1) {
error = nfsrpc_writeds(vp, uiop, iomode,
must_commit, stateidp, *dspp, off, xfer,
fhp, 0, 1, dp->nfsdi_vers,
dp->nfsdi_minorvers, tcred, p);
if (error == 0) {
NFSLOCKCLSTATE();
lyp->nfsly_flags |= NFSLY_WRITTEN;
NFSUNLOCKCLSTATE();
}
} else {
m = mp;
if (xfer < len) {
/* The mbuf list must be split. */
m2 = nfsm_split(mp, xfer);
if (m2 != NULL)
mp = m2;
else {
m_freem(mp);
error = EIO;
}
}
NFSCL_DEBUG(4, "mcopy len=%jd xfer=%jd\n",
(uintmax_t)len, (uintmax_t)xfer);
/*
* Do last write to a mirrored DS with this
* thread.
*/
if (error == 0) {
if (mirror < flp->nfsfl_mirrorcnt - 1)
error = nfsio_writedsmir(vp,
iomode, must_commit,
stateidp, *dspp, off,
xfer, fhp, m,
dp->nfsdi_vers,
dp->nfsdi_minorvers, drpc,
tcred, p);
else
error = nfsrpc_writedsmir(vp,
iomode, must_commit,
stateidp, *dspp, off,
xfer, fhp, m,
dp->nfsdi_vers,
dp->nfsdi_minorvers, tcred,
p);
}
NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
if (error != 0 && error != EACCES && error !=
ESTALE) {
NFSCL_DEBUG(4,
"DS layreterr for write\n");
nfscl_dserr(NFSV4OP_WRITE, error, dp,
lyp, *dspp);
}
}
}
NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
if (error == 0) {
len -= xfer;
off += xfer;
}
if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
NFSFREECRED(tcred);
}
NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
return (error);
}
/*
* The actual read RPC done to a DS.
*/
static int
nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
int attrflag, error, retlen;
struct nfsrv_descript nfsd;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
struct nfsrv_descript *nd = &nfsd;
struct nfssockreq *nrp;
struct nfsvattr na;
nd->nd_mrep = NULL;
if (vers == 0 || vers == NFS_VER4) {
nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
vers = NFS_VER4;
NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
if (flex != 0)
nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
else
nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
} else {
nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READ]);
NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READDS]);
NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
}
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
txdr_hyper(io_off, tl);
*(tl + 2) = txdr_unsigned(len);
nrp = dsp->nfsclds_sockp;
NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
if (nrp == NULL)
/* If NULL, use the MDS socket. */
nrp = &nmp->nm_sockreq;
error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
error);
if (error != 0)
return (error);
if (vers == NFS_VER3) {
error = nfscl_postop_attr(nd, &na, &attrflag, NULL);
NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
if (error != 0)
goto nfsmout;
}
if (nd->nd_repstat != 0) {
error = nd->nd_repstat;
goto nfsmout;
}
if (vers == NFS_VER3) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
*eofp = fxdr_unsigned(int, *(tl + 1));
} else {
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
*eofp = fxdr_unsigned(int, *tl);
}
NFSM_STRSIZ(retlen, len);
NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
error = nfsm_mbufuio(nd, uiop, retlen);
nfsmout:
if (nd->nd_mrep != NULL)
m_freem(nd->nd_mrep);
return (error);
}
/*
* The actual write RPC done to a DS.
*/
static int
nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
int32_t backup;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
struct nfssockreq *nrp;
struct nfsvattr na;
KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
nd->nd_mrep = NULL;
if (vers == 0 || vers == NFS_VER4) {
nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
vers = NFS_VER4;
if (flex != 0)
nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
else
nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
} else {
nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
}
txdr_hyper(io_off, tl);
tl += 2;
if (vers == NFS_VER3)
*tl++ = txdr_unsigned(len);
*tl++ = txdr_unsigned(*iomode);
*tl = txdr_unsigned(len);
nfsm_uiombuf(nd, uiop, len);
nrp = dsp->nfsclds_sockp;
if (nrp == NULL)
/* If NULL, use the MDS socket. */
nrp = &nmp->nm_sockreq;
error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
nd->nd_repstat);
if (error != 0)
return (error);
if (nd->nd_repstat != 0) {
/*
* In case the rpc gets retried, roll
* the uio fields changed by nfsm_uiombuf()
* back.
*/
uiop->uio_offset -= len;
uiop->uio_resid += len;
uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base - len;
uiop->uio_iov->iov_len += len;
error = nd->nd_repstat;
} else {
if (vers == NFS_VER3) {
error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
NULL);
NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
if (error != 0)
goto nfsmout;
}
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
rlen = fxdr_unsigned(int, *tl++);
NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
if (rlen == 0) {
error = NFSERR_IO;
goto nfsmout;
} else if (rlen < len) {
backup = len - rlen;
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base - backup;
uiop->uio_iov->iov_len += backup;
uiop->uio_offset -= backup;
uiop->uio_resid += backup;
len = rlen;
}
commit = fxdr_unsigned(int, *tl++);
/*
* Return the lowest commitment level
* obtained by any of the RPCs.
*/
if (committed == NFSWRITE_FILESYNC)
committed = commit;
else if (committed == NFSWRITE_DATASYNC &&
commit == NFSWRITE_UNSTABLE)
committed = commit;
if (commit_thru_mds != 0) {
NFSLOCKMNT(nmp);
if (!NFSHASWRITEVERF(nmp)) {
NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
NFSSETWRITEVERF(nmp);
} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF) &&
*must_commit != 2) {
*must_commit = 1;
NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
}
NFSUNLOCKMNT(nmp);
} else {
NFSLOCKDS(dsp);
if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF) &&
*must_commit != 2) {
*must_commit = 1;
NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
}
NFSUNLOCKDS(dsp);
}
}
nfsmout:
if (nd->nd_mrep != NULL)
m_freem(nd->nd_mrep);
*iomode = committed;
if (nd->nd_repstat != 0 && error == 0)
error = nd->nd_repstat;
return (error);
}
/*
* The actual write RPC done to a DS.
* This variant is called from a separate kernel process for mirrors.
* Any short write is considered an IO error.
*/
static int
nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
struct nfssockreq *nrp;
struct nfsvattr na;
nd->nd_mrep = NULL;
if (vers == 0 || vers == NFS_VER4) {
nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
vers = NFS_VER4;
NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
minorvers);
nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
} else {
nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
}
txdr_hyper(io_off, tl);
tl += 2;
if (vers == NFS_VER3)
*tl++ = txdr_unsigned(len);
*tl++ = txdr_unsigned(*iomode);
*tl = txdr_unsigned(len);
if (len > 0) {
/* Put data in mbuf chain. */
nd->nd_mb->m_next = m;
}
nrp = dsp->nfsclds_sockp;
if (nrp == NULL)
/* If NULL, use the MDS socket. */
nrp = &nmp->nm_sockreq;
error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
nd->nd_repstat);
if (error != 0)
return (error);
if (nd->nd_repstat != 0)
error = nd->nd_repstat;
else {
if (vers == NFS_VER3) {
error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
NULL);
NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
error);
if (error != 0)
goto nfsmout;
}
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
rlen = fxdr_unsigned(int, *tl++);
NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
rlen);
if (rlen != len) {
error = NFSERR_IO;
NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
len, rlen);
goto nfsmout;
}
commit = fxdr_unsigned(int, *tl++);
/*
* Return the lowest commitment level
* obtained by any of the RPCs.
*/
if (committed == NFSWRITE_FILESYNC)
committed = commit;
else if (committed == NFSWRITE_DATASYNC &&
commit == NFSWRITE_UNSTABLE)
committed = commit;
NFSLOCKDS(dsp);
if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF) &&
*must_commit != 2) {
*must_commit = 1;
NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
}
NFSUNLOCKDS(dsp);
}
nfsmout:
if (nd->nd_mrep != NULL)
m_freem(nd->nd_mrep);
*iomode = committed;
if (nd->nd_repstat != 0 && error == 0)
error = nd->nd_repstat;
return (error);
}
/*
* Start up the thread that will execute nfsrpc_writedsmir().
*/
static void
start_writedsmir(void *arg, int pending)
{
struct nfsclwritedsdorpc *drpc;
drpc = (struct nfsclwritedsdorpc *)arg;
drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
&drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
drpc->p);
drpc->done = 1;
NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
}
/*
* Set up the write DS mirror call for the pNFS I/O thread.
*/
static int
nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
{
int error, ret;
error = 0;
drpc->done = 0;
drpc->vp = vp;
drpc->iomode = *iomode;
drpc->must_commit = *must_commit;
drpc->stateidp = stateidp;
drpc->dsp = dsp;
drpc->off = off;
drpc->len = len;
drpc->fhp = fhp;
drpc->m = m;
drpc->vers = vers;
drpc->minorvers = minorvers;
drpc->cred = cred;
drpc->p = p;
drpc->inprog = 0;
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_writedsmir, drpc);
NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
}
if (ret != 0)
error = nfsrpc_writedsmir(vp, iomode, &drpc->must_commit,
stateidp, dsp, off, len, fhp, m, vers, minorvers, cred, p);
NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
return (error);
}
/*
* Free up the nfsclds structure.
*/
void
nfscl_freenfsclds(struct nfsclds *dsp)
{
int i;
if (dsp == NULL)
return;
if (dsp->nfsclds_sockp != NULL) {
NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
}
NFSFREEMUTEX(&dsp->nfsclds_mtx);
NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
for (i = 0; i < NFSV4_CBSLOTS; i++) {
if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
m_freem(
dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
}
free(dsp, M_NFSCLDS);
}
static enum nfsclds_state
nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
struct nfsclds **retdspp, uint32_t *sequencep)
{
struct nfsclds *dsp;
int fndseq;
/*
* Search the list of nfsclds structures for one with the same
* server.
*/
fndseq = 0;
TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
dsp->nfsclds_servownlen != 0 &&
!NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
dsp->nfsclds_servownlen) &&
dsp->nfsclds_sess.nfsess_defunct == 0) {
NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
TAILQ_FIRST(&nmp->nm_sess), dsp,
dsp->nfsclds_flags);
if (fndseq == 0) {
/* Get sequenceid# from first entry. */
*sequencep =
dsp->nfsclds_sess.nfsess_sequenceid;
fndseq = 1;
}
/* Server major id matches. */
if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
*retdspp = dsp;
return (NFSDSP_USETHISSESSION);
}
}
}
if (fndseq != 0)
return (NFSDSP_SEQTHISSESSION);
return (NFSDSP_NOTFOUND);
}
/*
* NFS commit rpc to a NFSv4.1 DS.
*/
static int
nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
NFSPROC_T *p)
{
uint32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
struct nfssockreq *nrp;
struct nfsvattr na;
int attrflag, error;
nd->nd_mrep = NULL;
if (vers == 0 || vers == NFS_VER4) {
nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
vers = NFS_VER4;
} else {
nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMIT]);
NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMITDS]);
}
NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
minorvers);
NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
txdr_hyper(offset, tl);
tl += 2;
*tl = txdr_unsigned(cnt);
nrp = dsp->nfsclds_sockp;
if (nrp == NULL)
/* If NULL, use the MDS socket. */
nrp = &nmp->nm_sockreq;
error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
nd->nd_repstat);
if (error != 0)
return (error);
if (nd->nd_repstat == 0) {
if (vers == NFS_VER3) {
error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
NULL);
NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
if (error != 0)
goto nfsmout;
}
NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
NFSLOCKDS(dsp);
if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
error = NFSERR_STALEWRITEVERF;
}
NFSUNLOCKDS(dsp);
}
nfsmout:
if (error == 0 && nd->nd_repstat != 0)
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* Start up the thread that will execute nfsrpc_commitds().
*/
static void
start_commitds(void *arg, int pending)
{
struct nfsclwritedsdorpc *drpc;
drpc = (struct nfsclwritedsdorpc *)arg;
drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
drpc->p);
drpc->done = 1;
NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
}
/*
* Set up the commit DS mirror call for the pNFS I/O thread.
*/
static int
nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
struct nfsfh *fhp, int vers, int minorvers,
struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
{
int error, ret;
error = 0;
drpc->done = 0;
drpc->vp = vp;
drpc->off = offset;
drpc->len = cnt;
drpc->dsp = dsp;
drpc->fhp = fhp;
drpc->vers = vers;
drpc->minorvers = minorvers;
drpc->cred = cred;
drpc->p = p;
drpc->inprog = 0;
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_commitds, drpc);
NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
}
if (ret != 0)
error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
minorvers, cred, p);
NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
return (error);
}
/*
* NFS Advise rpc
*/
int
nfsrpc_advise(vnode_t vp, off_t offset, uint64_t cnt, int advise,
struct ucred *cred, NFSPROC_T *p)
{
u_int32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
nfsattrbit_t hints;
int error;
NFSZERO_ATTRBIT(&hints);
if (advise == POSIX_FADV_WILLNEED)
NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
else if (advise == POSIX_FADV_DONTNEED)
NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
else
return (0);
NFSCL_REQSTART(nd, NFSPROC_IOADVISE, vp);
nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
txdr_hyper(offset, tl);
tl += 2;
txdr_hyper(cnt, tl);
nfsrv_putattrbit(nd, &hints);
error = nfscl_request(nd, vp, p, cred, NULL);
if (error != 0)
return (error);
if (nd->nd_repstat != 0)
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
#ifdef notyet
/*
* NFS advise rpc to a NFSv4.2 DS.
*/
static int
nfsrpc_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
struct nfssockreq *nrp;
nfsattrbit_t hints;
int error;
/* For NFS DSs prior to NFSv4.2, just return OK. */
if (vers == NFS_VER3 || minorversion < NFSV42_MINORVERSION)
return (0);
NFSZERO_ATTRBIT(&hints);
if (advise == POSIX_FADV_WILLNEED)
NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
else if (advise == POSIX_FADV_DONTNEED)
NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
else
return (0);
nd->nd_mrep = NULL;
nfscl_reqstart(nd, NFSPROC_IOADVISEDS, nmp, fhp->nfh_fh,
fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
vers = NFS_VER4;
NFSCL_DEBUG(4, "nfsrpc_adviseds: vers=%d minvers=%d\n", vers,
minorvers);
nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
txdr_hyper(offset, tl);
tl += 2;
*tl = txdr_unsigned(cnt);
nfsrv_putattrbit(nd, &hints);
nrp = dsp->nfsclds_sockp;
if (nrp == NULL)
/* If NULL, use the MDS socket. */
nrp = &nmp->nm_sockreq;
error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
NFSCL_DEBUG(4, "nfsrpc_adviseds: err=%d stat=%d\n", error,
nd->nd_repstat);
if (error != 0)
return (error);
if (nd->nd_repstat != 0)
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
return (error);
}
/*
* Start up the thread that will execute nfsrpc_commitds().
*/
static void
start_adviseds(void *arg, int pending)
{
struct nfsclwritedsdorpc *drpc;
drpc = (struct nfsclwritedsdorpc *)arg;
drpc->err = nfsrpc_adviseds(drpc->vp, drpc->off, drpc->len,
drpc->advise, drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers,
drpc->cred, drpc->p);
drpc->done = 1;
NFSCL_DEBUG(4, "start_adviseds: err=%d\n", drpc->err);
}
/*
* Set up the commit DS mirror call for the pNFS I/O thread.
*/
static int
nfsio_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
{
int error, ret;
error = 0;
drpc->done = 0;
drpc->vp = vp;
drpc->off = offset;
drpc->len = cnt;
drpc->advise = advise;
drpc->dsp = dsp;
drpc->fhp = fhp;
drpc->vers = vers;
drpc->minorvers = minorvers;
drpc->cred = cred;
drpc->p = p;
drpc->inprog = 0;
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_adviseds, drpc);
NFSCL_DEBUG(4, "nfsio_adviseds: nfs_pnfsio=%d\n", ret);
}
if (ret != 0)
error = nfsrpc_adviseds(vp, offset, cnt, advise, dsp, fhp, vers,
minorvers, cred, p);
NFSCL_DEBUG(4, "nfsio_adviseds: error=%d\n", error);
return (error);
}
#endif /* notyet */
/*
* Do the Allocate operation, retrying for recovery.
*/
int
nfsrpc_allocate(vnode_t vp, off_t off, off_t len, struct nfsvattr *nap,
int *attrflagp, struct ucred *cred, NFSPROC_T *p, void *stuff)
{
int error, expireret = 0, retrycnt, nostateid;
uint32_t clidrev = 0;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
struct nfsfh *nfhp = NULL;
nfsv4stateid_t stateid;
off_t tmp_off;
void *lckp;
if (len < 0)
return (EINVAL);
if (len == 0)
return (0);
tmp_off = off + len;
NFSLOCKMNT(nmp);
if (tmp_off > nmp->nm_maxfilesize || tmp_off < off) {
NFSUNLOCKMNT(nmp);
return (EFBIG);
}
if (nmp->nm_clp != NULL)
clidrev = nmp->nm_clp->nfsc_clientidrev;
NFSUNLOCKMNT(nmp);
nfhp = VTONFS(vp)->n_fhp;
retrycnt = 0;
do {
lckp = NULL;
nostateid = 0;
nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
stateid.other[2] == 0) {
nostateid = 1;
NFSCL_DEBUG(1, "stateid0 in allocate\n");
}
/*
* Not finding a stateid should probably never happen,
* but just return an error for this case.
*/
if (nostateid != 0)
error = EIO;
else
error = nfsrpc_allocaterpc(vp, off, len, &stateid,
nap, attrflagp, cred, p, stuff);
if (error == NFSERR_STALESTATEID)
nfscl_initiate_recovery(nmp->nm_clp);
if (lckp != NULL)
nfscl_lockderef(lckp);
if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
(void) nfs_catnap(PZERO, error, "nfs_allocate");
} else if ((error == NFSERR_EXPIRED ||
error == NFSERR_BADSTATEID) && clidrev != 0) {
expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
}
retrycnt++;
} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
error == NFSERR_STALEDONTRECOVER ||
(error == NFSERR_OLDSTATEID && retrycnt < 20) ||
((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
expireret == 0 && clidrev != 0 && retrycnt < 4));
if (error != 0 && retrycnt >= 4)
error = EIO;
return (error);
}
/*
* The allocate RPC.
*/
static int
nfsrpc_allocaterpc(vnode_t vp, off_t off, off_t len, nfsv4stateid_t *stateidp,
struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p,
void *stuff)
{
uint32_t *tl;
int error;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
nfsattrbit_t attrbits;
*attrflagp = 0;
NFSCL_REQSTART(nd, NFSPROC_ALLOCATE, vp);
nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
txdr_hyper(off, tl); tl += 2;
txdr_hyper(len, tl); tl += 2;
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
nfsrv_putattrbit(nd, &attrbits);
error = nfscl_request(nd, vp, p, cred, stuff);
if (error != 0)
return (error);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
error = nfsm_loadattr(nd, nap);
if (error == 0)
*attrflagp = NFS_LATTR_NOSHRINK;
} else
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* Set up the XDR arguments for the LayoutGet operation.
*/
static void
nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
int layoutlen, int usecurstateid)
{
uint32_t *tl;
NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
NFSX_STATEID);
*tl++ = newnfs_false; /* Don't signal availability. */
*tl++ = txdr_unsigned(layouttype);
*tl++ = txdr_unsigned(iomode);
txdr_hyper(offset, tl);
tl += 2;
txdr_hyper(len, tl);
tl += 2;
txdr_hyper(minlen, tl);
tl += 2;
if (usecurstateid != 0) {
/* Special stateid for Current stateid. */
*tl++ = txdr_unsigned(1);
*tl++ = 0;
*tl++ = 0;
*tl++ = 0;
} else {
*tl++ = txdr_unsigned(stateidp->seqid);
NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
*tl++ = stateidp->other[0];
*tl++ = stateidp->other[1];
*tl++ = stateidp->other[2];
}
*tl = txdr_unsigned(layoutlen);
}
/*
* Parse the reply for a successful LayoutGet operation.
*/
static int
nfsrv_parselayoutget(struct nfsmount *nmp, struct nfsrv_descript *nd,
nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp)
{
uint32_t *tl;
struct nfsclflayout *flp, *prevflp, *tflp;
int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
int m, mirrorcnt;
uint64_t retlen, off;
struct nfsfh *nfhp;
uint8_t *cp;
uid_t user;
gid_t grp;
NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
error = 0;
flp = NULL;
gotiomode = -1;
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
if (*tl++ != 0)
*retonclosep = 1;
else
*retonclosep = 0;
stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
(int)stateidp->seqid);
stateidp->other[0] = *tl++;
stateidp->other[1] = *tl++;
stateidp->other[2] = *tl++;
cnt = fxdr_unsigned(int, *tl);
NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
if (cnt <= 0 || cnt > 10000) {
/* Don't accept more than 10000 layouts in reply. */
error = NFSERR_BADXDR;
goto nfsmout;
}
for (i = 0; i < cnt; i++) {
/* Dissect to the layout type. */
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
3 * NFSX_UNSIGNED);
off = fxdr_hyper(tl); tl += 2;
retlen = fxdr_hyper(tl); tl += 2;
iomode = fxdr_unsigned(int, *tl++);
laytype = fxdr_unsigned(int, *tl);
NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
(uintmax_t)off, (uintmax_t)retlen, iomode);
/* Ignore length of layout body for now. */
if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
/* Parse the File layout up to fhcnt. */
NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
NFSX_HYPER + NFSX_V4DEVICEID);
fhcnt = fxdr_unsigned(int, *(tl + 4 +
NFSX_V4DEVICEID / NFSX_UNSIGNED));
NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
if (fhcnt < 0 || fhcnt > 100) {
/* Don't accept more than 100 file handles. */
error = NFSERR_BADXDR;
goto nfsmout;
}
if (fhcnt > 0)
flp = malloc(sizeof(*flp) + fhcnt *
sizeof(struct nfsfh *), M_NFSFLAYOUT,
M_WAITOK);
else
flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
M_WAITOK);
flp->nfsfl_flags = NFSFL_FILE;
flp->nfsfl_fhcnt = 0;
flp->nfsfl_devp = NULL;
flp->nfsfl_off = off;
if (flp->nfsfl_off + retlen < flp->nfsfl_off)
flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
else
flp->nfsfl_end = flp->nfsfl_off + retlen;
flp->nfsfl_iomode = iomode;
if (gotiomode == -1)
gotiomode = flp->nfsfl_iomode;
/* Ignore layout body length for now. */
NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
mtx_lock(&nmp->nm_mtx);
if (nmp->nm_minorvers > 1 && (flp->nfsfl_util &
NFSFLAYUTIL_IOADVISE_THRU_MDS) != 0)
nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
mtx_unlock(&nmp->nm_mtx);
flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
for (j = 0; j < fhcnt; j++) {
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
nfhlen = fxdr_unsigned(int, *tl);
if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
error = NFSERR_BADXDR;
goto nfsmout;
}
nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
M_NFSFH, M_WAITOK);
flp->nfsfl_fh[j] = nfhp;
flp->nfsfl_fhcnt++;
nfhp->nfh_len = nfhlen;
NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
}
} else if (laytype == NFSLAYOUT_FLEXFILE) {
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
NFSX_HYPER);
mirrorcnt = fxdr_unsigned(int, *(tl + 2));
NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
error = NFSERR_BADXDR;
goto nfsmout;
}
flp = malloc(sizeof(*flp) + mirrorcnt *
sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
flp->nfsfl_flags = NFSFL_FLEXFILE;
flp->nfsfl_mirrorcnt = mirrorcnt;
for (j = 0; j < mirrorcnt; j++)
flp->nfsfl_ffm[j].devp = NULL;
flp->nfsfl_off = off;
if (flp->nfsfl_off + retlen < flp->nfsfl_off)
flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
else
flp->nfsfl_end = flp->nfsfl_off + retlen;
flp->nfsfl_iomode = iomode;
if (gotiomode == -1)
gotiomode = flp->nfsfl_iomode;
flp->nfsfl_stripeunit = fxdr_hyper(tl);
NFSCL_DEBUG(4, "stripeunit=%ju\n",
(uintmax_t)flp->nfsfl_stripeunit);
for (j = 0; j < mirrorcnt; j++) {
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
k = fxdr_unsigned(int, *tl);
if (k < 1 || k > 128) {
error = NFSERR_BADXDR;
goto nfsmout;
}
NFSCL_DEBUG(4, "servercnt=%d\n", k);
for (l = 0; l < k; l++) {
NFSM_DISSECT(tl, uint32_t *,
NFSX_V4DEVICEID + NFSX_STATEID +
2 * NFSX_UNSIGNED);
if (l == 0) {
/* Just use the first server. */
NFSBCOPY(tl,
flp->nfsfl_ffm[j].dev,
NFSX_V4DEVICEID);
tl += (NFSX_V4DEVICEID /
NFSX_UNSIGNED);
tl++;
flp->nfsfl_ffm[j].st.seqid =
*tl++;
flp->nfsfl_ffm[j].st.other[0] =
*tl++;
flp->nfsfl_ffm[j].st.other[1] =
*tl++;
flp->nfsfl_ffm[j].st.other[2] =
*tl++;
NFSCL_DEBUG(4, "st.seqid=%u "
"st.o0=0x%x st.o1=0x%x "
"st.o2=0x%x\n",
flp->nfsfl_ffm[j].st.seqid,
flp->nfsfl_ffm[j].st.other[0],
flp->nfsfl_ffm[j].st.other[1],
flp->nfsfl_ffm[j].st.other[2]);
} else
tl += ((NFSX_V4DEVICEID +
NFSX_STATEID +
NFSX_UNSIGNED) /
NFSX_UNSIGNED);
fhcnt = fxdr_unsigned(int, *tl);
NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
if (fhcnt < 1 ||
fhcnt > NFSDEV_MAXVERS) {
error = NFSERR_BADXDR;
goto nfsmout;
}
for (m = 0; m < fhcnt; m++) {
NFSM_DISSECT(tl, uint32_t *,
NFSX_UNSIGNED);
nfhlen = fxdr_unsigned(int,
*tl);
NFSCL_DEBUG(4, "nfhlen=%d\n",
nfhlen);
if (nfhlen <= 0 || nfhlen >
NFSX_V4FHMAX) {
error = NFSERR_BADXDR;
goto nfsmout;
}
NFSM_DISSECT(cp, uint8_t *,
NFSM_RNDUP(nfhlen));
if (l == 0) {
flp->nfsfl_ffm[j].fhcnt
= fhcnt;
nfhp = malloc(
sizeof(*nfhp) +
nfhlen - 1, M_NFSFH,
M_WAITOK);
flp->nfsfl_ffm[j].fh[m]
= nfhp;
nfhp->nfh_len = nfhlen;
NFSBCOPY(cp,
nfhp->nfh_fh,
nfhlen);
NFSCL_DEBUG(4,
"got fh\n");
}
}
/* Now, get the ffsd_user/ffds_group. */
error = nfsrv_parseug(nd, 0, &user,
&grp, curthread);
NFSCL_DEBUG(4, "after parseu=%d\n",
error);
if (error == 0)
error = nfsrv_parseug(nd, 1,
&user, &grp, curthread);
NFSCL_DEBUG(4, "aft parseg=%d\n",
grp);
if (error != 0)
goto nfsmout;
NFSCL_DEBUG(4, "user=%d group=%d\n",
user, grp);
if (l == 0) {
flp->nfsfl_ffm[j].user = user;
flp->nfsfl_ffm[j].group = grp;
NFSCL_DEBUG(4,
"usr=%d grp=%d\n", user,
grp);
}
}
}
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
#ifdef notnow
/*
* At this time, there is no flag.
* NFSFLEXFLAG_IOADVISE_THRU_MDS might need to be
* added, or it may never exist?
*/
mtx_lock(&nmp->nm_mtx);
if (nmp->nm_minorvers > 1 && (flp->nfsfl_fflags &
NFSFLEXFLAG_IOADVISE_THRU_MDS) != 0)
nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
mtx_unlock(&nmp->nm_mtx);
#endif
flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
flp->nfsfl_fflags, flp->nfsfl_statshint);
} else {
error = NFSERR_BADXDR;
goto nfsmout;
}
if (flp->nfsfl_iomode == gotiomode) {
/* Keep the list in increasing offset order. */
tflp = LIST_FIRST(flhp);
prevflp = NULL;
while (tflp != NULL &&
tflp->nfsfl_off < flp->nfsfl_off) {
prevflp = tflp;
tflp = LIST_NEXT(tflp, nfsfl_list);
}
if (prevflp == NULL)
LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
else
LIST_INSERT_AFTER(prevflp, flp,
nfsfl_list);
NFSCL_DEBUG(4, "flp inserted\n");
} else {
printf("nfscl_layoutget(): got wrong iomode\n");
nfscl_freeflayout(flp);
}
flp = NULL;
}
nfsmout:
NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
if (error != 0 && flp != NULL)
nfscl_freeflayout(flp);
return (error);
}
/*
* Parse a user/group digit string.
*/
static int
nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
NFSPROC_T *p)
{
uint32_t *tl;
char *cp, *str, str0[NFSV4_SMALLSTR + 1];
uint32_t len = 0;
int error = 0;
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
len = fxdr_unsigned(uint32_t, *tl);
str = NULL;
if (len > NFSV4_OPAQUELIMIT) {
error = NFSERR_BADXDR;
goto nfsmout;
}
NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
if (len == 0) {
if (dogrp != 0)
*gidp = GID_NOGROUP;
else
*uidp = UID_NOBODY;
return (0);
}
if (len > NFSV4_SMALLSTR)
str = malloc(len + 1, M_TEMP, M_WAITOK);
else
str = str0;
NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
NFSBCOPY(cp, str, len);
str[len] = '\0';
NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
if (dogrp != 0)
error = nfsv4_strtogid(nd, str, len, gidp);
else
error = nfsv4_strtouid(nd, str, len, uidp);
nfsmout:
if (len > NFSV4_SMALLSTR)
free(str, M_TEMP);
NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
return (error);
}
/*
* Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
* so that it does both an Open and a Layoutget.
*/
static int
nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
struct ucred *cred, NFSPROC_T *p)
{
struct nfscllayout *lyp;
struct nfsclflayout *flp;
struct nfsclflayouthead flh;
int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
int layouttype, laystat;
nfsv4stateid_t stateid;
struct nfsclsession *tsep;
error = 0;
if (NFSHASFLEXFILE(nmp))
layouttype = NFSLAYOUT_FLEXFILE;
else
layouttype = NFSLAYOUT_NFSV4_1_FILES;
/*
* If lyp is returned non-NULL, there will be a refcnt (shared lock)
* on it, iff flp != NULL or a lock (exclusive lock) on it iff
* flp == NULL.
*/
lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, mode, &flp,
&recalled);
NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
if (lyp == NULL)
islocked = 0;
else if (flp != NULL)
islocked = 1;
else
islocked = 2;
if ((lyp == NULL || flp == NULL) && recalled == 0) {
LIST_INIT(&flh);
tsep = nfsmnt_mdssession(nmp);
layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
3 * NFSX_UNSIGNED);
if (lyp == NULL)
usecurstateid = 1;
else {
usecurstateid = 0;
stateid.seqid = lyp->nfsly_stateid.seqid;
stateid.other[0] = lyp->nfsly_stateid.other[0];
stateid.other[1] = lyp->nfsly_stateid.other[1];
stateid.other[2] = lyp->nfsly_stateid.other[2];
}
error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
newfhp, newfhlen, mode, op, name, namelen,
dpp, &stateid, usecurstateid, layouttype, layoutlen,
&retonclose, &flh, &laystat, cred, p);
NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
laystat, error);
laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
&stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
&islocked, cred, p);
} else
error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
if (islocked == 2)
nfscl_rellayout(lyp, 1);
else if (islocked == 1)
nfscl_rellayout(lyp, 0);
return (error);
}
/*
* This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
* enabled, only for the CLAIM_NULL case. All other NFSv4 Opens are
* handled by nfsrpc_openrpc().
* For the case where op == NULL, dvp is the directory. When op != NULL, it
* can be NULL.
*/
static int
nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
int *laystatp, struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfscldeleg *ndp = NULL;
struct nfsvattr nfsva;
struct nfsclsession *tsep;
uint32_t rflags, deleg;
nfsattrbit_t attrbits;
int error, ret, acesize, limitby, iomode;
*dpp = NULL;
*laystatp = ENXIO;
nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
0, 0);
NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
tsep = nfsmnt_mdssession(nmp);
*tl++ = tsep->nfsess_clientid.lval[0];
*tl = tsep->nfsess_clientid.lval[1];
nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
nfsm_strtom(nd, name, namelen);
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
nfsrv_putattrbit(nd, &attrbits);
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
iomode = NFSLAYOUTIOMODE_RW;
else
iomode = NFSLAYOUTIOMODE_READ;
nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
layouttype, layoutlen, usecurstateid);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0)
return (error);
NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
if (nd->nd_repstat != 0)
*laystatp = nd->nd_repstat;
if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
/* ND_NOMOREDATA will be set if the Open operation failed. */
NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
6 * NFSX_UNSIGNED);
op->nfso_stateid.seqid = *tl++;
op->nfso_stateid.other[0] = *tl++;
op->nfso_stateid.other[1] = *tl++;
op->nfso_stateid.other[2] = *tl;
rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
if (error != 0)
goto nfsmout;
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
deleg = fxdr_unsigned(u_int32_t, *tl);
if (deleg == NFSV4OPEN_DELEGATEREAD ||
deleg == NFSV4OPEN_DELEGATEWRITE) {
if (!(op->nfso_own->nfsow_clp->nfsc_flags &
NFSCLFLAGS_FIRSTDELEG))
op->nfso_own->nfsow_clp->nfsc_flags |=
(NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
M_NFSCLDELEG, M_WAITOK);
LIST_INIT(&ndp->nfsdl_owner);
LIST_INIT(&ndp->nfsdl_lock);
ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
ndp->nfsdl_fhlen = newfhlen;
NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
newnfs_copyincred(cred, &ndp->nfsdl_cred);
nfscl_lockinit(&ndp->nfsdl_rwlock);
NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
NFSX_UNSIGNED);
ndp->nfsdl_stateid.seqid = *tl++;
ndp->nfsdl_stateid.other[0] = *tl++;
ndp->nfsdl_stateid.other[1] = *tl++;
ndp->nfsdl_stateid.other[2] = *tl++;
ret = fxdr_unsigned(int, *tl);
if (deleg == NFSV4OPEN_DELEGATEWRITE) {
ndp->nfsdl_flags = NFSCLDL_WRITE;
/*
* Indicates how much the file can grow.
*/
NFSM_DISSECT(tl, u_int32_t *,
3 * NFSX_UNSIGNED);
limitby = fxdr_unsigned(int, *tl++);
switch (limitby) {
case NFSV4OPEN_LIMITSIZE:
ndp->nfsdl_sizelimit = fxdr_hyper(tl);
break;
case NFSV4OPEN_LIMITBLOCKS:
ndp->nfsdl_sizelimit =
fxdr_unsigned(u_int64_t, *tl++);
ndp->nfsdl_sizelimit *=
fxdr_unsigned(u_int64_t, *tl);
break;
default:
error = NFSERR_BADXDR;
goto nfsmout;
};
} else
ndp->nfsdl_flags = NFSCLDL_READ;
if (ret != 0)
ndp->nfsdl_flags |= NFSCLDL_RECALL;
- error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
- &acesize, p);
+ error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, false,
+ &ret, &acesize, p);
if (error != 0)
goto nfsmout;
} else if (deleg != NFSV4OPEN_DELEGATENONE) {
error = NFSERR_BADXDR;
goto nfsmout;
}
if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
nfscl_assumeposixlocks)
op->nfso_posixlock = 1;
else
op->nfso_posixlock = 0;
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
/* If the 2nd element == NFS_OK, the Getattr succeeded. */
if (*++tl == 0) {
error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
NULL, NULL, NULL, p, cred);
if (error != 0)
goto nfsmout;
if (ndp != NULL) {
ndp->nfsdl_change = nfsva.na_filerev;
ndp->nfsdl_modtime = nfsva.na_mtime;
ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
*dpp = ndp;
ndp = NULL;
}
/*
* At this point, the Open has succeeded, so set
* nd_repstat = NFS_OK. If the Layoutget failed,
* this function just won't return a layout.
*/
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
*laystatp = fxdr_unsigned(int, *++tl);
if (*laystatp == 0) {
error = nfsrv_parselayoutget(nmp, nd,
stateidp, retonclosep, flhp);
if (error != 0)
*laystatp = error;
}
} else
nd->nd_repstat = 0; /* Return 0 for Open. */
}
}
if (nd->nd_repstat != 0 && error == 0)
error = nd->nd_repstat;
nfsmout:
free(ndp, M_NFSCLDELEG);
m_freem(nd->nd_mrep);
return (error);
}
/*
* Similar nfsrpc_createv4(), but also does the LayoutGet operation.
* Used only for mounts with pNFS enabled.
*/
static int
nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
int *dattrflagp, void *dstuff, int *unlockedp, nfsv4stateid_t *stateidp,
int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
struct nfsclflayouthead *flhp, int *laystatp)
{
uint32_t *tl;
int error = 0, deleg, newone, ret, acesize, limitby;
struct nfsrv_descript nfsd, *nd = &nfsd;
struct nfsclopen *op;
struct nfscldeleg *dp = NULL;
struct nfsnode *np;
struct nfsfh *nfhp;
struct nfsclsession *tsep;
nfsattrbit_t attrbits;
nfsv4stateid_t stateid;
struct nfsmount *nmp;
nmp = VFSTONFS(dvp->v_mount);
np = VTONFS(dvp);
*laystatp = ENXIO;
*unlockedp = 0;
*nfhpp = NULL;
*dpp = NULL;
*attrflagp = 0;
*dattrflagp = 0;
if (namelen > NFS_MAXNAMLEN)
return (ENAMETOOLONG);
NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp);
/*
* For V4, this is actually an Open op.
*/
NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(owp->nfsow_seqid);
*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
NFSV4OPEN_ACCESSREAD);
*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
tsep = nfsmnt_mdssession(nmp);
*tl++ = tsep->nfsess_clientid.lval[0];
*tl = tsep->nfsess_clientid.lval[1];
nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
if ((fmode & O_EXCL) != 0) {
if (NFSHASSESSPERSIST(nmp)) {
/* Use GUARDED for persistent sessions. */
*tl = txdr_unsigned(NFSCREATE_GUARDED);
nfscl_fillsattr(nd, vap, dvp, 0, 0);
} else {
/* Otherwise, use EXCLUSIVE4_1. */
*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
*tl++ = cverf.lval[0];
*tl = cverf.lval[1];
nfscl_fillsattr(nd, vap, dvp, 0, 0);
}
} else {
*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
nfscl_fillsattr(nd, vap, dvp, 0, 0);
}
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
nfsm_strtom(nd, name, namelen);
/* Get the new file's handle and attributes, plus save the FH. */
NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
*tl++ = txdr_unsigned(NFSV4OP_GETFH);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
nfsrv_putattrbit(nd, &attrbits);
/* Get the directory's post-op attributes. */
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_PUTFH);
nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
nfsrv_putattrbit(nd, &attrbits);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
layouttype, layoutlen, usecurstateid);
error = nfscl_request(nd, dvp, p, cred, dstuff);
if (error != 0)
return (error);
NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
error);
if (nd->nd_repstat != 0)
*laystatp = nd->nd_repstat;
NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
6 * NFSX_UNSIGNED);
stateid.seqid = *tl++;
stateid.other[0] = *tl++;
stateid.other[1] = *tl++;
stateid.other[2] = *tl;
error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
if (error != 0)
goto nfsmout;
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
deleg = fxdr_unsigned(int, *tl);
if (deleg == NFSV4OPEN_DELEGATEREAD ||
deleg == NFSV4OPEN_DELEGATEWRITE) {
if (!(owp->nfsow_clp->nfsc_flags &
NFSCLFLAGS_FIRSTDELEG))
owp->nfsow_clp->nfsc_flags |=
(NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
M_NFSCLDELEG, M_WAITOK);
LIST_INIT(&dp->nfsdl_owner);
LIST_INIT(&dp->nfsdl_lock);
dp->nfsdl_clp = owp->nfsow_clp;
newnfs_copyincred(cred, &dp->nfsdl_cred);
nfscl_lockinit(&dp->nfsdl_rwlock);
NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
NFSX_UNSIGNED);
dp->nfsdl_stateid.seqid = *tl++;
dp->nfsdl_stateid.other[0] = *tl++;
dp->nfsdl_stateid.other[1] = *tl++;
dp->nfsdl_stateid.other[2] = *tl++;
ret = fxdr_unsigned(int, *tl);
if (deleg == NFSV4OPEN_DELEGATEWRITE) {
dp->nfsdl_flags = NFSCLDL_WRITE;
/*
* Indicates how much the file can grow.
*/
NFSM_DISSECT(tl, u_int32_t *,
3 * NFSX_UNSIGNED);
limitby = fxdr_unsigned(int, *tl++);
switch (limitby) {
case NFSV4OPEN_LIMITSIZE:
dp->nfsdl_sizelimit = fxdr_hyper(tl);
break;
case NFSV4OPEN_LIMITBLOCKS:
dp->nfsdl_sizelimit =
fxdr_unsigned(u_int64_t, *tl++);
dp->nfsdl_sizelimit *=
fxdr_unsigned(u_int64_t, *tl);
break;
default:
error = NFSERR_BADXDR;
goto nfsmout;
};
} else {
dp->nfsdl_flags = NFSCLDL_READ;
}
if (ret != 0)
dp->nfsdl_flags |= NFSCLDL_RECALL;
- error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
- &acesize, p);
+ error = nfsrv_dissectace(nd, &dp->nfsdl_ace, false,
+ &ret, &acesize, p);
if (error != 0)
goto nfsmout;
} else if (deleg != NFSV4OPEN_DELEGATENONE) {
error = NFSERR_BADXDR;
goto nfsmout;
}
/* Now, we should have the status for the SaveFH. */
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (*++tl == 0) {
NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
/*
* Now, process the GetFH and Getattr for the newly
* created file. nfscl_mtofh() will set
* ND_NOMOREDATA if these weren't successful.
*/
error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
if (error != 0)
goto nfsmout;
} else
nd->nd_flag |= ND_NOMOREDATA;
/* Now we have the PutFH and Getattr for the directory. */
if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (*++tl != 0)
nd->nd_flag |= ND_NOMOREDATA;
else {
NFSM_DISSECT(tl, uint32_t *, 2 *
NFSX_UNSIGNED);
if (*++tl != 0)
nd->nd_flag |= ND_NOMOREDATA;
}
}
if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
/* Load the directory attributes. */
error = nfsm_loadattr(nd, dnap);
NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
if (error != 0)
goto nfsmout;
*dattrflagp = 1;
if (dp != NULL && *attrflagp != 0) {
dp->nfsdl_change = nnap->na_filerev;
dp->nfsdl_modtime = nnap->na_mtime;
dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
}
/*
* We can now complete the Open state.
*/
nfhp = *nfhpp;
if (dp != NULL) {
dp->nfsdl_fhlen = nfhp->nfh_len;
NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
nfhp->nfh_len);
}
/*
* Get an Open structure that will be
* attached to the OpenOwner, acquired already.
*/
error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
(NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
cred, p, NULL, &op, &newone, NULL, 0, false);
if (error != 0)
goto nfsmout;
op->nfso_stateid = stateid;
newnfs_copyincred(cred, &op->nfso_cred);
nfscl_openrelease(nmp, op, error, newone);
*unlockedp = 1;
/* Now, handle the RestoreFH and LayoutGet. */
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
*laystatp = fxdr_unsigned(int, *(tl + 3));
if (*laystatp == 0) {
error = nfsrv_parselayoutget(nmp, nd,
stateidp, retonclosep, flhp);
if (error != 0)
*laystatp = error;
}
NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
error);
} else
nd->nd_repstat = 0;
}
}
if (nd->nd_repstat != 0 && error == 0)
error = nd->nd_repstat;
if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
nfscl_initiate_recovery(owp->nfsow_clp);
nfsmout:
NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
if (error == 0)
*dpp = dp;
else
free(dp, M_NFSCLDELEG);
m_freem(nd->nd_mrep);
return (error);
}
/*
* Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
*/
static int
nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
int *dattrflagp, void *dstuff, int *unlockedp)
{
struct nfscllayout *lyp;
struct nfsclflayouthead flh;
struct nfsfh *nfhp;
struct nfsclsession *tsep;
struct nfsmount *nmp;
nfsv4stateid_t stateid;
int error, layoutlen, layouttype, retonclose, laystat;
error = 0;
nmp = VFSTONFS(dvp->v_mount);
if (NFSHASFLEXFILE(nmp))
layouttype = NFSLAYOUT_FLEXFILE;
else
layouttype = NFSLAYOUT_NFSV4_1_FILES;
LIST_INIT(&flh);
tsep = nfsmnt_mdssession(nmp);
layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
dstuff, unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
&flh, &laystat);
NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
laystat, error);
lyp = NULL;
if (laystat == 0) {
nfhp = *nfhpp;
laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
layouttype, laystat, NULL, cred, p);
} else
laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
cred, p);
if (laystat == 0)
nfscl_rellayout(lyp, 0);
return (error);
}
/*
* Process the results of a layoutget() operation.
*/
static int
nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
{
struct nfsclflayout *tflp;
struct nfscldevinfo *dip;
uint8_t *dev;
int i, mirrorcnt;
if (laystat == NFSERR_UNKNLAYOUTTYPE) {
NFSLOCKMNT(nmp);
if (!NFSHASFLEXFILE(nmp)) {
/* Switch to using Flex File Layout. */
nmp->nm_state |= NFSSTA_FLEXFILE;
} else if (layouttype == NFSLAYOUT_FLEXFILE) {
/* Disable pNFS. */
NFSCL_DEBUG(1, "disable PNFS\n");
nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
}
NFSUNLOCKMNT(nmp);
}
if (laystat == 0) {
NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
LIST_FOREACH(tflp, flhp, nfsfl_list) {
if (layouttype == NFSLAYOUT_FLEXFILE)
mirrorcnt = tflp->nfsfl_mirrorcnt;
else
mirrorcnt = 1;
for (i = 0; i < mirrorcnt; i++) {
laystat = nfscl_adddevinfo(nmp, NULL, i, tflp);
NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
if (laystat != 0) {
if (layouttype == NFSLAYOUT_FLEXFILE)
dev = tflp->nfsfl_ffm[i].dev;
else
dev = tflp->nfsfl_dev;
laystat = nfsrpc_getdeviceinfo(nmp, dev,
layouttype, notifybit, &dip, cred,
p);
NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
laystat);
if (laystat != 0)
goto out;
laystat = nfscl_adddevinfo(nmp, dip, i,
tflp);
if (laystat != 0)
printf("nfsrpc_layoutgetresout"
": cannot add\n");
}
}
}
}
out:
if (laystat == 0) {
/*
* nfscl_layout() always returns with the nfsly_lock
* set to a refcnt (shared lock).
* Passing in dvp is sufficient, since it is only used to
* get the fsid for the file system.
*/
laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
layouttype, retonclose, flhp, lypp, cred, p);
NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
laystat);
if (laystat == 0 && islockedp != NULL)
*islockedp = 1;
}
return (laystat);
}
/*
* nfs copy_file_range operation.
*/
int
nfsrpc_copy_file_range(vnode_t invp, off_t *inoffp, vnode_t outvp,
off_t *outoffp, size_t *lenp, unsigned int flags, int *inattrflagp,
struct nfsvattr *innap, int *outattrflagp, struct nfsvattr *outnap,
struct ucred *cred, bool consecutive, bool *must_commitp)
{
int commit, error, expireret = 0, retrycnt;
u_int32_t clidrev = 0;
struct nfsmount *nmp = VFSTONFS(invp->v_mount);
struct nfsfh *innfhp = NULL, *outnfhp = NULL;
nfsv4stateid_t instateid, outstateid;
void *inlckp, *outlckp;
if (nmp->nm_clp != NULL)
clidrev = nmp->nm_clp->nfsc_clientidrev;
innfhp = VTONFS(invp)->n_fhp;
outnfhp = VTONFS(outvp)->n_fhp;
retrycnt = 0;
do {
/* Get both stateids. */
inlckp = NULL;
nfscl_getstateid(invp, innfhp->nfh_fh, innfhp->nfh_len,
NFSV4OPEN_ACCESSREAD, 0, NULL, curthread, &instateid,
&inlckp);
outlckp = NULL;
nfscl_getstateid(outvp, outnfhp->nfh_fh, outnfhp->nfh_len,
NFSV4OPEN_ACCESSWRITE, 0, NULL, curthread, &outstateid,
&outlckp);
error = nfsrpc_copyrpc(invp, *inoffp, outvp, *outoffp, lenp,
&instateid, &outstateid, innap, inattrflagp, outnap,
outattrflagp, consecutive, &commit, cred, curthread);
if (error == 0) {
if (commit != NFSWRITE_FILESYNC)
*must_commitp = true;
*inoffp += *lenp;
*outoffp += *lenp;
} else if (error == NFSERR_STALESTATEID)
nfscl_initiate_recovery(nmp->nm_clp);
if (inlckp != NULL)
nfscl_lockderef(inlckp);
if (outlckp != NULL)
nfscl_lockderef(outlckp);
if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
(void) nfs_catnap(PZERO, error, "nfs_cfr");
} else if ((error == NFSERR_EXPIRED ||
error == NFSERR_BADSTATEID) && clidrev != 0) {
expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
curthread);
}
retrycnt++;
} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
error == NFSERR_STALEDONTRECOVER ||
(error == NFSERR_OLDSTATEID && retrycnt < 20) ||
((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
expireret == 0 && clidrev != 0 && retrycnt < 4));
if (error != 0 && (retrycnt >= 4 ||
error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
error == NFSERR_STALEDONTRECOVER))
error = EIO;
return (error);
}
/*
* The copy RPC.
*/
static int
nfsrpc_copyrpc(vnode_t invp, off_t inoff, vnode_t outvp, off_t outoff,
size_t *lenp, nfsv4stateid_t *instateidp, nfsv4stateid_t *outstateidp,
struct nfsvattr *innap, int *inattrflagp, struct nfsvattr *outnap,
int *outattrflagp, bool consecutive, int *commitp, struct ucred *cred,
NFSPROC_T *p)
{
uint32_t *tl;
int error;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
struct nfsmount *nmp;
nfsattrbit_t attrbits;
uint64_t len;
nmp = VFSTONFS(outvp->v_mount);
*inattrflagp = *outattrflagp = 0;
*commitp = NFSWRITE_UNSTABLE;
len = *lenp;
*lenp = 0;
if (len > nfs_maxcopyrange)
len = nfs_maxcopyrange;
NFSCL_REQSTART(nd, NFSPROC_COPY, invp);
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
nfsrv_putattrbit(nd, &attrbits);
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_PUTFH);
nfsm_fhtom(nd, VTONFS(outvp)->n_fhp->nfh_fh,
VTONFS(outvp)->n_fhp->nfh_len, 0);
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_COPY);
nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID);
nfsm_stateidtom(nd, outstateidp, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, uint32_t *, 3 * NFSX_HYPER + 4 * NFSX_UNSIGNED);
txdr_hyper(inoff, tl); tl += 2;
txdr_hyper(outoff, tl); tl += 2;
txdr_hyper(len, tl); tl += 2;
if (consecutive)
*tl++ = newnfs_true;
else
*tl++ = newnfs_false;
*tl++ = newnfs_true;
*tl++ = 0;
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSWRITEGETATTR_ATTRBIT(&attrbits);
nfsrv_putattrbit(nd, &attrbits);
error = nfscl_request(nd, invp, p, cred, NULL);
if (error != 0)
return (error);
if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
/* Get the input file's attributes. */
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (*(tl + 1) == 0) {
error = nfsm_loadattr(nd, innap);
if (error != 0)
goto nfsmout;
*inattrflagp = 1;
} else
nd->nd_flag |= ND_NOMOREDATA;
}
/* Skip over return stat for PutFH. */
if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (*++tl != 0)
nd->nd_flag |= ND_NOMOREDATA;
}
/* Skip over return stat for Copy. */
if ((nd->nd_flag & ND_NOMOREDATA) == 0)
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
if (*tl != 0) {
/* There should be no callback ids. */
error = NFSERR_BADXDR;
goto nfsmout;
}
NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED +
NFSX_VERF);
len = fxdr_hyper(tl); tl += 2;
*commitp = fxdr_unsigned(int, *tl++);
NFSLOCKMNT(nmp);
if (!NFSHASWRITEVERF(nmp)) {
NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
NFSSETWRITEVERF(nmp);
} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
nd->nd_repstat = NFSERR_STALEWRITEVERF;
}
NFSUNLOCKMNT(nmp);
tl += (NFSX_VERF / NFSX_UNSIGNED);
if (nd->nd_repstat == 0 && *++tl != newnfs_true)
/* Must be a synchronous copy. */
nd->nd_repstat = NFSERR_NOTSUPP;
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
error = nfsm_loadattr(nd, outnap);
if (error == 0)
*outattrflagp = NFS_LATTR_NOSHRINK;
if (nd->nd_repstat == 0)
*lenp = len;
} else if (nd->nd_repstat == NFSERR_OFFLOADNOREQS) {
/*
* For the case where consecutive is not supported, but
* synchronous is supported, we can try consecutive == false
* by returning this error. Otherwise, return NFSERR_NOTSUPP,
* since Copy cannot be done.
*/
if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (!consecutive || *++tl == newnfs_false)
nd->nd_repstat = NFSERR_NOTSUPP;
} else
nd->nd_repstat = NFSERR_BADXDR;
}
if (error == 0)
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* Seek operation.
*/
int
nfsrpc_seek(vnode_t vp, off_t *offp, bool *eofp, int content,
struct ucred *cred, struct nfsvattr *nap, int *attrflagp)
{
int error, expireret = 0, retrycnt;
u_int32_t clidrev = 0;
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
struct nfsnode *np = VTONFS(vp);
struct nfsfh *nfhp = NULL;
nfsv4stateid_t stateid;
void *lckp;
if (nmp->nm_clp != NULL)
clidrev = nmp->nm_clp->nfsc_clientidrev;
nfhp = np->n_fhp;
retrycnt = 0;
do {
lckp = NULL;
nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
NFSV4OPEN_ACCESSREAD, 0, cred, curthread, &stateid, &lckp);
error = nfsrpc_seekrpc(vp, offp, &stateid, eofp, content,
nap, attrflagp, cred);
if (error == NFSERR_STALESTATEID)
nfscl_initiate_recovery(nmp->nm_clp);
if (lckp != NULL)
nfscl_lockderef(lckp);
if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
(void) nfs_catnap(PZERO, error, "nfs_seek");
} else if ((error == NFSERR_EXPIRED ||
error == NFSERR_BADSTATEID) && clidrev != 0) {
expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
curthread);
}
retrycnt++;
} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
error == NFSERR_BADSESSION ||
(error == NFSERR_OLDSTATEID && retrycnt < 20) ||
((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
expireret == 0 && clidrev != 0 && retrycnt < 4) ||
(error == NFSERR_OPENMODE && retrycnt < 4));
if (error && retrycnt >= 4)
error = EIO;
return (error);
}
/*
* The seek RPC.
*/
static int
nfsrpc_seekrpc(vnode_t vp, off_t *offp, nfsv4stateid_t *stateidp, bool *eofp,
int content, struct nfsvattr *nap, int *attrflagp, struct ucred *cred)
{
uint32_t *tl;
int error;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
nfsattrbit_t attrbits;
*attrflagp = 0;
NFSCL_REQSTART(nd, NFSPROC_SEEK, vp);
nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
txdr_hyper(*offp, tl); tl += 2;
*tl++ = txdr_unsigned(content);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
nfsrv_putattrbit(nd, &attrbits);
error = nfscl_request(nd, vp, curthread, cred, NULL);
if (error != 0)
return (error);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED + NFSX_HYPER);
if (*tl++ == newnfs_true)
*eofp = true;
else
*eofp = false;
*offp = fxdr_hyper(tl);
/* Just skip over Getattr op status. */
error = nfsm_loadattr(nd, nap);
if (error == 0)
*attrflagp = 1;
}
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* The getextattr RPC.
*/
int
nfsrpc_getextattr(vnode_t vp, const char *name, struct uio *uiop, ssize_t *lenp,
struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
int error;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
nfsattrbit_t attrbits;
uint32_t len, len2;
*attrflagp = 0;
NFSCL_REQSTART(nd, NFSPROC_GETEXTATTR, vp);
nfsm_strtom(nd, name, strlen(name));
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
nfsrv_putattrbit(nd, &attrbits);
error = nfscl_request(nd, vp, p, cred, NULL);
if (error != 0)
return (error);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
len = fxdr_unsigned(uint32_t, *tl);
/* Sanity check lengths. */
if (uiop != NULL && len > 0 && len <= IOSIZE_MAX &&
uiop->uio_resid <= UINT32_MAX) {
len2 = uiop->uio_resid;
if (len2 >= len)
error = nfsm_mbufuio(nd, uiop, len);
else {
error = nfsm_mbufuio(nd, uiop, len2);
if (error == 0) {
/*
* nfsm_mbufuio() advances to a multiple
* of 4, so round up len2 as well. Then
* we need to advance over the rest of
* the data, rounding up the remaining
* length.
*/
len2 = NFSM_RNDUP(len2);
len2 = NFSM_RNDUP(len - len2);
if (len2 > 0)
error = nfsm_advance(nd, len2,
-1);
}
}
} else if (uiop == NULL && len > 0) {
/* Just wants the length and not the data. */
error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
} else if (len > 0)
error = ENOATTR;
if (error != 0)
goto nfsmout;
*lenp = len;
/* Just skip over Getattr op status. */
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
error = nfsm_loadattr(nd, nap);
if (error == 0)
*attrflagp = 1;
}
if (error == 0)
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* The setextattr RPC.
*/
int
nfsrpc_setextattr(vnode_t vp, const char *name, struct uio *uiop,
struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
int error;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
nfsattrbit_t attrbits;
*attrflagp = 0;
NFSCL_REQSTART(nd, NFSPROC_SETEXTATTR, vp);
if (uiop->uio_resid > nd->nd_maxreq) {
/* nd_maxreq is set by NFSCL_REQSTART(). */
m_freem(nd->nd_mreq);
return (EINVAL);
}
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4SXATTR_EITHER);
nfsm_strtom(nd, name, strlen(name));
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(uiop->uio_resid);
nfsm_uiombuf(nd, uiop, uiop->uio_resid);
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
nfsrv_putattrbit(nd, &attrbits);
error = nfscl_request(nd, vp, p, cred, NULL);
if (error != 0)
return (error);
if (nd->nd_repstat == 0) {
/* Just skip over the reply and Getattr op status. */
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
NFSX_UNSIGNED);
error = nfsm_loadattr(nd, nap);
if (error == 0)
*attrflagp = 1;
}
if (error == 0)
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* The removeextattr RPC.
*/
int
nfsrpc_rmextattr(vnode_t vp, const char *name, struct nfsvattr *nap,
int *attrflagp, struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
int error;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
nfsattrbit_t attrbits;
*attrflagp = 0;
NFSCL_REQSTART(nd, NFSPROC_RMEXTATTR, vp);
nfsm_strtom(nd, name, strlen(name));
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
nfsrv_putattrbit(nd, &attrbits);
error = nfscl_request(nd, vp, p, cred, NULL);
if (error != 0)
return (error);
if (nd->nd_repstat == 0) {
/* Just skip over the reply and Getattr op status. */
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
NFSX_UNSIGNED);
error = nfsm_loadattr(nd, nap);
if (error == 0)
*attrflagp = 1;
}
if (error == 0)
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* The listextattr RPC.
*/
int
nfsrpc_listextattr(vnode_t vp, uint64_t *cookiep, struct uio *uiop,
size_t *lenp, bool *eofp, struct nfsvattr *nap, int *attrflagp,
struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
int cnt, error, i, len;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
nfsattrbit_t attrbits;
u_char c;
*attrflagp = 0;
NFSCL_REQSTART(nd, NFSPROC_LISTEXTATTR, vp);
NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
txdr_hyper(*cookiep, tl); tl += 2;
*tl++ = txdr_unsigned(*lenp);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
nfsrv_putattrbit(nd, &attrbits);
error = nfscl_request(nd, vp, p, cred, NULL);
if (error != 0)
return (error);
*eofp = true;
*lenp = 0;
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
*cookiep = fxdr_hyper(tl); tl += 2;
cnt = fxdr_unsigned(int, *tl);
if (cnt < 0) {
error = EBADRPC;
goto nfsmout;
}
for (i = 0; i < cnt; i++) {
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
len = fxdr_unsigned(int, *tl);
if (len <= 0 || len > EXTATTR_MAXNAMELEN) {
error = EBADRPC;
goto nfsmout;
}
if (uiop == NULL)
error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
else if (uiop->uio_resid >= len + 1) {
c = len;
error = uiomove(&c, sizeof(c), uiop);
if (error == 0)
error = nfsm_mbufuio(nd, uiop, len);
} else {
error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
*eofp = false;
}
if (error != 0)
goto nfsmout;
*lenp += (len + 1);
}
/* Get the eof and skip over the Getattr op status. */
NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
/*
* *eofp is set false above, because it wasn't able to copy
* all of the reply.
*/
if (*eofp && *tl == 0)
*eofp = false;
error = nfsm_loadattr(nd, nap);
if (error == 0)
*attrflagp = 1;
}
if (error == 0)
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
return (error);
}
/*
* Split an mbuf list. For non-M_EXTPG mbufs, just use m_split().
*/
static struct mbuf *
nfsm_split(struct mbuf *mp, uint64_t xfer)
{
struct mbuf *m, *m2;
vm_page_t pg;
int i, j, left, pgno, plen, trim;
char *cp, *cp2;
if ((mp->m_flags & M_EXTPG) == 0) {
m = m_split(mp, xfer, M_WAITOK);
return (m);
}
/* Find the correct mbuf to split at. */
for (m = mp; m != NULL && xfer > m->m_len; m = m->m_next)
xfer -= m->m_len;
if (m == NULL)
return (NULL);
/* If xfer == m->m_len, we can just split the mbuf list. */
if (xfer == m->m_len) {
m2 = m->m_next;
m->m_next = NULL;
return (m2);
}
/* Find the page to split at. */
pgno = 0;
left = xfer;
do {
if (pgno == 0)
plen = m_epg_pagelen(m, 0, m->m_epg_1st_off);
else
plen = m_epg_pagelen(m, pgno, 0);
if (left <= plen)
break;
left -= plen;
pgno++;
} while (pgno < m->m_epg_npgs);
if (pgno == m->m_epg_npgs)
panic("nfsm_split: eroneous ext_pgs mbuf");
m2 = mb_alloc_ext_pgs(M_WAITOK, mb_free_mext_pgs);
m2->m_epg_flags |= EPG_FLAG_ANON;
/*
* If left < plen, allocate a new page for the new mbuf
* and copy the data after left in the page to this new
* page.
*/
if (left < plen) {
pg = vm_page_alloc_noobj(VM_ALLOC_WAITOK | VM_ALLOC_NODUMP |
VM_ALLOC_WIRED);
m2->m_epg_pa[0] = VM_PAGE_TO_PHYS(pg);
m2->m_epg_npgs = 1;
/* Copy the data after left to the new page. */
trim = plen - left;
cp = (char *)(void *)PHYS_TO_DMAP(m->m_epg_pa[pgno]);
if (pgno == 0)
cp += m->m_epg_1st_off;
cp += left;
cp2 = (char *)(void *)PHYS_TO_DMAP(m2->m_epg_pa[0]);
if (pgno == m->m_epg_npgs - 1)
m2->m_epg_last_len = trim;
else {
cp2 += PAGE_SIZE - trim;
m2->m_epg_1st_off = PAGE_SIZE - trim;
m2->m_epg_last_len = m->m_epg_last_len;
}
memcpy(cp2, cp, trim);
m2->m_len = trim;
} else {
m2->m_len = 0;
m2->m_epg_last_len = m->m_epg_last_len;
}
/* Move the pages beyond pgno to the new mbuf. */
for (i = pgno + 1, j = m2->m_epg_npgs; i < m->m_epg_npgs; i++, j++) {
m2->m_epg_pa[j] = m->m_epg_pa[i];
/* Never moves page 0. */
m2->m_len += m_epg_pagelen(m, i, 0);
}
m2->m_epg_npgs = j;
m->m_epg_npgs = pgno + 1;
m->m_epg_last_len = left;
m->m_len = xfer;
m2->m_next = m->m_next;
m->m_next = NULL;
return (m2);
}
/*
* Do the NFSv4.1 Bind Connection to Session.
* Called from the reconnect layer of the krpc (sys/rpc/clnt_rc.c).
*/
void
nfsrpc_bindconnsess(CLIENT *cl, void *arg, struct ucred *cr)
{
struct nfscl_reconarg *rcp = (struct nfscl_reconarg *)arg;
uint32_t res, *tl;
struct nfsrv_descript nfsd;
struct nfsrv_descript *nd = &nfsd;
struct rpc_callextra ext;
struct timeval utimeout;
enum clnt_stat stat;
int error;
nfscl_reqstart(nd, NFSPROC_BINDCONNTOSESS, NULL, NULL, 0, NULL, NULL,
NFS_VER4, rcp->minorvers);
NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID + 2 * NFSX_UNSIGNED);
memcpy(tl, rcp->sessionid, NFSX_V4SESSIONID);
tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
*tl++ = txdr_unsigned(NFSCDFC4_FORE_OR_BOTH);
*tl = newnfs_false;
memset(&ext, 0, sizeof(ext));
utimeout.tv_sec = 30;
utimeout.tv_usec = 0;
ext.rc_auth = authunix_create(cr);
nd->nd_mrep = NULL;
stat = CLNT_CALL_MBUF(cl, &ext, NFSV4PROC_COMPOUND, nd->nd_mreq,
&nd->nd_mrep, utimeout);
AUTH_DESTROY(ext.rc_auth);
if (stat != RPC_SUCCESS) {
printf("nfsrpc_bindconnsess: call failed stat=%d\n", stat);
return;
}
if (nd->nd_mrep == NULL) {
printf("nfsrpc_bindconnsess: no reply args\n");
return;
}
error = 0;
newnfs_realign(&nd->nd_mrep, M_WAITOK);
nd->nd_md = nd->nd_mrep;
nd->nd_dpos = mtod(nd->nd_md, char *);
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
nd->nd_repstat = fxdr_unsigned(uint32_t, *tl++);
if (nd->nd_repstat == NFSERR_OK) {
res = fxdr_unsigned(uint32_t, *tl);
if (res > 0 && (error = nfsm_advance(nd, NFSM_RNDUP(res),
-1)) != 0)
goto nfsmout;
NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
4 * NFSX_UNSIGNED);
tl += 3;
if (!NFSBCMP(tl, rcp->sessionid, NFSX_V4SESSIONID)) {
tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
res = fxdr_unsigned(uint32_t, *tl);
if (res != NFSCDFS4_BOTH)
printf("nfsrpc_bindconnsess: did not "
"return FS4_BOTH\n");
} else
printf("nfsrpc_bindconnsess: not same "
"sessionid\n");
} else if (nd->nd_repstat != NFSERR_BADSESSION)
printf("nfsrpc_bindconnsess: returned %d\n", nd->nd_repstat);
nfsmout:
if (error != 0)
printf("nfsrpc_bindconnsess: reply bad xdr\n");
m_freem(nd->nd_mrep);
}
diff --git a/sys/fs/nfsserver/nfs_nfsdport.c b/sys/fs/nfsserver/nfs_nfsdport.c
index 1a72251530dd..8afcc9400f95 100644
--- a/sys/fs/nfsserver/nfs_nfsdport.c
+++ b/sys/fs/nfsserver/nfs_nfsdport.c
@@ -1,7128 +1,7128 @@
/*-
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 1989, 1993
* The Regents of the University of California. All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* Rick Macklem at The University of Guelph.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/capsicum.h>
#include <sys/extattr.h>
/*
* Functions that perform the vfs operations required by the routines in
* nfsd_serv.c. It is hoped that this change will make the server more
* portable.
*/
#include <fs/nfs/nfsport.h>
#include <security/mac/mac_framework.h>
#include <sys/callout.h>
#include <sys/filio.h>
#include <sys/hash.h>
#include <sys/sysctl.h>
#include <nlm/nlm_prot.h>
#include <nlm/nlm.h>
FEATURE(nfsd, "NFSv4 server");
extern u_int32_t newnfs_true, newnfs_false, newnfs_xdrneg1;
extern int nfsrv_useacl;
extern int newnfs_numnfsd;
extern struct mount nfsv4root_mnt;
extern struct nfsrv_stablefirst nfsrv_stablefirst;
extern SVCPOOL *nfsrvd_pool;
extern struct nfsv4lock nfsd_suspend_lock;
extern struct nfsclienthashhead *nfsclienthash;
extern struct nfslockhashhead *nfslockhash;
extern struct nfssessionhash *nfssessionhash;
extern int nfsrv_sessionhashsize;
extern struct nfsstatsv1 nfsstatsv1;
extern struct nfslayouthash *nfslayouthash;
extern int nfsrv_layouthashsize;
extern struct mtx nfsrv_dslock_mtx;
extern int nfs_pnfsiothreads;
extern struct nfsdontlisthead nfsrv_dontlisthead;
extern volatile int nfsrv_dontlistlen;
extern volatile int nfsrv_devidcnt;
extern int nfsrv_maxpnfsmirror;
extern uint32_t nfs_srvmaxio;
extern int nfs_bufpackets;
extern u_long sb_max_adj;
struct vfsoptlist nfsv4root_opt, nfsv4root_newopt;
NFSDLOCKMUTEX;
NFSSTATESPINLOCK;
struct nfsrchash_bucket nfsrchash_table[NFSRVCACHE_HASHSIZE];
struct nfsrchash_bucket nfsrcahash_table[NFSRVCACHE_HASHSIZE];
struct mtx nfsrc_udpmtx;
struct mtx nfs_v4root_mutex;
struct mtx nfsrv_dontlistlock_mtx;
struct mtx nfsrv_recalllock_mtx;
struct nfsrvfh nfs_rootfh, nfs_pubfh;
int nfs_pubfhset = 0, nfs_rootfhset = 0;
struct proc *nfsd_master_proc = NULL;
int nfsd_debuglevel = 0;
static pid_t nfsd_master_pid = (pid_t)-1;
static char nfsd_master_comm[MAXCOMLEN + 1];
static struct timeval nfsd_master_start;
static uint32_t nfsv4_sysid = 0;
static fhandle_t zerofh;
struct callout nfsd_callout;
static int nfssvc_srvcall(struct thread *, struct nfssvc_args *,
struct ucred *);
int nfsrv_enable_crossmntpt = 1;
static int nfs_commit_blks;
static int nfs_commit_miss;
extern int nfsrv_issuedelegs;
extern int nfsrv_dolocallocks;
extern int nfsd_enable_stringtouid;
extern struct nfsdevicehead nfsrv_devidhead;
static int nfsrv_createiovec(int, struct mbuf **, struct mbuf **,
struct iovec **);
static int nfsrv_createiovec_extpgs(int, int, struct mbuf **,
struct mbuf **, struct iovec **);
static int nfsrv_createiovecw(int, struct mbuf *, char *, struct iovec **,
int *);
static void nfsrv_pnfscreate(struct vnode *, struct vattr *, struct ucred *,
NFSPROC_T *);
static void nfsrv_pnfsremovesetup(struct vnode *, NFSPROC_T *, struct vnode **,
int *, char *, fhandle_t *);
static void nfsrv_pnfsremove(struct vnode **, int, char *, fhandle_t *,
NFSPROC_T *);
static int nfsrv_proxyds(struct vnode *, off_t, int, struct ucred *,
struct thread *, int, struct mbuf **, char *, struct mbuf **,
struct nfsvattr *, struct acl *, off_t *, int, bool *);
static int nfsrv_setextattr(struct vnode *, struct nfsvattr *, NFSPROC_T *);
static int nfsrv_readdsrpc(fhandle_t *, off_t, int, struct ucred *,
NFSPROC_T *, struct nfsmount *, struct mbuf **, struct mbuf **);
static int nfsrv_writedsrpc(fhandle_t *, off_t, int, struct ucred *,
NFSPROC_T *, struct vnode *, struct nfsmount **, int, struct mbuf **,
char *, int *);
static int nfsrv_allocatedsrpc(fhandle_t *, off_t, off_t, struct ucred *,
NFSPROC_T *, struct vnode *, struct nfsmount **, int, int *);
static int nfsrv_deallocatedsrpc(fhandle_t *, off_t, off_t, struct ucred *,
NFSPROC_T *, struct vnode *, struct nfsmount **, int, int *);
static int nfsrv_setacldsrpc(fhandle_t *, struct ucred *, NFSPROC_T *,
struct vnode *, struct nfsmount **, int, struct acl *, int *);
static int nfsrv_setattrdsrpc(fhandle_t *, struct ucred *, NFSPROC_T *,
struct vnode *, struct nfsmount **, int, struct nfsvattr *, int *);
static int nfsrv_getattrdsrpc(fhandle_t *, struct ucred *, NFSPROC_T *,
struct vnode *, struct nfsmount *, struct nfsvattr *);
static int nfsrv_seekdsrpc(fhandle_t *, off_t *, int, bool *, struct ucred *,
NFSPROC_T *, struct nfsmount *);
static int nfsrv_putfhname(fhandle_t *, char *);
static int nfsrv_pnfslookupds(struct vnode *, struct vnode *,
struct pnfsdsfile *, struct vnode **, NFSPROC_T *);
static void nfsrv_pnfssetfh(struct vnode *, struct pnfsdsfile *, char *, char *,
struct vnode *, NFSPROC_T *);
static int nfsrv_dsremove(struct vnode *, char *, struct ucred *, NFSPROC_T *);
static int nfsrv_dssetacl(struct vnode *, struct acl *, struct ucred *,
NFSPROC_T *);
static int nfsrv_pnfsstatfs(struct statfs *, struct mount *);
int nfs_pnfsio(task_fn_t *, void *);
SYSCTL_NODE(_vfs, OID_AUTO, nfsd, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
"NFS server");
SYSCTL_INT(_vfs_nfsd, OID_AUTO, mirrormnt, CTLFLAG_RW,
&nfsrv_enable_crossmntpt, 0, "Enable nfsd to cross mount points");
SYSCTL_INT(_vfs_nfsd, OID_AUTO, commit_blks, CTLFLAG_RW, &nfs_commit_blks,
0, "");
SYSCTL_INT(_vfs_nfsd, OID_AUTO, commit_miss, CTLFLAG_RW, &nfs_commit_miss,
0, "");
SYSCTL_INT(_vfs_nfsd, OID_AUTO, issue_delegations, CTLFLAG_RW,
&nfsrv_issuedelegs, 0, "Enable nfsd to issue delegations");
SYSCTL_INT(_vfs_nfsd, OID_AUTO, enable_locallocks, CTLFLAG_RW,
&nfsrv_dolocallocks, 0, "Enable nfsd to acquire local locks on files");
SYSCTL_INT(_vfs_nfsd, OID_AUTO, debuglevel, CTLFLAG_RW, &nfsd_debuglevel,
0, "Debug level for NFS server");
SYSCTL_INT(_vfs_nfsd, OID_AUTO, enable_stringtouid, CTLFLAG_RW,
&nfsd_enable_stringtouid, 0, "Enable nfsd to accept numeric owner_names");
static int nfsrv_pnfsgetdsattr = 1;
SYSCTL_INT(_vfs_nfsd, OID_AUTO, pnfsgetdsattr, CTLFLAG_RW,
&nfsrv_pnfsgetdsattr, 0, "When set getattr gets DS attributes via RPC");
/*
* nfsrv_dsdirsize can only be increased and only when the nfsd threads are
* not running.
* The dsN subdirectories for the increased values must have been created
* on all DS servers before this increase is done.
*/
u_int nfsrv_dsdirsize = 20;
static int
sysctl_dsdirsize(SYSCTL_HANDLER_ARGS)
{
int error, newdsdirsize;
newdsdirsize = nfsrv_dsdirsize;
error = sysctl_handle_int(oidp, &newdsdirsize, 0, req);
if (error != 0 || req->newptr == NULL)
return (error);
if (newdsdirsize <= nfsrv_dsdirsize || newdsdirsize > 10000 ||
newnfs_numnfsd != 0)
return (EINVAL);
nfsrv_dsdirsize = newdsdirsize;
return (0);
}
SYSCTL_PROC(_vfs_nfsd, OID_AUTO, dsdirsize,
CTLTYPE_UINT | CTLFLAG_MPSAFE | CTLFLAG_RW, 0, sizeof(nfsrv_dsdirsize),
sysctl_dsdirsize, "IU", "Number of dsN subdirs on the DS servers");
/*
* nfs_srvmaxio can only be increased and only when the nfsd threads are
* not running. The setting must be a power of 2, with the current limit of
* 1Mbyte.
*/
static int
sysctl_srvmaxio(SYSCTL_HANDLER_ARGS)
{
int error;
u_int newsrvmaxio;
uint64_t tval;
newsrvmaxio = nfs_srvmaxio;
error = sysctl_handle_int(oidp, &newsrvmaxio, 0, req);
if (error != 0 || req->newptr == NULL)
return (error);
if (newsrvmaxio == nfs_srvmaxio)
return (0);
if (newsrvmaxio < nfs_srvmaxio) {
printf("nfsd: vfs.nfsd.srvmaxio can only be increased\n");
return (EINVAL);
}
if (newsrvmaxio > 1048576) {
printf("nfsd: vfs.nfsd.srvmaxio cannot be > 1Mbyte\n");
return (EINVAL);
}
if ((newsrvmaxio & (newsrvmaxio - 1)) != 0) {
printf("nfsd: vfs.nfsd.srvmaxio must be a power of 2\n");
return (EINVAL);
}
/*
* Check that kern.ipc.maxsockbuf is large enough for
* newsrviomax, given the setting of vfs.nfs.bufpackets.
*/
if ((newsrvmaxio + NFS_MAXXDR) * nfs_bufpackets >
sb_max_adj) {
/*
* Suggest vfs.nfs.bufpackets * maximum RPC message for
* sb_max_adj.
*/
tval = (newsrvmaxio + NFS_MAXXDR) * nfs_bufpackets;
/*
* Convert suggested sb_max_adj value to a suggested
* sb_max value, which is what is set via kern.ipc.maxsockbuf.
* Perform the inverse calculation of (from uipc_sockbuf.c):
* sb_max_adj = (u_quad_t)sb_max * MCLBYTES /
* (MSIZE + MCLBYTES);
* XXX If the calculation of sb_max_adj from sb_max changes,
* this calculation must be changed as well.
*/
tval *= (MSIZE + MCLBYTES); /* Brackets for readability. */
tval += MCLBYTES - 1; /* Round up divide. */
tval /= MCLBYTES;
printf("nfsd: set kern.ipc.maxsockbuf to a minimum of "
"%ju to support %ubyte NFS I/O\n", (uintmax_t)tval,
newsrvmaxio);
return (EINVAL);
}
NFSD_LOCK();
if (newnfs_numnfsd != 0) {
NFSD_UNLOCK();
printf("nfsd: cannot set vfs.nfsd.srvmaxio when nfsd "
"threads are running\n");
return (EINVAL);
}
nfs_srvmaxio = newsrvmaxio;
NFSD_UNLOCK();
return (0);
}
SYSCTL_PROC(_vfs_nfsd, OID_AUTO, srvmaxio,
CTLTYPE_UINT | CTLFLAG_MPSAFE | CTLFLAG_RW, NULL, 0,
sysctl_srvmaxio, "IU", "Maximum I/O size in bytes");
#define MAX_REORDERED_RPC 16
#define NUM_HEURISTIC 1031
#define NHUSE_INIT 64
#define NHUSE_INC 16
#define NHUSE_MAX 2048
static struct nfsheur {
struct vnode *nh_vp; /* vp to match (unreferenced pointer) */
off_t nh_nextoff; /* next offset for sequential detection */
int nh_use; /* use count for selection */
int nh_seqcount; /* heuristic */
} nfsheur[NUM_HEURISTIC];
/*
* Heuristic to detect sequential operation.
*/
static struct nfsheur *
nfsrv_sequential_heuristic(struct uio *uio, struct vnode *vp)
{
struct nfsheur *nh;
int hi, try;
/* Locate best candidate. */
try = 32;
hi = ((int)(vm_offset_t)vp / sizeof(struct vnode)) % NUM_HEURISTIC;
nh = &nfsheur[hi];
while (try--) {
if (nfsheur[hi].nh_vp == vp) {
nh = &nfsheur[hi];
break;
}
if (nfsheur[hi].nh_use > 0)
--nfsheur[hi].nh_use;
hi = (hi + 1) % NUM_HEURISTIC;
if (nfsheur[hi].nh_use < nh->nh_use)
nh = &nfsheur[hi];
}
/* Initialize hint if this is a new file. */
if (nh->nh_vp != vp) {
nh->nh_vp = vp;
nh->nh_nextoff = uio->uio_offset;
nh->nh_use = NHUSE_INIT;
if (uio->uio_offset == 0)
nh->nh_seqcount = 4;
else
nh->nh_seqcount = 1;
}
/* Calculate heuristic. */
if ((uio->uio_offset == 0 && nh->nh_seqcount > 0) ||
uio->uio_offset == nh->nh_nextoff) {
/* See comments in vfs_vnops.c:sequential_heuristic(). */
nh->nh_seqcount += howmany(uio->uio_resid, 16384);
if (nh->nh_seqcount > IO_SEQMAX)
nh->nh_seqcount = IO_SEQMAX;
} else if (qabs(uio->uio_offset - nh->nh_nextoff) <= MAX_REORDERED_RPC *
imax(vp->v_mount->mnt_stat.f_iosize, uio->uio_resid)) {
/* Probably a reordered RPC, leave seqcount alone. */
} else if (nh->nh_seqcount > 1) {
nh->nh_seqcount /= 2;
} else {
nh->nh_seqcount = 0;
}
nh->nh_use += NHUSE_INC;
if (nh->nh_use > NHUSE_MAX)
nh->nh_use = NHUSE_MAX;
return (nh);
}
/*
* Get attributes into nfsvattr structure.
*/
int
nfsvno_getattr(struct vnode *vp, struct nfsvattr *nvap,
struct nfsrv_descript *nd, struct thread *p, int vpislocked,
nfsattrbit_t *attrbitp)
{
int error, gotattr, lockedit = 0;
struct nfsvattr na;
if (vpislocked == 0) {
/*
* When vpislocked == 0, the vnode is either exclusively
* locked by this thread or not locked by this thread.
* As such, shared lock it, if not exclusively locked.
*/
if (NFSVOPISLOCKED(vp) != LK_EXCLUSIVE) {
lockedit = 1;
NFSVOPLOCK(vp, LK_SHARED | LK_RETRY);
}
}
/*
* Acquire the Change, Size, TimeAccess, TimeModify and SpaceUsed
* attributes, as required.
* This needs to be done for regular files if:
* - non-NFSv4 RPCs or
* - when attrbitp == NULL or
* - an NFSv4 RPC with any of the above attributes in attrbitp.
* A return of 0 for nfsrv_proxyds() indicates that it has acquired
* these attributes. nfsrv_proxyds() will return an error if the
* server is not a pNFS one.
*/
gotattr = 0;
if (vp->v_type == VREG && nfsrv_devidcnt > 0 && (attrbitp == NULL ||
(nd->nd_flag & ND_NFSV4) == 0 ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_CHANGE) ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_SIZE) ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_TIMEACCESS) ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_TIMEMODIFY) ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_SPACEUSED))) {
error = nfsrv_proxyds(vp, 0, 0, nd->nd_cred, p,
NFSPROC_GETATTR, NULL, NULL, NULL, &na, NULL, NULL, 0,
NULL);
if (error == 0)
gotattr = 1;
}
error = VOP_GETATTR(vp, &nvap->na_vattr, nd->nd_cred);
if (lockedit != 0)
NFSVOPUNLOCK(vp);
/*
* If we got the Change, Size and Modify Time from the DS,
* replace them.
*/
if (gotattr != 0) {
nvap->na_atime = na.na_atime;
nvap->na_mtime = na.na_mtime;
nvap->na_filerev = na.na_filerev;
nvap->na_size = na.na_size;
nvap->na_bytes = na.na_bytes;
}
NFSD_DEBUG(4, "nfsvno_getattr: gotattr=%d err=%d chg=%ju\n", gotattr,
error, (uintmax_t)na.na_filerev);
NFSEXITCODE(error);
return (error);
}
/*
* Get a file handle for a vnode.
*/
int
nfsvno_getfh(struct vnode *vp, fhandle_t *fhp, struct thread *p)
{
int error;
NFSBZERO((caddr_t)fhp, sizeof(fhandle_t));
fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
error = VOP_VPTOFH(vp, &fhp->fh_fid);
NFSEXITCODE(error);
return (error);
}
/*
* Perform access checking for vnodes obtained from file handles that would
* refer to files already opened by a Unix client. You cannot just use
* vn_writechk() and VOP_ACCESSX() for two reasons.
* 1 - You must check for exported rdonly as well as MNT_RDONLY for the write
* case.
* 2 - The owner is to be given access irrespective of mode bits for some
* operations, so that processes that chmod after opening a file don't
* break.
*/
int
nfsvno_accchk(struct vnode *vp, accmode_t accmode, struct ucred *cred,
struct nfsexstuff *exp, struct thread *p, int override, int vpislocked,
u_int32_t *supportedtypep)
{
struct vattr vattr;
int error = 0, getret = 0;
if (vpislocked == 0) {
if (NFSVOPLOCK(vp, LK_SHARED) != 0) {
error = EPERM;
goto out;
}
}
if (accmode & VWRITE) {
/* Just vn_writechk() changed to check rdonly */
/*
* Disallow write attempts on read-only file systems;
* unless the file is a socket or a block or character
* device resident on the file system.
*/
if (NFSVNO_EXRDONLY(exp) ||
(vp->v_mount->mnt_flag & MNT_RDONLY)) {
switch (vp->v_type) {
case VREG:
case VDIR:
case VLNK:
error = EROFS;
default:
break;
}
}
/*
* If there's shared text associated with
* the inode, try to free it up once. If
* we fail, we can't allow writing.
*/
if (VOP_IS_TEXT(vp) && error == 0)
error = ETXTBSY;
}
if (error != 0) {
if (vpislocked == 0)
NFSVOPUNLOCK(vp);
goto out;
}
/*
* Should the override still be applied when ACLs are enabled?
*/
error = VOP_ACCESSX(vp, accmode, cred, p);
if (error != 0 && (accmode & (VDELETE | VDELETE_CHILD))) {
/*
* Try again with VEXPLICIT_DENY, to see if the test for
* deletion is supported.
*/
error = VOP_ACCESSX(vp, accmode | VEXPLICIT_DENY, cred, p);
if (error == 0) {
if (vp->v_type == VDIR) {
accmode &= ~(VDELETE | VDELETE_CHILD);
accmode |= VWRITE;
error = VOP_ACCESSX(vp, accmode, cred, p);
} else if (supportedtypep != NULL) {
*supportedtypep &= ~NFSACCESS_DELETE;
}
}
}
/*
* Allow certain operations for the owner (reads and writes
* on files that are already open).
*/
if (override != NFSACCCHK_NOOVERRIDE &&
(error == EPERM || error == EACCES)) {
if (cred->cr_uid == 0 && (override & NFSACCCHK_ALLOWROOT))
error = 0;
else if (override & NFSACCCHK_ALLOWOWNER) {
getret = VOP_GETATTR(vp, &vattr, cred);
if (getret == 0 && cred->cr_uid == vattr.va_uid)
error = 0;
}
}
if (vpislocked == 0)
NFSVOPUNLOCK(vp);
out:
NFSEXITCODE(error);
return (error);
}
/*
* Set attribute(s) vnop.
*/
int
nfsvno_setattr(struct vnode *vp, struct nfsvattr *nvap, struct ucred *cred,
struct thread *p, struct nfsexstuff *exp)
{
u_quad_t savsize = 0;
int error, savedit;
time_t savbtime;
/*
* If this is an exported file system and a pNFS service is running,
* don't VOP_SETATTR() of size for the MDS file system.
*/
savedit = 0;
error = 0;
if (vp->v_type == VREG && (vp->v_mount->mnt_flag & MNT_EXPORTED) != 0 &&
nfsrv_devidcnt != 0 && nvap->na_vattr.va_size != VNOVAL &&
nvap->na_vattr.va_size > 0) {
savsize = nvap->na_vattr.va_size;
nvap->na_vattr.va_size = VNOVAL;
if (nvap->na_vattr.va_uid != (uid_t)VNOVAL ||
nvap->na_vattr.va_gid != (gid_t)VNOVAL ||
nvap->na_vattr.va_mode != (mode_t)VNOVAL ||
nvap->na_vattr.va_atime.tv_sec != VNOVAL ||
nvap->na_vattr.va_mtime.tv_sec != VNOVAL)
savedit = 1;
else
savedit = 2;
}
if (savedit != 2)
error = VOP_SETATTR(vp, &nvap->na_vattr, cred);
if (savedit != 0)
nvap->na_vattr.va_size = savsize;
if (error == 0 && (nvap->na_vattr.va_uid != (uid_t)VNOVAL ||
nvap->na_vattr.va_gid != (gid_t)VNOVAL ||
nvap->na_vattr.va_size != VNOVAL ||
nvap->na_vattr.va_mode != (mode_t)VNOVAL ||
nvap->na_vattr.va_atime.tv_sec != VNOVAL ||
nvap->na_vattr.va_mtime.tv_sec != VNOVAL)) {
/* Never modify birthtime on a DS file. */
savbtime = nvap->na_vattr.va_birthtime.tv_sec;
nvap->na_vattr.va_birthtime.tv_sec = VNOVAL;
/* For a pNFS server, set the attributes on the DS file. */
error = nfsrv_proxyds(vp, 0, 0, cred, p, NFSPROC_SETATTR,
NULL, NULL, NULL, nvap, NULL, NULL, 0, NULL);
nvap->na_vattr.va_birthtime.tv_sec = savbtime;
if (error == ENOENT)
error = 0;
}
NFSEXITCODE(error);
return (error);
}
/*
* Set up nameidata for a lookup() call and do it.
*/
int
nfsvno_namei(struct nfsrv_descript *nd, struct nameidata *ndp,
struct vnode *dp, int islocked, struct nfsexstuff *exp,
struct vnode **retdirp)
{
struct componentname *cnp = &ndp->ni_cnd;
int i;
struct iovec aiov;
struct uio auio;
int lockleaf = (cnp->cn_flags & LOCKLEAF) != 0, linklen;
int error = 0;
char *cp;
*retdirp = NULL;
cnp->cn_nameptr = cnp->cn_pnbuf;
ndp->ni_lcf = 0;
/*
* Extract and set starting directory.
*/
if (dp->v_type != VDIR) {
if (islocked)
vput(dp);
else
vrele(dp);
nfsvno_relpathbuf(ndp);
error = ENOTDIR;
goto out1;
}
if (islocked)
NFSVOPUNLOCK(dp);
VREF(dp);
*retdirp = dp;
if (NFSVNO_EXRDONLY(exp))
cnp->cn_flags |= RDONLY;
ndp->ni_segflg = UIO_SYSSPACE;
if (nd->nd_flag & ND_PUBLOOKUP) {
ndp->ni_loopcnt = 0;
if (cnp->cn_pnbuf[0] == '/') {
vrele(dp);
/*
* Check for degenerate pathnames here, since lookup()
* panics on them.
*/
for (i = 1; i < ndp->ni_pathlen; i++)
if (cnp->cn_pnbuf[i] != '/')
break;
if (i == ndp->ni_pathlen) {
error = NFSERR_ACCES;
goto out;
}
dp = rootvnode;
VREF(dp);
}
} else if ((nfsrv_enable_crossmntpt == 0 && NFSVNO_EXPORTED(exp)) ||
(nd->nd_flag & ND_NFSV4) == 0) {
/*
* Only cross mount points for NFSv4 when doing a
* mount while traversing the file system above
* the mount point, unless nfsrv_enable_crossmntpt is set.
*/
cnp->cn_flags |= NOCROSSMOUNT;
}
/*
* Initialize for scan, set ni_startdir and bump ref on dp again
* because lookup() will dereference ni_startdir.
*/
ndp->ni_startdir = dp;
ndp->ni_rootdir = rootvnode;
ndp->ni_topdir = NULL;
if (!lockleaf)
cnp->cn_flags |= LOCKLEAF;
for (;;) {
cnp->cn_nameptr = cnp->cn_pnbuf;
/*
* Call lookup() to do the real work. If an error occurs,
* ndp->ni_vp and ni_dvp are left uninitialized or NULL and
* we do not have to dereference anything before returning.
* In either case ni_startdir will be dereferenced and NULLed
* out.
*/
error = lookup(ndp);
if (error)
break;
/*
* Check for encountering a symbolic link. Trivial
* termination occurs if no symlink encountered.
*/
if ((cnp->cn_flags & ISSYMLINK) == 0) {
if ((cnp->cn_flags & (SAVENAME | SAVESTART)) == 0)
nfsvno_relpathbuf(ndp);
if (ndp->ni_vp && !lockleaf)
NFSVOPUNLOCK(ndp->ni_vp);
break;
}
/*
* Validate symlink
*/
if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1)
NFSVOPUNLOCK(ndp->ni_dvp);
if (!(nd->nd_flag & ND_PUBLOOKUP)) {
error = EINVAL;
goto badlink2;
}
if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
error = ELOOP;
goto badlink2;
}
if (ndp->ni_pathlen > 1)
cp = uma_zalloc(namei_zone, M_WAITOK);
else
cp = cnp->cn_pnbuf;
aiov.iov_base = cp;
aiov.iov_len = MAXPATHLEN;
auio.uio_iov = &aiov;
auio.uio_iovcnt = 1;
auio.uio_offset = 0;
auio.uio_rw = UIO_READ;
auio.uio_segflg = UIO_SYSSPACE;
auio.uio_td = NULL;
auio.uio_resid = MAXPATHLEN;
error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
if (error) {
badlink1:
if (ndp->ni_pathlen > 1)
uma_zfree(namei_zone, cp);
badlink2:
vrele(ndp->ni_dvp);
vput(ndp->ni_vp);
break;
}
linklen = MAXPATHLEN - auio.uio_resid;
if (linklen == 0) {
error = ENOENT;
goto badlink1;
}
if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
error = ENAMETOOLONG;
goto badlink1;
}
/*
* Adjust or replace path
*/
if (ndp->ni_pathlen > 1) {
NFSBCOPY(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
uma_zfree(namei_zone, cnp->cn_pnbuf);
cnp->cn_pnbuf = cp;
} else
cnp->cn_pnbuf[linklen] = '\0';
ndp->ni_pathlen += linklen;
/*
* Cleanup refs for next loop and check if root directory
* should replace current directory. Normally ni_dvp
* becomes the new base directory and is cleaned up when
* we loop. Explicitly null pointers after invalidation
* to clarify operation.
*/
vput(ndp->ni_vp);
ndp->ni_vp = NULL;
if (cnp->cn_pnbuf[0] == '/') {
vrele(ndp->ni_dvp);
ndp->ni_dvp = ndp->ni_rootdir;
VREF(ndp->ni_dvp);
}
ndp->ni_startdir = ndp->ni_dvp;
ndp->ni_dvp = NULL;
}
if (!lockleaf)
cnp->cn_flags &= ~LOCKLEAF;
out:
if (error) {
nfsvno_relpathbuf(ndp);
ndp->ni_vp = NULL;
ndp->ni_dvp = NULL;
ndp->ni_startdir = NULL;
} else if ((ndp->ni_cnd.cn_flags & (WANTPARENT|LOCKPARENT)) == 0) {
ndp->ni_dvp = NULL;
}
out1:
NFSEXITCODE2(error, nd);
return (error);
}
/*
* Set up a pathname buffer and return a pointer to it and, optionally
* set a hash pointer.
*/
void
nfsvno_setpathbuf(struct nameidata *ndp, char **bufpp, u_long **hashpp)
{
struct componentname *cnp = &ndp->ni_cnd;
cnp->cn_flags |= (NOMACCHECK | HASBUF);
cnp->cn_pnbuf = uma_zalloc(namei_zone, M_WAITOK);
if (hashpp != NULL)
*hashpp = NULL;
*bufpp = cnp->cn_pnbuf;
}
/*
* Release the above path buffer, if not released by nfsvno_namei().
*/
void
nfsvno_relpathbuf(struct nameidata *ndp)
{
if ((ndp->ni_cnd.cn_flags & HASBUF) == 0)
panic("nfsrelpath");
uma_zfree(namei_zone, ndp->ni_cnd.cn_pnbuf);
ndp->ni_cnd.cn_flags &= ~HASBUF;
}
/*
* Readlink vnode op into an mbuf list.
*/
int
nfsvno_readlink(struct vnode *vp, struct ucred *cred, int maxextsiz,
struct thread *p, struct mbuf **mpp, struct mbuf **mpendp, int *lenp)
{
struct iovec *iv;
struct uio io, *uiop = &io;
struct mbuf *mp, *mp3;
int len, tlen, error = 0;
len = NFS_MAXPATHLEN;
if (maxextsiz > 0)
uiop->uio_iovcnt = nfsrv_createiovec_extpgs(len, maxextsiz,
&mp3, &mp, &iv);
else
uiop->uio_iovcnt = nfsrv_createiovec(len, &mp3, &mp, &iv);
uiop->uio_iov = iv;
uiop->uio_offset = 0;
uiop->uio_resid = len;
uiop->uio_rw = UIO_READ;
uiop->uio_segflg = UIO_SYSSPACE;
uiop->uio_td = NULL;
error = VOP_READLINK(vp, uiop, cred);
free(iv, M_TEMP);
if (error) {
m_freem(mp3);
*lenp = 0;
goto out;
}
if (uiop->uio_resid > 0) {
len -= uiop->uio_resid;
tlen = NFSM_RNDUP(len);
if (tlen == 0) {
m_freem(mp3);
mp3 = mp = NULL;
} else if (tlen != NFS_MAXPATHLEN || tlen != len)
mp = nfsrv_adj(mp3, NFS_MAXPATHLEN - tlen,
tlen - len);
}
*lenp = len;
*mpp = mp3;
*mpendp = mp;
out:
NFSEXITCODE(error);
return (error);
}
/*
* Create an mbuf chain and an associated iovec that can be used to Read
* or Getextattr of data.
* Upon success, return pointers to the first and last mbufs in the chain
* plus the malloc'd iovec and its iovlen.
*/
static int
nfsrv_createiovec(int len, struct mbuf **mpp, struct mbuf **mpendp,
struct iovec **ivp)
{
struct mbuf *m, *m2 = NULL, *m3;
struct iovec *iv;
int i, left, siz;
left = len;
m3 = NULL;
/*
* Generate the mbuf list with the uio_iov ref. to it.
*/
i = 0;
while (left > 0) {
NFSMGET(m);
MCLGET(m, M_WAITOK);
m->m_len = 0;
siz = min(M_TRAILINGSPACE(m), left);
left -= siz;
i++;
if (m3)
m2->m_next = m;
else
m3 = m;
m2 = m;
}
*ivp = iv = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK);
m = m3;
left = len;
i = 0;
while (left > 0) {
if (m == NULL)
panic("nfsrv_createiovec iov");
siz = min(M_TRAILINGSPACE(m), left);
if (siz > 0) {
iv->iov_base = mtod(m, caddr_t) + m->m_len;
iv->iov_len = siz;
m->m_len += siz;
left -= siz;
iv++;
i++;
}
m = m->m_next;
}
*mpp = m3;
*mpendp = m2;
return (i);
}
/*
* Create an mbuf chain and an associated iovec that can be used to Read
* or Getextattr of data.
* Upon success, return pointers to the first and last mbufs in the chain
* plus the malloc'd iovec and its iovlen.
* Same as above, but creates ext_pgs mbuf(s).
*/
static int
nfsrv_createiovec_extpgs(int len, int maxextsiz, struct mbuf **mpp,
struct mbuf **mpendp, struct iovec **ivp)
{
struct mbuf *m, *m2 = NULL, *m3;
struct iovec *iv;
int i, left, pgno, siz;
left = len;
m3 = NULL;
/*
* Generate the mbuf list with the uio_iov ref. to it.
*/
i = 0;
while (left > 0) {
siz = min(left, maxextsiz);
m = mb_alloc_ext_plus_pages(siz, M_WAITOK);
left -= siz;
i += m->m_epg_npgs;
if (m3 != NULL)
m2->m_next = m;
else
m3 = m;
m2 = m;
}
*ivp = iv = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK);
m = m3;
left = len;
i = 0;
pgno = 0;
while (left > 0) {
if (m == NULL)
panic("nfsvno_createiovec_extpgs iov");
siz = min(PAGE_SIZE, left);
if (siz > 0) {
iv->iov_base = (void *)PHYS_TO_DMAP(m->m_epg_pa[pgno]);
iv->iov_len = siz;
m->m_len += siz;
if (pgno == m->m_epg_npgs - 1)
m->m_epg_last_len = siz;
left -= siz;
iv++;
i++;
pgno++;
}
if (pgno == m->m_epg_npgs && left > 0) {
m = m->m_next;
if (m == NULL)
panic("nfsvno_createiovec_extpgs iov");
pgno = 0;
}
}
*mpp = m3;
*mpendp = m2;
return (i);
}
/*
* Read vnode op call into mbuf list.
*/
int
nfsvno_read(struct vnode *vp, off_t off, int cnt, struct ucred *cred,
int maxextsiz, struct thread *p, struct mbuf **mpp,
struct mbuf **mpendp)
{
struct mbuf *m;
struct iovec *iv;
int error = 0, len, tlen, ioflag = 0;
struct mbuf *m3;
struct uio io, *uiop = &io;
struct nfsheur *nh;
/*
* Attempt to read from a DS file. A return of ENOENT implies
* there is no DS file to read.
*/
error = nfsrv_proxyds(vp, off, cnt, cred, p, NFSPROC_READDS, mpp,
NULL, mpendp, NULL, NULL, NULL, 0, NULL);
if (error != ENOENT)
return (error);
len = NFSM_RNDUP(cnt);
if (maxextsiz > 0)
uiop->uio_iovcnt = nfsrv_createiovec_extpgs(len, maxextsiz,
&m3, &m, &iv);
else
uiop->uio_iovcnt = nfsrv_createiovec(len, &m3, &m, &iv);
uiop->uio_iov = iv;
uiop->uio_offset = off;
uiop->uio_resid = len;
uiop->uio_rw = UIO_READ;
uiop->uio_segflg = UIO_SYSSPACE;
uiop->uio_td = NULL;
nh = nfsrv_sequential_heuristic(uiop, vp);
ioflag |= nh->nh_seqcount << IO_SEQSHIFT;
/* XXX KDM make this more systematic? */
nfsstatsv1.srvbytes[NFSV4OP_READ] += uiop->uio_resid;
error = VOP_READ(vp, uiop, IO_NODELOCKED | ioflag, cred);
free(iv, M_TEMP);
if (error) {
m_freem(m3);
*mpp = NULL;
goto out;
}
nh->nh_nextoff = uiop->uio_offset;
tlen = len - uiop->uio_resid;
cnt = cnt < tlen ? cnt : tlen;
tlen = NFSM_RNDUP(cnt);
if (tlen == 0) {
m_freem(m3);
m3 = m = NULL;
} else if (len != tlen || tlen != cnt)
m = nfsrv_adj(m3, len - tlen, tlen - cnt);
*mpp = m3;
*mpendp = m;
out:
NFSEXITCODE(error);
return (error);
}
/*
* Create the iovec for the mbuf chain passed in as an argument.
* The "cp" argument is where the data starts within the first mbuf in
* the chain. It returns the iovec and the iovcnt.
*/
static int
nfsrv_createiovecw(int retlen, struct mbuf *m, char *cp, struct iovec **ivpp,
int *iovcntp)
{
struct mbuf *mp;
struct iovec *ivp;
int cnt, i, len;
/*
* Loop through the mbuf chain, counting how many mbufs are a
* part of this write operation, so the iovec size is known.
*/
cnt = 0;
len = retlen;
mp = m;
i = mtod(mp, caddr_t) + mp->m_len - cp;
while (len > 0) {
if (i > 0) {
len -= i;
cnt++;
}
mp = mp->m_next;
if (!mp) {
if (len > 0)
return (EBADRPC);
} else
i = mp->m_len;
}
/* Now, create the iovec. */
mp = m;
*ivpp = ivp = malloc(cnt * sizeof (struct iovec), M_TEMP,
M_WAITOK);
*iovcntp = cnt;
i = mtod(mp, caddr_t) + mp->m_len - cp;
len = retlen;
while (len > 0) {
if (mp == NULL)
panic("nfsrv_createiovecw");
if (i > 0) {
i = min(i, len);
ivp->iov_base = cp;
ivp->iov_len = i;
ivp++;
len -= i;
}
mp = mp->m_next;
if (mp) {
i = mp->m_len;
cp = mtod(mp, caddr_t);
}
}
return (0);
}
/*
* Write vnode op from an mbuf list.
*/
int
nfsvno_write(struct vnode *vp, off_t off, int retlen, int *stable,
struct mbuf *mp, char *cp, struct ucred *cred, struct thread *p)
{
struct iovec *iv;
int cnt, ioflags, error;
struct uio io, *uiop = &io;
struct nfsheur *nh;
/*
* Attempt to write to a DS file. A return of ENOENT implies
* there is no DS file to write.
*/
error = nfsrv_proxyds(vp, off, retlen, cred, p, NFSPROC_WRITEDS,
&mp, cp, NULL, NULL, NULL, NULL, 0, NULL);
if (error != ENOENT) {
*stable = NFSWRITE_FILESYNC;
return (error);
}
if (*stable == NFSWRITE_UNSTABLE)
ioflags = IO_NODELOCKED;
else
ioflags = (IO_SYNC | IO_NODELOCKED);
error = nfsrv_createiovecw(retlen, mp, cp, &iv, &cnt);
if (error != 0)
return (error);
uiop->uio_iov = iv;
uiop->uio_iovcnt = cnt;
uiop->uio_resid = retlen;
uiop->uio_rw = UIO_WRITE;
uiop->uio_segflg = UIO_SYSSPACE;
NFSUIOPROC(uiop, p);
uiop->uio_offset = off;
nh = nfsrv_sequential_heuristic(uiop, vp);
ioflags |= nh->nh_seqcount << IO_SEQSHIFT;
/* XXX KDM make this more systematic? */
nfsstatsv1.srvbytes[NFSV4OP_WRITE] += uiop->uio_resid;
error = VOP_WRITE(vp, uiop, ioflags, cred);
if (error == 0)
nh->nh_nextoff = uiop->uio_offset;
free(iv, M_TEMP);
NFSEXITCODE(error);
return (error);
}
/*
* Common code for creating a regular file (plus special files for V2).
*/
int
nfsvno_createsub(struct nfsrv_descript *nd, struct nameidata *ndp,
struct vnode **vpp, struct nfsvattr *nvap, int *exclusive_flagp,
int32_t *cverf, NFSDEV_T rdev, struct nfsexstuff *exp)
{
u_quad_t tempsize;
int error;
struct thread *p = curthread;
error = nd->nd_repstat;
if (!error && ndp->ni_vp == NULL) {
if (nvap->na_type == VREG || nvap->na_type == VSOCK) {
vrele(ndp->ni_startdir);
error = VOP_CREATE(ndp->ni_dvp,
&ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr);
/* For a pNFS server, create the data file on a DS. */
if (error == 0 && nvap->na_type == VREG) {
/*
* Create a data file on a DS for a pNFS server.
* This function just returns if not
* running a pNFS DS or the creation fails.
*/
nfsrv_pnfscreate(ndp->ni_vp, &nvap->na_vattr,
nd->nd_cred, p);
}
VOP_VPUT_PAIR(ndp->ni_dvp, error == 0 ? &ndp->ni_vp :
NULL, false);
nfsvno_relpathbuf(ndp);
if (!error) {
if (*exclusive_flagp) {
*exclusive_flagp = 0;
NFSVNO_ATTRINIT(nvap);
nvap->na_atime.tv_sec = cverf[0];
nvap->na_atime.tv_nsec = cverf[1];
error = VOP_SETATTR(ndp->ni_vp,
&nvap->na_vattr, nd->nd_cred);
if (error != 0) {
vput(ndp->ni_vp);
ndp->ni_vp = NULL;
error = NFSERR_NOTSUPP;
}
}
}
/*
* NFS V2 Only. nfsrvd_mknod() does this for V3.
* (This implies, just get out on an error.)
*/
} else if (nvap->na_type == VCHR || nvap->na_type == VBLK ||
nvap->na_type == VFIFO) {
if (nvap->na_type == VCHR && rdev == 0xffffffff)
nvap->na_type = VFIFO;
if (nvap->na_type != VFIFO &&
(error = priv_check_cred(nd->nd_cred, PRIV_VFS_MKNOD_DEV))) {
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
vput(ndp->ni_dvp);
goto out;
}
nvap->na_rdev = rdev;
error = VOP_MKNOD(ndp->ni_dvp, &ndp->ni_vp,
&ndp->ni_cnd, &nvap->na_vattr);
VOP_VPUT_PAIR(ndp->ni_dvp, error == 0 ? &ndp->ni_vp :
NULL, false);
nfsvno_relpathbuf(ndp);
vrele(ndp->ni_startdir);
if (error)
goto out;
} else {
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
vput(ndp->ni_dvp);
error = ENXIO;
goto out;
}
*vpp = ndp->ni_vp;
} else {
/*
* Handle cases where error is already set and/or
* the file exists.
* 1 - clean up the lookup
* 2 - iff !error and na_size set, truncate it
*/
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
*vpp = ndp->ni_vp;
if (ndp->ni_dvp == *vpp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
if (!error && nvap->na_size != VNOVAL) {
error = nfsvno_accchk(*vpp, VWRITE,
nd->nd_cred, exp, p, NFSACCCHK_NOOVERRIDE,
NFSACCCHK_VPISLOCKED, NULL);
if (!error) {
tempsize = nvap->na_size;
NFSVNO_ATTRINIT(nvap);
nvap->na_size = tempsize;
error = VOP_SETATTR(*vpp,
&nvap->na_vattr, nd->nd_cred);
}
}
if (error)
vput(*vpp);
}
out:
NFSEXITCODE(error);
return (error);
}
/*
* Do a mknod vnode op.
*/
int
nfsvno_mknod(struct nameidata *ndp, struct nfsvattr *nvap, struct ucred *cred,
struct thread *p)
{
int error = 0;
enum vtype vtyp;
vtyp = nvap->na_type;
/*
* Iff doesn't exist, create it.
*/
if (ndp->ni_vp) {
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
vput(ndp->ni_dvp);
vrele(ndp->ni_vp);
error = EEXIST;
goto out;
}
if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) {
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
vput(ndp->ni_dvp);
error = NFSERR_BADTYPE;
goto out;
}
if (vtyp == VSOCK) {
vrele(ndp->ni_startdir);
error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
&ndp->ni_cnd, &nvap->na_vattr);
VOP_VPUT_PAIR(ndp->ni_dvp, error == 0 ? &ndp->ni_vp : NULL,
false);
nfsvno_relpathbuf(ndp);
} else {
if (nvap->na_type != VFIFO &&
(error = priv_check_cred(cred, PRIV_VFS_MKNOD_DEV))) {
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
vput(ndp->ni_dvp);
goto out;
}
error = VOP_MKNOD(ndp->ni_dvp, &ndp->ni_vp,
&ndp->ni_cnd, &nvap->na_vattr);
VOP_VPUT_PAIR(ndp->ni_dvp, error == 0 ? &ndp->ni_vp : NULL,
false);
nfsvno_relpathbuf(ndp);
vrele(ndp->ni_startdir);
/*
* Since VOP_MKNOD returns the ni_vp, I can't
* see any reason to do the lookup.
*/
}
out:
NFSEXITCODE(error);
return (error);
}
/*
* Mkdir vnode op.
*/
int
nfsvno_mkdir(struct nameidata *ndp, struct nfsvattr *nvap, uid_t saved_uid,
struct ucred *cred, struct thread *p, struct nfsexstuff *exp)
{
int error = 0;
if (ndp->ni_vp != NULL) {
if (ndp->ni_dvp == ndp->ni_vp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
vrele(ndp->ni_vp);
nfsvno_relpathbuf(ndp);
error = EEXIST;
goto out;
}
error = VOP_MKDIR(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd,
&nvap->na_vattr);
VOP_VPUT_PAIR(ndp->ni_dvp, error == 0 ? &ndp->ni_vp : NULL, false);
nfsvno_relpathbuf(ndp);
out:
NFSEXITCODE(error);
return (error);
}
/*
* symlink vnode op.
*/
int
nfsvno_symlink(struct nameidata *ndp, struct nfsvattr *nvap, char *pathcp,
int pathlen, int not_v2, uid_t saved_uid, struct ucred *cred, struct thread *p,
struct nfsexstuff *exp)
{
int error = 0;
if (ndp->ni_vp) {
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
if (ndp->ni_dvp == ndp->ni_vp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
vrele(ndp->ni_vp);
error = EEXIST;
goto out;
}
error = VOP_SYMLINK(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd,
&nvap->na_vattr, pathcp);
/*
* Although FreeBSD still had the lookup code in
* it for 7/current, there doesn't seem to be any
* point, since VOP_SYMLINK() returns the ni_vp.
* Just vput it for v2.
*/
VOP_VPUT_PAIR(ndp->ni_dvp, &ndp->ni_vp, !not_v2 && error == 0);
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
out:
NFSEXITCODE(error);
return (error);
}
/*
* Parse symbolic link arguments.
* This function has an ugly side effect. It will malloc() an area for
* the symlink and set iov_base to point to it, only if it succeeds.
* So, if it returns with uiop->uio_iov->iov_base != NULL, that must
* be FREE'd later.
*/
int
nfsvno_getsymlink(struct nfsrv_descript *nd, struct nfsvattr *nvap,
struct thread *p, char **pathcpp, int *lenp)
{
u_int32_t *tl;
char *pathcp = NULL;
int error = 0, len;
struct nfsv2_sattr *sp;
*pathcpp = NULL;
*lenp = 0;
if ((nd->nd_flag & ND_NFSV3) &&
(error = nfsrv_sattr(nd, NULL, nvap, NULL, NULL, p)))
goto nfsmout;
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
len = fxdr_unsigned(int, *tl);
if (len > NFS_MAXPATHLEN || len <= 0) {
error = EBADRPC;
goto nfsmout;
}
pathcp = malloc(len + 1, M_TEMP, M_WAITOK);
error = nfsrv_mtostr(nd, pathcp, len);
if (error)
goto nfsmout;
if (nd->nd_flag & ND_NFSV2) {
NFSM_DISSECT(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
nvap->na_mode = fxdr_unsigned(u_int16_t, sp->sa_mode);
}
*pathcpp = pathcp;
*lenp = len;
NFSEXITCODE2(0, nd);
return (0);
nfsmout:
if (pathcp)
free(pathcp, M_TEMP);
NFSEXITCODE2(error, nd);
return (error);
}
/*
* Remove a non-directory object.
*/
int
nfsvno_removesub(struct nameidata *ndp, int is_v4, struct ucred *cred,
struct thread *p, struct nfsexstuff *exp)
{
struct vnode *vp, *dsdvp[NFSDEV_MAXMIRRORS];
int error = 0, mirrorcnt;
char fname[PNFS_FILENAME_LEN + 1];
fhandle_t fh;
vp = ndp->ni_vp;
dsdvp[0] = NULL;
if (vp->v_type == VDIR)
error = NFSERR_ISDIR;
else if (is_v4)
error = nfsrv_checkremove(vp, 1, NULL, (nfsquad_t)((u_quad_t)0),
p);
if (error == 0)
nfsrv_pnfsremovesetup(vp, p, dsdvp, &mirrorcnt, fname, &fh);
if (!error)
error = VOP_REMOVE(ndp->ni_dvp, vp, &ndp->ni_cnd);
if (error == 0 && dsdvp[0] != NULL)
nfsrv_pnfsremove(dsdvp, mirrorcnt, fname, &fh, p);
if (ndp->ni_dvp == vp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
vput(vp);
if ((ndp->ni_cnd.cn_flags & SAVENAME) != 0)
nfsvno_relpathbuf(ndp);
NFSEXITCODE(error);
return (error);
}
/*
* Remove a directory.
*/
int
nfsvno_rmdirsub(struct nameidata *ndp, int is_v4, struct ucred *cred,
struct thread *p, struct nfsexstuff *exp)
{
struct vnode *vp;
int error = 0;
vp = ndp->ni_vp;
if (vp->v_type != VDIR) {
error = ENOTDIR;
goto out;
}
/*
* No rmdir "." please.
*/
if (ndp->ni_dvp == vp) {
error = EINVAL;
goto out;
}
/*
* The root of a mounted filesystem cannot be deleted.
*/
if (vp->v_vflag & VV_ROOT)
error = EBUSY;
out:
if (!error)
error = VOP_RMDIR(ndp->ni_dvp, vp, &ndp->ni_cnd);
if (ndp->ni_dvp == vp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
vput(vp);
if ((ndp->ni_cnd.cn_flags & SAVENAME) != 0)
nfsvno_relpathbuf(ndp);
NFSEXITCODE(error);
return (error);
}
/*
* Rename vnode op.
*/
int
nfsvno_rename(struct nameidata *fromndp, struct nameidata *tondp,
u_int32_t ndstat, u_int32_t ndflag, struct ucred *cred, struct thread *p)
{
struct vnode *fvp, *tvp, *tdvp, *dsdvp[NFSDEV_MAXMIRRORS];
int error = 0, mirrorcnt;
char fname[PNFS_FILENAME_LEN + 1];
fhandle_t fh;
dsdvp[0] = NULL;
fvp = fromndp->ni_vp;
if (ndstat) {
vrele(fromndp->ni_dvp);
vrele(fvp);
error = ndstat;
goto out1;
}
tdvp = tondp->ni_dvp;
tvp = tondp->ni_vp;
if (tvp != NULL) {
if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
error = (ndflag & ND_NFSV2) ? EISDIR : EEXIST;
goto out;
} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
error = (ndflag & ND_NFSV2) ? ENOTDIR : EEXIST;
goto out;
}
if (tvp->v_type == VDIR && tvp->v_mountedhere) {
error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV;
goto out;
}
/*
* A rename to '.' or '..' results in a prematurely
* unlocked vnode on FreeBSD5, so I'm just going to fail that
* here.
*/
if ((tondp->ni_cnd.cn_namelen == 1 &&
tondp->ni_cnd.cn_nameptr[0] == '.') ||
(tondp->ni_cnd.cn_namelen == 2 &&
tondp->ni_cnd.cn_nameptr[0] == '.' &&
tondp->ni_cnd.cn_nameptr[1] == '.')) {
error = EINVAL;
goto out;
}
}
if (fvp->v_type == VDIR && fvp->v_mountedhere) {
error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV;
goto out;
}
if (fvp->v_mount != tdvp->v_mount) {
error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV;
goto out;
}
if (fvp == tdvp) {
error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EINVAL;
goto out;
}
if (fvp == tvp) {
/*
* If source and destination are the same, there is nothing to
* do. Set error to -1 to indicate this.
*/
error = -1;
goto out;
}
if (ndflag & ND_NFSV4) {
if (NFSVOPLOCK(fvp, LK_EXCLUSIVE) == 0) {
error = nfsrv_checkremove(fvp, 0, NULL,
(nfsquad_t)((u_quad_t)0), p);
NFSVOPUNLOCK(fvp);
} else
error = EPERM;
if (tvp && !error)
error = nfsrv_checkremove(tvp, 1, NULL,
(nfsquad_t)((u_quad_t)0), p);
} else {
/*
* For NFSv2 and NFSv3, try to get rid of the delegation, so
* that the NFSv4 client won't be confused by the rename.
* Since nfsd_recalldelegation() can only be called on an
* unlocked vnode at this point and fvp is the file that will
* still exist after the rename, just do fvp.
*/
nfsd_recalldelegation(fvp, p);
}
if (error == 0 && tvp != NULL) {
nfsrv_pnfsremovesetup(tvp, p, dsdvp, &mirrorcnt, fname, &fh);
NFSD_DEBUG(4, "nfsvno_rename: pnfsremovesetup"
" dsdvp=%p\n", dsdvp[0]);
}
out:
if (!error) {
error = VOP_RENAME(fromndp->ni_dvp, fromndp->ni_vp,
&fromndp->ni_cnd, tondp->ni_dvp, tondp->ni_vp,
&tondp->ni_cnd);
} else {
if (tdvp == tvp)
vrele(tdvp);
else
vput(tdvp);
if (tvp)
vput(tvp);
vrele(fromndp->ni_dvp);
vrele(fvp);
if (error == -1)
error = 0;
}
/*
* If dsdvp[0] != NULL, it was set up by nfsrv_pnfsremovesetup() and
* if the rename succeeded, the DS file for the tvp needs to be
* removed.
*/
if (error == 0 && dsdvp[0] != NULL) {
nfsrv_pnfsremove(dsdvp, mirrorcnt, fname, &fh, p);
NFSD_DEBUG(4, "nfsvno_rename: pnfsremove\n");
}
vrele(tondp->ni_startdir);
nfsvno_relpathbuf(tondp);
out1:
vrele(fromndp->ni_startdir);
nfsvno_relpathbuf(fromndp);
NFSEXITCODE(error);
return (error);
}
/*
* Link vnode op.
*/
int
nfsvno_link(struct nameidata *ndp, struct vnode *vp, struct ucred *cred,
struct thread *p, struct nfsexstuff *exp)
{
struct vnode *xp;
int error = 0;
xp = ndp->ni_vp;
if (xp != NULL) {
error = EEXIST;
} else {
xp = ndp->ni_dvp;
if (vp->v_mount != xp->v_mount)
error = EXDEV;
}
if (!error) {
NFSVOPLOCK(vp, LK_EXCLUSIVE | LK_RETRY);
if (!VN_IS_DOOMED(vp))
error = VOP_LINK(ndp->ni_dvp, vp, &ndp->ni_cnd);
else
error = EPERM;
if (ndp->ni_dvp == vp) {
vrele(ndp->ni_dvp);
NFSVOPUNLOCK(vp);
} else {
vref(vp);
VOP_VPUT_PAIR(ndp->ni_dvp, &vp, true);
}
} else {
if (ndp->ni_dvp == ndp->ni_vp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
if (ndp->ni_vp)
vrele(ndp->ni_vp);
}
nfsvno_relpathbuf(ndp);
NFSEXITCODE(error);
return (error);
}
/*
* Do the fsync() appropriate for the commit.
*/
int
nfsvno_fsync(struct vnode *vp, u_int64_t off, int cnt, struct ucred *cred,
struct thread *td)
{
int error = 0;
/*
* RFC 1813 3.3.21: if count is 0, a flush from offset to the end of
* file is done. At this time VOP_FSYNC does not accept offset and
* byte count parameters so call VOP_FSYNC the whole file for now.
* The same is true for NFSv4: RFC 3530 Sec. 14.2.3.
* File systems that do not use the buffer cache (as indicated
* by MNTK_USES_BCACHE not being set) must use VOP_FSYNC().
*/
if (cnt == 0 || cnt > MAX_COMMIT_COUNT ||
(vp->v_mount->mnt_kern_flag & MNTK_USES_BCACHE) == 0) {
/*
* Give up and do the whole thing
*/
if (vp->v_object && vm_object_mightbedirty(vp->v_object)) {
VM_OBJECT_WLOCK(vp->v_object);
vm_object_page_clean(vp->v_object, 0, 0, OBJPC_SYNC);
VM_OBJECT_WUNLOCK(vp->v_object);
}
error = VOP_FSYNC(vp, MNT_WAIT, td);
} else {
/*
* Locate and synchronously write any buffers that fall
* into the requested range. Note: we are assuming that
* f_iosize is a power of 2.
*/
int iosize = vp->v_mount->mnt_stat.f_iosize;
int iomask = iosize - 1;
struct bufobj *bo;
daddr_t lblkno;
/*
* Align to iosize boundary, super-align to page boundary.
*/
if (off & iomask) {
cnt += off & iomask;
off &= ~(u_quad_t)iomask;
}
if (off & PAGE_MASK) {
cnt += off & PAGE_MASK;
off &= ~(u_quad_t)PAGE_MASK;
}
lblkno = off / iosize;
if (vp->v_object && vm_object_mightbedirty(vp->v_object)) {
VM_OBJECT_WLOCK(vp->v_object);
vm_object_page_clean(vp->v_object, off, off + cnt,
OBJPC_SYNC);
VM_OBJECT_WUNLOCK(vp->v_object);
}
bo = &vp->v_bufobj;
BO_LOCK(bo);
while (cnt > 0) {
struct buf *bp;
/*
* If we have a buffer and it is marked B_DELWRI we
* have to lock and write it. Otherwise the prior
* write is assumed to have already been committed.
*
* gbincore() can return invalid buffers now so we
* have to check that bit as well (though B_DELWRI
* should not be set if B_INVAL is set there could be
* a race here since we haven't locked the buffer).
*/
if ((bp = gbincore(&vp->v_bufobj, lblkno)) != NULL) {
if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_SLEEPFAIL |
LK_INTERLOCK, BO_LOCKPTR(bo)) == ENOLCK) {
BO_LOCK(bo);
continue; /* retry */
}
if ((bp->b_flags & (B_DELWRI|B_INVAL)) ==
B_DELWRI) {
bremfree(bp);
bp->b_flags &= ~B_ASYNC;
bwrite(bp);
++nfs_commit_miss;
} else
BUF_UNLOCK(bp);
BO_LOCK(bo);
}
++nfs_commit_blks;
if (cnt < iosize)
break;
cnt -= iosize;
++lblkno;
}
BO_UNLOCK(bo);
}
NFSEXITCODE(error);
return (error);
}
/*
* Statfs vnode op.
*/
int
nfsvno_statfs(struct vnode *vp, struct statfs *sf)
{
struct statfs *tsf;
int error;
tsf = NULL;
if (nfsrv_devidcnt > 0) {
/* For a pNFS service, get the DS numbers. */
tsf = malloc(sizeof(*tsf), M_TEMP, M_WAITOK | M_ZERO);
error = nfsrv_pnfsstatfs(tsf, vp->v_mount);
if (error != 0) {
free(tsf, M_TEMP);
tsf = NULL;
}
}
error = VFS_STATFS(vp->v_mount, sf);
if (error == 0) {
if (tsf != NULL) {
sf->f_blocks = tsf->f_blocks;
sf->f_bavail = tsf->f_bavail;
sf->f_bfree = tsf->f_bfree;
sf->f_bsize = tsf->f_bsize;
}
/*
* Since NFS handles these values as unsigned on the
* wire, there is no way to represent negative values,
* so set them to 0. Without this, they will appear
* to be very large positive values for clients like
* Solaris10.
*/
if (sf->f_bavail < 0)
sf->f_bavail = 0;
if (sf->f_ffree < 0)
sf->f_ffree = 0;
}
free(tsf, M_TEMP);
NFSEXITCODE(error);
return (error);
}
/*
* Do the vnode op stuff for Open. Similar to nfsvno_createsub(), but
* must handle nfsrv_opencheck() calls after any other access checks.
*/
void
nfsvno_open(struct nfsrv_descript *nd, struct nameidata *ndp,
nfsquad_t clientid, nfsv4stateid_t *stateidp, struct nfsstate *stp,
int *exclusive_flagp, struct nfsvattr *nvap, int32_t *cverf, int create,
NFSACL_T *aclp, nfsattrbit_t *attrbitp, struct ucred *cred,
struct nfsexstuff *exp, struct vnode **vpp)
{
struct vnode *vp = NULL;
u_quad_t tempsize;
struct nfsexstuff nes;
struct thread *p = curthread;
if (ndp->ni_vp == NULL)
nd->nd_repstat = nfsrv_opencheck(clientid,
stateidp, stp, NULL, nd, p, nd->nd_repstat);
if (!nd->nd_repstat) {
if (ndp->ni_vp == NULL) {
vrele(ndp->ni_startdir);
nd->nd_repstat = VOP_CREATE(ndp->ni_dvp,
&ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr);
/* For a pNFS server, create the data file on a DS. */
if (nd->nd_repstat == 0) {
/*
* Create a data file on a DS for a pNFS server.
* This function just returns if not
* running a pNFS DS or the creation fails.
*/
nfsrv_pnfscreate(ndp->ni_vp, &nvap->na_vattr,
cred, p);
}
VOP_VPUT_PAIR(ndp->ni_dvp, nd->nd_repstat == 0 ?
&ndp->ni_vp : NULL, false);
nfsvno_relpathbuf(ndp);
if (!nd->nd_repstat) {
if (*exclusive_flagp) {
*exclusive_flagp = 0;
NFSVNO_ATTRINIT(nvap);
nvap->na_atime.tv_sec = cverf[0];
nvap->na_atime.tv_nsec = cverf[1];
nd->nd_repstat = VOP_SETATTR(ndp->ni_vp,
&nvap->na_vattr, cred);
if (nd->nd_repstat != 0) {
vput(ndp->ni_vp);
ndp->ni_vp = NULL;
nd->nd_repstat = NFSERR_NOTSUPP;
} else
NFSSETBIT_ATTRBIT(attrbitp,
NFSATTRBIT_TIMEACCESS);
} else {
nfsrv_fixattr(nd, ndp->ni_vp, nvap,
aclp, p, attrbitp, exp);
}
}
vp = ndp->ni_vp;
} else {
if (ndp->ni_startdir)
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
vp = ndp->ni_vp;
if (create == NFSV4OPEN_CREATE) {
if (ndp->ni_dvp == vp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
}
if (NFSVNO_ISSETSIZE(nvap) && vp->v_type == VREG) {
if (ndp->ni_cnd.cn_flags & RDONLY)
NFSVNO_SETEXRDONLY(&nes);
else
NFSVNO_EXINIT(&nes);
nd->nd_repstat = nfsvno_accchk(vp,
VWRITE, cred, &nes, p,
NFSACCCHK_NOOVERRIDE,
NFSACCCHK_VPISLOCKED, NULL);
nd->nd_repstat = nfsrv_opencheck(clientid,
stateidp, stp, vp, nd, p, nd->nd_repstat);
if (!nd->nd_repstat) {
tempsize = nvap->na_size;
NFSVNO_ATTRINIT(nvap);
nvap->na_size = tempsize;
nd->nd_repstat = VOP_SETATTR(vp,
&nvap->na_vattr, cred);
}
} else if (vp->v_type == VREG) {
nd->nd_repstat = nfsrv_opencheck(clientid,
stateidp, stp, vp, nd, p, nd->nd_repstat);
}
}
} else {
if (ndp->ni_cnd.cn_flags & HASBUF)
nfsvno_relpathbuf(ndp);
if (ndp->ni_startdir && create == NFSV4OPEN_CREATE) {
vrele(ndp->ni_startdir);
if (ndp->ni_dvp == ndp->ni_vp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
if (ndp->ni_vp)
vput(ndp->ni_vp);
}
}
*vpp = vp;
NFSEXITCODE2(0, nd);
}
/*
* Updates the file rev and sets the mtime and ctime
* to the current clock time, returning the va_filerev and va_Xtime
* values.
* Return ESTALE to indicate the vnode is VIRF_DOOMED.
*/
int
nfsvno_updfilerev(struct vnode *vp, struct nfsvattr *nvap,
struct nfsrv_descript *nd, struct thread *p)
{
struct vattr va;
VATTR_NULL(&va);
vfs_timestamp(&va.va_mtime);
if (NFSVOPISLOCKED(vp) != LK_EXCLUSIVE) {
NFSVOPLOCK(vp, LK_UPGRADE | LK_RETRY);
if (VN_IS_DOOMED(vp))
return (ESTALE);
}
(void) VOP_SETATTR(vp, &va, nd->nd_cred);
(void) nfsvno_getattr(vp, nvap, nd, p, 1, NULL);
return (0);
}
/*
* Glue routine to nfsv4_fillattr().
*/
int
nfsvno_fillattr(struct nfsrv_descript *nd, struct mount *mp, struct vnode *vp,
struct nfsvattr *nvap, fhandle_t *fhp, int rderror, nfsattrbit_t *attrbitp,
struct ucred *cred, struct thread *p, int isdgram, int reterr,
int supports_nfsv4acls, int at_root, uint64_t mounted_on_fileno)
{
struct statfs *sf;
int error;
sf = NULL;
if (nfsrv_devidcnt > 0 &&
(NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_SPACEAVAIL) ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_SPACEFREE) ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_SPACETOTAL))) {
sf = malloc(sizeof(*sf), M_TEMP, M_WAITOK | M_ZERO);
error = nfsrv_pnfsstatfs(sf, mp);
if (error != 0) {
free(sf, M_TEMP);
sf = NULL;
}
}
error = nfsv4_fillattr(nd, mp, vp, NULL, &nvap->na_vattr, fhp, rderror,
attrbitp, cred, p, isdgram, reterr, supports_nfsv4acls, at_root,
mounted_on_fileno, sf);
free(sf, M_TEMP);
NFSEXITCODE2(0, nd);
return (error);
}
/* Since the Readdir vnode ops vary, put the entire functions in here. */
/*
* nfs readdir service
* - mallocs what it thinks is enough to read
* count rounded up to a multiple of DIRBLKSIZ <= NFS_MAXREADDIR
* - calls VOP_READDIR()
* - loops around building the reply
* if the output generated exceeds count break out of loop
* The NFSM_CLGET macro is used here so that the reply will be packed
* tightly in mbuf clusters.
* - it trims out records with d_fileno == 0
* this doesn't matter for Unix clients, but they might confuse clients
* for other os'.
* - it trims out records with d_type == DT_WHT
* these cannot be seen through NFS (unless we extend the protocol)
* The alternate call nfsrvd_readdirplus() does lookups as well.
* PS: The NFS protocol spec. does not clarify what the "count" byte
* argument is a count of.. just name strings and file id's or the
* entire reply rpc or ...
* I tried just file name and id sizes and it confused the Sun client,
* so I am using the full rpc size now. The "paranoia.." comment refers
* to including the status longwords that are not a part of the dir.
* "entry" structures, but are in the rpc.
*/
int
nfsrvd_readdir(struct nfsrv_descript *nd, int isdgram,
struct vnode *vp, struct nfsexstuff *exp)
{
struct dirent *dp;
u_int32_t *tl;
int dirlen;
char *cpos, *cend, *rbuf;
struct nfsvattr at;
int nlen, error = 0, getret = 1;
int siz, cnt, fullsiz, eofflag, ncookies;
u_int64_t off, toff, verf __unused;
uint64_t *cookies = NULL, *cookiep;
struct uio io;
struct iovec iv;
int is_ufs;
struct thread *p = curthread;
if (nd->nd_repstat) {
nfsrv_postopattr(nd, getret, &at);
goto out;
}
if (nd->nd_flag & ND_NFSV2) {
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
off = fxdr_unsigned(u_quad_t, *tl++);
} else {
NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
off = fxdr_hyper(tl);
tl += 2;
verf = fxdr_hyper(tl);
tl += 2;
}
toff = off;
cnt = fxdr_unsigned(int, *tl);
if (cnt > NFS_SRVMAXDATA(nd) || cnt < 0)
cnt = NFS_SRVMAXDATA(nd);
siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
fullsiz = siz;
if (nd->nd_flag & ND_NFSV3) {
nd->nd_repstat = getret = nfsvno_getattr(vp, &at, nd, p, 1,
NULL);
#if 0
/*
* va_filerev is not sufficient as a cookie verifier,
* since it is not supposed to change when entries are
* removed/added unless that offset cookies returned to
* the client are no longer valid.
*/
if (!nd->nd_repstat && toff && verf != at.na_filerev)
nd->nd_repstat = NFSERR_BAD_COOKIE;
#endif
}
if (!nd->nd_repstat && vp->v_type != VDIR)
nd->nd_repstat = NFSERR_NOTDIR;
if (nd->nd_repstat == 0 && cnt == 0) {
if (nd->nd_flag & ND_NFSV2)
/* NFSv2 does not have NFSERR_TOOSMALL */
nd->nd_repstat = EPERM;
else
nd->nd_repstat = NFSERR_TOOSMALL;
}
if (!nd->nd_repstat)
nd->nd_repstat = nfsvno_accchk(vp, VEXEC,
nd->nd_cred, exp, p, NFSACCCHK_NOOVERRIDE,
NFSACCCHK_VPISLOCKED, NULL);
if (nd->nd_repstat) {
vput(vp);
if (nd->nd_flag & ND_NFSV3)
nfsrv_postopattr(nd, getret, &at);
goto out;
}
is_ufs = strcmp(vp->v_mount->mnt_vfc->vfc_name, "ufs") == 0;
rbuf = malloc(siz, M_TEMP, M_WAITOK);
again:
eofflag = 0;
if (cookies) {
free(cookies, M_TEMP);
cookies = NULL;
}
iv.iov_base = rbuf;
iv.iov_len = siz;
io.uio_iov = &iv;
io.uio_iovcnt = 1;
io.uio_offset = (off_t)off;
io.uio_resid = siz;
io.uio_segflg = UIO_SYSSPACE;
io.uio_rw = UIO_READ;
io.uio_td = NULL;
nd->nd_repstat = VOP_READDIR(vp, &io, nd->nd_cred, &eofflag, &ncookies,
&cookies);
off = (u_int64_t)io.uio_offset;
if (io.uio_resid)
siz -= io.uio_resid;
if (!cookies && !nd->nd_repstat)
nd->nd_repstat = NFSERR_PERM;
if (nd->nd_flag & ND_NFSV3) {
getret = nfsvno_getattr(vp, &at, nd, p, 1, NULL);
if (!nd->nd_repstat)
nd->nd_repstat = getret;
}
/*
* Handles the failed cases. nd->nd_repstat == 0 past here.
*/
if (nd->nd_repstat) {
vput(vp);
free(rbuf, M_TEMP);
if (cookies)
free(cookies, M_TEMP);
if (nd->nd_flag & ND_NFSV3)
nfsrv_postopattr(nd, getret, &at);
goto out;
}
/*
* If nothing read, return eof
* rpc reply
*/
if (siz == 0) {
vput(vp);
if (nd->nd_flag & ND_NFSV2) {
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
} else {
nfsrv_postopattr(nd, getret, &at);
NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
txdr_hyper(at.na_filerev, tl);
tl += 2;
}
*tl++ = newnfs_false;
*tl = newnfs_true;
free(rbuf, M_TEMP);
free(cookies, M_TEMP);
goto out;
}
/*
* Check for degenerate cases of nothing useful read.
* If so go try again
*/
cpos = rbuf;
cend = rbuf + siz;
dp = (struct dirent *)cpos;
cookiep = cookies;
/*
* For some reason FreeBSD's ufs_readdir() chooses to back the
* directory offset up to a block boundary, so it is necessary to
* skip over the records that precede the requested offset. This
* requires the assumption that file offset cookies monotonically
* increase.
*/
while (cpos < cend && ncookies > 0 &&
(dp->d_fileno == 0 || dp->d_type == DT_WHT ||
(is_ufs == 1 && ((u_quad_t)(*cookiep)) <= toff))) {
cpos += dp->d_reclen;
dp = (struct dirent *)cpos;
cookiep++;
ncookies--;
}
if (cpos >= cend || ncookies == 0) {
siz = fullsiz;
toff = off;
goto again;
}
vput(vp);
/*
* If cnt > MCLBYTES and the reply will not be saved, use
* ext_pgs mbufs for TLS.
* For NFSv4.0, we do not know for sure if the reply will
* be saved, so do not use ext_pgs mbufs for NFSv4.0.
*/
if (cnt > MCLBYTES && siz > MCLBYTES &&
(nd->nd_flag & (ND_TLS | ND_EXTPG | ND_SAVEREPLY)) == ND_TLS &&
(nd->nd_flag & (ND_NFSV4 | ND_NFSV41)) != ND_NFSV4)
nd->nd_flag |= ND_EXTPG;
/*
* dirlen is the size of the reply, including all XDR and must
* not exceed cnt. For NFSv2, RFC1094 didn't clearly indicate
* if the XDR should be included in "count", but to be safe, we do.
* (Include the two booleans at the end of the reply in dirlen now.)
*/
if (nd->nd_flag & ND_NFSV3) {
nfsrv_postopattr(nd, getret, &at);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
txdr_hyper(at.na_filerev, tl);
dirlen = NFSX_V3POSTOPATTR + NFSX_VERF + 2 * NFSX_UNSIGNED;
} else {
dirlen = 2 * NFSX_UNSIGNED;
}
/* Loop through the records and build reply */
while (cpos < cend && ncookies > 0) {
nlen = dp->d_namlen;
if (dp->d_fileno != 0 && dp->d_type != DT_WHT &&
nlen <= NFS_MAXNAMLEN) {
if (nd->nd_flag & ND_NFSV3)
dirlen += (6*NFSX_UNSIGNED + NFSM_RNDUP(nlen));
else
dirlen += (4*NFSX_UNSIGNED + NFSM_RNDUP(nlen));
if (dirlen > cnt) {
eofflag = 0;
break;
}
/*
* Build the directory record xdr from
* the dirent entry.
*/
if (nd->nd_flag & ND_NFSV3) {
NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
*tl++ = newnfs_true;
*tl++ = 0;
} else {
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = newnfs_true;
}
*tl = txdr_unsigned(dp->d_fileno);
(void) nfsm_strtom(nd, dp->d_name, nlen);
if (nd->nd_flag & ND_NFSV3) {
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
txdr_hyper(*cookiep, tl);
} else {
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(*cookiep);
}
}
cpos += dp->d_reclen;
dp = (struct dirent *)cpos;
cookiep++;
ncookies--;
}
if (cpos < cend)
eofflag = 0;
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = newnfs_false;
if (eofflag)
*tl = newnfs_true;
else
*tl = newnfs_false;
free(rbuf, M_TEMP);
free(cookies, M_TEMP);
out:
NFSEXITCODE2(0, nd);
return (0);
nfsmout:
vput(vp);
NFSEXITCODE2(error, nd);
return (error);
}
/*
* Readdirplus for V3 and Readdir for V4.
*/
int
nfsrvd_readdirplus(struct nfsrv_descript *nd, int isdgram,
struct vnode *vp, struct nfsexstuff *exp)
{
struct dirent *dp;
u_int32_t *tl;
int dirlen;
char *cpos, *cend, *rbuf;
struct vnode *nvp;
fhandle_t nfh;
struct nfsvattr nva, at, *nvap = &nva;
struct mbuf *mb0, *mb1;
struct nfsreferral *refp;
int nlen, r, error = 0, getret = 1, usevget = 1;
int siz, cnt, fullsiz, eofflag, ncookies, entrycnt;
caddr_t bpos0, bpos1;
u_int64_t off, toff, verf __unused;
uint64_t *cookies = NULL, *cookiep;
nfsattrbit_t attrbits, rderrbits, savbits;
struct uio io;
struct iovec iv;
struct componentname cn;
int at_root, is_ufs, is_zfs, needs_unbusy, supports_nfsv4acls;
struct mount *mp, *new_mp;
uint64_t mounted_on_fileno;
struct thread *p = curthread;
int bextpg0, bextpg1, bextpgsiz0, bextpgsiz1;
if (nd->nd_repstat) {
nfsrv_postopattr(nd, getret, &at);
goto out;
}
NFSM_DISSECT(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
off = fxdr_hyper(tl);
toff = off;
tl += 2;
verf = fxdr_hyper(tl);
tl += 2;
siz = fxdr_unsigned(int, *tl++);
cnt = fxdr_unsigned(int, *tl);
/*
* Use the server's maximum data transfer size as the upper bound
* on reply datalen.
*/
if (cnt > NFS_SRVMAXDATA(nd) || cnt < 0)
cnt = NFS_SRVMAXDATA(nd);
/*
* siz is a "hint" of how much directory information (name, fileid,
* cookie) should be in the reply. At least one client "hints" 0,
* so I set it to cnt for that case. I also round it up to the
* next multiple of DIRBLKSIZ.
* Since the size of a Readdirplus directory entry reply will always
* be greater than a directory entry returned by VOP_READDIR(), it
* does not make sense to read more than NFS_SRVMAXDATA() via
* VOP_READDIR().
*/
if (siz <= 0)
siz = cnt;
else if (siz > NFS_SRVMAXDATA(nd))
siz = NFS_SRVMAXDATA(nd);
siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
if (nd->nd_flag & ND_NFSV4) {
error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
if (error)
goto nfsmout;
NFSSET_ATTRBIT(&savbits, &attrbits);
NFSCLRNOTFILLABLE_ATTRBIT(&attrbits, nd);
NFSZERO_ATTRBIT(&rderrbits);
NFSSETBIT_ATTRBIT(&rderrbits, NFSATTRBIT_RDATTRERROR);
} else {
NFSZERO_ATTRBIT(&attrbits);
}
fullsiz = siz;
nd->nd_repstat = getret = nfsvno_getattr(vp, &at, nd, p, 1, NULL);
#if 0
if (!nd->nd_repstat) {
if (off && verf != at.na_filerev) {
/*
* va_filerev is not sufficient as a cookie verifier,
* since it is not supposed to change when entries are
* removed/added unless that offset cookies returned to
* the client are no longer valid.
*/
if (nd->nd_flag & ND_NFSV4) {
nd->nd_repstat = NFSERR_NOTSAME;
} else {
nd->nd_repstat = NFSERR_BAD_COOKIE;
}
}
}
#endif
if (!nd->nd_repstat && vp->v_type != VDIR)
nd->nd_repstat = NFSERR_NOTDIR;
if (!nd->nd_repstat && cnt == 0)
nd->nd_repstat = NFSERR_TOOSMALL;
if (!nd->nd_repstat)
nd->nd_repstat = nfsvno_accchk(vp, VEXEC,
nd->nd_cred, exp, p, NFSACCCHK_NOOVERRIDE,
NFSACCCHK_VPISLOCKED, NULL);
if (nd->nd_repstat) {
vput(vp);
if (nd->nd_flag & ND_NFSV3)
nfsrv_postopattr(nd, getret, &at);
goto out;
}
is_ufs = strcmp(vp->v_mount->mnt_vfc->vfc_name, "ufs") == 0;
is_zfs = strcmp(vp->v_mount->mnt_vfc->vfc_name, "zfs") == 0;
rbuf = malloc(siz, M_TEMP, M_WAITOK);
again:
eofflag = 0;
if (cookies) {
free(cookies, M_TEMP);
cookies = NULL;
}
iv.iov_base = rbuf;
iv.iov_len = siz;
io.uio_iov = &iv;
io.uio_iovcnt = 1;
io.uio_offset = (off_t)off;
io.uio_resid = siz;
io.uio_segflg = UIO_SYSSPACE;
io.uio_rw = UIO_READ;
io.uio_td = NULL;
nd->nd_repstat = VOP_READDIR(vp, &io, nd->nd_cred, &eofflag, &ncookies,
&cookies);
off = (u_int64_t)io.uio_offset;
if (io.uio_resid)
siz -= io.uio_resid;
getret = nfsvno_getattr(vp, &at, nd, p, 1, NULL);
if (!cookies && !nd->nd_repstat)
nd->nd_repstat = NFSERR_PERM;
if (!nd->nd_repstat)
nd->nd_repstat = getret;
if (nd->nd_repstat) {
vput(vp);
if (cookies)
free(cookies, M_TEMP);
free(rbuf, M_TEMP);
if (nd->nd_flag & ND_NFSV3)
nfsrv_postopattr(nd, getret, &at);
goto out;
}
/*
* If nothing read, return eof
* rpc reply
*/
if (siz == 0) {
vput(vp);
if (nd->nd_flag & ND_NFSV3)
nfsrv_postopattr(nd, getret, &at);
NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
txdr_hyper(at.na_filerev, tl);
tl += 2;
*tl++ = newnfs_false;
*tl = newnfs_true;
free(cookies, M_TEMP);
free(rbuf, M_TEMP);
goto out;
}
/*
* Check for degenerate cases of nothing useful read.
* If so go try again
*/
cpos = rbuf;
cend = rbuf + siz;
dp = (struct dirent *)cpos;
cookiep = cookies;
/*
* For some reason FreeBSD's ufs_readdir() chooses to back the
* directory offset up to a block boundary, so it is necessary to
* skip over the records that precede the requested offset. This
* requires the assumption that file offset cookies monotonically
* increase.
*/
while (cpos < cend && ncookies > 0 &&
(dp->d_fileno == 0 || dp->d_type == DT_WHT ||
(is_ufs == 1 && ((u_quad_t)(*cookiep)) <= toff) ||
((nd->nd_flag & ND_NFSV4) &&
((dp->d_namlen == 1 && dp->d_name[0] == '.') ||
(dp->d_namlen==2 && dp->d_name[0]=='.' && dp->d_name[1]=='.'))))) {
cpos += dp->d_reclen;
dp = (struct dirent *)cpos;
cookiep++;
ncookies--;
}
if (cpos >= cend || ncookies == 0) {
siz = fullsiz;
toff = off;
goto again;
}
/*
* Busy the file system so that the mount point won't go away
* and, as such, VFS_VGET() can be used safely.
*/
mp = vp->v_mount;
vfs_ref(mp);
NFSVOPUNLOCK(vp);
nd->nd_repstat = vfs_busy(mp, 0);
vfs_rel(mp);
if (nd->nd_repstat != 0) {
vrele(vp);
free(cookies, M_TEMP);
free(rbuf, M_TEMP);
if (nd->nd_flag & ND_NFSV3)
nfsrv_postopattr(nd, getret, &at);
goto out;
}
/*
* Check to see if entries in this directory can be safely acquired
* via VFS_VGET() or if a switch to VOP_LOOKUP() is required.
* ZFS snapshot directories need VOP_LOOKUP(), so that any
* automount of the snapshot directory that is required will
* be done.
* This needs to be done here for NFSv4, since NFSv4 never does
* a VFS_VGET() for "." or "..".
*/
if (is_zfs == 1) {
r = VFS_VGET(mp, at.na_fileid, LK_SHARED, &nvp);
if (r == EOPNOTSUPP) {
usevget = 0;
cn.cn_nameiop = LOOKUP;
cn.cn_lkflags = LK_SHARED | LK_RETRY;
cn.cn_cred = nd->nd_cred;
} else if (r == 0)
vput(nvp);
}
/*
* If the reply is likely to exceed MCLBYTES and the reply will
* not be saved, use ext_pgs mbufs for TLS.
* It is difficult to predict how large each entry will be and
* how many entries have been read, so just assume the directory
* entries grow by a factor of 4 when attributes are included.
* For NFSv4.0, we do not know for sure if the reply will
* be saved, so do not use ext_pgs mbufs for NFSv4.0.
*/
if (cnt > MCLBYTES && siz > MCLBYTES / 4 &&
(nd->nd_flag & (ND_TLS | ND_EXTPG | ND_SAVEREPLY)) == ND_TLS &&
(nd->nd_flag & (ND_NFSV4 | ND_NFSV41)) != ND_NFSV4)
nd->nd_flag |= ND_EXTPG;
/*
* Save this position, in case there is an error before one entry
* is created.
*/
mb0 = nd->nd_mb;
bpos0 = nd->nd_bpos;
bextpg0 = nd->nd_bextpg;
bextpgsiz0 = nd->nd_bextpgsiz;
/*
* Fill in the first part of the reply.
* dirlen is the reply length in bytes and cannot exceed cnt.
* (Include the two booleans at the end of the reply in dirlen now,
* so we recognize when we have exceeded cnt.)
*/
if (nd->nd_flag & ND_NFSV3) {
dirlen = NFSX_V3POSTOPATTR + NFSX_VERF + 2 * NFSX_UNSIGNED;
nfsrv_postopattr(nd, getret, &at);
} else {
dirlen = NFSX_VERF + 2 * NFSX_UNSIGNED;
}
NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
txdr_hyper(at.na_filerev, tl);
/*
* Save this position, in case there is an empty reply needed.
*/
mb1 = nd->nd_mb;
bpos1 = nd->nd_bpos;
bextpg1 = nd->nd_bextpg;
bextpgsiz1 = nd->nd_bextpgsiz;
/* Loop through the records and build reply */
entrycnt = 0;
while (cpos < cend && ncookies > 0 && dirlen < cnt) {
nlen = dp->d_namlen;
if (dp->d_fileno != 0 && dp->d_type != DT_WHT &&
nlen <= NFS_MAXNAMLEN &&
((nd->nd_flag & ND_NFSV3) || nlen > 2 ||
(nlen==2 && (dp->d_name[0]!='.' || dp->d_name[1]!='.'))
|| (nlen == 1 && dp->d_name[0] != '.'))) {
/*
* Save the current position in the reply, in case
* this entry exceeds cnt.
*/
mb1 = nd->nd_mb;
bpos1 = nd->nd_bpos;
bextpg1 = nd->nd_bextpg;
bextpgsiz1 = nd->nd_bextpgsiz;
/*
* For readdir_and_lookup get the vnode using
* the file number.
*/
nvp = NULL;
refp = NULL;
r = 0;
at_root = 0;
needs_unbusy = 0;
new_mp = mp;
mounted_on_fileno = (uint64_t)dp->d_fileno;
if ((nd->nd_flag & ND_NFSV3) ||
NFSNONZERO_ATTRBIT(&savbits)) {
if (nd->nd_flag & ND_NFSV4)
refp = nfsv4root_getreferral(NULL,
vp, dp->d_fileno);
if (refp == NULL) {
if (usevget)
r = VFS_VGET(mp, dp->d_fileno,
LK_SHARED, &nvp);
else
r = EOPNOTSUPP;
if (r == EOPNOTSUPP) {
if (usevget) {
usevget = 0;
cn.cn_nameiop = LOOKUP;
cn.cn_lkflags =
LK_SHARED |
LK_RETRY;
cn.cn_cred =
nd->nd_cred;
}
cn.cn_nameptr = dp->d_name;
cn.cn_namelen = nlen;
cn.cn_flags = ISLASTCN |
NOFOLLOW | LOCKLEAF;
if (nlen == 2 &&
dp->d_name[0] == '.' &&
dp->d_name[1] == '.')
cn.cn_flags |=
ISDOTDOT;
if (NFSVOPLOCK(vp, LK_SHARED)
!= 0) {
nd->nd_repstat = EPERM;
break;
}
if ((vp->v_vflag & VV_ROOT) != 0
&& (cn.cn_flags & ISDOTDOT)
!= 0) {
vref(vp);
nvp = vp;
r = 0;
} else {
r = VOP_LOOKUP(vp, &nvp,
&cn);
if (vp != nvp)
NFSVOPUNLOCK(vp);
}
}
/*
* For NFSv4, check to see if nvp is
* a mount point and get the mount
* point vnode, as required.
*/
if (r == 0 &&
nfsrv_enable_crossmntpt != 0 &&
(nd->nd_flag & ND_NFSV4) != 0 &&
nvp->v_type == VDIR &&
nvp->v_mountedhere != NULL) {
new_mp = nvp->v_mountedhere;
r = vfs_busy(new_mp, 0);
vput(nvp);
nvp = NULL;
if (r == 0) {
r = VFS_ROOT(new_mp,
LK_SHARED, &nvp);
needs_unbusy = 1;
if (r == 0)
at_root = 1;
}
}
}
/*
* If we failed to look up the entry, then it
* has become invalid, most likely removed.
*/
if (r != 0) {
if (needs_unbusy)
vfs_unbusy(new_mp);
goto invalid;
}
KASSERT(refp != NULL || nvp != NULL,
("%s: undetected lookup error", __func__));
if (refp == NULL &&
((nd->nd_flag & ND_NFSV3) ||
NFSNONZERO_ATTRBIT(&attrbits))) {
r = nfsvno_getfh(nvp, &nfh, p);
if (!r)
r = nfsvno_getattr(nvp, nvap, nd, p,
1, &attrbits);
if (r == 0 && is_zfs == 1 &&
nfsrv_enable_crossmntpt != 0 &&
(nd->nd_flag & ND_NFSV4) != 0 &&
nvp->v_type == VDIR &&
vp->v_mount != nvp->v_mount) {
/*
* For a ZFS snapshot, there is a
* pseudo mount that does not set
* v_mountedhere, so it needs to
* be detected via a different
* mount structure.
*/
at_root = 1;
if (new_mp == mp)
new_mp = nvp->v_mount;
}
}
/*
* If we failed to get attributes of the entry,
* then just skip it for NFSv3 (the traditional
* behavior in the old NFS server).
* For NFSv4 the behavior is controlled by
* RDATTRERROR: we either ignore the error or
* fail the request.
* Note that RDATTRERROR is never set for NFSv3.
*/
if (r != 0) {
if (!NFSISSET_ATTRBIT(&attrbits,
NFSATTRBIT_RDATTRERROR)) {
vput(nvp);
if (needs_unbusy != 0)
vfs_unbusy(new_mp);
if ((nd->nd_flag & ND_NFSV3))
goto invalid;
nd->nd_repstat = r;
break;
}
}
}
/*
* Build the directory record xdr
*/
if (nd->nd_flag & ND_NFSV3) {
NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
*tl++ = newnfs_true;
*tl++ = 0;
*tl = txdr_unsigned(dp->d_fileno);
dirlen += nfsm_strtom(nd, dp->d_name, nlen);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
txdr_hyper(*cookiep, tl);
nfsrv_postopattr(nd, 0, nvap);
dirlen += nfsm_fhtom(nd,(u_int8_t *)&nfh,0,1);
dirlen += (5*NFSX_UNSIGNED+NFSX_V3POSTOPATTR);
if (nvp != NULL)
vput(nvp);
} else {
NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
*tl++ = newnfs_true;
txdr_hyper(*cookiep, tl);
dirlen += nfsm_strtom(nd, dp->d_name, nlen);
if (nvp != NULL) {
supports_nfsv4acls =
nfs_supportsnfsv4acls(nvp);
NFSVOPUNLOCK(nvp);
} else
supports_nfsv4acls = 0;
if (refp != NULL) {
dirlen += nfsrv_putreferralattr(nd,
&savbits, refp, 0,
&nd->nd_repstat);
if (nd->nd_repstat) {
if (nvp != NULL)
vrele(nvp);
if (needs_unbusy != 0)
vfs_unbusy(new_mp);
break;
}
} else if (r) {
dirlen += nfsvno_fillattr(nd, new_mp,
nvp, nvap, &nfh, r, &rderrbits,
nd->nd_cred, p, isdgram, 0,
supports_nfsv4acls, at_root,
mounted_on_fileno);
} else {
dirlen += nfsvno_fillattr(nd, new_mp,
nvp, nvap, &nfh, r, &attrbits,
nd->nd_cred, p, isdgram, 0,
supports_nfsv4acls, at_root,
mounted_on_fileno);
}
if (nvp != NULL)
vrele(nvp);
dirlen += (3 * NFSX_UNSIGNED);
}
if (needs_unbusy != 0)
vfs_unbusy(new_mp);
if (dirlen <= cnt)
entrycnt++;
}
invalid:
cpos += dp->d_reclen;
dp = (struct dirent *)cpos;
cookiep++;
ncookies--;
}
vrele(vp);
vfs_unbusy(mp);
/*
* If dirlen > cnt, we must strip off the last entry. If that
* results in an empty reply, report NFSERR_TOOSMALL.
*/
if (dirlen > cnt || nd->nd_repstat) {
if (!nd->nd_repstat && entrycnt == 0)
nd->nd_repstat = NFSERR_TOOSMALL;
if (nd->nd_repstat) {
nfsm_trimtrailing(nd, mb0, bpos0, bextpg0, bextpgsiz0);
if (nd->nd_flag & ND_NFSV3)
nfsrv_postopattr(nd, getret, &at);
} else
nfsm_trimtrailing(nd, mb1, bpos1, bextpg1, bextpgsiz1);
eofflag = 0;
} else if (cpos < cend)
eofflag = 0;
if (!nd->nd_repstat) {
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = newnfs_false;
if (eofflag)
*tl = newnfs_true;
else
*tl = newnfs_false;
}
free(cookies, M_TEMP);
free(rbuf, M_TEMP);
out:
NFSEXITCODE2(0, nd);
return (0);
nfsmout:
vput(vp);
NFSEXITCODE2(error, nd);
return (error);
}
/*
* Get the settable attributes out of the mbuf list.
* (Return 0 or EBADRPC)
*/
int
nfsrv_sattr(struct nfsrv_descript *nd, vnode_t vp, struct nfsvattr *nvap,
nfsattrbit_t *attrbitp, NFSACL_T *aclp, struct thread *p)
{
u_int32_t *tl;
struct nfsv2_sattr *sp;
int error = 0, toclient = 0;
switch (nd->nd_flag & (ND_NFSV2 | ND_NFSV3 | ND_NFSV4)) {
case ND_NFSV2:
NFSM_DISSECT(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
/*
* Some old clients didn't fill in the high order 16bits.
* --> check the low order 2 bytes for 0xffff
*/
if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff)
nvap->na_mode = nfstov_mode(sp->sa_mode);
if (sp->sa_uid != newnfs_xdrneg1)
nvap->na_uid = fxdr_unsigned(uid_t, sp->sa_uid);
if (sp->sa_gid != newnfs_xdrneg1)
nvap->na_gid = fxdr_unsigned(gid_t, sp->sa_gid);
if (sp->sa_size != newnfs_xdrneg1)
nvap->na_size = fxdr_unsigned(u_quad_t, sp->sa_size);
if (sp->sa_atime.nfsv2_sec != newnfs_xdrneg1) {
#ifdef notyet
fxdr_nfsv2time(&sp->sa_atime, &nvap->na_atime);
#else
nvap->na_atime.tv_sec =
fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec);
nvap->na_atime.tv_nsec = 0;
#endif
}
if (sp->sa_mtime.nfsv2_sec != newnfs_xdrneg1)
fxdr_nfsv2time(&sp->sa_mtime, &nvap->na_mtime);
break;
case ND_NFSV3:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (*tl == newnfs_true) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
nvap->na_mode = nfstov_mode(*tl);
}
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (*tl == newnfs_true) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
nvap->na_uid = fxdr_unsigned(uid_t, *tl);
}
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (*tl == newnfs_true) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
nvap->na_gid = fxdr_unsigned(gid_t, *tl);
}
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (*tl == newnfs_true) {
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
nvap->na_size = fxdr_hyper(tl);
}
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
switch (fxdr_unsigned(int, *tl)) {
case NFSV3SATTRTIME_TOCLIENT:
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
fxdr_nfsv3time(tl, &nvap->na_atime);
toclient = 1;
break;
case NFSV3SATTRTIME_TOSERVER:
vfs_timestamp(&nvap->na_atime);
nvap->na_vaflags |= VA_UTIMES_NULL;
break;
}
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
switch (fxdr_unsigned(int, *tl)) {
case NFSV3SATTRTIME_TOCLIENT:
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
fxdr_nfsv3time(tl, &nvap->na_mtime);
nvap->na_vaflags &= ~VA_UTIMES_NULL;
break;
case NFSV3SATTRTIME_TOSERVER:
vfs_timestamp(&nvap->na_mtime);
if (!toclient)
nvap->na_vaflags |= VA_UTIMES_NULL;
break;
}
break;
case ND_NFSV4:
error = nfsv4_sattr(nd, vp, nvap, attrbitp, aclp, p);
}
nfsmout:
NFSEXITCODE2(error, nd);
return (error);
}
/*
* Handle the setable attributes for V4.
* Returns NFSERR_BADXDR if it can't be parsed, 0 otherwise.
*/
int
nfsv4_sattr(struct nfsrv_descript *nd, vnode_t vp, struct nfsvattr *nvap,
nfsattrbit_t *attrbitp, NFSACL_T *aclp, struct thread *p)
{
u_int32_t *tl;
int attrsum = 0;
int i, j;
int error, attrsize, bitpos, aclsize, aceerr, retnotsup = 0;
int moderet, toclient = 0;
u_char *cp, namestr[NFSV4_SMALLSTR + 1];
uid_t uid;
gid_t gid;
u_short mode, mask; /* Same type as va_mode. */
struct vattr va;
error = nfsrv_getattrbits(nd, attrbitp, NULL, &retnotsup);
if (error)
goto nfsmout;
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
attrsize = fxdr_unsigned(int, *tl);
/*
* Loop around getting the setable attributes. If an unsupported
* one is found, set nd_repstat == NFSERR_ATTRNOTSUPP and return.
*/
if (retnotsup) {
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
bitpos = NFSATTRBIT_MAX;
} else {
bitpos = 0;
}
moderet = 0;
for (; bitpos < NFSATTRBIT_MAX; bitpos++) {
if (attrsum > attrsize) {
error = NFSERR_BADXDR;
goto nfsmout;
}
if (NFSISSET_ATTRBIT(attrbitp, bitpos))
switch (bitpos) {
case NFSATTRBIT_SIZE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
if (vp != NULL && vp->v_type != VREG) {
error = (vp->v_type == VDIR) ? NFSERR_ISDIR :
NFSERR_INVAL;
goto nfsmout;
}
nvap->na_size = fxdr_hyper(tl);
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_ACL:
- error = nfsrv_dissectacl(nd, aclp, &aceerr, &aclsize,
- p);
+ error = nfsrv_dissectacl(nd, aclp, true, &aceerr,
+ &aclsize, p);
if (error)
goto nfsmout;
if (aceerr && !nd->nd_repstat)
nd->nd_repstat = aceerr;
attrsum += aclsize;
break;
case NFSATTRBIT_ARCHIVE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (!nd->nd_repstat)
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_HIDDEN:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (!nd->nd_repstat)
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_MIMETYPE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
i = fxdr_unsigned(int, *tl);
error = nfsm_advance(nd, NFSM_RNDUP(i), -1);
if (error)
goto nfsmout;
if (!nd->nd_repstat)
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(i));
break;
case NFSATTRBIT_MODE:
moderet = NFSERR_INVAL; /* Can't do MODESETMASKED. */
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
nvap->na_mode = nfstov_mode(*tl);
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_OWNER:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
j = fxdr_unsigned(int, *tl);
if (j < 0) {
error = NFSERR_BADXDR;
goto nfsmout;
}
if (j > NFSV4_SMALLSTR)
cp = malloc(j + 1, M_NFSSTRING, M_WAITOK);
else
cp = namestr;
error = nfsrv_mtostr(nd, cp, j);
if (error) {
if (j > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
goto nfsmout;
}
if (!nd->nd_repstat) {
nd->nd_repstat = nfsv4_strtouid(nd, cp, j,
&uid);
if (!nd->nd_repstat)
nvap->na_uid = uid;
}
if (j > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(j));
break;
case NFSATTRBIT_OWNERGROUP:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
j = fxdr_unsigned(int, *tl);
if (j < 0) {
error = NFSERR_BADXDR;
goto nfsmout;
}
if (j > NFSV4_SMALLSTR)
cp = malloc(j + 1, M_NFSSTRING, M_WAITOK);
else
cp = namestr;
error = nfsrv_mtostr(nd, cp, j);
if (error) {
if (j > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
goto nfsmout;
}
if (!nd->nd_repstat) {
nd->nd_repstat = nfsv4_strtogid(nd, cp, j,
&gid);
if (!nd->nd_repstat)
nvap->na_gid = gid;
}
if (j > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(j));
break;
case NFSATTRBIT_SYSTEM:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (!nd->nd_repstat)
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_TIMEACCESSSET:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
attrsum += NFSX_UNSIGNED;
if (fxdr_unsigned(int, *tl)==NFSV4SATTRTIME_TOCLIENT) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
fxdr_nfsv4time(tl, &nvap->na_atime);
toclient = 1;
attrsum += NFSX_V4TIME;
} else {
vfs_timestamp(&nvap->na_atime);
nvap->na_vaflags |= VA_UTIMES_NULL;
}
break;
case NFSATTRBIT_TIMEBACKUP:
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
if (!nd->nd_repstat)
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
attrsum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMECREATE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
fxdr_nfsv4time(tl, &nvap->na_btime);
attrsum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMEMODIFYSET:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
attrsum += NFSX_UNSIGNED;
if (fxdr_unsigned(int, *tl)==NFSV4SATTRTIME_TOCLIENT) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
fxdr_nfsv4time(tl, &nvap->na_mtime);
nvap->na_vaflags &= ~VA_UTIMES_NULL;
attrsum += NFSX_V4TIME;
} else {
vfs_timestamp(&nvap->na_mtime);
if (!toclient)
nvap->na_vaflags |= VA_UTIMES_NULL;
}
break;
case NFSATTRBIT_MODESETMASKED:
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
mode = fxdr_unsigned(u_short, *tl++);
mask = fxdr_unsigned(u_short, *tl);
/*
* vp == NULL implies an Open/Create operation.
* This attribute can only be used for Setattr and
* only for NFSv4.1 or higher.
* If moderet != 0, a mode attribute has also been
* specified and this attribute cannot be done in the
* same Setattr operation.
*/
if ((nd->nd_flag & ND_NFSV41) == 0)
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
else if ((mode & ~07777) != 0 || (mask & ~07777) != 0 ||
vp == NULL)
nd->nd_repstat = NFSERR_INVAL;
else if (moderet == 0)
moderet = VOP_GETATTR(vp, &va, nd->nd_cred);
if (moderet == 0)
nvap->na_mode = (mode & mask) |
(va.va_mode & ~mask);
else
nd->nd_repstat = moderet;
attrsum += 2 * NFSX_UNSIGNED;
break;
default:
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
/*
* set bitpos so we drop out of the loop.
*/
bitpos = NFSATTRBIT_MAX;
break;
}
}
/*
* some clients pad the attrlist, so we need to skip over the
* padding.
*/
if (attrsum > attrsize) {
error = NFSERR_BADXDR;
} else {
attrsize = NFSM_RNDUP(attrsize);
if (attrsum < attrsize)
error = nfsm_advance(nd, attrsize - attrsum, -1);
}
nfsmout:
NFSEXITCODE2(error, nd);
return (error);
}
/*
* Check/setup export credentials.
*/
int
nfsd_excred(struct nfsrv_descript *nd, struct nfsexstuff *exp,
struct ucred *credanon, bool testsec)
{
int error;
/*
* Check/setup credentials.
*/
if (nd->nd_flag & ND_GSS)
exp->nes_exflag &= ~MNT_EXPORTANON;
/*
* Check to see if the operation is allowed for this security flavor.
*/
error = 0;
if (testsec) {
error = nfsvno_testexp(nd, exp);
if (error != 0)
goto out;
}
/*
* Check to see if the file system is exported V4 only.
*/
if (NFSVNO_EXV4ONLY(exp) && !(nd->nd_flag & ND_NFSV4)) {
error = NFSERR_PROGNOTV4;
goto out;
}
/*
* Now, map the user credentials.
* (Note that ND_AUTHNONE will only be set for an NFSv3
* Fsinfo RPC. If set for anything else, this code might need
* to change.)
*/
if (NFSVNO_EXPORTED(exp)) {
if (((nd->nd_flag & ND_GSS) == 0 && nd->nd_cred->cr_uid == 0) ||
NFSVNO_EXPORTANON(exp) ||
(nd->nd_flag & ND_AUTHNONE) != 0) {
nd->nd_cred->cr_uid = credanon->cr_uid;
nd->nd_cred->cr_gid = credanon->cr_gid;
crsetgroups(nd->nd_cred, credanon->cr_ngroups,
credanon->cr_groups);
} else if ((nd->nd_flag & ND_GSS) == 0) {
/*
* If using AUTH_SYS, call nfsrv_getgrpscred() to see
* if there is a replacement credential with a group
* list set up by "nfsuserd -manage-gids".
* If there is no replacement, nfsrv_getgrpscred()
* simply returns its argument.
*/
nd->nd_cred = nfsrv_getgrpscred(nd->nd_cred);
}
}
out:
NFSEXITCODE2(error, nd);
return (error);
}
/*
* Check exports.
*/
int
nfsvno_checkexp(struct mount *mp, struct sockaddr *nam, struct nfsexstuff *exp,
struct ucred **credp)
{
int error;
error = VFS_CHECKEXP(mp, nam, &exp->nes_exflag, credp,
&exp->nes_numsecflavor, exp->nes_secflavors);
if (error) {
if (nfs_rootfhset) {
exp->nes_exflag = 0;
exp->nes_numsecflavor = 0;
error = 0;
}
} else if (exp->nes_numsecflavor < 1 || exp->nes_numsecflavor >
MAXSECFLAVORS) {
printf("nfsvno_checkexp: numsecflavors out of range\n");
exp->nes_numsecflavor = 0;
error = EACCES;
}
NFSEXITCODE(error);
return (error);
}
/*
* Get a vnode for a file handle and export stuff.
*/
int
nfsvno_fhtovp(struct mount *mp, fhandle_t *fhp, struct sockaddr *nam,
int lktype, struct vnode **vpp, struct nfsexstuff *exp,
struct ucred **credp)
{
int error;
*credp = NULL;
exp->nes_numsecflavor = 0;
error = VFS_FHTOVP(mp, &fhp->fh_fid, lktype, vpp);
if (error != 0)
/* Make sure the server replies ESTALE to the client. */
error = ESTALE;
if (nam && !error) {
error = VFS_CHECKEXP(mp, nam, &exp->nes_exflag, credp,
&exp->nes_numsecflavor, exp->nes_secflavors);
if (error) {
if (nfs_rootfhset) {
exp->nes_exflag = 0;
exp->nes_numsecflavor = 0;
error = 0;
} else {
vput(*vpp);
}
} else if (exp->nes_numsecflavor < 1 || exp->nes_numsecflavor >
MAXSECFLAVORS) {
printf("nfsvno_fhtovp: numsecflavors out of range\n");
exp->nes_numsecflavor = 0;
error = EACCES;
vput(*vpp);
}
}
NFSEXITCODE(error);
return (error);
}
/*
* nfsd_fhtovp() - convert a fh to a vnode ptr
* - look up fsid in mount list (if not found ret error)
* - get vp and export rights by calling nfsvno_fhtovp()
* - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
* for AUTH_SYS
* - if mpp != NULL, return the mount point so that it can
* be used for vn_finished_write() by the caller
*/
void
nfsd_fhtovp(struct nfsrv_descript *nd, struct nfsrvfh *nfp, int lktype,
struct vnode **vpp, struct nfsexstuff *exp,
struct mount **mpp, int startwrite, int nextop)
{
struct mount *mp, *mpw;
struct ucred *credanon;
fhandle_t *fhp;
int error;
if (mpp != NULL)
*mpp = NULL;
*vpp = NULL;
fhp = (fhandle_t *)nfp->nfsrvfh_data;
mp = vfs_busyfs(&fhp->fh_fsid);
if (mp == NULL) {
nd->nd_repstat = ESTALE;
goto out;
}
if (startwrite) {
mpw = mp;
error = vn_start_write(NULL, &mpw, V_WAIT);
if (error != 0) {
mpw = NULL;
vfs_unbusy(mp);
nd->nd_repstat = ESTALE;
goto out;
}
if (lktype == LK_SHARED && !(MNT_SHARED_WRITES(mp)))
lktype = LK_EXCLUSIVE;
} else
mpw = NULL;
nd->nd_repstat = nfsvno_fhtovp(mp, fhp, nd->nd_nam, lktype, vpp, exp,
&credanon);
vfs_unbusy(mp);
/*
* For NFSv4 without a pseudo root fs, unexported file handles
* can be returned, so that Lookup works everywhere.
*/
if (!nd->nd_repstat && exp->nes_exflag == 0 &&
!(nd->nd_flag & ND_NFSV4)) {
vput(*vpp);
*vpp = NULL;
nd->nd_repstat = EACCES;
}
/*
* Personally, I've never seen any point in requiring a
* reserved port#, since only in the rare case where the
* clients are all boxes with secure system privileges,
* does it provide any enhanced security, but... some people
* believe it to be useful and keep putting this code back in.
* (There is also some "security checker" out there that
* complains if the nfs server doesn't enforce this.)
* However, note the following:
* RFC3530 (NFSv4) specifies that a reserved port# not be
* required.
* RFC2623 recommends that, if a reserved port# is checked for,
* that there be a way to turn that off--> ifdef'd.
*/
#ifdef NFS_REQRSVPORT
if (!nd->nd_repstat) {
struct sockaddr_in *saddr;
struct sockaddr_in6 *saddr6;
saddr = NFSSOCKADDR(nd->nd_nam, struct sockaddr_in *);
saddr6 = NFSSOCKADDR(nd->nd_nam, struct sockaddr_in6 *);
if (!(nd->nd_flag & ND_NFSV4) &&
((saddr->sin_family == AF_INET &&
ntohs(saddr->sin_port) >= IPPORT_RESERVED) ||
(saddr6->sin6_family == AF_INET6 &&
ntohs(saddr6->sin6_port) >= IPPORT_RESERVED))) {
vput(*vpp);
nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
}
}
#endif /* NFS_REQRSVPORT */
/*
* Check/setup credentials.
*/
if (!nd->nd_repstat) {
nd->nd_saveduid = nd->nd_cred->cr_uid;
nd->nd_repstat = nfsd_excred(nd, exp, credanon,
nfsrv_checkwrongsec(nd, nextop, (*vpp)->v_type));
if (nd->nd_repstat)
vput(*vpp);
}
if (credanon != NULL)
crfree(credanon);
if (nd->nd_repstat) {
vn_finished_write(mpw);
*vpp = NULL;
} else if (mpp != NULL) {
*mpp = mpw;
}
out:
NFSEXITCODE2(0, nd);
}
/*
* glue for fp.
*/
static int
fp_getfvp(struct thread *p, int fd, struct file **fpp, struct vnode **vpp)
{
struct filedesc *fdp;
struct file *fp;
int error = 0;
fdp = p->td_proc->p_fd;
if (fd < 0 || fd >= fdp->fd_nfiles ||
(fp = fdp->fd_ofiles[fd].fde_file) == NULL) {
error = EBADF;
goto out;
}
*fpp = fp;
out:
NFSEXITCODE(error);
return (error);
}
/*
* Called from nfssvc() to update the exports list. Just call
* vfs_export(). This has to be done, since the v4 root fake fs isn't
* in the mount list.
*/
int
nfsrv_v4rootexport(void *argp, struct ucred *cred, struct thread *p)
{
struct nfsex_args *nfsexargp = (struct nfsex_args *)argp;
int error = 0;
struct nameidata nd;
fhandle_t fh;
error = vfs_export(&nfsv4root_mnt, &nfsexargp->export);
if ((nfsexargp->export.ex_flags & MNT_DELEXPORT) != 0)
nfs_rootfhset = 0;
else if (error == 0) {
if (nfsexargp->fspec == NULL) {
error = EPERM;
goto out;
}
/*
* If fspec != NULL, this is the v4root path.
*/
NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, nfsexargp->fspec);
if ((error = namei(&nd)) != 0)
goto out;
error = nfsvno_getfh(nd.ni_vp, &fh, p);
vrele(nd.ni_vp);
if (!error) {
nfs_rootfh.nfsrvfh_len = NFSX_MYFH;
NFSBCOPY((caddr_t)&fh,
nfs_rootfh.nfsrvfh_data,
sizeof (fhandle_t));
nfs_rootfhset = 1;
}
}
out:
NFSEXITCODE(error);
return (error);
}
/*
* This function needs to test to see if the system is near its limit
* for memory allocation via malloc() or mget() and return True iff
* either of these resources are near their limit.
* XXX (For now, this is just a stub.)
*/
int nfsrv_testmalloclimit = 0;
int
nfsrv_mallocmget_limit(void)
{
static int printmesg = 0;
static int testval = 1;
if (nfsrv_testmalloclimit && (testval++ % 1000) == 0) {
if ((printmesg++ % 100) == 0)
printf("nfsd: malloc/mget near limit\n");
return (1);
}
return (0);
}
/*
* BSD specific initialization of a mount point.
*/
void
nfsd_mntinit(void)
{
static int inited = 0;
if (inited)
return;
inited = 1;
nfsv4root_mnt.mnt_flag = (MNT_RDONLY | MNT_EXPORTED);
TAILQ_INIT(&nfsv4root_mnt.mnt_nvnodelist);
TAILQ_INIT(&nfsv4root_mnt.mnt_lazyvnodelist);
nfsv4root_mnt.mnt_export = NULL;
TAILQ_INIT(&nfsv4root_opt);
TAILQ_INIT(&nfsv4root_newopt);
nfsv4root_mnt.mnt_opt = &nfsv4root_opt;
nfsv4root_mnt.mnt_optnew = &nfsv4root_newopt;
nfsv4root_mnt.mnt_nvnodelistsize = 0;
nfsv4root_mnt.mnt_lazyvnodelistsize = 0;
}
static void
nfsd_timer(void *arg)
{
nfsrv_servertimer();
callout_reset_sbt(&nfsd_callout, SBT_1S, SBT_1S, nfsd_timer, NULL, 0);
}
/*
* Get a vnode for a file handle, without checking exports, etc.
*/
struct vnode *
nfsvno_getvp(fhandle_t *fhp)
{
struct mount *mp;
struct vnode *vp;
int error;
mp = vfs_busyfs(&fhp->fh_fsid);
if (mp == NULL)
return (NULL);
error = VFS_FHTOVP(mp, &fhp->fh_fid, LK_EXCLUSIVE, &vp);
vfs_unbusy(mp);
if (error)
return (NULL);
return (vp);
}
/*
* Do a local VOP_ADVLOCK().
*/
int
nfsvno_advlock(struct vnode *vp, int ftype, u_int64_t first,
u_int64_t end, struct thread *td)
{
int error = 0;
struct flock fl;
u_int64_t tlen;
if (nfsrv_dolocallocks == 0)
goto out;
ASSERT_VOP_UNLOCKED(vp, "nfsvno_advlock: vp locked");
fl.l_whence = SEEK_SET;
fl.l_type = ftype;
fl.l_start = (off_t)first;
if (end == NFS64BITSSET) {
fl.l_len = 0;
} else {
tlen = end - first;
fl.l_len = (off_t)tlen;
}
/*
* For FreeBSD8, the l_pid and l_sysid must be set to the same
* values for all calls, so that all locks will be held by the
* nfsd server. (The nfsd server handles conflicts between the
* various clients.)
* Since an NFSv4 lockowner is a ClientID plus an array of up to 1024
* bytes, so it can't be put in l_sysid.
*/
if (nfsv4_sysid == 0)
nfsv4_sysid = nlm_acquire_next_sysid();
fl.l_pid = (pid_t)0;
fl.l_sysid = (int)nfsv4_sysid;
if (ftype == F_UNLCK)
error = VOP_ADVLOCK(vp, (caddr_t)td->td_proc, F_UNLCK, &fl,
(F_POSIX | F_REMOTE));
else
error = VOP_ADVLOCK(vp, (caddr_t)td->td_proc, F_SETLK, &fl,
(F_POSIX | F_REMOTE));
out:
NFSEXITCODE(error);
return (error);
}
/*
* Check the nfsv4 root exports.
*/
int
nfsvno_v4rootexport(struct nfsrv_descript *nd)
{
struct ucred *credanon;
int error = 0, numsecflavor, secflavors[MAXSECFLAVORS], i;
uint64_t exflags;
error = vfs_stdcheckexp(&nfsv4root_mnt, nd->nd_nam, &exflags,
&credanon, &numsecflavor, secflavors);
if (error) {
error = NFSERR_PROGUNAVAIL;
goto out;
}
if (credanon != NULL)
crfree(credanon);
for (i = 0; i < numsecflavor; i++) {
if (secflavors[i] == AUTH_SYS)
nd->nd_flag |= ND_EXAUTHSYS;
else if (secflavors[i] == RPCSEC_GSS_KRB5)
nd->nd_flag |= ND_EXGSS;
else if (secflavors[i] == RPCSEC_GSS_KRB5I)
nd->nd_flag |= ND_EXGSSINTEGRITY;
else if (secflavors[i] == RPCSEC_GSS_KRB5P)
nd->nd_flag |= ND_EXGSSPRIVACY;
}
/* And set ND_EXxx flags for TLS. */
if ((exflags & MNT_EXTLS) != 0) {
nd->nd_flag |= ND_EXTLS;
if ((exflags & MNT_EXTLSCERT) != 0)
nd->nd_flag |= ND_EXTLSCERT;
if ((exflags & MNT_EXTLSCERTUSER) != 0)
nd->nd_flag |= ND_EXTLSCERTUSER;
}
out:
NFSEXITCODE(error);
return (error);
}
/*
* Nfs server pseudo system call for the nfsd's
*/
/*
* MPSAFE
*/
static int
nfssvc_nfsd(struct thread *td, struct nfssvc_args *uap)
{
struct file *fp;
struct nfsd_addsock_args sockarg;
struct nfsd_nfsd_args nfsdarg;
struct nfsd_nfsd_oargs onfsdarg;
struct nfsd_pnfsd_args pnfsdarg;
struct vnode *vp, *nvp, *curdvp;
struct pnfsdsfile *pf;
struct nfsdevice *ds, *fds;
cap_rights_t rights;
int buflen, error, ret;
char *buf, *cp, *cp2, *cp3;
char fname[PNFS_FILENAME_LEN + 1];
if (uap->flag & NFSSVC_NFSDADDSOCK) {
error = copyin(uap->argp, (caddr_t)&sockarg, sizeof (sockarg));
if (error)
goto out;
/*
* Since we don't know what rights might be required,
* pretend that we need them all. It is better to be too
* careful than too reckless.
*/
error = fget(td, sockarg.sock,
cap_rights_init_one(&rights, CAP_SOCK_SERVER), &fp);
if (error != 0)
goto out;
if (fp->f_type != DTYPE_SOCKET) {
fdrop(fp, td);
error = EPERM;
goto out;
}
error = nfsrvd_addsock(fp);
fdrop(fp, td);
} else if (uap->flag & NFSSVC_NFSDNFSD) {
if (uap->argp == NULL) {
error = EINVAL;
goto out;
}
if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
error = copyin(uap->argp, &onfsdarg, sizeof(onfsdarg));
if (error == 0) {
nfsdarg.principal = onfsdarg.principal;
nfsdarg.minthreads = onfsdarg.minthreads;
nfsdarg.maxthreads = onfsdarg.maxthreads;
nfsdarg.version = 1;
nfsdarg.addr = NULL;
nfsdarg.addrlen = 0;
nfsdarg.dnshost = NULL;
nfsdarg.dnshostlen = 0;
nfsdarg.dspath = NULL;
nfsdarg.dspathlen = 0;
nfsdarg.mdspath = NULL;
nfsdarg.mdspathlen = 0;
nfsdarg.mirrorcnt = 1;
}
} else
error = copyin(uap->argp, &nfsdarg, sizeof(nfsdarg));
if (error)
goto out;
if (nfsdarg.addrlen > 0 && nfsdarg.addrlen < 10000 &&
nfsdarg.dnshostlen > 0 && nfsdarg.dnshostlen < 10000 &&
nfsdarg.dspathlen > 0 && nfsdarg.dspathlen < 10000 &&
nfsdarg.mdspathlen > 0 && nfsdarg.mdspathlen < 10000 &&
nfsdarg.mirrorcnt >= 1 &&
nfsdarg.mirrorcnt <= NFSDEV_MAXMIRRORS &&
nfsdarg.addr != NULL && nfsdarg.dnshost != NULL &&
nfsdarg.dspath != NULL && nfsdarg.mdspath != NULL) {
NFSD_DEBUG(1, "addrlen=%d dspathlen=%d dnslen=%d"
" mdspathlen=%d mirrorcnt=%d\n", nfsdarg.addrlen,
nfsdarg.dspathlen, nfsdarg.dnshostlen,
nfsdarg.mdspathlen, nfsdarg.mirrorcnt);
cp = malloc(nfsdarg.addrlen + 1, M_TEMP, M_WAITOK);
error = copyin(nfsdarg.addr, cp, nfsdarg.addrlen);
if (error != 0) {
free(cp, M_TEMP);
goto out;
}
cp[nfsdarg.addrlen] = '\0'; /* Ensure nul term. */
nfsdarg.addr = cp;
cp = malloc(nfsdarg.dnshostlen + 1, M_TEMP, M_WAITOK);
error = copyin(nfsdarg.dnshost, cp, nfsdarg.dnshostlen);
if (error != 0) {
free(nfsdarg.addr, M_TEMP);
free(cp, M_TEMP);
goto out;
}
cp[nfsdarg.dnshostlen] = '\0'; /* Ensure nul term. */
nfsdarg.dnshost = cp;
cp = malloc(nfsdarg.dspathlen + 1, M_TEMP, M_WAITOK);
error = copyin(nfsdarg.dspath, cp, nfsdarg.dspathlen);
if (error != 0) {
free(nfsdarg.addr, M_TEMP);
free(nfsdarg.dnshost, M_TEMP);
free(cp, M_TEMP);
goto out;
}
cp[nfsdarg.dspathlen] = '\0'; /* Ensure nul term. */
nfsdarg.dspath = cp;
cp = malloc(nfsdarg.mdspathlen + 1, M_TEMP, M_WAITOK);
error = copyin(nfsdarg.mdspath, cp, nfsdarg.mdspathlen);
if (error != 0) {
free(nfsdarg.addr, M_TEMP);
free(nfsdarg.dnshost, M_TEMP);
free(nfsdarg.dspath, M_TEMP);
free(cp, M_TEMP);
goto out;
}
cp[nfsdarg.mdspathlen] = '\0'; /* Ensure nul term. */
nfsdarg.mdspath = cp;
} else {
nfsdarg.addr = NULL;
nfsdarg.addrlen = 0;
nfsdarg.dnshost = NULL;
nfsdarg.dnshostlen = 0;
nfsdarg.dspath = NULL;
nfsdarg.dspathlen = 0;
nfsdarg.mdspath = NULL;
nfsdarg.mdspathlen = 0;
nfsdarg.mirrorcnt = 1;
}
nfsd_timer(NULL);
error = nfsrvd_nfsd(td, &nfsdarg);
free(nfsdarg.addr, M_TEMP);
free(nfsdarg.dnshost, M_TEMP);
free(nfsdarg.dspath, M_TEMP);
free(nfsdarg.mdspath, M_TEMP);
} else if (uap->flag & NFSSVC_PNFSDS) {
error = copyin(uap->argp, &pnfsdarg, sizeof(pnfsdarg));
if (error == 0 && (pnfsdarg.op == PNFSDOP_DELDSSERVER ||
pnfsdarg.op == PNFSDOP_FORCEDELDS)) {
cp = malloc(PATH_MAX + 1, M_TEMP, M_WAITOK);
error = copyinstr(pnfsdarg.dspath, cp, PATH_MAX + 1,
NULL);
if (error == 0)
error = nfsrv_deldsserver(pnfsdarg.op, cp, td);
free(cp, M_TEMP);
} else if (error == 0 && pnfsdarg.op == PNFSDOP_COPYMR) {
cp = malloc(PATH_MAX + 1, M_TEMP, M_WAITOK);
buflen = sizeof(*pf) * NFSDEV_MAXMIRRORS;
buf = malloc(buflen, M_TEMP, M_WAITOK);
error = copyinstr(pnfsdarg.mdspath, cp, PATH_MAX + 1,
NULL);
NFSD_DEBUG(4, "pnfsdcopymr cp mdspath=%d\n", error);
if (error == 0 && pnfsdarg.dspath != NULL) {
cp2 = malloc(PATH_MAX + 1, M_TEMP, M_WAITOK);
error = copyinstr(pnfsdarg.dspath, cp2,
PATH_MAX + 1, NULL);
NFSD_DEBUG(4, "pnfsdcopymr cp dspath=%d\n",
error);
} else
cp2 = NULL;
if (error == 0 && pnfsdarg.curdspath != NULL) {
cp3 = malloc(PATH_MAX + 1, M_TEMP, M_WAITOK);
error = copyinstr(pnfsdarg.curdspath, cp3,
PATH_MAX + 1, NULL);
NFSD_DEBUG(4, "pnfsdcopymr cp curdspath=%d\n",
error);
} else
cp3 = NULL;
curdvp = NULL;
fds = NULL;
if (error == 0)
error = nfsrv_mdscopymr(cp, cp2, cp3, buf,
&buflen, fname, td, &vp, &nvp, &pf, &ds,
&fds);
NFSD_DEBUG(4, "nfsrv_mdscopymr=%d\n", error);
if (error == 0) {
if (pf->dsf_dir >= nfsrv_dsdirsize) {
printf("copymr: dsdir out of range\n");
pf->dsf_dir = 0;
}
NFSD_DEBUG(4, "copymr: buflen=%d\n", buflen);
error = nfsrv_copymr(vp, nvp,
ds->nfsdev_dsdir[pf->dsf_dir], ds, pf,
(struct pnfsdsfile *)buf,
buflen / sizeof(*pf), td->td_ucred, td);
vput(vp);
vput(nvp);
if (fds != NULL && error == 0) {
curdvp = fds->nfsdev_dsdir[pf->dsf_dir];
ret = vn_lock(curdvp, LK_EXCLUSIVE);
if (ret == 0) {
nfsrv_dsremove(curdvp, fname,
td->td_ucred, td);
NFSVOPUNLOCK(curdvp);
}
}
NFSD_DEBUG(4, "nfsrv_copymr=%d\n", error);
}
free(cp, M_TEMP);
free(cp2, M_TEMP);
free(cp3, M_TEMP);
free(buf, M_TEMP);
}
} else {
error = nfssvc_srvcall(td, uap, td->td_ucred);
}
out:
NFSEXITCODE(error);
return (error);
}
static int
nfssvc_srvcall(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
{
struct nfsex_args export;
struct nfsex_oldargs oexp;
struct file *fp = NULL;
int stablefd, i, len;
struct nfsd_clid adminrevoke;
struct nfsd_dumplist dumplist;
struct nfsd_dumpclients *dumpclients;
struct nfsd_dumplocklist dumplocklist;
struct nfsd_dumplocks *dumplocks;
struct nameidata nd;
vnode_t vp;
int error = EINVAL, igotlock;
struct proc *procp;
gid_t *grps;
static int suspend_nfsd = 0;
if (uap->flag & NFSSVC_PUBLICFH) {
NFSBZERO((caddr_t)&nfs_pubfh.nfsrvfh_data,
sizeof (fhandle_t));
error = copyin(uap->argp,
&nfs_pubfh.nfsrvfh_data, sizeof (fhandle_t));
if (!error)
nfs_pubfhset = 1;
} else if ((uap->flag & (NFSSVC_V4ROOTEXPORT | NFSSVC_NEWSTRUCT)) ==
(NFSSVC_V4ROOTEXPORT | NFSSVC_NEWSTRUCT)) {
error = copyin(uap->argp,(caddr_t)&export,
sizeof (struct nfsex_args));
if (!error) {
grps = NULL;
if (export.export.ex_ngroups > NGROUPS_MAX ||
export.export.ex_ngroups < 0)
error = EINVAL;
else if (export.export.ex_ngroups > 0) {
grps = malloc(export.export.ex_ngroups *
sizeof(gid_t), M_TEMP, M_WAITOK);
error = copyin(export.export.ex_groups, grps,
export.export.ex_ngroups * sizeof(gid_t));
export.export.ex_groups = grps;
} else
export.export.ex_groups = NULL;
if (!error)
error = nfsrv_v4rootexport(&export, cred, p);
free(grps, M_TEMP);
}
} else if ((uap->flag & (NFSSVC_V4ROOTEXPORT | NFSSVC_NEWSTRUCT)) ==
NFSSVC_V4ROOTEXPORT) {
error = copyin(uap->argp,(caddr_t)&oexp,
sizeof (struct nfsex_oldargs));
if (!error) {
memset(&export.export, 0, sizeof(export.export));
export.export.ex_flags = (uint64_t)oexp.export.ex_flags;
export.export.ex_root = oexp.export.ex_root;
export.export.ex_uid = oexp.export.ex_anon.cr_uid;
export.export.ex_ngroups =
oexp.export.ex_anon.cr_ngroups;
export.export.ex_groups = NULL;
if (export.export.ex_ngroups > XU_NGROUPS ||
export.export.ex_ngroups < 0)
error = EINVAL;
else if (export.export.ex_ngroups > 0) {
export.export.ex_groups = malloc(
export.export.ex_ngroups * sizeof(gid_t),
M_TEMP, M_WAITOK);
for (i = 0; i < export.export.ex_ngroups; i++)
export.export.ex_groups[i] =
oexp.export.ex_anon.cr_groups[i];
}
export.export.ex_addr = oexp.export.ex_addr;
export.export.ex_addrlen = oexp.export.ex_addrlen;
export.export.ex_mask = oexp.export.ex_mask;
export.export.ex_masklen = oexp.export.ex_masklen;
export.export.ex_indexfile = oexp.export.ex_indexfile;
export.export.ex_numsecflavors =
oexp.export.ex_numsecflavors;
if (export.export.ex_numsecflavors >= MAXSECFLAVORS ||
export.export.ex_numsecflavors < 0)
error = EINVAL;
else {
for (i = 0; i < export.export.ex_numsecflavors;
i++)
export.export.ex_secflavors[i] =
oexp.export.ex_secflavors[i];
}
export.fspec = oexp.fspec;
if (error == 0)
error = nfsrv_v4rootexport(&export, cred, p);
free(export.export.ex_groups, M_TEMP);
}
} else if (uap->flag & NFSSVC_NOPUBLICFH) {
nfs_pubfhset = 0;
error = 0;
} else if (uap->flag & NFSSVC_STABLERESTART) {
error = copyin(uap->argp, (caddr_t)&stablefd,
sizeof (int));
if (!error)
error = fp_getfvp(p, stablefd, &fp, &vp);
if (!error && (NFSFPFLAG(fp) & (FREAD | FWRITE)) != (FREAD | FWRITE))
error = EBADF;
if (!error && newnfs_numnfsd != 0)
error = EPERM;
if (!error) {
nfsrv_stablefirst.nsf_fp = fp;
nfsrv_setupstable(p);
}
} else if (uap->flag & NFSSVC_ADMINREVOKE) {
error = copyin(uap->argp, (caddr_t)&adminrevoke,
sizeof (struct nfsd_clid));
if (!error)
error = nfsrv_adminrevoke(&adminrevoke, p);
} else if (uap->flag & NFSSVC_DUMPCLIENTS) {
error = copyin(uap->argp, (caddr_t)&dumplist,
sizeof (struct nfsd_dumplist));
if (!error && (dumplist.ndl_size < 1 ||
dumplist.ndl_size > NFSRV_MAXDUMPLIST))
error = EPERM;
if (!error) {
len = sizeof (struct nfsd_dumpclients) * dumplist.ndl_size;
dumpclients = malloc(len, M_TEMP, M_WAITOK | M_ZERO);
nfsrv_dumpclients(dumpclients, dumplist.ndl_size);
error = copyout(dumpclients, dumplist.ndl_list, len);
free(dumpclients, M_TEMP);
}
} else if (uap->flag & NFSSVC_DUMPLOCKS) {
error = copyin(uap->argp, (caddr_t)&dumplocklist,
sizeof (struct nfsd_dumplocklist));
if (!error && (dumplocklist.ndllck_size < 1 ||
dumplocklist.ndllck_size > NFSRV_MAXDUMPLIST))
error = EPERM;
if (!error)
error = nfsrv_lookupfilename(&nd,
dumplocklist.ndllck_fname, p);
if (!error) {
len = sizeof (struct nfsd_dumplocks) *
dumplocklist.ndllck_size;
dumplocks = malloc(len, M_TEMP, M_WAITOK | M_ZERO);
nfsrv_dumplocks(nd.ni_vp, dumplocks,
dumplocklist.ndllck_size, p);
vput(nd.ni_vp);
error = copyout(dumplocks, dumplocklist.ndllck_list,
len);
free(dumplocks, M_TEMP);
}
} else if (uap->flag & NFSSVC_BACKUPSTABLE) {
procp = p->td_proc;
PROC_LOCK(procp);
nfsd_master_pid = procp->p_pid;
bcopy(procp->p_comm, nfsd_master_comm, MAXCOMLEN + 1);
nfsd_master_start = procp->p_stats->p_start;
nfsd_master_proc = procp;
PROC_UNLOCK(procp);
} else if ((uap->flag & NFSSVC_SUSPENDNFSD) != 0) {
NFSLOCKV4ROOTMUTEX();
if (suspend_nfsd == 0) {
/* Lock out all nfsd threads */
do {
igotlock = nfsv4_lock(&nfsd_suspend_lock, 1,
NULL, NFSV4ROOTLOCKMUTEXPTR, NULL);
} while (igotlock == 0 && suspend_nfsd == 0);
suspend_nfsd = 1;
}
NFSUNLOCKV4ROOTMUTEX();
error = 0;
} else if ((uap->flag & NFSSVC_RESUMENFSD) != 0) {
NFSLOCKV4ROOTMUTEX();
if (suspend_nfsd != 0) {
nfsv4_unlock(&nfsd_suspend_lock, 0);
suspend_nfsd = 0;
}
NFSUNLOCKV4ROOTMUTEX();
error = 0;
}
NFSEXITCODE(error);
return (error);
}
/*
* Check exports.
* Returns 0 if ok, 1 otherwise.
*/
int
nfsvno_testexp(struct nfsrv_descript *nd, struct nfsexstuff *exp)
{
int i;
/*
* Allow NFSv3 Fsinfo per RFC2623.
*/
if (((nd->nd_flag & ND_NFSV4) != 0 ||
nd->nd_procnum != NFSPROC_FSINFO) &&
((NFSVNO_EXTLS(exp) && (nd->nd_flag & ND_TLS) == 0) ||
(NFSVNO_EXTLSCERT(exp) &&
(nd->nd_flag & ND_TLSCERT) == 0) ||
(NFSVNO_EXTLSCERTUSER(exp) &&
(nd->nd_flag & ND_TLSCERTUSER) == 0))) {
if ((nd->nd_flag & ND_NFSV4) != 0)
return (NFSERR_WRONGSEC);
#ifdef notnow
/* There is currently no auth_stat for this. */
else if ((nd->nd_flag & ND_TLS) == 0)
return (NFSERR_AUTHERR | AUTH_NEEDS_TLS);
else
return (NFSERR_AUTHERR | AUTH_NEEDS_TLS_MUTUAL_HOST);
#endif
else
return (NFSERR_AUTHERR | AUTH_TOOWEAK);
}
/*
* This seems odd, but allow the case where the security flavor
* list is empty. This happens when NFSv4 is traversing non-exported
* file systems. Exported file systems should always have a non-empty
* security flavor list.
*/
if (exp->nes_numsecflavor == 0)
return (0);
for (i = 0; i < exp->nes_numsecflavor; i++) {
/*
* The tests for privacy and integrity must be first,
* since ND_GSS is set for everything but AUTH_SYS.
*/
if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5P &&
(nd->nd_flag & ND_GSSPRIVACY))
return (0);
if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5I &&
(nd->nd_flag & ND_GSSINTEGRITY))
return (0);
if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5 &&
(nd->nd_flag & ND_GSS))
return (0);
if (exp->nes_secflavors[i] == AUTH_SYS &&
(nd->nd_flag & ND_GSS) == 0)
return (0);
}
if ((nd->nd_flag & ND_NFSV4) != 0)
return (NFSERR_WRONGSEC);
return (NFSERR_AUTHERR | AUTH_TOOWEAK);
}
/*
* Calculate a hash value for the fid in a file handle.
*/
uint32_t
nfsrv_hashfh(fhandle_t *fhp)
{
uint32_t hashval;
hashval = hash32_buf(&fhp->fh_fid, sizeof(struct fid), 0);
return (hashval);
}
/*
* Calculate a hash value for the sessionid.
*/
uint32_t
nfsrv_hashsessionid(uint8_t *sessionid)
{
uint32_t hashval;
hashval = hash32_buf(sessionid, NFSX_V4SESSIONID, 0);
return (hashval);
}
/*
* Signal the userland master nfsd to backup the stable restart file.
*/
void
nfsrv_backupstable(void)
{
struct proc *procp;
if (nfsd_master_proc != NULL) {
procp = pfind(nfsd_master_pid);
/* Try to make sure it is the correct process. */
if (procp == nfsd_master_proc &&
procp->p_stats->p_start.tv_sec ==
nfsd_master_start.tv_sec &&
procp->p_stats->p_start.tv_usec ==
nfsd_master_start.tv_usec &&
strcmp(procp->p_comm, nfsd_master_comm) == 0)
kern_psignal(procp, SIGUSR2);
else
nfsd_master_proc = NULL;
if (procp != NULL)
PROC_UNLOCK(procp);
}
}
/*
* Create a DS data file for nfsrv_pnfscreate(). Called for each mirror.
* The arguments are in a structure, so that they can be passed through
* taskqueue for a kernel process to execute this function.
*/
struct nfsrvdscreate {
int done;
int inprog;
struct task tsk;
struct ucred *tcred;
struct vnode *dvp;
NFSPROC_T *p;
struct pnfsdsfile *pf;
int err;
fhandle_t fh;
struct vattr va;
struct vattr createva;
};
int
nfsrv_dscreate(struct vnode *dvp, struct vattr *vap, struct vattr *nvap,
fhandle_t *fhp, struct pnfsdsfile *pf, struct pnfsdsattr *dsa,
char *fnamep, struct ucred *tcred, NFSPROC_T *p, struct vnode **nvpp)
{
struct vnode *nvp;
struct nameidata named;
struct vattr va;
char *bufp;
u_long *hashp;
struct nfsnode *np;
struct nfsmount *nmp;
int error;
NFSNAMEICNDSET(&named.ni_cnd, tcred, CREATE,
LOCKPARENT | LOCKLEAF | SAVESTART | NOCACHE);
nfsvno_setpathbuf(&named, &bufp, &hashp);
named.ni_cnd.cn_lkflags = LK_EXCLUSIVE;
named.ni_cnd.cn_nameptr = bufp;
if (fnamep != NULL) {
strlcpy(bufp, fnamep, PNFS_FILENAME_LEN + 1);
named.ni_cnd.cn_namelen = strlen(bufp);
} else
named.ni_cnd.cn_namelen = nfsrv_putfhname(fhp, bufp);
NFSD_DEBUG(4, "nfsrv_dscreate: dvp=%p fname=%s\n", dvp, bufp);
/* Create the date file in the DS mount. */
error = NFSVOPLOCK(dvp, LK_EXCLUSIVE);
if (error == 0) {
error = VOP_CREATE(dvp, &nvp, &named.ni_cnd, vap);
vref(dvp);
VOP_VPUT_PAIR(dvp, error == 0 ? &nvp : NULL, false);
if (error == 0) {
/* Set the ownership of the file. */
error = VOP_SETATTR(nvp, nvap, tcred);
NFSD_DEBUG(4, "nfsrv_dscreate:"
" setattr-uid=%d\n", error);
if (error != 0)
vput(nvp);
}
if (error != 0)
printf("pNFS: pnfscreate failed=%d\n", error);
} else
printf("pNFS: pnfscreate vnlock=%d\n", error);
if (error == 0) {
np = VTONFS(nvp);
nmp = VFSTONFS(nvp->v_mount);
if (strcmp(nvp->v_mount->mnt_vfc->vfc_name, "nfs")
!= 0 || nmp->nm_nam->sa_len > sizeof(
struct sockaddr_in6) ||
np->n_fhp->nfh_len != NFSX_MYFH) {
printf("Bad DS file: fstype=%s salen=%d"
" fhlen=%d\n",
nvp->v_mount->mnt_vfc->vfc_name,
nmp->nm_nam->sa_len, np->n_fhp->nfh_len);
error = ENOENT;
}
/* Set extattrs for the DS on the MDS file. */
if (error == 0) {
if (dsa != NULL) {
error = VOP_GETATTR(nvp, &va, tcred);
if (error == 0) {
dsa->dsa_filerev = va.va_filerev;
dsa->dsa_size = va.va_size;
dsa->dsa_atime = va.va_atime;
dsa->dsa_mtime = va.va_mtime;
dsa->dsa_bytes = va.va_bytes;
}
}
if (error == 0) {
NFSBCOPY(np->n_fhp->nfh_fh, &pf->dsf_fh,
NFSX_MYFH);
NFSBCOPY(nmp->nm_nam, &pf->dsf_sin,
nmp->nm_nam->sa_len);
NFSBCOPY(named.ni_cnd.cn_nameptr,
pf->dsf_filename,
sizeof(pf->dsf_filename));
}
} else
printf("pNFS: pnfscreate can't get DS"
" attr=%d\n", error);
if (nvpp != NULL && error == 0)
*nvpp = nvp;
else
vput(nvp);
}
nfsvno_relpathbuf(&named);
return (error);
}
/*
* Start up the thread that will execute nfsrv_dscreate().
*/
static void
start_dscreate(void *arg, int pending)
{
struct nfsrvdscreate *dsc;
dsc = (struct nfsrvdscreate *)arg;
dsc->err = nfsrv_dscreate(dsc->dvp, &dsc->createva, &dsc->va, &dsc->fh,
dsc->pf, NULL, NULL, dsc->tcred, dsc->p, NULL);
dsc->done = 1;
NFSD_DEBUG(4, "start_dscreate: err=%d\n", dsc->err);
}
/*
* Create a pNFS data file on the Data Server(s).
*/
static void
nfsrv_pnfscreate(struct vnode *vp, struct vattr *vap, struct ucred *cred,
NFSPROC_T *p)
{
struct nfsrvdscreate *dsc, *tdsc = NULL;
struct nfsdevice *ds, *tds, *fds;
struct mount *mp;
struct pnfsdsfile *pf, *tpf;
struct pnfsdsattr dsattr;
struct vattr va;
struct vnode *dvp[NFSDEV_MAXMIRRORS];
struct nfsmount *nmp;
fhandle_t fh;
uid_t vauid;
gid_t vagid;
u_short vamode;
struct ucred *tcred;
int dsdir[NFSDEV_MAXMIRRORS], error, i, mirrorcnt, ret;
int failpos, timo;
/* Get a DS server directory in a round-robin order. */
mirrorcnt = 1;
mp = vp->v_mount;
ds = fds = NULL;
NFSDDSLOCK();
/*
* Search for the first entry that handles this MDS fs, but use the
* first entry for all MDS fs's otherwise.
*/
TAILQ_FOREACH(tds, &nfsrv_devidhead, nfsdev_list) {
if (tds->nfsdev_nmp != NULL) {
if (tds->nfsdev_mdsisset == 0 && ds == NULL)
ds = tds;
else if (tds->nfsdev_mdsisset != 0 && fsidcmp(
&mp->mnt_stat.f_fsid, &tds->nfsdev_mdsfsid) == 0) {
ds = fds = tds;
break;
}
}
}
if (ds == NULL) {
NFSDDSUNLOCK();
NFSD_DEBUG(4, "nfsrv_pnfscreate: no srv\n");
return;
}
i = dsdir[0] = ds->nfsdev_nextdir;
ds->nfsdev_nextdir = (ds->nfsdev_nextdir + 1) % nfsrv_dsdirsize;
dvp[0] = ds->nfsdev_dsdir[i];
tds = TAILQ_NEXT(ds, nfsdev_list);
if (nfsrv_maxpnfsmirror > 1 && tds != NULL) {
TAILQ_FOREACH_FROM(tds, &nfsrv_devidhead, nfsdev_list) {
if (tds->nfsdev_nmp != NULL &&
((tds->nfsdev_mdsisset == 0 && fds == NULL) ||
(tds->nfsdev_mdsisset != 0 && fds != NULL &&
fsidcmp(&mp->mnt_stat.f_fsid,
&tds->nfsdev_mdsfsid) == 0))) {
dsdir[mirrorcnt] = i;
dvp[mirrorcnt] = tds->nfsdev_dsdir[i];
mirrorcnt++;
if (mirrorcnt >= nfsrv_maxpnfsmirror)
break;
}
}
}
/* Put at end of list to implement round-robin usage. */
TAILQ_REMOVE(&nfsrv_devidhead, ds, nfsdev_list);
TAILQ_INSERT_TAIL(&nfsrv_devidhead, ds, nfsdev_list);
NFSDDSUNLOCK();
dsc = NULL;
if (mirrorcnt > 1)
tdsc = dsc = malloc(sizeof(*dsc) * (mirrorcnt - 1), M_TEMP,
M_WAITOK | M_ZERO);
tpf = pf = malloc(sizeof(*pf) * nfsrv_maxpnfsmirror, M_TEMP, M_WAITOK |
M_ZERO);
error = nfsvno_getfh(vp, &fh, p);
if (error == 0)
error = VOP_GETATTR(vp, &va, cred);
if (error == 0) {
/* Set the attributes for "vp" to Setattr the DS vp. */
vauid = va.va_uid;
vagid = va.va_gid;
vamode = va.va_mode;
VATTR_NULL(&va);
va.va_uid = vauid;
va.va_gid = vagid;
va.va_mode = vamode;
va.va_size = 0;
} else
printf("pNFS: pnfscreate getfh+attr=%d\n", error);
NFSD_DEBUG(4, "nfsrv_pnfscreate: cruid=%d crgid=%d\n", cred->cr_uid,
cred->cr_gid);
/* Make data file name based on FH. */
tcred = newnfs_getcred();
/*
* Create the file on each DS mirror, using kernel process(es) for the
* additional mirrors.
*/
failpos = -1;
for (i = 0; i < mirrorcnt - 1 && error == 0; i++, tpf++, tdsc++) {
tpf->dsf_dir = dsdir[i];
tdsc->tcred = tcred;
tdsc->p = p;
tdsc->pf = tpf;
tdsc->createva = *vap;
NFSBCOPY(&fh, &tdsc->fh, sizeof(fh));
tdsc->va = va;
tdsc->dvp = dvp[i];
tdsc->done = 0;
tdsc->inprog = 0;
tdsc->err = 0;
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_dscreate, tdsc);
NFSD_DEBUG(4, "nfsrv_pnfscreate: nfs_pnfsio=%d\n", ret);
}
if (ret != 0) {
ret = nfsrv_dscreate(dvp[i], vap, &va, &fh, tpf, NULL,
NULL, tcred, p, NULL);
if (ret != 0) {
KASSERT(error == 0, ("nfsrv_dscreate err=%d",
error));
if (failpos == -1 && nfsds_failerr(ret))
failpos = i;
else
error = ret;
}
}
}
if (error == 0) {
tpf->dsf_dir = dsdir[mirrorcnt - 1];
error = nfsrv_dscreate(dvp[mirrorcnt - 1], vap, &va, &fh, tpf,
&dsattr, NULL, tcred, p, NULL);
if (failpos == -1 && mirrorcnt > 1 && nfsds_failerr(error)) {
failpos = mirrorcnt - 1;
error = 0;
}
}
timo = hz / 50; /* Wait for 20msec. */
if (timo < 1)
timo = 1;
/* Wait for kernel task(s) to complete. */
for (tdsc = dsc, i = 0; i < mirrorcnt - 1; i++, tdsc++) {
while (tdsc->inprog != 0 && tdsc->done == 0)
tsleep(&tdsc->tsk, PVFS, "srvdcr", timo);
if (tdsc->err != 0) {
if (failpos == -1 && nfsds_failerr(tdsc->err))
failpos = i;
else if (error == 0)
error = tdsc->err;
}
}
/*
* If failpos has been set, that mirror has failed, so it needs
* to be disabled.
*/
if (failpos >= 0) {
nmp = VFSTONFS(dvp[failpos]->v_mount);
NFSLOCKMNT(nmp);
if ((nmp->nm_privflag & (NFSMNTP_FORCEDISM |
NFSMNTP_CANCELRPCS)) == 0) {
nmp->nm_privflag |= NFSMNTP_CANCELRPCS;
NFSUNLOCKMNT(nmp);
ds = nfsrv_deldsnmp(PNFSDOP_DELDSSERVER, nmp, p);
NFSD_DEBUG(4, "dscreatfail fail=%d ds=%p\n", failpos,
ds);
if (ds != NULL)
nfsrv_killrpcs(nmp);
NFSLOCKMNT(nmp);
nmp->nm_privflag &= ~NFSMNTP_CANCELRPCS;
wakeup(nmp);
}
NFSUNLOCKMNT(nmp);
}
NFSFREECRED(tcred);
if (error == 0) {
ASSERT_VOP_ELOCKED(vp, "nfsrv_pnfscreate vp");
NFSD_DEBUG(4, "nfsrv_pnfscreate: mirrorcnt=%d maxmirror=%d\n",
mirrorcnt, nfsrv_maxpnfsmirror);
/*
* For all mirrors that couldn't be created, fill in the
* *pf structure, but with an IP address == 0.0.0.0.
*/
tpf = pf + mirrorcnt;
for (i = mirrorcnt; i < nfsrv_maxpnfsmirror; i++, tpf++) {
*tpf = *pf;
tpf->dsf_sin.sin_family = AF_INET;
tpf->dsf_sin.sin_len = sizeof(struct sockaddr_in);
tpf->dsf_sin.sin_addr.s_addr = 0;
tpf->dsf_sin.sin_port = 0;
}
error = vn_extattr_set(vp, IO_NODELOCKED,
EXTATTR_NAMESPACE_SYSTEM, "pnfsd.dsfile",
sizeof(*pf) * nfsrv_maxpnfsmirror, (char *)pf, p);
if (error == 0)
error = vn_extattr_set(vp, IO_NODELOCKED,
EXTATTR_NAMESPACE_SYSTEM, "pnfsd.dsattr",
sizeof(dsattr), (char *)&dsattr, p);
if (error != 0)
printf("pNFS: pnfscreate setextattr=%d\n",
error);
} else
printf("pNFS: pnfscreate=%d\n", error);
free(pf, M_TEMP);
free(dsc, M_TEMP);
}
/*
* Get the information needed to remove the pNFS Data Server file from the
* Metadata file. Upon success, ddvp is set non-NULL to the locked
* DS directory vnode. The caller must unlock *ddvp when done with it.
*/
static void
nfsrv_pnfsremovesetup(struct vnode *vp, NFSPROC_T *p, struct vnode **dvpp,
int *mirrorcntp, char *fname, fhandle_t *fhp)
{
struct vattr va;
struct ucred *tcred;
char *buf;
int buflen, error;
dvpp[0] = NULL;
/* If not an exported regular file or not a pNFS server, just return. */
if (vp->v_type != VREG || (vp->v_mount->mnt_flag & MNT_EXPORTED) == 0 ||
nfsrv_devidcnt == 0)
return;
/* Check to see if this is the last hard link. */
tcred = newnfs_getcred();
error = VOP_GETATTR(vp, &va, tcred);
NFSFREECRED(tcred);
if (error != 0) {
printf("pNFS: nfsrv_pnfsremovesetup getattr=%d\n", error);
return;
}
if (va.va_nlink > 1)
return;
error = nfsvno_getfh(vp, fhp, p);
if (error != 0) {
printf("pNFS: nfsrv_pnfsremovesetup getfh=%d\n", error);
return;
}
buflen = 1024;
buf = malloc(buflen, M_TEMP, M_WAITOK);
/* Get the directory vnode for the DS mount and the file handle. */
error = nfsrv_dsgetsockmnt(vp, 0, buf, &buflen, mirrorcntp, p, dvpp,
NULL, NULL, fname, NULL, NULL, NULL, NULL, NULL);
free(buf, M_TEMP);
if (error != 0)
printf("pNFS: nfsrv_pnfsremovesetup getsockmnt=%d\n", error);
}
/*
* Remove a DS data file for nfsrv_pnfsremove(). Called for each mirror.
* The arguments are in a structure, so that they can be passed through
* taskqueue for a kernel process to execute this function.
*/
struct nfsrvdsremove {
int done;
int inprog;
struct task tsk;
struct ucred *tcred;
struct vnode *dvp;
NFSPROC_T *p;
int err;
char fname[PNFS_FILENAME_LEN + 1];
};
static int
nfsrv_dsremove(struct vnode *dvp, char *fname, struct ucred *tcred,
NFSPROC_T *p)
{
struct nameidata named;
struct vnode *nvp;
char *bufp;
u_long *hashp;
int error;
error = NFSVOPLOCK(dvp, LK_EXCLUSIVE);
if (error != 0)
return (error);
named.ni_cnd.cn_nameiop = DELETE;
named.ni_cnd.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
named.ni_cnd.cn_cred = tcred;
named.ni_cnd.cn_flags = ISLASTCN | LOCKPARENT | LOCKLEAF | SAVENAME;
nfsvno_setpathbuf(&named, &bufp, &hashp);
named.ni_cnd.cn_nameptr = bufp;
named.ni_cnd.cn_namelen = strlen(fname);
strlcpy(bufp, fname, NAME_MAX);
NFSD_DEBUG(4, "nfsrv_pnfsremove: filename=%s\n", bufp);
error = VOP_LOOKUP(dvp, &nvp, &named.ni_cnd);
NFSD_DEBUG(4, "nfsrv_pnfsremove: aft LOOKUP=%d\n", error);
if (error == 0) {
error = VOP_REMOVE(dvp, nvp, &named.ni_cnd);
vput(nvp);
}
NFSVOPUNLOCK(dvp);
nfsvno_relpathbuf(&named);
if (error != 0)
printf("pNFS: nfsrv_pnfsremove failed=%d\n", error);
return (error);
}
/*
* Start up the thread that will execute nfsrv_dsremove().
*/
static void
start_dsremove(void *arg, int pending)
{
struct nfsrvdsremove *dsrm;
dsrm = (struct nfsrvdsremove *)arg;
dsrm->err = nfsrv_dsremove(dsrm->dvp, dsrm->fname, dsrm->tcred,
dsrm->p);
dsrm->done = 1;
NFSD_DEBUG(4, "start_dsremove: err=%d\n", dsrm->err);
}
/*
* Remove a pNFS data file from a Data Server.
* nfsrv_pnfsremovesetup() must have been called before the MDS file was
* removed to set up the dvp and fill in the FH.
*/
static void
nfsrv_pnfsremove(struct vnode **dvp, int mirrorcnt, char *fname, fhandle_t *fhp,
NFSPROC_T *p)
{
struct ucred *tcred;
struct nfsrvdsremove *dsrm, *tdsrm;
struct nfsdevice *ds;
struct nfsmount *nmp;
int failpos, i, ret, timo;
tcred = newnfs_getcred();
dsrm = NULL;
if (mirrorcnt > 1)
dsrm = malloc(sizeof(*dsrm) * mirrorcnt - 1, M_TEMP, M_WAITOK);
/*
* Remove the file on each DS mirror, using kernel process(es) for the
* additional mirrors.
*/
failpos = -1;
for (tdsrm = dsrm, i = 0; i < mirrorcnt - 1; i++, tdsrm++) {
tdsrm->tcred = tcred;
tdsrm->p = p;
tdsrm->dvp = dvp[i];
strlcpy(tdsrm->fname, fname, PNFS_FILENAME_LEN + 1);
tdsrm->inprog = 0;
tdsrm->done = 0;
tdsrm->err = 0;
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_dsremove, tdsrm);
NFSD_DEBUG(4, "nfsrv_pnfsremove: nfs_pnfsio=%d\n", ret);
}
if (ret != 0) {
ret = nfsrv_dsremove(dvp[i], fname, tcred, p);
if (failpos == -1 && nfsds_failerr(ret))
failpos = i;
}
}
ret = nfsrv_dsremove(dvp[mirrorcnt - 1], fname, tcred, p);
if (failpos == -1 && mirrorcnt > 1 && nfsds_failerr(ret))
failpos = mirrorcnt - 1;
timo = hz / 50; /* Wait for 20msec. */
if (timo < 1)
timo = 1;
/* Wait for kernel task(s) to complete. */
for (tdsrm = dsrm, i = 0; i < mirrorcnt - 1; i++, tdsrm++) {
while (tdsrm->inprog != 0 && tdsrm->done == 0)
tsleep(&tdsrm->tsk, PVFS, "srvdsrm", timo);
if (failpos == -1 && nfsds_failerr(tdsrm->err))
failpos = i;
}
/*
* If failpos has been set, that mirror has failed, so it needs
* to be disabled.
*/
if (failpos >= 0) {
nmp = VFSTONFS(dvp[failpos]->v_mount);
NFSLOCKMNT(nmp);
if ((nmp->nm_privflag & (NFSMNTP_FORCEDISM |
NFSMNTP_CANCELRPCS)) == 0) {
nmp->nm_privflag |= NFSMNTP_CANCELRPCS;
NFSUNLOCKMNT(nmp);
ds = nfsrv_deldsnmp(PNFSDOP_DELDSSERVER, nmp, p);
NFSD_DEBUG(4, "dsremovefail fail=%d ds=%p\n", failpos,
ds);
if (ds != NULL)
nfsrv_killrpcs(nmp);
NFSLOCKMNT(nmp);
nmp->nm_privflag &= ~NFSMNTP_CANCELRPCS;
wakeup(nmp);
}
NFSUNLOCKMNT(nmp);
}
/* Get rid all layouts for the file. */
nfsrv_freefilelayouts(fhp);
NFSFREECRED(tcred);
free(dsrm, M_TEMP);
}
/*
* Generate a file name based on the file handle and put it in *bufp.
* Return the number of bytes generated.
*/
static int
nfsrv_putfhname(fhandle_t *fhp, char *bufp)
{
int i;
uint8_t *cp;
const uint8_t *hexdigits = "0123456789abcdef";
cp = (uint8_t *)fhp;
for (i = 0; i < sizeof(*fhp); i++) {
bufp[2 * i] = hexdigits[(*cp >> 4) & 0xf];
bufp[2 * i + 1] = hexdigits[*cp++ & 0xf];
}
bufp[2 * i] = '\0';
return (2 * i);
}
/*
* Update the Metadata file's attributes from the DS file when a Read/Write
* layout is returned.
* Basically just call nfsrv_proxyds() with procedure == NFSPROC_LAYOUTRETURN
* so that it does a nfsrv_getattrdsrpc() and nfsrv_setextattr() on the DS file.
*/
int
nfsrv_updatemdsattr(struct vnode *vp, struct nfsvattr *nap, NFSPROC_T *p)
{
struct ucred *tcred;
int error;
/* Do this as root so that it won't fail with EACCES. */
tcred = newnfs_getcred();
error = nfsrv_proxyds(vp, 0, 0, tcred, p, NFSPROC_LAYOUTRETURN,
NULL, NULL, NULL, nap, NULL, NULL, 0, NULL);
NFSFREECRED(tcred);
return (error);
}
/*
* Set the NFSv4 ACL on the DS file to the same ACL as the MDS file.
*/
static int
nfsrv_dssetacl(struct vnode *vp, struct acl *aclp, struct ucred *cred,
NFSPROC_T *p)
{
int error;
error = nfsrv_proxyds(vp, 0, 0, cred, p, NFSPROC_SETACL,
NULL, NULL, NULL, NULL, aclp, NULL, 0, NULL);
return (error);
}
static int
nfsrv_proxyds(struct vnode *vp, off_t off, int cnt, struct ucred *cred,
struct thread *p, int ioproc, struct mbuf **mpp, char *cp,
struct mbuf **mpp2, struct nfsvattr *nap, struct acl *aclp,
off_t *offp, int content, bool *eofp)
{
struct nfsmount *nmp[NFSDEV_MAXMIRRORS], *failnmp;
fhandle_t fh[NFSDEV_MAXMIRRORS];
struct vnode *dvp[NFSDEV_MAXMIRRORS];
struct nfsdevice *ds;
struct pnfsdsattr dsattr;
struct opnfsdsattr odsattr;
char *buf;
int buflen, error, failpos, i, mirrorcnt, origmircnt, trycnt;
NFSD_DEBUG(4, "in nfsrv_proxyds\n");
/*
* If not a regular file, not exported or not a pNFS server,
* just return ENOENT.
*/
if (vp->v_type != VREG || (vp->v_mount->mnt_flag & MNT_EXPORTED) == 0 ||
nfsrv_devidcnt == 0)
return (ENOENT);
buflen = 1024;
buf = malloc(buflen, M_TEMP, M_WAITOK);
error = 0;
/*
* For Getattr, get the Change attribute (va_filerev) and size (va_size)
* from the MetaData file's extended attribute.
*/
if (ioproc == NFSPROC_GETATTR) {
error = vn_extattr_get(vp, IO_NODELOCKED,
EXTATTR_NAMESPACE_SYSTEM, "pnfsd.dsattr", &buflen, buf,
p);
if (error == 0) {
if (buflen == sizeof(odsattr)) {
NFSBCOPY(buf, &odsattr, buflen);
nap->na_filerev = odsattr.dsa_filerev;
nap->na_size = odsattr.dsa_size;
nap->na_atime = odsattr.dsa_atime;
nap->na_mtime = odsattr.dsa_mtime;
/*
* Fake na_bytes by rounding up na_size.
* Since we don't know the block size, just
* use BLKDEV_IOSIZE.
*/
nap->na_bytes = (odsattr.dsa_size +
BLKDEV_IOSIZE - 1) & ~(BLKDEV_IOSIZE - 1);
} else if (buflen == sizeof(dsattr)) {
NFSBCOPY(buf, &dsattr, buflen);
nap->na_filerev = dsattr.dsa_filerev;
nap->na_size = dsattr.dsa_size;
nap->na_atime = dsattr.dsa_atime;
nap->na_mtime = dsattr.dsa_mtime;
nap->na_bytes = dsattr.dsa_bytes;
} else
error = ENXIO;
}
if (error == 0) {
/*
* If nfsrv_pnfsgetdsattr is 0 or nfsrv_checkdsattr()
* returns 0, just return now. nfsrv_checkdsattr()
* returns 0 if there is no Read/Write layout
* plus either an Open/Write_access or Write
* delegation issued to a client for the file.
*/
if (nfsrv_pnfsgetdsattr == 0 ||
nfsrv_checkdsattr(vp, p) == 0) {
free(buf, M_TEMP);
return (error);
}
}
/*
* Clear ENOATTR so the code below will attempt to do a
* nfsrv_getattrdsrpc() to get the attributes and (re)create
* the extended attribute.
*/
if (error == ENOATTR)
error = 0;
}
origmircnt = -1;
trycnt = 0;
tryagain:
if (error == 0) {
buflen = 1024;
if (ioproc == NFSPROC_READDS && NFSVOPISLOCKED(vp) ==
LK_EXCLUSIVE)
printf("nfsrv_proxyds: Readds vp exclusively locked\n");
error = nfsrv_dsgetsockmnt(vp, LK_SHARED, buf, &buflen,
&mirrorcnt, p, dvp, fh, NULL, NULL, NULL, NULL, NULL,
NULL, NULL);
if (error == 0) {
for (i = 0; i < mirrorcnt; i++)
nmp[i] = VFSTONFS(dvp[i]->v_mount);
} else
printf("pNFS: proxy getextattr sockaddr=%d\n", error);
} else
printf("pNFS: nfsrv_dsgetsockmnt=%d\n", error);
if (error == 0) {
failpos = -1;
if (origmircnt == -1)
origmircnt = mirrorcnt;
/*
* If failpos is set to a mirror#, then that mirror has
* failed and will be disabled. For Read, Getattr and Seek, the
* function only tries one mirror, so if that mirror has
* failed, it will need to be retried. As such, increment
* tryitagain for these cases.
* For Write, Setattr and Setacl, the function tries all
* mirrors and will not return an error for the case where
* one mirror has failed. For these cases, the functioning
* mirror(s) will have been modified, so a retry isn't
* necessary. These functions will set failpos for the
* failed mirror#.
*/
if (ioproc == NFSPROC_READDS) {
error = nfsrv_readdsrpc(fh, off, cnt, cred, p, nmp[0],
mpp, mpp2);
if (nfsds_failerr(error) && mirrorcnt > 1) {
/*
* Setting failpos will cause the mirror
* to be disabled and then a retry of this
* read is required.
*/
failpos = 0;
error = 0;
trycnt++;
}
} else if (ioproc == NFSPROC_WRITEDS)
error = nfsrv_writedsrpc(fh, off, cnt, cred, p, vp,
&nmp[0], mirrorcnt, mpp, cp, &failpos);
else if (ioproc == NFSPROC_SETATTR)
error = nfsrv_setattrdsrpc(fh, cred, p, vp, &nmp[0],
mirrorcnt, nap, &failpos);
else if (ioproc == NFSPROC_SETACL)
error = nfsrv_setacldsrpc(fh, cred, p, vp, &nmp[0],
mirrorcnt, aclp, &failpos);
else if (ioproc == NFSPROC_SEEKDS) {
error = nfsrv_seekdsrpc(fh, offp, content, eofp, cred,
p, nmp[0]);
if (nfsds_failerr(error) && mirrorcnt > 1) {
/*
* Setting failpos will cause the mirror
* to be disabled and then a retry of this
* read is required.
*/
failpos = 0;
error = 0;
trycnt++;
}
} else if (ioproc == NFSPROC_ALLOCATE)
error = nfsrv_allocatedsrpc(fh, off, *offp, cred, p, vp,
&nmp[0], mirrorcnt, &failpos);
else if (ioproc == NFSPROC_DEALLOCATE)
error = nfsrv_deallocatedsrpc(fh, off, *offp, cred, p,
vp, &nmp[0], mirrorcnt, &failpos);
else {
error = nfsrv_getattrdsrpc(&fh[mirrorcnt - 1], cred, p,
vp, nmp[mirrorcnt - 1], nap);
if (nfsds_failerr(error) && mirrorcnt > 1) {
/*
* Setting failpos will cause the mirror
* to be disabled and then a retry of this
* getattr is required.
*/
failpos = mirrorcnt - 1;
error = 0;
trycnt++;
}
}
ds = NULL;
if (failpos >= 0) {
failnmp = nmp[failpos];
NFSLOCKMNT(failnmp);
if ((failnmp->nm_privflag & (NFSMNTP_FORCEDISM |
NFSMNTP_CANCELRPCS)) == 0) {
failnmp->nm_privflag |= NFSMNTP_CANCELRPCS;
NFSUNLOCKMNT(failnmp);
ds = nfsrv_deldsnmp(PNFSDOP_DELDSSERVER,
failnmp, p);
NFSD_DEBUG(4, "dsldsnmp fail=%d ds=%p\n",
failpos, ds);
if (ds != NULL)
nfsrv_killrpcs(failnmp);
NFSLOCKMNT(failnmp);
failnmp->nm_privflag &= ~NFSMNTP_CANCELRPCS;
wakeup(failnmp);
}
NFSUNLOCKMNT(failnmp);
}
for (i = 0; i < mirrorcnt; i++)
NFSVOPUNLOCK(dvp[i]);
NFSD_DEBUG(4, "nfsrv_proxyds: aft RPC=%d trya=%d\n", error,
trycnt);
/* Try the Read/Getattr again if a mirror was deleted. */
if (ds != NULL && trycnt > 0 && trycnt < origmircnt)
goto tryagain;
} else {
/* Return ENOENT for any Extended Attribute error. */
error = ENOENT;
}
free(buf, M_TEMP);
NFSD_DEBUG(4, "nfsrv_proxyds: error=%d\n", error);
return (error);
}
/*
* Get the DS mount point, fh and directory from the "pnfsd.dsfile" extended
* attribute.
* newnmpp - If it points to a non-NULL nmp, that is the destination and needs
* to be checked. If it points to a NULL nmp, then it returns
* a suitable destination.
* curnmp - If non-NULL, it is the source mount for the copy.
*/
int
nfsrv_dsgetsockmnt(struct vnode *vp, int lktype, char *buf, int *buflenp,
int *mirrorcntp, NFSPROC_T *p, struct vnode **dvpp, fhandle_t *fhp,
char *devid, char *fnamep, struct vnode **nvpp, struct nfsmount **newnmpp,
struct nfsmount *curnmp, int *ippos, int *dsdirp)
{
struct vnode *dvp, *nvp = NULL, **tdvpp;
struct mount *mp;
struct nfsmount *nmp, *newnmp;
struct sockaddr *sad;
struct sockaddr_in *sin;
struct nfsdevice *ds, *tds, *fndds;
struct pnfsdsfile *pf;
uint32_t dsdir;
int error, fhiszero, fnd, gotone, i, mirrorcnt;
ASSERT_VOP_LOCKED(vp, "nfsrv_dsgetsockmnt vp");
*mirrorcntp = 1;
tdvpp = dvpp;
if (nvpp != NULL)
*nvpp = NULL;
if (dvpp != NULL)
*dvpp = NULL;
if (ippos != NULL)
*ippos = -1;
if (newnmpp != NULL)
newnmp = *newnmpp;
else
newnmp = NULL;
mp = vp->v_mount;
error = vn_extattr_get(vp, IO_NODELOCKED, EXTATTR_NAMESPACE_SYSTEM,
"pnfsd.dsfile", buflenp, buf, p);
mirrorcnt = *buflenp / sizeof(*pf);
if (error == 0 && (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS ||
*buflenp != sizeof(*pf) * mirrorcnt))
error = ENOATTR;
pf = (struct pnfsdsfile *)buf;
/* If curnmp != NULL, check for a match in the mirror list. */
if (curnmp != NULL && error == 0) {
fnd = 0;
for (i = 0; i < mirrorcnt; i++, pf++) {
sad = (struct sockaddr *)&pf->dsf_sin;
if (nfsaddr2_match(sad, curnmp->nm_nam)) {
if (ippos != NULL)
*ippos = i;
fnd = 1;
break;
}
}
if (fnd == 0)
error = ENXIO;
}
gotone = 0;
pf = (struct pnfsdsfile *)buf;
NFSD_DEBUG(4, "nfsrv_dsgetsockmnt: mirrorcnt=%d err=%d\n", mirrorcnt,
error);
for (i = 0; i < mirrorcnt && error == 0; i++, pf++) {
fhiszero = 0;
sad = (struct sockaddr *)&pf->dsf_sin;
sin = &pf->dsf_sin;
dsdir = pf->dsf_dir;
if (dsdir >= nfsrv_dsdirsize) {
printf("nfsrv_dsgetsockmnt: dsdir=%d\n", dsdir);
error = ENOATTR;
} else if (nvpp != NULL && newnmp != NULL &&
nfsaddr2_match(sad, newnmp->nm_nam))
error = EEXIST;
if (error == 0) {
if (ippos != NULL && curnmp == NULL &&
sad->sa_family == AF_INET &&
sin->sin_addr.s_addr == 0)
*ippos = i;
if (NFSBCMP(&zerofh, &pf->dsf_fh, sizeof(zerofh)) == 0)
fhiszero = 1;
/* Use the socket address to find the mount point. */
fndds = NULL;
NFSDDSLOCK();
/* Find a match for the IP address. */
TAILQ_FOREACH(ds, &nfsrv_devidhead, nfsdev_list) {
if (ds->nfsdev_nmp != NULL) {
dvp = ds->nfsdev_dvp;
nmp = VFSTONFS(dvp->v_mount);
if (nmp != ds->nfsdev_nmp)
printf("different2 nmp %p %p\n",
nmp, ds->nfsdev_nmp);
if (nfsaddr2_match(sad, nmp->nm_nam)) {
fndds = ds;
break;
}
}
}
if (fndds != NULL && newnmpp != NULL &&
newnmp == NULL) {
/* Search for a place to make a mirror copy. */
TAILQ_FOREACH(tds, &nfsrv_devidhead,
nfsdev_list) {
if (tds->nfsdev_nmp != NULL &&
fndds != tds &&
((tds->nfsdev_mdsisset == 0 &&
fndds->nfsdev_mdsisset == 0) ||
(tds->nfsdev_mdsisset != 0 &&
fndds->nfsdev_mdsisset != 0 &&
fsidcmp(&tds->nfsdev_mdsfsid,
&mp->mnt_stat.f_fsid) == 0))) {
*newnmpp = tds->nfsdev_nmp;
break;
}
}
if (tds != NULL) {
/*
* Move this entry to the end of the
* list, so it won't be selected as
* easily the next time.
*/
TAILQ_REMOVE(&nfsrv_devidhead, tds,
nfsdev_list);
TAILQ_INSERT_TAIL(&nfsrv_devidhead, tds,
nfsdev_list);
}
}
NFSDDSUNLOCK();
if (fndds != NULL) {
dvp = fndds->nfsdev_dsdir[dsdir];
if (lktype != 0 || fhiszero != 0 ||
(nvpp != NULL && *nvpp == NULL)) {
if (fhiszero != 0)
error = vn_lock(dvp,
LK_EXCLUSIVE);
else if (lktype != 0)
error = vn_lock(dvp, lktype);
else
error = vn_lock(dvp, LK_SHARED);
/*
* If the file handle is all 0's, try to
* do a Lookup against the DS to acquire
* it.
* If dvpp == NULL or the Lookup fails,
* unlock dvp after the call.
*/
if (error == 0 && (fhiszero != 0 ||
(nvpp != NULL && *nvpp == NULL))) {
error = nfsrv_pnfslookupds(vp,
dvp, pf, &nvp, p);
if (error == 0) {
if (fhiszero != 0)
nfsrv_pnfssetfh(
vp, pf,
devid,
fnamep,
nvp, p);
if (nvpp != NULL &&
*nvpp == NULL) {
*nvpp = nvp;
*dsdirp = dsdir;
} else
vput(nvp);
}
if (error != 0 || lktype == 0)
NFSVOPUNLOCK(dvp);
}
}
if (error == 0) {
gotone++;
NFSD_DEBUG(4, "gotone=%d\n", gotone);
if (devid != NULL) {
NFSBCOPY(fndds->nfsdev_deviceid,
devid, NFSX_V4DEVICEID);
devid += NFSX_V4DEVICEID;
}
if (dvpp != NULL)
*tdvpp++ = dvp;
if (fhp != NULL)
NFSBCOPY(&pf->dsf_fh, fhp++,
NFSX_MYFH);
if (fnamep != NULL && gotone == 1)
strlcpy(fnamep,
pf->dsf_filename,
sizeof(pf->dsf_filename));
} else
NFSD_DEBUG(4, "nfsrv_dsgetsockmnt "
"err=%d\n", error);
}
}
}
if (error == 0 && gotone == 0)
error = ENOENT;
NFSD_DEBUG(4, "eo nfsrv_dsgetsockmnt: gotone=%d err=%d\n", gotone,
error);
if (error == 0)
*mirrorcntp = gotone;
else {
if (gotone > 0 && dvpp != NULL) {
/*
* If the error didn't occur on the first one and
* dvpp != NULL, the one(s) prior to the failure will
* have locked dvp's that need to be unlocked.
*/
for (i = 0; i < gotone; i++) {
NFSVOPUNLOCK(*dvpp);
*dvpp++ = NULL;
}
}
/*
* If it found the vnode to be copied from before a failure,
* it needs to be vput()'d.
*/
if (nvpp != NULL && *nvpp != NULL) {
vput(*nvpp);
*nvpp = NULL;
}
}
return (error);
}
/*
* Set the extended attribute for the Change attribute.
*/
static int
nfsrv_setextattr(struct vnode *vp, struct nfsvattr *nap, NFSPROC_T *p)
{
struct pnfsdsattr dsattr;
int error;
ASSERT_VOP_ELOCKED(vp, "nfsrv_setextattr vp");
dsattr.dsa_filerev = nap->na_filerev;
dsattr.dsa_size = nap->na_size;
dsattr.dsa_atime = nap->na_atime;
dsattr.dsa_mtime = nap->na_mtime;
dsattr.dsa_bytes = nap->na_bytes;
error = vn_extattr_set(vp, IO_NODELOCKED, EXTATTR_NAMESPACE_SYSTEM,
"pnfsd.dsattr", sizeof(dsattr), (char *)&dsattr, p);
if (error != 0)
printf("pNFS: setextattr=%d\n", error);
return (error);
}
static int
nfsrv_readdsrpc(fhandle_t *fhp, off_t off, int len, struct ucred *cred,
NFSPROC_T *p, struct nfsmount *nmp, struct mbuf **mpp, struct mbuf **mpendp)
{
uint32_t *tl;
struct nfsrv_descript *nd;
nfsv4stateid_t st;
struct mbuf *m, *m2;
int error = 0, retlen, tlen, trimlen;
NFSD_DEBUG(4, "in nfsrv_readdsrpc\n");
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
*mpp = NULL;
/*
* Use a stateid where other is an alternating 01010 pattern and
* seqid is 0xffffffff. This value is not defined as special by
* the RFC and is used by the FreeBSD NFS server to indicate an
* MDS->DS proxy operation.
*/
st.other[0] = 0x55555555;
st.other[1] = 0x55555555;
st.other[2] = 0x55555555;
st.seqid = 0xffffffff;
nfscl_reqstart(nd, NFSPROC_READDS, nmp, (u_int8_t *)fhp, sizeof(*fhp),
NULL, NULL, 0, 0);
nfsm_stateidtom(nd, &st, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
txdr_hyper(off, tl);
*(tl + 2) = txdr_unsigned(len);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
NFSM_STRSIZ(retlen, len);
if (retlen > 0) {
/* Trim off the pre-data XDR from the mbuf chain. */
m = nd->nd_mrep;
while (m != NULL && m != nd->nd_md) {
if (m->m_next == nd->nd_md) {
m->m_next = NULL;
m_freem(nd->nd_mrep);
nd->nd_mrep = m = nd->nd_md;
} else
m = m->m_next;
}
if (m == NULL) {
printf("nfsrv_readdsrpc: busted mbuf list\n");
error = ENOENT;
goto nfsmout;
}
/*
* Now, adjust first mbuf so that any XDR before the
* read data is skipped over.
*/
trimlen = nd->nd_dpos - mtod(m, char *);
if (trimlen > 0) {
m->m_len -= trimlen;
NFSM_DATAP(m, trimlen);
}
/*
* Truncate the mbuf chain at retlen bytes of data,
* plus XDR padding that brings the length up to a
* multiple of 4.
*/
tlen = NFSM_RNDUP(retlen);
do {
if (m->m_len >= tlen) {
m->m_len = tlen;
tlen = 0;
m2 = m->m_next;
m->m_next = NULL;
m_freem(m2);
break;
}
tlen -= m->m_len;
m = m->m_next;
} while (m != NULL);
if (tlen > 0) {
printf("nfsrv_readdsrpc: busted mbuf list\n");
error = ENOENT;
goto nfsmout;
}
*mpp = nd->nd_mrep;
*mpendp = m;
nd->nd_mrep = NULL;
}
} else
error = nd->nd_repstat;
nfsmout:
/* If nd->nd_mrep is already NULL, this is a no-op. */
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
NFSD_DEBUG(4, "nfsrv_readdsrpc error=%d\n", error);
return (error);
}
/*
* Do a write RPC on a DS data file, using this structure for the arguments,
* so that this function can be executed by a separate kernel process.
*/
struct nfsrvwritedsdorpc {
int done;
int inprog;
struct task tsk;
fhandle_t fh;
off_t off;
int len;
struct nfsmount *nmp;
struct ucred *cred;
NFSPROC_T *p;
struct mbuf *m;
int err;
};
static int
nfsrv_writedsdorpc(struct nfsmount *nmp, fhandle_t *fhp, off_t off, int len,
struct nfsvattr *nap, struct mbuf *m, struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
struct nfsrv_descript *nd;
nfsattrbit_t attrbits;
nfsv4stateid_t st;
int commit, error, retlen;
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
nfscl_reqstart(nd, NFSPROC_WRITE, nmp, (u_int8_t *)fhp,
sizeof(fhandle_t), NULL, NULL, 0, 0);
/*
* Use a stateid where other is an alternating 01010 pattern and
* seqid is 0xffffffff. This value is not defined as special by
* the RFC and is used by the FreeBSD NFS server to indicate an
* MDS->DS proxy operation.
*/
st.other[0] = 0x55555555;
st.other[1] = 0x55555555;
st.other[2] = 0x55555555;
st.seqid = 0xffffffff;
nfsm_stateidtom(nd, &st, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
txdr_hyper(off, tl);
tl += 2;
/*
* Do all writes FileSync, since the server doesn't hold onto dirty
* buffers. Since clients should be accessing the DS servers directly
* using the pNFS layouts, this just needs to work correctly as a
* fallback.
*/
*tl++ = txdr_unsigned(NFSWRITE_FILESYNC);
*tl = txdr_unsigned(len);
NFSD_DEBUG(4, "nfsrv_writedsdorpc: len=%d\n", len);
/* Put data in mbuf chain. */
nd->nd_mb->m_next = m;
/* Set nd_mb and nd_bpos to end of data. */
while (m->m_next != NULL)
m = m->m_next;
nd->nd_mb = m;
nfsm_set(nd, m->m_len);
NFSD_DEBUG(4, "nfsrv_writedsdorpc: lastmb len=%d\n", m->m_len);
/* Do a Getattr for the attributes that change upon writing. */
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SIZE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEACCESS);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SPACEUSED);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &attrbits);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
NFSD_DEBUG(4, "nfsrv_writedsdorpc: aft writerpc=%d\n", nd->nd_repstat);
/* Get rid of weak cache consistency data for now. */
if ((nd->nd_flag & (ND_NOMOREDATA | ND_NFSV4 | ND_V4WCCATTR)) ==
(ND_NFSV4 | ND_V4WCCATTR)) {
error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0, NULL, NULL,
NULL, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL);
NFSD_DEBUG(4, "nfsrv_writedsdorpc: wcc attr=%d\n", error);
if (error != 0)
goto nfsmout;
/*
* Get rid of Op# and status for next op.
*/
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (*++tl != 0)
nd->nd_flag |= ND_NOMOREDATA;
}
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
retlen = fxdr_unsigned(int, *tl++);
commit = fxdr_unsigned(int, *tl);
if (commit != NFSWRITE_FILESYNC)
error = NFSERR_IO;
NFSD_DEBUG(4, "nfsrv_writedsdorpc:retlen=%d commit=%d err=%d\n",
retlen, commit, error);
} else
error = nd->nd_repstat;
/* We have no use for the Write Verifier since we use FileSync. */
/*
* Get the Change, Size, Access Time and Modify Time attributes and set
* on the Metadata file, so its attributes will be what the file's
* would be if it had been written.
*/
if (error == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0, NULL, NULL,
NULL, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL);
}
NFSD_DEBUG(4, "nfsrv_writedsdorpc: aft loadattr=%d\n", error);
nfsmout:
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
NFSD_DEBUG(4, "nfsrv_writedsdorpc error=%d\n", error);
return (error);
}
/*
* Start up the thread that will execute nfsrv_writedsdorpc().
*/
static void
start_writedsdorpc(void *arg, int pending)
{
struct nfsrvwritedsdorpc *drpc;
drpc = (struct nfsrvwritedsdorpc *)arg;
drpc->err = nfsrv_writedsdorpc(drpc->nmp, &drpc->fh, drpc->off,
drpc->len, NULL, drpc->m, drpc->cred, drpc->p);
drpc->done = 1;
NFSD_DEBUG(4, "start_writedsdorpc: err=%d\n", drpc->err);
}
static int
nfsrv_writedsrpc(fhandle_t *fhp, off_t off, int len, struct ucred *cred,
NFSPROC_T *p, struct vnode *vp, struct nfsmount **nmpp, int mirrorcnt,
struct mbuf **mpp, char *cp, int *failposp)
{
struct nfsrvwritedsdorpc *drpc, *tdrpc = NULL;
struct nfsvattr na;
struct mbuf *m;
int error, i, offs, ret, timo;
NFSD_DEBUG(4, "in nfsrv_writedsrpc\n");
KASSERT(*mpp != NULL, ("nfsrv_writedsrpc: NULL mbuf chain"));
drpc = NULL;
if (mirrorcnt > 1)
tdrpc = drpc = malloc(sizeof(*drpc) * (mirrorcnt - 1), M_TEMP,
M_WAITOK);
/* Calculate offset in mbuf chain that data starts. */
offs = cp - mtod(*mpp, char *);
NFSD_DEBUG(4, "nfsrv_writedsrpc: mcopy offs=%d len=%d\n", offs, len);
/*
* Do the write RPC for every DS, using a separate kernel process
* for every DS except the last one.
*/
error = 0;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
tdrpc->done = 0;
NFSBCOPY(fhp, &tdrpc->fh, sizeof(*fhp));
tdrpc->off = off;
tdrpc->len = len;
tdrpc->nmp = *nmpp;
tdrpc->cred = cred;
tdrpc->p = p;
tdrpc->inprog = 0;
tdrpc->err = 0;
tdrpc->m = m_copym(*mpp, offs, NFSM_RNDUP(len), M_WAITOK);
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_writedsdorpc, tdrpc);
NFSD_DEBUG(4, "nfsrv_writedsrpc: nfs_pnfsio=%d\n",
ret);
}
if (ret != 0) {
ret = nfsrv_writedsdorpc(*nmpp, fhp, off, len, NULL,
tdrpc->m, cred, p);
if (nfsds_failerr(ret) && *failposp == -1)
*failposp = i;
else if (error == 0 && ret != 0)
error = ret;
}
nmpp++;
fhp++;
}
m = m_copym(*mpp, offs, NFSM_RNDUP(len), M_WAITOK);
ret = nfsrv_writedsdorpc(*nmpp, fhp, off, len, &na, m, cred, p);
if (nfsds_failerr(ret) && *failposp == -1 && mirrorcnt > 1)
*failposp = mirrorcnt - 1;
else if (error == 0 && ret != 0)
error = ret;
if (error == 0)
error = nfsrv_setextattr(vp, &na, p);
NFSD_DEBUG(4, "nfsrv_writedsrpc: aft setextat=%d\n", error);
tdrpc = drpc;
timo = hz / 50; /* Wait for 20msec. */
if (timo < 1)
timo = 1;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
/* Wait for RPCs on separate threads to complete. */
while (tdrpc->inprog != 0 && tdrpc->done == 0)
tsleep(&tdrpc->tsk, PVFS, "srvwrds", timo);
if (nfsds_failerr(tdrpc->err) && *failposp == -1)
*failposp = i;
else if (error == 0 && tdrpc->err != 0)
error = tdrpc->err;
}
free(drpc, M_TEMP);
return (error);
}
/*
* Do a allocate RPC on a DS data file, using this structure for the arguments,
* so that this function can be executed by a separate kernel process.
*/
struct nfsrvallocatedsdorpc {
int done;
int inprog;
struct task tsk;
fhandle_t fh;
off_t off;
off_t len;
struct nfsmount *nmp;
struct ucred *cred;
NFSPROC_T *p;
int err;
};
static int
nfsrv_allocatedsdorpc(struct nfsmount *nmp, fhandle_t *fhp, off_t off,
off_t len, struct nfsvattr *nap, struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
struct nfsrv_descript *nd;
nfsattrbit_t attrbits;
nfsv4stateid_t st;
int error;
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
nfscl_reqstart(nd, NFSPROC_ALLOCATE, nmp, (u_int8_t *)fhp,
sizeof(fhandle_t), NULL, NULL, 0, 0);
/*
* Use a stateid where other is an alternating 01010 pattern and
* seqid is 0xffffffff. This value is not defined as special by
* the RFC and is used by the FreeBSD NFS server to indicate an
* MDS->DS proxy operation.
*/
st.other[0] = 0x55555555;
st.other[1] = 0x55555555;
st.other[2] = 0x55555555;
st.seqid = 0xffffffff;
nfsm_stateidtom(nd, &st, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
txdr_hyper(off, tl); tl += 2;
txdr_hyper(len, tl); tl += 2;
NFSD_DEBUG(4, "nfsrv_allocatedsdorpc: len=%jd\n", (intmax_t)len);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
nfsrv_putattrbit(nd, &attrbits);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
NFSD_DEBUG(4, "nfsrv_allocatedsdorpc: aft allocaterpc=%d\n",
nd->nd_repstat);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0, NULL, NULL,
NULL, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL);
} else
error = nd->nd_repstat;
NFSD_DEBUG(4, "nfsrv_allocatedsdorpc: aft loadattr=%d\n", error);
nfsmout:
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
NFSD_DEBUG(4, "nfsrv_allocatedsdorpc error=%d\n", error);
return (error);
}
/*
* Start up the thread that will execute nfsrv_allocatedsdorpc().
*/
static void
start_allocatedsdorpc(void *arg, int pending)
{
struct nfsrvallocatedsdorpc *drpc;
drpc = (struct nfsrvallocatedsdorpc *)arg;
drpc->err = nfsrv_allocatedsdorpc(drpc->nmp, &drpc->fh, drpc->off,
drpc->len, NULL, drpc->cred, drpc->p);
drpc->done = 1;
NFSD_DEBUG(4, "start_allocatedsdorpc: err=%d\n", drpc->err);
}
static int
nfsrv_allocatedsrpc(fhandle_t *fhp, off_t off, off_t len, struct ucred *cred,
NFSPROC_T *p, struct vnode *vp, struct nfsmount **nmpp, int mirrorcnt,
int *failposp)
{
struct nfsrvallocatedsdorpc *drpc, *tdrpc = NULL;
struct nfsvattr na;
int error, i, ret, timo;
NFSD_DEBUG(4, "in nfsrv_allocatedsrpc\n");
drpc = NULL;
if (mirrorcnt > 1)
tdrpc = drpc = malloc(sizeof(*drpc) * (mirrorcnt - 1), M_TEMP,
M_WAITOK);
/*
* Do the allocate RPC for every DS, using a separate kernel process
* for every DS except the last one.
*/
error = 0;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
tdrpc->done = 0;
NFSBCOPY(fhp, &tdrpc->fh, sizeof(*fhp));
tdrpc->off = off;
tdrpc->len = len;
tdrpc->nmp = *nmpp;
tdrpc->cred = cred;
tdrpc->p = p;
tdrpc->inprog = 0;
tdrpc->err = 0;
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_allocatedsdorpc, tdrpc);
NFSD_DEBUG(4, "nfsrv_allocatedsrpc: nfs_pnfsio=%d\n",
ret);
}
if (ret != 0) {
ret = nfsrv_allocatedsdorpc(*nmpp, fhp, off, len, NULL,
cred, p);
if (nfsds_failerr(ret) && *failposp == -1)
*failposp = i;
else if (error == 0 && ret != 0)
error = ret;
}
nmpp++;
fhp++;
}
ret = nfsrv_allocatedsdorpc(*nmpp, fhp, off, len, &na, cred, p);
if (nfsds_failerr(ret) && *failposp == -1 && mirrorcnt > 1)
*failposp = mirrorcnt - 1;
else if (error == 0 && ret != 0)
error = ret;
if (error == 0)
error = nfsrv_setextattr(vp, &na, p);
NFSD_DEBUG(4, "nfsrv_allocatedsrpc: aft setextat=%d\n", error);
tdrpc = drpc;
timo = hz / 50; /* Wait for 20msec. */
if (timo < 1)
timo = 1;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
/* Wait for RPCs on separate threads to complete. */
while (tdrpc->inprog != 0 && tdrpc->done == 0)
tsleep(&tdrpc->tsk, PVFS, "srvalds", timo);
if (nfsds_failerr(tdrpc->err) && *failposp == -1)
*failposp = i;
else if (error == 0 && tdrpc->err != 0)
error = tdrpc->err;
}
free(drpc, M_TEMP);
return (error);
}
/*
* Do a deallocate RPC on a DS data file, using this structure for the
* arguments, so that this function can be executed by a separate kernel
* process.
*/
struct nfsrvdeallocatedsdorpc {
int done;
int inprog;
struct task tsk;
fhandle_t fh;
off_t off;
off_t len;
struct nfsmount *nmp;
struct ucred *cred;
NFSPROC_T *p;
int err;
};
static int
nfsrv_deallocatedsdorpc(struct nfsmount *nmp, fhandle_t *fhp, off_t off,
off_t len, struct nfsvattr *nap, struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
struct nfsrv_descript *nd;
nfsattrbit_t attrbits;
nfsv4stateid_t st;
int error;
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
nfscl_reqstart(nd, NFSPROC_DEALLOCATE, nmp, (u_int8_t *)fhp,
sizeof(fhandle_t), NULL, NULL, 0, 0);
/*
* Use a stateid where other is an alternating 01010 pattern and
* seqid is 0xffffffff. This value is not defined as special by
* the RFC and is used by the FreeBSD NFS server to indicate an
* MDS->DS proxy operation.
*/
st.other[0] = 0x55555555;
st.other[1] = 0x55555555;
st.other[2] = 0x55555555;
st.seqid = 0xffffffff;
nfsm_stateidtom(nd, &st, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
txdr_hyper(off, tl); tl += 2;
txdr_hyper(len, tl); tl += 2;
NFSD_DEBUG(4, "nfsrv_deallocatedsdorpc: len=%jd\n", (intmax_t)len);
/* Do a Getattr for the attributes that change upon writing. */
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SIZE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEACCESS);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SPACEUSED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
nfsrv_putattrbit(nd, &attrbits);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
NFSD_DEBUG(4, "nfsrv_deallocatedsdorpc: aft deallocaterpc=%d\n",
nd->nd_repstat);
/* Get rid of weak cache consistency data for now. */
if ((nd->nd_flag & (ND_NOMOREDATA | ND_NFSV4 | ND_V4WCCATTR)) ==
(ND_NFSV4 | ND_V4WCCATTR)) {
error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0, NULL, NULL,
NULL, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL);
NFSD_DEBUG(4, "nfsrv_deallocatedsdorpc: wcc attr=%d\n", error);
if (error != 0)
goto nfsmout;
/*
* Get rid of Op# and status for next op.
*/
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (*++tl != 0)
nd->nd_flag |= ND_NOMOREDATA;
}
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0, NULL, NULL,
NULL, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL);
} else
error = nd->nd_repstat;
NFSD_DEBUG(4, "nfsrv_deallocatedsdorpc: aft loadattr=%d\n", error);
nfsmout:
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
NFSD_DEBUG(4, "nfsrv_deallocatedsdorpc error=%d\n", error);
return (error);
}
/*
* Start up the thread that will execute nfsrv_deallocatedsdorpc().
*/
static void
start_deallocatedsdorpc(void *arg, int pending)
{
struct nfsrvdeallocatedsdorpc *drpc;
drpc = (struct nfsrvdeallocatedsdorpc *)arg;
drpc->err = nfsrv_deallocatedsdorpc(drpc->nmp, &drpc->fh, drpc->off,
drpc->len, NULL, drpc->cred, drpc->p);
drpc->done = 1;
NFSD_DEBUG(4, "start_deallocatedsdorpc: err=%d\n", drpc->err);
}
static int
nfsrv_deallocatedsrpc(fhandle_t *fhp, off_t off, off_t len, struct ucred *cred,
NFSPROC_T *p, struct vnode *vp, struct nfsmount **nmpp, int mirrorcnt,
int *failposp)
{
struct nfsrvdeallocatedsdorpc *drpc, *tdrpc = NULL;
struct nfsvattr na;
int error, i, ret, timo;
NFSD_DEBUG(4, "in nfsrv_deallocatedsrpc\n");
drpc = NULL;
if (mirrorcnt > 1)
tdrpc = drpc = malloc(sizeof(*drpc) * (mirrorcnt - 1), M_TEMP,
M_WAITOK);
/*
* Do the deallocate RPC for every DS, using a separate kernel process
* for every DS except the last one.
*/
error = 0;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
tdrpc->done = 0;
NFSBCOPY(fhp, &tdrpc->fh, sizeof(*fhp));
tdrpc->off = off;
tdrpc->len = len;
tdrpc->nmp = *nmpp;
tdrpc->cred = cred;
tdrpc->p = p;
tdrpc->inprog = 0;
tdrpc->err = 0;
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_deallocatedsdorpc, tdrpc);
NFSD_DEBUG(4, "nfsrv_deallocatedsrpc: nfs_pnfsio=%d\n",
ret);
}
if (ret != 0) {
ret = nfsrv_deallocatedsdorpc(*nmpp, fhp, off, len,
NULL, cred, p);
if (nfsds_failerr(ret) && *failposp == -1)
*failposp = i;
else if (error == 0 && ret != 0)
error = ret;
}
nmpp++;
fhp++;
}
ret = nfsrv_deallocatedsdorpc(*nmpp, fhp, off, len, &na, cred, p);
if (nfsds_failerr(ret) && *failposp == -1 && mirrorcnt > 1)
*failposp = mirrorcnt - 1;
else if (error == 0 && ret != 0)
error = ret;
if (error == 0)
error = nfsrv_setextattr(vp, &na, p);
NFSD_DEBUG(4, "nfsrv_deallocatedsrpc: aft setextat=%d\n", error);
tdrpc = drpc;
timo = hz / 50; /* Wait for 20msec. */
if (timo < 1)
timo = 1;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
/* Wait for RPCs on separate threads to complete. */
while (tdrpc->inprog != 0 && tdrpc->done == 0)
tsleep(&tdrpc->tsk, PVFS, "srvalds", timo);
if (nfsds_failerr(tdrpc->err) && *failposp == -1)
*failposp = i;
else if (error == 0 && tdrpc->err != 0)
error = tdrpc->err;
}
free(drpc, M_TEMP);
return (error);
}
static int
nfsrv_setattrdsdorpc(fhandle_t *fhp, struct ucred *cred, NFSPROC_T *p,
struct vnode *vp, struct nfsmount *nmp, struct nfsvattr *nap,
struct nfsvattr *dsnap)
{
uint32_t *tl;
struct nfsrv_descript *nd;
nfsv4stateid_t st;
nfsattrbit_t attrbits;
int error;
NFSD_DEBUG(4, "in nfsrv_setattrdsdorpc\n");
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
/*
* Use a stateid where other is an alternating 01010 pattern and
* seqid is 0xffffffff. This value is not defined as special by
* the RFC and is used by the FreeBSD NFS server to indicate an
* MDS->DS proxy operation.
*/
st.other[0] = 0x55555555;
st.other[1] = 0x55555555;
st.other[2] = 0x55555555;
st.seqid = 0xffffffff;
nfscl_reqstart(nd, NFSPROC_SETATTR, nmp, (u_int8_t *)fhp, sizeof(*fhp),
NULL, NULL, 0, 0);
nfsm_stateidtom(nd, &st, NFSSTATEID_PUTSTATEID);
nfscl_fillsattr(nd, &nap->na_vattr, vp, NFSSATTR_FULL, 0);
/* Do a Getattr for the attributes that change due to writing. */
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SIZE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEACCESS);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SPACEUSED);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &attrbits);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
NFSD_DEBUG(4, "nfsrv_setattrdsdorpc: aft setattrrpc=%d\n",
nd->nd_repstat);
/* Get rid of weak cache consistency data for now. */
if ((nd->nd_flag & (ND_NOMOREDATA | ND_NFSV4 | ND_V4WCCATTR)) ==
(ND_NFSV4 | ND_V4WCCATTR)) {
error = nfsv4_loadattr(nd, NULL, dsnap, NULL, NULL, 0, NULL,
NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL);
NFSD_DEBUG(4, "nfsrv_setattrdsdorpc: wcc attr=%d\n", error);
if (error != 0)
goto nfsmout;
/*
* Get rid of Op# and status for next op.
*/
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (*++tl != 0)
nd->nd_flag |= ND_NOMOREDATA;
}
error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
if (error != 0)
goto nfsmout;
if (nd->nd_repstat != 0)
error = nd->nd_repstat;
/*
* Get the Change, Size, Access Time and Modify Time attributes and set
* on the Metadata file, so its attributes will be what the file's
* would be if it had been written.
*/
if (error == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
error = nfsv4_loadattr(nd, NULL, dsnap, NULL, NULL, 0, NULL,
NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL);
}
NFSD_DEBUG(4, "nfsrv_setattrdsdorpc: aft setattr loadattr=%d\n", error);
nfsmout:
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
NFSD_DEBUG(4, "nfsrv_setattrdsdorpc error=%d\n", error);
return (error);
}
struct nfsrvsetattrdsdorpc {
int done;
int inprog;
struct task tsk;
fhandle_t fh;
struct nfsmount *nmp;
struct vnode *vp;
struct ucred *cred;
NFSPROC_T *p;
struct nfsvattr na;
struct nfsvattr dsna;
int err;
};
/*
* Start up the thread that will execute nfsrv_setattrdsdorpc().
*/
static void
start_setattrdsdorpc(void *arg, int pending)
{
struct nfsrvsetattrdsdorpc *drpc;
drpc = (struct nfsrvsetattrdsdorpc *)arg;
drpc->err = nfsrv_setattrdsdorpc(&drpc->fh, drpc->cred, drpc->p,
drpc->vp, drpc->nmp, &drpc->na, &drpc->dsna);
drpc->done = 1;
}
static int
nfsrv_setattrdsrpc(fhandle_t *fhp, struct ucred *cred, NFSPROC_T *p,
struct vnode *vp, struct nfsmount **nmpp, int mirrorcnt,
struct nfsvattr *nap, int *failposp)
{
struct nfsrvsetattrdsdorpc *drpc, *tdrpc = NULL;
struct nfsvattr na;
int error, i, ret, timo;
NFSD_DEBUG(4, "in nfsrv_setattrdsrpc\n");
drpc = NULL;
if (mirrorcnt > 1)
tdrpc = drpc = malloc(sizeof(*drpc) * (mirrorcnt - 1), M_TEMP,
M_WAITOK);
/*
* Do the setattr RPC for every DS, using a separate kernel process
* for every DS except the last one.
*/
error = 0;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
tdrpc->done = 0;
tdrpc->inprog = 0;
NFSBCOPY(fhp, &tdrpc->fh, sizeof(*fhp));
tdrpc->nmp = *nmpp;
tdrpc->vp = vp;
tdrpc->cred = cred;
tdrpc->p = p;
tdrpc->na = *nap;
tdrpc->err = 0;
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_setattrdsdorpc, tdrpc);
NFSD_DEBUG(4, "nfsrv_setattrdsrpc: nfs_pnfsio=%d\n",
ret);
}
if (ret != 0) {
ret = nfsrv_setattrdsdorpc(fhp, cred, p, vp, *nmpp, nap,
&na);
if (nfsds_failerr(ret) && *failposp == -1)
*failposp = i;
else if (error == 0 && ret != 0)
error = ret;
}
nmpp++;
fhp++;
}
ret = nfsrv_setattrdsdorpc(fhp, cred, p, vp, *nmpp, nap, &na);
if (nfsds_failerr(ret) && *failposp == -1 && mirrorcnt > 1)
*failposp = mirrorcnt - 1;
else if (error == 0 && ret != 0)
error = ret;
if (error == 0)
error = nfsrv_setextattr(vp, &na, p);
NFSD_DEBUG(4, "nfsrv_setattrdsrpc: aft setextat=%d\n", error);
tdrpc = drpc;
timo = hz / 50; /* Wait for 20msec. */
if (timo < 1)
timo = 1;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
/* Wait for RPCs on separate threads to complete. */
while (tdrpc->inprog != 0 && tdrpc->done == 0)
tsleep(&tdrpc->tsk, PVFS, "srvsads", timo);
if (nfsds_failerr(tdrpc->err) && *failposp == -1)
*failposp = i;
else if (error == 0 && tdrpc->err != 0)
error = tdrpc->err;
}
free(drpc, M_TEMP);
return (error);
}
/*
* Do a Setattr of an NFSv4 ACL on the DS file.
*/
static int
nfsrv_setacldsdorpc(fhandle_t *fhp, struct ucred *cred, NFSPROC_T *p,
struct vnode *vp, struct nfsmount *nmp, struct acl *aclp)
{
struct nfsrv_descript *nd;
nfsv4stateid_t st;
nfsattrbit_t attrbits;
int error;
NFSD_DEBUG(4, "in nfsrv_setacldsdorpc\n");
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
/*
* Use a stateid where other is an alternating 01010 pattern and
* seqid is 0xffffffff. This value is not defined as special by
* the RFC and is used by the FreeBSD NFS server to indicate an
* MDS->DS proxy operation.
*/
st.other[0] = 0x55555555;
st.other[1] = 0x55555555;
st.other[2] = 0x55555555;
st.seqid = 0xffffffff;
nfscl_reqstart(nd, NFSPROC_SETACL, nmp, (u_int8_t *)fhp, sizeof(*fhp),
NULL, NULL, 0, 0);
nfsm_stateidtom(nd, &st, NFSSTATEID_PUTSTATEID);
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
/*
* The "vp" argument to nfsv4_fillattr() is only used for vnode_type(),
* so passing in the metadata "vp" will be ok, since it is of
* the same type (VREG).
*/
nfsv4_fillattr(nd, NULL, vp, aclp, NULL, NULL, 0, &attrbits, NULL,
NULL, 0, 0, 0, 0, 0, NULL);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
NFSD_DEBUG(4, "nfsrv_setacldsdorpc: aft setaclrpc=%d\n",
nd->nd_repstat);
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
return (error);
}
struct nfsrvsetacldsdorpc {
int done;
int inprog;
struct task tsk;
fhandle_t fh;
struct nfsmount *nmp;
struct vnode *vp;
struct ucred *cred;
NFSPROC_T *p;
struct acl *aclp;
int err;
};
/*
* Start up the thread that will execute nfsrv_setacldsdorpc().
*/
static void
start_setacldsdorpc(void *arg, int pending)
{
struct nfsrvsetacldsdorpc *drpc;
drpc = (struct nfsrvsetacldsdorpc *)arg;
drpc->err = nfsrv_setacldsdorpc(&drpc->fh, drpc->cred, drpc->p,
drpc->vp, drpc->nmp, drpc->aclp);
drpc->done = 1;
}
static int
nfsrv_setacldsrpc(fhandle_t *fhp, struct ucred *cred, NFSPROC_T *p,
struct vnode *vp, struct nfsmount **nmpp, int mirrorcnt, struct acl *aclp,
int *failposp)
{
struct nfsrvsetacldsdorpc *drpc, *tdrpc = NULL;
int error, i, ret, timo;
NFSD_DEBUG(4, "in nfsrv_setacldsrpc\n");
drpc = NULL;
if (mirrorcnt > 1)
tdrpc = drpc = malloc(sizeof(*drpc) * (mirrorcnt - 1), M_TEMP,
M_WAITOK);
/*
* Do the setattr RPC for every DS, using a separate kernel process
* for every DS except the last one.
*/
error = 0;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
tdrpc->done = 0;
tdrpc->inprog = 0;
NFSBCOPY(fhp, &tdrpc->fh, sizeof(*fhp));
tdrpc->nmp = *nmpp;
tdrpc->vp = vp;
tdrpc->cred = cred;
tdrpc->p = p;
tdrpc->aclp = aclp;
tdrpc->err = 0;
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_setacldsdorpc, tdrpc);
NFSD_DEBUG(4, "nfsrv_setacldsrpc: nfs_pnfsio=%d\n",
ret);
}
if (ret != 0) {
ret = nfsrv_setacldsdorpc(fhp, cred, p, vp, *nmpp,
aclp);
if (nfsds_failerr(ret) && *failposp == -1)
*failposp = i;
else if (error == 0 && ret != 0)
error = ret;
}
nmpp++;
fhp++;
}
ret = nfsrv_setacldsdorpc(fhp, cred, p, vp, *nmpp, aclp);
if (nfsds_failerr(ret) && *failposp == -1 && mirrorcnt > 1)
*failposp = mirrorcnt - 1;
else if (error == 0 && ret != 0)
error = ret;
NFSD_DEBUG(4, "nfsrv_setacldsrpc: aft setextat=%d\n", error);
tdrpc = drpc;
timo = hz / 50; /* Wait for 20msec. */
if (timo < 1)
timo = 1;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
/* Wait for RPCs on separate threads to complete. */
while (tdrpc->inprog != 0 && tdrpc->done == 0)
tsleep(&tdrpc->tsk, PVFS, "srvacds", timo);
if (nfsds_failerr(tdrpc->err) && *failposp == -1)
*failposp = i;
else if (error == 0 && tdrpc->err != 0)
error = tdrpc->err;
}
free(drpc, M_TEMP);
return (error);
}
/*
* Getattr call to the DS for the attributes that change due to writing.
*/
static int
nfsrv_getattrdsrpc(fhandle_t *fhp, struct ucred *cred, NFSPROC_T *p,
struct vnode *vp, struct nfsmount *nmp, struct nfsvattr *nap)
{
struct nfsrv_descript *nd;
int error;
nfsattrbit_t attrbits;
NFSD_DEBUG(4, "in nfsrv_getattrdsrpc\n");
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, (u_int8_t *)fhp,
sizeof(fhandle_t), NULL, NULL, 0, 0);
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SIZE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEACCESS);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SPACEUSED);
(void) nfsrv_putattrbit(nd, &attrbits);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
NFSD_DEBUG(4, "nfsrv_getattrdsrpc: aft getattrrpc=%d\n",
nd->nd_repstat);
if (nd->nd_repstat == 0) {
error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
NULL, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL,
NULL, NULL);
/*
* We can only save the updated values in the extended
* attribute if the vp is exclusively locked.
* This should happen when any of the following operations
* occur on the vnode:
* Close, Delegreturn, LayoutCommit, LayoutReturn
* As such, the updated extended attribute should get saved
* before nfsrv_checkdsattr() returns 0 and allows the cached
* attributes to be returned without calling this function.
*/
if (error == 0 && VOP_ISLOCKED(vp) == LK_EXCLUSIVE) {
error = nfsrv_setextattr(vp, nap, p);
NFSD_DEBUG(4, "nfsrv_getattrdsrpc: aft setextat=%d\n",
error);
}
} else
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
NFSD_DEBUG(4, "nfsrv_getattrdsrpc error=%d\n", error);
return (error);
}
/*
* Seek call to a DS.
*/
static int
nfsrv_seekdsrpc(fhandle_t *fhp, off_t *offp, int content, bool *eofp,
struct ucred *cred, NFSPROC_T *p, struct nfsmount *nmp)
{
uint32_t *tl;
struct nfsrv_descript *nd;
nfsv4stateid_t st;
int error;
NFSD_DEBUG(4, "in nfsrv_seekdsrpc\n");
/*
* Use a stateid where other is an alternating 01010 pattern and
* seqid is 0xffffffff. This value is not defined as special by
* the RFC and is used by the FreeBSD NFS server to indicate an
* MDS->DS proxy operation.
*/
st.other[0] = 0x55555555;
st.other[1] = 0x55555555;
st.other[2] = 0x55555555;
st.seqid = 0xffffffff;
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
nfscl_reqstart(nd, NFSPROC_SEEKDS, nmp, (u_int8_t *)fhp,
sizeof(fhandle_t), NULL, NULL, 0, 0);
nfsm_stateidtom(nd, &st, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
txdr_hyper(*offp, tl); tl += 2;
*tl = txdr_unsigned(content);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
NFSD_DEBUG(4, "nfsrv_seekdsrpc: aft seekrpc=%d\n", nd->nd_repstat);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED + NFSX_HYPER);
if (*tl++ == newnfs_true)
*eofp = true;
else
*eofp = false;
*offp = fxdr_hyper(tl);
} else
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
NFSD_DEBUG(4, "nfsrv_seekdsrpc error=%d\n", error);
return (error);
}
/*
* Get the device id and file handle for a DS file.
*/
int
nfsrv_dsgetdevandfh(struct vnode *vp, NFSPROC_T *p, int *mirrorcntp,
fhandle_t *fhp, char *devid)
{
int buflen, error;
char *buf;
buflen = 1024;
buf = malloc(buflen, M_TEMP, M_WAITOK);
error = nfsrv_dsgetsockmnt(vp, 0, buf, &buflen, mirrorcntp, p, NULL,
fhp, devid, NULL, NULL, NULL, NULL, NULL, NULL);
free(buf, M_TEMP);
return (error);
}
/*
* Do a Lookup against the DS for the filename.
*/
static int
nfsrv_pnfslookupds(struct vnode *vp, struct vnode *dvp, struct pnfsdsfile *pf,
struct vnode **nvpp, NFSPROC_T *p)
{
struct nameidata named;
struct ucred *tcred;
char *bufp;
u_long *hashp;
struct vnode *nvp;
int error;
tcred = newnfs_getcred();
named.ni_cnd.cn_nameiop = LOOKUP;
named.ni_cnd.cn_lkflags = LK_SHARED | LK_RETRY;
named.ni_cnd.cn_cred = tcred;
named.ni_cnd.cn_flags = ISLASTCN | LOCKPARENT | LOCKLEAF | SAVENAME;
nfsvno_setpathbuf(&named, &bufp, &hashp);
named.ni_cnd.cn_nameptr = bufp;
named.ni_cnd.cn_namelen = strlen(pf->dsf_filename);
strlcpy(bufp, pf->dsf_filename, NAME_MAX);
NFSD_DEBUG(4, "nfsrv_pnfslookupds: filename=%s\n", bufp);
error = VOP_LOOKUP(dvp, &nvp, &named.ni_cnd);
NFSD_DEBUG(4, "nfsrv_pnfslookupds: aft LOOKUP=%d\n", error);
NFSFREECRED(tcred);
nfsvno_relpathbuf(&named);
if (error == 0)
*nvpp = nvp;
NFSD_DEBUG(4, "eo nfsrv_pnfslookupds=%d\n", error);
return (error);
}
/*
* Set the file handle to the correct one.
*/
static void
nfsrv_pnfssetfh(struct vnode *vp, struct pnfsdsfile *pf, char *devid,
char *fnamep, struct vnode *nvp, NFSPROC_T *p)
{
struct nfsnode *np;
int ret = 0;
np = VTONFS(nvp);
NFSBCOPY(np->n_fhp->nfh_fh, &pf->dsf_fh, NFSX_MYFH);
/*
* We can only do a vn_set_extattr() if the vnode is exclusively
* locked and vn_start_write() has been done. If devid != NULL or
* fnamep != NULL or the vnode is shared locked, vn_start_write()
* may not have been done.
* If not done now, it will be done on a future call.
*/
if (devid == NULL && fnamep == NULL && NFSVOPISLOCKED(vp) ==
LK_EXCLUSIVE)
ret = vn_extattr_set(vp, IO_NODELOCKED,
EXTATTR_NAMESPACE_SYSTEM, "pnfsd.dsfile", sizeof(*pf),
(char *)pf, p);
NFSD_DEBUG(4, "eo nfsrv_pnfssetfh=%d\n", ret);
}
/*
* Cause RPCs waiting on "nmp" to fail. This is called for a DS mount point
* when the DS has failed.
*/
void
nfsrv_killrpcs(struct nfsmount *nmp)
{
/*
* Call newnfs_nmcancelreqs() to cause
* any RPCs in progress on the mount point to
* fail.
* This will cause any process waiting for an
* RPC to complete while holding a vnode lock
* on the mounted-on vnode (such as "df" or
* a non-forced "umount") to fail.
* This will unlock the mounted-on vnode so
* a forced dismount can succeed.
* The NFSMNTP_CANCELRPCS flag should be set when this function is
* called.
*/
newnfs_nmcancelreqs(nmp);
}
/*
* Sum up the statfs info for each of the DSs, so that the client will
* receive the total for all DSs.
*/
static int
nfsrv_pnfsstatfs(struct statfs *sf, struct mount *mp)
{
struct statfs *tsf;
struct nfsdevice *ds;
struct vnode **dvpp, **tdvpp, *dvp;
uint64_t tot;
int cnt, error = 0, i;
if (nfsrv_devidcnt <= 0)
return (ENXIO);
dvpp = mallocarray(nfsrv_devidcnt, sizeof(*dvpp), M_TEMP, M_WAITOK);
tsf = malloc(sizeof(*tsf), M_TEMP, M_WAITOK);
/* Get an array of the dvps for the DSs. */
tdvpp = dvpp;
i = 0;
NFSDDSLOCK();
/* First, search for matches for same file system. */
TAILQ_FOREACH(ds, &nfsrv_devidhead, nfsdev_list) {
if (ds->nfsdev_nmp != NULL && ds->nfsdev_mdsisset != 0 &&
fsidcmp(&ds->nfsdev_mdsfsid, &mp->mnt_stat.f_fsid) == 0) {
if (++i > nfsrv_devidcnt)
break;
*tdvpp++ = ds->nfsdev_dvp;
}
}
/*
* If no matches for same file system, total all servers not assigned
* to a file system.
*/
if (i == 0) {
TAILQ_FOREACH(ds, &nfsrv_devidhead, nfsdev_list) {
if (ds->nfsdev_nmp != NULL &&
ds->nfsdev_mdsisset == 0) {
if (++i > nfsrv_devidcnt)
break;
*tdvpp++ = ds->nfsdev_dvp;
}
}
}
NFSDDSUNLOCK();
cnt = i;
/* Do a VFS_STATFS() for each of the DSs and sum them up. */
tdvpp = dvpp;
for (i = 0; i < cnt && error == 0; i++) {
dvp = *tdvpp++;
error = VFS_STATFS(dvp->v_mount, tsf);
if (error == 0) {
if (sf->f_bsize == 0) {
if (tsf->f_bsize > 0)
sf->f_bsize = tsf->f_bsize;
else
sf->f_bsize = 8192;
}
if (tsf->f_blocks > 0) {
if (sf->f_bsize != tsf->f_bsize) {
tot = tsf->f_blocks * tsf->f_bsize;
sf->f_blocks += (tot / sf->f_bsize);
} else
sf->f_blocks += tsf->f_blocks;
}
if (tsf->f_bfree > 0) {
if (sf->f_bsize != tsf->f_bsize) {
tot = tsf->f_bfree * tsf->f_bsize;
sf->f_bfree += (tot / sf->f_bsize);
} else
sf->f_bfree += tsf->f_bfree;
}
if (tsf->f_bavail > 0) {
if (sf->f_bsize != tsf->f_bsize) {
tot = tsf->f_bavail * tsf->f_bsize;
sf->f_bavail += (tot / sf->f_bsize);
} else
sf->f_bavail += tsf->f_bavail;
}
}
}
free(tsf, M_TEMP);
free(dvpp, M_TEMP);
return (error);
}
/*
* Set an NFSv4 acl.
*/
int
nfsrv_setacl(struct vnode *vp, NFSACL_T *aclp, struct ucred *cred, NFSPROC_T *p)
{
int error;
if (nfsrv_useacl == 0 || nfs_supportsnfsv4acls(vp) == 0) {
error = NFSERR_ATTRNOTSUPP;
goto out;
}
/*
* With NFSv4 ACLs, chmod(2) may need to add additional entries.
* Make sure it has enough room for that - splitting every entry
* into two and appending "canonical six" entries at the end.
* Cribbed out of kern/vfs_acl.c - Rick M.
*/
if (aclp->acl_cnt > (ACL_MAX_ENTRIES - 6) / 2) {
error = NFSERR_ATTRNOTSUPP;
goto out;
}
error = VOP_SETACL(vp, ACL_TYPE_NFS4, aclp, cred, p);
if (error == 0) {
error = nfsrv_dssetacl(vp, aclp, cred, p);
if (error == ENOENT)
error = 0;
}
out:
NFSEXITCODE(error);
return (error);
}
/*
* Seek vnode op call (actually it is a VOP_IOCTL()).
* This function is called with the vnode locked, but unlocks and vrele()s
* the vp before returning.
*/
int
nfsvno_seek(struct nfsrv_descript *nd, struct vnode *vp, u_long cmd,
off_t *offp, int content, bool *eofp, struct ucred *cred, NFSPROC_T *p)
{
struct nfsvattr at;
int error, ret;
ASSERT_VOP_LOCKED(vp, "nfsvno_seek vp");
/*
* Attempt to seek on a DS file. A return of ENOENT implies
* there is no DS file to seek on.
*/
error = nfsrv_proxyds(vp, 0, 0, cred, p, NFSPROC_SEEKDS, NULL,
NULL, NULL, NULL, NULL, offp, content, eofp);
if (error != ENOENT) {
vput(vp);
return (error);
}
/*
* Do the VOP_IOCTL() call. For the case where *offp == file_size,
* VOP_IOCTL() will return ENXIO. However, the correct reply for
* NFSv4.2 is *eofp == true and error == 0 for this case.
*/
NFSVOPUNLOCK(vp);
error = VOP_IOCTL(vp, cmd, offp, 0, cred, p);
*eofp = false;
if (error == ENXIO || (error == 0 && cmd == FIOSEEKHOLE)) {
/* Handle the cases where we might be at EOF. */
ret = nfsvno_getattr(vp, &at, nd, p, 0, NULL);
if (ret == 0 && *offp == at.na_size) {
*eofp = true;
error = 0;
}
if (ret != 0 && error == 0)
error = ret;
}
vrele(vp);
NFSEXITCODE(error);
return (error);
}
/*
* Allocate vnode op call.
*/
int
nfsvno_allocate(struct vnode *vp, off_t off, off_t len, struct ucred *cred,
NFSPROC_T *p)
{
int error;
off_t olen;
ASSERT_VOP_ELOCKED(vp, "nfsvno_allocate vp");
/*
* Attempt to allocate on a DS file. A return of ENOENT implies
* there is no DS file to allocate on.
*/
error = nfsrv_proxyds(vp, off, 0, cred, p, NFSPROC_ALLOCATE, NULL,
NULL, NULL, NULL, NULL, &len, 0, NULL);
if (error != ENOENT)
return (error);
/*
* Do the actual VOP_ALLOCATE(), looping so long as
* progress is being made, to achieve completion.
*/
do {
olen = len;
error = VOP_ALLOCATE(vp, &off, &len, IO_SYNC, cred);
if (error == 0 && len > 0 && olen > len)
maybe_yield();
} while (error == 0 && len > 0 && olen > len);
if (error == 0 && len > 0)
error = NFSERR_IO;
NFSEXITCODE(error);
return (error);
}
/*
* Deallocate vnode op call.
*/
int
nfsvno_deallocate(struct vnode *vp, off_t off, off_t len, struct ucred *cred,
NFSPROC_T *p)
{
int error;
off_t olen;
ASSERT_VOP_ELOCKED(vp, "nfsvno_deallocate vp");
/*
* Attempt to deallocate on a DS file. A return of ENOENT implies
* there is no DS file to deallocate on.
*/
error = nfsrv_proxyds(vp, off, 0, cred, p, NFSPROC_DEALLOCATE, NULL,
NULL, NULL, NULL, NULL, &len, 0, NULL);
if (error != ENOENT)
return (error);
/*
* Do the actual VOP_DEALLOCATE(), looping so long as
* progress is being made, to achieve completion.
*/
do {
olen = len;
error = VOP_DEALLOCATE(vp, &off, &len, 0, IO_SYNC, cred);
if (error == 0 && len > 0 && olen > len)
maybe_yield();
} while (error == 0 && len > 0 && olen > len);
if (error == 0 && len > 0)
error = NFSERR_IO;
NFSEXITCODE(error);
return (error);
}
/*
* Get Extended Atribute vnode op into an mbuf list.
*/
int
nfsvno_getxattr(struct vnode *vp, char *name, uint32_t maxresp,
struct ucred *cred, uint64_t flag, int maxextsiz, struct thread *p,
struct mbuf **mpp, struct mbuf **mpendp, int *lenp)
{
struct iovec *iv;
struct uio io, *uiop = &io;
struct mbuf *m, *m2;
int alen, error, len, tlen;
size_t siz;
/* First, find out the size of the extended attribute. */
error = VOP_GETEXTATTR(vp, EXTATTR_NAMESPACE_USER, name, NULL,
&siz, cred, p);
if (error != 0)
return (NFSERR_NOXATTR);
if (siz > maxresp - NFS_MAXXDR)
return (NFSERR_XATTR2BIG);
len = siz;
tlen = NFSM_RNDUP(len);
if (tlen > 0) {
/*
* If cnt > MCLBYTES and the reply will not be saved, use
* ext_pgs mbufs for TLS.
* For NFSv4.0, we do not know for sure if the reply will
* be saved, so do not use ext_pgs mbufs for NFSv4.0.
* Always use ext_pgs mbufs if ND_EXTPG is set.
*/
if ((flag & ND_EXTPG) != 0 || (tlen > MCLBYTES &&
(flag & (ND_TLS | ND_SAVEREPLY)) == ND_TLS &&
(flag & (ND_NFSV4 | ND_NFSV41)) != ND_NFSV4))
uiop->uio_iovcnt = nfsrv_createiovec_extpgs(tlen,
maxextsiz, &m, &m2, &iv);
else
uiop->uio_iovcnt = nfsrv_createiovec(tlen, &m, &m2,
&iv);
uiop->uio_iov = iv;
} else {
uiop->uio_iovcnt = 0;
uiop->uio_iov = iv = NULL;
m = m2 = NULL;
}
uiop->uio_offset = 0;
uiop->uio_resid = tlen;
uiop->uio_rw = UIO_READ;
uiop->uio_segflg = UIO_SYSSPACE;
uiop->uio_td = p;
#ifdef MAC
error = mac_vnode_check_getextattr(cred, vp, EXTATTR_NAMESPACE_USER,
name);
if (error != 0)
goto out;
#endif
if (tlen > 0)
error = VOP_GETEXTATTR(vp, EXTATTR_NAMESPACE_USER, name, uiop,
NULL, cred, p);
if (error != 0)
goto out;
if (uiop->uio_resid > 0) {
alen = tlen;
len = tlen - uiop->uio_resid;
tlen = NFSM_RNDUP(len);
if (alen != tlen)
printf("nfsvno_getxattr: weird size read\n");
if (tlen == 0) {
m_freem(m);
m = m2 = NULL;
} else if (alen != tlen || tlen != len)
m2 = nfsrv_adj(m, alen - tlen, tlen - len);
}
*lenp = len;
*mpp = m;
*mpendp = m2;
out:
if (error != 0) {
if (m != NULL)
m_freem(m);
*lenp = 0;
}
free(iv, M_TEMP);
NFSEXITCODE(error);
return (error);
}
/*
* Set Extended attribute vnode op from an mbuf list.
*/
int
nfsvno_setxattr(struct vnode *vp, char *name, int len, struct mbuf *m,
char *cp, struct ucred *cred, struct thread *p)
{
struct iovec *iv;
struct uio uio, *uiop = &uio;
int cnt, error;
error = 0;
#ifdef MAC
error = mac_vnode_check_setextattr(cred, vp, EXTATTR_NAMESPACE_USER,
name);
#endif
if (error != 0)
goto out;
uiop->uio_rw = UIO_WRITE;
uiop->uio_segflg = UIO_SYSSPACE;
uiop->uio_td = p;
uiop->uio_offset = 0;
uiop->uio_resid = len;
if (len > 0) {
error = nfsrv_createiovecw(len, m, cp, &iv, &cnt);
uiop->uio_iov = iv;
uiop->uio_iovcnt = cnt;
} else {
uiop->uio_iov = iv = NULL;
uiop->uio_iovcnt = 0;
}
if (error == 0) {
error = VOP_SETEXTATTR(vp, EXTATTR_NAMESPACE_USER, name, uiop,
cred, p);
free(iv, M_TEMP);
}
out:
NFSEXITCODE(error);
return (error);
}
/*
* Remove Extended attribute vnode op.
*/
int
nfsvno_rmxattr(struct nfsrv_descript *nd, struct vnode *vp, char *name,
struct ucred *cred, struct thread *p)
{
int error;
/*
* Get rid of any delegations. I am not sure why this is required,
* but RFC-8276 says so.
*/
error = nfsrv_checkremove(vp, 0, nd, nd->nd_clientid, p);
if (error != 0)
goto out;
#ifdef MAC
error = mac_vnode_check_deleteextattr(cred, vp, EXTATTR_NAMESPACE_USER,
name);
if (error != 0)
goto out;
#endif
error = VOP_DELETEEXTATTR(vp, EXTATTR_NAMESPACE_USER, name, cred, p);
if (error == EOPNOTSUPP)
error = VOP_SETEXTATTR(vp, EXTATTR_NAMESPACE_USER, name, NULL,
cred, p);
out:
NFSEXITCODE(error);
return (error);
}
/*
* List Extended Atribute vnode op into an mbuf list.
*/
int
nfsvno_listxattr(struct vnode *vp, uint64_t cookie, struct ucred *cred,
struct thread *p, u_char **bufp, uint32_t *lenp, bool *eofp)
{
struct iovec iv;
struct uio io;
int error;
size_t siz;
*bufp = NULL;
/* First, find out the size of the extended attribute. */
error = VOP_LISTEXTATTR(vp, EXTATTR_NAMESPACE_USER, NULL, &siz, cred,
p);
if (error != 0)
return (NFSERR_NOXATTR);
if (siz <= cookie) {
*lenp = 0;
*eofp = true;
goto out;
}
if (siz > cookie + *lenp) {
siz = cookie + *lenp;
*eofp = false;
} else
*eofp = true;
/* Just choose a sanity limit of 10Mbytes for malloc(M_TEMP). */
if (siz > 10 * 1024 * 1024) {
error = NFSERR_XATTR2BIG;
goto out;
}
*bufp = malloc(siz, M_TEMP, M_WAITOK);
iv.iov_base = *bufp;
iv.iov_len = siz;
io.uio_iovcnt = 1;
io.uio_iov = &iv;
io.uio_offset = 0;
io.uio_resid = siz;
io.uio_rw = UIO_READ;
io.uio_segflg = UIO_SYSSPACE;
io.uio_td = p;
#ifdef MAC
error = mac_vnode_check_listextattr(cred, vp, EXTATTR_NAMESPACE_USER);
if (error != 0)
goto out;
#endif
error = VOP_LISTEXTATTR(vp, EXTATTR_NAMESPACE_USER, &io, NULL, cred,
p);
if (error != 0)
goto out;
if (io.uio_resid > 0)
siz -= io.uio_resid;
*lenp = siz;
out:
if (error != 0) {
free(*bufp, M_TEMP);
*bufp = NULL;
}
NFSEXITCODE(error);
return (error);
}
/*
* Trim trailing data off the mbuf list being built.
*/
void
nfsm_trimtrailing(struct nfsrv_descript *nd, struct mbuf *mb, char *bpos,
int bextpg, int bextpgsiz)
{
vm_page_t pg;
int fullpgsiz, i;
if (mb->m_next != NULL) {
m_freem(mb->m_next);
mb->m_next = NULL;
}
if ((mb->m_flags & M_EXTPG) != 0) {
KASSERT(bextpg >= 0 && bextpg < mb->m_epg_npgs,
("nfsm_trimtrailing: bextpg out of range"));
KASSERT(bpos == (char *)(void *)
PHYS_TO_DMAP(mb->m_epg_pa[bextpg]) + PAGE_SIZE - bextpgsiz,
("nfsm_trimtrailing: bextpgsiz bad!"));
/* First, get rid of any pages after this position. */
for (i = mb->m_epg_npgs - 1; i > bextpg; i--) {
pg = PHYS_TO_VM_PAGE(mb->m_epg_pa[i]);
vm_page_unwire_noq(pg);
vm_page_free(pg);
}
mb->m_epg_npgs = bextpg + 1;
if (bextpg == 0)
fullpgsiz = PAGE_SIZE - mb->m_epg_1st_off;
else
fullpgsiz = PAGE_SIZE;
mb->m_epg_last_len = fullpgsiz - bextpgsiz;
mb->m_len = m_epg_pagelen(mb, 0, mb->m_epg_1st_off);
for (i = 1; i < mb->m_epg_npgs; i++)
mb->m_len += m_epg_pagelen(mb, i, 0);
nd->nd_bextpgsiz = bextpgsiz;
nd->nd_bextpg = bextpg;
} else
mb->m_len = bpos - mtod(mb, char *);
nd->nd_mb = mb;
nd->nd_bpos = bpos;
}
/*
* Check to see if a put file handle operation should test for
* NFSERR_WRONGSEC, although NFSv3 actually returns NFSERR_AUTHERR.
* When Open is the next operation, NFSERR_WRONGSEC cannot be
* replied for the Open cases that use a component. Thia can
* be identified by the fact that the file handle's type is VDIR.
*/
bool
nfsrv_checkwrongsec(struct nfsrv_descript *nd, int nextop, enum vtype vtyp)
{
if ((nd->nd_flag & ND_NFSV4) == 0) {
if (nd->nd_procnum == NFSPROC_FSINFO)
return (false);
return (true);
}
if ((nd->nd_flag & ND_LASTOP) != 0)
return (false);
if (nextop == NFSV4OP_PUTROOTFH || nextop == NFSV4OP_PUTFH ||
nextop == NFSV4OP_PUTPUBFH || nextop == NFSV4OP_RESTOREFH ||
nextop == NFSV4OP_LOOKUP || nextop == NFSV4OP_LOOKUPP ||
nextop == NFSV4OP_SECINFO || nextop == NFSV4OP_SECINFONONAME)
return (false);
if (nextop == NFSV4OP_OPEN && vtyp == VDIR)
return (false);
return (true);
}
/*
* Check DSs marked no space.
*/
void
nfsrv_checknospc(void)
{
struct statfs *tsf;
struct nfsdevice *ds;
struct vnode **dvpp, **tdvpp, *dvp;
char *devid, *tdevid;
int cnt, error = 0, i;
if (nfsrv_devidcnt <= 0)
return;
dvpp = mallocarray(nfsrv_devidcnt, sizeof(*dvpp), M_TEMP, M_WAITOK);
devid = malloc(nfsrv_devidcnt * NFSX_V4DEVICEID, M_TEMP, M_WAITOK);
tsf = malloc(sizeof(*tsf), M_TEMP, M_WAITOK);
/* Get an array of the dvps for the DSs. */
tdvpp = dvpp;
tdevid = devid;
i = 0;
NFSDDSLOCK();
/* First, search for matches for same file system. */
TAILQ_FOREACH(ds, &nfsrv_devidhead, nfsdev_list) {
if (ds->nfsdev_nmp != NULL && ds->nfsdev_nospc) {
if (++i > nfsrv_devidcnt)
break;
*tdvpp++ = ds->nfsdev_dvp;
NFSBCOPY(ds->nfsdev_deviceid, tdevid, NFSX_V4DEVICEID);
tdevid += NFSX_V4DEVICEID;
}
}
NFSDDSUNLOCK();
/* Do a VFS_STATFS() for each of the DSs and clear no space. */
cnt = i;
tdvpp = dvpp;
tdevid = devid;
for (i = 0; i < cnt && error == 0; i++) {
dvp = *tdvpp++;
error = VFS_STATFS(dvp->v_mount, tsf);
if (error == 0 && tsf->f_bavail > 0) {
NFSD_DEBUG(1, "nfsrv_checknospc: reset nospc\n");
nfsrv_marknospc(tdevid, false);
}
tdevid += NFSX_V4DEVICEID;
}
free(tsf, M_TEMP);
free(dvpp, M_TEMP);
free(devid, M_TEMP);
}
extern int (*nfsd_call_nfsd)(struct thread *, struct nfssvc_args *);
/*
* Called once to initialize data structures...
*/
static int
nfsd_modevent(module_t mod, int type, void *data)
{
int error = 0, i;
static int loaded = 0;
switch (type) {
case MOD_LOAD:
if (loaded)
goto out;
newnfs_portinit();
for (i = 0; i < NFSRVCACHE_HASHSIZE; i++) {
mtx_init(&nfsrchash_table[i].mtx, "nfsrtc", NULL,
MTX_DEF);
mtx_init(&nfsrcahash_table[i].mtx, "nfsrtca", NULL,
MTX_DEF);
}
mtx_init(&nfsrc_udpmtx, "nfsuc", NULL, MTX_DEF);
mtx_init(&nfs_v4root_mutex, "nfs4rt", NULL, MTX_DEF);
mtx_init(&nfsv4root_mnt.mnt_mtx, "nfs4mnt", NULL, MTX_DEF);
mtx_init(&nfsrv_dontlistlock_mtx, "nfs4dnl", NULL, MTX_DEF);
mtx_init(&nfsrv_recalllock_mtx, "nfs4rec", NULL, MTX_DEF);
lockinit(&nfsv4root_mnt.mnt_explock, PVFS, "explock", 0, 0);
callout_init(&nfsd_callout, 1);
nfsrvd_initcache();
nfsd_init();
NFSD_LOCK();
nfsrvd_init(0);
NFSD_UNLOCK();
nfsd_mntinit();
#ifdef VV_DISABLEDELEG
vn_deleg_ops.vndeleg_recall = nfsd_recalldelegation;
vn_deleg_ops.vndeleg_disable = nfsd_disabledelegation;
#endif
nfsd_call_nfsd = nfssvc_nfsd;
loaded = 1;
break;
case MOD_UNLOAD:
if (newnfs_numnfsd != 0) {
error = EBUSY;
break;
}
#ifdef VV_DISABLEDELEG
vn_deleg_ops.vndeleg_recall = NULL;
vn_deleg_ops.vndeleg_disable = NULL;
#endif
nfsd_call_nfsd = NULL;
callout_drain(&nfsd_callout);
/* Clean out all NFSv4 state. */
nfsrv_throwawayallstate(curthread);
/* Clean the NFS server reply cache */
nfsrvd_cleancache();
/* Free up the krpc server pool. */
if (nfsrvd_pool != NULL)
svcpool_destroy(nfsrvd_pool);
/* and get rid of the locks */
for (i = 0; i < NFSRVCACHE_HASHSIZE; i++) {
mtx_destroy(&nfsrchash_table[i].mtx);
mtx_destroy(&nfsrcahash_table[i].mtx);
}
mtx_destroy(&nfsrc_udpmtx);
mtx_destroy(&nfs_v4root_mutex);
mtx_destroy(&nfsv4root_mnt.mnt_mtx);
mtx_destroy(&nfsrv_dontlistlock_mtx);
mtx_destroy(&nfsrv_recalllock_mtx);
for (i = 0; i < nfsrv_sessionhashsize; i++)
mtx_destroy(&nfssessionhash[i].mtx);
if (nfslayouthash != NULL) {
for (i = 0; i < nfsrv_layouthashsize; i++)
mtx_destroy(&nfslayouthash[i].mtx);
free(nfslayouthash, M_NFSDSESSION);
}
lockdestroy(&nfsv4root_mnt.mnt_explock);
free(nfsclienthash, M_NFSDCLIENT);
free(nfslockhash, M_NFSDLOCKFILE);
free(nfssessionhash, M_NFSDSESSION);
loaded = 0;
break;
default:
error = EOPNOTSUPP;
break;
}
out:
NFSEXITCODE(error);
return (error);
}
static moduledata_t nfsd_mod = {
"nfsd",
nfsd_modevent,
NULL,
};
DECLARE_MODULE(nfsd, nfsd_mod, SI_SUB_VFS, SI_ORDER_ANY);
/* So that loader and kldload(2) can find us, wherever we are.. */
MODULE_VERSION(nfsd, 1);
MODULE_DEPEND(nfsd, nfscommon, 1, 1, 1);
MODULE_DEPEND(nfsd, nfslockd, 1, 1, 1);
MODULE_DEPEND(nfsd, krpc, 1, 1, 1);
MODULE_DEPEND(nfsd, nfssvc, 1, 1, 1);