gems-kernel/source/THIRDPARTY/xnu/bsd/nfs/nfs_subs.c
2024-06-03 11:29:39 -05:00

3146 lines
73 KiB
C

/*
* Copyright (c) 2000-2020 Apple Inc. All rights reserved.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_START@
*
* This file contains Original Code and/or Modifications of Original Code
* as defined in and that are subject to the Apple Public Source License
* Version 2.0 (the 'License'). You may not use this file except in
* compliance with the License. The rights granted to you under the License
* may not be used to create, or enable the creation or redistribution of,
* unlawful or unlicensed copies of an Apple operating system, or to
* circumvent, violate, or enable the circumvention or violation of, any
* terms of an Apple operating system software license agreement.
*
* Please obtain a copy of the License at
* http://www.opensource.apple.com/apsl/ and read it before using this file.
*
* The Original Code and all software distributed under the License are
* distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
* EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
* INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
* Please see the License for the specific language governing rights and
* limitations under the License.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_END@
*/
/* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
/*
* 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. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. 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.
*
* @(#)nfs_subs.c 8.8 (Berkeley) 5/22/95
* FreeBSD-Id: nfs_subs.c,v 1.47 1997/11/07 08:53:24 phk Exp $
*/
#include <nfs/nfs_conf.h>
#if CONFIG_NFS_SERVER
/*
* 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 <sys/kauth.h>
#include <sys/mount_internal.h>
#include <sys/vnode_internal.h>
#include <sys/kpi_mbuf.h>
#include <sys/un.h>
#include <sys/domain.h>
#include <nfs/rpcv2.h>
#include <nfs/nfsproto.h>
#include <nfs/nfs.h>
#define _NFS_XDR_SUBS_FUNCS_ /* define this to get xdrbuf function definitions */
#include <nfs/xdr_subs.h>
#include <nfs/nfsm_subs.h>
#include <nfs/nfs_gss.h>
/*
* NFS globals
*/
struct nfsrvstats __attribute__((aligned(8))) nfsrvstats;
size_t nfs_mbuf_mhlen = 0, nfs_mbuf_minclsize = 0;
/* NFS debugging support */
uint32_t nfsrv_debug_ctl;
#include <libkern/libkern.h>
#include <stdarg.h>
static mount_t nfsrv_getvfs_by_mntonname(char *path);
void
nfs_printf(unsigned int debug_control, unsigned int facility, unsigned int level, const char *fmt, ...)
{
va_list ap;
if (__NFS_IS_DBG(debug_control, facility, level)) {
va_start(ap, fmt);
vprintf(fmt, ap);
va_end(ap);
}
}
#define DISPLAYLEN 16
static bool
isprint(int ch)
{
return ch >= 0x20 && ch <= 0x7e;
}
static void
hexdump(void *data, size_t len)
{
size_t i, j;
unsigned char *d = data;
char *p, disbuf[3 * DISPLAYLEN + 1];
for (i = 0; i < len; i += DISPLAYLEN) {
for (p = disbuf, j = 0; (j + i) < len && j < DISPLAYLEN; j++, p += 3) {
snprintf(p, 4, "%2.2x ", d[i + j]);
}
for (; j < DISPLAYLEN; j++, p += 3) {
snprintf(p, 4, " ");
}
printf("%s ", disbuf);
for (p = disbuf, j = 0; (j + i) < len && j < DISPLAYLEN; j++, p++) {
snprintf(p, 2, "%c", isprint(d[i + j]) ? d[i + j] : '.');
}
printf("%s\n", disbuf);
}
}
void
nfs_dump_mbuf(const char *func, int lineno, const char *msg, mbuf_t mb)
{
mbuf_t m;
printf("%s:%d %s\n", func, lineno, msg);
for (m = mb; m; m = mbuf_next(m)) {
hexdump(mbuf_data(m), mbuf_len(m));
}
}
/*
* functions to convert between NFS and VFS types
*/
nfstype
vtonfs_type(enum vtype vtype, int nfsvers)
{
switch (vtype) {
case VNON:
return NFNON;
case VREG:
return NFREG;
case VDIR:
return NFDIR;
case VBLK:
return NFBLK;
case VCHR:
return NFCHR;
case VLNK:
return NFLNK;
case VSOCK:
if (nfsvers > NFS_VER2) {
return NFSOCK;
}
return NFNON;
case VFIFO:
if (nfsvers > NFS_VER2) {
return NFFIFO;
}
return NFNON;
case VBAD:
case VSTR:
case VCPLX:
default:
return NFNON;
}
}
enum vtype
nfstov_type(nfstype nvtype, int nfsvers)
{
switch (nvtype) {
case NFNON:
return VNON;
case NFREG:
return VREG;
case NFDIR:
return VDIR;
case NFBLK:
return VBLK;
case NFCHR:
return VCHR;
case NFLNK:
return VLNK;
case NFSOCK:
if (nfsvers > NFS_VER2) {
return VSOCK;
}
OS_FALLTHROUGH;
case NFFIFO:
if (nfsvers > NFS_VER2) {
return VFIFO;
}
OS_FALLTHROUGH;
case NFATTRDIR:
if (nfsvers > NFS_VER3) {
return VDIR;
}
OS_FALLTHROUGH;
case NFNAMEDATTR:
if (nfsvers > NFS_VER3) {
return VREG;
}
OS_FALLTHROUGH;
default:
return VNON;
}
}
int
vtonfsv2_mode(enum vtype vtype, mode_t m)
{
switch (vtype) {
case VNON:
case VREG:
case VDIR:
case VBLK:
case VCHR:
case VLNK:
case VSOCK:
return MAKEIMODE(vtype, m);
case VFIFO:
return MAKEIMODE(VCHR, m);
case VBAD:
case VSTR:
case VCPLX:
default:
return MAKEIMODE(VNON, m);
}
}
/*
* Mapping of old NFS Version 2 RPC numbers to generic numbers.
*/
int nfsv3_procid[NFS_NPROCS] = {
NFSPROC_NULL,
NFSPROC_GETATTR,
NFSPROC_SETATTR,
NFSPROC_NOOP,
NFSPROC_LOOKUP,
NFSPROC_READLINK,
NFSPROC_READ,
NFSPROC_NOOP,
NFSPROC_WRITE,
NFSPROC_CREATE,
NFSPROC_REMOVE,
NFSPROC_RENAME,
NFSPROC_LINK,
NFSPROC_SYMLINK,
NFSPROC_MKDIR,
NFSPROC_RMDIR,
NFSPROC_READDIR,
NFSPROC_FSSTAT,
NFSPROC_NOOP,
NFSPROC_NOOP,
NFSPROC_NOOP,
NFSPROC_NOOP,
NFSPROC_NOOP
};
/*
* and the reverse mapping from generic to Version 2 procedure numbers
*/
int nfsv2_procid[NFS_NPROCS] = {
NFSV2PROC_NULL,
NFSV2PROC_GETATTR,
NFSV2PROC_SETATTR,
NFSV2PROC_LOOKUP,
NFSV2PROC_NOOP,
NFSV2PROC_READLINK,
NFSV2PROC_READ,
NFSV2PROC_WRITE,
NFSV2PROC_CREATE,
NFSV2PROC_MKDIR,
NFSV2PROC_SYMLINK,
NFSV2PROC_CREATE,
NFSV2PROC_REMOVE,
NFSV2PROC_RMDIR,
NFSV2PROC_RENAME,
NFSV2PROC_LINK,
NFSV2PROC_READDIR,
NFSV2PROC_NOOP,
NFSV2PROC_STATFS,
NFSV2PROC_NOOP,
NFSV2PROC_NOOP,
NFSV2PROC_NOOP,
NFSV2PROC_NOOP
};
/*
* initialize NFS's cache of mbuf constants
*/
void
nfs_mbuf_init(void)
{
struct mbuf_stat ms;
mbuf_stats(&ms);
nfs_mbuf_mhlen = ms.mhlen;
nfs_mbuf_minclsize = ms.minclsize;
}
static void
nfs_netopt_free(struct nfs_netopt *no)
{
if (no->no_addr) {
kfree_data(no->no_addr, no->no_addr->sa_len);
}
if (no->no_mask) {
kfree_data(no->no_mask, no->no_mask->sa_len);
}
kfree_type(struct nfs_netopt, no);
}
/*
* allocate a list of mbufs to hold the given amount of data
*/
int
nfsm_mbuf_get_list(size_t size, mbuf_t *mp, int *mbcnt)
{
int error, cnt;
mbuf_t mhead, mlast, m;
size_t len, mlen;
error = cnt = 0;
mhead = mlast = NULL;
len = 0;
while (len < size) {
nfsm_mbuf_getcluster(error, &m, (size - len));
if (error) {
break;
}
if (!mhead) {
mhead = m;
}
if (mlast && ((error = mbuf_setnext(mlast, m)))) {
mbuf_free(m);
break;
}
mlen = mbuf_maxlen(m);
if ((len + mlen) > size) {
mlen = size - len;
}
mbuf_setlen(m, mlen);
len += mlen;
cnt++;
mlast = m;
}
if (!error) {
*mp = mhead;
*mbcnt = cnt;
}
return error;
}
/*
* nfsm_chain_new_mbuf()
*
* Add a new mbuf to the given chain.
*/
int
nfsm_chain_new_mbuf(struct nfsm_chain *nmc, size_t sizehint)
{
mbuf_t mb;
int error = 0;
if (nmc->nmc_flags & NFSM_CHAIN_FLAG_ADD_CLUSTERS) {
sizehint = nfs_mbuf_minclsize;
}
/* allocate a new mbuf */
nfsm_mbuf_getcluster(error, &mb, sizehint);
if (error) {
return error;
}
if (mb == NULL) {
panic("got NULL mbuf?");
}
/* do we have a current mbuf? */
if (nmc->nmc_mcur) {
/* first cap off current mbuf */
mbuf_setlen(nmc->nmc_mcur, nmc->nmc_ptr - (caddr_t)mbuf_data(nmc->nmc_mcur));
/* then append the new mbuf */
error = mbuf_setnext(nmc->nmc_mcur, mb);
if (error) {
mbuf_free(mb);
return error;
}
}
/* set up for using the new mbuf */
nmc->nmc_mcur = mb;
nmc->nmc_ptr = mbuf_data(mb);
nmc->nmc_left = mbuf_trailingspace(mb);
return 0;
}
/*
* nfsm_chain_add_opaque_f()
*
* Add "len" bytes of opaque data pointed to by "buf" to the given chain.
*/
int
nfsm_chain_add_opaque_f(struct nfsm_chain *nmc, const u_char *buf, size_t len)
{
size_t paddedlen, tlen;
int error;
paddedlen = nfsm_rndup(len);
while (paddedlen) {
if (!nmc->nmc_left) {
error = nfsm_chain_new_mbuf(nmc, paddedlen);
if (error) {
return error;
}
}
tlen = MIN(nmc->nmc_left, paddedlen);
if (tlen) {
if (len) {
if (tlen > len) {
tlen = len;
}
bcopy(buf, nmc->nmc_ptr, tlen);
} else {
bzero(nmc->nmc_ptr, tlen);
}
nmc->nmc_ptr += tlen;
nmc->nmc_left -= tlen;
paddedlen -= tlen;
if (len) {
buf += tlen;
len -= tlen;
}
}
}
return 0;
}
/*
* nfsm_chain_add_opaque_nopad_f()
*
* Add "len" bytes of opaque data pointed to by "buf" to the given chain.
* Do not XDR pad.
*/
int
nfsm_chain_add_opaque_nopad_f(struct nfsm_chain *nmc, const u_char *buf, size_t len)
{
size_t tlen;
int error;
while (len > 0) {
if (nmc->nmc_left <= 0) {
error = nfsm_chain_new_mbuf(nmc, len);
if (error) {
return error;
}
}
tlen = MIN(nmc->nmc_left, len);
bcopy(buf, nmc->nmc_ptr, tlen);
nmc->nmc_ptr += tlen;
nmc->nmc_left -= tlen;
len -= tlen;
buf += tlen;
}
return 0;
}
/*
* Find the length of the NFS mbuf chain
* up to the current encoding/decoding offset.
*/
size_t
nfsm_chain_offset(struct nfsm_chain *nmc)
{
mbuf_t mb;
size_t len = 0;
for (mb = nmc->nmc_mhead; mb; mb = mbuf_next(mb)) {
if (mb == nmc->nmc_mcur) {
return len + (nmc->nmc_ptr - (caddr_t) mbuf_data(mb));
}
len += mbuf_len(mb);
}
return len;
}
/*
* nfsm_chain_advance()
*
* Advance an nfsm_chain by "len" bytes.
*/
int
nfsm_chain_advance(struct nfsm_chain *nmc, size_t len)
{
mbuf_t mb;
while (len) {
if (nmc->nmc_left >= len) {
nmc->nmc_left -= len;
nmc->nmc_ptr += len;
return 0;
}
len -= nmc->nmc_left;
nmc->nmc_mcur = mb = mbuf_next(nmc->nmc_mcur);
if (!mb) {
return EBADRPC;
}
nmc->nmc_ptr = mbuf_data(mb);
nmc->nmc_left = mbuf_len(mb);
}
return 0;
}
#if 0
/*
* nfsm_chain_reverse()
*
* Reverse decode offset in an nfsm_chain by "len" bytes.
*/
int
nfsm_chain_reverse(struct nfsm_chain *nmc, size_t len)
{
size_t mlen, new_offset;
int error = 0;
mlen = nmc->nmc_ptr - (caddr_t) mbuf_data(nmc->nmc_mcur);
if (len <= mlen) {
nmc->nmc_ptr -= len;
nmc->nmc_left += len;
return 0;
}
new_offset = nfsm_chain_offset(nmc) - len;
nfsm_chain_dissect_init(error, nmc, nmc->nmc_mhead);
if (error) {
return error;
}
return nfsm_chain_advance(nmc, new_offset);
}
#endif
/*
* nfsm_chain_get_opaque_pointer_f()
*
* Return a pointer to the next "len" bytes of contiguous data in
* the mbuf chain. If the next "len" bytes are not contiguous, we
* try to manipulate the mbuf chain so that it is.
*
* The nfsm_chain is advanced by nfsm_rndup("len") bytes.
*/
int
nfsm_chain_get_opaque_pointer_f(struct nfsm_chain *nmc, uint32_t len, u_char **pptr)
{
mbuf_t mbcur, mb;
uint32_t padlen;
size_t mblen, cplen, need, left;
u_char *ptr;
int error = 0;
/* move to next mbuf with data */
while (nmc->nmc_mcur && (nmc->nmc_left == 0)) {
mb = mbuf_next(nmc->nmc_mcur);
nmc->nmc_mcur = mb;
if (!mb) {
break;
}
nmc->nmc_ptr = mbuf_data(mb);
nmc->nmc_left = mbuf_len(mb);
}
/* check if we've run out of data */
if (!nmc->nmc_mcur) {
return EBADRPC;
}
/* do we already have a contiguous buffer? */
if (nmc->nmc_left >= len) {
/* the returned pointer will be the current pointer */
*pptr = (u_char*)nmc->nmc_ptr;
error = nfsm_chain_advance(nmc, nfsm_rndup(len));
return error;
}
padlen = nfsm_rndup(len) - len;
/* we need (len - left) more bytes */
mbcur = nmc->nmc_mcur;
left = nmc->nmc_left;
need = len - left;
if (need > mbuf_trailingspace(mbcur)) {
/*
* The needed bytes won't fit in the current mbuf so we'll
* allocate a new mbuf to hold the contiguous range of data.
*/
nfsm_mbuf_getcluster(error, &mb, len);
if (error) {
return error;
}
/* double check that this mbuf can hold all the data */
if (mbuf_maxlen(mb) < len) {
mbuf_free(mb);
return EOVERFLOW;
}
/* the returned pointer will be the new mbuf's data pointer */
*pptr = ptr = mbuf_data(mb);
/* copy "left" bytes to the new mbuf */
bcopy(nmc->nmc_ptr, ptr, left);
ptr += left;
mbuf_setlen(mb, left);
/* insert the new mbuf between the current and next mbufs */
error = mbuf_setnext(mb, mbuf_next(mbcur));
if (!error) {
error = mbuf_setnext(mbcur, mb);
}
if (error) {
mbuf_free(mb);
return error;
}
/* reduce current mbuf's length by "left" */
mbuf_setlen(mbcur, mbuf_len(mbcur) - left);
/*
* update nmc's state to point at the end of the mbuf
* where the needed data will be copied to.
*/
nmc->nmc_mcur = mbcur = mb;
nmc->nmc_left = 0;
nmc->nmc_ptr = (caddr_t)ptr;
} else {
/* The rest of the data will fit in this mbuf. */
/* the returned pointer will be the current pointer */
*pptr = (u_char*)nmc->nmc_ptr;
/*
* update nmc's state to point at the end of the mbuf
* where the needed data will be copied to.
*/
nmc->nmc_ptr += left;
nmc->nmc_left = 0;
}
/*
* move the next "need" bytes into the current
* mbuf from the mbufs that follow
*/
/* extend current mbuf length */
mbuf_setlen(mbcur, mbuf_len(mbcur) + need);
/* mb follows mbufs we're copying/compacting data from */
mb = mbuf_next(mbcur);
while (need && mb) {
/* copy as much as we need/can */
ptr = mbuf_data(mb);
mblen = mbuf_len(mb);
cplen = MIN(mblen, need);
if (cplen) {
bcopy(ptr, nmc->nmc_ptr, cplen);
/*
* update the mbuf's pointer and length to reflect that
* the data was shifted to an earlier mbuf in the chain
*/
error = mbuf_setdata(mb, ptr + cplen, mblen - cplen);
if (error) {
mbuf_setlen(mbcur, mbuf_len(mbcur) - need);
return error;
}
/* update pointer/need */
nmc->nmc_ptr += cplen;
need -= cplen;
}
/* if more needed, go to next mbuf */
if (need) {
mb = mbuf_next(mb);
}
}
/* did we run out of data in the mbuf chain? */
if (need) {
mbuf_setlen(mbcur, mbuf_len(mbcur) - need);
return EBADRPC;
}
/*
* update nmc's state to point after this contiguous data
*
* "mb" points to the last mbuf we copied data from so we
* just set nmc to point at whatever remains in that mbuf.
*/
nmc->nmc_mcur = mb;
nmc->nmc_ptr = mbuf_data(mb);
nmc->nmc_left = mbuf_len(mb);
/* move past any padding */
if (padlen) {
error = nfsm_chain_advance(nmc, padlen);
}
return error;
}
/*
* nfsm_chain_get_opaque_f()
*
* Read the next "len" bytes in the chain into "buf".
* The nfsm_chain is advanced by nfsm_rndup("len") bytes.
*/
int
nfsm_chain_get_opaque_f(struct nfsm_chain *nmc, size_t len, u_char *buf)
{
size_t cplen, padlen;
int error = 0;
padlen = nfsm_rndup(len) - len;
/* loop through mbufs copying all the data we need */
while (len && nmc->nmc_mcur) {
/* copy as much as we need/can */
cplen = MIN(nmc->nmc_left, len);
if (cplen) {
bcopy(nmc->nmc_ptr, buf, cplen);
nmc->nmc_ptr += cplen;
nmc->nmc_left -= cplen;
buf += cplen;
len -= cplen;
}
/* if more needed, go to next mbuf */
if (len) {
mbuf_t mb = mbuf_next(nmc->nmc_mcur);
nmc->nmc_mcur = mb;
nmc->nmc_ptr = mb ? mbuf_data(mb) : NULL;
nmc->nmc_left = mb ? mbuf_len(mb) : 0;
}
}
/* did we run out of data in the mbuf chain? */
if (len) {
return EBADRPC;
}
if (padlen) {
nfsm_chain_adv(error, nmc, padlen);
}
return error;
}
/*
* nfsm_chain_get_uio()
*
* Read the next "len" bytes in the chain into the given uio.
* The nfsm_chain is advanced by nfsm_rndup("len") bytes.
*/
int
nfsm_chain_get_uio(struct nfsm_chain *nmc, size_t len, uio_t uio)
{
size_t cplen, padlen;
int error = 0;
padlen = nfsm_rndup(len) - len;
/* loop through mbufs copying all the data we need */
while (len && nmc->nmc_mcur) {
/* copy as much as we need/can */
cplen = MIN(nmc->nmc_left, len);
if (cplen) {
cplen = MIN(cplen, INT32_MAX);
error = uiomove(nmc->nmc_ptr, (int)cplen, uio);
if (error) {
return error;
}
nmc->nmc_ptr += cplen;
nmc->nmc_left -= cplen;
len -= cplen;
}
/* if more needed, go to next mbuf */
if (len) {
mbuf_t mb = mbuf_next(nmc->nmc_mcur);
nmc->nmc_mcur = mb;
nmc->nmc_ptr = mb ? mbuf_data(mb) : NULL;
nmc->nmc_left = mb ? mbuf_len(mb) : 0;
}
}
/* did we run out of data in the mbuf chain? */
if (len) {
return EBADRPC;
}
if (padlen) {
nfsm_chain_adv(error, nmc, padlen);
}
return error;
}
/*
* Schedule a callout thread to run an NFS timer function
* interval milliseconds in the future.
*/
void
nfs_interval_timer_start(thread_call_t call, time_t interval)
{
uint64_t deadline;
clock_interval_to_deadline((int)interval, 1000 * 1000, &deadline);
thread_call_enter_delayed(call, deadline);
}
int nfsrv_cmp_secflavs(struct nfs_sec *, struct nfs_sec *);
int nfsrv_hang_addrlist(struct nfs_export *, struct user_nfs_export_args *);
int nfsrv_free_netopt(struct radix_node *, void *);
int nfsrv_free_addrlist(struct nfs_export *, struct user_nfs_export_args *);
struct nfs_export_options *nfsrv_export_lookup(struct nfs_export *, mbuf_t);
struct nfs_export *nfsrv_fhtoexport(struct nfs_filehandle *);
struct nfs_user_stat_node *nfsrv_get_user_stat_node(struct nfs_active_user_list *, struct sockaddr *, uid_t);
void nfsrv_init_user_list(struct nfs_active_user_list *);
void nfsrv_free_user_list(struct nfs_active_user_list *);
/*
* add NFSv3 WCC data to an mbuf chain
*/
int
nfsm_chain_add_wcc_data_f(
struct nfsrv_descript *nd,
struct nfsm_chain *nmc,
int preattrerr,
struct vnode_attr *prevap,
int postattrerr,
struct vnode_attr *postvap)
{
int error = 0;
if (preattrerr) {
nfsm_chain_add_32(error, nmc, FALSE);
} else {
nfsm_chain_add_32(error, nmc, TRUE);
nfsm_chain_add_64(error, nmc, prevap->va_data_size);
nfsm_chain_add_time(error, nmc, NFS_VER3, &prevap->va_modify_time);
nfsm_chain_add_time(error, nmc, NFS_VER3, &prevap->va_change_time);
}
nfsm_chain_add_postop_attr(error, nd, nmc, postattrerr, postvap);
return error;
}
/*
* Extract a lookup path from the given mbufs and store it in
* a newly allocated buffer saved in the given nameidata structure.
*/
int
nfsm_chain_get_path_namei(
struct nfsm_chain *nmc,
uint32_t len,
struct nameidata *nip)
{
struct componentname *cnp = &nip->ni_cnd;
int error = 0;
char *cp;
if (len > (MAXPATHLEN - 1)) {
return ENAMETOOLONG;
}
/*
* Get a buffer for the name to be translated, and copy the
* name into the buffer.
*/
cnp->cn_pnbuf = zalloc(ZV_NAMEI);
cnp->cn_pnlen = MAXPATHLEN;
cnp->cn_flags |= HASBUF;
/* Copy the name from the mbuf list to the string */
cp = cnp->cn_pnbuf;
nfsm_chain_get_opaque(error, nmc, len, cp);
if (error) {
goto out;
}
cnp->cn_pnbuf[len] = '\0';
/* sanity check the string */
if ((strlen(cp) != len) || strchr(cp, '/')) {
error = EACCES;
}
out:
if (error) {
if (cnp->cn_pnbuf) {
NFS_ZFREE(ZV_NAMEI, cnp->cn_pnbuf);
}
cnp->cn_flags &= ~HASBUF;
} else {
nip->ni_pathlen = len;
}
return error;
}
/*
* Set up nameidata for a lookup() call and do it.
*/
int
nfsrv_namei(
struct nfsrv_descript *nd,
vfs_context_t ctx,
struct nameidata *nip,
struct nfs_filehandle *nfhp,
vnode_t *retdirp,
struct nfs_export **nxp,
struct nfs_export_options **nxop)
{
vnode_t dp;
int error;
struct componentname *cnp = &nip->ni_cnd;
uint32_t cnflags;
char *tmppn;
*retdirp = NULL;
/*
* Extract and set starting directory.
*/
error = nfsrv_fhtovp(nfhp, nd, &dp, nxp, nxop);
if (error) {
goto out;
}
error = nfsrv_credcheck(nd, ctx, *nxp, *nxop);
if (error || (vnode_vtype(dp) != VDIR)) {
vnode_put(dp);
error = ENOTDIR;
goto out;
}
*retdirp = dp;
nip->ni_cnd.cn_context = ctx;
if (*nxop && ((*nxop)->nxo_flags & NX_READONLY)) {
cnp->cn_flags |= RDONLY;
}
cnp->cn_flags |= NOCROSSMOUNT;
cnp->cn_nameptr = cnp->cn_pnbuf;
nip->ni_usedvp = nip->ni_startdir = dp;
nip->ni_rootdir = rootvnode;
/*
* And call lookup() to do the real work
*/
cnflags = nip->ni_cnd.cn_flags; /* store in case we have to restore */
while ((error = lookup(nip)) == ERECYCLE) {
nip->ni_cnd.cn_flags = cnflags;
cnp->cn_nameptr = cnp->cn_pnbuf;
nip->ni_usedvp = nip->ni_dvp = nip->ni_startdir = dp;
}
if (error) {
goto out;
}
/* Check for encountering a symbolic link */
if (cnp->cn_flags & ISSYMLINK) {
if (cnp->cn_flags & (LOCKPARENT | WANTPARENT)) {
vnode_put(nip->ni_dvp);
}
if (nip->ni_vp) {
vnode_put(nip->ni_vp);
nip->ni_vp = NULL;
}
error = EINVAL;
}
out:
if (error) {
tmppn = cnp->cn_pnbuf;
cnp->cn_pnbuf = NULL;
cnp->cn_flags &= ~HASBUF;
NFS_ZFREE(ZV_NAMEI, tmppn);
}
return error;
}
/*
* A fiddled version of m_adj() that ensures null fill to a 4-byte
* boundary and only trims off the back end
*/
void
nfsm_adj(mbuf_t mp, int len, int nul)
{
mbuf_t m, mnext;
int count, i;
long mlen;
char *cp;
/*
* Trim from tail. Scan the mbuf chain,
* calculating its length and finding the last mbuf.
* If the adjustment only affects this mbuf, then just
* adjust and return. Otherwise, rescan and truncate
* after the remaining size.
*/
count = 0;
m = mp;
for (;;) {
mlen = mbuf_len(m);
count += mlen;
mnext = mbuf_next(m);
if (mnext == NULL) {
break;
}
m = mnext;
}
if (mlen > len) {
mlen -= len;
mbuf_setlen(m, mlen);
if (nul > 0) {
cp = (caddr_t)mbuf_data(m) + mlen - nul;
for (i = 0; i < nul; i++) {
*cp++ = '\0';
}
}
return;
}
count -= len;
if (count < 0) {
count = 0;
}
/*
* Correct length for chain is "count".
* Find the mbuf with last data, adjust its length,
* and toss data from remaining mbufs on chain.
*/
for (m = mp; m; m = mbuf_next(m)) {
mlen = mbuf_len(m);
if (mlen >= count) {
mlen = count;
mbuf_setlen(m, count);
if (nul > 0) {
cp = (caddr_t)mbuf_data(m) + mlen - nul;
for (i = 0; i < nul; i++) {
*cp++ = '\0';
}
}
break;
}
count -= mlen;
}
for (m = mbuf_next(m); m; m = mbuf_next(m)) {
mbuf_setlen(m, 0);
}
}
/*
* Trim the header out of the mbuf list and trim off any trailing
* junk so that the mbuf list has only the write data.
*/
int
nfsm_chain_trim_data(struct nfsm_chain *nmc, int len, int *mlen)
{
int cnt = 0;
long dlen, adjust;
caddr_t data;
mbuf_t m;
if (mlen) {
*mlen = 0;
}
/* trim header */
for (m = nmc->nmc_mhead; m && (m != nmc->nmc_mcur); m = mbuf_next(m)) {
mbuf_setlen(m, 0);
}
if (!m) {
return EIO;
}
/* trim current mbuf */
data = mbuf_data(m);
dlen = mbuf_len(m);
adjust = nmc->nmc_ptr - data;
dlen -= adjust;
if ((dlen > 0) && (adjust > 0)) {
if (mbuf_setdata(m, nmc->nmc_ptr, dlen)) {
return EIO;
}
} else {
mbuf_setlen(m, dlen);
}
/* skip next len bytes */
for (; m && (cnt < len); m = mbuf_next(m)) {
dlen = mbuf_len(m);
cnt += dlen;
if (cnt > len) {
/* truncate to end of data */
mbuf_setlen(m, dlen - (cnt - len));
if (m == nmc->nmc_mcur) {
nmc->nmc_left -= (cnt - len);
}
cnt = len;
}
}
if (mlen) {
*mlen = cnt;
}
/* trim any trailing data */
if (m == nmc->nmc_mcur) {
nmc->nmc_left = 0;
}
for (; m; m = mbuf_next(m)) {
mbuf_setlen(m, 0);
}
return 0;
}
int
nfsm_chain_add_fattr(
struct nfsrv_descript *nd,
struct nfsm_chain *nmc,
struct vnode_attr *vap)
{
int error = 0;
// XXX Should we assert here that all fields are supported?
nfsm_chain_add_32(error, nmc, vtonfs_type(vap->va_type, nd->nd_vers));
if (nd->nd_vers == NFS_VER3) {
nfsm_chain_add_32(error, nmc, vap->va_mode & 07777);
} else {
nfsm_chain_add_32(error, nmc, vtonfsv2_mode(vap->va_type, vap->va_mode));
}
nfsm_chain_add_32(error, nmc, vap->va_nlink);
nfsm_chain_add_32(error, nmc, vap->va_uid);
nfsm_chain_add_32(error, nmc, vap->va_gid);
if (nd->nd_vers == NFS_VER3) {
nfsm_chain_add_64(error, nmc, vap->va_data_size);
nfsm_chain_add_64(error, nmc, vap->va_data_alloc);
nfsm_chain_add_32(error, nmc, major(vap->va_rdev));
nfsm_chain_add_32(error, nmc, minor(vap->va_rdev));
nfsm_chain_add_64(error, nmc, vap->va_fsid);
nfsm_chain_add_64(error, nmc, vap->va_fileid);
} else {
nfsm_chain_add_32(error, nmc, vap->va_data_size);
nfsm_chain_add_32(error, nmc, NFS_FABLKSIZE);
if (vap->va_type == VFIFO) {
nfsm_chain_add_32(error, nmc, 0xffffffff);
} else {
nfsm_chain_add_32(error, nmc, vap->va_rdev);
}
nfsm_chain_add_32(error, nmc, vap->va_data_alloc / NFS_FABLKSIZE);
nfsm_chain_add_32(error, nmc, vap->va_fsid);
nfsm_chain_add_32(error, nmc, vap->va_fileid);
}
nfsm_chain_add_time(error, nmc, nd->nd_vers, &vap->va_access_time);
nfsm_chain_add_time(error, nmc, nd->nd_vers, &vap->va_modify_time);
nfsm_chain_add_time(error, nmc, nd->nd_vers, &vap->va_change_time);
return error;
}
int
nfsm_chain_get_sattr(
struct nfsrv_descript *nd,
struct nfsm_chain *nmc,
struct vnode_attr *vap)
{
int error = 0;
uint32_t val = 0;
uint64_t val64 = 0;
struct timespec now;
if (nd->nd_vers == NFS_VER2) {
/*
* There is/was a bug in the Sun client that puts 0xffff in the mode
* field of sattr when it should put in 0xffffffff. The u_short
* doesn't sign extend. So check the low order 2 bytes for 0xffff.
*/
nfsm_chain_get_32(error, nmc, val);
if ((val & 0xffff) != 0xffff) {
VATTR_SET(vap, va_mode, val & 07777);
/* save the "type" bits for NFSv2 create */
VATTR_SET(vap, va_type, IFTOVT(val));
VATTR_CLEAR_ACTIVE(vap, va_type);
}
nfsm_chain_get_32(error, nmc, val);
if (val != (uint32_t)-1) {
VATTR_SET(vap, va_uid, val);
}
nfsm_chain_get_32(error, nmc, val);
if (val != (uint32_t)-1) {
VATTR_SET(vap, va_gid, val);
}
/* save the "size" bits for NFSv2 create (even if they appear unset) */
nfsm_chain_get_32(error, nmc, val);
VATTR_SET(vap, va_data_size, val);
if (val == (uint32_t)-1) {
VATTR_CLEAR_ACTIVE(vap, va_data_size);
}
nfsm_chain_get_time(error, nmc, NFS_VER2,
vap->va_access_time.tv_sec,
vap->va_access_time.tv_nsec);
if (vap->va_access_time.tv_sec != -1) {
VATTR_SET_ACTIVE(vap, va_access_time);
}
nfsm_chain_get_time(error, nmc, NFS_VER2,
vap->va_modify_time.tv_sec,
vap->va_modify_time.tv_nsec);
if (vap->va_modify_time.tv_sec != -1) {
VATTR_SET_ACTIVE(vap, va_modify_time);
}
return error;
}
/* NFSv3 */
nfsm_chain_get_32(error, nmc, val);
if (val) {
nfsm_chain_get_32(error, nmc, val);
VATTR_SET(vap, va_mode, val & 07777);
}
nfsm_chain_get_32(error, nmc, val);
if (val) {
nfsm_chain_get_32(error, nmc, val);
VATTR_SET(vap, va_uid, val);
}
nfsm_chain_get_32(error, nmc, val);
if (val) {
nfsm_chain_get_32(error, nmc, val);
VATTR_SET(vap, va_gid, val);
}
nfsm_chain_get_32(error, nmc, val);
if (val) {
nfsm_chain_get_64(error, nmc, val64);
VATTR_SET(vap, va_data_size, val64);
}
nanotime(&now);
nfsm_chain_get_32(error, nmc, val);
switch (val) {
case NFS_TIME_SET_TO_CLIENT:
nfsm_chain_get_time(error, nmc, nd->nd_vers,
vap->va_access_time.tv_sec,
vap->va_access_time.tv_nsec);
VATTR_SET_ACTIVE(vap, va_access_time);
vap->va_vaflags &= ~VA_UTIMES_NULL;
break;
case NFS_TIME_SET_TO_SERVER:
VATTR_SET(vap, va_access_time, now);
vap->va_vaflags |= VA_UTIMES_NULL;
break;
}
nfsm_chain_get_32(error, nmc, val);
switch (val) {
case NFS_TIME_SET_TO_CLIENT:
nfsm_chain_get_time(error, nmc, nd->nd_vers,
vap->va_modify_time.tv_sec,
vap->va_modify_time.tv_nsec);
VATTR_SET_ACTIVE(vap, va_modify_time);
vap->va_vaflags &= ~VA_UTIMES_NULL;
break;
case NFS_TIME_SET_TO_SERVER:
VATTR_SET(vap, va_modify_time, now);
if (!VATTR_IS_ACTIVE(vap, va_access_time)) {
vap->va_vaflags |= VA_UTIMES_NULL;
}
break;
}
return error;
}
/*
* Compare two security flavor structs
*/
int
nfsrv_cmp_secflavs(struct nfs_sec *sf1, struct nfs_sec *sf2)
{
int i;
if (sf1->count != sf2->count) {
return 1;
}
for (i = 0; i < sf1->count; i++) {
if (sf1->flavors[i] != sf2->flavors[i]) {
return 1;
}
}
return 0;
}
/*
* Build hash lists of net addresses and hang them off the NFS export.
* Called by nfsrv_export() to set up the lists of export addresses.
*/
int
nfsrv_hang_addrlist(struct nfs_export *nx, struct user_nfs_export_args *unxa)
{
struct nfs_export_net_args nxna;
struct nfs_netopt *no, *rn_no;
struct radix_node_head *rnh;
struct radix_node *rn;
struct sockaddr *saddr, *smask;
struct domain *dom;
size_t i, ss_minsize;
int error;
unsigned int net;
user_addr_t uaddr;
kauth_cred_t cred;
uaddr = unxa->nxa_nets;
ss_minsize = sizeof(((struct sockaddr_storage *)0)->ss_len) + sizeof(((struct sockaddr_storage *)0)->ss_family);
for (net = 0; net < unxa->nxa_netcount; net++, uaddr += sizeof(nxna)) {
error = copyin(uaddr, &nxna, sizeof(nxna));
if (error) {
return error;
}
if (nxna.nxna_addr.ss_len > sizeof(struct sockaddr_storage) ||
(nxna.nxna_addr.ss_len != 0 && nxna.nxna_addr.ss_len < ss_minsize) ||
nxna.nxna_mask.ss_len > sizeof(struct sockaddr_storage) ||
(nxna.nxna_mask.ss_len != 0 && nxna.nxna_mask.ss_len < ss_minsize) ||
nxna.nxna_addr.ss_family > AF_MAX ||
nxna.nxna_mask.ss_family > AF_MAX) {
return EINVAL;
}
if (nxna.nxna_flags & (NX_MAPROOT | NX_MAPALL)) {
struct posix_cred temp_pcred;
bzero(&temp_pcred, sizeof(temp_pcred));
temp_pcred.cr_uid = nxna.nxna_cred.cr_uid;
temp_pcred.cr_ngroups = nxna.nxna_cred.cr_ngroups;
for (i = 0; i < (size_t)nxna.nxna_cred.cr_ngroups && i < NGROUPS; i++) {
temp_pcred.cr_groups[i] = nxna.nxna_cred.cr_groups[i];
}
cred = posix_cred_create(&temp_pcred);
if (!IS_VALID_CRED(cred)) {
return ENOMEM;
}
} else {
cred = NOCRED;
}
if (nxna.nxna_addr.ss_len == 0) {
/* No address means this is a default/world export */
if (nx->nx_flags & NX_DEFAULTEXPORT) {
if (IS_VALID_CRED(cred)) {
kauth_cred_unref(&cred);
}
return EEXIST;
}
nx->nx_flags |= NX_DEFAULTEXPORT;
nx->nx_defopt.nxo_flags = nxna.nxna_flags;
nx->nx_defopt.nxo_cred = cred;
bcopy(&nxna.nxna_sec, &nx->nx_defopt.nxo_sec, sizeof(struct nfs_sec));
nx->nx_expcnt++;
continue;
}
no = kalloc_type(struct nfs_netopt, Z_WAITOK | Z_ZERO | Z_NOFAIL);
no->no_opt.nxo_flags = nxna.nxna_flags;
no->no_opt.nxo_cred = cred;
bcopy(&nxna.nxna_sec, &no->no_opt.nxo_sec, sizeof(struct nfs_sec));
if (nxna.nxna_addr.ss_len) {
no->no_addr = kalloc_data(nxna.nxna_addr.ss_len, M_WAITOK);
bcopy(&nxna.nxna_addr, no->no_addr, nxna.nxna_addr.ss_len);
}
saddr = no->no_addr;
if (nxna.nxna_mask.ss_len) {
no->no_mask = kalloc_data(nxna.nxna_mask.ss_len, M_WAITOK);
bcopy(&nxna.nxna_mask, no->no_mask, nxna.nxna_mask.ss_len);
}
smask = no->no_mask;
sa_family_t family = saddr->sa_family;
if ((rnh = nx->nx_rtable[family]) == 0) {
/*
* Seems silly to initialize every AF when most are not
* used, do so on demand here
*/
TAILQ_FOREACH(dom, &domains, dom_entry) {
if (dom->dom_family == family && dom->dom_rtattach) {
dom->dom_rtattach((void **)&nx->nx_rtable[family],
dom->dom_rtoffset);
break;
}
}
if ((rnh = nx->nx_rtable[family]) == 0) {
if (IS_VALID_CRED(cred)) {
kauth_cred_unref(&cred);
}
nfs_netopt_free(no);
return ENOBUFS;
}
}
rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh, no->no_rnodes);
if (rn == 0) {
/*
* One of the reasons that rnh_addaddr may fail is that
* the entry already exists. To check for this case, we
* look up the entry to see if it is there. If so, we
* do not need to make a new entry but do continue.
*
* XXX should this be rnh_lookup() instead?
*/
int matched = 0;
rn = (*rnh->rnh_matchaddr)((caddr_t)saddr, rnh);
rn_no = (struct nfs_netopt *)rn;
if (rn != 0 && (rn->rn_flags & RNF_ROOT) == 0 &&
(rn_no->no_opt.nxo_flags == nxna.nxna_flags) &&
(!nfsrv_cmp_secflavs(&rn_no->no_opt.nxo_sec, &nxna.nxna_sec))) {
kauth_cred_t cred2 = rn_no->no_opt.nxo_cred;
if (cred == cred2) {
/* creds are same (or both NULL) */
matched = 1;
} else if (cred && cred2 && (kauth_cred_getuid(cred) == kauth_cred_getuid(cred2))) {
/*
* Now compare the effective and
* supplementary groups...
*
* Note: This comparison, as written,
* does not correctly indicate that
* the groups are equivalent, since
* other than the first supplementary
* group, which is also the effective
* group, order on the remaining groups
* doesn't matter, and this is an
* ordered compare.
*/
gid_t groups[NGROUPS];
gid_t groups2[NGROUPS];
size_t groupcount = NGROUPS;
size_t group2count = NGROUPS;
if (!kauth_cred_getgroups(cred, groups, &groupcount) &&
!kauth_cred_getgroups(cred2, groups2, &group2count) &&
groupcount == group2count) {
for (i = 0; i < group2count; i++) {
if (groups[i] != groups2[i]) {
break;
}
}
if (i >= group2count || i >= NGROUPS) {
matched = 1;
}
}
}
}
if (IS_VALID_CRED(cred)) {
kauth_cred_unref(&cred);
}
nfs_netopt_free(no);
if (matched) {
continue;
}
return EPERM;
}
nx->nx_expcnt++;
}
return 0;
}
/*
* In order to properly track an export's netopt count, we need to pass
* an additional argument to nfsrv_free_netopt() so that it can decrement
* the export's netopt count.
*/
struct nfsrv_free_netopt_arg {
uint32_t *cnt;
struct radix_node_head *rnh;
};
int
nfsrv_free_netopt(struct radix_node *rn, void *w)
{
struct nfsrv_free_netopt_arg *fna = (struct nfsrv_free_netopt_arg *)w;
struct radix_node_head *rnh = fna->rnh;
uint32_t *cnt = fna->cnt;
struct nfs_netopt *nno = (struct nfs_netopt *)rn;
(*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh);
if (IS_VALID_CRED(nno->no_opt.nxo_cred)) {
kauth_cred_unref(&nno->no_opt.nxo_cred);
}
nfs_netopt_free(nno);
*cnt -= 1;
return 0;
}
/*
* Free the net address hash lists that are hanging off the mount points.
*/
int
nfsrv_free_addrlist(struct nfs_export *nx, struct user_nfs_export_args *unxa)
{
struct nfs_export_net_args nxna;
struct radix_node_head *rnh;
struct radix_node *rn;
struct nfsrv_free_netopt_arg fna;
struct nfs_netopt *nno;
size_t ss_minsize;
user_addr_t uaddr;
unsigned int net;
int i, error;
if (!unxa || !unxa->nxa_netcount) {
/* delete everything */
for (i = 0; i <= AF_MAX; i++) {
if ((rnh = nx->nx_rtable[i])) {
fna.rnh = rnh;
fna.cnt = &nx->nx_expcnt;
(*rnh->rnh_walktree)(rnh, nfsrv_free_netopt, (caddr_t)&fna);
zfree(radix_node_head_zone, rnh);
nx->nx_rtable[i] = 0;
}
}
return 0;
}
/* delete only the exports specified */
uaddr = unxa->nxa_nets;
ss_minsize = sizeof(((struct sockaddr_storage *)0)->ss_len) + sizeof(((struct sockaddr_storage *)0)->ss_family);
for (net = 0; net < unxa->nxa_netcount; net++, uaddr += sizeof(nxna)) {
error = copyin(uaddr, &nxna, sizeof(nxna));
if (error) {
return error;
}
if (nxna.nxna_addr.ss_len == 0) {
/* No address means this is a default/world export */
if (nx->nx_flags & NX_DEFAULTEXPORT) {
nx->nx_flags &= ~NX_DEFAULTEXPORT;
if (IS_VALID_CRED(nx->nx_defopt.nxo_cred)) {
kauth_cred_unref(&nx->nx_defopt.nxo_cred);
}
nx->nx_expcnt--;
}
continue;
}
if (nxna.nxna_addr.ss_len > sizeof(struct sockaddr_storage) ||
(nxna.nxna_addr.ss_len != 0 && nxna.nxna_addr.ss_len < ss_minsize) ||
nxna.nxna_addr.ss_family > AF_MAX) {
printf("nfsrv_free_addrlist: invalid socket address (%u)\n", net);
continue;
}
if (nxna.nxna_mask.ss_len > sizeof(struct sockaddr_storage) ||
(nxna.nxna_mask.ss_len != 0 && nxna.nxna_mask.ss_len < ss_minsize) ||
nxna.nxna_mask.ss_family > AF_MAX) {
printf("nfsrv_free_addrlist: invalid socket mask (%u)\n", net);
continue;
}
if ((rnh = nx->nx_rtable[nxna.nxna_addr.ss_family]) == 0) {
/* AF not initialized? */
if (!(unxa->nxa_flags & NXA_ADD)) {
printf("nfsrv_free_addrlist: address not found (0)\n");
}
continue;
}
rn = (*rnh->rnh_lookup)(&nxna.nxna_addr,
nxna.nxna_mask.ss_len ? &nxna.nxna_mask : NULL, rnh);
if (!rn || (rn->rn_flags & RNF_ROOT)) {
if (!(unxa->nxa_flags & NXA_ADD)) {
printf("nfsrv_free_addrlist: address not found (1)\n");
}
continue;
}
(*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh);
nno = (struct nfs_netopt *)rn;
if (IS_VALID_CRED(nno->no_opt.nxo_cred)) {
kauth_cred_unref(&nno->no_opt.nxo_cred);
}
nfs_netopt_free(nno);
nx->nx_expcnt--;
if (nx->nx_expcnt == ((nx->nx_flags & NX_DEFAULTEXPORT) ? 1 : 0)) {
/* no more entries in rnh, so free it up */
zfree(radix_node_head_zone, rnh);
nx->nx_rtable[nxna.nxna_addr.ss_family] = 0;
}
}
return 0;
}
void enablequotas(struct mount *mp, vfs_context_t ctx); // XXX
static int
nfsrv_export_compare(char *path1, char *path2)
{
mount_t mp1 = NULL, mp2 = NULL;
if (strncmp(path1, path2, MAXPATHLEN) == 0) {
return 0;
}
mp1 = nfsrv_getvfs_by_mntonname(path1);
if (mp1) {
vfs_unbusy(mp1);
mp2 = nfsrv_getvfs_by_mntonname(path2);
if (mp2) {
vfs_unbusy(mp2);
if (mp1 == mp2) {
return 0;
}
}
}
return 1;
}
int
nfsrv_export(struct user_nfs_export_args *unxa, vfs_context_t ctx)
{
int error = 0;
size_t pathlen, nxfs_pathlen;
struct nfs_exportfs *nxfs, *nxfs2, *nxfs3;
struct nfs_export *nx, *nx2, *nx3;
struct nfs_filehandle nfh;
struct nameidata mnd, xnd;
vnode_t mvp = NULL, xvp = NULL;
mount_t mp = NULL;
char path[MAXPATHLEN], *nxfs_path;
int expisroot;
if (unxa->nxa_flags == NXA_CHECK) {
/* just check if the path is an NFS-exportable file system */
error = copyinstr(unxa->nxa_fspath, path, MAXPATHLEN, &pathlen);
if (error) {
return error;
}
NDINIT(&mnd, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1,
UIO_SYSSPACE, CAST_USER_ADDR_T(path), ctx);
error = namei(&mnd);
if (error) {
return error;
}
mvp = mnd.ni_vp;
mp = vnode_mount(mvp);
/* make sure it's the root of a file system */
if (!vnode_isvroot(mvp)) {
error = EINVAL;
}
/* make sure the file system is NFS-exportable */
if (!error) {
nfh.nfh_len = NFSV3_MAX_FID_SIZE;
error = VFS_VPTOFH(mvp, (int*)&nfh.nfh_len, &nfh.nfh_fid[0], NULL);
}
if (!error && (nfh.nfh_len > (int)NFSV3_MAX_FID_SIZE)) {
error = EIO;
}
if (!error && !(mp->mnt_vtable->vfc_vfsflags & VFC_VFSREADDIR_EXTENDED)) {
error = EISDIR;
}
vnode_put(mvp);
nameidone(&mnd);
return error;
}
/* all other operations: must be super user */
if ((error = vfs_context_suser(ctx))) {
return error;
}
if (unxa->nxa_flags & NXA_DELETE_ALL) {
/* delete all exports on all file systems */
lck_rw_lock_exclusive(&nfsrv_export_rwlock);
while ((nxfs = LIST_FIRST(&nfsrv_exports))) {
mp = vfs_getvfs_by_mntonname(nxfs->nxfs_path);
if (mp) {
vfs_clearflags(mp, MNT_EXPORTED);
mount_iterdrop(mp);
mp = NULL;
}
/* delete all exports on this file system */
while ((nx = LIST_FIRST(&nxfs->nxfs_exports))) {
LIST_REMOVE(nx, nx_next);
LIST_REMOVE(nx, nx_hash);
/* delete all netopts for this export */
nfsrv_free_addrlist(nx, NULL);
nx->nx_flags &= ~NX_DEFAULTEXPORT;
if (IS_VALID_CRED(nx->nx_defopt.nxo_cred)) {
kauth_cred_unref(&nx->nx_defopt.nxo_cred);
}
/* free active user list for this export */
nfsrv_free_user_list(&nx->nx_user_list);
kfree_data_addr(nx->nx_path);
kfree_type(struct nfs_export, nx);
}
LIST_REMOVE(nxfs, nxfs_next);
kfree_data_addr(nxfs->nxfs_path);
kfree_type(struct nfs_exportfs, nxfs);
}
if (nfsrv_export_hashtbl) {
/* all exports deleted, clean up export hash table */
hashdestroy(nfsrv_export_hashtbl, M_TEMP, nfsrv_export_hash);
nfsrv_export_hash = 0;
nfsrv_export_hashtbl = NULL;
}
lck_rw_done(&nfsrv_export_rwlock);
return 0;
}
error = copyinstr(unxa->nxa_fspath, path, MAXPATHLEN, &pathlen);
if (error) {
return error;
}
lck_rw_lock_exclusive(&nfsrv_export_rwlock);
/* init export hash table if not already */
if (!nfsrv_export_hashtbl) {
if (nfsrv_export_hash_size <= 0) {
nfsrv_export_hash_size = NFSRVEXPHASHSZ;
}
nfsrv_export_hashtbl = hashinit(nfsrv_export_hash_size, M_TEMP, &nfsrv_export_hash);
}
// first check if we've already got an exportfs with the given ID
LIST_FOREACH(nxfs, &nfsrv_exports, nxfs_next) {
if (nxfs->nxfs_id == unxa->nxa_fsid) {
break;
}
}
if (nxfs) {
/* verify exported FS path matches given path */
if (nfsrv_export_compare(path, nxfs->nxfs_path)) {
error = EEXIST;
goto unlock_out;
}
if ((unxa->nxa_flags & (NXA_ADD | NXA_OFFLINE)) == NXA_ADD) {
/* find exported FS root vnode */
NDINIT(&mnd, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1,
UIO_SYSSPACE, CAST_USER_ADDR_T(nxfs->nxfs_path), ctx);
error = namei(&mnd);
if (error) {
goto unlock_out;
}
mvp = mnd.ni_vp;
/* make sure it's (still) the root of a file system */
if (!vnode_isvroot(mvp)) {
error = EINVAL;
goto out;
}
/* if adding, verify that the mount is still what we expect */
mp = nfsrv_getvfs_by_mntonname(nxfs->nxfs_path);
if (mp) {
mount_ref(mp, 0);
vfs_unbusy(mp);
}
/* sanity check: this should be same mount */
if (mp != vnode_mount(mvp)) {
error = EINVAL;
goto out;
}
}
} else {
/* no current exported file system with that ID */
if (!(unxa->nxa_flags & NXA_ADD)) {
error = ENOENT;
goto unlock_out;
}
/* find exported FS root vnode */
NDINIT(&mnd, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1,
UIO_SYSSPACE, CAST_USER_ADDR_T(path), ctx);
error = namei(&mnd);
if (error) {
if (!(unxa->nxa_flags & NXA_OFFLINE)) {
goto unlock_out;
}
} else {
mvp = mnd.ni_vp;
/* make sure it's the root of a file system */
if (!vnode_isvroot(mvp)) {
/* bail if not marked offline */
if (!(unxa->nxa_flags & NXA_OFFLINE)) {
error = EINVAL;
goto out;
}
vnode_put(mvp);
nameidone(&mnd);
mvp = NULL;
} else {
mp = vnode_mount(mvp);
mount_ref(mp, 0);
/* make sure the file system is NFS-exportable */
nfh.nfh_len = NFSV3_MAX_FID_SIZE;
error = VFS_VPTOFH(mvp, (int*)&nfh.nfh_len, &nfh.nfh_fid[0], NULL);
if (!error && (nfh.nfh_len > (int)NFSV3_MAX_FID_SIZE)) {
error = EIO;
}
if (!error && !(mp->mnt_vtable->vfc_vfsflags & VFC_VFSREADDIR_EXTENDED)) {
error = EISDIR;
}
if (error) {
goto out;
}
}
}
/* add an exportfs for it */
nxfs = kalloc_type(struct nfs_exportfs, Z_WAITOK | Z_ZERO | Z_NOFAIL);
nxfs->nxfs_id = unxa->nxa_fsid;
if (mp) {
nxfs_path = mp->mnt_vfsstat.f_mntonname;
nxfs_pathlen = sizeof(mp->mnt_vfsstat.f_mntonname);
} else {
nxfs_path = path;
nxfs_pathlen = pathlen;
}
nxfs->nxfs_path = kalloc_data(nxfs_pathlen, Z_WAITOK);
if (!nxfs->nxfs_path) {
kfree_type(struct nfs_exportfs, nxfs);
error = ENOMEM;
goto out;
}
bcopy(nxfs_path, nxfs->nxfs_path, nxfs_pathlen);
/* insert into list in reverse-sorted order */
nxfs3 = NULL;
LIST_FOREACH(nxfs2, &nfsrv_exports, nxfs_next) {
if (strncmp(nxfs->nxfs_path, nxfs2->nxfs_path, MAXPATHLEN) > 0) {
break;
}
nxfs3 = nxfs2;
}
if (nxfs2) {
LIST_INSERT_BEFORE(nxfs2, nxfs, nxfs_next);
} else if (nxfs3) {
LIST_INSERT_AFTER(nxfs3, nxfs, nxfs_next);
} else {
LIST_INSERT_HEAD(&nfsrv_exports, nxfs, nxfs_next);
}
/* make sure any quotas are enabled before we export the file system */
if (mp) {
enablequotas(mp, ctx);
}
}
if (unxa->nxa_exppath) {
error = copyinstr(unxa->nxa_exppath, path, MAXPATHLEN, &pathlen);
if (error) {
goto out;
}
LIST_FOREACH(nx, &nxfs->nxfs_exports, nx_next) {
if (nx->nx_id == unxa->nxa_expid) {
break;
}
}
if (nx) {
/* verify exported FS path matches given path */
if (strncmp(path, nx->nx_path, MAXPATHLEN)) {
error = EEXIST;
goto out;
}
} else {
/* no current export with that ID */
if (!(unxa->nxa_flags & NXA_ADD)) {
error = ENOENT;
goto out;
}
/* add an export for it */
nx = kalloc_type(struct nfs_export, Z_WAITOK | Z_ZERO | Z_NOFAIL);
nx->nx_id = unxa->nxa_expid;
nx->nx_fs = nxfs;
microtime(&nx->nx_exptime);
nx->nx_path = kalloc_data(pathlen, Z_WAITOK);
if (!nx->nx_path) {
error = ENOMEM;
kfree_type(struct nfs_export, nx);
nx = NULL;
goto out1;
}
bcopy(path, nx->nx_path, pathlen);
/* initialize the active user list */
nfsrv_init_user_list(&nx->nx_user_list);
/* insert into list in reverse-sorted order */
nx3 = NULL;
LIST_FOREACH(nx2, &nxfs->nxfs_exports, nx_next) {
if (strncmp(nx->nx_path, nx2->nx_path, MAXPATHLEN) > 0) {
break;
}
nx3 = nx2;
}
if (nx2) {
LIST_INSERT_BEFORE(nx2, nx, nx_next);
} else if (nx3) {
LIST_INSERT_AFTER(nx3, nx, nx_next);
} else {
LIST_INSERT_HEAD(&nxfs->nxfs_exports, nx, nx_next);
}
/* insert into hash */
LIST_INSERT_HEAD(NFSRVEXPHASH(nxfs->nxfs_id, nx->nx_id), nx, nx_hash);
/*
* We don't allow/support nested exports. Check if the new entry
* nests with the entries before and after or if there's an
* entry for the file system root and subdirs.
*/
error = 0;
if ((nx3 && !strncmp(nx3->nx_path, nx->nx_path, pathlen - 1) &&
(nx3->nx_path[pathlen - 1] == '/')) ||
(nx2 && !strncmp(nx2->nx_path, nx->nx_path, strlen(nx2->nx_path)) &&
(nx->nx_path[strlen(nx2->nx_path)] == '/'))) {
error = EINVAL;
}
if (!error) {
/* check export conflict with fs root export and vice versa */
expisroot = !nx->nx_path[0] ||
((nx->nx_path[0] == '.') && !nx->nx_path[1]);
LIST_FOREACH(nx2, &nxfs->nxfs_exports, nx_next) {
if (expisroot) {
if (nx2 != nx) {
break;
}
} else if (!nx2->nx_path[0]) {
break;
} else if ((nx2->nx_path[0] == '.') && !nx2->nx_path[1]) {
break;
}
}
if (nx2) {
error = EINVAL;
}
}
if (error) {
/*
* Don't actually return an error because mountd is
* probably about to delete the conflicting export.
* This can happen when a new export momentarily conflicts
* with an old export while the transition is being made.
* Theoretically, mountd could be written to avoid this
* transient situation - but it would greatly increase the
* complexity of mountd for very little overall benefit.
*/
printf("nfsrv_export: warning: nested exports: %s/%s\n",
nxfs->nxfs_path, nx->nx_path);
error = 0;
}
nx->nx_fh.nfh_xh.nxh_flags = NXHF_INVALIDFH;
}
/* make sure file handle is set up */
if ((nx->nx_fh.nfh_xh.nxh_version != htonl(NFS_FH_VERSION)) ||
(nx->nx_fh.nfh_xh.nxh_flags & NXHF_INVALIDFH)) {
/* try to set up export root file handle */
nx->nx_fh.nfh_xh.nxh_version = htonl(NFS_FH_VERSION);
nx->nx_fh.nfh_xh.nxh_fsid = htonl(nx->nx_fs->nxfs_id);
nx->nx_fh.nfh_xh.nxh_expid = htonl(nx->nx_id);
nx->nx_fh.nfh_xh.nxh_flags = 0;
nx->nx_fh.nfh_xh.nxh_reserved = 0;
nx->nx_fh.nfh_fhp = (u_char*)&nx->nx_fh.nfh_xh;
bzero(&nx->nx_fh.nfh_fid[0], NFSV2_MAX_FID_SIZE);
if (mvp) {
/* find export root vnode */
if (!nx->nx_path[0] || ((nx->nx_path[0] == '.') && !nx->nx_path[1])) {
/* exporting file system's root directory */
xvp = mvp;
vnode_get(xvp);
} else {
NDINIT(&xnd, LOOKUP, OP_LOOKUP, LOCKLEAF, UIO_SYSSPACE, CAST_USER_ADDR_T(path), ctx);
xnd.ni_pathlen = (uint32_t)pathlen - 1; // pathlen max value is equal to MAXPATHLEN
xnd.ni_cnd.cn_nameptr = xnd.ni_cnd.cn_pnbuf = path;
xnd.ni_startdir = mvp;
xnd.ni_usedvp = mvp;
xnd.ni_rootdir = rootvnode;
while ((error = lookup(&xnd)) == ERECYCLE) {
xnd.ni_cnd.cn_flags = LOCKLEAF;
xnd.ni_cnd.cn_nameptr = xnd.ni_cnd.cn_pnbuf;
xnd.ni_usedvp = xnd.ni_dvp = xnd.ni_startdir = mvp;
}
if (error) {
goto out1;
}
xvp = xnd.ni_vp;
}
if (vnode_vtype(xvp) != VDIR) {
error = EINVAL;
vnode_put(xvp);
goto out1;
}
/* grab file handle */
nx->nx_fh.nfh_len = NFSV3_MAX_FID_SIZE;
error = VFS_VPTOFH(xvp, (int*)&nx->nx_fh.nfh_len, &nx->nx_fh.nfh_fid[0], NULL);
if (!error && (nx->nx_fh.nfh_len > (int)NFSV3_MAX_FID_SIZE)) {
error = EIO;
} else {
nx->nx_fh.nfh_xh.nxh_fidlen = nx->nx_fh.nfh_len;
nx->nx_fh.nfh_len += sizeof(nx->nx_fh.nfh_xh);
}
vnode_put(xvp);
if (error) {
goto out1;
}
} else {
nx->nx_fh.nfh_xh.nxh_flags = NXHF_INVALIDFH;
nx->nx_fh.nfh_xh.nxh_fidlen = 0;
nx->nx_fh.nfh_len = sizeof(nx->nx_fh.nfh_xh);
}
}
} else {
nx = NULL;
}
/* perform the export changes */
if (unxa->nxa_flags & NXA_DELETE) {
if (!nx) {
/* delete all exports on this file system */
while ((nx = LIST_FIRST(&nxfs->nxfs_exports))) {
LIST_REMOVE(nx, nx_next);
LIST_REMOVE(nx, nx_hash);
/* delete all netopts for this export */
nfsrv_free_addrlist(nx, NULL);
nx->nx_flags &= ~NX_DEFAULTEXPORT;
if (IS_VALID_CRED(nx->nx_defopt.nxo_cred)) {
kauth_cred_unref(&nx->nx_defopt.nxo_cred);
}
/* delete active user list for this export */
nfsrv_free_user_list(&nx->nx_user_list);
kfree_data_addr(nx->nx_path);
kfree_type(struct nfs_export, nx);
}
goto out1;
} else if (!unxa->nxa_netcount) {
/* delete all netopts for this export */
nfsrv_free_addrlist(nx, NULL);
nx->nx_flags &= ~NX_DEFAULTEXPORT;
if (IS_VALID_CRED(nx->nx_defopt.nxo_cred)) {
kauth_cred_unref(&nx->nx_defopt.nxo_cred);
}
} else {
/* delete only the netopts for the given addresses */
error = nfsrv_free_addrlist(nx, unxa);
if (error) {
goto out1;
}
}
}
if (unxa->nxa_flags & NXA_ADD) {
/*
* If going offline set the export time so that when
* coming back on line we will present a new write verifier
* to the client.
*/
if (unxa->nxa_flags & NXA_OFFLINE) {
microtime(&nx->nx_exptime);
}
error = nfsrv_hang_addrlist(nx, unxa);
if (!error && mp) {
vfs_setflags(mp, MNT_EXPORTED);
}
}
out1:
if (nx && !nx->nx_expcnt) {
/* export has no export options */
LIST_REMOVE(nx, nx_next);
LIST_REMOVE(nx, nx_hash);
/* delete active user list for this export */
nfsrv_free_user_list(&nx->nx_user_list);
kfree_data_addr(nx->nx_path);
kfree_type(struct nfs_export, nx);
}
if (LIST_EMPTY(&nxfs->nxfs_exports)) {
/* exported file system has no more exports */
LIST_REMOVE(nxfs, nxfs_next);
kfree_data_addr(nxfs->nxfs_path);
kfree_type(struct nfs_exportfs, nxfs);
if (mp) {
vfs_clearflags(mp, MNT_EXPORTED);
}
}
out:
if (mvp) {
vnode_put(mvp);
nameidone(&mnd);
}
unlock_out:
if (mp) {
mount_drop(mp, 0);
}
lck_rw_done(&nfsrv_export_rwlock);
return error;
}
/*
* Check if there is a least one export that will allow this address.
*
* Return 0, if there is an export that will allow this address,
* else return EACCES
*/
int
nfsrv_check_exports_allow_address(mbuf_t nam)
{
struct nfs_exportfs *nxfs;
struct nfs_export *nx;
struct nfs_export_options *nxo = NULL;
if (nam == NULL) {
return EACCES;
}
lck_rw_lock_shared(&nfsrv_export_rwlock);
LIST_FOREACH(nxfs, &nfsrv_exports, nxfs_next) {
LIST_FOREACH(nx, &nxfs->nxfs_exports, nx_next) {
/* A little optimizing by checking for the default first */
if (nx->nx_flags & NX_DEFAULTEXPORT) {
nxo = &nx->nx_defopt;
}
if (nxo || (nxo = nfsrv_export_lookup(nx, nam))) {
goto found;
}
}
}
found:
lck_rw_done(&nfsrv_export_rwlock);
return nxo ? 0 : EACCES;
}
struct nfs_export_options *
nfsrv_export_lookup(struct nfs_export *nx, mbuf_t nam)
{
struct nfs_export_options *nxo = NULL;
struct nfs_netopt *no = NULL;
struct radix_node_head *rnh;
struct sockaddr *saddr;
/* Lookup in the export list first. */
if (nam != NULL) {
saddr = mbuf_data(nam);
if (saddr->sa_family > AF_MAX) {
/* Bogus sockaddr? Don't match anything. */
return NULL;
}
rnh = nx->nx_rtable[saddr->sa_family];
if (rnh != NULL) {
no = (struct nfs_netopt *)
(*rnh->rnh_matchaddr)((caddr_t)saddr, rnh);
if (no && no->no_rnodes->rn_flags & RNF_ROOT) {
no = NULL;
}
if (no) {
nxo = &no->no_opt;
}
}
}
/* If no address match, use the default if it exists. */
if ((nxo == NULL) && (nx->nx_flags & NX_DEFAULTEXPORT)) {
nxo = &nx->nx_defopt;
}
return nxo;
}
/* find an export for the given handle */
struct nfs_export *
nfsrv_fhtoexport(struct nfs_filehandle *nfhp)
{
struct nfs_exphandle *nxh = (struct nfs_exphandle*)nfhp->nfh_fhp;
struct nfs_export *nx;
uint32_t fsid, expid;
if (!nfsrv_export_hashtbl) {
return NULL;
}
fsid = ntohl(nxh->nxh_fsid);
expid = ntohl(nxh->nxh_expid);
nx = NFSRVEXPHASH(fsid, expid)->lh_first;
for (; nx; nx = LIST_NEXT(nx, nx_hash)) {
if (nx->nx_fs->nxfs_id != fsid) {
continue;
}
if (nx->nx_id != expid) {
continue;
}
break;
}
return nx;
}
struct nfsrv_getvfs_by_mntonname_callback_args {
const char *path; /* IN */
mount_t mp; /* OUT */
};
static int
nfsrv_getvfs_by_mntonname_callback(mount_t mp, void *v)
{
struct nfsrv_getvfs_by_mntonname_callback_args * const args = v;
char real_mntonname[MAXPATHLEN];
size_t pathbuflen = MAXPATHLEN;
vnode_t rvp;
int error;
error = VFS_ROOT(mp, &rvp, vfs_context_current());
if (error) {
goto out;
}
error = vn_getpath_ext(rvp, NULLVP, real_mntonname, &pathbuflen,
VN_GETPATH_FSENTER | VN_GETPATH_NO_FIRMLINK);
vnode_put(rvp);
if (error) {
goto out;
}
if (strcmp(args->path, real_mntonname) == 0) {
error = vfs_busy(mp, LK_NOWAIT);
if (error == 0) {
args->mp = mp;
}
return VFS_RETURNED_DONE;
}
out:
return VFS_RETURNED;
}
static mount_t
nfsrv_getvfs_by_mntonname(char *path)
{
struct nfsrv_getvfs_by_mntonname_callback_args args = {
.path = path,
.mp = NULL,
};
mount_t mp;
int error;
mp = vfs_getvfs_by_mntonname(path);
if (mp) {
error = vfs_busy(mp, LK_NOWAIT);
mount_iterdrop(mp);
if (error) {
mp = NULL;
}
} else if (vfs_iterate(0, nfsrv_getvfs_by_mntonname_callback,
&args) == 0) {
mp = args.mp;
}
return mp;
}
/*
* nfsrv_fhtovp() - convert FH to vnode and export info
*/
int
nfsrv_fhtovp(
struct nfs_filehandle *nfhp,
struct nfsrv_descript *nd,
vnode_t *vpp,
struct nfs_export **nxp,
struct nfs_export_options **nxop)
{
struct nfs_exphandle *nxh = (struct nfs_exphandle*)nfhp->nfh_fhp;
struct nfs_export_options *nxo;
u_char *fidp;
int error;
struct mount *mp;
mbuf_t nam = NULL;
uint32_t v;
int i, valid;
*vpp = NULL;
*nxp = NULL;
*nxop = NULL;
if (nd != NULL) {
nam = nd->nd_nam;
}
v = ntohl(nxh->nxh_version);
if (v != NFS_FH_VERSION) {
/* file handle format not supported */
return ESTALE;
}
if (nfhp->nfh_len > NFSV3_MAX_FH_SIZE) {
return EBADRPC;
}
if (nfhp->nfh_len < (int)sizeof(struct nfs_exphandle)) {
return ESTALE;
}
v = ntohs(nxh->nxh_flags);
if (v & NXHF_INVALIDFH) {
return ESTALE;
}
*nxp = nfsrv_fhtoexport(nfhp);
if (!*nxp) {
return ESTALE;
}
/* Get the export option structure for this <export, client> tuple. */
*nxop = nxo = nfsrv_export_lookup(*nxp, nam);
if (nam && (*nxop == NULL)) {
return EACCES;
}
if (nd != NULL) {
/* Validate the security flavor of the request */
for (i = 0, valid = 0; i < nxo->nxo_sec.count; i++) {
if (nd->nd_sec == nxo->nxo_sec.flavors[i]) {
valid = 1;
break;
}
}
if (!valid) {
/*
* RFC 2623 section 2.3.2 recommends no authentication
* requirement for certain NFS procedures used for mounting.
* This allows an unauthenticated superuser on the client
* to do mounts for the benefit of authenticated users.
*/
if (nd->nd_vers == NFS_VER2) {
if (nd->nd_procnum == NFSV2PROC_GETATTR ||
nd->nd_procnum == NFSV2PROC_STATFS) {
valid = 1;
}
}
if (nd->nd_vers == NFS_VER3) {
if (nd->nd_procnum == NFSPROC_FSINFO) {
valid = 1;
}
}
if (!valid) {
return NFSERR_AUTHERR | AUTH_REJECTCRED;
}
}
}
if (nxo && (nxo->nxo_flags & NX_OFFLINE)) {
return (nd == NULL || nd->nd_vers == NFS_VER2) ? ESTALE : NFSERR_TRYLATER;
}
/* find mount structure */
mp = nfsrv_getvfs_by_mntonname((*nxp)->nx_fs->nxfs_path);
if (!mp) {
/*
* We have an export, but no mount?
* Perhaps the export just hasn't been marked offline yet.
*/
return (nd == NULL || nd->nd_vers == NFS_VER2) ? ESTALE : NFSERR_TRYLATER;
}
fidp = nfhp->nfh_fhp + sizeof(*nxh);
error = VFS_FHTOVP(mp, nxh->nxh_fidlen, fidp, vpp, NULL);
vfs_unbusy(mp);
if (error) {
return error;
}
/* vnode pointer should be good at this point or ... */
if (*vpp == NULL) {
return ESTALE;
}
return 0;
}
/*
* nfsrv_credcheck() - check/map credentials according
* to given export options.
*/
int
nfsrv_credcheck(
struct nfsrv_descript *nd,
vfs_context_t ctx,
__unused struct nfs_export *nx,
struct nfs_export_options *nxo)
{
if (nxo && nxo->nxo_cred) {
if ((nxo->nxo_flags & NX_MAPALL) ||
((nxo->nxo_flags & NX_MAPROOT) && !suser(nd->nd_cr, NULL))) {
kauth_cred_ref(nxo->nxo_cred);
kauth_cred_unref(&nd->nd_cr);
nd->nd_cr = nxo->nxo_cred;
}
}
ctx->vc_ucred = nd->nd_cr;
return 0;
}
/*
* nfsrv_vptofh() - convert vnode to file handle for given export
*
* If the caller is passing in a vnode for a ".." directory entry,
* they can pass a directory NFS file handle (dnfhp) which will be
* checked against the root export file handle. If it matches, we
* refuse to provide the file handle for the out-of-export directory.
*/
int
nfsrv_vptofh(
struct nfs_export *nx,
int nfsvers,
struct nfs_filehandle *dnfhp,
vnode_t vp,
vfs_context_t ctx,
struct nfs_filehandle *nfhp)
{
int error;
uint32_t maxfidsize;
nfhp->nfh_fhp = (u_char*)&nfhp->nfh_xh;
nfhp->nfh_xh.nxh_version = htonl(NFS_FH_VERSION);
nfhp->nfh_xh.nxh_fsid = htonl(nx->nx_fs->nxfs_id);
nfhp->nfh_xh.nxh_expid = htonl(nx->nx_id);
nfhp->nfh_xh.nxh_flags = 0;
nfhp->nfh_xh.nxh_reserved = 0;
if (nfsvers == NFS_VER2) {
bzero(&nfhp->nfh_fid[0], NFSV2_MAX_FID_SIZE);
}
/* if directory FH matches export root, return invalid FH */
if (dnfhp && nfsrv_fhmatch(dnfhp, &nx->nx_fh)) {
if (nfsvers == NFS_VER2) {
nfhp->nfh_len = NFSX_V2FH;
} else {
nfhp->nfh_len = sizeof(nfhp->nfh_xh);
}
nfhp->nfh_xh.nxh_fidlen = 0;
nfhp->nfh_xh.nxh_flags = htons(NXHF_INVALIDFH);
return 0;
}
if (nfsvers == NFS_VER2) {
maxfidsize = NFSV2_MAX_FID_SIZE;
} else {
maxfidsize = NFSV3_MAX_FID_SIZE;
}
nfhp->nfh_len = maxfidsize;
error = VFS_VPTOFH(vp, (int*)&nfhp->nfh_len, &nfhp->nfh_fid[0], ctx);
if (error) {
return error;
}
if (nfhp->nfh_len > maxfidsize) {
return EOVERFLOW;
}
nfhp->nfh_xh.nxh_fidlen = nfhp->nfh_len;
nfhp->nfh_len += sizeof(nfhp->nfh_xh);
if ((nfsvers == NFS_VER2) && (nfhp->nfh_len < NFSX_V2FH)) {
nfhp->nfh_len = NFSX_V2FH;
}
return 0;
}
/*
* Compare two file handles to see it they're the same.
* Note that we don't use nfh_len because that may include
* padding in an NFSv2 file handle.
*/
int
nfsrv_fhmatch(struct nfs_filehandle *fh1, struct nfs_filehandle *fh2)
{
struct nfs_exphandle *nxh1, *nxh2;
int len1, len2;
nxh1 = (struct nfs_exphandle *)fh1->nfh_fhp;
nxh2 = (struct nfs_exphandle *)fh2->nfh_fhp;
len1 = sizeof(fh1->nfh_xh) + nxh1->nxh_fidlen;
len2 = sizeof(fh2->nfh_xh) + nxh2->nxh_fidlen;
if (len1 != len2) {
return 0;
}
if (bcmp(nxh1, nxh2, len1)) {
return 0;
}
return 1;
}
/*
* Functions for dealing with active user lists
*/
/*
* Search the hash table for a user node with a matching IP address and uid field.
* If found, the node's tm_last timestamp is updated and the node is returned.
*
* If not found, a new node is allocated (or reclaimed via LRU), initialized, and returned.
* Returns NULL if a new node could not be allocated OR saddr length exceeds sizeof(unode->sock).
*
* The list's user_mutex lock MUST be held.
*/
struct nfs_user_stat_node *
nfsrv_get_user_stat_node(struct nfs_active_user_list *list, struct sockaddr *saddr, uid_t uid)
{
struct nfs_user_stat_node *unode;
struct timeval now;
struct nfs_user_stat_hashtbl_head *head;
/* seach the hash table */
head = NFS_USER_STAT_HASH(list->user_hashtbl, uid);
LIST_FOREACH(unode, head, hash_link) {
if ((uid == unode->uid) && (nfs_sockaddr_cmp(saddr, (struct sockaddr*)&unode->sock) == 0)) {
/* found matching node */
break;
}
}
if (unode) {
/* found node in the hash table, now update lru position */
TAILQ_REMOVE(&list->user_lru, unode, lru_link);
TAILQ_INSERT_TAIL(&list->user_lru, unode, lru_link);
/* update time stamp */
microtime(&now);
unode->tm_last = (uint32_t)now.tv_sec;
return unode;
}
if (saddr->sa_len > sizeof(((struct nfs_user_stat_node *)0)->sock)) {
/* saddr length exceeds maximum value */
return NULL;
}
if (list->node_count < nfsrv_user_stat_max_nodes) {
/* Allocate a new node */
unode = kalloc_type(struct nfs_user_stat_node,
Z_WAITOK | Z_ZERO | Z_NOFAIL);
/* increment node count */
OSAddAtomic(1, &nfsrv_user_stat_node_count);
list->node_count++;
} else {
/* reuse the oldest node in the lru list */
unode = TAILQ_FIRST(&list->user_lru);
if (!unode) {
return NULL;
}
/* Remove the node */
TAILQ_REMOVE(&list->user_lru, unode, lru_link);
LIST_REMOVE(unode, hash_link);
}
/* Initialize the node */
unode->uid = uid;
bcopy(saddr, &unode->sock, MIN(saddr->sa_len, sizeof(unode->sock)));
microtime(&now);
unode->ops = 0;
unode->bytes_read = 0;
unode->bytes_written = 0;
unode->tm_start = (uint32_t)now.tv_sec;
unode->tm_last = (uint32_t)now.tv_sec;
/* insert the node */
TAILQ_INSERT_TAIL(&list->user_lru, unode, lru_link);
LIST_INSERT_HEAD(head, unode, hash_link);
return unode;
}
void
nfsrv_update_user_stat(struct nfs_export *nx, struct nfsrv_descript *nd, uid_t uid, u_int ops, u_int rd_bytes, u_int wr_bytes)
{
struct nfs_user_stat_node *unode;
struct nfs_active_user_list *ulist;
struct sockaddr *saddr;
if ((!nfsrv_user_stat_enabled) || (!nx) || (!nd) || (!nd->nd_nam)) {
return;
}
saddr = (struct sockaddr *)mbuf_data(nd->nd_nam);
/* check address family before going any further */
if ((saddr->sa_family != AF_INET) && (saddr->sa_family != AF_INET6)) {
return;
}
ulist = &nx->nx_user_list;
/* lock the active user list */
lck_mtx_lock(&ulist->user_mutex);
/* get the user node */
unode = nfsrv_get_user_stat_node(ulist, saddr, uid);
if (!unode) {
lck_mtx_unlock(&ulist->user_mutex);
return;
}
/* update counters */
unode->ops += ops;
unode->bytes_read += rd_bytes;
unode->bytes_written += wr_bytes;
/* done */
lck_mtx_unlock(&ulist->user_mutex);
}
/* initialize an active user list */
void
nfsrv_init_user_list(struct nfs_active_user_list *ulist)
{
uint i;
/* initialize the lru */
TAILQ_INIT(&ulist->user_lru);
/* initialize the hash table */
for (i = 0; i < NFS_USER_STAT_HASH_SIZE; i++) {
LIST_INIT(&ulist->user_hashtbl[i]);
}
ulist->node_count = 0;
lck_mtx_init(&ulist->user_mutex, &nfsrv_active_user_mutex_group, LCK_ATTR_NULL);
}
/* Free all nodes in an active user list */
void
nfsrv_free_user_list(struct nfs_active_user_list *ulist)
{
struct nfs_user_stat_node *unode;
if (!ulist) {
return;
}
while ((unode = TAILQ_FIRST(&ulist->user_lru))) {
/* Remove node and free */
TAILQ_REMOVE(&ulist->user_lru, unode, lru_link);
LIST_REMOVE(unode, hash_link);
kfree_type(struct nfs_user_stat_node, unode);
/* decrement node count */
OSAddAtomic(-1, &nfsrv_user_stat_node_count);
}
ulist->node_count = 0;
lck_mtx_destroy(&ulist->user_mutex, &nfsrv_active_user_mutex_group);
}
/* Reclaim old expired user nodes from active user lists. */
void
nfsrv_active_user_list_reclaim(void)
{
struct nfs_exportfs *nxfs;
struct nfs_export *nx;
struct nfs_active_user_list *ulist;
struct nfs_user_stat_hashtbl_head oldlist;
struct nfs_user_stat_node *unode, *unode_next;
struct timeval now;
long tstale;
LIST_INIT(&oldlist);
lck_rw_lock_shared(&nfsrv_export_rwlock);
microtime(&now);
tstale = now.tv_sec - nfsrv_user_stat_max_idle_sec;
LIST_FOREACH(nxfs, &nfsrv_exports, nxfs_next) {
LIST_FOREACH(nx, &nxfs->nxfs_exports, nx_next) {
/* Scan through all user nodes of this export */
ulist = &nx->nx_user_list;
lck_mtx_lock(&ulist->user_mutex);
for (unode = TAILQ_FIRST(&ulist->user_lru); unode; unode = unode_next) {
unode_next = TAILQ_NEXT(unode, lru_link);
/* check if this node has expired */
if (unode->tm_last >= tstale) {
break;
}
/* Remove node from the active user list */
TAILQ_REMOVE(&ulist->user_lru, unode, lru_link);
LIST_REMOVE(unode, hash_link);
/* Add node to temp list */
LIST_INSERT_HEAD(&oldlist, unode, hash_link);
/* decrement node count */
OSAddAtomic(-1, &nfsrv_user_stat_node_count);
ulist->node_count--;
}
/* can unlock this export's list now */
lck_mtx_unlock(&ulist->user_mutex);
}
}
lck_rw_done(&nfsrv_export_rwlock);
/* Free expired nodes */
while ((unode = LIST_FIRST(&oldlist))) {
LIST_REMOVE(unode, hash_link);
kfree_type(struct nfs_user_stat_node, unode);
}
}
/*
* Maps errno values to nfs error numbers.
* Use NFSERR_IO as the catch all for ones not specifically defined in
* RFC 1094.
*/
static u_char nfsrv_v2errmap[] = {
NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO,
NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO,
NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR,
NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS,
NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO,
NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE,
};
/*
* Maps errno values to nfs error numbers.
* Although it is not obvious whether or not NFS clients really care if
* a returned error value is in the specified list for the procedure, the
* safest thing to do is filter them appropriately. For Version 2, the
* X/Open XNFS document is the only specification that defines error values
* for each RPC (The RFC simply lists all possible error values for all RPCs),
* so I have decided to not do this for Version 2.
* The first entry is the default error return and the rest are the valid
* errors for that RPC in increasing numeric order.
*/
static short nfsv3err_null[] = {
0,
0,
};
static short nfsv3err_getattr[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_setattr[] = {
NFSERR_IO,
NFSERR_PERM,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_INVAL,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOT_SYNC,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_lookup[] = {
NFSERR_IO,
NFSERR_NOENT,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_NOTDIR,
NFSERR_NAMETOL,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_access[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_readlink[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_INVAL,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_read[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_NXIO,
NFSERR_ACCES,
NFSERR_INVAL,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_write[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_INVAL,
NFSERR_FBIG,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_create[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_EXIST,
NFSERR_NOTDIR,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_NAMETOL,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_mkdir[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_EXIST,
NFSERR_NOTDIR,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_NAMETOL,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_symlink[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_EXIST,
NFSERR_NOTDIR,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_NAMETOL,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_mknod[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_EXIST,
NFSERR_NOTDIR,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_NAMETOL,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
NFSERR_BADTYPE,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_remove[] = {
NFSERR_IO,
NFSERR_NOENT,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_NOTDIR,
NFSERR_ISDIR,
NFSERR_ROFS,
NFSERR_NAMETOL,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_rmdir[] = {
NFSERR_IO,
NFSERR_NOENT,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_EXIST,
NFSERR_NOTDIR,
NFSERR_INVAL,
NFSERR_ROFS,
NFSERR_NAMETOL,
NFSERR_NOTEMPTY,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_rename[] = {
NFSERR_IO,
NFSERR_NOENT,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_EXIST,
NFSERR_XDEV,
NFSERR_NOTDIR,
NFSERR_ISDIR,
NFSERR_INVAL,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_MLINK,
NFSERR_NAMETOL,
NFSERR_NOTEMPTY,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_link[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_EXIST,
NFSERR_XDEV,
NFSERR_NOTDIR,
NFSERR_ISDIR,
NFSERR_INVAL,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_MLINK,
NFSERR_NAMETOL,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_readdir[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_NOTDIR,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_BAD_COOKIE,
NFSERR_TOOSMALL,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_readdirplus[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_NOTDIR,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_BAD_COOKIE,
NFSERR_NOTSUPP,
NFSERR_TOOSMALL,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_fsstat[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_fsinfo[] = {
NFSERR_STALE,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_pathconf[] = {
NFSERR_STALE,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
NFSERR_TRYLATER,
0,
};
static short nfsv3err_commit[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
NFSERR_BADTYPE,
NFSERR_TRYLATER,
0,
};
static short *nfsrv_v3errmap[] = {
nfsv3err_null,
nfsv3err_getattr,
nfsv3err_setattr,
nfsv3err_lookup,
nfsv3err_access,
nfsv3err_readlink,
nfsv3err_read,
nfsv3err_write,
nfsv3err_create,
nfsv3err_mkdir,
nfsv3err_symlink,
nfsv3err_mknod,
nfsv3err_remove,
nfsv3err_rmdir,
nfsv3err_rename,
nfsv3err_link,
nfsv3err_readdir,
nfsv3err_readdirplus,
nfsv3err_fsstat,
nfsv3err_fsinfo,
nfsv3err_pathconf,
nfsv3err_commit,
};
/*
* Map errnos to NFS error numbers. For Version 3 also filter out error
* numbers not specified for the associated procedure.
*/
int
nfsrv_errmap(struct nfsrv_descript *nd, int err)
{
short *defaulterrp, *errp;
if (nd->nd_vers == NFS_VER2) {
if (err <= (int)sizeof(nfsrv_v2errmap)) {
return (int)nfsrv_v2errmap[err - 1];
}
return NFSERR_IO;
}
/* NFSv3 */
if (nd->nd_procnum > NFSPROC_COMMIT) {
return err & 0xffff;
}
errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
while (*++errp) {
if (*errp == err) {
return err;
} else if (*errp > err) {
break;
}
}
return (int)*defaulterrp;
}
#endif /* CONFIG_NFS_SERVER */