freebsd-dev/usr.sbin/rpc.lockd/kern.c
Alfred Perlstein 64b829d192 leverage fixed fifo implementation (rev 1.56 of
src/sys/fs/fifofs/fifo_vnops.c) to serve locks better, my previous
workaround for lack of decent fifo system wasn't cutting it,
particularly the kernel would send a message down the fifo and
immediately close it, this would lead to delayed unlock requests
being seen by the lockd causing all sorts of badness.

Basically, don't reopen the fifo, just select(2) on it.
2001-11-08 10:37:32 +00:00

578 lines
15 KiB
C

/*-
* Copyright (c) 1997 Berkeley Software Design, Inc. 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.
* 3. Berkeley Software Design Inc's name may not be used to endorse or
* promote products derived from this software without specific prior
* written permission.
*
* THIS SOFTWARE IS PROVIDED BY BERKELEY SOFTWARE DESIGN INC ``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 BERKELEY SOFTWARE DESIGN INC 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.
*
* from BSDI kern.c,v 1.2 1998/11/25 22:38:27 don Exp
* $FreeBSD$
*/
#include <sys/param.h>
#include <sys/mount.h>
#include <sys/queue.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <err.h>
#include <errno.h>
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <syslog.h>
#include <unistd.h>
#include "nlm_prot.h"
#include <nfs/rpcv2.h>
#include <nfs/nfsproto.h>
#include <nfsclient/nfs_lock.h>
#include "lockd.h"
#include "lockd_lock.h"
#include <nfsclient/nfs.h>
#define nfslockdans(_v, _ansp) \
((_ansp)->la_vers = _v, \
nfsclnt(NFSCLNT_LOCKDANS, _ansp))
/* Lock request owner. */
typedef struct __owner {
pid_t pid; /* Process ID. */
time_t tod; /* Time-of-day. */
} OWNER;
static OWNER owner;
static char hostname[MAXHOSTNAMELEN + 1]; /* Hostname. */
int lock_request(LOCKD_MSG *);
int test_request(LOCKD_MSG *);
void show(LOCKD_MSG *);
int unlock_request(LOCKD_MSG *);
/*
* will break because fifo needs to be repopened when EOF'd
*/
#ifdef SETUID_DAEMON
#define lockd_seteuid(uid) seteuid(uid)
#else
#define lockd_seteuid(uid) (1)
#endif
#define d_calls (debug_level > 1)
#define d_args (debug_level > 2)
#define from_addr(sockaddr) \
(inet_ntoa((sockaddr)->sin_addr))
void
client_cleanup(sig, code)
int sig;
int code;
{
(void)lockd_seteuid(0);
(void)unlink(_PATH_LCKFIFO);
exit(-1);
}
/*
* client_request --
* Loop around messages from the kernel, forwarding them off to
* NLM servers.
*/
pid_t
client_request(void)
{
LOCKD_MSG msg;
fd_set rdset;
int fd, nr, ret;
pid_t child;
/* Recreate the NLM fifo. */
(void)unlink(_PATH_LCKFIFO);
(void)umask(S_IXGRP|S_IXOTH);
if (mkfifo(_PATH_LCKFIFO, S_IWUSR | S_IRUSR)) {
syslog(LOG_ERR, "mkfifo: %s: %m", _PATH_LCKFIFO);
exit (1);
}
/*
* Create a separate process, the client code is really a separate
* daemon that shares a lot of code.
*/
switch (child = fork()) {
case -1:
err(1, "fork");
case 0:
break;
default:
return (child);
}
signal(SIGHUP, client_cleanup);
signal(SIGTERM, client_cleanup);
/* Setup. */
(void)time(&owner.tod);
owner.pid = getpid();
(void)gethostname(hostname, sizeof(hostname) - 1);
/* Open the fifo for reading. */
if ((fd = open(_PATH_LCKFIFO, O_RDONLY | O_NONBLOCK)) < 0)
syslog(LOG_ERR, "open: %s: %m", _PATH_LCKFIFO);
/* drop our root priviledges */
(void)lockd_seteuid(daemon_uid);
for (;;) {
/* Wait for contact... fifo's return EAGAIN when read with
* no data
*/
/* Set up the select. */
FD_ZERO(&rdset);
FD_SET(fd, &rdset);
(void)select(fd + 1, &rdset, NULL, NULL, NULL);
/* Read the fixed length message. */
if ((nr = read(fd, &msg, sizeof(msg))) == sizeof(msg)) {
if (d_args)
show(&msg);
if (msg.lm_version != LOCKD_MSG_VERSION) {
syslog(LOG_ERR,
"unknown msg type: %d", msg.lm_version);
}
/*
* Send it to the NLM server and don't grant the lock
* if we fail for any reason.
*/
switch (msg.lm_fl.l_type) {
case F_RDLCK:
case F_WRLCK:
if (msg.lm_getlk)
ret = test_request(&msg);
else
ret = lock_request(&msg);
break;
case F_UNLCK:
ret = unlock_request(&msg);
break;
default:
ret = 1;
syslog(LOG_ERR,
"unknown lock type: %d", msg.lm_fl.l_type);
break;
}
if (ret) {
struct lockd_ans ans;
ans.la_msg_ident = msg.lm_msg_ident;
ans.la_errno = EHOSTUNREACH;
if (nfslockdans(LOCKD_ANS_VERSION, &ans)) {
syslog((errno == EPIPE ? LOG_INFO :
LOG_ERR), "process %lu: %m",
(u_long)msg.lm_msg_ident.pid);
}
}
} else if (nr == -1) {
if (errno != EAGAIN) {
syslog(LOG_ERR, "read: %s: %m", _PATH_LCKFIFO);
goto err;
}
} else if (nr != 0) {
syslog(LOG_ERR,
"%s: discard %d bytes", _PATH_LCKFIFO, nr);
}
}
/* Reached only on error. */
err:
(void)lockd_seteuid(0);
(void)unlink(_PATH_LCKFIFO);
_exit (1);
}
void
set_auth(cl, ucred)
CLIENT *cl;
struct ucred *ucred;
{
if (cl->cl_auth != NULL)
cl->cl_auth->ah_ops->ah_destroy(cl->cl_auth);
cl->cl_auth = authunix_create(hostname,
ucred->cr_uid,
ucred->cr_groups[0],
ucred->cr_ngroups-1,
&ucred->cr_groups[1]);
}
/*
* test_request --
* Convert a lock LOCKD_MSG into an NLM request, and send it off.
*/
int
test_request(LOCKD_MSG *msg)
{
CLIENT *cli;
struct timeval timeout = {0, 0}; /* No timeout, no response. */
char dummy;
if (d_calls)
syslog(LOG_DEBUG, "test request: %s: %s to %s",
msg->lm_nfsv3 ? "V4" : "V1/3",
msg->lm_fl.l_type == F_WRLCK ? "write" : "read",
from_addr((struct sockaddr_in *)&msg->lm_addr));
if (msg->lm_nfsv3) {
struct nlm4_testargs arg4;
arg4.cookie.n_bytes = (char *)&msg->lm_msg_ident;
arg4.cookie.n_len = sizeof(msg->lm_msg_ident);
arg4.exclusive = msg->lm_fl.l_type == F_WRLCK ? 1 : 0;
arg4.alock.caller_name = hostname;
arg4.alock.fh.n_bytes = (char *)&msg->lm_fh;
arg4.alock.fh.n_len = msg->lm_fh_len;
arg4.alock.oh.n_bytes = (char *)&owner;
arg4.alock.oh.n_len = sizeof(owner);
arg4.alock.svid = msg->lm_msg_ident.pid;
arg4.alock.l_offset = msg->lm_fl.l_start;
arg4.alock.l_len = msg->lm_fl.l_len;
if ((cli = get_client(
(struct sockaddr *)&msg->lm_addr,
NLM_VERS4)) == NULL)
return (1);
set_auth(cli, &msg->lm_cred);
(void)clnt_call(cli, NLM_TEST_MSG,
xdr_nlm4_testargs, &arg4, xdr_void, &dummy, timeout);
} else {
struct nlm_testargs arg;
arg.cookie.n_bytes = (char *)&msg->lm_msg_ident;
arg.cookie.n_len = sizeof(msg->lm_msg_ident);
arg.exclusive = msg->lm_fl.l_type == F_WRLCK ? 1 : 0;
arg.alock.caller_name = hostname;
arg.alock.fh.n_bytes = (char *)&msg->lm_fh;
arg.alock.fh.n_len = msg->lm_fh_len;
arg.alock.oh.n_bytes = (char *)&owner;
arg.alock.oh.n_len = sizeof(owner);
arg.alock.svid = msg->lm_msg_ident.pid;
arg.alock.l_offset = msg->lm_fl.l_start;
arg.alock.l_len = msg->lm_fl.l_len;
if ((cli = get_client(
(struct sockaddr *)&msg->lm_addr,
NLM_VERS)) == NULL)
return (1);
set_auth(cli, &msg->lm_cred);
(void)clnt_call(cli, NLM_TEST_MSG,
xdr_nlm_testargs, &arg, xdr_void, &dummy, timeout);
}
return (0);
}
/*
* lock_request --
* Convert a lock LOCKD_MSG into an NLM request, and send it off.
*/
int
lock_request(LOCKD_MSG *msg)
{
CLIENT *cli;
struct nlm4_lockargs arg4;
struct nlm_lockargs arg;
struct timeval timeout = {0, 0}; /* No timeout, no response. */
char dummy;
if (d_calls)
syslog(LOG_DEBUG, "lock request: %s: %s to %s",
msg->lm_nfsv3 ? "V4" : "V1/3",
msg->lm_fl.l_type == F_WRLCK ? "write" : "read",
from_addr((struct sockaddr_in *)&msg->lm_addr));
if (msg->lm_nfsv3) {
arg4.cookie.n_bytes = (char *)&msg->lm_msg_ident;
arg4.cookie.n_len = sizeof(msg->lm_msg_ident);
arg4.block = msg->lm_wait ? 1 : 0;
arg4.exclusive = msg->lm_fl.l_type == F_WRLCK ? 1 : 0;
arg4.alock.caller_name = hostname;
arg4.alock.fh.n_bytes = (char *)&msg->lm_fh;
arg4.alock.fh.n_len = msg->lm_fh_len;
arg4.alock.oh.n_bytes = (char *)&owner;
arg4.alock.oh.n_len = sizeof(owner);
arg4.alock.svid = msg->lm_msg_ident.pid;
arg4.alock.l_offset = msg->lm_fl.l_start;
arg4.alock.l_len = msg->lm_fl.l_len;
arg4.reclaim = 0;
arg4.state = nsm_state;
if ((cli = get_client(
(struct sockaddr *)&msg->lm_addr,
NLM_VERS4)) == NULL)
return (1);
set_auth(cli, &msg->lm_cred);
(void)clnt_call(cli, NLM_LOCK_MSG,
xdr_nlm4_lockargs, &arg4, xdr_void, &dummy, timeout);
} else {
arg.cookie.n_bytes = (char *)&msg->lm_msg_ident;
arg.cookie.n_len = sizeof(msg->lm_msg_ident);
arg.block = msg->lm_wait ? 1 : 0;
arg.exclusive = msg->lm_fl.l_type == F_WRLCK ? 1 : 0;
arg.alock.caller_name = hostname;
arg.alock.fh.n_bytes = (char *)&msg->lm_fh;
arg.alock.fh.n_len = msg->lm_fh_len;
arg.alock.oh.n_bytes = (char *)&owner;
arg.alock.oh.n_len = sizeof(owner);
arg.alock.svid = msg->lm_msg_ident.pid;
arg.alock.l_offset = msg->lm_fl.l_start;
arg.alock.l_len = msg->lm_fl.l_len;
arg.reclaim = 0;
arg.state = nsm_state;
if ((cli = get_client(
(struct sockaddr *)&msg->lm_addr,
NLM_VERS)) == NULL)
return (1);
set_auth(cli, &msg->lm_cred);
(void)clnt_call(cli, NLM_LOCK_MSG,
xdr_nlm_lockargs, &arg, xdr_void, &dummy, timeout);
}
return (0);
}
/*
* unlock_request --
* Convert an unlock LOCKD_MSG into an NLM request, and send it off.
*/
int
unlock_request(LOCKD_MSG *msg)
{
CLIENT *cli;
struct nlm4_unlockargs arg4;
struct nlm_unlockargs arg;
struct timeval timeout = {0, 0}; /* No timeout, no response. */
char dummy;
if (d_calls)
syslog(LOG_DEBUG, "unlock request: %s: to %s",
msg->lm_nfsv3 ? "V4" : "V1/3",
from_addr((struct sockaddr_in *)&msg->lm_addr));
if (msg->lm_nfsv3) {
arg4.cookie.n_bytes = (char *)&msg->lm_msg_ident;
arg4.cookie.n_len = sizeof(msg->lm_msg_ident);
arg4.alock.caller_name = hostname;
arg4.alock.fh.n_bytes = (char *)&msg->lm_fh;
arg4.alock.fh.n_len = msg->lm_fh_len;
arg4.alock.oh.n_bytes = (char *)&owner;
arg4.alock.oh.n_len = sizeof(owner);
arg4.alock.svid = msg->lm_msg_ident.pid;
arg4.alock.l_offset = msg->lm_fl.l_start;
arg4.alock.l_len = msg->lm_fl.l_len;
if ((cli = get_client(
(struct sockaddr *)&msg->lm_addr,
NLM_VERS4)) == NULL)
return (1);
set_auth(cli, &msg->lm_cred);
(void)clnt_call(cli, NLM_UNLOCK_MSG,
xdr_nlm4_unlockargs, &arg4, xdr_void, &dummy, timeout);
} else {
arg.cookie.n_bytes = (char *)&msg->lm_msg_ident;
arg.cookie.n_len = sizeof(msg->lm_msg_ident);
arg.alock.caller_name = hostname;
arg.alock.fh.n_bytes = (char *)&msg->lm_fh;
arg.alock.fh.n_len = msg->lm_fh_len;
arg.alock.oh.n_bytes = (char *)&owner;
arg.alock.oh.n_len = sizeof(owner);
arg.alock.svid = msg->lm_msg_ident.pid;
arg.alock.l_offset = msg->lm_fl.l_start;
arg.alock.l_len = msg->lm_fl.l_len;
if ((cli = get_client(
(struct sockaddr *)&msg->lm_addr,
NLM_VERS)) == NULL)
return (1);
set_auth(cli, &msg->lm_cred);
(void)clnt_call(cli, NLM_UNLOCK_MSG,
xdr_nlm_unlockargs, &arg, xdr_void, &dummy, timeout);
}
return (0);
}
int
lock_answer(int pid, netobj *netcookie, int result, int *pid_p, int version)
{
struct lockd_ans ans;
if (netcookie->n_len != sizeof(ans.la_msg_ident)) {
if (pid == -1) { /* we're screwed */
syslog(LOG_ERR, "inedible nlm cookie");
return -1;
}
ans.la_msg_ident.pid = pid;
ans.la_msg_ident.msg_seq = -1;
} else {
memcpy(&ans.la_msg_ident, netcookie->n_bytes,
sizeof(ans.la_msg_ident));
}
if (d_calls)
syslog(LOG_DEBUG, "lock answer: pid %lu: %s %d",
ans.la_msg_ident.pid,
version == NLM_VERS4 ? "nlmv4" : "nlmv3",
result);
ans.la_set_getlk_pid = 0;
if (version == NLM_VERS4)
switch (result) {
case nlm4_granted:
ans.la_errno = 0;
break;
default:
ans.la_errno = EACCES;
break;
case nlm4_denied:
if (pid_p == NULL)
ans.la_errno = EACCES;
else {
/* this is an answer to a nlm_test msg */
ans.la_set_getlk_pid = 1;
ans.la_getlk_pid = *pid_p;
ans.la_errno = 0;
}
break;
case nlm4_denied_nolocks:
ans.la_errno = EAGAIN;
break;
case nlm4_blocked:
return -1;
/* NOTREACHED */
case nlm4_denied_grace_period:
ans.la_errno = EAGAIN;
break;
case nlm4_deadlck:
ans.la_errno = EDEADLK;
break;
case nlm4_rofs:
ans.la_errno = EROFS;
break;
case nlm4_stale_fh:
ans.la_errno = ESTALE;
break;
case nlm4_fbig:
ans.la_errno = EFBIG;
break;
case nlm4_failed:
ans.la_errno = EACCES;
break;
}
else
switch (result) {
case nlm_granted:
ans.la_errno = 0;
break;
default:
ans.la_errno = EACCES;
break;
case nlm_denied:
if (pid_p == NULL)
ans.la_errno = EACCES;
else {
/* this is an answer to a nlm_test msg */
ans.la_set_getlk_pid = 1;
ans.la_getlk_pid = *pid_p;
ans.la_errno = 0;
}
break;
case nlm_denied_nolocks:
ans.la_errno = EAGAIN;
break;
case nlm_blocked:
return -1;
/* NOTREACHED */
case nlm_denied_grace_period:
ans.la_errno = EAGAIN;
break;
case nlm_deadlck:
ans.la_errno = EDEADLK;
break;
}
if (nfslockdans(LOCKD_ANS_VERSION, &ans)) {
syslog(((errno == EPIPE || errno == ESRCH) ?
LOG_INFO : LOG_ERR),
"process %lu: %m", (u_long)ans.la_msg_ident.pid);
return -1;
}
return 0;
}
/*
* show --
* Display the contents of a kernel LOCKD_MSG structure.
*/
void
show(LOCKD_MSG *mp)
{
static char hex[] = "0123456789abcdef";
struct fid *fidp;
fsid_t *fsidp;
size_t len;
u_int8_t *p, *t, buf[NFS_SMALLFH*3+1];
syslog(LOG_DEBUG, "process ID: %lu\n", (long)mp->lm_msg_ident.pid);
fsidp = (fsid_t *)&mp->lm_fh;
fidp = (struct fid *)((u_int8_t *)&mp->lm_fh + sizeof(fsid_t));
for (t = buf, p = (u_int8_t *)mp->lm_fh,
len = mp->lm_fh_len;
len > 0; ++p, --len) {
*t++ = '\\';
*t++ = hex[(*p & 0xf0) >> 4];
*t++ = hex[*p & 0x0f];
}
*t = '\0';
syslog(LOG_DEBUG, "fh_len %d, fh %s\n", mp->lm_fh_len, buf);
/* Show flock structure. */
syslog(LOG_DEBUG, "start %qu; len %qu; pid %lu; type %d; whence %d\n",
mp->lm_fl.l_start, mp->lm_fl.l_len, (u_long)mp->lm_fl.l_pid,
mp->lm_fl.l_type, mp->lm_fl.l_whence);
/* Show wait flag. */
syslog(LOG_DEBUG, "wait was %s\n", mp->lm_wait ? "set" : "not set");
}