File:  [ELWIX - Embedded LightWeight unIX -] / embedaddon / rsync / clientserver.c
Revision 1.1.1.4 (vendor branch): download - view: text, annotated - select for diffs - revision graph
Wed Mar 17 00:32:36 2021 UTC (3 years, 2 months ago) by misho
Branches: rsync, MAIN
CVS tags: v3_2_3, HEAD
rsync 3.2.3

/*
 * The socket based protocol for setting up a connection with rsyncd.
 *
 * Copyright (C) 1998-2001 Andrew Tridgell <tridge@samba.org>
 * Copyright (C) 2001-2002 Martin Pool <mbp@samba.org>
 * Copyright (C) 2002-2020 Wayne Davison
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 3 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License along
 * with this program; if not, visit the http://fsf.org website.
 */

#include "rsync.h"
#include "itypes.h"
#include "ifuncs.h"

extern int quiet;
extern int dry_run;
extern int output_motd;
extern int list_only;
extern int am_sender;
extern int am_server;
extern int am_daemon;
extern int am_root;
extern int msgs2stderr;
extern int rsync_port;
extern int protect_args;
extern int ignore_errors;
extern int preserve_xattrs;
extern int kluge_around_eof;
extern int munge_symlinks;
extern int open_noatime;
extern int sanitize_paths;
extern int numeric_ids;
extern int filesfrom_fd;
extern int remote_protocol;
extern int protocol_version;
extern int always_checksum;
extern int checksum_files;
extern int io_timeout;
extern int no_detach;
extern int write_batch;
extern int default_af_hint;
extern int logfile_format_has_i;
extern int logfile_format_has_o_or_i;
extern char *bind_address;
extern char *config_file;
extern char *link_by_hash_dir;
extern char *logfile_format;
extern char *files_from;
extern char *tmpdir;
extern char *early_input_file;
extern struct chmod_mode_struct *chmod_modes;
extern filter_rule_list daemon_filter_list;
#ifdef ICONV_OPTION
extern char *iconv_opt;
extern iconv_t ic_send, ic_recv;
#endif
extern uid_t our_uid;
extern gid_t our_gid;

char *auth_user;
int read_only = 0;
int module_id = -1;
int pid_file_fd = -1;
int early_input_len = 0;
char *early_input = NULL;
pid_t namecvt_pid = 0;
struct chmod_mode_struct *daemon_chmod_modes;

#define EARLY_INPUT_CMD "#early_input="
#define EARLY_INPUT_CMDLEN (sizeof EARLY_INPUT_CMD - 1)

/* module_dirlen is the length of the module_dir string when in daemon
 * mode and module_dir is not "/"; otherwise 0.  (Note that a chroot-
 * enabled module can have a non-"/" module_dir these days.) */
char *module_dir = NULL;
unsigned int module_dirlen = 0;

char *full_module_path;

static int rl_nulls = 0;
static int namecvt_fd_req = -1, namecvt_fd_ans = -1;

#ifdef HAVE_SIGACTION
static struct sigaction sigact;
#endif

static item_list gid_list = EMPTY_ITEM_LIST;

/* Used when "reverse lookup" is off. */
const char undetermined_hostname[] = "UNDETERMINED";

/**
 * Run a client connected to an rsyncd.  The alternative to this
 * function for remote-shell connections is do_cmd().
 *
 * After negotiating which module to use and reading the server's
 * motd, this hands over to client_run().  Telling the server the
 * module will cause it to chroot/setuid/etc.
 *
 * Instead of doing a transfer, the client may at this stage instead
 * get a listing of remote modules and exit.
 *
 * @return -1 for error in startup, or the result of client_run().
 * Either way, it eventually gets passed to exit_cleanup().
 **/
int start_socket_client(char *host, int remote_argc, char *remote_argv[],
			int argc, char *argv[])
{
	int fd, ret;
	char *p, *user = NULL;

	/* This is redundant with code in start_inband_exchange(), but this
	 * short-circuits a problem in the client before we open a socket,
	 * and the extra check won't hurt. */
	if (**remote_argv == '/') {
		rprintf(FERROR,
			"ERROR: The remote path must start with a module name not a /\n");
		return -1;
	}

	if ((p = strrchr(host, '@')) != NULL) {
		user = host;
		host = p+1;
		*p = '\0';
	}

	fd = open_socket_out_wrapped(host, rsync_port, bind_address, default_af_hint);
	if (fd == -1)
		exit_cleanup(RERR_SOCKETIO);

#ifdef ICONV_CONST
	setup_iconv();
#endif

	ret = start_inband_exchange(fd, fd, user, host, remote_argc, remote_argv);

	return ret ? ret : client_run(fd, fd, -1, argc, argv);
}

static int exchange_protocols(int f_in, int f_out, char *buf, size_t bufsiz, int am_client)
{
	int remote_sub = -1;
#if SUBPROTOCOL_VERSION != 0
	int our_sub = protocol_version < PROTOCOL_VERSION ? 0 : SUBPROTOCOL_VERSION;
#else
	int our_sub = 0;
#endif

	io_printf(f_out, "@RSYNCD: %d.%d\n", protocol_version, our_sub);
	if (!am_client) {
		char *motd = lp_motd_file();
		if (motd && *motd) {
			FILE *f = fopen(motd, "r");
			while (f && !feof(f)) {
				int len = fread(buf, 1, bufsiz - 1, f);
				if (len > 0)
					write_buf(f_out, buf, len);
			}
			if (f)
				fclose(f);
			write_sbuf(f_out, "\n");
		}
	}

	/* This strips the \n. */
	if (!read_line_old(f_in, buf, bufsiz, 0)) {
		if (am_client)
			rprintf(FERROR, "rsync: did not see server greeting\n");
		return -1;
	}

	if (sscanf(buf, "@RSYNCD: %d.%d", &remote_protocol, &remote_sub) < 1) {
		if (am_client)
			rprintf(FERROR, "rsync: server sent \"%s\" rather than greeting\n", buf);
		else
			io_printf(f_out, "@ERROR: protocol startup error\n");
		return -1;
	}

	if (remote_sub < 0) {
		if (remote_protocol == 30) {
			if (am_client)
				rprintf(FERROR, "rsync: server is speaking an incompatible beta of protocol 30\n");
			else
				io_printf(f_out, "@ERROR: your client is speaking an incompatible beta of protocol 30\n");
			return -1;
		}
		remote_sub = 0;
	}

	if (protocol_version > remote_protocol) {
		protocol_version = remote_protocol;
		if (remote_sub)
			protocol_version--;
	} else if (protocol_version == remote_protocol) {
		if (remote_sub != our_sub)
			protocol_version--;
	}
#if SUBPROTOCOL_VERSION != 0
	else if (protocol_version < remote_protocol) {
		if (our_sub)
			protocol_version--;
	}
#endif

	if (protocol_version >= 30)
		rl_nulls = 1;

	return 0;
}

int start_inband_exchange(int f_in, int f_out, const char *user, const char *host, int argc, char *argv[])
{
	int i, modlen;
	char line[BIGPATHBUFLEN];
	char *sargs[MAX_ARGS];
	int sargc = 0;
	char *p, *modname;

	assert(argc > 0 && *argv != NULL);

	if (**argv == '/') {
		rprintf(FERROR,
			"ERROR: The remote path must start with a module name\n");
		return -1;
	}

	if (!(p = strchr(*argv, '/')))
		modlen = strlen(*argv);
	else
		modlen = p - *argv;

	modname = new_array(char, modlen+1+1); /* room for '/' & '\0' */
	strlcpy(modname, *argv, modlen + 1);
	modname[modlen] = '/';
	modname[modlen+1] = '\0';

	if (!user)
		user = getenv("USER");
	if (!user)
		user = getenv("LOGNAME");

	if (exchange_protocols(f_in, f_out, line, sizeof line, 1) < 0)
		return -1;

	if (early_input_file) {
		STRUCT_STAT st;
		FILE *f = fopen(early_input_file, "rb");
		if (!f || do_fstat(fileno(f), &st) < 0) {
			rsyserr(FERROR, errno, "failed to open %s", early_input_file);
			return -1;
		}
		early_input_len = st.st_size;
		if (early_input_len > (int)sizeof line) {
			rprintf(FERROR, "%s is > %d bytes.\n", early_input_file, (int)sizeof line);
			return -1;
		}
		if (early_input_len > 0) {
			io_printf(f_out, EARLY_INPUT_CMD "%d\n", early_input_len);
			while (early_input_len > 0) {
				int len;
				if (feof(f)) {
					rprintf(FERROR, "Early EOF in %s\n", early_input_file);
					return -1;
				}
				len = fread(line, 1, early_input_len, f);
				if (len > 0) {
					write_buf(f_out, line, len);
					early_input_len -= len;
				}
			}
		}
		fclose(f);
	}

	server_options(sargs, &sargc);

	if (sargc >= MAX_ARGS - 2)
		goto arg_overflow;

	sargs[sargc++] = ".";

	while (argc > 0) {
		if (sargc >= MAX_ARGS - 1) {
		  arg_overflow:
			rprintf(FERROR, "internal: args[] overflowed in do_cmd()\n");
			exit_cleanup(RERR_SYNTAX);
		}
		if (strncmp(*argv, modname, modlen) == 0
		 && argv[0][modlen] == '\0')
			sargs[sargc++] = modname; /* we send "modname/" */
		else if (**argv == '-') {
			if (asprintf(sargs + sargc++, "./%s", *argv) < 0)
				out_of_memory("start_inband_exchange");
		} else
			sargs[sargc++] = *argv;
		argv++;
		argc--;
	}

	sargs[sargc] = NULL;

	if (DEBUG_GTE(CMD, 1))
		print_child_argv("sending daemon args:", sargs);

	io_printf(f_out, "%.*s\n", modlen, modname);

	/* Old servers may just drop the connection here,
	 rather than sending a proper EXIT command.  Yuck. */
	kluge_around_eof = list_only && protocol_version < 25 ? 1 : 0;

	while (1) {
		if (!read_line_old(f_in, line, sizeof line, 0)) {
			rprintf(FERROR, "rsync: didn't get server startup line\n");
			return -1;
		}

		if (strncmp(line,"@RSYNCD: AUTHREQD ",18) == 0) {
			auth_client(f_out, user, line+18);
			continue;
		}

		if (strcmp(line, "@RSYNCD: GSS") == 0) {
#ifdef GSSAPI_OPTION
			if (auth_gss_client(f_out, host) < 0)
				return -1;
			continue;
#else
			rprintf(FERROR, "GSSAPI is not supported\n");
			return -1;
#endif
		}

		if (strcmp(line,"@RSYNCD: OK") == 0)
			break;

		if (strcmp(line,"@RSYNCD: EXIT") == 0) {
			/* This is sent by recent versions of the
			 * server to terminate the listing of modules.
			 * We don't want to go on and transfer
			 * anything; just exit. */
			exit(0);
		}

		if (strncmp(line, "@ERROR", 6) == 0) {
			rprintf(FERROR, "%s\n", line);
			/* This is always fatal; the server will now
			 * close the socket. */
			return -1;
		}

		/* This might be a MOTD line or a module listing, but there is
		 * no way to differentiate it.  The manpage mentions this. */
		if (output_motd)
			rprintf(FINFO, "%s\n", line);
	}
	kluge_around_eof = 0;

	if (rl_nulls) {
		for (i = 0; i < sargc; i++) {
			if (!sargs[i]) /* stop at --protect-args NULL */
				break;
			write_sbuf(f_out, sargs[i]);
			write_byte(f_out, 0);
		}
		write_byte(f_out, 0);
	} else {
		for (i = 0; i < sargc; i++)
			io_printf(f_out, "%s\n", sargs[i]);
		write_sbuf(f_out, "\n");
	}

	if (protect_args)
		send_protected_args(f_out, sargs);

	if (protocol_version < 23) {
		if (protocol_version == 22 || !am_sender)
			io_start_multiplex_in(f_in);
	}

	free(modname);

	return 0;
}

#ifdef HAVE_PUTENV
static int read_arg_from_pipe(int fd, char *buf, int limit)
{
	char *bp = buf, *eob = buf + limit - 1;

	while (1) {
		int got = read(fd, bp, 1);
		if (got != 1) {
			if (got < 0 && errno == EINTR)
				continue;
			return -1;
		}
		if (*bp == '\0')
			break;
		if (bp < eob)
			bp++;
	}
	*bp = '\0';

	return bp - buf;
}
#endif

static void set_env_str(const char *var, const char *str)
{
#ifdef HAVE_PUTENV
	char *mem;
	if (asprintf(&mem, "%s=%s", var, str) < 0)
		out_of_memory("set_env_str");
	putenv(mem);
#endif
}

#ifdef HAVE_PUTENV
void set_env_num(const char *var, long num)
{
	char *mem;
	if (asprintf(&mem, "%s=%ld", var, num) < 0)
		out_of_memory("set_env_num");
	putenv(mem);
}
#endif

/* Used for "early exec", "pre-xfer exec", and the "name converter" script. */
static pid_t start_pre_exec(const char *cmd, int *arg_fd_ptr, int *error_fd_ptr)
{
	int arg_fds[2], error_fds[2], arg_fd;
	pid_t pid;

	if ((error_fd_ptr && pipe(error_fds) < 0) || pipe(arg_fds) < 0 || (pid = fork()) < 0)
		return (pid_t)-1;

	if (pid == 0) {
		char buf[BIGPATHBUFLEN];
		int j, len, status;

		if (error_fd_ptr) {
			close(error_fds[0]);
			set_blocking(error_fds[1]);
		}

		close(arg_fds[1]);
		arg_fd = arg_fds[0];
		set_blocking(arg_fd);

		len = read_arg_from_pipe(arg_fd, buf, BIGPATHBUFLEN);
		if (len <= 0)
			_exit(1);
		set_env_str("RSYNC_REQUEST", buf);

		for (j = 0; ; j++) {
			char *p;
			len = read_arg_from_pipe(arg_fd, buf, BIGPATHBUFLEN);
			if (len <= 0) {
				if (!len)
					break;
				_exit(1);
			}
			if (asprintf(&p, "RSYNC_ARG%d=%s", j, buf) >= 0)
				putenv(p);
		}

		dup2(arg_fd, STDIN_FILENO);
		close(arg_fd);

		if (error_fd_ptr) {
			dup2(error_fds[1], STDOUT_FILENO);
			close(error_fds[1]);
		}

		status = shell_exec(cmd);

		if (!WIFEXITED(status))
			_exit(1);
		_exit(WEXITSTATUS(status));
	}

	if (error_fd_ptr) {
		close(error_fds[1]);
		*error_fd_ptr = error_fds[0];
		set_blocking(error_fds[0]);
	}

	close(arg_fds[0]);
	arg_fd = *arg_fd_ptr = arg_fds[1];
	set_blocking(arg_fd);

	return pid;
}

static void write_pre_exec_args(int write_fd, char *request, char **early_argv, char **argv, int exec_type)
{
	int j = 0;

	if (!request)
		request = "(NONE)";

	write_buf(write_fd, request, strlen(request)+1);
	if (early_argv) {
		for ( ; *early_argv; early_argv++)
			write_buf(write_fd, *early_argv, strlen(*early_argv)+1);
		j = 1; /* Skip arg0 name in argv. */
	}
	if (argv) {
		for ( ; argv[j]; j++)
			write_buf(write_fd, argv[j], strlen(argv[j])+1);
	}
	write_byte(write_fd, 0);

	if (exec_type == 1 && early_input_len)
		write_buf(write_fd, early_input, early_input_len);

	if (exec_type != 2) /* the name converter needs this left open */
		close(write_fd);
}

static char *finish_pre_exec(const char *desc, pid_t pid, int read_fd)
{
	char buf[BIGPATHBUFLEN], *bp, *cr;
	int j, status = -1, msglen = sizeof buf - 1;

	if (read_fd >= 0) {
		/* Read the stdout from the program.  This it is only displayed
		 * to the user if the script also returns an error status. */
		for (bp = buf, cr = buf; msglen > 0; msglen -= j) {
			if ((j = read(read_fd, bp, msglen)) <= 0) {
				if (j == 0)
					break;
				if (errno == EINTR)
					continue;
				break; /* Just ignore the read error for now... */
			}
			bp[j] = '\0';
			while (1) {
				if ((cr = strchr(cr, '\r')) == NULL) {
					cr = bp + j;
					break;
				}
				if (!cr[1])
					break; /* wait for more data before we decide what to do */
				if (cr[1] == '\n') {
					memmove(cr, cr+1, j - (cr - bp));
					j--;
				} else
					cr++;
			}
			bp += j;
		}
		*bp = '\0';

		close(read_fd);
	} else
		*buf = '\0';

	if (wait_process(pid, &status, 0) < 0
	 || !WIFEXITED(status) || WEXITSTATUS(status) != 0) {
		char *e;
		if (asprintf(&e, "%s returned failure (%d)%s%s%s\n%s",
			     desc, status, status < 0 ? ": " : "",
			     status < 0 ? strerror(errno) : "",
			     *buf ? ":" : "", buf) < 0)
			return "out_of_memory in finish_pre_exec\n";
		return e;
	}
	return NULL;
}

static int path_failure(int f_out, const char *dir, BOOL was_chdir)
{
	if (was_chdir)
		rsyserr(FLOG, errno, "chdir %s failed", dir);
	else
		rprintf(FLOG, "normalize_path(%s) failed\n", dir);
	io_printf(f_out, "@ERROR: chdir failed\n");
	return -1;
}

static int add_a_group(int f_out, const char *gname)
{
	gid_t gid, *gid_p;
	if (!group_to_gid(gname, &gid, True)) {
		rprintf(FLOG, "Invalid gid %s\n", gname);
		io_printf(f_out, "@ERROR: invalid gid %s\n", gname);
		return -1;
	}
	gid_p = EXPAND_ITEM_LIST(&gid_list, gid_t, -32);
	*gid_p = gid;
	return 0;
}

#ifdef HAVE_GETGROUPLIST
static int want_all_groups(int f_out, uid_t uid)
{
	const char *err;
	if ((err = getallgroups(uid, &gid_list)) != NULL) {
		rsyserr(FLOG, errno, "%s", err);
		io_printf(f_out, "@ERROR: %s\n", err);
		return -1;
	}
	return 0;
}
#elif defined HAVE_INITGROUPS
static struct passwd *want_all_groups(int f_out, uid_t uid)
{
	struct passwd *pw;
	gid_t *gid_p;
	if ((pw = getpwuid(uid)) == NULL) {
		rsyserr(FLOG, errno, "getpwuid failed");
		io_printf(f_out, "@ERROR: getpwuid failed\n");
		return NULL;
	}
	/* Start with the default group and initgroups() will add the rest. */
	gid_p = EXPAND_ITEM_LIST(&gid_list, gid_t, -32);
	*gid_p = pw->pw_gid;
	return pw;
}
#endif

static int rsync_module(int f_in, int f_out, int i, const char *addr, const char *host)
{
	int argc;
	char **argv, **orig_argv, **orig_early_argv, *module_chdir;
	char line[BIGPATHBUFLEN];
#if defined HAVE_INITGROUPS && !defined HAVE_GETGROUPLIST
	struct passwd *pw = NULL;
#endif
	uid_t uid;
	int set_uid;
	char *p, *err_msg = NULL;
	char *name = lp_name(i);
	int use_chroot = lp_use_chroot(i);
	int ret, pre_exec_arg_fd = -1, pre_exec_error_fd = -1;
	int save_munge_symlinks;
	pid_t pre_exec_pid = 0;
	char *request = NULL;

	set_env_str("RSYNC_MODULE_NAME", name);

#ifdef ICONV_OPTION
	iconv_opt = lp_charset(i);
	if (*iconv_opt)
		setup_iconv();
	iconv_opt = NULL;
#endif

	/* If reverse lookup is disabled globally but enabled for this module,
	 * we need to do it now before the access check. */
	if (host == undetermined_hostname && lp_reverse_lookup(i))
		host = client_name(client_addr(f_in));
	set_env_str("RSYNC_HOST_NAME", host);
	set_env_str("RSYNC_HOST_ADDR", addr);

	if (!allow_access(addr, &host, i)) {
		rprintf(FLOG, "rsync denied on module %s from %s (%s)\n",
			name, host, addr);
		if (!lp_list(i))
			io_printf(f_out, "@ERROR: Unknown module '%s'\n", name);
		else {
			io_printf(f_out,
				  "@ERROR: access denied to %s from %s (%s)\n",
				  name, host, addr);
		}
		return -1;
	}

	if (*lp_link_by_hash_dir(i))
		link_by_hash_dir = lp_link_by_hash_dir(i);

	if (am_daemon > 0) {
		rprintf(FLOG, "rsync allowed access on module %s from %s (%s)\n",
			name, host, addr);
	}

	if (!claim_connection(lp_lock_file(i), lp_max_connections(i))) {
		if (errno) {
			rsyserr(FLOG, errno, "failed to open lock file %s",
				lp_lock_file(i));
			io_printf(f_out, "@ERROR: failed to open lock file\n");
		} else {
			rprintf(FLOG, "max connections (%d) reached\n",
				lp_max_connections(i));
			io_printf(f_out, "@ERROR: max connections (%d) reached -- try again later\n",
				lp_max_connections(i));
		}
		return -1;
	}

	read_only = lp_read_only(i); /* may also be overridden by auth_server() */
#ifdef GSSAPI_OPTION
	if (lp_use_gssapi(i))
		auth_user = auth_gss_server(f_in, f_out, i, host, addr, "@RSYNCD: GSS");
	else
#endif
		auth_user = auth_server(f_in, f_out, i, host, addr, "@RSYNCD: AUTHREQD ");

	if (!auth_user) {
		io_printf(f_out, "@ERROR: auth failed on module %s\n", name);
		return -1;
	}
	set_env_str("RSYNC_USER_NAME", auth_user);

	module_id = i;

	if (lp_transfer_logging(module_id) && !logfile_format)
		logfile_format = lp_log_format(module_id);
	if (log_format_has(logfile_format, 'i'))
		logfile_format_has_i = 1;
	if (logfile_format_has_i || log_format_has(logfile_format, 'o'))
		logfile_format_has_o_or_i = 1;

	uid = MY_UID();
	am_root = (uid == ROOT_UID);

	p = *lp_uid(module_id) ? lp_uid(module_id) : am_root ? NOBODY_USER : NULL;
	if (p) {
		if (!user_to_uid(p, &uid, True)) {
			rprintf(FLOG, "Invalid uid %s\n", p);
			io_printf(f_out, "@ERROR: invalid uid %s\n", p);
			return -1;
		}
		set_uid = 1;
	} else
		set_uid = 0;

	p = *lp_gid(module_id) ? conf_strtok(lp_gid(module_id)) : NULL;
	if (p) {
		/* The "*" gid must be the first item in the list. */
		if (strcmp(p, "*") == 0) {
#ifdef HAVE_GETGROUPLIST
			if (want_all_groups(f_out, uid) < 0)
				return -1;
#elif defined HAVE_INITGROUPS
			if ((pw = want_all_groups(f_out, uid)) == NULL)
				return -1;
#else
			rprintf(FLOG, "This rsync does not support a gid of \"*\"\n");
			io_printf(f_out, "@ERROR: invalid gid setting.\n");
			return -1;
#endif
		} else if (add_a_group(f_out, p) < 0)
			return -1;
		while ((p = conf_strtok(NULL)) != NULL) {
#if defined HAVE_INITGROUPS && !defined HAVE_GETGROUPLIST
			if (pw) {
				rprintf(FLOG, "This rsync cannot add groups after \"*\".\n");
				io_printf(f_out, "@ERROR: invalid gid setting.\n");
				return -1;
			}
#endif
			if (add_a_group(f_out, p) < 0)
				return -1;
		}
	} else if (am_root) {
		if (add_a_group(f_out, NOBODY_GROUP) < 0)
			return -1;
	}

	module_dir = lp_path(module_id);
	if (*module_dir == '\0') {
		rprintf(FLOG, "No path specified for module %s\n", name);
		io_printf(f_out, "@ERROR: no path setting.\n");
		return -1;
	}
	if (use_chroot) {
		if ((p = strstr(module_dir, "/./")) != NULL) {
			*p = '\0'; /* Temporary... */
			if (!(module_chdir = normalize_path(module_dir, True, NULL)))
				return path_failure(f_out, module_dir, False);
			*p = '/';
			if (!(p = normalize_path(p + 2, True, &module_dirlen)))
				return path_failure(f_out, strstr(module_dir, "/./"), False);
			if (!(full_module_path = normalize_path(module_dir, False, NULL)))
				full_module_path = module_dir;
			module_dir = p;
		} else {
			if (!(module_chdir = normalize_path(module_dir, False, NULL)))
				return path_failure(f_out, module_dir, False);
			full_module_path = module_chdir;
			module_dir = "/";
			module_dirlen = 1;
		}
	} else {
		if (!(module_chdir = normalize_path(module_dir, False, &module_dirlen)))
			return path_failure(f_out, module_dir, False);
		full_module_path = module_dir = module_chdir;
	}
	set_env_str("RSYNC_MODULE_PATH", full_module_path);

	if (module_dirlen == 1) {
		module_dirlen = 0;
		set_filter_dir("/", 1);
	} else
		set_filter_dir(module_dir, module_dirlen);

	p = lp_filter(module_id);
	parse_filter_str(&daemon_filter_list, p, rule_template(FILTRULE_WORD_SPLIT),
		XFLG_ABS_IF_SLASH | XFLG_DIR2WILD3);

	p = lp_include_from(module_id);
	parse_filter_file(&daemon_filter_list, p, rule_template(FILTRULE_INCLUDE),
		XFLG_ABS_IF_SLASH | XFLG_DIR2WILD3 | XFLG_OLD_PREFIXES | XFLG_FATAL_ERRORS);

	p = lp_include(module_id);
	parse_filter_str(&daemon_filter_list, p,
		rule_template(FILTRULE_INCLUDE | FILTRULE_WORD_SPLIT),
		XFLG_ABS_IF_SLASH | XFLG_DIR2WILD3 | XFLG_OLD_PREFIXES);

	p = lp_exclude_from(module_id);
	parse_filter_file(&daemon_filter_list, p, rule_template(0),
		XFLG_ABS_IF_SLASH | XFLG_DIR2WILD3 | XFLG_OLD_PREFIXES | XFLG_FATAL_ERRORS);

	p = lp_exclude(module_id);
	parse_filter_str(&daemon_filter_list, p, rule_template(FILTRULE_WORD_SPLIT),
		XFLG_ABS_IF_SLASH | XFLG_DIR2WILD3 | XFLG_OLD_PREFIXES);

	log_init(1);

	if (*lp_db_config(i)) {
		db_read_config(FLOG, lp_db_config(i));
		db_lax = lp_db_lax(i);
	}

#ifdef HAVE_PUTENV
	if ((*lp_early_exec(module_id) || *lp_prexfer_exec(module_id)
	  || *lp_postxfer_exec(module_id) || *lp_name_converter(module_id))
	 && !getenv("RSYNC_NO_XFER_EXEC")) {
		set_env_num("RSYNC_PID", (long)getpid());

		/* For post-xfer exec, fork a new process to run the rsync
		 * daemon while this process waits for the exit status and
		 * runs the indicated command at that point. */
		if (*lp_postxfer_exec(module_id)) {
			pid_t pid = fork();
			if (pid < 0) {
				rsyserr(FLOG, errno, "fork failed");
				io_printf(f_out, "@ERROR: fork failed\n");
				return -1;
			}
			if (pid) {
				int status;
				close(f_in);
				if (f_out != f_in)
					close(f_out);
				if (wait_process(pid, &status, 0) < 0)
					status = -1;
				set_env_num("RSYNC_RAW_STATUS", status);
				if (WIFEXITED(status))
					status = WEXITSTATUS(status);
				else
					status = -1;
				set_env_num("RSYNC_EXIT_STATUS", status);
				if (shell_exec(lp_postxfer_exec(module_id)) < 0)
					status = -1;
				_exit(status);
			}
		}

		/* For early exec, fork a child process to run the indicated
		 * command and wait for it to exit. */
		if (*lp_early_exec(module_id)) {
			int arg_fd;
			pid_t pid = start_pre_exec(lp_early_exec(module_id), &arg_fd, NULL);
			if (pid == (pid_t)-1) {
				rsyserr(FLOG, errno, "early exec preparation failed");
				io_printf(f_out, "@ERROR: early exec preparation failed\n");
				return -1;
			}
			write_pre_exec_args(arg_fd, NULL, NULL, NULL, 1);
			if (finish_pre_exec("early exec", pid, -1) != NULL) {
				rsyserr(FLOG, errno, "early exec failed");
				io_printf(f_out, "@ERROR: early exec failed\n");
				return -1;
			}
		}

		/* For pre-xfer exec, fork a child process to run the indicated
		 * command, though it first waits for the parent process to
		 * send us the user's request via a pipe. */
		if (*lp_prexfer_exec(module_id)) {
			pre_exec_pid = start_pre_exec(lp_prexfer_exec(module_id), &pre_exec_arg_fd, &pre_exec_error_fd);
			if (pre_exec_pid == (pid_t)-1) {
				rsyserr(FLOG, errno, "pre-xfer exec preparation failed");
				io_printf(f_out, "@ERROR: pre-xfer exec preparation failed\n");
				return -1;
			}
		}

		if (*lp_name_converter(module_id)) {
			namecvt_pid = start_pre_exec(lp_name_converter(module_id), &namecvt_fd_req, &namecvt_fd_ans);
			if (namecvt_pid == (pid_t)-1) {
				rsyserr(FLOG, errno, "name-converter exec preparation failed");
				io_printf(f_out, "@ERROR: name-converter exec preparation failed\n");
				return -1;
			}
		}
	}
#endif

	if (early_input) {
		free(early_input);
		early_input = NULL;
	}

	if (use_chroot) {
		/*
		 * XXX: The 'use chroot' flag is a fairly reliable
		 * source of confusion, because it fails under two
		 * important circumstances: running as non-root,
		 * running on Win32 (or possibly others).  On the
		 * other hand, if you are running as root, then it
		 * might be better to always use chroot.
		 *
		 * So, perhaps if we can't chroot we should just issue
		 * a warning, unless a "require chroot" flag is set,
		 * in which case we fail.
		 */
		if (chroot(module_chdir)) {
			rsyserr(FLOG, errno, "chroot %s failed", module_chdir);
			io_printf(f_out, "@ERROR: chroot failed\n");
			return -1;
		}
		module_chdir = module_dir;
	}

	if (!change_dir(module_chdir, CD_NORMAL))
		return path_failure(f_out, module_chdir, True);
	if (module_dirlen || (!use_chroot && !*lp_daemon_chroot()))
		sanitize_paths = 1;

	if ((munge_symlinks = lp_munge_symlinks(module_id)) < 0)
		munge_symlinks = !use_chroot || module_dirlen;
	if (munge_symlinks) {
		STRUCT_STAT st;
		char prefix[SYMLINK_PREFIX_LEN]; /* NOT +1 ! */
		strlcpy(prefix, SYMLINK_PREFIX, sizeof prefix); /* trim the trailing slash */
		if (do_stat(prefix, &st) == 0 && S_ISDIR(st.st_mode)) {
			rprintf(FLOG, "Symlink munging is unsafe when a %s directory exists.\n",
				prefix);
			io_printf(f_out, "@ERROR: daemon security issue -- contact admin\n", name);
			exit_cleanup(RERR_UNSUPPORTED);
		}
	}

	if (gid_list.count) {
		gid_t *gid_array = gid_list.items;
		if (setgid(gid_array[0])) {
			rsyserr(FLOG, errno, "setgid %ld failed", (long)gid_array[0]);
			io_printf(f_out, "@ERROR: setgid failed\n");
			return -1;
		}
#ifdef HAVE_SETGROUPS
		/* Set the group(s) we want to be active. */
		if (setgroups(gid_list.count, gid_array)) {
			rsyserr(FLOG, errno, "setgroups failed");
			io_printf(f_out, "@ERROR: setgroups failed\n");
			return -1;
		}
#endif
#if defined HAVE_INITGROUPS && !defined HAVE_GETGROUPLIST
		/* pw is set if the user wants all the user's groups. */
		if (pw && initgroups(pw->pw_name, pw->pw_gid) < 0) {
			rsyserr(FLOG, errno, "initgroups failed");
			io_printf(f_out, "@ERROR: initgroups failed\n");
			return -1;
		}
#endif
		our_gid = MY_GID();
	}

	if (set_uid) {
		if (setuid(uid) < 0
#ifdef HAVE_SETEUID
		 || seteuid(uid) < 0
#endif
		) {
			rsyserr(FLOG, errno, "setuid %ld failed", (long)uid);
			io_printf(f_out, "@ERROR: setuid failed\n");
			return -1;
		}

		our_uid = MY_UID();
		am_root = (our_uid == ROOT_UID);
	}

	if (lp_temp_dir(module_id) && *lp_temp_dir(module_id)) {
		tmpdir = lp_temp_dir(module_id);
		if (strlen(tmpdir) >= MAXPATHLEN - 10) {
			rprintf(FLOG,
				"the 'temp dir' value for %s is WAY too long -- ignoring.\n",
				name);
			tmpdir = NULL;
		}
	}

	io_printf(f_out, "@RSYNCD: OK\n");

	read_args(f_in, name, line, sizeof line, rl_nulls, &argv, &argc, &request);
	orig_argv = argv;

	save_munge_symlinks = munge_symlinks;

	reset_output_levels(); /* future verbosity is controlled by client options */
	ret = parse_arguments(&argc, (const char ***) &argv);
	if (protect_args && ret) {
		orig_early_argv = orig_argv;
		protect_args = 2;
		read_args(f_in, name, line, sizeof line, 1, &argv, &argc, &request);
		orig_argv = argv;
		ret = parse_arguments(&argc, (const char ***) &argv);
	} else
		orig_early_argv = NULL;

	/* The default is to use the user's setting unless the module sets True or False. */
	if (lp_open_noatime(module_id) >= 0)
		open_noatime = lp_open_noatime(module_id);

	munge_symlinks = save_munge_symlinks; /* The client mustn't control this. */

	if (am_daemon > 0)
		msgs2stderr = 0; /* A non-rsh-run daemon doesn't have stderr for msgs. */

	if (pre_exec_pid) {
		write_pre_exec_args(pre_exec_arg_fd, request, orig_early_argv, orig_argv, 0);
		err_msg = finish_pre_exec("pre-xfer exec", pre_exec_pid, pre_exec_error_fd);
	}

	if (namecvt_pid)
		write_pre_exec_args(namecvt_fd_req, request, orig_early_argv, orig_argv, 2);

	if (orig_early_argv)
		free(orig_early_argv);

	am_server = 1; /* Don't let someone try to be tricky. */
	quiet = 0;
	db_config = NULL;

	if (lp_ignore_errors(module_id))
		ignore_errors = 1;
	if (write_batch < 0)
		dry_run = 1;

	if (lp_fake_super(module_id)) {
		if (preserve_xattrs > 1)
			preserve_xattrs = 1;
		am_root = -1;
	} else if (am_root < 0) /* Treat --fake-super from client as --super. */
		am_root = 2;

	checksum_files = always_checksum ? lp_checksum_files(i)
					 : CSF_IGNORE_FILES;

	if (filesfrom_fd == 0)
		filesfrom_fd = f_in;

	if (request) {
		if (*auth_user) {
			rprintf(FLOG, "rsync %s %s from %s@%s (%s)\n",
				am_sender ? "on" : "to",
				request, auth_user, host, addr);
		} else {
			rprintf(FLOG, "rsync %s %s from %s (%s)\n",
				am_sender ? "on" : "to",
				request, host, addr);
		}
		free(request);
	}

#ifndef DEBUG
	/* don't allow the logs to be flooded too fast */
	limit_output_verbosity(lp_max_verbosity(module_id));
#endif

	if (protocol_version < 23 && (protocol_version == 22 || am_sender))
		io_start_multiplex_out(f_out);
	else if (!ret || err_msg) {
		/* We have to get I/O multiplexing started so that we can
		 * get the error back to the client.  This means getting
		 * the protocol setup finished first in later versions. */
		setup_protocol(f_out, f_in);
		if (!am_sender) {
			/* Since we failed in our option parsing, we may not
			 * have finished parsing that the client sent us a
			 * --files-from option, so look for it manually.
			 * Without this, the socket would be in the wrong
			 * state for the upcoming error message. */
			if (!files_from) {
				int i;
				for (i = 0; i < argc; i++) {
					if (strncmp(argv[i], "--files-from", 12) == 0) {
						files_from = "";
						break;
					}
				}
			}
			if (files_from)
				write_byte(f_out, 0);
		}
		io_start_multiplex_out(f_out);
	}

	if (!ret || err_msg) {
		if (err_msg) {
			while ((p = strchr(err_msg, '\n')) != NULL) {
				int len = p - err_msg + 1;
				rwrite(FERROR, err_msg, len, 0);
				err_msg += len;
			}
			if (*err_msg)
				rprintf(FERROR, "%s\n", err_msg);
			io_flush(MSG_FLUSH);
		} else
			option_error();
		msleep(400);
		exit_cleanup(RERR_UNSUPPORTED);
	}

#ifdef ICONV_OPTION
	if (!iconv_opt) {
		if (ic_send != (iconv_t)-1) {
			iconv_close(ic_send);
			ic_send = (iconv_t)-1;
		}
		if (ic_recv != (iconv_t)-1) {
			iconv_close(ic_recv);
			ic_recv = (iconv_t)-1;
		}
	}
#endif

	if (!numeric_ids
	 && (use_chroot ? lp_numeric_ids(module_id) != False && !*lp_name_converter(module_id)
		        : lp_numeric_ids(module_id) == True))
		numeric_ids = -1; /* Set --numeric-ids w/o breaking protocol. */

	if (lp_timeout(module_id) && (!io_timeout || lp_timeout(module_id) < io_timeout))
		set_io_timeout(lp_timeout(module_id));

	/* If we have some incoming/outgoing chmod changes, append them to
	 * any user-specified changes (making our changes have priority).
	 * We also get a pointer to just our changes so that a receiver
	 * process can use them separately if --perms wasn't specified. */
	if (am_sender)
		p = lp_outgoing_chmod(module_id);
	else
		p = lp_incoming_chmod(module_id);
	if (*p && !(daemon_chmod_modes = parse_chmod(p, &chmod_modes))) {
		rprintf(FLOG, "Invalid \"%sing chmod\" directive: %s\n",
			am_sender ? "outgo" : "incom", p);
	}

	start_server(f_in, f_out, argc, argv);

	return 0;
}

BOOL namecvt_call(const char *cmd, const char **name_p, id_t *id_p)
{
	char buf[1024];
	int got, len;

	if (*name_p)
		len = snprintf(buf, sizeof buf, "%s %s\n", cmd, *name_p);
	else
		len = snprintf(buf, sizeof buf, "%s %ld\n", cmd, (long)*id_p);
	if (len >= (int)sizeof buf) {
		rprintf(FERROR, "namecvt_call() request was too large.\n");
		exit_cleanup(RERR_UNSUPPORTED);
	}

	while ((got = write(namecvt_fd_req, buf, len)) != len) {
		if (got < 0 && errno == EINTR)
			continue;
		rprintf(FERROR, "Connection to name-converter failed.\n");
		exit_cleanup(RERR_SOCKETIO);
	}

	if (!read_line_old(namecvt_fd_ans, buf, sizeof buf, 0))
		return False;

	if (*name_p)
		*id_p = (id_t)atol(buf);
	else
		*name_p = strdup(buf);

	return True;
}

/* send a list of available modules to the client. Don't list those
   with "list = False". */
static void send_listing(int fd)
{
	int n = lp_num_modules();
	int i;

	for (i = 0; i < n; i++) {
		if (lp_list(i))
			io_printf(fd, "%-15s\t%s\n", lp_name(i), lp_comment(i));
	}

	if (protocol_version >= 25)
		io_printf(fd,"@RSYNCD: EXIT\n");
}

static int load_config(int globals_only)
{
	if (!config_file) {
		if (am_daemon < 0 && am_root <= 0)
			config_file = RSYNCD_USERCONF;
		else
			config_file = RSYNCD_SYSCONF;
	}
	return lp_load(config_file, globals_only);
}

/* this is called when a connection is established to a client
   and we want to start talking. The setup of the system is done from
   here */
int start_daemon(int f_in, int f_out)
{
	char line[1024];
	const char *addr, *host;
	char *p;
	int i;

	/* At this point, am_server is only set for a daemon started via rsh.
	 * Because am_server gets forced on soon, we'll set am_daemon to -1 as
	 * a flag that can be checked later on to distinguish a normal daemon
	 * from an rsh-run daemon. */
	if (am_server)
		am_daemon = -1;

	io_set_sock_fds(f_in, f_out);

	/* We must load the config file before calling any function that
	 * might cause log-file output to occur.  This ensures that the
	 * "log file" param gets honored for the 2 non-forked use-cases
	 * (when rsync is run by init and run by a remote shell). */
	if (!load_config(0))
		exit_cleanup(RERR_SYNTAX);

	if (lp_proxy_protocol() && !read_proxy_protocol_header(f_in))
		return -1;

	p = lp_daemon_chroot();
	if (*p) {
		log_init(0); /* Make use we've initialized syslog before chrooting. */
		if (chroot(p) < 0 || chdir("/") < 0) {
			rsyserr(FLOG, errno, "daemon chroot %s failed", p);
			return -1;
		}
	}
	p = lp_daemon_gid();
	if (*p) {
		gid_t gid;
		if (!group_to_gid(p, &gid, True)) {
			rprintf(FLOG, "Invalid daemon gid: %s\n", p);
			return -1;
		}
		if (setgid(gid) < 0) {
			rsyserr(FLOG, errno, "Unable to set group to daemon gid %ld", (long)gid);
			return -1;
		}
		our_gid = MY_GID();
	}
	p = lp_daemon_uid();
	if (*p) {
		uid_t uid;
		if (!user_to_uid(p, &uid, True)) {
			rprintf(FLOG, "Invalid daemon uid: %s\n", p);
			return -1;
		}
		if (setuid(uid) < 0) {
			rsyserr(FLOG, errno, "Unable to set user to daemon uid %ld", (long)uid);
			return -1;
		}
		our_uid = MY_UID();
		am_root = (our_uid == ROOT_UID);
	}

	addr = client_addr(f_in);
	host = lp_reverse_lookup(-1) ? client_name(addr) : undetermined_hostname;
	rprintf(FLOG, "connect from %s (%s)\n", host, addr);

	if (am_daemon > 0) {
		set_socket_options(f_in, "SO_KEEPALIVE");
		set_nonblocking(f_in);
	}

	if (exchange_protocols(f_in, f_out, line, sizeof line, 0) < 0)
		return -1;

	line[0] = 0;
	if (!read_line_old(f_in, line, sizeof line, 0))
		return -1;

	if (strncmp(line, EARLY_INPUT_CMD, EARLY_INPUT_CMDLEN) == 0) {
		early_input_len = strtol(line + EARLY_INPUT_CMDLEN, NULL, 10);
		if (early_input_len <= 0 || early_input_len > BIGPATHBUFLEN) {
			io_printf(f_out, "@ERROR: invalid early_input length\n");
			return -1;
		}
		early_input = new_array(char, early_input_len);
		read_buf(f_in, early_input, early_input_len);

		if (!read_line_old(f_in, line, sizeof line, 0))
			return -1;
	}

	if (!*line || strcmp(line, "#list") == 0) {
		rprintf(FLOG, "module-list request from %s (%s)\n",
			host, addr);
		send_listing(f_out);
		return -1;
	}

	if (*line == '#') {
		/* it's some sort of command that I don't understand */
		io_printf(f_out, "@ERROR: Unknown command '%s'\n", line);
		return -1;
	}

	if ((i = lp_number(line)) < 0) {
		rprintf(FLOG, "unknown module '%s' tried from %s (%s)\n",
			line, host, addr);
		io_printf(f_out, "@ERROR: Unknown module '%s'\n", line);
		return -1;
	}

#ifdef HAVE_SIGACTION
	sigact.sa_flags = SA_NOCLDSTOP;
#endif
	SIGACTION(SIGCHLD, remember_children);

	return rsync_module(f_in, f_out, i, addr, host);
}

static void create_pid_file(void)
{
	char *pid_file = lp_pid_file();
	char pidbuf[32];
	STRUCT_STAT st1, st2;
	char *fail = NULL;

	if (!pid_file || !*pid_file)
		return;

#ifdef O_NOFOLLOW
#define SAFE_OPEN_FLAGS (O_CREAT|O_NOFOLLOW)
#else
#define SAFE_OPEN_FLAGS (O_CREAT)
#endif

	/* These tests make sure that a temp-style lock dir is handled safely. */
	st1.st_mode = 0;
	if (do_lstat(pid_file, &st1) == 0 && !S_ISREG(st1.st_mode) && unlink(pid_file) < 0)
		fail = "unlink";
	else if ((pid_file_fd = do_open(pid_file, O_RDWR|SAFE_OPEN_FLAGS, 0664)) < 0)
		fail = S_ISREG(st1.st_mode) ? "open" : "create";
	else if (!lock_range(pid_file_fd, 0, 4))
		fail = "lock";
	else if (do_fstat(pid_file_fd, &st1) < 0)
		fail = "fstat opened";
	else if (st1.st_size > (int)sizeof pidbuf)
		fail = "find small";
	else if (do_lstat(pid_file, &st2) < 0)
		fail = "lstat";
	else if (!S_ISREG(st1.st_mode))
		fail = "avoid file overwrite race for";
	else if (st1.st_dev != st2.st_dev || st1.st_ino != st2.st_ino)
		fail = "verify stat info for";
#ifdef HAVE_FTRUNCATE
	else if (do_ftruncate(pid_file_fd, 0) < 0)
		fail = "truncate";
#endif
	else {
		pid_t pid = getpid();
		int len = snprintf(pidbuf, sizeof pidbuf, "%d\n", (int)pid);
#ifndef HAVE_FTRUNCATE
		/* What can we do with a too-long file and no truncate? I guess we'll add extra newlines. */
		while (len < st1.st_size) /* We already verified that st_size chars fits in the buffer. */
			pidbuf[len++] = '\n';
		/* We don't need the buffer to end in a '\0' (and we may not have room to add it). */
#endif
		if (write(pid_file_fd, pidbuf, len) != len)
			 fail = "write";
		cleanup_set_pid(pid); /* Mark the file for removal on exit, even if the write failed. */
	}

	if (fail) {
		char msg[1024];
		snprintf(msg, sizeof msg, "failed to %s pid file %s: %s\n",
			fail, pid_file, strerror(errno));
		fputs(msg, stderr);
		rprintf(FLOG, "%s", msg);
		exit_cleanup(RERR_FILEIO);
	}

	/* The file is left open so that the lock remains valid. It is closed in our forked child procs. */
}

/* Become a daemon, discarding the controlling terminal. */
static void become_daemon(void)
{
	int i;
	pid_t pid = fork();

	if (pid) {
		if (pid < 0) {
			fprintf(stderr, "failed to fork: %s\n", strerror(errno));
			exit_cleanup(RERR_FILEIO);
		}
		_exit(0);
	}

	create_pid_file();

	/* detach from the terminal */
#ifdef HAVE_SETSID
	setsid();
#elif defined TIOCNOTTY
	i = open("/dev/tty", O_RDWR);
	if (i >= 0) {
		ioctl(i, (int)TIOCNOTTY, (char *)0);
		close(i);
	}
#endif
	/* make sure that stdin, stdout an stderr don't stuff things
	 * up (library functions, for example) */
	for (i = 0; i < 3; i++) {
		close(i);
		open("/dev/null", O_RDWR);
	}
}

int daemon_main(void)
{
	if (is_a_socket(STDIN_FILENO)) {
		int i;

		/* we are running via inetd - close off stdout and
		 * stderr so that library functions (and getopt) don't
		 * try to use them. Redirect them to /dev/null */
		for (i = 1; i < 3; i++) {
			close(i);
			open("/dev/null", O_RDWR);
		}

		return start_daemon(STDIN_FILENO, STDIN_FILENO);
	}

	if (!load_config(1)) {
		fprintf(stderr, "Failed to parse config file: %s\n", config_file);
		exit_cleanup(RERR_SYNTAX);
	}
	set_dparams(0);

	if (no_detach)
		create_pid_file();
	else
		become_daemon();

	if (rsync_port == 0 && (rsync_port = lp_rsync_port()) == 0)
		rsync_port = RSYNC_PORT;
	if (bind_address == NULL && *lp_bind_address())
		bind_address = lp_bind_address();

	log_init(0);

	rprintf(FLOG, "rsyncd version %s starting, listening on port %d\n",
		rsync_version(), rsync_port);
	/* TODO: If listening on a particular address, then show that
	 * address too.  In fact, why not just do getnameinfo on the
	 * local address??? */

#ifdef HAVE_LIBSLP
	if (lp_use_slp() && register_services()) {
		rprintf(FINFO,
		    "Couldn't register with service discovery protocol, continuing anyway\n");
	}
#endif

	start_accept_loop(rsync_port, start_daemon);
	return -1;
}

FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>