main.c

Rsync 3.0.5 source code

/*
 * The startup routines, including main(), for rsync.
 *
 * Copyright (C) 1996-2001 Andrew Tridgell <[e-mail]>
 * Copyright (C) 1996 Paul Mackerras
 * Copyright (C) 2001, 2002 Martin Pool <[e-mail]>
 * Copyright (C) 2003-2008 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 "ifuncs.h"
#include "io.h"
#if defined CONFIG_LOCALE && defined HAVE_LOCALE_H
#include <locale.h>
#endif

extern int verbose;
extern int dry_run;
extern int list_only;
extern int am_root;
extern int am_server;
extern int am_sender;
extern int am_generator;
extern int am_daemon;
extern int inc_recurse;
extern int blocking_io;
extern int remove_source_files;
extern int need_messages_from_generator;
extern int kluge_around_eof;
extern int do_stats;
extern int got_xfer_error;
extern int module_id;
extern int copy_links;
extern int copy_dirlinks;
extern int copy_unsafe_links;
extern int keep_dirlinks;
extern int preserve_hard_links;
extern int protocol_version;
extern int file_total;
extern int recurse;
extern int xfer_dirs;
extern int protect_args;
extern int relative_paths;
extern int sanitize_paths;
extern int curr_dir_depth;
extern int curr_dir_len;
extern int module_id;
extern int rsync_port;
extern int whole_file;
extern int read_batch;
extern int write_batch;
extern int batch_fd;
extern int filesfrom_fd;
extern int connect_timeout;
extern pid_t cleanup_child_pid;
extern unsigned int module_dirlen;
extern struct stats stats;
extern char *filesfrom_host;
extern char *partial_dir;
extern char *dest_option;
extern char *basis_dir[];
extern char *rsync_path;
extern char *shell_cmd;
extern char *batch_name;
extern char *password_file;
extern char curr_dir[MAXPATHLEN];
extern struct file_list *first_flist;
extern struct filter_list_struct daemon_filter_list;

uid_t our_uid;
int local_server = 0;
int daemon_over_rsh = 0;
mode_t orig_umask = 0;
int batch_gen_fd = -1;

/* There's probably never more than at most 2 outstanding child processes,
 * but set it higher, just in case. */
#define MAXCHILDPROCS 7

#ifdef HAVE_SIGACTION
# ifdef HAVE_SIGPROCMASK
#  define SIGACTMASK(n,h) SIGACTION(n,h), sigaddset(&sigmask,(n))
# else
#  define SIGACTMASK(n,h) SIGACTION(n,h)
# endif
static struct sigaction sigact;
#endif

struct pid_status {
	pid_t pid;
	int status;
} pid_stat_table[MAXCHILDPROCS];

static time_t starttime, endtime;
static int64 total_read, total_written;

static void show_malloc_stats(void);

/* Works like waitpid(), but if we already harvested the child pid in our
 * remember_children(), we succeed instead of returning an error. */
pid_t wait_process(pid_t pid, int *status_ptr, int flags)
{
	pid_t waited_pid;
	
	do {
		waited_pid = waitpid(pid, status_ptr, flags);
	} while (waited_pid == -1 && errno == EINTR);

	if (waited_pid == -1 && errno == ECHILD) {
		/* Status of requested child no longer available:  check to
		 * see if it was processed by remember_children(). */
		int cnt;
		for (cnt = 0; cnt < MAXCHILDPROCS; cnt++) {
			if (pid == pid_stat_table[cnt].pid) {
				*status_ptr = pid_stat_table[cnt].status;
				pid_stat_table[cnt].pid = 0;
				return pid;
			}
		}
	}

	return waited_pid;
}

/* Wait for a process to exit, calling io_flush while waiting. */
static void wait_process_with_flush(pid_t pid, int *exit_code_ptr)
{
	pid_t waited_pid;
	int status;

	while ((waited_pid = wait_process(pid, &status, WNOHANG)) == 0) {
		msleep(20);
		io_flush(FULL_FLUSH);
	}

	/* TODO: If the child exited on a signal, then log an
	 * appropriate error message.  Perhaps we should also accept a
	 * message describing the purpose of the child.  Also indicate
	 * this to the caller so that they know something went wrong. */
	if (waited_pid < 0) {
		rsyserr(FERROR, errno, "waitpid");
		*exit_code_ptr = RERR_WAITCHILD;
	} else if (!WIFEXITED(status)) {
#ifdef WCOREDUMP
		if (WCOREDUMP(status))
			*exit_code_ptr = RERR_CRASHED;
		else
#endif
		if (WIFSIGNALED(status))
			*exit_code_ptr = RERR_TERMINATED;
		else
			*exit_code_ptr = RERR_WAITCHILD;
	} else
		*exit_code_ptr = WEXITSTATUS(status);
}

/* This function gets called from all 3 processes.  We want the client side
 * to actually output the text, but the sender is the only process that has
 * all the stats we need.  So, if we're a client sender, we do the report.
 * If we're a server sender, we write the stats on the supplied fd.  If
 * we're the client receiver we read the stats from the supplied fd and do
 * the report.  All processes might also generate a set of debug stats, if
 * the verbose level is high enough (this is the only thing that the
 * generator process and the server receiver ever do here). */
static void handle_stats(int f)
{
	endtime = time(NULL);

	/* Cache two stats because the read/write code can change it. */
	total_read = stats.total_read;
	total_written = stats.total_written;

	if (do_stats && verbose > 1) {
		/* These come out from every process */
		show_malloc_stats();
		show_flist_stats();
	}

	if (am_generator)
		return;

	if (am_daemon) {
		if (f == -1 || !am_sender)
			return;
	}

	if (am_server) {
		if (am_sender) {
			write_varlong30(f, total_read, 3);
			write_varlong30(f, total_written, 3);
			write_varlong30(f, stats.total_size, 3);
			if (protocol_version >= 29) {
				write_varlong30(f, stats.flist_buildtime, 3);
				write_varlong30(f, stats.flist_xfertime, 3);
			}
		}
		return;
	}

	/* this is the client */

	if (f < 0 && !am_sender) /* e.g. when we got an empty file list. */
		;
	else if (!am_sender) {
		/* Read the first two in opposite order because the meaning of
		 * read/write swaps when switching from sender to receiver. */
		total_written = read_varlong30(f, 3);
		total_read = read_varlong30(f, 3);
		stats.total_size = read_varlong30(f, 3);
		if (protocol_version >= 29) {
			stats.flist_buildtime = read_varlong30(f, 3);
			stats.flist_xfertime = read_varlong30(f, 3);
		}
	} else if (write_batch) {
		/* The --read-batch process is going to be a client
		 * receiver, so we need to give it the stats. */
		write_varlong30(batch_fd, total_read, 3);
		write_varlong30(batch_fd, total_written, 3);
		write_varlong30(batch_fd, stats.total_size, 3);
		if (protocol_version >= 29) {
			write_varlong30(batch_fd, stats.flist_buildtime, 3);
			write_varlong30(batch_fd, stats.flist_xfertime, 3);
		}
	}
}

static void output_summary(void)
{
	if (do_stats) {
		rprintf(FCLIENT, "\n");
		rprintf(FINFO,"Number of files: %d\n", stats.num_files);
		rprintf(FINFO,"Number of files transferred: %d\n",
			stats.num_transferred_files);
		rprintf(FINFO,"Total file size: %s bytes\n",
			human_num(stats.total_size));
		rprintf(FINFO,"Total transferred file size: %s bytes\n",
			human_num(stats.total_transferred_size));
		rprintf(FINFO,"Literal data: %s bytes\n",
			human_num(stats.literal_data));
		rprintf(FINFO,"Matched data: %s bytes\n",
			human_num(stats.matched_data));
		rprintf(FINFO,"File list size: %s\n",
			human_num(stats.flist_size));
		if (stats.flist_buildtime) {
			rprintf(FINFO,
				"File list generation time: %.3f seconds\n",
				(double)stats.flist_buildtime / 1000);
			rprintf(FINFO,
				"File list transfer time: %.3f seconds\n",
				(double)stats.flist_xfertime / 1000);
		}
		rprintf(FINFO,"Total bytes sent: %s\n",
			human_num(total_written));
		rprintf(FINFO,"Total bytes received: %s\n",
			human_num(total_read));
	}

	if (verbose || do_stats) {
		rprintf(FCLIENT, "\n");
		rprintf(FINFO,
			"sent %s bytes  received %s bytes  %s bytes/sec\n",
			human_num(total_written), human_num(total_read),
			human_dnum((total_written + total_read)/(0.5 + (endtime - starttime)), 2));
		rprintf(FINFO, "total size is %s  speedup is %.2f%s\n",
			human_num(stats.total_size),
			(double)stats.total_size / (total_written+total_read),
			write_batch < 0 ? " (BATCH ONLY)" : dry_run ? " (DRY RUN)" : "");
	}

	fflush(stdout);
	fflush(stderr);
}


/**
 * If our C library can get malloc statistics, then show them to FINFO
 **/
static void show_malloc_stats(void)
{
#ifdef HAVE_MALLINFO
	struct mallinfo mi;

	mi = mallinfo();

	rprintf(FCLIENT, "\n");
	rprintf(FINFO, RSYNC_NAME "[%d] (%s%s%s) heap statistics:\n",
		getpid(), am_server ? "server " : "",
		am_daemon ? "daemon " : "", who_am_i());
	rprintf(FINFO, "  arena:     %10ld   (bytes from sbrk)\n",
		(long)mi.arena);
	rprintf(FINFO, "  ordblks:   %10ld   (chunks not in use)\n",
		(long)mi.ordblks);
	rprintf(FINFO, "  smblks:    %10ld\n",
		(long)mi.smblks);
	rprintf(FINFO, "  hblks:     %10ld   (chunks from mmap)\n",
		(long)mi.hblks);
	rprintf(FINFO, "  hblkhd:    %10ld   (bytes from mmap)\n",
		(long)mi.hblkhd);
	rprintf(FINFO, "  allmem:    %10ld   (bytes from sbrk + mmap)\n",
		(long)mi.arena + mi.hblkhd);
	rprintf(FINFO, "  usmblks:   %10ld\n",
		(long)mi.usmblks);
	rprintf(FINFO, "  fsmblks:   %10ld\n",
		(long)mi.fsmblks);
	rprintf(FINFO, "  uordblks:  %10ld   (bytes used)\n",
		(long)mi.uordblks);
	rprintf(FINFO, "  fordblks:  %10ld   (bytes free)\n",
		(long)mi.fordblks);
	rprintf(FINFO, "  keepcost:  %10ld   (bytes in releasable chunk)\n",
		(long)mi.keepcost);
#endif /* HAVE_MALLINFO */
}


/* Start the remote shell.   cmd may be NULL to use the default. */
static pid_t do_cmd(char *cmd, char *machine, char *user, char **remote_argv, int remote_argc,
		    int *f_in_p, int *f_out_p)
{
	int i, argc = 0;
	char *args[MAX_ARGS];
	pid_t pid;
	int dash_l_set = 0;

	if (!read_batch && !local_server) {
		char *t, *f, in_quote = '\0';
		char *rsh_env = getenv(RSYNC_RSH_ENV);
		if (!cmd)
			cmd = rsh_env;
		if (!cmd)
			cmd = RSYNC_RSH;
		cmd = strdup(cmd); /* MEMORY LEAK */
		if (!cmd)
			goto oom;

		for (t = f = cmd; *f; f++) {
			if (*f == ' ')
				continue;
			/* Comparison leaves rooms for server_options(). */
			if (argc >= MAX_ARGS - MAX_SERVER_ARGS)
				goto arg_overflow;
			args[argc++] = t;
			while (*f != ' ' || in_quote) {
				if (!*f) {
					if (in_quote) {
						rprintf(FERROR,
						    "Missing trailing-%c in remote-shell command.\n",
						    in_quote);
						exit_cleanup(RERR_SYNTAX);
					}
					f--;
					break;
				}
				if (*f == '\'' || *f == '"') {
					if (!in_quote) {
						in_quote = *f++;
						continue;
					}
					if (*f == in_quote && *++f != in_quote) {
						in_quote = '\0';
						continue;
					}
				}
				*t++ = *f++;
			}
			*t++ = '\0';
		}

		/* check to see if we've already been given '-l user' in
		 * the remote-shell command */
		for (i = 0; i < argc-1; i++) {
			if (!strcmp(args[i], "-l") && args[i+1][0] != '-')
				dash_l_set = 1;
		}

#ifdef HAVE_REMSH
		/* remsh (on HPUX) takes the arguments the other way around */
		args[argc++] = machine;
		if (user && !(daemon_over_rsh && dash_l_set)) {
			args[argc++] = "-l";
			args[argc++] = user;
		}
#else
		if (user && !(daemon_over_rsh && dash_l_set)) {
			args[argc++] = "-l";
			args[argc++] = user;
		}
		args[argc++] = machine;
#endif

		args[argc++] = rsync_path;

		if (blocking_io < 0) {
			char *cp;
			if ((cp = strrchr(cmd, '/')) != NULL)
				cp++;
			else
				cp = cmd;
			if (strcmp(cp, "rsh") == 0 || strcmp(cp, "remsh") == 0)
				blocking_io = 1;
		}

		server_options(args,&argc);

		if (argc >= MAX_ARGS - 2)
			goto arg_overflow;
	}

	args[argc++] = ".";

	if (!daemon_over_rsh) {
		while (remote_argc > 0) {
			if (argc >= MAX_ARGS - 1) {
			  arg_overflow:
				rprintf(FERROR, "internal: args[] overflowed in do_cmd()\n");
				exit_cleanup(RERR_SYNTAX);
			}
			args[argc++] = *remote_argv++;
			remote_argc--;
		}
	}

	args[argc] = NULL;

	if (verbose > 3) {
		for (i = 0; i < argc; i++)
			rprintf(FCLIENT, "cmd[%d]=%s ", i, args[i]);
		rprintf(FCLIENT, "\n");
	}

	if (read_batch) {
		int from_gen_pipe[2];
		set_allow_inc_recurse();
		if (fd_pair(from_gen_pipe) < 0) {
			rsyserr(FERROR, errno, "pipe");
			exit_cleanup(RERR_IPC);
		}
		batch_gen_fd = from_gen_pipe[0];
		*f_out_p = from_gen_pipe[1];
		*f_in_p = batch_fd;
		pid = (pid_t)-1; /* no child pid */
#ifdef ICONV_CONST
		setup_iconv();
#endif
	} else if (local_server) {
		/* If the user didn't request --[no-]whole-file, force
		 * it on, but only if we're not batch processing. */
		if (whole_file < 0 && !write_batch)
			whole_file = 1;
		set_allow_inc_recurse();
		pid = local_child(argc, args, f_in_p, f_out_p, child_main);
#ifdef ICONV_CONST
		setup_iconv();
#endif
	} else {
		pid = piped_child(args, f_in_p, f_out_p);
#ifdef ICONV_CONST
		setup_iconv();
#endif
		if (protect_args && !daemon_over_rsh)
			send_protected_args(*f_out_p, args);
	}

	return pid;

  oom:
	out_of_memory("do_cmd");
	return 0; /* not reached */
}

/* The receiving side operates in one of two modes:
 *
 * 1. it receives any number of files into a destination directory,
 * placing them according to their names in the file-list.
 *
 * 2. it receives a single file and saves it using the name in the
 * destination path instead of its file-list name.  This requires a
 * "local name" for writing out the destination file.
 *
 * So, our task is to figure out what mode/local-name we need.
 * For mode 1, we change into the destination directory and return NULL.
 * For mode 2, we change into the directory containing the destination
 * file (if we aren't already there) and return the local-name. */
static char *get_local_name(struct file_list *flist, char *dest_path)
{
	STRUCT_STAT st;
	int statret;
	char *cp;

	if (verbose > 2) {
		rprintf(FINFO, "get_local_name count=%d %s\n",
			file_total, NS(dest_path));