Noted hardlink memory footprint reduction.

[rsync/rsync.git] / receiver.c

/* -*- c-file-style: "linux" -*-

   Copyright (C) 1996-2000 by Andrew Tridgell
   Copyright (C) Paul Mackerras 1996

   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 2 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, write to the Free Software
   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/

#include "rsync.h"

extern int verbose;
extern int recurse;
extern int delete_mode;
extern int protocol_version;
extern int csum_length;
extern struct stats stats;
extern int dry_run;
extern int am_server;
extern int relative_paths;
extern int preserve_hard_links;
extern int cvs_exclude;
extern int io_error;
extern char *tmpdir;
extern char *compare_dest;
extern int make_backups;
extern int do_progress;
extern char *backup_dir;
extern char *backup_suffix;
extern int backup_suffix_len;
extern int cleanup_got_literal;

static struct delete_list {
	DEV64_T dev;
	INO64_T inode;
} *delete_list;
static int dlist_len, dlist_alloc_len;

/* yuck! This function wouldn't have been necessary if I had the sorting
 * algorithm right. Unfortunately fixing the sorting algorithm would introduce
 * a backward incompatibility as file list indexes are sent over the link.
 */
static int delete_already_done(struct file_list *flist,int j)
{
	int i;
	STRUCT_STAT st;

	if (link_stat(f_name(flist->files[j]), &st)) return 1;

	for (i = 0; i < dlist_len; i++) {
		if (st.st_ino == delete_list[i].inode &&
		    (DEV64_T)st.st_dev == delete_list[i].dev)
			return 1;
	}

	return 0;
}

static void add_delete_entry(struct file_struct *file)
{
	if (dlist_len == dlist_alloc_len) {
		dlist_alloc_len += 1024;
		delete_list = realloc_array(delete_list, struct delete_list,
					    dlist_alloc_len);
		if (!delete_list) out_of_memory("add_delete_entry");
	}

	delete_list[dlist_len].dev = file->dev;
	delete_list[dlist_len].inode = file->inode;
	dlist_len++;

	if (verbose > 3)
		rprintf(FINFO,"added %s to delete list\n", f_name(file));
}

static void delete_one(char *fn, int is_dir)
{
	if (!is_dir) {
		if (robust_unlink(fn) != 0) {
			rprintf(FERROR, "delete_one: unlink %s failed: %s\n",
				full_fname(fn), strerror(errno));
		} else if (verbose) {
			rprintf(FINFO, "deleting %s\n", fn);
		}
	} else {
		if (do_rmdir(fn) != 0) {
			if (errno != ENOTEMPTY && errno != EEXIST) {
				rprintf(FERROR, "delete_one: rmdir %s failed: %s\n",
					full_fname(fn), strerror(errno));
			}
		} else if (verbose) {
			rprintf(FINFO, "deleting directory %s\n", fn);
		}
	}
}


static int is_backup_file(char *fn)
{
	int k = strlen(fn) - backup_suffix_len;
	return k > 0 && strcmp(fn+k, backup_suffix) == 0;
}


/* this deletes any files on the receiving side that are not present
 * on the sending side. For version 1.6.4 I have changed the behaviour
 * to match more closely what most people seem to expect of this option */
void delete_files(struct file_list *flist)
{
	struct file_list *local_file_list;
	int i, j;
	char *name, fbuf[MAXPATHLEN];
	extern int module_id;
	extern int ignore_errors;
	extern int max_delete;
	static int deletion_count;

	if (cvs_exclude)
		add_cvs_excludes();

	if (io_error && !(lp_ignore_errors(module_id) || ignore_errors)) {
		rprintf(FINFO,"IO error encountered - skipping file deletion\n");
		return;
	}

	for (j = 0;j < flist->count; j++) {
		if (!S_ISDIR(flist->files[j]->mode) ||
		    !(flist->files[j]->flags & FLAG_DELETE)) continue;

		if (protocol_version < 19 &&
		    delete_already_done(flist, j)) continue;

		name = f_name_to(flist->files[j], fbuf, sizeof fbuf);

		if (!(local_file_list = send_file_list(-1,1,&name)))
			continue;

		if (verbose > 1)
			rprintf(FINFO,"deleting in %s\n", name);

		for (i = local_file_list->count-1; i >= 0; i--) {
			if (max_delete && deletion_count > max_delete) break;
			if (!local_file_list->files[i]->basename) continue;
			if (protocol_version < 19 &&
			    S_ISDIR(local_file_list->files[i]->mode))
				add_delete_entry(local_file_list->files[i]);
			if (-1 == flist_find(flist,local_file_list->files[i])) {
				char *f = f_name(local_file_list->files[i]);
				if (make_backups && (backup_dir || !is_backup_file(f))) {
					(void) make_backup(f);
					if (verbose)
						rprintf(FINFO, "deleting %s\n", f);
				} else {
					int mode = local_file_list->files[i]->mode;
					delete_one(f, S_ISDIR(mode) != 0);
				}
				deletion_count++;
			}
		}
		flist_free(local_file_list);
	}
}


/*
 * get_tmpname() - create a tmp filename for a given filename
 *
 *   If a tmpdir is defined, use that as the directory to
 *   put it in.  Otherwise, the tmp filename is in the same
 *   directory as the given name.  Note that there may be no
 *   directory at all in the given name!
 *
 *   The tmp filename is basically the given filename with a
 *   dot prepended, and .XXXXXX appended (for mkstemp() to
 *   put its unique gunk in).  Take care to not exceed
 *   either the MAXPATHLEN or NAME_MAX, esp. the last, as
 *   the basename basically becomes 8 chars longer. In that
 *   case, the original name is shortened sufficiently to
 *   make it all fit.
 *
 *   Of course, there's no real reason for the tmp name to
 *   look like the original, except to satisfy us humans.
 *   As long as it's unique, rsync will work.
 */

static int get_tmpname(char *fnametmp, char *fname)
{
	char *f;
	int     length = 0;
	int	maxname;

	if (tmpdir) {
		strlcpy(fnametmp, tmpdir, MAXPATHLEN - 2);
		length = strlen(fnametmp);
		fnametmp[length++] = '/';
		fnametmp[length] = '\0';	/* always NULL terminated */
	}

	if ((f = strrchr(fname, '/')) != NULL) {
		++f;
		if (!tmpdir) {
			length = f - fname;
			/* copy up to and including the slash */
			strlcpy(fnametmp, fname, length + 1);
		}
	} else
		f = fname;
	fnametmp[length++] = '.';
	fnametmp[length] = '\0';		/* always NULL terminated */

	maxname = MIN(MAXPATHLEN - 7 - length, NAME_MAX - 8);

	if (maxname < 1) {
		rprintf(FERROR, "temporary filename too long: %s\n", fname);
		fnametmp[0] = '\0';
		return 0;
	}

	strlcpy(fnametmp + length, f, maxname);
	strcat(fnametmp + length, ".XXXXXX");

	return 1;
}


static int receive_data(int f_in,struct map_struct *mapbuf,int fd,char *fname,
			OFF_T total_size)
{
	int i;
	struct sum_struct sum;
	unsigned int len;
	OFF_T offset = 0;
	OFF_T offset2;
	char *data;
	static char file_sum1[MD4_SUM_LENGTH];
	static char file_sum2[MD4_SUM_LENGTH];
	char *map=NULL;

	read_sum_head(f_in, &sum);

	sum_init();

	for (i=recv_token(f_in,&data); i != 0; i=recv_token(f_in,&data)) {
		if (do_progress)
			show_progress(offset, total_size);

		if (i > 0) {
			if (verbose > 3) {
				rprintf(FINFO,"data recv %d at %.0f\n",
					i,(double)offset);
			}

			stats.literal_data += i;
			cleanup_got_literal = 1;

			sum_update(data,i);

			if (fd != -1 && write_file(fd,data,i) != i) {
				rprintf(FERROR, "write failed on %s: %s\n",
					full_fname(fname), strerror(errno));
				exit_cleanup(RERR_FILEIO);
			}
			offset += i;
			continue;
		}

		i = -(i+1);
		offset2 = i*(OFF_T)sum.blength;
		len = sum.blength;
		if (i == (int) sum.count-1 && sum.remainder != 0)
			len = sum.remainder;

		stats.matched_data += len;

		if (verbose > 3)
			rprintf(FINFO,"chunk[%d] of size %d at %.0f offset=%.0f\n",
				i,len,(double)offset2,(double)offset);

		if (mapbuf) {
			map = map_ptr(mapbuf,offset2,len);

			see_token(map, len);
			sum_update(map,len);
		}

		if (fd != -1 && write_file(fd,map,len) != (int) len) {
			rprintf(FERROR, "write failed on %s: %s\n",
				full_fname(fname), strerror(errno));
			exit_cleanup(RERR_FILEIO);
		}
		offset += len;
	}

	flush_write_file(fd);

	if (do_progress)
		end_progress(total_size);

	if (fd != -1 && offset > 0 && sparse_end(fd) != 0) {
		rprintf(FERROR, "write failed on %s: %s\n",
			full_fname(fname), strerror(errno));
		exit_cleanup(RERR_FILEIO);
	}

	sum_end(file_sum1);

	read_buf(f_in,file_sum2,MD4_SUM_LENGTH);
	if (verbose > 2) {
		rprintf(FINFO,"got file_sum\n");
	}
	if (fd != -1 && memcmp(file_sum1,file_sum2,MD4_SUM_LENGTH) != 0) {
		return 0;
	}
	return 1;
}


/**
 * main routine for receiver process.
 *
 * Receiver process runs on the same host as the generator process. */
int recv_files(int f_in,struct file_list *flist,char *local_name,int f_gen)
{
	int fd1,fd2;
	STRUCT_STAT st;
	char *fname, fbuf[MAXPATHLEN];
	char template[MAXPATHLEN];
	char fnametmp[MAXPATHLEN];
	char *fnamecmp;
	char fnamecmpbuf[MAXPATHLEN];
	struct map_struct *mapbuf;
	int i;
	struct file_struct *file;
	int phase=0;
	int recv_ok;
	extern struct stats stats;
	extern int preserve_perms;
	extern int delete_after;
	extern int orig_umask;
	struct stats initial_stats;

	if (verbose > 2) {
		rprintf(FINFO,"recv_files(%d) starting\n",flist->count);
	}

	while (1) {
		cleanup_disable();

		i = read_int(f_in);
		if (i == -1) {
			if (phase==0) {
				phase++;
				csum_length = SUM_LENGTH;
				if (verbose > 2)
					rprintf(FINFO,"recv_files phase=%d\n",phase);
				write_int(f_gen,-1);
				continue;
			}
			break;
		}

		if (i < 0 || i >= flist->count) {
			rprintf(FERROR,"Invalid file index %d in recv_files (count=%d)\n",
				i, flist->count);
			exit_cleanup(RERR_PROTOCOL);
		}

		file = flist->files[i];

		stats.num_transferred_files++;
		stats.total_transferred_size += file->length;
		cleanup_got_literal = 0;

		if (local_name)
			fname = local_name;
		else
			fname = f_name_to(file, fbuf, sizeof fbuf);

		if (dry_run) {
			if (!am_server && verbose) {	/* log transfer */
				rprintf(FINFO, "%s\n", fname);
			}
			continue;
		}

		initial_stats = stats;

		if (verbose > 2)
			rprintf(FINFO,"recv_files(%s)\n",fname);

		fnamecmp = fname;

		/* open the file */
		fd1 = do_open(fnamecmp, O_RDONLY, 0);

		if ((fd1 == -1) && (compare_dest != NULL)) {
			/* try the file at compare_dest instead */
			snprintf(fnamecmpbuf,MAXPATHLEN,"%s/%s",
						compare_dest,fname);
			fnamecmp = fnamecmpbuf;
			fd1 = do_open(fnamecmp, O_RDONLY, 0);
		}

		if (fd1 != -1 && do_fstat(fd1,&st) != 0) {
			rprintf(FERROR, "fstat %s failed: %s\n",
				full_fname(fnamecmp), strerror(errno));
			receive_data(f_in,NULL,-1,NULL,file->length);
			close(fd1);
			continue;
		}

		if (fd1 != -1 && S_ISDIR(st.st_mode) && fnamecmp == fname) {
			/* this special handling for directories
			 * wouldn't be necessary if robust_rename()
			 * and the underlying robust_unlink could cope
			 * with directories
			 */
			rprintf(FERROR,"recv_files: %s is a directory\n",
				full_fname(fnamecmp));
			receive_data(f_in, NULL, -1, NULL, file->length);
			close(fd1);
			continue;
		}

		if (fd1 != -1 && !S_ISREG(st.st_mode)) {
			close(fd1);
			fd1 = -1;
			mapbuf = NULL;
		}

		if (fd1 != -1 && !preserve_perms) {
			/* if the file exists already and we aren't preserving
			 * permissions then act as though the remote end sent
			 * us the file permissions we already have */
			file->mode = st.st_mode;
		}

		if (fd1 != -1 && st.st_size > 0) {
			mapbuf = map_file(fd1,st.st_size);
			if (verbose > 2)
				rprintf(FINFO,"recv mapped %s of size %.0f\n",fnamecmp,(double)st.st_size);
		} else
			mapbuf = NULL;

		if (!get_tmpname(fnametmp,fname)) {
			if (mapbuf) unmap_file(mapbuf);
			if (fd1 != -1) close(fd1);
			continue;
		}

		strlcpy(template, fnametmp, sizeof(template));

		/* we initially set the perms without the
		 * setuid/setgid bits to ensure that there is no race
		 * condition. They are then correctly updated after
		 * the lchown. Thanks to snabb@epipe.fi for pointing
		 * this out.  We also set it initially without group
		 * access because of a similar race condition. */
		fd2 = do_mkstemp(fnametmp, file->mode & INITACCESSPERMS);

		/* in most cases parent directories will already exist
		 * because their information should have been previously
		 * transferred, but that may not be the case with -R */
		if (fd2 == -1 && relative_paths && errno == ENOENT &&
		    create_directory_path(fnametmp, orig_umask) == 0) {
			strlcpy(fnametmp, template, sizeof(fnametmp));
			fd2 = do_mkstemp(fnametmp, file->mode & INITACCESSPERMS);
		}
		if (fd2 == -1) {
			rprintf(FERROR, "mkstemp %s failed: %s\n",
				full_fname(fnametmp), strerror(errno));
			receive_data(f_in,mapbuf,-1,NULL,file->length);
			if (mapbuf) unmap_file(mapbuf);
			if (fd1 != -1) close(fd1);
			continue;
		}

		cleanup_set(fnametmp, fname, file, mapbuf, fd1, fd2);

		if (!am_server && verbose) {	/* log transfer */
			rprintf(FINFO, "%s\n", fname);
		}

		/* recv file data */
		recv_ok = receive_data(f_in,mapbuf,fd2,fname,file->length);

		log_recv(file, &initial_stats);

		if (mapbuf) unmap_file(mapbuf);
		if (fd1 != -1) {
			close(fd1);
		}
		close(fd2);

		if (verbose > 2)
			rprintf(FINFO,"renaming %s to %s\n",fnametmp,fname);

		finish_transfer(fname, fnametmp, file);

		cleanup_disable();

		if (!recv_ok) {
			if (csum_length == SUM_LENGTH) {
				rprintf(FERROR,"ERROR: file corruption in %s. File changed during transfer?\n",
					full_fname(fname));
			} else {
				if (verbose > 1)
					rprintf(FINFO,"redoing %s(%d)\n",fname,i);
				write_int(f_gen,i);
			}
		}
	}

	if (delete_after) {
		if (recurse && delete_mode && !local_name && flist->count>0) {
			delete_files(flist);
		}
	}

	if (preserve_hard_links)
		do_hard_links();

	/* now we need to fix any directory permissions that were
	 * modified during the transfer */
	for (i = 0; i < flist->count; i++) {
		file = flist->files[i];
		if (!file->basename || !S_ISDIR(file->mode)) continue;
		recv_generator(local_name? local_name
			     : f_name_to(file,fbuf,sizeof fbuf), file, i, -1);
	}

	if (verbose > 2)
		rprintf(FINFO,"recv_files finished\n");

	return 0;
}